Genomic approaches in the search for molecular biomarkers in chronic kidney disease

Abstract

Background

Chronic kidney disease (CKD) is recognised as a global public health problem, more prevalent in older persons and associated with multiple co-morbidities. Diabetes mellitus and hypertension are common aetiologies for CKD, but IgA glomerulonephritis, membranous glomerulonephritis, lupus nephritis and autosomal dominant polycystic kidney disease are also common causes of CKD.

Main body

Conventional biomarkers for CKD involving the use of estimated glomerular filtration rate (eGFR) derived from four variables (serum creatinine, age, gender and ethnicity) are recommended by clinical guidelines for the evaluation, classification, and stratification of CKD. However, these clinical biomarkers present some limitations, especially for early stages of CKD, elderly individuals, extreme body mass index values (serum creatinine), or are influenced by inflammation, steroid treatment and thyroid dysfunction (serum cystatin C). There is therefore a need to identify additional non-invasive biomarkers that are useful in clinical practice to help improve CKD diagnosis, inform prognosis and guide therapeutic management.

Conclusion

CKD is a multifactorial disease with associated genetic and environmental risk factors. Hence, many studies have employed genetic, epigenetic and transcriptomic approaches to identify biomarkers for kidney disease. In this review, we have summarised the most important studies in humans investigating genomic biomarkers for CKD in the last decade. Several genes, including UMOD, SHROOM3 and ELMO1 have been strongly associated with renal diseases, and some of their traits, such as eGFR and serum creatinine. The role of epigenetic and transcriptomic biomarkers in CKD and related diseases is still unclear. The combination of multiple biomarkers into classifiers, including genomic, and/or epigenomic, may give a more complete picture of kidney diseases.

Introduction

Chronic kidney disease (CKD) is recognised as a global public health problem [1] with adjusted CKD prevalence ranging between 3.3 and 17.3% in adult European populations [2]. CKD is more prevalent in older persons and is associated with multiple co-morbidities including an increased risk of cardiovascular disease (CVD). Diabetes mellitus (DM) and hypertension are common aetiologies for CKD. Other common causes of CKD include autoimmune renal diseases such as IgA glomerulonephritis (IgAN), membranous glomerulonephritis (MGN) and lupus nephritis (LN). Autosomal dominant polycystic kidney disease (ADPKD) is the commonest genetic disorder causing CKD [3].

The 2012 Kidney Disease: Improving Global Outcomes (KDIGO) Clinical Practice Guideline for the Evaluation and Management of CKD, developed by the Kidney Disease Outcomes Quality Initiative (KDOQI) of the National Kidney Foundation (NKF), recommends the use of estimated glomerular filtration rate (eGFR) in the evaluation, classification, and stratification of CKD [3]. The guidelines, categorise CKD into five stages based on eGFR measurements (Table 1) [3, 4]. The most widely used eGFR equation is derived from four variables (serum creatinine (SCr), age, gender and ethnicity), and is recommended by the KDOQI guidelines for initial assessment of kidney function [4].

Table 1 Categorisation of chronic kidney disease according to estimated glomerular filtration rate [4]

Although SCr assays are routinely available in clinical practice there are some limitations to its use as creatinine is influenced by muscle mass, exercise, age, gender, and ethnicity. Furthermore, ~ 50% of kidney function can be lost before the SCr rises above the normal laboratory range [5, 6]. An eGFR equation, based on SCr (eGFRcrea), is not an accurate measure for early stages of CKD, in elderly individuals with low muscle mass and in those with extreme body mass index values [7]. A second equation, based on the measurement of serum cystatin C, has been proposed as an alternative [4]. The use of eGFR based on cystatin C (eGFRcys) is recommended by the KDOQI guidelines as a confirmatory test to diagnose CKD in those specific circumstances when eGFR-creatinine is less accurate [3, 4]. However, serum cystatin C measurements also have recognised limitations since the circulating cystatin C level can be increased by inflammation, steroid treatment and thyroid dysfunction [8]. Urinary albumin, the principal component of urinary protein in most kidney diseases, is also routinely used as a marker of kidney damage. Although albuminuria may be an early sign of kidney disease in glomerulonephritis (an can occur prior to a decrease in glomerular filtration rate) it is not a universal feature of CKD [3]. Repeated assessments, employing the existing eGFR equations and urinary albumin measurements, can help to identify persons with CKD. There is however an ongoing need to identify additional non-invasive biomarkers that are useful in clinical practice to help improve CKD diagnosis, inform prognosis and guide therapeutic management.

CKD is a multifactorial disease with associated genetic and environmental risk factors. The increasing need to identify CKD patients at earlier stages and improve stratification of their risk for progression to end-stage renal disease (ESRD) has prompted many studies of novel and existing biomarkers for kidney disease in large cohorts of patients. Various approaches have been employed, from candidate single gene studies to genome wide multi-omic studies. Among them, many genetic, epigenetic, and transcriptomic studies have been developed in the hope that the introduction of novel technologies and unbiased approaches would allow the identification of new biomarkers for CKD that would also contribute to biological understanding of renal disease.

In this review, we will summarise the most important studies in humans investigating genomic biomarkers for CKD in the last decade.

Genetic biomarkers

A genetic susceptibility to CKD exists although there is limited evidence for the individual gene variants responsible for this risk across the multiple aetiologies of CKD [9, 10]. Unlike Mendelian kidney disorders, such as ADPKD, where the causal variants are in the PKD1 and PKD2 genes [11], CKD is considered to be a polygenic disease in which many common, low-penetrance variants influence disease development and progression [12]. In the past, studies of candidate genes from pathways associated with renal function and linkage analysis in extended family pedigrees with multiple affected generations have been utilised to identify causal variants which contribute to CKD [13, 14]. However, both approaches have their drawbacks; candidate gene studies require knowledge of disease pathways and bias is introduced in gene selection, while genome-wide linkage analyses do not reflect genomic variation across all populations, as analyses are carried out within families [15]. More recently, genome-wide association studies (GWAS) have been employed to detect genetic variations in large numbers of unrelated individuals. For over 10 years, GWAS have been employed to identify single nucleotide polymorphisms (SNPs) associated with CKD and/or a range of traits associated with renal function. Serum creatinine (SCr), eGFR and urinary albumin-to-creatinine ratio (UACR), have been investigated using GWAS in different aetiologies of CKD and ESRD, such as diabetic kidney disease (DKD), MGN and IgAN. Genetic variants in over 50 loci have been associated with CKD or measures of renal function in different populations. Several of these gene variants are implicated in pathways driving the pathogenesis of CKD. The relevance of GWAS for the identification of SNPs associated with kidney transplantation outcomes and their potential use as genetic biomarkers for risk stratification or donor selection has been recently reviewed [16]. A small number of genetic variants were associated with graft function, T cell mediated rejection and tacrolimus trough levels, but these markers have not been validated in multiple different populations [16].

Various genetic loci have been associated with CKD (Table 2), but by far the most widely reported and consistently replicated genetic variants lie within the UMOD gene. UMOD codes for the protein uromodulin (previously known as Tamm-Horsfall [17], which is produced in the kidneys and is the most abundant urinary protein [18]. Loss of UMOD in mice leads to dysfunction of transmembrane solute transporters in the loop of Henle [19]. Mutations in UMOD have been associated with autosomal dominant tubulointerstitial kidney disease [20] which can progress to CKD [21]. Discovery analysis of 2388 CKD patients and 17,489 non-CKD controls and replication analysis of 1932 CKD patients and 19,534 controls of European ancestry, found UMOD variant rs12917707 to be significantly associated with both CKD and eGFR [22]. The same group confirmed these findings in a larger study in a total of 7173 European patients across the discovery and replication analyses [23]. A meta-analysis of population-based studies comprising over 130,600 individuals also found significant association with rs12917707 in 6271 cases of CKD (p = 3.7 × 10⁻¹⁶) and significant association in 2181 patients whose CKD was more severe (eGFR < 45 ml/min/1.73 m²; p = 1.1 × 10⁻⁰⁵) [24]. An additional variant (rs4293393) in the UMOD gene was significantly associated with CKD (p = 4.1 × 10⁻¹⁰) in a study of 3203 Icelandic CKD patients and 38,782 controls [25]. This variant was replicated in a larger study of 15,594 Icelandic CKD patients (p = 9.1 × 10⁻³⁸), in which another UMOD variant (rs11864909) was found to be significantly associated with CKD in a combined analysis (p = 2.2 × 10⁻¹⁹). Another UMOD variant, rs13329952 was significantly associated with CKD in a fixed-effects meta-analysis on a total of 151,137 individuals, of which 16,630 had CKD (p = 1.98 × 10⁻²⁵); variant rs12917707 was also replicated in this meta-analysis (p = 1.16 × 10⁻⁴¹) [26].

