Idelalisib Rescues Natural Killer Cells from Monocyte-Induced Immunosuppression by Inhibiting NOX2-Derived Reactive Oxygen Species

Chronic lymphocytic leukemia (CLL) is characterized by progressive clonal expansion of mature B lymphocytes (1). Therapy for CLL includes anti-CD20 IgG such as rituximab (RTX), ofatumumab (OFA), or glycoengineered obinutuzumab, which marks malignant cells for destruction by natural killer (NK) cells and myeloid cells (2, 3). NK cells and myeloid cells express Fc gamma receptors (FcγR) that attach the Fc portion of IgG to exert antibody-dependent cellular cytotoxicity (ADCC) or phagocytosis against foreign cells, including malignant B cells. ADCC exerted by NK cells has been proposed to contribute to the therapeutic efficacy of anti-CD20 in CLL and other B-cell malignancies (4–8). However, NK cells are dysfunctional in CLL (9–12), and patients may therefore benefit from therapies that augment NK cell survival and function.

Monocytes and other myeloid cells, including neutrophils, express the NOX2 isoform of NADPH oxidase, which generates reactive oxygen species (ROS). Exposure of monocytes and neutrophils to a variety of stimuli, including immobilized IgG antibodies, may activate NOX2 to generate ROS. Fc gamma receptor I (FcγRI, CD64), a high-affinity Fc receptor for IgG, is constitutively expressed by classical monocytes (CD14⁺/CD16⁻) and macrophages (13, 14). FcγRI cross-linking by IgG1 and IgG3 triggers intracellular signal transduction pathways that activates NOX2 to generate ROS (15, 16). Similarly, ligation of Fc gamma receptors (FcγRII, CD32 and FcγRIII, CD16) on the surface of monocytes and neutrophils by IgG can induce ROS production (17).

NOX2-derived ROS are pivotal for the elimination of ingested microbes (18) but may also be released extracellularly to suppress the function and viability of adjacent NK cells and T cells; these immunosuppressive properties comprise induction of PARP-dependent cell death characterized by nuclear accumulation of apoptosis-inducing factor and DNA degradation (19–23). Mice genetically deprived of NOX2 function (gp91^phox−/− or Nox2^−/−) show improved NK cell–mediated clearance of metastatic cells but also develop autoimmunity, thus supporting the idea that NOX2-derived ROS may control NK cell and T-cell function in vivo (18, 24, 25).

The class I phosphatidylinositol 3 kinase (PI3K) produces the PIP3 lipid that is critical for activation of Akt (protein kinase B; ref. 26). Class I PI3Ks are heterodimeric enzymes that comprise a regulatory and a catalytic subunit. Class IA PI3Ks may contain the catalytic isoforms p110α, p110β, or p110δ, whereas p110γ is the single isoform of class IB PI3K. p110α and p110β are widely distributed in mammalian tissues, whereas p110δ is highly enriched in leukocytes (27, 28). The PI3Kδ-selective inhibitor idelalisib (CAL-101) inhibits PI3K targets in B-cell receptor signaling to promote apoptosis in CLL cells, supporting that PI3Kδ function is critical for CLL cell survival (29, 30). Idelalisib, alone or in conjunction with anti-CD20, is approved in Europe and the United States for use in CLL and other B-cell malignancies but is associated with several toxicities such as autoimmunity and severe infections, including opportunistic infections (31–36). Details regarding mechanisms underlying these toxicities and the benefit of the anti-CD20/idelalisib interaction are not known.

PI3K is implicated in regulation of NOX2-derived ROS formation in myeloid cells, including cells of the monocyte/macrophage lineage (37, 38). Akt phosphorylates the p47^phox subunit of NOX2 (39) and thus contributes to the initiation of ROS production. Anti-CD20 triggers neutrophils and monocytes to release extracellular ROS that inhibit the survival and function of adjacent NK cells (10, 40).

