Genetic variability shapes the alternative pathway complement activity and predisposition to complement‐related diseases

Summary The implementation of next‐generation sequencing technologies has provided a sharp picture of the genetic variability in the components and regulators of the alternative pathway (AP) of the complement system and has revealed the association of many AP variants with different rare and common diseases. An important finding that has emerged from these analyses is that each of these complement‐related diseases associate with genetic variants altering specific aspects of the activation and regulation of the AP. These genotype–phenotype correlations have provided valuable insights into their pathogenic mechanisms with important diagnostic and therapeutic implications. While genetic variants in coding regions and structural variants are reasonably well characterized and occasionally have been instrumental to uncover unknown features of the complement proteins, data about complement expressed quantitative trait loci are still very limited. A crucial task for future studies will be to identify these quantitative variations and to determine their impact in the overall activity of the AP. This is fundamental as it is now clear that the consequences of genetic variants in the AP are additive and that susceptibility or resistance to disease is the result of specific combinations of genetic variants in different complement components and regulators (“complotypes”).


| B RIEF INTRODUC TI ON TO THE AMPLIFI C ATI ON NATURE OF THE AP
The complement system is a fundamental part of our innate immunity playing an essential role to fight pathogens and remove immune complexes and cell debris. Complement discriminates between self-components and pathogens, tagging the latter for elim-

I N V I T E D R E V I E W Genetic variability shapes the alternative pathway complement activity and predisposition to complement-related diseases Summary
The implementation of next-generation sequencing technologies has provided a sharp picture of the genetic variability in the components and regulators of the alternative pathway (AP) of the complement system and has revealed the association of many AP variants with different rare and common diseases. An important finding that has emerged from these analyses is that each of these complement-related diseases associate with genetic variants altering specific aspects of the activation and regulation of the AP. These genotype-phenotype correlations have provided valuable insights into their pathogenic mechanisms with important diagnostic and therapeutic implications. While genetic variants in coding regions and structural variants are reasonably well characterized and occasionally have been instrumental to uncover unknown features of the complement proteins, data about complement expressed quantitative trait loci are still very limited. A crucial task for future studies will be to identify these quantitative variations and to determine their impact in the overall activity of the AP. This is fundamental as it is now clear that the consequences of genetic variants in the AP are additive and that susceptibility or resistance to disease is the result of specific combinations of genetic variants in different complement components and regulators ("complotypes").

K E Y W O R D S
complement, genetics, susceptibility to disease This article is part of a series of reviews covering The Alternative Pathway or Amplification Loop of Complement appearing in Volume 313 of Immunological Reviews. that cleave C3 to generate C3b. This convertase-generated C3b can form more AP C3-convertase, providing the AP with the capacity to amplify exponentially. As a result, C3b and other C3-activated molecules (iC3b and C3dg) cluster in high amounts around the surfacebound C3-convertase providing the ligands for the complement receptor (CR)-mediated phagocytosis and, eventually, driving the activation of C5, which triggers inflammation and initiate the formation of the lytic membrane attack complex. 1,2 The lack of specificity and the amplification nature of the AP are both an advantage and a danger that requires strong regulation by a set of control proteins that collectively avoid damage to host cells and prevent the consumption of the AP components. These control mechanisms discriminate between host cells and pathogens so that the activity of complement in fighting microorganisms and removing cellular debris is not compromised. [3][4][5][6][7] In this review, I will describe the extension of AP genetic variability in normal and disease populations. I will discuss how genetic variants in components and regulators, determining differences in their activity and concentration, influence the overall activity of the AP, which results in either increased risk or protection from specific diseases. I will illustrate how the identification of disease-associated genetic variants in AP components and, eventually, their structural and functional characterization has been instrumental to improve our molecular understanding of how AP dysregulation contributes to disease. I will describe the additive effect of AP genetic variants and how they impact on the disease risk. And, finally, I will comment on challenges involved in the identification and classification of the AP genetic variants.