Table 2 Genomic-wide association studies in chronic kidney disease

There are ethnic differences in the incidence and prevalence of CKD; the incidence of ESRD is almost five times higher in African Americans than Americans of European descent [27]. A study of 1372 African American ESRD patients and 806 controls identified several variants in the myosin heavy chain type II isoform A (MYH9) gene as being specifically associated with non-diabetic ESRD in African Americans [28]. To further investigate this association, ESRD patients were sub-divided by presence of diabetes. No association was found in diabetic ESRD patients, but significance of MYH9 variants was retained for the non-diabetic ESRD group [28]. Similar association was found in variants spanning exon 14–23 of the MYH9 gene in 852 focal segmental glomerulosclerosis (FSGS) and 433 non-diabetic ESRD patients, but not for 476 diabetic-ESRD patients when compared to 222 controls [29]. The results for ESRD patients were consistent with the first analysis, and both studies described a specific association between variant rs735853 and non-diabetic ESRD patients, but not with diabetic ESRD patients [28, 29]. A study of 464 non-diabetic ESRD patients and 478 controls, replicated in 336 non-DM ESRD cases and 363 controls, identified 16 SNPs associated with non-diabetic ESRD, 12 of which were found in/near MYH9 [30]. To reinforce the specificity of these variants in African American individuals, a sufficiently powered study in 23,812 European patients found no association of MYH9 with eGFR or CKD [31]. Despite the lack of previous association with diabetic-ESRD, two subsequent studies found significant association of MYH9 with type 2 diabetes mellitus (T2DM)-ESRD [32, 33]. The MYH9 gene codes for myosin non-muscle myosin heavy chain IIA protein [34], a protein involved in cytokinesis, chemotaxis and differentiation of non-muscle cells [35]. Mutations in this gene are responsible for a number of autosomal dominant conditions and patients, such as those with Epstein Syndrome, can develop nephritis or other renal abnormalities [36]. This known link to renal function further supports the role of MYH9 as a genomic biomarker in CKD, although further studies in patients of European descent with CKD may be required to determine its suitability a biomarker across populations or should be limited to African Americans.

However, strong linkage disequilibrium found among variants in MYH9 and the adjacent APOL1 gene made it unclear as to which variants were associated disease [37]. With the release of data from the 1000 genomes project, two independent groups identified variants in APOL1 which were associated with non-diabetic ESRD [38, 39]. In a study of 1030 non-diabetic ESRD patients and 1025 non-diabetic non-ESRD controls, two variants in APOL1 (rs73885319 and rs71785313, referred to as G1 and G2, respectively) reached genome-wide significance (G1: p = 1.1 × 10⁻³⁹, G2: p = 8.8 × 10⁻¹⁸) [39]. The same study found similar association of G1 (p = 1.07 × 10⁻²³) and G2 (p = 4.38 × 10⁻⁷) with FSGS in 205 FSGS African Americans patients and 180 non-FSGS controls [39]. After adjusting for G1 and G2 using logistic regression, association with MYH9 was lost for both non-diabetic ESRD and FSGS, but association with variants in APOL1 remained when adjusting for variants in MYH9 [39]. An independent group was carrying out similar analysis at the same time in 346 African American and 84 Hispanic American ESRD patients and 147 African American and 378 Hispanic American controls [38]. In this study they also found that variants in APOL1 (rs73885319 and rs60910145) were much more strongly associated with ESRD than all previously reported variants in MYH9 [38]. The G1 and G2 risk alleles were also found to be associated with a greater risk of developing CKD (OR = 1.49, CI 1.02–2.07) and greater risk of progression to ESRD (OR = 1.88, CI 1.20–2.93) in 3067 African American individuals did not have CKD at baseline, of which 190 went on to develop CKD and 114 developed ESRD [40].

APOL1 codes for apolipoprotein L1 (APOL1), a protein which interacts with high density lipoprotein (HDL) cholesterol in plasma [41]. APOL1 was identified as a protective factor in human sleeping sickness, common in Sub-Saharan Africa [42], and it has been shown that only variants in APOL1 linked with renal dysfunction confer protection from disease [39]. Therefore, these variants may have been passed on due to their protective effect against human sleeping sickness in Africans, resulting in higher prevalence of APOL1 variants in individuals of African origin, compared to those of Europeans.

Other loci have shown association with CKD, such as PRKAG2 [23, 24, 26, 43] and WDR37 [23, 24, 26, 43], variants within these genes are reported less often than those in UMOD or MYH9/APOL1.

eGFR

Variants in the UMOD gene have demonstrated the ability not only to predict the risk of CKD, but also changes in eGFR. The rs12917707 variant was the first SNP in UMOD to reach genome-wide significance for both SCr eGFRcrea and eGFRcys [22]. Estimating GFR independently of creatinine discriminates between association derived from renal function and from creatinine production or metabolism. Meta-analysis of discovery (n = 19,877) and replication (n = 21,466) analyses of independent European patient cohorts showed significance of UMOD with eGFRcrea (p = 3.0T × 10⁻¹¹) and eGFRcys (p = 2.0 × 10⁻⁰⁷), showing UMOD is associated with renal function decline and not creatinine production or metabolism [22]. The rs12917707 variant has been associated with SCr/eGFR in patients of European ancestry in several other studies, both with and without diabetes [24, 44, 45]. It also showed a stronger association with older individuals when patient groups were stratified by age (p = 8.4 × 10⁻¹³) [24] and with significant changes in eGFR over time and kidney function decline in a total of 63,558 patients [45]. Another variant in UMOD (rs4293393) correlated with SCr in a combined analysis of 24,375 Icelandic and Dutch participants [25]. This variant was replicated for SCr in 194,286 Icelandic individuals (p = 2.48 × 10⁻³⁸), where additional UMOD variants (rs11864909; p = 4.05 × 10⁻²¹) and rs12917707 (p = 2.03 × 10⁻³⁶) also reached genome-wide significance [43]. More recently, rs12917707 and another UMOD variant (rs13329952, p = 9.47 × 10⁻⁴³) were found to be significantly associated with eGFRcrea. Variant rs12917707 showed association both in patients with diabetes (n = 118,365, p = 2.48 × 10⁻⁰⁸) and without (n = 11,522, p = 4.68 × 10⁻³⁶) [26].

In summary, as several different UMOD SNPs have been linked to both CKD and eGFR, UMOD may prove to be useful as a diagnostic and prognostic genomic biomarker of CKD. Indeed, levels of serum uromodulin have recently been shown to useful in disease detection as CKD patients were shown to have a significantly lower serum uromodulin concentration (p < 0.001) and measurement of serum UMOD was shown to significantly enhance performance of a CKD prediction model (p = 0.049) [46]. UMOD often exhibits one of the strongest associations with CKD [22, 26], suggesting that UMOD would likely be suitable as a diagnostic genomic biomarker of CKD.

