Background
Neuromyelitis optica (NMO) is an autoimmune inflammatory disease of the central nervous system (CNS) that causes optic neuritis and transverse myelitis, leading to loss of vision and motor function, reviewed in [
1‐
4]. Affected tissues in NMO generally show astrocyte damage with complement activation, inflammatory cell infiltration, and demyelination. Most NMO patients are seropositive for immunoglobulin G autoantibodies against aquaporin-4 (AQP4), a water channel expressed on astrocytes. Human NMO pathology and rodent models of NMO produced by passive transfer of anti-AQP4 autoantibody (AQP4-IgG) suggest a pathogenesis mechanism in which AQP4-IgG binding to AQP4 causes primary complement- and cell-mediated astrocyte toxicity, with a secondary inflammatory response leading to oligodendrocyte injury, demyelination, and axon loss.
NMO patients often manifest retinal abnormalities including thinning of the retinal nerve fiber layer (RNFL) and ganglion cell layer (GCL) following episodes of optic neuritis [
5‐
7]. During acute episodes of NMO optic neuritis, the inner nuclear layer (INL) typically thickens [
8,
9], often coincident with inner retinal microcystic changes [
10,
11]. It has been speculated that retinal injury in NMO may in part be a primary disease manifestation, although it has been difficult to resolve direct retinal injury from retrograde axon degeneration due to retrobulbar optic neuritis. These retinal abnormalities are also seen, albeit at lower frequency, following optic neuritis in multiple sclerosis and a variety of severe noninflammatory optic neuropathies [
12‐
15].
The cellular makeup and structure of the retina is quite different from the optic nerve, spinal cord, and brain, where there are myelinated nerve fibers and AQP4 expression on astrocytes [
16]. In the retina, AQP4 is expressed in two types of retinal glia: Müller cells, which are a specialized form of radial glia spanning the retinal inner limiting to outer limiting membranes, with the cell body lying within the INL, and astrocytes, which are mostly localized in the RNFL along with nonmyelinated axons of retinal ganglion cells (RGCs).
Here, to investigate the possibility that NMO autoantibodies could produce primary retinal injury, we exposed rat retinas to AQP4-IgG in vivo by intravitreal injection. Rats were chosen because they have human-like serum complement activity [
17] and have been used in various models in which AQP4-IgG produces characteristic NMO pathology in CNS tissues [
18,
19]. We found that intravitreally delivered AQP4-IgG efficiently bound to AQP4 on retinal Müller cells and, unexpectedly, produced unique complement-independent retinal injury, which was further characterized by studies on ex vivo retinal cultures exposed to AQP4-IgG.
Methods
Rats
Adult male Sprague-Dawley rats (age 8–10 weeks) were purchased from Charles River Laboratories (Wilmington, MA). CD59−/− rats were generated using CRISPR-Cas9 technology, which will be reported separately. Protocols were approved by the University of California San Francisco Committee on Animal Research and were in compliance with the ARVO Statement for the Use of Animals in Ophthalmic and Visual Research.
Intravitreal injections
Rats were anesthetized using 3 % isoflurane and topical proparacaine (0.5 %, Akorn, Lake Forest, IL), and their pupils were dilated with phenylephrine (2.5 %, Paragon BioTeck Inc., Portland, OR) and atropine (1 %, Akorn). In each eye, a full-thickness track was made through superotemporal sclera at the pars plana with a 30-gauge needle for intravitreal delivery of a 4-μL solution volume using a 33-gauge beveled needle attached to a 10-μL Hamilton syringe (Reno, NV). Microscopic examination was performed to confirm the absence of vitreous hemorrhage or gross lens trauma. Lubricating ophthalmic ointment was applied to protect the cornea until recovery from anesthesia.
