Binding of Regulatory Subunits of Cyclic AMP-Dependent Protein Kinase to Cyclic CMP Agarose

The bacterial adenylyl cyclase toxins CyaA from Bordetella pertussis and edema factor from Bacillus anthracis as well as soluble guanylyl cyclase α1β1 synthesize the cyclic pyrimidine nucleotide cCMP. These data raise the question to which effector proteins cCMP binds. Recently, we reported that cCMP activates the regulatory subunits RIα and RIIα of cAMP-dependent protein kinase. In this study, we used two cCMP agarose matrices as novel tools in combination with immunoblotting and mass spectrometry to identify cCMP-binding proteins. In agreement with our functional data, RIα and RIIα were identified as cCMP-binding proteins. These data corroborate the notion that cAMP-dependent protein kinase may serve as a cCMP target.


Introduction
Previous studies claimed that in addition to adenosine 39,59-cyclic monophosphate (cAMP) and (cytidine 39,59-cyclic monophosphate) cGMP [1,2], the cyclic pyrimidine nucleotide cytidine 39,59-cyclic monophosphate (cCMP) may play a role as second messenger molecule [3]. However, studies on cellular effects of cCMP were not reproducible [4] and technical problems hampered the determination of tentative cytidylyl cyclase activity in mammalian cells [5,6]. Moreover, a postulated cCMP-specific phosphodiesterase could not be identified so far [7]. In fact, several known phosphodiesterases do not cleave cCMP [8]. With refined radiometric and liquid chromatography-mass spectrometry (LC-MS)-based methods we could recently show that the highly purified bacterial adenylyl cyclase toxins CyaA from Bordetella pertussis and edema factor from Bacillus anthracis, in addition to cAMP, produce cCMP [9]. Furthermore, the highly purified soluble guanylyl cyclase a 1 b 1 along with cGMP, produces cCMP in a nitric oxide-dependent manner [10]. In addition, the regulatory subunits of cAMPdependent protein kinase A (PKA), RIa and RIIa, are activated not only by cAMP, but by cCMP as well [11]. These recent data indicate that cCMP may, indeed, play a role as second messenger.
The aim of our present study was to identify cCMP-binding proteins. As methodological approach, we synthesized and tested 29-6-aminohexylcarbamoyl-cCMP (29-AHC-cCMP) agarose and 4-6-aminohexyl-cCMP (4-AH-cCMP) agarose and a corresponding control agarose (Figure 1). In 29-AHC-cCMP agarose, the nucleoside 39,59-cyclic monophosphate (cNMP) is linked to the matrix via the 29-O-ribosyl group, and in 4-AH-cCMP agarose via the 4-NH group of the pyrimidine ring. Hence accessibility of the affinity ligand to proteins is different in the two matrices. Bound proteins were subsequently analyzed by immunoblotting and LC-MS. The cNMP-agarose approach is very useful at identifying cNMP-binding proteins [12]. Here, we show that in accordance with our enzymological data, cCMP-agarose binds RIa and RIIa.

Materials and Methods
Materials 29-AHC-cCMP agarose was synthesized by analogy to other 29-AHC-agarose matrices [13]. Syntheses of 4-AH-cCMP and 4-AH-cCMP agarose were in accordance to literature procedures [14,15]. Both cCMP agaroses were prepared with ligand densities of ,6 mMol/mL of settled gel. cCMP (purity . 99,8%) was from Biolog Life Science Institute (Bremen, Germany). Anti-RIa Ig (sc-136231) was purchased from Santa Cruz Biotechnology (Santa Cruz, CA, USA). This antibody also recognizes RIb. All other reagents and cell culture media were purchased from standard suppliers.

Cell Culture
B103 rat neuroblastoma cells (kindly provided by Dr. E. Zoref-Shani,, Tel-Aviv, Israel) [16] were cultured in MEM RAA medium supplemented with 10% (v/v) fetal bovine serum at 37uC and 5% (v/v) CO 2 . Human HeLa cervix carcinoma cells were obtained from the American Type Culture Collection and were cultured in DMEM medium supplemented with 10% (v/v) fetal bovine serum at 37uC and 5% (v/v) CO 2 . Human HEK293 embryonic kidney cells were from the American Type Culture Collection and were cultured in DMEM medium supplemented with 10% (v/v) fetal bovine serum, non-essential amino acids and sodium pyruvate at 37uC and 5% (v/v) CO 2 . HL-60 human promyelocytic leukemia cells (kindly provided by Dr. P. Gierschik, Ulm, Germany) [17] were cultured in RPMI 1640 medium supplemented with 10% (v/v) horse bovine serum, non-essential amino acids and sodium pyruvate at 37uC and 5% (v/v) CO 2 . J774 mouse macrophages [18] were obtained from Dr. I. Just, Hannover, Germany and were cultured in DMEM medium supplemented with 10% (v/v) fetal bovine serum and 2 mM Lglutamine at 37uC and 5% (v/v) CO 2 .

