IKKε and TBK1 in diffuse large B‐cell lymphoma: A possible mechanism of action of an IKKε/TBK1 inhibitor to repress NF‐κB and IL‐10 signalling

Abstract The IKK‐related kinases, IKKε and TBK1, have essential roles in innate immunity in part through modifying MYD88 signalling from the Toll‐like receptors to regulate NF‐κB signalling. We investigated the expression and function of IKKε and TBK1, in diffuse large B‐cell lymphoma (DLBCL). DLBCL cell lines and patient‐derived xenografts were used to determine their sensitivity to IKKε and TBK1 inhibitors. To understand the function of IKKε and TBK1 secreted factors were determined following administration of inhibitors. Gene expression microarrays were used to determine the transcriptional effects of inhibitors. Higher TBK1 mRNA levels associated with poorer clinical outcome but IKKε and TBK1 were expressed in both germinal centre and non‐germinal centre types of DLBCL. Survival of cell lines Ly10, Ly03 and Pfeiffer, and of some primary human lymphoma cells, was suppressed by a small molecule IKKε/TBK1 inhibitor, DMX3433. DMX3433 reduced IL‐10 production from Ly10 and repressed NF‐κB mediated transcription. Inhibition of IKKε and TBK1 warrants further investigation as a potential therapeutic route to suppress NF‐κB signalling in lymphoma.

may be associated with compensatory IKKα signalling arguing against the feasibility of this therapeutic route. 15 The standard-of-care treatment of DLBCL is combination immuno-chemotherapy with rituximab, cyclophosphamide, adriamycin, vincristine and prednisolone (RCHOP), and this is curative in a small majority of patients although at the cost of morbidity and mortality due to neutropenic sepsis. 16 Patients who are not cured clearly require different treatments and pre-clinical studies have suggested that small molecule inhibitors of Bruton's tyrosine kinase (BTK) 17 or phosphatidylinositol-4,5-bisphosphate 3-kinase catalytic subunit delta (PI3Kδ) 18 might be effective routes to inhibiting NF-κB signalling. The effects of PI3Kδ, however, may be transient and reduced by a compensatory increase in PI3Kα, 19 a result which seems to be reflected in a 30% response rate obtained in an early-stage clinical trial of the small molecule PI3Kδ inhibitor, parsaclisib, in refractory or relapsed DLBCL. 20 A phase III trial investigating the BTK inhibitor, ibrutinib, in combination with RCHOP in untreated DLBCL has been reported 21 but the addition of ibrutinib did not improve event-free survival. Other approaches to repressing NF-κB signalling are, therefore, needed.
IKK-related kinases IκB kinase ε (IKKε, also named IKKi or IKBKE) 22,23 and tank binding kinase 1 (TBK1) function in response to bacterial and viral signalling in normal immunity. TBK1 shows 64% homology to IKKε at the amino acid level. 24 There are many uncertainties about the function of these kinases 25 but it is clear that IKKε/TBK1 act downstream of Toll-like receptors (TLRs) and MYD88, to phosphorylate interferon regulatory factors (IRF3 and IRF7) and induce type 1 interferon genes. 26,27 IKKε and TBK1 also regulate NF-κB activity by directly phosphorylating RELA 24,28 controlling constitutive (but not cytokine-induced) RELA phosphorylation at serine 536 29 and promoting transactivation by p52 in the non-canonical pathway. 30 B-cell specific roles have also been suggested and one study showed that TBK1 repressed the adapter NIK to prevent IgA class switching. 31 The gene encoding IKKε, IKBKE, is amplified in 8 of 49 (16.3%) breast cancer cell lines and in over 30% of primary breast tumours 32 and is essential to the survival of breast cancer cell lines through activation of NF-κB suggesting that IKKε is a breast cancer oncogene.
Inhibition of IKKε and TBK1 is a potential route to repression of RELA activity but has not been previously considered as a potential therapeutic target in DLBCL. Here, we have employed small molecule inhibitors of IKKε/TBK1 to investigate the role of IKKε and TBK1 in DLBCL. are classified as Type III. 33

| Luminex assay
Using information from the literature, a Luminex panel was designed to account for the major growth factors and chemokines produced by B cells. Initial experiments comparing Ly03 and Ly10 showed that Ly10 produced the greatest amounts of growth factor in baseline conditions.

