Thin Glomerular Basement Membrane Phenotypes With No Identified Pathogenic COL4A3/A4/A5 Variant

Abstract

Introduction

Pathogenic (P) variants in COL4A3/A4/A5 genes are known to cause thin glomerular basement membrane (GBM) or Alport-related kidney disease; however, the exact diagnostic yield of genetic testing remains unknown.

Methods

In this retrospective genotype-phenotype correlation study, we screened the patient populations of 2 major US medical centers for individuals who underwent kidney biopsy, and who had documented genetic testing results on a large 385-kidney disease gene panel. We correlated GBM thickness, estimated glomerular filtration rate, proteinuria, and hematuria with genotyping results.

Results

We identified 115 patients with coexisting histopathology and genetic testing data, of which 49 had ultrastructural abnormalities of the GBM. Among those 49 cases, 9 had a heterozygous pathogenic or likely pathogenic (P/LP) variant in one of the COL4A genes, and 9 additional patients had COL4A variants of uncertain significance (VUS). Thirty-one patients with thin GBM were COL4A3/A4/A5 wildtype. One patient with a P variant in COL4A4 had no GBM abnormalities. Three COL4A VUS were upgraded to P/LP through experimental testing. Presence of P/LP variants in COL4A genes correlated with GBM thickness, but not with other clinical parameters. Among 31 thin GBM cases with no COL4A variant, we found variants in steroid-resistant nephrotic syndrome-, congenital anomalies of the kidneys and urinary tract (CAKUT)-, and autosomal dominant tubulointerstitial kidney disease (ADTKD) genes characterized as P/LP or as “high-risk VUS.” Immune-mediated glomerular injury was as frequent in biopsy specimens with thin GBM as with normal GBM.

Conclusion

In our study, almost two-thirds of patients with thin GBM have no variant in COL4A3/A4/A5 genes. Our data suggest that genetic testing may not obviate the need for kidney biopsy.

Graphical abstract

In the diagnostic workup of chronic kidney disease (CKD), hematuria, and proteinuria, renal histopathology plays a crucial role in defining parenchymal abnormalities. The benefits of a gold-standard diagnosis are carefully weighed against the risks of the kidney biopsy procedure. In the case of thin GBM- or Alport-related kidney disease, many clinical presentations are defined through chronic indolent hematuria with no development of proteinuria or advanced CKD, and many clinicians may forego a kidney biopsy. The increasing availability of genetic testing adds a new aspect of diagnostic precision. However, the role of systematic genetic testing in CKD is yet undefined, and it is unclear if genetic test results would obviate the need for a kidney biopsy. We hypothesized that thin GBM phenotypes strongly correlate with variants in COL4A3/A4/A5 genes, and we sought to investigate genotype-phenotype correlations in patients who underwent both kidney biopsy and genetic testing. In addition, we hypothesized that an unknown proportion of patients with variants in COL4A genes and thin GBM may have superimposed vascular or immunologic lesions that only histopathology could resolve.

Methods

Chart Review

We conducted a retrospective chart review at Beth Israel Deaconess Medical Center (BIDMC) and Brigham and Women’s Hospital (BWH), identifying 121 patients with available genetic testing and renal histopathology data, including electron microscopy. Genetic tests were obtained between January 1, 2020, and June 30, 2023 at BIDMC, and between January 1, 2020, and September 30, 2022 at BWH. Six patients were excluded from the study because of incomplete results. Of the remaining 115 patients, 79 came from BWH and 36 from BIDMC. The study was approved by institutional review boards at both medical centers (protocols BWH #2021P003586; BIDMC #2023P000145). We collected information on patient age, sex, family history, serum creatinine concentration, estimated glomerular filtration rate, presence of hematuria on urine dipstick, and urine protein-to-creatinine ratio. We manually reviewed all electron microscopy reports for quantitative assessment of GBM thickness and ultrastructural abnormalities, and we reviewed original electron micrographs together with the pathologist. Differences in the quantitative assessment of GBM thickness between both institutions are discussed in the Supplementary Materials. All genetic data were obtained with the Renasight 385-kidney-disease-gene panel (Natera, Inc.). We selectively analyzed variants in COL4A genes, CAKUT-, ADTKD-, and steroid-resistant nephrotic syndrome- or podocytopathy genes. Genes of interest are listed in Supplementary Table S1.

