NOTCH1 drives immune-escape mechanisms in B cell malignancies

NOTCH1 is a recurrently mutated gene in Chronic Lymphocytic Leukemia (CLL) and Mantle Cell Lymphoma (MCL). Functional studies to investigate its role have been hampered by the inability to genetically manipulate primary human lymphoma cells, attributed to low transduction-efficacy and procedure-associated toxicity. To overcome these limitations, we have developed a novel method to retrovirally transfer genes into malignant human B cells. We generated isogenic human tumor cells from patients with CLL and MCL, differing only in their expression of NOTCH1. Our data demonstrate that NOTCH1 facilitates immune-escape of malignant B cells by up-regulating PD-L1, partly dependent on autocrine interferon-γ signaling. In addition, NOTCH1 causes silencing of the entire HLA-class II locus via suppression of the transcriptional co-activator CIITA. These NOTCH1-mediated immune escape mechanisms are associated with the expansion of CD4+ T cells in vivo, further contributing to the poor clinical outcome of NOTCH1-mutated CLL and MCL.


Introduction
The landscape of somatic mutations present in malignant B cells from patients with CLL has been described in several pivotal sequencing studies, identifying more than 40 recurrent mutations mostly affecting oncogenes (NOTCH1, Wnt-signalling), tumor suppressors (TP53, ATM) and genes involved in RNA-processing (SF3B1, XPO1, RPS15) [1][2][3][4] . While the prognostic significance is known for some of these mutations, their specific contributions to the pathogenesis of the disease remains largely unknown. Attempts to address this experimentally have employed genetically engineered mouse models (GEMMs), most commonly the Eµ-TCL1 mouse model 5 , and human cell lines. While the former model proved to be useful in particular to recapitulate disease aspects which can only properly be studied in vivo (such as tumor-microenvironment interactions), significant limitations exist which prevent extrapolation of data from mice to human. In addition, the experimental manipulation of primary tumor cells from the Eµ-TCL1-model remains technically challenging and most commonly requires crossing of different GEMMs, which consumes time and resources. In contrast, cell lines are easy to manipulate and provide a sheer unlimited and immediate access to tumor cells. However, because they are commonly obtained from patients with end-stage, refractory diseases, cell lines frequently are EBV-positive 6,7 and often have been selected for decades to grow in minimal culture conditions, therefore losing the biological identity they are supposed to represent. Vigorous proliferation, absence of spontaneous apoptosis and aberrant homing in NSG mice are some examples for these discrepancies, limiting the conclusions one can draw from such experiments.
Few studies have used adenovirus vectors or their derivates to genetically manipulate primary CLL cells 8 . However, the lack of integration into the host genome results in only transient expression of a gene-of-interest (GOI) and precludes from studying effects in dividing cells or subsequent use of cells in in vivo studies. Alternative attempts using retro-or lentivirus vectors for gene transfer have been unsuccessful for decades, limited by low transfection efficacy (<1%) and substantial toxicity 9 , which made down-stream analyses impossible. To overcome these limitations, we have developed a novel method to effectively infect primary neoplastic human B cells. This method permits gene transfer with high transduction efficacy and minimal toxicity, which allows the functional investigations of genes recurrently mutated in primary malignant B cells. We used patient samples from CLL and MCL, two mature B neoplasms with partially overlapping biological features and clinical behaviors 10 that lack appropriate in vitro or in vivo models that span their clinico-biological spectrum.
We have employed this technique to interrogate the molecular functions of NOTCH1, which is one of the most commonly mutated genes in CLL and associated with a poor clinical outcome and a high frequency of Richter's transformation [11][12][13] . Most NOTCH1-activating mutations affect the PEST-domain, encoded by exon 34 14 . In addition, point mutations in the 3'-UTR have been identified and cause expression of a truncated protein 2 . Both scenarios result in an abnormally stable NOTCH1 protein, which continues to require ligand-binding in order to become transcriptionally active. Several groups have employed cell lines to study the biology of NOTCH1 in B cell malignancies and then associated these findings with data from primary cells [15][16][17][18] . While such approaches provided important insights into the role of NOTCH1 in CLL, it often remains unclear whether these findings report a direct consequence of activated NOTCH1 or are a mere correlation. Our method to retrovirally infect primary malignant B cells from patients with CLL and MCL to generate isogenic cells provides a unique opportunity to answer this question. Here we provide evidence of how NOTCH1 favors immune escape of tumor B cells and we address how cells with trisomy 12 may provide a selective advantage for NOTCH1-mutations.

