Human cactin interacts with DHX8 and SRRM2 to assure efficient pre-mRNA splicing and sister chromatid cohesion

The cactin protein family comprises evolutionarily conserved polypeptides involved in seemingly disparate cellular processes. Isolated at first as an antigen recognized by autologous antibodies in sera from human patients with renal-cell carcinoma (Scanlan et al., 1999), cactin was subsequently identified in and named after a two-hybrid screening performed using the Drosophila melanogaster I-κB protein Cactus as bait (Lin et al., 2000). Overexpression of Cactin in a Cactus-compromised background enhanced cactus phenotypes including embryonic lethality and embryo ventralization (Lin et al., 2000). Human cactin was later found to physically and functionally interact with IκB-like protein (IκBL) and to be part of a negative feedback loop that controls NFκB transcriptional response (Suzuki et al., 2016), suggesting a conserved function for cactins in immune response and development (Atzei et al., 2010a). To date, cactin orthologs have been studied also in Toxoplasma gondii, Litopenaeus vannamei, Arabidopsis thaliana, Caenorhabditis elegans, and Danio rerio (Atzei et al., 2010b; Baldwin et al., 2013; Cecchetelli et al., 2016; Doherty et al., 2014; LaBonty et al., 2014; Szatanek et al., 2012; Tannoury et al., 2010; Zhang et al., 2014). In these organisms, cactin loss-of-function is associated with compromised cell viability and proliferation, and with developmental defects, highlighting the essentiality of cactin proteins. In Schizosaccharomyces pombe, ablation of Cactin in fission yeast 1 (Cay1) only mildly affects proliferation in standard culture conditions. However, cay1Δ cells accumulate aberrant chromosome structures and stop dividing when shifted to cold temperatures (Lorenzi et al., 2015).

Although the molecular functions of cactins remain elusive, protein interaction and localization studies have suggested connections with pre-mRNA splicing. Mass-spectrometry-based analysis of purified human and Drosophila spliceosomes identified Cactin as a component of the catalytically active spliceosome complex C (Bessonov et al., 2008; Herold et al., 2009; Jurica et al., 2002; Rappsilber et al., 2002; Zhou et al., 2002). Moreover, analysis of the protein interactome of the C. elegans cactin ortholog, CACN-1, revealed interactions with several spliceosome components (Doherty et al., 2014). Similarly, Drosophila Cactin interacts with the core spliceosome factor SmB protein (Giot et al., 2003) and the cactin interactor IκBL forms complexes with spliceosomal proteins (An et al., 2013). In A. thaliana, cactin colocalizes with the two splicing factors RSP31 and SR45 within nuclear speckles (Baldwin et al., 2013) and RNPS1, the human ortholog of SR45, also interacts with cactin (Ewing et al., 2007). Functional studies in S. pombe further support a link with pre-mRNA splicing. cay1Δ cells inefficiently splice the pre-mRNA of the telomeric factor Rap1, which promotes heterochromatin establishment at telomeres and restricts telomerase-mediated telomere elongation (Lorenzi et al., 2015; Miller et al., 2005). As a consequence, in cay1Δ cells, Rap1 protein levels are severely diminished, heterochromatin-mediated telomere silencing is weakened and telomeres are excessively elongated by telomerase (Lorenzi et al., 2015). cay1Δ cells also accumulate unprocessed precursor transcripts from Tf2 retrotransposons, a feature shared with independent yeast splicing mutants (Lorenzi et al., 2015).

Here, we show that depletion of human cactin in various cultured immortal cell types leads to premature sister chromatid separation, genome instability and cell proliferation arrest. Deep sequencing of the transcriptome of cactin-depleted cells reveals that thousands of pre-mRNAs are incompletely spliced. In particular, we find that the pre-mRNA of sororin (also known as CDCA5), a cohesin-associated factor, is aberrantly spliced and sororin protein levels are diminished upon cactin depletion. Expression of a sororin cDNA in cactin-depleted cells is sufficient to largely restore normal chromatid cohesion, revealing a fundamental role for sororin dysfunction in the cellular defects associated with cactin deficiency. Lastly, we show that cactin physically and functionally interacts with the spliceosome-associated factors DHX8 and SRRM2. Our data indicate that cactin supports normal cellular physiology by promoting efficient splicing of a multitude of pre-mRNAs, and further support the emerging idea that cactin proteins are functional components of the splicing machineries of distant eukaryotes.

