Phosphorylation events associated with different states of activation of a hepatic cardiolipin/protease-activated protein kinase. Structural identity to the protein kinase N-type protein kinases.

Cardiolipin- or protease-activated protein kinase, isolated from rat liver cytosol and originally named liver PAK-1, was found to be the natural form of protein kinase N (PKN) by comparing the sequences of 43 tryptic peptides of the purified liver enzyme and determining the corresponding liver cDNA sequence. These analyses also identified (i) Arg-546 as the major site of proteolytic activation, (ii) the protease resistance of the C-terminal extension beyond the catalytic domain, and (iii) in vivo stoichiometric phosphorylation of Thr-778 in the mature enzyme. Homology modeling of the catalytic domain indicated that phosphothreonine 778 functions as an anchoring site similar to Thr-197 in cAMP-dependent protein kinase, which stabilizes an active site compatible with preferred substrate sequences of PAK-1/PKN. Sigmoidal autophosphorylation kinetics and increased S6-(229-239) peptide kinase activity following preincubation with ATP suggested phosphorylation-dependent activation of PAK-1/PKN. The onset of activation corresponded with phosphorylation of the regulatory domain site Ser-377 (located within a spectrin homology region), followed by Thr-504 (within a limited protein kinase C homology region), and, to a lesser extent, Thr-64 (in the RhoA-binding region). Several additional sites in the hinge region adjacent to a PEST protein degradation signal were selectively autophosphorylated following cardiolipin activation. Overall, these observations suggest that the regulation of this class of protein kinase involves complex interactions among phosphorylation-, lipid-, and other ligand-dependent activation events.

extraction procedure (16). First-strand cDNA was synthesized in a reaction based on 10 g of RNA, oligo(dT) primer, and Superscript RNase H Ϫ reverse transcriptase. PCR amplification was based on the cDNA template, Taq DNA polymerase, and degenerate oligonucleotide primers (synthesized on an Applied Biosystems 381A DNA synthesizer) corresponding to the tryptic peptide sequences of the PAK-1 catalytic domain (subdomains VI and XI) (3). The PCR-amplified DNA products were verified by DNA sequencing after subcloning into pGEM7zf(ϩ)derived T-vector (17).
cDNA Library Screening and DNA Sequence Analyses-Either PCRamplified PAK-1 cDNA probes or subcloned cDNA fragments, randomly labeled with Klenow DNA polymerase fragment (18), were used to detect PAK-1-positive clones during screening of the rat liver cDNA library. DNA sequencing was carried out either by the dideoxynucleotide chain termination reaction (Sequenase) or by dye primer PCR sequencing (Taq DNA polymerase) as described by the manufacturers. Sequences were confirmed by sequencing in both directions, overlapping regions of independently cloned DNA.
Homology Modeling of PAK-1/PKN-The three-dimensional structure of residues 616 -910 of PAK-1 in complex with Mg 2 ATP and the substrate peptide S6-(229 -239) was modeled on the x-ray structure of cyclic AMP-dependent protein kinase (cAPK) in complex with Mn 2 ATP and the peptide inhibitor PKI-(5-24) (21,22). Following alignment of the two sequences, the atomic coordinates for PAK-1/PKN, Mg 2 ATP, three coordinated water molecules, and the S6 peptide were copied from the x-ray structure of cAPK using the homology module of the InsightII program (Molecular Simulations, Inc., San Diego, CA). For loops with insertions or deletions or with proline substitutions, suitable templates were identified from searches of a subset of all protein structures in the Brookhaven Protein Data Bank (23). 2 Thr-778 was phosphorylated based on direct primary structural analyses (see "Results"). Also included were nine water molecules in cavities in the cAPK structure and a 5-Å layer of 1667 water molecules around the complex. The structure was minimized using the all-atom CHARMm22 force field with the program X-PLOR (24,25). Missing parameters for the phosphate group on Thr-778 and the triphosphate group were estimated from the x-ray structure and from similar parameters for AMP. Atomic charges for the triphosphate group were extrapolated from a one-point modifiedneglect-of-diatomic-overlap (MNDO) calculation on ATP. Because no van der Waals parameters are available for Mn 2ϩ , Mg 2ϩ was used instead. 2 The following protein and nucleic acid sequence data bases were used as sources for protein kinase sequences: SWISS-PROT, European Molecular Biology Laboratory, Postfach 102209, D-6900 Heidelberg, Federal Republic of Germany; GenBank, National Center for Biotech- nology 1. Primary structure of rat liver PAK-1/PKN. Nucleotides identified by rat liver PAK-1/PKN cDNA sequencing are numbered on the right, and predicted amino acid residues on the left. The mammalian mRNA-binding site consensus sequence (45), nucleotides Ϫ6 to ϩ4, is boxed. The nucleotide sequence is identical to the lung PKN sequence (5) except for nucleotides as marked (q, above). Underlined regions are identical to sequences of the PAK-1/PKN tryptic peptides (numbered in order T1-T43). The T27, T30, T31, T33, T34, T39, and T43 sequences were as reported previously (3). Arg-292 is marked with an asterisk to indicate an unresolved anomaly in T10 sequence analyses (see "Results"). Experimentally determined phosphorylation sites (phosphopeptide analyses) are denoted by white letters in black boxes. Leucine residues in leucine zipper-like sequences are marked by open boxes.
