High-throughput DNA sequencing of the moose rumen from different geographical locations reveals a core ruminal methanogenic archaeal diversity and a differential ciliate protozoal diversity

Moose rumen samples from Vermont, Alaska and Norway were investigated for methanogenic archaeal and protozoal density using real-time PCR, and diversity using high-throughput sequencing of the 16S and 18S rRNA genes. Vermont moose showed the highest protozoal and methanogen densities. Alaskan samples had the highest percentages of Methanobrevibacter smithii, followed by the Norwegian samples. One Norwegian sample contained 43 % Methanobrevibacter thaueri, whilst all other samples contained < 10 %. Vermont samples had large percentages of Methanobrevibacter ruminantium, as did two Norwegian samples. Methanosphaera stadtmanae represented one-third of sequences in three samples. Samples were heterogeneous based on gender, geographical location and weight class using analysis of molecular variance (AMOVA). Two Alaskan moose contained >70 % Polyplastron multivesiculatum and one contained >75 % Entodinium spp. Protozoa from Norwegian moose belonged predominantly (>50 %) to the genus Entodinium, especially Entodinium caudatum. Norwegian moose contained a large proportion of sequences (25–97 %) which could not be classified beyond family. Protozoa from Vermont samples were predominantly Eudiplodinium rostratum (>75 %), with up to 7 % Diploplastron affine. Four of the eight Vermont samples also contained 5–12 % Entodinium spp. Samples were heterogeneous based on AMOVA, principal coordinate analysis and UniFrac. This study gives the first insight into the methanogenic archaeal diversity in the moose rumen. The high percentage of rumen archaeal species associated with high starch diets found in Alaskan moose corresponds well to previous data suggesting that they feed on plants high in starch. Similarly, the higher percentage of species related to forage diets in Vermont moose also relates well to their higher intake of fibre.


Introduction
Previous investigations into the micro-organisms in the rumen of the moose have focused on bacteria using cultivation (Dehority, 1986) and high-throughput sequencing techniques (Ishaq & Wright, 2012, 2014a or on protozoa using light microscopy (Dehority, 1974;Krascheninnikow, 1955;Sládeček, 1946;Westerling, 1969) and high-throughput sequencing (Ishaq & Wright, 2014b). Methanogenic archaea in the rumen of moose have not previously been identified nor have methanogens or protozoa from moose been compared across samples from different geographical locations. Methanogens and protozoa in the rumen are often found in intracellular or extracellular symbiotic associations involving hydrogen transfer from protozoa to methanogens. Previously, protozoa from the genera Dasytricha, Entodinium, Polyplastron, Epidinium and Ophryoscolex have been shown to interact with methanogens from the orders Methanobacteriales and Methanomicrobiales (Finlay et al., 1994;Newbold et al., 1995;Sharp et al., 1998;Stumm et al., 1982;Vogels et al., 1980).
For domestic livestock, methanogenesis represents a loss of dietary efficiency as compounds such as acetate or hydrogen are sequestered by methanogens instead of being used by the host for production (i.e. live weight gain, milk production, wool production, etc.). Much research has been performed on methanogenesis and rumen microbial populations between domestic and wild ruminants, as wild ruminants (bison, elk and deer) are estimated to produce up to 0.37 Tg CO 2 Eq year 21 (Hristov, 2012;McAllister et al., 1996). This is a drastically lower figure than that for domestic livestock, at 141 Tg CO 2 Eq year 21 (EPA, 2014). Wild ruminants are presumed to produce less methane based on a presumed higher dietary efficiency and lower production demands. As a first step to better understanding methanogenesis in moose, the present study identified the methanogens present in the rumen of moose, as well as the protozoa that are potentially symbiotically associated with them.
