Aqp5 Is a New Transcriptional Target of Dot1a and a Regulator of Aqp2

Dot1l encodes histone H3 K79 methyltransferase Dot1a. Mice with Dot1l deficiency in renal Aqp2-expressing cells (Dot1lAC) develop polyuria by unknown mechanisms. Here, we report that Aqp5 links Dot1l deletion to polyuria through Aqp2. cDNA array analysis revealed and real-time RT-qPCR validated Aqp5 as the most upregulated gene in Dot1lAC vs. control mice. Aqp5 protein is barely detectable in controls, but robustly expressed in the Dot1lAC kidneys, where it colocalizes with Aqp2. The upregulation of Aqp5 is coupled with reduced association of Dot1a and H3 dimethyl K79 with specific subregions in Aqp5 5′ flanking region in Dot1lAC vs. control mice. In vitro studies in IMCD3, MLE-15 and 293Tcells using multiple approaches including real-time RT-qPCR, luciferase reporter assay, cell surface biotinylation assay, colocalization, and co-immunoprecipitation uncovered that Dot1a represses Aqp5. Human AQP5 interacts with AQP2 and impairs its cell surface localization. The AQP5/AQP2 complex partially resides in the ER/Golgi. Consistently, AQP5 is expressed in none of 15 normal controls, but in all of 17 kidney biopsies from patients with diabetic nephropathy. In the patients with diabetic nephropathy, AQP5 colocalizes with AQP2 in the perinuclear region and AQP5 expression is associated with impaired cellular H3 dimethyl K79. Taken together, these data for the first time identify Aqp5 as a Dot1a potential transcriptional target, and an Aqp2 binding partner and regulator, and suggest that the upregulated Aqp5 may contribute to polyuria, possibly by impairing Aqp2 membrane localization, in Dot1lAC mice and in patients with diabetic nephropathy.


Introduction
In addition to glucosuria, polyuria is the earliest clinical renal symptom in untreated or poorly controlled diabetes [1] and is not considered as a simple result of an osmotic diuresis due to the large solute load of urinary glucose [2,3]. However, the molecular mechanism(s) by which polyuria develops beyond glucosuria is not fully understood.
Aquaporins (AQPs) are members of the water channel family. Aqp1-4 are important for maintenance of normal urinary concentration and implicated in the renal water disorders [4][5][6][7]. Reduced expression and/or apical localization of Aqp2 under pathological conditions (i.e. nephrosis, hypokalemia, and Aqp2 mutations) results in polyuria. In contrast, nephrotic syndrome and congestive heart failure due to abnormal secretion of vasopressin increase apical Aqp2 levels, leading to excessive water reabsorption and hyponatremia (reviewed in [8]).
Aqp5 is expressed in eyes, salivary glands, lung and sweat glands [9][10][11]. A selective defect in lacrimal gland Aqp5 trafficking is responsible for Sjögren's syndrome characterized by dry eye and mouth [12]. While Aqp5 and Aqp2 are the closest homologs and share 66% sequence identity, Aqp5 is undetectable in normal mouse kidney by Northern analysis and immunoblotting (IB) [13].
Recently, we have reported generation of a Dot1l conditional knockout line using the LoxP-Cre system (Dot1l f/f ), which inactivate most of Dot1l function including the methyltransferase activity upon Cre-mediated recombination [23]. This line was used to generate connecting tube/collecting duct (CNT/CD)specific Dot1l-deficient (Dot1l f/f Aqp2:Cre or Dot1l AC ) mice by crossing them with Aqp2:Cre mice [29], which drive Cre recombinase expression under the control of regulatory elements of the mouse Aqp2 gene. Generation and characterization of Dot1l AC have been detailed in our recent manuscript [30].
Compared to Dot1l f/f controls, Dot1l AC mice have polyuria without severe impairment in maintaining normal electrolyte and acidbase balance [30]. In this report, we provide strong in vivo and in vitro evidence for the first time demonstrating that Dot1a downregulates Aqp5 and Aqp5 interacts with Aqp2 and impairs Aqp2 membrane localization. We also observed upregulated AQP5 and decreased H3m2K79 in kidney biopsies from patients with diabetic nephropathy (DN). The polyuria phenotype in Dot1l AC mice and in patients with DN may be partially attributable to upregulated Aqp5.

Results
Dot1l AC mice displayed polyuria without impaired Aqp2 expression Generation of Dot1l AC mice and description of their polyuria phenotype on a normal pellet Na + diet are detailed in our related manuscript [30]. Briefly, we used a Dot1l conditional knockout line (Dot1l f/f ) [23] and an Aqp2Cre line [29] to inactivate Dot1l and thus abolish histone H3 K79 methylation in Aqp2-expressing cells, which are located in the CNT/CD [30].
To further confirm the polyuria phenotype, we performed additional metabolic analysis. Dot1l AC vs. Dot1l f/f littermates after 24-h water deprivation (n = 14 mice/group) showed significantly increased normalized (to body weight) and slightly decreased urine osmolarity ( Figure 1A-B). Excretion of Na + and K + was 155616% and 146617% of Dot1l f/f mice, respectively, in Dot1l AC after the 24-h water deprivation. There were subtle, but not significant differences in all other urinary parameters ([Na + ], ) tested between the two groups ( Figure S1). The absolute urine volume was also significantly increased by 73%, 63% and 465% in Dot1l AC vs. Dot1l f/f mice in fed state, after 24-hour water deprivation, and after Streptozotocin (STZ)-induced diabetes, respectively ( Figure S2).

