TP53 mutations have been observed in diffuse large B-cell lymphoma (DLBCL), with a mean frequency of ~20%. Studies on TP53 mutations as prognostic markers have historically been controversial, and the results have not been consistent across different studies on DLBCL. Considering the complex pathophysiological mechanisms involved in DLBCL, we wondered whether the interaction of TP53 with other genetic variants could further promote the development of DLBCL, and thus be more prognostically predictive. Moreover, whether the genetic interactions between TP53 and other oncogenic mutations could shape the discrepant immune landscape in DLBCL remains unknown, as these genetic alterations usually drive the malignant phenotype and directly or indirectly affect the tumor microenvironment (TME) and support tumor survival.

In this study, we performed a comprehensive analysis of the genomic characteristics of TP53 through high-throughput sequencing in patients with de novo DLBCL. Patients' characteristics are reported in Table S1. Detailed methods are provided in the Supplementary Material. A total of 227 significantly mutated genes were identified (Table S2), of which TP53 was the second most frequently mutated gene, with a rate of 30% (53 of 176) and 62 sequence variants detected. Among these variants, 74% (n = 46/62) were missense mutations, and the remaining were inactivating frameshift indels (n = 7), nonsense mutations (n = 3), coding sequencing indels (n = 4), and splicing mutations (n = 2). Mutation patterns and distributions are shown in Figure S1 and Table S3. Importantly, most mutations (56/64, 87.5%) occurred in exons 5–8, which encoded the DNA-binding domain (DBD) region of TP53 (Figure S1C,E). Codons 175, 273, and 248 of the p53 protein had the highest mutation frequency, which are also the hot spots of TP53 mutation found in most human cancers (Figure S1D). Given that the DBD of TP53 is the functional central core domain and mutations in this region potentially have a strong impact on TP53 function, we mainly focused on mutations in this region. Patients were divided into TP53-MUT and TP53-WT groups according to TP53 mutation status in the DBD region. There were no differences in the number of small deletions/insertions and single nucleotide variants (SNVs) between the TP53-MUT and TP53-WT groups (Figure S2). Moreover, the tumor mutation burden (TMB) was similar between the two groups (Figure S2E). Clinical relevance between TP53 mutation and clinicopathological characteristics, such as age, sex, B symptoms, stage, number of extranodal sites, performance status, LDH level, and International Prognostic Index was not observed (Table S4). Note that TP53 mutations significantly enriched in the GCB subtype (p = .033, Table S4). Among the 176 patients, 155 patients having the complete follow-up data were enrolled in the survival analysis. Overall, patients with TP53-MUT tended to have inferior overall survival compared with patients with TP53-WT (median: 92.3 versus 110.8 months, respectively, p = .17, Figure 1A), but it did not reach statistical significance. A subgroup analysis showed that the potential predictive value of TP53 mutations was mainly attributed to the GCB subtype (Figure S3).

We next recognized the genomic variants that co-occur or are mutually exclusive with TP53. We observed that DDX3X, MYLK2, and FUT6 mutations co-occurred with TP53 mutations, and CD58 mutations were mutually exclusive with TP53 mutations (Figure 1B and Table S5). Specifically, patients were divided into four groups based on the mutation status of these genes. No significant difference was observed in survival among the three groups according to the combination of co-occurring mutation genes with TP53 (p = .37 for DDX3X; p = .11 for MYLK2; p = .54 for FUT6; Figure S4). However, we found that the combination of TP53 and CD58 mutations could significantly distinguish the prognosis of patients with DLBCL (p = .033, Figure 1C). Patients with both wild-type TP53 and CD58 had a better prognosis than patients with either of the two mutually exclusive modes of CD58 and TP53 mutations. Unexpectedly, patients with TP53 wild-type and CD58 mutations (TP53WT&CD58MUT) had worse survival than those with TP53 mutations and CD58 wild-type (TP53MUT&CD58WT). Because TP53 and CD58 mutations were mutually exclusive, only one patient harbored both TP53 and CD58 mutations, and the patient still alive at the last follow-up. The predictive value of TP53WT&CD58MUT group was also observed in patients with GCB-DLBCL (Figure S5). Moreover, the relationship of a mutually exclusive mutant between CD58 and TP53 and the prognostic significance of this interaction were validated using publicly available data from 1001 patients with DLBCL from the Duke University's cohort¹ (Figure S6).

