Identification of β-catenin-interacting proteins in nuclear fractions of native rat collecting duct cells

Abstract

The gene encoding the aquaporin-2 water channel is regulated transcriptionally in response to vasopressin. In the renal collecting duct, vasopressin stimulates the nuclear translocation and phosphorylation (at Ser⁵⁵²) of β-catenin, a multifunctional protein that acts as a transcriptional coregulator in the nucleus. The purpose of this study was to identify β-catenin-interacting proteins that might be involved in transcriptional regulation in rat inner medullary collecting duct (IMCD) cells, using experimental and computational approaches. We used a standard chromatin immunoprecipitation procedure coupled to mass spectrometry (ChIP-MS) in a nuclear fraction isolated from rat IMCD suspensions. Over four biological replicates, we reproducibly identified 43 β-catenin-binding proteins, including several known β-catenin-binding partners as well as novel interacting proteins. Multiple proteins involved in transcriptional regulation were identified (Taf1, Jup, Tdrd3, Cdh1, Cenpj, and several histones). Many of the identified β-catenin-binding partners were found in prior studies to translocate to the nucleus in response to vasopressin. There was only one DNA-binding transcription factor (TF), specifically Taf1, part of the RNA-polymerase II preinitiation complex. To identify sequence-specific TFs that might interact with β-catenin, Bayes’ theorem was used to integrate data from several information sources. The analysis identified several TFs with potential binding sites in the Aqp2 gene promoter that could interact with β-catenin in the regulation of Aqp2 gene transcription, specifically Jun, Junb, Jund, Atf1, Atf2, Mef2d, Usf1, Max, Pou2f1, and Rxra. The findings provide information necessary for modeling the transcriptional response to vasopressin.

the peptide hormone vasopressin regulates solute and water transport across the epithelium of the renal collecting duct. It controls water transport, in part through regulation of the water channel aquaporin-2 (AQP2) in at least two ways (33, 52): 1) vasopressin stimulates membrane trafficking events that redistribute AQP2 from the endosomal compartment to the apical plasma membrane of collecting duct cells; and 2) vasopressin increases AQP2 protein abundance, in part via regulation of Aqp2 gene expression (60). This paper addresses the latter mechanism.

The multifunctional protein β-catenin can function in the nucleus as a transcriptional coregulator that binds to and modulates the actions of DNA-binding transcription factors (6). It also has roles in the cytoplasm as part of the adherens junctions and as an element of the Wnt signaling complex (72). Recent studies in collecting duct cells have shown that vasopressin markedly increases the phosphorylation of β-catenin at Ser⁵⁵², a basophilic site that is a putative target of protein kinase A (3, 5, 58, 62), which differs from the sites phosphorylated by the protein kinase GSK3β in Wnt signaling (77). Furthermore, in mpkCCD cells, vasopressin-induced phosphorylation of β-catenin at Ser⁵⁵² was associated with translocation of β-catenin into the nucleus (62). In addition, a siRNA-mediated knockdown of β-catenin in mpkCCD cells markedly impaired the ability of vasopressin to increase AQP2 protein abundance, consistent with a role in the regulation of Aqp2 gene transcription (32). Also, it has been demonstrated that the soluble prorenin receptor can induce or enhance antidiuresis through β-catenin-dependent effects on Aqp2 gene expression (44). Thus, β-catenin is a prime candidate for a role in vasopressin-mediated transcriptional regulation.

One way to investigate β-catenin-mediated transcriptional regulation in the collecting duct is to identify proteins that bind to it in nuclear fractions in the absence and/or presence of vasopressin. To accomplish this in the rat inner medullary collecting duct (IMCD) in the present study, we carried out chromatin immunoprecipitation (ChIP) of β-catenin followed by identification of coprecipitated proteins by use of protein mass spectrometry (ChIP-MS). We reproducibly identified 43 β-catenin-binding proteins, which included a number of known β-catenin-binding partners as well as novel interacting proteins. We extended the analysis computationally through use of large-scale data integration techniques to identify sequence-specific transcription factors that are known to interact with β-catenin, that are expressed in IMCD cells, and that possess potential binding sites (conserved transcription factor-binding motifs) in the promoter of the Aqp2 gene.

METHODS

Preparation of Rat IMCD Suspensions and Nuclear Fractions

The study design was approved by the Animal Care and Use Committee of the Division of Intramural Research of the National Heart, Lung, and Blood Institute (Animal Study Protocol No. H-0110R3). In each of four experiments, 12 male Sprague-Dawley rats weighing 250–275 g were euthanized by rapid decapitation. Inner medullas from both kidneys were minced and rapidly subjected to treatment with a digestion solution (250 mM sucrose, 10 mM triethanolamine, pH 7.6) containing collagenase B (3 mg/ml) and hyaluronidase (3 mg/ml) for 75–90 min, as previously described, to produce IMCD suspensions (50). The resulting inner medullary suspension (whole IM) was subjected to three low-speed centrifugations (at 70 g, 20s) to separate the IMCD-enriched fraction in the pellet from the non-IMCD fraction in the supernatant. The IMCD-enriched pellet was resuspended in IMCD suspension fluid (in mM: 118 NaCl, 25 NaHCO₃, 5 KCl, 4 Na₂HPO₄, 2 CaCl₂, 1.2 MgSO₄, 5.5 glucose, and 5 Na acetate, 290 mosmol/kgH₂O) and divided into two equal-volume samples. One was exposed to the V2 receptor-selective vasopressin analog dDAVP (1 nM, Bachem) and the other to its vehicle for 30 min at 37°C under 95% air-5% carbon dioxide.

Nuclear pellets were isolated using a commercial kit (truChIP Chromatin Shearing Reagent Kit, Covaris, no. 520154) following the manufacturer’s protocol. Briefly, cells are fixed and cross-linked with formaldehyde, plasma membranes are lysed, allowing cytoplasmic elements to be washed out, and the nuclei left behind are harvested by centrifugation. The nuclear pellets were resuspended in 1 ml of shearing buffer provided in the kit and were sheared by ultrasound to break up DNA. The quality of sheared chromatin was confirmed by DNA gel electrophoresis on 2% agarose gel (E-Gel EX gels 2% agarose, Invitrogen). Protein amounts of the sheared chromatin samples were determined by BCA assay.

Immunoprecipitation

We used a standard ChIP protocol to immunoprecipitate β-catenin, using a commercial ChIP kit (SimpleChIP Enzymatic Chromatin IP Kits; Cell Signaling, no. 9003) following the manufacturer’s protocol. We used a Ctnnb1 antibody (Cell Signaling, no. 9562, 3.4 μg IgG per 700 μg protein at 4°C overnight with gentle rotation). For IP controls, we substituted an equal amount of rabbit preimmune IgG. The immunoprecipitated material consisted of protein plus DNA and is usually used for deep sequencing of DNA. However, in the present study, it was used instead for protein mass spectrometric identification of immunoprecipitated proteins.

In-Gel Digestion

Aliquots (50 μl) of immunoprecipitated material were loaded into individual lanes of a 12-well 7.5% Criterion TGX polyacrylamide gel (Bio-Rad) for SDS-PAGE. The gel was washed and stained with Coomassie blue dye for 10 min. Each lane was sliced into 2 mm × 9 mm rectangles above the nominal molecular mass of β-catenin (above 80 kDa) by using GridCutter (Gel Company). Each of four experiments, then, yielded 32 separate samples for LC-MS/MS analysis (8 minced gel-slice samples × 2 treatments × 2 IP conditions). The gel pieces were washed in 100 µl of 25 mM NH₄HCO₃ and then dehydrated with 25 mM NH₄HCO₃-50% ACN. These samples underwent reduction with DTT, alkylation with iodoacetamide, and trypsinization using 12.5 ng/μl Trypsin Gold (Promega).

After digestion, the peptides were extracted from the gel pieces by four successive washes in 50% ACN-0.5% formic acid. The extracted peptides were desalted using ZipTips (Millipore, ZTC18S096, tip size: P10) before MS analysis.

LC-MS/MS

All samples were analyzed on a nanoflow LC system (Eksigent, Dublin, CA) coupled to a tandem mass spectrometer (Orbitrap Elite; 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.

Peptide/Protein Identification

To maximize the number of peptide identifications, three algorithms were used to match spectra to peptides, specifically those coded by Mascot, SEQUEST, and InsPecT. 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 (13). To identify ambiguous identifications, we used an in-house program (coded in Java) called ProMatch (http://esbl.nhlbi.nih.gov/Bioinformatic%20Tools.htm). 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 out on the Biowulf Linux Cluster at the National Institutes of Health (https://hpc.nih.gov/). For InsPecT, the P value cut-off for the peptide identification was set at 0.01 (68). 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 (database at: https://esbl.nhlbi.nih.gov/IMCD-transcriptome/). Peptides that matched to more than one gene symbol were assigned to the gene symbol with the most abundant transcript in the IMCD on the basis of prior data (55).