| G ENE TI C VARIAB ILIT Y IN G ENE S EN COD ING AP PROTEIN S IN NORMAL P OPUL ATIONS
In recent years, the application of next-generation sequencing (NGS) techniques to the exome sequencing of tens of thousands of normal individuals belonging to different ethnic groups has provided invaluable information of the genetic variability in all genes of the human genome. Focusing on complement, this information reveals that variability in the genes encoding proteins of the AP in the normal population, as determined by population genetic resources such as the Genome Aggregation Database (gnomAD), 8 is considerable. Table 1  The numbers that result from the classification of these FH variants as pathogenic or benign based on prediction algorithms like the Combined Annotation Dependent Depletion (CADD) method (https:// cadd.gs.washi ngton.edu/info) 9,10 are equally important. For FH, 47.1% and 47.5% of the variants with MAF < 1% and MAF < 0.1%, respectively, are predicted as pathogenic (CADD PHRED C-score >15). This translates into 1.79% and 0.8% of individuals in the normal population being heterozygote for a potential pathogenic variant in FH with MAF < 1% and MAF < 0.1%, respectively ( Table 1). As similar results are obtained for the other AP genes, we can summarize by saying that genetic variability in the AP is such that approximately 12% of people in the normal populations are heterozygous for a potential pathogenic variant with MAF < 0.1% in at least one AP component or regulator.
These elevated figures may look paradoxical considering that most diseases associated with the AP genes variants are very rare diseases.
However, as I will describe in this review, the rarity of these diseases is mainly determined by their association with particular genetic variants causing specific functional alterations in a complement protein, by the concurrence of a particular set of genetic variations in different complement genes and, also important, by the fact that disease-associated AP variants are predisposition factors, which implies that the occurrence of the disease in carriers of these pathogenic variants requires non-genetic or environmental triggers.
This picture of genetic variability in AP genes in normal populations is nevertheless incomplete because it relates only to exome sequencing and it does not include the intronic and intergenic regions where we know there are sequence variations that influence the expression of the complement proteins. These expression quantitative trait loci (eQTL) added to those affecting the activity of the AP proteins shape both the overall activity of the AP and the predisposition to complement-related diseases.

| R ARE G ENE TI C VARIANTS IN THE AP AND PRED IS P OS ITI ON TO D IS E A S E
We have known for long time that genetic variants in the AP proteins and regulators associate with predisposition to a long list of different diseases, but it is only during the last two decades that we have started to unravel the peculiarities of these associations.
These studies have identified diseases that associate with genetic variants altering specific aspects of the activation and regulation of the AP, which has provided valuable insights into the pathogenic mechanisms underlying those pathologies and occasionally they have also revealed unknown features of the complement proteins involved.

| Rare loss-of-function mutations in complement regulators of the AP
Factor H, the key regulator of the AP, controls complement activation in the fluid phase and on cellular surfaces using distinct functional  Figure 1). [11][12][13][14] The mid-region of FH (SCR5-18) contains additional polyanion binding sites. However, their functional significance and contribution to complement regulation remain unclear. The current view is that this mid-region plays mainly a structural role by enabling FH to bend and bind simultaneously to different sites on C3b. 11,13,[15][16][17] Hundreds of potentially pathogenic rare FH genetic variants, most with MAF < 10 −4 , are present in heterozygosis in normal populations. They are distributed through the whole length of the CFH gene and include both missense and non-sense variants. Cohorts of patients with diseases like aHUS (atypical hemolytic uremic syndrome), C3G (C3-glomerulopathy), and AMD (age-related macular degeneration) are enriched in these rare genetic FH variants, but the type of genetic variant and their distribution in the CFH gene varies among the different pathologies. Missense genetic variants in the C-terminal region of FH, for example, are prototypical of aHUS. [18][19][20][21] Several years ago, the discovery of this association changed our understanding of the pathogenesis of aHUS revealing that this condition was not a consequence of the hypocomplementemia that characterizes many of these patients, but rather it was caused by the complement-mediated damage to the microvascular endothelium due to a failure to regulate complement activation in host surfaces. This was a decisive paradigm shift that prompted the use of anticomplement drugs to treat the disease. 22,23 Further studies with these FH C-terminal variants have also provided important structural and functional information regarding how FH interacts with the thioester domain (TED) of C3b 14,24,25 and with sialic acids 13,26 in cell surfaces.
In contrast with the particular association of the FH C-terminal variants with aHUS, CFH variants that impair expression of FH or eliminate the complement regulatory functions in the N-terminal region associate with a broad spectrum of pathologies like C3G, aHUS, AMD, and IgAN (IgA nephropathy). These pathologies share a common link to complement dysregulation, but the causes that trigger the complement dysregulation and where it occurs are different among them. Later in this review, I will discuss that the final pathological outcome in a heterozygote carrier of these CFH genetic variants is impinged by the concurrence with other complement genetic variants and with environmental factors (see Section 6). Importantly, in homozygosis or compound heterozygosis, these CFH variants that impair expression of FH or eliminate the complement regulatory functions of FH result in the complete consumption of C3 in plasma and the generation of massive amounts of activated C3 products that deposit in the kidney glomeruli, causing prototypically dense deposit disease, a rare form of C3G characterized by strong electron-dense deposits within the glomerular basement membrane. 27,28 In addition, because homozygosity for these FH variants causes a secondary C3 deficiency that severely impairs opsonophagocytosis, the individuals affected by this condition are also predisposed to severe infections, particularly by encapsulated bacteria.
Genetic variants in membrane cofactor protein (MCP; CD46) that decrease the expression of this protein in the cell surfaces or reduce its regulatory activity behave like the FH C-terminal variants described above; they do not impact significantly fluid phase complement regulation but impair the protection of self surfaces from complement damage. This is consistent with the crucial role of MCP as a cofactor in the FI-mediated inactivation of C3b and C4b deposited on host cells. In fact, the majority of pathogenic MCP variants described to date have been found in heterozygosis during the genetic screening of aHUS patients, 29,30 although they are not exclusive of this disease. 31 Homozygotes (or compound heterozygotes) for MCP pathogenic variants are extremely rare. They associate with unusually severe aHUS presentations and with common variable Immunodeficiency. 32

