Common nonmutational NOTCH1 activation in chronic lymphocytic leukemia

Significance

A pathogenetic role of NOTCH1 in chronic lymphocytic leukemia (CLL) has been implied by the presence of deregulating mutations in a relatively small fraction of cases. Our results now indicate that ∼50% of CLL cases devoid of mutations express the active form of NOTCH1 ICN1 (intracellular portion of NOTCH1), thus implicating a much broader role of this transcription factor in the disease. ICN1⁺ CLL cases display equivalent NOTCH1-dependent transcriptional responses regardless of the gene mutation status, indicating that the detection of ICN1 represents a reliable biomarker of NOTCH1 activation for diagnostic and therapeutic targeting. Finally, our results identify the NOTCH1-dependent transcriptional program in CLL cells, thus providing direct insights into the pathogenesis of a large fraction of CLL cases.

Abstract

Activating mutations of NOTCH1 (a well-known oncogene in T-cell acute lymphoblastic leukemia) are present in ∼4–13% of chronic lymphocytic leukemia (CLL) cases, where they are associated with disease progression and chemorefractoriness. However, the specific role of NOTCH1 in leukemogenesis remains to be established. Here, we report that the active intracellular portion of NOTCH1 (ICN1) is detectable in ∼50% of peripheral blood CLL cases lacking gene mutations. We identify a “NOTCH1 gene-expression signature” in CLL cells, and show that this signature is significantly enriched in primary CLL cases expressing ICN1, independent of NOTCH1 mutation. NOTCH1 target genes include key regulators of B-cell proliferation, survival, and signal transduction. In particular, we show that NOTCH1 transactivates MYC via binding to B-cell–specific regulatory elements, thus implicating this oncogene in CLL development. These results significantly extend the role of NOTCH1 in CLL pathogenesis, and have direct implications for specific therapeutic targeting.

Chronic lymphocytic leukemia (CLL) is a common hematologic tumor characterized by the clonal expansion of CD5⁺ B cells (1, 2). Recent investigations have provided a comprehensive picture of the CLL genome, revealing its relatively low burden of genetic lesions, with a small number of frequently mutated “driver” genes. CLL mutated genes include the NOTCH1 oncogene, the splicing regulator SF3B1, the tumor-suppressors TP53 and ATM, and several B-cell receptor (BCR)/NF-κB regulators, such as MYD88, BIRC3, and NFKBIE, among others (3–7).

NOTCH1, a well-known oncogene in T-cell acute lymphoblastic leukemia (T-ALL) (8, 9), has emerged as the most commonly mutated gene in CLL at diagnosis, accounting for ∼4–13% of patients (3–7). NOTCH1 encodes a transmembrane receptor that, upon binding to a ligand expressed on the surface of a “signal-sending” cell, undergoes a series of conformational changes and proteolytic cleavages, ultimately allowing the translocation of its intracellular, cleaved, and active portion (hereinafter referred to as “ICN1”) to the nucleus (10, 11). Once in the nucleus, ICN1 binds to the DNA-binding protein RBPJ, the main effector of NOTCH-signaling, and recruits a series of coactivator proteins to induce transcriptional activation of target genes (10, 11). NOTCH1 target genes mediate regulation of fundamental biological processes, such as development, cell differentiation, cell-fate decision, proliferation, and apoptosis (10, 11).

In contrast to T-ALL, where the majority of mutations are represented by constitutively activated ligand-independent alleles affecting the heterodimerization domain of the protein, most NOTCH1 mutational events in CLL are represented by PEST [proline (P), glutamic acid (E), serine (S), threonine (T)-rich protein sequence] -truncations removing the phosphodegron sequence required for FBXW7-mediated ICN1 proteasomal degradation; in a minority of cases, point mutations in the 3′UTR of the NOTCH1 mRNA lead to aberrant splicing events that also remove the PEST domain of the NOTCH1 protein (3–7). NOTCH1 mutations in CLL were shown to associate with poor prognosis, including a specific subset of patients carrying trisomy 12, disease progression, transformation to highly aggressive diffuse large B-cell lymphomas, termed Richter syndrome, and immunochemotherapy resistance (3, 4, 12–16).

Despite this potentially relevant role in the CLL clinical course, the oncogenic role of NOTCH1 in this disease remains poorly understood. Although few NOTCH1 targets have been shown to be overexpressed in NOTCH1-mutated cases compared with wild-type CLL (4, 17, 18), the full spectrum of genes controlled by NOTCH1 and their contribution to the disease pathogenesis have not been identified. Moreover, recent reports documented that CLL cells in the lymph node frequently express ICN1, independent of NOTCH1 PEST-truncation (19, 20), especially within the proliferation centers, which represent the key microanatomical sites of interaction of CLL cells with accessory cells and proliferation (21). Accordingly, these findings have been interpreted as the result of microenvironmental signals activating the NOTCH1 cascade. Conversely, the status of NOTCH1 activation in the peripheral blood (PB) compartment of CLL patients is less clear (4, 7, 18, 22).

To address these questions, we have analyzed the functional status of ICN1 in normal mature B cells and in a panel of PB CLL cells including both NOTCH1-mutated and wild-type cases. We report broader NOTCH1 activation, significantly extending beyond the mutated cases and the transcriptional consequences of this activation in leukemic B cells.

Results

NOTCH1 Is Activated in Naïve and Memory B Cells, the Putative Normal Counterparts of CLL.

To obtain a comparative baseline before investigating the activity of NOTCH1 in CLL, we first defined the expression and activation pattern of NOTCH1 in normal mature B-cell subsets. We performed gene-expression profiling as well as ICN1 immunoblot analysis in naïve, germinal center (GC) and memory B cells isolated from human tonsils (23). Although the phenotype of the B cell expanding to generate overt CLL remains a matter of debate, naïve and memory B cells are considered the most likely putative normal counterparts of this disease (24–26). The levels of NOTCH1 mRNA and of the cleaved and active intracellular portion of NOTCH1 ICN1 were abundant in naïve and memory B cells, whereas they were almost undetectable in GC B cells (Fig. 1 A and B). Immunofluorescence staining of human tonsillar biopsies confirmed these findings, revealing ICN1 nuclear staining in the B-cell fraction populating the mantle zone of the GCs, which is highly enriched in naïve B cells (Fig. 1C and Fig. S1).

Fig. 1.