Raw files, search results, and all spectra have been uploaded to the ProteomeXchange Consortium via the PRIDE partner repository with the data set identifier PXD005792. 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 were normalized by the total number of spectral counts mapped to all proteins. Relative abundance of a given protein was estimated by dividing the normalized spectral counts by the number of amino acids in the mapped protein. Normalized spectral counts from specific-antibody IP vs. preimmune IgG IP were expressed as a ratio of the two values. Vehicle 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.

Semiquantitative Immunoblotting

Sample proteins were resolved by SDS-PAGE on 7.5% Criterion TGX polyacrylamide gel (Bio-Rad) and transferred electrophoretically onto nitrocellulose membranes. Membranes were blocked for 1 h with Odyssey blocking buffer (Li-Cor), and probed overnight at 4°C with primary antibody against β-catenin (from Cell Signaling) at 1:1,000 dilution in Odyssey blocking buffer containing 0.1% Tween 20. After 1 h of incubation with secondary antibody (Li-Cor no. 926-68071, IRDye680RD Goat anti-Rabbit IgG) at 1:5,000 dilution in blocking buffer with 0.1% Tween 20, sites of antigen-antibody reaction were detected using Odyssey's infrared imager (Li-Cor).

Bioinformatics and Statistics

For analysis of the ChIP-MS data, several software tools were used. ABE (Automated Bioinformatics Extractor; Systems Biology Center, NHLBI, Bethesda, MD) was used to extract Gene Ontology terms (http://helixweb.nih.gov/ESBL/ABE/). Nuclear translocation data were mined from a publicly accessible web page: https://hpcwebapps.cit.nih.gov/ESBL/Database/QuantNucProteomics/ (temporary login: clp; temporary password: Esbl!@#$). Random selection of genes from a list was done in Microsoft Excel, using a random number generator RAN() to create an arbitrary order and by selecting those with the lowest random number values. Statistical analysis of mass spectrometry data used Minimum Bayes’ Factors in lieu of conventional P values to summarize strength of evidence (17, 18). Conventional Chi-square analysis (a web-based Chi-Square calculator http://www.socscistatistics.com/tests/chisquare/) was used to test whether a set of proteins with a given property is found more frequently in the interactome list than it would be in a list of randomly chosen proteins. Nuclear localization signal sequences (K-(R/K)-X-(R/K) in β-catenin-interacting proteins were identified using an in-house program ProMatch, which utilizes the Java Pattern Class using “Regular Expressions” (https://hpcwebapps.cit.nih.gov/ESBL/ProMatch/). Annotated domains of β-catenin-interacting proteins were identified from domain information extracted with ABE (https://hpcwebapps.cit.nih.gov/ESBL/ABE/). An interactive network of domain-protein was generated by Medusa (3.0) and Cytoscape (http://www.cytoscape.org/). Data regarding vasopressin responses in renal collecting duct cells were mined from prior publications using Biological Information Gatherer (BIG) (87) (https://big.nhlbi.nih.gov/index.jsp).

Large-Scale Data Integration using Bayes’ Theorem

We carried out a large-scale data integration analysis using Bayes’ Theorem to answer the question, “Which among the 1,375 DNA-binding transcription factors annotated in the rat genome are most likely to regulate Aqp2 gene expression in a β-catenin-dependent manner?” The method, as applied to physiological questions, was described previously in detail (40, 79). Briefly, to address the question, the method begins with a set of unbiased prior probabilities for each transcription factor (each 1/1,375). It then calculates new posterior probabilities for each transcription factor using likelihood estimates derived from experimentally determined data from relevant data sources. Posterior probabilities after integration of one data set become prior probabilities for the next step. The steps in the calculation, the data sources and the rationale for the use of the data are summarized in Table 7. A novel aspect of the analysis was the prediction of which transcription factors were most likely to bind to the promoter of the Aqp2 gene (defined as the 1,000 bp immediately upstream from the transcription start site). For transcription factor binding site (TFBS) prediction analysis, we used a web-based tool PROMO 3.0 (http://alggen.lsi.upc.es/cgi-bin/promo_v3/promo/promoinit.cgi?dirDB=TF_8.3) to identify binding motifs from the TRANSFAC database (http://gene-regulation.com/pub/databases.html#transfac). This identified sites by using specific rat transcription factors as exemplars. We mapped the transcription factors expressed in rat IMCD cells to these exemplars using the BLAST algorithm (criterion: E value < 0.001) (https://blast.ncbi.nlm.nih.gov/Blast.cgi) using the sequence corresponding to the DNA-binding region as inputs. As described in Table 8, the likelihood values for each transcription factor used in the Bayes’ analysis were the values from PROMO, giving the probability of finding the indicated motif in the Aqp2 promoter vs. the probability of finding the motif in random sequences of bases of the same length (values in parentheses in Fig. 6). The mapped transcription factors that are known to bind β-catenin were identified through application of STRING analysis (http://string-db.org/), BioGrid (https://thebiogrid.org/), the previously published β-catenin interactome in mouse embryonic stem cells (80), and PubMed searches.

Table 7.

Data sets used in Bayes’ analysis to address the question, “What transcription factors (TFs) are most likely among all known TFs to be involved in regulation of Aqp2 gene transcription?”

Data Set	Rationale	Assignment of Likelihood Values: P(B|A)	URL
All transcription factors (TFs) in rat	TFs that regulate AQP2 must include one or more of the TFs on this list.	1/1,375 of total transcription factors as initial probability of each transcription factor	http://www.bioguo.org/AnimalTFDB/
Rat IMCD transcriptome	The ability of a transcription factor to regulate transcription of Aqp2 gene depends on whether or not it is expressed in IMCD cells.	Complements of minimum Bayes’ factors (Z* = signal value/noise value where intrinsic noise is 0.4)	https://esbl.nhlbi.nih.gov/IMCD-transcriptome
Rat IMCD proteome	Ability of a transcription factor to regulate transcription of Aqp2 gene depends on whether or not it is expressed in IMCD cells.	Complements of minimum Bayes’ factors from χ2 test using number of data sources among 20 IMCD proteome databases published previously (null hypothesis: 0/20)	https://hpcwebapps.cit.nih.gov/ESBL/Database/IMCD_Proteome_Database/
Rat IMCD nuclear proteome	Ability of a transcription factor to regulate transcription of Aqp2 gene depends on whether or not it is present in nuclei of IMCD cells.	If found in both nuclear extract and nuclear pellet fractions, P = 0.95; if found one of two fractions, P = 0.8; if not found, P = 0.4	https://helixweb.nih.gov/ESBL/Database/IMCD_Nucleus/NE_NP_Webpage_TF.html
Transcription factor binding sites in promoter of rat Aqp2 gene	Ability of a transcription factor to regulate transcription of Aqp2 gene depends on whether or not it binds to specific DNA sites in the promoter of rat Aqp2 gene	Complements of minimum Bayes’ factors from the nucleotide probability of random expectation (expected occurrences of the match in a random sequence of the same length as the query sequence according to the dissimilarity index)	http://alggen.lsi.upc.es/cgi-bin/promo_v3/promo/promoinit.cgi?dirDB=TF_8.3

See Supplemental Table S1 (online) for full details of calculation.

Table 8.

Top-ranking TFs answer the question, “What TFs are most likely among all known TFs to be involved in regulation of Aqp2 gene transcription?”

Gene Symbol	Protein Name	TF Family	Predicted Factor	Initial Probability	IMCD Transcriptome	IMCD Proteome	IMCD Nuclear Proteome	TFBS at Rat AQP2 Promoter
Jund	Jun D proto-oncogene	TF_bZIP	T00031/T00164	0.0007	0.0048	0.0100	0.0106	0.0284
Mef2d	Myocyte enhancer factor 2D	SRF	T00766	0.0007	0.0042	0.0087	0.0093	0.0247
Crem	Camp responsive element modulator	TF_bZIP	T00164	0.0007	0.0048	0.0073	0.0092	0.0246
Nr1h2	Oxysterols receptor LXR-beta	THR-like	T00042/T00333	0.0007	0.0048	0.0073	0.0092	0.0246
Fosl2	Fos-like antigen 2	TF_bZIP	T00031/T00124	0.0007	0.0048	0.0073	0.0078	0.0207
Rxra	Retinoid X receptor alpha	RXR-like	T00042/T00333	0.0007	0.0048	0.0073	0.0078	0.0207
Nr2f6	Nuclear receptor subfamily 2, group F, member 6	RXR-like	T00042/T00333	0.0007	0.0048	0.0073	0.0078	0.0207
Usf2	Upstream transcription factor 2	bHLH	T00875/T02115	0.0007	0.0048	0.0073	0.0078	0.0207
Foxp1	Forkhead box P1	Fork head	T00371/T01049	0.0007	0.0033	0.0051	0.0065	0.0172
Hoxb8	Homeo box B8	Homeobox	T03458	0.0007	0.0031	0.0048	0.0060	0.0160
Esrra	Estrogen related receptor, alpha	ESR-like	T00042/T00333	0.0007	0.0030	0.0046	0.0058	0.0154
Atf3	Activating transcription factor 3	TF_bZIP	T00031/T00124	0.0007	0.0048	0.0045	0.0057	0.0152
Ddit3	DNA-damage inducible transcript 3	TF_bZIP	T00108/T00459/T00109	0.0007	0.0048	0.0045	0.0057	0.0152
Nr3c2	Mineralocorticoid receptor	ESR-like	T00042/T00333	0.0007	0.0048	0.0045	0.0057	0.0152
Usf1	Upstream transcription factor 1	bHLH	T00875/T02115	0.0007	0.0047	0.0044	0.0056	0.0148
Pou2f1	POU class 2 homeobox 1	Pou	T03458	0.0007	0.0031	0.0048	0.0051	0.0135
Max	MYC associated factor X	bHLH	USF-1/T02115	0.0007	0.0042	0.0040	0.0050	0.0134
Jun	Jun proto-oncogene	TF_bZIP	T00031/T00164	0.0007	0.0048	0.0090	0.0048	0.0128
Junb	Jun B proto-oncogene	TF_bZIP	T00031/T00164/T00124	0.0007	0.0048	0.0045	0.0048	0.0128
Pou3f3	POU class 3 homeobox 3	Pou	T03458	0.0007	0.0048	0.0045	0.0048	0.0128
Srebf1	Sterol regulatory element binding transcription factor 1	bHLH	T00875/T02115	0.0007	0.0048	0.0045	0.0048	0.0128
Mitf	Microphthalmia-associated transcription factor	bHLH	T00875/T02115	0.0007	0.0047	0.0044	0.0047	0.0125
Arntl	Aryl hydrocarbon receptor nuclear translocator-like	bHLH	T00875/T02115	0.0007	0.0038	0.0036	0.0045	0.0120

