Alpha-satellite RNA transcripts are repressed by centromere–nucleolus associations

Although originally thought to be silent chromosomal regions, centromeres are instead actively transcribed. However, the behavior and contributions of centromere-derived RNAs have remained unclear. Here, we used single-molecule fluorescence in-situ hybridization (smFISH) to detect alpha-satellite RNA transcripts in intact human cells. We find that alpha-satellite RNA-smFISH foci levels vary across cell lines and over the cell cycle, but do not remain associated with centromeres, displaying localization consistent with other long non-coding RNAs. Alpha-satellite expression occurs through RNA polymerase II-dependent transcription, but does not require established centromere or cell division components. Instead, our work implicates centromere–nucleolar interactions as repressing alpha-satellite expression. The fraction of nucleolar-localized centromeres inversely correlates with alpha-satellite transcripts levels across cell lines and transcript levels increase substantially when the nucleolus is disrupted. The control of alpha-satellite transcripts by centromere-nucleolar contacts provides a mechanism to modulate centromere transcription and chromatin dynamics across diverse cell states and conditions.


Introduction
Chromosome segregation requires the function of a macromolecular kinetochore structure to connect chromosomal DNA and spindle microtubule polymers. Kinetochores assemble at the centromere region of each chromosome. The position of centromeres is specified epigenetically by the presence of the histone H3-variant, CENP-A, such that specific DNA sequences are neither necessary nor sufficient for centromere function. (McKinley and Cheeseman, 2016). However, despite the lack of strict sequence requirements, centromere regions are typically characterized by repetitive DNA sequences, such as the alpha-satellite repeats found at human centromeres. Understanding centromere function requires knowledge of the centromere-localized protein components, as well as a clear understanding of the nature and dynamics of centromere chromatin. Although originally thought to be silent chromosome regions, centromeres are actively transcribed (Perea-Resa and Blower, 2018). Prior work has detected a-satellite transcription at centromere and pericentromere regions based on the localization of RNA polymerase II (Bergmann et al., 2012; and the production of centromere RNA transcripts Saffery et al., 2003;Wong et al., 2007). Centromere transcription and the resulting RNA transcripts have been proposed to play diverse roles in kinetochore assembly and function (Biscotti et al., 2015;Blower, 2016;Fachinetti et al., 2013;Ferri et al., 2009;Grenfell et al., 2016;Ideue et al., 2014;McNulty et al., 2017;Quénet and Dalal, 2014;Rosˇic´and Erhardt, 2016;Wong et al., 2007). However, due to limitations for analyses of centromere transcripts that average behaviors across alpha-satellite transcripts in Hela cells using smFISH. Indeed, for the ASAT probe sequences, we were able to visualize~3 foci/cell using antisense smFISH probes, similar to numbers using the sense probe set (four foci/cell) ( Figure 1C). Antisense transcription at the centromere has also been previously reported across a variety of species (Carone et al., 2009;Choi et al., 2011;Chueh et al., 2009;Ideue et al., 2014;Koo et al., 2016;Li et al., 2008;May et al., 2005).