| Taqman assay
Total mRNA was extracted from harvested cells using a RNeasy Mini Kit (Qiagen). Reverse transcription was carried out with the SensiFAST™ cDNA synthesis kit using the manufacturer's protocol (Bioline). Reactions were then carried out using Taqman primers for IL10 (Hs00961622_m1) and HPRT (Hs02800695) (Applied Biosystems).

| Immunohistochemistry and Immunofluorescence
A human DLBCL tissue microarray was used consisting of 72 cases of DLBCL (catalog number LY1001c; US Biomax Inc) of which 7 cases could not be used. The GC/non-GC status can be found at https:// www.biomax.us/tissu e-array s/Lymph oma/LY1001c.  (Table S1). Tumour growth was monitored twice weekly using a caliper and all efforts were made to minimize suffering. 35 Animals were euthanized by CO 2 inhalation. Characteristics of PDX models used (PDX0257, PDX2345, PDX2214 and PDX2318) are presented (Table S2).

| Patient-derived xenograft models
Ex vivo 2D cultures were set up at a cell concentration of 1 × 10 5 / mL in a 96-well plate. Viability following the addition of drug was measured at 24 hours using CellTiter-Glo ® (Promega).
To generate cell pellets, 2 × 10 6 cells were seeded in 1.9 mL of X-vivo 15 basal growth medium per well of a 6-well plate. Cells were then incubated overnight, followed by incubation for 24 hours with drug or vehicle control (DMSO). Post incubation, cell supernatant was removed, and the cells were harvested and centrifuged. Cell pellets were then stored at −80°C prior to shipping on dry ice.

| Gene expression microarray analysis
Total RNA was purified from PDX model cell pellets. RNA isolation was carried out by means of Trizol/chloroform phase separation followed by PureLink ® RNA Mini Kit (ThermoFisher Scientific) procedure. RNA quality was checked on a Bioanalyzer 2100 (Agilent).
All RNA samples had a RNA Integration Number (RIN) > 7. A total of 100 ng of total RNA were reverse transcribed, converted into complementary RNA (cRNA) and labelled with Cy3 using the LowInput QuickAmp Labeling Kit One-Color according to manufacturer's protocol (Agilent). Labelled cRNA was then hybridized over night at 65°C onto the SurePrint G3 Human Gene Expression v3 8 × 60 K Microarray and scanned on an Agilent DNA microarray C-scanner.
Extraction and quality check of the raw data were performed using the Agilent Feature extraction software version 10.5.1.1.

Quantile normalization of data was performed using Partek
Genomic suite (Partek Inc). Normalized data were then imported into Multiple Experiment Viewer (MeV), 36 and a two-class unpaired significance analysis for microarrays (SAM) analysis was performed.
Hierarchical clustering (distance metric selection by Pearson correlation, linkage by average linkage clustering) was carried out for significant genes (FDR q-value of <0.25 and a nominal P value of <.05). The data that support the findings of this study are openly available in Gene Expression Omnibus (GEO) at https://www.ncbi.

| Gene set enrichment analysis
Gene expression values from untreated and treated PDX models were analysed with the desktop GSEA software (Broad institute). The PDX expression set was interrogated for enrichment with hallmark gene sets in the Broad molecular signature database. Enrichment with pathways was determined using thresholds of a FDR q-value of <0.25, a normalized enrichment score <−1 or >1 and a nominal P-value of <.05.

| Kaplan-Meier survival analysis
Survival data were generated using SurvExpress 37  Data were then subjected to Kaplan-Meier survival analysis using GraphPad Prism 6.0.

