A Common RET Variant Is Associated with Reduced Newborn Kidney Size and Function

RESULTS

We scanned the NCBI dbSNP database (see Appendix) and identified two common variants (minor allele frequency >10%) of the RET gene coding region that might alter its tyrosine kinase receptor function during kidney development. In exon 11, an A/G substitution at mRNA position 2251 (rs1799939) occurs in approximately 16% of white alleles, changing gly691 to ser691. In exon 7, an A/G substitution at mRNA position 1476 (rs1800860) occurs in 25% of white alleles and lies within an exonic splice enhancer (ESE) sequence (Figure 1A). The critical position of this substitution suggests that it might modify pre-mRNA splicing. We used ESEfinder¹¹ to calculate the effect of the rs1800860 polymorphism on binding to proteins (SF2/ASF, SC35) of the mRNA spliceosome (Table 1). In each case, the 1476(A) allele reduces RET affinity score below the predicted threshold for effective binding to these components of the splicing machinery.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

1476A SNP in human RET mRNA is associated with an aberrant 190-bp transcript. (A) Genomic structure of the human RET gene, indicating the 1476(G/A) SNP within an ESE of exon 7. (B) Nested primers spanning exon 6 to exon 17 were used to amplify RET transcripts in reverse-transcribed RNA extracted from various human cell lines. Only the predicted wild-type 1252-bp transcript is evident in the homozygous (1476 G/G) Wit49 Wilms’ tumor cell line (lane 3); however, an aberrant 190-bp transcript is noted in heterozygous (1476G/A) SH-SY5Y neuroblastoma cells (lane 2) and in homozygous 1476(A/A) SK-N-BE(2) neuroblastoma cells (lane 1). The aberrant 190-bp transcript was not seen in four other 1476G/G Wilms’ tumor, renal cell carcinoma, and ovarian carcinoma cell lines (data not shown). (C) Sequence analysis of the aberrant 190-bp band demonstrates fusion of RET exon 7a to exon 14b with a heterogeneous 18-bp intervening sequence which consisted of nucleotides continuing beyond 1476(A) into exon 7b (upper sequence) or nucleotides from exon 14b upstream of mRNA position 2500 through 2518 (down sequence). This suggests aberrant splicing between sites after the ESE in exon 7 and two alternative cryptic splice sites (position 2500 and 2518) in exon 14b.

To confirm that the minor RET1476(A) allele alters normal mRNA splicing, we studied three RET-expressing human cell lines: (1) SK-N-BE(2) neuroblastoma cells (1476A/A); (2) SH-SY5Y neuroblastoma cells (1476G/A); and (3) Wit49 Wilms’ tumor cells (1476G/G). RET transcripts were amplified with nested primers spanning exon 6 through exon 15. Only the predicted wild-type 1252-bp transcript is evident in Wit49 (1476G/G) cells, whereas an aberrant 190-bp transcript is noted in the two cell lines bearing one or more 1476A alleles (Figure 1B). The 190-bp transcript was not seen in any of four other 1476(G/G) cell lines (data not shown). Sequencing of the 1252-bp band in the heterozygous (rs1800860) SH-SY5Y cell line demonstrated that normal splicing of exon 7 can occur with either nucleotide at mRNA position 1476 (Figure 2); however, when the minor transcript in SK-N-BE(2) or SH-SY5Y cells was sequenced, we also noted aberrant splicing from 1476A to two cryptic splice sites in exon 14 (Figure 1C). This del(7b-14a) transcript eliminates part of the cadherin-like domain 4 (CLD4), the entire cysteine-rich domain (CRD), transmembrane domain, and the first tyrosine kinase domain, undoubtedly rendering the protein product nonfunctional (Figure 3). The CLD4 and CRD are the extracellular sites of interaction with RET ligand (GDNF) and co-receptor (GFRα1).^12,13 To confirm that this aberrant splicing activity occurred in other cell lines, we also studied the heterozygous (1476G/A) Wilms’ tumor G401 cell line and found the same del(7b-14a) transcript (data not shown).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.

