PNAS202118483_proof.pdf - pnas.2118483119.pdf

Three-dimensional imaging for the quanti

fi

cation of spatial

patterns in microbiota of the intestinal mucosa

Octavio Mondrag



on-Palomino

a,1

, Roberta Poceviciute

, Antti Lignell

a,b,2

, Jessica A. Griffiths

, Heli Takko

, and Rustem F. Ismagilov

a,b,3

Edited by Jeffrey Gordon, Washington University in St. Louis School of Medicine, St. Louis, MO; received October 18, 2021; accepted March 7, 2022

Improving our understanding of host

–

microbe relationships in the gut requires the abil-

ity to both visualize and quantify the spatial organization of microbial communities in

their native orientation with the host tissue. We developed a systematic procedure to

quantify the three-dimensional (3D) spatial structure of the native mucosal microbiota

in any part of the intestines with taxonomic and high spatial resolution. We performed

a 3D biogeographical analysis of the microbiota of mouse cecal crypts at different stages

of antibiotic exposure. By tracking eubacteria and four dominant bacterial taxa, we

found that the colonization of crypts by native bacteria is a dynamic and spatially orga-

nized process. Cipro

fl

oxacin treatment drastically reduced bacterial loads and elimi-

nated Muribaculaceae (or all Bacteroidetes entirely) even 10 d after recovery when

overall bacterial loads returned to preantibiotic levels. Our 3D quantitative imaging

approach revealed that the bacterial colonization of crypts is organized in a spatial

pattern that consists of clusters of adjacent colonized crypts that are surrounded by

unoccupied crypts, and that this spatial pattern is resistant to the elimination of Muri-

baculaceae or of all Bacteroidetes by cipro

fl

oxacin. Our approach also revealed that the

composition of cecal crypt communities is diverse and that Lactobacilli were found

closer to the lumen than Bacteroidetes, Ruminococcaceae, and Lachnospiraceae, regard-

less of antibiotic exposure. Finally, we found that crypts communities with similar taxo-

nomic composition were physically closer to each other than communities that were

taxonomically different.

microbiota

j

quantitative biogeography

j

tissue clearing

The composition of resident microbial communities is driven by nutrient availability

–

3), the physical environment (4, 5), host

–

microbiota interactions (6, 7), and inter-

actions within the microbiota (8, 9). The sum of all these forces may shape the spatial

arrangement of intestinal microbes, and, in turn, the spatial structure of the microbiota

could in

fl

uence how host

–

microbe and microbe

–

microbe interactions occur (10). The

synergy between the microgeography of intestinal bacterial consortia and the interac-

tions of microbes with their environment or other microbes has been studied in vitro

using synthetic communities and computational simulations (11

–

15). In the context of

the gastrointestinal system, studying the connection between the native spatial structure

of the microbiota and its function naturally calls for three-dimensional (3D) imaging

strategies that enable the simultaneous visualization of bacterial communities and host

structures at multiple scales (16, 17). However, existing 3D imaging approaches remain

hindered by the opacity of intestinal tissues and their contents as well as their imperme-

ability to labeling probes. Methods have been developed to obtain cross-sectional slices

from paraf

fi

n- or plastic-embedded intestinal tissues (18

–

20). Thin sections eliminate

the optical and diffusion barriers that thick tissues present to imaging and molecular

staining, but they fragment host tissues and microbial assemblies. The advent of tissue-

clearing technologies has enabled the imaging of cellular structures in thick tissues such

as the brain (21, 22). However, the full potential of tissue-clearing techniques has yet

to be realized to quantify the composition and organization of the host

–

microbiota

interface with spatial resolution.

Sequencing of bacterial 16S ribosomal RNA (rRNA) genes has been effective at sur-

veying the composition of the bacterial microbiota in different compartments along

and across the gastrointestinal tract (GIT). Indeed, sequencing has revealed that the

mucosal microbiota is distinct and spatially heterogeneous, and bioinformatics tools

have enabled the inference of bacterial networks of interaction (23

–

33). However,

sequencing alone cannot be used to reconstruct the spatial distribution of bacteria rela-

tive to the host with high spatial resolution. Therefore, microscopic imaging of thin

sections of intestinal tissue is the de facto approach to study the

fi

ne spatial structure of

the microbiota and the host (2, 18, 19, 34). Thin-section imaging (TSI) is ordinarily

coupled with

fl

uorescence in situ hybridization (FISH), immunohistochemistry, and

Signi

fi

cance

Many human diseases are causally

linked to the gut microbiota, yet

the

fi

eld still lacks mechanistic

understanding of the underlying

complex interactions, because

existing tools cannot

simultaneously quantify microbial

communities and their native

context. In this work, we provide

an approach to tissue clearing and

preservation that enables 3D

visualization of the biogeography

of the host

–

microbiota interface.

