Targeting the complement system in neuromyelitis optica spectrum disorder

Abstract

Introduction:

Neuromyelitis optica spectrum disorder (NMOSD) is characterized by central nervous system inflammation and demyelination. In AQP4-IgG seropositive NMOSD, circulating immunoglobulin G (IgG) autoantibodies against astrocyte water channel aquaporin-4 (AQP4) cause tissue injury. Compelling evidence supports a pathogenic role for complement activation following AQP4-IgG binding to AQP4. Clinical studies supported the approval of eculizumab, an inhibitor of C5 cleavage, in AQP4-IgG seropositive NMOSD.

Areas covered:

This review covers in vitro, animal models and human evidence for complement-dependent and complement-independent tissue injury in AQP4-IgG seropositive NMOSD. Complement targets are discussed, including complement proteins, regulators and anaphylatoxin receptors, and corresponding drug candidates.

Expert opinion:

Though preclinical data support a central pathogenic role of complement activation in AQP4-IgG seropositive NMOSD, they do not resolve the relative contributions of complement-dependent vs. complement-independent disease mechanisms such as antibody-dependent cellular cytotoxicity, T cell effector mechanisms, and direct AQP4-IgG-induced cellular injury. The best evidence that complement-dependent mechanisms predominate in AQP4-IgG seropositive NMOSD comes from eculizumab clinical data. Various drug candidates targeting distinct complement effector mechanisms may offer improved safety and efficacy. However, notwithstanding the demonstrated efficacy of complement inhibition in AQP4-IgG seropositive NMOSD, the ultimate niche for complement inhibition is not clear given multiple drug options with alternative mechanisms of action.

1. Introduction

Multiple sclerosis and neuromyelitis optica spectrum disorder (NMOSD) are among a group of autoimmune neurological diseases characterized by inflammation and demyelination in spinal cord, optic nerve and brain. These conditions can produce severe neurological deficits including motor impairment, loss of visual function, cognitive dysfunction and others. Clinical features of NMOSD can include recurrent attacks of transverse myelitis and optic neuritis, with radiographic features that can include longitudinally extensive myelitis and typical features of optic neuritis [1]. A subset of NMOSD, called ‘seropositive NMOSD’, is unique among autoimmune disorders in that it involves a well-defined humoral immune mechanism targeting a small membrane protein, aquaporin-4 (AQP4) [1–3]. AQP4 functions as a bidirectional water transporting protein expressed at the plasma membrane of astrocytes throughout the central nervous system (CNS), as well as in several organs outside of the CNS including skeletal muscle and various epithelial cell types [4]. In AQP4-IgG seropositive NMOSD there are circulating immunoglobulin G (IgG) autoantibodies against AQP4 [5], called AQP4-IgG, which consists of a polyclonal mixture of IgG1-class antibodies having complement-dependent cytotoxicity (CDC) and antibody-dependent cellular cytotoxicity (ADCC) effector functions.

Abundant evidence supports a pathogenesis mechanism for AQP4-IgG seropositive NMOSD in which AQP4-IgG binding to astrocyte AQP4 produces astrocyte injury by complement and cellular mechanisms, leading to an inflammatory response, blood-brain barrier disruption, and ultimately oligodendrocyte and neuronal injury [3,6–8]. Other mechanisms may be involved as well such as AQP4-sensitized T cells [9–11], direct cellular injury by AQP4-IgG [12,13], and bystander cytotoxicity [14–16]. Therapy for NMOSD has included immunosuppressants, B cell depletion, and plasma exchange [1,17,18]. Recently, three antibody drugs have received FDA approval in the USA for treatment of NMOSD, including the inhibitor of C5 cleavage eculizumab [19], the interleukin-6 receptor blocker satralizumab [20], and the CD19 B-cell lineage targeting drug inebilizumab [21].

There is compelling evidence for complement activation as a major mechanism in the pathogenesis of AQP4-IgG seropositive NMOSD. Pathology in CNS tissues shows vasculocentric deposition of activated complement [8,22], and limited biomarker data suggest general complement activation in the CNS [23,24]. In in vitro model systems, including astrocyte and spinal cord slice cultures, CDC is produced by exposure to AQP4-IgG and human complement [25–27]. In rodent models, inflammatory demyelination with characteristic features of AQP4-IgG seropositive NMOSD pathology is seen following passive transfer of AQP4-IgG by various routes, sometimes together with human complement (reviewed in ref. [28]). Transgenic mice and rats lacking complement regulator protein CD59 have greatly increased NMOSD pathology in AQP4-IgG passive transfer models [29,30]. Though there are caveats in the interpretation of the various in vitro and animal studies (reviewed in ref. [1]), the consensus of findings support a major role for complement. Perhaps the most compelling evidence comes from the pivotal clinical trial that supported the approval of eculizumab involving 143 AQP4-IgG seropositive NMOSD patients, with inclusion criteria including at least two relapses in the preceding 12 months or 3 relapses within 24 months [19]. Using a time-to-relapse endpoint, the annualized relapse rate was reduced by more than 90% with eculizumab compared to placebo.

The focus of this review is on complement-related targets and drugs for treatment of AQP4-IgG seropositive NMOSD. Evidence for complement-dependent and complement-independent tissue injury in AQP4-IgG seropositive NMOSD is discussed based on in vitro, animal models and human data. Various complement targets are discussed, including complement proteins and regulators, and anaphylatoxin receptors, as well as complement targeted drugs and drug candidates. Finally, we opine on the rationale and value of targeting complement in AQP4-IgG seropositive NMOSD.

