193-nm Excimer Laser–Induced Fluorescence Detection of Fluoroquinolonesin Rabbit Corneas

Objective  To measure the 193-nm excimer laser–induced fluorescence of fluoroquinolone-treatedcadaver rabbit corneas.

Methods  Prior to ablation with a commercially available ophthalmic excimer laser(Nidek EC-5000; Nidek Technologies, Pasadena, Calif), 35 cadaver rabbit corneaswere treated with topical sterile balanced salt solution, 0.3% tobramycinsulfate, or the fluoroquinolones—0.3% ofloxacin, 0.5% levofloxacin,0.3% ciprofloxacin hydrochloride, or 0.3% gatifloxacin. The fluorescence generatedfrom each ablated corneal layer was measured and used to identify the presenceof antibiotic. This was achieved by training a partial least-squares modelto discriminate between the fluorescence spectra of antibiotic-treated andantibiotic-free (healthy) cornea. Antibiotic concentrations down to 0.06 μg/mLwere detected with high accuracy. Assuming a constant ablation rate of 0.3μm per laser pulse, the number of corneal layers ablated to reach antibiotic-freecornea is used to calculate the penetration depth of the antibiotic.

Results  The mean ± SD penetration to a detectable depth wasas follows: 0.3% ofloxacin, 7.1 ± 3.0 μm; 0.5% levofloxacin,6.7 ± 1.4 μm; 0.3% ciprofloxacin, 1.2 ± 0.6 μm;and 0.3% gatifloxacin, 7.0 ± 1.9 μm. The penetration depthof 0.3% tobramycin could not be determined because its fluorescence spectrumoverlapped with that of the native cornea.

Conclusions  Topical administration of fluoroquinolone-containing solutions resultsin measurable differences in laser-induced corneal fluorescence. Under theseexperimental conditions, 0.3% ofloxacin, 0.5% levofloxacin, and 0.3% gatifloxacinall appear to penetrate the epithelium significantly more than 0.3% ciprofloxacin(P<.02).

Clinical Relevance  Monitoring of laser-induced fluorescence may be helpful in determiningthe penetration depths and concentrations of topically applied fluoroquinoloneswithin the cornea.

With the increasing popularity of such techniques as laser in situ keratomileusis(LASIK) and photorefractive keratectomy, excimer laser refractive surgeryhas emerged as one of the most common procedures performed in the United States.In 2000 alone, more than 1.5 million of these procedures were performed inthe United States,¹ and they are predictedto surpass surgical procedures for cataract as the most common eye operationsconducted within the next several years. Over the years, numerous aspectsof excimer laser technology have been scrutinized, including safety and thepossible health adverse effects of laser application. One of the earliestconcerns regarding the use of photorefractive keratectomy was initially raisedby Loree et al² about the potential damageto surrounding ocular tissues. This group examined laser-induced fluorescenceproduced by the ablation of porcine corneas at 193-nm (argon-fluoride), 248-nm(krypton-fluoride), and 308-nm (xenon-chloride) wavelengths using an intensifieddiode array detector system. They reported a potentially increased risk ofsecondary cataract formation from the low-energy UV light fluorescence producedduring ablation of the cornea. Muller-Stolzenburg et al³ alsoexplored this issue in bovine corneas by placing a quartz fiber into the anteriorchamber of the treated eye; they also expressed a similar concern. However,2 subsequent and more complete analyses by Ediger⁴ andTuft et al⁵ using photodiode systems independentlyshowed a total UV energy dispersion that suggested a low risk of cataractfrom the laser-induced fluorescence. Unlike the slower photodiode systems,using an intensified charge-coupled detector system, we monitored real-time,193-nm laser-induced fluorescence spectra during transition from epitheliumto stroma at clinically relevant laser fluences. We were able to corroboratethese later studies by reporting qualitatively similar laser-induced fluorescencespectra.⁶

In related work, using a pulsed 193-nm excimer laser at ablative energieson cadaver human and rabbit corneas, we determined the UV and visible fluorescencespectra as ablation progressed through the entirety of the cornea^7,8 with the ultimate goal of developinga fluorescence-guided spectrometer to provide an objective means to controlexcimer laser ablation.

