|Molecular Vision 2002;
Received 26 August 2001 | Accepted 11 January 2002 | Published 25 January 2002
Evaluation of lectin staining in the diagnosis of fungal keratitis in an experimental rabbit model
M. Lourdes García,
J. M. Herreras, E. Dios, P. Argüeso,
IOBA, Instituto de Oftalmobiología Aplicada, Universidad de Valladolid, Spain
Correspondence to: J. M. Herreras, IOBA, Facultad de Medicina, C/Ramón y Cajal, Valladolid, 47005, Spain; email: firstname.lastname@example.org
Purpose: To assess the sensitivity, specificity, and reliability of peroxidase labeled lectin staining in the diagnosis of fungal keratitis in an experimental rabbit model.
Methods: Fungal keratitis by Candida albicans, Aspergillus fumigatus, and Fusarium solani was induced in rabbits. WGA-peroxidase staining of 660 corneal sections was performed. Fungal staining was evaluated independently by two observers. The test sensitivity, specificity, and reliability indexes were calculated.
Results: The sensitivity of the lectin staining test for Candida albicans was 100% (95% CI: 93.51-100.00), and specificity was 100% (95% CI: 93.51-100.00). The sensitivity of the test for Aspergillus fumigatus was 96.36% (95% CI: 86.46-99.35), and specificity was 100% (95% CI: 93.51-100.00). The sensitivity of the test for Fusarium solani was 96.36% (95% CI: 86.46-99.35) and specificity was 96.15% (95% CI: 85.74-99.31). There was also a high degree of test-retest and inter-rater concordance for all three fungi tested. The test-retest k reliability indexes were 0.9455, 0.9636, and 0.8879, for Candida albicans, Aspergillus fumigatus, and Fusarium solani, respectively. The inter-rater k reliability indexes were 0.9636, 0.9818, and 0.9252, for Candida albicans, Aspergillus fumigatus, and Fusarium solani, respectively.
Conclusions: WGA-peroxidase staining is a very sensitive, specific, and reliable test for the identification of fungi in an experimental rabbit model of fungal keratitis.
Fungal keratitis is an uncommon, but potentially serious ophthalmic infection . Frequently, the indolent nature and lack of pathognomonic clinical signs associated with these infections may delay diagnosis and initiation of appropriate therapy, possibly resulting in perforation and loss of the eye.
Corneal scrapings and cultures are frequently negative in diagnosing fungal keratitis, given the typically deep location of the microorganisms in the corneal stroma . Furthermore, cultures require a minimum of 48 to 72 h, and days to weeks in some cases. Definitive diagnosis frequently requires a corneal biopsy, obtaining a deep piece of tissue and histopathologic staining to detect the fungi . The most commonly used staining techniques to visualize fungi are Gomori methenamine silver and periodic acid-Schiff (PAS) [1,2]. These techniques are not absolutely specific for fungi. Methenamine silver staining can be very time consuming, taking up to three hours.
Lectins are proteins that bind specifically to carbohydrate groups, especially oligosaccharides and polysaccharides. Most lectins have more than one binding site, allowing their use in agglutination studies. The availability of a number of lectins with different carbohydrate specificities makes these proteins a useful tool in biochemistry, immunology and cellular biology.
There have been several studies on the biochemical composition of human [4-8] and animal [9,10] corneas using lectins, and significantly different results have been reported. Also several authors have published studies on fungal staining for the diagnosis of fungal keratitis using fluorescence conjugated lectins on a few specimens, also reporting differences in staining pattern for some fungi [11-14]. These differences have been explained based on variations in lectin concentrations and in fixation techniques. Lectins are very sensitive to small variations in tissue epitopes and the affinity of the lectin for its binding sugars in a tissue is also affected by surrounding carbohydrates.
Previous studies have utilized fluorescent labeled lectins  in the identification of microorganisms in a few specimens from corneal infections. We used peroxidase labeled lectins to provide rapid visualization of the three most common fungal pathogens in human keratitis  and we assessed the specificity and sensitivity of this test in an experimental model of fungal keratitis in the rabbit.
