Cell Communication and Signaling BioMed Central Review

Secreted factors and cell surface receptors can be internalized by endocytosis and translocated to the cytoplasm. Instead of being recycled or proteolysed, they sometimes translocate to the nucleus. Nuclear import generally involves a nuclear localization signal contained either in the secreted factor or its transmembrane receptor, that is recognized by the importins machinery. In the nucleus, these molecules regulate transcription of specific target genes by direct binding to transcription factors or general coregulators. In addition to the transcription regulation, nuclear secreted proteins and receptors seem to be involved in other important processes for cell life and cellular integrity such as DNA replication, DNA repair and RNA metabolism. Nuclear secreted proteins and transmembrane receptors now appear to induce new signaling pathways to regulate cell proliferation and differentiation. Their nuclear localization is often transient, appearing only during certain phases of the cell cycle. Nuclear secreted and transmembrane molecules regulate the proliferation and differentiation of a large panel of cell types during embryogenesis and adulthood and are also potentially involved in wound healing. Secreted factors such as CCN proteins, EGF, FGFs and their receptors are often detected in the nucleus of cancer cells. Nuclear localization of these molecules has been correlated with tumor progression and poor prognosis for patient survival. Nuclear growth factors and receptors may be responsible for resistance to radiotherapy.


Background
MAPKs catalyze the transfer of the γ-phosphate of adenosine triphosphate (ATP) to serine (S) or threonine (T) residues that precede proline (P) [1,2]; thus, these enzymes are termed proline-directed serine/threonine kinases. Although the sequences ST and TP are sufficient for phosphorylation to occur, the optimal sequence for phosphorylation by a MAPK is PX(S/T)P [1,3]. The majority of cellular proteins contain an SP or a TP sequence, yet, many of these proteins are not MAPK substrates [4], indi-cating that a mechanism exists for achieving substrate specificity for the MAPKs. This specificity is conferred by the substrate through a docking domain. In addition to underlying specificity, these docking interactions increase the catalytic efficiency of substrate phosphorylation [5][6][7].

MAPK docking sites
A MAPK docking site, distinct from the phosphoacceptor site, was first identified in c-Jun [8,9], a c-Jun N-terminal kinase (JNK) substrate; this site was designated the "δ domain". Subsequently, a JNK binding site in the transcription factor ATF-2 [10,11] and a motif termed the "dbox" of Elk-1 that binds ERK2 [4,12] were noted to be similar in sequence to the JNK binding site in c-Jun. Related motifs have been identified in a number of other proteins and have been given various names, including DEJL (docking sites for ERK and JNK, LXL) domain [4], kinase interaction motif (KIM) [13,14], MAPK-docking site [15,16], D box [5,12], D-site [17] and D-domain [6,[18][19][20]. It is important to note that, although these domains were identified based on the ability to bind one or more MAPK, there are differences in the consensus sequences used to identify each of them. For example, MacKenzie et al. [14] proposed a consensus KIM sequence of (V/L)X 2 (R/K)(R/K)X (3)(4)(5)(6) L, with V, L, R, and K representing the amino acids valine, leucine, arginine and lysine, respectively; Bardwell et al. [16] define a consensus MAPK-binding site sequence of (R/K) 2 X (2-6) (L/I)X(L/I), with I representing the amino acid isoleucine; and Kornfeld and colleagues [4] reported two consensus sequences for the DEJL domain: (K/R)X(X/K/R)(K/R)X (1)(2)(3)(4) (L/I)X(L/ I) and (K/R)(K/R)(K/R)X (1)(2)(3)(4)(5) (L/I)X(L/I). In the present studies we use the term D-domain and the consensus sequences reported by Kornfeld and colleagues [4].
Sharrocks and colleagues [21] report that D-domains are characterized by a cluster of basic residues positioned amino-terminal to an (L/I)X(L/I) motif followed by a triplet of hydrophobic amino acids that precedes a series of proline residues [17,21]. These investigators assessed the role of each of these regions in the binding of ERK2 and p38 to transcription factors, MEF2A, SAP-1, and Elk-1. They determined that mutation of the basic region of the transcription factors reduced their phosphorylation by both phospho-ERK2 and phospho-p38 [21]. This suggests that the basic residues are important for both ERK2 and p38 targeting of MAPK substrates. Mutation of the (L/ I)X(L/I) motif (also called the LXL motif) diminished phosphorylation of phospho-ERK substrates, whereas it is not required for phosphorylation of substrates by the MAPK, phospho-p38 [21]. It was also determined that the hydrophobic patch plays an important role in phosphorylation of the substrates by both phospho-ERK and phospho-p38; however, this patch is more important for p38 binding than ERK2 binding. Barsyte-Lovejoy et al. [21] concluded that the proline residues were not important in specificity determination of MAPK substrates. Therefore, the authors hypothesize that the proline residues may play a structural role within the motif. D-domains can show specificity for families of MAPKs; for example, the Elk-1 D-domain binds JNK and ERK, but not p38 [5,22]; both the SAP-1 and Elk-1 D-domains bind ERK2, whereas the SAP-1, but not Elk-1, D-domain binds p38α [22]. Other D-domains show specificity within a MAPK family; for example, the SAP-1 D-domain binds p38α and p38β, but not p38δ [22]. The D-domain can be positioned either N-or C-terminal to the phosphorylation site [7,12,16,19,23].
A second MAPK docking motif has also been identified: the FXFP motif, or DEF (docking site for ERK, FXFP) motif [4,18,22], where F and P represent the amino acids phenylalanine and proline, respectively. Binding and substrate phosphorylation can occur in the absence of the proline residue [14,18]; however, its presence does increase the effectiveness of the motif [18]. Thus, we chose to include the proline in our searches. The identity of the second (X) residue is highly variable [4,18]. In most, if not all, proteins, the FXFP motif is C-terminal to the phosphorylation site [4,18,19,24]. In general, it appears that the FXFP motif occurs more proximal to a phosphorylation site than is often the case for a D-domain [4,18,20,22,24]. The FXFP motif binds ERK2 and p38α [18,22], but not JNK3 [4] and p38β [4,22]. The FXFP motif and D-domain are each sufficient for MAPK docking; however, when both are present in a protein, they function additively [18,22].
We recently identified a D-domain sequence in PLC-γ1 and provided strong evidence that this sequence mediated an observed interaction between PLC-γ1 and phospho-ERK2 [25]. We have also reported that PLC-γ2, -β1, -β2, and -β4 each have at least one identifiable D-domain, as well co-immunoprecipitate with ERK2 [26]. Based on these observations, we have proposed that MAPK signaling and the metabolism of PIs are integrated. In order to substantiate this hypothesis, we sought to determine whether MAPK docking sites could be recognized in other enzymes that metabolize PIs; additionally, we sought support for extending this hypothesis to include the metabolism of IPs.

