Complement Activation by Post‐Translationally Modified Proteins: Links to Chronic Inflammation and Autoimmunity

3.1. Post‐Translational Modifications

A wide range of naturally occurring post‐translationally‐induced protein alterations have been identified in physiological and in pathophysiological processes. The post‐translational modifications that are formed on mature proteins can range from enzyme‐mediated and transient to non‐enzyme‐mediated and irreversible. Enzyme‐mediated, transient modifications generally target one specific amino‐acid residue in a specific protein, leading to its activation/deactivation. They include for example, the kinase‐mediated phosphorylation of an enzyme or a receptor and are an indispensable part of normal physiology. An example is the (auto)phosphorylation of components of the insulin signaling cascade [6]. Other types of enzyme‐mediated modifications include methylation, glycosylation, and ubiquitinylation, each with distinct physiological consequences. Together, these enzyme‐mediated, transient modifications comprise > 90% of all reported PTMs [7]. Non‐enzyme‐mediated but reversible modifications are also seen. These include protein S‐sulfenylation, S‐nitrosylation or S‐sulfhydration upon oxidation of Cys, yielding unstable PTMs [7]. These modifications provide a read‐out of oxidative stress and play a role as redox switches [8]. In addition to these reversible protein modifications, there is also a range of broad and irreversible protein modifications. These include, for instance, the advanced glycation end‐products (AGEs) that are formed on multiple amino acid residues in multiple proteins in a state of prolonged hyperglycemia [9, 10]. In general, the formation of specific, transient PTMs is enzyme‐mediated while formation of broad‐spectrum and irreversible PTMs is non‐enzyme mediated. More than 700 different PTMs have currently been identified [7], highlighting the enormous complexity of protein regulation in biological systems.

Non‐enzyme mediated, irreversible PTMs are often, although not exclusively, found in long‐lived proteins. These modifications accumulate over time and are increasingly perceived to be markers of “protein molecular aging” and hence considered hallmarks of aging [6, 11, 12]. Their formation results from the modification of accessible amino acids by reactive metabolites (Figure 2). These reactive metabolites are primarily generated during metabolism and include, for example, methylglyoxal (MGO) produced during glycolysis [12], acetaldehyde (AA) generated in alcohol metabolism [13] and cyanate formed during urea metabolism [14]. In addition, reactive metabolites can arise under conditions of oxidative stress, for example, malondialdehyde (MDA), which is generated during lipid peroxidation of unsaturated lipids in cell membranes [15] (Figure 2). These highly reactive compounds can then modify accessible amino acids, thereby generating PTMs. MGO predominantly modifies mostly arginine (Arg) and lysine (Lys), resulting in the formation of multiple advanced glycation end products (AGEs). MDA and AA together can modify Lys to Malondialdehyde Acetaldehyde (MAA), while carbamylation of Lys by cyanate generates the non‐standard amino acid homocitrulline [15, 16, 17, 18, 19]. While many irreversible PTMs are generated independently of enzymes, one important exception is citrullination, which is the irreversible modification of Arg to citrulline, catalyzed by peptidylarginine deaminase (PAD), particularly in response to high local calcium levels [20, 21].

These aforementioned protein adducts or PTMs are generally considered to be chemically irreversible [22]. Removal of PTM proteins can occur via proteasome‐dependent degradation of the proteins whose integrity is affected by the PTMs with subsequent removal of the modified amino acids via the kidney [11, 23, 24]. As discussed below in more detail, circulating PTM‐proteins can also be recognized by complement and/or antibodies that will stimulate clearance via liver and spleen (Figure 3).

The formation of PTMs on proteins can have multiple consequences. Firstly, when the PTM is formed on an amino acid in the functional domain of a protein, it can affect the physiological function of that specific enzyme or regulatory protein. In addition, and likely more relevant for chronic diseases (discussed later), PTMs can interfere with normal physiology through several mechanisms including receptor binding and activation, induction of an antibody response, and via the formation of protein crosslinks. Given that the diversity of PTMs and their (patho)physiological consequences extend far beyond the scope of one review, we herein focus on a subset of irreversible, non‐enzyme mediated PTMs that have been implicated in human diseases that are characterized by a state of chronic low‐grade inflammation.

3.2. Post‐Translational Modifications and Disease

Several irreversible, non‐enzyme mediated PTMs including AGE, MAA, carbamylated proteins (CarP), and citrullinated proteins (CitP) have been implicated in a plethora of chronic diseases including, but not limited to, cardiovascular disease (CVD) [25, 26, 27, 28] kidney disease [29], liver disease [30, 31, 32], rheumatoid arthritis [33, 34], and neurodegenerative diseases [20]. Their association with these chronic diseases is not unexpected given the lifelong accumulation of these PTMs. Over the past two to three decades, several mechanisms have been proposed to explain how these PTMs may contribute to chronic low‐grade inflammation and related disease.

