Compositional changes
We next explored if the increased gas production could be explained by
compositional differences resulting from the addition of the pre-adapted
inocula. At the end of the experiment, community composition had changed
in parallel across both the control and adaptation treatments (and the
pre-adaptation source communities for the adaptation treatment),
indicative of strong selection associated with the transition to a novel
environment. There was a 60% drop in the mean number of taxa detected
per community, compared to the ancestral community, but we did not find
a significant difference in diversity loss across treatments (one-way
ANOVA on ASV reduction from the ancestral sample;
F(2,32)=2.93, P value = 0.07; Figure 3A). In
particular, there were reductions in the abundance of taxa belonging to
the Firmicutes, Proteobacteria, Spirochaetes and Tenericutes.
Despite these parallel changes, treatments differed in terms of their
community composition (Figure 3B, PERMANOVA on Bray-Curtis
dissimilarities: F(2,32) = 15.8, R2 =
0.49, Unweighted UniFrac: F(2,32) = 5.6,
R2 = 0.26, Weighted UniFrac: F(2,32) =
26.9, R2 = 0.63, p-value = 0.001 and, 999 iterations
per test in all cases), with the frequency of methanogens, the organisms
directly responsible for methane production, remaining constant across
treatments. Most notably, control samples had a higher fraction of
Bacteriodia and Gamma-proteobacteria compared to both the Pre-Adaptation
and Adaptation samples (Figure 4). We note however that the magnitude of
compositional differences between control and adapted communities are
similar to that observed between pre-adapted versus control communities,
which did not differ in gas production.
The compositional changes and associated increases in gas production
resulting from the addition of the pre-adapted inocula could have arisen
through selection acting on novel ecological composition (ecology),
genetic variants (evolution) or both. Given that directly attributing
causal changes in composition and function to evolution in such complex
communities is not possible, we instead sought to determine if
ecological processes were sufficient to explain these changes. First,
our experimental design - adding only 1% of the pre-adapted communities
- should have had only a minimal direct impact on the ecology of the
recipient communities. To confirm this, we simulated the addition of 1%
of the pre-adapted communities to control communities and unsurprisingly
found that this caused non-significant changes in community composition
(adonis2, F1,23 = 0.017,
R2<0.001 P = 1). Furthermore, adapted
communities were no more similar to directly linked pre-adapted
communities, compared to unlinked (randomly selected pre-adapted
community) or control communities (Fig. 2C).
Previous work has shown that rare taxa can play an important role in
methane production in methanogenic and other communities . As such,
while adding only 1% of the pre-adapted community had no detectable
direct impact on community composition, it could have had a direct
effect on gas production if functionally important rare taxa increased
in frequency by orders of magnitude during pre-adaptation. To
investigate this possibility, we first identified taxa that had
significantly greater abundance in adapted versus control communities,
as their presence could explain between-treatment differences in gas
production. Next, we determined whether any of those taxa increased
between the start (i.e., the ancestral community) and end of the
pre-adaptation treatment to an extent where they could have enriched the
adaptation treatment. Using this two-pronged approach, we found one
candidate taxon, belonging to the Ruminococcus genus (Table 1) – this
taxon had a higher frequency in the adaptation versus control treatment
and was enriched over the course of the pre-adaptation treatment. It is
however highly unlikely that this organism could explain the difference
in gas production between control and adaptation treatments: it was only
detected in a subset of replicates in the adaptation treatment (10 out
of 12), so its presence cannot contribute to all observed increases in
gas production. The taxon is also present in the pre-adaptation
treatment (in 6 replicates) and we did not find a significant difference
in its abundance when comparing pre-adapted versus adapted communities
(t(21) = -0.21; P = 0.84, means of 92.8 for
pre-adaptation and 104.5 for adaptation treatment, 95% CI -138.1 to
114.1). Despite that, there is a significant difference in gas
production between those two treatments.