NOTCH1 is expressed and activated in naïve and memory B cells, putative normal counterparts of CLL. (A) Gene-expression profile analysis (HG-U133 Plus 2.0 Array) of NOTCH1, MYC, HES1, and BCL6 in normal mature naive, GC, and memory B-cell subpopulations isolated from human tonsils (23). Each column corresponds to an independent sample. The mRNA expression pattern of NOTCH1 in naïve and memory B cells is similar to that of MYC, typically expressed only in a small fraction of GC–B cells (69), and opposite to that of BCL6, a known GC master regulator (81). Moreover, NOTCH1 expression levels are concordant with those of HES1, a NOTCH1 target in multiple tissue types (11). (B) Immunoblot (IB) analysis of ICN1, BCL6, MYC, and control β-actin in mature B-cell subpopulations isolated from human tonsils. (C) Immunofluorescence (IF) staining of ICN1, the dark-zone GC-marker AID (82), and the B-cell–specific surface antigen CD20 in a human tonsil section. (D) Tracking of the HALLMARK_NOTCH_SIGNALING geneset from the Molecular Signatures Database v5.1 (software.broadinstitute.org/gsea/msigdb/index.jsp) in normal mature B-cell subpopulations by GSEA. Abbreviations: DZ, dark zone; LZ, light zone; M, mantle zone.

Fig. S1.

ICN1 is expressed in the B-cell fraction populating the mantle zone (M) of the germinal centers (GCs). (A) Double IF staining of ICN1 and AID in a representative GC in a human tonsil section. (B) Double IF staining of ICN1 and the B-cell–specific surface antigen CD20 at lower magnification (4×) in a human tonsil section.

Gene set enrichment analysis (GSEA) using the “Hallmark_Notch_Signaling” signature from the Molecular Signature Database (27) confirmed that ICN1 expression in naïve and memory B cells is associated with NOTCH1 transcriptional activity (Fig. 1D). Taken together, these data indicate that NOTCH1 is physiologically expressed and activated in the cells of origin of CLL.

PB CLL Cells Express ICN1 in both NOTCH1-Mutated and Wild-Type Cases.

We next investigated the incidence of NOTCH1 pathway activation in CLL by analyzing peripheral blood CLL cells (>70% purity in 90 of 93 cases analyzed by cytofluorimetry analysis for CD5⁺/CD19⁺) (Materials and Methods) from a cohort of primary CLL cases (n = 124). Twenty-two percent of these cases carried NOTCH1 PEST-truncating events (n = 27 of 124) (Dataset S1), representative of the prototypical p.P2515fs mutation (n = 17 of 29, 58.6%), frameshift deletions (n = 3 of 29, 10.3%), and nonsense mutations (n = 5 of 29, 17.2%). In addition, four cases carried 3′UTR NOTCH1 mutations known to lead to aberrant splicing events disrupting the PEST domain of the ICN1 protein (7). The majority of NOTCH1-mutated CLL cases carried unmutated IGHV genes (n = 23 of 24 with known IGHV status, 95.8%), as previously reported (3, 4, 12).

Notably, ICN1 was detectable by immunoblot analysis in 50.5% (n = 49 of 97) of NOTCH1–wild-type cases (Fig. 2 A and B and Fig. S2A). Among these, ICN1 expression occurred in 53.3% (n = 24 of 45) of IGHV mutated and 41.9% (n = 13 of 31) of IGHV-unmutated cases, respectively (Dataset S1).

Fig. 2.

Primary CLL cases express ICN1 because of NOTCH1 PEST-truncations or alternative mechanisms. (A) IB analysis of ICN1 and control β-actin in 10 representative PB CLL cases, 4 carrying NOTCH1 PEST-truncations (ΔPEST) and 6 NOTCH1–wild-type (WT), in the control T-ALL cell line CUTLL1 (83) and in MO1043 CLL cells cocultured with OP9 stromal cells expressing the NOTCH1 ligand DL1 (54). The full set of analyzed primary CLL cases, including those reported here, is displayed in Fig. S2. (B) Frequency of ICN1 positivity in 124 primary CLL cases. (C) IF staining of ICN1 in primary ICN1⁺ (pos) and ICN1⁻ (neg) CLL cells and in the control CUTLL1 T-ALL cell line in basal conditions (+) and upon Compound E (CpE, 24 h, 1 μM) treatment (−).

Fig. S2.

ICN1 expression analysis in a panel of primary CLL cases and PBMC. (A) IB analysis of ICN1 and control β-actin in a panel of 124 CLL PB primary CLL cases, (B) in primary NOTCH1–wild-type CLL cells treated with the γ-secretase inhibitor Compound E (CpE, 500 nM, 8 h) or control DMSO, and (C) in PBMC protein extracts and representative primary CLL cases expressing ICN1. Samples are color-coded based on the NOTCH1 mutational status [red, clonal NOTCH1 PEST-truncating events; orange, subclonal NOTCH1 PEST-truncating events; blue, RAG-mediated NOTCH1 translocation (83); and black, NOTCH1–wild-type]. Samples in gray were excluded from the analysis because of low quality of the protein lysate, low viability, or low leukemic representation. Color-coded arrows indicate cases subjected to RNA-Seq analysis: dark red denotes NOTCH1-mutated cases expressing ICN1; blue, NOTCH1–wild-type cases expressing ICN1; and green, ICN1⁻ NOTCH1–wild-type cases. Abbreviations: MO+DL1, MO1043 cells cocultured on OP9-DL1 cells (54); s.e., short exposure; l.e., long exposure.

As expected, all NOTCH1 PEST-disrupted CLL cases expressed a truncated active form of the protein (Fig. 2 A and B and Fig. S2A) (4, 7). The levels of ICN1 expression were variable across the panel, with NOTCH1-mutated cases often displaying higher levels of ICN1 compared with wild-type samples (Fig. S2A). ICN1 levels in NOTCH1–wild-type cases were responsive to NOTCH inhibition by the γ-secretase inhibitor Compound-E (28) (Fig. S2B).

Immunofluorescence analysis showed that strong nuclear ICN1 expression was detectable in virtually 100% of CLL cells in both NOTCH1-mutated and wild-type ICN1⁺ CLL cases (Fig. 2C). Moreover, ICN1 was not expressed in PB mononuclear cells (PBMCs) from healthy, age-matched elderly individuals (Fig. S2C). These observations exclude the possibility that ICN1 expression in NOTCH1–wild-type cases was because of residual contamination by normal cells. Altogether, these results suggest that the activation of the NOTCH1 oncogene in CLL is more common than what is currently known based on the frequency of NOTCH1 activating mutations (Discussion).

Identification of the CLL NOTCH1-Direct Transcriptional Program.

As a tool to further interrogate the functional activity of NOTCH1 in ICN1⁺ CLL cells and to shed light on the NOTCH1-controlled biological functions in CLL, we next investigated the NOTCH1-dependent CLL transcriptional program. We used a CLL cell line (MO1043) (29), which carries a hemizygous NOTCH1 PEST-truncation and expresses a truncated ICN1 that is responsive to NOTCH inhibition by the γ-secretase inhibitor Compound-E (28). Given the low intrinsic signaling activity of PEST-truncated alleles alone (8) (Fig. S3A), we sought to investigate NOTCH1-dependent programs using an inducible lentiviral system expressing either an HA-tagged constitutively active form of NOTCH1 (ICN1-HA) or control eGFP upon doxycycline addition (Fig. S3B) (30). Induction of NOTCH1 signaling was demonstrated by HA immunoblot and quantitative RT-PCR (qRT-PCR) for DTX1, a well-established NOTCH1-direct target gene (11) (Fig. S3 C–E). This experimental system was used to identify genetic elements bound and directly regulated by NOTCH1 by integrating RNA-Seq and NOTCH1 ChIP-Seq data (Datasets S1–S3).

