A noncoding mutation in the NOTCH1 gene initiates oncogenic NOTCH signaling via wild-type NICD stabilization in CLL Open Access

Chronic lymphocytic leukemia (CLL) has undergone a dramatic increase globally in the last several decades, which makes it the most common leukemia.¹ According to the National Cancer Institute’s Surveillance, Epidemiology, and End Results program, it is estimated that 0.6% of inhabitants with male dominance (1.7:1) will be diagnosed with CLL at a median age of 70 years (https://seer.cancer.gov/statfacts/html/clyl.html). Patients usually have an immunocompromised state and frequent infections due to clonal accumulation of dysfunctional CD5⁺ B cells within the peripheral blood, bone marrow, lymph nodes, and spleen.^2,3 A central question driving the attention of clinical hematologists is how to develop reliable tools distinguishing different prognostic subgroups thus optimizing the management of patients with CLL due to the indolent disease course. Immunoglobulin heavy chain variable (IGHV) status is one of the risk factors and suggested being evaluated at diagnosis.⁴ IGHV-nonmutated CLLs are supposed to be more aggressive due to higher frequency of adenopathy and tumor burden, correlated with more adherent and invasive cells.⁵ Because the NOTCH pathway is activated in almost half of all CLL cases, an integrated model included NOTCH1 mutations to stratify an intermediate-risk group with a 10-year survival rate of 37%.^6,7 

The NOTCH signaling pathway is required for the regulation of multiple processes, such as embryonic and adult development and tissue homeostasis, but it is also implicated in malignant transformation.^8,9 NOTCH signaling is critically involved in the maintenance of hematopoietic stem cells and in the specification of the T-lineage cells and especially seems to be important for the differentiation of antibody-secreting B cells.¹⁰ After NOTCH signaling is initiated upon binding of a ligand from a neighboring cell, NOTCH1 receptor is cleaved at 2 specific sites. The NOTCH1 intracellular domain (NICD) translocates from the cell membrane to the nucleus after sequential cleavage by “a disintegrin and metalloprotease” (ADAM; S2) and γ-secretase (S3). The interaction of NICD and the responsive elements is mediated by the DNA-binding protein RBP-J. In addition, mastermind like (MAML) and other transcriptional coactivators are then recruited to transcriptionally activate NOTCH target genes. The NICD, per se, is an unstable protein, and the turnover is regulated by CDK8-dependent phosphorylation and subsequent ubiquitination by FBXW7 E3 ubiquitin ligase at its C-terminal proline-glutamic acid-serine-threonine–rich PEST domain leading to proteasomal degradation.¹¹ However, ubiquitination of proteins is counteracted by deubiquitinases which catalyze hydrolysis and removal of ubiquitin chains.¹² Ubiquitin-specific peptidase 28 (USP28) is localized on the long arm of chromosome 11 (11q23). Interestingly, deletions in 11q [del(11q)] are detected in up to 25% of patients with CLL.^3,13 Thus, the correlation between del(11q) and USP28 and its function and effects on NICD degradation are of great interest in CLL research.

At diagnosis of CLL, the most frequently mutated gene is NOTCH1, ranging from 4% to 20%. NOTCH1 mutations are related to the risk of progression to Richter syndrome, which is consistent with the fact that 30% Richter-transformed CLL cases have mutated NOTCH1.^14,15 In CLL, most NOTCH1 mutations affect the C-terminal PEST domain, characterized by coding mutations and noncoding mutations which both cause protein truncation, loss of the PEST domain, and subsequent dysregulation of the FBXW7-dependent NICD degradation system. Unlike the NOTCH1 coding mutations, from which the truncated PEST domain leads to aberrantly increased NICD stability and activated NOTCH1 signaling, consequences of noncoding NOTCH1 mutations are not well understood.¹⁴ 

