Aquaporin-4 Peptide Injection in Mice Induces Retinal and Optic Nerve Alterations That Simulate Those of Neuromyelitis Optica Spectrum Disorder

Abstract

Purpose: To investigate functional and structural changes in the retinas and optic nerves of mice immunized with an aquaporin-4 (AQP4) peptide, which was previously shown to induce paralysis mimicking human neuromyelitis optica spectrum disorder (NMOSD).

Methods: Electroretinography and histological analyses were used to evaluate retinal function, and inflammatory cell infiltration or glial fibrillary acidic protein immunoreactivity in the retinas and optic nerves of AQP4-immunized mice. Additionally, the blood–retinal barrier function was assessed by Evans blue dye injection to measure in vivo retinal vascular permeability and by in vitro transendothelial electrical resistance using mouse primary retinal microvascular endothelial cells exposed to serum from AQP4-immunized mice.

Results: AQP4 immunization led to a significant reduction in b-wave and scotopic threshold response amplitudes, indicating impaired inner retinal function. Histological analysis revealed inflammatory cell infiltration at the optic nerve head. Whole-mounted retinal glial fibrillary acidic protein immunoreactivity showed aberrant Müller cell activation, particularly in the juxtapapillary region. AQP4-immunized mice exhibited increased retinal Evans blue dye leakage and mouse retinal microvascular endothelial cells exhibited reduced transendothelial electrical resistance, indicating blood–retinal barrier disruption.

Conclusions: AQP4 immunization induced functional impairment, inflammatory cell infiltration, glial cell activation, and blood–retinal barrier disruption in the retinas and optic nerves in mice, which mimics human NMOSD-associated optic neuritis.

Translational Relevance: Along with the previously reported development of paralysis, this study indicates that AQP4 peptide-immunized mice can be considered an animal model of NMOSD and a powerful tool for further understanding the pathophysiology of NMOSD-associated optic neuritis and for assessing the efficacy of drugs in its treatment.

Introduction

Neuromyelitis optica spectrum disorder (NMOSD) is a severe autoimmune disease accompanied by longitudinally transverse myelitis and optic neuritis.^1,2 Its pathogenesis is characterized by autoantibodies against aquaporin-4 (AQP4), a water channel protein that is expressed mainly at astrocyte endfeet.^3,4 It is also well-recognized that visual prognosis after the onset of optic neuritis is poorer in NMOSD than that with optic neuritis brought on by other etiologies such as multiple sclerosis.^5–9 We previously reported that, in NMOSD-associated optic neuritis, the degree of visual dysfunction is proportional to the thinning of the retinal nerve fiber layer and combined inner macular layers composed of retinal ganglion cells (RGCs) as measured using optical coherence tomography.⁵ However, the precise mechanisms through which the RGC and optic nerve damage is induced in NMOSD remain elusive. 

Numerous studies have already demonstrated that astrocytes and Müller cells are deeply involved in maintaining the proper function of the blood–retinal barrier (BRB).¹⁰ Cytokines secreted by astrocytes are essential for the formation of tight junctions in the endothelial cells of the superficial retinal vasculature.¹¹ Conversely, damage to astrocytes leads to the secretion of cytokines that disrupt the BRB.¹⁰ 

Many animal models mimicking the pathologies of human NMOSD have been reported^12–24; however, few of these models have sought to replicate the structural^18–21 and in particular the functional^22,23 alterations that occur in the optic nerve. In addition, most of the previous NMOSD models require passive or direct transfer of immunoglobulin G (IgG) from NMOSD patients^12–19 or AQP4-reactive T helper 17 cells from AQP4 peptide-immunized AQP4^−/− mice²⁰ into animals to produce NMOSD-mimicking pathologies. 

Our group has established a novel experimental autoimmune mouse model by intradermal immunization with AQP4 peptide.²⁴ This model exhibits reversible myelitis with limb paralysis, AQP4 loss, and complement deposition on the spinal cord. Although AQP4 peptide immunization was shown to induce the pathophysiology of NMOSD-associated myelitis, it remains unclear whether this model also reflects the pathophysiology of optic neuritis. 

In this study, we conducted electrophysiological and histochemical studies to identify functional and structural changes in the retinas and optic nerves of AQP4 peptide-immunized mice. These mice showed pathologies mimicking human NMOSD-associated optic neuritis. 