F I G U R E 1
The C-terminal region of FH. Structure of the last two SCRs of FH to illustrate the large surface area involved in the binding sites for the TED domain of C3 (in yellow) and sialic acids (in green). The TED domain (in purple) and a sialic acid molecule (in red) are depicted interacting with their respective binding sites. Because these binding sites play a crucial role in the regulation of complement activation by FH in host surfaces, the region cluster a multitude of pathogenic variants associated with aHUS, including S1191L and V1197A, the two prototypical variants that disrupt the sialic acid binding site (in orange) Decay accelerating factor (DAF; CD55) regulates C3 and C5 cleavage by accelerating the decay of the AP C3/C5 convertases.
DAF is a glycosylphosphatidylinositol (GPI)-anchored membrane protein, a biochemical peculiarity that among complement proteins is only shared with CD59. The most common genetic cause associated with alterations of these two (GPI)-anchored complement regulators are somatic mutations of the PIGA gene, which encodes a protein essential for the synthesis of the GPI anchor. When this happens in a clonal hematopoietic stem cell, it results in paroxysmal nocturnal hemoglobinuria, a rare disease that presents with hemolytic anemia, thrombosis and, eventually, bone marrow failure. 33 Recently, several homozygote carriers of non-sense or pathogenic missense variants in the DAF gene were identified by whole-exome sequencing of a cohort of patients with a very rare early-onset protein-losing enteropathy. 34 This was unexpected because individuals with a congenital DAF deficiency were known for a long time (they are the very rare Inab phenotype of the Cromer blood group 35 ), but no pathological consequence was associated with this phenotype. Consistent with the complement regulatory activities of DAF, the congenital DAF deficiencies result in complement over-activation on cell surfaces.
However, how complement dysregulation distinctively causes damage to the intestinal tissue, resulting in the protein loss enteropathy, is still unclear.
Factor I (FI) is crucial to regulate the activation of both the classical and the AP. In the presence of appropriate cofactors, it inactivates C3b and C4b. Most CFI variants described in the literature are heterozygous rare variants, mainly associated with AMD, aHUS and, occasionally, with different forms of C3G. [36][37][38][39] The majority are missense variants resulting in reduced plasma FI levels, but loss of functional activity has also been demonstrated for a few FI variants that express normally in plasma. 40,41 Heterozygotes for these pathogenic FI variants have impaired regulation of C3b which results in excessive AP activation and they normally develop diseases characterized by chronic inflammation or acute complement-mediated tissue damage. Like carriers of partial FH deficiencies, the final pathological outcome in heterozygote carriers of CFI variants is strongly contingent on the associated genetic background. 42 Like FH, the complete FI deficiency, resulting from homozygous or compound heterozygous variants in the CFI gene, causes secondary C3 deficiencies that normally associate with recurrent infections with encapsulated microorganisms, but since C3b cannot be proteolyzed to generate iC3b and C3dg, these individuals elude the inflammation and tissue damage that characterize the FH deficiency. 28,43

