The Diabetic Ocular Environment Facilitates the Development of Endogenous Bacterial Endophthalmitis

Abstract

Purpose.

We tested the hypothesis that changes in the diabetic ocular environment facilitate the development of endogenous bacterial endophthalmitis (EBE).

Methods.

C57BL/6J mice were rendered diabetic with streptozotocin (STZ) for 1, 3, or 5 months' duration. Diabetic and age-matched nondiabetic mice were tail vein–injected with 10⁸ CFU of Klebsiella pneumoniae, a common cause of EBE in diabetics. After either 2 or 4 days postinfection, the EBE incidence was assessed by electroretinography, histology, bacterial counts, and myeloperoxidase ELISAs. Blood-retinal barrier (BRB) permeability in uninfected diabetic mice also was determined.

Results.

No cases of EBE were observed among the 1-month diabetic group. Extending the time from diabetes induction to 3 months resulted in a 23.8% EBE incidence after 2 days, and a 22% incidence after 4 days. The incidence of EBE increased to 27% in the 5-month diabetic group. Infected eyes had an average 8.01 × 10² and 6.19 × 10⁴ CFU/eye for the 3- and 5-month diabetic groups, respectively. There was no significant difference in BRB permeability between control and 1-month uninfected diabetic mice. However, 3- and 5-month diabetic mice had significantly greater BRB permeability than control mice. These results suggested that increasing the time from STZ diabetes induction to 3 and 5 months resulted in an ocular environment more conducive to the development of EBE.

Conclusions.

These results demonstrated a correlation between an increase in BRB permeability and an increase in EBE incidence, supporting the hypothesis that diabetic ocular changes contribute to the development of EBE.

Endogenous bacterial endophthalmitis (EBE) almost always results in significant vision loss and the mechanisms underlying this disease are not known. This study identifies the diabetic ocular environment and retinal vascular permeability specifically as a correlate to the development of EBE.

Introduction

Endogenous bacterial endophthalmitis (EBE) is a potentially blinding form of intraocular infection resulting from hematogenous spread of bacteria from a focus of infection into the eye. A number of conditions predispose individuals to EBE, including intravenous drug abuse and treatment with immunosuppressive agents. However, the most common predisposing risk factor is diabetes.^1–6 During the course of diabetes, permeability changes occur in the blood-retinal barrier (BRB). Of particular interest is the role that the diabetic ocular environment may have in the migration of bacteria from the vasculature into the eye, resulting in EBE. At present, the mechanisms underlying movement of bacteria across the BRB and into the eye from the bloodstream are not well understood.

One of the most common causes of EBE in diabetics is the opportunistic pathogen Klebsiella pneumoniae, a Gram-negative bacterium found in the environment and a constituent of the gastrointestinal tract microbiota.^1–6 K. pneumoniae is a frequent cause of community-acquired and healthcare-associated infections, which include pneumonia, bacteremia, urinary tract, and wound infections.^7–10 K. pneumoniae is the most prevalent bacteria isolated from pyogenic liver abscesses in the United States, Trinidad, and Asian Pacific regions,^11–16 with diabetes as the major underlying condition associated with these abscesses.^17–23 Ten percent of pyogenic K. pneumoniae liver abscess cases are complicated by EBE.⁸ The visual outcome in patients with K. pneumoniae EBE is uniformly poor, ranging from hand motion visualization to evisceration or enucleation of the eye.^1–3 The prevalence of cases of K. pneumoniae EBE where diabetes is a predisposing risk factor led to the hypothesis that changes that occur within the ocular environment during diabetes contribute to the development of K. pneumoniae EBE. To test this hypothesis, we developed a streptozotocin (STZ)-induced diabetic mouse model of K. pneumoniae EBE. In this report, we demonstrated that the incidence of K. pneumoniae EBE was greater in STZ-induced diabetic mice than in control mice, and that the degree of BRB permeability may have facilitated a higher incidence of EBE. The results demonstrated a correlation between the duration of diabetes and the incidence of K. pneumoniae EBE, and also suggested that increases in BRB permeability may contribute to the enhanced incidence of K. pneumoniae EBE.

