A Novel Cell-Associated Protection Assay Demonstrates the Ability of Certain Antibiotics To Protect Ocular Surface Cell Lines from Subsequent Clinical Staphylococcus aureus Challenge^,

Abstract

In vivo effectiveness of topical antibiotics may depend on their ability to associate with epithelial cells to provide continued protection, but this contribution is not measured by standard antibiotic susceptibility tests. We report a new in vitro method that measures the ability of test antibiotics azithromycin (AZM), erythromycin (ERY), tetracycline (TET), and bacitracin (BAC) to associate with mammalian cells and to protect these cells from destruction by bacteria. Mammalian cell lines were grown to confluence using antibiotic-free medium and then incubated in medium containing a single antibiotic (0 to 512 μg/ml). After incubation, the cells were challenged with Staphylococcus aureus ocular isolates, without antibiotics added to the culture medium. Epithelial cell layer integrity was assessed by gentian violet staining, and the minimum cell layer protective concentration (MCPC) of an antibiotic sufficient to protect the mammalian cells from S. aureus was determined. Staining was also quantified and analyzed. Bacterial viability was determined by culture turbidity and growth on agar plates. Preincubation of Chang and human corneal limbal epithelial cells with AZM, ERY, and TET at ≥64 μg/ml provided protection against AZM-susceptible S. aureus strains, with increasing protection at higher concentrations. TET toxicity was demonstrated at >64 μg/ml, whereas AZM displayed toxicity to one cell line at 512 μg/ml. BAC failed to show consistent protection at any dose, despite bacterial susceptibility to BAC as determined by traditional antibiotic susceptibility testing. A range of antibiotic effectiveness was displayed in this cell association assay, providing data that may be considered in addition to traditional testing when determining therapeutic dosing regimens.

INTRODUCTION

Traditional antibiotic efficacy tests, such as MIC, evaluate the interaction between the pharmacologic agent and the bacterial cells in culture (12). Although this interaction is highly important and has guided clinical decision-making regarding antibiotic choice, these tests fail to incorporate information about the host tissue that may affect bacterial susceptibility to clinical therapy. While this phenomenon may be important in many tissue types, it is especially important for the eye, where antimicrobials may be delivered topically but may not remain at the site of infection long enough to provide adequate therapy without very frequent dosing.

When antibiotics are applied directly to the ocular surface, they may adhere to or become incorporated within epithelial cells. Since the tear film is made briskly and quickly circulates away from the eye via the nasolacrimal system (20), there is a theoretical advantage to antibiotics that have a prolonged tissue half-life due to tissue absorption. In the current study, we evaluate the efficacy of various antibiotics to control clinical ocular Staphylococcus aureus strains using a novel cell-associated assay. Specifically, we report the different extents to which azithromycin (AZM), erythromycin (ERY), tetracycline (TET), and bacitracin (BAC) protect Chang and human corneal limbal epithelial (HCLE) cell lines against challenge with S. aureus ocular isolates after all free drug had been removed from the cell culture. Antibiotic toxicity was also evaluated.

We chose S. aureus as a challenge in this assay because it is a major pathogen, associated with a variety of ocular infections, including blepharitis, conjunctivitis, keratitis, and endophthalmitis (1, 6, 7, 17, 19). Our approach was to evaluate possible treatments for conjunctivitis and blepharoconjunctivitis, and so the six strains chosen in this assay were isolated from patients with these conditions. The antibiotics that we evaluated include two readily available antibiotics that are marketed as ophthalmic ointments (erythromycin and bacitracin), one that has recently received Food and Drug Administration approval as an ophthalmic solution intended to treat bacterial conjunctivitis caused by S. aureus and other bacteria (azithromycin), and one that historically was a treatment for a number of ocular surface infections (tetracycline). This study therefore addresses the effects of a range of antibiotics from different drug classes in the protection of multiple ocular surface cell lines of likely clinical relevance, measuring the antibiotic's in vitro ability to protect epithelial cells against a clinically relevant infectious agent.

