Identification and Characterization of Novel Mutations in the Human Gene Encoding the Catalytic Subunit Calpha of Protein Kinase A (PKA)

The genes PRKACA and PRKACB encode the principal catalytic (C) subunits of protein kinase A (PKA) Cα and Cβ, respectively. Cα is expressed in all eukaryotic tissues examined and studies of Cα knockout mice demonstrate a crucial role for Cα in normal physiology. We have sequenced exon 2 through 10 of PRKACA from the genome of 498 Norwegian donors and extracted information about PRKACA mutations from public databases. We identified four interesting nonsynonymous point mutations, Arg45Gln, Ser109Pro, Gly186Val, and Ser263Cys, in the Cα1 splice variant of the kinase. Cα variants harboring the different amino acid mutations were analyzed for kinase activity and regulatory (R) subunit binding. Whereas mutation of residues 45 and 263 did not alter catalytic activity or R subunit binding, mutation of Ser109 significantly reduced kinase activity while R subunit binding was unaltered. Mutation of Cα Gly186 completely abrogated kinase activity and PKA type I but not type II holoenzyme formation. Gly186 is located in the highly conserved DFG motif of Cα and mutation of this residue to Val was predicted to result in loss of binding of ATP and Mg2+, which may explain the kinetic inactivity. We hypothesize that individuals born with mutations of Ser109 or Gly186 may be faced with abnormal development and possibly severe disease.


Introduction
Protein Kinase A (PKA) is a cyclic AMP (cAMP)-dependent Ser/Thr kinase involved in regulating a multitude of biological processes including cell growth and division, cell differentiation, as well as metabolism and immune responsiveness [1]. In its inactive state, PKA exists as a tetrameric holoenzyme consisting of two regulatory (R) and two catalytic (C) subunits. Four different genes, PRKAR1A, PRKAR1B, PRKAR2A, and PRKAR2B, encode the R subunit proteins RIa, RIb, RIIa, and RIIb, respectively, with several splice variants of the RIa and RIIa subunits [2,3]. Two principal genes, PRKACA and PRKACB, encode the C subunits Ca and Cb, respectively. Also, PRKX and the retroposons PRKY and PRKACG are identified as PKA C subunit genes. While no protein products for PRKY and PRKACG have been identified, PRKX is translated into a protein kinase which is inhibited by the R subunit in a cAMP-sensitive manner [4][5][6][7]. Both the PRKACA and PRKACB genes encode several protein products. Alternative use of two exons upstream of exon 2 in the PRKACA gene gives rise to two Ca variants with different N-termini. These proteins are designated Ca1 and Ca2 [8][9][10][11]. In the case of the PRKACB gene, several exons 59 of exon 2 are encoding a number of Cb splice variants designated Cb1, Cb2, Cb3, and Cb4 [12,13], as well as a number of Cb3 and Cb4 forms that contain N-terminal inserts from exons designated a, b, and c [14,15]. In the brain and nervous tissues of higher primates, a range of C variants lacking exon 4 encoded sequences are also identified [16].
The two major groups of the R subunits RI and RII form two types of the PKA holoenzymes termed PKA type I and PKA type II, respectively. The C subunits are the catalytically active components, and they become activated after dissociation from the R subunits in a cAMP-dependent manner. For specificity in the cAMP-PKA signaling pathway, the PKA holoenzymes can be located to specific subcellular structures via A Kinase Anchoring Proteins (AKAPs). AKAPs typically bind RII subunits with high affinity [8,[17][18][19]. In contrast, PKA type I holoenzymes tend to locate to the soluble fraction of the cell but can bind to dual-and RI-specific AKAPs [19]. Both the C and R subunits are differentially expressed in various tissues. Whereas Ca1 is ubiquitously expressed, Ca2 is exclusively found in sperm cells [8][9][10]. For the Cb variants, Cb1 is ubiquitously expressed, while the other Cb variants are more specifically expressed in certain tissues, such as lymphoid and neuronal tissues [14,15].