Methods

Animals

Six-week-old C57BL/6J mice were acquired from the Jackson Laboratory and used in accordance with institutional guidelines and the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. Mice were allowed to acclimate to the environment for two weeks before the establishment of diabetes. Mice were anesthetized with an intramuscular injection of 85 mg ketamine/kg and 14 mg xylazine/kg before tail-vein injections and electroretinography.

STZ Induction of Diabetes

To establish diabetes, 8-week-old male C57BL/6J mice were given a series of 2 STZ injections at 100 mg/kg one week apart.²⁴ STZ (Sigma-Aldrich, St. Louis, MO) was solubilized in freshly prepared 10 mM citrate buffer pH 4.5 and immediately injected intraperitoneally. Blood glucose levels were measured two weeks following the second STZ injection. All mice demonstrated blood glucose levels greater than or equal to 300 mg/dl. Mice were allowed to proceed for either 1, 3, or 5 months, and were the same age at the time of infection. Control animals were administered citrate buffer pH 4.5 only. Before injections with K. pneumoniae or Evans Blue dye, blood glucose levels were measured again to ensure a diabetic state.

Endogenous Bacterial Endophthalmitis Model

A clinical hypermucoviscosity (HMV) negative K. pneumoniae endophthalmitis isolate²⁵ was grown for 18 hours in brain heart infusion media (BHI; Difco Laboratories, Detroit, MI), and subcultured in prewarmed BHI to logarithmic phase. Bacteria then were centrifuged and resuspended in PBS. To establish endogenous endophthalmitis infections, all groups of mice were tail vein injected with 10⁸ CFU in 100 μL PBS. At 2 days postinfection, some eyes were harvested for bacterial quantitation as described below. At 4 days postinfection, retinal function of both eyes was assessed by electroretinography. Both eyes from each mouse were harvested for bacterial and polymorphonuclear leukocyte (PMN) quantitation or histology.

Electroretinography

To determine the degree of retinal function 4 days after mice were infected with K. pneumoniae, mice were dark adapted for at least 6 hours and then anesthetized as described above. Pupils were dilated with 10% topical phenylephrine (Akorn, Inc., Lake Forest, IL) and gold-wire electrodes were placed over both corneas with a drop of GONAK (Hypromeliose Ophthalmic Demulscen Solution 2.5%; Akorn, Inc.) to ensure maximum conductance. Scotopic ERGs were recorded for both eyes using a UTAS-E3000 Electrodiagnostic System (LKC Technologies, Inc., Gaithersburg, MD). The A-wave amplitude was measured from the prestimulus baseline to the A-wave trough, and the B-wave amplitude was measured from the A-wave trough to the B-wave peak. Percent ERG retention was calculated relative to mean baseline ERGs recorded from 6 eyes per group before infection.^26,27

Bacterial Quantitation

For each group of infected mice, eyes were selected at random for bacterial quantitation. Briefly, the left and right eyes from each mouse selected were enucleated, placed into separate tubes containing 400 μl of sterile PBS and 1.0 mm sterile glass beads (Biospec Products, Inc., Bartlesville, OK), and homogenized for 60 seconds at 5,000 revolutions per minute (rpm) in a Mini-BeadBeater (Biospect Products, Inc.). Eye homogenates were diluted serially and plated in triplicate on brain heart infusion (BHI) agar and the CFU per eye determined.^26,27

PMN Quantitation

PMN leukocyte infiltration into eyes was assessed indirectly by measurement of myeloperoxidase (MPO) concentrations in whole eye homogenates. Eyes were enucleated and homogenized with glass beads in PBS as described for bacterial quantitation. Homogenates were diluted 1:4 and analyzed for MPO by ELISA according to manufacturer's instructions (Mouse MPO ELISA Test Kit; Cell Sciences, Canton, MA).^27,28

