TLR9 expression in chronic lymphocytic leukemia identifies a promigratory subpopulation and novel therapeutic target

Chronic lymphocytic leukemia (CLL) is a common B-cell malignancy with a remarkably diverse clinical course. It is characterized by the accumulation of mature B-lymphocytes in the peripheral blood (PB), bone marrow, and secondary lymphoid organs such as the lymph nodes (LNs). Until recently, therapeutic options for patients with CLL were limited and poorly tolerated, but the introduction of Bruton’s tyrosine kinase (BTK) inhibitors¹  and drugs targeting phosphoinositide kinases (PI3Ks)²  and Bcl-2³  have revolutionized treatment. However, these agents are not curative, and some patients have short-term responses, develop resistance, or have to stop treatment because of side effects.⁴ 

CLL has been described as a 2-compartment disease because LN and PB CLL cells express distinct phenotypic and transcriptional features.^5,6  PB CLL cells are largely quiescent,⁷  whereas in the LN microenvironment, disease-promoting interactions occur.^5,6  It seems that the pathology of CLL is influenced by the ability of tumor cells to traffic between these 2 compartments,⁸  and recent LN emigrants can be identified by a CXCR4^dimCD5^bright phenotype.⁹  An important goal of CLL therapy is to eradicate tissue-resident CLL cells; “trapping” them in the PB deprives them of the proliferation and survival signals afforded by protective microenvironmental niches.¹⁰  Indeed, BTK and PI3Kδ inhibitors cause tumor cell tissue redistribution,^11-13  but heterogeneity in responses and lack of curative effect indicate that inhibition of these individual kinases alone is not sufficient to eliminate disease. The identification of novel targets and therapeutic combinations is therefore crucial for future cures; one such potential target is Toll-like receptor 9 (TLR9).

The TLR family of pattern recognition receptors is an integral component of the innate immune system and important mediators of the adaptive immune response. After B-cell receptor (BCR) activation and costimulatory T-cell interactions, TLR signaling has been coined the third essential B-cell activation signal following recognition of unmethylated cytosine guanine dinucleotide (CpG)-DNA.^14,15  Each TLR exhibits unique ligand specificity, but most signal via the myeloid differentiation primary response 88 (MyD88) adaptor protein and interleukin-1 receptor kinase family (IRAK1/2/4). This culminates in the activation of the transcription factors NF-κB¹⁶  and signal transducer and activator of transcription 3 (STAT3),¹⁷  which in turn induce a pro-inflammatory immune response and cellular expansion.

Activating mutations in MYD88 occur in ∼3% of patients with CLL,¹⁸  and TLR signaling has been associated with CLL cell proliferation, survival, immunogenicity, and resistance to chemotherapy.^7,19-23  Murine knockout experiments of the TLR decoy and inhibitory receptor (TIR8) showed an increased rate of CLL cell accumulation within the secondary lymphoid tissues,²⁴  and targeting IRAK4, a downstream target of MyD88, delayed CLL tumor development.²⁵  In contrast, Wang et al²⁶  showed that inhibition of TLR9 had a growth-promoting effect on splenic B cells in vivo on an MyD88 mutant background. These different findings support the rationale for further investigation of TLR9 signaling in CLL.

TLR9 is known to play a role in tumor cell trafficking, as evidenced by its association with metastasis in solid tumors.²⁷  Furthermore, CLL cells exhibit higher levels of endosomal TLR9 (eTLR9) compared with healthy B cells,²⁸  and several studies have shown that CpG-DNA can activate primary CLL cells.^7,21-23  It is noteworthy that the phenotype of TLR9-stimulated healthy B cells is similar to that of CLL cells²⁹ ; increased expression of CD5, CD23, CD25, and the T-cell–associated ZAP70 kinase, which is characteristically expressed in poor prognosis disease. Wagner et al³⁰  suggested that, in addition to BCR-independent effects, TLR9 signaling integrates into the adaptive immune response in ZAP70-positive CLL. All of these findings implicate TLR9 signaling in the pathology of CLL.

