Endogenous p21 levels protect genomic stability by suppressing both excess and restrained nascent DNA syntheses

Abstract

The rate of DNA synthesis is crucial for full DNA duplication. We report a key role of p21 in controlling this rate. During normal replication, p21 promotes nascent DNA synthesis alongside the DNA polymerase iota (Pol ι)/p53 complex. When p21 is down-regulated but detectable, nascent DNA tracks are longer and discontinuous and rely on primase and DNA polymerase (PrimPol). With the complete elimination of p21, nascent DNA tracks become shorter and continuous and depend on Pol kappa (κ). Endogenous p21 levels are critical for genomic stability, as both PrimPol- and Pol κ–mediated syntheses can induce chromosomal instability. The residual expression of p21 in p53-null cells influences the involvement of PrimPol or Pol κ in nascent DNA synthesis and subsequent chromosomal instability. Our results demonstrate that endogenous levels of p21 in cycling cells, insufficient for cyclin-dependent kinase inhibition, prevent genomic instability through proliferating cell nuclear antigen binding (PCNA), limiting PrimPol and Pol κ’s role in nascent DNA synthesis.

p21 acts as a rheostat of DNA replication speed controlling PrimPol- and Pol kappa–dependent nascent DNA synthesis.

INTRODUCTION

To ensure its timely finalization, DNA replication initiates at multiple sites throughout the eukaryotic genome (1, 2). The rate of nascent DNA elongation is crucial for the complete duplication of DNA. However, DNA synthesis occurs at a slower rate after the accumulation of DNA damage than under unperturbed conditions. Such a slowdown takes place because replicative DNA polymerases (R-Pols) are blocked by chemically modified nucleotides (3). At these sites, the DNA damage tolerance (DDT) arm within the DNA damage response network prevents the persistent stalling of replication forks. The loss of DDT factors slows down the rate of DNA synthesis, which is detrimental to the cells (4, 5). The negative consequences of reducing DNA replication rate include the increase in replication stress markers, such as phosphorylated histone H2AX (γH2AX), the accumulation of underreplicated DNA, chromosome instability (CIN), and cell death (6–8). Intriguingly, not only the reduction but also the increase in DNA replication rate has, counterintuitively, been associated with negative outcomes at the level of CIN and, eventually, cell fitness (8–10). In contrast to the well-characterized molecular bases linking slow nascent DNA replication rate with CIN and cytotoxicity, the triggers for the negative outcomes in the context of longer nascent DNA tracks have not been fully elucidated.

A well-characterized DDT event that aids DNA replication in cells treated with DNA-damaging agents is translesion DNA synthesis (TLS) involving specialized DNA Pols (S-Pols) such as Pol eta (η), Pol kappa (κ), Pol iota (ι), Rev1, and the B-Pol, Pol zeta (ζ) (11–14). S-Pols ensure the continuity of nascent DNA synthesis at damaged sites. However, their low processivity frequently causes the accumulation of shorter nascent DNA tracks after different types of DNA damage, as reported for ultraviolet radiation, cisplatin, and benzo(a) pyrene diolepoxide (BPDE) (15, 16). In contrast to their important contribution to damaged DNA replication, the S-Pols are not expected to make a substantial contribution to the elongation of most nascent DNA tracks in the absence of DNA damage. Notwithstanding, they also contribute to DNA synthesis under unperturbed conditions in limited scenarios, specifically at hard-to-replicate regions of the genome [e.g., (17, 18)]. Similarly, while the DNA primase PrimPol (primase and DNA-directed polymerase) does not participate in the unperturbed synthesis of undamaged DNA, it does contribute to the lengthening of replication tracks after DNA damage by ultraviolet radiation, hydroxyurea, and cisplatin (19–21).

S-Pols may also participate in non-TLS DDT. This is particularly relevant for one S-Pol, DNA Pol ι (Pol ι), which modulates nascent DNA elongation both in the absence or presence of augmented DNA damage (7, 10) and in the context of a complex containing the tumor suppressor p53 (7, 22). It has been proposed that the Pol ι/p53 complex generates idling events at stalled forks, thereafter activating a DDT pathway involving fork reversal factors that oppose TLS events (7, 10, 22, 23).

Regardless of the DDT pathway contributing to changes in the length of nascent DNA track by S-Pols, the latter are required for the genomic stability and the fitness of cells. Their depletion causes DNA breakage, accumulation of underreplicated DNA, CIN, and cell death, and this happens selectively when they are not expressed at the time their contribution to DNA replication is needed (10, 20, 24–27). Hence, the choice and timing of participation of S-Pols in DDT are key to the protection of the genome.

Another regulator of nascent DNA synthesis is the cyclin kinase inhibitor p21. While p21 inhibits cyclin-dependent kinases (CDKs) through its N terminus when highly expressed, its C terminus is a potent interactor of the clamp loader proliferating cell nuclear antigen (PCNA) (28, 29). We have previously shown that p21 negatively regulates the recruitment of all Y-Pols to DNA (15). On the other hand, physiological levels of p21 are required to grant the unperturbed rate of nascent DNA track elongation (24). Intriguingly, apparently contrasting reports indicate both longer and shorter nascent DNA tracks after p21 depletion (9, 24). While such results reinforce the role of endogenous p21 levels on DNA replication, the mechanisms controlling such different outcomes in replication rates are unknown. To address the conundrum, we tested the relevance of the extent of p21 down-regulation on the rate of nascent DNA synthesis, finding that this variable is the source of the diametrically opposed effects on nascent DNA elongation. Our data indicate that small variations in p21 levels impose changes in the choice of DDT pathways involving Pol ι/p53, PrimPol, Pol κ, and Pol η, which are relevant for the maintenance of the chromosomal stability of cells. Together, our findings unequivocally demonstrate that endogenous levels of p21, while residual from the perspective of CDK inhibition, do have an important PCNA-associated function that relates to the control of the rate of nascent DNA synthesis affecting the protection of chromosomal stability.

RESULTS

p21 down-regulation up- and down-modulates nascent DNA synthesis

We have previously reported that levels of p21, which are insufficient to promote CDK-dependent cell cycle arrest in the G₁ phase, are enough to facilitate the lengthening of nascent DNA tracks during unperturbed DNA replication (24). More recently, Bartek’s group reported the exact opposite effect, implying that p21 prevents rather than facilitates nascent DNA synthesis (9). Careful comparison of the two protocols revealed that different small interfering RNAs (siRNAs) were used in the two reports. We then performed experiments comparing sip21#1 from ref. (24) and sip21#2 from ref. (9), reproducing the effects reported in each original manuscript: shorter nascent DNA tracks after sip21#1 transfection and longer tracks after sip21#2 transfection [Fig. 1A; 5-chloro-2′-deoxyuridine (CldU) tracks and representative images in fig. S1]. We also noticed that sip21#1 caused a more pronounced p21 down-regulation than sip21#2 (Fig. 1A, Western blot below). Because the endogenous p21 levels are low in U2OS cells (compared with the augmentation observed when treating U2OS cells with a DNA-damaging agent, daunorubicin; fig. S2A) (30), we concluded that sip21#2 is inefficient. Transfection of cells with increasing concentrations of the two siRNAs reinforced the conclusion that sip21#1 was more efficient than sip21#2 in depleting endogenous p21 in untreated U2OS and HCT116 cells (fig. S2, B and C).

Fig. 1. Both longer and shorter nascent DNA tracks are revealed after p21 KD.