Retroviral transduction into primary human tumor B cells permits functional downstream analyses
In the past, transduction of CLL B cells using retroviral or lentiviral vectors has largely been unsuccessful mainly due to the fact that cells ex vivo are prone to apoptosis and are arrested in the G0/G1 phase of the cell cycle 9 . In order to overcome these limitations, we generated three different de-novo retrovirus-based envelopes able to infect mammalian cells of different tissue origins. Additionally, we engineered a new stroma-cell line from a subset of CD45 neg , Lin neg , Sca1 pos -bone marrow derived stromal cells, hereafter called MM1-cells. We initially tested different viral envelopes to genetically manipulate proliferating CLL cells, which were continuously cultured on MM1-cells after transduction. Initial attempts to transduce cells were unsuccessful due to a persistent high rate of apoptosis of also proliferative cells and required further manipulation of MM1-cells. MM1-cells constitutively expressing 3 additional prosurvival factors fully antagonized procedure-associated toxicity and permitted the successful transduction of primary CLL cells. While primary malignant B cells were effectively transduced with Env1 (24% in CLL and 12% in MCL), the Env2 infected a higher percentage of CLL and MCL cells with an average efficacy of 37% and 39%, respectively. In contrast, CLL cells were resistant to infections with the Env3, which displayed moderate efficacy to infect primary MCL cells (Fig. 1a). Importantly, our transfection method was not associated with increased cell death: after 7 days, CLL cell viability was >80% (Fig. 1b), similar to non-transduced, previously cryopreserved cells and cultured under identical conditions ( Supplementary Fig. 1a).
To test that transduced cells remained functional to allow down-stream analyses, cells were transduced with a dominant-negative TP53 (p53DD) and subsequently exposed to the anthracycline doxorubicin for 12 hours. P53DD significantly mitigated p21mRNA transcription induced by doxorubicin in cells carrying wild-type TP53, but significantly less so in p53deficient cells (Fig. 1c). Similarly, p53DD reduced Fludarabine-induced apoptosis (Fig. 1d). In addition to the interference with a major tumor suppressor, we overexpressed the protooncogene c-MYC in primary malignant B cells from CLL patients. Expectedly, ectopically expressed c-MYC induced a robust proliferative response in CLL cells, indicated by the increased number of cells transitioning through S-phase (Fig. 1e).
In conclusion, we have established a method to effectively infect primary tumor cells from patients with CLL and MCL with minimal toxicity, allowing to generate isogenic, patient specific tumor cells which differ only in the GOI.

NOTCH1 drives proliferation, CD38-expression and enhances B cell receptor signaling
We next used this method to investigate the role of NOTCH1 in primary CLL cells, which is frequently mutated in approximately 10% of untreated patients 14,19 . In order to simultaneously account for point-mutations, missense and frameshift mutations affecting exon 34 and for less common 3'-UTR NOTCH1 mutations, we expressed the coding sequence of NOTCH1-ICD, but lacking the PEST-domain, followed by IRES-GFP (hereafter named NOTCH1 DPEST ; Fig.   2a) in primary CLL cells. Importantly, since NOTCH1 DPEST also lacked the ectodomain, activation did not require binding of NOTCH ligands or cleavage of the extracellular domain for transactivation. Since CLL cells co-express NOTCH1 and NOTCH2 20 , this approach allowed us to investigate the function of NOTCH1 in isolation without simultaneously activating other NOTCH-receptors. To test that NOTCH1 DPEST was transcriptionally active, we first assessed the mRNA expression levels of HES1, HEY1 and DTX1, bona-fide NOTCH-target genes. Compared to empty vector (EV) control, NOTCH1 DPEST (N1 DPEST ) increased the abundancy of HES1, HEY1 and DTX1 mRNA by 3.0-, 3.4-and 3.6-fold, respectively (Fig. 2b).
Importantly, similar expression changes of HES1 and DTX1 were reported in ligand activated CLL cells carrying NOTCH1 mutations 17,18,21,22 , indicating that NOTCH1 DPEST has a similar activation potential than mutated, endogenous NOTCH1.
Several previous studies have suggested that CLL cells carrying NOTCH1 mutations have a proliferative advantage compared to wild-type CLL cells 16,22 . To test whether NOTCH1 is involved in cell cycle regulation, we assessed the number of GFP + cells in S-phase 6 days 6 after transduction. Compared to EV-control cells, primary CLL cells expressing NOTCH1 DPEST consistently showed a higher percentage of cells going through S-phase (Fig. 2c).
Accordingly, the fraction of GFP-positive CLL cells continuously increased only in the NOTCH1 DPEST -transduced cells but not in EV-controls, indicating that NOTCH1 positively affects cell cycle progression (Fig. 2d).
NOTCH1 mutations have also been associated with surface CD38 expression 23 . In keeping with the observation that CD38 expression is higher in CLL lymph nodes compared to peripheral blood cells 24 , co-culture of CLL cells on MM1-cells increased baseline expression of CD38 (data not shown). NOTCH1 DPEST consistently up-regulated CD38 expression further, suggesting that its expression is functionally dependent on NOTCH1 activation (Fig. 2e). Of note, NOTCH1 DPEST expression did not induce the expression of CD138, BLIMP1 or IRF4 ( Supplementary Fig. 1b,c), indicating that CLL B cells did not differentiate into antibodysecreting plasma cells 25 . Other studies have associated NOTCH1 mutations to the expression of CD49d, which equally and independently indicates a poor prognosis. In agreement with a previous study on MEC-1 cells 18 , we observed that NOTCH1 DPEST induced surface expression of CD49d only in non-trisomy 12 patients, which overall expressed much higher levels of CD49d (Fig. 2f). Lastly, we assessed B cell receptor (BCR)-responsiveness of CLL cells transduced with NOTCH1 DPEST or EV. Anti-IgM induced a stronger calcium-flux in cells expressing NOTCH1 DPEST compared to EV cells (Fig. 2g), supporting a recent study which demonstrated collaboration between NOTCH-and BCR-signaling 21 . In conclusion, retrovirally expressed NOTCH1 DPEST has biological activities similar to mutated NOTCH1 and recapitulates functions previously attributed to NOTCH-activation.