Microarray analysis identified Aqp5 as the most upregulated gene in Dot1l AC mice
To assess the effect of Dot1l inactivation on global gene expression and to identify the molecular defects leading to polyuria, we performed gene expression microarray analysis of Dot1l AC vs. Dot1l f/f mice (n = 4 mice/group), using the dual-color Agilent 4X44K Whole Mouse Genome Array system. With a minimal two-fold difference between the two genotypes as an arbitrary cut-off, we found 1359 up-and 627 down-regulated genes, respectively, in Dot1l AC vs. Dot1l f/f mice (Table S1 and S2). However, all of the known Aqps except Aqp5 had no or only subtle changes in their mRNA levels ( Figure 1C). Real-time RT-qPCR validated that the two genotypes were indistinguishable in the expression of Aqp2 and Aqp3, two important water channels in the Aqp2 + principal cells, where Dot1l deletion occurs ( Figure 1D). Immunoblotting analysis (IB) confirmed comparable Aqp2 protein levels between the two groups ( Figure 1E).
Aqp5 is the closest homolog of and adjacent to Aqp2 in the genome. Dot1l AC and Dot1l +/+ Aqp2Cre mice have two copies of the Aqp2Cre transgene, which carries the adjacent Aqp5 [29]. If Aqp5 expression were solely copy-number-dependent, Aqp5 would be 2fold higher in these animals than in Dot1l f/f mice. Unexpectedly, microarray analyses revealed Aqp5 as the most upregulated gene, with 26-fold higher mRNA level in Dot1l AC than in Dot1l f/f mice ( Figure 1C). Real-time RT-qPCR unearthed an even bigger difference (105-fold). Aqp5 mRNA level was also significantly increased in Dot1l +/+ Aqp2Cre vs. Dot1l f/f mice. However, such increase was much less prominent, compared to Dot1l AC mice ( Figure 2A). Subsequent agarose gel analysis revealed a single band from each of these RT-qPCR reactions ( Figure 2B). Sequencing analyses of a regular RT-PCR product from Dot1l f/f mice Aqp5 as a New Dot1a Target and Aqp2 Regulator PLOS ONE | www.plosone.org confirmed the identity of Aqp5 ( Figure 2C). This very low level of Aqp5 mRNA may explain why Aqp5 was undetectable by Northern and IB [13] and by immunofluorescence (IF) in normal mouse kidney ( Figure 2D).
Aqp5 is co-expressed with Aqp2 in Dot1l AC mice As shown in Figure 2D and Figure S3 and S4, IF with goat anti-Aqp2 (as the marker of PCs in which Dot1l deletion occurs) and rabbit anti-Aqp5 antibodies revealed that Dot1l f/f mice had robust Aqp2 and undetectable Aqp5 throughout the whole kidney. Very faint Aqp5 staining can be seen in Dot1l +/+ Aqp2Cre mice. Nevertheless, Aqp5 was readily detected and coexpressed with Aqp2 in most, but not all of the PC in Dot1l AC mice. The lack of Aqp5 in some of PC may result from absence of Cre expression [30], leaving Dot1l intact. Substantial Aqp5 was also visible in some of Aqp2 2 cells, which are most likely the intercalated cells derived from the canonical Aqp2-expressing cells as detailed in our related manuscript [30]. The affinity-purified anti-Aqp5 antibody was produced using a synthetic peptide corresponding to residues of 251-265 of rat Aqp5. This sequence is identical to mouse Aqp5 and has no significant homology to mouse Aqp2. This may explain why the Aqp5 antibody lacks the cross reactivity with mouse Aqp2 ( Figures 2D, S3, S4). The anti-Aqp5 antibody has been used to detect Aqp5 in different species including mouse [31] and duckling [32].

