The ubiquitously expressed nuclear protein NIPP1 (also known as PPP1R8) recruits phosphoproteins for regulated dephosphorylation by the associated protein phosphatase PP1. To bypass the PP1 titration artifacts seen upon NIPP1 overexpression, we have engineered covalently linked fusions of PP1 and NIPP1, and demonstrate their potential to selectively explore the function of the PP1:NIPP1 holoenzyme. By using inducible stable cell lines, we show that PP1–NIPP1 fusions cause replication stress in a manner that requires both PP1 activity and substrate recruitment via the ForkHead Associated domain of NIPP1. More specifically, PP1–NIPP1 expression resulted in the build up of RNA–DNA hybrids (R-loops), enhanced chromatin compaction and a diminished repair of DNA double-strand breaks (DSBs), culminating in the accumulation of DSBs. These effects were associated with a reduced expression of DNA damage signaling and repair proteins. Our data disclose a key role for dephosphorylation of PP1:NIPP1 substrates in setting the threshold for DNA repair, and indicate that activators of this phosphatase hold therapeutic potential as sensitizers for DNA-damaging agents.
Protein serine/threonine phosphatase 1 (PP1) belongs to the PPP-type family of protein phosphatases and catalyzes a considerable fraction of all eukaryotic protein dephosphorylation events (Bollen et al., 2010; Heroes et al., 2013). PP1 exists in a complex with one or two PP1-interacting proteins (PIPs) that function as subcellular targeting subunits, substrate specifiers and/or activity regulators (Bollen et al., 2010). One of the ∼200 known PIPs is NIPP1, for nuclear inhibitor of PP1, also known as PPP1R8, which forms a heterodimer with a substantial proportion of the nuclear pool of PP1 (Jagiello et al., 1995). NIPP1 consists of an N-terminal substrate-binding forkhead-associated (FHA) domain (Boudrez et al., 2000, 2002; Minnebo et al., 2013; Vulsteke et al., 2004), a central PP1-anchoring domain (O'Connell et al., 2012) and a C-terminal PP1-inhibitory domain (Beullens et al., 2000). It functions as a very specific and potent inhibitor of associated PP1 towards all tested substrates (Beullens et al., 1992; Vulsteke et al., 1997; Winkler et al., 2015), but allows the controlled dephosphorylation of phosphoproteins that are recruited via the FHA domain (Tanuma et al., 2008). Substrates only bind to the phosphate-binding loop of the FHA domain when they are phosphorylated on at least one threonine that is followed by a proline (pTP) (Boudrez et al., 2000, 2002; Minnebo et al., 2013; Vulsteke et al., 2004). However, effective dephosphorylation of FHA-binding ligands by PP1:NIPP1, at any serine or threonine residue, also requires the allosteric removal of the C-terminal inhibitory domain of NIPP1, which can, in vitro, be induced by the RNA-dependent phosphorylation of the C-terminus (Beullens et al., 2000). The validated FHA-binding ligands of NIPP1 include the pre-mRNA splicing factors SAP155 (also known as SF3B1) (Boudrez et al., 2002) and CDC5L (Boudrez et al., 2000), histone methyltransferase EZH2 (Minnebo et al., 2013; Roy et al., 2007) and protein kinase MELK (Vulsteke et al., 2004).
NIPP1 plays a role in transcriptional silencing, through recruitment of the EZH2-containing Polycomb repressive complex 2 (PRC2) to a subset of target genes, in a PP1-dependent manner (Minnebo et al., 2013; Van Dessel et al., 2010). However, NIPP1 has also been implicated in pre-mRNA splicing, possibly through the timely dephosphorylation of SAP155 and CDC5L mediated by associated PP1 (Beullens, 2002; Tanuma et al., 2008). Furthermore, NIPP1 regulates the protein kinase MELK, a poorly characterized regulator of cell proliferation (Liu et al., 2017; Vulsteke et al., 2004). A multifunctional role of NIPP1 is also indicated by the diverse phenotypes associated with its deletion or overexpression. The total deletion of NIPP1 in mice is embryonic lethal, at around the gastrulation stage, and this correlates with reduced cell proliferation (Van Eynde et al., 2004). Likewise, the deletion of NIPP1 in the testis results in a complete loss of germ cells, due to reduced proliferation and survival of progenitor germ cells (Ferreira et al., 2017). In contrast, the removal of NIPP1 from liver epithelial cells causes a gradual expansion of periportal progenitor cells (Boens et al., 2016). In HeLa cells, a very mild overexpression of NIPP1 (∼10%) induces transdifferentiation into smooth-muscle like cells in a PP1- and FHA-dependent manner (Van Dessel et al., 2015). However, a multifold overexpression of NIPP1 in HeLa cells causes an arrest in mid-mitosis due to the competitive removal of PP1 from other holoenzymes that are essential for progression through mitosis (Winkler et al., 2015).
To bypass PP1 titration artifacts associated with the overexpression of NIPP1, we have engineered covalently linked fusions of PP1 and NIPP1 to specifically delineate the function of this holoenzyme. Here, we show that stable cell lines that inducibly express a PP1–NIPP1 fusion have no mid-mitotic phenotype but develop replication stress, culminating in the accumulation of DNA double-strand breaks (DSBs). Site-directed mutagenesis revealed that these effects are critically dependent on both substrate recruitment via NIPP1 and the activity of NIPP1-associated PP1, indicating that they stem from a deficient dephosphorylation of (a subset of) substrates. Our data disclose a hitherto unrecognized key function of protein phosphatase PP1:NIPP1 in DNA damage signaling and repair.
