The anti-apoptotic protein B-cell lymphoma 2 (BCL-2) is a key regulator of the intrinsic apoptotic pathway,¹ and survival dependence on BCL-2 has been demonstrated in chronic lymphocytic leukemia (CLL).² Venetoclax is a highly selective, potent, orally bioavailable BCL-2 inhibitor³ that has demonstrated efficacy as monotherapy for patients with previously treated CLL in four clinical studies.^4-8 Because venetoclax is the first approved BH3 mimetic and an approved treatment option for CLL, strong interest exists in learning about potential mechanisms of resistance.

To investigate resistance to venetoclax in vitro, we performed a genome-wide CRISPR knockout screen. In the absence of a suitable CLL cell line, it was performed in SU-DHL4, a diffuse large B-cell lymphoma cell line. Upon venetoclax treatment, cells with potential genetic perturbations leading to drug resistance were enriched in the surviving population. Consistent with previously described functional genomic screens across multiple disease models,⁹ genes encoding members of the pro-apoptotic machinery, including BAX, BCL2L11 (BIM), and PMAIP1 (NOXA), were among the strongest hits associated with venetoclax resistance (Figure 1A). Loss of ID3 or NFKBIA was also associated with decreased venetoclax sensitivity, suggesting multiple possible mechanisms of resistance.¹⁰

To compare these results with potential mechanisms of resistance in patients treated with venetoclax, we performed whole-exome and targeted sequencing of DNA extracted from CD19+ B cells collected before treatment initiation and/or upon treatment discontinuation from 43 patients with relapsed/refractory CLL or small lymphocytic lymphoma. All patients were enrolled on a venetoclax monotherapy study (NCT01328626,⁵ n = 18 or NCT01889186,⁶ n = 25) and received treatment until disease progression or unacceptable toxicity. Selection for these analyses was based on sample availability. Baseline characteristics are shown in Table S1.

Median time on therapy was 20.1 months (range, 1.2–60.3) with a median duration follow-up of 61.3 months (95% CI, 48.6–73.4). Venetoclax was nearly as efficacious in this cohort as it was in the overall cohort^{5, 6} (and pooled data of 347 patients across three venetoclax studies¹¹), with an overall response rate (ORR) of 69.8%, a complete remission/complete remission with incomplete hematopoietic recovery (CR/CRi) rate of 11.6%, and median progression-free survival (PFS) of 21.9 months (Table S2). Although minimal residual disease (MRD) evaluation was not required for all patients, of the 12 patients with ≥ 1 MRD assessment, five achieved undetectable MRD at 10⁻⁴ in the peripheral blood, whereas seven had detectable MRD. Post-treatment specimens were collected at study discontinuation due to CLL progression (n = 29), Richter's transformation (n = 8), or other reasons (n = 6) (Table S3).

To investigate changes in clonal architecture after venetoclax treatment in CLL, we focused on three of the most mutated genes in CLL: NOTCH1, SF3B1, and TP53 (Figure 1B). Although some gains and losses of mutations in these genes were observed, most mutations present before treatment initiation were detected in post-venetoclax specimens. Gains and losses of NOTCH1 mutations were detected in post-treatment samples of four and two patients, respectively. One patient's CLL gained an SF3B1 mutation at progression, whereas in three patients, SF3B1-mutated clones present at commencement were no longer detected. We did not observe gains of TP53 mutations. For three patients, TP53-mutant clones present at commencement were not detected subsequently, supporting the notion that venetoclax-based treatments are efficacious in CLL patients with these mutations. Two patients gained BRAF V600E mutations, consistent with a recent report.¹² Overall, based on the markers evaluated, one-third (n = 14/43) of patients exhibited changes in clonal architecture during treatment.