Sequence of wild-type human RET exon 7 (NM_020630). After reverse transcription of mRNA from heterozygous (rs1800860) SH-SY5Y cells, exon 7 was amplified by PCR and the normal 1252-bp amplicon was sequenced. The presence of either G or A at position 1476 demonstrates that normal splicing of RET exon 7 can occur with both alleles.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3.

The aberrant del(7b-14a) RET transcript lacks key functional domains. CLD1 through 4, CRD, transmembrane segment (TM), juxtamembrane segment (JM), and two cytoplasmic tyrosine kinase domains (TK1 and TK2) are displayed adjacent to the corresponding exon structure of the wild-type RET transcript (WT). To the left is the aberrant del(7b-14a) transcript associated with 1476(A) alleles; deletion of extracellular domains for ligand (GDNF) and co-receptor (GFRα1), transmembrane segments, and a large portion of TK1 is predicted, rendering the transcript nonfunctional.

Because both the 1476A and 2251A single-nucleotide polymorphisms (SNP) are potentially hypomorphic variants of the RET gene, we hypothesized that either one could compromise nephrogenesis during kidney development, resulting in subtle renal hypoplasia in a significant fraction of newborns. To validate newborn kidney volume as a surrogate for congenital nephron number, we measured glomerular number in kidneys from 15 infants who died of congenital anomalies or sudden infant death syndrome at or before the age of 3 mo. Nephron number ranged from 246,181 to 1,106,062. A strong direct relationship (P = 0.019) between kidney mass and total nephron number in children aged ≤3 mo was identified (Figure 4). Regression analysis predicted an additional 23,459 (95% confidence interval 4590 to 42,238) glomeruli per gram of kidney mass in the range.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4.

Relationship between N_glom (with 95% confidence interval) and mass of right kidney at autopsy in infants aged ≤3 mo. N_glom was measured by the disector/fractionator method as described previously in autopsied kidneys from infants who died from various congenital anomalies or sudden infant death syndrome within the first 3 mo of life.

We then recruited 136 normal term newborn white infants born at the Royal Victoria Hospital in Montreal between 2004 and 2006, excluding any with low birth weight (<2500 g), renal malformation, maternal diabetes, or twin pregnancy. Total (combined) renal volume was estimated by ultrasonography within 48 h of birth and normalized for body surface; cord blood was obtained for assay of cystatin C as a surrogate of newborn GFR.^14–16 In previous studies, we found that newborn kidney volume correlated directly with body surface area and inversely with cord blood cystatin C.¹⁴ Clinical characteristics of the cohort are summarized in Figure 5. The distribution of cystatin and kidney size in our cohort approached normality (skewness <2 for all parameters).

• Download figure
• Open in new tab
• Download powerpoint

Figure 5.

Clinical characteristics of the newborns of white cohort (n = 136). (A) Total KidVol/BSA in the white newborn cohort (mean ± SD 132.16 ± 29.34 ml/m²). (B) Cord blood cystatin C (mean ± SD 1.93 ± 0.32 mg/L). (C) Clinical characteristics of the newborn cohort.

DNA extracted from cord blood was used to genotype each infant for the rs1800860 (1476G/A) and rs1799939 (2251G/A) SNP. Genotype distribution for each SNP conformed to the Hardy-Weinberg equilibrium (P = 0.39 and 0.47, respectively); SNP frequencies in our cohort (Table 2) were similar to those reported in the CAUC1 and CEU populations (NCBI dbSNP database; see Appendix). RET2251(G/A) SNP was associated with neither renal volume nor cord cystatin C; however, total kidney volume factored for body surface area (KidVol/BSA) in newborns bearing one or more 1476(A) alleles was 9.7% smaller than that in newborns with the homozygous 1476(G/G) genotype (P = 0.009). The 1476(A) allele was also associated with 9.2% increase in cord blood cystatin C concentration (P = 0.002), suggesting a comparable reduction in functional renal mass¹⁵ (Table 2).

