USP7 Cooperates with NOTCH1 to Drive the Oncogenic Transcriptional Program in T-Cell Leukemia

This article is featured in Highlights of This Issue, p. 1

Acute lymphoblastic leukemia (ALL) represents 25% of childhood tumors (1, 2). T-cell ALL (T-ALL; refs. 1–7) is the most aggressive subtype of ALL and these tumors exhibit a significant risk of recurrence due to tumor cells that are refractory to chemotherapy, a common characteristic of high-risk T-ALL. In this regard, 20% of pediatric, and more than 50% of adult patients, with T-ALL fail to achieve a complete remission after chemotherapy, making resistance to therapy the most substantial challenge in treatment (5, 8, 9). Immunotherapy using engineered T cells appears promising for some patients, but these experimental therapies also appear fraught with unacceptable toxicities. ALL, thus, represents a high unmet need. To this end, improvement in therapy, as well as personalized cancer medicine, should identify subgroups of patients, based on the molecular mechanism of resistance, which will benefit patients through the development of novel therapeutic strategies, including targeted agents in patients with high-risk T-ALL.

The NOTCH1 (1, 2, 10–26) signaling pathway is activated in more than 50% of T-ALL cases. NOTCH1 mutations were initially identified in the form of chromosomal translocation t(7;9)(q34;q34.3), which leads to the expression of a truncated and constitutively active form of NOTCH1 (10). In most T-ALL cases, NOTCH1 is activated via mutations that disrupt specific domains responsible for the nuclear levels of NOTCH1 protein. NOTCH1-positive leukemia presents with infiltration of bone marrow, spleen, and liver with leukemic blasts and lead to a fulminant terminal disease that, if untreated, can lead to death over the course of months. Seminal work has described the major oncogenic pathways (i.e., NOTCH1) in T-ALL (1, 2, 10–21). Chemical inhibitors or antibodies against NOTCH1 have been unsuccessful in the clinic, due to high levels of on-target toxicity, given the involvement of this protein in multiple physiologic processes (11, 27).

Others and we have previously shown that loss of the repressive mark trimethylation of lysine 27 on histone H3 (H3K27me3) is important for leukemia initiation and progression (15), and NOTCH1 recruitment to target genes and loss of H3K27me3 correlate during leukemogenesis (15). We also identified and characterized the pro-oncogenic role of the epigenetic modulator Jumonji D3 (JMJD3; refs. 28–31), which mediates NOTCH1 oncogenic function in T-ALL (16). JMJD3 is a NOTCH1 interactor and is recruited by NOTCH1 to oncogenic targets (16).

Here we sought out to identify additional NOTCH1 cofactor proteins with pro-oncogenic functions that can be targeted in T-ALL contexts without causing toxicity. In this study, we used a number of patient samples and cell lines of T-ALL, as well as preclinical models, including patient-derived xenografts. We identified a novel member of the NOTCH1 transcriptional complex in T-ALL, the first deubiquitinase (DUB) protein with critical role in leukemia, referred to as ubiquitin-specific peptidase 7 (USP7; refs. 32–43). We have shown that: (i) the NOTCH1 complex interacts with and is stabilized by USP7; (ii) USP7 levels are significantly increased in T-ALL tumor samples compared to normal T cells via a positive feedback loop with NOTCH1; (iii) a small chemical molecule that inhibits USP7 (USP7i), leads to a blockade in T-ALL cell growth, without associated toxicity in healthy animals. Thus, it is proposed to determine the mechanisms employed by USP7/NOTCH1 that leads to a T-ALL phenotype and most importantly, can targeting USP7 reverse this phenotype in patients.

The human T-ALL cell lines CUTLL1 (gift from Iannis Aifantis, New York University, New York, NY), LOUCY (gift from Pieter Van Vlierberghe, Cancer Research Institute, Belgium), CCRF-CEM (ATCC; #CCL-119), JURKAT (ATCC), HPB (HPB-ALL), and KOPTK1 (Iannis Aifantis' group) were cultured in RPMI1640 medium supplemented with 10% heat-inactivated FBS (Sigma-Aldrich), 2% penicillin/streptomycin (Gibco, Fisher Scientific), and 1% GlutaMAX (Gibco, Fisher Scientific). 293T cells (ATCC, #CRL-11268) were maintained in DMEM medium supplemented with 10% heat-inactivated FBS, 2% penicillin/streptomycin, and 1% GlutaMAX. Human pan T cells were purchased from AllCells.com. Primary human samples were collected by collaborating institutions with informed consent and analyzed under the supervision of the Institutional Review Board of Ghent University. Informed consent to use leftover material for research purposes was obtained from all of the patients at trial entry in accordance with the Declaration of Helsinki.

The following antibodies were used for Western blot analysis: mouse anti-actin (Millipore; clone C4), rabbit anti-JMJD3 (Cell Signaling Technology; #3457), rabbit anti-USP7 (Bethyl Laboratories, Montgomery, TX; #A300-033A-7), rabbit anti-cleaved NOTCH1 (Val1744; Cell Signaling Technology; #4147), rabbit anti-K48 specific ubiquitin (Millipore; 05-1307), rabbit anti-Lamin B1 (Proteintech; 12987-1-AP), rabbit anti-USP9X (Bethyl Laboratories; #A301-351A), anti-Flag (Sigma; M2 clone), anti-HA (Abcam; ab18181), and rabbit anti-USP24 (Bethyl Laboratories; #A300-938A).

The following antibodies were used for ChIP: rabbit anti-H3K27Ac (Cell Signaling Technology; #8173S), rabbit anti-H3K27me3 (Cell Signaling Technology; #9733S), rabbit anti-H2BK120ub (Cell Signaling Technology; #5546S), rabbit anti-H2AK119ub (Cell Signaling Technology; #8240S), rabbit anti-H3K79me2 (made by Ali Shilatifard's laboratory, Northwestern University), rabbit anti-H3K4me3 (Ali Shilatifard's laboratory), normal rabbit IgG (Millipore; #12-370), anti-Flag (Sigma; M2 clone), and rabbit anti-USP7 (Bethyl Laboratories; #A300-033A-7). The rabbit anti-C-MYC:(N-262; Santa Cruz Biotechnology) and goat anti-RUNX1 (Santa Cruz Biotech; SC- 8563) antibodies were used in published ChIP-seq studies we use in our analysis. A RUNX1 antibody from Cell Signaling Technology was used for our Western blot and ChIP-qPCR studies (AML1 antibody #4334). Secondary antibodies for Western blots were HRP-conjugated anti-rabbit and anti-mouse IgG (GE Healthcare).

Quick Start Bovine Gamma Globulin (BGG) Standard Set (protein standards) were purchased from Bio-Rad; benzonase, RNase A, dithiothreitol (DTT), EZview Red Anti-HA Affinity Gel (HA beads), and Influenza Hemagglutinin (HA) peptide were purchased from Sigma-Aldrich; NaV and NaF were purchased from New England BioLabs; Protein G Dynabeads were purchased from Life Technologies; IgG-free BSA was purchased from Jackson ImmunoResearch Laboratories (West Grove, PA); phenol chloroform was purchased from ThermoScientific; and proteinase K, Tousimis formaldehyde, and Peptide International Z-Leu-Leu-Leu-H aldehyde (MG132) were purchased from Fisher Scientific. The latest generation of Progenra inhibitors (USP7i), P217564, was a kind gift from Progenra (Malvern, PA). Several additional inhibitors from Progenra were used (P5091 and P22077) yielding results similar to P217564. All USP7 inhibitors used in the manuscript have been published previously as referred to in the manuscript. GSKJ4 was purchased from Cayman Chemical; #12073.

