Nudt21 regulates the alternative polyadenylation of Pak1 and is predictive in the prognosis of glioblastoma patients

Abstract

Alternative polyadenylation (APA) has emerged as a prevalent feature associated with cancer development and progression. The advantage of APA to tumor progression is to induce oncogenes through 3'-UTR shortening, and to inactivate tumor suppressor genes via the re-routing of microRNA competition. We previously identified the Mammalian Cleavage Factor I-25 (CFIm25) (encoded by Nudt21 gene) as a master APA regulator whose expression levels directly impact the tumorigenicity of glioblastoma (GBM) in vitro and in vivo. Despite its importance, the role of Nudt21 in GBM development is not known and the genes subject to Nudt21 APA regulation that contribute to GBM progression have not been identified. Here, we find that Nudt21 is reduced in low grade glioma (LGG) and all four subtypes of high grade glioma (GBM). Reduced expression of Nudt21 associates with worse survival in TCGA LGG cohorts and two TCGA GBM cohorts. Moreover, although CFIm25 was initially identified as biochemically associated with both CFIm59 and CFIm68, we observed three CFIm distinct subcomplexes exist and CFIm59 protein level is dependent on Nudt21 expression in GBM cells, but CFIm68 is not, and that only CFIm59 predicts prognosis of GBM patients similar to Nudt21. Through the use of Poly(A)-Click-Seq to characterize APA, we define the mRNAs subject to 3'-UTR shortening upon Nudt21 depletion in GBM cells and observed enrichment in genes important in the Ras signaling pathway, including Pak1. Remarkably, we find that Pak1 expression is regulated by Nudt21 through its 3'-UTR APA, and the combination of Pak1 and Nudt21 expression generates an even stronger prognostic indicator of GBM survival versus either value used alone. Collectively, our data uncover Nudt21 and its downstream target Pak1 as a potential "combination biomarker" for predicting prognosis of GBM patients.

Fig. 1

Reduced expression of Nudt21 in TCGA LGG and GBM cohorts and Nudt21 correlation with reduced survival in TCGA LGG cohort and two TCGA GBM patient cohorts. a Reduced expression of Nudt21 in TCGA low grade glioma patients and TCGA GBM patients based on RNAseq data. P = 0.08 grade II (n = 259) versus grade III (n = 266); p < 2.2e−16 grade II versus grade IV (n = 173) and p = 1.3e−10 grade III versus grade IV. b Kaplan–meier plot analysis showing that reduced expression of Nudt21 is associated with worse survival in the TCGA LGG grade II patients (n = 259, p = 0.012). c Downregulation of Nudt21 in four subtypes of GBM compared to normal brain samples, based on exon array data with Affymetrix human exon 1.0 ST platform. (Nor: Normal control (n = 11); Cla: Classical (n = 52); Mes: Mesenchymal (n = 56); Neu: Neural (n = 31); Pro: Proneural (n = 54), *p < 0.05, **p < 0.01). d, e Kaplan–meier plot analysis showing that reduced expression of Nudt21 is associated with worse survival in the TCGA GBM exon array cohort (p = 0.015, n = 348), and GDC TCGA GBM RNAseq cohort (p = 0.0073, n = 173)

Fig. 2

Three CFIm subunits form three distinct subcomplexes and CFIm25 stabilizes CFIm59 but not CFIm68 in GBM cell lines. a Coimmunoprecipitation (Co-IP) of CFIm complex members with an individual anti-CFIm subunit antibody. Lysates from LN229 and U251 cell lines were incubated with protein A/G agarose beads conjugated with anti-CFIm25, or anti-CFIm59, or anti-CFIm68 antibody. Anti-IgG serves as a control. Co-IP pull-down complex and the depletion supernatant were subjected to analysis using a 10% SDS-PAGE and the composition of CFIm subunits were detected by western blotting with indicated antibodies. A representative image is shown from three-independent experiments, CFIm subunits in Co-IP complex (upper panel) and in depletion supernatant (lower panel). b Interaction of CFIm subunits was determined by two yeast hybrid screen. Data show CFim25 interacts with itself and the two other individual CFIm sub-units and empty vector serves as a negative control. c Nudt21 knockdown decreases CFIm59 expression, but has no effect on CFIm68 expression in GBM cell lines. A representative data are shown from three-independent experiments