Both eyes in each rat received the same treatment to control for potential contralateral effects of a given treatment. The following experimental groups were evaluated: (i) no injection, or injection of (ii) saline; (iii) control human IgG (control-IgG; 40 μg; Pierce Biotechnology, Rockford, IL); (iv) purified human monoclonal recombinant AQP4-IgG (rAb-53, 40 μg, ref. [
20]); (v) recombinant monoclonal anti-AQP4 “aquaporumab”-lacking effector functions (AQP4-IgG
-CDC, 4 μg, ref. [
21]); or (vi) lipopolysaccharide (LPS) from
Escherichia coli (5 μg; Sigma-Aldrich, St. Louis, MO). In some experiments, rat complement was inactivated by intraperitoneal injection of cobra venom factor (CVF; 600 U/kg, Quidel Corporation, Santa Clara, CA) 24 h before and 48 h after intravitreal injection of AQP4-IgG, as described in [
19]. Rats were sacrificed 6 h, 24 h, 5 days, or 30 days after intravitreal injection. Globes were enucleated after transcardiac perfusion with phosphate-buffered saline (PBS) followed by 4 % paraformaldehyde, fixed for 4 h and left overnight at 4 °C in 30 % sucrose. The eyes were embedded in OCT and sectioned axially at 20-μm thickness.
Retinal explant cultures
Rats were deeply anesthetized with isoflurane and then decapitated. The freshly enucleated eyes were immersed in ice-cold Hank’s balanced salt solution (HBSS) containing 1 % penicillin-streptomycin. Using a dissecting microscope, a circumferential incision was made at the pars plana, followed by removal of the anterior segment, lens, and vitreous body. With Dumont forceps, the retinas were separated from the sclera and separated from the optic nerve head. Each retina was cut radially and separated into four quadrants, which were each transferred with inner retinal surfaces facing up onto 12-mm-diameter filters (0.4-μm pore; Sigma-Aldrich) in 12-well plates.
Retinal explants were maintained immersed in a thin layer of serum-free culture medium at an air/medium interface in a 5 % CO2 incubator at 37 °C. Culture media contained neuronal growth medium (Neurobasal A) supplemented with 2 % B27, 1 % N2, L-glutamine (0.8 mM), and 1 % penicillin-streptomycin. One half of the media was replaced after 24 h in culture. AQP4-IgG (final 20 μg/mL) was added to some wells after the initial 24 h in culture. Some explants were also incubated with dynasore hydrate (inhibitor of dynamin-dependent endocytosis; 50 μM) or chloroquine (inhibitor of lysosomal degradation; 10 μM). At 24 h later, explants were fixed in 4 % PFA for 24 h and then placed in 30 % sucrose for 24 h at 4 °C before embedding in OCT. Sections were cut at 10-μm thickness perpendicular to the full-thickness retina.
Immunofluorescence
Frozen sections were incubated in blocking solution (PBS, 1 % bovine serum albumin, 0.2 % Triton X-100) for 1 h prior to overnight incubation (4 °C) with primary antibodies against the following: AQP4 (1:200, Santa Cruz Biotechnology, Santa Cruz, CA), glial fibrillary acidic protein (GFAP, 1:100, Millipore), glutamine synthetase (GS, 1:500, Sigma-Aldrich), Brn3a (1:100, Santa Cruz Biotechnology), ionized calcium-binding adaptor molecule-1 (Iba1; 1:1000; Wako, Richmond, VA), C1q (1:50, Abcam, Cambridge, MA), or C5b-9 (1:50, Hycult Biotech, Uden, Netherlands), followed by appropriate species-specific Alexa Fluor-conjugated secondary antibody for 1 h at room temperature (1:200, Invitrogen, Carlsbad, CA). Rinsed sections were mounted with VECTASHIELD with 4’,6-diamidino-2-phenylindole (DAPI) (Vector Laboratories, Burlingame, CA). Staining with hematoxylin and eosin (H&E) was done using standard procedures. Sections were visualized on a Leica epifluorescence microscope (Wetzlar, Germany) or Nikon confocal fluorescence microscope (Melville, NY).
AQP4, GFAP, and Iba1 immunofluorescence were quantified in ×20 fields of central retina, 50 μm from the optic nerve head. AQP4 and GFAP fluorescence were defined using the polygon drawing tool and quantified using ImageJ (NIH, Bethesda, MD). For AQP4 quantification, retinal layers were segmented into RNFL + GCL, inner plexiform layer (IPL) + INL, and outer plexiform layer (OPL) + outer nuclear layer (ONL). GFAP was measured in two segments: (RNFL + GCL) and (IPL + INL + OPL + ONL). Data are presented as a percentage of area of immunofluorescence loss normalized to untreated retinas.