Materials and Methods

Animals

Female C57BL/6J mice (8 weeks old; The Jackson Laboratory Japan, Inc., Kanagawa, Japan, and Clea Japan, Inc, Tokyo, Japan) were used. All mice were fed ordinary laboratory chow and allowed free access to water under a constant light and dark cycle of 12 hours. 

All animal procedures were conducted in the Department of Surgery, Division of Ophthalmology, Kobe University Graduate School of Medicine (electrophysiological experiments) or the Product Research Department, Chugai Pharmaceutical Co., Ltd. (histochemical, immunohistochemical, and in vitro experiments) in accordance with the guidelines set forth in the ARVO Resolution on Care and Use of Laboratory Animals. In addition, all experimental protocols were approved either by the Animal Care Committee of Chugai Pharmaceutical Co., Ltd, and conformed to the Guide for the Care and Use of Laboratory Animals published by Institution of Laboratory Animal Research or approved by the Animal Care Committee of the Kobe University Graduate School of Medicine (P220307). 

Immunization

Mice were immunized intradermally at multiple places on day 0 with 200 µg of AQP4 p201-220 peptide trifluoroacetic acid salt (synthesized by the Peptide Institute, Inc., Osaka, Japan) emulsified in complete Freund's adjuvant (CFA), as previously reported.²⁴ CFA was made by supplementation of incomplete Freund's adjuvant (263910; BD, Frankin Lakes, NJ, USA) with 5 mg/mL Mycobacterium tuberculosis extract H37Ra (231141; BD). In addition, the mice received 300 ng pertussis toxin (181, List Biological Laboratories, Campbell, CA, USA) intravenously on day 0 and day 2.²⁴ Control mice were administered CFA and saline alone. 

Assessment for Clinical Score

Clinical scores were assigned according to the following scale:²⁴ 0 (no disease), 1 (limp tail), 2 (hind limb weakness), 3 (hind limb paresis), 4 (hind limb paralysis), 5 (hind limb and fore limb paralysis) and 6 (moribundity and death). Mice with a clinical score of 4 or higher were euthanized in accordance with humane end point criteria. 

Electroretinography

All animals were dark-adapted overnight before electroretinography (ERG) recording, and all procedures were performed under dim red light as previously reported.^25,26 Mice were anesthetized with an intraperitoneal injection of ketamine (100 mg/kg) and xylazine (10 mg/kg) and were positioned on a heating pad to maintain a body temperature of 37°C. After dilating the pupils with 0.4% tropicamide eye drops (Santen Pharmaceutical Co., Osaka, Japan), a contact lens electrode embedded with gold wire was placed on the cornea as an active electrode (MAYO Corporation, Aichi, Japan), and a chloride silver plate was placed in the mouth as a reference electrode. The ERGs were recorded using a commercially available instrument equipped with a Ganzfeld bowl (PuREC PC-100; MAYO Corporation). Scotopic recordings were obtained from overnight dark-adapted animals at the following increasing light intensities. Responses were amplified 10,000 times and bandpass-filtered at 0.3 to 500.0 Hz. For the recording of positive and negative scotopic threshold responses (STRs), serially increasing luminescence intensities of −6.1, −5.5, −5.1, −4.6, and −4.1 log sc td s were used. Responses were amplified differentially and bandpass filtered at 0.125 to 50.000 Hz, and responses from 80 repeated stimuli for each intensity were averaged. Photopic recordings were performed after 5-minute light adaptation intervals on a background light intensity of 1.5 log sc td. Responses were amplified differentially and bandpass-filtered at 0.3 to 300.0 Hz, and the responses from 30 repeated stimuli for each intensity were averaged. 

