A high-content RNAi screen reveals multiple roles for long noncoding RNAs in cell division

Genome stability relies on proper coordination of mitosis and cytokinesis, where dynamic microtubules capture and faithfully segregate chromosomes into daughter cells. With a high-content RNAi imaging screen targeting more than 2,000 human lncRNAs, we identify numerous lncRNAs involved in key steps of cell division such as chromosome segregation, mitotic duration and cytokinesis. Here, we provide evidence that the chromatin-associated lncRNA, linc00899, leads to robust mitotic delay upon its depletion in multiple cell types. We perform transcriptome analysis of linc00899-depleted cells and identify the neuronal microtubule-binding protein, TPPP/p25, as a target of linc00899. We further show that linc00899 binds TPPP/p25 and suppresses its transcription. In cells depleted of linc00899, upregulation of TPPP/p25 alters microtubule dynamics and delays mitosis. Overall, our comprehensive screen uncovers several lncRNAs involved in genome stability and reveals a lncRNA that controls microtubule behaviour with functional implications beyond cell division. Long noncoding RNAs (lncRNAs) regulate key steps of cell division. Here, the authors perform a comprehensive RNAi imaging screen targeting more than 2,000 human lncRNAs, and suggest a role of chromatin-associated linc00899 in regulation of cell division by suppressing the transcription of microtubule-binding protein TPPP/p25.

L ong noncoding RNAs (lncRNAs) are defined as RNAs longer than 200 nucleotides that lack functional open reading frames, and represent a major transcriptional output of the mammalian genome 1,2 . LncRNAs control numerous cellular processes including the cell cycle, differentiation, proliferation and apoptosis [3][4][5] and their deregulation is associated with human disease including cancer 6,7 . Several lncRNAs regulate the levels of key cell cycle regulators such as cyclins, cyclin-dependent kinases (CDK), CDK inhibitors and p53 8,9 . LncRNAs are also linked to cell division as they can regulate the levels of mitotic proteins 10,11 or by modulating the activity of enzymes involved in DNA replication and cohesion 12 . In addition, lncRNAs can control chromosome segregation by controlling kinetochore formation via centromeric transcription 13 or by acting as decoys for RNA-binding proteins involved in maintaining genome stability 14,15 . All of these lncRNA-dependent functions can occur through transcriptional and posttranscriptional gene regulation, chromatin organisation and/or posttranslational regulation of protein activity 5 . These mechanisms usually involve lncRNAs establishing interactions with proteins and/or nucleic acids, which allows lncRNA-containing complexes to be recruited to specific RNA or DNA targets 16 . Although lncRNAs represent >25% of all human genes (GENCODE v24), the biological significance of the majority of lncRNAs remains unknown.
Systematic screens in human cells identify protein-coding genes involved in cell survival, cell cycle progression and chromosome segregation [17][18][19][20] . Similar loss-of function screens are performed to identify lncRNAs with cell cycle functions. For example, CRISPR interference (CRISPRi) is used in highthroughput screens to identify lncRNA loci important for cell survival 21 and revealed lncRNAs whose functions were highly cell-type specific. Similar results are obtained in a CRISPR/ Cas9 screen targeting lncRNA splice sites 22 . In a recent RNA interference (RNAi) screen targeting human cancer-relevant lncRNAs, Nötzold and colleagues used time-lapse microscopy imaging of HeLa Kyoto cells and identify 26 lncRNAs linked to cell cycle regulation and cell morphology 23 . However, this screen only studied~600 lncRNAs in the genome with respect to a limited number of phenotypes.
With an aim to identify lncRNAs with functions in cell division, we perform a more comprehensive high-content imaging RNAi screen involving the depletion of 2231 lncRNAs in HeLa cells. We develop image analysis pipelines to quantify a diverse set of mitotic features in fixed cells, and discover multiple lncRNAs with roles in mitotic progression, chromosome segregation, and cytokinesis. We focus on linc00899, a hitherto uncharacterised lncRNA that regulates mitotic progression by repressing the transcription of the microtubule-stabilising protein TPPP/p25. Our study demonstrates the regulatory function of linc00899 in mitotic microtubule behaviour and provides a comprehensive imaging data resource for further investigation of the roles of lncRNAs in cell division.

Results
High-content RNAi screen identifies lncRNAs in cell division.
To identify lncRNAs involved in regulating cell division, we performed two consecutive RNAi screens (screen A and B). Briefly, we transfected HeLa cells with the human Lincode small interfering RNA (siRNA) library targeting 2231 lncRNAs ( Fig. 1a; Supplementary Data 1) and examined their effects using highcontent screening of mitotic phenotypes. Each lncRNA was targeted with a SMARTpool of four different siRNAs. Following 48-h incubation, cells were fixed and processed for immunostaining and subsequent automated image acquisition and analysis. In screen A, antibodies targeting CEP215 (to label centrosomes), α-tubulin (to label the microtubule cytoskeleton), phalloidin (to label the actin cytoskeleton) and Hoechst (to label nuclei) were used. In screen B (Fig. 1b-d), phospho-histone H3 (PHH3; to specifically label mitotic cells), α-tubulin, γ-tubulin (to label centrosomes) and Hoechst was used. We used these two screens as independent approaches to robustly identify lncRNAs with functions in mitotic progression, chromosome segregation and cytokinesis.
In each screen, we employed automated image analysis to segment the cells and developed in-house pipelines to quantify defects in each of abovementioned categories upon lncRNA depletion ( Supplementary Fig. 1a). First, for defects in mitotic progression, we determined the percentage of mitotic cells (also called mitotic index), because an increase in the mitotic index implies a delay or block in mitotic progression. We performed nuclear segmentation and computed the mitotic index ( Supplementary Fig. 1b) where mitotic cells were identified by the presence of mitotic spindle staining (detected by α-tubulin and CEP215) in screen A or by positive PHH3 staining of chromosomes in screen B. Second, for quantification of chromosome segregation defects, a category that includes chromatin bridges and lagging chromatids, we identified anaphase cells based on α-tubulin staining between the separating nuclei, in addition to Hoechst (DNA) signal ( Supplementary Fig. 1c). Third, to evaluate defects in the execution of cytokinesis, we segmented the cytoplasm of interphase cells and scored the number of cells with cytokinetic bridges based on α-tubulin staining (Supplementary Fig. 1d).
For each category, we ascertained lncRNAs for which depletion increased the frequency of defects relative to the mean across all lncRNAs in the siRNA library. Negative control siRNA and cells without any siRNA treatment were used as controls, for which we observed no systematic differences in the frequencies of each phenotype (Fig. 1b-d). We identified candidate lncRNAs involved in mitotic progression ( Fig. 1a; linc00899 and C1QTNF1-AS1), chromosome segregation ( Fig. 1b; PP7080 and linc00883) and cytokinesis ( Fig. 1c; linc00840 and loc729970). As a positive control for mitotic progression defects, we used a SMARTpool against the protein-coding gene Ch-TOG/CKAP5, whose depletion leads to mitotic delay and increased mitotic index 24 (Fig. 1b). For chromosome segregation, we successfully identified the lncRNA NORAD (Fig. 1c), depletion of which increases the rate of chromosome segregation errors 14,15 . Supplementary Data 2 contains raw data and computed Z-scores for each lncRNA and phenotype.
To confirm our findings, we conducted a validation screen targeting the top 25 lncRNA candidates identified in the initial screens for mitotic progression and cytokinesis (Supplementary Data 3). Depletion of each lncRNA was performed in two biological (and in total eight technical) replicates. For mitotic progression, this screen corroborated the increase in mitotic index following depletion of linc00899 and C1QTNF1-AS1 (Supplementary Fig. 2a). Although loc100289019-depleted cells also displayed an elevated mitotic index, levels of the lncRNA did not change upon RNAi 25 . For the cytokinesis category, we observed an increase in the number of cells with cytokinetic bridges after linc00840 depletion and a decrease after loc729970 depletion, but neither led to multinucleation ( Supplementary Fig. 2b, c). Furthermore, elevated mitotic index and cytokinesis defects were not associated with reduced cell viability for these lncRNAs ( Supplementary Fig. 2d). As positive controls, we used Ch-TOG and ECT2 (a key regulator of cytokinesis) 26 , the depletion of which led to expected phenotypes: an increased number of mitotic and multinucleated cells, respectively (Supplementary Fig. 2a-c).
Mitotic perturbations caused by depletion of the lncRNA candidates were further characterised by time-lapse microscopy imaging to investigate the dynamics of each phenotype. As expected, a marked mitotic delay was observed in HeLa cells depleted of linc00899 and C1QTNF1-AS1, lncRNAs associated with increased mitotic index ( Supplementary Fig. 3). Next, we depleted linc00883 and PP7080, lncRNAs with potential functions in chromosome segregation ( Supplementary Fig. 4). Using timelapse microscopy imaging of HeLa Kyoto cells stably expressing histone H2B-mCherry (a chromatin marker) and eGFP-α-tubulin (a microtubule marker) 18 , we found that depletion of linc00883 and PP7080 increased the rate of chromosome segregation errors to a similar extent as that of NORAD. We then depleted linc00840 and loc729970 ( Supplementary Fig. 5), lncRNAs from the cytokinesis category, and found that knockdown of linc00840 doubled the time required for cells to cleave the cytokinetic bridge, whereas knockdown of loc729970 resulted in shorter cytokinesis. Overall, our screen identified functions of lncRNAs in the control of cell division, supporting the idea that lncRNAs play an important role in cell cycle progression.
