Self-reinoculation with fecal flora changes microbiota density and composition leading to an altered bile-acid profile in the mouse small intestine

RESEARCH

Open Access

Self-reinoculation with fecal flora changes

microbiota density and composition

leading to an altered bile-acid profile in the

mouse small intestine

Said R. Bogatyrev

, Justin C. Rolando

and Rustem F. Ismagilov

1,2*

Abstract

Background:

The upper gastrointestinal tract plays a prominent role in human physiology as the primary site for

enzymatic digestion and nutrient absorption, immune sampling, and drug uptake. Alterations to the small intestine

microbiome have been implicated in various human diseases, such as non-alcoholic steatohepatitis and

inflammatory bowel conditions. Yet, the physiological and functional roles of the small intestine microbiota in

humans remain poorly characterized because of the complexities associated with its sampling. Rodent models are

used extensively in microbiome research and enable the spatial, temporal, compositional, and functional

interrogation of the gastrointestinal microbiota and its effects on the host physiology and disease phenotype.

Classical, culture-based studies have documented that fecal microbial self-reinoculation (via coprophagy) affects the

composition and abundance of microbes in the murine proximal gastrointestinal tract. This pervasive self-

reinoculation behavior could be a particularly relevant study factor when investigating small intestine microbiota.

Modern microbiome studies either do not take self-reinoculation into account, or assume that approaches such as

single housing mice or housing on wire mesh floors eliminate it. These assumptions have not been rigorously

tested with modern tools. Here, we used quantitative 16S rRNA gene amplicon sequencing, quantitative microbial

functional gene content inference, and metabolomic analyses of bile acids to evaluate the effects of self-

reinoculation on microbial loads, composition, and function in the murine upper gastrointestinal tract.

Results:

In coprophagic mice, continuous self-exposure to the fecal flora had substantial quantitative and

qualitative effects on the upper gastrointestinal microbiome. These differences in microbial abundance and

community composition were associated with an altered profile of the small intestine bile acid pool, and,

importantly, could not be inferred from analyzing large intestine or stool samples. Overall, the patterns observed in

the small intestine of non-coprophagic mice (reduced total microbial load, low abundance of anaerobic microbiota,

and bile acids predominantly in the conjugated form) resemble those typically seen in the human small intestine.

Conclusions:

Future studies need to take self-reinoculation into account when using mouse models to evaluate

gastrointestinal microbial colonization and function in relation to xenobiotic transformation and pharmacokinetics

or in the context of physiological states and diseases linked to small intestine microbiome and to small intestine

dysbiosis.

Keywords:

Microbial quantification, Metabolomics analyses, Mouse models, Small intestine microbiota, Bile acids,

Deconjugation, Coprophagy, Microbial colonization, 16S rRNA gene amplicon sequencing

Open Access

This article is distributed under the terms of the Creative Commons Attribution 4.0

International License (

http://creativecommons.org/licenses/by/4.0/

), which permits unrestricted use, distribution, and

reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to

the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver

(

http://creativecommons.org/publicdomain/zero/1.0/

) applies to the data made available in this article, unless otherwise stated.

* Correspondence:

rustem.admin@caltech.edu

Division of Biology and Biological Engineering, California Institute of

Technology, Pasadena, CA, USA

Division of Chemistry and Chemical Engineering, California Institute of

Technology, 1200 E. California Blvd, Pasadena, CA, USA

Bogatyrev

et al. Microbiome

(2020) 8:19

https://doi.org/10.1186/s40168-020-0785-4

Background

The small intestine is the primary site for enzymatic di-

gestion and nutrient uptake, immune sampling, and drug

absorption in the human gastrointestinal system. Its

large surface area vastly exceeds that of the large intes-

tine [

1], and thus may serve as a broad interface for

host-microbial interactions.

A growing body of scientific evidence highlights the

importance of the small intestine microbiome in normal

human physiology and response to dietary interventions

[

2, 3]. Alterations in the small intestine microbiome are

implicated in a number of human disorders, such as

malnutrition [

4, 5], obesity, and metabolic disease [

6], in-

flammatory bowel disease (IBD) and irritable bowel syn-

drome (IBS) [

–

9], and drug side effects [

10]. Despite

the apparent importance of the small intestine micro-

biome in human health, it remains understudied and

poorly characterized largely because of the procedural

and logistical complexities associated with its sampling

in humans (methods are too invasive and require spe-

cialized healthcare facilities). Moreover, microbial com-

position tends to differ substantially among the small

intestine, large intestine, and stool of the same animal or

human subject [

11, 12], which highlights the importance

of targeted sampling of the small intestine for analyses.

Mice are the predominant animal species of model or-

ganisms in the field of microbiome research. Compared

with other mammalian models, mice have a lower cost

of maintenance, their environment and diet can be easily

controlled, they are amenable to genetic manipulation,

there are numerous genetic mouse models already avail-

able, and propagation using inbred colonies reduces

inter-individual variability [

13]. Additionally, murine

germ-free (GF) and gnotobiotic technologies are well

established. Using mouse models enables interrogation

of the entire gastrointestinal tract (GIT) and examin-

ation of the changes in microbiome and host physiology

that occur in response to experimental conditions (e.g.,

dietary modifications, xenobiotic administration) or mi-

crobial colonization (e.g., monocolonization, colonization

with defined microbial consortia, human microbiota-

associated mice).

