USP24 Is a Cancer-Associated Ubiquitin Hydrolase, Novel Tumor Suppressor, and Chromosome Instability Gene Deleted in Neuroblastoma

Deubiquitinating enzymes (DUB) remove ubiquitin (Ub) from targeted proteins and thereby provide a counterregulating force to the impact of ubiquitin modification on protein activity, localization, and degradation. Approximately 100 DUBs have been identified through functional and bioinformatic analyses though well-defined mechanistic details for the roles of most were unknown (1). We and others have identified DUBs with oncogenic and tumor-suppressive activity, although few have been validated with physiologic models. USP24 has been linked with both tumor-suppressive and tumor-promoting activities, with a number of potential mechanisms identified in the regulation of the cell cycle, apoptosis, and the accuracy of mitosis. USP24 was a candidate hit in a screen designed to identify novel regulators of the spindle assembly (mitotic) checkpoint that ensures the accurate timing of anaphase onset relative to the attachment of spindle microtubules to chromatid kinetochores (2). Disruption of the mitotic checkpoint drives chromosome segregation errors and promotes the development of aneuploidy, the most common chromosome alteration in human cancer (3). Although aneuploidy has been implicated as contributing to tumor development, chromosome losses and gains do not typically produce oncogenic properties such as proliferation, migration, invasion, and others. Paradoxically, while most tumor cells have chromosome imbalances, ongoing loss or gain of chromosomes in mitosis often lead to short-term reductions in cellular fitness (4). Whether USP24 plays a role in mitosis, however, was not evaluated further. The physiologic relevance of these identified mechanisms and whether they impact tumorigenesis in vivo have not been determined.

Here, we report the generation of mutant mice lacking catalytically active USP24. We find that homozygous null mice were per-natal lethal with a tumor-prone phenotype observed in hemizygous animals following carcinogen exposure or with age. We identify a novel mechanism by which USP24 stabilizes mitotic levels of the microtubule associated collapsin response mediator protein 2 (CRMP2), leading to defective mitotic spindles, anaphase chromosome lagging, and aneuploidy. Reduced expression of USP24 is highly associated with poor survival in neuroblastoma, a highly aggressive childhood cancer. These data lead us to conclude that USP24 is a novel chromosome instability (CIN) gene and a haploinsufficient tumor suppressor in mice and likely in human neuroblastoma.

All procedures and experiments involving animals were approved by the Institution Animal Use and Care Committee of the Mayo Clinic. Usp24-mutant mice were derived from the XB614 embryonic stem cell clone generated by Bay Genomics. Cells were injected into blastocysts and chimeric mice were bred to C57BL/6 mice. Mice were genotyped by PCR using standard procedures. More details may be found in the Supplementary Materials and Methods. DMBA tumor susceptibility studies were performed as in ref. 5.

Mouse embryonic fibroblasts (MEF) were generated in our laboratory from day 13.5 embryos using standard techniques. The genotype of MEFs was assessed by PCR and the identity of MEF lines was subsequently assessed by immunoblotting in each experiment. All MEFs were cultured as described previously (6). Mycoplasma testing was not routinely performed on MEF cultures. All experiments using primary MEFs were performed between passages 2 and 5. Immortalization of MEFs was performed by lentiviral transduction with the SV40 large T antigen as described previously (7). The passage number of immortalized MEFs is not monitored. Chromosome counts on MEFs were performed as described previously (8). MEF transductions were conducted as described previously (9). Cell synchronization was performed as in ref. 10. Proliferation was monitored for three independent MEF lines for each genotype/condition in triplicate using an IncuCyte (Essen Bioscience) by monitoring five regions of each well every 2 hours. The device was maintained in a water-jacketed incubator with 5% carbon dioxide. The immunoreagents used in this study can be found in the Supplementary Materials and Methods. USP24 activity was monitored using ubiquitin vinyl sulfone (UbVS), SUMO-VS, ISG15-VS, and NEDD-VS (all obtained from Boston Biochem), as described previously (11).