(A) Labeling scheme and IdU track lengths from DNA fibers in U2OS cells transfected with siLuc and two different sip21. Cells were labeled with CldU and IdU for 10 and 30 min, respectively. Approximately 300 DNA fibers obtained from three independent experiments were analyzed for each condition. The bars on top of the distribution clouds indicate the median. Letters on top of the plot show statistical significance, as described in Materials and Methods. P < 0.001. Right panel: Representative DNA fibers (scale bar, 5 μm). Lower panel: Western blot revealed with a p21 antibody. Actin was used as a loading control. (B) Labeling scheme and IdU track lengths from DNA fibers in U2OS cells transfected with siLuc and two different concentrations of sip21#1: low (10 nM) and high (100 nM). Cells were labeled with CldU and IdU for 10 and 30 min, respectively. Approximately 300 DNA fibers obtained from three independent experiments were analyzed for each condition. The bars on top of the distribution clouds indicate the median. P < 0.001. Right panel: Representative DNA fibers (scale bar, 5 μm). Lower panel: Western blot revealed with a p21 antibody. Actin was used as a loading control. Samples siLuc and sip21#1-high used in this figure panel are the same used in (A). (C) Representative images of U2OS cells transfected as indicated (scale bar, 20 μm). Lower panel: Western blot revealed with a p21 antibody. Actin was used as a loading control. EV, empty vector. (D) IdU track lengths from DNA fibers in U2OS cells. Cells were labeled with CldU and IdU for 10 and 30 min, respectively. Approximately 300 DNA fibers obtained from three independent experiments were analyzed for each condition. The bars on top of the distribution clouds indicate the median. P < 0.001.

We speculated that the nascent DNA track length differences triggered by sip21#1 versus sip21#2 could depend on the efficiency of p21 down-regulation accomplished by each siRNA. We reasoned that, if that was the case, the results obtained with sip21#2 should have been recapitulated by lower amounts of sip21#1. To test this possibility, nascent DNA tracks were measured in samples subjected to transfection with increasing amounts of sip21#1. We obtained evidence of longer nascent DNA tracks when using sip21#1 at the lowest doses (5 and 10 nM) and shorter tracks at the highest doses (50 and 100 nM) (fig. S3A). We then compared a low dose (10 nM) and a 10 times higher dose (100 nM) of the same sip21#1 (designated as sip21#1-low and sip21#1-high, respectively). This resulted in longer and shorter nascent DNA tracks, respectively, in both U2OS and HCT116 cells (Fig. 1B and fig. S3B). While nascent DNA track lengthening was caused by sip21#1-low and sip21#2-high, yet at different siRNA concentrations, shorter nascent DNA tracks were observed only after the use of sip21#1-high but never when using the less efficient sip21#2. To verify that the results obtained with sip21#1-high were not associated with an off-target effect of sip21#1, we evaluated nascent track lengths after p21 knockout (KO) in different cell lines. We used U2OS (osteosarcoma used in previous figures), HCT116 (colorectal cancer used in previous figures), RPE-1 (normal cells from the retinal origin), and HT1080 (fibrosarcoma) p21 KO clones and control clones previously published: HCT116 (24), HT1080 (31), U2OS, and RPE-1 (32). In all cases and similarly to sip21#1-high, p21 KO caused the accumulation of shorter nascent DNA tracks (fig. S4, A to D). These results ruled out a potential off-target effect of sip21#1 and supported the likelihood that small variations in the extent of p21 down-regulation can either promote or restrain nascent DNA elongation. While sip21#1 promotes both types of modulations in a manner that depends on the dose of sip21#1 used, the less efficient sip21#2 achieves a partial p21 down-regulation only, selectively causing the lengthening of DNA tracks.

Having established that the down-regulation of p21, whether partial or complete, induces alterations in the length of nascent DNA tracks (longer and shorter), we set out to determine the contribution of p21 domains to this puzzling bimodal effect on DNA replication. Because sip21#1 is directed to the 3′ untranslated region of p21, mRNA corresponding to exogenously expressed p21 (that does not contain this sequence) is refractory to that sip21 (24). We modified such a plasmid to encode p21CDK−, i.e., p21 with a disrupted CDK binding site [amino acid residue mutations: W49R, F51S, and D52A (33)] that allows cell cycle progression in the presence of high levels of p21 (29). We also introduced a second set of mutations that disrupts the PCNA binding site [amino acid residue mutations: M147A, D149A, and F150A (33)] previously validated by us (29). Lentiviral transduction allowed the expression of such p21CDK− mutants either without (p21) or with (p21^PIPMut) additional point mutations in its PCNA interacting protein (PIP) motif (33) without disrupting the bipartite nuclear localization signal of p21 (28). The transduction efficiency was ~70% of total cells (Fig. 1C), and the nuclear localization of both p21 and p21^PIPMut was 100% (see representative images). We found that altered replication dynamics observed after complete and partial p21 down-regulation were prevented when expressing p21 but not p21^PIPMut (Fig. 1D). Thus, both the longer and shorter nascent DNA tracks after p21 down-regulation and depletion result from the loss of p21-PCNA interactions in cycling cells.

p21 levels rule over the DNA Pol choice

The DNA Pol in charge of DNA replication can modify the length of DNA tracks (34, 35). It is therefore possible that the differential effects that partial and complete p21 knockdown (KD) have on track lengths are the consequences of DNA replication by different DNA Pols. Our previous observations indicate that the shorter nascent DNA tracks can be associated, between other possibilities, with excess DNA synthesis by DNA Pol κ (Pol κ), an S-Pol involved in TLS, which is less processive than replicative Pols ε and δ (24). Depletion of Pol κ (Fig. 2A) achieved using an siRNA previously validated by us (24) confirmed that the short tracks induced by sip21#1-high were dependent on the excess participation of Pol κ in nascent DNA synthesis (Fig. 2B). Conversely, Pol κ was not responsible for the lengthening of DNA tracks observed after sip21#1-low transfection (Fig. 2B). Given that the DNA Pol and primase PrimPol is well known for lengthening nascent DNA tracks (10, 20, 36, 37), we also tested the KD of PrimPol (Fig. 2A) using an siRNA previously validated by us (10). PrimPol depletion did not affect replication tracks from control samples or those transfected with sip21#1-high but reverted the lengthening of DNA tracks observed after sip21#1-low transfection (Fig. 2B). As a control, we evaluated the Pol κ and PrimPol levels after sip21#1-low and sip21#1-high transfection, revealing no changes in the levels of the indicated DNA Pols (Fig. 2C). The effect of siPol κ and siPrimPol was also tested in those Western blots, confirming results in Fig. 2A (Fig. 2C). Corresponding results to those in Fig. 2B were obtained in HCT116 cells (fig. S5, A and B) and in U2OS cells transfected with sip21#1 and sip21#2, respectively (fig. S5C). In addition, in HCT116 p21 KO and HT1080 p21 KO cells, the short DNA tracks are Pol κ–dependent. Here, the nascent DNA track length was modulated by siPol κ but not by siPrimPol (fig. S5, D to G). These results demonstrate that physiological p21 levels prevent the excess participation of PrimPol and Pol κ in nascent DNA synthesis. When p21 levels are partially reduced (sip21#1-low or sip21#2), the involvement of PrimPol in DNA replication increases and the nascent DNA tracks are longer when compared to control tracks. In contrast, when p21 levels are undetectable (sip21#1-high or p21 KO), Pol κ–dependent DNA synthesis prevails and track lengths are shorter when compared to control tracks.

Fig. 2. p21 levels in cycling cells influence the DNA Pol choice at nascent DNA sites.

