PMCCPMCCPMCC

Search tips
Search criteria 

Advanced

 
Logo of elifeeLifeRecent contentAbout eLifeFor authorsSign up for alerts
 
eLife. 2017; 6: e26640.
PMCID: PMC5550278

Structural basis for interdomain communication in SHIP2 providing high phosphatase activity

Roger L Williams, Reviewing editor
Roger L Williams, Medical Research Council, United Kingdom;

Abstract

SH2-containing-inositol-5-phosphatases (SHIPs) dephosphorylate the 5-phosphate of phosphatidylinositol-3,4,5-trisphosphate (PI(3,4,5)P3) and play important roles in regulating the PI3K/Akt pathway in physiology and disease. Aiming to uncover interdomain regulatory mechanisms in SHIP2, we determined crystal structures containing the 5-phosphatase and a proximal region adopting a C2 fold. This reveals an extensive interface between the two domains, which results in significant structural changes in the phosphatase domain. Both the phosphatase and C2 domains bind phosphatidylserine lipids, which likely helps to position the active site towards its substrate. Although located distant to the active site, the C2 domain greatly enhances catalytic turnover. Employing molecular dynamics, mutagenesis and cell biology, we identify two distinct allosteric signaling pathways, emanating from hydrophobic or polar interdomain interactions, differentially affecting lipid chain or headgroup moieties of PI(3,4,5)P3. Together, this study reveals details of multilayered C2-mediated effects important for SHIP2 activity and points towards interesting new possibilities for therapeutic interventions.

DOI: http://dx.doi.org/10.7554/eLife.26640.001

Research Organism: E. coli, Human

Introduction

The levels of soluble inositol phosphates and membrane integrated phosphoinositide lipids in the cell are regulated by inositol kinases and phosphatases. They play key roles in a variety of cellular processes, including cell proliferation, survival and vesicular trafficking. The generation of the phosphoinositide phosphatidylinositol 3,4,5-trisphosphate (PI(3,4,5)P3) is triggered via stimulation of growth factor and cytokine receptors that activate type I phosphoinositide 3-kinase (PI3K) to produce PI(3,4,5)P3 from PI(4,5)P2. Upon stimulation, PI(3,4,5)P3 levels peak rapidly followed by a prompt return to basal levels (Stephens et al., 1993; Yip et al., 2008). The two main enzymes responsible for PI(3,4,5)P3 degradation are the tumor suppressor phosphatase and tensin homolog (PTEN), which removes the 3-phosphate (3 P) and the SH2-containing inositol 5-phosphatase (SHIP), which dephosphorylates the 5 P to generate PI(3,4)P2. Deregulation of PI(3,4,5)P3 levels is one of the most frequent causes of tumorigenesis (Song et al., 2012; Vanhaesebroeck et al., 2012).

The SHIP 5-phosphatase (5-Ptase) family consists of two members, SHIP1 and SHIP2. SHIP1 expression is limited to hematopoietic cells and spermatocytes, whereas SHIP2 is ubiquitously expressed (Liu et al., 1998; Pesesse et al., 1997). SHIP1 plays an important role in myeloid homeostasis and exhibits reduced levels or mutational inactivation in various leukemias and lymphomas, suggesting a role as tumor suppressor by negatively regulating the PI3K/Akt pathway (Brauer et al., 2012; Lo et al., 2009; Luo et al., 2003). SHIP2 polymorphisms associate with susceptibility to type 2 diabetes mellitus and hypertension (Hao et al., 2015; Kagawa et al., 2005; Kaisaki et al., 2004) as well as opsismodysplasia (Below et al., 2013; Huber et al., 2013; Iida et al., 2013), a rare but severe type of skeletal dysplasia. SHIP2 knockout mice were reported to be resistant to dietary obesity (Sleeman et al., 2005) and mice expressing catalytically inactive SHIP2 displayed several developmental defects and lower insulin secretion (Dubois et al., 2012). Several studies associate SHIP2 with signaling triggered by growth factors, including EGF and FGF (Jurynec and Grunwald, 2010; Olsen et al., 2006). Like SHIP1, SHIP2 also plays important roles in cancer, although its behavior is more complex and tissue dependent. Similar to SHIP1, SHIP2 functions as a tumor suppressor in glioblastoma, erythroleukemia and squamous cell carcinoma (Giuriato et al., 2002; Taylor et al., 2000; Yu et al., 2008), but is oncogenic in breast carcinomas and colorectal cancer (Hoekstra et al., 2016; Prasad et al., 2008). On the one hand, this ambivalent oncogenic behavior of SHIP2 in different cell types likely depends on how efficient the SHIP2 product, PI(3,4)P2, is further metabolized to PI(3)P, a reaction catalyzed by inositol polyphosphate 4-phophatases (INPP4). Accumulation of PI(3,4)P2 in a case where the INPP4 reaction is rate limiting can be oncogenic, since PI(3,4)P2 can still partially activate the Akt pathway (Li and Marshall, 2015), but is no longer accessible for degradation by PTEN. On the other hand, the SHIP product can be directly tumor suppressing by shutting down mTORC1 signaling (Marat et al., 2017).

SHIP1 and SHIP2 are large ~140 kDa multidomain proteins that share a very similar domain organization (Figure 1A). The SHIP Ptase domain belongs to a family of Mg2+-dependent inositol phosphatases with specificity for 5 P dephosphorylation of inositol rings. These 5-Ptases have different substrate recognition specificities (Jefferson and Majerus, 1996) with the SHIP 5-Ptase exhibiting a strong preference for 3-phosphorylated substrates (Chi et al., 2004; Pesesse et al., 1998; Trésaugues et al., 2014). Further, the inositol 5-Ptase family displays distant homology to Mg2+-dependent DNase I and Apurinic/Apyrimidinic (AP) endonucleases (Dlakić, 2000; Whisstock et al., 2000) and the conservation of catalytic residues suggests a common mechanism of catalysis (Whisstock et al., 2002). The existence of a C2 domain following the Ptase domain was previously predicted for SHIP1 (Ong et al., 2007), however, the sequence identity to other C2 domains is low and a C2 fold has not been structurally confirmed. The SHIP1 C2 domain was reported to allosterically upregulate Ptase activity by binding to the PI(3,4)P2 product, a feature not found to be conserved in SHIP2.

Figure 1.
Structure of SHIP2 Ptase-C2.

Here, we report the crystal structure of SHIP2 containing the 5-Ptase and a C2 domain (Ptase-C2), revealing that the two domains tightly associate via an extensive interface. We show that, both the Ptase and Ptase-C2 regions of SHIP2 bind phosphatidylserine (PS), which likely localizes the rigid Ptase-C2 region close to its substrate. Importantly, even in the absence of a localization effect, the C2 domain provides activating signals to the Ptase active site, resulting in a significant increase in catalytic turnover. Combining enzyme kinetics, molecular dynamics (MD) simulations and mutagenesis we identify distinct regions at the domain interface with different effects on SHIP2 activity. Hydrophobic Ptase-C2 interactions increase SHIP2 turnover with the lipid chain containing PI(3,4,5)P3 substrate, whereas polar interactions communicate from the domain interface to a substrate binding loop mainly affecting catalysis of the soluble inositol 1,3,4,5-tetrakisphosphate (IP4) substrate. Together, our data provide novel structural and mechanistic insights on a C2 assisted functioning of SHIP2, and by homology likely of SHIP1.

Results

Structure of SHIP2 Ptase-C2

A construct of human SHIP2 containing residues 420–878 corresponding to the Ptase and a C2 domain (Ptase-C2) was crystallized and diffraction data collected to 1.96 Å. The protein crystallized in the P212121 space group with eight molecules per asymmetric unit (for crystallographic and refinement statistics, see Table 1). Each of the eight molecules (A-H) contains a 5-Ptase and a C2 domain (Figure 1B). The overall structure of the Ptase domain in Ptase-C2 is similar to the one previously reported (Mills et al., 2012; Trésaugues et al., 2014), displaying a central β-sheet sandwich decorated by seven helices (α1-α7) and several loops (L1-L4) (Figure 1B). However, there are significant differences. Loop residues 587–594 (L3) are located at the interface with the C2 domain and are due to extensive interactions with the C2 domain well ordered (Figure 1C). Importantly, a loop proximal to the active site, spanning residues 674–684 (L4), was in previous Ptase structures disordered or closed over the active site (Mills et al., 2012; Trésaugues et al., 2014) (Figure 1D; hereafter referred to as L4-in conformation), whereas in our structure 6 of the Ptase-C2 monomers (A, B, C, E, F, G) exhibit a conformation where L4 points away from the active site (L4-out). These L4-out conformations are stabilized by interactions of R682 in L4 with one or two aspartic acids (D613/D615) adjacent to helix α5 (Figure 1E–F, hereafter referred to as singly or doubly bound L4-out conformations). In molecule D, R682 is half way between the L4-in and L4-out conformation and in monomer H R682 is disordered. Together, these conformations suggest extensive mobility of L4 with a displacement of the guanidinium group of the R682 side chain by ~20 Å between L4-in and L4-out conformations. L4-out conformations appear to be favored by the presence of the C2 domain, since they are only observed for Ptase-C2, although crystal packing analysis suggests ample space in Ptase crystals in this region. In L4-out structures where R682 is doubly bound to D613 and D615 (molecules B and F), a network of hydrogen bonds is established that includes H674, T683 and N684 on L4, rigidifying the loop (Figure 1F).

Table 1.
Diffraction and refinement statistics.

The C2 domain interacts with the Ptase on a face that is opposite to the active site (Figure 1B). The Ptase and C2 domains share an extensive interface of 940 Å2 (Figure 1C), resulting in stabilization of the Ptase (Supplementary file 1). The interface is constituted largely of hydrophobic interactions, with A592, F593, L597, Y697 and P698 in the Ptase and Y770, H820, L822, E836, V838 and A840 in the C2 domain forming a hydrophobic core (Figure 1C, Figure 1—figure supplement 1). A significant peripheral polar contact is formed by R649 in the Ptase, forming hydrogen bonds with E862 and a cation-π interaction with F855 in the C2 domain. Notably, R649 is located in a loop that connects to a stretch of three helices (α5–7, colored grey in Figure 1), which at the other end join to D613 and D615, the docking site for R682 in the L4-out conformation.

The C2 domain of SHIP2

Our structure confirms the presence of a β-sheet sandwich structure typical of C2 domains within residues 746–874 of SHIP2. Using the DaliLite server (Holm and Rosenström, 2010), we identified the closest structural homologues as the C2 domains of the itchy E3 ubiquitin protein ligase (pdb code 2NQ3, Z score = 12.7, RMSD over 106 residues = 2.1 Å) and dysferlin (pdb code 4IQH, Z score = 12.4, RMSD over 110 residues = 2.4 Å), which both belong to the PKC type C2 subfamily (Zhang and Aravind, 2010). A sequence-based similarity search using HH-pred (Söding et al., 2005) also identified the C2 domain of dysferlin as the closest related C2 domain (E-value = 0.049), although with 16% sequence identity the homology is low. The topology adopted by the SHIP2 C2 domain is as for itchy and dysferlin that of type II (or P-variant) (Corbalan-Garcia and Gómez-Fernández, 2014). PKC type C2 domains usually bind phospholipids, most frequently phosphatidylserine (PS), often in a Ca2+-dependent manner via acidic residues present in three Ca2+-binding loops (CBLs) at the tip of the C2 domain (Figure 2A–B) (Cho and Stahelin, 2006). Structure-based sequence alignment indicates that only 2 of 5 acidic residues on CBL1 and CBL3 are conserved between PKCα and SHIP2 (D829 and E832) and two are replaced by serines (S761 and S827). To test for lipid binding and calcium dependency, we performed surface plasmon resonance (SPR) experiments with immobilized PS vesicles and protein lipid overlay (PLO) assays. We found that both the Ptase and Ptase-C2 bind PS, with the C2 domain enhancing the interaction (Figure 2C, Figure 2—figure supplement 1). SPR experiments reveal that PS binding of Ptase-C2 exhibits a small increase of ~10% in presence of Ca2+ (Figure 2D), an effect that was not detected in PLO experiments (Figure 2—figure supplement 1). In summary, we conclude that both, the Ptase and Ptase-C2 bind PS and the small effect of Ca2+suggests that the interaction is dominated by direct PS binding to basic residues, rather than Ca2+-mediated binding to acidic residues and serines. Nevertheless, the low conservation of Ca2+-binding residues in CBL1 and CBL3 of the SHIP2 C2 domain might weakly contribute to Ca2+-mediated lipid binding (Figure 2D).

Figure 2.
The C2 domain of SHIP2 and lipid binding.

The C2 domain affects SHIP2 activity

Next, we tested the effect of the C2 domain on SHIP2 activity. For this, we employed a Malachite Green assay to compare activity of SHIP2 Ptase-C2 to that of the isolated Ptase. Activity was measured with two soluble substrates, IP4 and PI(3,4,5)P3-diC8 and in all cases we observe a behavior close to conventional Michaelis-Menten enzyme kinetics (Figure 3). We find that the C2 domain increases turnover rates for both substrates (Figure 3A–B, Table 2). Interestingly, the effect is significantly larger with PI(3,4,5)P3-diC8, where the C2 domain increases the kcat ~10 fold, while for IP4 the increase is ~1.5 fold. As a result, the C2 domain switches the substrate preference with the Ptase alone turning over IP4 twice as fast as PI(3,4,5)P3-diC8, whereas Ptase-C2 displays ~ 4 fold faster kinetics with PI(3,4,5)P3-diC8 compared to IP4 (Table 2). The KM is less affected by the C2 domain with a significant change only with PI(3,4,5)P3-diC8, where the absence of the C2 domain decreases the KM ~ 2 fold (Table 2). The fact that the distant C2 domain has specific and differential effects on catalysis of the two substrates, suggests the presence of an allosteric communication between the domain interface and the active site (Figure 1B).