The level of transcription for centromeric and pericentric satellite DNA has been proposed to vary between developmental stages and tissue types (Maison et al., 2010;Pezer and Ugarkovic´, 2008). In addition, changes in centromere and pericentromere transcription have been observed in cancers (Ting et al., 2011). Therefore, we next sought to analyze differences in smFISH foci across different cell lines using the ASAT and SF1 probe sets. We selected the chromosomally-unstable osteosarcoma cell line U2OS, the breast cancer cell line MCF7, and the immortalized, but non-transformed hTERT-RPE-1 cell line. We found that the levels of alpha-satellite transcripts varied modestly across cell lines ( Figure 1D,E; Figure 1-figure supplement 1C), with RPE-1 cells displaying overall lower levels of smFISH foci. As an additional confirmation of these behaviors, we tested the presence of alpha-satellite transcripts by RT-qPCR. Using a previously validated RT-qPCR primer pair against the alpha-satellite array on chromosome 21 (Molina et al., 2016;Nakano et al., 2003), we observed dramatically reduced levels of alpha-satellite transcripts in RPE-1 cells compared to HeLa cells (Figure 1-figure supplement 1D; Figure 1F). To test whether the transformation status of the cell line correlated with the level of smFISH foci, we eliminated the tumor suppressor p53 in RPE-1 cells using our previously-established inducible knockout strategy (McKinley and Cheeseman, 2017). Eliminating p53 did not substantially alter the levels of alpha-satellite smFISH foci in Rpe1 cells ( Figure 1D,E) indicating that other factors likely contribute to the observed cellular levels of alpha-satellite RNA transcripts. Together, this strategy provides the ability to quantitatively detect centromere and pericentromere-derived alpha-satellite RNA transcripts using smFISH probes against alpha-satellite sequences and demonstrates that human cell lines display varying levels of alpha-satellite transcripts.

Analysis of alpha-satellite transcript localization and cell-cycle control
We next sought to assess the localization of alpha-satellite RNA transcripts within a cell. Prior work suggested that non-coding centromere transcripts are produced in cis and remain associated with the centromere from which they are derived, including through associations with centromere proteins (McNulty et al., 2017). Other studies support the action of centromere-derived RNAs in trans (Blower, 2016), but again acting at centromeres. To investigate the distribution of the centromere transcripts, we performed combined immunofluorescence and smFISH to visualize alpha-satellite transcripts relative to centromeres and microtubules. In interphase cells, smFISH foci localized within the nucleus (Figure 2A). Thus, unlike many mRNAs, alpha-satellite-derived RNAs are not exported to the cytoplasm. Although we detected colocalization of alpha-satellite RNAs with a subset of centromeres in HeLa cells, only~10% of smFISH foci overlapped with centromeres ( Figure 2A,B). In mitotic cells, smFISH foci did not associate with chromatin ( Figure 2C). Instead, during all stages of Figure 1 continued represent the number of foci per cell for each cell test. Error bars represent the mean and standard deviation of at least 100 cells. (C) Detection of antisense alpha-satellite transcripts in HeLa cells for the ASAT smFISH probe sequences. Error bars represent the mean and standard deviation of at least 100 cells. (D) Images showing varying abundance of alpha-satellite RNA across cell lines (based on smFISH foci), with RPE-1 cells displaying overall lower levels of centromere smFISH foci. For the RPE-1 + p53 KO condition, p53 was eliminated using an established TP53 iKO cell line (McKinley and Cheeseman, 2017). (E) Left, quantification indicating the variation of smFISH foci across selected cell lines. Error bars represent the mean and standard deviation of at least 100 cells. Right, average smFISH foci/cell for multiple independent replicates to enable statistical comparisons. p-values represent T-tests conducted on replicates of smFISH foci numbers for each selected cell line. (F) Graph showing quantification of RT-qPCR for alpha-satellite transcripts from chromosome 21. Levels of chromosome 21 alpha-satellite RNAs was not detected in Rpe1 cells and was therefore set to 0 in the figure. The levels of alpha-satellite transcripts in RPE-1 cells are reduced compared to HeLa cells. A semi-quantitative assessment of the RT-PCR data (with no standard curve interpolation, see Figure 1-figure supplement 1D) indicated a~20-fold reduction in alpha-satellite transcripts in RPE-1 cells relative to HeLa. We performed three biological replicates of the RT-qPCR. Scale bars, 25 mm. The online version of this article includes the following source data and figure supplement(s) for figure 1: Source data 1. Source data for the RT-qPCR experiments shown in Figure 1F and  mitosis, alpha-satellite RNA transcripts appeared broadly distributed within the cytoplasm. Finally, as the cells exited mitosis into G1, the smFISH foci remained distinct from the chromosomal DNA and were thus excluded from the nucleus when the nuclear envelope reformed ( Figure 2D). Similar patterns of cell-cycle dependent localization changes with mitotic exclusion from chromatin have been reported for other cellular long non-coding RNAs (Cabili et al., 2015;Clemson et al., 1996).