Here, we assessed the role of PI3K, in particular PI3Kδ, in antibody-induced ROS formation and NK cell dysfunction. We observed that idelalisib prevented Akt-dependent phosphorylation of the p47^phox subunit of NOX2, suppressing anti–CD20-induced ROS formation by human monocytes. Idelalisib rescued NK cells from ROS-induced toxicity and improved NK cell–mediated ADCC against primary and cultured malignant B cells. In mice, systemic treatment with idelalisib reduced hematogenous metastasis by a mechanism that required intact NOX2 and presence of NK cells. Our findings provide additional insight into the mechanisms of idelalisib efficacy and toxicity.

The following compounds and reagents were used: RTX (Roche AB; Pharmacy catalog no. 494237), OFA (Novartis Pharmaceuticals Corporation; Pharmacy catalog no. 699855), cetuximab (Sigma-Aldrich; catalog no. MSQC18), the PI3Kδ inhibitor idelalisib (CAL101; Selleckchem; catalog no. S2226; ref. 30), the PI3Kβ inhibitor TGX-221 (Selleckchem; catalog no. S1169; ref. 37), the PI3Kα inhibitors HS-173 (ref. 41; Selleckchem; catalog no. S7356) and GDC-0326 (Selleckchem; catalog no. S8157; ref. 42), the PI3Kγ inhibitor AS-252424 (Selleckchem; catalog no. S2671; ref. 37), the pan-PI3K inhibitor LY294002 (Cell Signaling Technology; catalog no. 9901; ref. 43), the NOX2 inhibitor diphenylene iodonium (DPI; Sigma-Aldrich; catalog no. D2926; ref. 44), the ROS scavenger catalase (Worthington; catalog no. LS001872; ref. 45), DMSO (Sigma-Aldrich; catalog no. D2650–100ML; used for dissolving PI3K inhibitors and DPI), CellTrace violet stain (Thermo Fisher Scientific; catalog no. C34557), and anti–CD107a-PE-Cy7 Ab (BD Pharmingen; catalog no. 561348).

Blood samples were collected from 6 patients with CLL. The patients provided written-informed consent, and the study was performed in accordance with the Declaration of Helsinki. Patients were asymptomatic, Binet stage A and did not undergo treatment for CLL (including chemotherapy, corticosteroids, anti-CD20, or other B-cell–inhibitory treatment) at the time of participation and sampling. Leukopacks from anonymized healthy blood donors were obtained from the Blood Center at Sahlgrenska University Hospital. Peripheral blood mononuclear cells (PBMC) from patients with CLL and healthy donors were isolated using dextran sedimentation followed by density gradient centrifugation using Lymphoprep (Alere Technologies AS; catalog no. 1114547; ref. 22). NK cells (CD3⁻/CD56⁺; Miltenyi Biotec; catalog no. 130–092–657) and monocytes (CD14⁺/CD16⁻) from healthy donors were enriched from PBMCs by immunomagnetic isolation using an MACS NK cell isolation kit and a MACS monocyte isolation kit II (Miltenyi Biotec; catalog no. 130–091–153), respectively, according to the manufacturer's instructions. In these experiments, the purity of isolated NK cells and monocytes exceeded 96% and 92%, respectively (Supplementary Fig. S1A and S1B). Primary CLL cells were isolated from patient PBMCs using the MACS B-cell (B-CLL) isolation kit according to the manufacturer's instructions (purity >90%; Miltenyi Biotec; catalog no. 130–093–660; Supplementary Fig. S1C). The enriched B cells from patients were resuspended in Iscove's modified Dulbecco minimum essential medium supplemented with 10% human AB+ serum and used as target cells in assays of ADCC as described below. The 221 B lymphoblastoid cell line (46) was kindly provided by Professor Alessandro Moretta, University of Genoa, Genoa, Italy, in 2016 and grown in RPMI 1640 medium with 10% FCS and 1% l-glutamine. The cells carried lymphoblastic morphology and stained with human anti-CD20 using flow cytometry but were otherwise not authenticated. The 221 cells were checked for absence of mycoplasma using PCR before freezing aliquots. Each aliquot was thawed and cultured for no more than 2 months for each experiment.

