Rare genetic causes of complex kidney and urological diseases

Abstract

Although often considered a single-entity, chronic kidney disease (CKD) comprises many pathophysiologically distinct disorders that result in persistently abnormal kidney structure and/or function, and encompass both monogenic and polygenic aetiologies. Rare inherited forms of CKD frequently span diverse phenotypes, reflecting genetic phenomena including pleiotropy, incomplete penetrance and variable expressivity. Use of chromosomal microarray and massively parallel sequencing technologies has revealed that genomic disorders and monogenic aetiologies contribute meaningfully to seemingly complex forms of CKD across different clinically defined subgroups and are characterized by high genetic and phenotypic heterogeneity. Investigations of prevalent genomic disorders in CKD have integrated genetic, bioinformatic and functional studies to pinpoint the genetic drivers underlying their renal and extra-renal manifestations, revealing both monogenic and polygenic mechanisms. Similarly, massively parallel sequencing-based analyses have identified gene- and allele-level variation that contribute to the clinically diverse phenotypes observed for many monogenic forms of nephropathy. Genome-wide sequencing studies suggest that dual genetic diagnoses are found in at least 5% of patients in whom a genetic cause of disease is identified, highlighting the fact that complex phenotypes can also arise from multilocus variation. A multifaceted approach that incorporates genetic and phenotypic data from large, diverse cohorts will help to elucidate the complex relationships between genotype and phenotype for different forms of CKD, supporting personalized medicine for individuals with kidney disease.

Chronic kidney disease (CKD) is a broad term that encompasses many complex disorders that collectively affect over 1 in 10 individuals worldwide, resulting in substantial morbidity and mortality and high health-care costs¹. Despite the heavy economic and societal burden of CKD, its pathogenesis remains incompletely understood, impeding specific diagnosis and targeted therapy¹. In addition to known Mendelian nephropathies², multiple lines of evidence support a broader genetic contribution to the pathogenesis of CKD. Between 10 and 29% of adults with chronic kidney failure note a positive family history for nephropathy across different ethnicities and aetiologies^3-5. Furthermore, glomerular filtration rate has a heritability of approximately 30–60% in the general population^6-8, and other indices of kidney function, such as albuminuria and electrolyte excretion, show similarly significant heritability^9-11.

Prior research has shown that most common diseases result from a combination of genetic and environmental factors and have more complex genetic architectures than originally assumed. For example, investigations of epilepsy^12,13, autism^14,15 and coronary artery disease^16,17 have demonstrated that these complex conditions encompass many distinct disorders, including both monogenic and polygenic subtypes. In addition, these studies have shown that, contrary to earlier assumptions, genetic variants spanning a wide spectrum of allele frequencies mediate an individual’s risk of developing complex disorders^18,19. Typically, the effect of a variant on disease risk is inversely related to its frequency in the general population (FIG. 1). Thus, monogenic forms of disease are usually caused by rare or even private variants in a single gene that have a large effect size, whereas modifier alleles exhibit intermediate allele frequencies and more modest effect sizes. Polygenic subtypes of kidney disease usually reflect the collective contribution of multiple common variants, each of which has a small effect on disease risk; however, investigations of polygenic risk scores have shown that individuals who have many such common variants can have a similar disease risk to individuals with a monogenic form of the condition¹⁶ (BOX 1). Moreover, in some cases, relatively common variants have been shown to each have a large impact on disease risk, as illustrated by the association of APOL1 risk variants with various forms of non-diabetic CKD^20,21 and the association of risk alleles at the HLA and PLA2R1 loci with idiopathic membranous nephropathy^22,23.

Fig. 1 ∣. Chronic kidney disease as a complex disease.

Chronic kidney disease (CKD) is frequently considered to be a single disease entity, defined by persistent abnormalities of kidney structure and/or function²⁶. However, these phenomena can result from many genetically distinct aetiologies, with disease risk shaped by the contributions of variants of varying frequencies and effect sizes. Causal variants for CKD include rare or even private variants with large effect sizes, which are generally associated with monogenic CKD subtypes, and more common alleles, which have been associated both with clinically distinct CKD subtypes, such as focal segmental glomerulosclerosis (FSGS), IgA nephropathy (IgAN), and membranous nephropathy (MN), and with kidney function traits, such as estimated glomerular filtration rate, serum creatinine level and albuminuria²¹³. Although common alleles typically have small individual effect sizes, certain common alleles have a greater impact on disease risk, including APOL1 variants in FSGS and non-diabetic CKD (ND-CKD)^20,21 and HLA and PLA2R1 variants in MN²². Moreover, variants within a single gene can have a role in both monogenic and polygenic forms of nephropathy. For example, rare variants in UMOD result in the monogenic nephropathy of autosomal-dominant tubulointerstitial kidney disease, whereas more common non-coding UMOD variants are associated with kidney function traits, such as estimated glomerular filtration rate and serum creatinine level⁴⁶. In addition, the high genetic heterogeneity of CKD means that different single-gene mutations can yield the same clinical disease subtype as defined by clinical symptomatology and investigations of kidney function, imaging and/or histopathology. Conversely, the high phenotypic heterogeneity of CKD can also lead to mutations in the same gene, producing clinically distinct CKD subtypes. For example, the clinical disease entity of FSGS can result from mutations in any one of multiple different genes, including TRPC6, INF2, PAX2 and CLCN5. In many cases, mutations in these genes can also yield other clinical subtypes of CKD; for example, PAX2 mutations can cause congenital anomalies of the kidney and urinary tract (CAKUT) and CLCN5 mutations can cause tubulointerstitial disease (TID). Conversely, mutations in a single gene can produce a variety of CKD phenotypes. For instance, individuals with variants in HNF1B can present with CAKUT, TID or cystic kidney disease. As a result of such high phenotypic heterogeneity, many monogenic nephropathies can present as CKD of unknown aetiology (CKD-UE). Together, this genetic and phenotypic variability supports the notion that, as posited for other common diseases¹⁹, CKD might instead represent a wide array of rarer disorders, each of which has its own distinct genetic architecture.

Box 1 ∣. Opportunities for precision nephrology through polygenic risk scores and phenotypic risk scores.

The clinically non-specific and heterogeneous nature of many forms of chronic kidney disease (CKD) can make early diagnosis challenging. Moreover, such phenotypic heterogeneity can make it difficult to link genotype to phenotype, thereby impeding accurate prognosis and targeted management. Several findings point to the potential of genome-wide polygenic risk scores (GPS) and phenotypic risk scores (PheRS) to address these challenges and improve understanding of genotype–phenotype relationships in nephrology, supporting precision medicine for patients with kidney disease.

GPS represent a weighted sum of the risk alleles harboured by an individual for a given condition, with the weight assigned to a given allele corresponding to estimates of its effect on disease risk from prior genome-wide association studies (GWAS)²¹⁴. These scores are then refined by assessing their performance in predicting the relative risk of the disease of interest in a second (validation) cohort. Typically, such common risk alleles each account for a small fraction of total disease risk, limiting their predictive value in clinical practice. However, when aggregated into a GPS, they can detect individuals at risk of a variety of complex disorders, including coronary artery disease, type 2 diabetes mellitus and breast cancer¹⁶. Importantly, disease risk for individuals at the uppermost GPS percentiles (first and second percentiles) was found to be comparable with that of individuals harbouring causal mutations for rare monogenic forms of the condition. Moreover, as GPS are derived from common variants, they have the potential to identify a greater proportion of individuals at risk of CKD than does screening for monogenic forms alone and could also help to explain the variable penetrance and expressivity observed for many monogenic nephropathies. However, owing to the great aetiological heterogeneity of CKD, the risk alleles thus far identified by GWAS explain a miniscule fraction of its heritability and are also often unique to ethnic subpopulations, both of which make it difficult to create a GPS that is sufficiently robust and broadly generalizable for use in clinical practice²¹⁵. By focusing GWAS on specific (and thus more aetiologically homogeneous) forms of nephropathy in studies that include individuals of diverse ancestries, this challenge may be surmounted. In addition to being sensitive to ancestry, which limits their applicability across different ethnicities²¹⁶, the performance of GPS might be impacted by other factors, including age, gender and socioeconomic status²¹⁷. Thus, further research and integration of additional determinants of disease susceptibility, including demographic and lifestyle factors, into GPS may further augment their value, enabling prediction of absolute versus relative risk, and thereby supporting early disease detection and robust risk stratification in the greater CKD patient population.