2. Mechanisms of complement injury in AQP4-IgG seropositive NMOSD

2.1. AQP4, the target of AQP4-IgG autoantibodies in AQP4-IgG seropositive NMOSD

AQP4 is a plasma membrane water transporting protein expressed in astrocytes throughout the CNS, as well as in skeletal muscle and some epithelial cells in other organs such as kidney and stomach [31]. Phenotype studies in mice lacking AQP4 have indicated its role in brain water movement, neuroexcitation, astrocyte migration and neuroinflammation, which involve a variety of interesting mechanisms as reviewed elsewhere [32,33]. In astrocytes, AQP4 expression is polarized to foot-processes at the blood-brain barrier. Of relevance to AQP4-IgG seropositive NMOSD, AQP4 monomers, each ~30 kDa in molecular size, assemble in the plasma membrane as tetramers that further assemble as supramolecular crystal-like aggregates called orthogonal arrays of particles (OAPs) [34]. OAPs have a characteristic cobblestone appearance by freeze-fracture electron microscopy and now have been visualized by super-resolution microscopy [35]. Assembly of AQP4 in OAPs is required for complement activation, as AQP4-IgGs generally bind to AQP4 OAPs much better than to individual AQP4 tetramers [36]; also, binding of complement component C1q to the Fc portion of AQP4-IgG involves a multivalent interaction in which one C1q molecule binds to up to six closely associated AQP4 molecules on the cell plasma membrane [37].

2.2. Pathogenesis mechanisms in AQP4-IgG seropositive NMOSD

The pathogenesis of AQP4-IgG seropositive NMOSD is now fairly well-understood, as reviewed elsewhere [1,6,17]. Briefly, pathogenic AQP4-IgG autoantibodies are thought to be generated primarily in the periphery and, by so far incompletely understood mechanisms, access astrocytes in the CNS by passage across the blood-brain barrier. AQP4-IgG binds to extracellular epitopes on AQP4 at the astrocyte plasma membrane, which initiates astrocyte injury by CDC and ADCC mechanisms. In CDC, C1q binding to the Fc region of AQP4-IgG activates the classical complement pathway, producing pro-inflammatory anaphylatoxins and causing cellular injury by formation of the pore-like membrane attack complex (MAC), as discussed further in section 2.3. In ADCC, binding of leukocytes (neutrophils, macrophages, natural killer (NK) cells) to the Fc region of AQP4-IgG through Fcγ receptors causes their activation and degranulation. The inflammatory response created by astrocyte injury, cytokine release, leukocyte infiltration and microglial activation further disrupts the blood-brain barrier and ultimately injures oligodendrocytes and neurons. As discussed in sections 2.3 and 2.4, complement activation following AQP4-IgG binding to AQP4 on astrocytes may also cause cytotoxicity to nearby cells, including oligodendrocytes and neurons, by a bystander mechanism, and several complement-independent mechanisms, in addition to ADCC, may contribute to the pathogenesis of AQP4-IgG seropositive NMOSD.

2.3. Complement effector mechanisms in AQP4-IgG seropositive NMOSD

The complement system is part of the innate immune system that facilitates clearance of pathogens by antibodies and phagocytic cells. While complement normally has a beneficial role, its overactivation or misdirected activation can be deleterious. Fig. 1A diagrams the classical, alternative, and lectin complement pathways as relevant to AQP4-IgG seropositive NMOSD and its treatment by complement-targeted drugs. AQP4-IgG binding to AQP4 at the astrocyte plasma membrane initiates activation of the classical pathway by C1q binding to the Fc region of AQP4-IgG, resulting in assembly of the C3 convertase that cleaves C3 to C3a and C3b. Binding of C3b to factor Bb forms the C3 convertase for the alternative pathway, which amplifies the cascade. C3b also binds to the C3 convertase to form the C5 convertase that cleaves C5 to C5a and C5b. C5b complexes serially with C6, C7, C8 and C9 to form MAC (C5b-9), which acts as a permeability pore to cause cell injury [38,39].

Figure 1. Mechanisms of complement-induced cellular injury in AQP4-IgG seropositive NMOSD.

A. Schematic showing astrocyte target, leukocyte effector cell and bacterium, with major components of the classical complement pathway indicated. Complement activation is initiated by C1q binding to the Fc portion of AQP4-IgG when bound to AQP4 on astrocytes. Major effectors of injury and inflammation include the terminal membrane attack complex (MAC) and anaphylatoxins C3a and C5a. B. Schematic showing major regulators of the complement activation pathway. Complement regulators (white text) include soluble regulators (purple) and membrane-bound regulators (red). Blunt-end arrows indicate the targets of complement regulators. AQP4, aquaporin-4; AQP4-IgG, aquaporin-4-immunoglobulin G; C1-INH, C1-esterase inhibitor; C4BP, C4 binding protein; CFHR1, complement factor H related 1; FHL-1, factor-H-like protein 1; MAC, membrane attack complex: MASPs, mannose-binding lectin-associated serine proteases; MBL, maltose binding lectin.

In addition to MAC formation, complement activation produces fluid-phase anaphylatoxins C3a and C5a, which recruit and activate immune effector cells that express anaphylatoxin receptors C3aR and C5aR (CD88). C3aR and C5aR bind C3a and C5a, respectively, with binding affinity of approximately 1 nM. C3aR and C5aR are highly expressed on cells of myeloid lineage, including neutrophils, basophils, eosinophils, mast cells, macrophages and microglia. C3aR is also expressed on activated T and B cells [40]. In non-myeloid cells, C3aR and C5aR are expressed in neurons, reactive astrocytes, endothelial cells, and some epithelial and smooth muscle cells [41]. C3aR and C5aR signal transduction involves the pertussis-toxin-sensitive G-protein G_αi with downstream activation of intracellular calcium, PI3K, Akt and MAPK pathway, resulting in production of reactive oxygen species, pro-inflammatory mediators (IL-6, TNF-α, histamine) and adhesion molecules [40]. The activities of C3a and C5a in serum are controlled by carboxypeptidase N and B, which cleave their C-terminal arginines to reduce activity [42]. Involvement of anaphylatoxins in NMOSD is supported by correlations between plasma C3a levels and NMOSD disease activity as assessed by the expanded disability status scale [24], which has suggested the potential utility of C3a as a biomarker in NMOSD. A recent study suggested C3a as a key mediator of microglial activation and CNS pathology in early NMOSD using an experimental animal model of AQP4-IgG seropositive NMOSD produced by intrathecal AQP4-IgG administration [43].