Ca1 is the principal source of PKA activity in most tissues [11] and was the first protein kinase to be subjected to crystallization and 3D structure determination [20]. The Ca1 structure reveals a large C terminal lobe (C-lobe or large lobe) and a smaller N terminal lobe (N-lobe or small lobe). The large lobe is mainly composed of a-helices, and is involved in R subunit and substrate binding. It contains a number of residues involved in catalysis. The small lobe is mostly composed of b-strands and contains an ATP binding site. Between the large and small lobe is the active site cleft. The small and large lobes, as well as the active site, are known as the catalytic core and are encoded by residues 40-300 in PKA Ca1. The amino acid sequence and 3D structure of the catalytic core are conserved among all protein kinases. The catalytic core may, in addition to ATP, bind two Mg 2+ ions that are critical for catalysis [20]. The two Mg 2+ ions are designated activating and inhibitory Mg 2+ , or Mg1 and Mg2, respectively [21]. The binding affinity for Mg1 is higher than for Mg2 in the presence of ATP, while binding of Mg2 is believed to stabilize the protein, but also to inhibit catalysis [21][22][23][24]. Mg 2+ is also essential for high-affinity binding to inhibitors PKI and RIa subunits [25]. For catalytic activity, Ca has to be phosphorylated on residue Thr 197 , located in the activation loop of the large lobe. This phosphorylation affects the conformation of the conserved DFG (Asp 184 -Phe-Gly) motif in the active site [26]. In the active conformation, Asp 184 of the DFG motif coordinates the three phosphates of ATP via the Mg 2+ ions, positioning the c-phosphate of ATP optimally for catalysis. This central role of Asp 184 in catalysis depends on a conserved hydrogen bond between Asp 184 and Gly 186 backbone amide group. This interaction orients Asp 184 perfectly for coordination of the Mg 2+ ions and efficient ATP binding [27]. The Asp 184 -Gly 186 hydrogen bond is only established after phosphorylation of Thr 197 , which causes a twist of the peptide bond between Phe 185 and Gly 186 . In this way phosphorylation of Thr 197 serves as a regulatory mechanism for the activation of the C subunit. The Local Spatial Pattern (LSP) alignment method, developed by Kornev et al [27], has revealed two conserved spatial motifs in eukaryotic protein kinases. These are two hydrophobic structures called the Catalytic (C-) and Regulatory (R-) spines that have to be established in order for the kinase to become catalytically active. The C-spine is completed when the adenine nucleobase of ATP interacts with the active site. For the R-spine to be established, Thr 197 in the activation loop has to be phosphorylated. Evaluation of the C-and R-spines is a helpful way of predicting whether a kinase has a catalytically active or inactive conformation.
Despite its central role in a multitude of processes in the body, few diseases have been linked to defects or alterations in the PKA subunit genes or proteins. Germline mutations leading to premature stop codons in PRKAR1A have been identified in patients with Carney complex, a multiple neoplasia syndrome with skin pigmentations, cardiac and other myxomas, endocrine tumors, and psammomatous melanotic schwannomas [28,29]. To the best of our knowledge, no diseases have been linked to defects in any of the C subunit genes. Nevertheless, studies on mice that are homozygote knockout (KO) for the PRKACA gene reveal a severe phenotype. The majority of the Ca 2/2 mice die at birth or during the early postnatal period [30]. The offspring that survive beyond 2 months (, 11%, [31]), all show reduced growth, and the few males, but not females, that reach puberty are all 100% infertile [30,32]. Furthermore, mice with reduced PKA C gene transcription by only expressing one functional C subunit allele of Ca or Cb, show reduced PKA activity and neural tube defects [33].
Due to the severe phenotype of Ca KO mice we searched for mutations in the human PRKACA gene by genomic sequencing of 498 subjects. We identified two nonsynonymous point mutations in the PRKACA gene that result in amino acid switches in the Ca1 protein at residues 45 and 109. In addition, we searched for previously described nonsynonymous mutations in various human genomic DNA databases, and selected two of these, giving amino acid switches at residues 186 and 263, for further studies. These four mutations were introduced to Ca1-and Ca2-encoding plasmids and the proteins were expressed and analyzed with respect to kinase activities and R subunit binding in vitro and in vivo. Mutation of residues 109 and 186 were associated with significantly reduced and totally abrogated kinase activity, respectively. In addition, mutation of residue 186 rendered Ca incapable of forming PKA type I holoenzymes.

Analysis of Mutations
The PRKACA gene from 498 Norwegian donors deriving from three different control groups [34,35] was sequenced and analyzed for point mutations by Lark Technologies (Takeley, UK). Leukocyte DNA was isolated from thawed blood containing ethylenediaminetetraacetic acid. Using the Applied Biosystems 340A Nucleic Acid Extractor, DNA was extracted with chloroform/phenol followed by ethanol precipitation. Exons 2-10 of PRKACA were sequenced using the primers listed in table 1.

