PRAME induces genomic instability in uveal melanoma

PRAME is a CUL2 ubiquitin ligase subunit that is normally expressed in the testis but becomes aberrantly overexpressed in many cancer types in association with aneuploidy and metastasis. Here, we show that PRAME is expressed predominantly in spermatogonia around the time of meiotic crossing-over in coordination with genes mediating DNA double strand break repair. Expression of PRAME in somatic cells upregulates pathways involved in meiosis, chromosome segregation and DNA repair, and it leads to increased DNA double strand breaks, telomere dysfunction and aneuploidy in neoplastic and non-neoplastic cells. This effect is mediated at least in part by ubiquitination of SMC1A and altered cohesin function. PRAME expression renders cells susceptible to inhibition of PARP1/2, suggesting increased dependence on alternative base excision repair pathways. These findings reveal a distinct oncogenic function of PRAME than can be targeted therapeutically in cancer.


Introduction
Preferentially Expressed Antigen in Melanoma (PRAME) was initially discovered as a melanoma antigen recognized by autologous T cells 1 . PRAME is normally expressed predominantly in the testis and becomes aberrantly expressed in a wide variety of cancer types, often associated with poor outcome 2, 3, 4, 5, 6, 7 . One such cancer is uveal melanoma (UM), the most common primary malignancy of the eye, which results in metastatic disease in up to one half of patients 8 . About a quarter of UM tumors express PRAME, which is strongly associated with aneuploidy, metastasis and poor patient outcome 2,9,10 .
Numerous functions have been attributed to PRAME, but it remains unclear how PRAME drives cancer progression. PRAME can repress retinoic acid receptor signaling and is induced by pathogen associated molecular particles (PAMPs) and gamma-interferon in leukemia cells 4,11,12 . PRAME functions as a substrate recognition subunit for a Cullin2-based E3 ubiquitin ligase (CUL2) complex 12,13 , and it co-localizes on chromatin with NFY and EKC/KEOPS transcription factor complexes 13,14 . The tumor suppressor p14/ARF is one of few proteins that has been implicated as a substrate of PRAME-CUL2 15 . Here, we elucidate the cellular and genomic consequences of PRAME expression in UM cells and non-transformed human uveal melanocytes (UMCs), revealing new oncogenic functions of PRAME and suggesting therapeutic strategies for targeting PRAME.

PRAME is normally expressed in spermatogonia leading up to meiotic recombination.
Since PRAME expression is largely confined to the testis 9, 16 , we analyzed single-cell RNA sequencing (sc-RNAseq) data from human testis and found that PRAME expression peaks early in spermatogenesis, predominantly in spermatogonial stem cells and spermatogonia, leading up to activation of meiotic recombination (Fig. 1a). The pattern of PRAME expression closely parallels that of essential mediators of meiotic recombination, including ATM, MRE11, RAD50, NBN, BRCA1 and BRCA2 17,18 (Fig. 1b).
The PRAME promoter region is hypermethylated and silenced in somatic cells and PRAME(-) cancer cells, whereas it is hypomethylated in testis and PRAME(+) cancer cells 9,19 (Fig. 1c). Interestingly, the upstream untranslated region of the PRAME locus contains an unusually long stretch of approximately 36 putative G-quadruplex (G4) forming sequences ( Fig.   1c), which have been associated with transcriptional regulation and cancer 20 . G4 structures are frequently formed in stem cells at promoter and enhancer regions of pluripotency genes and then lost during differentiation with silencing of these genes 20 . G4 structures regulate transcription through multiple epigenetic mechanisms; they often co-localize with the activating histone mark H3K4me3 and the histone acetyltransferase p300, and they promote DNA hypomethylation by sequestering DNMT1 21 . Indeed, we found that UM cells expressing PRAME exhibited marked accumulation of H3K4me3 and H3K27ac surrounding the G4-forming stretch, similar to human testis, whereas these marks were absent in UM cells lacking PRAME expression (Fig. 1c). The highly regulated transcriptional control of PRAME and its restricted expression to meiotic cells suggest that context-inappropriate expression of PRAME in somatic cells has deleterious consequences and is strongly suppressed. PRAME activates meiotic genes and promotes chromosomal instability. To explore the transcriptional impact of PRAME expression, we generated cell lines to allow for ectopic expression and knockout of PRAME (Supplementary Figure 1). We then performed RNA-seq in (1) PRAME-negative Mel290 UM cells before and after induction of PRAME expression, and (2) PRAME-positive MP41 UM cells before and after CRISPR-mediated knockout of PRAME.