The objectives of this research were to identify the methanogens and protozoa present in the rumen of moose from Alaska, Vermont and Norway; to measure the density of methanogens and protozoa in these samples; to compare samples across geographical location, gender and weight class to determine possible trends; and to compare samples with published studies on wild and domesticated ruminants. It was hypothesized that moose may have fewer total methanogens than domestic ruminants due to a fast rate of passage through the gastrointestinal tract (Lechner et al., 2010). In previous studies, age (Godoy-Vitorino et al., 2010;Ishaq & Wright, 2014a;Li et al., 2012) and geographical location (Ishaq & Wright, 2014a;Sundset et al., 2007) have played a role in differentiating core bacterial microbiomes of various hosts, and it was also hypothesized that this would hold true for methanogens in the moose rumen. However, reindeer, which often share a similar diet or geographical location to moose, have been shown to have similar protozoal diversity across geographical locations, indicating that the host species may not have been isolated long enough to develop a unique profile regardless of geographical location of the host (Imai et al., 2004). As moose have not been isolated long, it was hypothesized that this would hold true for moose as well.

Impact Statement
For the first time, to the best of our knowledge, the methanogenic archaea in the moose rumen have been identified. Additionally, both methanogens and protozoa diversity have been compared from the rumens of moose across three geographical locations using highthroughput techniques. Rumen ciliate protozoa and methanogenic archaea often form symbiotic relationships for the transport of hydrogen from protozoa to methanogens. Understanding their impact to the moose rumen microbiota is critical to understanding the overall rumen function of wild moose. This information can be applied to further studies on methane production in rumens and potential mitigation strategies, such as altering the methanogen or protozoal diversity to reduce methane output. Surprisingly, there was low methanogen diversity and high protozoal diversity across moose in different locations where they have consumed different diets. Additionally, a large percentage of protozoal sequences could not be identified beyond genus or family, indicating a need for additional work into identifying novel species.
Sequence analysis. All sequences were analysed using mothur version 1.31 (Schloss et al., 2009) and are available under the NCBI Sequence Read Archive under Bioproject IDs PRJNA281249 for methanogens and PRJNA281109 for protozoa. For methanogens, sequence analysis was as described previously (Ishaq & Wright, 2014a), with the following modifications. Sequences were trimmed to a uniform length of 436 alignment characters (minimum 350 bases) and candidate sequences were aligned against the Ribosomal Database Project reference alignment integrated into mothur with the bacterial sequences removed. Sequences were classified using the k-nearestneighbour method against the full Ribosomal Database Project alignment, which had been modified to include species-level taxonomy. A 2 % genetic distance cut-off was used to designate species. For protozoa, sequence analysis was as previously described for primer set 1 (PSSU316F and GIC758R), using a 4 % genetic distance cut-off to designate species (Ishaq & Wright, 2014b). Sequences were subsampled evenly for each sample. The operational taxonomic unit (OTU) estimators CHAO (Chao & Shen, 2003) and ACE (http://chao.stat.nthu.edu.tw), Good's Coverage (Good, 1953), and the Shannon-Weaver Diversity Index (Shannon & Weaver, 1949) were calculated. An analysis of molecular variance (AMOVA) and UniFrac (Hamady et al., 2010) were used to compare the heterogeneity of samples. UniFrac measures overall phylogenetic tree distance between samples and will create a dendrogram which clusters samples. Principal coordinate analysis (PCoA) calculates the distance matrix for each pair of samples and then turns these distances into points in a space with a number of dimensions one less than the number of samples.
For protozoa, the primers PSSU316F and PSSU539R (59-ACTTGCCCTCYAATCGTWCT-39) (Sylvester et al., 2004) targeted the 18S rRNA gene, following the protocol by Sylvester et al. (2004), and internal standards for protozoa were created in the laboratory using fresh dairy cattle rumen contents which were filtered through one layer of cheesecloth to remove large particles, and then the protozoa were allowed to separate for 2 h at 39 uC. Once a protozoal pellet was visible, 50 ml was drawn from the bottom of the funnel and 1 vol. 100 % ethanol was added to fix the cells and DNA. The mix was centrifuged for 5 min at 2000 g, and the pellet was washed with TE buffer (1 M Tris/HCl, 0.5 M EDTA, pH 8.0) and then centrifuged again. Cells were counted microscopically using a Thoma Slide following the protocol by Dehority (1974) (R 2 50.998). Both protocols were followed by a melt curve, with a temperature increase of 0.5 uC every 10 s from 65 uC up to 95 uC to check for contamination.

