Identification of Acer2 as a First Susceptibility Gene for Lithium-Induced Nephrogenic Diabetes Insipidus in Mice

Significance Statement

Lithium causes nephrogenic diabetes insipidus and hypercalcemia in 20% and 10% of patients, respectively, and may lead to metabolic acidosis. To determine the potential role of genetic predisposition in these adverse effects, the authors fed mice from 29 genetically different inbred strains a normal or a lithium-supplemented diet. Some strains developed adverse effects, whereas others did not. Genome-wide association studies revealed eight loci containing different candidate genes that were associated with development of lithium-induced nephrogenic diabetes insipidus. Of these, Acer2 is specifically expressed in the collecting duct; mice lacking Acer2 were more likely to develop lithium-induced nephrogenic diabetes insipidus. These findings demonstrate the importance of genetic variation in susceptibility for lithium-induced adverse effects in mice, and the genes identified may facilitate subsequent identification of human susceptibility genes.

Abstract

Background

Lithium, mainstay treatment for bipolar disorder, causes nephrogenic diabetes insipidus and hypercalcemia in about 20% and 10% of patients, respectively, and may lead to acidosis. These adverse effects develop in only a subset of patients treated with lithium, suggesting genetic factors play a role.

Methods

To identify susceptibility genes for lithium-induced adverse effects, we performed a genome-wide association study in mice, which develop such effects faster than humans. On day 8 and 10 after assigning female mice from 29 different inbred strains to normal chow or lithium diet (40 mmol/kg), we housed the animals for 48 hours in metabolic cages for urine collection. We also collected blood samples.

Results

In 17 strains, lithium treatment significantly elevated urine production, whereas the other 12 strains were not affected. Increased urine production strongly correlated with lower urine osmolality and elevated water intake. Lithium caused acidosis only in one mouse strain, whereas hypercalcemia was found in four strains. Lithium effects on blood pH or ionized calcium did not correlate with effects on urine production. Using genome-wide association analyses, we identified eight gene-containing loci, including a locus containing Acer2, which encodes a ceramidase and is specifically expressed in the collecting duct. Knockout of Acer2 led to increased susceptibility for lithium-induced diabetes insipidus development.

Conclusions

We demonstrate that genome-wide association studies in mice can be used successfully to identify susceptibility genes for development of lithium-induced adverse effects. We identified Acer2 as a first susceptibility gene for lithium-induced diabetes insipidus in mice.

Lithium is the drug of choice for the treatment of bipolar disorders and is also regularly used to treat schizoaffective disorders and depression. Lithium is therefore often prescribed and used by 0.1% of the population.¹ Unfortunately, lithium treatment has different side effects, here called lithiopathies: within a few years, patients may develop the urine concentration defect nephrogenic diabetes insipidus (NDI), metabolic acidosis, and hypercalcemia. Despite these side effects, lithium remains the drug of choice because bipolar disorder has a larger effect on the patient’s quality of life than the lithiopathies, and for many patients there is no good alternative medication.² To lower the burden currently associated with lithium use, it is essential to understand the pathophysiology of, and the susceptibility for, the diverse lithiopathies.

Lithium-induced NDI (Li-NDI) develops in approximately 20% of patients and is thereby the most common cause of NDI. Li-NDI is caused by the inability of the kidney to concentrate its prourine and is characterized by polyuria and polydipsia.^3–5 Urine concentration is mediated by collecting duct principal cells that take up water from the prourine via the luminal water channel Aquaporin-2 (AQP2), which then exits through AQP3 and AQP4 at the basolateral membrane. Urine concentration is tightly regulated via the hormone arginine vasopressin, which is released from the pituitary in response to hypovolemia or hypernatremia and binds its type-2 receptor in the basolateral membrane of principal cells, leading to redistribution of AQP2 from intracellular vesicles to the apical membrane. In Li-NDI, AQP2 is downregulated in the short term, whereas prolonged treatment also reduces the ratio of principal cells to intercalated cells.^6,7 Lithium-induced AQP2 downregulation is a consequence of principal cell lithium entry through the epithelial sodium channel (ENaC).^8–10 Inside principal cells, lithium inhibits the activity of glycogen synthase kinase type 3.¹¹ Moreover, Li-NDI coincides with elevated urinary prostaglandin E 2 (PGE2) levels, which are known to reduce AQP2 abundance and elevate diuresis.¹²

Lithium also affects acid-base balance. Within days, lithium administration impairs acid excretion in rats, dogs, and humans.^13–16 Development of metabolic acidosis is mostly observed in animal models using acute supraclinical lithium doses,^13,14 whereas clinically relevant lithium doses have no or very mild effects on blood pH.^15–20 Rats treated with lithium for 1 month enhanced the number of acid-secreting α-intercalated cells and the abundance of its proton pump ATPase (H⁺-ATPase).^6,15 This is likely a compensatory mechanism to enhance acid excretion and may explain the enhanced net acid excretion in both rats and humans observed with long-term lithium treatment.¹⁸

In addition to effects on the kidney, lithium causes hyperparathyroidism and hypercalcemia in approximately 10% of patients.²¹ Interestingly, hypercalcemia is an independent risk factor for the development of NDI.²² The molecular cause of lithium-induced hyperparathyroidism and hypercalcemia is not well understood, but ex vivo studies demonstrated lithium directly increased the secretion of parathyroid hormone from isolated bovine parathyroid cells.^23,24

The fact that not all lithium-using patients develop the above-described lithiopathies indicates that humans vary in their susceptibility for the development of lithiopathies, which is likely partly determined by genetic factors. Genome-wide association (GWA) could be used to identify these factors, but large and long-term cohort studies are needed to perform such an approach in humans. Because commonly used rodents develop these lithiopathies consistently and in a much shorter time,^6,13,25–27 we used a GWA approach in mice to investigate the potential role of genetic factors in the development of lithiopathies and identify lithiopathy susceptibility genes.

