Serine/threonine phosphatases and aquaporin-2 regulation in renal collecting duct

Abstract

Phosphorylation of the aquaporin-2 (AQP2) water channel at four COOH-terminal serines plays a central role in the regulation of water permeability of the renal collecting duct. The level of phosphorylation at these sites is determined by a balance between phosphorylation by protein kinases and dephosphorylation by phosphatases. The phosphatases that dephosphorylate AQP2 have not been identified. Here, we use large-scale data integration techniques to identify serine-threonine phosphatases likely to interact with AQP2 in renal collecting duct principal cells. As a first step, we have created a comprehensive list of 38 S/T phosphatase catalytic subunits present in the mammalian genome. Then we used Bayes’ theorem to integrate available information from large-scale data sets from proteomic and transcriptomic studies to rank the known S/T phosphatases with regard to the likelihood that they interact with AQP2 in renal collecting duct cells. To broaden the analysis, we have generated new proteomic data (LC-MS/MS) identifying 4538 distinct proteins including 22 S/T phosphatases in cytoplasmic fractions from native inner medullary collecting duct cells from rats. The official gene symbols corresponding to the top-ranked phosphatases (common names in parentheses) were: Ppp1cb (PP1-β), Ppm1g (PP2C), Ppp1ca (PP1-α), Ppp3ca (PP2-B or calcineurin), Ppp2ca (PP2A-α), Ppp1cc (PP1-γ), Ppp2cb (PP2A-β), Ppp6c (PP6C), and Ppp5c (PP5). This ranking correlates well with results of prior reductionist studies of ion and water channels in renal collecting duct cells.

aquaporin-2 (AQP2) is an integral membrane protein that allows water molecules to cross the apical plasma membrane of collecting duct cells in the renal tubule of the mammalian kidney (22). In the kidney, its regulation is mainly controlled by the peptide hormone vasopressin, which is synthesized in the hypothalamus, then stored and secreted from the posterior pituitary. Vasopressin controls AQP2 in at least two ways (22): regulation of AQP2 levels in the plasma membrane via trafficking of AQP2 containing membrane vesicles, and control of the overall abundance of AQP2 present in the principal cells by regulating transcription of the Aqp2 gene.

A common process by which transporters are regulated involves the control of the level of phosphorylation. The putting on and taking off of phosphates at regulatory sites is performed, respectively, by protein kinases, which add phosphate groups, and protein phosphatases, which remove them. The overall level of phosphorylation, then, depends on the net activity of these two types of enzymes. There are at least four vasopressin-regulated phosphorylation sites in the AQP2 protein at Serine residues 256, 261, 264, and 269 (7, 11). Phosphorylation at these sites governs the trafficking of AQP2 into and out of the collecting duct apical plasma membrane (2, 12, 14, 21, 28). The phosphatases that reverse phosphorylation at these sites belong to the Serine/Threonine (S/T) phosphatase subgroup. Our goal here is to identify which of the serine/threonine (S/T) protein phosphatases coded by the mammalian genome are most likely to interact with AQP2 in the renal collecting duct. This analysis uses prior transcriptomic and proteomic data from inner medullary collecting ducts (IMCDs) isolated from rat kidneys (31), from cultured mouse mpkCCD cells (a vasopressin-responsive cell culture model of cortical collecting duct) (19, 35), from microdissected rat collecting duct segments (16), as well as new data presented in this paper profiling the cytoplasmic proteome of rat IMCDs.

The present paper has three aims: 1) to curate a list of S/T phosphatase catalytic subunits present in the mammalian genome and make the information available via a publicly accessible webpage; 2) to use Bayes’ rule to integrate available information from large-scale data sets to predict what S/T phosphatases may interact with AQP2 in renal collecting duct cells; and 3) to relate the phosphatase predictions made via Bayesian analysis to the reductionist literature on regulation of transport in collecting duct principal cells.

METHODS

Assembly of a Database of Mammalian Protein Phosphatase Catalytic Subunits

We assembled a list of the known serine/threonine (S/T) phosphatase catalytic subunits in human, mouse, and rat genomes. The original list was taken from DEPOD (the human DEPhOsphorylation Database at http://www.koehn.embl.de/depod/), and the amino acid sequences of the catalytic regions of representative members of each major family were used in multiple BLAST searches of proteomic (https://hpcwebapps.cit.nih.gov/ESBL/Database/IMCD_Proteome/index.html) and transcriptomic data (https://hpcwebapps.cit.nih.gov/ESBL/Database/Transcriptomic/IMCDdatabase.html) to identify other members relevant to collecting duct transport. Human gene symbols were converted to mouse symbols using the Automated Bioinformatics Extractor (ABE) tool (https://hpcwebapps.cit.nih.gov/ESBL/ABE/). The data were curated on an Excel spreadsheet, which was converted to an .html file, edited with Notepad⁺⁺, and posted on a permanent web server to make it available for public access at https://hpcwebapps.cit.nih.gov/ESBL/Database/Phosphatases/. This database was used as a beginning point in the identification of the S/T phosphatases most likely to interact with AQP2 in the renal collecting duct. The database excludes regulatory subunits.

Use of Bayes’ Theorem to Predict AQP2-Phosphatase Interactions

To determine the S/T phosphatases most likely to dephosphorylate AQP2 at any of its four vasopressin-regulated phosphorylation sites, we ranked the S/T phosphatases using Bayes’ theorem (1, 20) as illustrated in Fig. 1A. Here the prior probability vector, P(A), and the vector of likelihoods from any new data set, P(B|A), are used to calculate a new posterior probability vector P(A|B). The term P(B) is calculated as the scalar sum of probabilities of B over all A. For the calculations in this study, the dimension of the vectors is 38, reflecting all 38 S/T phosphatase catalytic subunits found in mammalian genomes. As shown in Fig. 1B, the application of each step using Bayes’ rule can be represented as a so-called "Bayesian operator" that receives two vectors of equal length as inputs and produces a new posterior probability vector of the same length. This operator can be applied multiple times in series to integrate multiple data sets. The posterior probability vector from each step is used as the prior probability vector for the next. To initiate the process on the first step, we set all values for the prior probability vector to 1/38, i.e., a uniform probability distribution over all S/T phosphatases.

Fig. 1.

Bayes’ rule is used to integrate multiple data sets. A: Bayes’ rule integrates 2 probability vectors (dimension = 38 for family of S/T phosphatases): a prior probability vector and a “likelihood vector” derived from values in a particular large-scale data set. The result is a single posterior probability vector that can be used as a new prior probability vector to integrate the next data set. B: the application of Bayes’ rule in this setting can be viewed as application of a “Bayesian operator” that converts 2 equal-length vectors to obtain a new vector of the same length. Use of this operator can be repeated multiple times to integrate different data sets into the overall calculation, using the posterior probability vector of 1 step as the prior probability vector of the next step.

