The following materials were purchased from the manufacturers indicated: HS, chondroitin sulfates A, B, and C (CS A, CS B, and CS C), Heparinases I and III, chondroitinase ABC, fluorescein isothiocyanate (FITC), and phospholipase C phosphatidylinositol-specific (PI-PLC) from Bacillus cereus, all from Sigma-Aldrich (St. Louis, MO, USA); 90% DMEM/F-12 + GlutaMAX™ [Dulbecco's Modified Eagle's medium F-12 Nutrient Mixture (Ham)], fetal bovine serum, penicillin-streptomycin, insulin, EGF (epidermal growth factor recombinant protein), and PBS (phosphate-buffered saline), from Gibco-Thermo Fischer Scientific (Waltham, MA, USA); HiTrap Desalting column and superose 12 from GE Healthcare Life Sciences (Little Chalfont, UK); Brain-Heart Infusion broth (BHI), from Pronadisa (Madrid, Spain); RNeasy Kit and RNase-Free DNase, from Qiagen (Hilden, Germany); High-Capacity cDNA Reverse Transcription Kit and PowerSYBR Green PCR Master Mix, from Applied Biosystems (Foster City, CA, USA); synthetic peptides were from Abyntek Biopharma (Derio, Spain); mouse monoclonal anti-syndecan 1 (CD138), from DakoCytomation (Carpinteria, CA, USA); and rabbit anti-syndecan 2, goat anti-syndecan 3, and rabbit anti-syndecan 4 polyclonal antibodies, from Santa Cruz Biotechnology (Santa Cruz, CA, USA).

The bacterial species used in this study were S. aureus (16491464), Staphylococcus epidermidis (16546925), Streptococcus pyogenes (16598698), Streptococcus pneumoniae (16171067), Escherichia coli (16553780), Kebsiella pneumoniae (16589956), Serratia marcescens (16583077), Neisseria gonorrhoeae (50264947), Pseudomonas aeruginosa (16531445), and Haemophilus influenzae (16091813), all of which were clinical isolates obtained from the Hospital Universitario Central de Asturias. All the bacteria were grown in Brain Heart Infusion (BHI broth) at 37°C in a shaking incubator, except S. pneumoniae, N. Gonorrhoeae, and H. Influenzae, which were grown in a 5% (v/v) CO₂ atmosphere without shaking.

HCE-2 [50.B1] ATCC CRL-11135 corneal cell line was grown in Dulbecco's Modified Eagle's minimal essential medium (DMEM). The culture broth was supplemented with 10% (w/v) fetal bovine serum, 1.5 units/ml of insulin, 10 ng/ml of epidermal growth factor (EGF), and with penicillin G/streptomycin (5000 IU/ml, 5000 μg/ml). Cultures were incubated at 37°C in a 5% (v/v) CO₂ atmosphere.

FITC labeling was performed on overnight cultures washed four times with PBS and resuspended in a 0.1 mg/ml FITC solution to an A₆₀₀ of 0.5; incubation in the dark at 37°C under agitation proceeded for 1 h and the bacterial suspensions were centrifuged, once again washed four times with PBS to eliminate the FITC excess, and resuspended in PBS to an A₆₀₀ of 0.5.

Adhesion of the different pathogenic bacteria to HCE-2 monolayers was performed in 24-well plates to 70–90% confluence. The media was aspirated, and the cells washed twice with PBS and then blocked with 10% fetal bovine serum in PBS for 2 h at 37°C in a 5% CO₂ atmosphere. After further washing with PBS, 100 μl of FITC-labeled bacteria in 500 μl of PBS buffer was added and the mixture incubated for 1 h at 37°C and 5% CO₂. Next, the wells were rinsed four times with 500 μl PBS to remove any unbound bacteria. At the end of the experiment, HCE-2 cells were disaggregated with 1% SDS and the fluorescence of the pathogens attached to them was quantified in a Perkin Elmer LS55 fluorometer set at 488 nM (excitation) and 560 nM (emission). Data for the different experiments were normalized using the adhesion value without any additive or treatment, which was given the arbitrary value of 1. Assays were performed at least in triplicate and the data are expressed as the mean ± SD.