To our knowledge, however, a key question that has not been addressedto date is whether the topical application of fluorescent medications to thecornea might alter the outcomes of these studies. Use of such agents priorto excimer laser ablation might alter the pattern and intensity of fluorescenceand related potential phototoxic reactions.

Among the most common ophthalmic drops used preoperatively and postoperativelyare fluoroquinolone antibiotic solutions.⁹ Infact, over the past decade, fluoroquinolones have become the antibiotic classof choice for most topical ophthalmic antibiotic indications. The fluoroquinolones,derivatives of nalidixic acid,¹⁰ are a fluorescentclass of drug, and, thus, could potentially be detected by fluorescence spectroscopy.¹¹

Another issue in photorefractive surgery that has been examined throughoutthe past dozen years is the risk of postoperative infection. Although a rarephenomenon (0.25%-1.2%),^12,13 infectiouskeratitis is a feared complication that is increasingly associated with unusualpathogens and pathogens resistant to the commonly used anti-infective agents.^9,14-20 Theemergence of resistant strains of Staphylococcus aureus, the most common cause of infectious keratitis, to the familiar fluoroquinoloneshas not helped alleviate these concerns. Although many techniques have beenused to assess the efficacy of anti-infective agents, drawbacks exist. Invitro tests fail to consider many factors such as drug penetrance into cornealtissue and the intrinsic anatomy of the eye. There have been numerous reportsthat have measured the antibiotic level in aqueous humor subsequent to aspiration.^21-27 Threegroups have determined the intracorneal concentration of drug only after excisionof corneal samples from patients undergoing penetrating keratoplasty.^28-30 Donnenfeld et al²⁸ enzymatically digested corneal samples. In articlesby McDermott et al²⁹ and Diamond et al,³⁰ the specimen corneas were mechanically pulverizedand analyzed spectroscopically after high-performance liquid chromatographyfor antibiotic concentration. To our knowledge, there have been no reportedstudies assessing the depth of penetration of fluoroquinolone antibioticsin corneal tissue as determined by real-time detection of excimer laser–inducedfluorescence during corneal ablation. This information is critical becauseit is the intracorneal level of antibiotic therapy that is thought to be themost important for optimal effectiveness against unforgiving pathogens involvedin infectious keratitis.

The purpose of this investigation was to measure the 193-nm excimerlaser–induced fluorescence spectrum of fluoroquinolone-treated cadaverrabbit cornea during excimer laser ablation. We sought to assess whether fluorescencedetection of fluoroquinolone antibiotics in the cornea was possible on a shot-to-shotdepth-dependent basis.

A schematic diagram of the experimental spectroscopy setup is shownin Figure 1. The main component is theArF excimer laser corneal surgery system (model EC-5000; Nidek TechnologiesInc, Pasadena, Calif) operating at 193 nm. The laser was operated in manualmode at 4 Hz, a spot size of 2 × 4.5 mm, and an average energy outputof 16.4 mJ/pulse. The laser-induced fluorescence generated from the ablatedcorneal layers is reflected by a 2-in off-axis parabolic mirror (AlSiO-coated,model A8037-331; Janos Technology Inc, Townsend, Vt) and focused in the entranceslit of the spectrograph (model SD2000; Ocean Optics, Dunedin, Fla). The long-passfilter commonly used to eliminate the excitation radiation in laser-inducedfluorescence spectroscopy was not needed because the mirror’s coatingdid not reflect the 193-nm radiation scattered from the surface of the cornea.The spectrometer was set to automatically measure and save the fluorescencespectrum. Because of the extremely high absorption of the 193 nm in the cornealtissue, the fluorescence signal is known to arise from the ablated corneallayer.^4,5 Assuming a linear rateof removal of tissue with each laser pulse, a constant ablation rate of 0.3 μm/pulsewas experimentally determined.

Fresh New Zealand white rabbit heads were obtained from a local slaughterhouseand kept at 4°C until the eyes were enucleated within 7 hours of deathto maximize preservation of the corneal epithelium. The globes were storedin a moist chamber at 4°C for approximately 4 hours until time of use.A total of 35 globes were used in this study.