Rodent models also have several well-recognized limita-

tions associated with their genetic, anatomical, and

physiological differences with humans [

13, 14]. Among

these limitations is the persistent tendency of rodents to

practice gastrointestinal auto- and allo-reinoculation with

large intestine microbiota (via fecal ingestion, or

coprophagy) in laboratory settings [

–

17]. This pervasive

behavior has been documented in classical studies using

observational techniques in both conventional and GF

mice [

18], in conventional mice maintained on standard

and fortified diets [

19], in animals with and without access

to food [

20], and across different mouse strains [

16, 21].

Multiple classical studies have attempted to evaluate

the effects of self-reinoculation on the structure of the

microbiota in the rodent small intestine [

–

24] and

large intestine and stool [

20, 23, 25, 26] using traditional

microbiological techniques, but reported conflicting re-

sults [

, 25, 26]. This lack of consensus may be attrib-

uted to the use of different methods for preventing

coprophagy (some of which are ineffective), non-

standardized diets, inter-strain or inter-species differ-

ences among the animal models, or other unaccounted

for experimental parameters. It has been also suggested

that repeated self-exposure in mice via coprophagy can

promote microbial colonization of the GIT by

“

exogen-

ous

”

microbial species, such as

Pseudomonas

spp. [

27].

All of these observations highlight the importance of

considering self-reinoculation in studies of gastrointes-

tinal microbial ecology in murine models. However, the

field currently lacks precise and comprehensive evalua-

tions of the effects of self-reinoculation on the spatial,

structural, and functional state of the gut microbiome

and its effects on murine host physiology. Current

microbiome studies in rodents either do not take self-

reinoculation into account, or assume it can be elimi-

nated by single housing of animals or housing them on

wire mesh floors (also referred to as

“

wire screens

”

“

wire grids

”

)[14]. Despite classical literature suggesting

these assumptions can be incorrect [

16, 21, 28

–

32], they

have not been tested on mice housed in modern facilities

using state-of-the-art quantitative tools.

Here, we explicitly test these assumptions about mur-

ine self-reinoculation to answer the following three

questions relevant to gastrointestinal microbiome re-

search: (1) Do quantitative 16S rRNA gene amplicon se-

quencing tools detect differences in small intestine

microbial loads between mice known to be coprophagic

and non-coprophagic? (2) Does coprophagy impact the

microbial composition of the small intestine? (3) Do dif-

ferences in microbiota density and composition associ-

ated with self-reinoculation in mice impact microbial

function (e.g., alter microbial metabolite production or

modifications) in the small intestine?

To answer these questions, we analyzed gastrointes-

tinal samples from mice under conditions known to pre-

vent coprophagy (fitting with

“

tail

”

“

fecal collection

”

cups [

16, 23, 26, 30, 33]) and typical laboratory condi-

tions in which mice are known to be coprophagic (hous-

ing in standard cages). We also included samples from

single-housed mice in standard and wire-floor cages. We

analyzed the quantitative and compositional changes in

the microbiome along the entire length of the mouse

GIT in response to self-reinoculation, computationally

inferred the changes in microbial function, and evaluated

the microbial function-related metabolite profiles in the

corresponding segments of the gut.

Bogatyrev

et al. Microbiome

(2020) 8:19

Page 2 of 22

Results

We first performed a pilot study to confirm that pre-

venting coprophagy in mice would result in decreased

viable microbial load and altered microbiota compos-

ition in the small intestine. We used a most probable

number (MPN) assay utilizing anaerobic BHI-S broth

medium to evaluate the live (culturable) microbial loads

along the entire GIT of mice known to be coprophagic

(housed in standard cages in groups,

= 5) and mice

known to be non-coprophagic (fitted with tail cups and

housed in standard cages in groups,

= 5). Consistent

with the published classical literature [

20, 24], we found

that coprophagic mice had significantly higher loads of

culturable microbes in their upper GIT than mice that

were non-coprophagic (Additional file

1: Figure S4A).

Moreover, the microbial community composition in the

proximal GIT, particularly in the stomach, of copropha-

gic mice more closely resembled the microbial compos-

ition of the large intestine (Additional file

1: Figure S4B)

as revealed by 16S rRNA gene amplicon sequencing

(

= 1 mouse analyzed from each group) and principal

component analysis (PCA) of the resulting relative abun-

dance data.

This pilot study confirmed that in our hands, tail cups

were effective at preventing the self-reinoculation of

viable fecal flora in the upper GIT of mice. These results

spurred us to design a rigorous, detailed study (Fig.

1)to

answer the three questions posed above using state-of-

the-art methods: quantitative 16S rRNA gene amplicon

sequencing (to account for both changes in the total mi-

crobial load and the unculturable taxa), quantitative

functional gene content inference, and targeted bile-acid

metabolomics analyses.

The study design (Fig.

1) consisted of six cages of four

animals each that were co-housed for 2

–

6 months and

then split into four experimental groups and singly

housed for 12

–

20 days. The four experimental condi-

tions were the following: animals fitted with functional

tail cups (TC-F) and singly housed in standard cages, an-

imals fitted with mock tail cups (TC-M) and singly

housed in standard cages, animals singly housed on wire

floors (WF), and control animals singly housed in stand-

ard conditions (CTRL). At the end of the study, gastro-

intestinal contents and mucosal samples were collected

from all segments of the GIT of each animal and we

evaluated total microbial loads (entire GIT) and micro-

biome composition (stomach (STM), jejunum (SI2), and

cecum (CEC)).

We chose the cecum segment of the large intestine for

quantitative 16S rRNA gene amplicon sequencing

Fig. 1

An overview of the study design and timeline.

