Molecular Vision 2007; 13:1169-1180 <>
Received 4 April 2007 | Accepted 1 July 2007 | Published 17 July 2007

LPS-stimulated inflammation and apoptosis in corneal injury models

Hong Liang,1,2 Françoise Brignole-Baudouin,2,3 Antoine Labbé,1,2 Aude Pauly,2 Jean-Michel Warnet,1,3 Christophe Baudouin1,2

1Department of Ophthalmology III, Quinze-Vingts National Ophthalmology Hospital and Ambroise Paré Hospital, APHP, University of Versailles, Paris, France; 2INSERM, UMR S 872, Cordeliers, University Paris Descartes, France; 3Department of Toxicology, Faculty of Biological and Pharmacological Sciences, University Paris Descartes, France

Correspondence to: Christophe Baudouin, MD, PhD, Department of Ophthalmology III, Quinze-Vingts National Ophthalmology Hospital, 28 Rue de Charenton, 75012, Paris, France; Phone: (33) 1 40 02 13 04; FAX: (33) 1 40 02 13 99; email:


Purpose: To evaluate and compare the proinflammatory and apoptotic effects of lipopolysaccharide (LPS) in three rabbit corneal injury models using a new in vivo confocal microscope (IVCM) and immunohistological techniques.

Methods: Adult male New Zealand albino rabbits were used in this study. Three corneal models were tested: corneal incision, corneal epithelium scraping, and corneal suture. Ten rabbits were used in each model and these three groups were subdivided into two subgroups: with or without LPS instillation (with saline used as control) for eight days. Rabbit corneas were analyzed in vivo by using the Rostock Cornea Module (RCM) of the Heidelberg Retina Tomograph (HRT)-II. Immunohistology was used to evaluate inflammatory, proliferating, and apoptotic cells in the different injury models following saline or LPS instillations.

Results: Clinically, LPS induced earlier and higher levels of inflammation and corneal neovascularization in eyes subjected to scraping and suturing compared to saline. The RCM/HRT successfully presented high-quality images allowing analysis of all pathological corneal layers. Compared to groups receiving saline, LPS caused earlier and greater surface and stromal inflammatory infiltration as well as neovascularization. Immunohistology was correlated with in vivo findings and confirmed these results by showing greater infiltration of KI 67+ proliferating cells, TUNEL+ apoptotic cells, and TNF-α+, TNFR1+, TLR4/MD2+, ICAM-1+, RLA-DR+, CD11b+, and CD11c+ inflammatory cells, in eyes receiving LPS compared to those receiving saline.

Conclusions: These results indicate that in various models of corneal injury, LPS is a potent proinflammatory stimulus and its exposure has major effects on determinants of inflammation, angiogenesis, and apoptosis.


Although less common than Gram-positive (GP) bacterial infections, Gram-negative (GN) bacterial infections often lead to more serious ocular surface inflammation with a high risk of visual impairment and even blindness [1,2]. GN bacterial infections are characterized by a rapidly progressing, suppurative stromal infiltration with marked mucopurulent exudates [3]. GN bacteria such as Pseudomonas aeruginosa (P. aeruginosa) are the most common bacteria isolated from contact lenses, which remain important causes of highly destructive keratitis and neovascularization [3,4]. Moreover, the successful treatment of GN bacteria is being challenged by increasing antibiotic resistance [5].

In addition to direct infectious diseases induced by GN species, lipopolysaccharide (LPS or endotoxin), a highly conserved molecular pattern of GN bacteria, has also been suggested to play a direct role in severe ocular surface pathologies [3] such as delayed corneal wound healing, complications after corneal surgery [6,7], or aggravation of certain infectious situations [8]. LPS induces the release of a number of proinflammatory cytokines, i.e. interleukin(IL)-1, IL-6, and tumor necrosis factor (TNF)α [9]. LPS also activates the innate immune system via CD14 and toll-like receptors (TLRs) [10]. Therefore, it is an important environmental agent that induces and aggravates ocular bacteria-related inflammatory reactions and is directly involved in keratitis, conjunctivitis, and uveitis. LPS also interferes with wound healing, which could cause severe concerns after refractive surgery. In a rabbit laser in situ keratomileusis (LASIK) model, LPS was capable of reproducing diffuse lamellar keratitis(DLK)-like inflammation by stimulating production of IL-8 when applied beneath the corneal flaps [6]. Conversely, endotoxin blockers such as polymyxin were effective in decreasing DLK incidence [7]. LPS-related TNF production was shown to inhibit growth factors and decrease collagen production, thus impairing the wound healing process in skin wounds [11].

Human cornea, conjunctiva, and uvea were shown to express LPS receptor proteins such as CD14 and TLR4 [12,13]. Corneal inflammatory response triggered by LPS is characterized by neutrophil (polymorphonuclear leukocytes, PMNs) and macrophage recruitment [14,15]. Recently, within an acute inflammation model of subconjunctival injection of LPS in rabbits using an in vivo confocal microscopy (IVCM) technique, we demonstrated the time sequence of inflammatory infiltration [16]. Co-injection of neutralizing anti-TNF-α in the conjunctiva could significantly reduce LPS-induced inflammation and apoptosis in the epithelium and substantia propria.