Molecular characterisation of linc00899 and C1QTNF1-AS1. Two lncRNAs associated with delayed mitotic progression, linc00899 and C1QTNF1-AS1, were selected for in-depth functional analysis. Linc00899 and C1QTNF1-AS1 are spliced and polyadenylated lncRNAs. Linc00899 (also known as loc100271722 or ENSG00000231711) is a multi-exonic intergenic lncRNA located on chromosome 22 and is~1.6 kb long. C1QTNF1-AS1 (also known as ENSG00000265096) is an lncRNA on chromosome 17 that is~1 kb long and is antisense to a protein-coding gene "C1q And TNF Related 1" (C1QTNF1/CTRP1). Both lncRNAs are annotated in GENCODE, show signs of active transcription (Fig. 2a) and have low protein-coding potential ( Supplementary Fig. 6a). Although we did not identify a syntenic ortholog for linc00899 in the mouse genome, short stretches of conserved regions 27 are present within exon 1 ( Supplementary  Fig. 6b). This places linc00899 in a group of lncRNAs with conserved exonic sequences embedded in a rapidly evolving transcript architecture 28 . Based on the syntenic position of proteincoding gene C1QTNF1, we found a mouse ortholog for C1QTNF1-AS1 (GM11747, ENSMUG000000086514) that is also antisense to the mouse C1qtnf1. Thus C1QTNF1-AS1 is an lncRNA that is conserved across mouse and human, while linc00899 contains short conserved stretches at its 5′ end representing possible functional domains 29,30 .
We then validated the expression of both lncRNAs in HeLa cells using a variety of techniques. cDNA generated from polyadenylated RNA was used for rapid amplification of cDNA ends (RACE) to define the locations of the 5′ cap and 3′ end identifying several isoforms for linc00899 and C1QTNF1-AS1 ( Supplementary Fig. 6c). The diversity of isoforms for both lncRNAs is consistent with a previous study on lncRNA annotation 27 . Expression data from ENCODE cell lines indicated that linc00899 and C1QTNF1-AS1 were present in both the nucleus and cytoplasm of HeLa-S3 cells (Fig. 2b), which we further confirmed by RNA fluorescence in situ hybridisation (RNA FISH) (Fig. 2c). Quantitative real-time polymerase chain reaction (qPCR) analyses on RNA extracted from different cellular fractions revealed that linc00899 but not C1QTNF1-AS1 is associated with chromatin (Fig. 2d). Although some linc00899 foci remain detectable in mitotic cells (Supplementary Fig. 6d), they do not associate with mitotic chromatin, arguing against a mitotic bookmarking role for these lncRNAs. linc00899 and C1QTNF1-AS1 are estimated to occur, on average, in five and two copies per cell, respectively, and do not show cell cycle dependency in their expression ( Supplementary Fig. 6e, f). In the Genotype-Tissue Expression (GTEx) RNA-seq dataset 31 from human tissue, C1QTNF1-AS1 was highly expressed in the adrenal gland and spleen, while linc00899 was broadly expressed in most of the tissues, with uterus being the highest ( Supplementary Fig. 6g).
Linc0889 and C1QTNF1-AS1 facilitate timely mitotic progression. To characterise the mitotic phenotype further, we depleted linc00899 and C1QTNF1-AS1 in HeLa and HeLa Kyoto cells and examined the effect on mitotic progression with immunofluorescence and time-lapse microscopy imaging, respectively. Quantification of lncRNA-depleted cells using RNAi revealed an increase in the mitotic index compared to cells treated with control siRNA (Fig. 3a, b). This was confirmed by time-lapse microscopy imaging of HeLa Kyoto cells where depletion of linc00899 and C1QTNF1-AS1 resulted in increased mitotic duration ( Supplementary Fig. 7a-c). We found that cells treated with control siRNA initiated anaphase onset at 40 ± 10 min (median ± SD), whereas the cells depleted of linc00899 and C1QTNF1-AS1 initiate anaphase onset at 100 ± 173 and 240 ± 146 min, respectively (Fig. 3c). Although unaligned chromosomes near spindle poles were detectable in a small population of linc00899-depleted cells (and these showed delays of >4 h), in the majority of cells bipolar spindle formation and chromosome congression occurred with normal kinetics, and cells exhibited a delay in the metaphase to anaphase transition ( We next asked whether these phenotypes could be recapitulated by a loss-of-function (LOF) method other than RNAi. We used locked nucleic acid (LNA) oligonucleotides as a complementary method to target lncRNA expression due to their higher efficiency in depleting nuclear compared to cytoplasmic lncRNAs 32,33 . We successfully depleted both linc00899 and C1QTNF1-AS1 with two different LNAs and selected the LNA with more efficient depletion (denoted LNA1) for subsequent studies (Fig. 3e). Similar to RNAi, we observed an increase in the mitotic index following LNA1-mediated depletion of linc00899 and C1QTNF1-AS1 (Fig. 3f)  The apparent increase in α-tubulin intensity at the spindle poles of C1QTNF1-AS1-depleted cells is most likely due to spindles being short and multipolar. Together, these data indicate that linc00899 and C1QTNF1-AS1 have biological functions in regulation of mitotic progression, albeit through different mechanisms.
To further confirm the phenotype of C1QTNF1-AS1 depletion, we inserted a polyadenylation poly(A) signal (pAS) 34 downstream of the C1QTNF1-AS1 transcriptional start site (TSS) using CRISPR-Cas9 gene editing. This method allows lncRNAs to be transcribed but terminates them prematurely due to the inserted pAS, preventing the expression of full-length lncRNA transcripts. We obtained four homozygous clones for C1QTNF1-AS1 with similar knockdown efficiency, all of which displayed mitotic delay similar to Fig. 1 Identification of lncRNAs involved in regulation of cell division. a Schematic representation of the high-throughput RNAi imaging screen for lncRNAs regulating three mitotic processes: mitotic progression, chromosome segregation and cytokinesis. The screen depleted each of 2231 lncRNAs in HeLa cells using the Human Lincode siRNA library (Dharmacon). b Z-scores for mitotic progression defects upon depletion of each lncRNA in the siRNA library. Each point corresponds to a single lncRNA where the Z-score was computed based on the mean mitotic index (MI). siRNAs against the proteincoding gene Ch-TOG were used as positive controls, in addition to negative control siRNAs (Ctl, from Ambion). Representative images from the top candidate (linc00899, in blue) are also shown, with PHH3 in yellow indicating mitotic cells (white circles). c Z-scores for chromosome segregation defects upon lncRNA depletion, similar to b. The Z-score per lncRNA was computed from the mean number of chromosome segregation errors (CSE). Here NORAD (grey) was used as a positive control. Top candidates are highlighted in purple. Representative images from one of the top candidates (linc00883, in purple) are also shown with staining for α-tubulin (red), PHH3 (green) and γ-tubulin (yellow). Inset depicts normal anaphase cell (blue area) or anaphase cell with CSE (yellow area). d Z-scores for cytokinesis defects upon lncRNA depletion, similar to a. The Z-score for each lncRNA was computed based on the mean number of cells with cytokinetic bridges (CB). Representative images from the top candidate (linc00840, in orange) are shown with staining for α-tubulin (red) and DNA (blue). CB are depicted in white circles. All Z-scores shown here are from screen B. Some of the top candidates are shown in colour and labelled in each plot. The scale bar for all images is 20 μm. Tables below each panel represent the raw data and calculated Z-scores for top lncRNA candidates for each category from two independent screens (A and B).
RNAi-and LNA-mediated depletion of C1QTNF1-AS1 (Supplementary Fig. 8). These results suggest that the C1QTNF1-AS1 transcript, and not transcription at the C1QTNF1-AS1 locus, is required for the regulation of mitotic progression. Similar experiments for linc00899 were hindered by the presence of over four copies of linc00899 in the HeLa genome (https://cansar.icr.ac.uk/ cansar/cell-lines/HELA/copy_number_variation/) preventing the generation of homozygous clones despite the use of two different guide RNAs (gRNAs) targeting different regions of linc00899.
Linc0889-dependent regulation of TPPP in mitotic progression. To reveal transcriptional regulatory functions of linc00899 and C1QTNF1-AS1, we performed RNA sequencing (RNA-seq) of cells following RNAi-and LNA-mediated depletion of each lncRNA. We selected the subset of differentially expressed genes (DEGs) that changed in a consistent direction with both LOF methods to minimise any potential method specific off-target effects as shown previously 25 . For C1QTNF1-AS1, a single DEG was detected, which was C1QTNF1-AS1 itself (Fig. 4a). These data argue against a transcriptional role for C1QTNF1-AS1 and suggest that its function in mitotic progression could depend on C1QTNF1-AS1 protein interactors. For linc00899, we identified eight DEGs in common across the two LOF methods (Fig. 4a), of which four were changing in the same direction with both methods (Fig. 4b). Since we were not able to validate changes in RAI14, DNAAF5 and ITGB1BP1

C1QTNF1-AS1 Linc00899
Relative expression  expression at the mRNA and protein levels ( Supplementary  Fig. 9), we decided to focus on TPPP/p25. TPPP/p25 is a tubulin polymerisation-promoting protein with established roles in microtubule dynamics and mitosis 35,36 . TPPP was upregulated upon RNAi-and LNA-mediated depletion of linc00899 (Fig. 4c, d), which we validated at the protein level in asynchronous (Fig. 4e, f) as well as in mitotic cells (Supplementary Fig. 10a).