One hundred million T-ALL cells were collected and washed with chilled PBS. Cells were resuspended in 5 volumes of Buffer A [10 mmol/L HEPES, 1.5 mmol/L MgCl₂, 10 mmol/L KCl, 1:100 protease inhibitor (Sigma-Aldrich; P8340), 1 mmol/L NaV, 1 mmol/L NaF, and 0.5 mmol/L DTT in H₂O], incubated on ice for 10 minutes, and lysed using a Dounce homogenizer. Nuclear pellets were resuspended in 1.5 mL of TENT buffer [50 mmol/L Tris pH 7.5, 5 mmol/L EDTA, 150 mmol/L NaCl, 0.05% (v/v) Tween 20, 1:100 protease inhibitor (Sigma-Aldrich, P8340), 1 mmol/L NaV, 1 mmol/L NaF, and 0.5 mmol/L DTT in H₂O] containing 5 mmol/L MgCl₂ and 100 units benzonase, and incubated at 4°C for 30 minutes, rotating. Lysates were passed through a 25^1/2G needle/syringe 5 times, and spun down at 4°C, 2000 RPM, for 7 minutes to remove debris. Protein G magnetic beads were added to the lysates to decrease nonspecific binding and incubated at 4°C for 30 minutes, rotating. Precleared lysates were then incubated with the appropriate antibody-conjugated beads (5 μg antibody per 100 million cells) at 4°C overnight, rotating. Beads were washed 4 times in TENT buffer at 4°C for 3 minutes, and protein complexes were eluted in 200-μL 0.1 mol/L glycine pH 2.5 for 10 minutes at 25°C, shaking. Twenty microliters of 1 mol/L Tris pH 8.0 was then added to the supernatants. For IP of HA-JMJD3, precleared lysates were incubated with HA-coated beads overnight, and protein complexes were eluted in 200-μL TENT buffer containing 400 μg/mL HA peptide (Sigma-Aldrich) at 4°C overnight, rotating.

Histone epiproteomics analysis was performed as previously described (44, 45). For the JMJD3 mass spectrometry, the affinity-purified proteins were reduced, alkylated, and loaded onto an SDS-PAGE gel to remove any detergents and LC/MS incompatible reagents. The gel plugs were excised, destained, and subjected to proteolytic digestion with trypsin. The resulting peptides were extracted and desalted as previously described (46). An aliquot of the peptides was analyzed with LC/MS coupled to a ThermoFisher Scientific Orbitrap QExactive Mass Spectrometer operated in data dependent mode. The data were searched against a UniProt human database, using Sequest within Proteome Discoverer. The results were filtered with a 1% FDR searched against a decoy database and for proteins with at least 2 unique peptides. Word clouds were generated using the statistical program R, and gene ontology was performed using the Gene Ontology Consortium (http://www.geneontology.org/).

As previously described (47), the affinity purified proteins were resuspended in NuPAGE LDS Sample Buffer (Novex). The samples were reduced with 2 μL of 0.2M dithiothreitol (Sigma) for 1 hour at 57 °C at pH 7.5. Next the samples were alkylated with 2 μL of 0.5M iodoacetamide (Sigma) for 45 minutes at room temperature in the dark. The samples were loaded on a NuPAGE 4% to 12% Bis-Tris Gel 1.0 mm (Life Technologies) and run for 8 minutes at 200 V. The gel was stained with GelCode Blue Stain Reagent (Thermo). The gel plugs were excised and destained for 15 minutes in a 1:1 (v/v) solution of methanol and 100 mmol/L ammonium bicarbonate. The buffer was exchanged and the samples were destained for another 15 minutes. This was repeated for another 3 cycles. The gel plugs were dehydrated by washing with acetonitrile, and further dried by placing them in a SpeedVac for 20 minutes. Three hundred nanograms of sequencing grade modified trypsin (Promega) was added directly to the dried gel plugs followed by enough 100 mmol/L ammonium bicarbonate to cover the gel pieces. The gel plugs were allowed to shake at room temperature and digestion proceeded overnight. The digestion was halted by adding a slurry of R2 50 μm Poros beads (Applied Biosystems) in 5% formic acid and 0.2% trifluoroacetic acid (TFA) to each sample at a volume equal to that of the ammonium bicarbonate added for digestion. The samples were allowed to shake at 4°C for 120 minutes. The beads were loaded onto C18 ziptips (Millipore), equilibrated with 0.1% TFA, using a microcentrifuge for 30 s at 6,000 rpm. The beads were washed with 0.5% acetic acid. Peptides were eluted with 40% acetonitrile in 0.5% acetic acid followed by 80% acetonitrile in 0.5% acetic acid. The organic solvent was removed using a SpeedVac concentrator and the sample reconstituted in 0.5% acetic acid.

An aliquot of each sample was loaded onto an Acclaim PepMap100 C18 75-μm × 15-cm column with 3-μm bead size coupled to an EASY-Spray 75-μm × 50-cm PepMap C18 analytical HPLC column with a 2-μm bead size using the auto sampler of an EASY-nLC 1000 HPLC (ThermoFisher) and solvent A (2% acetonitrile, 0.5% acetic acid). The peptides were eluted into a Thermo Fisher Scientific Orbitrap Q Exactive Mass Spectrometer increasing from 2% to 30% solvent B (90% acetonitrile, 0.5% acetic acid) in 60 minutes, followed by an increase from 30% to 40% solvent B in 10 minutes and 40% to 100% solvent B in another 10 minutes.

High-resolution full MS spectra were obtained with a resolution of 70,000, an AGC target of 1e6, with a maximum ion time of 120 milliseconds, and a scan range from 400 to 1,500 m/z. Following each full MS scan, 20 data-dependent MS-MS spectra were acquired. The MS-MS spectra were collected with a resolution of 17,500 an AGC target of 5e4, maximum ion time of 120 milliseconds, 1 microscan, 2 m/z isolation window, fixed first mass of 150 m/z, dynamic exclusion of 30 seconds, and normalized collision energy (NCE) of 27.

All acquired MS2 spectra were searched against a UniProt human database using Sequest within Proteome Discoverer (ThermoScientific). The search parameters were as follows: precursor mass tolerance ±10 ppm, fragment mass tolerance ± 0.02 Da, digestion parameters trypsin allowing 2 missed cleavages, fixed modification of carbamidomethyl on cysteine, variable modification of oxidation on methionine, and variable modification of deamidation on glutamine and asparagine. The results were filtered with a 1% FDR searched against a decoy database and for proteins with at least 2 unique peptides. Proteins enriched in the sample affinity purification over the empty vector control were further interrogated.

Reverse phase protein array (RPPA) was performed as described in Milani G. and colleagues, Oncotarget 2014, and Serafin V. and colleagues, Leukemia 2017. Briefly, cells were lysed in an appropriate lysis buffer with proteases and phosphatases inhibitors, serially diluted into 4-point dilution curves and printed on nitrocellulose-coated glass slides with the 2470 Aushon Arrayer (Aushon Biosystems).

To make total cell extracts, up to 10 million cells were collected and resuspended in 20-μL RIPA buffer [50 mmol/L Tris HCl pH 8.0, 150 mmol/L NaCl, 1% NP-40/IGEPAL, 0.5% sodium deoxycholate, 0.1% SDS, 1:100 protease inhibitor (Sigma-Aldrich, P8340), 1 mmol/L NaV, and 1 mmol/L NaF in H₂O] per 1 million cells. Cells were lysed on ice for 20 minutes, and spun down at 4°C, max speed, for 10 minutes to remove debris.

Protein concentrations were determined via Bradford assay. Samples and buffer were diluted 1:10 in H₂O. Two microliters of protein standards, H₂O, or diluted sample were added to wells of a 96-well plate in duplicate. Then, 2 μL of diluted buffer and 100-μL Quick Start Bradford 1× Dye Reagent (Bio-Rad) were added to each well, and absorbance was measured at 600 nm using the GloMax-Multi Detection System (Promega).

Up to 50-μg sample was boiled in 1× SDS loading dye (Bio-Rad) at 95°C for 10 minutes prior to loading into 4% to 15% Tris-glycine polyacrylamide gels (Bio-Rad). Eight microliters of PageRuler Plus Prestained Protein Ladder (10–250 kD; Fisher Scientific) was also loaded. Gels were run at 100 V until samples reached the separating part of the gel, and then were run at 130 V. Gels were transferred for 1.5 hours at 80 V or overnight at 35 to 40 V, and membranes were blocked in 5% milk in TBST (0.1% Tween 20 in 1× TBS) for 1 hour. Membranes were incubated at 4°C overnight with the appropriate antibody in TBST. Then, the membranes were washed 3 times for 10 minutes with TBST, incubated for 2 hours at 4°C with the appropriate secondary antibody, washed 3 times for 10 minutes with TBST, and developed using Clarity Western ECL Substrate (Bio-Rad), or SuperSignal West Femto Maximum Sensitivity Substrate (ThermoScientific) as needed, on a Bio-Rad ChemiDoc Touch Imaging System. Analysis was performed using Image Lab software (Bio-Rad).

Ten million T-ALL cells were cross-linked in 1 mL/million cells fixation buffer (1% formaldehyde, 1× PBS, and 1% FBS in H₂O) for 10 minutes at 25°C. Then, 1:12.5 glycine [2.5 mol/L] was added for 5 minutes. Pelleted cells were then lysed according to the type of ChIP performed.