The non-bonded interactions were calculated with a group-based cutoff at 7.5 Å. Peptide bonds were restrained to 180 Ϯ 3°; Mg 2 ATP was tethered to its original position; and tight nuclear Overhauser effect restraints were placed on the residues coordinated to the Mg 2ϩ ions. Further nuclear Overhauser effect restraints were placed on hydrogen bonds from the protein to ATP or to other residues in conserved secondary structure elements. The energy minimization was performed in four stages with decreasing numbers of restraints until the energy gradient was Ͻ0.1 kcal mol Ϫ1 K Ϫ1 .
Miscellaneous Methods-32 P-Labeled tryptic phosphopeptides were also characterized by isoelectric focusing on polyacrylamide gels using pH 3-10 gradients as described previously (26), except for detection of 32 P-phosphopeptides on the dried gels by PhosphorImager analyses (Molecular Dynamics, Inc.). Tryptic peptides were analyzed by matrixassisted laser desorption time-of-flight mass spectrometry, using a Finnigan LaserMat spectrometer with ␣-cyano-4-hydroxycinnamic acid as sample matrix (27). Protease (trypsin)-activated peptide kinase activity of liver PAK-1 was determined with ribosomal S6-(229 -239) peptide substrate (4, 15) as described previously (3). The Australian National Genomic Information Service programs FastA and BLAST were used to search for DNA and protein sequence homologies and were used for comparisons of PAK-1/PKN sequences with sequences in the SWISS-PROT and GenBank data bases. 2

RESULTS
Primary Structure of Hepatic PAK-1-The objective of the sequence analyses was to provide a structural framework for subsequent investigations of PAK-1/PKN structure/function relationships, particularly in relation to phosphorylation sites. The sequences of 43 tryptic peptides isolated from purified cytosolic 116-kDa PAK-1 (3), including the seven catalytic domain-derived peptides described previously (3), accounted for a total of 496 amino acid residues (Fig. 1). The individual peptide sequences closely matched the corresponding regions of the sequence predicted from the rat lung PKN cDNA sequence ( Fig.  1) and were distributed throughout the PKN sequence: from residue 12 (peptide T1) through to the C-terminal T43 tryptic peptide (residues 938 -946) (Fig. 1).