Aqp5 is expressed in developing Dot1l AC kidneys
To identify when Aqp5 is expressed during kidney development, we examined Aqp2 and Aqp5 co-expression in the kidneys from 3-, 11-, and 20-day old pubs. Like in adult Dot1l f/f mice, Aqp5 remained undetectable in all of the developing Dot1l f/f kidneys ( Figure S4). In Dot1l AC mice, Aqp5 was barely discernable at day 3, but became prominent at day 11 and day 20 in some Aqp2 + and Aqp2 2 connecting tube/collecting duct cells, as observed in the adult stage ( Figure S4). Representative IF images showing Aqp5 (green) expression in Aqp2 + (red) cells in mice as indicated. OM and IM: outer and inner medulla. *: An IC without Aqp5 expression, possibly due to lack of Aqp2Cre-mediated Dot1l ablation. Arrow: PC with strong Aqp2 and weak Aqp5, highlighting the lack of cross reactivity of the two antibodies. Aqp5 + Aqp2 2 cells are most likely the intercalated cells derived from the Aqp2-expressing progenitor cells or mature PC [30]. Arrowhead: Colocalization of Aqp5 with Aqp2, which is amplified in the inserts. Scale bar: 50 mM. For more images with lower magnification, see Figure S3. doi:10.1371/journal.pone.0053342.g002 Aqp5 as a New Dot1a Target and Aqp2 Regulator PLOS ONE | www.plosone.org AQP5 is pathologically expressed and colocalizes with AQP2 at the perinuclear region in patients with DN While polyuria in DN is clearly due to glucosuria, we hypothesize that abnormal co-expression of AQP5 with AQP2 occurs in patients with DN, which may make the polyuria phenotype even more prominent. Accordingly, we investigated AQP5 and AQP2 expression by IF in kidney biopsies from 17 patients with DN and 15 with minimal change disease (MCD). Since these MCD samples had no significant pathological changes in the tubules, as revealed by Hematoxylin-Eosin staining ( Figure 3A), they were considered as ''normal'' controls. All of the MCDs had undetectable AQP5, with AQP2 primarily seen at the apical side ( Figure 3A). Unlike MCDs, all DN samples showed various tubular abnormalities including dilated lumen, epithelial thinning, nuclear irregularity, eosinophilic cytoplasm, severe tubular basement membrane thickening, interstitial fibrosis, and mononuclear cell infiltrate ( Figure 3, B-D). We focused on the tubules containing at least one AQP5 + or AQP2 + cell. The tubules containing only AQP2 + -AQP5 + -or double positive cells were categorized as AQP2 + AQP5 2 , AQP2 2 AQP5 + , and AQ-P2 + AQP5 + , respectively. The AQP2 2 AQP5 2 tubules were excluded from analyses because of the uncertainty of their identities. In the AQP2 + AQP5 + tubules, these two proteins were typically colocalized as large discrete foci near the perinuclear region ( Figure 3, B & C). In AQP2 2 AQP5 + tubules, however, AQP5 distribution was broader and more homogenous ( Figure 3D). The detection of AQP2 + AQP5 2 and AQP2 2 AQP5 + tubules/cells demonstrated the specificity of each antibody used.
To evaluate the relative abundance of the three types of tubules, we counted their numbers in each field containing at least one AQP5 + or AQP2 + cell. Among 135 images from the 17 DN samples, 587 tubules were counted, yielding 79.6% AQ-P2 + AQP5 + , 16% AQP2 + AQP5 2 , and 6.4% AQP2 2 AQP5 + . A single image containing 12 AQP2 + AQP5 + tubules is shown in Figure 3B. Our data for the first time strongly suggest that AQP5 is pathologically expressed in renal PC and may inhibit AQP2 membrane localization by ''trapping'' it near the perinuclear region in DN patients ( Figure 3B and 3C).