Generation of Plasmids
The pDONR201-Ca1 WT vector has previously been described [40] and was used to make plasmids containing Ca1 Mut  Ca1 was cloned into the mammalian expression vector pGFP 2 -C3 (Perkin Elmer) as previously described [42], creating the BRET sensor construct GFP-Ca1 WT . Using Quikchange H II Site-Directed Mutagenesis Kit, GFP-Ca1 WT was mutated into GFP-Ca1 Mut , according to the manufacturer's protocol. RIa-Rluc and RIIa-Rluc constructs have previously been described [42].
For all mutagenesis reactions, the PCR reaction mixture was initially heated to 95uC for 30 s, followed by 16 cycles of 95uC for 30 s, 55uC for 1 min and 68uC for 7 min, before final elongation at 68uC for 2 min. Primers are listed in table 2. All plasmids were verified by sequencing (Eurofins MWG Operon).

Purification of Proteins
Ca1 WT and Ca2 WT were purified by affinity chromatography using PKI-peptide Affi-Gel, as previously described [41]. Ca2-Gly186Val was purified using a modified method described by Hemmer et al [43]. Briefly, BL21(DE3) cells transformed with either Ca2 Gly186Val or His-tagged RIIa Gly337Glu (provided by Antje Badel, University of Kassel, Germany) were cultured and protein expression induced with IPTG. After centrifugation, the bacterial pellets were resuspended and lysed, and the lysates were mixed in equimolar amounts forming PKA holoenzymes. The holoenzymes were then coupled to a Ni 2+ -resin binding the His-tagged R subunits. Following washing with 50 mM NaH 2 PO 4 (pH 8.0), 5 mM b-mercaptoethanol, and 25 mM KCl, Ca2 Gly186Val was eluted with the same buffer supplemented with 10 mM cAMP.
Phosphotransferase Assay 20 h post transfection, the HEK 293T cells were harvested and washed 3 x in phosphate buffered saline, then lysed in a potassium-phosphate buffer (5 mM K 2 HPO 4 , 1 mM EDTA, 250 mM sucrose and 0.5% Triton X-100, 1 mM phenylmethanesulfonyl fluoride, 1 mM Na 3 VO 4 and protease inhibitor cocktail (Sigma-Aldrich, cat. no. P8340)) by vortexing and 20 min on ice. Debris was pelleted by centrifugation at 16 000 g for 15 min at 4uC. After Bradford protein determination (Bio-Rad, cat. no. 500-0006), all samples were adjusted to equal concentrations. Phosphotransferase activities of lysates from Ca1 WT , Ca1 Mut , and mock transfected cells were determined by the phosphotransferase assay described by Witt and Roskoski [45,46]. Briefly, phosphotransferase activity against the PKA-specific substrate Kemptide (Leu-Arg-Arg-Ala-Ser-Leu-Gly, Sigma-Aldrich) was measured in a reaction mixture (14.3 mM Mg-acetate, 143 mM ATP, 7.5 mCi/mL c 232 P-ATP (PerkinElmer), 50 mM Tris-HCl, pH 7.4) in the presence or absence of cAMP or PKI. After incubation at 30uC for 9 min, the reactions were stopped by spotting onto p81 phosphocellulose papers followed by 4 615 min washing in 75 mM phosphorous acid. The filter papers were washed in 96% ethanol for 10 min and air dryed for approximately 1 h. Phosphotransferase activity was measured by liquid scintillation in 3 ml Opti-fluor (Packard BioScience, PerkinElmer). All experiments were repeated at least three times.

Immunoblotting
Samples used in phosphotransferase assays were evaluated for C subunit expression. Protein concentrations of each sample within an assay were adjusted to the same levels, as described above. Proteins were separated by SDS-PAGE and transferred to PVDF membranes. The membranes were blocked by drying and rehydrated in methanol before incubation with primary antibody Purified Mouse Anti-PKA [C] (1:250 dilution, BD Transduction Laboratories, cat. no. 610981) for 1 h, followed by washing with TBST. After 30 min incubation with secondary antibody HRP conjugated goat anti-mouse (1:2000 dilution, ICN Biomed, cat. no. 55563) and washing, proteins were detected by enhanced chemiluminescence (Pierce Biotechnology, Rockford, IL, USA) and the Syngene G:BOX imaging system.
