Soluble CLRs identify bacteria with specific glycan motifs in synthetic mixtures

The aim of this study was to establish a pipeline for identification of bacterial microbiota species that can be recognized by human APC-expressed CLRs. For this proof-of-concept study, we used the macrophage CLR MGL (CD301) and the Langerhans cell CLR langerin (CD207), which have distinct glycan binding specificities [21,22]. First, we validated that our approach could selectively identify and extract bacteria based on their surface glycan composition from a mixed bacterial community. To this end, we created a synthetic microbial community of equal proportions of four bacterial species, i.e., two Gram-positive (Streptococcus thermophilus and Listeria monocytogenes) and two Gram-negative bacteria (Escherichia coli and Alcaligenes faecalis) that did not bind to either recombinant human langerin or MGL (Fig 1A). We then added CLR-binding S. aureus bacteria to this mixture in a proportion of 20% relative to the total community. For MGL, we used S. aureus PS187, for which we previously showed that MGL binds GalNAc attached to surface-anchored wall teichoic acid (WTA) [23]. S. aureus strain N315 was used to validate langerin-based identification and sorting in line with our observations that langerin binds this strain through the β-linked GlcNAc on WTA [24].

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. CLRs can identify bacteria with specific glycan patterns in a synthetic microbial community.

A recombinant human MGL or langerin-FITC staining of the five bacterial species that were used to create a synthetic microbial mixture. S. aureus PS187 and N315 are included as CLR-binding positive controls. B Schematic overview of the procedure. Cell trace yellow (CTY)-stained S. aureus N315 or PS187 (identified in PE channel) were spiked into the synthetic microbial community and the mixed population was subsequently stained with either recombinant human MGL or langerin-FITC. The CLR-positive-and negative populations were sorted. C Proportion of CTY-stained S. aureus in the total, CLR-positive and CLR-negative sorted populations. The exact percentages of CTY+ bacteria are shown for each bar. Data shown is representative of two independent experiments.

https://doi.org/10.1371/journal.pone.0356059.g001

Both S. aureus strains were labeled with Cell Trace Yellow (CTY) before mixing the strains, allowing identification of these bacteria in the phycoerythrin (PE) channel and discriminate it from the synthetic bacterial mixture (S1 Fig). Subsequently, the bacterial mixture was stained with FITC-labeled his-MGL or langerin and FITC-positive and -negative bacteria were sorted using fluorescence-based cell sorting (Fig 1B). Re-analysis of sorted CLR-positive and -negative populations showed strong enrichment of CTY-stained S. aureus in the CLR-positive population and depletion in the CLR-negative populations (Fig 1C). These results show that fluorescence-based cell sorting of bacteria based on CLR-staining is glycan specific.

Bacterial species are present at different abundances in gut microbiota. Therefore, we aimed to show that CLR-based cell sorting would allow identification of low-abundant bacteria in a complex synthetic microbial community by decreasing the proportion of CTY-stained S. aureus PS187 or N315 cells to 10% and 1% of the total synthetic bacterial community. Similar to previous results, bacteria could be discriminated from the synthetic bacterial community by double staining even at lower abundance (S1 Fig). Sorting and re-analysis of CLR-positive and -negative bacteria showed that independent of the proportion of S. aureus added to the mixture, CTY-positive S. aureus were strongly reduced in the negative population (Fig 1C). Complementary, even at 1% abundance, the positively-sorted fractions were strongly enriched for S. aureus, with approximately 80% CTY-positive cells at 1% presence (Fig 1C). Taken together, these data indicate that bacterial strains with specific glycan motifs can be identified and sorted from more complex microbial communities.