Complement activation is modulated by regulator proteins expressed on cell plasma membranes as well as in the extracellular fluid phase [38,44] (Fig. 1B). Among the fluid-phase regulators, C1-esterase inhibitor (C1-INH) is a serine protease inhibitor that interferes with the initial steps of complement activation by reversibly binding to C1r and C1s in the classical pathway and MASP2 in the lectin pathway. C3 convertase is controlled by a set of regulator proteins, including complement factor I and its cofactors (C4BP, Factor H and FHL1). Vitronectin and clusterin bind to the terminal complement complexes (C5b-7, C5b-8 and C5b-9) to prevent their insertion into the cell plasma membrane.

The major membrane-bound complement regulator proteins in human include CD55, CD59, CD46 and CD35. CD55, CD46 and CD35 primarily target the C3 and C5 convertase enzymes [38] and CD59 targets the terminal MAC. CD55 (also called decay accelerating factor) prevents the assembly of the C3 convertase and enhances its degradation [45]. CD35 and CD46 are receptors for C3b and C4b that inhibit the assembly of the C3 convertase and accelerate its decay [46]. CD59 blocks MAC formation by inhibition of C9 incorporation to the C5b-8 complex [47].

The CNS is considered an immune privileged site because of limited access of peripheral complement proteins. Several complement proteins are locally synthesis by CNS cells including astrocytes, microglial and neurons [48]. Astrocytes and microglia synthesize and release fluid-phase regulators including C1-INH, factor I and factor H, and express the membrane-bound regulator proteins CD59, CD35 and CD55 [49]. Brain endothelial cells express CD46, CD55 and CD59. Oligodendrocytes are particularly susceptible to complement injury because they express relatively little CD55 and CD59 [49,50]. Neurons express MCP and CD59 at low levels and also show relatively high susceptibility to complement injury [51].

2.4. Complement bystander cell injury mechanism

As described in section 2.2, secondary injury to cells that do not express AQP4, such as oligodendrocytes, neurons and endothelial cells, may occur from the inflammatory environment produced by astrocyte injury. However, more direct mechanisms may be involved to explain the early and marked myelin loss and neuronal injury in AQP4-IgG seropositive NMOSD involving complement bystander injury to oligodendrocytes and neurons [14,15]. Fig. 2A shows CDC following exposure of an astrocyte-oligodendrocyte coculture to AQP4-IgG and human complement. Dead (red color) astrocytes are seen, as well as dead oligodendrocytes nearby dead astrocytes. Mechanistically, as diagrammed in Fig. 2B, the results support a mechanism involving local diffusion of the short-lived, soluble C5b67 complex produced by complement activation on astrocytes, resulting in MAC formation on nearby bystander cells. Bystander cell killing has important implications for complement-targeted therapeutics in AQP4-IgG seropositive NMOSD. There is also evidence for bystander injury by an ADCC mechanism in which leukocytes activated by AQP4-IgG binding to AQP4 on astrocytes injure nearby cells by targeted exocytosis of toxic granule contents [16].

Figure 2. Complement bystander injury.

A. Complement-mediated oligodendrocyte injury in oligodendrocyte-astrocyte cocultures. Cocultures incubated for 2 hours with 20 µg/ml AQP4-IgG (rAb-53) and 2% human complement, and then immunostained for GFAP (green) and MPB (blue), with dead cells shown with red fluorescence using fixable dead cell marker (left). Percentage dead oligodendrocytes at different distances from the center of dead astrocytes (mean ± S.E.M., n=6, * P < 0.01 comparing AQP4-IgG vs. control-IgG) (right). B. Schematic of complement bystander injury showing complement activation on astrocytes resulting in local diffusion of a short-lived, activated complement complex leading to MAC formation on nearby bystander cells. Reproduced with permission from [14]. AQP4, aquaporin-4; AQP4-IgG, aquaporin-4-immunoglobulin G; MAC, membrane attack complex; MBP, myelin binding protein.

3. Complement-independent pathogenesis mechanisms in AQP4-IgG seropositive NMOSD

3.1. Antibody-dependent cellular cytotoxicity

ADCC is likely an important complement-independent mechanism of cellular and tissue injury in AQP4-IgG seropositive NMOSD, which involves activation of effector leukocytes following interaction of their cell surface Fcγ receptors with AQP4-IgG bound to AQP4 on astrocytes. Activated granulocytes, macrophages and potentially NK cells release toxic granule contents and cause local cell killing. ADCC has been demonstrated in cell culture and spinal cord slice models of AQP4-IgG seropositive NMOSD [13,27]. In mice, which lack a functional classical complement pathway, intracerebral injection of AQP4-IgG and NK cells produced NMOSD-like pathology with loss of AQP4 and glial fibrillary acidic protein (GFAP), albeit minimal loss of myelin [52]. In a related study, intracerebral injection of an engineered mutated AQP4-IgG lacking ADCC effector function, together with human complement, produced much less pathology than the non-mutated AQP4-IgG and human complement [53], suggesting synergistic action of ADCC in complement-dependent NMOSD pathogenesis. In rats, which have an active complement system, intracerebral injection of AQP4-IgG produced characteristic NMOSD pathology around the injection site, surrounded by a penumbra with loss of AQP4 but not of GFAP or myelin [54]. AQP4 loss was also seen following intracerebral injection of AQP4-IgG injection in rats made complement-deficient with cobra venom toxin or when rats were administered AQP4-IgG lacking CDC effector function, but not seen with AQP4-IgG lacking both CDC and ADCC effector functions. Together, the animal data support the involvement of ADCC in the pathogenesis of AQP4-IgG AQP4-IgG seropositive NMOSD, but do not inform directly on the relative importance of ADCC vs. complement injury mechanisms in human NMOSD.