The aims of this study were threefold. First was to investigate the role of TLR9 stimulation in the activation and migration of CLL cells and whether patient plasma contains levels of unmethylated cell-free DNA (cfDNA), which could trigger this. Second, as aberrant surface expression of TLR9 has been shown in some poor prognosis solid tumors,²⁵  we investigated the expression and function of surface TLR9 (sTLR9) in CLL. Finally, we explored whether dual targeting of BTK and TLR9 has the potential to eliminate CLL cell trafficking and induce more sustained clinical responses.

More detailed methods are provided in the supplemental Materials and methods (available on the Blood Web site).

PB samples were obtained from CLL patients with informed consent in accordance with the Declaration of Helsinki. Matched LN fine needle aspirate sampling was undertaken simultaneously on patients with a palpable lymphadenopathy as previously described.⁶ 

Plasma cfDNA quantification was performed by using real-time polymerase chain reaction to amplify the CCR5 gene and determining the cycle threshold (C_t) value. An assay to detect unmethylated mitochondrial DNA (mtDNA) was designed, based on digesting DNA with the FspEI restriction endonuclease.

CLL PB mononuclear cells (PBMCs) were seeded at 3 × 10⁵ cells/150 μL of complete media (RPMI, 10% fetal calf serum, penicillin/streptomycin, l-glutamine [Sigma-Aldrich], and 5 μg/mL interleukin-4 [RayBiotech]). Cells were cultured ±1 μM ODN2006 (TLR9 agonist; InvivoGen) or 20 μL autologous plasma in duplicate and incubated for 24 hours (or 4 hours for phospho-STAT3/5 [p-STAT3/5] and phospho-p65 [p-p65] NF-κB) at 37°C/5% carbon dioxide. For the TLR9 blocking experiments, CLL cells were preincubated for 30 minutes with the TLR9 antagonist (5 μM ODN INH-18 [InvivoGen] or anti-TLR9 [eB72-1665; eBioscience]) at 5 μM/10⁶ PBMCs. For synergy, 2 μM ODN INH-18 and 1 μM ibrutinib (Selleck Chemicals) were used. Cells were harvested for migration or phenotyping assays as described in the following 2 sections.

Cells were labeled as described in the supplemental Materials and methods using antibody panels detailed in supplemental Table 1.

Transwell migration assays were performed by using 5-μm pore polycarbonate transwell inserts in 24-well plates. A total of 600 μL complete media + 100 ng/mL CXCL12 (BioLegend) were added to the basolateral chambers, and PBMCs were then transferred into the apical chambers and incubated for 4 hours at 37°C/5% carbon dioxide. Migrated and nonmigrated CLL cells were identified by CD19⁺CD5⁺ labeling and counted by using a BD Accuri flow cytometer.

A hollow fiber bioreactor system (FiberCell Systems, Inc) was previously adapted by our group to generate an in vitro model of circulating CLL.^6,8  PB-CLL cells were introduced into the model through the access ports in the circulating compartment and were allowed to circulate for 48 hours before samples were removed from port D (circulating) and port C (migrated). CLL cells were immunophenotyped as described in the "Surface and intracellular immunophenotyping" section.

PBMCs from 5 patients with CLL were antibody labeled (supplemental Table 1). Viable, single CLL cells were identified by CD19⁺CD5⁺ labeling, and both sTLR9^hi and sTLR9^lo populations were sorted by using a BD FACSMelody. RNA was extracted by using the RNeasy Micro Kit (Qiagen) as per the manufacturer’s instructions and immediately frozen at −80°C. mRNA TruSeq library generation and sequencing were performed by Qiagen.

For quantitative polymerase chain reaction, the TLR9 (Hs00370913_s1) and β-actin (Hs99999903_m1) TaqMan (Applied Biosystems) gene expression assays were used.

CLL cells from 7 different patients were xenotransplanted into NOD/Shi-scid/IL-2Rγ^null mice as previously described.^31-33  Details are given in the supplemental Materials and methods.

