Inocula characterization and effects on rumen fermentation and animal performance
This study investigated the effects of early-in-life inoculation of newborns goat kids with rumen fluid from adult goats adapted to forage (RFF) or concentrate-rich diets (RFC), autoclaved rumen fluid (AUT) or absence of inoculation as control (CTL). RFF inocula had greater pH and acetate molar proportion, whereas RFC had greater concentrations of DM, total VFA and propionate and butyrate molar proportions (Supplementary Table 1). The microbial composition also showed substantial differences between these inocula: RFF inocula had higher bacterial and methanogens richness and higher abundances of certain microbial taxa such as Clostridiales, Rikenellaeae, Methanomassiliicoccaceae, Dasytricha and Caecomyces. On the contrary, RFC inocula had a higher load of total bacteria, protozoa and anaerobic fungi, as well as higher abundances of certain microbial taxa such as Lachnospiraceae, Veillonellaceae, Methanobrevibacter, Polyplastron and Piromyces. AUT inocula had a concentration of fermentation products in between RFF and RFC values, but without viable cells, as noted by the undetectable concentration of microbial DNA and intact protozoal cells after optical inspection.
Goat kids were daily inoculated from birth to 11 weeks of age and reared under artificial milk feeding. Rumen samples were taken at 5, 7 and 9 weeks of age to describe the rumen microbiome before, during and after weaning, respectively. Inoculation with fresh rumen (RFF and RFC) fluid increased solid feed intake (Fig. 1A) and the rumen VFA concentration during the pre-weaning period (Fig. 1B). This inoculation favored the transition to a solid diet as greater average daily gain (ADG) and butyrate molar proportion were observed during the post-weaning period (week 8, Fig. 1C) in comparison with CTL and AUT kids.
Effect of inoculation on the multi-kingdom rumen community
The multi-kingdom analysis included all microbial OTUs from bacterial (88%), methanogens (2.7%), protozoal (2.2%) and anaerobic fungal origin (7.1%) in a combined community representing the entire rumen microbiome. Permutational analysis of variance (PERMANOVA, Table 1) showed that this multi-kingdom community was highly affected (P < 0.01) by the microbial inoculation (explaining 21.8% of the total variance), the age of the animals (16.2) and their interaction (9.9%). Pair-wise analysis also showed differences between all four inoculation treatments and sampling times being illustrated in the Principal Coordinate Analysis (PCoA, Suppl. Fig. S1). This graphical representation showed that PCO1 (explaining 19.4% of the variance) discriminated between control (right), AUT (center) and RFF and RFC (left) whereas the PCO2 (17% of the variance) discriminated between pre-weaning (top) and post-weaning (bottom). Samples from RFF and RFC animals positively correlated with OTUs belonging to Ruminococcaceae, Christensenellaceae, Clostridiales, Anaerovorax, SP3-e08, Mogibacterium, Bacteroidales and Prevotella at 5 weeks, with Entodinium at 7 weeks and with Prevotellaceae, Selenomonas, Lachnospiraceae and Succinivibrio at 9 weeks of age, indicating a successional colonization process. For a more detailed description of the rumen microbiome, the main microbial groups were separately analyzed.
Table 1
Effects of early-in-life rumen microbial inoculation and time on the rumen community structure.
