Pathological activation of the Toll-like receptor signaling adaptor protein MYD88 underlies many autoimmune and inflammatory disease states. In the activated B cell–like (ABC) subtype of diffuse large B cell lymphoma (DLBCL), the oncogenic MYD88 L265P mutation occurs in 29% of cases, making it the most prevalent activating mutation in this malignancy. IRAK4 kinase accounts for almost all of the biological functions of MYD88, highlighting IRAK4 as a therapeutic target for diseases driven by aberrant MYD88 signaling. Using innovative structure-based drug design methodologies, we report the development of highly selective and bioavailable small molecule IRAK4 inhibitors, ND-2158 and ND-2110. These small molecules suppressed LPS-induced TNF production, alleviated collagen-induced arthritis, and blocked gout formation in mouse models. IRAK4 inhibition promoted killing of ABC DLBCL lines harboring MYD88 L265P, by down-modulating survival signals, including NF-κB and autocrine IL-6/IL-10 engagement of the JAK–STAT3 pathway. In ABC DLBCL xenograft models, IRAK4 inhibition suppressed tumor growth as a single agent, and in combination with the Bruton’s tyrosine kinase (BTK) inhibitor ibrutinib or the Bcl-2 inhibitor ABT-199. Our findings support pharmacological inhibition of IRAK4 as a therapeutic strategy in autoimmune disorders, in a genetically defined population of ABC DLBCL, and possibly other malignancies dependent on aberrant MYD88 signaling.
Autoimmune disorders and B cell malignancies arise from pathological expansion of B lymphocytes. In autoimmune diseases, B cells react to self-antigens using both the B cell receptor (BCR) and TLRs (Leadbetter et al., 2002; Lau et al., 2005; Ehlers et al., 2006; Marshak-Rothstein, 2006; Green and Marshak-Rothstein, 2011). Activation of TLRs (except TLR3) recruits the signaling adaptor protein MYD88, which engages the interleukin-1 receptor–associated kinases IRAK4 and IRAK1, thereby activating the NF-κB and type-1 IFN pathways to promote survival and expansion of autoantibody-producing B cells (Leadbetter et al., 2002; Lau et al., 2005).
In the activated B cell–like (ABC) subtype of diffuse large B cell lymphoma (DLBCL), recurrent mutations targeting the MYD88 TIR domain occur in 39% of cases, with one dominant mutation, L265P, occurring in 29% of cases (Ngo et al., 2011). MYD88 L265P is absent or rare in most other subtypes of lymphoma, including the germinal center B cell–like (GCB) DLBCL subtype (Ngo et al., 2011). MYD88 L265P is unique among MYD88 mutant isoforms in its ability to coordinate a stable signaling complex involving IRAK4 and IRAK1, in which IRAK4 phosphorylates IRAK1, causing constitutive NF-κB activation, type I IFN signaling, and autocrine IL-6/IL-10 engagement of the JAK–STAT3 pathway (Ngo et al., 2011). The ability of mutant MYD88 to sustain the survival of ABC DLBCL cell lines requires the kinase activity of IRAK4, whereas IRAK1 kinase activity is dispensable; however, IRAK1 protein appears to perform a necessary scaffolding function (Ngo et al., 2011). Together, these data support the development of IRAK4-selective kinase inhibitors for the treatment of ABC DLBCL tumors expressing oncogenic MYD88 mutant isoforms.
The design of inhibitors of BCR signaling is a major focus of therapeutic development for the treatment of ABC DLBCL, given the dependence of ABC DLBCL on BCR signaling and the fact that gain-of-function mutations targeting the BCR subunits CD79A and CD79B occur frequently in this lymphoma subtype (Davis et al., 2010). Although the role of MYD88 in maintaining viability of ABC DLBCL lines is equally important (Ngo et al., 2011), the utility of small molecule therapeutics targeting MYD88 signaling remains largely undetermined. Given that IRAK4 mediates most, if not all, of the biological effects of MYD88, inhibition of IRAK4 is an attractive therapeutic approach to block pathological MYD88 signaling (Kawai et al., 1999; Suzuki et al., 2002; Kim et al., 2007). However, despite great interest in IRAK4 as a therapeutic target, the development of selective inhibitors has been confounded by the challenging structure of the IRAK4 catalytic domain. In addition, the generation of small molecules with the necessary properties suitable for in vivo testing has proven difficult (Wang et al., 2009; Chaudhary et al., 2015), leaving the potential of IRAK4 targeting in human malignancies relatively unexplored. In this study, we describe the discovery, properties, and initial in vivo pharmacological characterization of two compounds in the thienopyrimidine class that are potent, highly selective, and bioavailable small molecule IRAK4 inhibitors for the treatment of autoimmune disorders and B cell malignancies.
