Uromodulin regulates renal magnesium homeostasis through the ion channel transient receptor potential melastatin 6 (TRPM6)

Up to 15% of the population have mild to moderate chronic hypomagnesemia, which is associated with type 2 diabetes mellitus, hypertension, metabolic syndrome, and chronic kidney disease. The kidney is the key organ for magnesium homeostasis, but our understanding of renal magnesium regulation is very limited. Uromodulin (UMOD) is the most abundant urinary protein in humans, and here we report that UMOD has a role in renal magnesium homeostasis. Umod-knockout (Umod−/−) mice excreted more urinary magnesium than WT mice and displayed up-regulation of genes promoting magnesium absorption. The majority of magnesium is absorbed in the thick ascending limb. However, both mouse strains responded similarly to the diuretic agent furosemide, indicating appropriate function of the thick ascending limb in the Umod−/− mice. Magnesium absorption is fine-tuned in the distal convoluted tubule (DCT) via the apical magnesium channel transient receptor potential melastatin 6 (TRPM6). We observed decreased apical Trpm6 staining in the DCT of Umod−/− mice. Applying biotinylation assays and whole-cell patch-clamp recordings, we found that UMOD enhances TRPM6 cell-surface abundance and current density from the extracellular space. UMOD physically interacted with TRPM6 and thereby impaired dynamin-dependent TRPM6 endocytosis. WT mice fed a low-magnesium diet had an increased urinary UMOD secretion compared with the same mice on a regular diet. Our results suggest that increased urinary UMOD secretion in low-magnesium states reduces TRPM6 endocytosis and thereby up-regulates TRPM6 cell-surface abundance to defend against further urinary magnesium losses.

The regulation of renal Mg 2ϩ homeostasis and TRPM6 remains largely unknown. The majority of TRPM6 channel regulation occurs either by (i) receptor-dependent pathways via insulin, epidermal growth factor (EGF), and receptor for activated protein C kinase 1 (RACK1) altering TRPM6 phosphorylation or (ii) increased channel mRNA expression by estrogen (32)(33)(34)(35)(36)(37). Recently, the ADP-ribosylation factor-like GTPase 15 (ARL15) was also shown to influence Mg 2ϩ absorption via TRPM6 and to modify glucose and lipid metabolism (38). A new mechanism for renal Mg 2ϩ regulation was implied when mice fed a low-Mg 2ϩ diet displayed increased renal Uromodu-lin (Umod) gene expression (39). This suggested a role for UMOD as part of a physiological response to hypomagnesemia (39). In fact, two different mutant Umod mouse models display urinary Mg 2ϩ wasting (40,41). UMOD is the most abundant protein in human urine; it protects against urinary tract infections and renal calcium crystal formation and is a promoter for hypertension (42)(43)(44). In this study, we wanted to analyze the role for UMOD in Mg 2ϩ regulation and investigate the mechanism by which UMOD enhances TRPM6 cell-surface abundance from the extracellular space.

Umod ؊/؊ mice display urinary Mg 2؉ wasting and decreased apical Trpm6 abundance in the DCT
Because up-regulation of UMOD mRNA in hypomagnesic mice suggested a role for UMOD in Mg 2ϩ regulation, we studied whether Umod Ϫ/Ϫ mice excreted more urinary Mg 2ϩ than WT animals. Umod Ϫ/Ϫ mice had almost twice the urinary Mg 2ϩ compared with WT mice when fed a regular diet (Fig.  1A). Whereas there was no significant difference in serum Mg 2ϩ between WT and Umod Ϫ/Ϫ mice at baseline (Fig. 1B), Umod Ϫ/Ϫ mice displayed up-regulated mRNA expression of Trpm6, Hnf1b (hepatocyte nuclear factor 1b), Egf, Fxyd2b (Na ϩ -K ϩ -ATPase regulatory protein), and parvalbumin (PV), all of which promote renal Mg 2ϩ absorption (Fig. 1C), which is consistent with a Mg 2ϩ -deficient state. It has been proposed that Umod Ϫ/Ϫ mice have a dysfunctional TAL, as they have a blunted response to furosemide (45,46). As the majority of Mg 2ϩ is absorbed in the TAL, we tested the TAL of WT and Umod Ϫ/Ϫ mice for the response to furosemide. At baseline, we detected in 9-month-old Umod Ϫ/Ϫ mice significantly higher urinary calcium (Ca 2ϩ ) and Mg 2ϩ excretion (Table 1). However, we found no significant difference between WT and Umod Ϫ/Ϫ mice in response to furosemide administration regarding urine volume, sodium (Na ϩ ), potassium (K ϩ ), Ca 2ϩ , or Mg 2ϩ excretion (Table 1). We also detected appropriate response of the TAL to furosemide in 5-month-old Umod Ϫ/Ϫ mice (Table S1). As the TAL in Umod Ϫ/Ϫ mice responded well to furosemide, we wondered about other mechanisms for how UMOD could affect renal Mg 2ϩ regulation. Therefore, we tested whether a lack of apical Trpm6 localization in the DCT contributes to renal Mg 2ϩ wasting in Umod Ϫ/Ϫ mice. In WT and Umod Ϫ/Ϫ mice, immunofluorescent staining for the Na ϩ -Cl Ϫ cotransporter (Ncc) and calbindin-D28K confirmed correct identification of the DCT (b and c in Fig. 1D). Ncc and calbindin-D28K expression appear similar in WT and Umod Ϫ/Ϫ animals (b and c versus e and f in Fig. 1D). However, evaluation of Trpm6 staining in the Ncc-and calbindin-D28K-positive tubules revealed less apical Trpm6 abundance in Umod Ϫ/Ϫ compared with WT mice (a versus d in Fig. 1D), despite an increase in Trpm6 transcript in Umod Ϫ/Ϫ animals (Fig. 1C). Our data point to the DCT as the responsible nephron segment for urinary Mg 2ϩ wasting in Umod Ϫ/Ϫ mice and suggest a post-transcriptional mechanism for UMOD regulation of TRPM6.

