Role of adenylyl cyclase 6 in the development of lithium-induced nephrogenic diabetes insipidus

Despite lithium (Li⁺) usage being associated with renal, neurological, and endocrine side effects, the drug is still frequently used for the treatment of psychological disorders, e.g., bipolar disorder (1). The most common adverse effect of Li⁺ therapy, affecting approximately 50% of patients, is Li⁺-induced nephrogenic diabetes insipidus (Li-NDI), characterized by polydipsia, polyuria, and a limited renal response to the antidiuretic hormone, arginine vasopressin (AVP) (2). The polyuric effects of Li⁺ are associated with a marked decrease in the abundance of the water channel aquaporin-2 (AQP2) in collecting duct (CD) principal cells (PCs), accompanied by a large increase in the ratio of intercalated cells (ICs) to PCs along the CD system (3, 4). Furthermore, long-term (chronic) Li⁺ therapy increases the possibility of developing end-stage renal disease and is associated with an increased incidence of developing solid renal tumors (discussed in ref. 5). Despite these side effects, Li⁺ therapy is continued for the majority of patients with Li-NDI, as the patient’s quality of life is greatly affected by side effects, e.g., bipolar disorder symptoms. Thus, understanding the complete mechanism(s) of Li-NDI and developing ways to reverse the AVP resistance of the CD following Li⁺ treatment are the first steps to newer and safer therapies for psychological disorders.

A multitude of studies have been performed to identify the altered cellular signaling events that underlie Li-NDI (discussed in ref. 6). However, despite major progress, no clear consensus regarding the “trigger” for the onset of the condition exists. For example, data from large-scale proteomic (7) and metabonomic approaches (8), coupled with targeted studies in animal models and gene-modified mice, have illustrated changes in renal prostaglandins (9); purinergic signaling (10); and the activity of glycogen synthase kinase 3 (GSK3) (11), PKC (12), MAPKs, ERKs ERK1/2 and P38α (13), and the phosphatidylinositol signaling pathway and the Wingless-Int/β-catenin (Wnt/β-catenin) pathway (7) among others, during Li-NDI. However, due to its role as the major downstream signaling molecule of AVP and its capacity to modulate several of the altered signaling pathways described above, diminished levels/effects of cAMP are thought to be the primary cause of Li-NDI. This essential role of cAMP in Li-NDI is supported by the following: (a) Li⁺ can directly inhibit adenylyl cyclases (ACs) (14, 15); (b) the AVP-induced water permeability of isolated perfused rabbit cortical CDs is reduced in the presence of Li⁺ but overcome in the presence of exogenous cAMP (16); (c) in response to AVP or direct activation of AC by forskolin, cAMP accumulation in CDs isolated from Li⁺-treated rats is severely reduced (17); and (d) inhibition of the Li⁺ target GSK3β reduces AVP-induced AC activity and cAMP generation in CD cells (18).

AC6 is the most abundantly expressed AC isoform in the mouse inner medulla (IM), and we have previously demonstrated that mice lacking AC6 (AC6^–/–) have dramatically reduced IM cAMP levels, which are not significantly increased following forskolin or AVP stimulation (19). Furthermore, AC6^–/– mice have reduced AQP2 phosphorylation/trafficking and suffer from NDI, highlighting an essential role of AC6 for the AVP-cAMP axis and urinary concentration. Therefore, due to the proposed essential role of reduced cAMP levels in the onset of Li-NDI and evidence that Li⁺ preferentially inhibits AC6 (20), we hypothesized in this study that Li-NDI would not develop in total body AC6 knockout (AC6^–/–) mice. However, our results demonstrate that, in AC6^–/– mice and connecting tubule/PC-specific AC6 knockout (AC6^loxloxCre) mice, dietary Li⁺ administration has similar effects on renal water handling and AQP2 levels relative to control mice. Our studies demonstrate that in vivo AC6 and IM cAMP generation play a minor role in the development of Li-NDI but point toward AC6 playing a role for cellular proliferation.

Li⁺ reduces urine osmolality and increases water intake in AC6^–/– and WT mice. To test the contribution of AC6 to Li-NDI, AC6^–/– mice and WT littermates were initially fed a standard diet (baseline) and then shifted to a Li⁺-containing diet for 27 days. AC6^–/– and WT mice developed NDI, as evidenced by reduced urine osmolality and increased water intake (Figure 1, A and B, P < 0.05; for body weight and plasma [Li⁺], see Table 1). Baseline urine osmolality was lower in AC6^–/– mice compared with WT mice (Figure 1A, P < 0.05), a difference that remained for the duration of the Li⁺ administration. After the initiation of Li⁺ administration, AC6^–/– mice had significantly reduced urine osmolality on day 1, while a significant reduction in urine osmolality in WT mice was only observed from day 3 and onward (Figure 1A). Similarly, urine osmolality relative to baseline was significantly lower in AC6^–/– mice on days 1 and 2 (Figure 1C, P < 0.05), after which no significant differences were observed. In agreement with the urine osmolality data, baseline water intake was significantly higher in AC6^–/– mice, and this difference remained for the duration of the Li⁺ administration (Figure 1B, P < 0.05). Compared to baseline, Li⁺ induced a significantly higher water intake at an earlier time point in AC6^–/– mice than in WT mice (day 4 and day 6, respectively, Figure 1B, P < 0.05). For the majority of the time points examined, no major differences between the genotypes in the fractional water intake relative to baseline were observed (Figure 1D).