To address whether the mitotic delay in linc00899-depleted cells is due to increased TPPP levels, we performed single and double knockdowns of linc00899 and TPPP and analysed the mitotic progression (Fig. 4g). Whereas TPPP knockdown alone did not affect mitotic timing, depletion of linc00899 alone in HeLa cells led to a mitotic delay, consistent with results from HeLa Kyoto cells (Fig. 3c, g). However, cells with simultaneous depletion of linc00899 and TPPP progressed through mitosis with near-normal timing, a phenotype observed with both LOF methods. We confirmed that co-depletion of linc00899 and TPPP rescued the previously observed upregulation of the latter with qPCR ( Fig. 4h, i). Similar results were obtained with an additional LNA oligonucleotide targeting the first intron of linc00899 ( Supplementary Fig. 10b, c). These data suggest that linc00899 needs to be depleted at least by 50% to attain TPPP upregulation and mitotic delay in HeLa cells ( Supplementary Fig. 10d, e).
Linc00889 controls microtubule dynamics in the spindle. TPPP is known to localise to the mitotic spindle, and its overexpression influences microtubule dynamics and stability in mammalian cells [35][36][37][38] . Overexpression of TPPP increases tubulin acetylation and also microtubule stability via microtubule bundling 36 . Thus we tested whether depletion of linc00899, which leads to upregulation of TPPP, had a similar effect. Immunofluorescence of linc00899-depleted cells using antibodies against α-tubulin (a microtubule marker) and acetylated α-tubulin (a marker of long-lived microtubules) 39 showed a marked increase in acetylated α-tubulin levels (Fig. 5a), consistent with linc00899-depleted cells containing more long-lived microtubules. We next benchmarked the effects of linc00899 depletion on microtubules to those exerted by paclitaxel (taxol), a microtubule-stabilising agent that suppresses microtubule dynamics at nanomolar doses 40,41 . Cells were treated with 0.5-3 nM taxol, concentrations that suppress microtubule dynamics without affecting spindle morphology 42 because high doses of taxol (5 nM-1 µM) block microtubule depolymerisation leading to highly aberrant mitotic spindles and cell death 40 . We analysed cells after 1-or 20-h taxol treatment. We found that the impact of linc008999 depletion on acetylated α-tubulin levels was comparable to that of 3 nM taxol (Fig. 5b, c), suggestive of impaired microtubule dynamics in linc008999depleted cells.
As TPPP influences the growth velocity of microtubules by affecting their stability 36 , we examined the localisation of EB1 protein, which specifically associates with growing microtubule plus-ends where it regulates microtubule dynamics 43 . As expected, EB1 staining was apparent at the spindle pole and throughout the spindle in control cells but it was much reduced in intensity upon linc00899 depletion (Fig. 5d). Importantly, reduced EB1 signal was not due to diminished microtubule levels, because α-tubulin staining of the mitotic spindle was comparable between linc00899-depleted and control cells. Quantification of EB1 signal in mitotic cells confirmed the decrease in EB1 levels, indicating that linc00899 depletion lessens the number of growing microtubule ends, a phenotype consistent with a reduction in microtubule dynamics (Fig. 5e). Taken together, these data demonstrate that, by controlling TPPP expression levels, linc00899 limits the number of long-lived microtubules and maintains normal microtubule dynamics in mitotic HeLa cells.
To gain insight into why depletion of linc00899 and C1QTNF1-AS1 causes a mitotic delay, we examined the distribution of the kinetochore (CREST) and spindle (α-tubulin) markers in HeLa cells depleted of these lncRNAs (Fig. 6a). In contrast to controls, where most chromosomes were present in a narrow or wide metaphase plate,~20% of linc00899-depleted mitotic cells displayed narrow metaphase plates with single unattached kinetochore pairs near the poles (congression defect I; Fig. 6a and Supplementary Fig. 11). Congression defects appeared even more severe in C1QTNF1-AS1-depleted cells, which showed very wide metaphase plates with several clusters of chromosomes surrounding both spindle poles (congression defect II; Fig. 6a and Supplementary Fig. 11). Phenotypes were confirmed with LNAmediated depletion for both lncRNAs. The differences in mitotic phenotypes suggest that the mechanisms through which these two lncRNAs control mitotic progression are likely to be different.
Spindle assembly checkpoint (SAC) signalling at the kinetochores ensures that each and every chromosome is bi-oriented on the metaphase plate before anaphase is initiated 44 . To assess the status of SAC, we immunostained control and linc00899-depleted cells for the presence of Mad2, an essential component of the SAC signalling machinery 45,46 (Fig. 6b). Frequency of Mad2 on narrow metaphase plates appeared slightly reduced in linc00899-depleted cells when compared to controls (12% vs 20%). However, this may not be of biological significance because 89% of linc00899-depleted cells with narrow metaphase plates exhibited Mad2 signal on the sister kinetochores of these uncongressed chromosomes. Thus active SAC Fig. 3 Linc00899 and C1QTNF1-AS1 regulate mitosis through different mechanisms. a Expression of linc00899 and C1QTNF1-AS1 after RNAi depletion, as measured by qPCR. Results are also shown for negative control siRNAs (Ctl, from Ambion) and for cells treated with transfection reagent only (Cells). n = 3 (C1QTNF1-AS1) and 6 (linc00899) biological replicates, ****P < 0.0001 by two-tailed Student's t test. Expression values are presented relative to control siRNA. b Changes in the mitotic index (MI, based on PHH3 and Hoechst staining) after RNAi-mediated depletion of linc00899 and C1QTNF1-AS1 determined 48 h after siRNA transfection. n = 4 biological replicates, *P < 0.05 by Mann-Whitney test. c Quantification of mitotic duration from time-lapse microscopy imaging after depletion of linc00899 and C1QTNF1-AS1 in HeLa Kyoto cells. We analysed n = 183 for cells treated with Ctl si, n = 241 for Cells, n = 163 for linc00899 siRNAs and n = 147 for C1QTNF1-AS1 siRNAs. Bars show the median and the interquartile range from three biological replicates. ****P < 0.0001 by Mann-Whitney test. d Representative still images from time-lapse microscopy from c. Scale bar, 5 µm. e Expression of linc00899 and C1QTNF1-AS1 after depletion using two different LNA gapmers, as measured by qPCR. Results are also shown for negative control LNAs (Ctl LNA A and B) and for Cells. n = 3 biological replicates, ***P < 0.001 and ****P < 0.0001 by two-tailed Student's t test. Expression levels are presented relative to Ctl LNA A. f Changes in the MI after depletion of linc00899 and C1QTNF1-AS1 using LNA1 gapmers, as measured in b. n = 4 biological replicates, *P < 0.05 by Mann-Whitney test. g Quantification of mitotic duration as in c after depleting linc00899 and C1QTNF1-AS1 with LNA1 gapmers. We analysed n = 273 for Ctl LNA A, n = 312 for Cells, n = 138 for linc00899 LNA1 and n = 349 for C1QTNF1-AS1 LNA1. Bars show the median and interquartile range from three biological replicates. ****P < 0.0001 by Mann-Whitney test. h Representative still images from time-lapse microscopy from g. Scale bar, 5 µm. Data are shown as mean ± S.E.M for a, b, e, f. Mitotic duration in c, g was defined from NEBD (t = 0 min) to anaphase onset. Source data are provided as a Source Data file.