Live-cell imaging experiments were performed as described previously in detail (12). A lentiviral construct encoding YFP-tagged H2B (pTSIN-H2B-YFP) was used to visualize chromosomes by fluorescence microscopy (9, 13). Indirect immunofluorescence was performed as described with modifications (10). More detail is contained in the Supplementary Materials and Methods.

The full details for the proteomic steps can be found in the Supplementary Materials and Methods. Briefly, Usp24 wild-type and homozygous mutant MEFs (n = four independent lines each) were cultured in multiple 15-cm dishes and were incubated for 6 to 8 hours with nocodazole (100 ng/mL final) prior to isolation of mitotic cells by mitotic shake-off. Frozen cell pellets were solubilized in 1% SDS/20 mmol/L Tris pH 8, and equal amounts of protein were separated in the first dimension on 18 cm pH 3–10NL IPG strips. The focused strips were mounted on top of 8% to 14% Tris glycine HCl gels, where they were then separated in the second dimension. The resulting gels were fixed and stained with SyproRuby and high-resolution images were obtained using the Molecular Imager FX system (Bio-Rad). Images were analyzed using REDFIN software (Ludesi) and spots with a significant and 2-fold change in volume (P < 0.025, ANOVA) were excised from silver-stained gels followed by LC/MS-MS identification. Gene ontology analysis of the retrieved proteins was performed using the functionality of the STRING web tool (https://string-db.org/).

The two isoforms of CRMP2 (CRMP2A-FLAG and CRMP2B-HA; ref. 14) were cloned from mouse brain cDNA. Both isoforms were PCR amplified and inserted into a modified lentiviral TSIN vector at the AscI-XhoI site (9, 13). Final products were confirmed by Sanger sequencing. For gene expression analysis, total RNA was isolated using RNeasy Kits (Qiagen), then converted into cDNA using SuperScript III First-Strand Synthesis System (Invitrogen). Equal amount of cDNA was used in quantitative real-time PCR reactions, using TaqMan probes (Applied Biosystems) according to the manufacturer’s instructions.

The analysis of the impact of deubiquitinating enzymes on cancer outcomes was performed using the PREdiction of Clinical Outcomes from Genomic Profiles web tool (PRECOG; https://precog.stanford.edu/). All survival analyses were initially performed using the R2: Genomics Analysis and Visualization platform (http://r2.amc.nl). Where indicated, gene expression was stratified by percentiles or using the median expression. The CIN25 signature score was generated by determining the percentile rank expression for each the 25 genes within the signature score using the “sample Ranked GeneSetScores” tool within R2. A single score was generated for each tumor by averaging the percentile ranks for all genes in the signature. A “high” CIN25 score was defined as a greater than median CIN25 score. Copy number analyses of neuroblastoma were performed using the web tool (https://padpuydt.shinyapps.io/check_cn_in_hr_nb/) accompanying the published works (15, 16). Graphs were reproduced using GraphPad Prism by exporting data from R2.

Quantitative in vitro experiments involving MEFs were performed with a minimum of three independently derived MEF lines that were nonimmortalized or immortalized as indicated. Graphs of these data represent the mean ± SEM for the results obtained for each line, not the individual values. The test used is listed in each respective figure legend. Unless otherwise noted in each figure, a P value of less than 0.05 is used to denote significance. Unless otherwise noted in the figure legend, statistics were performed using the functions embedded in the software GraphPad Prism. Where indicated, calculations were performed using the statistical functions embedded in web tools including the Z-scores in the PRECOG tool (https://precog.stanford.edu/) and the genomic correlations on R2 website (http://r2.amc.nl), which were corrected for multiple comparisons using the FDR. The scan function to find the ideal cut-off values for survival based on gene expression use P values corrected by the Bonferroni method.