Ox‐LDL, a major culprit in the development and progression of atherosclerosis, carries multiple PTMs [35, 36, 37]. Binding of antibodies to ox‐LDL and other PTM tags them for subsequent (complement‐mediated) removal by scavenger cells. While this is basically a beneficial process, it can result in lipid overload of the scavenger cells. In the vascular wall, for example, LDL‐derived lipid accumulation in macrophages results in the formation of foam cells, a hallmark of atherosclerotic plaques.

The intracellular mitochondrially‐expressed C5aR1 has been proposed as a key regulator of the proinflammatory response in human atherosclerotic plaques [38], particularly in male mice. In the context of carbamylation, mass‐spectrometric analyses have been performed to identify the ‘carbamylome’ of the inflamed joint [34]. Carbamylated proteins were readily identified in the synovial fluid, in the synovial tissue and in the cartilaginous tissue. Most of the carbamylation in rheumatoid arthritis was found on the matrix molecules in the cartilage [34], a finding consistent with the concept that the half‐life of matrix molecules is much larger than that of circulating or cell bound protein. This prolonged residence time allows accumulation of PTMs over time.

The protein modifications AGE, MAA, and CarP all bind to the scavenger receptor SR1A on macrophages, eliciting an inflammatory response [39, 40, 41]. The binding of CitP to scavenger receptors and the resulting inflammatory response was enhanced by the simultaneous presence of MAA adducts [42]. AGEs can also activate the Receptor for AGEs (RAGE), which induces oxidative stress and an inflammatory response via, among others, activation of the Nuclear Factor‐kappaB pathway [43]. In addition, PTMs on specific proteins exert pro‐inflammatory effects; for example, the above‐described uptake of PTM‐modified‐LDL via SR‐A1 can contribute to the process of foam cell formation [44], the main initiating step for atherosclerosis. Moreover, citrullinated fibrinogen can activate macrophages via the TLR4 receptor, thereby contributing to an inflammatory response, although this is potentially only as part of an immune complex [45]. Citrullinated proteins have been identified as highly immunogenic, as illustrated by the pronounced presence of anti‐citrullinated protein antibodies (ACPA) in rheumatoid arthritis [46]. Notably, citrullinated proteins are not unique herein, as other PTMs also have been shown to induce an antibody response. Antibodies against AGEs, MAA, and CarP have been identified in patients with auto‐immune disease such as rheumatoid arthritis, but also in patients with CVD [47, 48] and Type 2 diabetes mellitus [49, 50] who were not diagnosed with an auto‐immune disorder.

It remains yet to be clarified to what extent these PTMs and the anti‐PTM antibodies in the circulation are risk factors via separate pathological pathways, whether they are part of a combined pathway, or whether one functions as a biomarker for the other.

3.3. Autoantibodies Against Post‐Translationally Modified Proteins

Autoantibodies directed against an endogenous protein are present in many autoimmune diseases. By binding to their target antigen, they can impair protein function, enhance protein function, or cause tissue damage by forming an immune complex and triggering effector functions by cells expressing Fc‐receptors or by activating complement. Over the years it has become clear that some autoantibodies bind specifically to the native protein structure, while other autoantibodies only bind to their target protein after these proteins have undergone a PTM [51].

The term anti‐modified protein antibodies (AMPA) is frequently used to describe the family of autoantibodies that recognize PTM‐proteins [52]. Anti‐PTM antibodies are not only seen in autoimmunity, but also in non‐autoimmune diseases [48, 53]. Several well‐known examples are present in rheumatoid arthritis wherein antibodies were identified that target citrullinated proteins (ACPA) [54]; antibodies that target carbamylated proteins (anti‐CarP) [55], and antibodies against Malonaldehyde Acetaldehyde modified proteins (anti‐MAA) [56]. In other conditions also other PTMs are targeted. For example, autoantibodies against acetylated proteins (AAPA) have been identified, not only in rheumatoid arthritis but also in SLE [57, 58], while antibodies against proteins with AGE‐adducts (anti‐AGEs) were originally described in diabetes [59].

In several diseases, patients can be positive for several anti‐PTM antibodies, hence targeting multiple distinct PTMs simultaneously. This allows clustering of individuals on the basis of their anti‐PTM antibody patterns [57, 58, 60]. The presence of anti‐PTM antibodies is generally associated with worse disease outcomes, for example, more joint damage in rheumatoid arthritis [55, 61, 62]. These anti‐PTM antibodies are reported to have the capacity to contribute to disease severity by triggering cellular activation through Fc‐receptor engagement [63] or by activating complement [64].

Despite our growing knowledge of the widespread presence of anti‐PTM antibodies, it is largely unknown why and how individuals make antibodies against PTM‐proteins and how their relationships are interconnected. Importantly, many of these autoantibodies can already be detected in the circulation of individuals many years prior to the onset of disease [65, 66, 67]. And while there is a degree of cross‐reactivity between anti‐PTM antibodies binding different PTMs [68, 69], clear differences are also observed with individuals being only positive for one anti‐PTM reactivity but negative for others [55, 58, 60]. In experimental animal models, it was shown that immunization with one PTM may trigger antibodies against multiple PTMs [52]. In addition, there may be a certain evolution from one to another anti‐PTM antibody depending on the order of exposures to different PTMs [70, 71].