Community1
|
Inoculation
|
Time
|
Interaction
|
Multi-kingdom
|
|
|
Variance (%)
|
21.8
|
16.2
|
9.90
|
Pseudo-F
|
12.3
|
13.7
|
2.79
|
P-value
|
< 0.001
|
< 0.001
|
< 0.001
|
Bacteria
|
|
|
|
Variance (%)
|
16.1
|
17.3
|
9.26
|
Pseudo-F
|
8.20
|
13.21
|
2.36
|
P-value
|
< 0.001
|
< 0.001
|
< 0.001
|
Methanogens
|
|
|
Variance (%)
|
24.9
|
11.5
|
9.07
|
Pseudo-F
|
12.5
|
8.70
|
2.28
|
P-value
|
< 0.001
|
< 0.001
|
0.002
|
Protozoa
|
|
|
|
Variance (%)
|
14.5
|
9.07
|
6.28
|
Pseudo-F
|
6.79
|
4.23
|
1.47
|
P-value
|
< 0.001
|
< 0.001
|
0.096
|
Anaerobic fungi
|
|
|
Variance (%)
|
30.4
|
1.73
|
4.41
|
Pseudo-F
|
8.45
|
1.44
|
1.23
|
P-value
|
< 0.001
|
0.192
|
0.193
|
1PERMANOVA based on the Bray-Curtis dissimilarity. |
Effect of inoculation on the rumen bacterial community
RFC kids had lower bacterial DNA concentration per gram of DM than the other three treatments (Fig. 2A). The sequencing analysis generated on average 14,451 ± 716 high quality bacterial sequences per sample. RFF kids, followed by RFC and AUT, showed the highest bacterial diversity in terms of OTUs (Fig. 2B) and Shannon index (Fig. 2C), whereas CTL kids showed the lowest bacterial diversity indexes. Venn diagrams showed that the core bacterial community was composed by 15 OTUs (Fig. 2D). The inoculation with fresh rumen promoted a large core bacterial community in RFF (composed by 202, 231 and 164 OTUs) and RFC kids (139, 190 and 159 OTUs at weeks 5, 7 and 9, respectively). Moreover, many of these OTUs (20–30%) were exclusively shared between these two treatments indicating a similar community structure. AUT kids had a medium size core community (144, 153 and 124 OTUs), whereas CTL kids had much smaller core community (53, 76 and 53 OTUs at week 5, 7 and 9, respectively), being most of these OTUs (up to 64%) common across all treatments.
PERMANOVA analysis showed that inoculation and sampling time had significant impacts on bacterial community structure, explaining 16–17% of the total variance (Table 1). PCoA illustrated these differences in the bacterial community structure (Fig. 2E) in which PCO1 captured a gradient of community development according to the age of the kids (from left to right), whereas PCO2 did so for the CTL kids (from up to down). This PCoA also identified relevant microbes which partially explained these differences. For example, samples taken at 9 weeks of age from AUT, RFF and RFC kids correlated with the presence of Prevotella OTUs, whereas those from CTL kids did so with Succinivibrio OTUs. Inocula samples clustered together with samples from RFF and RFC kids at 7 weeks of age, indicating a similar bacterial community and as expected, the effect of the inoculation differed with the age of the kids (interaction, P < 0.001. Table 1), indicating that specific analyses for each sampling time were needed.
During the pre-weaning period (week 5, Fig. 2F), inoculation with fresh rumen fluid showed a unique and different bacterial community as compared to AUT or CTL kids. This bacterial community correlated with indicators of the rumen microbial and functional development such as dry matter intake (DMI), forage intake, protozoal concentration and bacterial and protozoal richness. Most differences among treatments were also extended to the age at weaning (week 7, Fig. 2G). Again, clustered samples from these latter groups were related with higher bacterial and protozoal richness, as well as with the average daily gain (ADG-f) and feed efficiency (FE-f) during the following week, suggesting that this community structure minimized the weaning shock. During the post-weaning period (week 9, Fig. 2H) CTL kids retained a different bacterial community than that of other treatments, whereas the bacterial community of AUT kids became closer to those kids inoculated with fresh rumen fluid. Greater bacterial and protozoal richness were again associated with the bacterial community structure of inoculated groups, along with digestible cellulose (DCI) and hemicellulose (DHCI) and forage intake during the following week (Forage –f).
