Structural Features that Distinguish Inactive and Active PI3K Lipid Kinases

Abstract

PI3K lipid kinases signal through the PI3K/Akt pathway, regulating cell growth and proliferation. While the structural features that distinguish between the active and inactive states of protein kinases are well established, that has not been the case for lipid kinases, and neither was the structural mechanism controlling the switch between the two states. Class I PI3Ks are obligate heterodimers with catalytic and regulatory subunits. Here, we analyze PI3K crystal structures. Structures with the nSH2 (inactive state) are featured by collapsed activation loop (a-loop) and an IN kinase domain helix 11 (kα11). In the active state, the a-loop is extended and kα11 in the OUT conformation. Our analysis suggests that the nSH2 domain in the regulatory subunit regulates activation, catalysis and autoinhibition through the a-loop. Inhibition, activation and catalytic scenarios are shared by class IA PI3Ks; the activation is mimicked by oncogenic mutations and the inhibition offers an allosteric inhibitor strategy.

Introduction

Phosphatidylinositol 3-kinases (PI3Ks) are a family of lipid kinases essential for the PI3K/Akt pathway in proliferation, differentiation, cell mobility, and survival [1–3]. Their roles in the immune system and cancer drove the intense pharmaceutical efforts over the past decades to target them [4–8]. Class I PI3Ks phosphorylate phosphatidylinositol (4,5)-bisphosphate (PIP₂) to phosphatidylinositol (3,4,5)-trisphosphate (PIP₃) on the membrane, to which downstream kinases including Akt bind for signal transduction [9]. Class I PI3Ks are obligate heterodimers consisting of catalytic and regulatory subunits. They have two subclasses, IA (PI3Kα, PI3Kβ, and PI3Kδ) and IB (PI3Kγ) [3,10]. The p110 catalytic subunits in class I have five conserved domains. In class IA, p110 interacts with the p85 regulatory subunit, while in IB it binds to the p101/p84 regulatory subunits [11]. p85 in class IA and p101/p84 in class IB have different sequences/domains, following the distinct regulation of PI3K actions [11–13].

Class I PI3Ks are activated by receptor tyrosine kinases (RTKs) and Ras superfamily members [14,15]. Ras, a superfamily member, interacts with the Ras binding domain (RBD) of the p110 catalytic subunit to promote membrane localization of PI3Ks [16]. RTKs' phosphorylated tyrosine (pY) motifs have high binding affinity to the nSH2/cSH2 domains in the p85 regulatory subunits. The motifs compete with the p110 catalytic subunit whose interactions with the nSH2/cSH2 domains autoinhibit PI3K. Their binding releases the autoinhibition and activates PI3K [17,18]. While cSH2 autoinhibition applies to PI3Kβ and PI3Kδ [19,20], nSH2 plays a general and dominant role in inhibiting the basal activities of class IA [21,22]. Its release from the p110 catalytic subunit by either pY motifs in RTKs or oncogenic mutations dramatically increases PI3K activities [23]. The hydrogen/deuterium exchange mass spectrometry indicated structural change in PI3K upon nSH2 release and activation, illustrating the key events that can be mimicked by oncogenic mutations [24]. Recent computational studies showed that the disruption of the nSH2 interaction with the catalytic subunit triggers structural rearrangement between the iSH2 and kinase domains, shifting the PI3Kα ensemble toward the active state [25,26].

Structurally, the lipid kinase domains in PI3Ks resemble those in protein kinases. In protein kinases, the DFG motif, αC helix, and the activation loop (a-loop) collectively define the inactive and active states [27]. The structural features that distinguish the inactive and active states of PI3K lipid kinases have been elusive. This is likely due to the strong domain–domain packing of the kinase domain in the p110 catalytic subunit. nSH2 has no contact with the catalytic, ATP pocket in the kinase domain. It has also been unclear how the inhibitory nSH2 domain interferes with PI3K catalysis through the structural components in the kinase domain.