Once chromosome bi-orient, they come under tension from pulling forces generated by kinetochore microtubules. Given the central role of microtubule dynamics in this process, and its impairment in linc00899-depleted cells, we determined the interkinetochore distances (inter-KT) of chromosomes, which are known to report of the tension that kinetochore pairs are subjected to (Fig. 6c, d) 47 . Whereas in cells treated with negative control siRNA or control LNA, the mean inter-KT distance was 1.4 ± 0.2 µm, bi-oriented chromosomes in linc00899-depleted cells had a reduced mean inter-KT distance of 1.2 ± 0.2 µm  TPPP, a microtubule-stabilising protein, is a target of linc00899 in regulation of mitosis. a Venn diagram of DEGs detected by RNA-seq after depletion of linc00899 and C1QTNF1-AS1 using RNAi and LNA1 gapmers. The total number of genes was 17022. b Heat map of DEGs detected with RNAseq after depletion of linc00899 using RNAi and LNA gapmers. Three to four biological replicates were generated for each condition and compared to appropriate negative controls (Ctl)-A and B for LNA gapmers, Ambion (Amb) and Dharmacon (Dh) for RNAi. Only DEGs changing in the same direction with both LOF methods are shown. Linc00899 itself is shown as a reference. c Genome tracks of RNA-seq coverage for linc00899 before and after depletion of linc00899 using RNAi or LNA1 gapmers. The tracks were constructed from averages of 3-4 biological replicates for each condition. d Genome tracks of RNA-seq coverage for TPPP, as in c. e Representative western blot of TPPP levels after depletion of linc00899. β-Tubulin and p150 were used as two loading controls. f Densitometric analysis of TPPP levels in e. n = 6 biological replicates. **P < 0.01 and ***P < 0.001 by two-tailed Student's t test. g Quantification of mitotic duration of HeLa cells after single and double knockdown of linc00899 and TPPP using RNAi or LNAs. We analysed n = 153 for cells treated with transfection reagent alone (Cells), n = 187 for negative control siRNA (Ctl, from Ambion), n = 207 for TPPP siRNAs, n = 97 for linc00899 siRNAs, n = 204 for linc00899 and TPPP siRNAs, n = 182 for Ctl LNA A, n = 115 for linc00899 LNA1 and n = 246 for linc00899 LNA1 and TPPP siRNA. For each condition, we show the median with interquartile range from two biological replicates. ****P < 0.0001 by Mann-Whitney test. Mitotic duration was defined from NEBD (t = 0 min) to anaphase onset. h-i Expression of linc00899 and TPPP after single or double knockdown with RNAi (h) or LNA1 gapmers (i) to deplete linc000899. n = 4 biological replicates, *P < 0.05, **P < 0.01, ***P < 0.001 and ****P < 0.0001 by two-tailed Student's t test. Data are shown as mean ± S.E.M. for f, h, i. Source data are provided as a Source Data file.   , and (iv) congression defect II (brown) using α-tubulin (microtubules, in green) and CREST (kinetochore marker, in red). Data are shown as mean ± S.E.M from at least three biological replicates, *P < 0.05, **P < 0.01, ***P < 0.001 and ****P < 0.0001 by two-tailed Student's t test. Scale bar, 5 µm. b Mad2 is present on kinetochores of unaligned chromosomes in linc00899-depleted mitotic cells. Representative images of metaphase cells after RNAi-mediated depletion of linc00899. Cells were stained for Mad2 (SAC protein, in green) and with CREST (red). In cells treated with negative control siRNA (Ctl), Mad2-positive kinetochores were detectable in 20% of cells with narrow metaphase plates, whereas this figure was 12% in linc00899-depleted cells. In all, 89% of linc00899-depleted cells with congression defects were positive for Mad2. DNA was stained with Hoechst and is shown in blue in the merged images. Scale bar, 5 µm. c Representative images representing maximum intensity projections of interkinetochore (inter-KT) distance in HeLa cells after RNAi-and LNA-mediated depletion of linc00899. Cells were stained with CREST (red) and Hoechst (blue). Scale bar, 5 µm. ( Fig. 6e). Thus depletion of linc00899 leads to an overall decrease in tension across sister kinetochores.
In summary, we have demonstrated that linc00899 depletion suppresses microtubule dynamics, causes a chromosome congression defect and reduces tension on the metaphase plate. These defects are expected to preclude timely inactivation of the SAC, thus delaying anaphase onset.
Linc00899 regulates TPPP and mitosis in multiple cell lines. We next asked whether linc00899-mediated regulation of TPPP occurs in cell lines other than HeLa. RNAi-mediated depletion of linc00899 resulted in upregulation of TPPP in three normal diploid cell lines (hTERT-RPE1, retinal pigment epithelial cells; MCF10A, untransformed breast epithelial cells; HUVEC, primary human umbilical vein endothelial cells) where mitosis is~25 min (Fig. 7a, c, e). Elevated TPPP levels were accompanied with a mitotic delay from~8 to 16 min upon linc00899 knockdown (Fig. 7b, d, f). This delay was smaller compared to a mitotic delay in HeLa cells and is most likely due to the presence of at least 82 chromosomes in HeLa cells that need to be aligned at the metaphase plate, compared to 46 chromosomes present in normal diploid cells. Indeed, the mitotic timing in HeLa cells is at least 15 min longer than in RPE1, MCF10A and HUVEC cells. Thus linc00899 regulates mitotic progression by controlling TPPP levels in multiple cell lines.
In addition to its mitotic functions, TPPP is crucial for microtubule organisation in the brain. Although TPPP is present in multiple tissues, expression of TPPP is highest in the brain in both mouse 48 and human ( Supplementary Fig. 12a). In GTEx brain samples, high TPPP expression is accompanied by low levels of linc00899, and in multiple sclerosis patients TPPP and linc00899 expression were negatively correlated ( Supplementary  Fig. 12b, c). This suggests that the regulatory relationship observed in our study may be physiologically relevant in human brain tissue and neuropathological diseases.
Gain-of-function and rescue studies of linc00899. Linc00089 could regulate TPPP expression in cis (locally) or in trans (distally), as shown for multiple lncRNAs 3,5,49 . Despite linc00089 and TPPP being located on different chromosomes, a cis-acting mechanism whereby lncRNA interacts with its target site by being tethered to its site of synthesis cannot be excluded. To elucidate the mode of action by which linc00899 regulates TPPP, we compared the effects of ectopic and endogenous overexpression of linc00899 on TPPP levels. Despite an increase in linc00899 expression after its ectopic overexpression using the expression plasmid encoding linc00899 cDNA, no changes in TPPP levels were observed in HeLa and RPE1 cells (Fig. 8a). These results suggest that linc00899 is unlikely to function in trans to regulate TPPP. Indeed, rescue experiments using expression plasmid after RNAi-or LNA-mediated depletion of linc00899 failed to reduce TPPP levels (Fig. 8b). Since linc00899 and TPPP are present on different chromosomes, we tested the effect of linc00899 activation in the context of its normal genomic environment. For that purpose, we employed the CRISPR activation (CRISPRa) system 50 , which uses catalytically inactive dCas9 fused to transcriptional activator VP64, and gRNAs targeting different regions of the linc00899 promoter. We observed twofold overexpression of linc00899 from its endogenous locus, which led to downregulation of TPPP in the normal diploid RPE1 but not in HeLa cells (Fig. 8c). The lack of TPPP downregulation in HeLa cells may stem from the presence of multiple linc00899 and TPPP alleles; for instance, CRISPRa may not induce overexpression of all (four or more) linc00899 loci. In summary, our data suggest that linc00899 needs to be expressed from its own locus in order to repress TPPP expression.
Linc0889 binds and regulates transcription of TPPP. As shown in Fig. 2, linc00899 is a nuclear-and chromatin-enriched lncRNA, raising the possibility that it could directly regulate transcription of TPPP. To test this, we performed cleavage under targets and release using nuclease (CUT&RUN) 51 that allows genome-wide profiling of active and repressive histone modifications as well as transcription factors from low cell numbers. We observed an increase in trimethylation of lysine 4 on histone H3 (H3K4me3), a mark of active transcription, at the promoter of TPPP after RNAi-and LNA-mediated depletion of linc00899 (Fig. 9a). Thus elevated levels of TPPP mRNA in linc00899-depleted cells is likely to arise from increased transcription at the TPPP locus. No significant changes were seen in other active and repressive histone modifications at the TPPP locus ( Supplementary Fig. 13).
Based on these findings, we further hypothesised that linc00899 binds to regulatory regions of the TPPP locus. Therefore, we performed capture hybridisation analysis of RNA targets with sequencing (CHART-seq), a method used to identify lncRNA binding sites on chromatin [52][53][54][55] . We first mapped antisense oligonucleotides whose binding is accessible to linc00899 transcript ( Supplementary Fig. 14a) and using the antisense cocktail found approximately tenfold enrichment of linc00899 transcript compared to the control oligonucleotides (Fig. 9b). No enrichment upon pulldown was detected with the negative control transcript 5.8S. After sequencing the enriched genomic DNA using the CHART-seq protocol 56 , we identified a prominent linc00899binding site in the intron of TPPP ( Fig. 9c; Supplementary  Fig. 14b-d); however, we were not able to identify regions with sequence complementarity to the linc00899 transcript (Supplementary Fig. 15). Thus transcriptional upregulation of TPPP is likely to be mediated by linc00899 binding to the TPPP locus through protein interactions. This could occur through a 3D proximity-guided localisation mechanism, which allows lowabundant lncRNAs, such as linc00899, to identify its target genes even on different chromosomes 57 . Such proximity-guided search has been observed for Firre and CISTR-ACT lncRNAs 58,59 and could explain how linc00899, which is encoded on chromosome 22, may bind and regulate TPPP encoded on chromosome 5.
To corroborate that linc00899 act as a transcriptional repressor of TPPP, we performed co-RNA FISH using intronic probes against the premature TPPP and the mature linc00899 transcripts (Fig. 9d). TPPP has eight alleles in the HeLa cell genome (https:// cansar.icr.ac.uk/cansar/cell-lines/HELA/copy_number_variation), yet premature TPPP transcripts were detected only in one third of the cells and presented mostly as a single focus. Approximately 3% of cells showed colocalisation between mature linc00899 and premature TPPP transcript. Given that mature linc00899 transcripts can be detected in most cells, this low level of colocalisation is consistent with effective linc00899-mediated suppression of TPPP transcription.

Discussion
Previous studies have attributed the regulation of the cell cycle primarily to multi-protein networks. Here we performed a highcontent imaging screen to identify lncRNAs with functions in cell division. Development of in-house image analysis pipelines coupled with targeted validation of lncRNA-induced phenotypes allowed us to quantify the impact of lncRNA depletion on cell division. Among other lncRNAs, this study identified linc00899 and C1QTNF1-AS1 as lncRNAs involved in the control of mitotic progression.