The analysis of the relative abundances of the most predominant bacterial families and genera (Fig. 3A and Suppl. Table S2) indicated that 12 out of the 21 families identified showed significant differences based on the inoculation treatment, regardless of sampling time. Inoculation with fresh rumen fluid promoted the presence of certain bacterial taxa at week 5 (Elusimicrobia, Anaerobiospirillum, Catenisphaera, Clostridiaceae, Marvinbryantia, Saccharofermentans, Quinella, Selenomonas, Rhodocyclaceae, Succinomonas, Synergistes, Olicosphaeraceae, PeH15 and SP3-e08) which were not present in CTL kids, the first three being also absent in AUT kids. Most of these taxa were not detected at later sampling times in CTL kids indicating a microbial colonization delay. Inoculation with fresh rumen fluid also increased the abundance of various phyla (Firmicutes, Fibrobacteres, Tenericutes, Cyanobacteria and Elusimicrobia) and genera (Fibrobacter, Succiniclasticum, Eubacterium, Lachnoclostridium, Moryella, Oribacterium or Anaerotruncus) in comparison with CTL and AUT kids across sampling times. On the contrary, Bacteroidales, Alloprevotella and Coprococcus were most abundant in CTL and AUT groups across sampling times. Moreover, 17 out of the 21 taxa presented significant differences according to sampling time: Prevotellaceae and Succinivibrionaceae increased over time whereas Ruminococcaceae decreased. The interaction between inoculation and sampling time was significant in 10 out of 21 bacterial taxa indicating that the effects were more obvious before than after weaning. For example, CTL kids had a higher abundance of Bacteroides, Lachnospira, Streptococcus, Pasterurellaceae and Pyramidobacter than the rest of treatments at 5 weeks of age but differences disappeared after weaning.
Spearman correlations were performed to assess the potential implications of changes in rumen meta-taxonomics data on animal physiology (Table 2 and Suppl. Table S3). Bacterial richness, a number of bacterial phyla such as Fibrobacteres, Firmicutes, Tenericutes, Elusimicrobia, Cyanobacteria, Chloroflexi and Lentisphaerae, and several bacterial families such as Ruminococcaceae, Veillonellaceae, Rhodocyclaceae, Rikenellaceae, Defluviitaleaceae and Family_XIII positively correlated with various indicators of the rumen physiological development such as forage and solids intake, acetate molar proportion, presence of protozoa and higher bacterial, protozoal and methanogens diversity. Moreover, the abundance of Firmicutes and Veillonellaceae positively correlated with the ADG during the post-weaning period indicating a better transition from liquid to solid feed during the post-weaning period. On the contrary, the phylum Bacteroidetes, the families Bacteroidaceae, Comamonadaceae and Neisseriaceae and the order Bacteroidales showed a negative correlation with these indicators of the rumen physiological development as well as with animal performance during the post-weaning period.
Table 2. Correlations among the rumen microbiota and digestive physiology data.
1Microbial data were log10 transformed and only Spearman’s correlations coefficients ρ > 0.3 (green) or ρ < −0.3 (red) and p < 0.01 are shown (N=96). Parameters: milk, concentrate, forage and DM intake (g/d), rumen pH,total volatile fatty acids (mM), acetate (%), propionate, butyrate (%), odd and branched chain fatty acids (%), microbial concentration (log10 copies/mg DM), OTUs richness (-R), plasma β-hydroxybutyrate (mM), blood glucose (mg/dL), average daily gain (g/d) and ADG during the post weaning period (ADG-pw).
Effect of inoculation on the rumen archaeal community
The interaction between inoculation treatment and time on the concentration of methanogenic archaea in the rumen was significant (Fig. 4A). Although RFF kids, followed by AUT and RFC, showed the highest concentrations of methanogens at 5 weeks of age (and CTL the lowest) these differences tended to decrease as kids aged. Sequencing analysis generated an average of 14,180 ± 1,100 high-quality methanogens sequences per sample and showed that methanogens diversity increased with the age of the kids (Fig. 4B). Moreover, CTL kids showed lower methanogens diversity in terms of OTUs and Shannon index (Fig. 4C) than observed in other treatments across time points. The methanogens core community was composed by only two OTUs (both of them belonging to the genus Methanobrevibacter) which were shared across all treatments and time points (Fig. 4D). However, at week 9 new Methanobrevibacter and Methanosphaera OTUs appeared in this core community. Venn diagrams for individual time points revealed that the methanogens core community remained similar for CTL kids (5, 3 and 5 OTUs at week 5, 7 and 9, respectively) but increased for AUT (6, 12 and 17 OTUs), RFF (7, 9 and 15 OTUs) and RFC kids (7, 11 and 13 OTUs).
