Chronic lymphocytic leukemia (CLL) is a malignancy of mature B cells driven by B-cell receptor (BCR) signaling and activated primarily in the lymph node. The Bruton's tyrosine kinase (BTK) inhibitor ibrutinib effectively inhibits BCR-dependent proliferation and survival signals and has emerged as a breakthrough therapy for CLL. However, complete remissions are uncommon and are achieved only after years of continuous therapy. We hypothesized that other signaling pathways that sustain CLL cell survival are only partially inhibited by ibrutinib. In normal B cells, Toll-like receptor (TLR) signaling cooperates with BCR signaling to activate prosurvival NF-κB. Here, we show that an experimentally validated gene signature of TLR activation is overexpressed in lymph node–resident CLL cells compared with cells in the blood. Consistent with TLR activation, we detected phosphorylation of NF-κB, STAT1, and STAT3 in lymph node–resident CLL cells and in cells stimulated with CpG oligonucleotides in vitro. CpG promoted IRAK1 degradation, secretion of IL10, and extended survival of CLL cells in culture. CpG-induced TLR signaling was significantly inhibited by both an IRAK1/4 inhibitor and ibrutinib. Although inhibition of TLR signaling was incomplete with either drug, the combination achieved superior results, including more effective inhibition of TLR-mediated survival signaling. Our data suggest an important role for TLR signaling in CLL pathogenesis and in sustaining the viability of CLL cells during ibrutinib therapy. The combination of ibrutinib with a TLR pathway inhibitor could provide superior antitumor activity and should be investigated in clinical studies.
CLL relies on the concomitant cooperation of B-cell receptor and Toll-like receptor signaling; inhibition of both pathways is superior to inhibition of either pathway alone.
Chronic lymphocytic leukemia (CLL), the most common leukemia in Western countries, is characterized by the accumulation of mature, clonal B lymphocytes in the peripheral blood, bone marrow, and lymphoid organs (1, 2). CLL has a heterogenous clinical course, with some patients never needing any treatment whereas others progress rapidly and often require multiple rounds of treatment. Among the different prognostic markers, immunoglobulin heavy chain variable (IGHV) mutation status is particularly useful in distinguishing indolent and more progressive forms of CLL (3). Patients with IGHV-unmutated CLL progress faster, relapse earlier after chemotherapy, and have inferior overall survival compared with patients with IGHV-mutated CLL (4, 5). Further, in IGHV-unmutated CLL, BCR signaling, a key driver of CLL progression, is more active and tumor proliferation is increased compared with IGHV-mutated CLL (2).
When CLL cells are cultured in vitro, they undergo rapid apoptosis due to a lack of microenvironmental support. However, CLL cells can be rescued from spontaneous apoptosis when cultured in conditions that mimic microenvironmental signaling (6). Microenvironmental signals are initiated by communications with accessory cells including stromal cells, macrophages, and T cells. T cells have been shown to be necessary for the survival and proliferation of CLL cells in a patient-derived xenograft mouse model, supporting an important role for T cells in CLL pathogenesis (7). More recently, macrophages have been shown to support tumor growth in murine models, including the Eμ-TCL1 transgenic mouse model (8–10).
Comparative transcriptome analysis of CLL cells from lymph nodes, blood, and bone marrow revealed activation of BCR and canonical NF-κB signaling in lymphoid tissues of patients with CLL (11). In addition, both transcriptome analysis and in vivo labeling of proliferating cells using deuterium oxide (heavy water) identified the lymph node as the primary site of tumor proliferation (11, 12). Consistent with these data, activation of BCR signaling in vitro with anti-IgM improves CLL cell survival (13). BCR signaling is now established as a key therapeutic target in CLL and other B-cell malignancies (14). The Bruton's tyrosine kinase (BTK) inhibitor ibrutinib (IB) is approved for all lines of therapy in CLL and achieves high rates of durable remissions in the majority of patients. In patients treated with IB, on-target effects include rapid inhibition of BCR and NF-κB signaling in all disease compartments, including lymph nodes (15), inhibition of tumor proliferation (15–17), and increased apoptosis of CLL cells (17, 18). IB induces clinical responses in over 80% of patients across the spectrum of disease, including patients with relapsed or refractory disease and deletion 17p (19–21). However, deep remissions are uncommon and are achieved only after years of continuous therapy (20, 22). Further, more than half of patients with relapsed or refractory disease and high-risk cytogenetics develop resistance to single agent IB within a few years of starting treatment (20, 22, 23).
