Preclinical Evaluation of a Novel SHIP1 Phosphatase Activator for Inhibition of PI3K Signaling in Malignant B Cells

PI3K are key drivers of survival and growth of normal B cells (1). PI3K catalyze the phosphorylation of phosphatidylinositol-4,5-bisphosphate (PI(4,5)P₂) to generate phosphatidylinositol-3,4,5-triphosphate (PI(3,4,5)P₃) which serves as a docking site at the plasma membrane for various pleckstrin homology domain–containing proteins, including phospholipase Cγ2 (PLCγ2), Bruton's tyrosine kinase (BTK), and AKR mouse thymoma kinase (AKT). Once recruited, these effectors become fully activated and drive downstream signaling leading to cell growth and survival.

PI3K activity is countered by the phosphatases PTEN and Src homology domain 2-containing inositol-5′-phosphatase (SHIP)1/2. Whereas PTEN directly reverses PI3K activity to generate PI(4,5)P₂, SHIP1/2 catalyze conversion of PI(3,4,5)P₃ to phosphatidylinositol-3,4-bisphosphate (PI(3,4)P₂; ref. 2). PI(3,4)P₂ has second messenger activity but, in contrast to PI(3,4,5)P₃, appears to act predominantly to suppress signaling (3). In addition to their catalytic functions, SHIP1/2 may also modulate signaling by acting as scaffold proteins (2).

Multiple signaling pathways converge on PI3Ks in various subtypes of B-cell neoplasms, including signaling downstream of the B-cell receptor (BCR), a major driver and therapeutic target for malignant B cells (4). For example, PI3K is activated in diffuse large B-cell lymphoma (DLBCL) cells due to chronic or tonic BCR signaling (5) and, in chronic lymphocytic leukemia (CLL) and follicular lymphoma (FL) cells, following engagement of the BCR by (auto)antigen or environmental lectins, respectively (6–8). In CLL, BCR signaling capacity (including activation of PI3K) is most strongly retained in samples with unmutated IGHV genes (U-CLL) compared with samples with mutated IGHV genes (M-CLL) and associates with poor patient outcomes (9, 10). PI3K is also activated in malignant B cells downstream of other receptors, including chemokine receptors (11, 12). Finally, expression of PTEN is frequently downregulated through a variety of mechanisms in GCB-DLBCL (13).

The frequent activation of the PI3K pathway in B-cell neoplasms has led to the clinical development of PI3K inhibitors (PI3Ki). The PI3Kδ isoform–specific inhibitor idelalisib is approved for treatment of CLL, FL, and small lymphocytic lymphoma (14). Although idelalisib can be highly effective, its use is limited by substantial toxicity, especially in the frontline setting (15, 16). Moreover, idelalisib is relatively ineffective as a monotherapy in DLBCL. This lack of effectiveness appears to be due to activation of additional PI3K isoforms because dual inhibition of PI3Kα and δ is associated with increased cytotoxicity of DLBCL cell lines compared with PI3Kδ inhibition alone (2, 17). Responses can be enhanced using more broadly acting PI3Ki, such as the PI3Kδ/γ inhibitor duvelisib and the PI3Kα/δ inhibitor copanlisib, or by combinations of idelalisib with other targeted agents, including the BTK inhibitor, ibrutinib (17–21). However, there remains a clear need for novel strategies to target PI3K signaling, ideally combining broad activity against multiple subtypes of B-cell neoplasms with low toxicity, and with potential for effective combination with other targeted agents.

Here, we have investigated the effects of pharmacologic activation of SHIP1 as a novel strategy for inhibition of BCR-induced PI3K signaling in malignant B cells. BCR stimulation results in activation of SHIP1 as part of a negative feedback response to limit signaling (22, 23). Thus, deletion of Inppd5 (encoding SHIP1) in mouse B cells results in hyperactive BCR signaling (24). Enhanced BCR signaling in SHIP1-deficient animals appears to be a consequence of both increased PI(3,4)P₂ and reduced accumulation of PI(3,4,5)P₃ because functional inactivation of Plekha1 and Plekha2, which encode effectors that bind the SHIP1 product PI(3,4)P₂, has a similar phenotype to deletion of Inppd5 (3). Activation of SHIP1 following BCR stimulation is mediated by induced phosphorylation and recruitment to the plasma membrane. This includes SYK-dependent phosphorylation of SHIP1 at Y¹⁰²² which then mediates binding to the BCR via the Dok-3 adaptor protein (2, 23, 25–27). Importantly, SHIP1 is amenable to pharmacologic activation using chemical agonists (28). For example, AQX-MN100, a synthetic derivative of the natural product SHIP1 activator pelorol, induces allosteric activation of SHIP1 by binding to the C2 domain of SHIP1 (which lies adjacent to the C-terminal side of the catalytic site) and reduces AKT phosphorylation and induces apoptosis in multiple myeloma cell lines (29–31). However, the utility of pelorol and MN-100 is restricted by their extremely limited aqueous solubility (30). Moreover, effects of SHIP1 agonists specifically on BCR signaling and on the in vivo growth of neoplastic B cells have not been studied.

