Resident microbes shape the vaginal epithelial glycan landscape

Sialic acids are depleted from the vaginal glycocalyx in BV

To study the vaginal epithelial glycocalyx, the overall properties of epithelial glycans were first examined using transmission electron microscopy (TEM) using a classical method of ruthenium red and osmium tetroxide staining^64,65. BV status was determined using Nugent’s scoring method by analysis of Gram-stained slides of vaginal specimens⁶⁶. Epithelial cells were isolated from vaginal swab eluates of women with (Nugent score 7-10) or without BV (Nugent score 0-3) as described in the methods. In a pilot study, epithelial cells from a woman without BV (referred to as “No BV’’ in the figures hereafter) had a fuzzy layer of glycocalyx protruding from the VEC membrane (Fig. 1A). However, there was little to no glycocalyx evident on the cell surfaces from one BV specimen (Fig. 1B), suggesting that the vaginal glycocalyx may be degraded in BV. These observations propelled us to perform a more detailed characterization of the vaginal glycocalyx and its glycan components.

• Download figure
• Open in new tab

Fig. 1: Visualization of the vaginal epithelial glycocalyx in women with and without BV.

Representative transmission electron microscopy (TEM) images of vaginal epithelial cell (VEC) stained with ruthenium red and osmium tetroxide. One specimen from (A) a woman without BV (Nugent 0-3, referred to as No BV) and (B) a woman with BV (Nugent 7-10) were used for staining and microscopy. Arrowheads point to the glycocalyx observed on one VEC from the woman without BV. Data is representative of two independent experiments. Scale bar for images in the left panel is 2 µm and for those in the right panel is 500 nm.

The most abundant sialic acid in humans is N-acetylneuraminic acid (Neu5Ac)⁴². Initial investigation of epithelial surface sialylation performed in specimens without BV revealed prominent binding of the Maackia amurensis lectin-II (MAL-II) (Fig. 2A), which recognizes terminal α2-3-linked sialic acid residues (Neu5Acα2-3Gal)⁶⁷. Specificity was confirmed by pretreating No BV VECs with exogenous sialidase from Arthrobacter ureafaciens (A.u. sialidase), which eliminated MAL-II binding (Fig. S1) and there was minimal background in the secondary alone control (Neutravidin-OG488). Sialidase activity was measured in each vaginal specimen by incubating the eluate with fluorogenic substrate Neu5Ac-4-methyl umbelliferone (4MUSia). In agreement with previous studies^54,56, we found sialidase levels were significantly higher in samples from women with BV compared to women without BV (Fig. 2B). In parallel analyses, eight of nine BV specimens with high sialidase activity (rate of 4MUSia hydrolysis > 0.1), showed negligible MAL-II staining, suggesting extensive depletion of epithelial α2-3-linked sialic acids in BV. In some BV specimens with low sialidase activity, evident MAL-II binding suggested that sialic acid depletion is related to the presence of sialidases in vaginal fluids. The lectin analysis, together with the preliminary electron microscopic analysis, strongly suggest that sialic acids are depleted from the vaginal glycocalyx in women with BV.

• Download figure
• Open in new tab

Fig. 2: Depletion of sialic acids from vaginal glycans in women with BV.

(A) Representative confocal images of VECs stained with MAL-II lectin (Green) recognizing α2-3-linked sialic acids from individual women without BV (n=3) and with BV (n=12). Specimens were compared across varying levels of endogenous sialidase activity. Nuclei and bacterial cells (Blue) are stained with DAPI. Scale bars = 50 μm. (B) Quantification of endogenous sialidase activity in the vaginal swab eluates using the fluorogenic substrate Neu5Ac-4-methyl umbelliferone (4MU-sialic acid). Data shown is the rate of 4MU hydrolysis and the points represent values for individual women (n=16 without BV and n=20 with BV). Data in A and B is combined from 3 independent experiments. Images shown in A are from a subset of specimens used in B. (C and D) Fluorimetric quantification of 1,2-diamino-4,5-methylenedioxybenzene (DMB)-labelled sialic acid (Neu5Ac) in isolated VEC N- and O-glycans, using reversed-phase chromatography after mild acid hydrolysis. Graph shows sialic acid quantification on glycans derived from protein extracts of pools of VECs from women with BV (Nugent scores 7-10, n=7 pools from a total N=45 specimens, with 5 or 10 specimens in each pool) and without BV (Nugent scores 0-3, No BV, n=8 pools from a total of N=55 specimens, with 5 or 10 specimens in each pool). Glycans derived from No BV VEC pools pretreated with A.u. sialidase were included as a control (n=2 pools with 10 specimens in each pool). Data in C and D are from same pools of VECs and was combined from 2 independent experiments. Error bars show standard deviation for each group, Mann–Whitney U test was used. ***P < 0.001, ****P<0.0001. A.u. sialidase = sialidase from Arthrobacter ureafaciens. A total of N=151 specimens were used to generate these data. A subset of VEC pools from C and D were also used for studies reported in Fig. 3, Fig. 4 and Table 1. See methods for pooling rationale.

The repertoire of glycans on mammalian cell surfaces varies from tissue to tissue and its characterization can be complicated by branching and extensive heterogeneity⁶⁸. Two abundant glycan types modify proteins on mammalian surfaces. O-linked glycans are found on serine or threonine residues, whereas N-linked glycans are attached to asparagines. To quantitatively assess the degree of sialylation on N- and O-glycans, epithelial cells were pooled from multiple specimens to yield sufficient material for analysis. Glycans were released from equal amounts of extracted protein for each condition. N-glycans were released with PNGase-F under denaturing conditions, while O-glycans were released using reductive beta-elimination (see methods). Sialic acids were released from glycans by mild acid hydrolysis, fluorescent derivatized using 1,2-diamino-4,5-methylenedioxybenzene (DMB), and quantified by reverse-phase chromatography by comparison to synthetic standards. Biochemical measurements of sialic acids (Neu5Ac) in these preparations revealed that BV specimens had ~4.5-fold lower sialylation levels on N-glycans and ~9-fold lower sialylation levels on O-glycans (Figs. 2C, D). As a control, VEC pools from women without BV were pretreated with A.u. sialidase and washed to remove liberated sialic acid prior to lysate preparation, revealing BV-like N- and O-glycans with low levels of sialic acid.