In this study, we demonstrated that a novel in vitro assay, which we termed the cell-associated protection assay (CAPA), can be employed to measure the relative protective efficacy of an antibiotic based on its ability to associate with human ocular surface cell layers composed of epithelial cells. We demonstrated that certain antibiotics associated so closely with ocular surface cell lines that the epithelial cell layers were protected from clinical S. aureus challenge even after all free drug was vigorously washed away from the cell culture. This protection was observed throughout a 24-h S. aureus challenge, and several assays suggested that the majority of the bacteria had actually been killed by the cell-associated drugs. CAPA may be used to analyze the ability of antibiotics to continue providing effective antibacterial control throughout the day, between eye drop doses or other antibiotic applications, an analysis that could help guide dosing interval recommendations.

MATERIALS AND METHODS

Cell lines and culture conditions.

Chang conjunctival cells (clone 1-5c-4 [Wong-Kilbourne derivative of Chang conjunctiva]; ATCC CCL-20.2; American Type Culture Collection [ATCC], Manassas, VA) were maintained in Gibco medium 199 with 1% Penn-Strep, 10% fetal bovine serum (FBS), 5% sodium bicarbonate, and 0.1% gentamicin. When cells were plated for these experiments, outgrowth medium (OG) for Chang cells consisted of Gibco medium 199 with 10% FBS and 5% sodium bicarbonate but without antibiotics. The ATCC reports that the Chang cell line is contaminated with HeLa cells. Here, we use Chang cells as sample mammalian cells and draw no conclusions based upon their debatable ocular source.

HCLE cells (4) were obtained from Jes Klarlund with permission from Ilene Gipson. OG for HCLE cells consisted of keratinocyte-SFM (serum-free medium) with l-glutamine, supplemented with 25 μg/ml bovine pituitary extract (BPE), 0.2 ng/ml epidermal growth factor (EGF), and 1 mM CaCl₂, without any antibiotics.

Bacterial strains and growth medium.

Six clinical isolates of S. aureus were recovered from patients presenting with bacterial conjunctivitis and blepharoconjunctivitis to the UPMC Eye Center, with microbiologic testing performed in the Charles T. Campbell Ophthalmic Microbiology Laboratory, Pittsburgh, PA. Isolates were retrieved from a frozen −75°C retrospective clinical collection that was deidentified and stored for antibiotic validations. Isolates were selected from the collection to include four isolates that were macrolide (AZM, ERY) susceptible and two that were macrolide resistant, based on Kirby-Bauer disk diffusion testing. No other antibiotic susceptibilities were tested as part of the inclusion criteria, although the MIC of all isolates was later determined using Etests for each studied antibiotic (bioMérieux, Inc., Durham, NC) (Table 1). Bacteria were grown in tryptic soy broth (TSB) and maintained on tryptic soy agar (TSA) containing 5% sheep's blood (BD BBL Becton, Dickinson and Co., Sparks, MD).

Table 1.

Bacterial strains used in this study and their MICs and MCPCs

Straina	Antibiotic MICb [MCPCc] (μg/ml)
	AZM	ERY	TET	BAC
B1370	0.75 [32/8]	0.5 [64/8]	0.38 [32/16]	48 (R) [>512/>512]
B1382	64 (R) [>512/>512]	32 (R) [512/256]	0.75 [32/16]	>256 (R) [>512/>512]
B1391	0.75 [8/8]	0.38 [32/8]	0.75 [32/8]	48 (R) [>512/>512]
B1396	64 (R) [>512/>512]	24 (R) [256/>512]	0.5 [32/64]	>256 (R) [>512/>512]
B1412	1.5 [32/8]	0.5 [64/8]	0.5 [32/8]	14 [>512/>512]
B1415	1.5 [32/16]	0.38 [32/64]	0.38 [32/32	14 [>512/>512]

^aAll strains are conjunctivitis or blepharoconjunctivitis clinical isolates.

^bBacterial strains with MICs (μg/ml) for each antibiotic by Etest. Resistance is based on the CLSI (Clinical and Laboratory Standards Institute) systemic breakpoints for each antibiotic. (R) indicates resistance to the antibiotic by MIC; where this is not listed, the strain was susceptible to the given antibiotic by MIC.