Table 1 lists the data sets used and the methods for estimating the likelihood vector for each. Where possible we use the complement of the minimum Bayes’ factor as described by Goodman (10) to estimate likelihood values from quantitative data. Table 1 also includes the rationale for use of each data set. The rationale for use of data sets 1, 3, and 4 (Table 1) is that for a phosphatase to interact with AQP2 in collecting duct cells, its gene must be expressed. The assignment of likelihood values recognizes that "not found" does not necessarily mean "not expressed," so no absolute zero likelihood values are assigned for any S/T phosphatase. The rationale for application of data set 2 (Table 1) is “If a phosphatase plays a critical role in regulation of AQP2, we expect it to be expressed in all collecting duct segments.” This is simply application of the principle of Ockham’s razor or the Law of Parsimony (https://en.wikipedia.org/wiki/Occam's_razor). Specifically, when one considers the ways that AQP2 phosphorylation can be regulated, the simplest view is that the same phosphatases are involved in all AQP2-expressing renal tubule segments. The general rationale for data sets 5–8 (Table 1) is that for a phosphatase to interact with AQP2, it must be present in the same subcellular compartments as AQP2. This question is addressed in different ways using data from proteomic analysis of differential centrifugation fractions in mpkCCD cells (database 5), data from surface biotinylation in mpkCCD cells (database 6), proteomic analysis of cytoplasmic fractions from mpkCCD cells (database 7), and proteomic analysis of an IMCD cytoplasmic fraction (database 8). In this paper, we use the term “cytoplasmic fraction” to refer to all nonnuclear elements of the cells including membrane components, cytosol, and organelles. Since AQP2 is localized to the cytoplasmic fraction, we argue that interacting proteins must also reside in the cytoplasmic fraction.

Table 1.

Data sets used in Bayes' calculations in this study

Data Set	Rationale	Assignment of Likelihood Values: P(B|A)	Noise Threshold	URL
Rat IMCD transcriptome (data set 1)	The ability of a phosphatase to dephosphorylate aquaporin-2 depends on whether or not it is expressed in collecting duct cells.	Complements of minimal Bayes' factors (z* = value/noise)	0.4	https://esbl.nhlbi.nih.gov/IMCD-transcriptome/
Rat RNA-Seq (data set 2)	If a phosphatase plays a critical role in regulation of aquaporin-2, we expected it to be expressed in all collecting duct segments.	Complements of minimal Bayes' factors derived from χ2 statistic		https://hpcwebapps.cit.nih.gov/ESBL/Database/NephronRNAseq/All_transcripts.html
Mouse mpkCCD transcriptome (data set 3)	The ability of a phosphatase to dephosphorylate aquaporin-2 depends on whether or not it is expressed in collecting duct cells.	Complements of Minimal Bayes' Factors (z* = value/noise)	0.4	https://esbl.nhlbi.nih.gov/mpkCCD-transcriptome/
Mouse mpkCCD proteome (data set 4)	The ability of a phosphatase to dephosphorylate aquaporin-2 depends on whether or not it is expressed in collecting duct cells.	Complements of minimal Bayes' factors (z* = value/noise)	10th percentile value	https://hpcwebapps.cit.nih.gov/ESBL/Database/mpkCCD_Protein_Abundances/
Dot products from mpkCCD subcellular fraction proteomics (data set 5)	The ability of a phosphatase to dephosphorylate aquaporin-2 depends on whether or not it is found in the same subcellular compartment.	Complements of minimal Bayes' factors (z* = value/noise)	median dot product	https://hpcwebapps.cit.nih.gov/ESBL/Database/mpkFractions/proteomic_fractions_linear.html
Apical mpkCCD proteome (data set 6)	The ability of a phosphatase to dephosphorylate aquaporin-2 is greater if it is found in the apical region of the cell.	Complements of minimal Bayes' factors derived from χ2 statistic		http://sbel.mc.ntu.edu.tw/mpkCCDqAMP/qAMP.htm#SamePlace
mpkCCD cytoplasmic fraction proteome (Data set 7)	The ability of a phosphatase to dephosphorylate aquaporin-2 depends on whether or not it is found in the cytoplasm.	If found in cytoplasm fraction, P = 0.95; if found only in nuclear fraction, P = 0.2; if not found, P = 0.4		https://hpcwebapps.cit.nih.gov/ESBL/Database/mNPD/index.html
Rat IMCD cytoplasmic fraction proteome (data set 8)	The ability of a phosphatase to dephosphorylate aquaporin-2 depends on whether or not it is found in the cytoplasm.	If normalized signal >7, P = 0.95; if signal between 0 and 7, P = 0.75; if signal = 0, P = 0.5		This paper

Protein Mass Spectrometry

We used liquid chromatography–tandem mass spectrometry (LC-MS/MS) to analyze the cytoplasmic fraction of IMCD cells isolated from rat kidneys as described by Pickering et al. (23) and summarized as follows. Thirty male Sprague-Dawley rats were euthanized according to an approved National Heart, Lung, and Blood Institute Animal Care and Use Committee protocol (H-0110R3). IMCD suspensions were prepared as previously described (4). The kidney inner medullas were dissected, minced, and digested into suspensions by incubation at 37°C for 70–90 min in digestion solution (250 mM sucrose, 10 mM triethanolamine, pH 7.6) containing collagenase B (3 mg/ml; Roche, Indianapolis, IN) and hyaluronidase (3 mg/ml; Worthington, Lakewood, NJ). The resulting inner medullary suspension (whole IM) was subjected to three low-speed centrifugations (at 70 g, 20 s) to separate the IMCD-enriched fraction in the pellet from the non-IMCD fraction in the supernatant. The supernatants containing non-IMCD elements were discarded. The final pellet was resuspended in tubule suspension solution containing (in mM) 118 NaCl, 5 KCl, 4 Na₂HPO₄, 25 NaHCO₃, 2 CaCl₂, 1.2 MgSO₄, 5.5 glucose, and 5 sodium acetate (300 mOsm) split in half and exposed to either the vasopressin analog dDAVP (1 nM) or its vehicle for 30 min. This method has been previously shown to successfully isolate IMCD cells that are viable (6), vasopressin responsive (3), and largely free of non-IMCD cells (31). From this point, subcellular fractionation procedures followed Tchapyjnikov et al. (30) using the NE-PER Nuclear and Cytoplasmic Extraction Reagents kit (Life Technologies, #78835) to obtain fractions three fractions, the NE (nuclear extract), the NP (nuclear pellet), and the cytoplasm. Total protein concentrations were measured (BCA assay, Life Technologies, #23227). Mass spectrometric analyses of the NE and NP fractions were previously reported by Pickering et al. (23). For this study, we performed protein mass spectrometry of the cytoplasmic fraction only. Based on this procedure, we use the term "cytoplasm" in this paper to refer to all nonnuclear elements of the cells including membrane components, cytosol, and organelles.