In contrast to our findings that alpha-satellite transcripts are primarily separable from centromere loci, prior work from others found close associations between alpha-satellite transcripts and centromeres (Blower, 2016;Bobkov et al., 2018;McNulty et al., 2017;Rosˇic´et al., 2014). Based on these different behaviors, we hypothesize that the smFISH approach using the native fixation conditions detects mature alpha-satellite transcripts, but is unable to detect nascent RNAs in the process of transcription. Thus, once transcribed, alpha-satellite non-coding RNAs visualized by smFISH display nuclear localization, but are not tightly associated with the centromere regions from which they are derived. We next analyzed the temporal changes in alpha-satellite transcript numbers during the cell cycle. In contrast to other genomic loci, RNA Polymerase II is present at human and murine centromeres during mitosis Perea-Resa et al., 2020). In addition, centromere transcription during G1 has been proposed to play a role in CENP-A loading (Bobkov et al., 2018;Chen et al., 2015;Quénet and Dalal, 2014). Recent work measuring the levels of satellite transcripts originating from specific centromeres in human cells suggested the presence of stable RNA levels during the entire cell cycle (McNulty et al., 2017). smFISH provides the capacity to measure the levels of alpha-satellite transcripts in individual cells over the course of the cell cycle. We utilized combined immunofluorescence-smFISH to simultaneously label alpha-satellite RNA transcripts and microtubules, allowing us to distinguish between G1 cells (due to the presence of a mid-body), an S/ G2 interphase population, and mitotic cells. In contrast to previous observations, our analysis revealed that the transcripts detected by our smFISH method increased in S/G2 and remained stable throughout mitosis ( Figure 2E). We note that a G2/M peak of transcript levels has been reported for murine Minor Satellite transcripts (Ferri et al., 2009). However, as cells exited mitosis into G1, transcripts detected by smFISH were reduced ( Figure 2E). We speculate that this may result from the nuclear exclusion of the existing alpha-satellite transcripts, which would make this more susceptible to degradation by cytoplasmic RNAses. Thus, alpha-satellite transcript levels fluctuate over the cell cycle with G1 as a period of low transcript numbers, either indicating reduced transcription during this cell-cycle stage or the increased elimination of alpha-satellite-derived RNA transcripts.

Alpha-satellite RNAs are products of Pol II-mediated transcription
Previous studies have suggested that centromeres are actively transcribed by RNA polymerase II. RNA polymerase II localizes to centromeres in S. pombe, Drosophila melanogaster, and human cells, including at centromeric chromatin on human artificial chromosomes (HACs) and at neocentromeres (Bergmann et al., 2011;Catania et al., 2015;Chueh et al., 2009;Ferri et al., 2009;Li et al., 2008;Ohkuni and Kitagawa, 2011;Perea-Resa et al., 2020;Quénet and Dalal, 2014;Rosˇic´et al., 2014;Wong et al., 2007). However, it remains possible that additional polymerases contribute to the transcription of alpha-satellite regions. To determine the polymerases that are responsible for generating the alpha-satellite transcripts detected by our smFISH assay, we treated Hela cells with small-molecule inhibitors against all three RNA polymerases. We found a significant reduction in alpha-satellite smFISH foci following inhibition of RNA Polymerase II activity using the small-molecule THZ1 (Figure 3A-C; Figure 3-figure supplement 1A,B), which targets the RNA Pol II activator Cdk7 (Kwiatkowski et al., 2014). In contrast, we did not detect a reduction in smFISH foci following treatment with inhibitors against RNA polymerase I (small-molecule inhibitor BHM-21; Colis et al., 2014) or RNA polymerase III (ML-60218; Wu et al., RT-qPCR quantification reveals significantly