cCMP Agarose Affinity Chromatography
Cells were harvested and suspended in lysis buffer consisting of 40 mM b-glycerolphosphate, 100 mM NaF, 4 mM Na 3 VO 4 , 2% (m/v) Triton X-100, 100 mM NaCl, 60 mM NaPP i and 20 mM Tris/HCl, pH 7.5. Protein concentration was determined using the BCA protein assay. 29-AHC-cCMP agarose, 4-AH-cCMP agarose and EtOH-NH agarose (30 ml each) were  In competition experiments, cCMP (2 mM) was added to cCMP agarose samples. Input designates cell lysate before incubation with agarose. C, cell lysates of HeLa cells were incubated with 29-AHC-cCMP agarose or control agarose. RIa was detected by immunoblotting with an antibody. Numbers at the left margins of immunoblots designate markers of molecular mass standards. Representative immunoblots are shown. A and B were from the same experiment, different exposures were shown. Similar data were obtained in three independent experiments. doi:10.1371/journal.pone.0039848.g002 equilibrated three times with wash buffer consisting of 1 mM dithiothreitol, 1% (m/v) Triton X-100, 1 mM Na 3 VO 4 , 50 mM NaF, 154 mM NaCl and 20 mM Tris/HCl, pH 7.5. Agarose beads were incubated with 2 mg of cell lysate protein in wash buffer (total volume 500 ml) in the presence of 100 mM isobutyl-methylxanthine under rotation at 30 rpm at 4uC overnight. In order to detect non-specific binding, 2 mM cCMP was included in some samples. Samples were then centrifuged at 1,000 g for 3 min at 4uC, and beads were washed three times with 500 ml of wash buffer, followed by addition of 25 ml of 26 sample buffer. Samples were heated for 10 min at 95uC. For alkylation of cysteine residues 1 mL of an acrylamide solution (40%, m/v) was added and incubated at room temperature for 30 min. Proteins were subsequently separated by sodium dodecyl sulfate gel electrophoresis in gels containing 10% (m/v) acrylamide.

Immunoblotting
Gels were blotted onto nitrocellulose membranes. Membranes were incubated with anti-RIa Ig (1:500) over night, followed by a 2 h incubation with anti-mouse IgG from sheep (1:2,000). Bands were visualized using the Signal WestPico Luminol Enhancer and Stable Peroxidase Solution (Thermo Fisher Scientific, Rockford, IL, USA). The highly abundant proteins myosin-Ig, a-actinin-4 and cytoplasmic actin 1 bound to the 4-AH-cCMP agarose matrix nonspecifically. Proteins in the ,45 kDa region represent RIa (43 kDa) and RIIa (46 kDa), respectively, and bound to the matrix specifically since competition with cCMP (2 mM) eliminated these bands from the gel. Numbers at the right margin of the gel designate markers of molecular mass standards. After photography, the gel was cut into small pieces, and proteins were identified by MALDI and LC-MALDI mass spectrometry. doi:10.1371/journal.pone.0039848.g003

Sample Preparation for MS Analysis
Following photography for documentation, protein-containing gel lanes were cut into small pieces and destained with ACN (50%, v/v) in 20 mM NH 4 HCO 3 . Subsequently, ACN (100%) was added until gel pieces were dry and ACN was removed in a vacuum centrifuge. Trypsin was added at a concentration of 10 ng/mL in 20 mM NH 4 CO 3 and 10% (v/v) ACN and the protein digest was performed at 37uC over night. Peptides were extracted by incubation of samples with 50 ml of 10% (v/v) ACN and 0.5% (v/ v) trifluoroacetic acid (TFA) at room temperature and shaking at 300 rpm for 30 min. The supernatant fluid was transferred into a new vial, and the extraction was repeated twice using increasing concentrations of ACN (30%, 50%). Following vacuum drying, samples were dissolved in 5 ml of 5% (v/v) ACN and 0.2% (v/v) TFA for matrix-assisted laser desorption/ionization (MALDI)-MS analysis. Samples (0.5 ml) were spotted onto a MALDI target plate (AB Sciex, Darmstadt, Germany) and mixed with 0.8 ml a-cyano-4hydroxycinnamic acid (CHCA) (4 mg/mL in 50% ACN, 0.2% TFA) using the dried droplet method.