As many monogenic nephropathies can manifest with clinically non-specific or ambiguous presentations, identifying affected individuals can be difficult. A 2018 study²¹² identified PheRS as a promising means of surmounting this diagnostic challenge. In that study, researchers mapped diagnostic codes from individuals’ electronic health records (EHRs) to the phenotypic manifestations of various monogenic disorders. Using these data, they constructed PheRS for these disorders, which was defined as a weighted summary of various traits associated with a given condition, with each trait weighted according to its specificity for that condition (that is, its prevalence among affected versus unaffected individuals). Individuals with high PheRS were found to harbour rare, putatively pathogenic variants in genes for the associated monogenic disorders, which included multiple monogenic causes of CKD, such as DGKE-associated nephrotic syndrome and AGXT-associated primary hyperoxaluria. Importantly, review of these individuals’ EHRs demonstrated that, despite their consistent clinical features, the majority of individuals were not clinically recognized to have the associated monogenic disorder. Together, these findings illustrate the potential to empower diagnosis of rare conditions through the leverage of genomic sequence and EHR data, supporting early recognition and targeted management for individuals who may otherwise have gone undiagnosed.

The spectrum of CKD consists of many clinically distinct disorders, including congenital anomalies of the kidney and urinary tract (CAKUT), diabetic kidney disease, hypertensive nephropathy and glomerulopathies, such as IgA nephropathy and focal segmental glomerulosclerosis (FSGS). Many of these disorders can arise from genetic aetiologies, but they can also occur from secondary factors, including metabolic disease, immune dysfunction, infection and/or toxic injury. Moreover, the mode of inheritance and underlying molecular mechanisms vary considerably within many CKD subtypes^2,24,25. In epidemiological studies, however, CKD is often considered to be a single disease entity, defined as persistent abnormalities of kidney structure and/or function²⁶. Using this definition, genome-wide association studies (GWAS) in large population cohorts have detected numerous loci associated with CKD and/or variation in kidney function; however, these loci explain a minority of the overall heritability of these traits^8,27,28.

In this Review, we describe rare genetic causes of CKD and the insights that they offer into the genetic and phenotypic complexity of this group of disorders. We examine the phenotypic variability of recurrent hereditary nephropathies, including genomic disorders and monogenic aetiologies, and their underlying genetic mechanisms. To gain further understanding of the complexity of these diseases, we explore the phenotypes that result from pathogenic variants at multiple loci. Finally, we discuss novel approaches to help to address the challenges posed by the complexity of CKD, such as polygenic risk scores and phenotypic risk scores, which support a precision medicine approach for individuals with kidney disease.

CKD as a complex disease

Complex disorders are defined as conditions that have a strong heritable component, yet show considerable phenotypic heterogeneity owing to genetic and environmental modifiers^19,29. Thus, although CKD and features of renal function show substantial heritability, the phenotypes associated with rarer genetic forms of disease can be highly variable and may overlap, which can complicate diagnosis based on clinical symptomatology alone. For example, patients with nephronophthisis-related ciliopathies and those with monogenic renal tubulopathies can both present with enhanced renal echogenicity^30,31, and those with DGKE-associated atypical haemolytic uraemic syndrome, type IV collagen-associated nephropathy and Dent disease 2 caused by OCRL mutations can all present with steroid-resistant nephrotic syndrome (SRNS)³². Coupling genome-wide analysis with reassessment of clinical features consistent with the candidate genetic diagnosis (that is, ‘reverse phenotyping’) can help to surmount this diagnostic challenge and pinpoint individuals with phenocopies of the condition of interest. For instance, reverse phenotyping following exome sequencing among patients clinically diagnosed with SRNS identified variants diagnostic of other monogenic nephropathies that could phenocopy SRNS in 28% of cases³³. Similarly, exome sequencing and reverse phenotyping of patients with presumed CAKUT found mutations in phenocopy genes in 4 of the 9 patients (44%) with diagnostic mutations³⁴.

Sources of phenotypic complexity in rare genetic nephropathies include pleiotropy, incomplete penetrance and/or variable expressivity. For example, mutations in PAX2 have pleiotropic effects as they can cause papillorenal syndrome, whose manifestations include ophthalmological anomalies, high frequency hearing loss and renal disease, reflecting the fact that PAX2 encodes a transcription factor that is involved in the development of the eye, ear and urinary tract^35,36. However, owing to incomplete penetrance, not all individuals with a disease-causing PAX2 mutation will be symptomatic³⁶. Variable expressivity can result in further phenotypic complexity, with the spectrum of disease manifestations and severity varying between affected individuals with mutations in the same gene. Hence, patients with PAX2 mutations can manifest with a wide spectrum of kidney diseases: although some present with the classic PAX2-associated phenotype of CAKUT³⁵, some develop nephrotic syndrome^37,38, and others display no signs of abnormal renal structure and/or function³⁹. Such variability has even been reported among family members harbouring the same PAX2 mutation⁴⁰.

The exact mechanisms underlying pleiotropy, incomplete penetrance and/or variable expressivity of rare genetic kidney and urological disorders are incompletely understood. In addition to being shaped by the type of genetic mutation and environmental factors, these phenomena can be mediated by the actions of modifier genes⁴¹. In the case of some purportedly monogenic conditions, the observed manifestations can vary substantially depending on the genotype at other loci, suggestive of oligogenic inheritance. For example, in sickle cell disease, which results from mutations in the gene that encodes the β-haemoglobin subunit, HBB, common variants at two other loci — BCL11A and HBS1L-MYB — lead to greater expression of (normal) fetal haemoglobin and thereby confer a milder phenotype, with fewer pain crises⁴². Similarly, in patients with NPHP1-associated nephronophthisis, the prevalence of retinal degeneration is over sevenfold higher among individuals who are heterozygous for a common missense variant in AH1 than among those without this missense variant⁴³. Digenic inheritance has also been reported among patients with the renal tubulopathy Bartter syndrome type IV, with biallelic mutations in both CLCNKA and CLCNKB segregating with disease^44,45. Common variants can also yield a genetic architecture similar to that of digenic inheritance resulting from two rare mutations: for example, individuals who harbour homozygous variants at the HLA and the PLA2R1 loci have a nearly 80-fold higher risk of developing idiopathic membranous nephropathy than those who do not have these risk variants^22,23. Additional research will provide further insights into these and other complex models of inheritance and the role of modifier genes in mediating the clinical manifestations of nephropathies.

Insights from large-scale genetic studies

Hereditary kidney and urological disorders can arise from genomic disorders, which arise from structural variants, or from monogenic causes, which result from shorter variants (that is, single nucleotide variants or smaller insertions or deletions) in a single gene. Microarray and massively parallel sequencing (MPS)-based studies performed over the past 5 years (Supplementary Table 1) show that both genomic disorders and monogenic aetiologies contribute meaningfully across different clinical subtypes of CKD. Although common variants typically have small effects on disease risk, as noted above, exceptions do exist, including the APOL1 variants in FSGS and other forms of non-diabetic CKD^20,21 and the HLA and PLA2R1 variants in membranous nephropathy^22,23 (FIG. 1). Moreover, a single gene can contribute to both monogenic and polygenic forms of nephropathy. For instance, rare variants in UMOD are associated with a rare disease — autosomal-dominant tubulointerstitial kidney disease (ADTKD) — whereas more common UMOD variants confer increased risk of all-cause CKD⁴⁶ (FIG. 1).

Genomic disorders

Several studies that have assessed the prevalence of genomic disorders in large CKD paediatric cohorts^47-52 suggest that copy number variants (CNVs) in one or more syndrome-associated loci account for a meaningful proportion of disease cases. These investigations^47-52 demonstrate a particularly high burden of genomic disorders among individuals with CAKUT, with pathogenic CNVs detected in approximately 4–10% of patients. Importantly, this enrichment was noted both among individuals reported to have isolated renal anomalies and among those with syndromic disease, reflecting the sensitivity of nephrogenesis to alterations in gene dosage^53,54.