The expression of a PP1–NIPP1 fusion causes an interphase arrest
Since the multifold overexpression of NIPP1 competitively disrupts other PP1:PIP complexes (Winkler et al., 2015), this strategy cannot be used to delineate specific functions of the PP1:NIPP1 holoenzyme. To prevent such titration artifacts, we generated EGFP-tagged PP1–NIPP1 fusions joined via a flexible linker (Fig. 1A). In addition to a fusion between wild-type PP1γ (encoded by PPP1CC) and NIPP1 (denoted PP1–NIPP1), we engineered a fusion with a PP1-binding mutant of NIPP1 (PP1–NIPP1-Pm) through mutation of key residues in the PP1-anchoring and PP1-inhibitory domains of NIPP1. More specifically, PP1–NIPP1-Pm was obtained by mutation of residues Val201 and Phe203 in the RVxF-type PP1-anchoring motif and of the flanking basic residues 195–199 into alanine residues, as well as by a phospho-mimicking aspartate mutation of Tyr335 in the C-terminal PP1-inhibitory region (Beullens et al., 2000). EGFP trap experiments (hereafter called EGFR traps) performed in HEK293T cells that transiently expressed EGFP-tagged PP1–NIPP1 did not reveal an interaction of this fusion with endogenous PP1 or any of the four tested PIPs, namely PNUTS (also known as PPP1R10) (Allen et al., 1998), SIPP1 (WBP11) (Llorian et al., 2004), RepoMan (CDCA2) (Trinkle-Mulcahy et al., 2006) and Inhibitor-3/PPP1R11 (Zhang et al., 1998), indicating that the fusion formed a functional holoenzyme that does not competitively disrupt other PP1 holoenzymes (Fig. 1B). In contrast, traps of EGFP-tagged PP1–NIPP1-Pm also contained the examined PIPs, demonstrating that PP1 that cannot interact with the NIPP1 moiety of the fusion is available for interaction with other PIPs. To examine in an independent manner whether PP1–NIPP1 expression causes PP1 titration effects, as previously reported for NIPP1 overexpression (Winkler et al., 2015), we also examined, in the same lysates, how the binding of RepoMan to endogenous PP1 was affected by the ectopic expression of the EGFP fusions (Fig. 1B). The association of endogenous PP1 with immunoprecipitates of RepoMan was not affected by the expression of PP1–NIPP1, but was strongly reduced by expression of EGFP-tagged PP1, NIPP1 or PP1–NIPP1-Pm. The reduced binding of RepoMan to endogenous PP1 in the latter conditions can be explained by competition between endogenous PP1 and either EGFP–PP1 or EGFP–PP1–NIPP1-Pm for binding to RepoMan, and by competition between EGFP–NIPP1 and RepoMan for binding to the limited pool of endogenous PP1, as previously reported (Winkler et al., 2015). Finally, we found that the transient expression of PP1–NIPP1 was associated with hypophosphorylation of its putative substrates, namely SAP155, CDC5L, MELK and EZH2 (Fig. S1A), indicating that the fusion is a functional PP1 holoenzyme. Taken together, these data validate PP1–NIPP1 fusions as useful tools to explore functions of the PP1:NIPP1 holoenzyme.
By using the Flip-In T-REx system, we subsequently generated HeLa cell lines that inducibly and stably expressed EGFP-tagged PP1–NIPP1 fusions (Fig. 1C). In addition to the fusion between wild-type PP1 and NIPP1 (PP1–NIPP1), we made a cell line that expressed a fusion with hypo-active PP1 (PP1m–NIPP1) by mutation of an essential metal-coordinating residue in the active site into alanine (D64A). Mutation of this residue decreases the Vmax of PP1 but does not affect its Km (Zhang et al., 1996). Furthermore, we generated a cell line that expressed a substrate-binding mutant of the fusion (PP1–NIPP1m) by mutation of the phosphate-binding loop in the substrate-binding FHA domain (S68A/R69A/V70A/H71A) (Boudrez et al., 2000). The cell lines expressed the fusions in a doxycycline (Dox)-dependent manner and to a similar extent (Fig. 1D).
To examine the properties of the PP1–NIPP1 fusions, we first assessed the PP1 activity in EGFP traps using glycogen phosphorylase a as a substrate. The PP1–NIPP1 and PP1–NIPP1m fusions were inactive, but a phosphatase activity was revealed upon trypsinolysis (Fig. 1E), which destroys NIPP1 and releases the fully active catalytic core of PP1 (Beullens et al., 1998). This demonstrates that the PP1 moiety in the fusions was catalytically active and inhibited by the associated NIPP1. In contrast, a pretreatment of PP1m–NIPP1 with trypsin only generated ∼20% of the phosphorylase phosphatase activity that was associated with the other fusions, confirming that the D64A mutation strongly decreased the catalytic efficiency of the PP1 moiety. The activity that was still associated with PP1m–NIPP1 stemmed from PP1, as shown by its nearly complete inhibition by addition of Inhibitor-2/PPP1R2 (1±1% of control activity; n=3). As expected, EGFP traps of PP1–NIPP1 and PP1m–NIPP1 co-precipitated with SAP155, an established FHA ligand and substrate (Boudrez et al., 2002; Tanuma et al., 2008), but this interaction was not seen when the phosphate-binding loop of the FHA-domain was mutated, as in PP1–NIPP1m (Fig. 1F). Interestingly, PP1m–NIPP1 reproducibly bound much more SAP155 than did PP1–NIPP1, which can be accounted for by the reduced substrate-turnover rate of the hypo-active PP1m–NIPP1 fusion. NIPP1 is a nuclear protein and is enriched in the nuclear speckles though association with FHA ligands (Jagiello et al., 2000). Accordingly, we found that PP1–NIPP1 and PP1m–NIPP1 were nuclear and, after pre-extraction with Triton X-100, the remaining NIPP1 was associated with speckles (Fig. 1G). As expected PP1–NIPP1m was not associated with nuclear speckles. Collectively, these data show that the catalytic activity of PP1 and regulatory properties of NIPP1 are retained in EGFP-tagged PP1–NIPP1 fusions. Our results furthermore validate PP1m–NIPP1 and PP1–NIPP1m as hypo-active and substrate-binding mutants, respectively.