(A) Quantitative real-time PCR of PrimPol and Pol κ normalized to GAPDH in U2OS cells. Error bars represent the SD of three technical replicates from one experiment representative of two independent experiments. P < 0.01. (B) IdU track lengths from DNA fibers in U2OS cells. Cells were labeled with CldU and IdU for 10 and 30 min, respectively. Approximately 300 DNA fibers obtained from three independent experiments were analyzed for each condition. The bars on top of the distribution clouds indicate the median. P < 0.001. (C) Western blot revealing PrimPol and Pol κ levels in U2OS cells with the indicated siRNAs. Actin was used as a loading control. (D) Representative image of U2OS cells transfected with the V5-PrimPol expression plasmid with (positive) or without (negative) V5-PrimPol foci (scale bar, 5 μm). (E and F) Percentage (means + SD) of U2OS cells with V5-PrimPol foci. Approximately 750 nuclei per sample were analyzed in three independent experiments. P < 0.01. (G) Percentage (means + SD) of U2OS cells with V5- PrimPol or V5-PrimPol RBDm foci. Approximately 600 nuclei per sample were analyzed in two independent experiments. P < 0.01. (H) Percentage (means + SD) of U2OS cells with more than four RPA foci. Approximately 300 nuclei per sample were analyzed in two independent experiments. P < 0.01. The right panel shows representative images of U2OS cells with less than four RPA foci (negative) or with more than four RPA foci (positive). Scale bar, 10 μm. (I) Model. Partial silencing of p21 causes an increase in PrimPol-dependent nascent DNA synthesis. In contrast, complete silencing of p21 causes an augmentation of Pol κ–mediated DNA replication.

To confirm the association of PrimPol and Pol κ with DNA replication structures after partial and complete down-regulation of p21, respectively, we transiently expressed V5-PrimPol (Fig. 2D) and green fluorescent protein (GFP)-Pol κ (fig. S6A) and quantified cells with nuclear foci of each DNA Pol, as previously established. All expression vectors for GFP- or V5-tagged Pols were previously described (10, 24, 38, 39). Notably, the percentage of cells with PrimPol foci in the nucleus increased only when partial p21 expression was achieved, e.g., under sip21#1-low but not sip21#1-high conditions (Fig. 2E). Such excess in the focal organization of PrimPol was still evident when transducing cells with p21^PIPMut but was not observed under conditions of expression of p21 with a wild-type PIP (Fig. 2F). As expected from the known properties of PrimPol (21), its focal organization depended on its ability to bind replication protein A (RPA) under all conditions tested (Fig. 2G). Such a conclusion was reached when comparing V5-PrimPol and a mutant version with a mutated RPA binding domain (amino acids D519R and F522A; 5-PrimPol RBDm) (39). Intriguingly, the focal organization of RPA is the lowest when PrimPol-mediated DNA synthesis takes place under sip21#1-low conditions (Fig. 2H). These results suggest that, while PrimPol recruitment to replication sites depends on RPA binding, it does not linearly correlate with RPA focal organization.

Similar correlations between DNA track lengths and Pol κ focal organization were obtained when quantifying GFP-Pol κ foci (fig. S6B). While the percentage of cells with Pol κ foci increased only after sip21#1-high transfection (fig. S6B), such an increase was reverted by the expression of p21 but not p21^PIPMut (fig. S6C). Moreover, as we expected, sip21#2, which achieves only a partial down-regulation of p21, led to an augmentation of the number of cells with PrimPol but not Pol κ foci (fig. S6, D and E). These results supported the notion that p21 exquisitely controls the choice of DNA Pols in charge of DNA replication in a manner that depends on its levels (Fig. 2I).

Full, but not partial, p21 depletion stresses S phase transit

A solid amount of evidence demonstrated that shorter nascent DNA tracks are associated with replication stress (11). Longer tracks, on the other hand, were also suggested as a source of replication stress, particularly when exploring the role of p21 down-regulation and PARP (poly-ADP ribose polymerase) inhibition (PARPi) on DNA replication (9). We therefore compared the consequences of PARPi by olaparib, sip21#1-low, and sip21#1-high treatments. Olaparib treatment favors PrimPol focus formation and provokes the accumulation of DNA tracks of a length similar to that of sip21#1-low (Fig. 3, A and B). In addition, the long tracks in olaparib and sip21#1-low samples are fully dependent on PrimPol (Fig. 3B). Another similarity between olaparib and sip21#1-low treatments was the sensitivity of their nascent DNA tracks to S1 nuclease treatment (Fig. 3B). Intriguingly, the lengths of tracks detected following treatment with olaparib + S1 nuclease and sip21#1-low + S1 nuclease are shorter than those in control samples, potentially indicating similarly inefficient sealing of PrimPol-generated gaps under both conditions (Fig. 3B). While olaparib may affect the track length at the level of formation of leading strand gaps or accumulation of unsealed lagging strand tracks (40), our experimental conditions are insufficient to explore such details as DNA combing is necessary to reveal such information (41). In any case, olaparib and sip21#1-low show very similar effects at the level of nascent DNA synthesis.

Fig. 3. p21 down-regulation and PARPi cause a PrimPol-dependent augmentation in the length of nascent DNA tracks, but only PARPi increases γH2AX focal organization.

(A) Percentage (means + SD) of U2OS cells with V5-PrimPol foci. Approximately 600 nuclei per sample were analyzed in two independent experiments. PARPi (10 μM olaparib) treatment was used for 24 hours. P < 0.01. Samples siLuc, sip21#1-low, and sip21#1-high used in this figure panel are the same used in Fig. 2G. (B) IdU track lengths from DNA fibers in U2OS cells. Cells were labeled with CldU and IdU for 10 and 30 min, respectively. S1 nuclease was used to reveal DNA gaps. Approximately 200 DNA fibers obtained from two independent experiments were analyzed for each condition. The bars on top of the distribution clouds indicate the median. PARPi (10 μM olaparib) treatment was used for 24 hours. P < 0.001. (C) BrdU intensity from U2OS cells. Approximately 500 cells obtained from two independent experiments were measured for each condition. The bars on top of the distribution clouds indicate the median. PARPi (10 μM olaparib) treatment was used for 24 hours. P < 0.001. The right panel shows representative images corresponding to different levels of BrdU intensity (expressed in arbitrary units) in nuclei from U2OS (scale bar, 5 μm). (D) Percentage of S phase (BrdU-positive) U2OS cells (means + SD). Approximately 1500 cells per sample were analyzed in three independent experiments. P < 0.01. (E) Intensity plot obtained after quantification of γH2AX in U2OS cells. More than 600 cells per sample were analyzed in two independent experiments. The bars on top of the distribution clouds indicate the median. P < 0.001. The right panel shows representative images of U2OS cells transfected with siLuc (control), PARPi, sip21#1-low, and sip21#1-high (scale bar, 20 μm).

Despite the resemblances at the level of nascent DNA synthesis, a number of dissimilar events were documented after olaparib and sip21#1-low treatments immediately after nascent DNA synthesis. Both olaparib and sip21#1-low cause increased 5-bromo-2′-deoxyuridine (BrdU) incorporation in BrdU-positive cells (Fig. 3C), but the percentage of BrdU-positive cells increases after olaparib but not after sip21#1-low treatment (Fig. 3D), implying a longer transit through the S phase for olaparib treatment but not sip21#1-low treatment. Strengthening the differences between olaparib and sip21#1-low, the focal organization of γH2AX, which can be quantified by intensity (fig. S7) and interpreted as a marker of double-strand breaks (42), increases after olaparib and not sip21#1-low treatment (Fig. 3E). Together, these results demonstrate that long nascent DNA tracks are not always a prelude to replication stress and double-strand break formation in the S phase, as previously suggested (9). In contrast, sip21#1-high causes the accumulation of replication stress following the accumulation of shorter nascent DNA tracks (Fig. 3E).

