|Home | About | Journals | Submit | Contact Us | Français|
Vasopressin acts on the inner medullary collecting duct (IMCD) in the kidney to regulate water and urea transport. To obtain a ‘parts list’ of gene products expressed in the IMCD, we carried out mRNA profiling of freshly isolated rat IMCD cells using Affymetrix Rat 230 2.0 microarrays with approximately 31,000 features. 7913 annotated transcripts were found to be expressed above background in the IMCD cells. We have created a new online database (the “IMCD Transcriptome Database” †) to make the results publicly accessible. Among the 30 transcripts with the greatest signals on the arrays were three water channels: aquaporin-2, aquaporin-3, and aquaporin-4, all of which have been reported to be targets for regulation by vasopressin. In addition, the transcript with the greatest signal among members of the solute carrier (SLC) family of genes was the UT-A urea transporter (Slc14a2), which is also regulated by vasopressin. The V2 vasopressin receptor was strongly expressed, but the V1a and V1b vasopressin receptors did not produce signals above background. Among the 200 protein kinases expressed, the serum-glucocorticoid regulated kinase (Sgk1) had the greatest signal intensity in the IMCD. WNK1 and WNK4 were also expressed in the IMCD with a relatively high signal intensity as was protein kinase A (β-catalytic subunit). In addition, a large number of transcripts corresponding to AKAPs and 14-3-3 proteins (phospho-S/T binding proteins) were expressed. Altogether, the results combine with proteomics studies of the IMCD to provide a framework for modeling complex interaction networks responsible for vasopressin action in collecting duct cells.
The inner medullary collecting duct (IMCD) is the terminal portion of the collecting duct system of the kidney. The collecting duct system represents the final site of adjustment of urinary composition and volume, and therefore is critical for extracellular fluid homeostasis. An important regulator of collecting duct transport function is vasopressin, which controls both water and urea transport (21; 53)
To address the mechanisms of vasopressin signaling in the renal IMCD, we have been following a “systems–biology” approach consisting of identification of the component proteins via protein mass spectrometry-based analysis (2; 29; 30; 62; 71; 102), and antibody-based quantification of protein abundance (41), coupled to computational analysis of the identified proteins to discover signaling networks involved in IMCD regulation (30). An important product of these studies is the generation of a ‘parts list’, the IMCD proteome, that includes the signaling apparatus (http://dir.nhlbi.nih.gov/labs/lkem/rm/proteomics_db.asp). The IMCD proteome database currently is made up of 2338 proteins identified at high stringency.
Despite the success so far of the proteomics approach, proteomics methodologies have limitations that make it unlikely that all of the proteins that play important roles in vasopressin signaling will be found. Limited sensitivity is the chief factor. In the present study, in order to augment our enumeration of an ‘IMCD parts list’, we have pursued an analysis of the IMCD transcriptome using Affymetrix oligonucleotide arrays. The studies were carried out in native IMCD cells freshly isolated from rat inner medullas. The arrays contain approximately 31,000 features representing most of the expressed genes in the rat genome. Although the approach is comprehensive, the chief objective is to determine what transcripts are expressed in functional categories most relevant to vasopressin signaling related to short- and long-term regulation of water and urea transport in the IMCD.
The new data from the present study add to transcriptomic data obtained in microdissected cortical and outer medullary collecting ducts using a micro-SAGE technique (9). SAGE data, however, are not available for the IMCD. Early approaches to transcriptional profiling in IMCD involved large-scale sequencing of cDNAs prepared from microdissected IMCD segments, an approach that was informative but not comprehensive (78). A number of studies have also reported transcriptional profiling of whole inner medulla (7; 8; 51; 66; 80), which contain IMCDs as well as loops of Henle, vasa recta, capillaries and interstitial cells. IMCDs make up about 40% of the inner medullary tissue by volume (40). In the present study, we carry out transcriptional analysis of biochemically enriched IMCD cells isolated from rat renal inner medulla.
Pathogen-free male Sprague-Dawley rats weighing 100–120 grams (Taconic Farms Inc. Germantown, NY) were maintained on an autoclaved pelleted rodent diet and ad libitum drinking water (NHLBI Animal Care and Use Committee approved protocol number H-0110). The age of the rats was chosen to match the age of rats typically used for our isolated perfused tubule studies (12).
In order to assess purity of IMCD samples, aliquots from each cell suspension were loaded onto SDS-polyacrylamide gels and immunoblotting was carried out as described (12). An immunoblot for a collecting duct marker AQP2 (Figure 2 panel A) demonstrated that AQP2 was indeed highly enriched in the IMCD fraction and depleted from the non-IMCD fraction (Band densities: IMCD fraction, 412; non-IMCD fraction, 45). An immunoblot for AQP1, a marker of vasa recta and thin descending limbs of Henle (Figure 2 panel B), demonstrated that there was relatively little contamination of the IMCD fraction from non-IMCD elements (Band densities: IMCD fraction, 1.4; non-IMCD fraction, 29.4). Lack of contamination by outer medullary cells was demonstrated by immunoblotting using an antibody to NKCC2, a marker of the outer medullary thick ascending limb (Figure 2 panel C) (Band densities: IMCD fraction,0.9; non-IMCD fraction, 0.4; OMCD fraction, 11.3).
