Influence of Familial Renal Glycosuria Due to Mutations in the SLC5A2 Gene on Changes in Glucose Tolerance over Time

Subjects

A family with renal glycosuria from the Botnia region on the west coast of Finland has been followed clinically for the last 30 years [17]. In addition to renal glycosuria, 8 individuals developed impaired fasting glucose (IFG), impaired glucose tolerance (IGT) or type 2 diabetes (T2D) (Fig 1) and the family were therefore followed by repeated oral glucose tolerance tests (OGTT, 75 g glucose) as part of the Botnia Study [18]. To evaluate whether glucose loss in the urine in individuals with glycosuria would affect glucose tolerance we calculated the changes in the area under the OGTT curve (ΔAUC_OGTT) between the first and last visit (mean follow up time 10.5 years, range 3–22 years). AUC_OGTT was calculated with the trapezoidal method from the plasma glucose concentrations at 0, 30, 60 and 120 minutes during the OGTT. One female family member (No 45 in the pedigree, Fig 1) presented with marked glycosuria associated with symptoms of postprandial hypoglycemia and severe chronic urinary tract infections (predominantly candida infections). Under controlled conditions with a daily intake of 160 grams of carbohydrates, she excreted 55.2 g glucose/24 hours. She died from a gastric cancer at the age of 49 years. A paraffin embedded gastric biopsy was used for extraction of her DNA. Thin sections of the tissue were carefully scraped off the block using a scalpel and put into xylene for paraffin removal and the DNA was extracted using the QIAmp DNA FFPE Tissue Kit (Qiagen, Hilden, Germany) according to the manufacturer’s instructions.

The DNA extracted from the paraffin embedded tissue was unfortunately not of sufficient quality for sequencing; therefore two potentially functional mutations (c.300-303+2del and p.A343V) were imputed from sequencing data from her husband, children and grandchildren and then genotyped in the index case, family member 45, using a Taqman assay and a PCR with genotype specific primers (S2 Table).

To investigate the prevalence of glycosuria in the general population we extracted data on urine glucose excretions during OGTT from 4999 individuals part of the Botnia Study [18]. This was related to diagnosis of IGT/T2D using the Χ² test.

The study was approved by the ethics committee at Vaasa Central Hospital, and written informed consent was obtained from all participants or guardians on the behalf of the minors/children participants.

Whole exome sequencing and validation genotyping

DNA was extracted from blood samples and one μg of high quality DNA was used for sample preparation for sequencing using the TruSeq DNA sample preparation kit v2 (Illumina, CA, USA) before whole exome capture with the TruSeq Exome Enrichment Kit (Illumina, CA, USA) according to the manufacturer’s instructions. A paired end 2x100 base pair sequencing was then performed on a HiSeq 2000 (Illumina, CA, USA).

After base calling and de-multiplexing using CASAVA 1.8.2 (Illumina, CA, USA) the fastq-files were quality assessed with FastQC (http://www.bioinformatics.babraham.ac.uk/projects/fastqc/) and aligned against Hs37d5 using Burrows-Wheeler Aligner (BWA) v. 0.6.8 [19]. The Genome Analysis Tool Kit [20] (GATK) v. 1.6.2 was used on the resulting BAM-files for local realignment around known indels (IndelRealigner), for marking of duplicate reads (MarkDuplicates) and for base quality recalibration (CountCovariates and TableRecalibration). Variants were called using GATKs Unified Genotyper and variant recalibration (VariantRecalibrator and ApplyRecalibrations) [21]. This procedure was done according to the GATK best practices (https://www.broadinstitute.org/gatk/guide/best-practices).

As most mutations that cause FRG are located in the SLC5A2 gene [22], the initial filtering was restricted to variants within the region of this gene. Resulting variants were analysed with SnpEff v3.0 [23] and SIFT [24] to evaluate functionality and potential effect of the variants on the resulting protein. Only variants predicted to be detrimental to the protein were tested for association with renal glycosuria in the family using Fisher’s exact test. The ExPASy Translate tool [25] was used to predict the effect of mutations on the translation of the mRNA.

The selected variants were validated in 23 family members by genotyping using custom made TaqMan assays (Life Technologies, CA, USA) according to the supplier’s instructions.