AQP5 expression is inversely correlated with H3m2K79 in DN patients
To test the hypothesis that the pathological expression of AQP5 is associated with loss of H3m2K79 in DN patients, similar double IF was conducted to examine coexpression of AQP5 and H3m2K79. We chose H3m2K79 as an indicator of DOT1L function because all available antibodies against DOT1L failed to detect the target protein specifically by IF ( [33] and data not shown) and Dot1l is solely responsible for H3m2K79 in mouse kidney [30]. Moreover, H3m2K79 was used to assess inactivation of Dot1l function and thus disrupted Dot1l expression in the Dot1ldeficient embryos [18], in the heart of the cardiac-specific Dot1l knockout mice [21] or in the peripheral blood nucleated cells of Dot1l f/f Vav-Cre mice [34].
All MCDs had strong H3m2K79 staining and apparently no AQP5 throughout the biopsies. However, all of the DN samples significantly reduced H3m2K79 labeling throughout the whole biopsies in some cases or focally in others. Most of the AQP5 + cells displayed weak or no H3m2K79 labeling at all in their nuclei ( Figure 4, A-D). Within the tubules containing AQP5 + H3m2K79 2 cells, some cells apparently were negative for both antibodies (AQP5 2 H3m2K79 2 ). This suggests that loss of H3m2K79 is not always associated with ''de-silencing'' AQP5.
To more accurately assess the inverse correlation between AQP5 and H3m2K79, we measured their pixel intensities in 256 cells from all of the DN samples (see Materials and Methods). The data were best fit with a curve Y = 133.122.221X+0.021X 2 22.16X 3 , which yielded a determination coefficient (R 2 ) = 0.749 ( Figure 4E). Therefore, our data strongly suggest an inverse correlation between the two parameters in DN patients, that MCDs express high H3m2K79 without detectable AQP5 labeling, and that the undetectable H3m2K79 is not due to a general defect in protein synthesis and/or stability since robust Aqp5 expression can occur in the same cells in the patients with DN.
Dot1l AC vs. Dot1l f/f mice displayed impaired Dot1a binding and reduced H3m2K79 at the Aqp5 59 regulatory region To determine if Dot1l regulates Aqp5 transcription through modulation of H3m2K79 at the Aqp5 59 flanking region, chromatin immunoprecipitation (ChIP) assay coupled with realtime qPCR was pursued. We focused on the 6.5 kb region that spans from the very end of the last exon (exon 4) of Aqp2 to the translation start site ATG of Aqp5. The whole region was divided into 12 subregions named A-L ( Figure 5, A-B). ChIP with anti-Dot1l or anti-H3m2K79 detected substantial signals in F, I, K and L, with the strongest binding in F in Dot1l f/f mice. Such signals were significantly attenuated, but still detectable in Dot1l AC mice, possibly due to the fact that deletion of Dot1l occurs only in Aqp2expressing cells, rather than in all cells in the kidney. All other subregions showed no or weak ChIP signals in both genotypes. These observations collectively reinforce the specificity of Dot1l and H3m2K79 antibodies that have been used by others and us [24,27,28,35,36]. Our data also suggest that the association of Dot1a and H3m2K79 with some, but not all of the subregions is not simply due to non-specific DNA binding activity as reported for hDOTL1  in vitro [37]. Furthermore, ChIP with normal rabbit IgG yielded barely detectable background signals (data not shown). These data suggest that Dot1l downregulates Aqp5 mRNA expression, possibly by modulating H3m2K79 at the Aqp5 promoter.

Dot1a decreases Aqp5 expression in M1 and MLE-15 cells
To further evaluate whether Dot1a inhibits Aqp5 promoter activity in cultured cells, mouse cortical collecting duct M1 cells were transfected with two Dot1a-specific siRNAs: siRNA#1 and siRNA#2. Real-time RT-qPCR revealed that cells transfected with these two siRNAs increased Aqp5 mRNA to 964% and 1014% of control ( Figure 6A). Since M1 cells have a very low basal level of Aqp5 expression, overexpression of Dot1a apparently further reduced Aqp5 mRNA abundance. This was envisioned by non-specific rather than specific amplification, which made the real-time RT-qPCR measurements unreliable. Accordingly, we chose mouse lung epithelial MLE-15 cells that robustly express Aqp5 to more accurately investigate the effect of Dot1a overexpression. Dot1a-overexpressing vs. control cells showed significantly decreased expression of endogenous Aqp5 and activity of a luciferase reporter driven by a 4.3-kb upstream fragment of rat Aqp5 [38]. Luciferase activity was progressively decreased with increasing amounts of Dot1a constructs ( Figure 6, B-D). Sequential deletion analysis revealed that the fragment beginning at 2385 bp was apparently capable of and sufficient for Dot1a-mediated downregulation ( Figure 6E). However, the smallest construct starting at 2139 displayed just background luciferase activity with or without Dot1a overexpression. Therefore, Dot1a attenuates Aqp5 expression, possibly by direct or indirect binding to the Aqp5 promoter and modulating H3m2K79 associated with the Aqp5 promoter in M1 and MLE-15 cells and in mouse kidney.

AQP5 coimmunoprecipitates with AQP2
Colocalization of Aqp5 with Aqp2 in Dot1l AC kidney and in patients with DN suggests interactions between these proteins. To solidify this finding, we conducted co-immunopreciptation assays with green fluorescence protein (GFP)-AQP2 and FLAG-AQP5 expressed in 293T cells. FLAG-AQP5 was immunoprecipitated from its transfectants, but not from FLAG Vec-transfectants, confirming specificity of the FLAG antibody used for immunoprecipitation ( Figure 7A). Co-immunoprecipitation of GFP-AQP2 and FLAG-AQP5 occurred when they were co-expressed. Replacing one of them with the corresponding vector abolished the coimmunoprecipitation (Figure 7, A-B). In all cases, GFP-AQP2 and FLAG-AQP5 fusions were expressed at comparable levels among transfections (Figure 7, C-D).

AQP5 impairs AQP2 cell surface expression
Colocalization of AQP5 with AQP2 at the perinuclear region in patients with DN suggests that AQP5 may reduce AQP2 cell surface abundance. To test this hypothesis, we performed cell surface biotinylation assays using FLAG-AQP2 and Myc-AQP5 expressed in IMCD3 cells. Compared with Myc vector, Myc-AQP5 significantly decreased plasma membrane-associated FLAG-AQP2 to about 44% ( Figure 7E). When biotin labeling was omitted, plasma membrane-associated FLAG-AQP2 became undetectable, confirming the specificity of biotinylation ( Figure 7E).