A key feature that we highlight in this review is the observation that PTM‐proteins can also activate complement [72, 73, 74, 75, 76, 77] (Figure 3 and section 4.1). This provides additional immunostimulatory capacity to the PTM‐protein, thereby enhancing the chance of breaking tolerance. Complement activation not only results in the release of anaphylatoxins C3a and C5a, providing an inflammatory environment resulting in immune cell recruitment and antigen presentation, it also provides C3b/iC3b mediated opsonization. Moreover, C3d, the degradation product of C3b/iC3, can bind to Complement Receptor 2 present on both B cells and on follicular dendritic cells. This is important for antigen presentation and is reported to lower the threshold for breaking tolerance [78].

3.4. Technical Pitfalls and Considerations in Identification and Quantification of PTMs and Their Antibodies

The non‐enzymatic PTMs that are the focus of this review are modifications of single amino acids by small reactive metabolites and the resulting changes that are introduced will therefore also generally be small. To identify the non‐enzymatically post‐translationally modified amino acids on proteins, different techniques are available including mass‐spectrometry, ELISA and Western blot. It is important to realize that all these techniques have their benefits and pitfalls when identifying a type of PTM, or the effect of PTMs, on protein function [79]. While not the focus of this review, we highlight here several analytical considerations that may be useful in designing experiments to study PTM‐proteins and their interactions and for the interpretation of data in literature.

First, when producing PTM‐proteins it is important to simultaneously produce the non‐modified version of the same protein using an approach similar to a sham treatment in animal studies. This implies that the unmodified control protein should undergo a mock modification, in which all the reaction steps, the handling, and the incubations are identical except for the reagent essential for the modification. This approach ensures that observed effects can be attributed to the PTM itself rather than to differences in protein handling or quality. Second, to allow generalizable conclusions it is important to introduce the PTM to several different proteins because the protein structure and the amino acids surrounding the modified amino acid may influence its binding to anti‐PTM antibodies and/or its interaction with other proteins [77]. Third, when studying complement activation by PTM‐proteins in serum it is essential to also use a control condition consisting of that same human serum but after heat inactivation as a control in parallel experiments. This allows the researcher to evaluate if the presence of complement on PTM‐proteins is the consequence of complement activation or due to passive binding of complement proteins. Fourth, although the identification and quantification by mass‐spectrometry can be very precise, it is important to realize that several PTMs induce a shift in the molecular mass that is actually very similar to mass shifts induced by other modifications. For instance, the identification and quantification of carbamylation is relatively straightforward, whereas the evaluation of citrullination is much more difficult as this very small mass shift closely resembles several other modifications [79]. To be sure that it is actually a citrullinated peptide that is identified by mass‐spectrometry, both modified and unmodified peptides are required to confirm that the spectrum indeed matches the modified peptide. Moreover, some of the PTMs that are stable in their biological, physiological environment are not stable under conditions of sample preparation for mass spectrometry. Fifth, for many of the PTMs commercial antibodies are available, but many of these antibodies cross‐react to other modifications [79]. When ELISA or Western blot are used to identify or quantify PTM proteins, the use of anti‐PTM antibodies to detect the PTM proteins is unavoidable. In this context it is important to realize that the specificity of the detection assay is limited by the specificity of the anti‐PTM antibody that is used for the detection of the PTMs. For example, to enhance the sensitivity of the Western blot‐based identification of citrullinated proteins, the anti‐modified citrulline ‐Senshu method was developed, in which a chemical reaction on the blot combined with specific detection antibodies for this in situ‐modified citrulline enhances the sensitivity of the assay [80]. However, this comes at a cost since this method does not discriminate between citrulline and homocitrulline [81], posing substantial limitations in the interpretation of the data [79]. Sixth, in serum and plasma there is often a simultaneous presence of PTM‐protein and autoantibodies against these PTMs. The presence of autoantibodies against PTMs may interfere with the detection of those PTM, and vice versa, particularly in antibody‐based assays such as ELISA and Western blot. Importantly, the nature and ratio of these circulating PTMs and their antibodies differs according to different disease states. Use of multiple controls and, where possible, confirmation with different detection techniques are advised.

Taken together, these considerations highlight that it is very well possible to quantify PTM‐proteins and to study their interactions with naturally‐occurring autoantibodies and with other proteins in a reliable way, but only with carefully designed experiments and with cautious interpretation of published data.

4.1. Activation of Complement by Post‐Translationally Modified Proteins

Several studies have now described the capacity of PTM‐proteins to activate the complement system directly [72, 73, 74, 75, 76, 77] (Figure 3). Complement activation upon the direct binding of PTM‐proteins differs from complement activation driven by anti‐PTM autoantibodies binding to their PTM‐protein targets [64].

