Characteristics and prognosis of rrDLBCL with TP53 mutations and a high‐risk subgroup represented by the co‐mutations of DDX3X‐TP53

Abstract Background TP53 mutations have a prognostic significance in relapsed and refractory diffuse large B‐cell lymphoma (rrDLBCL) patients, and their treatment still faces a great challenge. This study aimed to evaluate the prognosis of patients with TP53 mutations (TP53mut) in the context of CAR‐T therapy (Chimeric antigen receptor T‐cell therapy) as well as explore the heterogeneity in their cohort and identify the possible risk factors. Methods A retrospective study was conducted to investigate the clinical characteristics of rrDLBCL patients with TP53 mutations and their prognostic factors, receiving CAR‐T therapy. And the expression level of TP53 and DDX3X, which was an important co‐mutation of TP53 revealed in the cohort, were explored in public databases and cell lines. Results The median overall survival time of 40 patients with TP53 mutations was 24.5 months, while their median progression‐free survival time after CAR‐T was 6.8 months. There were no significant differences in the ORR (objective remission rate, X 2 = 3.0498, p > 0.05) and PFS (after CAR‐T therapy) between the patients with wild‐type and mutated TP53 genes after CAR‐T therapy, while the OS of patients with TP53 mutations was significantly worse (p < 0.01). In patients with TP53 mutations, the performance status (ECOG score) was identified as the most important prognostic factor, while the efficacies of induction and salvage treatments were also correlated with the prognosis. Among molecular indicators, the co‐mutations of Chr‐17 and those located on the exon 5 of the TP53 gene showed a tendency for a worse prognosis. Moreover, the patients with TP53‐DDX3X co‐mutations were identified as a subgroup with an extremely bad prognosis. The expression levels of DDX3X and TP53 were explored in a public database and the cell lines with their co‐mutations, which indicated that inhibiting the DDX3X gene could affect the proliferation of rrDLBCL cells and the expression of TP53. Conclusions This study indicated rrDLBCL patients with TP53 mutations was still the group of poor prognosis in the CAR‐T therapy era. CAR‐T therapy can benefit some TP53mut patients, and the performance status (ECOG) might help predict their prognosis. The study also revealed a subgroup of TP53‐DDX3X co‐mutations in rrDLBCL, which showed a strong clinical significance.


| INTRODUCTION
TP53 mutations, a long-standing research hotspot, have been supported by abundant evidence for their prognostic significance in diffuse large B-cell lymphoma (DLBCL). 1,2 However, the treatment of DLBCL with TP53 mutations (TP53mut-DLBCL) is still an unresolved problem. Based on the strong prognostic significance of these mutations in DLBCL, researchers suggest that it might be necessary to treat TP53mut-DLBCL as a separate subtype. 3,4 In the past, due to the independence of TP53 mutations and the poor prognosis of TP53mut patients, the significance of exploring intragroup heterogeneity was unclear. However, the emergence of novel therapies, such as Venetoclax and chimeric antigen receptor T-cell therapy (CAR-T therapy), has changed this concept. 5 In particular, CAR-T therapy has significantly improved the prognosis of relapsed and refractory DLBCL (rrDLBCL) with an objective remission rate (ORR) of more than 80%. 6,7 Numerous CAR-T therapies have also been launched recently in China with an ORR reaching over 60%. 8,9 Therefore, the current study aimed to evaluate the prognosis of TP53mut patients in the context of CAR-T therapy and to explore the heterogeneity in their cohort and identify the possible risk factors. In addition, an exploratory study was conducted on the TP53-DDX3X (DEAD box protein 3, X-chromosomal) comutation subtype identified in this study.