Predicted Factor: TF considered as the most potential protein binding to specific region of the query sequence as predicted by ‘PROMO 3.0’ (http://alggen.lsi.upc.es/cgi-bin/promo_v3/promo/promoinit.cgi?dirDB=TF_8.3). TFs theoretically capable of binding to these sites were identified by BLAST search against all TFs. Final values after integrating all data sets are highlighted in boldface (last column).

Fig. 6.

Map of potential binding sites for sequence-specific transcription factors in the promoter of the Aqp2 gene. Designators (Txxxx format) refer to motifs found by PROMO (http://alggen.lsi.upc.es/cgi-bin/promo_v3/promo/promoinit.cgi?dirDB=TF_8.3) corresponding to TRANSFAC motifs. Values in parentheses refer to the probability of finding the indicated motif in a randomly ordered sequence of the same length and base composition as in the Aqp2 promoter. Assigned likelihood values in Bayes’ analysis are given at the right of gene symbols for specific transcription factors. Transcription factors shown in red are known to bind β-catenin.

RESULTS

To provide information needed for modeling of transcriptional regulation of the Aqp2 gene in the renal collecting duct, we used two approaches: 1) ChIP-MS analysis to identify proteins that complex with β-catenin in the nucleus of native rat renal collecting duct cells, and 2) large-scale data integration using Bayes’ theorem to identify sequence-specific transcription factors likely to bind β-catenin in the promoter of the Aqp2 gene.

ChIP-MS Analysis

We used an antibody to β-catenin and a standard ChIP protocol to isolate β-catenin and its interacting proteins in nuclear fractions of native rat IMCD cells. We combined this with protein mass spectrometry (LC-MS/MS) to identify and quantify the bound proteins. Figure 1 shows the characterization of the antibody with regard to its ability to recognize β-catenin in Western blotting and IP. Western blotting of the “Input” to the IP procedure identified a unitary band of molecular mass of ~90 kDa, which is similar to the nominal molecular mass of β-catenin, namely 85.4 kDa. The middle two lanes of Fig. 1 show that β-catenin was successfully immunoprecipitated with the β-catenin-specific antibody but not with preimmune IgG.

Fig. 1.

The β-catenin antibody recognizes β-catenin in immunoprecipitation (IP) samples from nuclear fraction of rat inner medullary collecting duct (IMCD) cells. “Input” is pre-IP sample; “IP” is bound material eluted from protein G beads after IP using β-catenin antibody or preimmune IgG as control. An equal amount of IgG was used for preimmune and β-catenin IPs. “Unbound” is material that did not attach to the protein G beads. Loading was on an equal-volume basis (15 µl per lane).

We then used the β-catenin antibody to carry out eight IP experiments consisting of four pairs of samples (vasopressin-treated vs. vehicle-treated native IMCD cells). We first addressed the question, “Among all of the samples, what proteins show significant enrichment in the β-catenin immunoprecipitation compared with the preimmune-IgG immunoprecipitation?” To do this, we used protein mass spectrometry to identify and quantify tryptic peptides from the immunoprecipitated proteins. Figure 2 shows a Venn diagram comparing the numbers of unique peptides that were identified by three search algorithms that are based on different principles, specifically Mascot, SEQUEST, and InsPecT. This figure shows that all three algorithms made unique identifications and thus contributed significantly to the data set of β-catenin-interacting proteins. The quantification of the abundance of a given protein was done by spectral counting, which was normalized by the total number of spectral counts for all proteins obtained for the sample. Table 1 summarizes the results, listing the 43 proteins identified as β-catenin-interacting proteins in all four experimental pairs. Among these, 13 were annotated as adherens junction proteins. Thirty-six were previously found in the nuclei of native rat IMCD cells by Pickering et al. (55). Five (Jup, Ppl, Ahnak, Sptbn2, and Ctnnb1) were previously found to translocate into the nucleus in response to vasopressin (62). Twenty-three were found to possess classical nuclear localization signals [motif: K(K/R)x(K/R)] (Table 2). Among the 43 β-catenin-interacting proteins, there was one transcription factor protein, Taf1, which is the largest component of transcription factor TFIID. The TFIID complex is composed of TATA-binding protein (Tbp) and a variety of Tbp-associated factors (78) and is a component of the RNA-polymerase II preinitiation complex. We also found a number of proteins known to be involved in transcriptional regulation (based on Biological Process Gene Ontology terms) including a few histones, Jup, Tdrd3, Lgr4, Cdh1, and Cenpj. The functional annotations (from UniProt) of these five proteins and Taf1 are shown in Table 3. In addition, several of the proteins in Table 1 are intermediate filament proteins, specifically Krt4, Krt19, Krt28, and Krt76, or intermediate filament binding proteins (Plec, Pkp2, Jup). Also there are proteins involved in the organization of the actin cytoskeleton (Vcl, Epb41l1, Plec, Ctnna1, Cdh1, Ahnak, Jup, H1f0, Hspb1, and Fat1) as well as several β-catenin-like proteins (Pkp2, Jup, Ctnna1, and Ctnnd1). β-Catenin-like proteins may be expected, due to the ability of β-catenin to heterodimerize with other catenins (57). In addition, we found β-catenin binding to its classical interactor, Cdh1 (E-cadherin), as well as another cadherin, Fat1. Interestingly, the analysis did not identify either of two other cadherins that are abundant in IMCD cells, specifically Cdh16 (“kidney cadherin”) and Cdh23 (39). Figure 3 shows a general picture of the β-catenin-interacting protein list from this study, indicating the relative number of publications that link β-catenin with each member. The figure includes proteins previously known to interact with β-catenin as well as novel interactors with no reported publications. Figure 4 summarizes the major Molecular Function Gene Ontology terms that mapped to the β-catenin interactome, specifically those related to transcriptional coactivator activity, cell-cell adhesion, and polyA RNA binding. The first two functional terms are consistent with generally recognized roles for β-catenin in all cells and support the validity of the methodology.

Fig. 2.

Nominal distribution of unique peptides found from each search algorithm is displayed as a Venn diagram. Mascot and SEQUEST searches were run using Proteome Discoverer 1.4 on desktop computers. The InsPecT search was carried on the Biowulf Linux Cluster at the National Institutes of Health (https://hpc.nih.gov). All three search programs added unique peptide identifications.

Table 1.