NOTCH1 noncoding mutations in the 3′ untranslated region (3′UTR) were first reported in 2015.¹⁶ According to their locations at chromosome 9 (Genome Reference Consortium Human Build 37 patch release 13), they are g.139390152 A>G, g.139390145 A>G, and g.139390143 A>C∗ (referred to NOTCH1 152, NOTCH1 145, and NOTCH1 143∗, respectively, in this study). Among them, the most prevalent one is NOTCH1 152 (74.5%), followed by NOTCH1 145 (21.3%) and NOTCH1 143∗ (4.2%).¹⁴ Therefore, in this study, most of our attention was driven to the mutation NOTCH1 152. In clinical trials, patients with CLL with noncoding NOTCH1 mutations exhibited adverse outcome, with shorter overall survival (OS) compared with a wild-type (wt) NOTCH1 cohort, of note, shorter time-to-first treatment (TTT) even compared with the NOTCH1 coding mutation group.^16-18 Furthermore, noncoding NOTCH1-mutated cases were related to low CD20 expression; therefore, distinguishing 1 subgroup which might have limited benefits from anti-CD20 treatment (rituximab).^2,19 In the era of precision medicine, deciphering mechanisms of altered NOTCH1 signaling by noncoding mutations in CLL might contribute to distinguish subgroups with diverse outcomes and to provide evidence for treatment decisions and new therapeutic response predictors to patients with resistance to current treatments.

Detailed information about topoisomerase I cloning, cell culture, quantitative polymerase chain reaction (qPCR), cloning of plasmids used in this study, immunofluorescence microscopy, luciferase assays, coimmunoprecipitations, nuclear extracts, western blotting, fluorescence-activated cell sorting analysis, Giemsa staining, and analysis of clinical data is available in the supplemental Methods, available on the Blood website.

Owing to the high sensitivity (98.7%) and specificity (96.5%) in predicting variant splicing efficiency, the maximum entropy model^20,21 was applied using MaxEntScan (MaxEntScan::scoresplice [mit.edu]) to calculate the splicing score of NOTCH1 canonical and cryptic splice sites (SSs).

In cooperation with the Institute of Clinical Transfusion Medicine and Immunogenetics, a joint venture of German Red Cross Blood Service Baden-Wuerttemberg–Hessen and the University Hospital Ulm, whole blood donated by healthy adults was processed as described²² to obtain peripheral blood mononuclear cells (PBMCs) using Ficoll-Hypaque (catalog no. L6115; Biochrom AG). Samples from patients with CLL were provided by the Division of Chronic Lymphocytic Leukaemia, Department of Internal Medicine III, Ulm University (approved ethics vote 459/19) after CD19 MicroBeads (catalog no. 130050301; Miltenyi Biotec) sorting as mentioned.²² 

The study was approved by the ethics committee of Ulm University (institutional review board statement: approved ethics vote 459/19) and conducted in accordance with the principles of the Declaration of Helsinki of the World Medical Association.

NOTCH1 3′UTR splicing-specific PCR primers were designed as follows: SP_forward: 5′-CCT GGC GGT GCA CAC TAT TCT G-3′, SP_reverse: 5′-CTT TTT GGA CTA TGC TCG TTC AAC TTC C-3′. The Taq DNA polymerase, recombinant kit (catalog no. 10342046; Invitrogen), was used to amplify target NOTCH1 fragments from complementary DNA.

The design of the CRISPR RNA was performed using the Integrated DNA Technologies genome-editing tool (https://eu.idtdna.com/pages). The CRISPR RNA (5′-TCT TTT TGG ATT TTG AAA AA-3′) with the highest on-target potential and the lowest off-target risk was selected for further procedures (see supplemental Methods).

Comparison of significances was performed in GraphPad Prism 9.0, and quantitative data of at least 3 independent experiments were represented by mean ± standard deviation. Significant differences were determined by P value <.05. Details on the statistical tests are found in the legends for Figures 2, 3, 5, 6, and 7 and supplementary Figures 6, 7, and 11.

To investigate the role of 3′UTR mutations in CLL, we analyzed baseline samples from 700 patients of the CLL11 trial (ClinicalTrials.gov identifier: NCT02035462), which randomly assigned untreated patients with CLL with need for treatment to 12 cycles of chlorambucil with/without CD20 antibody (supplemental Table 1). We identified a total of 18 3′UTR NOTCH1 mutations in 18 patients with CLL, of which 15 were g.139390152 A>G and 3 were g.139390145 A>G. The presence of 3′UTR mutations reveals correlation with β2-microglobulin >3.5mg/L and (very) high CLL–International Prognostic Index (both P = .05). All but 1 patient had undetermined IGHV (P < .01), whereas absence of del(13q) (P = .05) was more frequent in 3′UTR-mutated cases. No association was found with the high-risk markers del(17p) or TP53 mutation.

