CDH12 as a Candidate Gene for Kidney Injury in Posterior Urethral Valve Cases: A Genome-wide Association Study Among Patients with Obstructive Uropathies

Background Posterior urethral valves (PUVs) and ureteropelvic junction obstruction (UPJO) are congenital obstructive uropathies that may impair kidney development. Objective To identify genetic variants associated with kidney injury in patients with obstructive uropathy. Design, setting, and participants We included 487 patients born in 1981 or later who underwent pyeloplasty or valve resection before 18 yr of age in the discovery phase, 102 PUV patients in a first replication phase, and 102 in a second replication phase. Outcome measurements and statistical analysis Signs of kidney injury were defined as dialysis, nephrectomy, kidney transplantation, estimated glomerular filtration rate (eGFR) <60 ml/min/1.73 m2, high blood pressure, antihypertensive medication use, proteinuria, and/or one kidney functioning at <45%. We used χ2 tests to calculate p values and odds ratios for >600 000 single-nucleotide polymorphisms (SNPs) in the discovery sample comparing patients with and without signs of kidney injury within 5 yr after surgery. We performed stratified analyses for PUV and UPJO and Kaplan-Meier and Cox regression analyses in the discovery and two replication samples for the associated SNPs, and RNA and protein expression analyses for the associated gene in fetal tissues. Results and limitations Despite the small and nonhomogeneous sample, we observed suggestive associations for six SNPs in three loci, of which rs6874819 in the CDH12 gene was the most clear (p = 7.5 × 10–7). This SNP also seemed to be associated with time to kidney injury in the PUV discovery and replication samples. RNA expression analyses showed clear CDH12 expression in fetal kidneys, which was confirmed by protein immunolocalization. Conclusions This study identified CDH12 as a candidate gene for kidney injury in PUV. Patient summary We found that variants of the CDH12 gene increase the risk of kidney injury in patients with extra flaps of tissue in the urethra (posterior urethral valves). This is the first report on this gene in this context. Our study provides interesting new information about the pathways involved and important leads for further research for this condition.


Introduction
Posterior urethral valves (PUV) and ureteropelvic junction obstruction (UPJO) are congenital anomalies of the urinary tract that impair urinary flow. PUV affects only boys and occurs in one in 4000 live male births [1], while UPJO affects one in 500 children [2]. These obstructions may perturb kidney development and are referred to as obstructive uropathy [3]. Obstructive uropathy is the second most prevalent cause of end-stage kidney disease in children [4], but long-term kidney function is variable and cannot be predicted reliably.
Knowledge about the molecular pathways involved in the pathophysiology of kidney injury in congenital obstructive uropathy has mainly been derived from neonatal rats and mice with induced complete unilateral ureteral obstruction (UUO). Kidney response to experimental UUO involves oxidative stress, monocyte infiltration, uncontrolled apoptosis of tubular cells, fibroblast accumulation, and increased deposition of extracellular matrix [3]. This is similar to the response in human obstructed kidneys, where the extent of monocyte infiltration correlates with the intensity of tubulointerstitial damage, and inflammatory molecules, such as MCP1, are upregulated [5]. In addition to elements from proinflammatory pathways, several other molecules have been suggested to play a pathological role in kidney injury in obstructive uropathy in animal models (such as angiotensin II and Tgf-b) [3,6,7]. Several of these were confirmed in human expression studies (including TGF-b1, REN, and EGF) [5,6,8,9]. Finally, copy number variations (CNVs) >100 Kb were associated with kidney failure in PUV patients [10], while polymorphisms in ACE and AGTR2 have been associated with lower estimated glomerular filtration rate (eGFR), chronic kidney disease, and kidney scarring in PUV patients [11][12][13].
Most of the genetic studies described above focused on candidate genes, but a hypothesis-generating approach, such as a genome-wide association study (GWAS), may shed new light on pathways involved in the development of kidney injury in obstructive uropathy. In addition, genetic variants associated with kidney function decline could be used in prognostic modeling approaches to better predict long-term kidney function. As no such studies have been conducted so far, the aim of this study was to perform a GWAS for kidney injury in obstructive uropathy. We included patients with PUV and UPJO under the assumption that similar genetic variants are associated with kidney function decline in both obstructive uropathies.