Proteins that bind β-catenin in nuclear fractions of rat IMCD cells

Gene Symbol	Annotation	Mean Relative Abundance	Median Ab/Preimmune Ratio	P (χ2)	Adherens Junction?	Nucleus?	Nuclear Translocation in Response to Vasopressin?
Ctnnb1	Catenin beta-1	508	∞	0.0001	Yes	Yes	Yes
Ctnna1	Catenin alpha-1	337	∞	0.0001	Yes	Yes	No
Ctnnd1	Catenin delta-1	101	∞	0.0001	Yes	Yes	No
Cdh1	Cadherin-1	87	∞	0.0001	Yes	Yes	No
Lama1	Laminin subunit alpha-1	67	∞	0.0209	No	Yes	No
Hspd1	60 kda heat shock protein, mitochondrial	51	∞	0.0004	No	Yes	No
Tenm3	Teneurin-3	40	∞	0.0001	No	No	No
Plekha6	Pleckstrin homology domain-containing family A member 6	36	∞	0.0001	No	Yes	No
Akr1b1	Aldose reductase	34	∞	0.0004	No	Yes	No
Lrba	Lipopolysaccharide-responsive and beige-like anchor protein	33	∞	0.0001	No	Yes	No
Tpi1	Triosephosphate isomerase	33	∞	0.0209	No	Yes	No
Sptbn2	Spectrin beta chain, non-erythrocytic 2	32	∞	0.0001	Yes	No	Yes
Tdrd3	Tudor domain-containing protein 3	31	∞	0.0004	No	Yes	No
Taf1	Transcription initiation factor TFIID subunit 1	25	∞	0.0209	No	Yes	No
Vcl	Vinculin	20	∞	0.0019	Yes	Yes	No
Hspb1	Heat shock protein b-1	20	∞	0.0209	No	Yes	No
Fat1	Protocadherin Fat 1	18	∞	0.0209	No	Yes	No
Epb41l1	Band 4.1-like protein 1	16	∞	0.0209	Yes	Yes	No
Afdn	Afadin	16	∞	0.0019	No	No	No
Ugt1a5	UDP-glucuronosyltransferase 1–5	15	∞	0.0209	No	No	No
Pkp2	Plakophilin-2	12	∞	0.0209	Yes	Yes	No
Pithd1	PITH domain-containing protein 1	11	∞	0.0209	No	Yes	No
Etl4	Sickle tail protein homolog	10	∞	0.0019	No	Yes	No
Mdh2	Malate dehydrogenase, mitochondrial	9	∞	0.0004	No	Yes	No
Dnhd1	Dynein heavy chain domain-containing protein 1	6	∞	0.0209	No	No	No
Hnrnpm	Heterogeneous nuclear ribonucleoprotein M	4	∞	0.0001	No	Yes	No
Ahnak	Neuroblast differentiation-associated protein AHNAK	46	8.06	0.0004	Yes	Yes	Yes
Ccdc130	Coiled-coil domain-containing protein 130	109	3.82	0.0019	No	Yes	No
Tns3	Tensin-3	136	3.78	0.0019	No	Yes	No
Lgr4	Leucine-rich repeat-containing G-protein coupled receptor 4	325	3.06	0.0004	No	Yes	No
Cenpj	Centromere protein J	61	2.82	0.0019	No	Yes	No
Jup	Junction plakoglobin	206	2.64	0.0001	Yes	Yes	Yes
Plec	Plectin	30	2.19	0.0004	Yes	Yes	No
Itga1	Integrin alpha-1	123	1.87	0.0004	No	No	No
Alpk1	Alpha-protein kinase 1	92	1.82	0.0019	No	Yes	No
Ppl	Periplakin	14	1.82	0.0019	Yes	Yes	Yes
Krt4	Keratin, type II cytoskeletal 4	16	1.80	0.0209	No	No	No
Krt76	Keratin, type II cytoskeletal 2	75	1.76	0.0019	No	Yes	No
Hist1h4b	Histone H4	408	1.70	0.0019	No	Yes	No
H1f0	Histone H1.0	313	1.69	0.0019	No	Yes	No
H2afz	Histone H2A.Z	294	1.65	0.0019	No	Yes	No
Krt19	Keratin, type I cytoskeletal 19	382	1.64	0.0001	No	Yes	No
Krt28	Keratin, type I cytoskeletal 28	149	1.61	0.0019	No	Yes	No

Mean relative abundance is the normalized spectral counts divided by the number of amino acids (in millions) averaged over all positive samples. Ab/IgG ratio is the ratio of normalized spectral counts with the β-catenin-specific antibody divided by the normalized spectral counts with the same concentration of preimmune IgG. χ² P value is based on a comparison of the number of samples (max = 4) that had Ab/IgG ratio greater than 1 vs. the null hypothesis for vasopressin-treated or vehicle-treated cells. Presence in adherens junction was identified through extraction of Cellular Component GO terms. Presence in nucleus was identified by comparison with nuclear proteome of rat inner medullary collecting duct (IMCD) cells (55) (https://helixweb.nih.gov/ESBL/Database/IMCD_Nucleus/). Nuclear Translocation in Response to Vasopressin is based on data from Schenk et al. (62), with P < 0.05 for nuclear extract fraction or nuclear pellet fraction. (https://hpcwebapps.cit.nih.gov/ESBL/Database/QuantNucProteomics/).

Table 2.

IMCD nuclear β-catenin interacting proteins with nuclear localization signal sequences (K-(K/R)-X-(K/R)

Gene symbol	Annotation	RefSeq	NLS Location(s) and Sequence(s)
Afdn	Afadin	NP_037349.1	163–166:KKEK;166–169:KKEK;169–172:KKKR;171–174:KREK;223–226:KRRR;353–356:KKMK
Ahnak	Neuroblast differentiation-associated protein AHNAK	NP_001178880.1	2627–2630:KKSR;4550–4553:KKPK;4593–4596:KKSK;5275–5278:KKSK;5341–5344:KKPR;5418–5421:KKSR
Alpk1	Alpha-protein kinase 1	XP_017446832.1	1092–1095:KKGK
Ccdc130	Coiled-coil domain-containing protein 130	NP_001032733.1	137–140:KKEK
Cenpj	Centromere protein J	NP_001100735.1	941–944:KKER
Ctnna1	Catenin alpha-1	NP_001007146.1	51–54:KRGR;375–378:KKTR;873–876:KREK
Ctnnd1	Catenin delta-1	NP_001101210.1	622–625:KKGK;788–791:KKLR
Dnhd1	Dynein heavy chain domain-containing protein 1	XP_008758040.1	547–550:KKLK
Epb41l1	Band 4.1-like protein 1 isoform L	NP_067713.2	92–95:KKFK;1055–1058:KKPR
Etl4	Sickle tail protein homolog	XP_008770135.2	608–611:KKLR;690–693:KRVR
Fat1	Protocadherin Fat 1 precursor	NP_114007.1	1504–1507:KKFR
H1f0	Histone H1.0	NP_036710.1	14–17:KRAK;108–111:KKTK;136–139:KKPK;144–147:KKAK;155–158:KKAK;158–161:KKPK;174–177:KKAK
Hist1h4b	Histone H4	NP_073177.1	17–20:KRHR
Hnrnpm	Heterogeneous nuclear ribonucleoprotein M	NP_001103381.1	68–71:KRYR;217–220:KKLK
Hspd1	60 kDa Heat shock protein, mitochondrial	NP_071565.2	417–420:KKDR
Jup	Junction plakoglobin	NP_112309.2	83–86:KRVR
Plec	Plectin	NP_001157780.1	667–670:KKIK;783–786:KRAK;1130–1133:KKLR;1298–1301:KKPK;1811–1814:KRQR;3859–3862:KRCR;4122–4125:KRER
Ppl	Periplakin	NP_001100446.1	7–10:KKNK;822–825:KRAR
Sptbn2	Spectrin beta chain, non-erythrocytic 2	NP_062040.2	1258–1261:KRHR;2086–2089:KRKR;2385–2388:KKNK
Taf1	Transcription initiation factor TFIID subunit 1	NP_001178652.1	167–170:KKER;282–285:KRKK;285–288:KKHR;404–407:KKLR;557–560:KKSR;666–669:KKAK;865–868:KRLK;1270–1273:KKPK;1273–1276:KKMK;1374–1377:KKKR;1375–1378:KKRR
Tdrd3	Tudor domain-containing protein 3	NP_001012043.1	439–442:KRGK
Tns3	Tensin-3	NP_001163930.1	247–250:KKFR
Vcl	Vinculin	NP_001100718.1	975–978:KRIR

Protein sequences were analyzed using ProMatch (see methods).

Table 3.