Expression of endogenous C and GFP-C constructs in transfected COS-7 cells used for BRET assays were also assessed with immunoblotting. Proteins were detected similarly to the method described above, with the following modifications: blocking was performed with 5% milk powder for 1 h; primary antibody was rabbit anti-Ca (Santa Cruz, PKA cat (c-20) cat. no. SC-903) and secondary antibody anti-rabbit IgG (peroxidaselinked species-specific whole antibody (ECL) NA934, GE Healthcare, Freiburg, Germany).

Spectrophotometric Kinase Activity Assay
Kinetic activities of purified Ca2 WT and Ca2 Gly186Val were evaluated by the continuous enzyme-linked spectrophotometric method described by Cook et al. [22]. Activities were reported in U/mg, defined as mmolmg 21 min 21 . All experiments were repeated at least three times.

Molecular Representation and Simulation
Selected motifs, including the DFG motif and it's interactions with ATP and divalent cations were presented and analyzed with PyMOL [47], using the experimental structure described by Thompson et al [48] (PDB identifier 3FJQ). Simulated mutagenesis in PyMOL was performed on Gly 186 in the DFG motif. Relevant distances between the Gly 186 /Val 186 residue and surrounding atoms were calculated.

Results
We performed two independent studies of the PRKACA gene; 1) by sequencing genomic DNA from 498 Norwegian subjects and 2) by a bioinformatics analysis of DNA sequences from various populations submitted to publicly available databases (see Material and Methods). By the first approach, exons 2 through 10 of PRKACA were sequenced using exon-specific primers and genomic DNA extracted from leukocytes. In this way we detected five mutations in the PRKACA gene; one in each of exons 2, 3, 4, 5, and 8. Only two of the mutations translated into an amino acid switch. The corresponding nucleotides were located in exons 3 and 4, and are affecting Arg 45 and Ser 109 in the Ca1 sequence, while the three silent mutations were at Pro 33 , Gly 126 , and Pro 236 (Table 3, Fig. 1A). With the recent submission of a large amount of new mutations from several thousand individuals by exome and whole genome sequencing projects, four of these five mutations are now also present in dbSNP [36,49]. Our study thus confirms the presence of the mutations referred to as rs56105247 (at Pro 33 ), rs56085217 (Arg 45 ), rs78098302 (Gly 126 ), and rs137911238 (Pro 236 ) in the Norwegian population.
By the second approach, we detected another twelve nonsynonymous mutations (Table 4), of which the majority has been submitted within the last year. Among the mutations that were known at the initiation of this study, two were selected for further investigations. Firstly, the mutation rs35635531, resulting in a switch of Ser 263 to Cys, is the only mutation that has been submitted by three independent projects. Although the frequency of this mutation has not been determined in detail, it is likely to be relatively common in at least some human populations. Secondly, the mutation causing the Gly186Val mutation was considered particularly intriguing since this residue is located in the Mg 2+ positioning loop close to ATP and Mg 2+ in the active site cleft of Ca (Fig. 1B) and is part of the highly conserved DFG motif [26,50,51]. The positions of the four investigated mutations resulting in amino acid switches are shown relative to exon encoded sequence in Fig. 1A and in the 3D structure of Ca1 in Fig. 1B.
Mutation of Residues Ser 109 and Gly 186 , but not Arg 45 and Ser 263 in Ca1 Influences Catalytic Activity We first tested if any of the mutations influenced the catalytic activity of Ca. Site-directed mutagenesis was used to introduce the required amino acid switches Arg45Gln, Ser109Pro, Gly186Val, and Ser263Cys to Ca encoding expression vectors. Wild type (WT) and the four mutated Ca1 products (collectively termed Ca1 Mut ) were expressed in HEK-293T cells, and 20 h post transfection the cells were lysed. Expression of protein was first assessed by immunoblotting with a pan C antibody, revealing that both Ca1 WT (Fig. 1C, lane 1) and Ca1 Mut (Fig. 1C, lanes 2 to 5) were expressed at comparable levels and with similar apparent size as endogenous Ca1 (Fig. 1C, lane 6). We further determined their catalytic activity in the presence of the PKA-specific substrate Kemptide and c-32 P-ATP (Fig. 1C). In these experiments, lysates from untransfected cells (Fig. 1C, lane 6) were included to assess endogenous kinase activity. This demonstrated high catalytic activity of expressed Ca1 WT which was set to 100% (Fig. 1C, lane 1). According to this assay, the catalytic activities of Ca1 Arg45Gln and Ca1 Ser263Cys were shown not to be significantly different from Ca1 WT . On the other hand, the activity of Ca1 Ser109Pro was significantly lower compared to Ca1 WT (P ,0.05), and the activity of Ca1 Gly186Val was comparable to background levels (Fig. 1C,  lanes 4 and 6). This suggested that mutation of Gly 186 rendered Ca inactive and prompted us to focus further on revealing the mechanism for this inactivation. Accordingly, we transfected cells with Ca1 Gly186Val and Ca1 WT and compared catalytic activity in the absence and presence of cAMP and the PKA-specific inhibitor PKI and compared phosphotransferase activities to mock transfected cells (Fig. 2A). The activity of Ca1 WT was reduced to background level in the presence of PKI while activities of Ca1 Gly186Val were comparable to background levels independently of stimulation with cAMP or inhibition with PKI ( Fig. 2A, Mock compared to Ca1 Gly186Val ).