Fig. S3.

Establishment of ICN1-HA inducible MO1043 CLL cell line. (A) IB analysis of ICN1 and control β-actin in T-ALL cell lines (CUTLL1 and CEM) and MO1043 CLL cells treated with the γ-secretase inhibitor Compound E (CpE, 500 nM, 72 h) or control DMSO. (B) Diagrams of the pINDUCER vectors used to express ICN1-HA or control eGFP upon doxycycline addition (30). (C) IB analysis of HA (ICN1) and β-actin in MO1043-ICN1-HA whole cellular lysates (WCE) obtained upon doxycycline addition (1 μg/mL for 24 h, six independent inductions, a–f). (D) HA (ICN1) IB analysis of cytoplasmic (CE) and nuclear (NE) fractions of MO1043-ICN1-HA and control eGFP cells upon doxycycline addition (1 μg/mL for 24 h). β-Tubulin (β-tub) and Lamin-B1 IBs serve as controls for the purity of the subcellular fractions. (E) qRT-PCR analysis of DTX1 mRNA in MO1043-ICN1-HA cells upon doxycycline addition (1 μg/mL for 24 h, six independent inductions, a–f). Results are presented relative to those of induced MO1043-eGFP cells, set as 1. The bar graphs show the mean values, and the error bars represent the SD.

Unsupervised clustering of RNA-Seq data showed that the expression profiles of MO1043-ICN1-HA and -eGFP cells cluster separately (Fig. 3A). Supervised analysis revealed that ICN1-HA induction leads to up-regulation of over 700 transcripts [false-discovery rate (FDR) < 0.001, median fold-change 1.7, range = 1.1–170.4], including known NOTCH1 targets, such as HES/HEY family members, NRARP, DTX1, and NOTCH1 itself, as expected (Fig. 3B), as well as genes involved in immune and signaling pathways relevant for the development and activation of B cells (Datasets S4 and S5).

Fig. 3.

Identification of NOTCH1 direct targets in CLL. (A) Hierarchical clustering of RNA-Seq profiles of MO1043-ICN1-HA and -eGFP cells (Pearson correlation with average linkage, minimum log₂ expression 5 and minimum SD 1). (B) Scatter plot of log₂-transformed RNA-Seq FPKM values of differentially expressed genes between MO1043-ICN1-HA and -eGFP control CLL cells (FDR < 0.001). (C and D) Distribution of NOTCH1 binding sites (BS) in the genome of MO1043-ICN1-HA CLL cells. (E) Functional classification of NOTCH1-BS mapping to proximal promoters and distal regions of the genome based on their overlap with the H3K4me3, H3K4me, H3K27Ac and H3K27me3 histone marks. (F) Rank order of increasing H3K27Ac fold-enrichment at enhancer loci in in MO1043-ICN1-HA CLL cells. (G) Overlap between NOTCH1-BS and superenhancers identified with the ROSE algorithm (35, 36). (H) Representative examples of genes regulated by NOTCH1 via binding to superenhancer regions. (I) Intersection between RNA-Seq and ChIP-Seq data obtained in MO1043-ICN1-HA CLL cells. (J) Top three significantly (P = 1.00E-15) enriched transcription factor motifs lying ±200 bp of NOTCH1-BS. Abbreviations: NoExp, transcripts not expressed in MO1043-ICN1-HA cells; NoMov, transcripts not moving upon ICN1-HA expression; SEs, superenhancers; TF, transcription factor.

ChIP-Seq analysis identified a total of 4,737 NOTCH1 binding sites, mapping to promoters in ∼40% of the cases, and to intragenic or distal regulatory regions of the genome in ∼60% of the cases (Fig. 3 C and D). The integration of NOTCH1 ChIP-Seq profiles with those of the H3K4me3, H3K4me, H3K27Ac, and H3K27me3 histone modifications (Fig. S4) revealed that ∼94% of NOTCH1 proximal binding sites displayed chromatin marks characteristic of active promoters, whereas ∼37% of the distal ones were associated with putative active enhancers (Fig. 3E) (31–33). Analysis of H3K27Ac patterns across the genome identified 917 superenhancers (34), many of which involved genes defining key functions of B cells and displayed sequence motif enrichment of transcription factors known to be master regulators of B-cell identity (Fig. 3F, Fig. S5 A and B, and Datasets S6 and S7) (35–38). NOTCH1-binding at superenhancer regions (n = 698 of 4,737 binding sites, 14.7%) was prominently involved in the activation of genes implicated in B-cell differentiation and activation and antiapoptotic functions (Fig. 3 G and H and see below).

Fig. S4.

Distribution of histone modifications across the genome of MO1043-ICN1-HA cells. (A) Genomic distribution of histone marks-decorated regions with respect to the closest TSS (Left) and their distribution in the genome (Right). (B) Functional promoter status of genes based on the chromatin configuration of their TSSs and gene-expression levels (FPKM) of the corresponding transcripts (C) in MO1043-ICN1-HA CLL cells upon doxycycline addition (1 μg/μL for 24 h).

Fig. S5.

CLL superenhancers features. (A) Overlap of MO1043-ICN1-HA superenhancers with the H3K27me3 and H3K4me1 chromatin marks. (B) Expression levels (log₂ FPKM) of genes associated with superenhancers (SE) or regular enhancers (E) identified in MO1043-ICN1-HA cells. (C) Differential up-regulation of genes associated with NOTCH1 binding sites (BS) overlapping with superenhancer regions or located elsewhere in the genome (“other”) of MO1043-ICN1-HA cells. In A and C, P values are reported according to a two-tailed Fisher’s exact test. In B, the P value was calculated based on an unpaired unequal variance two-tailed Student's t test.

The intersection between the genes differentially expressed upon ICN1-HA induction as identified by RNA-Seq and the NOTCH1 binding profiles obtained through ChIP-Seq revealed a significant overlap between the two sets (P < 0.001), with ∼39% of genes induced by ICN1-HA (FDR < 0.001) being bound by NOTCH1 (Fig. 3I and Fig. S6). Notably, genes associated with NOTCH1 binding sites in superenhancer regions were more often up-regulated upon ICN1-HA induction compared with those associated with NOTCH1 binding sites in other genomic regions (52% vs. 29%, P < 0.001) (Fig. S5C). Motif enrichment analysis of sequences surrounding the NOTCH1 binding sites (±200 bp) and associated with significant (FDR < 0.001) up-regulation of the corresponding genes (n = 503) confirmed significant (P = 1.00E-15) enrichment of the DNA motifs of RBPJ, the main effector of NOTCH signaling (11), as well as of binding sequences of other potential cooperating cofactors, including NF-κB, PU.1, ETS, and STAT family members (Fig. 3J and Dataset S8).