The overlap of genes upregulated with PRAME expression and downregulated with PRAME knockout were enriched for pathways regulating meiosis, chromosome segregation, and DNA double strand break (DSB) repair (Fig. 2a,b). Within 4-7 days of enforced PRAME expression in the near-diploid Mel290 UM cell line, there was a significant increase in the number of micronuclei (Fig. 2c,d), which are associated with defects in mitotic chromosome segregation 22 . Within 4 weeks of PRAME expression, subclones appeared with new chromosome copy number aberrations (CNAs) demonstrated by flow cytometry and single-cell DNA sequencing (scDNA-seq) (Fig. 2e-i). In non-transformed diploid human UMCs, 7 months of enforced PRAME expression resulted in striking morphologic changes consistent with malignant transformation, including loss of contact inhibition, formation of multicellular spheroids, and aneuploidy ( Fig. 2j-l). 15% of analyzed metaphase spreads from PRAME expressing cells harbored an abnormal chromosome 6, monosomy 16, and a marker chromosome, whereas 85% showed these abnormalities plus additional copies of chromosome 6, 7, 12 and 22 Table 1). These findings demonstrate that PRAME expression in somatic cells can elicit aberrant activation of transcriptional programs resulting in genomic instability and aneuploidy. Consistent with these findings, PRAME expression is associated with increased aneuploidy in multiple TCGA cancer types (Supplementary Figure 2). PRAME alters cohesin complexes. To explore how PRAME induces aneuploidy, we immunoprecipitated ectopically expressed FLAG-tagged PRAME in Mel270 UM cells and searched for PRAME-interacting proteins using mass spectrometry ( Fig. 3a and Supplementary Table 2). Proteins pulled down with PRAME were enriched for components of the Cullin-RING-based E3 ubiquitin ligase (CRL) and COP9 signalosome complexes, consistent with findings in other cell types implicating PRAME as a CUL2 substrate recognition subunit 13 .
To identify potential substrates of the PRAME-CUL2 complex, we probed for changes in the ubiquitinated proteome following PRAME induction in Mel290 cells and PRAME knockout in MP41 cells (Fig. 3b,c and Supplementary Table 3). Differentially ubiquitinated proteins were enriched for pathways involved in nuclear envelope reformation, male gonad development, and cellular response to stress. We then expressed BioID2-tagged PRAME in Mel290 cells to perform proximity labeling followed by pull down and mass spectrometry to screen for additional protein-protein interactions (Supplementary Table 4). The intersection of proteins enriched in these three experiments yielded a list of 10 leading candidates for PRAME-CUL2 substrates, including SMC1A and other proteins involved in chromosome maintenance (e.g., HUWE1, NUP107 and SMHD1) (Figure 3d).