Methods

Experimental Animals Strain Survey

Female mice (7–9 weeks old) from 29 different inbred mouse strains (n=18 per strain) were obtained from The Jackson Laboratory (Bar Harbor, ME). Mice were housed in a climate-controlled facility with a 12-hour light/dark cycle. All mice had free access to food, water, and a salt (sodium chloride) lick throughout the experiment. At 10 weeks of age, 8 mice from each strain received normal rodent diet (7013, NIH-31 Modified; Harlan Laboratories, Madison, WI), while the other ten mice received the same diet supplemented with 40 mmol lithium chloride/kg. During a 20-week interval, the control/lithium diet was initiated for nine mice per strain (four control, five lithium) for a maximal number of three strains per week, while the second cohort of nine mice (four control, five lithium) of a strain started in a different week. At days 8–10 and 26–28, mice were individually housed in metabolic cages (3600M021; Tecniplast) and urine was collected during the last 24 hours of their stay. After their stay in the metabolic cages, blood was obtained through submandibular bleeding. At day 18, blood was collected in Microvette tubes (Sarstedt, Nürmbrecht, Germany) through submandibular bleeding, centrifuged at 3000 × g for 5 minutes to sediment the red blood cells, and serum was stored at −80°C. All animals were housed at The Jackson Laboratory, which is approved by the American Association for Accreditation of Laboratory Animal Care. All animal studies were approved by The Jackson Laboratory’s Institutional Animal Care and Use Committee.

Acer2 Knockout Mice

C57BL/6J-Acer2^em1Mvw/MvwJ mice (Acer2 knockout) (JR#026793) were obtained from The Jackson Laboratory, where they were generated as described.²⁸ After crossing Acer2^+/− mice, 10- to 12-week-old Acer^−/− (n=19) and their heterozygous (n=20) and wild-type (n=20) littermates were treated with a control or lithium diet. Each control or lithium treatment group consisted of five female and five male Acer2^+/+, Acer2^+/−, or Acer2^−/− mice, except for the Acer2^−/− control group, which consisted of four female and five male mice. To determine whether male mice reacted differently to lithium compared with female mice, a regression analysis was performed but no sex effects were found. After 8 days of treatment, mice were housed in metabolic cages and urine and blood were collected as described above. At day 10, mice were euthanized by cervical dislocation and the kidneys were rapidly removed. One kidney was sliced in two; one half was processed for immunohistochemistry (IHC), while the other half was stored at −80°C together with cortex and medulla material obtained from the second kidney.

Urine and Blood Analyses

Collected urine was centrifuged at 1000 × g for 5 minutes to remove sediment. Urine osmolality was measured using a Micro-Osmometer Model 3320 (Advanced Instruments, Norwood, MA) and pH was measured using an Oakton pH Spear (Eutech Instruments). Blood obtained in the strain survey was immediately analyzed for sodium ions (Na⁺), ionized blood calcium (iCa²⁺), hematocrit, hemoglobin, pH, base excess, partial pressure of carbon dioxide (pCO₂), partial pressure of oxygen, bicarbonate (HCO₃⁻), total carbon dioxide, and oxygen saturation (sO₂) using EG7+ cartridges and the i-STAT Clinical Analyzer (Abbott BV, Hoofddorp, The Netherlands). Serum lithium concentration was measured using The Medimate MiniLab (Enschede, The Netherlands).

IHC

Kidneys were immersion fixed in 3.5% paraformaldehyde in 0.1 M phosphate buffer for 3 days. Tissues were then dehydrated in ethanol, incubated in xylene, and embedded in paraffin. Paraffin sections (approximately 2 µm) were labeled with antibodies recognizing either AQP2²⁹ (dilution 1:8000) or H⁺-ATPase⁷ (dilution 1:1500). Labeling was visualized using peroxidase-conjugated goat anti-rabbit secondary antibody (Dako, Glostrup, Denmark) and 3,3′-diaminobenzidine. Cell counting was performed on the microscope in the inner medulla on sections labeled for H⁺-ATPase and was blinded to the group. The numbers of positive (labeled) and negative (unlabeled) cells were counted in tubules containing at least one positive cell and with a visible lumen.