Patients with trisomy-12 or del13q present a common NOTCH1 transcriptome
To define the global transcriptional program controlled by NOTCH1 and contributing to these phenotypic and proliferative effects, we performed RNA-seq on 13 primary CLL samples, either transduced with an EV-control or NOTCH1 DPEST . Only patients who had a deletion of chromosome 13q (del13q) or carried an additional copy of chromosome 12 (tri12), assessed by conventional FISH-analyses, were included. Pairwise analysis of all 13 patient samples identified 1636 differentially expressed genes of which 979 were upregulated and 657 were downregulated (applying a cut-off of Log2FC>0.5, present in at least half of all samples) (Fig.   3a). Gene Set Enrichment Analysis (GSEA) of these differentially expressed (DE) genes identified gene clusters in canonical Notch-signaling (e.g., HES4, SEMA7A, CD300A and DTX1), B cell activation/ BCR-signaling (e.g., FYN, BLNK and CR2) and MAPK-activation (e.g., MAPK8 and MAP2K6), in keeping with previous reports based on the ectopic expression of NOTCH1 in lymphoma cell lines 15,16 . Unexpectedly, NOTCH1 repressed genes were strongly enriched in antigen-processing and presentation, predominately belonging the family of MHC class II genes (Fig. 3b).
While NOTCH1 mutations have been identified in less than 15% of treatment naïve, unselected patients 1,19 , this frequency is significantly higher in patients with trisomy12, in which NOTCH1 mutations can be found in 40-50% of patients 26,27 . Importantly, the presence of NOTCH1 mutations in tri12-CLL is associated with high rates of transformation into Richter's syndrome 11,28 . The underlying reasons for this peculiar association are unknown, but it suggests that NOTCH1 may regulate distinct, transformation-favoring genes in cells carrying an extra chromosome 12. To address this hypothesis, we separately analyzed the gene expression profiles induced by NOTCH1 in 7 tri12 and 6 del13q patients. Pairwise analysis (applying a cut-off of Log2FC>0.5 in at least half of all samples and of Log2FC>0 in all samples) identified 410 NOTCH1-regulated genes in tri12 (268 up-regulated/142 downregulated) and 418 genes in del13q (236 up-regulated/182 down-regulated) patient samples.
DE genes in each group were then used to validate the NOTCH1-transcriptome on a cohort of NOTCH1-mutated tri12 and del13q patients 29 and showed a significant enrichment in the respected genotype (Fig. 3c). The comparison of NOTCH1-induced genes did not identify a distinct expression profile in tri12 compared to del13q cells, as shown by the correlation of the respective logFCs in both sets of deregulated genes (Fig. 3d). Interestingly, the slope of the tri12-specific genes was clearly smaller than that of del13q-specific genes, suggesting that NOTCH1 DPEST induced a higher amplitude of gene activation/deactivation in the tri12 genetic context (Fig. 3d, black vs grey line). Furthermore, this analysis recognized 130 genes that were de-regulated at similar levels by NOTCH1 DPEST in all 13 samples (Fig. 3d, red line).
Subjecting this gene-set to GSEA identified transcriptional changes in gene clusters involved in signaling, stress-response and antigen-presentation ( Supplementary Fig. 2).
In addition to a clear transcriptional modulation, we next assessed whether NOTCH1 ΔPEST was also able to induce epigenetic programming at the level of chromatin regulation. For this, we performed ChIP-seq for histone H3 lysine 27 acetylation (H3K27ac), a bona fide mark for active regulatory elements 30 , in five paired CLL samples (expressing either EV-control or NOTCH1 ΔPEST ). Unsupervised principal component analysis revealed that the first components of the chromatin activation variability were patient specific. However, the fifth component, which explains 3.6% of the total variability, remarkably separated controls from NOTCH1 ΔPEST expressing samples, regardless of their genetic background (Fig. 3e). This NOTCH1 ΔPEST -associated signature was composed of 587 H3K27ac peaks. Of them, 422 peaks were located at active chromatin states of CLL reference epigenome samples 31 and at 8 the promoter or gene body of an annotated gene (see materials and methods). Furthermore, at those differentially acetylated peaks we observed a consistent increase or decrease in H3K27ac signal in all cases with NOTCH1 ΔPEST regardless of the cytogenetic background (Fig.   3f). Consistently with the RNAseq data, we identified chromatin activation at several NOTCH1 target genes such as HES1 (Fig. 3g). These data demonstrate that NOTCH1 mutations not only drive transcriptional changes but also induce an aberrant epigenetic programing of CLL cells.
Collectively, these results indicate that NOTCH1 activation positively regulates gene expression important for B cell activation while simultaneously repressing genes required for antigen-presentation. These effects were not qualitatively different in tri12 cells, but here NOTCH1 effects seemed to be more enhanced compared to del13q cells.