| IKBKE and TBK1 mRNA expression and prognosis in DLBCL
Analysis of a publicly available gene expression data set (GSE10846 38 ) showed that, across the entire cohort of patients with DLBCL, those expressing TBK1 at levels greater than the median had significantly worse overall survival (log-rank test, P = .028) ( Figure 1A). IKBKE levels did not associate with survival ( Figure S2). TBK1 mRNA levels did not associate with differences in patient characteristics in respect of age, clinical stage and performance score (Table S3). Inspection of mRNA levels showed no significant TBK1 differences across patient groups ( Figure 1B) while IKBKE levels were higher in unclassified (Mann-Whitney test: P = <.0001) and GC-DLBCL (P = .0002) than in ABC-DLBCL.
Overall despite mRNA levels of IKBKE associating with specific subgroups of DLBCL, this correlation does not appear to influence patient survival. However, independent of COO subgroups higher TBK1 mRNA levels associated with poorer clinical outcomes.

| IKKε and TBK1 expression in primary DLBCL
Next, using a tissue microarrays (n = 65) ( Figure 1C (Table S4). IKKε and TBK1 are, therefore, found in both GC-and non-GC cases with a greater proportion of non-GC cases than GC cases showing expression.
Overall higher TBK1 mRNA levels associate with poor clinical outcome but there is no difference in TBK1 levels between GC-and non-GC-DLBCL. However, TBK1 protein levels varied from undetectable to strongly expressed on the TMA. This suggests post-transcriptional control contributes to the regulation of TBK1 protein levels.  Figure 2C).

| IKKε and TBK1 expression in DLBCL cell lines and sensitivity to small molecule IKKε/ TBK1 inhibitors
To support the idea that loss of viability was due to on-target effects on IKKε and TBK1, we determined the sensitivity of Ly10 and Ly03 to a set of nine IKKε/TBK1 inhibitors (described in patent PCT/ GB2014/050521) with pIC50 for the purified enzymes varying from 5.8 to 8.6 (for IKKε) and from 6 to 8.3 (for TBK1). There were significant associations between inhibition (pIC50) of TBK1 ( Figure 2D) or IKKε ( Figure 2E) activity and pEC50 for reduction in cell viability suggesting that loss of viability was indeed through effects on IKKε and TBK1.
In order to detect early changes in phosphorylated proteins Western blots were carried out at time-points up to 8 hours for Ly10 ( Figure 2F) and Ly03 ( Figure 2G). DMX3433 caused a reduction in phosphorylated RELA (S536) in both cell lines. We also noted complex effects on STAT3 phosphorylation with an early reduction in STAT3 phosphorylation at tyrosine 705 but an increase at 4 hours in both Ly03 and Ly10 which was sustained in Ly03.