To unveil the functions exerted by cactin we transfected cervical carcinoma HeLa and osteosarcoma U2OS cells with two independent siRNAs directed against the 3′-UTR of cactin mRNA (siCacD and siCacE) and harvested cells 24, 48 and 72 h later. Approximately 70% and 80% protein depletion was achieved in HeLa and U2OS cells, respectively, at the latest time points (Fig. 1A, Fig. 2A), when cell proliferation was already evidently impaired. Fluorescence-activated cell sorting (FACS) of propidium-iodide-stained cells confirmed a proliferation defect in cactin-depleted cells, which ultimately accumulated in S and G2/M phases in HeLa cells (Fig. 1B) and in G2/M phase in U2OS cells (Fig. 2B). As shown by indirect immunofluorescence, cactin-depleted cells displayed increased frequencies of nuclear foci containing the DNA damage marker 53BP1 (Fig. 1C, Fig. 2C). Similarly, another DNA damage marker, histone H2AX phosphorylated at serine 139 (γH2AX), accumulated upon cactin depletion, mostly in cells in late S and G2/M phases, as shown by flow cytometric analysis combining anti-γH2AX antibodies and 5-ethynyl-2′-deoxyuridine (EdU) incorporation (Fig. 1D, Fig. 2D). We then performed pulse-field gel electrophoresis (PFGE) of undigested genomic DNA and uncovered broken DNA in cactin-depleted cells (Fig. S1A), implying that 53BP1 foci and γH2AX accrue at least in part in response to DNA double-stranded breaks. Because Cay1 deficiency in fission yeast is associated with aberrant telomere maintenance (Lorenzi et al., 2015), the DNA damage observed in cactin-depleted cells could be elicited by dysfunctional telomeres. However, in cactin-depleted cells: (1) 53BP1 foci did not preferentially colocalize with telomeres (Fig. S1B); (2) the levels of several telomeric factors, including human RAP1 (officially known as TERF2IP), were not noticeably altered (Fig. S1C); and (3) telomere length and silencing were normal (data not shown). Thus, contrarily to what was observed in fission yeast, cactin deficiency seems not to majorly disturb telomere maintenance in human cells.

The fact that nucleotide incorporation defects preceded accumulation of γH2AX (Fig. 1D, Fig. 2D), together with the appearance of DNA breaks (Fig. S1A) suggested that cell cycle arrest in cactin-depleted cells could be triggered by activation of checkpoints responding to DNA damage caused by impaired DNA replication. Indeed, KAP1 phosphorylation at serine 824 and CHK1 phosphorylation at serine 345, which are markers for activation of the DDR checkpoint kinases ATM and ATR, respectively (Liu et al., 2000; Ziv et al., 2006), mildly accumulated in cactin-depleted cells (Fig. S1D). Similarly, phosphorylation of the single-stranded DNA-binding protein RPA32 at serine 33, a sensitive marker for DNA replication defects and S-phase checkpoint activation (Olson et al., 2006), was detected, although at very low levels, in cactin-depleted cells (Fig. S1D,E). We thus conclude that cactin depletion induces mild replication stress and activation of DNA damage checkpoints, which are likely to contribute to cell cycle progression impairment. Nevertheless, given the very low levels of phosphorylation of KAP1, CHK1 and RPA32, we do not exclude that other mechanisms not relying on DNA damage checkpoint activation could contribute to the observed proliferation defects.

We then tested whether cactin is required for normal chromosome segregation, as is the case in fission yeast cells grown in the cold (Lorenzi et al., 2015). Scoring of DAPI-stained cactin-depleted cells revealed nuclear morphology aberrations, with polylobate and fragmented nuclei progressively accruing over the experimental time course (Fig. 1E, Fig. 2E). Furthermore, metaphase chromosome spread analysis showed a prominent defect in sister chromatid cohesion, with up to ∼90% of metaphases from depleted cells comprising prematurely separated and often hyper-condensed sister chromatids (Fig. 3A,B). Taken together, these results reveal that cactin is essential to support cellular proliferation, genome stability, proper nuclear morphology and chromosome segregation. All phenotypes are true outcomes of cactin deficiency because: (1) they were induced by independent siRNAs, targeting different regions of cactin mRNA; (2) retrovirus-mediated overexpression of cactin from a UTR-less cDNA averted those phenotypes in U2OS and in human embryonic kidney HEK 293T cells expressing a short hairpin RNA (shRNA) targeting the same sequence as siCacD (Fig. 3A,B; Fig. S2A-E).

Data from different organisms suggest that cactin proteins function in pre-mRNA splicing (An et al., 2013; Baldwin et al., 2013; Bessonov et al., 2008; Doherty et al., 2014; Ewing et al., 2007; Giot et al., 2003; Herold et al., 2009; Jurica et al., 2002; Lorenzi et al., 2015; Rappsilber et al., 2002; Zhou et al., 2002). To directly test the requirement of cactin for pre-mRNA splicing, we deep-sequenced poly(A)+ RNA from U2OS cells transfected with two cactin shRNA-expressing plasmids (shCacC and shCacD), as well as from shRNA-control-transfected cells. Both shRNAs were functional as they efficiently depleted cactin and led to accumulation of G2/M cells, abnormal nuclei and 53BP1 foci (Fig. S2G-I). RNA sequencing analysis revealed a mild overall increase in the relative abundance of introns in both depleted samples (Fig. 4A). Specifically, a total of 32,661 introns from 7950 different genes were significantly differentially retained (adjusted P<0.1) in samples depleted using both shRNAs as compared with samples from shRNA-control-transfected cells (Fig. S3A; Table S1). The vast majority of these (>91% in each case) showed increased retention in the cactin-depleted samples. Affected genes participate in different cellular processes, with a striking enrichment in nucleic acid synthesis and regulation, and in ubiquitin ligase activity process (Fig. S3B). This shows that cactin supports pre-mRNA splicing of a multitude of pre-mRNAs and suggests that cactin supports normal cellular physiology by regulating various cellular pathways.

