Interferon-stimulated gene 15 accelerates replication fork progression inducing chromosomal breakage

Timely and accurate DNA replication in dividing cells is crucial to maintain the integrity of the human genome. However, due to constitutive growth signaling and defective DNA repair, cancer cells may exhibit replication stress, a phenomenon characterized by perturbation of error-free DNA replication and slowing or stalling of replication fork progression and DNA synthesis, inducing genomic instability and tumorigenesis (Zeman and Cimprich, 2014). Replication stress can arise as consequence of normal cellular events involving DNA (i.e., replication–transcription collisions and replication of special DNA structures, such as telomeres, fragile sites, and G-quadruplex), upon exposure to external agents, including irradiation or chemotherapeutic drugs, or after oncogene activation (Muñoz and Méndez, 2017). Although replication stress has been proven to induce genomic instability and tumorigenesis, recent studies have shown that enhancing replicative stress to induce catastrophic failure of cancer cell proliferation may provide a powerful therapeutic approach (Forment and O’Connor, 2018).

The mechanisms that underlie the cellular DNA damage response and DNA replication stress are complex and tightly controlled by posttranslational protein modifications, including phosphorylation, acetylation, methylation, poly-(ADP-ribosyl)ation, and modifications by the ubiquitin system (Wang et al., 2017). Ubiquitin-like modifiers (UBLs) are small polypeptides whose three-dimensional structures are strikingly similar to that of ubiquitin, although the similarity in their amino acid sequences to ubiquitin significantly varies (Kerscher et al., 2006). Ubiquitin and UBLs have pivotal roles in the cellular response to various forms of stress and mainly act via covalent conjugation to target proteins. This kind of protein modification can affect stability, subcellular localization, activity, and overall function (Wang et al., 2017). The role of ubiquitin and the UBLs SUMO and NEDD8 in the control of cell cycle and DNA damage response has been extensively studied (Brown and Jackson, 2015; Dantuma and van Attikum, 2016). However, the function of most UBLs, including ISG15 (the first UBL identified), in genome stability is largely unknown.

As part of the innate immunity, ISG15 is robustly induced by type I and III IFNs, in order to protect the host during pathogen infection (Loeb and Haas, 1992; Perng and Lenschow, 2018). ISG15 can modify many cellular proteins, in a process called ISGylation, by conjugating its C-terminal glycine residue to lysines on the targets, yet the fate of this modification is still largely unknown. Increasing evidence suggests that ISG15 can regulate host response also by acting as a free intracellular molecule. An example is the stabilization of USP18 by noncovalent binding of ISG15, which is essential to prevent aberrant IFN signaling in humans (Zhang et al., 2015). Furthermore, unconjugated ISG15 can be secreted and function as a cytokine (D’Cunha et al., 1996; Dos Santos and Mansur, 2017; Swaim et al., 2017). ISG15 expression can also be induced independently of IFNs via the activity of p53 upon exposure to DNA-damaging agents and irradiation or in condition of telomere shortening (Jeon et al., 2012; Liu et al., 2004; Lou et al., 2009; Park et al., 2016; Park et al., 2014). Interestingly, it was shown that reversible PCNA ISGylation relays a signaling pathway to turn off error-prone translesion synthesis after DNA lesion bypass for suppressing UV-induced mutagenesis as well as for resuming normal DNA replication (Park et al., 2014).

Elevated levels of ISG15 expression occur in many types of cancer (Andersen et al., 2006; Bektas et al., 2008; Desai et al., 2006, 2012; Ina et al., 2010; Jinawath et al., 2004; Laljee et al., 2013; Li et al., 2014; Padovan et al., 2002; Talvinen et al., 2006), and in some cases, the robust expression of ISG15 was reported to support tumor growth (Burks et al., 2014; Forys et al., 2014; Hadjivasiliou, 2012). In spite of the increasing interest on ISG15 and its clear correlation with human malignancies, its role in tumorigenesis is still controversial and largely unexplored (Han et al., 2018; Villarroya-Beltri et al., 2017), and its mechanism of action is far from being clarified.

Here, we show that high levels of ISG15 expression, which occur upon type I IFN (IFN-β) treatment and in many human tumors, are detrimental for the cell, leading to accelerated and deregulated DNA replication fork progression, which ultimately results in extensive chromosomal lesions. This effect is largely independent of ISG15 conjugation activity and relies on the noncovalent functional interaction with RECQ1, a key helicase involved in replication fork restart after stalling.

To gain insight into the potential effect of ISG15 in the regulation of genome stability, we developed different systems to modulate ISG15 expression. To reproduce conditions of high levels of ISG15 expression irrespective of IFN stimulation, we generated a human osteosarcoma (U2OS) Flp-In T-REx (FIT) cell line, which inducibly expresses FLAG-ISG15 upon doxycycline treatment. In this cell system, ISG15 levels upon induction are comparable to those observed upon IFN-β treatment (Fig. 1 A). Moreover, we established ISG15 knockout (KO) in U2OS FIT cell lines (referred as U2OS FIT ISG15/KO) via CRISPR/Cas9-based genome editing (Fig. S1, A and B), to use as control in different experiments.

To assess the localization of ISG15 in cells, we performed subcellular fractionation and found that ISG15 is detectable not only in cytosolic and nuclear soluble fractions but also in chromatin fractions (Fig. S1 C). To validate the presence of ISG15 in chromatin compartments, we employed the isolation of proteins on nascent DNA (iPOND) technique, which allows the isolation of proteins bound, directly or indirectly, to nascent DNA at the replication forks (Sirbu et al., 2011). Using this method, we were able to confirm the localization of ISG15 on chromatin (Fig. 1 B; 5-ethynyl-2-deoxyuridine [EdU] +Thy chase, +click) and, most important, at the replication forks (Fig. 1 B; EdU +click). Finally, we adopted a cell-based method, namely the proximity ligation assay (PLA), to visualize and measure the localization of ISG15 at the replication forks by monitoring the close association of ISG15 with the proliferating cell nuclear antigen (PCNA) and newly synthesized DNA, labeled by EdU. Quantitative imaging allowed us to evaluate the high number of PLA signals (measured as foci counts per nucleus) for ISG15/PCNA and ISG15/EdU in cells expressing high levels of FLAG-ISG15 (Fig. 1, C–F). Interestingly, low ISG15/PCNA PLA signal appeared also in control cells (empty vector [EV]) likely unveiling the interaction of PCNA with endogenous ISG15, which is expressed at low basal levels in U2OS cells. Accordingly, ISG15/PCNA PLA signals are absent in the EV sample when the FLAG antibody is used instead of ISG15 antibody (Fig. S1 D). Importantly, no signal was detected in ISG15 KO cells or upon staining with single antibodies (recognizing PCNA, ISG15, or EdU), confirming the specificity of the system (Fig. 1, D and F). These results reveal that ISG15 localizes on chromatin at the replication forks, suggesting a possible function in modulating DNA replication.

