Focused CRISPR-Cas9 genetic screening reveals USO1 as a vulnerability in B-cell acute lymphoblastic leukemia

Post-transcriptional gene regulation, including that by RNA binding proteins (RBPs), has recently been described as an important mechanism in cancer. We had previously identified a set of RBPs that were highly dysregulated in B-cell acute lymphoblastic leukemia (B-ALL) with MLL translocations, which carry a poor prognosis. Here, we sought to functionally characterize these dysregulated RBP genes by performing a focused CRISPR dropout screen in B-ALL cell lines, finding dependencies on several genes including EIF3E, EPRS and USO1. Validating our findings, CRISPR/Cas9-mediated disruption of USO1 in MLL-translocated B-ALL cells reduced cell growth, promoted cell death, and altered the cell cycle. Transcriptomic analysis of USO1-deficient cells revealed alterations in pathways related to mTOR signaling, RNA metabolism, and targets of MYC. In addition, USO1-regulated genes from these experimental samples were significantly and concordantly correlated with USO1 expression in primary samples collected from B-ALL patients. Lastly, we found that loss of Uso1 inhibited colony formation of MLL-transformed in primary bone marrow cells from Cas9-EGFP mice. Together, our findings demonstrate an approach to performing focused sub-genomic CRISPR screens and highlight a putative RBP vulnerability in MLL-translocated B-ALL, thus identifying potential therapeutic targets in this disease.

B-ALL is the most common type of leukemia in the pediatric population, and is characterized by a number of recurrent chromosomal rearrangements [1][2][3][4] . Among these, the t (4;11) MLL-AF4 (KMT2A-AFF1) translocation gives rise to a highly aggressive form of B-ALL 5,6 . Patients with MLL-rearranged B-ALL have a dismal prognosis, with 5-year event-free survival rates hovering at 33.6% for infants 7 and 50% for older children and adults 8 . Most of these patients are resistant to conventional treatment with chemotherapy and steroids 9 , with bone marrow transplantation being the only curative therapeutic alternative 10 . Although recent developments such as CAR-T therapy 11 and anti-CD19 based therapy such as Blinatumomab 12 have raised hope for such patients 13 , antigen escape and lineage infidelity in MLL-translocated leukemia have proved problematic 14 . Therefore, there is an urgent need to better characterize potential therapeutic targets with high specificity.
The MLL-AF4 translocation engenders a unique transcriptional profile, as the fusion protein juxtaposes a histone methyltransferase (MLL, also known as KMT2A) with a protein that is involved in transcriptional regulation (AF4, or AFF1). Recently our lab carried out a study examining the expression of RBPs in B-ALL 15 , including both known and predicted RBPs 16 . In our analysis, we identified 36 RBPs that are highly upregulated in MLL-AF4 translocated B-ALL 15 . To study the importance of these genes in B-ALL, we implemented the powerful gene editing technique, CRISRP/Cas9, to perform a rapid and medium-throughput assessment of gene function. Genome wide CRISPR/Cas9 screens on AML cell lines have identified multiple gene targets critical for cell proliferation and survival 17 , but similar studies have not been performed in MLL-AF4 translocated B-ALL.  Fig. 2A & 2B). To query their regulation by the MLL-AF4 fusion protein, we analyzed their expression following inhibition with I-BET151 18 , a BRD4 BET domain inhibitor, MI-503 menin-MLL1 inhibitor 19 and EPZ5676, a DOT1L inhibitor 20 , all known to inhibit MLL-dependent gene expression regulation. With increasing doses of I-BET151, MI-503 and EPZ5676 there was a decrease in the USO1 mRNA expression level in SEM ( Fig. 2A), and RS4;11 cells ( Fig. 2B), but not in NALM6 cells (Fig. 2C as well as Supplementary Fig. 2C, 2D & 2E), suggesting that USO1 expression is MLL-AF4 dependent. A consistent reduction in EIF3E and EPRS was not observed in the MLL translocated cell lines. We further queried the MLL dependence of USO1 by over-expressing the MLL-Af4 transgene 21 in murine bone marrow cells. Western Blot and RT-qPCR analysis showed upregulation of MLL1 and USO1 in MLL-Af4 transduced cells compared to control cells ( Fig. 2D and Supplementary Fig. 2F). Using publicly available data 21 , we found that the USO1 and EIF3E genes demonstrated multiple MLL-AF4 binding sites within the 5'UTR, the