Background
Neuromyelitis optica (NMO) is a severe autoimmune disease of the central nervous system (CNS) associated with predilection for the spinal cord and optic nerve. Histopathology and animal models demonstrate that complement-dependent cytotoxicity (CDC) plays a key role in the initiation of the NMO pathology [
1].
The complement system consists of over 30 components that are normally found inactive in serum and is an essential immune regulator of host defense to infection, cell integrity, and tissue homeostasis. The complement cascade can be activated by antigen-antibody complexes (classic pathway) or the antibody-independent lectin (lectin pathway). Additionally, the alternative pathway serves to amplify the classic and lectin pathways. All three pathways converge to trigger activation of terminal complement components resulting in cell lysis [
2]. In the CNS, the complement cascade regulates synaptic refinement and neuronal survival during development [
3,
4] and plays a role in maintaining brain homeostasis in adulthood [
5]. There is mounting evidence that complement synthesis and activation are increased in neurodegenerative and neuroinflammatory environments [
6].
NMO lesions show perivascular deposition of immunoglobulin and activated complement [
7,
8], and activated complement proteins are elevated in the serum and cerebrospinal fluid (CSF) of NMO patients [
9,
10]. Serum autoantibodies (NMO-IgG) against the water channel aquaporin-4 (AQP4) are found in most NMO patients [
11]. In the CSF of NMO-IgG positive patients, there exists a dynamic and enriched population of expanded plasmablast clones producing AQP4-specific antibodies [
12]. Both serum NMO-IgG and CSF-derived AQP4-specific monoclonal recombinant antibodies recapitulate NMO pathology when microinjected into the CNS, added to ex vivo spinal cord slices in the presence of human complement proteins [
13], introduced into rodent models of CNS inflammation [
12,
14], or injected into mice pretreated with Freund’s complete adjuvant [
15]. Astrocytes are selectively targeted in NMO, as evidenced by the extensive loss of immunoreactivity for the astrocytic proteins AQP4 and glial fibrillary acidic protein (GFAP) [
7]. Moreover, inhibition of the classic complement pathway blocks neuropathology in both in vitro and in vivo models of NMO [
16].
In developing experimental mouse models of NMO in vivo, ex vivo, or in vitro, co-administration of AQP4-IgG and normal human serum as the source of human complement (HC) is required to produce NMO pathological lesions [
17]. The weak activity of intrinsic mouse complement [
18] and the presence of complement inhibitor(s) in the mouse serum [
19] may combine to prevent rigorous complement activation in the absence of added HC in NMO models. Additional effects of HC on other CNS cells have not been characterized in NMO animal models.
In this study, we investigated the toxic effects of HC on murine neurons, astrocytes, differentiated oligodendrocytes (OLs), and oligodendrocyte progenitors (OPCs) in the presence or absence of AQP4-IgG in several in vitro and ex vivo CNS culture systems (monocultures, neuroglial mixed cultures, and organotypic cerebellar slice cultures). Our results reveal that neurons, OLs, and OPCs derived from mice display differential sensitivity to HC depending on the complexity of the neuroglial environment, and that AQP4-IgG-targeted astrocyte destruction results in secondary oligodendrocyte and neuronal loss.
Methods
Recombinant antibodies
The recombinant monoclonal human anti-AQP4 antibody #53 (rAb #53) was constructed from a clonally expanded plasmablast recovered from the cerebrospinal fluid (CSF) of a seropositive NMO patient. The control human antibody was generated from a CSF plasmablast sorted from a chronic meningitis patient. Recombinant antibodies were produced using a dual vector transient transfection system and purified with protein A-sepharose (Sigma-Aldrich, St. Louis, MO) as previously described [
12].
