Endotoxin-induced uveitis is primarily dependent on radiation-resistant cells and on MyD88 but not TRIF

The anterior segment of the eye is especially sensitive to endotoxin and MyD88 pathway activation, but remarkably, the cells that are most sensitive are not bone marrow-derived.

Abstract

TLR4 activation by LPS (endotoxin) is mediated by the MyD88 and TRIF intracellular signaling pathways. We determined the relative activation of these pathways in murine ocular tissue after LPS exposure. Additionally, we explored whether BM-derived or non-BM-derived cells were the major contributors to EIU. Mice deficient in TRIF or MyD88 and their congenic (WT) controls received 250 ng ultrapure LPS ivt at 0 h. Ocular inflammation was assessed by histological analysis at 4, 6, and 24 h, and additionally, in MyD88^−/− mice, intravital microscopy was performed at 4 h and 6 h to assess adherent, rolling, and infiltrating cells in the iris vasculature and tissue. Cytokines associated with the MyD88 and TRIF intracellular signaling pathways were analyzed in ocular tissue at 4 h. BM chimeric mice (WT→WT, TLR4^−/−→WT, WT→TLR4^−/−) received 250 ng LPS by ivt injection, and ocular tissues were examined by histology at 6 h. Lack of MyD88 resulted in a markedly diminished cellular response and reduced production of MyD88-related cytokines 4 h post-LPS treatment. In contrast, lack of TRIF led to reduced production of TRIF-related cytokines and no change in the cellular response to LPS. Therefore, the MyD88 pathway appears to be the dominant TLR4 pathway in EIU. Only WT → TLR4^−/− chimeric mice were resistant to EIU, and this suggests, surprisingly, that non-BM-derived (radiation-resistant) cells in the eye play a greater role than BM-derived cells.

Introduction

The immune system is usually conceptualized as having two arms: an innate component and an adaptive component. The innate immune system contributes immensely to the pathogenesis of immune-mediated diseases, even among diseases such as systemic lupus erythematosus or rheumatoid arthritis, which are characterized by autoantibodies that presumably result from an adaptive immune response. The innate immune system includes PRRs, which respond rapidly to potential pathogens. These receptors include TLRs including TLR4, which recognize bacterial endotoxin or LPS. Another family of receptors in the innate immune system is called NLRs. Within the eye, TLRs and NLRs have been implicated in the inflammatory response [1, 2]. The importance of NLRs in the uveal tract is emphasized by the observation that the change in a single base in the NLR, NOD2, results in an autosomal-dominant form of uveitis known as Blau syndrome [3].

The evidence that implicates the innate immune system in autoimmune disease includes the following. 1) Polymorphisms or copy number variation in TLRs affect susceptibility to several immune-mediated diseases, including systemic lupus erythematosus [4]. 2) TLR inhibitors have shown utility in treating autoimmune disease or immune-mediated disease models [5]. 3) Bacterial flora are strongly implicated in diseases such as rheumatoid arthritis [6] and inflammatory bowel disease [7]. 4) The severity of several animal models of immune-mediated disease can be reduced dramatically by eliminating bowel flora [8]. 5) TLRs for nucleic acids are strongly implicated in the pathogenesis of systemic lupus [9]. 6) Mutations in NLRs are the cause of widespread inflammation, such as Blau syndrome or neonatal onset multisystem disease [3, 10]. 7) The majority of animal models of autoimmune disease requires an adjuvant, which generally works by activating the innate immune system.

TLR4 was the first receptor in the TLR family to be characterized [11]. It is unique among TLRs, as it signals through two well-characterized adaptor molecules—MyD88 and TRIF [12, 13]. The different adaptor molecules lead to distinct cytokines and transcription factors [12, 14, 15]. MyD88, for example, induces the synthesis of the transcription factor NF-κB and the cytokine IL-6. The TRIF pathway induces the transcription factor, IFN response factor 3, as well as NF-κB, and such cytokines as type I IFNs, MCP-1, and RANTES. Not all cells express TLR4. In the eye, it is most associated with the myeloid lineage of BM-derived cells and tissue-resident epithelial and endothelial cells [16–18]. Pathway activation by LPS varies according to the responding cell type [19].