As a first approach to examine the effects of the expression of PP1–NIPP1 fusions, we used time-lapse imaging to measure the duration to mitotic entry after a release from a thymidine-induced G1/S-phase arrest, as assessed from the rounding up of the cells (Fig. 1H). For these experiments, we only analyzed cells that showed a bright EGFP signal. About 90% of the EGFP-expressing control cells (Ctr) or the non-induced cells entered mitosis within 20 h, as compared to only ∼5% of the PP1–NIPP1-expressing cells, hinting at a delayed mitotic entry phenotype (Fig. 1H; Fig. S1B, Movies 1–4). Strikingly, PP1–NIPP1-expressing cells often transiently rounded up but then flattened again and did not go through mitosis within the 60 h of filming. This phenotype was strongly reduced or absent in cells that expressed the mutated PP1–NIPP1 fusions. The data were identical for two independent clones of each stable cell line (not shown). A delayed mitotic entry was also observed in asynchronous cells expressing PP1–NIPP1 (Fig. S1C, Movies 5,6), showing that this phenotype is not an artifact caused by the pre-treatment with thymidine. Interestingly, the expression of PP1m–NIPP1 in asynchronous cells was often associated with a prolonged duration of mitosis (Fig. S1D), which was rarely seen after expression of the other fusions. This may be due to sequestration of FHA ligands by this fusion (Fig. 1F), in particular the splicing factors CDC5L and SAP155, which are essential for progression through mitosis (Mu et al., 2014; Paulsen et al., 2009; Sundaramoorthy et al., 2014).
Consistent with a delayed mitotic entry phenotype, MTT assays showed that cells expressing PP1–NIPP1 had a reduced proliferation, and this phenotype was also less strong with the mutant fusions (Fig. S1E,F). Taken together, these data show that the expression of PP1–NIPP1 induces an interphase arrest, which is less pronounced with PP1m–NIPP1 and nearly absent with PP1–NIPP1m, indicating that it requires an active PP1 moiety and a functional substrate recruitment domain.
The PP1–NIPP1 induced interphase arrest is associated with deregulated gene expression
Since PP1:NIPP1 has established functions in transcription and pre-mRNA splicing (see Introduction), we subsequently investigated whether the expression of the PP1–NIPP1 fusion affects the cellular transcriptome in a way that could explain the observed interphase arrest. We performed total RNA sequencing (RNA Seq) of the stable cell lines in the absence or presence of Dox and after a 1 h release from a double thymidine arrest. Induction of the fusions with Dox did not alter the proportion of cells in each cell cycle stage, as indicated by nearly identical levels of cyclin E1 and cyclin A2 transcripts (Fig. S2A). We noted striking differences when PP1–NIPP1 was expressed (Fig. 2A). These differences were not seen after expression of the mutant transgenes, indicating that PP1–NIPP1 affects the cellular transcriptome in a manner that depends on the PP1 activity and a functional FHA domain. Data analysis [false discovery rate (FDR) <0.01 and a cut-off of 2-fold change] identified 152 upregulated and 293 downregulated genes following the induction of PP1–NIPP1 (Fig. 2B; Fig. S2B,C, Table S1). An Ingenuity Pathway analysis (IPA) (FDR<0.01 and a cut-off of 1.5-fold change) showed that the Gene Ontology (GO) terms ‘cell cycle’, ‘DNA replication, recombination and repair’ and ‘RNA post-transcriptional modification’ were among the most affected molecular and cellular functions (Fig. 2C). Most genes associated with ‘cell cycle’ (76.9%) and ‘DNA replication, recombination and repair’ (76.2%) were downregulated, whereas genes associated with ‘RNA post-transcriptional modification’ (63.5%) were mostly upregulated. We validated the RNA sequencing data for five upregulated and five downregulated genes by quantitative real-time PCR (qRT-PCR) (Fig. 2D). Thus, PP1–NIPP1 expression perturbs cell cycle progression and this is linked with an altered expression of proteins implicated in the processing of nucleic acids.
PP1–NIPP1 reduces replication-fork progression
The interphase arrest induced by PP1–NIPP1 (Fig. 1H) was associated with a reduced expression of key regulators of DNA replication and repair (Fig. 2), hinting at a problem with progression through S-phase. To test this hypothesis, PP1–NIPP1-expressing cells were released from a double thymidine arrest (DTA, 0 h), pulsed with BrdU for 1 h, and subsequently left to cycle for up to 7 h (i.e. 2, 4 and 8 h including BrdU labeling) in a medium that contained nocodazole, but lacked thymidine and BrdU (Fig. 3A). Nocodazole was added to prevent the spreading of the BrdU signal to the next generation. The harvested cells were analyzed by FACS for BrdU (Fig. 3B). Based on the condition without BrdU treatment, the gates were set to quantify early (1), mid (2) and late S-phase (3), as well as the S-G2 transition (4) (Fig. S3A). In the control cultures (no Dox), BrdU-positive cells progressed through S-phase and, after 8 h, most cells were at the S-G2 transition or in G2/M. Consistent with our hypothesis, PP1–NIPP1-expressing cells showed an accumulation in early, mid and late S-phase, in particular at the later time-points (Fig. 3B–E). An accumulation of cells in S-phase was confirmed in asynchronous cells following the induction of PP1–NIPP1 (Fig. S3B–D).
To compare the kinetics of replication before and after expression of PP1–NIPP1, we monitored the progression of individual replication forks after the successive labeling of newly synthesized DNA with the halogenated nucleotide analogs IdU and CldU (Fig. 3F). Expression of PP1–NIPP1 resulted in a lower average length of CldU-labeled fibers (Fig. 3G,H), a higher frequency of slow-progressing forks (Fig. 3I), and the accumulation of stalled forks (Fig. 3J,K), which are all hallmarks of replication stress (Gaillard et al., 2015).