reduced levels of chromosome 21 alpha-satellite transcripts of cells treated by the Cdk7 inhibitor THZ1 for 5 hr, but increased levels following RNA polymerase I inhibition (24 hr treatment) when compared to control HeLa cells. The levels of alpha-satellite RNA from chromosome 21 detected was outside of our quantifiable range in cells treated with CDK7 inhibitor and thus was set to 0. The mean of 3 biological replicates was plotted and error bars represent the standard deviation. P-value represents the results of a T-test. The online version of this article includes the following source data and figure supplement(s) for figure 3: Source data 1. Source data for the RT-qPCR experiments shown in Figure 3D. . Instead, as discussed below, we found dramatically increased alpha-satellite smFISH foci following RNA polymerase I inhibition. Consistent with the effects of RNA polymerase I and II inhibition on alpha-satellite transcript levels as detected by smFISH, RT-qPCR analyses indicated substantially decreased chromosome 21 alpha-satellite transcripts following CDK7 inhibition, but increased levels following RNA polymerase I inhibition ( Figure 3D). This indicates that the alpha-satellite RNA transcripts detected by smFISH are products of RNA Pol II-mediated transcription.

Functional analysis of the protein requirements for alpha-satellite transcripts
We next sought to determine the requirements for the production of alpha-satellite transcripts. Centromere DNA functions as a platform for assembly of the kinetochore structure (McKinley and Cheeseman, 2016), an integrated scaffold of protein interactions that mediates the connection between the DNA and microtubules of the mitotic spindle. One possibility to explain the observed transcription of centromere regions, including at neocentromere loci lacking alpha-satellite sequences, is that centromere and kinetochore components act to recruit the RNA Polymerase machinery.
To test this, we selectively eliminated diverse centromere and kinetochore components using a panel . Eliminating these centromere and kinetochore components did not prevent the presence of alpha-satellite RNA-smFISH foci ( Figure 4A). In contrast, the number of foci/cell increased in many of these inducible knockout cell lines, from moderate increases in most knockout cell lines to a substantial increase in CENP-C inducible knockout cells ( Figure 4A). This suggests that centromere components are not required for the specific recruitment of RNA Polymerase II to centromere regions, although active centromeres may act to retain RNA Polymerase II during mitosis due to the persistence of sister chromatid cohesion (Perea-Resa et al., 2020).
We also tested the contribution of non-centromere-localized cell division components to alphasatellite transcription. Because of its DNA-based nature, the centromere is subject to cell-cycle-specific challenges that include chromatin condensation, cohesion, and DNA replication. We thus sought to assess whether disruption of any of these complexes would influence alpha-satellite RNA transcript levels. To do this, we targeted proteins involved in centromere regulation (Sgo1 and BubR1), DNA replication (Mcm6, Gins1, Orc1, and Cdt1), sister chromatid cohesion (ESCO2, Scc1), chromosome condensation (Smc2, CAPG, CAPG2, TOP2A), and nucleosome remodeling (SSRP1). Strikingly, despite the diverse roles of these proteins in different aspects of centromere function, none of these inducible knockouts resulted in reduced levels of ASAT alpha-satellite transcripts as detected by smFISH analysis (Figure 4-figure supplement 1B). Instead, in many cases we detected a modest increase in alpha-satellite smFISH foci in the inducible knockout cells. Overall, our results indicate alpha-satellite transcription does not require the presence of specific DNA binding proteins, DNA structures, or cell division components, and instead that multiple factors act to restrict transcription at centromeres.