Multiple variants (rs17319721 [22, 24, 25, 43, 44], rs2137154 [43], rs9992101 [43] and rs13146355 [43] in SHROOM3 have also been associated with eGFR in patients of European ancestry [22, 24, 25, 43, 44]. SHROOM3 encodes the shroom3 protein, which has been shown to regulate cell shape by coordinating cytoskeleton assembly [47] and whose overexpression (due to presence of variant rs17319721) is associated with increased allograft fibrosis in renal transplant patients [48]. The SHROOM3 rs17319721 variant was first identified in patients within four cohorts from Cohorts for Heart and Aging Research in Genomic Epidemiology (CHARGE) consortium [22]. Meta-analysis of 19,877 patients of European ancestry from CHARGE, to two additional population-based studies (n = 21,466) were used for discovery and replication, respectively [22]. After UMOD, the most significant genome-wide association with eGFR (estimated using SCr) was with SHROOM3, found in both discovery (p = 9.7 × 10⁻⁰⁸) and combined discovery and replication meta-analyses (p = 1.2 × 10⁻¹²) [22]. The same variant showed association with SCr in a study of 22,256 Icelandic patients (p  =  0.00057) [25], and association with eGFR was observed in a combined discovery and replication meta-analysis of 74,354 and 56,346 patients of European ancestry (p = 1.5 × 10⁻²¹) [24]. In the same study, association of SHROOM3 (rs17319721) and eGFR remained significant after stratifying patients by age, sex, diabetes and hypertension status, but a parallel investigation in patients of African ancestry yielded no association, highlighting the genetic differences between patients of different ethnic origins [24]. The rs17319721 variant in SHROOM3 was associated with eGFR in a study of 3028 Caucasian patients with type 2 diabetes (p = 3.18 × 10⁻⁰³), showing a larger impact on baseline eGFR when patients presented with albuminuria [44]. In a study of 81,656 Icelandic patients and their 112,630 relatives (correction made for relatedness), genome-wide significance was reached for each variant with SCr, in both discovery and replication analyses (rs17319721: p = 4.8 × 10⁻¹⁰, rs2137154: p = 4.7 × 10⁻¹³, rs9992101: p = 1.0 × 10⁻¹⁰ and rs13146355: p = 6.5 × 10⁻¹²) [43].

A number of other genes have shown specific association with eGFR, such as STC1 [23,24,25,26, 43], SLC22A2 [23, 24, 26, 31, 43] and WDR37 [23, 24, 26, 43, 49] are thought to be involved in creatinine secretion, rather renal function, therefore not being representative of CKD. This highlights the importance of functional analysis of significantly associated variants to determine their role in CKD. A second measure of eGFR, such as serum cystatin-C, may need to be incorporated into future studies to control for variants associated with creatinine secretion.

Diabetic kidney disease

The Engulfment and Cell Motility 1 gene (ELMO1) has been studied extensively in DKD (Table 3). Located on chromosome 7, ELMO1 codes for a protein of the same name involved in a pathway promoting phagocytosis of apoptotic cells [50]. A variant in intron 18 of chromosome 7p14 showed significant association with DKD (p = 5 × 10⁻⁵) in 560 Japanese patients and 360 T2DM controls [51]. As DKD often leads to ESRD [52], a follow-up study of ELMO1 gene variants in African American patients with T2DM-ESRD was carried out [53]. Combined analyses for a total of 1135 T2DM-ESRD patients was performed compared to a combination of healthy (n = 596), non-diabetic ESRD (n = 326) and T2DM controls (n = 328) in order to determine which SNPs were specifically associated with T2DM-ESRD and not the individual T2DM or ESRD traits [53]. A total of 13 variants in ELMO1 showed significant association with T2DM-ESRD, 11 of which were found in intron 13 [53]. These variants showed no association in patients with T2DM or ESRD singularly, indicating these SNPs are specific for DKD [53]. To extrapolate these findings in Japanese and African American patients, an additional study was carried out in patients of European ancestry, to determine if association of ELMO1 and DKD was independent of ethnic origin. In a study of advanced DKD, defined as the presence of persistent proteinuria or ESRD, 284 DKD patients with proteinuria and 536 patients with DKD-ESRD were analysed for association of 359,193 SNPs, 106 in ELMO1, versus 885 type 1 diabetes mellitus (T1DM) controls [54, 55]. Eleven nominally significant SNPs were identified within four novel genomic loci, including variants in FRMD3 (p = 5.0 × 10⁻⁷) and CARS (p = 3.1 × 10⁻⁶), which were also shown to contribute significantly to time needed to develop DKD (FRMD3, p = 0.02 and CARS, p = 0.01) [54].

Table 3 Genomic-wide association studies in diabetic kidney disease

While eight ELMO1 variants showed nominal association with DKD, and variant rs7785934 was found to be significantly associated with the ESRD group, three nominally associated variants (rs1558688, rs741301, and rs7799004) showed opposite genetic effect compared to the previous Japanese cohort study [54]. Ethnic differences may account for differences among studies, as a second, larger study in 1154 patients and 1988 controls of European ancestry from the GENIE consortium also failed to find genome-wide association between DKD and ELMO1, but found similar association for other 11 “top-hits” from a previous analysis in European patients [56]. Although performed in an animal model, diabetic mice with increased expression of ELMO1 show renal dysfunction similar to that seen in human DKD [57]. When the same mutation was induced in both diabetic and non-diabetic mice, ELMO1 expression was almost twofold higher in the diabetic mice [57]. This may explain why ELMO1 variants are significantly associated specifically with DKD, but further functional analysis in humans is required.

This study by the GENIE consortium also examined the erythropoietin gene (EPO) promoter, previously found associated (variant rs1617640; p = 2.76 × 10⁻¹¹) in a total of 1618 diabetic ESRD patients with proliferative diabetic retinopathy and 954 controls from three independent North American cohorts of European American ancestry [58]. The EPO variant rs1617640 was not replicated (although direction of effect was consistent) in the GENIE consortium patients, but the fixed-effects meta-analysis of the original and replication cohorts reached genome-wide significance (p = 2×10⁻⁹) [56].

Other traits and kidney diseases

Albuminuria, often measured using UACR, has been previously associated with increased risk of progression to ESRD [59]; therefore, genetic variants associated with increased UACR may help predict CKD progression. A meta-analysis of 63,153 individuals of European ancestry from the CKDGen consortium identified association between variant rs1801239 in the CUBN gene and both UACR (p = 1.1 × 10⁻¹¹) and microalbuminuria (p = 0.001), defined as UACR > 25 mg/g in women and > 17 mg/g in men [60]. Cubilin, the protein encoded by CUBN, when expressed in the kidney, complexes with membrane transporter megalin to reabsorb urinary albumin [61]. Cubilin has been suggested as a biomarker for renal cell carcinoma but may also hold promise as a useful addition to a biomarker panel for CKD [62]. As variants detected often differ among patients of different ethnicities, the rs1801239 variant in the CUBN gene was examined in 6981 African American individuals from the CARe consortium, 1159 of which had microalbuminuria. Significant association was again found with both UACR (p = 0.005) and the presence of microalbuminuria, indicating the influence on albuminuria is independent of ethnicity. However larger studies in a wider range of ethnically diverse populations may provide confirmation [60].

As well as showing significant association with eGFR, SHROOM3 (rs17319721) was also identified as being significantly associated with UACR in a combined meta-analysis of 31,580 and 27,746 Caucasian patients from independent cohorts within the CKDGen Consortium [60]. As this SHROOM3 variant has also been associated with larger impact on eGFR in patients with albuminuria [4] it may imply that shroom3 has a functional role in proteinuria.

GWAS have also been used to explore other kidney diseases, as shown in Table 4. Association of HLA-DQA1 and PLA2R1 has been shown in different studies of MGN patients. Both genes were first associated with MGN in three independent cohorts of Caucasian patients versus race-matched controls [63]. Replication of these findings also revealed that while PLA2R1 only showed association with MGN (p = 1.9 × 10⁻⁸), but HLA-DQA1 was significantly associated with other renal immune disorders in addition to MGN (p = 5.9 × 10⁻²⁷): LN (p = 2.8 × 10⁻⁶), T1DM with CKD (p = 6.9 × 10⁻⁵) and FSGS (p = 5.1 × 10⁻⁵) [64]. As IgAN is more common in Asian individuals [65], many studies have focused on studies of Asian cohorts, to determine if genetic variants give rise to higher disease incidence (Table 4). A combined analysis of 3144 patients and 2822 controls from one European and two Han Chinese cohorts identified five genomic loci significantly associated with IgAN, most of which are involved in innate immune responses [66]. These results were replicated in a meta-analysis of 12 cohorts comprising 5372 cases and 5383 controls of either European, East Asian or African American ethnicity [67]. Twelve SNPs within the five previously reported genomic loci (CFHR3/R1, HLA DQB1/DRB1, TAP2/PSMB9, DPA1/DPB2, HORMAD2) were found to reach genome-wide significance (p < 5.0 × 10⁻⁸) [67]. A study comprising 7658 cases and 12,954 controls of East Asian or European origin replicated all previously reported IgAN-associated variants and identified several new ones [68]. These include new variants within the previously identified HLA-DQ/DR locus (rs7763262, p = 1.8 × 10⁻³⁸) [66], and the newly identified rs11574637 variant in ITGAX-ITGAM (p = 2.8 × 10⁻¹). Variants within the HLA and ITGAX-ITGAM loci have been identified in patients with other autoimmune disorders, such as rheumatoid arthritis [69] and systemic sclerosis [70], respectively. A study of 8313 Han Chinese cases and 19,680 controls found significant association the ST6GAL1 (p = 7.27 × 10⁻¹⁰), ACCS (p = 3.93 × 10⁻⁹), ODF1-KLF10 (p = 1.41 × 10⁻⁹) genes, as well as three independent variants in the DEFA gene (rs2738058, p = 1.15 × 10⁻¹⁹; rs12716641, p = 9.53 × 10⁻⁹; rs9314614, p = 4.25 × 10⁻⁹) [71]. The association of a previously reported ITGAX-ITGAM variant was also validated (p = 2.26 × 10⁻¹⁹) [68, 71].