PP1–NIPP1 reduces the level of DNA-damage signaling and repair proteins
We proceeded to explore the causes of replication stress induced by the expression of PP1–NIPP1. Replication stress can be elicited by a shortage of proteins required for DNA replication, chromatin compaction and DNA repair (Gaillard et al., 2015; Zeman and Cimprich, 2014). An IPA analysis (FDR<0.01 and a cut-off of 1.5-fold change) of the RNA Seq data revealed that the GO terms ‘Role of BRCA1 in DNA damage response’, ‘Role of CHK proteins in cell cycle checkpoint control’ and ‘Hereditary breast cancer signaling’ were among the most affected pathways following PP1–NIPP1 expression (Fig. 4A). 83% of the genes that were affected in these pathways were downregulated by PP1–NIPP1 (Table S2), as is illustrated for 12 genes in Fig. 4B,C. The altered expression of these genes was independently confirmed by qRT-PCR (Fig. 4D). These data demonstrate that PP1–NIPP1 mainly decreased the expression of genes implicated in DNA processing. The downregulated genes encode proteins involved in DNA replication (e.g. RFC4, POLN and PRIMPOL), chromatin remodeling (e.g. SMARCD2 and RBBP4), DNA damage signaling (e.g. E2F1 and RBL2), DNA repair (e.g. BMI1, FANCD2, FANCL and BRIP1) and the DNA damage response (DDR) (e.g. ATM and BRCA1).
To examine whether PP1–NIPP1 also affects DNA-processing proteins at the protein level, we focused on the DDR. More specifically, we investigated how PP1–NIPP1 expression affects the protein level of key sensors (RPA2 and MRE11), transducers (ATM, ATR and DNA-PK) and effectors (CHK1, CHK2 and BRCA1) of the DDR (Fig. 4E; Fig. S4A–D). The sensors RPA2 and MRE11 (Fig. S4A,B), the transducer ATM, and the effectors CHK1 (Fig. 4F) and BRCA1 (Fig. S4C,D) were significantly downregulated following PP1–NIPP1 expression, hinting at a globally reduced DDR. These effects were not detected following expression of PP1m–NIPP1 or PP1–NIPP1 m (Fig. 4F, Fig. S4), indicating that they require NIPP1-associated PP1 and a functional FHA domain.
PP1–NIPP1 induces R-loops and chromatin compaction
The loss of BRCA1, FANCL and FANCD2 induced by the expression of PP1–NIPP1 (Fig. 4B–D; Fig. S4C,D) is expected to result in the accumulation of R-loops (Bhatia et al., 2014; García-Rubio et al., 2015; Hatchi et al., 2015; Schwab et al., 2015). R-loops are hybrids of DNA and non-spliced RNA, and form obstacles for the replication machinery, causing replication fork stalling and collapse, and culminating in single-strand or double-strand DNA breaks (Aguilera and García-Muse, 2012; Costantino and Koshland, 2015; Skourti-Stathaki and Proudfoot, 2014). To examine whether PP1–NIPP1 causes the accumulation of nucleoplasmic R-loops, we performed immunostainings with the RNA–DNA hybrid-recognizing monoclonal antibody S9.6 (Boguslawski et al., 1986). Following fixation and permeabilization with Triton X-100, a nuclear staining was mainly detected in nucleoli and diffusely distributed nuclear foci (Fig. 5A), in accordance with published data (Aguilera and García-Muse, 2012; Bhatia et al., 2014). However, we also noted a cytoplasmic staining, which probably reflects mitochondrial DNA replication (Brown et al., 2008). The nucleolar staining was not R-loop-specific as it was lost after a pretreatment with RNase A, which destroys single-stranded RNA (ssRNA). Strikingly, the non-nucleolar nuclear staining was even increased by a pretreatment with RNase A, which can be explained by a reduced competition of ssRNA for binding to the antibody (Zhang et al., 2015). In contrast, the non-nucleolar staining was lost by a preincubation with the RNA–DNA-hybrid specific RNase H, indicating that they represent real R-loops. To quantify the nucleoplasmic R-loops, we subtracted the cytoplasmic and nucleolar signals, revealing a doubling in nucleoplasmic R-loop staining following the expression of PP1–NIPP1 (Fig. 5B,C).
Since R-loops are known to contribute to heterochromatin formation (Castellano-Pozo et al., 2013; Groh et al., 2014; Skourti-Stathaki et al., 2014), and PP1–NIPP1 affects the expression of key chromatin remodeling factors (Fig. 4C), we also examined whether PP1–NIPP1 affects heterochromatinization. We indeed found that PP1–NIPP1 expression resulted nearly in a doubling in staining intensity for the facultative heterochromatin marker H3K27me3 (Fig. 5D,E) and the constitutive heterochomatin marker H3K9me3 (Fig. 5F,G), hinting at increased chromatin compaction.
PP1–NIPP1 hampers the repair of DNA DSBs
The observed effects of PP1–NIPP1 expression on chromatin structure and replication fork dynamics (Figs 3 and 5) are typically associated with the accumulation of DSBs. Indeed, stalled replication forks are often converted into DSBs by endonuclease-mediated DNA cleavage (Zeman and Cimprich, 2014). R-loops can be processed into DSBs by the nucleotide-excision repair endonucleases XPF and XPG (Sollier et al., 2014). In addition, heterochromatin represents a barrier for the repair of DSBs (Lemaître and Soutoglou, 2014). Finally, the actual repair of DSBs is expected to be hampered due to the decreased level of DSB-repair proteins, such as ATM (Fig. 4E,F) or BRCA1 (Fig. S4C,D). Consistent with these findings, we noted that the expression of PP1–NIPP1, but not the mutant fusions, resulted in a massive increase in the number of γH2Ax-containing cells (Fig. 6A–C, Fig. S5B) as well as in an augmented number of γH2Ax (Fig. 6D) and 53BP1 foci per cell (Fig. S5A), which serve as markers of DSBs. This response was strongly reduced or absent in cells that expressed mutated PP1–NIPP1 fusions. PP1–NIPP1 did not colocalize with γH2Ax (Fig. 6A), suggesting that it is not targeted to DNA lesions and causes DSBs through an indirect mechanism. Strikingly, the expression of PP1m–NIPP1 resulted in an increased size of nuclear speckles (SC35, also known as SRSF2, staining), which represent storage sites for splicing factors. We speculate that this increased speckle size results from the sequestration of hyperphosphorylated splicing factors SAP155 and CDC5L by PP1m–NIPP1 (Fig. 1F). Hyperphosphorylation of splicing factors has indeed been associated with enlarged nuclear speckles (Lamond and Spector, 2003; Misteli and Spector, 1996).