In the context of treatment efficacy (supplemental Table 3), undetectable minimal residual disease rate was higher in 3′UTR cases (P = .04). However, 3-year progression-free survival (PFS) rate was lower (15.4%) as compared with wt NOTCH1 (24.4%) and close to patients with coding NOTCH1 mutations (18.2%, cases with concurrent 3′UTR and coding NOTCH1 mutations excluded; supplemental Figure 1). The observation of an enhanced response to therapy coupled with a more rapid disease recurrence indicates a more aggressive form of CLL; however, this does not depend on known co-occurring high-risk gene mutations. To comprehensively understand the molecular underpinnings of this aggressive CLL phenotype, we performed a detailed analysis of different 3′UTR NOTCH1 mutations and their impact on aberrant splicing events.

Noncoding mutations in the human NOTCH1 gene result in aberrant splicing events and lead to excision of 158 bases of the NOTCH1 coding sequence (Figure 1A).¹⁶ However, the detailed regulation of the splicing events and the function of the resulting NOTCH1 proteins have not been analyzed, so far.

For this, we first calculated specific splicing scores within the human NOTCH1 gene without any mutation using the MaxEntScan software in which higher scores predict higher splicing efficiencies. Adjacent 5′SS reveals scores of 7.23 and 7.75, whereas 3′SS reveals the scores 9.04 and 11.86, respectively (Figure 1B). Calculation of the scores of all 5′SS and 3′SS in the human NOTCH1 gene shows a distinct variance with the mean of 8.2 (5′SS) and 9.2 (3′SS; Figure 1C; supplemental Table 4). Interestingly, the NOTCH1 152 mutation created an acceptor site, scored from 1.02 (wt) to 9.77 which is even higher than the average of all NOTCH1 canonical acceptor scores (Figure 1C-D; supplemental Tables 4 and 5). The NOTCH1 152 score ranges between acceptor scores of exons 32 to 33 (9.04) and exons 33 to 34 (11.86), followed by NOTCH1 143∗ (6.38) and NOTCH1 145 (5.40), representing scores still higher than the minimum canonical value (4.60).

Next, we developed a specific PCR protocol (Figure 2A; supplemental Methods) to monitor aberrantly spliced transcripts together with unspliced NOTCH1 complementary DNA in human samples. Spliced fragments (size ∼200 base pairs) were amplified in samples from patients with CLL with NOTCH1 152, NOTCH1 145, and NOTCH1 143∗ mutations. Sequencing data from isolated fragments confirmed predicted SSs (Figures 1A and 2B). Interestingly, the A>C mutation at position 143 of the NOTCH1 gene induces a cryptic splice acceptor which generates a spliced transcript identical to that, which we already detected in healthy PBMCs (Figure 2B, upper). This cryptic splice acceptor (“141”) exhibiting a splicing efficiency score of 1.37 (as predicted by MaxEntScan) was increased to 6.38 by the 143 A>C mutation (Figures 1D and 2B; supplemental Table 5). Remarkably, the NOTCH1 141 splicing event detected in healthy donors seems to be associated with the lymphoid lineage, because splicing PCR signals were barely detectable in cell lines from the myeloid lineage (supplemental Figure 2).

Spliced NOTCH1 152-specific transcripts were also detected by quantitative PCR in patients with CLL (Figure 2C). Importantly, we could also determine unspliced NOTCH1 messenger RNA (mRNA) derived from the wt allele due to the heterozygotic status of the NOTCH1 152 mutation (Figure 2C). Thus, in corresponding patients with CLL, NOTCH1 152 mRNAs are robustly coexpressed with wt NOTCH1 transcripts.