In this study, we used a novel small-molecule SHIP1 agonist, AQX-435, to investigate the effect of SHIP1 activation in preclinical models of B-cell neoplasms. AQX-435 was selected from a panel of new pelorol-related compounds, synthesized to improve the aqueous solubility and other drug-like properties of this class of agonist (32). We focused on CLL and DLBCL, both of which are characterized by PI3K activation, and demonstrated that AQX-435 reduces PI3K activation downstream of the BCR and induces apoptosis of malignant B cells in vitro, and reduces lymphoma growth in vivo, alone or in combination with ibrutinib. Overall, our results demonstrate that SHIP1 activation is a potential new therapeutic approach for B-cell neoplasms and can be combined effectively with ibrutinib.

Idelalisib and ibrutinib were from SelleckChem, and Q-VD-OPh was from Sigma. Details for synthesis of AQX-435 are provided (32). All compounds were dissolved in DMSO for in vitro experiments. DMSO was used as a solvent control in all in vitro experiments.

Analysis of CLL samples was performed following informed consent and in accordance with Ethics Committee approvals and the Declaration of Helsinki. Heparinized peripheral blood mononuclear cells (PBMC) were obtained from patients attending clinic at the Southampton General Hospital (Supplementary Table S1). All patients provided written-informed consent. IGHV mutation status, expression of cell surface IgM, CD5, CD19, CD23 and CD38, and ZAP70 were determined as described (9, 33). CD49d expression was analyzed by flow cytometry using a FITC-labeled anti-CD49d mouse monoclonal antibody (Biolegend) and calculated as percentage of CD5⁺CD19⁺ cells. sIgM signaling capacity was determined by measuring the percentage of cells with increased intracellular Ca²⁺ (iCa²⁺) following stimulation with soluble goat F(ab′)₂ anti-IgM (Southern Biotech; ref. 33).

CLL cells were analyzed using either snap-frozen pellets prepared immediately following isolation of PBMCs (for experiments shown in Fig. 1A and Supplementary Fig. S1) or following recovery of cryopreserved cells (for all other experiments). We previously showed that recovery does not significantly affect anti-IgM responses (9, 33, 34). PBMCs were also isolated from healthy individuals and cryopreserved before use (34). Cryopreserved PBMCs were recovered and rested for 1 hour at 37°C prior to use. Cell viability was determined by trypan blue exclusion and was ≥90% in all cases. For BCR stimulation, samples were incubated with soluble or bead-bound goat F(ab′)₂ anti-human IgM, goat F(ab′)₂ anti-human IgD, or control F(ab′)₂ (all Southern Biotech) as described (35). CD40L/IL4 stimulation was as described (36).

TMD8 (kind gift of Professor L.M. Staudt) and WSU-FSCCL and RL (Liebniz Institute DSMZ, Germany) were cultured in RPMI-1640 medium (Sigma) supplemented with 10% (v/v) FBS (BioSera) and penicillin, streptomycin, and glutamine (all Sigma). Cell line identity was routinely confirmed using short tandem repeat analysis (Powerplex 16 System, Promega), and absence of mycoplasma was confirmed using Mycoplasma PCR detection kit (Applied Biological Materials). Cell lines were typically cultured for a maximum of 6 to 8 weeks.

SDS-PAGE was performed using equal protein loading following quantitation of protein content using the BioRad Protein Assay. Immunoblot analysis was performed using the following antibodies: anti-ERK1/2, anti-phosphorylated ERK1/2 (T²⁰²/Y²⁰⁴), anti-SHIP1, anti-phosphorylated SHIP1 (Y¹⁰²²), anti-SYK, anti-phosphorylated-SYK (Y^525/526), anti-AKT, anti-phosphorylated AKT (T³⁰⁸ and S⁴⁷³; all Cell Signaling Technology), anti-MYC (Calbiochem), anti–μ-heavy chain (Jackson ImmunoResearch Laboratories), anti-PARP (BD Biosciences), anti–β-actin (Sigma), and anti-HSC70 (Santa Cruz Biotechnologies). Secondary antibodies were horseradish peroxidase–conjugated anti-rabbit, mouse, or goat antibodies (GE Healthcare), and images were captured using the ChemiDoc-It Imaging System with a BioChemi HR camera (UVP). Immunoblot signals were quantified using ImageJ (http://imagej.nih.gov/ij/). Expression of phosphorylated proteins was normalized to the equivalent total protein expression, whereas expression of MYC was normalized to β-actin. PARP cleavage was quantified by the proportion of cleaved PARP as a percentage of total PARP expression (i.e., cleaved plus uncleaved).