To determine whether ISG15 plays a role in DNA replication, we labeled newly replicated DNA by providing cells with halogenated nucleotides and performed the DNA fiber spreading assay (Jackson and Pombo, 1998) in different experimental conditions to evaluate replication fork progression at the single-molecule level. Strikingly, cells expressing high levels of ISG15 exhibited longer newly replicated tracks during the labeling period (ISG15 + doxycycline) compared with control cells (EV, − doxycycline and + doxycycline; Fig. 1 G). In line with this, by measuring EdU incorporation in S phase using FACS, we observed an increased rate of DNA synthesis in cells with high levels of ISG15, without affecting the frequency of origin firings (Fig. S1, E and F). While analyzing replication forks that diverge from the same replication origin, we observed no fork asymmetry, which is indicative of frequent fork pausing usually associated with replication stress, as observed upon mild treatment with the topoisomerase-1 inhibitor camptothecin (CPT; 50 nM; Fig. S1, G and H). The effect of ISG15 levels on DNA replication is dose dependent and already detectable at early time points of doxycycline induction (4 and 8 h; Fig. 1, H and I). Importantly, high levels of ISG15 do not induce expression or stabilization of its deconjugating enzyme USP18 at the mRNA or protein level (data not shown), thereby excluding any contribution of USP18 to the replication phenotype observed. Taken together, these data show that ISG15 is located at DNA replication forks, where it increases the fork progression rate in a dose-dependent manner.

Since type I IFN is one of the main physiological inducers of ISG15 expression, we tested the effects of IFN-β stimulation on DNA replication in U2OS cells. To prevent the cytotoxic effects of IFNs on cell viability, we limited the treatment to 2 h and then chased in IFN-β–free media for different time points, tested the induction of ISG15 (Fig. 2 A), and measured the rate of replication fork progression. Remarkably, we observed that treatment with IFN-β recapitulates the increased DNA replication fork speed observed in U2OS FIT cells upon doxycycline induction, which consistently correlates with ISG15 expression levels (Fig. 2, A and B). To assess the specific contribution of ISG15 over the many factors regulated by IFNs, we tested the effect of IFN-β on replication fork progression in U2OS FIT ISG15/KO cells. Notably, we found that the accelerated fork progression rate was abrogated in cells lacking ISG15 (Fig. 2, C and D) and restored upon doxycycline-dependent reexpression of FLAG-ISG15 in U2OS FIT ISG15/KO cells (Fig. 2, E and F; see Materials and methods for details). These data clearly indicate that the increase in replication fork progression, observed upon IFN-β treatment, relies on ISG15 expression. Additionally, to test if this is a general effect and not restricted to U2OS cells, we generated ISG15 KO in MCF7 cells (human breast cancer; MCF7 ISG15/KO; Fig. S2 A), using the same experimental pipeline as for U2OS FIT cells, and measured replication fork progression upon IFN-β stimulation in parental (WT) and MCF7 ISG15/KO cells. In line with the data obtained in U2OS cells, we found that IFN-β accelerates fork progression in MCF7 in a ISG15-dependent manner (Fig. 2 G). Again, accelerated fork progression in MCF7 ISG15/KO cells was restored by stable reexpression of ISG15 (Fig. 2 H). Our findings provide strong evidence that accelerated DNA replication fork progression is promoted by physiological IFN-β–mediated overexpression of ISG15.

Since ISG15 expression is often up-regulated in cancer, we aimed to investigate whether the rate of DNA replication is regulated by ISG15 in other cancer cells in addition to U2OS and MCF7. We analyzed the levels of ISG15 in a panel of cancer cell lines using U2OS as reference system for ISG15 expression (Fig. S2 B). We selected three cell lines (HeLa from cervical cancer and M059K and T98G from glioblastoma) that exhibit relatively high levels of ISG15 and efficient ISG15 knockdown upon transient transfection of siRNAs targeting ISG15 (Figs. 2 I and S2 C). Although HeLa, M059K, and T98G cells showed intrinsic differences in DNA replication fork progression, depletion of ISG15 leads to a 30–40% reduction in replicated track length in all of these cell types (Fig. 2 J). Overall, these data show that ISG15 expression levels affect DNA replication fork progression in cancer cells of various origin.

ISG15 function was mainly studied as protein modifier able to covalently conjugate to target proteins, but it can also act as free molecule by interacting with proteins noncovalently (Dos Santos and Mansur, 2017; Swaim et al., 2017; Zhang et al., 2015). To assess whether the conjugation ability of ISG15 is required for its influence on replication, we generated U2OS FIT and U2OS FIT ISG15/KO cell lines that express a mutant of ISG15 lacking the C-terminal diglycine motif that is required for covalent modification of target proteins (ISG15ΔGG; Fig. 3 A). Overexpression of ISG15ΔGG largely recapitulated the replication phenotype observed upon overexpression of WT ISG15 (Fig. 3 B; and Fig. S3, A and B). Likewise, ISG15ΔGG also accelerated replication fork progression in the context of ISG15 KO cells and localizes in close proximity to PCNA, as revealed by PLA analysis (FLAG/PCNA), although to a lesser extent than WT ISG15 (Fig. 3, B and C), suggesting possible additional roles for ISG15 conjugation in DNA synthesis. Analogous to WT ISG15, expression of ISG15ΔGG did not induce fork asymmetry or alterations in cell cycle distribution (Fig. S3, C and D).