Figure 3.
Enzyme kinetics of SHIP2.
Table 2.
Enzymatic parameters are calculated by fitting the Michaelis-Menten equation to substrate titrations. Enzyme concentrations used for activity measurements were 400 nM, or if this caused saturated signals 50 nM (*).

The effect of the C2 domain was further investigated by mutagenesis. In order to disrupt the hydrophobic core of the domain interface we mutated F593 and L597 to aspartates (FLDD mutant). Further, we mutated R649 in the Ptase to alanine (R649A) to prevent polar interactions with E862 and F855 in the C2 domain (Figure 1C). Crystal structures of Ptase-C2 FLDD confirm that few hydrophobic interactions remain intact between the Ptase and C2 domains (Figure 1—figure supplement 2). The FLDD mutations also disrupt interdomain hydrogen bonds between R649 and E862. When comparing the catalytic turnover of the FLDD mutants we find that the C2 domain no longer has a significant effect with either substrate (Figure 3C–D, Table 2), indicating that the FLDD mutations largely mimic the absence of the C2 domain. We note that the FLDD mutations increase intrinsic Ptase activity with PI(3,4,5)P3-diC8, but importantly the presence of the C2 does not cause any further change. Interestingly, the polar R649A mutation specifically eliminates the effect of the C2 domain on IP4 kinetics, while retaining most of its effect on PI(3,4,5)P3 (Figure 3E–F). Together, these results suggest that the polar R649-E682/F855 interactions between Ptase and C2 domains mainly affect SHIP2 activity with IP4, whereas hydrophobic interactions increase activity towards PI(3,4,5)P3.

Next, we tested whether PS binding alters the C2-mediated effects. We find that soluble PS-diC8 has little effect on IP4 kinetics and increases activity ~40% with PI(3,4,5)P3-diC8 (Figure 4A–B), both for Ptase and Ptase-C2. If instead a lipid vesicle embedded PI(3,4,5)P3-diC16 substrate is used, the presence of 30% (mol/mol) PS-diC16 in the vesicles increases SHIP2 activity ~2–3 fold for both, Ptase and Ptase-C2 (Figure 4C). The PS-mediated increase in activity on vesicles is maintained for the FLDD mutant, whereas as with soluble substrates, the increase in activity due to the C2 domain is abrogated by the interface mutations (Figure 4D). In summary, these data show that the effect of PS is amplified for SHIP activity on vesicles, possibly due to a localization or positioning effect and this effect is independent and additive to the enhancing allosteric signaling of the C2 domain.

Figure 4.
Effect of PS binding on SHIP activity.

Communication between the C2 domain and the active site

In order to obtain insight on the path of communication between the C2 domain and the active site we performed unbiased molecular dynamics (MD) simulations of the SHIP2 Ptase in presence or absence of the C2 domain. Simulations were started in the doubly bound L4-out conformation, where R682 interacts with D613 and D615 (based on molecule B in the crystal structure). The simulations indicate that the C2 domain induces several dynamic changes in the Ptase domain, as can be seen from root mean square fluctuations (RMSF) throughout the simulation (Figure 5A). Not surprisingly, L3 is strongly stabilized due to its direct interaction with the C2 domain. On the other hand, L2 (residues 531–539) and L4 fluctuate stronger in presence of the C2 domain. RMSF analysis separated into different L4 states (R682 unbound, singly or doubly bound) indicate that higher L4 fluctuations occur when L4 is detached from D613/D615 (Figure 5—figure supplement 1A). Non-loop regions fluctuate little overall but fluctuations are increased in a stretch containing helices α5 and α6 and the proximal D613/D615 docking site of R682 (shaded grey in Figure 5A–B). Principal component analysis shows as one of the principal motions α5 moving away from α6 (Figure 5A insert, Videos 12). In order to analyze which regions deviate during the simulations most from the doubly bound L4-out starting conformation, we calculated root mean square deviations (RMSD) with respect to the L4-out starting structure, analyzed per residue and averaged over time. We find a significant difference in the region of α5, α6 and the proximal D613/D615, which in the presence of the C2 domain remain closer to the R682 doubly bound starting structure, whereas in absence of the C2 domain this region deviates significantly from the starting conformation (Figure 5B, also observed in Videos 12). The same RMSD analysis of different L4 states indicates that this occurs specifically in conformations where R682 is singly bound to either D613 or D615 (Figure 5—figure supplement 1E). Interestingly, different α5 helix positions are also observed in the Ptase-C2 crystal structure, confirming its mobility, and in agreement with MD simulations helix α5 is shifted specifically in singly bound R682 conformations by ~1.5 Å (Figure 5C). The C2 induced increase of L4 fluctuations when L4 is unbound (Figure 5—figure supplement 1A) together with stabilized D613/D615 when R682 is singly docked (Figure 5—figure supplement 1E) could potentially favor the transition from L4-in to L4-out and might be a reason why the L4-out conformation is only observed in Ptase-C2 crystal structures, but not in Ptase-only or Ptase-C2 FLDD structures.

Video 1.

Principal component analysis of Ptase simulation.

Movements according to the first four eigenvectors from the principle component analysis of the Ptase simulation are shown. The starting conformation in the doubly bound L4-out conformation, based on molecule B of the Ptase-C2 crystal structure, is superimposed and colored blue.

DOI: http://dx.doi.org/10.7554/eLife.26640.018

Video 2.

Principal component analysis of Ptase-C2 simulation.

Movements according to the first four eigenvectors from the principle component analysis of the Ptase-C2 simulation are shown. The starting conformation in the doubly bound L4-out conformation, based on molecule B of the Ptase-C2 crystal structure, is superimposed and colored blue.

DOI: http://dx.doi.org/10.7554/eLife.26640.019

Figure 5.
SHIP2 dynamics.

On the other hand, once R682 detaches from D613/D615 we observe a more sustained release of L4 in MD simulations where the C2 domain is present. While in absence of the C2 domain R682 experiences only brief releases followed by immediate rebinding, in the Ptase-C2 simulation R682 remains fully detached from D613/D615 for ~1 μs of the 4.35 μs simulation (Figure 5D). Indeed, cluster analysis of snapshots from the simulation reveals that for the Ptase alone all of the main clusters remain in a L4-out conformation, whereas for Ptase-C2 the third most populated cluster has R682 detached from D613/D615 (Figure 5—figure supplement 1G–H). Together, these data support a communication between the C2 domain and the active site in the Ptase by modulating the dynamics of helices α5–7 and L4 to facilitate transitions between L4-in and L4-out conformations. Structurally, this suggests a communication via R649, since this residue connects helices α5–7 to the C2 domain.

To understand how L4 dynamics could affect catalysis, we next analyzed the mode of substrate binding in SHIP2. Co-crystallization experiments using the catalytically inactive SHIP Ptase-C2 D607A mutant with Mg2+ and substrate failed to reveal clear electron density for the substrate. Therefore we generated a model based on the homologue INPP5B crystal structure containing the PI(4)P product (Trésaugues et al., 2014) and insights from AP endonucleases, which are thought to share a common phosphate hydrolysis mechanism (Whisstock et al., 2002). We modeled the substrate according to a Michaelis-Menten complex, where all catalytic residues are in place for catalysis (Figure 6A–B). In this complex the catalytic base D607 stabilizes a water molecule for 5 P attack. According to our model a key 4 P interacting residue is N684 on L4. N684 can adopt two conformations (Figures 1D–F and and5C)5C) and needs to point towards the active site to interact with the substrate 4 P, which is seen in unbound or singly bound, but not doubly bound L4 conformations. In our model L4 adopts the L4-in conformation seen in 2 SHIP2 Ptase structures (PDB: 3NR8 and 4A9C; Mills et al., 2012; Trésaugues et al., 2014). In this conformation R682 is ideally positioned to interact with the 3 P of PI(3,4,5)P3. We performed unbiased MD simulations of SHIP Ptase or Ptase-C2 starting with the PI(3,4,5)P3-diC8 or IP4 substrates bound according to our model shown in Figure 6A–B. Interestingly, the C2 domain significantly restricts the displacement of the headgroup of both substrates away from the Michaelis-Menten complex (Figure 6C–D, Videos 34).

Video 3.

Simulation of Ptase with substrate.

The first 100 ns of the trajectory from the simulation of Ptase bound to PI(3,4,5)P3-diC8 and Mg2+ are shown.

DOI: http://dx.doi.org/10.7554/eLife.26640.022

Video 4.

Simulation of Ptase-C2 with substrate.

The first 100 ns of the trajectory from the simulation of Ptase-C2 bound to PI(3,4,5)P3-diC8 and Mg2+ are shown.

DOI: http://dx.doi.org/10.7554/eLife.26640.023

Figure 6.
Model of substrate bound SHIP2.

Although our co-crystallization experiments did not reveal bound substrate they did exhibit electron density corresponding to a Mg2+ ion and a phosphate group (Figure 6—figure supplement 1A). The Mg2+ ion is on average displaced by 1.3 Å and the phosphate by 1.7 Å compared to the Mg2+ and 5 P positions in our model (Figure 6—figure supplement 1B). The position of the phosphate does not leave space for an attacking water bound to D607, hence this could represent the post-catalysis positions of the Mg2+ and 5 P.

Probing the communication path by mutagenesis

The simulations and crystal structures indicate that a potential allosteric path linking the C2 domain to the active site could lead via R649 to helices α5–7 and at the other end of the helices through D613 and D615 to R682 and L4. To test this hypothesis, we generated the following mutants along the path: R649A (already generated as an interface mutant), R691A, D613A/D615A, R682A, N684A as well as active site mutants D607A and R665A (Figure 7A). All mutants were generated for Ptase and Ptase-C2 with the exception of R691A and D613A/D615A, which appear to destabilize the isolated Ptase and only yielded soluble protein for Ptase-C2. Circular dichroism (CD) analysis indicates that most of the mutations do not alter protein structure, apart from R691A in Ptase-C2 and N684A in the Ptase (Figure 7—figure supplement 1). The mutants were tested for changes in activity with IP4 and PI(3,4,5)P3-diC8 (Figure 7, Table 3). Interestingly, D613A/D615A mutated Ptase-C2 behaves similar to the R649A mutant, in that both only affect turnover with IP4, but have little effect on PI(3,4,5)P3 kinetics (Figures 3E–F and 7B–C). This similitude in behavior supports an allosteric communication via R649 and D613/D615. Further, we find that the R682A and N684A mutants have lost most of their activity (Table 3; note that for N684A this is also the case for the CD-stable Ptase-C2 form), supporting a key role for these residues in substrate binding via 3 P and 4 P, respectively (Figure 6A). R691 links the Ptase core β-sandwich with helices α5–7 (Figure 7A) and its mutation inactivates Ptase-C2, which together with the changed CD spectra indicates a structural role for this residue. Unexpectedly, R665A mainly affects the Ptase domain, but has little effect on Ptase-C2 activity (Figure 7D–E, Table 3). According to our model, R665 stabilizes a water molecule in the coordination sphere of the Mg2+ ion (Figure 6B). This fits with MD simulations showing C2-mediated stabilization of the substrate headgroup (Figure 6C–D), apparently making the R665-mediated stabilization dispensable in presence of the C2 domain. Together, these results support an allosteric communication from the C2 domain via R649 and helices α5–7 to L4, which by changing their dynamic behavior affects enzyme kinetics with the IP4 headgroup. In addition, they support an overall stabilization of the Ptase active site by the C2 domain allowing both substrates to bind in a catalytically productive mode.

Figure 7.
SHIP2 mutational analysis.
Table 3.
Enzymatic parameters are calculated by fitting the Michaelis-Menten equation to substrate titrations. Enzyme concentrations used for activity measurements were 400 nM, or if this caused saturated signals 50 nM (*). NA, not analyzed; ND, not determinable. ...

Lastly, to probe how C2 interactions affect activity of full-length SHIP2 in a cellular setting we introduced the FLDD mutations or a C2 deletion ([increment]C2) into the SHIP2 sequence and transiently expressed the full-length proteins in HEK293 cells (Figure 8—figure supplement 1). Expressing wild-type SHIP2 causes a significant reduction in downstream Akt phosphorylation levels (Figure 8). Removing the C2 domain abolishes observed SHIP2 activity, whereas the FLDD mutations partially reduce activity of the wild-type protein. This confirms that C2 interactions observed in the crystal structure are important for SHIP2 activity in cells.

Figure 8.
SHIP2 cellular activity.

Discussion

Of the enzymes controlling PI(3,4,5)P3 levels (PI3K, PTEN and SHIP1/2), the SHIP enzymes lag far behind in terms of a detailed understanding of their functioning and regulation. Surprisingly, even though the Ptase domains of PTEN and SHIP2 bear no homology at the sequence or structural level, they conceptually share an important feature in that they are rigidly linked to a C2 domain (Lee et al., 1999). Despite the low homology between the PTEN and SHIP2 C2 domains (4% identity in a structural alignment) they use a common face for Ptase interactions. Indeed, a corresponding surface is also used by the C2 domain of PLCδ1 to form extensive interactions with its catalytic domain (Essen et al., 1996). It appears therefore, that the C2 domain presents an ideal scaffold to dock lipid modifying enzymes onto the membrane and via rigid interactions orient the catalytic domain for productive substrate attack. For SHIP2, we show that the C2 domain is essential for cellular function and the rigid interface enhances its efficiency (Figure 8). Assuming that the SHIP2 C2 domain engages the membrane via motifs known to bind lipids in other C2 domains (Cho and Stahelin, 2006) and taking into account several positively charged residues on the Ptase and C2 domains facing the membrane, we provide a model of SHIP2 Ptase-C2 docked to a lipid bilayer (Figure 9A). Although, we cannot rule out domain rearrangements between the Ptase and C2 domains upon lipid binding, the extensive domain interface between the two domains argues against large changes and suggests that the Ptase-C2 portion of SHIP2 represent the minimal fully active catalytic unit. Moreover, our model indicates that the rigid Ptase-C2 fragment can bind the membrane and engage substrate without the need of extensive extraction of the PI(3,4,5)P3 substrate from the lipid bilayer. In SHIP2, PS binding occurs through both, the C2 and the Ptase domains and, as shown previously, interactions with the Ptase enhance SHIP2 activity (Vandeput et al., 2006). A similar scenario has indeed been observed for PTEN, where regions outside the C2 domain also contribute to membrane interactions and binding of acidic phospholipids to the N-terminus enhances activity (Campbell et al., 2003; Walker et al., 2004). The fact that the PS effect is larger when embedded in vesicles compared to soluble PS-diC8 (Figure 4) suggests that PS mediated SHIP2 localization and/or positioning on vesicles does contribute to activity on lipid membranes, however, binding via the Ptase appears sufficient for this effect.