Methanogens
A total of 141 368 sequences, of which 47 370 were unique, passed quality assurance steps. For each sample, between 22 and 330 OTUs were assigned using a 2 % genetic distance cut-off (Wright et al., 2009) Vermont samples contained the highest mean density of methanogens at 1.3e+10 cells ml -1 , followed by Alaskan samples and Norwegian samples (5.19e+09 and 3.58e+09 cells ml -1 , respectively) ( Table 3). Whilst there was a positive correlation between individual methanogen and protozoal density in moose, it was not significant (R 2 50.38) (data not shown). Two of three Alaskan moose, as well as two of six Norwegian moose had a larger proportion of methanogens belonging to the SGMT clade (Methanobrevibacter smithii, Methanobrevibacter gottschalkii, Methanobrevibacter millerae, and Methanobrevibacter thauri). All eight Vermont moose, one Alaskan moose and four Norwegian moose had greater proportions of members of the RO clade (Methanobrevibacter ruminantium and Methanobrevibacter olleyae) (Fig. 2). There was also no trend seen between moose age and methanogen density (R 2 50.015, data not shown).
Alaskan samples had the highest percentages of Methanobrevibacter smithii (16-36 %), followed by the Norwegian samples (10-24 %) (Fig. 2) This was also confirmed using PCoA for gender, location, and weight class (Fig. 1B, D, F). No clustering bias was seen with respect to storage technique of samples.
Vermont samples contained the highest mean density of protozoa at 4.70e+06 cells ml -1 , followed by Alaskan samples and Norwegian samples (3.83e+06 and 5.17e+04 cells ml -1 , respectively) ( Table 3). Whilst there was a positive correlation between individual methanogen and protozoal density in moose, it was not significant (R 2 50.38) (Data not shown). There was also no trend seen between moose age and protozoal density (R 2 50.107, data not shown).
Protozoa were identified using a previously described reference alignment and taxonomy of valid protozoal sequences (Ishaq & Wright, 2014b) (Fig. 3). Two Alaskan moose contained w70 % Polyplastron multivesiculatum and one contained w75 % Entodinium spp. Protozoa from Norwegian moose belonged predominantly (w50 % of total sequences) to the genus Entodinium, especially Entodinium caudatum (Fig. 3). A large proportion of sequences in Norwegian moose (25-97 % of total sequences) could not be classified beyond the family Ophryoscolecidae (Fig. 3). Protozoa from Vermont samples were predominantly composed of Eudiplodinium rostratum (w75 % of total sequences). Vermont samples also contained up to 7 % Diploplastron affine (Fig. 3). Many other species were identified in moose, with v1 % each of the following identified: Anoplodinium denticula-

Geographical location
The present study represents the first insight into the methanogenic archaeal diversity in the rumen of the moose. Although distinct in terms of proportion of methanogenic taxa present in each of the three moose populations, the samples were not statistically different between geographical populations. This suggests that moose have a core methanogen microbiome, as has been suggested for protozoa in other host species (Imai et al., 2004). It is possible that, whilst diet is a significant factor in determining the micro-organisms present in the rumen, there are other factors, such as body/rumen temperature or rumen pH, which are selecting for similar methanogen species in moose from different geographical locations on different diets. There was also no trend seen between moose age and methanogen density in the present study, despite clear trends between age and density in previous studies (Saengkerdsub et al., 2007;Skillman et al., 2004). This may be due to a relatively small sample size or trends may be indistinct once the moose rumen reaches developmental maturity before its first year.
Given the markedly different protozoal populations found in Alaska, Vermont and Norway, as well as the AMOVA analysis confirming statistically different groups, it may be concluded that moose do not have a typical protozoal diversity as do reindeer from various locations (Imai et al., 2004). It is estimated that moose reached North America across the Bering Strait from Asia some 14 000-11 000 years ago, but it was not until relatively recently that moose dispersed to peripheral (i.e. coastal) regions and began to diversify genetically (Hundertmark et al., 2003). Despite the relatively recent diversification of moose subspecies in North America, moose have been geographically isolated long enough to form distinct rumen protozoal populations.