Statistical analyses were performed by using GraphPad Prism 7.0 (GraphPad Software). Data were assessed for normality, and the appropriate descriptive statistics and tests were used. Unless otherwise stated, results are presented as mean ± standard deviation, paired Student t test, or Wilcoxon matched-pairs signed-rank, and Pearson’s or Spearman correlation coefficient depending on whether the data were Gaussian.

We initially investigated whether plasma from patients with CLL contains cfDNA that could potentially trigger TLR9 signaling and, if so, whether these levels correlated with established prognostic markers. To test this theory, we measured the cfDNA levels in plasma from 37 patients for whom prognostic data were available. cfDNA levels varied between patient samples but showed a positive correlation with CD38 expression (r = 0.67; P < .001) (Figure 1A). In addition, patients with a lymphocyte doubling time (LDT) <12 months had a mean 8-fold higher cfDNA than those with longer LDT (P = .030) (Figure 1B), and those with β₂-microglobulin (B2M) >3.5 mg/L had a mean 3.5-fold higher cfDNA than those with B2M <3.5 mg/L (P = .002) (supplemental Figure 2A). There was a trend for raised cfDNA in immunoglobulin heavy chain variable region (IGHV) unmutated cases (poor prognosis) compared with mutated cases (P = .080) (supplemental Figure 2B) but no difference between ZAP70-positive and ZAP70-negative cases (supplemental Figure 2C). Notably, it was dynamic prognostic markers such as CD38³⁴  and LDT that exhibited the strongest relationship with cfDNA levels. It is therefore possible that high leukemic burden generates raised cfDNA, which in turn induces CLL proliferation via an autostimulation loop. In support of this, we found a positive correlation between cfDNA and white blood cell count (r = 0.71; P < .0001) (Figure 1C) and a strong association between raised cfDNA and a shorter time to first treatment (hazard ratio, 4.0; P = .003) (Figure 1D).

Because only unmethylated DNA, present in eukaryotic cell mitochondria (mtDNA), can trigger TLR9 signaling, we quantified this in the plasma of patients with CLL and age-matched healthy control subjects. Total mtDNA was 12.9-fold higher in plasma from patients with CLL (P = .001) (supplemental Figure 2D), and unmethylated mtDNA was 28.1-fold higher (P < .0001) (Figure 1E). Even when normalized to total mtDNA, the unmethylated mtDNA was 2.1-fold greater in CLL plasma (P < .001) (Figure 1F). It is well established that CLL cells proliferate when stimulated through their TLRs, but this is the first time, to the best of our knowledge, that a link between clinical CLL proliferation and the TLR9 trigger, unmethylated DNA, has been reported. We next investigated whether TLR signaling could induce a more activated and migratory phenotype in CLL cells.

We initially stimulated CLL cells using a TLR9-specific oligonucleotide, B-CpG ODN2006. After a 24-hour stimulation period, primary CLL cells were assessed for markers associated with poor prognosis and cell adhesion/migration. All 15 CLL samples exhibited a marked increase in the expression of CD38 (mean, 2.7-fold; P = .020), CD49d (mean, 1.9-fold; P = .05), and CD69 (mean, 2.1-fold; P = .005) (Figure 2A) after ODN2006 stimulation. Because phosphorylation of p65 NF-κB and tyrosine 705 on STAT3 are associated with migration and poor prognosis in lymphoma³⁵  and CLL,^36,37  and STAT5 mutations are associated with aggressive lymphoma,³⁸  we investigated phosphorylation of these transcription factors after TLR9 stimulation. Both p-STAT3 and p-p65 NF-κB were upregulated in all 24 samples studied (mean, 2.1- and 1.7-fold, respectively) (Figure 2B). In contrast, p-STAT5 exhibited a mean-fold increase of just 1.1, with 7 of 24 samples showing no change or reduced phosphorylation.

