Animal board invited review: genetic possibilities to reduce enteric methane emissions from ruminants

Measuring and mitigating methane (CH4) emissions from livestock is of increasing importance for the environment and for policy making. Potentially, the most sustainable way of reducing enteric CH4 emission from ruminants is through the estimation of genomic breeding values to facilitate genetic selection. There is potential for adopting genetic selection and in the future genomic selection, for reduced CH4 emissions from ruminants. From this review it has been observed that both CH4 emissions and production (g/day) are a heritable and repeatable trait. CH4 emissions are strongly related to feed intake both in the short term (minutes to several hours) and over the medium term (days). When measured over the medium term, CH4 yield (MY, g CH4/kg dry matter intake) is a heritable and repeatable trait albeit with less genetic variation than for CH4 emissions. CH4 emissions of individual animals are moderately repeatable across diets, and across feeding levels, when measured in respiration chambers. Repeatability is lower when short term measurements are used, possibly due to variation in time and amount of feed ingested prior to the measurement. However, while repeated measurements add value; it is preferable the measures be separated by at least 3 to 14 days. This temporal separation of measurements needs to be investigated further. Given the above issue can be resolved, short term (over minutes to hours) measurements of CH4 emissions show promise, especially on systems where animals are fed ad libitum and frequency of meals is high. However, we believe that for short-term measurements to be useful for genetic evaluation, a number (between 3 and 20) of measurements will be required over an extended period of time (weeks to months). There are opportunities for using short-term measurements in standardised feeding situations such as breath ‘sniffers’ attached to milking parlours or total mixed ration feeding bins, to measure CH4. Genomic selection has the potential to reduce both CH4 emissions and MY, but measurements on thousands of individuals will be required. This includes the need for combined resources across countries in an international effort, emphasising the need to acknowledge the impact of animal and production systems on measurement of the CH4 trait during design of experiments.


Implication
Measuring and mitigating methane (CH 4 ) emissions from livestock is of increasing importance for the environment and for policy making. Potentially, the most sustainable way of reducing enteric CH 4 emission from ruminants is through the estimation of genomic breeding values to facilitate genetic selection. Enteric CH 4 emissions are difficult and expensive to measure, thus genomic prediction could provide significant,

Introduction
Climate change is of growing international concern and it is well established that the release of greenhouse gases (GHG) is the driving factor (IPCC, 2006). Globally, livestock farming contributes~9% to 11% of total anthropogenic GHG emissions (Smith et al., 2007;Tubiello et al., 2013). Of the various GHG, methane (CH 4 ) is the most important agricultural contributor, with a global warming potential 25 times that of carbon dioxide (CO 2 ) (Forster et al., 2007).
Globally, in the year 2010, GHG emissions from the agriculture sector accounted for 4.6 GtCO 2 eq, of which enteric fermentation (emissions of CH 4 from ruminant animals) contributed 2 GtCO 2 eq (Tubiello et al., 2013), with an annual increase of 0.95% between 1961 and 2010. Non-dairy cattle were the single largest source of enteric CH 4 , followed by dairy cattle, buffaloes, sheep and goats (FAOSTAT, 2013). Enteric CH 4 emissions from ruminant livestock (cattle, sheep and goats) account for 2% to 12% of gross energy intake (Blaxter, 1962;Johnson and Johnson, 1995). Although CH 4 production is an energy loss to ruminants, it can also be considered a small price to pay for their adaptation to digest cellulose-based feeds. Sources of systematic variation in CH 4 production by an individual animal include: total feed intake, the nutrient composition of the feed eaten, the proportion and rate of fermentation of that feed in the rumen, feeding frequency (for recent reviews see Hristov et al., 2013a and2013b), rumen volume and rate of passage of digesta from the rumen (Goopy et al., 2014), physiological state of the animal and variation between individual animals including that between sire families (Pinares-Patiño et al., 2013a).
Production of CH 4 (and other GHGs) per unit of animal product (e.g. milk, meat) has declined over the past 50 years in most ruminant livestock industries in developed countries due to ongoing improvements in animal productivity. For example, the carbon footprint, in terms of CO 2 eq/kg of milk produced, of the US dairy industry in 2007 was 37% of that in 1944 (Capper et al., 2009). Productivity improvements included a change of breed type of the dairy cow (to Holstein), improved genetics within the Holstein breed and a shift from a forage based to total mixed ration feeding system (see Capper et al., 2009). Similarly, analysis of the carbon footprint of total US beef production indicates a reduction of CO 2 eq of 16% per kg of beef produced in 2007 compared with 1977 (Capper, 2011), due to a reduction in total feedstuff used, changed industry structure, improved nutritional management and improved herd genetics.
Most of the mitigation potential in the livestock sector is found in the developing countries. However, for these countries, it is important to combine development and mitigation strategies, like adapted selection programmes and feeding strategies, as a lot can still be achieved in developing countries by increasing lifetime production of animals (Gerber et al., 2013).
The extent to which genetic improvement can contribute to improvement in individual animal milk production and consequent impacts on GHG emissions has been highlighted by Wall et al. (2010). They described how systematic improvement in environmental outcomes has resulted from productivity improvements and discussed how direct and indirect measures of emissions can be incorporated into breeding objectives to reduce emissions.
There are many potential methods to reduce enteric CH 4 emissions per head and thereby intensity of CH 4 production per unit product. These include: changing feed type (e.g. from pasture to concentrate feed or to new pasture varieties); use of supplements that reduce CH 4 emissions (fats, oils, plant extracts and nitrate); improving productivity through management change including use of growth enhancers and improved genetics; immunisation against methanogens and selective breeding of animals with low CH 4 emissions, through either reduced feed intake per product or reduced CH 4 production per feed consumed, without compromising production characteristics (Martin et al., 2010;Wall et al., 2010). The aim of this review is to provide an overview of possibilities and some of the remaining issues that need to be addressed to realise these possibilities to genetically reduce enteric CH 4 emissions by livestock.