Table 4 Genomic-wide association studies in other kidney diseases

Conclusions

Although several genes with strong associations to renal diseases, eGFR and UACR, have been identified, the development of a universal diagnostic gene biomarker panel is however challenged by the variability and ethnic specificity of many of the variants identified. While GWAS may provide insights to genetic causes of disease, little information regarding the functional implications of the identified SNPs can be derived. Further analyses such as epigenomic and transcriptomic profiling, coupled with protein and metabolite analysis may give a more complete picture of CKD. A multi-omic approach may add functional and physiological evidence for the use of GWAS-identified variants as biomarkers or lead to the identification of biomarkers of a different nature that have not been detected through GWAS.

Epigenetic biomarkers

Epigenetics is defined as, “the study of changes in gene function that are mitotically and/or meiotically heritable and that do not entail a change in DNA sequence” [72]. Epigenetics acts as a bridge between genotype and phenotype, helping to explain why some genetic alterations may not necessarily result in an altered phenotype. It is therefore extremely important to consider epigenomic biomarkers alongside genomic and transcriptomic analyses to discover the role of different pathways in disease, as change in gene expression may not be linked solely to genetic aberrations and therefore will not be detected using purely genetic analysis.

Epigenetics can be subdivided by mechanism. Covalent modifications of DNA and histone proteins, such as DNA methylation, and the action of non-coding RNAs, such as microRNAs (miRNAs), all seek to interfere with gene expression without changing the genetic sequence of the gene in question, including any of its promotor or enhancer regions. Epigenetic regulation of gene expression results in altered levels of target gene mRNA available for translation, and thus affects protein synthesis. Epigenetic regulation is required to maintain the regulation of gene activity, transcription and contributes to the maintenance of genomic stability [73]. However, epigenetic dysregulation has previously been linked to diseases like Prader–Willi syndrome, an imprinting disorder, Fragile X syndrome, caused by loss of genomic stability and multiple different cancers [74]. Evidence has suggested that loss of epigenetic control may be associated with CKD, with differential expression of microRNAs [75] and differential DNA methylation detected in kidney fibrogenesis, a hallmark of CKD [76].

DNA methylation

DNA methylation is an example of a covalent modification to the DNA sequence that has the ability to alter gene transcription, especially if found near the promoter or enhancer regions of genes [77, 78]. DNA methylation is the transfer of a methyl group (-CH₃) from S-adenosylmethionine to the fifth carbon of a cytosine nucleotide in the DNA sequence [79], catalysed by DNA methyltransferases (DNMTs) [80, 81]. It is the presence of these cytosine-bound methyl groups that can either prevent the binding of transcription factors in the promoter region of genes, or attract and bind repressor proteins, both of which lead to a reduction in gene transcription. The majority of methylated cytosines are followed by a guanine nucleotide and these sites have been denoted as CpG sites [82]. Most of CpG sites are methylated, except in regions of high CpG density, known as CpG islands. CpG islands exhibit much lower levels of DNA methylation than expected, with methylation found to be significantly lower in these regions when compared to other genomic CpG sites [83]. CpG islands are often found at gene promoters, and their methylation regulates gene expression [84]. DNA methylation has been shown to be involved in maintaining X-inactivation [85], the silencing of imprinted genes [86], and plays a role in embryonic development; knockout of DNMT enzymes in mice results in embryonic lethality [87, 88]. The role of DNA methylation in diseases like cancer is clear, with colorectal, breast and renal cancers all showing high levels of epigenetic dysregulation [89].

Role of DNA methylation in CKD

Multiple approaches have been taken to further understand the role of epigenetic aberrations in the pathogenesis of CKD (Table 5). Global methylation profiling [90,91,92,93], epigenome-wide association studies [94,95,96,97,98,99,100,101] and candidate gene studies [91, 102, 103] have been carried out in patients with CKD, DKD and ESRD. Detection of differential DNA methylation in these studies may lead to new diagnostic and prognostic epigenetic biomarker panels being developed.

Table 5 Methylation studies in chronic kidney disease

Global methylation and CKD

The influence of total genomic methylation levels on CKD has been analysed by global methylation analysis. Global methylation can be assessed by different methodologies; enzymatic digestion with methylated and unmethylated cytosine-sensitive enzymes generates a methylation ratio using the luminometric methylation assay (LUMA), percentage methylation can be established using a combination of liquid chromatography and mass spectrometry (LC-MS) or surrogate markers of whole genome methylation may be used [104]. An example of the latter is pyrosequencing the long-interspersed nuclear element 1 (LINE-1), which is considered a reliable marker of global methylation [104, 105].

In the first study of global methylation in CKD, LUMA was used to evaluate 155 patients with CKD (stage 3–5) and 36 healthy controls [90]. A differential methylation ratio was found for inflammation in late-stage CKD patients CKD (p = 0.0001) [90]. While this study showed no significant difference for CKD itself, other studies attempted to demonstrate differential global methylation in CKD, with mixed success. Pyrosequencing of the LINE-1 element in 22 ESRD patients and 26 healthy controls showed that this retrotransposon was significantly hypermethylated (p < 0.01) in ESRD patients [91]. Although all ESRD patients also had hypertension, this is not known to be associated with global hypermethylation, being either associated with global loss of methylation [106] or with no difference in methylation at the LINE-1 transposable element [107]. While this study found significant hypermethylation associated with ESRD, another study focusing on the association between DNA methylation and renal function in 93 CKD patients (stage 2–4) found no significant association of DNA methylation with renal function [93]. Indeed, a more-recent study found significant DNA hypomethylation (p = 0.0001) in 30 Stage 3–4 CKD patients and demonstrated the ability of cholesterol lowering agents to significantly increase DNA methylation [92].

No clear association of CKD with global changes in DNA methylation has been established from these studies. Published studies of global methylation have reported on patient groups with small numbers (n < 100) and such studies lack power to detect significant differences in global DNA methylation.

Site-specific differential methylation in chronic kidney disease

In the last 10 years, research has been undertaken to identify individual genes or genomic regions whose differential methylation is associated with CKD; several groups have utilised large, array-based methods to analyse DNA methylation at an epigenome-wide level. The Illumina Infinium HumanMethylation 450 K BeadChip (450 K Array) is used to assess methylation status of 485,577 genomic CpG sites [108], while next-generation methylation sequencing (Methyl-Seq) assesses the methylation status of 813,267 CpG sites [109]. These large, array-based methods allow much higher sample throughput, contrasting with traditional techniques for assessing methylation like pyrosequencing, where only a fraction of the CpG sites can be assessed in the same time-frame.

Using the 450 K array, 23 unique genes were identified in a sample of 255 CKD patients that contained more than one significantly differentially methylated CpG site (p < 10⁻⁸) [97]. Of these genes, six were selected to undergo RNA analysis. While some genes, such as ELMO1, showed altered levels of expression in CKD patients, these differences did not reach significance. So while this may suggest that DNA methylation plays a role in regulating their expression, further analysis is required in a larger number of patients, as only two Stage 5 CKD patients and two controls were used in expression studies [97]. Further study in this group of patients was focused on the differential methylation of the UMOD and related genes UMODL1 and UMODL1-AS1 genes [100], previously associated with CKD in GWAS [23, 25]. Indeed, 11 CpGs across the three genes were found to be significantly differentially methylated in patients with CKD (p < 10⁻⁸) [100].