2.
Patients and methods Outcome measurements and statistical analysis: Signs of kidney injury were defined as dialysis, nephrectomy, kidney transplantation, estimated glomerular filtration rate (eGFR) <60 ml/min/1.73 m 2 , high blood pressure, antihypertensive medication use, proteinuria, and/or one kidney functioning at <45%. We used x 2 tests to calculate p values and odds ratios for >600 000 singlenucleotide polymorphisms (SNPs) in the discovery sample comparing patients with and without signs of kidney injury within 5 yr after surgery. We performed stratified analyses for PUV and UPJO and Kaplan-Meier and Cox regression analyses in the discovery and two replication samples for the associated SNPs, and RNA and protein expression analyses for the associated gene in fetal tissues. Results and limitations: Despite the small and nonhomogeneous sample, we observed suggestive associations for six SNPs in three loci, of which rs6874819 in the CDH12 gene was the most clear (p = 7.5 Â 10 -7 ). This SNP also seemed to be associated with time to kidney injury in the PUV discovery and replication samples. RNA expression analyses showed clear CDH12 expression in fetal kidneys, which was confirmed by protein immunolocalization.
Conclusions: This study identified CDH12 as a candidate gene for kidney injury in PUV. Patient summary: We found that variants of the CDH12 gene increase the risk of kidney injury in patients with extra flaps of tissue in the urethra (posterior urethral valves). This is the first report on this gene in this context. Our study provides interesting new information about the pathways involved and important leads for further research for this condition.

Dutch replication sample
For replication purposes, we included patients treated in the Isala clinics, Zwolle or University Medical Center (UMC), Groningen, The Netherlands.
Owing to registration issues, we were only able to identify patients treated in 2002 or later.

European replication sample
The European replication sample was derived from the 4C study [16]. In brief, a cohort of children aged 6-17 yr with chronic kidney disease (eGFR 10-45 ml/min/1.73 m 2 ) was enrolled between 2010 and 2012 and followed until 2018 at 39 pediatric nephrology centers in eight European countries. We selected children with PUV as the underlying diagnosis for whom genotype data, longitudinal eGFR measurements, and information on treatment changes were available.

Kidney injury
The medical files for the Dutch patients were scrutinized for information about kidney function and for clinical descriptions.
Patients were defined as having signs of kidney injury in case of dialysis, nephrectomy, kidney transplantation, eGFR <60 ml/min/

Statistical analysis
We excluded patients with <5 yr of follow-up from our discovery sample and compared patients with and without signs of kidney injury within 5 yr after surgery using the PLINK whole-genome data analysis toolset (www.cog-genomics.org/plink/1.9/). Analyses were adjusted for the first four principal components to account for population stratification. We  (https://training.cochrane.org/online-learning/core-softwarecochrane-reviews/revman/revman-5-download) using the inverse-variance method and a random-effects model.

Expression analyses
We used the Expression Atlas (www.ebi.ac.uk/gxa/home) in February 2019 to see which tissues CDH12 is expressed in [17]. We tested expression in adult human kidney and brain and in commercially available mRNA from human fetal kidneys (Clontech Laboratories, Mountain View, CA, USA) via quantitative polymerase chain reaction. We used data from our transcriptomic analysis of human organogenesis covering 15 embryonic tissues and organs 33-55 d after conception [18] to identify embryonic tissues in which the gene is expressed. PUV = posterior urethral valve; UPJO = ureteropelvic junction obstruction; eGFR = estimated glomerular filtration rate. a Urinary tract infections after surgery were not taken into account. b Kidney abnormalities defined as the presence of kidney scars or cysts, kidney dysplasia, increased echogenicity, and/or loss of parenchyma. c Bladder dysfunction defined as training or use of medication to improve bladder function, intermittent catheterization, or vesicostoma/ureterocutaneostoma. d Signs of kidney injury developed during the maximum follow-up period of 10 yr after surgery. Some patients developed more than one of these signs, which is why the numbers do not add up to the total number of patients with signs of kidney injury.
We used the Protein Atlas (www.proteinatlas.org/) in February 2019 [19] and performed immunohistochemistry of human fetal tissues as previously described [20], using a primary antibody to CDH12 (ab71055; Abcam, Cambridge, UK) at a dilution of 1:100. Tissue processing was as previously described [21].
Finally, we checked the GTEx portal (https://gtexportal.org/home/) in February 2019 to see whether the SNPs that we identified are expression quantitative trait loci (eQTL) influencing expression of CDH12.

Discovery sample
AGORA contained DNA from 433 eligible patients. We identified 247 additional patients, of whom 151 (61%) participated ( Supplementary Fig. 1). We excluded 40 patients who did not have their first surgery (pyeloplasty or PUV resection) in the Radboudumc, four patients with end-stage kidney disease before surgery, and five patients without follow-up information. Of the remaining 535 samples, 12 failed genotyping and 36 failed sample quality control (one sample with a call rate <97%, one discordant sex information, four related individuals, and 30 ethnic outliers; Supplementary Fig. 2). Genotyping information for 487 patients was available for the discovery analyses (Table 1).