Review of these investigations also demonstrates a high level of heterogeneity with respect to both the spectrum of genomic disorders and the associated phenotypes observed. A microarray study of 178 individuals with all-cause CAKUT⁵⁰, for instance, identified seven previously described genomic disorders, each of which was unique to a single patient. Similarly, assessment of nearly 3,000 patients with CAKUT⁴⁹ detected 99 distinct genomic disorders among 159 patients, comprising 45 different known genomic disorders in 112 individuals and an additional 47 individuals harbouring 54 novel, large, rare, genic copy number variants. Of these 99 genomic disorders, 80 (81%) were singleton genetic diagnoses. Moreover, many of the recurrent genetic diagnoses spanned a diverse array of CAKUT phenotypes. For example, 16p11.2 microdeletions were identified in individuals with upper urinary tract anomalies, such as renal agenesis and obstructive uropathy, and in those with lower urinary tract anomalies, such as vesicoureteral valves and duplicated collecting systems.

Importantly, other studies have also noted enrichment of genomic disorders among individuals with other (non-CAKUT) kidney phenotypes, supporting a broader contribution of genomic imbalances to CKD. For example, a study of 419 paediatric patients with CKD of any cause identified pathogenic CNVs that were diagnostic for a known genomic disorder in 5% of non-CAKUT cases, and large, rare, gene-disrupting CNVs that were likely pathogenic in an additional 2.5% of patients⁴⁸. These cases spanned a wide range of clinical disease subtypes, including glomerulopathy and thrombotic microangiopathy, as well as congenital renal disease. Furthermore, a 2019 assessment of nearly 400,000 adults from the UK Biobank, who were recruited for population sequencing analyses and unselected for CKD, found that multiple genomic disorders that had been previously noted to be enriched among patients with CAKUT, such as alterations at the 17q12 and 16p11.2 loci, were also significantly associated with all-cause kidney failure⁵⁵.

In addition, owing to ascertainment bias in genetic testing practices, the above rates may in fact underestimate the true contribution of genomic disorders to CKD. Many of the genomic disorders noted above also involve extra-renal manifestations — such as neuro-developmental delay and congenital heart disease — that have been first-line indications for microarray testing. In these cases, patients’ renal function may have gone untested, thereby impeding accurate assessment of the prevalence of nephropathy. Accordingly, retrospective analyses of patients with genomic disorders noted to be enriched in CKD have revealed higher rates of kidney anomalies than previously thought. For example, in a medical record review of 42 patients with the 16p11.2 microdeletion syndrome, 9 of the 15 (60.0%) patients who had undergone nephrological work-up had CAKUT — a substantially higher prevalence than the 5.4% of cases with the 16p11.2 microdeletion syndrome noted in DECIPHER, a database that contains phenotypic data submitted for individuals with various genomic disorders⁴⁹. Similarly, retrospective review of the medical records of 1,267 patients with 22q11.2 deletion syndrome demonstrated genitourinary anomalies in 162 (15.1%) of the 1,073 patients for whom renal imaging data were available⁵⁶.

Of the genomic disorders identified in individuals with CKD, CNVs at the 17q12, 16p11.2 and 22q11.2 loci are among the most common, estimated to collectively occur in 2.9% of patients with CAKUT⁵³. As these loci contain multiple genes that have pleiotropic effects, each of these disorders has diverse, multiorgan manifestations; however, they share the feature of being associated with highly variable neurodevelopmental, cardiac and renal anomalies⁵³ (Fig. 2). This observed phenotypic complexity hinders full understanding of the molecular pathogenesis underlying these disorders and thereby impedes accurate clinical prognosis and targeted management. In some instances, phenotypic heterogeneity might in part result from the differing effects of a loss (for example, a deletion) versus a gain (for example, a duplication) at the locus. For example, deletions at the 16p11.2 locus lead to macrocephaly and obesity, whereas duplications at this same locus are associated with microcephaly and being underweight⁵⁷. However, the highly variable penetrance and expressivity^47-49 of these syndromes suggests that the phenotypic heterogeneity is a consequence of more complex genetic mechanisms than differences in gene dosage alone. Studies published over the past 4 years have utilized a multidisciplinary approach, integrating additional genetic studies, bioinformatics and functional modelling, to investigate the molecular pathogenesis of deletions at these loci and pinpoint genetic drivers of their renal and extra-renal manifestations.

Fig. 2 ∣. Genetic drivers of the complex phenotypes of common genomic disorders in chronic kidney disease.

Copy number variants (CNVs) at the 17q12, 16p11.2 and 22q11.2 loci are common genomic disorders detected among individuals with chronic kidney disease, altogether occurring in approximately 2.9% of individuals with congenital anomalies of the kidney and urinary tract⁵³. These disorders are characterized by highly heterogeneous neurodevelopmental, cardiac and renal involvement, reflecting genetic pleiotropy and variable penetrance and expressivity. Multidisciplinary investigations integrating genetic, bioinformatic and functional studies have pinpointed key genetic mechanisms underlying the renal and extra-renal manifestations of these disorders. a ∣ In patients with deletions in 17q12, HNF1B haploinsufficiency mediates renal and endocrine anomalies and DNA methylation contributes to variable neuropsychiatric dysfunction. b ∣ In patients with 16p11.2 microdeletion, the degree of TBX6 gene inactivation mediates congenital anomalies of the kidney and urinary tract phenotypes. c ∣ In patients with deletion at the 22q11.2 locus, epistatic interactions between CRKL and other genes in the 22q11.2 locus (SNAP2 and AIFM3) are thought to drive the renal anomalies. GI, gastrointestinal; GU, genitourinary.

17q12 deletions.

Deletions at the 17q12 locus were first detected among individuals with renal cysts and/or CAKUT and maturity onset diabetes of the young^58,59. They were subsequently found to be enriched among patients with neurological disorders, such as autism, schizophrenia and epilepsy; other manifestations can include dysmorphic facial features, liver disease, and cardiac, musculoskeletal and gastrointestinal anomalies^60-62 (Fig. 2). A reciprocal duplication of this locus is less common, and more commonly presents with neuropsychiatric disease, with CAKUT reported in less than 25% of cases⁶³.

The classic 17q12 deletion syndrome interval contains 15 genes, including HNF1B. In addition to the known importance of HNF1B in kidney and pancreas development^64,65, findings of heterozygous intragenic HNF1B loss-of-function mutations among individuals with CAKUT and maturity onset diabetes of the young^66,67 supported the notion that HNF1B haploinsufficiency drives the observed kidney and endocrine anomalies^58,59. Although HNF1B participates in hindbrain development in mouse and zebrafish models^68,69, the higher rates of neurodevelopmental and neuropsychiatric disease among patients with 17q12 deletions than among individuals with intragenic loss-of-function HNF1B mutations⁷⁰ suggested that the neurological phenotype does not result from HNF1B haploinsufficiency alone. Two other genes in the 17q12 locus, LHX1 and ACACA, are thought to contribute to the neurological anomalies observed, based on the known role of LHX1 in brain development^71,72 and the identification of smaller CNVs involving ACACA in patients with autism⁷³. However, the dearth of individuals with neurological disease associated with mutations in either of these candidate genes and the variable severity of neurological disease among 17q12 deletion carriers⁷⁴ support a more complex disease model.

An investigation published in 2018 suggested that this variability in neurological disease severity might reflect the role of epigenetic factors⁷⁵. Analysis of methylation patterns revealed that these differed between patients with 17q12 deletions and those with HNF1B intragenic mutations. Within the 17q12 locus, top signals included sites upstream of LHX1 and ACACA; in addition, differential methylation was observed in another region, upstream of SLC1A3, a gene for which duplications have been associated with autism and attention deficit hyperactivity disorder⁷⁶. Although further study is needed to fully understand the genetic basis of 17q12 deletion-associated neurological disease, these findings support a model wherein HNF1B haploinsufficiency is further mediated by altered methylation of other genes, both within and outside of the locus.

16p11.2 microdeletions.

16p11.2 microdeletions were first recognized as a recurrent genomic disorder among individuals with autism spectrum disorder^77,78 and were subsequently found to have extra-neurological phenotypes including obesity, congenital heart defects, vertebral anomalies, macrocephaly and hearing impairment^79-82 (FIG. 2). Although a 2012 investigation of 285 individuals with the 16p11.2 microdeletion reported CAKUT to be relatively infrequent, noting it in just 3.4% of the cases assessed⁸⁰, 16p11.2 microdeletions have been detected in multiple patients with CAKUT in microarray studies of paediatric CKD cohorts^48,52, pointing to a potential association.