The relationship between rat liver PAK-1 and lung PKN was confirmed by cloning and determining the sequences of liver PAK-1 cDNA (see "Experimental Procedures"), which matched the complete 946-amino acid coding sequence for rat lung PKN (5), except for amino acid substitutions at positions 510 (liver PAK-1 Gln 3 lung PKN His), 722 (Val 3 Gly), and 808 (Leu 3 Phe) (Fig. 1). The liver PAK-1/PKN Leu-808 assignment was confirmed by the sequence of the matching tryptic peptide, T34 (Fig. 1). The Val-722 assignment was consistent with the presence of a chemically similar amino acid at the corresponding position in other members of the PKN/PRK family. 2 Other variations detected were Val-650 (liver PAK-1/PKN) rather than Leu-650 (lung PKN) and Asn-654 rather than Asp-654, but only in one out of three independently isolated cDNA clones encompassing this region; in the other two clones, these assignments corresponded exactly to PKN. The Leu-650 and Asp-654 assignments were preferred as these residues are conserved in all known mammalian PKN/PRKs. Alignment of the PAK-1/ PKN tryptic peptide and liver cDNA-predicted sequences revealed only two differences, at positions 292 and 298 within the T10 peptide sequence (Fig. 1). The T10 peptide (35 pmol of peptide recovered from ϳ50 pmol of enzyme) was only isolated as a phosphopeptide from autophosphorylated PAK-1 (see below). The trypsin resistance of the third unidentified amino acid residue in peptide T10 and its sequencing characteristics, including phosphate release and the absence of an identifiable 3-phenyl-2-thiohydantoin-derivative signal at the third sequencer cycle, suggested a phosphorylated residue and were inconsistent with the Arg-292 assignment from the cDNA analyses ( Fig. 1) (5). The identity of this residue in the purified liver enzyme remained unresolved. Position 298 was identified as either Ala or Arg in two independently isolated cDNA clones; preference for Ala was based on the sequence of the corresponding T10 tryptic peptide (also Ala in lung PKN at position 298). Overall, these minor differences appeared most likely to reflect rat strain or cloning-generated variations.
Possible Regulatory Role of 116-kDa PAK-1/PKN Autophosphorylation Events-Indication that phosphorylation events regulate the catalytic activity of purified liver PAK-1/PKN came initially from the observation of a sigmoidal time curve of autophosphorylation in the absence of cardiolipin with a doubling of the reaction rate after an initial lag period of 10 min (Fig. 2). A further indication was the observed increase in the catalytic activity of the 116-kDa enzyme toward the substrate peptide S6-(229 -239) in the absence of any other form of activation (32% increase; p Ͻ 0.025) following preincubation with ATP for 10 min (Table I). This activation was also reflected in the appreciably diminished trypsin activability of the enzyme following the preincubation with ATP (Table I). The ATP effect appeared to be directed toward the regulatory domain as it was not observed with the trypsin-activated enzyme (Table I). A different time course of autophosphorylation was observed in the presence of cardiolipin, which was characterized by an appreciably greater initial rate, with no evidence of sigmoidal kinetics and a maximum stoichiometry appreciably greater  peptide kinase activity Purified liver PAK-1/PKN was treated with and without trypsin for 3 min at 30°C, and the reaction was terminated with soybean trypsin inhibitor as described previously (3). This was followed by a second incubation for 15 min at 30°C in buffer containing 5 mM MgCl 2 (see "Experimental Procedures") in the presence or absence of 0.2 mM ATP. Aliquots were removed, adjusted to 10 mM EDTA, and stored on ice for 10 min before assay of S6-(229 -239) peptide kinase activity for 5 min as described under "Experimental Procedures"; results are expressed as the means Ϯ S.E. of four determinations. Percentage activation of PAK-1/PKN by trypsin is given in parentheses. The ratios of peptide kinase activities for both trypsin-treated and untreated samples are expressed as ϩATP/ϪATP activity ratios.