While searching for mis-spliced transcripts that could account for the cellular defects arising upon cactin depletion, the cohesin accessory factor sororin (CDCA5 in Table S1) caught our attention because erroneous splicing of its pre-mRNA was previously reported to cause premature separation of sister chromatids (Oka et al., 2014; Sundaramoorthy et al., 2014; van der Lelij et al., 2014; Watrin et al., 2014), a major phenotype induced by cactin depletion. In our RNA-seq data, reads corresponding to sororin intron 1 were over-represented in cactin-depleted samples as compared with control samples, as well as with sororin introns 2–5 both in cactin-depleted and control samples (Fig. 4B,C). Consistently, in northern blot analysis of total RNA from depleted cells, full-length, unprocessed sororin pre-mRNA only mildly accumulated, whereas the majority of sororin transcripts were slightly longer than in control samples (Fig. 4D). Reverse-transcription (RT)-PCR analysis confirmed robust retention of intron 1 upon depletion of cactin but, unexpectedly, not of the core spliceosome component SF3A-3 (Fig. 4E). By contrast, intron 5 was retained at similar levels in cactin- and SF3A-3-depleted cells (Fig. 4E). Confirming the specificity of our findings, UTR-less cactin expression alleviated sororin intron 1 retention in shCacD-transfected HEK 293T cells (Fig. S2F). Lastly, sororin intron 1 splicing was executed efficiently in different phases of the cell cycle (Fig. S2L,M), ruling out the possibility that intron 1 retention in cactin-depleted cells is an indirect consequence of G2/M arrest. Thus, cactin is essential for splicing of sororin pre-mRNA and cactin depletion acutely affects splicing of intron 1. Retention of sororin intron 1 is expected to deplete functional sororin proteins in cells either resulting from a failure in mRNA export from the nucleus or, if the aberrant mRNA is exported to the cytoplasm, from the introduction of a premature stop codon that would allow translation of a truncated protein of 117 amino acids instead of 467 (Oka et al., 2014; Sundaramoorthy et al., 2014; van der Lelij et al., 2014; Watrin et al., 2014). Indeed, western blot analysis revealed that the levels of full-length sororin were approximately halved in cactin-depleted samples when compared with control samples (Fig. 4F).

As mentioned above, sororin is a cohesin-associated factor and defects in sororin splicing strongly impair sister chromatid cohesion (Oka et al., 2014; Sundaramoorthy et al., 2014; van der Lelij et al., 2014; Watrin et al., 2014), as is also the case for cactin depletion. To investigate to what extent sororin mis-splicing contributed to the cohesion defects induced by cactin depletion, we depleted cactin in HeLa cells infected with lentiviruses carrying an intron-less and UTR-less sororin cDNA under the control of a doxycycline (dox)-inducible promoter (pL-Sor). As shown by northern blot, a short sororin transcript corresponding to spliced and UTR-less sororin (herein referred to as ectopic RNA or eRNA), was readily induced by dox (Fig. 5A). Dox treatments did not affect endogenous sororin intron 1 retention when cactin was depleted either in pL-Sor cells or empty vector control cells (pL-ev; Fig. 5A). Expression of sororin eRNA almost completely averted premature sister chromatid separation (PSCS) and to a large extent abnormally shaped nuclei. By contrast, G2/M cell accumulation was only partly rescued and 53BP1 focus frequencies were left essentially unchanged (Fig. 5B-D). Hence, cactin supports normal chromosome cohesion and nuclear structure largely by promoting accurate splicing of sororin pre-mRNA.

To further shed light on the functions associated with cactin we decided to characterize its protein interactome, as this has not yet been directly explored. We generated a HEK 293T cell line ectopically expressing a dox-inducible version of cactin N-terminally tagged with Strep–HA (SHA–cactin; Fig. S4A). Mass spectrometric analysis of proteins co-purified with SHA–cactin identified a number of putative interactors participating in diverse cellular processes (Fig. S4B-C, Table S2). We depleted the majority of the identified factors in HeLa cells using siRNAs and analyzed sororin splicing by northern blot and RT-PCR (Fig. 6A,B). Despite variable efficiencies of siRNA-mediated depletion of target mRNAs (Fig. 6B), depletion of the ATPase/RNA helicase DEAH-box helicase 8 (DHX8) and the SR protein serine/arginine repetitive matrix 2 (SRRM2), both functional components of the spliceosome, stabilized intron-1-retaining sororin transcripts (Fig. 6A,B). Although we do not exclude that sub-optimal depletion of other tested factors might have not been sufficient to induce detectable defects in sororin pre-mRNA splicing, we focused on DHX8 and SRRM2 as our data suggested both physical and functional interactions with cactin.