RESULTS AND DISCUSSION
Discovery of potent and selective IRAK4 inhibitors
Using a published cocrystal structure of IRAK4 (Wang et al., 2006), we conducted a virtual screen of a commercially available library of 1.3 million compounds. Confirmation of IRAK4 inhibition by several virtual screen hits resulted in the identification of a compound containing a thienopyrimidine motif. Elaboration of this early hit led to the identification of two highly potent and selective IRAK4 inhibitors, ND-2158 and ND-2110, which are competitive inhibitors that bind in the ATP pocket and are built upon novel structural scaffolds (Fig. 1 A). The in vitro inhibitory constants (Ki) versus IRAK4 kinase for ND-2158 and ND-2110 were 1.3 nM and 7.5 nM, respectively (Fig. 1 B). When tested against 334 kinases, these inhibitors were highly selective for IRAK4 (Fig. 1 C and Table S1) and exhibited attractive drug-like properties including solubility, cell permeability, and a pharmacokinetic profile suitable to pharmacologically interrogate IRAK4-dependent mechanisms in cells and animal disease models (Fig. 1 D).
IRAK4 inhibitors block TNF production, collagen-induced arthritis, and gout formation in mice
The importance of TLR signaling in inflammatory cell responses (Kawai et al., 1999; Suzuki et al., 2002; Kim et al., 2007) prompted us to investigate the use of IRAK4 inhibitors in this setting. ND-2158 or ND-2110 potently blocked production of the proinflammatory cytokine TNF by human white blood cells (WBCs) stimulated with the TLR4 agonist LPS or the TLR9 agonist CpG; however, the structurally related compound ND-1659, which does not inhibit IRAK4 kinase activity, was inactive (Fig. 1, A and B; and Fig. 2 A). ND-2158 or ND-2110 also significantly reduced serum TNF levels in a dose dependent manner in vivo (Fig. 2 B and not depicted). In mouse models of arthritis and inflammatory gout, signaling through the MYD88–IRAK4 complex is required for development of overt disease (Choe et al., 2003; Chen et al., 2006; Ehlers et al., 2006; Pierer et al., 2011). In a collagen-induced arthritis mouse model (Bendele, 2001), pharmacological IRAK4 inhibition was effective within 2 d of initiating treatment (Fig. 2 C, dark blue squares). Inflammation reduction was examined by histological evaluation of inflammatory cell influx, cartilage damage, bone resorption, and periosteal bone formation (Fig. 2 D). In a mouse model of MYD88-dependent inflammatory gout (Chen et al., 2006), IRAK4 inhibition blocked leukocyte infiltration in a dose-dependent manner, with the higher doses achieving more than twofold reduction in leukocyte influx compared with vehicle control (Fig. 2 E). Together, our data demonstrate that these IRAK4 inhibitors can interrupt IRAK4-dependent responses in vitro and in vivo.
Pharmacological IRAK4 inhibition in
MYD88 L265P–expressing ABC DLBCL
Activating oncogenic mutations in MYD88, particularly the L265P variant, are pathognomonic for ABC DLBCL (Ngo et al., 2011), leading us to explore the activity of small molecule IRAK4-selective inhibitors in this malignancy. The in vitro viability of multiple ABC DLBCL lines was reduced by ND-2158 and ND-2110, but not the control compound ND-1659, whereas GCB DLBCL lines with WT MYD88 were unaffected (Fig. 3 A and not depicted). ND-2158 and ND-2110 targeted the subset of ABC DLBCL lines with MYD88 L265P mutations, but not ABC DLBCL lines with other MYD88 mutations (S222R and S219C) or those with WT MYD88 (Fig. 3 A and not depicted). In a mouse xenograft model created using the ABC DLBCL line OCI-Ly10 (MYD88 L265P+), the IRAK4 inhibitor ND-2158 retarded tumor growth and was well tolerated (Fig. 3, B and C) and reduced IRAK4 phosphorylation (Thr-345/Ser-346) in the tumors (Fig. 3 D).
