Episode Transcript
[00:00:03] Picture two genes that work as partners. They sit at the Same locus on chromosome 2. Each one encodes a chain of the same collagen, the one that holds the kidney's filter together. If you inherit one damaged copy of either gene, clinical guidelines have generally treated the two as interchangeable. Now, a study of more than 900,000 people has tested that assumption. For some kinds of damage it does not hold.
[00:00:28] The same type of variant carries a very different risk, depending on which of the two partners it hits.
[00:00:39] This is base by base. Every episode is built with AI tools from a real Open Access paper curated by a human, and every claim is checked against the paper before we publish. You'll find this episode's fact check score on our website.
[00:00:53] Now back to the story.
[00:01:00] Genes are COL4A3 and COL4A4, together with a third gene, COL4A5. They encode the alpha 3, alpha 4 and alpha 5 chains of type 4 collagen. These chains are made by glomerular podocytes, cells of the kidney's filtration barrier. The three chains trimerize, and the collagen they form is essential for the glomerular basement membrane. When these genes carry pathogenic variants, the result can be Alport syndrome, a hereditary kidney disord.
[00:01:34] Allport syndrome usually begins with blood in the urine, called hematuria, typically first noted during childhood. Podocyte loss, protein in the urine and progressive kidney damage can follow. Kidney failure can occur as early as childhood in some people and as late as old age in others. Hearing loss and eye abnormalities can come with it.
[00:01:56] So here is the if a variant breaks the alpha 3 chain, is that the same as a variant that breaks the alpha 4 chain? And how would anyone measure the difference in a population that was not selected for kidney disease?
[00:02:14] Today we celebrate the work of Konstantinos Dzumkas, Gabriel Dr. Omid Szedeghi Olavije and Daniel Gale from the Centre for Kidney and Bladder Health at University College London, who have advanced our understanding of collagen 4 kidney disease.
[00:02:30] This base by base episode is based on their Open Access article Population scale genomics reveals divergent pathogenicity of variant classes across paralogous collagen 4 genes, published in the American Journal of human genetics in December 2026.
[00:02:50] Alport syndrome comes in more than one form. The X linked form is caused by COL4A5 variants. It is most severe in males who carry a single X chromosome, and they face a high lifetime risk of kidney failure. Heterozygous females usually have milder disease, although they can develop chronic kidney disease and even kidney failure. The autosomal recessive form needs two damaged copies of COL4A3 or COL4A4. It affects both sexes similarly and typically causes early onset severe kidney disease.
[00:03:28] Then there are heterozygotes, people with a single damaged copy of COL4A3 or COL4A4. They are strongly associated with microscopic hematuria. When a kidney biopsy is performed it often shows a thinned basement membrane. These variants are now recognized as risk factors for proteinuria and even kidney failure, with a lifetime risk of kidney failure that is thought to reach up to 3%.
[00:03:55] And these heterozygotes are not rare. Analyses of cohorts not enriched for kidney disease suggest that up to 1 in 106 people carry such a variant.
[00:04:07] To follow this study we need the shape of the protein. Type 4 collagen has a long triple helical collagenous domain built from a repeating motif of glycine than any amino acid than any amino acid.
[00:04:22] Glycine is the smallest amino acid and it is needed for the three chains to wind tightly.
[00:04:28] Replacing it with a bulkier amino acid is the commonest type of mutation in Alport syndrome. At the far end of each chain sits the nc1 domain, a non collagenous region that is crucial to start trimer formation that gives three classes of damage glycine substitutions in the helix missense variants in the NC1 domain which can block the trimer from forming and truncating variants which cut the protein short in people with 2 damaged copies or in males with COL4A5 variants. Truncating variants are associated with a younger age at kidney failure than glycine substitutions in heterozygotes. The picture is less clear.
[00:05:12] Much of the literature comes from families with kidney disease who were more likely to be tested. That is Ascertainment bias and there is an assumption sitting underneath all of this. Clinical guidelines and variant classification frameworks have generally treated COL4A3 and COL4A4 as functionally interchangeable. Those are the paralogues in the title two related genes that that encode partner chains of the same complex.
[00:05:41] The authors point out that this assumption had never been tested at population scale. If it is wrong, risk estimates, variant classification and genetic counseling for many families could be affected.
[00:05:52] Testing it requires a very large and unselected population.
[00:06:02] The main data set was the UK Biobank, a prospective cohort of approximately 500,000 mostly healthy adults aged 40 to 69 years. When they were recruited, participants gave samples and questionnaires and their records are linked to hospital data over time.