Mice from two age cohorts (4-month-old and 8-month-old) were raised co-housed (four

mice to a cage) for 2

–

6 months. One mouse from each cage was then assigned to one of the four experimental conditions: functional tail cups

(TC-F), mock tail cups (TC-M), housing on wire floors (WF), and controls housed in standard conditions (CTRL). All mice were singly housed and

maintained on each treatment for 12

–

20 days (

= 24, 6 mice per group).

Samples were taken from six sites throughout the gastrointestinal

tract. Each sample was analyzed by quantitative 16S rRNA gene amplicon sequencing of lumenal contents (CNT) and mucosa (MUC) and/or

quantitative bile-acid analyses of CNT. Panel

is adapted from [

13, 34])

Bogatyrev

et al. Microbiome

(2020) 8:19

Page 3 of 22

because the analysis of the contents of this section can

provide a complete snapshot of the large intestine and

fecal microbial diversity in response to environmental

factors [

–

37]. Cecal contents also enabled us to collect

a more consistent amount of sample from all animals

across all experimental conditions (whereas defecation

may be inconsistent among animals at the time of ter-

minal sampling).

Self-reinoculation increases microbial loads in the upper

gut

To answer our first question (Can quantitative sequen-

cing tools detect the difference in 16S rRNA gene DNA

copy load in the upper GIT of mice known to be copro-

phagic and non-coprophagic?), we analyzed total quanti-

fiable microbial loads across the GIT using 16S rRNA

gene DNA quantitative PCR (qPCR) and digital PCR

(dPCR). Preventing self-reinoculation in mice equipped

with functional tail cups dramatically decreased the lu-

menal microbial loads in the upper GIT but not in the

lower GIT (Fig.

2a). Total quantifiable microbial loads in

the upper GIT were reduced only in mice equipped with

functional tail cups. All other experimental groups of

singly-housed animals (those equipped with mock tail

cups, housed on wire floors, or housed on standard

woodchip bedding) that retained access to fecal matter

and practiced self-reinoculation had similarly high mi-

crobial loads in the upper GIT, as expected from the

published literature [

16, 21, 28

–

32].

Across all test groups, mucosal microbial loads in the

mid-small intestine demonstrated high correlation

(Pearson

’

= 0.84,

= 2.8 × 10

−

) with the microbial

loads in the lumenal contents (Fig.

2b).

Stomach (STM) and small intestine (SI1, SI2, and SI3)

samples from one (out of six) of the TC-F mice showed

higher microbial loads compared with the other TC-F

mice. The total microbial load in the upper GIT in this

TC-F mouse was similar to mice from all other groups

(TC-M, WF, CTRL), which emphasizes the crucial im-

portance of performing analyses of both microbial load

and composition (discussed below) on the same samples.

Self-reinoculation substantially alters the microbiota

composition in the upper gut but has less pronounced

effects in the large intestine

To answer our second question (Does self-reinoculation

with fecal microbiota impact upper GIT microbial com-

position?), we performed quantitative 16S rRNA gene

amplicon sequencing [

38, 39] (Barlow JT, Bogatyrev SR,

Ismagilov RF: A quantitative sequencing framework for

absolute abundance measurements of mucosal and lu-

menal microbial communities, submitted) on the stom-

ach (STM), jejunum (SI2), and cecum (CEC) samples.

Qualitative sequencing revealed dramatic overall changes

in the upper GIT microbiota caused by self-

reinoculation (Fig.

3). An exploratory PCA performed on

the multidimensional absolute microbial abundance pro-

files highlights the unique and distinct composition of the

upper GIT microbiome of non-coprophagic mice (Fig.

a). It is noteworthy that the stomach (STM) and small in-

testine (SI2) microbiota in all coprophagic mice clustered

closer to the large intestine microbiota, suggesting the

Fig. 2

Quantification of microbial loads in lumenal contents and mucosa of the gastrointestinal tracts (GIT) of mice in the four experimental

conditions: functional tail cups (TC-F), mock tail cups (TC-M), housing on wire floors (WF), and controls housed in standard conditions (CTRL).

Total 16S rRNA gene DNA copy loads, a proxy for total microbial loads, were measured along the GIT of mice of all groups (STM = stomach; SI1 =

upper third of the small intestine (SI), SI2 = middle third of the SI, SI3 = lower third of the SI roughly corresponding to the duodenum, jejunum,

and ileum respectively; CEC = cecum; COL = colon). Multiple comparisons were performed using a Kruskal

–

Wallis test, followed by pairwise

comparisons using the Wilcoxon

–

Mann

–

Whitney test with false-discovery rate (FDR) correction. Individual data points are overlaid onto box-and-

whisker plots; whiskers extend from the quartiles (Q2 and Q3) to the last data point within 1.5 × interquartile range (IQR).

Correlation between

the microbial loads in the lumenal contents (per gram total contents) and in the mucosa (per 100 ng of mucosal DNA) of the mid-SI.

= 6 mice

per experimental group

Bogatyrev

et al. Microbiome

(2020) 8:19

Page 4 of 22

Fig. 3

(See legend on next page.)

Bogatyrev

et al. Microbiome

(2020) 8:19

Page 5 of 22

similarity was due to persistent self-reinoculation with the

large intestine microbiota (Fig.

3a).

Self-reinoculation had differe

ntial effects across microbial

taxa (Fig.