Total RNA of IMCD and non-IMCD cells was extracted using Trizol reagent (Invitrogen, Carlsbad, CA) following the manufacturer’s protocol. Full details are reported in the Supplementary Methods. Typically, for IMCD, the yield was 1 μg per kidney and, for non-IMCD, the yield was 0.7 μg per kidney.
Hybridization probes (labeled cDNAs corresponding to the transcripts) were prepared and hybridized to the Affymetrix Rat Genome 230 2.0 microarrays (Affymetrix Inc, Santa Clara, CA) according to the manufacturer’s protocols. See Supplementary Methods for full details.
The total number of genes with the Affymetrix Detection p-value consistent with a “Present Call” (detected) was about 17,000 from 31,000 features in the Affymetrix Rat Genome 230 2.0 array for both IMCD and non-IMCD samples. These 17,000 features corresponded to about 10,200 distinct genes. The remaining 14,000 features were identified to have “Absent” (not detected) or “Marginal” Calls based on the Detection p-value from Affymetrix GCOS software. Intensities of each probe set were normalized by median-transformed normalization.
The IMCD and non-IMCD samples contain a relatively small amount of contaminating cells from non-IMCD elements and IMCD elements, respectively (see AQP1 and AQP2 immunoblots in Figure 2). We calculated corrected values for mRNA signals using estimates of cross contamination. The appropriate equations are (see Supplementary Methods for derivation):
IMCDcorrected = (d·IMCDobserved — b·non–IMCDobserved)/(ad–bc)(Equation 1)
non–IMCDcorrected = (a·non–IMCDobserved — c·IMCDobserved)/(ad–bc)(Equation 2)
IMCDobserved = the normalized signal intensity for a given feature for IMCD sample;
non–IMCDobserved = the normalized signal intensity for the same feature for the corresponding non–IMCD sample;
a = fraction of the total signal in IMCD samples from IMCD cells
b = fraction of the total signal in IMCD samples from non–IMCD cells
c = fraction of the total signal in non–IMCD samples from IMCD cells
d = fraction of the total signal in non–IMCD samples from non–IMCD cells.
Note also that a + b = 1 and c + d = 1.
The cross contamination values a, b, c, and d could in principle be estimated from the AQP1 and AQP2 immunoblots but a more general approach is to use the array data to determine the required cross contamination correction. Supplementary Figure 3 shows a histogram of the ratios of normalized, but uncorrected, IMCD-to-non-IMCD signal ratios for individual features on the array. In theory, the maximum and minimum ratios should be bounded at a distinct value determined by the degree of cross contamination. The maximum and minimum ratios in principle could then be used to estimate the degree of cross contamination. However, as seen in Supplementary Figure 3, there is considerable splay in the tails of the ratio histogram, indicating that a direct choice of single unique values for cross contamination is not feasible. Consequently, we chose an alternative approach in which we chose a set of transcripts corresponding to IMCD and non-IMCD marker proteins. These are transcripts known to be exclusively or predominantly expressed in the IMCD (“IMCD markers) and transcripts known to be expressed in non-IMCD structures but not in the IMCD itself (“non-IMCD markers”). Thus, we choose values for a, b, c, and d such that the mean value for [non-IMCDcorrected] is zero for a selected set of IMCD markers and the mean value for [IMCDcorrected] is zero for a selected set of non-IMCD markers. For IMCD markers, we have chosen aquaporin-3 (Aqp3), gamma ENaC (Scnn1g), ClC-K2 (Clcnkb), the vasopressin V2 receptor (Avpr2) and frizzled receptor 1 (Fzd1). For non-IMCD markers, we have chosen UT-B (Slc14a1, expressed in vasa recta), NHE3 (Slc9a3, expressed in thin descending limb of Henle), von Willebrand’s factor (Vwf, expressed in endothelial cells of vasa recta) and EEAC1 (Slc1a1, expressed in thin limbs of Henle’s loop). Calculated in this way, the resulting values for b, c, and d were: a = 0.80 ± 0.00, b = 0.20 ± 0.00, c = 0.27 ± 0.01 and d = 0.73 ± 0.01.
Equations 1 and 2 were applied to each feature to gain corrected signal intensities that would be expected if the samples were pure IMCD and non-IMCD suspensions. The signal intensities for IMCD and non-IMCD are reported as mean ± SE for all replicates.