Effect of IRAK4 inhibitors on oncogenic MYD88 signaling in ABC DLBCL
IRAK4 phosphorylates three residues within its activation loop: Thr-342, Thr-345, and Ser-346 (Cheng et al., 2007). This autophosphorylation event is the first step required for IRAK4 activation and leads to IRAK1 phosphorylation and the constitutive NF-κB signaling that sustains ABC DLBCL survival (Cheng et al., 2007; Ngo et al., 2011; Ferrao et al., 2014). When we treated ABC DLBCL lines (MYD88 L265P+) with ND-2158, ND-2110, or the negative control compound ND-1659, coimmunoprecipitation analysis revealed that IRAK4 associates with IRAK1 and is autophosphorylated in this complex (Fig. 3 E). Pretreatment of cells with IRAK4 inhibitors decreased the level of IRAK4 autophosphorylation (Thr-345/Ser-346) relative to cells treated with the inactive compound ND-1659 (Fig. 3 E).
Oncogenic MYD88 signaling engages the NF-κB pathway by activating IκB kinase β (IKKβ)–mediated phosphorylation and degradation of IκB proteins (Ngo et al., 2011). Treatment of ABC DLBCL lines with the IRAK4 inhibitors suppressed IKK activation, as indicated by decreased phosphorylation of IKKβ (Fig. 3 F), and stabilization of an IκBα-luciferase reporter, which measures IKK activity (Fig. 3 G; Lam et al., 2005). Moreover, in four MYD88 L265P+ ABC DLBCL lines, ND-2158 or ND-2110 decreased transcription of an NF-κB–dependent luciferase reporter (Fig. 3 H) and reduced secretion of IL-6 and IL-10 (Fig. 3 H), two cytokines that depend on MYD88-mediated NF-κB activation for their production (Ngo et al., 2011).
To explore the biological consequences of inhibiting IRAK4 activity in ABC DLBCL, we profiled gene expression changes over time in four MYD88 L265P+ ABC DLBCL lines treated with ND-2158 or the negative control ND-1659. We defined an IRAK4 gene expression signature consisting of 366 down-modulated genes, which was highly enriched for genes whose expression depends on MYD88 activity (Fig. 3 I and Table S2; sevenfold enrichment, P = 2.3 × 10−21; Ngo et al., 2011). When the overlap between the IRAK4 signature and previously defined gene expression signatures was examined, the most significant enrichment was with signatures reflecting NF-κB signaling (Fig. 3 I and Fig. S1 and Table S2; 13-fold enrichment, P = 5.3 × 10−40).
Autocrine IL-6 and IL-10 production in ABC DLBCL drives JAK–STAT3 signaling and promotes ABC DLBCL survival (Lam et al., 2008). ND-2158 reduced gene expression of IL6 and IL10 (Fig. 3 J), and accordingly, a signature of JAK kinase signaling in ABC DLBCL overlapped significantly with the IRAK4 signature (Fig. 3 I, Fig. S1, and Table S2; fivefold enrichment, P = 1.7 × 10−11). MYD88 signaling also induces IFN-β production in ABC DLBCL cells (Ngo et al., 2011; Yang et al., 2012), and a subset of genes induced by type I IFN signaling were also components of the IRAK4 signature (Fig. 3 I, Fig. S1, and Table S2; fivefold enrichment, P = 7.2 × 10−13). Thus, pharmacological IRAK4 inhibition blunts multiple biological responses downstream of oncogenic MYD88 signaling in ABC DLBCL.
Combined inhibition of oncogenic MYD88 and BCR signaling
Bruton’s tyrosine kinase (BTK) is a promising therapeutic target in ABC DLBCL (Wilson et al., 2015). BTK is a crucial kinase linking proximal BCR signaling to downstream IKK activation and maintains NF-κB activity (Davis et al., 2010). In many ABC DLBCL lines, both chronic active BCR signaling and constitutive MYD88 signaling converge on IKK, suggesting that dual inhibition of these two pathways could have synergistic effects on ABC DLBCL killing. Combination of the BTK inhibitor ibrutinib with ND-2158 enhanced killing of BCR-dependent ABC DLBCL that harbor coincident MYD88 L265P mutations and activating mutations in the BCR subunits CD79A and CD79B (OCI-Ly10 and TMD8). In contrast, no increased killing was observed for a GCB DLBCL line (OCI-Ly19) that does not rely on either MYD88 or BCR signaling (Fig. 4 A). Isobologram analysis confirmed that ND-2158 and ibrutinib were synergistic (Fig. 4, B and C). The combination was also more effective in inhibiting IKK activity (Fig. 4 D) and enhancing apoptosis (Fig. 4 E). ND-2158 and ibrutinib also cooperated to suppress the in vivo growth of OCI-Ly10 xenografts (Fig. 4 F), laying the foundation for combined use of ibrutinib and IRAK4 inhibitors in future clinical trials.