The availability of clinically annotated public gene expression datasets allowed for a virtual screen of DUBs as potential oncogenes and tumor suppressors. To understand the impact of DUBs in cancer, we used the Prediction of Clinical Outcomes from Genomic Profiles database (PRECOG; precog.stanford.edu; ref. 17). In this dataset, the impact of gene expression on patient survival (greater or less than the median) is calculated across 39 classes of cancer and expressed as a Z-score. Positive Z-scores indicate that high expression is associated with poor outcomes, whereas negative Z-scores indicate that low expression correlates with poor outcomes. We profiled the meta Z-score for 95 known or predicted DUBs. Using a cutoff of P < 0.05 (Z-score less than –1.96 or greater than +1.96) higher expression of 16 DUBs and lower expression of 20 was associated with poor outcomes (Supplementary Table S1). This list includes some with established involvement in cancer including UCHL1, TNFAIP3 (A20), and USP1, as well as others with literature supporting a role in malignancy (18–20).

Because of its impact on children, we focused on the aggressive cancer neuroblastoma. Performing a similar analysis, the higher expression of 31 DUBs and the lower expression of 15 was associated with poor outcomes (Supplementary Table S2). Notably, the genes with the highest and lowest Z-scores (USP44 and USP24, respectively; Fig. 1A) were the top two hits in a genome-wide shRNA screen aimed at discovering new regulators of the mitotic (spindle assembly) checkpoint (2). We previously showed that disruption of USP44 leads to whole chromosome aneuploidy in humans and mice (10). We therefore focused on USP24. With respect to all annotated genes, USP24 is in the lowest 5% with a meta-Z-score of −3.54, indicating that reduced expression is on average associated with poor outcomes (P < 0.001). Examining the Z-scores for USP24 across individual cancers, there were significantly positive Z-scores in germ cell tumors and Ewing sarcoma (Supplementary Table S3). There is also significantly negative Z-scores in follicular lymphoma, multiple myeloma, brain astrocytoma, and most prominently in neuroblastoma (Fig. 1B; Supplementary Table S3).

To determine the impact of USP24 expression on clinical outcomes in neuroblastoma, we used to the R2: Genomics Analysis and Visualization Platform (http://r2.amc.nl). We examined event-free (EFS) and overall survival (OS) based on USP24 expression in the SEQC neuroblastoma dataset (n = 498; see Supplementary Table S4 for details on all datasets used) dividing the groups by the median USP24 expression. As indicated by the PRECOG data, there was a significant difference with lower USP24 expression associated with poor EFS and OS (Supplementary Fig. S1A and S1B). A similar result was obtained in two distinct datasets (Primary NRC n = 283, TARGET n = 249; Supplementary Table S1; Supplementary Fig. S1C–S1F) despite differences in rates of MYCN amplification and metastatic disease. Using the scan function of R2 to find the cutoff with the greatest survival separation (smallest P value) in SEQC, there was a progressive dose-dependent effect with decreasing USP24 expression associated with maximum effects (EFS) at the 19th percentile (Fig. 1C and D). There was also a strong correlation of reduced USP24 (< 20 percentile) with metastatic disease, and with MYCN amplification (Fig. 1E and F). While MYCN amplification itself is a potent predictor of poor outcomes, the impact of USP24 reduction on survival persists in patients with MYCN nonamplified tumors (Supplementary Fig. S2). These data indicate that reduced expression of USP24 is strongly associated with poor survival in neuroblastoma.

Chromosome alterations are common in neuroblastoma. Whole chromosome aneuploidy is generally regarded as a positive prognostic feature, whereas segmental chromosome gains and losses are an aggressive feature of high-risk disease (21, 22). At the cellular level, aneuploidy has a negative impact on the proliferation of cells (4). Its role therefore is largely thought to be in the initial formation of tumors and in promoting clonal evolution rather than having an impact on short-term measures of malignancy. Perhaps consistent with this notion, we found no difference in proliferation, migration, or invasion in two neuroblastoma cell lines (SH-SY5Y and IMR32) following USP24 depletion. As these are negative results, however, we sought more evidence linking USP24, aneuploidy, and neuroblastoma biology.