The 1476(A) allele was widely distributed in our population; combined renal volume in infants with a 1476(A) allele ranged from 16.06 to 44.48 ml versus 15.01 to 50.33 ml in infants who were homozygous for the more common 1476(G) allele. Thus, the association between 1476(A) allele and renal volume cannot be attributed to a few infants with very small kidneys. Similarly, the distribution of left/right kidney volume ratios was similar in newborns with one or more 1476(A) alleles (0.93 ± 0.22 SD) versus newborns homozygous for the 1476(G) allele (0.96 ± 0.18 SD) (P > 0.05). Thus, the association between 1476(A) and renal volume was not due to a few infants with unilateral renal hypoplasia.

To confirm that the RET1476(A) allele compromises expression of wild-type receptor mRNA, we examined allele-specific mRNA expression in the heterozygous SH-SY5Y and G401 cell lines. RET exon 7 was amplified from genomic DNA and cellular mRNA, using high-sensitivity sequencing technology to quantify each allele-specific amplicon as described by others.¹⁷ mRNA expression from the 1476(A) allele was substantially reduced in both SH-SY5Y (by 25%) and G401 (by 50%) cells, compared with the expression level from the 1476(G) allele (Figure 6).

• Download figure
• Open in new tab
• Download powerpoint

Figure 6.

Comparison of allele-specific mRNA expression in rs1800860(G/A) heterozygous G401 and SH-SY5Y cells. Peak area ratio of G/A alleles was measured in cDNA and genomic DNA from two cell lines heterozygous for the rs1800860 SNP. The G/A allelic ratio (2.09 ± 0.12) in cDNA was 2.1 times greater than the ratio (1.01 ± 0.02) in genomic DNA from G401 cells (*P = 0.004). Similarly, G/A ratio was 1.4 times greater in cDNA (1.41 ± 0.14) than in genomic DNA (0.98 ± 0.03) for SH-SY5Y cells (**P = 0.03).

Our previous studies demonstrated an association between a common (18.5% of white individuals) human PAX2 haplotype (AAA) and a 10% reduction in newborn renal size.¹⁶ Interestingly, RET transcription is regulated by PAX2 gene dosage.¹⁰ When we analyzed our cohort for both genes, renal volume in the 17 of 136 newborns carrying both the RET1476(A) and PAX2(AAA) minor alleles was 23% lower than that in infants with the RET1476(G/G), PAX2(GGG) haplotype (Figure 7).

• Download figure
• Open in new tab
• Download powerpoint

Figure 7.

Comparison of KidVol/BSA in newborns with various combinations of hypomorphic RET^1476(A) and PAX2^AAA alleles. All 136 newborns in our cohort were genotyped for both the RET(1476G/A) SNP and the hypomorphic PAX2^AAA haplotype previously described by Quinlan et al.¹⁵ KidVol/BSA among newborns bearing one or more hypomorphic RET^1476(A) alleles (n = 46) was 90.3% of that for infants with the major RET^1476(G)/PAX2^GGG alleles (n = 41; P = 0.01). Similarly, KidVol/BSA among newborns with one or more hypomorphic PAX2^AAA alleles (n = 14) was 89.5% of RET^1476(G)/PAX2^GGG newborns (P = 0.04). In the subset of newborns bearing both hypomorphic RET^1476(A) and PAX2^AAA alleles (n = 17), KidVol/BSA was only 77% of that in wild-type RET^1476(G)/PAX2^GGG infants (P = 0.00067).

