Page 15 - Read Online

Mueller et al. Microbiome Res Rep 2024;3:33 https://dx.doi.org/10.20517/mrr.2024.09 Page 9 of 18

existing metagenomic sequencing tools, enabling us to achieve species-level resolution of the Akkermansia
in these published datasets.

To determine if existing strain-finding programs can distinguish between Akkermansia species, we used the
POMMS stool samples [20,40] , for which we have matching 16S rRNA amplicon and shotgun metagenomic
sequencing datasets. We also previously isolated Akkermansia strains from several of these samples . We
[20]
identified Akkermansia phylogroups from metagenomic sequencing data using a combination of three
strain-finding programs, StrainPhlAn3, StrainR, and SMEG [53,55,56] .

[69]
StrainPhlAn3 is broadly used to identify bacterial strains from metagenomic data . To identify
Akkermansia phylogroups, we ran metagenomic sequencing data generated from POMMS samples through
our StrainPhlAn3 workflow, which included 31 samples from which we had previously identified and
isolated the dominant Akkermansia species [20,70] . StrainPhlAn3 frequently failed to detect Akkermansia in
samples where the relative abundance was less than 0.5% of total bacterial sequences and was able to assign
a phylogroup to only 85 out of 146 POMMS metagenomes with levels of Akkermansia detectable by
MetaPhlAn3 [Supplementary Figure 4A].

SMEG is designed to identify bacterial strains and measure their growth rate based on metagenomic
sequences by calculating the coverage of SNPs closer to the origin of replication as compared to those closer
to the terminus region . We generated a SMEG database using 34 Akkermansia genomes representing
[56]
different species and strains and used the growth_est module to predict Akkermansia species. SMEG was
able to assign a species or phylogroup to 108 out of 146 POMMS metagenomic samples with detectable
(greater than 0% relative abundance) levels of Akkermansia, which allowed for improved sensitivity in
assigning phylogroups and for the estimation of growth rates in vivo. The results of a simple linear
regression indicate that the relative abundance of A. muciniphila is a significant predictor of its estimated
growth rate [R = 0.1281, F(1,52) = 7.642, P = 0.0079] [Figure 3A]. However, no significant association
2
2
between growth rate and relative abundance was found for either A. massiliensis [R = 0.0043,
F(1,22) = 0.0956, P = 0.7601] or A. biwaensis [R = 0.0094, F(1,5) = 0.0473, P = 0.8365]. Several samples
2
display high growth rates despite having a low abundance of A. massiliensis or A. biwaensis. Additionally,
the relative abundance of A. muciniphila explains only 12.81% of the variation in the estimated growth rate.
Thus, other factors, such as predation by phages, interbacterial competition, or differential adherence to the
GI epithelia, may play significant roles in determining the prevalence of Akkermansia species in stool.

StrainR is designed for the relative quantification of highly related strains from metagenomic datasets . We
[55]
generated a StrainR database using one representative Akkermansia genome per species and phylogroup,
and StrainR assigned a species or phylogroup to 139 out of 146 POMMS metagenomic samples with
detectable levels of Akkermansia, nearly doubling the sensitivity of StrainPhlAn3. Out of those 146 POMMS
samples with detectable levels of Akkermansia, 138 (94.52%) were found to contain a single Akkermansia
species. The remaining eight samples all contained A. muciniphila in addition to A. massiliensis and/or
A. biwaensis. Of the samples containing only A. muciniphila, 92 were predicted to contain both AmIa and
AmIb. The ratio of AmIa and AmIb was consistent (~23%/72%) and independent of the total relative
abundance of A. muciniphila in the samples [Supplementary Figure 4B]. This could either mean that
StrainR is unable to reliably distinguish between AmIa and AmIb or the less likely scenario that AmIa and
AmIb strains co-occur at fixed ratios.

To validate the effectiveness of these methods, the Akkermansia species and subspecies in POMMS patients
were compared against strains isolated from the corresponding stool samples. Out of 31 POMMS