For histone ChIPs, cells were lysed in 375 μL of nuclei incubation buffer [15 mmol/L Tris pH 7.5, 60 mmol/L KCl, 150 mmol/L NaCl, 15 mmol/L MgCl₂, 1 mmol/L CaCl₂, 250 mmol/L Sucrose, 0.3% NP-40, 1 mmol/L NaV, 1 mmol/L NaF, and 1 EDTA-free protease inhibitor tablet (Roche)/10 mL in H₂O] for 10 min on ice. Nuclei were washed once with digest buffer [10 mmol/L NaCl, 10 mmol/L Tris pH 7.5, 3 mmol/L MgCl₂, 1 mmol/L CaCl₂, 1 mmol/L NaV, 1 mmol/L NaF, and 1 EDTA-free protease inhibitor tablet (Roche)/10 mL in H₂O] and resuspended in 57-μL Digest Buffer containing 4.5 units MNase (USB) for 1 hour at 37°C. MNase activity was quenched for 10 minutes on ice upon the addition of EDTA to a final concentration of 20 mmol/L. Pelleted nuclei were lysed in 300-μL Nuclei Lysis Buffer [50 mmol/L Tris-HCl pH 8.0, 10 mmol/L EDTA pH 8.0, 1% SDS, 1 mmol/L NaV, 1 mmol/L NaF, and 1 EDTA-free protease inhibitor tablet (Roche)/10 mL in H₂O] using a Bioruptor Pico (Diagenode) for 5 minutes (30 seconds on, 30 seconds off). Lysate was centrifuged at max speed for 5 minutes to remove debris, and 9 volumes of IP Dilution Buffer [0.01% SDS, 1.1% Triton X-100, 1.2 mmol/L EDTA pH 8.0, 16.7 mmol/L Tris-HCl pH 8.0, 167 mmol/L NaCl, 1 mmol/L NaV, 1 mmol/L NaF, and 1 EDTA-free protease inhibitor tablet (Roche)/10 mL in H₂O] were added to the supernatant. Fifty microliters of protein G magnetic beads, blocked with IgG-free BSA, were added to the sample and incubated at 4°C for 30 minutes, rotating. 1% of the precleared sample was kept out as input, and the remaining sample was split into 3 tubes. Fifty microliters of protein G magnetic beads conjugated to 15 μL of the appropriate antibody were added to each tube, and incubated at 4°C overnight, rotating. Bead-bound complexes were washed for 5 minutes each in 1 mL of low salt buffer [20 mmol/L Tris-HCl pH 8.0, 150 mmol/L NaCl, 2 mmol/L EDTA, 1% (w/v) Triton X-100, and 0.1% w/v SDS in H₂O], high salt buffer [20 mmol/L Tris-HCl pH 8.0, 500 mmol/L NaCl, 2 mmol/L EDTA, 1% (w/v) Triton X-100, and 0.1% w/v SDS in H₂O], LiCl buffer [10 mmol/L Tris-HCl pH8.0, 250 mmol/L LiCl, 1 mmol/L EDTA, 1% (w/v) NP-40, and 1% (w/v) deoxycholic acid in H₂O], and TE.

For epigenetic regulator and transcription factor ChIPs, cells were lysed in 1-mL LB1 buffer (50 mmol/L HEPES-KOH pH 7.5, 140 mmol/L NaCl, 1 mmol/L EDTA, 10% glycerol, 0.5% NP-40, 0.25% Triton X-100, 1:100 protease inhibitor [Sigma-Aldrich, P8340), 1 mmol/L NaV, and 1 mmol/L NaF in H₂O] for 10 minutes at 4°C, rotating. Nuclei pellets were resuspended in LB2 Buffer [10 mmol/L Tris-HCl pH 8.0, 200 mmol/L NaCl, 1 mmol/L EDTA, 0.5 mmol/L EGTA, 1:100 protease inhibitor (Sigma-Aldrich, P8340), 1 mmol/L NaV, and 1 mmol/L NaF in H₂O] and incubated for 10 minutes at 25°C, rotating. Finally, nuclei pellets were lysed in 300 μL LB3 buffer [10 mmol/L Tris-HCl pH 8.0, 100 mmol/L NaCl, 1 mmol/L EDTA pH 8.0, 0.5 mmol/L EGTA pH 8.0, 0.1% sodium deoxycholate, 0.5% N-lauroylsarcosine, 1:100 protease inhibitor (Sigma-Aldrich, P8340), 1 mmol/L NaV, and 1 mmol/L NaF in H₂O] using a Bioruptor Pico (Diagenode) for 5 minutes (30 seconds on, 30 seconds off). 1% Triton X-100 was added, and samples were centrifuged at max speed to remove debris. Fifty microliters of protein G magnetic beads, blocked with IgG-free BSA, were added to the sample and incubated at 4°C for 30 minutes, rotating. One percent of the precleared sample was kept out as input, and the remaining sample was used for IP. Fifty microliters of protein G magnetic beads conjugated to 5 to 10 μg of the appropriate antibody were added to each tube, and incubated at 4°C overnight, rotating. Bead-bound complexes were washed for 5 minutes each in 1 mL of RIPA wash buffer (50 mmol/L HEPES-KOH pH 7.6, 300 mmol/L LiCl, 1 mmol/L EDTA, 1% NP-40, and 0.7% sodium deoxycholate in H₂O), 5 times, and 1 time in 1-mL TE wash buffer (10 mmol/L Tris-HCl pH 8.0, 1 mmol/L EDTA, and 50 mmol/L NaCl in H₂O).

To elute bead-bound complexes, 50 μL of elution buffer (for 10 mL total volume, 7.5-mL H₂O, 0.5 mL 20% SDS, and 2-mL 0.5M sodium bicarbonate) was added to each sample, and samples were incubated at 65°C for 15 minutes, shaking at 1,000 RPM on a thermomixer (ThermoScientific). Elution was repeated a second time, and then 100-μL RNase Buffer (12-μL of 5M NaCl, 0.2-μL 30 mg/mL RNase, and 88-μL TE) was added to each ChIP and input sample. Samples were incubated at 37°C for 20 minutes, followed by the addition of PK Buffer (2.5-μL 20 mg/mL proteinase K, 5-μL 20% SDS, and 92.5-μL TE) overnight at 65°C. An equal volume of phenol chloroform solution was added to the samples, which were vortexed for 1 minute and transferred to MaXtract High Density tubes (Qiagen). Samples were centrifuged for 8 minutes at max speed, and the upper phase was transferred to new tubes containing 1.5-μL 20 mg/mL glycogen. Then, 30-μL sodium acetate and 800-μL 100% ethanol were added, and tubes were incubated on dry ice to 30 to 60 minutes. DNA pellets were washed in 70% ethanol, air-dried, and resuspended in 30-μL H₂O.

SRMs capable of discriminating 19 acetylation sites from 13 different histone peptides were developed with 2 to 4 fragment ions for each species, as described previously (44). Various histone peptides with different modifications were purchased or custom synthesized from Anaspec or Genescript to assist the method development. Data were analyzed in Skyline with Savitzky–Golay smoothing (34). All labeled and unlabeled peaks were grouped as a whole for manual peak determination to avoid bias. Total peak areas from SRMs were used for quantification (as listed in ftp://PASS01134:[email protected]/). To access files via FTP, use credentials:

Servername: ftp.peptideatlas.org

Username: PASS01134

Password: HG6634ca

Libraries were prepared using Agencourt AMPure XP beads (Beckman Coulter) and the KAPA HTP Library Preparation Kit (KAPA Biosystems), according to the manufacturer's protocol (v4.15). DNA fragment size was determined using High Sensitivity DNA Chips read on a 2100 Bioanalyzer (Agilent Technologies). Libraries were sequenced on the Illumina NextSeq 500 (50bp single reads). FASTQ reads were first trimmed from the 3′ end using Trimmomatic (48) version 0.33, such that the Phred33 score of all the nucleotides was above 30 and all reads shorter than 20bp were discarded. The resulting reads were then aligned to the hg19 genome using bowtie version 1.1.2 with the following parameters: bowtie -p 5 -m 1 -v 2 -S (49). Next, the bam file was sorted and bigwig tracks were created by extending each read by 150 bp and using the GenomicAlignments (50) package in R in order to calculate the coverage of reads in counts per million (CPM) normalized to the total number of reads for each sample in the library.

Peaks were called using MACS (51) version 1.4.2 with the default parameters, using the specific aligned ChIP-Seq data for a particular antibody as the treatment sample and an aligned bam file of the unprecipitated input as a control sample. Only the top 10,000 peaks ordered by peak score (which is proportional to the FDR of the peak) were chosen for further analysis.

