Aurora Kinase A Is Involved in Controlling the Localization of Aquaporin-2 in Renal Principal Cells

The cAMP-dependent aquaporin-2 (AQP2) redistribution from intracellular vesicles into the plasma membrane of renal collecting duct principal cells induces water reabsorption and fine-tunes body water homeostasis. However, the mechanisms controlling the localization of AQP2 are not understood in detail. Using immortalized mouse medullary collecting duct (MCD4) and primary rat inner medullary collecting duct (IMCD) cells as model systems, we here discovered a key regulatory role of Aurora kinase A (AURKA) in the control of AQP2. The AURKA-selective inhibitor Aurora-A inhibitor I and novel derivatives as well as a structurally different inhibitor, Alisertib, prevented the cAMP-induced redistribution of AQP2. Aurora-A inhibitor I led to a depolymerization of actin stress fibers, which serve as tracks for the translocation of AQP2-bearing vesicles to the plasma membrane. The phosphorylation of cofilin-1 (CFL1) inactivates the actin-depolymerizing function of CFL1. Aurora-A inhibitor I decreased the CFL1 phosphorylation, accounting for the removal of the actin stress fibers and the inhibition of the redistribution of AQP2. Surprisingly, Alisertib caused an increase in actin stress fibers and did not affect CFL1 phosphorylation, indicating that AURKA exerts its control over AQP2 through different mechanisms. An involvement of AURKA and CFL1 in the control of the localization of AQP2 was hitherto unknown.


Introduction
Arginine-vasopressin (AVP)-dependent water reabsorption in renal-collecting duct principal cells relies on the tight regulation of the cyclic adenosine monophosphate (cAMP) signaling pathway. The activation of vasopressin V2 receptors (V2R) on the basolateral plasma membrane of the principal cells by AVP induces an elevation of intracellular cAMP AURKA phosphorylates the actin regulatory protein cofilin-1 (CFL1) and thereby affects actin remodeling during mitosis [33]. We had previously carried out an siRNA screening in immortalized mouse medullary collecting duct (MCD4) cells to identify kinase-related genes involved in the control of AQP2. The screening identified 19 kinases whose knockdown inhibited the cAMP-induced redistribution of AQP2, amongst them AURKA [34]. We therefore further investigated a potential role of AURKA in the control of AQP2. We report here an involvement of AURKA and CFL1 in the control of the localization of AQP2 in two renal principal cell models. Within this context we synthesized new AURKA-selective small molecule inhibitors, which provide valuable insights into the structure-activity relationship of the Aurora-A inhibitor I. To our knowledge, a connection between CFL1 as actin regulator and AQP2 trafficking or a role of AURKA in this process is hitherto unknown.

Inhibition of AURKA Prevents the cAMP-Induced Redistribution of AQP2
In our previous siRNA screening we had identified AURKA as a potential regulator of the AQP2 redistribution downstream of PKA [34]. To validate the inhibitory effect of the knockdown of AURKA on the AQP2 redistribution, we examined the effect of a selective AURKA inhibitor (Aurora-A inhibitor I; TC-S 7010) on the AQP2 localization in our two cell models for AQP2 trafficking, MCD4 and primary rat inner medullary collecting duct (IMCD) cells ( Figure 1). Aurora-A inhibitor I is a potent inhibitor of AURKA with an IC 50 of 3.4 nM in a cell-free assay (Table 1; [28]). It is 1000-fold more selective for AURKA over AURKB. Aurora-A inhibitor I induces G2/M cell cycle arrest, multinucleation, and apoptosis in ARID1A-deficient colorectal cancer cells [35].
MCD4 and primary IMCD cells were treated with Aurora-A inhibitor I prior to their stimulation. Unlike primary IMCD cells, which can be stimulated with AVP, MCD4 cells lack the V2R. Their stimulation is achieved by the cAMP-elevating agent forskolin, which directly activates adenylyl cyclases. In resting cells, AQP2 is mainly localized on intracellular vesicles in the perinuclear area. In response to forskolin or AVP, AQP2 is redistributed to the plasma membrane. The AQP2 localization was analyzed by immunofluorescence microscopy (Figure 1). The AQP2 fluorescence signals at the plasma membrane and perinuclear area of the cells were determined, and the ratios of plasma membrane to perinuclear AQP2 fluorescence intensities were calculated. A ratio >1 indicated a predominant localization at the plasma membrane, while ratios <1 indicated a predominant intracellular localization.
The AURKA-selective inhibitor was effective in concentrations greater than 1 µM and inhibited the redistribution in both cell models. We did not observe the inhibitory effect in all cells most likely because the expression of AURKA is cell cycle-dependent [18], and unsynchronized MCD4 cells might be differentially affected by AURKA inhibition depending on their cell cycle state, density, and general proliferation rate.
The use of compound concentrations in the low µM range is common for the principal models in this study. Other examples include the antimycotic drug fluconazole [36], inhibitors of the interaction of AKAP-Lbc with RhoA [37], or Rho inhibitors [4]. The apparently high concentrations are most likely due to the robust plasma membranes of the cells, which, in situ in the kidney are exposed to various kinds of adverse environmental factors, including nephrotoxic agents in serum or urine. In the further experiments in primary IMCD cells, compounds were applied in the low µM concentration range, which did not alter the cell shape. were left untreated or treated with DMSO (0.3%, 1 h) or Aurora-A inhibitor I (1 µM, 10 µM or 30 µM, 1 h) alone or in combination with forskolin (Fsk; 30 µM, 30 min) or arginine-vasopressin (AVP; 100 nM, 30 min) where indicated. AQP2 was detected by immunofluorescence microscopy using specific primary antibody (AQP2, E-2) and Cy3-coupled anti-mouse secondary antibody (red). Nuclei were stained with DAPI (cyan). Shown are representative images of three independent experiments; scale bar 30 µm. Magnified views (zoom) are from the indicated boxes. The intensities of plasma membrane and perinuclear immunofluorescence signals arising from AQP2 were determined. The ratios of plasma membrane to perinuclear fluorescence signal intensities were calculated. Ratios >1 indicate a predominant localization at the plasma membrane. Shown are means ± SEM of three independent experiments with a total of 75 cells per condition. Statistically significant differences compared to DMSO-treated control cells are indicated, **** p < 0.0001, *** p < 0.001.