UMOD increases TRPM6 current density and TRPM6 cell-surface abundance
First, we determined in HEK293 cells whether UMOD increases TRPM6 mRNA or protein expression. Relative TRPM6 mRNA and protein expression was not enhanced in cells co-transfected with TRPM6 and UMOD compared with control (Fig. 2,  A and B). UMOD did not increase total TRPM6 protein abundance at 24, 48, or 72 h (Fig. S1). To study whether UMOD has an effect on TRPM6 channel current density, we co-transfected HEK293 cells with TRPM6 and control or UMOD. Co-expression of TRPM6 and UMOD increased TRPM6 current density significantly by a factor of 4.6 compared with control ( Fig. 2, C  and D). For the subsequent experiments, UMOD enhanced TRPM6 current density in the range of 1.9 -4.7-fold compared with control. This variability is most likely due to different transfection efficiency with each experiment. To test whether UMOD enhances TRPM6 cell-surface abundance, we performed biotinylation assays. TRPM6 cell-surface abundance was about 2.7-fold higher in cells co-transfected with UMOD compared with control plasmid (Fig. 2E). The increase of TRPM6 cell-surface abundance with UMOD was reasonably close to the observed increases in TRPM6 current density recorded for this work. Our data indicate that UMOD increases TRPM6 current density by enhancing the number of channels in the plasma membrane.

Extracellular UMOD increases TRPM6 current density in a dose-dependent fashion
The majority of UMOD is synthesized in the TAL, whereas TRPM6 is mostly localized in the more distal segment of the DCT (47)(48)(49)(50). Proteolytic cleavage between the UMOD zona pellucida domain and the glycosylphosphatidylinositol (GPI) anchor releases UMOD into the ultrafiltrate (51). Therefore, we tested the hypothesis that UMOD enhances TRPM6 from the extracellular space. HEK293 cells were transfected either with control, WT UMOD, or C150S mutant UMOD (a known human mutation that results in UMOD retention in the endoplasmic reticulum and decreased UMOD secretion as previously shown by us) (52,53). UMOD-and control-containing supernatant was harvested after 48 h and transferred onto TRPM6-expressing cells as described previously (52). After incubation overnight, WT UMOD-containing supernatant increased TRPM6 current density significantly compared with control-or C150S UMOD-containing supernatant (Fig. 3A). We confirmed an extracellular mechanism for up-regulation of TRPM6 current density by UMOD by performing a dose-response curve applying increasing dosages of purified UMOD. The half-maximal concentration for UMOD-mediated TRPM6 up-regulation was ϳ1 g/ml (Fig. 3B). This value is estimated to be 10-fold lower than estimates of human urinary UMOD concentration and therefore is within physiological concentration ranges for UMOD (54,55). Moreover, the addition of anti-UMOD antibody to TRPM6-transfected HEK293 cells incubated with UMOD-containing supernatant abrogated UMOD-mediated TRPM6 up-regulation (Fig. 3C). Collectively, these results strongly support a model by which UMOD

UMOD enhances renal magnesium absorption via TRPM6
enhances TRPM6 current density through an extracellular, apical interaction within tubules.

UMOD up-regulation requires almost all UMOD domains except the GPI anchor
UMOD secondary structure contains an N-terminal leader peptide that is crucial for secretion of UMOD, followed by four EGF-like domains that are important for protein-protein interaction, a D8C cysteine-rich sequence of unknown significance, a zona pellucida (ZP) domain that is crucial for UMOD multimerization, and finally a GPI anchor domain (Fig. 4A). To determine which UMOD domain is required for TRPM6 upregulation, we created different UMOD constructs and tested their effect on TRPM6 current density. Protein expression of different plasmids was assayed by Western blotting and titrated so that there was an equivalent amount of UMOD proteins in each condition. Full-length WT UMOD up-regulated TRPM6, whereas C150S mutant UMOD did not change TRPM6 current density compared with control ( Fig. 4B). UMOD plasmids containing the EGF-like domains (L180X) or the EGF-like domains plus the D8C domains (C297X) had no effect on TRPM6 current density. However, the UMOD construct containing the EGF-like, D8C, and ZP domains (R586X) increased TRPM6 current density significantly (Fig. 4B). The ZP domain is essential and sufficient for formation of UMOD polymers (56). Our results show that the EGF-like, D8C, and ZP domains are required for TRPM6 up-regulation and suggest that UMOD multimerization may be involved in TRPM6 up-regulation (Fig. 4B).