| Rare gain-of-function mutations in the components of the AP convertase
Activation of C3 into C3b causes huge displacement of the TED domain that exposes the reactive thioester to nucleophilic reagents and generates a new surface area in C3b containing the binding sites for FB that mediate formation of the AP pro-convertase C3bB.
Binding of FB to C3b also results in a large conformational change in FB that exposes a site that is cleaved by FD releasing the Ba fragment and yielding the active AP C3-convertase C3bBb. 44,45 Modulation of the activity of the C3bBb convertase, either prolonging its half-life on pathogen surfaces where activation must proceed (by properdin), or accelerating its spontaneous decay and inactivating C3b on host surfaces to avoid inflammation and tissue damage (by FH, MCP, DAF, CR1, and FI), is critical for the correct functioning of the AP.
Complete deficiencies of C3, FB, FD and properdin caused by homozygote (or compound heterozygote) pathogenic variants in these genes are very rare and associate with recurrent infections. [46][47][48] More interesting are, however, the genetic variants in the C3 and CFB genes that increase the functional activity of AP. These gain-of-function (GoF) variants have been described in association with aHUS, C3G, and AMD. 36,[49][50][51][52] The functional characterization of some of the GoF variants in FB (i.e., D279 G, F286L, K323E, and K350N) has shown that they enhance formation of the C3bBb convertase or increase its resistance to accelerated decay by complement regulators ( Figure 2). 49,51 Similarly, it has been shown that GoF variants in C3 alter the sensitivity of C3b to inactivation by FH, CR1 and MCP, and confer the AP C3 convertase resistance to accelerated decay by FH and DAF. 15,36,50,[52][53][54] Both C3 and FB GoF variants cause hyper-activation of the AP, which results in consumption of C3 and FB, increasing complement-mediated inflammation and tissue damage. An important finding was that the C3 GoF variants associated with C3G affects regulation by FH and CR1 (fluid phase regulation), while the C3 GoF variants associated with aHUS affect primarily the inactivation of C3b by MCP (cell surface regulation), 54,55 which again illustrates that the pathogenic mechanisms of a particular disease implicates specific aspects of the activation and regulation of the AP ( Figure 3). The key contribution of these GoF variants to the disease phenotype is further illustrated by the remarkably reproducible and characteristic presentation of aHUS in carriers of two C3 GoF variants (R161W and I1147T) that are relatively prevalent in Europe and Japan. [56][57][58] In a different context, the C3-923delDG variant associated with C3G is also remarkable. It is a deletion of two amino acids (Asp923, Gly924) in the MG7 domain of C3 that makes the corresponding C3b and C3bBb convertase resistant to inactivation by FH (and CR1). Paradoxically C3-923delDG renders C3 resistant to cleavage by the AP C3 convertase and has been crucial to identify a region in the surface of the MG7 domain that is very likely a contact surface between the C3 substrate and the C3b molecule in the AP C3 convertase and may represent a therapeutic target for inhibition of C3 activation ( Figure 3). 52