3.2. Direct cellular actions of AQP4-IgG autoantibody

Binding of AQP4-IgG autoantibodies to AQP4 on astrocytes has a variety of cellular actions independent of CDC and ADCC effector mechanisms, some of which may be involved in the pathogenesis of AQP4-IgG seropositive NMOSD. Studies in AQP4-expressing cell models have reported enhanced AQP4 endocytosis following AQP4-IgG binding [13,26], though animal studies did not support this action in vivo [55]. A more recent study in primary cultures of human astrocytes showed that NMOSD patient sera caused AQP4 clustering and degradation, with changes in astrocyte morphology and adherence [56], though these actions were only partially reversed by removal of IgG, making the results difficult to interpret. Other studies reported AQP4-IgG inhibition of AQP4 water permeability [57], though subsequent work refuted this finding [58]. There is evidence that AQP4-IgG, in the absence of complement, may induce cytokine release by astrocytes [43]. In a rat study of retinal NMOSD mechanisms, delivery of AQP4-IgG by intravitreal injection resulted in its deposition on retinal Müller cells [59]. By day 5 there was reduced retinal AQP4 expression, increased GFAP expression and mild inflammation, with loss of retinal ganglion cells at day 30. Notably, complement deposition was not seen, and similar retinal pathology was seen in cobra venom factor-treated rats and in normal rats receiving a mutated AQP4-IgG lacking complement effector function. AQP4-IgG thus produced primary, complement-independent retinal pathology, with mechanistic studies in ex vivo retinal cultures showing beneficial effects of inhibitors of endocytosis or lysosomal acidification. Changes in the human anterior visual pathway in AQP4-IgG seropositive NMOSD are consistent with these observations [60,61]. Together, the data support several potential complement-independent pathogenic effects of AQP4-IgG.

3.3. T cell effector mechanisms in AQP4-IgG seropositive NMOSD

Based on experiments in animal models, it is possible that T cell effector mechanisms may play a role in the pathogenesis of AQP4-IgG seropositive NMOSD, though so far there is no direct evidence to support their significance in humans. Early work showed that animals with experimental autoimmune encephalomyelitis (EAE), a condition caused by a T cell response against myelin, did worse when administered AQP4-IgG [62,63]. Several studies have utilized AQP4-specific T cells in an attempt to induce NMO pathology. In one study, AQP4-specific T cells were generated by immunization of Lewis rats with AQP4_207–232 peptide, and then isolated and transferred to naïve rats [64]. The recipient rats showed inflammation in brain and spinal cord but no astrocyte or myelin injury, which are the characteristic features of AQP4-IgG seropositive NMOSD. A subsequent study reported that highly encephalitogenic AQP4_268–285 -specific T cells from Lewis rats deeply infiltrated into brain parenchyma following intraperitoneal administration to rats [65]. Administration of AQP4-IgG following the T cells produced loss of AQP4 and GFAP in spinal cord and brain. In mouse studies, AQP4-reactive T cells were generated in AQP4 peptide-immunized AQP4 knockout mice, and Th17 polarized in vitro [11]. Intravenous transfer of these T cells to wild type mice produced tail and hind limb weakness, with demyelination and T cell infiltration in spinal cord, optic nerve and brain, though no loss of AQP4. A similar study compared Th1- vs. Th17-polarized AQP4-specific T cells from AQP4 null mice [10], showing greater clinical disease with Th17-polarized T cells in recipient wild type mice, with T cell infiltration in spinal cord and optic nerve. Together, these animal models demonstrate that AQP4-specific T cells can produce some pathological features of NMOSD, but do not inform on human NMOSD.

4. Complement-targeted therapeutics

4.1. C5 convertase inhibition

Eculizumab (Soliris®) is a humanized monoclonal antibody that prevents cleavage of C5 into C5a and C5b. Eculizumab was initially approved for treatment paroxysmal nocturnal hemoglobinuria and atypical hemolytic-uremic syndrome [66]. In the phase 3 clinical study (PREVENT study, prevention of relapses in NMOSD) AQP4-IgG seropositive NMOSD patients receiving eculizumab had a significantly lower relapse rate of 0.02 per year compared with 0.35 per year with placebo [19]. However, eculizumab is associated with an increase risk of meningococcal infection, with meningococcal vaccination required for patients prior to starting eculizumab [67]. In the PREVENT study all subjects received meningococcal vaccine and no meningococcal infection occurred. Based on the PREVENT study eculizumab was approved in the US, EU, Canada and Japan for AQP4-IgG seropositive NMOSD patients who are in a relapsing course of disease. Recognized limitations and caveats of eculizumab include its action late in the complement cascade, high cost, and non-ideal pharmacokinetics. Eculizumab treatment requires intravenous administration at a dose of 900 mg weekly for the first four doses, followed by 1200 mg every 2 weeks.