Inhibition of AURKA but Not AURKB Prevents the cAMP-Induced Redistribution of AQP2
Next, we sought to define whether the observed inhibitory effect on the AQP2 redistribution was due to inhibition of AURKA and not one of its paralogues. We thus treated our cells with another commercially available AURKA-selective inhibitor, Alisertib (MLN8237), as well as an AURKB-selective inhibitor, Barasertib-HQPA (AZD-1152-HQPA; Figure 2, Table 1). Alisertib has an IC 50 of 1.2 nM in a cell-free assay and >200-fold higher selectivity for AURKA over AURKB [38]. Aurora-A inhibitor I has antiproliferative effects [38,39]. Barasertib-HQPA is a potent AURKB inhibitor with an IC 50 of 0.37 nM in a cell-free assay and a~3800-fold selectivity for AURKB compared to AURKA [40,41]. We excluded AURKC-selective inhibitors from testing as the expression of AURKC has been mainly reported for germ cells [18]. We thus expected that the biological effects we observed in our renal cell models could only be elicited by the activity of AURKA or AURKB.
We treated MCD4 cells with the inhibitors and analyzed the AQP2 localization by immunofluorescence microscopy (Figure 2A). Alisertib significantly inhibited the forskolininduced redistribution of AQP2, whereas Barasertib-HQPA did not. Intrigued by the AURKA-specific effect, we continued by testing the effect of the inhibitors in our primary cell model. In primary IMCD cells, AURKA inhibition by Alisertib significantly inhibited the redistribution compared to control cells treated with the compound solvent, DMSO. No inhibitory effect on the AQP2 redistribution was visible after inhibiting AURKB ( Figure 2B). The observed inhibitory effects on the AQP2 redistribution elicited by two structurally different AURKA inhibitors (Table 1), Aurora-A inhibitor I and Alisertib, demonstrated an AURKA-specific role in the control of the AQP2 localization.  DMSO (0.3% or 0.2%, 1 h), Alisertib (AURKA inhibitor; 30 µM or 20 µM, 1 h), or Barasertib-HQPA (AURKB inhibitor; 30 µM or 20 µM, 1 h) alone or in combination with forskolin (Fsk; 30 µM, 30 min) or arginine-vasopressin (AVP; 100 nM, 30 min) where indicated. AQP2 was detected by immunofluorescence microscopy as described in the legend of Figure 1. Shown are representative images of four independent experiments; scale bar 30 µm. AQP2 was quantified using a segmentationbased approach [42]. Ratios > 1 indicate a predominant localization at the plasma membrane. Shown are means ± SEM of four independent experiments with ≥25 cells per image per condition. Statistically significant differences are indicated, **** p < 0.0001, ** p < 0.01, * p < 0.05.

Aurora-A Inhibitor I Derivatives Show Varying Effects on the AQP2 Localization
We aimed to further elucidate the relationship between the chemical structure of Aurora-A inhibitor I and its biological activity as an inhibitor of AURKA and the AQP2 redistribution. Preparation of Aurora-A inhibitor I derivatives by medicinal chemistry [28] yielded a new set of compounds (Table 2). We analyzed their effects on the AQP2 localization by immunofluorescence microscopy. In MCD4 cells, the derivatives showed varying inhibitory effects on the forskolin response ( Figure 3). In the presence of forskolin in combination with either ERJ2-04, -06, -11, or -12, the localization of AQP2 was not different from that in cells treated with the compound solvent, DMSO. Therefore, being amongst the most effective inhibitors of the AQP2 redistribution, ERJ2-04, -06, -11, and -12 were chosen for further testing in primary IMCD cells ( Figure 4). There was no significant difference between resting and AVP-stimulated cells that had been treated with ERJ2-04, -06, or -11 but AQP2 fluorescence intensity ratios significantly differed for ERJ2-12-treated cells. A significant decrease in the AQP2 fluorescence intensity ratio compared to DMSO-treated control cells was detected upon treatment with ERJ2-11. We occasionally observed prominent AQP2 speckles in the perinuclear area upon treatment with this compound (not shown), which might explain the shift towards lower AQP2 intensity ratios. immunofluorescence microscopy as described in the legend of Figure 1. Shown are repre images of four independent experiments; scale bar 30 µm. AQP2 was quantified segmentation-based approach [42]. Ratios > 1 indicate a predominant localization at the membrane. Shown are means ± SEM of four independent experiments with ≥25 cells per im condition. Statistically significant differences are indicated, **** p < 0.0001, ** p < 0.01, * p <

