Mondragon‐Palomino et al. Supplement p. 1
Supplemental information for
3D imaging for the quantifi
cation of spatial patterns
in microbiota of the intestinal mucosa
Octavio Mondragón-Palomino
1,β
, Roberta Poceviciute
1
, Antti Lignell
1,2,α
, Jessica A. Griffiths
2
, Heli Takko
1
,
and Rustem F. Ismagilov
1,2
*
1
Division of Chemistry and Chemic
al Engineering, California Inst
itute of Technology
2
Division of Biology and Biological Engineering, California Inst
itute of Technology
1200 E. California Blvd., Pasadena
, CA, United States of Americ
a
* Correspondence to:
rustem.admin@caltech.edu
.
α
Current address: Department of C
hemistry, University of Helsin
ki, Helsinki, Finland
β
Current address: Laboratory of P
arasitic Diseases, National In
stitute of Allergy and Infect
ious Diseases, Bethesda,
MD, United States of America
Supplementary Materials and Methods
Composition of acrylamide monomer mix
Tissue preservation and clearing for imaging
HCR staining of bacterial 16S rRNA
Media for bacterial culture
Optimization of lysozyme treatment for HCR
Controls for in situ HCR
In vitro assay to find formamide concentrations for st
ringent hybridization of taxon-specific HCR probes, as
well as to quantify their sensitivity and specificity
Processing and analysis of in situ imaging
Supplementary Figures S1-S22
Supplementary Video captions S1-S4
Supplementary Tables S1-S2
Supplementary References
Contributions of non-corresponding authors
Mondragon‐Palomino et al. Supplement p. 2
Supplementary Materials and Methods
Composition of acrylamide monomer mix
The following reagents were used for the preservation of expose
d intestinal tissue: Acrylamide solution 40% in
water (#01697, Sigma Aldrich, Saint Louis, MO, USA), 2% Bis (#1
610142, Bio-Rad, Hercules, CA, USA),
Paraformaldehyde 32% (#15714-S, Electron Microscopy Sciences, B
risbane, CA, USA), Polymerization thermal
initiator VA044 (#
NC0632395,
Wako Chemicals, Richmond, VA, USA). The acrylamide monomer mix
for the
protective gel layer on the mucosa requires the crosslinker bis
-acrylamide to become rigid. The final
concentrations of reagents for the gel layer were: 4% Acrylamid
e, 0.08% Bis-acrylamide, 4% Paraformaldehyde,
2.5 mg/mL VA044, 1X PBS. The final concentrations of reagents f
or acrylamide embedding of the rest of the
sample were: 4% Acrylamide, 0.0% Bis-acrylamide, 4% Paraformald
ehyde, 2.5 mg/mL VA044, 1X PBS.
Tissue preservation and clearing for imaging
To prepare tissue samples for imaging of the mucosal microbiota
(Fig. 1-5), 4 mice of 20–21 weeks of age (strain
C57BL/6J) and 1 germ-free mouse of 11 weeks of age were euthani
zed and their GIT tissues harvested. Mice
for the antibiotic-challenge experiments were 14-15 weeks of ag
e (strain C57BL/6J) at day one of the experiment.
Each mouse received an intraperitoneal injection of 220 μL of a
10X dilution of the sedative Fatal-Plus (Vortech
Pharmaceuticals, Dearborn, MI, USA). Once a mouse was anestheti
zed, we performed transcardial perfusion
with sterile, ice-cold 1X PBS for 20 min at a rate of 4-5 mL/mi
n to euthanize the mouse and clear its vasculature
of blood. During perfusion, the exposed viscera were kept wet w
ith sterile 1X PBS, and covered with a small bag
of ice. After perfusion, the viscera were quickly removed and k
ept in a dry, sterile, tube in ice. In a biosafety
cabinet, the GIT was isolated from the mesentery, liver and att
ached fat. The jejunum and duodenum were also
removed and discarded. To preserve the external muscle layer of
the intestines in its distended form, we fixed
the remaining GIT (from ileum to rectum) for 3 min in ice-cold
4% paraformaldehyde (15714-S Paraformaldehyde
32%, Electron Microscopy Sciences, Hatfield, PA, USA) and then
washed it in ice-cold 1X PBS for 3 min to stop
fixation. After fixation, the distal colon and the ileum were r
emoved and discarded. The cecum and the proximal
colon were separated and kept in sterile containers on ice.
The cecum and the proximal colon were cut open longitudinally a
nd the bulk contents cleared with sterile
tweezers. The remainder of the GIT contents were removed by gen
tly dripping sterile, cold 1x PBS on the
exposed surfaces. Any intestinal contents that remained attache
d to the tissue surface after PBS treatment were
retained. The proximal colon was then cut into two segments. On
e contained all the folds and the other segment
was a transition from the cecum to the colon and contained no f
olds. The cecum tissue was split into four
segments: the end tip, the middle, the top left, and the top ri
ght. Each segment was placed into a pool of PBS
(0.5 mL) on a glass slide, which was contained by a silicon iso
lator (#666503; Grace Bio-Labs, Bend, OR, USA)
to keep the tissue in place and prevent tissue desiccation.