In this work, we analyze the crystal structures of class I PI3K isoforms, aiming to better understand PI3K autoinhibition, activation, and catalysis. Our results suggest that the nSH2 domain mediates PI3K functions through a-loop dynamics. The collapsed a-loop and the kα11 IN conformation appear to be the distinctive structural features for the inactive PI3K conformation. They become extended and OUT in the activated PI3Ks upon nSH2 release. Our results further show that this mechanism is adopted by oncogenic mutations and outline the allosteric PI3K drug discovery strategy.

Results

The crystal structures of class I PI3Ks

As of now (April 1st, 2020), a total of 203 crystal structures of class I PI3Ks have been solved and deposited in the Protein Data Bank (PDB), 51 for PI3Kα (50 human and 1 mouse), 2 for PI3Kβ (2 mouse), 99 for PI3Kγ (93 human and 6 pig), and 51 for PI3Kδ (8 human and 43 mouse). There are two types of PI3K structures, (i) dimeric, with both catalytic and regulatory subunits and (ii) monomeric, with the N-terminal adaptor-binding domain (ABD)-truncated p110 catalytic subunit (Table S1). Class IA isoforms (PI3Kα, PI3Kβ, and PI3Kδ) have both types. The class IB (PI3Kγ) structures are monomeric. The interactions between the catalytic and regulatory subunits mediate PI3K inhibition and activation. In this work, we focus on the dimeric structures of the class IA PI3K isoforms. Class IB PI3K and other monomeric PI3K structures will not be discussed further.

PI3Kα, PI3Kβ, and PI3Kδ share sequence and structural similarities. Sequence alignment shows that the p110 catalytic subunits in PI3Kα, PI3Kβ, and PI3Kδ have 347 identical and 320 similar residues, leading to a ~31.4% sequence identity and ~60.4% sequence similarity (Figure S1). The p110 catalytic subunits in class IA PI3K isoforms are structurally conserved. The superimposed class IA PI3K dimeric structures show that the secondary structures of the ABD, RBD, C2, and helical domain are mostly conserved among isoforms (Figure S2A). The kinase domain exhibits structural variations in the C-terminal regulatory arch [28] (Figure S2B). The structures of the regulatory subunits in class IA PI3K are diverse; while iSH2 is always present, the inhibitory nSH2 and cSH2 domains are missing in some structures, leading to three types of the regulatory subunits, niSH2, iSH2, and icSH2 (Table S2).

iSH2 movements in PI3Ks upon nSH2 release

To explore the PI3K conformations, we calculated the domain–domain distances in p110 catalytic subunits. The sequence/structure convergence was evaluated and employed to ensure fair comparisons among isoforms (details in Materials and Methods). The distance profiles indicate variations in the distances between the C2 and kinase domains (Figure S3). When we labeled the C2-kinase domain distances based on the type of regulatory subunit involved, we identified a correlation between the presence of nSH2 and the C2-kinase domain distances (Figure 1(a)). In PI3K structures with the nSH2 domain, the center of mass distances between the C2 and kinase domains are ~42.9–44.1 Å (except 3HHM and 3HIZ, which will be discussed below). Upon nSH2 release, the distances increase to ~45.0–45.5 Å. This indicates that nSH2 release leads to the movement of C2 relative to the kinase domain (Figure 1(b)).

Figure 1.

Structural rearrangement of PI3Ks upon nSH2 release. (a) PI3Ks in the absence of nSH2 (marked by the red labels) reveal larger C2–kinase domain distance than those containing the nSH2 domain (marked by the black labels). The calculated distance denotes the center of mass distance between the C2 and the kinase domain. The orange labels denote the PI3Kα H1047R mutants with the nSH2 domain. (b) Superimposition of the crystal structures of PI3K in the absence (red cartoon) and presence (gray cartoon) of nSH2. The movements of iSH2 and C2 domains relative to the kinase domain are visible. KD, kinase domain.

