Review: Comparative methane production in mammalian herbivores

Abstract

Methane (CH₄) production is a ubiquitous, apparently unavoidable side effect of fermentative fibre digestion by symbiotic microbiota in mammalian herbivores. Here, a data compilation is presented of in vivo CH₄ measurements in individuals of 37 mammalian herbivore species fed forage-only diets, from the literature and from hitherto unpublished measurements. In contrast to previous claims, absolute CH₄ emissions scaled linearly to DM intake, and CH₄ yields (per DM or gross energy intake) did not vary significantly with body mass. CH₄ physiology hence cannot be construed to represent an intrinsic ruminant or herbivore body size limitation. The dataset does not support traditional dichotomies of CH₄ emission intensity between ruminants and nonruminants, or between foregut and hindgut fermenters. Several rodent hindgut fermenters and nonruminant foregut fermenters emit CH₄ of a magnitude as high as ruminants of similar size, intake level, digesta retention or gut capacity. By contrast, equids, macropods (kangaroos) and rabbits produce few CH₄ and have low CH₄ : CO₂ ratios for their size, intake level, digesta retention or gut capacity, ruling out these factors as explanation for interspecific variation. These findings lead to the conclusion that still unidentified host-specific factors other than digesta retention characteristics, or the presence of rumination or a foregut, influence CH₄ production. Measurements of CH₄ yield per digested fibre indicate that the amount of CH₄ produced during fibre digestion varies not only across but also within species, possibly pointing towards variation in microbiota functionality. Recent findings on the genetic control of microbiome composition, including methanogens, raise the question about the benefits methanogens provide for many (but apparently not to the same extent for all) species, which possibly prevented the evolution of the hosting of low-methanogenic microbiota across mammals.

Implications

This work reviews existing data on in vivo methane emissions in mammalian herbivores, demonstrating no constraint of methane physiology on body size, and no consistent difference between ruminants and nonruminants, or between foregut and hindgut fermenters. However, it singles out three groups – horses, kangaroos and rabbits – as model animals in which to investigate adaptations for low methane emissions.

Introduction

Methanogenesis is ubiquitous

The presence of methanogenic archaea appears to be nearly inevitable in the digestive tracts of animals. Methanogenesis has been reported in a wide array of arthropods (Hackstein and Stumm, 1994) and in the faeces of a large number of vertebrates, including not only herbivores, but also carnivorous reptiles and myrmecophageous mammals (Hackstein and Van Alen, 1996; Lambert and Fellner, 2012). Reports stating that various vertebrate groups do not produce methane (CH₄) or do not harbour methanogens were met by opposite evidence. This includes the ostrich (Struthio camelus) (Swart et al., 1993; Miramontes-Carrillo et al., 2008; Matsui et al., 2010), kangaroos (Dellow et al., 1988; Vendl et al., 2015), mammalian carnivores (Hackstein and Van Alen, 1996; Middelbos et al., 2008; Tun et al., 2012), sea cows (Marsh et al., 1978; Goto et al., 2004), colobus monkeys (Bauchop and Martucci, 1968; Ohwaki et al., 1974) and arvicoline rodents (Hackstein and Van Alen, 1996; this study). Supplementary Material Table S1 gives an exemplaryoverview over mammal species in which CH₄ emissions or methanogen presence has been detected. It generally appears prudent to assume that all mammals harbour some methanogens, and produce some CH₄, until consistently proven otherwise. Hence, differences between species are most likely only of a quantitative nature.

Why harbour methanogens?

The likely ubiquitous presence of methanogens in digestive microbiota raises the question about their value for the host. Is their presence the consequence of a convergent adaptation of herbivores in the sense that they provide an adaptive advantage? Or can their presence simply not be avoided because the host animal does not have the means to control the composition of its microbiota? In trying to answer these questions, the loss of ingested energy via CH₄ and the function of methanogens as efficient hydrogen (H₂) removers need to be weighed against each other.

In humans, the presence of methanogens/CH₄ emissions is linked to longer digesta retention and higher human body mass (BM) (Nakamura et al., 2010), suggesting that methanogens improve the digestive efficiency of the gut microbiota. The presence of CH₄ delays peristaltic action in the exenterated dog or guinea pig small intestine (Pimentel et al., 2006; Jahng et al., 2012). For the large intestine – the main site of microbial action in monogastric species – no direct in vitro effect of CH₄, but a passage-accelerating effect of H₂, was demonstrated in guinea pigs (Jahng et al., 2012). By removing H₂, methanogens might thus prevent expeditious gut clearance. In mice, the presence of methanogens led to a more efficient carbohydrate digestion and higher body fat stores (but not higher BM) at similar food intakes, compared to animals without methanogens or those inoculated with a sulphate-reducing bacterium as an alternative H₂ sink (Samuel and Gordon, 2006). In rats, reducing methanogens led to a less efficient use of carbohydrates, but no change in BM over 6 weeks (Yang et al., 2016). Reducing methanogens/CH₄ emissions in humans has been associated with a decrease in digesta retention (Ghoshal et al., 2018). Accordingly, the main mode of action of methanogens appears to be H₂ removal that facilitates a higher rate of acetate production than with any other H₂ sinks, and a putative increase in digesta retention. Concerns about human obesity and constipation notwithstanding, the presence of methanogens in monogastric animals could be interpreted as facilitating an efficient resource use.