In addition to CLL, IB has demonstrated efficacy in Waldenstrom macroglobulinemia (WM) and activated B-cell diffuse large B-cell lymphoma (ABC-DLBCL), both of which rely on NF-κB signaling to promote malignant growth and survival. In WM, activating mutations in myeloid differentiation primary response gene (MYD88; L265P) that promote BTK-dependent prosurvival signaling are a hallmark of the disease and are associated with response to IB (24, 25). MYD88 is an adaptor molecule essential for Toll-like receptor (TLR) signal transduction and is also frequently mutated in lymphoma (26). TLRs are a family of immune-receptors that react with pathogen associated molecular patterns such as LPS or unmethylated DNA to activate innate immune responses. TLR signaling has been shown to cooperate with BCR signaling in activating autoreactive B cells (27). Crosstalk between BCR and TLR signaling has also been reported in ABC-DLBCL, where it was associated with clinical response to IB (28). Although MYD88 mutations are uncommon in CLL, our previous study identified gene signatures indicative of active TLR signaling in CLL cells residing in lymphoid tissues (11). Based on these observations, we hypothesized that TLR signaling could cooperate with BCR signaling to enhance CLL proliferation and survival, particularly in microenvironmental niches, and might also sustain CLL cell survival in BTK-inhibited cells during IB therapy.
Materials and Methods
Peripheral blood (PB) and lymph node (LN) samples were obtained from patients with CLL with written informed consent in accordance with the Declaration of Helsinki, applicable federal regulations, and requirements from the local Institutional Review Board. Patient characteristics for samples utilized are described in Supplementary Table S1. Mononuclear cells were isolated through density gradient centrifugation by using lymphocyte separation medium (Ficoll Lymphocyte Separation Media; ICN Biomedicals) and cryopreserved in 90% FBS (Sigma-Aldrich) and 10% dimethyl sulfoxide (DSMO; Sigma). Through previously described IGHV sequencing, leukemic samples were classified as mutated (<98% homology to germline) or unmutated (≥98% homology; ref. 29). Because patients with IGHV-unmutated CLL cells typically have more aggressive disease and have been previously shown to be more activated by TLR ligation than IGHV-mutated CLL cells, we focused our in vitro analysis on IGHV-unmutated CLL (2, 13, 30–33).
Affymetrix array data on 17 patients with paired PB and LN samples were previously published (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE21029; ref. 11). Detected transcripts were annotated for coding genes. Normalized and log2 transformed data were used for downstream data analysis. A two-way ANOVA model was used to account for subject batch effects as previously described (11). Genes with 1.5-fold-change in expression between LN and PB with an FDR less than 20% were considered differentially expressed and selected for further investigation. Overrepresentation of experimentally derived gene sets among differentially expressed genes was estimated using hypergeometric testing. JMP 12 (www.jmp.com) and R software (https://cran.r-project.org) were used in data analysis. Gene sets with an FDR q-value <0.1 were considered overrepresented. Heatmap of differentially expressed upregulated genes in the LN overlapping with the TLR-CpG signature was produced using Gene Cluster 3.0 v.1.52 (http://bonsai.hgc.jp/∼mdehoon/software/cluster/) and Java TreeView 3.0 software (http://bonsai.hgc.jp/∼mdehoon/software/cluster/manual/TreeView.html).
Reagents and cell culture
CLL PB mononuclear cells (PBMC) were cultured in RPMI medium 1640 (Thermo Fisher Scientific) supplemented with 10% FBS (Sigma-Aldrich), and 1% penicillin streptomycin glutamine (Thermo Fisher Scientific). CLL PBMCs were incubated with or without single-strand, phosphorothioate-modified CpG- oligonucleotides (5′-TCGTCGCTGTCTCCG-3′) (CpG; Integrated DNA Technologies) at a dose of 1 μmol/L for between one and 24 hours as indicated. In select experiments, CLL PBMCs were pretreated with or without 1 μmol/L IB (Selleckchem), 10 μmol/L IRAK1/4 Inhibitor (IRAKi; Catalog No. 407601, EMD Millipore) or the combination for 1 hour, followed by incubation with either 1 μmol/L CpG or 20 μg/mL αIgM (Jackson ImmunoResearch) for the indicated length of time. Based on recombinant kinase assays, IRAKi, at doses up to at least 10 μmol/L, preferentially inhibits IRAK1 and IRAK4 (34); with the exception of FMS (CSF1R), which is not expressed on CLL cells (35). In DLBCL cell lines, dependent on TLR signaling, in vitro doses of IRAKi below 10 μmol/L showed minimal effects (26). We performed dose–response experiments of the IRAKi, both by Western blot analysis and by flow cytometry, evaluating changes in degradation of IRAK1, phosphorylation of STAT3, and expression of the NF-κB regulated activation marker CD69. We found that doses lower than 10 μmol/L did not consistently inhibit TLR signaling in CLL cells (Supplementary Fig. S1); thus, we chose to use the IRAKi at 10 μmol/L based on this data and previously published data suggesting that the drug is still selective at this dose (36).