Overview of phosphatidylinositol and inositol phosphate metabolism and signaling
Eight PIs and more than 20 IPs have been identified [27][28][29][30]. Several reviews of the metabolism and cellular roles of these molecules have appeared [27][28][29][30][31][32][33][34][35]. As the physiologic functions of the PIs and IPs were not a primary focus of the present studies, we will not summarize this information here; instead, the interested reader is directed to the sources cited above; we acknowledge that this is only a partial listing of the reviews that have been written on these subjects. PIs are substrates for a variety of phospholipases, acyl transferases, kinases and phosphatases, while IPs are metabolized by a series of kinases and phosphatases. Of these enzymes, we have limited the scope of the present studies to kinases and phosphatases. In order to assist the reader in understanding the reactions catalyzed by the enzymes which we analyzed, the pathways for the metabolism of PIs and IPs by various kinases and phosphatases are shown in Figures 1A and 1B, respectively. We use the IUPAC-IUB nomenclature for the identification of phosphatidylinositol (PtdIns) and inositol (Ins) phosphates (P): numbers are used to designate the carbon atoms to which phosphate groups are bound and the total number of phosphate groups is designated by a subscript, with the exception that no subscript is employed to designate the presence of a single phosphate group. For example, phosphatidylinositol 4-phosphate is designated PtdIns4P and inositol 1,4,5-trisphosphate is designated Ins(1,4,5)P 3 . For a review of inositol phosphate chemistry the reader is referred to recent articles by Shears [36] and Irvine [37].