To examine whether PP1–NIPP1 also affects the DNA repair capacity, we made use of previously described stable U2OS cell lines with an integrated EGFP reporter that contains a consensus site for the I-SceI endonuclease (Gunn and Stark, 2012). The reporter gene in these cell lines only generates EGFP protein when the I-SceI-induced DSB is repaired by non-homologous end joining (NHEJ) repair (Fig. 7A), homologous recombination (HR) repair (Fig. 7B), or single-strand annealing (SSA) repair (Fig. 7C). We used FACS analysis to quantify the accumulation of EGFP following the co-transfection of the endonuclease HA-tagged I-SceI and either a FLAG-tagged PP1–NIPP1 fusion or the empty vector (Fig. 7A–C). All three major pathways for the repair of DSBs were severely compromised by the expression of PP1–NIPP1 in this semi-quantitative assay, but significantly less so after the expression of the mutated fusions. We confirmed that I-SceI and the PP1–NIPP1 fusions were expressed in these cell lines (Fig. 7D).
DSBs cause a cell cycle arrest and, when the breaks are massive or irreparable, this results in a permanent cell cycle arrest (senescence) or apoptosis (Ciccia and Elledge, 2010). The senescence markers p16 and p27 were downregulated after the induction of PP1–NIPP1 and not affected by the mutant fusions (Fig. S6A,B), arguing against a PP1–NIPP1 mediated development of a senescence phenotype. However, we noted an accumulation of caspase-3 (Fig. S6C,D) and an increased amount of cleaved PARP (Fig. S6A,E) when PP1–NIPP1 or PP1m–NIPP1 were expressed, hinting at apoptosis. Thus, at least in some cells, the expression of PP1–NIPP1 causes an accumulation of DNA damage to an extent that it induces apoptosis (Movies 2 and 6). We speculate that the induction of apoptosis by PP1m–NIPP1 is related to titration of FHA ligands (Fig. 1F), resulting in a prolonged mitotic arrest (Fig. S1D) and sometimes culminating in cell death.
NIPP1 is a nuclear PIP that is associated with a major fraction of chromatin-associated PP1. We have previously demonstrated that a multifold overexpression of NIPP1 causes a mid-mitotic-arrest (Winkler et al., 2015). This phenotype could be explained by the competitive disruption of PP1 holoenzymes that are essential for progression through mitosis. Here, we show that the expression of a PP1–NIPP1 fusion does not cause such PP1 titration artifacts and can be used to delineate the biological functions of this holoenzyme. Indeed, we have demonstrated that a PP1–NIPP1 fusion forms a functional and properly localized holoenzyme that does not bind endogenous PP1 or PIPs (Fig. 1; Fig. S1). It seems likely that the same strategy, that is, the expression of PP1–PIP fusions, can also be adopted to selectively study the function of other PP1 holoenzymes. Moreover, such fusions can be mutated in specific regulatory or catalytic sites, as we have demonstrated for PP1–NIPP1, enabling an exploration of the contribution of individual subunits or their ligands to the observed phenotype(s).
The expression of PP1–NIPP1 caused replication stress, as indicated by slower replication fork progression and the accumulation of stalled replication forks (Fig. 3; Fig. S3) and DSBs (Fig. 6, Fig. S5). Moreover, the capacity to repair DSBs was reduced (Fig. 7A–D). We have studied the effects of NIPP1 complexed to PP1γ, which also exists in vivo (Comerford et al., 2006; Fardilha et al., 2004; Trinkle-Mulcahy et al., 2001). However, in some cell types NIPP1 preferentially binds to PP1β (encoded by PPP1CB) (Jagiello et al., 1995; Verheyen et al., 2015). Although PP1 isoforms are at least partially functionally redundant, it cannot be excluded that the phenotypes of expression of either PP1β–NIPP1 or PP1γ–NIPP1 are slightly different.
The severe phenotype associated with the expression of PP1–NIPP1 can, at least partially, be explained by a lower level of key DNA replication and repair factors (Fig. 4; Fig. S4). In addition, the accumulation of R-loops and enhanced chromatin compaction after the expression of PP1–NIPP1, which are possibly also a consequence of the deregulated expression of RNA- or DNA-processing proteins, are likely to have contributed to the severity of the observed phenotype. Importantly, the effects of PP1–NIPP1 expression were largely lost with a substrate-binding mutant (PP1–NIPP1m) and were less pronounced with a phosphatase-activity mutant (PP1m–NIPP1) (Fig. 7E). It should be pointed out that PP1m–NIPP1 is still partially active (20% of the wild-type fusion; Fig. 1E), explaining why this mutant fusion still caused a mild phenotype. In addition, some of the effects caused by induction of PP1m–NIPP1, such as enlarged nuclear speckles (Fig. 6A) and mitotic arrest (Fig. S1D), could be due to remaining phosphatase activity (Fig. 1E) and/or its ability to trap and sequester substrates (Fig. 1F). Collectively, our data strongly suggest that the phenotype observed upon PP1–NIPP1 expression was caused by a precocious dephosphorylation of (a subset of) PP1–NIPP1 substrates (Fig. 7E). Nevertheless, NIPP1 was not recruited to laser-induced sites of DNA damage (Landsverk et al., 2010) and did not colocalize with γH2Ax foci (inset in Fig. 6A), indicating that it has an indirect role in DNA repair.
