Current techniques for studying gut microbiota are unable to answer some important microbiology questions, like how different bacteria grow and divide in the gut. We propose a method that integrates the use of sequential d-amino acid–based in vivo metabolic labeling with fluorescence in situ hybridization (FISH), for characterizing the growth and division patterns of gut bacteria. After sequentially administering two d-amino acid–based probes containing different fluorophores to mice by gavage, the resulting dual-labeled peptidoglycans provide temporal information on cell wall synthesis of gut bacteria. Following taxonomic identification with FISH probes, the growth and division patterns of the corresponding bacterial taxa, including species that cannot be cultured separately in vitro, are revealed. Our method offers a facile yet powerful tool for investigating the in vivo growth dynamics of the bacterial gut microbiota, which will advance our understanding of bacterial cytology and facilitate elucidation of the basic microbiology of this gut “dark matter.”
The biological diversity and complexity of the mammalian gut microbiota present formidable obstacles to the investigation and comprehension of these intimate microbial neighbors. To fully comprehend the physiological and pathological functions executed by gut microbiotas, it is critical to understand the basic microbiology of these microbes, such as how different bacteria grow and divide in the gut (1). However, even after nearly two decades of ever-increasing gut microbiota research, many of these questions remain unaddressed (2). The difficulty of separate culture of many gut bacterial species in vitro, which is partially why these microbes are often referred to as the “dark matter” in the gut (3), prevents researchers from further investigating these bacteria in the laboratory. After extensive efforts in optimizing culture conditions for different gut bacteria by microbiologists, gradually more species can be cultured individually in vitro (4). Nonetheless, for the bacteria that can be cultured and investigated in vitro, to what extent the microbial knowledge obtained from in vitro studies can be translated into in vivo situations remains debatable (1). Therefore, a method that can be used to directly probe and investigate the gut microbes in vivo is highly desirable (5).
One approach that could directly probe the indigenous activities of gut microbes uses a traditional isotope-based labeling strategy. N15-tagged amino acids were given to mice by intravenous injection, and the microbial N15 signals in the gut acquired by bacterial foraging on host proteins were then detected using nanoscale resolution secondary ion mass spectrometry (6). Consequently, the metabolic activities and host-protein usage preference of different gut bacteria could be assessed. Nonetheless, the data acquisition of this isotope labeling approach is restricted to the highly specialized mass spectrometry–based technique, and the knowledge obtained was limited to a specific metabolic pathway of the bacteria. A different chemical approach, fluorescent d-amino acid (FDAA)–based metabolic labeling, has been highly valuable in bacteriology studies, because of the labeling specificity [targeting bacterial peptidoglycan (PGN), thus no labeling on eukaryotes], speed (labeling within minutes under optimized conditions), and ease of use (reading fluorescent signals with routine analytical equipment) (7, 8). Recently, it has been demonstrated that FDAA probes could label mouse gut microbiota in vivo with high efficiency, and we further established a STAMP (sequential tagging with d-amino acid–based metabolic probes) strategy for recording the survival of transplanted microbiota in the receiving mouse using the labeling signals of the two FDAA probes (9, 10).
Here, to develop a method that can directly probe and visualize the growth and division modes of different gut bacteria in situ, we propose a STAMP-based and fluorescence in situ hybridization (FISH)–facilitated probing strategy. In this method, two FDAAs with different fluorophores were sequentially administered to mice by gavage to record in vivo bacterial growth and division processes during the labeling period. Consequently, identifying the labeled bacteria with FISH probes at different taxonomic levels (genus and species) allowed us to determine how different bacterial taxa, particularly those that cannot be cultured in vitro, grow and divide in the mammalian gut.
FDAA sequential labeling of mouse gut microbiota
DAAs are essential building blocks for bacterial PGN synthesis; the sidechain-functionalized DAAs (i.e., FDAA) are well-tolerated by the enzymes (d, d or l, d-transpeptidases) involved in cell wall construction (7). Chronological use of multiple FDAAs has provided valuable information of the temporal PGN synthesis in several model bacterial species in vitro (10). Here, to study the in vivo growth and division patterns of different bacterial taxa in gut microbiota, we sequentially applied two FDAA probes, TAMRA-amino-D-alanine (TADA) and Cy5-amino-D-alanine (Cy5ADA), containing TAMRA (tetramethylrhodamine) or Cy5 (Cyanine 5) on side chains, in two gavages to label the mouse gut microbiota (scheme shown in Fig. 1A). The probes were given at an interval of 3 hours, and the collected cecal microbiota showed strong two-color labeling (Fig. 1B) with high coverage (fig. S1) after 6 hours of the first gavage without apparent alterations of the microbiota composition (fig. S2).