Because patients with CLL have elevated levels of the physiological TLR9 ligand unmethylated DNA (Figure 1), we next stimulated CLL cells with autologous plasma, with or without the TLR9 antagonist ODN INH-18. CLL cells were split into 4 fractions: untreated, ODN2006 activated, plasma activated, and plasma activated after preincubation with ODN INH-18. Autologous plasma increased both p-STAT3 and p-p65 NF-κB in all cases (1.5-fold [P < .001] and 1.4-fold [P < .01], respectively), which was abrogated in the presence of the TLR9 inhibitor (P < .0001 and P < .01) (Figure 2C). Again, there was no increase in p-STAT5. These data show that CLL cfDNA can induce the activation of both STAT3 and p65 NF-κB via a TLR9-specific autostimulation loop.

We and others previously reported a link between CLL cell migration and expression of CD38, CD49d, and STAT3.^8,36,39  We therefore investigated whether TLR9 stimulation increased migration. Up to 2-fold greater migration was seen in TLR9-stimulated CLL cells compared with unstimulated controls (P < .001) (Figure 2D). We have previously shown that distinct CLL subpopulations have a greater propensity to migrate,⁶  and thus we investigated whether there was a link between levels of TLR9 and migration.

Basal levels of endosomal TLR9 (eTLR) were quantified in permeabilized unstimulated CLL cells and correlated to the expression of known CLL adhesion/migration markers; mean fluorescence intensity (MFI) values of TLR9, CD38, and CD49d were obtained by gating the viable singlet CD19⁺CD5⁺ CLL cells of 27 patients (gating strategy shown in supplemental Figure 3A). We found no significant correlation between the expression of eTLR9 and CD49d and only a weak correlation with CD38 (supplemental Figure 3B). However, this analysis revealed that within each patient sample, CLL cells with the highest CD38 and CD49d expression had a much greater proportion of TLR9^bright cells (Figure 3A). We therefore gated on the CD19⁺CD5⁺CD38⁺ or CD19⁺CD5⁺CD49d⁺ CLL cells and found striking correlations between the MFI of CD38 (r = 0.93; P < .001) and CD49d (r = 0.73; P = .011) with eTLR9 (Figure 3B). Furthermore, we showed that constitutive expression of p-STAT3 (Tyr 705) and p-p65 NF-κB was elevated in CD38^hiCD49d^hiTLR9^bright CLL cells (Figure 3C). In addition, there was a correlation between CLL cell migration and basal levels of eTLR9 (r = 0.72; P < .001) (Figure 3D), suggesting that, along with the BCR, TLR9 signaling plays an important role in modulating CLL cell migration. Although synergy between the 2 signaling pathways has long been established in normal B cells,^40,41  and suggested in CLL,³⁰  this is the first time that a link between TLR9 levels and CLL migration has been made.

Sustained surface expression of TLR9 (sTLR9) has been previously reported and was linked to tumorigenesis and cancer progression.^25,42  Because we previously identified that many patients with CLL include a small subpopulation with a more migratory phenotype (CD38^hi and CD49d^hi), we gated on these cells and quantified sTLR9. sTLR9 was detected on this subpopulation in all 29 CLL samples tested and, as with eTLR9, there was a strong correlation with CD38 (r = 0.77; P = .001) and CD49d (r = 0.91; P < .001) (Figure 4A-B). For confirmation, we performed intra-patient analysis to compare CD38 and CD49d expression in the gated sTLR9^hi and sTLR9^lo subpopulations. sTLR9^hi CLL cells had 2.5-fold and 3.6-fold higher expression of CD38 and CD49d, respectively, compared with their matched sTLR9^lo cells (P < .001) (Figure 4C). In addition, sTLR9^hi cells exhibited a CXCR4^dimCD5^bright phenotype, which is a phenotype associated with LN trafficking⁹  (supplemental Figure 4).