Biochemical assays and reagents
Lactate dehydrogenase (LDH) cytotoxicity assays were performed on cell culture supernatants using the LDH cytotoxicity detection kit (TakaRa Bio, Shiga, Japan). Except for IncuCyte live imaging (see below), propidium iodide (PI) (Sigma) was used at 5 μg/ml in the culture medium to label dead cells in cell and organotypic slice cultures. Normal, C1q-depleted, C3-depleted, C4-depleted, C5-depleted, and C9-depleted human serum (Complement Technology, Tyler, TX) were used as the source of human complement. Pooled and individual normal human (Sigma and Innovative Research), rat (Complement Technology), and mouse (Complement Technology and Innovative Research) sera were used as sources of complement for comparative complement experiments. Serum proteins were obtained by passing the serum through Amicon Ultra 3KD centrifugal filters (Millipore, Billerica, MA). Protein A/G plus agarose (Thermo Fisher Scientific, Waltham, MA) was incubated with human serum overnight at 4 °C, then centrifuged and discarded to deplete IgGs in the serum. Heat-inactivated serum was obtained by incubating normal serum at 58 °C for 30 min.
Primary cell cultures
Primary astrocyte, differentiated oligodendrocyte (OL), oligodendrocyte progenitor (OPC), and neuronal monocultures, and neuroglia mixed cultures were prepared as described from mice (CD-1 or PLP-eGFP [
20]) or rats (Sprague Dawley) (Harlan, Indianapolis, IN) with some modifications [
21‐
24]. Mixed glial cultures were prepared from P0-1 dissociated cortices and plated in poly-D-lysine-coated 24-well plates at a density of one brain per plate. Plates were used for experiments when astrocyte layer was confluent and oligodendrocyte cell bodies were apparent (day 5–8 in culture). Purified oligodendrocyte monocultures were prepared by shaking flasks of mixed glial cultures overnight at 200 rpm to detach the OPCs. OPCs were plated on poly-D-lysine/laminin coated dishes and maintained in media containing 10 ng/ml PDGF (platelet-derived growth factor)/FGF (fibroblast growth factor) (Peprotech, Rocky Hill, NJ) for 48–72 h prior to experiments. To differentiate OPCs into OLs, PDFG/FGF was withdrawn after 48 h and media supplemented with 40 ng/ml T3 (Sigma). OLs were used for experiments 24–72 h after exposure to T3. For neuroglial mixed cultures, P0-1 cortices were dissected, dissociated in 0.25% trypsin, and cells were plated on poly-D-lysine/Laminin coated coverslips (BD Biosciences, Franklin Lakes, NJ) in Neurobasal medium containing B27 supplement, 0.05 mM L-glutamine, penicillin-streptomycin (all from Life Technologies) and 5% heat-inactivated FBS (Sigma). Neuronal monocultures were prepared from E15-17 mouse cortex and cultured in Neurobasal medium in the presence of B27 supplement, 0.05 mM L-glutamine and penicillin-streptomycin. On day 2–5 in culture, 2 μM ara-C (Sigma) was added to inhibit the growth of glial cells. Human cortical neuron cultures (iCell Neurons) were cultured in iCell proprietary maintenance medium purchased from Cellular Dynamics (Madison, WI). Cell cultures were maintained at 37 °C in 5 or 8.5% CO2 atmospheric conditions.
Cerebellar slice culture
Sagittal slices (300 μm) were prepared from the cerebella of C57Bl6 or PLP-eGFP mice at P10 and cultured on MilliCell 0.4 μm membrane inserts (Millipore) for 10–14 days in media containing 25% inactivated horse serum [
25]. Prior to treatment, slices were switched to a serum-free Neurobasal medium supplemented with B27, 2 mM GlutaMAX, and 28 mM D-glucose (Sigma).
Immunostaining
Differential labeling of cell surface and internalized proteins in cell culture
Cell cultures were washed with fresh culture medium, then incubated with 20 μg/ml recombinant antibody at 4 or 37 °C for 1 h. Cells were then washed and fixed in 4% paraformaldehyde (Thermo Scientific) in phosphate-buffered saline (PBS) and rinsed and blocked with 3% normal goat serum (NGS) and 2% bovine serum albumin (BSA) in PBS without Triton X-100 (all from Sigma). Cells were then probed with Alexa-fluor 488 goat anti-human IgG secondary antibodies (1:500). Subsequently, cells were blocked overnight with unlabeled goat anti-human IgG at 0.25 mg/ml at room temperature and then washed, postfixed, and permeabilized with blocking buffer containing 0.1% Triton X-100. Alexa-fluor 594 goat anti-human IgG (1:500) was used to probe for intracellular recombinant antibody. After extensive washing, slides were mounted with ProLong Gold anti-fade reagent with DAPI (Life Technologies).