We first validated our mass spectrometry results by performing coimmunoprecipitation experiments using antibodies against endogenous proteins and HEK 293T cellular extracts. DHX8 and, albeit at low levels, SRRM2, were detected in protein fractions immunoprecipitated with anti-cactin antibodies, and cactin and DHX8 coimmunoprecipitated with endogenous SRRM2 (Fig. S4D). This shows that cactin physically interacts with DHX8 and with SRRM2, and that DHX8 physically interacts with SRRM2, possibly pointing to the existence of cellular protein complexes comprising the three factors. Our data also suggest that the interaction of SRRM2 with cactin-containing complexes might occur only transiently or at low levels. We then analyzed the cellular distribution of the three factors by indirect immunofluorescence following pre-extraction of soluble material. All three proteins localized within the nucleus and were excluded from the nucleolus. Cactin and DHX8 were evenly distributed within nuclei. SRRM2 formed prominent speckles generally devoid of the other two factors, but it was also diffused throughout the nucleus, covering regions containing cactin and DHX8 (Fig. S4E,F). Thus, interactions between the three factors could conceivably happen in the nucleus.

Finally, we depleted cactin, DHX8 and SRRM2 either singly or in different combinations in HeLa cells and analyzed sororin splicing, PSCS, 53BP1 foci, abnormal nuclei and cell cycle progression. DHX8 depletion largely recapitulated cactin depletion with a partly additive effect observed on accumulation of abnormal nuclei and G2/M cells when DHX8 and cactin were co-depleted (Fig. 6C-E). In SRRM2-depleted cells all examined aberrations accumulated at lower levels than in cactin- or DHX8-depleted cells. No additive effect was observed when we co-depleted SRRM2 and either cactin or DHX8 (Fig. 6C-E). Interestingly, depleting cactin or DHX8 strongly destabilized SRRM2 total levels and accumulation within speckles (Fig. 6C,F); depleting SRRM2 alone, and even more in combination with cactin, destabilized DHX8 total cellular levels (Fig. 6C); and depleting SRRM2 in combination with DHX8 destabilized cactin total cellular levels (Fig. 6C). We propose that cactin, DHX8 and SRRM2 form functionally relevant cellular complexes supporting pre-mRNA splicing and that complex assembly contributes towards stabilizing the levels of the three factors within cells. Notably, DHX8 was previously found to interact with cactin in a high-throughput study (Huttlin et al., 2015), and MOG-5, the C. elegans ortholog of DHX8, was previously isolated as a component of the CACN-1 interactome (Doherty et al., 2014). These data further corroborate our results and suggest that cactin–DHX8 interactions are conserved throughout evolution.

Using a protein depletion approach, we have shown that cactin is essential for normal physiology of human cells. Moreover, consistent with previous reports implicating cactins in pre-mRNA splicing, we find that cactin promotes faithful splicing of thousands of different transcripts. The exact mechanism by which cactin promotes splicing will be the subject of future investigations; nevertheless, its physical interaction with DHX8 and SRRM2, and other members of the spliceosome machinery from different organisms (Bessonov et al., 2008; Doherty et al., 2014; Ewing et al., 2007; Giot et al., 2003; Herold et al., 2009; Jurica et al., 2002; Rappsilber et al., 2002; Zhou et al., 2002), suggests that it is a component of the active spliceosome. Prp22p and Cwc21p, the putative budding yeast orthologs of DHX8 and SRRM2, respectively, participate in the selection of intronic 3′ end splice sites (3′ss) (Gautam et al., 2015; Schwer and Gross, 1998; Semlow et al., 2016). It is conceivable that erroneous 3′ss selection could occur when cactin is depleted. Moreover, the helicase activity of Prp22p is required for proper release of mature mRNAs after completion of the splicing reaction (Company et al., 1991; Schwer, 2008; Schwer and Gross, 1998), suggesting that cactin could also participate in spliceosome disassembly and mRNA export. Further supporting a direct role for cactins in pre-mRNA splicing, mouse cactin has recently been identified as an RNA binding protein in a large-scale study where protein-RNA photocrosslinking was combined with quantitative mass spectrometry (He et al., 2016).

Impaired splicing of pre-mRNAs coding for factors involved in different cellular processes could already explain the diversity of the defects arising upon cactin inactivation. We have demonstrated that complete splicing of sororin pre-mRNA is sufficient to largely explain how cactin promotes proper chromosome separation and, to some extent, cell cycle progression. Aberrant splicing of transcripts involved in suppressing DNA damage accumulation or in DNA damage repair could explain why cactin-depleted cells show marks of damaged DNA. Consistently, accumulating evidence indicates that in the absence of functional splicing cells are unable to properly repair broken DNA. For example, genome-wide siRNA screens have shown that depletion of various factors involved in splicing and RNA processing impairs DNA double-strand break repair, thus inducing genomic instability (Adamson et al., 2012; Paulsen et al., 2009). Moreover, splicing promotes proper function of the E3 ubiquitin ligase RNF8 at sites of DNA damage (Pederiva et al., 2016). According to our transcriptome analysis, several transcripts encoding factors supporting genome stability are mis-spliced when cactin is depleted. Amongst these, introns from RAD50, RAD52, UPF1, RIF1 and RNF8 itself, all involved in maintaining genome integrity, are over-represented in libraries from cactin-depleted cells (see Table S1). In a similar manner to the approach that we have used for sororin, expression of intron-less cDNAs for each of these factors in combination with cactin depletion will help in understanding to what extent they are involved in the DNA damage response arising when cactin is depleted.