Aurora-A Inhibitor I Derivatives Show Varying Effects on the AQP2 Localization
We aimed to further elucidate the relationship between the chemical stru Aurora-A inhibitor I and its biological activity as an inhibitor of AURKA and th redistribution. Preparation of Aurora-A inhibitor I derivatives by medicinal ch [28] yielded a new set of compounds (Table 2). We analyzed their effects on th localization by immunofluorescence microscopy. In MCD4 cells, the derivatives varying inhibitory effects on the forskolin response ( Figure 3). In the presence of f in combination with either ERJ2-04, -06, -11, or -12, the localization of AQP2 different from that in cells treated with the compound solvent, DMSO. Therefor amongst the most effective inhibitors of the AQP2 redistribution, ERJ2-04, -06, -1 12 were chosen for further testing in primary IMCD cells ( Figure 4). There significant difference between resting and AVP-stimulated cells that had been with ERJ2-04, -06, or -11 but AQP2 fluorescence intensity ratios significantly diff ERJ2-12-treated cells. A significant decrease in the AQP2 fluorescence intensi compared to DMSO-treated control cells was detected upon treatment with ERJ2 occasionally observed prominent AQP2 speckles in the perinuclear area upon tr with this compound (not shown), which might explain the shift towards lowe intensity ratios.   Figure 1. Shown are representative images of four independent experiments; scale bar 30 µm. AQP2 was quantified using a segmentation-based approach [42]. Ratios > 1 indicate a predominant localization at the plasma membrane. Shown are means ± SEM of four independent experiments with ≥25 cells per image per condition. Statistically significant differences between resting and forskolin-stimulated cells are indicated, *** p < 0.001, ** p < 0.01, * p < 0.05. Additional images can be found in Appendix A ( Figure A1).

Cell Viability Is Affected by Some of the Aurora-A Inhibitor I Derivatives
Antiproliferative effects elicited by Aurora-A inhibitor I had previously been described for the colon cancer cell lines HCT116 and, to a smaller extent, HT-29 [28,35]. None of our inhibitors induced apparent cell shape changes in MCD4 or primary IMCD cells. However, in order to investigate the effects on the viability of our test cells upon inhibition of AURKA, we carried out MTS assays ( Figure 5). MTS assays rely on the conversion of a tetrazolium compound into a colored formazan product by NADPH or NADH in metabolically active cells [43,44]. They are based on the assumption that the production is proportional to the number of viable cells. Figure 5. Effects of Aurora-A inhibitor I and its derivatives on the viability of MCD4 cells. Cell proliferation was assessed using the CellTiter 96 ® Aq ueous One Solution Cell Proliferation Assay (Promega; Walldorf, Germany). MCD4 cells were left untreated (Utr) or treated with DMSO (0.3%, 1 h) or compound (30 µM, 10 µM or 1 µM, 1 h). As negative controls, i.e., to induce maximal toxicity, cells were incubated with ethanol (EtOH). Shown are means ± SEM of three or six independent experiments. Statistically significant differences compared to untreated cells are indicated, **** p < 0.0001, ** p < 0.01, * p < 0.05.
Amongst the AURKA inhibitors with inhibitory properties on the AQP2 redistribution, Aurora-A inhibitor I and ERJ2-06 did not elicit any significant effects on the viability of MCD4 cells within the 1 h of treatment. Cell viability was reduced by 10 or 30 µM of ERJ2-03, -04, or -11. ERJ2-03 and -04 both have a morpholine moiety and ERJ2-11 a piperazine moiety. As morpholine and piperazine are well-known drug-like features, these moieties are unlikely to cause the reduction in cell viability, suggesting that another part of these compounds exerts the deleterious effect on MCD4 cell viability [45,46]. This is supported by the observation that ERJ2-12, which also has a piperazine fragment, does not affect the viability of MCD4 cells. With ERJ2-03 not being an inhibitor of the AQP2 redistribution ( Figure 3), reduced viability of the MCD4 cells did not necessarily seem to be accompanied by an impaired AQP2 trafficking. Incubation with ERJ2-05 or -08 partly increased the viability compared to untreated control cells. This could be due to interference of the compound with the colorimetric assay itself or could be due to off-target effects in the cells. The observation that the Aurora-A inhibitor I did not affect viability of MCD4 cells was reassuring that AURKA is directly involved in the control of AQP2.

