Protein therapeutics and their lessons: Expect the unexpected when inhibiting the multi-protein cascade of the complement system

Summary

Over a century after the discovery of the complement system, the first complement therapeutic was approved for the treatment of paroxysmal nocturnal hemoglobinuria (PNH). It was a long-acting monoclonal antibody (aka 5G1-1, 5G1.1, h5G1.1, and now known as eculizumab) that targets C5, specifically preventing the generation of C5a, a potent anaphylatoxin, and C5b, the first step in the eventual formation of membrane attack complex. The enormous clinical and financial success of eculizumab across four diseases (PNH, atypical hemolytic uremic syndrome (aHUS), myasthenia gravis (MG) and anti-aquaporin-4 (AQP4) antibody-positive neuromyelitis optica spectrum disorder (NMOSD)) has fueled a surge in complement therapeutics, especially targeting diseases with an underlying complement pathophysiology for which anti-C5 therapy is ineffective. Intensive research has also uncovered challenges that arise from C5 blockade. For example, PNH patients can still face extravascular hemolysis or pharmacodynamic breakthrough of complement suppression during complement-amplifying conditions. These “side” effects of a stoichiometric inhibitor like eculizumab were unexpected and are incompatible with some of our accepted knowledge of the complement cascade. And they are not unique to C5 inhibition. Indeed, “exceptions” to the rules of complement biology abound and have led to unprecedented and surprising insights. In this review, we will describe initial, present and future aspects of protein inhibitors of the complement cascade, highlighting unexpected findings that are redefining some of the mechanistic foundations upon which the complement cascade is organized.

1). The early beginnings of the Complement system

One of the oldest immunological principles is to label dangerous microbial intruders as foreign by decorating them with “signatures” called opsonins that mark them for efficient neutralization. Proteins with a reactive thioester moiety are ideal for this purpose and are an integral component in this strategy, which has been adopted by rather ancient organisms like insects but is also seen in higher mammals^1,2. A thioester domain (TED) enables covalent attachment to hydroxyl or amine groups on nearby surfaces, thus fixing the protein (and the opsonin’s effector functions) covalently to that surface, be it a microbe threating a host or even the host cells themselves³. Such indiscriminate deposition of a thioester-containing protein, well-described for the complement protein C3, has been referred to as “promiscuous binding to surfaces”⁴. The longer the half-life of the activated state of a thioester protein, the higher is the likelihood of promiscuous binding to nearby surfaces. In the case of the complement protein C3, activation and deposition of the TED-containing opsonins can even be a self-propagating process with the deposition of one C3b opsonin leading to an enzymatic cascade (or amplification loop) that deposits more C3b opsonins on the same surface, increasing opsonin density and promoting execution of the opsonin’s effector functions⁵.

During the late 19^th century, researchers discovered in the sera of higher mammals a heat labile fraction in blood plasma that contains an immunological effector function which kills bacteria⁶. From 1886 onwards Fodor, Nuttall, and Buchner all observed bactericidal activity in normal serum (reviewed in ⁷). Then Bordet showed that two components are required for this bactericidal activity in serum, one that is heat-labile and one that is heat-stable (i.e. antibodies). Ehrlich came to similar conclusions showing that erythrocyte lysis also requires heat-stable and heat-labile components in serum. He coined the name complement for the heat-labile component that complements the activity of the antibodies.

2). The definition of the complement pathways

2.1). Activation and amplification

When complement was first studied, the diverse methods of initiating a complement response were not recognized. Rather, the complement cascade was referred to as the classical pathway activation route and centered around the early and hence “classical” pathway components C1, C2, C4 and C3⁴. The delineation of discrete complement pathways principally came with the discovery of the alternative activation mode in which activity occurs in absence of the early classical pathway components C1, C2, C4. The first reports of an activation mode consuming complement component C3 but not C1, C2 and C4 goes back to the second decade of the 20^th century. Cobra venom factor (CVF) and zymosan were the components used in these studies (reviewed in ⁷). However, the concept of a truly alternative activation mode characterized by ‘direct C3 activation’ was only established much later⁸. Seminal findings for establishing the alternative pathway as stand-alone initiation pathway were the discoveries of properdin in 1954⁹ and more than a decade later, in 1968, the finding that bacterial lipopolysaccharides (LPS) consume C3-9 without consuming the early classical pathway (CP) components¹⁰. At this time, alternative pathway convertases had not been discovered, and the pivotal and central role of C3 in the complement cascade was not truly appreciated. Instead, C3 was considered one of the six terminal complement components including C3 to C9 (without C4)¹⁰. Later, it was established that components C6-C9, in addition to activated C5 can form membrane attack complex (MAC) pores on cells (in the complete absence of early components C1, C2, C4 and C3)¹¹. This finding proved that C3 was not a constituent of the terminal pathway (TP). The central role for C3 in the complement cascade was recognized when it became clear that the generation of C3b, by proteolytic activation of C3, is a self-propagating process in serum producing increasing amounts of C3b molecules via the amplification C3 convertase (today known as C3bBb which is stabilized by properdin)^5,12-14.

The key feature in initiating the alternative pathway (AP) and its amplification loop is the generation of initial C3b molecules by the early AP convertase C3(H2O)Bb¹⁵. Conformational, nonenzymatic auto-activation of C3, also known as C3 tick-over, starts with the hydrolysis of the internal thioester bond in C3, leading to a large conformational change that yields C3(H2O) triggering the assembly the early AP convertase C3(H2O)Bb^12,16-19. C3 tick-over is indiscriminate and thus can lead to initial C3b deposition on any surface, be it host or foreign. Therefore, discrimination between non-activating and activating surfaces by the AP occurs after the initial deposition of C3b by pre-existing regulators of defense⁵ (see below for more detail). The discovery of a third activation pathway, the lectin pathway (LP), which is closely related to the CP, emphasized the central position of C3 activation by convertases for the complement cascade²⁰: at the step of proteolytic C3 activation all three initiation pathways converge. Convertase-driven C3 activation within the AP fuels the C3b feedback cycle⁴, however, more generally speaking any convertase-driven C3 activation may be summarized with the term C3 activation cycle²¹. Naturally, where activation and enzymatic amplification occur, regulation is needed to keep a physiological balance and protect healthy host surfaces from collateral damage.

2.2). Regulation

The complement cascade is regulated from its initiation until its termination. Non-activating host surfaces are furnished with pre-existing regulators of defense (reviewed in ²²). These regulators are either surface-fixed membrane proteins or soluble plasma proteins and can be subdivided into non-convertase-directed and convertase-directed regulators (Table 1 and Table 2, reproduced from²²). The non-convertase directed regulators are either protease inhibitory proteins that impede the activity of activated complement enzymes within the CP and LP, or they are proteins that interfere with the successful assembly of membrane attack complex pores, blocking lytic pore formation. These non-convertase-directed inhibitors form stable (or even irreversible) stoichiometric complexes with their targets and thus only inhibit once. This means that stoichiometric inhibitors get consumed when they inhibit their target. In contrast, convertase-directed inhibitors act on the convertase enzymes as well as on the opsonins C3b and C4b, which act as platforms for the formation of new convertases.

Table 1.

Non-convertase directed regulators of complement activation (reproduced from reference 23)

Regulator	Function	regulated pathway	main regulatory compartment
C1 inhibitor (C1-INH)	Inactivates C1r and C1s, MASP-1 and MASP-2	CP/LP	fluid
sMAP	Binding to MBL, competition with MASPs	LP	fluid
MAP-1	Binds to MBL/ficolins,	LP	fluid
Vitronectin (S protein)	Binds to C5b-7/8/9	TP (MAC formation)	fluid
Clusterin	Binds to C5b-7/8/9	TP (MAC formation)	fluid
CD59	Binds to C8 and C9	TP (MAC formation)	surface

Table 2.

Convertase directed regulators of complement activation (reproduced from reference 23)

Regulator	regulatory activity decay / cofactor		regulated pathway	main regulatory compartment
CR1	DAA	CA	CP/LP & AP	surface
DAF	DAA	-	CP/LP & AP	surface
MCP	-	CA	CP/LP & AP	surface
C4BP	DAA	CA	CP/LP	fluid/surface
Factor H	DAA	CA	AP	fluid/surface
FHL-1	DAA	CA	AP	fluid/surface
Factor I	Protease for degradation of C3b or C4b in presence of a cofactor		CP/LP & AP	fluid (on surface only in conjunction with cofactor)

Convertases amplify any initial complement activation trigger and therefore require tight regulation to tailor the resultant C3b opsonin densities according to the biological response needed: e.g. no/little opsonization on healthy host cells, limited opsonization on apoptotic cells and high C3b opsonin densities allowing the transition to TP activation on microbial intruders^23-25. All but one convertase-directed regulator is encoded by the regulators of complement activation (RCA) gene cluster entailing complement receptor 1 (CR1, CD35), decay accelerating factor (DAF, CD55), membrane cofactor protein (MCP, CD46), C4 binding protein (C4BP) and Factor H (FH) (as well as its alternatively spliced relative, Factor H-like protein 1 (FHL-1))^22,26,27. The serum protease Factor I (FI) completes the set of convertase-directed regulators. Only FI is classified as an enzyme, however any of the convertase-directed inhibitors has an ‘enzyme-like’ mode of action as the RCA decay bimolecular convertase complexes and/or are cofactors that enable FI to cleave the convertase platform molecules C4b and C3b²⁸ (Figure 1A,B). Together the convertase-directed regulators execute the decay acceleration of convertases and the breakdown of C3b (and C4b), thus forming the C3 regulation cycle²¹ (Figure 1C). Both the C3 activation and the C3 regulation cycle operate in opposition, ideally arriving at a balanced complement response according to the level of complement effector functions needed for any given physiological situation.

Figure 1. Concepts of complement activation and regulation.

A) ‘Catalytic’ inhibitory mode of convertase-directed regulators. The decay accelerating activity of the RCA proteins is illustrated. The decay accelerator binds to the convertase complex and dissociates the convertase into its single components. Convertase activity is lost since the decayed convertase is not able to reassemble. However, C3b or C4b remain intact and could form new convertases if there is an ample supply of Factor B and Factor D (AP) or C2 and activated C1s (CP), respectively. The decay accelerator is recycled and can perform another cycle of convertase decay. B) Illustrates cofactor activity of RCA proteins. Once the cofactor is bound to C3b and/or C4b, Factor I is able to cleave within the CUB domains of C3b and C4b. Any of the C3b and C4b cleavage products cannot form new convertases. C) C3 activation and regulation cycles rotate in opposite directions and either promote or hinder the main three complement effector functions (opsonization, anaphylatoxin release, and membrane attack complex formation).

Panel C is reproduced with permission from Schmidt et al²¹.