CLR-sequencing on human microbiota

We have thus far stained and identified known CLR-binding S. aureus strains in a synthetic community of limited diversity. To allow identification of new bacterial species that are recognized by human MGL or langerin, we coupled CLR-based binding and sorting to 16S rRNA gene sequencing. We first stained isolated human fecal microbiota with the DNA stain SYTO60 to determine the bacterial presence and purity of our isolation protocol. On average, 85% of events was SYTO60 positive (S2A Fig), implying that our isolation protocol removes most debris and non-bacterial particles. Having validated our bacterial purity, we next assessed the interaction between FITC-labeled MGL or langerin upon incubation of the isolated fecal material. The isolated fecal bacteria were stained with FITC-labeled MGL or langerin and the CLR-positive populations were obtained through fluorescent cell sorting. Re-analysis of the CLR-positive sorted fractions showed 89% and 94% SYTO+ events for langerin and MGL, respectively, indicating the high presence of bacterial particles in these samples (S2A Fig). Next, we tested the effect of the calcium chelator EGTA on CLR binding to fecal microbiota, as CLR interactions are calcium dependent [25]. We observed donor specific differences in EGTA-mediated blocking of MGL and langerin binding of microbiota (S2B and C Figs). EGTA blocking appeared more effective on MGL than on langerin binding and the blocking was more effective for donors with higher CLR bound populations. This indicates the potential of non-specific binding to a subset of bound bacteria after fluorescence-based sorting. Therefore, post-sorting validation is crucial for discriminating biological binding from non-specific binding.

Having established that our pipeline allows identification of CLR-reactive bacterial strains/species with little non-bacterial impurities, we applied the CLR-seq pipeline to isolated fecal microbiota samples from three individual healthy human donors. Fecal microbiota samples were incubated with MGL or langerin and positive fractions were sorted using fluorescence-activated cell sorting. Additionally, we identified and sorted IgA-positive bacteria from the same fecal samples, as it was previously shown that this pipeline allows identification of bacterial species with immunostimulatory properties [14,15]. We observed differences in the proportion of IgA-, MGL- and langerin-reactive bacteria of the healthy donors, where the proportion of CLR-reactive bacteria was smaller compared to the IgA-positive fraction (Fig 2A).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. Characteristics of CLR- and IgA-positive populations from fecal microbiota.

A MGL-, langerin-, and IgA-positive bacterial populations were collected from fecal microbiota samples (n = 3 healthy donors), using fluorescence-based cell sorting, and analyzed with 16S rRNA gene sequencing. B Multilevel principle component analysis of all sequenced samples with donor-specific effects removed prior to ordination. C Relative abundances at the genus level for genera that were among the 10 most abundant genera in at least one sequenced sample.

https://doi.org/10.1371/journal.pone.0356059.g002

Subsequently, DNA was extracted from the total unsorted microbiota input as well as IgA- and CLR-positive sorted fractions and analyzed by 16S rRNA (V3-V4 regions) gene sequencing on isolated DNA from these fractions to identify the bacteria species in the enriched fractions. The reads were mapped against zero-radius operational taxonomic unit (zOTU) sequences, after which taxonomy was assigned to obtain the bacterial abundances. No contaminating signatures in the post-sorting reagents were identified. We observed a decrease in species diversity based on the Shannon Diversity index post-sorting when compared to the unsorted, total populations (S3A Fig). Importantly, the sequencing data varied most strongly between the total unsorted microbiota samples and the sorted IgA and CLRs fractions based on a multilevel principal component analysis (PCA) with removal of donor-specific effects prior to ordination (Fig 2B). Based on the 16S rRNA gene sequencing, relative abundances were determined at various taxonomic levels for the total populations and sorted samples. Despite interindividual variation between donors at the genus taxonomic level, some shared changes in relative abundances could be observed (Fig 2C and S1 Table). For all three donors, the relative abundance of Bacteroides bacteria was decreased for MGL-, langerin-, and IgA sorted samples. In contrast, the proportion of Butyricicoccus and Coprococcus 1 was increased in all three MGL sorted donor samples. The langerin samples for donor 1 and 2 showed an increase in relative abundance for Blautia and Faecalibacterium bacteria, respectively. When examining the sample composition at the taxonomic species level, we observed a large presence of specifically Blautia obeum in donor 1 and a Faecalibacterium sp. (zOTU_579) in donor 2 of the langerin sorted samples (S3B Fig and S2 Table).