An aminoglycoside (tobramycin sulfate) and 4 different fluoroquinolones(ofloxacin, levofloxacin, ciprofloxacin hydrochloride, and gatifloxacin) weretested, as these are some of the most common agents used for surgical prophylaxis.The applied solution concentrations were as follows: 0.3% tobramycin (Tobrex;Alcon Laboratories Inc, Fort Worth, Tex), 0.3% ofloxacin (Ocuflox; AllerganInc, Irvine, Calif), 0.5% levofloxacin (Quixin; Santen USA Inc, Napa, Calif),0.3% ciprofloxacin (Ciloxan; Alcon Laboratories Inc), and 0.3% gatifloxacin(Zymar; Allergan Inc). Sterile balanced salt solution (BSS; Alcon LaboratoriesInc) was used as a negative control. Four drops of BSS (n = 5),tobramycin sulfate (n = 5), ofloxacin (n = 5), levofloxacin(n = 3), ciprofloxacin hydrochloride (n = 3), or gatifloxacin(n = 4) were applied topically to separate corneas and allowed torest for 10 minutes. Three drops of the appropriate solution were then administereda second time and globes were allowed to rest for an additional 5 minutes.To wash away excess antibiotic, especially at the corneal surface, each globewas rinsed with approximately 2 mL of lactated Ringer solution (Baxter HealthcareCorp, Deerfield, Ill) subsequent to antibiotic administration and prior tolaser ablation. After washing, the globes were mounted on the laser platformand the corneas were ablated to perforation and the associated fluorescencespectra were measured.

In a separate preparation, each of the above antibiotics was dilutedto 0.06 μg/mL and topically applied to the corneas of another set of globes(n = 2 for each antibiotic, 10 total) at room temperature and allowedto rest for 5 minutes. Three drops of the appropriate solution were then administereda second time and globes were allowed to rest for an additional 30 seconds.Accordingly, the anterior (<1 μm) of the antibiotic-treated corneawould have an antibiotic concentration equal to the known concentration ofthe topically applied antibiotic solution of 0.06 μg/mL. The globes wereimmediately mounted on the laser platform and the corneas were ablated forabout 5 seconds; the resulting fluorescence spectra were measured. Assumingan ablation rate of 0.3 μm/pulse, the first 2 fluorescence spectra generatedby the first 2 ablative pulses would be arising from the anterior 0.6-μmcorneal layer and have the characteristic spectral features of corneal tissuesaturated with 0.06 μg/mL of the specific antibiotic.

An aliquot (ie, 5 drops) of each antibiotic solution and the BSS wasplaced in an aluminum container at the focal point of the laser. The solutionwas irradiated with the excimer laser and the generated autofluorescence wasmeasured.

The measured fluorescence spectra (250-650 nm) were corrected for thedark current of the detector and the background light. Each spectrum was smoothedusing a 5-point moving average window and resampled every 5 nm to reduce thenumber of spectral points from 379 to 76.

For each antibiotic tested, a partial least-squares (PLS) model wasdeveloped for the discrimination between healthy and antibiotic-treated cornealtissue. The model is basically a numerical matrix that when multiplied bythe fluorescence spectrum of an ablated corneal layer would yield a numericalvalue that is indicative of the antibiotic concentration in that layer. Detaileddescription of the PLS technique is available elsewhere.^31-33 Thetraining inputs of the model were the first 200 fluorescence spectra acquiredfrom the anterior 60 μm of healthy cornea (antibiotic free), and the first2 fluorescence spectra acquired from the anterior 0.6 μm of the corneastreated with the diluted antibiotic (ie, 0.06 μg/mL). The training outputsof the model were −1 for the antibiotic-free cornea and 1 for antibioticpresent with a concentration of 0.06 μg/mL and above. The PLS modelingalgorithm is implemented using the scientific programming language MATLAB(version 5.3; The MathWorks Inc, Natick, Mass). The model’s accuracyin detecting the antibiotics in the cornea was determined using the methodof cross validation. In the latter, one input-output pair is excluded fromthe input-output data set used in training the model. The trained model isthen used to predict the output (ie, antibiotic present/absent) for the excludedinput (ie, spectrum) and the result is compared with the actual output. Thisprocess is repeated for each input-output pair in the training data set andthe model’s prospective accuracy is calculated by:

% Accuracy = Number of Correctly Predicted Output/Number of Input-Output Pairs in the Data Set

Because the tested input-output pair was not used in training the model,this prospective accuracy is unbiased and represents the accuracy with whichthe model would detect the antibiotic’s presence or absence in the cornea.A separate model is developed for each antibiotic to be tested for cornealpenetration.