To establish whether sTLR9 is functional, we split CLL patient samples into 2 fractions; one was untreated, and one was preincubated with an anti-TLR9 antibody to block sTLR9. Both fractions were then stimulated with ODN2006, enabling eTLR9 to be activated as normal but not sTLR9 in the blocked fraction. As expected, both fractions showed evidence of activation, including increased CD49d, CD38, and CD69. However, in the blocked fraction, the expression of these antigens was lower: 63%, 77%, and 67% of the unblocked fraction, respectively (supplemental Figure 5A). Migration assays comparing ODN2006-stimulated CLL cells, with or without sTLR9, or total TLR9 (tTLR9) blockade, showed that blocking sTLR9 resulted in a 14% reduction in migration (P < .01). However, these CLL cells exhibited significantly more migration than unstimulated CLL cells, indicating that sTLR9 blockade alone cannot prevent TLR9 signaling (Figure 4D). In contrast, tTLR9 blockade caused almost total abrogation of ODN2006-induced migration (P < .01) back to that of the unstimulated fraction.

We proceeded to show that sTLR blockade reduced p-p65 NF-κB by 22% and p-STAT3 by 18%, but tTLR9 blockade reduced both to levels below that of the unstimulated cells (supplemental Figure 5B). Importantly, these findings were mirrored using autologous plasma stimulation; tTLR9 and sTLR9 blockade reduced p-p65 NF-κB in all samples but tTLR9 blockade to the resting levels or below in 7 of 10 and sTLR9 blockade in none of the 10. Similarly, p-STAT3 was reduced by both and to the resting level or below in 8 of 12 samples by tTLR9 blockade but in only 1 of 12 after sTLR9 blockade (Figure 4E). Finally, a striking correlation was noted between levels of eTLR9 and sTLR9, implicating sTLR9 as an extracellular marker of extensive eTLR9 signaling (r = 0.9; P < .0001) (Figure 4F).

We next investigated whether sTLR9^hi and sTLR9^lo CLL cells had different transcriptional signatures. We sorted paired fractions from 5 different patients and performed RNA-sequencing. After false discovery rate correction, 653 differentially expressed genes retained significance (q < 0.05): https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE147561. Gene ontology and gene set enrichment analysis, using WebGestalt (WEB-based GEne SeT AnaLysis Toolkit),⁴³  confirmed that sTLR9^hi CLL cells showed increased expression of genes associated with raised metabolic activity (glycolysis), lymphocyte activation, adhesion/migration (Rho GTPase cytoskeleton regulation and IGTB1; part of the CD49d complex), and inflammation mediated by chemokines and cytokines (Figure 5). Most importantly, in the sTLR9^hi fraction, components of the TLR signaling pathway were highly overrepresented, with a normalization enrichment score of 2.08; the overexpressed gene list included the obligate TLR signaling molecule IRAK4.⁴⁴  Furthermore, there was a striking overrepresentation of NF-κB (eg, NFKBIA, NFKBIE, RelA, NFKB2, IKBKB, TRAF6) and STAT3-regulated genes (JAK2 and JUNB). Finally, as shown by Wagner et al,³⁰  there was a link between TLR9 and BCR signaling as the sTLR9^hi population overexpressed the BCR signaling gene PLCγ2.

The RNA-sequencing data suggest that sTLR9^hi expression represents CLL cells undergoing enhanced TLR signaling with a pro-inflammatory and promigratory transcriptional signature. We therefore investigated whether migrated CLL cells showed enrichment for a sTLR9^hi phenotype. CLL cells from 10 patients were put through our circulation model for 48 hours,⁸  and the sTLR9 expression was compared between those that migrated vs those that remained circulating. sTLR9 was higher in the migrated cells from 9 of 10 patients tested (P = .004) (Figure 6A; supplemental Table 2). Importantly, we were also able to show that LN-derived CLL cells were enriched for the sTLR9^hi phenotype in matched fine needle aspirates and PB from CLL patients with palpable LNs. CLL cells from all 7 patients tested had higher expression of sTLR9 in the LNs compared with their PB (P = .016) (Figure 6B-C).