Our results revealed that linc00899 controls mitotic progression by regulating TPPP, a protein that binds and stabilises the microtubule network at all stages of the cell cycle [35][36][37][38]48,60 . TPPP also binds to and inhibits histone deacetylase 6, an enzyme responsible for tubulin deacetylation. This binding results in increased tubulin acetylation 36 , a phenotype also observed upon linc00899 depletion. Fine-tuning TPPP protein levels seems particularly important for mitosis. Indeed, TPPP overexpression in human cells suppresses microtubule growth velocity and normal microtubule dynamics, thereby impeding timely spindle assembly and cell division 36 . Previous studies have shown that TPPP levels are subject to regulation by microRNAs 61 and protein kinases 48,60,62 ; our study now reveals lncRNA-mediated transcriptional control as an additional regulatory layer. Our mechanistic studies indicate that linc00899 regulates TPPP transcription. In particular, this notion is supported by our findings that (i) linc00899 is a nuclear and chromatin-enriched lncRNA, (ii) linc00899 needs to be expressed from its own locus in order to repress TPPP expression, (iii) H3K4me3 levels increase at the TPPP promoter in linc00899-depleted cells, and (iv) linc00899 binds at the TPPP genomic locus. It is possible that linc00899 contributes to the repressive chromatin landscape at the TPPP locus by altering local chromatin accessibility as observed with other lncRNAs 63 . The mechanism whereby linc00899 binds to the TPPP genomic locus and represses its transcription remains to be fully defined. Given the limited colocalisation between linc00899 and the premature TPPP transcript, it is more likely that linc00899 uses "proximity-guided search" where transcription site of linc00899 resides in close spatial proximity of TPPP in the nucleus. This would allow linc00899 to act   immediately upon its transcription and suppress TPPP expression possibly via interactions with protein complexes, such as chromatin regulators 57 . It is also possible that the linc00899-mediated regulation of TPPP transcription depends on linc00899 release from the chromatin, with linc00899 target specificity being guided by the pre-established chromosomal proximity, as shown for lncRNA A-ROD 64 . As we did not observe a strong sequence complementarity between the linc00899 transcript and the TPPP DNA sequence, we excluded the possibility of direct binding of linc00899 to TPPP locus. Instead, our data suggest that the linc00899 function could be mediated through linc00899-protein interactions. Further studies will be required to determine the in vivo linc00899-protein interactome and the relevance of these interactions in TPPP transcriptional regulation and cell division. TPPP is crucial for microtubule organisation in the brain and for local microtubule nucleation and myelin sheath elongation 65 . In Drosophila, TPPP mutants (also known as Ringmaker) exhibit defects in axonal extension 66 . In mammals, it is primarily expressed in oligodendrocytes where it stabilises microtubule networks, and its depletion in progenitors inhibits oligodendrocyte differentiation 61 . Indeed, TPPP-deficient mice display convulsive seizures and motor coordination deficits 65 due to hypomyelination in multiple brain regions, consistent with a defect in myelinating oligodendrocytes 67 .  Supplementary Fig. 14a), compared to a mixture of two control oligonucleotides containing parts of the linc00899 sense sequence, as measured by qPCR. The 5.8S was used as a negative control region. Data are shown mean ± S.E.M. n = 4 biological replicates. **P < 0.01 by two-tailed Student t test. c Profiling of linc00899 binding by CHART-seq after RNase H elution at the TPPP locus in HeLa cells. The region with the most significant increase in linc00899 binding compared to the sense control was present in the second intron of TPPP (arrow). All binding sites were detected using an empirical FDR threshold of 30%. d Validation of linc00899 colocalisation with the TPPP locus by co-RNA FISH in HeLa cells. Linc00899 (exonic probes against the mature transcript) colocalised to the genomic region of TPPP (intronic probes against the premature transcript) in~3% of the cells (n = 376). The nucleus was stained with DAPI (blue). Scale bar, 5 μm. Source data are provided as a Source Data file.
In humans, an inverse correlation between TPPP and linc00899 across all tissues is consistent with a regulatory relationship, with the highest expression of TPPP (and lowest expression of linc00899) being observed in the brain. This suggests that linc00899-dependent suppression of TPPP could be used to finetune TPPP expression and hence microtubule behaviour in a developmental stage-and tissue-specific manner. Intriguingly, altered TPPP protein levels have been observed in a number of neurodegenerative disorders, including multiple sclerosis and Parkinson's disease 68,69 .
In this study, we have comprehensively explored the activity of thousands of lncRNAs in cell cycle regulation and identified an assortment of lncRNAs that are involved in controlling mitotic progression, chromosome segregation and cytokinesis. While our analysis encompassed several cellular features, the imaging data from our screen allows us to extract phenotypes at any stage of cell division upon lncRNA depletion. As interest in the regulatory functions of lncRNAs increases, we anticipate that our data will serve as a powerful resource for prioritising lncRNA candidates for further studies in the RNA and cell cycle fields. High-content imaging screen: Lincode siRNA library. The Lincode siRNA Library (GE Dharmacon, G-301000) is a collection of siRNA reagents targeting 2231 human lncRNAs (1860 unique human lncRNA genes and 371 lncRNA transcripts associated with protein-coding genes). The design of this library is based on RefSeq version 54 and the siRNAs are arrayed as SMARTpools. The library was purchased at 0.1 nmol in a 96-well format. The library was diluted to a 5 μM stock with 1× siRNA buffer (GE Dharmacon, B-002000-UB-100) and arrayed onto black 384-well PerkinElmer Cell Carrier plates (Perkin Elmer, 6007550) using the Echo Liquid Handler (Labcyte). The black 384-well PerkinElmer Cell Carrier plates were prepared in advance and stored at −80°C. The final siRNA concentration per well was 20 nM. An siRNA targeting exon 1 of lncRNA GNG12-AS1 (Silencer select, Life Technologies, S59962) 70 and SMARTpool siRNAs targeting protein-coding gene CKAP5/Ch-TOG (GE Dharmacon, L-006847-00) were also included on each plate. CKAP5/Ch-TOG was used as a positive control as its depletion leads to mitotic delay and increased mitotic index 24 . ECT2 SMARTpool siRNAs (GE Dharmacon, L-006450-00-0005) were used as a positive control in the third validation screen as its depletion results in multinucleated cells 26 .

Methods
High-content imaging screen: reverse transfection. To redissolve the siRNA in the black 384-well PerkinElmer Cell Carrier plates (10 plates in total), 5 μl of OptiMEM medium (Thermo Fisher Scientific, 31985047) was added. The plates were centrifuged (1 min, 900 × g) and incubated at room temperature (RT) for 5 min. Lipofectamine RNAimax (Thermo Fisher Scientific, 13778150) was added in OptiMEM medium to a final concentration of 8 μl Lipofectamine to 1 ml Opti-MEM medium and incubated at RT for 5 min. Five μl of OptiMEM/Lipofectamine mix was then added to the plates. Plates were centrifuged and incubated at RT for 20 min. In the meantime, HeLa cells were trypsinised and counted using an automated cell counter (Countess, Thermo Fisher Scientific). Cells were centrifuged at 1000 × g for 4 min, the medium was removed and the cells were resuspended in OptiMEM medium to a final concentration of 2000 cells/well. Ten μl of cell suspension was added to the plates, and plates were centrifuged and incubated at 37°C for 4 h, before adding 10 μl of DMEM+30% FBS+3% P/S (penicillin/streptomycin, P/S) (final concentration 10% FBS, 1% P/S). Plates were centrifuged and incubated at 37°C for 48 h before fixation. A Multidrop Combi Reagent dispenser (Thermo Fisher Scientific) was used throughout the transfection protocol to ensure even liquid addition.
For screen B, the same fixation protocol as described for screen A was used. For permeabilisation, PBS/0.05% sodium dodecyl sulfate was used for 20 min at RT. Blocking was performed as described above. For the immunostaining, cells were incubated with primary antibodies against γ-tubulin (Sigma, GTU88, dilution 1:1000) and phospho-histone H3 serine 10 (PHH3, Millipore, 06-570, dilution 1:2000) for 2 h at RT and washed three times with 1× PBS before incubation with secondary antibodies (Thermo Fisher Scientific, A-31571 and A-21206, both at dilution 1:1000). After three 1× PBS washes, the cells were stained with α-tubulin (Serotec, MCA78G, dilution 1:500) and incubated with secondary antibody (Thermo Fisher Scientific, A-21434, dilution 1:1000). All primary and secondary antibodies were diluted in 1% BSA/PBS. The fixation and staining for both replicates was carried out at the Institute for Cancer Research (ICR, London) using the PerkinElmer Cell:Explorer system coupled to automated liquid handling equipment. Solutions were dispensed using a Multidrop Combi Reagent dispenser (Thermo Fisher Scientific) and aspirated/washed using a Biotek washer with 96 pins. All plates were imaged using the PerkinElmer Opera high-content confocal screening platform with spinning disc. Thirty fields of view per well were captured using a ×20 air objective, numerical aperture (NA) 0.45.
High-content imaging screen: image analysis. All image analysis was performed using custom workflows created with the Columbus software (PerkinElmer). Several output parameters were evaluated from high-content images: mitotic index (number of cells in mitosis), multinucleation index (number of multinucleated cells), number of viable cells, number of chromosome segregation errors (chromatin bridges and lagging chromatids) and number of cells with cytokinetic bridges (Supplementary Fig. 1). These are defined in more detail below.