To determine if Ca Gly186Val was completely inactive we next expressed and purified recombinant Ca WT and Ca Gly186Val . We used recombinant WT and mutated Ca2 since Ca2 is more easily produced than Ca1 due to higher solubility [41]. Ca2 will therefore give a higher protein yield. Ca1 and Ca2 may also be interchanged in these experiments because studies suggest that they are kinetically indistinguishable [41]. WT and mutated recombinant Ca2 (Ca2 WT and Ca2 Gly186Val ) were produced in BL21(DE3) cells, purified and visualized by SDS-PAGE (Fig. 2B).
(In order to simplify the nomenclature we refer to the mutations according to their Ca1 numbering also when introduced to Ca2. The actual position of the mutated residues in the Ca2 protein is achieved by subtracting the number 8). Ca2 Gly186Val was purified by running protein extracts over an RIIa Gly337Glu affinity column and Ca2 bound to RIIa was eluted with 10 mM cAMP (see Material and Methods). The purification process was evaluated after each step by SDS-PAGE analysis (Fig. 2B, lanes 1 to 14). Purified Ca2 Gly186Val and Ca2 WT were tested for catalytic activity employing the Cook assay [22] that showed a specific activity for Ca2 WT at 20.9 6 2.1 U/mg (95% confidence interval) (Fig. 2C, Ca2 WT ). In comparison, catalytic activity by Ca2 Gly186Val was undetectable demonstrating that a Gly 186 to Val mutation renders the kinase catalytically inactive (95% confidence interval of Ca2 Gly186Val : 0.55 6 0.96 U/mg) (Fig. 2C, Ca2 Gly186Val ).

Mutation Gly186Val Alters the Catalytic Core of Ca1
We next investigated the Gly186Val mutation in a 3D structure model of Ca1 in an attempt to understand the molecular nature of the kinetic inactivity (Fig. 3). Figure 3A shows selected conserved motifs surrounding the DFG motif where residue 186 is located. The chain of interactions leading into the DFG motif from phosphorylated Thr 197 is depicted. In the active conformation of Ca1 WT , the C-and R-spines are assembled and the structure is optimized for catalysis (Fig. 3B). Simulated mutagenesis of Gly186Val revealed three possible rotamers. For representation in the figures, the rotamer with the least steric hindrance was selected. Our Ca1 Gly186Val model suggests an altered conformation of the DFG motif, which we hypothesize may disturb the structure of the spines (Fig. 3B, lower right box). In all published experimental structures of catalytically active kinases, there is a conserved hydrogen bond between the side chain of Asp 184 and the amide group of Gly 186 [50] (Fig. 3C, left panel, dashed line). In figure 3C (right panel) it is shown that the presence of the hydrophobic Val side chain leads to steric hindrance and an unfavourable binding of the cation Mn1 (Cc1-Mn1 distance 2.7 Å ) compared to Gly (Ca-Mn1 distance 4.3 Å ). The simulated distances between Cc1, Cc2 and Mn1 as well as ATP for this rotamer and the two other rotamers of Ca1 Gly186Val were reduced (results not shown). According to these data we hypothesize that by replacing Gly 186 with Val, the bulky side chain of Val will make it less likely for Ca to bind Mg1/Mn1 in its optimal position. This will also prevent ATP binding, together suggesting why Ca2-Gly186Val lacks kinase activity.