Overlap between 2 ChIP-Seq peak datasets were calculated using the mergePeaks tool in the HOMER (52) tools. In brief, the intersection between 2 ChIP-Seq peak datasets was determined by calculating the number of peaks in 1 set that overlap with the peaks in the second set. Statistical significance of the overlaps between ChIP-Seq datasets was done by using reservoir sampling (53) with a set of 161 transcription factor peak binding sites downloaded from http://hgdownload.cse.ucsc.edu/goldenpath/hg19/encodeDCC/wgEncodeRegTfbsClustered/ (wgEncodeRegTfbsClusteredV3.bed.gz, ENCODE data). The tool for implementing the sampling schema is called poverlaps and can be downloaded from github at https://github.com/brentp/poverlap. A total of 100 permutations per pair-wise combination of transcription factor overlaps was used. Venn diagrams of overlaps were generated using an online Venn diagram generator (http://jura.wi.mit.edu/bioc/tools/venn.php).

HOMER, with the standard default parameters, was used to determine the enrichment for known and unknown motifs in the USP7 ChIP-Seq peaks. Pathway enrichment analysis of USP7-bound genes were determined using GREAT (Genomic Regions Enrichment of Annotations Tool; ref. 54).

Heatmap representation of log2 fold-changes in epigenetic marks between DMSO and USP7 inhibition were visualized using ngs.plot (55). In brief, log2 fold-changes of H3K27me3, H2Aub, H2Bub, and H3K79me2 upon USP7 inhibition versus DMSO treatment at the top 10,000 USP7 peaks were visualized. A k-means clustering algorithm option was used with the number of clusters set to 5. Correlated gene expression changes were calculated by determining the log2 fold-change in gene expression upon USP7 or JMJD3 inhibition of the nearest protein-coding gene to the USP7 peak. Gene expression changes were visualized using Java TreeView (56).

RNA was isolated from T-ALL cells using the Aurum Total RNA Mini Kit (Bio-Rad), and was quantified on a Qubit 3.0 Fluorometer (Life Technologies) using the Qubit RNA HS (High Sensitivity) Assay Kit (Life Technologies) according to the manufacturers' instructions. cDNA was made using the High Capacity cDNA RT Kit (Applied Biosystems), according to the manufacturer's protocol. qPCR reactions were carried out using iTaq Universal SYBR Green Supermix (Bio-Rad) and the following primers: USP7: 5′-CGGTGTTGTGTCCATCACTC-3′ (forward) and 5′-AGTTGAGCGAGCCCGAG-3′ (reverse); NOTCH3: 5′-GTAGAGGGCATGGTGGAAGA-3′ (forward) and 5′-AAGTGGTCCAACAGCAGCTT-3′ (reverse); DTX1: 5′-CTCGCCACTGCTATCTACCC-3′ (forward) and 5′-CGTGCCGATAGTGAAGATGA-3′ (reverse); and HES1: 5′-GCAGATGACGGCTGCGCTGA-3′ (forward) and 5′-AAGCGGGTCACCTCGTTCATGC-3′ (reverse). Samples were run on a CFX Connect Real-Time System (Bio-Rad) under the following conditions: 95°C for 10 seconds (denaturation), 60°C for 30 seconds (annealing), and 72°C for 30 seconds (elongation), for 40 cycles. Analyses were performed using GraphPad Prism software (GraphPad Software). Statistical comparisons were made using the Student unpaired, 2-sided t test. P values <0.05 were considered statistically significant. Relative mRNA abundance refers to the levels of the gene of interest normalized to housekeeping genes GAPDH or G6PD. All values were then normalized to the control sample.

Analysis of microarray data from GEO was done using the NCBI GEO2R online tool for microarray analysis. Adjusted P value calculations were done using Benjamini and Hochberg (False discovery rate or FDR) option. An FDR of <0.05 was considered to be statistically significant.

This assay was performed as previously described, using recombinant enzymes and a panel of USP7 inhibitors, including P217564 (57). In brief, in vitro recombinant enzymes in 20 mmol/L Tris-HCl (pH 8.0), 2 mmol/L CaCl2, and 2 mmol/L b-mercaptoethanol were incubated with dose ranges of P5091 for 30 minutes in a 96-well plate before the addition of Ub-PLA2 and NBD C6-HPC (Invitrogen) or Ub-EKL and EKL substrate, as previously described (58). The liberation of a fluorescent product within the linear range of the assay was monitored using a Perkin Elmer Envision fluorescence plate reader. Activity-based DUB probes were used for the evaluation of the in vivo activity of DUBs. These are based on the sequence of Ub as the DUB-targeting motif and comprise a reactive C-terminal warhead such as vinyl methyl ester (VME), and an N-terminal epitope tag (59).

293T cells that reach up to 70% confluency were used for transfection. 48 hours after transfection, 293T cells were collected for the subsequent experiment as required. The following USP7-specific shRNA (Sigma-Aldrich, MISSION system) was used: shUSP7.1:5′-CCGGCCTGGATTTGTGGTTACGTTACTCGAGTAACGTAACCACAAATCCAGGTTTTT-3′ and shUSP7.2: 5′-CCGGCCAGCTAAGTATCAAAGGAAACTCGAGTTTCCTTTGATACTTAGCTGGTTTTT-3′. Retroviral NOTCH1 plasmid (gift from Iannis Aifantis, New York University) and retroviral JMJD3 plasmids (from addgene and Paul Khavari's group, wild-type #21212 and mutant #21214) were used for retrovirus production. Retrovirus was used to infect T-ALL cells as previously described (16).

A total of 500,000 cells were plated in wells of a 24-well plate, and treated with DMSO or USP7i as indicated in the Figure Legends. Cells were counted each day via trypan blue, and inhibitor/medium was changed. When cells reached a confluency of >1 million cells/mL, cells were diluted 1:2, and this dilution was factored into the cell numbers for viability assays. For apoptosis analysis, cells were stained with LIVE/DEAD Fixable Near-IR Dead Cell Stain (Life Technologies) according to the manufacturer's protocol, except that cells were stained for 20 minutes at 4°C, prior to staining with PE-conjugated Annexin V (Life Technologies) in Annexin V Binding Buffer (BD Biosciences), according to the manufacturers' instructions. To measure cell cycle, cells were fixed in 100-μL Fix and Perm Medium A (Life Technologies) for 15 minutes, washed with PBS, and incubated with 100-μL Fix and Perm Medium B (Life Technologies) + 1 μg/mL DAPI (Invitrogen) for 1 hour at 4°C. Flow cytometry was performed on an LSR II (BD Biosciences) and analyses were performed using FlowJo software (Tree Star). Statistical analyses were performed using GraphPad Prism software (GraphPad Software). Comparisons were made using the Student unpaired, 2-sided t test. P values <0.05 were considered statistically significant.

A total of 200 million JURKAT/CUTLL1 T-ALL cells were treated in duplicate with either DMSO or 5 μmol/L USP7i for 8 hours. Cell pellets were collected, washed with PBS, and frozen. Ubiquitinated substrates were pulled down using agarose-TUBE beads, eluted, and treated with the DUB USP2 to remove ubiquitin prior to Western blotting as previously described (57). The wild-type and mutant USP7-expressing plasmids are purchased by addgene (#46751 and #46752 correspondingly, from the Maertens and Peters labs).

RNA was extracted from up to 10 million T-ALL cells using the Aurum Total RNA Mini Kit (Bio-Rad) according to the manufacturer's instructions. RNA was quantified on a Qubit 3.0 Fluorometer (Life Technologies) using the Qubit RNA HS (High Sensitivity) Assay Kit (Life Technologies) according to the manufacturer's instructions. RNA integrity and DNA fragment size were determined using RNA Nano Chips and High Sensitivity DNA Chips read on a 2100 Bioanalyzer (Agilent Technologies). Libraries were prepared using Agencourt AMPure XP beads (Beckman Coulter) and the TruSeq RNA Library Prep Kit v2 (Illumina), according to the manufacturer's low sample (LS) protocol, and sequenced on the Illumina NextSeq 500 (50 bp single reads). FASTQ reads were aligned to the hg19 genome using tophat version 2.1.0 (60) using the following options—no-novel-juncs—read-mismatches 2—read-edit-dist 2—max-multihits 5. The generated bam files were then used to count the reads only at the exons of genes using htseq-count (61) with the following parameters -q -m intersection-nonempty -s reverse -t exon. Differential expression analysis was done using the R package edgeR (62). Bigwig tracks of RNA-Seq expression were generated by using the GenomicAlignments package in R in order to calculate the coverage of reads in CPM normalized to the total number of uniquely mapped reads for each sample in the library.