^cBacterial strains with MCPCs (μg/ml), indicating the minimum concentration of antibiotic for which cell layer protection was evident in the CAPA. The left number is for Chang cells, and the right number is for HCLE cells.

Experimental drugs.

AZM 1% solution was provided by Inspire Pharmaceuticals (Durham, NC). TET was purchased from Sigma-Aldrich (St. Louis, MO), and a 1% solution was prepared in 95% ethanol. Stocks of ERY (Sigma-Aldrich) and BAC (Sigma-Aldrich) were created prior to each set of experiments and stored as a 1,024-μg/ml solution.

In vitro CAPA. (i) Step 1.

Flat-bottom 96-well tissue culture plates (Costar 3595; Corning Inc., Corning, NY) containing Chang or HCLE cells were grown to confluence without antibiotics, as described above.

(ii) Step 2.

The medium from each well was removed, and 200 μl of OG medium with the test antibiotics (0 to 512 μg/ml) was added to each well according to the key detailed in Fig. 1A. Each concentration was plated in triplicate wells on each of the 8 plates used per experiment. Plates were incubated at 37°C in 5% CO₂.

Fig. 1.

Different antibiotics and different ocular cell lines exhibit contrastive staining zones in the cell-associated protection assay. (A) Assay plate legend showing the concentration of antibiotics, with which each well was incubated before free antibiotic was washed away and before challenge with S. aureus. TET (0 μg/ml) was supplemented with ethanol to a concentration equal to the ethanol concentration in 512 μg/ml TET. AZM, azithromycin; ERY, erythromycin; TET, tetracycline; BAC, bacitracin. (B and C) Gentian violet staining of a representative experiment using Chang (B) and HCLE (C) cells after pretreatment with study antibiotics and subsequent inoculation with S. aureus strain B1412. Dark-purple-stained wells are where pretreatment of the monolayer with the indicated antibiotic concentration was sufficient to afford protection to the monolayer from S. aureus infection. Clear wells did not have sufficient cell-associated antibiotic to protect the wells from bacteria, and the mammalian monolayer was obliterated, or antibiotic toxicity destroyed the cell layers. The MCPCs for Chang cells for S. aureus strain B1412 were 32 μg/ml for AZM, 64 μg/ml for ERY, 32 μg/ml for TET, and >512 μg/ml for BAC. The MCPCs for HCLE cells for S. aureus strain B1412 were 8 μg/ml for AZM, 8 μg/ml for ERY, 8 μg/ml for TET, and >512 μg/ml for BAC.

(iii) Step 3.

Plates containing the epithelial cell cultures were removed from incubation after 24 h with the antibiotics, medium was removed, and all wells were washed twice with 200 μl phosphate-buffered saline (MP Biomedicals, Solon, OH).

(iv) Step 4.

Bacterial inocula were 1 × 10⁵ CFU in 200 μl of OG medium per well (confirmed range by colony counts: 3.68 × 10⁴ to 4.76 × 10⁵ CFU/well). The bacterial strains are listed in Table 1. Plates were incubated for 24 h at 37°C in 5% CO₂. Two plates were not inoculated with bacteria, one as a mock infection plate, and the other for toxicity assays noted below.

(v) Step 5.

Remaining medium was removed from all wells, and plates were washed twice with water to remove unattached cells and debris. Epithelial cell layers were then fixed and stained with a gentian violet solution (0.5% gentian violet, 0.9% NaCl, 1.85% formaldehyde, and 50% ethanol) and allowed to dry. Plates were scanned to provide qualitative data on the survival of intact epithelial cell monolayers.

MCPC determination.

To assign the minimum cell layer protective concentration (MCPC) of an antibiotic, we observed scanned plate images for wells in which gentian violet staining was apparent. With Chang cells, the cell layers without antibiotic treatment were destroyed, and no staining was apparent, so that any purple staining was counted as a positive. Two plates from independent experiments were observed, each with three replicate wells for each bacterium-antibiotic concentration combination. At least two of the three replicate wells in one experiment had to agree in order to consider an antibiotic concentration as having a protective effect, and if the MCPCs of the two experiments differed, then the lower value was used. For the HCLE cells, the cell layers were often more resistant to bacterial challenge so that there was more background gentian violet staining, making identification of wells with increased protection more difficult. For HCLE cells, three independent experimental plate scans were observed. In this case, we recorded for the lowest concentration with a difference in staining compared to the no-antibiotic treatment. As with the Chang cell MCPC prediction, at least two wells at a concentration were required for MCPC estimation, and the lowest value was taken.