Follow-up case-control genetic investigations and functional modelling have demonstrated that CAKUT is a major feature of the 16p11.2 microdeletion syndrome and identified TBX6 as a key driver of the renal phenotypes observed⁴⁹. In that study⁴⁹, a comparison of nearly 3,000 individuals with all-cause CAKUT with over 21,000 population controls found that 16p11.2 deletions were significantly enriched among patients with CAKUT. A retrospective review of clinical data from a separate cohort of 42 individuals with the 16p11.2 microdeletion syndrome further supported this association, revealing genitourinary anomalies in 9 of the 15 (40%) individuals with genitourinary tract imaging available⁴⁹. Analysis of the 9 patients with CAKUT who were found to harbour 16p11.2 microdeletions demonstrated a minimal region of overlap containing 19 genes, including TBX6 (REF.⁴⁹). In addition to being highly haploinsufficient, TBX6 was identified as a top candidate gene for 16p11.2-associated renal anomalies for two reasons. First, TBX6 has a known role in the development of the primitive streak^83,84, which is the primary source of the embryonic mesoderm — a precursor tissue in the context of genitourinary development — and is expressed in the mesonephric duct and metanephric blastema, which give rise to the trigone of the urinary bladder and renal tubules, respectively⁸⁵. Second, intragenic loss-of-function mutations in TBX6 had previously been detected in individuals with vertebral anomalies⁸⁶, which are clinically associated with CAKUT⁸⁷. Subsequent animal studies further supported TBX6 as a genetic driver of the renal manifestations in this syndrome^49,88. In mice, inactivation of TBX6 resulted in CAKUT, with the degree of TBX6 gene inactivation mediating the penetrance and severity of the renal anomalies observed^49,88.

Reports of CAKUT among individuals harbouring heterozygous loss-of-function TBX6 mutations^88,89 further support a dosage effect of TBX6 on 16p11.2 microdeletion-associated renal manifestations. A 2019 case report⁸⁹ of a Chinese patient with a duplex kidney and collecting system and mild vertebral anomalies detected two variants: an intronic variant that was experimentally demonstrated to be a loss-of-function mutation, resulting in altered splicing and premature protein truncation; and a known hypomorphic allele comprising three common polymorphisms previously found in Chinese individuals with congenital scoliosis⁸⁶. Her unaffected father was homozygous for the hypomorphic allele and her unaffected mother was heterozygous for the intronic variant, with the other allele being wild type. A subsequent investigation detected loss-of-function TBX6 mutations in 7 individuals⁸⁸; upon reverse phenotyping, 6 (86%) had phenotypes consistent with or suggestive of CAKUT. Although additional investigations will provide further insight, findings to date support gene dosage as a key mediator of TBX6-associated kidney disease in patients with 16p11.2 microdeletion, with the degree of TBX6 inactivation shaping its observed penetrance and expressivity.

22q11.2 deletions.

22q11.2 deletions are associated with DiGeorge syndrome, velocardiofacial syndrome and conotruncal anomaly face syndrome, the manifestations of which include congenital heart defects, immunological disease, hypocalcaemia, craniofacial and musculoskeletal anomalies, malignancies and neuropsychiatric disease; CAKUT has been reported in approximately 30% of patients⁹⁰ (FIG. 2). The 22q11.2 locus is a 3-Mb region that includes four sets of low copy number repeats (LCRs), designated LCR22A, LCR22B, LCR22C and LCR22D, in which deletions can occur via meiotic non-allelic homologous recombination^91,92. Heterozygous 2.5-Mb deletions spanning regions A–D (LCR22A–D) are detected in over 90% of affected individuals; the remainder harbour smaller deletions⁹¹. Prior studies identified TBX1, located in the LCR22A–B interval, as a candidate gene for DiGeorge syndrome-associated cardiac^93,94 and potentially neuropsychiatric disease⁹⁵. However, the fact that the syndrome can result from smaller 22q11.2 deletions that do not include TBX1 supports a role for other genes in the pathogenesis of the DiGeorge phenotype, including its renal manifestations.

Two studies in 2017 integrated case–control genetic analyses and animal models to identify candidate genetic drivers of 22q11.2 deletion-associated kidney anomalies^47,96. Comparison of individuals with CAKUT with unselected population controls demonstrated a significant case-level enrichment for 22q11.2 deletions⁴⁷; retrospective analyses showing high rates of genitourinary anomalies among individuals with a molecular diagnosis of DiGeorge syndrome (that is, harbouring 22q11.2 deletions)^47,96 further supported this association. Comparison of the 22q11.2 deletions observed among patients with CAKUT revealed a minimal region of overlap in the LCRC–D interval, which includes CRKL. CRKL is a highly haploinsufficient gene that had been previously shown to confer DiGeorge-like features, including neurocristopathies⁹⁷ and congenital heart defects⁹⁸, in knockout mice; however, kidney phenotypes were not assessed in these previous mouse model studies. MPS-based sequencing demonstrated an excess of rare, putatively damaging intragenic CRKL variants among patients with CAKUT relative to those observed in exome data from population-based controls⁴⁷, further supporting CRKL as a candidate gene. Subsequent studies in mice and zebrafish^47,96 confirmed that CRKL is a genetic driver of CAKUT in DiGeorge syndrome and also suggested that the CAKUT phenotype is mediated by epistatic interactions between CRKL and AIFM3 and SNAP29 (REF.⁴⁷) — two other genes within the LCRC–D interval that are shared among 22q11.2 deletion-carrying patients with CAKUT.

Monogenic aetiologies

MPS-based studies demonstrate that diagnostic yield differs considerably between various CKD subtypes (Supplementary Table 1), highlighting corresponding variability in genetic architecture. For example, detection rates of 60% or higher have been reported for renal cystic ciliopathies, compared with 5–20% for CAKUT and 10–30% for glomerulopathies, supporting the notion that the contribution of known monogenic causes differs for different forms of nephropathy. Importantly, high genetic heterogeneity has been noted across clinical disease categories, with sequencing studies of individuals sharing a single clinical diagnosis, such as SRNS or renal cystic ciliopathies, revealing the existence of many distinct single-gene aetiologies and a high rate of singleton genetic diagnoses. A review of large-scale MPS-based studies of patients with CKD of ‘unknown aetiology’ published over the past 2 years (Supplementary Table 1) further illustrates this heterogeneity⁹⁹. Of the 443 individuals collectively included in the six studies assessed, monogenic forms of nephropathy were identified in 93 (21%), encompassing 47 distinct genes. Across these investigations, the majority of patients in whom a monogenic form of nephropathy was identified were singletons, with each study reporting that 61% or more of the monogenic nephropathies identified were unique to one individual. Together, the high detection rate and heterogeneity of genetic findings among patients with CKD of unknown aetiology reinforce the concept that CKD is a complex disease, with many monogenic aetiologies noted to result in clinically ambiguous presentations that cannot be resolved using traditional diagnostics alone.

Investigations of individuals clinically diagnosed with CKD of various causes further support this concept. These studies report diagnostic findings for monogenic forms of kidney disease in 10–43% of patients^100-105, consistent with a contribution of monogenic aetiologies to CKD across clinical disease subtypes (Supplementary Table 1). In addition to demonstrating a high level of genetic heterogeneity, with 55–77% of all diagnostic genetic findings occurring as singletons, these analyses also highlight the phenotypic complexity of CKD, with up to 78% of genetic findings recurring in patients within different clinical diagnostic categories. For example, diagnostic mutations in COL4A3, COL4A4 and COL4A5 are found not only in patients with clinically diagnosed Alport syndrome, thin basement membrane disease and hereditary FSGS^106-110 but also among patients clinically diagnosed with other forms of CKD, such as CAKUT, cystic renal disease and hypertension-attributed nephropathy, and among patients with CKD of unknown aetiology^{100,101,103,104,111}. This apparent phenotypic broadening may in part reflect that many monogenic nephropathies, including those associated with mutations in the type IV collagen genes^{101,103,107,108,112}, can present with non-specific manifestations, which might impede accurate clinical diagnosis using traditional modalities alone. Wider usage of genetic testing could therefore contribute to such observations of a more diverse spectrum of phenotypes for type IV collagen-associated renal disease^{100,101,103,104,111}. In addition to encompassing clinically distinct CKD subtypes, variation within a single gene may have different effects on disease risk. For example, although rare variants in COL4A3 are associated with an increased risk of type IV collagen-associated nephropathy, a 2019 study identified a common missense variant in COL4A3 that had a protective association with diabetic kidney disease, wherein mutation carriers showed lower rates of albuminuria and kidney failure than non-carriers¹¹³. Moreover, although certain monogenic nephropathies are recurrently found by exome sequencing of individuals with all-cause CKD, genetic findings among these diagnostic cases encompass a wide range of clinical CKD subtypes, and for each CKD subtype, many distinct single-gene aetiologies are observed (FIG. 3).