Figure 9.
Model of SHIP2 catalytic cycle.

Independent to a role in membrane binding, our data suggest that the C2 domain provides important allosteric interdomain effects to increase Ptase activity. In contrast to a traditional concept of allosteric signaling, effects by the C2 domain appear to be transmitted via changes in protein dynamics and active site stabilization rather than large conformational changes. Our data suggest that hydrophobic and polar regions at the Ptase-C2 domain interface differentially affect the lipid chains or head group of the PI(3,4,5)P3 substrate. The fact that the R649A mutant, which presumably retains hydrophobic Ptase/C2 interactions, only removes C2 effects on IP4 but not PI(3,4,5)P3 kinetics, indicates that R649 communicates with the PI(3,4,5)P3 headgroup and hydrophobic interactions with PI(3,4,5)P3 lipid chains. Since the PI(3,4,5)P3 lipid chains are predicted to interact with L2 (Figure 6A), this suggests a communication between hydrophobic C2 interactions and L2.

Further, our study has identified a specific path of polar connections, linking the C2 domain to the Ptase active site. This ‘polar’ path leads via C2 interactions with R649 in the Ptase to helices α5–7 and at the other end of the helices to D613/D615 and R682 in L4. Mutation of R691 at the interface between helices α5–7 and the Ptase core inactivates SHIP likely due to structural destabilization (Table 3, Figure 7—figure supplement 1B), and similar inactivating mutations at this interface have been associated with opsismodysplasia (P659S and W688C) and protection from diabetes (L632I) (Huber et al., 2013; Kagawa et al., 2005). More specific disruption of this path, by either R649A or D613A/D615A mutations specifically removes C2 effects on IP4 (Figures 3E–F and 7B–C). In a cellular setting this path could be relevant, firstly for SHIP2 activity on intracellular IP4, whose cellular roles are only emerging for example in T-cell signaling (Huang and Sauer, 2010). Secondly, it is likely also important when the SHIP2 enzyme approaches the PI(3,4,5)P3 substrate in membranes with the lipid chains buried in the lipid bilayer.

A critical observation in our study is that L4, proximal to the active site, can switch between a closed (L4-in) and an open (L4-out) conformation (Figure 1D–F) and our simulations indicate that the dynamics of this loop and the region capturing the loop in the open conformation is altered by the C2 domain (Figure 5, Figure 5—figure supplement 1). We propose a catalytic cycle for SHIP2 where efficient opening and closing of L4 is important for high turnover (Figure 9B, Video 5): To initiate the cycle, L4 needs to open to allow substrate entry to the active site. Initial interactions of the PI(3,4,5)P3 head group include 4 P binding to N684, which prevents the formation of a hydrogen bond network seen in the R682 doubly bound L4-out conformation (Figure 1F). With this hydrogen network disrupted L4 is more mobile (see e.g. Figure 1E) facilitating the switch to L4-in. We propose closure of L4 over the substrate as an important second step in catalysis, allowing R682 interactions with the substrate 3 P. With the substrate tightly bound and correctly positioned, catalysis proceeds. Once 5 P cleavage has occurred we propose that L4 needs to open again for product release. In absence of 3 P and 5 P interactions, reestablishment of the hydrogen network as seen in the R682 doubly bound L4-out conformation that orients the N684 sidechain towards H674 (Figure 1F), might aid ejection of the product, hence closing the catalytic cycle.

Video 5.

The SHIP2 catalytic cycle.

The conformational changes between L4-out doubly bound, L4-out singly bound and L4-in conformations are shown as morphs between observed crystal structures and assembled in sequence according to the model of the catalytic cycle explained in Figure 9B.

DOI: http://dx.doi.org/10.7554/eLife.26640.034

It has been previously proposed that the 3 P specificity of SHIP2 is conferred by R682 (Trésaugues et al., 2014). Our model shows that R682 in the L4-in conformation is well placed to interact with the 3 P of PI(3,4,5)P3 (Figure 6A) and mutation of R682 confirms an important role for this residue in catalysis (Table 3). In other 5-Ptases the loop corresponding to L4 is significantly shorter and residues corresponding to R682 (K516 in INPP5B) and N684 (R518 in INPP5B) on L4 interact with the 4 P and belong to a set of residues defined as the P4 interacting motif (P4IM) (Trésaugues et al., 2014). Possibly, R682 in SHIP2 might also contribute to 4 P interactions, which could account for the observed PI(4,5)P2 activity of SHIP2 (Elong Edimo et al., 2016; Nakatsu et al., 2010; Taylor et al., 2000), however, several reports demonstrate a clear preference for PI(3,4,5)P3 (Chi et al., 2004; Pesesse et al., 1998; Trésaugues et al., 2014; Vandeput et al., 2007). Another difference in SHIP2 is the somewhat weaker Mg2+ coordination compared to other 5-Ptases. An asparagine in INPP5B (N275) providing a Mg2+coordination is replaced by a glycine in SHIP2 (G434) and the coordination is likely provided by a water molecule (Figure 6B). This water can be stabilized by R665, which appears important only in absence of active site stabilization by the C2 domain (Figures 6C–D and 7D–E, Table 3).

Other head group interactions are conserved among 5-Ptases and the mechanism initiating 5 P attack is well described and shared among 5-Ptases as well as the AP endonucleases (Dlakić, 2000; Whisstock et al., 2000, 2002). However, the exact mechanism of catalysis is still being debated. Different models have for AP nucleases proposed the involvement of either one or two static, or one moving metal ion (Beernink et al., 2001; Mol et al., 2000; Oezguen et al., 2007). A 2-metal hypothesis for AP nucleases is supported by a crystal structure of APE1, which has two Pb2+ ions bound 5 Å apart from each other, with one ion at the site seen in 5-Ptases (A-site) and the other at a more buried site (B-site) (Beernink et al., 2001). Similar to a recent structure of INPP5B (Mills et al., 2016), we observe in our Ptase-C2 D607A structure the Mg2+ displaced on average by 1.3 Å from the A towards the B-site and a phosphate group shifted by 1.7 Å from the 5 P location towards the attacking water (Figure 6—figure supplement 1B). As suggested by Mills et al. (2016) this likely represents the post-cleavage position for Mg2+ and the 5 P, since there is no longer space for the attacking water. In contrast, for OCRL a free phosphate is observed at the pre-cleaved 5 P site and the Mg2+ at the A-site (Trésaugues et al., 2014), hence there appear to be subtle differences in the post-cleavage positioning of the Mg2+ and phosphate for different 5-Ptases. We note that the movement of the Mg2+ seen in INPP5B and SHIP2 from the A towards the B-site is opposite to the B to A-site movement proposed to occur during catalysis in AP nucleases (Oezguen et al., 2007).

SHIP1 shares an overall 45% sequence identity with SHIP2 (65% within the 5-Ptase and 43% in the C2 domain), suggesting that many of the mechanisms described here are conserved among the two enzymes. Residues at the Ptase-C2 domain interface are mostly conserved, suggesting that the interface as well as a possible communication to L2, which is fully conserved, are also present in SHIP1. The polar path is partially conserved with R649 replaced in SHIP1 by H632, D613/D615 are with D593 and E598 similar and R682 is substituted by K665. The equal length of L4 suggests that it will be able to sample a similar space and likely K665 provides the 3 P specificity in SHIP1.

In conclusion, our study provides important insights into how the C2 domain assists SHIP2 catalysis via a combination of membrane positioning, active site stabilization and allosteric signaling, which together is crucial for efficient cellular functioning (Figure 8). Since SHIP enzymes are highly implicated in disease, this will likely aid the design of novel strategies to target SHIP enzymes with high specificity.

Materials and methods

Protein expression and purification

The experimental details regarding the cloning, expression and purification of human SHIP2 Ptase-C2 have been described in (Le Coq et al., 2016). The Ptase was cloned, expressed and purified following the same protocol, whereas the isolated C2 could not be purified due to insolubility of the protein.

Crystallization, data collection

The crystallization conditions for the Ptase-C2 have been previously described (Le Coq et al., 2016). Briefly, crystals were grown by mixing an equal volume of the protein with the precipitant solution (100 mM Bis-Tris propane, pH 7.0, 0.2 M NaNO3, 20% (w/v) PEG3350, 0.025% (v/v) CH2Cl2). Crystals were cryoprotected in the precipitant solution with 25% ethylene glycol and then flash-frozen. SHIP2 Ptase-C2 D607A (5.5 mg/mL) in the presence of PI(3,4,5)P3-diC8 and Mg2+ (0.450 mM and 0.750 mM respectively) and SHIP2 Ptase-C2 FLDD (5.32 mg/mL) crystallization conditions were initially rescreened. For SHIP2 Ptase-C2 D607A mixed with PI(3,4,5)P3diC8 and Mg2+ conditions were optimized to 0.1 M Bis-Tris propane, pH 7.0, 0.4 M KSCN, 20% (w/v) PEG3350 for SHIP2 Ptase-C2 D607A mixed with PI(3,4,5)P3diC8 and Mg2+. The crystals were flash-frozen in precipitant with additional 25% ethylene glycol, 0.220 mM of PI(3,4,5)P3diC8 and 0.750 mM of Mg2+. For SHIP2 Ptase-C2 FLDD crystals were grown in 50 mM Hepes pH 7.1, 17.5% PEG 1000 and cryopreserved in additionally 25% ethylene glycol (I2 crystals) or 0.1 M PCB buffer pH 7 (sodium propionate, sodium cacodylate, and BIS-TRIS propane; 2:1:2 molar ratio), 25% PEG 1500 and cryopreserved in additionally 25% ethylene glycol (P21 crystals).

Structure determination, Refinement, and Analysis

Crystallographic data for all protein crystals were measured at 100K, on beamlines ID23-2 (λ = 0.873 Å) and ID29 (λ = 0.976 Å) at European Synchrotron Radiation Facility (ESRF, Grenoble) and BL13-XALOC-BL13 (ALBA, Barcelona, for preliminary data collection) using a Pilatus 6M detector. XDS (Kabsch, 2010) was used for indexing and integration and AIMLESS (Evans and Murshudov, 2013) for scaling all diffraction data. The best dataset for SHIP2 Ptase-C2 WT diffracted to a resolution of 1.96 Å and reveals a P212121 space group with unit cell parameters a = 136.04 Å, b = 175.84 Å, c = 176.89 Å. The Matthews coefficient is 2.54 Å3.Da−1 corresponding to 8 molecules in the asymmetric unit (asu) with a solvent content of 51.7% (Le Coq et al., 2016). The crystal structure of human SHIP2 Ptase model (PDB: 3NR8) was used as search probe for molecular replacement (MR) using Phaser (McCoy et al., 2007), which identified 8 clear solutions for the Ptase domain. MR using various available C2 structures as search probes failed. C2 domains where therefore built using ARP/wARP (Langer et al., 2008) and manual building with COOT (Emsley and Cowtan, 2004; Emsley et al., 2010). Translation/Libration/Screw (TLS) and Maximum likelihood of restrained refinement were performed using REFMAC5 (Murshudov et al., 1997). Final R-factors for SHIP2 Ptase-C2 WT are 18.0/20.8 (Rwork/Rfree). SHIP2 Ptase-C2 D607A crystallized in the same space group as wild type SHIP2 Ptase-C2, but Phaser (McCoy et al., 2007) was used to generate the initial model since it produced better initial R values than directly using the SHIP2 Ptase-C2 WT model. Refinement was performed as described above. A Mg2+ and a PO43- were placed in the active site of 6 molecules in the asu (monomers C-H). SHIP2 Ptase-C2 FLDD crystallized in two different space groups, one in P21 (a = 44.03, b = 81.12, c = 128.90, β = 92.85) with 2 molecules per asu and the other in I2 (a = 43.74, b = 73.43, c = 158.0, β = 90.7) with 1 molecule per asu. The Ptase-C2 WT model was used in Phaser (McCoy et al., 2007) to provide initial phases. Initial rebuilding was performed with ARP/wARP (Langer et al., 2008) and further refinement was performed as described above. All data processing and structure refinement statistics are summarized in Table 1. All residues in the final models lay within the favored region of the Ramachandran plot with the exception of the model of SHIP2 Ptase-C2_D607A which has 0.2% Ramachandran outliers.

CD spectroscopy

Circular dichroism was used to assess and compare the overall protein fold of the different constructs. Far-UV CD spectra between 250 and 200 nm of protein at 0.25 mg/mL were run at 20°C for each sample.

Thermal melting

Tm’s were measured using a ThermoFluor assay. Protein at 1–10 μM was mixed with SYPRO Orange and subjected to a temperature gradient of 0.5°/min from 20 to 95°C and fluorescence recorded at 570 nm.

Lipid binding

For the PLO assay dipalmitoyl phosphatidylserine or dipalmitoyl phosphatidylcholine (Echelon Biosciences, Salt Lake City, UT) were spotted on a nitrocellulose membrane, dried and blocked with 5% skimmed milk. The membranes were then incubated for 1 hr with 2.5 μg/ml of GST fused SHIP Ptase or Ptase-C2 in blocking solution in presence or absence of 1 mM CaCl2, washed and developed using an HRP conjugated anti-GST antibody.