The substrates of PP1–NIPP1 that are implicated in the development of replication stress remain to be identified. They could include both established and hitherto unknown substrates. Strikingly, all known substrates of PP1–NIPP1 can be (in)directly linked to DNA damage and repair signaling. Thus, an enhanced dephosphorylation of the pre-mRNA splicing factors SAP155 and CDC5L by PP1–NIPP1 expression may contribute to deficient pre-mRNA splicing, resulting in the accumulation of R-loops (Fig. 5). However, SAP155 is also a key component of a BRCA1-associated complex that promotes the splicing and nuclear export of transcripts encoding DDR proteins (Savage et al., 2014; Vohhodina et al., 2017). Likewise, CDC5L has been identified as a direct interactor and activator of the DNA-damage checkpoint kinase ATR (Zhang et al., 2009). The chromatin targeting and/or activation of the histone methyltransferase EZH2 may (partially) account for facultative chromatin condensation after expression of PP1–NIPP1. However, EZH2 has recently also been shown to recruit the endonuclease MUS81 to stalled replication forks (Rondinelli et al., 2017). Moreover, EZH2 is a substrate for protein kinase MELK (Liu et al., 2017), another NIPP1 FHA ligand.
In conclusion, we have demonstrated here that PP1–NIPP1 reduces the expression of DNA damage signaling and repair proteins associated with replication stress and reduced the capacity for DNA repair (Fig. 7E). PP1–NIPP1 thereby joins a growing list of PP1 holoenzymes that downregulate or reverse the DDR (reviewed by Shimada and Nakanishi, 2013). These include PP1–RepoMan, which inactivates protein kinase ATM (Peng and Maller, 2010), PP1–PNUTS, which antagonizes the DNA-activated protein kinase (Zhu et al., 2017), PP1–BRCA1, which contributes to the recovery from the repair of DSBs (Hsu, 2007; Winter et al., 2007), and PP1–Rif1, which destabilizes stalled replication forks (Alver et al., 2017). Small-molecule activators of these PP1 holoenzymes have therapeutic potential because cancer cells have often one or more DNA damage repair pathways inactivated, making them more vulnerable to inhibitors of remaining, compensatory, mechanisms (Zhang et al., 2016). We envisage that activators of PP1:NIPP1 could be designed by targeting the PP1-inhibitory domain of NIPP1, and could be used for cancer therapy as such or in combination with clinically used DNA-damaging drugs or radiotherapy.
MATERIALS AND METHODS
Anti-EGFP (SC-8334, 1:500) and anti-BRCA1 (SC-642, 1:1000) antibodies were purchased from Santa Cruz Biotechnology. Anti-RepoMan (HPA030049, 1:1000) and anti-FLAG (F1804, 1:5000) antibodies, RNase A, camptothecin (CPT), MTT [3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide], nocodazole and doxycycline were obtained from Sigma-Aldrich. Anti-GAPDH (2118, 1:5000), anti-phospho-threonine-proline (pTP) (9391, 1:1000), anti-ATR (2790, 1:1000), anti-CHK2 (2662, 1:500), anti-PARP (9542, 1:1000), anti-p27 (3698, 1:1000) and anti-cleaved caspase 3 [9664, 1:200 for immunofluorescence (IF)] antibodies were purchased from Cell Signaling Technology, and anti-53BP1 (A300-272A, 1:1000 for IF) antibody was from Bethyl Laboratories. RNase H was purchased from Applied Biosystems. The anti-NIPP1 (1:5000), anti-Inhibitor-3 (1:500), anti-PNUTS (1:1000), anti-SIPP1 (1:1000) and anti-PP1 (1:2500) antibodies were made in-house, as previously described (Boudrez et al., 2000; Lesage et al., 2004; Llorian et al., 2005; Van Dessel et al., 2010). Anti-SAP155 (D221-3, 1:5000), anti-SC35 (556363, 1:200 for IF), anti-HA (MMS-101P, 1:500 for IF), anti-DNA-PK (MA5-13404, 1:5000), anti-CHK1 (A300-161A, 1:500), anti-p16 (10883-1-AP, 1:1000) and anti-DNA:RNA hybrid antibodies (S9.6, ENH001, 1:200) were purchased from MBL International, BD Pharmingen, Covance, Thermo Scientific, Bethyl Laboratories, Proteintech and Kerafast, respectively. The anti-nucleolin (ab22758, 1:1000 for IF), anti-BrdU (BrdU, pulse-chase experiment in Fig. 3 ab6326, 1:200), anti-RPA (ab16855, 1:1000), anti-ATM (ab17995, 1:1000) and anti-MRE11 (ab214, 1:1000) antibodies were from Abcam. Anti-γH2Ax (05-636, MABE205, 1:200 for IF, 1:300 for FACS), anti-H3K27me3 (07-449, 1:500 for IF) and anti-H3K9me3 (07-442, 1:500 for IF) antibodies were obtained from Upstate. Secondary horseradish peroxidase (HRP)-conjugated antibodies were purchased from Agilent-Dako and secondary Alexa-Fluor-488-, 555- and 633-conjugated antibodies were from Invitrogen. Thymidine was purchased from Acros Organics and microcystin-LR from Cayman Chemicals. The signal enhancer HIKARI kit was purchased from Nacalai Tesque.
