The mutants gip1, gip2, gip1gip2 and their corresponding genetic background Col-0, WS, and Col-0 x WS have been described previously (Janski et al., 2012; Batzenschlager et al., 2017). Arabidopsis seedlings were grown in vitro on ½ Murashige and Skoog medium (SERVA Electrophoresis) in presence of 1% sucrose supplemented with 1.2% agar at 20°C under long day conditions (16-h light 70 μmol m^–2 s^–1 of fluorescent lighting/8-h dark). The gip1gip2 mutant was complemented by the expression of a pGIP1::GIP1-GFP construct as previously described (Janski et al., 2012; Batzenschlager et al., 2015). For sensitivity tests to genotoxic drugs, seeds were initially sowed on ½ MS agar and after 5 days on growth, seedlings were transferred on media supplemented with the genotoxins. Drug concentrations were 10 μM for Bleomycin (BLM) (Laboratoire Thissen, Belgium) and 50 μM for cisplatin (CP) (Sigma, St. Louis, United States). Seedlings were treated during 16 h by 50 μM CP in ½ MS for cytological analyses.

Nuclei were isolated from 9-day-old seedlings using a Chopping solution and comet assays were performed as described (Roa et al., 2009; Lang et al., 2012). The quantification of the comet figures was related to an arbitrary scoring of the comet figures as described previously (Collins, 2004). In each assay, 100 comets were scored and the results represent the mean values from three independent experiments.

Nine-day-old Arabidopsis seedlings were fixed in 4% PFA and processed as described previously on nuclei from squashed root tip (Batzenschlager et al., 2015). The primary anti-γ-H2AX antibody (diluted at 1/500) produced by Davids Biotechnology (Regensburg, Germany; Friesner et al., 2005), the rat anti-RAD51 antibody (diluted at 1/500) (Kerzendorfer et al., 2006) and when needed the monoclonal antibody directed against GFP (diluted at 1/500) (Invitrogen, Thermo Fisher Scientific) were incubated overnight at 4°C. Signals were revealed accordingly with the following secondary antibodies: Alexa fluor-488 goat anti-rabbit (diluted at 1/200), the Alexa fluor-488 goat anti-mouse (diluted at 1/200), the Alexa fluor-568 goat anti-rat (diluted at 1/300) and the Cy5 goat anti-rabbit (diluted at 1/300) (Life Technologies, Thermo Fisher Scientific). Root tips were mounted in antifade Vectashield (Vector Laboratories), with DAPI (2 μg/ml).

Confocal images were recorded with a Zeiss LSM 780 microscope equipped with an oil 63 × /1.4 NA lens. The excitation and emission wavelengths for Alexa 488 were 488 and 510 nm, respectively. To reveal Alexa 568, the excitation and emission wavelengths were 555 and 617 nm, respectively. For DAPI observations, the excitation and emission wavelengths were 405 and 500 nm, respectively. Cy5 was combined with DAPI, with excitation and emission wavelengths of 571 and 735 nm, respectively. Observations were performed in multi-tracking mode using 405-, 488-, or 561-nm laser excitation. Images were processed using the ImageJ software.

Total RNA was extracted from 9-day-old Arabidopsis thaliana seedlings using the Nucleospin RNA Plant kit (Macherey-Nagel, Hoerdt, France) according to manufacturer’s instructions after grinding with glass beads (1, 7/2 mm) in a Precellys^®24 grinder (Bertin Technologies, Saint-Quentin-en-Yvelines, France) at 5,500 rpm, 2 × 30 s. For RT-qPCR, 2.5 μg of RNA were used to synthesize cDNA using specific primers, random hexamer primers (IDT) and the protocol “SuperScript^® IV (SSIV) First-strand and cDNA Synthesis Reaction” (Invitrogen). RT-qPCR was performed on a LightCycler 480 II (Roche) with SYBR Green Master Mix (Bio-Rad), according to the manufacturer’s instructions. All primers used are described in Supplementary Table 1. Quantification was done using the ΔΔCt method and normalized to ACTIN2 (Livak and Schmittgen, 2001).