We combine this tool with

sequencing and multiplexed

microbial labeling to provide the

fi

eld with a platform on which to

discover patterns in the spatial

distribution of microbes. We

validated this platform by

quantifying bacterial distribution

in cecal mucosa at different stages

of antibiotic exposure. This

approach may enable researchers

to formulate and test new

hypotheses about host

–

microbe

and microbe

–

microbe

interactions.

Author contributions: O.M.-P., R.P., J.A.G., and R.F.I.

designed research; O.M.-P., R.P., J.A.G., and H.T.

performed research; O.M.-P., R.P., J.A.G., and H.T.

contributed new reagents/analytic tools; O.M.-P., R.P.,

A.L., J.A.G., and H.T. analyzed data; O.M.-P. and R.F.I.

wrote the paper; O.M.-P. and R.F.I. conceived the

study; and R.F.I. supervised the research.

The authors declare no competing interest.

This article is a PNAS Direct Submission.

This open access article is distributed under

Creative

Commons Attribution License 4.0 (CC BY)

Present address: Laboratory of Parasitic Diseases,

National Institute of Allergy and Infectious Diseases,

Bethesda, MD 20892.

Present address: Department of Chemistry, University

of Helsinki, 00560 Helsinki, Finland.

To whom correspondence may be addressed. Email:

rustem.admin@caltech.edu.

This article contains supporting information online at

http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.

2118483119/-/DCSupplemental

Published April 27, 2022.

PNAS

2022 Vol. 119 No. 18 e2118483119

https://doi.org/10.1073/pnas.2118483119

1of12

RESEARCH ARTICLE

MICROBIOLOGY

OPEN ACCESS

Downloaded from https://www.pnas.org by George Porter on April 28, 2022 from IP address 131.215.225.183.

other labeling methods that link the molecular identity of bac-

teria and host elements to their location. For example, TSI has

been used to study the spontaneous segregation of

Escherichia

coli

and mucolytic bacteria in the colonic mucus layer (35), by

measuring the distance of different bacterial taxa from the

epithelial surface (19), such as during in

fl

ammation (36). In

notable recent examples of the quantitative application of TSI,

semiautomated computational image analysis was used to mea-

sure the thickness of the colonic mucus layer and the proximity

of bacteria to the host as a function of diet (18), and highly

multiplexed FISH was used to investigate the microscopic

spatial structure of microbiota in the distal colon (20).

Although TSI is valuable to investigate the biogeography of

the intestines and the microbiota, it is unable to completely

capture the spatial structure of bacterial communities in the

gut. The

fi

rst limitation of TSI is that it sets 2D bounds on the

spatial exploration of a heterogeneous, 3D system. TSI sections

are typically 5

mto10

m thick, whereas topographic epithe-

lial features and mucosal microbial communities can be one to

four orders of magnitude larger. Mucosal bio

fi

lms can be hun-

dreds of microns long (37), and bacterial colonies in the

colonic crypts have a heterogeneous taxonomic composition

with a 3D spatial structure that cannot be charted unless the

entire crypt (diameter 50

m) is imaged (30, 38).

Quantitative descriptions of the 3D spatial structure of

native bacterial bio

fi

lms with taxonomic resolution are chal-

lenging to develop because of the natural opacity of the intesti-

nal tissue and contents, and the complex composition of the

microbiota, in which potentially hundreds of bacterial species

coexist. Moreover, a quantitative description of a diverse and

spatially heterogeneous system requires abundant data that can

only be obtained through unrestricted optical access to samples.

Tissue-clearing techniques have been developed for some tissues

and organs (including brain, heart, kidney, lung, stomach, and

sputum) (22, 39

–

42). However, the direct application of tissue-

clearing techniques typically results in the loss of the delicate

mucus layer and associated bacterial communities (43). CLAR-

ITY (clear lipid-exchanged acrylamide-hybridized rigid imag-

ing/immunostaining/in situ hybridization-compatible tissue

hydrogel) and PACT (passive CLARITY technique) techniques

involve multiple mechanically stressful sample-preparation steps

to transform the cellular matrix of tissue into an acrylamide gel

(21, 22, 40). Moreover, application of CLARITY or PACT to

whole-mount tissues would irreversibly deform them and

destroy the patterns of bacterial colonization on the mucosa.