Quantifying enteric CH 4 emission
There are three levels in which a CH 4 trait can be defined; first, the farm system level which uses information on the number of animals present within a system boundary with a related estimate of CH 4 emissions per head, calculated for example from the Intergovernmental Panel on Climate Change (2006) Tier 2 calculations. These calculations have embedded within them a number of assumptions about the factors which affect CH 4 emission per head, that is feed intake, feed quality and CH 4 yield. Second, the animal production level which uses information about productivity per head that is milk yield or kg carcass weight, from individual animals to give us CH 4 intensity (g CH 4 /kg product). Finally, at the animal level, individual CH 4 emissions and feed intake measurements to enable genetic progress on CH 4 yield (MY; g CH 4 /kg dry matter intake (DMI)), or residual feed intake (RFI; MJ/day), which is the difference between net energy intake and calculated energy requirements for maintenance as a function of live weight and for fat and protein corrected milk yield.

Methodologies for measurement of CH 4 from ruminants
The respiratory chamber (RC) system is often viewed as a 'gold standard' for emission measurement. There is little question RC measurements accurately quantify CH 4 output over the 1 to 3-day measurement period typically used, and they achieve this by frequently measuring emissions. Pickering, Oddy, Basarab, Cammack, Hayes, Hegarty, Lassen, McEwan, Miller, Pinares-Patiño and de Haas The variability in emission rate resulting from eructation events, animal position and feed intake that occur in 24 h, are typically damped within the large chamber volume. Feeding in RCs can also cause a reduction in feed intake (relative to pre-chamber intakes) and completely eliminates diet selection and feeding pattern which has strong genetic control and may well be a means by which animal genetics moderates emission in the grazing environment (Hegarty, 2004). The RCs rarely monitor CH 4 outflow on a second by second basis, the chambers used to estimate CH 4 parameters do so by measuring volume of air flow coupled with intermittent sampling (at 3 to 13 min) of gas for determination of CH 4 concentrations. This means that hourly measurements described here consist of averages of 4 to 20 measurements each taken over a few seconds (albeit averaged via dilution in a large volume that is the chamber). As shown by Pinares-Patiño et al. (2013a) a 1 to 3-day collection only poorly describes the CH 4 phenotype of an animal over a year or a lifetime and could benefit from repeated measurements. In reality, CH 4 is largely emitted intermittently via brief eructations or burps lasting only seconds, albeit with a basal level of emission.
The sulfur hexafluoride (SF 6 ) technique is one tool that offers field measurement over a longer time, but requires insertion of rumen boluses, daily animal handling and laboratory measurement of gases (McGinn et al., 2006). Moreover, the sampling procedures provide an average CH 4 output for periods of typically 24 h, but can be repeated over periods of 5 to 10 days, or until the rate of release of SF 6 from the permeation tube is no longer stable. While repeatability of daily CH 4 production is being improved as the methodology is refined (Deighton et al., 2013), SF 6 remains a very demanding method to get accurate emission measures over multiple days in individual animals.
Other systems that measure (or estimate) emissions over multiple short periods per day with minimal operator input have been developed. These include measuring all emissions from animals in short-term confinement; that is, Portable Accumulation Chambers (PAC; Goopy et al., 2011), monitoring eructations in feeding stations (Negussie et al., 2012) or voluntary milking systems for dairy cattle (Garnsworthy et al., 2012;Lassen et al., 2012), or Greenfeed monitors (GEM). A hand-held laser has been used to estimate CH 4 flux indirectly from dairy cattle (Chagunda et al., 2013). All of these methods, except PAC and GEM, measure concentrations, and assume that they have a constant recovery or little drift, and are therefore accurately reflecting gross flux from the animal over the recorded period. Similarly, all short-term estimates also assume that there is a high genetic correlation with longer term measurements and that this is essentially independent of when the animals are recorded. Average CH 4 emissions in various units, heritability estimates, where known, and various repeatability estimates for example across days, across periods and across rounds are shown in Supplementary Table S1 for cattle and in Supplementary  