| Rare genetic variants in the FH-related proteins (FHRs)
Several non-sense genetic variants the CFHR1-5 genes, leading to This is the case of genetic variants in the C-terminal region of FHR-1 that provide it with the capacity to bind sialic acids. These variants associate specifically with aHUS because their pathological consequences are equivalent to those of the FH C-terminal pathogenic variant. 26,59,60 FHR-1 and FH have virtually identical C-terminal regions, with only two amino acid differences between them (Leu290 and Ala296 in FHR-1 are Ser1181 and Val1187 in FH). These two amino acid substitutions suffice to eliminate from FHR-1 the capacity to bind sialic acids, which prevents that FHR-1 hampers the complement regulatory role of FH in host surfaces. 26 aHUS-associated variants CFHR1 Leu290Ser,Ala296Val and CFRH1 Leu290Val are pathogenic because they restore in FHR-1 the capacity to bind sialic acids, making FHR-1 an strong competitor of FH for binding to surface-bound C3b and dysregulating the AP in host tissues. 26 Potential pathogenic GoF variants in FHR-5 with increased binding to C3b and other ligands have also been described associated with aHUS, but detailed functional studies are still pending. 61 Another interesting association is that of the FHR-1, FHR-2, and FHR-5 variants carrying a duplication of the N-terminal dimerization domain with C3G. [62][63][64][65][66][67][68] The classic example is an FHR-5 protein encoded by a CFHR5 gene with an internal duplication resulting in a duplication of SCR1 and SCR2 (FHR-5[1-2]-FHR-5) that was identified in several Greek Cypriot patients with C3GN and a common ancestry (often called CFHR5 nephropathy). 62 Other FHR proteins with duplicated dimerization domains have been identified associated with C3G in small families and include: and FHR-1(1-2)-FHR-1. 64 Different studies have tried to explain why these peculiar FHR variants associate with C3G. It was initially though that by forming multimeric complexes these variants would outcompete binding of FH to surface-bound C3b (FH de-regulation), promoting complement activation. 63 However, a competition with FH would justify better an association with aHUS than with C3G. 26 Notably, recent data generated for the FHR-1(1-2)-FHR-1 variant suggest these FHRs variants dysregulate complement at C3-opsonized surfaces by promoting complement activation and further deposition of C3-activated fragments without interfering the binding of FH to C3b. 64 In summary, variants that confer to is still uncertain, mainly because we lack the necessary understanding of the physiological role of these proteins to perform proper functional assays and also because their identification requires special techniques and they are poorly represented in the available genetic databases of normal populations, which are generally limited to exome sequencing. As an example of the diversity of these CFHR1-5 structural variants, Figure 4 summarizes those that we have identified in our aHUS (n = 1151) and C3G (n = 373) cohorts.
The most common structural variant described in the CFH-CFHRs gene family is the 84 kb deletion of CFHR3 and CFHR1 (Δ CFHR3-CFHR1 ), which has an allele frequency ranging from 2% to 51%, depending on ethnicity. 70 Δ CFHR3-CFHR1 strongly associates with protection from AMD, IgAN, and C3G but confers risk to systemic lupus erythematosus. 64,[71][72][73] In fact, it has been shown that the prevalence of these diseases in human populations correlates well with the allele frequencies of the Δ CFHR3-CFHR1 polymorphism. 70 The reason for these associations is still unclear, although, as discussed above, some interesting hypotheses are emerging. Δ CFHR3-CFHR1 is not a risk factor for aHUS, 74,75 as it was previously reported. 76 In homozygosis, however, is a relevant finding because it is strongly associated with the presence of auto-antibodies against the C-terminal region of FH, the most important acquired factor associated with the development of aHUS in children; 3%-15% in European cohorts 77,78 and as much as 56% of aHUS patients in India 79 have such auto-antibodies. Δ CFHR1-CFHR4 (deletion of CFHR1 and CFHR4) is also found with relatively high frequency in genetic screenings, but no associations with disease have been reported for this deletion. Like Δ CFHR3-CFHR1 , finding Δ CFHR1-CFHR4 in homozygosis or in heterozygosis F I G U R E 3 Gain-of-function variants in C3. A, C3 variants are indicated in the crystal structure of the C3b molecule to illustrate that aHUS-associated variants (black circles) and C3G-associated variants (red circles) are allocated in separate regions of the C3b molecule. 52,53,155 C3b structural domains are depicted in colors. The common variant R102G (yellow circle) is located at the interface between the MG1 and TED domains. C3 variants D1115N (associated with aHUS) and C3-923delDG (associated with C3G) are highlighted with an asterisk to indicate that they have been introduced in mice and replicate the aHUS and C3G phenotypes, respectively. 156,157 B, Functional analysis of the C3G-associated C3-923delDG and aHUS-associated C3-I1157T variants. SDS-PAGE gels illustrate their resistance to FI-mediated proteolysis using FH (C3-923delDG) or MCP (C3-I1157T) as a cofactor (indicated with red arrows). Experimental details are described in Martinez-Barricarte et al. 52,53 Below, a model of the interaction of FH and MCP with C3b provides a structural rationalization of the differential regulation by FH and MCP. 158,159 C, Top view of the hypothetical complex between the C3-convertase (C3bBb) 154  Complete deficiencies of CR1 proteins have not been reported in humans and the first case of a genetic CR2 deficiency was described in 2012. 86 The patient, a 28-year-old man, was a compound

| HAPLOT YPE S IN THE RC A G E N E CLUS TER
The Regulators of Complement Activation (RCA) gen cluster

| The CFH-CFHRs haplotypes
Four haplotypes, CFH-H1, CFH-H2, CFH-H3, and CFH-H4, explain more than 90% of the genetic variability at the CFH gene region and extend also into the CFHRs gene region ( Figure 5). The CFH-H1 haplotype carries the Tyr402His (rs1061170) polymorphism in SCR7 of FH and is strongly associated with the development of AMD. 91 Structural and functional studies have tried to understand the mechanism by which the 402His allele may impact AMD risk. These studies have shown that amino acid 402His is directly involved in a GAG binding site spanning SCR6-8 of FH 92 suggesting that switching between histidine and tyrosine at this position may alter the ligand specificity resulting in failure to recruit FH to sites in the retina where complement is activated by the accumulation of endogenous compounds such as C-reactive protein, heparan sulfates, or malondialdehyde. [92][93][94] Whilst these ideas are suggestive, it has been recently shown that the AMDassociated CFH-H1 haplotype extend into the CFHR4 locus and includes one or more eQTL that influences the expression levels of the FHR-4 protein. 82  (rs800292). The FH-62Ile allele was originally reported to be protective for AMD 91 and later for aHUS and other diseases. 95 The functional impact of the FH-62Ile variant is subtle (20%-50% enhanced regulatory activities compared with FH-62Val), 96 but the amplification nature of the complement system and the combination of this FH variant with other variants in complement components and regulators will amplify this small effect resulting in significant differences in complement activity that justify its association with disease (see Section 5.1). The CFH-H2 haplotype includes also an eQTL (rs1410996) that associates with reduced levels of FHR-4, 97 which again suggest that the protection effect reported for this haplotype may be the sum of distinct variants at different genes.
The CFH-H3 haplotype, originally described as a combination of CFH SNPs that confer increased risk to aHUS, 98 was later extended to include polymorphisms in the CFHR3 and CFHR1 downstream genes. 74 This extended CFH-H3 haplotype strongly associates with risk for aHUS and carries the rs426736 variant that confers protection against meningococcal disease (MD). 99 Analysis of plasma levels of FH and FHR-3 proteins in carriers of this extended haplotype showed slightly reduced levels of FH and twofold elevated levels of FHR-3 compared with non-carriers, 100,101 which may explain the opposite impact of this haplotype in aHUS and MD. N meningitidis recruits FH via the surface lipoprotein fHbp 102, 103 and it has been shown that FHR-3 competes with FH for binding to fHbp on the bacterial surface, influencing its survival in plasma. 104 Since the ability of N meningitidis to evade the host complement system is determined by the relative levels of Interestingly, approximately 15% of the individuals with Δ CFHR3-CFHR1 carry that protective variant together with the also protective FH-62Ile variant in a fifth (CFH-H5) haplotype that has a frequency of 2% in the Spanish population.