NK cells and monocytes were cocultured in 96-well plates overnight at 37°C and 5% CO2 in the presence or absence of IgG isotype mAbs (10 μg/mL, RTX and OFA), with PI3K isoform–selective inhibitors (idelalisib, TGX-221, HS-173, GDC-0326, or AS-252424), the pan-PI3K inhibitor LY294002 (5 μmol/L), the NOX2 inhibitor DPI (3 μmol/L), the ROS scavenger catalase (200 U/mL), or DMSO (control), as indicated in figure legends. In some experiments, RTX and OFA Fab fragments (10 μg/mL; ref. 10) were used to define the role of the Fc portion. To determine the effect of antibody-independent ROS-induced cell death, NK cells were cocultured overnight with monocytes in 96-well low-adherence plates in the absence of antibodies. In this model, low-adherence plates favor cell–cell interactions and allow spontaneous ROS production from monocytes toward the NK cells. NK cell viability was assessed by flow cytometry (using a 4-laser BD LSR Fortessa and BD FACS Diva software, BD Biosciences) after staining with the Live/Dead Fixable Dead Cell Stain kit (Thermo Fisher Scientific; catalog no. L10120) according to the manufacturer's instructions. Cell death was measured as the percentage of the Live/Dead-positive cells and confirmed using the altered scatter characteristics displayed by dead lymphocytes, i.e., a reduced forward scatter and an increased side scatter signal (19).

To determine NK cell cytotoxicity against malignant B cells after exposure to ROS-producing monocytes, NK cells and monocytes were first cocultured in 96-well plates overnight with or without immobilized RTX (10 μg/mL) and in the presence or absence of idelalisib (0.1–0.2 μmol/L), LY294002 (5 μmol/L), or catalase (200 U/mL) using the diluent (sodium chloride or DMSO) as control. After 16 hours, Cell Trace violet–labeled primary CLL cells or 221 B-lymphoblastoid cells coated with RTX (10 μg/mL) were added at an effector to target cell ratio of 2:1 (NK:CLL) together with anti–CD107a-PE-Cy7. After another 4 hours of incubation, plates were centrifuged at 500G for 5 minutes, the supernatant was discarded, and cells were stained with the Live/Dead Far Red Dead Cell Stain. Cell death in target cells and NK cells as well as NK cell degranulation was assayed by flow cytometry using the aforementioned anti-CD107a. Lysed primary CLL and 221 target cells were identified as Cell Trace violet–positive Live/Dead Far Red–positive cells. Dead NK cells were identified as Cell Trace violet–negative Live/Dead Far Red–positive cells (for gating strategy of lysed and dead cells, see Supplementary Fig. S1D and S1E). NK cell degranulation was estimated as the percentage of viable NK cells (gated as Live/Dead Far Red negative, non–Cell Trace violet-labeled cells within the FSC/SSC gate comprising lymphocytes) that stained positively for CD107a (47) using gating strategies as previously described (48). In some experiments, NK cells were directly cocultured with RTX-coated primary CLL cells in the presence of monocytes for 4 hours followed by assay of ADCC and NK cell degranulation.

ROS production from monocytes was measured by isoluminol-enhanced chemiluminescence that only captures extracellular ROS (21). In brief, 2 × 10⁵ purified monocytes from healthy donors were added to 96-well plates in the presence of isoluminol (10 mg/mL) and horseradish peroxidase (HRP, 4 U/mL, Sigma-Aldrich; catalog no. 516531–25KU) and presence or absence of IgG isotype mAbs (RTX, OFA, or cetuximab) or Fab fragments, as indicated in figure legends. The release of extracellular ROS (light emission) was continuously monitored using a BMG FLUOStar Microplate Reader (BMG Labtech). ROS production (chemiluminescence) was calculated as the AUC. In some experiments, monocytes were stimulated with phorbol myristate acetate (PMA, 50 nmol/L, an activator of protein kinase C, Sigma-Aldrich; catalog no. P1585–1MG; ref. 49) and assayed for extracellular ROS production. In mouse experiments, Gr1⁺ cells were isolated through magnetic separation from harvested bone marrow samples (50) followed by the isoluminol-based ROS production assay using the BMG FLUOStar Microplate Reader as described above.

