DDX3X loss is an adverse prognostic marker in diffuse large B-cell lymphoma and is associated with chemoresistance in aggressive non-Hodgkin lymphoma subtypes

Atish Kizhakeyil, Nurmahirah Binte Mohammed Zaini, Zhi Sheng Poh, Brandon Han Siang Wong, Xinpeng Loh, Aik Seng Ng, Zun Siong Low, Praseetha Prasannan, Chun Gong, Michelle Guet Khim Tan, Chandramouli Nagarajan, Dachuan Huang, Pang Wan Lu, Jing Quan Lim, Sharon Barrans, Choon Kiat Ong, Soon Thye Lim, Wee Joo Chng, George Follows, Daniel J. Hodson, Ming Qing Du, Yeow Tee Goh, Suat Hoon Tan, Nicholas Francis Grigoropoulos and Navin Kumar Verma


Main text
Non-Hodgkin a diverse group of malignancies, encompassing the common diffuse large B-cell lymphoma (DLBCL) to the rarer T-cell lymphomas. Chemoresistance is a major barrier to treatment and the mechanisms through which it occurs are incompletely understood [1]. Although efforts are made to target frequently dysregulated pathways in NHL subtypes, these diseases still evolve into aggressive forms resistant even to newer therapies [2]. DEAD box helicase 3, X-linked (DDX3X) is an ATPdependent RNA helicase and is involved in multiple cancer-related cellular processes, including transcriptional regulation, cell adhesion, signal transduction and stemness [3]. It plays tumor suppressive or oncogenic roles depending on tumor type, and is implicated in various cancer types, including glioma, medulloblastoma, squamous cell carcinoma, hepatocellular carcinoma, lung and breast cancer [3]. A role for DDX3X in NHL subtypes remains unclear.
Here, we present our novel findings that DDX3X mutations are associated with poor prognosis in DLBCL. We demonstrate that mutation/loss of DDX3X in cell lines originated from DLBCL, cuteneous T-cell lymphoma (CTCL) and NK-cell/T-cell lymphoma (NKTCL) results in increased STAT3/p42/p44 phosphotylation and the development of chemoresistance. Intriguingly, DDX3X mutated/depleted B-and T-cell lineage NHL cell subtypes remain sensitive to pharmacological inhibitors of STAT3.
in 49 out of 1343 (3.6%) DLBCL cases. Moreover, available database in the ICGC [5] and the COSMIC [6] portals revealed mutations in DDX3X in 63 out of 1319 (4.7%) and 160 out of 5160 (3.1%) DLBCL cases, respectively. Besides DLBCL, DDX3X was also recurently mutated in other NHL subtypes, including Burkitt lymphoma (BL) and NKTCL [7]. Altogether, we identified 165 missense, 81 truncating, and 3 in-frame indel mutational variants in the DDX3X in NHL subtypes (Fig. 1A). DLBCL patients with DDX3X mutations (14 cases out of 223 from 3 different cohorts) had significantly worse median overall survival (41.13 months) compared to cases with wild-type DDX3X (211.07 months) (Fig. 1B). The 5-year overall survival of DLBCL patients with DDX3X lesions was only 22% compared to 72% (p = 0.021) for patients without DDX3X alterations (Fig. 1B). Data availability for other end-points, such as progression-free survival or event-free survival, was insufficient for a robust analysis. Notably, we found several other mutations that co-existed in DDX3X mutant DLBCL cases (Table  S2).
We performed whole exome sequencing and detected DDX3X mutations in 4 out of 9 relapsed/ refractory DLBCL patients treated with R-CHOP or similar regimens. Five damaging DDX3X mutations were identified (PolyPhen-2 score = 1, Table  S1) in the catalytic domains -R296C and V375fs*8 in the "helicase ATP-binding" and R475C, R475H and R534H in the "helicase C-terminal" (Fig. 1C). Targeted sequencing of DDX3X on exons 8-15 in samples from 158 unselected DLBCL patients showed damaging DDX3X variants in 5 subjects (3, 95% confidence interval 1-7%) (Fig. 1C). All the identified DDX3X mutations were confirmed somatic and occurred in residues that are highly conserved across various species. These data indicate that DDX3X mutations are important determinants of response to chemotherapy in DLBCL.

DDX3X mutation/loss in NHL cell subtypes increases resistance to antineoplastic drugs
To determine the functional role of DDX3X mutations in drug resistance, we stably expressed mutant DDX3X-R475C in U2392 cells using CRISPR knock-in technique [8] and knocked-down DDX3X in a panel of NHL cell lines [DLBCL (U2392, BJAB), BL (Raji,), NKTCL (SNK1, SNK6, NKYS), and CTCL (HuT78, MJ, MyLa)] using DDX3X-targeted shRNA or siRNA (Fig. S1). Cells were treated with IC 50 concentrations (of wild-type cells) of commonly used histone deacetylase (HDAC) inhibitors (vorinostat, panobinostat, trichostatin, romidepsin), STAT3 inhibitors (stattic, WP1066) or doxorubicin and cell viability was analyzed by MTS-based assay. We observed significant resistance to doxorubicin in DDX3X-mutant/depleted U2932 and BJAB cells and to HDAC inhibitors in HuT78 and SNK1 cells ( Fig. 1D;  Fig. S2). Importantly, DDX3X-mutant/depleted NHL cell subtypes remained sensitive to pharmacological STAT3 inhibition (Fig. 1D), although no synergism of drug action was identified in these cell-types as analyzed by determining "Combination Index" using checkerboard assay and the CompuSyn software (Fig. 1E). These results strongly support the clinical data suggesting that DDX3X mutations cause true chemoresistance and worse overall survival, rather than simply being surrogate markers in NHL subtypes, including DLBCL, NKTCL and CTCL.