The cytoplasmic fraction was concentrated using Amicon centrifugal filters, Ultracel 3K (Millipore) to ~400 μg in 110 μl of solution. A 5× concentrate of Laemmli solution (7.5% SDS, 30% glycerol, 50 mM Tris, pH 6.8) with bromophenol blue (1 part 5× concentrate: 4 parts protein sample) was added to the samples. One-dimensional SDS-PAGE was performed using 12% polyacrylamide SDS-Tris-Glycine gels (Bio-Rad). Molecular weight markers (Precision Plus, Bio-Rad) were run in separate lanes. The gels were run for 45 min at 3.00 A and 200 V and were then stained with Imperial protein stain (Thermo Scientific). The gels were destained for 30 min in MS-grade water (JT Baker #JTB-9831–02) and then sliced into 39 equal-sized pieces. Each gel piece was diced into 1.5 mm³ blocks using a razor blade. These samples underwent reduction with DTT, alkylation, and in gel trypsinization as described (24) except for the use of 12.5 ng/μl Trypsin Gold (Promega).

After digestion, the peptides were extracted from the gel pieces using four successive washes in 50% ACN/0.5% formic acid. Samples were dried and peptides were then suspended in 0.1% formic acid in MS-grade water. Samples were desalted using C-18 spin columns (Thermo Scientific, # 89870) according to the protocol provided with the kit. The resulting samples were dried (Speed-Vac) for storage. Immediately before mass spectrometry analysis, the samples were re-dissolved in 0.1% formic acid in MS-grade water.

LC-MS/MS.

The samples were analyzed on a nanoflow LC system (Eksigent, Dublin, CA) coupled to a tandem mass spectrometer (Orbitrap Velos Pro Hybrid; Thermo Scientific, San Jose, CA). The sample loading onto a peptide trap cartridge (Agilent Technologies, Palo Alto, CA) occurred at a flow rate of 6 μl/min. The trapped peptides were then fractionated with a reversed-phase PicoFrit column (New Objective, Woburn, MA) using a linear gradient of 5–35% ACN in 0.1% FA. The gradient time was 45 min at a flow rate of 0.25 μl/min. Precursor mass spectra (MS1) were acquired in the Orbitrap at 60,000 resolution, and product mass spectra (MS2) were acquired with the ion trap.

To maximize the number of peptide identifications, three algorithms were used to match spectra to peptides, viz. those coded by Mascot (27), SEQUEST (34) and InsPecT (29). The posttranslational modifications allowed were a fixed carbamidomethyl modification on cysteine, variable deamination modifications on asparagine and glutamine, and a variable oxidation modification on methionine. False discovery rate (FDR) at a peptide level was set to 0.01 based on target-decoy analysis (8). To identify ambiguous identifications, we used an in-house program (coded in Java) called ProMatch (30) (https://esbl.nhlbi.nih.gov/Bioinformatic%20Tools.htm). Peptides that only matched to one gene symbol were extracted as a “unique” identification. The peptides that matched to more than one gene symbol were separated as “multiple” identifications. To reconcile the multiple identifications, transcriptomic data from Affymetrix array profiling of rat renal IMCD transcripts was used (31). (Database available at: https://dir.nhlbi.nih.gov/papers/lkem/imcdtr/.). If a given gene was not expressed, based on the transcriptomic data (median normalized value <0.4), its protein product was dropped from consideration. The SEQUEST and Mascot searches were executed within Proteome Discoverer, and peptides with FDR < 0.01 and peptide rank = 1 were retained for further analyses. The InsPecT search was carried on the Biowulf Linux Cluster at the National Institutes of Health (https://hpc.nih.gov/systems). Raw files (35.5 GB), search results, and all spectra have been uploaded to the ProteomeXchange Consortium (33) via the PRIDE partner repository with the data set identifier PXD005488. These data are accessible at http://www.ebi.ac.uk/pride/archive/.

Spectral counting.

The number of peptide matches for a given gene symbol were counted and designated as the “spectral count” for that gene symbol. This was the sum of all spectra matching to the gene symbol in all gel slices. The spectral counts in the control samples and the dDAVP samples were compared using the Fisher exact test with the contingency table consisting of the control count, dDAVP count, total count of all peptides in control, and total count of all peptides in dDAVP. For this analysis, peptides that matched to more than one gene symbol were assigned to the gene symbol with the most abundant transcript in the IMCD based on prior data (31). Spectral counting data were used in this study for the Bayes’ theorem-based analysis by extracting data for all S/T phosphatases and mapping the spectral counting values to likelihood values as indicated in Table 1.

RESULTS

Database of Mammalian S/T Protein Phosphatase Catalytic Subunits

The goal was to identify S/T phosphatases that are candidates for roles in the regulation of the AQP2 water channel as well as other proteins involved in renal collecting duct transport. For this, we used Bayes’ rule to integrate data from multiple sources to rank all S/T phosphatases with regard to probability of interacting with aquaporin-2. To do such an analysis, a list of all S/T phosphatase catalytic subunits present in mammalian genomes was needed. However, because of the lack of a readily accessible, fully curated list of known S/T phosphatases, we compiled our own list (methods) containing 38 mammalian S/T phosphatase catalytic subunits. This data set has been made available as a freely accessible database located at https://hpcwebapps.cit.nih.gov/ESBL/Database/Phosphatases/. A link is provided to allow users to download the data into an electronic spreadsheet. The webpage can be searched using the browser’s intrinsic search command.

LC/MS-MS Identification of Cytoplasmic Proteome in Rat IMCD

LC-MS/MS analysis of the cytoplasmic fraction isolated from freshly isolated rat IMCD suspensions identified 47,916 distinct peptides from proteins coded by 4,538 distinct genes. We provide the data as a publically accessible web page from which the information can be downloaded (https://hpcwebapps.cit.nih.gov/ESBL/Database/IMCDCytoplasm/). From this, we extracted spectral counting data for all 38 S/T phosphatases to be used in the Bayes’ theorem-based analysis (Table 2). In brief, we found 22 out of the 38 S/T phosphatases in the rat IMCD cytoplasmic fraction. None of these showed a significant change in abundance in response to the vasopressin analog dDAVP. Interestingly, of the 85 proteins that did show significant changes in abundance (Table 3), eight were ubiquitin E3 ligases (P < 0.0001, χ² = 51.0 vs. 46 E3 ligases among 4,492 unregulated proteins). Those that were decreased in abundance were Herc1, Herc2, Mycbp2, Rnf213, and Ubr4. Those that were increased were Trim25, Rnf31, and Ube3c.