Consumption of hydrogen peroxide was determined as described (22, 51). In brief, compounds were diluted in Krebs–Ringer glucose (KRG) buffer and incubated with 50 μmol/L H₂O₂ for 15 minutes using catalase (200 U/mL) as the positive control. Remaining H₂O₂ was determined fluorimetrically by measurement of oxidation of p-hydroxyphenyl acetic acid (Sigma-Aldrich; catalog no. H50004; 0.5 mg/mL) catalyzed by HRP (4 U/mL), using the BMG FLUOStar Microplate Reader (excitation, 320 nm; emission 400 nm).

Cell lysates from monocytes were prepared as described (22). Briefly, monocytes were disrupted in RIPA buffer (Sigma-Aldrich; catalog no. R0278–50ML) containing a protease inhibitor cocktail (Sigma-Aldrich; catalog no. P8340). The protein concentration of the cell lysate was determined using the Pierce BCA protein assay kit (Thermo Fisher Scientific; catalog no. 23227). The lysate was mixed with NuPAGE LDS sample buffer (Invitrogen; catalog no. NP0007) and NuPAGE sample reducing agent (Invitrogen; catalog no. NP0004), heated at 70°C for 10 minutes, and resolved on 4%–12% NuPAGE Novex Bis-Tris precast gels (Invitrogen). After electrophoresis, proteins were transferred to nitrocellulose membranes using the iBlot Gel Transfer Device (Invitrogen). Membranes were blocked with buffered saline containing detergent and Hammersten casein solution (Invitrogen; catalog no. WB7050). For detection of phosphorylated Akt and p47^phox, membranes were incubated with a rabbit phospho-Akt (Ser473) polyclonal antibody (Cell Signaling Technology; catalog. no. 9271; 1:1,000) or with a rabbit phospho-p47^phox (Ser304) polyclonal antibody (Abcam; catalog no. ab63554; 1:500), respectively, followed by incubation with an HRP-conjugated goat anti-rabbit (DAKO; catalog no. P0448; 1:1,000). Bound antibodies were visualized by enhanced chemiluminescence (Chemi-Doc, Bio-Rad) with Clarity Western ECL Substrate (Bio-Rad; catalog no. 1705060). Band intensity was quantified using the NIH ImageJ software (52). The nitrocellulose membranes were then stripped with NewBlot IR Stripping buffer (LI-COR; catalog no. 928–40028) for 15 minutes at room temperature and reprobed with rabbit anti total-p47^phox (Cell Signaling Technology; catalog no. 4312; 1:1,000) or with rabbit GAPDH (D16H11) monoclonal antibody (Cell Signaling Technology; catalog no. 5174; 1:1,000). Cell lysate content of GAPDH was used as the loading control.

F_ab2 fragments of RTX and OFA were generated as previously described (10). Briefly, these Fab fragments were prepared by pepsin digestion using the Pierce F(ab)2 preparation kit (Thermo Fisher Scientific; catalog no. 44988) according to the manufacturer's instructions. Digestion and purity were confirmed by SDS-PAGE gel.

C57BL/6 mice were obtained from Charles River Laboratories, and B6.129S6-Cybb^tm1Din (Nox2^−/− or Nox2-KO) mice, which lack functional ROS-forming NOX2 (25), were obtained from The Jackson Laboratory. The wild-type (WT) and Nox2-KO mice were kept under pathogen-free conditions according to guidelines issued by the University of Gothenburg. B16F10 cells were confirmed by use of PCR to be of mouse origin, and no mammalian interspecies contamination was detected (Idexx Bioanalytics). The B16F10 cells showed minor genetic changes (addition of an allele at marker MCA-5–5), but were otherwise identical to the genetic profile established for this cell line (Idexx Bioanalytics). The B16F10 cells were checked for absence of mycoplasma using PCR before freezing aliquots and were cultured in Iscove's medium containing 10% FCS (Sigma-Aldrich), 2 mmol/L l-glutamine, 1 mmol/L sodium pyruvate, 100 U/mL penicillin, and 100 mg/mL streptomycin at 37°C for 1 week prior to start of experiments. Genetic depletion of Nox2 entails reduced B16F10 metastasis (18, 25). To achieve comparable numbers of metastases, we injected 10 × 10⁴ B1610 cells in 0.1 mL KRG intravenously (tail vein injection) per WT mouse and 15 × 10⁴ cells in 0.1 mL per Nox2-KO mouse. After B16F10 cell injection, mice were treated i.p. with idelalisib (40 mg/kg in 0.1 mL diluted in DMSO) or vehicle (control) for 5 consecutive days. Three weeks after tumor injection, mice were euthanized, and visible pulmonary metastatic foci were enumerated by light microscopy. In NK depletion experiments, mice were injected i.p. with 250 μg NK1.1 ab (Invivomab in 0.1 mL) 1 day prior to inoculation as described (25). This procedure depleted >95% of NK1.1⁺ cells (out of all CD45⁺ cells) from peripheral blood for at least 6 days (Supplementary Fig. S1F).