Assessment of GBM Thickness on Electron Microscopy Samples

Electron microscopy samples processed at the BWH Department of Pathology had GBM harmonic mean thickness assessed using the orthogonal intercept method,¹ based on > 100 randomly selected segments. Samples processed at the BIDMC Department of Pathology had GBM thickness measured using the method of Haas² (Supplementary Methods and Supplementary Figure S1). Electron microscopy samples were diagnosed as thin GBM nephropathy (TBMN) or Alport-related kidney disease, if there was GBM attenuation (mean harmonic thickness ≤ 264 nm, assessed using the orthogonal intercept method; or average thickness ≤ 250 nm, assessed using the Haas method). In addition, at the discretion of the pathologist, the diagnosis of TBMN or Alport-related kidney disease was made in some cases of characteristic ultrastructural GBM changes (splitting, lamination, and basket weaving), even if the measured thickness was slightly greater than the cutoff thickness.

Results

In TBMN or Alport-related kidney disease, the causative link to variants in COL4A3/A4/A5 genes is well-known.³ According to the current diagnostic paradigm, we hypothesized that a positive genetic test for a COL4A3/A4/A5 gene variant would predict TBMN or Alport-related kidney disease, and vice versa. To this end, we screened medical records for patients who underwent genetic testing on the Natera Renasight kidney-disease-gene panel at 2 major US academic medical centers between January 1, 2020 and June 30, 2023 (BIDMC), or September 30, 2022 (BWH), respectively, resulting in 578 total cases (319 BWH and 259 BIDMC). Among those, we found 121 patients who had documented kidney biopsy results (Figure 1a). We manually reviewed all patient charts and rejected 6 cases because of incomplete or unanalyzable data and included the remaining 115 cases into our study. The exact indications for genetic testing were not always clear from the chart review, because large-panel genetic testing was often used as an open-ended screening tool, rather than to answer a specific diagnostic question. Of note, 83.5% of the patients were given a histopathologic diagnosis first, and underwent genetic testing later; in the other 16.5% of cases, the decision for a kidney biopsy was made with previous genetic testing results already available.

Figure 1.

Study population and summary genotypes and phenotypes. (a) Flow diagram of the study, and breakdown of genotypes among “normal GBM” and “TBMN/Alport” subcohorts. (b) Identical prevalence of GN in patients with thin and normal GBM. Among the 18 patients with thin GBM and GN, we found IgA nephropathy (n = 8), lupus nephritis (n = 5), membranous nephropathy (n = 3), and antineutrophil cytoplasmic autoantibody–associated vasculitis (n = 2). (c) About two-thirds of patients with histopathological evidence of thin and/or ultrastructurally distorted GBM have wildtype COL4A3/A4/A5 genotypes. One patient with a known pathogenic variant in COL4A4 had no evidence of basement membrane abnormalities. ADTKD, autosomal dominant tubulointerstitial kidney disease; BIDMC, Beth Israel Deaconess Medical Center; CAKUT, congenital anomalies of the kidneys and urinary tract; GBM, glomerular basement membrane; GN, glomerulonephritis; P/LP, pathogenic or likely pathogenic; TBMN, thin GBM nephropathy; VUS, variants of uncertain significance.

We found ultrastructural abnormalities of the GBM in 49 biopsy specimens, consistent with a pathologic diagnosis of TBMN or Alport-related kidney disease (Table 1). Thirty-eight of 49 patients (77.6%) were female, 10 of 49 (20.4%) were African American. In 42 cases, GBM thickness was quantified, revealing a median value of 238 nm (interquartile range: 213–260 nm); in 7 other samples, exact GBM thickness could not be determined, but GBM was characteristically attenuated and ultrastructurally disorganized. Histopathology correlated with clinical phenotypes of hematuria in 36 of 49 cases (73.5%), albuminuria stage A2 or above in 40 of 49 cases (81.6%), and CKD stage G3 or above in 23 of 49 cases (46.9%). None of the patients had known ophthalmic or audiologic symptoms typical of Alport syndrome. Twenty-two of 49 patients (44.9%) had a known family history of CKD. Eighteen of 49 patients (36.7%) had a concurrent histopathologic diagnosis of glomerulonephritis. GN in the context of thin GBM was as frequent as in the context of normal GBM (25/66 cases; 37.9%; Figure 1b).