Aurora-A Inhibitor I and Its Derivatives Show Favorable Selectivity Profiles
Next, we confirmed the specificity of our new Aurora-A inhibitor I derivatives by evaluating binding to 45 different kinases in a thermal shift assay [47] ( Figure 6A). The binding of a compound to a kinase domain induces a measurable shift in its melting temperature (∆Tm). ∆Tm values are unique for each kinase, i.e., binding affinities cannot be directly derived, and ∆Tm values have to be interpreted in the context of a positive control such as staurosporine, a global kinase inhibitor. Our measurements confirmed a strong affinity of the tested compounds to AURKA, underpinned by ∆Tm values in the range of the positive control. Aurora-A inhibitor I and ERJ2-06 induced the highest ∆Tm. None of the compounds bound to AURKB. Moderate ∆Tm values were measured for a small number of other serine/threonine kinases such as AP2-associated protein kinase 1 (AAK1), Cyclin G-associated kinase (GAK), and Unc-51like kinase 3 (ULK3). However, those shifts were not in the range of the positive control. To our knowledge, a role of these kinases in the control of AQP2 has not been investigated.
In quantitative phosphoproteomic analyses, AAK1 was shown to be phosphorylated upon V2R activation in renal collecting duct cells [48,49]. However, its inhibition by our AURKA inhibitors is unlikely because inhibition of clathrin-mediated endocytosis in renal principal cells causes an accumulation of AQP2 in the plasma membrane [50]. In a transcriptional profiling study, only GAK and the ULK3 homologue ULK1 were found to be expressed in rat IMCD cells [51]. GAK also regulates clathrin-mediated endocytosis [52,53], and its inhibition by AURKA inhibitors as a reason for the inhibition of the AQP2 redistribution is unlikely for the same reason as for AAK1: GAK inhibition would be expected to cause an accumulation of AQP2 in the plasma membrane [50]. In addition, GAK also contributes to centrosome maturation and mitotic progression [54]. Due to its high plasticity, the kinase domain of GAK is known to be predestined to off-targeting by different clinical kinase inhibitors [55] and thus potentially also by our AURKA inhibitors. Conformational plasticity of the kinase domain has also been reported for ULK3 [56]. ULK3 is involved in Sonic Hedgehog (SHH) signaling [57,58] and the induction of autophagy as a consequence of cellular senescence [59].
The selectivity of the Aurora-A inhibitor I derivatives against the three mammalian AURK paralogues was determined by a NanoBRET assay, by which we measured the apparent affinities of our compounds to the respective kinases in HEK293T cells overexpressing the individual kinases ( Figure 6B). Most compounds bound AURKA with affinities in the nanomolar range and displayed higher selectivity for AURKA over the two AURK paralogues. The most potent inhibitor of the AQP2 redistribution, ERJ2-06, showed the strongest affinity for AURKA binding (165 nM). Surprisingly, ERJ2-12 did not bind to any of the AURK paralogues, despite inducing a moderate temperature stabilization in the thermal shift assay ( Figure 6A). This might be due to limited cell penetration of this inhibitor and provides a rationale for the ambiguous results observed for MCD4 and primary IMCD cells (Figures 3 and 4). Moreover, a high cellular ATP concentration acts as a competitive binder to the kinase domains.
On the structural level, the ortho-chlorophenyl on the right-hand side of Aurora-A inhibitor I and ERJ2-02 to -08 appeared to be very important for the compound's activity as an AURKA inhibitor, possibly as it allows a conformationally preferred conformation. Derivatives lacking this group were poor or weaker binders of AURKA (Table 2, Figure 6B). Substitution of the ortho-chlorophenyl by an unsubstituted phenyl rendered the molecules less active (ERJ2-10 and -11). However, ERJ2-11 still potently inhibited the forskolin response in MCD4 cells (Figure 3). Derivatives lacking the ortho-chlorophenyl or unsubstituted phenyl, e.g., ERJ2-12, were inactive. Substitutions of the terminal fragment at the left-hand side were of little importance, and most variations were tolerated. This was in line with previously published work on ortho-chlorophenyl substituted pyrimidines [60]. We next carried out PKA activity assays to rule out off-target effects of the compounds on PKA ( Figure 7A). None of the tested compounds significantly reduced the forskolin response in MCD4 cell lysates compared to forskolin-stimulated control lysates. Global effects of the derivatives on phosphatase activity in MCD4 cells were assessed by a para-Nitrophenylphosphate (pNPP)-based activity assays. No significant differences compared to control lysates were detected ( Figure 7B). . cAMP (1 µM) was added to induce maximal PKA activity. PKA catalytic (C) subunit (2 µg/mL) was used in positive controls and omitted in negative controls. Relative PKA activity was determined by calculating the ratio of phosphorylated to non-phosphorylated peptides, which were semi-quantitatively analyzed by densitometry. Shown are representative agarose gels and means ± SEM of three independent experiments. Statistically significant differences are indicated, *** p < 0.001, ** p < 0.01, * p < 0.05. (B) Phosphatase activity in MCD4 cell lysates was determined by para-Nitrophenylphosphate (pNPP)-based activity assays. Cells were left untreated or treated with DMSO (0.3%, 1 h) or compound (30 µM, 1 h) alone or in combination with forskolin (Fsk; 30 µM, 30 min). Cell lysates were incubated with a pNPP substrate solution and absorbance measured at 405 nm. Shown are means ± SEM of two to five independent experiments.