Partial and complete p21 depletion triggers PrimPol- and Pol κ–mediated chromosome instabilities, albeit with differences at the level of replication stress–associated markers

We then wondered about the biological relevance of the inhibitory effect on PrimPol and Pol κ exerted by p21. Dysregulation of DNA replication dynamics reportedly causes incomplete finalization of DNA replication and accumulation of CIN in G₂ or M phases (43–45). We therefore performed analyses of micronuclei and aberrant anaphases, two types of CIN with important roles in the modulation of tumor immunity, tumor invasion, and adaptation to treatment (46). Partial (sip21#1-low) and complete (sip21#1-high) p21 KD triggered the marked accumulation of micronuclei, which depended on PrimPol when p21 KD was partial and on Pol κ after complete p21 KD (Fig. 4A). The formation of micronuclei takes place at the end of the M phase and can be triggered by the accumulation of anaphase aberrations and other mitotic defects (44, 46). Similar to the results obtained when quantifying micronuclei, partial and complete p21 KD caused the marked accumulation of anaphase bridges and acentric/lagging chromosomes, which depended on PrimPol after partial p21 KD and on Pol κ after complete p21 KD (Fig. 4B). Under physiological p21 levels, the basal levels of micronuclei and anaphase aberrations were independent of PrimPol or Pol κ (Fig. 4, A and B). Collectively, these findings demonstrate that p21 prevents PrimPol or Pol κ from synthesizing nascent DNA, thereby safeguarding a DNA replication program free of CIN.

Fig. 4. PrimPol causes γH2AX-independent CIN in samples with partial p21 KD, while Pol κ causes γH2AX-associated CIN in samples with complete p21 KD.

(A) Percentage (means + SD) of binucleated U2OS cells with micronuclei. Approximately 450 binucleated cells per sample were analyzed in three independent experiments. P < 0.01. The right panel shows representative images of binucleated U2OS cells without (negative) or with (positive) micronuclei (scale bar, 10 μm). (B) Percentage of anaphases with aberrations (means + SD) of chromosome bridges and lagging/acentric chromosomes. Approximately 150 anaphases per sample were analyzed in three independent experiments. P < 0.01. The right panel shows representative images of U2OS normal anaphasic cells and anaphase aberrations (anaphase bridges and lagging/acentric chromosomes) in a z-stack (scale bar, 5 μm). (C) Intensity plot obtained after quantification of γH2AX in U2OS cells. More than 1000 cells per sample were analyzed in three independent experiments. The bars on top of the distribution clouds indicate the median. P < 0.001. (D) Percentage (means + SD) of PICH-positive anaphases. More than 150 anaphases per sample were analyzed in three independent experiments. P < 0.01. The right panel shows representative images of U2OS anaphases without (negative) or with (positive) PICH-positive ultrafine bridges (scale bar, 5 μm). (E) Percentage (means + SD) of EdU-negative U2OS cells with more than four 53BP1 nuclear bodies (53BP1-NBs). Approximately 600 nuclei per sample were analyzed in three independent experiments. P < 0.01. The right panel shows representative images of EdU-negative U2OS cells without (negative) and with (positive) more than four 53BP1 nuclear bodies (scale bar, 10 μm).

The vast majority of reports concerning CIN documents replication stress as a preceding event (5, 47, 48). Only few exceptions report a type of CIN that is not preceded by replication stress (6, 10, 49, 50). We then tested whether PrimPol or Pol κ depletion had an effect on the intensity of γH2AX after complete p21 KD and found a selective contribution of Pol κ to γH2AX accumulation (Fig. 4C). We further noticed that other features that were previously reported to stem from persistent replication stress such as the formation of ultrafine bridges in the M phase or the accumulation of cells with p53-binding protein 1 (53BP1) nuclear bodies in the G₁ phase (43, 44, 51, 52) were more pronounced under conditions of complete but not partial p21 KD (Fig. 4, D and E). Similar results were obtained in a second cell line, HCT116, when exploring micronuclei and γH2AX and 53BP1 nuclear bodies (fig. S8, A to C). Moreover, such observations were in line with results obtained when using sip21#2 (fig. S8, D and E). In conclusion, long and short nascent DNA tracks caused by partial and complete p21 depletion, respectively, trigger two types of CIN, which mechanistically differ at the level of accumulation of the replication stress marker γH2AX.

p21 and Pol ι/p53 prevent PrimPol-mediated DNA replication

So far, our results indicate that PrimPol participates in DNA replication only when p21 is down-regulated. In a recent article from our laboratory, we reported that Pol ι prevents PrimPol-dependent DNA replication (10), which was observed under conditions of physiological p21 expression. Hence, we set out to explore the relationship between Pol ι and p21 regarding their individual contributions to restricting PrimPol-mediated nascent DNA elongation. Because the Pol ι–dependent antagonistic effect on PrimPol-dependent DNA synthesis relies on the formation of the Pol ι/p53 DDT complex (7, 10, 22, 23), we used HCT116 p53 KO, p21 KO, and parental cells for such investigations. In agreement with previous reports (53, 54), p53-independent p21 expression was detected in p53 KO cells when comparing these levels to those of p21 KO cells (Fig. 5A). When measuring the nascent DNA track length as a function of PrimPol and Pol κ down-regulation (fig. S5, A and D, and Fig. 5B), we observed that, in p53 KO cells, nascent DNA tracks were longer than those in control and p21 KO samples, which depended on PrimPol (Fig. 5C). While this result aligns with previous experiments of this manuscript demonstrating that PrimPol is dysregulated when p21 levels are low, it is unclear whether p53 exerts a direct negative effect on PrimPol or if it is mediated by p21. To discriminate between these possibilities, we augmented p21 levels in p53 KO cells. We reasoned that, if p53 controlled nascent DNA replication exclusively through the transcriptional up-regulation of p21, then PrimPol-mediated replication should be eliminated when reexpressing p21 despite the depletion of p53. Instead, track lengths were not modified by p21 reexpression, implying that, in addition to p21, the participation of p53 is needed to restrain PrimPol-mediated track lengthening (Fig. 5D).

Fig. 5. p21 and Pol ι/p53 prevent PrimPol-mediated nascent DNA replication.

(A) Western blot revealing p53 and p21 in parental HCT116, HCT116 p53 KO, and HCT116 p21 KO cells. Actin was used as a loading control. (B) Quantitative real-time PCR of PrimPol and Pol κ normalized to GAPDH. Error bars represent the SD of three technical replicates within one experiment from two independent experiments. P < 0.01. (C) IdU track lengths from DNA fibers in HCT116, HCT116 p53 KO, and HCT116 p21 KO cells. CldU: 10 min; IdU: 30 min. Approximately 300 DNA fibers obtained from three independent experiments were analyzed for each condition. The bars indicate the median. P < 0.001. (D) IdU track lengths from DNA fibers in HCT116, HCT116 p53 KO, and HCT116 p21 KO cells. CldU: 10 min; IdU: 30 min. Approximately 200 DNA fibers obtained from two independent experiments were analyzed for each condition. The bars indicate the median. P < 0.001. The lower panel shows a Western blot revealed with a p21 antibody. Actin was used as a loading control. (E) IdU track lengths from DNA fibers in U2OS cells. CldU: 10 min; IdU: 30 min. Approximately 150 DNA fibers obtained from two independent experiments were measured for each condition. The bars indicate the median. P < 0.001. The lower panel shows p53 and p21 levels in a Western blot. Actin was used as a loading control. (F) IdU track lengths from DNA fibers in U2OS WT (clone 18) and U2OS Pol ι KO (clones 13 and 24) cells. CldU: 10 min; IdU: 30 min. Approximately 200 DNA fibers obtained from two independent experiments were analyzed for each condition. The bars indicate the median. P < 0.001. (G) Model. p21 regulates PrimPol through Pol ι. However, p21 regulates Pol κ more strongly, and it can only synthesize DNA when p21 is completely depleted.