Histology

Because homogenization of eyes for bacterial quantitation precludes subsequent histologic analysis on eyes that are confirmed culture positive, eyes from each group were selected randomly and fixed in Excalibur's Alcoholic Z-Fix for 24 hours, and then embedded in paraffin. Sections were stained with hematoxylin and eosin (H&E), and tissue Gram stain. Livers were fixed in 10% neutral buffered formalin for 24 hours, embedded in paraffin, and sections stained with hematoxylin and eosin.^26,27

Vascular Permeability

Albumin leakage from blood vessels into the retina was measured to quantify vascular permeability using a modified Evans blue protocol.²⁹ Mice were anesthetized as described above and 15 mg Evans Blue dye (Sigma-Aldrich) per kg were injected into the tail vein. After two hours, mice were perfused with 1% paraformaldehyde in citrate buffer (pH 4.2) for 1 minute, and retinas harvested and placed in 150 μL formamide. Retinas were incubated at 70°C for 18 hours to extract the dye and centrifuged at 70,000 rpm (TLA 100.3 rotor; Beckman Coulter, Fullerton, CA) for 20 minutes. The optical density at 620 nanometers (OD₆₂₀) of 100 μL of the supernatants was measured and the concentration of Evans blue was calculated from a standard curve of Evans Blue in formamide. The pellets then were solubilized in 1 mL of 0.2% SDS in PBS and the protein concentrations measured using a BCA protein assay. The concentration of Evans Blue in each sample then was normalized to the total protein per sample. Results were expressed in micrograms of Evans blue/mg total protein content.

Statistics

All values represent the mean ± SD of the number of eyes per group indicated in the figure legends for each assay. Two-tailed, two-sample t-tests were used for statistical comparisons between groups. A P value of ≤ 0.05 was considered significant.

Results

Diabetes and Incidence of K. pneumoniae EBE

To test the hypothesis that changes in the diabetic ocular environment facilitated the development of K. pneumoniae EBE, we used a mouse model of STZ -induced diabetes.²⁴ Male C57BL/6J mice were injected with 100 mg/kg STZ and diabetes was allowed to progress for one month. One-month diabetic and age-matched sham injected control mice were tail vein-injected with 10⁸ CFU of a clinical ocular isolate of K. pneumoniae. After 2 days postinfection, the incidence of EBE was evaluated. No difference was observed in the incidence of EBE between the one-month diabetic and control mice (Table 1). We surmised that the changes in the architecture of the eye observed during diabetes progression had not yet occurred after one month, even though the mean blood glucose level of these mice was >400 mg/dl. We sought to determine whether extending the time from diabetes induction to 3 months might facilitate the changes necessary for EBE development. After 2 days postinfection, we observed a 23.8% incidence of infection among 3-month diabetic mice, while no cases of EBE were observed in control mice (Table 1).

Table 1.  .

Incidence of K. pneumoniae EBE at 2 Days Postinfection in STZ-Induced Diabetic or Age Matched Controls (1 Month or 3 Months in Duration) Infected with HMV− K. pneumoniae

Duration of Diabetes	1 Month	1 Month	3 Months	3 Months
Group	Diabetic	Control	Diabetic	Control
Total mice	15	10	25	25
N surviving after 2 days postinfection	15	10	21	25
Mice with EBE	0	0	5	0
% Infected of survivors	0	0	23.8	0

To determine whether extending the time from diabetes induction to 5 months might result in greater incidence of EBE, we repeated the above-described experiments and included a 5-month diabetic group. Mice in all groups were age-matched at the time of infection and had similar blood glucose levels of >300 mg/dl. As shown in Table 2, the incidence of K. pneumoniae EBE among 1- and 3-month groups was similar to that observed previously, even though infection was allowed to proceed for 4 days. We observed no cases of EBE among 1-month diabetic mice, but a 22% incidence among 3-month diabetic mice. Of interest was the slight increase in incidence observed in the 5-month STZ group to 27%, suggesting that increasing the time from STZ diabetes induction to 5 months resulted in an ocular environment more conducive to the development of EBE. The mean CFU per eye was 8.01 × 10² (s = 1.08 × 10³) for the 3-month and 6.19 × 10⁴ (s = 1.30 × 10⁵) for the 5-month diabetic groups.