The iSH2 domain in the p85 regulatory subunit interacts with the C2 domain. When the C2 domain moves, the iSH2 moves with it, apart from the C-lobe of kinase domain (Figure S4). Upon nSH2 release, the termini of the iSH2 domain lost densities in some PI3K crystal structures (PDB: 5DXH, 2Y3A), indicating that the iSH2 becomes flexible (Figure S4A and B). In PI3K structures containing the nSH2 domain, the iSH2 interacts with the first basic box (xKxK) in the a-loop (Figure S5A). These interactions are largely disrupted upon nSH2 release. The basic residues in the a-loop are released from the iSH2 interactions, shifting towards the kinase domain surface with the side-chains rotations (Figure S5B). The flexible iSH2 can coordinate the membrane binding surface of PI3Ks. We have sampled these conformational changes in PI3Kα activation in the previous simulations [25]. These changes can collectively promote the membrane localization of PI3Ks, in line with experiments [29].

The collapsed/extended a-loop in PI3Ks upon nSH2 release

Kinase domains experience structural rearrangement upon nSH2 release. While N-lobes show minor variation, C-lobes exhibit two distinctive conformations (Figure 2). In PI3K structures containing the nSH2 domain, the a-loop in the kinase domain's C-lobe adopts a collapsed conformation, in which the first basic box (xKxK) is compact, adopting a helical structure and the C-terminal residues form a U-shaped motif (U-motif) (Figure 2(a)). In PI3K structures without the nSH2 domain, the a-loops becomes extended, and the basic boxes lose the helical structure and become flat. The U-motifs at the C-terminal flip (Figure 2(b)). Upon nSH2 release, the kα11 in the regulatory arch undergoes a structural change from an IN to an OUT conformation (Figure 2(c)). This indicates that nSH2 release activates PI3Ks by inducing a transition of the a-loop from a collapsed to an extended conformation.

Figure 2.

Two distinct conformations of PI3K kinase domain upon nSH2 release. (a) For PI3K containing the nSH2 domain, nSH2 restricts the a-loop into a collapsed conformation and kα11 in the IN conformation. (b) For PI3K missing the nSH2 domain, the a-loop becomes extended, and kα11 exhibits the OUT conformation. (c) The structural features of a-loop and kα11 are summarized and compared. The a-loop contains the first basic box (B1), the second basic box (B2), and the U-motif (U). For PI3K crystal structures, the collapsed and extended a-loop conformations are marked by gray and red bars, respectively. The unfilled bars indicate the incomplete density in crystal structures. For kα11, the IN and OUT conformations are denoted by gray and red spheres, respectively. The gray and red labels denote PI3K crystal structures in the presence and absence of nSH2, respectively. The orange labels denote the PI3Kα H1047R mutants with nSH2.

nSH2 has direct contacts with the a-loop through its acidic motif (residues 335–349) (Figure 3(a)). The motif contains consecutive negatively charged residues (D337, E341, E342, and E345) that generate an intense negative surface, which accommodates the basic residues in the a-loop and confines the collapsed a-loop conformations in the inactive state. In most structures, the nSH2 acidic residues interact with the basic residues in the second basic box (KRER) of the a-loop (Figure 3(b)). The multiple binding models and the absence of density in some structures (Figure S6A) suggest that the electrostatic interactions between the nSH2 acidic motif and the basic residues in the a-loop are nonspecific and dynamic. The dynamic interactions are also supported by the nSH2/a-loop interfaces being exposed to the solvent (Figure S6B and C). The a-loop becomes extended when the inhibitory interaction with nSH2 is dismissed (Figure 3(c)).

Figure 3.

Structural insights into the PI3K activation by nSH2 release. (a) Structural comparison of PI3Ks in the presence (gray cartoon) and absence (red cartoon) of nSH2. In the cartoon, the crystal structure of ATP is depicted (PDB: 1E8X). (b) For PI3Ks containing the nSH2 domain, the negatively charged acidic motif of nSH2 (as illustrated by the red surface) interacts with the a-loop through the second basic box (KRER), confining the a-loop into a collapsed conformation. (c) The a-loop becomes extended when PI3K releases the nSH2 domain.