Different from the results obtained in p53 KO cells, in their p21 KO counterparts, the expression of exogenous p21 caused the reconstitution of normal track lengths (Fig. 5D). These results indicate that, while p21 alone restricts Pol κ participation in DNA synthesis, both p21 and p53 are needed to restrict PrimPol-dependent DNA synthesis. In addition, in U2OS cells, p53 depletion by siRNA causes the lengthening of DNA tracks (Fig. 5E), which we previously demonstrated depends on PrimPol (10), but such a lengthening does not happen if cells are depleted from p21 (Fig. 5E). Hence, both p21 and p53 expressions are needed to prevent untimely PrimPol-dependent nascent DNA synthesis.

We then wondered whether p53 prevented PrimPol-mediated DNA synthesis in a manner that depends on Pol ι, as suggested previously by us (10). If that is the case, the PrimPol-mediated track lengthening should be similarly favored in Pol ι KO samples and in p21-depleted samples. Hence, we used U2OS Pol ι KO clones previously characterized by us (10) as well as sip21#1 and siPrimPol to achieve low and high p21 down-regulation in combination with PrimPol depletion (fig. S9, A and B, and Fig. 5F). Quantification of track lengths confirmed that sip21#1-low and Pol ι KO caused the lengthening of nascent DNA tracks (Fig. 5F). The combination of sip21#1-low and Pol ι KO did not cause further changes in the average track lengths, hence indicating epistasis. PrimPol KD confirmed that the lengthening of tracks observed after sip21#1-low transfection in Pol ι KO cells and sip21#1-low/Pol ι KO samples was fully dependent on PrimPol (Fig. 5F). Last, when applying sip21#1-high conditions, nascent DNA tracks were shorter in the wild-type control cells and the two Pol ι KO clones. PrimPol depletion did not modify these short tracks (Fig. 5F), which is consistent with the fact that short tracks are dependent on Pol κ and not PrimPol (Fig. 2B). Hence, p21 and Pol ι/p53 are both required to prevent PrimPol participation in nascent DNA replication. Single p53, Pol ι, or p21 KD is sufficient to promote PrimPol-dependent nascent DNA synthesis.

We propose that p21 favors Pol ι/p53–mediated nascent DNA synthesis, which counteracts PrimPol, while p21 displaces Pol κ from replication sites (Fig. 5G). When p21 levels are reduced, the Pol ι/p53–mediated nascent DNA synthesis is lost, but the prevention of Pol κ–mediated synthesis is still active, creating a window of opportunity for PrimPol-mediated replication. Only severe p21 KD by sip21#1-high or p21 KO promotes Pol κ–dependent nascent DNA synthesis that restricts the PrimPol-mediated events.

We also explored the relevance of p53 for the prevention of CIN and replication stress. p53 KO cells accumulated micronuclei, which was dependent on PrimPol expression (fig. S10A). In these cells, we did not observe γH2AX accumulation (fig. S10B), which is consistent with a PrimPol-dependent effect observed under sip21#1-low conditions (Fig. 4C). In contrast, p21 KO cells accumulated micronuclei triggered by Pol κ expression (fig. S10A), which was preceded by γH2AX accumulation (fig. S10B), recapitulating results with sip21#1-high (Fig. 4, A and C). While PrimPol-mediated replication did not induce γH2AX in p53 KO cells, the depletion of p21 in these samples caused an increased in γH2AX (fig. S10, C and D) that depends on unleashed Pol κ (fig. S10E). Similar results were recapitulated in U2OS cells depleted from p53 and p21, hence implying that the levels of p21 expressed in those p53 down-regulation scenarios suffice to prevent Pol κ– but not PrimPol-dependent DNA replication and genomic instability. Together, these results demonstrate that p21 and Pol ι/p53 act in concert to prevent PrimPol-mediated DNA replication and the consequential PrimPol-mediated CIN augmentation.

Pols ƞ and κ prevent PrimPol-mediated DNA replication after p21 KO

So far, our results demonstrate that, while PrimPol-mediated replication is evident after the partial depletion of p21, Pol κ–mediated DNA replication is observed only when p21 levels are undetectable. If our model is correct, we should then expect to observe a restoration of PrimPol-dependent replication, i.e., track lengthening, under conditions of sip21#1-high/siPol κ. Such PrimPol-mediated DNA synthesis should result from the lack of p21 expression and the elimination of Pol κ that competes with PrimPol. Instead, under conditions of sip21#1-high/siPol κ, PrimPol-mediated replication is undetectable (Fig. 2B). This result suggests that PrimPol is unable to participate in DNA replication because of the action of yet another unknown factor. The lack of participation of PrimPol in DNA replication under conditions of sip21#1-high/siPol κ is also suggested by the inability of PrimPol to assemble in nuclear foci (Fig. 6A).

Fig. 6. Pols κ and η prevent PrimPol-mediated nascent DNA replication after complete p21 KD.

(A) Percentage (means + SD) of U2OS cells with V5-PrimPol foci. Approximately 500 nuclei per sample were analyzed in two independent experiments. P < 0.01. (B) Western blot revealing Pol η and p21 in parental HCT116 and HCT116 p21 KO cells. Actin was used as a loading control. (C) Percentage (means + SD) of HCT116 and HCT116 p21 KO cells with V5-PrimPol foci. Approximately 500 nuclei per sample were analyzed in two independent experiments. P < 0.01. (D) IdU track lengths from DNA fibers in HCT116 and HCT116 p21 KO cells. Cells were labeled with CldU and IdU for 10 and 30 min, respectively. Approximately 200 DNA fibers obtained from two independent experiments were analyzed for each condition. The bars on top of the distribution clouds indicate the median. P < 0.001. (E) Model. p21 controls PrimPol through Pol ι. However, at high levels, p21 primarily antagonizes Pol κ. Only when p21 is completely depleted, Pols κ and η prevent PrimPol DNA replication.

Because p21 potently displaces all S-Pols from PCNA, we reasoned that perhaps not only Pol κ but also another S-Pol acts in concert with Pol κ to prevent PrimPol-dependent DNA synthesis under conditions of sip21#1-high. We reasoned that Pol η could have a role as we previously showed that Pol η depletion promotes PrimPol-dependent events (55). We then tested whether PrimPol nuclear focus accumulation was restored under conditions of concomitant p21 KO/siPol κ/siPol η (Fig. 6B) and found that this was the case (Fig. 6C). However, such an increase in the number of cells with PrimPol foci was not observed under conditions of p21 KO/siPol κ and p21 KO/siPol η (Fig. 6C), indicating that depletion of both Pols κ and η was needed to restore PrimPol-mediated nascent DNA synthesis in p21 KO cells. We also tested the effects on the nascent DNA synthesis of concomitant Pol κ and η depletion and found that PrimPol-mediated track lengthening was detected in p21 KO cells simultaneously transfected with siPol κ/siPol η but not with single siPol κ or siPol η (Fig. 6D). Such a result indicated that Pols κ and η in concert prevent PrimPol focal organization.

We also tested the role of Pol ι in the prevention of PrimPol-dependent DNA replication. While, under conditions of sip21#1-low, Pol ι was required to prevent the augmentation of cells with PrimPol foci, under conditions of sip21#1-high, there was no contribution of Pol ι to the regulation of PrimPol (fig. S11, A and B). We concluded that PrimPol-dependent DNA synthesis is tightly regulated and can happen only under conditions of p21KD. PrimPol-mediated DNA synthesis is prevented by p21 and Pol ι/p53 under conditions of unperturbed expression of endogenous p21, while PrimPol is controlled by the action of Pols κ and η when p21 is depleted (by KO or efficient KD; Fig. 6E).