New drug candidates are in development that address some of these limitations (Fig. 3). Ravulizumab and crovalimab are modified forms of eculizumab in which the antigen-binding region has been altered to reduce binding affinity to C5 at reduced endosomal pH to facilitate dissociation of the antibody-antigen complex in endosomes and recycling of free antibody [68]. This modification enables less frequent or lower antibody dosing. Intravenously administration of ravulizumab in PNH has shown the similar efficacy to eculizumab, with less frequent dosing of every 8 weeks for ravulizumab vs. 2 weeks for eculizumab [69]. Crovalimab is administered by subcutaneous injection at 340 mg every two weeks compared to 1200 mg for eculizumab [70]. Currently, ravulizumab is approved for paroxysmal nocturnal hemoglobinuria (PNH) and is in phase 3 trials for NMOSD (CHAMPION trial, NCT04201262) and generalized myasthenia gravis (NCT03920293); crovalimab is in a phase 3 trial for PNH (NCT04432584). LGF316 is an another monoclonal IgG1 targeting C5 with a high affinity of 12 pM being studied in PNH (NCT02534909).

Figure 3. Complement drug targets and drug candidates.

Schematic of the complement pathways indicating sites of action of eculizumab, the complement inhibitor approved for AQP4-IgG seropositive NMOSD (purple), several approved drugs for non-NMOSD indications (orange), and experimental/pre-clinical drug candidates (green). See text for further explanation. MAC, membrane attack complex.

Other drug candidates targeting C5 are under development, including zimura, a PEGylated C5 aptamer that prevents C5 cleavage with good stability and pharmacokinetics [71]. Currently, zimura is in a phase 2/3 clinical study for age-related macular degeneration (NCT02686658). Nomacopan, a 17-kDa protein C5 inhibitor that binds to C5 and prevents its cleavage at different site from that of eculizumab, is suitable for small-volume subcutaneous injection and hence for self-administration [72]. Nomacopan is in a phase 2 study for PNH (NCT02591862). Because of its different binding site on C5, nomacopan is effective in PNH patients who carry a genetic variant of C5, missense mutation p.Arg885His, that interferes with eculizumab binding to C5 [73]. A limitation of nomacopan, however, is its short half-life requiring daily administration [74].

4.2. Therapeutics targeting other complement proteins

4.2.1. Antibody therapeutics targeting C1

An investigational complement-targeted biologic has been tested in in vitro and animal models of AQP4-IgG seropositive NMOSD – a neutralizing monoclonal antibody against C1q having 11 nM binding affinity [75]. The antibody blocked AQP4-IgG-induced CDC in cell culture models (Fig. 4A) and prevented astrocyte damage and demyelination in mouse spinal cord slice cultures incubated with AQP4-IgG and human complement. In mice administered AQP4-IgG and human complement by intracerebral injection, the antibody greatly reduced NMOSD pathology, including loss of astrocyte markers AQP4 and GFAP, and myelin loss (Fig. 4B). As discussed in section 2.3, targeting of C1q may be highly effective in NMO as it inhibits an early activation step in the classical complement pathway just after AQP4-IgG binding to AQP4, hence blocking the generation of anaphylatoxins and MAC without interference with the lectin complement activation pathway.

Figure 4. Neutralizing C1q monoclonal antibody C1q^mAb prevents AQP4-IgG and complement-induced injury in astrocyte cultures and in an experimental mouse model of AQP4-IgG seropositive NMOSD.

A. Complement-dependent cytotoxicity in primary cultures of murine astrocytes incubated with 10 µg/ml monoclonal AQP4-IgG rAb-53, 5% human complement and C1q^mAb (S.E., n = 4). Percentage cytotoxicity determined using Alamar blue assay. B. C1q^mAb prevents NMOSD pathology in mice following intracerebral injection of rAb-53 and complement. Brains were injected with 3 μl human complement and 1.3 µg C1q^mAb (left), 0.9 μg rAb-53 (middle), and 0.9 μg rAb-53 + 1.3 μg C1q^mAb (right). GFAP, AQP4 and MBP immunofluorescence shown at 3 days after injection. Yellow lines represent the needle tract, and white lines indicate areas of loss of AQP4, GFAP, and myelin. Adapted with permission from ref. [75]. AQP4, aquaporin-4; AQP4-IgG, AQP4-immunoglobulin G; C1q^mAb, monoclonal C1q antibody; GFAP, glial fibrillar acidic protein; MBP, myelin basic protein.

Sutimlimab, a humanized mouse monoclonal antibody targeting C1s, is under development [76]. A phase 1 study showed efficacy of sutimlimab in preventing hemolytic anemia in cold agglutinin disease with 6 weeks intravenous administration of 60 mg/kg weekly (NCT02502903). Sutimlimab is currently in a phase 3 study for primary cold agglutinin disease (NCT03347396). Recently, Sanofi developed an antibody targeting activated C1s, BIVV020, which is currently in a phase 1 study (NCT04269551).

4.2.2. C1-esterase inhibitor

C1-esterase inhibitor is a plasma protein with serine protease inhibition activity that acts on the kallikrein, coagulation and fibrinolytic systems, as well as on the complement pathway [77]. C1-esterase purified from human serum is approved for treatment of hereditary angioedema. Motivated by its inhibitory action on C1r and C1s proteases and hence on complement activation in the classical pathway, a phase 1b open-label non-controlled trial (NCT01759602) was done for acute NMOSD relapses, which demonstrated safety in ten patients administered 2000 units of C1-esterase inhibitor daily for three days [78]. Follow-on clinical testing has not been reported. In cell cultures, C1-esterase inhibition of AQP4-IgG-dependent CDC was demonstrated, though minimal effect was seen at concentrations achievable at doses used in humans [79]. In rats, intravenous C1-esterase inhibitor at a dose 30-fold greater than that approved for use in hereditary angioedema inhibited serum complement activity by <5 % and did not reduce pathology in a rat model of AQP4-IgG seropositive NMOSD produced by intracerebral injection of AQP4-IgG. It thus appears unlikely that C1-esterase inhibitor will be useful for therapy of NMOSD.