[00:06:18] The authors used whole genome sequencing data from this cohort to reduce confounding by population structure. The main analyses were restricted to people of inferred European ancestry. For replication, they turned to a second independent cohort, the all of US Research program.
[00:06:37] Next came the phenotypes defined to be specific. Glomerular hematuria was taken from hospital diagnosis codes. People with other obvious causes of blood in the urine, such as urinary tract infections, kidney stones or genitourinary cancers, were excluded.
[00:06:54] Proteinuria combined diagnosis codes with a urine measurement taken at the baseline visit the albumin to creatinine ratio.
[00:07:03] Severe chronic kidney disease used codes for advanced stages and for kidney replacement therapy.
[00:07:09] Each definition came with a matching set of control subjects people without those codes.
[00:07:16] The analysis ran at three levels. First, every variant was tested on its own. Many rare variants appear in only one to three people, so the authors used Fisher's exact test on 2x2 tables and kept coding variants whose confidence interval for the odds ratio excluded one.
[00:07:34] Second, rare variants were grouped into the three classes for each gene separately. That made six genetic predictors. Glycine substitutions and NC1 missense variants also had to pass a pathogenicity score threshold, a CADD score above 20.
[00:07:51] Each of those six groups was then tested against against a long list of diagnoses in what is called a Phenome wide association study in the UK Biobank. That meant a panel of 1,363 diagnoses codes which included the predefined proteinuria and severe kidney disease traits.
[00:08:10] The models adjusted for age, sex and the first 10 genetic principal components. Then the authors compared the two genes directly for each variant class. They took the the effect size in Col4a3, subtracted the effect size in Col4a4 and tested the difference.
[00:08:29] The third level was genome wide. The authors ran a genome wide association study for glomerular hematuria looking for any locus anywhere in the genome. Then they repeated it with a twist. They built a weighted burden score for each person from the rare cold 4A3 and COL4A4 variants they carried and added it to the model. If a signal disappears after that adjustment, it is accounted for by the combined effects of those rare variants. If new signals appear, there could be modifiers elsewhere in the genome.
[00:09:05] Start with single variants. The hematuria analysis included 9,100 affected people and 335,746 control subjects. A total of 117 rare coding variants in the two genes were significantly associated with hematuria, 105 of them changed the protein Glycine substitutions were the most frequent Class 33 variants across hematuria and proteinuria. The authors identified several 64 Col4A3 and 92 Col4A4 variants, I.e. a variant by variant evidence base measured in an unselected population.
[00:09:50] Proteinuria told a similar story at the single variant level, 16,074 affected people were included and 86 coding variants showed significant associations, 68 of them protein channels changing again glycine substitutions predominated. Severe chronic kidney disease had fewer cases and 62 coding variants showed statistical evidence of association.
[00:10:14] Only five variants were significant across all three phenotypes. They were all glycine substitutions in the triple helix or one recurrent truncating variant, a founder variant that comes up again later in this story.
[00:10:29] Then the authors counted how many carriers in each group actually had the phenotype. Here the two genes began to separate for proteinuria. Among carriers of truncating variants, the case rate was 52.7% in COL4A4 and 17.1% in COL4A3.
[00:10:50] These are raw case rates not yet adjusted for anything.
[00:10:56] The grouped tests put that difference on firmer ground. Glycine substitutions behaved alike in both genes. They were strongly associated with proteinuria and with unspecified hematuria. Truncating variants were a different matter. 119truncating variants in COL4A3 carried by 732 people showed no significant association with any diagnosis. After correction for multiple testing, 87 truncating variants in COL4A4 carried by 713 people were associated with proteinuria. Several hematuria codes, chronic renal failure and gout.
[00:11:37] There was an obvious objection 1 col 4 a 4 truncating variant is a founder variant enriched in people of British ancestry with a minor allele frequency of 0.13% in the UK Biobank. Was it driving everything?
[00:11:53] The authors removed it. The number of carriers fell from 713 to 175. The associations with proteinuria and unspecified hematuria remained highly significant and of similar magnitude. The broader associations, including gout, no longer passed the threshold, so the founder variant contributed to the breadth of the signal, but the core associations were not purely attributable to it.
[00:12:20] NCN1 missense variants repeated the pattern. In COL4A3 there were 94 variants carried by 2,030 people and they showed no significant associations.
[00:12:34] In COL4A4 there were 103 three variants carried by 1,256 people were associated with proteinuria and unspecified hematuria.
[00:12:45] Notice the numbers. COL4A3 had more carriers, which should make an association easier to detect, not harder, and the predicted pathogenicity was comparable with mean CAD scores of 24.26 and 23.98.