PrimPol and Pol κ cause CIN in p53-null cells

p53 is frequently mutated or depleted in cancer cells, leaving residual levels of p21 in cells (56). Given the relevance of the levels of p21 in the DNA Pol choice during DNA replication and CIN prevention in HCT116 cells, we decided to test whether that was valid for cancer cells that lost p53 expression during carcinogenesis. To explore so, we used two p53-null cancer cell lines: the lung carcinoma H1299 and ovarian cancer SKOV3 cell lines (Fig. 7A). While H1299 cells express residual levels of p21, SKOV3 cells are characterized by even lower and almost undetectable levels of p21 (Fig. 7A). We depleted PrimPol and Pol κ in these cells (Fig. 7B) and found that H1299 cells with detectable p21 expression replicate nascent DNA in a manner that depends on PrimPol, while SKOV3 cells, with undetectable p21, have a DNA replication program that depends on Pol κ (Fig. 7, C and F). When exploring the accumulation of CIN, we found that the number of cells with micronuclei was reduced after PrimPol and Pol κ depletion in H1299 and SKOV3 cells, respectively (Fig. 7, D and G), hence implying that H1299 and SKOV3 cells accumulate DNA replication–associated CIN. Further down-regulation of p21 in H1299 cells with sip21#1-high caused a shift in the DNA Pol in charge of micronucleation (from PrimPol to Pol κ; Fig. 7E). In contrast, SKOV3 cells that express almost null levels of p21 were not further affected by sip21#1-high with micronuclei dependent on Pol κ under both control and sip21#1-high conditions (Fig. 7H). Then, while p21 is not lost during tumorigenesis, the modulation of p21 levels that results from the elimination of the p53 tumor suppressor reveals a crucial rheostat role of p21 in the choice of the DNA Pol in charge of DNA replication, with detectable consequences at the levels of the accumulation of CIN.

Fig. 7. Nascent DNA track lengths and CIN in p53-null cell lines are ruled by the levels of p21 and downstream participation of PrimPol or Pol κ in DNA replication.

(A) Western blot revealing p53 and p21 levels in H1299 and SKOV3 cells. Actin was used as a loading control. sip21#1 (100 nM) was transfected into each cell line to achieve complete p21 down-regulation. (B) Quantitative real-time PCR of PrimPol and Pol κ normalized to GAPDH in H1299 and SKOV3 cells. Error bars represent the SD of three technical replicates from one experiment representative of two independent experiments. P < 0.01. (C) IdU track lengths from DNA fibers in H1299 cells. Cells were labeled with CldU and IdU for 10 and 30 min, respectively. Approximately 200 DNA fibers obtained from two independent experiments were analyzed for each condition. The bars on top of the distribution clouds indicate the median. P < 0.001. (D) Percentage (means + SD) of binucleated H1299 cells with micronuclei. Approximately 300 binucleated cells per sample were analyzed in two independent experiments. P < 0.01. (E) Percentage (means + SD) of binucleated H1299 cells with micronuclei. sip21#1 (100 nM) was transfected to achieve complete p21 down-regulation. Approximately 300 binucleated cells per sample were analyzed in two independent experiments. P < 0.01. (F to H) The IdU track length from DNA fibers and the percentage of binucleated SKOV3 cells with micronuclei treated with siRNAs were determined as in (C) to (E). P values for (F) to (H) are <0.001, 0.01, and 0.01, respectively.

DISCUSSION

p21 is a well-established CDK inhibitor that must increase considerably for effective inhibition (57–59). It also displaces PCNA partners from its interdomain connecting loop (58, 60). This study shows that, even at low levels, p21 regulates the nascent DNA synthesis pathway choice through interaction with PCNA. Normal p21 levels favor a Pol ι/p53–mediated nascent DNA synthesis, causing the lowest CIN levels in cells. p21 down-regulation leads to a shift to PrimPol-mediated nascent DNA synthesis, increasing DNA replication rates. In this context, PrimPol’s role in DNA replication results in replication stress–independent CIN. Complete p21 elimination further alters the pathway in charge of nascent DNA synthesis, leading to Pol κ–dependent replication, reduced DNA replication rates, and increased γH2AX and CIN.

p21 rules over the DNA Pol choice at replication forks

The levels of p21 in cycling cells have long been interpreted as residual and insufficient to perform any biologically relevant function. From the perspective of its interaction with PCNA, overexpression of a p21 mutant with a disrupted CDK binding site but an unaltered PCNA binding domain did not modify bulk DNA replication or nucleotide excision repair, implying that physiological p21 levels would not suffice to modulate PCNA-related activities in cycling cells (28, 58). Such a view was challenged by the discovery of a distinct role of endogenous p21 in the modulation of nascent DNA elongation, which was initially reported by us (24) and later on by the group of Bartek (9). However, the results from those manuscripts were opposed in terms of the length of the nascent DNA tracks documented after p21 down-regulation. Later reports (61, 62) reproduced both our results (24) and Maya Mendoza et al.’s (9) results, respectively, suggesting that the explanation of such a puzzling difference went beyond technical mistakes. Herein, we thoroughly investigated such a conundrum, providing a detailed mechsanistic explanation for such differences. We found PrimPol and Pol κ as the effectors of these two DNA replication phenotypes and revealed their relevance in the context of p53 loss. Our results demonstrate that the p21 levels in unperturbed cycling cells, which have repeatedly been reported as undetectable in the literature, have a function in the DNA Pol choice and the protection of the genome.

Cross-talk between p21 and other components of the replisome

As shown in Fig. 8A, the experiments presented here imply that, in the absence of exogenous stress, p21 levels in cells promote Pol ι/p53–dependent DNA replication events, disfavoring PrimPol and Pol κ participation in DNA replication. Our results have also identified the PCNA binding site of p21 as the critical p21 region relevant for such a hierarchy in the DNA Pol choice. In addition, we have previously shown that p21 is not an efficient inhibitor of R-Pols. A massive overexpression of a p21 mutant that does not bind CDKs but interacts with PCNA does not prevent DNA replication by replicative Pols (28). In notable opposition, endogenous levels of p21 in cycling cells suffice to prevent DNA replication by S-Pols (15, 24). We have previously speculated that the reason behind such differences is related to the number of PIPs in each enzyme. While p21 encodes the strongest PIP box, replicative Pols have a multisubunit organization with multiple PIPs that may prevent their displacement from the PCNA trimer (with three docking sites for PIPs) by p21 (58). In contrast, S-Pols harbor a single PIP with an affinity that is lower to the one of p21. This may explain why a complete down-regulation of p21 is needed to reveal Pol κ–mediated DNA synthesis.

Fig. 8. Model depicting the role of p21 levels in the control of DNA replication.

(A) Left panel: The endogenous levels of p21 in cells undergoing DNA replication inhibit Pol κ and PrimPol participation in nascent DNA elongation, ensuring a safe rate of DNA replication (green region in the speed meter). Genomic stability is at its highest. Middle panel: Partial depletion of p21 increases the nascent DNA elongation rate because of excess PrimPol participation in DNA replication (dangerously high region in the speed meter). PrimPol-mediated DNA replication does not induce replication stress but induces genomic instability. Right panel: Complete elimination of p21 reduces the nascent DNA elongation rate because of excess participation of Pol κ in DNA replication. Pol κ–mediated DNA replication triggers replication stress and genomic instability. (B) The participation of PrimPol in DNA synthesis is subject to strong regulation. p21 prevents PrimPol from participating in nascent DNA synthesis in concert with the Pol ι/p53 DNA replication complex. When p21 is completely down-regulated, Pols κ and η act in concert to prevent PrimPol-dependent DNA synthesis. Only when p21 is partially down-regulated, a window of opportunity for PrimPol is created because the Pol ι/p53 complex is not favored and the participation of Pols κ and η in DNA replication is still prevented by p21.