The relationship between DNA methylation and CKD progression has also been assessed in Stage 2–4 CKD patients. Twenty patients with “rapid disease progression” and 20 patients with “stable kidney function,” determined using a mixed effects linear regression model using two measures of eGFR, had their methylation status assessed using the 450 K Array [98]. Although no significant association with CKD progression could be found after correction for multiple testing, several candidate genes (including UMODL1) that had previously been related with CKD in the literature [96, 98, 100] showed a trend towards association with rapid CKD progression.

Differential methylation of single genes has also been studied, particularly when they are candidate to have a functional role in the development of CKD. In a study of 96 ESRD patients/96 controls, the specific methylation of the methylenetetrahydrofolate reductase (MTHFR) gene promoter was found to be significantly associated with ESRD (p = 0.003) and significant correlation was observed between a lower eGFR and MTHFR promoter methylation (p = 0.026) [102]. The hypermethylation of the MTHFR promoter is thought to lead to the transcriptional silencing of MTHFR, preventing remethylation of S-adenosylhomocysteine to S-adenosylmethionine, which is required to donate methyl groups necessary in DNA methylation [110]. Loss of MTHFR activity is characterised by a build-up of S-adenosylhomocysteine, resulting in hyperhomocysteinaemia in patients—which is strongly associated with CKD [111]. However, MTHFR was also highlighted in two other analyses, with Sapienza et al. and Ko et al. both finding significant hypomethylation of the MTHFR gene in patients with ESRD and DKD and solely DKD, respectively [95, 96]. This contrasts with the hypermethylation of the gene promoter found in the original study. Therefore, as the evidence of MTHFR as a biomarker for CKD remains conflicting, its usefulness in any panel of non-invasive epigenetic biomarkers remains controversial [95, 96, 102].

Significant hypomethylation at the promoter of SHC1, formerly known as P66SHC (Gene ID: 6464), was found in 22 patients with ESRD (p < 0.01) [91]. SCH1 is involved in the regulation of reactive oxygen species [112] and may be linked to CKD due to oxidative stress has been shown increased in patients with CKD [113].

Klotho (KL) is another gene that has been assessed on an individual basis in CKD patients. KL is highly expressed in kidney tissues, and causes a syndrome similar to CKD when it is knocked out in mice [114]. Both KL mRNA and protein production have been found to be reduced significantly in patients suffering from ESRD [115]. Using pyrosequencing, six CpG sites across the KL gene in peripheral blood DNA and DNA extracted from renal tissue of 47 patients were analysed [103]. Forty-seven renal cell carcinoma biopsies and peripheral blood DNA of 48 healthy patients acted as non-CKD control samples. KL was significantly hypermethylated in approximately 70% of all renal and blood samples taken from patients (p < 0.001), and KL promoter hypermethylation also found to be inversely correlated with eGFR [103]. For this reason, KL has been further investigated in animal and in vitro models to determine if KL promoter hypermethylation is involved in the development of CKD and altered renal fibrosis was noted in cells with KL hypermethylation [116, 117].

Association with diabetic kidney disease

To date, several studies have investigated the role of methylation in DKD and diabetic ESRD, as shown in Table 6 [94,95,96, 101].

Table 6 Methylation studies in diabetic kidney disease

Two different studies have used the Illumina Infinium HumanMethylation27 BeadChip (27 K Array) to assess the methylation status of 27,578 CpG sites in patients with DKD and diabetic ESRD [94, 95]. Bell et al. examined differences in DNA methylation in genomic DNA extracted from whole blood from 96 DKD patients and 96 controls [94], while Sapienza et al. focussed on differential methylation in DNA extracted from saliva of 23 diabetic ESRD patients and 23 controls [95]. Bell et al. found over 400 differentially methylated CpGs originally, which were reduced to 19 when corrected for multiple testing. Sapienza et al. found 187 genes containing more than one significantly differentially methylated CpG, but several of these genes were eliminated as false-positives after correction. There were no common genes identified between these two groups. Although they used the same methylation analysis, this is perhaps not surprising, since Sapienza studied patients with diabetic ESRD, whereas Bell only analysed patients with DKD, and both studies were very small. Ethnic differences in cohorts may also have played a role; Bell et al. analysed samples from Irish patients [94], where Sapienza examined African American and Hispanic American patients [95]. Indeed, when Hispanic patients were removed from the analysis, the number of genes detected as being differentially methylated was reduced, showing that ethnicity may influence on the genes detected [95]. While these studies have not provided clear differentially methylated candidate genes, they have highlighted the need for studies with greater statistical power and that the future of epigenome-wide studies may lie in the analysis of specific phenotypes within well-defined patient populations.

By far the most extensive list of differentially methylated genes in DKD has been produced by Ko et al. where 14 patients were compared to 14 controls in a discovery analysis, which was followed up by a replication study consisting of 21 patients and 66 controls [96]. There were marked differences in phenotypes between the discovery and replication patient groups; the discovery cohort had both hypertension and DKD patients, and controls were healthy subjects, whereas the replication cohort consisted of a DKD patient group and non-DKD control group, including patients with hypertension and diabetes. They identified 1061 unique differentially methylated genes in the replication cohort, 98% of which had also been found in the discovery analysis. Of the 2% of genes only found in the discovery cohort, CUX1, PTPRN2 and STK24 had previously been associated with CKD [97] and the association of SALL1 and WHSC1 were later confirmed in cells extracted from non-diabetic ESRD patients [99]. These genes were not found to be associated in the replication analysis, suggesting that even subtle phenotypic differences may give rise to differences in DNA methylation. Further analysis would be required to determine the contribution of specific phenotypes to the significance associated with CpGs in these genes. The larger number of significant differentially methylated CpGs detected in this study may also be due to the use of renal tubule cells as opposed to genomic DNA from whole blood extracted in all previous epigenome-wide analyses.

RUNX3 and PHB have been described in multiple studies analysing differential methylation in DKD. RUNX3 was found to be significantly differentially methylated by Ko et al. and Bell et al. and in fact was the only unique gene that remained significant after the application of a more stringent false discovery rate threshold [94, 96]. RUNX3 is a transcription factor that works in balance with the STAT5 transcription factor in the renal fibrosis pathway [118], suggesting that altered methylation and thus expression of RUNX3 may lead to aberrant renal fibrosis, a hallmark of CKD [119]. PHB, a mitochondrial gene, was identified by both Ko et al. and Swan et al. Swan et al. found PHB significantly differentially methylated in 150 DKD patients, 50 of whom were diabetic ESRD patients, and significance was reached for PHB in both groups when compared to diabetic controls [101]. The sub-analysis in mitochondrial genes showed 51 genes differentially methylated in DKD versus the diabetic control group (p < 10⁻⁸), and 374 genes were identified between the ESRD and diabetic control group [101]. Forty-three of these genes overlapped with the 51 genes found associated with DKD, but PHB was the only gene overlapping with other studies [96, 101]. PHB has been involved in regulating cellular processes, such as transcription and apoptosis [120] and its dysregulation increases renal fibrosis and oxidative stress [121], which have previously been associated with CKD [113, 119].

In a non-population based study, further analysis of the epigenetic changes involved in ESRD was investigated. The genome-wide methylation of monocytes treated with either serum from patients with ESRD or healthy controls was assessed [99]. Although only performed in 5 patients and 5 healthy age- and sex-matched controls, differentially methylated CpGs were detected between monocytes that had differentiated in the presence of ESRD patient or control serum. The top 50 genes were listed and five of these (GPR39, HDAC1, PRKCE, SLC43A2 and ST3GAL5) overlapped with genes identified as significant in the replication analysis by Ko et al. [96, 99]. While this study seeks to replicate the changes in DNA methylation seen in uraemic ESRD patients rather than measuring DNA methylation in patients directly, overlapping results in DKD patients may prompt further investigation in this area in a large patient cohort to confirm results at population level.