Search strategy
We obtained the primary sequences of human and mouse kinases and phosphatases that control the phosphoryla-A: Pathways of metabolism of phosphatidylinositols in animals. B: Pathways of inositol phosphate metabolism Figure 1 A: Pathways of metabolism of phosphatidylinositols in animals. This figure is based on similar figures in Toker [28] and Parker [29] and information provided in the Results and Discussion section. We are not aware of evidence that PtdIns5P is a substrate for a PI 3-kinase, producing PtdIns(3,5)P 2 ; therefore, an arrow has not been included for this reaction. B: Pathways of inositol phosphate metabolism. The pathways of the metabolism of inositol phosphates in animal cells are shown; additional pathways are present in plants and slime mold [30]. The PLC-catalyzed synthesis of 1,2-diacylglycerol (1,2-DAG) and Ins(1,4,5)P 3 is also shown. The figure is based on similar figures found in Irvine and Schell [30], Shears [36], and Irvine [37], as well as information provided in the Results and Discussion section.  [4]: (K/R)X(X/K/ R)(K/R)X (1-4) (L/I)X(L/I) or (K/R)(K/R)(K/R)X (1)(2)(3)(4)(5) (L/ I)X(L/I). We note that this strategy most likely failed to identify all possible sequences that may function as MAPK docking sites. For example, both the human and mouse Class I PI 3-kinase β include the sequence LILRRHGNLFI, which contains a KIM, as defined by MacKenzie et al [16], and a MAPK-docking site, as defined by Bardwell et al. [16], but not a D-domain according to the criteria that we employed. Similarly, the human and mouse Ins(1,3,4)P 3 5/6-kinase/Ins(3,4,5,6) 1-kinase contain the sequence LCRKRGXEVVQLNL (X is M in human and I in mouse), which fits the consensus sequences for a KIM and a MAPKbinding site, but not a D-domain. We chose to use the criteria of Kornfeld and colleagues based on its successful application in the identification of D-domains in PLC isozymes [25,26]. Although the list that we have compiled is likely to be incomplete, it does serve as a useful first approximation. We also searched each of the enzyme sequences for potential MAPK phosphorylation sites (i.e., (S/T)P sequences) and MAPK optimal phosphorylation sequences, PX(S/T)P. Finally, we note that we did not analyze the sequences of enzymes that control the metabolism of the diphosphorylated IPs.

Presentation of the data
In order to facilitate the presentation of the data, we separated human from mouse enzymes, kinases from phosphatases, and enzymes in which we identified a Ddomain and/or FXFP motif from those in which we did not. Human and mouse kinases having a D-domain and/ or FXFP motif are contained in Tables 1 and 3, respectively, whereas human and mouse kinases that do not have either of these sequences are presented in Tables 2  and 4, respectively. Similarly, human and mouse phosphatases containing a D-domain and/or FXFP motif are listed in Tables 6 and 8, respectively, while human and mouse phosphatases that are devoid of these sequences are contained in Tables 7 and 9, respectively.
In the following discussion we use the terms "alternative pair" and "overlapping" in referring to relationships of Ddomains. An alternative pair of D-domains has the same amino-terminus and two possible carboxyl-termini: e.g., 15RRRDAVIAL and 15RRRDAVIALGI in human PI 4kinase α (Table 1). We use the term overlapping to identify D-domains that have distinct amino-termini and carboxyl-termini, but share a region of sequence: e.g., 376RNSKGERLLL and 379KGERLLLYI found in human PI4P 5-kinase type Iα (Table 1).

PI 3-kinase
Phosphatidylinositol 3-kinase (also called phosphoinositide 3-kinase, PtdIns 3-kinase, PI 3-kinase, and PI3K) isozymes catalyze the phosphorylation of the 3-position of the inositol ring of phosphatidylinositols [32]. Three classes (I, II, and III) of PI 3-kinase have been identified [38,39]. These classes are differentiated on the basis of their subunit composition, substrate specificity and mechanisms of regulation. The Class I enzymes are heterodimers of a regulatory subunit, of various sizes, and a catalytic subunit of approximately 110 kDa. Three distinct forms of the catalytic subunit (p110α, p110β, and p110δ) and five forms of the regulatory subunit (p85α, p85β, p55α, p55γ, and p50α) have been identified. The class IB isoform consists of a p101 regulatory subunit coupled to a p100γ catalytic subunit. Class I enzymes catalyze the synthesis of PtdIns3P, PtdIns(3,4)P 2 , and PtdIns(3,4,5)P 3 [40,41]; PtdIns(4,5)P 2 may be the preferred substrate in vivo [42]. Class II isozymes are monomeric catalytic subunits (α, β, and γ) containing a carboxyl-terminal C2 domain; these enzymes may be referred to as PI3K-C2. PtdIns and PtdIns4P, and under certain conditions PtdIns(4,5)P 2 , are substrates for the Class II PI 3-kinase [39,43]. A single Class III isozyme, which is specific for PtdIns, has been identified [44].
The Class I p110α and p110γ human (Table 1) and mouse (Table 3) proteins contain D-domains, whereas the p110β Table 3: The Class II PI3-kinase C2-α contains a D-domain that is conserved in the human (Table 1) and mouse (Table 3) isozymes. PI 3-kinase C2-α was the only PI 3-kinase isozyme in which we found an optimal MAPK phosphorylation sequence (Tables 1, 2, 3, 4). The human and mouse PI 3-kinase C2-α proteins each contain three such sequences; a sequence alignment revealed that two of these (PLTP and PATP) are conserved, while the third, 118 PVTP and 1374 PFSP in human and mouse, respectively, is unique. Both the human (Table 1) and mouse (Table 3) PI 3-kinase C2-γ contain D-domains. A sequence align-ment of these two proteins revealed that only the last Ddomain ( 1347 KHMKNIHL in human and 1409 KHLKNIHL in mouse) aligns; all of the other identified D-domains are unique to one or the other protein. These results indicate the MAPK-dependent regulation of the human and mouse Class II PI 3-kinases may be significantly different and, thus, caution should be exercised when comparing studies on these two proteins.