To test the splicing efficacy initiated by such NOTCH1 3′UTR mutations, we developed a splicing reporter system including the NOTCH1 152 and the “physiological” (cryptic) NOTCH1 141 sites (Figure 3A). An NICD construct fused to enhanced green fluorescent protein (EGFP) without 3′UTR served as a positive control (Figure 3B, NICD EGFP). Expressed NICD proteins are localized in the nucleus (Figure 3B, α-NOTCH1). The cryptic splice acceptor at position 141 induced splicing and expression of the fusion protein (Figure 3B, NICD 141 wt; Figure 3C, bar 2) which was also detected by western blotting (Figure 3D, upper [lanes 3-4] and lower [lane 4]). After mutation of the cryptic splice acceptor (141 G>A), splicing was prevented and no EGFP expression could be detected (Figure 3B, NICD 141 G>A; Figure 3C, bar 3; Figure 3D, upper [lane 5]). The patient-specific NOTCH1 152 mutation resulted in efficient splicing (Figure 3B, NICD 152 A>G; Figure 3C, bar 4; Figure 3D, upper [lane 6] and lower [lane 5]). Mutation of the cryptic splice donor (Figure 3A, donor mutated [dmut], GG>AC) abrogated the splicing process of all constructs, and EGFP expression could not be measured any longer (Figure 3B, 3 bottom rows; Figure 3C, bars 5-7; Figure 3D, lower [lanes 1-3]), thereby confirming the predicted 5′SS MaxEntScan score after mutation (wt, 5.40 vs dmut, −9.33; supplemental Table 6). Taken together, our results clearly demonstrate that a functional cryptic splice donor exists at position “g.139390681” but a cryptic splice acceptor at position “g.139390141” does exist in the wt NOTCH1 gene. In addition, a 3′SS acceptor, which is 11 nucleotides apart, is converted into a highly effective SS after the A>G mutation at position “g.139390152” in patients with CLL.

The PEST domain of NOTCH1 is missing in all splice variants (Figure 4A), whereas a novel carboxy terminus is generated in NICD 152/143. The variant 143∗ (A>C) directly enhances the 141 cryptic splice acceptor (MaxEntScan score 1.37 vs 6.38; Figures 1D and 2B, patients 3 and 4; supplemental Figure 3), whereas the variant 143 (A>G) shares most of C-terminal amino acids with NICD 152 (Figure 4A). As described previously, NICD 141 and NICD 143∗ shared exactly the same sequence due to use of the same 3′SS (also found in supplemental Figure 3), hereafter, referred to as NICD 141.

Unlike NICD 141 and NICD 145, spliced NICD 152/143 were almost of the same size as NICD wt (supplemental Figure 4), which makes it difficult to distinguish the aberrant NOTCH1 152 from NOTCH1 wt in patients with CLL owing to the monoallelic mutation. Using Basic Local Alignment Search Tool for Proteins, we did not find specific peptides of the aberrant C terminus of the NOTCH1 152 protein within the human proteome. Indeed, using mass spectrometry, we could detect the NICD 152 specific peptides (SSWGVQAAPFPR, PARPLHQHHLPVR; supplemental Figure 5; supplemental Table 9). Therefore, the spliced NICD 152 variant expressed from a transfected NICD+3′UTR 152 construct was stable enough to be detected by mass spectrometry.

We used a NICD 152 specific peptide “NPKRNDVGEGKLNE” as an antigen to generate a NICD 152/143 specific polyclonal antibody (α-NICD 152; Figure 4). The α-NICD 152 antibody was highly specific to NICD 152/143 but did not recognize other NICD variants and/or endogenous NICD (wt) proteins in different cell lines. Importantly, the antibody could also detect subsequent spliced NICD 152 and even NICD 143 from transfected NOTCH1 constructs with the corresponding 3′UTR mutations, although splicing efficiency of the latter was low (Figure 4B), according to the MaxEntScan score of NOTCH1 143 (4.09; Figure 1D). After immunoprecipitation, but not directly from the whole cell lysates (supplemental Figure 12A), the α-NICD 152 was able to detect specific protein signals at the expected size in patients with NOTCH1 152 mutated CLL although with varying intensities due to the heterogeneity and varying mutation frequencies in patient samples. In lysates from patients with wild-type NOTCH1 and patients harboring the NOTCH1 c.7544_7545delCT (ΔCT) mutation, no specific signal was detectable. A commercially available antibody was used as a control (Figure 4C).