Mitotic and multinucleation index. Nuclei were first segmented using Hoechst staining (which defines the total cell number). The false positives (e.g. dead cells) were discarded based on the nucleus area, α-tubulin and γ-tubulin/ CEP215 staining intensity. Multinucleated non-dividing cells were retained as a separate subpopulation using a two-step detection process: binucleated cells were isolated using size, aspect ratio and roundness parameters of close nucleus pairs. Other multinucleated cells were then identified among remaining cells for which αor γ-tubulin intensity was low in the perinuclear region. Further identification of mitotic cells/stages was accomplished using filters based on Hoechst for the nucleus shape and size, in combination with high PHH3 (screen B) or high α-tubulin and low CEP215 staining intensity (screen A). Notably, nuclei of cells in anaphase/ telophase stage of mitosis were small, had elongated shape and exhibited low Hoechst integrated intensity (low amount of DNA among mitotic cells). The distance between both nuclei of cells in anaphase/telophase stage was the main criteria to discard two daughter non-mitotic cells (maximum of 0.65 and 2.6 µm, respectively). From all these sub-populations, we calculated mitotic and multinucleation index relative to the total number of live cells.
Chromosome segregation errors. We started from the previous identified subpopulation of cells in anaphase/telophase stage. We filtered cells according to αtubulin staining intensity between nuclei, as cells in anaphase have lower α-tubulin intensity compared to the cells in telophase. This allowed us to identify only the cells in anaphase. To calculate the number of anaphase cells with chromosome segregation errors, the inter-nuclei space was used as the measuring area to calculate the remaining Hoechst signal. This captures both chromatin bridges and lagging chromatids.
Number of viable cells. The total number of viable cells was determined after removal of dead cells and cell debris with anaphase and telophase cells counted as one (despite exhibiting two nuclear segments). The same rule was applied for multinucleated cells.
NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-14978-7 ARTICLE Number of cytokinetic bridges. Cytokinetic bridges were defined as elongated high-intensity objects split into two parts that are positive for α-tubulin staining. We used prior segmentation of the cytoplasm and the Spot Finder feature to identify bridge half parts and sorted them as doublets by calculating the distance between them. We discarded the false positive candidates based on the shape criteria and γ-tubulin staining. The final number of cytokinetic bridges was divided by the total number of viable cells.
To minimise the variation in the cell density between different wells among all ten 384-well plates, we divided the output numbers by the total number of cells per well (dead cells were not included). Multinucleated cells were considered to be single cells during counting. The ratios were then normalised between screen plates by calculating the average value per plate and finally the grand average of all ten plates, giving a reference mean ratio. Per-well ratios were scaled so that the perplate average was equal to the reference. Z-scores (z) were calculated as follows for each parameter: where x represents the ratio for the feature of interest (e.g. mitotic/multinucleation index), μ represents the reference ratio and σ represents the standard deviation of ratios across wells. All the scripts used for the image analysis are available at https://github.com/MarioniLab/LncScreen2018.
High-content imaging screen: third-pass validation screen. The third-pass validation screen was performed in two replicates with four technical replicates using the top 25 candidates from each of the categories (mitotic progression, cytokinesis) using the same antibodies as in screen B. For validation of CHART-sequencing, Stellaris FISH Probes for intronic region of TPPP (Q670) were combined with linc00899 exonic probe (Q570) at a final concentration of 250 nM per probe set. To score whether TPPP (intronic signal) and linc00899 (exonic signal) colocalize, we only considered cells in which both signals were present. The sequences of RNA FISH probes are presented in Supplementary Data 4.
siRNA and LNA depletion experiments. HeLa cells were transfected with Lipofectamine RNAiMax reagent (Thermo Fischer Scientific) following the manufacturer's instructions. All experiments were done 48 h after transfection. The pool of four siRNA sequences (SMARTpool, Thermo Fischer Scientific) and LNA Gapmers (Exiqon) were used at a final concentration of 50 and 25 nM, respectively. For double knockdown experiments, HeLa cells (10,000 cells/well) were plated on 8-well chamber slides (Ibidi, 80826) for time-lapse microscopy imaging or in 6-well for RNA extraction (Corning, 120,000 cells/well). The cells were transfected the next day with either negative control siRNA (Ctl, from Ambion), a SMARTpool of siRNAs targeting linc00899 or TPPP or siRNAs targeting linc00899 in combination with TPPP. The same final concentration of 50 nM was achieved for both single and double knockdown by adding equal amount of control siRNA sequence (Ctl) to the single SMARTpool targeting linc00899 or TPPP separately. siRNA and LNA sequences are listed in Supplementary Tables 2 and 3, respectively.
Time-lapse microscopy imaging. HeLa (10,000 cells/well), HeLa Kyoto (10,000 cells/well), RPE1 (10,000 cells/well), MCF10A (15,000 cells/well) and HUVEC cells (15,000 cells/well) were cultured in 8-well chamber slides (Ibidi) with 200 μl/well of the corresponding medium. Imaging was performed in their corresponding medium. Time-lapse microscopy imaging was performed for all cell lines 48 h after transfection with RNAi or LNA gapmers. Mitotic duration was measured as the time from nuclear envelope breakdown until anaphase onset, based on visual inspection of the images. Cytokinesis was measured from anaphase onset to abscission completion. Live-cell imaging was performed using a Zeiss Axio Observer Z1 microscope equipped with a PL APO 0.95NA ×40 dry objective (Carl Zeiss Microscopy) fitted with a LED light source (Lumencor) and an Orca Flash 4.0 camera (Hamamatsu). Four positions were placed per well and a z-stack was acquired at each position every 10 min (HeLa, HeLa Kyoto) or 5 min (MCF10A, HUVEC, RPE1) for a total duration of 12 h. To detect chromosome segregation errors (chromatin bridges and/or lagging chromatids), HeLa Kyoto cells were imaged every 4 min with only 2 positions/well. Voxel size was 0.325 µm × 0.325 µm × 2.5 µm. Zen software (Zeiss) was used for data collection and analysis. Throughout the experiment, cells were maintained in a microscope stage incubator at 37°C in a humidified atmosphere of 5% CO 2 .
Cell cycle synchronisation. HeLa cells were grown to 50% confluency and then synchronised with thymidine for 16 h (2 mM, Sigma, T1895). The cells were washed three times with 1× PBS and then released into thymidine-free medium for 5 h. The cells were released for the indicated timepoints for RNA collection. Treatment of HeLa cells with monastrol (100 µM for 16 h; Tocris, 1305) coupled with the mitotic shake-off was used to isolate mitotic cells.
Copy number evaluation. To calculate the linc00899 and C1QTNF1-AS1 copy number, a standard curve of Ct values was generated by performing qPCR on a dilution series of known concentration of linc00899 or C1QTNF1-AS1 DNA templates (Labomics). cDNA was prepared from RNA (1 µg) extracted from the known number of HeLa cells (500,000 cells). The observed Ct values were fitted on the standard curve and the number of lncRNA molecules per cell was calculated. The final value was multiplied by 2 to account for the fact that cDNA is single stranded and DNA templates used to make the standard curve were double stranded.
RNA library preparation, sequencing and analysis. RNA-seq libraries were prepared from HeLa cells using TruSeq Stranded Total RNA Kit with Ribo-Zero Gold (Illumina, RS-122-2303). We generated four biological replicates for RNAiand LNA-mediated depletion of linc00899 and C1QTNF1-AS1. Indexed libraries were PCR-amplified and sequenced using 125 bp paired-end reads on an Illumina Hiseq 2500 instrument (CRUK CI Genomics Core Facility). Each library was sequenced to a depth of 20-30 million read pairs. Paired-end reads were aligned to the human genome hg38 with subread and the number of read pairs mapped to the exonic regions of each gene was counted for each library by using the feature-Counts 72 function in Rsubread v1.30.0 with Ensembl GRCh38 version 91. Only alignments with mapping quality scores >10 and with the first read pair on the reverse strand were considered during counting. Approximately 80% of read pairs contained one read that was successfully mapped to the reference, and 74% of all read pairs in each library were assigned into exonic regions. Any outlier samples with very low depth (resulting from failed library preparation or sequencing) were removed prior to further analysis.
The DE analysis was performed using the limma package v3.36.0 73 . First, lowly expressed genes with average counts <3 were filtered out. Normalisation was performed using the trimmed mean of M-values (TMM) method to remove composition biases. Log-transformed expression values with combined precision/ array weights were computed with the voomWithQualityWeights function 74 . The experimental design was parametrised using an additive model with a group factor, where each group was comprised of all samples from one batch/treatment combination, and an experiment factor, representing samples generated on the same day. Robust empirical Bayes shrinkage 75 was performed using the eBayes function to stabilise the variances. Testing for DEGs was performed between pairs of groups using the treat function 76 with a log-fold change threshold of 0.5. Here the null hypothesis was that the absolute log 2 -fold change between groups was ≤0.5. All pairwise contrasts involved groups from the same batch to avoid spurious differences due to batch effects. For each contrast, genes with significant differences in expression between groups were detected at an false discovery rate (FDR) of 5%. To identify DEGs that were consistent across LOF methods, an intersection-union test was performed 77 on one-sided P values in each direction, followed by an FDR correction. Coverage tracks for each library were generated using Gviz as previously described.