Next, in a chemical safety cabinet, tissue samples from each mo
use were put in a petri dish, placed in an ice
box, and fixed for 1 h by adding 1 mL of ice-cold 4% paraformal
dehyde (PFA) to each PBS pool. We replaced
with fresh 4% PFA every 15 min. For the tissues used in the ant
ibiotic-challenge experiments, we improved this
step by closing the chamber with a glass slide and fixed tissue
s for 60 min without replenishing PFA. After
fixation, tissues were flipped over onto the pool of 4% PFA (so
that the muscle side was facing up). To increase
the volume of the pools in which the tissues were submerged, we
stacked onto each slide an additional two
silicon isolators. We added more 4% PFA, covered each pool with
a silicon membrane to avoid evaporation,
placed them in Petri dishes and transferred the slides into an
ice box. The ice box was then placed in an
anaerobic chamber along with the bis-acrylamide monomer mix (Su
pplementary Materials and Methods). In the
anaerobic chamber, we removed the 4% PFA in which the tissues w
ere floating and substituted it with 2 mL of
the monomer mix. The tissues were left in the monomer mix on ic
e for about 15 min so that the components of
the mix could penetrate the bacterial biofilms and other conten
ts on the tissues. The monomer mix was removed
using a pipette and substituted with 1 mL of fresh mix. Finally
, we removed 900 μL. We covered the pools with
Mondragon‐Palomino et al. Supplement p. 3
plastic membranes (#664475; Grace Bio-Labs), and added a Kimwip
e imbibed with PBS to maintain the humidity
in the petri dish. We sealed each petri dish with parafilm and
put all the Petri dishes into an incubator set to 37
°C in the anaerobic chamber for 3 h to allow the acrylamide lay
er to form at the glass-tissue interface. We
removed the Petri dishes from the incubator and the anaerobic c
hamber and added a few droplets of 1X PBS
onto each tissue to keep them humid. The Petri dishes were refr
igerated (4 °C) overnight. The next day, the Petri
dishes containing the tissue samples were put in a box with ice
and brought back into the anaerobic chamber,
where each tissue was embedded in an acrylamide matrix without
bisacrylamide. This step is necessary to turn
the tissue into a hydrogel. Embedding lasted 3 h, after which t
he excess acrylamide mix was removed and the
tissue was polymerized for 3 h at 37 °C. The tissues were taken
out of the incubator and the anaerobic chamber,
and stored at 4 °C with a few droplets of sterile PBS.
Tissue samples were removed from the glass slides with a steril
e razor-blade and glued (Gluture 503763; World
Precision Instruments, Sarasota, FL, USA) onto a piece of semi-
rigid plastic (Polypropylene film 160364-46510;
Crawford Industries, Crawfordsville, IN, USA) that was previous
ly cleaned of RNAse (RNaseZap, AM9780;
ThermoFisher Scientific), sterilized with 70% ethanol, and trea
ted with oxygen plasma for 3 min to enhance
adherence.
After samples were turned into a hydrogel and before they were
passively cleared, we permeabilized bacteria
according to the parameters prescribed by the optimization of l
ysozyme treatment (Supplementary Materials and
Methods). Samples were pre-incubated in 10 mM Tris-HCl (#AM9856
, Invitrogen, Carlsbad, CA, USA) for 1 h at
room temperature, then treated with lysozyme at a concentration
of 5 mg/mL in 10 mM Tris-HCl, pH = 7.9, at 37
°C for 7 h for thin samples without much materials left on thei
r surface, and for 13h for samples with abundant
contents left on their surface. Lysozyme treatment was stopped
by washing excess enzyme overnight in 1X PBS
at room temperature in gentle shaking. Permeabilized samples we
re enclosed in tissue cassettes and cleared
for 4 d in 8% w/v sodium dodecyl sulfate (SDS) in PBS, pH = 8.3
at 37 °C. SDS was vigorously stirred. pH was
adjusted daily. SDS was removed by washing in stirred 1X PBS fo
r 2 d at 25 °C. Total DNA was stained with
DAPI (3 μg/mL in PBS) for 1 d. Host mucus was stained by submer
ging samples in a solution of WGA in 1XPBS
at a concentration of 50 μg/mL.
Distal ileum samples for imaging (Fig. 3) were obtained from on
e 9-month-old C57BL/6J male mouse and were
processed similarly to tissues from the cecum and proximal colo
n.
HCR staining of bacterial 16S rRNA
To fluorescently tag 16S rRNA transcripts from mucosal bacteria
, we incorporated HCR labeling of RNA to the
workflow. HCR is executed in two stages: detection and amplific
ation. In the detection stage, one or multiple
HCR probes hybridize to homologous RNA transcripts. In the ampl
ification stage, a unique initiator sequence
encoded in each probe selectively hybridizes to matching DNA ha
irpin pairs. The HCR seeded by the initiator
sequence concatenates the matching fluorescently labeled hairpi
ns into a long double-stranded DNA molecule.
Independent probes and orthogonal hairpins enable multiplexed f
luorescent labeling of RNA markers.