To analyze adhesion to the ECM, HCE-2 monolayers under the same conditions as described above were lifted with 0.5 mM EDTA in PBS. The ECM (material remaining in the wells after the cells had been lifted) was washed twice with PBS, and each well was visually inspected to ensure that no cells remained. The bacteria were then added and the assays were performed as described above.

The cell cultures in 24-well plates at about 70% confluence were incubated in medium containing rhodamine B 50 μg/ml or genistein 30 μM (final concentration) overnight at 37°C. The cultures were washed twice with PBS, and submitted to adherence assays, as described in the previous paragraph. Efficiency of inhibition was tested by determining the concentration of GAGs in the treated cells vs. the control cells as previously described (Barbosa et al., 2003).

Hydrolysis of HS from cell cultures was achieved by incubation at 37°C for 3 h in a 5% CO₂ atmosphere in DMEM minimal medium with a mix of 500 mU/ml (final concentration) of each of heparinase I and III. Digestion of CS chains was carried out by 3 h of incubation under the same conditions, but using 250 mU/ml (final concentration) of chondroitinase ABC. The hydrolysis of both GAGs was performed through simultaneous incubation with heparinase I and III and chondroitinase ABC under the same conditions. The reactions were stopped with two washes in PBS buffer and the cell cultures were immediately submerged in DMEM and adherence assays with pathogens carried out as described above. The efficiency of digestions was established by determining the concentration of GAGs in treated cells vs. control cells as previously described (Barbosa et al., 2003).

The experiments examining the effect of GAGs on adherence interference were performed through the addition of either HS, CS A, CS B, and CS C, or a mixture of all four, at concentrations ranging between 0.01 and 200 nM to the labeled bacteria before their addition to the monolayers. However, the assays comparing HS of high and low molecular weight (MW), and those examining the differences between poorly and highly sulfated HS domains used a concentration of 10 nM.

The effect of peptides which include consensus heparin binding sequences on adherence interference experiments was initially calibrated using the peptides QKKFKN and FKKKYGKS at concentrations ranging between 1 and 500 nM, and measuring the interference they produced on the binding of Pseudomonas aeruginosa and Stapylococcus aureus. Based on these data, the effect on all pathogens was analyzed at a concentration of 20 nM of each peptide. The peptides were added to the monolayers before the addition of the labeled bacteria.

Molecular size fractionation of HS was carried out using size-exclusion chromatography. One hundred microliters of a 1 mg/ml HS solution including 10 nCi of [³H]HS were applied to a superose 12 column connected to an FPLC system (GE), and eluted with 50 mM of pH 8.0 Tris-HCl buffer and 150 mM of NaCl at a flow rate of 1 ml/min. One milliliter fractions were collected, and HS eluting from the column was determined by measuring radioactivity in aliquots of 200 μl of each fraction. Fractions including molecules of either high or low MW were pooled separately and then precipitated with 85% ethanol for 2 h at −80°C.

Highly sulfated HS domains were obtained by digestion of HS chains with heparinase III, and poorly sulfated HS domains were obtained by digestion with heparinase I. In both cases 18 μg HS were digested with 100 mU enzime in a 200 μl total volume of sodium acetate buffer 100 mM pH 6.8 including CaCl2 10 mM.

RNA was isolated from corneal cells using the RNeasy kit following the manufacturer's specifications. Samples were subjected to treatment with RNase-free DNase during the purification process itself. The concentration of RNA obtained was determined spectrophotometrically by measuring absorbance using a Picodrop Microliter UV/Vis pectrophotometer (Picodrop Limited, UK).

cDNA synthesis was carried out using the High Capacity cDNA Transcription Kit following the manufacturer's specifications. The reaction products were cleaned using the PCR Clean-Up GenElute kit following the manufacturer's instructions.

qRT-PCR reactions were carried out as previously described (Fernández-Vega et al., 2013), and actin was used as a control gene to compare run variation and to normalize individual gene expression. The expression values of the genes were calculated as 2^−ΔCt (relative to actin as the housekeeping gene).