However, our characterization of the protein interactome of cactin suggests the intriguing and not mutually exclusive possibility that the pleiotropy of cactin inactivation defects derives from concomitant impairment of pathways that do not intersect with pre-mRNA splicing. We found that cactin physically interacts with RAD50 and Mre11 (Fig. S4C, Table S2), which form the MRN complex together with NBS1. MRN supports genome stability by mediating DNA double-strand break repair and DNA recombination (Stracker and Petrini, 2011). It is therefore possible that cactin deficiency causes genome instability not only by affecting splicing but also by directly interfering with MRN functions.

We have also found that cactin physically interacts with CUL7 (Fig. S4C, Table S2), a large E3 ubiquitin ligase conserved only in vertebrates. CUL7 forms the so-called 3M complex together with the two other core components, OBSL1 and CCDC8. Mutually exclusive mutations in one of the three genes are associated with the 3M syndrome, a rare hereditary disorder characterized by severe pre- and post-natal growth retardation (Clayton et al., 2012). The 3M complex also interacts with several additional proteins, including FBXW8, a substrate adaptor for CUL7 (Yan et al., 2014). OBSL1, CCDC8 and FBXW8 were all identified in our cactin interactome (Table S2), making cactin a possible component of the 3M complex. In this regard, cactin was not identified in a mass-spectrometry-based study where CUL7, OBSL1 and CCDC8 were over-expressed in HEK 293 cells and used as bait (Hanson et al., 2014). Because we performed our experiments in HEK 293T cells (HEK 293 cells stably expressing the SV40 large T antigen), it is possible that the SV40 large T antigen, which is known to associate with CUL7 and inhibit its ubiquitin ligase activity (Hartmann et al., 2014), stimulates cactin interaction with the 3M complex. Alternatively, cactin might be associated with the other elements of the 3M complex only transiently or at very low levels and therefore escaped detection in experiments where CUL7, OBSL1 and CCDC8 were used as bait (Hanson et al., 2014). Because the 3M complex supports microtubule and genome integrity (Yan et al., 2014), the genomic instability and cell proliferation defects observed in cells depleted for cactin could be linked to impaired 3M complex function. Our data also imply the possible existence of so-far-unexplored links between cactin and developmental disorders or between the 3M syndrome and pre-mRNA splicing defects. Although preliminary experiments did not reveal defects in sororin intron 1 splicing when CUL7 was siRNA-depleted (Fig. 4D,E and data not shown), the latter hypothesis is consistent with the enrichment of several pre-mRNA splicing factors in the CUL7, OBSL1 and CCDC8 interactome (Hanson et al., 2014).

Overall, cactin seems to be a versatile, multifunctional player participating in several cellular processes essential for cell viability and/or proliferation, and chromosome stability at large. Our study, besides corroborating the essentiality of cactin for cell physiology and its direct connections with pre-mRNA splicing, expands our understanding of the broad spectrum of cellular and organismal defects caused by cactin deficiency across eukaryotes.

A plasmid containing full-length cactin (C19orf29) cDNA was purchased from Origene and utilized for successive plasmid constructions. pB-Cac and pSHA-Cac were generated by ligating an N-terminally HA–Strep-tagged cactin cDNA into the pBABE-Hygro and the pcDNA5/FRT/TO-Hygro (Life Technologies) plasmids, respectively. pL-Sor was generated by ligating sororin cDNA PCR-amplified from reverse-transcribed HeLa total RNA into the pLVX-Puro vector (Clontech).