Table 2.  .

Incidence of K. pneumoniae EBE at 4 Days Postinfection in STZ-Induced Diabetic Mice (1 Month, 3 Months, and 5 Months in Duration) Infected with an HMV− K. pneumoniae

Duration of Diabetes	1 Month	3 Months	5 Months
N of mice infected with K. pneumoniae	27	35	32
N surviving after 4 days postinfection	24	32	30
Mice with EBE	0	7	8
% Infected of survivors	0	22	27
Mean CFU/eye	0	8.01 × 102 (± 1.08 × 103)	6.19 × 104 (± 1.30 × 105)

Retinal Function and Inflammation

To determine whether retinal function had been affected during infection, mice were subjected to electroretinography 4 days after infection with K. pneumoniae. No changes from preinfection baseline retinal function were observed among the 1-month diabetic mice. We also did not observe significant decreases in the percent A- or B-wave amplitudes retained among the animals in the 3- or 5-month diabetic groups that were culture confirmed with EBE (data not shown). These results are not surprising given the low number of CFU of this HMV− strain detected in the eyes of infected mice after 4 days. Previous studies of this HMV− strain in a direct injection model of endophthalmitis demonstrated that to observe significant retinal function decline, orders of magnitude higher concentrations of bacteria in the eye are necessary.^25,28 Moreover, we detected essentially no infiltration of PMNs as measured indirectly by MPO concentrations in culture-confirmed, infected 3- or 5-month diabetic eyes. No significant differences were observed in the concentration of MPO in the eyes of 3- or 5-month diabetic animals with EBE versus the eyes of 1-month diabetic eyes without EBE (data not shown). Concentrations of MPO ranged from 2.5 to 3.0 ng per eye for the 1-month diabetic eyes without EBE, and the 3- and 5-month diabetic mice with EBE. We previously have shown that between 9 and 21 hours after direct infection of the vitreous with K. pneumoniae, MPO concentrations ranged from 10 to 60 ng per eye²⁵ in comparison.

Correlation between EBE and Underlying Disease State

To determine if there was a correlation between the presence of liver abscesses and EBE among infected animals, histologic analysis of liver sections from representative animals from each group was performed. Interestingly, no abscesses were observed among the 1-month diabetic animals. However, liver abscesses were observed in 5 of 7 mice in the 3-month diabetic group, and in 2 of 5 in the 5-month diabetic group that were analyzed (Fig. 1). Of the 5 animals in the 3-month group that had liver abscesses (Figs. 1B, 1C), 1 animal had culture-confirmed EBE, and both animals with liver abscesses (Figs. 1D, 1E) in the 5-month group had EBE. The other 3 animals in that group were confirmed to have EBE; however, no liver abscesses were observed. These results are not suggestive of a direct correlation between the presence of liver abscesses and EBE occurrence, but demonstrated that in this diabetic model, liver abscesses do occur, consistent with reports of other bacterial species.^30,31

Figure 1.  .

Histologic sections of livers from diabetic mice with culture-confirmed EBE. Representative 5 μm sections from a 1-month diabetic mouse without EBE (A), the one 3-month diabetic mouse (B, C), and the two 5-month diabetic mice with culture-confirmed EBE (D, E) are shown. Arrows indicated location of abscesses. H&E, magnification 10× (A, B, D, E) and 20× (C).