Functional studies expressing SGLT2 mutation in HEK293-cells

We also tested whether the missense mutation in the SLC5A2 gene (see below) influenced glucose transport and reabsorption in the tubule by expressing the wild type (WT) and mutant form of the protein (SGLT2) in HEK 293 [26] cells lacking the natural SGLT2 [27]. The WT cDNA sequence (S1 table) of SLC5A2 was synthesized and mutated (GenScript USA Inc., NJ, USA) and the resulting cDNA was cloned into the neomycin resistance containing pCMV6-Neo vector (OriGene Technologies, Inc., MD, USA). The constructs were separately transfected into the HEK293 cell line in triplicates, and successfully transfected cells were selected using 0.8 mg/ml of G418 (Merk Millipore, MA, USA), an antibiotic often used to select for neomycin resistant cells, for 14 days. The cells were subsequently cultured in D-MEM (Cat# 41966029, Life Technologies, CA, USA) supplemented with 10% fetal bovine serum and 1% penicillin/streptomycin with a 50 μg/ml maintenance dose of G418. SLC5A2 mRNA expression was measured by SYBR-green qPCR using SLC5A2 mRNA specific primers designed to span exon junctions (S2 Table).

To evaluate the effect of the missense mutation in SLC5A2 on glucose transport the transfected HEK293 cells were assayed using the Glucose Uptake Cell based Assay Kit (Cayman Chemical, MI, USA). Triplicates of the WT and mutant SGLT2 cell lines and the untransfected HEK 293 cell line were seeded at 2x10⁴ cells per well in a 96-well plate and incubated over night. The relative mRNA expression was measured to compare the transfection and transcription efficiency of the six cell lines. The next morning, the cells were incubated in glucose and serum free medium containing 150 μg/ml of the fluorescent glucose analogue 2-deoxy-2[(7-nitro-2,1,3-benzoxadiazol-4-yl)amino]-D-Glucose (2-NBDG) for 30 minutes. The cells were then washed and treated according to the manufacturer’s instructions before measuring fluorescence of the cells using an Infinite 200 Pro plate reader (Tecan, Männedorf, Switzerland). Untransfected cells were treated with the glucose uptake inhibitor Apigenin to control for unspecific uptake and the signal was used as background in the measurements. Relative glucose uptake was calculated as fluorescence normalized against WT-transfected signal and the difference between the WT-transfected cells and the mutant-transfected cells was tested with Student’s t-test.

Screening for SGLT2 mutations in the Botnia Study

To study how common these mutations in the SLC5A2 gene are in the general population we screened 2584 samples from the Botnia study [18] for the c.300-303+2del mutation by genotyping them using a custom TaqMan assay (Life Technologies, CA, USA). Secondly, we screened whole exome sequencing data from the GOT2D consortium for possibly damaging variants in a total of 2760 samples.

SLC5A2 and SLC5A1 expression in human pancreatic islets

As the expression of both SLC5A2 and SLC5A1 has been suggested to be affected in islets from diabetic patients [14] we studied the expression of these two genes in 132 human islet samples from donors with a wide range of glycemia [28] using STAR[29]-aligned RNA-sequence data by the voom [30] and limma [31] R-packages. Expression values were expressed as the logarithmic count per million (log2CPM). The expression of SLC5A2 and SLC5A1 was correlated with insulin and glucagon expression levels and SLC5A2 expression with glucagon secretion at 1 mM glucose using Pearson correlation. All p-values were FDR corrected. The islets were provided by the Human Tissue Lab EXODIAB/LUDC through the Nordic Transplantation Program (http://www.nordicislets.org).

Mutations in the SLC5A2 gene were associated with renal glycosuria in the family

By sequencing the SLC5A2 gene in 23 members of the FRG family three novel and one previously reported variant were found, i.e. 3 single nucleotide variants (SNVs) and one 6 base pair deletion (Table 1). Two SNVs were intronic in no known regulatory region, and the third SNV was only seen in one of 23 samples, and therefor they were not considered any further. The deletion (c.300-303+2del) removed the four last base pairs in exon three as well as two base pairs of the conjunctive splice site and was by the ExPASy Translate tool [25] predicted to cause a premature stop codon, resulting in a shorter 117 residue peptide (S1 Fig).

This heterozygous deletion (c.300-303+2del) segregated with glycosuria in 10 of 14 family members (Fig 1). All carriers of the deletion (c.300-303+2del) excreted glucose in urine (defined as having at least one positive quantitative or qualitative urine glucose tests) compared with 5 out of 14 non-carriers (p = 0.003, Fig 2A). While this segregation was compatible with co-dominant inheritance, the index case (family member 45) with the most severe form of glycosuria, excreting 55.2 g glucose/24 hours compared to other family members excreting between 0.2–20.7 g/24 hours, had only one copy of the mutation and, in contrast to our expectations, was not homozygous for the deletion (see below).

In order to impute the genotype of the index case we genotyped her husband, her two children and three grandchildren. Her husband did not carry any mutation in the SLC5A2 gene but 4 variants were found in the children and the grandchildren (Table 1). Three of the variants were intronic whereas the fourth variant (p.A343V) was a missense mutation in exon 9 of the SLC5A2 gene also found in its heterozygous form in the daughters and one grandchild. Both daughters presented with glycosuria while the grandchild did not (Table 2).