| Patients
In this targeted sequencing retrospective study, the patients were recruited from a collaborative cohort of Peking University Third Hospital and Beijing Boren Hospital. The original cohort included patients with rrDLBCL and mediastinal large B-cell lymphoma (PMBCL), and the study protocol was reviewed and approved by the ethics committees of the above-mentioned two centers. In this study, the patients were further screened with the specified inclusion and exclusion criteria. The inclusion criteria were as follows: (1) the patients were diagnosed with DLBCL lymphoma at the Department of Pathology, Peking University Third Hospital using newly obtained biopsy specimens; (2) the patients were 18 years or older; and (3) the patients with rrDLBCL confirmed by two investigators (G.F and S.H) by reviewing the medical history and histopathological report. Relapsed DLBCL referred to the relapse condition after achieving complete response (CR) by the initial treatment. Refractory DLBCL was defined as the patients, receiving a standard induction regimen (R-CHOP regimen) for at least four courses, did not achieve the complete clinical remission of lymph nodes and/or involved organs (Partial remission [PR] was not achieved), or new lesions appeared during treatment. The evaluation criteria were based on the 2014 Lugano criteria. 10 The exclusion criteria were as follows: (1) the patients with Tly (transforming lymphoma) and PMBCL; (2) the patients, who failed to complete the induction regimen due to reasons other than disease progression; (3) among the patients with recurrence DLBCL cases, the samples obtained before recurrence were excluded, and those derived from the biopsy results of recurrence were included; and (4) the samples with double-hit or triple-hit lymphoma were excluded from the TP53mut-related prognosis and survival analyses. The sequencing results of 96 patients were used for statistical analysis of mutation sites, while 83 patients, including 40 patients with TP53 mutations, were screened for prognostic and survival analyses along with their follow-up results.

| METHODS OF TARGETED SEQUENCING
A core panel of 339 genes, which were selected based on their prior implications in the pathogenesis of hematologic diseases, was analyzed in rrDLBCL disease. The complete set of biotinylated long oligonucleotide probes to capture the coding exons of all the included 339 genes was purchased from Roche NimbleGen. The DNA libraries were generated using 225-ng genomic DNA extracted from the frozen tumor tissues. Five 10-μmthick sections per sample were used for the tumor DNA extraction using the QIAamp DNA Mini Kit following the manufacturer's instructions (Qiagen). The probe pool was hybridized to the tumor DNA (250 ng) upstream and benefit some TP53mut patients, and the performance status (ECOG) might help predict their prognosis. The study also revealed a subgroup of TP53-DDX3X comutations in rrDLBCL, which showed a strong clinical significance.

K E Y W O R D S
chimeric antigen receptor T-cell therapy, DDX3X mutation, diffuse large B-cell lymphoma, sequencing, TP53 gene downstream of each gene of interest. The pooled DNA libraries were loaded onto the cBot System for cluster generation followed by 2 × 150 paired-end sequencing using the NextSeq550 sequencer (both from Illumina). The average depth of coverage across the targeted regions was approximately 3000 bp. The paired-end sequencing reads, which were in FASTQ format, were mapped to the human genome (NCBI build 37) using BWA control software with default parameters. The mutational signatures were called using the output (BAM files) of the previous analysis with Vardict and MuTect2 software and stored as the final VCF file.

| Cell lines
The human DLBCL cell lines, including OCI-LY3, SU-DHL-6, OCI-LY7, and WSU-DLCL2 cell lines, were obtained from Meisen-CTCC (Zhejiang Meisen Cell Technology Co., Ltd). After obtaining informed consent, the blood samples were obtained from healthy subjects in EDTA-containing tubes, from which, the peripheral blood mononuclear cells (PBMCs) were isolated. The SU-DHL-6 and WSU-DLCL2 cells were cultured in RPMI 1640 medium, containing 10% FBS, while the OCI-LY3 and OCI-LY7 cells were cultured in RPMI 1640 medium, containing 20% FBS (Gibco). All the cells were cultured at 37°C in a humidified chamber with a 5% CO 2 concentration.

| Cell Counting Kit-8 (CCK-8) assay
Cell proliferation was determined using CCK-8 reagent (Dojindo). After overnight incubation, the cells were seeded into 96-well plates at a density of 5000 cells/well and treated with various concentrations of RK-33 (1, 2.5, 5, 10, and 20 μM) followed by incubation at 37°C with 5% CO 2 0, 24, and 48 h. Then, the cells were incubated for 2 h with 10 μL of CCK-8 solution. DMSO (dimethyl sulfoxide) was used as a control treatment. The optical density at 450 nm (OD450) values of each well was measured using a microplate reader.