Rumen methanogens community structure was greatly affected by inoculation, sampling time and their interaction (Table 1, Fig. 4E), however, PERMANOVA indicated that the proportion of the variance explained by the inoculation treatment was twice more than that explained by sampling time (24.92 vs 11.54%). PCoA showed that PCO1 separated samples from CTL (right) and from the rest of treatments (left), whereas PCO2 disaggregated samples between pre-weaning (bottom) and post-weaning (top). Moreover, samples collected from RFF and RFC at 5 weeks positively correlated with OTUs belonging to Methanimicrococcus, Methanophanus and Groups 8, 9 and 10.
The study of the methanogens community structure at 5, 7 and 9 weeks of age using Distance-Based Redundancy Analyses (Fig. 4F) showed a general pattern characterized by a separation through axis 1 between samples from CTL kids (right) and those from RFF and RFC kids (left). At week 5, RFF and RFC samples positively correlated with the presence of higher diversity levels for bacteria, methanogen and protozoa and higher DMI and blood beta-hydroxybutyrate indicating a more microbiological and functional rumen development. At weeks 7 (Fig. 4F) and 9 (Fig. 4H), RFF and RFC samples correlated with higher diversity indexes protozoal concentration, forage intake and ADG and FE during the following week after weaning whereas CTL samples clustered on opposite direction indicating a more undeveloped methanogens community.
The analysis of the relative abundances of the 15 most predominant methanogen species (Fig. 3B and Suppl. Table S4) showed differences according to the inoculation treatment (13 species) and sampling time (11 species), however a significant interaction was found for most of them. At 5 weeks of age, nearly the entire methanogens community (99.7%) in CTL kids was formed by Methanobrevibacter gottschalkii and Group8_sp, however these two species only represented 39.0, 22.9 and 29.1% in AUT, RFF and RFC. On the contrary, RFF and RFF had increased abundances of Group9_sp (34.4%), Methanomicrobium mobile (9.7%), Methanobrevibacter ruminantium (3.8%) and Methanobassiliicoccaceae spp (18.3%), whereas AUT kids were more abundant in Group10_sp (14.7%). At week 7, a consistent presence of Methanimicrococcus blatticola, Methanomicrobium mobile, Methanosphaera, Group12_sp and Group10_sp was detected in AUT, RFF and RFC but were absent in CTL kids which still retained higher numbers of M. gottschalkii (39.6%). This over-representation was even bigger at 9 weeks of age (59.9%), whereas inoculated kids were more abundant in a greater number of methanogen species (e.g. Group8_sp and Group_9sp).
The abundance of M. gottschalkii (and Group8_sp) was negatively correlated with forage intake, presence of protozoa and ADG before and after weaning, indicating the presence of an immature methanogens community (Table 2). On the contrary, M. blatticola, M. mobile, M. ruminantium, Methanosphaera and Group9_sp were positively correlated with indicators of a greater rumen microbiological (protozoal, methanogens and bacterial concentration and diversity) and physiological development (forage intake and ADG before and after weaning).
Effect of inoculation on the rumen protozoal community
Control kids remained protozoa-free over the entire duration of this study. Inoculation with fresh rumen fluid promoted a higher concentration of rumen protozoa (Fig. 5A) at week 5 but these differences tended to be smaller as time progressed (interaction P < 0.05). The 18S amplicon sequencing yielded an average of 18,164 ± 644 sequences per sample and diversity analysis showed that RFF and RFC had higher protozoal diversity in terms of OTUs (Fig. 5B) and Shannon index (Fig. 5C) than AUT kids across sampling times. A total of 14 protozoal OTUs formed the core community shared between AUT, RFF and RFC kids across time points (Fig. 5D). The treatment-specific core community increased over time for RFC (20, 23 and 25 OTUs at week 5, 7 and 9, respectively), and for RFF kids (25, 27 and 19 OTUs) since most OTUs were shared across these two treatments. On the contrary, the protozoal core community was smaller and remained constant over time for AUT kids.
