DNA damage promotes microtubule dynamics through a DNA-PK-AKT axis for enhanced repair

DNA double-strand breaks (DSBs) greatly threaten the integrity of eukaryotic genomes, and incorrectly repaired DSBs lead to chromosomal aberrations and genome instability. To counteract the deleterious effects of DSBs, two major DSB repair pathways exist, canonical nonhomologous end joining (c-NHEJ) and homologous recombination (HR; Jackson and Bartek, 2009; Lukas and Lukas, 2013). HR operates relatively slower and is restricted to the S and G2 phases during the cell cycle, when sister chromatids are available as repair templates. In contrast to HR, c-NHEJ is a relatively fast and efficient process and functions throughout the cell cycle. In G1, DSBs are mainly repaired by c-NHEJ. Key components in c-NHEJ are the Ku70/Ku80 heterodimer, which could form a complex at DNA breaks with the DNA-dependent protein kinase catalytic subunit (DNA-PKcs), generating the DNA-PK holoenzyme (Jette and Lees-Miller, 2015). In G1 phase, c-NHEJ shows biphasic kinetics involving a fast and a slow process in response to ionizing radiation (IR)–induced DSBs (Biehs et al., 2017; Löbrich and Jeggo, 2017). The DNA ligase 4 complex, including XRCC4, XLF, and PAXX, carries out the direct ligation step of the two broken DNA ends in the later stages of c-NHEJ (Biehs et al., 2017; Ochi et al., 2015). The nuclease Artemis does not involve the fast end joining but is required for the slow end resection–dependent process (Biehs et al., 2017; Riballo et al., 2004). Mre11 exonuclease, EXD2, and Exo1 are also required for this end resection–dependent slow NHEJ in G1 (Riballo et al., 2004). The slow NHEJ may contribute to the genomic instability in G1 (Biehs et al., 2017; Löbrich and Jeggo, 2017).

As DSBs are the most deleterious form of DNA damages, c-NHEJ and HR are highly regulated to avoid ectopic repair. End resection is required for HR in S or G2 cells, while the inappropriate resection in G1 impedes the initiation of the NHEJ repair process. 53BP1 is a crucial factor for c-NHEJ and limits the 5′ resection of the broken ends in a cell cycle–dependent manner. 53BP1-bound Rif1 and Rev7-shieldin complex executes the inhibition of 5′ end resection in G1 (Dev et al., 2018; Ghezraoui et al., 2018; Gupta et al., 2018; Mirman et al., 2018; Noordermeer et al., 2018; Xu et al., 2015). Interestingly, recent findings suggest that DSB-induced phosphorylation of CtIP by Plk3 in G1 could mediate CtIP-BRCA1 interaction, which regulates end resection–dependent slow c-NHEJ (Barton et al., 2014; Biehs et al., 2017; Löbrich and Jeggo, 2017). As both fast NHEJ and slow NHEJ contribute to the DSB repair in G1 cells, most DSBs should be repaired by fast NHEJ to avoid slow NHEJ–induced genomic instability. The underlying mechanism that regulates the choice between fast and slow NHEJ in G1 or G0 cells is still not clear.

DNA damage increases chromatin mobility, both locally at DSBs and genome wide (Hauer and Gasser, 2017). DSB mobility is regulated by several factors, including 53BP1, LINC (linker of nucleoskeleton and cytoskeleton) complex, microtubule, nuclear actin, Lamin A/C, and IFFO1 (Caridi et al., 2018; Lawrimore et al., 2017; Li et al., 2019; Lottersberger et al., 2015; Schrank et al., 2018). For instance, the increase of DSB mobility requires 53BP1 and dynamic microtubules, which act through the LINC complex and kinesins on the nuclear envelope (Lawrimore et al., 2017; Lottersberger et al., 2015). In G1, mobile DSBs could increase their exploration and promote end joining (Lottersberger et al., 2015). However, mobility of DSBs should be tightly regulated, as increased mobility of DSBs can also be a source of genomic translocation (Li et al., 2019; Roukos et al., 2013). As microtubule dynamics are one of the factors for DSB mobility (Lottersberger et al., 2015), the regulation of microtubule dynamics is crucial for DSB mobility and repair. Thus, we set out to study whether the microtubule dynamics will change after DNA damage and, if so, what is the underlying mechanism.

The centrosome is the major microtubule organizing center and comprises of a pair of centrioles and the surrounding pericentriolar material (PCM), which contains the key microtubule nucleation molecules, including γ-tubulin ring complex (Moritz et al., 2000; Zheng et al., 1995), NEDD1, and pericentrin (PCNT; Woodruff et al., 2014; Zhang et al., 2009). The centrosome undergoes a maturation process as cells progress toward mitosis, in which PCM increases in size and promotes its microtubule nucleation capacity (Palazzo et al., 2000). The communication between the centrosome and DNA damage response (DDR) has been reported in different conditions (Mullee and Morrison, 2016). Several DDR proteins, such as ATM, ATR, DNA-PKcs, CHK1 and CHK2 kinases, and the BRCA1 ubiquitin ligase complex, have been found at the centrosomes (Mullee and Morrison, 2016). Some centrosome proteins, such as centrin2 and PCNT, have been shown to stimulate DDR and are required for efficient nucleotide excision repair (Griffith et al., 2008; Mullee and Morrison, 2016; Nishi et al., 2005). DNA-damaging treatment causes significant alterations in centrosome structures and promotes centrosome duplication, leading to multicentrosomal cells (Bourke et al., 2007; Dodson et al., 2007; Löffler et al., 2013). Although the relationship between DNA damage and the centrosome has been widely studied, the short-term effect of DSBs on the centrosome and microtubule dynamics is still a mystery, as most of the studies focused on the long-term effect (>12 h) on the centrosome after DNA-damaging treatment. Considering the fast NHEJ process only lasts around 4 h (Löbrich and Jeggo, 2017), we examined the short-term effect of DSBs on the centrosome and microtubule dynamics.

In this study, we found that DSBs promoted microtubule dynamics in G1 or G0 phase cells. We named this specific microtubule response as DSB-induced microtubule dynamics stress response (DMSR). Alteration of DMSR affects the c-NHEJ process and leads to genomic instability.