UMOD enhances TRPM6 current density by impairing TRPM6 endocytosis
Because our data were in line with an extracellular role of UMOD in TRPM6 up-regulation, we tested the hypothesis that UMOD may influence TRPM6 at the cell surface by interfering with TRPM6 endocytosis by studying the role of dynamin-2, a crucial protein for protein retrieval from the cell membrane. Expression of dynamin-2 in the DCT was described (57). When cells were co-transfected with WT dynamin-2, TRPM6 current density was up-regulated by UMOD (Fig. 6A). Cells co-transfected with dominant-negative (DN) dynamin-2, which impairs constitutive endocytosis, displayed enhanced TRPM6 current density even without UMOD (Fig. 6A). Stimulation of DN dynamin-2-transfected cells with UMOD did not result in any further increase of TRPM6 current density, indicating that  Table 1 with n ϭ 14 for each group). B, despite higher urinary Mg 2ϩ excretion in Umod Ϫ/Ϫ mice they have no lower serum Mg 2ϩ (WT 1.72 Ϯ 0.08 mg/dl versus Umod Ϫ/Ϫ 1.67 Ϯ 0.08 mg/dl, not significant (n.s.), n ϭ 6 for each group). The results are representative of two similar experiments. C, compared with WT mice, in Umod Ϫ/Ϫ mice, mRNA expression of genes involved in Mg 2ϩ regulation, such as Trpm6, Hnf1b, Egf, Fxyd2b, and PV, were up-regulated, consistent with a Mg 2ϩ -deficient state (n ϭ 4 for each group). D, Trpm6, Ncc, and calbindin immunostaining in the mouse renal cortex. a-c, tubular apical staining of Trpm6 (green), Ncc (red; a marker of DCTs), and cytosolic calbindin-D28K (blue) expression in WT mice. Compared with WT mice, Umod Ϫ/Ϫ mice (d-f) had less apical Trpm6 abundance, whereas Ncc and calbindin-D28K expression remain unchanged and confirm the correct location of the DCT (e and f). Scale bar, 20 m. Table 1 Urine and serum chemistries of 9-month-old WT and Umod ؊/؊ mice at baseline and after furosemide treatment WT and Umod Ϫ/Ϫ mice were kept in metabolic cages. Urine was obtained by bladder massage. Data showing the change after furosemide challenge for WT and Umod Ϫ/Ϫ mice (in columns five and six) are shown as percentage from baseline obtained before furosemide challenge and 4 h after furosemide treatment. Data are normalized per hour. WT and Umod Ϫ/Ϫ mice did not respond significantly differently to furosemide challenge regarding urine output, urinary Na ϩ excretion, K ϩ excretion, Ca 2ϩ excretion, or Mg 2ϩ excretion. Data show mean Ϯ S.D., n ϭ 14 for each group. * in Umod Ϫ/Ϫ column, p Ͻ 0.05; # in Umod Ϫ/Ϫ column, p Ͻ 0.01 compared with WT column. § in WT ϩ furosemide column, p Ͻ 0.005 compared with WT column. $ in WT ϩ furosemide column, p Ͻ 0.0001 compared with WT column. $ in Umod Ϫ/Ϫ ϩ furosemide column, p Ͻ 0.0001 compared with Umod Ϫ/Ϫ column. § in Umod Ϫ/Ϫ ϩ furosemide column, p Ͻ 0.005 compared with Umod Ϫ/Ϫ column. Hct, hematocrit. ND, not determined.

UMOD enhances renal magnesium absorption via TRPM6
UMOD increases TRPM6 by interfering with dynamin-2-dependent TRPM6 endocytosis. This conclusion is further supported by experiments using dynasore, an inhibitor of dynamin GTPase activity and blocker of dynamin-dependent endocytosis (58) (Fig. 6B). Previously, we have shown that the urinary protein Mucin-1 (MUC1) requires additional urinary proteins, such as galec-tin-3, a urinary lectin, for MUC1-related stimulation of the renal calcium channel TRPV5 (59). Galectins form a group of 15 members and bind ␤-galactosides (60,61). We tested the hypothesis that galectin-3 also is a mediator of TRPM6 up-regulation by UMOD. We used an siRNA knockdown approach that is efficient in reducing endogenous galectin-3 protein expression as previously described (59). However, galectin-3