Table 2.

S/T phosphatases found in cytoplasmic fraction of rat IMCD by protein mass spectrometry

Gene Symbol	Amino Acids, n	Spectral Counts in Control	Relative Abundance*	Annotation	dDAVP: Control Ratio	P (Fisher exact)
Ppp1cb	327	18	55.05	serine/threonine-protein phosphatase PP1-beta catalytic subunit	0.94	0.5384
Ppp2cb	309	12	38.83	serine/threonine-protein phosphatase 2A catalytic subunit beta isoform	1.08	0.4675
Ppm1g	305	8	26.23	protein phosphatase 1G	1.67	0.2080
Pdp1	563	10	17.76	[Pyruvate dehydrogenase [acetyl-transferring]]-phosphatase 1, mitochondrial isoform d	1.10	0.4701
Ppp6c	119	2	16.81	serine/threonine-protein phosphatase 6 catalytic subunit	0.88	0.5254
Ppp5c	307	5	16.29	serine/threonine-protein phosphatase 5	1.50	0.3572
Pptc7	261	4	15.33	protein phosphatase PTC7 homolog	0.80	0.5199
Ppp3ca	330	4	12.12	serine/threonine-protein phosphatase 2B catalytic subunit alpha isoform	1.00	0.6543
Ppp1ca	337	4	11.87	serine/threonine-protein phosphatase PP1-alpha catalytic subunit	1.25	0.4801
Ctdsp1	542	6	11.07	carboxy-terminal domain RNA polymerase II polypeptide A small phosphatase 1	0.75	0.5177
Ppm1k	372	4	10.75	protein phosphatase 1K, mitochondrial	0.75	0.5177
Ppp3cb	499	4	8.02	serine/threonine-protein phosphatase 2B catalytic subunit beta isoform	1.50	0.4848
Ppp1cc	521	4	7.68	serine/threonine-protein phosphatase PP1-gamma catalytic subunit	1.00	0.6543
Ilkap	307	2	6.51	integrin-linked kinase-associated serine/threonine phosphatase 2C	1.00	0.6995
Ppm1a	465	3	6.45	protein phosphatase 1A	1.00	0.6995
Ppm1b	382	2	5.24	protein phosphatase 1B	0.67	0.5152
Ppm1f	392	2	5.10	protein phosphatase 1F	2.00	0.4879
Ctdp1	244	1	4.10	RNA polymerase II subunit A C-terminal domain phosphatase	2.00	0.4879
Ppp4c	525	2	3.81	serine/threonine-protein phosphatase 4 catalytic subunit	1.50	0.4848
Ctdnep1	309	1	3.24	CTD nuclear envelope phosphatase 1	1.00	0.7580
Ctdspl	450	1	2.22	CTD small phosphatase-like protein	1.00	0.6995
Ppp2ca	969	1	1.03	serine/threonine-protein phosphatase 2A catalytic subunit alpha isoform	2.00	0.4879

^*Number of spectra in control (non-dDAVP) normalized by number of amino acids and multiplied by 1,000.

Table 3.

Proteins that change in of abundance in cytoplasmic fraction of rat IMCDs in response to dDAVP for 30 min (P < 0.1 by Fisher exact test)