Quantification of stained cell layers and microscopy.

Glacial acetic acid (200 μl per well of 30% [vol/vol]; Fisher Scientific) was then added to each well to solubilize the gentian violet. After gentle vortexing, 20 μl from each well was transferred to corresponding wells in new plates that contained 180 μl of 95% ethanol. A single well in each plate was filled with 200 μl of 95% ethanol and used as a blank, and plates were read for absorbance at 590 nm on the Synergy 2 microplate reader (BioTek, Winooski, VT).

In addition to gentian violet staining, cell layers were assessed by microscopy, and representative photomicrographs were obtained. Phase-contrast images were obtained using a Nikon Eclipse TE2000-U microscope equipped with a CoolSnap HQ charge-coupled device camera, and images were acquired using Metamorph software. Photomicrographs were taken of plates of cells challenged using bacterial strain B1370 before we determined that it was BAC resistant, as evaluated by MIC determination (Table 1); micrographs were not taken of plates challenged with other bacterial strains.

Mathematical modeling.

Mathematical modeling was used to analyze gentian violet staining data. Models were fitted for the logarithm of output (staining) against the 4th root of the antibiotic dose. Various kinds of nonlinear regression models (including sigmoid and bell-shaped curves) and linear regression models (including linear, quadratic, or cubic curves) were fitted. Statistically significant dose-dependent relationships were determined based on the statistical significance of the coefficients estimated.

When a statistically significant dose-dependent relationship was found, a 50% effectiveness dose was calculated, i.e., the theoretical antibiotic dose at which a half maximal gentian violet staining was attained. To determine this metric, the output values of the lower and upper asymptotes were determined within the dose range of 0 to 512 μg/ml (4th root of dose 0 to 4.76 μg/ml), and the halfway point on the output axis (epithelial cell integrity) was determined, such that the corresponding antibiotic dose(s) on the dose axis could be calculated. All statistical computations and graphics were produced using the R statistical programming language and environment (14).

Culture turbidity and bacterial growth.

After overnight culture incubation, culture turbidity was assessed as an indicator of bacterial growth by reading absorbance at 600 nm by using a Synergy 2 microplate reader, using wells with OG medium as a blank. To determine whether there was viable S. aureus remaining in the inoculated wells, a 48-pronged, alcohol- and flame-sterilized plate replicator (Dan-Kar, Woburn, MA) was used to transfer ∼2 μl of each culture well to blood agar plates. These blood agar plates were incubated at 37°C overnight and subsequently graded for degree of S. aureus growth (0, no growth; 1, 1 to 5 colonies; 2, 6 or more colonies; 3, confluent growth). Wells that showed obvious contamination with microbes other than S. aureus were excluded from further analysis, as were wells displaying visible turbidity prior to bacterial inoculation. Control experiments indicated that epithelial cell layers were necessary for the observed antibacterial activity.

Antibiotic toxicity assays.

For antibiotic toxicity plates, after the overnight epithelial cell incubation with antibiotics, 200 μl of OG was added instead of a bacterial culture. These plates were incubated for 20 to 24 h at 37°C in 5% CO₂ and then assessed for cytotoxicity using alamarBlue (Invitrogen, Camarillo, CA). Healthy cells are able to reduce alamarBlue, shifting supernatant color from blue to pink and increasing its fluorescence over 100-fold. First, to serve as a negative control for reducing power, 2 μl of Promega lysis solution (Madison, WI) was added to 3 wells to lyse cell layers. After 15 min, medium was removed from all wells, and 200 μl of a 4% alamarBlue solution in clear OG (without phenol red) was added to each well. The plate was returned to the incubator for 2 h at 37°C in 5% CO₂, and then fluorescence was measured using the Synergy 2 microplate reader (excitation filter, 500/27; emission filter, 620/40).