Immunoblotting

Whole-kidney material (from half kidney) and medulla were homogenized in 1 ml and 300 µl of ice-cold homogenization buffer and diluted in Laemmli buffer, respectively, as described. SDS-PAGE, blotting, and blocking of the polyvinylidene difluoride membranes were done as described,³⁰ using affinity-purified rabbit precarboxy tail AQP2 antibody recognizing amino acids 236–255 (1:2000) and goat anti-rabbit IgGs coupled to horseradish peroxidase (Sigma, St. Louis, MO) as secondary antibodies.³¹ Densitrometric analysis was done using Bio-Rad quantification equipment (690c densitometer, Chemidoc XRS; Bio-Rad) and software (QuantityOne; Bio-Rad Laboratories GmbH, München, Germany). Equal loading of the samples was confirmed by staining blots with Coomassie blue.

PGE2 Assay and Analysis of Ceramide Levels

PGE2 metabolites (PGEMs) were measured in urine samples (day 9–10, stored at −80°C) using the PG E Metabolite EIA kit (Cayman Chemical, Ann Arbor, MI) according to the manufacturer’s instructions. Control samples were diluted by a factor of 300. The dilution rate of the lithium samples was between 40 and 1000, depending on the urine production of the different strains. The absorbance was measured on a SpectraMax 190 Microplate Reader (Molecular Devices, Sunnyvale, CA) at a wavelength of 405 nm. Renal ceramide levels were determined as described.³²

Analysis of mRNA Expression along the Mouse Nephron

Different nephron segments were dissected from collagenase-treated kidneys of six 8- 10-weeks-old C57BL/6J males (Charles River Breeding Laboratories), fed ad libitum with a standard diet (A04, SAFE; Epinay, France) as described.³³ Total RNA was extracted, using the RNeasy Micro Kit (Qiagen, Hilden, Germany) from pools of approximately 50 nephron segments, and reverse transcribed using a first strand cDNA synthesis kit for RT-PCR (Roche Diagnostics). Real-time PCR was performed using a cDNA quantity corresponding to 0.1 mm of nephron segments using the LightCycler 480 SYBR Green I Master qPCR kit (Roche Diagnostics). In each experiment, a standardization curve was made using serial dilutions of a standard cDNA stock solution made from whole-kidney RNA. The amount of PCR product was calculated as the percentage of standard DNA and gene expression was normalized as a function of that of the housekeeping gene Rpl26, as done previously. The use of Rpl26 as a housekeeping gene was validated using housekeeping gene Ppia, demonstrating highly similar relative expression levels for the different nephron segments (data not shown).

GWA Mapping Analysis

Genome-wide analysis was performed using the Efficient Mixed Model Association (EMMA; http://mouse.cs.ucla.edu/emma) method, which uses a linear mixed model algorithm to control for population structure and genetic relatedness.³⁴ Groups with a sample size of less than three were not included in the analysis because the reliability of such data were considered as not sufficient. We used urine volume and osmolality ratios from day 10 and 28 as phenotype input, which consisted of the absolute urine volume/osmolality values of the control versus lithium treatment group, and an NDI data set containing binary data comprising both phenotypes. Strains that developed NDI, defined by a more than twofold increased ratio of urine output and >2.5-fold decreased ratio of urine osmolality, were marked with “1;” whereas strains without NDI, defined by a less than twofold increased ratio of urine output and <2.5-fold decreased ratio of urine osmolality, were marked with “0.” These ratio thresholds were based on the definition of diabetes insipidus in patients, in which similar ratios are used for urine volume and osmolality as compared with normal levels.³⁵ All strains that did not fall into the above-mentioned criteria were not included in the analysis. For the exact number of mice per strain for each analysis, see Supplemental Table 1. The analysis was carried out using a panel of 4,016,612 single-nucleotide polymorphisms (SNPs), which were previously identified in four wild-derived and 11 classic strains³⁶ and subsequently imputed with high confidence for the less densely typed classic inbred strains.³⁷ Each SNP was evaluated individually and those without variation in the set of inbred strains were automatically filtered out by the EMMA server. P values were recorded as the strength of the genotype-phenotype association and were transformed using −log₁₀ (P value) in the scan plot. We only considered P values exceeding the threshold of 1×10⁻⁶, because this threshold resulted in few significant associations by comparing GWA analyses from different data sets with similar variation. Because of the lack of many GWA studies (GWAS) in mice, it remains difficult to implement a commonly accepted threshold, however this P value is also implemented in other studies.^38,39 Associations exceeding the threshold were excluded from further analysis if only one or two strains shared the associated haplotype. Loci were established by putting together all associations with the same haplotype within a distance of 1.0 Mb of each other, whereas SNPs with P values <1×10⁻⁵ were also included to determine the boundaries of a locus. Loci that only contained one or two SNPs were excluded from further analysis. Genome sequences within the candidate regions were compared between the different strains based on their haplotype distribution using the Sanger institute Mouse Genomes database (www.sanger.ac.uk/resources/mouse/genomes/), which also contains whole-genome sequences of ten strains that were included in the GWA analysis. Whole-genome sequences of these strains, also representing other strains with the same haplotype, were then assessed for potential deleterious effects of nonsynonymous variations on the structure or function of the protein using this website via the Variant Effect Predictor software from ENSEMBL.⁴⁰ A SIFT score of <0.05 was considered deleterious.⁴⁰

Statistical Analysis

Data are presented as mean with SEM. t tests (one-tailed test with unequal variances) were used to evaluate the differences between control and lithium-treated mice of 29 strains. Only strains with a sample size of three or more were included in the analysis. A one-way ANOVA with Bonferroni correction was applied for the mRNA expression data and the data on the Acer2 mutant mice. A threshold of P<0.05 was used to test for significance which, if required, was corrected using the Bonferroni multiple-comparisons procedure by dividing 0.05 by the number of included strains or genes.