Immunohistochemistry

Mice were euthanized under isoflurane, and eyeballs enucleated. For histopathological analysis, eyeballs were post-fixed in Davidson's solution overnight followed by 10% neutral buffered formalin overnight, and then embedded in paraffin. The sections were stained with hematoxylin and eosin. Immunohistochemistry for complement component 3 (C3) was performed using anti-C3 antibody (EPR19394; Abcam, Cambridge, UK) with the VENTANA automated slide stainers (Ventana Medical Systems Inc, Tucson, AZ, USA). Infiltrating cells and C3 deposition were morphologically evaluated by pathologists (S.O. and K.Y.), and retinal thickness was measured using ImageJ (NIH, Bethesda, MD, USA). For glial fibrillary acidic protein (GFAP) staining, eyeballs were post-fixed with phosphate-buffered saline (PBS)-diluted 4% paraformaldehyde (15710; Electron Microscopy Sciences, Hatfield, PA, USA) for 2 hours on ice. After washing in PBS, retinas were flat mounted and permeabilized with blocking buffer (10% fetal bovine serum [F2442; Merck, Darmstadt, Germany] and 0.2% Triton X-100 [X100; Merck] in PBS) for 1 hour at room temperature. Retinas were then incubated with anti-GFAP monoclonal antibody (1:250, 13-0300; Thermo Fisher Scientific, Waltham, MA, USA) at 4°C overnight.²⁶ After washing with PBS, the retinas were incubated with secondary antibody (1:250, A48262; Thermo Fisher Scientific) and Hoechst (1:1000, 346-07951; Dojindo Laboratories, Kumamoto, Japan) at 4°C overnight. PBS-washed samples were then mounted with ProLongGlass Antifade Mountant (P36980; Thermo Fisher Scientific) and observed under a confocal microscope (A1R; Nikon, Tokyo, Japan). 

Retinal Vascular Permeability

For Evans blue imaging, after 14 days from immunization, mice were given 10 minutes of intravenous administration of 2% v/w Evans blue (056-04061; FUJIFILM Wako Pure Chemical Corporation, Osaka, Japan) dissolved in saline, and were then euthanized. Retinas were dissected and mounted after 30 minutes of post-fixation with 4% paraformaldehyde (09154-85; Nacalai Tesque, Kyoto, Japan). 

For mouse IgG staining, retinas were dissected from the eyes after 2 hours of 4% paraformaldehyde fixation. Retinas were washed three times with PBS and blocked for 1 hour with a blocking buffer. The blocking buffer was then exchanged for a primary antibody for Alexa Fluor 488–conjugated anti-mouse IgG (Ab150117; Abcam, Cambridge, UK) diluted in blocking buffer (1:200) at 4°C. The next day retinas were reacted overnight with Hoechst 33342 after washing three times with PBS. Retinas were then washed again three times with PBS and mounted with ProLong Glass Antifade Mountant. Samples were observed under a fluorescence microscope (BZ-X810; Keyence, Osaka, Japan). 

The entire flat-mounted retina was used as the region of interest, and the average intensity per pixel of Evans blue or anti-mouse IgG fluorescence was quantified by ImageJ. 

Transendothelial Electrical Resistance (TEER) Measurement

Primary mouse retinal microvascular endothelial cells (mRMECs) were purchased from Cell Biologics (mRMECs, C57-6065; Cell Biologics, Chicago, IL, USA). mRMECs were cultured on cell culture inserts (3 µm pore size, 9323012; SABEU, Northeim, Germany) with Endothelial Cell Medium (M1168; Cell Biologics), and TEER was measured over time by cellZscope+ (nanoAnalytics, Münster, Germany). Two days after initiation of measurements, 10% v/w medium-diluted serum from CFA-administered or AQP4-immunized mice was added to the upper compartment of the wells. Individual TEER values were normalized to values just before the addition of serum. 

Statistical Analyses

All data are expressed as mean ± standard deviation. The n values refer to the number of individual animals in each group on which experiments were performed. The statistical significance of differences was determined by using an unpaired t test for the comparison of the two groups. In cases where both eyes were from an identical mouse, Satterthwaite's t test using restricted maximum likelihood estimation was used. These tests were conducted after confirming the normality of the sample distribution by the F-test or Shapiro–Wilk test. P values of less than 0.05 were considered significant. Statistical analyses were performed using JMP version 15.0.0 software (SAS Institute, Cary, NC, USA). A mixed effects model or generalized mixed effects model was used to conduct repeated comparisons, using R (version 4.3.2) with the lmerTest (version 3.1.3) packages. 

Results

AQP4 Peptide Immunization Reduced b-Wave and STR Amplitude of ERG, but not a-Wave Amplitude

As reported previously,²⁴ intradermal immunization with AQP4 peptide led to development of autoimmune myelitis in all AQP4-immunized mice (Fig. 1A). 