3). ‘Prominent’ complement-mediated diseases, pathophysiology & approved complement inhibitors

Too much or too little complement activation can be associated with pathology. Complement-mediated diseases may be sub-grouped into those that are intrinsically caused by too much or too little complement activation, and those in which complement plays ‘only’ a non-causative and secondary role. A secondary role may, nonetheless, contribute substantially to the pathophysiology. Examples for the latter case include autoantibody-mediated diseases or any pathology associated with ischemia-reperfusion injury. After reperfusion, strong initial CP and LP activation occurs on the altered cell surfaces affected during ischemia, with the AP amplification loop intensifying the initial activating trigger. Mutations or polymorphisms that affect the expression and/or the function of complement proteins substantially causing illness may be classified as intrinsic complement diseases. The hematologic condition of paroxysmal nocturnal hemoglobinuria (PNH) is one such example.

3.1). The model disease paroxysmal nocturnal hemoglobinuria (PNH)

In 1882, four years before the first reports on bactericidal (complement) activity in serum were published, Strübing described a patient suffering from the AP-mediated disease paroxysmal nocturnal hemoglobinuria (PNH)²⁹. In 1911, van den Berg demonstrated that hemolysis of PNH erythrocytes requires a heat-labile fraction in serum and thus was considering the complement system as source for the hemolytic reaction. However, at the time, two characteristics had to be fulfilled to prove complement activity: i) the activity should be lost upon heat-inactivation and ii) the activity could be restored by adding a small proportion of fresh serum (reviewed in ³⁰). These two conditions are still relevant today for CP activation, which typically can be performed in highly diluted serum. In contrast, AP activity relies on the amplification loop, which does not function at high serum dilutions (a point which we will come back to later). In van den Berg's experiments, the amount of fresh serum added to the heat-inactivated serum was too small to restore AP activation and so he erroneously ruled out a role for complement in the pathophysiology of PNH. In contrast, although the AP was not discovered in 1939, Ham and Dingle cautiously concluded that complement or a complement-like substance was involved in the pathophysiology of PNH³¹.

The pathophysiological mechanisms underlying PNH were intriguing and it was ultimately discovered that the susceptibility of the PNH erythrocytes for AP-mediated lysis is due to the absence of complement regulatory proteins on PNH erythrocytes^32,33. This absence was consequently attributed to the deficiency of GPI-anchors on the affected erythrocytes³⁴. GPI-deficiency (or a reduced number GPI anchors depending on the activity phenotype of the mutated enzyme phosphatidylinositol N-acetylglucosaminyltransferase subunit A; pig-a) was found to be caused by somatic mutations in the PIG-A gene that reduces GPI-linked proteins at cell surface. It is lack of CD55 that results in reduced regulation of convertases, and lack of CD59 that renders the cells susceptible to MAC lysis³⁵. Erythrocytes (also healthy ones) are generally more susceptible to complement attack because of their reduced number of complement regulator (naturally lacking the regulator CD46) and their inability to actively remove MAC structures(reviewed in ^36,37).

PNH erythrocytes still have two remaining complement regulators that protect them from the non-directed activation mode of the AP, the soluble plasma protein FH and the integral transmembrane protein CR1, which occurs at numbers of only about 100-1000 molecules per erythrocyte according to the ‘quantitative’ CR1 polymorphism (reviewed in ³⁸). FH can associate with polyanionic host markers like sialic acids on host cell surfaces, which is important in the protection of PNH erythrocytes³⁹. As a consequence, PNH erythrocytes do not lyse because of low-level tick-over activation of C3 due to reduced albeit sufficient protection by the regulators FH and CR1. However, when a PNH patient experiences complement amplifying conditions, e.g. infections, pregnancy or surgery, the enhanced complement activation meets the under-protected PNH erythrocytes leading to large scale lysis of PNH erythrocytes (reviewed in⁴⁰). The requirement for stronger complement activation explains why hemolysis occurs paroxysmally in PNH: complement amplifying conditions like infections or surgery also happen paroxysmally. Once hemolysis occurs, it induces, among other symptoms, anemia and smooth muscle dystonia and thrombosis. The latter is the major cause for morbidity and mortality in PNH patients. In a recent review, Schmidt et al. delineate why complement activation leading to cytolysis in PNH (and aHUS; see also below) associates with thrombosis while widespread complement activation in the kidney disease C3 glomerulopathy (C3G) does not²¹. Interconnections between the systems of hemostasis and complement have been reported several times based on in vitro studies and animal models (reviewed in ²¹), but the most compelling evidence originates from clinical evidence. Historically, PNH patients regularly succumbed due to ongoing thrombotic complications despite vigorous anticoagulation^41,42. While anticoagulation failed to prevent thrombosis in PNH, complement inhibition did. Therapy with the first-in-class complement inhibitor eculizumab (approved in 2007), which inhibits the terminal and cytolytic complement pathway, proved to almost eradicate the occurrence of thrombotic events in PNH^41,43-45. Once more, research into PNH pushed the boundaries of our understanding of the complement system, this time pioneering the successful clinical use of complement intervention and definitively establishing PNH as a complement model disease. Ravulizumab, a second generation version of eculizumab with improved pharmacokinetic properties has been approved recently and a whole pipeline of other C5 inhibitors has entered clinical development emphasizing the impact eculizumab has had in this clinical area^46-48 (see section 4 and 5 below).

3.2). aHUS – another intrinsic complement disease

Hemolytic-uremic syndrome (HUS) was first described by Gasser in 1955⁴⁹. The disease is characterized by the triad of microangiopathic hemolytic anemia, thrombocytopenia and acute renal failure, and is the most common cause for renal failure in children. It is typically associated with diarrhea secondary to Shiga-toxin producing Escherichia coli (STX-HUS)⁵⁰. Although there may be a role for complement in the pathophysiology in STX-HUS, the context is not well-established and in any case STX-HUS is not initiated by aberrant complement activation. In contrast, the atypical from of HUS (aHUS) is mediated by AP complement activation, although the underlying genetic variants in complement genes predispose to aHUS rather than cause the disease (i.e. penetrance is highly variable). aHUS is characterized by the same clinical triad as is STX-HUS, and somewhat similar to PNH, an environmental trigger like an infection or surgery (i.e. a complement amplifying condition; see PNH section above) is needed to offset the weakened balance between C3 activation and regulation. Complement activation in aHUS is concentrated on the endothelial cell lining of glomeruli (summarized in ²¹). The glomerulus may be especially prone to complement activation/amplification because of the relative lack of complement regulators at the glycocalyx and the glomerular basement membrane (GBM), which is not a phospholipid bilayer but composed of extracellular matrix. These structures lack membrane-fixed complement regulators and instead rely on regulation by the soluble regulator FH.

Because complement activation/amplification occurs locally in the glomeruli, systemic complement levels are affected only in a smaller proportion of patients. For example, only 30-50% of aHUS patients have reduced C3 concentration^51,52, and FH levels are normal in about 90% of aHUS patients with identified FH mutations⁵³. However, the first hint that complement was involved in the pathophysiology of aHUS came from aHUS patients with low C3 or FH concentrations^54-56. Later, elegant studies established that the concentrations of FH (and also other complement components) are not always reduced^57,58, but mutations alter the function of the affected proteins. Therefore, determining the concentrations of complement components is not generally required for the diagnosis of aHUS (but it may be helpful in some individual cases).

aHUS is therefore an archetypical example of an imbalance between C3 activation and C3 regulation (Figure 1C), with disease reflecting uncontrolled TP effector functions. Gain-of-function mutations (heterozygous) in C3 or Factor B (FB) set the ‘C3 activation cycle’ in overdrive, while loss-of-function mutations (heterozygous) in the regulators FH, FI and/or CD46 weaken the ‘C3 regulation cycle’, with the end consequence in both cases being uncontrolled TP activity. Autoantibodies against the C-terminal domains of FH, or fusion proteins of FH and Factor H-related proteins (FHRs) can also impact C3 regulation and predispose to aHUS. The degree of functional alteration associated with a certain mutation in a complement gene and the strength of the environmental trigger act synergistically to increase the risk for developing disease (elegantly summarized here ⁵⁹).

Historically, plasma exchange or infusion was found to be beneficial in some aHUS patients but the success of this type of treatment was eventually demonstrated to depend on the underlying mutated complement genes. In addition, treatment can result in unintended consequences as for example in giving a patient with a truncating CFH mutation fresh frozen plasma, which not only replenishes nonfunctional FH with functional FH, but also provides C3 in the setting of partially consumed C3. The end result may be no net gain.

In 2011, eculizumab, the anti-C5 antibody already approved for PNH therapy, was approved for the treatment of aHUS⁶⁰. The pivotal study provided strong clinical evidence that activation of C5is the driver of the TMA typical for aHUS. Consistent with this cause-and-effect relationship, patients with non-complement-mediated aHUS such as that characterized by recessive mutations in diacylglycerol kinase epsilon (DGKE; which is not part of the complement cascade) are non-responsive to eculizumab^61,62. In aggregate, both PNH and aHUS are primarily mediated by excessive AP activity classifying both of them as typical complement-mediated diseases.

3.3). Complement activating autoantibodies of the IgG type are the primary cause for myasthenia gravis (MG) and neuromyelitis optica spectrum disorder (NMOSD) which secondarily trigger major complement-mediated damage

In contrast to PNH and aHUS, the diseases MG and NMOSD are primarily caused by autoantibodies. MG patients exhibit IgG autoantibodies directed against the acetylcholine receptors at the neuromuscular junction. In about 85% of these patients, the autoantibodies are of the subtypes IgG₁ and IgG₃, which strongly activate the CP^63,64. About 75% of NMOSD patients exhibit complement-fixing autoantibodies directed against aquaporin-4 on astrocytes⁶⁵. The primary therapeutic strategy is to suppress autoantibody production in these patients using immunosuppressant drugs, although this strategy does not always work sufficiently.

Autoantibodies can drive disease through different mechanisms of action. For example, IgG autoantibodies that bind to their target receptors may interfere with ligand-receptor signaling by blocking ligand access or inducing surface aggregation of receptors. However, the complement fixing autoantibodies in MG and NMOSD are known to lead to C5 activation and MAC assembly. In NMOSD, MAC deposition damages or destroys astrocytes eventually leading to secondary demyelination and loss of neurons. Pro-inflammatory effects triggered by C5a also lead to the activation of eosinophils and neutrophils, which migrate into the cerebrospinal fluid to promote further inflammation⁶⁶. In MG, MAC formation leads to destruction of the plasma membrane of the neuromuscular synapse causing disruption of signal transmission and a decrease in functioning acetylcholine receptors, eventually cumulating in structural changes within the neuromuscular junction (reviewed in ⁶⁷).

Although MG and NMOSD are not intrinsically complement mediated, in a proportion of patients it is the TP that drives the pathophysiology due to complement-fixing autoantibodies. These patients are clearly expected to benefit from anti-complement therapy. Also in these two neurological diseases, the therapeutic antibody eculizumab pioneered the field and proved clinically that the complement TP is a major driver of pathology in MG and NMOSD. Eculizumab was found to be efficacious in patients suffering from anti-acetylcholine receptor antibody positive generalized MG and anti-aquaporin-4 antibody-positive NMOSD^68-71.