The collapsed/extended structural transitions allow the first basic box (xKxK) in the a-loop to approach the catalytic ATP pocket in the kinase domain (Figure 4(a)). The basic residues in the a-loop mediate substrate specificity [30]. The substrate, PIP₂, has three phosphates, at positions 1, 4 and 5 (Figure S7A). The structures indicate that the lysine residue in the P-loop (K776 in PI3Kα, K776 in PI3Kβ, and K755 in PI3Kδ) is available for coordinating the PIP₂ phosphate, in line with the significance of K776 in PI3Kα catalysis [31] (Figure S7B). The other two basic residues in the kinase domain for PIP₂ binding are more likely to be the lysine residues in the first basic box of the a-loop (Figures 4(b) and S7B). The modeling rationalizes the scenario. The two lysine residues in the first basic box interact with the phosphates at positions 4 and 5 and the lysine residue in the P-loop binds to the phosphate at position 1. The OH group at position 3 points to the γ- phosphate of ATP. The overall structure fits into the membrane binding surface defined by the mutation sites of H1047R and E726K in PI3Kα (Figure 4(c)). In most PI3K structures that lack nSH2, the first basic box of the a-loop lost density, implying its flexibility for accommodating PIP₂ mobility in the membrane (Figure 3(c)). These results support our suggestion that a kinase domain with an extended a-loop may represent a catalytically active conformation of PI3Ks.

Figure 4.

A plausible catalytic scenario for PI3Ks. (a) Structural transition of the a-loop from a collapsed to an extended conformation allows the first basic box (xKxK denoted by blue spheres) to approach ATP in the kinase domain. (b) The basic residues in the first basic box and the lysine residue in the P-loop may accommodate the phosphate groups of PIP₂ substrate for catalysis. The side-chains of lysine residues in the first basic box are rotated to accommodate the PIP₂ substrate (the original positions are shown as transparent sticks). (c) This description well fits into the PI3K membrane binding surface. In the cartoon, the crystal structure of ATP is depicted (PDB: 1E8X), and the substrate of PIP₂ analog is presented (PDB: 4OVV). The catalytic scenario is modeled based on PI3K crystal structure in the absence of nSH2 (PDB: 2Y3A). The positions of H1047R and E726K are modeled based on the PI3Kα mutant structure (PDB: 3HHM).

The IN/OUT regulatory arch in PI3Ks upon nSH2 release

The regulatory arch at the C-terminal kinase domain involves kα10, kα11, and kα12 [28]. In PI3K crystal structures, the structural difference of kα10 is insignificant (Figure 5(a) and (b)). kα12 is ordered in PI3Kβ and disordered in PI3Kα and PI3Kδ. Upon nSH2 release, kα11 presents a notable conformational change. In PI3K structures containing the nSH2 domain, kα11 adopts the IN conformation; upon nSH2 release, it reorients to an OUT conformation (Figure 5(a)). The kα11 structural transition leads to dramatic changes in local hydrophobic interactions. The conserved phenylalanine residues (F954 in PI3Kα, F952 in PI3Kβ, and F932 in PI3Kδ) in the a-loop flip, making the C-terminal U-motif point down, facilitating the extended a-loop conformation in PI3K activation (Figure 5(c) and (d)).

Figure 5.

Structural transition of the regulatory arch in the PI3Kα activation by nSH2 release. The structural transition of regulatory arch upon (a) nSH2 release and (b) H1047R mutation. (c and d) The transition couples with the reorientations of the phenylalanine residues in the U-motif of a-loop (F954 in PI3Kα, F952 in PI3Kβ, F932 in PI3Kδ) and the residues in regulatory arch (M1043 and H1047 in PI3Kα, F1039 and L1043 in PI3Kβ, F1019, and L1023 in PI3Kδ). (e) This provides an explanation for M1043V/I and H1047R oncogenic mutations in PI3Kα.