Using the relative abundances from the 16S rRNA gene sequencing analysis and fraction sizes of the sorted populations, positive probability scores were calculated for all zOTUs as described previously [26]. This method is less influenced by the pre-sort taxonomic composition of the microbiota than conventional approaches. To avoid the introduction of potential contaminants during sample preparation, collection or sequencing, we only included zOTUs that were detected in the total population of all three donors and had a relative abundance of >0.1% in at least two donors. Probability scores of zOTUs in the MGL-, langerin-, or IgA-positive samples did not correlate with their relative abundance in the total population for any of the donors, suggesting that species abundance in the unsorted microbiota population does not influence CLR- or IgA-binding (S4 Fig).

Plotting the individual values for the zOTUs with the highest and lowest average probability scores showed the individual differences in binding of specific species (Fig 3). In the zOTUs with the highest average MGL binding (Fig 3A), donor 2 often showed the highest score, whereas for langerin (Fig 3B) and IgA (S5 Fig) this varied more among the donors. For both MGL and langerin, the zOTUs with the lowest average positive probability scores consisted mostly of Bacteroides sp. with a number of these being completely depleted in the sorted sample, as indicated by a positive probability score of 0.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 3. CLRs bind to specific human gut bacteria species.

Probability scores were calculated for CLR binding based on the 16S rRNA gene sequencing relative abundances. Shown are the 10 zOTUs with either the highest (grey background) or lowest average probability scores (white background) for individual donors for A MGL or B langerin. Open circles are used for donors where the zOTU had < 0.1% relative abundance in the total population.

https://doi.org/10.1371/journal.pone.0356059.g003

Marvinbryantia sp. zOTU_690 and Ruminococcaceae UCG-005 sp. zOTU_141 were among the zOTUs with the highest average probability scores for MGL and langerin. Interestingly, these species also showed high IgA binding (S5 Fig). Complementary, zOTUs with low probability scores for CLR binding also had low IgA probability scores (S5 Fig). Including all zOTUs in the sorted fractions, the average probability scores of the CLR- and IgA-positive zOTUs correlated significantly (S6 Fig).

Validation of CLR binding to species identified by CLR-seq

Interestingly, we observed Blautia sp. zOTUs in both the highest and lowest average langerin probability scores. This indicates that CLR-seq may discriminate within the taxonomic genus level between species expressing different glycan profiles [27]. We therefore deemed it important to validate a number of species in the CLR-binding fraction as monoculture.

Bifidobacterium bifidum, a gut commensal with probiotic characteristics and present in the intestinal tracts of nearly every individual [28], and Dorea formicegenerans, another intestinal commensal bacterium, were among the species with the highest average probability scores for MGL (Fig 3A). To validate the CLR-seq results of these donors, MGL binding was tested on monocultures of a B. bifidum and D. formicegenerans strain. MGL interacted with B. bifidum and this binding was dependent on the carbohydrate recognition domain (CRD) of MGL and bivalent cations, since MGL binding was abrogated by the addition of GalNAc and significantly reduced in the presence of EGTA, respectively (Fig 4A and S7A Fig). No binding of MGL to the D. formicenegerans monoculture was observed (Fig 4A and S7A).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 4. Validation of CLR binding to bacterial species identified through CLR-seq.

Binding of A recombinant human MGL and B recombinant human langerin-FITC to bacterial species monocultures identified by CLR-seq. CLRs were used at a concentration of 25 µg/mL, similar to the concentration used during sorting. Binding was blocked using GalNAc or mannan for MGL and langerin, respectively, and with the calcium chelator EGTA. Statistical comparisons indicated are the blocked or unstained/α-his FITC antibody-stained bacteria compared to the CLR stained control. Data are shown as the mean of the geometric mean fluorescence intensity (FI) ± standard error of mean from three independent experiments. *, p < 0.05; ***, p < 0.001; ****,p < .0001.

https://doi.org/10.1371/journal.pone.0356059.g004

For langerin binding, the gut commensal bacterium Anaerostipes hadrus was among zOTUs with the highest average probability scores (Fig 3B). Therefore, we tested binding of this CLRs to an A. hadrus monoculture (Fig 4B and S7B Fig). We observed langerin-FITC binding, although this binding was not abrogated by the addition of EGTA, and even increased by the addition of mannan, suggesting that this binding was non-specific.