4.2. Post‐Translational Modifications on Complement Proteins

Besides activation of complement via the above‐mentioned binding of complement recognition molecules to PTM‐proteins, PTMs can also be formed on the complement factors and complement regulators themselves. Human serum complement components, including C3, can be carbamylated in vitro [99], indicating that in their native conformation at least part of their lysine residues are accessible for post‐translational modification. Note that the overrepresentation of information on C3 in this literature on PTMs on complement is likely due to its relatively high concentration in plasma which facilitates the detection and analysis of PTMs that are formed on this molecule. Another example is glycation (i.e., the formation of AGE‐adducts) on CD59, which is the complement inhibitor that prevents C5b‐9 assembly. Glycation of CD59 has been shown to impair its protective function, as reflected in increased deposition of C5b‐9 in the kidneys in diabetes patients [100]. Similarly, glycation also alters the activity of decay accelerating factor (DAF/CD55) in persons with diabetes [101]. Moreover, hyperglycemia‐induced AGE‐adducts may play a role in the C3‐mediated resistance to fibrinolysis of blood clots in diabetes patients [102, 103].

Taken together, these data show that the formation of PTMs on complement proteins, particularly when they are generated in the active sites of complement regulators, can disturb the delicate balance between complement activation and inhibition.

5.1. Complement in Adipose Tissue

Adipose tissue is nowadays accepted as a key regulator of metabolism and homeostasis. Oxidative stress, particularly in the visceral and other ectopic depots, is a hallmark of dysregulated adipose tissue metabolism [104]. Nutrient excess may overcharge the intrinsic metabolic capacity of adipocytes thereby inducing endoplasmic reticulum (ER) stress, mitochondrial dysfunction, and oxidative stress [105]. The oxidative stress will be further aggravated through the presence of the damaged mitochondria [106]. Moreover, hypoxia that occurs in the enlarged adipose tissue depots in obesity may confer intracellular stress, thereby eliciting an inflammatory response [107]. The adipose tissue produces an array of complement components, particularly components of the classical and the alternative activation pathways. Obese individuals have increased circulating concentrations of C3a and C5a which elicit an inflammatory response [108]. Complement activation in dysfunctional adipose tissue links obesity‐induced systemic derangements including low‐grade inflammation, insulin resistance, and dyslipidemia [109]. The liver can also contribute to the obesity‐induced increase in circulating complement. The liver is a major source of systemically present complement factors, including the central component C3. The liver is also directly exposed to exogenous (dietary) fatty acids from the intestine and endogenous fatty acids from the visceral adipose tissue which can overload the hepatocytes leading to hepatic fat accumulation and inflammation. Complement production in the liver is increased in conditions where fat accumulates in the liver such as metabolic‐associated fatty liver disease (MAFLD) [110], previously referred to as non‐alcoholic (NA) FLD [111].

Control and function of the complement system depend on a delicate balance [5] which, once disturbed, will perpetuate to affect different cardiometabolic systems. There are several reasons why the intrinsic balance of the complement system may be particularly vulnerable to obesity‐induced metabolic changes. First, many complement components are synthesized in adipocytes and in the stromal cells of the adipose tissue; second, their expression and production are altered in obese adipose tissue; third, also plasma concentrations of multiple complement components are increased in obesity, although it is not known if this is via direct production and secretion of complement by the adipose tissue, or via indirect pathways (e.g., the liver). Notably, adipose tissue, particularly in obese individuals, is not a single depot but is partly deposited as ectopic depots adjacent to many organs including, but not limited to, perivascular, epicardial, perirenal as well as intrahepatic and intramuscular fat depots [112]. Concentrations of certain complement factors are associated with fat accumulation in general, while others are more strongly associated with these visceral and ectopic fat depots [109]. Dysfunctional adipose tissue metabolism can also affect body composition. In rheumatoid arthritis this is reflected in a relative increase in ectopic fat combined with a decrease in muscle mass, reminiscent of what is seen in aging [113]. Pericardial fat in rheumatoid arthritis is associated with cardiovascular disease [114]. Moreover, in experimental mouse models, C3 and C4 contributed to aorta stiffness through thiol‐ether binding to collagen and elastin in the vascular wall [115]. These authors pointed at the perivascular adipose tissue (PVAT) depot that surrounds the aorta as the most likely source of these complement components and suggested that PVAT‐derived complement may contribute to “outward‐in” remodeling of the vasculature. Likewise, PVAT‐derived C3 was shown to modulate the migration and differentiation of adventitial fibroblasts [116]. Complement has also been implicated in remodeling of the non‐vascular extracellular matrix in a range of other organs including the eye [117], the lung [118] and the kidney [119].