3c), which could be classified into three main cat-

egories depending on the pattern of their change as follows:

“

Fecal taxa

”

(e.g.,

Clostridiales

Bacteroidales,

Erysipelotrichales

) that either dropped significantly

or disappeared (fell below the lower limit of

detection [LLOD] of the quantitative sequencing

method [

38] (Barlow JT, Bogatyrev SR, Ismagilov

RF: A quantitative sequencing framework for

absolute abundance measurements of mucosal and

lumenal microbial communities, submitted)) in the

upper GIT of non-coprophagic mice;

“

True small intestine taxa

”

(e.g.,

Lactobacillales

)

that remained relatively stable in the upper GIT in

non-coprophagic mice;

3. Taxa that had lower absolute abundance in the

cecum (e.g.,

Bacteroidales

Erysipelotrichales

Betaproteobacteriales

) of non-coprophagic (com-

pared with coprophagic) mice.

Overall, the composition of the small intestine micro-

biota of coprophagic mice was consistent with that pre-

viously reported in literature [

35]. The upper GIT

microbiota in non-coprophagic mice was dominated by

Lactobacilli

(Fig.

3c), known to be a prominent micro-

bial taxon in human small intestine microbiota [

3, 40,

]. Importantly, the compositional analysis showed that

the single TC-F mouse that had high microbial loads in

its stomach and small intestine had a microbial compos-

ition in those segments of the GIT similar (i.e., domi-

nated by

Lactobacillales

) to all other TC-F mice, and

very distinct from all coprophagic mice (Fig.

3b, c). The

PCA showed that the stomach and mid-small intestine

of this mouse clustered with the stomach and mid-small

intestine of all other TC-F mice (Fig.

3a).

Changes in the small intestine microbiota lead to

differences in inferred microbial functional gene content

We hypothesized that the quantitative and qualitative

changes in the small intestine microbiota induced by

self-reinoculation may result in altered microbial

function [

42, 43] and an altered metabolite profile, either

indirectly, as a result of functional changes in the micro-

biota, or directly via re-ingestion of fecal metabolites. To

understand how such alterations to microbiota would

impact microbial function in the small intestine, we next

aimed to predict how the absolute abundances of func-

tional microbial genes would be affected. We coupled

the pipeline for microbial functional inference based on

the 16S rRNA marker gene sequences (PICRUSt2) [

44,

] with our quantitative 16S rRNA gene amplicon se-

quencing approach [

38] (Barlow JT, Bogatyrev SR, Isma-

gilov RF: A quantitative sequencing framework for

absolute abundance measurements of mucosal and lu-

menal microbial communities, submitted). We focused

our analysis on microbial functions that would be highly

relevant to small intestine physiology: microbial conver-

sion of host-derived bile acids and microbial modifica-

tion of xenobiotics.

We found that the inferred absolute abundances of a

number of microbial gene orthologs implicated in en-

zymatic hydrolysis of conjugated bile acids (bile salt

hydrolase, BSH [

–

48]) and xenobiotic conjugates (e.g.,

beta-glucuronidase, arylsulfatase [

49, 50]) in the stomach

and the small intestine of coprophagic mice were dra-

matically higher (in some cases by several orders of mag-

nitude) than in non-coprophagic mice (Fig.

4). This

difference was not observed in the cecum.

Changes in the small intestine microbiota induced by

self-reinoculation alter the bile acid profile

Bile acids are a prominent class of host-derived com-

pounds with multiple important physiological functions

and effects on the host and its gut microbiota [

51, 52].

These host-derived molecules are highly amenable to

microbial modification in both the small and large intes-

tine [

53]. The main microbial bile acid modifications in

the GIT include deconjugation, dehydrogenation, dehy-

droxylation, and epimerization [

52]. Thus, we next per-

formed quantitative bile acid profiling along the entire

GIT to evaluate the effects of self-reinoculation on bile

acid composition.

The small intestine is the segment of the GIT that har-

bors the highest levels of bile acids (up to 10 mM) and

where they function in lipid em

ulsification a

nd absorption

(See figure on previous page.)

Fig. 3

Compositional and quantitative 16S rRNA gene amplicon sequencing analysis of the gut microbiota.

Principal component analysis (PCA)

of the log

-transformed and standardized (mean = 0, SD = 1) absolute microbial abundance profiles in the stomach, mid-small intestine, and

cecum. Loadings of the top contributing taxa are shown for each principal component.

Mean relative and absolute abundance profiles of

microbiota in the mid-SI (order level) for all experimental conditions. Functional tail cups (TC-F), mock tail cups (TC-M), housing on wire floors

(WF), and controls housed in standard conditions (CTRL). N = 6 mice per experimental group, 4 of which were used for sequencing.

Absolute

abundances of microbial taxa (order level) compared between coprophagic and non-coprophagic mice along the mouse GIT. *Chloroplast and

*Richettsiales (mitochondria) represent 16S rRNA gene DNA amplicons from food components of plant origin. Multiple comparisons were

performed using the Kruskal

–

Wallis test

Bogatyrev

et al. Microbiome

(2020) 8:19

Page 6 of 22

Fig. 4

Inference of microbial genes involved in bile-acid and xenobiotic conjugate modification along the GIT of coprophagic and non-

coprophagic mice. Inferred absolute abundance of the microbial genes encoding (

) bile salt hydrolases (cholylglycine hydrolases), (

) beta-

glucuronidases, and (

) arylsulfatases throughout the GIT (

STM

stomach,

SI2

middle third of the small intestine (SI) roughly corresponding to the

jejunum,

CEC

cecum). KEGG orthology numbers are given in parentheses for each enzyme. In all plots, individual data points are overlaid onto

box-and-whisker plots; whiskers extend from the quartiles (Q2 and Q3) to the last data point within 1.5 × interquartile range (IQR). Multiple

comparisons were performed using the Kruskal

–

Wallis test; pairwise comparisons were performed using the Wilcoxon

–

Mann

–

Whitney test with

FDR correction.