3.4). Cold-agglutinins disease is caused by IgM autoantibodies followed by complement activation

Cold agglutinins disease (CAD) is a rare autoimmune hemolytic anemia. The cold agglutinins are autoantibodies that recognize the erythrocyte surface of the host. Once bound to their target, IgM autoantibodies strongly activate the CP leading to massive C3b opsonization and MAC-mediated hemolysis that ultimately results in severe anemia and thrombotic complication (reviewed in ^21,72). Complement involvement in CAD has long been recognized⁷³. There are two aspects of complement-mediated loss of erythrocytes. The first is intravascular lysis by TP-mediated MAC formation on erythrocytes. The second occurs when CP activation does not result in MAC formation but still leads to deposition of C3b opsonins on erythrocytes, as is the case when minor disruptions between the C3 activation and regulation occur. In this scenario, the C3b-density may not be enough to trigger MAC-mediated hemolysis, but will nevertheless tag erythrocytes for phagocytosis by the reticuloendothelial system (RES).

The impact of complement inhibitors has been investigated in CAD patients in several studies. In a phase two clinical trial, eculizumab was found to reduce the median lactate dehydrogenase level from 572 U/L to 334 U/L, in line with a significant but incomplete reduction in hemolysis (and ultimately transfusion requirement) of the enrolled CAD patients⁷⁴. Another inhibitory approach blocked the complement cascade early using a monoclonal antibody that targets C1s. The approach was effective and the FDA has now approved this therapy for the treatment of CAD^75,76. C1s is a zymogen within the C1 complex, which comprises the soluble pattern recognition molecule of C1q, and two molecules of the zymogens C1r and C1s. C1s becomes proteolytically activated by C1r after a conformational change in C1q. The conformational change of C1q takes place upon recognizing the Fc-part of surface deposited IgG (all subtypes to various extents except from subtype IgG₄) or IgM antibodies⁷⁷. Activated C1 then cleaves and activates C4 and C2.

Given the low serum concentration of C1s (~34 μg/ml; ~0.4 μM), it is remarkable that the dose of the anti-C1s antibody sutimlimab is rather high with 6.5 - 7.5 g per patient every other week⁷⁶. Technically speaking, however, a protein inhibitor of activated C1s (and C1r of the CP and the analogous MASP-1, MASP-2 of the LP²²), C1 inhibitor, has been available for clinical use (in Europe) for decades as plasma-purified therapeutic protein (reviewed in ⁷⁸). The physiological regulator C1 esterase inhibitor (C1-INH) is a ‘serpin’, which inhibits proteases of the complement, the coagulation and the kinin system. Thus far, the clinical need for plasma purified or recombinant (which has become available) C1-INH is not in a complement disease but rather in hereditary or acquired angioedema, which is characterized by uncontrolled generation of bradykinin. However, clinical trials have been or are underway to investigate if C1-INH may be efficacious in complement diseases with a substantial CP component, such as antibody-mediated rejection after kidney transplantation or overshooting inflammation after severe traumatic brain injury (e.g. see ^79,80). Future studies will show whether C1-INH, which inhibits more than one physiological pathway, warrants in depth investigation as a complement inhibitor in diseases with a substantial CP involvement now that a specific monoclonal antibody directed against C1s has been approved for clinical use.

4). Specific aspects about some protein inhibitors targeting complement in certain diseases

Although characterized by different etiologies, it is uncontrolled TP activation that drives the major pathophysiological events in PNH, aHUS, MG and NMOSD, as confirmed by clinical evidence demonstrating efficacy for anti-C5 therapy in these four diseases (see also above). However, there are other diseases with a clear and substantial complement component in their pathophysiology (regardless of being intrinsically or secondarily mediated by complement activation) that have not responded convincingly to complement inhibition. For these diseases, inhibiting complement earlier in the disease progression, for a longer period within a pathophysiological time course, or at a different step within the complement system may provide better clinical results. With more than 40 effector and regulatory proteins, the complement cascade holds multiple opportunities to inhibit specific steps that may be more suitable in certain pathological settings than targeting only the terminal pathway. The route of administration and especially the mode of action (local vs systemic) will likely impact the efficacy of each complement inhibitor.

4.1. Age-related macular degeneration and complement inhibition

One disease in which complement intervention has not yet produced a substantial clinical benefit sufficient enough for therapeutic approval by regulatory bodies is age-related macular degeneration (AMD). Next to lifestyle factors, the risk for developing AMD is impacted by genetic factors such as sequence variations in ARMS2/HTRA1 and several complement genes, including FH, FI, FB, C3, C9 and the complement factor H-related proteins (FHR) (reviewed in ^81-83). In aggregate, about 70% of the risk of developing advanced AMD can be attributed to genetic factors and about 60% of these genetic risk variants are associated with genes of the AP, emphasizing the role complement has in the pathogenesis of AMD⁸².

Given the success of the therapeutic anti-C5 antibody eculizumab in PNH, a trial using eculizumab was designed for AMD. Launched in 2009, the COMPLETE study (Complement inhibition with eculizumab for the treatment of nonexudative AMD; NCT00935883) evaluated the effect of systemically administered eculizumab on the growth of geographic atrophy (GA) in patients with AMD⁸⁴. Thirty patients were enrolled. Drug substance or saline (placebo) was administered intravenously for six months. As expected, systemic administration of eculizumab was well tolerated, however it failed to significantly decrease the growth rate of GA. The power of this study is rather limited and generalized conclusions are not possible. For this reason, additional trials will be required to address whether i) systemic or local administration (or both), or ii) proximal or terminal complement inhibition is needed to halt progression of GA. Or whether GA, once that it has occurred, even responds to complement inhibition. The phase III CHROMA (NCT02247479) and SPECTRI (NCT02247531) trials have addressed whether intravitreal application and more proximal complement inhibition will bring a benefit. These trials were initiated after the phase II MAHALO trial (NCT01229215) showed that selective complement Factor D (FD) inhibition with lampalizumab (a Fab fragment of a humanized anti-human FD monoclonal antibody blocking the FD exosite⁸⁵) reduced the rate of GA enlargement (with 123 patients being enrolled in total)⁸⁶. From 2014 onwards, nearly 1900 patients enrolled in CHROMA and SPECTRI have been treated for almost two years with 10mg of lampalizumab or placebo (IVT dosing) every 4-6 weeks to assess whether progression of GA can be slowed⁸⁷. 1733 participants completed the trial through 48 weeks. Intravitreal administration of lampalizumab did not reduce GA enlargement versus sham during 48 weeks of treatment. Local and proximal complement inhibition ‘did not do the trick’, but this outcome does not mean that proximal complement inhibition applied locally is destined to fail in AMD. Questions that arise include: Is the window of treatment opportunity with a complement inhibitor closed when GA has occurred? And is local and systemic complement inhibition needed simultaneously? Also, is it possible that FD is not a suitable target?

Judged by its properties, FD would appear to be a suitable target. Its plasma concentration is the lowest of all AP proteins (about 75-100 nM; 1.8 to 2.4 μg/ml at a MW of about 24 kDa). In addition, it has been described as the rate limiting enzyme of the AP^88,89, although this has been questioned recently⁹⁰..

Several caveats must also be considered when targeting FD. First, FD catalyzes the ‘promotion’ of the enzymatically very weakly (determined in murine studies⁹¹) active AP proconvertases (C3(H₂O):FB and C3b:FB) into active C3 converting enzymes. Indeed, even miniscule amounts of functional FD (about 0.5% of normal concentrations) lead to substantial convertase formation and amplification, emphasizing the critical need for complete inhibition if FD is targeted. Second, the systemic synthesis rate of FD is high at an estimated rate of 1.33 mg/kg/day in humans⁹². However, in the eye, FD ocular influx and ocular production are substantially lower making FD targeting more feasible in this tissue compartment⁹³. Third, there is gathering (unexpected) evidence that C3(H₂O)FB, without being activated by FD, entails a weak C3 convertase activity⁹¹. Lastly, although having an 800-fold lower activity than FD, pro-FD has a reasonable enzymatic activity for activating the proconvertase C3b:FB⁹⁴. This finding has no implications on how the complement cascade runs physiologically, but it may very well have important implications within the disease setting of C3G and other rare complement-mediated diseases (as discussed in the next section).

4.2. TP inhibition and beyond in C3 glomerulopathy

A prototypical AP-mediated disease is C3 glomerulopathy (C3G), a rare cause of glomerulonephritis that presents with proteinuria, hematuria, and hypertension (reviewed here⁹⁵). The glomerular inflammation culminates in irreversible tissue damage and in 50% of patients, end-stage renal disease (ESRD). In many patients, genetic or acquired drivers of disease are identified that result in either unnaturally stable C3 convertases or impaired convertase regulation thus introducing a profound imbalance between C3 activation and regulation (reviewed in ^21,95). In contrast to other complement diseases, the AP dysregulation in C3G affects primarily the fluid phase with secondary damage to surfaces like the glycocalyx of the glomerular endothelium. C3G patients with, for example, FH abnormalities have mutations or autoantibodies that either predominantly affect the N-terminal FH domains, which harbor the AP regulatory functions as opposed to the C-terminal FH host recognition domains which are mainly affected in aHUS^96,97, or lead to a complete absence of FH⁹⁵, explaining the prominent fluid-phase dysregulation with C3 consumption in plasma and strong C3 deposition in glomeruli, which can in some cases lead to TP activation^98-102. Results of elegant animal studies indicate that liberation of the anaphylatoxin C5a has a pathophysiological role in C3G while cytolysis by MAC formation has not^103,104 (reviewed in ²¹). This finding is in agreement with data demonstrating that C5b-9 complexes in G3G glomeruli remain non-lytic due to inhibition by the soluble regulators vitronectin (which is also called S protein) and clusterin¹⁰¹.

No anti-complement therapeutic has yet been approved for the treatment of C3G. Several small case studies have used eculizumab to block C5 activation and have demonstrated only a very modest improvement: some C3G patients benefitted to a certain degree while substantial proportions had no observable clinical response (discussed in ⁹⁵). This outcome is in line with biomarker data in patients and animal studies indicating C3 deposition in the glomerular microenvironment drives C3G pathophysiology with potential contributions from the anaphylatoxin C5a but not from the clusterin and/or vitronectin quenched, non-cytolytic terminal complement complex (TCC or sC5b-9)^101,103.

The focus of therapeutic intervention in C3G has shifted from inhibiting the TP (eculizumab) to inhibiting C3 activation (Table 3). The notable exception is avacopan, which blocks the signaling of the anaphylatoxin C5a which is implicated in the pathophysiology of C3G at least to some extent (see above). Inhibitors that stop C3 activation either target C3 (two different peptide inhibitors of C3) or the enzymes FD and FB (different small molecules). These are stoichiometric inhibitors that bind in a 1:1 fashion to their (activated) target and inhibit as long as they remain bound. As this review focuses on protein inhibitors, we refer to other reviews in this series for detailed coverage of small molecules and peptide inhibitors of the complement system. The only non-stoichiometric inhibitors that were or are being investigated as complement inhibitors in C3G are the protein TP10 (also known as soluble complement receptor 1) and the RNA interference (RNAi) approach using ARO-C3, which is designed to silence expression of C3.