ISG15 consists of two tandem ubiquitin-like domains bearing the typical β–grasp folds (Narasimhan et al., 2005), though the sequence homology with ubiquitin is quite low (Fig. S3 E). Intriguingly, while analyzing the crystal structure of ISG15, we observed that the N-terminal lobe of ISG15 contains a hydrophobic surface, centered on L10, L72, and V74 (referred as LLV), reminiscent of the hydrophobic patch characteristic for ubiquitin (L8, I44, and V70; Fig. 3 D). This patch is strictly required for ubiquitin functions and constitutes the recognition site of most ubiquitin-binding domains (Hicke et al., 2005). Therefore, we tested whether the LLV patch is required for ISG15 function. We found that single mutations in the LLV patch significantly reduced fork acceleration induced by high levels of ISG15, while the triple mutant suppressed it completely (LLVAAA; Fig. 3, E and F) without major alterations of cellular localization (Fig. S3 F). To exclude that mutations targeting the LLV patch generally affect ISG15 protein folding, we tested the ability of L72A and the LLVAAA mutants to conjugate to target proteins. When coexpressed in HEK293T cells together with the ISG15 conjugation machinery (UBE1L as E1, UBCH8 as E2, and HERC5 as E3), both ISG15 variants were as efficiently conjugated as WT ISG15 (Fig. S3 G).

To pinpoint factors potentially involved in the replication function of ISG15, we performed mass spectrometry analysis to search for ISG15 binding partners, rather than targets, by analyzing the chromatin factors interacting with ISG15ΔGG (Fig. 3 A; and Fig. S4, A–C). Under these conditions, we found a limited number of potential interaction partners of ISG15 (Table S1). The best candidates were ranked on the basis of their function and chromatin localization. One of the most promising factors identified was RECQ1, a key DNA helicase that binds a variety of DNA structures, including DNA replication forks and Holliday junctions (Popuri et al., 2008; Sharma et al., 2005), and promotes the branch migration and restart of DNA replication forks upon fork stalling (Berti et al., 2013). To validate RECQ1 interaction, we performed coimmunoprecipitation in HEK293T cells overexpressing both HA-RECQ1 and myc-ISG15 and observed a relatively modest interaction (Fig. S4 D), while no clear association was found using recombinant proteins (data not shown), likely suggesting that ISG15–RECQ1 interaction is difficult to detect and study by standard biochemistry. Therefore, to better explore the ISG15–RECQ1 interaction, we exploited the NanoBRET assay, based on the bioluminescence resonance energy transfer optimized to study dynamic interactions between proteins in a cellular context, due to the distance constraint of energy transfer of ∼5 nm (Machleidt et al., 2015). Importantly, this assay revealed a clear association of RECQ1 with ISG15, both as WT and as conjugation-defective variant, whereas no signal was observed with p53, an unrelated protein that we used as a control for specificity (Figs. 4 B and S4 E). To further corroborate this result, we performed PLA on U2OS FIT cells expressing ISG15 WT and ISG15ΔGG using antibodies against ISG15 and endogenous RECQ1, and we obtained remarkably strong signals in both cases, providing further evidence that RECQ1 is indeed in a protein complex with ISG15 (Fig. 4, C and D).

Next, we addressed the functional link between ISG15 and RECQ1 by assessing whether the expression of RECQ1 is required for the replication phenotype observed in cells expressing high levels of ISG15. Notably, depletion of RECQ1 completely abolished the accelerated replication fork progression induced by high levels of ISG15 and ISG15ΔGG (Fig. 4, E and F), suggesting that ISG15 may regulate RECQ1 function by unleashing its restart activity. To test whether ISG15 promotes RECQ1-dependent fork restart, we measured the DNA replication restart after fork stalling using an established DNA fiber protocol that includes a prolonged treatment with hydroxyurea (HU; 4 mM, 4 h) between nucleotide analogue–labeling periods. In line with previous results (Berti et al., 2013; Zellweger et al., 2015), RECQ1-depleted cells were only partially defective in fork restart due to the contribution of alternative restart pathways (Thangavel et al., 2015). Notably, cells expressing high levels of ISG15 display accelerated restart of stalled forks, which is suppressed by RECQ1 depletion, leading to a marked fork restart defect (Fig. 4, G and H). Collectively, these data strongly suggest that ISG15 promotes the fork restart activity of RECQ1, without affecting RECQ1 protein levels (Fig. S4 F).

We next addressed whether high levels of ISG15 could also lead to unrestrained replication fork progression in conditions of mild DNA replication stress, which is typically associated with early tumorigenesis or chemotherapeutic treatments (Berti and Vindigni, 2016; Macheret and Halazonetis, 2015). Cells expressing high levels of ISG15 and control cells were challenged by mild doses of genotoxic agents, which cause replication fork slowing without detectable DNA damage (Zellweger et al., 2015), and then subjected to DNA fiber analysis. The ratio between 5-iodo-2-deoxyuridine (IdU) and 5-chloro-2'-deoxyuridine (CldU) was measured to assess the replication slowdown observed upon treatment with genotoxic stress (see scheme in Fig. 5 A). As expected, replication fork progression was significantly reduced in control cells treated with a low dose (50 nM) of CPT (EV + CPT). In contrast, cells expressing high levels of either WT or ISG15ΔGG were less sensitive to CPT treatment and displayed unrestrained fork progression (Fig. 5 A). Similar effects were observed upon treatment with mild doses of the DNA cross-linking agent cisplatin (cis-diamminedichloridoplatinum(II), CDDP; 1 µM), but not with HU (0.5 mM), which, at this concentration, directly prevents fork progression by depleting nucleotides (Fig. S4 G). Remarkably, the replication rate observed in cells with high levels of ISG15 and treated with CPT was comparable to that measured in unperturbed control cells (Fig. 5 B), indicating that high ISG15 levels impose sustained fork progression even in conditions of replication stress. Moreover, the unrestrained fork progression observed in condition of high levels of ISG15 upon CPT treatment was also abrogated by loss of RECQ1 (Fig. 5 C), further suggesting that ISG15 unleashes RECQ1 restart activity, even in the context of replication stress.

Accelerated replication fork progression has been reported to be detrimental for cells (Maya-Mendoza et al., 2018). Thus, we asked whether the deregulated replication fork progression observed in context of high levels of ISG15 results into detectable genomic lesions. To address this point, we took advantage of the neutral comet assay, which allows measuring the formation of DNA double-strand breaks in different experimental conditions. Strikingly, we found that high levels of either ISG15 or ISG15ΔGG mutant were sufficient to induce the accumulation of DNA lesions in cells, and this effect was increased upon treatment with low-dose CPT (Figs. 5 D and S4 H). This adverse effect on DNA integrity is usually repaired in normal conditions. However, in cells expressing high levels of ISG15, increased chromosomal abnormalities were detected during mitosis, mostly visible as regions of decondensed chromatin along metaphase chromatids (Fig. 5, E and F). To examine whether the detrimental phenotype promoted by ISG15 impacts cell viability in response to different drugs, we performed clonogenic survival assays. Remarkably, the expression of high levels of ISG15 sensitized cells to low doses of CPT and the poly (ADP-ribose) polymerase (PARP) inhibitor (PARPi) olaparib (Fig. 5, G and H). This result supports and helps explain previous observations, showing that high levels of ISG15 increase the sensitivity to CPT in breast cancer cells (Desai et al., 2008) and correlate with high sensitivity to irinotecan (a clinically used CPT derivative) in gastric cancer (Shen et al., 2013).