CLL samples were incubated with bead-bound F(ab′)₂ anti-IgM or bead-bound control F(ab′)₂ at 4°C or 37°C. Cells were collected by centrifugation and lysed using an NP-40 lysis buffer [0.025 mol/L Tris-HCl, 0.15 mol/L NaCl, 1 mmol/L EDTA, 1% (v/v) NP-40, and 5% (v/v) glycerol with added phosphatase and protease inhibitors (Sigma)]. After 1 hour on ice, cells were briefly sonicated and then incubated with RNase-free DNase (Qiagen) for 15 minutes at room temperature. Beads and associated proteins were collected by centrifugation and washed 6 times with cold lysis buffer and boiled in SDS-PAGE sample buffer before analysis using immunoblotting.

NOD.Cg-Prkdc^scidIl2rg^tm1Wjl/SzJ mice (NSG mice) were purchased from The Jackson Laboratory or Charles River Laboratories and maintained in a pathogen-free environment in individually ventilated cages and provided with sterilized food and water. Preliminary toxicity experiments were performed to confirm acceptable tolerability of vehicles/compounds. No adverse effects were observed in these initial assessments (data not shown). For efficacy analysis, adult female NSG mice were inoculated subcutaneously with 10⁷ lymphoma cells in PBS. Therapy was initiated when all mice developed palpable tumors (termed day 1) and mice were stratified so that treatment arms contained mice with comparable starting calculated tumor volumes. For TMD-based xenografts, AQX-435 was formulated in 4% ethanol, 0.8% 1-methyl-2-pyrrolidinone (Sigma), 12% polyethylene glycol 200 (Sigma), and 83.2% PBS, and administered at 10 mg/kg (i.p.; 0.3 mL per mouse). Animals were dosed in 7-day cycles each comprising 5 days of AQX-435 (10 mg/kg; i.p.) followed by 2 days with no drug. For patient-derived xenograft (PDX) models, AQX-435 was formulated in 15% cremophor EL/10% dimethylacetamide (both Sigma) in saline (which allowed us to increase the administered dose to 50 mg/kg) and was administered daily (i.p.; 0.3 mL per mouse). Ibrutinib was formulated in 1% hydroxy-propyl-β-cyclodextrin (Sigma) and administered in all experiment at 25 mg/kg by oral gavage (po; 0.2 mL per mouse). The number of animals per experimental group was 6 for TMD8 xenografts, 8 for VFN-D1 and VFN-D2, and 9 for VFN-D3. Tumor growth was assessed using a digital caliper in three perpendicular dimensions and volume calculated according to the following formula: π/6 × length × width × height. Mouse weights were also recorded. All animals were euthanized when the largest dimension of any tumor exceeded 2 cm. Tumors were then resected and weighed. For immunoblot analysis of AKT phosphorylation in TMD8 tumors, animals (n = 3 per arm) were treated with AQX-435 or vehicle on days 1, 2, 4, and 5. Tumors were resected 6 hours following final drug/vehicle administration, disrupted using collagenase/DNase (both Sigma), and lymphoma cells isolated using Lymphoprep (Stemcell Technologies).

Animal experiments were approved by (i) the Institutional Animal Care and Use Committee, the Research and Higher Education section of the Ministry of Education, Youth and Sports of the Czech Republic under MSMT-32441/2018-5 or (ii) by local ethical committees, reporting to the Animal Welfare Ethical Review Board, under UK Home Office license PPL# P4D9C89EA and were fully compliant with the principles of the NC3Rs and the ARRIVE guidelines.

Additional methods are provided in the Supplementary File.

We first characterized SHIP1 expression and Y¹⁰²² phosphorylation in unstimulated primary CLL cells. Analysis was performed using a cohort of samples comprising both U- and M-CLL, and a range of anti–IgM signal-responsiveness (by analysis of iCa²⁺ mobilization). SHIP1 was expressed and constitutively phosphorylated in all cases (Fig. 1A; Supplementary Fig. S1). Although levels varied between samples, we did not observe statistically significant correlations between SHIP1 expression/phosphorylation and IGHV mutation status, signaling capacity, or other prognostic markers (CD38, CD49d, or ZAP70 expression; Supplementary Fig. S1).

Previous studies have shown that stimulation of the BCR of human or mouse B cells results in rapid phosphorylation of SHIP1, including on Y¹⁰²² (Y¹⁰²⁰ in mouse) and relocalization of SHIP1 to the plasma membrane where it is able to access its substrate (2, 23, 25–27). We therefore investigated the effect of anti-IgM on SHIP1 phosphorylation and subcellular localization using a cohort of samples all of which were considered to be anti–IgM-responsive.