Results

Li-NDI Develops Differently among Inbred Mouse Strains

To identify susceptibility genes for different lithiopathies, female mice from 29 different inbred strains (for abbreviations of the strains and number of mice per strain, see Supplemental Table 1) were treated with a control or lithium diet, starting at 10 weeks of age. Following lithium treatment, mice were housed in metabolic cages at days 8–10 and 26–28 to collect urine during the last 24 hours of each period, whereas blood was collected at three different time points (Figure 1A). The wild-derived strains CAST, PWD, and WSB demonstrated a lot of food spoilage in the urinary collection system of the metabolic cages. Therefore, our read out of urine volume and osmolality for the wild-derived strains are less reliable. At 10 days of lithium treatment, urine output was significantly increased in 17 strains (e.g., C3H, SM, BUB, and SWR), whereas urine output in other strains (e.g., BALB, LP, and A) was not affected by lithium (Figure 1B). Consistent with the NDI phenotype, most mice with an increased urine volume exhibited a decreased urine osmolality (Supplemental Figure 1A) and increased water intake (Supplemental Figure 2A). Prolonging lithium treatment of up to 28 days also resulted in large differences in urine output (Figure 1C), urine osmolality (Supplemental Figure 1B), and water intake (Supplemental Figure 2B) between strains. Serum lithium levels varied from 0.34 to 0.71 mM among the strains (Figure 1D) and did not correlate with the lithium effect on urine output or urine osmolality at day 10 or 28 (Supplemental Figure 3, A–D). Moreover, in most strains, the extent of Li-NDI on days 10 and 28 was similar (Supplemental Figure 4).

Figure 1.

The effect of lithium on urine production differs among mouse strains. (A) Mice from 29 different inbred strains were treated with a control (n=8) or lithium (n=10) diet for 10 and 28 days and were housed in metabolic cages from days 8 to 10 and 26 to 28. At (B) day 10 and (C) 28, 24-hour urine production was determined, and (D) at day 18 blood was collected to determine serum lithium levels. In a small number of strains, including WSB, NZO, and PWD, it was not possible to collect blood in the required number of mice for the statistical analysis. The order of strains on x axis (B) is based on lithium effect on urine production (starting with largest absolute increase); the order in (C and D) is based on (B). Control (Ctr), n=3–8; Lithium treated (Li⁺), n=3–10. *P<0.002, significantly different from control.

Lithium Does Not Affect Blood Hematocrit, Na⁺, or PGE2 Excretion in Most Strains

To investigate whether the extent of Li-NDI affected volume status or blood electrolyte concentrations, we determined blood hematocrit and Na⁺ levels after 10 and 28 days of lithium treatment. Lithium did not significantly alter hematocrit in any of the strains; but after treatment for 28 days, Na⁺ concentrations were increased in LP mice and decreased in C57L mice (Supplemental Table 2). Hematocrit and blood Na⁺ levels were unaltered in the vast majority of strains. As PGE2 has been described as one of the most important independent factors in Li-NDI development, we measured urinary PGEM levels from day 9 to 10. However, none of the strains demonstrated a significant difference in urinary PGEM content (Supplemental Figure 5).

The Effect of Lithium on the Acid-Base Balance

To investigate whether lithium affected the acid-base balance, blood pH, pCO₂, and HCO₃⁻, urinary pH was measured after 10 and 28 days of lithium treatment. With lithium treatment, blood pH was only significantly decreased in BUB and P mice after 10 and 28 days, respectively (Figure 2). The effects of lithium on blood pH per strain from day 10 correlated with day 28 (Supplemental Figure 6A), but did not correlate with the extent of Li-NDI (Supplemental Figure 6, B and C). Lithium did not affect blood pCO₂ in any strain at 10 or 28 days (Supplemental Figure 7) and only decreased blood HCO₃⁻ in the BUB strain after 10 days of lithium treatment (Supplemental Figure 8). After both 10 and 28 days, lithium increased urine pH in most strains, including the BUB strain (Supplemental Figure 9, A and B), suggesting a decreased ability to excrete protons/reabsorb HCO₃⁻. However, the pH of a solution is affected by dilution. Indeed, the lithium-induced increase in urine pH strongly correlated with NDI development (Supplemental Figure 9, A and B). Correcting urine pH for urine output by analyzing total free hydrogen ion (H⁺) excretion revealed there were no significant differences at day 10; whereas after 28 days of lithium treatment, free H⁺ excretion was increased in eight strains (Supplemental Figure 10).

Figure 2.