Our previous data have shown that chromosome instability occurs during mitosis in the gip1gip2 double mutant, resulting in the appearance of micronuclei in interphase cells and ploidy defects (Janski et al., 2012; Batzenschlager et al., 2015). Because micronuclei may result as well from mis-repaired and/or unrepaired DNA DSBs (Fenech et al., 2011; Ye et al., 2019) we decided to explore whether DSB repair is affected in gip1gip2. Firstly, DSBs were detected and quantified using the neutral comet assay. Our analyses were performed on isolated nuclei from gip1gip2 seedlings showing a strong phenotype (Batzenschlager et al., 2015), the single mutants gip1 and gip2 and on nuclei of the corresponding WT controls (i.e., Col-0xWS, WS, and Col-0). Compared to their respective controls, comet tails appeared increased in both single and double gip mutants (Figure 1A). In order to quantify more precisely their respective genomic instability, DNA damage was evaluated upon visual scoring of each individual comet tail as previously described (Roa et al., 2009). While gip1 and gip2 presented about 1.4 times more DSBs than their respective WT control, gip1gip2 showed 2.4 times more DSBs than Col-0xWS (Figure 1B). Altogether, our data suggest that GIP1 and GIP2 may have some overlapping functions to maintain genome stability and to limit DNA damages in interphase nuclei under normal growth conditions.

To further analyze the significant increase of DSBs in gip1gip2 and determine their localization at the cellular level in root meristematic nuclei, we performed immunolocalization with an antibody directed against γ-H2AX, a phosphorylated form of the H2AX histone variant known to be associated with sites of DSBs (Lang et al., 2012). Compared to the low occurrence of nuclei showing γ-H2AX foci in WT, the gip1gip2 double mutant showed a significantly higher proportion of nuclei with foci and, frequently, several foci per nucleus (Figures 2A–D). The foci were found in the nucleoplasm, around the nucleolus and at the nuclear periphery. Among them, 22% were associated to the chromocenters (Figures 2B–D, see white arrows).

The γ-H2AX foci were observed in 19% of the nuclei in the WT control (hereafter named NPF for Nuclei Presenting Foci), while the proportion of NPF was 3.2 times higher in the double mutant (Figure 2E). Moreover, compared to the WT control in which a mean of 1 focus per nucleus was evaluated, the number of foci per nucleus was significantly increased in the double mutant, reaching a mean of 6 ± 1 foci/nucleus (Figure 2F). While the foci mean diameter was relatively homogenous in the WT control with a mean of 0.3 μm ranging from 0.1 to 0.7 μm, it appeared larger in the double mutant with a mean of 0.5 μm and strongly heterogeneous with diameters ranging from 0.2 to 2.5 μm (Figure 2G).

Previously, gip1gip2 plants were described to display ploidy instability (Janski et al., 2012; Batzenschlager et al., 2015). To disentangle appearance of γ-H2AX foci from ploidy defects, we next focused our analysis on 2C flow sorted nuclei from root seedlings (Figures 2H–M). Proportions of NPF in the WT control (18%) and the gip1gip2 double mutant (62%) (Figure 2K) were similar as for root meristematic nuclei (Figure 2E). In addition, both the number of foci per nucleus and their individual sizes were significantly higher in the double mutant compared to the WT control (Figures 2L,M). Indeed, WT nuclei with γ-H2AX associated signal presented rarely more than 1 focus, while in the double mutant the number of foci per nucleus could reach up to 12 (Figure 2L). Similarly, the diameter of the foci was rather constant in WT control nuclei, with an average around 0.2 μm, while in gip1gip2 it varied between 0.12 to more than 1.2 μm with a mean diameter of 0.3 μm (Figure 2M). Interestingly, among large foci (n = 81; Figures 2I,J), 70% were detected at the nuclear periphery and 30% around the nucleolus. Smaller and less intense ones were distributed randomly through the nucleoplasm.

Consistent with the higher rate of DNA lesions measured by the comet assay, these results altogether indicate that gip1gip2 over-accumulates γ-H2AX foci that are heterogenous in size. These foci are notably found at the nuclear periphery close to the chromocenters.

Since we observed a constitutively high level of DSBs and numerous γ-H2AX foci in gip1gip2, we next investigated which DNA repair process was impaired. Firstly, we have checked the expression level of genes involved in the DSBs repair using quantitative real-time PCR (RT-qPCR). To this end, we selected genes encoding key factors involved in the two main DSB repair pathways: (i) the HR, restricted to cells in the S/G2 phase to resolve stalled replication fork and (ii) the non-homologous end-joining pathways (NHEJ) involved in repair of most DSBs (West et al., 2004). As representative of the HR pathway, we choose genes encoding HR regulators such as BRCA1 (Lafarge and Montane, 2003) and RAD9 (Heitzeberg et al., 2004), as well as the gene encoding the HR effector RAD51 (Bleuyard et al., 2006). As indicative of the NHEJ pathway, KU70 and KU80 genes were analyzed for the classical NHEJ (c-NHEJ; Charbonnel et al., 2010), and the poly (ADP-ribose) polymerase encoding gene PARP2 for the KU-independent backup-NHEJ (b-NHEJ) pathway (Jia et al., 2013). The expression of all tested genes was up-regulated in gip1gip2, indicating the activation of the DDR. However, contrary to those involved in NHEJ pathways, the genes representative of the HR pathway were the most induced in gip1gip2 compared to WT (Figure 3). Indeed, BRCA1 and RAD9 encoding upstream regulators of HR, showed the highest fold change in their expression level, ranging between 6 and 9, respectively. The RAD51 expression showed also a significant increase compared to WT with a 3.4-fold change, while the expression of NHEJ pathway genes (KU70, KU80 and PARP2) remained only weakly induced (between 1.6- and 1.8-fold change).