Table S2 for sheep. There are a wide array of variables including; system (RC, SF 6 , laser, GEM or PACs), diet (composition and particle size), feeding level (ad libitum or at a proportion of maintenance) and experimental period. Despite this, gross CH 4 production and repeatability estimates are not so different. However, MY is variable with a noticeable difference between studies where animals are fed at a proportion of maintenance versus those that are fed ad libitum. Those fed at maintenance are theoretically estimating CH 4 emission per live weight as much as CH 4 emission per unit intake; CH 4 emissions increases with live weight, and thus the ratio measure could be similar across time points in maintenance fed studies.
In summary, daily enteric emission is principally constrained by the quantity and fermentability of the feed consumed; but an understanding of within-day and between-day variances is required to ensure the emission data collected reflects the long-term CH 4 phenotype of a ruminant. When collecting records for selective breeding, it will often be a choice between accuracy of the phenotype and number of records. In the case of gross CH 4 production the most accurate method would be the RC method, but in order to generate enough data to do selective breeding and make recordings in practice, this method has limitations. Alternately, compared to RC, spot breath samples taken during milking in dairy cattle might be less accurate phenotypes for selective breeding, but can generate a large number of individual animal records. A genetic and environmental correlation structure between these methods together with 1 h RC methods, SF 6 and other methods is needed and would allow merging of data to generate enough data for use in selective breeding.

Implications for measurement
Three messages on repeatability emerge from Supplementary Table S1 and S2. The repeatability of daily CH 4 emissions is highest between RC measures made on consecutive days, but diminishes as time between measures increases. Repeatability of CH 4 emission is lower for short term measurement systems (e.g. PACs) relative to RC system. Consequently, more measures will be required from shortterm measurement methods to capture variation within a day, but multiple samples across many days offers additional information about the robustness of the emissions phenotype that is not normally obtained by RC studies made only over 1 to 3 days. So far, we have not been able to source sufficient structured data from these methods and protocols to develop a common procedure for measurement of rate of CH 4 emissions capable of being used for genetic selection. McEwan et al. (2012) assessed the usefullness of multiple 1 h measures of emissions compared to 22 h RC measures using 684 sheep and found a high genetic correlation between 24 h emission measure and a 1 h emission measure (0.89 for g CH 4 /day and 0.76 for MY). They estimated there is little difference in estimates of CH 4 emissions and MY by measuring animals twice in a RC, 14 days apart, or by measuring an animal four times for 1 h, 14 days apart. Such assessments indicate that using a range of measurement technologies is possible, but the intensity of sampling Genetic possibilities to reduce enteric methane required and number of animals needing to be measured will be different for each system used.
It has been calculated that 3 × 1 h PAC measurements will be as useful at describing CH 4 production rate as one RC measure for 1 day (Bickell et al., 2011). Defining this comparability is a key requirement for developing measurement protocols of equivalent power to use in genetic selection.
Pinares-Patiño et al. (2011) showed that groups of animals selected to be high or low MY when consuming 2.2 × maintenance lucerne pellets retained their ranking when fed lucerne and concentrate pellets. Subsequently they (C.S. Pinares-Patiño personal communication) demonstrated that with five different diets the groups remained different in MY, although individuals in the groups sometimes re-ranked (Table 1). Similar results were obtained by Michal et al. (2013) from growing beef heifers fed three different diets. This suggests that using a standard diet to assess rank of animals for MY is useful and the rankings are likely to hold across a range of production diets. The data also suggest that the differences in MY between animals in high and low MY groups (and therefore individuals) are greater when they are eating a more digestible diet. This suggests that the discriminatory power of a phenotype test could be expanded by feeding a mixed ration of forage and concentrate, although this requires testing with more animals.