Previous studies have identified BCL2 mutations associated with relapse on venetoclax.^13-15 In our cohort, no BCL2 mutations were found at study entry, but BCL2 mutations were acquired in CLL cells from 15 of 43 patients (35%), 14 of whom experienced CLL progression (Figure 1C). Of 29 patients with CLL progression, BCL2 mutations were identified in 10 patients who were on treatment for ≥24 months. Overall, 30 distinct BCL2 mutations were identified, with multiple mutations identified in nine patients (Figure 1D,E). The most frequent mutations of any description observed were at residue D103 (11 occasions) and G101 (8 occasions). Additionally, mutations were identified at F104 (one occasion), A113 (three occasions), and V156 (four occasions). A 4–amino acid (RRYR) insertion at residue 110 was identified in samples from three patients. Except for the insertion, all BCL2 alterations were missense mutations. In agreement with previous findings,^{13, 14} most BCL2 mutations were found in minor clones (Figure 1F). Only 1 BCL2 mutation (D103Y) had a variant allele frequency (VAF) > 20%, whereas most SF3B1 and TP53 mutations were in larger clones (VAF > 20%).

Mutations in apoptosis regulators other than BCL2 were identified, including PMAIP1 and BAX, two of the strongest hits identified in the CRISPR screen. All PMAIP1 mutations were loss-of-function, including three nonsense mutations and one mutation leading to loss of translation-initiating methionine, M1. One BAX mutation was nonsense; the other mutation replaced leucine at amino acid 181 with histidine. Although both BAX mutations had a VAF < 10%, three of four PMAIP1 mutations had a VAF > 20%, suggesting that PMAIP1 mutations could perhaps be a key driver of venetoclax resistance, although further study is needed. In addition to a BCL2 mutation, one patient also had a mutation in the pro-apoptotic gene BAD (G111R) with a VAF >25%; the functional relevance of this is unknown. Three PMAIP1 mutant patients also had BCL2 mutations present at the end of treatment, albeit at different allele frequencies, supporting the idea of multiple potential mechanisms of resistance within an individual patient.

Although initial clinical response to venetoclax-based therapies are similar in CLL patients regardless of TP53 disruption via del(17p) and/or TP53 mutation, patients with del(17)p/TP53 mutation had shorter response duration.^{11, 16, 17} In our cohort, 78% of patients (n = 29/37) with CLL progression or Richter's transformation had a del(17p)/TP53 mutation at baseline. Additionally, 87% of patients (n = 13/15) with BCL2 mutations had evidence of TP53 disruption.

To provide a deeper understanding of the characteristics of the patients available for these exploratory analyses, the outcomes are reported. Among the eight patients who progressed with Richter's transformation, the ORR was 25.0% with a median PFS of 5.3 months (Table S2). While two patients achieved partial remission (PR), most patients were refractory to venetoclax prior to diagnosis of Richter's transformation. One patient who acquired a BCL2 mutation achieved a PR but relapsed after 6 months. In contrast, the ORR for patients who eventually relapsed due to CLL progression was 86.2%, with seven achieving a deep remission (CR/CRi/nodular partial remission) and 18 achieving a PR. Median duration of response (DOR) was 23.7 months and median PFS was 21.9 months. Among patients who relapsed with CLL progression, median PFS for patients with a BCL2 mutation at relapse was 27.3 months but for patients without a BCL2 mutation median PFS was 10.8 months. Similar results were demonstrated for patients with a mutation in any apoptotic gene; median PFS for patients with versus without an acquired mutation in any apoptotic gene was 27.3 and 10.8 months, respectively. Lastly, for patients who acquired a mutation in BCL2 or other apoptotic genes, the DOR was longer compared with DOR in patients without a mutation (Table S2). Given the observation that acquisition of mutations in apoptotic genes did not portend a shorter response, we evaluated gene expression changes in the pre- and post-progression specimens.