After the study of LPS in conjunctiva, we wished to continue our research by seeking LPS potentiation of neovascularization, inflammation, apoptosis, and cicatrization in various models involving the cornea. We studied the effects of LPS in three types of corneal injuries in rabbits: deep transversal incision, epithelium scraping, and corneal suture. In the literature, the incision models were used to mimic radial keratotomy or simple corneal incision while scraping was used as a photorefractive keratectomy (PRK) model and suture placements were considered to mimic corneal allograft suture and a model for inducing corneal neovascularization. The effects of LPS on cornea were previously studied following injection into the corneal stroma [14,15]. Here, we used topical instillation of LPS, considering that direct contact between LPS and the wounded tissue would be more similar to the environmental stimulus. We used LPS derived from Escherichia coli (E. coli) because on corneal fibroblasts, it had a greater effect than P. aeruginosa. Probably because LPS from P. aeruginosa has a smaller lipid A component [17-19]. LPS was instilled directly onto the ocular surface in these three types of corneal wounds and saline instillation was used as control. We quantified in vivo the inflammatory processes using IVCM and completed with standard immunohistology to confirm and analyze the inflammatory and apoptotic effects of LPS in corneal injuries.



Adult, male, 10-month-old New Zealand albino rabbits weighing 2.5-3 kg with ocular surface integrity observed by a slit lamp biomicroscope were used. All animals were treated according to the Association for Research in Vision and Ophthalmology Resolution on the Human Use of Animals in Vision Research under the supervision of a health authority-accredited staff member for animal care and management. Considering the damage to the eyes of rabbits, especially those in LPS-instilled scraping and suture models, we used just one eye of each rabbit to do the experiments. Before all experiments, rabbits were anesthetized by subcutaneous injection of ketamine (35 mg/kg, Imalgéne 500; Merial, Lyon, France) and xylazine (5 mg/kg; Bayer, Puteaux, France). A total of 30 rabbits were divided into six groups. In each group, five rabbits were used for IVCM observation at 30 min (M30), 1 h, 4 h (H1 and H4), and from day(D)1 to D9; three rabbits were then observed from D10 to D30. Two rabbits from each treatment group were sacrificed for immunohistological procedures at D9, a time point chosen for optimally observing neovascularization according to a previous preliminary study in three rabbits (data not shown). The scores were assessed by two ophthalmologists in a masked manner. The final score was the average of these two scores.

Corneal models and saline or lipopolysaccharide instillation

Three corneal models consisting of corneal incision, corneal epithelium scraping, and corneal suture were reproduced in one eye of 30 rabbits (ten for each type of injury). A simple linear incision was made in the cornea by a cut using a sterile 25-G needle from the limbus and directed toward the center of the cornea. Incision length (7 mm) was measured with a caliper and depth (midstroma) was controlled under an operating microscope. Epithelial scraping was performed using a mechanical scraping of the corneal epithelium with a sterile blade under an operating microscope. The superior corneal epithelium was removed, leaving an untreated area 0.5 mm along the limbus. In the corneal suture model, three 10-0 nylon nonabsorbable monofilament sutures were placed in the rabbit cornea at midstromal depth, parallel to the limbus, and approximately 5 mm along the limbus at the two, ten, and twelve o'clock positions.

The three corneal injury models were divided into two subgroups: with LPS instillation (from Escherichia coli, Sigma Aldrich, St. Louis, MO) and endotoxin-free saline as control. We administered 25 μl of endotoxin-free saline or 25 μl of LPS solution at a concentration of 50 μg/ml [6] onto the rabbit's ocular surface immediately after corneal surgery and then every day from D1 to D8 at the end of IVCM examination. Saline and LPS instillations were then stopped between D9 and D30 in the three remaining animals in each group in order to observe corneal recovery.

Clinical findings and Draize test

Using slit lamp microscopy, the eyes were examined for irritation and scored according to a weighted scale for grading the severity of ocular lesions (modified scores from the Draize test) [20]. The conjunctiva was evaluated for degrees of redness, swelling (chemosis), and discharge. The cornea was evaluated for degree and area of opacity. The iris was assessed for increased prominence of iridal folds, congestion, swelling, etc. The maximum total score possible was 110 (conjunctiva=20, cornea=80, iris=10).

Neovascularization scoring

Corneal neovascularization was observed by slit lamp microscopy after corneal surgeries from D1 to D9 (five eyes in each group) then until D30 (three eyes). To give an overall assessment of its extent, neovascularization was scored from zero to four according to a previous study [21]. The scores were recorded by the length of newly formed corneal vessels: zero, no vessels; one, vessels only in the peripheral cornea and extension less than 1 mm; two, vessel extension less than 2 mm from the limbus but higher than 1 mm; three, vessel extension between 2 and 3 mm; four, vessel extension greater than 3 mm.

In vivo confocal microscopy evaluation

A recently developed laser scanning IVCM, the Heidelberg Retina Tomograph (HRT) II/Rostock Cornea Module (RCM, Heidelberg Engineering GmbH, Heidelberg, Germany), was used, as described previously, on humans [22,23] and on animals [16,20,24]. Briefly, the HRT II camera was disconnected from the head rest and maintained in a vertical position. The objective of the microscope, magnification 60X, numerical aperture 0.90 (Olympus, Hamburg, Germany), covered by a polymethyl methacrylate cap, was used to evaluate the corneal injuries. Images comprise 384x384 pixels covering an area of 400x400 μm with a transversal optical resolution of approximate 1 μm/pixel and an acquisition time of 0.024 s (Heidelberg Engineering). The x-y position and the depth of the optical section were controlled manually and the focus position (μm) was automatically calculated by the HRT II/RCM.