Proximity ligand assay
Activation of TLR signaling was evaluated utilizing a proximity ligand assay as previously described (37). Briefly, CLL LN cells were applied to a ibidi μ-slide with 15 wells (Sigma) and allowed to adhere for 30 minutes. The cells were fixed with 4% PFA for 10 minutes, permeabilized with 90% methanol for 20 minutes, and blocked with Duolink block solution for 1 hour. Primary antibodies were prepared in Duolink antibody dilution buffer and incubated overnight at 4°C (1:200 dilution; Sigma). Cells were counterstained with CD20 488 nm (eBioscience). Primary antibodies included, TLR9 (rabbit), IRAK1 (rabbit), and pIκBα (mouse) from Cell Signaling Technology and MYD88 (rabbit) from Abcam. The cells were incubated with secondary antibodies (Duolink PLUS and MINUS; Sigma) at 37°C for 1 hour, followed by a ligation mixture (Sigma) at 37°C for 30 minutes and an amplification mixture (Sigma) at 37°C for 100 minutes, with washes in between. The wells were mounted with Prolong Gold mounting medium with DAPI and visualized on a Zeiss LSM 880 NL Airyscan microscope (Zeiss). Positive and negative controls for the assay were previously published (34).
Expression of intracellular proteins was evaluated as previously described (38). Briefly, anti-CD19 (APC or PECy5; BD Biosciences) was used to define the CLL population. Cells were then fixed with 4% paraformaldehyde (Electron Microscopy Sciences), permeabilized with either 90% methanol or 80% ethanol (Sigma) at −20°C, and stained with intracellular stains. These included IgG1-isotype control and IRAK1 (PE; Cell Signaling Technologies) and IgG1, IgG2a, and IgG2b-isotype controls and pNF-κB(pP65(S529)), pSTAT3(Y705), pSTAT1(S727), pSTAT3(S727), pBTK(Y223), pPLCγ2(Y759), and pERK(T202/Y204; Alexa Fluor 488; BD Biosciences). In select experiments, anti-CD3 (BD Biosciences) was added as an exclusion control for B-cell selection. Cells were analyzed on a BD FACS Canto II flow cytometer using FlowJo software (Version 10; TreeStar). Mean fluorescence intensity (MFI) shown is isotype subtracted. To allow for analysis of multiple samples in one flow tube, cells were barcoded with pacific blue (Invitrogen), as previously described (39). Cell surface activation markers and cellular viability were evaluated as previously described (15, 18). Briefly, PBMCs were stained with anti-CD19 and CD3 (APC or FITC; BD Biosciences) and either one of anti-IgG1-isotype control, CD40, CD54, CD69, and CD86 (PE; BD Biosciences) or Annexin-V (PE; BD Biosciences) and ViViD Live-Dead stain (Pacific Blue; Invitrogen).
Western blot analysis
Immunoblot experiments were done as previously described (40). Whole cell lysates were made from PBMCs in a RIPA lysis buffer using PhosSTOP inhibitor cocktail and complete Protease Inhibitor Cocktail (Roche Applied Science). In select experiments utilizing LN-resident CLL cells, CD19 purification was done using a CD19+ MACS Selection Kit as described by the manufacture (Miltenyi). Proteins were run on Novex NuPAGE SDS-PAGE (Invitrogen), and subsequently transferred to polyvinyidene fluoride membranes and immunoblotted using the indicated antibodies and horseradish peroxidase-labeled secondary antibodies (GE Healthcare Biosciences). Primary antibodies used include anti-IRAK1, phospho-STAT3(Y705), total STAT3, phospho-IκBα(S32/36), total IκB, and GAPDH (all from Cell Signaling Technologies), and γ-tubulin (Sigma). Images were developed using LAS-4000 imaging system (Fuji Film).