The effect of lithium on blood pH differs among mouse strains. Mice from 29 different inbred strains were treated with a control or lithium diet for 10 and 28 days. At (A) day 10 and (B) 28, blood was sampled by cheek puncture and pH was determined for 25 strains at both time points. In the remaining strains it was not possible to collect blood in the required number of mice for the statistical analysis. Order strains on x axis (A) is based on lithium effect on blood pH (starting with largest relative decrease); the order in (B) is based on (A). Control (Ctr), n=3–8; lithium treated (Li⁺), n=3–10. *P<0.002, significantly different from control.

Lithium Induces Hypercalcemia in SWR, A, C3H, and CBA Mice

To determine whether lithium induced hypercalcemia in one or more strains, iCa²⁺ was analyzed. After 10 days, lithium significantly increased blood iCa²⁺ levels in SWR, A, C3H, and CBA mice; whereas after 28 days no significant changes were observed (Figure 3). The effects of lithium on iCa²⁺ were not correlated with serum lithium levels nor with the development of Li-NDI (Supplemental Figure 11, A–D).

Figure 3.

The effect of lithium on blood Ca²⁺ differs among mouse strains. Mice from 29 different inbred strains were treated with a control or lithium diet for 10 and 28 days. At (A) day 10 and (B) 28, blood was sampled by cheek puncture and blood calcium ions (Ca²⁺) was determined, as depicted for 26 and 25 strains at day 10 and 28, respectively. In the remaining strains it was not possible to collect blood in the required number of mice for the statistical analysis. Order strains on x axis (A) is based on lithium effect on blood Ca²⁺ (starting with largest relative increase); the order in (B) is based on (A). Control (Ctr), n=3–8; lithium (Li⁺), n=3–10. (A) *P<0.002 (A) and (B) *P<0.002, significantly different from control.

GWAS

We thus identified many strains that developed Li-NDI, whereas only a few strains developed metabolic acidosis or hypercalcemia with lithium. As GWA requires variation in the parameter of interest between most of the strains in the study, GWA was only possible for Li-NDI. The wild-derived strains (CAST, WSB, PWD) were excluded from this analysis due to the increased food spoilage in the urinary collection system of the metabolic cages. To identify loci associated with Li-NDI, genome-wide analysis was performed for urine volume and osmolality ratios, consisting of the absolute urine volume/osmolality values of the control versus lithium treatment group, and an NDI data set containing binary data composed of both phenotypes (Table 1). Genome-wide scans from day 10 and 28 demonstrated various associations exceeding our threshold of 1×10⁻⁶ (Figure 4). Haplotypes of all significant associations were determined. Although the genome-wide scan of the urine volume analysis of day 28 revealed many loci exceeding the threshold, all were excluded from further analysis due to the presence of only one or two strains in a haplotype block. This led to the identification of eight gene-containing loci with up to 13 genes in the interval on chromosome 2 (Chr 2), Chr 4, Chr 6, Chr 8, and Chr 19 (Table 2). Of these genome-wide scans, QQ plots were generated and genomic inflation was calculated (Supplemental Figure 12).

Table 1.

Urine volume and osmolality ratios with included strains for binary analysis

Strains	Day 10 NDI			Day 28 NDI
	Urine Volume (Li/Ctr)	Urine Osmolality (Ctr/Li)	NDI (Binary Data)	Urine Volume (Li/Ctr)	Urine Osmolality (Ctr/Li)	NDI (Binary Data)
129	1.9	1.4	0	2.2	2.0	—
A	1.2	1.5	0	1.7	2.0	0
B10	6.1	3.6	1	4.0	2.5	1
B6	7.8	5.7	1	4.6	4.2	1
BALB	1.4	1.4	0	1.1	1.0	0
BLKS	2.4	2.5	—	1.6	1.7	0
BTBR	1.8	1.9	0	2.9	2.7	1
BUB	4.6	3.9	1	3.1	3.0	1
C3H	10.5	9.3	1	14.2	11.9	1
C57L	2.9	2.6	1	1.9	1.3	0
CBA	3.0	2.0	—	2.0	1.5	—
D2	1.3	1.6	0	1.3	1.5	0
FVB	10.4	8.6	1	6.7	4.8	1
KK	3.4	2.4	—	2.3	2.6	1
LP	1.2	1.5	0	2.4	1.9	—
MRL	2.2	2.3	—	2.6	2.3	—
NOD	3.0	2.8	1	2.3	2.4	—
NON	7.2	3.7	1	15.8	4.8	1
NZO	1.6	2.1	0	2.2	1.7	—
NZW	1.6	1.2	0	1.3	1.1	0
P	1.9	1.9	0	2.2	2.9	1
PL	2.7	2.4	—	2.2	1.8	—
RIIIS	2.5	2.3	—	0.8	1.7	0
SJL	3.6	3.1	1	6.9	6.6	1
SM	15.6	10.2	1	11.1	4.2	1
SWR	6.2	5.0	1	5.5	4.1	1

Ctr, control; Li, lithium treated.

Figure 4.