To dig deeper, we utilized a CIN gene signature that predicts poor outcomes across a number of human cancers (23). Consistent with those findings, the CIN signature (CIN25) was associated with a poor survival in neuroblastoma (Fig. 2A). Reduced USP24 was also significantly correlated with the CIN25 gene signature in neuroblastoma (Fig. 2B). To better understand the relationship between USP24 and segmental chromosome aberrations, we performed a genome-wide analysis to identify genes that correlate with USP24 expression (SEQC n = 498, Supplementary Table S4, FDR P < e−30). Mapping the resulting list to chromosome locations revealed a highly significant correlation between USP24 and genes located on chromosome 1p (chr. 1p; Supplementary Fig. S3A). The USP24 locus itself is located at chr. 1p32.3, and as expected, the highest correlation occurred at the USP24 locus (Supplementary Fig. S3B). Chr 1p is deleted in approximately one third of patients with aggressive high-risk neuroblastoma with a critical deleted region at 1p36 (16, 22). In a copy number analysis of 556 high-risk tumors (de Preter, Supplementary Table S4), USP24 was deleted in 129 (23.4%) of samples, whereas the overall incidence of chr. 1p deletion was 44.6% (Supplementary Fig. S4; ref. 16). We analyzed the impact of chr. 1p loss on USP24 levels using a dataset with paired gene expression and copy number analysis performed by array comparative genomic hybridization (Westermann, n = 105, Supplementary Table S4; ref. 24). In these data, a substantial proportion of tumors with chr. 1p loss also had a reduced copy number of the USP24 locus (Fig. 2C). Accordingly, there was a significant reduction of USP24 expression in tumors with chr. 1p loss, as well as a correlation between copy number loss at the USP24 locus and its transcript levels (Fig. 2D). To determine whether USP24 reductions have an impact on survival in neuroblastoma independently of chr. 1p loss, we used five surrogate genes located on chr. 1p that highly correlate with chr. 1p loss (ATPIF1, MEAF6, PRDM2, SRRM1, ZCCHC17; SEQC n = 498, Supplementary Table S4). Low levels (<25th percentile) of each had a significant correlation with poor survival as would be expected with the loss of chr. 1p. When only tumors with higher expression of these genes (>25th percentile, associated with favorable survival) were included as a surrogate for normal chr. 1p, USP24 reductions continued to have a significant impact on survival (Supplementary Fig. S5A–S5F). These data indicate that USP24 reductions occur commonly in neuroblastoma, often in association with loss of chr. 1p, and also indicate that USP24 reduction itself is a negative prognostic factor in neuroblastoma.

As aneuploidy is thought to play a role in the initial development of tumors, we generated mice with a genetrap predicted to interrupt the Usp24 gene to examine its physiologic role in tumorigenesis. The genetrap insertion interrupts a predicted armadillo fold and leads to a truncated protein, preserving 1,447 amino acids (160 kDa) upstream of the insertion (Fig. 3A and B). Importantly, the truncation would result in the loss of the USP_3 domain that is required for catalytic activity. After confirming germline transmission, we intercrossed Usp24^+/− mice but did not identify any Usp24^−/− animals at day 7 to 10 (Supplementary Table S5). We did observe, however, that each litter had dead pups in the cages shortly after birth. When observing timed pregnancies, we found that all pups were born alive but that a subset would die within a few hours of birth, all genotyped as Usp24^−/−. There were no overt physical malformations noted in the Usp24^−/− pups. Whole-mount histologic analysis revealed no obvious changes in major organs. We backcrossed the Usp24^+/− mice more than 10 generations onto C57BL/6 and could again identify no living homozygous mutant pups. We conclude that loss of the Usp24 gene results in perinatal lethality, although the cause is unknown.