CENP-C acts to repress alpha-satellite RNA levels
Of proteins that we tested, eliminating CENP-C had a particularly substantial effect on the number of smFISH foci ( Figure 4A). To confirm this behavior following the loss of CENP-C, we repeated these experiments for both the ASAT and SF1 smFISH probes ( Figure  in a strong increase in the number of ASAT smFISH foci ( Figure 4D). Moreover, we observed a substantial increase in steady state alpha-satellite RNA levels in HeLa CENP-C inducible knockout cells based on RT-qPCR ( Figure 4E). We also note that recent work found that CENP-C overexpression resulted in decreased RNA Polymerase II occupancy at centromere regions (Melters et al., 2019). This increase in alpha-satellite transcripts in cells depleted for CENP-C depends on RNA Polymerase II, as THZ1 treatment resulted in a clear reduction in smFISH foci for the ASAT and SF1 probe sets in both control cells and CENP-C inducible knockout cells ( Figure 4C; Figure 4-figure supplement 1C). The changes in smFISH foci that we observe in the CENP-C and RNA Polymerase I-inhibited cells likely reflects the number of independent and diffusible transcripts, as we did not detect a corresponding change in smFISH focus intensity ( Figure 4F,G). Importantly, despite the increased numbers of smFISH foci in CENP-C inducible knockout cells, we found that alpha-satellite RNA transcripts displayed a similar half-life for their turnover based on RT-qPCR in control HeLa cells and CENP-C inducible knockout cells based on their loss following RNA Polymerase II inhibition (THZ1 treatment; Figure 4H). As the half-life of the alpha-satellite smFISH foci is similar in each case, this suggests that increased numbers of smFISH foci reflects increased transcription of alpha-satellite DNA instead of the increased stability of alpha-satellite transcripts. As eliminating CENP-C potently disrupts the localization of all centromere proteins (McKinley et al., 2015), this suggests that centromere and kinetochore formation could act as a physical block to restrict the passage of RNA polymerase through the centromere, downregulating alpha-satellite transcript levels. Alternatively, kinetochore proteins could act to create a repressive environment for transcription (see below).

Centromere-nucleolar associations act to repress alpha-satellite transcript levels
In the functional analysis described above, we were surprised that most perturbations resulted in increased centromere smFISH RNA foci instead of a loss of signal. The largest increases were observed for the depletion of CENP-C ( Figure 4A-C) and the inhibition of RNA Polymerase I ( Figure 3B). RNA Polymerase I transcribes rDNA, but also has an important role in assembling the nucleolus, which creates a repressive transcriptional environment. Given reported connections between the centromere and nucleolus in prior work (Ochs and Press, 1992;Padeken et al., 2013;Wong et al., 2007), we hypothesized that alpha-satellite transcription occurs at a basal level, but that this transcription is repressed by associations between the centromere and the nucleolus. To test this model, we first visualized centromeres and nucleoli in human cells. In HeLa and RPE-1 cells, a subset of centromeres overlap with the nucleolus, as marked with antibodies against Ki-67 ( Figure 5A,B) or Fibrillarin ( Figure 5-figure supplement 1A). However, other centromeres are present outside of the nucleoli within the rest of the nucleus. This contrasts with work in Drosophila cells, where centromeres from all four chromosomes are found in close proximity surrounding the Source data 1. Source data for the RT-qPCR experiments shown in Figure 4D and H. nucleolus (Padeken et al., 2013). Importantly, we observed an inverse relationship between the fraction nucleoli-localized centromeres and the numbers of alpha-satellite smFISH foci. First, we observed an increased fraction of nucleoli-localized centromeres in RPE-1 cells compared to HeLa cells ( Figure 5A,B), correlating with the reduced numbers of alpha-satellite smFISH foci in RPE-1 cells ( Figure 1F). Similarly, we found that CENP-C inducible knockout cells displayed a reduced fraction of nucleoli-localized centromeres ( Figure 5C,D), again correlating with the increased alpha-satellite smFISH foci in these cells ( Figure 4C). In contrast, we did not detect a change in centromerenucleolar associations in the CENP-B inducible knockout ( Figure 5-figure supplement 1B), which does not substantially alter smFISH foci numbers ( Figure 4A). When the different conditions affecting the nucleolus are compared, there is a clear inverse relationship between nucleolar-localized centromeres and the number of ASAT smFISH foci per cell ( Figure 5E).