Depletion of negatively charged N-linked glycans on epithelial cells in BV

To further characterize epithelial N-glycans, we employed anion exchange chromatography to separate neutral and negatively charged glycans. This analysis distinguishes between different types of N-glycans [e.g. high mannose (neutral) or mono-, bi-, tri- and tetra-antennary sialylated glycans]. N-glycans were isolated from protein extracts of VECs using PNGase-F, labelled with 2-amino benzamide (2-AB) and analyzed by high-performance anion exchange chromatography (HPAEC) with fluorimetric detection (Fig. 3A). N-glycan standards isolated from the glycoproteins RNaseB and bovine fetuin were analyzed in parallel, illustrating a clear separation between neutral and charged glycans. Neutral N-glycan structures were abundant in all women. However, peaks corresponding to negatively charged N-glycans were smaller and fewer in samples derived from BV specimens compared to those without the condition (Fig. 3B). When the “normal” N-glycans were experimentally treated with sialidase, most of the negatively charged (sialylated) peaks disappeared, and the N-glycans adopted a BV-like profile (Fig. 3C). These results further demonstrate that sialylated epithelial N-glycans are found at substantially lower levels in BV. Consistent with analyses presented in Fig. 2C, these data further underscore that sialylated N-glycans are depleted from vaginal epithelial cells in BV.

• Download figure
• Open in new tab

Fig. 3: Sialylated vaginal epithelial N-linked glycans are depleted in BV.

(A) Schematic for 2-amino benzamide (2AB) profiling of N-linked glycans by high performance anion exchange chromatography (HPAEC). Glycans were released using PNGase-F from protein extracts derived from VECs pools and fluorescently labelled with 2-AB prior to analysis by HPAEC. (B) HPAEC profiles of 2-AB labeled N-glycans derived from (i) protein standards, with well-known glycan structures, Rnase B and Bovine Fetuin, (ii) 2 pools of No BV VECs (n=10 specimens/pool), (ii) 2 pools of BV VECs (n=10 specimens/pool). (C) HPAEC profiles of N-glycans derived from (i) Rnase B and Bovine Fetuin, and (ii) pools of No BV VECs (same VEC pools as used in B, n=10 specimens/pool) pretreated with commercially available sialidase (+A.u. sialidase, dotted line) or with buffer alone (-A.u. sialidase, solid line). Structures of different types of N-glycans (high mannose, mono- and di-sialylated) are depicted following the NCBI Symbol Nomenclature for Glycans (yellow circle, galactose; green circle, mannose; blue square, N-acetylglucosamine; purple diamond, sialic acid). A.u. sialidase = sialidase from Arthrobacter ureafaciens. A total of N=40 specimens were used to generate these data. VEC pools generated from these specimens were also used for studies reported in Fig. 2C, 2D and Fig. 4. See methods for pooling rationale.

O-glycans of human vaginal epithelial cells

In women without BV, measurements of sialic acid levels (Fig. 2D) suggest that the vaginal glycocalyx contains abundant sialylated O-glycans. To identify the relevant O-glycan structures of VECs, mass spectrometric (MS) analysis was performed on permethylated epithelial O-glycans derived from women without BV.

The most commonly found O-glycans in mammals are the mucin type O-glycans⁶⁹. Biosynthesis of mucin type O-glycosylation is initiated by addition of an alpha-linked N-acetylgalactosamine (α-GalNAc) residue to a serine or threonine residue in the peptide backbone^69,70. Subsequently, the core GalNAc can be extended with other sugars to form several different core structures. Analysis of VEC O-glycans using matrix-assisted laser desorption/ionization-time of flight mass spectrometry (MALDI-TOF) revealed that the vaginal glycocalyx contains multiple sialylated O-glycan structures. Among these, we observed a small fraction of di-sialylated core 1 (Galβ1-3GalNAc-Ser/Thr, m/z 1256) structures, and relatively high abundance of differentially sialylated and fucosylated core 2 structures (Fig. 4A). The core-2 structures are generated by the addition of a β1-6 linked N-acetylglucosamine (GlcNAc) to the GalNAc in core 1. Specifically, in “normal” VEC O-glycans the highest intensity peak was consistent with a di-sialylated and fucosylated core 2 ion with m/z 1879, containing a potential sialyl Lewis X (sLeX, also known as CD15s) on one antenna. To verify the presence of Lewis X (CD15) structure, we used anti-Lewis X antibody to probe VECs isolated from No BV vaginal specimens with or without prior treatment with exogenous sialidase (A.u. sialidase). Analysis by confocal imaging and flow cytometry confirmed the presence of Lewis X structure on the VECs of women without BV (Fig. S2). A significant increase in anti-Lewis X binding was observed when VECs were treated with A.u. sialidase, suggesting that most of the Lewis X antigens present on “normal” VECs were sialylated, a conclusion also supported by mass spectrometry.

• Download figure
• Open in new tab

Fig. 4: MALDI-TOF spectra of permethylated O-glycans present in vaginal epithelial glycocalyx of women with and without BV.

(A) Mass spectra revealed presence of sialylated core-1 and core-2 glycans in VECs. Asialoglycans dominate in BV (ion m/z = 983), and sialylated glycans dominate under “normal” conditions (ion m/z=1879). Pools of VECs from women without BV pretreated with exogenous sialidase (A.u. sialidase) were included as control. Two different pools were analyzed for each condition (n=10 specimens/pool). Structures of different types of O-glycans are depicted following the NCBI Symbol Nomenclature for Glycans (red triangle, fucose; yellow circle, galactose; blue square, N-acetylglucosamine; yellow square, N-acetylgalactosamine; purple diamond, sialic acid). Arrowheads point to assigned O-glycan peaks with highest intensity in the spectra. Data is representative of 3 independent experiments. A total of N=60 specimens were used to generate these data. VEC pools generated from these specimens were also used for studies reported in Fig. 2C, 2D, Fig. 3 and Table 1. See methods for pooling rationale.

O-glycan structures were assigned by tandem MS (MS/MS) analysis of selected precursor ions using a linear ion trap mass spectrometer (LTQ-MS). Fragment ions are described based on the nomenclature proposed by Domon and Costello⁷¹. The MS/MS fragmentation spectra of permethylated O-glycans contained molecular ions characteristic of core 1 and fucosylated core 2 O-glycans. We fragmented the sialylated O-glycan precursor ion with m/z 1256, yielding a Y-ion corresponding to sialylated T-antigen (Neu5Ac-Galβ1-3GalNAc-Ser/Thr, m/z 881) and a Z-ion corresponding to sialyl-Tn antigen (Neu5Ac2-6GalNAc-Ser/Thr, m/z 659) with much lower intensity (Fig. S3A) confirming a core 1 structure. Similarly, fragmentation of the most abundant di-sialylated core 2 ion (m/z 1879) yielded daughter ions with m/z 1504, 1673, 1282 and 1129 confirming a bi-antennary glycan (Fig. S3A). Presence of Lewis X on VECs (Fig. S2) suggests that the fucose is linked to GlcNAc.