Most studies on CH₄ in mammals stem from domestic ruminants, where CH₄ is, by contrast, mainly considered an unavoidable, unwelcome loss of energy and contribution to greenhouse gas emissions. As in humans, intra-specific variation in CH₄ production in ruminants has been linked to digesta retention: animals with longer digesta retention, either due to more voluminous rumens at similar intake, or due to lower intake at uncontrolled rumen capacity, generally produce more CH₄ per ingested DM (Pinares-Patiño et al., 2003; Goopy et al., 2014; Hammond et al., 2014; Barnett et al., 2015; Cabezas-Garcia et al., 2017). Here, however, it is the retention time that is considered the cause, where more time available is considered responsible for more fermentative digestion with an ensuing increased CH₄ production. When manipulating digesta retention by other means than varying intake, namely by the addition of weights into the rumen, a shorter retention time in the rumen was correspondingly associated with a lower CH₄ production (Okine et al., 1989). On the other hand, when manipulating the CH₄ available in the rumen and accounting for variation in food intake in a cross-over study, a higher presence of CH₄ was associated with shorter retention and increased motility, possibly indicating a mechanism that aims at keeping losses at bay (Dittmann et al., 2016). These findings apparently contradict those made in monogastrics. From settings without cross-over design, no effect of CH₄ inhibition on digesta retention was reported (Nolan et al., 2010; Knight et al., 2011). Evidently, more work is required to understand the effects of CH₄ on peristalsis and digesta kinetics in ruminants.

In ruminants, the absence of methanogens in gnotobiotically raised animals (Fonty et al., 2007), or the chemical inhibition of methanogenesis, do not appear to have evident negative effects. Although on roughage diets food intake may be reduced, this does not translate into BM losses, but may, on the contrary, be linked with higher feed conversion efficiency (McCrabb et al., 1997; Hristov et al., 2015; Dittmann et al., 2016). However, natural variation in residual BM gain or residual feed efficiency was not related to CH₄ production (Freetly et al., 2015; McDonnell et al., 2016; Alemu et al., 2017), and selection for high feeding efficiency might even be associated with increased CH₄ yields (Flay et al., 2019). Yet, CH₄ inhibition has been reported to facilitate higher milk or milk protein yields (Abecia et al., 2012; Hristov et al., 2015). Therefore, the presence of methanogens in ruminants is considered somewhat similar to the presence of protozoa – most likely unavoidable, but with no or only minor disadvantages when lacking. The possibility remains that putatively positive effects of methanogens would only be detectable, for both monogastrics and foregut-fermenting animals, under certain conditions of reduced quality, natural forages.

The comparative approach

With respect to CH₄ physiology, the classification of species and species groups as emitters/non-emitters, or as high v. low emitters, has a certain tradition (Crutzen et al., 1986; Hackstein and Van Alen, 1996). Differences in the methanogenic potential of the microbiome of different herbivore species have been long acknowledged (Jensen, 1996; Fievez et al., 2001; Ouwerkerk et al., 2009), but the causes of these differences remain elusive. An evident outcome of comparative studies are predictive equations related to the scaling of CH₄ emissions with BM (Franz et al., 2011; Pérez-Barbería, 2017), with the intention to reconstruct CH₄budgets for fossil megafaunas (Smith et al., 2010; Wilkinson et al., 2012; Smith et al., 2016). Additionally, comparative approaches have favoured a dichotomy either between foregut-fermenting and monogastric herbivores (Jensen, 1996; Smith et al., 2015), or between ruminating and non-ruminating herbivores (Franz et al., 2011).