Immunostaining of fixed cells and brain sclices
Cell cultures or brain slices were fixed in 4% paraformaldehyde in PBS, rinsed, permeabilized in 0.1% (for cells) or 3% (for slices) Triton X-100 in PBS, and then blocked with 5% normal goat serum (NGS) or normal donkey serum (NDS) in PBS containing 0.1% (for cells) or 0.3% (for slices) Triton X-100 for 1 h. Samples were then incubated with primary antibodies in blocking buffer, washed, and then probed with secondary antibodies (Alexa Fluor secondary antibodies, all from Life Technologies). After washing, the coverslips were mounted with ProLong Gold anti-fade reagent (Life Technologies). The brain slices were mounted with Fluoromount-G (Southern Biotech, Birmingham, AL), placed under the glass covers, and sealed. Primary antibodies used in the immunostaining were mouse anti-NeuN (Millipore), mouse anti-β-tubulin (Sigma), rabbit anti-GFAP (DAKO, Carpinteria, CA), mouse anti-Olig1(Millipore), mouse anti-O4 IgM (R&D Systems, Minneapolis, MN ), rabbit anti-LAMP1(Abcam, Cambridge, UK), rabbit anti-Calbindin (Millipore), mouse anti-C3d (a gift from Dr. Joshua Thurman, University of Colorado), rabbit anti-MAC complex (Abcam), and rabbit anti-NG2 and rabbit anti-Olig2 (both are gifts from Dr. Charles Stiles, Harvard University).
Microscopy
Fluorescence images were acquired by Zeiss fluorescence microscope with Axiovision software (Zeiss, Jena, Germany). Confocal images were acquired by Leica SP5 laser scanning microscope (Leica Microsystems GmbH, Wetzlar, Germany) and Zeiss LSM 780 microscope or Olympus FV-1000 confocal microscope (Olympus, Center Valley, PA).
IncuCyte live cell imaging
Live cell imaging was performed using IncuCyte Zoom from Essen BioScience (Ann Arbor, MI). Cells were grown and scanned on 24-well plates in the cell culture incubator. Each well was scanned with a 10× objective lens in nine randomly selected positions at 15 min intervals with high definition phase contrast and epifluorescence microscopy using the following filter sets: 460/524 nm (green fluorescence, to detect enhanced green fluorescent protein, eGFP) and 585/635 nm (red fluorescence, to detect the DRAQ7 dead cell nuclei; Cell Signaling Technology, Danvers, MA). Image processing and cell counting were performed using IncuCyte software. Oligodendrocyte lineage cells in the neuroglial mixed cultures were labeled with eGFP in PLP-eGFP mice [
20]. In cell culture, OLs and OPCs were defined by their distinct morphologies. In the neuroglia mixed cultures, the majority of oligodendrocytes were bipolar and tripolar OPCs, while a minority of the cells displayed a highly complex network of processes and O4 antigen consistent with differentiating OLs [
26]. Neurons and astrocytes were identified by their distinct morphologies. DRAQ7+ cells were classified as dead neurons or astrocytes based on the different signal intensity and area. Dead oligodendrocytes were identified as DRAQ7+ cells with reduced eGFP signal. The identification of different cell types in mixed cultures was confirmed by immunostaining with NeuN (neurons), AQP4 and GFAP (astrocytes), Olig1 (a general oligodendrocyte lineage marker), and O4 (differentiating OLs) (Additional file
1: Figure S1). To assess the total number of the cells in the cultures, cells were stained with Cell Trace Calcein red-orange (Life Technologies) and then scanned and counted using IncuCyte software.
Quantification and statistical analysis
Cell counts in IncuCyte images were quantified by IncuCyte software. In each well of the 24-well plate, signals were measured in nine scanned positions and summated. PLP-eGFP positive cell bodies in the brain slices were imaged using a Zeiss fluorescence microscope with 20× objective. Images were quantified with ImageJ (National Institutes of Health open source). For each slice, 2–3 images were taken, quantified, and averaged. Slices from three independent experiments were analyzed. Statistical analyses were performed by unpaired Student’s t test for single comparisons or by two-way ANOVA for grouped comparisons using GraphPad Prism software. Data are expressed as means ± SD of independent experiments (n ≥ 3). Significance is reported for p < 0.05.