Breeding to reduce CH 4 emissions from livestock
Genetic selection provides a reliable route towards permanent and cumulative reductions in quantitative traits such as enteric CH 4 emissions.
To justify investment of effort and money in developing protocols for measurement of emissions to support genetic improvement in a CH 4 trait, it is worth summarising evidence supportive of this breeding strategy (Lassey et al., 1997). Genetic diversity in a range of digestive parameters likely to be associated with enteric CH 4 production was apparent when reviewed in 2002 (Hegarty, 2004). The prospect for selection for a CH 4 trait was initially investigated by multiple groups; some identified variation in CH 4 traits amenable to animal selection (Robinson et al., 2010) and some did not (Münger and Kreuzer, 2008). More recent research in 530 beef animals  and 1225 sheep (Pinares-Patiño et al., 2011 and is increasingly supportive of CH 4 traits being heritable with improvement by direct selection achievable. Based on records of 1277 pedigreed sheep, estimated heritability and repeatability of CH 4 across days, rounds and years, using the total 24 h measurement were 0.29 ± 0.05 and 0.13 ± 0.03 for gross CH 4 production (g/day), and MY (g /kg DMI), respectively (Pinares-Patiño et al., 2013a). There were high repeatabilities across consecutive days. Across rounds and across years the repeatability estimates were lower than for consecutive days, but, relatively stable. Estimation of genetic and phenotypic correlations with some of the main New Zealand production traits; weaning weight at 3 months, live weight at 8 months, fleece weight at 12 months (FW12), eye muscle depth and dag score (accumulation of faeces on the perineum region) at 3 or 8 months of age show that correlations with MY are low or close to zero, the only exception was FW12. The negative genetic and phenotypic correlations of FW12 with MY (−0.32 ± 0.11 and −0.08 ± 0.03, respectively) imply that selecting for increased hogget fleece weight would in part result in lower CH 4 yield.
Results from Donoghue et al. (2013) on Australian Angus beef cattle showed very similar heritabilities. Based on 530 pedigreed cattle, fed at a proportion of maintenance (1.2 ×), heritability estimates for gross CH 4 production (L/day), and MY (L/kg DMI) were 0.40 ± 0.11 and 0.19 ± 0.10, respectively. Genetic and phenotypic correlations of gross CH 4 production with eye muscle area were 0.17 ± 0.29 and −0.01 ± 0.05, respectively. With MY, the genetic and phenotypic correlations were −0.02 ± 0.30 and −0.03 ± 0.05, respectively.
Both studies are based on 24 h RC measurement with known feed intake. However, the cost of routinely measuring CH 4 emissions using RC is thought to be prohibitive for a testing programme using industry animals. Therefore, protocols for measuring or estimating CH 4 production and feed intake are required that need less time and cost. It has to be kept in mind that phenotype recording of feed intake or DMI is most limiting in commercial condition and generally only recorded on experimental farms.
In the longer term, it may be possible to incorporate genomic information to estimate genomic breeding values (GEBVs) for CH 4 emissions into breeding schemes (Meuwissen et al., 2013). For GEBVs to be implemented, a reference population of several thousand genotyped industry relevant animals, with the CH 4 phenotype measured, is required to provide initial estimates of the contribution of each genomic region to the expression of the phenotype under investigation (Calus et al., 2013). Similarly, selection on GEBVs for correlated indicator traits can be used where it is impractical to directly measure CH 4 on enough animals to establish a reference population. Finally, there must be an  (and/or social) incentive to breed animals with the trait which is incorporated in the selection objective, so that the CH 4 trait receives the appropriate weighting in any breeding programme. There is already on-going improvement in emissions intensity that is CH 4 emissions per unit product, arising from genetic selection for current production traits (Capper et al., 2009;Wall et al., 2010;Hayes et al., 2013). One could therefore argue that further research investment into this area (i.e. selection for reduced intensity of CH 4 emissions) is not necessary. However, selection solely on productivity traits such as live weight gain and/or milk production will increase feed intake and CH 4 emissions per animal and hence total CH 4 emissions unless a physical or economic constraint is imposed on total emissions. For dairy products, there is a market constraint on total production which has resulted in an increase in productivity per cow and a decrease in number of animals. This may suit some industries, but poses the question 'is it possible to increase productivity and reduce CH 4 emissions per animal at the same time?' This could be achieved by reducing MY that is CH 4 per unit feed consumed, and/or decreasing DMI provided that there is no concomitant reduction in productivity or increase in feed consumption. Selection on MY provides options to either reduce emissions while holding net enterprise feed consumption constant, or alternatively, allowing intake to increase supporting a production boost per animal without raising total emissions. Early results from a number of studies around the world, suggest that MY is both a heritable and repeatable trait (e.g. Pinares-Patiño et al., 2013a). However, the means by which the host influences fermentation in the gut to affect CH 4 production is still largely unknown. The extent to which genetic selection can be used to reduce MY is also not known. The methods by which CH 4 emissions of individual animals can be measured are an important factor because the method used to measure the CH 4 trait will also influence the resulting genetic parameters and is therefore an integral part of the selection programme. Besides, caution should be taken for ratio traits, as the genetic parameters may not truly represent the trait under consideration, because there is always extra variability of the denominator trait.
It is also important to remember that fertility and longevity have a huge aspect in the overall environmental impact of livestock, and therefore improved fertility and longevity through breeding and management will also be important mitigation strategies (Cottle et al., 2011).