The general strategy employed in this study was to use Affymetrix microarrays to profile gene expression at an mRNA level in rat IMCD cells, focusing on the general question: What gene products can be identified in the IMCD that could be components of the pathways and networks involved in regulation of aquaporin-2 and osmotic water permeability in the IMCD? The data presented are calculated from the results of 6 microarray chip sets, i.e. 3 pairs of IMCD vs. non-IMCD transcripts. Total RNA samples for each pair were isolated from 7 rats (14 inner medullas). A full listing of the results is shown in Supplementary Table S1 and the full results have also been deposited in the GEO database (GSE7891: at http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?token=rbyjrsiskikmeze&acc=GSE7891). 7913 annotated transcripts (compared to 10,206 total transcripts above background) were found to be expressed above background in the IMCD cells.
We have created a new online database (the “IMCD Transcriptome Database” at http://dir.nhlbi.nih.gov/papers/lkem/imcdtr/) to provide ready access to these data. This site also includes the data in a downloadable format and a BLAST facility for searching for a particular transcript in the database.
The variability of the normalized, corrected signals for the three experiments is shown as scattergrams in Supplementary Figure 4. As can be seen, the distribution of IMCD and non-IMCD signals were reproducible with more than 85% of IMCD features and more than 80% of non-IMCD features showing coefficients of variation (CV = variance/mean ×100%) less than 15%. Supplementary Figure 5 shows mean corrected, normalized signals for IMCD plotted against the non-IMCD values for the same features. Values that deviate from the diagonal indicate transcripts selectively expressed in IMCD (above line) or non-IMCD elements of inner medulla (below line). Supplementary Figure 6 shows the corrected values for IMCD plotted against the corresponding non-corrected values. As can be seen, the corrections were generally relatively small. Only 0.8 percent of the values had correction factors of 50% or more of the uncorrected values.
In the following, we present a series of tables showing, for specific categories of gene products, what transcripts in IMCD cells are associated with the highest signals on the microarrays along with comments regarding gene products that are thought to be involved in vasopressin action in the IMCD. We also include significant negatives in which corrected signals in the IMCD were indistinguishable from zero. The classification of expressed genes in each table was based on classification terms from the PANTHER Database (83) and the Rat Genome Database (87). We emphasize that the signals from the array provide a general index of relative expression levels for individual transcripts but do not necessarily correlate with the absolute level of the specific mRNA in the cells. Furthermore, transcript level does not necessarily correlate with protein expression level owing to differences in translation rate and protein stability among various proteins (7).
Table 1 summarizes the microarray results for G-protein coupled receptors. Normalized and corrected intensity values reported in all tables are calculated from the raw data as described in the Methods section. The value of 1.0 represents the median value for all positive transcripts in the IMCD. A number of receptors with well established roles in the IMCD were identified and had substantial signals on the arrays. These were the vasopressin V2 receptor, the frizzled receptor 1, the PGE2 receptor EP1, the endothelin B receptor, the adenosine A1 receptor, the adrenergic receptor 2a, the frizzled receptor 6, the adrenergic receptor 1d, the PGE2 receptor EP4, and the nucleotide receptor P2Y2. Of these, only the vasopressin V2 receptor (53) and the prostaglandin E2 receptor EP4 (5) are known to be coupled to the heterotrimeric G-protein Gs and to generate cyclic AMP when bound to ligand. Presumably, PGE2 does not increase water permeability in the IMCD because of the prevailing inhibitory effect of the EP1 receptor, which signals via Gαq. In addition to receptors with well-established roles, a number of other G-protein coupled receptors were identified in the IMCD (Table 1), providing hypotheses for further study. For example, the chemokine orphan receptor 1 (Cmkor1) is the highest ranked receptor from the point of view of normalized intensity.
In our previous study, vasopressin-induced cAMP production was inhibited by epinephrine or the α2-agonist clonidine, but not by the α1-agonist, phenylephrine (46). This finding is consistent with the present finding of dominant expression of the α2a-adrenergic receptor relative to the α1-adrenergic receptor as shown in Table 1. In contrast, β-adrenergic receptor expression was below background.
The adenosine receptor A1 is associated with a high signal intensity on the IMCD arrays (Table 1). It has been shown that adenosine analogues can decrease water permeability induced by arginine vasopressin (18). Our previous experiments using the adenosine A1 agonist NPEA (N6-2-phenylethyladenosine) showed an increase in intracellular Ca2+ concentration and this effect was inhibited by the selective adenosine A1 antagonist 8-phenyltheophylline (17).
Endothelin-1 can induce diuresis in part by inhibiting water reabsorption induced by arginine vasopressin in the IMCD (60). In the IMCD, it has been demonstrated by immunofluorescence and RT-PCR that the endothelin B receptor is the predominant endothelin receptor subtype in agreement with our microarray data (42; 81).
Purinergic P2Y2 receptors have been demonstrated in the rat IMCD (38). Binding of ATP or UTP to P2Y2 receptors increase intracellular calcium in the IMCD, decreasing water permeability (17). This inhibition of water permeability might occur through activation of protein kinase C and reduction of cellular cAMP levels (37).
Arginine vasopressin receptors are of two types, V1 and V2 receptors. From the microarray analysis, the signal intensity corresponding to the V1a and V1b receptors in the IMCD cells was negligible. In contrast, V2 receptor signal intensity was relatively high. This evidence supports our previous finding that V2 receptors, but not V1a receptors, are expressed in the IMCD (16; 45).