A subset of ABC DLBCLs with chronic active BCR signaling require spleen tyrosine kinase (Syk) to deliver signals from the BCR to downstream survival pathways, including NF-κB and PI(3) kinase (Davis et al., 2010). ND-2158 and the selective Syk inhibitor PRT062607 (Coffey et al., 2012) were synergistic in killing ABC DLBCL lines that depend on both BCR and MYD88 signaling (OCI-Ly10, TMD8, and HBL1), but a control GCB DLBCL line was unaffected (Fig. 5, A–C). The observed cytotoxicity was associated with increased apoptosis and decreased IKK activity in cells treated with the combination compared with either drug alone (Fig. 5, D and E).
Combined inhibition of MYD88 and Bcl-2 signaling in ABC DLBCL
The antiapoptotic protein Bcl-2 provides a powerful survival signal in ABC DLBCL and is characteristically expressed at high levels as the result of transcriptional up-regulation and amplification of the BCL2 locus (Alizadeh et al., 2000; Lenz et al., 2008). The Bcl-2 inhibitor ABT-199 synergistically enhanced the ability of ND-2158 to decrease viability of ABC DLBCL lines (Fig. 6 A) and promote apoptosis, as indicated by more cells expressing activated caspase-3 and cleaved PARP (Fig. 6 B). Of note, this synergy was observed in the CARD11 mutant ABC line OCI-Ly3 that does not rely on BCR signaling and is insensitive to ibrutinib (Fig. 6 A; Davis et al., 2010). In vivo, ND-2158 or ABT-199 moderately suppressed growth of OCI-Ly3 xenografts as single agents, but together they produced near complete inhibition of tumor growth (Fig. 6 C). In addition to Bcl-2, ABC DLBCLs express three other antiapoptotic Bcl-2 family members: BCL2L1, MCL1, and BCL2A1 (encoding Bcl-xL, Mcl-1, and A1, respectively). The IRAK4 inhibitor ND-2158 decreased BCL2L1 and BCL2A1 mRNA levels but increased BCL2 expression (Fig. 6 D), suggesting that the synergy between ND-2158 and ABT-199 could be caused by combined inhibition of all three family members (Bcl-xL, A1, and Bcl-2).
In summary, the potent and selective IRAK4 kinase inhibitors reported here provide insights into disease mechanisms and suggest new therapeutic strategies for patients with inflammatory diseases and B cell malignancies. Recently, two papers have been published that describe IRAK4 inhibitors with in vivo efficacy in LPS-induced TNF models and in an antibody-induced mouse arthritis model of inflammation (Tumey et al., 2014; McElroy et al., 2015). The compounds described by Tumey et al. (2014) are flat tetracyclic aromatic compounds that possess high protein binding (>99%), low permeability, and high efflux, which is reflected in poor oral bioavailability. McElroy et al. (2015) describe an amidopyrazole IRAK4 inhibitor with good solubility and modest protein binding, which exhibits paw size reduction in a mouse model of arthritis. However, a subsequent publication by the same group (Lim et al., 2015) describes the difficulty in obtaining desired drug-like properties (selectivity, permeability, and cell potency) in this series of compounds. In contrast, our thienopyrimidine compounds described herein possess excellent selectivity and solubility, low protein binding, good permeability and cell potency, and in vivo activity in several autoimmune and tumor efficacy models.
Despite the advent of anti-TNF biologics and JAK kinase inhibitors, many patients with autoimmune or chronic inflammatory diseases must still be treated with broad-acting immunosuppressive agents. We have shown that pharmacological inhibition of IRAK4 ameliorates clinical progression of collagen-induced arthritis and blocks gout formation in mouse models.
Among mature B cell malignancies, mutations affecting MYD88 have emerged as one of the most prevalent gain-of-function genetic alterations. Our work establishes that oncogenic MYD88 signaling is a tractable therapeutic target in human lymphomas and supports the development of IRAK4 inhibitors for treatment of the genetically defined subset of ABC DLBCL harboring the MYD88 L265P mutation. Our systematic exploration of small molecule combinations also revealed the value of IRAK4 blockade in enhancing the efficacy of small molecules that target orthogonal ABC DLBCL survival pathways. Thus, IRAK4 inhibitors may have clinical utility in ABC DLBCL, either alone or in combination with pharmacological blockade of the BCR or Bcl-2 pathways. The therapeutic potential of IRAK4 inhibitors may also extend beyond ABC DLBCLs because MYD88 L265P is prevalent in multiple mature B cell lymphoma subtypes, including Waldenström’s macroglobulinemia (91%; Treon et al., 2012), primary central nervous system lymphoma (75%; Kraan et al., 2013), primary cutaneous DLBCL–leg type (69%; Pham-Ledard et al., 2012), primary testicular DLBCL (68%; Kraan et al., 2014), MALT lymphoma (6%; Ngo et al., 2011; Li et al., 2012), and chronic lymphocytic leukemia (3%; Puente et al., 2011).