UniProt annotations of proteins involved in transcriptional regulation

Gene Symbol	UniProt Functional Annotation
Jup	Common junctional plaque protein. Membrane-associated plaques are architectural elements in an important strategic position to influence the arrangement and function of both the cytoskeleton and the cells within the tissue. The presence of Jup in both the desmosomes and in the intermediate junctions suggests that it plays a central role in the structure and function of submembranous plaques. Acts as a substrate for VE-PTP and is required by it to stimulate VE-cadherin function in endothelial cells. Can replace β-catenin in E-cadherin/catenin adhesion complexes, which are proposed to couple cadherins to the actin cytoskeleton. [Translocates to nucleus in response to vasopressin (62). Jup can activate the Wnt signaling cascade directly without interaction of β-catenin, and has multiple functions as a transcriptional activator and a cell adhesion molecule like β-catenin (45).]
Tdrd3	Scaffolding protein that specifically recognizes and binds dimethyl arginine-containing proteins. In nucleus, acts as a coactivator: recognizes and binds asymmetric dimethylation on the core histone tails associated with transcriptional activation (H3R17me2a and H4R3me2a) and recruits proteins at these arginine-methylated loci. May play a role in the assembly and/or disassembly of mRNA stress granules and in regulation of translation of target mRNAs by binding Arg/Gly-rich motifs (GAR) in dimethyl arginine-containing proteins.
Lgr4	Receptor for R-spondins that potentiates the canonical Wnt signaling pathway and is involved in the formation of various organs. Upon binding to R-spondins (RSPO1, RSPO2, RSPO3, or RSPO4), associates with phosphorylated LRP6 and frizzled receptors that are activated by extracellular Wnt, triggering the canonical Wnt signaling pathway to increase expression of target genes. Its function as activator of the Wnt signaling pathway is required for the development of various organs, including liver, kidney, intestine, bone, reproductive tract, and eye. May also act as a receptor for norrin (NDP); such results, however require additional confirmation in vivo. Involved in kidney development; required for maintaining the ureteric bud in an undifferentiated state.
Cdh1	Cadherins are calcium-dependent cell adhesion proteins. They preferentially interact with themselves in a homophilic manner in connecting cells; cadherins may thus contribute to the sorting of heterogeneous cell types. CDH1 is involved in mechanisms regulating cell-cell adhesions and mobility and proliferation of epithelial cells. It is a ligand for integrin α-E/β-7.1 E-Cad/CTF2 promotes nonamyloidogenic degradation of Aβ precursors. Has a strong inhibitory effect on APP C99 and C83 production. [Cleaved Cdh1 translocates to nucleus to regulate transcription (15).]
Cenpj	Plays an important role in cell division and centrosome function by participating in centriole duplication. Inhibits microtubule nucleation from the centrosome.
Taf1	Largest component and core scaffold of the TFIID basal transcription factor complex. Contains novel NH2- and COOH-terminal Ser/Thr kinase domains, which can autophosphorylate or transphosphorylate other transcription factors. Phosphorylates TP53 on Thr55, which leads to MDM2-mediated degradation of TP53. Phosphorylates GTF2A1 and GTF2F1 on Ser residues. Possesses DNA-binding activity. Essential for progression of the G1 phase of the cell cycle. Exhibits histone acetyltransferase activity toward histones H3 and H4.

Square brackets indicate additional material culled from the literature.

Fig. 3.

β-Catenin interactome includes members that have been previously studied relative to β-catenin and members with novel relationships to β-catenin. The value listed for each interacting protein is the number of publications found via a Pubmed search (http://www.ncbi.nlm.nih.gov/pubmed) of the gene symbol for that protein and the term “β-catenin” or “Ctnnb1”. The lengths of the edges connecting the protein to β-catenin correspond to the reciprocal of the number of publications.

Fig. 4.

Classification of β-catenin-binding proteins based on Molecular Function Gene Ontology terms. A: bar graph representation. B: 3 terms stood out, specifically transcriptional coactivator activity, polyA RNA binding, and cell-cell adhesion.

What proteins show a change in binding to β-catenin in response to vasopressin?

Next, we asked, “Among the interacting proteins found in Table 1, which show changes in binding in response to dDAVP treatment of the cells?” The dDAVP/Vehicle ratio values are shown in Table 4. Focusing only on those with a dDAVP/Vehicle ratio greater than 2 or less than 0.5, five had minimum Bayes’ factors less than 0.01 (increased: Lrba, Sptbn2, Tenm3, and Krt4; decreased: Plec) and nine had minimum Bayes’ factors less than 0.1.

Table 4.

Proteins that show a change in binding to β-catenin in response to vasopressin

Unique Symbol	Protein Name	Average (dDAVP/Vehicle)*	MBF‡	Odds Ratio?†
Lrba	Lipopolysaccharide-responsive and beige-like anchor	∞	0	∞
Sptbn2	Spectrin beta chain, non-erythrocytic 2	∞	0	∞
Tenm3	Teneurin-3	∞	0	∞
Krt4	Keratin, type II cytoskeletal 4	16.4	3.3E-05	31,000
Epb41l1	Band 4.1-like protein-1	13.7	0.042	24
Ccdc130	Coiled-coil domain-containing protein 130	6.0	0.44	2.3
Hnrnpm	Heterogeneous nuclear ribonucleoprotein M	5.4	0.54	1.8
Tns3	Tensin-3	0.5	0.081	12
Itga1	Integrin alpha-1	0.4	0.092	11
Tpi1	Triosephosphate isomerase	0.4	0.23	4.3
Alpk1	Alpha-protein kinase-1	0.4	0.14	7.2
Plec	Plectin	0.4	8.4E-15	1.2E+14
Hspb1	Heat shock protein beta-1	0.4	0.51	1.9
Lama1	Laminin subunit alpha-1	0.4	0.61	1.6
Ugt1a5	UDP-glucuronosyltransferase 1–5	0.3	0.51	1.9
Mdh2	Malate dehydrogenase, mitochondrial	0.2	0.072	14
Pkp2	Plakophilin-2	0.2	0.51	1.9
Cenpj	Centromere protein J	0.2	0.23	4.3

^*dDAVP/Control Ratio is the average ratio of normalized spectral counts (≥2 or ≤0.5) over 4 dDAVP:Vehicle pairs. dDAVP:Vehicle ratios were formed from spectral counts normalized to total counts for the sample but without any correction for spectral counts obtained from the preimmune IgG sample.

^‡Minimum Bayes’ Factor calculated as exp(−t²/2), where t is the Student’s t statistic for vasopressin vs. vehicle comparison. Boldface indicates MBF < 0.1.

^†Pr(difference)/Pr(no difference), where Pr is probability.

What protein domains are represented in β-catenin-binding partners?

Analysis of protein domains can be informative regarding protein function. Figure 5 summarizes the protein domains represented among proteins coimmunoprecipitated with β-catenin. The most prevalent domain was the armadillo domain (ARM), originally described in the β-catenin homolog in Drosophila (49). This includes sites in β-catenin itself as well as in three other catenins (δ-catenin, plakophilin-2, and Jup) and Lrba (lipopolysaccharide-responsive and beige-like anchor protein). The latter is a member of the LYST family of proteins (9), involved in coupling signal transduction and endosomal trafficking in cells. The ARM domain may play a role in nuclear translocation (16). Another prevalent domain among β-catenin-interacting proteins is the spectrin repeat domain, present in laminin subunit α1, β-spectrin, plectin, periplakin, and the intermediate filament protein keratin type I-28. Several β-catenin-interacting proteins have multiple binding domains consistent with roles in protein scaffolding.

Fig. 5.

Domain map of β-catenin-interacting proteins identify proteins with related domain structure and scaffold proteins. Blue nodes indicate domains; yellow nodes indicate β-catenin interacting proteins. Domains were identified using the Automated Bioinformatics Extractor (ABE) (see methods).

Relationship between membership in the β-catenin interactome and vasopressin-regulated phosphorylation of nuclear proteins.

We mapped the β-catenin interactome to a list of proteins in the nucleus of cultured mpkCCD cells previously identified to undergo changes in phosphorylation in response to vasopressin (5) (Table 6). Remarkably, among the 43 proteins in the IMCD nuclear β-catenin interactome, there were increases in phosphorylation at 11 distinct phosphorylation sites in seven different interacting proteins. In contrast, when 43 proteins were randomly selected from the IMCD nuclear proteome (55), none showed increased phosphorylation in the data from Bolger et al. (5) (Chi-square, P < 0.05). Thus, it appears that there may be a functional connection between vasopressin-mediated phosphorylation in IMCD cells and being a nuclear binding partner for β-catenin.

Table 6.

Several β-catenin binding partners have been shown to undergo increases in phosphorylation in the nucleus of mouse mpkCCD cells in response to the vasopressin analog dDAVP (5)

Nuclear Fraction	Gene Symbol	Protein Name	RefSeq	Phosphorylation Site	Log2 (dDAVP/Vehicle) Phosphorylation (means ± SE)	P Value
NP	Ahnak	AHNAK nucleoprotein	NP_033773	S5522	0.86 ± 0.17	0.000070
NE	Ahnak	AHNAK nucleoprotein	NP_033773	S5522	0.70 ± 0.22	0.000785
NP	Ahnak	AHNAK nucleoprotein	NP_033773	S5525	0.79 ± 0.12	0.000357
NE	Ahnak	AHNAK nucleoprotein	NP_033773	S5525	0.77 ± 0.27	0.001428
NP	Ctnnb1	Catenin beta-1	NP_031640	S552	1.79 ± 0.11	0.000000
NP	Ctnnb1	Catenin beta-1	NP_031640	S675	0.28 ± 0.03	0.000004
NP	Ctnnb1	Catenin beta-1	NP_031640	T551	1.83 ± 0.39	0.000092
NP	Ctnnd1	Catenin delta-1	NP_031641	S862	2.23 ± 1.18	0.030759
NP	Ctnnd1	Catenin delta-1	NP_031641	S864	1.78 ± 0.57	0.005715
NP	Jup	Junction plakoglobin	NP_034723	S665	0.60 ± 0.12	0.000067
NE	Krt19	Keratin, type I cytoskeletal 19	NP_032497	S67	2.00 ± 0.25	0.000004
NP	Krt19	Keratin, type I cytoskeletal 19	NP_032497	S67	1.50 ± 0.36	0.000186
NP	Pkp2	Plakophilin-2	NP_080439	S151	0.44 ± 0.19	0.004469
NE	Ppl	Periplakin	NP_032935	S1655	0.66 ± 0.50	0.087711

NE, nuclear extract; NP, nuclear pellet. Data are from phosphoproteomic analysis of mpkCCD collecting duct cells published by Bolger et al. (5).