In PI3Kα, kα11 contains two oncogenic mutations, H1047R and M1043V/I. H1047R is one of the hotspot mutations in PI3Kα. It activates PI3Kα independently of Ras, suggesting its role in enhancing the PI3K membrane interactions [32]. In wild-type PI3Kα, H1047 points inward. The longer side-chain of this mutation incurs steric hindrance, driving R1047 outward, accumulating a positive charge for membrane interaction [33]. Crystal structures of PI3Kα H1047R mutants have been solved (PDB: 3HHM, 3HIZ) [34]. nSH2 was included, which, in principle, should lead to an inactive conformation. However, the presence of the nSH2 domain fails to fully confine the inactive conformation in the two structures (Figure 5(e)). The a-loops are extended, the kα11 points out and the C2-kinase domain distances are larger than in other inactive PI3K structures with nSH2 domain (Figures 1(a) and 2(c)). This implies that in addition to enhancing membrane localization, H1047R may shift the PI3K ensemble toward an active conformation for catalysis by affecting the regulatory arch.

The structural transition of kα11 upon activation by nSH2 release couples with the movement of M1043 (F1039 in PI3Kβ and F1019 in PI3Kδ), explaining the role of the M1043V/I oncogenic mutation in promoting PI3Kα activation. In the inactive state, M1043 points to the U-motif, occupying the space that accommodates the phenylalanine residues (F954 in PI3Kα, F952 in PI3Kβ, and F932 in PI3Kδ) in the active state (Figure 5C). In PI3K structures with nSH2 released or oncogenic mutations, M1043 (F1039 in PI3Kβ and F1019 in PI3Kδ) moves out, which invokes a hydrophobic pocket for the phenylalanine residues in the extended a-loop (Figure 5(c)–(e)). When M1043 is mutated to valine/isoleucine, the steric hindrance would be smaller, increasing the population of the extended a-loop for catalysis.

Discussion

PI3Ks are lipid kinases with substantial roles in PI3K/Akt signaling pathways [3]. Their over-activation and oncogenic mutations are frequent in cancer [35]. PI3Ks phosphorylate PIP₂ to PIP₃, delivering downstream signals [36]. The interactions of the p85 regulatory subunit stabilize and inhibit the basal activity of the p110 catalytic subunit. The release of the nSH2 domain in p85 from p110 activates PI3Ks [37]. Here, we analyzed PI3K crystal structures and found that nSH2 mediates PI3K inhibition, activation, and catalysis via direct control of the a-loop dynamics (Figure 6). In response to nSH2 release, the a-loop in the C-lobe kinase domain adopts two distinct conformations: collapsed and extended that correspond, respectively, to the inactive and active states. The collapsed a-loop in inactive PI3Ks is confined by the negative surface of the acidic motif of the nSH2 domain and is coupled with an IN kα11 conformation. Upon nSH2 release, the active PI3K structures have an extended a-loop and an OUT kα11 conformation. The structural transition of kα11 upon activation by nSH2 release suggests a mimic scenario of oncogenic mutations (M1043V/I and H1047R) in promoting the catalytically active conformation of PI3Ks. This scenario gains consistent support by PI3K crystal structures.

Figure 6.

Schematic illustration of the PI3K activation by nSH2 release.

The crystal resolutions for the inactive (2.20 to 3.55 Å) and active (2.21 to 3.30 Å) PI3Ks are comparable, supporting the conformational change upon activation (Table S2). As an example, the inactive (5UL1, collapsed a-loop with full density) and active (5DXH, extended a-loop with partial density) PI3Kα structures with identical crystal resolution (3.0 Å) present the a-loops with distinctive densities and conformations. The crystal packings are highly diverse in PI3K structures. The PI3K structures show two conformations (inactive versus active) with the diverse crystal packings. In many inactive and active PI3K structures (taking inactive (4L1B) PI3Kα conformation and active (5DXH) PI3Kα conformation as the examples in Figure S8), the key activation components in the kinase domain C-lobe, a-loop and kα11, barely involve in the crystal packing, which implies that the role of crystal packing in modulating the PI3K conformational change is not major.