Up-regulation of ISG15 and increased ISGylation of target proteins are well-characterized IFN-mediated responses to pathogen infection (Perng and Lenschow, 2018) but are also associated with pathological conditions observed in many types of cancer. Depending on the context, either oncogenic or tumor suppressive effects were reported (Han et al., 2018). Here, we describe a novel unexpected function of the UBL ISG15 in the regulation of DNA synthesis. We demonstrate that ISG15 localizes at DNA replication forks in close proximity to PCNA and newly synthesized DNA. High levels of ISG15 expression, as observed upon IFN induction or in our inducible cell line, deregulate replication fork progression, leading to genomic instability.

Keeping replication fork speed under tight control is essential to preserve genome stability. While slowing down replication fork progression is frequently observed when cells experience replication stress (Zeman and Cimprich, 2014), an increase in fork progression rate is far more uncommon. In the present study, we found that expression of high levels of ISG15, as well as its conjugation-defective form, ISG15ΔGG, causes faster and unrestrained DNA replication fork progression. In the presence of high levels of ISG15, the fork progression rate observed upon mild genotoxic stress is comparable to that of untreated control cells, strongly indicating that ISG15 dampens the active slowing of replication fork progression in response to genotoxic agents. As a result of this desensitization, cells accumulate chromosomal breakage. In line with this, we found that the rate of replication fork progression of different cancer cells constitutively expressing high levels of ISG15 is markedly reduced upon depletion of ISG15, further supporting the role of ISG15 as a critical modulator of replication fork progression across different cellular systems.

The rate of replication fork progression may be affected by multiple factors. Since the effect of ISG15 in replication fork progression appears partly independent on its conjugation, we ran a mass spectrometric analysis searching for ISG15-interacting proteins potentially involved in the increased replication rate. Although very few chromatin-associated factors were found, three of them (RECQ1, DEK, and SMCE1) are proteins known to bind four-way junction structures, such as reversed replication forks that typically form in conditions of replication stress (Quinet et al., 2017). In the present study, we focused on the DNA helicase RECQ1, as its role in replication fork restart upon fork stalling was already established (Berti et al., 2013). Although a direct interaction between ISG15 and RECQ1 proved difficult to monitor and study, the two proteins were detected in close proximity at replication forks and displayed a clear functional interaction, as the accelerated replication fork progression induced by high ISG15 levels is fully dependent on RECQ1. However, as suggested by the mass spectrometry data, ISG15 may have a more general role in regulating DNA replication at these specific replication intermediates (i.e., the four-way junctions). Despite their close proximity, it is still unclear how ISG15 affects RECQ1’s activity. Although other scenarios cannot be currently excluded, we envision that ISG15 may directly modulate the ability of RECQ1 to bind and unwind different DNA structures at stalled forks; alternatively, by binding to stalled forks, ISG15 may favor RECQ1-mediated fork restart. Further studies will need to address whether and how ISG15 regulates formation, stability, and/or restart of reversed replication forks and whether additional factors are involved.

The detrimental effects of high ISG15 levels on DNA replication and genome stability recall recent data showing a high speed of fork progression upon PARPi (Maya-Mendoza et al., 2018). Albeit not directly discussed in this report, the effect of PARPi effect on DNA replication rate also relies on deregulation of RECQ1, which is a key target of PARP1 negative regulation in this context (Berti et al., 2013). In line with this, we observed that the effect of ISG15 on replication fork progression was not further increased by treatment with PARPi (data not shown), suggesting that ISG15 and PARPi may affect replication fork progression with a similar mechanism. As also supported by recent studies using mouse models of chemoresistance (Gogola et al., 2019), PARP1-modulated function of RECQ1 in replication fork restart is emerging as a key regulatory process for the efficacy of anticancer treatments targeting replication. The increased activity of RECQ1 by high ISG15 levels may thus represent an important vulnerability that can be exploited for genotoxic anticancer treatments. Furthermore, the evaluation of ISG15 levels in tumor samples may represent a predictive parameter to stratify patients in personalized cancer therapy.

Why did cells evolve such a potentially harmful system to counteract infection? A possible safe-lock strategy resides on the limited availability of free intracellular ISG15. During pathogen infection, the activation of the IFN pathway leads to the expression of several genes, including the ISG15 conjugation machinery (E1, E2, and E3) that promotes extensive ISGylation of its target proteins and consequently reduces the free intracellular pool of ISG15. The detrimental effects may thus arise only when ISG15 expression and conjugation are uncoupled, resulting in abnormally high levels of free ISG15, which promote deregulated replication events.

Our data uncover the first physiological response (i.e., IFN-β stimulation) directly modulating the velocity of replication fork progression, via the induction of ISG15 expression. Recent literature demonstrates that defects in processing DNA replication stalled forks lead to accumulation of cytosolic DNA and to activation of the cGAS–STING pathway, resulting in the activation of the type I IFN pathway with consequent expression of ISG15 (Coquel et al., 2018). Thus, ISG15 overexpression appears as central player in this emerging field, being at the same time a modulator of DNA replication fork speed and a consequence of replication stress.

Overall, ISG15 function in regulating replication fork progression and genome stability contributes to explain the complex role of the IFN system, and of ISG15 itself, in tumorigenesis and cancer therapy (Han et al., 2018). A key challenge for future studies will be to understand if ISG15, via interaction or conjugation to protein partners, plays additional roles in the maintenance of genome integrity, impacting other fundamental aspects of DNA replication and/or the DNA damage response and repair, and whether its expression levels might be used to predict sensitivity to therapeutic treatments.