The two-color fluorescence imaging showed a great morphological diversity of the gut microbes. Different distributions of the two colors among various bacteria revealed their distinct dividing patterns and growth rates. Because the second probe used in the sequential labeling was Cy5ADA, the PGN sites with more active constructions had stronger labeling of Cy5 (shown in red), and PGN synthesized earlier had more TAMRA signals (shown in green). Thus, the color distributions of red/green provided a chronological account of the PGN synthesis in each bacterium. Most of the bacteria grew in a dispersed model with relatively strong FDAA labeling throughout the cell (Fig. 1B) and divided in binary fission with one or more red-labeled septums in the middle of the bacteria (Fig. 1C, nos. 1 and 2). Some bacteria were only labeled with the first probe (Fig. 1C, no. 3), suggesting that these cells might have different growth rates during the two labeling steps. It is worth noting that many bacteria showed asymmetric labeling (Fig. 1C, nos. 4 and 5). Some only had one red pole, with two halves having very different labeling intensities (no. 4). One explanation is that these were the daughter cells from a binary fission, and the red pole was from the newly separated septum. It is also possible that some of these bacteria might grow in an asymmetric or polarized manner, like the zonal-apical growth observed in Agrobacterium tumefaciens (7). Moreover, it was also common to see the two daughter cells with different growth rates, where only one of the two cells showed a new septum (Fig. 1C, no. 5). Besides these long rod/spindle bacteria, PGN growth of small rod/coccus bacteria could also be readily observed (Fig. 1C, no. 6). Many of the classical growth and division modes of bacteria could be observed in the labeled gut microbiota, including diffuse synthesis of PGN (Fig. 2A), spiral synthesis (Fig. 2B), division dominated by septum synthesis (Fig. 2C), polar growth (Fig. 2D), and division by stalk/budding formation (Fig. 2E).
Identification of the bacterial growth patterns on the genus level
With the rich microbiological information obtained from the FDAA-labeled gut microbiota, taxonomically identifying individual bacterium in the view field became highly desirable. Toward this end, we resorted to FISH, a classical method for determining taxonomic information in complex bacterial samples. To facilitate the FISH probe selection and design, we first did 16S ribosomal DNA (rDNA) and metagenomic sequencing of the labeled microbiotas (combined cecal microbiotas of five mice) to identify the bacterial composition (fig. S3 and table S2). To have a reasonably high coverage of the microbiota, we first used 15 FISH probes to stain some of the most abundant genera (covering ~71% of classified bacteria in the microbiota; table S1), the sequences of which were either based on previous reports or designed in this study. The FDAA-labeled microbiota was split into dozens of small aliquots and labeled by different FISH probes separately. It is worth mentioning that the microbiotas collected from small intestines also showed clear two-color FDAA labeling (fig. S4). However, because the amount of bacteria was very small, we did not perform any FISH staining.
Of the 15 FISH probes covering bacteria from nine families, seven stained Gram-positive and eight stained Gram-negative bacteria. Representative images of the labeled bacteria in each genus from multiple FISH experiments were presented. As expected, because of their thinner PGN, Gram-negative bacteria (Fig. 3, A to F) showed weaker FDAA labeling than the Gram-positive bacteria (Fig. 3, G to L), and most of them were short rods (1 to 2 μm). Of note, we also identified two Gram-negative genera, Helicobacter and Desulfovibrio, that could not be labeled by FDAAs (fig. S5). Extended labeling time (two gavages of TADA at an interval of 6 hours) still did not lead to any FDAA labeling (fig. S6). These bacteria may either have very slow PGN synthesis rates or have lipopolysaccharides/capsules impermeable to FDAA probes. It is also possible that they have atypical PGN structures or transpeptidases that cannot tolerate the incorporation of FDAAs, which merits further studies to extend our understanding of the structure and synthesis of PGN in different bacteria.