Additional work has been carried out by several groups to determine the functional implications of the significant differences in methylation in CKD patients [96, 97]. Ko et al. performed RNA and Gene Ontology analysis on the replication cohort and found that over 40% of the significantly methylated genes showed significant changes in gene expression. Many of these genes, such as TGFBR3 and SMAD6, belong to the TGF-β pathway, whose involvement in CKD development is well-established [96]. Smyth et al. carried out similar analysis on two patients and controls and although no significant changes in gene expression were found, pathway analysis linked several genes to the mucin type O-glycan biosynthesis pathway, dysregulation of which has been shown to result in kidney abnormalities [97, 122]. However, despite this additional work, no genes or CpG sites have been established as a clear candidates to be taken forward into biomarker development [96, 97].

Conclusions

Altogether, these findings suggest a role for DNA methylation analysis in the diagnosis and prognosis of CKD and related diseases, but also highlight the difficulties associated with establishment of definitive epigenomic biomarkers. Among all the methylation markers described in the literature, only a few genes appear more than once. MTHFR gene is the only epigenetic biomarker proposed in more than two studies, but with inconsistent results regarding the mechanism underlying its influence [95, 96, 102]. Therefore, the role of MTHFR as a potential biomarker in CKD is yet to be elucidated. Further investigations using a larger number of CpG sites, phenotypes and ethnicities in larger cohorts is required to allow accurate, phenotype-specific epigenetic biomarkers to be established. Deeper understanding of the role DNA methylation plays in disease development may lead to the discovery of potential points of therapeutic intervention, as well as the discovery of diagnostic and prognostic biomarkers. To better understand this, the identification of specific methylation regions in CKD is essential, to serve both as potential candidates for epigenomic biomarkers, but to also to give an indication of the genes affected in different functional pathways, giving rise to the elucidation of mechanisms involved in the development of CKD.

microRNA

MiRNAs are 21–25 nucleotides length small non-coding RNA molecules, partially complementary to one or more messenger RNA (mRNA), which downregulate gene expression by translational repression, mRNA cleavage, and deadenylation. Targeted mRNAs include those involved in biological processes such as inflammatory response, cell–cell interaction, apoptosis and intra-cellular signalling [123]. MiRNAs have been widely investigated as diagnostic and prognostic biomarkers in CKD, with a particular interest in their potential use as non-invasive markers.

Chronic kidney disease

Prediction of CKD

Several studies have explored the use panels of differentially expressed miRNAs as potential biomarkers for CKD although these reported on relatively small numbers of patients (Table 7).

Table 7 miRNA studies in chronic kidney disease

Recently, a panel composed of 16 miRNAs upregulated (let-7c-5p, miR-222–3p, miR-27a-3p, miR-27b-3p, miR-296-5p, miR-31-5p, miR-3687, miR-6769b-5p, and miR-877-3p) or downregulated (miR-133a, miR-133b, miR-15a-5p, miR-181a-5p, miR-34a-5p, miR-181c-5p, and miR1-2) has been identified in urine exosomes of 15 CKD patients compared to 10 healthy controls [124]. A similar 14 miRNAs panel (miR-29c-5p, miR-345-5p, miR-142-3p, miR-339-3p were upregulated and miR-17-5p, miR-130a-3p, miR-15b-5p, miR-106b-3p, miR-106a-5p, miR-16-5p, miR-181a-5p, miR-1285-3p, miR-15a-5p, miR-210-3p were downregulated) has been described in 15 African American CKD patients with treated hypertension compared to 15 controls (non-CKD patients with treated hypertension) [125]. In contrast with these results, downregulation of members of miR-29 and miR-200 was found in urine exosomes from 32 CKD patients when compared to 7 healthy controls [126].

Circulating miR-122 has been identified as a biomarker for ESRD, being substantially downregulated in 17 pre-haemodialysis patients compared to 22 healthy controls (19-fold lower), 30 patients with CKD (21.7-fold lower) and 15 transplanted patients [127]. Another biomarker differentially expressed in dialysis patients is miRNA-155, whose serum levels were found increased in 82 dialysis patients compared with 16 healthy subjects (P < 0.05) [128]. Interestingly, miR-155, along with other miRNAs involved in CVD, such as miR-21, miR-26b, or miR-146b, had been found differentially expressed in 10 clinically stable haemodialysis patients compared to 10 healthy controls [129]. In a larger study, circulating levels of miR-125b, miR-145 and miR-155 were found significantly decreased in 90 CKD and 10 haemodialysis patients compared to those in 8 healthy volunteers [130]. Decreased circulating levels of the three miRNA were also associated with progressive loss of eGFR [130].

The miRNA transcriptome has also been studied in platelets from five chronic haemodialysis patients and five stage 4 CKD uremic patients and compared with five age- and sex-matched healthy subjects with normal renal function [131]. Upregulation of miR-19b, involved in platelet reactivity, was found in uraemic CKD patients.

eGFR and CKD stratification

Total circulating small RNA level has been found decreased in 53 patients with impaired kidney function (CKD stages 3–5) compared to 22 controls (p < 0.0001, r = 0.553; Spearman rho) [132]. Furthermore, total small RNA concentration was threefold higher in plasma of normal/stage 3 patients compared to stage 4/ESRD (p < 0.0001). Specifically, the expressions of circulating miR-16, miR-21, miR-155, miR-210 and miR-638 were inversely correlated with eGFR [132]. Urinary miRNA level and kidney function was only associated for miR-638, which showed a significant increase in patients with stage 4 CKD compared to normal and stage 3 CKD patients (p < 0.006) [132]. In agreement with these results, decreased circulating levels of miR-125b, miR-145 and miR-155 were associated with progressive loss of eGFR by multivariate analyses in as study comprising 100 CKD and 10 haemodialysis patients of Caucasian and African American origin [130]. eGFR also inversely correlated with urinary level of miR-155 in 60 nephrolithiasis patients compared with 50 controls [133].

Another miRNA example associated with eGFR is miR-29c, which correlated with both eGFR (r = 0.362; p < 0.05) and degree of tubulointerstitial fibrosis (r = − 0.359; p < 0.05) in a group of 32 CKD patients [126].

More recently, 384 urinary and 266 circulatory miRNAs were found differentially expressed in 28 CKD patients when grouped by eGFR ≥ 30 vs. < 30 ml/min/1.73 m² [134]. Among them, TGF-β signaling-related mRNA targets suggest that specific urinary and plasma miRNA profiles may act as diagnostic and prognostic biomarkers in CKD [134]. Urine downregulation of Let-7a and upregulation of miR-130 may identify eGFR < 30 ml/min/1.73 m² patients, whereas upregulation of miR-1825 and miR-1281 in both urine and plasma, and plasma miR-423 downregulation would serve as an indicator of decreased eGFR [134].

Diabetic kidney disease

Prediction of DKD

Table 8 shows miRNA studies focused on DKD. The development of microalbuminuria was associated with a urinary miRNA signature composed of miRNAs known to be involved in the pathogenesis and progression of diabetic renal disease (miR-105-3p, miR-1972, miR-28-3p, miR-30b-3p, miR-363-3p, miR-424-5p, miR-486-5p, miR-495, miR-548o-3p and for women miR-192-5p, miR-720) in 17 T1DM patients exhibiting microalbuminuria compared to 10 T1DM without renal disease [135]. The same group had described a decrease in urine expression of miRNA-221-3p in a former study of T1DM patients with [10] or without DKD [10, 136]. In these patients, a panel of 10 miRNAs (underexpressed: miR-221-3p; overexpressed: miR-619, miR-486-3p, miR-335-5p, miR-552, miR-1912, miR-1224-3p, miR-424-5p, miR-141-3p, miR-29b-1-5p) was identified for patients with overt DKD [136]. Furthermore, downregulation of miR-323b-5p and upregulation of miR-122-5p was associated with persistent microalbuminuria. Appearance of microalbuminuria was associated with decreased miR-323b-5p and increased miR-429 urine levels [136]. Consistent with these findings, downregulation of miR-155 and miR-424, and upregulation of miR-130a and miR-145 was found in urinary exosomes from 12 microalbuminuric T1DM compared with 12 normoalbuminuric patients [137].

Table 8 miRNA studies in diabetic kidney disease

eGFR

In the largest miRNA study to date in DKD patients, downregulation of miR-2861, miR-1915-3p, and miR-4532 was associated with eGFR (p < 0.01) and interstitial fibrosis/tubular atrophy (p < 0.05) [138]. In a separate study, miR-192 downregulation was correlated with tubulointerstitial fibrosis and low eGFR in 22 DKD patients especially in those who had stage 5 CKD or required renal replacement therapy within 6 months of their renal biopsy [139].