We designed HCR probes (Supplementary Tables S1-S2) and used th
em to image the location of total bacteria
and specific taxa on intestinal tissue. We used the eubacterial
probe eub338-B4 and B4-Cy3B hairpins. Samples
were pre-hybridized in a solution of 2xSSC (#V4261, saline sodi
um citrate, Promega Corp., WI, USA) and 10%
dextran sulfate sodium (#D8906, Sigma, MO, USA) for 1 h at room
temperature. Next, we treated samples for
16 h at 46 °C in a buffer consisting of 2xSSC, 10 %w/v dextran
sulfate sodium, 15% formamide (#BP227100,
Fisher Scientific, NH, USA) and probe eub338-B4 at a final conc
entration of 10 nM. The unbound probe was
washed off for 1 h in a solution of 2xSSCT (2xSSC, 0.05 % Tween
20) and 30% formamide, followed by another
wash in 2xSSCT for 1 h. Samples were pre-amplified with in a bu
ffer of 5xSSC and 10 % w/v dextran sulfate for
1 h. The pre-amplification time was used to aliquot, heat-shock
(90 s at 95 °C), and cool down hairpins to room
temperature (at least 30 min in the dark). The amplification st
ep was carried out in a buffer that consists of a
solution with 5xSSC, 10 % w/v of dextran sulfate and a final co
ncentration of 120 nM of each hairpin (B4-Cy3B-
H1, B4-Cy3B-H2 for Fig. 1-5, and B1-A514-H1, B1-A514-H2 for Fig
. 6-7). The amplification reaction lasted 16-
Mondragon‐Palomino et al. Supplement p. 4
20 h at room temperature in the dark. Hairpins that did not par
ticipate in the reaction were washed out in 2xSSCT
for at least 1 h at room temperature with gentle shaking.
Multiplexed fluorescent labeling of 16S rRNA transcripts of muc
osal bacteria by HCR was executed analogously
to the monochromatic staining. However, because taxon-specific
probes have different melting temperatures,
hybridization reactions were executed over 3 d, starting with t
he probes that required the highest formamide
concentration and finishing with the probes that required the l
east formamide. Because the detection sequence
cfb560 is degenerate and usually produces a low signal, we only
considered the sequences (cfb560a, cfb560b)
that target bacteria we found through sequencing. Hybridization
reactions for multiplexed imaging (Fig. 5) were
as follows: muc1437-B4 and the suite lgc354a-b-c-B5 at 10% form
amide, clept1240-B3 at5% formamide, and
cfb560a-B2, cfb560b-B2 and lac435-B1 at 0% formamide. Similarly
, for cecal tissues of mice unexposed to and
recovered from ciprofloxacin, formamide concentrations were: lg
c354a-b-c-B5, eub338-B1, and clept1240-B3 at
10% formamide, lac435-B3 and muc1437-B4 at5% formamide, and cfb
560a-B2 and cfb560b-B2 at 0%
formamide. Samples were “pre-hybridized” in a solution of 2xSS
C (#V4261, saline sodium citrate, Promega
Corporation) and 10% dextran sulfate sodium (#D8906, Sigma) for
1 h at room temperature. Next, samples were
treated for 16 h at 46 °C in a buffer consisting of 2xSSC, 10 %
w/v dextran sulfate sodium, formamide at the
specified concentration (#BP227100, Fisher Scientific) and each
probe at a final concentration of 10 nM.
Unbound probes were washed off for 1 h in a solution of 2xSSCT
(SSC with % Tween 20) and 30% formamide,
followed by another wash in 2xSSCT for 1h (or 3 h for the last
wash on day 3). Probe clept1240-B3 was used at
a 2X concentration because it has one degenerate base. After al
l probes were hybridized to samples, the
amplification stage was carried out in a single step. HCR hairp
in pairs were assigned to fluorophores as follows:
B1-A514, B2-A647, B3-A594, B4-Cy3B, B5-A488. The HCR hairpin pa
irs for the ciprofloxacin challenge
experiments were as follows: B1-A514, B2-A546, B3-A594, B4-A647
, B5-A488. The amplification buffer consists
of a solution of 5xSSC, 10% of dextran sulfate and 120 nM of ea
ch hairpin. The amplification reaction was run
for 20 h at room temperature in the dark. Hairpins that did not
participate in the reaction were washed out in
2xSSCT for 4 h at room temperature with gentle shaking.
HCR probes were ordered as individual 250 nmol scale DNA oligos
purified by standard desalting (Integrated
DNA Technologies, IA, USA). Hairpins were ordered from Molecula
r Instruments, a Caltech facility within the
Beckman Institute. All the solutions were made with DNase/RNase
-free distilled water (#10977023, Invitrogen).
Media for bacterial culture
The following media were used to culture bacteria for
in vitro
assays.
Escherichia coli
was cultured in LB media
(LB Broth, #240230, Difco, Becton, Dickinson and Company, NJ, U
SA) and LB agar.
Clostridium scindens
was
cultured in a mix of 50% Shaedler media (#cm0497, Oxoid, Thermo
Fisher, Waltham, MA, USA) and 50% MRS
media (Lactobacilli MRS Broth, #288130, Difco), and in Schaedle
r agar.
Lactobacillus AN10
was cultured in a
mix of 50% Shaedler media and 50% MRS media, and in MRS agar.
Bacteroides fragilis
and
Faecalibacterium
prausnitzii
were cultured in LYBHI
11
media (brain-heart infusion medium supplemented with 0.5% yeas
t extract,
Difco, Detroit, USA), and in LYBHI agar.
Akkermansia muciniphila
was cultured in LYBHI media supplemented
with hog-mucus.