To identify the relevant phosphorylation sites, the 32 P-labeled tryptic phosphopeptides, generated from 116-kDa PAK-1/PKN autophosphorylated in the absence of cardiolipin (4), were separated by RP-HPLC (3) and subjected to automated phosphopeptide sequence analysis (19,20). The tryptic phos-phopeptide map for the enzyme, autophosphorylated for sufficient time (20 min) to effect an increase in catalytic activity (Fig. 2), contained a major early eluting phosphopeptide peak (peak 1) plus several relatively small later eluting peaks (Fig.  3). The map for PAK-1/PKN autophosphorylated for 40 min was similar except for a selective increase in the peak 3 phosphopeptide(s). As an illustration of this, the recoveries from ϳ60 pmol of PAK-1/PKN for peak 1 peptide(s) were 37 and 38 pmol of peptide-bound phosphate for 20-and 40-min reactions, respectively, and for the peak 3 peptide, 13 and 24 pmol, respectively. Overall, this pattern suggested ordered phosphorylation events, with complete phosphorylation of the peak 1 peptide site occurring within the initial 20 min, followed by a relatively delayed phosphorylation of the peak 3 and other sites (5). The single peak 1 peptide was identified as the T14 peptide, Ser 375 -Gly-Ser(PO 4 )-Leu-Ser-Gly-Arg 381 containing the major Ser-377 phosphorylation site (Figs. 1 and 4), while the peaks 2 and 3 peptides were identified as Ala 62 -Thr-Xaa-Asp-Leu-Gly-Arg 68 (T2; Fig. 1) and Gln 500 -Gln-Gly-Gln-Thr(PO 4 )-Phe-Gln-Arg 507 (T19; Fig. 1), respectively. In the case of peptide T2, radiochemical sequencing was not attempted due to insufficient radioactivity; however, the detection of authentic 3-phenyl-2-thiohydantoin-Thr (Thr-63) at cycle 2, but not cycle 3, indicated that Thr-64 corresponded to the phosphorylation site. Three additional later eluting phosphopeptides, corresponding to peptide T20 (residues 525-543), a tryptic variant of peptide T20, and peptide T23 (residues 569 -610) (peaks 5, 4, and 6, respectively; Fig. 3), were found to contain the phosphorylation sites Ser-534, Ser-577, and Ser-579 ( Figs. 1 and 4A). These same phosphopeptides and their tryptic variants were also observed in RP-HPLC tryptic maps of PAK-1/PKN autophosphorylated in the presence of cardiolipin, except that the T2 peptide (RP-HPLC peak 2; Fig. 3) appeared to be missing, suggesting that cardiolipin suppressed the phosphorylation of the Thr-64 site (data not shown). A selective cardiolipin-dependent increase in the later eluting peaks (peaks 3-7; Fig. 3) was evident (data not shown). To overcome variations in RP-HPLC-based tryptic mapping due to interference by residual detergent in the tryptic digest (3), a more reproducible tryptic mapping procedure was investigated, involving in-gel tryptic digestion of SDS-PAGE-separated autophosphorylated PAK-1/ PKN (3) and subsequent isofocusing of the tryptic phosphopeptides on polyacrylamide gels (Fig. 5). Cardiolipin activation enhanced all of the detectable tryptic phosphopeptide bands of PAK-1/PKN, other than band 4, corresponding to the T14 peptide containing the Ser-377 phosphorylation site (Fig. 5). Peptides containing the cardiolipin-regulated sites that were recovered from RP-HPLC in sufficient quantities for sequence analyses included the T19, T20, and T23 peptides ( Fig. 1) with the phosphorylation sites Ser-504, Ser-534, Ser-577, and Ser-579 ( Figs. 1 and 4A). In addition, the T10 peptide (Fig. 1) chromatographed with a major cardiolipin-dependent peak of 32 P radioactivity during RP-HPLC (data not shown), which, as discussed above, gave rise to anomalous sequencing results, suggesting an unidentified phosphorylated residue at the third position, but which corresponded to the position of Arg-292 in the cDNA-predicted sequence (Fig. 1).  4A). Screening the sequencing and mass spectrometric characteristics of the various tryptic peptides derived from the 55-kDa fragment identified peptide T33 as a candidate phosphopeptide (Fig. 1). The detection of threonine adducts in the absence of authentic 3-phenyl-2-thiohydantoin-threonine signals at cycle 3 of the T33 sequence analyses suggested that Thr-778 was phosphorylated (30) (data not shown). Signals characteristic of authentic unmodified 3-phenyl-2-thiohydantoin-threonine signals were detected at all other threonine positions in this peptide. Matrix-assisted laser desorption time-of-flight mass spectrometric analysis of the pyridylethylated peptide identified a molecular ion (MH ϩ ) with an m/z of 2721, which, when compared with the theoretical m/z of 2640 for unmodified T33, confirmed the presence of a single phosphoryl group (additional mass of 80 Da) on Thr-778. The sequencing yields, comparable with those of other 55-kDa PAK-1/PKN tryptic peptides, indicated that this site was quantitatively phosphorylated in vivo. The Thr-778 site was found to be distinct from in vitro 55-kDa FIG. 5. Tryptic phosphopeptide mapping by isoelectric focusing. PAK-1/PKN was autophosphorylated for 40 min at 37°C in the presence (lane 1) or absence (lane 2) of 10 g/ml cardiolipin, isolated by SDS-PAGE, and subsequently digested with trypsin, and the 32 P-labeled tryptic phosphopeptides were analyzed by isoelectric focusing on polyacrylamide gels and subsequent PhosphorImager detection. Lane 3, pI standards: pI 8.0 peptide (upper band) and pI 4.5 peptide (lower band) (26). The major phosphopeptides in the minus-cardiolipin digest were Ser-Gly-Ser(PO 4 )-Leu-Ser-Gly-Arg (band 4) and Gln-Gln-Gly-Gln-Thr(PO 4 )-Phe-Gln-Arg (band 11; paraglutamate-cyclized NH 2 terminus). PAK-1/PKN autophosphorylation sites, which included the Ser-577 and Ser-579 autophosphorylation sites, also phosphorylated in whole PAK-1/PKN, plus the Ser-612 site (sites identified using strategy similar to that described for intact PAK-1/PKN (data not shown); results summarized in Fig. 4A).