UMOD enhances renal magnesium absorption via TRPM6
siRNA had no effect on TRPM6 current density up-regulation by UMOD, indicating that galectin-3 is not required for UMOD up-regulation of TRPM6 (Fig. 6C).
Another candidate for TRPM6 regulation by UMOD was galectin-1, which, like galectin-3, is a constituent of large crosslinked complexes that form lattice networks (62,63). Galectin-1 forms homodimers and binds to TRPV5 N-glycan. Klotho, an antiaging hormone that also functions as a sialidase and modifies a variety of ion channels, enhances TRPV5 current density in a galectin-1-dependent fashion (63). In previous work (59), we demonstrated efficient reduction of endogenous galectin-1 protein applying an siRNA approach. We found that knockdown of galectin-1 abolished the stimulatory effect of UMOD toward TRPM6, consistent with an important role in UMODmediated TRPM6 up-regulation (Fig. 6D).
Because galectin-1 binds to N-glycans and UMOD-mediated up-regulation of TRPV5 requires the TRPV5's N-glycan, we wondered whether TRPM6 is also N-glycosylated. Using computational tools, we identified 17 potential N-glycosylation sites within the TRPM6 protein (64). We focused on the TRPM6 Asn-787 N-glycan site because it is the only N-glycan located in an extracellular loop of TRPM6, thus allowing extracellular proteins to interact with this N-glycan. We analyzed the effect of UMOD on WT and N-glycan-deficient TRPM6. In contrast to its effect on WT TRPM6, UMOD did not enhance current density of N-glycan-deficient N787A TRPM6, consistent with a role for the presumed TRPM6 N-glycan in UMOD up-regulation (Fig. 6E).

Urinary UMOD secretion is increased in low-Mg 2؉ states
As UMOD modified TRPM6 current density, we wondered whether a low-Mg 2ϩ state itself could modify urinary UMOD secretion. In humans, urinary UMOD secretion is modified by dietary salt intake and antidiuretic and thyroid hormones (65)(66)(67)(68). We placed WT mice first on a regular Mg 2ϩ diet for 2 weeks and evaluated 24-h urinary UMOD secretion. Subsequently, the same animals were fed a low (50 ppm)-Mg 2ϩ diet for 2 weeks, and 24-h urinary UMOD secretion was measured. The 24-h urine sample was studied by semiquantitative Western blotting, and tested urine volume was normalized to urine creatinine. Our studies showed that animals fed a low-Mg 2ϩ diet had a significantly elevated, 2-fold stronger, urinary UMOD protein band compared with urine obtained on a regular diet (Fig. 7, A and B). These data are consistent with the systemic Mg 2ϩ state influencing the urinary UMOD secretion and point toward a feedback mechanism targeting systemic Mg 2ϩ homeostasis.