| The MCP haplotype
The strong LD in the human MCP/CD46 gene region also reduces the genetic variability within this region to a couple of SNP haplotype blocks. One of these, the MCP ggaac haplotype, is an important risk factor for aHUS, particularly in concurrence with pathogenic GoF variants in C3 or CFB. 49,53,106 The functional analysis of two SNPs included in this haplotype that are located in the CD46 promoter region demonstrated a reduced transcriptional activity compared to the prevalent MCP aaggt haplotype, which suggest MCP ggaac may associate with slightly decreased levels of CD46 in the endothelial cell surfaces. 106 In this respect, the observation that the severity and penetrance of aHUS in carriers of C3 and (1) rs1061170C corresponds to FH-402His and it confers risk for AMD. 91 (2) rs61818925G associates with increased FHR-4 levels, which confers risk for AMD. 82,125 (3) rs800292A corresponds to FH-62Ile and it is protective for AMD. 91 (4) rs1410996A associates with decreased FHR-4 levels and it is protective for AMD. 97 (5) rs6677604A is a proxi of Δ CFHR3-CFHR1 and it is protective for AMD, IgAN, and C3G and confers risk for SLE. 64,71-73 (6) "-" indicates Δ CFHR3-CFHR1 . (7) In homozygosis strongly associates with anti-FH autoantibodies. 160 (8) rs570618, a proxi of rs1061170C, also associates with increased levels of FHR-1, -2, -3 and -4. 82,129,130 (9) rs10922109A, a proxi of rs1061170C, also associates with increased levels of FHR-1, -2, -3 and -4 82,129,130 isolated kidney transplant without anticomplement prophylaxis from a donor negative for the MCP ggaac risk haplotype. 107 Although the evidence is anecdotal, it may be worth exploring the idea that eluding the MCP ggaac risk haplotype may prevent the recurrence of aHUS after kidney transplantation in carriers of CFB and C3 GoF variants.

| FUN C TIONAL P OLYMORPHIS MS AND E XPRE SS I ON QUANTITATIVE TR AIT LO CI (e QTL) IN THE AP
In addition to genetic variants in AP components causing dramatic impact in the expression and/or function of the protein, which