MATERIALS AND METHODS
Synthesis of ND-2158 and ND-2110 (examples 106 and 29) has been reported previously (Corin et al., 2012).
Kinase activity was evaluated using radioisotope-based enzymatic assays, and the experiments were conducted at Reaction Biology Corp. (Ma et al., 2006). IC50 and Ki of ND-2158 and ND-2110 were obtained in dose inhibition assays against IRAK4 and IRAK1. Broad kinase selectivity was evaluated against 334 kinases at one concentration (10 µM) of ND-2158 or ND-2110. IC50 and Ki were subsequently determined for kinases that showed ≥75% inhibition at 10 µM of the compounds. Dose–response curves were generated with Prism software (GraphPad Software) using four-parameter fit.
In vitro solubility of ND-2110 and ND-2158
300 µM of either ND-2110 or ND-2158 was prepared in Dulbecco’s phosphate-buffered saline (D-PBS), pH 7.4, containing 0.1% DMSO. Samples were incubated at 25°C for 2 h with shaking using a plate shaker. At the end of the incubation, samples were filtered, the filtrate was diluted 100-fold with methanol, and the drug level was determined using the API 4000 Q TRAP LC/MS/MS system (Applied Biosystems).
DLBCL cell lines were cultured in RPMI 1640 medium (Gibco) supplemented with penicillin/streptomycin (Gibco) and 20% fetal bovine serum. For the OCI series of DLBCL cell lines, Iscove’s medium supplemented with 100 U/ml penicillin/streptomycin and 20% fresh human plasma was used. Cells were propagated in humidified, 5% CO2 incubators at 37°C. THP-1 cells from ATCC (TIB-202) were cultured in THP-1 media containing RPMI 1640, 10% fetal bovine serum, 100 U/ml penicillin/streptomycin, and 50 µM 2-Mercaptoethanol (Invitrogen).
LPS, CpG, and R848-induced cytokine release
Human WBCs were isolated from the Buffy coat by lysing red blood cells. Cells were preincubated for 1 h with ND-2158, ND-2110, or ND-1659 at concentrations shown in Fig. 2 A, with a final concentration of 0.3% DMSO, before stimulation with the indicated agonists. WBCs were plated in RPMI 1640 containing 100 U/ml penicillin/streptomycin and stimulated with 1 µg/ml LPS (InvivoGen), 0.5 µM CpG (InvivoGen), or 1 µM of the TLR7/8 inhibitor R848 (InvivoGen) for 20 h (Jurk et al., 2011; Matera et al., 2012). For R848-induced cytokine release assays (Fig. 1 D), human whole blood was collected in sodium heparin vacutainer tubes and diluted with an equal volume of RPMI 1640 for 30 min. Blood was then stimulated with 1 µM R848 for 20 h, and plasma was collected at the end of stimulation for TNF analysis. THP-1 cells were plated in THP-1 media (RPMI 1640, 10% fetal bovine serum,100 U/ml penicillin/streptomycin, and 50 µM 2-Mercaptoethanol) and stimulated with 300 ng/ml LPS for 5 h. All assays were performed in 96-well plates. TNF production in the WBC culture supernatant or human whole blood was detected using the TNF ELISA kit (R&D Systems). TNF production in the THP-1 cell supernatant (Fig. 1 D) was detected using the TNF HTRF kit (Cisbio), as per manufacturer instructions.
Gene expression array and gene expression signature enrichment
Four ABC DLBCL cell lines (OCI-Ly10, TMD8, HBL1, and OCI-Ly3), were treated with either 10 µM ND-2158 or 10 µM of the structurally related negative control compound ND-1659 for 6, 12, 24, or 36 h in culture. Cells were harvested at each time point, and RNA for gene expression was prepared from 5 million cells using TRIzol (Invitrogen) extraction, followed by clean up using the RNeasy kit (QIAGEN). Gene expression profiling was performed using two-color human Agilent 4 × 44K gene expression arrays (Agilent Technologies), as described by the manufacturer, comparing signal from control compound-treated (ND-1659) control cells (Cy3) with cells treated with ND-2158 (Cy5) for the indicated times. Array elements were filtered for those meeting confidence thresholds for spot size, architecture, and level above local background. These criteria are a feature of the Agilent Technologies gene expression software package for Agilent 4 × 44K arrays. Gene expression signature enrichment was performed by comparing changes in gene expression from at least two ABC DLBCL cell lines, occurring in at least two time points after ND-2158 treatment. Gene expression signature enrichment analysis was performed as described previously (Yang et al., 2012). Gene expression data has been deposited under GEO accession no. GSE63029.