| DMX3433 alters the output of growth factors and chemokines
NF-κB drives production of some cytokines, and these, in turn, activate STATs. Using multiplex ELISA to detect a panel of growth factors and chemokines, we investigated the hypothesis that DMX3433 modifies secretion of these proteins in cell lines Ly03 and Ly10, which are the cell lines most sensitive to DMX3433 and, in contrast to other DLBCL cell lines, are known to secrete cytokines 39 (Figures 3A and S3). 8 of the 15 proteins in our panel were detectable after 48 hours culture without inhibitor, in either Ly03 or Ly10 supernatants but only two of these (IL-10 and CCL22) were detectable in supernatants from both cell lines. Ly10, which demonstrated the greatest effects of inhibitor to amounts of secreted protein, was used for further studies with DMX3433. This showed significant repression of CCL3 (paired t test, P = .042 at 1 µmol/L and P = .033 at 2 µmol/L), CCL4 (P = .032 at 1 µmol/L and P = .023 at 2 µmol/L) and IL-10 (P = .031 at 1 µmol/L and P = .032 at 2 µmol/L) ( Figure 3B). IL10 signalling is required for DLBCL survival. 40 In order to investigate the hypothesis that autocrine IL-10 signalling is regulated by IKKε/TBK1 and contributes to STAT3 phosphorylation, we added CCL3, CCL4 or IL-10 to Ly10 cell culture in the presence or absence of DMX3433 ( Figure 4A). In the absence of DMX3433 phosphorylation of STAT3 at tyrosine 705 was induced by IL-10 but not by CCL3 or CCL4. DMX3433 abolished STAT3 phosphorylation, which was restored by IL-10 suggesting that abolition of tyrosine phosphorylation is mediated principally through this cytokine in Ly10 ( Figure 3C). For each plot, the blue line represents patients with mRNA levels greater than the median and the red line those patients with levels less than the median. Data on expression and survival is from GSE10846 38 using TBK1 probeset 218520_at and IKBKE probeset 204549_at. Higher TBK1 levels are associated with poor clinical outcome (log-rank test, P = .042), which is not observed for IKBKE levels (P = .88). B, Expression of IKBKE and TBK1 mRNA in ABC, GC and unclassified (Un) subgroups of DLBCL. Data taken from, 38 GEO dataset GSE10846. Median is indicated by the horizontal bar. For IKBKE median levels of GC cases (Mann-Whitney test: P = .0002) and unclassified (P = <.0001) are significantly higher than levels in ABC cases. C, Immunofluorescence images from representative cases on the DLBCL TMA. The antibody used is indicated to the right either anti-TBK1 (yellow) or anti-IKKε (green). Both have also been stained with DAPI to define cell nuclei (blue). Representative examples are shown with their position on the microarray indicated by the letter and number above the images. IKKε hi /TBK1 lo (5G), IKKε lo /TBK1 hi (7A), IKKε hi /TBK1 hi (8F) and IKKε lo /TBK1 lo (6A). D, Histograms showing numbers of DLBCL cases (defined by expression of IKKε and TBK1) of GC (blue) or non-GC (red) type in the tissue microarray of DLBCL. E, Western blots showing IKKε and TBK1 expression in a panel of DLBCL cell lines. GAPDH is a loading control. Molecular weight (kDa) is indicated to the left. The numbers below the IKKε and TBK1 autorads are ratios of IKKε:GAPDH and TBK1:GAPDH respectively. F, Expression of the NF-κB component RELA and phosphorylated RELA together with STAT3 and phosphorylated STAT3 (Y705) in a panel of DLBCL cell lines. GAPDH is a loading control. The numbers below the pRELA and pSTAT3 autorads are ratios of pRELA/GAPDH:RELA/ GAPDH and pSTAT3/GAPDH:STAT3/GAPDH respectively. Uncropped autorads are in the Appendix repression by the inhibitor is required for reduction in levels of the cytokine in culture supernatants ( Figure 3D).

| Sensitivity of primary human DLBCL to IKKε/ TBK1 inhibitors
Next, we determined the effects of DMX3433 on primary human DLBCL. Four patient-derived xenografts (PDX) chosen to include two non-GC-DLBCL (PDX0257 and PDX2345) and two GC-DLBCL (PDX2214 and PDX2318) were cultured and treated with varying concentrations of drug. PDX0257 and PDX2214 showed the greatest sensitivity to the inhibitor with EC50 of 2.1 and 1.8 µmol/L, respectively, but PDX2318 had an EC50 of 6.6 µmol/L and for PDX2345 EC50 was not reached ( Figure 4A). Neither cell-of-origin classification nor expression of TBK1 or IKKε appeared to associate with sensitivity to DMX3433 but both sensitive models showed SOCS1 mutations (Table S2)