To study the effect of DNA damage on the microtubule network, we first synchronized U2OS cells in G1 phase by a double thymidine treatment in which cells have relatively low microtubule dynamics. Synchronized cells were treated with 2 Gy IR and released for the indicated time points (Fig. S1 A). The centrosome-dependent microtubule polymerization rate was determined by microtubule length in a microtubule regrowth assay. Intriguingly, the rate of microtubule polymerization started to rise at 2 h and peaked at 4 h after IR. This effect gradually diminished at 6 h after IR (Fig. 1, A and B). The same effect of DSBs on microtubules was also observed in MCF7 (Fig. 1, C and D), HeLa cells (Fig. 1 E), and nontransformed RPE-1 (retinal pigment epithelium) cells (Fig. 1 F), indicating that the observations were a general response of microtubule dynamics to DNA damage. The DDR activation and DNA damage repair kinetics were determined by 53BP1 or γH2AX foci formation (Fig. 1 G), which represented DSB sites in the nucleus. We hypothesized that the increased microtubule polymerization rate was caused by IR-induced DSBs. To test our hypothesis, we treated cells with bleomycin, which is a radiomimetic agent that causes DSBs directly (Robles and Adami, 1998). The same effect was observed in G1 cells treated with bleomycin (Fig. 1 H). On the contrary, UV treatment, which does not directly lead to DSBs (Rastogi et al., 2010), did not affect the microtubule nucleation in G1 cells. These results suggested that the DNA damage–induced promotion of microtubule polymerization was DSB specific. As the centrosome in G1 phase cells normally has low microtubule nucleation activity and the effect lasts only around 4–6 h after DNA damage treatment (Fig. 1, B and F), we speculate that this effect was a DMSR accompanied by a short-term increase of centrosome-dependent microtubule polymerization. To examine whether DMSR was a stress response, we treated G1 phase cells with different doses of IR and found that, although DMSR could be observed when cells were treated with 1 Gy IR, the extent of DMSR rose significantly when we elevated the dose of IR treatment (Fig. 1 I). The same effect was also found in MCF7 cells (Fig. 1, C and D). These results implied that DMSR is a DSB dose-dependent stress response in the microtubule network, which lasts around 4–6 h after DNA damage treatment and could be only induced by DSBs.

Most microtubules originate from the centrosome, and microtubule regrowth assay is normally used to determine the extension rate of microtubule polymerization from the centrosome. As the rate of microtubule polymerization is distinct from centrosome-dependent microtubule nucleation capacity, we studied the relationship between DMSR and these two aspects of microtubule dynamics. From the inverted grayscale images of microtubule regrowth assay (Fig. S1, B and C), we observed that IR treatment clearly led to an increased number of microtubules that originated from the centrosome, although the exact number of microtubules in each time point is hard to quantitate. Due to the difficulty in quantifying the microtubule numbers in microtubule regrowth assay, we employed RPE-1 cells exogenously expressing GFP-tagged EB3 (end-binding protein 3) to check microtubule dynamics. GFP-EB3 served as a marker for the plus-end tips of each growing microtubule (Komarova et al., 2005). Thus, the mobility of the GFP-EB3 signal indicates the extension rate of microtubule polymerization, while the intensity of GFP-EB3 signals that originated from the centrosome within a specific period could be used to measure the centrosome-dependent microtubule nucleation capacity.

Using RPE-1 cells expressing GFP-EB3, the effect of DMSR on the extension rate of microtubule polymerization was first confirmed by time-lapse fluorescent imaging experiment (Fig. 2, A and B; Fig. S1 D; and Videos 1, 2, and 3). Representative GFP-EB3 comet tracks are presented in different colors for each treatment, and, obviously, the comet tracks in bleomycin- or IR-treated cells are longer than the tracks in untreated cells (0–15 s; Fig. 2 A). The measurement of the velocity of EB3 comets further supported that the mobility of GFP-EB3 increased significantly in IR- or bleomycin-treated cells (Fig. 2 B), indicating DSBs lead to increased rates of microtubule polymerization.

As GFP-EB3 locates on the microtubule plus end, GFP-EB3 was used to track the newly nucleated microtubules and to determine the emanation rate from the centrosome (Colello et al., 2012). Thus, we analyzed the GFP-EB3 signals originating from the centrosome within 30 s through a live-imaging time-lapse experiment (Fig. 2 C; Videos 4, 5, 6, and 7). In total, 30 images were recorded for every 1 s and overlaid in one image. The centrosomal GFP-EB3 intensity in the overlaid image was used to indicate the number of newly nucleated microtubules that originated from the centrosome within 30 s (Fig. 2 C, right). Remarkably, IR treatment led to increased centrosomal intensity of GFP-EB3, which started from 1 h and, with continuous effect, lasted until 4 h after IR (Fig. 2, C and D), suggesting that IR treatment increased the centrosome-dependent microtubule nucleation capacity. Bleomycin treatment displayed the same enhanced intensity of centrosomal GFP-EB3 (Fig. 2 E; Videos 8 and 9). These data demonstrated that DMSR is accompanied with both increased a microtubule polymerization rate and enhanced microtubule nucleation capacity.

Next, we asked whether DMSR occurred in another cell cycle phase besides G1. First, we examined DMSR in G0 cells. U2OS cells were synchronized in G0 by serum starvation and then treated with bleomycin for 2 h. After release from bleomycin treatment for the indicated time points, DMSR was checked with the microtubule regrowth assay (Fig. S2 A). Quantitative analysis of microtubule length confirmed that DMSR also happened in G0 cells and that DMSR lasted for around 4–6 h after DNA damage (Fig. 3, A and B). The DNA damage repair kinetics after bleomycin treatment were determined by 53BP1 foci formation (Fig. 3 C). We observed the same microtubule dynamics in G0 cells after IR-induced DSBs (Fig. 3 D). To determine whether DMSR happened in S phase cells, U2OS cells were released from double thymidine block for 3 h or 5 h to allow cells to enter S phase. The procedure for this experiment was interpreted in Fig. S2 B, and the cell cycle stage was determined by flow cytometry. DMSR was then examined by microtubule regrowth assay 2 h after IR. From the flow cytometry data, we found that cells were still in G1 phase after being released from IR for 2 h (Fig. S2 B), and, accordingly, DMSR occurred as indicated by increased microtubule length (Fig. 3 E, lane 2). Intriguingly, the extent of DMSR significantly decreased in S phase cells when the cells were released from double thymidine block for 3 h, and DMSR totally disappeared when cells were released for 5 h from double thymidine block (Fig. 3 E).

As a complementary approach, we also analyzed DMSR in asynchronously dividing cells. DMSR was determined by microtubule regrowth assay at 1 h or 2 h after IR. To differentiate G1 from S/G2 cells, we coimmunostained β-tubulin with cyclin A in microtubule regrowth assay, which is restricted to S/G2 (Fig. 3 F; Escribano-Díaz et al., 2013). Quantitative analysis of microtubule length indicated that DMSR only happened in cyclin A–negative cells (Fig. 3 G). These results suggest that DMSR is restricted in G0 or G1 cells.

As c-NHEJ is the predominant DSB repair pathway in G1 or G0 cells, we hypothesized that the c-NHEJ pathway may participate in DMSR. Consistent with our hypothesis, the DNA-PK inhibitor, but not ATM or ATR inhibitor treatment, abolished DMSR (Fig. 4 A and Fig. S3 A). We got the same result with DNA-PKcs depletion (Fig. 4 B). The Ku70/Ku80 heterodimer is required for DNA-PK activation and c-NHEJ initiation. DMSR disappeared after depletion of Ku70 or Ku80, which further confirmed the role of DNA-PK in DMSR (Fig. 4 B). As DNA-PKcs autophosphorylation on Ser2056 could be a marker for DNA-PK activation, we probed the whole U2OS cell extracts with an anti-pS2056 DNA-PKcs antibody and found that activation of DNA-PK was observed 2 h after IR and gradually decreased from 6 h (Fig. S3 B), which was consistent with the time course of DMSR. The efficiency of DNA-PKcs, Ku70, or Ku80 depletion was checked by Western blot or quantitative PCR (Fig. S3 C). All these results indicate that DNA-PK activity is important for DMSR.