Table 2 shows heterotrimeric G-protein subunits expressed in the IMCD. The α subunits expressed include three inhibitory α subunits (Gαi) which regulate adenylyl cyclases in an inhibitory fashion, the stimulatory α subunit (Gαs) which stimulates adenylyl cyclases, Gαq which couples to phospholipase Cβ (PLCβ) isoforms and Gα12, an activator of Rho-GEF (43) and a regulator of the actin cytoskeleton. All are potentially involved in vasopressin signaling.
Only 3 adenylyl cyclase transcripts of the 10 known genes were identified in the IMCD samples, namely Adcy6 and Adcy3 previously identified in the IMCD (28), as well as Adcy9 as shown in Table 3. In addition, soluble guanylyl cyclase 1 was identified, which functions as a nitric oxide receptor and transducer. Interestingly, there was no significant signal for any of the three nitric oxide synthase (NOS) isoforms in the IMCD samples, although NOS3 (endothelial NOS) was detected in the non-IMCD samples (Table 16). However, previous qualitative RT-PCR studies in microdissected IMCD segments demonstrated the presence of NOS1, 2 and 3 mRNA (96). Furthermore, both NOS1 (102) and NOS2 (91) protein has been detected in the IMCD by protein mass spectrometry.
As shown in Table 4, the predominant cyclic nucleotide phosphodiesterase transcripts correspond to the type 4 and type 1 classes, consistent with pre-existing findings (15; 75). Type 4 phosphodiesterases (PDE4A, PDE4B, PDE4C, and PDE4D, coded by separate genes) are cAMP–specific and inhibitable by rolipram. Type 1 phosphodiesterases hydrolyze both cAMP and cGMP and are calmodulin-dependent. No significant signal was found for PDE5, which has been implicated in regulation of aquaporin-2 trafficking in more proximal collecting duct segments (4).
There are more than 500 genes coding for protein kinases in mammalian genomes (48). As shown in Table 5, a large number of these are expressed in the IMCD. There were 93 serine/threonine kinases with signals above the median value in the IMCD (corrected signal above 1.0, Table 5A). Serum-glucocorticoid regulated kinase, sgk1, a serine/threonine kinase closely related to protein kinase B (Akt1), is the kinase with greatest signal intensity in the IMCD (Table 5A). This kinase is strongly regulated by the mineralocorticoid hormone aldosterone (11). WNK1 and WNK4, two kinases implicated in regulation of ion transport in more proximal parts of the renal tubule (95; 99), are also expressed in the IMCD with a relatively high signal intensity. Protein kinase A (catalytic β subunit), long recognized as a major element of vasopressin signaling, is also associated with a high signal intensity on the IMCD arrays. Additional protein kinases with proposed roles in vasopressin signaling have relatively high signal intensities including protein kinase B (Akt1), protein kinase C (iota, zeta, nu and epsilon), casein kinase isoforms, calmodulin-dependent kinases, myosin light chain kinase, Rho-kinase, two G-protein coupled receptor kinases (GRK2 and GRK6), various kinases in MAP kinase pathways, and cyclic-GMP-dependent kinases. Obviously, a large number of kinases were detected whose roles in the IMCD have not been investigated.
Relative to the number of serine/threonine kinases in IMCD, There are relatively few tyrosine kinases (Table 5B and 5C). Tyrosine kinases are of two types: non-receptor tyrosine kinases and receptor tyrosine kinases. Of the non-receptor tyrosine kinases, Frk (a src-family homolog) is strongly expressed in IMCD relative to non-IMCD elements. Also strongly represented is Janus kinase 1 (Jak1), which regulates transcription through activation of Stat1 and/or Stat2. Of the receptor tyrosine kinases, two are heavily expressed in the IMCD relative to non-IMCD elements, namely fibroblast growth factor receptor 2 (Fgfr2) and epidermal growth factor receptor (Egfr). Most of the detectable tyrosine kinases in the IMCD have been identified previously in fetal kidney including Jak1, Frk, Ptk2, Fgfr2, Egfr, Ryk, Ddr1 and Fgfr1 (36). Tyrosine kinases are associated with regulation of cell proliferation and differentiation.
The serine/threonine phosphatases found in the IMCD are listed in Table 6A. Focusing on the catalytic subunits, the highest signal intensities in the IMCD appear to be the PP1 phosphatases of which three different isoforms are represented namely, α, β and γ. The catalytic subunit of calcineurin, a Ca2+-calmodulin dependent phosphatase (PPP3CA), is also expressed as is the catalytic subunit of the magnesium-dependent phosphatase, PP2C. For all of these catalytic subunits, significant expression was found in both the IMCD and non-IMCD fractions, consistent with their ubiquitous roles in cell signaling (69). In contrast, the PP2A catalytic subunit signal on the IMCD arrays was indistinguishable from background, although several of its regulatory subunits were present. Calcineurin has been reported to be expressed in the IMCD and to be colocalized with PKCζ, the RII subunit of PKA, and AQP2 in endosomes of the IMCD (35). It has been reported that the α isoform of the calcineurin catalytic subunit is required for normal intracellular trafficking of AQP2 (23). Finally, there are many more genes in the genome that code for Ser/Thr phosphatase regulatory subunits than for catalytic subunits, a number of which were found to be expressed in the IMCD.