= 4 mice per group

Fig. 5

Bile acid profiles in gallbladder bile and in lumenal contents along the entire GIT.

Total bile acid levels (conjugated and unconjugated;

primary and secondary) and

the fraction of unconjugated bile acids in gallbladder bile and throughout the GIT (

STM

stomach;

SI1

upper third

of the small intestine (SI),

SI2

middle third of the SI,

SI3

lower third of the SI roughly corresponding to the duodenum, jejunum, and ileum

respectively;

CEC

cecum;

COL

colon). In all plots, individual data points are overlaid onto box-and-whisker plots; whiskers extend from the

quartiles (Q2 and Q3) to the last data point within 1.5 × interquartile range (IQR). Multiple comparisons were performed using the Kruskal

–

Wallis

test; pairwise comparisons were performed using the Wilcoxon

–

Mann

–

Whitney test with FDR correction.

= 6 mice per group

Bogatyrev

et al. Microbiome

(2020) 8:19

Page 7 of 22

[54

–

56]. Given these high concen

trations of bile acid sub-

strates, we specifically wished to analyze whether the differ-

ences we observed in small int

estine microbiota (Figs.

2 and

) between coprophagic and no

n-coprophagicmicewould

result in pronounced effects on microbial deconjugation of

bile acids. We also wished to test whether any differences

in bile acid deconjugation w

ere in agreement with the

differences in the absolute BSH gene content we inferred

(Fig.

4a) from the absolute microbial abundances (Fig.

3c).

We first confirmed that in all four experimental groups,

total bile acids levels (conjugated and unconjugated; pri-

mary and secondary) across all sections of the GIT were

highest in the small intestine (Fig.

5a). We then compared

the levels of conjugated and unconjugated (Fig.

5b) as well

as primary (host-synthesized) and secondary (microbe-

modified) bile acids (Additional file

1: Figure S5) between

coprophagic and non-coprophagic mice.

Across all sections of the GIT and in the bile, non-

coprophagic mice (TC-F) had significantly lower levels of

unconjugated bile acids compared with coprophagic mice

(Fig.

5b). Consistent with the computational inference in

Fig.

4a (performed on mid-SI samples only), in all three

sections of the small intestine of non-coprophagic mice

(TC-F), the levels of unconjugated bile acids were substan-

tially lower than in coprophagic mice. Almost 100% of the

total bile acid pool remained in a conjugated form in the

small intestine of non-coprophagic mice.

In all groups of coprophagic mice (TC-M, WF, and

CTRL) the fraction of unconjugated bile acids gradually

increased from the proximal to distal end of the small

intestine. Gallbladder bile acid profiling (Fig.

5b) con-

firmed that bile acids were secreted into the duodenum

predominantly in the conjugated form in all coprophagic

mice. This pattern is consistent with the hypothesis that

the exposure of bile acids to microbial deconjugation ac-

tivity increases as they transit down a small intestine

with high microbial loads (Fig.

2a) [ 54].

In the large intestine, non-coprophagic (TC-F) mice car-

ried a smaller fraction of unconjugated bile acids com-

pared with all coprophagic experimental groups (Fig.

5b).

Bile acid deconjugation in the small intestine of copro-

phagic mice was uniform for all glyco- and tauro-

conjugates of all primary and secondary bile acids measured

in our study (Additional file

1: Table S7), suggesting a

broad-specificity BSH activity was provided by a complex

fecal flora in the small intestine of those animals.

In the gallbladder bile and across all segments of the

GIT from the stomach to the cecum, non-coprophagic

mice had a statistically significantly lower fraction (but not

lower absolute levels) of total secondary bile acids (conju-

gated and unconjugated) than coprophagic mice (Add-

itional file

1: Figure S5). This change was uniform for the

entire secondary bile acid pool of those analyzed (Add-

itional file

1: Table S7). The only segment of the gut in

which the difference in the fraction of secondary bile acids

was not statistically significant between coprophagic and

non-coprophagic mice was the colon. In fact, the differ-

ences in the fractions of total unconjugated and total sec-

ondary bile acids between coprophagic and non-

coprophagic mice would have gone largely undetected had

we only analyzed colonic contents or stool. These findings

further highlight the importance of the comprehensive

spatial interrogation of the complex crosstalk between the

microbiota and bile acids in the gastrointestinal tract.

Discussion

In this study, we used modern tools for quantitative micro-

biota profiling and showed tha

t when self-reinoculation

with fecal flora is prevented, the mouse small intestine har-

bors dramatically lower densities of microbiota and an al-

tered microbial profile. Consistent with published literature

[

16, 21, 28

–

32], we confirmed that single housing on wire

floors failed to prevent mice from practicing coprophagy

and that only functional tail cups reliably prevented the

self-reinoculation with fecal flora.

Despite its effectiveness, the tail cup approach has lim-

itations. Tail cups in their current design may not be

suitable for female rodents due to anatomical differences

leading to urine entering and remaining inside the de-

vices [

57]. Animals need to be singly housed to prevent

them from gnawing on each other

’

s tail cups and caus-

ing device failure or injury. The tail cup approach may

be hard to implement in younger and actively growing

mice (e.g., before or around weaning). Some mice in our

study developed self-inflicted skin lesions from over-

grooming at the location where the tail cups come in

contact with the body at the animal

’

s hind end. Thus, we

concluded that the approach in its current implementa-

tion is limited to 2

–

3 weeks in adult animals.