first exon, and the first intron ( Fig. 2E and Supplementary Fig. 2G). In contrast, the EPRS gene did not show strong MLL-AF4 binding sites ( Supplementary Fig. 2H). EIF3e and EPRS were previously reported to be "common essential genes", per the depmap portal 22 , whereas USO1 was not such a gene, suggesting its potential utility as a novel clinical target. To confirm our findings of USO1 dependence on MLL-AF4, we designed chromatin immunoprecipitation (ChIP) experiments, designing primers for the regulatory regions of USO1 including the 5' UTR within the first exon. We found significant enrichment of the USO1 regulatory regions in both the MLL1 and AF4 pulldowns (Fig. 2F). Treatment of the cells with I-BET151 inhibited the association of both MLL1 and AF4 with the promoter region of USO1 (Fig. 2G). Together, these findings indicate that USO1 is a direct target of the MLL-AF4 transcriptional program. USO1 depletion alters B-ALL cell proliferation, survival and cell cycle. USO1 expression is upregulated in several types of cancer including B-ALL with MLL-AF4 translocations [23][24][25] . To characterize the functional role of USO1 in B-ALL suggested by our CRISPR screen, we utilized the previously mentioned twovector lentivirus system to transduce SEM cells with three different lentiviral constructs containing sgRNAs that target different regions of the USO1 gene (Fig. 3A, B). We found that two sgRNAs (sg2 and sg3) caused a significant downregulation of USO1 protein and mRNA expression by approximately 80% in SEM cells (Fig. 3C, D). Cas9-mediated frameshift mutation in USO1 was confirmed using TIDE assay 26 (Supplementary Fig. 3A). The lack of complete ablation was a reproducible finding in all cell lines we tested, as multiple single-cell cloning experiments failed to produce cells with complete knockout and multiple bulk cultures experiments showed  (Fig. 3F, G). Increased cell death was also observed in the USO1-depleted cells by Annexin V staining (Fig. 3H, I). Interestingly, USO1-depleted cells treated with I-BET151 also showed increased cell cycle arrest and apoptosis ( Supplementary Fig. 3D-3G), suggesting an additive effect with this inhibitor of BRD4. To confirm our findings in an orthogonal system, we introduced siR-NAs targeting USO1 using nucleofection. In these short-term assays, we found that there was partial reduction of USO1 mRNA, as expected with siRNA-mediated knockdown ( Supplementary Fig. 4A) and increased Annexin V staining ( Supplementary Fig. 4B, C). There were also increased cells in Sub-G0/G1 and a modest reduction in cell proliferation as measured by the MTS assay ( Supplementary Fig. 4E, F). Together, these observations confirm the importance of USO1 in regulation of cell cycle and survival of B-ALL cells.   www.nature.com/scientificreports/ USO1 impacts pathways related to cellular proliferation and RNA homeostasis. To evaluate the effect of USO1 depletion on the overall gene expression pattern in B-ALL cells, we performed RNA-Seq on SEM cells that were depleted of USO1 by CRISPR/Cas9 (three biological replicates per condition). Following differential expression analysis of USO1-depleted and control cells, we utilized the Metascape algorithm and Gene set enrichment analysis (GSEA) to assess pathway enrichment 27,28 . We found that significantly downregulated pathways included MTOR, ERB2, and Hallmark Hypoxia, while upregulated terms included Metabolism of RNA and Hallmark MYC targets ( Fig. 4A and Supplementary Fig. 5A). Analysis of Gene Ontology -Molecular Function gene sets revealed positive enrichment of several gene sets related to RNA homeostasis in the USO1depleted cells ( Supplementary Fig. 5B). Selected examples of up-and down regulated genes (log2FC > 1.5 and Padj. value < 0.01), from the enriched pathway identified by GSEA are highlighted in the Volcano plot for RNAseq data (Fig. 4B). RT-qPCR was used to confirm these same significantly upregulated (TFRC, MRPS12, and PSMD1) (Fig. 4C) and downregulated genes (BAPIP2, ABCA1 and BTAF1) (Fig. 4D) in USO1-depleted SEM cells. This led us to hypothesize that there were alterations in expression and activation of MTOR in USO1depleted cells using western blotting. We observed there was a mild downregulation of p-MTOR (Ser2481) in the USO1-depleted cells (Fig. 4E). Hence, it appears that USO1 regulates several pathways that are known to play a role in cell survival and cell death 29 .