Table 6B summarizes the protein tyrosine phosphatases (PTPs) found in the IMCD. The protein tyrosine phosphatases can be classified into transmembrane receptor-like protein tyrosine phosphatases and cytosolic protein tyrosine phosphatases (84). There are 16 PTP transcripts identified in the IMCD samples with signal intensity above median. Ten coded for cytosolic PTPs and 5 for transmembrane receptor-like PTPs.
Dual specificity phosphatases in the IMCD are summarized in Table 6C. These phosphatases dephosphorylate a phosphotyrosine and a neighboring phosphothreonine. Among these, the transcript corresponding to Dusp1 displays the strongest intensity in the IMCD. Its function is to dephosphorylate the MAP kinases ERK1 and ERK2 at the sites phosphorylated by dual specificity kinases (MAP kinase kinases or MEKs) (73). Two dual specificity phosphatases are expressed in the IMCD but not in non-IMCD cell types, namely Dusp2 and Dusp5.
A-kinase anchoring proteins (AKAPs) are scaffold proteins that facilitate localized signaling through creation of binding complexes that bring enzymes (mainly kinases and phosphatases) into the vicinity of their targets. Table 7 shows the AKAPs whose signal intensity in the IMCD samples were above background. Those expressed at highest levels were AKAP-7, -8, -9, -11 and -13. AKAP7 (aka AKAP18δ) has been shown to be colocalized with AQP2 intracellularly in the IMCD (27). Both AKAP7 and AQP2 were localized predominantly at the apical plasma membrane when the IMCD cells were stimulated with dDAVP.
14-3-3 proteins are phospho-serine/phospho-threonine binding proteins. By virtue of their binding specificities, they have been proposed to play various phosphorylation-dependent regulatory roles. There are seven isoforms of 14-3-3, viz. β, γ, ε, η, σ, τ and ζ (Note: α and δ are phosphorylated forms of β and ζ isoforms (97; 98)). In the present study, 5 of 7 isoforms of 14-3-3 were identified in the IMCD with high signal intensity (Table 15). Functional roles of 14-3-3 in the IMCD have not been elucidated yet. However, 14-3-3 proteins have been found to modulate the function of the ubiquitin ligase Nedd4-2 in cultured collecting duct cells by binding to Nedd4-2 after sgk1-mediated phosphorylation (3).
Table 8 shows the phospholipases in IMCD with signals above background for either IMCD or non-IMCD samples. Among the phospholipases with the highest signal intensities was cytosolic phospholipase A2 (PLA2) which hydrolyzes the ester bond in SN-2 position of the core glycerol. This reaction is important because it liberates arachidonic acid, an unsaturated fatty acid that is the substrate for prostaglandin synthesis. Vasopressin has been seen to be variably associated with stimulation of prostaglandin E2 synthesis in collecting ducts, which provides a modulating influence on water permeability (68). Prostaglandin E2 synthesis depends on expression of one of two cyclooxygenases (COX-1 and COX-2). COX-1 exhibited a very high signal on the arrays, with a signal that is in the top 100 among transcripts in the IMCD. Both COX-1 and COX-2 are selectively expressed in the IMCD. As noted above, three different PGE2 receptors are expressed in the IMCD consistent with the view that PGE2 works as an autocoid.
As shown in Table 8, several isoforms of phospholipase C (PLC) are also expressed in the IMCD. PLC cleaves phosphatidylinositol 4,5-bisphosphate (PIP2) to yield two second messengers, diacylglycerol and inositol 1,4,5 tris-phosphate (IP3). The beta isoforms are activated by heterotrimeric G-protein alpha subunits Gαq and Gα11. There are two PLC-β isoforms expressed in the IMCD, PLCβ1 and PLCβ4. PLCβ1 is expressed selectively in IMCD cells versus non-IMCD cells. PLCγ isoforms are regulated via tyrosine phosphorylation, typically following growth factor receptor activation (65). PLCγ1, but not PLCγ2, appears to be expressed in IMCD substantially above ackground. PLCδ4 is also selectively expressed in the IMCD over non-IMCD cell types.
Table 9 lists transcripts corresponding to small GTP-binding proteins expressed on the IMCD arrays including Rab proteins, Arf/Sar proteins, Rho/Rac/CDC42 proteins, Ras and Ras-like proteins, and Ran proteins (77).