Uveitis or intraocular inflammation is a leading cause of blindness [20]. Uveitis may occur as an immune-mediated disease, which affects only the eyes, but it also occurs frequently in association with multiorgan immune-mediated diseases, such as sarcoidosis, ankylosing spondylitis, juvenile idiopathic arthritis, Behçet's disease, and multiple sclerosis. Intriguingly, rats develop an acute anterior uveitis within 24 h of the injection of bacterial endotoxin into a distant site such as the footpad. Mice also develop EIU, but an ivt injection of LPS, rather than a systemic injection, is needed to induce a substantial cellular infiltrate within the eye [21]. Although EIU was characterized 30 years ago [22], fundamental questions about this model remain. In this report, we address two of these questions: which pathway, MyD88 or TRIF, contributes primarily to EIU, and which cells within the anterior segment of the eye account for this disease model?

MATERIALS AND METHODS

Animals

Age-matched (8–10 weeks) female BALB/c WT mice and TRIF-deficient or TLR4-deficient (C57BL/10ScSn) mice and their C57BL/6 WT controls were purchased from Jackson Laboratories (Bar Harbor, ME, USA). The TLR4-deficient mice used in our studies have been backcrossed seven generations by Jackson Laboratories onto the C57BL/6 strain. MyD88-deficient mice on a BALB/c background were obtained from Drs. Daniel Goldstein (Yale University School of Medicine, New Haven, CT, USA) and Shizuo Akira (Osaka University, Japan) and bred in a facility approved by the Association of Assessment and Accreditation of Laboratory Animal Care International. Mice were provided free access to autoclaved food and water. Procedures were carried out in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals, the Association for Research in Vision and Ophthalmology Statement for the Use of Animals in Ophthalmic and Vision Research, and guidelines designated by Oregon Health and Science University Institutional Animal Care and Use Committee policies (Portland, OR, USA).

Production of BM chimeras

BM radiation chimeras were created using WT or TLR4^−/− mice as BM donors, as described previously [23]. Briefly, the femur and tibia of donor mice were harvested, and the shafts were centrifuged at 10,000 rpm for 60 s at 4°C. Following erythrocyte lysis, the pellet was resuspended in sterile DMEM at the desired concentration of BM cells. WT or TLR4^−/− mice were irradiated with two doses of 6 Gy, 4 h apart, and received an injection of 2 × 10⁶ BM cells (in 150 μl) from donor WT or TLR4^−/− mice into the lateral tail vein (∼2 h after the second dose of irradiation). Antibiotic water (1.1 g/l neomycin; Sigma-Aldrich, St. Louis, MO, USA) was given to recipient mice for 7 days before and 2 weeks after irradiation.

To confirm chimerism and to ensure irradiation doses used were adequate to remove circulating leukocytes from recipient mice, we completed a pilot study in which irradiated CD45.2 C57BL/6 mice were reconstituted using BM from CD45.1 C57BL/6 mice. Single-cell suspensions from splenic tissues were prepared from untreated and recipient CD45.2 mice for flow cytometric analysis. Briefly, cells were resuspended in FACS buffer (1% BSA/PBS/10 mM NaN₃), and aliquots were prepared for further staining. Aliquots of splenic single-cell suspensions were blocked with 5% normal rat serum for 15 min and then stained with mouse PE-anti-CD45.1 (1/100; BD Biosciences, San Jose, CA, USA) and FITC-anti-CD45.2 (1/100; BD Biosciences). All antibodies were diluted in PBS containing 1% BSA. Negative controls and single fluorochrome controls were performed to allow accurate calibration. Isotype control antibodies were used to ensure the specific staining of each antibody. All samples were analyzed by flow cytometry using CELLQuestPro software (BD Biosciences). Flow cytometry (Fig. 1) revealed that by 8 weeks, ∼90% of splenocytes in CD45.2 mice were CD45.1⁺ cells.