Rabbits were positioned on their side and held in place to align the central cornea parallel to the objective tip. A drop of gel tear substitute (Lacrigel®, carbomer 0.2%; Europhta, Monaco) was placed on the tip of the objective lens. For all eyes, at least ten confocal microscopic images of each layer were recorded by focusing the microscope on the cornea.

Using the Cell Count® software associated with the HRT II/RCM, we counted the number of infiltrating inflammatory cells (lymphocytes, polymorphonuclear cells, or dendritic-like cells) in corneal stroma and basal epithelium as in our previous study [16]. The number of marks of each image was counted by the computer and cellular densities were expressed as cells per square millimeter (cells/mm2). Cellular densities were calculated on the five to ten confocal microscopy images for each corneal layer of each eye and the final scores were the average of three (D9 to D30) to five (M30 to D9) rabbits.

Immunohistology on cryosections

Two rabbits in each group were euthanized with a lethal dose of pentobarbital at D9, the day following the last saline or LPS instillation. Enucleated eyes were fixed in 4% paraformaldehyde, embedded in an optimal cutting-temperature (OCT) compound (Tissue-Tek®; Miles Inc., Bayer Diagnostic, Puteaux, France). The 8 μm cryosections were incubated with antibodies to TNF-α (1:50; clone 6402; R&D Systems, Wiesbaden-Nordenstadt, Germany), TNFR1 (1:50; clone 16803; R&D Systems), TLR4/MD2 (1:50; Santa Cruz Biotech, Santa Cruz, CA), class II antigen RLA-DR (1:50; clone TAL.1B5; DAKO, Copenhagen, Denmark), ICAM-1 (1:200; clone 6.5B5; DAKO), CD11b, CD11c (Immunotech, Marseilles, France), KI 67 (1:50; Nuclear Antigen Ki-67, Immunotech), and with mouse IgG1 as negative control (Immunotech). Sections were stained with Alexa Fluor®488 goat anti-mouse as secondary antibody (1:250; Molecular Probes, Montluçon, France) for one h and later with propidium iodide (PI, Sigma Chemical Co., Saint Louis, MO). Images were digitized using an Olympus BX-UCB fluorescent microscope (Olympus, Melville, NY), equipped with DP70 Olympus digital camera and image analysis software, to determine the total number of cells positive to the different markers. Cells were counted in a masked manner in at least five 100x100 μm areas for corneal stroma and 100 μm long for epithelial layers.

Apoptosis evaluation with terminal deoxynucleotidyl transferase-mediated dUTP-nick end labeling staining

A terminal deoxynucleotidyl transferase-mediated dUTP-nick end labeling (TUNEL) assay (Roche Diagnostics, Meylan, France) was used according to the manufacturer's instructions. Cryosections were permeabilized with a 0.1% Triton X-100-0.1% sodium citrate (2V:1V) solution for two min and then incubated with an apoptosis detection kit including the 10 μl TUNEL enzymes and 90 μl TUNEL label at 37 °C for one h. After three washes in PBS, the slides were stained with DAPI, observed, and counted under the fluorescence microscope.

Statistical analysis

The groups were compared using the Mann-Whitney test for Draize test and neovascularization scores and factorial analysis of variance (ANOVA) followed by the Bonferroni/Dunnet method for cell counts (Statview V; SAS Institute Inc., Cary, NC).


Saline- or lipopolysaccharide-instilled corneal incision models

Clinically, 4 h after corneal incision, compared to saline instillation (Figure 1A), LPS-instilled corneas presented more redness and purulent secretions (Figure 1B). At D9, the saline plus incision eyes returned to a normal aspect (Figure 1C) whereas LPS plus incision eyes still presented slight corneal opacity (Figure 1D) with delayed corneal wound healing. Between D1 and D15, according to the Draize test, LPS instillation provoked more ocular inflammation than did saline (Figure 2A, p<0.05). The saline plus incision eyes presented an almost normal ocular surface aspect at D4, which remained until the end of the experiment with no obvious ocular irritation. However, in the LPS plus incision model, animals presented a higher Draize score during instillation time, which started to return to a normal aspect at D15 and continued returning normal until the end of the experiment. In the two incision models, no corneal neovascularization was observed at D9 (Figure 3) or during the entire experimental process.