5.2. Complement and the Inflammasome

Chronic low‐grade inflammation lies at the core of multiple diseases including cardiometabolic, neurodegenerative and autoimmune disease and their comorbidities. Activation of the inflammasome is a major pathway via which complement activation contributes to inflammation. Complement and the inflammasome are the two major danger‐sensing effector systems of innate immunity. An extensive overview of their key roles in the inflammatory response was recently published [120]. Via bidirectional crosstalk, complement and inflammasome activation are mutual amplifiers: An activated complement can sensitize and induce inflammasome activation while the effector molecules of the inflammasome, for example, IL‐1β, induce the production of complement [120]. C3a and C5a [121, 122] and sublytic C5b‐9 [123] can all activate the NLRP3 inflammasome in macrophages thereby contributing to an inflammatory response and ROS production. Complement activation has been implicated in the initiation of inflammation via activation (e.g., via binding of C3a and C5a to their receptors) and inhibition (e.g., via CD46) of the NLRP3 inflammasome [121, 122, 124, 125]. Sublytic C5b‐9/MAC on immune cells is another complement‐derived inflammasome‐activating signal [123, 126]. The classical pathway initiator C1q has been implicated in promoting as well as mitigating inflammasome activity, depending on the context. C1q was reported to contribute to inflammasome activation in disease states such as rheumatoid arthritis [127], Alzheimer's dementia [128] and stroke [129, 130]. On the other hand it prevents inflammasome activation through its role in the clearance of apoptotic cells and the concomitant induction of an anti‐inflammatory response in macrophages [131], as was seen in the early stages of atherosclerosis [132]. In obesity and MAFLD, small amounts of gut‐derived lipopolysaccharide (LPS) can sustain a state of low‐grade inflammation [133], at least in part via non‐canonical inflammasome activation, with direct activation of caspase [134]. Microcrystals and extracellular ATP can activate both complement and the inflammasome [135, 136]. Activation of the recently identified complosome, the intracellular complement system, can also contribute to inflammasome response via directly activating the inflammasome and via lowering its activation threshold [136, 137]. The joint stimuli for activation of complement and the inflammasome include danger‐associated molecular patterns (DAMPS) and pathogen‐associated molecular patterns (PAMPS), mitochondrial damage, cellular stress and particle matter [120]. Notably, these DAMPS will include PTM. Cell stress can be a consequence of PTMs on cellular protein and PTMs can also induce the accumulation of particulate matter such as amyloid Aβ in Alzheimer's dementia. The receptor for AGE (RAGE) is a key pattern recognition receptor that induces the inflammasome [138]. Note that the cycle of amplification between complement and the inflammasome will be more prominent in the context of a relative shortage of functional complement inhibitors.

5.3. Complement in Chronic Low‐Grade Inflammation: Focus on Cardiometabolic Disease, Neurodegeneration and Rheumatoid Arthritis

Given their close co‐existence and the intricate relationship between complement activation and inflammation, it is hard to determine the exact order of events for their effects on the development of low‐grade inflammation‐related diseases. We and others have shown that circulating complement activation products such as C3a, C5a, and soluble C5b‐9 are associated with systemic low‐grade inflammation [109]. Early experimental studies in mouse strains that were genetically deficient for the receptors C3a and C5a support that complement activation is essential for, and precedes, the development of high fat diet‐induced adipose tissue inflammation and insulin resistance [139, 140]. An early study by Muscari et al. identified systemic C3 as a risk factor for incident CVD [141]. In line with this, higher plasma concentrations of C3 predicted future weight gain in a large observational human study that included only male participants [142], and we showed that C3 (and C4) concentrations were associated with incident metabolic syndrome [143]. C3 is also associated with the development of incident Type 2 diabetes mellitus, independent of the degree of low‐grade inflammation [144]. At present, possibilities for pharmacological modulation of complement activation on diseases related to low‐grade inflammation are limited given the high cost of the currently available compounds and high chance of unwanted side effects due to systemic chronic complement inhibition.

C1q and C1q‐induced phagocytosis are considered anti‐inflammatory, which is illustrated by the observation that C1q limits inflammasome activation in macrophages [131]. These effects of C1q are consistent with the finding that C1q deficiency is associated with autoimmunity [145]. However, while generally considered to induce an anti‐inflammatory state, C1q can also provoke a pro‐inflammatory response when activating the downstream effectors of the complement cascade including the amplification loop and the terminal pathway. C1q was shown to contribute to ongoing local inflammation in Alzheimer's dementia and atherosclerosis, among others by binding to oxidized (i.e., post‐translationally modified) LDL and steering the attraction of immune cells [146]. This classical pathway‐mediated proinflammatory process was ameliorated by the binding of lipoprotein apoE to C1q [146]. C1q can also, via its globular head domains, bind to the receptor for AGEs (RAGE) thereby enhancing C1q‐mediated phagocytosis [147], reflecting an effect of C1q that is independent of the activation of complement. RAGE is a multiligand receptor and activation of RAGE on adipocytes and macrophages which, via its canonical ligand, AGE, induces a proinflammatory response [9]. Binding of another RAGE‐ligand, that is, the high mobility group box 1 protein (HMGB1) which is an intracellular danger system released by apoptotic cells to RAGE on macrophages, also induces a proinflammatory response and an M1‐like macrophage phenotype [148]. Notably, this response is modulated by C1q wherein the combined binding of C1q and HMGB1 to RAGE on monocytes induces an anti‐inflammatory M2‐like phenotype [149]. More specifically, binding of HMGB1‐only, in the absence of C1q, results in RAGE‐mediated leukotriene production, while binding of HMGB1 plus C1q induces the production of the specialized pro‐resolving lipid mediators lipoxin A4, resolvin D1, and resolvin D2 through a RAGE‐ and LAIR‐1‐dependent pathway [150]. It remains to be evaluated if C1q can also modulate the effect of AGE‐induced RAGE signaling.