Most of the labeled Gram-positive bacteria (Fig. 3, G to L) were spindle shaped and divided by binary fission, but the distributions of the two colors in each genus were quite different, suggesting distinct arrangements of divisome and/or elongasome in various genera (11). For example, the bacteria shown in Fig. 3 (G to J) all belonged to the family Lachnospiraceae, but the two-color images presented different patterns. In Lachnospiracea incertae sedis (LIS; Fig. 3G), FDAAs were labeled in a striped manner, but in Lachnoclostridium (Fig. 3H), Roseburia (Fig. 3I), and Marvinbryantia (Fig. 3J), the labeling was more diffuse. Moreover, we also identified some bacteria with polar growth, with one (Fig. 3F) or two poles (Fig. 3L) strongly labeled, a phenomenon that was previously only observed in some Alphaproteobacteria and Actinobacteria (12). The consistent bacterial labeling patterns and morphogenesis in each genus (fig. S7) also verified the specificities of the newly designed FISH probes. Of note, the specificities of several of the newly designed FISH probes could not be verified (listed in table S2). These probes either stained bacteria with inconsistent FDAA-labeling patterns or had a higher labeling ratio than the taxon’s relative abundance determined by 16S rDNA sequencing. FISH sequences with improved specificities or a more stringent staining protocol are warranted for these bacterial groups.
Identification of the bacterial growth patterns on the species level
Of the 15 genera examined, Clostridium is known for being highly polyphyletic (13). We observed different labeling patterns within this genus (Fig. 4A). On the basis of the metagenomic sequencing results of the microbiota (table S2), we selected three species in this genus and labeled them with corresponding FISH probes designed for each. Varied labeling patterns were observed (Fig. 4B), with asymmetrical growth noticed in two (KNHs209 and ASF502) of the three species. Because Clostridium sp. ASF502 had never been cultured individually in vitro, these data showcased the potential of using our method for studying unculturable bacterial species in the gut microbiota.
Encouraged by the labeling results of Clostridium, we then set out to examine more species in the gut microbiota. Nine were selected, including six species that have been cultured in vitro: Akkermansia muciniphila (14), Parabacteroides distasonis (15),
Alistipes putredinis (16), Lactobacillus johnsonii (17), Lactobacillus brevis (18), and segmented filamentous bacteria (SFB; Candidatus Savagella) (19); and three unculturable taxa: Oscillibacter sp. 1-3, Eubacterium sp. 14-2, and Dorea sp. 5-2. The growth patterns of A. muciniphila, P. distasonis, A. putredinis, Oscillibacter sp., Eubacterium sp., and Dorea sp. were readily identified by our method (Fig. 5, A to F). However, because of the difficulties of performing FISH with Lactobacillus (20), the required step of lysozyme digestion destroyed most of the FDAA labeling signals from the bacteria, leaving only labeled septums (fig. S8). An improved FISH protocol for Lactobacillus is required for more effective analysis of these important gut bacteria by our method.
As a heavily studied species with fundamental functions of degrading mucin, producing bioactive molecules and immune modulation (21), A. muciniphila cells were clearly observed dividing by binary fission (Fig. 5A). The imaged short rods of P. distasonis did not seem to be dividing (Fig. 5B). A. putredinis showed strong labeling at one pole, indicative of its polar growth (Fig. 5C). The three species that could not be cultured in vitro belonged to the order Clostridiales. Oscillibacter (Fig. 5D) and Eubacterium (Fig. 5E) might have diffuse synthesis of PGN (scheme shown in Fig. 2A), but Dorea seemed to have a more prominent septum (Fig. 5F), the synthesis of which might dominate cell division (scheme shown in Fig. 2C). Consistent modes of labeling and cellular morphologies in each species were observed in each species (fig. S9), supporting the specificities of these new FISH probes.