Antibody inhibition assays were carried out in 24-well plates and grown to 80% confluence. The media was aspirated, and the cells were washed twice with PBS and then blocked with 10% fetal bovine serum in PBS for 2 h at 37°C in a 5% CO₂ atmosphere. After further washing with PBS, anti syndecan-1, anti syndecan-2, anti syndecan-3, or anti syndecan-4, or a mixture of all four, in PBS were added and incubated for 1 h. All the antibodies were diluted 1:100, except anti-SDC2, which was diluted 1:250. After the treatments, adherence assays were performed as previously indicated.

PI-PLC was used to enzymatically remove glypicans from the cell membrane. Cells were grown in 24-well plates to 80% confluence. After washing with PBS, the cells were incubated in the absence or presence of 80 mU/ml PI-PLC for 40 min at 37°C in a 5% CO₂ atmosphere. After the treatment, adherence assays were performed as detailed previously.

Human lentiviral constructs for NDST1 (TRCN0000008648), HS2OST1 (TRCN0000034936), and HS6ST1 (TRCN0000034564) were purchased from Thermo Fischer Scientific. Human lentiviral constructs for NDST1 and HS6ST1 were selected to include sequences shared by NDST2 and HS6ST2 respectively. Lentiviral particles were produced by cotransfection of 293T cells with the shRNA plasmid and packaging vectors pMD2.G and psPAX2 (gifted by Didier Trono; Addgene plasmids # 12259 and #12260, respectively) according to the vendor's instructions. Knockdown efficiency was determined by qRT-PCR analysis as previously described (Fernández-Vega et al., 2013).

All analyses were performed using the Statistics for Windows program (Statsoft Inc; Tulsa, OK). Mean values between two samples were compared using the Mann-Whitney U-test, and between multiple samples using the Kruskal-Wallis test. P < 0.05 was accepted as significant. All data are presented as means ± standard error.

To study whether GAGs mediate the binding of different common corneal pathogens, including both Gram-positive and Gram-negative bacteria, to corneal epithelial cells, their biosynthesis was inhibited using either rhodamine B or genistein. The results showed a decrease in the adherence of all the bacteria investigated, suggesting that GAGs are involved in a general way in the binding of pathogenic microorganisms to the corneal epithelium (Figure 1A). However, the observed effect was highly dependent on two factors: the inhibitor molecule used and the Gram nature of the bacterium analyzed. Rhodamine B diminished the binding of Gram-positive bacteria by 60% (±5.9), while Gram-negative binding was decreased by only 23% (±8.9). Moreover, GAG-devoid cells, obtained through treatment with genistein, reduced Gram-positive and Gram-negative bacteria binding by 44 (±3.7) and 60% (±14.8), respectively (Figure 1B); However the peculiar behavior of P. aeruginosa, whose binding decreased by only about 13%, demonstrates the high deviation of the result obtained for the Gram-negatives, to the extent that the exclusion of this microorganism results in a reduction of 65% (±5). Statistical analysis showed significant values for rhodamine B (p < 0.01) and a close to significant effect for genistein (p = 0.08, p < 0.01 if Pseudomonas is excluded), suggesting that inhibition by both molecules differentially affects the binding of microorganisms depending on their Gram nature. What is more, the two molecules produce opposite effects, such that inhibition by rhodamine strongly affects binding of Gram-positive bacteria, while genistein's effect is more intense with Gram-negatives (Figure 1B).