Figure 1

Li⁺ reduces urine osmolality and increases water intake in AC6^–/– and WT mice. Following an initial baseline period on standard diet, mice lacking adenylyl cyclase 6 (AC6^–/–) and WT mice were fed a Li⁺-containing diet for 27 days, during which physiological parameters were measured. (A) Urine osmolality (osm), (B) water intake, (C) urine osmolality relative to baseline conditions (percentage of baseline), and (D) water intake relative to baseline conditions (percentage of baseline). Baseline values are averages of the last 5–7 days before switching from standard diet to Li⁺ diet. Sample size for urine osmolality: AC6^–/–, n = 5–8; WT, n = 6–8. Sample size for water intake: AC6^–/–, n = 3–5; WT, n = 4. Values are mean ± SEM. All statistical comparisons were performed using 2-way repeated-measurements ANOVA followed by Holm-Sidak post-hoc tests. *P < 0.05, baseline versus Li⁺ administration; ^#P < 0.05, AC6^–/– versus WT at individual days.

Table 1

Body weight and plasma [Li⁺] in AC6^–/– mice versus WT mice and AC6^loxloxCre mice versus AC6^loxlox mice under baseline conditions and Li⁺ administration

AC6^loxloxCre mice have greater water intake and lower urine osmolality under baseline conditions. AC6 expression is heterogeneous in the kidney tubule (21), and we have previously demonstrated roles of AC6 in cells of the proximal tubule, thick ascending limb, and distal convoluted tubule (22, 23). Therefore, to isolate a potential role of AC6 in kidney and CD PCs, the main “target” cell for Li⁺ effects (24), we generated a conditional mouse model with AC6 deletion controlled by AQP2-dependent Cre recombinase expression (25). After several attempts with antibodies from various commercial sources, we identified an antibody that, in cortex homogenates from WT mice, detected AC6 as a smear centered at 150 kDa (26, 27). No signal was detectable in AC6^–/– mice, demonstrating for the first time to our knowledge specificity of an AC6 antibody in Western blots (Figure 2A). In AC6^loxloxCre mice under baseline conditions, AC6 protein was approximately 50% reduced in IM homogenate compared with AC6^loxlox mice (Figure 2B, P < 0.05), a finding consistent and indicative of efficient AC6 knockout in PCs. No significant differences were detected in AC6 levels in the cortex and outer medulla/cortex (OM/cortex) homogenates (Figure 2B), presumably due to AC6 expression remaining in non-AQP2-expressing cells, which compose the vast majority of the cortex/outer stripe OM. Although labeling of AC6, using IHC, was unsuccessful (our unpublished observations), clear nuclear IHC labeling of Cre recombinase in both cortical, OM and IM collecting ducts in AC6^loxloxCre mice suggested successful deletion of AC6 in AQP2-expressing cells throughout the CD system (Figure 2, C–F). This is in agreement with previous IHC characterization of the Aqp2-Cre transgenic mouse line, demonstrating Cre expression specifically in connecting tubule (CNT) cells and CD PCs (28). Initial physiological characterization of body weight–matched mice (AC6^loxloxCre: 31.1 ± 2.3 g, ref. 9; AC6^loxlox: 33.3 ± 1.2 g, ref. 17) revealed lower urine osmolality and higher water intake in AC6^loxloxCre mice compared with AC6^loxlox mice (Figure 2G, P < 0.05). No significant differences were detected in food intake and plasma osmolality between the groups (Figure 2G).

Figure 2

AC6^loxloxCre mice have greater water intake and lower urine osmolality under baseline conditions. A mouse model with deletion of adenylyl cyclase 6 (AC6) in aquaporin-2–expressing (AQP2-expressing) cells (AC6^loxloxCre) was generated as described in Methods. (A) AC6 was detected using immunoblotting as a broad smear in cortex homogenates from WT mice, whereas no signal was detectable in mice lacking AC6 (AC6^–/– mice), confirming antibody specificity. (B) Under baseline conditions, AC6 protein signal was approximately 50% less intense in inner medulla (IM) homogenates from AC6^loxloxCre mice compared with AC6^loxlox control mice, whereas no significant reductions could be observed in cortex or outer medulla (OM)/cortex homogenates. The samples were used for immunoblotting on Figure 5. (C–F) Cre recombinase IHC showed clear nuclear labeling in cortical and IM collecting ducts of AC6^loxloxCre mice. Scale bar: 50 μm. (G) Under baseline conditions, AC6^loxloxCre mice had lower urine osmolality and higher water intake than AC6^loxlox mice, while no significant differences were found in food intake or plasma osmolality. Values are mean ± SEM. Numbers in parentheses indicate sample sizes. Statistical comparisons were performed using Student’s 2-tailed t tests (B [cortex and OM/cortex homogenates] and G [water intake, food intake, and plasma osmolality]) and Satterthwaite’s 2-tailed unequal variance t test (B [IM homogenate] and G [urine osmolality]). *P < 0.05. Data on urine osmolality and water intake are equivalent to baseline data presented in Figure 3 and Supplemental Figure 1. Plasma osmolality data originate from a different cohort of mice.