To further investigate O-glycan structure, we performed linkage analysis of charged O-glycans from VECs confirming the presence of 3-GalNAc-itol, 3,6-GalNAc-itol, and 4-GlcNAc (Table 1 and Fig. S4), which is consistent with core 1 and core 2 O-glycan structures detected in MS/MS analysis (Fig. S3). Detection of both 3-linked Gal and 6-linked Gal is consistent with α2-3, as well as α2-6 sialic acid decoration. Additionally, the presence of terminal fucose and 3,4-GlcNAc supports the presence of Lewis antigen(s) on VECs.

Analysis of O-glycans isolated from BV vaginal specimens using MALDI-TOF MS revealed much lower intensity (< 30%) of sialyl Lewis X containing core 2 ion (m/z 1879), as compared to No BV O-glycans (Fig. 4A). This is consistent with the depletion of sialic acids seen in O-glycans in women with BV (Fig. 2D). Additionally, corroborating these data we also observed a concurrent higher intensity of a core 2 ion lacking sialic acids (m/z 983) in BV O-glycans. In support of this assignment, MS/MS analysis (by LTQ-MS) of the most abundant O-glycan ion (m/z 983) yielded daughter ions with m/z 747 (Z-ion), 520 (Y-ion, T-antigen), and 486 (B-ion, LacNAc) confirming the presence of a core 2 structure with LacNAc on one antenna and T-antigen on the other (Fig. S3A). Further, MS/MS analysis of the precursor ion with m/z 1157 yielded several fucose containing daughter ions with m/z 921 (most abundant, Fucose{Gal-GlcNAc}), 694 (Fucose-Gal-GalNAc with reducing end) and 660 (Fucose{Gal-GlcNAc}), suggesting the presence of multiple fucosylated glycoforms (Fig. S3B). In order to verify if these changes in the molecular ion intensity were due to de-sialylation of glycans, we analyzed O-glycans isolated from VECs from women without BV that were pretreated with A.u. sialidase (Fig. 4A). We found that the MALDI-TOF O-glycoprofile of sialidase pretreated VECs was similar to that obtained from women with BV. The highest peak in the O-glycans of sialidase pretreated samples corresponded to the desialylated core 2 ion (m/z 983), while the relative intensity of molecular ion m/z 1879 (sialic acid containing core 2 ion) was less than 30%. These data confirm that there is a decrease in abundance of sialylated epithelial O-glycans in BV, and suggest a possible role of sialidases in the alteration of vaginal epithelial glycan landscape.

Exposure of carbohydrates underlying sialic acids on vaginal epithelial cells in BV

Sialic acids are most often present at the terminal position of glycans and can protect the underlying sugars from degradation by glycoside hydrolases^73,74. In both N- and O-linked glycans, sialic acids often cap galactose (Gal) residues attached to GlcNAc or GalNAc. To evaluate the impact of BV on such glycans, if any, we evaluated terminal Gal residues on the vaginal cell surface using peanut agglutinin (PNA), a lectin that recognizes T antigen (Galβ1-3GalNAc) when the Gal is not capped by sialic acids (Fig. 5A)⁷⁵. To distinguish if the PNA epitopes on vaginal epithelial surfaces were masked by sialic acids or simply not present at all, PNA binding was compared between cells from the same sample that were either untreated or pretreated with exogenous sialidase (A.u. sialidase). We observed minimal PNA binding to untreated VECs in No BV samples (Fig. 5B). However, a significant increase in PNA binding was observed when the VECs from the same sample were treated with sialidase. This suggests that the PNA epitope is present on No BV VECs, but is typically masked by sialic acids. In striking contrast, cells from BV-positive women showed considerable PNA binding even in the absence of sialidase treatment (Fig. 5B). Furthermore, PNA binding to VECs from women with BV did not increase upon sialidase treatment, suggesting that the PNA epitopes are mostly not masked by sialic acids in BV. These data corroborate the analytical observation of hypo-sialylated epithelial glycans in BV, and strongly support the conclusion that uncapped Gal residues (as part of the T antigen) are present on epithelial cell surface in BV. Control experiments confirmed PNA selectivity and ruled out non-specific binding (Fig. S5). As expected, PNA binding to sialidase-treated epithelial cells from BV-positive women could be out-competed with excess lactose (Galβ1-4-Glc), but not with sucralose. Notably, counter-staining with DAPI revealed that the cells from BV-positive women had large numbers of bacteria covering their surfaces (Fig. S6), lending them a pink/purple coloration (Fig. 5C). In contrast, the A.u. sialidase treated cells from BV-negative women had few adherent bacteria and hence stained red. The BV epithelial cells coated in bacteria are reminiscent of clue cells, a diagnostic feature of BV also observed by wet mount microscopy²⁸.

• Download figure
• Open in new tab

Fig. 5: Exposed galactose on the vaginal epithelial cell surface in BV.

(A) Schematic: PNA-Rhodamine (PNA-Rh) binds to galactose residues, not masked by sialic acids, on VECs treated with or without sialidase. A.u. sialidase = sialidase from Arthrobacter ureafaciens. (B) Representative confocal images of PNA (Red) stained VECs (n=3 individual specimens either with BV or without BV) pretreated with exogenous sialidase (+A.u. sialidase) or with buffer alone (-A.u. sialidase, untreated control). Nuclei and bacterial cells (Blue) are stained with DAPI. Images shown are representative of multiple fields of view for each specimen. Scale bars = 30 μm. (C) Representative image of a BV epithelial cell with surface bacteria observed as blue puncta close to the cell membrane redolent of clue cells. Scale bars = 10 μm. (D-E) Flow cytometry quantification of PNA binding to BV and No BV VECs pretreated with exogenous sialidase (+A.u. sialidase) or with buffer alone (untreated control). (D) Representative histogram overlays are shown from n=6 individual specimens. (E) Estimation of percentage cell surface exposed galactose residues, calculated as percentage ratio of mean fluorescence intensity (MFI) of PNA binding to VECs from the same sample in the absence versus presence of A.u. sialidase (n=15 samples per group). When possible, data points on the graph represent epithelial cells from individual women (n=22, black/blue circles). However, some data points represent pooled samples from 2-5 women (when recovery yielded insufficient cell quantity for flow cytometry) (n=8 pools, circles filled with red). Data in A-E was combined from 3 independent experiments. Error bars on the graph show standard deviation for each group. Statistical analysis by Mann–Whitney U, ****P<0.0001. The contingency table below the graph shows sensitivity and specificity values determined using two-sided Fisher’s exact test, with a galactose exposure threshold of 60% compared to the Nugent method. A total of N=46 specimens from individual women were used to generate the data in A-E. See methods for pooling rationale.