On the contrary, depletion of Artemis, another key factor in NHEJ, did not affect DMSR positively or negatively (Fig. 4 B). Previous reports showed that IR-induced DSBs are repaired by fast and slow c-NHEJ processes in G1 phase cells, and Artemis nuclease is specifically required for the slow but not fast NHEJ process (Biehs et al., 2017; Löbrich and Jeggo, 2017). The slow c-NHEJ process in G1 phase cells is accompanied by end resection, which depends on CtIP, EXD2, EXO1, and NBS1 (Biehs et al., 2017). Depletion of these proteins did not have an effect on DMSR (Fig. S3 D), suggesting that end resection–dependent slow NHEJ was dispensable for DMSR.

DSBs could be repaired by c-NHEJ or HR, and the choice was tightly regulated, especially in G1 cells, since there is no DNA template required for HR repair in G1. 53BP1 and Shieldin complex promotes c-NHEJ and inhibits HR (Dev et al., 2018; Findlay et al., 2018; Ghezraoui et al., 2018; Gupta et al., 2018; Mirman et al., 2018; Noordermeer et al., 2018; Tomida et al., 2018). Thus, we examined the effect of 53BP1 and the Shieldin complex on DMSR and found that depletion of 53BP1 or Shieldin complex, such as with FAM35A, REV7/RINN2, or RINN1, could affect DMSR (Fig. 4 C; and Fig. S3, E and F), indicating that 53BP1 and Shieldin complex were involved in DMSR.

The complex of Ligase 4, XRCC4, and XLF is responsible for direct DSB ligation, which is the final step of c-NHEJ (Pannunzio et al., 2018). To further study the effect of c-NHEJ on DMSR, we knocked down Ligase 4, XRCC4, or XLF by siRNA to sustain the active c-NHEJ process and found that depletion of these proteins significantly enhanced the extent of DMSR compared with scrambled siRNA–treated cells (negative control [NC]; Fig. 4 D; and Fig. S3, G and H). Furthermore, DMSR could still be obviously observed at 8 h after IR in siLigase 4–treated cells (Fig. 4 E), indicating that DMSR was prolonged. Consistent with this result, DNA damage repair kinetics in Ligase 4–depleted cells were greatly delayed (Fig. S3 I), indicating that accumulation of DSBs in IR-treated Ligase 4–depleted cells may enhance DMSR. Meanwhile, we found that in XLF, XRCC4, or Ligase 4–depleted cells without IR treatment, microtubule dynamics were higher than in control U2OS cells (Fig. 4, D and E), indicating that accumulated DSBs caused by inactive ligation might continuously activate the NHEJ pathway and cause sustained DNA-PK activation, which further leads to increased microtubule dynamics. These results demonstrated that prolonged c-NHEJ caused by Ligase 4, XRCC4, or XLF depletion enhances and prolongs DMSR.

Using time-lapse fluorescent imaging experiments, the velocity of the GFP-EB3 track was also measured in cells treated with scrambled siRNA, siLigase 4, or si53BP1. The velocity of EB3 comets increased in cells treated with scrambled siRNA after IR. The extent of the increase of the DSB-induced velocity was further enhanced in cells treated with siLigase 4. On the contrary, we did not observe obvious DSB-induced changes in the velocity of GFP-EB3 in si53BP1-treated cells (Fig. 4 F). The centrosomal GFP-EB3 intensity was determined as in Fig. 2 C and was also increased in siLigase 4–treated cells but not in si53BP1-treated cells (Fig. 4 G), indicating that centrosome-dependent microtubule nucleation capacity was enhanced in siLigase 4 but not in si53BP1 cells. All these results suggested that DMSR may be activated when fast NHEJ fails to repair DSBs.

As the centrosome is the major microtubule organization center, we further investigated the effect of DSBs on the centrosome to explore the underlying mechanism of DMSR. We observed a dramatic accumulation of centrosome-associated proteins on the centrosome in G1 cells at 4 h after IR, including PCMs, such as PCNT, γ-tubulin, and NEDD1 (Fig. 5 A). Intriguingly, DMSR was abolished after NEDD1 or PCNT depletion (Fig. 5 B and Fig. S4 A). PCNT-depleted cells remained in G1 phase at 4 h post-IR as determined by flow cytometry (Fig. S4 B). These results indicate that DSBs induce DMSR through the accumulation of PCM proteins at the interphase centrosome.

The accumulation of PCM proteins on the centrosome is typically a hallmark of centrosome maturation during G2 phase and mitosis. The normal centrosome maturation process in G2 cells depends on Polo-like kinase 1 (PLK1) and results in an increased ability of the centrosome to nucleate microtubules (Barr et al., 2004). To examine whether DMSR is operated through the classical centrosome maturation mechanism, we inhibited PLK1 with a PLK1 inhibitor and found there were no detectable changes in DSB-induced PCM protein accumulation at the centrosome (Fig. 5 C and Fig. S4 C), and DMSR was also not affected (Fig. 5 D), suggesting that DSB-induced PCM recruitment at the interphase centrosome depends on different pathways.

Based on the fact that DSBs lead to accumulation of PCMs at the interphase centrosome and an elevated microtubule polymerizing rate in G1 or G0 cells, we uncovered a novel DSB-induced interphase centrosome maturation process. Actually, this interphase centrosome maturation induced by DSBs is similar to bacterial lipopolysaccharide (LPS)–induced centrosome changes (Vertii et al., 2016). P38 and JNK MAPKs have been implicated in the interphase centrosome maturation in LPS-stimulated cells. To examine whether these two processes share a similar mechanism on interphase centrosome maturation, we chemically inhibited P38 and did not observe an obvious effect on DMSR (Fig. 5 D and Fig. S4 C), suggesting that DSB-induced interphase centrosome maturation may have a distinct mechanism from LPS-induced centrosome maturation.

As we found that DNA-PK and c-NHEJ are required for DMSR, we next examined the function of DNA-PK and c-NHEJ on PCM recruitment at the centrosome. Although DNA-PK has been reported as an important regulator of mitotic spindle formation (Shang et al., 2010), the role of DNA-PK on the interphase centrosome is still unknown. DNA-PK inhibitor (Fig. 5 E) and 53BP1 depletion (Fig. 5 F) affected PCM recruitment at the centrosome after IR, indicating that c-NHEJ may participate in DSB-induced interphase centrosome maturation. The major subdistal appendage proteins, Ninein and CEP170, which are crucial for microtubule anchoring, also dramatically accumulated at the interphase centrosome after IR (Fig. S4 D). DNA-PK inhibition or 53BP1 knockdown abolished the DSB-induced accumulation of Ninein and CEP170 (Fig. S4 E), indicating that DSBs induce comprehensive changes on the interphase centrosome through DNA-PK, which leads to DMSR.