Table 3.

Clinical trials investigating complement inhibitors in C3G listed with any status except of status ‘unknown’ on https://www.clinicaltrials.gov.

Complement target	Name of complement inhibitor	Clinical Phase	reported statuses	ClinicalTrials.gov identifiers (and date first posted)
Factor D	BCX9930 (small molecule)	II	active, not recruit.	NCT05162066 (2021)
Factor D	Danicopan (ACH-0144471) (small molecule)	II II II	completed completed completed	NCT03124368 (2017) NCT03369236 (2017) NCT03459443 (2018)
Factor B	Iptacopan (LNP023) (small molecule)	II II III II	open-label exten. completed recruiting expanded access	NCT03955445 (2019) NCT03832114 (2019)* NCT04817618 (2021) NCT05222412 (2022)
C3	Cp40 (AMY-101) (peptide)	I	completed	NCT03316521 (2017)
C3	Pegcetacoplan (APL-2) (pegylated peptide)	II II III	active, not recruit. recruiting recruiting	NCT03453619 (2018) NCT04572854 (2020)# NCT05067127 (2021)
C3	ARO-C3 (RNAi)	I	recruiting	NCT05083364 (2021)
Complement convertases	TP10 (protein)	I	withdrawn	NCT02302755 (2014)
C5a	Avacopan (small molecule)	II	completed	NCT03301467 (2017)
C5	Eculizumab (protein)	I	completed	NCT01221181 (2010)

^*addresses transplanted and not transplanted patients

^#addresses post-transplant recurrence of C3G

The attractiveness of non-stoichiometric inhibitors lies in their ‘potentiated’ inhibitory mode – one molecule of inhibitor can inhibit multiple target proteins. The RNAi approach abolishes or vastly reduces systemic expression of C3 in the liver, however in the backdrop of local C3 synthesis in different tissues (among them are kidneys¹⁰⁵), it remains to be seen how efficiently C3 production can be suppressed throughout the body. RNAi silencing of C5 has set a precedence¹⁰⁶. In a myasthenia gravis animal model, it was shown that C5 silencing achieved a potent and rather durable effect. Treated animals demonstrated reduced weakness and weight loss as well as reduction of complement-mediated destruction of neuromuscular junctions. However, although C5 was virtually absent in serum, the serum retained substantial levels of hemolytic activity, with about 20-40% of complement-mediated hemolytic activity remaining in the virtual absence of C5. It has been demonstrated that as little 1% of residual C5 is enough to achieve hemolytic activity of about 25%, highlighting how crucial complete C5 inhibition is to stop fast cytolytic processes in vulnerable cells¹⁰⁷.

These dynamics also highlight that the success of a complement inhibitor and its mode of action depend largely on the underlying compartment/cell type together with the specific pathophysiology of the disease. To stop intravascular hemolysis of vulnerable erythrocytes, C5 silencing alone is not sufficient. Considering the findings with C5 silencing, even if C3 is virtually removed, no hard conclusions can be drawn on what C3 silencing could achieve in C3G. With only residual C3 levels available, C5 activation would likely not be possible in the fluid phase, removing C5a signaling largely form the pathophysiological course (although C5 activation remains possible if C3b deposition locally were to reach high densities). More unpredictable are any effects due to local C3 production in the kidney. It must be assumed that the underlying genetic (or acquired) defects in the patients will also influence how these variables impact on the C3G pathophysiology making prognostic outlooks very imprecise.

TP10 or sCR1 is the soluble form of the natural convertase-directed complement inhibitor CR1 (Table 2). It has cofactor and decay accelerating activity and hence acts as an enzyme-like inhibitor of complement activation that is not consumed (Figure 1A,B). Promising in vitro data has shown that AP dysregulation can be ameliorated in sera of C3G patients including sera that contains nephritic factors¹⁰⁸. Animal models using FH-deficient mice (transgenic for human CR1) confirmed this finding demonstrating normalization of the serum C3 concentrations and clearance of iC3b from glomerular basement membranes¹⁰⁸. TP10 was used in a pediatric patient with end stage renal failure for a short term (FDA-approved compassionate-use investigational new drug)¹⁰⁸. In this limited study, short-term treatment increased the C3 levels and normalized TP activity as judged by sC5b-9 levels. A trial studying TP10 in a C3G patient collective has been withdrawn. Recently, a shorter variant of sCR1 with somewhat improved activity has entered preclinical development and promises potential future insights on whether a natural inhibitor, which operates in an enzyme-like fashion, can efficiently correct complement dysregulation in C3G patients and thus provide a clinical benefit¹⁰⁹.

5). Expected and unexpected findings from therapy with complement inhibitors

“The unexpected always happens,” a quote from Bram Stoker’s Dracula, offers sanguine advice to many aspects in the preclinical development and clinical use of complement inhibitors. Although unexpected findings can be unwelcome at first pass, they often offer unique and important insights into different aspects of the complement cascade. Few important illustrative examples are summarized here.

5.1). C5 inhibition and unexpected findings associated with this inhibition - explained

The first-in-class complement inhibitor eculizumab transformed the lives of patients with PNH, normalizing what was heretofore a dramatically shortened life expectancy. Eculizumab and then ravulizumab are now standard-of-care to treat classical PNH^{41-45,47,48,110,111}. And the aHUS story is very similar. TP activation is the hallmark of aHUS and its inhibition by eculizumab revolutionized the treatment of aHUS dramatically improving outcomes for aHUS patients⁶⁰.

After initial demonstration of the efficacy of eculizumab in PNH and aHUS patients, clinical complement analytics and patient monitoring revealed a few oddities that could not be explained based on the mechanism by which eculizumab inhibits C5 activation. Thus, while eculizumab worked well, it did not work completely as expected. Eculizumab has picomolar affinity for C5 with extremely low off-rates, and clearly the expectation was that eculizumab-bound C5 cannot be activated to form functional MACs¹¹². Wehling et al. found that sera from aHUS patients on eculizumab reliably inhibited CP-mediated hemolysis ex vivo, but not AP-mediated hemolysis¹¹³. In fact, up to 20% of TP activity remained in some patient sera when evaluated under AP assay conditions. In addition, elevated sC5b-9 complexes could be detected in some patients. That CP activity was completely blocked indicated this phenomenon not to be due to an under-dosing (also called pharmacokinetic breakthrough). All C5 molecules therefore were predicted to be complexed by eculizumab. This finding was seen as an oddity, and while the majority of patients responded well to anti-C5 therapy, aHUS patients with persistently elevated sC5b-9 levels showed some deterioration as measured by increases in proteinuria¹¹³.

With PNH patients, two clinical oddities were observed. While hemoglobin levels rose and lactate dehydrogenase (LDH) levels fell substantially under eculizumab therapy, as expected, a considerable number of PNH patients did not reach normal LDH or hemoglobin levels despite dramatic improvement^111,114. A significant proportion of these patients still suffered from PNH typical symptoms like fatigue. Although anti-C5 therapy mitigated most symptoms, they did not necessarily completely disappear. Therapy with ravulizumab, which effectively works like eculizumab but has improved pharmacokinetic properties, improved these laboratory parameters to a small degree¹¹⁵, reflecting the fact that pharmacokinetic breakthrough (i.e. an unintended underdosing of the drug) is less likely under ravulizumab.

5.1.1. PNH & extravascular hemolysis

One reason why anti-C5 therapy could not completely normalize hemoglobin values was due to extravascular hemolysis. Risitano et al. spearheaded this research, and others supported the finding that PNH erythrocytes become C3-fragment positive when patients are on anti-C5 therapy (resulting in a positive direct antiglobulin test for C3d)^116-118. C3d-coated PNH erythrocytes are cleared by spleen and liver, and removed from the circulation by phagocytes¹¹⁶. Of note, extravascular hemolysis will not increase plasma LDH levels but will result in fewer erythrocytes and thus lower hemoglobin levels. Normally, human erythrocytes are protected from AP-mediated attack by the integral membrane protein CR1 and the GPI-anchored proteins CD55 and CD59 (see Table 2). In addition, the soluble regulator FH recognizes sialic acid moieties on the erythrocyte surface^119-123. However, on PNH erythrocytes the GPI-anchored proteins are lacking, which leaves PNH erythrocytes vulnerable but not completely defenseless. With the remaining CR1 and FH, they can usually withstand the low-level tick-over autoactivation of C3 by the AP, however during complement amplifying conditions such as infections or surgery, the stronger-than-normal complement activation ‘spills’ over to the PNH erythrocytes, which can lead to larger scale hemolysis. As can be expected, if the TP is blocked by a therapeutic, proximal complement activation (prior C5 activation in the cascade) still occurs during complement amplifying conditions fixing C3 opsonins on PNH RBCs. In contrast, during conditions in which complement is not blocked by external agents, the under-regulated PNH erythrocytes that were to start fixing ‘detectable’ amounts of C3 would immediately progress to forcefully trigger the AP amplification loop leading to instantaneous lysis via MAC. Hence C3 positive PNH erythrocytes are only observable when PNH patients receive C5 inhibitors. Under eculizumab treatment, C5 activation and TP activity is blocked leading to the accumulation of C3-fragments on the surface of PNH erythrocytes. This is why PNH erythrocytes, which traditionally were known not to be C3-positive in clinical tests, transform into a C3 positive status when C5 activity is inhibited therapeutically. C3 positivity refers to a covalent fixing of either C3b, iC3b, C3dg or C3d on a cell surface. C3b and especially iC3b coatings are considered to aid the phagocytic uptake. However, usually, the C3dg or C3d (see also below) fragment is detected on PNH erythrocytes of patients on anti-C5 therapy (with C3d being the end product of C3b degradation: C3b is cleaved by FI into iC3b and then C3dg, which in turn is finally processed to C3d by plasma proteases removing a small peptide stretch²⁸). But does opsonization with C3dg or C3d even trigger phagocytosis of tagged cells? Traditionally, C3b and its first inactivation product iC3b, but not C3dg or C3d, were considered tags for phagocytic uptake. However, Bajic et al. showed that CR3 engages with the C3d fragment¹²⁴. This observation was confirmed by Lin et al. who demonstrated in vitro that C3d-coated PNH erythrocytes do indeed become phagocytosed in a CR3-dependent manner¹²⁵. Hence, both, iC3b and C3dg/C3d need to be considered as C3 opsonins triggering phagocytic uptake.