CHART sequencing and analysis. CHART enrichment and RNase H (NEB, M0297S) mapping was performed as previously described 52,56 . Briefly, five 150mm dishes of HeLa cells were used to prepare the CHART extract for each pulldown. In the first sonication step, the samples were sonicated using Covaris S220 (Covaris, 500217) in microTUBEs (Covaris, 520045) in a final volume of 130 µl (Programme conditions: 20% duty cycle, 200 bursts/cycle, Intensity 175, 8 min). The extracts were hybridised with a linc00899 oligonucleotide cocktail (mix of 4 oligos used at final concentration of 25 µM) overnight at RT. MyOne streptavidin beads C1 (Thermo Fischer Scientific, 65001) were used to capture complexes overnight at RT by rotation. Bound material was washed five times and RNase H (NEB) was added for 30 min at RT to elute RNA-chromatin complexes. To increase recovery yield, the remaining beads were saved and the bound material was eluted with Proteinase K (Thermo Fischer Scientific, 25530-049) in 100 µl of XLR buffer for 60 min at 55°C. The supernatant was then collected and heated for additional 30 min at 65°C. The RNase H eluate samples were also treated with Proteinase K (Thermo Fischer Scientific, 25530-049) and cross-links were reversed (55°C for 1 h, followed by 65°C for 30 min). One fifth of the total sample was used to purify RNA using the miRNeasy Kit (QIAGEN, 217004) and to calculate RNA enrichment, while the rest of the sample was used to purify DNA using Phenol-ChCl 3 :isoamyl (Thermo Fischer Scientific, 15593-031) extraction and ethanol precipitation. The final CHART DNA was eluted in 12.5 µl of 1× TE buffer (low EDTA, pH 7.4). Ten µl from this reaction was used for sonication using Covaris LE220 (programme 250 bp/10 s) to obtain average fragment size of 200-300 bp. For this, microTUBE-15 beads strips (520159, Covaris) were used in a total volume of 15 µl. After sonication, the DNA was measured by Nanodrop (input, diluted 1:20) and Qubit High Sensitivity DNA Assay (eluates were undiluted). CHART material (5 ng) was used for library preparation using the ThruPLEX DNA-Seq library preparation protocol (Rubicon Genomics, UK). For inputs, six PCR cycles were performed, while for eluates eight PCR cycles were performed owing to the lower quantity of input DNA. Library fragment size was determined using a 2100 Bioanalyzer (Agilent). Libraries were quantified by qPCR (Kapa Biosystems). Pooled libraries were sequenced on a HiSeq4000 (Illumina) according to the manufacturer's instructions to generate paired-end 150 bp reads (CRUK CI Genomics Core Facility). CHART-seq was performed from five biological replicates using a mix of two sense control oligos for linc00899 as a negative control for the linc00899 pulldown. Reads were aligned to the hg38 build of the human genome using subread in paired-end genomic mode. A differential binding (DB) analysis was performed using csaw to identify lncRNA-binding sites in the pulldown compared to the sense control. Filtered windows were obtained as described for the CUT&RUN data analysis. Normalisation factors were computed by binning read pairs into 5-kbp intervals and applying the TMM method without weighting, to account for composition biases from greater enrichment in the antisense pulldown samples. The filtered windows and normalisation factors were then used in a DB analysis with the quasi-likelihood framework in edgeR. This was performed using an additive design for the generalised linear model fit, containing terms for the batch and the pulldown (five batches in total, sense and antisense pulldowns in each batch). A P value was computed for each window by testing whether the pulldown term was equal to zero, i.e. no difference in coverage between the antisense pulldown and sense control. The above analysis was repeated using window sizes from 150 to 1000 bp to obtain DB results at varying resolutions. These results were consolidated into a single list of DB regions using consolidateWindows as previously described, yielding a combined P value for each region. The combined P values across all regions were then used to compute the empirical FDR (eFDR), defined as the ratio of the number of clusters that exhibited increased coverage in the sense control (i.e. false positives) to the number of clusters with increased coverage in the antisense pulldown (i.e. potentially true discoveries). Putative lncRNA-binding sites were defined at an eFDR of 30%. Coverage tracks for each library were generated using Gviz as previously described. The list of CHART probes and CHART primers is provided in Supplementary Tables 5 and 6, respectively.
CUT&RUN (cleavage under targets and release using nuclease). Chromatin profiling was performed according to Skene et al. 51 with minor modifications. HeLa cells were plated in 30-mm dishes (220,000 cells/dish) and transfected the next day with either negative control siRNA, linc00899 siRNA pool, negative control LNA A or an LNA gapmer targeting linc00899 (LNA1). Two biological replicates were performed per condition. Cells were washed once with PBS, spun down at 600 × g for 3 min in swinging-bucket rotor and washed twice with 1.5 ml Wash buffer (20 mM HEPES-KOH (pH 7.5), 150 mM NaCl, 0.5 mM Spermidine and 1× cOmplete™ EDTA-free protease inhibitor cocktail (Roche)). During the cell washes, concanavalin A-coated magnetic beads (Bangs Laboratories, BP531) (10 µl per condition) were washed twice in 1.5 ml Binding Buffer (20 mM HEPES-KOH (pH 7.5), 10 mM KCl, 1 mM CaCl, 1 mM MnCl 2 ) and resuspended in 10 µl Binding Buffer per condition. Cells were then mixed with beads and rotated for 10 min at RT, and samples were split into aliquots according to the number of antibodies profiled per cell type. We used 80-100,000 cells per chromatin mark. Cells were then collected on magnetic beads and resuspended in 50 µl Antibody Buffer (Wash buffer with 0.05% Digitonin and 2 mM EDTA) containing one of the following antibodies in 1:100 dilution: H3K4me3 (Millipore 05-1339 CMA304, Lot 2780484), H3K27me3 (Cell Signaling #9733 S C36B11, Lot 8), H3K36me3 (Active Motif Cat#61101, Lot 32412003), H3K27ac (Abcam ab4729, Lot GR3211741-1) and Goat Anti-Rabbit IgG H&L (Abcam ab97047, Lot GR254157-8). Cells were incubated with antibodies overnight at 4°C rotating, and then washed once with 1 ml Digitonin buffer (Wash buffer with 0.05% Digitonin). For the mouse anti-H3K4me3 antibody, samples were incubated with 50 µl of a 1:100 dilution in Digitonin buffer of secondary rabbit anti-mouse antibody (Invitrogen, A27033, Lot RG240909) for 10 min at RT and then washed once with 1 ml Digitonin buffer. Samples were then incubated in 50 µl Digitonin buffer containing 700 ng/ml Protein A-MNase fusion protein (kindly provided by Steven Henikoff) for 10 min at RT followed by two washes with 1 ml Digitonin buffer. Cells were then resuspended in 100 µl Digitonin buffer and cooled down to 4°C before addition of CaCl 2 to a final concentration of 2 mM. Targeted digestion was performed for 30 min on ice until 100 µl of 2× STOP buffer (340 mM NaCl, 20 mM EDTA, 4 mM EGTA, 0.02% Digitonin, 250 mg RNase A, 250 µg Glycogen, 15 pg/ml yeast spikein DNA (kindly provided by Steven Henikoff)) was added. Cells were then incubated at 37°C for 10 min to release cleaved chromatin fragments, spun down for 5 min at 16,000 × g at 4°C and collected on magnet. Supernatant containing the cleaved chromatin fragments were then transferred and cleaned up using the Zymo Clean & Concentrator Kit. Library preparation 78 was performed using the Thru-PLEX® DNA-Seq Library Preparation Kit (Rubicon Genomics) with a modified library amplification programme: extension and cleavage for 3 min at 72°C followed by 2 min at 85°C, denaturation for 2 min at 98°C followed by four cycles of 20 s at 98°C, 20 s at 67°C, and 40 s at 72°C for the addition of indexes. Amplification was then performed for 12 cycles of 20 s at 98°C and 15 s at 72°C (14 cycles were used for the Goat Anti-Rabbit antibody due to the lower yield of input DNA). Double-size selection of libraries was performed using Agencourt AMPure XP beads (Beckman Coulter, A63880) according to the manufacturer's instructions. Average library size was determined using an Agilent Tapestation DNA1000 High Sensitivity Screentape and quantification was performed using the KAPA Library Quantification Kit. CUT&RUN libraries were sequenced on a HiSeq2500 using a paired-end 125 bp run at the CRUK CI Genomics Core Facility. Reads were aligned to a reference genome containing the hg38 build of the human genome and the R64 build of the yeast genome, using subread v1.6.1 79 in paired-end genomic mode. A DB analysis was performed using the csaw package v1.14.0 80 to identify changes in histone mark enrichment in linc00899-depleted cells compared to the negative controls for each LOF method. Coverage was quantified by sliding a window across the genome and counting the number of sequenced fragments overlapping the window in each sample. Windows were filtered to retain only those with average abundances that were fivefold greater than the expected coverage due to background non-specific enrichment. Normalisation factors were computed by applying the TMM method 81 without weighting to the filtered windows. The filtered windows and normalisation factors were then used in a DB analysis with the quasi-likelihood framework in the edgeR package v3.22.0 82 . This was performed using an additive design for the generalised linear model fit, containing terms for NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-14978-7 ARTICLE the batch and the depletion status (two batches in total, depletion and negative control from LNA and RNAi in each batch). A P value was computed for each window by first testing whether for a depletion effect in each LOF method and then taking the larger of the two P values 77 across both LOF methods. The above analysis was repeated using window sizes from 150 to 1000 bp to obtain DB results at varying resolutions. To consolidate these results into a single list of DB regions, overlapping windows of all sizes were clustered together based on their genomic locations, using a single-linkage approach in the consolidateWindows function. For each cluster of windows, a combined P value was computed using Simes' method 83 . This represents the evidence against the global null hypothesis for each cluster, i.e. that none of the constituent windows are DB. The Benjamini-Hochberg method was used to define putative DB regions at an FDR of 5%. Coverage tracks were generated by computing the number of sequencing fragments per million overlapping each based, using library sizes adjusted according to the TMM normalisation factors for each library. Tracks were visualised using the Gviz package 84 .