In the corneal incision model, IVCM (Figure 4) showed that the earliest cellular changes (H1-H4) were keratocytes adjacent to the cut edge becoming invisible (H4; Figure 4A). Infiltration of inflammatory cells started already at H1 in the LPS-instilled incision. Within three to four h, abundant inflammatory cells appeared near the incision (H4; Figure 4B). However, the saline-instilled model showed mild and delayed inflammatory infiltrates. We counted inflammatory cell infiltrates at H4 in the cornea stroma (Figure 5A) and found a highly significant difference between the two models: 745.8±55.5 cells/mm2 for LPS plus incision versus 55.3±11.5 cells/mm2 for saline plus incision (p<0.0001). By D1-D2, corneal reepithelization had already occurred in the lesional zone both in saline- (D1; Figure 4A) and LPS-instilled eyes (D1; Figure 4B). At this time, more numerous inflammatory cells could still be found inside at the edge and outside the incisional lesion in eyes that had received LPS (953.3±127.0 cells/mm2 for LPS versus 144.3±20.2 cells/mm2 for saline, p<0.0001). At D9, compared to the early time points, the inflammatory infiltrates decreased but still remained significantly different: 320.5±26.0 cells/mm2 for LPS plus incision (D9; Figure 4B) versus 70.3±11.1 cells/mm2 for the saline plus incision model (D9; Figure 4A; p<0.001). Later, from D15 to the end of the experiment, there was no obvious inflammatory infiltration observed near the two incisions (Figure 5A). In basal epithelial layers, the inflammatory cell counts were consistent with those found in the stroma (Figure 6): from D2 to D6, LPS plus incision induced significantly more inflammatory infiltration than did saline plus incision (p<0.01 from D2 to D6). From D9 to D30, no obvious inflammatory infiltration in the subepithelial layer was observed in the two incision models.

Immunohistology was used at the end of instillation to investigate markers of inflammation, proliferation, and apoptosis in rabbit corneas. Figure 7A shows the results in cryosections of saline plus incision and LPS plus incision eyes at D9. There were no significant differences between the two groups in inflammatory (TNF-α, TNFR1, TLR4/MD2, RLA-DR, ICAM-1, CD11b, and CD11c), proliferating (KI 67), or apoptotic (TUNEL) cells. We observed that even in LPS plus incision eyes, there were few cells positive for TLR4/MD2 (Figure 8B) or TUNEL (Figure 8D) immunostainings.

Saline- or lipopolysaccharide-instilled corneal scraping models

In scraping models at H4, compared to saline-treated corneas (Figure 1E), the LPS-instilled eyes (Figure 1F) were characterized by severe inflammatory reactions with chemosis, redness, and especially abundant purulent secretions in the ocular surface and adjacent to the eyelid. During the entire instillation period (H1 to D8), LPS induced more ocular inflammation and redness than did saline (Figure 2B; p<0.05). From D4 to D9 (Figure 1G for saline plus scraping at D9, and Figure 1H for LPS plus scraping at D9), LPS also induced a significantly higher neovascularization score (Figure 3A) than did saline (p<0.05). After stopping saline or LPS instillations, there was no difference in neovascularization scores in the two models until the end of the experiment.

In the scraping model eyes observed by IVCM, we found that at H4 the first reactions in response to mechanical scraping were keratocytes becoming larger and more reflective and the presence of numerous hyper-reflective stellate structures (H4; Figure 4C). Compared to saline-treated eyes at H4, the LPS-instilled eyes presented more inflammatory infiltration with long strands of inflammatory cells stretching from the scraping periphery toward the lesion center (H4; Figure 4D). From H4 to D4, at all time points, LPS induced significantly more inflammatory cell infiltrates in corneal stroma than did saline (Figure 5B, p<0.0001). From D6 (1,096.5±147.2 cells/mm2 in LPS plus scraping versus 894±114.1 cells/mm2 in saline plus scraping) to the end of the experiments (D30), differences were no longer significant between saline- and LPS-treated scraping models (Figure 5B). At D9, the two models developed neovascularization (D9; Figure 4C,D) but with no difference in inflammatory infiltration cells until the end of observation times (Figure 5B). In basal epithelial layers, we observed numerous dendritic-like inflammatory cells during the entire LPS instillation procedure. Their number reached maximal values at D9 (Figure 6): 265±9 cells/mm2 in LPS plus scraping versus 114±17 cells/mm2 in saline plus scraping (p<0.0001). A significant difference between saline- or LPS-instilled inflammatory infiltrates in basal epithelial layers was observed throughout the experimental period even at D30, 22 days after stopping LPS instillation (Figure 6, p<0.0001).

Histologically, all inflammatory, proliferating, and apoptotic markers showed significant differences between saline- and LPS-instilled scraping eyes (Figure 7B, p<0.01). We observed that there were more TLR4/MD2-positive cells, especially in the corneal stroma, in the eyes of LPS plus scraping (Figure 8F) than those of saline plus scraping (Figure 8E). These cells could be keratocytes or inflammatory cells. The other inflammatory markers such as TNFα, TNFR1, RLA-DR, ICAM-1, CD11b, and CD11c were also found more abundantly in corneal layers in LPS-associated scraping than in saline-treated eyes. KI-67-positive cells were especially numerous in basal epithelial layers in the two scraping models, but the number was much higher in LPS plus scraping (Figure 7B, 740±48 cells/mm2 versus 460±74 cells/mm2 in saline plus scraping, p<0.01). At D9, TUNEL-positive apoptotic cells in saline plus scraping (Figure 8G) were notably located in the basal epithelial layer or anterior stroma whereas, in the eyes subjected to LPS-instilled scraping, the cells were found more abundantly and were located in all corneal (Figure 8H). These apoptotic cells were possibly keratocytes or inflammatory cells, but were not endothelial cells of new blood vessels (double staining of TUNEL and CD31 immunostaining, data not shown).