The Inhibition of AURKA Affects the Organization of the Actin Cytoskeleton in Both MCD4 and Primary IMCD Cells
Actin remodeling is orchestrated by a variety of actin regulatory proteins including AURKA [33]. We therefore investigated whether the observed inhibitory effect on the AQP2 redistribution by AURKA inhibition could be explained by its role in the regulation of the actin cytoskeleton. For this, we stained F-actin in our cells using a fluorophore-coupled phalloidin to investigate the inhibitors' effects on the actin cytoskeleton by immunofluorescence microscopy.
Resting principal cells exhibit a network of F-actin-containing stress fibers, which depolymerizes upon cAMP elevation [4]. Accordingly, forskolin and AVP decreased the actin stress fibers in both MCD4 ( Figure 8) and primary IMCD cells (Figure 9). Treatment with Aurora-A inhibitor I led to the depolymerization of actin stress fibers even in the absence of a cAMP-elevating agent. Comparable effects were observed for ERJ2-04 and ERJ2-06. Surprisingly, the other AURKA inhibitor, Alisertib, and also ERJ2-11 promoted or maintained the formation of actin stress fibers. The phenotypes of cells treated with ERJ2-12 and Barasertib-HQPA resembled those of the control cells. We concluded that the AURKA inhibitors may act via different mechanisms on the AQP2 localization.
2.7. The Inhibition of AURKA by Aurora-A Inhibitor I Decreases the Phosphorylation of the AURKA Substrate CFL1 in Both MCD4 and Primary IMCD Cells CFL1 is a substrate of AURKA and is involved in the control of the actin cytoskeleton [33]. In its active unphosphorylated form, CFL1 depolymerizes F-actin. We hypothesized that the inhibition of AURKA by our inhibitors might lead to a reduction in phosphorylated (p-)CFL1. This would cause a depolymerization of the actin cytoskeleton even in the absence of forskolin/AVP as observed for Aurora-A inhibitor I, ERJ2-04 and -06 (Figures 8 and 9). However, it would not account for the effects observed after treatment with ERJ2-11 and Alisertib.
We prepared lysates of MCD4 cells that had been incubated with the AURK inhibitors and analyzed the samples by Western blotting (Figure 10). Indeed, in accordance with its actin stress fiber-depolymerizing effect, Aurora-A inhibitor I significantly decreased the CFL1 phosphorylation both in the absence and presence of forskolin. Significant additive effects in combination with forskolin were observed for Aurora-A inhibitor I, ERJ2-04, -06, and -11. AURKA inhibition by the compounds most likely induced a priming effect on the CFL1 dephosphorylation, which is enforced by forskolin. Alisertib elicited a similar effect in forskolin-stimulated cells. However, in contrast to Aurora-A inhibitor I-treated cells, CFL1 phosphorylation was not reduced in resting cells. Compounds that did not inhibit AURKA (ERJ2-12 and Barasertib-HQPA) had no effect on the CFL1 phosphorylation.   . Cell lysates were prepared and CFL1, phosphorylated (p-)CFL1, and cadherin as loading control were detected by Western blotting using specific antibodies. The signals were semi-quantitatively analyzed by densitometry and normalized to the loading control. Shown are representative blots and means ± SEM of three to five independent experiments. Statistically significant differences compared to DMSO-treated control cells are indicated, *** p < 0.001, ** p < 0.01, * p < 0.05.
In primary IMCD cell lysates, no significant changes in the CFL1 phosphorylation could be detected ( Figure 11). However, densitometric analysis of the Western blot samples revealed trends comparable to the results obtained in MCD4 cells ( Figure 10). Relatively low levels of CFL1 phosphorylation were detected after incubation of the cells with Aurora-A inhibitor I or one of its active derivatives.
Taken together, the effects on the CFL1 phosphorylation upon treatment of MCD4 and primary IMCD cells with Aurora-A inhibitor I and its structurally closely related derivatives ERJ2-04 and -06 were in line with their depolymerizing effect on the actin cytoskeleton (Figures 8 and 9). An exception is ERJ2-11, which showed a tendency to promote or maintain the formation of actin stress fibers. This could be explained by structural differences as ERJ2-11 lacks the ortho-chlorophenyl as described before. The two selective AURKA inhibitors Aurora-A inhibitor I and Alisertib differ in their chemical structure, which most likely explains their different effects on the actin cytoskeleton and CFL1 phosphorylation. Collectively, our data show that AURKA exerts the control over the AQP2 localization through different mechanisms: It modulates the actin cytoskeleton through CFL1 phosphorylation (Aurora-A inhibitor I) and through modulating another target and thereby another signaling pathway (Alisertib).