GWA mapping in lithium-treated mice reveals various significant associations. Mice were treated with lithium for 10 and 28 days and associations between their genotype and urine outcome (A and D, respectively), urine osmolality (B and E, respectively), or a combination of both data sets (C and F, respectively; see also Table 1) are displayed in Manhattan plots. Associations with a P value of <10⁻⁶ were considered significant. In (C), peaks at Chr 4, Chr 7, Chr 8, and Chr 19 with a P value <1×10⁻¹⁰⁰ were not shown to visualize peaks with higher P values. The arrow with number indicates the precise place and number of peaks that have been removed.

Table 2.

Genetic peak locations associating with development of Li-NDI

Duration Treatment	Chromosome	Locus (Mb)a	P Value	Data File	Genes in Intervalb
10 and 28 d	19	18.56–18.67	1.7×10−07	Binary	Ostf1
10 d	2	143.86–145.38	5.2×10−08	Binary	Banf2, Snx5, 8430406107Rik, Ovol2, Csrp2bp, Dzank1, Polr3f, Rbbp9, Sec23b, Gm561, Dtd1, 1700010M22Rik, Slc24a3
	4	75.64–76.97	1.0×10−266	Binary, urine osmolality	Ptprd
	4	85.03–86.57	4.2×10−07	Binary	AK044374, Adamtsl1, Fam154a, AK015482, Rraga, Haus6, Scarna8, Plin2, Dennd4c, Rps6, Acer2
	6	76.49–77.07	5.2×10−07	Binary	Ctnna2
	8	126.36–126.93	5.2×10−07	Binary	1700054N08Rik, Acta1, Nup133, AK013187, Abcb10, Taf51, mKIAA0133, Mir1967, Urb2, Galnt2, AK085459, Pgbd5
	19	44.78–45.07	4.4×10−08	Binary	Pax2, Fam178a, Sema4g
	19	48.57–48.61	2.3×10−239	Binary, urine osmolality	Sorcs3

^aNational Center for Biotechnology Information m37 assembly.

^bBased on University of California Santa Cruz Genome Browser (genome.ucsc.edu).

Expression of Candidate Genes in Renal Tubule

To determine which candidate genes are involved in the development of Li-NDI, we narrowed the candidate list by selecting for genes that are involved in the regulation of glycogen synthase kinase type 3, osmoregulation in the collecting duct, or cell proliferation via literature search in PubMed. This resulted in the identification of Ovol2, Rbbp9, Ptprd, Plin2, Acer2, Pax2, Urb2, and Galnt2.^41–45 Because Li-NDI is a disorder of the renal connecting tubule and collecting duct, we then determined their segment-specific expression using reverse transcription quantitative PCR on RNA isolated from different segments of adult C57BL/6J mice. Although all genes were expressed in connecting tubules and collecting duct, only Acer2 expression was enriched in the collecting duct because its levels were significantly higher in these tubules than in all other segments (Figure 5). Convincingly, this segment-specific distribution of Acer2 expression was similar to that found in rat.⁴⁶ The locus containing Acer2 harbored numerous variations among the strains. By analyzing whole-genome sequence data from ten strains (representing both haplotypes) via the Sanger institute Mouse Genomes database, we found that none of these variations resulted in an amino acid change predicted to be deleterious for protein structure or function. Thus, rather than a protein-inactivating mutation, it suggests the expression of Acer2 might be associated with the susceptibility to Li-NDI development.

Figure 5.

Expression of Ovol2, Rbbp9, Ptprd, Plin2, Acer2, Pax2, Urb2, and Galnt2 differs in the mouse kidney. mRNA was isolated from different nephron segments from C57BL/6J mice, and subjected to reverse transcription quantitative PCR to determine mRNA levels of Ovol2, Rbbp9, Ptprd, Plin2, Acer2, Pax2, Urb2, and Galnt2 along the nephron. The determined mRNA levels were normalized to that of housekeeping gene Rpl26. Values are mean±SEM from six mice. Indicated segments are proximal convoluted (S1) and straight (S3) tubule (PCT, PST), medullary (MTAL) and cortical (CTAL) thick ascending limb of the loop of Henle, distal convoluted tubule (DCT), connecting tubule (CNT), and the cortical (CCD) and outer medullary (OMCD) collecting duct. *P<0.006, significantly different from PCT, PST, MTAL, CTAL, and DCT.

