Introduction

Estimates of the prevalence of Alport syndrome vary from one in 5,000 to one in 53,000 of the population [1, 2], but the frequent finding of likely pathogenic COL4A5 variants in normal reference datasets and cohorts with renal failure suggests that X-linked disease affects closer to closer to one in 5,000 [3, 4]. This is consistent with Alport syndrome being the second commonest cause of inherited kidney failure after autosomal dominant Polycystic Kidney Disease. It also means that many variants are undetected and that even apparently normal individuals have pathogenic variants in the COL4A3–COL4A5 genes.

Inheritance of Alport syndrome is X-linked (OMIM301050), autosomal recessive (OMIM 203780), digenic (typically with a pathogenic variant in each of COL4A3 and COL4A4 [5]) or autosomal dominant (OMIM 104200, or thin basement membrane nephropathy, due to heterozygous COL4A3 or COL4A4 variants [6]). The typical clinical features of Alport syndrome are persistent microscopic haematuria, end-stage kidney failure, and often a family history of haematuria or renal failure [7]. Hearing loss and non-nephrotic range proteinuria are common. Ocular abnormalities such as lenticonus with abnormal lens protrusion, or fleck retinopathy may be present [7]. Clinical features overlap between boys and men with X-linked disease, and in boys and girls, men and women with recessive or digenic inheritance [5, 8]. Many women with X-linked Alport syndrome, and males and females with heterozygous COL4A3 or COL4A4 variants, or sometimes digenic COL4A3 and COL4A4 variants, have only microscopic haematuria [5, 8, 9].

In February 2020, a group of 47 medical and scientific specialists from three continents, all with an interest in the molecular diagnostics Alport syndrome (the ‘Alport Variant Collaborative’), met in London to review the current recommendations for genetic testing for Alport syndrome and to further refine the American College of Medical Genetics/Association of Molecular Pathologists (ACMG/AMP) criteria for evaluating variants in the COL4A3–COL4A5 genes that have been published previously [8, 10]. The group comprised adult (n = 12) and paediatrics (n = 6) nephrologists, histopathologists (n = 2), geneticists (n = 17), laboratory scientists (n = 4), researchers (n = 2), industry scientists (n = 3) and a representative of the Alport UK Patient Support group (n = 1). Attendees had submitted difficult-to-evaluate COL4A3–COL4A5 variants, which were precirculated, and they spent two sessions at the meeting examining these variants, and discussing the evidence for pathogenicity. Their conclusions have been summarised into the following guidelines for genetic testing for Alport syndrome for use by molecular testing laboratories. A draft of these guidelines was then sent to all attendees, modified based on their comments, and the final document approved by the group prior to submission. Decisions about individual variants are ongoing with further clinical data being sought in order to reach at least two-thirds consensus on the assertions.

Broadening the phenotype

Increasingly, pathogenic variants in COL4A3–COL4A5 are found in individuals with proteinuria (‘nephrotic syndrome’, ‘nephrotic range proteinuria’, ‘steroid-resistant nephrotic syndrome’) or with the renal biopsy finding of focal and segmental glomerulosclerosis (FSGS)[11, 12]. FSGS is one of the commonest causes of glomerular disease resulting in progressive renal failure, and pathogenic COL4A3–COL4A5 variants are the commonest cause of adult-onset FSGS ranging from 5% to 20% of cases with sporadic or familial disease respectively [11, 12]. In such cases the glomerular basement membrane does not necessarily have the lamellation typical of Alport syndrome and COL4A3–COL4A5 variants may be associated with an older age at end-stage kidney failure [13]. Pathogenic variants in any of the COL4A3–COL4A5 genes are also found in up to 10% of individuals with renal failure where the cause was not known [3]. In addition families with IgA glomerulonephritis not uncommonly have mutations in the COL4A3–COL4A5 genes [14]. Finally there are reports of COL4A3–COL4A5 variants being associated with renal cysts where autosomal dominant polycystic kidney disease has been excluded [15, 16].

Thus, not only individuals with the typical features of Alport syndrome should undergo genetic testing, but also those with proteinuria or FSGS, with familial IgA glomerulonephritis or with end-stage kidney failure where there is no obvious cause (Table 1). For individuals with proteinuria or FSGS, or end-stage kidney failure, the presence of haematuria or a family history of haematuria, and in the case of IgA disease, a family history of haematuria, all make the detection of a pathogenic COL4A3–COL4A5 variant more likely. However the meeting considered there was currently insufficient evidence to recommend that individuals with cystic kidney disease should also be tested for COL4A3–COL4A5 variants.