A recent study on the Raf kinase where Gly 596 , which is the counterpart of Gly 186 in Ca1, was mutated to Arg revealed that some Raf Gly569Arg kinase activity was achieved at high concentrations of ATP [52]. Based on this study and considering the molecular structure depicted in figure 3 we speculated if increasing the concentrations of ATP and/or Mg 2+ would be sufficient to restore at least some kinase activity in Ca2 Gly186Val . Kinase activity of purified Ca2 Gly186Val and Ca1 WT were measured against increasing doses of Mg 2+ and ATP, keeping either the ATP or Mg 2+ concentration constant at 143 mM or 14.3 mM, respectively. For Ca1 WT high phosphotransferase activity was detected at concentrations of Mg 2+ between 0.2 and 100 mM and ATPconcentrations between 1 and 143 mM. At none of these concentrations was activity of Ca2 Gly186Val detectable.

Ca2 Gly186Val Associates with RIIa but not RIa
Ca2 Gly186Val was purified by affinity chromatography with RIIa in which Ca Gly186Val was released by addition of cAMP. This suggests that mutation of Ca1 at residue 186 does not interfere with RII association or cAMP activation. However, ATP is required for association between C and RI [53,54] and given the fact that Ca2 Gly186Val probably does not bind ATP efficiently, it may be suggested that Ca Gly186Val will only associate with RII and not RI. To investigate this we determined the binding of Ca1 Gly186Val to RIa and RIIa using a Bioluminescence resonance energy transfer (BRET) assay. In these experiments, holoenzyme formation was investigated for all four Ca Mut proteins to test if any of the mutations would influence R subunit binding and cAMP sensitivity. COS-7 cells were cotransfected with either Ca1 WT or Ca1 Mut N-terminally coupled to Green Fluorescent Protein (GFP), GFP-Ca1 WT and GFP-Ca1 Mut , respectively, and RIa and RIIa C-terminally coupled to Renilla luciferase (Rluc),RIa-Rluc and RIIa-Rluc, respectively. Immunoblotting shows the expression of both GFP-Ca1 WT and GFP-Ca1 Mut (Fig. 4A, lanes 2 to 6). Figure 4, lane 1 depicts the expression of endogenous C subunit in cells transfected with Rluc alone serving as background. In the case of holoenzyme formation with RIa subunits, measurements were normalized by setting the BRET signal in holoenzyme with GFP-Ca1 WT to 100% (Fig. 4B, WT ''2''). Increasing the intracellular concentration of cAMP with forskolin (fsk) and IBMX (''+'') leads to the dissociation of the holoenzyme complex and the reduction of the BRET signal by approximately 50% as reported previously [55]. The same was true for holoenzymes formed with RIa-Rluc and GFP-Ca1 Arg45Gln , GFP-Ca1 Ser109Pro , and GFP-Ca1 Ser263Cys , respectively (Fig. 4B, ''+''). In contrast, the GFP-Ca1 Gly186Val showed only residual binding to RIa in resting cells (Fig. 4B, Gly186Val ''2''), and only marginal activation was detected after stimulation with fsk and IBMX (Fig. 4B, Gly186Val ''+'').

Discussion
Here we have examined the human PRKACA gene for mutations by sequencing genomic DNA from 498 Norwegian individuals and by searching for earlier reported mutations in publicly available databases. By genomic sequencing, we identified two nucleotide changes that resulted in the mutations Arg45Gln and Ser109Pro in the Ca1 protein. In public databases we identified two interesting mutations that would lead to residue switches, Gly186Val and Ser263Cys.