• Download figure
• Open in new tab

Fig. 6: Epithelial cells emulate BV phenotypes when treated with Gardnerella sialidases.

(A-D) Cells from the same pool of VECs, from women with or without BV, were treated with (i) vector control (Control) or (ii) commercial Arthrobacter ureafaciens sialidase (A.u.) or (iii) recombinant Gardnerella NanH2 sialidase (NanH2) or (iv) recombinant Gardnerella NanH3 sialidase (NanH3) as indicated in the labels. (A) Fluorimetric quantification of sialic acid (Neu5Ac) released from No BV VECs treated with either vector control or sialidase enzymes. (B, C) Flow cytometric analysis of PNA binding to VECs. Galactose exposure was assessed by comparing PNA binding to A.u. sialidase treated cells from the same pool. (B) Galactose exposure on No BV VECs. (C) Galactose exposure on BV VECs (n=3 pools, with 2-3 specimens in each pool). (D) Representative confocal images of VECs from women with and without BV stained with PNA (Red) lectin under each condition. Nuclei and bacterial cells (Blue) are stained with DAPI. Each row shows analysis of one pooled specimen. Scale bars are 50 µm. Data in A-D combined from 2 independent experiments. For No BV groups treated with vector control or A.u. sialidase or G.v. NanH2 sialidase - n=10 pools, with 2 – 5 specimens in each pool; for the No BV group treated with G.v. NanH3 sialidase - n=8 pools,, with 2 – 5 specimens in each pool. **p<0.01 (Wilcoxon signed rank test). A total of N=38 specimens from individual women were used to generate the data in A-D. Data in 7A, 7B, and 7D is from the same No BV VEC pools. Data in 7C and 7D is from the same BV VEC pools. See methods for pooling rationale.

To further quantitatively assess the amount of cell surface galactose exposure on epithelial cells from women with and without BV, we measured PNA binding to VECs using flow cytometry. Analysis of the same gated population of cells (Fig. S7) from BV and No BV samples showed that PNA binding to cells from BV-positive women was higher compared to the No BV controls in the absence of any exogenous sialidase treatment (Fig. 5D). In contrast, epithelial cells from all women were labeled with PNA after sialidase treatment. The extent of exposure of Gal residues in each group was calculated as the percentage ratio of PNA binding to cells in absence versus presence of treatment with A.u. sialidase. Using this metric, the differences between BV-negative and BV-positive samples became even more striking. VECs from BV-positive women had a significantly higher (p<0.0001) portion of uncapped Galβ1-3GalNAc (PNA recognized) epitopes compared to cells from BV negative women (Fig. 5E).

• Download figure
• Open in new tab

Fig. 7: The epithelial glycocalyx is largely absent in BV and Gardnerella sialidase diminishes the appearance of the glycocalyx.

Representative transmission electron microscopy (TEM) images of VECs from (A) No BV specimens - glycocalyx appears as a fuzzy layer close to epithelial cell membrane (indicated with arrowheads), (B) BV specimens, and (C) No BV specimens - treated with recombinant Gardnerella NanH2 sialidase. (D, E) Abundance of the glycocalyx was scored on a 0-4 scale, where four is “abundantly present” and zero is “none visible”. A total of ten TEM images were scored from each specimen (scoring rubric in Fig. S8). Data points in the graph represent the average of three scores of each image. For (i) No BV; n=40 images from n=4 specimens (same data in shown in D and E), (ii) BV; n=30 images from n=3 specimens, and (iii) No BV treated with NanH2; n=30 images from n=3 specimens. VECs from three of the No BV specimens were divided in half; one half was treated with recombinant Gardnerella NanH2. Images were acquired in a blinded fashion and scored by three observers blinded to the status of the sample. All images were acquired at 25,000 X. Scale bars are 500 nm. ****p<0.0001 (Kruskal-Wallis with Dunn’s multiple comparisons test). A total of N=7 specimens were used to generate these data.

As noted previously⁵⁴, samples from individual BV-positive women also had higher sialidase activity (Fig. S7). Correlation analysis, using data from individual un-pooled samples, revealed a direct relationship between the level of sialidase activity and the percentage of exposed galactose on the epithelial cell surface in BV-positive women (Fig. S7), analogous to the results presented in Fig 2A. Furthermore, in most of the BV samples with high sialidase activity (relative activity >0.5) more than 90% of galactose was found to be exposed in VEC glycans. Interestingly, in a small proportion of BV samples, high galactose exposure (>60%) was also observed on cells from vaginal swabs with very low or no detectable sialidase activity. This suggests that the phenotype of higher PNA binding to VECs may be a more sensitive indicator of BV compared to sialidase, persisting even when sialidase activity (sialidase) is not detectable. In fact, analysis of the sensitivity and specificity of PNA-binding as a potential diagnostic was promising. Based on the distribution of data from samples in both patient groups, we considered 60% galactose exposure as a threshold and compared the data to the Nugent’s method⁶⁶. Using this small sample set, we found that epithelial cell-PNA-binding analysis had a specificity of 100% and sensitivity of 80% (Fig. 5E), suggesting this may be a useful area for diagnostic development.

Gardnerella sialidases hydrolyze sialic acids from vaginal epithelial cells and generate BV-like glycan phenotypes

So far we have employed a commercially available recombinant sialidase from A. ureafaciens as a control, to demonstrate the potential impact of exogenous sialidase on glycan phenotypes. Next we tested if the more relevant sialidases from Gardnerella, often one of the most abundant microbes in BV, could generate BV-like features of glycans.