AQP5 and AQP2 colocalize and partially reside in the ER/ Golgi in IMCD3 cells
Like in patients with DN ( Figure 3, B & C), colocalization of AQP2 and AQP5 as large discrete foci in the perinuclear region was also observed when they were co-expressed as GFP and red fluorescence protein (RFP) fusions in IMCD3 cells and examined by confocal microscopy. The colocalization was apparently not a result of overexpression because neither GFP-AQP2 nor RFP-AQP5 colocalized with overexpressed RFP or GFP alone, Aqp5 as a New Dot1a Target and Aqp2 Regulator PLOS ONE | www.plosone.org respectively (Figure 8, A-C). Furthermore, over 95% of transfected cells were doubly transfected. GFP-AQP2 seldom formed large discrete foci in the absence of RFP-AQP5. Based on these two observations, we wanted to determine if AQP2/AQP5 complexes are localized in the ER/Golgi. IMCD3 cells were co-transfected with GFP-AQP2, Myc-AQP5 and DsRed-ER as endoplasmic reticulum marker [39]. Alternatively, the cells were transfected without DsRed-ER and stained with an antibody specific for the cis-Golgi marker GM130. Confocal microscopy revealed that GFP-AQP2/Myc-AQP5 complex evidenced by large discrete foci partially colocalized with the markers, suggesting that some of the complexes were targeted to and/or retained in ER and cis-Golgi compartments (Figure 8, D & E).

Discussion
In this report, we uncover a new mechanism by which Dot1l regulates water homeostasis, and identify Aqp5 as a potential novel target of Dot1a, a binding partner of Aqp2, a negative trafficking regulator of Aqp2, and thus the potential missing component linking disrupted Dot1l to polyuria.
We provide evidence showing that Aqp5 is a new Dot1aregulated gene. Genetic inactivation, and siRNA-mediated depletion of Dot1l led to endogenous Aqp5 upregulation. The reverse correlation of AQP5 expression with H3m2K79 in DN further suggests the importance of H3m2K79 in this regulation. In contrast, Dot1a overexpression inhibited expression of endogenous Aqp5 and all but the smallest Aqp5 promoter-luciferase constructs. These results suggest that exogenously expressed constructs and the endogenous Aqp5 responded similarly to changes in Dot1a abundance. Unlike the 2389 construct, the 2139 construct apparently lacks Dot1a-mediated repression and basal promoter activity, indicating the existence of a critical ''Dot1a-responsible'' cis-element and elements essential for the minimal promoter activity between 2389 and 2139 region.
TNFa and lipopolysaccharide inhibit Aqp5 expression in salivary and parotid gland cells, respectively [40,41]. TNFamediated repression involves suppression of histone H4 acetylation [40]. Lipopolysaccharide enhances NF-kB, but not AP-1 binding at the Aqp5 promoter [41]. Interestingly, the AP-1 and NF-kB sites reside in subregions K and L, which were shown to bind Dot1l and H3m2K79 in ChIP assays ( Figure 5). Therefore, it would be interesting to know how the binding of these transcription factors is coordinately regulated in the Aqp5 promoter, and to investigate if the 4 subregions (F, I, K and L) are critical to an insulator element between Aqp2 and Aqp5. Dot1a may also indirectly inhibit Aqp5 expression by upregulating another factor that represses Aqp5. In either case, upregulation of Aqp5 may contribute to the polyuria phenotype in Dot1l AC mice and in DN patients.
Aqp5 promotes outflow of water in the lacrimal and salivary glands. Since Aqp5-mediated water transport in secretory epithelia occurs in the secretory direction in response to an osmotic gradient, water transport via Aqp5 in kidney would also move down an osmotic gradient. Thus, like Aqp2, Aqp5 would be expected to mediate water uptake. Obviously, the polyuria phenotype does not support this hypothesis. Identification of Aqp5 as a negative regulator of Aqp2 apical localization offers a clue for an opposite role.
In contrast, the importance of the interaction between WT AQP2 and some AQP2 mutants is clearly demonstrated in autosomal dominant nephrogenic diabetes insipidus (NDI). All dominant mutations identified so far occur within the C-terminus of AQP2. The AQP2 mutants form heterotetramers with WT AQP2 and thus prevent WT AQP2 from reaching the plasma membrane. The WT AQP2 is retained either in the Golgi, late endosomes, lysosomes, or the basolateral plasma membrane [50]. In several aspects, AQP5 seems to behave like the AQP2 mutants. First, the high sequence homology with 66% amino acid identity offers a potential molecular basis for AQP2-AQP5 interaction. Such interaction may interfere with their homotetramer formation and facilitate hybrid arrangement. Secondly, the C-termini of AQP2 and AQP5 are important for trafficking, but not for tetramer formation [51]. They are not conserved between the two proteins, mimicking the difference between the WT and mutant AQP2 forms.
Our finding that AQP5 is readily detectable and colocalizes with AQP2 at the perinuclear region suggests that the apical localization of AQP2 may be impaired and contributory to the deterioration of glucosuria-induced polyuria in patients with DN. Earlier studies show that compensatory Aqp2 protein abundance increases in rats with streptozotocin-induced diabetic mellitus [52][53][54], which apparently differs from the present results. All of these animals studies were conducted in a relatively short period ranging from day 5 to day 21 following STZ treatment [52][53][54]. While diabetic mellitus was clearly induced, no pathological data were shown in any of these studies to demonstrate tubular abnormalities as seen in the patients with DN, raising the possibility that renal pathological changes in these diabetic rats, particular in terms of tubular aberration, are not comparable with those in the patients with DN. It is likely that the increased Aqp2 expression is a result of compensatory mechanism in the early stages of the diseases progression. Such a compensatory mechanism may be lost in the advanced stage of DN development.
As an aldosterone downregulated gene [24,27,28], Dot1l may also links excessive aldosterone to polyuria via Aqp5 and Aqp2. Long-term aldosterone infusion decreases Aqp2 apical localization and causes polyuria in several animal models [36,55,56]. It is worthy to determine if Dot1a is really downregulated and Aqp5 upregulated in these experimental conditions.
Although our data strongly support the formation and localization of AQP2/AQP5 complex in the ER/Golgi as the potential underlying mechanism by which Aqp5 plays an inhibitory role in Aqp2 membrane localization and thus in the development of polyuria, conclusive demonstration requires blocking Aqp5 in Dot1l AC mice. Genetic inactivation using Aqp5 2/2 mice, pharmacological blocking with Aqp5-specific inhibitors, and siRNA-mediated in vivo depletion of Aqp5 offer three different strategies. To our knowledge, there are currently no Aqp5-specific inhibitors. Since extensive filtration and uptake of siRNA take place in the proximal tubule cells, but not in the distal Chromatin from Dot1l f/f and Dot1l AC mice on the normal Na + pellet diet (n = 6 mice/group) was immunoprecipitated by the rabbit antibodies specific for Dot1l (A) and H3m2K79 (B), followed by real-time qPCR with primers amplifying subregions A-L covering almost the whole region between the end of Aqp2 exon 4 and the translation start site ATG of Aqp5 as shown at the bottom. There is a 525-bp gap between J and K, which cannot be reliably amplified and was excluded from further analysis. Relative ChIP efficiency was defined as the immunoprecipitated amount of materials present as compared to that of the initial input sample. *: P,0.05 vs. Dot1l f/f within the same subregion. doi:10.1371/journal.pone.0053342.g005 Aqp5 as a New Dot1a Target and Aqp2 Regulator PLOS ONE | www.plosone.org tubule cells [57], where Dot1l-deletion-dependent Aqp5 expression occurs, siRNA-mediated in vivo knockdown of Aqp5 may not be effective.