To assess the functional relationship between the nucleolus and alpha-satellite transcription, we generated inducible knockout cell lines for the nucleolar components Fibrillarin and Ki-67 using our established inducible Cas9 knockout system (McKinley and Cheeseman, 2017). Induction of these knockouts resulted in increased levels of alpha-satellite transcripts, particularly for Ki-67 ( Figure 5figure supplement 1C). Although Ki-67 plays important roles in nucleogenesis, mitotic chromosome structure, and transcription of cell-cycle targets (Booth et al., 2014;Cuylen et al., 2016;Sobecki et al., 2016;Sun et al., 2017), deletion of Ki-67 is not lethal (Sobecki et al., 2016). Therefore, we additionally generated a stable Ki67 knockout cell line in HeLa cells ( Figure 5-figure supplement 1D). Ki-67 knockout cells proliferated normally, but displayed a 2-3 fold increase in alphasatellite transcript levels for both the ASAT and SF1 smFISH probes ( Figure 5F,G). However, we note that we did not detect a significant change in alpha-satellite transcript levels based on RT-qPCR ( Figure 5-figure supplement 1E). The discrepancy between the smFISH and qPCR may represent differences between single molecule and bulk assays, or technical considerations as the smFISH probes recognize alpha-satellite RNAs derived from multiple chromosomes, whereas the RT-qPCR experiments detect alpha-satellite transcripts from only chromosome 21. Despite the increased alpha-satellite transcript levels detected by smFISH, we did not detect notable consequences to centromere protein levels (based on the localization of CENP-A; Figure 5-figure supplement 1D,F) or chromosome mis-segregation (not shown). The combination of these data supports a model in which a properly functioning nucleolus and nucleolar-centromere connections act to limit centromere and pericentromere transcript levels.

Roles for alpha-satellite transcripts and centromere transcription
Together, this work defines the parameters for the production of alpha-satellite RNA transcripts and demonstrates that centromere-nucleolar connections act to restrict alpha-satellite transcription. The nature of the behavior that we observed for alpha-satellite smFISH foci, including the lack of persistent localization to centromeres or mitotic structures, is inconsistent with a direct, physical role for these transcripts in cell division processes. Instead, we propose that the process of ongoing transcription at centromeres itself, rather than the presence of alpha-satellite-derived RNAs, is an important feature of centromere biology. Active transcription could act to promote the dynamics of centromeric chromatin, resulting in the gradual turnover of DNA-bound proteins, including Figure 5 continued colocalization of centromeres with the nucleolus, as marked with antibodies against Ki-67 and CENP-A. Scale bar, 10 mm. (D) Quantification reveals that depletion of CENP-C results in a reduced fraction of nucleoli-localized centromeres (32.8%) compared to control cells (44.6%). The asterisk indicates that the data from control cells is repeated from (B). Error bars represent the mean and standard deviation of 25 cells. (E) Graph showing the relationship between the number of ASAT smFISH foci (summarized from data in Figures 1-4) and the fraction of nucleolar-localized centromeres in the indicated conditions. RNA Polymerase I inhibition should eliminate nucleolar function, and so is listed as '0' for nucleolar centromeres. Dashed line shows a linear fit trendline. (F) smFISH analysis reveals an increase of alpha-satellite transcripts in Ki67 knockout cells (right) when compared to control (left). Scale bar, 25 mm. (G) Quantification reveals a 2-3 fold increase in alpha-satellite transcript levels for both the ASAT and SF1 smFISH probes in Ki67 stable knockout cells. Error bars represent the mean and standard deviation of at least 100 cells. Right, graph showing replicates of the indicated data. P-values indicate T-tests for ASAT and SF1 replicates for Ki67 knockout cells compared to the corresponding control. The online version of this article