We isolated MEFs from day 13.5 embryos and found mutant alleles at the expected frequencies (Supplementary Table S5). Using an antibody targeting an epitope at the C-terminus of USP24, we observed that there was no detectable full-length protein in Usp24^−/− MEFs (Fig. 3C). An antibody raised against an N-terminal epitope, however, detected a protein in all genotypes near the expected mass for full-length USP24 (Fig. 3C). The genetrap vector encodes a β-geo fusion protein that is predicted to be spliced in-frame after exon 37. The predicted fusion has nearly the same mass of USP24 (Fig. 3B). Confirming this, we were able to detect the USP24-β-geo fusion using an antibody recognizing β-galactosidase in Usp24^+/− and Usp24^−/− cells (Fig. 3C). We next examined the catalytic activity of the wild-type and mutant USP24 using the activity-based probe UbVS (11). While wild-type USP24 reacted with UbVS, the USP24-β-geo fusion did not (Fig. 3D). We conclude that the Usp24⁻ (null) allele encodes a USP24-β-geo fusion protein and that homozygous null cells have no catalytically active USP24.

A shRNA screen identified USP24 as potentially required for efficient signaling in the spindle assembly (mitotic) checkpoint (2). As defects in this pathway were associated with premature mitotic exit, chromosome missegregation, and aneuploidy, we monitored MEFs by live-cell video microscopy to examine mitotic accuracy. Cells were transduced with histone H2B-YFP to facilitate the visualization of chromosomes. We observed a significant increase in the rate of chromosome missegregation in Usp24^−/− MEFs compared with wild-type (Fig. 4A). Among the defects, there was a significant increase in chromosome lagging (Fig. 4A and B). We performed metaphase chromosome counts on MEFs and found a significant increase in whole chromosome aneuploidy in Usp24^−/− MEFs, with most errors involving the gain or loss of 1–3 chromosomes (Fig. 4C and D). This led us to consider whether massive aneuploidy may contribute to death of Usp24^−/− pups. We performed metaphase chromosome counts on liver cells from day 19.5 embryos and we found a significant and similar increase in aneuploidy in both Usp24^+/− and Usp24^−/− (Fig. 4E). A similar rate of aneuploidy was also observed in 5-month-old Usp24^+/− splenocytes (Fig. 4F). These data indicate that USP24 is involved in maintaining mitotic accuracy and that complete or partial reduction in its levels lead to aneuploidy in vivo. We conclude, however, that catastrophic aneuploidy is unlikely to cause the neonatal lethality of Usp24^−/− mice.

Erroneous attachments where one kinetochore is connected to microtubules from more than one centrosome, known as a merotelic attachment, is an important cause of chromosome lagging (25). We found no evidence of centrosome amplification or separation defects in Usp24^−/− cells, two common causes of merotelic attachments. To identify potential mechanisms for the chromosome lagging, we isolated cells arrested in mitosis using nocodazole and mechanical dissociation (shake-off; n = 4 independent immortalized MEF lines each). Two-dimensional SDS-PAGE was performed and the stained gels were analyzed digitally. There were 24 spots with reduced volume in Usp24^−/− MEFs compared with Usp24^+/+ and 23 reduced in Usp24^+/+ compared with Usp24^−/− MEFs. The spots were excised and proteins were identified by mass spectrometry (Supplementary Table S6). Gene ontogeny (GO) analysis revealed significant enrichment for several biological processes, molecular functions, and cellular components (Supplementary Tables S7–S9). Focusing on proteins with a potential to influence the mitotic spindle, we identified reductions in the level of the tubulin binding protein collapsin response mediator protein 2 (CRMP2; protein symbol Dpysl2) in Usp24^−/− MEFs (Fig. 5A). This was confirmed in subsequent independent immunoblots on extracts of synchronized MEFs (Fig. 5B). We confirmed that the transcript for CRMP2 (encoded by Dpysl2) is not reduced in Usp24^−/− MEFs (Supplementary Fig. S6). CRMP2 exists in two isoforms that differ by the presence of a 114 residue N-terminal extension CRMP2A (73 kDa) compared with CRMP2B (62 kDa; ref. 26). Although there were variable decreases observed in both isoforms, the most reproducible changes were seen in the smaller CRMP2B isoform.