Our device design reduced the risk of tail injury and ne-

crosis described in previous works [

33] and allows for

emptying the cups only once every 24 h to reduce hand-

ling stress. Because host stress can affect the microbiota

[

58] and other physiological parameters, we included a

mock tail cup group. Both TC-F and TC-M mice demon-

strated a similar degree of weight loss (Additional file

Figure S3A) when compared with the WF and CTRL mice

despite similar food intake rates across all four groups

(Additional file

1: Figure S3B). Mice fitted with mock tail

cups (TC-M) had microbial patterns and bile acid profiles

similar to the CTRL mice, thus the effects on the upper

GIT microbiota and bile acid profiles that we observed in

non-coprophagic (TC-F) mice are not attributable to stress.

We believe that the tail cup approach is implementa-

ble in gnotobiotic settings (e.g., flexible film isolators

and individually ventilated cages), which can aid studies

that involve association of mice with defined microbial

communities or with human-derived microbiota.

Bogatyrev

et al. Microbiome

(2020) 8:19

Page 8 of 22

The non-coprophagic mouse model may be more

relevant to humans

Using quantitative microbiota profiling, our study dem-

onstrated that preventing self-reinoculation dramatically

reduced the total levels of several prominent taxonom-

ical groups of obligate anaerobes (e.g.,

Clostridiales

Bac-

teroidales

Erysipelotrichale

) in the upper gastrointestinal

microbiota of conventional mice. Despite these differ-

ences in taxa, levels of

Lactobacillales

in the small intes-

tine and cecum, but not in the stomach, remained

similar between coprophagic and non-coprophagic ani-

mals (Fig.

3c). The physiological significance of the

maintained persistent population of

Lactobacillales

the upper gastrointestinal tract (e.g., stomach or small

intestine) and their overall consistent presence along the

entire GIT [

14, 59] for the host is not fully understood.

However,

Lactobacilli

colonization in the stomach and

small intestine has been shown to promote resistance to

colonization by pathogens (reviewed in [

60, 61]).

Compared with conventional (coprophagic) mice, the

non-coprophagic mice displayed features of the small in-

testine microbiota and bile acid profiles that are more

similar to the patterns seen in the small intestine of

humans: orders of magnitude lower microbiota density,

reduced abundance of obligate anaerobic flora and dom-

inance of

Lactobacillales

, and a higher ratio of conju-

gated bile acids. These findings highlight the need to

understand and control self-reinoculation in mouse

models used to answer questions relevant to host-

microbiota interactions in human health.

Self-reinoculation and microbial ecology in the mouse GIT

We observed that within the approximately 2-week

timeframe of our study, the taxonomical diversity of the

mouse large intestine microbiome was stable in the ab-

sence of persistent microbial self-reinoculation: all taxo-

nomical groups at the order level observed in the cecum

of coprophagic mice were present in the cecum of non-

coprophagic mice, and vice versa.

The trending changes in the absolute abundances of

several taxa (

Bacterodales, Erysipelotrichales

, and

Beta-

proteobacteriales

) in the large intestine of non-

coprophagic mice may be the result of eliminated self-

reinoculation and/or the consequence of the altered pro-

file of bile acids entering the cecum from the small in-

testine, or other undetected changes in the biochemical

environment. Additionally, changes in the absolute

abundance of some taxa may lead to changes in the ab-

solute abundances of other metabolically coupled taxa. It

has been previously suggested that the degree of bile

acid deconjugation may alter the microbiota profile [

46].

Erysipelotrichales

(including

Turicibacter

spp

) in the

mouse ileum and cecum have been shown to be posi-

tively correlated with unconjugated ileal and cecal [

62]

and plasma [

63] bile acids.

Bacteroidales

(including

Muribaculum

spp

) in the cecum increased upon dietary

supplementation of unconjugated cholic and cheno-

deoxycholic acids [

64] and

Betaproteobacteriales

(includ-

ing

Parasutterella

spp.) were positively correlated with

unconjugated primary and secondary bile acids [

65, 66]

in mice. Thus, the decrease in the fraction of unconju-

gated bile acids in the large intestine of non-coprophagic

mice (Fig.

5b) may be responsible for the decreased ab-

solute abundance of these three taxonomic groups in the

cecum of non-coprophagic mice. It is of note that most

of the published reports describe correlations between

bile acid profiles and microbiota composition based on

relative abundance data and without accounting for the

inherent compositionality of relative abundance data

[

67], which is known to introduce inaccuracies in the

correlation analysis [

64, 68]. For improved correlation

analysis, our study reports absolute abundances of the

taxa, which could lead to discrepancies between such

correlations observed in this study and previously pub-

lished studies.

Stability of complex microbiomes in response to per-

turbations with and without continuous species reintro-

duction is an important subject of research in microbial

ecology [

69, 70]. Eliminating fecal ingestion provides a

way to study stability and recovery of the mouse gut

microbiota (e.g., in response to dietary change or anti-

biotic exposure [

71]) in a way more relevant to modern

humans. Thus, the non-coprophagic mouse model can

significantly aid such research.

Self-reinoculation with fecal flora leads to altered bile

acid profiles in the GIT

We demonstrated that changes to small intestine micro-

biota density and composition had pronounced effects

on microbial function resulting in increased bile acid

deconjugation in that segment of the GIT. Bile acid

deconjugation is a microbiota-mediated process that in

healthy humans is conventionally believed to take place

in the distal small intestine (ileum) and in the large in-

testine [

72] such that sufficient lipid emulsification (with

conjugated bile acids) and absorption can take place in

the small intestine by the time digesta reaches the ileum

[

73]. As a result of the much higher bile acid concentra-

tions in the small intestine compared with the large in-

testine, altered deconjugation of bile acids in the small

intestine may have more wide-ranging effects on the en-

tire enterohepatic system. Our data indicate that bile

acid deconjugation can take place in any segment of the

small intestine of conventional healthy mice as a func-

tion of the microbial density and composition (Figs.