Understanding animal variation in CH 4 production over time
Sources and transfer of CH 4 within the ruminant While CH 4 is produced in both the reticulo-rumen and the hindgut, some transfer within the animal occurs before the CH 4 is emitted. For example, in ewes eating lucerne, 97.5% of CH 4 emission was voided via the oesophagus and lungs and only 2.5% via flatus, despite 23% of CH 4 production occurring in the lower gut, presumably because of absorption of hindgut CH 4 into the blood (Murray et al., 1976). Cattle studies have shown the proportion of CH 4 derived from the hindgut increases with feeding level (Hofmeyr et al., 1984). Most of the CH 4 leaving the rumen in oesophageal eructation is subsequently drawn into the lungs and then emitted in exhaled breath; although some rumen produced CH 4 is also absorbed into the blood and diffuses into the lungs without passing up the oesophagus. This has been confirmed by dosing and radiotracer studies (Dougherty et al., 1964;Heywood and Wood, 1985). The fraction of CH 4 absorbed into the bloodstream from the gastrointestinal tract decreases as volume of eructated gas increases; also when an animal is not ruminating (Hoernicke et al., 1965); and also after feeding (Hoernicke et al., 1965). Studies with tracheotomised cattle have revealed that before feeding, 25% to 94% of the total CH 4 emission (flatus not included) was by exhalation, whereas after feeding exhalation is reduced to 9% to 43% of emissions.
Cattle eructate every 40 to 90 s and take between 25 and 40 breaths per minute (Mortola and Lanthier, 2005), although the frequency of eructation peaks is reduced when drinking (Hegarty, 2013). As breathing frequency in cattle oscillates within a day and varies largely between animals (Piccione et al., 2004), differences in gas excretion mechanisms (eructation, tracheal inhalation, exhalation and expiration) might differ considerably among individual animals.
While the proportion of CH 4 entering the lungs by absorption or by inhalation varies, the important value is the absolute quantity and constancy of CH 4 leaving the mouth and nose. Large oscillations in CH 4 release rate (but not necessarily methanogenesis rate) are observed during CH 4 measurements. Animal position and activity is known to affect pooling of gas in the rumen (McCauley and Dziuk, 1965), and pooling of gas in the rumen may be part of the reason that variable short term CH 4 production rates are seen during RC studies even from animals fed at 2 h intervals (e.g. Figure 1a: Nolan et al., 2010;Figure 1b: Mathers and Walters, 1982). Enteric CH 4 production rate varies widely over 2 h intervals (Figure 1b), potentially contributing to a highly variable estimate of emission rate if measurements are short term. Mathers and Walters (1982) acknowledged 'violent shortterm variations were evident in the plots of the observations', so emission rates were averaged, over various periods, to generate smoother emission profiles. Poor in-chamber mixing of air can cause similar variability in emission rates assessed over the short term (Gardiner and Coleman, 2013).

Diurnal and longer term emission cycles
In the grazing environment, ruminants are considered to ingest most of their feed in morning and late-afternoon feeding sessions (see Gregorini, 2012 for recent review). Emulation of this pattern in RCs (Robinson, 2009) shows a biphasic diurnal CH 4 emission pattern, consistent with timing of feed intake, but there was no difference in either total daily emission or MY when feed was provided in a single meal or as four equal meals in the morning and four equal meals in the afternoon. Murray et al. (2001) found a Genetic possibilities to reduce enteric methane similar pattern of biphasic emissions in grazing sheep using a polytunnel.
A number of studies offer evidence of repeatability of emissions over prolonged periods, but the repeatability is confounded by the variations in pasture cover that occur with changes in season (Knight et al., 2008;Münger and Kreuzer, 2008), so do not reflect innate repeatability of emission by the animal as would occur if the same diet was fed for a prolonged period.
Recent sheep genetics research provides evidence of repeatability over extended time intervals when a consistent diet is fed (Pinares-Patiño et al., 2013a) and confounding with changes in feed composition do not occur.