Fig. S6.

Significant enrichment of top NOTCH1-bound genes in MO1043-ICN1-HA compared with control -eGFP cells. (A) GSEA enrichment plot depicting significant enrichment of a geneset composed by the top 400 NOTCH1-bound genes (i.e., top genes ranked based on ChIP-Seq P values) in MO1043-ICN1-HA CLL cells compared with -eGFP controls and (B) corresponding leading edge genes.

The NOTCH1-Dependent CLL Signature Is Detectable in NOTCH1–Wild-Type CLL Cases Expressing ICN1.

To identify bona fide direct NOTCH1 targets to be used as a tool to interrogate primary CLL cells (NOTCH1 CLL signature), we selected genes bound by NOTCH1 and the transcripts of which were up-regulated by ICN1-HA with a FDR < 0.001. Two-hundred and ninety-one genes met these criteria, including NOTCH1 target genes, such as HES1, DTX1, JAG1, and NOTCH1 itself, among others, as well as NF-κB, antiapoptotic, and cytokine-chemokine genes (Dataset S9).

To determine whether NOTCH1–wild-type ICN1⁺ CLL cases display evidence of NOTCH1 signaling activation analogous to NOTCH1-mutated cases, we explored the presence of the NOTCH1 CLL signature identified above in RNA-Seq data from 49 PB CLL samples fully characterized in terms of NOTCH1 mutation and ICN1 protein expression. This panel included 10 NOTCH1-mutated cases expressing ICN1 (including three cases carrying NOTCH1 3′UTR events), 13 NOTCH1–wild-type cases expressing ICN1, and 26 cases devoid of both NOTCH1 mutations and ICN1 expression (Fig. S2A). GSEA analysis (27) revealed significant enrichment of the NOTCH1 CLL signature both in NOTCH1-mutated (P = 0.002) and NOTCH1–wild-type (P = 0.04) cases expressing ICN1 compared with ICN1⁻ cases (Fig. 4A). Leading edge genes (n = 90) determining a significant (P = 0.03) enrichment of the NOTCH1 CLL signature in ICN1⁺ cases were overall expressed at similar levels in NOTCH1-mutated and wild-type ICN1⁺ cases [average fragments per kilobase of transcript per million mapped reads (FPKM) 54.3 and 51.3, respectively], with few genes expressed at higher levels in the NOTCH1-mutated ones (n = 9 of 90, P < 0.05) (Fig. 4B and Dataset S10).

Fig. 4.

The NOTCH1 CLL signature is enriched in primary CLL cases expressing ICN1. (A) GSEA enrichment plots depicting significant enrichment of the NOTCH1 CLL signature in NOTCH1-mutated (M) and wild-type (WT) primary CLL cases expressing ICN1⁺ (ICN1-pos) compared with ICN1⁻ (ICN1-neg) cases, and heatmap of RNA-Seq profiles of corresponding leading edge genes (n = 90) (B).

Thus, ICN1 is also functionally active in ICN1⁺ CLL cases devoid of NOTCH1 mutations, indicating that they are functionally equivalent in terms of NOTCH1-dependent transcriptional responses to NOTCH1-mutated ones (Discussion).

NOTCH1 Regulates Genes with Key Functions in B-Cell Physiology.

Functional annotation of the full set of genes bound (ChIP-Seq) and dynamically connected (RNA-Seq) to NOTCH1 revealed that NOTCH1 directly regulates general functions involved in cell proliferation and survival (Datasets S11 and S12). The former included CCND3, which encodes a cyclin necessary for G1/G2 transition (39) via direct binding to the gene promoter, consistent with a previous report in T-ALL (40). Among the latter, BCL2 and MCL1, two antiapoptotic genes with a well-established role in the pathogenesis of CLL, emerged as novel targets of NOTCH1, likely regulated through long-range dynamic interactions (24, 41–43).

The NOTCH1 transcriptional program included also a cadre of genes with specific functions in B-cell physiology (Fig. S7). Among these are BCR signaling pathway genes, including upstream pathway members (e.g., LYN, SYK, BLK, BLNK, CR2, and PIK3CD), as well as downstream effectors, such as MAPK (e.g., MAP3K1, KRAS, and RRAS) and NF-κB cascade members (e.g., IKBKB, NFKB1, and the CBM signalosome complex member MALT1) (44). NOTCH1 also appears to activate the NF-κB target NFKBIA, which encodes the NF-κB repressor IκBα (44), and PTPN6, encoding SHP-1, an important negative modulator of antigen-receptor signaling in lymphocytes (45), suggesting a role of NOTCH1 in a delayed negative-feedback of activation of this cascade (46). CXCR4, which encodes a chemokine receptor relevant for the chemotaxis of CLL cells toward microenvironmental cells expressing the CXCL12 ligand (47), emerged as a novel NOTCH1 target in CLL. This axis is fundamental for the exit of CLL cells from lymph nodes and, accordingly, the expression of the CXCR4 receptor has been shown to associate with a higher risk of lymphoid organ infiltration and poor disease outcome (48). Finally, among several NOTCH-pathway related genes, NOTCH1 induced the expression of JAG1, which encodes for a ligand of NOTCH1 reported to be expressed on the surface of CLL cells (22), suggesting a positive feed-forward loop in signaling activation.

Fig. S7.

Representative examples of NOTCH1-direct target genes. Representative ChIP-Seq plots depicting NOTCH1 binding and histone marking patterns at genes that are bound by NOTCH1 and up-regulated in MO1043-ICN1-HA CLL cells. The y axes in the ChIP-Seq plots indicate fragment density in reads per million (rpm).

NOTCH1 Transactivates MYC in CLL.

MYC is a central oncogene in human malignancy, an established NOTCH1 target in T-ALL and is involved in CLL progression (3, 13, 49–51). Thus, we investigated the relationship between NOTCH1 activation and MYC gene expression in CLL. Consistent with previous reports (52), we observed two putative superenhancers located ∼500 kb upstream of the MYC oncogene both in MO1043-ICN1-HA cells and primary CLL cases (Fig. 5 A and B). These superenhancers were also present in CD19⁺ and CD20⁺ B cells, small lymphocytic lymphoma, and mantle cell lymphoma (52), but not in the majority of ∼90 distinct tissue types (37), suggesting context specificity (Fig. 5C and Fig. S8A). NOTCH1 ChIP-Seq analysis of MO1043-ICN1-HA cells revealed the presence of multiple NOTCH1 peaks and RBPJK conserved motifs in this region. Specifically, one superenhancer within region 8q24 (chr8:128191039–128239723, hg19) contained one NOTCH1 binding site and two RBPJK motifs, whereas a second one (chr8:128299403–128321023, hg19), contained four NOTCH1 binding sites and three RBPJK motifs (Fig. 5B). Local ChIP analysis of NOTCH1 and the H3K27Ac mark in MO1043-ICN1-HA cells confirmed NOTCH1 binding at these epigenetically active superenhancers (Fig. 5D).