Table 1.

Demographics and baseline clinical parameters

Characteristics	Normal GBM (n = 66)	TBMN/Alport (n = 49)
Age, sex, race
Age at biopsy, median, (IQR)	39.5 (28, 56) yrs	45 (35–63) yrs
Female (n, [%])	39 (59.1%)	38 (77.6%)
African American (n, [%])	16 (24.2%)	10 (20.4%)
Baseline clinical parameters
Family history of CKD	16 (24.2%)	22 (44.9%)
Estimated GFR (median, [IQR])	61 (41–87) ml/min	52 (23–88) ml/min
Albumin-to-creatinine ratio (median, [IQR])	639 (114–1591) mg/g	439 (108–1311) mg/g
Protein-to-creatinine ratio (median, [IQR])	1.02 (0.28–3.14) g/g	0.74 (0.22–1.97) g/g
Hematuria on dipstick	40 (60.6%)	36 (73.5%)
Trace or 1+	22 (33.3%)	13 (26.5%)
2+ and 3+	18 (27.3%)	23 (46.9%)

CKD, chronic kidney disease; GBM, glomerular basement membrane; GFR, glomerular filtration rate; IQR, interquartile range; TBMN, thin GBM nephropathy.

The subpopulation with TBMN or Alport-related kidney disease (n = 49) is compared with the subpopulation with ultrastructurally normal glomerular basement membranes (n = 66).

We sought to analyze the genotypes of patients with thin GBM phenotypes, across the 385 renal disease genes covered by the Natera Renasight panel. Among the 49 patients with thin GBM, 9 (18.4%) were found to have a monoallelic diagnostic variant (P/LP) in 1 of the 3 COL4A genes, and 9 additional patients had 11 COL4A gene VUS (Table 2). In the 18 patients with COL4A gene variants, we did not find any P/LP variants in other kidney disease genes. For completeness, we report on numerous coexisting VUS in CAKUT- and podocytopathy genes, although their contribution to the basement membrane phenotype remains unclear (Supplementary Table S2). Thirty-one of 49 patients (63.3%) with thin GBM had wildtype COL4A3/A4/A5 genotypes. One additional patient with a known heterozygous P variant in COL4A4 had no GBM abnormalities (Figure 1c). In summary, in all but 1 patient (18/19; 94.7%) with variants in COL4A genes, GBMs were thin or disorganized. In contrast, an ultrastructural thin GBM or Alport syndrome phenotype predicted the presence of a COL4A gene variant in only 36.7% (18/49) of cases.

Table 2.

Genotypes, clinical, and ultrastructural phenotypes in patients with COL4A3/A4/A5 gene variants