In the present study, data were collated on CH₄ emissions from in vivo measurements of herbivores fed diets consisting of forages, such as pasture, or grass or lucerne hay in whole, chopped or pelleted form, without the addition of concentrates, drawing on a collection of literature data and our own measurements. Objectives included a test for scaling of CH₄ emissions with BM, not only using all available emission data, but also those data for which food intake had been recorded in parallel, to assess whether CH₄ emissions actually scale differently than food intake and thus represent a putative disadvantage at increasing herbivore body size (Franz et al., 2010; Franz et al., 2011), or whether CH₄ emissions and food intake scale in parallel. Other objectives were testing differences between foregut and hindgut fermenters, and between ruminating and non-ruminating herbivores. Based on intra-specific findings in domestic ruminants, negative relationships on an inter-specific level between the level of total food intake and CH₄ yield were expected, as well as positive relationships between the digesta mean retention time (MRT) or gut capacity and the CH₄ yield.

Methods

A literature data compilation (Franz et al., 2010; Franz et al., 2011) was expanded to comprise the sources indicated in Supplementary Material Table S2. Only diets made of forages (either fed whole, chopped or pelleted) were accepted, and the minimum information required was BM and CH₄production. A variety of measurement techniques was included, mainly chamber respirometry, and the SF₆ tracer method where CH₄ production is estimated from the ratio of CH₄ to a tracer gas that is released from the rumen at a known rate. When available, data on DM intake (DMI), diet composition (including NDF), digestibility coefficients, MRTs of solute or particle markers in the gastrointestinal tract and CO₂ production were noted. Whenever possible, missing values were calculated (e.g., if CH₄ yield per DMI and DMI were specified, absolute CH₄ emission was calculated). Transformations of units (grams to litres, or litres to joules) were made using standard conversion factors (Brouwer, 1965).

Hitherto unpublished results derived from experiments (of which in some cases other measures than CH₄ have been published) were included as well. They include CH₄ measurements in various ruminants (gazelles down to a body size of dikdik Madoqua saltiana of approximately 1.5 kg), hystricomorph and arvicoline rodents, a giant rabbit breed, and in some cases retention time data for specimens whose intake, digestibility and CH₄ measurements have already been reported (Supplementary Material Table S2). The experimental methods followed those described, for example, in Dittmann et al. (2014) or Hagen et al. (2019). With data on DMI, DM digestibility and small particle MRT in the total gastrointestinal tract, the DM gut fill of the specimen was calculated according to the occupancy principle (Holleman and White, 1989; linear approach). Species were classified as ruminants or nonruminants, and as foregut or hindgut fermenters. The full dataset is provided as a supplement.

First, all available data for a set of measures (starting from absolute CH₄ emissions and BM) or the corresponding averages per species were used. Subsequently, entries that did not qualify for the next step, for example, did not include data on food intake, were excluded, and again analyses were performed on the total of the remaining data and the corresponding species averages. Results are thus reported for absolute CH₄ emission (the whole dataset), for CH₄ yield (in % of gross energy intake (GEI)), for CH4 : CO2 ratios, for parallel measures of DMI, for parallel measures of intake and fibre digestibility, and for parallel measures of intake and digesta retention.

To assess scaling relationships and their exponents, log-transformed data were submitted to linear regression in R v 3.3.2 (R_Core_Team, 2015) with the package ‘nlme’ (Pinheiro et al., 2011) in generalised least squares (GLS), for all available data and the species averages, indicating the 95% confidence intervals (CIs) for parameter estimates. To account for phylogeny, species averages were additionally analysed by phylogenetic generalised least squares (PGLS) with package ‘caper’ (Orme et al., 2013), using a mammalian supertree (Fritz et al., 2009), pruned to include the relevant taxa in our dataset. To retain the two rabbit breeds, the wapiti (contrasted to the red deer) and the alpaca, these groups were linked to the closest relatives of their original species in the tree (Bunolagus, Rucervus and Vicugna, respectively). The strength of the phylogenetic signal (λ, varying from 0 to 1, indicating phylogenetic structure in the dataset) was estimated by maximum likelihood. The analyses were repeated including species classification as ruminant or nonruminant (Rum) and as foregut or hindgut fermenter (Fore), first including the respective interactions with the independent variable (e.g., when relating absolute CH₄ production to BM, Rum, Fore, and the BM–Rum and BM–Fore interactions were included). If the interactions were not significant, the model was repeated without them. Scaling relationships for four (non-exclusive) herbivore groups (functional ruminants including camelids, taxonomic ruminants without camelids, foregut fermenters including ruminants and hindgut fermenters) are given in the Supplementary Material (for species overlap between groups, see Supplementary Material Table S2). For the dataset that included measurements of MRT, individual relationships were only calculatedfor small (lagomorphs, rodents) and large (all other) herbivores. The significance level was set to 0.05. Additional explanations on the statistical approach are given in the Supplementary Material S1.