We examined how PRAME may specifically affect the function of SMC1A, a core subunit of the cohesin complex that is essential for proper chromosome segregation in dividing cells 23 . PRAME expression resulted in increased ubiquitination of SMC1A at K998, and its protein levels were not decreased (Supplementary Fig. 3), suggesting that PRAME does not promote proteasomal degradation of SMC1A. Instead, we hypothesized that PRAME may affect the function of SMC1A by altering its interaction with other cohesin components. To investigate this possibility, we immunoprecipitated SMC1A and performed mass spectrometry in Mel290 cells with or without induction of PRAME ( Fig. 3e and Supplementary Table 5). PRAME expression resulted in decreased interaction between SMC1A and several proteins, including STAG2, a cohesin subunit associated with centromere cohesion 24,25,26 . This finding was orthogonally validated in co-immunoprecipitation experiments, wherein PRAME significantly decreased the interaction between SMC1A and STAG2 (Figure 3f). PRAME induces genomic instability. Cohesin complexes play a key role in sister chromatid cohesion, telomere maintenance and DNA damage repair 23,27 , which are critical to maintaining genomic stability. Using fluorescence in situ hybridization (FISH) to probe telomere integrity, we found that induction of PRAME in both UMC and Mel290 cells caused an increase in telomere loss and doublet formation (Figure 4a,b), consistent with fragile telomeres and a defect in DNA repair activity 28,29 . Hence, we determined the activity of the two main DSB repair pathways, homologous recombination (HR) and non-homologous end-joining (NHEJ), by evaluating the accumulation of HR and NHEJ core factors BRCA1 and 53BP1, respectively in cells that were either unchallenged or treated with ionizing radiation (IR) to induce DSBs. As expected, 4 Gy of IR induced comparable levels of gH2AX staining, a marker of DSBs, in both UMC and Mel290 cells (Fig. 4c,d). Interestingly, PRAME expression increased the number of gH2AX foci in the absence of IR in Mel290 cells, indicating that PRAME alone can induce the accumulation of DSBs. Further, recruitment of BRCA1, but not 53BP1, was significantly increased by PRAME expression, with or without IR in both UMC and Mel290. Since BRCA1 recruitment is a hallmark of HR-mediated DNA damage repair, we further evaluated the effect of PRAME expression on HR in Mel290 cells using a direct repeat green fluorescent protein (DR-GFP) HR reporter assay 30 . PRAME expression resulted in a significant increase in HR activity in the presence or absence of SceI endonuclease activity (Figure 4e,f). Since PRAME expression leads to an increase in DSBs, we hypothesized that it may also result in susceptibility to inhibitors of poly(ADP-ribose) polymerase 1/2 (PARP1/2), which are central to DNA damage repair pathways 31 . Indeed, PRAME expression rendered UMCs and Mel290 cells susceptible to the PARP1/2 inhibitor talazoparib (Figure 4g,h). In an orthotopic mouse model of UM using MP41 cells, which express high endogenous levels of PRAME, treatment with talazoparib significantly reduced metastasis (Figure 4i). These findings indicate that PRAME dysregulation in somatic cells results in chromosomal instability that can be targeted pharmacologically.

Discussion
PRAME is a cancer-testis antigen that is normally expressed in the testis and is aberrantly overexpressed in many cancer types 9, 16, 32 . Here, we shed new light on the functions of PRAME and reveal how misexpression of PRAME promotes genomic instability and aneuploidy, which are hallmarks of cancer 33,34 . We found that PRAME is normally expressed preferentially in spermatogonial stem cells and spermatogonia at meiotic crossing-over, where homologous chromosomes undergo programmed DSBs, exchange of genetic material, and repair by HR 17 .
Genes coordinately expressed with PRAME in spermatogenesis include essential mediators of HR, including ATM, MRE11, RAD50, RPA1, NBN, BRCA1 and BRCA2 17,18 . PRAME expression in UM cells and non-transformed UMCs activated transcriptional programs involved in meiosis, DNA repair and chromosomal segregation. Further, PRAME expression led to increased DSBs, HR and telomere instability, as well as the formation of micronuclei, a hallmark of chromosomal instability associated with chromosome segregation errors and aneuploidy 22,35 .
Taken together, these findings suggest that PRAME plays a role in regulating normal meiotic recombination and causes genomic instability when mis-expressed in somatic cells, which could explain the association between PRAME expression and aneuploidy in UM and numerous other cancer types (Supplementary Fig. 1) 2,9 . PRAME misexpression appears to induce these genomic aberrations at least in part by altering the function of cohesin complexes, which align sister chromatids during HR, telomere maintenance, and chromosome segregation 23,27,36,37 . PRAME expression resulted in the ubiquitination of SMC1A, a core member of the cohesin complex 23 . This ubiquitination did not result in a reduction of SMC1A protein levels, but rather, in altering the interaction of SMC1A with STAG2, another component of the cohesin complex. STAG2 mutations are associated with DNA damage repair defects and aneuploidy in numerous cancer types 38,39,40 . Ubiquitination modulates cohesin function during meiosis and mitosis 41,42,43 , and further work is needed to elucidate the role of PRAME-mediated ubiquitination in tumorigenesis.