Next, we aimed to uncover the downstream factor of DNA-PK on DMSR. Protein kinases, such as CHK1, CHK2, and AKT, which regulate the centrosome structure or functions, have been reported as downstream factors of DNA-PK (Bozulic et al., 2008; Buttrick et al., 2008; Li and Stern, 2005; Lin et al., 2014; Löffler et al., 2007; Wang et al., 2015). CHK1 and CHK2 are key components of DNA damage–activated checkpoint signaling response (Bartek and Lukas, 2003). AKT is the key factor in the phosphoinositide 3-kinase (PI3K) pathway and regulates the activation of the major signals for cell growth, survival, and metabolism (Carnero et al., 2008). To determine which factor is the downstream of DNA-PK in DMSR, G1 cells were pretreated with AKT, CHK1, or CHK2 inhibitor before IR treatment. DMSR was examined by microtubule regrowth assay 4 h after IR. AKT inhibition, but not CHK1 or CHK2 inhibition, resulted in an obvious decrease of DMSR (Fig. 6 A). Depletion of AKT affected DMSR (Fig. 6, B and C) and the accumulation of PCM proteins at the interphase centrosome after IR (Fig. 6 D), suggesting that AKT might be the downstream factor of DNA-PK after IR. We also confirmed the effect of AKT depletion on DMSR by a time-lapse live-imaging experiment and found that loss of AKT led to decreased velocity of EB3 comets (Fig. 6 E) and the microtubule nucleation capacity of the centrosome (Fig. 6 F).

The phosphorylation of two key residues on AKT—Thr308 in the T-loop of the catalytic protein kinase core and Ser473 in a C-terminal hydrophobic motif—are required for AKT activation (Alessi et al., 1996). To determine the mechanism of AKT activation in DMSR, we separated cells into cytoplasmic and nuclear fractions. Nuclear pSer473 AKT increased after IR, which was consistent with previous reports. Intriguingly, in cytoplasmic fraction, pThr308 AKT, but not pSer473 AKT, significantly increased at 4 h after IR in G1 phase cells (Fig. 6 G). Cytoplasmic pThr308 signal increased at 2 h, peaked at 4 h, and started to decrease at 6 h after IR (Fig. 6 H), which was consistent with the time course of DMSR. Furthermore, DSB-induced cytoplasmic AKT Thr308 phosphorylation was abrogated by DNA-PK inhibition (Fig. 6 I). As PDK1 has been reported as the main protein kinase for pThr308 phosphorylation, we chemically inhibited PDK1 and found that, although the basal level of pThr308 AKT significantly decreased, the DSB-induced increase of cytoplasmic pThr308 AKT still existed (Fig. 6 J), indicating that DSB-induced pThr308 AKT is PDK1 independent. PDK1 inhibition did not affect DMSR (Fig. 6 K), implying that PDK1 is dispensable for DMSR. These results suggest that AKT could be the downstream factor of DNA-PK during DMSR.

As the importance of centrosome maturation on DMSR, we speculate that Thr308 phosphorylation may regulate DSB-induced interphase centrosome maturation and DMSR. Given that AKT has been reported to locate at the centrosome (Buttrick et al., 2008; Buttrick and Wakefield, 2008; Wakefield et al., 2003), we analyzed the effect of DMSR on the centrosomal localization of AKT. Although we could observe the centrosomal localization of AKT in RPE-1 cells, we encountered difficulties in quantitating the centrosomal AKT intensity due to high background of AKT immunofluorescence caused by the universal distribution of AKT in cytosol (Fig. S5 A). Thus, we could not draw a conclusion on the effect of DSBs on the accumulation of AKT on the centrosome. pT308 AKT has been reported to accumulate on mitotic centrosomes and regulate spindle assembly (Wakefield et al., 2003). Thus, we analyzed whether pThr308 AKT localized on the centrosome in interphase cells and found that pT308 AKT clearly located on the interphase centrosome and the signal increased at 4 h after IR in G1 cells (Fig. 7, A and B). Furthermore, DNA-PK inhibition could abolish the increase of pT308 AKT on the centrosome after IR, implying that DMSR resulted in the increased accumulation of pT308 AKT on the centrosome (Fig. 7, A and B). The fact that depletion of AKT eliminated the pT308 AKT signal on the centrosome (Fig. S5 B) suggested the specificity of the interphase centrosomal pT308 AKT signal. Although PDK1 inhibition alone did not affect the DNA damage–induced accumulation of pT308 AKT signal on the interphase centrosome, PDK1 inhibition combined with DNA-PK inhibition greatly impaired the centrosomal localization of pT308 AKT (Fig. 7, C and D), indicating that DNA damage–induced accumulation of pT308 AKT on the interphase centrosome could be regulated by DNA-PK.

To explore how DNA-PK regulates the accumulation of pT308 AKT on the centrosome, we analyzed the cytoplasmic and nuclear distribution of DNA-PKcs and pS2056 DNA-PKcs after IR and found that both of them gradually increased after IR in the cytoplasmic portion of RPE-1 (Fig. 8 A) and HeLa cells (Fig. 8 B). Intriguingly, DNA damage–induced cytoplasmic pS2056 DNA-PKcs diminished in 53BP1-depleted cells, suggesting that 53BP1 may promote DMSR through regulating the cytoplasmic distribution of activated DNA-PK. DMSR was affected when protein export was blocked by leptomycin B treatment (Fig. S5 C), implying the transportation of activated DNA-PK from nuclear to cytoplasm may be involved in DMSR. Moreover, we found that 53BP1 depletion abolished DNA damage–induced accumulation of pT308 AKT on the interphase centrosome, which is consistent with its role in the cytoplasmic distribution of activated DNA-PK (Fig. 8, C and D).

All these results implied that AKT was involved in DMSR and that DNA-PK may contribute to DSB-induced centrosomal accumulation of pThr308 AKT in G1 cells.

Microtubule dynamics promote DSB mobility through kinesins and the LINC complex, which is important for NHEJ repair (Lottersberger et al., 2015), implying that DMSR may function in DSB repair through regulating DSB mobility. First, we studied the role of DMSR on DSB mobility. As PCM proteins are important for DMSR, we depleted PCM proteins, such as PCNT and NEDD1, to abrogate DMSR and examined DSB mobility in these cells. As 53BP1 Tudor domain (TD) foci (1220–1711 aa) have been widely used to measure the dynamics of DSBs (Li et al., 2019; Zgheib et al., 2009), we measured the mobility of GFP-53BP1 TD foci at 1 h after IR in G1 cells with indicated treatments. Nocodazole treatment, which could disassemble the microtubule network, significantly affected DSB mobility as shown by the ensemble mean-square displacement (eMSD) of 53BP1 foci in HeLa cells, which is consistent with a previous report (Fig. 9 B; Lottersberger et al., 2015). Interestingly, PCNT or NEDD1 knockdown decreased the DSB mobility as with nocodazole treatment (Fig. 9, A and B; and Videos 10 and 11), indicating that PCM proteins may contribute to DSB mobility in G1 cells. These results demonstrated that PCM protein–dependent DMSR may regulate DSB mobility.