For SPR measurements vesicles containing 30% PS (16:0; Echelon Biosciences) and 70% PC (from chicken egg; Avanti Polar Lipids, Alabaster, AL) or 30% PC (16:0, Avanti Polar Lipids) and 70% PC (chicken egg) were prepared at a final total lipid concentration of 1.5 mM. Organic solvent was removed by rotary evaporation for 1.5 hr at 45°C. The lipid film was resuspended in SPR running buffer (20 mM HEPES, pH 7.5, 150 mM NaCl, 1 mM TCEP) and subjected to six cycles of freeze-thaw and passed 15 times through a membrane with 100 nm pore size, using a mini extruder (Avanti Polar Lipids). SPR experiments were performed on a Biacore X100 instrument (GE Healthcare, Chicago, IL). A L1 sensor chip was washed with two 1 min injections of a 2:3 (v/v) isopropanol: 50 mM NaOH solution at a flow rate of 10 µl/min and then coated by injecting PS containing vesicles in the active flow cell (Fc2) and PC vesicles in the reference flow cell (Fc1), for 15 min at a flow rate of 2 μl/min, followed by two 1 min injections of 10 mM NaOH solution at 10 μl/min to remove loosely bound vesicles and for stabilization. SHIP2 Ptase or Ptase-C2 proteins were injected for 2 min at a flow rate of 30 μl/min in running buffer. SPR responses plotted in Figure 2C-insert and Figure 2D correspond to response units (RUs) at 10 s after injection, when a relatively constant steady state binding phase is reached (dashed line in Figure 2C). Injections in presence of Ca2+ were performed with 0.5 mM CaCl2. For all data shown in Figure 2C–D, buffer injections (with or without calcium, accordingly) are subtracted.

Enzyme kinetics

Activity measurements were performed using a Malachite Green phosphatase activity assay. Proteins were incubated for 2 min at 23°C with the substrate in a reaction buffer containing 20 mM Hepes, 150 mM NaCl, 2 mM MgCl2, 1 mM TCEP, pH 7 in a total volume of 25 μL. The reaction was quenched by addition of 5 μL of 0.5 M EDTA, pH 8. Subsequently, 25 μL of the reaction was mixed with 100 μL of Malachite Green solution (Echelon Biosciences) and left to incubate for 15 min at room temperature and the optical density was measured at 620 nm. Each data point is measured at least in triplicates. For activity assays shown in Figures 4A–B 100 μM soluble PS-diC8 or PC-diC8 were included.

For activity measurements with vesicles two sets of PI(3,4,5)P3 vesicles were prepared at a total lipid concentration of 4.5 mM, containing either 10% (mol/mol) PI(3,4,5)P3 (16:0; Echelon Biosciences), 30% (mol/mol) PS (16:0, Echelon Biosciences) and 60% (mol/mol) PC (from chicken egg, Avanti Polar Lipids) or for control vesicles PS was replaced by 30% (mol/mol) PC (16:0, Avanti Polar Lipids). The organic solvent was removed by rotary evaporation for 90 min at 45°C. The lipid film lipid was subsequently resuspended in 20 mM HEPES, 150 mM NaCl, 1 mM TCEP, pH 7 and sonicated for 35 min. To calculate the dilution of vesicles in experiments shown in Figure 4C–D, only half of the total PI(3,4,5)P3-diC16 concentration, present in the outer leaflet of vesicles, was considered.

Cell assay

HEK293 cells were maintained in Dulbecco’s modified Eagle’s medium (DMEM) containing 10% fetal bovine serum (FBS) and supplemented with antibiotics. The cells were transiently transfected with the empty vector (pOPINJ) or vector expressing GST fused full-length SHIP2 (WT, FLDD or [increment]C2). The transient transfection was performed using polyethylenimine (Hsu and Uludağ, 2012). After 48 hr in DMEM with 10% FBS the cells were harvested and levels of total Akt, phospho-T308 of Akt (Cell signaling, Danvers, MA, AB_2255933, AB_329827) and SHIP2 (Abcam, UK, AB_2686895) expression levels were assessed by immunoblotting. Quantifications were performed with ImageJ on eight independent experiments, each in triplicates (n = 24). Standard errors in Figure 8 are calculated using values of pAkt/Akt ratios and for Figure 8—figure supplement 1 SHIP2/Vinculin (AB_477629) ratios. The significance is calculated using a two-tailed unpaired Student t test assuming Gaussian distribution.

MD simulations

For apo simulations, the initial model was based on molecule B of the SHIP Ptase-C2 WT structure, which has a L4-out conformation. Missing residues in the linker (residues 731–746) were modelled with ModLoop (Fiser et al., 2000) and energy minimized. For substrate bound models L4 was replaced with an L4-in conformation from the SHIP2 Ptase (3NR8). The position of PI(3,4,5)P3, Mg2+ and catalytic waters were based on the INPP5B crystal structure bound to PI(3,4)P2 (3MTC, adding the 5 P and 3 P) and based on AP endonucleases (Whisstock et al., 2002). Lipid parameters for IP4 and PI(3,4,5)P3 and a net charge of −5 for protein bound IP4/PI(3,4,5)P3 were based on previous studies (Rosen et al., 2011; Slochower et al., 2013). For Ptase simulations the C2 and linker residues (733-874) were omitted. Hydrogens were added using the H++ server (Gordon et al., 2005). The systems were solvated in a cubic simulation cell of 80 Å edge length, constructed from an equilibrated box of TIP3P water molecules (Mahoney and Jorgensen, 2000). The size of the simulation cell was chosen such that a distance of at least 10 Å between the surface of the protein and the cell boundaries was maintained. The systems were neutralized with Na+/Cl-. All calculations were carried out with periodic boundary conditions. All complexes were minimized in three stages. In the first stage, the water molecules and the ions were energy minimized, while harmonic restraints of 1000 kJ/(mol·nm2) were applied to the protein and the lipid; 500 steps of steepest descent and 500 steps of conjugate gradient minimization were carried out. In the second stage the lipid harmonic restrains were removed and another 500 steps of minimization were carried out. In the last stage, all restrains were removed and the entire system was minimized for 2000 steps. The resulting configurations were used for MD production: the lipid systems were simulated for a total of 300 ns each, and the non-lipid simulations for 3 × 1.5 μs each (three independent simulations for each system). The first 50 ns of each simulation were considered equilibration time and omitted from the analysis. All minimization and MD production were performed using version 4.6 of the molecular simulations package GROMACS (Hess et al., 2008; Van Der Spoel et al., 2005) with the AMBER99SB*-ILDN force field (Lindorff-Larsen et al., 2010) with the dihedral corrections of Best Hummer (Best and Hummer, 2009). The isobaric-isothermal (NPT) ensemble was employed for all MD calculations at 300 K with a velocity-rescale thermostat (Bussi et al., 2007) and a time step of 2 fs.

RMSF and RMSD analysis

RMSF and RMSD were calculated on the Cα of the 5-Ptase domain from the total 4.35 μs MD simulations, using the g_rmsf and g_rms tools from the GROMACS package. RMSDs were calculated per residue, using the L4-out structure as reference, and averaged over time. For plots shown in Figure 5—figure supplement 1A–F, RMSF and RMSD were calculated separately over snapshots, in which R682 was unbound, singly bound to either D613 or D615, or doubly bound to D613 and D615. Bonds were considered for minimal R682-D613/D615 distances below 4 Å (excluding hydrogens), hence including hydrogen bond and electrostatic bonding character.

Principal component analysis

PCA was performed on all Cα atoms, applying g_covar and g_anaeig tools from GROMACS package on the full MD trajectories, to calculate, diagonalize and analyze the covariance matrix respectively. PCA identifies collective motions associated with the largest variance and the lowest frequencies shown in the first few eigenvectors, thus, we analyzed the first eight PC of each simulation.

Cluster analysis

Cluster analysis was performed on all non-hydrogen atoms using g_cluster with the gromos algorithm (Daura et al., 1999) and a RMSD cutoff of 0.15 nm. The structure with the smallest average distance to the others was written as the representative structure in each cluster.

Acknowledgements

We thank José Terrón Bautista for help with MD analysis. We thank the ESRF and ALBA for providing the synchrotron-radiation facilities and the staff for their assistance in the data collection. We are grateful to the Barcelona Supercomputing Centre and National Supercomputing Centre (BSC-CNS) for allocating computer time to run the reported simulations. The work was supported by the Spanish Ministry of Economy, Industry and Competitiveness (MEIC) Grants BFU2010-15923 (DL) and MEIC Project Retos BFU2016-77665-R co-funded by the European Regional Development Fund (ERDF) (DL), the Comunidad Autónoma de Madrid Grant S2010/BMD-2457 (DL), and by the National Cancer Research Centre. DL is also a recipient of awards from the Volkswagen Foundation (Az: 86 416–1) and Worldwide Cancer Research (15-1177).

Funding Statement

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Funding Information

This paper was supported by the following grants:

  • Ministerio de Economía, Industria y Competitividad BFU2010-15923 to Daniel Lietha.
  • Comunidad de Madrid S2010/BMD-2457 to Daniel Lietha.
  • Ministerio de Economía, Industria y Competitividad Project Retos BFU2016-77665-R to Daniel Lietha.
  • http://dx.doi.org/10.13039/501100008530European Regional Development Fund Project Retos BFU2016-77665-R to Daniel Lietha.

Additional information

Competing interests

The authors declare that no competing interests exist.

Author contributions

JLC, Conceptualization, Data curation, Formal analysis, Validation, Investigation, Visualization, Methodology, Writing—original draft, Writing—review and editing.

MC-A, Data curation, Formal analysis, Investigation, Visualization, Writing—review and editing.

JVV, Investigation, Methodology, Writing—review and editing.

CMS, Investigation, Writing—review and editing.

LHG, Formal analysis, Investigation.

RC-O, Formal analysis, Investigation, Writing—review and editing.

ND, Formal analysis, Supervision, Writing—review and editing.

DL, Conceptualization, Data curation, Formal analysis, Supervision, Funding acquisition, Investigation, Visualization, Writing—original draft, Project administration, Writing—review and editing.

Additional files

10.7554/eLife.26640.035

Supplemental file 1.

Thermal melting (Tm) of SHIP2 WT and mutants.

Tm values are determined by thermofluor. NA, not analyzed; WS, weak signal.