Large-Scale Data Integration Using Bayes’ Theorem

The ChIP-MS analysis did not identify any sequence-specific transcription factors of the type that bind to regulatory elements in the promoter to activate or repress gene expression. This was likely due to the low abundances of proteins in this class relative to other types of proteins in the nucleus (29, 30), presumably placing them below the detection limit for mass spectrometry in the type of experiment described in this paper. Modeling the transcription network involved in Aqp2 gene regulation, however, requires identification of the sequence-specific transcription factors most likely to interact with the Aqp2 promoter. To identify sequence-specific transcription factors that might interact with β-catenin in the promoter region of the Aqp2 gene, we followed two steps. 1) We utilized large-scale data integration techniques using Bayes’ theorem (40, 79) to identify which of the 1,375 transcription factors annotated in the rat genome were most likely to bind to the Aqp2 promoter; and 2) we looked for prior evidence of interactions between these transcription factors and β-catenin. The algorithm for the calculation is summarized in Table 7, and the 23 top-ranked transcription factors are listed in order in Table 8. The predicted binding sites in the Aqp2 promoter are shown in Fig. 6. The transcription factors labeled in red in Fig. 6 are highly ranked transcription factors that are known to bind β-catenin and include Jun, Junb, Jund, Atf1, Atf2, Mef2d, Usf1, Max, Pou2f1, and Rxra.

DISCUSSION

Regulation of Aqp2 gene expression is a central process in the regulation of water excretion by the kidney, and failures of this process result in disorders of water balance (36). The molecular machine that links the vasopressin receptor to an increase in Aqp2 gene expression is, however, largely unknown. Model building is an essential technique in the analysis of complex systems like the transcriptional regulatory network in collecting duct cells. Modeling occurs in at least two steps: 1) identification of key components of the system, and 2) identification of the causal interactions among the components to create a predictive model. This paper begins to address the first step.

We previously identified an important transcriptional coregulator, β-catenin, that is regulated by vasopressin in two ways. 1) Vasopressin binding to the V2 receptor in collecting duct cells increases the phosphorylation of β-catenin at Ser⁵⁵²; and 2) vasopressin stimulates the translocation of β-catenin into the nuclei of collecting duct cells. These effects are likely to be independent of the role of β-catenin in the canonical Wnt pathway, which is a key component of collecting duct development (63). Thus, we make the assumption that β-catenin may be involved in regulation of Aqp2 gene transcription and ask the question, “What proteins interact with β-catenin in nuclear fractions of collecting duct cells; and how might the interactions affect Aqp2 gene transcription?” To address these, we Semployed both an experimental and a computational approach, yielding potentially complementary information.

Nuclear Proteins That Bind β-Catenin

The experimental approach combined chromatin immunoprecipitation with modern mass spectrometry to identify 43 β-catenin-interacting proteins in the nuclei of IMCDs from rats. We used a standard approach for enrichment of nuclei from IMCDs and ChIP as provided in a commercial kit. The principle of the nuclear isolation is to use gentle detergent treatment to remove the plasma membrane, allowing liberation of cytosol and all membrane structures not tethered to the nucleus. What is left behind is generally nucleus and a good share of the endoplasmic reticulum, which includes the outer nuclear membrane. Since the outer membrane of the nucleus is part of the endoplasmic reticulum, it is likely that some of the identified β-catenin interactions occur within the endoplasmic reticulum rather than on chromatin. However, because the approach identified a number of proteins already known to interact with β-catenin in the nucleus and identified many of the proteins previously seen to translocate to the nucleus in response to vasopressin (62), it is reasonable to surmise that some of the novel β-catenin-interacting proteins also play roles in the nucleus, especially those that are known to be nuclear resident proteins. With regard to junctional proteins that were isolated, it seems possible that the identified interactions occur either in the endoplasmic reticulum or in the nucleus. The literature is replete with examples of junctional proteins that migrate into the nucleus (2, 12, 48, 70), so that the finding of interactions in nuclei cannot be regarded as unexpected.

A previous study also identified β-catenin-interacting proteins using a similar ChIP-MS protocol in mouse embryonic stem cells (80). Among the 43 β-catenin-interacting proteins found in the present study, 25 were identified in the previous study and 18 were unique to our study [These are: Tenm3, Plekha6, Lrba, Sptbn2, Tdrd3, Taf1, Epb41l1, Afdn, Ugt1a5, Pkp2, Pithd1, Etl4, Dnhd1, Ccdc130, Lgr4, Cenpj, Alpk1 and H1f0 (see Table 5 for annotations.)]. Many of the identified β-catenin-interacting proteins had been found in other studies to undergo vasopressin-mediated translocation into the nucleus (62) and/or vasopressin-mediated changes in phosphorylation (5), consistent with an important role for β-catenin in vasopressin-mediated regulation of transcription in renal collecting duct cells.

Table 5.

UniProt annotations for proteins that show a change in binding to β-catenin in response to vasopressin

Gene Symbol	UniProt Annotation
Plec	Interlinks intermediate filaments with microtubules and microfilaments and anchors intermediate filaments to desmosomes or hemidesmosomes. Could also bind muscle proteins such as actin to membrane complexes in muscle. May be involved not only in the filaments network but also in the regulation of their dynamics. Structural component of muscle. Isoform 9 plays a major role in the maintenance of myofiber integrity.
Cdh1	See Table 3.
Krt4	Cytokeratin-4.
Lrba	May be involved in coupling signal transduction and vesicle trafficking to enable polarized secretion and/or membrane deposition of immune effector molecules. [Involved in endosomal traffic and may be involved in efficient delivery of cell surface proteins to the lysosome (11).].
Sptbn2	Probably plays an important role in membrane skeleton; interacts with Epb41l1. [Spectrin is a cytoskeletal protein that lines the intracellular side of the plasma membrane in eukaryotic cells. Spectrin forms pentagonal or hexagonal arrangements, forming scaffolding and playing an important role in maintenance of plasma membrane integrity and cytoskeletal structure (27); interacts with Epb41l1].
Tenm3	Integral membrane protein involved in development. Promotes homophilic cell adhesion. May function as a cellular signal transducer.

Square brackets indicate additional material culled from the literature.

One particularly interesting interacting protein is Taf1 (TATA-binding protein-associated factor). It is a component of the transcriptional initiation complex TFIID (TFIID subunit 1), and theoretically, regulation of transcriptional initiation could be exerted through its binding to β-catenin. Recently, Sandoval et al. (60) demonstrated that vasopressin causes an almost global increase in transcriptional initiation in mpkCCD collecting duct cells. In previous studies, Taf1 abundance in the nuclear extract fraction of mpkCCD was seen to increase by 68% in response to vasopressin (https://hpcwebapps.cit.nih.gov/ESBL/Database/QuantNucProteomics/), providing a potential explanation for the increase in transcriptional initiation. Whether this translocation event is mediated by Taf1 binding to β-catenin remains unknown. However, another member of the RNA-polymerase II preinitiation complex, Tbp, is known to bind directly to β-catenin, allowing β-catenin to play a role in transcriptional initiation (24).

β-Catenin Interacting Proteins with Indirect Roles in Transcriptional Regulation

Several of the 43 β-catenin-interacting proteins in Table 1 have been previously shown to be involved in transcriptional regulation. Aside from Taf1, several other β-catenin-binding partners identified in this study may have indirect roles in transcriptional regulation, namely Tdrd3, Jup, Cdh1, and Cenpj. We discuss these in the following.

Tdrd3 (Tudor domain-containing protein) is a transcriptional coactivator that binds to dimethyl arginine moieties on histone tails and also interacts with the asymmetrically dimethylated COOH-terminal domain of RNA Pol II (65). TDRD3 protein was found to be mainly recruited near the transcriptional start sites of many genes in a recent ChIP-Seq analysis (83) and, along with Taf1, could play a role in the global effect of vasopressin to increase transcriptional initiation (60).

Jup (junction plakoglobin), also known as γ-catenin, is a paralog of β-catenin. Considering it contains the conserved 12 armadillo repeats as β-catenin, Jup may fulfill some of the same roles, including as a transcriptional coregulator. Interestingly, as seen with Ser⁵⁵² of β-catenin, vasopressin increases phosphorylation of Jup in mouse mpkCCD cells at Ser⁶⁶⁵ (5), a basophilic site (RKRVS). This sequence is similar to that seen with many protein kinase A substrates (66, 75, 82).

Cdh1 (E-cadherin) is the prototypical epithelial cadherin. It is an integral membrane protein that is the central component of the adherens junction complex, which includes β-catenin. There are two possible explanations for finding E-cadherin in the nuclear fraction. 1) Being an integral membrane protein, β-catenin is translated on the rough endoplasmic reticulum; it may therefore be present in intact form in the nuclear fraction by virtue of the fact that the endoplasmic reticulum is an extension of the outer nuclear membrane, allowing it to be isolated with the nuclei. 2) Cdh1 may be present in the nuclear fraction when cleaved proteolytically to yield a polypeptide that is not membrane bound (43) and can be translocated to the nucleus, where it can alter intranuclear signaling (15).