Cell culture supernatant was collected from CLL PBMCs cultured with or without 1 μmol/L CpG after 1, 6, or 24 hours. The percentage of B cells was confirmed to be ≥97% by flow cytometry. Levels of cytokine secretion were measured using human IL10 Quantikine ELISA Kits (R&D Systems) according to the manufacturer's instructions. The reaction was read at an absorbance of 450 nm (Wallac Victor, PerkinElmer).
Statistics for bioinformatic analysis is detailed above. All other statistics were determined using GraphPad PRISM (GraphPad Software Inc.) by Wilcoxon matched-pairs signed rank test after determining normality using JMP software (SAS Institute). P values ≤0.05 were considered statistically significant.
TLR signaling is activated in lymph node–resident CLL cells
We previously reported upregulation of BCR and NF-κB signaling in CLL cells in the LN compared with the PB (11). In addition, a gene set representing components of the TLR signaling pathway was also modestly upregulated in the LN. Here, we used recently described, experimentally validated gene signatures to further probe our dataset, including one specific for TLR signaling in CLL B cells (Supplementary Table S2; refs. 41, 42). Among 511 genes upregulated in the LN compared with the peripheral blood (1.5-fold change; FDR 0.2; Supplementary Table S3), the TLR signature was highly overrepresented (FDR 5.1 × 10−13 by hypergeometric analysis; Fig. 1A; Table 1) and ranked second among the 10 most significantly enriched gene signatures equally with the BCR and IL6 gene signatures (Table 1). Only a gene signature downregulated by HIF1α activation scored even higher, likely reflecting an increase in the active phenotype of CLL cells in the LN as compared with the more quiescent phenotype of circulating cells in the blood. We found an overrepresentation of TLR genes in patients with CLL with both an IGHV-unmutated and IGHV-mutated CLL; however, patients with IGHV-unmutated CLL displayed a higher magnitude of TLR activation (Fig. 1B; P < 0.05). Additional significantly enriched gene signatures represent NF-κB, STAT3, CD40, and IL10 signaling as well as T-cell activation, and cell proliferation (Table 1). Many TLR-regulated genes encode chemokines and cytokines (43). We found a trend for higher expression in the LN compared with the PB, with CCL4, CCL3, CCL5, TNF, and IL10, demonstrating significant increases, whereas IL12A was significantly decreased (Fig. 1C; P < 0.05).
|Gene set name .||No. genes in gene seta .||No. detected genes in gene set (K)b .||No. DE genes in gene set (k)c .||P value .||FDR q value .|
|Gene set name .||No. genes in gene seta .||No. detected genes in gene set (K)b .||No. DE genes in gene set (k)c .||P value .||FDR q value .|
aThe number of genes in the reference publication (Supplementary Table S2) used to identify the gene set.
bThe number of genes from the reference publication identified in this data set.
cThe number of genes from the reference publication that were found to be differentially expressed (DE) between the LN and PB (fold change 1.5 and FDR 20%).
Ten distinct receptors with different ligand specificities initiate TLR signaling through IL1R-associated kinase 1/4 (IRAK1/4) and MYD88. We found that most TLRs were expressed at higher levels in the LN than in the PB, with significant enrichment of TLR 1, 2, 6, and 10 and a trend for higher expression in TLR 8 and 9 (Fig. 1D; P < 0.05).
To further evaluate TLR signaling in CLL cells, we utilized a proximity ligand assay (PLA), which identifies proteins within tens of nanometers of each other, and may be part of signaling complexes (34). A TLR9:pIκBα PLA produced fluorescent puncta in the cytoplasm of CLL cells harvested from the LN, indicating active TLR9 signaling (Fig. 2A). We confirmed this data by also evaluating MYD88:pIκBα and IRKA1:pIκBα PLA, both of which also produced fluorescent puncta in the cytoplasm of CLL cells (Supplementary Fig. S2). Together this suggests a multiprotein complex, indicative of active TLR/NF-κB signaling, is present in primary CLL cells harvested from the lymph node.