Discussion
The majority of filtered Mg 2ϩ is absorbed in the TAL (50 -70%) in a paracellular fashion due to the lumen-positive potential that is created by electroneutral absorption of Na ϩ , K ϩ , and 2 Cl Ϫ via NKCC2 and subsequent K ϩ secretion through ROMK back into the lumen (26). UMOD has been shown to up-regulate NKCC2 and ROMK and is therefore thought to contribute to Mg 2ϩ absorption in the TAL (44,45,69,70).
Our data are consistent with an additional role for UMOD in Mg 2ϩ absorption in the DCT. This is the first study to show that UMOD from the extracellular space can directly stimulate the cell-surface abundance of TRPM6 channels and thus enhances TRPM6 current density. In our model, UMOD is secreted from the TAL and up-regulates TRPM6 channels in the DCT from the urinary space by forming a carbohydrate lattice with a newly identified TRPM6 N-glycan (Fig. 8). Lattice formation requires galectin-1 and interferes with TRPM6 endocytosis, thereby increasing the TRPM6 cell-surface abundance and allowing for enhanced tubular Mg 2ϩ absorption to compensate for low-Mg 2ϩ states.
Published data on urinary Mg 2ϩ excretion in Umod Ϫ/Ϫ mice has been ambiguous. Mo et al. (43) showed no increase in urinary Mg 2ϩ excretion in 2-3-month-old Umod Ϫ/Ϫ mice, but Liu et al. (71) described increased urinary Mg 2ϩ excretion at 5-8 months of age. However, in a recent publication, the same group reports high urinary Mg 2ϩ excretion in 1-month-old Umod Ϫ/Ϫ mice and a lower degree of urinary Mg 2ϩ wasting in 12-month-old Umod Ϫ/Ϫ animals (46). It is possible that these results are conflicting because urinary Mg 2ϩ concentration was assessed rather than a measurement of urinary Mg 2ϩ excretion (e.g. correction for urinary volume and collection time). Our results, which were performed at 5 and 9 months of age, show higher urinary Mg 2ϩ and Ca 2ϩ excretion (e.g. corrected for volume and time) in the Umod Ϫ/Ϫ mice compared with control mice (Fig. 1A, Table 1, and Table S1).
UMOD has been shown to up-regulate NKCC2 and ROMK in the TAL (44,45,69,70). Mutig et al. (45) and Liu et al. (46) also demonstrated a blunted response in Umod Ϫ/Ϫ mice after furosemide treatment, supporting the hypothesis of a defective TAL in Umod Ϫ/Ϫ mice. Therefore, our finding that 9-monthold WT and Umod Ϫ/Ϫ mice responded in a similar fashion to furosemide was unexpected (Table S1 and Table 1). We considered that there may be a different response to furosemide dependent on the age of the Umod Ϫ/Ϫ mice, given the fact that most published furosemide challenges in Umod Ϫ/Ϫ mice were performed at 3 months (45,69,70). However, 5-and 9-month- Figure 2. UMOD enhances TRPM6 whole-cell current density and TRPM6 apical cell-surface abundance but not TRPM6 mRNA or protein. A, we compared TRPM6 mRNA expression in HEK293 cells when co-transfected with control or UMOD using quantitative RT-PCR. TRPM6 mRNA expression was not increased with UMOD co-transfection. Experiment was repeated in triplicate. B, Western blotting of TRPM6 protein expression at 48 h after co-transfection with control or UMOD. TRPM6 protein was not increased when co-transfected with UMOD. TRPM6 protein band density relative to ␤-actin is shown below. The experiment was repeated in triplicate. C, in HEK293 cells, co-expression of TRPM6-GFP with WT UMOD for 48 h increased the TRPM6 current density compared with control (196 Ϯ 22 pA/pF versus 42 Ϯ 13 pA/pF, p Ͻ 0.0001). TRPM6 current density (current normalized to cell-surface area, pA/pF; mean Ϯ S.D. (error bars)) was evoked by test pulses from Ϫ100 to 100 mV, with ϩ20-mV increments for 400 ms. The steady-state current-voltage (I-V) relation curve showed characteristic outwardly rectifying TRPM6 currents in cells co-transfected with TRPM6 with control versus UMOD (n ϭ 5 for each group). The experiment was performed in triplicate. D, scatter plot shows TRPM6 current density (pA/pF) at ϩ100 mV co-transfected with control or UMOD. E, UMOD effect on TRPM6 cell-surface abundance measured by biotinylation assay. TRPM6 abundance at plasma membrane (top) and total lysates (middle) were analyzed by immunoblot analysis using antibody against TRPM6. The antibody did not detect protein signal in the untransfected cells (data not shown). For detection of TRPM6 in the surface (biotinylated TRPM6, top row) and lysate (lysate TRPM6, second row), 250 and 20 g of protein were separated by SDS-gel electrophoresis, respectively. The experiment was repeated in duplicate. N/A, not applicable.

UMOD enhances renal magnesium absorption via TRPM6
old WT and Umod Ϫ/Ϫ animals showed a very similar response to furosemide, thus limiting the possibility of an age-dependent furosemide response in Umod Ϫ/Ϫ mice. A number of other experimental variables may account for the observed differences to furosemide in Umod Ϫ/Ϫ mice, such as genetic background, different Umod Ϫ/Ϫ mouse model, mouse strain, furosemide dose, urine collection method, and correction for urine volume or body weight. Consistent with our previous data on UMOD-mediated enhancement of TRPV5 cell-surface abundance, the 9-month-old Umod Ϫ/Ϫ mice had hypercalciuria, and there was a clear trend to higher urinary calcium excretion in the 5-month-old Umod Ϫ/Ϫ animals ( Table 1 and Table S1) (52). Including our other experiments, our results are consistent with an effect of UMOD in the DCT. Finally, a recent publication describes a smaller degree (10 -15% compared with TAL) of UMOD secretion in the DCT (72).
UMOD is the most abundant protein in urine. Our estimated EC 50 for UMOD-mediated up-regulation of TRPM6 current density is 10-fold higher than for TRPV5, but significantly lower than the estimated UMOD concentration in the DCT, suggesting near-maximally active TRPM6 channels in most conditions (52). We think that our results are more within the dynamic range of UMOD concentrations, as UMOD has to multimerize via the ZP domains (Figs. 4 and 8) to increase TRPM6 current density, and because we expect the effective concentration of macromolecular UMOD to be lower than that of UMOD monomers.
Interestingly, although serum Mg 2ϩ concentration in Umod Ϫ/Ϫ mice was not appreciably changed, several genes promoting Mg 2ϩ absorption were up-regulated (Fig. 1C). Measurement of serum Mg 2ϩ is known to be an insensitive marker of systemic Mg 2ϩ status, as only about 1% of total body Mg 2ϩ is in the extracellular fluid, with the remainder equally distributed in bone and soft tissue (77)(78)(79)(80). Our results are similar to those of another study showing up-regulated colonic Trpm6 mRNA but no effect on serum Mg 2ϩ concentration in mice fed omeprazole, a proton pump inhibitor that causes Mg 2ϩ deficiency, or a low-Mg 2ϩ diet (81,82). A recent publication (38) suggested that urinary Mg 2ϩ excretion in humans may be a better marker of disturbed Mg 2ϩ homeostasis. One may wonder how the Umod Ϫ/Ϫ mice maintain a normal serum  5 for each group). B, increasing concentrations of purified UMOD (final concentration in the medium is shown) enhanced TRPM6 current density in a dose-dependent fashion. TRPM6-transfected cells were treated with purified UMOD overnight. C, effect of anti-UMOD antibody (anti-UMOD Ab) on TRPM6 up-regulation by purified UMOD. Anti-UMOD antibody was added to culture medium of cells transfected with TRPM6 and treated with control or purified UMOD (n ϭ 5 for each group). All experiments in Fig. 3 were performed in triplicate. Error bars, S.D.