DOI: http://dx.doi.org/10.7554/eLife.26640.035

References

  • Beernink PT, Segelke BW, Hadi MZ, Erzberger JP, Wilson DM, Rupp B. Two divalent metal ions in the active site of a new crystal form of human apurinic/apyrimidinic endonuclease, Ape1: implications for the catalytic mechanism. Journal of Molecular Biology. 2001;307:1023–1034. doi: 10.1006/jmbi.2001.4529. [PubMed] [Cross Ref]
  • Below JE, Earl DL, Shively KM, McMillin MJ, Smith JD, Turner EH, Stephan MJ, Al-Gazali LI, Hertecant JL, Chitayat D, Unger S, Cohn DH, Krakow D, Swanson JM, Faustman EM, Shendure J, Nickerson DA, Bamshad MJ, University of Washington Center for Mendelian Genomics Whole-genome analysis reveals that mutations in inositol polyphosphate phosphatase-like 1 cause opsismodysplasia. The American Journal of Human Genetics. 2013;92:137–143. doi: 10.1016/j.ajhg.2012.11.011. [PubMed] [Cross Ref]
  • Best RB, Hummer G. Optimized molecular dynamics force fields applied to the helix-coil transition of polypeptides. The Journal of Physical Chemistry B. 2009;113:9004–9015. doi: 10.1021/jp901540t. [PMC free article] [PubMed] [Cross Ref]
  • Brauer H, Strauss J, Wegner W, Müller-Tidow C, Horstmann M, Jücker M. Leukemia-associated mutations in SHIP1 inhibit its enzymatic activity, interaction with the GM-CSF receptor and Grb2, and its ability to inactivate PI3K/AKT signaling. Cellular Signalling. 2012;24:2095–2101. doi: 10.1016/j.cellsig.2012.07.017. [PubMed] [Cross Ref]
  • Bussi G, Donadio D, Parrinello M. Canonical sampling through velocity rescaling. The Journal of Chemical Physics. 2007;126:014101 doi: 10.1063/1.2408420. [PubMed] [Cross Ref]
  • Campbell RB, Liu F, Ross AH. Allosteric activation of PTEN phosphatase by phosphatidylinositol 4,5-bisphosphate. Journal of Biological Chemistry. 2003;278:33617–33620. doi: 10.1074/jbc.C300296200. [PubMed] [Cross Ref]
  • Chi Y, Zhou B, Wang WQ, Chung SK, Kwon YU, Ahn YH, Chang YT, Tsujishita Y, Hurley JH, Zhang ZY. Comparative mechanistic and substrate specificity study of inositol polyphosphate 5-phosphatase schizosaccharomyces pombe synaptojanin and SHIP2. Journal of Biological Chemistry. 2004;279:44987–44995. doi: 10.1074/jbc.M406416200. [PubMed] [Cross Ref]
  • Cho W, Stahelin RV. Membrane binding and subcellular targeting of C2 domains. Biochimica Et Biophysica Acta (BBA) - Molecular and Cell Biology of Lipids. 2006;1761:838–849. doi: 10.1016/j.bbalip.2006.06.014. [PubMed] [Cross Ref]
  • Corbalan-Garcia S, Gómez-Fernández JC. Signaling through C2 domains: more than one lipid target. Biochimica Et Biophysica Acta (BBA) - Biomembranes. 2014;1838:1536–1547. doi: 10.1016/j.bbamem.2014.01.008. [PubMed] [Cross Ref]
  • Daura X, Gademann K, Jaun B, Seebach D, van Gunsteren WF, Mark AE. Peptide folding: when simulation meets experiment. Angewandte Chemie International Edition. 1999;38:236–240. doi: 10.1002/(SICI)1521-3773(19990115)38:1/2<236::AID-ANIE236>3.0.CO;2-M. [Cross Ref]
  • Dlakić M. Functionally unrelated signalling proteins contain a fold similar to Mg2+-dependent endonucleases. Trends in Biochemical Sciences. 2000;25:272–273. doi: 10.1016/S0968-0004(00)01582-6. [PubMed] [Cross Ref]
  • Dubois E, Jacoby M, Blockmans M, Pernot E, Schiffmann SN, Foukas LC, Henquin JC, Vanhaesebroeck B, Erneux C, Schurmans S. Developmental defects and rescue from glucose intolerance of a catalytically-inactive novel Ship2 mutant mouse. Cellular Signalling. 2012;24:1971–1980. doi: 10.1016/j.cellsig.2012.06.012. [PubMed] [Cross Ref]
  • Elong Edimo W, Ghosh S, Derua R, Janssens V, Waelkens E, Vanderwinden JM, Robe P, Erneux C. SHIP2 controls plasma membrane PI(4,5)P2 thereby participating in the control of cell migration in 1321 N1 glioblastoma cells. Journal of Cell Science. 2016;129:1101–1114. doi: 10.1242/jcs.179663. [PubMed] [Cross Ref]
  • Emsley P, Cowtan K. Coot: model-building tools for molecular graphics. Acta Crystallographica Section D Biological Crystallography. 2004;60:2126–2132. doi: 10.1107/S0907444904019158. [PubMed] [Cross Ref]
  • Emsley P, Lohkamp B, Scott WG, Cowtan K. Features and development of coot. Acta Crystallographica Section D Biological Crystallography. 2010;66:486–501. doi: 10.1107/S0907444910007493. [PMC free article] [PubMed] [Cross Ref]
  • Essen LO, Perisic O, Cheung R, Katan M, Williams RL. Crystal structure of a mammalian phosphoinositide-specific phospholipase C Delta. Nature. 1996;380:595–602. doi: 10.1038/380595a0. [PubMed] [Cross Ref]
  • Evans PR, Murshudov GN. How good are my data and what is the resolution? Acta Crystallographica Section D Biological Crystallography. 2013;69:1204–1214. doi: 10.1107/S0907444913000061. [PMC free article] [PubMed] [Cross Ref]
  • Fiser A, Do RK, Sali A. Modeling of loops in protein structures. Protein Science. 2000;9:1753–1773. doi: 10.1110/ps.9.9.1753. [PubMed] [Cross Ref]
  • Giuriato S, Blero D, Robaye B, Bruyns C, Payrastre B, Erneux C. SHIP2 overexpression strongly reduces the proliferation rate of K562 erythroleukemia cell line. Biochemical and Biophysical Research Communications. 2002;296:106–110. doi: 10.1016/S0006-291X(02)00787-8. [PubMed] [Cross Ref]
  • Gordon JC, Myers JB, Folta T, Shoja V, Heath LS, Onufriev A. H++: a server for estimating pKas and adding missing hydrogens to macromolecules. Nucleic Acids Research. 2005;33:W368–W371. doi: 10.1093/nar/gki464. [PMC free article] [PubMed] [Cross Ref]
  • Hao YM, Liu QJ, Wang RY, Cao YP, Zhang Y, Zuo LF. Single nucleotide polymorphisms on SHIP2 is associated with type 2 diabetes mellitus in chinese han population. European Review for Medical and Pharmacological Sciences. 2015;19:129–137. [PubMed]
  • Hess B, Kutzner C, van der Spoel D, Lindahl E. GROMACS 4: algorithms for highly efficient, load-balanced, and scalable molecular simulation. Journal of Chemical Theory and Computation. 2008;4:435–447. doi: 10.1021/ct700301q. [PubMed] [Cross Ref]
  • Hoekstra E, Das AM, Willemsen M, Swets M, Kuppen PJ, van der Woude CJ, Bruno MJ, Shah JP, Ten Hagen TL, Chisholm JD, Kerr WG, Peppelenbosch MP, Fuhler GM. Lipid phosphatase SHIP2 functions as oncogene in colorectal cancer by regulating PKB activation. Oncotarget. 2016;7:73525-73540. doi: 10.18632/oncotarget.12321. [PMC free article] [PubMed] [Cross Ref]
  • Holm L, Rosenström P. Dali server: conservation mapping in 3D. Nucleic Acids Research. 2010;38:W545–W549. doi: 10.1093/nar/gkq366. [PMC free article] [PubMed] [Cross Ref]
  • Hsu CY, Uludağ H. A simple and rapid nonviral approach to efficiently transfect primary tissue-derived cells using polyethylenimine. Nature Protocols. 2012;7:935–945. doi: 10.1038/nprot.2012.038. [PubMed] [Cross Ref]
  • Huang YH, Sauer K. Lipid signaling in T-cell development and function. Cold Spring Harbor Perspectives in Biology. 2010;2:a002428 doi: 10.1101/cshperspect.a002428. [PMC free article] [PubMed] [Cross Ref]
  • Huber C, Faqeih EA, Bartholdi D, Bole-Feysot C, Borochowitz Z, Cavalcanti DP, Frigo A, Nitschke P, Roume J, Santos HG, Shalev SA, Superti-Furga A, Delezoide AL, Le Merrer M, Munnich A, Cormier-Daire V. Exome sequencing identifies INPPL1 mutations as a cause of opsismodysplasia. The American Journal of Human Genetics. 2013;92:144–149. doi: 10.1016/j.ajhg.2012.11.015. [PubMed] [Cross Ref]
  • Iida A, Okamoto N, Miyake N, Nishimura G, Minami S, Sugimoto T, Nakashima M, Tsurusaki Y, Saitsu H, Shiina M, Ogata K, Watanabe S, Ohashi H, Matsumoto N, Ikegawa S. Exome sequencing identifies a novel INPPL1 mutation in opsismodysplasia. Journal of Human Genetics. 2013;58:391–394. doi: 10.1038/jhg.2013.25. [PubMed] [Cross Ref]
  • Jefferson AB, Majerus PW. Mutation of the conserved domains of two inositol polyphosphate 5-phosphatases. Biochemistry. 1996;35:7890–7894. doi: 10.1021/bi9602627. [PubMed] [Cross Ref]
  • Jurynec MJ, Grunwald DJ. SHIP2, a factor associated with diet-induced obesity and insulin sensitivity, attenuates FGF signaling in vivo. Disease Models & Mechanisms. 2010;3:733–742. doi: 10.1242/dmm.000703. [PubMed] [Cross Ref]
  • Kabsch W. XDS. Acta Crystallographica. Section D, Biological Crystallography. 2010;66:125–132. doi: 10.1107/S0907444909047337. [PMC free article] [PubMed] [Cross Ref]
  • Kagawa S, Sasaoka T, Yaguchi S, Ishihara H, Tsuneki H, Murakami S, Fukui K, Wada T, Kobayashi S, Kimura I, Kobayashi M. Impact of SRC homology 2-containing inositol 5'-phosphatase 2 gene polymorphisms detected in a japanese population on insulin signaling. The Journal of Clinical Endocrinology & Metabolism. 2005;90:2911–2919. doi: 10.1210/jc.2004-1724. [PubMed] [Cross Ref]
  • Kaisaki PJ, Delépine M, Woon PY, Sebag-Montefiore L, Wilder SP, Menzel S, Vionnet N, Marion E, Riveline JP, Charpentier G, Schurmans S, Levy JC, Lathrop M, Farrall M, Gauguier D. Polymorphisms in type II SH2 domain-containing inositol 5-phosphatase (INPPL1, SHIP2) are associated with physiological abnormalities of the metabolic syndrome. Diabetes. 2004;53:1900–1904. doi: 10.2337/diabetes.53.7.1900. [PubMed] [Cross Ref]
  • Langer G, Cohen SX, Lamzin VS, Perrakis A. Automated macromolecular model building for X-ray crystallography using ARP/wARP version 7. Nature Protocols. 2008;3:1171–1179. doi: 10.1038/nprot.2008.91. [PMC free article] [PubMed] [Cross Ref]
  • Le Coq J, Heredia Gallego L, Lietha D. Expression, purification, crystallisation and X-ray crystallographic analysis of a truncated form of human src homology 2 containing inositol 5-Phosphatase 2. The Protein Journal. 2016;35:225–230. doi: 10.1007/s10930-016-9665-y. [PubMed] [Cross Ref]
  • Lee JO, Yang H, Georgescu MM, Di Cristofano A, Maehama T, Shi Y, Dixon JE, Pandolfi P, Pavletich NP. Crystal structure of the PTEN tumor suppressor: implications for its phosphoinositide phosphatase activity and membrane association. Cell. 1999;99:323–334. doi: 10.1016/S0092-8674(00)81663-3. [PubMed] [Cross Ref]
  • Li H, Marshall AJ. Phosphatidylinositol (3,4) bisphosphate-specific phosphatases and effector proteins: a distinct branch of PI3K signaling. Cellular Signalling. 2015;27:1789–1798. doi: 10.1016/j.cellsig.2015.05.013. [PubMed] [Cross Ref]
  • Lindorff-Larsen K, Piana S, Palmo K, Maragakis P, Klepeis JL, Dror RO, Shaw DE. Improved side-chain torsion potentials for the Amber ff99SB protein force field. Proteins: Structure, Function, and Bioinformatics. 2010;78:1950–1958. doi: 10.1002/prot.22711. [PMC free article] [PubMed] [Cross Ref]
  • Liu Q, Shalaby F, Jones J, Bouchard D, Dumont DJ. The SH2-containing inositol polyphosphate 5-phosphatase, ship, is expressed during hematopoiesis and spermatogenesis. Blood. 1998;91:2753–2759. [PubMed]
  • Lo TC, Barnhill LM, Kim Y, Nakae EA, Yu AL, Diccianni MB. Inactivation of SHIP1 in T-cell acute lymphoblastic leukemia due to mutation and extensive alternative splicing. Leukemia Research. 2009;33:1562–1566. doi: 10.1016/j.leukres.2009.04.032. [PMC free article] [PubMed] [Cross Ref]
  • Luo JM, Yoshida H, Komura S, Ohishi N, Pan L, Shigeno K, Hanamura I, Miura K, Iida S, Ueda R, Naoe T, Akao Y, Ohno R, Ohnishi K. Possible dominant-negative mutation of the SHIP gene in acute myeloid leukemia. Leukemia. 2003;17:1–8. doi: 10.1038/sj.leu.2402725. [PubMed] [Cross Ref]
  • Mahoney MW, Jorgensen WL. A five-site model for liquid water and the reproduction of the density anomaly by rigid, nonpolarizable potential functions. The Journal of Chemical Physics. 2000;112:8910–8922. doi: 10.1063/1.481505. [Cross Ref]
  • Marat AL, Wallroth A, Lo WT, Müller R, Norata GD, Falasca M, Schultz C, Haucke V. mTORC1 activity repression by late endosomal phosphatidylinositol 3,4-bisphosphate. Science. 2017;356:968–972. doi: 10.1126/science.aaf8310. [PubMed] [Cross Ref]
  • McCoy AJ, Grosse-Kunstleve RW, Adams PD, Winn MD, Storoni LC, Read RJ. Phaser crystallographic software. Journal of Applied Crystallography. 2007;40:658–674. doi: 10.1107/S0021889807021206. [PubMed] [Cross Ref]
  • Mills SJ, Persson C, Cozier G, Thomas MP, Trésaugues L, Erneux C, Riley AM, Nordlund P, Potter BV. A synthetic polyphosphoinositide headgroup surrogate in complex with SHIP2 provides a rationale for drug discovery. ACS Chemical Biology. 2012;7:822–828. doi: 10.1021/cb200494d. [PMC free article] [PubMed] [Cross Ref]
  • Mills SJ, Silvander C, Cozier G, Trésaugues L, Nordlund P, Potter BV. Crystal structures of type-ii inositol polyphosphate 5-phosphatase INPP5B with synthetic inositol polyphosphate surrogates reveal new mechanistic insights for the inositol 5-Phosphatase family. Biochemistry. 2016;55:1384–1397. doi: 10.1021/acs.biochem.5b00838. [PMC free article] [PubMed] [Cross Ref]
  • Mol CD, Izumi T, Mitra S, Tainer JA. DNA-bound structures and mutants reveal abasic DNA binding by APE1 and DNA repair coordination [corrected] Nature. 2000;403:451–456. doi: 10.1038/35000249. [PubMed] [Cross Ref]
  • Murshudov GN, Vagin AA, Dodson EJ. Refinement of macromolecular structures by the maximum-likelihood method. Acta Crystallographica Section D Biological Crystallography. 1997;53:240–255. doi: 10.1107/S0907444996012255. [PubMed] [Cross Ref]
  • Nakatsu F, Perera RM, Lucast L, Zoncu R, Domin J, Gertler FB, Toomre D, De Camilli P. The inositol 5-phosphatase SHIP2 regulates endocytic clathrin-coated pit dynamics. The Journal of Cell Biology. 2010;190:307–315. doi: 10.1083/jcb.201005018. [PMC free article] [PubMed] [Cross Ref]
  • Oezguen N, Schein CH, Peddi SR, Power TD, Izumi T, Braun W. A "moving metal mechanism" for substrate cleavage by the DNA repair endonuclease APE-1. Proteins: Structure, Function, and Bioinformatics. 2007;68:313–323. doi: 10.1002/prot.21397. [PubMed] [Cross Ref]
  • Olsen JV, Blagoev B, Gnad F, Macek B, Kumar C, Mortensen P, Mann M. Global, in vivo, and site-specific phosphorylation dynamics in signaling networks. Cell. 2006;127:635–648. doi: 10.1016/j.cell.2006.09.026. [PubMed] [Cross Ref]
  • Ong CJ, Ming-Lum A, Nodwell M, Ghanipour A, Yang L, Williams DE, Kim J, Demirjian L, Qasimi P, Ruschmann J, Cao LP, Ma K, Chung SW, Duronio V, Andersen RJ, Krystal G, Mui AL. Small-molecule agonists of SHIP1 inhibit the phosphoinositide 3-kinase pathway in hematopoietic cells. Blood. 2007;110:1942–1949. doi: 10.1182/blood-2007-03-079699. [PubMed] [Cross Ref]
  • Pesesse X, Deleu S, De Smedt F, Drayer L, Erneux C. Identification of a second SH2-domain-containing protein closely related to the phosphatidylinositol polyphosphate 5-phosphatase SHIP. Biochemical and Biophysical Research Communications. 1997;239:697–700. doi: 10.1006/bbrc.1997.7538. [PubMed] [Cross Ref]
  • Pesesse X, Moreau C, Drayer AL, Woscholski R, Parker P, Erneux C. The SH2 domain containing inositol 5-phosphatase SHIP2 displays phosphatidylinositol 3,4,5-trisphosphate and inositol 1,3,4,5-tetrakisphosphate 5-phosphatase activity. FEBS Letters. 1998;437:301–303. doi: 10.1016/S0014-5793(98)01255-1. [PubMed] [Cross Ref]
  • Prasad NK, Tandon M, Badve S, Snyder PW, Nakshatri H. Phosphoinositol phosphatase SHIP2 promotes cancer development and metastasis coupled with alterations in EGF receptor turnover. Carcinogenesis. 2008;29:25–34. doi: 10.1093/carcin/bgm213. [PubMed] [Cross Ref]
  • Rosen SA, Gaffney PR, Gould IR. A theoretical investigation of inositol 1,3,4,5-tetrakisphosphate. Phys. Chem. Chem. Phys. 2011;13:1070–1081. doi: 10.1039/C0CP00956C. [PubMed] [Cross Ref]
  • Sleeman MW, Wortley KE, Lai KM, Gowen LC, Kintner J, Kline WO, Garcia K, Stitt TN, Yancopoulos GD, Wiegand SJ, Glass DJ. Absence of the lipid phosphatase SHIP2 confers resistance to dietary obesity. Nature Medicine. 2005;11:199–205. doi: 10.1038/nm1178. [PubMed] [Cross Ref]
  • Slochower DR, Huwe PJ, Radhakrishnan R, Janmey PA. Quantum and all-atom molecular dynamics simulations of protonation and divalent ion binding to phosphatidylinositol 4,5-bisphosphate (PIP2) The Journal of Physical Chemistry B. 2013;117:8322–8329. doi: 10.1021/jp401414y. [PMC free article] [PubMed] [Cross Ref]
  • Song MS, Salmena L, Pandolfi PP. The functions and regulation of the PTEN tumour suppressor. Nature Reviews Molecular Cell Biology. 2012;13:283–296. doi: 10.1038/nrm3330. [PubMed] [Cross Ref]
  • Stephens L, Eguinoa A, Corey S, Jackson T, Hawkins PT. Receptor stimulated accumulation of phosphatidylinositol (3,4,5)-trisphosphate by G-protein mediated pathways in human myeloid derived cells. The EMBO Journal. 1993;12:2265–2273. [PubMed]
  • Söding J, Biegert A, Lupas AN. The HHpred interactive server for protein homology detection and structure prediction. Nucleic Acids Research. 2005;33:W244–W248. doi: 10.1093/nar/gki408. [PMC free article] [PubMed] [Cross Ref]
  • Taylor V, Wong M, Brandts C, Reilly L, Dean NM, Cowsert LM, Moodie S, Stokoe D. 5' phospholipid phosphatase SHIP-2 causes protein kinase B inactivation and cell cycle arrest in glioblastoma cells. Molecular and Cellular Biology. 2000;20:6860–6871. doi: 10.1128/MCB.20.18.6860-6871.2000. [PMC free article] [PubMed] [Cross Ref]
  • Trésaugues L, Silvander C, Flodin S, Welin M, Nyman T, Gräslund S, Hammarström M, Berglund H, Nordlund P. Structural basis for phosphoinositide substrate recognition, catalysis, and membrane interactions in human inositol polyphosphate 5-phosphatases. Structure. 2014;22:744–755. doi: 10.1016/j.str.2014.01.013. [PubMed] [Cross Ref]
  • Van Der Spoel D, Lindahl E, Hess B, Groenhof G, Mark AE, Berendsen HJ. GROMACS: fast, flexible, and free. Journal of Computational Chemistry. 2005;26:1701–1718. doi: 10.1002/jcc.20291. [PubMed] [Cross Ref]
  • Vandeput F, Backers K, Villeret V, Pesesse X, Erneux C. The influence of anionic lipids on SHIP2 phosphatidylinositol 3,4,5-trisphosphate 5-phosphatase activity. Cellular Signalling. 2006;18:2193–2199. doi: 10.1016/j.cellsig.2006.05.010. [PubMed] [Cross Ref]
  • Vandeput F, Combettes L, Mills SJ, Backers K, Wohlkönig A, Parys JB, De Smedt H, Missiaen L, Dupont G, Potter BV, Erneux C. Biphenyl 2,3',4,5',6-pentakisphosphate, a novel inositol polyphosphate surrogate, modulates Ca2+ responses in rat hepatocytes. The FASEB Journal. 2007;21:1481–1491. doi: 10.1096/fj.06-7691com. [PubMed] [Cross Ref]
  • Vanhaesebroeck B, Stephens L, Hawkins P. PI3K signalling: the path to discovery and understanding. Nature Reviews Molecular Cell Biology. 2012;13:195–203. doi: 10.1038/nrm3290. [PubMed] [Cross Ref]
  • Walker SM, Leslie NR, Perera NM, Batty IH, Downes CP. The tumour-suppressor function of PTEN requires an N-terminal lipid-binding motif. Biochemical Journal. 2004;379:301–307. doi: 10.1042/bj20031839. [PubMed] [Cross Ref]
  • Wallace AC, Laskowski RA, Thornton JM. LIGPLOT: a program to generate schematic diagrams of protein-ligand interactions. "Protein Engineering, Design and Selection". 1995;8:127–134. doi: 10.1093/protein/8.2.127. [PubMed] [Cross Ref]
  • Whisstock JC, Romero S, Gurung R, Nandurkar H, Ooms LM, Bottomley SP, Mitchell CA. The inositol polyphosphate 5-phosphatases and the apurinic/apyrimidinic base excision repair endonucleases share a common mechanism for catalysis. Journal of Biological Chemistry. 2000;275:37055–37061. doi: 10.1074/jbc.M006244200. [PubMed] [Cross Ref]
  • Whisstock JC, Wiradjaja F, Waters JE, Gurung R. The structure and function of catalytic domains within inositol polyphosphate 5-phosphatases. IUBMB Life. 2002;53:15–23. doi: 10.1080/15216540210814. [PubMed] [Cross Ref]
  • Yip SC, Eddy RJ, Branch AM, Pang H, Wu H, Yan Y, Drees BE, Neilsen PO, Condeelis J, Backer JM. Quantification of PtdIns(3,4,5)P(3) dynamics in EGF-stimulated carcinoma cells: a comparison of PH-domain-mediated methods with immunological methods. Biochemical Journal. 2008;411:441–448. doi: 10.1042/BJ20071179. [PMC free article] [PubMed] [Cross Ref]
  • Yu J, Ryan DG, Getsios S, Oliveira-Fernandes M, Fatima A, Lavker RM. MicroRNA-184 antagonizes microRNA-205 to maintain SHIP2 levels in epithelia. PNAS. 2008;105:19300–19305. doi: 10.1073/pnas.0803992105. [PubMed] [Cross Ref]
  • Zhang D, Aravind L. Identification of novel families and classification of the C2 domain superfamily elucidate the origin and evolution of membrane targeting activities in eukaryotes. Gene. 2010;469:18–30. doi: 10.1016/j.gene.2010.08.006. [PMC free article] [PubMed] [Cross Ref]
2017; 6: e26640.