RESULTS

In vitro epithelial assay to measure the cell-associated efficacy of antibiotics.

A new in vitro method was devised to measure the contribution of an antibiotic's ability to interact with epithelial cells and protection of the cell layer from infectious challenge. The key component to the CAPA is that after the cell layer is incubated with antibiotics, the layer is vigorously washed and challenged by pathogenic bacteria in antibiotic-free medium. Therefore, any protective effect occurs from antibiotics that associate with the epithelial cell surface or enter the cells. We tested this model using four antibiotics, six ocular clinical isolates of S. aureus, and two mammalian cell lines. We report the results of the above experiments for one representative bacterial strain, B1412 (Fig. 1 and 2). Strain B1412 was chosen as the representative strain because it was susceptible to all studied antibiotics by MIC (Table 1). Results for the additional tested strains are shown in Fig. S1 and Tables S1 and S2 in the supplemental material.

Fig. 2.

Quantitative analysis of gentian violet staining for Chang and HCLE cells challenged with S. aureus strain B1412. Gentian violet stain was solubilized and measured at A₅₉₀. (A) CAPA with Chang cells. The experiment was performed twice on different days (n ≥ 6 total wells per condition), and the average is shown. (B) CAPA with HCLE cells. The experiment was performed three times on different days (n ≥ 9 total wells per condition), and the average is shown. Error bars, standard errors of the means.

Epithelial cell layer protection as a function of pretreatment with antibiotics.

Visual inspection of gentian violet staining of the Chang and HCLE cells shows that without antibiotic pretreatment, mammalian cell layers were destroyed by the bacterial challenge (Fig. 1B and C). AZM and ERY show some degree of protection throughout the tested range when cell layers were challenged with macrolide-susceptible S. aureus strains: TET shows protection in the middle of the tested range, with apparent toxicity at higher doses, and BAC shows no obvious protective efficacy at any tested concentration with either mammalian cell type (Fig. 1B and C). Results similar to those using the Chang cell line were measured with the A549 lung carcinoma cell line using AZM and ERY and a subset of the S. aureus strains (data not shown).

Microscopic analysis of protected and challenged epithelial cells supports that the gentian violet staining corresponds to intact monolayers of epithelial cells and that biofilms evident in low-antibiotic wells do not contribute to the A₅₉₀ measurements (data not shown), as there was negligible staining in the wells with 0 μg/ml antibiotic (Fig. 1). This is consistent with previous reports showing that S. aureus requires specific medium conditions to form wash-resistant biofilms capable of being stained with gentian violet (16).

Whereas Chang cells were highly susceptible to challenge by all of the S. aureus strains used, HCLE cells not exposed to antibiotics showed a range of damage by the different bacterial strains; this difference can be clearly seen by the basal levels of staining in the zero-antibiotic treatment groups when gentian violet staining was quantified, described below (Fig. 2; see also Fig. S1 in the supplemental material). Nevertheless, the range in which pretreatment with antibiotics confers protection to the cell layer against bacterial challenge can clearly be determined.

To characterize the extent of cell layer protection, we established a metric that we titled the minimum cell layer protective concentration (MCPC). The MCPC represents the minimum concentration of a given antibiotic that protected the cell layer from destruction by S. aureus challenge in the CAPA assay (Table 1). It was noted that AZM exhibited the same or lower MCPC values than ERY (Table 1), despite the fact that the strains had lower MIC values to ERY (Table 1). All bacterial strains had low MCPCs with TET, although these were somewhat difficult to determine with HCLE cells, as the staining differences were subtle. BAC up to 512 μg/ml produced no measure of protection, so a value of >512 μg/ml was designated.

Quantification and modeling of gentian violet staining.

Gentian violet was solubilized and measured to quantify the observed cell layer protection effect, both to validate the MCPC values and to further characterize the ability of antibiotics to associate with and protect mammalian cells. The combined solubilized gentian violet data from replicate experiments shows that Chang cell layer integrity is enhanced by AZM (32 to 512 μg/ml), ERY (64 to 512 μg/ml), and TET (32 to 128 μg/ml) when susceptible bacteria were used (Fig. 2A; see also Fig. S1 in the supplemental material). There is a drop-off in staining from TET between 64 μg/ml and 128 μg/ml with all S. aureus strains, suggesting that TET begins to show toxicity at >64 μg/ml. BAC shows little difference from untreated controls at any concentration even though strain B1412 is susceptible to BAC (Fig. 2A; see also Fig. S1 in the supplemental material).