NOTCH1 represses MHC class II genes via down-regulation of CIITA
Our RNAseq analyses indicated that NOTCH1 is associated with reduced expression of genes important for antigen-presentation, including HLA-DM, -DR, -DP and -DQ, suggesting silencing of the MHC class II locus on chromosome 6. Indeed, H3K27ac CHIP-seq analysis confirmed that this gene repression was due to epigenetic silencing of the entire HLA-locus Dividing patients into either CIITA low or high expresser, based on the overall CIITA mRNA abundance, we discovered that those patients with low CIITA levels had a significantly more active disease and required treatment sooner. Importantly, NOTCH1 exon 34 mutations were twice as common in the CIITA low expresser group compared to high expresser (Fisher t-test, p> 0.032) (Fig. 4i).
In addition, we assessed the significance of CIITA expression in a cohort of pre-treated CLL patients from the CLL-2H study 39 . To also consider NOTCH1-activation in the absence of exon 34 mutations 15 , we first stratified 337 treatment-naïve patients based on the expression of canonical target genes HES1/2, HEY1/2 and CIITA expression (median high vs. median low).
This analysis identified a group of patients with high and low expression of CIITA in both NOTCH1-activated and non-activated groups, defining 4 patient cohorts ( Supplementary Fig.   3b,c). We applied these expression thresholds to gene expression data generated from PBMCs in a cohort of fludarabine-resistant CLL treated in the CLL-2H study. These analyses indicated that high expression of canonical NOTCH-target genes was not per se associated with an unfavorable prognosis, but significantly impacted in the overall survival in combination with low levels of CIITA expression (Fig. 4j). Importantly, within the CIITA low expresser, NOTCH1 mutations were significantly more frequent in the NOTCH high versus NOTCH low group.
Collectively, these data demonstrate that low levels of CIITA are associated with a more aggressive disease, in particular for patients with activated NOTCH-signaling. Furthermore, this analysis also suggests that NOTCH1 can be activated in the absence of NOTCH1mutations, as previously reported 15 .