| Western blotting
Total cell protein was extracted using RIPA lysis buffer containing phosphatase inhibitor and protease inhibitor. Protein samples at an equal loading quantity were separated by 10% SDS-PAGE before transferring onto nitrocellulose membranes for Western blotting. Transferred membranes were individually blocked in TBST containing 5% skimmed milk, followed by overnight incubation at 4°C with primary antibodies of p53 and βactin (Abcam); Immunodecorated membranes were washed with TBST buffer, followed by a 1-hour incubation step at room temperature with anti-mouse secondary antibodies and a TBST-washing step afterward before fluorescent signal analysis using Odyssey infrared imaging system (LI-COR).

| Statistical analyses
Using the R language survival package (version 4.2.0), the Log-Rank test and COX regression analysis were performed for the survival and multivariate analyses, respectively. Chi-squared test and Fisher's exact test were used for the comparison of clinical information between the different subtypes, and a t-test was used to compare the PCR (polymerase chain reaction) results using the R package. The visualization of relevant results was performed using the R package or GraphPad Prism 8.0.

| Incidence
Among the total 96 patients, 46 patients carried TP53 mutations, including 45 predicted deleterious TP53 mutations (Combined Annotation Dependent Depletion [CADD] scores 11 >10; 38 mutations had CADD scores >20). The complete follow-up data were available for 85 of the total patients, as listed in Table 1, and the high-frequency mutations in these patients are shown as a heatmap in Figure 1A. The main co-mutations along with TP53 mutations were present in the following genes: lysine-specific methyltransferase 2D (KMT2D) ( Analyzing the types of TP53 mutations indicated that there were four deletion mutations, one insertion mutation, and 34 missense mutations in the exons 5 to 10. In addition, there were two nonsense mutations and three splice site mutations. Most of the mutations (40/45 88.9%) were present in the DNA-binding domain of TP53; their specific location distribution is shown in Figure 1B. Among the 34 missense mutations, 15 mutations were transversions (Tv), while 19 mutations were transitions (Ti) (Data S2). The distribution of Ti/Tv types in the samples is presented in Figure 1C. Among these missense mutations, GC>AT (Ti), GC>TA (Tv), AT>GC (Ti), AT>CG (Tv), GC>CG (Tv), and AT>TA (Tv) were observed in 50%, 14.6%, 14.6%, 12.5%, 4.2%, and 4.2% of the samples, respectively. This distribution was similar to those observed in previous studies. Among all the 45 patients with TP53 mutations, only seven patients had two different TP53 mutations (Data S2).
Most of the mutations in the coding sequence of the TP53 gene identified in the rrDLBCL patients in the current study had already been recorded in the UMD library 12 and other previous TP53 mutation databases, 13 except for a few structural mutations, such as p.R335_ E336insG, p.G245_M246delinsV, p.H233Lfs*14, and p.S149Yfs*20, etc. All these missense mutations had been predicted as harmful using numerous prediction tools, such as PolyPhen-2 (Polymorphism Phenotyping v2), SIFT (Sorting Intolerant From Tolerant), Mut_ass (Mutationassessor), and PROVEAN (Protein Variation Effect Analyzer) (Data S2). In the current study, the CADD tool was used to screen the mutations with scores ≥10 for subsequent analysis. The amino acid positions 273 (three cases of p.R273H and one case of p.R273C) and 306 (three cases of p.R306X) were the most common mutation sites, which have also been previously identified as mutation hotspots. 14 Other mutations, such as p.Y236D, p.T211P, and p.N131Y, had not been reported in lymphomas and were predicted as harmful using the above tools (Table S1). As its main co-mutation, DDX3X showed no significant hot spot in the cohort, and the mutation sites of nine patients were different (p.T450S, p.T384A, p.K387X, and p.R528H, see Table S2).