Fig. 3 ∣. Genetic and phenotypic heterogeneity observed on exome sequence analysis of patients with all-cause chronic kidney disease.

a ∣ Genetic spectrum of positive cases from exome sequence analysis of 3,315 patients with all-cause chronic kidney disease¹⁰⁰. Of the 66 distinct monogenic disorders detected, findings in six genes accounted for 63% of all genetic diagnoses, with autosomal-dominant polycystic kidney disease resulting from mutations in PKD1 (75 cases; 24%) or PKD2 (22 cases; 7%); type IV collagen-associated nephropathy secondary to mutations in COL4A3 (27 cases; 9%), COL4A4 (21 cases; 7%) or COL4A5 (44 cases; 14%); and autosomal-dominant tubulointerstitial kidney disease caused by mutations in UMOD (9 cases; 3%). The remaining 37% of genetic diagnoses encompassed 60 different monogenic disorders. b ∣ The 60 other monogenic disorders spanned a diverse range of clinical chronic kidney disease subtypes, each of which included many distinct single-gene aetiologies, including cystic or congenital renal disease (30 genes; 50%); glomerulopathy (10 genes; 17%); tubulointerstitial disease (13 genes; 22%); and other causes of nephropathy (7 genes; 12%). Percentages do not total 100 because of rounding.

The variable detection rates of monogenic causes of kidney disease observed across studies of all-cause CKD probably reflect the corresponding variability in the clinical characteristics of the patients studied. For example, higher yields have been reported by studies of individuals preselected for features consistent with hereditary forms of nephropathy, such as familial and/or early-onset disease and/or a clinical diagnosis of a form of CKD thought to arise primarily from monogenic causes (for example, cystic or congenital renal disease). In a broad study of a largely adult cohort of over 3,300 individuals with all-cause renal disease¹⁰⁰, diagnostic yield was nearly 10% — a rate similar to that observed for other complex diseases for which genetic testing is frequently used, such as all-cause cancer^114,115. Moreover, diagnostic yield was higher among patients with familial CKD and/or those clinically diagnosed with cystic or congenital renal disease, consistent with the view that monogenic aetiologies contribute to nephropathy to a greater extent among patients with these features.

The majority of genetic studies to date have assessed individuals clinically diagnosed with CKD subtypes for which monogenic causes are thought to contribute substantially, such as SRNS or cystic renal disease. The knowledge gained from these investigations supports the use of precision diagnostics for patients clinically diagnosed with these forms of CKD. However, the majority of individuals with CKD are sporadic adult patients clinically diagnosed with CKD thought to be secondary to acquired causes, such as hypertension- or diabetes-associated renal disease, and scant data exist on the contribution of monogenic causes of disease in this population. Currently, a diagnosis of diabetic kidney disease or hypertensive kidney disease is often made presumptively on the basis of clinical history, without a confirmatory renal biopsy. Accordingly, review of biopsy samples from individuals clinically diagnosed with diabetic or hypertensive renal disease demonstrate that a substantial fraction of these individuals have biopsy findings consistent with other forms of CKD^116-120. These findings support the hypothesis that a proportion of patients with presumed diabetic kidney disease or hypertensive kidney disease represent cases of undiagnosed genomic disorders or monogenic nephropathy. The lower yield of genetic findings observed for patients diagnosed with diabetic kidney disease or hypertensive kidney disease in the large-scale diagnostic sequencing studies of patients with all-cause CKD¹⁰⁰, and the fact that relatively few causal loci with large effect size have been identified by genetic association studies in large cohorts^121-124, both suggest that in many cases, these forms of CKD may have a complex, polygenic basis. Additional studies are needed to better understand the genetic architecture of the variety of clinically defined subtypes of CKD, including the extent to which monogenic causes contribute to the pathogenesis of each.

Complexities of genetic analysis in CKD

Analytical challenges

Although MPS-based analysis of exonic (coding) regions frequently provides a genetic diagnosis, it has important technical limitations for identifying disease-causal variation. Thus, if a monogenic disease is strongly suspected but no diagnostic variants are detected by exome sequencing, these limitations should be considered and addressed as needed. First, some regions, such as the CAKUT-causing gene GREB1L, are not targeted by all exome sequencing kits^125-127; thus, a negative test result should prompt assessment of whether all nephropathy-associated coding regions were adequately captured by the sequencing approach. In addition, as exome sequencing generally has more variable coverage and lower analytical sensitivity for insertion and deletion variants than does genome sequencing, it can therefore miss coding variants and moderate sized deletions and duplications that can be detected by genome sequencing¹²⁸. Moreover, exome sequencing can miss certain classes of functional variants that can cause monogenic disease, such as intronic variants that affect splicing, mobile genetic elements that result in retrotransposition and mosaic variants^129-133. Usage of genome sequencing and long-read technologies with specialized bioinformatics approaches can support identification of these types of variants.

In addition, integration of other omics technologies can further empower these analyses, helping to achieve a genetic diagnosis among patients for whom exome sequencing alone is insufficient^{130,131,134-136}. For instance, although non-coding variants can be detected via genome sequencing, interpretation of their pathogenicity remains challenging. Integration of transcriptome sequencing with genome sequencing data can help to identify splice-altering intronic variants^129,137 and has been shown to identify diagnostic variants in 7.5% and candidate causal variants in an additional 16.7% of patients with other, non-renal rare diseases who had negative findings by exome sequencing alone¹³⁸. In one study, for example, among 4 patients with suspected type VI collagen-related muscular dystrophy, for whom targeted sequencing of the disease-associated genes (COL6A1, COL6A2 and COL6A3) had failed to detect a causative variant, use of exome sequencing, genome sequencing and RNA sequencing identified a splice-altering intronic variant in COL6A1 that resulted in insertion of 24 amino acids into the highly conserved N-terminal triple-helical collagenous domain of the encoded protein¹²⁹. In this study, subsequent investigation of another 637 individuals with genetically unresolved type IV collagen-related muscular dystrophy detected an additional 27 patients who harboured this variant. Importantly, the mutant transcript was found to be present at appreciable levels in muscle, but at much lower levels in cultured skin fibroblasts, highlighting the importance of tissue-specific transcriptome analysis.

Phenotypes from multilocus variation

Several large-scale genome-wide sequencing studies of individuals with a variety of complex phenotypes have demonstrated that at least 5% of patients in whom diagnostic genetic findings are identified harbour dual genetic diagnoses, with pathogenic variants in two different disease-associated genes^139-141. These findings challenge the principle of diagnostic parsimony that is frequently taught in clinical medicine. In support of these findings, sequencing investigations of individuals with CKD have also identified patients with diagnostic variants spanning multiple monogenic nephropathies^{25,48,102,103}. Similarly, a large microarray study of patients with all-cause CAKUT⁴⁹ found that 6.9% of individuals harboured multiple CNVs diagnostic for different genomic disorders. Importantly, these individuals manifested more severe forms of CAKUT and higher rates of extra-renal anomalies than those with CNV findings diagnostic for a single genomic disorder, suggesting that the aggregate burden of genomic disorders mediated the penetrance and expressivity of the disease.

Although the scale of the sequencing studies to date does not enable examination of genetic interactions between monogenic disorders, these observations collectively suggest that biological interactions might underlie the joint occurrence of these disorders. Phenotypes resulting from multilocus variation may present as a more severe form of a single, clinically defined disease, overlapping symptoms that are potentially attributable to one of a number of genetic conditions detected, or as an entirely novel condition that involves features unique to each of these genetic conditions¹³⁹. These complicated presentations might delay diagnosis. Hence, dual diagnoses should be suspected and genome-wide approaches should be considered when a hereditary disease is suspected but the patient’s presentation cannot be explained by a single diagnosis.