Figure 1.

 

AQP4 peptide immunization reduced the amplitudes of ERG responses. (A) AQP4-immunized mice developed limb paralysis. Black circles, CFA-administered group; white squares, AQP4-immunized group. Mean ± standard deviation, n = 5. (B) Representative waves from each group. Gray lines, CFA-administered group; red lines, AQP4-immunized group. (C and D) Averaged amplitudes of b-wave + positive STR (C) and a-wave (D), respectively. Black circles, CFA-administered group; red squares, AQP4-immunized group. Mean ± standard deviation, n = 10 eyes from 5 mice. *P < 0.05, **P < 0.01, ***P < 0.001 vs. CFA by Satterthwaite's t test.

Compared with controls, AQP4-immunized mice exhibited significantly reduced b-wave amplitudes elicited by stimulus of 10 cd s/m² (1.85 log sc td s) (P < 0.05), positive STR elicited by stimuli of 6.0 × 10⁻⁶ and 1.0 × 10⁻⁵ cd s/m² (−4.37 and −4.15 log sc td s; P < 0.01 and P < 0.05, respectively), and negative STR elicited by stimuli of 3.0 × 10⁻⁶ and 1.0 × 10⁻⁵cd s/m² (−4.67 and −4.15 log sc td s, respectively; P < 0.05 for both stimuli). In comparison, there were no significant changes in a-wave amplitude elicited by stimulus of 1.85 log sc td s between controls and AQP4-immunized mice (P = 0.27) (Figs. 1B and 1C). These results demonstrate that AQP4 peptide immunization adversely affects inner retinal layer functions associated with bipolar cells, RGCs, and Müller cells, but does not impact photoreceptors, which correlates with the retinal dysfunction seen with NMOSD-associated optic neuritis in humans.^5,27,28 

AQP4 Peptide Immunization Induced Infiltration of Inflammatory Cells Into the Head of the Optic Nerve and Its Surroundings

Retinas of AQP4-immunized mice were stained with hematoxylin and eosin. Unexpectedly, there was no significant difference in the thickness of each retinal layer between CFA-administered mice and AQP4-immunized mice (Supplementary Table S1). No infiltrating inflammatory cells were observed in the retinas of AQP4-immunized mice; however, in four out of five AQP4-immunized mice, infiltrating inflammatory cells, mainly granulocytes, were observed in the optic nerve head parenchyma and its surrounding connective tissues (Fig. 2), which implies that AQP4 peptide immunization may induce infiltration of inflammatory cells into the optic nerve head. 

Figure 2.

 

In AQP4-immunized mice, infiltrating inflammatory cells were observed in the optic nerve head, but not in the retina. (Top) Hematoxylin and eosin staining of retinal sections. Scale bars, 500 µm. (Middle) Enlargement of the area within the black circle in top (optic nerve head). Scale bars, 50 µm. (Bottom) Enlargement of the area within the white circle in the top image (retina). Scale bars, 50 µm.

It is well-known that the anti-AQP4 antibody induces pathology through complement-dependent cytotoxicity. Previously, we reported that AQP4-immunized mice developed myelitis with complement deposition²⁴; however, whether it is recognized in the retina remains unclear. Therefore, retinal and optic nerve deposition of complement C3 was visualized with immunohistochemistry. No C3 deposition was observed in the retinas of AQP4-immunized mice (data not shown). In contrast, C3 deposition was observed in the optic nerve parenchyma of these mice, consistent with the observed inflammatory cell infiltration (Supplementary Fig. S1). 

AQP4 Peptide Immunization Activated Müller Cells in the Juxtapapillary Region

We suspected that one of the major causes of the observed attenuation of b-wave amplitude may be due to aberrant activation and dysfunction of Müller cells. To test if this were the case, the flat-mounted retinas were subjected to immunostaining for GFAP (Fig. 3). It is well-known that, under normal conditions, GFAP immunoreactivity is confined to astrocytes resident in the superficial layer and appears as a tentacle-like pattern, whereas under conditions of retinal or optic nerve stress, GFAP immunoreactivity is induced in activated Müller cells and appears as a punctate or granular pattern of GFAP immunostaining in the deeper layers of flat-mounted retina.^29,30 Figure 3A shows representative maximum intensity projection images of flat-mounted retinas with GFAP immunostaining, where all slices of images at different microscopical depths of focus were superimposed to integrate the z-axis directional image information. 