We then explored whether the cooperation of the mutually exclusive mutations between TP53 and CD58 may profoundly influence the microenvironment in DLBCL. We found that the overall TMB was significantly higher in the TP53WT&CD58MUT group than in the TP53MUT&CD58WT group (p = .0177, Figure 1D), while there was no difference in TMB when dividing patients only according to TP53 mutation status (p = .5348, Figure S2E). In addition, the ESTIMATE immune scores in the TP53WT&CD58MUT group were significantly higher than that in the TP53MUT&CD58WT groups (p = .0047) (Figure 1D). Moreover, the exhausted T cell, macrophage cell, NK cell, and Th1 cell enriched in the TP53WT&CD58MUT group (Figure 1E). The difference between the two groups was mainly due to the combined influence of the mutation pattern “TP53WT&CD58MUT,” but rather only affected by the CD58 mutations, given the immune cell infiltration was similar when dividing patients just according to CD58 mutation status (Figure S7). Furthermore, the co-inhibitory receptors such as PD-1, TIM3, and LAG3 were preferentially expressed in the TP53WT&CD58MUT group (Figure 1F and Table S6). However, there was no difference in the expression of co-stimulatory molecules (Table S6). Inhibitory immunomodulators were also significantly upregulated in the TP53WT&CD58MUT group when comparing with the TP53WT&CD58WT group (Figure S8), suggesting the unique immune phenotype in the TP53WT&CD58MUT group. The findings that high immune scores and abundant infiltrating exhausted T cells in the TP53WT&CD58MUT group were validated in an independent external cohort from the REMoDL-B trail (N = 400)² (Figure S9 and Table S7).

Finally, we investigated the differentially biological pathways between the TP53WT&CD58MUT and the TP53MUT&CD58WT groups. Five hundred differentially expressed genes were identified with a false discovery rate less than 0.05 and |log2foldchange| > 1. One hundred genes were significantly upregulated and 400 genes were significantly downregulated in the TP53WT&CD58MUT group (Figure 1G). Figure S10 presented the enriched gene ontology terms in the TP53WT&CD58MUT group, including cytokine and chemokine production, binding and activity, and interferon-γ pathways. GSEA showed significantly activated interferon-α and interferon-γ responses and IL-6/JAK/STAT3 signaling in the TP53WT&CD58MUT group (Figure 1H). The immune landscape of patients with TP53WT&CD58MUT harbored was summarized and conceptualized in Figure 1I.

The value of TP53 gene alterations in predicting survival in DLBCL remains controversial, even in the era of sequencing. In the L.M. Staudt' study, four prominent genetic subtypes were identified based on the molecular classifications. Notably, TP53 was not significantly enriched in one of these subtypes, although TP53 was the much frequently mutated gene (25.2%). On the basis of this classification, L.M. Staudt and colleague further distinguished ST2, A53, and mixed subtypes, and the survival of A53, characterized by inactivation of TP53, was intermediate in this model.³ The C2 molecular subtype identified in Chapuy's study corresponding to A53 also showed a tendency for a poor prognosis.⁴ A subsequent study demonstrated that the impact of TP53 mutations on survival was relied on the genetic context of the lymphoma, conferring no effect in the SOCS1/SGK1 clusters and NOTCH2 subtype and inferior prognosis in the MYD88 subtype.⁵ These results demonstrated that TP53 alterations have limited ability to identify a subset of patients at high risk. DLBCL is highly molecular heterogeneity and there are still a proportion of patients who could not be accurately classified according to the existing molecular classifications. In this study, we found that patients with TP53WT&CD58MUT had the worst outcomes. Interestingly, this subtype of patients harbored enhanced immune escape capacity, giving the abundant infiltration of exhausted T cells and multiple upregulated inhibitory immunomodulatory molecules. Persistent interferon signaling could augment the expression of T-cell inhibitory immune checkpoints such as PD-1, TIM-3, and LAG-3 through JAK/STAT pathway.⁶ We found that interferon responses and JAK/STAT pathway enriched in the TP53WT&CD58MUT group. Moreover, inhibitory immunomodulatory molecules were preferentially expressed in this group. It suggests that interferon/JAK/STAT pathway-mediated up-regulated expression of inhibitory immunomodulatory molecules could be one potential mechanism through which the inactivation of CD58 facilitated immune evasion and accelerated tumor growth in DLBCL. Despite many of such immune dysregulations had an adverse prognosis; they provided new opportunities for anti-tumor immunotherapy of the subset of DLBCL patients with TP53WT&CD58MUT. Historically, the response rate to anti-PD-1/PD-L1 therapy in unselected DLBCL patients was generally low. Consequently, patients with TP53WT&CD58MUT may be optimal candidates for novel immunotherapy in clinical trials.