Here, we developed an advanced tissue-clearing technique that

preserved the spatial structure of the mucosal microbiota and the

host tissue, including the delicate mucus layer. We combined this

method with sequencing of 16S rRNA genes, ampli

fi

ed in situ

labeling of rRNA, spectral imaging, and statistical analyses. This

method is capable of revealing patterns in the composition of the

microbiota with taxonomic and spatial resolution. We used this

methodology to test the effects of antibiotic on the bacterial coloni-

zation of the intestinal mucosa. We were able to quantify patterns

in the spatial structure of the mucosal microbiota of the cecum at

multiple scales and at different stages of antibiotic exposure.

Results

Sample Preparation, Staining, and Imaging.

To achieve unre-

stricted optical access to the mucosa, we developed a tissue clar-

fi

cation method that exposes the intestinal mucosa in a fully

laid out display (Fig. 1). Mounting tissue samples

fl

at enabled

us to image any point of the mucosa using a standard confocal

microscope, and clearing the tissue increased the depth of imag-

ing with refractive index

–

matching long-working-distance

objectives (

SI Appendix

Supplementary Materials and Methods

However, to achieve optical transparency of exposed intestinal

tissues, we had to solve multiple experimental challenges.

To maintain the spatial integrity of bacteria and mucus during

whole-mount sample preparation, we developed a method that

addresses the preservation of the materials on the tissue surface

separately from the preservation of the rest of the sample, and

that minimizes the duration of steps that can dislodge mucus and

bio

fi

lms. The overall work

fl

ow of our method (Fig. 1

), which

we developed in a murine model, was as follows: After careful dis-

section and removal of intestinal contents, tissues were

fi

xed in

paraformaldehyde for 1 h to prevent biochemical decay. Next, we

created a capillary layer of acrylamide mix between the exposed

mucosa and the glass bottom of a shallow chamber. Upon heat-

ing, the acrylamide mix polymerized into a surface gel layer with

a thickness on the order of 100

l

m. Once the mucosal surface of

the sample was protected, the remainder of the tissue was embed-

ded and gelled. Finally, the uncovered surface of the sample (the

muscle side) was glued to a rigid,

fl

at, plastic substrate to keep

the sample

fl

at (Fig. 1

). In this con

fi

guration, samples could

be passively cleared, stained, and imaged without damaging

the mucosal surface. A detailed description of the work

fl

ow is

available here (

Materials and Methods

To locate bacteria in situ, we

fl

uorescently labeled bacterial

16S rRNA transcripts through hybridization chain reaction

(HCR) (44, 45) (

Materials and Methods

). Standard FISH probes

are labeled with up to two

fl

uorophores, which produce a

fl

uores-

cent emission that is suf

fi

ciently intense to image bacteria on thin

sections. However, bacteria in the mammalian gut can be found

in thick bio

fi

lms, epithelial crypts, or across the epithelial barrier,

all of which obscure visibility. Therefore, we used HCR for label-

ing because it increases the intensity of

fl

uorescence by at least

one order of magnitude compared with FISH probes (44).

The method presented here enables the mapping of bacteria

on the mucosa at multiple length scales. To reveal patterns of col-

onization over spatial scales on the order of centimeters, tissue

samples were imaged in a laser-scanning confocal microscope at

low magni

fi

cation (5

), and the images were tiled (Fig. 1

). To

image the detailed spatial structure of bacterial bio

fi

lms with

micrometer resolution (Fig. 1

and

Movie S1

), we mounted

samples in a refractive index

–

matching solution (RIMS;

1.46) and used a 20

CLARITY objective with a collar for the

compensation of spherical aberrations (

Materials and Methods

Sensitivity and Specificity of Bacterial Staining.