Revealing the in vivo growth of SFB
Another group of bacteria worth special note is SFB. As one of the most extensively studied bacteria in gut microbiota, SFB have been found to be critical in the induction of T helper cell 17, alongside many other immunity-related effects (22, 23). In the ~90-μm SFB shown here (Fig. 6), we could clearly see the FDAA-labeled segments differentiated at distinct stages (Fig. 6A), intrasegmental bodies (Fig. 6B), a needle-like holdfast (Fig. 6C), a triseptum at an asymmetric division location (Fig. 6C), and a symmetric division locus (Fig. 6D) (19). However, no obvious FDAA labeling of SFB intrasegmental bodies was noticed. Previously, it was reported that the PGN of Bacillus subtilis endospores lacked the d-Ala at the fifth locus of the peptide stem (24), which was the labeling target of FDAA in Gram-positive bacteria (7). This finding might explain the absence of FDAA labeling signals in SFB intrasegmental bodies. Of note, strong TADA and very weak Cy5ADA labeling were observed in the segments where intrasegmental bodies were found (Fig. 6, C and D, yellow arrows). This staining pattern suggests that metabolism of segmental PGN might be halted during formation of intrasegmental bodies, thus leading to reduced decay of the TADA signals (used in the first gavage) and much weaker labeling of Cy5ADA (used in the second gavage). Alternating intensities of the two colors were observed in some SFB cells on their neighboring septums (fig. S10). This pattern suggests that these segments were probably dividing during the two labeling steps, and that neighboring septums formed at different times were labeled with different concentrations of the two FDAAs. The phenotypes of SFB have been heavily studied mostly by scanning electron microscopy and Gram staining for nearly 50 years (25, 26). Our labeling strategy offers new perspectives on these important bacteria and will be a useful tool for further understanding the microbiology of SFB in vivo.
How PGN is constructed is one of the central topics in bacteriology. Sequential FDAA labeling has been used in investigating PGN synthesis in many model bacterial species (27). The strategy proposed here, to probe a large number of species in the microbiota collectively, greatly improves the efficiency of bacterial morphogenesis studies. Two-color imaging of the labeled gut microbiota recorded in situ growth and division patterns of the highly diverse gut bacteria during the 6 hours of labeling, giving us a unique opportunity for a direct look at how this “gut dark matter” actually grows and multiplies in vivo. The highly distinct labeling patterns of different gut bacteria offer a gold mine for microbial cytologists, where new bacterial growth and division patterns may be discovered. Moreover, FDAA-based visualization of several gut bacterial groups that have been heavily studied will enable further understanding of the activities of these important microbes in the mammalian gut. It is worth mentioning that among the 15 genera that were FDAA labeled, many species of Prevotella, Roseburia, Oscillibacter, Anaerotruncus, and Barnesiella were on the “most wanted” taxa list with high priority from the National Institutes of Health Human Microbiome Project (28). The information on in situ growth patterns captured by our method will provide valuable information about these understudied bacteria.
Further optimization of the FDAA probes, for instance, use of smaller fluorophores, for example, may improve labeling coverage, especially for those Gram-negative genera that could not yet be labeled. Optimized labeling protocols (e.g., varied time intervals between administration of the two probes) may assist characterization of the morphogenesis of different bacteria growing at different rates. Superresolution microscopy using FDAA-containing compatible fluorophores will also help in elucidating the distinct arrangements of PGN synthesis machineries in different bacteria, which will greatly enrich our knowledge of microbial cytology. Besides mouse gut the microbiota, studies of other complex microbial systems, such as the microbiota from other animal hosts and environmental microbiotas, including those in soil, water sediments, etc., may also benefit from this labeling and imaging strategy.
MATERIALS AND METHODS
FDAA probe was purchased from Chinese Peptide Company (Hangzhou, China). FISH probes and paraformaldehyde were from Sangon Biotech (Shanghai, China). Other chemicals, not noted above, were from Sigma-Aldrich (St. Louis, MO, USA).
C57BL/6 specific pathogen–free mice (male, 6 weeks old) were obtained from Jie Si Jie Laboratory Animals (Shanghai, China). Mice were bred in the Renji Hospital animal facility in a temperature-controlled (25°C) environment with a 12-hour light/dark cycle, receiving a standard chow diet and free access to clean water. All animal experiments were performed in accordance with guidelines approved by the Institutional Animal Care and Use Committee of the Shanghai Jiao Tong University School of Medicine.