UMOD enhances renal magnesium absorption via TRPM6
Mg 2ϩ concentration despite increased urinary Mg 2ϩ excretion. Probably, there are redundant renal and nonrenal mechanisms to ensure Mg 2ϩ homeostasis in Umod Ϫ/Ϫ mice, as Trpm6 is also localized in the cecum and colon (34,81). Therefore, frank hypomagnesemia may only occur with the challenge of a low-Mg 2ϩ diet. Mg 2ϩ depletion alone was reported to enhance  Figure 4. The EGF-like, D8C, and ZP domains of UMOD are required for up-regulation of TRPM6. A, UMOD contains a leader peptide, which is important for proper secretion of UMOD; four (I-IV) EGF-like domains, which are crucial for protein-protein interaction; a D8C domain, which is enriched with cysteine residues and of unknown significance; and the ZP domain, which is required for UMOD multimerization. For UMOD secreted into urine, all three protein domains could be important, as UMOD is cleaved at the C terminus between the ZP domain and the GPI anchor. We created deletion constructs with a premature stop after the EGF-like (L180X), the D8C domain (C297X), and the ZP domains (R586X). Amino acids are provided above the model for the protein domains. B, co-transfection of TRPM6 and WT UMOD confirmed the up-regulation of TRPM6 current density in contrast to control or UMOD C150S mutant. The plasmids containing only the EGF-like (UMOD L180X) or the EGF-like plus D8C domains (UMOD C297X) had no effect on TRPM6 current density. Only the plasmid containing all three major domains (the EGF-like, D8C, and ZP domains) but not the GPI anchor (UMOD R586X) was able to increase TRPM6 current density in a significant fashion. Protein expression of different plasmids was assayed by Western blotting and titrated so that there was an equivalent amount of UMOD proteins in each condition (n ϭ 5 for each group). The experiment was performed in duplicate.

UMOD enhances renal magnesium absorption via TRPM6
renal Trpm6, PV, and Hnf1b mRNA expression in WT mice (34,82). Increased mRNA gene expression of these genes points toward a compensatory mechanism to cope with a low-Mg 2ϩ state.
It is tempting to consider that some of the complications of chronic hypomagnesemia could be due to TRPM6 dysregulation caused by insufficient UMOD secretion in low-Mg 2ϩ states. Remarkably, similar to hypomagnesemia, either low serum or urinary UMOD concentrations have been associated with a higher risk of cardiovascular disease, type 2 diabetes mellitus, and nephrolithiasis (21,(83)(84)(85). Interestingly, the degree of UMOD secretion depends on specific SNPs in UMOD and the UMOD promoter (44,86,87). Individuals carrying SNPs resulting in a reduced baseline UMOD secretion may have an impaired ability to respond appropriately to low-Mg 2ϩ states with increased urinary UMOD secretion. This would expose carriers of such UMOD SNPs to chronic, mild to moderate hypomagnesemia and subsequent complications.

Materials and DNA constructs
UMOD purified from human urine was purchased from SunnyLab (Sittingbourne, UK). The rabbit polyclonal anti-GFP-peroxidase antibody was obtained from Invitrogen/ Molecular Probes, Inc. (Eugene, OR). Human TRPM6 was cloned into the bicistronic vector pCINeo/IRES-GFP (32). TRPM6 N787A was obtained from WT TRPM6 using the XL site-directed mutagenesis kit from Agilent (Santa Clara, CA). To test the predicted N787A TRPM6 N-glycan site, WT and N787A TRPM6 were subcloned into pcDNA DEST53, which contains a GFP tag (Thermo Fisher Scientific). Protein G immobilized on agarose, mouse monoclonal anti-HA anti-Input CoIP a n t i -I g G a n t i -I g G a n t i -H . Immunoprecipitated proteins were identified using Western blotting (WB) and specific antibodies as shown on the right. Cell lysate is shown at the left of each immunoprecipitation experiment. 400 g of lysate was used for immunoprecipitation, and 4 g of protein was loaded for input. TRPM6 is detected with precipitation of the UMOD-TRPM6 complex using anti-HA antibody (lane 4). B, to confirm UMOD-TRPM6 physical interaction, we transfected either GFP or TRPM6-GFP with UMOD. Anti-GFP antibody was used for immunoprecipitation. Immunoprecipitated proteins were identified using Western blotting and specific antibodies as shown on the right. 400 g of lysate was used for immunoprecipitation, and 8 g of protein was loaded for input. Cell lysate is shown at the left of each immunoprecipitation experiment. Whereas GFP (25 kDa) as a negative control did not precipitate with UMOD (lane 3), UMOD-HA was detected with precipitation of the UMOD-HA-TRPM6-GFP complex. Samples for UMOD and GFP input (B, left) were run on the same Western blot as the remaining samples but not adjacent to each other. Therefore, they are separated by white bars. Both experiments were performed in duplicate.