TP53mut patients
The prognostic significance of TP53 mutations for the entire rrDLBCL cohort (n = 83) was evaluated. Among the patients included for survival analysis, the median follow-up time was 3.54 years, and the 2-year survival   Figure 2A). In the multivariate analysis, including included basic clinical variables, the TP53 mutations (HR = 2.11, p < 0.01) and good performance status (ECOG score <2, HR = 0.37, p < 0.01) were independent prognostic factors, and the global p-value of the model was less than 0.01 ( Figure 2B). Only the TP53 mutations (Data S2) showed significant prognostic significance (p < 0.01) in multivariate analysis among the high-frequency mutations (>15%). The clinical information of the 40 patients with TP53 mutations is listed in Table 2.
Among the two major co-mutations, the patients with DDX3Xmut and TP53mut co-mutations showed a very poor prognosis (6/40, 15%, 5 of GCB, and 1 of Non-GCB), which was significantly different from that of the patients with TP53mut-DDX3Xwt (p < 0.01, Figure 3A). On the  Figure 3B). Similarly, the other high-frequency co-mutations, such as KMT2D, CREBBP, and MYD88 mutations, did not show significant prognostic significance. Among the patients with TP53 mutations, about one-third of the patients had co-mutations located on chromosome 17, including CD79B (cluster of differentiation 79B), GNA13 (G protein subunit alpha 13), and STAT3 (signal transducer and activator of transcription 3) genes. These patients showed a worse prognosis (p = 0.054, Figure 3C). The patients with mutations located in the exon 5 of the TP53 gene also showed a worse prognosis as compared to those with mutations in other regions (p = 0.056, Figure 3D).
Among the main clinical indicators, the performance status (ECOG score) (p < 0.05, Figure 4A), the efficacy of induction therapy (p < 0.05, Figure 4B), and the efficacy of salvage therapy (p < 0.05) could significantly predict the prognosis of patients with TP53 mutations. The COX regression analysis of the performance status (ECOG), induction therapy efficacy, and DDX3X and Chr17 co-mutations also showed that these factors could still significantly predict the prognosis of patients with TP53 mutations (ECOG score <2, HR = 0.41, p < 0.05, Figure S1).

| Efficacy of CAR-T therapy in patients with TP53 mutations
In this study, 65 patients, receiving CAR-T therapy, were followed up. For these patients, the TP53 mutations (p < 0.01), DDX3X mutations (p < 0.001), performance status (p < 0.01), and salvage treatment effect (p < 0.01) had significant prognostic significance. Among the CAR-T therapy-receiving patients, there were only four patients with DDX3X mutations, and no one among them achieved PR, and their median survival time was only 12.5 months. The multivariate analysis of CAR-T therapy-receiving patients also showed that the TP53 mutations (HR = 1.9, p < 0.05), performance status (ECOG score <2, HR = 0.4, p < 0.01), and salvage treatment effect (HR = 1.8, p < 0.05) were still the most important predictors of OS in the CAR-T therapy-receiving patients ( Figure 5A). These variables were also the best predictors of achieving CR by CAR-T therapy.
A total of 80% (32/40) of patients with TP53 mutations received CAR-T therapy. The median OS of patients with TP53 mutations was 24.5 months, and the median PFS after CAR-T therapy was 6.8 months. Although CAR-T therapy significantly improved the prognosis of patients with TP53 mutations as compared to those who did not receive CAR-T therapy (p < 0.01, Figure 5B), the overall ORR was only 56.2% (18/32), which was slightly lower than the ORR for the entire CAR-T therapy-receiving cohort (45/65, 69.2%). The ORR of the patients with wild-type TP53 was 81.8% (27/33) as compared to those receiving CAR-T therapy. There was no significant difference in ORR (χ 2 = 3.0498, p > 0.05, Figure 5C) and PFS (after CAR-T therapy, p > 0.1) between the patients with wildtype and mutated TP53 gene. However, there was still a significant difference in the OS times of the patients between the two groups, showing a worse OS of the patients with TP53 mutations in the CAR-T therapy-receiving cohort (p < 0.01, Figure 5D). Seven patients with TP53 mutations achieved CR (7/32, 21.9%) with a median OS time of 35 months and a median PFS time of 30 months after CAR-T therapy. Nine patients, who did not achieve PR in salvage treatment, finally achieved remission through CAR-T therapy and accounted for 47.4% (9/19) of the patients, showing salvage treatment failure (SD or PD).