PERMANOVA revealed that the inoculation and the sampling time greatly modified the protozoal community structure explaining 14.5% and 9.07% of the total variance, respectively (Table 1). Pair-wise comparisons and PCoA analysis (Fig. 5E) showed that inoculation with fresh rumen fluid promoted a protozoal community similar to the observed in the inocula and positively correlated with the presence of 8 different protozoal OTUs, whereas the community in the AUT kids only correlated with Entodinium OTUs. The analysis of the protozoal community at different time points showed that RFF and RFC always shared a similar protozoal community which was positively correlated with indicators of a rumen microbiological (higher bacterial and protozoal richness) and functional development (higher DM, forage and concentrate intakes, and higher VFA, butyrate and ADG). This protozoal community differed to that observed in AUT kids at 5 (Fig. 5F) and 7 weeks (Fig. 5G) but not at 9 weeks of age (Fig. 5H), indicating a delay in the rumen protozoal colonization in AUT kids.
Analysis of the protozoa relative abundances (Fig. 6A and Suppl. Table S5) showed a progressive decrease over time in the entodiniomorphids (family Ophryoscolecidae) and an increase in holotrichs protozoa (family Buetschliidae). AUT kids had increased numbers of Entodinium, whereas RFF and RFC were more abundant on Diplodinium, Enoploplastron, Isotricha and Dasytricha, these differences being greater before than after weaning. The abundance of most protozoal species was positively correlated with DM intake and bacterial, methanogens and protozoal diversities. Abundances of Ophryoscolex, Isotricha and Dasytricha were also correlated with higher VFA concentration and ADG during the post-weaning period as indicators of rumen development.
Effect of inoculation and age on the rumen fungal community
Inoculation with fresh rumen fluid increased the anaerobic fungal concentration at week 5 in comparison to CTL and AUT kids (Fig. 7A), some of these differences persisted at weaning but disappeared after weaning (interaction, P < 0.001). Fungal sequencing yielded an average of 6,548 ± 618 high quality sequences per sample for those taken at 5 and 7 weeks of age. However, the number of reads observed at 9 weeks was unexpectedly low and this time point was not further considered. No differences were found in the anaerobic fungal richness or alpha-diversity levels across treatments (Fig. 7B and 7C).
An absence of a core anaerobic fungal community was observed since no OTUs were shared across treatments and time points (Fig. 7D). Despite that, a treatment-specific core community was observed in CTL (6 and 5 OTUs at week 5 and 7, respectively), RFF (2 and 2 OTUs) and RFC (11 and 3 OTUs) but not in AUT kids. PERMANOVA revealed that the fungal community structure was highly affected by the inoculation (P < 0.001) explaining 29.8% of the total variation, whereas no effect was observed for the age of the kids (Table 1). PCoA analysis (Fig. 7E) showed a clear separation between kids inoculated with fresh rumen fluid (right) and those from CTL and AUT (left). Moreover, a different fungal community was observed between RFF (top) and RFC samples (bottom) being these latter samples correlated with the presence of several Piromyces OTUs.
Inoculation with rumen fluid greatly modified the fungal colonization process in terms of taxa abundance (Fig. 6B and Suppl. Table S6). CTL kids had an anaerobic fungal community composed by only 3 taxa: Neocallimastigaceae spp (76%), Neocallimastix (10%) and Caecomyces (14%). AUT kids showed lower numbers of Neocallimastigaceae spp and Neocallimastix than CTL kids, but higher numbers of Caecomyces, Piromyces, Orpinomyces and Capnodiales. Moreover, RFF showed the highest abundance of Orpinomyces, whereas RFC did so for Piromyces. Correlation analysis showed that Caecomyces negatively correlated with the presence of protozoa (Table 2), whereas Orpinomyces positively correlated with indicators of the rumen physiological (concentrate and DM intakes) and microbiological development (higher bacterial and protozoal richness). Piromyces was correlated with even more indicators of the rumen microbiological development (including methanogens and protozoal concentrations and methanogens richness).