Li⁺ reduces urine osmolality and increases water intake in AC6^loxloxCre and AC6^loxlox mice. To examine the contribution of AC6 to Li-NDI specifically in CNT cells and CD PCs, AC6^loxloxCre and AC6^loxlox mice were initially fed a standard diet (baseline) and then were either shifted to a Li⁺-containing diet (Figure 3) or maintained on standard diet (Supplemental Figure 1; supplemental material available online with this article; https://doi.org/10.1172/jci.insight.91042DS1) for a total of 15 days. Both genotypes developed NDI in response to Li⁺, as evidenced by reduced urine osmolality and increased water intake (Figure 3, A and B). Baseline urine osmolality was lower in AC6^loxloxCre mice compared with AC6^loxlox mice (Figure 3A, P < 0.05), and this difference between the groups remained throughout the period of the Li⁺ administration (Figure 3A). In response to Li⁺, both AC6^loxloxCre and AC6^loxlox mice reduced their urine osmolality, but the time frame for these reductions to be significant was different between the groups (day 6 and day 3, respectively, Figure 3A, P < 0.05). No significant differences were detected between groups in the fractional urine osmolality relative to baseline (Figure 3C), suggesting an equivalent response to Li⁺ administration. Baseline water intake tended to be higher in AC6^loxloxCre mice (Figure 3B, P = 0.112), and it remained significantly higher for the duration of the Li⁺ administration (days 1–15, Figure 3B, P < 0.05). Water intake was increased in both AC6^loxloxCre and AC6^loxlox mice (day 5 and 9, respectively, Figure 3B, P < 0.05), with fractional water intake relative to baseline remaining higher in AC6^loxloxCre mice (Figure 3D, P < 0.05). Other physiological parameters on day 15 (study endpoint) are shown in Table 1 and Figure 4.

Figure 3

Li⁺ reduces urine osmolality and increases water intake in AC6^loxloxCre and AC6^loxlox mice. Following an initial baseline period on standard diet, mice with deletion of adenylyl cyclase (AC6) in aquaporin-2–expressing (AQP2-expressing) cells (AC6^loxloxCre mice) and AC6^loxlox control mice were fed a Li⁺-containing diet for 15 days. AC6^loxloxCre and AC6^loxlox mice were simultaneously maintained on the standard diet (Supplemental Figure 1). Physiological parameters were measured daily. (A) Urine osmolality (osm), (B) water intake, (C) urine osmolality relative to baseline conditions (percentage of baseline), and (D) water intake relative to baseline conditions (percentage of baseline). Baseline values are averages of the last 1–2 days before switching from standard to Li⁺ diet. Water intake and urine osmolality data on day 15 are equivalent to data presented in Figure 4. Sample size for urine osmolality: AC6^loxloxCre, n = 4–5; AC6^loxlox, n = 9. Sample size for water intake: AC6^loxloxCre, n = 3; AC6^loxlox, n = 4. Values are mean ± SEM. All statistical comparisons were performed using 2-way repeated-measurements ANOVA followed by Holm-Sidak post-hoc tests. *P < 0.05, baseline versus Li⁺ administration; ^#P < 0.05 AC6^loxloxCre versus AC6^loxlox.

Figure 4

Li⁺ induces lower urine osmolality and higher water intake in AC6^loxloxCre and AC6^loxlox mice compared with standard diet. Following an initial baseline period on standard diet, mice with deletion of AC6 in AQP2-expressing cells (AC6^loxloxCre mice) and AC6^loxlox control mice were either fed a standard diet or a Li⁺-containing diet for 15 days (study endpoint). Presented are summary data for (A) body weight, (B) food intake, (C) water intake, (D) urine osmolality, (E) plasma osmolality, (F) hematocrit, and (G) plasma [Li⁺]. Data on water intake and urine osmolality are equivalent to data on day 15 presented in Figure 3 and Supplemental Figure 1. Values are mean ± SEM. Numbers in parentheses indicate sample sizes. All statistical comparisons were performed using 2-way ANOVA followed by Holm-Sidak post-hoc tests. *P < 0.05.