5.1.2. PNH & ongoing intravascular hemolysis (i.e. pharmacodynamic breakthrough)

Another reason (in addition to extravascular hemolysis) why anti-C5 therapy does not completely normalize hemoglobin values was suspected to be connected to the slightly higher than normal LDH levels found in a substantial number of PNH patients treated with eculizumab or ravulizumab. LDH is a marker for intravascular hemolysis and slightly increased LDH levels indicate residual intravascular hemolysis, as proposed by Risitano¹¹¹. The hallmark of this low-level hemolysis is that it happens constantly although there is an excess of eculizumab (or another C5 inhibitor) over C5.¹²⁶. Harder et al. proposed that in cases of ‘transient’ massive complement activation (also referred to as complement amplifying conditions, e.g. during infections), the constant low-level C5 activity is drastically increased and results in a larger scale of intravascular hemolysis of PNH erythrocytes¹²⁶.

Clinicians have observed hemolytic episodes during infections or surgery and have referred to the phenomenon as pharmacodynamic breakthrough as it was not associated with an under-dosing of the drug but instead with complement amplifying conditions (Figure 2A)^127,128. This phenomenon was documented convincingly by Harder et al.¹²⁶ who summarized the clinical course in a small cohort of PNH patients¹²⁶. LDH deterioration either associated with being at the end of a dosing interval (suggesting pharmacokinetic breakthrough) or with the occurrence of a clinically proven infection at any time during the eculizumab dosing interval (suggesting pharmacodynamic breakthrough). In one patient with a LDH spike during an infection, it was experimentally determined that a surplus of eculizumab over C5 was present.

Figure 2. Residual C5 activity.

A) LDH as a measure of intravascular hemolysis. LDH course in a subset of five PNH patients is shown. The dashed black line indicates the upper limit of normal for LDH (250 u/l). The patients in this subset were all on eculizumab therapy. High LDH values were measured when infections could be documented concomitantly. For patient #01 constantly high LDH values were measured, which exacerbated during signs of severe infection. A dashed circle indicates a LDH peak value at a time point for which signs of infections were documented. (Units: LDH u/l; CRP: mg/l). B) A structural cartoon of C5 and the three C5 inhibitors eculizumab (Fab fragment depicted), coversin (or OmCI) and RaCI. The cartoon highlights the C5 domains to which the individual C5 inhibitors bind. Also, the C5a scissile bond just C-terminal of the C5a domain is shown. This cartoon figure is based on structural data by Jore et al¹³⁰. C) CP-mediated lysis of sheep RBCs (shRBCs). Hemolysin-treated shRBCs were mixed with 5% serum in the presence of specific inhibitors. Released hemoglobin was measured as a marker of hemolysis (the average of 3 independent assays with standard deviation [SD] is shown). D) AP-mediated lysis of rabbit erythrocytes. Rabbit erythrocytes were incubated in 25% human serum in presence of inhibitors or controls (average of 3 independent assays with SD is shown). E) TP activity of serum from PNH patients on eculizumab. Rabbit RBCs were incubated in 25% human serum derived from PNH patients on eculizumab with or without addition of other complement inhibitors as indicated (average of 2 independent assays with SD is shown). F) Residual hemolysis was tested in sera of 13 PNH patients. All but one patient was on eculizumab therapy (one patient did not receive anti-complement therapy). A standard rabbit RBC assay with the sera of the patients was performed. Hemolysis was measured in serum or in serum that had been supplemented in vitro with eculizumab to reach a final concentration of “in vitro added Eculizumab” of 1.6μM (final serum content was 25%). #01–09 and #11–13 were on Eculizumab treatment; serum from these patients contained eculizumab prior to the in vitro addition of extra eculizumab. Patient #10 was not on eculizumab treatment.

Sources of figures: A) and F) reproduced form Harder et al (2019)¹²⁶ under the Creative Commons CC-BY license (CC-BY 4.0). B) is an adapted figure from Jore et al¹³⁰ with novel elements. C), D) and E) panels are reproduced from Harder et al¹⁰⁷ with permission.

The phenomenological distinction between pharmacokinetic (too little drug) and pharmacodynamic breakthrough hemolysis (hemolytic activity despites enough drug) is very important, as it indicates whether administration of an extra dose of eculizumab may be expected to provide a clinical benefit or not¹¹¹. In aggregate, pharmacodynamic breakthrough means continuous low-level residual C5 activity (with low-level clinical implications) which can get exacerbated transiently during complement amplifying conditions - it cannot be alleviated with additional administration of the same C5 inhibitor. This insight was a landmark concept for understanding the intricacies of treating PNH patients with C5 inhibitors, however, what pharmacodynamic breakthrough meant in terms of complement mechanisms was not clear at this point. The two important mechanistic questions associated with this phenomenon were whether the observed residual hemolytic activity could be explained by residual C5 activity (proving that it is indeed complement mediated), and if so, how an C5 inhibitor with very high affinity for C5 could fail to completely inhibit C5 function.

5.1.3. Mechanistic aspects of residual C5 activity in presence of stoichiometric C5 inhibitors

To address the first mechanistic question, the C5 inhibitors eculizumab and coversin were investigated in vitro in standard CP- and AP-mediated hemolysis assays¹⁰⁷. Coversin is the recombinant version of the Ornithodoros moubata complement inhibitor (OmCI), a natural C5 inhibitor found in the saliva of this soft tick¹²⁹. Structural data show that Coversin and eculizumab bind to C5 at two widely separated sites and can bind simultaneously to one C5 molecule^130,131 (Figure 2B). A PASylated version of coversin, which effectively is a genetic fusion with a conformationally disordered polypeptide composed of the amino acids proline, serine and alanine, has a higher hydrodynamic radius and a longer plasma half-life and was also investigated^107,132. In the CP assay (at a low serum percentage of 5% or lower), eculizumab and coversin completely inhibit hemolysis (Figure 2C,D), however in the AP assay in 25% serum, substantial levels of residual hemolysis of about 40%, 30% and 25% can be observed for surplus amounts (over C5 in the assay) of coversin, eculizumab and PASylated coversin, respectively. An extremely interesting observation was that the simultaneous application of two orthogonal C5 inhibitors, eculizumab and coversin, was able to completely inhibit hemolysis. This observation proves that the residual hemolytic activity in presence of one C5 inhibitor indeed corresponds to residual C5 activity and thus is clearly complement mediated. In contrast, the use of the proximal AP inhibitor miniFH alone was sufficient to efficiently inhibit AP-mediated hemolysis^107,133.

That two orthogonal C5 inhibitors applied simultaneously achieve complete C5 inhibition was not only true for the combination of the C5 inhibitor pair of eculizumab and coversin, but also for the pair of eculizumab and a commercially available anti-C5 monoclonal antibody (only available as in vitro reagent), which on its own only inhibited C5 activity minimally. The finding that three different C5 inhibitors (of which one is used in the clinic, one is in clinical evaluation, and one is a reagent) inhibited C5 lytic function only partially, is a strong indication that residual C5 activation is a class effect of stoichiometric C5 inhibitors. Any binary combination of the three single C5 inhibitors, however, completely inhibited C5 activation. Harder et al.¹⁰⁷ observed that the effect of residual C5 activity is also present in an ex vivo assay using sera from PNH patients on eculizumab (Figure 2E). In this setting, only C5 double inhibition (by adding coversin into patient serum that already contains eculizumab) or addition of the proximal inhibitor miniFH completely protected the strongly AP-activating rabbit erythrocytes from lysis, confirming the in vitro results.

Indications that different C5 inhibitors do not completely protect from AP-mediated hemolysis of rabbit erythrocytes, while complete inhibition in the CP assay could be achieved, had already been reported. However, it has not been discussed or investigated further prior to the publication by Harder et al¹⁰⁷. Nunn et al show such a behavior for coversin (then called OmCI)¹²⁹, Blom et aldemonstrated the same behavior for eculizumab and coversin (then called OmCI)¹³⁴, and Jore et al showed this same effect for yet another class of C5 inhibitors, called Rhipicephalus appendiculatus C5 inhibitors (RaCIs), which are also derived from the saliva of ticks¹³⁰. The residual hemolytic effects in these early reports were rather mild and were not further pursued, probably because the effect had been small (but still distinct) due to the low serum percentage. The higher serum concentration used in the AP assays by Harder et al (25% serum content) magnified this effect. However, it was noticeable that higher serum content (2.5, 3.75 and 5%) increased even the very low-level residual hemolysis in this specialized convertase hemolytic assay investigated by Blom et al¹²⁰.

In aggregate, these data show that stronger complement activation (i.e. higher serum concentration¹³⁴, stronger activating cells as in the comparison of PNH-RBCs to rRBCs¹⁰⁷, or an alloantibody titer titration in a sensitive mixed CP/AP assay in 50% serum¹⁰⁷) leads to higher residual C5 activity and hence higher residual cytolysis. Naturally, due to the high sensitivity of erythrocytes for MAC-mediated hemolysis, the effect of residual lytic C5 activity is especially emphasized on erythrocytes. Residual C5 activation also applies to nucleated cells, but on this cell type was not sufficient to yield cell lysis. However, activation of nucleated cells did occur when they were damaged/stimulated by residual C5 activity²⁵. To assess the residual C5 activation quantitatively Harder et al.¹⁰⁷ performed their standard AP assay with rabbit erythrocytes in C5-depleted serum and titrated back purified C5 protein. As little as 1% of the normal C5 concentration in 25% serum (the final serum content in the assay) was necessary to achieve a 30% level of hemolysis. By analogy, this finding would suggest that about 1% of the C5 within the eculizumab inhibited serum must assume an activated state, as in presence of surplus amounts of eculizumab the residual hemolysis was 30%.

5.1.4. Residual C5 activity and individual patient-specific aspects

In a follow up publication, Harder et al¹²⁶ investigated whether the residual levels of C5 activation were similar across patients or depended on patient-specific factors. They investigated sera from 12 PNH patients on eculizumab therapy using the AP rabbit erythrocyte assay with and without the addition of additional eculizumab¹²⁶. As control, one PNH patient not on any anti-complement therapy was included Figure 2F). The addition of extra eculizumab to patient serum showed a constant but overall very marginal effect to further reduce residual hemolysis. Remarkably, the levels of residual C5 activity across the 12 patients on eculizumab therapy spread from almost 40% to about 10%. Potential explanations for this range of residual lytic activity could be patient-specific differences in serum levels of eculizumab (as the extra addition of eculizumab marginally reduced the residual hemolytic levels further ex vivo) and/or single nucleotide polymorphisms (SNP) in the complement genes that encode C5 and/or complement components that form the convertases. The latter would be in line with the concept of the complotype, which describes that a set of SNPs in complement genes can have an additive effect on complement activity^135,136.