Osteosarcoma U2OS (a kind gift from Massimo Lopes, IMCR, Zürich, Switzerland), cervical carcinoma HeLa (ATCC), human embryonic kidney expressing SV40 large T antigen HEK 293T (ATCC) and Flp-In T-Rex 293 (Thermo Fisher Scientific) cells were cultured in high-glucose D-MEM (Invitrogen) supplemented with 10% TET-free FBS (Pan Bio Tech) and 100 U/ml penicillin–streptomycin (Sigma-Aldrich). All cell lines were routinely tested for bacteria and mycoplasma contaminations. For constitutive expression of ectopic cactin, U2OS and HEK 293T cells were infected with retroviruses prepared according to standard protocols using pB-Cac, followed by selection with 200 µg/ml Hygromycin B (Fluka). For inducible expression of ectopic cactin, Flp-In T-Rex 293 cells were transfected with pSHA-Cac, followed by selection with 200 µg/ml Zeocin (Invitrogen). For inducible expression of ectopic sororin, HeLa cells were infected with lentiviruses prepared according to standard protocols using pL-Sor, followed by selection with 1 µg/ml puromycin (Sigma-Aldrich) and clonal cell line isolation. For transcription induction, cells were treated with 500 ng/ml doxycycline (Sigma-Aldrich). For DNA damage induction, cells were treated with 5 µg/ml bleomycin (Sigma-Aldrich) for 2 h or with 100 nM camptothecin (Sigma-Aldrich) for 12 h or 500 nM for 4 h. For cell cycle synchronization, cells were treated for 20 h with 2 mM thymidine (Sigma-Aldrich) and then released into normal medium. 4 h after release, cells were incubated with 100 ng/ml nocodazole for 12 h, collected by mitotic shake-off and released into normal medium. shRNA-expressing plasmids were generated in pSUPER-Puro (Azzalin and Lingner, 2006) and transfected using Lipofectamine 2000 reagent (Invitrogen), followed by selection with 1 µg/ml puromycin (Sigma-Aldrich). mRNA target sequences were as follows: 5′-GCTTCGAGTGGAACAAGTACAAC-3′ (shCacC); 5′-ATGGAATTGGCCTATTGGCAAGA-3′ (shCacD); 5′-GGAAATAGTTCCGTTTGTTTCTC-3′ (shCacE); 5′-GGAGGAGCTCAATGCCATT-3′ (shSF3A-3). siRNAs were purchased from Qiagen and transfected at 20–50 nM using Lipofectamine RNAiMAX (Invitrogen). siRNAs were as follows: siCtrl (#1027310); siCacD (5′-ATGGAATTGGCCTATTGGCAA-3′); siCacE (5′-AAATAGTTCCGTTTGTTTCTC-3′); siCUL7 (#SI00357000); siOBSL1 (#SI04339951); siCCDC8 (#SI05077086); siCK2α (#SI02660504); sicXorf56 (#SI03195731); siUBR5 (#SI03074225); siUSP9X (#SI00066584); siDDX1 (#SI00299978); siDHX8 (#SI03019373); siSRRM2 (#SI04173995); siSRPK1 (#SI02223109).

Approximately 1×10⁸ Flp-In T-Rex 293 cells expressing HA–Strep-tagged cactin were harvested and resuspended in ice-cold HNN lysis buffer [50 mM HEPES, 150 mM NaCl, 50 mM NaF, 0.5% NP-40, 1 mM PMSF, 1× phosphatase and 1× protease inhibitor cocktails (Sigma-Aldrich), 10 µg/ml Avidin (IBA)]. After 10 min incubation on ice, lysates were centrifuged at high speed and supernatants were loaded on Bio-Spin chromatography columns (Bio-Rad) containing 200 µl of 50% slurry Strep-Tactin sepharose beads (IBA) equilibrated in HNN lysis buffer. Beads were washed twice with HNN lysis buffer and three times with HNN buffer (50 mM HEPES, 150 mM NaCl, 50 mM NaF) followed by elution with 0.5 mM Biotin (Sigma-Aldrich) in HNN buffer. Eluted proteins were boiled in 2× Laemmli buffer at 95°C for 5 min and loaded onto gradient polyacrylamide gels (Invitrogen). Gels were fixed and stained with AgNO₃ according to standard protocols and bands were excised and sent for mass-spectrometry-based analysis to Alphalyse (Denmark).

Approximately 5×10⁶ cells were harvested and resuspended in ice-cold lysis buffer [50 mM Tris pH 7.4, 1.5 mM MgCl₂, 1 mM EDTA, 150 mM NaCl, 0.5% Triton X-100, 1 mM DTT, 1× phosphatase and 1× protease inhibitor cocktails (Sigma-Aldrich)] followed by incubation on ice for 30 min. Supernatants were recovered by high-speed centrifugation and pre-cleared by incubation with 20 µl of 50% slurry Protein A/G agarose beads (Santa Cruz Biotechnology) at 4°C on a wheel. Beads were eliminated by high-speed centrifugation and 500 µg of proteins were incubated with 4 µg of primary antibodies for 4 h at 4°C. Antibodies were as follow: mouse monoclonal anti-HA (Santa Cruz Biotechnology, sc-57592); mouse monoclonal anti-Myc (Cell Signaling, 2276); rabbit polyclonal anti-cactin (Bethyl Laboratories, A303-349A); rabbit polyclonal anti-CUL7 (GeneTex, GTX113906); mouse monoclonal anti-SRRM2 (Santa Cruz Biotechnology, sc390315). Lysates were incubated with 30 µl of 50% slurry Protein A/G beads for ∼16 h at 4°C on a rotating wheel. Beads were washed three times with IP buffer (50 mM Tris pH 7.4, 1.5 mM MgCl₂, 1 mM EDTA, 150 mM NaCl, 0.5% Triton X-100), once with wash buffer (50 mM Tris pH 7.4, 1.5 mM MgCl₂, 1 mM EDTA, 150 mM NaCl) and then boiled in 2× Laemmli buffer for 95°C for 5 min.