Results

Complete dataset (absolute CH₄ emission and body mass)

Using the complete dataset and species averages, absolute CH₄ emission (L/day) had a significant phylogenetic signal and scaled to BM^0.84, with the 95% CI for the exponent ranging from 0.77 to 0.92 (Table 1). Using all individual data instead of species averages led to a slightly steeper scaling (BM^0.96) that did not include linearity in the 95% CI. In GLS, when using species averages, being a ruminant had a positive, and being a foregut fermenter a negative effect on CH₄ emissions. Nevertheless, the 95% CI of the parameter estimates for the scaling relationships of ruminants, nonruminants and hindgut fermenters overlapped (Supplementary Material Table S3). The 95% CI of the scaling exponent included linearity only in taxonomic ruminants. The single elephant measurement, horses, macropods and rabbits were on a generally lower level than ruminants, whereas many hystricomorph rodents as well as the nonruminant foregut fermenters peccary and pygmy hippo had levels similar to those of similar-sized ruminants (Figure 1a).

Table 1 Scaling relationships in mammalian species between CH₄ (in L/day or as % GEI) or the CH₄ : CO₂ ratio and body mass (BM, kg) according to y = a BM^b

GEI = gross energy intake; BM = body mass; 95% CI = 95% confidence interval.

Using either all individual values (all) in generalised least squares (GLS), or species averages (av) in GLS or in phylogenetic generalised least squares (PGLS, indicating the phylogenetic signal λ). Significant parameter estimates as well as λ significantly different from 0 are set in bold. Results of additional models that include whether a species is a ruminant or nonruminant (Rum) or a foregut or hindgut fermenter (Fore) are indicated in their direction (if interaction terms were nonsignificant, models were repeated without them). Scaling relationships for individual species groups are given in Supplementary Material Tables S3 to S5.

Figure 1 Relationship of body mass (BM) and (a) absolute daily CH₄ emission, (b) CH4 yield (in % gross energy intake), (c) the CH₄ : CO₂ ratio in the data collection of the present study. Domestic ruminants comprise goat, sheep and cattle. For a complete list of species, cf. Supplementary Material Table S2; for statistics, Table 1. Note that while horses, macropods (kangaroos) and rabbits generally have lower values than ruminants, hystricomorph rodents as well as the nonruminant foregut fermenters peccary and hippo are in the ruminant range. GEI = gross energy intake.

CH₄ yield in % gross energy intake

Whereas CH₄ (in % of GEI) showed some scaling with BM when all individual data or species averages were used in GLS, indicating that ruminants were on a generally higher level, this was not the case when accounting for phylogeny, indicating no scaling within related groups (Table 1). Correspondingly, there was also no scaling of CH₄ (in %GEI) in any of the four individual herbivore groups (Supplementary Material Table S4). Generally, CH₄ losses (in %GEI) appear to be constrained to a maximum of 10% (Figure 1B).

CH₄ : CO₂ ratio

The CH₄ : CO₂ ratio showed some BM scaling when phylogeny was not controlled for. There were significant interactions between BM and digestion types when all individual data were used, and indication that ruminants generally have higher ratios using species averages, even when accounting for phylogeny (Table 1). Within herbivore groups, this measure again showed no scaling (Supplementary Material Table S5). Ruminants, camelids and hystricomorph rodents as well as hippos can achieve high ratios of 0.06 and higher, whereas equids, macropods and rabbits appear limited to lower ratios; therefore, average estimates for nonruminants or hindgut fermenters are lower than those of ruminants (Supplementary Material Table S5). At BM below 1 kg, measured ratios appear limited to below 0.04 (Figure 1C).

CH₄ in relation to DM intake

Dry matter intake scaled to an exponent close to metabolic BW, with scaling exponents for ruminants being higher than those of nonruminants or hindgut fermenters (Tables 2 and Supplementary Material Table S6). Apart from the comparatively high intake in the smallest species, arvicoline rodents, no deviation from the overall pattern was apparent (Supplementary Material Figure S1A). The absolute CH₄ emission scaled very similarly to DMI with overlapping 95% CI for the scaling exponents when using species averages, when controlling for phylogeny or for individual herbivore groups (Tables 2 and Supplementary Material Table S6). Correspondingly, the scaling of absolute CH₄ emissions to DMI included linearity in the 95% CI for the scaling exponent in all groups (Tables 2 and Supplementary Material Table S6). Only when using all individual data or species averages without controlling for phylogeny, this scaling exceeded linearity, due to an effect of ruminants in these datasets. Similar to the CH₄ yield (in % gross energy), the CH₄ yield (per DMI) did not scale with BM, again except when using all individuals or uncontrolled averages, because of the ruminants (Supplementary Material Figure S1B). When using all individuals, and when controlling averages for phylogeny, there was a negative relationship between the relative DMI and CH₄ yield (per DMI), with ruminants on a higher level than nonruminants (Table 2). In datasets using all individuals, this effect was driven by the ruminants, and in datasets using species averages, by the arvicoline rodents (Supplementary Material Figure S1C).