Starting with the fluorescence spectrum produced by the ablation ofthe topmost corneal layer, the model processes each subsequent spectrum toestimate the presence or absence of antibiotic in the corresponding corneallayer. As described earlier, the model is trained to detect antibiotics withconcentrations of 0.06 μg/mL or greater. The sequential number of the firstspectrum indicating antibiotic-free cornea is multiplied by the assumed ablationrate (ie, 0.3 μm/pulse or spectrum) to calculate the maximum penetrationdepth of the antibiotic.

The fluorescence spectra of 0.3% ofloxacin, 0.5% levofloxacin, 0.3%gatifloxacin, and 0.3% ciprofloxacin are shown in Figure 2. Each antibiotic exhibited a characteristic fluorescencespectrum with peaks between 400 and 500 nm. Levofloxacin produced the highestfluorescence intensity followed by ciprofloxacin, gatifloxacin, and ofloxacin,respectively. The spectral peak of ofloxacin, levofloxacin, and gatifloxacinoccurred at 483 nm, while ciprofloxacin exhibited double peaks at 433 and483 nm.

The fluorescence spectrum of 0.3% tobramycin is shown in Figure 2E. In contrast to the other 4 antibiotics, tobramycin exhibitedits peak fluorescence at about 300 nm. The fluorescence spectrum of tobramycincoincided with the autofluorescence spectrum of the corneal structural proteins(ie, elastin and collagen), which prevented its spectroscopic detection inthe corneal tissue.

The fluorescence spectra of corneal ablation are plotted vs ablationdepth in the 3-dimensional plots shown in Figure3. The corneas treated with 0.3% tobramycin exhibited a single-peakfluorescence spectrum that is similar to that of healthy cornea. On the otherhand, the corneas treated with 0.3% ofloxacin, 0.5% levofloxacin, 0.3% gatifloxacin,and 0.3% ciprofloxacin exhibited a characteristic double-peak fluorescencespectrum. The first peak at about 300 nm coincides with the natural fluorescenceof the structural proteins of the corneal tissue while the second peak atabout 483 nm coincides with the characteristic fluorescence of the antibiotic.

The fluorescence spectra of the diluted antibiotics in BSS are shownin Figure 4. The results indicate thateven at this low concentration (0.06 μg/mL), each antibiotic generatesmore than 4 times the fluorescence produced by BSS at the antibiotic’smaximum emission wavelength.

The PLS model was able to detect the presence of antibiotics in concentrationsof 0.06 μg/mL and above with a prospective accuracy of 90%. The maximumpenetration depth of the antibiotic into the corneal tissue is given in the Table. The mean ± SD penetrationto a detectable depth was as follows: 0.3% ofloxacin, 7.1 ± 3.0 μm;0.5% levofloxacin, 6.7 ± 1.4 μm; 0.3% ciprofloxacin, 1.2 ± 0.6 μm;and 0.3% gatifloxacin, 7.0 ± 1.9 μm. The prospective accuracyof the above measurements is about 90% with a detection sensitivity of 0.06 μg/mL.

Under these experimental conditions, none of the fluoroquinolone antibioticstested appear to have diffused significantly past the intact epithelium. Ofloxacin,levofloxacin, and gatifloxacin all seem to penetrate the epithelium significantlymore than ciprofloxacin (P<.02). Tobramycin absorptioncould not be detected because of the model’s inability to discriminatebetween its fluorescence spectrum and that of the cornea due to their closeresemblance.

We have determined that applied fluoroquinolone antibiotic preoperativelyprior to laser refractive surgery can certainly alter the excimer laser–inducedfluorescence patterns observed during surgery. Most of the doped fluorescenceoccurs between 400 to 500 nm. It is not clear that observed intensities atthese wavelengths in vitro translate into toxic irradiation in the clinicalsetting. Prospective human studies using clinical drug concentrations willneed to be performed to answer this question.