Two out of the four mutations (Ca1 Ser109Pro and Ca1 Gly186Val ) resulted in significantly reduced kinase activity. It should, however, be noted that comparing the exact activities of Ca1 WT and Ca1 Mut was a challenge due to differences in transfection efficiencies. We compensated for this by adjusting activities to immunoreactivity in the immunoblots. This demonstrated reduced kinase activity for Ca1 Ser109Pro , whereas Ca1 Gly186Val was kinase inactive. As Gly 186 is absolutely conserved (Supplementary Fig. S1) and is part of the DFG motif [27] we investigated the mutation affecting this residue in more detail. The importance of Gly 186 conservation was underscored since mutation was associated with complete abrogation of kinase activity. The exact mechanism that leads to complete inactivation in this case is not known. However, Ca Gly186Val was successfully purified by RIIa-affinity chromatography followed by elution with cAMP. This, together with the fact that Ca Gly186Val forms cAMP-sensitive holoenzymes with RIIa in vivo suggests that lack of enzymatic activity is not caused by an overall misfolding of the protein. We rather suggest that lack of kinase activity is due to less extensive molecular changes, affecting only parts of the protein structure. The introduction of a bulky residue in place of Gly 186 is likely to affect many critical factors necessary for kinase activity. The most apparent explanation is that the aliphatic Val 186 side chain leads to less efficient binding of Mg1/Mn1 due to displacement of the cation itself or water molecules that are solvating Mg1/Mn1. Loss of Mg1/Mn1 would in turn disable ATP binding, thus explaining the kinase inactivity. This hypothesis is supported by our BRET results, which show exclusive binding to RII but not RI subunits. Moreover, we were unable to purify Ca Gly186Val using PKI affinity chromatography (results not shown). It has previously been shown that the RI subunits, as well as PKI, need ATP to bind C with high affinity [53,54]. According to our model, it is not unlikely that the hydrogen bond between residues 184 and 186 fails to form in Ca Gly186Val , leading to displacement of Asp 184 and inability to position Mg1 or Mn1 and thereby ATP in the active site. The inability to bind ATP implicates that the C-spine is not established, which is necessary for the active conformation. An alternative explanation for the inactivity is that Val 186 , instead of suppressing binding of Mg1/Mn1 and ATP, is rather displaced, leading to a conformational change in the DFG motif itself. This would lead to a malformed R-spine, which is also thought to be incompatible with kinase activity [26]. Enumeration is given as in Table 3, and sequence conservation is illustrated in Supplementary Fig. S1. Reference identifiers from dbSNP [36,49]  Reduced kinase activity was also observed for Ca1 Ser109Pro . Ser 109 is located in the middle of b-strand 4 in the small lobe and this b-strand stabilizes the N-terminal end through targeting Ser 109 to Thr 37 [48]. It is well known that the N-terminal tail is important for C subunit stability, and deletion of the N-terminus has previously been demonstrated to lead to a significant reduction in thermal stability [56]. We speculate that mutation of Ser 109 to Pro leads to partial loss of N terminal structure and destabilization of the kinase, and that this may in part provide an explanation for the reduced kinase activity. A second consequence of mutating Ser 109 may be associated with our recent findings that Ser 109 belongs to a series of signature residues that can be used to distinguish the Ca from Cb ortholog (unpublished results). Due to this, the Ser109Pro mutation may result in alteration of Ca-specific functions which do not include holoenzyme formation since our BRET-results showed that Ca1 Ser109Pro formed holoenzymes with both RIa and RIIa with comparable affinities as the Ca1 WT . Full comprehension of the reduced kinase activity and other features associated with mutation at Ser 109 merits further investigation.
Neither mutations of Arg45Gln nor Ser263Cys influenced apparent kinase activity. Arg 45 is located near a recently identified conserved pocket in the N lobe known as the N lobe cap, which is above the crucial amino acids Ala 70 and Lys 72 [48]. Ala 70 is part of the C-spine while Lys 72 is directly involved in ATP binding. It might have been expected that mutation of Arg 45 would affect kinase activity, also because this residue is highly conserved in metazoan PKA Ca/Cb homologs (Supplementary Fig. S1) and is clearly under strong purifying selection. The fact that we did not observe any change in phosphotransferase activity suggests that the mutation does not influence the positioning of Ala 70 and Lys 72 . Despite of this, it may be that mutation of residue 45 affects other features necessary for function of the kinase. It could be speculated that the N lobe cap may also be a docking site for proteins inhibiting kinase activity by disturbing either the C-spine or Lys 72 . This may indicate that the Arg45Gln mutation could be involved in deregulated PKA activity due to altered interactions with so far unidentified interaction partners docking to the N lobe cap. This binding partner does not include the R subunit since the BRET experiments of Ca1 Arg45Gln showed no differences in binding and release upon fsk/IBMX stimulation of RIa and RIIa subunits compared to Ca1 WT . This is also consistent with the localization of Arg 45 far from the R subunit docking site.