Correlation between EBE and Retinal Vascular Permeability

Because a greater incidence of K. pneumoniae EBE was observed in the 3- and 5-month diabetic groups than in the 1-month diabetic group, we hypothesized that the diabetic retinal environment at 3 and 5 months facilitated a higher incidence of EBE. To determine the extent of BRB permeability in these groups of diabetic mice, an Evans Blue dye assay was used to measure albumin leakage into the retina.²⁹ There was no difference in BRB permeability between 1-month control and 1-month diabetic mice (Fig. 2, P = 0.23). However, 3-month diabetic mice had significantly greater BRB permeability than nondiabetic control mice (Fig. 2, P = 0.02), suggesting that changes in vascular permeability occurred in the eyes of 3-month diabetic mice. An even greater degree of vascular permeability was observed among the 5-month diabetic mice relative to the control animals (Fig. 2, P = 0.000004). Taken together, these results showed a correlation between an increase in BRB permeability and an increase in EBE incidence, supporting our initial hypothesis that changes in the diabetic eye contribute to acquisition of EBE.

Figure 2.  .

Duration from diabetes induction correlates with vascular permeability. Albumin leakage from blood vessels into the retina was measured to quantify vascular permeability using a modified Evans blue protocol. The concentration of Evans blue was calculated from a standard curve and normalized to the total protein per sample. Results were expressed in micrograms of Evans blue per milligrams of total protein content. Bars: mean ± SD for N ≥ 5 animals for all groups. A two-tailed t-test was used to assess significance between group (control versus 1 month P = 0.23, *control versus 3 month P = 0.02, **control versus 5 month P = 0.000004).

Discussion

EBE results from bacteria seeding the bloodstream from a distant focus of infection and migrating across the BRB into the posterior segment of the eye. The often poor visual outcome and potential for bilateral blindness make EBE one of the most destructive and devastating infections of the eye. Approximately 41% of cases of EBE result in a final visual acuity of count fingers or better, 26% of infected eyes lose all useful vision, and 29% required surgical removal.^1–3 K. pneumoniae causes the majority of Gram-negative cases of EBE.^1–6 K. pneumoniae has become the leading cause of pyogenic liver abscess in patients in Taiwan^10,32 and there are increasing numbers of reports of K. pneumoniae EBE resulting from septicemia following pyogenic liver abscesses in diabetics.^7–10,32,33 K. pneumoniae liver abscesses also are being reported with increased frequency in North America.³⁴

Diabetes is the primary risk factor for contracting bacterial EBE.^1–6 Diabetes is a metabolic disease due to dysregulation of glucose metabolism that compromises the immune system and predisposes the individual to an increased risk of infection. Diabetes also compromises the eye architecturally, resulting in changes that may enhance the ability of organisms to invade the eye during systemic infection. Among patients with diabetes, specifically type II diabetes, diabetic retinopathy is a serious and common complication that impacts the retinal vasculature, glia, and retinal neurons.^35,36 During the early stages, death of pericytes and thickening of the basement membrane occur, leading to capillary leakage and occlusion. These changes ultimately result in macular edema, formation of new vessels, and neuroretinal degeneration.^37–42 Increases in VEGF, and the cytokines IL-1β, IL-6, IL-8, and TNFα in diabetic retinas lead to angiogenesis and a breakdown in the BRB, implying an association with inflammation.⁴³ Intercellular adhesion molecule 1 (ICAM-1) and CD18 are upregulated, and contribute to increased leukocyte adhesion to the endothelial wall, leading to endothelial cell injury and death.^44–47 Elevated expression of matrix metalloproteases in the retina also may facilitate increases in BRB permeability by degradation of tight junction complexes.⁴⁸ RPE and vascular endothelium, barrier cells of the BRB, also undergo changes in hyperglycemic environments in vitro and in vivo, including increases in VEGF that may contribute to vascular permeability.^49,50 It currently is unknown whether these changes are linked to the development of EBE, and the pathogenic mechanisms underlying the migration of bacteria toward and into the eye, inflammation, and vision loss have not been addressed to our knowledge.