Superenhancers (SEs) were determined using a H3K27Ac ChIP-Seq dataset in JURKAT cells treated with DMSO or USP7 inhibitor. Aligned bam files of H3K27Ac ChIP-Seq signal and input control were analyzed using the Rank Ordering of Super-Enhancers algorithm (ROSE; refs. 63, 64). A stitching distance of 12.5Kb were used to stitch together enhancer regions, and regions within 2.5Kb of a transcriptional start site were ignored in order to prevent promoter bias. The initial set of enhancer regions used in the analysis was determined by merging the peaks of H3K4me1 and H3K27Ac in CUTLL1 cells (ChIP-Seq dataset for H3K4me1 signal and H3K27Ac was downloaded from GEO with the accession numbers GSM732910 and GSM1252938 for H3K4me1 and H3K27Ac, respectively).

Gene set enrichment analysis (GSEA) was done using the Broad Institute GSEA software (http://software.broadinstitute.org/gsea/index.jsp; refs. 65, 66). RNA-Seq gene expression data were used to create a preranked order of genes, where genes were sorted using the P value of the differential expression between USP7 inhibition and DMSO treatment, and gene up- or downregulation upon USP7 inhibition was determined. In brief, the genes were ranked by calculating –log(P value)*sign(logFC), and a PreRanked GSEA analysis was done using the Hallmark database and the standard weighted enrichment statistic with 5,000 permutations. Enriched datasets with a FDR < 0.05 were considered statistically significant.

All mice were housed in a barrier facility, and procedures were performed as approved by the Northwestern University Institutional Animal Care and Use Committee (protocol Ntziachristos #IS00002058 and Mazar #IS00000556) and the Ghent University Animal Ethics Committee (protocol #ECD16/56).

For CUTLL1 T-ALL subcutaneous studies, 0.7 million (PLKO.1 and shUSP7.1) cells were injected subcutaneously into the right flank of 8-week-old NOD.Cg-Prkdcscid male mice (#005557; Jackson Laboratories, Portage, MI) with an equal volume of BD Matrigel, at 50 μL of cells to 50-μL Matrigel. Body weight and tumor size (via calipers) were measured 3 times per week.

For JURKAT T-ALL subcutaneous studies, 0.7 million cells were injected subcutaneously into the right flank of 8-week-old NOD.Cg-Prkdcscid male mice (#005557; Jackson Laboratories) with an equal volume of BD Matrigel, at 50 μL of cells to 50-μL Matrigel. The following day, mice were randomly assigned to treatment groups (7 mice/group), and were treated with either DMSO or 10 mg/kg USP7i 3 times per week intraperitoneally for 8 doses. One week later, mice were treated 5 times per week intraperitoneally for 17 days. Body weight and tumor size (via calipers) were measured 3 times per week.

For JURKAT T-ALL intravenous transplantation studies, 1 million cells were injected via tail vein into 8-week-old NOD.Cg-Prkdcscid male mice (#005557; Jackson Laboratories) in 100-μL PBS. Animals were monitored by IVIS every 3 days until luciferase signal was detected, and then animals were randomly assigned to treatment groups (9 mice/group). Mice were treated 3 times per week intraperitoneally with DMSO or 10 mg/kg USP7i (in 10% DMSO + 2% Tween 80) for up to 10 doses. IVIS images were taken twice per week (Perkin Elmer) and animal weight was measured 3 times per week to ensure accurate dosing.

For primagraft studies, 1.2 million primary human T-ALL cells (from the spleen of leukemic primary xenograft engrafted with cells of T-ALL patient FV, N16/0267) in 150-μL PBS were injected via tail vein into 1.5-month-old female NGS mice (Jackson Laboratories). Two weeks after transplantation, blood was collected via tail nick for analysis of hCD45 expression (engraftment) using PE-conjugated mouse anti-human CD45 antibody (Miltenyi Biotec; clone 5B1). Flow cytometry data were analyzed using Flowjo software. One week later, mice were randomly divided into 2 groups (7 mice/group) and treated with either DMSO or 10 mg/kg USP7i in 2% Tween 80 + 10% DMSO in PBS. Treatment was administered on days 1 to 5 and 8 to 11 intravenously, and on days 12 and 15 to 22 intraperitoneally. Blood was collected via tail nick on days 17 and 22 to measure tumor burden by staining for hCD45.

For toxicity studies, 8-week-old NOD.Cg-Prkdcscid male mice (#005557; Jackson Laboratories) were randomly assigned to treatment groups (3 mice/group). Mice were treated with either vehicle (10% DMSO, 2% Tween-80, and 10% captisol in PBS), 10 mg/kg USP7i, 10 mg/kg USP7i + 10 mg/kg GSKJ4, 10 mg/kg USP7i + 20 mg/kg GSKJ4, or 10 mg/kg USP7i + 50 mg/kg GSKJ4 i.v. daily for 5 days to evaluate treatment toxicity.

For combination treatment studies, 1 million JURKAT T-ALL cells were injected subcutaneously into the right flank of 8-week-old NOD.Cg-Prkdcscid male mice (#005557; Jackson Laboratories) with an equal volume of BD Matrigel, at 50 μL of cells to 50-μL Matrigel. After tumors reached 150 to 200 mm³, mice were randomly assigned to treatment groups and treated intravenously with vehicle (10% DMSO, 2% Tween 80, and 10% captisol in PBS; n = 10), 10 mg/kg USP7i (n = 6 per inhibitor), or 10 mg/kg USP7i + 50 mg/kg GSKJ4 (n = 10) 5 times the first week, and then 3 times per week until tumors reached 1.5 cm³.

Analyses were performed using GraphPad Prism software (GraphPad Software). Statistical comparisons were made using the Student unpaired, 2-sided t test. P values <0.05 were considered statistically significant.

Mouse blood samples were collected via cardiac puncture and run on a Hemavet 950 FS (Drew Scientific, Inc.) to obtain blood cell counts.

This was performed as described in the past by Milano and colleagues (67), and Real and colleagues (11).

A total of 3,000 cells per well were seeded using a microplate Dispenser (MultiFlo, BioTek) in 384-well clear bottom, black wall plates (Corning). Drugs were added using the Tecan D300e digital dispenser (Tecan). After 72-hour incubation, alamarBlue cell viability reagent (ThermoFisher) was added and viability was quantified by measuring fluorescence in a plate reader (Tecan Infinite m1000 pro, E_x: 530 nm; E_λ: 590). Synergy analysis was conducted using Compusyn software (68).

The studies were performed as previously described (69). JURKAT cells were treated with compound or DMSO for 5 hours. Cells were washed once with PBS, harvested, and lysed in 1% NP40 lysis buffer (50 mmol/L Tris-HCl, pH 7.5, 150 mmol/L NaCl, 1% NP40, 10% Glycerol, 1 mmol/L PMSF, 2 mmol/L β-ME). Twenty micrograms of lysate per sample was incubated at 37 °C for 30 minutes with Ub-VME at a final concentration of 400 nmol/L in a total reaction volume of 10 μL. The reaction was stopped by adding 2 μL of 6× SDS sample buffer and boiling for 5 minutes, followed by SDS-PAGE electrophoresis, and transfer to PVDF membranes. The blots were probed with antibodies against USP7 and USP24 to determine the formation of DUB-Ub-VME complex.

To identify potential NOTCH1/JMJD3-associated proteins and understand NOTCH1 complex stability in leukemia, we retrovirally expressed hemaglutinin (HA) tagged-JMJD3 in CCRF-CEM T-ALL cells. We performed LC/MS-MS following immunoprecipitation (IP) of HA-JMJD3 (Fig. 1A). One hundred and one proteins were uniquely associated with HA-JMJD3, and these proteins were enriched in members of the ubiquitin-specific protease (USP) family of DUBs, including USP7, USP9X, and USP24 (Fig. 1B; Supplementary Table S1; Supplementary Fig. S1A). Indeed, gene ontology analysis of the JMJD3 interactome showed enrichment in proteins involved in various metabolic processes, cell cycle progression and members of the proteasome complex (Supplementary Fig. S1B).