These dual and opposing effects of C1q on phagocytosis and inflammation may underlie the U‐shaped relationship of systemic C1q levels with incident CVD that we observed in a human cohort [151]. Notably, a similar U‐shaped relationship was reported for the association of MBL with carotid intima‐media thickness in patients with rheumatoid arthritis [152]. It appears that—at least locally, at the tissue level—there is an optimum for complement activation, balancing the protective effects related to the removal of dead cells, debris, and protein depositions with the detrimental effects that likely result from ‘too pronounced’ local complement activation and the subsequent inflammatory response. Thus far, such U‐shaped relationships were not reported for the activation of the alternative pathway in human studies. In fact, activation of the alternative pathway was generally associated with the development of disease. Factor D, the rate limiting enzyme in the activation of the alternative pathway, was in a large prospective cohort (men only) shown to be associated with stroke [153] but not with coronary heart disease [154]. In a cross‐sectional analysis of an unrelated cohort, we showed that factor D is positively associated with low‐grade inflammation and endothelial dysfunction which translated to more CVD, manifesting primarily as cerebral vascular disease in the male participants [155]. This difference between the initiators of the classical and the lectin pathway versus the activation of the alternative pathway makes sense given the prominent roles of the former in direct opsonization and phagocytosis and the generation of the prominent anaphylatoxin C3a in the latter.

In experimental models, antagonists of C3aR and the C5aR1 attenuate development of dementia and atherosclerosis in suitable models [156, 157]. Some complement‐mediated effects on disease may be sex‐specific, as illustrated by the results of an experimental mouse model showing that deletion of C5aR1 results in a reduction of atherosclerosis in male, but not in female mice [158]. The cholesterol crystals in the atherosclerotic plaques are coated with complement, suggesting that they are primed for complement‐mediated phagocytosis while concomitant generation of anaphylatoxins triggers the (local) inflammatory response [122, 159]. The observation that activation of C3, C5, and the terminal pathway was – in experimental mouse models‐ consistently shown to be proatherogenic suggests that the observed protective effects of the classical and the lectin pathway in atherosclerotic disease lie in their proximal steps.

An emerging comorbidity of (obesity‐associated) cardiometabolic diseases such as CVD and Type 2 diabetes is neurodegeneration, reflected in cognitive decline and dementia, which may at least partly be attributable to systemic and local inflammation, including neuroinflammation, but also to underlying vascular pathology, as in vascular dementia. Complement may contribute to these pathological processes via local production and activation of complement and via induction of endothelial injury [160]. Local complement activation was reported in the key lesions involved in the pathology of these diseases, that is, atherosclerotic plaques in the vascular wall [161], amyloid Aβ plaques in the brain [162] and islet amyloid polypeptide (IAPP) in the pancreatic beta cells [163]. Accumulation of intra‐ an extracellular debris can act as a damage signal which is detected by complement and subsequently triggers an enhanced inflammatory signal. Experimental mouse models suggested a role for the C3‐C3a‐C3aR axis in Aβ‐induced dementia [164], as well as in disruption of blood–brain barrier integrity [165]. Moreover, concentrations of C3 in plasma and cerebrospinal fluid are higher in patients with Alzheimer's Dementia than in controls [166] and plasma C3 predicts CVD in Alzheimer's dementia [167]. In contrast, low as opposed to high C3 concentrations also associated with a higher risk to develop Alzheimer's dementia, especially in carriers of the E4 allele [168]. Likewise, there is experimental data suggesting the complete absence of full complement activation through C3 deficiency is detrimental as well, because C3 deficiency resulted in accelerated disease in a dementia‐prone [169] and in an atherosclerosis‐prone [170] mouse model. These latter effects might be related to the proposed role of complement in neuroplasticity and repair in normal neuronal development [171].

Vascular calcification is yet another disease state that is characterized by accumulation of insoluble deposits. Prevalence of vascular calcification increases with age, is particularly high in chronic kidney disease [172] and is also a common comorbidity in rheumatoid arthritis [173]. In the process of vascular calcification, calcium is deposited in the vascular wall, thereby contributing to the stiffening of the arteries. Vascular stiffness and vascular calcification are multifactorial. An array of complement factors, indicative of overall complement involvement and activation, was identified in the calcified microdomains in the affected arteries [174, 175], and the plasma concentration of C3a was also associated with abdominal and coronary calcification [176, 177]. It is not yet clear if, as has been suggested, complement actively and directly contributes to the calcification process via osteogenic transdifferentiation of vascular smooth muscle cells [178] or rather that complement is activated on the surface of the hydroxyapatite crystals, similar to what is seen for cholesterol crystals in the atherosclerotic plaque [122].