Harvesting of tissues for sequencing
Four 4-month-old adult male and female specific-pathogen-free (
SPF) mice were euthanized by CO
2
inhalation
according to approved IACUC protocol #1646 and #1769 and follow
ing all guidelines and standard operating
procedures (SOPs) of the Caltech Institutional Animal Care and
Use Committee (IACUC). The gastrointestinal
tract, from the stomach to the rectum, was dissected and stored
in a sterile container on ice. The cecum of mice
was cut open with sterile instruments on an ice-cold sterile su
rface inside a biosafety cabinet. The bulk of cecal
contents was removed with sterile tweezers, stored in sterile t
ubes, and kept at -20 °C. The cecal tissue was
Mondragon‐Palomino et al. Supplement p. 5
kept flat on a cold and sterile surface while it was cleaned wi
th ice-cold and sterile 1X phosphate-buffered saline.
PBS 1X was obtained from a 10X dilution of phosphate buffered s
aline 10X (Corning, 46-013-CM) in ultra-pure
DNase/RNase-free distilled water (10977023; ThermoFisher, Walth
am, MA, USA). After removing contents from
the cecum, the cecal mucosa was harvested by scraping it with s
terilized microscopy glass plates. Samples were
stored in sterile tubes at -20 °C. Cecal contents and tissue sc
rapings were sent to Zymo Research (Irvine, CA,
USA) for 16S rRNA gene sequencing and bioinformatics analyses.
For the antibiotic challenge experiment, fecal pellets for sequ
encing of bacterial 16S rRNA genes were collected
from all mice at days 0 (no antibiotic), 4 (end of antibiotic a
dministration) and 14 (end of 10-day recovery). For
each time point, the samples of two mice per cage were weighted
and processed as described elsewhere (Said’s
paper) to obtain DNA extracts, which were sent to Zymo Research
(Irvine, CA, USA) for 16S rRNA gene
sequencing and bioinformatics analyses.
Optimization of lysozyme treatment for HCR
A. Preparation of acrylamide gels pads with embedded bacteria
Bacteria were cultured to exponential phase at 37 °C in anaerob
ic conditions. From this culture, a dense (10
9
–
10
10
cells/mL) suspension of cells was prepared in PBS. This suspen
sion was spiked into the monomer mix with
bisacrylamide (see
Composition of acrylamide monomer mix
) to the final cell density of ~ 5 x 10
7
cells/mL. The
mix of monomer and cells sat on ice for 15 min before being dis
pensed into pools made of silicone isolators (13
mm diameter x 0.8 mm depth; #666507; Grace BioLabs) glued to mi
croscope slides. We pipetted 106 μL into
each pool, and polymerized at 37 °C for 3 h in anaerobic condit
ions. The next day, the original silicone isolators
were replaced with larger ones (20 mm diameter x 2.6 mm depth;
#666304; Grace-Bio Labs). The new pools
with the gels were filled with a monomer mix with no bisacrylam
ide (see
Composition of acrylamide monomer
)
and incubated on ice for 3 h in anaerobic conditions. Next, the
monomer mix was removed and the gels were
polymerized at 37 °C for 3 h in anaerobic conditions. Bacteria
in the gel pads were predigested in lysozyme
buffer (10 mM Tris, pH=8.0) at room temperature for 1 h, and th
en digested with lysozyme (1, 2.5 or 5 mg/mL
lysozyme in 10 mM Tris, pH = 8.0) at 37 °C for 6 h. Lysozyme wa
s washed away with PBS at room temperature
overnight. The gel pads were cleared with 8% SDS in 1xPBS, pH =
8.3, at 37 °C for 2 d following a 1x PBS wash
at 25 °C for another 2 d. Bacteria were hybridized with a eubac
terial HCR probe (eub338-B5) in hybridization
buffer (Materials and Methods) with 15% formamide and amplified
for 16 h (Materials and Methods). Finally,
DNA was stained with DAPI (5 μg/mL) overnight.
B. Imaging of gel-embedded bacteria
Gel pads were mounted in 1x PBS and imaged with an upright lase
r-scanning confocal microscope (LSM880,
Carl Zeiss AG, Germany) using a long-working-distance water-imm
ersion objective (W Plan Apochromat 20X/1.0
DIC Korr UV Vis IR, #421452-9700; Carl Zeiss AG). Fluorophores
were excited using two lasers with λ = 488
nm and λ = 405 nm. Imaging settings were the same across all ge
l pads. Images were processed in commercial
software for 3D image analysis (Imaris, Bitplane AG, Switzerlan
d). Cell surfaces were identified by their
fluorescence in the 405 nm channel. Finally, for the identified
cells, mean cell fluorescence intensity in the 488
nm channel was computed.