[00:13:03] The direct comparison made it formal for glycine substitutions. The effect sizes in the two genes were equivalent for both hematuria and proteinuria. For truncating variants, COL4A4 had significantly larger effects than COL4A3 for both traits and that difference persisted after excluding the founder variant.
[00:13:25] NC1 missense variants showed the same pattern with significantly larger effects in COL4A4 for both traits. The authors summarize it plainly. Truncating and NC1 variants confer significantly greater risk in COL4A4, while glycine substitutions confer equivalent risk in either gene.
[00:13:47] Would it hold in a second cohort with many contributing healthcare sites in all of us? The replication used to participants of European ancestry aged 40 or older, about 160,000 people.
[00:14:01] Glycine substitutions in both genes again associated with hematuria. Truncating and NC1 variants in COL4A3 again showed no significant associations while those in COL4A4 did. In the direct comparison for hematuria, the gene specific difference replicated for both both classes. Proteinuria was harder to define consistently in all of us, and the NC1 difference did not replicate for that trait.
[00:14:31] Finally, the genome wide scan with 8,224 affected people and 41,120 matched controls. The COL4A3 and COL4A4 locus was the strongest hematuria signal in the genome. Its lead variant was that same founder variant.
[00:14:52] Two other regions also stood out HLAB on chromosome 6 and THSD4 on chromosome 15.
[00:15:00] Then the authors added the weighted burden score. The signal at the collagen locus was completely abolished, HLAB and THSD4 remained and no new locus appeared.
[00:15:17] What does that last result mean? The genetic architecture of hematuria in this population looks largely explained by the direct effects of rare variants in these two genes with no evidence of independent transacting modifiers. The authors add a careful the variant weights were estimated without accounting for local haplotypes types for the founder variant. Some of the attributed effect could reflect linked common variants, so the abolished signal means it is accounted for by the rare variant effects. Not necessarily that common variants at this locus do nothing.
[00:15:53] The clinical message is direct gene identity, not variant class alone should inform risk stratification, variant interpretation and gender genetic counseling for people with one damaged copy of these genes. A truncating variant in COL4A3 and a truncating variant in COL4A4 may look like the same kind of finding on a report in this population, they did not carry the same risk.
[00:16:19] The authors go further. Paralogous genes encoding subunits of the same complex can differ substantially in their tolerance to loss of function variance, a principle that may apply to other gene families.
[00:16:34] Why would the two chains differ? The authors offer two possibilities, and they call the first one speculative alpha the four may play a disproportionately important role in selecting chains or stabilizing the assembled protomer. The second involves alpha 5.
[00:16:50] Alpha 3 and alpha 5 both belong to the odd class class of collagen 4 chains, which share more sequence similarity, and the alpha 5 NC1 domain can form complexes on its own, so alpha 5 might partially substitute for a missing or damaged alpha 3, while nothing comparable compensates for the loss of alpha 4. Neither idea has been directly tested.
[00:17:14] The limitations matter here. The analyses were restricted to people of European ancestry and the findings may not generalize. Phenotypes came from hospital codes and available laboratory data. Asymptomatic microscopic hematuria, the cardinal early sign, is usually found in primary care and rarely generates the codes used here. The UK Biobank is a healthy volunteer cohort, so the case rates should be read as conservative estimates of penetrance and for severe chronic kidney disease the numbers were small smaller, so the authors do not extend their tolerance claim to that outcome.
[00:17:54] So here is where this leaves us. Two partner genes build the same collagen in the kidney filter. Glycine substitutions in the helix carried equivalent risk in both but truncating and NC1 missense variants were strongly associated with hematuria and proteinuria in circumstances Col4A4 and substantially attenuated or absent in Col4A3 that held after removing a founder variant, and it was largely replicated in a second cohort. The kind of damage alone did not predict the risk which gene carried it mattered too.
[00:18:30] These findings came from population data, not from biopsies or protein studies, and the mechanism is still over. Open direct evidence would require looking at which collagen chains are deposited in the kidney filters of people with these variants. Until then, the result is a clear caution for anyone reading a genetic report that two genes treated as equivalent may not be if partner genes can differ this much in how they tolerate damage, what does this mean for the future of variant interpretation in other protein complexes?
[00:19:06] This episode was based on an Open Access article under the CCBY 4.0 license. You can find a direct link to the paper and the license in our episode description. If you enjoyed this, follow or subscribe in your podcast app and leave a five star rating. If you'd like to support our work, use the donation link in the description. Thanks for listening and join us next time as we explore more science base by base.