Figure 3.

 

AQP4 peptide immunization induced punctate patterns of GFAP, implying juxtapapillary Müller cell activation. (A) Representative images of the retina around the optic nerve head, created by maximum intensity projection. Scale bars, 100 µm. (B) Single-slice images of the deep layer with the greatest number of GFAP-immunoreactive dots. Scale bars, 100 µm. (C) The number of punctate or granular-patterned areas of GFAP immunoreactivity in the single-slice images. Black circles, CFA-administered group; white squares, AQP4-immunized group. Mean ± standard deviation, n = 17 eyes from 10 mice in the CFA group, 14 eyes from 7 mice in the AQP4 group. *P < 0.05 vs. CFA by Satterthwaite's t test. (D) Logistic mixed-effects model regression of clinical scores and the GFAP-immunoreactive dot count (C).

CFA-administered mice exhibited a tentacle-like pattern of GFAP immunoreactivity rooted at the optic nerve head, which is intensively immunostained as a circular shape (Fig. 3A, left). There is almost no punctate or granular immunoreactivity. These findings are consistent with the fact that, under normal physiological conditions, GFAP is expressed in astrocytes but not in Müller cells.^28,29 In contrast, AQP4-immunized mouse retinas prominently exhibited punctate or granular patterns of GFAP immunoreactivity around the optic disc, with less intense tentacle-like GFAP immunoreactivity seen in the peripheral retina (Fig. 3A, right). These findings imply aberrant Müller cell activity with mildly reduced astrocyte GFAP immunoreactivity. 

To quantify the degree of Müller cell activity, we counted the punctate or granular GFAP-immunoreactive dots in the Z-stack slice showing the highest GFAP intensity within a 640 × 640 µm area centered on the optic disc center (Fig. 3B). The AQP4-immunized mice were found to show a significantly greater dot number (163.0 ± 126.8) than the CFA-administered mice (75.47 ± 23.73; P < 0.05) (Fig. 3C). Furthermore, these dot counts significantly correlated with clinical score reflecting spinal cord injury (P < 0.05) (Fig. 3D); that is, the higher the clinical score, the more punctate or granular GFAP-immunoreactive dots, indicating that the degree of Müller cell activation corresponded with the degree of spinal cord dysfunction. 

AQP4-immunized Mice Exhibited Increased Retinal Vascular Permeability

Because Müller cells express AQP4 on their endfeet, which are integral to the structure of the BRB, we investigated BRB function in AQP4-immunized mice. For quantification of the blood–brain barrier (BBB) disruption, the fluorescent intensity of leaked Evans blue or IgG is often used.²⁹ To investigate whether AQP4 peptide immunization affects vascular permeability, the mean fluorescent intensity of Evans blue on flat-mounted retinas was evaluated (Fig. 4A). Based on the observation that ERG amplitude decreased during the onset of myelitis, evaluation of Evans blue dye leakage was also performed within a similar timeframe. The mean dye intensity in AQP4-immunized mice was found to be significantly greater than that in CFA-administered mice (Fig. 4B). In parallel with this increased vascular permeability, the mean intensity of mouse IgG in the retina significantly increased in AQP4-immunized mice compared with CFA-administered mice (Fig. 4C). Given that a disrupted BRB likely facilitates access of IgG to the retinal parenchyma, we consider that the increased IgG intensity indicates increased deposition of autoantibodies within the retina in this model. 

Figure 4.

 

The barrier function of retinal blood vessels was decreased upon AQP4 peptide immunization. (A) Representative images of Evans blue leakage on flat-mounted retinas. Scale bars, 100 µm. (B) Quantitative analysis of mean Evans blue intensity for individual retinas. The entire retina was defined as the ROI, and mean pixel intensity within this ROI was quantified using ImageJ software. Black circles, CFA-administered group; white squares, AQP4-immunized group. Mean ± standard deviation, n = 8 eyes from 4 mice. *P < 0.05 vs. CFA by Satterthwaite's t test. (C) Representative images of total mouse IgG leakage on flat-mounted retinas. Scale bars, 100 µm. (D) Quantitative analysis of mean mouse IgG intensity for individual retinas. Black circles, CFA-administered group; white squares, AQP4-immunized group. Mean ± standard deviation, n = 8 eyes from 4 mice. *P < 0.05 vs. CFA by Satterthwaite's t test. ROI, region of interest.