Cenpj (centromere protein J) plays a key role in cell division as a component of a ternary complex that mediates centriole duplication. In interphase, it has been shown to play a role as a transcriptional coregulator that binds to the transcription factors STAT5 (54) and NF-κB (37). The latter is known to be involved in regulation of Aqp2 gene transcription (21). Cenpj has previously been shown to bind to band 4.1 (Epb41) (28), a paralog of another protein identified here as a β-catenin-interacting protein, specifically band 4.1-like protein-1.

Sequence-Specific Transcription Factors Likely to Bind the Promoter of the Aqp2 Gene and β-Catenin

The computational approach employed in this study used modern large-scale data integration techniques to identify a set of 10 sequence-specific transcription factors that are likely to bind to the promoter of the Aqp2 gene and to bind β-catenin, specifically Jun, Junb, Jund, Atf1, Atf2, Mef2d, Usf1, Max, Pou2f1, and Rxra. The first five of these are b-ZIP family transcription factors that are predicted to bind to one or more of three AP-1/CREB-motifs (T00031, T00124, T00164; Fig. 6). AP-1 components have previously been proposed to be involved in regulation of Aqp2 gene expression (64, 84). Medf2d is predicted to bind to a SRF motif (T00766, Fig. 6) that has been described previously (68, 86). Usf1 and Max bind E-box elements (T00875, T02115; Fig. 6), which may be involved in global amplification of gene expression (42). The canonical heterodimerization partner of Max is Myc, which was found to have extremely low protein and mRNA abundance. Pou2f1 is a homeobox transcription factor that is predicted to bind to a well-described Hox-binding site (T03458; Fig. 6) in the Aqp2 promoter. Finally, Rxra is a ligand activated nuclear receptor predicted to bind to a conserved nuclear receptor binding site (T00042; Fig. 6).

It is worthwhile to point out a couple of interesting aspects of the general analysis of TFs that could regulate Aqp2 gene transcription that are independent of β-catenin. First, Creb1, which is often listed as a likely regulator of Aqp2 gene expression in review articles (22, 46, 76), is ranked very low relative to other b-ZIP family transcription factors, including Crem, Jun family (Jun, Junb, and Jund), Fos family (Fos, Fosb, Fosl1, and Fosl2), and Atf family (Atf1, Atf2, and Atf3). Its low ranking is primarily because Creb1 mRNA was not detected in the transcriptome data set from microarray assay of rat IMCD, in contrast to other Creb proteins (Creb3, Crebl1, and Crebl2). Second, the analysis predicts that the mineralocorticoid receptor Nr3c2 may play a role in regulation of Aqp2 gene expression. This is a ligand-activated transcription factor that is known to regulate expression of the gene coding for the α subunit of the epithelial sodium channel (ENaC) (4). Although several studies on mineralocorticoid-mediated Aqp2 gene expression have been reported (23, 31, 38, 53), addressing the possibility of the importance of the mineralocorticoid receptor relative to other transcription factors has not been established.

Previous studies have established the idea that multiple transcription factors are required to exert precise transcriptional control (19). Aqp2 is a cell type-specific gene that is expressed virtually exclusively in collecting duct principal cells and IMCD cells (52). Moreover, Aqp2 gene expression is also strongly regulated by vasopressin as a result of selective transcriptional elongation (60). Both cell type-specific expression and regulation by vasopressin are likely to occur largely as a consequence of binding of as-yet-unknown combinations of sequence-specific transcription factors to the gene promoter.

Transcription Factors That May Regulate Aqp2 Gene Transcription Independently of β-Catenin

Although the main focus of this paper is on β-catenin-interacting proteins, for completeness it is worth mentioning that several transcription factors may be involved in vasopressin-regulated Aqp2 gene transcription independently of β-catenin. These include some of the AP-1 components (64), nuclear receptors (8, 74), members of the NFAT family (20, 21, 41), GATA family (71, 85), and Ets (E26 transformation-specific) family (68, 86). See supplemental discussion, linked to this paper online, for details.

β-Catenin-Interacting Proteins and Vasopressin-Mediated Regulation of Nuclear Translocation

In a prior study, β-catenin was seen to translocate into the nuclei of mouse collecting duct mpkCCD cells in response to vasopressin in association with an increase in β-catenin phosphorylation at Ser⁵⁵² (62). It is interesting that several of the identified β-catenin-interacting proteins also translocated into the nucleus in response to vasopressin as determined by quantitative mass spectrometry of nuclear fractions (62) or the 1,000-g fraction from differential centrifugation of mpkCCD cell homogenates (81). These are phakophilin-2 (Pkp2), the multifunctional protein Annak, junction plakoglobin (Jup or γ-catenin), periplakin (Ppl), β-spectrin (Sptbn2), α-catenin, δ-catenin, and E-cadherin (Table 9). β-Catenin, which lacks a classical nuclear localization signal sequence, is believed to move into the nucleus by binding directly to the nuclear pore machinery via its multiple ARM domains (14). Among the proteins that are present in the β-catenin interactome and translocate into the nucleus in response to vasopressin, Jup, Pkp2, and δ-catenin also have ARM domains (Fig. 5). It seems possible that some or all of these components move together into the nucleus bound to β-catenin. Indeed, it has been suggested that binding relationships among different types of catenins that exist at the adherens junction and desmosomes may be conserved in nuclei as part of their coordinate roles in transcriptional regulation (47). β-Catenin contains multiple ARM domains that can interact with the nuclear pore complex, facilitating its entry (despite the absence of a nuclear localization signal) (14), and this interaction may hypothetically facilitate nuclear entry of its binding partners, as has been demonstrated for the transcription factor Lef1 (1). Among the proteins previously found to translocate to the nucleus in response to vasopressin, phakophilin-2 (Pkp2), junction plakoglobin (Jup), and δ-catenin, all possess ARM domains. In mouse mpkCCD cells, vasopressin markedly increases phosphorylation of β-catenin at Ser⁵⁵² (3, 5, 58, 62), a basophilic site (sequence RRTS) that is believed to be a target for protein kinase A (34). This site is located within the 10th ARM domain, i.e., within the region thought to interact with the nuclear pore complex. Thus, it is possible that vasopressin-induced nuclear translocation of β-catenin is triggered by the increase in Ser⁵⁵² phosphorylation. In addition, the ARM10-12 domains of β-catenin interact with other cofactors, such as CREBBP and p300 (51), which are involved in transcriptional control through histone acetylation. CREBBP has been found to be translocated into the nucleus in response to vasopressin in cultured collecting duct cells (62), although it was not identified as a β-catenin-binding partner in the present study. β-Catenin-mediated transactivation occurs, in part, through its ability to bind CREBBP (25, 67), which acetylates histone H3 at Lys²⁷, thereby increasing chromatin accessibility locally (69).

Table 9.