ID	Age (yrs)	Sex (M/F)	GBM thickness (nm)	ACRa (mg/g)	eGFR (ml/min)	Foot process effac.b	Coexisting GN	FH CKD	Variant in COL4A Genec	ClinVar ID	Allelic frequency gnomAD	ACMG designationc (Experimental test – this study)
T1	38	F	181	2271	31	1–2	–	–	COL4A3 c.3134G>A (p.Gly1045Asp)	n/a	n/a	Likely pathogenic
T2	39	F	207	1.4a	103	0	IgAN	+	COL4A3 c.872G>A (p.Gly291Glu)	562337	n/a	Likely pathogenic
T3	56	M	215	313	32	1	–	–	COL4A3 c.934G>A (p.Gly312Arg)	n/a	1.20e-6	Likely pathogenic
T4	45	F	217	1219	91	1	–	+	COL4A3 c.4001G>A (p.Gly1334Glu)	1455985	3.84e-6	Pathogenic
T5	47	F	199	115	83	1–2	–	+	COL4A4 c.4129C>T (p.Arg1377X)	17407	8.21e-6	Pathogenic
T6	61	F	223	359	82	1	–	+	COL4A4 c.1405G>A (p.Gly469Arg)	1708954	2.74e-6	Likely pathogenic
T7	30	F	Normal	1265	26	1	–	–	COL4A4 c.3022G>A (p.Gly1008Arg)	552195	1.92e-5	Pathogenic
T8	64	M	209	878	22	0	–	+	COL4A5 c.1871G>A (p.Gly624Asp)	24455	4.63e-5	Pathogenic
T9	53	F	249	1464	43	1	–	+	COL4A5 c.4315+1G>A	24725	1.67e-6	Pathogenic
T10	36	F	n/a	1070	16	2	–	+	COL4A5 c.4706G>A (p.Arg1569Gln)	24757	n/a	Pathogenic
T11	32	F	n/a	0.9a	132	1	IgAN	+	COL4A3 c.2155T>C (p.Ser719Pro) COL4A3 c.4664C>T (p.Ala1555Val)d	2203280 104409	6.43e-6 4.52e-5	VUS (not pathogenic) VUS (pathogenic)
T12	76	F	238	4545	95	2–3	Focal proliferative	–	COL4A4 c.3034T>C (p.Phe1012Leu) COL4A5 c.830C>T (p.Pro277Leu)	n/a 1698603	n/a 2.74e-6	VUS (likely not pathogenic) VUS (inconclusive)
T13	26	F	187	54	> 75	1	–	–	COL4A3 c.547-7T>G (p.?)e	n/a	n/a	VUS (likely pathogenic)
T14	71	F	190	171	97	1	ANCA/MPO	–	COL4A3 c.685C>T (p.Arg229Trp)	1025386	4.65e-5	VUS (borderline pathogenic)
T15	57	F	263	18	49	1	–	–	COL4A3 c.4523A>G (p.Asn1508Ser)	334786	3.72e-4	VUS (not pathogenic)
T16	20	F	295	47	126	1	–	+	COL4A3 c.1978C>A (p.Pro660Thr)	599113	1.42e-5	VUS (not pathogenic)
T17	33	F	191	2206	33	1–2	IgAN	–	COL4A4 c.2144C>T (p.Ala715Val)	667226	2.84e-4	VUS (not pathogenic)
T18	68	F	260	920	39	1	–	–	COL4A4 c.1203A>G (p.Ala401=)f	422613	6.40e-6	VUS (pathogenic)
T19	63	F	225	117	56	1	–	–	COL4A5 c.3041C>G (p.Pro1014Arg)	2166888	2.20e-5	VUS (borderline pathogenic)

ACMG, American College of Medical Genetics and Genomics; ACR, albumin-to-creatinine ratio; ANCA, antineutrophil cytoplasmic autoantibody eGFR, estimated glomerular filtration rate; F, female; FH CKD, familial hypercholesterolemia chronic kidney disease; GBM, glomerular basement membrane; GN, glomerulonephritis; IgAN, IgA nephropathy; M, male; MPO, myeloperoxidase; n/a, not applicable; VUS, variants of uncertain significance.

Ten patients (20.4%) tested positive for a heterozygous pathogenic or likely pathogenic variant in COL4A3 (n = 4) or COL4A4 (n = 3), or a single pathogenic or likely pathogenic variant in COL4A5 (n = 3; 1 male, 2 female), respectively. One of the patients with a COL4A4 pathogenic variant had an ultrastructurally normal GBM (T7). Nine additional patients had a single (n = 7) or 2 (n = 2) heterozygous VUS in one of the COL4A3/A4/A5 genes. Among the VUS, 2 were noncoding, and the remainder were missense variants. Three VUS were determined as pathogenic through experimental evidence in a minigene splicing assay, or in a collagen domain trimerization assay (Supplementary Methods, Supplementary Table S3 and Supplementary Figure S2).