Figure 1. Flow cytometric analysis of splenocytes confirmed chimerism in irradiated CD45.2-recipient mice, which were reconstituted using BM from CD45.1 mice.

As expected, the untreated host mice lacked CD45.1+ cells. Eight weeks after irradiation and reconstitution, only 4.87% of splenocytes in the chimera were of CD45.2-recipient origin, and nearly 90% of splenocytes were CD45.1⁺ cells. Data are representative from one single study out of 10.

Induction of EIU

Mice were administered 250 ng ultrapure Escherichia coli 0111:B4 LPS (Invivogen, San Diego, CA, USA) in 4 μl pyrogen-free PBS by local ivt injection while anesthetized with 1.7% isoflurane in oxygen. Ultrapure LPS was used to avoid the activation of TLRs other than TLR4 by contaminants commonly found in standard LPS preparations. Control mice were injected with PBS alone. For BM chimera studies, EIU was induced bilaterally 8 weeks after irradiation and BM reconstitution.

Intravital microscopy

Ocular inflammation in BALB/c and MyD88^−/− mice was evaluated at 4 h and 6 h after ivt LPS injection using intravital microscopy [24], which measures the leukocyte response within the iris vasculature and extravascular tissue. Briefly, mice were injected with 35 mg/kg rhodamine 6G (Sigma-Aldrich) to label all circulating leukocytes and anesthetized with 1.7% isoflurane in oxygen. Digital videos of 10-s duration from three independent regions of the iris vasculature were captured using a black and white video camera (Kappa Scientific, Gleichen, Germany) on an epifluorescence microscope (modified Orthoplan; Leica, Wetzlar, Germany). The diameter and length of each vessel segment or iris tissue and leukocyte phenotype (rolling, adhering, infiltrating) were quantified off-line with ImageJ analysis software (developed by Wayne Rasband, NIH, Bethesda, MD, USA), as described previously [24, 25]. Intravital microscopy was not performed on pigmented mice, as the pigment in the iris interferes with the detection of fluorescence.

Histological assessment of EIU

Eyes were harvested at 4, 6, and 24 h post-ivt LPS injections. Eyes were formalin-fixed overnight, embedded in paraffin, and sectioned at 7 μm through the pupillary-optic nerve axis at four different depths. Sections were stained with H&E, and the numbers of infiltrating leukocytes within the aqueous humor were counted in four sections across the four different levels by two masked observers. Adobe Photoshop was used to perform final image processing (Adobe Sytems, San Jose, CA, USA).

Cytokine analysis

Eyes, spleen, and plasma were harvested 4 h post-saline or LPS treatment. Protein was extracted from eye tissue and spleen as described previously [26]. Briefly, pooled samples were homogenized in lysis buffer containing a protease inhibitor cocktail (Roche, Mannheim, Germany) and protein concentrations determined using a BCA kit (Pierce-Endogen, Rockford, IL, USA). Plasma samples were centrifuged in filter tubes (Millipore, Billerica, MA, USA; pore size 0.22 μm) to remove any insoluble material. Equal amounts of protein were assessed to measure the production of MyD88-associated cytokines, (IL-6 and IL-1β), TRIF-related cytokines (MCP-1 and RANTES), as well as MIP-2/CXCL2 and keratinocyte-derived chemokine/CXCL1 by standard ELISA (Duoset or Quantikine ELISA kits, R&D Systems, Minneapolis, MN, USA) or by multiplex ELISA (Luminex, Millipore). Luminex data were analyzed using BeadView Multiplex Analysis software (version 1.0, Upstate, Lake Placid, NY, USA).

Statistical analysis

Data are represented as mean ± sem. Mean differences were compared, and statistical differences were determined by Student's t test (Prism; GraphPad Software, La Jolla, CA, USA). Differences were considered statistically significant when P < 0.05.