One-way ANOVA followed by Bonferroni's correction was used for multiple comparisons, Student t test was used in all figures, and asterisks are used as follows: *, P < 0.05; **, P < 0.01; ***, P < 0.001; and ****, P < 0.0001. All indicated P values are two-sided. The study was approved by the Ethical Review Board of Gothenburg (application no. 312-13), and all experiments were performed in accordance with the Declaration of Helsinki. Animal experiments were approved by the Research Animal Ethics Committee at the University of Gothenburg, Sweden (application no. 86-14).

PI3K is implicated in the regulation of NOX2-derived ROS formation in myeloid cells, but the role of PI3K isoforms in NOX2 regulation is incompletely understood (37, 38). In agreement with a previous publication (10), immobilized RTX but not RTX F_ab fragments triggered robust ROS production from human monocytes (Fig. 1A and B). This implies that these cells are induced to generate ROS by binding to the F_c-part of anti-CD20. The PI3K p110δ inhibitor idelalisib inhibited RTX-induced formation of extracellular ROS by monocytes (Fig. 1A and B). Dose-response analysis suggested that idelalisib inhibited RTX-induced ROS formation with an IC₅₀ of approximately 25 nmol/L (Supplementary Fig. S2A). The pan-PI3K inhibitor LY294002 and the NOX2 inhibitor DPI (Fig. 1B) exerted similar inhibition of anti–CD20-induced ROS formation.

We next defined a putative link between the PI3Kδ and NOX2 intracellular transduction pathways. Akt is a downstream target of PI3K that may activate NOX2 by phosphorylating Ser304 and Ser328 of the cytosolic subunit p47^phox (39). To determine the involvement of PI3Kδ in anti–CD20-induced NOX2 activation, human monocytes were stimulated with RTX in the presence or absence of idelalisib to assay Akt and p47^phox phosphorylation by Western blot. Figure 1C–F show RTX-induced phosphorylation of Akt at Ser473 and p47^phox at Ser304 in monocytes. The antibody-induced p47^phox phosphorylation was inhibited by idelalisib; as expected, idelalisib did not inhibit extracellular ROS production induced by PMA, which directly activates protein kinase C (Fig. 1G).

By releasing ROS, monocytes have been reported to trigger cell death by parthanatos in adjacent antineoplastic lymphocytes (23). Subsets of NK cells (CD56^dim/CD16⁺ “cytotoxic” NK cells) are particularly sensitive to ROS-induced immunosuppression and death, presumably because CD56^dim/CD16⁺ NK cells contain low levels of intracellular ROS-protective thiols (53). We determined whether CD20 antibodies affected NK cell viability in monocyte/NK cell cocultures and whether PI3K inhibitors prevented NK cell death. The addition of immobilized RTX triggered extensive NK cell death after overnight incubation with monocytes (Fig. 2A). NK cell death was prevented by the pan-PI3K inhibitor LY294002, the NOX2 inhibitor DPI, and the ROS scavenger catalase, implying that PI3K and NOX2 mediated cell death. Notably, no substantial increase in NK cell death was observed after overnight culture of NK cells with immobilized RTX in the absence of monocytes.