As DSB mobility contributes to the high efficiency of NHEJ, we speculated that DMSR contributed to c-NHEJ repair. As HeLa cells were used in previous reports to study end resection–dependent slow NHEJ (Barton et al., 2014; Biehs et al., 2017), we employed HeLa cells in the following experiments to study the role of DMSR on c-NHEJ. First, we examined the NHEJ efficiency in G1 cells by γH2AX immunofluorescence and found that, when cells were depleted with PCNT or NEDD1, the percentage of cells with >10 γH2AX foci was higher than that of control cells after IR, indicating that NHEJ efficiency was affected by PCNT (Fig. 9 C) or NEDD1 (Fig. S5 D) depletion. These results were confirmed by 53BP1 foci formation analysis in shPCNT and shNEDD1 G1 HeLa cells (Fig. 9 D and Fig. S5 E). c-NHEJ in G1 phase cells comprise a fast and a slow process and the latter depends on end resection. To determine whether the end resection process in slow NHEJ was also affected by PCNT depletion, RPA2, which coats single-strand DNA after end resection, was stained in synchronized G1/S HeLa cells at 2 h after IR. Intriguingly, the percentage of cells with RPA2 foci increased in PCNT-depleted (Fig. 9 E) or NEDD1-depleted cells (Fig. S5 D), indicating that, when fast NHEJ was affected, more DSBs were processed by end resection. These results suggest that, when DMSR was abolished through PCM protein depletion, the extent of end resection–dependent slow NHEJ increased in G1 cells. As slow NHEJ will cause genomic instability, we next examined the effect of PCM protein depletion on genomic instability. HeLa cells treated with scrambled shRNA, shPCNT, or shNEDD1 were synchronized in G1/S and exposed to 2 Gy IR, then released from double thymidine block. 8 h after IR, the cells were treated with nocodazole for 45 min to accumulate mitotic cells. Collected mitotic cells were analyzed by chromosome spread assay to calculate chromosome breaks. NEDD1 or PCNT depletion did not cause obvious chromosomal breaks in cells without IR treatment (Fig.9 F and Fig. S5 F). Remarkably, IR treatment led to more chromosomal breaks in NEDD1- or PCNT-depleted cells than shcon (negative control scrambled shRNA) cells, indicating that DMSR may contribute to maintain genomic stability by facilitating c-NHEJ repair.

In this study, we uncovered that DSBs actively promote microtubule dynamics in G0/G1 cells through DMSR. DMSR is accompanied by PCM accumulation and interphase centrosome maturation, which required functional c-NHEJ, DNA-PK, and AKT. The 53BP1-Shieldin complex may also regulate DMSR through facilitating the activation of DNA-PK during c-NHEJ. DSB-induced interphase centrosome maturation leads to increased centrosome-dependent microtubule nucleation and polymerization, which could promote DSB mobility and facilitate c-NHEJ. Thus, we reveal a closed feedback loop between DSBs and microtubule dynamics during DSB repair (Fig. 10).

Several studies have reported the relationship between the centrosome and DNA damage. Most of them focused on the long-term effect of DNA damage, which induces centrosome overduplication (Antonczak et al., 2016; Bourke et al., 2007; Dodson et al., 2004; Löffler et al., 2013; Mullee and Morrison, 2016; Sugihara et al., 2006). In this study, we uncovered DSB-induced short-term (within 6 h)– and cell phase (G1 or G0)–specific effects on the centrosome. Previous notions suggest that abnormal centrosomes lead to abnormal spindle assembly and chromosomal separation in mitosis, which causes genomic instability (Mullee and Morrison, 2016). For example, PCNT mutations lead to genomic instability, and NEDD1 depletion results in cell senescence (Antonczak et al., 2016; Griffith et al., 2008; Manning and Kumar, 2010). Our study discovered that, during DMSR, both the rate of microtubule polymerization and the capacity of centrosome-dependent microtubule nucleation were promoted through DSB-induced accumulation of PCM proteins on centrosomes, which could promote DSB mobility and facilitate c-NHEJ repair. Thus, our new findings may provide a new explanation for these phenomena caused by centrosome protein defects.

DMSR is a short-term stress response and only occurs in G0/G1 cells, implying that DMSR is highly regulated. First, DMSR is restricted to 6 h after DNA damage. Microtubules are important cytoskeletons and usually serve as cargo transportation roads (Ross et al., 2008). Microtubules also play important roles in cell migration and the organization of many cellular components, including the ER and Golgi apparatus (Gurel et al., 2014). Prolonged changes in microtubule dynamics may affect the normal functions of these cellular apparatuses. Thus, the duration of DMSR should be restricted to a limited time period to avoid deleterious side effects. Second, DMSR happens specifically in G0/G1 cells, during which c-NHEJ is the predominant DSB repair pathway. Centrosome-dependent microtubule polymerization and nucleation capacity is weak in G1, while massive centrosome maturation happens at the G2/M transition for the following spindle formation in mitosis (Barr et al., 2004). In G1 cells, DSB-induced interphase centrosome maturation may be required for properly sustaining DSB mobility during c-NHEJ. In S/G2 phase, HR coexists with NHEJ, and elevated DSB mobility may cause genomic translocation (Li et al., 2019; Roukos et al., 2013). Thus, DMSR wisely disappears in S or G2 cells to balance the efficiency of NHEJ and accuracy of HR. Third, the mechanism of DSB-induced interphase centrosome maturation is different from the centrosome maturation in the G2/M transition. For instance, PLK1 is dispensable for DMSR. PLK1-dependent centrosome maturation may gradually become predominant from the G1 to G2 phase and compete with DNA-PK-AKT–mediated centrosome maturation. This may also explain why DMSR gradually disappears when cells enter S phase.

DSBs in G1 can be repaired by NHEJ in two ways: the fast NHEJ, which is an end resection–independent process, and slow NHEJ process, which is end resection dependent (Löbrich and Jeggo, 2017). Depletion of CtIP, EXO1, or Artemis, which is associated with slow NHEJ, does not affect DMSR, indicating that slow NHEJ is dispensable for DMSR. Instead, depletion of Ligase 4, XLF, or XRCC4, which are crucial for fast NHEJ, leads to persistent DMSR in G1, suggesting that a prolonged c-NHEJ process leads to overactivated DMSR. DMSR appears at around 1 h after DNA damage and diminishes at around 6 h, which overlaps with the fast NHEJ repair period (1–4 h after DNA damage; Löbrich and Jeggo, 2017). Furthermore, loss of DMSR by PCM protein depletion leads to increased end resection in G1, implying that DMSR may inhibit the end resection process in slow NHEJ by elevating DSB mobility. We hypothesize that, when fast NHEJ failed to repair DSBs properly in G1 cells, DMSR could be activated to inhibit end resection and avoid slow NHEJ. Thus, DMSR may be important for maintaining genome stability in G1 cells through facilitating fast NHEJ.