Although mainly localized in the nuclei such as other variants (NICD wt, coding mutants: NICD ΔG and NICD ΔCT; supplemental Figures 4B and 6A), NICD 152, NICD 145, and NICD 143 activate the RBPJ-responsive reporter gene (supplemental Figure 6B) only weakly, whereas, NICD 141 was more active. In parallel, western blots with the same constructs excluded bias from different transfection efficiencies (supplemental Figure 6C). Lower transactivation capacities of spliced NICD variants were not caused by different affinities to the transcription complex, as they demonstrated comparable binding capacities to the MAML1-RBPJ complex (MAML1, endogenous; Figure 5A; MAML1, overexpressed; supplemental Figure 6D) when compared with NICD wt.

The low transcriptional activity of the spliced NICD variants (supplemental Figure 6B) could not explain the aggressive phenotype of NOTCH1 3′UTR–mutated CLL entities. Therefore, we evaluated the activity of FBXW7 and USP28 (Figure 5B) because these components of the NICD degradation system were previously proven to contribute to malignant transformation in CLL^26,27 development. Surprisingly, NICD 152 and NICD 143 (Figure 5C) demonstrated higher binding capacity to overexpressed and endogenous USP28 compared with NICD wt (with faint signal), whereas NICD ΔCT showed undetectable affinity to USP28 (Figure 5D-E). The shortened spliced NICD 145 (supplemental Figure 4B) also showed lower affinity to endogenous USP28 (Figure 5E) suggesting that the aberrant C terminus of NICD 152 triggers a strong affinity to USP28. Therefore, we used GFP proteins fused to the C-termini of NICD variants (Figure 5F) and performed coimmunoprecipitation experiments with FBXW7 and USP28. Although lacking the FBXW7 binding site (FLTPSPE; Figure 5C),²⁸ the C terminus of NICD 152 still binds to FBXW7 comparable to wt NICD (Figure 5H) and has an even higher binding affinity to USP28 (Figure 5G).

Having demonstrated that the specific C terminus of NICD 152 gives rise to a high-affinity interaction with the NICD degradation system, we next asked for the functional consequences. For this purpose, we conducted an ubiquitinylating competition assay (Figure 6A) which revealed that precipitated NICD wt was less ubiquitinylated on increasing NICD 152 expressions in HEK 293 cells and NOTCH1 knockout (KO) HEK 293 cells.²⁴ Thus, decreased ubiquitylation of NICD wt in the context of simultaneous NICD 152 expression might lead to stabilized NICD wt protein, a phenotype possibly reflecting the situation of patients with CLL with monoallelic NOTCH1 152 mutation.

In line with this hypothesis, reporter gene assays demonstrated that the addition of the NICD 152 specific C terminus is able to activate NICD wt-dependent transcription when compared with NICD wt only or after cotransfection of the NICD wt/NICD ΔCT C-termini (Figure 5I). Furthermore, on overexpressing NICD 152 in HEK 293 wt, HeLa wt, and NOTCH1 KO HEK 293 cells²⁴ (Figure 6B-C; supplemental Figure 7), our NICD wt protection model was proven by significantly accumulated active NICD wt. Unlike in the whole cell extract, nuclear NICD 152 protein was not stable. Nevertheless, addition of proteasome inhibitor MG132 during the extraction process contributed to proper signals of NICD 152 in HEK 293 wt cells (supplemental Figure 8), which in turn indirectly confirmed our hypothesis that NICD 152 might be ubiquitinylated by FBXW7 and degraded by the proteasome, thereby stabilizing active NICD wt and allowing it to act as an oncogenic factor.

To further validate the oncogenic activity of NICD 152, we introduced this type of mutation in HG3 cells by CRISPR/Cas9. Indeed, we could identify stabilized nuclear NICD wt (Figure 7A-B; supplemental Figure 9) in CRISPR/Cas9 edited HG3 cell lines (HG3 11 and HG3 149) which carry a heterozygous NOTCH1 152 mutation. The specific phenotype was confirmed by genomic sequencing (supplemental Figure 9A), splicing PCR assays (supplemental Figure 9B), and most importantly, the presence of spliced NOTCH1 152 protein (supplemental Figure 9C). In contrast to HG3 wt, both HG3 149 and HG3 11 demonstrated a more adherent phenotype, colony formation activity, and characteristics of enhanced mobility (supplemental Figure 10). In addition, we determined an elevated expression of classic NOTCH1 target genes according to Fabbri et al⁷ and Ryan et al³⁰ in HG3 149 (HEY1, HEY2, DTX1, NRARP1, NEDD9, ATF5, JAK1, ADAM19, TMOD1, and IL6R; supplemental Figure 11) as an immediate consequence of accumulated NICD. Interestingly, we also uncovered the stabilization of other FBXW7 targets (c-Jun, c-Myc, and cyclin E) in HG3 149 cells (supplemental Figure 9E) suggesting also NOTCH1 indirect/independent effects.