To confirm the role of CRMP2 loss in the mitotic phenotypes associated with USP24 loss, we cloned the two isoforms of CRMP2 and reintroduced one or the other isoform into Usp24^−/− MEFs and again performed chromosome counts. While introducing either isoform into Usp24^+/+, MEFs did not change the rate of aneuploidy, expression of CRMP2B, but not CRMP2A reduced the aneuploidy in Usp24^−/− MEFs to near wild-type levels (Fig. 5C). Interestingly, in multiple independent experiments, we observed that the level of CRMP2 expressed did not exceed the level seen in wild-type. This suggests that there may be tight regulation of CRMP2 levels. We next examined the localization of CRMP2 on the spindle using immunofluorescence microscopy. Consistent with it having a role at mitotic microtubules, CRMP2 exhibited a strong colocalization on the spindle as visualized by α-tubulin. Notably, there was a significant reduction in CRMP2 spindle localization in Usp24^−/− MEFs compared with Usp24^+/+ that was rescued by reintroducing CRMP2B, or by incubating cells with the proteasome inhibitor MG132 (Fig. 6A and B). CRMP2 binds to tubulin heterodimers and promotes plus-end microtubule polymerization (27). We therefore examined whether CRMP2 loss affected the integrity of the mitotic spindle. We found a significant reduction in the immunodetection of α-tubulin in mitotic spindles and found significant increase in spindle length (interpolar distance) in Usp24^−/− MEFs (Fig. 6C–E). Both findings were restored with the reintroduction of CRMP2B. Similarly, depletion of CRMP2 (Dpysl2) in wild-type MEFs was itself able to reproduce the spindle changes and aneuploidy observed in Usp24-null cells (Supplementary Fig. 7SA–S7D). We conclude that USP24 prevents the degradation of CRMP2 and that CRMP2B is required for optimal assembly of the mitotic spindle. Repeated attempts to isolate ubiquitinated CRMP2 were unsuccessful, so we were unable to directly test whether USP24 deubiquitinates CRMP2. Because CRMP2 is known to be modified by SUMO (28), we assayed the specificity of USP24 toward ubiquitin and ubiquitin-like substrates using activity-based probes: UbVS, SUMO-VS, ISG15-VS, and NEDD-VS. We observed selectivity of USP24 toward UbVS, but not with other VS-coupled substrates (Supplementary Fig. S8).

Given that we found that USP24 is required to maintain levels of CRMP2 on mitotic spindles, we wondered whether reduced CRMP2 (DPYSL2) also correlates with poor outcomes in neuroblastoma. Comparing survival based on the median expression of DPYSL2, there was a significant effect with reduced expression predicting poor survival (Supplementary Fig. S9). Importantly, either when combining the impact of DPYLS2 and USP24, those with low levels of both transcripts had the worst outcomes. Importantly, higher DPYSL2 or USP24 had an intermediate survival compared with those with low, or normal, levels of both transcripts (Supplementary Fig. S9). This suggests that low CRMP2, either secondary to reduced USP24 protein or DPYSL2 transcript levels, predicts poor outcomes in this disease.