Sensitive and

speci

fi

cidenti

fi

cation of mucosal bacteria through

fl

uorescence

imaging was accomplished by optimizing HCR tagging and con-

trolling for off-target effects (

Materials and Methods

and

Appendix

Supplementary Materials and Methods

and Figs. S1

–

and S6

–

). Fluorescent tagging through HCR was achieved by

making the bacterial cell wall permeable to DNA probes and

HCR hairpins. However, the acrylamide gel sheet that we created

to protect the mucosal surface of samples formed a barrier for the

diffusion of lysozyme (Fig. 2

) that digests the bacterial peptido-

glycan. Poor permeabilization of bacteria limits the sensitivity of

imaging to bacteria closer to the mucosal surface and impedes the

detection of bacteria deep in the tissue samples. To determine the

correct concentration of lysozyme for optimal permeabilization of

the cell wall, we created acrylamide gel slabs and embedded them

with gram-positive (

Clostridium scindens

) and gram-negative (

Bac-

teroides fragilis

) bacteria. The purpose of these gels was to mimic

the geometry and composition of the acrylamide layer on tissue

2of12

https://doi.org/10.1073/pnas.2118483119

pnas.org

Downloaded from https://www.pnas.org by George Porter on April 28, 2022 from IP address 131.215.225.183.

samples.Thegelslabswereobtainedbyusingthesameprocedure

as in the preservation and clearing of tissues, had similar dimen-

sions to tissue samples, and were exposed to lysozyme on one

side only (

SI Appendix

Supplementary Materials and Methods

and

Figs. S1 and S2

). The duration of the treatment with lysozyme

was kept constant at 6 h, and we varied the concentration of lyso-

zyme in the range 1 mg/mL to 5 mg/mL to determine the opti-

mal concentration for bacterial permeabilization. Bacteria were

tagged with an HCR probe that included a eubacterial detection

sequence (eub338), and we imaged from the surface of the gels

to a depth of 600

m(Fig.2

). We measured the intensity of

HCR tagging of bacteria, which were identi

fi

ed with the blue-

fl

uorescent DNA intercalated dye DAPI. The sensitivity of our

method was de

fi

ned as the proportion of bacteria down to 600

mwitha

fl

uorescent signal-to-background ratio of

≥

20 (Fig.

). At a lysozyme concentration of 5 mg/mL, sensitivity was

94%, and it dropped to

∼

50% for 1 mg/mL.

Nonspeci

fi

c detection and ampli

fi

cation are potential sources

of background signal in HCR. Control experiments showed that,

in the absence of a target (germ-free [GF]

eub338) or a detect-

ing probe (speci

fi

c-pathogen-free [SPF]

non338), there was no

ampli

fi

cation, whereas, when both the target and the probe were

present (SPF

eub338), there was ampli

fi

cation (Fig. 2

–

)

(

SI Appendix

Supplementary Materials and Methods

). Plotting the

intensity values showed that in situ HCR tagging of bacteria pro-

duced a signal that is 8.5 to 9 times as strong as the background

in 90% of bacteria (Fig. 2

General 3D Spatial Organization of Bacteria in the IIeum,

Cecum, and Proximal Colon.

To evaluate our 3D imaging meth-

ods, we imaged bacteria, mucus, and the host epithelium in

disparate sections of the GIT with different biological functions,

mucosal topographies, and amounts of mucosal materials (46, 47).

Proximal colon.

At low magni

fi

cation (5

), we observed the

crests and valleys of the epithelial folds and that most of the

mucosa was covered by food particles and mucus (

SI Appendix

Fig. S5

). At higher magni

fi

cation (20

), our method enabled

the exploration of the 3D organization of the host

–

microbiota

interface in the proximal colon (Fig. 3

and

SI Appendix

, Fig.

). The 3D imaging can be analyzed through digital cross-

sections with arbitrary orientation and thickness. Examining

digital cross-sections, we found that bacteria were mixed with

mucus threads and granules in a layer that had an average

thickness of 125

m (Fig. 3

B and C

and

Movie S2

). We also

found that bacteria were separated from the epithelium by a

single layer of mucus with an average thickness of 22

m. The

3D imaging provides the ability to examine tissues in their

totality through computational 3D rendering. Thus, we were

we able to scan the tissue and

fi

nd rare but conspicuous loca-

tions where bacteria had penetrated the mucus layer or crossed

it and reached a crypt and the subepithelial space (Fig. 3

–

Ileum.

At low magni

fi

cation (5

), imaging revealed that bacteria

were not uniformly distributed throughout villi and were mostly

found as part of large agglomerations of food particles and mucus

that adhere to the epithelium (

SI Appendix

,Fig.S5

). At higher

magni

fi

cation (20

), 3D imaging showed that bacteria were con-

tained by mucus to a layer near the top of villi (Fig. 3

and

Cecum.