The Inhibition of AURKA and AURKB Has no Effect on the Phosphorylation of LIM Kinase-1 (LIMK-1) in Primary IMCD Cells
Besides AURKA, LIMK-1 as well as phosphatases such as slingshot phosphatase 1 (SSH-1) and chronophin (CIN) regulate CFL1 activity by phosphorylation or dephosphorylation, respectively [33,[61][62][63][64]. In primary IMCD cell lysates we could not detect significant changes in the phosphorylation of LIMK-1 upon treatment of the cells with the compounds (Figure 12). In contrast to AVP, forskolin induced a significant increase in the LIMK-1 phosphorylation. This can most likely be explained by the compartmentalization of cAMP signaling: Forskolin was previously shown to induce higher cAMP levels and relative PKA activity compared to AVP in primary IMCD cells. This was due to global activation of PKA, in contrast to focal activation induced by AVP [65]. On the level of phosphatases, we could not detect global effects of the compounds on phosphatase activity ( Figure 7B). . Cell lysates were prepared and LIMK-1, phosphorylated (p-)LIMK-1 (threonine (Thr) 508), and Hsp60 as loading control were detected by Western blotting using specific antibodies. The signals were semi-quantitatively analyzed by densitometry and normalized to the loading controls. Shown are representative blots and means ± SEM of three independent experiments. Statistically significant differences are indicated, ** p < 0.01.

Discussion
In this study, we showed that two structurally different AURKA inhibitors, Aurora-A inhibitor I and Alisertib, inhibit the AQP2 redistribution in two renal principal cell models. Our findings demonstrate that AURKA is involved in the control of the AQP2 redistribution via at least two different pathways. We propose the mechanisms depicted in Figure 13.