Blasticidin (InvivoGen; catalog number ant-bl-1), hygromycin B (InvivoGen; catalog number ant-hg-05), doxycycline (Sigma-Aldrich; catalog number D9891), Geneticin (G418; Thermo Fischer Scientific; catalog number 10131035), EdU (Sigma-Aldrich; catalog number 900584), CldU (Sigma-Aldrich; catalog number C6891), IdU (Sigma-Aldrich; catalog number I7125-5G), ProLong Gold Antifade Reagent (Thermo Fisher Scientific; catalog number P36930), VECTASHIELD Antifade Mounting Medium (Vector Laboratories; catalog number H-1200), SYBR Gold Nucleic Acid Gel Stain (Invitrogen; catalog number S11494), CPT (Sigma-Aldrich; catalog number C9911), HU (Sigma-Aldrich; catalog number H8627), cisplatin (CDDP; Sigma-Aldrich; catalog number 232120), Nocodazole (Sigma-Aldrich; catalog number M1404), Olaparib (Selleckchem; catalog number S1060), SeaPlaque GTG Agarose (Lonza; catalog number 50111), FuGENE HD Transfection Reagent (Promega; catalog number E2311), and JetPRIME transfection reagent (Polyplus; catalog number 114–07) were used.

MCF7, HeLa, T98G, M059K, and HEK293T cells were grown in DMEM, 10% FBS. The U2OS FIT cell line was grown in DMEM, 10% FBS tetracycline-free, complemented with blasticidin (10 µg/ml) and hygromycin B (100 µg/ml) when bearing integration.

U2OS FIT cells were lysed in plates with hypotonic lysis buffer A (0.01 M Hepes, pH 7.5, 0.05 M NaCl, 0.3 M sucrose, 0.5% Triton X-100, protease inhibitor cocktail [1:100], 1 mM PMSF, 10 mM N-Ethylmaleimide (NEM), 10 µM PJ-34, and 75 µM Tannic acid) and incubated 15 min on ice. Cytosolic fraction was isolated after centrifuging for 5 min at 1,500 g. Samples were incubated for 10 min in nuclear lysis buffer B (0.01 M Hepes, pH 7, 0.2 M NaCl, 1 mM EDTA, 0.5% NP-40, protease inhibitor cocktail [1:100], 1 mM PMSF, 10 mM NEM, 10 µM PJ-34, and 75 µM Tannic acid). Samples were centrifuged for 2 min at 16,000 g,and the nuclear fractions were isolated. Pellets were resuspended in chromatin lysis buffer C (0.01 M Hepes, 0.5 M NaCl, 1 mM EDTA, 1% NP-40, protease inhibitor cocktail [1:100], 1 mM PMSF, 10 mM NEM, 10 µM PJ-34, and 75 µM Tannic acid), sonicated for 15 min at low amplitude, and centrifuged at 16,000 g, and the chromatin fractions were isolated from the supernatant.

HEK293T cells, transfected with plasmids coding for myc-ISG15 and HA-FLAG-RECQ1 or the EV, were lysed in YY buffer (50 mM Hepes, pH 7.5, 150 mM NaCl, 10% glycerol, 1% Triton X-100, 1 mM EDTA, and 1 mM EGTA) added with protease inhibitor cocktail, 1 mM PMSF, 10 mM NEM, 50 mM NaF, and 10 mM NaPyr, benzonase (100 U/ml). Cell extracts were incubated for 2 h at 4°C on a wheel with anti-HA resins (Sigma-Aldrich). Resins were washed four times with HNTG buffer (20 mM Hepes, pH 7.5, 150 mM NaCl, 10% glycerol, and 0.1% Triton X-100) before elution with Laemmli buffer at 95°C.

ISG15 KO was generated in U2OS FIT (Thermo Fisher Scientific; catalog number R78007) and MCF7 cells using a CRISPR/Cas9^D10A nickase system as described previously (Chiang et al., 2016). Short guide RNAs (sgRNAs) were designed using UCSC Genome Browser (GRCh38/hg38) mapping at the end of the gene between the intron/exon region shared among all ISG15 splice variants (chromosome 1: 1,013,559–1,013,605, 47 bp in GRCh38/hg38) using the online tools CRISPR Design (http://www.crispr.mit.edu) and WTSI Genome Editing (http://www.sanger.ac.uk/htgt/wge/). sgRNA sense (5′-TTA​CCA​TGG​CTG​TGG​GCT​GT-3′) and sgRNA antisense (5′-CAG​ATG​TCA​CAG​GTG​GGG​GG-3′; Microsynth). sgRNAs were cloned into All-in-One EGFP vector (AIO-GFP; Addgene; Steve Jackson Lab Plasmids, #74119) that was transfected into cells with FuGENE transfection reagent. GFP-positive cells were sorted by FACS and single-cell plated in 96-well plates. Clones were grown and ISG15 expression was tested by immunoblotting. 42 and 41 single clones were screened for U2OS FIT and MCF7, respectively.

Asynchronous, subconfluent cells were labeled for 30 min with 30 µM of the thymidine analogue CldU and then washed with warm PBS and labeled for 30 min with 250 µM of another thymidine analogue, IdU, alone or in combination with genotoxic agents (50 nM CPT, 1 µM CDDP, or 0.5 mM HU). To evaluate fork restart, after CldU labeling, cells were released for 4 h in fresh media containing 4 mM HU, washed with PBS, and then labeled with IdU. Cells were collected in cold PBS, mixed 1:2 with unlabeled cells, and lysed for 8 min in 200 mM Tris-HCl, pH 7.5, 50 mM EDTA, and 0.5% [wt/vol] SDS directly on a glass slide. Slides were tilted at a 45°C to stretch the DNA fibers, air-dried, and fixed in 3:1 methanol/acetic acid overnight at 4°C. The DNA fibers were denatured for 80 min with 2.5 M HCl and blocked for 40 min with 2% BSA/PBS-Tween. Incorporated CldU and IdU tracks were stained for 2.5 h with anti-BrdU primary antibodies recognizing CldU (Abcam; catalog number ab6326) or IdU (BD Biosciences; catalog number 347580) and stained for 2 h with secondary antibody anti-mouse Alexa Fluor 488 (Invitrogen) and anti-rat Cy3 (Jackson ImmunoResearch; catalog number 712–166-153). Slides were mounted with ProLong Gold Antifade Reagent. Microscopy was done using an Olympus IX81 microscope with a charge-coupled device camera (Hamamatsu). To assess fork progression, IdU and CldU track length of DNA fiber molecules was measured using ImageJ64 software. Value in pixels was converted to micrometers considering the objective lens used during acquisition (63×; conversion factor: 1 pixel = 10.54 µm). Length was converted into fork speed considering that 1 µm DNA is composed of ∼2.59 kb (Jackson and Pombo, 1998), and nucleotide incorporation lasted for 30 min.