Gene Symbol	Control Spectral Counts	dDAVP Spectral Counts	dDAVP: Control Ratio	P (Fisher exact)	Annotation
Aftph	0	5	∞	0.0288	aftiphilin
Tcaf2	0	5	∞	0.0288	TRPM8 channel-associated factor 2
Rnf31	0	5	∞	0.0288	E3 ubiquitin-protein ligase RNF31
Krt84	0	4	∞	0.0586	keratin, type II cuticular Hb4
Arhgef1	1	6	6.00	0.0574	rho guanine nucleotide exchange factor 1
Trim25	4	15	3.75	0.0079	E3 ubiquitin/ISG15 ligase TRIM25
Ube3c	2	7	3.50	0.0821	ubiquitin-protein ligase E3C
Spag9	5	16	3.20	0.0110	C-Jun-amino-terminal kinase-interacting protein 4
Ctnnbl1	3	9	3.00	0.0655	beta-catenin-like protein 1
Nt5c2	4	10	2.50	0.0803	5′-nucleotidase, cytosolic II
Rufy1	4	10	2.50	0.0803	RUN and FYVE domain-containing protein 1
Htatsf1	5	12	2.40	0.0630	HIV Tat-specific factor 1 homolog
Ptpn23	5	11	2.20	0.0937	tyrosine-protein phosphatase non-receptor type 23
Atp6v1h	9	17	1.89	0.0722	V-type proton ATPase subunit H
Capn5	8	15	1.88	0.0915	calpain-5
Cul5	8	15	1.88	0.0915	cullin-5
Ap2m1	14	23	1.64	0.0783	AP-2 complex subunit mu
Eps8l1	13	21	1.62	0.0974	epidermal growth factor receptor kinase substrate 8-like 1
Pfas	13	21	1.62	0.0974	phosphoribosylformylglycinamidine synthase
Actn1	20	30	1.50	0.0824	alpha-actinin-1
Fbn1	65	80	1.23	0.0871	fibrillin-1
Plec	267	208	0.78	0.0102	plectin
Ahnak	230	179	0.78	0.0156	neuroblast differentiation-associated protein AHNAK
Lrba	124	91	0.73	0.0255	lipopolysaccharide-responsive and beige-like anchor protein
Utrn	156	112	0.72	0.0089	utrophin
Tln2	49	34	0.69	0.0819	talin-2
Dync1h1	280	191	0.68	0.0001	cytoplasmic dynein 1 heavy chain 1
Prpf8	49	33	0.67	0.0650	pre-mRNA-processing-splicing factor 8
Tpr	54	35	0.65	0.0391	nucleoprotein TPR
Vps13a	58	35	0.60	0.0164	vacuolar protein sorting-associated protein 13A
Akap12	40	24	0.60	0.0398	A-kinase anchor protein 12
Mtor	37	22	0.59	0.0441	serine/threonine-protein kinase mTOR
Acsl1	24	14	0.58	0.0863	long-chain-fatty-acid–CoA ligase 1
Anxa3	24	14	0.58	0.0863	annexin A3
Piezo1	35	20	0.57	0.0379	piezo-type mechanosensitive ion channel component 1
Syne2	137	75	0.55	0.0000	nesprin-2
Macf1	200	105	0.53	0.0000	microtubule-actin cross-linking factor 1
Ubr4	139	71	0.51	0.0000	E3 ubiquitin-protein ligase UBR4
Map1b	32	16	0.50	0.0193	microtubule-associated protein 1B
Mycbp2	15	7	0.47	0.0773	E3 ubiquitin-protein ligase MYCBP2
Erc1	13	6	0.46	0.0949	ELKS/Rab6-interacting/CAST family member 1
Dst	104	42	0.40	0.0000	dystonin
Igf2r	35	14	0.40	0.0027	cation-independent mannose-6-phosphate receptor
RGD1307100	26	10	0.38	0.0074	fragile site-associated protein homolog
Sorl1	36	13	0.36	0.0010	sortilin-related receptor
Htt	37	13	0.35	0.0007	huntingtin
Dync2h1	15	5	0.33	0.0245	cytoplasmic dynein 2 heavy chain 1
Birc6	53	17	0.32	0.0000	baculoviral IAP repeat-containing protein 6
Pnn	10	3	0.30	0.0521	pinin
Fryl	27	8	0.30	0.0013	protein furry homolog-like
Rnf213	51	13	0.25	0.0000	E3 ubiquitin-protein ligase RNF213
Nup133	16	4	0.25	0.0072	nuclear pore complex protein Nup133
Herc2	12	3	0.25	0.0204	E3 ubiquitin-protein ligase HERC2
Mrpl40	8	2	0.25	0.0606	39S ribosomal protein L40, mitochondrial
Zzef1	10	2	0.20	0.0220	zinc finger ZZ-type and EF-hand domain-containing protein 1
Tacc2	24	4	0.17	0.0001	transforming acidic coiled-coil-containing protein 2
Dmxl1	18	3	0.17	0.0010	dmX-like protein 1
RGD1562629	12	2	0.17	0.0076	neurobeachin
Mdn1	6	1	0.17	0.0680	midasin
Neb	6	1	0.17	0.0680	nebulin
Prkdc	6	1	0.17	0.0680	DNA-dependent protein kinase catalytic subunit
Usp10	6	1	0.17	0.0680	ubiquitin carboxyl-terminal hydrolase 10
Vps13b	6	1	0.17	0.0680	vacuolar protein sorting-associated protein 13B
Herc1	19	3	0.16	0.0006	E3 ubiquitin-protein ligase HERC1
Scfd2	7	1	0.14	0.0388	sec1 family domain-containing protein 2
Zw10	7	1	0.14	0.0388	centromere/kinetochore protein zw10 homolog
Fry	15	2	0.13	0.0015	protein furry homolog
Wdfy3	16	2	0.13	0.0008	WD repeat and FYVE domain-containing protein 3
Lrp2	21	1	0.05	0.0000	low-density lipoprotein receptor-related protein 2
Dhx30	10	0	0.00	0.0011	putative ATP-dependent RNA helicase DHX30
Lrp1	10	0	0.00	0.0011	prolow-density lipoprotein receptor-related protein 1
Upf2	6	0	0.00	0.0172	regulator of nonsense transcripts 2
Acss2	4	0	0.00	0.0666	acetyl-coenzyme A synthetase, cytoplasmic
Ankrd17	4	0	0.00	0.0666	ankyrin repeat domain-containing protein 17
Atrx	4	0	0.00	0.0666	transcriptional regulator ATRX
Cntrl	4	0	0.00	0.0666	centriolin
Dmxl2	4	0	0.00	0.0666	dmX-like protein 2
Hal	4	0	0.00	0.0666	histidine ammonia-lyase
Kidins220	4	0	0.00	0.0666	kinase D-interacting substrate of 220 kDa
Krt17	4	0	0.00	0.0666	keratin, type I cytoskeletal 17
LOC100294508	4	0	0.00	0.0666	dyslexia-associated protein KIAA0319-like protein homolog
Snx13	4	0	0.00	0.0666	sorting nexin-13
Tmem126b	4	0	0.00	0.0666	complex I assembly factor TMEM126B, mitochondrial
Wdr18	4	0	0.00	0.0666	WD repeat-containing protein 18
Xpo5	4	0	0.00	0.0666	exportin-5

Bayes’ Theorem-Based Analysis of S/T Phosphatase Catalytic Subunits

Which of the 38 S/T phosphatases in the mammalian genome could be responsible for regulation of AQP2 and other transporters in the renal collecting duct? Using Bayes’ rule, we started with equal a priori probabilities for each phosphatase in the curated database (Table 4, column 1) equal to 1/38 and sequentially updated all probabilities (columns 2–9) using the large-scale data sets listed in Table 1. After integration of all data sets, we obtained a final ranked list of S/T phosphatases (Table 4, column 10) most likely to interact with AQP2 in the renal collecting duct, headed by (in order) Ppp1cb (tied for 1st), Ppm1g (tied for 1st), Ppp1ca, Ppp3ca, Ppp2ca, Ppp1cc, Ppp2cb, Ppp6c, and Ppp5c. A listing of the key characteristics of these S/T phosphatases is given in Table 5. These are discussed further in the discussion.

Table 4.

Ranking of S/T phosphatase catalytic subunits with regard to likelihood of interacting with aquaporin-2 in renal collecting duct