5.1.5. Conclusion, clinical evidence for C5 double inhibition, and outlook

In conclusion, pharmacodynamic breakthrough under anti-C5 therapy is C5 dependent. Further addition of the same C5 inhibitor only marginally changes the levels of ongoing residual complement activity. However, addition of an orthogonal complement inhibitor (another C5 inhibitor or a proximal complement inhibitor) provides complete complement control. A clinical trial investigating the switch of PNH patients from the C5 inhibitor eculizumab to the C5 inhibitor crovalimab provided the first clinical data that double C5 inhibition indeed stops residual C5 activity in vivo¹³⁷. When crovalimab was administered at the end of the natural eculizumab dosing interval, patients still had significant amounts of eculizumab in their circulation, resulting in C5 double inhibition during the switchover period, with a slight but notable higher hemoglobin and lower LDH and reticulocyte levels – all indicative of a better hematological response. Fifty days after the switch (expecting only single C5 inhibition by crovalimab), the hematological parameters approached those levels recorded during eculizumab therapy before the switch. In the future, double C5 inhibition may be important for patients who suffer from very strong complement activating conditions and consequently achieve suboptimal benefit from a single C5 inhibitor.

It is also remarkable that not all C5 inhibitors yield identical levels of residual C5 activity in vitro (e.g. comparing eculizumab to coversin)^107,126. Of note, residual TP activity under C5 inhibition only happens when complement is activated very strongly. The more powerfully complement is activated, the less effective is terminal pathway inhibition by diverse anti-C5 agents¹⁰⁷. That sera from PNH patients on eculizumab exhibit rather different levels of residual C5 activity ex vivo could potentially also have clinical implications and should be investigate in more detail. The summarized studies provided evidence that pharmacodynamic breakthrough is a complement-dependent phenomenon that centers on residual C5 activity, which becomes exacerbated during complement-amplifying conditions.

5.2). Mechanistic investigation into residual C5 activity in the presence of C5 inhibitors sheds light onto the nature of the C5 convertase of the AP

How could residual C5 activity be maintained in presence of excess amounts of different C5 inhibitors, which bind their target with picomolar or nanomolar affinity^112,132,138? Solving this question involves not only the C5:inhibitor complexes alone but how these complexes interact with the C5 convertases, which in turn focuses on the question on the molecular composition of the C5 convertase and how it operates.

Despite the well-defined pathways and functional roles of complement in textbooks, recent basic, translational, and clinical research has challenged the dogma of how the terminal pathway is initiated by proteolytic activation of C5. The role of a well-defined trimolecular convertase C3bBb3b (and in analogy also C4b2a3b; some experts re-designated the CP C3 convertase as C4b2b¹³⁹, however in this review the classical nomenclature C4b2a is used) is called into question. From a complement conceptional point of view, the role for a trimolecular convertase has never been compatible with what is known about complement dynamics of TP activation. If it were true that a bimolecular C3 convertase recruits any additionally produced C3b to form a trimolecular convertase, then C5 activation could/should occur immediately after C3 convertases have processed the first few C3 substrates. However, this temporal sequence is not what occurs. Efficient C5 activation only occurs on surfaces after strong complement activation has taken place^24,140. C5 activation in the fluid phase does not happen, unless rather high concentrations of soluble C3b are present – and even under these conditions C5 activation in the fluid phase is minimal²⁴.

Physiologically, this sequence of events appears very reasonable. C3 activation and opsonization maintains homeostasis and is a required constant process^141,142. C3 activation therefore should not easily drift into C5 activation, which has inflammatory (C5a) and cytolytic (MAC) roles. As such, a strong checkpoint must be maintained to control the shift from the proximal pathways to the terminal and inflammatory one. And this checkpoint appears to be the accumulated C3b density on an activation surface. If a high C3b density is reached, C5 activation becomes possible. To reach this tipping point on a host surface, which is equipped with pre-existing regulators of defense²², C3b deposition must happen extremely fast. Otherwise, the host C3 regulation cycle will process the C3b opsonins into the late stage opsonins iC3b and C3dg, which do not fuel convertase activity (Figure 1). What mechanistic data support this concept?

Berends et al found that C5 directly binds to high densities of C3b on beads and can be activated upon addition of Factor D and Factor B²⁴. Strikingly, these data are consistent with a not much noticed report by Vogt et al in 1978 emphasizing the finding that C5 can directly bind to C3b in the absence of complement convertases¹⁴⁰. The structural report by Jore et al shows that three different C5 inhibitors bind to completely orthogonal sites on the surface of C5 but share the ability to inhibit proteolytic C5 activation (also discussed above)¹³⁰. This finding conflicts with the generally proposed steric-inhibition model of C5 inhibitors (Figure 3), which suggests that C5 inhibitors hinder the access of the substrate C5 to its convertase. Instead, the mentioned structural data suggest that a C5 priming event on C3b is needed for proteolytic activation of C5 by bimolecular complement convertases (e.g. C3bBb), which was also proposed by Vogt et al in 1978. Only upon structural changes in C5 does the scissile bond at the C-terminus of C5a become available for processing by C3bBb. By several different approaches using cellular and purified protein systems, Harder et al¹⁰⁷ and Mannes et al²⁵ confirmed the findings by Berends et al²⁴ and Vogt et al¹⁴⁰, and showed that only high densities of C3b on a surface (in the absence of convertases) are capable of reversibly recruiting C5. It was also demonstrated that C4b shares this functionality with C3b (see also below C3 bypass activation of C5 below)²⁵. Moreover, Mannes et al demonstrated that C3bBb (on an SPR sensorchip surface) is able to bind and proteolytically process C3, but is unable to bind C3b (which was applied in the fluid phase on to a sensorchip surface deposited with C3bBb), questioning the existence of the trimolecular convertase C3bBbC3b, at least in this setting²⁵. Mannes et al also demonstrates that soluble bimolecular C3bBb convertases are fully capable of proteolytically processing C5 once that C5 had been recruited to a surface densely packed with C3b (or C4b). This finding is fully compatible with the report by Vogt et al¹⁴⁰. Vogt et al¹⁴⁰, Berends et al²⁴ and Mannes et al²⁵ consistently demonstrate that no covalent linkage between a convertase unit (either C3bBb or C4b2a) and an additional C3b molecule is needed to obtain efficient convertase activity, but demonstrate that additional C3b molecules are required to make up a C3b dense surface that captures and prepares C5 for the convertase units. However, the three reports do not technically disprove that a covalent linkage of a C3b molecule to C3bBb/C4b2a may yield a functional C5 convertase as had been proposed before^143,144.

Figure 3. Structures of C5 convertase and C5:inhibitor complexes.

A) A structural model showing C5 (red; with anaphylatoxin domain C5a in dark red) and the convertase C3bBb in sand and gold colour, respectively. The model was derived by utilising pymol and aligning the complex structure of C3b:Bb:SCIN (pdb code: 2WIN¹⁷⁹) on to the CVF:C5 complex structure (pdb-code: 3PRX ¹⁸⁰). The anaphylatoxin domain C5a in C5 is positioned near the catalytic unit of Bb. B) As in (A), but the two tick-derived C5 inhibitors OmCI and RaCI are shown bound to C5 (pdb-code: 6RQJ ¹⁸¹). The C5 inhibitors OmCI and RaCI bind on to C5 patches that are far away from the C5a cleavage site. Thus, the blocking mode of these inhibitors does not involve steric competition with convertase binding to C5. C) As in (A) but this time C5 is shown in its complex structure with a Fab fragment of Eculizumab (pdb-code: 5I5K ¹³¹). In contrast to OmCI and RaCI, eculizumab clearly competes with the convertase for C5 binding. D) A close-up image of the C5a-Bb area derived from (B). C5, OmCI and RaCI are shown in cartoon representation. Bb is shown in surface representation with the three amino acids of the catalytic triad being displayed in blue. The scissile bond is buried in a α-helix in C5 and oriented in a way that it is not accessible to the catalytic unit of Bb indicating that C5 must change its conformation (in a so-called priming step) to prepare for the scissile bond to be cleaved by Bb. Any C5 inhibitor is proposed to interfere with such structural (priming) rearrangements and/or sterically hinder the access to Bb, as in the case of eculizumab.

Taken together, the surfaces of cells very densely packed with C3b (and/or C4b molecules) acquire the ‘new’ property of binding C5. The level of C5 recruitment to these surfaces correlates with the level of deposited C3b and/or C4b molecules. C5 can bind efficiently (with affinities of 2.5 or 1.2 μM, respectively²⁵) to surfaces covered with C3b or C4b, suggesting that at least two neighboring C3b/C4b molecules are required to efficiently recruit C5. In this model, one molecule of C5 is thought to interact with two different binding patches on each of the two neighboring C3b/C4b molecules. This model proposes that simultaneous binding of C5 to two neighboring C3b/C4b molecules captures and prepares C5 for proteolytic activation by a convertase enzyme. Zwarthoff et al and Mannes et al could narrow down at least two regions on C3b that interact with C5 by employing biophysical and well-defined cellular complement assays with purified components^25,145. One is the C3b patch of macroglobulin domains 3 to 6 (MG3-6 site) where CRIg binds. The second site is an area at or around the C3b CUB domain (where Factor B binds)²⁵. As hinted above, in terms of the biological regulation of the complement cascade, the requirement of neighboring surface-attached C3b/C4b molecules for C5 activation appears to represents a safety mechanism: only strong complement activation leading to high C3b (and/or C4b) densities allows for the transition from the proximal complement cascade (C3 activation) to the terminal pathway (C5 activation), which holds the highest inflammatory potential of the cascade (i.e. C5a anaphylatoxin release and membrane attack complex formation).

In conclusion, several lines of evidence suggest that only high densities of C3b/C4b molecules deposited on a surface lead to efficient recruitment and priming of C5. The priming step is proposed to involve conformational changes in C5 that expose the scissile peptide bond to allow access to the convertase^130,140. These concepts are summarized in a reorganization of the complement cascade stressing the significance of the C5 priming step and the consecutive proteolytical activation of primed C5 by bimolecular convertases in Figure 4²⁵. Several important mechanistic questions about C5 priming remain. These entail molecular requirements of C5 priming like the exact density and spacing between C3b and/or C4b molecules, the stoichiometry of the interaction, and the involvement of particular domains of C3 and C5. Even though not all molecular details of C5 priming are available, the concept has been instrumental in understanding how eculizumab-bound C5 can retain cytolytic activity. In fact, it appears that most (if not all) single stoichiometric C5 inhibitors are limited by some degree of residual C5 activity.

Figure 4. A new schematic drawing of the complement cascade.

The new feature of this complement scheme is the absence of trimolecular C5 convertases. Instead, it is envisaged that the bimolecular C3 convertases C3bBb (or C3bBbP) and C4b2a proteolytically activate C5 once it is conformationally prepared for cleavage (or primed) on highly dense clusters of C3b alone or C3b and C4b.

Figure is reproduced with permission from Mannes et al²⁵.

5.2.1. Conformational activation of C5 and its resistance to C5 inhibitors

Cells exposed to strong complement activation in presence of C5 inhibitors fix many more C3b opsonins than under non-inhibiting conditions. As the transition to initiate the TP is blocked by the C5-inhibitors, the cells under attack survive long enough to have the AP amplification loop deposit unnatural high amounts of C3b molecules on their surfaces¹⁰⁷. So, the better (or more efficiently) the C5 inhibitor works, the longer the cell under complement attack will survive and the denser the C3b deposition will become.