U2OS FIT cells (and U2OS FIT ISG15/KO cells) carrying EV, FLAG-ISG15, or FLAG-ISG15ΔGG were generated by Flp recombinase–mediated integration (Thermo Fisher Scientific; Flp-In T-REx Core Kit, catalog number K650001). Briefly, cells were transfected with pcDNA5 EV, pcDNA5 FLAG-ISG15, or pcDNA5 FLAG-ISG15ΔGG using FuGENE Transfection Reagent together with pOG44 vector (Thermo Fisher Scientific; catalog number V600520; 9:1 pOG44/pcDNA5) and 48 h after transfection selected for 2 wk with hygromycin B. Single clones were isolated in plate with cloning cylinders and expanded. ISG15 expression was tested by immunofluorescence after 48 h of doxycycline (1 µg/ml). Five clones that were comparable in ISG15 expression were pooled together and used for experiments where we referred to them as EV, ISG15, and ISG15ΔGG. MCF7 ISG15/KO cells were transfected with pcDNA3.1 empty, pcDNA3.1 FLAG-ISG15 using FuGENE Transfection Reagent. 24 h after transfection, antibiotic Geneticin (800 µg/ml) was added onto the cells to select cells that randomly integrated the vector in the genome.

Cells were labeled for 30 min with 10 µM EdU, harvested, and fixed for 10 min in 4% formaldehyde/PBS. Cells were blocked for 15 min with 1% BSA/PBS, pH 7.4. Incorporated EdU was labeled with click reaction according to the manufacturer’s instructions (Thermo Fisher Scientific; Click-iT Plus EdU Cell Proliferation Kit for Imaging, catalog number C10640). Total DNA was stained with 1 µg/ml DAPI. Samples were treated for 15 min with 100 µg/ml RNaseA and analyzed on an Attune NxT Flow Cytometer (Thermo Fisher Scientific) and analyzed using FlowJo software V.10.0.8 (FlowJo).

Whole-cell extracts were prepared in lysis in buffer 1% SDS and 50 mM Tris-HCl, pH 8.0, prewarmed at 95°C. After sonication and clearing (15 min, 16,000 rcf), lysates were analyzed by SDS-PAGE. 20 µg protein was solved in 8% or 12% acrylamide gel and transferred onto a nitrocellulose membrane. The following antibodies were used for immunoblotting: human-ISG15 1:1,000 (provided by K.P. Knobeloch), RECQ1 1:40,000 (Bethyl Laboratories; catalog number A300-450A), FLAG M2 1:1,000 (Sigma-Aldrich; catalog number F3166), pSTAT1 (Tyr701) 1:1,000 (Cell Signaling; catalog number 7649), STAT1 p84/p91 1:1,000 (Santa Cruz; catalog number sc-464), tubulin 1:8,000 (Sigma-Aldrich), Lamin A 1:1,000 (Sigma-Aldrich; catalog number L1293), GAPDH 1:50,000 (Millipore; catalog number MAB374), H3 1:5,000 (Abcam; catalog number ab1791), PCNA 1:1,000 (Santa Cruz; catalog number sc-56), and myc (9E10) 1:1,000 (Santa Cruz; catalog number sc-40).

Chromatin fractions, prepared as above without the sonication step, were extracted from U2OS FIT cells expressing the EV or FLAG-ISG15ΔGG (72 h of doxycycline induction, 1 µg/ml). 2 mg chromatin fractions were incubated for 2 h 4°C on a wheel with anti-FLAG resins (Sigma-Aldrich; catalog number A2220). Resins were washed four times with HNTG buffer (20 mM Hepes, pH 7.5, 150 mM NaCl, 10% glycerol, and 0.1% Triton X-100) before elution with 0.1 M glycine (pH 2.2) for 5 min on ice. Eluted samples were precipitated in 10% TCA and washed twice with cold acetone. Dry pellets were dissolved in a buffer containing 10 mM Tris and 2 mM CaCl₂, pH 8.2, and then 0.5 µg of trypsin was added. After digestion, samples were dried, dissolved in 20 µl 0.1% formic acid, and subjected to liquid chromatography with tandem mass spectrometry (LC/MS/MS) at the Functional Genomic Center Zurich. All MS/MS samples were analyzed using Mascot (Matrix Science, London, UK; version 2.5.1.3). Mascot was set up to search the SwissProt_autoup_20180912 database (selected for Homo sapiens, unknown version, 20,395 entries). Mascot was searched with a fragment ion mass tolerance of 0.030 D and a parent ion tolerance of 10.0 PPM. Oxidation of methionine was specified in Mascot as a variable modification. Scaffold (Proteome Software; version Scaffold_4.8.9) was used to validate MS/MS based peptide and protein identifications. Peptide identifications were accepted if they could be established at greater than 21.0% probability to achieve a false discovery rate (FDR) less than 0.1% by the Scaffold Local FDR algorithm. Protein identifications were accepted if they could be established at greater than 94.0% probability to achieve an FDR <1.0% and contained at least two identified peptides. Protein probabilities were assigned by the Protein Prophet algorithm (Nesvizhskii et al., 2003). Proteins that contained similar peptides and could not be differentiated based on MS/MS analysis alone were grouped to satisfy the principles of parsimony. Proteins sharing significant peptide evidence were grouped into clusters.

U2OS FIT and MCF7 cells were treated for 2 h in complete media complemented with 30 U/ml IFN-β (PeproTech; catalog number 300-02BC), washed with warm PBS, and released in fresh media for 46 h unless otherwise explained; ISG15 induction and IFN-β activity (phosphorylation of Tyr 701 of STAT1, pSTAT1; Cell Signaling; catalog number 7649) were tested by immunoblotting.