| Functional polymorphisms
I have already described two of these polymorphisms in the AP (CFH-Y402H and CFH-V62I) and described the functional changes that explain their associated with increased risk and protection from AMD and other diseases (Section 4.1). Other common polymorphisms that are also strongly associated with protection and increased risk for AMD are FB-R32Q 108 and C3-R102G. 109 Interestingly, genetic data for some of these polymorphisms associated with AMD indicate that these common variants conferring risk and protection combine to create a gradient of risk for AMD in the population. 110 Functional and structural studies have also revealed the bases for the association of the FB-32Q and C3-102G variants with disease. The FB-32Q variant decreases risk from AMD because it results in a reduced activity of the AP as a consequence of the decreased interaction between the Ba fragment of FB and C3b, which impacts the formation of the AP pro-convertase C3bB. 111 Similarly, the C3-102G variant associates with increased risk to AMD because it is less susceptible than C3-102R to inactivation by complement regulators, which increases the activity of the AP. 112 The biochemical and structural analyses of the C3-R102G polymorphism are particularly interesting because they added an unanticipated complexity in complement regulation revealing that the conformational flexibility of C3b impact the interactions of complement regulators with C3b. 53,113,114 Notably, residue Arg102 at the MG1 domain of C3b ( Figure 3A) is involved in a salt bridge with residue Glu1032 at the TED domain that hold together the TED domain and the MG ring in C3b, a conformation that is critical for the interaction with FH. In C3-102G, this salt bridge is lost, altering the TED-MG1 separation and the regulation by FH. 113,114 A crucial finding of these functional studies was, however, to observe that when the aggregate effect of the combination of variants conferring risk (FH-62 V, FB-32R, and C3-102G) was tested experimentally, they resulted in a significant difference in complement activity vs the low-risk combination (FH-62I, FB-32Q, and C3-102R) that exceeded the subtle functional alterations of the individual complement variants (Figure 6). These high and low AP activity variant combinations should represent the extremes of a continuum in AP activity that fit very well with the additive risk effect observed at the genetic level. Individuals at the high end of complement activity should be more prone to chronic inflammation, which explains the association with AMD, whereas those with low activity may be protected from it, but likely at the cost of increased susceptibility to infection.
Since the CFH, CFB, and C3 genes segregate independently, we should expect that the prevalence of the different combinations of genotypes for the three common polymorphisms are in correspondence with their individual allele frequencies. In the Spanish population, for example, the prevalence of triple homozygotes for the alleles that associate with increased AP activity (CFH-62VV, CFB-32RR, and C3-102GG) is expected to be one in 68, whereas being homozygote for the three alleles associated with reduced AP activity (CFH-62II, CFB-32QQ, and C3-102RR) is only of 1 in 2928. 115 The low prevalence of the low activity combination suggests a negative selective pressure by pathogens on the variants that compose this combination and some very early observations may support this hypothesis. 116 Ironically, what may have once been a disadvantage to escape childhood infections has now become an advantage to evade AMD for an increasingly aged population in the developed world.
The conclusion that the effects of AP variants are additive, and their different combinations result in distinct AP activities, prompt us to redefine Chester Alper's early term "complotype" 117 to refer to combinations of variants in complement components and regulators that result in distinct complement activities. 111,112,115 As I will discuss in Section 6, these "complotypes" are relevant to predict disease risk and should also be of help to assist clinical decision-making in complement-related diseases.