Kinases (mouse) that have an FXFP motif and/or D-domain sequence. The number of potential MAPK phosphorylation sites, sequences fitting the optimal MAPK phosphorylation consensus sequence of PX(S/T)P, and FXFP motifs and D-domain sequences in kinases that use phosphatidylinositols or inositols as substrates are listed in the
The human (Table 1) and mouse (Table 3) Class III PI 3kinase contains a conserved D-domain. Neither of these proteins contains an FXFP motif or an optimal phosphorylation sequence for MAPKs.
PI4Kα (also called PI 4-kinase 230) contains three alternative pairs of D-domains that are conserved in human (Table 1) and mouse (Table 3) sequences. PI4Kα also contains an FXFP sequence and two optimal MAPK phosphorylation sequences that are conserved in human and mouse proteins. PI 4-kinase β (also called PI 4-kinase 92) contains three D-domains that are conserved in the human (Table 1) and mouse (Table 3) isozymes. The first of these ( 232 RGTKLRKLIL) overlaps with an identifiable bipartite nuclear localization sequence motif [45]. The human and mouse PI 4-kinase-β do not contain an FXFP motif or optimal MAPK phosphorylation sequence. The    (Table 4), PI 4kinase type II-β contains a single D-domain, while neither protein contains a sequence fitting the MAPK optimal phosphorylation sequence or the FXFP motif.

PIP kinase
Phosphatidylinositol phosphate kinases (PIP kinases, PIPKs) utilize PtdIns3P, PtdIns4P, and PtdIns5P as substrates, catalyzing the synthesis of PtdIns(3,4)P 2 , PtdIns(3,5)P 2 , and PtdIns(4,5)P 2 ( Table 5). These enzymes are also able to catalyze the formation of PtdIns5P and PtdIns(3,4,5)P 3 ( Table 5). Three types (I, II and III) of PIP kinase are defined based on primary sequence, substrate specificity, and subcellular localization [51]. These families are often designated by the reaction that they catalyze most efficiently. Thus, the Type I PIP kinases are also termed PI4P 5-kinase, the Type II PIP kinases are termed PI5P 4-kinases, and the Type III PIP kinases are called PI3P 5-kinases. Three forms (termed α, β, and γ), each with multiple splice variants, of Type I and Type II PIP kinase have been identified [52][53][54][55]. It should be noted that the nomenclature for the αand β-forms of the human and mouse Type I PIP kinase are reversed: that is, the human Type Iα enzyme corresponds to the mouse Type Iβ enzyme, and vice versa.
Each of the Type I isozymes contain at least one identifiable D-domain that was conserved in humans and in mice. In human PI4P 5-kinase Type Iα (Table 1) an alternative pair of D-domains ( 379 KGERLLLYI and 379 KGERLLLYIGI) overlaps with a different D-domain ( 376 RNSKGERLLL). This observation raises the intriguing possibility that these sequences are specific for binding of a particular MAPK, or MAPK family, and that there is competition among MAPKs for binding, with the likelihood that a mechanism(s) exists for the regulation of this binding. Further, the MAPK that is bound may affect the kinetics of the signal that is generated and, thus, the downstream response. The identified alternative pair of D-domains in PI4P 5kinase type Iγ (PIPKIγ) (Tables 1 and 3) may be responsible for binding phospho-ERK1, which has been shown to catalyze the phosphorylation of serine 650 in PIPKIγ [56]. Phosphorylation of serine 650 inhibits the binding of talin to PIPKIγ and may play a role in synaptic neurotransmission and focal adhesion disassembly during mitosis. None of the Type I PIP kinases that we analyzed contains an FXFP motif (Tables 1 and 3). Each of the Type I PIPKs that we analyzed contains an optimal MAPK phosphorylation site that is conserved in human and mouse proteins. In the case of PIPKIγ, this is not serine-650. However, it is possible that other MAPKs are capable of phosphorylating threonine-512 (human)/threonine-511 (mouse).
The Type II isozymes (Tables 2 and 4) were found to be devoid of D-domain sequences and FXFP motifs. However, the Type IIα and Type IIβ isozymes do contain an optimal MAPK phosphorylation sequence that is conserved in human (Table 2) and mouse (Table 4) protein sequences.
The PIKfyve human (Table 1) and mouse (Table 3) sequences each contain six MAPK optimal phosphorylation sequences. This is the most that we found in any of the sequences that we analyzed. These enzymes contain three conserved D-domain sequences, but are devoid of an FXFP motif.