DMSR is accompanied by pT308 AKT accumulation on interphase centrosomes. Although the localization of AKT and pT308 AKT on mitotic centrosomes has been reported (Buttrick et al., 2008; Buttrick and Wakefield, 2008; Wakefield et al., 2003), the localization of pT308 AKT on the interphase centrosome during DMSR is a novel finding. Interestingly, PDK1 is dispensable for DSB-induced interphase centrosomal accumulation of pThr308 AKT, which is consistent with the report that the localization of pT308 AKT on mitotic centrosomes is PI3K independent (Wakefield et al., 2003). Previous reports showed that both DNA-PK and ATM may contribute to Ser473 AKT phosphorylation (Bozulic et al., 2008; Fraser et al., 2011). Although we found that DNA-PK may contribute to the DMSR-related cytoplasmic pThr308 AKT, whether DNA-PK directly phosphorylates AKT on Thr308 is still unknown. It is also possible that DNA-PK may regulate the cytoplasmic distribution of pT308 AKT through phosphorylating AKT-binding proteins, or DNA-PK may phosphorylate AKT at other sites that facilitate the centrosomal accumulation of pT308 AKT. Further studies are needed to illustrate the relationship between DNA-PK and AKT during DMSR. As to how pT308 AKT regulates centrosome functions, several studies have reported that AKT could phosphorylate several substrates, including GSK-3 (Buttrick et al., 2008; Buttrick and Wakefield, 2008; Wakefield et al., 2003), TEIF (Telomerase transcriptional element-interacting factor; Zhao et al., 2014), and Inversin (Suizu et al., 2016), to regulate interphase or mitotic centrosomes functions, respectively. Whether these AKT substrates are involved in DMSR and what is the exact role of pT308 AKT in DMSR still need further studies.

53BP1 has been proposed to promote DSB mobility to facilitate c-NEHJ repair, but the mechanism is still missing (Dimitrova et al., 2008; Lottersberger et al., 2015). In our study, we found that DNA-PK, 53BP1, and the Shieldin complex were important regulators of DMSR. Inhibition of DNA-PK abolished the S2056 phosphorylation of human DNA-PKcs, suggesting S2056 is the autophosphorylation site (Chen et al., 2005). Thus, pS2056 has been widely used as the marker for DNA-PKcs activation. By probing pS2056 DNA-PKcs, we found that DNA-PK activation was delayed in both cytoplasmic and nuclear fractions in 53BP1-depleted G1 HeLa and RPE-1 cells after IR. Thus, 53BP1 may contribute to DMSR through facilitating DNA-PK activation. We hypothesize that the increase of the 5′ end resection in 53BP1-depleted G1 cells may switch off DNA-PK and subsequently DMSR, which might lead to precocious activation of slow NHEJ. We also found that, although at a relatively lower level, the activation of DNA-PK (pS2056) still could be observed in 53BP1-depleted G1 cells after IR. One possibility is that when 53BP1 is depleted, end resection happens in a subset of DSBs, and DNA-PK still could be activated by the remaining unresected DSBs, which leads to a decreased DNA-PK activation and defective DMSR. Interestingly, although 53BP1/Rif1/Shieldin could inhibit 5′ end resection at DSBs to promote c-NHEJ, 53BP1, but not Rif1 and Rev7, is required for DSB mobility (Boersma et al., 2015; Zimmermann et al., 2013). Meanwhile, Dimitrova et al. (2008) reported that both ATM and 53BP1 are required for the increased telomere mobility after their deprotection. On the contrary, our data show that DMSR depends on DNA-PK but not ATM, and DMSR also could be affected by Rev7 depletion. These discrepancies suggest that DSB mobility may be regulated by several different pathways, and more evidence is needed to illustrate the regulation and function of DSB mobility in the future.

Another remaining question is how signals from nuclear DDR are transported to the cytoplasmic centrosome. DNA-PKcs could locate in both the nucleus and cytoplasm (Poruchynsky et al., 2015; Saji et al., 2005; Wang et al., 2013). It is possible that activated DNA-PKcs is exported from nucleus to cytoplasm after IR and subsequently transported to the centrosome through microtubules during DMSR. In fact, microtubule localization of DDR-related proteins, including DNA-PKcs, has been observed (Poruchynsky et al., 2015). Disruption of nuclear export by leptomycin B treatment abolishes DMSR, implying that nuclear export might play a role in DMSR.

Gene transcription in normal cells could induce massive endogenous DSBs, which are mainly repaired by c-NHEJ (Dellino et al., 2019). The elegantly regulated DMSR revealed an underlying relationship between DSBs and the centrosomes in quiescent (G0) and nontransformed (RPE-1) interphase cells. Aberrant centrosomes may affect c-NHEJ repair of endogenous DSBs generated from transcription in normal cells or end-differentiated cells, which may lead to cancer-associated genome instability. Thus, our findings may shed light on the explanation of the cancerous transition of end-differentiated cells.

Cells were grown on coverslips and treated with cold medium on ice for 30 min to depolymerize microtubules. Microtubule regrowth was allowed in prewarmed medium for 1 or 2 min (as indicated in figures) in 37°C. Cells were fixed with PHEM buffer (Pipes 60 mM, Hepes 25 mM, EGTA 10 mM, MgCl₂ 2 mM, pH 6.9) containing 4% PFA and 0.5% Triton X-100 for 15 min at room temperature. After washing several times with PBS, cells were stained with anti–β-tubulin (Frbio; 1:200) antibody. Images were gathered through Z-stack by Olympus IX83 microscopy and Andor’s Zyla 5.5 camera, and deconvolution analysis was done by cellSens software. Z-stacks at 0.34-µm steps were acquired by Olympus with 60× oil objective lens (NA, 1.35). For measuring the microtubule length in each microtubule regrowth experiment, three microtubules in each cell were counted and the average value was used as the microtubule length for the cell. More than 50 cells were measured for each sample in all the experiments. Quantitation of the microtubule length in each experiment was repeated by three members in the laboratory and we got the same tendency. The quantitative results from one person were shown in each figure. All statistical analysis was done using GraphPad Prism 8 software.