| D ISCUSS I ON
We found that IKKε and TBK1 are expressed in both GC-and non-GC-DLBCL while there were also cases of both subgroups that express no detectable protein. Therefore, expression of these proteins does not segregate with COO subgroups. Recent genetic analysis made possible by comprehensive CRISPR screens suggests that ABC-and GC-DLBCL have some driver mutations in common. 4 Our results suggest that IKKε and TBK1 are similarly expressed in both COO subgroups.
Analysis of the publicly available gene expression data shows no significant variation in TBK1 mRNA levels across COO subgroups ( Figure 1B Figure S1). IKKε/TBK1 have been demonstrated to be part of a complex that activates NF-κB. 22 There is, therefore, likely to be cross-talk between the MYD88 and IKKε/ TBK1 pathways such that both contribute to NF-κB signalling. Ly10 and Ly03 ( Figure 1F) show constitutive RELA phosphorylation consistent with their known activating MYD88 change p.Leu265Pro and IKKε/TBK1 inhibition reduced RELA phosphorylation. Collectively, our work showed effects of inhibitors on RELA phosphorylation and gene expression, which supports a role for IKKε/TBK1 in contributing to NF-κB activity in some DLBCL and we note that Ly10, which was used for the gene expression studies has constitutive NF-κB activity due to MYD88 mutation. Plausibly IKKε/TBK1 are downstream of MYD88 in this cell line. Further work will be needed to establish if MYD88 (p.Leu265Pro) associates with IKKε/TBK1 signalling and sensitivity to IKKε/TBK1 inhibition. However, we found that one cell line sensitive to DMX3433, Pfeiffer, as well as the less sensitive cell lines, did not have detectable phosphorylated RELA (or mutated MYD88).
We do not have a clear explanation for why DMX3433 reduced viability of these lines but this may be mediated through effects on other pathways such as interferon signalling and STAT1 or STAT2. 43,44 Although gene expression profiling supported a role for IKKε/ TBK1 in repressing NF-κB activity we observed that mRNA levels of some type I interferon genes was increased by DMX3433 in primary DLBCL cases. This is consistent with results of a recent study 45 showing that STAT3 suppresses apoptosis by inhibition of IRF7, STAT1, STAT2 and IRF9 mRNA expression to prevent type I interferon responses in DLBCL. An implication of our work is that one route leading to reduction in viability following inhibition of IKKε/TBK1 is repression of STAT3 phosphorylation, and increased interferon signalling. Proliferation and survival of subgroups of ABC-DLBCL are dependent on STAT3 signalling 13,46 and a proportion of GC-DLBCL cases express pSTAT3. 46 The mechanism of expression and the clinical importance of pSTAT3 in the GC-DLBCL subgroup have not been investigated in part because of the lack of representative DLBCL cell lines. STAT3 signalling is associated with proliferation and poor clinical outcome in DLBCL. 47 Earlystage clinical trials of STAT3 inhibitors in a variety of disease settings have been reported 48 and although STAT3 inhibitors do not yet have approvals to treat cancer it is a potential therapeutic target for subgroups of DLBCL.
Biomarkers predicting response to treatment are necessary in order to direct therapeutic agents to patient groups that will benefit the most. We found that CCL3 and CCL4 (as well as IL-10) were highly expressed by Ly10 and repressed by DMX3433 but levels of another chemokine, CCL22, were induced by the inhibitor. CCL22 is repressed by type I interferon responses 49 and it is, therefore, possible that the increase we observed with DMX3433 is an on-target effect. Circulating levels of CCL3 and CCL4 associate with clinical outcome in ABC-DLBCL but not in GC-DLBCL. 50 Higher levels of circulating IL-10 correlate with shorter event-free survival and higher lactose dehydrogenase (LDH) levels 51 and the genes for the IL10 receptors (IL10RA and IL10RB) are amplified in 21% (IL10RA) and 10% (IL10RB) of DLBCL patients, respectively, and associated with higher IL10RA and IL10RB mRNA than normal subjects. 40 Levels of CCL3, CCL4 or IL-10 may, therefore, be suitable biomarkers predicting sensitivity to IKKε/TBK1 inhibitors. The most sensitive PDX models, PDX0257 and PDX2214, show SOCS1 mutations, which could enhance STAT3 signalling.