UMOD enhances renal magnesium absorption via TRPM6
body, and mouse monoclonal anti-␤-actin-peroxidase antibody were purchased from Sigma. Cappel antibody against human uromucoid (Tamm-Horsfall glycoprotein) from goat was purchased from MP Biochemicals (Solon, OH). Donkey antibody against goat IgG was obtained from Santa Cruz Biotechnology, Inc. (Dallas, TX).

UMOD enhances renal magnesium absorption via TRPM6
and WT or dominant-negative (K44A) rat dynamin II or control vectors, as indicated, in each experiment. In each experiment, the total amount of DNA for transfection was balanced by using empty vectors. For galectin-1 and galectin-3 siRNA (Dharmacon, Lafayette, CO) transfections, we utilized 40 and 20 nmol, respectively (59).

Quantitative RT-PCR studies
Total RNA was isolated from kidneys from WT and Umod Ϫ/Ϫ mice using miRNeasy Mini kits from Qiagen (Ger-mantown, MD). First-strand cDNA was synthesized by an iScript TM cDNA synthesis kit (Bio-Rad). Relative transcript expression was measured by quantitative real-time PCR using iTaq TM Universal SYBR Green Supermix (Bio-Rad). Samples were run on a CFX96 real-time PCR detection system (Bio-Rad). 18S RNA was used to normalize for expression of mRNA. Primers for quantitative RT-PCR were described previously (82). Data were analyzed using the Bio-Rad CXF software.  UMOD is secreted in the TAL. Urinary secretion of UMOD is enhanced in low-Mg 2ϩ states. UMOD forms multimers through the ZP domains and interacts through a predicted N-glycan with TRPM6. Galectin-1 either stabilizes UMOD multimerization or UMOD-TRPM6 interaction, thus promoting lattice formation. The latter impairs dynamin-2-dependent endocytosis of TRPM6, thereby increasing the TRPM6 cellsurface abundance. This allows the nephron to compensate for low-Mg 2ϩ states.

Whole-cell patch-clamp recording
In brief, whole-cell patch clamp recording, co-immunoprecipitation, and immunofluorescent studies were performed in a similar fashion as outlined previously (52,59). Approximately 48 h after transfection, cells were dissociated and placed in a chamber for ruptured whole-cell recordings as described previously. Transfected cells were identified for recording by their GFP fluorescence. TRPM6 bath solution contained 140 mM NaCl, 5 mM CsCl, 2 mM CaCl 2 , 1 mM MgCl 2 , 10 mM glucose, 10 mM HEPES (pH 7.4 with NaOH). The pipette TRPM6 solution contained 120 mM CsCl, 10 mM NaCl, 1 mM HEDTA, and 10 mM HEPES (pH 7.2 with CsOH). Whole-cell patch clamp pipettes were pulled from borosilicate glass (Dagan Corp., Minneapolis, MN) and had resistance between 1.5 and 3 megaohms. The cell membrane capacitance and series resistance were monitored and compensated (Ͼ75%) electronically using an Axopatch 200B amplifier (Axon Instruments, Foster City, CA). The voltage protocol consists of a 0-mV holding potential and successive voltage sets (400-ms duration) from Ϫ100 to ϩ100 mV in ϩ20 increments. Current densities were obtained by normalizing current amplitude (obtained at ϩ100 mV) to cell capacitance. Data acquisition was performed using ClampX version 9.2 software (Axon Instruments). Currents were lowpass filtered at 2 kHz using an 8-pole Bessel filter in the clamp amplifier, sampled every 0.1 ms (10 kHz) with the Digidata-1440 interface, and stored directly to a computer hard drive.