Phenotypic heterogeneity

The genetic mechanisms of phenotypic heterogeneity in CKD can be explored by considering three disorders that are observed frequently in MPS-based studies of all-cause nephropathy^100-104: autosomal-dominant polycystic kidney disease (ADPKD), type IV collagen-associated nephropathy and ADTKD. Although these monogenic nephropathies are traditionally considered to be clinically homogeneous conditions, investigations over the past 10 years have demonstrated substantial genetic and phenotypic complexity.

Autosomal-dominant polycystic kidney disease.

ADPKD classically presents with bilaterally enlarged, multicystic kidneys. Extra-renal manifestations include liver cysts, intracranial aneurysms and, less commonly, cysts in other organs and cardiac disease¹⁴². However, the severity of the renal and extra-renal manifestations vary substantially, from very early onset ADPKD, which presents in utero with oligohydramnios and hyperechogenic enlarged kidneys, to isolated bilateral renal cysts with normal kidney function¹⁴³. Although previously thought to arise from mutations in only two genes — PKD1 and PKD2 — the disease has been found to be genetically as well as phenotypically heterogeneous, with identification of novel causal genes, including GANAB¹⁴⁴, DNAJB11 (REF.¹⁴⁵) and ALG9 (REF.¹⁴⁶) in the past 5 years. ADPKD is traditionally diagnosed by imaging; however, the genetic and phenotypic heterogeneity of ADPKD, as well as findings of a strong genotype–phenotype relationship, have led some clinicians and researchers to call for a genetics-based framework for the clinical classification of ADPKD¹⁴³. Importantly, however, genetic diagnosis of ADPKD remains challenging, as PKD1 — the most commonly mutated gene in ADPKD — is large, and has a high guanine–cytosine (GC) content with a high degree of pseudohomology. Use of long-range PCR or genome sequencing can increase the technical sensitivity of genetic testing for PKD1 (REF.¹⁴⁷). Moreover, achieving a genetic diagnosis may require assessment of variants other than protein-altering mutations in the coding regions, as synonymous variants¹⁴⁸ and intronic variants¹⁴⁹ that lead to altered splicing of PKD1 can also result in ADPKD; integration of RNA-based functional assays can help to identify such variants and ascertain their effect on the encoded polycystin 1 (PC1) protein.

Investigations of familial cases of ADPKD¹⁵⁰ and longitudinal studies of ADPKD case cohorts^151,152 highlight the importance of gene- and allele-level variation in mediating disease severity, as well as the importance of identifying other genetic modifiers. ADPKD is thought to arise from insufficiency of the PC1 or polycystin 2 (PC2) proteins, which normally inhibit cyst formation¹⁴². Accordingly, disease most frequently results from loss-of-function mutations in the genes that encode these proteins — PKD1 and PKD2, respectively — which together account for at least 90% of ADPKD cases¹⁵³. The finding that cysts only develop in a proportion of renal tubular epithelial cells has led to the proposal that cystogenesis in ADPKD occurs through a two-hit model^154,155, whereby in addition to a heterozygous germline mutation, cyst formation requires a second somatic mutation in the normal allele, resulting in full insufficiency of PC1 or PC2.

In addition to mutations in PKD1 and PKD2, mutations in genes involved in the maturation and trafficking of PC1 (REFS¹⁴²,¹⁵⁶), including GANAB¹⁴⁴, DNAJB11 (REF.¹⁴⁵) and ALG9 (REF.¹⁴⁶), can lead to ADPKD. However, despite the fact that mutations in these genes disrupt a common biological pathway, the phenotypes associated with mutations in these genes vary greatly. For example, mutations in PKD1 generally confer a severe form of the disease with enlarged, multicystic kidneys, hepatic cysts and a high risk of progression to kidney failure by 60 years of age, whereas mutations in GANAB¹⁴⁴ seem to cause mild cystic renal disease with variable liver involvement that rarely progresses to kidney failure^143,144. In addition to differing according to the gene involved, ADPKD phenotypes also vary by mutation type. For instance, protein-truncating mutations in PKD1 are associated with more severe disease, characterized by larger height-adjusted total kidney volume, lower estimated glomerular filtration rate and earlier age of onset of kidney failure^151,157, than are non-truncating variants, reflecting that protein-truncating mutations inactivate PC1 to a greater extent. Such strong genotype–phenotype correlations have enabled the creation of prognostic algorithms such as the predicting renal outcomes in ADPKD (PROPKD) score¹⁵⁸, which integrates knowledge of the causal gene (for instance, PKD1 versus PKD2) and mutation type (for instance, truncating versus non-truncating mutations) with other clinical features, such as gender and age at onset of hypertension, to predict renal outcomes in affected individuals.

ADPKD also shows substantial intrafamilial variability. For example, a 2019 study of over 600 families reported that 12% of families had members with mild disease and members with severe disease¹⁵⁰. This heterogeneity might arise in part through complex genetic mechanisms. For example, individuals can present with ADPKD without a clinically apparent family history of disease as a result of parental mosaicism¹⁵⁹. As mosaic variants are variably distributed and can be present in only a small fraction of an individual’s total cells, resulting in a low signal-to-noise (that is, the mutant to normal allele) ratio, they can be missed by traditional Sanger sequencing¹⁶⁰. Screening of multiple cell types using MPS with very high sequencing coverage and read depths can help to surmount these challenges, supporting the detection of mosaic variants^147,160. Digenic inheritance can also lead to intrafamilial variability. Thus, co-inheritance of PKD1 and PKD2 variants can yield severe, early-onset ADPKD, whereas family members who harbour only one of the two variants typically present with milder disease^161,162. Similarly, in families with PKD1- and PKD2-associated ADPKD, individuals who have mutations in genes associated with other hereditary nephropathies, such as HNF1B¹⁶³ and COL4A1 (REF.¹⁶⁴), display more severe cystic disease and earlier onset of kidney failure than those who have only the PKD1 or PKD2 mutation.

Type IV collagen-associated nephropathy.

As described earlier, mutations in COL4A3, COL4A4 and COL4A5 are associated with type IV collagen-associated nephropathies, which include Alport syndrome, thin basement membrane disease and FSGS^106-110. These genes encode the collagen IV α3, α4 and α5 proteins, which together form the collagen IV α3α4α5 trimer — a key constituent of basement membranes in the glomerulus, eye and inner ear¹⁶⁵. Accordingly, pathogenic mutations in any one of the three genes can result in disease, which classically features haematuric nephropathy with progression to renal failure, ocular anomalies and sensorineural hearing loss^107,108. MPS-based sequencing of the type IV collagen genes has been recommended as a time- and cost-effective means of diagnostic screening¹⁰⁸. However, this approach has limited technical sensitivity for certain types of causal mutation, such as deep intronic variants that alter splicing and small genic deletions, and can also miss mosaic variants owing to insufficient sequencing depth of coverage¹⁰⁸. Thus, more comprehensive high-coverage sequencing of non-coding regions, transcriptome analysis and functional splicing assays can help to resolve diagnosis in individuals who remain undiagnosed using standard MPS-based analysis of coding regions of type IV collagen genes^166,167.

The phenotypic spectrum resulting from mutations in type IV collagen is diverse, with presentations ranging from isolated microhaematuria with normal renal function to early-onset kidney failure with ocular and audiological involvement^107,108. This phenotypic heterogeneity is partially attributable to effects at the genotype level. Among individuals with kidney disease resulting from mutations in COL4A3 or COL4A4, those with (two) biallelic variants show more severe disease, with earlier-onset kidney failure and more frequent extra-renal involvement, than those who are heterozygous for a single variant^168,169. Similarly, among individuals with kidney disease due to mutations in COL4A5, (hemizygous) males are more severely affected than (heterozygous) females^170,171. Although variable, case reports¹⁷² and studies in mouse models¹⁷³ suggest that phenotypes among heterozygous females are shaped by the pattern of X chromosome inactivation, with greater inactivation of the X chromosome harbouring the wild-type allele associated with more severe disease than with lesser inactivation of this chromosome¹⁷². Phenotypic variability can also occur among males with COL4A5-associated disease caused by somatic mosaicism, with these males presenting with atypically mild forms of kidney disease¹⁷⁴. Mutation type also seems to have an important role in determining phenotype. Among individuals with type IV collagen variants in the same gene, those who harbour protein-truncating mutations or glycine substitutions in the collagenous domain generally manifest more severe disease than those who harbour mutations that result in premature protein truncation or a collagenous domain glycine substitution^175,176. These more severe outcomes reflect the fact that these types of mutation either lead to absence of the encoded type IV collagen protein as a consequence of nonsense-mediated decay, or to structural instability resulting from interruption of the highly conserved Gly-Xaa-Yaa repeats, thereby impeding normal trimer assembly^107,108,165.