Saline- or lipopolysaccharide-instilled corneal suture models

As in the eyes that underwent scraping, four h after corneal sutures, compared to eyes receiving saline (Figure 1I), the LPS-treated suture eyes (Figure 1J) were characterized by a more pronounced inflammatory reaction with chemosis, redness, and substantial purulent secretions. The Draize test (Figure 2C) showed that ocular inflammation from H1 to D6 were significantly greater in LPS plus suture than in saline plus suture eyes (p<0.05). At D9, the LPS plus suture eyes (Figure 1L) still showed more active inflammation than did saline-treated eyes (Figure 1K). During the observation time, LPS plus suture induced an earlier onset of neovascularization (starting at D3-D4) than did saline (starting at D5-D6; Figure 3B). During the experiments, LPS treatment induced higher neovascularization scores (p<0.05 versus saline treatment from D9 to D20). Twelve days after stopping instillations (at D20), we still found higher neovascularization scores in LPS-instilled eyes (Figure 3B; p<0.05 at D20 versus saline treatment). At D30, the two models presented almost the same neovascularization scores.

In the saline plus suture eyes, no obvious inflammatory infiltration was observed at H4 by IVCM (H4; Figure 4E). However, numerous inflammatory cells, dendritiform in shape or round and hyper-reflective and most likely PMNs, accumulated near the LPS-instilled suture at H4 (H4; Figure 4F) and their number increased at D1 (Figure 5C; p<0.0001). Elongated, bright, and spindle-shaped structures were also found near the suture one day (D1; Figure 4E) after suture placement, suggesting migratory fibroblast-like cells in response to a suture-induced mechanical attraction. In LPS-instilled eyes, because of the strong global hyper-reflectivity of the tissue resulting from the brightness of abundant inflammatory cells, we could not clearly observe these elongated spindle-shaped fibroblast-like structures. Inflammatory cells also arranged in line beside the suture, possibly influenced by migratory fibroblast arrangement (D1; Figure 4F). On D9, in the saline plus suture eyes near the suture point, aligned inflammatory cells were observed with no obvious neovascularization in the suture area (D9; Figure 4E). However, at that time, neovascularization in LPS plus suture eyes had already reached and even passed the suture (D9; Figure 4F) with more numerous inflammatory cells (Figure 5C; 1,161.3±145.1 cells/mm2 versus 862.3±52.5 cells/mm2 in saline plus suture eyes, p<0.01). From H4 to D20, LPS-instilled suture treatment always induced more inflammatory cells in the cornea stroma than did saline (p<0.01). After that time point, we counted no difference in stromal inflammatory infiltrates in the two models until the end of the experiment (D30). In basal epithelial layers, consistent with the eyes subjected to scraping, LPS induced more inflammatory cell infiltration (Figure 6, 386±11 cells/mm2 at D9) than did saline (157±17 cells/mm2, p<0.0001 at D9) and this difference in inflammatory infiltration lasted until the end of the experiment (p<0.0001 at all time points).

Immunohistology also showed important inflammatory, proliferative, and apoptotic markers in LPS-instilled sutured eyes (Figure 7C, p<0.0001 compared to saline-instilled sutured eyes). In corneal cryosections of saline plus suture eyes (Figure 8I), TLR4/MD2-positive cells were observed especially in the neovascularization area. However, in LPS plus suture eyes (Figure 8J), they were found not only in the neovascularization area but also in the other corneal stroma layers. In LPS plus suture eyes, the other inflammatory markers (TNFα, TNFR1, RLA-DR, ICAM-1, CD11b, and CD11C) were also found in the corneal epithelium and stroma especially near the neovascularization area and their numbers were higher than those found in saline-instilled sutured eyes (Figure 7C, p<0.0001). The proliferative cell marker, KI-67, was found chiefly located near the nylon suture and neovascularization areas and the number of positive cells was higher in LPS plus suture (987±154 cells/mm2) than in the saline plus suture eyes (566±99 cells/mm2, p<0.0001). TUNEL-positive apoptotic cells were observed mainly near the nylon stitch in the saline plus suture eyes (Figure 8K) whereas they were observed more abundantly not only near the nylon suture but also at a distance in the corneal stroma in the LPS-instilled eyes (Figure 8L).


Our design of this work was aimed to mimic the clinical differences between an infected and non-infected corneal wound. In a previous study, an intrastromal injection of LPS in rabbits induced severe keratitis characterized by edema and PMN infiltration [14]. Pseudomonas aeruginosa endotoxin-induced keratitis in a mouse-ablated epithelium model was regulated by TLR4-dependent expression of PECAM-1 and MIP-2, which are essential for recruitment of neutrophils [8]. Compared with these previous studies, we observed in vivo inflammatory infiltration using IVCM, a very promising technique meeting the criteria of the guidelines for the design of animal experiments that recommend minimizing the number of animals and refining the tests used on animals (Statement for the Use of Animals in Ophthalmic and Visual Research). The HRT II/RCM used for IVCM offers a high definition providing histology-like images and could help us choose the maximum reaction time point and then optimize the time chosen for the histological studies requiring animal sacrifice. The high-resolution images given by HRT II/RCW provided non-invasive evaluations at a cellular level in these animal models. This technique can be performed repeatedly in vivo to follow the course of a disease or a healing process. We observed a strong correlation between standard immunohistology with non-invasive HRT and its major advantage for cellular analyses of healthy or pathological corneas by providing in vivo histologic-like images. HRT II/RCM clearly showed that in corneal tissue, LPS derived from E. coli was a powerful inflammatory cell inducer in the corneal stroma and also in basal epithelial layers. IVCM images showed that, compared to saline instillation, this special layer was more sensitive to LPS instillation as we found an abundant number of inflammatory cells especially dendritic-like cells, even 20 days after stopping LPS instillation. We thus observed that the significant increase in inflammatory infiltrates induced by LPS instillation remained for a longer duration in basal epithelial layers (until D30) than in the stroma (until D20). We observed, however, that the total inflammatory cell counts were always higher in the corneal stroma than in the basal corneal epithelium.