As with moose rumen bacteria in a previous study (Ishaq & Wright, 2014a), Alaskan moose shared a large number of methanogenic sequences, followed by the Norwegian samples, and females shared more archaeal and protozoal sequences than males. Unlike previously (Ishaq &   Wright, 2014a), the 202-300 kg weight class shared the greatest number of archaeal and protozoal sequences of all the weight classes. Whilst the total methanogen OTUs were higher than usually reported, our numbers were not outside the range of those previously reported for other ruminants (7168 OTUs, with a range of 788-2758 OTUs; Piao Hailan et al., 2014). The present study retained singletons and doubletons to prevent the removal of rare taxa.

Diet
Methanosphaera stadtmanae has previously been associated with diets including fruit, as they require methanol (a byproduct of pectin fermentation), and has been previously seen in omnivores (Dridi et al., 2009;Facey et al., 2012) and the rumen of various hosts (Cersosimo et al., 2015;Cunha et al., 2011;Snelling et al., 2014). Nordic blueberries have been found in the diet of Norwegian moose (Shipley et al., 1998;Wam & Hjeljord, 2010) and often bear fruit year-round. Nordic blueberries contain an average of 0.7 g fat and it is possible that an increase in dietary fat from berries reduced methanogen density (Dohme et al., 2001) in Norwegian moose in the present study.
Methanobrevibacter smithii, unlike many other methanogens, has been shown to grow at less than neutral pH ( Rea et al., 2007), is often associated with high-calorie diets in ruminants (Carberry et al., 2014;Zhou et al., 2010), has been associated with high-efficiency animals (Zhou et al., 2010), has been shown to influence weight gain in rats (Mathur et al., 2013) and has been shown to improve polysaccharide fermentation by bacteria (Joblin et al., 1990;Samuel & Gordon, 2006). Conversely, Methanobrevibacter ruminantium has been associated with a highforage diet in ruminants (Zhou et al., 2010). Previously, Alaskan moose were speculated to be on a high starch/ energy diet and showed a much higher proportion of Bacteroidetes, especially Prevotella spp. (Ishaq & Wright, 2014a), which are associated with protein and starch digestion in the gastrointestinal tract. Methanobrevibacter smithii improves polysaccharide digestion by bacteria (Joblin et al., 1990;Samuel & Gordon, 2006).
Previously, Vermont moose were presumed to be on a high forage/low energy diet (Ishaq & Wright, 2012, 2014a, which may account for the relatively high proportions of Methanobrevibacter ruminantium in the present study. Whilst Methanobrevibacter ruminantium does not use acetate for methanogenesis, it can use formate, which is created during acetogenesis. An increase in plant cell wall digestion increases the amount of acetate produced in the rumen, which can increase methanogenesis by providing methyl groups (Johnson & Johnson, 1995). Roughage diets in livestock have been shown to increase methane emissions (Liu et al., 2012), even when the roughage diets are not associated with altered methanogen densities (Liu et al., 2012;Zhou et al., 2010).
Previously, domestic steers fed a roughage diet had a mean density of 1.34e+09 cells ml -1 for methanogens, which were predominantly Methanobrevibacter spp. (Denman et al., 2007), and which was comparable to the present study. Holstein dairy cattle on a high-forage diet had a mean density of 6.04e+05 cells ml -1 for protozoa (Sylvester et al., 2004), which is similar to densities in moose. It was also shown that densities decreased on a lowforage diet and the dominant genus was Entodinium spp. (Sylvester et al., 2004).