To assess whether the gene expression changes caused by USO1-depletion had any clinical relevance, we turned to the Target Phase II Acute lymphoblastic leukemia dataset, accessed via the cBioPortal interface 30,31 . We calculated correlation coefficients between USO1 and 24,278 genes detected by RNA-seq in 203 samples from 154 patients. Using a cut-off q value of 0.001, we then overlapped genes up-or down-regulated in USO1depleted with genes that had either a positive or negative correlation co-efficient. We found a highly significant overlap between genes that were positively correlated with USO1 across the B-ALL samples and downregulated in USO1-depleted cells (Hypergeometric test; p < 0.0001), but not in genes that were upregulated in USO1-depleted cells (non-significant p value) (

USO1 inhibits the MLL-AF4-driven leukemogenesis in primary murine bone marrow cells.
Having demonstrated that USO1 is required for the survival and growth of MLL-AF4 + cells in culture, we wanted to determine if USO1 was required in an experimentally induced primary cell model of MLL-AF4driven leukemia. Briefly, Lincells from the bone marrow of Cas9-EGFP mice 32 were transduced with MLL-Af4 retrovirus 33 (Fig. 5A) and selected with G418. We confirmed expression of the MLL-Af4 transgene, finding overexpression by RT-qPCR and western blot analysis (Fig. 5B). As expected, we observed rapid proliferation and expansion of the Lin -Cas9 MLL-Af4 cells. In order to deplete Uso1 from these cells, we designed and cloned three Uso1 murine specific sgRNAs (msg2 and msg3) into our internally designed MSCV.EFs.mCherry retroviral vector. To determine their effectiveness, we sorted GFP + mCherry + 70Z/3 cells transduced with the MSCV retroviruses, briefly expanded them in culture, and then queried USO1 expression. Western blot and RT-qPCR demonstrated that msg2 and msg3 both resulted in significant reduction of USO1 protein and mRNA expression (Fig. 5C, D). Uso1-depleted 7OZ/3 cells also had reduced cell growth by MTS assay which correlated with the different levels of USO1 depletion for msg2 and msg3 (Fig. 5E). From these data, the msg3 retrovirus was selected for transduction of Lin -Cas9 MLL-Af4 cells. After sorting GFP + mCherry + transduced Lin -Cas9 MLL-Af4 cells (Fig. 5F), USO1-depletion was confirmed by western blot, in which protein expression was reduced to 26% (Fig. 5G), and the cells were subsequently used in a colony formation assay. Functionally, USO1-depletion resulted in significantly fewer colonies in Lin -Cas9 MLL-Af4 cells compared to NT control cells at 12 days. This change was maintained at several different starting numbers of Lin -Cas9 MLL-Af4 cells (Fig. 5H). Hence, the function of USO1 is preserved in not only in human B-ALL cell lines but also in primary murine MLL-Af4 transformed cells.

Discussion
The molecular mechanism of MLL-AF4 driven leukemogenesis remains incompletely understood, and this subtype of B-ALL is highly aggressive 5,7 . Although there are several fusion partners for MLL in acute leukemia, downstream transcriptional dysregulation is a common feature 15,34,35 . In this study, we sought to understand whether overexpression of putative RBPs, which we identified previously, contributes to the pathogenesis of MLL-AF4 + B-ALL 15 . We performed a focused sub-genomic CRISPR/Cas9 dropout screen to specifically address whether these putative RBPs had a functional role in leukemia cell growth. Indeed, we identified three genes (EIF3E, EPRS, and USO1 ) that appeared to be required in the MLL-AF4 + cell line, SEM. These genes showed slightly higher rates of dropout in SEM cells than in NALM6 cells. Of these, USO1 expression showed a dependence on MLL-AF4, whereas EIF3E and EPRS did not show the same dependence and were previously reported to be "common essential genes", per the depmap portal 22 . Follow-up studies confirmed a role for USO1, both in cell lines and in a model of MLL-AF4 driven leukemia in primary murine bone marrow cells. RNA sequencing revealed that USO1 regulates numerous pathways, including mTOR, MYC targets, as well as elements of RNA homeostasis.