DDX3X mutation/loss enhances proliferation and migratory potential of NHL cell subtypes
DDX3X mutation/depletion in U2932, HuT78 and SNK1 cells significantly increased their proliferation rates in comparison to controls, as determined by MTS-based assay over 7 days (Fig. 1F). In contrast, forced overexpression of wild-type DDX3X in U2932 and HuT78 cells significantly decreased proliferation (Fig. S3). Real-time monitoring of cells transmigrating through Matrigel, using impedance-based measurements, showed significantly increased migratory/invasive potentials of DDX3X-mutant/depleted NHL cell subtypes (U2932, HuT78 and SNK1) (Fig. 1G). Western immunoblot analysis of DDX3X-mutant/depleted U2932 and HuT78 cells detected increased vimentin expression in these cell-types (Fig. S4). Overexpression of vimentin has been associated with aggressive transformation in lymphoma [9].

DDX3X mutation/loss in NHL sybtype-derived cells increase cyclin-D1 expression
To further investigate the role of DDX3X in NHL subtypes, we carried out whole RNA-seq analysis of mutant DDX3X-R475C U2392, DDX3X-depleted HuT78 and DDX3X-depleted SNK1 cells in comparison to their wild-type and identified 451, 1682 and 441 differentially expressed genes (DEGs), respectively ( Fig. 2A; Table S3-S5). We next performed gene network analysis using IPA ® (a web-based application that enables analysis, integration, and interpretation of RNA-seq and other 'omics' datasets), DAVID (a biological module-centric algorithm that functionally analyzes large gene lists), and GSEA (an analytical method that determines whether an a priori defined gene sets shows significant differences between two biological states). This multimodal analysis of DEGs showed that DDX3X mutation/loss in NHL cells enriched cyclin-D1 and JAK-STAT3 pathways ( Fig. 2B; Fig. S5; Table S6). Using RT-qPCR and Western-immunoblot analysis, we verified that cyclin-D1 mRNA and protein levels were significantly elevated in DDX3X-mutant/depleted U2392 and Raji cells relative to control (Fig. 2C,D; Fig. S6). A subset of NHL patients, including 2.1% of DLBCL cases, has been found to overexpress cyclin-D1 [10] and the upregulation of cyclin-D1 has been associated with doxorubicin resistance in gastric cancer [11].

Conclusions
Despite being uncommon, the biological effects DDX3X mutations/loss in DLBCL and other NHL subtypes (NKTCL, CTCL), suggest a worse overall survival. Given the high incidence of DLBCL [13,14], the absolute number of patients with mutated DDX3X is likely to be substantial. Our results also suggest that STAT3 inhibition may be a rational choice for chemoresistant NHL (DLBCL, NKTCL, CTCL) patients with mutated DDX3X. The STAT3 inhibitor AZD9150 is currently in phase 1b clinical trial in a subset of patients with heavily pre-treated lymphoma [15]. Further studies are needed to improve our understanding of the consequences of DDX3X mutation/loss, which would help improve risk stratification of aggressive NHL subtypes and identify new therapeutic options for patients with poor prognosis.

Supplementary Information
The online version contains supplementary material available at https:// doi. org/ 10. 1186/ s12943-021-01437-0. p-values were calculated by 1000-gene-set two-sided permutation tests. Supplementary Fig. 6. Upregulation of cyclin-D1 in BL cells upon DDX3X depletion. Control and DDX3X-depleted Raji cells (using shRNA#1 or shRNA#2) were lysed, and cellular lysates were analyzed for the expression levels of DDX3X and cyclin-D1 by Western immunoblotting. Blots were re-probed with anti-β-actin to confirm equal loading. Data represent at least 3 independent experiments. Values in the relative densitometry graphs are mean ± SEM; *, p < 0.05. Supplementary Fig. 7. Effect of DDX3X loss on STAT3 regulation in CTCL cells. (A) HuT78 cells were treated with DDX3X siRNA and STAT3 GapmeR. Cellular lysates were collected and Western immunoblotted to determine the expression levels pSTAT3(Y705), STAT3, DDX3X, and GAPDH (loading control). Relative densitometry graph of immunoblots is presented. *, p < 0.01. (B) Cell lysates of HuT78 cells were immuno-precipitated (IP) using anti-DDX3X antibody or IgG control. Subsequently the immuno-precipitates were resolved on SDS-PAGE and subjected to immunoblotting with anti-STAT3 and anti-DDX3X antibodies. Data represent at least 3 independent experiments.