Fig. 3

Identification of key signaling pathways whose APA is regulated by Nudt21 in GBM cells. a Knockdown of Nudt21 in LN229 was validated by western blotting for three-independent biological replicates. b Preparation of PAC-seq library by poly azido-nucleotides. The pooled PAC-seq libraries from 200 bp to 300 bp were isolated from a 2% agarose gene and sequenced using Miseq. c Detection of 3′-UTR APA events in Nudt21 KD cells. The total six bar coded libraries were loaded on a flowcell (illumine V3 kit) and analyzed with a Miseq and total sequence reads were trimmed and aligned to USCS human genome browser. 82% genes (n = 5487) with one poly(A) site; 8% genes (n = 531) with two more poly(A) sites no APA and 10% genes (n = 695) with 3′-UTR APA. d Identification of 3′-UTR shortening genes. Filtering of mRNA that had more than 1 poly(A) site that showed a greater than twofold change in site usage and a change of at least 20% in the percent distal poly(A) site usage (% dPAS) upon Nudt21 KD

Fig. 4

3′-UTR shortening of Pak1 and Pak2 regulated by Nudt21. a Two poly(A) sites of Pak1 were detected in 3′-UTR and PAC-seq reads density in proximal poly(A) site (pPAS) was increased in Nudt21 KD cells compare to control cells. A representative 3′-UTR APA profile of Pak1 was shown. b Three poly(A) sites were found in 3′- UTR of Pak2 and the sequence reads in distal poly(A) site (dPAS) was significantly decreased in Nudt21 KD cells compare to control cells. A representative 3′-UTR APA profile of Pak2 is shown. Numbers on y-axis indicated PAC-Seq read coverage. c Diagram of RNase H dependent APA (RHAPA) for validation of Pak1 3′-UTR APA. D Usage of 3′-UTR PPAS of Pak1 was increased in both in LN229 and U251 with Nudt21 KD. Data from real time PCR are shown as fold change of usage in pPAS/dPAS. Data are shown as mean ± standard deviation. (p < 0.001 in LN229 and p = 0.041 in U251, two side T- test). e Diagram of RHAPA for validation of Pak2 3′-UTR APA. f Usage of pPAS of Pak2 was increased in both in LN229 and U251 with Nudt21 KD. Data are shown as fold change of usage in pPAS/ dPAS. Experiments were performed in triplicate with the data shown as mean ± standard deviation. (p = 0.022 in LN229 and p = 0.025 in U251, two side T-test)

Fig. 5

Upregulation of Pak1 regulated by Nudt21 and promoting GBM cell proliferation. a Pak1 expression increased in LN229 with Nudt21 KD. Cell lysate from siRNA-control and siRNA-Nudt21 was separated in 10% SDS-PAGE and protein expression was detected by western blot with indicated antibodies. A representative profile of Pak1 was shown from three-independent experiments. b Pak1 3′-UTR-mediated luciferase activity was increased in LN229 cells with Nudt21 KD. Data are shown as average luciferase activity ± standard deviation from three-independent experiments in triplicate (p = 0.028 and p = 0.05 two side T-test). c, d Effects of knockdown of Nudt21 and/or Pak1 on colony formation in LN229 cells and U251 cells. The ability of colony formation was evaluated by soft agar assays. Comparison of the difference of this ability between groups of negative control (NC), knockdown of Nudt21, knockdown of Pak1 or combination of knockdown of Nudt21 and Pak1 was performed. *p < 0.05 was assumed significant difference between the siRNA knockdown group cells versus control siRNA transfected cells. e, f Effects of knockdown of Nudt21 and /or Pak1 on cell migration in LN229 and U251 cells. LN229 and U251 cells were transfected with siRNA for Nudt21 and/or Pak1 for 24 h and the transfected cells were seeded into the transwell inserts of a CytoSelect 24-well cell migration plate. Qualification data showed the mean of migrating cells ± standard deviations in different groups of NC, knockdown of Nudt21, knockdown of Pak1 or combination of knockdown of Nudt21 and Pak1. *p < 0.05 was assumed significant difference between the knockdown group cells versus control siRNA transfected cells

Fig. 6

Induced expression of Pak1 predicts worse survival in two TCGA GBM cohorts. a, c Kaplan–meier plots analyses indicated that increased expression of Pak1 predicts worse survival in the TCGA GBM exon array cohort (p = 0.023, n = 348), and GDC TCGA GBM RNAseq cohort (p = 0.0047, n = 173). b, d Kaplan–meier plot analyses showed that combination of high level of Pak1 expression and low level of Nudt21 expression stratified prognosis of GBM patients in TCGA GBM exon array cohort (p = 0.00017, n = 348) and GDC TCGA GBM RNAseq cohort (p = 0.0015, n = 173)