Ser 263 is highly conserved among metazoa (Supplementary Fig.  S1) and is part of the H helix at the very lower end of the large lobe. This site is close to the R subunit docking sites, and the  (pThr 197 ) to the DFG motif are represented as stick models [50]. Mn 2 ATP (green), the DFG motif (teal), Gly-rich loop (salmon), catalytic loop (yellow), and activation loop (purple) are also highlighted. B. Overall structure of Ca1 WT with the conserved structural motifs the C-and R-spine structural motifs highlighted. The boxed segments depict spatial relations between residue 186 (Gly or Val) and ATP, divalent cations, and the C-and R-spines. C. DFG motif in Ca1 WT (left) and Ca1 Gly186Val (right) and its relations to Mn1 and ATP. Residues are represented as stick models with carbon (orange), oxygen (red) and nitrogen (blue) atoms. The hydrogen bond between the side chain of Asp 184 and the amide group of Gly 186 (dashed line) is predicted to be broken in Ca1 Gly186Val due to the Val side chain. The models are based on the structure with PDB identifier 3FJQ [48]. doi:10.1371/journal.pone.0034838.g003 mutation would not be expected to interfere directly with catalytic activity as was also demonstrated here. Rather it could be speculated that a shift from Ser to Cys could influence the mechanism regulating R-C interaction in vivo. However, our BRET experiments showed no difference in holoenzyme formation and dissociation of Ca1 Ser263Cys to neither RIa nor RIIa subunits compared to Ca1 WT . Although no effect of a Ser263Cys mutation in Ca was detected, the high degree of conservation of this residue and strong purifying selection is an indicator of a hereto unknown functional importance of Ser 263 .
Among the mutations identified in the sequencing of the PRKACA gene, Arg45Gln and Ser109Pro were identified in one and two samples, respectively. Of the two mutations identified in the database search only the Ser263Cys mutation was identified by two independent submitters. To what extent this indicates anything about the prevalence of the different mutations, remains to be verified. However, Gly in the DFG motif is a relatively frequent site of disease-causing mutation in various protein kinases which is most likely due to kinase inactivation [57]. As described above, homozygote targeting mutation of the PRKACA gene in mouse is associated with high pre-and postnatal lethality, most probably due to lack of kinase activity at critical steps in embryonic development [30,32]. Based on this, it is expected that any mutation affecting Ca activity may lead to a severe phenotype and possibly disease in human. Since the Ca1 Gly186Val mutation resulted in catalytic inactivation and partial lack of holoenzyme formation, homozygote mutation for Gly 186 in vivo may functionally be considered a gene KO and hence may be incompatible with normal development and life. The same would most likely be the case for the frame shift mutation detected in exon 8. The fact that Ca Gly186Val exclusively forms holoenzyme with RII subunits suggest that individuals with heterozygote mutation of Gly 186 may have reduced levels of PKA type I holoenzymes, in addition to harboring type II holoenzymes occupied by inactive C subunits, which can hypothetically cause an unbalance in PKA signaling.
The other investigated mutations are likely more compatible with normal development and may not be associated with disease since neither of them influenced holoenzyme formation and mutation of Ser 109 only partly reduced catalytic activity. Despite this there is a possibility that homozygote mutation of Ser 109 may be associated with disease since experiments on mice have demonstrated that reduced C subunit gene expression can lead to spinal neural tube defects [33]. In total, 13 nonsynonymous point mutations in the PRKACA gene were identified in the present study. For example the Gly186Val mutation was only identified in a single EST sequence, and no frequency data was available. Hence, the information on the prevalence of the various mutations is therefore limited and a full comprehension of their existence in patients or patient groups remains to be elucidated. Finally, it is also worth mentioning that several thousand regions of the human genome have structural variation in large segments termed Copy Number Variation (CNV) [58]. In this paper they report that three out of 95 individuals were found to have a loss of ,160 kb which included the whole PRKACA gene as well as up-and downstream genes. Due to this it may be speculated that a combination of CNV deletion at the PRKACA locus and a heterozygote loss-offunction mutation of Ca could be associated with disease due to severe reduction in C subunit gene dose. To what extent this is a cause of disease remains to be determined. Supporting Information Figure S1 Multiple sequence alignment of human PKA Ca (top row) and homologous sequences from a number of metazoan species for the sequence segments containing the 13 mutations discussed in the present study. Enumeration is according to the human PKA Ca1 splice variant and the mutated residues are highlighted. The sequences were obtained from the NCBI (http:// www.ncbi.nlm.nih.gov) and UniProt (http://www.uniprot.org) protein sequence databases with the following identifiers: P17612, P05132, P00517, NP_001003032, Q90WN3, A3KMS9, NP_001003470, P22694, P68181, P05131, XP_867543, XP_422379, Q7ZWV0, Q3ZB92, Q7T374, XP_001175934, XP_002740161, CAG44453, XP_393285, XP_968170, NP_476977. (TIF)