Dynamic NOTCH1 (de)ubiquitination plays a pivotal role in the regulation of NOTCH1 protein levels in leukemia, and ubiquitin ligases such as FBXW7 have been demonstrated to regulate NOTCH1 degradation and are tumor suppressors in T-ALL (70, 71). To this end, we hypothesized that USPs might be acting as oncogenic co-factors for the NOTCH1 complex, controlling complex stability and, ultimately, transcriptional potency. Ubiquitination occurs through a sequence of enzymatic reactions involving the activity of 2 E1 ubiquitin-activating enzymes, multiple E2 enzymes and hundreds (∼700) of E3 enzymes in humans. E3 ligases determine complex specificity (72). In contrast, DUBs (73, 74) mainly include the ubiquitin-specific protease superfamily (USP/UBP, 58 members), the ovarian tumor (OTU, 14) superfamily, the Machado-Josephin domain (MJD, 5) superfamily, the ubiquitin C-terminal hydrolase (UCH, 4) superfamily (all the aforementioned categories are cysteine proteases), and the Jab1/Mov34/Mpr1 Pad1 N-terminal+ (MPN+) (JAMM, 14) domain superfamily proteins that binds zinc (metalloproteases).

USP7 (32–43) was a top interacting partner of JMJD3. USP7 might also interact with NOTCH1 (75) and has a demonstrated role in other hematologic malignancies, in multiple myeloma in particular (76). We confirmed the interaction of NOTCH1 and JMJD3 with USP7 in reciprocal IP experiments (Fig. 1C and D). USP24 and USP9x (77), in contrast, have not been found to interact with NOTCH1 (75). Indeed, our pull-down studies for NOTCH1, followed by detection of USP24 and USP9x in Western blot experiments, showed that neither USP9x nor USP24 interact with NOTCH1 (Supplementary Fig. S1C and S1D). Furthermore, our studies show that, in contrast to USP7 and NOTCH1, both USP9x and USP24 are mainly localized in the cytoplasm (Supplementary Fig. S1E).

Given the potential participation of USP7 in NOTCH1 protein complexes, we studied the levels of USP7 mRNA in cancer using data from 176 cancer cell lines (cancer cell line encyclopedia). Our analysis showed that USP7, like NOTCH1, is significantly overexpressed in T-ALL compared with other hematologic and solid tumors (Fig. 2A). We then used reverse phase protein array (RPPA) analysis (78, 79) to quantify NOTCH1 and USP7 protein levels in a panel of 64 pediatric leukemia patient samples. We noticed that USP7 and intracellular or total NOTCH1 levels significantly correlate in T-ALL (Fig. 2B; Supplementary Fig. S2A). We then hypothesized that NOTCH1 might directly bind and control the USP7 gene locus. Analysis of ChIP-Seq data in CUTLL1 T-ALL cells showed that NOTCH1 directly binds the promoter of the USP7 gene (Fig. 2C). We also demonstrated that inhibition of NOTCH1, using an established method for NOTCH1 inhibition, using gamma secretase inhibitors (GSI; refs. 11, 15), decreased NOTCH1 binding enrichment on the USP7 gene locus as well as other well-characterized NOTCH1 targets, DTX1 and NOTCH3 (Supplementary Fig. S2B). Furthermore, GSI treatment or NOTCH1 silencing using short hairpins against NOTCH1 (shNOTCH1), led to downregulation of USP7 mRNA and protein levels and other well-characterized NOTCH1 targets (Fig. 2D; Supplementary Fig. S2C–S2E). In contrast, NOTCH1 overexpression leads to a transcriptional upregulation of USP7 (Supplementary Fig. S2F).

As USP7 has a characterized enzymatic activity as deubiquitinating enzyme, we hypothesized that it might control NOTCH1 protein levels through active deubiquitination. We silenced USP7 in T-ALL using short hairpin RNAs (shUSP7) and examined NOTCH1 protein levels, to observe a decrease in the levels of NOTCH1, suggesting that USP7 might control NOTCH1 protein stability (Fig. 3A). We also noticed a marked inhibition of T-ALL growth upon shUSP7 compared with the control (Fig. 3A; Supplementary Fig. S3A and S3B). USP7 silencing led to increased apoptosis and a decrease in S phase of T-ALL cells, as demonstrated using shUSP7 (Supplementary Fig. S3C and S3D). We then took advantage of established USP7-small molecule inhibitors (USP7i, see also Materials and Methods), developed by Progenra Inc., and successfully used in preclinical models of multiple myeloma and other types of cancer and in immunologic contexts (57, 76, 80–84), to block USP7 activity in leukemia cells. The efficacy and specificity of those compounds has been previously demonstrated in several systems (57, 76, 80–84). USP7i significantly inhibited the growth of various T-ALL lines, including CUTLL1, JURKAT, and CEM, in the micromolar (μmol/L) range of concentrations, similarly to shUSP7 (Fig. 3B and C; Supplementary Fig. S4A and 5A). The compound inhibits USP7 activity in a dose-dependent manner from 1 to 5 μmol/L (Supplementary Fig. 4B), without significantly blocking total DUB activity in T-ALL, even at 5 μmol/L concentrations (Supplementary Fig. S4C). Treatment of T-ALL cells with USP7i for 1 and 3 days coupled to apoptosis and cell-cycle analysis, using Annexin V and DAPI staining respectively, led to a significant increase in apoptosis and a decrease in S phase (Supplementary Fig. S5B and S5C), similarly to the shUSP7.

To confirm the role of USP7 in NOTCH1 stabilization, we expressed wild-type or a catalytic mutant of USP7 together with NOTCH1, and measured NOTCH1 expression in 293T cells. Indeed, NOTCH1 was stabilized in the presence of wild-type USP7 (Fig. 3D, right). Similarly, we show that only the wild-type but not the catalytically inactive USP7 can efficiently deubiquitinate NOTCH1 (Fig. 3D, left). These results show that USP7 and its catalytic activity are critical for NOTCH1 deubiquitination and stabilization. To evaluate the role of USP7 in NOTCH1 ubiquitination in T-ALL, we inhibited USP7 catalytic activity, pulled down ubiquitinated proteins using tandem ubiquitin binding entity (TUBE) technology (85), and measured NOTCH1 levels (Supplementary Fig. S6A). We then performed similar cell treatments followed by Flag-tagged NOTCH1 pull-down and detection of ubiquitination using antibodies against lysine 48 (K48)-linked poly-ubiquitin chains, a type of polyubiquitination coupled to protein degradation through proteasomal activity (86). We demonstrated an increase of NOTCH1 poly-ubiquitination upon USP7i treatment (Fig. 3E). NOTCH1 is a critical target of USP7, as NOTCH1 overexpression partially rescues USP7i-associated blockade in T-ALL growth and expression of NOTCH1 targets in T-ALL cells (Supplementary Fig. S7A–S7C). Similar to NOTCH1, we demonstrated a significant decrease in the levels of JMJD3 upon silencing of USP7 (Supplementary Fig. S6B). Treatment with USP7i led to a similar decrease in NOTCH1 and JMJD3 protein levels, starting at 1 μmol/L USP7i (Fig. 3F). Likewise, inhibition of USP7 in T-ALL cells leads to an increasing of K48-linked ubiquitination of JMJD3 (Supplementary Fig. S6C). JMJD3 overexpression using a retroviral system could not rescue the effect of the drug on leukemic cells (Supplementary Fig. S7D), potentially due to the simultaneous requirement for high levels of NOTCH1 expression for leukemia maintenance.

Because of the activity of intracellular NOTCH1 (N1-IC) on chromatin, we evaluated genome-wide USP7 binding via chromatin immunoprecipitation combined with sequencing (ChIP-Seq). We identified ∼10,000 high-confidence peaks for USP7 that localized near promoters (Fig. 4A) and had a strong overlap with NOTCH1 binding sites (Fig. 4B). Both USP7 and NOTCH1 bound to classical NOTCH1 targets, such as DTX1 and NOTCH3, as well as NOTCH1 itself (Supplementary Fig. S8A). Pathway analysis of the ChIP-Seq data showed a striking enrichment for NOTCH1 signaling (Fig. 4C). Furthermore, NOTCH1 recruited USP7 to chromatin, as treatment with GSI for depletion of intracellular NOTCH1 levels significantly decreased chromatin-bound USP7 levels (Fig. 4D). USP7 controls NOTCH1 stability and inhibition of USP7 led to a reduction of global NOTCH1 levels and depletion of NOTCH1 from chromatin (Fig. 4D). We further performed USP7 and NOTCH1 ChIP studies focusing on USP7 targets upon USP7 inhibition. Our results show that NOTCH1 and USP7 binding to chromatin is significantly affected upon USP7i treatment whereas binding of another T-ALL transcription factor, RUNX1 (used as control) remained largely unaffected (Supplementary Fig. S8B). Global cellular levels of USP7 did not change over the period of treatment (Fig. 4D). These findings suggest that USP7 and NOTCH1 form a positive feedback loop (see also our discussion), where NOTCH1 induces USP7 gene expression and then recruits USP7 to chromatin, resulting in increased expression of NOTCH1 targets, through stabilization of the NOTCH1 complex.