NF-κB and IκBα reporter assays
The generation of stable NF-κB firefly luciferase reporter ABC DLBCL lines (OCI-Ly10, TMD8, HBL1, and OCI-Ly3) has previously been described (Ceribelli et al., 2014). For drug response experiments, cells were seeded at a density of 2.5 × 105 cells per ml in 96-well plates. Cells were treated with drugs for the indicated times (as described in the figure legends). Firefly luciferase and Renilla activity was measured using the Dual-Glo Luciferase Reporter Assay System (Promega) according to the manufacturer’s protocol. Luminescence from equivalent amounts of lysate was read in triplicate on a Microtiter Plate Luminometer (Dyn-Ex Technologies). Relative NF-κB activity was determined by normalizing titers of firefly to Renilla luciferase. Generation of the IκBα photinus-luciferase reporter cell line has been described previously (Lam et al., 2005). Relative IκBα activity was determined by normalizing titers of firefly to Renilla luciferase.
Immunoblotting and coimmunoprecipitation
Cells were lysed in an endogenous lysis buffer (20 mM Tris-HCl, pH 7.6, 150 mM NaCl, 1 mM EDTA, 1% Triton X-100, 30 mM NaF, and 2 mM sodium pyrophosphate) supplemented with complete protease inhibitor cocktail (Roche), phosphatase inhibitor tablet (Roche), 1 mM DTT, 1 mM Na3VaO4, and 1 mM PMSF. Samples were cleared by centrifugation at 14,000 rpm, and protein concentration was measured using the BCA Protein Assay kit (Thermo Fisher Scientific). For immunoblotting, total proteins were separated on 4–12% SDS–polyacrylamide gels and transferred to nitrocellulose membranes. For immunoprecipitation experiments, samples were first precleared with Protein G Dynabeads (Invitrogen) for 60 min. Cleared lysates were incubated overnight with an anti-IRAK1 antibody (Santa Cruz Biotechnology, Inc.), followed by Protein G Dynabeads for an additional 2 h at 4°C. Immunoprecipitates were washed five times with 0.5 M NaCl lysis buffer, separated by SDS-PAGE, transferred to nitrocellulose, and analyzed by immunoblotting. Antibodies used were as follows: anti-IKKβ, anti–phospho-IKKβ, anti–phospho-IRAK4 (pThr-345/Ser-346), and anti-IRAK4 (total) from Cell Signaling Technologies; and anti–β-actin and anti-IRAK1 from Santa Cruz Biotechnology, Inc. Secondary HRP-conjugated antibodies were obtained from GE Healthcare.
MTS viability assay
Approximately 104 cells (per well) of each DLBCL cell line were cultured in triplicate in a 96-well plate, in the presence of vehicle (DMSO) or the indicated drugs on day 0. On day 2, each well was replenished with freshly prepared vehicle or drugs dissolved in culture medium. On day 4, cell viability was assayed by adding 20 µl/well of 3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulphophenyl)-2H tetrazolium and an electron coupling reagent (phenazine methosulphate; Promega) and incubating at 37°C for 2 h. Cell viability was determined by measuring the absorbance at 490 nm using a 96-well plate reader. The background was subtracted using a medium-only control. For single agent drug response experiments, the data were normalized to DMSO-treated cells. For combination drug response experiments, data were first normalized to DMSO-treated cells and then to cells treated with the IRAK4 inhibitor ND-2158 alone. A shift of the drug combination curves to the left indicates a more than additive effect of the combination.
Measurement of apoptosis
The activation of apoptosis was measured by intracellular flow cytometry methods (Sciammas et al., 2006). DLBCL lines were cultured for 72 h in the presence of the indicated drugs. Apoptosis was measured by intracellular FACS for activated caspase-3 and cleaved PARP. Apoptotic cells were defined as the percentage of cells double-positive for activated caspase-3 and cleaved PARP. Antibodies used were as follows: anti–active caspase-3 (PE; BD) and anti–cleaved PARP (Asp-214; Alexa Fluor 647; BD), according to the manufacturer’s instructions.