Li⁺ induces greater reductions in AQP2 and pS256-AQP2 abundances than genetic deletion of AC6. The molecular response to 15 days of Li⁺ administration in AC6^loxloxCre and AC6^loxlox mice was examined by immunoblotting kidney homogenates for various membrane proteins important for water and NaCl transport. In the IM, AC6 levels were significantly lower in AC6^loxloxCre mice (Figure 5, P < 0.05; see complete unedited blots in the supplemental material.). Li⁺ administration further reduced the IM AC6 levels in both genotypes, suggesting an effect of Li⁺ on AC6 abundance in AQP2-negative tissues/cells, such as ICs, thin descending or ascending limbs, blood vessels, or connective tissue. In agreement with our previous studies on AC6^–/– mice (19), AC6^loxloxCre mice had significantly lower IM abundances of AQP2, pS256-AQP2, and pS269-AQP2 compared with AC6^loxlox mice (Figure 5, P < 0.05). Dietary Li⁺ intake further reduced AQP2 and pS256-AQP2 levels in AC6^loxloxCre and AC6^loxlox mice (Figure 5, P < 0.05), whereas pS269-AQP2 abundance following Li⁺ administration was reduced only in AC6^loxlox mice to levels similar to AC6^loxloxCre mice under baseline conditions. The levels of pS9-GSK3β, an inhibitory site on GSK3β that may be involved in development of Li-NDI, were significantly higher in Li⁺-administered AC6^loxloxCre and AC6^loxlox mice (Figure 5, P < 0.05). In cortex homogenates, significant reductions in AQP2 and pS256-AQP2 abundances were detected following Li⁺ administration (Figure 6, P < 0.05), alongside decreased levels of cleaved αENaC and γENaC (Figure 6, P < 0.05). Similar changes in AQP2 profiles were observed in the OM/cortex (Figure 7, P < 0.05). Neither Li⁺ diet nor genotype had any effect on the abundances of NCC, NKCC2, and pT96/T101-NKCC2 (Figures 6 and 7, P < 0.05). Combined, these data indicate that dietary Li⁺ intake has a greater effect on AQP2 and pS256-AQP2 abundances and the levels of cleaved αENaC and γENaC than genetic deletion of AC6 alone.

Figure 5

Li⁺ reduces inner medullary AQP2 and pS256-AQP2 abundances in AC6^loxloxCre and AC6^loxlox mice. (A) Semiquantitative immunoblotting of inner medulla (IM) homogenates from mice with deletion of adenylyl cyclase 6 (AC6) in aquaporin-2–expressing (AQP2-expressing) cells (AC6^loxloxCre) and AC6^loxlox control mice using antibodies targeting various proteins modulated during Li⁺-induced nephrogenic diabetes insipidus. (B) Summary data demonstrate that AC6, AQP2, pS256-AQP2, and pS269-AQP2 levels were significantly lower in AC6^loxloxCre mice relative to AC6^loxlox mice under standard conditions. Dietary Li⁺ administration further reduced AQP2 and pS256-AQP2 levels in AC6^loxloxCre and AC6^loxlox mice, indicating AC6-independent effects. pS9-glycogen synthase kinase 3 β (GSK3β) and proliferating cell nuclear antigen (PCNA) levels were significantly increased following Li⁺ administration in both genotypes. The samples from AC6^loxloxCre and AC6^loxlox mice on standard diet were used for immunoblotting in Figure 2. Values are mean ± SEM. Sample sizes: AC6^loxlox standard = 5, AC6^loxlox Li⁺ = 6, AC6^loxloxCre standard = 4, AC6^loxloxCre Li⁺ = 5. All statistical comparisons were performed using 2-way ANOVA followed by Holm-Sidak post-hoc tests. *P < 0.05.

Figure 6

Li⁺ reduces cortical AQP2, pS256-AQP2, cleaved αENaC, and cleaved γENaC abundances in AC6^loxloxCre and AC6^loxlox mice. (A) Semiquantitative immunoblotting of cortex homogenates from mice with deletion of adenylyl cyclase 6 (AC6) in aquaporin-2–expressing (AQP2-expressing) cells (AC6^loxloxCre mice) and AC6^loxlox controls using antibodies targeting various proteins modulated during Li⁺-induced nephrogenic diabetes insipidus. (B) Summary data demonstrate lower abundances of AQP2, pS256-AQP2, cleaved α epithelial Na channel (αENaC), and cleaved γENaC in response to Li⁺ administration in both genotypes. The abundances of these proteins were higher in AC6^loxloxCre mice fed a standard diet than in AC6^loxlox mice fed a Li⁺ diet, indicating AC6-independent effects. Values are mean ± SEM. Sample sizes: AC6^loxlox standard = 5, AC6^loxlox Li⁺ = 6, AC6^loxloxCre standard = 4, AC6^loxloxCre Li⁺ = 5. All statistical comparisons were performed using 2-way ANOVA followed by Holm-Sidak post-hoc tests. *P < 0.05.