includes the following source data and figure supplement(s) for figure 5: nucleosomes. For example, prior work using artificial tethering of chromatin and transcription factors to chromosome regions has suggested that centromere function requires an intermediate level of transcription with strongly repressive or activating states incompatible with centromere function (Molina et al., 2017;Molina et al., 2016;Nakano et al., 2008). In addition, our recent work found that transcription was required to promote the gradual turnover of CENP-A nucleosomes in nondividing cells, resulting in continued 'rejuvenation' of centromere proteins (Swartz et al., 2019). As part of that work, we also found a similar behavior for non-centromere chromatin, with transcription acting to drive the turnover of histone H3 on the chromosome arms (Swartz et al., 2019), suggesting that this may be a general feature of non-coding regions. Together, we propose that basal centromere transcription acts to promote the turnover of DNA-bound proteins, providing a mechanism to ensure refresh CENP-A chromatin. Importantly, changes in nuclear and nucleolar organization and in centromere-nucleolar associations across cell types, between cell states (including both dividing and quiescent cells), and in disease states has the potential to create consequential changes to centromere transcription and centromere protein dynamics. Inducible knockouts for nucleolar components in HeLa cells were created as described previously (McKinley and Cheeseman, 2017). Briefly, sgRNA sequences were cloned into pLenti-sgRNA (Wang et al., 2015), and used to generate lentiviruses for stable infection in cells harboring inducible Cas9 (HeLa cells -cTT20; RPE-1 cell -cTT33). Cells were then selected with puromycin as described previously (McKinley and Cheeseman, 2017). Additional cell cycle and chromosome inducible knockouts were from McKinley and Cheeseman, 2017. For the Ki67 stable knockout cell line, the HeLa cell inducible knockout version was induced with Dox and subsequently sorted by FACS to create clonal cell lines, which were screened using immunofluorescence against Ki67.

Materials and methods
Single-molecule RNA fluorescence in-situ hybridization (smFISH) Custom Stellaris RNA-FISH probes labeled with Quasar dyes (i.e., Quasar570 or Quasar670) were designed against specific centromere RNAs and purchased from Biosearch Technologies (Petaluma, CA) and PixelBiotech GmbH (Schriesheim, Germany). To conduct single-molecule FISH, cells were grown on poly-L lysine coverslip in 12-well plates were washed with PBS and fixed with 4% paraformaldehyde in 1X PBS containing RVC (Ribonucleoside Vanadyl Complex) for 10 min at room temperature (RT). After washing cells twice with 1X PBS, cells were permeabilized in 70% ethanol for at least 20 min at 4˚C. Cells were pre-incubated with 2X SSC; 10% deionized formamide for 5 min, and incubated with hybridization mix (0.1 mM RNA-FISH, 10% deionized formamide, in Hybridization Buffer (Biosearch Technologies)) overnight at 37˚C in the dark. Finally, cells were washed twice with 10% deionized formamide in 2X SSC for 30 min at 37˚C and once with Wash B (Biosearch Technologies) for 5 min at RT. For experiments with immunofluorescence coupled to smFISH, HeLa cells grown on poly-L lysine coverslip in 12-well plates were washed with PBS and fixed with 4% paraformaldehyde in 1X PBS containing RVC (Ribonucleoside Vanadyl Complex) for 10 min at RT. After washing cells with 1X PBS, cells were permeabilized for 5 min at RT with 0.1% Triton-X in PBS with RVC. After washing, primary and secondary antibody incubation was performed at RT for 1 hr in PBS and RVC. Antibody concentrations: DM1a (anti-tubulin; Sigma): 1:10000, ACA (anti-centromere antibodies -human auto-immune serum; Antibodies, Inc): 1:1000, CENP-A (Abcam, ab13939) at 1:1000, Fibrillarin (Abcam, ab5821) at 1:300, and Ki67 (Abcam, ab15580) at 1:100. For smFISH, cells were fixed again for 10 min with 4% PFA in 1x PBS. Hybridization was performed as above. Coverslips were mounted on cells with Vectashield containing Hoechst.
