Report

Mechanism of Two Classes of Cancer Mutations in the Phosphoinositide 3-Kinase Catalytic Subunit

See allHide authors and affiliations

Science  13 Jul 2007:
Vol. 317, Issue 5835, pp. 239-242
DOI: 10.1126/science.1135394

Abstract

Many human cancers involve up-regulation of the phosphoinositide 3-kinase PI3Kα, with oncogenic mutations identified in both the p110α catalytic and the p85α regulatory subunits. We used crystallographic and biochemical approaches to gain insight into activating mutations in two noncatalytic p110α domains—the adaptor-binding and the helical domains. A structure of the adaptor-binding domain of p110α in a complex with the p85α inter–Src homology 2 (inter-SH2) domain shows that oncogenic mutations in the adaptor-binding domain are not at the inter-SH2 interface but in a polar surface patch that is a plausible docking site for other domains in the holo p110/p85 complex. We also examined helical domain mutations and found that the Glu545 to Lys545 (E545K) oncogenic mutant disrupts an inhibitory charge-charge interaction with the p85 N-terminal SH2 domain. These studies extend our understanding of the architecture of PI3Ks and provide insight into how two classes of mutations that cause a gain in function can lead to cancer.

Phosphoinositide 3-kinases and their lipid product, phosphatidylinositol-(3,4,5)-trisphosphate [PtdIns(3,4,5)P3], play key roles in a variety of cellular processes (13). Aberrations in PtdIns(3,4,5)P3 levels, either through activation of PI3Ks or through inactivation of lipid phosphatase PTEN, occur frequently in numerous forms of cancers. For example, recent data suggest that at least 50% of human breast cancers involve mutations in either PI3Kα or PTEN (4, 5). Broad-spectrum PI3K inhibitors such as LY294002 or wortmannin result in increased apoptosis, decreased proliferation, and reduced metastasis in vitro and in tumor models [reviewed in (68)]. Understanding the structural mechanisms of PI3K regulation may facilitate development of isozyme-specific therapeutics.

The class IA PI3Ks are obligate heterodimers (9), consisting of a p110 catalytic subunit and a regulatory subunit. Any of the three class IA catalytic subunits (p110α, p110β, and p110δ) can bind any of the p85-related regulatory subunits. Regulatory subunits have multiple roles in the function of PI3K: down-regulation of the basal activity, stabilization of the catalytic subunit, activation downstream of receptor tyrosine kinases, and sequential activation by tyrosine kinases and Ras (1013). Common to all regulatory subunits are two SH2 domains (nSH2 and cSH2) that flank an intervening domain (iSH2), and common to all catalytic subunits are the N-terminal adaptor-binding domain (ABD), the Ras-binding domain (RBD), the putative membrane-binding domain (C2), and the helical and catalytic domains (Fig. 1A). The iSH2 domain is responsible for tight binding to the ABD (14). The nSH2 and cSH2 domains bind phosphorylated tyrosines in Tyr-X-X-Met motifs found in activated receptors and adaptor proteins, and this interaction activates the heterodimeric PI3K. The nSH2-iSH2 unit constitutes the minimal fragment capable of regulating the PI3K activity (15): It both inhibits the basal activity and facilitates activation by binding phosphotyrosine peptides. In contrast, the isolated iSH2 only minimally affects the PI3K activity, although it tightly binds the p110 subunit.

Fig. 1.

Structure of a class IA PI3K heterodimerization core. (A) Domain organization of PI3K catalytic (classes IA and IB) and regulatory (class IA only) subunits. N and C, N- and C-terminal lobes of the kinase domain, respectively; GAP, Rho-GAP domain. (B) Ribbon representation of ABD/iSH2 heterodimer. The three ABD residues identified as somatic mutations in colon cancers (shown in spheres) are not in the ABD/iSH2 interface. (C) The ABD has a ubiquitin-like fold. (D) A cross-section through the ABD/iSH2 interface showing the surface complementarity. (E) Underside view of the heterodimer interface showing residues that contribute substantially to the binding. Cα atoms of residues forming prominent interactions with the iSH2 (contributing more than nine interatomic contacts, interatomic distance < 3.8 Å) are shown as large spheres. Smaller spheres represent residues involved in less extensive interactions (four to six interatomic contacts).