In recent years, PRAME has drawn increasing attention as an important cancer driver and potential target of immunotherapy 32,44,45 . Our findings suggest additional strategies for targeting PRAME through the DNA damage repair vulnerabilities it creates. PRAME expression results in an increase in DSBs despite increased HR, suggesting an increased dependency on alternative base excision repair pathways. Accordingly, we found that PRAME expression renders cells vulnerable to inhibitors of PARP1/2, which are enzymes critical to alternative DNA repair pathways 31 . The newly described functions of PRAME in this study open new avenues for investigating its role in cancer progression and for leveraging therapeutic vulnerabilities created by its expression.    Genes were hierarchical clustering based on their average RNA expression.
Chromatin immunoprecipitation and DNA sequencing. Chromatin immunoprecipitation (ChIP) followed by next-generation sequencing (ChIP-seq) was performed using 20 million cells per experiment. Cells were crosslinked for 7 minutes with 1% formaldehyde, then DNA was sonicated to an average fragment size of 200-500 base pairs using a Covaris M220 sonicator.

Plasmids and lentiviral expression vectors.
The pLV-TET-PRAME-V5 vector was created by PCR amplification of human full-length PRAME cDNA fragment (Horizon # MHS6278-202802292) and subsequent recombination into a pLV-TET-C-V5 plasmid encoding C-terminal V5. Plasmids containing FLAG-tagged PRAME and BioID2-tagged PRAME were created by substituting the C-terminal V5 tag with FLAG or FLAG-BioID2 fragments, respectively, using restriction digestion and T4 ligation. MP41s with knockout of PRAME were created using lentiviral particles encoding spCAS9 (Addgene plasmid no. 50661) and guide RNA-encoding plasmid (Addgene plasmid no. 64114). Guide RNA against PRAME 5'-GGGACAGGATACAGCACGT-3' and 5'-CCGGCAGTTAGTTATTGAG-3' directed the CRISPRmediated deletion of the first exon of the PRAME gene. The pLV-TRPM1-LUC vector expressing luciferase under TRPM1 promoter was synthesized by VectorBuilder. The lentiviral plasmids were packaged into lentiviral particles by transient co-transfection into HEK293T cells with pMD2G and psPAX2 packaging plasmids using JetPrime reagent (Polyplus). The lentivirus transduced cells were selected with puromycin for 7 days, clonally selected for optimal PRAME knockout or overexpression upon induction with 1 µg/ml doxycycline. Gene knockout and overexpression was verified with western blot (Supplementary Fig. 2). Biotinylation was visualized on an Amersham Imager 680 (GE Healthcare). Images were quantified using ImageJ software. Western blot was probed using antibody against STAG2 (Santa Cruz, sc-81852). The same membrane was stripped with Restore Plus Western blot stripping buffer (Thermo Scientific, 46430) and re-probed with SMC1A antibody (Bethyl, A300-055AA) to ensure equal protein loading. Anti-rabbit HRP-linked IgG (Cell Signaling, 7074P2) was used as a secondary antibody. Experiments were repeated at least three times. Homologous recombination assay. Mel290-EV and Mel290-PRAME cells (1x10 6 cells per sample) were transfected with pDR-GFP (Addgene plasmid # 26475) using jetPRIME DNA and/or siRNA transfection reagent (Polyplus) and selected for puromycin resistance (2 μg/ml).