| Recurrent TP53-DDX3X comutations and their expression levels in cell lines
The possible adverse prognostic effects of DDX3X mutations have already been discussed in the previous sections. In this study cohort, there were only nine patients with DDX3X mutations, and their median survival time was only 15.27 months. Six of the nine patients had TP53 co-mutations and could not get PR from salvage therapy, including CAR-T therapy. The DDX3X mutations were identified in the Sanger sequencing results provided by the Cell Model Passport: 15 Figure S2b,c). The expression levels of DDX3X were higher €n the two DDX3Xmutant cell lines as compared to those in OCI-LY3 (DDX3Xwt-TP53wt); however, the expression level of DDX3X in SU-DHL-6 cell line (TP53mut-DDX3Xwt) was also high ( Figure 6A,B). The CCK-8 analysis showed that the DDX3X inhibitor (RK-33 16,17 ) could inhibit the proliferation of DDX3X-mutant cell lines of OCI-LY7 and WSU-DLCL-2, especially that of OCI-LY7 cell line ( Figure 6C,D). RT-PCR analysis showed that the expression level of TP53 in the OCI-LY7 cell line decreased significantly after 48 h of RK-33 treatment (p < 0.001); However, this phenomenon was not observed in the WSU-DLCL-2 cell line ( Figure 6E,F). In the Western blot experiment, OCI-LY7 also responded more obviously to RK-33, and p53 showed a downward trend after adding inhibitor for 24-48 h ( Figure 6G).