A number of studies have identified a high frequency of somatic point mutations in the gene encoding the p110α catalytic subunit in different human cancers (16, 17). An increased lipid kinase activity in vitro and the ability to induce oncogenic transformation in vivo were shown both for the most frequently mutated, “hotspot” residues (16, 1820) and for 14 r are cancer-specific mutations in p110α (21). The cancer-specific mutations can be grouped into four classes defined by the four domains of the catalytic subunit in which they occur—the ABD, C2, helical, and catalytic domains—and it has been proposed that these classes may increase PI3K activity by different mechanisms (21, 22). Hotspot mutations in the catalytic domain cluster around the “activation loop” involved in substrate recognition (23) and are likely to share a common mechanism (21). The ABD binds tightly to the regulatory subunit, the C2 domain is thought to interact with the plasma membrane, and the helical domain appears to act as a rigid scaffold around which the RBD, C2, and catalytic domains are mounted (24). The catalytic and C2 domain mutations may up-regulate PI3K by increasing the affinity for substrate-containing membranes. However, it is not immediately clear what might be the mechanism of helical domain mutations. We used structural and biochemical approaches to understand the basis for gain-of-function mutations in the ABD and helical domains of p110α. Because the ABD is the only p110 domain for which there is no known structure, we crystallized it in a complex with the iSH2 domain from p85. This structure suggested a rough preliminary model for the p110/p85 heterodimer, which led us to hypothesize that the nSH2 domain might contact the helical domain. To understand how helical domain oncogenic mutations function, we created a series of site-specific mutations in the nSH2 domain, resulting in an adaptor subunit that specifically counteracts the p110 helical domain hotspot E545K mutant.

We determined the crystal structure of a complex between the bovine p110α ABD (residues 1 to 108) and the human p85α iSH2 domain (residues 431 to 600) at 2.4 Å resolution (25), revealing the interaction of the small, globular ABD (35 by 25 by 15 Å) with one end of the long, rodlike iSH2 (115 Å long) (Fig. 1B). The ABD is similar to many ubiquitin-like domains: It superimposes on ubiquitin with a root mean square deviation of 1.4 Å for 60 out of 76 residues (Fig. 1C). The β-grasp fold of both the ABD and ubiquitin is a common fold, and neither sequence nor function suggests a common origin for the ABD and ubiquitin. The ABD ranks as one of the least conserved domains among class I catalytic subunits. Conversely, the iSH2 domain is highly conserved in vertebrates, sharing >90% sequence similarity from human to zebrafish (fig. S1). Despite a lack of sequence similarity, secondary and tertiary structure predictions for the class IB p110γ ABD are consistent with a ubiquitin-like fold, similar to the class IA p110 ABD. However, the greater sequence divergence of the p110γ ABD makes it unable to bind the iSH2 domain. Consistent with previous studies (26), the p85α iSH2 domain consists of two helices, α1 and α2, that form a long, antiparallel coiled coil followed by the short α3 helix (Fig. 1B).

The ABD/iSH2 interface is large, burying 2284 Å2 surface area, with the concave face of the ABD β sheet interacting with the coiled-coil helices of the iSH2 (Fig. 1, B and E). Most of the interactions with the ABD are formed by the iSH2 helix α1, which contacts the ABD with seven of its turns, whereas only three turns of iSH2 helix α2 interact with the ABD. Central to the ABD/iSH2 interface is the surface encompassing the strands β1 to β2 of the ABD, which contains the conserved 25-Leu-Leu-Pro-X-Gly-ϕϕ-31 motif, where ϕ denotes a hydrophobic residue (Fig. 1, D and E). This motif forms a loop that through both its side chains and backbone contacts two conserved iSH2 residues, Gln497 and Asn527, one on each long helix of the domain. Among all residues at the interface, the side chains of these two polar iSH2 residues engage in the greatest number of intersubunit contacts. Overall, about three-quarters of the ABD/iSH2 interactions are van der Waals contacts, and most of them involve exposed hydrophobic side chains (fig. S1). At the periphery of the interface are two salt links, Glu23ABD/Arg534iSH2 and Arg79ABD/Glu496iSH2; however, site-specific mutagenesis of the individual charged residues to Ala did not eliminate binding in vitro. This is consistent with the very high affinity of the iSH2 domain for the ABD (14, 27). Other iSH2-contacting sites include the α1/β3, the β3/β4, and the β4/β5 loops of ABD.

Among rare cancer-specific mutations in p110, there are three residues that map to the ABD—Arg38, Arg88, and Pro104. Recently, it was shown the R38H (28) mutation induces oncogenic transformation of avian cells, although with weak efficiency (21). None of these residues are at the interface between the ABD and the iSH2 domain. Analysis with ConSurf (29) shows that there is a highly conserved ABD surface patch outside the conserved ABD/iSH2 interface (Fig. 2A). This patch, formed by the β4/h3 region and the β2/α1 loop in the ABD (Fig. 2 and fig. S1), consists predominately of polar residues, including two of the three ABD residues whose mutation is linked with cancer, Arg38 and Arg88. It represents a plausible docking site for other domains in the rest of the p110/p85 complex. Mutations in this surface may affect the relationship of the ABD with respect to the catalytic core. This could reorient the adaptor subunit with respect to the p110 subunit or change the orientation of the enzyme on the membrane.

Fig. 2.

A model of the minimal p110/p85 complex, showing the locations of rare cancer-specific mutations in the ABD. (A) Two ABD residues that are mutated in colon cancers map to a conserved, non–iSH2-binding surface patch (purple). (B) A view of the ABD/iSH2 complex rotated 90° around a vertical axis, showing conserved residues on the ABD surface and cancer-specific mutation sites (boxed).