Harder et al¹⁰⁷ demonstrated that C5 complexed by different C5-inhibitors retains at least a proportion of its original activity to adhere to a C3b-deposited surface (in absence of convertases). Mannes et al showed that the C5 molecules recruited to a densely deposited C3b surface i) still contain the C5a moiety and hence truly represent C5 and not C5b molecules, ii) bind eculizumab in a C5-dependent manner, and iii) upon addition of C6-C9 and in absence of any convertases but in presence of eculizumab lyse cells²⁵. This outcome signifies that eculizumab-bound C5 can become ‘trapped’ on an extremely dense C3b surface and eventually assumes a conformation that can assemble cytolytic MACs. In these studies, the presence of the C5a moiety confirmed that C5 was captured and not C5b. After the lysis reaction with C6-C9, the C5a moiety disappeared, indicating that it was activated C5 that initiated cytolysis.

This sequence of events was transformed to a purified system on an SPR chip (working only with purified proteins) to cross-validate results obtained in the more complex cellular system. Mannes et al proposed that eculizumab-bound C5 that interacts intensively with an unnaturally highly dense C3b surface stays in a primed conformation for extensive time, and eventually transitions to a C5b-like conformation. This transition is reminiscent of the conformational autoactivation of C3 into C3(H₂O), and hence the conformational activation of C5 into a C5b-like conformation on C3b-dense surfaces was referred to as C5_conf²⁵. That C5 has a somewhat labile inactive conformation is well documented. Chemical stress (low pH)¹⁴⁶, oxygen radicals^147,148 or physical stress¹⁴⁹ trigger non-enzymatic C5 activation resulting in a C5b-like conformation (i.e. C5_conf) that is capable of initiating reactive lysis²⁵. Of note, eculizumab binds C5, C5b6 and sC5b-9 with nearly identical high affinity, explaining why eculizumab is observable when C5:eculizumab complexes adhere on highly dense C3b surfaces. Eculizumab, which only binds to the single C5 domain (i.e. MG7), is expected to remain bound to C5 when C5 transitions to its primed and then eventually into its C5_conf state. This reflects the concept that eculizumab introduces steric hindrance for the convertase, while coversin (OmCI) and RaCI mainly stabilize the unproductive, non-primed C5 convertase by binding to (or ‘non-covalently cross-linking’) multiple C5 domains simultaneously (Figure 2B, 3B). These insights help to explain hypothetically why two orthogonal C5 inhibitors applied simultaneously can completely inhibit C5 activation and even abolish residual C5 activity despite the presence of highly dense C3b surfaces. Two orthogonal C5 inhibitors can lock C5 in its unproductive state. While this hypothesis explains the presented data, it must be confirmed. However, if all residual C5 activation is to be prevented, two orthogonal stoichiometric C5 inhibitors (like eculizumab and coversin, or RaCI or potentially another anti-C5 antibody) should be used^107,150.

5.3). C3 bypass activation of C5

It is well established that terminal pathway activation is initiated only when bimolecular C3 convertases (C3bBb or C4b2a of the AP and CP/LP, respectively) associate with additional C3b molecules to shift the substrate specificity of the convertases from C3 to C5 (disregarding the non-physiological CVF convertase CVFBb)²³. This certainly has been and probably still is the most prevalent concept about the complement cascade. The recent publication by Zhang et al convincingly challenges this dogma with data showing that C5 activation occurs subsequent to CP activation and in absence of C3 in several ex vivo and in vivo experiments utilizing C3-depleted sera or C3 knock out animals (Figure 5)¹⁵¹. Given the high homology between the convertase formation routes across the alternative and classical/lectin pathway, it may not be surprising that a dense C4b surface can substitute for a dense C3b one and prime C5 for proteolytic cleavage by a complement convertase (see also above). Mannes et al demonstrated that a C4b surface binds C5 molecules in a similar fashion to a C3b surface²⁵. Exclusive deposition of C4b on an SPR chip or C4b on an erythrocyte surface is sufficient for C5 capture. In fact, the affinity of C5 for C4b is slightly higher than for C3b²⁵.

Figure 5. Comparison of the classical pathway–mediated hemolysis between the C3-sufficient and C3-Dpl human or C3-deficient rat sera.

For the measurements of classical pathway–mediated hemolysis, antibody-sensitized sheep erythrocytes (EshA) were incubated with different concentrations of NHS, C3-Dpl human sera (A,C), or WT C3-deficient (C3KO) rat sera (B,D) in GVB buffer, and hemolysis was quantitated by measuring levels of released hemoglobin in the supernatants. These data show significantly reduced classical pathway-mediated hemolysis in the absence of C3 when lower concentrations of sera were used (A-B), but comparable and almost complete hemolysis when higher concentrations of sera were used (C-D). (*P, 0.05).

Figure reproduced from Zhang et al with permission¹⁵¹.

In healthy individuals, any CP/LP activation leads to C3b generation via the AP amplification loop. It must be assumed that it is easier for C3/C3b than for C4/C4b to reach sufficient densities to prime C5, since the serum concentration of C3 is at least twice as high as C4 and C3b can self-propagate through the AP amplification loop. It has to be stressed that in contrast to activation of C5 by the AP, C5 activation exclusively by the CP/LP bears only little physiological relevance since C3 deficiency in man is extremely rare. However, therapeutic intervention at the level of C3 by complement inhibitors that bind native C3 and block its activation create a clinical scenario that resembles the physiological situation of C3 deficiency. An example of such inhibitors is found within the compstatin family^152-155: One compstatin member (pegcetacoplan) is approved for treatment of PNH¹⁵⁶. In addition, compstatin family members are presently evaluated for their therapeutic potential in clinical trials with two of them being investigated for diseases mediated by CP complement activation, i.e. ABO incompatible kidney transplantation and autoimmune hemolytic anemia⁴⁶. For these scenarios it is crucial to know whether such C3 inhibition can protect sufficiently from C5 activation by the CP/LP. In vitro assays demonstrated that a compstatin analog completely blocked AP-mediated lysis of rabbit erythrocytes but failed to protect CP-mediated lysis of sensitized sheep erythrocytes (Figure 6)²⁵. In the assay, 80% serum had been supplemented with 16 μM compstatin, which exceeded the C3 concentration at least twofold. It is mandatory to use high serum contents, as the typically very low serum percentages used for the common type of CP assays are not suitable to demonstrate this effect. This result is in agreement with Zhang et al¹⁵¹ and probably reflects the lower C4 concentration in serum (in comparison to C3). Mannes et al proposed to designate C5 activation in absence of C3 as ‘C3 bypass activation of C5’.

Figure 6. Strong CP activation leads to hemolysis despite of C3 peptide inhibitor Cp40.

CP- and AP-mediated lysis in presence of Cp40. 80% NHS naturally containing antibodies against the Forssman antigen (which is present on shRBCs) was mixed with Cp40 to obtain final Cp40 concentrations of 0.25-16 μM, as indicated. AP activity was determined by mixing rabbit erythrocytes with 80% NHS supplemented with 5 mM Mg-EGTA in presence of the same range of Cp40 concentrations. A serum mix with PBS instead of Cp40 served as positive control, while 5 mM EDTA and 5 mM Mg-EGTA dissolved in PBS (in case of CP; for each final concentration) served as negative controls (PBSE: PBS containing EDTA). Released hemoglobin was measured as a marker of hemolysis (average of 3 independent assays with SD is shown).

Figure is reproduced with permission from Mannes et al²⁵.

C3 bypass activation of C5 contradicts a central dogma of complement biology and has little physiological relevance but is important for the development of complement therapeutics and thus has immense translational importance. For example, it explains why application of compstatin in different in vitro and ex vivo assays does not always prevent TP activation completely^157,158. It is also important to recognize that several in vivo studies have utilized C3-deficient animals or treatment with cobra venom factor (CVF) to deplete wildtype animals of C3 to interrogate the pathophysiology of different diseases in which complement participation was expected. If C5 activation products had been detected in these preclinical models, the interpretation was that an extrinsic pathway of complement activation could be at play^159,160. While this may still be possible, in light of the C3 bypass activation of C5, the role of any extrinsic pathway would be much less pronounced. Other studies for example have used autoantibody-mediated arthritis models or infectious disease models in animals with genetic (knockout) or pharmacological (CVF) C3 depletion^161,162. That C3 bypass activation of C5 occurs challenges the conclusions of these studies and in hindsight, it may be feasible to reevaluate their conclusions. Perhaps the conditions investigated could still be amenable to therapeutic complement intervention considering the existence of a C3 bypass activation route. Moreover, there are further, probably underappreciated, reports available showing that C5 can be activated without participation of C3. Kitamura et al¹⁶³ reported that complement immune hemolysis is possible in absence of C3. Sano et al¹⁶⁴ demonstrated that sera from C3-deficient patients retain some hemolytic activity that is C3-independent.

5.4). FB enzymatic activity in absence of FD

After uncovering several unexpected findings about the central complement proteins C3 and C5, it comes as no surprise that unexpected findings can be uncovered more proximal in the complement cascade as well. Indeed, it appears that the unexpected must be expected when inhibiting the multi protein cascade of the complement system. Unexpected activation routes can be exposed, especially when one of the many enzymatic steps within the complement cascade is completely or nearly completely blocked. For example, Zhang et al evaluated whether absence or inhibition of the rate limiting enzyme of the AP, Factor D (FD), would alleviate disease burden in a well-established C3G mouse model⁹¹. The underlying cause of C3G in some patients is the absence of FH⁹⁵, and FH KO mice are a well-established model in which to study C3G pathophysiology¹⁰³. To simulate the therapeutic effect of different FD inhibitors, Zhang et al generated FH/FD double-KO mice and compared C3 glomerulopathy between the FH-KO and the FH/FD double-KO mice. Remarkably and unexpectedly, the C3G phenotype seen in the FD-KO mice was not rescued by the complete absence of FD. Investigation into this phenomenon revealed that serum from the FH/FD double-KO mice is capable of depositing small amounts of C3b on surfaces that activate the AP. Despite AP activation being lower than normal, it was sufficient to result in complement-mediated rabbit erythrocyte hemolysis. Mechanistically, AP activity on surfaces of the FH/FD double-KO mice results from the slow tick-over activation mode of C3, yielding C3(H₂O). In line with this, the tissues of the FH/FD double-KO mice exhibited abundant amounts of C3(H₂O) and its inactivation product iC3(H₂O)⁹¹. That C3(H2O) can be adsorbed non-covalently (as the thioester is already lysed) to lipid and other surfaces (e.g. activated platelets) had been reported before¹⁶⁵.