Members of the Rab family of proteins play a critical role in intracellular membrane trafficking. Large numbers of transcripts corresponding to Rab proteins were identified on the IMCD expression arrays (Table 9A) including many of those previously identified in a proteomic analysis of aquaporin-2-containing vesicles in IMCD cells (2). Most of the Rab proteins were found both in IMCD and non-IMCD cells, although Rab15 and Rab31 appear to be selectively expressed in the IMCD. Based on studies in non-renal cells, Rab15 is believed to be expressed in early endosomes and recycling endosomes (74). Rab31 is thought to be an early endosomal Rab based on its similarity to Rab5 (72).
ADP-ribosylation factors (Arfs) regulate coat protein-mediated vesicle budding, vesicular trafficking, and actin cytoskeleton assembly. Transcripts corresponding to most of the known Arf and Arl (ADP-ribosylation factor-like) proteins are expressed in the IMCD (Table 9B). Arf1 has been shown to be important for Golgi to ER transport. Arf6 is important for endocytosis (10; 14). Both Arf1 and Arf6 displayed high signal intensity on the arrays. Transcripts for the Arf-like proteins Sar1a and Sar1b were also found in the IMCD.
Rho/Rac/CDC42 GTPases are involved in regulation of the actin cytoskeleton and cell mobility. CDC42, Rnd1, Rac1, and RhoA are the four members of this family with the greatest signals on the arrays (Table 9C). In IMCD cells in primary culture, cAMP was found to inhibit Rho activation, thereby stimulating actin depolymerization and aiding translocation of aquaporin-2 to the plasma membrane (90). Rnd1 is expressed at a very high level in IMCD relative to other inner medullary cell types, specifically about 20-fold higher. Rnd1 has also been reported to promote disassembly of actin filaments (55). However, whether it plays a role in aquaporin-2 regulation has not yet been reported. A number of other Rho-like GTPases were also associated with high signal intensities in the IMCD including RhoB, RhoC, RhoG, RhoQ, RhoT2, and Rnd3.
The Ras-like protein family of small GTPases includes Rap, Rheb and Ral as well as Ras itself. They play an important role in regulating cell proliferation and differentiation. Among the Ras and Ras-like protein transcripts expressed in the IMCD, Rheb was associated with the greatest signal on the arrays (Table 9D). Rheb is a small GTPase that binds and activates the mTOR kinase (1). Another transcript with a relatively high signal intensity was K-Ras. This isoform of Ras has been shown to be an aldosterone-stimulated protein in collecting duct cells (93). Transcripts corresponding to other Ras-like proteins expressed in the IMCD include RalA, RalB, Rap1a, Rap1b, Rasd1, and Rras2. The Ral proteins have been found in aquaporin-2 containing intracellular vesicles in the IMCD (2). These proteins are associated with the so-called ‘exocyst complex’ involved in basolateral targeting of proteins. We have identified RalA in the basolateral plasma membrane of IMCD cells (unpublished data). Rap1 is involved in cell morphogenesis, cell differentiation, cell adhesion and cytoskeleton organization. It has been demonstrated that SPA-1 (signal-induced proliferation-associated 1), a GTPase-activating protein specific to Rap1, binds to AQP2 in the renal medulla (56). We identified SPA-1 in the IMCD in this study (Supplementary Table S1). Studies in SPA-1 knockout mice showed a defect in AQP2 trafficking to the apical plasma membrane (58). Rap1 is also a target of Epac, a RapGEF, that is activated by cAMP. Epac has been found to be important for regulation of vasopressin-simulated calcium mobilization in collecting duct cells (101), which is necessary for AQP2 trafficking (13). Transcripts corresponding to Epac1 but not Epac2 were identified in the IMCD in our array studies (Supplementary Table S1). One of Rap1’s targets is Raf1, a kinase in the MAP kinase pathway (see above). A full listing of transcripts corresponding to small GTP-binding protein GEFs, GAPs and interacting proteins found in the IMCD is reported in (Supplementary Table S4).
SNARE proteins play an important role in membrane fusion in eukaryotic cells. Syntaxin 7 and 12 were the SNARE protein transcripts with the greatest signals in the IMCD as shown in Table 10. Syntaxin 7 is present in the late endosome while syntaxin 12 is present in the recycling endosome (63), an organelle that may be important in AQP2 recycling to the plasma membrane. Plasma membrane syntaxins include syntaxins 2, 3 and 4. Among these, only syntaxin 4 was associated with a strong signal in the IMCD. This syntaxin is present in the apical plasma membrane and has been proposed to be involved in AQP2 trafficking (47). In the current study, VAMP3 (cellubrevin) had the strongest signal intensity among VAMP/synaptobrevin transcripts. This protein was previously identified in the IMCD by immunogold electron microscopy (22). VAMP2 and VAMP4 transcripts were also found to be expressed in the IMCD above the median value. VAMP2 colocalizes with aquaporin-2 in AQP2-containing intracellular vesicles (2; 54). An additional subfamily of SNARE proteins is the SNAP25 homologues. SNAP25 is neural-specific and its expression level was very low in the IMCD. SNAP23 protein has previously been identified in the IMCD on the basis of immunorecognition (32) and in the present study we found that it is expressed in the IMCD above the median expression level for all IMCD transcripts.