4.2.3. Fc multimers

Synthetic Fc multimers are another protein biologic with inhibitory actions on the complement pathway. Motivated by the multiple anti-inflammatory actions of the Fc region of human IgG, which is the basis for intravenous human immunoglobulin (IVIG) treatment of various autoimmune disorders, several Fc multimer therapeutics are under development that target the Fcγ and neonatal Fc receptors [80], some with demonstrated efficacy in animal models of arthritis, idiopathic thrombocytopenic purpura and inflammatory neuropathies. One such Fc multimer that has been tested in preclinical experiments in AQP4-IgG seropositive NMOSD is recombinant Fc hexamers, which consist of the IgM µ-tailpiece fused with the Fc region of human IgG1 and containing the multimer-stabilizing Fc mutation L309C [81]. Fc hexamers are reported to deplete early components of the classical complement system and bind with high avidity to Fcγ receptors [80]. Fc hexamers strongly inhibited AQP4-IgG-induced CDC with >500-fold greater potency than IVIG or monomeric Fc fragments, and blocked AQP4-IgG-induced ADCC as well. In rats administered AQP4-IgG by intracerebral injection, Fc hexamers full prevented the development of NMOSD pathology. If approved for other indications, Fc-based therapeutics are attractive candidates for repurposing in AQP4-IgG seropositive NMOSD based on their inhibitory effect on the complement system as well as other beneficial anti-inflammatory actions.

4.2.4. Therapeutics targeting other complement components

C3 is considered another potential target for complement modulator therapy because of its central role in complement activation. Compstatin is a cyclic tridecapeptide discovered using a phage-display approach that binds to C3 and prevents its cleavage by the C3 convertase. A compstatin derivative, POT-4, showed safety in a phase 1 clinical trial for age-related macular degeneration (NCT00473928), but failed a phase 2 trial (NCT01603043) because of adverse events. Because of the poor stability and solubility in compstatin and its derivatives in vivo [82], a synthetic peptide analog of compstatin that is conjugated to a ~40 kDa polyethylene glycol polymer, called APL-2, has better stability and solubility. In a phase 1 trial, daily subcutaneous injection of 180–360 mg APL-2 over 28 days significantly reduced C3 activity without serious adverse events [83]. Currently, APL-2 is being tested in PNH (NCT03500549). Another compstatin-related small molecule inhibitor, AMY-101, had a good pharmacological profile in a primate study [82], and is in a phase 2 clinical trial for gingivitis (NCT03694444) and acute respiratory distress syndrome due to coronavirus (NCT04395456).

C2/C4 is another potential therapeutic target in NMOSD as it is involved in early activation in both the classical and lectin pathways, while leaving the alternative pathway intact. The plasma concentration of C2 is lower than that of other complement proteins, making it potentially easier to neutralize [84]. ARGX-117, a humanized monoclonal antibody targeting C2 that prevents its interaction C4b and consequent assembly of the C3 convertase, inhibited IgG-induced hemolysis in an in vitro model of autoimmune hemolytic anemia with EC₅₀ ~30 μg/ml [85]. Targeting C6 offers an alternative strategy to prevent MAC assembly without affecting C5 activation and C5a generation. A humanized C6 monoclonal antibody has been developed recently that binds to both free C6 and the C5b6 complex. The antibody prevented C7 binding to C5b6 and inhibited MAC formation in vitro with K_d of 1–2 nM [86]. No clinical or pharmacological data have been reported for these newer biologics.

Therapeutics are also under development that target the alternative pathway. Lampalizumab is an antigen-binding fragment (Fab) of a humanized monoclonal antibody against factor D, which is involved in generation of the C3 convertase. In a phase 2 study lampalizumab showed benefit in age-related macular degeneration [87]. ACH-447, an orally bioavailable small molecule inhibitor of factor D that prevent the formation of C3 convertase [88], is currently in a phase 2 study for PNH (NCT03053102). A small molecule targeting factor B, LNP023, prevents the cleavage of C3b-bound factor B to generate C3 convertase. LNP023 prevented hemolysis of PNH erythrocytes in vitro [89] and is currently in a phase I trial for PNH (NCT03439839)

4.3. Therapeutics targeting anaphylatoxin C3a/C5a receptors

Targeting C3a/C5a is an attractive strategy to inhibit the inflammatory response in NMOSD without blocking MAC formation, hence reducing the risk of infection. Avacopan (CCX168) is an oral C5a inhibitor that binds to C5a with sub-nanomolar activity that prevents binding C5a to its receptor C5aR. Oral administration of 30 mg avacopan twice a day blocked C5a activity and showed benefit in anti-neutrophilic cytoplasmic autoantibody (ANCA)-associated vasculitis [90]; currently, avacopan is in a phase 3 trial for this disorder (ADVOCATE study, NCT02994927).

Vilobelimab (IFX-1) is a neutralizing monoclonal antibody targeting human C5a. In a phase 1 study vilobelimab inhibited C5a activity with good pharmacokinetic profile (NCT01319903) [91]. Currently, vilobelimab is in clinical trials for ANCA-associated vasculitis (NCT03712345), hidradenitis suppurativa [92], pyoderma gangraneosum [93] and severe COVID-19 (PANAMO study, NCT04333420).