C. Results
The Gram-positive bacterium
Clostridium scindens
was efficiently permeabilized in gel pads subjected to the
treatment of 5 mg/mL of lysozyme (Fig. 2b-2c). To assess whethe
r lysozyme treatment may affect HCR staining
of Gram-negative bacteria, a lysozyme treatment optimization ex
periment was carried out using the model Gram-
negative bacterium
Bacteroides fragilis
. Exponential phase
B. fragili
s cells were embedded into acrylamide gel
pads, treated for 6 h with four concentrations of lysozyme (no
lysozyme control, 1.0 mg/mL, 2.5 mg/mL, and 5.0
mg/mL) and cleared with 8% SDS for 2 d. 16S rRNA was stained by
HCR using universal detection probe
eub338, and DNA was stained with DAPI. Bacteria in gel pads we
re imaged in a confocal microscope (LSM
880, Carl Zeiss AG) from the surface of the gel down to 600 μm
into each gel. Confocal images and the results
Mondragon‐Palomino et al. Supplement p. 6
from image analysis are shown in Supplementary Fig. S1. A 3D re
ndering of confocal images in the DAPI channel
(Supplementary Fig. S1a, S1d, S1g and S1j) shows that DAPI stai
ning did not require permeabilization of the
peptidoglycan layer, justifying the choice of using the DAPI ch
annel to define the surface of bacterial cells. When
lysozyme treatment was omitted,
B. fragilis
cells at the surface of the gel were stained poorly by HCR
(Supplementary Fig. S1b). Image analysis was in agreement with
these visual inspections; ECDF curves shifted
progressively to higher Signal/Background with depth (Supplemen
tary Fig. S1c). Across all 100-μm thick slices,
a substantial fraction of cells (>20%) were fainter than the se
t background value (Supplementary Fig. S1c). The
lowest lysozyme concentration (1 mg/mL) was sufficient to impro
ve HCR staining of
B. fragilis
(Supplementary
Fig. S1e-f). Although cells at the surface of the gel pad appea
red brighter, >99% of all cells across the entire 600
μm were brighter than the set background value (Supplementary F
ig. S1f). Lysozyme concentrations 2.5 and
5.0 mg/mL did not deteriorate HCR staining (Supplementary Fig.
S1h-i and S1k-l). These results showed that a
treatment of 5 mg/mL of lysozyme for 6 h pearmeabilized the pep
tidoglycan layer of Gram-positive and Gram-
negative bacterial cells; thus, we used this treatment as a ref
erence for
in situ
experiments.
In vitro
assays to find formamide concentra
tions for stringent hybridization of
taxon-specific HCR probes, as well
as to quantify their sensitivity and specificity
We created an
in vitro
assay to determine the adequate formamide concentration for th
e hybridization of each
HCR probe to its ideal target (Supplementary Table S1 and Suppl
ementary Fig. S3a), as well as to test the
probes’ sensitivity and specificity (Supplementary Fig. S3b). T
he assay consists of regularly spaced shallow
acrylamide gels on a glass slide. Bacteria are embedded in the
gels, which are then surrounded by individual
silicone wells.
A. Preparation of bacteria embedded in shallow acrylamide gels
Bacteria were grown anaerobically at 37 °C to OD600 0.2–0.24 fr
om overnight cultures (see
Media for bacterial
culture
). Cultures were pelleted and resuspended in a preparation of g
el mix with bisacrylamide (see
Composition of acrylamide monomer mix
). In anaerobic conditions, 3.8 μL of the acrylamide with bacte
ria were
pipetted into each well of a SecureSeal
imaging spacer (#470352; Grace BioLabs) that had been glued to
a clean
glass slide. Wells were sealed with a silicone membrane (#66447
5; Grace Bio-Labs). The slide was flipped
upside down for 5 min so that bacterial cells could settle on t
he surface, and then the slide was placed in a sealed
petri dish and placed in an anaerobic incubator for 2 h at 37 °
C. Once gels solidified, a silicone isolator (#665101;
Grace BioLabs) was added to each slide to create a pool around
each gel. Next, bacteria were treated with a
solution of 1 mg/mL lysozyme in 10 mM Tris balanced to pH = 8 f
or 2.5 h at 30 °C. Gels were washed twice with
1x PBS for 10 min and 30 min. In agreement with clearing method
s, the glass slide was submerged in a solution
of 4% SDS in 1x PBS at 37 °C for 2 h. The silicone wells were r
emoved and the SDS solution was gently rinsed
with 27 °C 1x PBS. Slides were further washed in 1xPBS for 10 m
in and overnight at room temperature. Slides
were dried out and another silicone isolator was applied around
gels. Probes were hybridized in 2x SSC (saline
sodium citrate) with 10% dextran sulfate, 0-60% formamide, and
final probe concentration of 10 nM. The
hybridization buffer was pipetted into the silicone isolator we
lls, covered with a hybridization film (#716024; Grace
BioLabs) and put in a sealed petri dish. Glass slides were incu
bated at 46 °C for 12 h. Unbound probes were
washed three times with 2x SSCT (2x SSC, 0.05% Tween 20), and 3
0% formamide for 10 min. Three additional
10 min long washes were done in a buffer of 2x SSCT.
In the amplification step, hairpins were heat-shocked at 95 °C
for 90 s and cooled down at room temperature for
30 min. Gels were covered with the amplification buffer of 2x S
SC, 10 % w/v dextran sulfate and hairpins to a
final concentration of 120 nM, and covered with a hybridization
film. The amplification reaction was carried out
at room temperature for 12 h. Unbound hairpins were washed out
in a solution of 2x SSCT three times for 10
min. Three additional 10-min-long washes were done in a buffer
of 2x SSC. Finally, bacterial DNA was stained
with DAPI for 1 h, followed by a 30 min wash in 1x PBS. Slides
were dried out and another imaging spacer was
applied around the gels (#654008; Grace BioLabs). Gels were mou
nted in 1x PBS and covered with a glass
coverslip. Bacteria on the upper surface of the gels were image
d with an oil-immersion objective (Plan
Mondragon‐Palomino et al. Supplement p. 7
Apochromat 63X/1.4 Oil DIC, #420782-9900-799; Carl Zeiss AG) in
an upright confocal microscope (LSM 880,
Carl Zeiss AG).