Consistent with TLR signaling, we detected significantly higher phosphorylation levels of NF-κB(P65; S529) in the LN compared with patient-matched blood samples using flow cytometry (Fig. 2B; P = 0.002 and Supplementary Fig. S3). Similarly, pSTAT3(Y705) was also significantly increased in LN-resident CLL cells as compared with circulating CLL cells (Fig. 2C; P = 0.004). To further demonstrate the increase in STAT and NF-κB signaling between CLL cells in the LN and PB, we performed Western blot analysis for phosphorylation of STAT3(Y705) and IκBα. Fig. 2D depicts two representative patients demonstrating an increase in phosphorylation compared with total protein. Finally, we evaluated changes in pSTAT1(S727) and pSTAT3(S727). We found that both were significantly increased in LN-resident CLL cells as compared with circulating CLL cells (Fig. 2E and F; P ≤ 0.02). Together, these results indicate increased expression of TLR regulated genes and an increase in signaling downstream of TLR activation in the LN, suggesting that TLR signaling could contribute to proliferation and survival of CLL cells.
CpG activates TLR signaling in CLL cells in vitro
We used unmethylated CpG oligodeoxynucleotide (CpG) to activate TLR signaling in CLL cells cultured in vitro (31). Upon TLR9 ligation, MYD88 is activated, which recruits IRAK4, which in turn phosphorylates IRAK1, leading to its rapid degradation (44). We measured IRAK1 levels in CLL cells by flow cytometry after 5 hours of CpG stimulation. IRAK1 expression was significantly reduced in cells treated with CpG compared with untreated controls, consistent with IRAK1 degradation (Fig. 3A; P < 0.0001). Western blot analysis in three representative patient samples confirmed a decrease in total IRAK1 protein expression after CpG treatment (Fig. 3B). Further downstream, we found a significant increase in pNF-κB(P65; S529) in CLL cells treated with CpG compared with untreated controls (Fig. 3C; P < 0.0001). In response to CpG treatment secretion of IL10 was significantly increased within 5 to 6 hours, followed by further increases after longer treatment duration (Fig. 3D; P = 0.02). IL10 secretion has been reported to lead to autocrine stimulation of IL10 receptors expressed on CLL cells (45). Consistent with consequent activation of JAK/STAT signaling, pSTAT3(Y705) was significantly increased in CpG stimulated cells compared with untreated controls (Fig. 3E, P < 0.0001). Congruent with this, patients with early increases in IL10 tended to display more phosphorylation of STAT3 (Supplementary Fig. S4). In a subset of patients, we corroborated our flow cytometry data using immunoblot analysis for pSTAT3(Y705), which demonstrated a similar range of phosphorylation levels between the two approaches (Supplementary Fig. S5). In addition, we investigated the link between IRAK1 and pSTAT3(Y705) through immunoblot analysis. In most patients, the CpG dependent decrease in IRAK1 was accompanied with an increase in pSTAT3(Y705) with patients demonstrating a greater loss in IRAK1 expression showing a larger increase in STAT3(Y705) phosphorylation as shown in Supplementary Figs. S5B and S6A. We further demonstrated a connection between phosphorylation of IκBα and both IRAK1 and phosphorylation of STAT3 (Supplementary Fig. S6B and S6C). Concurrently, both pSTAT1(S727) and pSTAT3(S727) also significantly increased in CpG treated samples compared with untreated controls (Fig. 3F and G; P < 0.0001). Altogether, these results suggest that we can recapitulate the TLR activation observed in the LN microenvironment in vivo.
We next assessed expression of activation markers: CD40, CD54, CD69, and CD86 by flow cytometry on CLL cells after 24 hours of CpG stimulation. We found a significant increase in the expression of all four activation markers after CpG stimulation as compared with untreated controls (Supplementary Fig. S7A; P < 0.0001). Furthermore, CpG stimulation supported IGHV-unmutated CLL cell survival in vitro, resulting in significantly higher CLL viability in CpG treated cells compared with untreated controls (Supplementary Fig. S7B; P < 0.0001). Overall, CpG-induced TLR signaling activates CLL cells in vitro and promotes cell survival.