| Complement eQTL
Expression quantitative trait loci are genomic loci that have been associated with variations in mRNA or protein expression levels.
Nowadays, there is a great interest in uncovering these eQTLs because it is thought that they likely explain many of the genetic variants located in non-coding regions of the human genome that have been associated with disease in genome-wide association studies (GWAS). eQTL are also the likely explanation to the large variations in the expression levels of the complement proteins in humans, but data on these eQTLs are scarce and almost limited to variations within the RCA gene cluster.
One of the first studies trying to identify the factors influencing the fivefold range of variation of the FH plasma levels in humans was performed almost 20 years ago. 118 In that study, we applied variancecomponent methods 119  The results indicated that 62% of the FH phenotypic variance is due to genetic effects and provided suggestive evidence of three genomic regions including potential eQTL, one of them within the RCA gene cluster in 1q32. Another example of early complement eQTL are genetic variants at the CR1 locus determining the expression levels of this complement regulator on erythrocytes. 84,85 Nowadays, three SNPs in strong LD (rs11118133, rs3811381, and rs2274565) allow discrimination of two alleles resulting in high (CR1-H) and low (CR1-L) expression of CR1 in erythrocytes. 85,120 The CR1-L allele is a risk factor to experience extravascular hemolysis under eculizumab treatment. 121 It has also been suggested that reduced CR1 expression on erythrocytes leading to impaired amyloid clearance is the mechanism by which the rs6656401 SNP impacts Alzheimer's disease 122 and a similar association with low CR1 expression alleles has been described in preeclampsia. 123 Other examples of complement eQTL already mentioned in this review are the SNPs in the MCP ggaac haplotype that show a reduced CD46 transcriptional activity 106 and the SNPs in the promoter and intergenic regions of the CFH, CFHR3, and CFHR4 genes that influence the plasma levels of the FH, FHR-3, and FHR-4 proteins. 82,98,100,101 Since 2005, numerous GWAS studies have been carried out trying to delineate the genetic predisposition to AMD which has resulted in the identification of numerous SNPs conferring risk or protection to the disease located in intronic or intragenic regions.
As an effort to correlate disease associations with gene expression and to provide an explanation about how disease-associated SNPs located in these non-coding regions cause phenotypic changes, a number of recent studies have explored the contribution of these AMD-associated variants to modulate expression of complement genes, in plasma, liver and retinal cells and tissue, 124-127 stablishing a correlation between some of these SNPs with complement F I G U R E 6 Combinations of common variations in C3, FB, and FH dramatically alter AP activity. A, Figure depicts the functional analysis of the common variants in C3, FB, and FH that have been found associated with increased risk to AMD. Individually, each of these variants shows small but consistent differences in hemolytic activities, compared with the normal allele. The hemolysis assays shown here were performed in normal human serum (NHS) depleted of FH, FB, or C3 that was reconstituted with the normal or AMD-associated variant of the corresponding protein, as described. 96,111,112 B, Additivity of the three polymorphisms was investigated by comparing hemolytic activities of the variant set promoting more AP amplification with that causing less amplification. A NHS depleted of C3, FB, and FH proteins and reconstituted with the appropriate C3, FB, and FH variants was used in these experiments as described. 112 Figure illustrates that the combination of variants C3-102G, FB-32R and FH-62 V shows sixfold increased (EC 50 = 50 nmol/L FB vs EC 50 = 288 nmol/L FB) complement activity in these hemolytic assays compared with that of variants C3-102R, FB-32Q, FH-62I. (Figure was adapted from data published in references 96,111,112 ) expression levels. One of the most significant of these correlations is that of rs6677604, 128 which is protective in AMD with the protective allele in strong LD with Δ CFHR3-CFHR1 and, obviously, correlate with decreased expression of the FHR-1 and FHR-3 proteins. This rs6677604 SNP may also influence expression of CFH and other CFHR genes, but this has to be confirmed. 125 Also, within the RCA gene cluster, there are variants at the with rare pathogenic variants. The CFH-CFHRs and MCP ggaac haplotypes at the RCA gene cluster are also crucial components of these complotypes as we know that the combination of rare pathogenic variants with these haplotypes are often decisive to define the risk of protection from disease. 106 The CFH-CFHRs locus is a major genetic factor in AMD with both risk and protective variants. 71,91,[135][136][137]  To evaluate the susceptibility to aHUS in individuals carrying different loads of genetic risk factors, we have recently analyzed the penetrance of the disease in 372 relatives of aHUS patients carrying 1 or 2 rare complement pathogenic variants. 144 Our data confirmed that the main driver of aHUS in these pedigrees is the pathogenic mutation and that penetrance of the disease raises with the genetic load of risk factors (Figure 7). A detailed age-adjusted analysis showed a relatively low aHUS penetrance of 9.6% at age 48 years for relatives carrying a single pathogenic variant that increases to 36% at age 44 years in carriers of two pathogenic variants. Notably, carrying both the MCP ggaac and the CFH-H3 haplotypes, in addition to one or two pathogenic variants, raises the aHUS penetrance to 18.8% (age 37 years) and 100% (age 44 years), respectively (Figure 7). 144 150 But functional studies are labor-intensive and complicated, and for these reasons variant effect is typically inferred using pathogenicity prediction algorithms and allele frequency data. Both approaches have major limitations and the level of evidence they provide is not as strong (ACMG criterion PP3, supporting evidence of pathogenicity). 150 As discussed in Section 2, the genetic variability of complement genes in the general population is such that the probability to find a pathogenic variant by chance is substantial.

| IDENTIFI C ATI ON AND CL A SS IFI C ATI ON OF COMPLEMENT G ENE VARIANTS
Therefore, an additional question relevant for the interpretation of findings in the genetic screening of patients with complement-related diseases is whether the AP functional alterations expected from the identified variants fit the pathogenesis of the disease.
To identify the strengths and weaknesses of the current pathogenicity prediction algorithms and provide recommendations to aid the appropriate classification of novel FH variants as they are identified, we have recently characterized functionally 105 genetic variants of FH associated with aHUS. 153 These analyses indicate that rarity in normal databases can be misleading for variant classification. While it is true that pathogenic variants tent to be rarer than benign variants, 21.5% of the benign variants are absent in gnomAD.
The data also identify important limitations in applying prediction algorithms to FH variants, as only 74% were classified correctly applying the standard CADD PHRED C-score > 15. Although a differential adjustment of the prediction algorithms to accommodate the peculiarities of the distinct FH regions improves overall predictions to 85%, our final conclusion was that functional analysis of the variants remains the gold standard to provide an accurate classification. between complement activation and regulation. These models will be decisive to determine disease risk and for decision-making in patients' management. Finally, and not discussed here, an intracellular role has been described for some AP proteins and it has been postulated that these newly unraveled activities are as relevant as the well stablished activities of the complement system mentioned in this review. It is yet unclear whether genetically determined differences in the activity and expression levels of the AP complement proteins have consequences for these non-canonical intracellular complement functions.

ACK N OWLED G EM ENTS
Many thanks to Hector Martin-Merinero, Sheila Pinto, and Angela Ruiz for their assistance in the elaboration of this review and to Prof.

CO N FLI C T O F I NTE R E S T
I have no conflict of interest to disclose.

DATA AVA I L A B I L I T Y S TAT E M E N T
Data sharing not applicable to this article as no datasets were generated or analysed during the current study.