RESULTS

Cytokine production following local LPS treatment is reduced in the absence of MyD88 and TRIF

We examined the ocular production of a variety of cytokines from whole eye tissue in MyD88^−/− and TRIF^−/− mice and their congenic WT control mice, BALB/c and C57BL/6, respectively, 4 h following ivt LPS treatment (Fig. 2). The local injection of saline induced minimal, if any, production of cytokines within the eye (Fig. 2). Following an ivt injection of LPS, TRIF- and MyD88-related cytokines were produced in abundance in ocular tissue (Fig. 2). As would be expected from known pathways, the production of the MyD88-associated cytokine IL-6 was markedly less in MyD88-deficient mice after local LPS treatment. Similarly, the production of IL-1β was also significantly less in MyD88^−/− mice, although the production of this cytokine was also lower in TRIF^−/− mice following ivt LPS treatment. TRIF deficiency also resulted in a significant reduction in the ocular production of the TRIF-associated cytokines, MCP-1 and RANTES, following locally administered LPS. The expression of MIP-2/CXCL2 was induced in ocular tissue by LPS administration and was markedly reduced in the absence of MyD88 (Fig. 2).

Figure 2. Differential cytokine production in the eyes of TRIF KO, MyD88 KO, and congenic control mice following LPS challenge.

Cytokine levels (MIP-2, IL-6, IL-1β, MCP-1, and RANTES) were measured by ELISA 4 h after ivt injection of 250 ng LPS (n=6) or saline (n=3); *P < 0.05.

The absence of MyD88 diminishes the recruitment of leukocytes within the iris and aqueous humor in response to LPS

To determine the effect of MyD88 deficiency on the intravascular and cellular responses in the eye to locally administered LPS, MyD88^−/− mice and BALB/c WT mice were administered 250 ng LPS by ivt injection. The intravascular response in the iris was measured by intravital microscopy at 4 h and 6 h post-local LPS treatment. The numbers of rolling, adherent, and infiltrating cells were markedly reduced in MyD88-deficient mice compared with WT control mice at both time-points following LPS injection (Fig. 3A). These observations in vivo were confirmed by histological assessment that demonstrated significantly fewer infiltrating cells in the aqueous humor of MyD88^−/− mice at the 4 h and 6 h time-points (Fig. 3B). Although some infiltrating cells were observed in MyD88-deficient mice at 24 h, this number of cells in the aqueous humor was markedly lower than in WT mice (Fig. 3B and C). These data show that the intravascular response to LPS in the iris and the cellular response to LPS in the aqueous humor are MyD88-dependent.

Figure 3. The absence of MyD88 results in a diminished ocular cellular response to local LPS treatment.

Intravital microscopy and histological analysis followed local administration of 250 ng LPS in BALB/c WT and MyD88 KO mice. (A) Rolling, adhering, and infiltrating cells were markedly reduced in the absence of MyD88 at 4 h and 6 h after ivt injection of LPS. *P < 0.00005; 6 h, P < 0.0005 (n=12 eyes/group). (B) Cellular infiltration of the aqueous humor in WT and MyD88 KO mice at 4, 6, and 24 h following ivt LPS treatment. Cellular response was abolished in MyD88 KO mice at 4 h and 6 h, but some cells were evident in the aqueous humor at 24 h. Images taken at 200× original magnification. (C) The number of infiltrating cells in the aqueous humor was higher in WT compared with MyD88 KO mice at 4, 6, and 24 h after ivt injection of LPS; n = 10–12 eyes/treatment group/time-point.

The absence of TRIF does not alter the ocular inflammatory response to locally administered LPS

To examine the effect of TRIF deficiency on the ocular response to local LPS treatment, TRIF^−/− and C57BL/6 WT mice received 250 ng LPS by ivt injection. The cellular response was examined by histological analysis, and the numbers of infiltrating cells in the anterior chambers were counted. Histopathology and cell counts revealed that WT and TRIF-deficient mice had similar cellular responses in the eye to locally administered LPS (Fig. 4).