Ethics Statement
All animal studies were performed in accordance with NIH Guides for the Care and Use of Laboratory Animals and were approved by the University of the University of Texas Health Science Center at Houston Animal Welfare Committee. For kidney biopsy studies, approval of the protocol (11-0144) was obtained from the University of Texas Health Science Center at Houston Institutional Review Board. No participants were specifically recruited for this project to collect the kidney biopsy samples. Kidney archival biopsy residual specimens from patients with the diagnosis of DN and patients with MCD were exclusively used. The Institutional Review Board determined that this project is qualified for exempt status according to 45 CFR 46.101(b), CATEGORY #4: Research, involving the collection or study of existing data, documents, records, pathological specimens, or diagnostic specimens, if these sources are publicly available or if the information is recorded by the investigator in such a manner that subjects cannot be identified directly or through identifiers linked to the subjects. Accordingly, the written or verbal informed consent of the ''participants'' to participate in this study was waived. The Department of Pathology and Laboratory Medicine, University of Texas Medical School at Houston provided the PI with these archival kidney biopsy samples as unidentified materials. MCDs without obvious tubular abnormalities were selected and used to serve as ''normal'' controls.

STZ treatment
Two-month-old male Dot1l f/f and Dot1l AC mice received 5 consecutive daily intraperitoneal injections of STZ-Na-Citrate solution (50 mg/kg), according to the Low-Dose Streptozotocin Induction Protocol (Mouse) established by Animal Models of Diabetic Complications Consortium (http://www.diacomp.org/ shared/protocols.aspx?model = 9). Blood glucose was monitored with a glucometer, using a one-drop blood sample collected through the tip of the tail. Once the blood glucose level reaches .200 mg/dl, mice were considered as diabetic and maintained for additional three months. Detailed characterization of diabetic Dot1l f/f and Dot1l AC mice will be reported elsewhere.