4.4. Therapeutics targeting complement regulator proteins

As discussed in section 2.3, complement regulator proteins play an important role in modulating complement activity in the fluid phase and on target cells, and hence represent an alternative complement-related drug target in AQP4-IgG seropositive NMOSD. CD55 and CD59 are the major complement regulator proteins expressed on astrocytes [48]. CD55 inhibits the assembly and accelerates the decay of C3 and C5 converting enzymes, blocking the generation of anaphylatoxins C3a and C5a, as well as MAC formation, while CD59 selectively inhibits MAC formation. Several lines of evidence support the importance of astrocyte complement regulator proteins in AQP4-IgG seropositive NMOSD, including greatly increased AQP4-IgG/complement-dependent NMOSD pathology in CD59 knockout mice and rats [29,30], and with CD59 small interfering RNAs [56], and a mathematical modeling study [94]. Thus, increasing the expression or activity of complement regulator proteins on astrocytes or other nearby cells in the CNS, including oligodendrocytes, neurons and endothelial cells, is a logical extension of complement inhibitor therapy for AQP4-IgG seropositive NMOSD, with the potential advantage of avoiding the global immunosuppressive actions of complement inhibition. Upregulation of CD55 or CD59 in bystander cells may reduce their direct injury by the complement bystander mechanism.

Motivated by reports of drug action to increase CD59 expression on some endothelial cells [95], a cell-based assay was used to screen approved and investigational drugs, and nutraceuticals, to identify upregulators of endogenous CD55 and CD59 in a human astrocyte cell line [96]. Screening identified transcriptional upregulators of CD55 but not of CD59, most notably the commonly used statins atorvastatin, simvastatin and lovastatin. Atorvastatin increased CD55 expression by 3–4-fold at 24 hours in a human astrocyte cell line and in primary rodent astrocyte cultures (Fig. 5A), conferring significant protection against AQP4-IgG-induced CDC. Mechanistic studies revealed that CD55 upregulation by atorvastatin occurred at the transcriptional level and involved inhibition of the geranylgeranyl transferase pathway. Oral atorvastatin at 10–20 mg/kg/day for 3 days strongly increased CD55 expression in mouse brain and spinal cord, and reduced NMOSD pathology following intracerebral AQP4-IgG injection (Fig. 5B). These studies provided proof-of-concept for the potential therapeutic utility of upregulation of a complement regulator protein in AQP4-IgG seropositive NMOSD. Atorvastatin is a potentially attractive candidate because of its established safety profile and CNS penetration.

Figure 5. Atorvastatin increases CD55 expression in astrocyte culture and reduces injury in an experimental model of AQP4-IgG seropositive NMOSD.

A. Primary cultures of murine astrocyte were incubated with 0, 2 or 5 µM atorvastatin for 24 hours. CD55 immunofluorescence shown. B. Mouse model of AQP4-IgG seropositive NMOSD produced by passive transfer of AQP4-IgG. Mice received atorvastatin 20 mg/kg/day for 3 days prior to intracerebral injection of 7.5 µg AQP4-IgG and 1 µl human complement (or vehicle control), and sacrificed on day 6. AQP4, MBP and Iba-1 immunofluorescence in brain at day 6. Loss of AQP4 in area bounded by dashed curve. Arrows denote loss of MBP and increased Iba-1. Reproduced with permission from [96], under the following creative commons license: http://creativecommons.org/publicdomain/zero/1.0/. AQP4, aquaporin-4; Iba-1, ionized calcium-binding adaptor protein-1; MBP, myelin binding protein.

Viral gene delivery is an alternative approach to increase protein expression. An adeno-associated virus 2 (AAV2) that drives the expression of CD59, AAVCAGsCD59 [97], is currently in a phase 1 study for dry age-related macular degeneration (NCT03144999).

5. Conclusion

A substantial body of in vitro, animal models and clinical data provides a logical foundation to support complement inhibition therapy for AQP4-IgG seropositive NMOSD. Complement activation following the binding of pathogenic AQP4-IgG autoantibodies to AQP4 on astrocytes produces an inflammatory response leading to injury to oligodendrocytes, endothelial cells and neurons by mechanisms involving cytokine release, anaphylatoxin generation, astrocyte cytotoxicity and bystander cytotoxicity. The approved drug eculizumab interferes with astrocyte and bystander cytotoxicity by inhibiting MAC formation, and reduces leukocyte chemotaxis by inhibiting the generation of anaphylatoxin C5a. Many newer therapeutics that target different components of the complement pathway are in development for autoimmune and other disorders, as summarized in Fig. 3, and may offer advantages over eculizumab in terms of efficacy, safety and pharmacodynamics. In addition, there are alternative complement-related therapeutic strategies, such as upregulation of astrocyte complement regulator proteins.

6. Expert opinion

Compelling evidence supports the central involvement of complement activation in the pathogenesis of AQP4-IgG seropositive NMOSD. Though in vitro and animal models data do not resolve the relative importance of complement-dependent vs. complement-independent disease mechanisms, the pivotal phase 3 clinical study demonstrating efficacy of eculizumab [19] suggests that complement inhibition is largely sufficient for prevention of relapses in AQP4-IgG seropositive NMOSD. Post-marketing data for a larger treatment population should further inform on the longer-term efficacy and safety of eculizumab, including its use in combination with other NMO drugs with different mechanisms of action. With regard to AQP4-IgG seronegative NMOSD, there are some patients with myelin oligodendrocyte glycoprotein (MOG) antibodies that manifest unique clinical and pathological features quite different from AQP4-IgG seropositive NMOSD [98–100]. Complement-targeted therapy for MOG antibody-positive disease could potentially have benefit based on the finding of complement deposition in human pathological specimens [1]. The pathogenesis mechanisms of seronegative patients without AQP4-IgG or MOG antibodies remain unclear and there is no evidence to support the involvement of complement activation, though upon future investigations there may be subsets of these patients that might benefit from complement inhibition therapy.

Given the many available therapeutic options to treat NMO, including immunosuppressants, plasma exchange and B cell depletion, and the newly approved antibody therapeutics targeting IL-6 receptors and CD19, a consequential question is if and under what circumstances complement inhibition therapy is preferable to other treatment options. A secondary question, if complement-targeting therapy is used, is whether alternatives to eculizumab, with different complement targets and pharmacodynamic profiles, might offer improvements in safety and/or efficacy.