Table 1 Recommendations for genetic testing for Alport syndrome.
Full size table

Genetic testing of the COL4A3–COL4A5 genes

Finding a COL4A3–COL4A5 variant that affects the structure or function of a collagen IV α chain confirms the diagnosis of Alport syndrome, and indicates that other family members should be investigated. Genetic testing also identifies the mode of inheritance, and which family members who are at risk of being affected. End-stage kidney failure is very likely in hemizygous COL4A5 males (90% by the age of 40 years) and individuals with biallelic defects in COL4A3 or COL4A4, but much less common in females with a heterozygous COL4A5 variant (15–30% by 60 years) and a person with a heterozygous pathogenic variant in COL4A3 or COL4A4 [17]. The nature of individual variants also indicates the likelihood of early-onset kidney failure and extra-renal features such as ocular abnormalities [18, 19]. Hearing loss is common with X-linked and recessive disease [19]. Targeted diagnostics can be used to identify other affected family members, to ensure that an affected person does not act as a kidney donor, and in preimplantation genetic diagnosis [8]. Where a variant is identified in a person referred with FSGS for testing, the finding of a causative pathogenic variant in one of the COL4A3–COL4A5 genes means that treatment with corticosteroids or immunosuppressants is not usually useful [4, 20, 21], and that this type of FSGS does not typically recur after renal transplantation.

MPS usually identifies at least 80% of disease-causing variants in COL4A3–COL4A5 where Alport syndrome is suspected clinically [22] because of microscopic haematuria or progressive kidney failure, together with a positive family history of haematuria or renal failure. Where a causative variant or variants are not found, it is important to ensure that the genes are sufficiently ‘covered’ by the sequencing technique. Other explanations for the inability to detect pathogenic variants are the large genomic imbalances encompassing one or more exons (such as deletions, duplications or inversions) that MPS detects less well; a location deep within an intron or regulatory element or affecting a non-canonical splice site (for example, 5 nucleotides from the intron-exon boundary); there being insufficient evidence for pathogenicity and the variant being assessed as a Variant of Uncertain Significance (VUS); the variant affecting a processing enzyme or chaperone rather than a COL4 gene, as occurs in about 10% of cases of osteogenesis imperfecta; [23] and rarely because of mosaicism. Sometimes the disease is simply not Alport syndrome. Phenocopies of Alport syndrome where there is predominantly haematuria include Fechtner syndrome (OMIM155100), fibronectin glomerulopathy (OMIM 601894), Nail-Patella syndrome (OMIM 161200), Hereditary Angiopathy, Nephropathy and muscle Cramps (OMIM 611773), Dense deposit disease (OMIM 134370) and CFHR5 deficiency (OMIM 614809). However increasingly pathogenic variants are found in genes affected in proteinuric diseases.

More pathogenic variants are known for COL4A5 than for COL4A3 and COL4A4, and these are missense in 60%, nonsense in 10%, canonical splice sites in 10% and frameshifts in 20% (databases.lovd.nl/shared/genes/COL4A5). About half the pathogenic COL4A3–COL4A5 variants found in diagnostic laboartories are novel. There appear to be some changes in the types of pathogenic variants found in the last years. Fewer large deletions have been reported possibly because MPS detects these less well, or because the more severe disease that these people have means that they have already been studied. In addition, more non-canonical splice site variants, and non-Gly substitutions in the collagenous and carboxy NC1 domains have been reported.

Definition of pathogenicity

The definition of pathogenicity has become less clear for sequence variants in COL4A3–COL4A5. Previous meetings have concluded that the demonstration of a pathogenic variant was the best evidence for the diagnosis of Alport syndrome [8, 10]. Despite the introduction of the standards and guidelines for the interpretation of sequence variants (ACMG/AMP criteria [24]) there is sometimes insufficient data for a conclusion of pathogenicity. The association with haematuria or kidney failure, and a family history of haematuria or renal failure may be the strongest evidence for the diagnosis of Alport syndrome. Microscopic haematuria is found in 95% of those with a pathogenic COL4A5 variant [8], and at least 67% with a pathogenic heterozygous COL4A3 or COL4A4 variant [6]. End-stage kidney failure occurs in all males, but many fewer females with a pathogenic COL4A5 variant [9]. The situation is more complicated in recessive Alport syndrome by the requirement for two COL4A3 or COL4A4 variants in trans [8]. Renal failure is less common, but recognised increasingly, with pathogenic heterozygous COL4A3 or COL4A4 variants [17].