Specifically, we investigated if Gardnerella sialidases degrade glycan components of the epithelial surface, hydrolyzing sialic acids and yielding underlying Galβ1-3GalNAc epitopes recognized by PNA. Briefly, pools of epithelial cells from women without BV were treated with commercial A.u. sialidase or recombinant enzymes from Gardnerella (NanH2 or NanH3)⁶². In parallel, a proportion of VECs from the same pool were mock treated using material derived from identically prepared E. coli containing the empty vector. Sialic acids hydrolyzed from epithelial cells were measured by fluorescent derivatization and HPLC as described earlier for analysis of sialylated N- and O-glycans. Cells were also subjected to PNA staining. The data show that Gardnerella sialidases NanH2 and NanH3 were adept at removing sialic acids from epithelial cells (Fig. 6A), revealing PNA-reactive underlying Galβ1-3GalNAc epitopes measured by flow cytometry (Fig. 6B). Consistent with findings in Fig. 5, BV samples analyzed in parallel had high levels of PNA binding that could not be further increased with treatment by any of the sialidases (Fig. 6C). Confocal imaging of PNA-stained cells further underscored the generation of a BV-like phenotype when No BV samples were treated with Gardnerella sialidase (Fig. 6D). As observed earlier (Fig. 5C), cells from BV-positive samples appeared pink/purple in coloration due to adherent bacteria that were stained blue with DAPI, whereas sialidase treated cells from BV-negative samples had red coloration. Overall, these data suggest that sialidases produced by BV-associated bacteria (e.g. Gardnerella) result in degradation of vaginal epithelial glycans.

The glycocalyx is largely absent in BV and is diminished by Gardnerella sialidase

Finally, we used classical methods to assess the impact of BV on the vaginal epithelial glycocalyx. VECs were prepared for transmission electron microscopy (TEM) by staining with ruthenium red and osmium tetroxide. Ruthenium red is a polycationic dye that stains acidic polysaccharides and therefore can facilitate visualization of sialic acid containing and other negatively charged glycans on cell surfaces in conjunction with osmium tetroxide, which attaches to vicinal dihydroxylated molecules. Micrographs were collected by a single blinded microscopist and scored by three blinded observers, using the scoring method described in Fig. S8, from zero (“none visible”) to four (“highly abundant”). As observed in initial studies (Fig. 1), compared to women without BV glycocalyx was diminished in women with BV (Figs. 7A, B, and Fig. S8). A striking 95% of the epithelial images from women without BV were scored as having a visible glycocalyx (38/40 score >1). In contrast, epithelial cells of BV specimens were devoid of glycocalyx ~63% of the time (19/30 had a 0 score from all three observers) (Fig. 7D).

In parallel with the TEM analysis of glycocalyx in women with and without BV, we also scored blinded images of No BV VECs that had been treated with recombinant purified Gardnerella sialidase (NanH2)⁶². We observed that the treatment of “normal” (No BV) VECs with Gardnerella sialidase led to a significantly diminished glycocalyx. Similar to what was observed in women with BV (Figs. 7B, D and Fig. S8), NanH2 led to a sharp reduction in the appearance of the glycocalyx (Figs. 7C, E). Cells treated with Gardnerella sialidase had a glycocalyx score less than two in 29/30 (~97%) images (Fig. 7E). In, contrast, the untreated group had a score of 2 or greater in 32/40 (~80%) images.

Vaginal specimen source

Previously collected vaginal swabs leftover from the inaugural contraceptive CHOICE project at Washington University¹⁰² were used in this study. CHOICE received informed consent in writing from all participants at enrollment, as well as permission to use vaginal samples for future studies. This study was reviewed and approved by the Institutional Review Board at Washington University School of Medicine (IRB ID # 201108155). CHOICE inclusion criteria were: reported sexual activity in the past six months or anticipated sexual activity with a male partner and the desire to prevent pregnancy by use of contraception. Exclusion criteria were ages outside the 14-45 year age range and history of tubal ligation or hysterectomy, as previously described¹⁰².

Specimen collection, handling, and Nugent scoring

Upon enrollment, women were given a double-headed rayon swab inside a collection tube (Starplex Scientific Inc, Etobicoke, Ontario, Canada) and were instructed on a self-collection protocol. Immediately after collection, a single technician rolled one swab onto a glass slide. Remaining swab material was stored at −80 °C. Slides were heat-fixed, Gram stained and subjected to Nugent scoring. In general, the Nugent scoring⁶⁶ system uses weighted quantitation of bacterial morphotype to reflect the overall character of vaginal bacteria present. Nugent scores are reflected on a scale of 0 to 10 combined from three components for each woman: 1) a score of 0-4 reflecting the presence of rod-shaped Gram-positive lactobacilli where 0 indicates highest numbers, 2) a score of 0-4 reflecting the presence of Gram-negative and Gram-variable morphotypes where 4 indicates the highest numbers, and 3) a score of 0-2 reflecting the presence of curved rods. A woman with score of 0-3 is considered not to have BV (without BV/No BV, “normal”), and a woman with score of 7-10 is considered BV-positive.

Specimen selection

A total of 258 specimens from individual women were used in the experiments reported here. Inclusion criteria for this sub-study were an available specimen Nugent score of 0-3 (No BV) or 7-10 (BV). Samples represent a cross-section of women from the same approximate (2 year) collection time window of CHOICE enrollment. Demographic characteristics of the participants in this sub study are presented in Table S1. We did not exclude specimens from women who were pregnant, menstruating, or those with sexually transmitted infections.

Specimen processing

For all analyses of vaginal epithelial cells, vaginal swabs were selected based on Nugent score from reproductive age women with BV (Nugent score 7-10) or without BV (Nugent score 0-3). No attempts were made to exclude swabs from women who had clinical indications of sexually transmitted infection, although rates were very low. Vaginal swabs were stored at −80 °C until processed further. Swabs were cut to approximately ~3.5-4.0 cm and enclosed in a cryotube. Swabs were removed from the freezer onto wet ice and eluted by immersing in 1 mL of 100 mM sodium acetate (pH 5.5, swab elution buffer (SEB) incubated in an ice bucket for 1 h followed by shaking at 500 rpm, using a IKA MS 3 digital, for 10 min at room temperature. The individual swab eluates were transferred to microcentrifuge tubes and kept on ice. To check for sialidase activity, 50 µL of eluate was resuspended and transferred to 96 well round bottom black polypropylene plates. Sialidase assay was performed as described below. In parallel, a second elution was performed by again immersing the same swabs in 1 mL of SEB in individual cryo tubes, followed immediately by shaking at 500 rpm for 10 min at room temperature. Cells were isolated from the swab eluates by centrifugation at 1000 x g for 5 min at room temperature after each elution step. Microscopic examination of the isolated cells confirmed the morphology as polygonal squamous epithelial cells with a diameter of 50-70 µm and no other morphologies were observed.