For centrosome-associated protein immunofluorescence, U2OS cells were grown on glass coverslips and fixed and permeabilized in methanol for 5 min. For phospho-AKT (Thr308), γH2AX, and RPA2, U2OS cells were permeabilized with 0.5% Triton X-100 for 5 min, then fixed in methanol for 5 min. The fixed coverslips were incubated with primary antibodies in PBS with 1% BSA at 37°C for 1 h, then washed three times with PBS. Cells were then incubated with secondary antibodies for 1 h at 37°C and stained with Hoechst33342 or DAPI. Digital images were captured on an Olympus IX83 microscopy with 60× oil objectivelens (NA, 1.35) and Andor’s Zyla 5.5 sCMOS camera and cellSens Dimension software in the same exposure time in each experiment. For quantitation of the centrosome-related protein fluorescence intensity, the fluorescence image was registered and converted to gray 8-bit, and centrosome particles were measured by using the Particles Analysis function of ImageJ. The concentration of primary antibodies we used for immunofluorescence is: anti–β-tubulin (Fribo; 1:200); anti–γ-tubulin, anti-RPA2, and anti-DNA-PKcs (Proteintech; 1:100); anti-NEDD1 (Abcam; 1:50); anti-pericentrin and γH2AX (Abcam; 1:500); anti-CEP170 (Proteintech; 1:200); anti-NINEIN (ABclonal; 1:300); anti-53BP1 (Cell Signaling Technology [CST]; 1:500); anti-AKT1 (Proteintech; 1:50); and anti-pT308 AKT, anti-pT473 AKT, and anti-pS8-RPA2 (CST; 1:100). The secondary antibodies TRITC-Rb/M and FITC-Rb/M were purchased from Jackson ImmunoResearch, and the using concentration is 1:300.

HeLa cells were synchronized in G1/S phase by double thymidine and treated with 2 Gy IR. After washing three times with PBS, cells were cultured in fresh medium for 8 h to allow cells to enter mitosis. Cells were then treated with 330 µM nocodazole for 45 min to harvest mitotic cells. Mitotic cells were collected and dehydrated with 55 mM KCl for 20 min at 37°C, prefixed with 40% fixative (methanol/acetic acid, 3:1) for 5 min, and centrifuged at 500 g for 5 min. After removing supernatant, cells were fixed twice by fixative at room temperature for 30 min, then cell suspension was dropped on the slide. The slide with spread chromosomes was dried at room temperature and stained with 1 µg/ml Hoechst33342 for 10 min.

U2OS, HeLa, RPE-1, and MCF7 cells were synchronized in G1/S by double thymidine treatment. Briefly, cells were treated with 2.5 µM thymidine for 16 h, washed three times with PBS, and cultured with fresh medium for 9 h. Cells were then treated with 2.5 µM thymidine for another 16 h to allow cells to be blocked in G1/S phase. For G0 phase, U2OS or RPE-1 cells were treated by serum starvation for 40 h.

For cell cycle analysis, cells were synchronized, collected using pancreatic enzyme, and washed in PBS. Cells were fixed in cold 70% ethanol overnight at 4°C, spun at 500 g in a centrifuge, and the supernatant carefully discarded. Then the cell pellet was washed two times by PBS. To stain DNA, the cell pellet was resuspended with PBS containing 50 µg/ml propidium iodide, 100 µg/ml RNaseA, 0.2% Triton X-100, and incubated at 4°C for 30min. Stained cells were sent to flow cytometry (BD; LSRFortessaX20), and cell cycle was analyzed by using FlowJo software.

U2OS, RPE-1, HeLa, and 293T cells were cultured in DMEM containing 10% FBS and 1% penicillin/streptomycin at 37°C and 5% CO₂. For microtubule regrowth assay, centrosomal protein immunofluorescence, and pS2056 DNA-PKcs analysis, U2OS or RPE-1 cells were irradiated with 2 Gy or 5 Gy and recovered for different time points (1, 2, 4, 6, and 8 h). For pT308/S473 AKT analysis, RPE-1 cells were exposed to 10 Gy IR. For chromosome spread, HeLa cells were exposed to 2 Gy IR. The following chemicals were used: DNA-PKcs inhibitor (MCE; ku57788; 1 µM), ATM inhibitor (MCE; ku55933; 10 µM), ATR inhibitor (MCE; ADZ6783; 1 µM), AKT inhibitor (TargetMol; MK-2206; 5 µM), CHK1 inhibitor (MCE; SB 218078; 10 µM), CHK2 inhibitor (MCE; CCT241533 hydrochloride; 5 µM), PLK1 inhibitor (MCE; BI2536; 100 nM), p38/MAPK inhibitor (MCE; SB 203580; 10 µM), PDK1 inhibitor (MCE; BX-795, 10 µm; BX-912, 10 µM), bleomycin (MCE; 5 µg/ml), mitomycin C (MCE; MMC, 5 µM), camptothecin (Sigma; CPT, 1 µM), cisplatin (Sigma; 2.5 µM), and etoposide (Sigma; 1 µM). For microtubule regrowth assay and immunofluorescence, cells were preincubated with indicated inhibitors for 1 h before DNA damage treatment. For Western blot analysis of AKT phosphorylation, cells were preincubated with DNA-PKcs inhibitor for 3 h or PDK1 inhibitor for 1 h before IR treatment.

For RNAi experiments, siRNAs were transfected into U2OS or RPE-1 cells using Lipofectamine RNAiMAX according to manufacturer’s instructions (Invitrogen). Cells were synchronized in G1/S phase by double thymidine treatment after 8 h transfection, or cells were collected and analyzed by quantitative RT-PCR or Western blot after 48-h transfections to determine the knockdown efficiency. As an NC, cells were also transfected with an equal amount of scrambled siRNA (NC siRNA) or scrambled shRNA (NC shRNA). The target sequence of each siRNA or shRNA is listed in Table S1. The effect of gene depletion was examined by Western blot or quantitative RT-PCR, and most of the results are shown in Fig. S3.

For quantitative RT-PCR, total RNA was extracted by using the Trizol RNA extraction protocol. One microgram of total RNA was reverse transcribed into cDNA by using the HIScriptII One Step RT-PCR Kit (Vazyme). Gene expression was analyzed by real-time quantitative PCR by using the SYBR Green quantitative PCR Mix (Monad) real-time PCR system. PCR reaction ran for 40 cycles at 95°C for 5 s, 60°C for 1 min, and 72°C for 30 s. Each cDNA sample was run with triplicates. The mRNA level of each sample for each gene was normalized to that of the GAPDH mRNA. The relative mRNA level was presented as unit values of 2^(Ct[GAPDH]–Ct [gene of interest]).

Transient plasmids transfection was performed with polyethylenimine. For the lentivirus package, shRNA plasmids and lentivirus plasmids (pSPAX2 and pMD2G) were cotransfected into 293T cells using polyethylenimine. After 56-h transfections, the supernatants containing packaged lentivirus were harvested to infect HeLa cells with 8 µg/ml polybrene. Stable cell pools were selected in medium containing 2 µg/ml puromycin.