The mechanism favoring PrimPol-dependent DNA synthesis at intermediate levels of p21 is more puzzling. PrimPol does not have a protein interaction region (PIR); furthermore, PrimPol is recruited to the single-stranded DNA binding protein RPA as shown previously (36, 63) and here. Hence, p21 prevents PrimPol-mediated DNA replication indirectly. As shown in Fig. 8B, when p21 levels are those corresponding to endogenous expression, the Pol ι/p53 complex prevents PrimPol-mediated DNA synthesis. p53 binds RPA, and Pol ι binds PCNA (22, 64). A dual docking in the Pol ι/p53 complex may displace PrimPol from the replisome. Furthermore, endogenous p21 aids PCNA ubiquitination (65), and PCNA polyubiquitination favors the Pol ι/p53 DDT pathway (7, 23). When p21 levels are undetectable, the S-Pols κ and η prevent PrimPol-mediated DNA synthesis. Pol κ is the most processive S-Pol (11, 66, 67). We speculate that, when accessing replisomes first, Pols κ and η overload the DNA synthesis sites with RPA, preventing PrimPol recruitment because of its limited capacity to displace RPA (68, 69). While such a hypothesis must be tested, the identification of protein networks that limit PrimPol activity (p21-Pol ι-p53 and Pol κ-Pol η) may be timely given the prominent role that PrimPol is gaining in the study of cancer genesis and treatment (70–72).

Replication stress–free or replication stress–triggered CIN after p21 KD

Nascent DNA track measurement represents a valuable variable relevant for the study of cancer genesis and treatment (73, 74). The reduced length of nascent tracks was early on associated with transient and permanent fork stalling, single-stranded DNA accumulation, increased checkpoint activation, and entrance in G₂-M phases with increased levels of underreplicated DNA (75, 76). In line with such a model, efficient p21 KD or KO causes a switch to Pol κ–dependent shorter nascent DNA tracks, associated with replication stress in the S phase, a well-characterized source of CIN (77). Later on, it was proposed that, when the nascent DNA replication rate increases above a threshold, replication forks become more fragile and checkpoint signals augment (9). Here, we show that this is not always the case, yet the consequences are negative for other reasons. After partial p21 KD, the contribution of PrimPol to nascent DNA replication does not cause replication stress or DSB formation, yet it causes augmentation of chromosomal instability including micronuclei, structures that can trigger severe and rapid chromosomal rearrangements (78–80). The replication stress–independent CIN triggered by PrimPol is reminiscent of the CIN profile reported for Pol ι depletion, which is also dependent on PrimPol dysregulation (10). It is possible that, under unperturbed conditions, DNA synthesis by PrimPol does not interfere with bulk DNA replication in the S phase, causing however a fast but incomplete and checkpoint-blind DNA synthesis (10). Hence, during unperturbed DNA replication, genomic instability is actively prevented by p21.

Turning to the biological relevance of these findings, we observe that, in p53-null cells, residual levels of p21 play a role in the selection of the DNA Pol responsible for DNA replication and the CIN-triggering mechanism. Not only p53 levels but also other genomic alterations can alter the endogenous p21 levels in cells, altering the replication rate and thereafter also affecting the levels of CIN [e.g., (62)]. Other molecules central to this study have well-documented roles in the genesis of cancer. Mutational signatures in cancer point toward TLS involvement in its genesis (8, 81, 82). Moreover, there is evidence of up- and down-regulation of Pol ι in cancer (32, 65, 72, 83), and p53 has a well-documented role in the prevention of chromothripsis (84). Thus, the implication of our findings is that endogenous p21 in unstressed cells is a master controller of replication dynamics, a process with strategic implications for the genomic stability of cells.

MATERIALS AND METHODS

Cell culture and chemicals

U2OS, HCT116, SKOV3, and H1299 cells were obtained from American Type Culture Collection. U2OS WT KO (clone 18), U2OS p21 KO (clone 1), U2OS Pol ι KO (clones 13 and 24), and RPE-1 p21 KO cells were generated by the CRISPR-Cas9 technology at The Francis Crick Institute and are a gift from J. Diffley. HT1080 and HT1080 CRISPR/Cas9 p21 KO (clone 1) cells were a gift from C. Prives (Columbia University). HCT116 p21 KO and HCT116 p53 KO cells were a gift from B. Vogelstein (John Hopkins). All cell lines were grown in Dulbecco’s modified Eagle’s medium (Invitrogen) with 10% fetal bovine serum (Natocor). Samples were minimally amplified and frozen after the reception and were used for limited passages after thawing of those primary stocks. None of the cell lines used were in the list of commonly misidentified cell lines maintained by the International Cell Line Authentication Committee updated at http://iclac.org/databases/cross-contaminations/. Samples are routinely tested for Mycoplasma by polymerase chain reaction (PCR). Daunorubicin (Sigma-Aldrich) and olaparib (Selleckchem) were used at the concentrations and times specified in the respective figures.

siRNAs and vector expression plasmids

Transfections of siRNAs and plasmid expression vectors were performed using jetPRIME (Polyplus) according to the manufacturer’s instructions. Cells were harvested 48 hours after transfection. siRNAs were purchased from Dharmacon or Eurofins Genomics, except for sip21#2, which was purchased from Ambion. siLuc (100 nM): 5′-CGUACGCGGAAUACUUCGA-3′ (24); sip21#1 (100 and 10 nM): 5′-GGAACAAGGAGUCAGACAU-3′ (24); sip21#2 (Ambion, s416) (9); siPrimPol (100 nM): 5′-GAGGAAACCGUUGUCCU CAGUGUAU-3′ (20); siPol κ (100 nM): 5′-AAGAUUAUGAAGCCCAUCCAA-3′ (85); sip53 (100 nM): 5′-GACUCCAGUGGUAAUCUA-3′ (10); siPol η (100 nM): 5′-UAAACCUUG UGCAGUUGUA-3′ (86). V5-PrimPol, V5-PrimPol RBDm, and GFP-Pol κ were gifts from J. Méndez [Spanish National Cancer Research Centre (CNIO), Madrid 28029, Spain] and A. Lehmann (38), respectively. V5-PrimPol and V5-PrimPol RBDm were cloned into the pLenti CMV/TO Puro vector (plasmid no. 17482, Addgene, third generation) to generate viruses and infect cells to obtain a more homogeneous V5-PrimPol expression.

Lentiviral infection

The p21 expression vector (CS2-p21) and p21^PIPMut [CS2-p21 with the M147A, D149A, and F150A mutations, which were shown to disrupt the PCNA/p21 interaction (29)] have three mutations in its CDK binding domain (W49R, F51S, and D52A), which disrupt the CDK2/p21 interaction (29). Primers used to generate p21 mutations were described and validated in our previous work (29). p21 and p21^PIPMut were cloned into the lentiviral transfer plasmid pLenti CMV/TO Puro. Lentivirus cloning, production, and infection were thoroughly described in (24). When transfection was required, cells were infected 5 hours after transfection. Immunodetection of p21 (1:1000, Santa Cruz Biotechnology, sc-6246, RRID: AB_628073) was conducted exactly as previously described (24). V5-PrimPol and V5-PrimPol RBDm lentiviruses were produced, and cells were infected exactly as with the p21 lentivirus. Nuclei were stained with 4′,6-diamidino-2-phenylindole (DAPI; Sigma-Aldrich). Images were acquired with a ZEISS Axio Observer 3 microscope and processed with ImageJ software (ImageJ 1.52a).