In our work, we observed a 22% to 27% incidence of EBE among 3- and 5-month diabetic mice, but no incidence of EBE among control, nondiabetic, and 1-month diabetic mice. Additionally, we demonstrated that increases in vascular permeability correlated with time from diabetes induction. While these infections were low-grade and did not entirely mimic human disease in terms of PMN infiltration and functional loss,^25,28 this model provides a solid framework to study and gain a better understanding of the role that diabetes has in development of EBE, and of the interplay of host and bacterial factors that contribute to this devastating disease. In our model, we used an HMV− strain of K. pneumoniae to establish the model and provide baseline data for future comparisons with HMV+ K. pneumoniae. Our laboratory has demonstrated in a mouse model of experimentally-induced endophthalmitis, in which bacteria are introduced into the vitreous by direct injection, that the HMV phenotype contributes to pathogenesis. As reported by Wiskur et al., an HMV+ strain caused significantly greater inflammation and greater loss of visual function than an HMV− strain,²⁵ the same strain we used in our study. Further, the HMV+ strain was not readily cleared from the eye in this model.²⁵ More recently, our group used the experimentally-induced endophthalmitis model to compare directly a wild type HMV+ K. pneumoniae strain with an isogenic mutant defective in magA, the gene primarily responsible for the HMV phenotype.²⁸ The studies showed that mice infected with the magA− strain showed a significantly improved visual outcome relative to mice infected with the wildtype strain.²⁸ Both of these studies clearly established a role for the HMV phenotype in K. pneumoniae endophthalmitis. As this direct injection model circumvents systemic infection, it remains to be seen whether the HMV phenotype is important for crossing the BRB and establishing EBE. The HMV phenotype might enable bacteria to more readily cross the BRB, adhere to structures within the eye, and/or evade PMN clearance. Future studies in our laboratory will address this hypothesis.

To our knowledge, no studies have examined whether patients suffering from bacterial EBE had a history of diabetes-induced vision changes or diabetic retinopathy before infection. However, if barrier function of the BRB were altered and the immune response was unable to suppress transient bacteremia, bacteria could circumvent the immune response, reach the retinal vasculature, and gain access into the eye through a permeable BRB. While we did not observe EBE in nondiabetic mice, it might be possible that K. pneumoniae contributed directly to increasing vascular permeability. It has been shown that Neisseria meningitidis, especially unencapsulated strains, induce the upregulation of the vascular adhesion molecules CD62E, ICAM-1, and VCAM-1, leading to increased binding of leukocytes to the endothelium, and ultimately vascular damage and permeability.⁵¹ While it has not been shown that K. pneumoniae similarly induces upregulation of vascular adhesion molecules, we cannot rule out the possibility that K. pneumoniae contributes to the pathogenesis of EBE by this mechanism. We further acknowledge that systemic infection with K. pneumoniae might elicit increases in vascular permeability through the liberation of TNFα, as do other encapsulated organisms.^52,53 However, as stated, we have not observed any incidence of EBE among control animals, so it is unlikely that systemic infection alone caused the permeability necessary for EBE development.

In addition to changes in BRB permeability during the course of diabetes, suppression of the innate immune system also contributes to an increased risk of infection, and potentially to EBE pathogenesis. PMNs are necessary for clearance of pathogens from the eye during acute bacterial endophthalmitis.^25–29 However, PMNs from diabetic mice and humans display defective abilities to phagocytize and kill bacteria, including Staphylococcus aureus and K. pneumoniae.^54,55 The possibility arises that the inability of PMNs to clear pathogens adequately from the bloodstream and the eye also may contribute to the development of EBE. However, this hypothesis has not been tested, and future studies will be necessary to address the role of the innate immune system in EBE pathogenesis.

To our knowledge, this is the first reported experimental model of endogenous bacterial endophthalmitis. Using this model, we have demonstrated a correlation between the duration of diabetes and the incidence of K. pneumoniae EBE. We also have shown that increases in retinal vascular permeability correlate with the time from diabetes induction. This model will prove useful in elucidating the mechanisms of EBE development and will allow the role of bacterial virulence traits to be dissected, as well as explore host factors and components of the immune system that are involved in EBE pathogenesis. Future studies will continue to explore the relationship between changes in the diabetic ocular environment and the development of EBE.