| DISCUSSION
In this study, the heterogeneity of prognosis in the rrDLBCL patient with TP53 mutations in the context of CAR-T therapy was explored. The follow-up data showed that the CAR-T therapy had a considerable effect on the rrDLBCL with TP53 mutations, about half of the patients, who did not achieve remission in salvage treatment, benefited from CAR-T therapy. The CAR-T thereby significantly improved the prognosis of patients with CR; however, these patients accounted for only about 1/5 of all the patients with TP53 mutations. The OS of patients with TP53 mutations in the CAR-T therapy cohort was poor, and their ORR was also lower than that of the patients with wild-type TP53. This study indicated that CAR-T therapy might not be the final solution for patients with TP53; however, it might act as a "relay statio" with great potential. Only eight patients in this cohort did not receive CAR-T therapy, which might bring bias and cause difficulties in interpretation. Therefore, the results of a subgroup study based on molecular biology in a large-scale prospective study should be performed in the future. The results showed that performance status was the most important prognostic factor among the clinical indicators, and the efficacy of front-line treatment showed prognostic significance as well. The analysis of clinical variables suggested that the impact of performance status on the prognosis of patients with TP53 mutations before CAR-T therapy should be fully considered. Among the molecular biological markers, the status of DDX3X and Chr17 co-mutations might affect the prognosis of rrDLBCL patients with TP53 mutations. The Chr17 comutations included the mutations in CD79B, GNA13, and STAT3 genes, which have been proven to be correlated with the prognosis of DLBCL in a previous study. 18 It was speculated that CD79B might play a certain role in prognosis, thereby affecting the prognosis analysis results of the Chr17 group. First, the CD79B mutations were the most common (5/14, 35.7%) mutations among Chr17 comutations. Second, the patients with CD79B and TP53 co-mutations showed a tendency to relapse or progress earlier after the CAR-T therapy (PFS after CAR-T therapy, p-value of K-M survival analysis was <0.05). The different TP53 mutation sites were generally discrete in the cohort, and it was difficult to analyze the prognostic significance of each site. Only the mutations on exon 5 showed a tendency of poor prognosis; however, the tendency was statistically insignificant (p = 0.056).
DDX3X is a ubiquitously expressed RNA helicase, which is involved in the multiple stages of RNA biogenesis. It has about 34%, and 15% incidence in Burkitt lymphoma and adult Burkitt lymphoma, respectively. 19 A study showed that the loss-of-function mutations in DDX3X were also enriched in the MYC-translocated DLBCL and revealed functional correlations between mutant DDX3X and MYC. 20 Researchers suggested that this was relatively common in the so-called molecular highgrade B-Cell Lymphoma among DLBCL patients. 21 As sex chromosome-specific genes, the functional difference between DDX3X and DDX3Y might explain the gender differences in the incidence rate of Burkitt lymphoma. 22 However, in this study, the proportion of men and women with DDX3X mutations was almost similar (5/4, 1.25: 1). All the patients with DDX3X mutations showed poor prognosis; only 2 of them had MYC split/rearrangement. Since the patients with double-hit and triple-hit lymphomas were excluded from this study, it was speculated that identifying the DDX3X mutations at the time of relapse might be valuable for evaluating the prognosis of rrDLBCL patients. The prognostic potential of DDX3X mutations has also been emphasized in previous studies; 23 however, the studies conducted on the response of patients with DDX3X mutations to CAR-T therapy are still lacking. The current study suggested that DDX3X had a prominent prognostic significance in rrDLBCL, and the patients with DDX3X mutations might not benefit from salvage treatments, including CAR-T therapy. It is worth mentioning that the frequency of co-mutation of TP53-DDX3X reported in this study reached 13.3% in the original cohort (n = 96, and 15% in the cohort for survival analysis), which might be overestimated. Because the prognosis of patients in the cohort was generally poor, and only two centers participated in the study, which could bring deviation.
The expression levels of DDX3X in different tumors are inconsistent. 24 Even in the published studies on DLBCL, there are contrasting results about the function of DDX3X, indicating its complexity. Kizhakeyil et al.
established the DDX3X-mutant and DDX3X-knockout cell lines to observe their effects on the proliferation, apoptosis, and other phenotypes of the cell lines. The results showed that the overexpression of wild-type DDX3X in the U2932 and HuT78 cells significantly decreased their proliferation. 23 Lacroix et al revealed that DDX3X was required for the lymphoid differentiation and MYC-driven lymphomagenesis, which indicated that inhibiting the expression of the DDX3X gene could be a treatment strategy for the MYC-driven B-cell lymphoma. 25 The current study analyzed the expression levels of DDX3X in different DLBCL cell lines using RT-PCR. As compared to the cell lines with the wild-type DDX3X gene (OCI-LY3/SU-DHL-6), those with the mutated DDX3X gene (OCI-LY7/WSU-DLCL-2) showed an upregulated expression level of DDX3X. On the contrary, the expression level of DDX3X in the cell lines with mutated TP53 gene (SU-DHL-6/OCI-LY7/WSU-DLCL2) was upregulated, suggesting that the expression of TP53 might affect that of DDX3X. Therefore, more samples from real patients are still needed to observe the specific phenotypic effects of DDX3X mutations in DLBCL. DDX3X can bind to both the wild-type and mutant TP53 proteins in tumors. When DNA damage occurs, DDX3X can still bind to the wild-type TP53 and stabilize its protein level, thereby promoting TP53-mediated apoptosis. 24 At the same time, a study suggested that TP53 might directly regulate the transcription of the DDX3X gene in lung cancer, and the relevant research results also supported the existence of the TP53-DDX3X pathway. 26 In this study, both the cell model database and the clinical cohort identified the TP53-DDX3X co-mutations as a recurrent phenomenon. The CCK-8 assay indicated that the inhibitor of DDX3X (RK-33) decreased the proliferation of DLBCL cell lines having TP53-DDX3X co-mutations. However, the response to RK-33 at the expression level was only observed in the OCI-LY7 cell line, showing the downregulation of the TP53 expression level, which could be purely due to off-target toxicity of RK-33. At present, the characteristics of DDX3X in lymphoma are still very mysterious and have not been fully revealed. The current experimental results are not enough to answer questions about it. Our research team will conduct more studies to evaluate the functional correlations between the two mutations in the future.

| CONCLUSIONS
Overall, this study investigated the clinical characteristics of rrDLBCL patients with TP53 mutations in the context of CAR-T therapy, which was still the group of patients with poor prognosis in the CAR-T therapy era. CAR-T therapy can benefit some patients with TP53 mutations, and the performance status (ECOG), an important clinical indicator, might help predict their prognosis. At the same time, this study also revealed a subgroup of TP53-DDX3X co-mutations in rrDLBCL, which showed a strong clinical significance. Although these co-mutations were uncommon in DLBCL, their response to treatment was extremely poor, and even CAR-T therapy could not reverse their prognosis. Therefore, it might be very necessary to detect DDX3X mutations when the disease relapses/progresses. Furthermore, as a potential target for TP53 regulation, DDX3X might have a promising potential in cancer research, and its studies might bring benefits to DLBCL patients.