Metabolic balance studies
All experimental mice were 2-5 months old, and sex-and agematched. After acclimation for 3-7 days to Tecniplast metabolic cages (Exton) with free access to water and normal Na + (0.4%) diet (fed state), mice were subsequently subject to water deprivation for 24 h. Twenty-four-hour urine samples were collected and analyzed as we reported before [36]. In brief, twenty-four-hour urine in fed state was collected daily for at least three consequtive days. For each mouse, the urine data from multiple days were pooled to calculate the final average to represent that mouse and counted as 1 (n = 1). To minimize circadian effects, urine collection were conducted around 5:00 p.m. each day. For water deripvation studies, water deprivation and urine collection were conducted samultaneiously for 24 hours. For STZ-induced diabetic mice, twenty-four-hour urine collection was conducted in three consequtive days at the end of the experiment (three months after blood glucose reached .200 mg/dl). To ensure complete and accurate urine collection, bladder voiding was conducted before the collection started and urine discarded. At the end of collection time point, bladder voiding was repeated and the urine collected. To minimize vaporization and thus urine volume loss, a fixed amount of corn oil was added to each of the containers so that the collected urine samples were covered by the oil during the collection period.

Urine measurements
Urinary [Na + ], [K + ], and [creatinine] were measured with an analyzer (Roche Cobas Integra 400 plus) in the Clinical Pathology Laboratory, Department of Veterinary Medicine and Surgery, University of Texas MD Anderson Cancer Center as described [36]. Urine osmolarity was determined by vapor pressure (Wescor Vapro Vapor Pressure Osmometer 5520, Scimetrics, Houston, TX, USA) [36].

Microarray analysis
Microarray experiments were carried out using the dual-color Agilent 4X44K ''Whole Mouse Genome Array'' system (Agilent) and the manufacturer's protocol followed. Detailed protocol can be found at Agilent website: www.Agilent.com. Briefly, total kidney RNA of Dot1l f/f and Dot1l AC mice (n = 4 mice/genotype) was isolated using Trizol (Invitrogen). Equal amounts of total RNA from each mouse of the same genotype were mixed to represent the corresponding genotype. 200 ng of the mixed RNAs was used and labeled with either Cy3-or Cy5-CTP. After 17 hr hybridization at 65uC, the arrays were washed and scanned with Agilent's dual-laser based scanner. Then, Feature Extraction software GE2-v5_95 was used to link a feature to a design file and determine the relative fluorescence intensity between the two samples. Genelists were created using pValue information from the internal replicates within the microarray. Data were deposited to Gene Expression Omnibus, with access number GSE40090.

Immunofluorescence and confocal staining
Mouse kidney tissues were fixed in 4% paraformaldehyde overnight at 4uC and embedded in paraffin. After boiling in antigen-retrieval buffer, paraffin sections were blocked with 5% BSA/0.5% Triton X-100 in PBS. Each primary antibody was diluted in 5% BSA/0.5% Triton X-100 in PBS and incubated overnight at 4uC. Following four 5-minute washes in PBS, the sections were incubated with Dylight 594-AffiniPure goat antichicken IgG (Jackson ImmunoResearch LABORATORIES, Inc.), Alexa Fluor 488-conjugated goat anti-rabbit IgG or Alexa 594conjugated goat anti-mouse IgG (Invitrogen), depending on the Aqp5 as a New Dot1a Target and Aqp2 Regulator PLOS ONE | www.plosone.org species from which the primary antibody was generated. Nuclei were visualized using DAPI (1:1000). The sections were mounted in VECTASHIELD HardSet Mounting Medium (H-1400, VECTOR LABROTORIES), and examined under an epifluorescence microscope (Olympus IX71) and confocal microscope (510 Meta, Zeiss LSM) as previously described [26,58].
For in vitro experiments, transfected IMCD3 cells with the plasmids as indicated in figure legends were fixed with 4% fresh prepared paraformaldehyde for 20 min, washed with PBS 365 min, and incubated in permealization buffer (0.1% Triton 6100, 5% Glycine in PBS) for 15 min. After 365 min washing with PBS, cells were blocked with Serum-free Protein Block (Dako) for 1 hr. Incubation of the primary and secondary antibodies, DAPI staining, mounting and microscopic examination were conducted as described above.