Epithelial cells obtained from vaginal swabs were processed individually for detection of glycocalyx by transmission electron microscopy (TEM) and sialic acid analysis by MAL-II staining. For lectin analyses which required larger numbers of cells, like quantification of PNA binding by flow cytometry, VECs were pooled from multiple swabs in the same group (Nugent 0-3 or Nugent 7-10) to have sufficient packed cell pellet volume (~50 – 100 µL). For analytical studies focused on N- and O-glycan characterization, frozen VEC pellets were sent to the University of California San Diego GlycoAnalytics core. An additional selection criterion was applied to these samples: swabs in the BV group with detectable high sialidase activity (relative sialidase activity >3 for most specimens) as estimated by sialidase assay on the first swab eluate, were pooled together. For each study condition, approximately 100-200 µL of packed cell volume was obtained by pooling VECs isolated from 5 or 10 vaginal swab eluates in the same group.

Fluorometric assay for sialidase activity

To measure sialidase activity, 50 µL of vaginal swab eluate was diluted 3-fold into 100 mM sodium acetate, pH 5.5, containing 200 µM 4-methylumbelliferyl-Neu5Ac (4-MU-Sialic acid) (Goldbio) in a 96 well round bottom black polypropylene plates and sealed with an optical clear film. Substrate hydrolysis was monitored using the fluorescence of 4-MU (Excitation 365 nm, Emission 440 nm) in a Tecan M200 plate reader every 2 min for 2 h at 37 °C.

Preparation of Gardnerella sialidases NanH2 and NanH3

Gardnerella NanH2 and NanH3 (truncated versions lacking the transmembrane segments of both proteins and the signal sequence of NanH2), along with the vector control, were expressed in E. coli and purified by nickel chromatography as previously described⁶². In experiments with recombinant Gardnerella sialidases NanH2/NanH3, and sialidase from Arthrobacter ureafaciens (A.u. sialidase), enzyme activities were normalized as described earlier⁶². Similar amounts of enzyme activity were used to investigate their ability to cleave sialic acid from VECs.

Visualization of glycocalyx on VECs by transmission electron microscopy (TEM)

VECs eluted from individual vaginal specimens of women with or without BV were washed with phosphate buffered saline (PBS) and processed for glycocalyx staining. For treatment of VECs with recombinant Gardnerella sialidase NanH2, cells obtained from individual specimens of women without BV were divided equally into two microcentrifuge tubes such that one half was treated with NanH2 in 300 µL SEB, and the other was mock treated with the same amount of buffer alone. Cells were kept at 37 °C on a rotor. After 1 h, cells were pelleted by centrifugation at 2000 g for 5 min and washed with 1 mL PBS twice. VECs were processed for glycocalyx detection using ruthenium red and osmium tetroxide (OsO₄) as described by Fassel and Edmiston⁶⁵. Briefly, VECs isolated from swab eluates were fixed and stained using 100 mM sodium cacodylate buffer (Polysciences Inc., Warrington, PA), pH 7.2, containing 0.15% ruthenium red, 2% paraformaldehyde and 2.5% glutaraldehyde for 1 h at RT. VECs were then washed with 100 mM cacodylate buffer containing 0.15% ruthenium red. Secondary fixation was performed with 1% OsO₄ and 0.15% ruthenium red in 100 mM cacodylate buffer for 1 h (Polysciences Inc.). Samples were then rinsed extensively in deionized water (dH2O) prior to en bloc staining with 1% aqueous uranyl acetate for 1 h (Ted Pella Inc., Redding, CA). Following several rinses in dH20, samples were dehydrated in a graded series of ethanol and embedded in Eponate 12 resin (Ted Pella Inc.). Sections of 95 nm were cut with a Leica Ultracut UCT ultramicrotome (Leica Microsystems Inc., Bannockburn, IL), stained with uranyl acetate and lead citrate, and viewed on a JEOL 1200 EX transmission electron microscope (JEOL USA Inc., Peabody, MA) equipped with an AMT 8 megapixel digital camera and AMT Image Capture Engine V602 software (Advanced Microscopy Techniques, Woburn, MA). For glycocalyx scoring, the microscopist and scorers were blinded to the experimental treatment, as well as the BV status of the sample. 10 images of the epithelial cell membrane were acquired from each specimen at 25,000 X magnification. Images were labelled randomly to facilitate the blinded glycocalyx scoring. Images were scored between zero to five based on the abundance of glycocalyx, with zero being least abundant or nearly absent and five being highly abundant. Data from three independent scorers was combined to compute the average glycocalyx abundance score for each image.

Lectin staining of vaginal epithelial cells

Cells were isolated from vaginal swab eluates of women with and without BV as described above. Cells were derived from individual women whenever possible; when cell numbers were insufficient for flow cytometry, cells from 2-5 individuals were pooled for analysis. Cells were washed three times with PBS containing 0.08% sodium azide (wash buffer). Prior to lectin staining, cell pellets for each sample (with ~50 µL packed cell volume) was re-suspended in 1 mL wash buffer, divided equally into 2 tubes and centrifuged at 2000 x g for 2 min. Supernatants were removed, pellets were resuspended in 300 µL of SEB (untreated controls) or SEB containing 15 mU of exogenous sialidase from Arthrobacter ureafaciens (A.u. sialidase) (EY Labs). Cells were incubated at 37 °C on a rotor to avoid clumping. After 1 h, cells were washed with 1 mL of lectin staining buffer (LSB) (10 mM HEPES, pH 7.5, 0.15 M sodium chloride and 0.1 mM calcium chloride). Cells were stained with peanut agglutinin (PNA) by incubating with 10 µg/mL of PNA-rhodamine (PNA-Rh) (Vector Lab) in 300 µL LSB in an ice bucket with intermittent mixing (every 10-15 min). After approximately 1 h, PNA stained cells were washed with LSB, re-suspended in 150 µL LSB and analyzed by flow cytometry. About 10,000 events were acquired for the same gated population for each sample (see Fig. S7A>). Untreated cells were used to assess auto fluorescence; PNA-Rh binding was analyzed in the FL-2 channel on a BD FACSCalibur flow cytometer. Exposed galactose was calculated as a percentage of PNA-binding mean fluorescence intensity (MFI) observed to untreated cells compared to the PNA-binding MFI to VECs from the same sample pretreated with A.u. sialidase.