Fig. 5.

A NOTCH1-bound superenhancer region regulating MYC expression is recurrently duplicated in CLL. (A) NOTCH1 occupancy profiles and histone marks patterns in the 8q24 region encompassing the MYC locus (chr8:128000000–129000000, hg19) in primary CLL cases and MO1043-ICN1-HA cells, with corresponding peaks depicted in the box below the ChIP-Seq plots. The y axes in the ChIP-Seq plots indicate fragment density in reads per million (rpm). The two boxes below the called peaks represent segmentation data (7, 53) visualized using IGV (2.3.59), with red denoting a region of CN gain, blue a CN loss, and white depicting a normal (diploid) CN. Individual genes in the region are aligned in the Bottom panel. (B) Schematic representation of the distribution of superenhancers, NOTCH1 binding sites, and RBPJ motifs in the 8q24 region encompassing the MYC locus. (C) In the heatmap, rows correspond to normal or malignant B cells (52, 84) and two control T-ALL cell lines, and columns represent the two superenhancers identified in the 8q24 region encompassing the MYC locus, color-coded based on their presence or absence in the displayed cell type (light gray, absent; black, present). (D) ChIP-qPCR analysis of NOTCH1 and H3K27Ac at the MYC-associated superenhancer regions identified in MO1043-ICN1-HA CLL cells; results are presented relative to those obtained with IgG (IgG; control) and to a distal actin locus, set as 1. (E) qRT-PCR analysis of MYC and HES1 mRNA expression in three representative primary CLL cases, upon NOTCH1 signaling induction via coculture on stromal OP9-DL1 cells in the presence or absence of the γ-secretase inhibitor Compound E (CpE, 24 h, 1 μM, Left and Center), or upon basal NOTCH1 signaling inhibition in the presence of CpE (Right). Results are represented relative to those of CLL cells cocultured on OP9 stromal cells (Left), on OP9-DL1 stromal cells in the presence of CpE (Center), or with vehicle DMSO (Right), set as 1. The full set of analyzed primary CLL cases, including those represented here, is displayed in Fig. S9. The bar graphs in D and E show the mean values, and the error bars represent the SD between triplicates. Abbreviations: B-LCL, B-lymphoblastoid cell line; DLBCL, diffuse large B-cell lymphoma; MCL, mantle cell lymphoma; SE, superenhancer; SLL, small lymphocytic lymphoma.

Fig. S8.

Focal copy number gains recurrently affect the NOTCH1-bound 8q24 B-cell–specific superenhancer region in CLL. (A) Overlap between NOTCH1-bound 8q24 superenhancers observed in CLL and other tissues. In the heatmap, rows correspond to different tissues (normal and malignant) reported in the dbSUPER database (84) and columns represent the two superenhancers identified in the 8q24 region encompassing the MYC locus, color-coded based on their presence or absence in the displayed cell type (light gray, absent; black, present). (B) Frequency of NOTCH1 mutations and MYC CN gains (including gains encompassing only the MYC-associated superenhancer region) in a panel of 452 primary CLL cases, as reported in Puente et al. (7) and of MYC CN gains (including gains encompassing only the MYC-associated superenhancer region) in a panel of 353 primary CLL cases (53). (C) Graphic display of CN data from 30 patients harboring CN gains involving the 8q24 region encompassing the newly identified MYC-associated superenhancer regions in CLL. Segmentation data were visualized using IGV (2.3.59), where each track represents one sample, and white denotes a normal (diploid) CN, red a region of CN gain and blue a CN loss. Individual genes in the region are aligned in the Bottom panel, and the red boxed area highlights the minimal common region (MCR) of CN gain. In the bottom are highlighted the locations of NOTCH1 binding sites, RBPJK motifs (RBP_Jkappa V$RBPJK_Q4 and V$RBPJK_01 from the TRANSFAC database) and the superenhancers identified in CLL. (D) Heatmap showing the distribution of NOTCH1 mutations and MYC CN gains identified in n = 71/452 primary CLL cases, as reported in Puente et al. (7). In the heatmap, each column corresponds to a different case, and the two Bottom rows represent NOTCH1 mutations (M) and MYC alterations (act), color-coded based on their presence or absence in the displayed case (light gray, absent; black, present). The Top row shows the IGHV mutational status of the displayed cases (M, mutated; NA, not available; UM, unmutated).

Intriguingly, our meta-analysis of copy number (CN) data obtained in two independent collections of CLL patients (7, 53) revealed that this region is recurrently affected by duplications in CLL (n = 15 of 452 and n = 15 of 353), including focal ones specifically encompassing this superenhancer cluster (Fig. 5A and Fig. S8 B and C), suggesting that these genetic lesions may affect MYC expression in CLL. Moreover, we observed that NOTCH1 mutations and MYC locus duplications, including those focally affecting only this superenhancer region, were largely mutually exclusive in primary CLL patients (7) (Fig. S8D), similarly to CLL cells that have undergone Richter syndrome transformation (3, 13).

To demonstrate a direct functional relationship between NOTCH1 and MYC in CLL, we first investigated whether NOTCH1 activation could promote MYC transcription in ICN1⁻ or ICN1-low primary CLL cells cocultured with stromal cells expressing the NOTCH1 ligand Delta-1 (OP9-DL1) (54). CLL cells cocultured with OP9-DL1 cells showed an increased level of MYC RNA, which was specifically dependent upon NOTCH1 expression because it was abrogated in the presence of the NOTCH-inhibitor Compound E (Fig. 5E and Fig. S9). Reciprocally, NOTCH1 inhibition by Compound E induced MYC down-regulation in ICN1-positive CLL cases (Fig. 5E and Fig. S9). These results indicate that NOTCH1 controls MYC expression in ICN1⁺ CLL cells.

Fig. S9.

MYC RNA levels are responsive to modulation of NOTCH1 signaling activation in primary CLL cases. qRT-PCR analysis of MYC and HES1 mRNAs expression in primary CLL cases upon ICN1 induction via coculture on stromal OP9-DL1 cells in the presence or absence of the γ-secretase inhibitor Compound E (CpE, 1 μM, 24 h, top five graphs), or upon basal NOTCH1 signaling inhibition in the presence of CpE (Bottom graph). Results are represented relative to those of CLL cells cocultured on OP9 stromal cells, on OP9-DL1 stromal cells in the presence of CpE, or with vehicle DMSO, set as 1. ICN1-positive cases depicted in the Bottom panel (n = 6) include 3 NOTCH1-mutated and 3 NOTCH1–wild-type cases.