Fungal keratitis was induced in pigmented rabbits (Oryctolagus cuniculus). The most common pathogens responsible for fungal keratitis in patients were used: Aspergillus fumigatus, Candida albicans, and Fusarium solani. The inoculation procedure used was adapted from the method described by Ishibashi . Prior to inoculation, rabbits underwent general anesthesia with intramuscular ketamine (0.85 mg/kg) and xylacaine (0.15 ml/kg). Ringer's lactate (0.2 ml) was injected intrastromally using a 29 gauge needle to create a cavity in the corneal stroma, and subsequently, a second needle was used to inject 0.2 ml of a suspension containing the fungus into the cavity. Only one type of fungus was inoculated in each eye. Triamcinolone (10 mg) was injected subconjunctivally at the time of inoculation. Microorganisms were provided by the Microbiology Department of Hospital Clinico de Valladolid. Candida albicans specimens were type 14053 ATCC (American Type Culture Collection). Aspergillus fumigatus and Fusarium solani were isolated from human samples and cultured in Sabouraud dextrose agar at 37 °C.
Animal care guidelines published by the Institute for Laboratory Animal Research were followed. Only right eyes were used. Animals were euthanised under anesthesia and corneal buttons were obtained the fifth day after inoculation. Experimental procedures were in full compliance with ARVO guidelines on uses of animals in research.
Corneal buttons were immediately frozen in liquid nitrogen, and stored at -80 °C until being processed. Frozen corneal sections were cut at 8 mm, fixed in cold acetone (-20 °C) for 5 min and air dried.
Some corneas were fixed in 10% buffered formalin, embedded in paraffin, sectioned and placed onto glass slides. Tissue sections were deparaffinized by first immersing them in xylene for 10 min. This was followed by rapid rehydration in a series of graded alcohols.
Half the sections were stained using methenamine silver, and the other half with a lectin. We compared fungal staining with different peroxidase labeled lectins in this experimental model of fungal keratitis.
Sections were rinsed in Tris Buffered Saline (TBS) containing 1 mM CaCl2, 1 mM MgCl2, and 1 mM MnCl2, and incubated with 3% hydrogen peroxide solution for 5 min to block endogenous peroxidase activity. The sections were then immersed in a solution of 5% BSA (bovine serum albumin) for 10 min, rinsed in TBS and incubated with lectin. We used lectins conjugated with peroxidase, biotin, and digoxigenin. Detection of biotin and digoxigenin was achieved by using Streptavidin peroxidase and an anti-digoxigenin antibody directly conjugated to peroxidase, respectively (Table 1).
We assayed incubation times of 30 and 60 min, and lectin concentrations of 1, 5, 10, 50, 100, 500, and 1000 mg/ml, on control corneas and on corneas infected by each of the three species of fungi. In the second part of our experiment, in order to evaluate internal and external validity of this diagnostic test, we used Wheat Germ Agglutinin (WGA; 5 mg/ml), allowing 20 min for incubation.
For pokeweed mitogen (PKM), a second incubation (15, 30, and 60 min) with Streptavidin conjugated with peroxidase (5, 10, and 50 mg/ml) was necessary. For lectins labeled with digoxigenin a second incubation with antibodies anti-digoxigenin conjugated with alkaline phosphatase was done.
Staining was revealed using DAB (diaminobenzidine, brown) or AEC (aminoethylcarbazol, red) as chromogen for peroxidase, and NBT (nitroblue tetrazole chloride, black) for alkaline phosphatase. Glass coverslips were mounted using an aqueous medium for AEC and a lipophilic medium for the other chromogens.
Negative controls were performed for every lectin by incubating the lectin with its corresponding blocking sugar at 0.5 M concentration (Table 1) and substituting the lectin with TBS prior to the staining of the tissue.
Staining with methenamine silver method
Slides were stained according to standard techniques . The whole procedure required over 3 h.
Validity and reliability
We compared the staining obtained with WGA peroxidase to the staining obtained using methenamine silver, the current gold standard for the identification of fungi in corneal sections.
For every fungus, 220 correlative cryosections were obtained. A total of 660 corneal sections were analyzed. Sample size was estimated  for the desired sensitiviy and specificity at a significance level of a=0.05. In order to assess expected sensitivity and specificity, a preliminary study with 120 slides was performed. Correlative sections were evaluated; the odd numbers were stained with methenamine silver and the even numbers with WGA peroxidase. Half of the slides were positive for fungi and half of them were negative, according to methenamine silver staining. The slides stained with WGA were masked, randomized, and evaluated by two observers trying to identify yeast and hyphae in the tissue. Results were compared to those for methenamine silver, which is considered the gold standard.