Monocytes from healthy subjects predominantly express the p110δ (class IB) isoform of PI3K but may also express α, β, and γ (class IA) isoforms (Supplementary Fig. S2B). Supplementary Table S1 shows the previously reported IC₅₀s (in cell-free experiments) of isoform-selective class I PI3K inhibitors. To determine effects of the PI3K inhibitors on ROS production and NK cell death triggered by RTX, we used each inhibitor at its reported IC₅₀ value. Figure 2B shows a representative experiment of the effects of PI3K inhibitors on RTX-induced ROS production from monocytes, and Fig. 2C shows box-plots summarizing results achieved in six donors. In these experiments, idelalisib consistently reduced or inhibited RTX-induced ROS formation, whereas α and γ inhibitors did not. The p110β inhibitor TGX-221 tended to inhibit ROS formation (P = 0.07; Fig. 2B and C), which may be explained by the earlier reported affinity of TGX-221 for p110δ at higher concentrations (Supplementary Table S1). Effects of PI3K inhibitors on NK cell death largely mirrored those observed in assays of RTX-induced ROS formation. Hence, idelalisib prevented NK cell death with an IC₅₀ of approximately 25 nmol/L, which was partly mimicked by a p110β inhibitor but not by α or γ inhibitors (Fig. 2D; Supplementary Fig. S2D). None of the PI3K inhibitors displayed ROS-scavenging properties, thus supporting that effects on ROS production resulted from the targeting of NOX2 (Supplementary Fig. S3).

OFA, a humanized antibody against CD20 (54), shared the property of RTX to induce ROS production from monocytes in an idelalisib-reversible fashion (Supplementary Fig. S4A and S4B). Idelalisib also prevented the ROS-dependent NK cell death induced by OFA (Supplementary Fig. S5A and S5B). In addition, the IgG1 antibody cetuximab (directed against the EGFR; ref. 55) efficiently triggered idelalisib-reversible ROS production from monocytes (Supplementary Fig. S2C), thus implying that IgG-induced ROS formation occurred independently of antigen specificity.

Next, we addressed the ability of idelalisib to restore NK cell cytotoxic responses. Freshly isolated NK cells were cocultured with RTX-coated primary CLL cells in the presence or absence of monocytes. After 4 hours, we determined the percentage of degranulating CD107⁺ NK cells in the culture (48) and determined the proportion of lysed CLL cells. As shown in Fig. 3A and B, idelalisib partially prevented loss of ADCC and degranulation responses in ROS-exposed NK cells. No substantial induction of NK cell death was observed in these short assays (Supplementary Fig. S6F). In a next series of experiments, we asked if NK cells that had been rescued from ROS-induced death by idelalisib remained functional in terms of cytotoxicity against malignant B cells. We therefore generated cocultures of NK cells and monocytes and added immobilized RTX and idelalisib followed by assay of NK cell–mediated ADCC, using anti-CD20 as the linking antibody. NK cells rescued from ROS-mediated cell death by idelalisib or other ROS inhibitors (Supplementary Fig. S6A) remained functional and exerted ADCC and degranulation against 221 cells (Fig. 3C and D). Similar results were obtained using ADCC experiments with primary CLL cells as target cells (Fig. 3E and F; gating strategy and representative FACS plots are shown in Supplementary Fig. S1D and S1E). Effects of idelalisib on NK cell degranulation and ADCC were mimicked by the pan-PI3K inhibitor Ly294002 and the ROS scavenger catalase (Fig. 3C–F). Exposure of NK cells to idelalisib in the absence of monocytes in the overnight culture did not affect their ADCC against RTX-coated primary CLL cells or 221 cells (Supplementary Fig. S6B and S6C). Incubation of NK cells with RTX in the absence of monocytes in the overnight culture did not significantly affect NK cell viability (Supplementary Fig. S6D). Monocytes triggered only minor RTX-dependent ADCC against CLL cells (Supplementary Fig. S6E). Collectively, these results implied that idelalisib exerted potent NOX inhibition while protecting NK cells against ROS-induced inactivation.

We asked if idelalisib may also prevent myeloid cell–mediated immunosuppression in the absence of ROS-inducing antibodies or other NOX2 stimuli. We first observed that idelalisib inhibited the constitutive ROS production by unstimulated human monocytes (Supplementary Fig. S7A–S7C). Next, NK cells were cocultured overnight with monocytes (in the absence of antibodies) followed by assay of ADCC against primary human CLL cells. Idelalisib prevented NK cell death in these experiments (Supplementary Fig. S7D) and maintained NK cell cytotoxicity against primary CLL cells (Supplementary Fig. S7E and S7F), thus suggesting that idelalisib preserved NK cell function and viability also in the absence of exogenous antibodies.