Discussion

CLL cells in the lymph node are known to display frequent NOTCH1 activation independent of mutation, as documented by their frequent immunohistochemical positivity for ICN1 in lymph node sections of both NOTCH1-mutated and wild-type cases (19, 20). Conversely, few reports have shown activation of NOTCH1 in PB CLL samples, and the results involved a relatively low number of cases and a possibly low sensitivity of detection, leading to a significant underestimation of the ICN1⁺ cases (4, 7, 18, 22). Our finding that ∼50% of PB CLL cases lacking NOTCH1 mutations express ICN1 strongly suggests that the activation of this pathway is more common than what is predicted by the frequency of classic NOTCH1 PEST-truncations. These findings have implications for the mechanisms leading to NOTCH1 activation in normal and transformed cells, for the understanding of the pathogenesis of CLL, and the development of anti-NOTCH1 targeted therapies.

Mechanisms of NOTCH1 Activation in PB CLL Cells.

The common detection of ICN1 in the nucleus of CLL cells within lymph nodes in both NOTCH1-mutated and wild-type cases has been interpreted as an induction by microenvironmental interactions with cells expressing NOTCH1 ligands (19, 20). Conversely, the frequent detection of ICN1 in NOTCH1–wild-type PB CLL cells shown herein raises the issue, shared also by NOTCH1-mutated cases carrying activation-dependent alleles, of the mechanisms leading to signaling induction. Although it is plausible that CLL cells may continuously recirculate between the PB and secondary niches, such as the lymph nodes, thus being exposed to signals triggering NOTCH1 activation, several observations seem to disfavor the possibility that the ICN1 presence observed in these cells is because of residual signaling form the nodal environment. It has been shown that perinodal CLL cells rapidly lose nuclear ICN1 expression once they move beyond the lymph node capsule (19). The half-life of ICN1 is short and variable, from less than an hour to a few hours in several tested cell types (55), including primary CLL cells, but it is significantly shorter than the long life of CLL cells in the PB, which has been estimated to be of at least several days (56). These observations, together with the fact that ICN1 is expressed in virtually all CLL cells in ICN1⁺ cases (Fig. 2C), are consistent with the continuous induction of the NOTCH1 cascade in CLL cells in the blood stream. However, future studies should compare ICN1 levels in the peripheral blood and in the nodal CLL compartment of the same individual to achieve a better understanding of the dynamics of ICN1 expression in different anatomic compartments within the same CLL patient.

It is conceivable that sustained NOTCH1 activation in the PB may be mediated by cell-autonomous and ligand-dependent mechanisms. Among the former, activation of alternative cryptic NOTCH1 promoters has been reported in human and, more frequently, murine T-ALL (57–59). However, this mechanism was preliminarily excluded by the analysis of H3K27Ac ChIP-Seq data in a panel of nine CLL cases, all of which displayed high levels of the H3K27Ac mark decorating only the canonical 5′ transcriptional start site (TSS) of the NOTCH1 gene (Fig. S10). Additional ligand-independent mechanisms include activation by other signaling pathways, as reported for the T-cell receptor pathway in T cells (60), or aberrant vesicle trafficking of the NOTCH1 receptor (61). Ligand-dependent mechanisms include binding to ligands expressed by endothelial cells in the blood vessels or by the CLL cells themselves (18, 22).

Fig. S10.

H3K27Ac ChIP-Sequencing plots at the NOTCH1 locus in nine primary CLL cases. The y axes in the ChIP-Seq plots indicate fragment density in reads per million. M, mutated; WT, wild-type.

Role of NOTCH1 in B-Cell Development and CLL Pathogenesis.

Our results indicate that NOTCH1 displays a stage-specific expression pattern in mature B cells, being expressed and activated in naïve and memory B cells, which are considered the cells of origin of CLL (1). The enrichment of our NOTCH1 CLL signature in these normal subpopulations compared with GC B cells suggests that the biological programs orchestrated by NOTCH1 in CLL are similar to those already active in the putative normal B-cell counterparts of the disease. Thus, the finding of NOTCH1 activation in CLL cells reflects the constitutive, dysregulated expression of a physiologic signal and its corresponding gene-expression program rather than an ectopic program associated with transformation. The set of NOTCH1-direct transcriptional target genes suggests a broad program aimed at promoting the survival and proliferation of mature B cells by supporting BCR and cytokine signaling and their downstream effectors, such as PI3K and NF-κB pathways. NOTCH1 direct targets specifically relevant for the B-cell phenotype appear to be regulated via direct activation of the corresponding promoters, as well as via long-range interactions occurring at superenhancer sites, consistent with the role of these large regulatory elements in orchestrating the expression of cell-type specific genes (37).

Role of NOTCH1-Induced MYC Expression.

An important component of the NOTCH1-controlled program in CLL cells is the transactivation of the MYC oncogene. Our data suggest that this transactivation is mediated by the binding of NOTCH1 to B-cell–specific superenhancers located ∼500 kb upstream of the MYC locus. This region interacts with the MYC promoter in small lymphocytic lymphoma and mantle cell lymphoma, and also leads to MYC transcriptional activation in Epstein–Barr-transformed lymphoblastoid cells as a result of Epstein–Barr virus nuclear antigen 2 binding (52, 62). The focal recurrent duplications of this locus observed in CLL (7, 53), analogous to what is observed in other malignancies in which CN gains affect the tissue-specific enhancers involved in MYC expression (50, 63, 64), suggest that the gain of context-specific superenhancers represents a common mechanism for up-regulating MYC expression in distinct tumor types. However, the detection of this superenhancer cluster also in CLL cases devoid of ICN1 expression (Fig. 5A), and the observation that MYC mRNA levels were not significantly different between ICN1⁺ and ICN1⁻ CLL cases indicate that other transcription factors are likely involved in the regulation of this locus in B cells, as previously suggested for EBF and RELA in Epstein–Barr-transformed lymphoblastoid cells (62).

Clinical Implications.

The observation that the NOTCH1 CLL signature is enriched in ICN1⁺ CLL cases independent of NOTCH1 mutation significantly increases the fraction of CLL cases that may be dependent on constitutive NOTCH1 activity. NOTCH1 mutations are known to associate with adverse CLL clinical and biological features, including an unmutated IGHV status (12), and predict a poor outcome when found in CLL patients at diagnosis (12). Conversely, our results showed that ICN1 expression in cases devoid of NOTCH1 mutations occurred at similar frequencies in IGHV-mutated and IGHV-unmutated CLL cases. However, the relatively small and heterogeneous cohort of patients analyzed in this study did not provide us with the statistical power of establishing reliable correlations between ICN1 expression and the clinical course of the disease. Thus, dedicated prospective clinical studies are warranted to assess the biological and prognostic associations of ICN1 expression rather than NOTCH1 mutations in CLL alone, especially in the context of anti-CD20–based therapies. This analysis may allow further refinement of the recent mutation/cytogenetic hierarchical model of classification of patients with CLL in distinct risk classes (65). Finally, we propose that ICN1 expression may also represent a more reliable biomarker of NOTCH1 activation in the testing of prognostic criteria and therapeutics agents targeting NOTCH1 (28, 66).

Materials and Methods

Cell Lines and Isolation of Human B-Cell Subsets.