LC Analysis
Peptide separation was performed by reversed phase chromatography using a nano-LC system (Dionex, Idstein, Germany) which consists of an autosampler (Famos), a loading pump (Switchos), a gradient pump (Ultimate) and a microfraction collector (Probot). An aliquot of up to 20 mL of each sample was injected onto a C18 trap column (PepMap 300 mm65 mm, 3 mm, TFA and a flow rate of 30 mL/min. Peptides were eluted onto a separation column (PepMap, C18 reversed phase material, 75 mm6150 mm, 3 mm, 100 Å , Dionex) and separated using eluent A with 5% (v/v) acetonitrile in 0.1% (v/v) TFA and eluent B with 80% (v/v) acetonitrile in 0.1% (v/v) TFA with a gradient from 10% to 40% eluent B in 134 min and 40% to 100% eluent B in 10 min. Samples were spotted directly onto a MALDI target plate (AB Sciex) that had been prespotted with CHCA matrix as described above. A sheath liquid of 50% (v/v) ACN was applied and subsequently spots were recrystallized using 50% (v/v) ACN and 0.1% (v/v) TFA.  Various cell types were cultured, harvested, lyzed and analyzed by gel electrophoresis (see Figure 3). Gels were cut into small pieces and subsequently analyzed by MALDI-MS/MS. Figures 4 and 5 show representative MS spectra for peptide precursors from HEK cells, and Tables 2 and 3

Identification of PKA RIa by Immunoblotting
The cNMP agarose affinity approach has already been proven to be successful at identifying cNMP-binding proteins [12,15]. PKA RIa is expressed in many cell types [1]. We probed both 29-AHC-cCMP agarose and 4-AH-cCMP agarose in HeLa cells, a widely used cell culture model (Figure 2A and 2B). Both matrices bound RIa as assessed by immunoblotting. Binding was specific since cCMP strongly inhibited RIa binding to cCMP matrices, and the control agarose devoid of the cCMP moiety did not bind RIa. In J774 mouse macrophages, 29-AHC-cCMP agarose also bound RIa in a specific manner as assessed by the use of cCMP as competing ligand and control agarose ( Figure 2C). 4-AH-cCMP agarose was more effective than 29-AHC agarose at binding RIa (compare Figure 2A versus Figure 2B and 2C). Therefore, all further experiments were performed with 4-AH-cCMP agarose. Figure S1 shows the sequence alignment of human RIa and RIIa. The sequence identity between the two isoforms amounts to 38%, but the amino acid sequences are sufficiently different from each other to allow for unequivocal protein identification by peptide analysis via MALDI-MS/MS. Figure 3 shows the Coomassie Blue-stained gel of cell lysates of HL-60 cells following incubation with 4-AH-cCMP agarose. The gel shows two bands in the ,45 kDa region that were competed for by cCMP. The gel was cut into thin slices, proteins were digested and peptides were analyzed by MALDI-MS/MS. This analysis showed that highly abundant proteins, i.e. myosin-Ig, a-actinin-4 and cytoplasmic actin bound non-specifically to 4-AH-cCMP agarose, i.e. the binding of these proteins was not competed for by cCMP (Figure 3). In contrast, the bands in the ,45 kDa region competed for by cCMP were identified as RIa and RIIa. Figure 4 and 5 show representative peptide precursor MS spectra for RIa and RIIa from HEK293 cells, respectively. Table 1 provides a summary of the MALDI-MS/MS analysis of the ,45 kDa region of HeLa cells, HEK293 cells, HL-60 cells and B103 cells. In all four cell types, RIa and RIIa were identified with sequence coverages ranging from 9-27%, the number of identified peptides ranging from 3-9 and highly significant combined Mascot score ranging from 80-428. Tables 2 and 3 list the amino acid sequences of peptides analyzed in Figures 4 and 5.

Identification of RIa and RIIa by MALDI-MS/MS
We further refined the analysis of proteins bound to 4-AH-cCMP agarose by separating peptides of the 45 kDa region using reversed phase chromatography prior to MALDI-MS/MS (LC-MALDI). Tables 4, 5, 6 show that in this analysis, RIa and RIIa were unequivocally identified in B103 cells, HEK293 cells and HL-60 cells, the number of identified peptides ranged from 5-19 and peptide ion scores of individual peptides ranged from 26-159.  Table 3. MS/MS analysis results of the peptide precursors shown in Figure 5.

