We propose three extrinsic “brakes” on the basal activity of p110β exerted by the nSH2, cSH2, and iSH2 domains of the regulatory subunit ( and Movie S1
). The nSH2 interacts with the C2, helical, and the kinase domains of p110 and provides one brake. The second brake is the contact of the cSH2 with the C lobe of the kinase domain. This inhibitory cSH2 interaction distinguishes p110β from p110α (Yu et al., 1998
), and our structure suggests a possible explanation for this difference. The third brake is provided by the iSH2, which nestles under the arch formed by the catalytic subunit and forms inhibitory contacts with the C2 domain (Wu et al., 2009
). Its mutations in several types of tumors lead to dramatic upregulation of all class IA isozymes. While p110β activity is inhibited by the engagement of three brakes, p110α appears to have only two brakes (nSH2 and iSH2), suggesting that p110α is poised to have greater activity when one of the brakes is lost by mutation. The inhibition of p110β by both SH2 domains of the p85 subunit may explain observations that p110β is less responsive to RTK stimulation in some cells (Kurosu et al., 1997; Maier et al., 1999; Guillermet-Guibert et al., 2008
A “Three-Brake” Regulatory Model for Inhibition of p110β Basal Activity
The inhibitory effect of the cSH2 could be explained by its contacts with the C-terminal region of the kinase domain. In our crystal structure, the kinase domain of the catalytic subunit exhibits the signatures of an inactive conformation. The contacts of the cSH2 with the regulatory elements surrounding the activation and catalytic loops, i.e., helices Kα11/Kα12 and Kα7/Kα8 forming the double-layer regulatory arm, could possibly pin the catalytic and activation loops into an inactive conformation and prevent the C-terminal helix from swinging out to its active conformation. By affecting the C-terminal helix conformation, the cSH2 may also affect membrane binding. In the primordial class III PI3K Vps34, the equivalent C-terminal helix interacts with membranes and its deletion abolishes enzyme activity (Miller et al., 2010
). Similarly, we show here that the C terminus in p110β is critical for lipid kinase activity.
Interestingly, the primary interaction site of p110β with the cSH2 is part of a regulatory “square” composed of helices Kα10, Kα11, and Kα12, which were described as three sides of an imaginary rectangle (Lempiäinen and Halazonetis, 2009
). While the cSH2 contacts Kα11 and Kα12, the nSH2 interacts with Kα10. As this square envelops the key loops for catalysis, any changes in the conformation of the square could possibly lead to allosteric regulation of the active site, either changing the affinity for lipid substrate or facilitating phosphoryl transfer. In fact, somatic mutations in p110α in human tumors cluster on the perimeter of the square, suggesting the importance of this square in the regulation of the enzyme (Figure S5
). Other proteins may also regulate PI3Ks through interactions with the regulatory square (Figure S6
The inhibition of p110β by the cSH2 can be released upon binding to RTK phosphopeptides. However, the way phosphopeptide breaks the contact between p110 and the SH2 domains is distinct for the nSH2 and cSH2 (Movie S2
). The pY-binding site on the nSH2 is buried in the interface with the catalytic subunit, thus pY binding breaks the contact by direct competition. In contrast, the pY-binding site on the cSH2 is exposed, and displacement of the cSH2 from the catalytic subunit requires five or more residues C-terminal to the pY. This means that various tyrosine-phosphorylated receptor kinases and adaptor proteins upstream of PI3Ks could have different potencies to displace the cSH2 inhibitory contact.
We find that the activity of p110β/p85-niSH2 and p110β/p85-icSH2 complexes with soluble lipid substrate as well as with water (ATPase activity in the absence of a lipid substrate) is enhanced in the presence of pY2. This suggests that neither the nSH2 nor the cSH2 exerts its inhibition exclusively by affecting membrane binding. At least part of the activation is due to conformational changes in the catalytic elements caused by the dislodging of the SH2 domains from the catalytic subunit upon binding to phosphopeptide. Nevertheless, the highest level of activation is observed with membrane substrates, suggesting changes in membrane binding could also contribute to maximal activation.
Overexpression of wild-type p110β causes oncogenic transformation of cells, whereas wild-type p110α does not (Kang et al., 2006
). There is more than one structural feature of the catalytic subunit that could contribute to intrinsic oncogenic potential. One of them is the presence of Leu1043 in the C terminus of p110β (and p110δ), equivalent to His1047 in p110α. We have shown that the L1043H-p110β mutation decreases enzyme activity in the presence of RTK phosphopeptide. Conversely, naturally occurring mutations of this residue to a leucine in p110α display increased activity and are highly oncogenic. It seems that p110α has an intrinsic inhibitory mechanism involving His1047 at the elbow region, whereas p110β and p110δ need an extrinsic brake, provided by the cSH2, to inhibit their activity. Other structural features, such as the more basic character of the putative membrane-binding CBR loops in the C2 domain of p110β and p110δ versus p110α, could also affect oncogenic potentials. Indeed, mutations of basic residues in CBR3 of p110δ (Denley et al., 2008
) or mutation of a basic residue in CBR1 of p110β (Dbouk et al., 2010
) dramatically reduced their transformation potency. The iSH2 inhibitory interaction with the C2 domain can also provide a differential brake on p110 activity. It was recently shown that the iSH2-C2-mediated inhibition is more pronounced in p110α than in p110β (Dbouk et al., 2010
). This constitutes an SH2-independent brake on p110 activity.
In addition to its lipid kinase activity, PI3Ks also exhibit kinase activity toward peptides (Bondeva et al., 1998
) and water. The ability to bind different substrates suggests plasticity in the active site, which is reflected by the weak density of the activation loop (i.e., substrate-binding loop) in our structure. Thus, we expect that the conformation of the activation loop observed in our structure would be changed upon substrate binding.
Besides regulation by interaction with RTKs, the activity of the p110 catalytic subunit could also be modulated by the posttranslational modifications of the regulatory subunits. When we map the reported p85 modification sites on the structure, it is clear that many of the modified residues are at the interface with the catalytic subunit and could thereby influence the p110 activity (Figure S7
). There are several phosphorylation sites in the cSH2. One of them is the Tyr688 in p85α, whose phosphorylation by the Src family kinases leads to upregulation of PI3K activity, and this activation is reversed by dephosphorylation of Tyr688 by Shp1 (Cuevas et al., 2001; Chan et al., 2002
). Our structure shows that this tyrosine is close to the cSH2/catalytic subunit interface and suggests that its phosphorylation could disinhibit p110β.
Our work illuminates unexpected aspects of p110β inhibition by the p85 regulatory subunit and activation by RTKs. The p85 SH2 domains can contribute to the well-known isoform-dependent specificities in PI3K signaling, despite being shared by all class IA catalytic subunits. Further regulatory complexity that needs to be unraveled is the mechanism of unique regulation of p110β by Gβγ heterodimers.