The epithelial layer of the murine cecum is organized as a

regular array of recessed mucus-secreting glands known as crypts

(48). At low magni

fi

cation (5

), imaging showed that bacteria in

the cecal mucosa formed colonies that were associated with one

or multiple crypts (Fig. 4

–

). However, the colonization of

crypts was not homogeneous across the tissue. Colonized crypts

were spatially clustered and surrounded by crypts with few or no

bacteria. In contrast, mucus was somewhat evenly distributed

across crypts. The 3D imaging at higher magni

fi

cation (20

)

con

fi

rmed that not all crypts were occupied by bacterial colonies,

but that all crypts secreted mucus (Fig. 4

–

Quantification of the Composition and Spatial Structure of

the Microbiota of Crypts.

As shown in our 3D imaging of the

mucosa (Figs. 3 and 4), bacteria occupy habitats with different

epithelium

bacteria

plastic

tissue

stomach

colon

cecum

Longitudinal dissection

and cleaning

plastic

glass

Gelling of the surface

acrylamide layer

Bonding of tissue

to a rigid substrate

gel

Tissue clearing

Tagging of

bacteria and

the host

outer muscle

Tissue before clearing

Tissue after 4 days of clearing

mucus

epithelium

bacteria

Bacteria

Epithelium

2.5 mm

100

Fig. 1.

Sample preparation and imaging for 3D mapping of the mucosal microbiota

’

sspatialstructure.(

)Thework

fl

ow of the method has

fi

ve key steps in which

a section of intestinal tissue is prepared for whole-mount confocal imaging of the mucosal microbiota. (

) A sample of preserved murine cecal tissue before and

after 4 d of lipid removal. The dimensions and shape of the sample were not visibly altered by clearing. (Scale bars: 1 cm.) (

) Maximum intensity projection of a

tiled image of a typical intestinal tissue sample after the method. The image of the cecum was obtained by stitching multiple

fi

elds of view acquired with a 5

objec-

tive that is not

fl

at-

fi

eld corrected. Bacteria were stained by HCR with a eubacterial detection probe, and host nuclei were stained with DAPI. (

) The 3D rendering

of the confocal imaging of the area enclosed in the dashed white square in

shows the location of bacteria with respect to each other and the host.

PNAS

2022 Vol. 119 No. 18 e2118483119

https://doi.org/10.1073/pnas.2118483119

3of12

Downloaded from https://www.pnas.org by George Porter on April 28, 2022 from IP address 131.215.225.183.

geometries along the mouse GIT. In the proximal colon, bacte-

ria accumulated in a layer that ran parallel to the epithelium,

whereas, in the cecum, bacteria were split into colonies that

were associated with crypts. The microbiota of the cecal

mucosa and of intestinal crypts are diverse (30). However, the

spatial structure of these communities remains unexplored.

To explore the spatial order in the microbiota of cecal crypts,

we extended our imaging method to enable multiplexed imag-

ing of bacterial targets (

Materials and Methods

and

SI Appendix

Supplementary Materials and Methods

). First, to identify the

taxa we should target for imaging, we sequenced the 16S rRNA

gene of the microbiota of the cecum (Fig. 5

), and searched

the literature for FISH probes that could speci

fi

cally detect bac-

teria belonging to the

fi

ve taxonomic groups that comprised

∼

76% of the sequenced reads: Bacteroidetes, Lactobacillaeae,

Ruminoccocaceae, Lachnospiraceae, and Verrucomicrobiaceae

(Fig. 5

). We tested, in vitro, the sensitivity and speci

fi

city of

the selected detection sequences in HCR (

Materials and Meth-

ods

and

SI Appendix

Supplementary Materials and Methods

and

Figs. S3 and S4

). We performed HCR with taxon-speci

fi

probes, targeting four species of bacteria that were representa-

tive of the target taxonomic groups. We used an additional

probe for

E. coli

because it was not found in the sequencing of

the cecal mucosa and thus served as a further control for the

speci

fi

city of our probes (

SI Appendix

, Table S1

). Finally, HCR

probes for multiplex in situ imaging were designed by pairing a

unique HCR hairpin pair to each detection sequence that

detected at least 84% of its ideal target bacterium while being

insensitive to the rest of the bacterial targets, with the exception

of the detection sequence cfb560 that cross-reacts with 0.3% of

E. coli

targets (Fig. 5

). The promiscuous lab158 probe was

rejected in favor of the orthogonal lgc354 suite.