MO1043 cells (29) were cultured in Iscove's Modified Dulbecco's Medium (Life Technologies) supplemented with 20% (vol/vol) FBS (Sigma-Aldrich), penicillin (100 U/mL), and streptomycin (100 μg/mL). The identity of the cell line was verified by CN analysis using the Genome-Wide Human SNP Array 6.0 (Affymetrix), as previously reported (3). HEK 293T cells were cultured in Dulbecco’s Modified Eagle Medium (Life Technologies) with 10% (vol/vol) FBS, 100 U/mL penicillin, and 100 μg/mL streptomycin. OP9 and OP9-DL1 cells were grown in Minimum Essential Medium Alpha Medium (Corning) supplemented with 20% (vol/vol) FBS, 100 U/mL penicillin, 100 μg/mL streptomycin, and 2 mM glutamine (Thermo Fisher Scientific) (54). Human GC B cells, naive B cells, and memory B cells were isolated from reactive tonsils as described previously (23). Compound E was obtained from Enzo Life Sciences and used at a final concentration of 500 nM to 1 μM in DMSO vehicle.

Protein Extraction and Immunoblot Analysis.

Whole-cell extracts were obtained using Nonidet P-40 lysis buffer (150 mM NaCl, 1.5% (vol/vol) Nonidet P-40, 50 mM Tris⋅HCl pH 8.0, 2 mM EDTA pH 8.0) supplemented with proteinase inhibitor mixture (Sigma-Aldrich), according to a previously described protocol (67). Protein lysates were resolved on 4–12% Tris-Glycine gels (Novex, Life Technologies). Subcellular fractionation was performed as previously described (68). Samples were incubated with primary antibodies overnight at 4 °C. The following primary antibodies were used: rabbit monoclonal anticleaved NOTCH1 (clone D3B8, Cell Signaling Technology), mouse monoclonal anti-MYC (clone 9E10, Santa Cruz), mouse monoclonal anti-BCL6 (clone GI191E/A8, Cell Marque), rabbit monoclonal anti-HA (clone C29F4, Cell Signaling Technology), mouse monoclonal anti–β-actin (clone AC-15, Sigma), rabbit polyclonal anti–β-tubulin (H-235, Santa Cruz). Horseradish peroxidase-conjugated secondary antibodies and ECL or West Dura reagent (Thermo Fisher Scientific) were used for signal detection.

Immunofluorescence Analysis.

Immunofluorescence analysis of ICN1, AID, and CD20 was performed on formalin-fixed paraffin-embedded material from human tonsils and primary peripheral blood CLL cells according to standard procedures using the following antibodies: rabbit monoclonal anticleaved NOTCH1 (clone D3B8, Cell Signaling Technology), rat monoclonal anti-AID (clone mAID-2, eBioscience), and mouse monoclonal anti-C20 (clone L26, Thermo Fisher) (69).

ChIP.

MO1043-ICN1-HA cells (cells induced to express ICN1-HA with 1 μg/mL of doxycycline for 36 h) were cross-linked with 1% formaldehyde for 10 min at room temperature, quenched by the addition of glycine to a final concentration of 0.125 M and frozen at −80 °C. Cross-linked chromatin was fragmented by sonication with the Bioruptor sonicator (Diagenode) to achieve fragment sizes of ∼200–500 bp using the following sonication buffer: 10 mM Tris⋅HCl pH 8.0, 100 mM NaCl, 1 mM EDTA pH 8.0, 0.5 mM EGTA pH 8.0, 0.1% Na-Deoxycholate, and 0.5% N-lauroylsarcosine. Sheared chromatin was incubated overnight with 10 μL of NOTCH1 antisera (70), 4 μg of antibodies to H3K27Ac (Active Motif, cat#39133) or H3K4me3 (Abcam, cat#ab8580), or 2 μg of antibodies to H3K4me1 (Abcam, cat#ab8895) or H3K27me3 (Active Motif, cat#39157) (ENCODE Project: genome.ucsc.edu/ENCODE/antibodies.html). Protein A magnetic beads were added for 4 h at 4 °C, followed by sequential washes at increasing stringency and reverse cross-linking. After RNase and proteinase K treatment, ChIP DNA was purified using the MiniElute Reaction Clean Up Kit (Qiagen) and quantified by Quant-iT PicoGreen dsDNA Reagent (Life Technologies). Validation via quantitative PCR analysis (qChIP-PCR) was performed with the Power SYBR green PCR Master Mix using the 7300 Real Time PCR system (Applied Biosystems). Oligonucleotides used for qChIP-PCR are listed in Dataset S13.

ChIP-Seq Library Preparation and Illumina Sequencing.

ChIP-Seq libraries were constructed starting from 4 ng of ChIP or Input DNA as reported in Blecher-Gonen et al. (71). Libraries were quantified using the KAPA SYBR FAST Universal qPCR Kit (KAPA Biosystems), normalized to 10 nM, pooled, and sequenced with the Illumina HiSEq. 2000 instrument as single-end 100-bp reads.

ChIP-Seq Analysis.

Sequencing data were acquired through the default Illumina pipeline using Casava v1.8. Reads were aligned to the human genome (UCSC hg19) using the Bowtie2 aligner v2.1.0 (72), allowing up to two mismatches to cope with human variation. Duplicate reads (i.e., reads of identical-length mapping to exactly the same genomic locations) were removed with SAM tools v0.1.19 using the rmdup option (73). Read counts were normalized to the total number of reads aligned in a library (reads per million). Peak detection was done using the ChIPseeqer v2.0 algorithm (74), enforcing a minimum fold-change of 2 between ChIP and input reads, a minimum peak width of 100 bp, and a minimum distance of 100 bp between peaks. The threshold for statistical significance of peaks was set at 10⁻⁵ for NOTCH1, and 10⁻¹⁵ for H3K4me1, H3K4me3, and H3K27Ac (Dataset S2). Peaks within 1 kb of centromeric or telomeric regions were removed. H3K4me1 and H3K27Ac peaks were stitched together into regions if located within ±2 kb and ±12.5 kb of each other, respectively, unless they started within a 2-kb window around the TSS. H3K27me3 peaks were called using the RSEG algorithm (75) with 100-bp bin size; only peaks above 5 kb in size were considered.

Motif Enrichment Analysis.

Regions within 200 bp of the center of each binding site were searched for motifs from the TRANSFAC 2010 Database. Motifs were represented as position weighted matrices. Using a moving window, motifs were scored against a reference DNA sequence using a log odds ratio comparing the motif’s score to a hypothetical score where every base is equally probable. Motifs scoring higher than a given threshold were considered as potentially bound in a location. To determine enrichment in a set of peaks/locations a hypergeometric model was used comparing motifs bound in the peaks to a GC and length controlled set of random genomic sequences.

Definition of Functional Chromatin States of NOTCH1-Bound Genomic Loci.