For analysis of SHIP1 phosphorylation, cells were stimulated with soluble or bead-bound anti-IgM. Bead-bound anti-IgM induces stronger and longer-lasting signaling compared with soluble anti-IgM, thereby allowing us to probe potential effects of signal strength (35). Both forms of anti-IgM induced AKT phosphorylation and, as expected, the response was stronger in cells treated with bead-bound anti-IgM (Fig. 1B). In most samples, SHIP1 Y¹⁰²² phosphorylation was not substantially affected at time points up to 60 minutes post stimulation (Fig. 1B; Supplementary Fig. S2A and S2B). However, there was some variation between samples, and SHIP1 phosphorylation was increased by approximately 50% in one sample and reduced by approximately 65% in another. Thus, overall, anti-IgM does not consistently promote SHIP1 Y¹⁰²² phosphorylation in signal-competent CLL samples, although SHIP1 Y¹⁰²² phosphorylation may be altered (up or down) in a subset of samples. SHIP1 phosphorylation was also essentially unaltered following treatment of two anti-IgM nonresponsive samples with soluble or bead-bound anti-IgM (Supplementary Fig. S2C).

Analysis of SHIP1 subcellular localization by microscopy was performed using soluble anti-IgM only because strong autofluorescence of anti–IgM-coated beads precluded the use of this reagent in these experiments. Although phospho-SHIP1 was predominantly localized in the cytoplasm of CLL cells without stimulation (Fig. 1C), anti-IgM triggered relocalization of phospho-SHIP1 to the plasma membrane. Confocal microscopy confirmed the membrane localization of phospho-SHIP1 in anti–IgM-treated cells and additionally demonstrated substantial colocalization of phospho-SHIP1 and sIgM (Fig. 1D).

We performed “pull-down” experiments to determine whether colocalization of phospho-SHIP1 and sIgM was due to association of SHIP1 with activated sIgM complexes. μ-heavy chain precipitations were performed at 4 and 37°C to investigate both constitutive and activation-induced association (Fig. 1E). Immunoblotting revealed increased binding of both phospho-SYK and phospho-SHIP1 to sIgM in cells at 37°C, whereas μ-heavy chain was precipitated equally under both conditions.

Overall, these results demonstrate that SHIP1 is constitutively expressed and phosphorylated on Y¹⁰²² in unstimulated CLL cells. sIgM activation does not alter SHIP1 Y¹⁰²² phosphorylation but leads to recruitment of phospho-SHIP1 to sIgM.

We selected the novel SHIP1 agonist AQX-435 to investigate the effects of SHIP1 activation (Supplementary Fig. S3). AQX-435 is structurally related to pelorol and the closely related compound MN-100, which was synthesized to remove pharmacologically undesirable catechol and ester functionalities (30, 37). However, both pelorol and MN-100 have extremely poor water solubility, and AQX-435 was produced as part of a panel of new analogues synthesized to improve solubility and other drug-like properties (30, 32). AQX-435 was selected from this panel based on its enhanced drug-like properties (including improved aqueous solubility) and the results of initial screens demonstrating effective inhibition of AKT phosphorylation in primary CLL cells (data not shown). We did not use AQX-1125 (which is structurally unrelated to pelorol and was developed as an anti-inflammatory agent) as a tool compound, because it is a relatively weak SHIP1 activator compared with pelorol-related compounds (38, 39). Some results have previously been reported in abstract form where AQX-435 was referred to as AQX-C5 (40, 41).

We first investigated the effect of AQX-435 on anti–IgM-induced AKT phosphorylation using a set of sIgM signal-responsive CLL samples. We analyzed both AKT T³⁰⁸ and S⁴⁷³ phosphorylation, although for quantitation we focused on S⁴⁷³ due to increased sensitivity of detection of phosphorylation at this site. Immunoblotting demonstrated that AQX-435 caused a dose-dependent decrease in phosphorylation of AKT at both sites (Fig. 2A). As expected, inhibition of AKT phosphorylation was also observed with idelalisib, although responses were more dramatic that observed with AQX-435 (Fig. 2A; Supplementary Fig. S4). AQX-435 also reduced AKT phosphorylation following ligation of sIgD which is coexpressed with sIgM on CLL cells (Supplementary Fig. S5A and S5B).

In addition to effects on AKT phosphorylation, AQX-435 significantly reduced anti–IgM-induced phosphorylation of ERK1/2 (Fig. 2B) which is susceptible to inhibition with idelalisib in CLL cells (12) and induction of MYC (Fig. 2C), a downstream target for PI(3,4,5)P₃-mediated signaling. By contrast, AQX-435 did not inhibit anti–IgM-induced SYK phosphorylation which is upstream of PI(3,4,5)P₃ (Fig. 2D) or anti–IgM-induced iCa²⁺ flux which was also only modestly reduced by idelalisib (Supplementary Fig. S6A and S6B). AQX-435 also did not alter sIgM expression or increase SHIP1 Y¹⁰²² phosphorylation (Supplementary Fig. S6C and S6D).

Therefore, AQX-435 decreases PI(3,4,5)P₃-mediated activation of AKT and downstream MYC induction, without substantial effects on upstream SYK phosphorylation or sIgM expression.