To avoid the interference of contaminants in our 16S rRNA gene sequencing analysis, we had removed zOTUs that were not detected in the total population of their respective donor. However, when examining the removed zOTUs more closely, we observed consistent presence of various zOTUs in the CLR-positive fractions of our three donors. For both MGL and langerin-positive fractions, we detected the genus Aquamicrobium, which was first described in activated sewage sludge [29]. In our 16S rRNA gene sequencing data, this genus consisted of a single zOTU attributed to two species, A. lusatiense and A. defluvii. Additionally, we identified two zOTUs at the species level within the genus Corynebacterium with consistent presence in the langerin-positive fraction. These were C. accolens, an abundant nasal commensal, and C. appendicis, which was first isolated from the abdominal swab of an appendicitis patient [30,31]. To verify CLR binding to the mentioned species, an individual strain of each species was grown as a monoculture and incubated with recombinant human MGL or langerin, corresponding to the sorted fractions in which they were identified (S7 Fig). MGL bound significantly to A. defluvii and binding was abrogated by the addition of GalNAc and EGTA (Fig 4A and S7A Fig), suggesting that the recognition by MGL occurred through the CRD and was calcium dependent. In contrast, only minor MGL binding to A. lusatiense was observed, which was not blockable by either GalNAc or EGTA and therefore concluded to be non-specific (Fig 4A, S7A Fig). For langerin, we observed significant binding to A. defluvii and a loss of binding by the addition of mannan or EGTA (Fig 4B and S7B Fig). No langerin binding to A. lusatiense was observed (S7B Fig). C. appendicis and C. accolens both bound to recombinant human langerin in a CRD- and calcium-dependent manner (Fig 4B and S7B Fig), although the binding appeared to be saturated at a lower concentration for C. appendicis. The combined results of CLR binding to the individual species identified by 16S rRNA gene sequencing highlights the potential of CLR-seq to specifically identify bacteria that can be recognized by CLRs.

Fecal donor samples

The fecal donor samples used were obtained retrospectively from the Amsterdam University Medical Center (Amsterdam, the Netherlands) and accessed for research purposes between 01-05-2022 and 31-07-2025, where they were previously collected as part of the PIMMS study (NL67136.018.18) [45]. This study was approved by the Medical Ethical Reviews Committee at Amsterdam UMC (previously Amsterdam Medical Center (AMC)). Information that could identify the individual participants was not accessible during data collection.

Bacterial strains and culture conditions

The bacterial strains used in this study are listed in Supporting information S1 File. Staphylococcus aureus and Aquamicrobium strains were grown in Tryptic Soy broth (TSB; Oxoid) and Escherichia coli in Luria Bertani (LB; Oxoid) broth. Streptococcus thermophilus was grown in Todd-Hewitt broth supplemented with 1% yeast extract (THY; Oxoid) in the presence of 5% CO₂ and Listeria monocytogenes in Brain Heart Infusion (BHI; Oxoid) broth. Overnight cultures were diluted to an optical density 600 nm (OD600) of 0.05 in fresh medium and grown to mid-exponential phase, after which bacteria were harvested by centrifugation. Alcaligenes faecalis and the Corynebacterium species were grown on Columbia blood agar (Oxoid), with Corynebacterium appendicis grown in a microaerophilic environment. Bacteria were collected from plate using an inoculation loop, and washed with Tris-sodium-magnesium buffer (TSM: 20 mM Tris; Roche, 150 mM NaCl; Merck, 2 mM CaCl₂; Merck, 2 mM MgCl₂; Merck, pH 7.0) + 0.1% Bovine Serum Albumin (BSA). Bifidobacterium bifidum and Dorea formicegenerans were isolated from human feces as described previously [14]. Subsequently, they were grown in an anaerobic chamber in Gut Microbiota medium (GMM) [46]. Overnight cultures were diluted in fresh medium to an OD600 of 0.05 and grown for 6 hours, after which they were harvested by centrifugation. Anaerostipes hadrus was cultured in modified yeast extract, casitone and fatty acid (YCFA) medium [47] supplemented with 30 mM glucose. Overnight cultures were diluted in fresh medium (1:10) and grown for 4 hours, after which they were harvested by centrifugation. All species were grown at 37°C, except for the Aquamicrobium species, which were grown at 30°C.