Significant NOTCH1-bound regions occurring at proximal promoters (i.e., within −2/+1 kb from the TSS of an annotated gene) were classified as active if overlapping with H3K4me3, but not H3K27me3, poised if occupied by both H3K4me3 and H3K27me3, and silenced if decorated only by H3K27me3. Distally NOTCH1-bound genomic regions (intergenic or intragenic) were classified as active enhancers if occupied by H3K4me and H3K27Ac, but not H3K4me3, poised if occupied by H3K4me and H3K27me3, and silenced or primed if occupied only by H3K27me3 or H3K4me, respectively. For the identification of superenhancers, we applied the ROSE algorithm (https://bitbucket.org/young_computation/rose) to our H3K27Ac ChIP-Seq datasets (MO1043-ICN1-HA and primary CLL cells). Occupancy of NOTCH1 at superenhancers was then determined based on the overlap between NOTCH1 peaks and genomic regions identified by ROSE. NOTCH1-bound superenhancers were assigned to the nearest expressed and transcriptionally active gene (i.e., distance from superenhancer center to TSS marked by H3K4me3) as the most likely candidate target gene (38).

Primary CLL Cases.

Primary CLL cells from the PB of CLL patients (n = 124) were obtained from the Feinstein Institute for Medical Research and the Division of Hematology and the Department of Translational Medicine and the Amedeo Avogadro University of Eastern Piedmont. Diagnosis of CLL was based on International Workshop on Chronic Lymphocytic Leukemia-National Cancer Institute Working Group criteria (76) and confirmed by a flow cytometry score >3. The percentage of tumor cells of CLL cases was estimated by cytofluorimetry analysis for CD5⁺/CD19⁺ PB cells of 93 of 124 CLL cases, and it was ≥70% in 90 cases and between 57% and 60% in 3 cases. CLL cases included in the RNA-Seq panel are highlighted in Fig. S2A. ICN1⁺ NOTCH1–wild-type cases were selected for RNA-Seq analysis based on a ratio of ICN1 expression > 0.1 compared with the levels observed in the CUTLL1 cell line in the low-exposure immunoblot image. Quantitation of signal intensity was obtained with the ImageJ software (https://imagej.nih.gov/ij/) by subtracting the background signal measured above each band from the signal measured in each band; areas of the same size (set on the image of ICN1 in the CUTLL1 cell line) were used for all measurements. Values were expressed as ratio relative to the CUTLL1 protein sample, set at 1, after normalization for the β-actin loading control. The study was approved by the Institutional Review Board of Columbia University, by the Ethical Committee of the Azienda Ospedaliera Maggiore della Carità di Novara, Amedeo Avogadro University of Eastern Piedmont, and by the Northwell Health’s Institutional Review Board and was conducted according to the principles of the World Medical Association Declaration of Helsinki.

DNA Extraction, IGHV Mutational Status, and Sanger Sequencing of NOTCH1.

Genomic DNA was extracted with the QIAamp DNA Mini Kit (Qiagen) and verified for integrity by gel electrophoresis. IGHV mutational status was performed as previously described (3, 13). The NOTCH1 gene portion encoding the PEST domain and the 3′UTR of the NOTCH1 gene were screened by Sanger targeted sequencing, as previously reported (3, 7).

Gene-Expression Profiling of Human Mature B-Cell Subsets.

Raw expression values of GeneChip Human Genome U133 Plus 2.0 (Affymetrix) data from normal mature B-cell subsets were normalized using the Robust Multiarray Averaging algorithm in GenePattern (https://www.broadinstitute.org/cancer/software/genepattern/), and multiple probes corresponding to the same gene were collapsed to a single probe based on the maximum t-statistic/maximum SD.

RNA Extraction, cDNA Synthesis, and Quantitative Real-Time PCR.

Total RNA was extracted from primary CLL cases with the RNeasy Mini Kit (Qiagen) with on-column DNase treatment. cDNA synthesis was performed using the SuperScript First-Strand Synthesis System (Life Technologies), according to the manufacturer's instructions. The ABsolute QPCR SYBR green mix (Thermo Scientific) was used to amplify specific cDNA fragments with the oligonucleotides listed in Dataset S13, in the 7300 Real-Time PCR system (Applied Biosystems). Data were analyzed by the change-in-threshold (2^−ΔΔCT) method (77), using GAPDH as a housekeeping reference gene.

RNA-Sequencing of ICN1-HA and Control eGFP MO1043 Cells and Primary CLL Cases.

Four MO1043-ICN1-HA and 4 MO1043-eGFP replicates and 49 primary CLL cases were subjected to RNA-Sequencing. Briefly, poly-A pull-down was performed to enrich mRNAs from total RNA samples and libraries were prepared using the Illumina TruSeq RNA prep kit and sequenced using the Illumina HiSeq.2000 instrument at the Columbia Genome Center. MO1043-ICN1-HA and MO1043-eGFP samples were multiplexed to obtain an average of 33,022,292 single-end 100-bp reads per sample; primary CLL samples were multiplexed to obtain an average of 61,266,691 paired-end 100-bp reads per sample (Dataset S3). Real-time analysis (Illumina) was used for base calling and bcl2fastq (v1.8.4) for converting BCL to fastq format, coupled with adaptor trimming. Reads were mapped to the reference genome (Human: NCBI/build37.2) using Tophat (v2.0.4) with 4 mismatches (–read-mismatches = 4) and 10 maximum multiple hits (–max-multihits = 10) (78). The relative abundance of genes was assessed using cufflinks (v2.0.2) with default settings (79). Hierarchical clustering of MO1043-ICN1-HA and -eGFP profiles was performed using the Pearson correlation average linkage, filtering for genes with a minimum log₂-transformed expression value of 5 and a minimum SD of 1. Genes differentially expressed between MO1043-ICN1-HA and -eGFP profiles were determined by an unpaired unequal variance two-tailed Student’s t test using a FDR ≤ 0.001 (after Benjamini–Hochberg correction) (80). For visualization of gene-expression intensity, expression data were converted to z-scores.

GSEA.

Gene-expression profile data from mature B cells, MO1043-ICN1-HA and MO1043-eGFP cells, and from primary CLL cases were analyzed for enrichment in NOTCH1-related gene sets with GSEA-2.0 and 1,000 phenotype permutations (27). Enrichments were considered significant with a P < 0.05 after correction for multiple hypothesis.

Functional Categories and Pathways Analyses of the NOTCH1-Regulated Genes.

Genes directly regulated by NOTCH1 in CLL were assigned to functional categories or annotated pathways using the publicly available bioinformatic tool DAVID 2008 6.7 (Database for Annotation, Visualization and Integrated Discovery, https://david-d.ncifcrf.gov) and the Molecular Signatures Database from the Broad Institute (MSigDBv5.1, CP Geneset, https://www.broadinstitute.org/gsea/msigdb/index.jsp). Only pathways relevant for B-cell biology, based on current knowledge, were selected for further discussion.

Statistical Analyses.

Statistical analysis was performed using the GraphPad Prism 5 software (GraphPad Software). The specific test adopted for each analysis is described in each figure legend.