Combination therapy is an effective strategy to boost responses, and we therefore analyzed the effects of combined treatment with AQX-435 and the BTK inhibitor, ibrutinib, on anti–IgM-induced AKT phosphorylation. AQX-435 was tested at 30 μmol/L, and ibrutinib was used at 100 nmol/L, a concentration that is readily achieved in the plasma of treated patients and is sufficient to fully occupy the BTK active site and inhibit BTK autophosphorylation in cells in vitro (GP, Rachel Arthur, Adam Linley; submitted for publication). Like AQX-435, ibrutinib only partially inhibited AKT phosphorylation in anti–IgM-treated cells (Fig. 3A and B). However, the combination of AQX-435 and ibrutinib resulted in almost complete inhibition of anti–IgM-induced AKT phosphorylation compared with either agent alone.

We also investigated effects of AQX-435 and/or ibrutinib on the calcium response pathway by analyzing both iCa²⁺ mobilization and phosphorylation of PLCγ2 (which catalyzes production of the iCa²⁺-liberating second messenger inositol triphosphate). Consistent with previous experiments, AQX-435 had relatively little effect on anti–IgM-induced PLCγ2 phosphorylation and iCa²⁺ release, whereas both responses were very effectively inhibited by ibrutinib (Supplementary Fig. S7A and S7B). Addition of AQX-435 did not interfere with ibrutinib-mediated inhibition of the calcium response. Therefore, the combination of AQX-435 and ibrutinib results in a more pronounced inhibition of anti-IgM signaling responses compared with either agent alone.

PI(3,4,5)P₃-dependent signaling is a key driver of cell survival, and we next analyzed the effect of AQX-435 on CLL cell viability using a cohort of 24 CLL samples. AQX-435 reduced CLL cell viability in a dose-dependent manner (Fig. 4A). AQX-435–induced cell killing was somewhat greater in samples from some poor prognosis subsets (U-CLL and CD38 positive; Supplementary Fig. S8A). AQX-435–induced apoptosis was mediated via caspases because AQX-435 induced PARP cleavage (Fig. 4B and C) and caspase inhibition reduced AQX-435–induced apoptosis (Fig. 4D). In contrast to AQX-435, and consistent with previous findings (42), idelalisib had only very modest proapoptotic effects in CLL cells even when tested at concentrations up to 10 μmol/L to ensure full inhibition of AKT phosphorylation (Supplementary Fig. S8B).

CLL cells are exposed to a range of microenvironmental factors in vivo which can promote malignant cell survival (43). To determine whether AQX-435 could also induce apoptosis in the presence of these signals, we performed experiments with CD40L and IL4 which promotes CLL cell survival in vitro and may mimic potential supportive effects of T cells in vivo (36, 44, 45). Although CD40L/IL4 decreased spontaneous apoptosis of CLL cells, AQX-435 was still able to significantly increase apoptosis in the presence of CD40L/IL4 (Fig. 4E). AQX-435 also overcame the survival promoting effects of bead-bound anti-IgM stimulation (Fig. 4F).

Finally, we investigated effects of AQX-435 on viability of B cells from healthy individuals. Although AQX-435 significantly reduced viability of normal B cells (Fig. 4G), these cells were significantly less sensitive than CLL cells. For example, when comparing DMSO-normalized data (to account for variation in spontaneous apoptosis between normal B cells and CLL cells), the difference in reduction in viability for cells treated with 30 μmol/L AQX-435 was highly statistically significant (P = 8.4 × 10⁻⁶; Student t test; mean ± SD of 17 ± 7% for normal B cells versus 52 ± 25% for CLL cells).

Having evaluated effects of AQX-435 in CLL, we next investigated whether the inhibitory effects of AQX-435 on BCR signaling extended to other B-cell neoplasms. We focused on DLBCL which is the most common lymphoma subtype and is also frequently characterized by active PI3K signaling (5, 13). Moreover, there is a particularly pressing need for new therapeutic options in DLBCL where responses to kinase inhibitors, including idelalisib, are much less dramatic than those observed in CLL.

We selected the ABC-DLBCL cell line TMD8 for initial studies because it depends on BCR signaling for growth (46) and confirmed like that, like idelalisib, AQX-435 inhibited anti-IgM-induced AKT phosphorylation in these cells (Fig. 5A and B). In addition, AQX-435 reduced anti–IgM-induced AKT phosphorylation in a second DLBCL cell line, WSU-FSCCL, derived from a transformed FL (Supplementary Fig. S9).

Next, we screened a panel of DLBCL cell lines, including examples of both GCB and ABC-DLBCL, for AQX-435–induced growth inhibition. Applying a cutoff of 5 μmol/L, 10 of 11 of the cell lines were considered sensitive to AQX-435–induced growth inhibition (Fig. 5C; Supplementary Table S2). Using Annexin V staining, we confirmed that AQX-435 induced TMD8 cell apoptosis in vitro with an IC₅₀ of approximately 2 μmol/L (Supplementary Fig. S10A). Thus, in addition to primary CLL cells, AQX-435 reduces anti–IgM-induced AKT phosphorylation and induces apoptosis in DLBCL cells.