RGCs were counted at day 30 after intravitreal injection of AQP4-IgG or control-IgG as the density of Brn3a-positive nuclei in fluorescence micrographs of retinal flat mounts. After transcardiac perfusion, the eyes were enucleated and the retinas were removed and immunostained in culture wells on a shaker. Four radial relaxing incisions were made, and the retinas were flattened and coverslipped with VECTASHIELD mounting media. A total of 12 nonoverlapping images (×20 magnification), each including nonoverlapping posterior, middle, or anterior retina of one quadrant, were taken using the epifluorescence microscope. Brn3a-positive nuclei were counted semi-automatically using cell-counting plugins from ImageJ software.
Optical coherence tomography and fundoscopy
Rats were induced and maintained under isoflurane gas anesthesia, and the eyes were numbed and dilated as before. The corneas were lubricated with 2.5 % Goniovisc (Accutome Inc., Malvern, PA) for direct contact with the imaging lens. The eyes were examined using the Micron III retinal imaging system (Phoenix Research Labs, Pleasanton, CA), and raw fundus photographs were captured. Spectral-domain optical coherence tomography (SD-OCT) horizontal line scans were acquired on a micron image-guided SD-OCT system (Phoenix Research Labs) by averaging 10 scans. Potential inner retinal atrophy was evaluated in vivo as longitudinal changes in thickness of the ganglion cell complex (GCC, encompassing RNFL, GCL, and IPL). GCC and total retinal thicknesses at single points 1500 μm from the nasal and temporal optic nerve head margins were averaged as a single point per eye.
Statistics
Data are presented as mean ± S.E.M. The two-sided Student’s t test was used for direct comparisons between two means. When there were three or more groups, analysis was done using a one-way ANOVA with Dunnett post hoc test. Analysis was performed using Prism 5 GraphPad Software package (GraphPad Software, San Diego, CA). Significance levels were set at p < 0.05 (*) and p < 0.01 (**).
Discussion
We found that exposure of retinal Müller cells to AQP4-IgG can produce primary retinal pathology in the absence of NMO optic neuritis. For these studies, we administered a purified, recombinant AQP4-IgG to rat eyes by an intravitreal route. AQP4-IgG diffused through the posterior segment where it accessed the ciliary epithelium and retinal Müller cells, without significant diffusion to AQP4 on retrobulbar optic nerve astrocytes. We found that intravitreal injection of AQP4-IgG produced pathology with loss of Müller cell AQP4, a gliotic response with increased GFAP, microglial activation with minimal leukocyte infiltration, and mild RGC loss and thinning of the GCC. An interesting and unexpected observation was that the changes were complement independent, which contrasts with the generally accepted NMO pathogenesis mechanism in the optic nerve, spinal cord, and brain. These findings have potential implications for pathogenesis mechanisms and therapy of retinal abnormalities in NMO.
The reduced Müller cell AQP4 expression following exposure to AQP4-IgG in ex vivo retinal cultures supports an endocytic retrieval mechanism with lysosomal targeting to account for the loss of AQP4. The comparable reduction in Müller cell AQP4 following intravitreal administration of an engineered AQP4-IgG lacking complement and cellular effector functions supports the conclusion that AQP4-IgG itself causes AQP4 internalization, as do the retinal culture studies in which neither complement nor effector cells were added. AQP4-IgG-induced internalization of AQP4 is likely cell-type-, polarization-, and, perhaps, even species-specific. While AQP4-IgG produces rapid AQP4 internalization in several AQP4-transfected cell lines, and moderately fast AQP4 internalization in nonpolarized astrocyte primary cultures, little AQP4-IgG-dependent AQP4 internalization was seen in astrocytes in the mouse brain in vivo [
23,
24]. Though Müller cell AQP4 expression is polarized, the biology of Müller cells is different from that of astrocytes in the brain, spinal cord, and optic nerve. It is not clear from the studies here whether the loss of RGCs is a consequence of the loss of Müller cell AQP4 expression, the gliotic response, and/or intrinsic retinal inflammation.
The largely complement-independent retinal pathology following intravitreal AQP4-IgG administration contrasts with that seen in other CNS tissues. In humans, centrovascular deposition of activated complement is a characteristic pathological feature in NMO [
25], and an initial open-label clinical study of a complement inhibitor showed efficacy in reducing attacks of NMO optic neuritis and transverse myelitis in seropositive NMO patients [
26]. Pathology following passive transfer of AQP4-IgG to the mouse and rat brain, optic nerve, and spinal cord is predominantly complement dependent, as omission of added complement to mice [
27‐
29] or inactivation of endogenous complement in rats [
18,
19] prevents the development of NMO pathology. The complement-independent retinal injury found here suggests that complement-targeted therapeutics may have limited efficacy in some NMO patients with anterior visual pathway involvement.