Decision letter

Roger L Williams, Reviewing editor
Roger L Williams, Medical Research Council, United Kingdom;

In the interests of transparency, eLife includes the editorial decision letter and accompanying author responses. A lightly edited version of the letter sent to the authors after peer review is shown, indicating the most substantive concerns; minor comments are not usually included.

Thank you for submitting your article "Structural basis for interdomain communication in SHIP2 providing high phosphatase activity" for consideration by eLife. Your article has been favorably evaluated by John Kuriyan (Senior Editor) and three reviewers, one of whom is a member of our Board of Reviewing Editors. The following individual involved in review of your submission has agreed to reveal their identity: Evzen Boura (Reviewer #3).

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

Previous work has established the structure of the SHIP2 phosphatase domain and its relationship with other members of the 5-phosphatase family. This manuscript has established the structure of a pptase-C2 domain construct and examined the role of the C2 domain in catalysis. The current study represents a significant advance in understanding SHIP2. The structure shows that the C2 domain makes an extensive interaction surface with the phosphatase domain. A similar interaction of the C2 domain with the catalytic domain has been observed in other lipid-modifying enzymes, such as PLCδ1 and PTEN. The manuscript establishes that the C2 domain acts as an allosteric activator by stabilising the pptase domain and contributing to membrane binding. The manuscript outlines two distinct pathways of stabilisation contributing to the effect of the C2 domain, a polar pathway manifest on activity with the soluble substrate IP4 and a hydrophobic pathway that specifically affects interactions with the acyl chains of PIP3.

The combination of structures and molecular dynamics provides a compelling model for the catalytic cycle of SHIP2 encountering substrate in lipid membranes, with the L4 loop changing conformation between an initial encounter state and a state in which substrate is engaged in the active site. This is an important contribution to our understanding of this family of enzymes. It also provides a detailed picture of how the C2 domain of SHIP2 participates in membrane binding and catalysis.

Several points made by the reviewers (see below) should be taken into account in the revised manuscript. The reviewers have found the allosteric mechanism for the effect of the C2 domain based on crystallography and MD are convincing, but it would be helpful if you could provide a movie that portrays this mechanism. In the revision, you should take account of the suggestions by the reviewers, in particular, in the discussion of an effect on kcat and KM as being indicative of allostery (reviewer 1).

An important concern that has been raised by the reviewers is that the lipid blot demonstration of lipid affinity is not adequate for the work. These blots are subject to a range of artifacts. It would be better if you could measure the affinity using lipid vesicles and the FRET assay suggested by reviewer 3. This assay has been described by a number of workers (see, for example, PMID 22949682). Also, the methods should be revised to make clear exactly what lipids are being used and how the lipids are prepared for the assay. While the reviewers believe that assays with substrate in lipid vesicles would be an informative addition to the manuscript, it may be that this is beyond the scope of the present work, and this should not be regarded as essential.

In an earlier study of several phosphatases (Nordlund and colleagues), the L4 loop of SHIP2 was referred to as the P4-interacting motif (P4IM). It would be helpful for readers familiar with the previous study if the current manuscript also at least mentions this nomenclature.

It is claimed at the end of the manuscript that the cellular assays presented show that the C2 domain has an additional role, likely in membrane localization. This statement should be clarified, or if the statement is not based on the results, this should be explained.

The manuscript claims that "both, the Ptase and C2 domains of SHIP2 bind PS in a Ca2+ independent manner" This is misleading. The C2 domain on its own was never tested. It could be that the interaction depends entirely on the pptase domain, but that the pptase is only oriented and stabilised by the C2 domain. The statement consistent with the results would be "both the Ptase only and the Ptase-C2 constructs bind PS in a Ca2+-independent manner."

Common membrane-binding mechanisms are asp/glu residues binding Ca2+, which in turn binds to membranes or lys/arg residues that bind directly to the negatively charged lipids. Figure 2B shows that the SHP2 residues in the key positions are ser or asp/glu residues, but Figure 2C shows that Ca2+ does not affect binding. It would be helpful if you briefly discuss this point.

For assistance to color blind readers, it would be helpful if the explanatory keys in Figure 6 are placed next to the curve that they describe, or that there is a line between the curve and the key. There are shapes for markers on the curves in the figure, but these are typically small in the final version and may not be clear on their own.

In Figure 7, since the ratio pAkt/Akt is not the directly observed quantity, the manuscript should make clear how the errors in this ratio were determined. Error propagation in the ratio based on the standard deviations of the two observed quantities should have been carried out. The reviewers are concerned that differences in activity are complicated by differences in expression. To allay these concerns, the statistical methods should be described in the Materials and methods.

The WT X-ray dataset appears to be the same data set that was described in the earlier crystallisation report, however, the I/σ and highest resolution shell limits are not the same. It appears that the data have been reprocessed, and this is not surprising. However, it should be made clear how the resolution limits were chosen. Since the I/σ for the highest-resolution shells differ for the various datasets, it does not seem that this was the decisive criterion. The correlation coefficients cc1/2 and cc* have been proposed as a criterion to help make resolution cutoff in a statistically reliable manner. The table should also include cc1/2.

Reviewer #1:

Previous work has established the structure of the SHIP2 phosphatase domain and its relationship with other members of the 5-phosphatase family. The current study by Lietha and colleagues has established the structure of pptase-C2 domain construct and examined the role of the C2 domain in catalysis. A previous report by the same group described crystallisation and diffraction of the pptase-C2 construct and demonstrated its enzymatic activity. The current study represents a significant advance in understanding this enzyme. The structure shows that the C2 domain makes an extensive interaction surface with the phosphatase domain. A similar interaction of the C2 domain with the catalytic domain has been observed in other lipid-modifying enzymes, such as PLCδ1 and PTEN. The authors establish that the C2 domain acts as an allosteric activator by stabilising the pptase domain and contributing to membrane binding. The authors are suggesting two distinct pathways of stabilisation contributing to the effect of the C2 domain, a polar pathway manifest on activity with the soluble substrate IP4 and a hydrophobic pathway that specifically affects interactions with the acyl chains of PIP3.

The study provides a compelling model for the catalytic cycle of SHIP2 encountering substrate in lipid membranes, with the L4 loop changing conformation between an initial encounter state and a state in which substrate is engaged in the active site. This is an important contribution to our understanding of this family of enzymes. It also provides a detailed picture of how the C2 domain of SHIP2 participates in membrane binding and catalysis. This is particularly useful, since a similar catalytic domain/C2 module is present in other lipid-modifying enzymes.

In a previous study by Nordlund and colleagues, structures of SHIP2, OCRL, and INPP5B pptase domains were determined and compared. In this previous study, it was shown that a motif referred to as the P4-interacting motif (P4IM) interacts with 3- and 4-phosphates of the substrates. In the manuscript of Lietha and colleagues, the same feature is referred to as loop L4. To simplify things for the readers, it would be helpful if the authors of the current manuscript use the P4IM notation or at least point out that this is contained in L4. The presence of the C2 domain results in a large change in the conformation of the L4 loop, and the authors propose that this change likely constitutes an important component of the catalytic cycle of SHIP2 on membranes.