Acer2 Expression Determines Susceptibility to the Development of Li-NDI

To investigate the role of Acer2 in the development of Li-NDI, we obtained an Acer2 mutant mouse strain²⁸ and treated Acer2^−/− animals and Acer2^+/− and Acer2^+/+ littermates for 10 days with a control or lithium diet. In the control groups, 24-hour urine volume was not different between the different genotypes (Figure 6A). With lithium treatment, however, urine volume was 1.8-times higher in Acer2^−/− mice as compared with their wild-type littermates, whereas urine volume of Acer2^+/− mice was in between that of Acer2^+/+ and Acer2^−/− mice. Urine osmolality of control diet mice was similar between the different genotypes, whereas urine osmolality of lithium-treated Acer2^−/− mice was significantly lower as compared with wild-type controls, and urine osmolality of Acer2^+/− mice was again between that of the Acer2^+/+ and Acer2^−/− mice (Figure 6B). Importantly, serum and urine lithium levels were not different among the different groups (Supplemental Figure 13, A and B). To determine whether the effect of Acer2 on urine output and osmolality was due to changes in AQP2 abundance, AQP2 localization, or principal cell–intercalated cell ratio, we performed AQP2 Western blotting and IHC. Although AQP2 abundance in the whole kidney did not significantly differ among lithium-treated Acer2^+/+ and Acer2^−/− mice (data not shown), we found that AQP2 abundance was reduced in the medulla of control and lithium-treated Acer2^−/− mice as compared with their wild-type control mice (Figure 6, C and D). IHC for AQP2 in the inner stripe of the outer medulla (Figure 6E) and the inner medulla (Supplemental Figure 13C) did not demonstrate any difference in AQP2 localization between the Acer2^+/+ and Acer2^−/− mice. In addition, there were no significant changes in the fraction of H⁺-ATPase–positive cells (marker of intercalated cells) in the inner medulla between the four groups (Acer2^+/+ control, 0.29±0.02 [n=5]; Acer2^−/− control, 0.28±0.02 [n=5]; Acer2^+/+ Li, 0.32±0.04 (n=5); Acer2^−/− Li, 0.21±0.04 [n=4]) (Supplemental Figure 14). To reveal the mechanism by which an altered expression of Acer2, an alkaline ceramidase, would affect the susceptibility for Li-NDI, we tested whether Acer2 mutant mice demonstrated altered ceramide levels in medulla tissue, however we did not find significant differences (Supplemental Figure 15).

Figure 6.

The development of Li-NDI is aggravated in Acer2 knockout mice. Acer2^+/+, Acer2^+/−, and Acer2^−/− mice were bred and, at an age of 10 weeks, treated with lithium for 10 days. The development of Li-NDI was studied by housing the mice during the last 48 hours in metabolic cages and analyzing (A) urine output and (B) urine osmolality of the urine from the last 24 hours. (C) Furthermore, kidneys were isolated and subjected to Western blotting to determine AQP2 abundance. Molecular masses of proteins are indicated on the left (in kDa). (D) The signals for nonglycosylated (29 kDa) and complex-glycosylated (40–45 kDa) AQP2, as indicated, were densitometrically semiquantified. Coomassie staining of the blots confirmed loading of protein equivalents. (E) Furthermore, IHC was performed to determine AQP2 abundance and localization. Control, Ctr; Hom, homozygous knockout mice; ISOM, inner stripe of the outer medulla; Li, lithium treated; mOsm, milliosmole; WT, wild type.

Discussion

The Genetic Background of a Mouse Strain Determines Its Susceptibility to Develop Li-NDI, Metabolic Acidosis, or Hypercalcemia

A substantial number of patients with bipolar disorder receiving lithium treatment develop side effects such as NDI, hypercalcemia, and possibly metabolic acidosis. In this study, we treated 29 different mouse inbred strains with a control or lithium diet and analyzed the development of these side effects after 10 and 28 days. Although lithium did not cause side effects in some strains, other strains developed side effects. Because treatment and housing conditions were identical and blood lithium concentrations were similar, the different susceptibility of these strains to develop side effects must be due to differences in their genetic background.

NDI was found in many strains after 10 or 28 days of lithium treatment, although there were also strains in which lithium did not affect urine concentration. In most strains, Li-NDI development after 10 days did not differ much from 28 days. During treatment all mice had access to a salt lick. We observed that strains with severe Li-NDI used more of the salt lick than strains without Li-NDI. Increased salt use as a consequence of volume loss due to Li-NDI development is likely the underlying reason. Furthermore, as we did not measure the actual salt intake, we cannot exclude the possibility that differences in salt craving between strains might have had a confounding effect on Li-NDI development, as very low salt intake attenuates the development of Li-NDI.⁴⁷

A small but significant decrease in blood pH was identified in BUB and P mice after 10 and 28 days, respectively, of lithium treatment. The decreased blood pH in BUB mice coincided with a significant decrease in blood bicarbonate, whereas blood pCO₂ was not changed, demonstrating that the small decrease in blood pH was indeed from metabolic origin. In contrast, in P mice blood bicarbonate at 28 days was not affected, whereas blood pCO₂ levels seemed to increase. This effect might result from the observed sensitivity of this strain to the repetitive blood draws (maximum 150 µl per draw), because these mice needed more time to recover than all other strains. Moreover, the mice on lithium treatment seemed more affected than the control mice. The relatively mild effects on blood pH might be due to the rather low serum lithium levels (0.34–0.71 mM) in our study, although most other studies with blood lithium levels around 1 mM also do not find a lithium-induced metabolic acidosis.^15–20 Finally, we found that lithium treatment for 28 days increased the total free urinary H⁺ levels in eight strains. Because we did not determine urinary net acid excretion, we cannot determine whether the elevated H⁺ levels coincided with a net overall increase in acid production. This is likely because our recent study using C57BL/6J mice treated with lithium in identical conditions revealed increased urine ammonium levels,⁴⁸ a phenomenon also found in recent studies with rats.^18,49 Altogether, it is likely that lithium impairs acid excretion, followed by a compensatory increase in production of titratable acids and acid excretion.