Figure 7

Li⁺ reduces outer medullary/cortical AQP2 and pS256-AQP2 abundances in AC6^loxloxCre and AC6^loxlox mice. (A) Semiquantitative immunoblotting of outer medulla/cortex (OM/cortex) homogenates from mice with deletion of adenylyl cyclase 6 (AC6) in aquaporin-2–expressing (AQP2-expressing) cells (AC6^loxloxCre mice) and AC6^loxlox controls using antibodies targeting various proteins modulated during Li⁺-induced nephrogenic diabetes insipidus. (B) Summary data demonstrate lower abundances of AQP2 and pS256-AQP2 in response to Li⁺ administration in both genotypes. The abundances of these proteins were higher in AC6^loxloxCre mice fed a standard diet than in Li⁺-administered AC6^loxlox mice, indicating AC6-independent effects. Values are mean ± SEM. Sample sizes: AC6^loxlox standard = 5, AC6^loxlox Li⁺ = 6, AC6^loxloxCre standard = 4, AC6^loxlox Li⁺ = 5. All statistical comparisons were performed using 2-way ANOVA followed by Holm-Sidak post-hoc tests. *P < 0.05.

Li⁺ increases the number of PCNA-positive cells in IMCD₁ in AC6^loxloxCre and AC6^loxlox mice and enhances IC numbers in AC6^loxloxCre mice. Li⁺ administration is known to induce cell proliferation (3, 29, 30). We examined the contribution of AC6 to this phenomenon by determining the levels of proliferating cell nuclear antigen (PCNA) in kidneys from AC6^loxloxCre and AC6^loxlox mice fed a standard diet or a Li⁺-containing diet for 15 days. Li⁺ administration increased the numbers of PCNA-positive cells in both genotypes, as demonstrated by greater total PCNA protein abundance in IM homogenates (Figure 5, P < 0.05) and a higher proportion of cells in the IMCD₁ (at the base of IM) expressing PCNA (Figure 8, A–I). Furthermore, a higher proportion of the PCNA-positive cells was AQP2 positive versus H⁺-ATPase B1 subunit positive in both genotypes (Figure 8J, P < 0.05). However, this difference was less pronounced in AC6^loxloxCre mice, in which fewer of the PCNA-labeled cells were AQP2 positive and more PCNA-labeled cells were H⁺-ATPase B1 subunit positive compared with AC6^loxlox mice (Figure 8J, P < 0.05). The Li⁺-administered AC6^loxloxCre mice showed a lower proportion of AQP2-positive cells and a higher proportion of H⁺-ATPase B1 subunit–positive cells in IMCD₁ compared with mice fed a standard diet (Figure 8K, P < 0.05), whereas these changes were not observed in AC6^loxlox mice (Figure 8K). In terms of PCNA-positive cells, we corrected for the difference in the number of AQP2-positive and H⁺-ATPase B1 subunit–positive cells by calculating the proportion of AQP2-positive and H⁺-ATPase B1 subunit–positive cells that were expressing PCNA. AC6^loxlox mice showed a higher proportion of AQP2-positive cells expressing PCNA compared with H⁺-ATPase B1 subunit–positive cells (Figure 8L, P < 0.05), whereas no effect was found in AC6^loxloxCre mice (Figure 8L). In both genotypes, H⁺-ATPase B1 subunit–positive cells were frequently situated next to each other in rows in the IMCD₁ (Figure 8, F, H and M), a finding which was not observed in mice fed a standard diet (Figure 8, E, G and M). There were no clear differences in the number of cells coexpressing AQP2 and the H⁺-ATPase B1 subunit under standard conditions or after Li⁺ administration (data not shown) (31, 32).

Figure 8

Li⁺ increases the number of PCNA-positive cells in IMCD₁ in AC6^loxloxCre and AC6^loxlox mice and enhances intercalated cell numbers in AC6^loxloxCre mice. (A–D) Immunohistochemical labeling of proliferating cell nuclear antigen (PCNA) at the base of the inner medulla (IM) in mice with deletion of adenylyl cyclase 6 (AC6) in aquaporin-2–expressing (AQP2-expressing) cells (AC6^loxloxCre mice) and AC6^loxlox control mice following Li⁺ administration. Scale bar: 100 μm. (E–H) Immunohistochemical triple-fluorescence labeling in IM₁ for AQP2 (red), H⁺-ATPase B1 subunit (H⁺, gray), and PCNA (green). White arrows indicate PCNA/AQP2-positive cells, arrowheads indicate PCNA/H⁺-ATPase B1 subunit–positive cells, and yellow arrows indicate H⁺-ATPase–positive cells situated in rows. Scale bar: 20 μm. (I) Percentage of cells in IMCD₁ that were PCNA positive. (J) Percentage of PCNA-positive cells in IMCD₁ that were AQP2- positive or H⁺-ATPase B1 subunit positive. (K) Percentage of cells in IMCD₁ that were AQP2 positive or H⁺-ATPase B1 subunit positive. (L) Percentage of H⁺-ATPase B1 subunit– and AQP2-positive cells that showed PCNA labeling. (M) Percentage of H⁺-ATPase B1 subunit–positive cells situated in rows. Values are mean ± SEM. Mean number of cells counted per mouse ± SEM (n) were as follows: AC6^loxlox standard: 254 ± 19 (5), AC6^loxlox Li⁺: 372 ± 24 (6), AC6^loxloxCre standard: 302 ± 30 (4); and AC6^loxloxCre Li⁺: 441 ± 93 (5). Statistical comparisons were performed using 2-way ANOVA followed by Holm-Sidak post-hoc tests (I and K), 2-way repeated-measurements ANOVAs followed by Holm-Sidak post-hoc tests (J and L), or Student’s 2-tailed t test (M). *P < 0.05.