In the first paragraph of the subsection “The C2 domain affects SHIP2 activity” it states that changing kcat without much change in KM suggests that the activation is allosteric. The most common definition of allosteric is an effect on activity due to interaction outside the catalytic site. This can be an effect on KM (K-type allostery), or it can be an effect on Vmax (V-type), or come combination of these. However, simply an effect on kcat does not suggest an influence directed from outside the active site. A kcat-only effect could be obtained by forming additional interactions with the transition state, but not with the substrate, but this would not be an allosteric effect. What the authors mean is that if the C2 domain has any influence at all, it must be allosteric, since the C2 domain is not near the active site. This should be restated. There is also a grammatic problem with the sentence in its current form (should be "suggests" instead of "suggesting").

In the last paragraph of the subsection “The C2 domain affects SHIP2 activity”, describing Figure 3G, it is not clear what the authors have done in the experiments that they describe as adding "PC" and "PS". What exactly are the lipids that they have used? If they are synthetic, they should state what the acyl chains are. If they are from a natural source, they should quote the dominant component of the source. It is particularly important the length and saturation of the acyl chains. Also, there is no description of how the lipid substrates were presented. Were they prepared as unilamellar vesicles? Were the multilammelar vesicles? Were they SUVS? LUVS? GUVS? The reason that this is important is that the authors are using a di-C8 PIP3 lipid that is soluble on its own, so it is not clear that this lipid would effectively and completely partition into membranes of lipid vesicles. If the lipid substrate is both in the bulk phase and in the membranes, it makes interpretation of the results more complex. Ideally, it would be helpful if the authors had used long-chain PIP3 incorporated into unilamellar vesicles for at least some of their assays, in order to strengthen their conclusions for the behavior of SHIP2 in cells.

In the aforementioned paragraph, it is proposed that PS orients the enzyme on PIP3-containing membranes. This is an interesting proposition that could be tested if the substrate were PIP3 in vesicles, however, the authors are using di-C8 PIP3, which is soluble and not necessarily entirely present in membranes. It is hard to assess what is happening in Figure 3G, which is intended to show the effect of PS. In Figure 3H, Both the pptase and pptase-C2 constructs are activated by PS, so there is little indication in this that the C2 domain is contributing to the activation by PS. Figure 2C suggests that the pptase-C2 construct binds better to PS than the pptase domain alone. However, from 3G, this does not translate to PS-enhanced activity. This is perhaps not surprising, since the di-C8 PIP3 substrate is soluble and perhaps not fully incorporated into the undescribed PS-containing membranes. This experiment would be far more informative if the study were using PIP3/PC/PS unilamellar vesicles with long-chain fatty acids on all of the lipids. I would not regard these vesicle assays as essential for the paper, but they would enhance the study.

In the last paragraph of the subsection “Probing the communication path by mutagenesis”, the authors state that their cellular results confirm that C2 interactions observed in the crystal structure are important for SHIP2 activity in cells and that the C2 domain has an additional role, likely in membrane localization. It is not clear how these results show that the C2 domain has an additional role in cells. This should be clarified or the statement removed.

Reviewer #2:

These authors examine interdomain regulation for the SHP2 phosphatase. In particular, how does the C2 domain affect the phosphatase domain? They report a number of x-ray crystal structures as well as circular dichroism spectra and phosphatase kinetics. Finally, they refine the results with molecular dynamics simulations.

They find two conformations: loop L4 in and out. They propose that the C2 domain influences this conformational change. They do a good job with a schematic for their mechanism in Figure 8. This study includes a very large number of experiments and addresses an important question. However, there are a number of points that could be improved. My suggestions are as follow:

1) Figure 2. I found this figure confusing. The most common mechanisms of binding negatively charged lipids are asp/glu residues binding Ca2+ or lys/arg residues that directly to the negatively charged lipids. Figure 2B shows that the SHP2 residues in the key positions are ser or asp/glu residues. There are no positively charged amino acids so one would predict a role for Ca2+. But Figure 2C shows that Ca2+ does not affect binding. This says to me that Figure 2B is misleading. The binding site must be elsewhere in the C2 domain.

2) Figure 6—figure supplement 1. I am partly color blind. It would be convenient if the figure legend identified the mutants that have different spectra.

3) Figure 7. This experiment examines two mutants in HEK293 cells. This figure surprised me in two ways. The recombinant SHIP2 is a dark band. The endogenous SHIP2 is not detected in the control lane so the SHIP2 is grossly over expressed but there is only a 40% decrease in pAkt. Second, the deletion of the C2 domain clearly reduces activity, but the FLDD mutation has only a minute affect, which they say is statistically significant. They do not explain the statistical analysis. How was the error for pAkt propagated with the error in Akt? I am just not convinced that the small effect of the FLDD mutation is real and not due to error or differences in expression. They have a supplemental figure for expression. The FLDD is slightly lower but they say that it is not statistically significant. I need a little bit more convincing, and the statistics should be in Materials and methods.

Figure 8. I liked this schematic. Very helpful.

Reviewer #3:

Le Coq et al. Report the crystal structure of a SHIP2 construct that, unlike previously reported structures, contains the membrane targeting C2 domain. The reported structural analysis together with molecular dynamics simulations reveals how the C2 domain (and thus membrane binding) allosterically controls the enzymatic activity of SHIP2. The findings are important, most of experiments are well done, manuscript is well written and the reasoning is sound and square. The manuscript clearly deserves to be published in eLife. However, there is one major issue (and several small technical issues) that should be improved before final acceptance. The additional control experiments should not take more than a day or two of a skilled experimentalist.

1) A protein lipid overlay assay is a useful screening technique, however, no conclusions should be derived from this method as it is notoriously known to be artifact prone. The authors need to verify their results using liposomes with/out PS and with/out Ca2+. Many reliable standard assays exist. Since SHIP2 contains trypthophan residues I suggest to use a FRET based assay (W residues as donors and dansyl-PE as acceptors) to measure binding.

2) Biochemical (enzymatic) analysis is done using soluble substrates only. To explore the effect of allosteric regulation these data need to be compared with membrane embedded substrates (liposomes of different lipid composition containing PI(3,4,5,)P3).

Several points made by the reviewers (see below) should be taken into account in the revised manuscript. The reviewers have found the allosteric mechanism for the effect of the C2 domain based on crystallography and MD are convincing, but it would be helpful if you could provide a movie that portrays this mechanism.

We now provide 5 videos:

Video 1 and 2 show the 4 principal motions (eigenvectors) extracted from principal component analysis of the MD simulations of the Ptase and Ptase-C2.

Video 3 and 4 show substrate interactions during a part of the simulations with Ptase and Ptase-C2.

Video 5 illustrates the catalytic cycle presented in Figure 9B and the conformational changes between L4-out and L4-in conformations.

In the revision, you should take account of the suggestions by the reviewers, in particular, in the discussion of an effect on kcat and KM as being indicative of allostery (reviewer 1).

We have corrected our statement according to the comment of reviewer 1 (see below, under replies to reviewer #1”).

An important concern that has been raised by the reviewers is that the lipid blot demonstration of lipid affinity is not adequate for the work. These blots are subject to a range of artifacts. It would be better if you could measure the affinity using lipid vesicles and the FRET assay suggested by reviewer 3. This assay has been described by a number of workers (see, for example, PMID 22949682).

As requested, we performed binding experiments with lipid vesicles. We initially tested the FRET assay suggested by reviewer 3. Under replies to reviewer #3 we explain and show data indicating that this assay is not suitable to measure SHIP interactions with PS vesicles.

We therefore performed lipid vesicle interaction studies by surface plasmon resonance using a Biacore X100 instrument and a L1 sensor chip with immobilized PS vesicles. These data are shown in Figures 2C-D of the revised manuscript and the lipid blot data is now presented as Figure 2—figure supplement 1.

Also, the methods should be revised to make clear exactly what lipids are being used and how the lipids are prepared for the assay.

We apologize for not specifying the type of lipids used in activity assays. In the previous submission we used soluble short chain lipids PS-diC8 or PC-diC8. This is now clearly stated in the main text (subsection “The C2 domain affects SHIP2 activity”, last paragraph), Figure 4A-B and its legend and the Materials and methods section. In the revised manuscript we further added experiments with substrate and PS incorporated in vesicles (Figure 4C-D, see point below).

While the reviewers believe that assays with substrate in lipid vesicles would be an informative addition to the manuscript, it may be that this is beyond the scope of the present work, and this should not be regarded as essential.

We agree that this provides an informative addition and have performed activity assays with substrate in vesicles (Figure 4C-D). These data show that the C2 domain also enhances SHIP2 activity on vesicles and that PS incorporated in PIP3 vesicles has a larger activating effect than soluble PS-diC8.

In an earlier study of several phosphatases (Nordlund and colleagues), the L4 loop of SHIP2 was referred to as the P4-interacting motif (P4IM). It would be helpful for readers familiar with the previous study if the current manuscript also at least mentions this nomenclature.

In the fifth paragraph of the Discussion we now refer to the P4IM terminology. However, since 2 residues of the P4IM are located outside L4 (for details see replies to reviewer #1”) and because Arg682 in SHIP2 likely has an important role in P3 substrate interactions, we believe it would be misleading to use the P4IM terminology for SHIP throughout the manuscript.

It is claimed at the end of the manuscript that the cellular assays presented show that the C2 domain has an additional role, likely in membrane localization. This statement should be clarified, or if the statement is not based on the results, this should be explained.

This statement was based on the observation that deleting the C2 domain has a larger cellular effect than the FLDD mutations. However, we agree this statement could be misleading. We now simply state: “This confirms that C2 interactions observed in the crystal structure are important for SHIP2 activity in cells.”

The manuscript claims that "both, the Ptase and C2 domains of SHIP2 bind PS in a Ca2+ independent manner" This is misleading. The C2 domain on its own was never tested. It could be that the interaction depends entirely on the pptase domain, but that the pptase is only oriented and stabilised by the C2 domain. The statement consistent with the results would be "both the Ptase only and the Ptase-C2 constructs bind PS in a Ca2+-independent manner."

We have rewritten the statement as suggested (subsection “The C2 domain of SHIP2”).

Common membrane-binding mechanisms are asp/glu residues binding Ca2+, which in turn binds to membranes or lys/arg residues that bind directly to the negatively charged lipids. Figure 2B shows that the SHP2 residues in the key positions are ser or asp/glu residues, but Figure 2C shows that Ca2+ does not affect binding. It would be helpful if you briefly discuss this point.

New SPR data have in fact revealed that there is a weak increase of 10% in Ptase-C2 binding to PS in presence of Ca2+ (Figure 2D). Firstly, we now show in Figure 2B both, acidic and basic residues potentially involved in lipid binding. Further, in the subsection “The C2 domain of SHIP2”, we now discuss that the interaction appears to be dominated by direct binding to basic residues, but the small Ca2+ effect we observe could indicate that the weakly conserved acidic residues and serines on CBL1 and 3 of SHIP2 potentially make a small Ca2+ dependent contribution to PS binding.

For assistance to color blind readers, it would be helpful if the explanatory keys in Figure 6 are placed next to the curve that they describe, or that there is a line between the curve and the key.

There are shapes for markers on the curves in the figure, but these are typically small in the final version and may not be clear on their own.

From the specific comments of reviewer 2 we understand that the comment concerns Figure 6—figure supplement 1 (now Figure 7—figure supplement 1, according to new numbering). As requested by reviewer 2, we have now clearly marked and labelled the CD curves that show a different behavior.

In Figure 7, since the ratio pAkt/Akt is not the directly observed quantity, the manuscript should make clear how the errors in this ratio were determined. Error propagation in the ratio based on the standard deviations of the two observed quantities should have been carried out. The reviewers are concerned that differences in activity are complicated by differences in expression. To allay these concerns, the statistical methods should be described in the Materials and methods.

We now describe in the Materials and methods section the statistical method used. We do not apply error propagation of the pAkt and total Akt measurements, because the two quantities are not “uncorrelated”, but rather are used for normalisation (see replies to reviewer #2 for more details).

In order to show higher significance that is not related to differences in SHIP expression, we performed 5 additional experiments (each in triplicates), therefore the data we now present in Figure 8 is from 8 independent experiments, each in triplicates (n=24). Taking all data together we still find significance for pAkt/Akt levels, but no significance for differences in SHIP expression levels.

The WT X-ray dataset appears to be the same data set that was described in the earlier crystallisation report, however, the I/σ and highest resolution shell limits are not the same. It appears that the data have been reprocessed, and this is not surprising. However, it should be made clear how the resolution limits were chosen. Since the I/σ for the highest-resolution shells differ for the various datasets, it does not seem that this was the decisive criterion. The correlation coefficients cc1/2 and cc* have been proposed as a criterion to help make resolution cutoff in a statistically reliable manner. The table should also include cc1/2.

We apologize for this and thank the reviewers for pointing this out. Indeed, the data were reprocessed since it was published in the Protein Journal. Different definitions of I/σ in older versions of scala (used to cut data) and newer versions of aimless (reported in the table) caused the high I/σ values (see replies to reviewer #3 for a more detailed explanation). We have now reprocessed the data to higher resolution and cut the data at a mean (I/sd), reported in aimless, of ~2.

CC(1/2) values are now included in Table 1.

Reviewer #1:

[…] In a previous study by Nordlund and colleagues, structures of SHIP2, OCRL, and INPP5B pptase domains were determined and compared. In this previous study, it was shown that a motif referred to as the P4-interacting motif (P4IM) interacts with 3- and 4-phosphates of the substrates. In the manuscript of Lietha and colleagues, the same feature is referred to as loop L4. To simplify things for the readers, it would be helpful if the authors of the current manuscript use the P4IM notation or at least point out that this is contained in L4. The presence of the C2 domain results in a large change in the conformation of the L4 loop, and the authors propose that this change likely constitutes an important component of the catalytic cycle of SHIP2 on membranes.