IB inhibits TLR signaling in CLL cells
BCR and TLR signaling have been shown to cooperate in activating autoreactive B cells (27). Furthermore, somatic mutations in TLR and BCR signaling pathways have been implicated as driver mutations in WM and ABC-DLBCL, respectively. We therefore sought to investigate the activity of TLR and BCR signaling inhibitors in CLL cells using IB and a preclinical IRAK1/4 inhibitor (IRAKi). We pretreated CLL PBMCs with the respective inhibitor or vehicle control for 1 hour before adding CpG for an additional 5 hours. As a readout of TLR signaling we measured IRAK1 levels by flow cytometry. IB clearly inhibited TLR signaling as evidenced by a significant reduction in CpG-induced IRAK1 degradation (Fig. 4A; P = 0.005). As expected, the IRAKi also inhibited CpG induced IRAK1 degradation (Fig. 4A; P = 0.0001). In comparison, IRAKi had a stronger effect on IRAK1 stabilization than IB (P = 0.002). In two representative patients, we used Western blot analysis to confirm IRAK1 degradation in response to CpG, and stabilization by the respective inhibitor (Fig. 4B).
Next, we evaluated the effect of the inhibitors on downstream signaling pathways activated by TLR engagement, including NF-κB and JAK/STAT. Both IB and the IRAKi significantly reduced the effects of CpG stimulation on NF-κB signaling (Fig. 4C; P ≤ 0.006). Similarly, we found that both IB and the IRAKi significantly inhibited JAK/STAT signaling as measured by expression of pSTAT3(Y705) (Fig. 4D; P ≤ 0.0002). In summary, IB inhibited upstream and downstream effects of TLR activation.
We observed a significant reduction in CpG-induced upregulation of CD54 and CD69 with both IB and the IRAKi when compared with CpG stimulated cells (Supplementary Fig. S8A; P ≤ 0.0004). Similarly, both IB and the IRAKi significantly reduced the CpG-induced prosurvival effect, resulting in lower cell viability (Supplementary Fig. S8B; P ≤ 0.005). Together, these data indicate that IB can at least partially inhibit upstream and downstream effects of TLR activation, resulting in decreased cell activation and survival.
IB inhibits both BCR and TLR signaling in CLL
To assess activity of TLR signaling, we quantified CpG-induced pSTAT3(Y705), pSTAT3(S727), and pSTAT1(S727) by flow cytometry. Both IB and the IRAKi significantly inhibited CpG-induced TLR signaling (Fig. 5A; P ≤ 0.002). Only pSTAT3(Y705), which was more potently inhibited by IB, demonstrated a significant difference in reduction of signaling between IB and the IRAKi (Fig. 5A). IB reduced pSTAT3(Y705) phosphorylation levels by a median of 18% (IQR: 17%–34%) whereas the IRAKi produced similar results, reducing phosphorylation levels by a median of 14% (IQR: 10%–20%; Fig. 5A, left; P = 0.03). Similarly, IB reduced pSTAT1(S727) and pSTAT3(S727) phosphorylation levels by a median of 15% (IQR: 10%–20%) and 12% (IQR: 6%–22%), respectively; whereas the IRAKi reduced phosphorylation levels by a median of 18% (IQR: 15%–23%) and 9% (IQR: 5%–20%), respectively (Fig. 5A, middle and right; P = 0.2 and 0.4, respectively).
Given evidence for cross-talk between TLR and BCR pathways, we next sought to compare the effects of IB and the IRAKi on BCR signaling (27, 28). As expected, stimulating CLL cells with αIgM induced phosphorylation of BTK, PLCγ2, and ERK (Supplementary Fig. S9A–S9C; P = 0.001). IB effectively inhibited IgM-induced BCR signaling evidenced by significant reductions in pBTK [by 26% (IQR: 19%–30%)], pPLCγ2 [by 29% (IQR: 21%–35%)], and pERK [by 29% (IQR: 16%–44%); Fig. 5B; P = 0.001). In contrast, the IRAKi achieved only modest reductions (by <10%) in pPLCγ2 and pERK (P = 0.01 compared with no inhibitor treatment) and borderline reductions in pBTK (Fig. 5B; P = 0.08). Finally, in a subset of patients with matched PB and LN samples we evaluated the effect of IB and the IRAKi on our panel of TLR and BCR markers (Supplementary Fig. S10A and S10B). We found a significant increase in phosphorylation levels of all markers between the cells harvested from the PB and LN after 30 minutes in culture, suggesting that the TLR pathway remained activated in the LN-resident cells. IB significantly reduced the phosphorylation levels of signaling markers in both TLR and BCR pathways (Supplementary Fig. S10A and S10B; P < 0.05). As observed with in vitro CpG stimulation, the IRAKi predominately reduced TLR signaling markers (Supplementary Fig. S10A and S10B). These findings indicate that although there is definitive crosstalk between these two pathways in CLL, inhibiting BCR signaling with IB has a stronger effect on the TLR pathway than the reverse with the IRAKi.