NOTCH1 up-regulates PD-L1 and impairs T cell activation
NOTCH1-dependent suppression of CIITA and further downstream HLA-class II genes indicated a mechanism for an escape from immune surveillance. To provide further evidence for this, we co-cultured primary CLL cells, either transfected with EV or NOTCH1 DPEST , with Jurkat T cells, expressing luciferase under the control of the NFAT responsive element. Under these conditions, T cell activation was strictly dependent on the presence of the anti-CD19/ anti-CD3 bi-specific antibody Blinatumomab. Notably, expression of NOTCH1 DPEST in primary CLL cells mitigated the activation of Jurkat T cells (Fig. 5a). While these results indicated that NOTCH1 mutations permit immune escape of CLL cells, they also suggested that mechanisms other than HLA-class II down-regulation contribute. To identify potential candidates capable of downregulating TCR activity following NOTCH1-activation we performed quantitative proteomic analysis on primary CLL cells transduced with NOTCH1 ΔPEST or EV-control. For the simultaneous analysis of both groups of samples, we applied Tandem Mass Tags (TMT-6plex), which allows for the quantitative comparison between replicates and conditions. Total proteomic investigation of 3 patients recognized 7876 unique proteins and as a result of pairwise analysis, we identified 385 differentially regulated proteins (average Log2FC>0.5 and all 3 patients with Log2FC>0; Fig. 5b). Proteins expressed at a higher level included positive controls such as NOTCH1 and CD38. Interestingly, NOTCH1 DPEST increased the expression of CD27 and CD274 (PD-L1) in CLL cells, which both can impair T cell activation. Assessment of PD-L1 expression on additional 20 CLL patients, transfected with NOTCH1 DPEST or EV, invariably showed an up-regulation of PD-L1 through activated NOTCH1. Similar to CLL, primary MCL cells also showed a trend towards an increased expression of PD-L1 following transfection with NOTCH1 DPEST (Fig. 5c). Notably, we were unable to recapitulate this phenotype in the CLL cell lines MEC-1 and Hg-3 ( Supplementary   Fig. 3d), further emphasizing the limitations inherent to studies with cell lines. To demonstrate that NOTCH1 DPEST functions similarly to ligand-activated, mutated NOTCH1, we performed a reverse experiment by treating NOTCH1-mutated CLL with g-secretase inhibitors (GSI) to block NOTCH-activation. GSI-treatment caused a significant down-regulation of PD-L1 in primary NOTCH1-mutated CLL (Fig. 5d).
Importantly, we observed that the constitutive expression of PD-L1 on quiescent cells was minimal, but significantly up-regulated on activated, cycling CLL cells (Fig. 5e), in keeping with a report showing strong expression of PD-L1 on CLL cells in proliferative centers in lymph nodes 40 . To test whether CLL cells recently egressed from lymph nodes had higher expression levels of PD-L1, we assessed its expression on peripheral blood cells expressing CXCR4 dim /CD5 bright 41 . Unexpectedly, we did not find a different expression of PD-L1 compared to CXCR4 bright /CD5 dim cells, which was overall very low ( Fig. 5f). These data demonstrate that the upregulation of PD-L1 in proliferative centers is short-lived.
PD-L1 is also post-transcriptionally regulated through cyclinD-Cdk4 activity, causing cell-cycle dependent oscillations of PDL-L1 expression with a peak expression during M-and early G1phase 42 . Since NOTCH1 DPEST provided a significant proliferative advantage for primary CLL cells (Fig. 2c,d), we hypothesized that the increased expression of PD-L1 could be attributed to an increased proliferation, rather than being a specific NOTCH1-response. To address this, we assessed the expression of PD-L1 on cycling CLL cells. For this, cells were stained with an anti-PD-L1 antibody, followed by fixation and staining with Edu to distinguish cells in G0/1 from those in S-or G2/M-phase. In keeping with cell-cycle modulated expression of PD-L1, we observed a significant down-regulation on CLL cells going through S-phase, which recovered with entry into G2/M-phase. Notably, the expression of PD-L1 was consistently increased in NOTCH1 DPEST -transduced CLL cells compared to EV-controls (Fig. 5g), indicating that the NOTCH1-induced expression of PD-L1 is not dependent on cell proliferation. In addition, although retroviral expression of c-MYC caused rigorous proliferation of primary CLL cells (Fig. 1g), c-MYC did not affect PD-L1 expression (Fig. 5h), providing further evidence for a cell-cycle independent regulation of PD-L1 by NOTCH1.
Besides the post-transcriptional regulation of PD-L1, its expression is induced by a variety of pro-inflammatory cytokines, of which interferons are strong inducers. Importantly, CLL cells can produce and secrete IFN-g, which provides an anti-apoptotic signal through autocrine stimulation 43 . To investigate whether this feed-back loop was affected by NOTCH1-activation, we analyzed the expression of IFN-g and its receptor in NOTCH1 DPEST transduced cells.
NOTCH1 not only up-regulated IFN-g mRNA, but also the IFN-g receptor (Fig. 5i), suggesting a contribution of autocrine secreted IFN-g to NOTCH1-mediated expression of PD-L1. Indeed, CLL cells cultured in the presence of an antibody blocking the IFN-g receptor showed a significantly down-regulation of PD-L1 (Fig. 5j), which was still significantly higher than PD-L1 levels of EV-control cells, indicating that this pathway only partly contributed to the NOTCH1mediated up-regulation of PD-L1.
In conclusion, these results demonstrate that NOTCH1 signaling promotes escape from immune-surveillance through transcriptional regulation of HLA-class II genes and PD-L1.