One of the challenges of genomic-scale CRISPR screens is that genes with a small average effect size on the phenotype of interest can be quite difficult to identify and can frequently be "drowned out" by genes with a larger effect size 36 . As we were particularly interested in the role that RBPs play in B-ALL, and their effect size in cell lines is unknown, we chose to perform a sub-genomic essentiality screen targeting a pre-defined set of RBPs known to be highly expressed in B-ALL. Therefore, under the hypothesis that a significant proportion of the  (H). Colony formation assay using Lin -Cas9 MLL-Af4 cells in methylcellulose assay as described in methods with titration of input cell number (t test, ***P < 0.001; ****P < 0.0001). www.nature.com/scientificreports/ sgRNAs would have a negative effect on cell viability, our design included a substantial number of both positive and negative control sgRNAs in order to properly model the null effect distribution and compare the effect size of functional RBPs to that of known essential genes. Secondly, the use of a non-MLL-AF4 control cell line allowed us to identify those proteins whose expression might be most important in MLL-AF4 leukemia. Supporting the idea that such specific effects may not be seen in genome scale studies, USO1 dropout was detected in one prior genome-scale analysis, whereas it was not observed in two other genome scale studies that queried vulnerabilities in MLL-translocated leukemia 17,37,38 . Thus, our approach to perform this type of CRISPR/Cas9 screen may help inform the design of future forward genetic screens. Our group is interested in understanding RNA binding proteins in B-ALL, and we recently described the functional role of IGF2BP3 in pathologic expansions of cells within the hematopoietic system and its requirement for survival and growth in B-ALL cell lines 15 . Here, we focused on USO1, based on its identification in this screen as well as a prior study in MV-4-11 AML cells that had identified USO1 as a factor required for AML survival 17 . A recent study reported a KMT2A(MLL)-USO1 fusion gene in a secondary AML, hinting at a further connection between MLL-driven leukemia and USO1 39 . More generally, USO1 has been shown to be of functional importance in cancer [23][24][25] . USO1 was recently reported to have RNA-binding function in studies utilizing high-throughput biochemical techniques 16,40,41 , despite a canonical role in the regulation of vesicular transport 42,43 . Adding to these prior descriptions of USO1 function, we validated USO1 as a MLL-AF4-induced gene, and found that it was functionally required in cell lines and in primary bone marrow for MLL-AF4 dependent phenotypes. This work firmly establishes the significance of this protein in acute leukemia, which was not previously appreciated.
Recent studies in multiple myeloma have shown that USO1-deficient cells showed have reduced cell proliferation and increased apoptosis via regulation of Erk pathway activity 24 . Transcriptome analysis of USO1-depleted SEM cells in our study demonstrated a mixed picture, with the downregulation of certain cancer and cell growthrelated pathways, including mTOR and ERB2, but concurrent downregulation of RNA metabolism and MYC targets. Curiously, there was also upregulation of the mTORC1 hallmark pathway in USO1-depleted cells (as opposed to downregulation of mTOR generally), perhaps indicating a specific effect on mTORC2. Nonetheless, we observed decreased phospho-mTOR, which is consistent with the effect seen on cell growth and cell cycle. Additionally, our data suggests that USO1 is associated with other molecular functions of gene regulation, such as RNA homeostasis. Interestingly, the pathways noted to be deregulated show similarity to those deregulated upon inhibition of the RNA demethylase FTO 44 . It is important to note, however, that we have not characterized its function as a RBP. It is tempting to speculate that USO1 is a bifunctional protein with roles in vesicular transport and RNA binding, perhaps in a manner similar to YBX1 45 . YBX1 appears to bind to and sort microRNAs, specifically, miR-223, into exosomes. By regulating this process, YBX1 can impact cellular homeostasis. Hence, further work to assess the molecular role of USO1 as a putative RBP in MLL-AF4 translocated leukemia is warranted.