For MAL-II staining, cells from individual specimens were incubated with 5 µg/mL of MAL-II biotin (Vector Laboratories) in 300 µL LSB in an ice bucket with intermittent mixing (every 10-15 min). After ~1 h, cells were washed with an avidin staining buffer (ASB) (1% BSA, 0.05% NaN₃ in PBS). To detect biotin, cells were then stained with 2 µg/mL Neutravidin-OregonGreen 488 in 300 µL ASB for ~45 min at 4 °C degrees in an ice bucket. Cells stained with secondary only were included as control. Finally, the MAL-II stained cells and controls were washed with PBS and processed further for imaging.

For confocal microscopy, lectin-stained cells were re-suspended in PBS containing 600 nM DAPI (Invitrogen) and incubated in an ice bucket for 15 min. 10 µL of the cell suspension was then transferred to a microscope slide, covered with a cover slip and sealed with a transparent nail polish. Images were captured from at least three independent fields for each sample using LSM 880 confocal laser scanning microscope (Carl Zeiss) using 40 X objective lens with immersion oil.

Detection of Lewis X (CD15) on VECs

Cells were isolated from vaginal swab eluates of women without BV as described above. Cells were washed three times with PBS containing 0.08% sodium azide (wash buffer). Prior to staining with anti-CD15, that recognizes Lewis X antigen, the cell pellet for each sample was resuspended in 1 mL wash buffer, divided equally into 2 tubes and centrifuged at 2000 x g for 2 min. Supernatants were removed, and the pellet was resuspended in 300 µL of SEB or SEB containing 15 mU of A.u. sialidase (EY Labs). Cells were incubated in this buffer at 37 °C on a rotor to avoid clumping. After 1 hour, cells were washed with 1 mL of LSB. To block non-specific binding, cells were incubated with 5 µg/mL of Fc Block (BD, 564220) in 600 µL of LSB for 15 min on a rotor at room temperature. Next, each specimen was divided equally into three tubes, 200 µL each, for staining with either 5.6 µg/mL secondary antibody alone (anti-mouse IgM-AlexaFluor488, Jackson ImmunoResearch) or 5 µg/mL of isotype control (mouse IgM, BD 555581) or 5 µg/mL of primary antibody (anti-Lewis X (Human CD15), BD 555400, clone HI98 (also known as HIM1)) followed by the secondary antibody. Cells stained with anti-Lewis X or isotype control were incubated in an ice bucket for 1 h with intermittent mixing (every 10-15 min). After washing the cells with 1 mL of LSB, a second staining step was performed in a similar manner by adding the secondary antibody to all the samples. After 40 min cells were washed two times with LSB, re-suspended in 150 µL LSB and analyzed by flow cytometry. Unstained cells were processed similarly in parallel without any antibody staining. About 10,000 events were acquired for each sample; anti-Lewis X binding was analyzed in the FL-1 channel on a BD FACSCalibur flow cytometer. Secondary alone showed little binding to the VECs and overlapped with unstained controls (data not shown). For confocal microscopy, samples were processed and imaged as described above.

Preparation of crude cell lysate for glycan extraction

All analyses of VEC N- and O-glycans were performed at GlycoAnalytics Core at the University of California San Diego, California, USA. VECs isolated from vaginal swab (100 µL of packed cell pellet) were re-suspended in 300 µL of ultrapure water (Invitrogen) with 5 µL EDTA free protease inhibitor (EMD-Millipore Corp.) and sonicated in an ultrasonic water bath (Fisher Scientific Sonicator Model FS-9) until it formed a uniform suspension in water. Protein was quantified using BCA assay and further extraction of glycans was performed using a defined amount of protein as described in the methods below.

Isolation and purification of N-glycans

N-linked glycans were isolated from glycoproteins in the crude cell lysate by overnight incubation with glycerol free recombinant PNGase-F (P0709S, NEB) at 37 °C. Briefly, cell extract for 500 µg of protein was mixed with a denaturing buffer (NEB kit) and boiled at 100 °C for 15 min. Proteins were solubilized by incubation with NP-40 for 1h at room temp, with intermittent vortexing every 15 min. N-glycans were extracted using PNGase-F and purified using a tandemly connected Sep-Pak C18 and porous graphitized carbon (PGC) cartridge equilibrated in water. PGC bound glycans were eluted with 30% acetonitrile containing 0.1% trifluoroacetic acid (TFA) in water and then lyophilized. The purified pool of glycans were processed further as required for each analysis.

Isolation and purification of O-glycans

To release O-linked glycans by reductive β-elimination, crude cell lysate for 500 µg of protein was incubated with 0.05 M NaOH / 1 M NaBH₄ at 45 °C for 16 h with constant stirring. Samples were then neutralized using ice-cold 30% aqueous acetic acid, followed by removal of sodium salts using cation exchange resin (Dowex 50Wx8, H⁺ Form, Bio-Rad 142-1451). The samples were co-evaporated using a 9:1 (v/v) methanol/acetic acid mixture and with absolute methanol to remove borate. The extracted O-glycans were purified using Sep-Pak C18 column, where the flow through from Sep-Pak C18 was lyophilized and used for further analysis¹⁰³. Additionally, for linkage analysis by gas chromatography-mass spectrometry (GC-MS), charged O-linked glycans were separated (described below) from neutral O-glycans to reduce interference from contaminating neutral saccharides in the vaginal specimens.

Sialic acid analysis on N- and O-glycans by reverse phase-liquid chromatography

For sialic acid analysis, glycans derived from 25 μg of protein extract were heated to 80 °C in 2 M acetic acid for 3 h. The released sialic acids were collected by ultrafiltration through a 3,000 MWCO filter and derivatized with 1,2-diamino-4,5-methylenedioxybenzene (DMB) (Sigma Aldrich), as described earlier¹⁰⁴. The DMB-derivatized sialic acids were eluted isocratic with acetonitrile (8%) and methanol (7%) in Milli-Q^® water and analyzed either by HPLC using Acclaim 120 C18, 5 µM column (4.6 x 250 mm, Thermo-Dionex) at a flow rate of 0.9 mL/min over 30 min or using an Acquity™ UPLC system, with fluorescence detector set at λ_ex 373 nm, λ_em 448 nm. For UPLC, samples were eluted over 10 min using Acquity UPLC® BEH C18, 1.7 µM column (2.1 x 50 mm, Waters) at a flow rate of 0.2 mL/min. The DMB-derivatized sialic acids were identified and quantified by comparing elution times and peak areas to known standards using Chromeleon^® software. Data shown is from two independent experiments that were conducted ~three years apart. For the first experiment, VECs from 10 specimens were combined to obtain each pool as a greater biomass was required to complete all the analytical studies (shown in Fig. 2–4) on the same set of samples. The second experiment was conducted to specifically validate the preliminary findings of this experiment using a larger set of samples, for which VECs from 5 specimens were combined to obtain each pool.