The combined solubilized gentian violet data suggest that HCLE cell layer integrity is enhanced by pretreatment with AZM (16 to 128 μg/ml) and ERY (32 to 512 μg/ml) and to a minor extent by TET at 32 μg/ml (Fig. 2B). BAC shows little or no protection at any concentration. The five additional tested bacterial strains showed comparable results, with similar concentrations of antibiotics providing epithelial cell layer protection when the strain was susceptible to that antibiotic by MIC (see Fig. S1 in the supplemental material).

Mathematical modeling was applied to the gentian violet spectrophotometric data so that 50% effectiveness doses could be determined from antibiotic-bacterium strain combinations where protection was observed. This 50% effectiveness dose corresponds to the theoretical antibiotic concentration that should yield 50% maximal gentian violet staining in that experiment (see Table S1 in the supplemental material). The results of this modeling predict that for strain B1412 in the Chang cell line, the 50% dose was lowest for AZM (see Table S1). For the other macrolide-susceptible bacterial strains, TET and AZM have lower 50% predicted doses than ERY with respect to protecting the Chang cell line. With HCLE cells, AZM had lower 50% protection doses than ERY in all macrolide-susceptible bacterial strains. TET provided protection only against B1412 (and not other bacterial strains), although it did so efficiently, exhibiting the lowest 50% effectiveness dose for this strain (see Table S1). The MCPC values (Table 1) and the 50% effectiveness doses (see Table S1) exhibit a similar pattern in all cases except when tetracycline was used on HCLE cells.

Culture turbidity and bacterial cell growth.

The simplest model for the mechanism of protection of epithelial layers from S. aureus infection in this cell protection assay is that epithelial cells secrete antibiotics into the medium at sufficient concentrations to prevent proliferation of S. aureus to numbers toxic to the epithelial cells. If this model is true, then there should be fewer bacteria in wells that exhibit protection. Culture turbidity provided a measurement of bacterial density within the supernatant (Fig. 3A), and growth on blood agar plates from well aliquots provided confirmation that this assessment correlated with bacterial survival (Fig. 3B). Figure 3A shows a representative culture turbidity assay performed with HCLE cell lines after inoculation with strain B1412. In both Chang and HCLE cell lines, AZM, ERY, and TET show an inverse relationship between antibiotic concentration and culture turbidity. These results are correlated with bacterial growth on blood agar plates. Figure 3B shows 1-day growth from one replicate of the HCLE experiment with strain B1412. Bacterial growth from three experiments was graded (Fig. 3C), and AZM and ERY show less growth at ≥128 μg/ml, and TET shows less growth at 256 to 512 μg/ml. BAC had no effect on bacterial growth to the sensitivity of either assay. Combined turbidity and blood agar data are correlative to the epithelial protection assay results above, and Chang cells show similar results (data not shown).

Fig. 3.

Analysis of bacterial growth and viability. (A) Turbidity of bacteria and HCLE cultures 24 h after inoculation with S. aureus strain B1412. Epithelial cells had been pretreated with the antibiotics listed at the displayed concentrations in micrograms per milliliter. Similar results were found using Chang cells. Error bars indicate one standard deviation. (B) Growth of S. aureus strain B1412 after plate replication from an inoculated HCLE culture that had been pretreated with the study antibiotics according to the legend for Fig. 1A. [Ab], antibiotic concentration. (C) Median graded growth of B1412 from blood agar plates as shown in panel B. 0, no growth; 1, 1 to 5 CFU; 2, 6+ CFU, but not confluent; 3, confluent. Similar results were measured with Chang cells. The experiment was repeated 3 times on different days; n = 9 for each data point.

Toxicity assays.