^aAlternatively, urine protein-to-creatinine ratio, [g/g].

^b0 = no effacement, 1 = mild/segmental, 2 = moderate/segmental, 3 = severe/diffuse.

^cPathogenic or likely pathogenic alleles (known and experimentally determined) are depicted in bold.

^dThis variant results in an amino acid substitution in the C4 domain (II) of COL4A-3. It exhibited reduced extracellular collagen domain trimerization in the NanoLuc to 68.5% of wildtype levels, and was considered pathogenic.

^eThis variant creates a novel splicing acceptor site, resulting in transcripts carrying additional 6 bases in exon 10 and insertion of 2 amino acids (Phe, Lys) between aa 182 and 183. The insertion occurs after 4 and before 13 Gly-X-Y repeats, likely disrupting the triple helix, and considered likely pathogenic.

^fThis variant leads to the usage of a cryptic splicing donor site in intron 19, resulting in transcripts carrying additional 34 bases in exon 19. This translates into a frameshift and is considered pathogenic.

Because many COL4A3/A4/A5 variants associated with thin GBM phenotypes were VUS, we sought to experimentally test their pathogenicity through in vitro functional assays. We hypothesized that the 2 noncoding variants, COL4A3 c.547-7T>G (p.?) and COL4A4 c.1203A>G (p.A401=; T13 and T18, Table 2) affect RNA splicing, and we employed both variants in a reverse transcription polymerase chain reaction–based splicing reporter minigene assay.^4,5 We found that both noncoding variants are “pathogenic” or “likely pathogenic,” through the creation of novel splice donor or acceptor sites. The aberrant splicing outcomes include a 2–amino-acid in-frame insertion, predicted to disrupt the collagen triple helix, in the case of variant T13 (now “LP”), or a frame-shifting and thus truncating event in the case of variant T18 (now “P”). The 9 missense variants found in the remaining 7 patients were tested for their impact on collagen trimerization in a split-Nano-Luciferase-reporter assay.^6,7 P variants disrupting collagen trimerization exhibit impaired reconstitution of Nano Luciferase from its precursor fragments, resulting in reduction of measured Nano Luciferase activity. One missense variant (T11) was found to be pathogenic, the others were borderline pathogenic (T14, T19) or not pathogenic. In summary, as a result of this study, 3 of 11 VUS in COL4A genes were reclassified as P/LP variants (Table 2, Supplementary Table S3, Supplementary Figure S2).

We correlated GBM thickness with genotype, stratifying the thin GBM cohort into carriers of COL4A P/LP variants, VUS, and wildtype. Patients with known or experimentally proven P variants had the thinnest GBM (median GBM thickness: 212 nm [interquartile range: 201–221.5]), whereas the median GBM thickness was 247 nm (interquartile range: 218.5–268.5) in the COL4A wildtype cohort (P < 0.05; Figure 2a). We did not find statistically significant differences in estimated glomerular filtration rate or proteinuria between the cohorts (Figure 2b and c).

Figure 2.

Clinical phenotypes of patients with thin glomerular basement membranes (GBMs). (a) GBM thickness as a function of genotype∗ (median: 212 nm with diagnostic variants in COL4A genes, 231.5 nm with COL4A VUS, 247 nm in thin GBM nephropathy (TBMN) with COL4A wildtype). (b) eGFR as a function of genotype∗ (median: 41 ml/min with diagnostic variants in COL4A genes, 95 ml/min with COL4A VUS, 57.5 ml/min in thin GBM pathology with COL4A WT). (c) UPCR as a function of genotype∗ (median: 1.45 g/g with diagnostic variants in COL4A genes, 0.35 g/g with COL4A VUS, 0.7 g/g in thin GBM pathology with COL4A wildtype). ∗COL4A VUS with experimentally determined pathogenicity were subtracted from the group of reported COL4A VUS and included in the group of known diagnostic variants (according to the Natera Renasight panel diagnosis). eGFR, estimated glomerular filtration rate; GBM, glomerular basement membrane; n.s., not significant; P/LP, pathogenic or likely pathogenic; UPCR, urine protein-to-creatinine ratio; VUS, variants of uncertain significance; WT, wild type.