Sequential labeling of microbiotas with FDAA probes
The C57BL/6 mice were sequentially administered by two different FDAA probes (200 μl, 1 mM TADA or Cy5ADA in distilled H2O) through oral gavage with an interval of 3 hours. Their small intestine and cecal microbiotas were collected using a previously reported protocol (9). Briefly, mice were euthanized by cervical dislocation, and the small intestine and cecum were dissected separately and finely minced with a pair of 11.43-cm iris scissors in 1 ml of phosphate-buffered saline (PBS). The tissues and digesta were then filtered with a 40-μm cell strainer to remove most of the nonbacterial materials. The filtrates were then centrifuged. The bacterial pellets (whitish-colored) were washed three times with 1.5 ml of PBS by centrifugation (15,000g, 3 min) and then resuspended in PBS for subsequent experiments.
In vitro culture of soil microbiota
Five grams of sediment collected from the Yangtze Estuary were homogeneously resuspended in 50 ml of sterile physiological water. Ten-fold serial dilutions of the sediment suspension were performed to 10−3. The serial dilutions (100 μl) were dispersed on Gause’s synthetic agar medium [containing 2% soluble starch, 0.1% KNO3, 0.05% NaCl, 0.05% K2HPO4, 0.05% MgSO4, 0.001% FeSO4, and 2% agar (pH 7.2)] and then incubated at 30°C. After 3 days, bacterial cultures (10−2) with the most phenotypic diversity were collected, washed three times with 1.5 ml of PBS by centrifugation (15,000g, 3 min) and resuspended in sterile PBS for subsequent experiments.
FISH probe design
Candidate FISH probes were identified using a k-mer–based algorithm similar to KASpOD (29). The sequenced 16S ribosomal RNA (rRNA) genes were downloaded from the Ribosomal Database Project (RDP) (30) and SILVA (31) databases. Here, the missing 16S rDNA sequences of some target groups were downloaded from National Center for Biotechnology Information (32) and added manually. The final pool was consisted of 3,200,588 sequences. Target groups were defined as sequence subclasses consisted of 16S rDNA sequences from target family/genus. The nontarget groups were defined as sequence subclasses consist of 16S rDNA from nontarget bacteria, which located in the same order/family with targets. For each group, the fully overlapping k-mers were then clustered at an 88% identity clustering threshold to get the degenerate consensus k-mer. Coverage and specificity evaluation of each degenerate consensus k-mer from the target group were performed by a coverage assessment against the target and nontarget groups, respectively. Consensus k-mers with best coverage and specificity was finally used as candidate probes. The related Perl scripts are available from https://github.com/songjiajia2018/Pdesign/archive/master.zip.
The newly designed probes were added to the microbiota suspensions for test using probes EUB338 and NONEUB as positive and negative controls, respectively. There are a number of parameters that can be adjusted, such as temperature and formamide concentration, to choose the best stringency for hybridization. According to the method described by Manz et al. (33), varying the concentration of formamide in the hybridization buffer at a constant hybridization temperature (46°C), ranging from 0 to 70% (in 5% increments), was used to evaluate the optimal stringency of each probe designed in this study. Subsequently, an equally stringent 30 min posthybridization wash for each hybridization was performed at 48°C. Confocal or flow cytometry can be used to quantify the bacteria FISH signal intensity at different formamide concentrations. The highest formamide concentration before losing the specific hybridization signal was regarded as optimal hybridization stringency for the test probe.
To test the labeling specificities of the newly designed FISH sequences, the probes were separately tested against a fixed soil microbiota sample that did not share any genera with the mouse gut microbiota, using protocols described below. The presented 16 new FISH probes all showed negative labeling in the test (fig. S11). Further specificity confirmation tests were performed by flow cytometry and confocal fluorescence imaging. Flow cytometry was carried out to analyze the labeling ratios of some genera (Clostridium, Barnesiella, and Alloprevotella) in the microbiota, using a previously published method (34). These tests showed results consistent with their relative abundances deduce by 16S rDNA sequencing (fig. S12), indicative of high labeling specificities. Other genera and species tagged with the new FISH probes were analyzed by confocal fluorescence microscopy to assess whether the cell morphologies and labeling patterns of the stained bacteria were consistent in each group, which could further verify the specificities of the new FISH probes.