In this study, we observed the enhancement of inflammation, apoptosis, and neovascularization under the influence of LPS from E. coli instillation in three types of corneal injury. The three types of injury corresponding to clinical ophthalmological conditions, i.e. incision, scraping, and intrastromal suture, were all sensitive to LPS instillation, showing more inflammatory cell infiltrates than to saline instillation with IVCM and immunohistology. Among the three types of corneal injury, the suture model showed higher inflammatory infiltration than did the scraping and incision models. This could be due to the persistence of the non-absorbable suture, which could prolong the duration of LPS contact at the corneal surface and within the stroma. Even though the deep transversal incision model also injured the stroma, corneal reepithelialization at D1-D2 seemed to prevent the effect of LPS instillation.

TNF-α and TLRs appear to be the most important mediators associated with LPS-induced inflammation and apoptosis in various in vivo and in vitro systems. In the endotoxic shock mouse model, LPS induced disseminated endothelial apoptosis predominantly resulting from autocrine secretion of TNF [25,26]. In a previous study, we showed that in the conjunctival tissue, LPS could induce substantial ocular surface inflammation. Anti-TNF-α neutralizing antibodies could in large part inhibit LPS-induced inflammation and apoptosis thus, emphasizing the cascade of interactions between LPS and TNF-α [16]. Here, we showed that LPS instillation clearly worsened the corneal injuries in whatever model was used, possibly via TNF-related interactions. Upregulation of TLR4 on inflammatory cells, CD4+ lymphocytes, eosinophils, and mast cells was found in vernal keratoconjunctivitis (VKC), suggesting a role played by TLRs and therefore by LPS in allergic keratitis and conjunctivitis. In the ocular epithelia, TLRs might participate in the defense against environmental microbial agents [27]. Another in vitro study showed that human conjunctival epithelial cell lines (HCEC Chang and IOBA-NHC cell lines) lack LPS responsiveness due to this deficient expression of MD2 (an accessory molecule required for TLR4 signaling), and the response to LPS can be restored by interferon-γ (IFNγ) priming or MD2 supplementation but not by TNF-α [2]. TLR4, but not TLR2, was shown to act as an apoptosis-promoting signal in cultured microglia [28]. In mice and bone-marrow-derived macrophage, apoptosis induced by different bacterial pathogens was dependent on activation of TLR4 [29]. In our study, TLR4/MD2 was expressed at a higher level in corneal injuries with LPS instillation than in those instilled with saline. We assume that TLR4/MD2 complex expression responded to LPS stimulation in corneal injury models, most likely as the injury induces inflammatory cascades stimulating both MD2 and TLRs. This could be one explanation for the accentuation of inflammation, apoptosis, and neovascularization induced by LPS.

Intrastromal injection of LPS in rats has been shown to stimulate corneal neovascularization more than injection of saline [30,31]. The reason for a high frequency of allograft rejection in these rat models was the consistent and prolonged neovascularization of allografts that were placed into intralamellar pockets formed by intrastromal injection of 10-25 μl of 100 μg/ml LPS [31]. In rabbits, implants containing 100 ng of LPS also enhanced angiogenic response in cornea [32]. Harmey et al. reported that in murine lung tumor cells, LPS exposure could increase angiogenesis and vascular permeability. This LPS-related angiogenesis was principally related to two inducers, TNF and TLRs [33]. LPS can induce proliferation of endothelial cells and initiate angiogenesis directly through TNF receptor-associated factor 6 (TRAF6)-dependent signaling pathways [34]. Primary cultures of human limbal fibroblasts (PCHLFs) participated in the vascular endothelial growth factor (VEGF) production induced by LPS through the stimulation of TLR4 [35]. Murine macrophage VEGF expression is synergistically upregulated by LPS, acting through TLR4 receptors [36]. Consistent with these previous studies, in corneal injury models, we found that LPS instillation could accelerate and amplify the neovascularization process. This LPS-related angiogenesis was accompanied by more TNF and TLR4/MD2 immunopositive cells observed in corresponding cryosections.

LPS stimulates directly or indirectly both TNF- and TLR-associated cascades. TLRs may respond to bacteria by inducing the expression of cytokines such as TNF-α, IL-1, IL-6, and IL-12 [37,38]. This LPS/TLR/TNF system is also implicated in apoptosis and may play a major role in ocular surface diseases.