DNA fiber spreading

DNA fiber spreading was conducted exactly as previously described (24). 5′-Iododeoxyuridine (IdU) and CldU were simultaneously detected with mouse α-BrdU (Becton Dickinson, 347580, RRID: AB_400326, 1:50) and donkey α-mouse Alexa 546 (Molecular Probes, A10036, RRID: AB_11180613, 1:150) or rat α-BrdU (Abcam, ab6326, RRID: AB_305426, 1:500) and donkey α-rat Alexa 488 (Molecular Probes, A21208, RRID: AB_141709, 1:500), respectively. Images were acquired with a ZEISS LSM 5 Pascal confocal microscope and processed with ImageJ software (ImageJ 1.52a). Some control samples were shared as a reference by more than one experiment and thereafter used in more than one figure panel, which is clarified in figure legends when it applies.

S1 nuclease DNA fiber assay

DNA fiber spreading with the single-stranded DNA–specific S1 nuclease was performed as previously described (87). Images were acquired with a ZEISS Axio Observer 3 microscope and processed with ImageJ software (ImageJ 1.52a).

Western blotting

Cells were lysed and harvested with Laemmli buffer, followed by 8-min incubation at 99°C. The following antibodies were used: α-p21 1:1000 (Santa Cruz Biotechnology, sc-6246, RRID: AB_628073), α-p53 1:1:1 (phosphate-buffered saline:purified immunoglobulin from clone DO-1:purified immunoglobulin from clone PAB-1801), α-Pol η 1:1000 (Santa Cruz Biotechnology, sc-5592, RRID: AB_2167006), α-Pol κ 1:1000 (Bethyl Laboratories, A301-977A, RRID: AB_1548020), α-PrimPol 1:1000 (Proteintech 29824-1-AP, RRID: AB_2918349), and α-Actin 1:20000 (Sigma-Aldrich, A2066, RRID: AB_476693). Incubations with secondary antibodies (Jackson ImmunoResearch) and enhanced chemiluminescence detection (GE Healthcare) were performed according to the manufacturers’ instructions. Western blot images were acquired with an ImageQuant LAS4000 and ImageQuant 800 (GE Healthcare) or by x-ray films. In all Western blots of the manuscript, molecular weights are shown on the left (K is kDa).

Quantitative real-time PCR

Quantitative PCR was conducted exactly as previously described (85). Primer sequences were as follows: glyceraldehyde-3-phosphate dehydrogenase (GAPDH): 5′-AGCCTCCCGCTTCGCTCTCT-3′ (forward) and 5′-GAGCGATGTGGCTCGGCTGG-3′ (reverse) (85); PrimPol: 5′-TGTGGCTTTGGAGGTTACTGA-3′ (forward) and 5′-TTCTACTGAAGTGCCGATACTGT-3′ (reverse) (10); Pol κ: 5′-AGGCCAGGATTGATGACAAAGT-3′ (forward) and 5′-GGAAGGATTATTGCACTTGCCT-3′ (reverse). Threshold cycles of target genes and the reference gene were acquired with LightCycler 480 System (Roche). In all instances, the quantitative PCR graphs represent one of two independent experiments, each with three technical replicates.

Micronucleus assay

Twenty-four hours after transfection, cells were replated at low density. Twenty-four hours after replating, cytochalasin B (4.5 μg/ml; Sigma-Aldrich) was added to the media, and after 36 hours, cells were fixed with 4% paraformaldehyde/sucrose for 20 min. DAPI (Sigma-Aldrich) staining served to visualize nuclei. Only binucleated cells were measured per sample. Representative images were acquired with a ZEISS LSM 510 confocal microscope and processed with ImageJ software (ImageJ 1.52a).

Anaphase aberration assay

Asynchronous cells were fixed with 4% paraformaldehyde/sucrose for 20 min. DAPI (Sigma-Aldrich) staining served to visualize anaphases. z-Stacks were acquired with a ZEISS LSM 880 confocal microscope. Maximum intensity projections were generated using Black ZEN Imaging Software (ZEISS).

Immunostaining and fluorescence detection

Immunodetection of γH2AX (Millipore, 05-636, 1:1000, RRID: AB_309864) was conducted exactly as previously described (24). Nuclei were stained with DAPI (Sigma-Aldrich). Images were acquired with a ZEISS Axio Observer 3 microscope, and the intensity of relative γH2AX per each nucleus was analyzed by ImageJ software (ImageJ 1.52a). The detection of the chromatin-bound fraction of RPA (Calbiochem, NA-18, RRID: AB_10682810, 1:1000) and BrdU (Sigma-Aldrich, B9285, 1/500) was conducted exactly as previously described (10). Nuclei were stained with DAPI (Sigma-Aldrich). Images were acquired with a ZEISS Axio Observer 3 microscope, and the intensity of relative γH2AX per each nucleus was analyzed by ImageJ software (ImageJ 1.52a). Representative images were acquired with a ZEISS LSM 510 confocal microscope and processed with ImageJ software (ImageJ 1.52a).

Ultrafine bridge and 53BP1 nuclear body detection

Immunodetection of ultrafine bridges by PLK1-Interacting Checkpoint Helicase (α-PICH) (Abnova, H00054821-B01P, RRID: AB_1573438, 1:100) and 53BP1 nuclear bodies by α-53BP1 (Santa Cruz, sc-22760, RRID: AB_2256326, 1:1500) was conducted exactly as previously described (6). Nuclei were stained with DAPI (Sigma-Aldrich). Ultrafine bridges were measured in asynchronous cells. 53BP1 foci were measured in G₁ cells by discrimination of cells in the S phase. The S phase was identified through the administration of a 15-min pulse of 20 μM 5-ethynyl-2′-deoxyuridine (EdU), which was subsequently detected using the Click-iT EdU Alexa Fluor 555 Imaging Kit (Life Technologies), in accordance with the instructions provided by the manufacturer. Images of cells in the interphase were acquired with a ZEISS LSM 510 confocal microscope. Images of anaphase cells were acquired with a ZEISS LSM 880 confocal microscope. Images were processed with ImageJ software (ImageJ 1.52a).

V5-PrimPol, V5-PrimPol RBDm, and GFP-Pol κ focus detection

GFP-Pol κ foci were detected by GFP autofluorescence exactly as previously described (24). Immunodetection of V5-PrimPol and V5-PrimPol RBDm foci was detected by α-V5 (Bethyl, A190-120A, 1:750). For the microscopy analysis of V5-PrimPol foci, we extracted soluble proteins with 0.05% Triton X-100 in phosphate-buffered saline on ice for 2 min and 30 s. Nuclei were stained with DAPI (Sigma-Aldrich). Representative images were acquired with a ZEISS LSM 510 confocal microscope and processed with ImageJ software (ImageJ 1.52a).

Statistical analysis

GraphPad Prism 8 was used for statistical analyses and drawing all plots. Frequency distributions were analyzed with a one-way analysis of variance (ANOVA; followed by a Bonferroni post test), and data shown as the means (+SD) of independent experiments were analyzed with repeated-measures ANOVA (followed by a Newman-Keuls post test). In all plots, different letters indicate groups that are significantly different. Thus, if two samples share the same letter, they are not significantly different, while if two samples do not share any letter, they are significantly different. ***P < 0.001 was considered significant for dot plots and **P < 0.01 for the means of independent experiments. No power analysis was performed for the sample size estimation, albeit a rigorous empirical analysis was performed before deciding on the sample size for each experimental setting. For each experimental setting, the sample size was defined as the minimal number of events that provided a significant difference that did not vary when increasing the sample size. In addition, the sample size is in the same range as the one used in other manuscripts from other laboratories, which have evaluated those variables. The sample size for each experimental setting is detailed in figure legends and supplementary figure legends.

Supplementary Materials

The PDF file includes:

Figs. S1 to S11

Source Western blots

Source representative images

Legend for data S1

Other Supplementary Material for this manuscript includes the following:

Data S1