IL-6 and IL-10 ELISA
Cytokine measurement was performed on ABC DLBCL lines cultured for 48 h with ND-2158, ND-2110, ND-1659, or vehicle (DMSO) control. The concentration of IL-6 in cell culture supernatants was measured using the BD OptEIA human IL-6 ELISA kit (BD). The concentration of IL-10 in cell culture supernatants was measured using the BD OptEIA human IL-10 ELISA kit (BD), according to the manufacturer’s protocol. Each experiment was performed in triplicate.
CalcuSyn and isobologram analysis
A crossover dose response grid was set up to measure cell viability effects of drug combinations. Combination index (CI) and fraction affected (Fa) for each dose combination was calculated from the experimental data by using the software package CalcuSyn (Biosoft). Drug synergism, addition, and antagonism are defined by CI values of <1.0, 1.0, and >1.0, respectively. Fa represents the inhibitory effect for each combination. Normalized isobolograms were generated to describe combined drug effect. Each isobologram contains a line defining additively, adopted to distinguish synergistic and antagonistic interactions of the drug combinations (Chou, 2010).
Prism 6 software (GraphPad Software) was used for statistical analysis of data, except for tumor xenograft experiments, which were analyzed using Spotfire S+ 8.2 (TIBCO Software Inc.). For tumor xenograft drug efficacy experiments, tumor volume was modeled using a mixed effects model, assuming a linear relationship between time and log tumor volume, with individual mice representing random effects. P-values represent the significance of the time and treatment interaction term on tumor volume.
Mouse pharmacokinetic analysis
Pharmacokinetic experiments for ND-2110 were performed by dosing DBA1 mice either i.v. at 3 mg/kg in 10% 2-hydroxypropyl-β-cyclodextrin, or by oral gavage at 10 mg/kg in 0.5% methylcellulose. Blood was collected into sodium heparin tubes before dose and at 0.083, 0.25, 0.5, 1, 2, 4, 6, and 8 h after dosing for the i.v. administration arm or 0.25, 0.5, 1, 2, 4, 6, and 8 h after dosing for the oral gavage administration arm. Plasma was prepared and ND-2110 level was quantified by LC/MS/MS with API 5000 Triple Quadropole System (Applied Biosystems). Pharmacokinetic experiments for ND-2158 were performed by dosing male C57BL/6 mice either i.v. or i.p. at 3 mg/kg in 10% 2-hydroxypropyl-β-cyclodextrin. Blood was collected into sodium heparin tubes before dose and at 0.083, 0.25, 0.5, 1, 2, 4, 6, 8, 12, and 24 h after dosing. Plasma was prepared and ND-2158 level was quantified by LC/MS/MS using the API 4000 Q TRAP system (Applied Biosystems). Pharmacokinetic parameters were calculated using Phoenix WinNonlin.
Tumor xenograft models
All animal experiments were performed in accordance with the National Cancer Institute/National Institutes of Health Animal Care and Use Committee (NCI ACUC) guidelines and approval. Female NOD/SCID mice (6–8 wk of age) were inoculated by s.c. injection of a tumor cell suspension containing Matrigel (Matrigel Basement Membrane Matrix; BD)/PBS at a ratio of 1:1. For the OCI-Ly10 ABC DLBCL line, 10 × 106 cells were inoculated. For the OCI-Ly3 ABC DLBCL line, 106 cells were inoculated. Mice were monitored daily for palpable tumor mass. When tumors reached a mean of 150 mm3, treatment was initiated. ND-2158 was prepared in 10% hydroxypropyl-β-cyclodextrin in ddH20. ND-2158 (100 mg/kg/day) or its vehicle was administered i.p. twice per day. Ibrutinib (Pharmacyclics) was prepared in DMSO/10% hydroxypropyl-β-cyclodextrin in ddH20 (5:95). Ibrutinib (9 mg/kg/day) or its vehicle was administered i.p. once per day. For ND-2158/ibrutinib combination experiments, ibrutinib was administered 4 h after ND-2158. ABT-199 (Selleckchem) was prepared in Phosal 50 PG/polyethylene glycol-400/ethanol (6:3:1) as previously described (Souers et al., 2013). ABT-199 (50 mg/kg/day) or its vehicle was administered once per day by oral gavage. For ND-2158/ABT-199 combination experiments, ABT-199 was administered 4 h after ND-2158. Tumor growth was monitored daily by measuring tumor size in two orthogonal dimensions by digital caliper measurement. Tumor volume was calculated by applying the following equation: tumor volume = (length × width2)/2. The statistical significance of tumor volume was modeled using a mixed effects model, assuming a linear relationship between time and log tumor volume, with individual mice representing random effects. P-values reported represent the significance of the time and treatment interaction term on tumor volume.