Dual targeting of TLR signaling with IB and IRAKi is superior to single agents
Although IB and IRAKi both reduced TLR signaling, complete inhibition was not achieved. We therefore investigated the effect of the combination. Concurrent treatment with IB and the IRAKi was significantly more potent than either drug alone, effectively inhibiting both upstream and downstream effects of TLR activation (Fig. 6). While CpG-induced IRAK1 degradation was inhibited by both IB and IRAK1, the combination was significantly more effective than either single agent (Fig. 6A; P ≤ 0.04). Similarly, CpG-induced phosphorylation of STAT3(Y705) and STAT1(S727) was inhibited by both single-agents, and the combination was again significantly more potent (Fig. 6B and C; P ≤ 0.006). In fact, in combination treated cells, STAT1 and STAT3 phosphorylation was reduced to levels not significantly different from unstimulated cells (P = 0.4). Consistent with the observed effects on TLR signaling, the combination of IB and IRAKi significantly reduced the viability of CpG-stimulated CLL cells compared with either single-agent and unstimulated controls (Fig. 6D; P = 0.01).
Most patients with CLL treated with IB achieve remissions; however, residual disease remains detectable for years (20, 22). The persistence of CLL cells despite virtually complete inhibition of BTK and tumor proliferation suggest that BTK-independent pathways can maintain tumor viability (15, 46). Here we focused on the role of TLR signaling in CLL pathogenesis and specifically, whether it can be inhibited by IB. A role of TLR signaling in the pathogenesis of lymphoma, WM, and CLL is supported by the presence of gain-of-function mutations in MYD88, an adaptor molecule essential for TLR signal transduction (26, 47, 48). Further, CpG motifs present in unmethylated DNA of bacteria or released from apoptotic cells activate TLR9, one of the TLRs universally expressed on CLL cells as well as on normal memory B cells (49, 50). Thus, TLRs are not only potent activators of innate immune responses but also contribute to activation of autoreactive B cells, including CLL cells. Crosstalk between BCR and TLR pathways has been implicated in aggressive lymphoma and in the activation of normal B cells, as well as CLL cells (27, 51). Taking advantage of experimentally validated pathway-specific gene signatures (41, 42), we provide evidence for active TLR signaling in CLL cells in the lymph node. Notably, the TLR gene signature is part of an overall activation signature expressed in lymph node resident CLL cells that is also highly enriched in genes regulated by BCR, NF-ĸB, and JAK/STAT pathways, the latter indicative of cytokine signaling. CpG-dependent activation of TLR signaling in vitro recapitulated activation of NF-ĸB and upregulation of IL10 with consequent activation STAT1 and STAT3 and improved CLL cell survival. Thus, TLR and concurrent cytokine signaling are activated within the tissue microenvironment and, in agreement with a prior in vitro study, may cooperate to improve CLL cell survival (30).
In CLL, IB has been shown to induce direct tumor cell death at a rate of less than 2% per day in vivo, suggesting a slow attrition of apoptotic cells when BTK signaling alone is inhibited (17, 18). We sought to test whether the combination of IB with an IRAKi could potentiate the apoptotic effects of BTK inhibition alone. We found that treatment with either IB or an IRAK1/4 inhibitor decreased CpG-induced TLR signaling and JAK/STAT signaling, cell activation, and survival. However, simultaneous treatment with both IB and the IRAKi resulted in a greater inhibition of TLR signaling and cell survival than with either single agent. In contrast to IB, the IRAKi had virtually no effects on IgM-induced BCR signaling. In summary, although IB effectively inhibited BCR signaling it could only partially inhibit TLR signaling, and combination treatment with the IRAKi was required to achieve maximal apoptosis of cells concurrently activated through BCR and TLR engagement.
IGHV-unmutated CLL cells are more responsive to BCR activation than those with IGHV-mutated CLL. This is mirrored by TLR activation both in vivo and in vitro. In vivo, we observed a higher TLR signature score in IGHV-unmutated patients compared with IGHV-mutated patients in the lymph node compared with the peripheral blood. Similarly, in vitro the two subtypes demonstrate differential responses to CpG stimulation. Thus, for our in vitro assays we focused on IGHV-unmutated CLL, the more aggressive CLL subtype that has consistently been found to derive activation, survival, and proliferation signals from TLR activation. In contrast, IGHV-mutated CLL cells are more likely to enter an apoptotic state when stimulated with CpG in vitro, in the absence of concurrent BCR activation or coculture with cytokines (30, 32). These in vitro observations raise the possibility that such costimulatory signals may sustain survival of IGHV-mutated CLL cells in vivo. In support, although there is a difference in the magnitude of TLR activation, both IGHV-unmutated and IGHV-mutated patients display an overrepresentation of TLR signature genes in the lymph node.