DISCUSSION

During renal development, the branching ureteric bud expresses high levels of the transcription factor PAX2.¹ Brophy and colleagues^10,18 recently showed that, among its many functions, PAX2 directly activates transcription of genes for both RET, a tyrosine kinase receptor, and its ligand, GDNF. Epithelial cells expressing RET receptors cluster at the tip of each ureteric bud branch as it undergoes branching morphogenesis.⁸ When activated by GDNF from nearby mesenchyme, the RET receptor heterodimerizes with GFRα1, stimulating cell proliferation, migration, and survival via several intracellular signals, including the RAS/mitogen-activated protein kinase, phosphatidylinositol 3-kinase/AKT, and RAC1/JUN NH(2) terminal kinase pathways.^19–21 Loss of RET, GDNF, or GFRα1 results in renal agenesis, because of inhibition of ureteric bud growth and branching.^9,22,23

Because RET integrity is critical for branching nephrogenesis, we hypothesized that heterozygous RET mutations might partially compromise the extent of ureteric bud branching during development, leading to suboptimal nephron number. In mutant mice, heterozygous null alleles reduce nephron number by approximately 22% at postnatal day 15.¹⁰ Thus, a heterozygous hypomorphic RET allele such as the 1476(A) variant should fit at the milder end of this spectrum and might plausibly produce the observed 10% decrease in newborn nephron number.

Mouse studies suggest a graded relationship between kidney size and the level of RET function; Ret knockout mice are anephric, whereas severely hypomorphic Ret alleles such as Tyr1062Phe²⁴ or RetDN²⁵ produce severe renal hypoplasia. Thus, only a subtle effect on kidney size would be expected for a RET polymorphism predicted to have only a modest effect on total RET mRNA level; however, the relationship between kidney volume and nephron number is difficult to establish in mice; nephrogenesis continues for at least 2 wk after birth and potentially overlaps with a period of postnatal compensatory hypertrophy. For example, Clarke et al.¹⁰ noted that heterozygous Ret(+/−) postnatal day 15 mice had a 22% reduction in nephron number but only a 10% reduction in kidney volume (NS). To establish that newborn kidney size was a valid surrogate for nephron number in humans, we measured glomerular number and kidney weight in newborn infants who died before 3 mo of age. This showed a strong correlation between renal mass and nephron number in the newborn period.

In this study, we identified a common SNP within the ESE of exon 7, which modifies the fidelity of mRNA splicing to the normal splice site in exon 8. An adenine nucleotide at position 1476 of this ESE increases the risk for aberrant splicing to alternative sites in exon 14. When this occurs, exons containing the crucial GFRα1 binding site, transmembrane, and first tyrosine kinase domains all are deleted, undoubtedly rendering the RET receptor dysfunctional. On the basis of our studies in cultured heterozygous human cells, the presence of a 1476A allele reduces wild-type mRNA expression by approximately 38%; in a heterozygous cell, this would amount to an overall reduction of functional RET transcript by 19%. Arguably, branching morphogenesis of the ureteric bud during fetal kidney development might be reduced in proportion to this reduction of RET expression. In our cohort of normal white infants from Montreal, the 1476(A) minor RET allele, infants with one or more 1476(A) alleles exhibited a 9.7% reduction in newborn kidney volume (normalized for body surface area) and 9.2% reduction in newborn renal function. Thus, the degree of renal hypoplasia observed in infants (9 to 10%) with one or more 1476(A) alleles is roughly commensurate with the reduction in functional RET mRNA expression (19%) measured in vitro.

To put this in a clinical perspective, the report of Keller et al.⁵ suggested that adults with essential hypertension are born with 47% fewer nephrons compared with those who remain normotensive. Thus, the effect of the 1476(A) allele could account for only one fifth of this clinically relevant congenital nephron deficit. Clearly, additional genes are involved in setting nephron number during development.

Heterozygous null mutations of RET have been reported in humans with Hirschprung disease. If arborization of the ureteric bud is proportional to the level of RET expression during renal development, then one might expect that congenital nephron number in Hirschprung disease should be reduced. There are reports of unilateral renal aplasia and various renal malformations in Hirschprung disease,²⁶ but nephron number has not been carefully assessed. By the time patients with Hirschprung disease are identified, there has been ample time for compensatory renal hypertrophy to erase the initial relationship between kidney size and congenital nephron number. Thus, in the report of Keller et al.,⁵ describing 50% reduction in nephron number among people with essential hypertension, adult kidney mass was similar to that in normal control subjects. In contrast, we measured renal volume and function at birth before the relationship can be masked by postnatal compensatory hypertrophy.¹⁰ Indeed, our findings from autopsied newborns suggested that kidney weight correlates with congenital nephron number (adjusted for age) for up to 3 mo.