5.4. Complement in Healthy and Unhealthy Aging

With increasing age, the balance in complement activation shifts towards a state of less rigorous checkpoint control by complement regulators and consequently towards a state of continuous ‘low‐grade’ complement activation that contributes to the process of ‘inflammaging’ [171]. In the process of healthy aging, such a state of mildly enhanced complement activation may boost the immune surveillance role of the complement system in defense and repair, which might be beneficial in the prevention of aging‐related diseases [179]. Notwithstanding, complement activation has extensively been implicated in progression of diseases of the elderly. C1q increases strongly with increasing age and complement activation via the classical pathway has been suggested to contribute to sarcopenia [180], a condition that is also common in rheumatoid arthritis [181]. In experimental models it was shown that during aging, chronic complement activation and signaling via C3aR contributes to deterioration of the integrity of the blood–brain barrier [165] and signaling via C5aR1 contributes to neuroinflammation and progression of cognitive decline [156]. At the same time, complete absence of C3 in animal models of Alzheimer's dementia also associated with worse neurodegeneration and more Aβ amyloid accumulation, suggesting compromised clearance of amyloid deposits [169].

Taken together, there are some positive effects of complement and complement activation that, when preserved, will protect from development of disease during healthy aging. On the other hand, tipping of this balance towards the non‐beneficial effects of advanced complement activation likely results in a state of ‘unhealthy and accelerated aging’ and to the development of (aging‐related) chronic diseases.

6.1. Intervention Through Lifestyle and the Environment

Accumulation of PTMs during the course of life is, among others, driven by a person's life‐long environmental, lifestyle and biological exposome. The environment contributes to PTM formation for example, through air pollution [182], while lifestyle components such as exercise, diet and alcohol intake, or smoking habits also have profound effects on PTM formation. They thereby contribute to the biological exposome which is further shaped by genetic and epigenetic factors.

Oxidative stress is a common consequence of an unfavorable environmental, lifestyle and biological exposome. It is characterized by the presence of reactive oxygen species (ROS) which induce the generation of reactive carbonyl species, one of which is MDA [15]. Moreover, intracellular metabolism of glucose (glycolysis) generates MGO as a by‐product [12]. MDA and MGO are highly‐reactive metabolites that will readily engage in the generation of MAA and AGE, respectively. This suggests that reducing the bioavailability of these highly‐reactive catalysts of PTM formation is an attractive target to control the production of PTMs. One potential approach to reduce the bioavailability of MDA and MGO is supplementation with dicarbonyl scavengers such as the vitamin B6 vitamer pyridoxamine [12, 183] or the flavonoid polyphenol and antioxidant quercetin [184, 185]. Additional, less focused approaches are regular moderate exercise and a healthy diet such as the Mediterranean diet, both of which may help reduce the total oxidative stress burden [25, 186, 187]. Moderate‐to‐high alcohol consumption can also add to oxidative stress and detoxification of alcohol in the liver generates AA, which combines with MDA to generate MAA. This makes reduction of alcohol consumption an attractive potential strategy to reduce MAA‐adducts. Moreover, cigarette smoke has been identified as a direct source of AA [188], which makes smoking cessation an opportunity to reduce MAA formation in smokers. Also for carbamylation, smoking is reported as an important contributor, so smoking cessation will be beneficial to reduce the exposure to many PTMs [19].

The dietary measures that are available to reduce urea in the circulation to prevent the formation of carbamylated proteins basically boil down to the general recommendations for management of kidney function including optimal hydration, moderate intake of protein and salt, and high fiber intake. AGEs are present in foods, particularly in (highly) processed foods [189]. To prevent the accumulation of AGEs, an additional strategy lies in the prevention of their entrance into the body. Herein, it must be noted that it is not yet known if the biological effects of exogenous (i.e., ingested via the diet) and endogenous (i.e., generated inside the body) AGEs are comparable. Notwithstanding, a recent meta‐analysis of 14 human interventions with diets high and low in AGEs showed, despite high heterogeneity, an overall detrimental effect of dietary AGEs on metabolic qualifiers of Type 2 diabetes mellitus that is, fasting glucose, fasting insulin, HOMA‐IR [190]. Notably and perhaps counterintuitively, recent observational studies on the effect of MGO in the diet showed a beneficial effect of higher MGO intake on incident Type 2 diabetes mellitus [191] and CVD [192]. The mechanism for this has not yet been elucidated but might be related to hitherto unanticipated beneficial effects of MGO on the gut microbiome.

Taken together it appears that for mitigating the accumulation of non‐enzymatic irreversible PTMs, following the general advice for a healthy lifestyle as recommended by the WHO [193] is a sensible approach. Whether the addition of specific compounds such as scavengers of PTM precursors will be of added value, particularly in persons at a higher risk of developing chronic non‐communicable diseases, needs further evaluation in future randomized controlled trials.