We further reviewed the 31 patient samples with thin GBM and COL4A wildtype genotypes, identifying 25 cases with variants in genes related to CAKUT, ADTKD, and glomerular disease (Supplementary Table S4): Two patients had dominant genetic kidney disease (ADTKD-UMOD, and FSGS-INF2, respectively). Three additional patients had biallelic APOL1 high-risk genotypes, and 1 was a heterozygous carrier of the NPHS2 R229Q allele. The remaining 19 patients in this subcohort had 39 VUS that were individually reassessed based on evolutionary conservation and analysis of the protein structure context. As a result of our manual adjudication, we declared 5 of 39 variants as “high-risk VUS”; that is, potentially pathogenic, including variants in UMOD, ROBO2, GLA, MAGI2, and DLC1 (Supplementary Tables S4 and S5; Supplementary References S1–S12). Sample electron micrographs of pathologically attenuated GBMs in patients with COL4A wildtype genotypes are depicted in Figure 3.^8,9

Figure 3.

Sample ultrastructural phenotypes of thin glomerular basement membranes (GBMs) in COL4A wildtype patients. (a–f) Sample electron micrographs of GBMs in patients with COL4A wildtype genotypes; scale bar, 500 nm (identifier as per Supplementary Table S4; missense variants in dominant and recessive steroid-resistant nephrotic-syndrome- or podocytopathy genes): (a) Patient T21, aged 65 years, GBM mean harmonic thickness 231 nm, UMOD p.Y133C, FN1 p.S2296L, TTC21B p.K1048E. The UMOD variant is considered potentially pathogenic or “high-risk variants of uncertain significance (VUS)” and is expected to act in a dominant mode. (b) Patient T26, aged 27 years, GBM mean harmonic thickness 223 nm, INF2 p.R218Q. This is a pathogenic variant in a dominant gene, and the correlation between INF2 variants and thin GBM nephropathy phenotypes has been described previously.⁸ (c) Patient T23, aged 26 years, GBM mean harmonic thickness 191 nm, ROBO2 p.P1077R, SMARCAL1 p.A127P. The ROBO2 variant is considered potentially pathogenic or “high-risk VUS” and is expected to act in a dominant mode. Beyond its known role as a congenital anomalies of the kidneys and urinary tract gene, ROBO2 has more recently been recognized as a podocyte and basement membrane gene.^9,S6^,S7 (d) Patient T36, aged 45 years, average GBM thickness 291 nm, GLA p.N320D, TTC21B p.R863W. The variant in GLA is considered potentially pathogenic or “high-risk VUS” and is expected to act in a dominant mode. Upon reexamination of the electron micrographs, no specific ultrastructural abnormalities of Fabry disease were identified. (e) Patient T32, aged 32 years, average GBM thickness 286 nm, DLC1 p.D1134G. This variant is considered potentially pathogenic or “high-risk VUS.” The patient has nephrotic syndrome and a focal segmental glomerulosclerosis–like lesion. Notice foot process effacement (arrowheads). (f) Patient T35, aged 68 years, average GBM thickness 254 nm, DLC1 p.R1425Q. Causative links between GBM phenotypes and variants in UMOD, ROBO2, GLA, or DLC1 have not been described in the existing literature. In Supplementary Table S4, we present a comprehensive summary of manually adjudicated VUS across all examined genes.

Discussion

Systematic genetic testing in kidney disease is being increasingly used and has the potential to change diagnostic routines and guide individualized therapy. In this study, we sought to correlate genotypes, as revealed by the Natera Renasight gene panel, with ultrastructural phenotypes of the GBM. Our goal was to better define the genetic background of TBMN or Alport-related kidney disease and to investigate the role of genetic testing and histopathology in the diagnostic workup of CKD. Our single most surprising finding was the absence of pathologic variants in COL4A3/A4/A5 genes in almost two-thirds of our patients, which is a major deviation from the current diagnostic paradigm.