In conclusion, our results suggest that TP53 mutation alone is insufficient to effectively differentiate the risk of DLBCL. The mutually exclusive patterns between TP53 and CD58 mutations accurately stratified patients with DLBCL to permit the optional immunotherapy.

在弥漫性大 B 细胞淋巴瘤 (DLBCL) 中观察到TP53突变，平均频率约为 20%。关于TP53突变作为预后标志物的研究在历史上一直存在争议，并且在 DLBCL 的不同研究中结果并不一致。考虑到 DLBCL 涉及的复杂病理生理机制，我们想知道TP53与其他遗传变异的相互作用是否可以进一步促进 DLBCL 的发展，从而更具预后预测性。此外， TP53之间是否存在遗传相互作用和其他致癌突变可能会塑造 DLBCL 中差异性的免疫格局仍然未知，因为这些遗传改变通常会驱动恶性表型并直接或间接影响肿瘤微环境 (TME) 并支持肿瘤存活。

在本研究中，我们通过高通量测序对新发 DLBCL 患者的TP53基因组特征进行了全面分析。患者的特征见表 S1。补充材料中提供了详细的方法。共鉴定出 227 个显着突变基因（表 S2），其中TP53是第二个最常见的突变基因，发生率为 30%（176 个中的 53 个），检测到 62 个序列变体。在这些变异中，74% ( n  = 46/62) 是错义突变，其余是失活的移码插入缺失 ( n  = 7)、无义突变 ( n  = 3)、编码测序插入缺失 ( n  = 4) 和剪接突变(n  = 2)。突变模式和分布如图 S1 和表 S3 所示。重要的是，大多数突变（56/64, 87.5%）发生在外显子 5-8 中，其编码TP53的 DNA 结合域（DBD）区域（图 S1C，E）。p53 蛋白的密码子 175、273 和 248 的突变频率最高，这也是在大多数人类癌症中发现的 TP53 突变的热点（图 S1D）。鉴于TP53的 DBD是功能性中心核心结构域，该区域的突变可能对 TP53 功能产生强烈影响，我们主要关注该区域的突变。根据TP53将患者分为 TP53-MUT 组和 TP53-WT 组DBD 区域的突变状态。TP53-MUT 和 TP53-WT 组之间的小缺失/插入和单核苷酸变异 (SNV) 的数量没有差异（图 S2）。此外，两组之间的肿瘤突变负荷（TMB）相似（图S2E）。未观察到 TP53 突变与临床病理学特征之间的临床相关性，例如年龄、性别、B 症状、分期、结外部位数量、体能状态、LDH 水平和国际预后指数（表 S4）。请注意，TP53突变在 GCB 亚型中显着富集（p = .033，表 S4）。176例患者中，有完整随访资料的155例患者进入生存分析。总体而言，与 TP53-WT 患者相比，TP53-MUT 患者的总生存期往往较差（中位数：分别为 92.3 个月和 110.8 个月，p  = .17，图 1A），但没有达到统计学意义。亚组分析显示TP53突变的潜在预测价值主要归因于 GCB 亚型（图 S3）。

我们接下来认识到与TP53共存或互斥的基因组变异。我们观察到DDX3X、MYLK2和FUT6突变与TP53突变同时发生，而CD58突变与TP53突变相互排斥（图 1B 和表 S5）。具体来说，根据这些基因的突变状态将患者分为四组。根据与TP53共同发生的突变基因的组合，三组之间的生存率没有显着差异（ DDX3X 为p  = .37； MYLK2 为p  = .11；p = .54 对于 FUT6；图 S4)。然而，我们发现TP53和CD58突变的组合可以显着区分 DLBCL 患者的预后（p  = .033，图 1C）。具有野生型TP53和CD58的患者比具有两种相互排斥的CD58和TP53突变模式的患者具有更好的预后。出乎意料的是，TP53野生型和CD58突变（TP53WT&CD58MUT）患者的生存率比TP53突变和CD58野生型（TP53MUT&CD58WT）的患者更差。因为TP53和CD58突变是相互排斥的，只有一名患者同时携带 TP53和CD58突变，并且该患者在最后一次随访时仍然存活。在 GCB-DLBCL 患者中也观察到 TP53WT&CD58MUT 组的预测价值（图 S5）。^{此外，使用来自杜克大学队列1}的 1001 名 DLBCL 患者的公开数据验证了CD58和TP53之间互斥突变体的关系以及这种相互作用的预后意义（图 S6）。