NOTCH1-activation in CLL cells favors expansion of CD4 + cells in vivo
Our data demonstrated that PD-L1 expression was significantly up-regulated in cycling CLL cells, further enhanced through NOTCH1-activation. To provide in vivo evidence supporting this finding, unselected lymph node specimens from CLL/SLL patients were retrieved from the files of the Institute of Pathology, Würzburg, Germany and stained for NOTCH1. We found nuclear expression of NOTCH1 in 12% of all samples (Fig. 6a). Although genomic data for these samples were not available, this frequency is expected based on the occurrence of NOTCH1 mutations in an unselected patient cohort 19 . For multiparameter analysis of the lymph node microenvironment, we employed imaging mass cytometry (IMC) and applied a panel of isotype-labeled antibodies against B-and T cell epitopes on paraffin-embedded tissues (Fig. 6b). Following cell segmentation using the CellProfiler software we identified areas with high Ki67 signal using HistoCat software 44 to specifically gate on proliferative centers (PCs). IMC-single cell data identified a high percentage of T cells present in PCs (Fig.   6c). Analysis of PD-L1 expression on CD19 + cells revealed a stronger signal in PC areas compared to non-PC areas in all samples, in agreement with published data 40 . Further analyses were restricted to B cells in PCs and showed that nuclear NOTCH1 expression was associated with higher PD-L1 and Ki67 signals, compared to NOTCH1 negative samples (Fig.   6d). Assessment of T cells in PCs also showed a significantly higher infiltration of CD4 + cells in NOTCH1-positive samples, associated with higher expression of Ki67, suggesting that NOTCH1-expression in CLL cells promotes T cell expansion. Notably, PD-1 levels on CD4 + cells were similar between NOTCH-positive and negative samples (Fig. 6e). Similar to CD4 + cells, CD8 + cells were also more abundant in PC of NOTCH1-positive patient samples and they also expressed higher levels of PD-1 (Fig. 6f) in contrast to CD4 + cells. These results confirmed our in vitro data of NOTCH1-mediated regulation of PD-L1 and indicated that NOTCH1 supports proliferation of CD4 + and CD8 + cells, with the latter having a more exhausted phenotype.

13
To provide further experimental evidence for this hypothesis, we injected NOD.Cg-Prkcd scid Il2rd tm1Wjl / Szj (NSG) mice intraperitoneally with isogeneic primary CLL cells, carrying either NOTCH1 DPEST or EV-control. Prior to this, autologous T cells were isolated with anti-CD3 beads, cultured for 7 days and then co-injected at a ratio of 20:1 (CLL:T cell) (Fig. 6g).
Assessment of engrafted CLL cells showed a moderate, but not significant, increase in the tumor burden of NOTCH1-transduced cells after 3 weeks (Fig. 6h). Importantly, we observed a significant increase of CD4 + T cell in the peritoneal cavity of mice diseased with NOTCH1 DPEST -expressing CLL cells, demonstrating that NOTCH1 promotes expansion of CD4 + cells (Fig. 6i). This is further supported through clinical data from untreated CLL patients, showing a higher ratio of CD4 + /CD8 + cells in the peripheral blood of NOTCH1 mutated patients compared to NOTCH-wild type patients (Fig. 6j).