Another interesting observation was increased GFAP expression in Müller cells following intravitreal AQP4-IgG administration, indicating a prominent gliotic response. Müller cell gliosis may be a consequence of intraretinal inflammation, Müller cell injury, and/or AQP4 loss. The retinal pathology seen here of AQP4 loss with increased GFAP and with minimal early complement deposition corresponds to the pathology seen in active human NMO lesions found in the brain and spinal cord classified as type 4 pathology [
30]. Similarly, two separate models of chronic intrathecal AQP4-IgG infusion have shown loss of AQP4 expression in the spinal cord with reactive gliosis in the absence of complement activation [
31,
32]. Thus, under conditions in which AQP4-IgG does not activate complement, the astrocytic response can be associated with GFAP upregulation rather than downregulation as seen when complement is activated. Release of soluble factors by activated microglia and Müller cells such as the proinflammatory cytokines TNF and monocyte chemoattractant protein MCP-1, which are thought to contribute to retinal degeneration in diabetic and other retinopathies [
33], may contribute to the retinal pathology. Chronic functional loss of AQP4 water permeability in Müller cells was found in AQP4 knockout mice to produce abnormalities in retinal signal transduction as seen by electroretinography [
34], which may reflect abnormal potassium and glutamate homeostasis. It is not clear whether the Müller cell gliosis and RGC loss produced by intravitreal AQP4-IgG is due to acute and partial loss of Müller cell water-transporting function.
Though the AQP4-expressing ciliary epithelium strongly bound AQP4-IgG following intravitreal AQP4-IgG administration, no pathology was seen at the light microscopic level. This is consistent with the near absence, albeit for a few case reports of NMO-related myositis [
35,
36], of NMO pathology in peripheral organs in which AQP4 is expressed, including the skeletal muscle, stomach, kidney, airways, and exocrine glands. High-dose systemic administration of AQP4-IgG to mice or rats produces prompt and extensive AQP4-IgG deposition on peripheral AQP4-expressing organs, but with no pathology [
18,
37]. Humans have been reported with circulating AQP4-IgG for at least a decade preceding clinical NMO disease [
38]. Why peripheral, AQP4-expressing tissues are not damaged in seropositive NMO, despite their exposure to AQP4-IgG, remains unclear. It has been speculated that the unique cellular and physical environment in the CNS may be responsible for the development of NMO pathology in CNS but not peripheral tissues, as might the presence of complement inhibitor or other anti-inflammatory mechanisms in peripheral tissues.
During the completion of our manuscript, a paper by Zeka et al. [
39] was published reporting that rats receiving AQP4-specific T cells developed retinitis with T cell infiltration and axonal pathology and that loss of AQP4 on Müller cells was seen when AQP4-IgG was delivered systemically along with the AQP4-specific T cells. Whether the retinal changes were primary or secondary to optic nerve pathology is unclear, as is its relevance to retinal abnormalities in human NMO.
Though our study demonstrates that exposure of the retina to AQP4-IgG can produce primary retinal injury with associated RGC loss, it is not known whether such a mechanism occurs in human NMO. Whether AQP4-IgG can access and bind to AQP4 on Müller cells beyond the blood-retinal barrier has not been established. There are no reports to our knowledge of retinal abnormalities in NMO without a history of optic neuritis, making it difficult to resolve primary vs. secondary AQP4-IgG-induced retinal injury in human NMO. Retinal pathology has been identified in postmortem tissue of patients with NMO, with mild loss of calbindin-positive horizontal cells and moderate loss of Müller cells and scattered loss of AQP4 immunoreactivity [
40]. There were also Iba1
+ microglia in the inner retina with few CD45
+ cells and little complement deposition. Our animal model recapitulates many of these features of retinal pathology in human NMO.
Acknowledgements
We thank Dr. Jeff Bennett (U. Colorado) for providing rAb-53 antibody, Dr. Xiaoming Yao (UCSF) for providing CD59−/− rats, and Hao Yiu and Dr. Marcel Alavi (UCSF) for their assistance in the acquisition and analysis of the OCT data.