HEK293T cells were transfected with calcium phosphate using 2 µg pcDNA3 carrying EV, FLAG-ISG15 WT, or FLAG-ISG15 mutants (ΔGG, L72A, LLVAAA) with or without ISGylation machinery components: 4 µg UBE1L (E1), 2 µg of UBCH8 (E2), and 4 µg HERC5 (E3) and then collected for Western blot analysis after 48 h.

iPOND was performed as previously described (Sirbu et al., 2011). Briefly, exponentially growing cells HEK293T cells were transfected with EV or myc-ISG15 with calcium phosphate. After 48 h, cells were incubated for 10 min with 10 µM EdU, cross-linked with 1% formaldehyde, harvested, and permeabilized. For thymidine-chase controls (Thy-chase), cells were incubated for 10 min in 10 µM EdU, washed and incubated for 1 h with medium containing 10 µM thymidine, and then cross-linked for 5 min with 1% formaldehyde in PBS. Biotin azide was covalently attached to EdU within newly replicated DNA using Click reaction, and EdU-containing DNA was precipitated using streptavidin–agarose beads (Millipore; catalog number 69203). EdU coprecipitates were boiled in Laemmli buffer and then analyzed by immunoblotting.

Asynchronously and subconfluent cells were incubated for 16 h in fresh medium containing 200 ng/ml nocodazole. Cells were collected by trypsinization and swollen for 20 min at 37°C with 75 mM KCl. The swollen mitotic cells were fixed in 3:1 methanol/acetic acid solution and subsequently spread dropwise onto prehydrated glass microscopy slides and air-dried overnight. The slides were mounted the following day using VECTASHIELD Antifade Mounting Medium with DAPI. Images of randomly selected metaphases were acquired by Leica DM6 B upright digital microscope equipped with a DFC360 FX Leica camera. Images were analyzed using ImageJ64, and chromatid breaks/gaps were counted.

HEK293T cells were cotransfected with HaloTag-Fusion construct (2 µg; Promega) and varying concentrations of Nluc-fusion vector (MDM2, 200 ng; RECQ1, 20 ng; Promega). After 20 h, cells were mixed with the HaloTag binding ligand 618 and replated 2 × 10⁴ cells per well on a white flat-bottom 96-well plate and incubated for 18–24 h. Luciferase substrate was added (Furimazine) to each well, and then Luminescence signal (counts/second) was measured using the Tecan Spark Machine. NanoBRET ratio was calculated by dividing the acceptor signal by the donor signal. Subsequently, the no-ligand control was subtracted from the ligand sample.

U2OS FIT were induced for 7 d with doxycycline 1 µg/ml and treated for 1 h with 50 nM CPT and washout (3 h). Cells were collected and resuspended in cold PBS. 2 × 10⁴ cells were mixed with 0.8% wt/vol low melting point, previously equilibrated to 37°C, and then spread onto a comet slide (Trevigen; CometAssay Kit, catalog number 4250–050-ESK). Slides were incubated for 20 min at 4°C to allow solidification of the low melting point. Cells were lysed overnight in lysis buffer (Trevigen). Slides were incubated in cold electrophoresis buffer (300 mM sodium acetate and 100 mM Tris, pH 8.3) for 1 h at 4°C and then subjected to electrophoresis for 30 min at 21 V/300 mA. Samples were rinsed twice in water, fixed in 70% ethanol for 20 min at 4°C, and then dried at 37°C. DNA was stained with SYBR Gold (Thermo Fisher Scientific). Microscopy was performed on a Leica DM6 B upright digital research microscope equipped with a DFC360 FX Leica camera at 10× magnification. The images were analyzed using the Open Comet plugin (http://www.cometbio.org) for Fiji.

FIT cells were induced 48 h with doxycycline (1 µg/ml) and grown on sterile 12-mm diameter glass coverslip, washed with cold PBS, and preextracted and fixed in 100% cold MeOH for 10 min. After washing three times with PBS, cells were permeabilized for 10 min at room temperature in 0.3% Triton X-100 in PBS and washed twice in PBS. Coverslips were then incubated with primary antibodies overnight: FLAG 1:1,000 (Sigma-Aldrich; catalog number F7425), PCNA 1:100 (Santa Cruz; catalog number sc-56), ISG15 1:1,000 (kindly provided by K.P. Knobeloch, Freiburg, Germany), and RECQ1 (A-9) 1:500 (Santa Cruz; catalog number sc-166388). 25 µM EdU was added to media for 10 min before fixing (10 min) MeOH and permeabilizing (10 min) with 0.3% Triton X-100; EdU was linked to biotin-NaN3 with click chemistry and then immunolabeled with anti-BIOTIN antibody Mo 1:500 (Jackson ImmunoResearch, Catalog number 200–002-211). After PBS washes, cells were incubated with mouse PLUS probe (Sigma-Aldrich, Catalog number DUO82001) and rabbit MINUS probe (Sigma-Aldrich; catalog number DUO82005) for 1 h at 37°C. Ligation was performed in ligation buffer (Sigma-Aldrich; catalog number DUO82009-1000Rxn) with Ligase (Sigma-Aldrich; catalog number DUO82027-1EA) for 30 min at 37°C and followed by amplification using Amplification Buffer Far Red (Sigma-Aldrich; catalog number DUO82028Rxn) and Polymerase (Sigma-Aldrich; catalog number DUO82028-1EA) for 100 min at 37°C. Cells were then washed with 0.2 M Tris, 0.1 M NaCl buffer and incubate for 15 min at RT with DAPI (0.5 µg/ml). Coverslips were mounted using ProLong Gold Antifade Reagent.

Automated multichannel wide-field microscopy for quantitative image-based cytometry (QIBC) was performed on an Olympus ScanR Screening System equipped with wide-field optics, a UPLSAPO 20× (0.75 NA), an inverted motorized Olympus IX83 microscope, a motorized stage, infrared (IR)-laser hardware autofocus, a fast emission filter wheel with single-band emission filters, and a 12-bit digital monochrome Hamamatsu ORCA-FLASH 4.0 V2 sCMOS (scientific complementary metal–oxide–semiconductor) camera (2,048 × 2,048 pixels). Images containing ≥1,000 cells per condition were acquired under nonsaturating conditions, and identical settings were applied to all samples within one experiment. Images were processed and analyzed with the inbuilt Olympus ScanR Image Analysis Software (version 3.0.0), a dynamic background correction was applied, nuclei segmentation was performed using an integrated intensity-based object detection module using the DAPI signal, and foci segmentation was performed using an integrated spot-detection module. Fluorescence intensities were quantified and are depicted as arbitrary units. These values were exported and analyzed with Spotfire data visualization software (TIBCO software version 7.0.1; https://www.tibco.com/products/tibco-spotfire). Within one experiment, similar cell numbers were compared for the different conditions. To visualize discrete data in scatterplots, mild jittering was applied to demerge overlapping data points. Representative scatterplots and quantifications of independent experiments are shown.