To examine the impact of USP24 loss on tumorigenesis, we generated cohorts of Usp24^+/− and Usp24^+/+ mice to examine carcinogen-induced and spontaneous tumorigenesis (Fig. 7A). The carcinogen dimethylbenz[a]anthracene (DMBA) is commonly used a general indicator of tumor susceptibility in mouse models (5, 12). To determine the impact of carcinogens on Usp24^+/− and Usp24^+/+ mice, we exposed newborn pups to a single dose of DMBA and scored for the development of skin tumors at 5 months of age. There was a significant increase in the incidence of tumors in Usp24^+/− mice (Fig. 7B). We then generated cohorts of mice and allowed spontaneous aging. At 18 months, there was again a significant increase in tumors observed in Usp24^+/− mice compared with Usp24^+/+ mice with lung tumors, small round blue cell tumors, lipomas, and liver tumors observed (Fig. 7C and D). We next examined the proliferation of Usp24^+/+ and Usp24^−/− MEFs using automated live-cell photography. There was a significant slowing of proliferation in the Usp24-null MEFs that was rescued with reintroduction of CRMP2B (Supplementary Fig. S10A). Introducing CRMP2B into wild-type MEFs had no effect on their proliferation. Strikingly, when proliferation was monitored in low-passage nonimmortalized MEF lines, we observed no significant difference in proliferation (n = 6 each; Supplementary Fig. S10B).

Deubiquitinating enzymes selectively regulate the ubiquitination status of their substrates, and therefore reverse ubiquitin-induced impacts on those proteins. While much progress has been made in understanding the biochemical roles of DUBs, many enzymes remain poorly understood. The large DUB USP24 was first identified as a potential susceptibility gene in Parkinson disease and has subsequently been implicated in cancer. Here, we provide the first evidence that USP24 plays a physiologic role in cancer and provide novel mechanistic insight into its tumor-suppressive functions.

In a bioinformatic analysis of human cancers, we find that USP24 reductions were common and strongly predictive of poor outcomes in four cancers, most notably neuroblastoma. This is a highly aggressive embryonal malignancy that mostly affects children (22). Aneuploidy has an important impact on survival in this disease, with differing effects depending on whether the aneuploidy involves whole chromosome gains or losses, segmental chromosome aberrations, or both. Unlike in most cancers, isolated whole chromosome aneuploidy (usually hyperdiploidy) is associated with better outcomes, whereas segmental alterations such as loss of chr. 1p, 11q, or gains at 17q were associated with poor survival (15, 16, 22, 29). The USP24 gene is located on human chr. 1p32 spanning over 149 kb, with 68 exons. This is proximal to the minimum deleted region seen in neuroblastoma, which is at chr. 1p36 (30). Copy number analysis, however, indicates that the USP24 locus is also deleted in a substantial fraction of tumors carrying deletions on chr. 1p. Our data also suggest that USP24 deletion has an independent impact on neuroblastoma outcomes and therefore may also contribute to the poor outcomes seen in association with chr. 1p deletions. Genomic evidence suggests that those patients whose tumors have both segmental and whole chromosome alterations have particularly poor outcomes, particularly those with coexisting MYCN amplification (29). The association of USP24 reductions in tumors with chr. 1p deletion, MYCN amplification, and above median CIN25 gene signature scores, may in part explain the poor outcomes seen in these patients. It is noteworthy that there has been increasing interest in the role of CIN in the pathogenesis of neuroblastoma (31). Our work underscores the need for a more systematic study of how CIN may influence the development and clinical behavior of neuroblastoma.

As USP24 was identified in a screen for ubiquitin pathway regulators of the mitotic (spindle assembly) checkpoint, we sought to understand the impact of USP24 loss on mitosis. Consistent with these data, we find that USP24 is required for accurate mitosis, with its loss associated with anaphase chromosome lagging and whole chromosome aneuploidy in MEFs and primary splenocytes. Anaphase chromosome lagging is a mitotic defect that results from improper attachment of spindle microtubules to the kinetochores of replicated sister chromatids. In particular, attachment of microtubules emanating from more than one spindle pole to one kinetochore, a defect known as a merotelic attachment, leads the affected chromatid to remain in the midzone between the spindle poles due to opposing forces exerted by each centrosome (25). Defects in the mitotic spindle, usually resulting from centrosome separation defects or centrosome amplification, were primary causes of merotelic attachments (32–34).