Overall, our study successfully queried the functional relevance of a set of genes identified from primary patient samples using expression profiling. Here, we provide a rubric for how to functionally analyze a prioritized list of genes in leukemogenesis, or in other pathogenetic processes. In addition, we establish a role for the putative RBP, USO1, in leukemogenesis. Given the broader range of cancer types that show USO1 dysregulation, our work may have implications beyond those in B-ALL. Furthermore, understanding how non-canonical RBP may participate in leukemogenesis may open up new avenues in developing novel strategies for the diagnosis, prognosis, and treatment of B-ALL. Sub-genomic CRISPR screen. A sub-genomic CRISPR/Cas9 screen was designed to target 36 RBP genes and 12 "positive control genes", and included 28 negative control (or non-targeting) single guide RNAs (sgR-NAs). The positive control genes, representing known vulnerabilities in MLL-translocated acute leukemia, were selected from top 100 genes dysregulated in Genome wide CRISPR screen in MV-4-11 cells 17 . These genes are expected to "drop out" in a CRISPR screen of MLL-translocated leukemia, but it is not known whether they are specific for MLL-translocated leukemia. This design was adapted to provide enough non-targeting controls in the context of a sub-genomic screen, where a significant proportion of targeting sgRNAs may be expected to change. Five sgRNAs were designed for each RBP or positive control genes, using sgRNA design tools from Broad Institute 46 . pLKO5.sgRNA.EFS.tRFP is a lentiviral vector, which contains EF-1 alpha binding sequence (EFS) upstream of tRFP, was obtained from Addgene (#57,823) 47 . The 268 pooled sgRNA were then cloned into pLKO5.sgRNA.EFS.tRFP lentiviral vector using standard protocols 17 . Prior to CRISPR/Cas9 screening, B-ALL cell lines with MLL-AF4 translocation (SEM) 33 and without MLL-AF4 translocation (NALM6) 48 were stably transduced with pLentiCas9-GFP 49 lentivirus and sorted on GFP positivity, with subsequent confirmation of Cas9 expression (Supplementary Fig. 1A). pLKO5.sgRNA.EFS.tRFP lentiviral pool titers were calculated from SEM and NALM6 cell transduction. For experiments, bulk GFP + SEM and NALM6 cells were infected at < 0.3 MOI and 2 × 10 6 cells GFP + tRFP + were sorted by FACS after 48 h of infection ( Supplementary Fig. 1B). Genomic DNA (gDNA) isolated from 10 6 cells was used for construction of the Reference (REF)  www.nature.com/scientificreports/ the other 10 6 cells were cultured and expanded. Cells were split every five days and 10 6 cells were reseeded for culture 50 to maintain a sgRNA representation of 3700X. Following 28 days of culture, cells were harvested and gDNA was isolated for the Depletion (DEP) library sample preparation.

Methods
Library preparation, DNA sequencing, and analysis. Sequencing libraries were prepared from both the Reference (REF) and Depleted (DEP) genomic DNA (gDNA) samples obtained at days 0 and 28 of CRISPR screen experiment, respectively 17,50 . Libraries were prepared from 200 ng of input DNA, by using Q5 highfidelity DNA polymerase (#M0492S, NEB) and Illumina adapted primers to amplify the sgRNA target region from the gDNA, as previously described 17 (Supplementary Fig. 1C). The purified PCR product was quantified using Qubit and quality control was done using Bioanalyzer and sequenced on HiSeq 3000 at the Technology Center for Genomic and Bioinformatics at UCLA. Adapter sequences were removed using in-house scripts. Candidate reads (those containing a valid primer sequence and with a minimum length of 20 bp after trimming) were aligned to the sgRNA library using bowtie v0.12.8 51 with a maximum tolerance of one mismatch. Counts tables for both individual sgRNAs and gene-level summaries were compiled from non-ambiguous hits for both the Reference and Depletion libraries in each experiment and for each cell line. Count tables were processed with DESeq2 48 to obtain variance-stabilized normalized abundance and rank sgRNAs and genes based on differential abundance (moderated fold change and adjusted Wald test p value). Cell proliferation, cell cycle and apoptosis assays. Cell proliferation assay was performed using standard MTS assay protocol. 5000 cells were plated in 100 µL volume of media in a single well of 96-well tissue culture plates. Cells were harvested at different time points of day 0, day1, day 3 and day 5. MTS reagent mix was prepared by adding 100 µL of PMS solution (0.21 mg/mL) to 1 ml of MTS reagent (0.33 mg/mL) and 20 µL reagent mix was added to each well. The plate was incubated at 37 °C for 2 h and absorbance was taken at 490 nm in a microplate reader. Cell cycle analysis was performed using propidium iodide (PI). Cells were harvested and washed in PBS and fixed in 70% ethanol overnight at − 20 °C. Fixed cells were washed with PBS and centrifuged at 2500 rpm. PI solution (2 mg/10 mL) was diluted in PBS and added with 0.2 mg/mL of DNase free RNase A. Nearly, 300 µL of the PI solution was added to each tube and incubated at RT for 2 h. The stained samples were analyzed by flow cytometer.

Cell line treatment with transcription inhibitors.