Protease-resistant Catalytic Domain of PAK-1/PKN and Its
Three-dimensional Model of Catalytic Domain of PAK-1/ PKN-To gain further insight into the properties of the catalytic domain, particularly in relation to the role of phosphorylated Thr-778 and the restricted substrate specificity of the enzyme, a three-dimensional model was built based on the x-ray structure of cAPK (21,22). Sequence alignment of residues 616 -910 of PAK-1/PKN with the corresponding residues 40 -329 of cAPK reveals a sequence identity of 41% (Fig. 6), which indicates that the two sequences will have a similar fold (31). The stereochemical quality of the model after energy minimization is good and is comparable to cAPK, as judged by analysis with the program Procheck (32). Profile energy analysis with ProsaII (33) shows that the structure has good side chain packing, although the total energy profile is slightly less favorable than for the crystal structure of cAPK (data not shown).  (47)). The ␣-helices and ␤-strands are indicated as described for Fig. 6. The loops are colored orange and red; the S6-(229 -239) peptide substrate is yellow. Phosphothreonine 778, ATP, and the two magnesium ions are drawn as ball-and-stick models with the atoms colored according to their types: carbon, green; nitrogen, blue; oxygen, red; phosphorus, magenta; and magnesium, gray. B, the active site of PAK-1/PKN is shown in stereo view as a molecular surface with bound Mg 2 ATP (mostly hidden in the ATP-binding cleft) and is colored according to the electrostatic potential of the kinase domain (prepared with Grasp (48)). Blue and red indicate positive and negative electrostatic potentials, respectively. S6-(229 -239) substrate is rendered with a tube for the backbone and rods for side chains. Atom colors used for the peptide and ATP are the same as described for A, with the addition of white for hydrogen atoms.
The proposed structure of the catalytic domain of PAK-1/ PKN consists of the core kinase domain (residues 619 -878) plus a long C-terminal extension after helix I that packs onto the surface of the core (Fig. 7A). The kinase domain consists of two lobes between which ATP binds in a deep cleft. The phosphate group on Thr-778 binds in a similar pocket as Thr-197 in cAPK and interacts with the side chains of Arg-743 and Lys-767 (Arg-165 and Lys-189 in cAPK). There is no equivalent in PAK-1/PKN for the third basic residue in cAPK that interacts with the phosphate group (His-87).