Discussion
These studies demonstrated that activated human complement HC is injurious to mouse neurons and OLs in a variety of culture conditions. We found that complement activation is necessary for cell injury in primary cell and slice cultures and occurs through the binding and activation of the classical pathway by anti-murine immunoglobulin in human sera. The susceptibility of neurons and glia to HC in the culture system lessens as the CNS cells coalesce into more complex networks. It may be that the different culture conditions (i.e., mono- or mixed cell cultures or tissues) alter the expression profiles of individual cell types resulting in loss of antigen expression, masking of target epitopes, and/or acquisition of complement inhibition. Each of these mechanisms may play a role in cell injury in models of NMO lesions and need to be examined relative to pathology in human tissue.
In primary cell monocultures, the addition of human serum alone is sufficient to cause rapid and substantial oligodendrocyte and neuronal cell death. In the neuroglial mixed cultures, the effects of activated complement are likely reduced through protective oligodendroglial and neuronal interactions with astrocytes. As a consequence, astrocyte injury and loss driven by AQP4-specific rAb and complement-mediated cytotoxicity rapidly increase the level of neuronal and oligodendroglial damage. Interestingly, OPCs in the neuroglial cultures are resistant to complement attack following astrocyte damage. In cerebellar tissue slices, cell-cell interactions are more complex, and complement-mediated cell death of neurons and OLs are further reduced. While astrocyte destruction similarly increases oligodendroglial and neuronal cell death, the susceptibility of these populations to complement cytotoxicity independent of astrocyte loss is much less than that observed in cell cultures. OPCs are the notable exception, becoming sensitized in the presence of astrocyte damage in organotypic slice culture (Table
1). Our results suggest that the increasing complexity of the neuroglia environment, particularly in slice cultures, protects neurons and oligodendrocytes from complement-induced cytotoxicity, and astrocytes play a central role in modulating complement injury. In the CNS of NMO patients, additional factors other than CDC, such as excitotoxicity, inflammatory cytokines, antibody-dependent cell-mediated cytotoxicity, anaphylaxotoxins, and disrupted astrocyte physiology may play critical roles in lesion formation (reviewed in [
28]). Whether exposure of tissue epitopes allows CDC to contribute to oligodendrocyte and neuronal injury in human NMO lesions remains to be determined. Conversely, the use of complement from multiple species will be necessary to validate the relevance of oligodendroglial and neuronal damage observed across in vitro, ex vivo, and in vivo NMO models.
Astrocytes are a major producer of complement in the healthy and diseased CNS [
29]. While astrocytes are not sensitive to HC alone, astrocyte destruction occurs readily through the production of terminal complement complexes (MAC complex) with AQP4-specific antibodies. Indeed, the current and previous studies demonstrate that inactivation of the classical or alternative pathway affects AQP4-IgG mediated astrocyte cytotoxicity [
16,
30]. Antibodies against factor B [
31], a key regulator of alternative pathway, or HC depleted of factor B, decrease astrocyte damage [
16].
Neurons in monoculture were very sensitive to activated complement, in agreement with the previous observations that neurons, unlike most peripheral cell types, express complement receptors and regulators, but do not express the most known complement inhibitors [
32,
33]. Neuronal damage in our in vitro and ex vivo models occurred as the direct result of activated complement. This may, in part, explain why persistent neurologic disability is strongly linked with acute relapses, as activated complement is prominent feature of acute NMO lesions [
1,
34,
35].
Consistent with previous reports [
36], our results demonstrated that oligodendrocytes were vulnerable to activated complement. Both in cell culture and brain slices, OLs (O4+) were susceptible to the addition of HC independent of the presence of astrocytes. Oligodendrocytes, unlike astrocytes, lack the complement regulatory protein CD59, which inhibits the final formation of membrane attack complex [
37‐
39]. Complement resistance in rat oligodendrocytes can be restored by the incorporation of purified rat CD59 into the cell membrane. Furthermore, neutralization of rat CD59 on complement-resistant astrocytes rendered them susceptible to CDC [
37]. When studied in mixed cultures or slice cultures with astrocytes and neurons, OL susceptibility to complement was partially decreased (Table
1). One explanation is that the presence of astrocytes or neurons may reduce the amount of the activated complement available for OL binding, mask OL membrane epitopes, or alter OL membrane epitope expression. Alternatively, astrocytes or neurons may secrete, exchange, or modulate complement regulatory protein expression on the oligodendrocyte surface. It has been reported that in a mixed culture system of rat oligodendrocytes and dorsal root ganglion neurons, OLs are remarkably less sensitive to complement. Incubation of OLs with neuron-conditioned medium afforded a similar protection against complement-mediated lysis [
40].