Inositol monophosphatase
Inositol monophosphatase is a Mg 2+ -dependent enzyme that catalyzes the hydrolysis of Ins1P, Ins3P and Ins4P, as well as several related compounds, but does not hydrolyze Ins2P [74][75][76][77][78]. Inositol monophosphatase has received significant attention as a potential site of action for lithium in the treatment of bipolar disorder. However, recent studies have identified a number of other lithium targets, as well [79].
Three forms of inositol monophosphatase have been identified: A1 (IMPA1), A2 (IMPA2) and A3 (IMPA3). Of these three forms, only inositol monophosphatase A1 contains an identifiable D-domain (Tables 6 and 8). The human and mouse sequences differ by two amino acids in this region resulting in an alternative pair of D-domains in the mouse isoform (and 261 RIAKEIEI and 261 RIAKEIEIIPL) and a single D-domain in the human isoform ( 261 RIAKEIQVIPL). We did not find an optimal sequence for MAPK-catalyzed phosphorylation or an FXFP motif in Table 7:

Isozyme
GenBank Accession # # amino acids # (S/T)P sites Site ( any of the human or mouse inositol monophosphatase isozymes.
When we analyzed the human (Table 7) and mouse (Table 9) myotubularin protein sequences, we did not find either an FXFP or D-domain sequence. However, the human sequence does contain two optimal MAPK phosphorylation sequences not present in the mouse sequence. We identified a conserved D-domain in the human (Table 6) and mouse (Table 8) MTMR1 protein, but did not find an optimal sequence for MAPK phosphorylation or FXFP motif in either protein. Neither the human (Table 7) nor the mouse (Table 9) MTMR2 has an optimal sequence for MAPK phosphorylation, a Ddoman, or an FXFP motif. Although the human (Table 7) and mouse (Table 9) MTMR3 and MTMR6 each contain a sequence fitting the optimal sequence for MAPK phosphorylation, they do not contain an identifiable Ddomain or FXFP motif.