B. Formamide curves
We next established the range of formamide concentrations that
would yield stringent hybridization of taxon-
specific HCR probes. Two slides of gels were prepared for each
target bacterium (Supplementary Fig. S3a).
One slide was used to quantify the efficiency of hybridization
for concentrations of formamide in 15% steps from
0-60 %. The second slide was used to refine the coarse measurem
ents in 5% steps around the maximum of the
coarse curve. Each slide was prepared once. We obtained one st
ack of images from each gel. Stringent
hybridization was obtained around the concentration of formamid
e that produced the strongest average
fluorescence. To determine the optimal formamide concentration
for the hybridization of probe muc1437 for
A.
muciniphila
, we tested the concentrations 0%, 5%, and 10%. We found that 5
% formamide provided stronger
signal than 0% and 10%. As a result, we set the concentration o
f formamide at 5% in hybridization reactions for
the antibiotics resistance-recovery experiments, as opposed to
the 10% that was previously used.
C. Sensitivity and specificity of taxon-specific HCR probes
To quantify the sensitivity and specificity of taxon-specific H
CR probes, one multi-species glass slide was
prepared for each tested probe (Supplementary Fig. S3b). Hybrid
ization was carried out at a formamide
concentration within the intervals prescribed by the formamide
bar plots (gam42a: 5%, eco630: 10%, lac435:
0%, lgc354: 5%, cfb560: 0%, lab158: 10%, clept1240: 0%.)(Supple
mentary Fig. S3b). Each hybridization
experiment was carried out in one gel. One stack of images was
stained from each gel. We tested the specificity
of lac435 , lgc354 , cfb560 , lab158 , clept1240 against
A. muciniphila
, without any detectable overlap.
D. Image Processing
Images were analyzed using commercial software (Imaris, Bitplan
e, Belfast, UK). Image stacks were 3D-
rendered and surfaces were created over individual bacteria usi
ng the fluorescent signal from the DAPI stain.
Because bacterial DNA is found throughout the cell, surfaces de
rived from DAPI fluorescence encompass entire
cells. For each cellular volume, the software computed the aver
age fluorescent intensity for two channels: the
eubacterial channel (eub338-B5/Alexa488) and the channel for a
taxon-specific sequence (B4/Cy3B).
Formamide plots (Supplementary Fig. S4) were obtained by plotti
ng the mean and standard deviation of the
fluorescence intensity in the Cy3B channel for each concentrati
on of formamide.
To quantify the sensitivity and specificity of taxon-specific p
robes (Fig. 5b), images were processed as described
in the previous paragraph. For each probe we set a fluorescenc
e detection threshold such that 85% of the ideal
target bacterium was detected (e.g., the lac435 probe’s ideal t
arget was
C. scindens
). For each non-ideal target
(for example lac435 should not target
B. fragilis
although it may bind to a small number of cells), off -target
hybridization is quantified as the fraction of bacteria above t
he fluorescence detection threshold.
Acquisition, processing and analysis of
in situ
imaging
A. Objectives and laser wavelengths
For large-scale acquisition, we used either of two objectives:
Plan-Neofluar 5X/0,15 (#440320, Carl Zeiss AG),
or EC Plan-Neofluar 5X/0.16 (#420330-9901, Carl Zeiss AG). For
imaging at 20X magnification, we used one
CLARITY optimized objective with an adjustable correction colla
r for compensation of spherical aberrations: Clr
Plan-Neofluar 20x/1.0 Corr nd=1.45 M32 85mm (#421459-9970-000,
Carl Zeiss AG). Fluorophores were excited
with laser light of the following wavelengths: 405 nm, 488 nm,
561 nm, 633 nm. Spectral acquisition was used
only for imaging samples with multiplexed HCR.
B. Imaging of the host-microbiota interface in the proximal col
on
Mondragon‐Palomino et al. Supplement p. 8
In one sample of the proximal colon (Supplementary Fig. S5), fo
ur areas corresponding to the tops of intestinal
folds were imaged (Materials and Methods). The resulting image
stacks contained three channels: one for DNA
(DAPI), one for bacteria (HCR staining), and one for mucus (WGA
lectin). To quantify the thickness of the layers
of mucus at the top of intestinal folds, image stacks were 3D r
endered in a commercial software (Vision4D 3.0,
Arivis AG, Germany). Next, the maximum intensity projections of
two digital cross sections (7 μm), along and
across the longitudinal axis of the folds, were obtained. The t
hickness of the internal mucus layer was measured
(n = 85) from the edge of the epithelium to the edge of the int
ernal mucus layer. The thickness of the external
mucus layer was measured (n = 75) from the end of the internal
mucus layer to the bright edge of the external
mucus layer (Fig. 3b).
A second sample of the proximal colon from a different mouse wa
s imaged (Supplementary Fig. S9) to show the
procedure is repeatable and produces consistent results. As in
Fig. 3, imaging in 3D shows that bacteria colonize
profusely the outer mucus barrier between luminal contents and
the epithelium. Bacteria are mostly segregated
to the outer mucus layer, but manage to contact the epithelial
layer at points where the inner mucus layer is thin.