The rate of eGFR decline positively correlated with the urinary miR-21 (r = 0.301; p = 0.026) and miR-216a (r = 0.515; p < 0.0001) in a combined analysis of patients with diverse CKD aetiologies including IgAN, diabetic glomerulosclerosis and hypertensive nephrosclerosis [140]. However, the separate analysis by disease only found urinary miR-216a levels correlated with the rate of eGFR decline in patients with hypertensive nephrosclerosis (r = 0.588; p = 0.005) and diabetic glomerulosclerosis (r = 0.605; p = 0.010) [140]. Other urinary miRNA biomarkers correlated with eGFR in IgAN patients are miR-200b (r = 0.512; p < 0.001) and miR-429 (r = 0.425; p = 0.005) [141]. In systemic lupus erythematous (SLE) patients, eGFR was correlated with urinary miR-146a in two studies from the same group (r = 0.242; p = 0.008) [142, 143]. Serum miR-146a also inversely correlated with proteinuria (r = − 0.341; p = 0.031) and the SLE Disease Activity Index (r = − 0.465; p = 0.003) in 40 SLE patients [142].

Primary glomerulonephritis

Although many studies have tried to find a specific signature of miRNAs in primary glomerulonephritis (Table 9), most of them are small sized cross-sectional designs which lack the statistical power and appropriate treatment of important issues, as correction for multiple testing, complicating the generalisation of their conclusions and limiting the extrapolation of their results.

Table 9 miRNA studies in primary glomerulonephritis

Overexpression of miR-25-3p, miR-144-3p or miR-486-5p in urinary erythrocytes was recently proposed as non-invasive diagnostic biomarkers for IgAN nephropathy in a study of 93 IgAN patients compared to 82 normal subjects and 40 disease controls [144].

Downregulation of miR-30d, miR-140-3p, miR-532-3p, miR-194, miR-190, miR-204 and miR-206 was associated with progression to ESRD or doubling of SCr in 43 patients with various glomerular diseases and decreased expression of miR-206 and miR-532-3p was confirmed in a validation cohort of 29 patients [123]. When the miRNA expression in the urinary sediment of 56 patients who had undergone kidney biopsy (17 diabetic glomerulosclerosis, 17 IgAN and 22 with hypertensive nephrosclerosis) was quantified, decreased miR-15 was associated with diabetic glomerulosclerosis, whereas increased miR-17 levels were shown in IgAN [140].

Other miRNA biomarkers proposed as indicators of IgAN are urinary and intra-renal overexpression of miR-146a and miR-155, described in 43 patients with IgAN compared to 13 healthy volunteers [145], urinary downregulation of miR-200a, miR-200b and miR-429 of patients with IgAN [141] and upregulation of miR-148b in PBMCs of patients with IgAN [146].

In 16 patients with FSGS, urinary downregulation of miR-1915 and miR-663 and miR-155 upregulation was found in patients when compared to five individuals with minimal change disease and five healthy controls (p < 0.001 and p < 0.005, respectively) [147]. Urinary levels of miR-663 were inversely correlated with eGFR (r = − 0.50; p < 0.04). Proteinuria correlated with higher plasma levels of miR-342 (r = 0.73; p = 0.02) and higher urine levels of miR-155 (r = 0.72; p = 0.03) and miR-663 (r = 0.45; p = 0.04), and with lower urinary levels of miR-1915 (r = 0.95, P = 0.003). These patients also showed upregulation of miR-30b, miR-30c, miR-34b, miR-34c and miR-342 in plasma and mir-1225-5p in urine [147]. In another study, an initial screening stage including plasma from nine FSGS patients and nine healthy controls identified a panel of three upregulated miRNAs (miR-125b, miR-186, and miR-193a-3p); the panel was subsequently confirmed in 32 FSGS patients and 30 healthy controls [148]. Increased plasma levels of miR-125b and miR-186 were also associated with nephrotic proteinuria [148]. Other proposed biomarkers for FSGS and kidney injury are miR-10a and miR-30d, which were found increased about 13- and 10-fold, respectively in the urine from 10 FSGS patients compared to 16 healthy controls [149].

Other kidney diseases

Table 10 shows miRNA studies developed in patients with other kidney diseases. Lupus disease activity has been associated with other biomarkers, such as miR-221 and miR-222, in the urinary sediment of 12 patients with LN [150]. These biomarkers have also been associated with macroalbuminuria in DKD patients [136] and CKD [124, 125]. Serum downregulation and urine upregulation of miR-146a and miR-155 was described in SLE patients compared to controls in two studies from the same group [142, 143]. Also in LN, downregulation of miR-3201 and miR-1273e (> threefold; p < 0.0001) and association with endocapillary glomerular inflammation was found in 89 LN patients (p < 0.01) [138].

Table 10 miRNA studies in other kidney diseases

Urinary overexpression of miR-142-3p, associated with CKD [125], has also been found in 41 kidney transplant recipients (23 patients with acute rejection and 18 with acute tubular necrosis when compared to eight stable patients (p < 0.001 and p < 0.005, respectively) [151]. Peripheral blood analysis in these patients also showed overexpression of miR-142-3p as a marker of acute tubular necrosis (p < 0.05 for both acute tubular necrosis/stable and acute tubular necrosis/acute rejection comparisons) [151].

Serum and urinary levels of miR-155 were found significantly higher in 60 nephrolithiasis patients compared with 50 controls [133]. miR-155 has also been associated with eGFR in CKD [130, 132, 133], proteinuria in FSGS [147], microalbuminuria in DKD [137], and proposed as diagnostic biomarker of CKD [129, 130, 148], FSGS [147] and IgAN [145].

Conclusions

In the last decade, there have been major research efforts to find miRNA biomarkers capable of identifying and stratifying kidney diseases and associated phenotypic features, such as eGFR and proteinuria. Arguably the most promising miRNA biomarker associated with kidney diseases is miR-155, which is inversely correlated with eGFR, and found differentially expressed in CKD, IgAN, FSGS and nephrolithiasis. MiR-29 has also been found in several studies associated with CKD, overt albuminuria in DKD and nephrotic syndrome. However, the findings regarding whether its expression is up- or downregulated in CKD are conflicting, although these differences may be due to the sample source used, blood or urine exosomes. Other miRNA biomarkers proposed for kidney disease (miR-16, miR-17, miR-21, miR-30b, miR-122, miR-130, miR-142-3p, miR-146, miR-200, miR221, miR222, miR-486, miR-1915) have not been confirmed in more than two independent studies, indicating that further investigation needs to be done to clarify their role.

Transcriptomics

Transcriptomics is the study of the complete set of RNA transcripts produced by the genome, in a specific cell or under certain circumstances [152]. The sum of RNA transcripts including mRNAs, ribosomal and transfer RNA, and regulatory noncoding RNAs comprise the transcriptome [153]. Throughout the past two decades, transcriptomics studies on renal disease have been carried out using different approaches, either studying the entire transcriptome or focusing on individual biomarkers. Since 1994, over 80 papers have tried to identify key regulators of different pathways which influence renal disease, including CKD, ESRD, DKD, Primary Glomerulonephritis, SLE and various other renal conditions. Most of the studies were performed on small numbers, with only three studies analysing results from over 100 patients (Table 11). A recent integrative bioinformatics analysis of 250 gene expression datasets of healthy renal tissues and those with various types of established CKD, including DKD, hypertensive nephropathy, and glomerulonephritis, retrieved from the Gene Expression Omnibus repository (https://www.ncbi.nlm.nih.gov/geo/) has identified nine genes significantly differentially expressed both in diseased glomeruli and tubules (IFI16, COL3A1, ZFP36, NR4A3, DUSP1, FOSB, HBB, FN1, PTPRC) [154]. Gene Ontology, pathway enrichment analysis and protein interaction network showed dysregulation of several metabolic, immune response, signalling pathways, platelet dysfunction, and extracellular matrix (ECM) organization in the glomeruli and signalling pathways in the tubules, mainly associated with apoptosis, PI3K-Akt, MAPK, and TGF-β [154].