Western blot analysis was carried out according to standard procedures using the following antibodies: rabbit polyclonal anti-cactin (Abnova, PAB23952, diluted 1:1000); rabbit polyclonal anti-sororin (kind gift from Jan-Michael Peters, IMP, Vienna, Austria; 1:1000); rabbit polyclonal anti-UPF1 (Chawla et al., 2011) (raised against a peptide corresponding to the UPF1 C-terminal sequence ERAYQHGGVTGLSQY at PolyPeptide Group; 1:1000); mouse monoclonal anti-golgin 97 (Molecular Probes, A-21270, 1:1000); rabbit polyclonal anti-lamin-B1 (GeneTex, GTX103292, 1:1000); mouse monoclonal anti-actin (Abcam, ab8224, 1:5000); rabbit polyclonal anti-DHX8 (Bethyl Laboratories, A300-624A-M, 1:1000); mouse monoclonal anti-SRRM2 (Santa Cruz Biotechnology, sc-390315, 1:200); mouse monoclonal anti-TRF2 (Millipore, 05-521, 1:1000); mouse monoclonal anti-TRF1 (Abcam, ab10579, 1:1000); rabbit polyclonal anti-RAP1 (Bethyl Laboratories, A300-306A, 1:500); mouse monoclonal anti-TPP1 (Abnova, H0006557-M02, 1:500); mouse monoclonal anti-HA (Santa Cruz Biotechnology, sc-57592, 1:2000); mouse monoclonal anti-Myc (Cell Signaling, 2276, 1:2000); rabbit polyclonal anti-KAP1 (Bethyl Laboratories, A300-274A, 1:2000); rabbit polyclonal anti-phospho-KAP1 S824 (Bethyl Laboratories, A300-767A, 1:1000); mouse monoclonal anti-CHK1 (Santa Cruz Biotechnology, sc-8408, 1:1000); rabbit monoclonal anti-phospho-CHK1 S345 (Cell Signaling, 2348, 1:500); rabbit polyclonal anti-phospho-RPA32 (S33) (Bethyl Laboratories, A300-246A, 1:1000); HRP-conjugated goat anti-mouse and anti-rabbit IgGs (Bethyl Laboratories, A90-116P and A120-101P, 1:5000). Signals were acquired using FluorChem HD2 apparatus (Alpha Innotech).

Indirect immunofluorescence analysis was carried out according to standard protocols. For indirect immunofluorescence and FISH, cells grown on coverslips were subjected to pre-extraction in CSK buffer (100 mM NaCl, 300 mM sucrose, 3 mM MgCl₂, 10 mM PIPES pH 6.8, 0.5% Triton-X) for 7 min on ice, fixed with 2% formaldehyde for 10 min at room temperature, permeabilised with 0.5% Triton X-100 for 10 min at room temperature, and re-fixed with methanol for 20 min at −20°C. Fixed cells were incubated in blocking solution (5% BSA, 20 mM glycine, 1× PBS) supplemented with RNaseA (20 µg/ml, Roche) for 30 min at 37°C, followed by incubation with primary and then secondary antibodies diluted in 5% BSA, 1× PBS at room temperature for 1 h each. Cells were again fixed with 4% formaldehyde for 10 min at room temperature and DNA was denatured by incubating coverslips in hybridization mix [10 mM Tris pH 7.2, 70% formamide, 0.5% blocking solution (Roche)] containing telomeric or centromeric PNA probes for 5 min at 80°C on a heating plate. Hybridizations were carried out for 2 h at room temperature and were followed by two washes with wash 1 solution (10 mM Tris pH 7.2, 70% formamide, 0.1% BSA) and one with wash 2 solution (0.1 M Tris pH 7.2, 0.15 M NaCl, 0.08% Tween-20) for 10 min each at room temperature. DNA was counterstained with DAPI (100 ng/ml, Sigma-Aldrich). Antibodies used for indirect immunofluorescence analysis were: rabbit polyclonal anti-cactin (Abnova, PAB23952, diluted 1:1000); rabbit polyclonal anti-53BP1 (Abcam, ab21083, 1:1000); rabbit polyclonal anti-DHX8 (Bethyl Laboratories, A300-624A-M, 1:1000); mouse monoclonal anti-SRRM2 (Santa Cruz Biotechnology, sc-390315, 1:1000); rabbit polyclonal anti-phospho-RPA32 S33 (Bethyl Laboratories, A300-246A, 1:1000). PNA FISH probes were: 5′-Cy3-OO-CCCTAACCCTAACCCTAA-3′ (TelC-Cy3, Panagene); 5′-FAM-AAACTAGACAAGCATT-3′ (Cent-FAM, PNAbio). For metaphase spread analysis, cells were treated with 200 ng/ml colcemid (Sigma-Aldrich) for 2–4 h, harvested by mitotic shake-off and incubated in hypotonic solution (0.075 M KCl) for 9 min at 37°C. Metaphase chromosomes were resuspended in cold methanol:acetic acid solution (3:1) by vortexing and incubated overnight at −20°C. Metaphases were spread on glass slides and counterstained with DAPI (100 ng/ml, Sigma-Aldrich). Images were acquired with an Olympus IX 81 microscope equipped with a Hamamatsu ORCA-ER camera using a 60×1.42NA oil PlanAPoN objective, or with a Deltavision Multiplexed system (Applied Precision) with an Olympus 1X 71 (inverse) microscope, Roper CoolSnap HQ2 camera and a 60×1.4NA oil DIC PlanAPoN objective. Images were analyzed using ImageJ (NIH) and Adobe Photoshop.