We analyzed in vivo antilymphoma effects of AQX-435 using a panel of mice bearing TMD8 or DLBCL PDX tumors. AQX-435 significantly reduced the volume of TMD8 tumors in vivo (Fig. 5D; Supplementary Table S7) and tumor weight at the end of the experiment (Fig. 5E). Efficacy of AQX-435 was very similar to ibrutinib. There was no evidence for toxicity in terms of reduced body weight (Supplementary Fig. S10B) or gross signs of toxicity with either agent. We also examined the levels of AKT phosphorylation in lymphoma cells from animals treated with AQX-435 or vehicle (n = 3 for each; Fig. 5F; Supplementary Fig. S10C). We were unable to obtain intact protein from one tumor from the AQX-435–treated mice. However, AKT phosphorylation in the remaining two tumors was substantially reduced (∼50%) compared with controls.

Next, three different PDX models derived from patients with treatment-refractory DLBCL (VFN-D1, VFN-D2) or DLBCL at diagnosis (VFN-D3) were used to assess antitumor activity of AQX-435. Experiments were performed using AQX-435 as a single agent, but also in combination with ibrutinib because previous studies have shown increased efficacy of ibrutinib with PI3Ki (47). Detailed molecular characterization of the PDX models is given in Supplementary Tables S3–S6 and Supplementary Fig. S11.

Consistent with the differing cell-or-origin and genetic make-up, response to AQX-435 and/or ibrutinib differed between the models. AQX-435 had substantial inhibitory effects on tumor growth in VFN-D2 and VFN-D3 but was less effective in VFN-D1 (Fig. 6A–C; Supplementary Table S7). Ibrutinib was essentially inactive in VFN-D2 (GCB subtype) and had similar activity to AQX-435 in the other two models (both non-GCB subtype). Importantly, the combination of AQX-435 and ibrutinib further reduced tumor volume in the non-GCB PDX models VFN-D1 and VFN-D3 (but not in GCB PDX model VFN-D2 where ibrutinib was essentially inactive as a monotherapy). Enhanced activity of the AQX-435/ibrutinib combination was particularly striking in the VFN-D1 model, resulting in profound and durable (>14 days) tumor regression. Effects of AQX-435 (±ibrutinib) on tumor volume were accompanied by significant reductions in tumor weight at the end of the experiments in all three models (Fig. 6D), and there was no evidence for toxicity in terms of reduced body weight (Supplementary Fig. S12). Thus, AQX-435 has promising activity against DLBCL and can be combined effectively with ibrutinib to enhance antitumor activity.

In this study, we explored the novel strategy of phosphatase activation to inhibit BCR signaling using the SHIP1 activator AQX-435. Previous studies using the SHIP1 agonist MN-100 have demonstrated induction of apoptosis in multiple myeloma cell lines in vitro (29, 31). However, the effects of pharmacologic SHIP1 activation on BCR signaling and in vivo antitumor activity of SHIP1 agonists have not been reported. We show for the first time that SHIP1 activation using AQX-435 effectively inhibits PI(3,4,5)P₃-mediated signaling downstream of the BCR in CLL and DLBCL cells. Moreover, AQX-435 promotes apoptosis of primary CLL cells with relative specificity compared with normal counterparts, and is well tolerated and efficacious in in vivo models of DLBCL, including in combination with ibrutinib. Thus, SHIP1 activation is a potential new therapeutic approach to inhibit inappropriate PI3K-mediated signaling, with activity across several distinct subtypes of B-cell neoplasms.

AQX-435 was selected from a panel of new synthetic pelorol analogues synthesized to generate new SHIP1 agonists with improved water solubility and other drug-like properties compared with previous agents, including pelorol itself and MN-100 (30, 32, 37). In particular, AQX-435 has much improved water solubility compared with previous pelorol-related SHIP1 agonists and retains good ability to suppress PI(3,4,5)P₃-mediated signaling in intact cells. Moreover, it can be formulated for in vivo administration to achieve doses that are sufficient to limit lymphoma growth without substantial toxicity.

Profiling of primary samples demonstrated that SHIP1 was widely expressed and phosphorylated in CLL cells. Levels were variable but were not clearly different between disease subtypes. We were unable to confirm results of a previous study demonstrating that SHIP1 expression/phosphorylation was higher in ZAP70-positive versus -negative cases (48). However, quantitative analysis in that study was based on only a small number of samples. Moreover, although miR-155 suppresses SHIP1 expression and is expressed more highly in U-CLL (49), we also did not observe a correlation between SHIP1 expression and IGHV mutation status. Although we cannot exclude the possibility that larger studies may reveal some quantitative differences in SHIP1 expression/phosphorylation between disease subtypes, constitutive SHIP1 expression/phosphorylation largely appears to be a general feature of CLL cells.