Both clathrin heavy chain and light chain polypeptides were identified in the IMCD (Table 11). The clathrin adaptors, AP1, AP2 and AP3 are also expressed in the IMCD. These clathrin adaptor proteins function in post-Golgi trafficking pathways. AP1 regulates vesicle transport between the trans-Golgi network and endosomes. AP2 regulates endocytosis (6) and AP3 is involved in vesicle trafficking from early endosomes to late endosomes or lysosomes (59). In addition, three clathrin adaptors called epsins are strongly expressed in the IMCD namely epsin 4 (enthoprotin), epsin 3 and epsin 2. These proteins bind both to ubiquitylated integral membrane proteins and to PIP2 (in addition to clathrin) and may be involved in regulated endocytosis of apical plasma membrane transporters in the collecting duct (94). Finally, an adaptor called phosphatidylinositol binding clathrin assembly protein or CALM, a PIP2-binding clathrin adaptor was found at a relatively high signal intensity level on the IMCD arrays.
Dynamins are GTPases (although not ‘small’ GTPases) that are associated with clathrin-coated pits and are involved in the abscission process forming clathrin-coated vesicles. In this study, dynamin-1 was expressed below background and dynamin-2 and -3 were expressed at very low levels only (0.21 and 0.45, respectively, Table 15). In contrast, the dynamin-like GTPases Mx1 and Mx2 were associated with high signal intensities in the IMCD (7.60 and 16.14, respectively). Their locations and roles in the IMCD cell have not been reported.
γ- and β-actin, the two chief forms of non-muscle actin were associated with the highest actin signals on the IMCD arrays. These forms have been seen previously in proteomic studies of the IMCD (2; 62). Vasopressin induces actin filament depolymerization in the IMCD (70), a process that has been shown to be required for AQP2 translocation from intracelluar to apical plasma membrane in cultured CD8 cells (79).
ERM proteins (ezrin, radixin, and moesin) are actin-binding proteins that link actin to the plasma membrane and are involved in signal transduction and cytoskeleton remodeling (85). In the present study, ezrin (Vil2) and radixin (Rdx) were found to be expressed in IMCD at high signal intensity. On the contrary, moesin mRNA was not detected. A previous study (66) reported a marked increased in the expression of ezrin mRNA in the inner medullas of lithium-treated rats, which was associated with an increase in ezrin protein in the IMCD. The authors proposed that this protein plays a role in the nephrogenic diabetes insipidus seen in response to lithium. Moesin has been reported to mediate trafficking of AQP2 to plasma membrane in rabbit CD8 cells (79). Our finding of low or nonexistent levels of moesin mRNA in the IMCD suggests that moesin’s role in AQP2 trafficking may not be universal.
Several actin-binding proteins have been identified that are AQP2-binding partners in rat inner medulla including α2-spectrin, α-tropomyosin 5b (tropomyosin 3γ), gelsolin, and α-actinin (57) Transcripts corresponding to all of these proteins showed high signal intensity on our microarrays.
Myosin regulatory light chain B and non-muscle myosin II-B (heavy chain 10) were associated with the greatest signals among myosin gene products. Also associated with relatively high signals in IMCD were myosin II-A (heavy chain 9) and unconventional myosins 1B, 1C, 1D, and 1E, as well as myosin 5B. Myosin regulatory light chain has been demonstrated to be phosphorylated in the IMCD in response to vasopressin (12). This protein regulates the non-muscle myosins II-A and II-B, both of which are expressed in the IMCD. As demonstrated previously, myosin II-B is collecting duct selective in the inner medulla (12). Myosin II-A has been demonstrated to be bound to AQP2 in cell lysates of kidney inner medulla (57). Recently myosin 5B has been reported to be associated with Rab11-positive AQP2-containing endosomes, presumably recycling endosomes (52).
Table 12C shows the transcripts corresponding to the microtubule proteins and microtubule-associated proteins including molecular motors that are expressed above background on the arrays. Microtubule disruption with colchicine and other agents reduces the hydro-osmotic action of vasopressin in collecting ducts (61) through effects on AQP2 trafficking (67). In addition, it has been shown that both dyneins (microtubule-based molecular motors) and dynactin in the inner medulla were associated with trafficking of AQP2-containing vesicles to the apical plasma membrane (49). In the present study, multiple dynein and dynactin isoforms were found in the rat IMCD. Cytoplasmic dyneins are multi-subunit proteins consisting of a combination of light chains, heavy chains, and intermediate chains. The predominant dyneins in the IMCD are cytoplasmic dyneins with signal intensity of light chain-1 > intermediate chain-1 > heavy chain-2 polypeptides. Cytoplasmic dynein heavy polypeptide-2 is selectively expressed in IMCD over non-IMCD elements of the inner medulla.