Although our sample size of 115 biopsy cases is small, this is, to our knowledge, the largest study systematically correlating renal disease gene panel testing results with ultrastructural kidney biopsy phenotypes. Further limitations of our study are its retrospective design and its observational nature. We are unable to assess case by case whether genetic tests were justified to help solve a specific differential diagnosis, or if tests were ordered indiscriminately, with an open-ended diagnostic question. Because our cases were identified through screening of a genetic test result database, there is a certain selection bias, and it is unknown how well our study population represents TBMN or Alport-related kidney disease in general. With that in mind, the true frequency of COL4A (and other gene) mutations in TBMN or Alport-related kidney disease remains elusive and requires further study. All examined cases are singletons, and we did not collect family pedigree information. A more comprehensive genotype-phenotype correlation study would require genetic tests and kidney biopsies on first-degree relatives, and segregation analysis across pedigrees, which is hardly feasible. Beyond the experimentally proven pathogenic COL4A alleles, we cannot claim causality for any other of the identified CAKUT- or podocytopathy variants. We therefore avoid using uniform American College of Medical Genetics and Genomics nomenclature¹⁰ in our variant adjudication, labeling plausible VUS as “high risk VUS,” in contrast to “likely pathogenic.”

Despite the limitations, we point out several relevant observations. First, an unexpectedly high proportion of thin GBM phenotypes was negative for COL4A gene variants. It is known that various genetic mechanisms may cause GBM abnormalities,^3,11 including podocytopathy gene variants,^8,12 or combinations of COL4A variants with heterozygous variants in recessive genes, such as MYO1E¹³ and LAMB2.¹⁴ A large list of candidate genes was described in a recent systems biology study,⁹ including many genes that are not known to play a role in human kidney disease, and that are not part of the 385-gene panel we used. We now report the correlation of several monoallelic variants in CAKUT-, ADTKD-, and podocyte disease genes with a thin GBM phenotype, including UMOD, ROBO2, GLA, NPHS2, MAGI2, and DLC1 (the latter 3 are recessive). It is conceivable that low-penetrance mutations in dominant genes or certain heterozygous recessive disease alleles may exhibit a phenotype that is only detectable through electron microscopy. Further studies are needed to determine causality. In addition, we speculate that the absence of audiologic or ophthalmic symptoms in our patient population is related to the low proportion of patients with COL4A gene variants, and that extrarenal manifestations are less common with variants in CAKUT- or podocytopathy genes. Second, in contrast to other reports,¹⁵ we find no evidence that thin GBMs predispose for immune-mediated injury, or, conversely, that GN may lead to thin GBM–like abnormalities. In our cohort, we find the same proportion of GN in patients with thin and normal GBM, suggesting that pathologic mechanisms are independent. Third, given the genetic complexity of thin GBM and the possibility of coexisting immunologic disease, our data suggest that genetic testing and kidney biopsy are complementary. A negative genetic test for a pathologic COL4A variant does not exclude TBMN or Alport-related kidney disease, and a positive test would not have revealed coexisting GN in more than one-third of our cases. We therefore conclude that both genetic testing and kidney biopsy will remain necessary for a precision diagnosis.

Disclosure

PGC and AJB have received honoraria for advisory or consultation activities from Natera, Inc. All the other authors declared no competing interests.

Supplementary Material

Supplementary File (PDF)

Supplementary Methods. Supplementary References. Figure S1. Comparison of GBM thickness measurements at BWH and BIDMC. Figure S2. Split-Nano-Luciferase assay of missense COL4A3/A4/A5 VUS. Table S1. Genes included in the variant analysis of patients with Alport-like ultrastructural phenotypes. Table S2. Additional variants in CAKUT and SRNS or podocytopathy genes in patients with Alport-like ultrastructural phenotypes and COL4A3/A4/A5 gene variants. Table S3. Experimental testing of pathogenicity of COL4A variants of uncertain significance. Table S4. Variants in ADTKD, CAKUT, and SRNS or podocytopathy genes, in patients with Alport-like ultrastructural phenotypes and COL4A wildtype genotypes. Table S5. AlphaFold identifiers for reference protein structure models. STROBE checklist.