Validity and reliability of the tests were estimated [17-21]. The first observer re-evaluated the same slides, two months later, to estimate the reproducibility of the procedure.
For every fungus, sensitivity, specificity, and likelihood ratios were estimated. Likelihood ratio of a positive result (LR+) is an estimation of how likely a positive result is among those who suffer the disease versus among those that are healthy. Values above 10 are considered significant. Likelihood ratio of a negative result (LR-) is an estimation of how likely a negative result is among those who suffer the disease versus those who are healthy. Values below 0.1 are considered significant.
We found important differences in fungal and corneal staining with different lectin concentrations (Table 2, Table 3, and Table 4). In some cases one of the chromogens yielded a more defined staining pattern, in general DAB caused more background staining than AEC. The identification of the microorganisms was easier in frozen sections than in paraffin specimens (Table 2 and Table 3).
We concluded that WGA peroxidase (5 mg/ml) was the lectin that yielded the best results in our study, in terms of less corneal background staining and easier fungal detection (Figure 1, Figure 2, and Figure 3).
In preliminary studies on non-infected corneal tissue, we had found that WGA peroxidase (1 mg/ml) would stain corneal epithelium brightly, barely stain Descemet membrane and the endothelium, and would not stain to stroma or keratocytes. A higher concentration of WGA (5 mg/ml) caused a slightly more evident staining of corneal stroma, but allowed for a better identification of the three fungi, especially Aspergillus and Fusarium. As for Candida, only some of the yeast present in the tissue would stain with WGA, but pseudohyphae were easily detected (Figure 1). Incubation time with WGA could be reduced to 20 min without losing fungal decoration.
Validity and reliability
In order to assess expected sensitivity and specificity, a preliminary study with 120 slides was performed. The sample size was estimated to be 51 slides, for each fungus, for each staining method, and for positive and negative results. Allowing for accidental losses of slides, sample size was increased to 55. A total of 660 slides were analyzed.
The first observer was able to identify Candida albicans in all corneal sections that contained the fungus. There were no false positive or false negative results (Table 5). Internal validity parameters were estimated according to the method described by Koopman . The sensitivity of the test for the detection of Candida albicans in the corneal tissue was 100% (95% CI: 93.51-100.00) and the specificity was 100% (95% CI: 93.51-100.00). Likelihood ratios were also calculated: LR+=110.0 (95% CI: 12.18-1043.45), LR-=0.00 (95% CI: 0.001-0.084).
Table 6 shows the results obtained for the identification of Aspergillus fumigatus. There were 53 of 55 slides in which evidence of fungus was found by the first observer. There were two false negatives, corresponding to two slides in which fungi were located in the epithelium, and hyphae staining were masked by WGA binding to epithelial cells. After comparing the slide to the corresponding methenamine silver stained slide, we determined that the observer failed to recognize fungi that had the same color as the epithelial tissue surrounding them. There were no false positive results. In no case did the first observer mistakenly identify fungi in slides that did not contain fungal structures by methenamine silver staining. The sensitivity of WGA peroxidase staining for the identification of Aspergillus fumigatus was calculated to be 96.36% (95% CI: 86.46-99.35), specificity was 100% (95% CI: 93.51-100.00), LR+ was 106.0 (95% CI: 11.84-1014.86) and LR- was 0.036 (95% CI: 0.01-0.12).
Results obtained with WGA staining for the identification of Fusarium solani were slightly worse than for the other two fungi, and they are displayed in Table 7. There were two slides in which the first observer failed to detect fungal hyphae in the tissue. After comparing the slide to the corresponding methenamine silver stained slide, we determined that there was poor staining of the fungi. There were two slides that were erroneously identified as infected tissue; these false positives were due to artifacts in the corneal stroma resembling filamentous structures, which appeared in two sections of the same slide.
Three slides were accidentally broken and could not be used. The sensitivity of WGA staining for the identification of Fusarium solani was 96.36% (95% CI: 86.46-99.35), specificity was 96.15% (95% CI: 85.74-99.31), LR+ was 25.05 (95% CI: 7.41-90.88) and LR- was 0.037 (95% CI: 0.01-0.12).
Inter-rater and test-retest reliability
The 327 WGA stained slides were masked and evaluated by a second observer. These results were compared to the evaluation of the first observer (Table 8, Table 9, and Table 10). For each fungus, concordance was assessed with the k index and McNemar test. The 327 WGA stained slides were masked and evaluated by the first observer two months later. These results were compared to the initial evaluation by the first observer. (Table 11, Table 12, and Table 13). For each fungus, concordance was assessed with the k index and McNemar test.