Table 2 Scaling relationships in mammalian species between DMI or CH₄ (in L/day or as L/kg DMI) and BM or absolute or relative DMI according to y = a BM^b

DMI = dry matter intake; BM = body mass; 95% CI = 95% confidence interval.

From experiments where CH₄ and intake were measured in parallel (n = 573, 34 species), using either all individual values (all) in generalised least squares (GLS), or species averages (av) in GLS or in phylogenetic generalised least squares (PGLS, indicating the phylogenetic signal λ). Significant parameter estimates as well as λ significantly different from 0 are set in bold. Results of additional models that include whether a species is a ruminant or nonruminant (Rum) or a foregut or hindgut fermenter (Fore) are indicated in their direction if they were significant. Scaling relationships for individual species groups of this dataset are given in Supplementary Material Table S6.

CH₄ in relation to digested fibre

In the dataset for which NDF intake and digestibility were available, DMI and absolute CH₄ emissions again scaled closely to each other, and the scaling of NDF intake was similar as that of DMI, as was the scaling of digestible NDF intake (Supplementary Material Tables S7 and S8). Neither NDF digestibility (Supplementary Material Figure S2A) nor CH₄ yield (per kg digested NDF) scaled with BM (Supplementary Material Tables S7 and S8), with the exception of the functional ruminants, in which this scaling was negative (due to the inclusion of Bactrian camels in this subset). Absolute CH₄ emissions scaled to digestible NDF intake below linearity when controlling for phylogeny (Supplementary Material Table S7), also in ruminants but not in nonruminants or hindgut fermenters (Supplementary Material Table S8). There was a negative relationship between NDF digestibility and CH₄ yield (per digested NDF) when accounting for phylogeny (Supplementary Material Table S7), and the effect occurred also within the nonruminants and hindgut fermenters (Supplementary Material Table S8). In all groups, there was a negative relationship between the relative intake of digestible NDF and CH₄ yield (per digested NDF) (Supplementary Material Figure S2B, Tables S7 and S8).

CH₄ in relation to digesta retention and gut capacity

In the dataset of experiments in which the MRT of small particles (and solute markers) had been measured in parallel to CH₄ measurements, DMI and absolute CH₄ emissions again scaled nearly identically when controlling for phylogeny (Supplementary Material Table S9). The scaling of small particle MRT with BM differed distinctively between models that did not and did account for phylogeny (Supplementary Material Table S9), because MRT increased particularly with BM among rodents but less so across larger herbivores (Supplementary Material Figure S3A and Table S10). Dry matter gut fill scaled slightly below linearity (Supplementary Material Table S9 and Figure S3B). While absolute CH₄ emissions scaled positively with MRT in all datasets, MRT was less clearly related to CH₄ yield (per DMI) (Supplementary Material Tables S9 and S10). Although there was a positive relationship between the two measures (Supplementary Material Table S9), rabbits, macropods and horses had lower CH₄ yields at similar MRT than other species, and in particularly hystricomorph rodents were very variable in this relationship (Supplementary Material Figure S3C). The relative DM gut fill did not scale with CH₄ yield (Supplementary Material Tables S9 and S10), and again, rabbits, macropods and horses had lower CH₄ yields at similar relative gut fills than other species (Supplementary Material Figure S3D).

In an expanded model, using CH₄ yield (per DMI) as the dependent variable and not only small particle MRT but also the ratio of MRTparticles:MRTsolute as a covariable, both MRT and the ratio had a positive relationship when using all individuals (P < 0.001 and 0.014, respectively), but only MRT was significant when using species averages (P = 0.049 and P = 0.635), and there was no significance when accounting for phylogeny (P = 0.168 and P = 0.975). The effect in the dataset with individual data was considered due to the presence of rodents and in particular lagomorphs, that have very small MRTparticles:MRTsolute ratios due to their wash-back colonic separation mechanism. When repeating the analyses for large herbivores only, neither MRT nor the ratio were significant in any model (individual data, species averages, phylogeny control; P always >0.233).

Discussion

The present study represents a compilation of in vivo measurements of CH₄ emissions in mammals and shows up scaling relationships with BM, food intake, digesta retention times and gut capacity that lead to modifications of existing concepts. This holds in particular for the presumed dichotomy between foregut and hindgut fermenters, or between ruminant and nonruminant herbivores, which is not consistent in the dataset. Rather, individual herbivore species – in particular horses, macropods (kangaroos) and rabbits stand out as peculiar with respect to their low CH₄ emissions.