Indirect selection to reduce emissions
Measuring CH 4 emissions directly from animals is difficult and thereby hinders direct selection on reduced CH 4 emission. However, improvements can be made through selection on associated traits (e.g. RFI), volatile fatty acids (VFA), milk composition or through selection on CH 4 predicted from feed intake and diet composition.

VFAs
The rumen microbial population converts the host ingested food in the rumen into CO 2 , hydrogen (H 2 ), VFA and microbial cells. The host absorbs the VFA across the rumen for its own use and rumen methanogens act on the H 2 to produce CH 4 . High H 2 concentrations are thought to stimulate methanogenesis while suppressing production of acetate and VFA in general, while low H 2 concentrations will stimulate VFA production, especially acetogenesis but suppress methanogenesis. VFA are thus a potential proxy for estimating CH 4 emissions. For sheep, Pinares-Patiño et al. (2013b) measured 1081 animals for VFA soon after exit from RCs. There were high genetic correlations (>0.78) of MY with log e mM VFA concentrations. Genetic correlations are lower, but still moderate, when VFAs were expressed as molar %.
For cattle, Herd et al. (2013) measured VFAs and other parameters from 532 young Angus bulls and heifers soon after exit from the RCs (at least 12 h post feed consumption). Pearson correlation coefficients with CH 4 production (L/day), MY (L/kg DMI) and CH 4 intensity (L/kg live weight gain) were estimated. There were correlations of 0.40 with MY and CH 4 emission intensity, but correlations with gross CH 4 production were almost zero. Other studies (Robinson et al., 2010), suggest that VFA concentration has limited utility in predicting CH 4 emissions, although VFA production rate may be useful (McPhee and Hegarty, 2008). This contrasting evidence indicates considerable work is still required before the utility of VFA as an indicator of CH 4 emissions can be realised.
Prediction form mid-infrared spectra of milk samples Mid-infrared spectra (MIR) of milk samples are generated routinely by national and commercial laboratories for prediction of milk composition during milk recording. Therefore, any approach that utilizes this information can immediately be implemented but also applied retrospectively to already analyzed samples with the spectral data stored. In vivo experiments performed using the SF 6 method showed that it is possible to estimate CH 4 emissions of lactating dairy cows from MIR spectra of milk samples (Dehareng et al., 2012). A possible delay between a variation in CH 4 emission and an onset in milk response was mentioned by these authors. These preliminary results suggest the possibility to predict individual CH 4 emissions, allowing at least inventory type of assessments at a farm level or at a regional scale. With more collaboration and additional data, an improved equation could be generated. Predictions could then become robust enough to use MIR spectra to identify individually low-CH 4 -emitting cows and to develop selection and management tools to reduce CH 4 emissions.