For imaging, slides were imaged using a DeltaVision Core microscope (Applied Precision/GE Healthsciences) with a CoolSnap HQ2 CCD camera and 60x and 100 Â 1.40 NA Olympus U-Pla-nApo objective. smFISH foci were counted per nucleus from z-projected images using CellProfiler (Carpenter et al., 2006). Prior to projection, each z-section was examined and the appropriate z-slices were projected (max intensity projection). For the analysis of smFISH foci, we used a projection of the entire cell volume. For the analysis of the overlap between smFISH foci in the nucleus and the presence of centromeres in the nucleolus, we analyzed individual z sections. Image files were processed using Deltavision software or Fiji (Schindelin et al., 2012).

DNA FISH
Custom multiplexing single-molecule FISH (smFISH) HuluFISH probes labeled with Atto565 were designed against specific centromere RNAs and purchased from PixelBiotech GmbH (Schriesheim, Germany). To conduct DNA FISH, cells were grown on poly-L lysine coverslip in 12-well plates were washed with PBS and fixed with 4% paraformaldehyde in 1X PBS for 10 min at room temperature (RT). After washing the cells with 1X PBS for 10 min, cells were permeabilized in 70% ethanol at À20C overnight. Cells were treated with RNase A (Qiagen) at 1:1000 for 30 min at 37˚C followed by denaturation in 70% deionized formamide; 2X SSC buffer at 75˚C. The samples were then dehydrated in series of cold ethanol washes (70%, 90% and 100%) for 2 min each and air dried. Cells were washed with Hulu Wash (PixelBiotech) twice for 10 min at room temperature and incubated with hybridization mix (0.5 mL HuluFISH probes in 1X HuluHyb solution) overnight at 30˚C in the dark. Finally, cells were washed twice with Hulu Wash for 30 min at room temperature.

Reverse transcription and quantitative real-time PCR
Cells were harvested 24 hr after BMH-21 addition, 5 hr after THZ1 treatment, or 4 days after doxycycline addition to induce the CENP-C knockout. RNA was purified using TRIzol reagent (Life Technologies) according to manufacturer's instructions. 2 mg of total RNA was used in the cDNA synthesis reaction with the Maxima First Strand cDNA Synthesis kit for RT-qPCR (Thermo Scientific). We used twice the recommended volume of dsDNase and allowed the DNase treatment to proceed for 30 min at 37˚C. The increased concentration of dsDNase and length of reaction was critical for the complete removal of genomic DNA. The cDNA was subjected to quantitative real-time PCR using the SYBR green PCR mastermix (Thermo Scientific) according to manufacturer's protocol. Standard curves were used for quantitative assessment of RNA levels and centromeric RNA levels were normalized to GAPDH mRNA. If the levels of centromeric RNA fall far below of the linear range in the standard curve, we noted that the RNA was not detectable and was set to 0 in the figures. After normalizing centromeric RNA levels to GAPDH, all of the data was normalized to HeLa. For RNA half-life experiments, HeLa or CENP-C inducible knockout cells (induced for 4 days) were treated with THZ1 and RNA was isolated the indicated times after THZ1 inhibition for RT-qPCR analysis as described above. The chromosome 21 alpha-satellite RNA levels were normalized to a stable mRNA, GAPDH. To calculate half-life the data was fit to an initial plateau followed by single exponential decay using GraphPad Prism. GAPDH primers: 5'-TCGGAGTCAACGGATTTGGT-3' and 5'-TTCCCG TTCTCAGCCTTGAC-3'. chromosome 21 (CH21) primers: 5'-GTCTACCTTTTATTTGAATTCCCG-3' and 5'-AGGGAATGTCTTCCCATAAAAACT-3' (Nakano et al., 2003;Molina et al., 2016).