4-phosphatase
Two forms (Types I and II) of inositol polyphosphate 4phosphatase have been identified [95][96][97]. These enzymes cleave the 4-phosphate from Ins(1,3,4)P 2 , Ins (3,4)P 2 , and PtdIns(3,4)P 2 . The Type I 4-phosphatase has been reported to localize to endosomes, where it plays an important role in the generation of PtdIns3P [98]. In growth factor-stimulated cells the Type I 4-phosphatase also localizes to plasma membrane ruffles, where it hydrolyzes PtdIns(3,4,)P 2 , thereby regulating the association of PtdIns(3,4)P 2 -binindg proteins with the plasma membrane [98]. We identified a conserved D-domain in human (Table 6) and mouse (Table 8) Type I 4-phosphatase. However, we did not identify an optimal phosphorylation sequence or an FXFP motif in this isoform. In contrast, we did not identify a D-domain in the human Type II 4-phosphatase (Table 7), although it does contain a consensus MAPK phosphorylation site. Human Type II 4-phosphatase also does not contain an FXFP motif.
The Sac phosphatase domain, a region of sequence homology found in several yeast, plant and animal proteins, can hydrolyze the 3-, 4-, or 5-position phosphate from PIs, although vicinal phosphate groups are resistant to hydrolysis [99,100]; thus, PtdIns3P, PtdIns4P and PtdIns(3,5)P 2 are substrates, whereas PtdIns(4,5)P 2 is not. The PtdIns4P phosphatase activity, but not PtdIns3P or PtdIns(3,5)P 2 phosphatase activities, of mammalian Sac1 complements phenotypic defects observed in yeast having deletions of Sac1p [101], indicating that the 4-phosphatase activity of these proteins is the most important in vivo. The human (Table 6) and mouse (Table 8) Sac1 proteins each contain five identifiable D-domains. Two of these D-domains are overlapping sequences ( 345 KNMRWDRLSI and 348 RWDRLSILL) and two are an alternative pair of D-domains ( 517 RDWKFLAL and 517 RDWKFLALPI). Neither the human nor mouse protein contains an optimal phosphorylation site for MAPKs or an FXFP motif.
We did not identify a D-domain, FXFP motif or optimal MAPK phosphorylation sequence in either the human (Table 7) or mouse (Table 9) Type I 5-phosphatase. Sev-eral of the Group II 5-phosphatases (Tables 6 and 8) contain an identifiable D-domain. OCRL is the gene responsible for occulocerebrorenal dystrophy or Lowe's syndrome, when mutated [27]. Both the human (Table 6) and mouse (Table 8) OCRL proteins contain a D-domain, but do not contain an FXFP motif or an optimal sequence for MAPK phosphorylation. Synaptojanin 1 and synaptojanin 2 are neuronal proteins that play a role in synaptic vesicle trafficking. They contain both a 5-phosphatase domain and a Sac phosphatase domain [99]. The 5-phosphatase domain is responsible for the reported PtdIns(4,5)P 2 -hydrolyzing activity of synaptojanins, while the Sac domain of synaptojanins accounts for their  ability to also hydrolyze other PIs, such as the PtdIns4P product generated by the action of its 5-phosphatase domain [105]. Both synaptojanin 1 and 2 contain optimal phosphorylation sites for MAPKs. However, only synaptojanin 2 contains a D-domain, which is conserved in the human (Table 6) and mouse (Table 8) proteins. The synaptojanin sequences that we searched do not contain an FXFP motif. The human ( Table 6) and mouse (Table 8) 75-kDa inositol polyphosphate 5-phosphatases (inositol polyphosphate 5-phosphatase B) contain two conserved D-domains, but are devoid of an FXFP motif and an optimal sequence for MAPK phosphorylation.

mouse) that have an FXFP motif and/or D-domain sequence. The number of potential MAPK phosphorylation sites, sequences fitting the optimal MAPK phosphorylation consensus sequence of PX(S/T)P, and FXFP motifs and D-domain sequences in phosphatases that use phosphatidylinositols or inositols as substrates are listed in the
The Group III SH2-containing inositol 5'-phosphatase 1 (SHIP1) is a hematopoietic-specific enzyme that hydrolyzes both PtdIns (3,4,5) 3 and Ins(1,3,4,5)P 4 [106]. In addition, SHIP1 is able to hydrolyze the 4-phosphate from PtdIns (4,5) in vitro, thereby generating PtdIns5P [107]. The human ( Table 6) and mouse (Table 8) SHIP1 contain an identifiable D-domain. The human SHIP1 contains four optimal MAPK phosphorylation sequences; corresponding sequences for two of these are present in the mouse protein, whereas two are unique to the human protein. SHIP1 does not contain a sequence conforming to an FXFP motif. The distribution of SHIP2 is more ubiquitous than is that of SHIP1 [106]. The human SHIP2 is devoid of a D-domain or an FXFP motif (Table 7). It does contain four sequences that fit the optimal phosphorylation sequence for a MAPK. The human ( Table 6) and mouse (Table 8) Group IV 5-phosphatase contain a conserved optimal sequence (PRSP) for MAPK phosphorylation; the human protein contains an additional optimal sequence of MAPK phosphorylation ( 55 PATP). Both the human and mouse Group IV 5-phosphatase contain an alternative pair of D-domains, as well as two other Ddomains; neither contains an FXFP motif.