In Tresaugues et al. (2014) the P4-interacting-motif (P4IM) is defined to be composed of the INPP5B residues Tyr502, Lys503, Arg518, and Lys516. Of these, only Arg518 (Asn684 in SHIP2) and Lys516 (Arg682 in SHIP2) are contained within loop L4. As suggested we point out that residues Asn684 and Arg682 in SHIP2 correspond to residues in the P4IM (Discussion, fifth paragraph). However, since 2 residues of the P4IM are located outside L4 and because Arg682 in SHIP2 likely has an important role in P3 substrate interactions, we believe it would be misleading to use the P4IM terminology in SHIP throughout the manuscript.

In the first paragraph of the subsection “The C2 domain affects SHIP2 activity” it states that changing kcat without much change in KM suggests that the activation is allosteric. The most common definition of allosteric is an effect on activity due to interaction outside the catalytic site. This can be an effect on KM (K-type allostery), or it can be an effect on Vmax (V-type), or come combination of these. However, simply an effect on kcat does not suggest an influence directed from outside the active site. A kcat-only effect could be obtained by forming additional interactions with the transition state, but not with the substrate, but this would not be an allosteric effect. What the authors mean is that if the C2 domain has any influence at all, it must be allosteric, since the C2 domain is not near the active site. This should be restated. There is also a grammatic problem with the sentence in its current form (should be "suggests" instead of "suggesting").

We thank the reviewer for pointing this out. We now simply state “The fact that the distant C2 domain has specific and differential effects on catalysis of the two substrates, suggests the presence of an allosteric communication between the domain interface and the active site.”

In the last paragraph of the subsection “The C2 domain affects SHIP2 activity”, describing Figure 3G, it is not clear what the authors have done in the experiments that they describe as adding "PC" and "PS". What exactly are the lipids that they have used? If they are synthetic, they should state what the acyl chains are. If they are from a natural source, they should quote the dominant component of the source. It is particularly important the length and saturation of the acyl chains. Also, there is no description of how the lipid substrates were presented. Were they prepared as unilamellar vesicles? Were the multilammelar vesicles? Were they SUVS? LUVS? GUVS? The reason that this is important is that the authors are using a di-C8 PIP3 lipid that is soluble on its own, so it is not clear that this lipid would effectively and completely partition into membranes of lipid vesicles. If the lipid substrate is both in the bulk phase and in the membranes, it makes interpretation of the results more complex.

We apologize for not specifying the type of lipids used. We now state in the main text of the Results section (subsection “The C2 domain affects SHIP2 activity”, last paragraph), Figure 4A-B and its legend and the Materials and methods section that initially we used soluble PS-diC8 and PC-diC8 in these experiments. The revised manuscript now also includes data with PS vesicles (new Figure 4C-D, see point below).

Ideally, it would be helpful if the authors had used long-chain PIP3 incorporated into unilamellar vesicles for at least some of their assays, in order to strengthen their conclusions for the behavior of SHIP2 in cells.

As suggested, we performed additional experiments with long-chain PIP3 embedded in vesicles containing 30% PS or PC. The new data are shown in Figure 4C-D and show that the C2 domain also enhances activity of SHIP in vesicles and this effect is abrogated by FLDD interface mutants. The effect of PS inclusion in vesicles is described in the point below.

In the aforementioned paragraph, it is proposed that PS orients the enzyme on PIP3-containing membranes. This is an interesting proposition that could be tested if the substrate were PIP3 in vesicles, however, the authors are using di-C8 PIP3, which is soluble and not necessarily entirely present in membranes. It is hard to assess what is happening in Figure 3G, which is intended to show the effect of PS. In Figure 3H, Both the pptase and pptase-C2 constructs are activated by PS, so there is little indication in this that the C2 domain is contributing to the activation by PS. Figure 2C suggests that the pptase-C2 construct binds better to PS than the pptase domain alone. However, from 3G, this does not translate to PS-enhanced activity. This is perhaps not surprising, since the di-C8 PIP3 substrate is soluble and perhaps not fully incorporated into the undescribed PS-containing membranes. This experiment would be far more informative if the study were using PIP3/PC/PS unilamellar vesicles with long-chain fatty acids on all of the lipids. I would not regard these vesicle assays as essential for the paper, but they would enhance the study.

As explained above, we have now performed such studies and show the results in Figure 4C-D. Indeed, when PS is incorporated in vesicles together with the PIP3 substrate, the PS effect is larger than with soluble PS-diC8 and PIP3-diC8. This supports the idea that PS in membranes can orient the SHIP2 enzyme, however, this does not appear to be solely achieved by binding of the C2 domain to PS, since PS also activates the Ptase domain alone. This is consistent with a significant PS binding of the Ptase observed in Figure 2C and Figure 2—figure supplement 1.

In the last paragraph of the subsection “Probing the communication path by mutagenesis”, the authors state that their cellular results confirm that C2 interactions observed in the crystal structure are important for SHIP2 activity in cells and that the C2 domain has an additional role, likely in membrane localization. It is not clear how these results show that the C2 domain has an additional role in cells. This should be clarified or the statement removed.

We agree this statement could be misleading. It was based on cell biology data showing that deletion of C2 has a much stronger effect than the FLDD mutation. However, new data presented in Figure 4C-D suggest that membrane localization (at least via PS) is not solely due to the C2 domain. Possibly, the intrinsic activating effect of FLDD mutations could contribute to the difference between the DC2 and FLDD mutants in cellular experiments. To avoid any confusion, we now simply state: “This confirms that C2 interactions observed in the crystal structure are important for SHIP2 activity in cells.”

Reviewer #2:

[…] This study includes a very large number of experiments and addresses an important question. However, there are a number of points that could be improved. My suggestions are as follow:

1) Figure 2. I found this figure confusing. The most common mechanisms of binding negatively charged lipids are asp/glu residues binding Ca2+ or lys/arg residues that directly to the negatively charged lipids. Figure 2B shows that the SHP2 residues in the key positions are ser or asp/glu residues. There are no positively charged amino acids so one would predict a role for Ca2+. But Figure 2C shows that Ca2+ does not affect binding. This says to me that Figure 2B is misleading. The binding site must be elsewhere in the C2 domain.

We agree and thank the reviewer for pointing this out. In the revised manuscript we show in Figure 2B both, the acidic residues in CBL1 and 3 and basic residues on C2 and Ptase domains expected to face the membrane. New SPR data have in fact revealed that there is a weak increase in Ptase-C2 binding to PS in presence of Ca2+ (new Figure 2D), an effect we don’t observe for the Ptase. In the subsection “The C2 domain of SHIP2”, we now discuss that the PS interaction appears to be dominated by direct interactions with basic residues, but possibly the weakly conserved acidic residues and serines on CBL1 and 3 of SHIP2 could be responsible for the small Ca2+ dependent contribution to PS binding we observe.

2) Figure 6—figure supplement 1. I am partly color blind. It would be convenient if the figure legend identified the mutants that have different spectra.

We have clearly labelled the curves that show a different behavior. According to the new numbering this is now in Figure 7—figure supplement 1.

3) Figure 7. This experiment examines two mutants in HEK293 cells. This figure surprised me in two ways. The recombinant SHIP2 is a dark band. The endogenous SHIP2 is not detected in the control lane so the SHIP2 is grossly over expressed but there is only a 40% decrease in PAkt.

Actually, we do detect a very weak band corresponding to endogenous SHIP2 in the control lane. However, since SHIP expression with the pOPINJ vector adds an N-terminal GST tag (this is now clearly stated in the Materials and methods section and legend of Figure 8—figure supplement 1), endogenous SHIP2 runs at a lower molecular weight, close to that of the SHIP2 DC2 construct. Nevertheless, it is correct that native SHIP levels appear low and the transfected constructs are strongly overexpressed.

Regarding the limited decrease in pAkt, we believe there could be several reasons explaining this behavior. As described in the Introduction, the effect of SHIP on Akt phosphorylation is complex and likely depends on several other factors, including: 1) Whether the SHIP product PI(3,4)P2 is efficiently further degraded to PI(3)P (since PI(3,4)P2 has still partial activity towards Akt); 2) What the relative contributions are of SHIP and PTEN in PIP3 degradation under the chosen condition and cellular context and how these activities compare to PI3K activity generating PIP3; 3) How efficiently is the overexpressed SHIP2 targeted to the membrane under the experimental conditions? We believe any of these factors could cause a limited effect of the expressed SHIP proteins on reducing Akt phosphorylation.

Second, the deletion of the C2 domain clearly reduces activity, but the FLDD mutation has only a minute affect, which they say is statistically significant. They do not explain the statistical analysis. How was the error for PAkt propagated with the error in Akt?

We do not apply error propagation of the pAkt and total Akt measurements, because the two quantities are not “uncorrelated”. The typical error propagation for a ratio Q according to the formula:

(in our case Q=pAkt/Akt; δ(X)=error in measurement of X, pAkt or Akt)

assumes that the quantities are uncorrelated. However, in our case dividing by total Akt is a “normalization” of the phospho-Akt signal and only the ratio is expected to have similar values, not necessarily the individual values. To our knowledge, and after talking to several cell biologists in our institute, it is standard practice in cell biology to calculate errors on the ratio of normalized quantities, without error propagation.

I am just not convinced that the small effect of the FLDD mutation is real and not due to error or differences in expression. They have a supplemental figure for expression. The FLDD is slightly lower but they say that it is not statistically significant. I need a little bit more convincing, and the statistics should be in Materials and methods.

In order to show higher significance that is not related with differences in SHIP expression, we performed an additional 5 independent experiments, each in triplicates (i.e. total 8 experiments, n=24). Admittedly, the only partial reduction in pAkt levels upon SHIP expression limits the window showing clear significance and the FLDD mutations having only a partial reduction in activity further limits the significance. Nevertheless, after 8 triplicate experiments (n=24), our results do still show significance between WT and FLDD mutants. Taking all experiments together the total SHIP expression levels are no longer smaller for the FLDD mutant (actually slightly higher, but without significance), making it unlikely that differences in pAkt levels are due to different SHIP expression levels. We would also like to point out that in the revised manuscript we include activity data with PIP3 embedded in vesicles (Figure 4C-D) that further support that FLDD mutations affect SHIP activity in a native membrane environment.

We now clearly state in the Materials and methods section how we analyze the statistics for cell biology experiments, i.e. that standard errors are calculated on values of pAkt/Akt ratios and significance is calculated using a two-tailed unpaired Student t test assuming Gaussian distribution.

Reviewer #3:

Le Coq et al. Report the crystal structure of a SHIP2 construct that, unlike previously reported structures, contains the membrane targeting C2 domain. The reported structural analysis together with molecular dynamics simulations reveals how the C2 domain (and thus membrane binding) allosterically controls the enzymatic activity of SHIP2. The findings are important, most of experiments are well done, manuscript is well written and the reasoning is sound and square. The manuscript clearly deserves to be published in eLife. However, there is one major issue (and several small technical issues) that should be improved before final acceptance. The additional control experiments should not take more than a day or two of a skilled experimentalist.

1) A protein lipid overlay assay is a useful screening technique, however, no conclusions should be derived from this method as it is notoriously known to be artifact prone. The authors need to verify their results using liposomes with/out PS and with/out Ca2+. Many reliable standard assays exist. Since SHIP2 contains trypthophan residues I suggest to use a FRET based assay (W residues as donors and dansyl-PE as acceptors) to measure binding.

As requested, we performed binding experiments with lipid vesicles. We initially tested the suggested FRET assay using energy transfer between excited tryptophans and dansyl-PE. However, instead of a FRET induced increase at the dansyl emission in presence of protein, we observed a decrease (see Author response image 1). The reason for is not exactly clear to us, but appears to result from quenching of liposome fluorescence by protein (apparently also observed at low level in Baskaran et al. (2012) Mol Cell 47, 339-48). As a result, this assay is not suitable to measure SHIP interactions with PS vesicles.

Author response image 1.
Tryptophan-dansyl FRET assay.

In an alternative approach, we performed lipid vesicle interaction studies by surface plasmon resonance (SPR) using a Biacore X100 instrument and a L1 sensor chip with immobilized PS and PC vesicles. These data are shown in Figure 2C-D of the revised manuscript and the lipid blot data is now presented as Figure 2—figure supplement 1. Shown are SPR responses resulting from differences between the active flow cell coated with 30% (mol/mol) PS vesicles and the reference flow cell containing 100% PC vesicles. These data confirm a stronger interaction with Ptase-C2 comparted to the isolated Ptase, although the difference is smaller compared to PLO experiments. Unfortunately, we were not able to extract dissociation constants or kinetic kon/koff values from SPR sensorgrams. Binding and dissociation appear to follow non-trivial kinetics, which we were not able to fit with standard models. On the other hand, plotting steady-state values versus protein concentration suggests that KD values are significantly above the highest protein concentration used (10µM).

Injecting higher protein concentrations caused irreversible deposition of protein on the surface making responses unreliable. Interestingly, using injections at 5 µM protein we detected a weak but significant effect of Ca2+ only with Ptase-C2, increasing the binding by ~10%. We do not claim that this modest effect is biologically significant, but as discussed in the subsection “The C2 domain of SHIP2”, this effect might result from the low conservation of Ca2+ binding residues in CBL1 and 3 of the SHIP2 C2 domain.

2) Biochemical (enzymatic) analysis is done using soluble substrates only. To explore the effect of allosteric regulation these data need to be compared with membrane embedded substrates (liposomes of different lipid composition containing PI(3,4,5,)P3).

As suggested, we performed enzyme activity assays with vesicles containing PIP3. These data are presented in Figure 4C-D of the revised manuscript. The data show that the C2 domain also promotes SHIP2 activity on vesicles and this effect is removed by the FLDD interface mutants. We further show that PS in vesicles promotes SHIP activity in an independent and additive manner to the C2 effect.


Articles from eLife are provided here courtesy of eLife Sciences Publications, Ltd