Generation of C1QTNF1-AS1 CRISPR poly(A) site clones. Insertion of the transcriptional termination signal pAS 34 into the first exon of C1QTNF1-AS1 was performed using a published CRISPR/CAS9 protocol 85 . Briefly, the CRISPR pAS guide oligonucleotides were phosphorylated (T4 poly nucleotide kinase; NEB), annealed and ligated (Quick Ligase Kit, NEB) into BbsI (NEB) digested pX458 vector (Addgene, #48138). The oligonucleotides (Sigma-Aldrich, diluted in low TE EDTA buffer at 100 µM) used for the gRNA target sequence are listed in Supplementary Table 7. All inserts were verified with Sanger sequencing. HeLa cells were plated in 6-well plates (200,000 cells/well) and, on the following day, transfected with either 2.5 µg of empty PX458 vector as a control or 2.5 µg of PX458 vector where guide oligonucleotides targeting C1QTNF1-AS1 were cloned together with a symmetric single-stranded oligonucleotide donor 86 (ssODN; IDT Inc., 4 µl from 10 µM stock) containing the pAS flanked by homology arms (75 bp) to the target site. Lipofectamine 3000 reagent (Thermo Fisher Scientific, L3000015) was used as transfection reagent according to the manufacturer's instructions. The sequence of the symmetric ssODNs is as follows (5′ to 3′): 75 bp homology left arm-aataaaagatctttattttcattagatctgtgtgttggttttttgtgtg-75 bp homology right arm CTCACTGCGGGGCTCGGGAAGGAGGAAAGGAGTGAGCATGTCCTGCTCC TGCATGTCCCTGCTTAAGCTCAGGACaataaaagatctttattttcattagatctgtgtgttggtttt ttgtgtgCCCTTCCAGGCCAAGGACCCCAGCATAGACCCCAGGACAGGGCCCCA AGGATCCCTGGCTCATGAGAGCGGCTTGC The exon 1 of C1QTNF1-AS1 is shown in upper case. The pAS site is in bold and contains BglII restriction sites (AGATCT) that were used for identification of the positive clones. The PAM site (TGG) was abrogated to avoid further cleavage by Cas9.
After 48 h, the GFP-positive cells were sorted using BD FACSAria Ilu (CRUK Flow Cytometry Core Facility) and plated on 96-well plates with DMEM supplemented with 20% FBS. Single-cell clones were expanded and genomic DNA was extracted using Direct PCR Lysis reagent (Viagen, 201-Y; 25 µl/well) with Proteinase K (Thermo Fisher Scientific, 25530-049; 0.4 mg/ml). The plates were incubated for 1 h at 55°C, followed by 45 min incubation at 85°C. Two µl of genomic DNA was used for screening by PCR amplification (Phusion® High-Fidelity PCR Master Mix with HF Buffer, NEB, M0531S) of the targeted genomic region. PCR conditions for C1QTNF1-AS1-guide 70 were: 98°C 30 s, 98°C 30 s, 70°C 30 s, 72°C 30 s (repeat 30×), 72°C 90 s, 4°C forever. After PCR, 5 µl was used for BglII digestion (NEB, 2 h at 37°C) and the products were loaded on a 2% agarose gel to confirm the pAS insertion ( Supplementary Fig. 8a). The rest of PCR reaction (15 µl) was also loaded on the gel as a control. Uncropped scan of the gel is provided in the Source Data file. The same genomic PCR product was also ligated into pJET1.2/Blunt and transformed into bacteria (CloneJET; Thermo Fisher Scientific, K1231). To ensure representation by both alleles, plasmids were isolated and sequenced from a minimum of 5-10 bacterial colonies. This method revealed that C1QTNF1-AS1 pAS clones were homozygously targeted clones. Four clones were analysed (clone 95, 136, 153 and 169) by qPCR and live-cell imaging for the mitotic phenotype, along with the wild-type controls (clone 2 and clone 5) that were transfected with an empty vector PX458. We also attempted to perform pAS insertion into exon 1 of linc00899 using two different sgRNAs. However, we failed to obtain homozygous targeted clones, most likely due to the presence of four copies of linc00899 in HeLa cells (https://cansar.icr.ac.uk/cansar/cell-lines/HELA/ copy_number_variation/). The list of PCR primers for CRISPR pAS insertion is provided in Supplementary Table 8.
Mapping of 5′ and 3′ ends of linc00899 and C1QTNF1-AS1 by RACE. HeLa total RNA (1 µg) was extracted and 5′ and 3′ RACE was performed using the Smarter RACE Kit (Clontech, 634858). cDNA was synthesised using 5′ and 3′ RACE CDS primers and SMARTer IIA oligo for template switching for 5′ RACE. cDNA ends were then amplified by touchdown PCR. The first PCR (touchdown) used the following conditions: 5  For linc00899, the first PCR used Universal Primer A mixed with linc00899-5R outer 1 = GATTACGCCAAGCTTacatcccggttcccacgaaaagcaacc for the 5′ ends or linc00899-3Routerinner3 = GATTACGCCAAGCTTccagggaggggaaaggagtcggcaat for the 3′ ends. The second (nested) PCR used Nested Universal Primer A and linc00899-5Routerinner1 = GATTACGCCAAGCTTggagcaggcgaagagggagtgagggg for the 5′ ends or linc00899-3R outerinner1 = GATTACGCCAAGCTTggtcacagc ctagccaagcccagcca for the 3′ ends.
For linc00899, nine different 5′ sites and four different 3′ sites were identified. For C1QTNF1-AS1, six different 5′ sites and four different 3′ sites were identified. The fragments were purified after agarose gel electrophoresis with the QiaQuick Gel Extraction Kit (QIAGEN), cloned in pRACE vector as per the kit instructions and transformed into Stellar component cells (Clontech, 636763) Based on our sequencing results, C1QTNF1-AS1 transcript varied from 864 to 952 bp in HeLa cells. We sequenced at least 10 clones per 5′ and 3′ RACE.
Lentiviral overexpression of lncRNAs. To produce lentivirus, HEK293T cells were first transfected with 15 µg of DNA, composed of 9 µg of the lentiviral vector DNA containing the transgene, 4 µg of psPAX.2 and 2 µg of pMD2.G in the final transfection volume of 1.5 ml (including 45 µl of Trans-Lt1 transfection reagent, Mirus) using OptiMEM medium (Thermo Fisher Scientific). Viral supernatant was collected 48 and 72 h post-transfection, spun down at 1800 × g for 5 min at +4°C and filtered through a 45-µm filter. For overexpression of linc00899 in HeLa and RPE1 cells, linc00899 sequence (Labomics) and negative control vector (scrambled linc00899 sequence) were cloned into pLenti6.3/TO/V5-DEST vector (also known as lincXpress; kindly provided by John Rinn, University of Colorado) using the Gateway cloning strategy 70 . To overexpress linc00899 from its endogenous locus using CRISPRa, we used pHAGE EF1alpha dCAS9-VP64-HA (Addgene, #50918). Two gRNA sequences targeting different regions of linc00899 (guides 1 and 2) were cloned into pU6-sgRNA EF1Alpha-puro-T2A-BFP (Addgene, #60955). All clones were verified by Sanger sequencing using mU6 forward primers (the mU6 sequence is provided in Supplementary Table 8). HeLa and RPE1 cells were transduced with lentivirus containing dCAS9-VP64 and two gRNA targeting linc00899 or with lentivirus contacting negative gRNA (NC2) in the presence of polybrene (5 µg/ml, Sigma). Twenty-four hours after viral transduction, the medium was exchanged and RNA was collected 48 h later for qPCR analysis. The list of CRISPRa guide sequences and primers to overexpress linc00899 from the lincXpress vector are provided in Supplementary Tables 9 and 10, respectively.
Image processing and quantification. Imaging of fixed cells was performed on a Leica Sp8 confocal microscope using a 100 ×1.4 numerical aperture Leica oil objective. Images were taken at identical exposure times within each experiment and were imported into ImageJ and Photoshop (CS6, Adobe). Images shown here represent three-dimensional maximum intensity projections. To analyse the ratio of α-tubulin to acetylated α-tubulin, raw integrated intensities were measured (ImageJ) over the total z-stack plane using a circle selection around the mitotic spindle (based on α-tubulin). The measured intensity values were divided by the area of the selection and the background signal was subtracted. Afterwards, the mean of the signal intensity over all z-stacks was calculated, and the ratio of the signal of acetylated-tubulin to α-tubulin was determined. The total EB1 levels were quantified as described above. Inter-KT distance was measured between single kinetochore pairs visible in the same z-stack plane using ImageJ. We calculated the distance between at least seven kinetochore pairs/cell from the single focal plane using the manual line measurement tool. Subcellular fractionation. The cells from a 150-mm dish were used to isolate RNA from cytoplasmic, nucleoplasmic and chromatin fractions by TRIzol extraction 70 . Turbo DNA-free Kit (Thermo Fischer Scientific, AM1907) was used to remove any traces of DNA. Expression of target genes in each fraction was analysed by qPCR. Data were normalised to the geometric mean of GAPDH and ACTB levels in each cellular compartment. U1 small nuclear RNA was used as a positive control for chromatin fraction.
Statistical analysis. The statistical significance of data was determined by twotailed Student's t test in all experiments using GraphPad Prism unless indicated otherwise. P values >0.05 were considered statistically not significant.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.