The clinical effects of SOCS1 mutations are controversial but a recent retrospective study of elderly patients who had received R-CHOP suggested that the presence of mutations confers a poor prognosis. One possibility prompted by our work is that SOCS1 mutation might be a marker for response to IKKε/TBK1 inhibitors.
A limitation of our work is that we focused on the Ly10 DLBCL cell line and used relatively few PDX models. Limitations on the amount of material available from the PDX models also meant that we could not carry out a more extensive analysis. Ly10 is favourable because it produces relatively large amounts of growth factors and has been used by others for this purpose 39 but it is clearly not representative of DLBCL. In view of the heterogeneity of DLBCL, larger studies using both cell lines and primary cells will be required.
Data presented here demonstrate that the innate immunity components, IKKε and TBK1 are expressed in primary DLBCL but in a manner that is not strictly dependent on COO subgroup. IKKε/TBK1 inhibitors suppress survival of some DLBCL cell lines and focusing on the most sensitive cell line, Ly10, we show that an IKKε/TBK1 inhibitor suppresses RELA phosphorylation, IL-10 production and STAT3 phosphorylation. We observed that both the sensitive DLBCL cell lines and the sensitive PDX models had high STAT3 expression, a F I G U R E 4 Pre-clinical testing of DMX3433 in PDX models of DLBCL. A, Dose-response curves showing effects of DMX3433 on cell survival determined by ATP luminescence on primary human lymphoma cells cultured for 24 h. Four models PDX0257, PDX2214, PDX2345 and PDX2318 were investigated. IC50s (µmol/L) are presented for three of the models but PDX2345 was essentially insensitive. Mean ± SEM. Assays were run in technical triplicate. B, Western blots showing baseline IKKε and TBK1 levels. Phosphorylated (tyrosine 705) and total STAT3 and phosphorylated (serine 536) and total RELA are also shown. Molecular weights (kDa) are indicated to the left. β-actin is a loading control. Uncropped autorads are in the Appendix. C, Gene set enrichment analysis. Using the entire dataset untreated cells are enriched for the Hallmark gene set "TNFA signalling through NF-κB" (MSigDB M5890) (FDR = 0.199) and "IFNA response" (M5911) (FDR = 0.005). The leading edge genes for each gene set are presented to the right as a heat map. Lanes highlighted in orange are untreated and lanes in grey are treated. Due to insufficient sample, one treated PDX model (PDX2345) was not included on the treated lanes previously defined characteristic of some DLBCL in both GC-and ABC-DLBCL subgroups. 13,46 Further work will be necessary to characterize in detail the subgroups of DLBCL that are most sensitive to inhibition of IKKε/TBK1.

ACK N OWLED G EM ENTS
We would like to thank Drs. Gary Newton, Claire Tebbutt and Stuart Crosby (Domainex Ltd.) for biochemical assay measurements and technical assistance. We thank Drs. Nic Sylvius and Spencer Gibson, (CBS Genomics Core Facility, University of Leicester) for carrying out gene expression microarray hybridization.

CO N FLI C T O F I NTE R E S T
The authors declare no competing financial interests other than that TP and KLC are employees of Domainex Ltd. and TP is an inventor of patent PCT/GB2014/050521 that claims inter alia DMX3433 as an IKKε/TBK1 inhibitor.

E TH I C S A PPROVA L A N D CO N S E NT TO PA RTI CI PATE
All animal studies were conducted at Crown Bioscience HuPrime SPF animal facility (CrownBio, Beijing, China) under sterile conditions and were in strict accordance with the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health.
Protocols for all studies were approved by the Committee on the Ethics of Animal Experiments of Crown Bioscience, Inc (Crown Bioscience IACUC Committee). The patient-derived xenograft models were obtained from Crown Bioscience.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are openly available in Gene Expression Omnibus (GEO) at https://www.ncbi.nlm.nih. gov/geo/, reference number GSE121159.