Prediction from feed intake and diet composition
The objective of a Dutch study was to establish phenotypic and genetic variation in predicted CH 4 output, and to determine the potential that genetic selection has in reducing CH 4 emissions in dairy cattle (de Haas et al., 2011). Records on daily feed intake, weekly live weights and weekly milk productions were available from 588 heifers. Along with RFI, predicted CH 4 emissions (PME, g/day) and fat and proteincorrected milk production (FPCM, kg/day) were estimated. The estimated heritabilities for PME and RFI were 0.35 and 0.40, respectively. The positive phenotypic and genetic  Pickering, Oddy, Basarab, Cammack, Hayes, Hegarty, Lassen, McEwan, Miller, Pinares-Patiño and de Haas correlations between RFI and PME indicated that cows with lower RFI have lower PME as well (estimates ranging from 0.18 to 0.84 in different periods of the lactation). However, the association between these indicator traits and true CH 4 output is unknown. It is still possible to decrease CH 4 production of a cow by selecting more efficient (low RFI) cows, and the genetic variation suggests that reductions in the order of 11% to 26% in 10 years are theoretically possible, and in a genomic selection programme even higher. However, as stated previously, it is essential to ensure selection on production does not increase feed intake and CH 4 emissions per animal and hence total CH 4 emissions.
CH 4 in a genomic selection programme CH 4 emissions (as g CH 4 /day or MY) certainly fit the description of hard to measure traits. Methods currently available are expensive and time consuming (RCs and SF 6 ) and subject animals to artificial environments. Those that measure animals in production situations (pasture, feedlot or dairy feeding station) sample CH 4 for only a part of a day and require repeat measurements (PACs, Sniffers or GEM) and in some cases calculation back to known standard procedures. Those methods of estimating CH 4 emissions that rely on computation of differences between feeding standards and production account for only part of the potential variation in CH 4 emissions between animals.
Genomic selection opens the possibility to efficiently select for hard to measure traits. It is progressively being used to increase rate of genetic progress for production traits that are measured late in life (e.g. meat yield and quality), expensive to measure (e.g. RFI) and are sex linked (e.g. milk production and quality). In the dairy and increasingly in the beef and sheep industries leading sires are routinely genotyped and GEBVs are used in making selection decisions. It is doubtful that adding the cost of genotyping onto a population in which CH 4 is measured would be cost effective, but by using industry animals which have measured production traits and have been genotyped it would be possible to estimate GEBVs for CH 4 emissions. This is predicated on having a large reference population, where CH 4 emission levels can be measured cheaply and genome wide DNA marker effects have been estimated, to establish the prediction equation for marker effects.
The key question is how large does this reference population have to be, that is, how many animals need to be measured for CH 4 and genotyped with the genome wide marker panels? Daetwyler et al. (2008), Goddard (2008) and Hayes et al. (2009) have all derived deterministic formula to estimate the accuracy of GEBV that could be achieved given the size of the reference population, the heritability of the trait and the effective population size. The accuracy of genomic selection for selection candidates (i.e. animals with a genotype, but no measured phenotype) with increasing size of reference population is shown in Figure 2. This was derived from the heritability of MY of 0.13 (Pinares-Patiño et al., 2013a) and an effective population size of 150 using the procedure described by Hayes et al. (2009). This graph assumes perfect linkage disequilibrium between the single nucleotide polymorphisms (SNP) and quantitative trait loci, which is unlikely for the current available chips and thus the graph will asymptote to the proportion of variance explained. For example, for dairy cattle using the Bovine 50 K SNP chip this would be 90%. The estimates also assume unrelated individuals, if individuals were related, particularly the selection candidates and the reference population, the accuracy would be greater, as this is effectively reducing the effective population size. Finally, if the individuals in the reference population were progeny tested, this would make the 'heritability' of the trait much higher and thus would require fewer animals genotyped to achieve the same accuracy, however, the total number of animals measured for CH 4 to achieve the same accuracy would stay the same.
Because MY is a new trait, it would be anticipated that even low initial accuracy will be useful to industry. As further animals are phenotyped the GEBVs would become increasingly useful. It remains to be determined if MY is independent of other (production) traits. If it is, then adding information from the GEBVs for MY into a selection index is relatively straightforward.
The number of animals with phenotypes in the reference population required to obtain GEBVs of high accuracy for MY is large and almost certainly exceed the resources available in any one country. To overcome these limitations an international effort is required to bring together data on production, feed intake and CH 4 emissions of ruminants.

Potential reduction in CH 4
Although genetic selection is possible, the potential magnitude of selection for MY is unknown. Pinares-Patiño et al. Genetic possibilities to reduce enteric methane one generation of selection for and against MY. The extent to which variation in MY can be exploited, depends on the stability of the underpinning relationships with production traits. The best way to incorporate this is with a selection index that includes traits related to production, functional traits and environmental impact. This will result in a slower response to selection for all traits, but in a good overall response to the overall breeding goal. The mechanisms that contribute to genetic variation in MY of individual animals may include: reduced fermentation of organic matter in the rumen (due to shorter retention time of digesta; Pinares-Patiño et al., 2011 and smaller rumen volume; Goopy et al., 2014), instability of fermentation (natural occurring defaunation; Faichney and Graham, 1996), different microbial populations in the rumen and potentially reductive acetogenesis (inferred from Faichney and Graham, 1996). The extent to which these combine to produce natural variation in MY is unknown, but data from measurement of MY in sheep using RCs suggest that the coefficient of variation is 10.3% (Pinares-Patiño et al., 2013a) and for cattle 14% . It would not be unreasonable to anticipate a response to long term selection to exceed 2 standard deviations from the mean, suggesting that a reduction of up to 25% in MY may be feasible through selection of livestock for low MY. Combined with potential reduction in CH 4 emissions due to selection for low RFI, this suggests that a reduction in CH 4 emissions of 40% to 45% may be possible through selection of individual animals on components that directly affect CH 4 production. Differences in feed intake of 1.17 kg/day between beef cattle selected for and against RFI were observed after 2.4 generations, equivalent to a difference of 18 g CH 4 /day around a mean 180 g CH 4 /day or a 10% difference (Hegarty et al., 2007). It remains to be seen if this is independent of productive traits, although in practice selection for reduced feed intake and CH 4 emissions will be conducted using an index that includes production traits.