To investigate transcriptional changes that occurred, we performed RNA-seq on 16 pairs of pre- and post-venetoclax samples from patients with CLL progression. Comparison of samples identified changes in expression of over 800 genes (FDR < 0.05). Several genes that play a role in IL-8, Tec kinase, and phospholipase C signaling were upregulated in patients at relapse (Figure 1G). Comparison of pre- and post-treatment samples between patients who were on study for different durations of time did not yield any differentially expressed genes. We were interested in looking at changes in expression of BCL2 family genes because they are known to be able to affect venetoclax resistance in vitro (Figure 1H). A trend of decreased BCL2 expression in post-venetoclax samples was observed, although the differences did not reach statistical significance. Concomitantly, we observed a statistically significant increase in expression of MCL1, BCL2L1 (BCL-X_L), and BCL2A1 (BFL-1), three genes whose protein products can inhibit apoptosis when BCL-2 is effectively targeted by venetoclax.

In addition to previously published mutations in BCL2, we identified novel mutations in PMAIP1, BAX, and BAD in CLL patients who developed resistance to venetoclax monotherapy. This is the first confirmation of the top hits from an in vitro screen with patient samples from clinical trials. Our findings suggest that inactivation of BAX, BAD, or PMAIP1 may be sufficient to render CLL cells resistant to venetoclax treatment, identifying these factors as potential critical regulators of venetoclax response. This is the first report to evaluate mRNA expression differences in this patient population. Observed gene expression changes indicated that tumor cells may upregulate anti-apoptotic BCL-2 proteins other than BCL-2 to avoid apoptosis induction by venetoclax. Overall, our data suggest that resistance to venetoclax monotherapy in CLL patients is multifactorial and highly dependent on mutational and gene expression changes that alter the balance between the pro- and anti-apoptotic machinery. Additional studies are needed to determine the clinical relevance of minor BCL2 mutant clones in venetoclax resistance and whether similar mechanisms of resistance arise in patients treated with fixed-duration venetoclax combination therapies.

抗凋亡蛋白 B 细胞淋巴瘤 2 (BCL-2) 是内在凋亡途径的关键调节因子，¹并且在慢性淋巴细胞白血病 (CLL) 中已证实对 BCL-2 的生存依赖。² Venetoclax 是一种高选择性、强效、口服生物可利用的 BCL-2 抑制剂³，在四项临床研究中已证明其作为单一疗法对既往接受过治疗的 CLL 患者有效。^4-8由于维奈托克是第一个获批的 BH3 模拟物和获批的 CLL 治疗选择，因此人们对了解潜在的耐药机制存在浓厚的兴趣。

为了研究体外对维奈托克的耐药性，我们进行了全基因组 CRISPR 基因敲除筛选。在没有合适的 CLL 细胞系的情况下，它在 SU-DHL4（一种弥漫性大 B 细胞淋巴瘤细胞系）中进行。在维奈托克治疗后，具有导致耐药性的潜在遗传扰动的细胞在存活的群体中富集。与先前描述的跨多种疾病模型的功能基因组筛选一致，编码促凋亡机制成员的^{9 个}基因，包括BAX、BCL2L11 (BIM) 和PMAIP1 (NOXA)，是与维奈托克抗性相关的最强命中（图 1A） . ID3或NFKBIA丢失也与维奈托克敏感性降低有关，表明有多种可能的耐药机制。¹⁰

为了将这些结果与接受维奈托克治疗的患者的潜在耐药机制进行比较，我们对从 43 名复发/难治性 CLL 或小细胞淋巴细胞白血病患者的治疗开始前和/或治疗停止时收集的 CD19+ B 细胞提取的 DNA 进行了全外显子组和靶向测序淋巴细胞性淋巴瘤。所有患者均参加维奈托克单药治疗研究（NCT01328626，5 n = ¹⁸  或 NCT01889186，6 ⁿ  = 25）并接受治疗直至疾病进展或出现不可接受的毒性。这些分析的选择基于样本可用性。基线特征如表 S1 所示。

