Page 21 - Read Online

Ladeira et al. Microbiome Res Rep 2023;2:9 https://dx.doi.org/10.20517/mrr.2023.01 Page 3 of 15

study characterized the strain dynamics, pangenome, and genomic diversity of the main Bifidobacterium
[6]
species from the human gut in early life with MAGs . However, functional pangenomic analyses of
Bifidobacterium in the adult human gut are lacking.

We performed an exploratory analysis of the ecology of Bifidobacterium based on data from a public
database containing human gut microbiome data, mostly for adult subjects. We first confirmed the
previously reported associations with health, age, and other factors. We then identified Bifidobacterium
partitions of the gut differing in terms of the abundance of Bifidobacterium, species composition, gut
microbiome features, and health status. Finally, using MAG-based pangenomic analysis, we showed that the
prevalence of some functional features of some Bifidobacterium species differed between health-associated
Bifidobacterium partitions. This study paves the way for more precise approaches to guide the selection of
Bifidobacterium strains for gut microbiome complementation in adulthood and, ultimately, human health.

METHODS
Pooled metagenomic studies dataset
We extracted taxonomic data from the curatedMetagenomicData (cMD) R package (Pasolli et al.) (version
3.0, release 2021), which consists of manually curated metadata together with all the taxonomic read counts
aggregated per species with MetaPhlAn3, for 86 studies (17,959 samples). Gut metagenomes with more than
five million reads were retained, and one duplicate study (referred to as “LeChatelierE_2013”) was excluded.
The read counts for the samples were sum-collapsed by genus. The resulting feature table was rarefied to a
depth of 1.000.000 counts per sample for alpha diversity analysis. Filtering for origin (stools), with the
selection of one fecal sample per subject (highest number of reads), resulted in 9,515 unique samples (61
studies). This dataset was used for a global description of the abundance of Bifidobacterium and the
prevalence of Bifidobacterium species across different metadata curated in cMD: age, lifestyle, antibiotic use
status, and health status.

The age categories were as follows: newborn (< 1 year of age), child (age ≥ 1 year and < 12 years), school-age
individuals (age ≥ 12 and < 19 years); adult (age ≥ 19 years), senior (> 65 years).

Lifestyle was classified as westernized or non-westernized, and antibiotic use was classified as yes (the
month preceding stool sample collection) or no.

Health-related metadata were aggregated into six categories as follows: control: subject known to be healthy;
adenoma: patients with all types and subtypes of adenoma; colorectal: patients with colorectal cancers
including metastases; metabolic: patients with metabolic conditions including atherosclerotic cardiovascular
disease, hypercholesterolemia, hypertension, type 2 diabetes, and impaired glucose tolerance; bowel: patients
with inflammatory bowel disease (IBD); arthritis: patients with rheumatoid arthritis or Behçet’s disease
(BD).

Bifidobacterium-based clustering of the gut microbiome
Samples were partitioned by applying Dirichlet’s Multinomial Mixture (DMM) modeling to the microbiota
[24]
data for 32 detected Bifidobacterium species with counts across cMD. We filtered the 9.515 datasets as
follows to obtain a final dataset relating to 5.329 subjects for DMM: (1) We retained individuals who had
not had antibiotic treatment as declared in the cMD (antibiotic use = no) (N = 216) or without information
(N = 3.571) to prevent bias in the diversity calculation; (2) We excluded subjects with a total count < 500, to
overcome Bifidobacterium underdetection issues (N = 19); (3) We excluded subjects with no
Bifidobacterium species total reads count as a DMM standard (N = 380).