Cells were plated and transfected the following day with siRNA oligonucleotides targeting ISG15 (5′-GCA​ACG​AAU​UCC​AGG​UGU​C-3′), RECQ1 (5′-UUACCAGUUACCAGCAUUAUUdTdT-3′), or luciferase (5′-CUUACGCUGAGUACUUCGAdTdT-3′) at a final concentration of 40 nM. Transfections were performed using JetPRIME according to the manufacturer’s instruction 48 h after transfection for ISG15 knockdown. For RECQ1 knockdown, transfection was repeated after 24 h. Transfection medium was replaced with complete medium after 24 h and protein depletion confirmed 48 h after the second transfection by immunoblotting.

pcDNA3.1 FLAG-ISG15 was mutagenized using the following primers: L10A forward, 5′-GAC​CTG​ACG​GTG​AAG​ATG​GCG​GCG​GGC​AAC​GAA​TTC​C-3′; L10A reverse, 5′-GGA​ATT​CGT​TGC​CCG​CCG​CCA​TCT​TCA​CCG​TCA​GGT​C-3′; L72A forward, 5′-GGC​CCC​GGC​AGC​ACG​GTC​GCG​CTG​GTG​GTG​GAC​AAA​TG-3′; L72A reverse, 5′-CAT​TTG​TCC​ACC​ACC​AGC​GCG​ACC​GTG​CTG​CCG​GGG​CC-3′; V74A forward, 5′-GGC​AGC​ACG​GTC​CTG​CTG​GCG​GTG​GAC​AAA​TGC​GAC​G-3′; and V74A reverse, 5′-CGT​CGC​ATT​TGT​CCA​CCG​CCA​GCA​GGA​CCG​TGC​TGC​C-3′; pcDNA3.1 FLAG-ISG15L10A was mutagenized using the following primers: 3X forward, 5′-GGC​AGC​ACG​GTC​GCG​CTG​GCG​GTG​GAC​AAA​TGC​GAC​G-3′; 3X reverse, 5′-CGT​CGC​ATT​TGT​CCA​CCG​CCA​GCG​CGA​CCG​TGC​TGC​C-3′. PfuTurbo DNA polymerase (Agilent Technologies; Pfu, catalog number 600250) was used for the reaction and checked on a 0.8% agarose gel. Template was digested with 1 µl DpnI (New England Biolabs; catalog number R0176S) and TOP10 (bacteria; Thermo Fisher Scientific; catalog number C4040-03) were transformed with mutagenized vector overnight at 37°C. Mutated vectors were sequenced using cytomegalovirus (CMV) primer (5′-CGC​AAA​TGG​GCG​GTA​GGC​GTG-3′).

Cells were seeded in 96-well plate and treated the day after with the genotoxic agents CPT (1, 5, 25, 50, and 150 nM) and Olaparib (0.5, 1, 5, and 10 µM) for 7 to 10 d. Resulting colonies were fixed with 100% cold MeOH and stained with 0.05% crystal violet in 100% MeOH for up to 2 h. Excess was washed out and cells were destained for 30 min with MeOH. Crystal violet in suspension was analyzed by measuring the absorbance at 600 nm using SpectraMaxi3.

The number of forks (fiber assay), nuclei (comet assay) or metaphase scored in the shown replicate and number of biological replicates is defined in the figure legends. Results were analyzed in GraphPad using a Mann–Whitney U test (two-tailed P value; P value > 0.05 was considered not significant). Flow cytometry data were analyzed using FlowJo software V.10.0.8 (https://www.flowjo.com/). The intensity values of EdU-positive cells per sample were extracted from the raw data and subjected to statistical analysis using GraphPad Prism 7 (two-tailed P value). In the neutral comet assay, double strands were evaluated measuring Olive tail moment (Olive moment), a parameter that includes the tail length and the fraction of total DNA in the tail, using the Open Comet plugin (http://www.cometbio.org/) for Fiji. The results were analyzed using GraphPad Prism7 using a Mann–Whitney test. Results were displayed as scatterplots with mean and SD. In the clonogenic assay, absorbance of each sample (technical triplicate) was normalized on untreated samples and on the corresponding treated EV sample.

Fig. S1 shows the pipeline for the generation of the ISG15 KO, analysis of representative clones in U2OS, and additional experiments on the ISG15 localization at the replication forks and the effect of its deregulated expression on DNA synthesis. Fig. S2 shows analysis of representative ISG15 KO clones in MCF7 and expression of ISG15 in different cancer cell lines. Fig. S3 reports additional experiments on the effect of the expression of the conjugation-defective form of ISG15 (ISG15ΔGG) in DNA synthesis and cell cycle profile, as well as a sequence alignment of human ISG15 and ubiquitin and further characterization of the ISG15 mutants tested in the replication phenotype. Fig. S4 shows the immunoprecipitation experiment performed for mass spectrometry studies, additional details of the ISG15–RECQ1 interaction, and the detrimental effects of expression of ISG15 and ISG15ΔGG in DNA replication fork progression and DNA damage upon a mild dose of replication stress. Table S1 reports the interacting factors identified by mass spectrometry.

We thank Massimo Lopes for critical reading of the manuscript and all members of Penengo and Lopes laboratories for technical support and helpful discussions, Jana Krietsch and Davide Eletto for helping with cell sorting and screen of KO cell lines, Alessandro Vindigni (Washington University, St. Louis, MO) for providing RECQ1 constructs, and the Functional Genomic Center of the University of Zurich for mass spectrometry analysis.

This work was supported by Helmut Horten Stiftung, Krebsliga Schweiz (KFS-4577-08-2018), and Swiss National Science Foundation research grants 310030_184966 and 31003A_166370 to L. Penengo.

The authors declare no competing financial interests.

Author contributions: M.C. Raso performed DNA fiber experiments, metaphase spreads, and comet and survival assays; M.C. Raso and N. Djoric generated the ISG15 KO cell lines and performed iPOND and cell fractionation and PLA studies; M.C. Raso performed the statistical analysis and contributed to the experimental design; F. Walser performed QIBC analysis; F.M. Schmid prepared samples for the mass spectrometry analysis; S. Burger for technical assistance; S. Hess and K.-P. Knobeloch performed the NanoBRET analysis; and L. Penengo conceived the project, designed experiments, and wrote the manuscript, supported by M.C. Raso.