We assayed different conditions for the lectins that according to the literature have been most consistently useful in the detection of fungi. We selected the lectin that allowed for the brightest staining and we tested this as a diagnostic test in fungal keratitis.
WGA is a lectin obtained from Triticum vulgaris. It has affinity for N-acetylglucosamine oligosacharides and sialic acid. Chitin is a unique polysaccharide of N-acetylglucosamine, present in the wall of fungi, algae, and arthropods, which has not been found in mammals. The high density of N-acetylglucosamine in fungi and the low density of sialic acid and N-acetylglucosamine in corneal tissue make WGA a lectin potentially very useful for the identification of fungi in corneal stroma from mammals, since WGA would bind specifically to the fungus, and poorly to corneal stroma.
Fluorescein conjugated WGA (FITC-WGA) has been assayed in the past for the identification of chitin in the walls of filamentous fungi, such as Aspergillus, Fusarium, and Rhizopus; poorer results were reported for Candida albicans . Biochemical studies suggest that the cell wall of Candida has an external layer poor in N-acetylglucosamine residues and an internal wall that contains this sugar. A study  reported that blastoconidias and pseudohyphae were able to stain brightly with WGA in specimens obtained after 7 h incubation. This pattern of staining was different in specimens obtained after 18 h culture, pseudohyphae and septae would still bind WGA, but not blastoconidias. These results suggest that as the fungus matures, it elaborates a superficial layer that is able to mask chitin oligomers. According to this, the yeast that stained in our study were probably the youngest microorganisms. In the clinic, this would mean that WGA staining would be stronger in infections caused by most virulent fungi, with a more active growing of new hyphae. This preferential staining of younger forms of hyphae has also been described for filamentous fungi, but does not interfere with the detection of mature microorganisms in corneal tissue, according to the results of our study.
Usually lectin binding is better in frozen tissue than in deparaffinized histopathologic specimens. This is due to the fact that during the processing of the tissues for paraffin inclusion and deparaffination many carbohydrate structures are sequestered and some glycolipid residues are lost.
Previous studies have used fluorescein-conjugated lectins, requiring a fluorescent microscope for the identification of lectin binding to the tissue. Also, it has been suggested that a property of fungi, auto fluorescence, could interfere with the interpretation of the binding of FITC labeled lectins to fungi. On the other hand, peroxidase labeled lectins can be easily revealed using a chromogen and visualized with a light microscope. We used peroxidase labeled lectins to provide an easy visualization of the three most common fungal pathogens in human keratitis and we assessed the internal and external validity of this diagnostic test in an experimental model of fungal keratitis in rabbit.
Sensitivity and specificity of WGA peroxidase labeled staining were above 95% for all three fungi. According to these results WGA peroxidase can be considered a very sensitive and specific test for the identification of fungi in cryosections of an experimental model of mycotic keratitis.
These results may have been favored by the high number of microorganism present in the tissue, especially for Candida albicans. As is shown in Figure 1, only some of the yeast and most pseudohyphae stained brightly; possibly, if there had been fewer microorganisms in the tissue, results would have been poorer. Aspergillus fumigatus and Fusarium solani could be more readily identified than Candida albicans since hyphae stained with more intensity (Figure 2, Figure 3).
Further studies including infectious keratitis of different etiologies and human tissues would be necessary to assess the value of this staining technique in human mycotic keratitis.
The k concordance index were higher than 0.9 in all but one case, and McNemar test resulted in p>0.05 in every case, meaning that agreement between different observers and at different times was quite strong, despite some discrepancies [17,19-21]. These results allow us to consider this technique as a very reliable and reproducible test for the identification of fungi in an experimental model of fungal keratitis in the rabbit.
WGA peroxidase labeled staining could be an alternative for the identification of fungi on frozen sections of mycotic keratitis, especially when rapid identification is necessary in order to begin specific treatment. Frozen sections are a much quicker procedure than paraffin embedding. This staining technique would allow identification of fungal etiology one hour after surgery. If methenamine silver staining was used, results would not be available for four hours, which could delay specific therapy until the following day.