Ribosomal S6-(229 -239) peptide, a preferred substrate for PAK-1/PKN (2)(3)(4), was modeled onto the PKI inhibitor peptide in the cAPK structure (Fig. 7, A and B). The oxygen atom of the S6 Ser-236 hydroxyl group is at a distance of 3.3 Å from the ␥-phosphorus atom of ATP (Fig. 7B). Its hydrogen atom forms a hydrogen bond with the predicted catalytic base, Asp-744. The basic residues at positions PϪ6 to PϪ3 of the substrate, previously shown to be important for PAK-1/PKN substrate efficacy (2), make contact with the acidic residues Asp-748, Glu-784, and Asp-909 on the surface of PAK-1/PKN. It is noteworthy that a residue outside the core kinase domain, Asp-909, can contribute to the binding of the substrate peptide and thus influence enzyme specificity. The leucine residues at the PϪ2 and Pϩ1 positions, S6 Leu-234 and S6 Leu-237, sit in two separate hydrophobic pockets of PAK-1/PKN made up of Phe-785 and the aliphatic part of the side chains of Lys-746 and Thr-781 and of Phe-779, Pro-783, and Leu-786, respectively. S6 Arg-238 at the Pϩ2 position forms an ion pair with Glu-660 at the beginning of helix C. This last interaction may be specific for the PAK/PKN/PRK family of kinases and influences both the substrate affinity and the substrate specificity. Mutation of Arg-238 to alanine reduces the K m by 20-fold, and the peptide is phosphorylated preferentially on Ser-235 instead of Ser-236 (2). In this case, the substrate is shifted one residue in the substrate-binding site. Arginines 231 and 232 can still contact acidic residues, and Arg-233 would reside in the deep PϪ2 pocket, at the bottom of which it can interact with Glu-811 (Fig.  7B). However, Leu-234 and Leu-237 would no longer reside in the hydrophobic pockets, possibly explaining the relatively unfavorable kinetics for this substrate (2). DISCUSSION The structural analyses of rat liver PAK-1/PKN have established that the predicted (34) catalytic domain consensus sequence (PAK-1/PKN-(619 -878)) is closely correlated with the structure of the minimum length catalytically active fragment (PAK-1/PKN-(612-946)) generated by limited tryptic digestion. In this fragment, the catalytic domain is extended at the C terminus by 68 residues. The trypsin resistance of the extension suggests a close interaction with the catalytic domain structure, possibly influencing substrate specificity (see above) and stabilizing the active conformation of the domain in a way similar to the proposed role of the phosphorylated C-terminal extensions of the PKCs (35,36). Consistent with this suggestion is the close packing of at least half of the C-terminal extension on the surface of the active conformation of the catalytic domain predicted in the three-dimensional molecular model (Fig. 7A). The active conformation also appears to be maintained by phosphothreonine 778 within the activation loop, a role similar to that of cAPK phosphothreonine 197 in the anchoring/alignment of catalytic domain residues functionally important in catalysis (29). The location of phosphothreonine 778 within a pocket that allows electrostatic interaction with Arg-743 and Lys-767 is consistent with the requirements of the active forms of cAPK and several other serine/threonine protein kinases (29). A requirement for the in vivo Thr-778 phosphorylation event creates the potential for regulation by an unidentified PAK-1 kinase and/or phosphatase similar to the regulation of PKC, mitogen-activated protein kinase, Cdk2, and p70 S6K kinase by phosphorylation events in their activation loops (29,37,38), although it is noted that, in the case of cAPK, Thr-197 is phosphorylated shortly after polypeptide synthesis and, thereafter, does not appear to exhibit rapid turnover (29).
Activation of PAK-1/PKN through interaction with cardiolipin, site-specific phosphorylation, or proteolytic cleavage involves suppression of the effects of the N-terminal negative regulatory domain (3,4). The unusual structural characteristics of this domain, which presumably determine function, include a striking similarity (16 -24% amino acid identity) between extended sequences within the N-terminal leucine zipper-based heptad repeat region of the regulatory domain (residues 1-330) (5,8) and the coiled-coil motifs of contractile proteins (e.g. myosins, tropomysin, and kinesins). 2 There is also an adjacent spectrin homology region (PAK-1/PKN-(300 -400), 20% identity to Drosophila spectrin-(2221-2333)). 2 The only structural regions conserved in the regulatory domains of PAK-1/PKN and the PKCs (14,36,39) are the PKC-⑀/PKC-V o -like sequence (Fig. 4A) and a sequence within the heptad repeats resembling a region of the yeast PKC-like kinases (40). The PKC V o region contains the pseudosubstrate sequence implicated in the intrasteric regulation/autoinhibition of the inactive forms of PKCs (36,39,41). The Thr-504 autophosphorylation site within the PAK-1/PKN pseudosubstrate-like sequence (Fig. 4A) has the potential to regulate any autoinhibitory function of this sequence. However, it should be noted that other potential autoinhibitory sequences are present, including the PAK-1/PKN-(39 -50) sequence, a peptide analogue of which has recently been shown to inhibit human recombinant PKN (42).