Altogether, our studies point out a direct enhancing role of LPS in ocular surface inflammation, apoptosis, and neovascularization. We assume a potentially important relationship between bacterial components and inflammatory reactions, leading to persistent neovascularization that may severely compromise visual function. Our findings emphasize the importance of an aggressive treatment of corneal bacterial infections to avoid permanent damage and irreversible visual function. Infected corneal wounds should be treated rapidly and aggressively to get rid of the proinflammatory and proapoptotic effects of LPS and minimize future irreversible corneal damage. Further studies will investigate the possible efficacy of blocking strategies for LPS, TNF, or TLR4 in such severe corneal inflammatory models.


Granted by Quinze-Vingts National Ophthalmology Hospital and INSERM, UMR S 872


1. Saint Andre A, Blackwell NM, Hall LR, Hoerauf A, Brattig NW, Volkmann L, Taylor MJ, Ford L, Hise AG, Lass JH, Diaconu E, Pearlman E. The role of endosymbiotic Wolbachia bacteria in the pathogenesis of river blindness. Science 2002; 295:1892-5.

2. Talreja J, Dileepan K, Puri S, Kabir MH, Segal DM, Stechschulte DJ, Dileepan KN. Human conjunctival epithelial cells lack lipopolysaccharide responsiveness due to deficient expression of MD2 but respond after interferon-gamma priming or soluble MD2 supplementation. Inflammation 2005; 29:170-81.

3. Hazlett LD. Corneal response to Pseudomonas aeruginosa infection. Prog Retin Eye Res 2004; 23:1-30.

4. Cowell BA, Willcox MD, Hobden JA, Schneider RP, Tout S, Hazlett LD. An ocular strain of Pseudomonas aeruginosa is inflammatory but not virulent in the scarified mouse model. Exp Eye Res 1998; 67:347-56.

5. Yeh DL, Stinnett SS, Afshari NA. Analysis of bacterial cultures in infectious keratitis, 1997 to 2004. Am J Ophthalmol 2006; 142:1066-8.

6. Asano-Kato N, Toda I, Shimmura S, Noda-Tsuruya T, Fukagawa K, Yoshinaga M, Matsukawa A, Tsubota K. Detection of neutrophils and possible involvement of interleukin-8 in diffuse lamellar keratitis after laser in situ keratomileusis. J Cataract Refract Surg 2003; 29:1996-2000.

7. Morck DW, Holland SP, Ceri H, Hancock R, Scott-Gough M, Nugyen V, Keith EJ, Lee TL. Use of polymyxin as an endotoxin blocker in the prevention of diffuse lamellar keratitis in an animal model. J Refract Surg 2005 Mar-Apr; 21:152-7.

8. Khatri S, Lass JH, Heinzel FP, Petroll WM, Gomez J, Diaconu E, Kalsow CM, Pearlman E. Regulation of endotoxin-induced keratitis by PECAM-1, MIP-2, and toll-like receptor 4. Invest Ophthalmol Vis Sci 2002; 43:2278-84.

9. Shalaby MR, Waage A, Aarden L, Espevik T. Endotoxin, tumor necrosis factor-alpha and interleukin 1 induce interleukin 6 production in vivo. Clin Immunol Immunopathol 1989; 53:488-98.

10. Ulevitch RJ, Tobias PS. Recognition of gram-negative bacteria and endotoxin by the innate immune system. Curr Opin Immunol 1999; 11:19-22.

11. Kawaguchi H, Hizuta A, Tanaka N, Orita K. Role of endotoxin in wound healing impairment. Res Commun Mol Pathol Pharmacol 1995; 89:317-27.

12. Song PI, Abraham TA, Park Y, Zivony AS, Harten B, Edelhauser HF, Ward SL, Armstrong CA, Ansel JC. The expression of functional LPS receptor proteins CD14 and toll-like receptor 4 in human corneal cells. Invest Ophthalmol Vis Sci 2001; 42:2867-77.

13. Chang JH, McCluskey P, Wakefield D. Expression of toll-like receptor 4 and its associated lipopolysaccharide receptor complex by resident antigen-presenting cells in the human uvea. Invest Ophthalmol Vis Sci 2004; 45:1871-8.

14. Trinkaus-Randall V, Leibowitz HM, Ryan WJ, Kupferman A. Quantification of stromal destruction in the inflamed cornea. Invest Ophthalmol Vis Sci 1991; 32:603-9.

15. Carlson EC, Drazba J, Yang X, Perez VL. Visualization and characterization of inflammatory cell recruitment and migration through the corneal stroma in endotoxin-induced keratitis. Invest Ophthalmol Vis Sci 2006; 47:241-8.

16. Liang H, Baudouin C, Labbe A, Pauly A, Martin C, Warnet JM, Brignole-Baudouin F. In vivo confocal microscopy and ex vivo flow cytometry: new tools for assessing ocular inflammation applied to rabbit lipopolysaccharide-induced conjunctivitis. Mol Vis 2006; 12:1392-402 <>.

17. Kumagai N, Fukuda K, Fujitsu Y, Lu Y, Chikamoto N, Nishida T. Lipopolysaccharide-induced expression of intercellular adhesion molecule-1 and chemokines in cultured human corneal fibroblasts. Invest Ophthalmol Vis Sci 2005; 46:114-20.

18. Kulshin VA, Zahringer U, Lindner B, Jager KE, Dmitriev BA, Rietschel ET. Structural characterization of the lipid A component of Pseudomonas aeruginosa wild-type and rough mutant lipopolysaccharides. Eur J Biochem 1991; 198:697-704.