Our data indicate a previously unrecognized role for the tubulin assembly protein CRMP2 in the assembly of spindle microtubules. This protein, encoded by the DPYSL2 gene, has been implicated in a number of neurologic conditions including Alzheimer disease, amyotrophic lateral sclerosis, multiple sclerosis, and in bipolar depression (35). The molecular functions of CRMP2 in these conditions includes its well described role in tubulin polymerization, but also a distinct function regulating calcium and sodium channel activity (36). CRMP2 is SUMOylated at lysine 374, which is important in its roles related to ion channel regulation, but is thought to not influence tubulin polymerization activity (35). However, our finding that USP24 does not react with a SUMO activity based probe makes it unlikely that this modification plays a role in the phenotypes we observe. A related protein CRMP4 (encoded by DPYSL3) has previously been implicated in mitosis (37, 38). This protein has been also observed to localize to the mitotic spindle, and when depleted, leads to the development of polyploid cells and chromosome misalignment, a state in which chromosomes do not align at the metaphase plate (37, 38). This defect, which we did not observe in Usp24^−/−-mutant MEFs, is more characteristic of cells lacking aurora B activity. Furthermore, a recent report finds that reduced DPYSL3 leads to increased migration of claudin-low (but not claudin-normal) breast cancer cell lines (38). We found no effect of reduced USP24 on migration or invasion of neuroblastoma cell lines. Given that reductions in either of these related proteins leads to differing phenotypes, we conclude that they perform mechanistically distinct functions, although tissue or genetic context-specific dependencies cannot be excluded.

USP24 has been found to have other tumor-suppressive activities. A recent publication found that increased EGF or KRAS signaling suppresses USP24, thereby enhancing tumorigenesis (39). The tumor-suppressive mechanisms proposed for USP24 include stabilization of E2F4, BAX, and securin to restrain cell-cycle progression and promote apoptosis. Most of those experiments were performed in human lung cancer cells, and it is unclear whether the behavior may different in different models. We do not observe a proliferative advantage in nonimmortalized Usp24^−/− MEFs or in neuroblastoma cells, where we deplete USP24. We did however observe a significant increase in proliferation in MEFs immortalized by transduction with the SV40 large T antigen, which was rescued by reintroducing CRMP2B. If the primary mechanism of USP24 loss in tumorigenesis is related to aneuploidy, it is unlikely that USP24-deficient cells would change their proliferation, migration, or invasion in the short time span of our in vitro assays. If any, there may actually be an impairment in proliferation due to the proteotoxic stress imposed by unbalanced protein synthesis that results from large chromosome gains or losses (40–42). However, it is possible that when USP24 is depleted in cells where the p53 and RB pathways were dampened (e.g., by the large T antigen) that additional mechanisms such as those mentioned above come into play.

P.J. Galardy reports grants from Howard Hughes Medical Institute and Fraternal Order of Eagles during the conduct of the study, as well as other funding from Abbvie and Abbott labs outside the submitted work. No disclosures were reported by the other authors.

T. Bedekovics: Data curation, formal analysis, investigation, methodology, writing-review and editing. S. Hussain: Data curation, investigation, writing-review and editing. Y. Zhang: Investigation, writing-review and editing. A. Ali: Investigation, writing-review and editing. Y.J. Jeon: Investigation. P.J. Galardy: Conceptualization, resources, formal analysis, supervision, funding acquisition, writing-original draft, writing-review and editing.

We acknowledge the laboratory of Dr. Jan van Deursen who helped with the generation of Usp24-mutant mice and provided critical feedback of our experiments. The work was supported by grants to P.J. Galardy including from the Howard Hughes Medical Institute (HHMI), Fraternal Order of Eagles, and a Mayo Clinic Career Development award.

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.