The finding of TLR activation in CLL cells from the lymph node corroborates the idea that CLL cells in the microenvironment behave differently than CLL cells in circulation. Our group has previously demonstrated that CLL cells in the LN display increased BCR and NF-κB signaling, suggesting CLL cells are more antigen exposed in the microenvironment (11). Similarly, our finding of activation TLR signaling, in specific TLR9 activation, suggests that CLL cells in the LN may be exposed to increased levels of microbial DNA (known to contain CpG motifs). Understanding the mechanisms of activation of CLL cells in the lymph node microenvironment may allow for the development of novel therapeutics and more rational combination therapy.
Our observation that IB could only partially inhibit TLR signaling but was not sufficient to completely abrogate TLR signaling or prosurvival effects is consistent with the clinical experience with IB in WM (24). Notably in WM, the decrease in IgM paraprotein is faster and more pronounced than the eradication of tumor in the bone marrow. Also, similar to CLL, although sustained on continuous therapy, most WM responses are partial with lingering tumor cells for the duration of treatment (24). Overall, the persistence of residual disease in WM patients with activated TLR signaling on IB is further indication that BTK inhibition alone is insufficient to completely block prosurvival signals.
In conclusion, our data warrants further investigation of a combination of BTK and TLR inhibitors in CLL. We demonstrate that although IB can act as partial TLR inhibitor, combining both IB and a direct IRAK1/4 inhibitor achieves a deeper inhibition of TLR signaling, increasing the potential for faster and/or deeper responses. Although we cannot formally exclude that off-target effects of the IRAKi used at 10 μmol/L could contribute to the observed response, the selectivity of the IRAKi is comparable to other kinase inhibitors (34, 38, 52, 53). Furthermore, CSFR1, one of the few kinases inhibited by IRAKi at concentrations below 10 μmol/L is not expressed in CLL cells (35). Nevertheless, the combination of BTK and TLR inhibitors should be further assessed using clinical grade TLR antagonists. In addition to efforts directed at IRAK4 inhibitors, several TLR antagonists with alternative mechanism of action have been developed, with some of these currently being investigated in clinical studies in autoimmune diseases and B-cell malignancies (54, 55). The mechanism by which IB inhibits TLR signaling in CLL was outside the scope of this work; however, data in DLBCL suggest a substantial cross-talk between BCR and TLR signaling, with the formation of a My-T-BCR (MYD88, TLR9, and BCR) supercomplex on endolysomes that drive prosurvival NF-κB, predicting IB-responsiveness (37). Future studies should further evaluate this in CLL.
Disclosure of Potential Conflicts of Interest
A. Wiestner reports receiving a commercial research grant from Pharmacyclics, Acerta Pharma, Merck, and is an consultant/advisory board member at Pharmacyclics. No potential conflicts of interest were disclosed by the other authors.
Conception and design: E.L. Dadashian, A.L. Shaffer III, A. Wiestner, S.E.M. Herman
Development of methodology: E.L. Dadashian, A.L. Shaffer III, S.E.M. Herman
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): E.L. Dadashian, E.M. McAuley, A.L. Shaffer III, J.R. Iyer, M.J. Kruhlak, S.E.M. Herman
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): E.L. Dadashian, E.M. McAuley, D. Liu, A.L. Shaffer III, M.J. Kruhlak, A. Wiestner, S.E.M. Herman
Writing, review, and/or revision of the manuscript: E.L. Dadashian, E.M. McAuley, J.R. Iyer, L.M. Staudt, A. Wiestner, S.E.M. Herman
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A.L. Shaffer III, R.M. Young
Study supervision: A. Wiestner, S.E.M. Herman
We thank our patients for participating and donating samples to make this research possible. This research was supported by the Division of Intramural Research of the National Heart Lung and Blood Institute, NIH (Grant No. ZIA HL002346-13; primary recipient, A. Wiestner).
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.