Discussion
For many years, research on cCMP barely progressed because of non-reproducible results [3,4] technical difficulties in determination of the activity of cCMP-forming enzymes [5,6] lack of sufficiently sensitive and specific cCMP detection techniques and absence of experimental tools to detect cCMP-binding proteins [3]. Recently, we could unequivocally demonstrate that certain bacterial adenylyl cyclase toxins also produce cCMP [9] and recombinant soluble guanylyl cyclase a 1 b 1 does so, too [10]. Moreover, we showed that the recombinant regulatory subunits RIa and RIIa of PKA bind cCMP, resulting in dissociation of the R subunits from the catalytic subunits and subsequent protein phosphorylation [11]. Thus, a functional effect of cCMP on clearly defined proteins was finally shown.
Considering the success of the cNMP agarose approach to identify cNMP-binding proteins [12,15] the recent results on cCMP synthesis and cCMP effects on PKA prompted us to synthesize and test two cCMP agaroses (Figure 1) in order to identify cCMP-binding proteins. The application of both cCMP agaroses was straightforward, EtOH-NH agarose and competition with cCMP serving as specificity control (Figure 2 and 3). In immunoblotting experiments we detected RIa (Figure 2). In MALDI-MS/MS analysis, a traditional approach analyzing gel slices (Figure 3, 4 and 5 and Tables 1, 2, 3) and in a more advanced approach applying additional reversed phase chromatography prior to MS analysis (Tables 4, 5, 6), we unequivocally identified RIa and RIIa in several cell types as proteins specifically binding to 4-AH-cCMP agarose.
We were somewhat surprised that the cCMP-agarose approach worked so well considering the fact that cCMP is only a low-potency activator of PKA [11]. RIa appears to possess considerable conformational flexibility since the attachment of the affinity ligand to the matrix, either via the 29-O-ribosyl group or the 4-NH group of the pyrimidine base worked. The higher efficacy of 4-AH-cCMP agarose compared to 29-AHC-cCMP agarose at binding RIa can be explained by the fact that the 29-OH group of cNMPs is important for interaction with the protein [19]. Thus, our data provide a compelling example for the notion that low-affinity interactions between a protein and a ligand cannot necessarily be dismissed as non-specific. Exceedingly high affinity of a protein to a ligand may  impede with subsequent dissociation of the protein from the affinity matrix [12,15]. Evidently, in cCMP agarose matrices, steric ligand accessibility and the balance between sufficient binding affinity and subsequent protein elution are quite right. In intact cells, cCMP, due to its stability (see discussion below) [8] may accumulate in specific PKA-containing cell compartments so that sufficiently high cCMP concentrations for PKA activation build up. In fact, in a recent study, we have shown that in certain cells, overall cCMP concentrations are in the range of ,30 pmol/10 6 cells which is just three-fold lower than the corresponding cAMP concentration [20].
In previous studies we showed that cCMP induces vasodilatation and inhibition of platelet aggregation via cGMP-dependent protein kinase (PKG) and that cCMP also binds to purified PKG [11,21]. However, in none of the cell types studied here and with none of the experimental approaches did we identify PKG as protein binding to cCMP agarose. This apparent discrepancy may be due to the fact that the expression of PKG is too low in the cell types studied. As a consequence, binding of PKG to cCMP agarose may be below the detection limit of the currently available mass spectrometers. Thus, in future studies, PKG-enriched cells such as platelets and smooth muscle cells will have to be examined. Alternatively or additionally, there may be steric conflicts in the binding of PKG to the two cCMP agarose matrices. A hint towards steric problems may be the fact that in contrast to the situation with PKA, cCMP is only a partial activator of PKG [11]. Accordingly, it will be necessary to develop affinity matrices with different ligand densities, space lengths between the agarose and the cNMP and different attachment positions of the cNMP to the linker. Figure 1 illustrates some of the chemical possibilities to optimize affinity matrices.
It is also noteworthy that our studies did not identify cNMPdegrading phosphodiesterases as target proteins for cCMP. Previous studies claimed the existence of a specific cCMPdegrading phosphodiesterase [7] but its molecular identity remained elusive. Rather, in a recent study, we examined a broad panel of human phosphodiesterases and found none of them to cleave cCMP [8]. Our negative cCMP affinity matrix data regarding phosphodiesterases fit to the functional data. These data raise the question through which mechanism cCMP is inactivated if it is, indeed, a second messenger. Transmem-brane export may be an inactivation mechanism but the affinity of the interaction of such transporters with cCMP may be too low to be detected by our affinity ligand approach [22,23]. In fact, transporters of the MRP family accept structurally very diverse substrates so that a specific interaction with an affinity ligand cannot necessarily be expected [23]. Lastly, in our study, we did neither detect Epac nor cNMP-regulated ion channels as cCMP-binding proteins [24,25]. As is the case for PKG and phosphodiesterases, such negative data do not exclude the existence of other cCMP-binding proteins. These proteins may simply have gone unnoticed in our analysis for various technical reasons including suitability of affinity matrices and sensitivity of MS detection methods.
In conclusion, in this study we provided proof of principle that the use of cCMP affinity matrices is a useful approach to identify cCMP-binding proteins. We anticipate that the systematic application of this approach in terms of the development of multiple matrices and the analysis of multiple cell types, together with refined LC-MS techniques, will lead to the identification of additional cCMP-binding proteins, some of which may turn out to be specific for cCMP. Figure S1 Sequence comparison of RIa and RIIa. Amino acid sequences of human RIa and RIIa were aligned, using the one-letter code. Sequences were aligned in http://www.uniprot. org/blast/. Sequence identity amounts to 38%. (JPG)