The structural insights proposed in this work are in line with experimental data. PI3K crystal structures suggest that nSH2 release triggers an overall structural rearrangement for activation. iSH2 in the p85 regulatory subunit moves away from the kinase domain along with the C2 domain, releasing the basic residues in the a-loop for enhancing the electrostatic interactions with the membrane. The enhanced interactions of PI3Ks with the membrane, particularly lipid bilayers containing anionic lipids including PIP₂, have been indicated in the experiments [23,29]. In the hydrogen/deuterium exchange mass spectrometry study, H1047R mutation decreased the protection of the a-loop. With the nSH2 release by the platelet-derived growth factor receptor peptide, the activation by H1047R mutation became more profound, consistent with the activation mechanism [24]. The nSH2 release by the platelet-derived growth factor receptor peptide increased the a-loop signals, but not significantly, which is possibly because the full release of nSH2 by the peptide may require the membrane [24]. In both inactive and active conformations, the a-loop is not fully buried. The inhibitory interactions between the nSH2 and the a-loop in the inactive PI3K conformation are formed through solvent-exposed electrostatic forces, which are dynamic and nonspecific.

PI3Ks have basal activity. The interactions of the inhibitory nSH2 with the p110 catalytic subunit are meta-stable, in line with the principles of autoinhibition [38,39]. The crystal structures also suggest that the H1047R hotspot mutation in the kinase domain may allow the a-loop to bypass the nSH2 inhibition for a catalytically active conformation by inducing the OUT kα11 conformation. In the inactive PI3Kα, H1047 is surrounded by the hydrophobic residues. Its fluctuation in the ensemble may facilitate the OUT kα11 conformation in basal activity. This mechanism is shared by class IA PI3K isoforms. The key structural components in their scenarios are conserved in PI3Kα, PI3Kβ, and PI3Kδ, and across species. The extended a-loop, the OUT kα11 conformation, and the iSH2 movements are identified in human PI3Kα, mouse PI3Kβ, and human PI3Kδ structures. Class IB PI3K (PI3Kγ) has a similar p110 catalytic subunit; but whether the mechanism also applies to class IB PI3K needs to be further explored.

PI3Kα contains oncogenic mutations at the ABD/ kinase domain interface, ABD–RBD linker and C2/ iSH2 interfaces [11,40]. Conformational changes in these regions in PI3K activation by nSH2 release have been indicated in experiments, in line with the roles of oncogenicmutations in PI3K activation [24]. Structural changes in these regions are observed in some PI3K crystal structures (Figure S2). We expect that conformational changes observed in the PI3K crystal structures may be more profound in the solvent/cellular environments.

The mechanism proposed here outlines novel PI3K allosteric inhibitor design strategy. Most PI3K inhibitors/drugs target the ATP pocket in the kinase domain, competing with the high affinity ATP and frequently lead to toxicity. The mechanism suggests that iSH2 movement is critical in the active conformation and membrane interactions for activation. The role of iSH2 dynamics in PI3K activation has also been implicated by the large number of oncogenic mutations in C2 and ABD. The iSH2 domain is loaded by C2 and ABD that are far from the catalytic site in the kinase domain. The modulation of iSH2 movement could be the guiding principle for PI3K allosteric inhibitor design. Inhibitors stabilizing the C2 and ABD may reduce the iSH2 movement for PI3K activation and allosterically tamper PI3K activity. The pockets at the ABD/kinase domain interface (involving R115–M123) [41], and the C2/helical domain interface (involving V461–S464) may be the potential targeting sites.

Conclusions

In this work, we analyze PI3K crystal structures, focusing on the roles of nSH2 in PI3K inhibition and activation and the structural determinants of the inactive and active PI3K conformations. We observe that inactive PI3Ks have a collapsed a-loop and an IN kα11 conformation. Upon activation by nSH2 release, the a-loop becomes extended and the kα11 rotates out to facilitate the structural transition of the a-loop for catalysis. nSH2 regulates the a-loop dynamics via an acidic motif. Those actions lead us to propose an allosteric PI3K drug discovery strategy. They also suggest a novel mimic mechanism of how M1043V/I and H1047R oncogenic mutations activate PI3K.