中位治疗时间为 20.1 个月（范围，1.2-60.3），中位随访时间为 61.3 个月（95% CI，48.6-73.4）。维奈托克在该队列中几乎与在整个队列中一样有效^5、6（以及三项维奈托克研究中 347 名患者的汇总数据¹¹），总缓解率 (ORR) 为 69.8%，完全缓解/完全缓解造血不完全恢复 (CR/CRi) 率为 11.6%，中位无进展生存期 (PFS) 为 21.9 个月（表 S2）。尽管并非所有患者都需要进行微小残留病 (MRD) 评估，但在 MRD 评估≥ 1 的 12 名患者中，有 5 名患者在 10 -4 时达到无法检测的^MRD在外周血中，而 7 个具有可检测的 MRD。由于 CLL 进展（n  = 29）、里氏转化（n  = 8）或其他原因（n  = 6），在研究中止时收集治疗后标本（表 S3）。

为了研究 CLL 中维奈托克治疗后克隆结构的变化，我们关注了 CLL 中突变最严重的三个基因：NOTCH1、SF3B1和TP53（图 1B）。尽管在这些基因中观察到了一些突变的得失，但在维奈托克治疗后的标本中检测到了治疗开始前存在的大多数突变。在四名和两名患者的治疗后样本中分别检测到NOTCH1突变的增加和丢失。一名患者的 CLL 在进展时获得了SF3B1突变，而在三名患者中，不再检测到在开始时存在的SF3B1突变克隆。我们没有观察到TP53的收益突变。对于三名患者，随后未检测到开始时存在的TP53突变克隆，这支持了基于维奈托克的治疗对具有这些突变的 CLL 患者有效的观点。两名患者获得了BRAF V600E突变，这与最近的一份报告一致。¹²总体而言，根据评估的标记物，三分之一 ( n  = 14/43) 的患者在治疗期间表现出克隆结构的变化。

之前的研究已经确定了与维奈托克复发相关的BCL2突变。^13-15在我们的队列中，在研究开始时未发现BCL2突变，但在来自 43 名患者中的 15 名 (35%) 的 CLL 细胞中获得了BCL2突变，其中 14 名经历了 CLL 进展（图 1C）。在 29 名 CLL 进展患者中，10 名接受治疗 ≥ 24 个月的患者发现BCL2突变。总体而言，30 个不同的BCL2确定了突变，在 9 名患者中发现了多个突变（图 1D，E）。观察到的任何描述的最常见突变位于残基 D103（11 次）和 G101（8 次）。此外，在 F104（1 次）、A113（3 次）和 V156（4 次）处发现了突变。在三名患者的样本中鉴定出残基 110 处的 4-氨基酸 (RRYR) 插入。除插入外，所有BCL2改变均为错义突变。与之前的发现一致，在次要克隆中发现了^{13、14 个}大多数BCL 2 突变（图 1F）。只有 1 个BCL2突变 (D103Y) 的变异等位基因频率 (VAF) > 20%，而大多数SF3B1和TP53突变发生在较大的克隆中（VAF > 20%）。

鉴定了BCL2以外的细胞凋亡调节因子的突变，包括PMAIP1和BAX，这是 CRISPR 筛选中鉴定出的两个最强命中。所有PMAIP1突变都是功能丧失，包括三个无义突变和一个导致翻译起始蛋氨酸 M1 丢失的突变。一个BAX突变是无意义的；另一个突变用组氨酸取代了第 181 位氨基酸的亮氨酸。尽管两个BAX突变的 VAF < 10%，但四个PMAIP1突变中的三个的 VAF > 20%，表明PMAIP1突变可能是维奈托克耐药性的关键驱动因素，但还需要进一步研究。除了BCL2突变外，一名患者的促凋亡基因BAD (G111R)也发生突变，VAF >25%；其功能相关性未知。三名PMAIP1突变患者在治疗结束时也存在BCL2突变，尽管等位基因频率不同，这支持了个体患者体内存在多种潜在耐药机制的观点。