Discussion
Numerous sequencing studies have identified many mutations recurrently found in malignant B cells from CLL and MCL-patients 1,19,34 . To translate this knowledge into patient care, functional studies are needed to understand the mechanisms governed by these mutations and to identify downstream effects amenable for therapeutic interventions. Here we provide a method to functionally interrogate gene-mutations in primary human malignant B cells. For a disease such as MCL or for studying tumor cells with structural chromosomal abnormalities, for which no animal models exist, this method is indeed a unique opportunity to decipher the underlying disease biology.
We applied this technique to address the question why CLL and MCL patients carrying NOTCH1 mutations have a dismal prognosis. Previous studies had approached this question through investigations in cell lines, commonly derived from therapy-resistant patients.
Undoubtably, these studies have made significant contributions to our understanding of the NOTCH1 biology and described that it promotes proliferation, BCR-signaling, MAPK-signaling and chemotaxis in CLL cells 15-17,21,45 . Our studies with primary malignant B cells indeed confirmed these findings, but also identified yet unappreciated roles of NOTCH1.
Since the majority of NOTCH1 mutations in CLL and MCL do not affect protein binding to DNA but instead impair its proteasomal degradation by truncating the PEST domain 2,19,34,46 , several aspects need to be considered to fully comprehend when, where and how NOTCH1 signaling drives disease progression. The preserved DNA-binding function of mutated NOTCH1 suggests that it regulates the expression of identical genes than wild-type NOTCH1 and that disease-promoting events are rather caused through secondary events attributed to signal-persistence. Thus, mutated NOTCH1 still requires binding of NOTCH-ligands, expressed in trans, for signaling. This apparent ordinary ligand-receptor signaling is complicated by an unusual redundancy of ligands, expressed not only on cells in the tumor microenvironment, but also on malignant B cells themselves 33 , as well as co-expression of other NOTCHreceptors on tumor B cells 20 . The simultaneous expression of both NOTCH receptors and ligands on the same cell is known to lead to cis-inhibition 47 . NOTCH-signal strength is therefore dependent on the balance between cis-inhibition and trans-activation. Unfortunately, in CLL as well as in MCL, very limited knowledge exists about the expression of NOTCHligands and receptors in distinct niches in vivo. Extrapolating from our recent in vitro data, ligand and receptor expression are also likely to be dynamic in vivo and regulated by NOTCHsignaling itself 33 . Another unknown variable, essential for understanding how NOTCH1 mutations modulate disease biology is the length of time a tumor cell resides in one tissue before migrating to another, which will almost certainly impact on the activation of NOTCH1 in tumor cells as signaling is cell-contact dependent.
We believe this complexity of NOTCH1-signalling is important for understanding the recently reported activation of NOTCH1 in 50% of CLL patients, based on the presence of NOTCH1-ICD protein, although PEST-truncating mutations of NOTCH1 were only found in 22% in the same study 15 . While the reasons for the discrepancy between the presence of NOTCH1-ICD protein and gene-mutations remains elusive, these data also indicate that carrying a NOTCH1 mutation is fundamentally different from expressing NOTCH1-ICD as only gene mutations appear to predict for a poor clinical outcome. Furthermore, these data also hint on the importance of the tumor microenvironment for NOTCH1-activation, suggesting that mutations of oncogenes still rely on signals from non-malignant cells to fully unfold their detrimental effects. Our limited knowledge about these factors contributing to the activation of mutated NOTCH1 in CLL suggest that the lymph node environment is predominantly important for its activation 17,22 .
The negative prognostic effect of NOTCH1 mutations in CLL becomes even more evident if they occur on the background of trisomy 12. NOTCH1 mutations are enriched in patients with trisomy 12 26,27 , suggesting that this chromosomal aberration provides a selective advantage for NOTCH1 mutations. In addition, the risk for patients with trisomy 12 for transformation into clonally related Richter's syndrome is 10-times higher for NOTCH1-mutated patients compared to wild-type, suggesting that NOTCH1-signaling drives genomic instability and clonal evolution 11,48 . Our method to generate isogenic primary CLL cells provided an opportunity to directly address this question. Unexpectedly, our experiments did not identify a distinct NOTCH1-regulated gene-set, present only in trisomy 12 cells, but rather indicated a higher amplitude of gene-regulation in trisomy 12 compared to del13q cells. This observation raises a further question of what other factors determine the selective advantage for clones concurrently harboring trisomy12 and NOTCH1-mutations? A possible explanation is the observation that trisomy 12 CLL cells have an increased expression of CD29, CD49d and ITGB7, which occurs independently of NOTCH1 mutations 49 and allows for an improved adherence to cells of the microenvironment. As an immediate consequence and since NOTCH1-mutated cells are still dependent on ligand-binding for activation, trisomy 12 cells may experience prolonged NOTCH1-signaling. Therefore, and based on our data, we propose that in the subgroup of patients with trisomy 12 the selective advantage for NOTCH-mutations is based on enhanced, ligand-mediated NOTCH1-activation, rather than due to a specific genetic program governed by NOTCH1.
Gene repressive functions of NOTCH1 have previously been underappreciated. Our data indicated that NOTCH1 signaling permits immune escape of malignant B cells through downregulation of HLA-class II expression. The prognostic significance of HLA-class II expression is well documented for DLBCL 50 and PMBCL 37 and shows that short overall survival is associated with low expression levels in these entities. Similar to our study, low HLAexpression levels were not due to large genetic deletions on chromosome 6 but correlated with CIITA expression levels 36 , pointing to transcriptional de-regulation of HLA class II genes in high grade lymphoma.
Our data suggest that CIITA expression levels are a novel prognostic marker also for indolent B cell malignancies and show that NOTCH1 is a strong suppressor of CIITA transcription. NOTCH1-mediated control of CIITA expression has not previously been reported and the underlying mechanisms of this regulation remain to be defined. Since we observed that NOTCH1-signaling also up-regulated IFN-g, which itself activates CIITA transcription 51 , the transcriptional repression of CIITA by NOTCH1 likely involves epigenetic silencing of its promotors as shown for other hematological malignancies 52 .
The NOTCH1-mediated modulation of surface receptors regulating the interaction with T cells expectedly has effects on the composition of the tumor microenvironment. We found significantly more cycling CLL cells in proliferative lymph nodes of NOTCH1-expressing cells compared to non-expresser, associated with an increased number of CD4 + and CD8 + T cells.
This association is likely to be mediated through the recruitment of T cells through the secretion of CCL3 and CCL4, derived from activated CLL cells 53,54 . Notably, the increased number of CD8 + T cells, but not of CD4 + T cells, was associated with higher expression of PD-1 in NOTCH1 positive samples, indicating terminal differentiation and exhaustion of CD8 + cells. The role of CD4 + T cell-subsets in CLL is far from being completely understood, but the collective evidence indicates that CD4 + T cells overall are tumor-promoting. This conclusion is based on the dependency of CLL cells on autologous T cells to engraft in NSG mice 55 , in vitro growth promoting effects of CLL-specific Th1-cells 56 and correlation between higher CD4 + cell counts and shorter PFS and OS 57 . The relative contribution of individual CD4 + subsets is less clear, but numerous studies suggest that this phenotype may predominantly be driven by Our data predict that PD-1/ PD-L1 inhibition could be more efficacious in NOTCH1 mutated patients and future prospective studies are needed to address this.