The most frequent side effect of Li⁺ therapy is NDI. Li⁺ interferes with the signal-transduction pathway of AVP, resulting in reduced AVP sensitivity in CD PCs and diminished levels of AQP2 expression (33). This reduced expression, coupled with reduced phosphorylation (34) at essential activation sites (35), is likely to be the major cell biological mechanism underlying Li-NDI. However, how Li⁺ therapy inhibits the signal-transduction pathway of AVP is not clear. One mechanism postulated for these effects is that Li⁺ inhibits the activity of ACs, leading to decreased cAMP production and ultimately decreased gene transcription and cAMP-dependent phosphorylation of AQP2. However, the major finding of the current study that Li⁺ treatment induces NDI of a similar magnitude in AC6^–/– mice and control mice suggests that in vivo AC6 and medullary cAMP generation play a minor role in the development of Li-NDI.

One of the major hurdles for determining a direct role of ACs and cAMP generation for the development of Li-NDI in vivo has been the lack of a suitable animal model. For example, several previous studies implicating a role for reduced cAMP in the onset of Li-NDI did not measure cAMP levels in vivo (12, 36), whereas a direct cause-effect correlation cannot be concluded from other models with diminished cAMP levels following Li⁺ treatment (17) due to compensatory mechanisms, such as a negative-feedback response of cells to increased AVP levels (37). A strength of the current study was the use two different mouse models that lack AC6, the AC isoform mediating the majority of AVP-induced IM cAMP formation, allowing us to assess directly a role of cAMP in Li-NDI.

Similarly to AC6^–/– mice (Figure 1), our conditional mouse model (AC6^loxloxCre mice) with AC6 deletion specifically in CNT cells and CD PCs demonstrated key features of NDI, including significantly higher water intake and lower urine osmolality under baseline conditions. A minor limitation of our constitutive knockout model is that we cannot completely exclude that developmental and physiological compensatory mechanisms play a role in some of the observed effects. However, with a phenotype closely resembling that of AC6^–/– mice, the new model allowed us to examine the effect of Li-NDI following loss of AC6 specifically in the renal CD (19).

The current data using AC6^loxloxCre mice, coupled with our previous data (19), strengthen the theory that AC6 is responsible for AVP-mediated cAMP formation in the CD and, subsequently, the majority of cAMP-mediated effects in PCs. However, in AC6^–/– and AC6^loxloxCre mice, dietary Li⁺ intake for between 15 and 27 days resulted in further reductions in urine osmolality and increases in water intake. Although there was a delay of approximately 2 days in the renal response to dietary Li⁺ intake in control mice (which we hypothesize is due to the time required for medullary osmolality gradients to be diminished), the overall reductions in urine osmolality and increases in water intake in both AC6 knockout models, relative to their respective baseline period, were of a similar magnitude to their respective control mice. Furthermore, Li⁺ administration resulted in greater reductions in AQP2 and pS256-AQP2 abundances than observed with genetic deletion of AC6; hence, the isolated effect of Li⁺ was greater than the effect of deleting AC6, suggesting that AC6- and cAMP-independent Li⁺ effects on the signal-transduction pathway modulate AQP2 expression. Together, these data indicate that development of Li-NDI is independent of the actions of AC6 in CD PCs and, thus, cAMP. The current data are in line with a previous study performed in Brattleboro rats with clamped blood dDAVP levels, in which during Li-NDI there was no difference in dDAVP-generated cAMP generation or AQP2 levels in IMs compared with control Brattleboro rats (38).