6.2. Medical Interventions—Towards Therapeutic Antibodies

The treatment of autoimmune diseases, including those characterized by the presence of AMPA, is largely based on non‐specific systemic immunosuppression. This type of treatment does alleviate the symptoms for the patients, reduces inflammation and limits organ damage, yet rarely leads to cure, and importantly for this review, it does not lead to a reduction in the disease‐provoking AMPA [194]. Even B cell‐targeted therapies, although effective in for example rheumatoid arthritis, do not lead to disappearance of AMPA [195]. Other treatments, such as plasmapheresis, may give a temporal drop in AMPA levels but do not address their production and hence AMPA will likely return after treatment. Novel therapies, derived from the oncology field, use Chimeric Antigen Receptor T cells (CAR‐T cells) to kill all the B cells in a patient [196]. This treatment was successfully tested in patients suffering from severe treatment refractory SLE and is now tested also in many other rheumatic conditions. Following systemic B cell depletion in these SLE patients, there was a complete disappearance of autoantibodies that also did not reappear once new B cells were generated [196]. Also, the symptoms of these individuals disappeared, indicating a state of cure in severe refractory SLE [196]. Because this treatment eliminated all autoantibodies in these SLE patients, it is likely that also AMPA were reduced, but so far no papers specifically reported on this. At present, only a few reports on treatment with CAR‐T cells of rheumatoid arthritis patients who were unresponsive to conventional treatment showed clear disappearance of ACPA [197, 198]. For other AMPA, this has not yet been studied, but it is conceivable that all AMPA will respond similarly to this CAR‐T treatment. Clearly, broad application of CAR‐T cell therapy is currently still limited because of the invasiveness of this intervention, with serious risks, side effects and major costs.

There are currently around 14 inhibitors on the market that are directed against 8 different complement targets [199]. These drugs are mainly used in rare diseases that are not antibody‐mediated, such as atypical Hemolytic Uremic Syndrome (aHUS) and Paroxysmal Nocturnal Hematuria (PNH). Only two of the currently approved complement inhibitory drugs are used in autoantibody driven diseases like Cold Agglutinin Disease and ANCA vasculitis [1]. However, these diseases are driven by antibodies against erythrocytes and neutrophil proteins, respectively, and are not diseases typically driven by PTMs or by anti‐PTM antibodies. Systemic complement inhibition is currently also tested in several trials in SLE, and some small trials have been performed in rheumatoid arthritis, but so far with disappointing results, in combination with unfavorable increase in the risk of infections [200].

To enhance complement inhibition while mitigating the unwanted increase in infections, several strategies are being developed to move away from systemically acting complement inhibitors towards targeted local complement inhibition [201, 202]. These approaches can roughly be divided into 2 lines of complement inhibitors. Firstly, an approach in which the complement inhibitors localize to tissue‐bound complement activation fragments, so the place where complement has already been activated, where they can inhibit further downstream activation [201, 203]. And secondly an approach in which complement inhibitors are specifically targeted to specific tissues or antigens where they can inhibit ongoing complement activation and prevent complement activation from occurring [202]. The latter approach has now been used to bring complement inhibitors specifically to PTM‐proteins. One line of research is focused on targeting complement inhibitors specifically to ischemia‐related neo‐antigens, as reviewed by Tomlinson [204]. The other line of research is focused on the use of bi‐specific antibodies to bring endogenous complement inhibitory proteins to specific PTM‐modified tissues or PTM‐proteins [96, 202, 205]. In these bispecific antibodies one arm of the antibody binds to a disease‐related tissue antigen while the other recruits one of the circulating endogenous complement inhibitors, Factor H or C4b‐binding protein (C4BP) [202]. As a proof‐of‐concept, the first bi‐specific antibodies were directed against model antigens such as dinitrophenyl (DNP) and biotin, but recently the first PTM‐reactive bi‐specific antibody has been made based on an anti‐CarP antibody, targeting carbamylated proteins [96]. These in vitro data indicate that it is indeed possible to bring the complement inhibitors Factor H or C4BP in close proximity to carbamylated proteins and in addition to inhibit local complement activation that was induced by anti‐CarP autoantibodies derived from rheumatoid arthritis patients [96]. These approaches to achieve local targeted complement inhibition have the potential to allow effective inhibition at the site where pathological complement activation is taking place while leaving the systemic complement system intact to fight infections elsewhere in the body [202]. This results in a much more favorable risk profile, and likely reduced costs, thus allowing complement inhibitory treatment in a wider set of clinical conditions. These bi‐specific antibodies may also allow for interventions in conditions where PTM‐driven complement activation contributes to chronic low‐level inflammation, for example, by targeting the endothelium. The modular nature of bispecific antibodies allows exchange of the anti‐CarP targeting arm for another anti‐PTM targeting arm. In addition, because of the cross‐reactive nature of some anti‐PTM antibodies, it may be possible to design a bi‐specific antibody that will target more than one PTM.

Collectively, changing lifestyle and environmental factors will reduce the generation of new PTM‐proteins and this reduction may be further enhanced by supplementation with several types of scavengers. In addition, targeted therapeutic interventions may reduce local complement‐driven inflammation that is induced by PTMs and dampen the vicious circle of chronic low‐grade inflammation as a driver of cardiometabolic and autoimmune disease and the generation of additional PTMs.