To investigate the role of TLR9 in CLL cells in vivo, we established CLL murine xenografts TLR9^hi and TLR9^lo CLL cells into NOD/Shi-scid/IL-2Rγ^null mice. Of the 7 different patient samples transplanted, 4 were TLR9^hi and 3 were TLR9^lo. CLL cells from the TLR9^hi samples exhibited considerably higher levels of splenic engraftment (mean, 10.2 ± 11.5%) and more rapid disease progression compared with the TLR9^lo samples, which all exhibited engraftment levels <1% (mean, 0.3 ± 0.3%; P = .018) (Figure 6D; supplemental Figure 6). Furthermore, a positive correlation (r² = 0.23; P = .01) was seen between TLR9 expression and the extent of splenic engraftment (supplemental Figure 7). These data support the notion that TLR9 contributes to CLL migration and engraftment in vivo.

The data presented above provide compelling evidence to support inhibition of TLR9 as a potential therapeutic strategy to reduce the activation of STAT3 and NF-κB and prevent CLL cell migration. Although we have shown that sTLR9 is functional in CLL, its blockade is not sufficient to prevent TLR9 signaling. We therefore investigated whether inhibition of tTLR9 synergizes with ibrutinib to reduce migration. To test this theory, an experimentally determined fixed molar ratio of 2:1 (ODN INH-18:ibrutinib) was used. In this model, a clinically relevant dose of 1 μM of ibrutinib⁴⁵  was found to reduce migration of activated CLL cells in 5 of 6 patients when used as a single agent; the addition of ODN INH-18 (2 μM) had a marked synergistic effect. In all 6 patients, migration was reduced to a level significantly below that seen with either drug alone; median combination index at the half maximal effective dose for the combination of the 2 drugs was 0.2, indicating a strong synergistic effect (Figure 7A-B). The combination also synergistically reduced p-p65 NF-κB and p-STAT3 but not p-STAT5 (Figure 7C-F; supplemental Figure 8). These data clearly support our hypothesis that inhibition of the TLR9 pathway, along with BCR signaling, is a potential therapeutic strategy, especially given a report that TLR9 signaling provides a common escape mechanism for CLL cells after treatment with ibrutinib and venetoclax.⁴⁶ 

Over the last decade, targeting BCR-related signaling pathways has led to improved therapies for CLL. However, despite these advances, CLL remains incurable. The identification of MyD88-activating mutations in ∼3% of CLL cases,¹⁸  supported by the potential for TLR9 to be triggered by autoantigens/DNA,⁴⁷  has prompted new research into the role of TLR signaling in CLL.⁴⁸ 

Here, we show for the first time that autologous plasma from patients with CLL contains a disproportionately high level of unmethylated mtDNA, which can trigger TLR9 signaling, resulting in downstream phosphorylation of STAT3 and NF-κB. Importantly, all of these effects can be reversed by TLR9 inhibition. Furthermore, we found that patients with aggressive disease have increased levels of cfDNA, which correlate strongly with the dynamic prognostic markers CD38, B2M, net proliferation (LDT), tumor burden, and shorter time to first treatment. These strong correlations suggest that cfDNA may promote tumor proliferation. There was also a trend toward higher cfDNA and unmutated IGHV status; although mutational status itself is not dynamic, it is recognized that an unmutated IGHV correlates with raised CD38,⁴⁹  which in turn is regulated by CLL cell interactions in the lymphoid tissue microenvironment.³⁴  It is therefore possible that raised cfDNA promotes CLL cell activation and trafficking to pro-proliferative lymphoid tissues. Although raised levels of cfDNA did not correlate with ZAP70 expression, it remains possible that the adapter molecule function of ZAP70^50,51  may enable ZAP70-positive CLL cells to signal via TLR9, regardless of cfDNA levels.

CD38 and CD49d are associated with each other^52,53  and CLL cell homing,⁵⁴  where it is known that the malignant clone differs from normal B cells in that α4β1 (CD49d) engagement is essential for transendothelial migration.³⁹  Here, we show that TLR9 stimulation increases CD38, CD49d, and migration. Within a patient’s CLL clone, eTLR9^bright cells exhibited higher expression of CD38, CD49d, p-STAT3, and p-p65 NF-κB. Furthermore, in contrast to TLR9^lo CLL cells, TLR9^hi CLL cells exhibited good engraftment in NOD/Shi-scid/IL-2Rγ^null mice. Taken together, these findings both support and expand on our previous study⁶  and suggest an important role for TLR9 in vivo.