2a,

and 5b), which is consistent with previous findings in

animal models and in humans with small intestinal bac-

terial overgrowth (SIBO) [

–

78].

Bogatyrev

et al. Microbiome

(2020) 8:19

Page 9 of 22

Strikingly, the very low degree of bile acid deconjugation

in the small intestine of non-coprophagic mice in our

study resembles profiles seen in germ-free animals [

–

], gnotobiotic animals colonized only with microbes

incapable of deconjugating bile acids [

–

85], and

antibiotic-treated animals [

–

88]. Our observations sug-

gest a mechanistic link between the small intestine micro-

biota density and composition and the bile acid

modification in this segment of the GIT. The small intes-

tine of healthy human subjects is believed to harbor bile

acids predominantly in the conjugated form [

89], which

further substantiates that (compared with coprophagic

mice) the small intestine of non-coprophagic mice is more

similar to the small intestine of a healthy human.

Although microbiota density and composition in the

large intestine of coprophagic and non-coprophagic

mice were largely similar, non-coprophagic mice had a

higher fraction of bile acids that remained in the conju-

gated form in the large intestine (Fig.

4b), likely as a re-

sult of the bile acids entering the large intestine from

the ileum predominantly in a conjugated form. Add-

itionally, across all study groups, the total concentrations

of bile acids in the small intestine were ~ 10-fold greater

than in the large intestine. We therefore infer that in

coprophagic mice, a greater absolute amount of bile

acids underwent deconjugation in the small intestine

than in the large intestine; i.e., in coprophagic mice, the

small intestine contaminated with high loads of fecal

flora was the primary site of bile acid deconjugation.

Regulation of bile acid deconjugation activity in the

gut is considered a potential health-promoting modality

in a number of contexts, including lowering blood chol-

esterol levels (reviewed in [

–

92]). BSH-active probio-

tics can be a promising delivery vehicle for promoting

increased bile acid deconjugation in the gut. Our study

emphasizes the importance of controlling for self-

reinoculation when using mice to study the effects of

BSH-active microbial strains or probiotics [

48, 93

–

98]

(especially those with high selectivity for particular bile

acid conjugates [

47, 82, 85]) because conventional

(coprophagic) mice already have pronounced BSH activ-

ity in their small intestines. A non-coprophagic mouse

may be a better animal model in such studies.

Our findings also have implications for the use of con-

ventional (coprophagic) mice in diet studies. Deconjugated

bile acids are less effective than conjugated bile acids at

lipid emulsification and fat micelle formation [

74, 99]. In-

creased bile acid deconjugation in the small intestine of

animals and humans can lead to lipid malabsorption and

fat-soluble vitamin deficiency, and in extreme scenarios

even to steatorrhea [

77, 100 ]. Previous research has shown

that the small intestine microbiota plays an important role

in mediating the effect of high-fat diets on the host [

101 ];

our results suggest that future studies of the microbiota-

mediated effects of high-fat diets need to consider in-

creased microbial bile acid deconjugation in the mouse in-

testine due to self-reinoculation with fecal flora.

Bile acid deconjugation is considered to be obligatory

[

84, 102 , 103 ] before the secondary bile acid metabolism

(believed to be predominantly occurring in the large in-

testine [

72]) can take place. These reactions in many

cases are carried out by different members of the micro-

biota. Thus, the reduction of the deconjugation activity

in the small intestine of non-coprophagic mice and con-

sequently lower availability of free primary bile acids for

further microbial modification can explain the decrease

in the secondary bile acid fraction (percentage of all bile

acids) in the bile acid pool across the GIT and gallblad-

der bile of non-coprophagic mice in our study. A similar

but more pronounced trend has been observed in rabbits

[

104]. Reduced oral intake and recycling of fecal second-

ary bile acids as a result of eliminating coprophagy may

also be a contributing factor to the lower fraction of sec-

ondary bile acids in the total bile acid pool in the entero-

hepatic circulation in these animals.

Total bile acid levels in the stomach were similar in

coprophagic and non-coprophagic mice (and agree with

literature [

104, 105 ]); however, bile acid profiles (includ-

ing the fraction of total unconjugated and total second-

ary bile acids) were substantially different. Surprisingly,

in all coprophagic mice the fraction of unconjugated bile

acids in the stomach appeared to be intermediate be-

tween the profiles in the small intestine and in the large

intestine (Fig.

5b), suggesting that the bile acids in the

stomach of coprophagic mice could be accumulating

from bile acids re-ingested in feces and bile acids

refluxed from the duodenum. This pattern was not ob-

served in non-coprophagic mice, suggesting that

coprophagy may alter the bile acid profile in the upper

GIT both directly (via re-ingestion of fecal metabolites)

and indirectly (via altered microbiota function).

Inferences about microbial function in bile acid and drug

modification

Our quantitative functional gene inference analysis pre-

dicted differential absolute abundance of the BSH ortho-

logs between the small intestine of coprophagic and

non-coprophagic mice (Fig.