(B) Aurora-A inhibitor I inhibits AURKA and decreases the CFL1 phosphorylation in stimulated cells.
It thereby promotes the F-actin depolymerization. (C) Alisertib inhibits AURKA and increases the amount of F-actin stress fibers without affecting the CFL1 phosphorylation, indicating that another signaling pathway is involved.
AURKA inhibition by Aurora-A inhibitor I elicited a depolymerization of the actin cytoskeleton by reducing the AURKA-catalyzed phosphorylation of CFL1 (Figure 10), which increases its activity [33]. The observed depolymerization of F-actin-containing stress fibers was consistent in both our primary and permanent cell models (Figures 8 and 9) and explains the inhibitory effect on the AQP2 redistribution, because actin stress fibers function as tracks that are necessary for the transport of AQP2-bearing vesicles to the plasma membrane [7,8].
The mode of action of Alisertib and its inhibitory effect on the redistribution of AQP2 cannot be explained by this mechanism. However, it apparently also involves the actin cytoskeleton because an increase in F-actin-containing stress fibers compared to control cells was observed (Figures 8 and 9). F-actin has a dual role in the regulation of the AQP2 redistribution. It not only acts as tracks but in the cell periphery as cortical F-actin also as a physical barrier for AQP2-bearing vesicles, preventing them to reach the plasma membrane [4][5][6]. The partial depolymerization of F-actin, for example, by cytochalasin D, is sufficient to induce the redistribution of AQP2 to the plasma membrane, supporting the notion that cortical F-actin acts as a barrier for this translocation process [4,5]. Alisertib apparently strengthens this physical barrier.
The small GTPase RhoA promotes the formation of cortical F-actin and thereby inhibits the redistribution of AQP2. PKA inhibits RhoA activity, causing the depolymerization of cortical F-actin. This enables AQP2-bearing vesicles to reach the plasma membrane [4,5,66,67]. AURKA phosphorylates Drok, the Rho-kinase orthologue in Drosophila, in vivo and thereby inactivates it [68]. Further, the vertebrate Rho-kinase isoform p160ROCK is phosphorylated by AURKA [69]. Thus, the control over the AQP2 localization by AURKA could involve RhoA and/or its downstream effector ROCK.
Downstream of RhoA/ROCK, actin remodeling is orchestrated by a complex interplay of LIMK-1 and AURKA [33,63,64,70], as well as phosphatases such as SSH-1 and CIN [61,62], which all regulate the activity of CFL1. A functional interaction has been reported for LIMK-1 and AURKA. During mitotic spindle formation, both kinases reciprocally serve as one other's substrate and activate each other by phosphorylation [71]. We could not detect significant effects on the phosphorylation of LIMK-1 upon AURKA inhibition ( Figure 12). LIMK-1 phosphorylates and thereby inactivates CFL1 [63,64]. A similar role has been suggested for AURKA. The kinase directly phosphorylates CFL1 at multiple sites, including at Ser3, and thereby regulates the organization of the actin cytoskeleton in early mitotic stages [33]. In this context, AURKA elicits an inhibitory effect on the CFL1 activity by blocking its binding to actin filaments. Opposite effects have been reported for cell migration and metastasis of various cancers such as papillary thyroid [72] and breast cancer [70], where overexpressed AURKA induced an up-regulation of SSH-1 expression, thus favoring the dephosphorylation and activation of CFL1. In M12 cells, AURKA inhibition by Alisertib increased the phosphorylation of CFL1, possibly as a consequence of decreased SSH-1 levels [33]. We could not detect an increased CFL1 phosphorylation upon treatment with Alisertib by Western blotting (Figures 10 and 11). However, the results from this previous study could provide an explanation for the pronounced actin stress fibers detected in both resting and stimulated Alisertib-treated cells (Figures 8 and 9) Taken together, the involvement of a target other than CFL1 in the control of the AQP2 redistribution seems likely because Alisertib did not elicit the same effects on the actin cytoskeleton and the CFL1 phosphorylation as Aurora-A inhibitor I despite also inhibiting the AQP2 redistribution. Belonging to different classes of compounds, the structural differences of Aurora-A inhibitor I and Alisertib most likely explain their distinct effects on the AURKA kinase activity and the affected downstream signaling pathway.
AURKA kinase activity is spatially and temporally regulated by complex mechanisms [17,29] and determined by the phosphorylation of the threonine (Thr) 288 residue in its activation loop, both by autophosphorylation or phosphorylation by other kinases [73,74]. Thr288 is part of a PKA consensus motif, and AURKA has been shown to become phosphorylated and thereby activated by PKA in vitro [74]. Further, AURKA is activated by mitogen-and stress-activated protein kinase (MSK1) in vitro [73]. We could not detect the phosphorylation of AURKA at Thr288 in our two cell models, possibly because its abundance was below the detection limit. Problems with the detection of p-AURKA signals have been previously reported and were explained by solubility issues, possibly by association of p-AURKA with insoluble cytoskeletal elements [75]. Moreover, most available AURKA and p-AURKA antibodies are reactive towards human AURKA and can thus not be used for the detection in mouse or rat cell lysates as it was the case in this study. Therefore, it is currently unclear whether V2R and subsequent PKA activation lead to the phosphorylation/activation of AURKA.
Different effects of Aurora-A inhibitor I and Alisertib in cells have been reported [75]. For instance, the two inhibitors differed in their affinity towards the complex of AURKA and its allosteric activator targeting protein for Xenopus kinesin-like protein 2 (TPX2). The affinity of Alisertib for AURKA was 5.8-fold decreased in the presence of TPX2, while the binding of Aurora-A inhibitor I to the complex was 2.4-fold increased [75]. These observations can be explained by the presence of different conformational states: The activation loop of the kinase is framed by conserved DFG and APE motifs (one-letter amino acid code) and can adopt different conformations. In the inactive DFG-out state, an additional hydrophobic binding site is exposed next to the ATP binding site [76,77]. Binding of Alisertib to AURKA promoted the inactive DFG-out state, which is counteracted by TPX2 binding to AURKA, which induced the active DFG-in state [78][79][80]. In contrast, binding of Aurora-A inhibitor I to the AURKA binding pocket is less affected as it has little contact with the structural elements, which move upon binding of TPX2 [28,75]. These conformational changes in the kinase induced by inhibitor binding might account for a conformational selectivity, determining the kinase's substrates and which downstream signaling pathways it induces [78]. Such observations may explain why Aurora-A inhibitor I and Alisertib act via different pathways and why ERJ2-11, despite sharing structural similarity with Aurora-A inhibitor I, ERJ2-04 and -06, elicits different effects on the Factin stress fibers. A group of ortho-chlorophenyl substituted pyrimidines, structurally resembling our ERJ compounds, was shown to bind in a DFG-out state [60]. ERJ2-11 lacks the ortho-chlorophenyl that might affect its DFG binding mode.
Based on the discovered different modes of action of Aurora-A inhibitor I and Alisertib with regard to inhibiting the AQP2 redistribution, we analyzed the effect of our compounds on other kinases using a thermal shift assay ( Figure 7A). Profiling against a panel of kinases revealed possible interactions with the serine/threonine kinases AAK1, GAK, and ULK3. The role of these proteins in the control of the localization of AQP2 is unknown. They could potentially constitute additional components of the machinery controlling AQP2 in renal principal cells. Phosphoproteomic studies revealed an AVP/V2R-dependent signaling network in renal-collecting duct cells that involves the regulation of a plethora of kinases, partly in a PKA-independent manner [48,49,81]. We had previously identified 19 potential regulators of the AQP2 redistribution via siRNA-mediated knockdown of 719 kinaserelated genes, one of which was AURKA [34]. However, a comprehensive understanding of the complete signaling pathways involving such a kinase network in the control of AQP2 in renal principal cells is lacking.
A few examples indicate that AURKA does not only play a role in the kidney during mitosis. AURKA activates the vacuolar H + -ATPase (V-ATPase) in Caki-2 cells, a human kidney carcinoma cell line [82]. Inhibition of V-ATPase with 4-acetyldiphyllin (4AD) inhibited the cAMP-dependent AQP2 redistribution. An increase in the intravesicular pH and an accumulation of AQP2 within the Golgi compartment was induced by 4AD [83]. Thus, the inhibitory effect of AURKA may involve V-ATPase. Other non-mitotic functions have been reported for AURKA [84]. For instance, there is an interphase pool of AURKA in ciliated cells, regulating cilia disassembly [85], and neurons, contributing to neurite outgrowth [86]. Our finding that AURKA is involved in the control of AQP2 adds another example to the repertoire of AURKA's non-mitotic functions and supports the notion that AURKA exerts physiological roles outside of the mitotic context. Moreover, the discovery of AU-RKA inhibitors for interfering with the AQP2 trafficking expands the toolbox of inhibitors that directly interfere with vasopressin-mediated water transport [87]. The discovery of novel inhibitors of the redistribution of AQP2 also strengthens arguments that modulating the localization of an aquaporin for therapeutic reasons is feasible. The inhibition of the translocation of AQP2 into the plasma membrane may be an approach to reduce edema in heart failure where elevated levels of AVP cause a predominant localization of AQP2 in the plasma membrane of the renal principal cells [2,88]. Recently, the translocation of AQP4 was prevented with the approved drug trifluoperazine [89]. The drug prevented the localization of AQP4 at the blood-spinal cord barrier and ablated CNS edema in a rat model. Of note, the cAMP-PKA pathway plays a critical role in the control of the localization of both AQP2 and AQP4, underpinning that the pathway is a viable therapeutic target despite being ubiquitous [88].