		Column 1	Column 2	Column 3	Column 4	Column 5	Column 6	Column 7	Column 8	Column 9	Column 10
Protein Name	Mouse Gene Symbol	Initial Probability	Rat Transcriptome	Rat RNA-Seq	Mouse Transcriptome	Mouse Proteome	Proteomics (dot product)	Apical mpkCCD Proteome	mpkCCD Cytoplasm	Rat IMCD Cytoplasm	Final
serine/threonine-protein phosphatase PP1-beta catalytic subunit	Ppp1c b	0.0263	0.0440	0.0518	0.0586	0.0629	0.0906	0.1446	0.2420	0.2482	0.2482
protein phosphatase 1G	Ppm1 g	0.0263	0.0440	0.0518	0.0586	0.0629	0.0906	0.1446	0.2420	0.2482	0.2482
serine/threonine-protein phosphatase PP1-alpha catalytic subunit	Ppp1ca	0.0263	0.0440	0.0518	0.0586	0.0629	0.0906	0.1179	0.1974	0.2024	0.2024
serine/threonine-protein phosphatase 2B catalytic subunit alpha isoform isoform 2	Ppp3ca	0.0263	0.0440	0.0462	0.0522	0.0560	0.0807	0.1051	0.0741	0.0759	0.0759
serine/threonine-protein phosphatase 2A catalytic subunit alpha isoform	Ppp2ca	0.0263	0.0440	0.0518	0.0586	0.0629	0.0906	0.0428	0.0717	0.0580	0.0580
serine/threonine-protein phosphatase PP1-gamma catalytic subunit	Ppp1 ml	0.0263	0.0440	0.0518	0.0586	0.0629	0.0906	0.1446	0.0510	0.0522	0.0522
serine/threonine-protein phosphatase 2A catalytic subunit beta isoform	Ppp2cb	0.0263	0.0440	0.0518	0.0586	0.0629	0.0906	0.1179	0.0416	0.0426	0.0426
serine/threonine-protein phosphatase 6 catalytic subunit	Ppp6c	0.0263	0.0440	0.0518	0.0577	0.0619	0.0892	0.0422	0.0149	0.0152	0.0152
serine/threonine-protein phosphatase 5	Ppp5c	0.0263	0.0403	0.0475	0.0456	0.0490	0.0445	0.0210	0.0148	0.0152	0.0152
protein phosphatase 1B isoform 4	Ppm1b	0.0263	0.0440	0.0518	0.0586	0.0629	0.0356	0.0169	0.0119	0.0096	0.0096
protein phosphatase 1A	Ppm1a	0.0263	0.0368	0.0386	0.0437	0.0468	0.0670	0.0317	0.0112	0.0090	0.0090
serine/threonine-protein phosphatase 4 catalytic subunit	Ppp4c	0.0263	0.0384	0.0329	0.0372	0.0399	0.0574	0.0272	0.0096	0.0077	0.0077
serine/threonine-protein phosphatase 2B catalytic subunit beta isoform isoform 3	Ppp3cb	0.0263	0.0440	0.0377	0.0426	0.0458	0.0530	0.0251	0.0088	0.0071	0.0071
carboxy-terminal domain RNA polymerase II polypeptide A small phosphatase 1	Ctdsp1	0.0263	0.0440	0.0518	0.0586	0.0628	0.0094	0.0045	0.0031	0.0032	0.0032
integrin-linked kinase-associated serine/threonine phosphatase 2C	Ilkap	0.0263	0.0437	0.0515	0.0551	0.0592	0.0120	0.0100	0.0035	0.0029	0.0029
CTD small phosphatase-like protein	Ctdspl	0.0263	0.0440	0.0462	0.0522	0.0560	0.0062	0.0030	0.0021	0.0021	0.0021
protein phosphatase PTC7 homolog	Pptc7	0.0263	0.0044	0.0052	0.0059	0.0056	0.0003	0.0001	0.0001	0.0001	0.0001
protein phosphatase 1D	Ppm1d	0.0263	0.0440	0.0462	0.0473	0.0051	0.0007	0.0003	0.0001	0.0001	0.0001
protein phosphatase 1F	Ppm1f	0.0263	0.0044	0.0038	0.0010	0.0011	0.0002	0.0001	0.0001	0.0000	0.0000
[Pyruvate dehydrogenase [acetyl-transferring]]-phosphatase 1, mitochondrial isoform d	Pdp1	0.0263	0.0044	0.0024	0.0003	0.0003	0.0001	0.0000	0.0000	0.0000	0.0000
protein phosphatase 1H isoform 1	Ppm1 h	0.0263	0.0415	0.0436	0.0049	0.0053	0.0001	0.0001	0.0000	0.0000	0.0000
ubiquitin-like domain-containing CTD phosphatase 1	Ublcp1	0.0263	0.0440	0.0462	0.0500	0.0536	0.0001	0.0000	0.0000	0.0000	0.0000
pyruvate dehydrogenase [acetyl-transferring]-phosphatase 2, mitochondrial	Pdp2	0.0263	0.0440	0.0242	0.0027	0.0003	0.0000	0.0000	0.0000	0.0000	0.0000
protein phosphatase 1M isoform 1	Ppm1m	0.0263	0.0044	0.0013	0.0009	0.0001	0.0000	0.0000	0.0000	0.0000	0.0000
CTD small phosphatase-like protein 2 isoform c	Ctdspl2	0.0263	0.0044	0.0013	0.0015	0.0002	0.0000	0.0000	0.0000	0.0000	0.0000
protein phosphatase 1K, mitochondrial precursor	Ppm1k	0.0263	0.0044	0.0024	0.0003	0.0000	0.0000	0.0000	0.0000	0.0000	0.0000
CTD nuclear envelope phosphatase 1	Ctdnep1	0.0263	0.0044	0.0052	0.0006	0.0001	0.0000	0.0000	0.0000	0.0000	0.0000
serine/threonine-protein phosphatase 2B catalytic subunit gamma isoform isoform 3	Ppp3 ml	0.0263	0.0044	0.0013	0.0002	0.0002	0.0000	0.0000	0.0000	0.0000	0.0000
PH domain leucine-rich repeat-containing protein phosphatase 1	Phlpp1	0.0263	0.0044	0.0024	0.0003	0.0000	0.0000	0.0000	0.0000	0.0000	0.0000
serine/threonine-protein phosphatase with EF-hands 1	Ppef1	0.0263	0.0044	0.0013	0.0002	0.0000	0.0000	0.0000	0.0000	0.0000	0.0000
PH domain leucine-rich repeat-containing protein phosphatase 2	Phlpp2	0.0263	0.0044	0.0013	0.0002	0.0000	0.0000	0.0000	0.0000	0.0000	0.0000
protein phosphatase 1E	Ppm1e	0.0263	0.0044	0.0013	0.0002	0.0000	0.0000	0.0000	0.0000	0.0000	0.0000
serine/threonine-protein phosphatase with EF-hands 2	Ppef2	0.0263	0.0044	0.0013	0.0002	0.0000	0.0000	0.0000	0.0000	0.0000	0.0000
protein phosphatase 1J	Ppm1j	0.0263	0.0044	0.0013	0.0002	0.0000	0.0000	0.0000	0.0000	0.0000	0.0000
probable protein phosphatase 1N	Ppm1n	0.0263	0.0044	0.0013	0.0002	0.0000	0.0000	0.0000	0.0000	0.0000	0.0000
RNA polymerase II subunit A C-terminal domain phosphatase	Ctdp1	0.0263	0.0292	0.0160	0.0096	0.0103	0.0000	0.0000	0.0000	0.0000	0.0000
carboxy-terminal domain RNA polymerase II polypeptide A small phosphatase 2 isoform a	Ctdsp2	0.0263	0.0044	0.0013	0.0015	0.0003	0.0000	0.0000	0.0000	0.0000	0.0000
protein phosphatase 1L	Ppm1l	0.0263	0.0394	0.0217	0.0171	0.0001	0.0000	0.0000	0.0000	0.0000	0.0000

Chart gives posterior probabilities after indicated elements of data from Table 1 were assimilated into the Bayes’ theorem-based analysis. Bayes’ rule operator is commutative.

Table 5.