Expression and purification of recombinant human MGL

The extracellular domain of human MGL (CLEC10A) was expressed and purified as described previously [48]. A gBlock encoding the open reading frame of human MGL and a C-terminal LPETGG-6xHis tag were cloned into a pcDNA34 expression vector (Invitrogen). Expression of recombinant hMGL-his proteins was performed in Expi293F cells (Life Technologies), which were cultured in Expi293 Expression Medium (Life Technologies). The protein was purified from the supernatant using affinity chromatography (ÄKTA Pure, GE Healthcare Life Sciences) using a Nickel column (GE Healthcare Life Sciences) as previously described [49]. Eluate was dialyzed against 300 mM NaCl 50 mM Tris pH 7.8 at 4°C.

Staining and flow cytometric analysis of bacterial strains

Bacteria were collected from broth culture by centrifugation or from plate with an inoculum loop after overnight growth, washed with and resuspended in TSM + 0.1% BSA after centrifugation (3,600 x g, 10 min, 4°C). Bacterial abundance was determined by optical density at 600 nm or using flow cytometry with Precision Count Beads (Biolegend 424902) using a BD FACSymphony or BD FACSCanto.

For langerin and MGL specificity analysis in synthetic microbial communities, S. aureus strains N315 and PS187 (Supplementary information S1 File) were used as positive controls, respectively [23,50]. Both S. aureus strains were fluorescently-labeled with Cell Trace Yellow (CTY; Invitrogen) as per manufacturer’s instructions and mixed at desired relative abundances in the synthetic microbial community consisting of 10⁷ bacterial cells (quantified by flow cytometry). Subsequently, the bacterial mixture was stained with either FITC-labeled recombinant langerin (kindly provided by Prof. Christoph Rademacher, University of Vienna, Austria) [22] or his-tagged recombinant human soluble MGL at 25 µg/mL in TSM + 0.1% BSA for 30 min at 37°C (900 rpm shaking, covered from light). After washing with TSM + 0.1% BSA, hMGL binding was detected using an anti‐hisTag FITC‐conjugated antibody (1:20 in TSM + 0.1% BSA, Invitrogen; clone AD1.1.10), incubated for 20 minutes at 4°C, covered from light. FITC-positive and -negative bacteria were sorted and analyzed using a Sony SH800 flow cytometric cell sorter.

To determine binding of CLR constructs to bacterial monocultures, 10⁶ CFU (determined by OD600) or bacterial cells (quantified by flow cytometry) were stained with langerin-FITC or his-hMGL as described above. MGL binding was blocked by adding N-acetyl-galactosamine (GalNAc; 100 mM; Sigma-Aldrich A2795) and langerin binding was blocked through the addition of mannan (20 µg/mL; Sigma-Aldrich M7504). To test for calcium dependent binding, the calcium chelator EGTA (Sigma Aldrich, 10 mM) was added. After washing with TSM + 0.1% BSA, bacteria were fixed with phosphate buffered saline (PBS) + 1% formaldehyde (Sigma-Aldrich) and measured by flow cytometry (BD FACS Canto).

Fecal bacteria isolation and staining

Fecal microbiota was isolated as previously described [14] with some modifications. Approximately 100 mg frozen fecal material was resuspended in 1 mL of ice-cold PBS for 15 minutes on ice. Next, the suspension was transferred to a Fast Prep Lysing Matrix D tube containing 1.4 mm zirconium-silicate beads (MP Biomedicals) and homogenized by beat-beating for 10 seconds at 6,000 rpm (MagNa Lyser; Roche). To separate the fecal bacteria from the debris, the homogenized suspension was centrifuged (50 x g, 15 minutes, 4°C) and 100 µL supernatant was transferred to a clean 1.5 mL Eppendorf tube and washed with 1 mL PBS + 1% BSA (8,000 x g, 5 min, 4°C). The fecal pellet was resuspended in 1 mL PBS + 1% BSA and filtered using a 70 µm mini cell strainer (PluriSelect), after which 20 µL was set aside as pre-sort sample.