Because we had observed that cecal crypts are colonized in patches

(Fig. 4

), we performed multiplexed HCR on several cecum sam-

ples and imaged the most abundant target taxon (Bacteroidetes) at

low resolution (5

)(notshown)tolocatepatches.

Within one patch of crypts, we obtained spectral imaging at

higher magni

fi

cation (20

), which was processed computation-

ally to remove the

fl

uorescent spectral overlap (

SI Appendix

Supplementary Materials and Methods

). The 3D spectral imag-

ing with linear deconvolution of the cecal mucosa clearly

showed multispecies colonization (Fig. 5

and

Movie S3

), and

distinguished the location of different taxa in dense cryptal col-

onies (Fig. 5

–

). We analyzed the taxonomic composition of

a subset of 57 abundantly colonized crypts using commercial

3D image analysis software (

Materials and Methods

and

Movie

). We measured the abundance (number of voxels) and the

position of the target taxa inside crypts. Accordingly, the crypt

microbiota was 65% Bacteroidetes, 18% Lachnospiraceae, 13%

Ruminococcaceae, and 3% Bacilli, with an insigni

fi

cant pro-

portion of

Akkermansia

. Also, in this small set of crypts, the

taxa were arranged in different depths within each crypt, with

Bacilli, Lachnospiraceae, and Ruminococcaceae found closer to

the luminal end (Fig. 5

–

Distance From Surface of Gel (

[ lysozyme ] = 1 mg/mL

DAPI

HCR

Distance (

100

150

200

150

300

450

600

Mucosa

[ lysozyme ] = 5 mg/mL

DAPI

HCR

SPF tissue + unspecific probe

GF tissue + eubacterial probe

SPF tissue + eubacterial probe

0 1020304050607080

HCR signal / Background signal

0.1

0.3

0.5

0.7

0.9

Cumulative Proportion of Bacteria

Background signal

from no-target control (D)

Background signal

from unspecific

probe control (E)

HCR signal / Background signal

Cumulative Proportion of Bacteria

1 mg/mL

2 mg/mL

5 mg/mL

100

150

200

250

300

0.2

0.4

0.6

0.8

< 300

> 300

Fig. 2.

Sensitivity and speci

fi

city of

fl

uorescence imaging of bacteria embedded in acrylamide gels using dual embedding. (

) Maximum intensity projection

of a digital cross-section (152

m) of intestinal tissue. The thickness of the protective acrylamide gel layer was revealed by blue-

fl

uorescent beads on its sur-

face. The layer of gel was a diffusive barrier for lysozyme during HCR staining of bacteria. (

) Maximum intensity projections of digital cross-sections (50

of gel slabs seeded with bacteria. The effect of lysozyme concentration on the sensitivity of HCR staining is illustrated. At a suboptimal concentrat

ion of lyso-

zyme (1 mg/mL), only bacteria near the surface of the gel could be detected, whereas a concentration of lysozyme of 5 mg/mL enabled the detection of bac-

teria throughout the gel. (

) Experimental cumulative distributions of HCR staining of bacteria embedded in gel slabs that were treated with different lyso-

zyme concentrations. At a lysozyme concentration of 5 mg/mL,

∼

94% of bacteria within 600

m of the surface had an HCR signal-to-background signal ratio

≥

20 (vertical dashed line). (

–

) Maximal intensity projections of representative luminal views of proximal colon tissue that was used to test the speci

fi

city

of HCR staining of bacteria in situ. (

) HCR with a eubacterial detection sequence (eub338) on GF tissue, (

) HCR with a nonspeci

fi

c control probe (non338)

on tissue from mice with a microbiota (SPF), and (

) HCR with a eubacterial detection sequence (eub338) on tissue from mice with a microbiota. (Scale bars:

100

m.) (

) Experimental cumulative distribution of the HCR signal-to-background signal ratio from controls for in situ HCR staining of bacteria in

–

. Three

fi

elds (

3) of view from each sample (

–

) were acquired. The average intensity of the background signal was calculated from the controls with no target

and a nonspeci

fi

c probe. In

, bacteria were segmented with an intensity

fi

lter to obtain their average HCR

fl

uorescence.