To date, however, our understanding of the molecular basis of the phenotypic heterogeneity found among individuals heterozygous for COL4A3 and COL4A4 mutations is limited. Although heterozygotes generally show milder disease than those with autosomal-recessive or X-linked disease^168,177,178, some individuals with heterozygous mutations show phenotypes of similar severity to those of patients with biallelic or hemizygous mutations^179-181. Interestingly, a 2019 population-based genetic study suggested that type IV collagen mutations have highly variable penetrance as well as expressivity, reporting a heterozygous protein-truncating variant in COL4A3 among population controls without any apparent association with kidney impairment¹⁸². As heterozygous type IV collagen mutations encompass a diverse phenotypic spectrum and have a variable impact on kidney disease risk, interpretation of their pathophysiological relevance is challenging. Additional investigations, such as analyses that integrate genome-wide sequence data with longitudinal phenotypic data from patients with CKD and unselected population controls, as well as saturation editing^183,184 studies to assess the functional consequences of a given mutation, will help to clarify the role of these variants in different forms of CKD, facilitating genetic diagnosis and supporting development of targeted therapies.

Autosomal-dominant tubulointerstitial kidney disease.

ADTKD describes a set of rare nephropathies characterized by the non-specific findings of progressive CKD with tubulointerstitial fibrosis on renal biopsy^185,186. To date, mutations in any one of five genes — UMOD, MUC1, REN, HNF1B and SEC61A1 — have been implicated in ADTKD^185,186; an ADTKD-like phenotype has also been noted among families with segregating mitochondrial gene mutations¹⁸⁷. Although defined by a common histopathology, ADTKD can vary substantially by molecular subtype, corresponding to biological functions specific to the given gene. For example, individuals with disease caused by mutations in the gene that encodes renin, REN (ADTKD-REN), display the additional features of early-onset anaemia as well as mild hypotension and hyperkalaemia, reflecting the involvement of the renin-angiotensin-aldosterone system in erythropoietin production, blood-pressure regulation and potassium homeostasis^185,188. Similarly, the phenotype resulting from mutations in SEC61A1 can include primary immunodeficiency^189,190, consistent with the role of the encoded protein in plasma cell differentiation and immunoglobulin synthesis¹⁹⁰.

Given its non-specific symptomatology and histopathology, ADTKD can be difficult to detect using traditional clinical diagnostics alone, supporting the use of genetic testing¹⁸⁵. Moreover, the fact that some of the major known causal genes have mutational hot spots can facilitate diagnostic interpretation of variants. For example, the majority of disease-causal variants in ADTKD-UMOD are missense mutations that cluster in exons 3–4 (REF.⁴⁶). However, in certain cases, causal variants cannot be detected using standard MPS-based testing. For example, ADTKD-MUC1 results from a cytosine insertion mutation in a long, GC-rich variable-number tandem repeat sequence of the MUC1 gene¹⁹¹. The high GC content and repetitive nature of the region impede detection using MPS-based sequencing and, instead, molecular diagnosis requires other assays, including long-range PCR¹⁹¹ and immunostaining for the encoded (mucin 1) protein¹⁹². Broader adoption of these alternative methodologies will facilitate identification of disease cases, supporting further longitudinal study of ADTKD and providing greater insight into genotype–phenotype correlations for the disorder.

Towards targeted therapies

Despite CKD being a collectively common disease, there is a paucity of targeted therapies. In part, this paucity reflects the challenge posed by the high aetiological and phenotypic heterogeneity of the disease, which makes it difficult to identify, understand and treat disease subtypes on the basis of their molecular pathogenesis^193,194. Genomics and other omics technologies have the potential to surmount this challenge by supporting discernment of different CKD subtypes at the molecular level, thereby enabling focused studies of affected individuals and the development of targeted therapies.

The use of pharmacological chaperone therapy, whereby small-molecule ligands known as pharmacological chaperones are used to stabilize and correct misfolded proteins, in Fabry disease^195-197 illustrates the potential of this type of targeted approach. Fabry disease results from mutations in GLA, which encodes the α-galactosidase enzyme. Fabry disease encompasses a diverse clinical spectrum: patients can present with the more severe, classic phenotype, a systemic disorder whose manifestations include progressive kidney failure, cardiac disease, cerebrovascular disease, peripheral neuropathy, skin lesions and ocular anomalies, or with the milder, non-classic phenotype, characterized by isolated cardiac disease and/or kidney disease. To date, 268 known disease-causal mutations that cause the enzyme to be misfolded but enable it to retain residual catalytic activity, and thus remain amenable to rescue by the pharmacological chaperone migalastat, have been identified¹⁹⁸. Importantly, these migalastat-responsive GLA mutations are found across the phenotypic spectrum of Fabry disease, among both patients with the classic phenotype and those with the non-classic phenotype, making it difficult to discern who may be a suitable candidate for pharmacological chaperone therapy on the basis of clinical presentation alone. Accordingly, a cellular assay¹⁹⁸ screen for patients with migalastat-responsive mutations has been used in clinical trials to identify which patients could benefit from this therapy^195-197.

In addition to targeting the encoded protein, precision therapies can also directly act at the genomic or transcriptomic level. For example, in sickle cell disease, which arises from the haemoglobin S (p.Glu6Val) variant in HBB, gene therapy via lentiviral transfer of a modified copy of HBB that encodes an anti-sickling allele has been shown to result in full remission¹⁹⁹. Among patients with hereditary transthyretin amyloidosis, use of oligonucleotide agents that target the mutant TTR RNA transcript has been shown to halt or reverse the progression of multiple disease manifestations, by inhibiting the synthesis and thereby systemic accumulation of misfolded transthyretin^200,201. In primary hyperoxaluria type 1, in which impaired glyoxylate metabolism leads to systemic oxalosis, CRISPR–Cas9-mediated gene editing²⁰² and mRNA silencing using small interfering RNAs²⁰³ are being investigated as potential approaches with which to inhibit synthesis of glycolate oxidase, the enzyme involved in glyoxylate production. Growing usage of genomic sequencing and other omics tools among patients with CKD will support the further development of precision therapeutics in nephrology.

Conclusions

Genomics medicine aims to provide personalized care based on an individual’s genetic information. However, understanding the relationship between genotype and phenotype for many forms of CKD is complicated by the complexity of rare genetic causes of CKD, impeding accurate assessment of an individual’s risk of disease and the identification of approaches for targeted management. Many monogenic nephropathies demonstrate variable and multisystem manifestations, reflecting phenomena including genetic pleiotropy, dosage sensitivity, incomplete penetrance and/or variable expressivity. Such phenotypic heterogeneity can hinder our ability to pinpoint a causal aetiology. Moreover, as patients can present atypically, individuals with monogenic forms of CKD can go clinically unrecognized. The high genetic heterogeneity observed within a single disease subtype also complicates the choice of optimal genetic testing modality. Thus, in many cases, genome-wide approaches such as exome sequencing or genome sequencing may provide superior diagnostic sensitivity over targeted panels¹⁰⁰, as the patient may have a subtype of CKD other than that clinically suspected or have one that has a multitude of potential causal genes. In either of these situations, use of a targeted gene panel may lead to a diagnostic genetic variant being missed, as the gene causal for the patient’s disease may not be included on the panel. However, the expanded scope of analysis offered by exome or genome sequencing also confers a greater burden of variant interpretation, which is augmented by the phenotypic complexity of CKD. The abundance of rare, putatively deleterious variations within any human genome results in a high potential for false-positive results^204,205. Moreover, both the variable penetrance and expressivity of many rare genetic causes of CKD and the lack of detailed, longitudinal clinical data available for control cohorts can make it difficult to discern whether findings of such variants among self-declared healthy individuals indicate that they are in fact benign, or instead result from cases of undiagnosed monogenic nephropathy²⁰⁶. Beyond impeding diagnosis, such phenotypic heterogeneity can make it difficult to provide a clear prognosis upon detecting a disease-causing variant. As many investigations of rare genetic causes of CKD have been performed in highly preselected populations, comprising individuals with clinical features consistent with the disorder under study, knowledge of the true prevalence, penetrance and frequency of the manifestations of these variants — which often encompass multiple organ systems — remains incomplete. This limited knowledge prevents accurate assessment of disease severity and rate of progression and thereby hinders targeted work-up, surveillance and management.