Expectations of methods for measuring CH 4
The key requirements of a methodology for measurement of CH 4 production and MY of individual animals for genetic selection are, first, the methodology must provide a reliable measure of the true CH 4 emission by the individual for the period of measurement and suitable for the production system under target. This requires that the recovery of CH 4 emissions by the measurement procedure be consistent and preferably 100%. The RC, PACs, GEMs and SF 6 all potentially meet these criteria (Table 2). Methods where recovery is <100% might be useful if they show consistent recovery and capture diurnal variance in emissions rate. These include GEMs and sniffers which permit losses of CH 4 between animal and sensor.
Second, the period of measurement (of CH 4 and for MY, feed intake) and number of measurement periods should be sufficient to reliably rank sires for estimation of breeding values. In practice, this means multiple measures per animal. The optimal period and number of measurements will be determined by the pedigree structure of the data and the purpose of research. The repeatability of CH 4 measurements in PACs is only slightly less than in RCs (Table 2; Pinares-Patiño et al., 2013a). There is limited data to reliably estimate repeatability of CH 4 emissions using SF 6 and GEMs (Table 1), but it is anticipated that it would be less than in RCs. Having more progeny per sire will increase the accuracy of the estimate of sire EBVs and having more sires will improve the accuracy of the initial estimates of heritability. Finally, the measurement must be robust over time, as low cost as possible, not unduly influence animal behaviour and permit a high rate of data capture with low labour requirements. Ideally it should replicate the normal production system as far as possible.

Conclusions
There is potential for adopting genetic selection and in the future genomic selection, for reduced CH 4 emissions in ruminants. From this review it has been observed that direct measurement of CH 4 emissions from RC, SF 6 or PAC has proven underlying animal genetic variability. Subsequently, indirect indicators were explored through genetic correlations with CH 4 trait. It can be concluded that indirect and genomic selection might be possible options for near future selection. CH 4 emissions are a heritable and repeatable trait. CH 4 emissions are strongly related to feed intake both in the short term (minutes to several hours) and over the medium term (days). When measured over the medium term, MY is a heritable and repeatable trait albeit with less genetic variation than for total CH 4 emission (g/day). CH 4 emissions of individual animals are moderately repeatable across diets, and across feeding levels, when measured in RCs. Repeatability is less when short-term measurements are  Pickering, Oddy, Basarab, Cammack, Hayes, Hegarty, Lassen, McEwan, Miller, Pinares-Patiño and de Haas used, possibly due to variation in time and amount of ingested feed before the measurement. However, repeated measurements add value; it is preferable the measures be separated by at least 3 to 14 days. This needs to be investigated further. Given the above issue can be resolved, short-term (over minutes to hours) measurements of CH 4 emissions show promise. Finally, we believe that for shortterm measurements to be useful for genetic evaluation, a number (between 3 and 20) of measurements will be required over an extended period of time (weeks to months). There are opportunities for using short-term measurements in standardised feeding situations such as breath 'sniffers' attached to milking parlours or total mixed ration feeding bins, to measure CH 4 . We anticipate these are also subject to the caveats above about the use of short-term measurements. The measurement 'protocol' (i.e. how the animal and its feeding behaviour are managed before measurement) is more important than the technology used to make the CH 4 measurement. While there is evidence that correlated and predictor traits exist for CH 4 emissions the current level of knowledge is insufficient to recommend their use in genetic selection to reduce CH 4 emissions. Genomic selection has the potential to reduce CH 4 emissions and MY, however, measurements on thousands of individuals will be required. This includes the need to combined resources across countries in an international effort, emphasising the need for acknowledging the impact of the animal and production system on measurement of the CH 4 trait during design of experiments. The 'size of the prize' when combining lower MY with selection for low RFI may result in a reduction in CH 4 emissions of 40% to 45% and may be possible through selection of individual animals on components that directly affect CH 4 production.
In summary we consider genetic and genomic selection offers a significant opportunity to reduce CH 4 emissions from ruminants. However attention needs to be directed to a number of issues if short-term low-cost measurements are to be implemented in industry.