The structural features that distinguish between the active and inactive states of protein kinases are well established; however, that has not been the case for lipid kinases. Our analysis of the available PI3K crystal structures of different isoforms from different classes and species, provides this important characterization and the role of the nSH2 in switching between them, which is mimicked by oncogenic mutations. The definition of the key features provided here permits detailed grasp of the activation, catalysis, and autoinhibition of lipid kinases whose signaling roles are vastly important in transducing external incoming cues at the membrane downstream to transcription factors and the cell cycle via small signaling lipids.

Methods and Materials

The coordinates of the PI3K crystal structures were obtained from the PDB. The sequence alignment was performed based on the human PI3Kα (PIK3CA), human PI3Kβ (PIK3CB), and human PI3Kδ (PIK3CD) by Clustal Omega (version 1.2.4) multiple sequence alignment algorithm [42]. The sequence/structure dual convergence was evaluated to ensure the fair comparisons among the class IA PI3K isoforms. In the calculations of the domain–domain distances in PI3K structures, the residues (Cα) in the domain were selected only if they can be aligned in the sequence alignment protocol and present clear density in all dimeric PI3Kα, PI3Kβ, and PI3Kδ crystal structures. The crystal structure of PI3Kβ (PDB: 2Y3A) was from mouse, which lacked six residues at the N-terminal compared to the human sequence. Its sequence was aligned to the human sequence. Table S3 listed the residues selected for the calculation of the domain–domain distances in three class IA PI3K isoforms.

PI3K structures containing the nSH2 domain

There are 39 PI3K crystal structures containing nSH2 (PDB: 2RD0, 3HHM, 3HIZ, 4A55, 4JPS, 4L1B, 4L23, 4L2Y, 4OVU, 4OVV, 4WAF, 4YKN, 4ZOP, 5FI4, 5ITD, 5SW8, 5SWG, 5SWO, 5SWP, 5SWR, 5SWT, 5SX8, 5SX9, 5SXA, 5SXB, 5SXC, 5SXD, 5SXE, 5SXF, 5SXI, 5SXJ, 5SXK, 5UK8, 5UKJ, 5UL1, 5XGH, 5XGI, 5XGJ, 6NCT). Most of those crystal structures (except 3HHM and 3HIZ for PI3Kα mutants) are featured by collapsed a-loop and IN kα11, indicating an inactive conformation. In two structures (PDB: 2RD0, 4A55), the nSH2 domain was included in the crystallization but lacked the full density, indicating the flexibility of the nSH2. Structural comparisons showed that despite the flexibility, the included nSH2 confined an inactive conformation. The overall structures of 2RD0 and 4A55 were similar to other PI3K structures with nSH2, with the collapsed a-loop and IN kα11. In 4A55, the C-terminal residues (U-motif) in the a-loop was slightly different due to the bound inhibitors nearby (not shown in the aligned clusters). 3HHM and 3HIZ were of PI3K oncogenic mutants, in which the H1047R hotspot mutation in the kinase domain made the a-loop escape from the nSH2 inhibition and achieve a catalytically active conformation.

PI3K structures missing the nSH2 domain

There are 10 PI3K crystal structures in the absence of nSH2 (5DXH for PI3Kα; 2Y3A for PI3Kβ; and 5DXU 5M6U 5T8F 5UBT 5VLR 6G6W 6PYR 6PYU for PI3Kδ). These structures exhibited an extended a-loop and OUT kα11 conformation, indicating an active conformation. In the PI3Kβ crystal structure, cSH2 was included and bound to the kinase domain. It blocked the membrane binding surface in the kinase domain, which may affect the membrane localization of PI3Kβ.

Abbreviations used:

PI3K

phosphatidylinositol 3-kinase

RTK

receptor tyrosine kinase

RBD

Ras binding domain

PDB

Protein Data Bank

ABD

adaptor-binding domain.