4a). This approach has limi-

tations associated with incomplete gene annotations,

limited ability to infer metagenomes from the marker

gene sequences when multiple microbial strains with

similar 16S rRNA gene sequences exist [

44, 45], diffi-

culty to predict the exact gene expression and enzyme

activity and specificity. To test our prediction about BSH

we employed a targeted bile acid metabolomic analysis

of mouse gastrointestinal samples and observed the dif-

ferences in the small intestine bile acid deconjugation be-

tween coprophagic and non-coprophagic mice (Fig.

5b)

Bogatyrev

et al. Microbiome

(2020) 8:19

Page 10 of 22

that were in agreement with the differences in the in-

ferred BSH gene abundances in the small intestine of

those two types of animals (Fig.

4a). Interestingly,

despite similar inferred BSH gene abundance in the

cecum of coprophagic and non-coprophagic mice, the

fraction of unconjugated bile acids in the cecum and

colon of non-coprophagic mice was statistically sig-

nificantly lower (Fig.

5b) compared with coprophagic

mice.

Michaelis

–

Menten constants (

) for many known

BSH isoforms are in the range of hundreds of nanomoles

[

72]

—

similar to the levels of bile acids observed in the

small intestine of all groups of mice in this study (Fig.

a). Total bile acid levels in the cecum and colon were

~ 10-fold lower than those in the small intestine, and

thus they were ~ 10-fold lower than the BSH

(Fig.

a). The predominantly conjugated form of the bile acids

arriving into the cecum from the small intestine of non-

coprophagic mice and their absolute concentration (~

10-fold lower than BSH

) can potentially explain the

lower degree of bile acid deconjugation in the large in-

testine of these animals (compared with coprophagic

mice) on the timescale of normal gastrointestinal transit.

This highlights the importance of considering func-

tional inference (based on either taxonomy or in silico

hidden state prediction [

44, 45, 106, 107]) in the context

of a variable biochemical environment (e.g., substrate

availability) and the host gastrointestinal physiology (se-

cretion, gastrointestinal transit, absorption and trans-

port, etc.) and warrants functional validation (e.g.,

metabolomics). Additionally, the validity of functional

inference based on 16S rRNA gene sequence counts

(from next generation sequencing) versus absolute 16S

rRNA gene sequence abundances (this study) should be

further explored in future work.

We next explored the effects of self-reinoculation on

the absolute abundance of microbial gene orthologs im-

plicated in xenobiotic modification [

108] in the small in-

testine, as microbiota-dependent drug modification and

toxicity in the small intestine have been previously ob-

served in rodents [

109

–

119]. Many drugs administered

to humans and mice both via enteral and parenteral

routes after reaching the systemic circulation are trans-

formed by the liver into conjugates (e.g., glucuronic acid,

sulphate, or glutathione conjugates) and excreted with

bile into the GIT lumen. Such transformations are be-

lieved to reduce the small intestine reabsorption of xe-

nobiotics and promote their excretion from the body

with stool. Alterations in the small intestine microbiota

may also lead to increased hydrolysis of such conjugates

by microbial enzymes and promote the local toxicity of

the drug and enable its re-uptake from the small intes-

tine (i.e., undergo enterohepatic circulation) [

10, 116 ],

resulting in an increase in the xenobiotic flux through

the liver [

120, 121] and to an overall microbiota-

dependent change in drug pharmacokinetics.

As with the inferred differential BSH absolute abun-

dances (correlating activity of which we confirmed with

the bile acid deconjugation measurements), our analysis

predicted differences in the absolute abundance (Fig.

4b,

c) of the microbial gene orthologs responsible for drug

conjugate hydrolysis (e.g., beta-glucuronidases, sulfohy-

drolases) between the small intestine of coprophagic and

non-coprophagic mice. If this prediction is further ex-

perimentally confirmed, it would imply that self-

reinoculation must be controlled for or taken into ac-

count when investigating drug pharmacology in mice.

Relevance of self-reinoculation in probiotics research

Many studies on probiotics and their effects on host ani-

mal physiology rely on repeated oral administration of

live probiotic microorganisms to rodents. Our study sug-

gests that self-reinoculation with live fecal flora in la-

boratory mice could both interfere with and introduce

inconsistencies in live probiotic administration regimens.

As has been stated earlier, particular attention should be

given to self-reinoculation and its effects on the small in-

testine bile acid profile in studies aiming to evaluate the

health effects of probiotics and other therapeutic modal-

ities [

48, 90

–

98] targeting bile acid deconjugation and

metabolism.

Relevance of mouse models in human microbiota

research

The role of mouse models in human microbiota research

remains a subject of debate [

13, 14, 122]. At the same

time, the field is recognizing the importance of reprodu-

cibility in gut microbiota research that uses mouse

models [

58, 122]. Several recent studies have highlighted

the variability in lab-mouse microbiota related to animal

strains and sources of origin [

36, 123

–

127]. Others have

attempted to catalog a

“

normal

”

“

core

”

gut micro-

biome [

128, 129] and its spatial organization [

35, 36]

and function [

130] in laboratory and wild mice. Recently,

the small intestine microbiome has become the focus of

studies conducted in mice in the context of host physi-

ology [

101] and disease [

4, 131]. Yet, little attention has

been given to the impact of self-reinoculation on the gut

microbiota spatial structure and function, or to how

study outcomes might be affected by controlling (or not

controlling) for this experimental parameter in mouse

models.

Self-reinoculation in rodents may affect not only their

native microbiota, but also individual microbial colo-

nizers [

24] (e.g., in gnotobiotic animals) and complex

xenomicrobiota (e.g., in human microbiota-associated

(HMA) mice). HMA mice have emerged as an important

research model for dissecting the mechanistic

Bogatyrev

et al. Microbiome

(2020) 8:19

Page 11 of 22