Besides reduced AQP2 expression, in response to Li⁺, the kidney CD “remodels,” with an increased ratio of ICs to PCs (3). It is well established that Li⁺ induces an increase in the numbers/percentages of both ICs and PCs expressing the proliferation marker, PCNA (3, 30). The increased PCNA expression is reduced by inhibiting the mTOR pathway, a controller of cell proliferation, whereas the inhibitory effect of Li⁺ on GSK3β and AQP2 is unaffected by rapamycin (39). mTOR may also stimulate checkpoint kinase 1 (Chk1) and thereby prevent cell cycle progression through the G₂ phase (30, 40). Hence, this mechanism may underlie the higher number of PCs expressing PCNA during Li⁺ administration. Li⁺ also induces renal distal tubular acidosis (41, 42), a condition that in itself may result in IC proliferation (43). Thus, it could be speculated that, during Li⁺ administration, PCNA-positive ICs are dividing in response to metabolic acidosis, while the majority of PCs expressing PCNA are arrested in the G₂ phase due to direct effects of Li⁺ (30). Of note, after 14 days of NH₄Cl-induced acidosis, the percentage of ICs expressing PCNA in the OMCD was reported to be approximately 5-fold higher compared with the percentage of PCs (43). This is in contrast to the Li⁺-induced (15 days) PCNA expression in our AC6^loxlox control mice, in which the percentage of PCNA-positive PCs was significantly higher than the percentage of PCNA-positive ICs. Thus, Li⁺-administered mice may show a higher number of PCs expressing PCNA compared with what would be expected during normal acidosis.

On standard diet, AC6^loxloxCre mice had a higher IC-to-PC ratio compared with AC6^loxlox mice, yet AC6^loxloxCre mice did not present ICs in rows following Li⁺ treatment, indicating the occurrence of two separate underlying mechanisms. It could be speculated that, while the Li⁺-induced increase in IC-to-PC ratio was due to proliferation of ICs, the increased IC-to-PC ratio induced by AC6 deletion could be due to interconversion of PCs to ICs, as previously reported in K⁺ depletion experiments (32). Thus, the combined effect of Li⁺ and AC6 deletion could have a positively additive effect on the IC-to-PC ratio. Whether AC6 in the kidney is an indicator/modulator of differentiation as described in the mouse colon (44), in which AC6 is downregulated in the most differentiated surface cells, remains to be determined. However, based on the current data, it is plausible that AC6 plays distinctive roles in epithelial cell proliferation and may even be a determinant of the conversion of PC to IC (45). Examination of Li⁺ effects in cell lineage–tracing models, or such models with an AC6-deficient background, would be a powerful approach to examine these possibilities (46, 47).

As highlighted in the Introduction, Li⁺ acts as a counterregulator of AVP by modulating a variety of different pathways (7–13), but it is often thought that cAMP is the second messenger mediating these events. An example of this comes from recent studies indicating a role for the ubiquitously expressed GSK3, which is effectively inhibited by Li⁺, in the onset of Li-NDI. GSK3β is suggested to be important for modulating AC activity, cAMP generation, and AVP sensitivity in the CD (18). However, GSK3α knockout mice, despite having impaired AVP-induced cAMP responses in CDs, are only mildly resistant to the onset of Li-NDI (48). Similarly, in the current study, although we saw clear inhibition of GSK3 activity (as evidenced by increased phosphorylation at an inhibitory site) following Li⁺ administration, the developed Li-NDI must be cAMP independent. If Li-NDI can develop in the absence of AC6 and changes in cAMP, what is the mechanism? It is plausible that, due to its pleiotropic actions, inhibition of GSK3 will affect a number of cAMP-independent intracellular signaling cascades and is thus still a central player in development of Li-NDI. Furthermore, very recent studies have demonstrated that cAMP-independent pathways from the vasopressin type 2 receptor for AQP2 membrane targeting exist (49), which may also be disrupted by Li⁺ treatment. Further studies are required to address the cAMP-independent signaling pathways involved in the development of Li-NDI.

In summary, our study shows that AC6 function and reduced cAMP levels are not major factors in the onset of Li-NDI. However, the data postulate a new function for AC6, in which it plays an instrumental role in determining the IC-to-PC ratios that occur during Li-NDI.

Generation and maintenance of AC6 knockout mouse models. AC6^–/– mice were described previously (50), and WT littermates were used as controls. A conditional mouse model with AC6 constitutively deleted specifically in CNT cells and CD PCs (termed AC6^loxloxCre mice) was generated by breeding transgenic mice in which exons 3–12 of the Adcy6 gene were flanked by loxP sites (termed AC6^loxlox, ref. 51) with mice expressing Cre recombinase (codon improved) under the regulatory elements of the Aqp2 gene (25, 28). Mice were housed at approximately 20°C under a 12-hour-light/dark cycle in standard mouse cages with free access to tap water and standard rodent chow (0.8% NaCl, TD.7001, Harlan Teklad). Genotyping was performed by PCR analyses of ear biopsies (for primers, see Table 2).