In addition to the toxicity results implied above, two dedicated toxicity assays were performed during each run of the experiments without bacterial challenge. First, mock experiment plates (no S. aureus challenge) were stained with gentian violet as described above (Fig. 1B and C). Visual inspection of these gentian violet-stained cells showed clear cytotoxicity from TET only, an effect seen with both Chang and HCLE cell lines (data not shown).

Solubilized gentian violet testing for mock-infected Chang cells showed TET toxicity at >64 μg/ml, but no other antibiotic tested showed staining at any concentration that was clearly different from that of untreated cells (Fig. 4A). HCLE cell testing showed a moderate reduction in staining when treated with AZM at high concentrations, ERY at 8 to 16 μg/ml, and a stronger effect from TET at concentrations of 64 to 512 μg/ml (Fig. 4B).

Fig. 4.

Toxicity controls show tetracycline toxicity. (A and B) Gentian violet from toxicity plates was solubilized and measured spectrophotometrically. Reduced staining indicates loss of cell layer integrity. (C and D) alamarBlue vital staining where reduced fluorescence compared to the mock group (0 antibiotic) indicates loss of epithelial layer viability. Lysis indicates minimum fluorescence from a detergent-destroyed monolayer. (A and C) Chang cells; each data point represents the average of results from two separate experiments performed on different days, with measurements from at least 5 wells. (B and D) HCLE cells; each data point represents the average of results from three separate experiments performed on different days, with measurements from at least 7 wells. Error bars, standard errors of the means.

Whereas gentian violet staining can be used to assess cell layer integrity, it does not necessarily correlate to cell layer viability. To more directly assess cell layer viability, the alamarBlue vital stain was used (Fig. 4C and D). alamarBlue testing confirmed the results found with gentian violet staining, showing the toxicities of TET at similar concentrations between the two assays. With both Chang (Fig. 4C) and HCLE (Fig. 4D) cells, the only antibiotic that conferred a dose-dependent loss of cell viability was TET. The toxicity of TET to Chang cells reduced viability staining to levels similar to those of the negative control (lysis), in which the cell layer had been killed using detergent (Fig. 4C).

DISCUSSION

The primary endpoint of infectious disease therapy is preservation of the host tissue, with microbial killing as a means to that goal. Our study and new model system underscore important differences in the ability of antibiotics to protect cells of the ocular surface. Although the S. aureus strain B1412 was susceptible to every study antibiotic, some antibiotics showed the in vitro ability to associate with epithelial cells, thereby prolonging their effective period of administration and providing superior protection against subsequent microbial challenge in this assay.

We conclude that the protection from microbial challenge is due to the antimicrobial effects of the antibiotics, as opposed to an unknown mammalian cell response to the antibiotic that leads to protection, for example, the upregulation of β-defensin production. This conclusion is based upon the observation that mammalian cell lines were not protected against bacterial strains that were antibiotic resistant. For example, when AZM-resistant bacterial strain B1382 (MIC = 64 μg/ml) was used, no tested concentration of AZM was able to protect Chang or HCLE cells from the bacteria (Fig. 1 and 2; see also Fig. S1 in the supplemental material).

The tendency of certain antibiotics, including AZM, ERY, TET, and BAC, to associate with human cells has been studied. Macrolide antibiotics, such as AZM and ERY, have been documented to reach high concentrations in bronchoalveolar cells, lung tissue, leukocytes, lung epithelial cells, and alveolar macrophages (8, 13, 21, 22). The ability of these drugs to penetrate cells may be important in fighting obligate intracellular pathogens (2). AZM was found to have a particularly long tissue half-life, estimated at greater than 2 days in mammalian phagocytic cells, and AZM accumulation was more than 20-fold higher than ERY accumulation in human polymorphonuclear neutrophils (9). With respect to the eye, a recent study described AZM reaching high concentrations in the human conjunctiva and persisting in the cornea for at least 7 days after treatment, with a half-life of 65.7 h (18). Another study showed that ERY was taken up by rabbit primary corneal epithelial cells, and its efflux was inhibited by steroids (10).