Upon 60% confluence, the cells were transfected with a plasmid expressing the restriction enzyme I-SceI (pCBASce1) (Addgene plasmid # 26477) using jetPRIME DNA and/or siRNA Spectrum (PerkinElmer, US). Mice were then divided with equal numbers of males and females into vehicle-and talazoparib-treatment groups. Talazoparib was dissolved in 10% n,ndimethylacetamide (DMAc) (Sigma) in 6% solutol (ChemScene) in saline (Aspen) and administered intraperitoneally (IP) every other day at 0.7mg/kg for 60 days starting 15 days after suprachoroidal injections. Vehicle consisting of the same solution lacking talazoparib was given on the same schedule. Four weeks after suprachoroidal injections, mice were anesthetized with ketamine/xylazine, and the right eyes were enucleated to prevent local discomfort from tumor growth in the eye. A toe-pinch was performed to confirm appropriate level of anesthesia, after which the eye and lid was cleaned using aseptic techniques. Two drops of proparacaine hydrochloride ophthalmic solution 0.5% (Bausch Lomb) were instilled in the right eye before surgery as a local anesthetic, and the left eye was lubricated (Optixcare) to prevent ulceration.
Next, proptosis was gently induced in the right eye using an open forceps until the globe was anterior to the lids and the optic nerve was accessible. The optic nerve and surrounding blood vessels were clamped for 5 minutes with a hemostat for hemostasis, and the eye was detached using surgical scissors. Enucleated eyes were placed in 10% formalin. The right eyelids were sutured together, and erythromycin ophthalmic ointment (Bausch Lomb) was applied to prevent infection. Following the surgery, mice were given one subcutaneous injection of Meloxicam SR (2 mg/kg, ZooPharm) and did not require additional analgesia based on their behavior and activity level. All mice were monitored closely and recovered fully without complications. Five months after suprachoroidal injections, mice were euthanized and necropsies were performed.
Liver, lungs, kidneys, and spleen were collected, placed in 10% formalin, and sent for histopathological analysis. All experimental animal protocols used in this study were approved by the Institutional Animal Care and Use Committee at University of Miami (protocol #15-197).
All activities were performed in compliance with federal, state, and institutional regulations.
Micronuclei staining. Mel290 cells with inducible FLAG-tagged PRAME expression construct (TET-FLAG-PRAME) were grown on #1.5 coverslips, and half of replicates were treated with 1 µg/mL doxycycline for the indicated amounts of time. Cells were fixed in 4% paraformaldehyde in 1X PBS for 20 minutes and washed three times with 1X PBS (5 minutes/wash). Cells were permeabilized with 0.1% Triton X-100 in 1X PBS for 10 minutes and washed in 1X PBS.
Coverslips were mounted using ProLong Diamond Antifade Mountant with DAPI, and cells were imaged using a Zeiss LSM 700 confocal microscope. At least 200 cells were counted for each condition and the experiments were repeated in triplicate.
Calculation of fraction genome altered. Bulk RNA sequencing expression and fraction genome altered data from the TCGA cohort was collected for each patient across 32 cancers via the Cancer Genomics Data Server R package (cgdsr Version 1.3). For each cancer, we fitted a LOESS model of PRAME expression and identified the inflection point before expression became aberrantly high, as previously described 2,9 . This inflection point provided the PRAME expression threshold for "positive" PRAME "positive Once samples were labeled as PRAME "HIGH" or "LOW", we conducted a one-way Wilcox test to identify significant association (pvalue < 0.05) of PRAME "HIGH" samples with increased fraction genome altered (FGA).
Single-cell DNA sequencing. Single-cell DNA sequencing was performed using the Chromium platform (10X Genomics). Single-cell suspensions were counted using both the Cellometer K2 Fluorescent Viability Cell Counter and hemocytometer, and cell counts were adjusted to 1,000 cells/μl. Samples were run using the Chromium Single Cell DNA Library & Gel Bead Kit with a target capture of 500 cells. Samples were processed on Chromium Single Cell C and D Chips (10X Genomics) according to the manufacturer's protocol and subsequently run on a thermocycler. Single-cell genomic DNA libraries were sequenced on the NextSeq 500 sequencer using 300-cycle high-output flow cells.
Single-cell CNV analysis. Raw BCL files for the DNA sequencing data were processed using Cellranger DNA (version 1.0.0). The "mkfastq" command was used to generate FASTQ files and the "cnv" command was used to generate CNV data aligned to the 10X Genomics GRCh37