4of12

https://doi.org/10.1073/pnas.2118483119

pnas.org

Downloaded from https://www.pnas.org by George Porter on April 28, 2022 from IP address 131.215.225.183.

The Biogeography of the Mucosal Microbiota Is Robust to

Taxonomic Changes Driven by Ciprofloxacin.

To investigate the

robustness of clustered crypt colonization and the particular role

of Muribaculaceae (formerly S24-7 family) on the colonization of

crypts, we used the broad-spectrum antibiotic cipro

fl

oxacin.

Recently, it was shown that Muribaculaceae can become unde-

tectable in the feces of conventional mice up to 10 d after stop-

ping a 5-d treatment with cipro

fl

oxacin, whereas other taxa seem

to recover to their initial numbers earlier (49). Accordingly, we

administered 4 mg of cipro

fl

oxacin twice per day for 4 d and

allowed the microbiota to recover for 10 d (Fig. 6

The effect of cipro

fl

oxacin on the composition and spatial

organization of the mucosal microbiota was investigated

through quantitative sequencing of the 16S rRNA gene of the

fecal microbiota (Fig. 6

and

) and 3D imaging (Fig. 6

–

) of the cecal crypts of two sets of mice. Mice of cohorts

A and B were of the same age and from the same room at the

facility of origin, and they had a similar fecal microbiota com-

position before the administration of cipro

fl

oxacin (Fig. 6

and

Materials and Methods

). After 4 d of twice-daily administration

of oral cipro

fl

oxacin, we imaged the microbiota of the cecal

mucosa of three mice from three separate cages (cohort A). The

remaining mice from each cage were given a 10-d-long postan-

tibiotic recovery period. After this period, we imaged the

microbiota of the cecal mucosa of another set of three mice

from three separate cages (cohort A). To control for the effects

of antibiotic, we imaged the microbiota of the cecal mucosa of

two control mice that were not exposed to the drug (cohort B).

We quanti

fi

ed the absolute total abundance of bacteria in

feces through qPCR and then quanti

fi

ed the absolute abundance

of individual bacterial families by multiplying the absolute total

abundance by the proportions obtained from the sequencing of

16S rRNA gene (50) (

Materials and Methods

and

SI Appendix

Tables S3 and S4

)(Fig.6

). The median of the total bacterial

load in feces was reduced by cipro

fl

oxacin by more than three

orders of magnitude among the three cages of cohort A (

(Fig. 6

, Total, blue). Ten days after discontinuing the antibi-

otic, the average bacterial load in feces recovered to the same

order of magnitude as the preantibiotic abundance (

(Fig. 6

,Total,green).

= 22 ± 9.7

= 125 ± 53

H - I

Overlay of C and D

Bacteria

Epithelium

Host mucus

Code of colors for all panels in this figure

Lumen

Fig. 3.

Spatial structure of the host

–

microbiota

interface of the murine proximal colon and distal

ileum after being processed with the method

presented here (Fig. 1). (

) The 3D rendering of

confocal imaging (20

) of the crest of a fold in

the proximal colon. The epithelium (blue) was

covered by a mix of mucus (green) and bacteria

(orange). (

–

) Maximum intensity projection of

the digital cross-section (7

m) depicted in

Mucus and bacteria were organized in well-

fi

ned layers. Two layers of mucus separated

most of bacteria from the mucosa and from the

luminal contents (removed from this area of the

sample). The thin layer of mucus that separated

the epithelium from the majority of the micro-

biota in the lumen could be crossed by bacteria

in healthy tissue. (

) Zoom-in view from

.Ovals

are examples of bacteria inside and across the

thin mucus layer that lines the epithelium.

(

) Maximum intensity projection of a digital

cross-section (7

m) from the same sample as in

. Inside the oval is another example of bacteria

crossing the thin mucus layer and the epithe-

lium. (

) The 3D rendering of confocal imaging

(20

) of villi of the small intestine covered with

mucus and bacteria. (

and

) Maximum intensity

projections of the digital cross-section (16

depicted in

. Bacteria accumulated on mucus

around the top of villi. (All scale bars: 100

m.)

PNAS

2022 Vol. 119 No. 18 e2118483119

https://doi.org/10.1073/pnas.2118483119

5of12

Downloaded from https://www.pnas.org by George Porter on April 28, 2022 from IP address 131.215.225.183.