The transcripts corresponding to transport proteins found in the IMCD above background are reported in Table 13. As can be seen in Table 13A, three water channels are associated with extremely high signal intensities, namely AQP4, AQP2 and AQP3. The signal intensities for these aquaporins were ranked 3rd, 10th and 14th, respectively, among the 7913 annotated transcripts found in the IMCD. In contrast, AQP1 was present only in the non-IMCD samples from the inner medulla.
Table 13B shows the ion channels and transporters expressed in the IMCD. The chloride channel ClC-Kb is expressed in the basolateral plasma membrane of cells throughout the distal nephron from thick ascending limb to the collecting duct and is mutated in some forms of Bartter’s Syndrome (92). Its transcript was found at a moderate level in IMCD but not in non-IMCD components of the inner medulla. Note that the levels of the epithelial sodium channel (ENaC) α, β, and γ transcripts are relatively low, consistent with the low level of ENaC protein expressed in this segment (24).
Table 13C shows the non-mitochondrial solute carrier (SLC) transcripts found in the IMCD. Among these, the transcript with the highest signal was the vasopressin-regulated urea transporter UT-A (Slc14a2). This protein is key to the accumulation of urea in the renal inner medulla (19; 20). Its manner of regulation by vasopressin is controversial with one group claiming that the protein undergoes regulated trafficking (39) while another finds an absence of regulated trafficking by vasopressin (33).
Supplementary Table S6 reports the mitochondrial transporters and carriers detected. As may be expected, many of these were components of the F0-F1 ATP synthase responsible for proton gradient-driven ATP synthesis.
We identified about 824 transcription factors expressed in IMCD above background. A list of the transcription factors identified with signal intensity above median is shown in Supplementary Table S7. Only a few of these have been studied in the context of AQP2 regulation.
Long-term increases in ambient vasopressin in animals or cultured cells results in increased AQP2 protein abundance that is mediated at least in part by enhanced AQP2 gene transcription rates (53). A cAMP responsive element (CRE) plays a critical role (31; 50; 100). It has been widely assumed that regulation of AQP2 transcription involves the transcription factor CREB. However, in this study, we found that CREB itself (now called Creb1) is associated with an extremely low signal intensity on the arrays (Table 14A). This finding raises the possibility that other CREB-family members may be responsible for the long-term regulation of AQP2 gene transcription. The CREB-like transcriptional factors in the IMCD associated with high signal intensities are Atf3 and Atf4 (Creb2). All of these factors share the same recognition site TGACGTCA, which is present in the AQP2 promoter region. Atf3 is thought to be a stress-response transcription factor (25). We hypothesize that Atf4 (Creb2) mediates the effects of cAMP on AQP2 gene expression.
Previous studies have demonstrated that Fos expression is markedly increased in rat inner medulla in response to vasopressin (7). Furthermore, c-Jun phosphorylation was found to be increased in the same study. These two proteins together constitute the transcription factor AP1. They were both found on our arrays with very high signal intensity (Table 14A)
Kruppel-like factors (KLF) are transcription factors in the zinc finger family that regulate cell differentiation, cell proliferation and development with a DNA-binding site motif of CACCC (34; 86). KLF12 was found to be expressed in the IMCDs of young mouse kidneys (76). In the current study in mature rats, the KLF12 intensity on the arrays was, however, below background. Other KLF transcription factors are expressed in the IMCD including KLF5, KLF6, KLF9, and KLF10.
Rat GATA3 but not GATA2 can be amplified by RT-PCR from the collecting ducts of rat kidneys (89). Overexpression of rat GATA3 in primary culture of mouse OMCD cells can increase AQP2 activities, indicating that GATA3 could interact with GATA motifs in the AQP-2 promoter (89). Also, both GATA2 and GATA3 were detected at high signal intensity in the IMCD, but not in the non-IMCD portions of the inner medulla (Table 14B). GATA1 signal intensity was below background. Sp1 is also a member of the KLF transcription factor family. In the present study, Sp1 showed a moderate signal intensity in the IMCD.
TonEBP (Nfat5) is a Rel/Dorsal transcription factor that regulates cellular accumulation of organic osmolytes and HSP70. These mechanisms are critical to the process that protects cells from the hypertonic stress of the renal medulla. In the present study, TonEBP was indeed found to be expressed in the IMCD at a transcript level, although, the signal intensity was below the median on the arrays (Supplementary Table S1). Mice deficient in the TonEBP gene display severe atrophy of the renal medulla, presumably because the cells failed to adapt to the hyperosmolality (44). In response to long term hypertonicity in mpkCCD cells, TonEBP mediates expression of UT-A and possibly AQP2 independent of vasopressin level (26). However, AQP2 was not found to be strongly regulated by local osmolality in the renal inner medulla (82).
We thank Tony Cooper, Nalini Raghavachari for help with the Affymetrix array hybridization and data reporting, Guozhong Ma for setting up the online database, and Peter Munson and Xiuli Xu for advice regarding normalization of array data. This study was funded by the Intramural Budget of the National Heart, Lung, and Blood Institute (Project HL001285, M.A. Knepper, principal investigator).