Dutch replication sample
For replication purposes, we identified 131 patients treated in the Isala clinics and 270 in UMC Groningen. Of these patients, 87 (66%) and 164 (61%), respectively, participated. We excluded three patients without follow-up information and 26 samples that failed quality control (two samples with a call rate <97%, one related individual, and 23 ethnic outliers; Supplementary Fig. 2). We used genotyping information for 222 patients in the replication analyses.

European replication sample
The European replication sample was derived from a study including 704 children [16]. PUV was the underlying diagnosis for 119 children, and all necessary information was available for 102 of these.

Discovery analyses
We excluded 130 patients with <5 yr of follow-up and compared 141 patients with and 216 patients without signs of kidney injury within 5 yr after surgery. Six SNPs in three loci reached suggestive genome-wide significance (p < 1 Â 10 -5 ; Fig. 1 and Supplementary Table 1), with rs6874819 on chromosome 5 being the most significant SNP (p = 7.5 Â 10 -7 ). The other signals were from chromosomes 8 (rs2957086; p = 9.3 Â 10 -6 ) and 13 (rs9580025 and rs2148707; both p = 8.2 Â 10 -6 ). Genotype frequencies were in Hardy-Weinberg equilibrium in controls (p values 0.23-0.95), except for rs2957086 (p = 0.04). After imputation, rs6874819 remained the SNP with the strongest signal. Subgroup analyses for the six SNPs in the PUV (n = 235) and UPJO (n = 122) groups separately showed that the signals on chromosomes 5 and 13 were stronger in the PUV group, while the signal on chromosome 8 was stronger in the UPJO group (Supplementary Table 2).
Kaplan-Meier and Cox regression analyses were performed to include 130 patients with <5 yr of follow-up. These analyses included 316 PUV and 171 UPJO patients, of whom 93 and 90, respectively, developed signs of kidney injury within 10 yr after surgery. Variant genotypes of the SNPs on chromosomes 5 and 13 were statistically significantly associated with time to the development of signs of kidney injury among PUV patients ( Fig. 2A for rs6874819), whereas the results for UPJO patients were less consistent ( Table 2).
In an additional analysis, we adjusted the Cox regression analyses for rs6874819 among the discovery PUV patients for the possible prognostic factors from

Replication analyses
Of the 222 Dutch replication samples, 78 were genotyped on the same platform as the discovery samples, while 144 were genotyped on the Global Screening array. Only two of the six suggestive SNPs (rs6874819 and rs2148707) were directly genotyped with the latter, so we used imputed [ ( F i g . _ 1 ) T D $ F I G ]  Table 2 and Fig. 2C).

Meta-analysis
We performed Cox regression analyses for the allelic effects in the discovery and replication samples, and subsequently performed a meta-analysis. This result was statistically significant at a genome-wide level, with a HR for the G allele of 2.3 (95% confidence interval 1.7-3.0; p = 4.1 Â 10 -9 ; Table 3).

Expression analyses
Visualization of the association results using imputed genotypes for the discovery sample in LocusZoom [22] showed that the peak on chromosome 5 was located in an intron of the CDH12 gene ( Supplementary Fig. 3). All studies in the Expression Atlas [17] revealed that CDH12 shows the highest expression in human adult brain, whereas it is not expressed in adult kidney. The NIH Epigenomics Roadmap (http://www.roadmapepigenomics.org/) showed that CDH12 is expressed in human fetal kidneys [23]. We confirmed that mRNA isolated from human adult kidney shows no CDH12 expression, whereas adult brain tissue does. Fetal kidneys showed expression levels comparable to those in adult brain. Transcriptomic analyses of 15 human embryonic tissues and organs 33-55 d after conception revealed that CDH12 is widely expressed in embryonic tissues, with highest levels in retinal pigmented epithelium, followed by the kidneys [18].
Immunohistochemical staining confirmed CDH12 expression in the fetal retina and kidney. At approximately 12 wk after conception, CDH12 was expressed in the ureteric bud, the precursor of the collecting duct, and in early condensates, where the ureteric bud induces mesenchymal cells to condense and start to form epithelia. In addition, CDH12 expression was localized to proximal tubules within the developing human kidney at 12 and 15 wk after conception (Fig. 3).
The GTEx portal showed that the associated area was an eQTL for CDH12 in the adrenal gland. Several SNPs in the associated area resulted in higher expression of CDH12 ( Supplementary Figs. 4 and 5). Two of these SNPs were [ ( F i g . _ 3 ) T D $ F I G ]  Table 3).