Plasma membrane localization is critical for SHIP1 function because it is required for substrate access and, in normal B cells, BCR stimulation induces rapid phosphorylation of SHIP1, including on Y¹⁰²², and relocalization to the plasma membrane (2, 23, 25–27). Anti-IgM stimulation of CLL cells also triggered plasma membrane relocalization of SHIP1 and association with activated BCRs. However, it was interesting to note that SHIP1 was constitutively phosphorylated on Y¹⁰²² in CLL cells and that phosphorylation was not further increased by stimulation in vitro. It is possible that this constitutive phosphorylation of SHIP1 is due to (auto)antigen engagement of CLL cell BCRs in vivo, or engagement of other receptors which can induce Y¹⁰²² phosphorylation, such as FcγRIIB (2). Constitutive Y¹⁰²² phosphorylation may therefore prime CLL cells for full activation following BCR activation, but membrane recruitment may require additional steps. One possibility is that activation of the BCR triggers recruitment of preexisting phospho-SHIP1 via binding of the SHIP1 SH2 domains to tyrosine phosphorylated residues in the ITAM motifs of CD79A/B (the signal transduction components of the BCR; refs. 2, 27).

Consistent with shared ability to reduce concentrations of PI(3,4,5)P₃, AQX-435 and idelalisib exerted similar effects on BCR signaling in primary CLL cells (i.e., inhibition of AKT and ERK phosphorylation without effects on upstream SYK phosphorylation or parallel iCa²⁺ mobilization). However, higher concentrations were required for AQX-435 action compared with idelalisib. Differences in potency/efficacy of AQX-435 and idelalisib are likely to be multifactorial, but may reflect, at least in part, the differing modes of action of these agents. Thus, increasing idelalisib concentrations will ultimately lead to full PI3K blockade once all of the kinase active sites have been occupied. By contrast, the inhibitory effects of AQX-435 are limited by the availability of SHIP1; once the available SHIP1 is fully activated, addition of further compound will not enhance inhibitory effects. It is also important to consider that, in contrast to PI3Ki, SHIP1 does not block production of PI(3,4,5)P₃, but accelerates its breakdown. Thus, even when SHIP1 is fully activated, there remains a transient “window” in which PI(3,4,5)P₃ may engage downstream effectors prior to its breakdown. However, although relatively high AQX-435 concentrations were required to inhibit signaling in CLL cells, low μmol/L concentrations were sufficient for growth-inhibitory effects in most lymphoma cell lines analyzed and for induction of apoptosis of TMD8 cells, and our analysis clearly demonstrated that it was possible to achieve concentrations sufficient to reduce AKT phosphorylation and tumor growth in vivo with no evidence for substantial toxicity. Notably, AQX-435 had broad activity against a panel of DLBCL cell lines; the ability of SHIP1 activation to counter PI(3,4,5)P₃-mediated signaling regardless of the specific PI3K isoform(s) that mediate PI(3,4,5)P₃ synthesis may account for this activity, whereas inhibition of multiple PI3K isoforms is generally required for effective growth inhibition (17).

Consistent with a key role for PI(3,4,5)P₃-mediated signaling in cell survival, AQX-435 induced caspase-dependent apoptosis of CLL cells and overcame survival-promoting effects of both anti-IgM and CD40L/IL4. Interestingly, CLL cells were significantly more susceptible to AQX-435–induced apoptosis than nonmalignant B cells. Future experiments will explore whether the enhanced sensitivity of CLL cells may be linked to constitutive Y¹⁰²² phosphorylation. AQX-435 also appeared to have stronger proapoptotic effects than idelalisib which only very modestly induced cell killing even at relatively high concentrations. The increased proapoptotic activity of AQX-435 compared with idelalisib could be mediated by combined depletion of PI(3,4,5)P₃ and accumulation of PI(3,4)P₂.

We extended our analysis to DLBCL, including for in vivo assessment. It was important to assess activity of AQX-435 in this setting, because, in contrast to CLL, targeted agents (including idelalisib) are relatively ineffective as a monotherapies in DLBCL, and effective combinations are important to boost therapeutic responses. We demonstrated that AQX-435 alone reduced lymphoma growth in vivo using the TMD8 cell line and three new DLBCL PDX models. Enhanced activity of the combination of AQX-435 and ibrutinib was observed in 2 of the 3 PDX models and was particularly striking in VFN-D1 which showed clear evidence for tumor regression with the combination. The mechanism underlying the enhanced efficacy of the combination of AQX-435 and ibrutinib is unclear but may stem from the more pronounced inhibition of AKT phosphorylation by the combination compared with either agent alone, as observed in CLL cells. Variable drug responsiveness between the PDX models may stem from differences in cell-of-origin (i.e., GCB vs. non-GCB), and it was notable that growth of VFN-D2 tumors, derived from GCB-DLBCL, appeared to be unaffected by ibrutinib because patients with GCB-DLBCL are less responsive to ibrutinib therapy compared with ABC-DLBCL (50).