The manual application of the ACMG/AMP guidelines is time-consuming, subject to error and yields inconsistent results [25]. The tools themselves may be flawed [26]. Several web-based tools, such as Alamut (www.interactive-biosoftware.com/alamut-visual/), Varsome (www.varsome.com) [27] and the ClinGen pathogenicity calculator (calculator.clinicalgenome.org) [25], may help in classifying variants by summarising the results for ACMG/AMP criteria such as the computational criteria (Polyphen2, SIFT, Mutation taster, conservation scores, and occurrence in gnomAD and other reference databases) in one site [24]. The ACMG/AMP recognise that “those working in specific disease groups should continue to develop more focused guidance regarding the classification of variants in specific genes“, and thus the Chandos House meeting assessed the ACMG criteria in the context of Alport syndrome and the COL4A3- COL4A5 genes and made the following recommendations (Table 2) including the identification of significant functional domains and mutational hotspots (PM1); the degree of conservation of the protein sequences in different species (PP3); the ‘well-established’ functional assays that indicate pathogenicity (PS3, BS3); whether pathogenic variants are absent from databases with healthy control individuals (PM2) [28] and the threshold in these databases above which variants are likely to be benign (BA1, BS1).

Table 2 Standards and guidelines for the interpretation of sequence variants in COL4A3–COL4A5 in Alport syndrome (modified ACMG/AMP criteria).
Full size table

The Chandos House meeting did not revise the weighting system for the ACMG/AMP criteria for Alport variants [24]. Nor did it did it reach consensus on who had the responsibility for requesting a review of Variants of Uncertain Significance for a reclassification to Likely Pathogenic/Pathogenic or Likely Benign/Benign and how often this assessment should be undertaken. Many laboratories considered that it was the clinician’s responsibility to request a VUS review based on the ongoing need for a diagnosis or the availability of new clinical information. On the other hand, clinicians considered that laboratories were more aware of advances in algorithms for defining pathogenicity. Finally, the ACMG/AMP recommendations for reporting incidental findings in Clinical Exome and Genome Sequencing do not include pathogenic variants in the COL4A3–COL4A5 genes [29] despite these being common and clinically significant.

Modification of the ACMG/AMP criteria for the COL4A3–COL4A5 genes and rationale

Types of pathogenic variants

Missense variants are the commonest changes found in COL4A3–COL4A5. For the collagen IV α5 chain, more missense variants are found in the large intermediate collagenous domain (84–88% depending on the database) which is more than expected based on length (p ≤ 0.006, LOVD, Table 3) and many of which are Glycine substitutions (95% (278/294) in ARUP, arup.utah.edu/database/). Missense variants in the carboxy non-collagenous domain are also common (11–14%). Missense variants that are not Glycine substitutions are probably the most difficult to classify, but even some Glycine substitutions, especially those adjacent to non-collagenous interruptions, are more difficult. In-silico tools used to assess the effect of a particular amino acid substitution on the protein structure may not take into account the character of the Gly-Xaa-Yaa triplets (a Glycine followed by two other amino acids) typical of the collagen chains and as a result their predictions of the effect on structure may be inaccurate.

Table 3 Location of missense variants in the collagen IV α5 chain in different variant databases.
Full size table

Most nonsense variants in the COL4A3–COL4A5 genes result in nonsense-mediated mRNA decay and loss of the corresponding collagen IV α chains from affected membranes. Because the α5 chain is part of the α3α4α5 heterotrimer, the α3 and α4 chains with which it normally forms the heterotrimer are imperfectly incorporated or not incorporated at all [30], and there is a compensatory increase in the expression of the α1α1α2 network [31]. This means that males with a COL4A5 nonsense variant typically have no α3α4α5 (or α5α5α6) network in affected membranes. In females with COL4A5 heterozygous nonsense variants, however, this loss is segmental depending on lyonisation since clusters of cells produce normal amounts of the α5 chain [32]. In males and females with COL4A3 or COL4A4 heterozygous nonsense variants, affected cells result in less of the α3α4α5 network and membrane thinning.

Variants in COL4A5 have been collated into a number of databases (Table 4), from the literature and submissions from individual laboratories. Currently all databases record variant assessments made by the submitting laboratories. The role of such databases in variant assertions is controversial [33, 34] but they still represent useful repositories. ClinVar is the only database that independently evaluates variants but it has no expert panel assessments for the COL4A3–COL4A5 variants. Most laboratories also consult their own databases of variants that they have assessed previously.