Microbial interactions
Methanobrevibacter ruminantium, found in Vermont moose, has previously been associated with higher densities of Polyplastron, Eudiplodinium and Entodinium (Ohene-Adjei et al., 2007;Sharp et al., 1998;Vogels et al., 1980). Vermont samples were dominated by Polyplastron multivesiculatum and Eudiplodinium maggii, whilst Norwegian samples were dominated by Entodinium spp. Previously, using light microscopy, moose were shown to have primarily Entodinium spp., including Entodinium dubardi and Entodinium longinucleatum in Alaska (Dehority, 1974), Entodinium dubardi and other Entodinium spp. from Slovakia, and Entodinium dubardi and Epidinium caudatum in Finish Lapland (Westerling, 1969). More recently, using high-throughput sequencing, moose in Alaska were shown to have a high percentage of Polyplastron multivesiculatum, and well as a variety of Entodinium and other species (Ishaq & Wright, 2014b), which was also shown in the present study. The Norwegian samples had a high percentage of Entodinium caudatum, Entodinium furca dilobum, and other Entodinium species, giving them a similar profile to moose samples from Alaska (Dehority, 1974), Finland (Westerling, 1969) and Slovakia (Sládeček, 1946) using light microscopy. The Norwegian samples also contained a large proportion of sequences which could not be identified beyond the family level, indicating that these moose host novel ciliate species or that no 18S rRNA sequences exist for previously identified species.
It has been shown that protozoal density affects methanogen density (Morgavi et al., 2012;Newbold et al., 1995), as the two microbial communities are often symbiotically associated with one another. In the present study, there was a trend towards positively correlated methanogen and protozoal density in individual moose, but it was not significant (R 2 50.38). Polyplastron, Eudiplodinium maggii and Entodinium caudatum have been shown to have w40 % association with methanogens (Vogels et al., 1980). More specifically, Polyplastron was recently shown to associate with Methanosphaera stadtmanae and Methanobrevibacter ruminantium (Ohene-Adjei et al., 2007).
Methanogen and protozoal densities in reindeer from Norway (Sundset et al., 2009b) averaged very closely to densities found in Norwegian moose, which were lower than in Alaskan and Vermont moose. In addition to the possibility that dietary fat was reducing methanogens, another possible reason for the low methanogen density in Norwegian moose is the presence of bacterial competitors (Wright & Klieve, 2011), such as acetate-utilizing, hydrogen-utilizing or sulphate-reducing bacteria, which sequester free CO 2 and H 2 in the rumen. Very few sequences of Acetitomaculum or Eubacterium (acetateutilizing) were identified, but they were found in Norwegian samples previously (Ishaq & Wright, 2014a). Sulphate-reducing bacteria, many of which belong to the Clostridium class of the phylum Firmicutes, were previously found in Norwegian reindeer (Sundset et al., 2007), but were not found in abundance in Norwegian moose. Likewise, only a few Desulfovibrio spp. were found in Norwegian moose, although the phylum Proteobacteria was in largest abundance in Norwegian moose, so perhaps more sulphate-reducing bacteria exist which would not be classified down to family (Ishaq & Wright, 2014a).
However, acetate-producing (acetogens) or hydrogenproducing bacteria could have potentially contributed to a higher methanogen density in Alaskan or Vermont samples. Acetogens, such as the phylum Actinobacteria, were found in low quantities across the board, with the exception of NOM4R (6 %) (Ishaq & Wright, 2014a). However, specific acetogens species (members of Sporomusa, Moorella, Clostridium, Acetobacterium and Thermoanaerobacter) were not identified, and their respective families were not found in large abundance in any sample. Of several known hydrogen-producing genera, Selenomonas and Streptococcus were found in low numbers, and Bacteroides and Succinomonas were not found. However, the order Bacteroidales was previously found in abundance in Alaskan samples (36-83 % of sequences), Norwegian samples (0.7-54 %) and Vermont samples (7-30 %) (Ishaq & Wright, 2014a).
If Alaskan moose were indeed consuming a diet relatively high in starch, the resulting reduction in pH would not be detrimental to Methanobrevibacter smithii density, which would in turn support bacterial polysaccharide fermentation (Joblin et al., 1990;Samuel & Gordon, 2006). Concentrate diets in livestock have been shown to reduce the total number of methanogens by decreasing the pH and increasing the food passage rate; however, even a forage diet higher in starch would not contain enough readily fermentable starches to produce the same effect in these moose.