Characteristics of top-ranked S/T phosphatases in collecting duct

Gene Symbol	Common Name	Functional Annotation (UniProt)	Evidence in Collecting Ducts from -Omic Studies
Ppp1cb	PP1-β	Protein phosphatase that associates with over 200 regulatory proteins to form highly specific holoenzymes which dephosphorylate hundreds of biological targets.	Ppp1cb protein has a half-life of 16.1 h in mouse mpkCCD cells in the absence of vasopressin and 9.8 h in presence of vasopressin (26). It was identified as attached to the apical plasma membrane by surface biotinylation in cultured mouse mpkCCD cells (19).
Ppm1g	PP2C	Member of the PP2C family of Ser/Thr protein phosphatases that are known to be negative regulators of cell stress response pathways.	Based on single-tubule RNA-Seq, the mRNA level for Ppm1g is highest in the CCD (cortical collecting duct) (16). Ppm1g protein has a half-life of 22.8 h in mouse mpkCCD cells in the presence and absence of vasopressin (26). It was identified in or attached to the apical plasma membrane by surface biotinylation in cultured mouse mpkCCD cells (19).
Ppp1ca	PP1-α	Protein phosphatase that associates with over 200 regulatory proteins to form highly specific holoenzymes which dephosphorylate hundreds of biological targets.	Ppp1ca protein has a half-life of 17.8 h in mouse mpkCCD cells in the absence of vasopressin and 12.9 h in presence of vasopressin (26). It was identified as attached to the apical plasma membrane by surface biotinylation in cultured mouse mpkCCD cells (19).
Ppp3ca	PP2B	Calcium-dependent, calmodulin-stimulated protein phosphatase. Many of the substrates contain a PxIxIT motif. (Also called “Serine/threonine-protein phosphatase 2B catalytic subunit alpha isoform”).	Ppp3ca protein has a half-life of 35.5 h in mouse mpkCCD cells in the absence of vasopressin (26). It was identified as attached to the apical plasma membrane by surface biotinylation in cultured mouse mpkCCD cells (19).
	Calcineurin
Ppp2ca	PP2A-α	PP2A is the major phosphatase for microtubule-associated proteins (MAPs). PP2A can modulate the activity of phosphorylase B kinase, casein kinase 2, mitogen-stimulated S6 kinase, and MAP-2 kinase. Activates RAF1 by dephosphorylating it at "Ser-259."	Based on single-tubule RNA-Seq, the mRNA level for Ppp2ca protein is highest in the OMCD (outer medullary collecting duct) (16). Ppp2ca protein has a half-life of 20.4 h in mouse mpkCCD cells in the absence of vasopressin and is essentially unchanged by vasopressin (26).
Ppp1cc	PP1-γ	Protein phosphatase that associates with over 200 regulatory proteins to form highly specific holoenzymes which dephosphorylate hundreds of biological targets.	Based on single-tubule RNA-Seq, the mRNA level for Ppp1cc protein is highest in the CCD (cortical collecting duct) (16). Ppp1cc protein has a half-life of 11.2 h in mouse mpkCCD cells in the absence of vasopressin and is essentially unchanged by vasopressin (26). It was identified as attached to the apical plasma membrane by surface biotinylation in cultured mouse mpkCCD cells (19).
Ppp2cb	PP2A-β	PP2A can modulate the activity of phosphorylase B kinase, casein kinase 2, mitogen-stimulated S6 kinase, and MAP-2 kinase.	Ppp2cb protein has a half-life of 67.1 h in mouse mpkCCD cells in the absence of vasopressin and only 4.1 h in the presence of vasopressin (26).
Ppp6c	PP6C	A component of a signaling pathway regulating cell cycle progression. NH2-terminal domain restricts G1 to S phase progression in cancer cells, in part through control of cyclin D1. Downregulates MAP3K7 kinase activation by dephosphorylation of MAP3K7.	Ppp6c protein has a half-life of 36 h in mouse mpkCCD cells in the absence of vasopressin (26).
Ppp5c	PP5	Serine/threonine-protein phosphatase that dephosphorylates a myriad of proteins. May modulate TGF-beta signaling pathway by the regulation of SMAD3 phosphorylation and protein expression levels.	Ppp5c protein has a half-life of 25.6 h in mouse mpkCCD cells in the absence of vasopressin and is essentially unchanged by vasopressin (26).

Sensitivity analysis.

To test the relative information content of each data set used for the ranking of S/T phosphatases with regard to likelihood of interacting with AQP2 in collecting duct cells, we performed a sensitivity analysis in which we dropped each data set from the analysis in turn and recalculated the rankings (Table 6). For all analyses with dropped data sets (except for small effects of set 7 removal), the top five phosphatases were unchanged relative to the ranking using all data sets. This indicates that no one data set had undue influence on the rankings. Set 7 is the result of proteomic profiling of the cytoplasmic fraction isolated from mouse mpkCCD cells. When this set was dropped, Ppp3ca and Ppp2ca fell slightly in ranking, while Ppp1cc and Ppp2cb rose slightly.

Table 6.

Sensitivity analysis

Gene Symbol	All Sets	Remove Set 1	Remove Set 2	Remove Set 3	Remove Set 4	Remove Set 5	Remove Set 6	Remove Set 7	Remove Set 8
Ppp1cb	1	1	1	1	1	1	1	1	1
Ppm1g	2	2	2	2	2	2	3	3	2
Ppp1ca	3	3	3	3	3	3	2	4	3
Ppp3ca	4	4	4	4	4	4	5	6	4
Ppp2ca	5	5	5	5	5	5	4	8	5
Ppp1cc	6	6	6	6	6	6	6	2	6
Ppp2cb	7	7	7	7	7	7	7	5	7
Ppp6c	8	9	8	9	8	13	8	7	8
Ppp5c	9	8	9	8	9	11	9	11	9
Ppm1b	10	11	13	10	10	10	10	13	10

Rankings of phosphatases obtained by dropping the indicated data sets and repeating the analysis. The set designator corresponds to values given parenthetically in first column of Table 1.

Comparison of S/T phosphatase rankings from rat data alone vs. mouse data alone.

The Bayes’ theorem-based ranking of S/T phosphatases with regard to likelihood of interacting with AQP2 in collecting duct cells combined data from rat and mouse studies. We asked the question: “How well do the rat data alone or the mouse data alone match the ranking obtained with the combined rat and mouse data?” As shown in Table 7, although there is some divergence among lower ranked S/T phosphatases, the top three S/T phosphatases from rat data alone or the mouse data alone matched the top three phosphatases from the Bayes’ analysis using the combined data. These three phosphatases were Ppp1cb, Ppm1g, and Ppp1ca. Since, the rat data and the mouse data were acquired independently, the agreement of species-separated analyses attests to the robustness of the overall method. It also supports the view that the mpkCCD cell line (source of mouse data) is an appropriate cell model for native collecting ducts (source of rat data) with regard to the S/T phosphatases that interact with AQP2.

Table 7.