Surface biotinylation assay
Biotinylation studies were performed as outlined previously (52). Specifically, for biotinylation of cell-surface TRPM6, cells were washed with ice-cold PBS and incubated with 0.75 ml of PBS containing 0.75 mg/ml EZ-Link-NHS-SS-biotin (Thermo Scientific, Rockford, IL) for 1 h at 4°C. After being quenched with glycine (100 mM), cells were lysed in radioimmune precipitation buffer (150 mM NaCl, 50 mM Tris-HCl, 5 mM EDTA, 1% (v/v) Triton X-100, 0.5% deoxycholate, and 0.1% (w/v) SDS) containing protease inhibitor mixture (Roche Applied Science). Biotinylated proteins were precipitated by streptavidin-agarose beads (Thermo Scientific). Beads were subsequently washed four times with PBS containing 1% (v/v) Triton X-100. Biotinlabeled proteins were eluted in sample buffer, heated at 50°C for 5 min, separated by SDS-PAGE electrophoresis, and transferred to nitrocellulose membranes for Western blotting. TRPM6 proteins on the membrane were detected using a mouse monoclonal anti-TRPM6 (human) antibody (Santa Cruz Biotechnology). Biotinylation experiments were performed three times with similar results.

Co-immunoprecipitation and immunoblotting
HEK293 cells were co-transfected with HA-tagged UMOD and either TRPM6 or GFP-tagged TRPM6 using Lipofectamine 2000 and incubated for 48 h at 37°C. Cells were collected and lysed using a needle syringe. The protein concentration in the cell lysate was determined using a protein assay (DC Protein Assay, Bio-Rad). Samples were adjusted to the same concentration with buffer. Samples were resolved on a 4 -20% gradient gel and processed for immunoblotting using specific antibodies. For immunoprecipitations, we incubated 700 g of cell lysate with 3 g of anti-HA, anti-TRPM6, or anti-GFP antibodies overnight at 4°C. Species-specific anti-IgG (Santa Cruz Biotechnology) was used as control. Antigen-antibody complex was then loaded onto Protein G beads (GE Healthcare, Little Chalfont, UK) by slow rotation overnight, washed, and eluted with Laemmli sample buffer. Samples were denatured at 65°C for 10 min and subjected to SDS-PAGE and immunoblotting. Antibodies against human TRPM6 (Santa Cruz Biotechnology) and UMOD (MP Biochemicals, Solon, OH) were used for Western blotting.

Immunofluorescent staining
Male WT mice were sacrificed at the age of 3 months. Anesthetized mice were perfused with 4% (w/v) paraformaldehyde in PBS (pH 7.4). Kidneys were harvested and sectioned. Sections were blocked with 10% (v/v) donkey sera in PBS, and immunofluorescence was performed with guinea pig anti-Trpm6 antibody (1:50), mouse anti-calbindin-D28K (1:100), and rabbit anti-Ncc (1:200; Millipore) overnight at 4°C (34). Fluorescent images were obtained using a Zeiss LSM510 confocal microscope (Zeiss, Jena, Germany). All animal experiments were performed in compliance with relevant laws and institutional guidelines and were approved by the University of Texas Southwestern Medical Center at Dallas Institutional Animal Care and Use Committee.

Metabolic cage studies in mice
Urinary excretion of Mg 2ϩ was determined in 9-month-old WT and Umod Ϫ/Ϫ mice on a 129/SeEv background (43). Animals were acclimated for 3 days in metabolic cages before experiments. Daily urine volume was measured after bladder massage (89). Urine Mg 2ϩ was analyzed by the University of Texas Southwestern metabolic phenotyping core. For furosemide challenge and urinary UMOD secretion studies, see the supporting information.

Furosemide challenge and urinary UMOD secretion in mice
Baseline 24-h urine collections were obtained. Complete bladder emptying was ensured by bladder massage. Animals were injected with a single dose of 15 mg/kg furosemide intraperitoneally (90). Urine was collected after 4 h with bladder massage (89). Urine output, sodium, potassium, calcium, and magnesium were analyzed by the University of Texas Southwestern metabolic phenotyping core. Furosemide response was assessed individually for every mouse by determining the hourly excretion of urine volume and electrolytes before and after furosemide treatment. To calculate the furosemide response, urinary data (e.g. volume and electrolyte excretion) obtained before and 4 h after treatment were corrected for collection time. For every individual animal, data points obtained after furosemide treatment were divided by the data points before treatment and multiplied by 100. To study the urinary UMOD secretion in mice, WT mice were fed a low (50 ppm)-Mg 2ϩ diet (Teklad, TD.02073) for 2 weeks followed by 2 weeks of a regular diet. At the end of each dietary treatment, urine was collected as outlined above. Urine creatinine was determined by the University of Texas Southwestern O'Brien Kidney Center. Urine volume was normalized for urinary creatinine, and

UMOD enhances renal magnesium absorption via TRPM6
UMOD secretion was studied in a semiquantitative fashion by Western blotting using anti-Thp (G-20) antibody (1:200) from Santa Cruz Biotechnology.

Statistical analysis
Student's t tests were used to test whether there are significant differences in the continuous outcomes between two study groups. For multiple comparisons, one-way analysis of variance studies followed by the Student-Newman-Keuls method, allowing for pairwise multiple comparisons, were performed. Data are reported as means Ϯ S.D. p Ͻ 0.05 was considered statistically significant.