To investigate further the adherence of pathogens to the GAGs, corneal cell surface GAGs were removed by digestion with bacterial lyases, and the effect of this on the binding of the different bacteria was determined. Treatment with heparinase I and III or chondroitinase ABC reduced the adherence of the pathogens in a similar manner, with no statistically significant differences (Figure 1C). However, there were differences when the effects on either Gram-positive or Gram-negative pathogens were analyzed separately; treatment with heparinases or chondroitinase reduced the adherence of Gram-positive bacteria 35.5 (±7.05) and 40% (±5.2), respectively, while the reduction for Gram-negatives were 22 (±2.6) and 22.5% (±4.9); the differences being statistically significant for both treatments (p < 0.01; Figure 1D). In the combined digestion employing both heparinases and chondroitinase, adherence was significantly reduced compared to in digestions using a single enzyme (p < 0.05 in all cases), and statistically significant differences between Gram-positive and Gram-negative were also detected (p < 0.01l Figure 1D).

To analyze the role that different species of GAGs play in the adhesion of pathogens to corneal cells, adherence interference experiments were performed using HS, CS A, CS B, CS C, and a mixture of all four, as described in the Material and Methods section. In all cases, the presence of GAG molecules decreased the adherence of bacteria to corneal cells in a dose dependent manner (Figure 2A). Furthermore, the effect was dependent on the particular species of GAG used. In all cases, HS was the most effective interfering molecule, followed by CS B, CS A and, lastly, CS C, except in the case of the two species of Streptococcus examined, where CS A showed a greater effect than CS B. When a mixture of all GAGs was used, the effect was similar to that obtained using HS alone, except in the cases of E. coli, S. marcescens, P. aeruginosa, and H. influenzae, which the effect of the combined treatment was greater (Figure 2A).

Taken together, the results evidence that both, HS and the mixture of GAGs, have the highest inhibitory effect, although the differences were not statistically significant (p = 0.1); nevertheless, the two treatments did display differences when compared to other species analyzed (p = 0.002, 0.04, and < 0.001 for HS compared to CS A, CS B, and CS C, respectively; Figure 2B). When the inhibitory effect of the different GAGs was analyzed separately for Gram-positive and Gram-negative bacteria, no significant differences was detected except in the case of the mixture of all the species (p = 0.049; Figure 2C).

To analyze whether pathogens adhere primarily to corneal epithelial cells or to the ECM, we prepared wells by detaching the cells with EDTA, as described in Material and Methods. The values of the adhesion of bacteria to the ECM were much lower than those observed for cells (usually below 5%). Adherence interference experiments carried out using the different GAG species at 40 nM showed that the adherence of bacteria decreased by around 10% with each of the CS types, by 21% for HS and by 26% for the GAG mixture, all much lower than the values obtained for the inhibition of adhesion to cells (Figure 2D).

Using size exclusion chromatography, HS was separated into fractions of low (< 10 kDa) and high (>10 kDa) MW (Figure 3A). The effect of both fractions on the binding of pathogens to corneal cells was analyzed by adherence interference experiments. The results showed that high MW HS decreased the adherence of bacteria around 20%, and this value was independent of the Gram type of the microorganism (Figures 3B,C). Interestingly though, low MW HS did display different, and opposite, effects depending on whether bacteria were Gram-negative or –positive: the former diminishing their adhesion about 40%, and the latter around 20% relative to controls (Figures 3B,C), and this difference was statistically significant (p < 0.001).

HS chains are composed of highly sulfated NS-domains interspaced with poorly sulfated NA-domains. These two domain types were isolated by treatment with specific heparinases, and then used to carry out adherence interference experiments (Figure 3D). NS-domains in general were found to be twice as effective in terms of inhibiting the binding of the bacteria (p < 0.01; Figure 3E); however, there were no significant differences between the effect produced on Gram-positive or Gram-negative bacteria by the two types of domain (Figure 3F).