With regard to the first question, notwithstanding cost considerations with eculizumab (~500,000 US dollars annually), published clinical data [19] suggest efficacy and reasonable safety, though several caveats in the trial have been noted [64,101], including patient population size and inclusion criteria, issues with relapse adjudication, limited follow-up, and lack of quality-of-life assessment. While the NMOSD clinical trial reported no cases of meningococcal meningitis in the treatment group there was a significantly greater incidence of upper respiratory infections and one patient died of pulmonary empyema. Recent clinical trials have demonstrated efficacy and safety of satralizumab and inebilizumab [20,21], though it is difficult to make direct comparisons with eculizumab because of differences in trials design. Rituximab, a CD20-targeted monoclonal antibody used widely in NMOSD, also appears to have very good efficacy [102]. Thus, the selection of an appropriate therapy can be challenging with so many available treatment options with very different mechanisms of action. On the basis of in vitro and animal models data, the reported high clinical efficacy of C5 convertase inhibition by eculizumab may be somewhat surprising given the predicted contributions of complement-independent pathogenesis mechanisms, including ADCC, T cell effector mechanisms and direct AQP4-IgG cytotoxicity. Also, inhibition of C5 cleavage by eculizumab does not reduce formation of anaphylatoxin C3a and its downstream effects. Regarding safety, inhibition of C5 cleavage impairs bacterial killing via the complement lectin pathway, which is responsible for infectious adverse events, most notably meningococcal meningitis [103]. Though additional clinical data are needed, perhaps complement inhibition therapy may be most appropriate as adjunctive treatment for serious disease exacerbations that do not respond adequately to steroids and plasma exchange, or for prevention of disease relapses where there is a history of poor response to other drugs and cumulative neurological impairment.

Notwithstanding considerations of how complement-targeted therapy would best fit into the treatment paradigm for AQP4-IgG seropositive NMOSD, various alternatives to eculizumab warrant evaluation. Second-generation inhibitors of C5 cleavage such as ravulizumab and crovalimab offer improved pharmacokinetics. Biologics and small molecules targeting different components of the complement pathway may theoretically offer improved efficacy and safety compared to drugs targeting C5. Of interest are drugs targeting earlier components of the classical complement pathway that would block the generation of both anaphylatoxins C3a and C5a without inhibition of bacterial killing by the lectin pathway. It is noted, however, that chronic inhibition of early components of the classical complement pathway may produce a different set of potentially serious adverse events; for example, C1q knockout in mice is associated with lupus-like autoimmunity [104] by mechanisms that may involve impaired clearance of immune complexes and dying cells, and impaired B and T cell tolerization [105]. It is also noted that while the product of C3 cleavage, C3a, has a proinflammatory role in NMOSD, C3b is involved immune complex clearance and microbial opsonization. Though additional preclinical studies are needed, pharmacological upregulation of astrocyte complement regulator proteins CD55 or CD59 is an intriguing possibility as it may confer astrocyte cytoprotection without interfering with the general complement pathway. Combining drugs that inhibit complement activation and protect against astrocyte injury may be particularly effective. Finally, we note that drugs targeting the complement pathway are combinable with approved and investigational drugs having different targets and mechanisms of action, such as drugs that reduce AQP4-IgG or its interaction with AQP4, inhibit immune effector mechanisms, reduce neuronal injury, or promote remyelination.

Article highlights.

• Neuromyelitis optica spectrum disorder (NMOSD) is an inflammatory demyelinating disease of the central nervous system that can cause significant neurological deficits.

• Pathogenesis of AQP4-IgG seropositive NMOSD involves complement activation produced by binding of AQP4-IgG autoantibodies to AQP4 water channels on astrocytes, which leads to inflammation and demyelination.

• In vitro, animal models and human clinical data support a central role of complement in AQP4-IgG seropositive NMOSD and hence the therapeutic utility of complement inhibition.

• Complement-independent pathogenesis mechanisms may also have a pathogenic role in AQP4-IgG seropositive NMOSD, including antibody-dependent cellular cytotoxicity, AQP4-sensitized T cells, and direct AQP4-IgG-induced astrocyte injury.

• Eculizumab, a humanized monoclonal antibody that binds to C5 and prevents its cleavage, has been approved for treatment of AQP4-IgG seropositive NMOSD.

• Alternative complement-related drugs are in development that target different components of the complement system, complement regulator proteins, and receptors involved in complement effector functions.

Abbreviations:

AAV2

Adeno associated virus 2

ADCC

antibody-dependent cellular cytotoxicity

ANCA

antineutrophilic cytoplasmic autoantibody

AQP4

aquaporin-4

AQP4-IgG

AQP4-immunoglobulin G

C1-INH

C1-esterase inhibitor

C3aR

C3a receptor

C4BP

C4 binding protein

C5aR

C5a receptor

CDC

complement-dependent cytotoxicity

CFHR1

complement factor H related 1

CNS

central nervous system

EAE

experimental autoimmune encephalomyelitis

EndoS

endoglycosidase S

FHL-1

factor-H-like protein 1

GFAP

glial fibrillary acidic protein

Iba-1

ionized calcium-binding adaptor protein-1

IgG

immunoglobulin G

IVIG

intravenous human immunoglobulin G

MAC

membrane attack complex

MBL

maltose binding lectin

MBP

myelin basic protein

MOG

myelin oligodendrocyte glycoprotein

NK cell

natural killer cell

NMOSD

neuromyelitis optica spectrum disorder

OAP

orthogonal arrays of particles

PNH

paroxysmal nocturnal hemoglobinuria