We found that the change in the abundance of bacteria after

recovery from cipro

fl

oxacin was not uniform across taxa and the

cages of cohort A (Fig. 6

). As expected, the family Muribacula-

ceae was undetected in all recovered mice (

SI Appendix

Supplementary Materials and Methods

and Tables S3 and S4

Compared with the control mice not treated with cipro

fl

oxacin,

the absolute abundance of

Bacteroides,

Verrucomicrobia, Rumino-

coccaceae, Lactobacillaceae, and other families were not statisti-

cally signi

fi

cantly different after 10 d of recovery. The dominant

Bacteroides

species in two of the three cages after recovery was

Bac-

teroides thetaiotaomicron

,whereas,inthethirdcage,noBacteroi-

detes was detected (

SI Appendix

Supplementary Materials and

Methods

and Table S4

Four days of twice-daily cipro

fl

oxacin reduced the size of

bacterial colonies in crypts to zero (Fig. 6

), but, after 10 d

without antibiotics, bacterial colonies grew to a size comparable

to colonies that were unexposed to cipro

fl

oxacin (

SI Appendix

Supplementary Materials and Methods

and Figs. S10

–

S12

). To

quantify the recovery of the mucosal microbiota after cipro

fl

ox-

acin, we compared the spatial distribution of bacteria of crypts

in unexposed (no cipro

fl

oxacin) and recovered (10 d after with-

drawing cipro

fl

oxacin) mice.

At the level of single crypts, we compared the abundance of

bacteria (volume) and their proximity to the host (depth) in the

cryptsofcontrolmice(noantibiotic)andthecryptsofmicethat

had recovered from antibiotic for 10 d (

SI Appendix

,Figs.

S10

–

S12

).Thevolumeofbacterialcolonieswasmeasuredasthe

voxel count of the objects segmented in the eubacterial channel

(

Materials and Methods

and

SI Appendix

)(Fig.6

), and the

depth of bacteria was measured as the relative position of their

center of mass with respect to the luminal opening of crypts (Fig.

). We found that, although bacteria recolonized crypts after a

10-d recovery period, the volume and depth of crypt colonies

were both less than the volume and depth of crypt colonies in

themicethatwerenotexposedtocipro

fl

oxacin. The median vol-

ume of eubacteria in recolonized crypts was less than half of the

median volume of unexposed colonies (45%), and the median

position of eubacteria in unexposed crypts was 2.5

l

m deeper.

Images of the bacterial colonies of cecal crypts (

SI Appendix

,Figs.

S10

–

S12

) illustrate the varied colonization of crypts.

In unexposed and recovered mice, we measured the spatial

distribution of three taxonomic groups across and within single

crypts: Bacilli, Bacteroidetes, and Clostridiales. Within Clostri-

diales, we considered the closely related Lachnospiraceae and

F-H

Overlay of B, C, and D

Bacteria

Epithelium

Host mucus

Code of colors for all panels in this figure

Overlay of G and H, and epithellum

Lumen

Fig. 4.

Multiscale imaging showed that the

cecal mucosa was colonized in clusters. (

–

)

Tiled image of luminal imaging of a tissue

sample from the cecum. The image was

obtained by stitching multiple

fi

elds of view

acquired at 5

magni

fi

cation. Bacteria were

stained by HCR with a eubacterial detection

probe (orange), the DNA of host cells was

stained with DAPI (blue), and the host mucus

was stained with WGA lectin (green). The epi-

thelium of the cecum was lined with crypts,

some of which were isolated and some of

which were connected to other crypts by

crevices. The colonization of the mucosal

crypts was discontinuous. Clusters of colo-

nized crypts were separated by areas with

fewer bacteria. The spatial distribution of

mucus was more uniform. (Scale bar: 1 mm.)

(

) The 3D rendering of confocal imaging (20

)

of the cecal mucosa enclosed in the dashed

square area in

–

) Maximum intensity

projection of the digital cross-section (70

is indicated by a dashed line in

. Dashed lines

and

indicate the approximate location

of crypts in

. Bacteria that colonized the

cecum occupied the crypts and the mucus

these glands secrete. All crypts produced

mucus, but not all crypts were colonized by

bacteria. (Scale bar: 75

m.)

6of12

https://doi.org/10.1073/pnas.2118483119

pnas.org

Downloaded from https://www.pnas.org by George Porter on April 28, 2022 from IP address 131.215.225.183.