Table 11 Transcriptomics studies in kidney diseases

Chronic kidney disease

The association of mRNA expression profiles with the risk of CKD has been investigated in different such as human renal biopsies, urine or peripheral blood samples.

Urine samples were used to investigate the correlation of the mRNA expression of a panel of target genes with renal function in 29 CKD patients (12 IgAN and 17 glomerulosclerosis) and 10 healthy controls using qRT-PCR [155]. Overexpression of TGF-β and MCP-1 was associated with glomerulosclerosis, and overexpression of collagen IV with both types of CKD. TGF-β1 plays an important role in the pathogenesis of CKD as it promotes inflammatory cell infiltration, tubular cell atrophy, mesangial cell hypertrophy and podocyte apoptosis [156], key features of CKD [157]. Urinary expression of connective tissue growth factor (CTGF) (r = − 0.471, p = 0.010) and collagen I (r = − 0.399, p = 0.032) were inversely correlated with the rate of eGFR decline after 12 months of follow-up, while TGF-β and MCP-1 did not have an influence on eGFR. CKD is characterized by the accumulation of ECM components in the glomeruli (glomerular fibrosis, glomerulosclerosis) and the tubular interstitium (tubulointerstitial fibrosis) [156]. TGF-β, IL-6 and MCP-1 expression was correlated to TLR4, which was significantly upregulated in human kidney biopsies from 70 CKD patients showing severe proteinuria and chronic ischaemic renal damage [158]. Both urinary and tissue MCP-1 also showed an increase in patients with acute renal inflammation in 174 patients with a variety of kidney diseases [159].

IL-6 expression was higher in 75 stage 2–5 CKD patients compared with 33 normal subjects, although it was not dependent on stage [160]. These CKD patients also showed upregulation of TNF-α, which was inversely correlated with eGFR [160].

The mRNA levels of COX6C, COX7C, ATP5ME, and UQCRH in peripheral blood mononuclear cells (PBMC) were all significantly higher in CKD stage 4–5 patients compared to nine CKD stage 2–3 patients and eight healthy controls [161].

Intrarenal mRNA expression of epidermal growth factor (EGF) in the tubulointerstitial compartment of kidney biopsies was identified as a potential predictive biomarker of eGFR at the time of the biopsy in 55 CKD patients [162].

Leukocyte angiotensin II AT1 receptor mRNA expression was inversely correlated with renal function in 20 CKD patients compared to 10 healthy subjects (r² = 0.15, P < 0.03) [163].

Increased levels of versican expression positively correlated to histologic damage scores and to renal function in proteinuric kidney disease and impaired renal function in 74 renal biopsies from patients with various proteinuric kidney diseases. Versican isoforms V0 and V1 significantly correlated with SCr levels [164].

In 64 human kidney biopsies from CKD patients, mRNA expression of peroxisome proliferator-activated receptor gamma (PPARγ) correlated inversely with renal function [165]. In 35 CKD patients, urinary mRNA of CTGF and nephroblastoma overexpressed (NOV) gene were overexpressed compared to 12 healthy controls [166].

End-stage renal disease

In 2006, Szeto analysed the role of urinary mRNA expression of 11 target genes as non-invasive markers on risk stratification of CKD patients in 131 patients with CKD followed until a primary endpoint of doubling SCr concentration or ESRD [155]. Urinary mRNA expression of hepatocyte growth factor was an independent predictor of the primary endpoint after adjustment for clinical and histological factors such as eGFR, gender and age and tubulointerstitial fibrosis [155].

Overexpression of IFN-γ, perforin, and granzyme B in CD4⁺CD28^null cells was associated with CKD in a study of 25 stage 4–5 CKD patients and 8 healthy subjects [167]. Higher expression of S100A12 in uraemic leukocytes has also been proposed as biomarker for ESRD in 40 stage 4–5 CKD patients compared to 20 healthy individuals (78.5 ± 70.5 vs. 23.7 ± 19.2 ng/ml; p = 0.0035) [168].

Diabetic kidney disease

Diabetes, as one of the leading causes of ESRD [169], has also been the focus of transcriptomic biomarkers studies.

The role of two isoforms of protein kinase C (PKC)-alpha and beta, was investigated in renal biopsies using reverse transcription-PCR in 20 patients with T2DM. PKC-alpha gene expression was increased in diabetic patients with CKD [170].

Peripheral blood samples were analysed for expression of 35 gene transcripts. The results showed that serum MCP-1, FGF-2, VEGF, and EGF were all elevated at every stage of DKD. However, serum IL2RA was the only mediator to show a linear increase with the disease severity consistent with decreasing eGFR. The peripheral blood samples showed elevated levels of ICAM1, TNF-α, TGF-β, IL-8, IL17RA, IFNγ, and MYD88 at all stages of disease in this sample of 20 patients with clinical or biopsy confirmed DKD age, race and gender matched with healthy volunteers [169].

Other renal diseases

Other CKD aetiologies, such as LN, have been investigated as potential beneficiaries of RNA expression biomarkers. CTGF mRNA expression was significantly correlated with TGF-β1, but inversely correlated with baseline eGFR, being higher in CKD stage 3–5 patients compared to stage 1–2 CKD in 39 patients with LN [171]. CTGF expression was positively associated with TGF-β1 [171].

Higher TGF-β1 expression was suggested to be associated with IgAN progression, it was significantly associated with eGFR at the time of biopsy in 51 patients [172]. Upregulation of TNFRSF13B and TNFRSF17 in B-lymphocytes appeared to be a trend in IgAN patients however statistical significance was not reached [173].

TNFAIP3 expression has been found downregulated in CD4+ T cells from SLE patients compared to normal controls [174].

Conclusion

Many studies have attempted to find transcriptomic biomarkers associated with a higher risk of CKD or biomarkers that predict changes in eGFR or proteinuria. However, most of the studies are limited in sample size and power and the biomarkers are not replicated in other studies. Expression of some cytokines as TGF-α and TGF-β1 and the matricellular protein CTGF have been proposed in different studies as potential biomarkers of renal function in CKD, but the evidence is not overwhelming.

Concluding remarks

The proliferation of studies on omics-related biomarkers in the past decade reflects the need for novel, valid, non-invasive tools capable of identifying persons at risk of CKD and helping to target the management of renal disease. Identification of undiagnosed disease, stratification of patients and monitoring of kidney disease are fields that would benefit of such validated biomarkers.

Several genes, including UMOD, SHROOM3 and ELMO1 have been strongly associated with renal diseases, and some of their traits, such as eGFR and creatinine. The role of epigenetic and transcriptomic biomarkers in CKD and related diseases is still unclear. Multiple putative biomarkers have been identified but these are mainly derived from small, single centre studies. Confirmation of the utility of such biomarkers is needed in separate populations and in larger cohorts.

It is very unlikely that a single biomarker can be identified that can improve CKD risk prediction beyond current clinically used tests such as serum creatinine or urinary albumin excretion. Future progress in this area will come from combining multiple biomarkers into classifiers, including genomic, epigenomic, proteomic and/or metabolomic profiles, to allow more precision in the diagnosis of CKD and in assessing the prognosis for renal disease.

Screening for CKD by eGFR and/or albuminuria in high-risk populations, e.g. DM or hypertension has been suggested to be cost-effective, whereas screening in the general population is only cost-effective in certain situations, for instance older patients or those with rapid CKD progression who could be targeted for RAAS inhibitors for renal and cardiovascular risk reduction [175]. The use of complementary molecular biomarkers to help in the identification of patients with potentially progressive diseases, may delay the need of renal replacement therapy, a very expensive treatment which consumes a high percentage of healthcare budgets. An important factor to take into consideration is the small effect that individual-omic variants identified may have on complex traits, such as CKD and other kidney diseases, and consequently the elevated number of patients needed to test to obtain benefit. It will be necessary to integrate panels of several biomarkers to increase the prediction of CKD progression. Although the cost of conventional biomarkers is insignificant compared to molecular biomarkers, the construction of disease-customised chips for routine use may reduce the prices considerably and make them cost-effective, especially if they are designed to identify patients with a poor prognosis.

Genomics biomarkers face additional methodological challenges ahead, including complexity and cost-effectiveness, the establishment of adequate gold standards, standardisation of the different technologies and validation of the biomarkers in clinical trials, that need to be addressed properly before any clinical implementation can be incorporated into guidelines.