β-Catenin-interacting proteins: vasopressin responses in prior studies

Gene Symbol	Annotation	Regulation by Vasopressin
Ctnna1	Catenin alpha-1	dDAVP at a concentration of 0.1 nM for 1 h increases abundance of Ctnna1 in nuclear pellet of mouse mpkCCD to 150% (62). dDAVP at a concentration of 1 nM for 0–15 min decreases phosphorylation of Ctnna1 protein at S643 protein in rat IMCD to 70% (26).
Ctnnd1	Catenin delta-1	Ctnnd1 protein undergoes a reciprocal abundance change in the 17,000 g membrane fraction [down] and the 1,000-g pellet fraction [up] in response to vasopressin (81). dDAVP at a concentration of 0.1 nM for 1 h increases phosphorylation of Ctnnd1 protein at S862 in nuclear pellet of mouse mpkCCD to 460% (5). dDAVP at a concentration of 0.1 nM for 1 h increases phosphorylation of Ctnnd1 protein at S864 in nuclear pellet of mouse mpkCCD to 350% (5). dDAVP at a concentration of 1 nM for 0.5 h decreases phosphorylation of Ctnnd1 protein at S899 in mouse mpkCCD to 80% (58).
Cdh1	Cadherin-1	Cdh1 protein undergoes a reciprocal abundance change in the 200,000-g cytosolic fraction [down] and the 1,000-g pellet fraction [up] in response to vasopressin (81).
Lama1	Laminin subunit alpha-1	None
Hspd1	60 kda heat shock protein, mitochondrial	None
Tenm3	Teneurin-3	None
Plekha6	Pleckstrin homology domain-containing family A member 6	dDAVP at a concentration of 1 nM for 0–15 min increases phosphorylation of Plekha6 at S970 in rat IMCD to 140% (26).
Akr1b1	Aldose reductase	dDAVP at a concentration of 5 ng/h to rats for 72 h decreases abundance of Akr1b1 protein in rat IMCD to 50% (73).
Lrba	Lipopolysaccharide-responsive and beige-like anchor protein	dDAVP at a concentration of 5 ng/h to rats for 72 h decreases abundance of Lrba protein in rat IMCD to 70% (56). dDAVP at a concentration of 0.1 nM for 120 h decreases abundance of Lrba protein in mouse mpkCCD to 80% (35). dDAVP at a concentration of 1 nM for 0–15 min increases phosphorylation of Lrba protein at T1231 protein in rat IMCD to 140% (26).
Tpi1	Triosephosphate isomerase	dDAVP at a concentration of 5 ng/h to rats for 72 h decreases abundance of Tpi1 protein in rat IMCD to 40% (56).
Sptbn2	Spectrin beta chain, non-erythrocytic 2	Sptbn2 protein undergoes a reciprocal abundance change in the 200,000-g cytosolic fraction [down] and the 1,000-g pellet fraction [up] in response to vasopressin (81). dDAVP at a concentration of 5 ng/h to rats for 72 h increases abundance of Sptbn2 protein in rat IMCD to 160% (56).
Tdrd3	Tudor domain-containing protein 3	None
Taf1	Transcription initiation factor TFIID subunit 1	None
Vcl	Vinculin	dDAVP at a concentration of 0.1 nM for 120 h decreases abundance of Vcl protein in mouse mpkCCD to 80% (35).
Hspb1	Heat shock protein b-1	dDAVP at a concentration of 5 ng/h to rats for 72 h increases abundance of Hspb1 protein in rat IMCD to 250% (56). dDAVP at a concentration of 1 nM for 0–15 min increases phosphorylation of Hspb1 protein at S13 protein in rat IMCD to 120% (26). dDAVP at a concentration of 1 nM for 0–15 min increases phosphorylation of Hspb1 protein at S85 protein in rat IMCD to 130% (26).
Fat1	Protocadherin Fat 1	None
Epb41l1	Band 4.1-like protein 1 isoform S	None
Afdn	Afadin	None
Ugt1a5	UDP-glucuronosyltransferase 1–5	None
Pkp2	Plakophilin-2	dDAVP at a concentration of 0.1 nM for 1 h increases abundance of Pkp2 in nuclear pellet of mouse mpkCCD to 200% (62). Pkp2 protein undergoes a reciprocal abundance change in the 17,000-g membrane fraction [down] and the 1,000-g pellet fraction [up] in response to vasopressin (81). Pkp2 protein half-life increases from 17.5 to 29.0 h in mouse mpkCCD cells in response to vasopressin (61).
Pithd1	PITH domain-containing protein 1	None
Etl4	Sickle tail protein homolog	None
Mdh2	Malate dehydrogenase, mitochondrial	None
Dnhd1	Dynein heavy chain domain-containing protein 1	None
Hnrnpm	Heterogeneous nuclear ribonucleoprotein M	None
Ahnak	Neuroblast differentiation-associated protein AHNAK	Ahnak protein undergoes a reciprocal abundance change in the 200,000-g cytosolic fraction [down] and the 1,000-g pellet fraction [up] in response to vasopressin (81). dDAVP alters phosphorylation of Ahnak at multiple sites in rat IMCD (26). dDAVP at a concentration of 0.1 nM for 1 h increases abundance of Ahnak in nuclear pellet of mouse mpkCCD to 150% (62). dDAVP aalters phosphorylation of Ahnak protein at multiple sites in nuclei of mouse mpkCCD (5).
Ccdc130	Coiled-coil domain-containing protein 130	None
Tns3	Tensin-3	dDAVP at a concentration of 1 nM for 0–15 min increases phosphorylation of Tns3 protein at S864 protein in rat IMCD to 260% (26).
Lgr4	Leucine-rich repeat-containing G-protein coupled receptor 4	None
Cenpj	Centromere protein J	None
Jup	Junction plakoglobin	dDAVP at a concentration of 0.1 nM for 1 h increases abundance of Jup in nuclear pellet of mouse mpkCCD to 160% (62). Jup protein undergoes a reciprocal abundance change in the 17,000-g membrane fraction [down] and the 1,000-g pellet fraction [up] in response to vasopressin (81). Jup protein half-life increases from 13.3 to 23.6 h in mouse mpkCCD cells in response to vasopressin (61). dDAVP at a concentration of 0.1 nM for 1 h increases phosphorylation of Jup protein at S665 in nuclear pellet of mouse mpkCCD to 150% (5). dDAVP at a concentration of 0.1 nM for 120 h decreases translation rate of Jup protein in mouse mpkCCD to 50% (61).
Plec	Plectin	dDAVP at a concentration of 0.1 nM for 1 h increases abundance of Plec in nuclear pellet of mouse mpkCCD to 150% (62). dDAVP at a concentration of 1 nM for 0–15 min decreases phosphorylation of Plec protein at S21 protein in rat IMCD to 80% (26). dDAVP at 0.5 ng/h for 7 days increases mRNA abundance of Plec in AQP1 KO mouse renal inner medulla to 150% (7).
Itga1	Integrin alpha-1	Itga1 protein half-life decreases from 52.8 to 10.6 h in mouse mpkCCD cells in response to vasopressin (61).
Alpk1	Alpha-protein kinase 1	None
Ppl	Periplakin	dDAVP at a concentration of 0.1 nM for 1 h increases abundance of Ppl in nuclear extract of mouse mpkCCD to 170% (62). Ppl protein undergoes a reciprocal abundance change in the 200,000-g cytosolic fraction [down] and the 1,000-g pellet fraction [up] in response to vasopressin (81). AVP at 10 nM for 4 h decreases mRNA abundance of Ppl in mouse mpkCCDcl4 (SAGE) to 0% (59). dDAVP at a concentration of 1 nM for 0.5 h decreases phosphorylation of Ppl protein at S14 in mouse mpkCCD to 20% (5).
Krt4	Keratin, type II cytoskeletal 4	None
Krt76	Keratin, type II cytoskeletal 2	None
Hist1h4b	Histone H4	None
H1f0	Histone H1.0	dDAVP at a concentration of 0.1 nM for 120 h decreases abundance of H1f0 protein in mouse mpkCCD to 70% (35).
H2afz	Histone H2A.Z	None
Krt19	Keratin, type I cytoskeletal 19	dDAVP at a concentration of 5 ng/h to rats for 72 h increases abundance of Krt19 protein in rat IMCD to 170% (56). dDAVP at a concentration of 1 nM for 0–15 min increases phosphorylation of Krt19 protein at S67 protein in rat IMCD to 150% (26). dDAVP at a concentration of 0.1 nM for 1 h increases phosphorylation of Krt19 protein at S67 in nuclear pellet of mouse mpkCCD to 400% (5). dDAVP at a concentration of 0.1 nM for 1 h increases phosphorylation of Krt19 protein at S67 in nuclear extract of mouse mpkCCD to 280% (5).
Krt28	Keratin, type I cytoskeletal 28	None

Data mined from prior publications using Biological Information Gatherer (BIG) (see methods).

Significance of the Study

Prior studies have identified β-catenin as a target for vasopressin regulation via phosphorylation (3, 5, 58, 62) and nuclear translocation (62). Other studies have shown that vasopressin regulates transcription of Aqp2 and other genes in collecting duct cells (60). Given known roles in transcriptional regulation, it seems likely that β-catenin is involved in the transcriptional response to vasopressin. It is likely that β-catenin interacts with specific proteins in the nucleus that mediate changes in transcription, including the proteins identified in this study. It is appears unlikely that these proteins include any members of the TCF/LEF family of transcription factors that are involved in canonical β-catenin signaling, since there are no obvious TCF/LEF-binding motifs in the promotor of the Aqp2 gene (68, 86), and none of the proteins found in this study are TCF/LEF family members. However, the proteins identified herein provide new candidates for vasopressin-regulated, β-catenin-mediated transcriptional regulation to be pursued in future studies using both computational and experimental approaches.

GRANTS

This work was primarily funded by the Division of Intramural Research, National Heart, Lung, and Blood Institute (Project ZIA HL-001285 and ZIA HL-006129, M.A. Knepper). Protein mass spectrometry was done in the NHLBI Proteomics Core Facility (M. Gucek, Director). J. Hwang was supported by the American Physiological Society Undergraduate Summer Research Fellowships Program (May–August, 2016). H. J. Jung is the recipient of the Korean Visiting Scientist the Training Award, supported by the Korea Health Industry Development Institute of the Korea Ministry of Health and Welfare in cooperation with National Institutes of Health (HI13C1211).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

J.R.H., C.-L.C., and H.J.J. performed experiments; J.R.H., C.-L.C., B.M., M.A.K., and H.J.J. analyzed data; J.R.H., C.-L.C., M.A.K., and H.J.J. interpreted results of experiments; J.R.H., C.-L.C., B.M., and H.J.J. prepared figures; J.R.H., C.-L.C., B.M., M.A.K., and H.J.J. edited and revised manuscript; J.R.H., C.-L.C., B.M., M.A.K., and H.J.J. approved final version of manuscript; C.-L.C., M.A.K., and H.J.J. conceived and designed research; C.-L.C., M.A.K., and H.J.J. drafted manuscript.