Comparison of Bayes’ theorem-based rankings for S/T phosphatases using rat data alone vs. mouse data alone

Protein Name	Gene Symbol	Posterior Prob. (all data)	Rank (all data)	Posterior Prob. (rat alone)	Rank (rat alone)	Posterior Prob. (mouse alone)	Rank (mouse alone)
Serine/threonine-protein phosphatase PP1-beta catalytic subunit	Ppp1cb	0.2482	1	0.1283	1	0.2359	1
Protein phosphatase 1G	Ppm1g	0.2482	2	0.1283	3	0.2359	2
Serine/threonine-protein phosphatase PP1-alpha catalytic subunit	Ppp1ca	0.2024	3	0.1283	1	0.1923	3
Serine/threonine-protein phosphatase 2B catalytic subunit alpha isoform	Ppp3ca	0.0759	4	0.0482	8	0.0809	4
Serine/threonine-protein phosphatase 2A catalytic subunit alpha isoform	Ppp2ca	0.0580	5	0.1013	4	0.0699	5
Serine/threonine-protein phosphatase PP1-gamma catalytic subunit	Ppp1cc	0.0522	6	0.0270	10	0.0497	6
Serine/threonine-protein phosphatase 2A catalytic subunit beta isoform	Ppp2cb	0.0426	7	0.0270	10	0.0405	7
Serine/threonine-protein phosphatase 6 catalytic subunit	Ppp6c	0.0152	8	0.0270	10	0.0145	11
Serine/threonine-protein phosphatase 5	Ppp5c	0.0152	9	0.0495	6	0.0158	8
Protein phosphatase 1B	Ppm1b	0.0096	10	0.0426	9	0.0116	13
Protein phosphatase 1A	Ppm1a	0.0090	11	0.0159	16	0.0146	10
Serine/threonine-protein phosphatase 4 catalytic subunit	Ppp4c	0.0077	12	0.0135	18	0.0147	9
Serine/threonine-protein phosphatase 2B catalytic subunit beta isoform	Ppp3cb	0.0071	13	0.0155	17	0.0118	12

Table includes only the top 13 phosphatases from the full analysis using data from both species.

DISCUSSION

In this study, we used Bayes’ rule to integrate several large-scale data sets from proteomics and transcriptomics studies to identify serine/threonine phosphatase catalytic subunits most likely to interact with AQP2 and possibly other transport proteins in the renal collecting duct. The analysis did not consider tyrosine phosphatases or dual-specificity phosphatases. The properties of the nine top-ranked S/T phosphatases are shown in Table 5. In the following, we discuss these phosphatases with respect to the reductionist literature. One problem with organizing such a discussion is the unfortunate discordance between the commonly used terminology for phosphatases and that used for official gene symbols. To facilitate the discussion, we have included the common terms used in most of the prior papers in Table 5 along with their associated gene symbols.

Several prior studies have examined the roles of S/T phosphatases in the regulation of the AQP2 water channel. Valenti et al. (32) found that the PP2A-selective phosphatase inhibitor okadaic acid promoted a 60% increase in AQP2 phosphorylation at Ser256 and increased AQP2 translocation to the apical plasma membrane in cultured collecting duct cells. PP2A corresponds to the gene symbols Ppp2ca (rank 5) and Ppp2cb (rank 7) and is calcium insensitive. Similarly, calyculin (a nonselective PP1 and PP2A inhibitor, increased AQP2 phosphorylation at Ser256 and Ser264 (without a significant effect on Ser261 and Ser269) and also increased the abundance AQP2 in the plasma membrane (25). Jo et al. (13) discovered a protein complex in endosomes from the renal IMCD consisting of an AKAP, the regulatory subunit of protein kinase A RII, protein kinase C zeta, and the calcium-calmodulin regulated phosphatase PP2B (also called calcineurin; gene symbol: Ppp3ca; rank 4 in Table 5). AQP2 present in these endosomes was found to be dephosphorylated in vitro by addition of exogenous PP2B. Gooch et al. (9) studied AQP2 phosphorylation and trafficking in calcineurin knockout mice, showing that vasopressin-mediated phosphorylation of AQP2 at Ser256 was paradoxically decreased compared with wild-type littermates. However, in the calcineurin knockout mice, there was a striking lack of apical accumulation of AQP2 in response to vasopressin, indicating that AQP2 exocytosis or endocytosis may have been affected. Consistent with this finding, a calcineurin inhibitor, FK-506, did not alter apical plasma membrane accumulation of AQP2 in native rat IMCD cells, despite increasing AQP2 phosphorylation at Ser261 and Ser264 (25). The water permeability response in isolated perfused collecting ducts is markedly reduced by inhibitors of calmodulin (5, 6), and a portion of this effect could have been mediated by loss of calcineurin action. Li et al. (17) showed that hypertonicity promotes the nuclear translocation of calcineurin-regulated NFATc proteins (in addition to the calcineurin insensitive Nfat5) with the subsequent induction of AQP2 expression. Possible roles in AQP2 regulation for three top ranked phosphatases in our Bayesian analysis (Ppp1cb, Ppm1g, Ppp1ca) have not, to our knowledge, been investigated. However, work by Kubokawa et al. (15) suggests a role for PP-2C, corresponding to the second ranked phosphatase Ppm1g, in regulation of the Kcnj1 potassium channel (ROMK) in the renal cortical collecting duct. Furthermore, Lin et al. (18) have implicated PP1 [corresponding to Ppp1cb (rank 1) and Ppp1ca (rank 3)] in the regulation of ROMK in the cortical collecting duct. Specifically, they found that c-Src regulates the interaction between WNK4 and SGK1 by increasing PP1 binding to WNK4, thereby decreasing WNK4 phosphorylation and regulating ROMK activity in cortical collecting duct. Finally, several of the top-ranked phosphatases were found in a comprehensive surface-biotinylation/mass-spectrometry study of mpkCCD cells to be apically located peripheral-membrane proteins (Table 5) (19), i.e., vicinal to AQP2, ROMK, and other apical transporters. These are Ppp1ca, Ppp1cb, Ppp1cd, Ppm1g, and calcineurin.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the author(s).

AUTHOR CONTRIBUTIONS

S.L., C.M.P., C.-L.C., E.N.U., and M.A.K. conception and design of research; S.L., C.M.P., and E.N.U. performed experiments; S.L., V.R., C.R.G., C.M.P., C.-L.C., and M.A.K. analyzed data; S.L., V.R., C.M.P., C.-L.C., and M.A.K. interpreted results of experiments; S.L. and M.A.K. prepared figures; S.L. and M.A.K. drafted manuscript; S.L., V.R., C.R.G., C.M.P., C.-L.C., E.N.U., and M.A.K. edited and revised manuscript; S.L., V.R., C.R.G., C.M.P., C.-L.C., E.N.U., and M.A.K. approved final version of manuscript.