Finally, we analyzed the abundance of cleaved NICD proteins in patients with CLL (Figure 7C). Stabilized NICD was clearly detected in samples from patients with the coding NOTCH1 ΔCT mutation (Figure 7C, lanes 6-10) and, most importantly, also in patients with the noncoding NOTCH1 152 mutation (Figure 7C, lanes 11-15). Lysates from NOTCH1 KO HEK 293 cells²⁴ expressing NICD wt and variants missing the Val1754 residue served as a negative control. Because NICD 152 protein is instable also in both HG3 clones (supplemental Figure 9C) and patient samples (supplemental Figure 12A) and can only be detected after immunoprecipitation (Figure 4C and supplemental Figure 9C), therefore, in contrast to the truncated proteins in “ΔCT” samples, stabilized NICD found in “152” mutated patient samples is most likely the wt protein. This supports our notion that the NICD 152 protein acts as a sponge thereby affecting stability of NICD wt protein in trans.

NOTCH1 mutations in CLL are found in up to 20% of patients, stratifying subgroups with poor prognosis and higher risk of Richter transformation. Stabilized NICD variants and constitutively active NOTCH1 signaling due to a PEST domain truncation are supposed to contribute to CLL oncogenesis.¹⁴ Recent studies^16,19,31 concerning noncoding NOTCH1 mutations hypothesized that the oncogenic activity of those was also initiated by the accumulation of NICD 152 through impaired degradation, due to the PEST deletion and subsequent constitutively activated NOTCH1 signaling. But none of them directly dissected the NOTCH1 spliced variants and their function in the heterozygotic background.

Our work uncovered a shortened spliced NOTCH1 variant in healthy condition, which is shared by patients with CLL harboring the NOTCH1 143∗ A>C mutation (Figures 2B and 4B; supplemental Figure 3). The corresponding NICD 141 variant might have oncogenic activity during initiation and development of CLL, but its physiological presence, although with low amount, is surprising. Unfortunately, quantitative data are still missing from patients with NOTCH1 143∗-mutated CLL due to their low frequency.¹⁴ Nevertheless, our splicing PCR result reveals that NOTCH1 141 is much more abundant in patients with NOTCH1 143∗-mutated CLL compared with that in the healthy circulating PBMCs, and oncogenic activity of NICD 141 might be dose dependent. NOTCH1 signaling was reported to be responsible for the regulation of T-cell dependent and T-cell independent B-cell responses during immune reaction. A low expression of a PEST deleted, and therefore more stable variant (NICD 141), might be necessary for recirculating B cells in response to inflammation.¹⁰ In the physiological context, this splice variant is expressed in the wt situation. Therefore, NICD 141 expression may be controlled by a cell-specific regulation of alternative splicing processes and thus also be switched off again. However, further cell type-specific splicing PCR needs to be performed to figure out the origin and function of the physiologically spliced NOTCH1 141.

We identified that this NICD variant was associated with unexpected significantly lower transactivation activity compared with the wt NICD but exhibited aggravated affinities to the NICD degradation system components (FBXW7 and USP28). Mutations in FBXW7 and its antagonist USP28 were reported to contribute to tumorigenesis of CLL.^26,27 We suggest that in patients with CLL expressing NICD 152, malignant transformation is induced by aberrantly stabilized NICD wt because the ubiquitin ligase FBXW7 and deubiquitinase USP28 are sequestered by this NICD splice variant and might form a tripartite complex (NICD 152-USP28-FBXW7).²⁵ Alternatively, NICD 152 mimics the complete deletion of USP28 by absorbing it and promotes FBXW7 autoubiquitination. Therefore, the discrepancy between former studies^16,19,31 and us might be due to the different NOTCH1 antibodies used. The elevated NICD signals from publications^16,19,31 in patients with NOTCH1 3′UTR–mutated CLL could be stabilized NICD wt but recognized by antibodies unable to distinguish NICD 152 and NICD wt because of similar sizes (Figure 4; supplemental Figures 4-6) and the instability of nuclear NICD 152 (supplemental Figure 8).