HEK293T/17 cells were grown in high-glucose DMEM (Sigma-Aldrich), supplemented with 10% fetal bovine serum, 100 U/ml penicillin and 100 μg/ml streptomycin (Sigma-Aldrich). U2OS cells were cultured identically except for the use of low-glucose DMEM (Sigma-Aldrich). HeLa Flip-In T-REx cells that stably expressed EGFP or EGFP-PP1–NIPP1 fusions were generated as previously reported (Van Dessel et al., 2015). These cell lines were regularly tested for mycoplasma contamination. For the creation of covalently linked PP1–NIPP1 fusions, a flexible linker with the sequence LSGGGGSGGGGSGGGGSGGGGSAAA (Arai et al., 2001) was attached to the C-terminus of EGFP-tagged rat PP1γ, using the adapter-duplex cloning strategy. Subsequently, bovine NIPP1 was added and EGFP–PP1γ–NIPP1 was subcloned into the pcDNA5/FRT/TO plasmid. Single-cell clones were picked and expanded to obtain monoclonal cell lines. The indicated mutants were created using the QuikChange mutagenesis kit (Agilent Technologies) and verified by sequencing. The culture conditions were as previously described (Van Dessel et al., 2015; Winkler et al., 2015). To induce the expression of the stable cell lines, doxycycline (Dox, 10 ng/ml) was added to the culture medium for 48 h, unless stated otherwise. Transfections were performed with jetPRIME (Polyplus, NY) and X-tremeGene9 (Roche) transfection reagent for up to 72 h.
For a double thymidine arrest, cells were consecutively cultured for 18 h with 2 mM thymidine, 9 h without thymidine, and 16 h with 2 mM thymidine. Unless indicated otherwise, the arrested cells were harvested 4 h after the double thymidine block. Cell proliferation in HeLa Flip-In T-REx cells was measured with the MTT assay, which actually quantifies metabolic activity. Cells were seeded in duplicate in 48-well plates and cultured in the absence or presence of Dox for 24, 48 or 72 h. MTT (0.33 mg/ml) was added 1 h before measurement of the absorbance at 550 nm. Cell proliferation was expressed as a percentage of the metabolic activity at 24 h.
Immunofluorescence studies and time-lapse imaging were carried out as previously described (Winkler et al., 2015). For time-lapse imaging under asynchronous conditions, cells were induced the night before imaging was started. Images and movies were processed with the ImageJ software (National Institutes of Health) using the ‘Max intensity’ or ‘sum slices’ feature of z-project. Confocal images were acquired with identical illumination settings. Contrast and brightness were adjusted using linear operations applied to the entire image.
The RNase A and RNase H treatment was performed before fixation of the cells with 4% formaldehyde (Fig. 5A). Following pre-extraction of the cells with 0.3% Triton X-100, they were consecutively treated for 10 min with DNase-free RNase A in PBS or for 20 min with 3.3 U RNase H in a buffer containing 20 mM Hepes-KOH at pH 8, 50 mM KCl, 4 mM MgCl2, 1 mM DTT and 50 μg/ml BSA.
R-loops (Fig. 5C) were quantified as previously reported (Sollier et al., 2014). Briefly, the DAPI signal was used to create a mask of the nucleus, and the nucleolin signal was used to create a mask of the nucleoli. Subsequently, the total corrected nuclear and nucleolar R-loop fluorescence was calculated as previously described (McCloy et al., 2014). The nucleolar R-loop fluorescence was subtracted from the nuclear R-loop fluorescence to obtain the nucleoplasmic R-loop signal. For the quantification of H3K9me3 and H3K27me3 levels (Fig. 5E–G), the total corrected fluorescence for H3K9me3, H3K27me3 and EGFP was calculated as described previously (McCloy et al., 2014). The EGFP signal reflects PP1–NIPP1 expression. For the Dox condition, only the H3K9me3 and H3K27me3 signals of cells with an EGFP total corrected fluorescence value above 10,000 (high PP1–NIPP1 expression) were taken into account.
Single-cell DNA fiber analysis
Single-cell DNA fiber analysis was performed as described previously (Schwab and Niedzwiedz, 2011). Briefly, asynchronous PP1–NIPP1 stable cell lines were either not induced or induced for 72 h with Dox. Cells were then sequentially labeled with 25 µM IdU and 250 µM CldU for 30 min each. After trypsinization, the cells were spotted onto microscope slides and lysed with a buffer containing 200 mM Tris-HCl pH 7.5, 50 mM EDTA and 0.5% SDS. The slides were tilted (15°) to allow lysates to move down the slide. The DNA spreads were air-dried and fixed with 3:1 methanol/acetic acid. After treatment with 2.5 M HCl, slides were immunostained with anti-BrdU antibody BD-347580 (BD Biosciences, 1:200) and ab6326 (Abcam, 1:200) to detect IdU and CldU, respectively. The frequency of stalled forks and fork progression rate were calculated as previously described (Nieminuszczy et al., 2016).
DNA reporter assays and FACS analysis
U2OS cells with integrated reporters for DSB repair were as previously described (Gunn and Stark, 2012). The reporter cells were transiently co-transfected with FLAG empty vector (EV) or FLAG–PP1–NIPP1 fusions and I-SceI plasmid. For the generation of FLAG-tagged PP1–NIPP1 fusions, cDNA was cloned via XhoI and SacII into an EGFP backbone in which the EGFP tag had been replaced by a triple FLAG sequence. At 72 h after transfection, cells were fixed for 10 min with 4% formaldehyde and EGFP-expressing cells were quantified with the Attune cytometer, using the Attune FACS software.
For the BrdU-propidium iodide (PI) pulse-chase FACS experiments (Fig. 3A–E), stable PP1–NIPP1 cells were synchronized with a double thymidine arrest, and the expression of PP1–NIPP1 was induced or not with Dox together with the addition of thymidine. Following release from the double thymidine arrest cells were pulse-labeled with 10 μM BrdU (Sigma) for 1 h and then left to cycle for an additional 1, 3 or 7 h in thymidine- and BrdU-free medium in the presence of 50 μg/ml nocodazole. For the asynchronous BrdU-FACS experiment (Fig. S3B–D) control and PP1–NIPP1 stable cells were induced or not induced with Dox for 48 h and pulsed for 1 h with 10 μM BrdU before harvesting. Cells were trypsinized, permeabilized with 70% ethanol and stored at −20°C. For the BrdU staining, cells were incubated in 2 M HCl plus 0.5% Triton X-100 for 30 min and subsequently neutralized with 0.1 M sodium tetraborate for 2 min. The samples were then washed once with PBS containing 1% BSA and incubated for 1 h with anti-BrdU antibodies. After two subsequent washing steps with PBS containing 1% BSA for 1 h with Alexa Fluor 488-conjugated antibodies, the nuclei were stained with 20 μg/ml PI (Invitrogen) in PBS containing 10 μg/ml RNase A and analyzed with the CANTO II HTS cytometer. FlowJo® software was used for the data analysis.