Table 4 Comparison of variant databases for COL4A5.
Full size table

Functional studies

The in vitro or in vivo functional studies used in pathogenicity assessments must be specific for the individual gene variant and not for the gene as a whole. One of the commonest in vitro tests for Alport syndrome is immunohistochemistry of a renal or skin biopsy to demonstrate that the collagen IV α5 chain is absent in X-linked and autosomal recessive Alport syndrome [35, 36]. However this does not confirm that a particular variant is pathogenic but only that the individual has a collagen IV defect consistent with the diagnosis of Alport syndrome.

Among the functional studies that support a variant having a damaging effect on the gene or gene product, in vitro trimerization assays that examine secretion into the extracellular space [37] correlate well with clinical variant severity, especially in distinguishing pathogenic from non-pathogenic Glycine variants in COL4A5. In addition, for non-canonical splice sites, minigene assays or sequencing of urinary podocyte or fibroblast mRNA can be used, but are currently only available in specialised laboratories [38]. Increasingly, COL4A5 variants affecting non-canonical splice sites, up to 11 nucleotides from the intron-exon boundary, are identified in individuals with X-linked Alport syndrome. Sometimes synonymous variants within the coding region result in splicing changes that are also disease-causing. Potential splicing changes can be examined with on-line prediction tools such as MaxEntScan or SpliceSiteFinder, but eventually require confirmation with in vitro assays.

Functional assays used in research include evaluation of abnormalities in electrophoresis and thermal stability of collagen IV (previously common in collagen type I); [39, 40] and mRNA quantitation [41]. Again these assays are not specific for individual gene variants unless a zebrafish, mouse or normal cell line has been engineered to incorporate the variant being assessed.

Critical and well-established domains and mutational hotspots

Collagen IV differs from most other collagens in forming a chicken-wire network rather than fibrillar structure. The COL4A1–COL4A6 genes code for the collagen IV α1–α6 chains. Collagen IV is a heterotrimer with three of the 6 α chains producing a triple helix (α1α1α2, α3α4α5 or α5α5α6) [42].The α3α4α5 network is the main component of the glomerular basement membrane, encoded by the COL4A3, COL4A4 and COL4A5 genes respectively. The amino acid sequences of these chains are each highly conserved in different species, and also between the individual α5, α3 and α4 chains (Fig. 1, Supplementary Figs. 1 and 2).

Fig. 1: Collagen IV α5 chain.
figure 1

a Schematic of collagen IV α5 chain with amino and carboxy non-collagenous domains, and intermediate collagenous sequence with Gly-Xaa-Yaa repeats and 23 non-collagenous interruptions. Gly substitutions adjacent to the interruptions are often hypomorphic with a milder clinical phenotype. b Clustal sequence demonstrating that the collagen IV α5 chain is highly conserved between different species both in the collagenous and in the carboxy non-collagenous domains. This sequence also includes a non-collagenous interruption as well as the carboxy terminus (both in boxes). c Clustal sequence demonstrating that the human collagen IV α5, α3 and α4 chains are also conserved.

Full size image

Each collagen IV α chain has an amino non-collagenous (NC) domain; an intermediate collagenous domain; and a carboxy NC domain (Fig. 1). The intermediate collagenous domain has the sequence Gly-Xaa-Yaa where X and Y are often Proline or Hydroxyproline. The presence of a Glycine in every third position in the Gly-Xaa-Yaa repeats is critical to triple helix formation because Glycine is the smallest amino acid and fits within the helix interior. The Proline residues contribute flexibility. The three α3, α4 and α5 chains bind together through disulphide bonds at the Cysteine residues and by a sulfilimine bond between Lysine and Methionine residues all within the carboxy NC domain [43]. There are also multiple non-collagenous interruptions in the collagenous domains that confer flexibility and facilitate network formation (Tables 5–7) [44]. Each interruption has adjacent Glycine residues, and some incorporate a Glycine. However the biochemistry is even more complicated because while the transcript for individual α chains is read off forwards 5’ to 3’, from the amino to the carboxy terminus, the three chains bind through disulphide and sulphilimide bonds in the carboxy terminus from which the triple helix winds up ‘backwards’ [45, 46]. Substitution of these critical position 1 Glycine residues within the Gly-Xaa-Yaa sequence distorts and disrupts helix formation. Glycine substitutions in other inherited collagen diseases are typically pathogenic, for example, in COL1A1 in osteogenesis imperfecta [47]. The effect of Glycine substitutions in positions 2 or 3 or within the interruption is not known. The local environment is also important and Glycine substitutions adjacent to interruptions, where there is more chain flexibility, appear to be less pathogenic [48, 49].