Ureteric bud cells express high levels of PAX2 during renal development; homozygous Pax2 mutant mice are anephric, and heterozygous Pax2 null mutants exhibit significant renal hypoplasia.^27–29 We recently identified a fairly common (allele frequency 0.2) hypomorphic human PAX2^AAA allele, which reduces PAX2 expression by 40% compared with the major wild-type PAX2^GGG allele; thus, total PAX2 mRNA in a heterozygous cell is approximately 80% of normal.¹⁰ Total newborn kidney volume among infants with one or more PAX2^AAA alleles was approximately 10% smaller than in PAX2^GGG/ PAX2^GGG infants.¹⁶ Fifteen percent of the newborn cohort in this study were compound PAX2^AAA/RET^1476A heterozygotes. In this subgroup, newborn KidVol/BSA was 23% smaller than in infants with homozygous wild-type RET^1476G/PAX2^GGG alleles. This corresponds very nicely with the observations of Clarke et al.,¹⁰ who found additive effects of compound heterozygosity for Pax2 and Ret alleles on nephron number in mutant mice. Thus, together, the two hypomorphic RET and PAX2 alleles could account for up to half of the nephron deficit (48% of control) associated with essential hypertension reported by Keller et al.⁵

On the basis of HapMap linkage data in the white (CEPH) population, the 1476(A) SNP is not in high genetic linkage disequilibrium with most other portions of the RET gene; however, to rule out the unlikely possibility that some other site within the RET gene is responsible for reduced kidney size, we screened for association with 22 other haplotype-tagging SNPs spanning all major linkage blocks and 10 kb to either side of the coding sequence. No other haplotype-tagging SNPs were significantly associated with either kidney size or cystatin C level (data not shown).

Interestingly, we identified a second RET SNP (rs1799939) that alters a glycine at amino acid position 691 to a serine (RET2251G/A); however, this SNP was not associated with newborn kidney size or cystatin C. Because the G/A substitution lies within a linker region between the RET transmembrane and tyrosine kinase domains, the 2251A isoform may retain sufficient signaling activity to permit normal nephrogenesis.

In conclusion, we identified a polymorphic variant (RET^1476A) of the human RET receptor that increases the risk for aberrant mRNA splicing and causes decreased expression of the functional wild-type allele. The RET^1476A is associated with a reduction (approximately 10%) of newborn kidney volume and an increase (approximately 9%) in umbilical cord cystatin C. Because kidney size and function were assessed at birth, before the period of postnatal hypertrophy, these measurements likely reflect congenital nephron endowment. In mice, homozygous RET mutations block ureteric bud outgrowth, and heterozygous RET null mutations have been shown to interfere with optimal nephrogenesis.¹⁰ Similarly, we previously showed that heterozygous PAX2 mutations interfere with ureteric bud branching and reduce congenital nephron number in mice; a common polymorphic variant of PAX2 causes 10% reduction in human newborn kidney volume.¹⁶ Among the 15% of normal newborns who inherit both a hypomorphic RET^1476(A) and a hypomorphic PAX2^AAA allele, kidney volume is reduced by 23% of wild-type controls. Clarke et al.¹⁰ also noted a synergistic effect of Pax2 and Ret mutations on ureteric bud branching and nephrogenesis in embryonic mouse kidney explants. Our observations suggest a model in which common polymorphic variants of genes involved in renal branching morphogenesis account for subtle renal hypoplasia in the normal human population.

CONCISE METHODS

APPENDIX

DISCLOSURES

None.