尽管 CLL 患者对基于维奈托克治疗的初始临床反应相似，无论TP53是否因 del(17p) 和/或TP53突变而中断，但具有 del(17)p/ TP53突变的患者的反应持续时间较短。^{11, 16, 17}在我们的队列中，78% 的患有 CLL 进展或里氏转化的患者 ( n = 29/37)在基线时 具有 del(17p)/ TP53突变。此外，87% 的BCL2突变患者 ( n  = 13/15)有TP53中断的证据。

为了更深入地了解可用于这些探索性分析的患者的特征，报告了结果。在发生里氏转化的 8 名患者中，ORR 为 25.0%，中位 PFS 为 5.3 个月（表 S2）。虽然有两名患者实现了部分缓解 (PR)，但大多数患者在诊断为里氏转化之前对维奈托克无效。一名获得BCL2 的患者突变达到 PR 但在 6 个月后复发。相比之下，由于 CLL 进展而最终复发的患者的 ORR 为 86.2%，其中 7 例达到深度缓解（CR/CRi/结节性部分缓解），18 例达到 PR。中位反应持续时间 (DOR) 为 23.7 个月，中位 PFS 为 21.9 个月。在 CLL 进展的复发患者中，复发时具有BCL2突变的患者的中位 PFS为 27.3 个月，而对于没有BCL2突变的患者，中位 PFS 为 10.8 个月。任何凋亡基因发生突变的患者都得到了类似的结果；具有或不具有任何凋亡基因获得性突变的患者的中位 PFS 分别为 27.3 和 10.8 个月。最后，对于获得BCL2突变的患者或其他凋亡基因，与无突变患者的 DOR 相比，DOR 更长（表 S2）。鉴于观察到凋亡基因突变的获得并不预示反应较短，我们评估了进展前和进展后标本中的基因表达变化。

为了研究发生的转录变化，我们对来自 CLL 进展患者的 16 对维奈托克前后样本进行了 RNA-seq。样本比较确定了 800 多个基因表达的变化 (FDR < 0.05)。在 IL-8、Tec 激酶和磷脂酶 C 信号传导中起作用的几个基因在复发患者中上调（图 1G）。对接受不同持续时间研究的患者的治疗前和治疗后样本进行比较，未发现任何差异表达的基因。我们有兴趣观察BCL2家族基因表达的变化，因为已知它们能够在体外影响维奈托克耐药性（图 1H）。BCL2呈下降趋势观察到 venetoclax 后样本中的表达，尽管差异未达到统计学显着性。同时，我们观察到MCL1、BCL2L1 (BCL-X _L ) 和BCL2A1 (BFL-1)的表达在统计学上显着增加，当 BCL-2 被维奈托克有效靶向时，这三个基因的蛋白产物可以抑制细胞凋亡。

除了先前公布的BCL2突变外，我们还在对维奈托克单药治疗产生耐药性的 CLL 患者中发现了PMAIP1、BAX和BAD的新突变。这是对来自临床试验患者样本的体外筛选结果的首次确认。我们的研究结果表明BAX、BAD或PMAIP1的失活可能足以使 CLL 细胞对维奈托克治疗产生耐药性，从而将这些因素确定为维奈托克反应的潜在关键调节因子。这是第一份评估该患者群体中 mRNA 表达差异的报告。观察到的基因表达变化表明，肿瘤细胞可能会上调 BCL-2 以外的抗凋亡 BCL-2 蛋白，以避免维奈托克诱导细胞凋亡。总体而言，我们的数据表明，CLL 患者对维奈托克单药治疗的耐药性是多因素的，并且高度依赖于改变促凋亡机制和抗凋亡机制之间平衡的突变和基因表达变化。需要更多的研究来确定次要BCL2的临床相关性venetoclax 耐药性的突变克隆以及在接受固定持续时间的 ve​​netoclax 联合疗法治疗的患者中是否出现类似的耐药机制。