Discussion
We identified for the first time CHD12 as a gene potentially involved in the development of kidney injury in patients with obstructive uropathies. This result was seen among PUV patients, but not clearly among UPJO patients, possibly because of the different timing of PUV and UPJO development, or because both kidneys are exposed to high urinary pressure in PUV, whereas generally only one kidney is in UPJO.
The association signal on chromosome 5 among PUV patients was the most clear and the other suggested SNPs were not associated with kidney injury in the Dutch replication sample. Detection of the signal for CDH12 despite our small and nonhomogeneous sample suggests that it may be an important factor. This was confirmed by detection of the signal in the European PUV replication sample, which was a very different patient group. Although the primary results for rs6874819 were not statistically significant at a genome-wide level, the result for the allelic meta-analysis among PUV patients was. In addition, all results show the same direction of effect. The fact that our association signal is located in an eQTL and that CDH12 is expressed in fetal kidneys further enhances the reliability of our finding.
In an additional analysis, we adjusted our analyses in the European replication sample for eGFR at study entry, which changed the effect of rs6874819 only slightly. Since eGFR at study entry correlates with GFR at birth, this finding suggests that the effect of rs6874819 on kidney injury we observed is not sufficiently explained by effects of CHD12 on fetal kidney development. Our hypothesis is that the damage arises in the period when affected kidneys are exposed to elevated pressure, which is from early fetal development up to the postnatal period preceding surgery. The damage may not be severe enough to display signs of kidney injury at that point in time, but as the patients grow and need greater filtration capacity, the affected kidneys may fall short and the previously acquired damage begins to show.
CDH12 is a member of the neural cadherin (N-cadherin) gene family, which encode transmembrane calcium-de-  pendent homophilic adhesion receptors that play an important role in cell recognition and sorting during development [24]. CDH12 has not been described in relation to kidney development before, but it is involved in the progression of several cancers [24][25][26][27]. In addition, CDH12 may play a role in monocyte infiltration. Grandaliano et al [5] showed that the extent of monocyte infiltration in kidney tissue of children with UPJO correlated with the degree of tubulointerstitial damage, suggesting that monocyte infiltration is an important event in the development of kidney injury in obstructive uropathy. They also showed a striking increase in MCP1 gene expression in these kidney tissues, which is in line with the suggestion by Diamond et al [28] that upregulation of Mcp1, in response to an increase in tubular pressure, stimulates monocyte influx in a rat model of UUO. Niu et al [29] found that MCP1 action in inducing angiogenesis is mediated via induction of a new transcription factor that they named MCP1-induced protein (MCPIP). Chromatin immunoprecipitation analysis showed that CDH12 is a target of MCPIP [29]. Ma et al [27] confirmed that CDH12 expression is modulated by MCP1 in colorectal cancer cells. A drawback of our study is the relatively small sample size. We may have missed less strong but still important associations because our discovery analyses only had enough power (80%) to detect variants with an allele frequency of 30% and genotype relative risk of 2.7 or higher under a multiplicative disease model. In addition, use of a more homogeneous cohort of, for example, only prenatally detected PUV, or a more homogeneous endpoint, such as the eGFR decrease in the first yr of life, may have been better. Instead, we used any sign of kidney injury as the outcome to achieve larger numbers, but did not have enough patients with specific signs to perform separate analyses for subgroups. Nevertheless, when we performed additional analyses that excluded one kidney functioning at <45% from the definition of signs of kidney injury (the diagnosis that explains why more UPJO than PUV patients have signs of kidney injury), results for PUV patients in the discovery sample were very comparable to the initial results. Using a more homogeneous cohort would result in patients being more comparable for other factors that could influence the prognosis of PUV [30]. However, when we adjusted the Cox regression analyses for PUV patients of the discovery sample for the possible prognostic factors from Table 1, the results were very comparable to the unadjusted results for rs6874819. This suggests that the effect of rs6874819 is independent from these factors and that it may influence the risk of signs of kidney injury in all PUV patients, regardless of severity or management.

Conclusions
We hypothesize that MCP1 pathway upregulation due to an increase in urinary tract pressure in PUV patients is augmented in carriers of the CDH12 variant, which facilitates monocyte infiltration and the development of progressive kidney fibrosis and exacerbates kidney injury. This hypothesis is in line with the importance of monocyte infiltration in the development of kidney injury in obstructive uropathy and provides leads for further research. The role of CDH12 in monocyte infiltration and the development of kidney injury in PUV, and the predictive value of rs6874819 for the prognosis of individual PUV patients are important future research topics, as is the generalizability of our findings to other forms of obstructive uropathy.
Financial disclosures: Loes F.M. van der Zanden certifies that all conflicts of interest, including specific financial interests and relationships and