Flow cytometry based γH2Ax quantification was performed using a method adapted from Marti et al. (2006) with slight modifications (Marti et al., 2006). Briefly, fixed cells were washed twice with PBS and incubated overnight with anti-γH2Ax antibodies. Following incubation, cells were washed twice with PBS and incubated with Alexa-Fluor-405-conjugated goat anti-mouse-IgG antibodies for 1 h at room temperature. After two washes with PBS, cells were resuspended in RNase A/propidium iodide (PI) solution. As a positive control of γH2Ax induction, non-induced control stable cells (EGFP-expressing stable cell line) were treated with 10 µM camptothecin (CPT) for 1 h before harvesting (Fig. S5B). Samples were acquired with the flow cytometer BD CANTOII-HTS using DiVa software and then analyzed with FlowJo software. γH2Ax-positive cell populations were quantified with appropriated gating on the population of interest, as described in the figure legend.
HEK293T/17, Hela Flp-In T-REx and U2OS cells were lysed with 0.3-M salt lysis buffer (50 mM Tris-HCl pH 7.4, 0.5% Triton and 300 mM NaCl) or RIPA lysis buffer (50 mM Tris-HCl pH 8, 150 mM NaCl, 1 mM EDTA, 0.5% deoxycholate, 1% Triton-X 100 and 0.1% SDS), supplemented with protease inhibitors (0.5 mM phenylmethanesulfonyl fluoride, 0.5 mM benzamidine, 5 µM leupeptin) and phosphatase inhibitors (25 mM NaF and 1 mM orthovanadate). For Fig. S1A, the lysates were additionally supplemented with 0.5 μM microcystin-LR. After 20 min incubation on ice, samples were centrifuged for 10 min at 1000 g (0.3 M salt lysis) or 3000×g (RIPA lysis). EGFP trapping and immunoprecipitation of endogenous RepoMan was performed as previously described (Winkler et al., 2015). EGFP traps were assayed for glycogen phosphorylase phosphatase activity before and after trypsinolysis. The EGFP trap for PP1m–NIPP1 was also assayed in the presence of 0.5 µM recombinant Inhibitor-2 (Beullens et al., 1992).
SDS-PAGE gel electrophoresis was performed with 10% or 4–12% Bis-Tris or 3-8% Tris-acetate gels (NuPAGE, Invitrogen). Immunoblots were visualized using ECL reagent (Perkin Elmer) or a signal enhancer HIKARI kit in an ImageQuant LAS4000 imaging system (GE Healthcare). The signals were quantified using the Image QuantTL software. Uncropped images of immunoblots are shown in Fig. S7.
RNA sequencing and qRT-PCR
Total RNA was isolated from cells using the GenElute™ Mammalian Total RNA Miniprep kit from Sigma. For RNA sequencing, the indicated stable cells were synchronized with a double thymidine arrest and released for 1 h. The expression of the transgenes was either not induced or induced with Dox for 13 h before the cells were harvested for total RNA isolation and subjected to RNA sequencing analysis. Library preparation, sequencing and statistical analysis of the RNA sequencing data were performed by the VIB Nucleomics Core (www.nucleomics.be) KU Leuven, as described previously (Ferreira et al., 2017).
Complementary DNA (cDNA) was synthesized from 2 µg of total RNA using RevertAid Premium Reverse Transcriptase and RiboLock RNase inhibitor enzymes (Fermentas, GmBH, St Leon-Rot, Germany) and oligo dT primers (Sigma). 1.2% of the cDNA was PCR-amplified in duplicate using SYBR Green qPCR Mix from Invitrogen (Paisley, UK) and a Rotorgene detection system from Corbett Research (Cambridge, UK). All values were normalized to the housekeeping gene HPRT1. qRT-PCR primers are listed in Table S3.
The statistical significance between experimental and control groups was calculated with the GraphPad Prism software using two-tailed unpaired or paired Student's t-test or Mann–Whitney test, as indicated. If unequal variances were assumed, an unpaired t-test with Welch's correction was applied. *P<0.05; **P<0.01; ***P<0.001.
We thank Fabienne Withof, Annemie Hoogmartens and Nicole Sente for technical assistance. The U2OS DNA-reporter cells were provided by Dr Anna Sablina (VIB, Leuven, BE), with permission from Prof. Dr Jeremy Stark (City of Hope, CA, USA). RNA sequencing library preparation, and statistical analysis was performed by the VIB Nucleomics Core, KU Leuven (www.nucleomics.be). Dr Rekin's Janky is acknowledged for his help with generating the heatmap.
Conceptualization: C.W., M. Bollen, A.V.E.; Methodology: C.W., R.R., M. Bollen; Software: R.R.; Validation: C.W., D.W., M. Bollen; Formal analysis: C.W., D.W.; Investigation: C.W.; Data curation: C.W., D.W., A.V.E.; Writing - original draft: C.W., A.V.E., M. Bollen; Writing - review & editing: C.W., M. Bollen; Visualization: C.W.; Supervision: M. Beullens, A.V.E., M. Bollen; Project administration: M. Bollen; Funding acquisition: M. Bollen.
This work was supported by Fonds Wetenschappelijk Onderzoek (grant G.078717N), the Stichting Tegen Kanker (Belgian Foundation Against Cancer), and Onderzoeksraad KU Leuven (GOA 15/016).
Gene expression data are available in the Gene Expression Omnibus under the accession number GSE104320.
The authors declare no competing or financial interests.