Annexin V staining was performed to study the apoptosis in the cells using standard protocol. Briefly, cells treated with inhibitors/siRNAs were harvested and washed in PBS before resuspending in binding buffer (10 6 / mL). 100 µL of cell suspension was stained with 0.5 µL of Annexin V antibody conjugated to Pacific blue and incubated at RT for 30 min. After incubation, 300 µL of binding buffer was added to the sample and analyzed by flow cytometer. siRNA knockdown of cell lines. siRNA transfection was performed using standard Nucleofection program provided by the manufacturer. SEM cells were Nucleofected using the 4D Nucleofector System (Lonza, Cologne, Germany). Cells were washed with phosphate-buffered saline and then resuspended in nucleofection solution (SF Cell Line 4D-Nucleofector X Kit, Lonza, Cologne, Germany), at a final concentration of 2 × 10 6 cells/100 µL reaction. Cells were nucleofected with 30 pmol of control siRNA, USO1 siRNA1, or USO1 siRNA 4, in 100 µL cuvettes using program CV-104. Immediately after nucleofection, 500 µL of pre-warmed, antibiotic-free media was added to the cuvette and incubated for 10 min at RT. After incubation cells were transferred to a 12 well plate containing 1.5 mL of media. Nucleofected cells were maintained at 37 °C and 5% CO 2 prior to harvesting for analysis.
RT-qPCR assays. Previous protocols were adapted for RT-qPCR, based on our prior work 15 . A full list of RT-qPCR primers is presented in Table S2. For normalization, we utilized RT-qPCR primers for 18S (human) and L32 (mouse).

Immortalization of Linbone marrow cells. All mice used in this study were obtained from Jackson
Labs and were genotyped according to JAX protocols and maintained in the UCLA Division of Laboratory Animal Medicine. Bone marrow cells from C57BL6/J Cas9-EGFP mice 53 (Jackson Laboratories, # 026,179) were isolated by flushing the bones from the mice and creating a single cell suspension. Cells were incubated with a lineage antibody cocktail and depleted for Lineage + cells using MACS technology (Miltenyi Biotech). Lincells were spin-infected and transduced with MLL-Af4 retroviral preparation 33 . The MSCV-MLL-flag-Af4 plasmid was the kind gift via MTA by Dr. Michael Thirman (University of Chicago, Department of Medicine). After four rounds of transduction, cells were selected in 400 µg/mL G418 supplemented media for 7 days.
MSCV sgRNA vectors. We generated a novel MSCV vector that can overexpress an individual sgRNA in addition to an mCherry reporter. In brief, the MSCV.mU6.sgRNA-EFs.mCherry.v1 retroviral vector was constructed by replacing a 2.1 kb EGFP-PGK.Puro fragment from the pMGP vector with a 2.7 kb sequence containing mU6.BbsI-stuffer-BbsI-scaffold-spacer-EFs.mCherry via BglII/ClaI digest. The sgRNA scaffold and EF-1α short (EFs) promoter elements were derived from the pLentiCRISPRv2 vector. The mU6 promoter was designed from the GenBank sequence NC_000076.6 (nt 79,908,880-79,909,195). A silent mutation was incorporated into the mCherry reporter element to remove an internal BbsI restriction site. The 1.2 kb stuffer sequence was derived from portions of the 1.8 kb firefly luciferase gene. The sgRNA sequences targeting mouse were designed as above and directionally cloned between the mU6 promoter and sgRNA scaffold sequence via BbsI. Detailed methods and vector maps are available upon request.
Colony forming unit assay. A colony forming unit assay was performed to study the effect of USO1depletion on the colony forming potential of Lin -Cas9 MLL-Af4 cells. The assay was performed using the Methocult colony forming media (STEMCELL Technologies, #M3434) 54 . Briefly, approximately 5,000 Lin -Cas9 MLL-Af4 USO1 depleted cells were mixed in 3.2 ml of overnight thawed Methocult media and plated in two 35 mm dishes along with the NT controls and cultured for 12 days. After 12 days of culture, individual 35 mm dishes were counted for both total number and morphologic subtypes of colonies formed by USO1 depletion and NT control cells.  55 was used to align the reads to a genome index that included both the genome sequence (GRC38 human primary assembly) and the exon/intron structure of known human gene models (Gencode v29 genome annotation). Alignment files were used to generate strand-specific, gene-level count summaries with STAR's built-in gene counter. Independent filtering was applied as before 56,57 : genes with less than 6 total counts across all samples, count outliers, or low mappability (< 50 bp) were filtered out for downstream analyses. Expression estimates were computed in units of fragments per kilobase of mappable length and million counts (FPKMs). Differential expression analyses between USO1 depletion and non-targeted controls was performed with DESeq2 27 and genes were ranked based on moderated fold change and adjusted Wald test p value. Functional enrichment for selected genes was performed with Metascape 36 .