To measure the speed of microtubule growth in live cells, RPE-1 cells stably expressing GFP-EB3 were seeded to a 20-mm glass culture dish in DMEM with 10% FBS and synchronized in G1/S phase by double thymidine block. Imaging was gathered using Airyscan of Zeiss LSM880 confocal with 63× oil objective lens (NA, 1.4) equipped with an environment chamber at 37°C and ZEN (black) gathering software. Time-lapse images were acquired with 1.3 pinhole and 1-s interval per frame over 30 s with 512 × 512 pixels in final size. To track the trajectory and analyze the movement velocity of GFP-EB3, ImageJ software with the Kymograph plug-in was used as described (Mangeol et al., 2016). These images within 30 s were used to make maximum intensity projections by KymographClear and got the overlaid images. Briefly, a series of images from each time-lapse experiment was opened by ImageJ with KymographClear, and images captured at different time points were overlaid to define the tracks of GFP-EB3 comets movement. Tracking diagrams were made using the Manual Tracking plug-in for ImageJ software and Adobe Photoshop, and the movies of the overlaid images were made by Adobe Premiere. Quantitative velocity measurements of the distribution of EB3-GFP tracks were done by plotting the average pixel intensities along a thick line with the width of 5 pixels using normalized projection images spanning 30 s in ImageJ software with the KymographDirect plug-in. The following parameters were used in KymographDirect: time per frame = 1,000 ms; pixel size = 100 nm; particle width = 3 pixels; maximum number of intermediate lines = 100; search window width = 2; and line width = 5 pixels. Finally, the forward motion of each kymograph was analyzed using a noise-reduction algorithm with a 0.5 intensity threshold under the above KymographDirect parameters. Statistical analyses were performed by using an unpaired two-tailed Student’s t test using GraphPad Prism software. Each experiment was repeated three times and one representative quantitative result from these experiments is shown in the figures.

To analyze the DSB mobility, PCNT- or NEDD1-depleted HeLa cells were transfected with GFP-53BP1 TD plasmids. Cells were cultured in DMEM with 10% FBS. After synchronization in G1/S by double thymidine, cells were treated with 2 Gy IR. Imaging was gathered using Olympus IX83 microscopy (Olympus America) with 60× oil objective lens (NA, 1.35) and Andor’s Zyla 5.5 sCMOS camera and cellSens Dimension software at 1 h after IR. The IR-induced DSB mobility was analyzed following the basic procedures described in a previous report (Lottersberger et al., 2015). Briefly, 5-µm Z-stacks at 0.5-µm step images were acquired with 50-ms exposure time every 30 s per frame over 10 min with 2056 × 2056 pixels in final size. After deconvolution, image stacks were average projected. To get the track of 53BP1 foci in different treatment conditions, cells were registered by the tracking plugin with ImageJ, and then particles were detected by the LoG Detector and tracked using TrackMate plugin.

The position $r$ from the same molecule in adjacent frames in the same cells were linked by standard algorithms using Trackpy, from which the trajectories of individual molecules (t) were obtained. The eMSD of every 53BP1 foci in cells was calculated by Python tracking packing Trackpy using the following equation: $[Δr2(τ)]=[(r(t+τ)−r(t))2].$ The eMSD data were then averaged from multiple movies for the same sample at the same time point (http://soft-matter.github.io/trackpy/v0.4.2/).

For distinguishing phosphorylation of AKT in the nuclear and cytoplasmic fraction, RPE-1 cells were collected and extracted by using a nuclear and cytoplasmic protein extraction kit (Beyotime; P0027), according to the manufacturer’s instructions. Briefly, cells were collected by PBS that contained 10 mM EDTA and washed one time by PBS. Cells were lysed in ice-cold cytoplasmic protein extraction fraction A for 15 min and centrifuged at 15,000 g for 5 min; the supernatant fraction was cytoplasmic protein. The pellet was then resuspended in the ice-cold nuclear protein extraction buffer by vortexing, incubated on ice for 30 min and centrifuged at 15,000 g for 10 min; the supernatant fraction was the nuclear protein. To quantify protein concentration, a BCA kit for protein determination (ZOMANBIO; ZD301-2) was used. For the Western blot assays, the proteins were separated by SDS-PAGE and transferred onto polyvinylidene difluoride membranes (Millipore). The membranes were blocked with 5% milk in PBST (PBS with Tween 20), probed with primary and then secondary antibodies, and finally exposed using ECL (Bio-Rad; US EVERBRIGHT). The concentration of primary antibodies AKT1, β-tubulin, β-actin, pericentrin, and 53BP1 was 1:3,000, and the concentration of another primary antibody was 1:1,000. The secondary antibody HRP-Rb/M was purchased from Jackson ImmunoResearch and the concentration was 1:20,000.

All statistical induction and statistical analysis were done using GraphPad Prism software. Mean values were compared by t tests and nonparametric tests for the variance analysis between different groups. Differences were considered significant for P values < 0.05 (****, P < 0.0001; ***, P < 0.001; **, P < 0.01; *, P < 0.05; mean ± SD). All the experiments were repeated three times. For the results shown in box plots when PCM protein intensity was analyzed: center line, median; box limits, 25th and 75th percentile; whiskers, maximum and minimum. When GFP-EB3 velocity and intensity was analyzed: center line, median; box limits, 25th and 75th percentile; whiskers, fifth and 95th percentile.

Fig. S1 contains the experiment procedure for Fig. 1 A and supporting images for Fig. 2. Fig. S2 summarizes the experiment procedures for Fig. 3, A and B, and Fig. 3 D. Fig. S3 contains supporting information for Fig. 4, including Western blots showing the efficiency of indicated siRNA and typical images showing DMSR under indicated conditions. Fig. S4 shows the supporting data for Fig. 5, including the centrosomal accumulation of Ninein and CEP170 during DMSR. Fig. S5 contains supportive data for Fig. 6 and additional images of 53BP1 foci images for Fig. 7. Videos 1, 2, and 3 correspond to Fig. 2 A, showing the GFP-EB3 comet movement under indicated treatment. Videos 4, 5, 6, and 7 correspond to Fig. 2 C, showing the GFP-EB3 movement. Videos 8 and 9 correspond to Fig. 2 E, showing the GFP-EB3 in untreated (Video 8) and bleomycin-treated (Video 9) G1 RPE-1 cells. Videos 10 and 11 correspond to Fig. 9 A, showing the DSB mobility through GFP–53BP1TD foci in shcon (Video 10) or shPCNT (Video 11) G1 HeLa cells. Table S1 contains the sequence information for siRNA or shRNA and the catalog number for the chemicals used in this study.

We thank Jiwei Chang (Medical Research Institute, Wuhan University, Wuhan, China) for the help on data analysis.

This work was supported by grants from the National Key Research and Development Program of China (2018YFC1003400), the National Natural Science Foundation of China (31770868), Wuhan University (2042018kf0215), and the Medical Science Advancement Program (Basic Medical Sciences) of Wuhan University (TFJC2018005) to Q. Chen.

The authors declare no competing financial interests.

Author contributions: Q. Chen and X. Zhang designed and supervised the study; S. Ma performed the experiments and data analysis; C. Liu and Z. Rong generated constructs; C. Liu, Z. Rong, and X. Qin contributed to the statistical analysis; Q. Chen, X. Zhang, and S. Ma wrote the paper.