We and others have shown that NOTCH1 activity leads to eviction of the repressive mark H3K27me3 from the promoters of oncogenic targets (15, 16, 87). We hypothesized that inhibition of USP7 would lead to an increase in H3K27me3 levels through reduction of JMJD3 protein levels. We assessed the levels of H3K27me3 on USP7 targets upon USP7i treatment for 24 hours. Indeed, USP7 and NOTCH1 target loci presented a significant gain of H3K27me3 upon USP7i treatment (Fig. 4E; Supplementary Fig. S8C). As USP7 has been shown to control histone ubiquitination (88, 89), we also examined histone H2B lysine 120 ubiquitination (H2Bub) levels following USP7 inhibition. We showed that USP7 colocalized with genes exhibiting a gain of H2B ubiquitination upon USP7 inhibition (Fig. 4E; Supplementary Figs. S8C and S9; refs. 90, 91). These changes in H2Bub and H3K27ac/me3 seemed to localize to USP7 targets and not a genome-wide phenomenon, as global levels of these marks were not significantly altered following treatment with USP7i as we demonstrated using global bottom-up proteomics (Supplementary Fig. S10; ref. 92). Changes in H2A lysine 119 ubiquitination (H2Aub) on downregulated loci were minimal but detectable, potentially associated to H3K27me3 changes (Fig. 4E; Supplementary Figs. S8C and S9; ref. 93).

Our data suggest that USP7 regulates gene expression of critical NOTCH1 targets. To delineate the molecular effects of USP7 activity, we performed global expression analysis using RNA-Seq upon treatment with USP7i, and comparison with GSI (NOTCH1 inhibitor), and GSKJ4 (JMJD3 inhibitor). The transcriptional signatures of NOTCH1, USP7, and JMJD3 chemical inhibition overlapped significantly, both with regards to downregulated as well as upregulated genes (Fig. 5A and B, top; Supplementary Fig. S11A–S11C). Specifically, inhibition of USP7 led to downregulation of NOTCH1 targets, including NOTCH3 and DTX1 (Fig. 5B and C; Supplementary Fig. S11C). Similar to enzymatic inhibition, USP7 silencing led to downregulation of NOTCH1 targets (Fig. 5A and B; bottom) and significant overlap with USP7i treatment (Supplementary Fig. S11D). Gene set enrichment analysis upon USP7 inhibition showed that major oncogenic pathways (NOTCH1, MYC, DNA damage response, and metabolism) were downregulated (Fig. 5D; Supplementary Fig. S12). Together, these data show that USP7 enzymatic activity is important for sustaining oncogenic activity in T-ALL.

It is well known that NOTCH1 regulates oncogenic enhancers (94, 95). Analysis of USP7 ChIP-Seq data showed that leukemia-specific SEs (65, 96), as determined by H3K27 acetylation signal intensity (Supplementary Fig. S13A), were enriched for binding of USP7 and NOTCH1: 192 out of 254 leukemia SEs were cobound by USP7 and NOTCH1. In agreement with the association of SEs with transcriptional activation, we found that the USP7-bound genes were more highly expressed compared to the rest of the transcriptome (Supplementary Fig. S13B). We hypothesized that combinatorial inhibition of USP7 and the SE component BRD4 using JQ1 inhibitors might be an additive or synergistic effect in inhibiting T-ALL growth. Indeed, our studies using combinatorial inhibition of the bromodomain proteins that recognize acetylated histones (JQ1; refs. 97, 98) and USP7 showed a synergistic effect (Supplementary Fig. S13C) on T-ALL inhibition. This was further underlined by our targeted and genome-wide analysis of USP7 interactions showing that USP7 interacts with Mediator 1, a main component of SEs (Ntziachristos group, unpublished), suggesting that USP7 may regulate gene activation via interactions with the T-ALL transcriptional machinery in SE areas.

These findings show that USP7 inhibition is a valid therapeutic tool in high-risk leukemia, where other treatments have failed to lead to significant disease regression. Silencing of USP7 in human: mouse xenograft models, using CUTLL1 T-ALL cells, led to a significant inhibition of tumor growth in animals (Fig. 6A) coupled to an extension of mouse survival (Fig. 6B). This significant finding led us to assess the potential of chemical inhibition of USP7 alone or in combination with JMJD3, using the GSKJ4 compound, to block leukemia growth in vivo. Because of the role of the H3K27me3 epigenetic mark in the regulation of NOTCH1 targets in T-ALL progression, we and others have successfully used GSKJ4 to target JMJD3 in T-ALL in the past with decent effect on leukemia inhibition (16, 99). We initially evaluated inhibitor toxicity by injecting mice intravenously with vehicle, USP7i alone (10 mg/kg body mass), or USP7i with increasing concentrations of GSKJ4 (up to 50 mg/kg body mass). The GSKJ4 concentrations used have been previously shown to successfully inhibit tumor growth (99, 100). Immunocompromised mice treated daily for 5 days showed no signs of treatment-associated toxicity, as determined by complete blood cell count, body weight, and analysis of spleen and liver (Supplementary Fig. S14A and data not shown). We also examined USP7i-treated animals for gastrointestinal (GI) toxicity as GSI treatment for direct NOTCH1inhibition leads to goblet cell metaplasia in animals. Our studies show no signs of toxicity associated with USP7i inhibition in the GI track (Supplementary Fig. S14B). We then transplanted luciferase-expressing JURKAT T-ALL cells either intravenously or subcutaneously into immunocompromised mice (15, 16). Upon tumor detection using bioluminescence imaging (IVIS), mice were randomized into different groups that were treated with vehicle or USP7i (10 mg/kg). In both models, tumors in USP7i-treated animals showed a significant growth disadvantage compared to the control group (Fig. 6C; Supplementary Fig. S15A). Similar results were obtained using a primagraft intravenous model using a patient sample with mutations affecting the gene of the main ubiquitin ligase for NOTCH1, FBXW7, and high NOTCH1 expression levels as well as CCND3 mutation (Supplementary Table S2), where T-ALL progression was detected in the peripheral blood using the marker human CD45 (Supplementary Fig. S15B and S15C). To further potentiate the therapeutic impact of USP7i in T-ALL, we investigated the effect of combinatorial USP7 and JMJD3 inhibition on leukemia growth by treating mice with USP7i (10 mg/kg body mass) and GSKJ4 (50 mg/Kg body mass) in a subcutaneous model of T-ALL and monitored animal survival (Fig. 6D). Similar to our previous studies (Fig. 6C; Supplementary Fig. S15B), we found that USP7i as a single therapy significantly inhibits tumor growth (Fig. 6D, right) and extends mouse survival (Fig. 6D, left). Moreover, combinatorial inhibition significantly decreased tumor growth, more effectively than vehicle or USP7 inhibition alone (Fig. 6D, right) without any associated toxicity based on our analysis of mouse weight (Supplementary Fig. S15D) and histology studies (not shown). Moreover, we demonstrated significant survival differences between mice injected with vehicle or those injected with a combination of USP7 and JMJD3 inhibitors (Fig. 6D, left). USP7i/GSKJ4 treatment yielded a prolonged mouse survival compared to vehicle or USP7i therapy (Fig. 6D, left and data not shown), lending rationale to the use of combinatorial drug treatments against epigenetic regulators in T-cell leukemia.

Therapeutic targeting of high-risk ALL has been challenging, rendering this disease an unmet clinical need. In this study, we performed a series of epigenetic, biochemical, functional and pharmacologic studies to show that the DUB USP7 functionally and physically interacts with and controls the NOTCH1 pathway and ultimately the oncogenic transcriptional circuitry in T-cell ALL. USP7 inhibition of its enzymatic activity could be a valid therapeutic target in this disease.