Additional sequence analyses
We examined each of the D-domains that we identified to determine if it overlaps with a KIM [14] or fits the MAPKdocking site consensus sequence defined by Bardwell and colleagues [15,16]. We found only one instance of overlap with a KIM: in the human PI 4-kinase α, the sequence 12 LDERRRDAVIALGI not only contains the alternative pair of D-domains that we identified (Table 1), but also contains a KIM. Examination of the sequences found in Tables 1, 3, 6, and 8 revealed that, in several instances, the D-domain sequence fits the MAPK-docking sequence of Bardwell and colleagues. It should be noted that in several proteins we identified, but did not catalog, one or more sequence that conformed to the MAPK-docking motif of Bardwell and colleagues but did not conform to the more restrictive sequence that we used for a D-domain.
Finally, we also analyzed the sequences of PtdIns synthase (CDP-1,2-diacyl-sn-glycerol:myo-inositol 3-phosphatidyltransferase) isozymes, which catalyze the production of PtdIns from cytidine diphosphodiacylglycerol and myoinositol. We did not identify either an FXFP or a Ddomain sequence in human or mouse PtdIns synthase sequences (GenBank:NP_006310, NP_665695, and NP_620093; data not shown). Further, we identified a MAPK phosphorylation site only in the human isoform 2.
These results indicate that the isozymes that catalyze PtdIns synthesis are not likely to be directly regulated by MAPK; however, they do not rule out the possibility that MAPKs may control PtdIns synthesis via an effect on myoinositol (e.g., via an effect on inositol monophosphatase A1) or cytidine diphosphodiacylglycerol levels.