Cells were fixed in ice-cold 70% ethanol for 15 min at −20°C, and then incubated with 25 µg/ml RNaseA (Sigma-Aldrich) diluted in 1× PBS for 20 min at 37°C. Cells were stained with 20 µg/ml propidium iodide (Sigma-Aldrich) for 10 min at room temperature, sorted and analyzed with a BD FACSCalibur (BD Biosciences) using FlowJo software. Flow cytometric analysis for DNA synthesis rates and antibody staining was performed according to previously published protocols (Neelsen et al., 2013). 10 µM halogenated nucleotide EdU (Click-IT EdU Flow Cytometry Assay Kit, Invitrogen) was added to cultured cells for 30 min before harvesting. Mouse monoclonal anti-γH2AX (Millipore, 05-636, 1:500), Click-It reaction mix containing Alexa Fluor 488 Azide (Invitrogen) and DAPI (1 µg/ml, Sigma-Aldrich) were used. Stained cells were sorted and analyzed using a Fortessa flow cytometer (BD Bioscience) and FlowJo software.

PFGE analysis was performed according to previously published protocols (Toller et al., 2011). Briefly, cells were harvested and embedded in 1% molten LMT agarose (Bio-Rad). Solidified plugs were incubated at 37°C for 36–72 h in lysis buffer (100 nM EDTA pH 8.0, 0.2% sodium deoxycholate, 1% sodium lauryl sarcosine) containing Proteinase K (1 mg/ml, Appli Chem). Plugs were embedded into a 0.9% pulse-field-certified agarose gel (Bio-Rad) and subjected to pulse-field electrophoresis in a CHEF DR III apparatus (Bio-Rad) at 14°C, stained with ethidium bromide (0.5 µg/ml) and imaged using a Gel Doc 2000 imaging system (Bio-Rad).

Total RNA was isolated using a Nucleo Spin RNA Kit Plus (Macherey-Nagel) or TRIzol reagent (Invitrogen) and treated at least once with DNaseI (Qiagen) according to manufacturer's instructions. For RT-PCR, 1 µg of RNA was reverse-transcribed using Superscript II reverse transcriptase (Invitrogen) and random hexamers followed by PCR amplification using Taq polymerase (New England Biolabs) or real-time PCR amplification using the Light Cycler 498 SYBR Green I master mix (Roche) on a Rotor-Gene Q instrument (Qiagen) using actin mRNA as normalizer. Oligonucleotide pairs are shown in Table S3. For northern blot analysis, 10-15 µg of RNA were denatured at 65°C for 5 min and subjected to electrophoresis in denaturing 1.2% agarose gels. RNA was transferred onto nylon membranes and hybridized using radioactively labeled probes for 16 h at 50–60°C. The sororin probe was a random-primer-labeled PCR amplification product from the pL-Sor plasmid. The 18S rRNA probe was a 5′-end-labeled oligonucleotide. Radioactive signals were detected using a Typhoon FLA 9000 imager (GE Healthcare). For transcriptome analysis, 5 µg of RNA were used for library preparation and high-throughput Illumina sequencing at Fasteris SA (Plan-les-Ouates, Switzerland). Sequencing was performed with a single-end protocol providing reads of 100 bp in length. Reads were trimmed with sickle software (available from https://github.com/najoshi/sickle) and aligned to the human reference genome (https://www.gencodegenes.org; GENCODE GRCh37 v19) using HISAT2 software (v2.0.4) (Kim et al., 2015). DEXSeq (v1.20.0) (Anders et al., 2012) was used to investigate the presence of intron retention. From the GENCODE catalog, we retained exons from known, protein-coding, non-automatically annotated transcripts, and processed the resulting GTF file with the Python scripts provided with DEXSeq in order to generate disjoint exon bins. The ‘flattened’ annotation was extended with intron bins, defined as intragenic regions disjoint from any retained exon. DEXSeq was used to quantify the abundance of each exonic and intronic bin in this extended annotation, and to test each bin for differential usage between the control group and each shRNA-transfected group. Only results for intronic bins were kept for interpretation. A fragments per kilobase per million mapped reads (FPKM) estimate was generated for each intronic and exonic bin by dividing the counts obtained from DEXSeq (after adding 1) by the size of the bin and the total number of read counts for the corresponding sample. Relative intron expression was calculated for each intron bin by dividing its estimated expression with the average expression of the exon bins from the same gene. Alignment files generated by HISAT2 were filtered to retain only reads with mapping quality above 10 and used to visualize the coverage of selected genes with the Gviz R package (v 1.18.0) (Hahne and Ivanek, 2016). Gene ontology enrichment analysis was performed using the PANTHER (protein analysis through evolutionary relationships) classification system (http://www.pantherdb.org).

Samples sizes (n) were derived from experiments with independent cell cultures. Experimental data are presented as the mean±s.e.m. Statistical calculations were performed in Microsoft Excel.

We thank Jan-Michael Peters, Massimo Lopes, Ulrike Kutay (IBC, Zürich, Switzerland) for reagents; Harry Wischnewski, Catherine Brun and Federica Richina for help with experiments; Rajika Arora for critical reading of the manuscript; members of the Azzalin laboratory for discussions.