Quantification of pixel intensity of AQP5 and H3m2K79 immunofluorescence staining in patients with DN
Kidney biopsies were immunostained under identical conditions. All digital images were acquired using the same settings including the same exposure time and amplification. Tubules with at least one AQP5 + cell were included for this analysis. All other AQP5 2 tubules were excluded from the analyses. Among AQP5 + tubules, only AQP5 2 H3m2K79 + , AQP5 + H3m2K79 2 , and AQP5 + H3m2K79 + cells were considered. AQP5 2 H3m2K79 2 cells were excluded as we already pointed out that loss of H3m2K79 might not always be associated with AQP5 expression. Spots of positive AQP5 staining without a clear DAPI-stained nucleus were excluded due to uncertainty of their H3m2K79 status. Under these restricted criteria, the pixel intensities over background of both AQP5 and H3m2K79 staining in the selected cells were quantified using ImageJ64 software. , lack of colocalization when GFP-AQP2 (B) and RFP-AQP5 (C) was coexpressed with RFP and GFP alone, respectively. (D) Representative confocal microscopy images showing AQP5-AQP2 complex partially locates in the ER in IMCD3 cells. IMCD3 cells were co-transfected with the plasmids as indicated and examined by confocal microscopy. DsRed-ER was used as an endoplasmic reticulum (ER) marker [39] and shown in red. GFP-AQP2 is shown in green. Myc-AQP5 was not directly detected, but can be inferred by its interaction and colocalization with AQP2 as large discrete foci seen in A. AQP2 alone seldom forms large discrete foci. In addition, more than 95% of transfected IMCD3 cells were always doubly transfected. Therefore, the presence of the large discrete foci is indicative of AQP2-AQP5 complex. (E) As in D except that DsRed-ER was omitted and cells were stained with an antibody specific for GM130, a cis-Golgi marker. Scale bar: 40 mM. doi:10.1371/journal.pone.0053342.g008 Cell surface biotinylation assay This assay was done using the cell surface protein isolation kit (Pierce), following the manufacturer's instructions. Briefly, transiently transfected IMCD3 cells were biotinylated with Sulfo-NHS-SS-Biotin in cold PBS at 4uC for 30 min and the reaction stopped by addition of Quenching solution. Cells were washed with cold TBS and lysed with Lysis buffer containing proteinase inhibitor cocktail at 4uC for 30 min. Cell lysates were centrifuged for 2 min at 10,000 g. Supernatants were incubated with NeutrAvitin Agarose for 60 min at room temperature. Beads were washed four times with Wash buffer. Bound proteins were eluted by heating for 5 min at 95uC in 26 SDS-PAGE loading buffer (2% SDS, 0.0625 M Tris pH 6.8, 20% glycerol, 0.01% bromophenol blue, 5% b-mercaptoethanol). Samples were analyzed by immunoblotting with antibodies as indicated in the figure legends.
Immunoblotting, luciferase assay, real-time RT-qPCR, and ChIP assay These assays were conducted according to our published protocols [24,25,28,36], For ChIP, a modified protocol described in our recent publication [33] was utilized. The sequences of the primers were listed in Table S3 and S4.

Statistical analysis
Quantitative data are presented as mean6SEM. Student's t-test was used with the statistical significance set at P,0.05. Figure S1 Additional urine metabolic analyses of Do-t1l AC and control mice after water deprivation for 24 h. Dot1l f/f (f/f) and Dot1l AC (AC) mice were fed the normal Na + diet (0.4% Na + ) in metabolic cages, deprived of water for 24 h, and analyzed for the parameters as indicated. n = 14 mice/group. *P,0.05 vs. Dot1l f/f . (DOC) Figure S2 Absolute urine volumes. Dot1l f/f (f/f) and Dot1l AC (AC) mice with free access to water and regular diet (A), after 24-hr water deprivation (B), and after blood glucose reaching 200 mg/dl induced by STZ injection (C) were analyzed for the absolute urine volume. In each case, n = 4-14 mice/genotype. *P,0.05 vs. Dot1l f/f . (DOC) Figure S3 Additional IF images showing that Aqp5 was significantly upregulated in the kidney of Dot1l AC mice on the normal Na + pellet diet. (A-B) Representative IF images showing Aqp5 (green) expression in Aqp2 + (red) cells in mice as indicated. Note: Some cells displayed Aqp5 + Aqp2 2 phenotype. These cells are most likely the intercalated cells derived from the Aqp2-expressing progenitor cells or mature PC [30]. OM and IM: outer and inner medulla. Detection of Aqp5 + Aqp2 2 and Aqp5 2 Aqp2 + cells demonstrates the specificity of the two antibodies. Scale bar: 50 mM. (DOC) Figure S4 Aqp5 is expressed in the developing Dot1l AC kidneys. Representative IF images showing detectable Aqp5 in some Aqp2 + and Aqp2 2 connecting tube/collecting duct cells of Dot1l AC mice at day 11 and day 20, but not at day 3. Aqp5 is not detectable in Dot1l f/f mice at all stages as indicated. Scale bar: 100 mM.

(DOC)
Table S1 Upregulated Genes in Dot1l AC vs. Dot1l f/f mice. Total kidney RNA of Dot1l f/f and Dot1l AC mice (n = 4 mice/genotype) was subjected to microarray analyses. There were 1359 genes represented by 1423 unique probes. These genes were upregulated with at least $2-fold higher mRNA levels in Dot1l AC vs. Dot1l f/f mice. (DOC)  Table S1, total kidney RNA of Dot1l f/f and Dot1l AC mice (n = 4 mice/genotype) was subjected to microarray analyses. There 627 genes represented by 680 unique probes. These genes were downregulated with at least $2-fold lower mRNA levels in Dot1l AC vs. Dot1l f/f mice. (DOC)