We used genetically Nox2-deficient and NK cell–depleted mice to determine if the NOX2-inhibitory properties of idelalisib translated into improved NK cell–mediated clearance of malignant cells in vivo. Pharmacologic or genetic inhibition of NOX2 reduces lung metastasis from B16F10 melanoma cells in mice; in this model, NK cells are significant mediators of tumor cell clearance (25). Ex vivo experiments verified that idelalisib inhibited constitutive ROS production by unstimulated murine myeloid Gr1⁺ cells (B6 strain; Fig. 4A). Gr1⁺ cells from Nox2-KO mice have no detectable ROS production by chemiluminescence (25). To assess effects of idelalisib on B16F10 metastasis, B6 mice were inoculated i.v. with B16F10 cells and treated with idelalisib (40 mg/kg x V) or vehicle followed by enumeration of melanoma metastases in lungs 3 weeks later (Fig. 4B). We observed significantly reduced melanoma metastasis in the idelalisib-treated WT mice. This effect was absent in Nox2-KO mice and in NK cell–depleted mice (Fig. 4C–E; Supplementary Fig. S1F), suggesting that the antimetastatic properties of idelalisib were NK cell–dependent and resulted from regulation of NOX2.

Idelalisib is used in conjunction with RTX in patients with B-CLL and B cell lymphoma. Idelalisib inhibits PI3K targets in BCR signaling to promote apoptosis of malignant B cells (30). Here, we presented results suggesting a supplementary antineoplastic mechanism for idelalisib, where it prevents Akt-dependent phosphorylation of the NOX2 component p47^phox in human myeloid cells and blocks the formation and release of immunosuppressive ROS, thus improving NK cell survival and cytotoxicity against leukemic cell lines and primary leukemic cells. Studies in experimental animals showed that idelalisib reduced ROS formation from Gr1⁺ cells and improved NK cell–dependent clearance of metastatic cells in WT mice but not in mice genetically deficient of Nox2. These data imply that idelalisib may exert antineoplastic effects in vivo by targeting NOX2. Notably, higher doses of idelalisib were required to inhibit ROS production from murine Gr1⁺ cells in vitro compared with human monocytes. This may imply that human monocytes are more sensitive to idelalisib than their murine counterparts.

Our study aimed at defining the interplay between PI3K and NOX2 that leads to NOX2 activation and subsequent ROS production. Exposure of phagocytes to any of a variety of stimuli, including immobilized IgG antibodies, may activate the enzyme to generate large amounts of extracellular superoxide anion. A pivotal aspect of NOX2 activation is the phosphorylation of the cytosolic subunit p47^phox (56, 57). Monocytes from patients with p47^phox-deficient chronic granulomatous disease (CGD) do not generate ROS via NOX2 and are therefore highly susceptible to bacterial infections. Patients with CGD are also prone to develop autoimmunity, presumably as a consequence of deficient ROS-mediated inactivation of lymphocyte-mediated autoimmunity (18, 58). There is an established role for PKC in p47^phox phosphorylation (56, 59, 60), but studies using PI3K inhibitors imply that the PI3K pathway also activates NOX2 (61, 62). We observed significant phosphorylation of Akt at Ser473 and p47^phox at Ser304 in monocytes shortly after stimulation with RTX, which was significantly inhibited by idelalisib. The Akt and p47^phox phosphorylation coincided with peak ROS production in stimulated monocytes, thus supporting the involvement of the PI3K–AKT pathway in antibody-induced NOX2 activation.

The role of PI3K isoforms in regulating NOX2-derived ROS has remained unknown. We used structurally distinct isoform-specific inhibitors of class I PI3K (Supplementary Table S1; Supplementary Fig. S8) and found that inhibition of the p110δ isoform using idelalisib efficiently blocked RTX-induced ROS production by human monocytes, whereas inhibitors of other isoforms exerted minor (p110β) or no (p110α and p110γ) effects on ROS production. Corresponding results were obtained in studies of protection of NK cells from ROS-induced cell death. These findings support that idelalisib primarily acts by blocking p110δ in reducing ROS formation and protecting NK cells. However, the potency and specificity of PI3K inhibitors may be context-dependent, and our results do not formally exclude that inhibition of isoforms of class I PI3K other than p110δ may have contributed to the observed ROS-inhibitory and immunostimulatory effects of idelalisib.