Figure 4. The absence of TRIF does not alter the cellular response to LPS in the eye.

(A) Cellular infiltration of the anterior segment at 4, 6, and 24 h after 250 ng ivt LPS injection was similar in WT and TRIF KO mice. (B) The number of cells in the aqueous humor following local injection of LPS was not altered in the absence of TRIF. Each dot in the scattergram represents an individual eye. At least six mice have been studied for each time-point. In some cases, both eyes are not shown, as the sections from one were inadequate for histological interpretation.

Non-BM-derived resident ocular cells with TLR4 deficiency prevent the development of EIU

We next set out to determine whether BM-derived (radiation-sensitive) or non-BM-derived (radiation-resistant) cells were key contributors in the response to local LPS treatment within the eye. We produced three different types of BM chimeric mice using irradiated C57BL/6 WT mice and TLR4^−/− mice: WT recipient mice, which received BM from WT donors (WT→WT), and TLR4^−/− mice (TLR4^−/−→WT) or TLR4^−/−-recipient mice, which were reconstituted with BM from WT donor mice (WT→TLR4^−/−). Chimeric mice were treated with 250 ng LPS by ivt injection, 8 weeks after irradiation, and BM reconstitution, as previous studies have shown almost complete turnover of BM-derived cells within the uveal tract and cornea of the eye by 8 weeks following BM replacement [27, 28]. Non-irradiated TLR4^−/− mice were unresponsive to ivt LPS treatment, whereas WT → WT chimeric mice produced a normal cellular response in the eye, 6 h after local administration of LPS (Fig. 5). Surprisingly, irradiated WT mice, which had received BM from TLR4^−/− mice (TLR4^−/−→WT), also produced a normal cellular response to LPS, whereas WT → TLR4^−/− chimeric mice were hyporesponsive to LPS treatment (Fig. 5). These data suggest that radiation-resistant, non-BM-derived ocular cells, such as the iris endothelial cells or nonpigmented ciliary body epithelial cells, play a greater role in the development of EIU than BM-derived macrophages and DCs residing in the eye.

Figure 5. Induction of EIU requires TLR4 in radiation-resistant cells and not in BM-derived cells.

(A) TLR4^−/− mice were unresponsive to local LPS treatment. Six hours after ivt LPS injection, C57BL/6 WT mice, which had received BM from WT mice (WT→WT), displayed a normal response. There was a markedly diminished response in TLR4^−/− mice, which had received BM from WT mice (WT→TLR4^−/−). The cellular response was normal in WT mice, which had received BM from TLR4^−/− mice (TLR4→WT). (B) The numbers of cells infiltrating the anterior chambers were counted and compared between BM chimeric groups. The number of infiltrating cells was reduced significantly in WT → TLR4^−/− chimeric mice. Each dot in the scattergram represents an individual eye.

DISCUSSION

EIU has become a standard rodent model, which has been the subject of more than 400 publications based on a January 2011 keyword search of the National Library of Medicine. Although the effects of endotoxin in the eye have been appreciated for decades, basic questions about the ocular sensitivity to endotoxin remain. TLR4 signaling is unique in that it can activate two TLR adaptor molecule pathways controlled by MyD88 and controlled by TRIF. Although it is known that endotoxin induces the expression of MyD88 in the rat uveal tract [29], the relative contribution of those two pathways in EIU was previously unstudied. Our data clearly show that locally injected LPS induces the MyD88 and TRIF pathways within the eye.

In studies of mice that lack MyD88 or TRIF, we found that MyD88-dependent and not TRIF-dependent cytokines are necessary for EIU. At first glance, this is contradictory to studies that have implicated MCP-1, a TRIF-dependent cytokine, in EIU. For example, MCP-1 KO mice develop reduced eye inflammation after LPS challenge [30]. TRIF KO mice, however, are deficient in other cytokines, in addition to MCP-1. In the model of experimental autoimmune uveitis, another TRIF-dependent cytokine, RANTES, is thought to be anti-inflammatory [31]. Thus, although some TRIF-dependent cytokines might contribute to EIU, their effects might be counterbalanced by other TRIF-dependent cytokines that are anti-inflammatory, such as RANTES and IFN-β. By eliminating inflammatory and anti-inflammatory cytokines, the net effect on the inflammation may be neutral.

