Transcriptome 3′end organization by PCF11 links alternative polyadenylation to formation and neuronal differentiation of neuroblastoma

Diversification at the transcriptome 3′end is an important and evolutionarily conserved layer of gene regulation associated with differentiation and dedifferentiation processes. Here, we identify extensive transcriptome 3′end-alterations in neuroblastoma, a tumour entity with a paucity of recurrent somatic mutations and an unusually high frequency of spontaneous regression. Utilising extensive RNAi-screening we reveal the landscape and drivers of transcriptome 3′end-diversification, discovering PCF11 as critical regulator, directing alternative polyadenylation (APA) of hundreds of transcripts including a differentiation RNA-operon. PCF11 shapes inputs converging on WNT-signalling, and governs cell cycle, proliferation, apoptosis and neurodifferentiation. Postnatal PCF11 down-regulation induces a neurodifferentiation program, and low-level PCF11 in neuroblastoma associates with favourable outcome and spontaneous tumour regression. Our findings document a critical role for APA in tumorigenesis and describe a novel mechanism for cell fate reprogramming in neuroblastoma with potentially important clinical implications. We provide an interactive data repository of transcriptome-wide APA covering > 170 RNAis, and an APA-network map with regulatory hubs.

a Heat maps of clustered TREND-affected genes (number on the right, y-axis, scaling of the graphs corresponds to number of TREND-affected genes) grouped per functional category of depleted TREND-regulators (x-axes; hierarchical clustering according to shortening index is based on Pearson's correlation coefficient and complete linkage method). Identity of genes and TREND-signatures are displayed in detail in the TREND-DB web explorer ([http://shiny.imbei.uni-mainz.de:3838/trend-db], see also Supplementary Table 2). b Age indexing of genes with discovered TREND-isoforms (middle panel) and screened TREND-regulators (lower panel) implying a high conservation of TREND (human gene age assignments obtained from 1 ; 'ancient genes' are genes with gene age >450 million years). c Enrichment analysis of age index of TREND-affected genes upon siRNA-depletion of 174 TREND-regulators revealing that dynamic changes at the RNA 3'end are mostly found among ancient genes (orange and purple depict significant over-and underrepresentation, respectively, of genes of a particular age group (y-axis); depleted TREND-regulators are shown on the x-axis; hyper-geometric test enrichment p-values; only p-values below 0.05 are coloured). d Components of the CFIm (NUDT21 and CPSF6) and CFIIm complexes (PCF11) pervasively regulate TREND in neuroblastoma in a unidirectional manner (total number of genes affected by TREND (y-axis), Supplementary Table 2).

Supplementary Figure 4
Protein profiling revealing protein abundance changes of TREND regulators during neuronal differentiation.
Total BE(2)-C protein lysate after 7 days of ATRA differentiation analysed by western blotting (equal loading is represented by in-gel and Ponceau S staining on the membrane, top panel on the right). A merge integrating the fold-regulation of protein abundance and global effect on TREND is depicted in Fig. 3a (error bars show s.e.m. for 2 independent replicates).

Supplementary Figure 5 PCF11 is a key driver of TREND in neuroblastoma.
a Depletion of top TREND-regulators alone and in combination with PCF11 reveals a key role for PCF11 in the hierarchy regulating TREND. The heat map reflecting TREND-changes per gene shows substantial TREND-lengthening (red) and/or abrogation of TRENDshortening (blue; 'native' CPSF6-phenotype) upon co-depletion of PCF11. Of note, big clusters of genes showing TREND-lengthening are unique to dual depletions (highlighted) indicating that 3'elongated transcript isoforms (upon PCF11-depletion) harbour cis-elements responsive to the depletion of the co-depleted processing factor. This 'TREND-facilitating' role of PCF11 may reflect its known function in RNA Pol II pausing and/or termination control 2-4 thereby regulating the exposition of weak or strong polyadenylation signals to the trans-acting 3'end processing machinery (see discussion). b Minimal PCF11-alterations affect TREND most significantly (reflected by substantial lengthening of representative indicator transcripts OAZ1 and GPI, compare quantity of long (L) and short (S) transcript isoforms in northern blotting and lengthening index depicted in the lower panel; reduction of the specific PCF11 siRNA concentration by dilution down to 1:9, specific versus unspecific control siRNA, respectively, is shown in the lanes 1-5). c Lack of protein abundance changes of other core 3'end processing components upon PCF11-depletion suggest a direct TRENDregulation in neuroblastoma via PCF11 (representative loading control is shown in the lower right panel). puromycin (n = 3) and lometrexol (n = 2) (right panel, error bars show s.e.m. for replicates). c PCF11-depletion in a complementary CHP-134 neuroblastoma model leads to downregulation of MYCN and up-regulation of TUBB3 (left) and results in neurodifferentiation (micrographs on the right, scale bar 100 µm; see also main Figures e.g. Fig. 4). d PCF11 overexpression (OE, stably expressing cell line) antagonizes ATRA induced neurodifferentiation (i.e. MYCN down-regulation on the left and inhibited morphological changes on the right; micrographs scale bar 100 µm).

Set of genes significantly affected by TREND (BH p ≤ 0.05 in at least 3 out of 5 PCF11 knockdown experiments)
Average Shortening Index (see Methods) with standard error of the mean (SEM) was calculated based on at least 3 independent PCF11 knockdown experiments. Color-highligted genes resemble a 'TREND-operon' (54 genes), which show a change in protein abundance (revealed by mass spectrometry, p ≤ 0.05, n = 3)

TREND-affected genes and statistical significance (p-value) of difference between long vs short TREND-isoform abundance ratio in stage 4s samples (n=48) compared to stage 4 samples (n=65) (Data downloaded from GSE49710 and RMA normalized)
Pairwise Student's t -Test was performed to identify significant differences between cohorts Relative proportion of number of genes affected by lengthening, shortening or not significantly affected (NS) by TREND is summarized in Figure 6b UBR1