然后，我们探讨了TP53和CD58之间互斥突变的合作是否会深刻影响 DLBCL 的微环境。我们发现 TP53WT&CD58MUT 组的总体 TMB 显着高于 TP53MUT&CD58WT 组（p = .0177，图 1D），而仅根据TP53突变状态 划分患者时，TMB 没有差异（ p  = .5348，图 S2E)。此外，TP53WT&CD58MUT 组的 ESTIMATE 免疫评分显着高于 TP53MUT&CD58WT 组（p = .0047) (图 1D)。此外，耗竭的 T 细胞、巨噬细胞、NK 细胞和 Th1 细胞在 TP53WT&CD58MUT 组中富集（图 1E）。两组之间的差异主要是由于突变模式“TP53WT&CD58MUT”的综合影响，而仅受CD58突变的影响，因为仅根据CD58划分患者时免疫细胞浸润相似突变状态（图 S7）。此外，PD-1、TIM3 和 LAG3 等共抑制受体在 TP53WT&CD58MUT 组中优先表达（图 1F 和表 S6）。然而，共刺激分子的表达没有差异（表S6）。与 TP53WT&CD58WT 组相比，TP53WT&CD58MUT 组的抑制性免疫调节剂也显着上调（图 S8），表明 TP53WT&CD58MUT 组中的独特免疫表型。TP53WT&CD58MUT 组的高免疫评分和大量浸润耗尽的 T 细胞的发现在来自 REMoDL-B 试验的独立外部队列 ( N  = 400) ²中得到验证（图 S9 和表 S7）。

最后，我们研究了 TP53WT&CD58MUT 和 TP53MUT&CD58WT 组之间的差异生物学途径。鉴定出 500 个差异表达基因，错误发现率小于 0.05，|log2foldchange| > 1. TP53WT&CD58MUT组100个基因显着上调，400个基因显着下调（图1G）。图 S10 展示了 TP53WT&CD58MUT 组中丰富的基因本体术语，包括细胞因子和趋化因子的产生、结合和活性以及干扰素-γ 途径。GSEA 在 TP53WT&CD58MUT 组中显示出显着激活的干扰素-α 和干扰素-γ 反应以及 IL-6/JAK/STAT3 信号传导（图 1H）。图 1I 总结并概念化了携带 TP53WT&CD58MUT 患者的免疫状况。

TP53基因改变在预测 DLBCL 存活率方面的价值仍然存在争议，即使在测序时代也是如此。在 LM Staudt 的研究中，根据分子分类确定了四种主要的遗传亚型。值得注意的是，TP53在其中一种亚型中没有显着富集，尽管TP53是突变频率最高的基因（25.2%）。在此分类的基础上，LM Staudt 及其同事进一步区分了 ST2、A53 和混合亚型，并且以TP53失活为特征的 A53 的存活率在该模型中处于中等水平。³ Chapuy 研究中确定的与 A53 相对应的 C2 分子亚型也显示出预后不良的趋势。⁴随后的一项研究表明，TP53突变对生存的影响取决于淋巴瘤的遗传背景，对 SOCS1/SGK1 簇和 NOTCH2 亚型没有影响，而 MYD88 亚型的预后较差。⁵这些结果表明，TP53改变识别高风险患者子集的能力有限。DLBCL具有高度的分子异质性，仍有一部分患者无法按照现有的分子分类进行准确分类。在这项研究中，我们发现 TP53WT&CD58MUT 患者的预后最差。有趣的是，这种亚型患者具有增强的免疫逃逸能力，使耗尽的 T 细胞和多种上调的抑制性免疫调节分子大量浸润。持续的干扰素信号可以通过 JAK/STAT 通路增强 T 细胞抑制性免疫检查点如 PD-1、TIM-3 和 LAG-3 的表达。⁶我们发现干扰素反应和 JAK/STAT 通路在 TP53WT&CD58MUT 组中富集。此外，抑制性免疫调节分子在该组中优先表达。这表明干扰素/JAK/STAT通路介导的抑制性免疫调节分子的上调表达可能是CD58失活促进免疫逃避和加速DLBCL肿瘤生长的一种潜在机制。尽管许多此类免疫失调都有不良预后；他们为患有 TP53WT 和 CD58MUT 的 DLBCL 患者亚群的抗肿瘤免疫治疗提供了新的机会。从历史上看，未经选择的 DLBCL 患者对抗 PD-1/PD-L1 治疗的反应率普遍较低。因此，TP53WT&CD58MUT 患者可能是临床试验中新型免疫疗法的最佳候选者。

总之，我们的结果表明仅TP53突变不足以有效区分 DLBCL 的风险。TP53和CD58突变之间的相互排斥模式准确地对 DLBCL 患者进行分层，以允许可选的免疫治疗。