As the SNV was not found in the father of the offspring, it is likely that the missense mutation (p.A343V) is inherited from the index case suggesting that she would be compound heterozygous for the deletion (c.300-303+2del) and the missense mutation (p.A343V). To test this we performed TaqMan genotyping of the 6 bp deletion and PCR using genotype specific primers for the missense mutation of DNA from the index case. These analyses confirmed our hypothesis that she indeed was compound heterozygous for mutations in the SLC5A2 gene as she carried both the 6 bp deletion (c.300-303+2del) and the missense mutation (p.A343V, S2 Fig).

Glucose transport in kidney cells (HEK293) is impaired in carriers of the p.A343V mutation

To confirm that the missense p.A343V mutation affected the glucose transport we expressed the mutation in a human kidney cell line, HEK293, lacking the endogenous SGLT2 protein. The cells expressing the p.A343V transporter showed decreased capability to transport glucose over the cell membrane in vitro relative to the cells expressing the WT transporter (n = 12, p = 0.02, Fig 3).

Influence of the SLC5A2 mutations on change in the glucose tolerance over time

Of 13 glycosuric family members, 5 had normal whereas 8 had abnormal glucose tolerance. To confirm that the glycosuria was due to a reduced threshold for glucose reabsorption in the kidney in these family members and not only a corollary of elevated plasma glucose we compared the prevalence of glycosuria between individuals with normal and abnormal glucose tolerance (IGT or T2D) and between those with the SLC5A2 mutations (6 of whom also had abnormal glucose tolerance). Among 4999 individuals from the Botnia study, 95% of all glycosuric individuals had IGT or T2D and, vice versa, 64% of patients with IGT/T2D had glycosuria (Χ² p<0.0001). Although glycosuria was also present in 9.5% of the individuals with normal glucose tolerance, these individuals had significantly higher glucose area under the curve during OGTT compared to those without glycosuria (n = 323, 1005 mmol·l^-1·min compared to 812 mmol·l^-1·min, p = 0.001). To explore how common the c.300-303+2del mutation was in Finland we genotyped 2584 individuals from Botnia for the c.300-303+2del mutation and identified 10 additional individuals from 5 families with the deletion. Next we analysed whole exome sequence data from the GOT2D consortium including individuals from the Botnia region. Of the 2760 screened individuals two carried the c.300-303+2del mutation and they both came from the Botnia region. Both were glycosuric but they had also diabetes. In addition, in the GOT2D data set we found 36 other missense variants in the SLC5A2 gene, 7 of which were previously reported in dbSNP [32].

To test whether chronic glycosuria as seen with the use of SGLT2 inhibitors would prevent diabetes or deterioration of the glucose tolerance we assessed changes in the glucose tolerance over time in glycosuric carriers and non-carriers of the mutation. As seen in Fig 2, there was no protective effect of the loss of glucose in the urine with respect to change in glucose tolerance (p = 0.67, Fig 2D, S3–S4 Figs) or HbA_1c (p = 0.09 Fig 2B) over time in the mutation carriers. There was also no difference in glucose tolerance at the last visit (p = 0.64, Fig 2C). Since treatment with SGLT2 inhibitors is associated with decreased body weight, an increase in the haematocrit values [33] and an increase in total, LDL and HDL cholesterol concentration [34] we also examined changes in BMI, hematocrit and cholesterol between carriers and non-carriers. We did not observe any differences in the change in BMI (p = 0.33, Fig 2E and S5 Fig), cholesterol (p = 0.14) or haematocrit (p = 0.10) between the two groups. There was furthermore no difference in the change in triglycerides (p = 0.88) or HDL cholesterol (p = 0.20) between the groups.

SLC5A1, but not SLC5A2 expression correlates with glucagon in human pancreatic islets

The mRNA expression of the SGTL2 (SLC5A2) and SGLT1 (SLC5A1) genes was analysed in RNA extracted from human islet donors (n = 131) using RNA sequencing. Expression of SLC5A1 (mean log2CPM = 5.26) was much higher than the very low expression of SLC5A2 (mean log2CPM = -3.19. As a comparison glucagon has a mean expression of log2CPM = 15.02 and insulin log2CPM = 10.16). As disruption of SLC5A2 by siRNA has been shown to influence glucagon (GCG) secretion [14] we further explored this by co-expression analyses. Expression of SLC5A2 did show a strong correlation with the insulin gene (n = 131, r = 0.63, q-value<0.0001), but not with the glucagon gene (Fig 4B, n = 131, r = -0.15, q-value = 0.19) or glucagon secretion at 1 mM glucose (n = 54, r = -0.27, p = 0.11). SLC5A1 correlated with glucagon (Fig 4A, n = 131, r = 0.53, q-value<0.0001), but not with insulin (n = 131, r = -0.15, q-value = 0.11) expression.