19. Kropinski AM, Jewell B, Kuzio J, Milazzo F, Berry D. Structure and functions of Pseudomonas aeruginosa lipopolysaccharide. Antibiot Chemother 1985; 36:58-73.

20. Pauly A, Brignole-Baudouin F, Labbe A, Liang H, Warnet JM, Baudouin C. New tools for the evaluation of toxic ocular surface changes in rat. Invest Ophthalmol Vis Sci. In press 2007.

21. Suzuki T, Sano Y, Kinoshita S. Effects of 1alpha,25-dihydroxyvitamin D3 on Langerhans cell migration and corneal neovascularization in mice. Invest Ophthalmol Vis Sci 2000; 41:154-8.

22. Bochert R, Zhivov A, Kraak R, Stave J, Guthoff RF. Contribution to comprehension of image formation in confocal microscopy of cornea with Rostock cornea module. Br J Ophthalmol 2005; 89:1351-5.

23. Messmer EM, Zapp DM, Mackert MJ, Thiel M, Kampik A. In vivo confocal microscopy of filtering blebs after trabeculectomy. Arch Ophthalmol 2006; 124:1095-103.

24. Labbe A, Liang H, Martin C, Brignole-Baudouin F, Warnet JM, Baudouin C. Comparative anatomy of laboratory animal corneas with a new-generation high-resolution in vivo confocal microscope. Curr Eye Res 2006; 31:501-9.

25. Haimovitz-Friedman A, Cordon-Cardo C, Bayoumy S, Garzotto M, McLoughlin M, Gallily R, Edwards CK 3rd, Schuchman EH, Fuks Z, Kolesnick R. Lipopolysaccharide induces disseminated endothelial apoptosis requiring ceramide generation. J Exp Med 1997; 186:1831-41.

26. Xaus J, Comalada M, Valledor AF, Lloberas J, Lopez-Soriano F, Argiles JM, Bogdan C, Celada A. LPS induces apoptosis in macrophages mostly through the autocrine production of TNF-alpha. Blood 2000; 95:3823-31.

27. Bonini S, Micera A, Iovieno A, Lambiase A, Bonini S. Expression of Toll-like receptors in healthy and allergic conjunctiva. Ophthalmology 2005; 112:1528; discussion1548-9.

28. Jung DY, Lee H, Jung BY, Ock J, Lee MS, Lee WH, Suk K. TLR4, but not TLR2, signals autoregulatory apoptosis of cultured microglia: a critical role of IFN-beta as a decision maker. J Immunol 2005; 174:6467-76.

29. Hsu LC, Park JM, Zhang K, Luo JL, Maeda S, Kaufman RJ, Eckmann L, Guiney DG, Karin M. The protein kinase PKR is required for macrophage apoptosis after activation of Toll-like receptor 4. Nature 2004; 428:341-5.

30. Gebhardt BM. Factors affecting corneal allograft rejection in inbred rats. Transplant Proc 1981; 13:1091-3.

31. Peterson RJ, Kwedar SA, Moticka EJ. An improved method of intralamellar keratoplasty in rats. Invest Ophthalmol Vis Sci 1987; 28:281-6.

32. BenEzra D, Griffin BW, Maftzir G, Aharonov O. Thrombospondin and in vivo angiogenesis induced by basic fibroblast growth factor or lipopolysaccharide. Invest Ophthalmol Vis Sci 1993; 34:3601-8.

33. Harmey JH, Bucana CD, Lu W, Byrne AM, McDonnell S, Lynch C, Bouchier-Hayes D, Dong Z. Lipopolysaccharide-induced metastatic growth is associated with increased angiogenesis, vascular permeability and tumor cell invasion. Int J Cancer 2002; 101:415-22.

34. Pollet I, Opina CJ, Zimmerman C, Leong KG, Wong F, Karsan A. Bacterial lipopolysaccharide directly induces angiogenesis through TRAF6-mediated activation of NF-kappaB and c-Jun N-terminal kinase. Blood 2003; 102:1740-2.

35. Rodriguez-Martinez S, Cancino-Diaz ME, Miguel PS, Cancino-Diaz JC. Lipopolysaccharide from Escherichia coli induces the expression of vascular endothelial growth factor via toll-like receptor 4 in human limbal fibroblasts. Exp Eye Res 2006; 83:1373-7.

36. Leibovich SJ, Chen JF, Pinhal-Enfield G, Belem PC, Elson G, Rosania A, Ramanathan M, Montesinos C, Jacobson M, Schwarzschild MA, Fink JS, Cronstein B. Synergistic up-regulation of vascular endothelial growth factor expression in murine macrophages by adenosine A(2A) receptor agonists and endotoxin. Am J Pathol 2002; 160:2231-44.

37. Means TK, Golenbock DT, Fenton MJ. Structure and function of Toll-like receptor proteins. Life Sci 2000; 68:241-58.

38. Underhill DM, Ozinsky A. Toll-like receptors: key mediators of microbe detection. Curr Opin Immunol 2002; 14:103-10.

Liang, Mol Vis 2007; 13:1169-1180 <>
©2007 Molecular Vision <>
ISSN 1090-0535