Conclusion
We found a high degree of conservation of D-domain sequence and location in human and mouse proteins. Notable exceptions were PI 3-kinase C2-γ, PI 4-kinase type IIβ, and inositol polyphosphate 1-phosphatase. For each of these proteins, either the human or mouse protein contained one or more D-domains that were not found in the other, with the differences in PI 3-kinase C2-γ being the most striking.
Other than within subtypes of an isozyme, we found no evidence of sequence conservation of D-domains in enzymes that metabolize PIs and IPs. That is, the primary sequences of the D-domains that we found were quite variable. This indicates that, both within a family and across families of these enzymes, there may be specificity for binding interactions with MAPKs. Detailed studies examining the relative specificities of each of these proteins for individual MAPKs, if they indeed bind MAPKs, and the contribution of individual amino acids within these sequences to MAPK binding, will provide important information for the development of tools aimed at modifying the integration of MAPK and signaling via PIs and IPs.
When D-domains were found in more than one location in a protein, each D-domain had a unique sequence. This is noted because it is not a universal property of proteinprotein interaction domains when present in multiples in a protein-e.g., an SH 2 -binding motif is commonly present in multiple copies in docking proteins [108]. Within a protein, it is possible that each D-domain binds a unique MAPK, or set of MAPKs, allowing for interaction with more than one MAPK signaling pathway, and, thus, the regulation of the metabolism of a particular PI/IP by various combinations of stimuli, or the formation of differing combinations of signaling complexes allowing for the generation of various downstream signals. Further, it is also possible that each interaction is independently regulated. Therefore, it is reasonable to think that these D-domains do not simply serve to amplify signaling through a single signaling pathway, but, instead, allow for the integration of multiple MAPK pathways with the metabolism of a specific PI or IP.
The frequency of occurrence of an identifiable FXFP motif in the enzymes that we analyzed was significantly less than that of a D-domain. There were only five enzymes in which we found a motif conforming to the sequence FXFP; four of these were in kinases (PI 3-kinase δ, PI 4kinase α, Ins(1,4,5)P 3 3-kinase B, and Ins(1,3,4,5,6) 2kinase) and only a single one was found in a phosphatase (PTEN). It is noteworthy that of all the enzymes that we analyzed, only PI 4-kinase α and Ins(1,4,5)P 3 3-kinase B contain both a D-domain and an FXFP sequence. Finally, we found several sequences fitting an FXF motif (data not shown). As noted in the Introduction, the FXF sequence has been reported to be sufficient for MAPK binding, indicating that these sites may also bind MAPKs.
At this time, we can only speculate on the physiologic significance of the presence of MAPK binding domains in enzymes that control the metabolism of PIs and IPs. We have previously shown that phospho-ERK2-dependent phosphorylation of PLC-γ1 opposes tyrosine kinasedependent activation of PLC-γ1 [25]. Similarly, MAPKs may regulate (either stimulating or inhibiting) the catalytic activity, or specificity, of kinases and phosphatases that are involved in the metabolism of PIs or IPs, and thereby exert regulatory actions on PI-and/or IP-depend-ent signaling pathways. Intriguing possibilities exist when a kinase and phosphatase are present in the same complex and one or both of them bind a MAPK. For example, the p85 subunit of Class I PI 3-kinase has been reported to form a complex with Type I inositol polyphosphate 4phosphatase [109], SHIP1 5-phosphatase [110], and Type IV 5-phosphatase [111]. In these complexes, MAPKs may regulate the relative level or turnover of the substrates and products; for example, by enhancing PI3-kinase activity and associated SHIP1 (or Type IV) 5-phosphatase activity, it would be possible to increase PtdIns(3,4)P 2 levels without increasing PtdIns(3,4,5)P 3 levels. Other scenarios (e.g., delayed activation kinetics of the associated 5-phosphatase) are also imaginable, resulting in a transient rise in PtdIns(3,4,5)P 3 with a delayed elevation in PtdIns(3,4,)P 2 levels. Similarly, complexes of the Type I inositol polyphosphate 4-phosphatase and PI 3-kinase could produce locally elevated levels of PtdIns3P, without elevating PtdIns(3,4)P 2 levels, or elevate PtdIns(3,4)P 2 and, with a delay, PtdIns3P. It is also possible that the interaction between a MAPK and a PI/IP kinase or phosphatase may recruit the MAPK to a multimolecular signal transduction complex containing components of pathways that regulate the activity of the MAPK (e.g., binding of a MAPK to PI 3-kinase may act to recruit the MAPK to a growth factor signaling complex containing Ras and a MAPK kinase) or target the MAPK to a particular subcellular localization (e.g., binding to Type I inositol polyphosphate 4-phosphatase may act to target the MAPK to endosomes). It should be noted that a protein can serve as a MAPK substrate without directly binding the MAPK: c-Jun-bound proteins that lack a JNK binding site can be phosphorylated by JNK [112]. Thus, our inability to identify a MAPK binding site in a protein does not preclude it from being a MAPK substrate. For example, the MAPK optimal phosphorylation site identified in Type IIα PIPK could be phosphorylated by a MAPK bound to Type Iα PIPK, which has been shown to co-immunoprecipitate with Type IIα PIPK [113]. Similarly, although we did not identify a Ddomain or FXFP motif in the 3-phosphatase myotubularin, it may be a MAPK substrate when bound to the 3phosphatase adaptor protein (3-PAP) subunit [114], which has four recognizable D-domains (data not shown). In fact, human myotubularin does contain two optimal MAPK phosphorylation sequences ( However, to our knowledge, a direct interaction between a PI 3-kinase and a MAPK has not been demonstrated, and the mechanisms underlying the apparent PI 3-kinase-dependent regulation of MAPKs remain speculative. In the case of PI 3kinase γ, the effect could be mediated by MEK-1, a MAPK kinase which has been shown to be an in vitro substrate of PI 3-kinase γ [115]. Wen et al [65] have shown that ERK1/ 2 activation in response to suboptimal stimulation of thymocytes is dependent on Ins(1,4,5)P 3 3-kinase B; they propose a model in which Ins(1,4,5)P 3 3-kinase Bdependent production of Ins(1,3,4,5)P 4 acts to sequester an Ins(1,3,4,5)P 4 -binding GTPase-activating protein 1 [116], promoting Ras-dependent activation of ERK1/2. SHIP1, which contains an identifiable D-domain, but no FXFP motif, has been shown to be a negative regulator of JNK activation in B cells [117], ERK1/2 activation in the erythropoietin-dependent cell lineAS-E2 [118], and MAPK (ERK1/2, JNK and p38) activation in RAW264.7 macrophages [119]. In the latter case, the action was shown to be independent of the SHIP1 5-phosphatase activity [119].
Interestingly, SHIP2, which does not have an identifiable D-domain or FXFP motif, has been reported to not exert an effect on cellular MAPKs [120,121]. PTEN, which does not contain a D-domain sequence but does contain an FXFP motif, has been reported to inhibit insulin-stimulated ERK1/2 activation in MCF-7 epithelial breast cancer cells [122,123]. Weng et al. [122] concluded that the effect of PTEN is the result of PTEN-dependent dephosphorylation of the insulin receptor substrate 1, and consequent coupling to ERK1/2 activation. In contrast to the studies of Eng and colleagues, Tang et al. [121] reported that short interfering RNA-induced reductions of PTEN expression in 3T3-L1 adipocytes did not affect insulin-dependent signaling to ERK1/2.
In conclusion, there appear to be a plethora of potential sites of crosstalk between MAPK signaling pathways and the enzymes controlling cellular PIs and IPs, and, thus, the signaling pathways that are regulated by the levels of these intracellular signals. We hope that the identification of these sites of signal integration will initiate a series of studies aimed at determining whether these interactions occur in vivo and the physiologic relevance of each to cellular responding.