Peroxidase labeled WGA could be an easy to perform and quick second staining procedure in cases of non conclusive results, since artifacts do not appear in two different staining techniques. It can be argued that staining corneal scrapings would be more convenient for the clinician, avoiding the need for a corneal biopsy. However, the limitations of this kind of diagnostic approach have been proven before . Not so much because of the quality, sensitivity or specificity of the staining method used to detect the fungus, but because these fungal infections are usually deep in the stroma, and superficial scrapings produce only corneal tissue or debris without microorganisms.
1. Arffa RC. Infectious keratitis: fungal and parasitic. In: Arffa RC, editor. Grayson's diseases of the cornea. St Louis: Mosby; 1991. p. 199-213.
2. Naumann G, Green WR, Zimmerman LE. A histopathologic study of 73 cases. Am J Ophthalmol 1967; 64:668-82.
3. Foster CS. Fungal keratitis. Infect Dis Clin North Am 1992; 6:851-7.
4. Brandon DM, Nayak SK, Binder PS. Lectin binding patterns of the human cornea. Comparison of frozen and paraffin sections. Cornea 1988; 7:257-66.
5. Rittig M, Brigel C, Lutjen-Drecoll E. Lectin-binding sites in the anterior segment of the human eye. Graefes Arch Clin Exp Ophthalmol 1990; 228:528-32.
6. Bishop PN, Bonshek RE, Jones CJ, Ridgway AE, Stoddart RW. Lectin binding sites in normal, scarred, and lattice dystrophy corneas. Br J Ophthalmol 1991; 75:22-7.
7. Philipp W. Altered lectin binding sites in keratoconus corneas. Curr Eye Res 1992; 11:397-409.
8. Tuori A, Virtanen I, Uusitalo R, Uusitalo H. Lectin binding in normal, scarred and keratoconus corneas. Cornea 1998; 17:88-98.
9. Panjwani N, Baum J. A histochemical comparison of human stromal glycoconjugates with eight other species. Distinct species-dictated differences in binding sites of Griffonia simplicifolia I. Histochemistry 1988; 89:41-5.
10. Panjwani N, Moulton P, Alroy J, Baum J. Localization of lectin binding sites in human, cat and rabbit corneas. Invest Ophthalmol Vis Sci 1986; 27:1280-4.
11. Karayannopoulou G, Weiss J, Damjanov I. Detection of fungi in tissue sections by lectin histochemistry. Arch Pathol Lab Med 1988; 112:746-8.
12. Slifkin M, Doyle RJ. Lectins and their application to clinical microbiology. Clin Microbiol Rev 1990; 3:197-218.
13. Robin JB, Arffa RC, Avni I, Rao NA. Rapid visualization of three common fungi using fluorescein-conjugated lectins. Invest Ophthalmol Vis Sci 1986; 27:500-6.
14. Pistole TG. Interaction of bacteria and fungi with lectins and lectin-like substances. Annu Rev Microbiol 1981; 35:85-112.
15. Ishibashi Y, Kaufman HE. Corneal biopsy in the diagnosis of keratomycosis. Am J Ophthalmol 1986; 101:288-93.
16. García del Moral R, Aguilar M. Métodos para la detección de microorganismos. In: García del Moral R, editor. Laboratorio de anatomía patológica. New York: Interamericana/McGraw-Hill; 1993. p. 288-289.
17. Feuer EJ, Kessler LG. Test statistic and sample size for a two-sample McNemar test. Biometrics 1989; 45:629-36.
18. Koopman PAR. Confidence intervals for the ratio of two binomial proportions. Biometrics 1984; 40:513-7.
19. Feinstein AR, Cicchetti DV. High agreement but low kappa: I. The problems of two paradoxes. J Clin Epidemiol 1990; 43:543-9.
20. Cicchetti DV, Feinstein AR. High agreement but low kappa: II. Resolving the paradoxes. J Clin Epidemiol 1990; 43:551-8.
21. Donner A, Li KY. The relationship between chi-square statistics from matched and unmatched analyses. J Clin Epidemiol 1990; 43:827-31.
22. Stoddart RW, Herbertson BM. The use of fluorescein-labelled lectins in the detection and identification of fungi pathogenic for man: a preliminary study. J Med Microbiol 1978; 11:315-24.
23. Ponton J, Jones JM. Analysis of cell wall extracts of Candida albicans by sodium dodecyl sulfate-polyacrylamide gel electrophoresis and Western blot techniques. Infect Immun 1986; 53:565-72.