In vivo LPS-induced cytokine production model
In vivo experiments of LPS-induced TNF production were conducted at ChemPartners (Shanghai, China). Female Lewis rats (n = 10 per group) were dosed i.p. with ND-2158 (1, 3, and 30 mg/kg) or vehicle (10% hydroxypropyl-β-cyclodextrin in saline). Dexamethasone (Dex) was administered by oral gavage in 0.5% methylcellulose saline. 30 min after dosing, the rats were i.v. administered 0.1 mg/kg LPS in PBS. 1 h later, rats were bled and serum analyzed for TNF production by ELISA (eBioscience).
Collagen-induced arthritis model
Male DBA mice were immunized with type II collagen emulsified in complete Freund’s adjuvant on days 0 and 21, according to the published protocol (Brand et al., 2004). Mice were randomly enrolled with disease onset (n = 8 per group, except for n = 4 naive controls). Disease onset was defined as four digits in one paw showing swelling/redness (as a symptom of arthritis; Brand et al., 2004). Day 1 indicates the first treatment day, and mice were evaluated daily for clinical scores until day 11. ND-2158 or its vehicle (10% hydroxypropyl-β-cyclodextrin in ddH20) was administered i.p. Dex served as a positive control for blocking collagen-induced arthritis. ND-2158 was dosed at 30 mg/kg i.p. twice daily and Dex at 0.1 mg/kg i.p. twice daily. Statistics are as follows: Student’s t test IRAK4 inhibitor versus vehicle: *, P < 0.05 days 2–11; Dex versus vehicle: *, P < 0.05 day 1; **, P < 0.01 day 2; ***, P < 0.001 days 3–5; ****, P < 0.0001 days 6–11. Histological analysis was conducted on harvested tissue sections. Histology scores are as follows: 0 = normal, 1 = minimal, 2 = mild, 3 = moderate, 4 = marked, 5 = severe.
Monosodium urate (MSU) air pouch model for gout
BALB/c mice were distributed randomly into 10 mice per group. ND-2158 or vehicle (10% hydroxypropyl-β-cyclodextrin prepared in ddH2O) was administered i.p. (3, 10, and 30 mg/kg). After the first dose, the mice were anesthetized, and the nape of the neck was shaved to generate an air pouch by injecting 6 ml of air using a 23-gauge needle (Chen et al., 2006). Mice were subsequently dosed once daily with ND-2158 or vehicle for 6 d. 30 min after the final dose, 3 ml MSU suspension was injected into the air pouch. 4 h after the MSU injection, exudate was removed from the pouch, and an aliquot was used for WBC count and differential analysis.
Online supplemental material
Fig. S1 shows gene expression changes in four ABC DLBCL lines induced by a time course of treatment with the IRAK4 inhibitor ND-2158. Table S1, included as a separate Excel file, shows the percent inhibition in kinase activity of IRAK4 inhibitors ND-2110 or ND-2158 tested across 334 kinases. Table S2, included as a separate Excel file, lists gene signature enrichment for genes down-modulated by a time course of ND-2158 treatment in ABC DLBCL.
We acknowledge Leah Frye, Jeremy Greenwood, Matt Wessel, Mee Shelley, Shawn Watts, and Ramy Farid (Schrödinger); Alison Bendele and Phillip Bendele (Bolder BioPath); Haiching Ma (Reaction Biology); Xinyi Huang, Lingling Liu, Lezhen Li, Ronggang Liu, and Congran Feng (Pharmaron); Feilan Wang and Yajun Xu (Chempartners); and Eric Smith (Nimbus Therapeutics) who helped prepare figures.
This research was supported by the Intramural Research Program of the National Institutes of Health (NIH), National Cancer Institute, Center for Cancer Research. We thank George Wright for statistical analysis (Biometric Research Branch, National Cancer Institute/NIH).
P.N. Kelly, Y. Yang, A.L. Shaffer III, L. Rui, and L.M. Staudt declare no competing financial interests. D.L. Romero, D. Chaudhary, W. Miao, W.F. Westlin, and R. Kapeller are employees of Nimbus Therapeutics. S. Robinson is an employee of Schrödinger. The authors declare no additional competing financial interests.
L. Rui’s present address is Dept. of Medicine, University of Wisconsin-Madison, Madison, WI 53792.