In order to further determine which DNA repair pathway(s) were more specifically affected in gip1gip2, we next tested this double mutant in root growth assays in presence of the DNA damaging agents bleomycin (BLM) or cisplatin (CP). While the DSBs induced by BLM are classically repaired by both HR and NHEJ, the intra- and inter-strand DNA cross-links bridges induced by CP are preferentially repaired through HR during DNA replication (Fuertes et al., 2003). Therefore, we tested WT and gip1gip2 root growth on media containing 10 μM BLM and 50 μM CP, respectively, as previously described (Roa et al., 2009; Biedermann et al., 2017). Contrary to WT plants in which root growth was similarly slowed down by BLM and CP, gip1gip2 root growth appeared more affected upon CP treatment than upon BLM treatment (Figure 4 and Supplementary Figure 1). The higher sensitivity to CP (twice more compared to WT) than to BLM (decrease of only 1.2-fold) in gip1gip2 is indicative of a predominant defect in HR repair.

Together, the major up-regulation of genes related to HR in gip1gip2 and the increased sensitivity of the mutant to CP, suggest that GIP1 and GIP2 may be important to regulate genome maintenance through HR in somatic cells.

In interphase nuclei from root meristems, the gip1gip2 mutant accumulated a high number of γ-H2AX foci. This is indicative of defects in DNA repair, notably through HR, as shown by the sensitivity of the mutant to CP. Thus, we further investigated the nuclear distribution of the HR effector RAD51. We performed co-immunolabeling on WT and gip1gip2 root tip nuclei using together specific antibodies against RAD51 and γ-H2AX (Figure 5). As expected, a few nuclei showed RAD51 foci in WT (5.5%, n = 109) which colocalized systematically with γ-H2AX foci (Figures 5A,B) as described previously (Biedermann et al., 2017). Such colocalization is further supported by fluorescence profiles presented in Figures 5A,B. In agreement with the increased number of γ-H2AX foci (Figure 2), the number of RAD51 foci was also significantly increased in gip1gip2, with 36% of nuclei showing RAD51 foci with an average of four per nucleus. Interestingly, the number of RAD51 foci per nucleus was significantly lower than the number of γ-H2AX foci and surprisingly most of the RAD51 foci showed no clear colocalization with γ-H2AX foci in gip1gip2 (73%, n = 224) (Figures 5C–E). In addition, 61% of the RAD51 foci were mainly found at/or very close to the bright DAPI-stained chromocenters (Figures 5C–E, see white arrows) without the presence of γ-H2AX or no clear colocalization with γ-H2AX. Since a substantial decrease in the activity of GIP1 and GIP2 in the double knockdown mutant gip1gip2 may impair RAD51 and γ-H2AX colocalization, we further explored the localization of GIP1-GFP upon genotoxic stress using CP. Using live cell imaging, GIP1-GFP in the nucleus appeared to be located close to the NE, as previously described (Janski et al., 2012; Batzenschlager et al., 2013; Batzenschlager et al., 2015) and this location was maintained when seedlings were treated by 50 μM CP for 16 h (Figures 6A,B). As expected, compared to the control (Figure 6C), we observed that all RAD51 foci induced by CP colocalized with γ-H2AX foci using immunolabeling (Figure 6D, merge RAD51 and γ-H2AX), while 23% (n = 444) of the RAD51 foci also colocalized with GIP1-GFP at the nuclear periphery (Figure 6D, merge RAD51 and GIP1-GFP with corresponding colocalization profiles).

Here, we showed that RAD51 and γ-H2AX foci are mostly not colocalizing in gip1gip2 and that RAD51 foci are surprisingly located more frequently at chromocenters in the mutant. Interestingly in response to CP, RAD51, besides being systematically colocalized with γ-H2AX foci in the complemented mutant, partly colocalizes with GIP1-GFP at the nuclear periphery.