Table 2

Genotyping primers

Animal experimental protocols and urine and plasma analyses. All experiments were conducted in regular cages. Following an initial baseline period on standard rodent chow (0.8% NaCl, TD.7001, Harlan Teklad), AC6^–/– and WT mice were shifted to a Li⁺-containing diet (0.2% LiCl [40 mmol/kg], Harlan Teklad, TD09326) for 27 days. Body weight as well as water and food intake was determined, and urine was collected via reflex urination. Similarly, AC6^loxloxCre and AC6^loxlox mice were followed under an initial baseline period on standard rodent chow and then either maintained on the standard diet or shifted to the Li⁺ diet for a period of 15 days. Mice were not provided with salt blocks during the Li⁺ administration, because this may lead to extensive salt intake. Except for one female in the AC6^loxloxCre standard group and one female in the AC6^loxloxCre Li⁺ group, all mice were males. On day 15, mice were anesthetized with isoflurane, blood was sampled from the retrobulbar plexus, and one of the kidneys was removed for generation of immunoblotting samples. For IHC, mice were subsequently perfused through the left ventricle with 4% paraformaldehyde and the kidney was removed and post-fixed in 4% paraformaldehyde at 37^oC for 1 hour, followed by post fixation at 4°C for 24 hours. Urine and plasma osmolalities were measured by vapor pressure (Vapro, Wescor), while plasma [Li⁺] was determined using a commercial assay (Infinity Lithium Reagent, Beckman Coulter).

Semiquantitative immunoblotting. The kidney was dissected on ice into IM, cortex, and the remaining tissue (enriched for OM/cortex) and homogenized in cold buffer containing 250 mmol/l sucrose, 10 mmol/l triethanolamine, protease inhibitors (Sigma-Aldrich and Roche Applied Science, respectively), and phosphatase inhibitors (Thermo Scientific). The homogenates were centrifuged at 1,000 g for 10 minutes, after which the supernatants were processed for immunoblotting as previously described (52). PVDF membranes were incubated with primary antibodies targeting AC5/6 (sc-590, Santa Cruz; dilution 1:800), AQP2 (H7661, dilution 1:1,500) (53), AQP2 (9398, dilution 1:1,000) (54), pS256-AQP2 (K0307, dilution 1:500) (55), pS269-AQP2 (dilution 1:500) (54), H⁺-ATPase B1 subunit (H7659, dilution 1:2,000) (3), PCNA (P8825; Sigma-Aldrich; dilution 1:3,000), αENaC (dilution 1:1,000, provided by Johannes Loffing, Institute of Anatomy, Zürich, Switzerland) (56), γENaC (5361, dilution 1:200) (57), NCC (SPC-402D; StressMarq Biosciences Inc.; dilution 1:1,000), NKCC2 (1495, dilution 1:100) (58, 59), pT96/T101-NKCC2 (9934, dilution 1:50) (60), GSK-3α/β antibody (0011-A, Santa Cruz, dilution 1:200), and pS9-GSK3β (5B3, 9323; Cell Signaling Technology Inc.; dilution 1:500). Secondary antibodies were from DAKO, and sites of antibody/antigen interaction were visualized using the Enhanced Chemiluminescence System (GE Healthcare) and an ImageQuant LAS 4000 imager (GE Healthcare). Densitometric analyses were performed using Image Studio Lite (Qiagen).

IHC. Kidney tissue was processed for paraffin embedding and labeling as previously described (61). Two-μm sections were single immunolabeled for light microscopy (Cre recombinase [Covance; dilution 1:5,000] and PCNA [P8825; Sigma-Aldrich, dilution 1:20,000]) and triple immunolabeled for confocal laser scanning microscopy (PCNA [dilution 1:5,000], AQP2 [7661, dilution 1:1,000] [ref. 54], and biotinylated H⁺-ATPase B1 subunit [7659, dilution 1:50][ref. 3]). Imaging was performed using a Leica DMRE light microscope equipped with a digital camera (Leica) and a Leica TCS SL laser scanning confocal microscope and Leica confocal software (Leica). Brightness was digitally enhanced on presented images. Cell counting was performed on fluorescence images acquired from the first 250 μm of IM₁, starting at the transition from the inner stripe OM.

Statistics. Pairwise comparisons of data meeting the statistical assumptions of normality and variance homogeneity were performed using Student’s 2-tailed t test, while data only meeting assumptions of normality were analyzed using Satterthwaite’s 2-tailed unequal variance t test. Comparisons of more than two groups were performed using either a 2-way repeated-measurements ANOVA or a 2-way ANOVA. Data not meeting the assumptions of normality and variance homogeneity were ln or square root transformed, and proportional data were either logit or arcsin transformed (62). Analyses were carried out using Stata 12.0 (StataCorp) or SigmaPlot 12.0 (Systat Software Inc.) for Windows. Values are presented as individual data points and mean ± SEM. P values of less than 0.05 were considered significant.

Study approval. All animal experimentation was conducted in accordance with the Guide for the Care and Use of Laboratory Animals (National Academies Press. 2011) and was approved by the local institutional animal care and use committee (VA San Diego Healthcare System).