Given the previously unrecognized importance of deubiquitination in acute leukemia progression, we characterized a novel role for USP7 in controlling NOTCH1 stability T-ALL, providing the first evidence that the ubiquitination–methylation axis can serve as an oncogenic switch and therapeutic target in T-ALL. USP7 is the first DUB identified and thoroughly characterized to regulate the NOTCH oncogenic pathway. Our studies showed that USP7, NOTCH1, and JMJD3 act in a positive feedback loop (Supplementary Fig. S16), where the NOTCH1/JMJD3 complex induces expression of USP7, and it subsequently results in USP7 recruitment to target genes. Ultimately, USP7 recruitment leads to stabilization of the oncogenic complex and activation of targets through demethylation of H3K27 and deubiquitination of H2B, as well as changes in H3K27Ac at SEs. Past and ongoing studies in our group aim at mapping NOTCH1 areas that interact with USP7. NOTCH1 molecules interact with and are controlled by the E3 ubiquitin ligase FBXW7 via their PEST domain and engineered PEST mutants (theonine 2512 to alanine, T2512A) for NOTCH1, are impervious to FBXW7 regulation (70, 71). In contrast, we demonstrate that this NOTCH1 mutant interacts with USP7, suggesting that other parts of the NOTCH1 intracellular fragment might control its binding to and regulation via USP7 (Supplementary Fig. S17A). To this end, we generated a series of NOTCH1 truncations to demonstrate that the ankyrin domain of NOTCH1 (amino acids 1850–2123) mediates interaction with USP7 (Supplementary Fig. S17B). Although further studies might be required to further map the exact interacting amino acid regions on USP7 and NOTCH1, our findings potentially suggest a novel mode of post-translational regulation of NOTCH1 independent of the PEST domain and FBXW7 ligase. Of note, there are no hitherto identified ankyrin mutants for NOTCH1 in T-ALL (10). We mined 264 T-ALL patient data available through the pediatric cancer genome project (PeCan) data portal (https://pecan.stjude.cloud/home), in our search for NOTCH1 mutations. No mutations in the ankyrin domain of NOTCH1 were identified, in agreement with published evidence.

A better understanding of H3K27me3 demethylation and H2B deubiquitination changes and whether levels of those 2 marks are intertwined provide us with information useful in designing epigenetic therapies. Intriguingly, and in contrast to the previously characterized pro-oncogenic role of USP7 in T-ALL and multiple myeloma (76), neuroblastoma (81), chronic lymphocytic leukemia (101), and other solid tumors (102–104), there are mutations affecting USP7 in pediatric cancers, TAL1-positive leukemias in particular (105–107). Moreover, it is known that USP7 can have different substrates, including PTEN (35), and the p53–HDM2 axis (36–38, 108, 109). Thus, similarly to what we and others have shown for another epigenetic modulator, the histone demethylase UTX, which can play dual roles (oncogene or tumor suppressor) in NOTCH1- and TAL1-postitive contexts of T-ALL (16, 99, 69), the role of USP7 might also be context-specific. These context-specific roles of the epigenetic modulators are another intriguing aspect of their biology that we can exploit to develop elegant targeted therapeutic approaches in cancer. Further research is needed to study the role of USP7 in leukemia contexts other than NOTCH1-positive T-ALL, such as TAL1-positive leukemia.

Our findings strongly suggest that USP proteins, and USP7, in particular, may be exploited for pharmacological inhibition in certain T-ALL patients. Previous studies using the same USP7 inhibitor backbone in mouse models showed that the drug inhibits USP7 in specific physiological mouse systems and in mouse tumor cells (57, 81, 110). This is not surprising as the USP7 gene/protein is significantly conserved between human and mouse. Although further studies are needed to directly compare the effect of the drug on human and mouse tissues, these published and our findings suggest that USP7 compounds might have significant efficacy with minimal toxicity in clinical trials against patients with T-ALL. The therapeutic window for USP7 inhibition in NOTCH1+ T-ALL might be explained by the significantly higher levels of USP7 in this disease compared to physiologic tissues. Combinations of targeted therapies are hailed as the future of cancer therapy, with hematological malignancies leading the way (i.e. bortezomib, lenalidomide, and dexamethasone for multiple myeloma, and rituximab and ibrutinib for chronic lymphocytic leukemia). As epigenetic modulators are 1 of the main gene families controlling tumor biology, we feel our study can provide insights in the use of DUBs inhibitors and combination with chemotherapy or other epigenetics inhibitors for therapy in leukemia. Three recent studies characterized small molecule inhibitors with efficacy against USP7 in different cancer contexts, including colorectal carcinoma, osteosarcoma and prostate cancer (111–113). Thus, combined inhibition of USP7 and JMJD3 may represent an attractive therapeutic approach for T-ALL.

N. Hijiya reports receiving speakers bureau honoraria from Novartis. N. L. Kelleher is an employee of Quest Diagnostics and Thermo Fisher, and is a consultant/advisory board member for Quest Diagnostics. J. Weinstock is an employee of Progenra. No potential conflicts of interest were disclosed by the other authors.

The raw files and the Proteome discoverer results are now accessible at MassIVE accession: MSV000081771 and ProteomeXchange accession: PXD008330. For now, the log in information for reviewers at MassIVE is:

Username: MSV000081771_reviewer

Password: reviewersonly

The raw MS and skyline files for the epiproteomics analysis are deposited to PeptideAtlas (http://www.peptideatlas.org/)

Official URL for this dataset: ftp://PASS01134:[email protected]/ To access files via FTP, use credentials:

Servername:ftp.peptideatlas.org

Username: PASS01134

Password:HG6634ca

Conception and design: Q. Jin, K.M. Arcipowski, Y. Zhu, B.T. Gutierrez Diaz, M.R. Johnson, A.G. Volk, S. Kumar, J.D. Crispino, P. Ntziachristos

Development of methodology: Q. Jin, Y. Zhu, B.T. Gutierrez Diaz, K.K. Wang, A.G. Volk, I. Kandela, N.L. Kelleher, S. Kumar, P. Ntziachristos

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Q. Jin, K.M. Arcipowski, Y. Zhu, K.K. Wang, M.R. Johnson, A.G. Volk, F. Wang, C. Grove, H. Wang, I. Sokirniy, P.M. Thomas, Y. Ah Goo, N.A. Abshiru, N. Hijiya, S. Peirs, N. Vandamme, G. Berx, S. Goosens, S.A. Marshall, E.J. Rendleman, Y.-H. Takahashi, R. Rawat, B. Ueberheide, C. Mantis, I. Kandela, J.-P. Bourquin, B. Bornhauser, S. Bresolin, M. Paganin, B. Accordi, G. Basso, P. Ntziachristos

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Q. Jin, C.A. Martinez, K.M. Arcipowski, B.T. Gutierrez Diaz, K.K. Wang, A.G. Volk, F. Wang, I. Sokirniy, P.M. Thomas, Y. Ah Goo, N. Hijiya, S. Peirs, S.A. Marshall, E.T. Bartom, C.K. Collings, S. Kelly, B. Ueberheide, I. Kandela, S. Bresolin, B. Accordi, G. Basso, N.L. Kelleher, S. Kumar, J.D. Crispino, P. Ntziachristos

Writing, review, and/or revision of the manuscript: Q. Jin, C.A. Martinez, K.M. Arcipowski, B.T. Gutierrez Diaz, M.R. Johnson, A.G. Volk, Y. Ah Goo, N. Hijiya, S. Peirs, R. Rawat, B. Accordi, S. Kumar, J.D. Crispino, P. Ntziachristos

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): Q. Jin, Y. Zhu, B.T. Gutierrez Diaz, K.K. Wang, M.R. Johnson, J. Wu, C. Grove, L. Wang, A. Strikoudis, A. Shilatifard, P. Ntziachristos

Study supervision: Q. Jin, N.L. Kelleher, S. Kumar, P. Ntziachristos

Other (synthesize compounds): C. Grove

Other (performed experiments and contributed ideas): R. Rawat

Other (performed experiments): V. Serafin

Other (clinical data): G. Basso

Other (design some of the compounds which were used): J. Weinstock

This work was supported by the NIH T32CA080621-11 Oncogenesis and Developmental Biology Training Grant and Cancer Smashers Postdoctoral Fellowship (to K.M. Arcipowski), and by the National Cancer Institute (R00CA188293-02), the American Society of Hematology, the Leukemia Research Foundation, the St. Baldrick's Foundation, the H Foundation, the Gabrielle's Angel Foundation, and the Zell Foundation (to P. Ntziachristos). This work is also supported by AIRC IG 19186 grant to G. Berx. High-throughput sequencing data have been deposited into Gene Expression Omnibus, accession number GSE97435. Proteomics services were performed by the Northwestern proteomics core, generously supported by NCI CCSG P30 CA060553 awarded to the Robert H Lurie Comprehensive Cancer Center and the National Resource for Translational and Developmental Proteomics supported by P41 GM108569. For some of the in vivo studies we collaborated with the Developmental Therapeutics Core (DTC), also supported by Cancer Center Support Grant P30 CA060553 from the National Cancer Institute.

We want to thank all members of the Ntziachristos laboratory for critical review of the manuscript and their comments. We want to thank Miklos Bekes and Tony Huang (New York University) for providing USP7-expressing plasmids.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.