The Barrier Function of Retinal Vascular Endothelial Cells Was Impaired by Serum From AQP4-immunized Mice

To investigate whether serum obtained from AQP4-immunized mice can induce breakdown of the BRB, we evaluated whether the AQP4-immunized mouse serum can change the TEER in commercially available primary mRMECs exposed to the serum (Fig. 5A). After 24 hours of exposure to AQP4-immunized serum, the TEER of mRMECs was decreased significantly relative to that of CFA serum-treated cells (Fig. 5B), indicating that serum of AQP4-immunized mice contains a factor or factors that increases retinal vascular permeability. 

Figure 5.

 

In vitro analysis demonstrated that serum from AQP4-immunized mice reduced the barrier function of retinal endothelial cells. (A) Change over time of transendothelial resistance (TEER) in mouse primary retinal microvascular endothelial cells (mRMECs) exposed to the serum of CFA-administered mice and AQP4-immunized mice. Black circles, CFA-administered group; white squares, AQP4-immunized group. Mean ± standard deviation, n = 6 for CFA serum-stimulated wells; 7 for AQP4 serum-stimulated wells. (B) TEER changes at 24 hours after serum exposure of (A). Black circles, CFA-administered group; white squares, AQP4-immunized group. Mean ± standard deviation, n = 6 for CFA serum-stimulated wells; n = 7 for AQP4 serum-stimulated wells. **P < 0.01 by Student's t test.

Discussion

Developing an experimental model of NMOSD is essential for understanding the underlying pathological mechanisms, which would facilitate identifying specific therapeutic targets. 

Several substantial studies of rodent models involving the passive or direct transfer of serum or IgG collected from NMOSD patients (NMO-IgG), with or without human complement, and with or without prior development of experimental encephalomyelitis have been reported.^12–18 These models induce permanent AQP4 and GFAP loss together with immunoglobulin and complement deposition, inflammatory cell infiltration, and neuronal loss primarily in the brain and spinal cord. Another methodology that induces encephalomyelitis experimentally is the passive transfer of AQP4-reactive T helper 17 cells to mice.²⁰ This model exhibits reversible paralysis, inflammatory cell infiltration into the central nervous system (CNS), including the optic nerve, and an acute increase in the thickness of inner retinal layers. 

However, although optic neuritis is one of the core clinical manifestations of NMOSD,^1,2 only a few studies have focused on optic nerve pathology, particularly assessment of in vivo visual function, in NMOSD model animals (Table). We previously demonstrated that the direct administration of serum from NMOSD patients to the optic nerve results in axonal damage and retrograde degeneration of RGCs.²¹ Zhang et al.²² demonstrated that the injection of serum from NMOSD patients into the subarachnoid space developed pathologies mimicking NMOSD-associated optic neuritis, along with reduced amplitude of visual evoked potential and pupillary light reflex. More recently, direct injection of anti-AQP4 monoclonal antibodies into the optic nerve has been reported to show macroglial and microglial activation and inflammatory reaction in the optic nerve.²³ 

Table.

 

Summary of Published and Present Studies Investigating Retino-Optic Nerve Pathology in NMOSD Animal Models

 

Although these models are all powerful tools for understanding the pathogenesis of autoantibodies or the contribution of leukocytes, none of them, with the exception of the study by Sagan et al.,²⁰ where AQP4-reactive Th cells are transferred to develop the model, allow us to explore the mechanisms of antibody production and various associated immune activities. 

The present study clearly demonstrated that AQP4 peptide immunization led to a decline in inner retinal layer function—as revealed electroretinographically by reduced positive and negative STRs and b-wave amplitudes—which was synchronized with the onset of myelitis symptoms. Thus, AQP4 peptide immunization alone is able to induce visual dysfunction via RGC functional impairment like that in patients with NMOSD. 

Furthermore, AQP4 peptide immunization induced the infiltration of inflammatory cells into the optic nerve head, complement deposition in the optic nerve, and a concentric increase in punctate or granular patterned GFAP immunoreactivity around the optic nerve head. However, no increase in cellular infiltration or granular GFAP immunoreactivity was observed in the peripheral retinal areas. 