In addition to identifying the essential LPS-activated signaling pathway(s), we wanted to discover the cell type(s) required for this signaling. The iris has a substantial population of BM-derived macrophages and DCs [23]. As the monocyte/macrophage is arguably the most responsive cell to LPS, we anticipated that the absence of TLR4 in BM-derived cells in the iris would diminish EIU. A recent study about the cornea indicated that BM-derived cells were critical in the response of that ocular tissue to LPS [32]. Another recent report documented that the number of TLR4-expressing cells increases in the anterior uveal tract after LPS injection [17]. To our surprise, inflammation was absent in mice with BM derived from WT mice but with epithelial and endothelial cells that were TLR4-negative. The importance of radiation-resistant or non-BM-derived cells in the response to LPS has precedence. For example, other studies using BM chimeric mice have shown that nonhematopoietic cell TLR4 signaling is essential for the development of LPS-induced ileus [15]. Likewise, TLR4 signaling in endothelial cells is critical to LPS-induced inflammation in the lung [33]. Although the direct activation of endothelial cells by LPS in the CNS mediates rolling and adhesion in the cerebral vasculature, leukocyte infiltration of the CNS tissue requires microglial activation by intracerebroventricular injection of LPS [34, 35]. On the other hand, although mice that exclusively express TLR4 on endothelial cells can effectively respond to and clear bacteria following systemic infection, neutrophil accumulation within lung tissue following intratracheal administration of LPS is dependent on BM-derived cells [33]. Immunohistology has shown that TLR4 is expressed by epithelial cells in the anterior segment of the eye [16].

Gene therapy has been successfully applied to genetic ocular disease and might one day succeed in the treatment of ocular inflammatory disease. As vectors can preferentially target different cell populations, it is critical to learn which cells are responsible for inflammation within a tissue milieu.

A limitation of our study is that we did not demonstrate directly that our radiation protocol created chimerism within the eye. The pigmentation of the iris interfered with attempts to verify chimerism by immunohistology (unpublished observations). We did, however, use a radiation protocol, which has been previously reported to establish ocular chimerism [23], and we confirmed that the radiation doses chosen created chimerism within the spleen. Our TLR4 KO mice had been backcrossed seven generations unto C57BL/6 mice by Jackson Laboratories. Ten generations are ideal to insure histocompatibility. Graft-versus-host disease can result in uveitis [36], but we found no evidence of background intraocular inflammation in WT mice that received TLR4 marrow, nor did we observe rash or diarrhea, which might have suggested graft-versus-host illness. Furthermore, host-versus-graft reaction would have impaired the health and survival of irradiated mice. For these reasons, we think that it is highly likely that chimerism was established. Some macrophages might produce cytokines that are primarily anti-inflammatory. A comparison of the inflammation in WT mice that received TLR4 KO marrow with WT mice that received WT marrow is not consistent with the net effect of BM-derived cells within the iris being primarily anti-inflammatory.

Thus, our study shows two important and surprising observations. Although the TRIF-dependent cytokine, MCP-1, has been convincingly implicated in EIU [30], the net effect of all TRIF-dependent cytokines does not enhance EIU. Second, the cells primarily responsible for EIU are radiation-resistant and therefore, not likely to be derived from BM. This information is important for the design of anti-inflammatory therapies that target specific cell types.

AUTHORSHIP

J.K. designed and performed most experiments, analyzed data, and wrote the manuscript. S.T. and S.G. performed experiments. S.R.P. contributed to study design and edited the manuscript. H.L.R. and J.T.R. designed experiments, analyzed data, and edited the manuscript.