A minimal regulatory p85 fragment consisting of nSH2 and iSH2 (p85ni) readily inhibits activity of the wild-type p110α (Fig. 3A), suggesting an undefined inhibitory p85-p110 contact (30). Hypothesizing that the helical domain might be a good candidate for the inhibitory nSH2 binding, we looked at the effect of oncogenic mutants in the helical domain. We found that the p85ni regulatory fragment does not inhibit p110α helical domain oncogenic mutants E542K, E545K, or Q546K in vitro (Fig. 3A). Because all of these helical domain mutations increase the positive charge on the surface of the helical domain, we speculated that there might be an inhibitory interaction between the nSH2 and the wild-type helical domain involving a charge-charge interaction between an acidic patch on the helical domain and a basic residue on the surface of the nSH2 domain. We singly mutated each of the lysine or arginine residues on the surface of the nSH2 domain into glutamate. Among these mutants, only the p85ni-R340E and p85ni-K379E mutants did not inhibit the wild-type p110α (Fig. 3B). The loss of inhibition for R340E and K379E suggests that these residues are involved in an inhibitory contact with the catalytic subunit. Consistent with this, the p85ni-K379E mutant inhibits the p110α-E545K oncogenic mutant, whereas the wildtype p85ni does not, suggesting that the charge reversal has reestablished the critical inhibitory interaction (Fig. 3C).

Fig. 3.

Helical domain oncogenic mutants of p110α eliminate an autoinhibitory contact with the p85 nSH2 domain. (A) PI3K activity of the three p110α helical domain mutants was not inhibited by the p85ni fragment. Equal amounts of protein were assayed and the catalytic subunits had similar specific activities (fig. S3) (25). WT, wild type. (B) Charge-reversal mutagenesis screening of basic surface residues in the nSH2. (C) Simultaneous charge-reversal in p110α and p85ni restored inhibition, suggesting direct interaction between the p110α helical domain and the p85α nSH2 domain. A titration series was carried out for both the wild-type (left) and E545K mutant (right) p110 subunits to ascertain that the inhibition was at saturating levels of p85ni. (D) Binding of IRS-1 phosphotyrosine peptide activates wild-type p110α/p85ni but not p110-E545K/p85ni (25). [(A) to (D)] Lipid kinase assays were performed with Myc-p110α produced in transfected human embryonic kidney 293 cells (A) or baculovirus-infected Sf9 cells [(B) to (D)], incubated with or without excess p85ni [expressed in Escherichia coli (25)]. Basal activity refers to p110α activity in the absence of p85ni. Results are means from three to four experiments [(A) and (B)] or are representative of two experiments, each carried out in triplicate [(C) and (D)].

The p85 residues Arg340 and Lys379 are part of the phosphopeptide-binding surface on the nSH2 domain. Both are in direct contact with bound phosphopeptide in crystal structures of two nSH2/phosphotyrosine-peptide complexes (31), suggesting that the phosphopeptide activates the enzyme by competing with the inhibitory contact. A tandem Tyr-X-X-Met phosphopeptide from the Tyr608/Tyr628 region of insulin receptor substrate 1 (IRS-1) roughly doubled the activation of the wild-type p110α/p85ni heterodimer but did not activate the p110α-E545K/p85ni complex (Fig. 3D). Figure 4 suggests how binding of the nSH2 domain to the p110α helical domain and to the phosphopeptide would be mutually exclusive. Currently, it is not clear how the nSH2 inhibits the wild-type enzyme. The helical domain interacts with the catalytic domain across a broad surface. The contact of the nSH2 domain with the helical domain might trigger a conformational change in the catalytic domain to affect the activity allosterically, or it may influence the orientation of the catalytic domain on the substrate-containing membrane.

Fig. 4.

A model showing the inhibitory contact between the nSH2 domain and the helical domain of p110α near the site of the helical domain hotspot mutations. [The p110α catalytic core was modeled on the p110γ catalytic core (24).] Arg340 and Lys379 are part of a highly conserved phosphopeptide-binding surface on nSH2. Binding of nSH2 to the p110α helical domain and to phosphopeptide are mutually exclusive (boxed image).

The ABD/iSH2 crystal structure completes the architectural account of the whole p110 catalytic subunit and shows that oncogenic mutations in the ABD are not in the interface with the p85 regulatory subunit. Our results also suggest an unexpected function for the helical domain. A rationally designed nSH2 mutant counteracts the gain-of-function oncogenic p110α-E545K mutation, supporting the proposal that the p85 inhibition of p110α occurs through a charge-charge interaction between the helical domain and the p85ni. The crystallographic structures combined with the mutational analyses have enabled us to propose a rough partial model for the p110/p85 complex (25), but ongoing studies by nuclear magnetic resonance, electron paramagnetic resonance, and crystallography will be required to refine any such model.

Supporting Online Material

www.sciencemag.org/cgi/content/full/317/5835/239/DC1

Materials and Methods

Figs. S1 to S5

Table S1

References

References and Notes

View Abstract

Navigate This Article