In this pilot study, we observed a detection sensitivity of better than0.06 μg/mL. In future experiments, we will need to determine the lowerlimits of detection in titration studies. However, even in these preliminarystudies, the lowest concentration of fluoroquinolone antibiotics tested, 0.06 μg/mL,is well below the minimum inhibitory concentrations of many ocular bacterialisolates.³⁴ Anticipated improvements in ourfluorescence collection and detection systems should result in even betterdetection limits than our current arrangement. Thus, we should be able todetect these antibiotics in the cornea at the level of their minimum inhibitoryconcentrations for many more bacterial species. Moreover, different than mostcurrent techniques used to determine the total corneal antibiotic depot, weshould be able to determine these values in a depth-dependent manner in vivo.Knowledge of the antibiotic concentration gradients will be invaluable duringlamellar corneal surgery, for example, LASIK and treatment of corneal infectionsat specific depths. If fluoroquinolones do not penetrate much past the epitheliumimmediately after application to intact epithelium, the epithelial levelsof fluoroquinolone would increase the fluorescence during transepithelialablation. This should not be a concern with LASIK.

Under our in vitro experimental conditions, the current topical fluoroquinoloneantibiotics did not penetrate very deeply past the corneal surface. Thus,repeated and frequent application might be needed to accumulate enough ofthe drug in the deeper layers of the cornea to reach therapeutic concentrations.Live animal and human trials are needed to determine antibiotic penetrationin vivo under normal conditions with intact epithelium, Bowman layer, andanterior stroma. Possible explanations for greater penetration of antibioticsdetected in living corneas include the repeated administration in many invivo studies and coapplication of epithelium-damaging drugs such as topicalanesthetics, as well as the intraoperative dehydration that occurs when theeyelids are opened with an eyelid speculum under the operating microscope.The resulting dehydration will cause the cornea to imbibe an antibiotic-containingsolution. Because we did not first desiccate our corneas under the operatingmicroscope, our experimental situation did not completely mimic the clinicalenvironment.

There are 2 major factors that enhance the statistical reliability (significance)of our small data set in this pilot study. First, the measured penetrationdepths for each drug seemed to be significantly close with narrow standarddeviations. Second, the number of corneal samples does not affect the accuracyof the penetration-depth measurement because this measurement is determinedfrom the hundreds of fluorescence spectra acquired during the corneal ablationprocess. The cross-validated accuracy of the penetration-depth measurementsexceeded 90%.

Initially, 3 globes were tested for each antibiotic. Based on grossvisual inspection of the initial unprocessed data, a few extra globes wereadded to validate that the fluoroquinolones did not penetrate past the epithelium(Table). The final spectral analysisof the large data sets for each globe proved that all of the measured spectrawere valid.

If we truly only achieve therapeutic levels of fluoroquinolone antibioticin the epithelium after topical drug administration shortly before surgery,then we may need to reconsider our standards of preoperative- and postoperative-dosingregimens. Methods to increase the corneal drug depot include increasing appliedtopical drug concentrations and frequency of dosing, and application onlyafter creation of a surgical wound when an intact epithelial barrier is nolonger present. Among the antibiotics tested, only 0.5% levofloxacin was notused near its maximum solubility concentration.

Finally, with respect to photorefractive keratectomy, if the therapeuticantibiotic depot does reside for the most part in the epithelium, then preoperativeantibiotic administration provides little advantage, as the epithelium isablated away during the initial stages of surgery. During this ablation ofdoped epithelium, one must also be wary of the generated secondary fluorescencederived from the fluoroquinolone dopant. In LASIK, however, the epithelium,and, thus, possibly the drug depot, remains intact for the most part. Becauseinfections after anterior segment surgery are typically caused by the patient’sown resident flora that may be recovered from the conjunctivae, eyelids, andnasopharynx, it would still be helpful in any case to apply these drops preoperatively,as they may be effective at reducing conjunctival flora.

Correspondence: Roy S. Chuck, MD, PhD, WilmerOphthalmological Institute, John Hopkins University, 3-127 Jefferson Building,600 N Wolfe St, Baltimore, MD 2/287-9278 (rchuck1@jhmi.edu).

Submitted for Publication: June 5, 2003; finalrevision received March 9, 2004; accepted April 29, 2004.

Financial Disclosure: None.

Funding/Support: This study was supported bygrants from Nidek Technologies Inc, Pasadena, Calif (Drs Chuck and Shehada)and grant EY00412-01A1 from the National Eye Institute, National Institutesof Health, Bethesda, Md (Dr Chuck).