Limitations

Typical limitations of comparative data compilations occur, where measurements from different researchers are compiled in one catalogue. Because of the well-known effect that diets with concentrates lead to lower CH₄ emissions, the present data compilation included only experiments in which animals were fed diets without concentrates, excluding for example the hyrax or the sloth (cf. Supplementary Material Table S1). The case of the hyrax data (von Engelhardt et al., 1978) provides an instructive example: the animals of that study had not only been fed mixed diets, but also been fasted prior to respiration measurements; nevertheless, the data had been used prominently by ourselves (Franz et al., 2011) and others (Smith et al., 2010; Smith et al., 2015) for the establishment of hindgut fermenter-specific regression equations.

There was evident variation in the fibre content of the roughage diets used in the different experiments. Across experiments, such variation may contribute to reductions of CH₄ yield per digestible fibre, when fibre digestibility varied as a function of diet. Potentially, additional analyses could account for variation in dietary fibre levels, or in variation between CH₄ measuring methods. Both would have made the approach in the present study additionally complex, but the information is retained in the supplementary datafile and can be used in further approaches.

Scaling relationships may not be appropriately captured in simple allometric power functions, as in the log–log regressions performed in the present study. Various examples of ‘quadratic’ scaling exist (Müller et al., 2012), including measures of food intake and digesta retention in mammalian herbivores (Müller et al., 2013). The data from the present study, which are largely independent from those used by Müller et al. (2013), indicate a similar pattern with a particularly high intake, and short retention, in arvicoline rodents (Supplementary Material Figures S1A and S3A), and in the different scaling exponents for several measurements between small and large herbivores (Supplementary Material Table S10). While our approach of simple allometries, using various measurements taken simultaneously for the same animals, can be used to compare species groups, a more detailed model using quadratic scaling may be more appropriate to extrapolate data for species not measured, but within the BM range of our data compilation. Actually, Smith et al. (2015) developed a complex regression equation for the extrapolation of CH₄ emissions of hindgut-fermenting mammals. However, that regression estimation should be considered with caution, because the hyrax data mentioned above (taken from fasted animals) represent the data entry of the lowest BM in their collection. In a similar way, predictive equations for carbon isotope signature related to CH₄ physiology for hindgut and foregut fermenters presented by Tejada-Lara et al. (2018) depend critically on individual data points at the low end of the body size range.

One important limitation in the current dataset is the derivation of gut capacity as DM gut fill from data on intake, digestibility and retention following Holleman and White (1989). Although the method has been validated, and deviation from measurements by dissection can be explained by variable food intake prior to slaughter (Munn et al., 2012), there is a systematic underestimation of the real DM gut fill in ruminants if retention time is measured (as in the current dataset) by a small particle marker (Munn et al., 2015). However, Supplementary Material Figure S3D indicates that even if data for ruminants would be increased (following Munn et al., 2015, by a factor of 1.2), the explanatory power of gut fill for differences in CH₄ physiology would not increase.

Scaling with body mass: no size constraint

In previous work of our group (Franz et al., 2010; Franz et al., 2011) and of others (Smith et al., 2010), linear scaling relationships between CH₄ emissions and BM had been suggested. Given that energy intake by herbivores is generally not assumed to scale linearly with BM but to a lower (allometric) exponent, the logic implication was to construe a CH₄-driven body size limit for herbivores, because at some point, energetic losses as CH₄ would become prohibitive (Clauss and Hummel, 2005; Clauss et al., 2013). Correspondingly, positive scaling relationships of CH₄ yield (per unit of energy intake) were detected (Franz et al., 2011), with the attractive side effect that the models indicated that ruminants would reach any putative threshold at lower BM than hindgut fermenters, thus offering an explanation for an apparent, intrinsic body size limitation in ruminants (Clauss et al., 2003).

This concept needs to be revised. An analysis of ruminant data by Pérez-Barbería (2017) already questioned whether CH₄ emissions really scaled higher than food intake. Consistent with those findings, our dataset not only suggests that CH₄ yield is not related to BM, but that in subsets where food intake was measured in parallel to CH₄ emissions, both measures show a more or less identical scaling (Tables 2, Supplementary Material Tables S6, S7, S9 and S10). Correspondingly, when scaling CH₄ emissions against intake, a linear relationship results. Current evidence suggests that the process of methanogenesis resulting from fermentative digestion does not represent a body size limitation.