Our results demonstrate that OPCs remain resistant to activated complement independent of the astrocyte damage in mixed cultures. Unexpectedly, in organotypic cerebellar slice cultures, astrocyte destruction enhanced OPC loss. Although the neuroglia mixed cultures are prepared from the cortex and the organotypic slices from cerebellum, the differential sensitivity of OPCs to astrocyte damage does not appear to result from distinctions between cortical and cerebellar glia. In cell cultures prepared from cerebellum, there are no significant differences in astrocyte death or OPC survival in the presence of NMO rAb #53 and HC (unpublished data). Thus, it appears that astrocyte damage in the complex neuroglia network of cerebellar slices sensitizes OPCs to injury. In addition, activated microglia, inflammatory mediators, and excitotoxic molecules may combine to produce a toxic milieu for OPCs in brain slices. Alternatively, the state of differentiation of OPCs in that environment may result in novel susceptibility to neuroinflammatory injury.
Following astrocyte damage, loss of oligodendrocytes, both mature cells and progenitors, is observed earlier than neuronal death in cerebellar slices. This is consistent with reports that the loss of OLs and OPCs are early pathologic features after astrocyte depletion in human and experimental NMO lesions [
41]. Vulnerability of oligodendrocytes is found in many neurological conditions. In patients with relapsing-remitting multiple sclerosis (RRMS), oligodendrocytes in affected tissue show apoptotic features and myelin sheaths stain with activated complement [
42]. In a rat model of hypoxia-ischemic injury and in a mouse model of amyotrophic lateral sclerosis (ALS), extensive degeneration of differentiated oligodendrocytes (O4+) is found before the loss of OPCs [
43,
44]. Although new oligodendrocytes are formed from NG2+ OPCs in the spinal cord of ALS mice, they fail to mature [
44]. These observations in combination with our results suggest a selective vulnerability of mature oligodendrocytes to multiple forms of injury. With extreme conditions, such as massive astrocyte destruction in NMO, OPCs may also be affected.
Taken together, our findings demonstrate that in mouse monoculture, activated human complement is targeted to neurons, OLs and OPCs through the action of anti-murine immunoglobulins. In mixed cultures, surrounding astrocytes ameliorate CDC; targeted destruction of astrocytes with AQP4-specific recombinant monoclonal antibodies restores neuronal and oligodendroglia susceptibility to HC. Increasing complexity of neuroglial networks in murine cerebellar slices further reduces the effects of activated human complement alone and alters the secondary effects of astrocyte injury on OPC survival. Experimental models of NMO require complement activation to initiate CNS lesions. Initial models involved administration of AQP4-IgG to rats with pre-existing neuroinflammation produced by experimental autoimmune encephalomyelitis [
12,
14,
45]. Although some pathology resulted from the influx of myelin-reactive T cells, lesions demonstrated hallmark features of NMO pathology including perivascular astrocyte destruction and complement deposition. Subsequent passive-transfer models of NMO in mice required intracranial injection of NMO-IgG and human complement to recapitulate key pathological findings in human lesions [
30,
46,
47]. The use of human complement was required due to the presence of inhibitors in mouse complement preparations [
18,
19]. The current studies indicate that one should use caution in interpreting the cause of secondary injury to CNS cells following AQP4-IgG mediated CDC when using human serum in murine models.
Acknowledgements
We are grateful to the shared resource in Protein Production/Mab/Tissue Culture Core (supported by the Cancer Center Support Grant (CCSG) P30CA046934), University of Colorado Cancer Center for assistance with IncuCyte live imaging. And we are indebted to Hannah Schumann and Andre Navarro for the preparation of isotype control, NMO anti-AQP4 #53 rAbs, and anti-C3d antibody.