The outer mucus layer is interspersed by spherical objects that
are consistent with DAPI staining of mammalian
cells (cyan circles in panels e-g), and larger objects that we
hypothesize are food particles (cyan shade on top
of outer mucus-bacteria layer Fig. S9e-g).
C. Linear unmixing of spectral imaging
Computational linear unmixing of spectral imaging was performed
to determine the relative contribution from
each fluorophore for every pixel of in situ multiplexed imaging
of bacteria. Linear unmixing requires the emission
spectrum of every fluorophore that was employed in the staining
of samples including DAPI and the suite of
Alexa fluorophores of HCR. Spectra were acquired independently
but in similar optical conditions as described
in our in situ imaging of bacteria. We used E. coli bacteria em
bedded in thick acrylamide gels that were prepared
with the procedure described in “Optimization of lysozyme treat
ment for HCR” (Supplementary Materials and
Methods). Two gels of 13-mm diameter were split into six smalle
r gels that were taken through our standard
HCR protocol. Each gel with E. coli was hybridized with a diffe
rent probe. Each HCR probe consisted of the
eubacterial detection sequence eub338 and a different initiator
sequence. Each initiator sequence matched a
different hairpin/fluorophore set. Probes: eub338-B1 (A514), eu
b338-B2 (A647), eub338-B3 (A594), eub338-B4
(A546), eub338-B5 (A488) and eub338-B4 (Cy3B). Bacterial DNA wa
s not stained with DAPI. The emission
spectrum of DAPI was acquired directly from the tissue samples.
Gels were mounted in a RIMS solution with n
~ 1.46. Imaging of bacteria in gels was carried out using a las
er-scanning confocal microscope with parallel
spectral acquisition (LSM880, Carl Zeiss AG), and with the same
objective as imaging of tissue samples (Clr
Plan-Neofluar 20x/1.0 Corr nd=1.45 M32, Carl Zeiss AG). We extr
acted the spectral references from the imaging
of bacteria using commercial software (Zen 2.3 SP1, Carl Zeiss
AG). Finally, the spectral references and the
same software were used to perform linear unmixing of in situ i
mages.
D. Multiplexed imaging of cecal crypts
The size of cecal tissue samples was approximately ¼ of the tot
al size of the cecum. Because there are
thousands of crypts in a sample, it was not practical to image
them all at 20X magnification (20 crypts/field of
view, 425x425 μm
2
). Inst
ead, we examined samples thoroughly at 20X magnification and im
aged only the areas
where crypts were colonized, making the best effort to capture
most colonized crypts. Therefore, imaged crypts
were representative of the region of the cecum that we imaged a
nd we believe that ¼ of the cecum may be
representative of the rest of the organ, but that would need to
be further investigated. We imaged the top portion
of the cecum, which we have found to be consistently colonized
by bacteria in C57BL/J6 mice of ages 13-20
weeks. In the context of the antibiotic challenge experiments,
in the control group, we quantified 296 colonized
crypts (2 mice) and in the recovery group we quantified 199 col
onized crypts (3 mice). In unexposed mice, we
quantified 160 crypts (3 mice) after 4 days of ciprofloxacin. H
owever, the number of crypts that were visually
Mondragon‐Palomino et al. Supplement p. 9
examined during image acquisition is perhaps two orders of magn
itude larger. It is also important to mention that
we focused on a very specific form of colonization of the mucos
a, namely the colonization of individual crypts,
but the mucosa is colonized much more abundantly in the crevice
s that are formed when multiple crypts merge
at the luminal side (Fig. 1e).
Multiplexed confocal spectral images of cecal mucosa at 20X mag
nification were taken through linear unmixing
(Materials and Methods) and analyz
ed computationally to measure
the abundance and location of bacterial taxa
that were labelled by HCR. The resulting data files contained i
mage stacks with seven
channels. Five channels
corresponded to the probe/fluorophore pairs that were used in H
CR (lcg354/A488, lac435/A514, muc1437/Cy3B,
clept1240/A594 and cfb560/A647, or eub338/A514, lcg354/A488, la
c435/A594, muc1437/A647, clept1240/A594
and cfb560/A546), one channel corresponded to the fluorescent D
NA marker DAPI, and one channel stored
pixels that were not assigned to any of the other six channels
in linear unmixing and thus captured undefined
content. Image stacks were uploaded to commercial software Visi
on4D (Vision4D 3.0, Arivis AG) and saved in
the native
sis
format. Because tissue was sometimes very tilted with respect t
o the plane of imaging, image
stacks were rotated so that crypts were approximately aligned w
ith the spatial z axis. Rotated stacks were
cropped manually to remove areas without data.