Chemical Synthesis of Novel Aurora-A Inhibitor I Derivatives
Instructions on the chemical synthesis of novel Aurora-A inhibitor I derivatives can be found in the Supplementary Materials (File S1).

Cell Cultures
Primary rat IMCD cells were obtained and cultured as previously described [36,83,88,89]. MCD4 cells [34,36,90] were grown in DMEM/F-12 GlutaMAX TM (Life Technologies, Darmstadt, Germany; #31331) supplemented with 5% (v/v) FCS and 5 µM of dexamethasone. MCD4 cells were used at 80-90% confluency. Their doubling time is~26 h. The same batch of cells was used for the entire study. Cells were cultured up to passage 20-25. Changes in morphology and AQP2 expression were regularly monitored using an inverted microscope. Primary IMCD cells were seeded to be fully confluent (95-100%). Mycoplasma contamination of MCD4 cells was excluded prior to their freezing and storage in liquid nitrogen using commercially available kits.

Animal Treatment
The local authority (Landesamt für Gesundheit und Soziales, Berlin, Germany) and MDC had approved the use of animals for obtaining primary IMCD cells (Anzeige der Tötung von Wirbeltieren zu wissenschaftlichen Zwecken nach § 4 (3) TSchG). Rats (10-12 weeks, male, Wistar Han) were obtained from Charles River and kept on standard diet and tap water.

Immunofluorescence Microscopy of MCD4 and Primary IMCD Cells
Cells were grown on 12 mm coverslips. Immunofluorescence staining was carried out as previously described [7,34,36,83,91]  Cells were imaged using a confocal laser scanning microscope (LSM) 710 (Carl Zeiss, Jena, Germany) with a Plan-Apochromat 63×/1.40 oil objective. Image analysis was performed using the software Fiji by Image J (Version 2.1.0) [92]. Where indicated, AQP2 localization was quantified using a segmentation-based approach as previously described [42]. For this, masks of the plasma membrane and perinuclear regions were generated and used for the determination of the fluorescence signal intensity arising from AQP2 in the respective cellular area.
Three to five fields of view were taken per sample and experiment. The sample's identity was only revealed after taking representative images. Non-specific binding and background autofluorescence were assessed when establishing the staining protocol by omitting either the primary or secondary antibody and by an unlabeled sample.

MTS Assays
The CellTiter 96 ® Aqueous One Solution Cell Proliferation Assay (Promega, Walldorf, Germany) was used according to the manufacturer's instructions and as previously described [37]. Cells were seeded in a 96-well format (30,000 cells/well) and grown over night. Cells were treated with compounds as indicated. As a control, ethanol (EtOH) was used to induce maximal toxicity. Wells containing medium only served as blanks. After 50 min, the MTS reagent was added to the cell suspension, and incubation at 37 • C continued for another 10 min. The absorbance at 490 nm was measured using a Spark ® 20M Multimode Reader.

Differential Scanning Fluorimetry-Based Selectivity Screening against a Curated Kinase Library (Thermal Shift Assay)
The assay was performed as previously described [47,95]. Briefly, recombinant protein kinase domains (2 µM) were mixed with compound (10 µM) in 20 mM HEPES, pH 7.5 and 500 mM NaCl. SYPRO Orange (5000×, Invitrogen/Thermo Fisher Scientific, Dreieich, Germany) was added as a fluorescence probe (1 µL/mL). Temperature-dependent protein unfolding profiles were measured using the QuantStudio™ 5 real-time PCR machine (Thermo Fisher Scientific). Excitation and emission filters were set to 465 nm and 590 nm, respectively. The temperature was raised with a step rate of 3 • C per minute. Data points were analyzed with the internal software (Thermal Shift Software TM Version 1.4, Thermo Fisher Scientific) using the Boltzmann equation to determine the inflection point of the transition curve.

PKA Activity Assay
PKA activity in MCD4 cell lysates was analyzed using the PepTag ® Non-Radioactive Protein Kinase Assay (Promega, Walldorf, Germany) according to the manufacturer's instructions and as previously described [34,36,83,98].

Statistical Analyses
Data were analyzed using the GraphPad Prism (Version 9.2.0) software. Brown-Forsythe and Welch ANOVA tests with Games-Howell correction or One-Way ANOVA tests with Tukey or Dunnett correction were used. Significant differences are indicated as **** p < 0.0001, *** p < 0.001, ** p < 0.01, or * p < 0.05. Mean plus standard error of the mean (SEM) were plotted.