Our data suggest a mechanism by which idelalisib may restore NK cell function in an environment of NOX2-mediated oxidative stress. Although this study was focused on protection of NK cells against NOX2-derived ROS, subsets of T cells are sensitive to this mechanism of immunosuppression (63). In this context, global genetic depletion of the p110δ PI3K in mice impedes the development of murine CLL by restoring T-cell–mediated immunity (64). Further studies are required to determine if idelalisib, or other PI3Kδ inhibitors, may promote T-cell–mediated antitumor immunity by reducing NOX2-induced immunosuppression.

Our findings may be relevant beyond B-cell malignancies. For example, NOX2-dependent immunosuppression occurs in acute myeloid leukemia (21, 65). p110δ inhibitors are also effective in nonhematopoietic cancer. Thus, genetic or pharmacologic inactivation of p110δ was proposed to target the function of regulatory T cells, thus unleashing antitumor responses in experimental solid cancer (66). Using an in vivo melanoma model, we observed that inactivation of p110δ by idelalisib significantly reduced B16 lung metastasis formation in an NK cell–dependent and NOX2-dependent manner. These results thus support that pharmacologic inhibition of p110δ may be useful in nonhematopoietic cancer (66) and imply that idelalisib may exert antineoplastic efficacy by preventing NOX2-mediated immunosuppression.

Patients treated with idelalisib are at risk for several toxicities such as autoimmunity and severe infections, including opportunistic infections (31–36). Reduced NOX2 function (in human CGD and in Nox2^−/− mice) makes patients susceptible to microbial infections and autoimmunity (18). It may thus be speculated that the NOX2-inhibitory properties of idelalisib contribute to these toxicities. Notably, idelalisib reduced ROS formation from monocytes at concentrations similar to those detected in blood of idelalisib-treated patients (34), thus implying that idelalisib therapy is likely associated with significant NOX2 inhibition.

This study suggests an antineoplastic mechanism for PI3K p110δ blockade beyond interference with BCR signaling in malignant B cells. By targeting the p110δ/Akt signal transduction pathway that drives NOX2-dependent release of immunosuppressive ROS from myeloid cells, p110δ inhibitors may exert activity in a broader range of malignancies than previously appreciated.

A. Martner and K. Hellstrand are authors of patents PCT/US2019/064151, PCT/US2019/045161, PCT/US2018/043693, and PCT/US2018/043543. A. Martner, F.B. Thorén, and K. Hellstrand are authors of patents PCT/US2019/038172 and WO2017137989A4. J. Aurelius, A. Martner, F.B. Thorén, and K. Hellstrand are authors of patent PCT/2018/040037. No potential conflicts of interest were disclosed by the other authors.

A.A. Akhiani: Conceptualization, data curation, validation, investigation, methodology, writing–original draft, project administration, writing–review and editing. A. Hallner: Conceptualization, investigation, methodology, writing–original draft. R. Kiffin: Methodology. E. Aydin: Methodology. O. Werlenius: Data curation, methodology. J. Aurelius: Supervision. A. Martner: Conceptualization, resources, supervision, methodology. F.B. Thorén: Conceptualization, resources, formal analysis, supervision, investigation, methodology, writing–original draft, project administration, writing–review and editing. K. Hellstrand: Conceptualization, resources, data curation, supervision, funding acquisition, investigation, writing–original draft, project administration, writing–review and editing.

This work was supported by the Swedish Research Council, the Swedish Cancer Society, the Swedish state via the ALF agreement, the IngaBritt and Arne Lundberg Research Foundation, the Clas Groschinsky Foundation, the Åke Wiberg Foundation, the Assar Gabrielsson Foundation, the Lions Cancer Foundation West, the Wilhelm and Martina Lundgren Research Foundation, and BioCARE.

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.