Under normal conditions, GFAP is radially expressed in retinal astrocytes on the retinal surface.^29,30 In rodent models of elevated intraocular pressure,²⁹ diabetes,³⁰ and excitotoxicity,³¹ the upregulation of GFAP in Müller cells is known to increase granular GFAP immunoreactivity in flat-mounted retinas. However, it is noteworthy that, with AQP4 peptide immunization, the increase in granular GFAP immunoreactivity was confined to the peripapillary region, where increased cell numbers of infiltrating cells were observed in hematoxylin and eosin staining. This finding is in remarkable contrast with previous studies^29,30,32 that reported enhanced granular GFAP immunoreactivity throughout the entire retina. 

Our study also found that AQP4 peptide immunization caused widespread retinal vascular hyperpermeability in vivo, endothelial tight junction dysfunction in vitro, and possible IgG deposition in the retinal parenchyma. The BBB plays a crucial role in protecting against the entry of inappropriate factors into the CNS, and is composed of endothelial cells, pericytes, and astrocytes. In NMOSD, disruption of the BBB results in the penetration of anti-AQP4 antibodies into the CNS, where they bind to AQP4 protein, leading to complement- and antibody-dependent astrocytic cytotoxicity and CNS inflammation with leukocyte infiltration or oligodendrocyte injury-induced demyelination.^9–23,32,33 In addition, astrocyte dysfunction is fatal for the barrier function of the BRB,^10,11 as well as BBB.³⁴ Previous studies have shown that IgG from NMOSD patients affects astrocytes to promote interleukin-6 production, causing further permeability.³⁵ Presumably, interleukin-6–dependent BBB loosening may facilitate the access of anti-AQP4 antibodies to the astrocyte. The BRB, with a similar barrier function on the retina, might express the same dysfunction in our model. The observed changes in GFAP expression around the optic nerve head and the increased retinal vascular permeability after AQP4 immunization further support these findings, suggesting that astrocyte damage induced by AQP4 immunization contributes to BRB disruption. Moreover, the finding that serum from AQP4-immunized mice reduced the TEER of mRMEC in vitro implies that blood components exposed to the AQP4 antigen may damage retinal vascular endothelial cells, subsequently attacking the astrocytes that envelop the retinal vasculature in a retrograde manner. This process could potentially trigger a vicious cycle that accelerates further BRB disruption. 

In summary, the current AQP4 peptide immunization model replicates key features of NMOSD-associated optic neuritis in humans, including BRB breakdown, IgG and complement deposition, Müller cell activation, and RGC dysfunction, all synchronized with myelitis-related symptoms and pathologies. 

This study has limitations. The disease course of the present model is essentially reversible with respect to myelitis; that is, the limb paresis recovers over time in many cases, which is an atypical presentation course in human NMOSD patients. The present study is a transverse study, so further research is needed to determine whether the optic nerve lesions also spontaneously recover. Another limitation is the possibility of underestimation of the C3 deposition on the retina. As previously reported,³⁶ rodent immunohistochemistry that is fixed with Davison's solution often displays lower sensitivity to various antibody staining. Likewise, it should be noted that our results possibly did not fully reflect the intrinsic staining. 

In conclusion, AQP4 peptide immunization produced acute, albeit short-term, myelitis-like and optic neuritis-like symptoms and pathologies mimicking NMOSD with no requirement for either passive or direct transfer of human NMO-IgG and complement. With further refinement, this model may serve as a useful animal model of NMOSD-associated spinal cord and optic nerve lesions for understanding the pathogenesis of NMOSD and assessing possible treatment modalities. 

Acknowledgments

The authors thank Shinichi Onishi and Keigo Yorozu for their histopathological evaluation. 

Funded by Chugai Pharmaceutical Co., Ltd. departmental resources. 

Disclosure: M. Okuda-Arai, None; S. Mori, Chugai Pharmaceutical (R); Y. Katsura, Chugai Pharmaceutical Co., Ltd. (E); S. Miyake, Chugai Pharmaceutical Co., Ltd. (E); K. Serizawa, Chugai Pharmaceutical Co., Ltd. (E); M. Nakamura, Chugai Pharmaceutical Co., Ltd. (F)