Two peptides were designed which included consensus heparin binding sequences of different lengths, the shorter being QKKFKN, and the longer, FKKKYGKS. The effect of the concentration of the two peptides on bacterial adherence to HCE-2 cells was determined using the Gram-positive S. aureus and the Gram-negative P. aeruginosa. In the case of S. aureus, both peptides decreased adherence in a dose-dependent manner up to concentrations of 10–20 nM, their effects diminishing at higher concentrations. In contrast, P. aeruginosa decreased its adherence in the presence of the short peptide up to 20 nM, while the longer peptide produced progressive interference, more noticeable at higher concentrations (Figure 4A). On the basis of these data, subsequent experiments were carried out at a concentration of 20 nM for both peptides.

Results showed that the longer peptide decreased adherence by about 20–30% for all the bacteria analyzed (Figure 4B). However, the shorter peptide had a similar effect on Gram-positive bacteria, but decreased the adherence of Gram-negative bacteria about 50–60%, the difference between the two groups being statistically significant (p < 0.001; Figure 4C).

To find out which species of HSPGs were present on the membrane of the HCE-2 cells, the transcription levels of all the isoforms of syndecans and glypicans were analyzed by qRT-PCR reactions. The results showed the existence of transcripts for each of the four syndecan isoforms, although their respective levels varied in total by around three orders of magnitude, SDC4 and SDC1 being the most abundant, and SDC2 the most scarce (Figure 5A). In contrast, transcripts were not detected for all isoforms of glypicans and some appeared at only very low levels, GPC1 being the most abundant species (Figure 5A).

To investigate the role of glypicans in the binding of pathogens, they were removed from the cell surface using PI-PLC, an enzyme that cleaves their glycosylphosphatidylinositol-anchor. The results showed that this treatment only slightly decreased bacterial adherence, and no differences between Gram-positive and Gram- negative were detected (Figure 5B).

Since the four isoforms of syndecans were all transcribed in HCE-2 cells, their individual involvement in adherence was analyzed by blocking them using specific antibodies for each species in separate experiments. The results obtained were variable and, curiously, were not related to the apparent levels of SDC isoforms derived from their transcription levels (Figure 5B). Interestingly, when antibodies against the four isoforms were used simultaneously in the experiment, adherence decreased sharply, irrespective of the Gram nature of the pathogen (Figure 5B).

To further establish the influence of sulfation at specific positions on the disaccharide unit of HS chains (Figure 6A), we used RNA interference to silence the expression of the genes involved in these reactions. Downregulation of N-sulfation of glucosamine residues was carried out by silencing NDST1, which reduced its expression with an efficiency of over 80%, and also reduced the expression of NDST2 about 70%, without affecting C5-glucuronate epimerization (GLCE), 2-O-sulfation (HS2ST1) or 6-O-sulfation (HS6ST1 and HS6ST2). Downregulation of 2-O-sulfation of uronic acid residues was performed by silencing HS2ST1, which reduced its expression with over 70% efficiency without affecting N-sulfation or 6-O-sulfation. Finally, downregulation of 6-O-sulfation of glucosamine residues was carried out by silencing HS6ST1, reducing its expression with an efficiency of over 80%, as well as the expression of HS6ST2 by around 70%, without N-sulfation or 2-O-sulfation being affected (Figure 6B).

Analysis of the adhesion of the pathogens to the knockout clones showed that the decrease in sulfation reduced the binding of all microorganisms tested. Interestingly, in all the cases, a decrease in N- or 6-O-sulfation, both of which involve the glucosamine residue, affected adherence to a greater extent than a reduction in 2-O sulfation, which involves the uronic acid residue (Figure 6C). When the effect on either Gram-positive or -negative bacteria was analyzed separately, N- and 6-O-sulfation varied considerably depending on the bacterium involved and differences between Gram type were not significant. In contrast, 2-O sulfation presented more homogenous results within each Gram type, with Gram-negative bacteria adherence being reduced by around 40% on average and Gram-positive by < 30%, a difference that was statistically significant (p < 0.01; Figure 6D).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.