Overall, our results demonstrate that SHIP1 activation is a potential new therapeutic approach for B-cell neoplasms characterized by PI3K activation, alone or in combination with ibrutinib. It has broad growth-inhibitory activity against DLBCL, which may reflect its ability (unlike idelalisib) to reduce PI(3,4,5)P₃-mediated signaling independent of the particular PI3K isoform(s) that catalyze PI(3,4,5)P₃ accumulation. It also has stronger proapoptotic activity which may relate to its ability to drive accumulation of PI(3,4)P₂. Further experiments, particularly using immunocompetent animals, will be required to determine whether the differing modes of action of idelalisib and AQX-435 (which appears to cause a more nuanced attenuation of PI(3,4,5)P₃-mediated signaling) translate into reduced toxicity, and the relative effectiveness of AQX-435 versus PI3Ki in combination with ibrutinib.

C.T. Williamson is an employee of Aquinox Pharmaceuticals. G. Lenz is a paid consultant for Bayer, AstraZeneca, Janssen, Roche, NanoString, Celgene, Abbvie, Bristol-Myers Squibb, and Gilead; reports receiving commercial research grants from Bayer, Aquinox, AstraZeneca, Janssen, Gilead, Verastem, and MorphoSys; reports receiving speakers bureau honoraria from Roche, Gilead, Janssen, Abbvie, and Celgene; and reports receiving other remuneration from Morphosys. J. Pettigrew is an employee of Aquinox Pharmaceuticals. C. Harwig is an employee of Aquinox Pharmaceuticals. M. Cragg is a paid consultant for Bioinvent; reports receiving commercial research grants from Bioinvent, GlaxoSmithKline, Roche, iTeos, and UCB; and reports receiving speakers bureau honoraria from Baxalta, Boehringer, and Merck GDa. F. Forconi is a paid consultant for Menarini; reports receiving commercial research grants from Gilead; reports receiving speakers bureau honoraria from Abbvie and Jannsen-Cilag; and reports receiving other remuneration from AstraZeneca. A.J. Steele holds ownership interest (including patents) in Portola Pharmaceuticals. J. Cross is an employee of Aquinox Pharmaceuticals. L. Mackenzie is an employee of Aquinox Pharmaceuticals. P. Klener reports receiving commercial research grants from Aquinox Pharmaceuticals. G. Packham is a paid consultant for and reports receiving commercial research grants from Aquinox Pharmaceuticals. No potential conflicts of interest were disclosed by the other authors.

Conception and design: C.T. Williamson, G. Lenz, J. Pettigrew, C. Harwig, F.K. Stevenson, J. Cross, L. Mackenzie, G. Packham

Development of methodology: E.A. Lemm, B. Valle-Argos, L.D. Smith, M.J. Carter, N.J. Weston-Bell, J. Pettigrew

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): E.A. Lemm, B. Valle-Argos, L.D. Smith, J. Richter, Y. Gebreselassie, M.J. Carter, M. Svaton, N.J. Weston-Bell, L. Karydis, C.T. Williamson, G. Lenz, C. Harwig, F. Forconi, P. Klener

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): E.A. Lemm, B. Valle-Argos, L.D. Smith, J. Richter, Y. Gebreselassie, J. Karolova, M. Svaton, K. Helman, N.J. Weston-Bell, G. Lenz, M. Cragg, A.J. Steele, P. Klener, G. Packham

Writing, review, and/or revision of the manuscript: E.A. Lemm, B. Valle-Argos, L.D. Smith, J. Richter, G. Lenz, C. Harwig, F.K. Stevenson, M. Cragg, F. Forconi, A.J. Steele, L. Mackenzie, P. Klener, G. Packham

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): B. Valle-Argos, F. Forconi, L. Mackenzie

Study supervision: J. Cross, G. Packham

Other (Coinventor on patent for AQX-435): C. Harwig, L. Mackenzie

This work was supported by grants from Cancer Research UK (C2750/A23669), Bloodwise (14045, 16004, 16003, 18009), Kay Kendall Leukaemia Fund (KKL0168), Aquinox Pharmaceuticals (Canada) Inc., the Southampton Experimental Cancer Medicine and Cancer Research Centres (C24563/A15581; C34999/A18087), UNCE/MED/016, GACR17-14007S, and the Keanu Eyles Haematology Fellowship for the Cancer Immunology Centre.

We thank the patients involved in this study for the kind gift of samples and Dr. Ian Tracy, Isla Henderson, Dr. Kathy Potter, and Carina Mundy for characterization and storage of the CLL samples, and Dr. Patrick Duriez for providing Annexin V. We thank Professor L. Staudt for the kind gift of TMD8 cells.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.