The lack of clear-cut specificity for acidic phospholipid and unsaturated fatty acid activators of the PAK-1/PKN/PRK1 enzymes (3, 4, 9 -11) and the apparent inaccessibility of the cytosolic enzyme to the most potent in vitro phospholipid activator, mitochondrially located cardiolipin (3,4), suggest that these lipids may not be the physiological activators. Presumably, cardiolipin mimics the action of some yet to be identified lipid second messenger or other activating ligand. The limited sequence similarity between the PAK-1/PKN-(50 -77) sequence located within the first heptad repeat region and the cardiolipin-binding presequence (residues 2-25) of yeast cytochrome oxidase subunit IV (28) (29% sequence identity; Fig. 4C) suggests a possible cardiolipin-binding/effector site. Consistent with an interaction of cardiolipin with this site is the apparent suppression of Thr-64 phosphorylation by cardiolipin in vitro. The basic region (residues 39 -61) adjacent to this proposed binding site could carry out a function similar to that of the cluster of basic residues in the C2 domain of the PKCs implicated in the binding of acidic phospholipid head groups (12,13) and is also of interest because of its possible function as a pseudosubstrate inhibitory sequence (42). Interestingly, these segments are contained within the region (PAK-1/PKN-(34 -103)) implicated in RhoA interaction (12,13). The location of the highly conserved Thr-64 (5, 6, 8) phosphorylation site within this region, albeit only weakly autophosphorylated in vitro, provides for an additional potential component of regulation, particularly in regard to RhoA binding (12,13).
The conservation of the Ser-377 phosphorylation site in all of the structurally defined PKNs and PRK (5,6,9) 2 suggests that it could be of general importance as a primary site of activation. An interesting possibility is that the transphosphorylation of Ser-377, either by PAK-1/PKN itself or by some other kinase(s), might contribute to the regulation of the enzyme in vivo. Why liver PAK-1/PKN is not fully phosphorylated at Ser-377 in vivo, given its spontaneous and quantitatively efficient Ser-377 autophosphorylation activity in vitro, is unclear. The PAK-1/PKN purification strategy did not eliminate the possibility of dephosphorylation of this site during purification and hence its availability for rephosphorylation in vitro. Alternatively, the phosphorylation of Ser-377 might be suppressed by an interaction of the cytosolic enzyme with another cellular component in vivo, analogous to the RhoA interaction, which is disrupted during enzyme isolation.
While proteolytic activation is not considered likely to constitute a physiological activation mechanism for PAK-1/PKN (3), the phenomenon could be relevant to the down-regulation of the enzyme in vivo. The sites of proteolytic activation, Arg-546 and Lys-611 (Fig. 4A), fall within a region rich in serine, threonine, and proline residues (PAK-1/PKN-(532-610)), which, by analogy with the PKC V3 hinge region (14,39), is defined as the hinge region of PAK-1/PKN (Fig. 4A). The presence of a high PEST score sequence within this region (PAK-1/PKN-(586 -609)), characteristic of proteins highly susceptible to intracellular proteolytic degradation (43), suggests a downregulation mechanism analogous to that operating with the PKCs (39), whereby ligand activation of the enzyme induces proteolytic lability, hence down-regulation via a conformational change exposing the PEST sequence. The location of a cluster of phosphorylation sites, Ser-534, Ser-577, and Ser-579, adjacent to the PEST sequence raises the further possibility that the proteolytic susceptibility of the hinge region might be regulated by these phosphorylation events.
The major sites phosphorylated under autophosphorylation conditions conform to the Arg-X-X-Thr 64 /Ser 377 -X-X-Gly-Arg-Ser motif, resembling motifs for a number of basic amino acid residue-directed protein kinases (44). The motif is approximated to within the structure of the efficiently phosphorylated S6-(229 -239) peptide, shown herein to be compatible with the molecular surface model of the PAK-1/PKN catalytic site (Fig.  7B), wherein the substrate's basic residues are positioned for efficient interaction with acidic clusters in the enzyme. Several other PAK-1 phosphorylation sites are not in close proximity to N-terminal basic residues and bear no structural resemblance to any other defined protein kinase motif. Exactly how PAK-1/ PKN selects such structurally diverse autophosphorylation substrate sequences, by either intramolecular or transphosphorylation events, cannot be readily explained based on our current knowledge of the structure and function of protein kinases.