The heterogenous and variable disease course of CLL requires accurate evaluation at diagnosis to distinguish subgroups with different prognosis in terms of TTT, PFS, and OS. With the current stratification system,^2,6 subgroups with adverse prognosis should be intensively monitored. However, detection of noncoding NOTCH1 mutations which is related to significant shorter TTT usually demands whole genome sequencing or whole exome sequencing which is not available in regular clinical practice. The novel C terminus of NICD 152 and a specific α-NICD 152 antibody allows to distinguish CLL entities with this dominant mutation at diagnosis easily and to classify them to a more aggressive phenotype¹⁷ (supplemental Table 1) with an intensive follow-up protocol, although its predictive significance for Richter transformation is currently not determined. In a subgroup analysis of the CLL11 trial,³² patients with CLL with NOTCH1 3′UTR mutations demonstrated better response to the type II CD20 antibody obinutuzumab in terms of negativity of minimal residual disease and prolonged PFS compared with rituximab. However, obinutuzumab-containing regimens were not able to benefit treatment-naïve CLL in terms of OS³³ and diffuse large B-cell lymphoma (entity after Richter transformation) in terms of PFS³⁴ compared to a rituximab-containing therapy. The inclusion of NOTCH1 noncoding mutations in the CLL stratification panel would give additional evidence to decision propensity for obinutuzumab in clinical practice, while balancing between increased risk of infusion-related reactions, secondary malignancies,³⁵ and clinical benefit.

Furthermore, low CD20 expression in CLL with noncoding NOTCH1 mutations can be partly rescued by γ-secretase inhibitor,¹⁹ which is consistent with higher nuclear NICD levels and NOTCH1 signaling activity in this study (Figure 6B-C and 7; supplemental Figures 7 and 11), providing a solution for anti-CD20 immunotherapy resistance. Meanwhile, histone deacetylases are reported to be involved in stabilized NICD-mediated repression of CD20 expression and histone deacetylase inhibitor treatment upregulates CD20 as well.³⁶ Furthermore, improved understanding of RNA splicing makes it possible to target the aberrant splicing events of NOTCH1 3′UTR mutations, such as inhibiting SF3B1. Mutations of SF3B1 are usually concurrent with noncoding NOTCH1 mutations³¹ and vital for splice acceptor selection.^37,38 Other options are targeting pathologically spliced mRNA with small molecules³⁹ and antisense oligos preventing use of alternative SSs.⁴⁰ 

In summary, our study provides the opportunity to stratify a CLL subgroup characterized by noncoding NOTCH1 mutations, which may benefit from more precise clinical management and treatment.

The authors thank Eva-Maria Rump and Sabine Schirmer for excellent technical assistance.

The work was funded by the Deutsche Forschungsgemeinschaft (DFG; German Research Foundation no. 427512076 [F.O.]) and the German Cancer Aid (grant numbers 70114289 [F.O.] and 70114291 [D.M.]). M.G. was supported by the China Scholarship Council (file no. 202006090055). T.B. acknowledges funding by the Landes-Offensive zur Entwicklung Wissenschaftlich-Ökonomischer Exzellenz Research Cluster Initiative CANcer-lung Crosstalk and the Excellence Cluster for Cardio Pulmonary System in Giessen. F.F. is supported by a research grant from the University Medical Center Giessen and Marburg. F.O. and B.B. are supported by the Collaborative Research Centers 1074, 1279 (DFG no. 217328187 and 316249678 [F.O.]), and 1506 (DFG no. 450627322 [F.O. and B.B.]).

Contribution: F.O. and M.G. conceived the project; M.G., T.M., and F.O. designed the project; T.M., K.S., and U.P. generated the CRISPR cell lines; M.G., T.M., and A.S.E. performed the experiments; E.T., S.S., H.D., B.B., K.F., and A.P. analyzed the data; D.M., F.F., and T.B. shared the reagents; and M.G., B.B., and F.O. wrote the manuscript, including comments from all authors.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Franz Oswald, Department of Internal Medicine I, University Hospital Ulm, Albert-Einstein-Allee 23, 89081 Ulm, Germany; email: franz.oswald@uni-ulm.de.