Finally, elevated blood calcium ion levels were found in SWR, A, C3H, and CBA mice after 10, but not after 28, days of lithium treatment. It must be noted that the development of hypercalcemia is likely dependent on blood lithium levels, which were rather low in our study. Altogether we identified different mouse strains with an increased susceptibility to develop lithium-induced hypercalcemia, which will facilitate follow-up studies on this topic.

Identification of Acer2 as a First Susceptibility Gene for Li-NDI Using Haplotype Association Mapping and Acer2 Knockout Mice

Genome-wide analysis on data of Li-NDI resulted in the identification of eight gene-containing loci which were significantly associated with the susceptibility of strains to develop NDI. Having selected Acer2 as a most promising candidate to be involved in Li-NDI, we investigated its potential role in Li-NDI. At similar blood lithium levels, Acer2 knockout mice were more susceptible to develop Li-NDI than their wild-type littermates because the homozygous knockout mice demonstrated a significantly higher urine output and lower urine osmolality after lithium treatment. These effects coincided with a further reduction of AQP2 abundance in both control as lithium-treated conditions. The fact that AQP2 expression was already lowered in the control situation but did not have strong effects on urine output or osmolality in these conditions suggests that Acer2 might affect water balance, which can be compensated for in the control situation, but becomes problematic during lithium conditions.

ACER2 belongs to the family of ceramidases which remove fatty acids from the lipid molecule ceramide, thereby producing sphingosine, which can be phosphorylated to form sphingosine-1-phosphate (S1P).⁵⁰ Ceramide, sphingosine, and S1P have important signaling functions which affect cellular processes like proliferation, apoptosis, and differentiation.^50,51 In this study, we did not find altered ceramide levels in medulla tissue of Acer2 knockout mice. The absence of any difference in ceramide levels might be due to other ceramidases regulating renal ceramide levels. Despite the specific expression of Acer2 in the collecting duct, and the absence or noncollecting duct–specific expression of Acer1 and Acer3, respectively (data not shown), we cannot be sure that collecting duct ACER2 regulates Li-NDI. Rather, a recent publication demonstrated that mice with Acer2 knockout in whole-body or specifically in hematopoietic cells exhibit decreased S1P blood levels.²⁸ Importantly, S1P also increases sodium excretion,^52,53 likely by inhibition of ENaC.⁵² The reduced S1P blood levels in the Acer2 knockout mice might lead to ENaC activation which, during lithium treatment, might lead to an enhanced uptake of lithium and thus an increased susceptibility to Li-NDI. Follow-up studies should investigate whether S1P indeed plays an essential role in the development of Li-NDI and whether its application might attenuate Li-NDI.

From GWAS in Mice to Determining Susceptibility for Lithium-Induced Side Effects in Patients

In this study, we identified that Acer2 plays a role in Li-NDI, however many other genes remain to be investigated. This might lead to the identification of novel pathways in the development of Li-NDI. Having established a more complete set of pathways involved in Li-NDI, the next step would be to perform GWAS in the human population and find the variations that significantly correlate with the development of lithium-induced side effects. Using the insight of important proteins and pathways in Li-NDI, as established by previous, current, and future animal studies, this would enhance the chances of successfully performing an expensive GWAS on patients treated with lithium.

Disclosures

None.

Funding

This project received support from a grant from the Society of Experimental Laboratory Medicine to Dr. Deen and a Marie Curie fellowship PIOF-GA-2012-332395 and a Niels Stensen Fellowship to Dr. de Groot.

Supplemental Material

This article contains the following supplemental material online at http://jasn.asnjournals.org/lookup/suppl/doi:10.1681/ASN.2018050549/-/DCSupplemental.

Supplemental Figure 1. The effect of lithium on urine osmolality in 29 mouse strains.

Supplemental Figure 2. The effect of lithium on water intake in 29 mouse strains.

Supplemental Figure 3. Relationship between serum lithium levels and development of Li-NDI.

Supplemental Figure 4. Relationship Li-NDI development between day 10 and 28.

Supplemental Figure 5. The effect of a 10-day lithium treatment on urinary PGEM levels in various mouse strains.

Supplemental Figure 6. Relationship between blood pH and Li-NDI development.

Supplemental Figure 7. The effect of lithium on blood pCO₂ in various mouse strains.

Supplemental Figure 8. The effect of lithium on blood HCO₃⁻ in various mouse strains.

Supplemental Figure 9. The effect of lithium on urine pH in 29 mouse strains.

Supplemental Figure 10. The effect of lithium on free urinary H⁺ content in 29 mouse strains.

Supplemental Figure 11. Relationship between effect of lithium on blood Ca²⁺ and development of Li-NDI.

Supplemental Figure 12. Genomic inflation.

Supplemental Figure 13. Serum and urine lithium content and inner medullary AQP2 localization in ACER2 knockout mice.

Supplemental Figure 14. The density of intercalated cells is not different for wildtype and Acer2 knockout mice.

Supplemental Figure 15. Ceramide levels in medullary tissue of male and female Acer2 Ko mice.

Supplemental Table 1. Names of the 29 inbred mouse strains, their abbreviations and number of mice used per analysis.

Supplemental Table 2. Blood Na⁺ and Hct levels in 29 different mouse strains treated for 10 and 28 days with lithium.