2-Aminobenzamide (2-AB) labeling of purified N-glycans

The purified N-glycans were labeled with 2-AB as described earlier¹⁰⁵. Briefly, samples were dissolved in 10 μL solution of 0.44 M 2-AB in acetic acid-DMSO mixture (35:65 v/v) containing 1 M sodium cyanoborohydride. Samples were then incubated at 65 °C for 2.5 h. 2-AB labeled glycans were purified using GlycoClean S cartridge (GLYKO) following the manufacturer’s protocol.

2-AB profiling with High-Performance Anion-Exchange Chromatography (HPAEC)-Fluorescence detection

Profiling of 2-AB labeled glycans was obtained using a Dionex CarboPac^® PA1 (4 X 250 mm) anion exchange column along with a guard column (4 X 50 mm) at flow rate of 1 mL/min. Glycans were separated in 100 mM sodium hydroxide with a sodium acetate gradient of 0-250 mM in 0-75 min. Data was collected using the Dionex ICS-3000 HPLC system with Ultimate 3000 fluorescence detector (Dionex) set at λ_ex 330 nm at λ_em 420 nm, with sensitivity 7. The data was processed using Chromeleon™ software (Thermo Scientific). Glycans from RNAse B and bovine fetuin were processed in parallel and labeled as known standards with high mannose and complex type N-glycans.

Separation of O-glycans by charge

Anionic glycans were captured using liquid chromatography-Amino (LC-NH₂) cartridges (Supelco, Part no. 504483, 1 mL). Briefly, total O-glycans (reconstituted in water after purification as described above) were loaded on the cartridge and the column was washed with 3 mL of ultrapure water (Invitrogen) which was collected as flow-through. Anionic species were eluted with 2 mL of 100 mM ammonium acetate and lyophilized.

Per-O-methylation of glycans

Purified O-glycans, dried in glass reaction tubes, were dissolved in 350 µL of anhydrous DMSO by stirring for 1 h. To this solution was added a freshly prepared slurry (50 µL) of powdered NaOH in anhydrous DMSO, followed by immediate addition of 500 µL of methyl iodide. This reaction mixture was stirred vigorously at room temperature for 40 min. The reaction was quenched by adding 1 mL of ice-cold water followed by extraction of methylated glycans using chloroform. Finally, the chloroform was dried using a stream of nitrogen.

MALDI-TOF profiling of O-glycans

Permethylated glycans were dissolved in absolute methanol and mixed 1:1 with super-DHB matrix solution and spotted on the matrix-assisted laser desorption/ionization (MALDI) plate. The samples were air-dried and analyzed in positive-reflectron mode on a MALDI-qTOF mass spectrometer (Applied Biosystems QSTAR)¹⁰⁶. Structures for were manually annotated and verified using the glycobioinformatics tool GlycoWorkBench^107,108. The proposed assignments for the selected peaks were based on the knowledge of mammalian glycan biosynthesis pathways. The proposed structures were confirmed using MS/MS and GC-MS linkage analysis as described below.

MS/MS analysis of O-glycans

All spectra were acquired in the positive ionization mode over a range of 250-2000 amu on LTQ-XL Orbitrap Discovery Electrospray Ionization Mass Spectrometer (Thermo Scientific). Permethylated glycans were dissolved in absolute methanol, followed by addition of 1mM sodium hydroxide (NaOH), and direct injection into the mass spectrometer with a constant flow rate of 5 mL/min. The ion spray voltage was optimized at 3.7 kV with a sheath gas flow rate 12 mL/min. All MS and MS(n) spectra were collected in positive ionization mode with resolution of 30000. The other instrument parameters for positive-ion detection were: capillary temperature 145 °C; capillary voltage 47 V. Tandem mass spectrometry (MSn) was performed via collision-induced dissociation (CID) on selected parent ions with collision energy of 15-20%.

Glycan linkage analysis by GC-MS

To determine linkage positions of vaginal epithelial O-glycans, permethylated glycans (permethylation described above) were hydrolyzed with 4 N TFA at 100 °C for 4 h, followed by removal of acid using nitrogen flush. The sample was then reduced using ~5 mg sodium borodeuteride (NaBD₄) solution in 1 M ammonium hydroxide, and acetylated using pyridine and acetic anhydride (1:1 v/v) mixture at 100 °C for 1 h¹⁰⁹. Partially methylated acetylated alditol (PMAA) derivatives were analyzed by GC-MS^72,110 using Restek-5ms capillary column (30 m x 0.25 mm x 0.25 µm) on an Agilent Technologies 7820A GC System 5975 Series Gas Mass Selective Detector (GC-MS). Identifications are achieved by using a combination of retention times and the mass fragmentation pattern in EI-MS mode.

Lectin analysis of VECs treated with Gardnerella sialidases

Prior to lectin staining, the cell pellet for each pooled sample (with ~100 µL of packed cell volume) was re-suspended in 1 mL wash buffer, divided equally into 4 tubes and centrifuged at 2000 x g for 2 min. Supernatants were removed, and the pellet was resuspended in 300 µL of SEB or SEB containing either similar amounts of enzyme activity of NanH2, NanH3, A.u. sialidase or similar volume of preparation from vector control. Cells were incubated in this buffer at 37 °C on a rotor to avoid clumping. After 1 h, supernatants were collected for sialic acid analysis by DMB-HPLC, and cells were washed with 1 mL of LSB. For staining with PNA, cells were incubated with 5 µg/mL of PNA-AlexaFluor647 (AF647) in 300 µL LSB in an ice bucket with intermittent mixing (every 10-15 min). After approximately 1h, PNA stained cells were washed with LSB, re-suspended in 150 µL LSB and analyzed by flow cytometry. About 10,000 events were acquired for each sample; PNA-AF647 binding was analyzed in the FL-4 channel on a BD FACSCalibur flow cytometer. For confocal microscopy, lectin-stained cells were processed for imaging as described above.