Tetracycline and related antibiotics were found to be concentrated severalfold by alveolar macrophages but to a lesser extent than ERY, whereas other drugs, such as penicillin G, cefamandole, and gentamicin exhibited very little uptake (8). Tetracycline was also reported to accumulate in lung epithelial cells, red blood cells, and neutrophils, with neutrophil uptake hypothesized to account for TET's immune modulatory effects (3, 22).

Very few studies have looked at the uptake of bacitracin by cells. In one study, the uptake of fluorescently labeled bacitracin in Swiss albino mouse dermal fibroblasts was found to be negligible without the use of ethosomes or liposomes as drug delivery mechanisms (5). Bacitracin was also stated to have no ability to penetrate through the cornea to the anterior chamber, unlike vancomycin, chloramphenicol, fluoroquinolones, macrolides, and aminoglycoside antibiotics, suggesting that it cannot pass through cell layers (15). Together, these previous antibiotic accumulation studies correlate with our findings that macrolide and tetracycline antibiotics exhibit cellular protection and cellular accumulation, whereas bacitracin has previously shown no accumulation and, here, no protection.

The effect of cellular antibiotic uptake in an in vitro system analyzing antibiotic efficacy has not been previously described in the literature. It is striking that BAC, a commonly prescribed antibiotic, shows scant if any ability to protect the tested epithelial cells against bacterial challenge in this pretreatment assay, whereas AZM and ERY show a remarkable ability to protect the cell lines even at very low pretreatment concentrations which are far below the toxicity threshold. These data suggest that AZM and ERY may be superior to BAC for the treatment of ocular surface S. aureus infections and that these drugs may also be superior to TET, since TET was toxic within or very near to the therapeutic concentration range for both cell lines and all bacterial strains tested. A study using human gingival epithelial cells showed a similar trend, with TET showing higher levels of cytotoxicity than both AZM and ERY (11). Cell association could not overcome resistance demonstrated by MIC, but MIC testing would have predicted success with BAC treatment, a result called into question by the cell association assay.

We describe a new metric for antibiotic characterization, the MCPC. This value can be used in conjunction with MIC values when determining antibiotic choice and dosing patterns. The assay is reasonably inexpensive, easy to perform, and flexible. Our results suggest that while the CAPA can be performed effectively with a wide range of mammalian cell lines, to employ the simple MCPC method, some care must be taken in choosing a cell line appropriate for the assay, ensuring that the bacterial challenge strain is sufficiently pathogenic to eliminate background staining. Preliminary experiments with several cell lines and with other antibiotics have shown that other antibiotic classes are capable of providing epithelial cell protection against bacterial challenge. The incubation time for antibiotic loading can be altered to further test the time it takes for a certain antibiotic to associate, and other alterations of the assay to test antibiotic disassociation from tissue can easily be adapted from the primary assay. The benefit of the MCPC value is that it provides a metric for antibiotic association to tissue without the use of more expensive approaches, such as mass spectroscopy or high-performance liquid chromatography. One limitation to the MCPC value is that it does not indicate antibiotic toxicity to a particular cell line; however, toxicity can be observed when a protection zone fades at high concentrations, as exhibited by TET in Fig. 1B.

The utility of this antibiotic testing approach will be in further testing of various antibiotics to correlate with clinically significant preventive and therapeutic efficacy against ocular or other infections. Pretreatment effects as demonstrated in this assay may be a reasonable paradigm to evaluate the utility of an antibiotic in the environment of the ocular surface. An active tear film turnover and drainage through the nasolacrimal system should quickly eliminate free drug from the environment of the ocular surface. The treatment interval would have to be very frequent to overcome this loss, unless epithelial cells or other surface reservoirs can hold on to the antibiotic and then continue to elute the drug. Our results demonstrate that it is possible to evaluate antibiotic association with epithelial cells in a clinically appropriate way, although the clinical relevance remains to be demonstrated. The data presented in this study directly pertain to in vitro antibiotic association with particular cell types, and it may not be possible to generalize these results or to assume an equivalent in vivo effect. However, it must be stated that the data presented here study only antibiotic association in vitro using particular cell types and may not represent the in vivo effectiveness of particular antibiotics. Furthermore, inflamed tissue may exhibit differential interactions with antibiotics, and this will be the subject of subsequent studies using this model.