Digestive strategy: no clear dichotomies

In contrast to a seemingly clear dichotomy between ruminant and nonruminant herbivores (Crutzen et al., 1986; Franz et al., 2011) or between ruminants/foregut fermenters and hindgut fermenters (Smith et al., 2010; Smith et al., 2015; Tejada-Lara et al., 2018), the current data indicate that no simple categories might apply to classify herbivores in relation to CH₄. The historical intermingling of the terms ‘ruminants’ and ‘foregut fermenters’ notwithstanding (Clauss et al., 2010), the CH₄ literature does not treat these categories consistently, for example, when the hippopotamus (a nonruminant foregut fermenter) CH₄ emission is extrapolated based on ‘nonruminants’ by Crutzen et al. (1986) and based on ‘ruminants’ by Smith et al. (2015). No simple rule can be established: Among the nonruminant foregut fermenters, macropods have particularly low comparative CH₄ emissions, peccaries and hippos are at the lower range of ruminants, and the sloth – excluded from the present data analysis to remain consistent as to the diets used – is in the upper range of ruminants even on a low-fibre diet (Vendl et al., 2016b). Similarly, no clear pattern seems to apply for the hindgut fermenters. Although the dichotomy between horses and ruminants is evident, additional measurements of other hindgut fermenters, mainly by our group (Table 2), do not yield a simple pattern, but indicate that several hystricomorph rodent species can have CH₄ emissions of a magnitude expected for ruminants of similar size. In Figure 1A, the absolute CH₄ emissions of the smallest gazelle, the dikdik, and the hystricomorph rodent nutria, are identical. While a focus on ruminants with respect to mitigation strategies follows logically from their great relevance as food producing production animals with an enormous greenhouse gas footprint (Steinfeld et al., 2006), these findings indicate that one should not consider the ruminant digestive tract as the only one capable of harbouring intensely productive methanogenic microbiota.

One possible approach to determine the CH₄ strategy of a larger number of species could be the diet–bioapatite carbon isotope enrichment offset. Studies in which such data have been applied to herbivores have implied a difference in CH₄ emissions between herbivore digestion types (Codron et al., 2018; Tejada-Lara et al., 2018). However, the data from Codron et al. (2018) suggest little differences between the hindgut fermenters rock hyrax, black and white rhinoceros (Diceros bicornis, Ceratotherium simum), warthog (Phacochoerus africanus) or even a zebra species (Equus quagga) and ruminants. The data from Tejada-Lara et al. (2018) potentially corroborate the finding of CH₄ emission in arvicoline rodents (voles) of the present study and indicate little difference between the hindgut fermenters African elephant (Loxodonta africana), the black rhinoceros, the horse, a zebra species (Equus burchelli) and the pig and the ruminants giraffe (Giraffa camelopardalis), Bactrian camel and guanaco (Lama guanicoe). At other places in the body size range of that study, little differences appear evident in foregut-fermenting sloth species (Choloepus hoffmanni and Bradypus variegatus) paired each with a hindgut fermenter (the rabbit and the koala, respectively). These datasets might be better comprehensible if no clear separation between general digestion types would be assumed.

Relationships with intake and digesta retention

Abandoning clear categories of CH₄-producing digestion types raises the question whether other rules can be gleaned from the comparative dataset. There are well-established relationships between food intake or digesta retention and CH₄ yield in domestic ruminants (Okine et al., 1989; Lassey et al., 1997; Barnett et al., 2012; Hammond et al., 2014; Barnett et al., 2015) and in individual groups of nondomestic species (Frei et al., 2015; Vendl et al., 2015; Vendl et al., 2016a). In addition, the intra-individual variation in CH₄ emission in domestic ruminants is explained by differences in digesta retention, possibly linked to digestive tract capacity (Pinares-Patiño et al., 2003; Goopy et al., 2014; Cabezas-Garcia et al., 2017). Therefore, species were expected to vary in their CH₄ yield depending on their species-specific food intake levels, digesta retention times and DM gut fill. The first two predictions were met (Supplementary Material Figures S1C and S3B); however, the respective datasets failed to provide an explanation why certain species, such as horses, macropods or rabbits, had lower CH₄ yields at similar intake levels or retention times than others.

Additionally, the combination of an apparent evolution of a high fluid throughput or ‘digesta washing’ in macropods, and results from various in vitro studies led to an additional hypothesis (Vendl et al., 2015): A large difference in the MRT of fluids and particles (measured as a high MRTparticle:MRTsolute ratio and leading to a high outwash rate of microbes from the fermentation chamber) might create conditions favourable of a microbiome tuned towards growth rather than CH₄ production. However, in spite of a similar pattern of fluid and particle retention in pygmy hippos as in macropods (Supplementary Material Figure S4), no similarity in CH₄ emissions were evident between these species. When adding the ratio between particle and fluid retention to a regression of CH₄ yield against particle retention, no significant relationship resulted.