For the initial small data set in SPF antibiotic-naïve mice, th
e spatial analysis included ~60 crypts from three
fields of view obtained from one
cluster of crypts in a sample
of the cecum. For the antibiotic
challengeexperiments, the spatial analysis included 296 coloniz
ed crypts from two untreated mice and 199
colonized crypts from three antibiotic-treated mice that were i
maged after 10 days of recovery. The internal
volume of crypts was segmented manually using the “Draw Objects
Tool.” The manual segmentation of crypts
was guided by the DAPI channel, which showed the location of nu
clei on the epithelial wall of crypts. To restrict
the analysis to bacteria inside crypts, we used the segmented i
nternal volumes of crypts as a mask on the
channels with HCR staining (i.e., the fluorescence intensity va
lue of voxels outside crypts was set to zero in the
five HCR channels). Next, bacterial channels were segmented wit
h an “Intensity Threshold” filter. In the output
of this operation, a bacterial cell or group of bacteria in eac
h channel (a segment) was defined as a set of
contiguous pixels with intensities that fell within a range (mi
nimum and maximum bounds, hereafter Min and
Max) where at least one pixel had an intensity equal to a core
value (required core intensity, RCI). For the initial
small data set in SPF antibiotic-naïve mice, segmentation param
eters (Min, Max and RCI) were estimated by
measuring the intensity of a subset of pixels in each channel t
hroughout every stack and defined RCI as the
mean of pixel intensities, and Min as the difference between th
e mean and the standard deviation of intensities.
Max was set equal to the maximum intensity of bacteria in the c
hannel. Next, we filtered out segments that were
<18 voxels. Channel cfb560/A647 required further manual curatio
n to remove segments that were not likely to
be bacteria due to their size and location. Finally, to determi
ne which bacterial segments were located within
each crypt we combined all bacteri
al segments into a single lis
t and used the “Segment Colocalization” operation.
Bacterial segments were considered the “Subjects,” and the manu
ally segmented crypts were used as
“References.” The “Colocalization Measure” required that “Subje
cts” (bacteria) were completely within the
“References” (crypts). The identities of bacterial segments and
their crypt-specific assignment were stored at the
end of the pipeline. The final result of the image-processing p
ipeline is shown in Supplementary Video S4. For
the antibiotic challenge experim
ents, segmentation parameters (
min, max and RCI) were set manually for each
image stack and stored in the segmentation pipelines saved with
in the corresponding analysis pipelines. In the
eubacterial channel, we filtered out segments that were <13 vox
els. A size filter was not applied to the rest of
the channels. Minimal manual curation of segmentation in the eu
bacterial channel was required in a reduced
number of crypts. These crypts typically had the lowest amounts
of bacteria and the poorest signal/noise ratio,
and were constrained to one sample. Objects that were segmented
in the taxon-specific channels were validated
as bacteria if they co-localized with an object in the eubacter
ial channel (we used the “Colocalization Measure”
that “Subjects” were partially within the eubacterial object “R
eferences”). The exception to this rule was the
channel (muc1437/A647) because the rRNA of
A. muciniphila
is poorly hybridized by the eubacterial probe
eub338. The eubacterial masking is illustrated in Supplementary
Fig. S6-8.
Mondragon‐Palomino et al. Supplement p. 10
The “Intensity Threshold” filter produced large bacterial segme
nts that spanned multiple fields of view. We
exported the identity (imaging channel), volume (voxel count),
center of mass (z coordinate in μm), and first/last
plane of all segments (bacteria and crypts) into a MATLAB-reada
ble file. From this output, we obtained the
abundance (voxel count) of bacteria for each crypt, as well as
the position (center of mass) of each bacterial
segment in the framework of the corresponding crypt. The spatia
l reference z = 0 μm in each crypt was set at
the luminal end of the crypt segment.
Tissues obtained at the end of the administration of ciprofloxa
cin (day 4) were stained with an eubacterial probe
and all crypts there seemed to be empty except for a few. Three
of these crypts were found in the sample from
the cage that was less harshly affected by ciprofloxacin (by th
e total load of bacteria in feces) (Fig S13a-b).
However, fluorescent signal was also observed underneath the ep
ithelium. To clarify this, we performed two-
color HCR tagging with the eubacterial eub338 probe in the 2 sa
mples from the other 2 cages (Fig. S13c-f). To
thoroughly quantify the abundance of bacteria in antibiotic tre
ated tissues, we picked randomly ~5 crypts per
field of view from all antibiotic-treated samples for a total o
f 160 crypts, and processed the imaging of the
eubacterial signal in the same way as we did for colonized cryp
ts in other tissues. The crypts with putative
bacteria (Fig. S13a-b) were included. The image processing pipe
line confirmed that most crypts were empty and
did not yield any segmented objects within most crypts, and in
the crypts where objects were found no overlap
between segmented objects in the eub338-A514 and eub338-A633 ch
annels was found. Therefore, we
concluded that the signal in the first sample (Fig. S13a-b) was
not of bacterial origin, but noise in the A633
channel and set the volume of bacteria in crypts as zero (Fig.
6d).
E. Controls for
in situ
HCR
We performed HCR
in situ
to quantify the intensity of fluorescence due to nonspecific d
etection and amplification
(Fig. 2d). We used an HCR probe with a nonspecific eubacterial
detection sequence (non338) (Supplementary
Table S1) on one tissue sample from the proximal colon of a mou
se with a microbiota (specific pathogen free,
SPF), and an HCR probe with a eubacterial detection sequence (e
ub338) on one tissue sample from the proximal
colon of a germ-free (GF) mous
e. The HCR reactions for these co
ntrol experiments followed the same steps as
the procedure to stain mucosal bacteria with a single eubacteri
al probe.
Mondragon‐Palomino et al. Supplement p. 11
Figure S1
.
Lysozyme treatment optimization using
Bacteroides fragilis
as a model Gram-negative bacterium.
Exponential-phase
B. fragilis
was embedded into four acrylamide gel pads and each pad was tr
eated with lysozyme at a different concentration (
a-c
: no-