Fluorescence in situ hybridization
An equal volume of 4% paraformaldehyde was added to the bacterial suspensions in PBS and incubated at room temperature for 1.5 hours to fix the bacteria. The samples were then washed twice with PBS and resuspended in 50% ethanol-PBS (v/v) and stored at −20°C for >24 hours. After washing with PBS, the bacterial pellets were resuspended in a hybridization buffer [0.9 M NaCl, 20 mM tris (pH 7.5), 0.01% SDS, and formamide, if required] (table S1). The FISH probe was then added with a final concentration of 5 ng/μl and incubated overnight at required temperature (table S1) using a ThermoMixer (Eppendorf, Hamburg, Germany). After hybridization, bacteria were then washed two times (15 min) with washing buffer [0.9 M NaCl, 20 mM tris (pH 7.5), and 0.01% SDS]. Bacteria were then resuspended in tris buffer [20 mM tris and 25 mM NaCl (pH 7.5)] before analysis with fluorescence microscopy and flow cytometry. FISH probe sequences that have been previously reported (35–44) are listed in table S1.
Confocal fluorescence microscopy
A bacterial suspension was added to an agarose layer [1.5% (w/v) in PBS, ~1-mm thick] and covered with a glass coverslip. Confocal fluorescence imaging was performed on a TCS SP8 laser scanning confocal microscope (Leica, Solms, Germany). Samples were excited with 488 nm for FAM (carboxyfluorescein), 555 nm for TAMRA, and 639 nm for Cy5 fluors, and the emission was detected using the corresponding emission filters. Deconvolution of the images was performed using Huygens Essential Deconvolution software (Scientific Volume Imaging B.V., Hilversum, The Netherlands) using a theoretical point spread function.
Flow cytometry analyses of the FDAA-labeled microbiota samples were carried out on a CytoFLex flow cytometer (Beckman Coulter Life Sciences, Indianapolis, IN, USA). FlowJo (V 10.0.8r1) was used for data analyses. Labeled microbiota were identified with flow cytometry plots of logFSC versus logSSC and then gated on fluorescence. For each sample, 15,000 events were collected for analysis with debris and doublets excluded.
DNA from the microbiotas was extracted either using a stool DNA Kit or bacterial DNA Kit (Omega Bio-tek, Norcross, GA, USA) according to the manufacturer’s protocol. The 16S rDNA sequencing and metagenomic sequencing were performed by Majorbio Bio-Pharm Technology Co. Ltd. (Shanghai, China). For 16S rDNA sequencing, the V3-V4 hypervariable regions of the 16S rDNA were amplified by polymerase chain reaction and subsequently paired-end sequenced (2 × 300) on an Illumina MiSeq platform (Illumina, San Diego, USA) according to the standard protocols. The taxonomy of each 16S rRNA gene sequence was analyzed by RDP Classifier (http://rdp.cme.msu.edu/) against the SILVA (SSU123) 16S rDNA database with a confidence threshold of 80%. For metagenomic sequencing, DNA was fragmented to an average size of about 400 bp using Covaris M220 (Gene Company Limited, China) for paired-end library construction and subsequently paired-end sequenced on an Illumina HiSeq4000 platform. BLASTP (Version 2.2.28+, http://blast.ncbi.nlm.nih.gov/Blast.cgi) was used for taxonomic annotations by aligning nonredundant gene catalogs against the integrated NR (non-redundant protein sequence) database with e-value cutoff of 1 × 10−5.
Acknowledgments: Funding: We are grateful to the National Natural Science Foundation of China (grants 21807070, 21735004, 21775128, and 21705024) and the Innovative Research Team of High-Level Local Universities in Shanghai (SSMU-ZLCX20180701) for financial support. Author contributions: L.L., W.W., and C.Y. designed the study. L.L., Q.W., J.S., Y.D., and J.G. performed and analyzed experiments. Y.S. contributed intellectually to the analysis and interpretation of the data. L.L., W.W., and C.Y. wrote the manuscript. Competing interests: The authors declare that they have no competing interests. Data and materials availability: The 16S rDNA and shotgun sequencing data of the cecal microbiotas have been deposited in the Sequence Read Archive with BioSample accessions SAMN14694443 and SAMN14694444, respectively. In addition, the 16S rDNA sequencing data of the soil microbiota cultured in vitro has also been deposited in the Sequence Read Archive with BioSample accession SAMN14734540. Additional data related to this paper may be requested from the corresponding authors.
- Copyright © 2020 The Authors, some rights reserved; exclusive licensee American Association for the Advancement of Science. No claim to original U.S. Government Works. Distributed under a Creative Commons Attribution NonCommercial License 4.0 (CC BY-NC).