Clauss and Hummel (2017) suggested that adaptations in so-called ‘cattle-type’ ruminants towards a high fluid throughput might have a similar effect, and that variation in the amount of fluid passing through the rumen, because of differences in saliva production, might contribute to inter-individual variation in CH₄ emissions. The authors specifically referred to evidence from breeding experiments against the susceptibility to frothy bloat in cattle, which is considered linked to low saliva production of individual animals (Gurnsey et al., 1980; Morris et al., 1997). The assumption of Pinares-Patiño et al. (2008) that bloat-susceptible cattle, with less fluid flow, have higher CH₄ yields, based on the observation of slightly higher proportions of CH₄ in rumen gases of bloat-susceptible animals (Moate et al., 1997), would match this hypothesis. However, the same authors demonstrated no difference in CH₄ yields between cattle of low and high bloat susceptibility (Pinares-Patiño et al., 2008). Additionally, Grandl et al. (2018) did not find an association of inter-individual CH₄ yield differences and the MRTparticle:MRTsolute ratio in cattle. Should the CH₄-saving effect of increased fluid throughput that is so evident in in vitro fermentation experiments (Isaacson et al., 1975; Pfau et al., 2019) occur in herbivores in vivo, it remains to be demonstrated.

Relationship with fibre digestion

Similar to the moderate scaling of digesta retention with BM detected in the present study, the absence of a body size effect on the digestibility of NDF across herbivore species is in congruence with previous findings (Müller et al., 2013; Steuer et al., 2014). The results indicate that the amount of CH₄ produced per unit of fibre digested is not necessarily constant. The scaling relationship between digestible NDF intake and absolute CH₄ emission included linearity in nonruminants and hindgut fermenters, but this was not the case in ruminants (Supplementary Material Table S8), where digested NDF translated to a less-than-linear scaling (albeit on a generally higher level) into CH₄. In all data subsets, the relative intake of digestible NDF was negatively related to the CH₄ yield per unit of digested fibre, and in some subsets, a similar negative CH₄ yield scaling was evident with NDF digestibility. While it cannot be excluded that these patterns reflect differences in diet, they also hold for some groups in Supplementary Material Figure S2B fed a consistent diet for the measurements, and thus resemble a finding made in three different cattle feeding groups by Grandl et al. (2018). It is tempting to speculate that conditions leading to a higher fibre digestibility are linked to microbiota functions that are characterised by fibre digestion at reduced CH₄ production. While promising, the pattern again cannot explain fundamental differences in CH₄ emission between all species.

Conclusion and outlook

While many of the species investigated by our team have only been assessed once, results on domestic ruminants, South American camelids, horses, macropods and rabbits have been reproduced in more than a single study, suggesting that CH₄-related characteristics represent repeatable, species-specific characteristics. Given the absence of generalisable patterns, all that is left is the suggestion that yet-to-be-defined, species-specific characteristics determine the composition and activity of the microbiota of herbivores, and that across species, these characteristics may not be linked to digesta retention mechanisms of gut anatomy. Such a host specificity has been shown within cattle (Weimer et al., 2010; Weimer et al., 2017) or humans (Goodrich et al., 2014), or across species in a famous cross-over experiment with mice and fish microbiota (Rawls et al., 2006). Differences in the methanogenic potential of the microbiota of cattle have been identified (Zhou et al., 2009; Danielsson et al., 2012; Ben Shabat et al., 2016), and studies indicate the heritability of the microbiome composition, that is, its genetic control (Goodrich et al., 2014; Roehe et al., 2016). These results open the possibility of selecting domestic animals for their microbiome composition. Yet, they also raise again the question of the Introduction why, if microbiome and hence methanogen control by the host is possible, evolutionary adaptations have not led to the exclusion but to a prominent role of methanogens in many herbivore species.

Our study suggests that CH₄ production may be more uniform across many herbivore species, that perceived differences between digestion types are due to a historical focus on certain animal groups, and that ruminants should possibly not be considered peculiar in this respect. Further studies elucidating the peculiar conditions that result in the comparatively low CH₄ emissions in macropods, equids, and, possibly, rabbits are warranted. For a reliable reconstruction of past CH₄ budgets, the use of domestic equids as model animals for ‘hindgut fermenters’ must be questioned, and additional measurements in other, non-rodent hindgut fermenters would be needed. This could be rhinoceroses, tapirs, (forage-fed and non-fasted) hyraxes, or pigs fed forage-only diets. Finally, the relevance of methanogens for the digestive physiology of primates remains to be explored.