Table 5 Non-collagenous domains and interruptions in the collagen IV α5 chain [44].
Full size table
Table 6 Non-collagenous domains and interruptions in the collagen IV α3 chain [56].
Full size table
Table 7 Non-collagenous domains and interruptions in the collagen IV α4 chain [57].
Full size table

Proline substitutions in the collagen IV α5 chain are reported pathogenic much less often than Glycine substitutions (databases.lovd.nl/shared/genes/COL4A5). Proline confers flexibility and its replacement with another amino acid is less problematic than for Glycine. However, Proline residues are also found in both the integrin binding sites and in the GlyProProGlyProPro binding site for glycoprotein VI which is needed for platelet activation [50], but no Proline substitutions have been described in these residues in the COL4A3–5 genes.

Interestingly, pathogenic variants have been reported for 10 of the 12 Cysteine residues that are critical for cross-linking in the carboxy NC domain (databases.lovd.nl/shared/genes/COL4A5). No substitutions have been described for the Lysine and Methionine residues in the NC domain that form the sulfilimine bond.

Thus, the Chandos House meeting recommended that most Glycine residues in the collagenous domain of the collagen IV α5, α3 and α4 chains should be recognised as critical residues equivalent to a functional domain. However some Glycine substitutions adjacent to NC interruptions represent hypomorphic variants with a milder clinical phenotype [49, 51] such as p.(Gly624Asp) in COL4A5. Clinical information is critical for the complete evaluation of variants, and where there are discrepancies, the evidence from many laboratories, such as segregation of haematuria with the variant in multiple families or within many members of the same family, may confirm a pathogenic nature. The 12 Cysteines in the carboxy NC domain should also be considered critical domains, and any substitutions are likely to be pathogenic.

Hypomorphic variants

Hypomorphic variants are increasingly described in the COL4A5 gene [51,52,–53] although there are, as yet, no definitions or criteria for their identification. They may result in haematuria only, later-onset kidney failure, or GBM thinning rather than lamellation in a male. Variants in the COL4A3–COL4A5 genes may also modify the pathogenic effects of genes encoding other podocyte or glomerular membrane proteins.

The COL4A5 hypomorphic variant, NM_000495.5:c.1871G>A substitution (rs104886142); p.(Gly624Asp) results in late-onset kidney failure and is the commonest pathogenic variant causing X-linked Alport syndrome in Central and Eastern Europe [51]. It has previously been considered benign [10] or a VUS. This variant is located adjacent to a non-collagenous interruption in the Gly-X-Y sequence which reduces the effect of the Gly substitution (Table 5). The Chandos House meeting considered that hypomorphic variants, such as p.(Gly624Asp) may still be actionable, requiring renin-angiotensin- aldosterone blockade consistent with published guidelines for other COL4A5 variants [8, 54].

The nature of other variants such as NM_000495.5:c.2858G>T (rs78972735; p.(Gly953Val)) in COL4A5 [55], and NM_000092.5:c.1634G>C (rs1800516; p.(Gly545Ala)) and NM_000092.4(COL4A4):c.2996G>A (rs13027659 p.(Gly999Glu)) in COL4A4 is less clear. All are very abundant in certain ancestries. p.(Gly953Val) is also located immediately adjacent to an interruption in the collagenous domain of the collagen IV α5 chain, but it is not on its own associated with haematuria, kidney failure or GBM lamellation [55], and has conflicting interpretations of pathogenicity in Clin Var (www.ncbi.nlm.nih.gov/clinvar/variation/24573/). Both p.(Gly545Ala) and p.(Gly999Glu) affect Gly residues within the collagenous domain of the collagen IV α4 chain itself. These are considered Benign** (www.ncbi.nlm.nih.gov/clinvar/variation/255015/) and with Conflicting interpretation of pathogenicity* (www.ncbi.nlm.nih.gov/clinvar/variation/191312/) respectively. Thus interpretation of pathogenicity is complicated, and variants while themselves not causing haematuria may still worsen renal failure progression.

Conclusions

The COL4A3–COL4A5 genes have particular characteristics that must be considered in assessing variants for a pathogenic or benign nature. Most notable are the importance of the position 1 Glycine substitutions in the collagenous domains, Cysteine substitutions in the carboxy NC domains and the recognition of hypomorphic variants associated with a milder clinical phenotype with microscopic haematuria only, late-onset renal failure or GBM thinning rather than lamellation. Evidence is emerging for the importance of other non-Glycine substitutions especially in the carboxy NC domains but these await further evaluation.