TLR9 is usually triggered by pathogenic DNA, which is structurally identical to self-DNA except for the presence of unmethylated CG dinucleotides. CG motifs are underrepresented within the mammalian genome, and the majority exist within the heavily methylated gene promoter regions.⁵⁵  However, mtDNA is almost entirely unmethylated^55,56  and given that exposed mtDNA is a natural and extracellular-residing product of cell death, it is vital that it does not come into contact with functional TLR9. Therefore, in normal resting conditions, TLR9 is sequestered within the endoplasmic reticulum and transported to the endosomal membrane in its inactive pre-receptor form after the endocytosis of CpG DNA.^50,57,58  TLR9 is then transported via the Golgi-dependent secretory pathway and is temporarily expressed at the cell surface before re-internalization. TLR9 functionality is not believed to be enabled until it is successfully integrated and cleaved within the endosomal membrane.⁵⁹  Such tight regulation of TLR9 is fundamental to the distinction of self/non–self-DNA and ensures that functional TLR9 only encounters pathogen-derived nucleic acids. However, a hallmark of CLL is the large tumor turnover,⁶⁰  and it is possible that extensive apoptosis, and the generation of excessive cfDNA, could overwhelm this regulatory system and trigger autologous TLR9 activation.

Surface TLR9 has been described in solid tumors and linked to tumorigenesis and cancer progression.²⁵  We hypothesize that the high cell turnover seen in malignancies such as CLL could trigger dysregulation of the distinction of self/non–self-DNA and thus result in sustained surface expression of TLR9. In support of this hypothesis, we described sTLR9, on a small population of CLL cells within patients, which is notably higher on those that migrate or reside within the LNs. Importantly, we also showed that blocking of sTLR9 marginally reduces CLL activation and migration, indicating some functional properties. However, its strong correlation with eTLR9 indicates that expression is likely a consequence of sustained eTLR9 signaling in the most aggressive CLL cell subpopulation, and this theory is supported by the striking TLR gene set enrichment seen in sorted sTLR9^hi cells. Importantly, raised sTLR9 could potentially represent a biomarker for patients who would particularly benefit from TLR9 inhibition. Finally, we showed that total blockade of TLR9 is highly synergistic with ibrutinib and reduces p-p65 NF-κB– and p-STAT3–driven migration down below that which either drug alone can achieve.

These findings strongly support the rationale for further investigations into the role of TLR9 in CLL. It is clear that blockade of BTK or PI3K alone is not curative, and these tumor cells have mechanisms for escape and clonal evolution that involve TLR9 signaling.⁴⁶  Therefore, simultaneous targeting of multiple signaling pathways could provide an opportunity for total tumor eradication. In this regard, we have shown that there is a strong biological rationale for dual targeting of BTK and TLR9.

This work was funded by a Blood Cancer UK Programme Continuity grant (18005), the Sussex Cancer Fund, the Crawley Leukaemia Research, and the Michael Chowen Research Fund, all awarded to C.P. and A.G.S.P.

Contribution: E.K., E.C., E.H., N.P.-P., E.W., B.C., E.H.P., T.A.B., S.M., R.S., M.S., A.A., C.O., N.D., T.S., T.L., and C.I.J. performed experiments, analyzed data, and revised the manuscript; S.D., R.J., and T.C. provided vital reagents and revised the manuscript; and A.G.S.P., C.D.F., and C.P. conceived the study, analyzed data, and drafted the manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Andrea G. S. Pepper, Department of Clinical and Experimental Medicine, Brighton and Sussex Medical School, Medical Research Building, Biology Rd, Falmer, BN1 9PS United Kingdom; e-mail: a.pepper@bsms.ac.uk.