Since in the FH/FD double-KO mice C3 is not rapidly consumed in the fluid phase by the AP amplification loop (due to the lack of FD), substantial amounts of C3 are left to undergo tick-over activation with C3(H₂O) being continuously adsorbed on tissue surfaces. Such adsorbed C3(H₂O) on its own, however, is unlikely to be the sole explanation for the remaining tissue damage in the FH/FD double-KO mice. Accumulated C3(H₂O) in the tissues naturally associates with FB to form the early AP proconvertase, C3(H2O)FB, which in turn generates C3b. The FH/FD double-KO model shows that prolonged existence of C3(H2O)FB enables FB to assume an open and enzymatically active conformation even in absence of proteolytic activating by FD. In essence, this conformational variability allows low level AP activation that is restricted to the tissue surfaces resulting in considerable AP-mediated damage. This study demonstrates elegantly just how important it is to understand the impact of complement blockade at a certain step before moving forward to clinical investigations.

5.5). Learning from the unexpected

Scientific oddities that do not make sense at first glance hold potential for new discoveries. In the complement field, this has been the case when it was recognized that: i) C5 can be activated in absence of C3; ii) C5 can undergo conformational activation in the presence of stoichiometric C5 inhibitors; and iii) C3 and C5 can be activated in absence of FD in the murine C3G model. The greatest value of these novel insights undoubtedly lies in the opportunity to optimize therapeutic strategies or in the preclinical development of new drugs. Figure 7 summarizes how the new insights discussed here have necessitated that we rethink some of the fundamental complement concepts. And undoubtedly there will be new lessons to be learned.

Figure 7. Graphical summary of how research into unexpected findings has helped to redefine some basic complement concepts.

A) C3 activation in a murine C3G model in the absence of FD. FH knockout mice are a versatile and justified model to study C3G pathophysiology, which results in mild AP-mediated glomerulopathy. In an attempt to stop complement AP pathology, FD was knocked out. Unexpectedly this resulted in denser C3 deposition and more severe kidney injury. Fluid phase consumption in the FH-KO/FD-KO was impaired as well as degradation of C3(H₂O) into iC3(H₂O). This in turn resulted in the accumulation of C3(H₂O) in the glomerular environment where C3(H₂O) could recruit FB to the glomerular surface. Prolonged lifetime of the early AP proconvertase C3(H₂O)FB (in absence of FH) enabled FB to assume an open, enzymatic active conformation although it had not been proteolytically activated by FD. C3 activation therefore was focused on the glomerular microenvironment leading to more severe damage in FD deficient mice. B) The data summarized propose re-assessments of several complement concepts. The existence of trimolecular C5 convertases is challenged. Instead, it is proposed that C5 is primed (prepared) by a dense C3b (and/or C4b) cluster for proteolytical activation by a bimolecular convertase. In the absence of functional C3 (during C3 inhibition), C3 can still be activated as long as the CP/LP are triggered very strongly. In presence of stoichiometric C5 inhibitors, cells surfaces can fix unnaturally high densities of C3b which recruit C5:C5-inhibitor complexes. The multiple interactions with and long residence time on the C3b cluster of the C5:C5-inhibitor complex induces the conformational activation of C5. As a result lytic activity occurs despites the presence of a C5 inhibitor and although C5a has not been cleaved.

Panel B is reproduced with permission from the graphical abstract by Mannes et al²⁵ containing a modification of the title.

6). The evolution of therapeutic complement inhibitory proteins

Only 15 years have passed since the first complement therapeutic eculizumab was approved for the treatment of PNH, during which time the field of complement therapeutics has evolved to adapt to new challenges (Figure 8). Complement therapeutics are approved for five indications, four of which are treated with C5 inhibition. One of the complexities noticed relatively early after eculizumab had been approved was that some PNH patients suffered from pharmacokinetic breakthrough. The dose of 900 mg every two weeks was not adequate for all PNH patients. Adaptions to the dose were made possible and for the second indication, aHUS, the dose was set to 1200 mg every second week. The pharmacokinetic properties of eculizumab were improved when a second-generation version of the antibody, called ravulizumab, was approved for the therapy of PNH in 2019. This longer-acting complement C5 inhibitor is administered every eight weeks and achieves complete and sustained inhibition of C5¹¹⁵. As expected, the incidence of overall breakthrough hemolysis was reduced, too. This dramatic half-life extension was made possible via engineering the epitope binding site within the variable domain of the heavy chain and the Fc-part of the original antibody. Two amino acid substitutions were introduced at each site¹⁶⁶. The substitutions in the Fc-part aimed at increasing recycling via the FcRn cycle¹⁶⁷, while the changes in the epitope binding region aimed at releasing the target protein C5 during FcRn cycling in the early endosome. The high affinity of eculizumab for C5 at both, the physiological pH of 7.4 and the slightly acidic pH of 6 (within the endosome) meant that FcRn recycled eculizumab:C5 complexes could exit the cell after their take up. This is also the reason why C5:eculizumab complexes accumulated in eculizumab treated patients resulting in higher-than-normal total C5 concentrations (which, however, were fully complexed by the antibody). Returning eculizumab:C5 complexes from the endosome into circulation does not help with maintaining effective drug concentrations. Sheridan et al¹⁶⁶ substituted two amino acids in the eculizumab complementarity-determining regions with histidine residues, which become increasingly protonated when the pH drops from 7 to 6 in the endosome. This introduces charged histidine residues within the complementarity-determining regions, which are not compatible with high affinity binding to C5. The result is that ravulizumab can be recycled without its C5 cargo. Free ravulizumab is returned into the circulation, while C5 is degraded within the endosome.

Figure 8. Evolution of complement therapeutics starting with the first approval of eculizumab for PNH in 2007.

Regulatory approved complement therapeutics are listed according to the timescale of their approval. Selected aspects that were noticed during therapy and/or were explained mechanistically and most likely had or have an influence on the development of the next generation anti-complement drugs are highlighted. This is not an exhaustive display of all developments. Pharmacokinetic (PK); Pharmacodynamic (PD).

Extravascular hemolysis represented another challenge that had to be tackled to improve therapy for PNH patients. As C5 inhibition alone does not stop the ongoing C3b deposition process, preclinical development focused on approaches that either were thought to completely inhibit all three initiation pathways of the complement cascade (e.g. compstatin derivatives) or that selectively inhibited the AP. With pegcetacoplan the former approach has recently entered the clinic as a therapy for PNH. The clinical trial demonstrated superiority to eculizumab in improving hematological parameters, which can be explained by the compstatin derivative controlling intravascular as well as extravascular hemolysis¹⁵⁶. The later approach of selectively inhibiting the AP appears especially attractive for any AP-mediated disease. Blocking solely the AP leaves the complement immune surveillance by the CP and LP mostly intact (but would still impair CP/LP activity to some extent due to the absence of C3b amplification by the AP).

TT30, a fusion protein of complement receptor 2 (CR2) and FH¹⁶⁸, and miniFH, a fusion protein of N- and C-terminal domains of FH¹³³, have demonstrated an impressive efficiency in completely protecting PNH erythrocytes (including protection from C3b deposition) from complement attack in vitro. A side-by-side in vitro comparison showed that the CR2-FH fusion approach was less active than the miniFH fusion¹⁶⁹. Both approaches follow the idea that targeting late stage opsonins like iC3b and C3dg can help in effectively protecting cell surfaces that constantly fix these opsonins to their surface¹⁷⁰. A second theme in common between TT30 and miniFH is the idea that the complement cascade should not be inhibited completely at one single step, which can have somewhat unpredictable consequences (see above). Therefore, instead of relying on a stoichiometric inhibitor such as an antibody or a peptide, miniFH was engineered to act like a natural convertase-directed regulator. TT30 works in the same fashion. The attractiveness of this mode of inhibiting the complement cascade is that: i) the cascade is not blocked completely (but stringently modulated) allowing some effector functions to take place; ii) the inhibitors only bind to activated complement components; and iii) the inhibitors are not consumed in stable 1:1 or 1:2 complexes as they are fully recycled due to their enzymatic-like activity (Figure 1). TT30 was previously evaluated in a phase 1 clinical trial for treatment of PNH but the trial has been terminated (NCT01335165)¹⁷¹. Different versions of miniFH are still under preclinical investigation and demonstrate promising activities in some murine models of complement-mediated diseases^172,173. Another strategy to stop extravascular hemolysis under anti-C5 therapy is to add a FD inhibitor as add-on therapy to the anti-C5 strategy, a strategy presently under clinical investigation (NCT04469465). Finally, with the stoichiometric peptide inhibitors of the C3 family a completely new class of complement inhibitors has evolved (Figure 8).

Evolution also occurred at another level: The C5 inhibitory approach has broadened in terms of indications and now includes the two autoantibody diseases, MG and NMOSD. The latest news is the approval of the C1s inhibitor sutimlimab, which represents a novelty in two ways. First, it is the first-in-class inhibitor that therapeutically address a complement initiation pathway. Second, it is the first complement inhibitor approved for CAD. Diversification through evolution has already gained momentum and considering the many new strategies in preclinical and clinical development (reviewed in ^46,174,175), one can certainly expect an exponential development of complement therapeutics in the not so distant future.

Two strategies are envisaged to be extremely fruitful for the next generation of complement inhibition: First, the introduction of cell- or tissue-targeted complement inhibitors into the clinic. The prominent inflammatory effector functions of the complement cascade require C5 activation, which in turn requires a surface densely deposited with C3b and/or C4b (Figure 4). Therefore, it seems attractive to locate complement inhibitors directly to the site where activation occurs resulting in less interference with the systemic immune surveillance by complement. Exciting indications for such approaches would certainly be organ transplantation, systemic lupus erythematosus, catastrophic antiphospholipid syndrome, C3G and AMD, among others.

Second, complement therapeutics not only entails inhibiting the complement system but also harnessing the cytolytic potential to clear unwanted cells. To this end, exciting structural studies of the antibody:C1 complex have paved the way for the development of antibodies that efficiently hexamerize on a target potently triggering the CP^77,176. This elegant strategy is expected to push the C3 activation cycle so much into overdrive that C3 regulation is simply overwhelmed (Figure 1C), allowing removal of altered/dangerous host cells. A related but orthogonal strategy utilizes engineered gain-of-function mutations in C2¹⁷⁷. Combining complement activating antibodies with gain-of-function versions of C2 boosts the complement effector functions of the complement activation antibodies. Whether physiological processes that rely on CP/LP activation, as for example the C3 opsonization of apoptotic bodies^141,142, may also be impacted substantially by the short term application of more active C2 molecules remains to be seen.

Finally, we have to keep in mind that there is extensive cross-talk between the complement system and other branches of the immune system²³, and there are undoubtedly complexities we do not yet understand. Recently, it has been demonstrated that C3 opsonized surfaces can trigger the upregulation of the checkpoint ligand PD-L1, underscoring complement’s role in helping to maintain tissue homeostasis¹⁷⁸. Two things are certain about the ‘complement future’. We will see more complement therapeutics in the clinic and unexpected findings will continue.