At the cellular level, numerous γ-H2AX foci are detected in gip1gip2 and their heterogenous size may reach up to 2.5 μm in diameter (Figure 2), indicating a permanent unrepaired DNA damage as described (Costes et al., 2006). This is contrasting with the low number of γ-H2AX foci which are smaller and more homogenous in size in WT, where active DNA repair occurs. As the mutant is mostly sensitive to cisplatin, this is indicative of a defective HR repair during DNA replication, leading to the accumulation of DSBs. Some of the γ-H2AX foci in gip1gip2 are found at the nuclear periphery close to the chromocenters, which are mainly constituted by repetitive sequences and pericentromeric heterochromatin (Fransz et al., 2002). As centromeric heterochromatin organization was affected in gip1gip2 (Batzenschlager et al., 2015), we cannot exclude defect in heterochromatin replication coupled with defective DNA repair. In this respect, we observed twice more late S-phase replicating nuclei in gip1gip2 compared to WT (Supplementary Figure 2A), with the presence of γ-H2AX foci in the vicinity of late S-phase replicating DNA, mainly corresponding to pericentromeric heterochromatin (Supplementary Figure 2B). This defect in DNA repair could also explain the accumulation of G2-stalled 4C nuclei in gip1gip2 (Janski et al., 2012; Batzenschlager et al., 2015). Alternatively, unscheduled RAD51 removal from DNA may also lead to its overaccumulation at stalled replication forks, as described in USO2 cells (Parplys et al., 2015). Thus, such overaccumulation leads to a decreased HR efficiency and an increased genomic instability. Together, besides the activation of the HR pathway, defects in DNA replication as well as abnormal RAD51 loading may result in the increased genomic instability observed in gip1gip2.

In plants, γ-H2AX form foci at DSB sites, where HR proteins such as RAD51 and RAD54 are sequentially recruited (Biedermann et al., 2017; Horvath et al., 2017; Meschichi et al., 2021). Moreover, RAD54 forms foci at DSBs that are restricted to S and G2 phase cells in roots (Hirakawa et al., 2017) and localized at the nuclear periphery close to the nuclear envelope at high frequency (Hirakawa and Matsunaga, 2019). Interestingly, this peripheral location was reduced in the double mutant for the plant nucleoskeleton components CRWN1 and CRWN4 after γ-irradiation (Hirakawa and Matsunaga, 2019). Furthermore, an interaction between RAD51 and RAD54 was already detected in plants (Klutstein et al., 2008) and RAD54 may contribute to the removal of RAD51 at replication forks, as demonstrated in human cells (Mason et al., 2015). Together, we cannot exclude that, like for RAD51 which partially colocalized with GIP1 (Figure 6D), the function of RAD54 might also be affected in gip1gip2 due to its possible impaired localization. Moreover, considering the crucial role of NPCs for DNA repair in animal and yeast (Freudenreich and Su, 2016), defects in the shape and the distribution of NPCs reported in gip1gip2 (Batzenschlager et al., 2013) may also partially explain the defective HR repair we observed here. Together, GIP proteins may contribute to the spatio-temporal recruitment of RAD51 for efficient HR in genomic maintenance at the nuclear periphery.

As GIPs are important to recruit microtubule (MT) nucleation complexes (Janski et al., 2012), we cannot exclude a role of either MTs or the γ-tubulin complex in the regulation of DNA repair. Indeed, MTs connected to the LINC complex were described to induce chromatin mobility around DSBs to promote efficient DNA repair in mammals (Lottersberger et al., 2015). Alternatively, an additional specific intranuclear function of GIPs and some of the MT nucleation complex proteins such as γ-tubulin, independent of MT dynamics and/or MT nucleation, may exist as well. Indeed, plant γ-tubulin was found at the inner membrane of NE in association with SUN1 (Chumova et al., 2019), and GIPs are also located on both sides of the NE (Batzenschlager et al., 2013; Batzenschlager et al., 2014).

GIPs are found at the nuclear periphery close to chromocenters (Batzenschlager et al., 2015). Interestingly, RAD51 localization at DNA lesions is mediated by RBR1 (Biedermann et al., 2017) which can form foci with γ-H2AX in the close vicinity of chromocenters (Horvath et al., 2017). In addition, both RAD51 and GIPs were shown to be substrates of the cell cycle regulators B1-type cyclin complexes (Motta et al., 2021), which activity drives the recruitment of RAD51 at DNA damage sites (Weimer et al., 2016). In line with this and similarly as our report on gip1gip2, cycB1 mutants are mainly sensitive to cisplatin (Weimer et al., 2016). Altogether these data suggest that a dynamic crosstalk may involve GIPs as well as the cell cycle regulators RBR1 and B1-type cyclin complexes to allow proper loading of RAD51 at DNA lesions. However, the underlying mechanisms need to be further explored.