Addressing these challenges will require a multifaceted approach. Use of unbiased, genome-wide sequencing approaches and reverse phenotyping will increase diagnostic sensitivity and help to distinguish hereditary nephropathies that overlap in clinical presentation. Inclusion of multiple types of omics data in the diagnostic work-up, as well as in the research-level study of individuals with suspected monogenic nephropathies, will not only support the identification of disease-causing variants but also offer deeper insight into the molecular pathogenesis of these disorders^129,207. Refinement of existing guidelines for clinical sequence interpretation by working groups, such as those led by ClinGen²⁰⁸, will also facilitate the discernment of pathogenic versus benign variation. Similarly, as further large-scale genetic studies are performed in nephrology, regular review of the literature by such expert panels will help the field to maintain an up-to-date understanding of the contribution of the genetic and phenotypic heterogeneity of CKD (Supplementary Table 1) and support the creation of evidence-based guidelines for the use of genetic testing in nephrology. Moreover, periodic reanalysis of sequence data may enable the reclassification of variants of unknown significance or the detection of pathogenic variants in newly discovered genes, as illustrated by studies from the past few years, where reanalysis of exome data identified disease-causal variants in patients who had been previously sequenced, with non-diagnostic results^209-211. In addition, novel techniques that integrate genetic and phenotypic data from large, diverse cohorts, including genome-wide polygenic risk scores¹⁶ and phenotypic risk scores²¹² (BOX 1), will deepen our understanding of genotype–phenotype relationships, enabling accurate assessment of disease risk across different subtypes of CKD. Incorporation of these emerging methodologies with traditional tools such as clinical history and renal histopathology will help to unravel the complex phenotypes encompassed by CKD, empowering both genetic diagnosis and novel gene discovery for patients with nephropathy.

Key points.

• Chronic kidney disease (CKD) is a complex disorder comprising many rarer, pathophysiologically distinct conditions that encompass both monogenic and polygenic forms, which share the common feature of leading to persistent anomalies in renal structure and/or function.

• Rare hereditary causes of CKD often show high phenotypic heterogeneity, which can result from pleiotropy, incomplete penetrance or variable expressivity.

• Microarray and massively parallel sequencing studies have shown that both genomic disorders and monogenic diseases account for a meaningful proportion of cases across different clinical subtypes of CKD.

• Copy number variants at the 17q12, 22q11.2 and 16p11.2 loci are recurrent genomic disorders among patients with CKD and display diverse and highly variable multiorgan manifestations, which can reflect gene dosage sensitivity and epistatic and epigenetic effects.

• Although classically considered to be clinically homogeneous, common monogenic causes of CKD, such as autosomal-dominant polycystic kidney disease, type IV collagen-associated nephropathy and autosomal-dominant tubulointerstitial kidney disease, display variable penetrance and expressivity, in part due to gene-and allele-level variation.

• Multilocus variation, involving variants for multiple genetic conditions, can confer complex phenotypes and has been detected in at least 5% of positive cases from genome-wide testing.

• Novel techniques that integrate genetic sequencing, experimental assays and clinical data, such as reverse phenotyping, functional screening of potentially pathogenic variants, and genetic and phenotypic risk scores, will support greater understanding of the phenotypic complexity of different forms of CKD.

Glossary

Heritability

The proportion of interindividual variation in a trait that is caused by genetic factors.

Private variants

Genetic variants unique to a single individual (that is, variants not observed in other individuals).

Phenocopy

A phenotype that bears resemblance to the phenotype resulting from a particular genotype but occurs in an individual who does not harbour that genotype.

Pleiotropy

Variation in a single gene influencing multiple phenotypic traits.

Incomplete penetrance

Penetrance is the proportion of individuals with a particular genotype who display the associated phenotype. In the case of incomplete penetrance, not all individuals with the genotype manifest the associated phenotype.

Variable expressivity

Expressivity refers to the extent to which individuals with a given genetic disease display the associated phenotype and thus reflects the range of phenotypes associated with the genotype. Variable expressivity describes a situation wherein affected individuals display the associated phenotype to differing extents.

Papillorenal syndrome

Also known as renal coloboma syndrome. A disorder resulting from pathogenic variants in PAX2 characterized by kidney disease and ophthalmological anomalies; other manifestations may include sensorineural hearing loss, soft skin and ligamentous laxity.

Oligogenic inheritance

A form of inheritance in which transmission of a phenotypic trait is mediated by multiple different genetic loci

Digenic inheritance

A form of inheritance in which transmission of a phenotypic trait is mediated by two different genetic loci.

Structural variants

Large (≥1-kb) DNA variants; these alterations can be balanced (for example, inversions or reciprocal translocations, with no overall change in the amount of DNA at the relevant locus) or imbalanced (for example, copy number variants), which lead to gain or loss of DNA at the relevant locus.

Single nucleotide variants

Alterations of single bases (nucleotides) in a DNA sequence; single nucleotide variants can lead to a different amino acid sequence in the encoded protein (non-synonymous variants) or leave the sequence unchanged (synonymous variants).

Insertions or deletions

Gains or losses in the number of bases in a DNA sequence versus the reference sequence at that site, producing a different amino acid sequence in the encoded protein.

Copy number variants

(CNVs). Structural variants that lead to gain (duplications) or loss (deletions) of DNA at the relevant locus.

Genic copy number variants

Copy number variants that contain protein-coding regions of the genome (that is, genes).

Locus

A specific site in the genome (plural: loci).

Haploinsufficiency

A condition resulting from inactivation of one copy of a gene for which two copies are needed for normal gene function; for such haploinsufficient genes, gene function is thus altered in heterozygotes, as the remaining (functional) copy does not produce sufficient gene product for normal function.

Meiotic non-allelic homologous recombination

Exchange of genetic material (DNA) between homologous regions located on different loci during meiosis.

Epistatic interactions

Interaction between multiple different loci, wherein the phenotypic impact of the genotype at one locus depends on the genotype at other loci.

Mobile genetic elements

DNA sequences that can move from their original site and integrate into another, different region of the genome

Retrotransposition

Insertion of a DNA sequence at a new site in the genome using an RNA intermediate: the DNA sequence is first transcribed into RNA, and the RNA is then reverse transcribed into DNA, which is then inserted at the new site.

Mosaic variants

Genetic variants that are present in a mosaic form — that is, the genotype at that locus is present within a certain proportion of an individual’s cells, such that they have multiple genetically different cell populations, with different genotypes at that locus.

Pseudohomology

The state in which the DNA sequence of a (protein-coding) gene is highly similar to the DNA sequence of a pseudogene (that is, a copy of the gene that is not transcribed or translated, and thus does not yield a protein), because they both originate from the same ancestral gene.

Synonymous variants

DNA sequence changes in the protein-coding (exonic) regions of the genome that do not alter the amino acid in the associated encoded proteins.

Sanger sequencing

A method of DNA sequencing that uses labelled chain-terminating dideoxynucleotides to detect the nucleotides in the DNA strand being sequenced. This method yields a sequence chromatogram, which can subsequently be analysed to identify genetic variants.

Sequencing coverage and read depths

In this Review, sequencing coverage refers to the percentage of bases in the DNA region targeted by sequencing that is sequenced a given number of times. Sequencing depth denotes the average number of times that a given nucleotide is read in a set of DNA sequence reads. Higher coverage and depth means that more of the targeted genomic region has been sampled a greater number of times, increasing the technical accuracy of the resulting data.

Hemizygous

The condition in which an individual has a single copy of a pair of chromosomes or a segment of a chromosome pair, rather than two copies. This occurs for genes located on the X chromosome among human males, as their sex chromosomes comprise one X chromosome and one Y chromosome

Saturation editing

A technique that replaces a DNA sequence in a gene of interest with variant DNA sequences encoding alternative amino acids in order to generate all possible amino acid substitutions at that site in the protein. The impact of each of the mutations on wild-type (non-mutated/normal) protein function can then be evaluated through various high-throughput screening assays (for example, for a gene encoding a kinase enzyme, assessing its ability to phosphorylate its substrate).

Transthyretin amyloidosis

A disorder resulting from systemic amyloid deposition in organs including the heart, liver, nervous system and kidney owing to mutations in TTR.