To determine CLR-specific binding of fecal bacteria, the filtered microbiota suspension was centrifuged and stained with 25 µg/mL recombinant human langerin-FITC or his-tagged human MGL in TSM + 1% BSA, and incubated 30 minutes at 37°C with agitation (600 rpm). MGL-stained bacteria were stained with anti-his-FITC after washing (1:20 in TSM + 1% BSA; Invitrogen, clone AD1.1.10) for 20 minutes at 4°C, covered from light. For IgA sorting, fecal bacteria were stained with PE anti-human IgA (Miltenyi Biotec, clone IS11-8E10; 1:10 in PBS + 1% BSA) and incubated 30 minutes on ice. Samples were washed with and resuspended in TSM + 1% BSA and analyzed by flow cytometry measuring 20,000 events (FACSymphony A1, BD Biosciences), or sorted based on fluorescent staining using fluorescence-activated cell sorting (Sony SH800). For cell sorting, the percentage of positive events was measured and 200,000 positive events per staining were collected in PBS + 1% BSA, pelleted by centrifugation (8,000 x g, 5 min, 4°C), resuspended in 50 µL PBS and stored at −20°C until DNA extraction.

DNA extraction, 16S rRNA gene sequencing

Fecal DNA extraction, 16S rRNA gene sequencing analysis, and bioinformatics analysis were performed as described previously [51]. Briefly, total genomic DNA was isolated from the pre-sorted and sorted microbiota samples at the Microbiota Centre of Amsterdam (MiCA, Amsterdam, the Netherlands). Isolation was performed using a repeated bead beating protocol and the Maxwell RSC Whole Blood DNA kit [52]. DNA concentration was measured in a 96-wells plate using the Qubit dsDNA BR Assay. Amplification of the V3-V4 regions of the 16S rRNA gene sequences was performed with the 341F-805R primers in a single step PCR protocol. Sample collection kits containing only solubilizing buffer with no stool samples were used as negative control and were followed for the same extraction steps. Samples were purified using Ampure XP beads, measured and sequenced equimolarly using Illumina MiSeq (v3 chemistry with 2  ×  250 cycles).

Amplicon sequences were parsed using a vsearch (v2.15.2) based pipeline [53]. Paired end reads were merged, with max differences set to 100 and allowing for staggered overlap. Zero-radius Operational Taxonomic Units (zOTUs) were inferred from reads with lower than 1.5 expected error rate using the cluster_unoise with centroids algorithm with a minsize of 4, after which chimeras were removed using the uchime3 denovo method. For each sample zOTU abundances were determined by mapping the merged reads against zOTU sequences using usearch_global with a 0.97 distance cut off. Taxonomy was assigned using R (4.2.0) and the dada2 [54] assign taxonomy function using the silva (v132) [55] reference database. A phylogenetic tree was generated use mafft (v7.310) [35] and Fasttree (2.1.11) [56].

The abundance and zOTU tables were used for downstream analysis using R’s phyloseq package [57]. Probability scores were calculated using Microsoft Excel using relative abundances and sorted fraction size, similarly as described previously [26] and zOTUs that were present in the sorted sample of all three individual donors were included. For visualization of 16S rRNA gene sequencing analysis, R-studio (version 4.4.1) or Graphpad Prism (version 10.2.0) was used.

Statistical analysis

Visualization and statistical analysis were performed using Graphpad Prism (version 10.2.0). Differences in CLR binding at specific concentrations and potential blocking were calculated with a one-way analysis of variance (ANOVA) for each individual species, followed by Bonferroni’s test for multiple comparison. For CLR binding concentration curves, a two-way ANOVA followed by Bonferonni’s multiple-comparison test was used. A p-value of <0.05 was considered significant.