Figure 6: NOTCH1-activation is associated with T cell proliferation in vivo
a. IHC staining of NOTCH1 in CLL lymph node biopsies. Five out of 42 samples showed nuclear expression of NOTCH1, representing 12% of cases from an unselected cohort (bar graph). Representative IHC images of positive and negative NOTCH1 nuclear staining are shown. b. Graphical scheme of the procedure used for the mass cytometry analysis. In brief, samples negative (n=3) or positive (n=4) for nuclear NOTCH1 IHC staining were processed and stained with a panel of metal-tagged antibodies. Following Hyperion Imaging System analysis, the data was processed using Cell Profiler and Histocat software. c. Multiplexed IMC image example of a NOTCH1 positive specimen with an enlarged area of the proliferative center (PC). Scale bar = 244µm. Representative t-SNE plots generated with Histocat showing CD19, CD4 and CD8 cell populations are displayed on the right. d. Intensity of cellular signal per given cell was calculated using the HistoCat software.
PD-L1 signal in CD19 gated cells following tNSE analysis in PC and non-PC area (i), PD-L1 (ii) and KI67 (iii) in CD19-gated cells following tNSE analysis in PC of samples with a positive or negative NOTCH1 nuclear staining. e. Total CD4 signal in PC (i) and KI67 (ii) and PD-1 (iii) in CD4-gated cells following tNSE analysis in PCs of NOTCH1 positive or negative samples. f. Total CD8 signal in PC (i) and PD-1 (ii) in CD8-gated cells following tNSE analysis in PCs of NOTCH1 positive or negative samples. g. Graphical scheme of the in vivo experiment using transduced CLL cells and autologous T cell to generate NSG chimera. CLL cells were transduced with an empty vector or NOTCH1 ΔPEST and intraperitoneally injected into NSG mice. Autologous T cells were cultured with IL-2, a-CD3 and a-CD28 for 7 days prior to injection. h. Engraftment of human CD19 + cells in the peritoneal cavity (left) and spleen (right) of mice injected with NOTCH1 ΔPEST (n=8) or an empty vector (n=8) transduced CLL cells.
In total 16 mice were analyzed using cells from 3 independent donors. i. Quantification of the ratio of human autologous CD4 + /CD8 + T cells in the peritoneal cavity of mice injected with CLL cells with or without NOTCH1 ΔPEST . Mean value was obtained from 3 independent experiment. In total analysis was performed on 6 mice injected with 3 independent CLL donor cells. j. Ratio of human CD4 + /CD8 + T cells in the peripheral blood of a cohort of naive CLL patients with NOTCH1 wild type or mutated Cohorts are shown as median (D-E) and as mean ± SEM (H-J). *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001 and not significant (ns) P > 0.05.

Primary cells and cell culture
After patients' informed consent and in accordance with the Helsinki Declaration, peripheral blood was obtained from patients with a diagnosis of CLL or MCL. Studies were approved by the Cambridgeshire Research Ethics Committee (07/MRE05/44).
PBMCs were isolated from heparinized blood samples from patients by centrifugation over a Microplate Reader.

Expression analysis/qPCR
Total RNA was isolated using the RNeasy Mini Kit (Qiagen, Manchester, UK), and complementary DNA (cDNA) was obtained using the qScriptTM cDNA SuperMix kit (QuantaBio, Beverly, MA, USA). Quantitative reverse-transcription polymerase chain reaction (RT-qPCR) was performed on isolated mRNA using the fast SYBR reagents and the Applied Biosystems™ QuantStudio™ 12K system. Target gene expression levels were normalized to GAPDH and values are represented as fold change relative to control using the ΔΔCt method.

Mass spectrometry
Following cell sorting 10 6 cell pellets were collected for mass spectrometry analysis. Protein isolation and TMT-6plex labelling was performed as described previously 65 . TMT mix was fractionated on a Dionex Ultimate 3000 system at high pH using the X-Bridge C18 column Raw data of each population was extracted into excel files and plotted using GraphPad Prism 9.0 (GraphPad Software, La Jolla, USA).

RNA-seq
Total RNA was isolated from GFP + CD19 + sorted cells using the RNeasy Mini Kit (Qiagen, Manchester, UK). Samples (25 ng total RNA) were then processed for NGS sequencing using  Clinical impact for cases with high NOTCH-pathway activation (averaged expression levels of HES1/2, HEY1/2, expression above median expression level was defined as high NOTCHpathway activation) and corresponding high or low CIITA expression levels (median high vs. median low CIITA) was assessed using gene expression data generated from PBMCs in a cohort of fludarabine resistant CLL patients.