Abstract
The alternative reading frame (ARF) protein is crucial in the cellular response to oncogenic stress, being likewise the second most frequently inactivated gene in a wide spectrum of human cancers. ARF is usually sequestered in the nucleolus by the well-known oncogenic nucleophosmin (NPM) protein and is liberated in response to cell damage to exhibit its tumor-suppressor ability. However, the mechanism underlying ARF activation is unknown. Here we show that mitochondria-to-nucleus translocation of cytochrome c upon DNA damage leads to the break-off of the NPM-ARF ensemble and subsequent release of ARF from the nucleoli. Our structural and subcellular data support a molecular model in which the hemeprotein triggers the extended-to-compact conformation of NPM and competes with ARF for binding to NPM.
Main Text
The alternative reading frame (ARF) protein of the INK4a-ARF gene constitutes a pivotal stress signaling and response pathway (1). Stress raises DNA damage levels in cells, in turn activating the ARF/p53 route, which initiates adaptive responses (2). INK4a-ARF is the second most frequently inactivated gene in human cancers (3). In mammalians, ARF (p19ARF in mice, p14ARF in humans) regulates p53 levels by inactivating the p53 ubiquitin ligase HDM2 (human double minute clone 2, MDM2 in mice) (4).
ARF is however usually sequestered in the nucleolus by nucleophosmin (NPM) – a.k.a. B23, NO38 or numatrin. NPM (5) is a histone chaperone that participates in ribosome synthesis (6), cytoplasmic-nuclear shuttling (7), chromatin remodeling (8) and mitotic spindle assembly (9). NPM is also a key factor favoring DNA repair and promoting cell survival (10). Expectedly, NPM mutations or dysregulation often correlate with oncogenesis (5,10). Under homeostasis, NPM sequesters ARF in the nucleolus, thus enabling p53 ubiquitination and subsequent proteasomal degradation (3,11,12). Upon stress and/or DNA damage, NPM releases ARF, which then binds to HDM2 to diminish p53 degradation (10). In fact, cells enter apoptosis after knocking down NPM (13). However, the molecular mechanism triggering NPM dissociation from ARF remains unknown.
In this work, NPM emerges as a novel target of mitochondrial cytochrome c (Cc) in human cells undergoing DNA damage. A recent proteomic analysis showed that Cc targets several histone chaperones upon DNA damage (14,15). Indeed, Cc moves into the cell nucleus and binds SET/TAF-Iβ once DNA lesions are detectable without stimulation of death receptors or stress-induced pathways (16). In plants, Cc likewise targets histone chaperone NRP1, analogue to human SET/TAF-Iβ, under similar conditions (17,18).
Here, we found that mitochondrial endogenous Cc is translocated into the nucleus of tumor and non-tumor cells following DNA breaks. Further, Cc competes with ARF for binding to the N-terminal oligomerization domain of NPM. To understand the molecular basis of such competition, we determined the structure of the complex between NPM and Cc. Notably, Cc lodges in the cavity formed by the NPM arms, the acidic disordered middle region of NPM, stabilizing the NPM-Cc complex. Our biophysical and structural analyses reveal that a single molecule of Cc can release up to five molecules of ARF out of NPM. Further liquid-liquid phase separation assays demonstrate that NPM undergoes phase transitions via heterotypic interactions with Cc, as does NPM with ARF. Noteworthy, Cc contains lysine-rich motifs rather than the canonical, linear arginine-rich motifs of well-known membrane-less organelle components, such as ARF, and forms nucleolar-like droplets with NPM. Even more interesting is the fact that Cc can enter the nucleoli to break off the preformed NPM-ARF ensembles and induce ARF release, thereby explaining the molecular mechanism underlying ARF activation upon DNA damage.
Nucleophosmin as a nuclear target of cytochrome c upon DNA breaks
First, to ascertain whether mitochondrial Cc can reach nuclei upon DNA damage, we immunodetected endogenous Cc in cells treated with either camptothecin (CPT) or doxorubicin. Immunostaining of HeLa cells treated with these drugs revealed the release of Cc from mitochondria to the cytoplasm and the nucleus (Fig. 1a, upper). Mitochondria were labeled with mitotracker—whose accumulation depends on mitochondrial transmembrane potential—because Cc release from mitochondria precedes the loss of such potential, as previously described (19) (Fig. 1a and Fig. S1a). We further explored the Cc subcellular location in non-cancer cells treated similarly. We observed a parallel behavior in CV-1 in origin, carrying SV40 cells (Cos-7, Fig. 1a, lower), as well as in mouse embryonic fibroblasts (MEF) (Fig. S1a, left). These results were also evident when applying an alternative immunofluorescence method in HeLa cells (Fig. S1a, right). In a previous report (16), we performed subcellular fractioning assays to demonstrate that endogenous Cc translocates into the nuclei upon DNA breaks induced by CPT, doxorubicin or indotecan. On the contrary, Cc was unable to be translocated into cell nuclei upon treatment with other apoptosis-inducing agents which do not induce DNA damage, namely tumor necrosis factor (TNF)-related apoptosis-inducing ligand (TRAIL) and staurosporine (STP) (16). More interestingly, we showed that Cc translocation into nucleus in response to DNA breaks occurs prior to caspase-3 activation (16).
We extended our previous proteomic analysis of the Cc-centered interactome—based on tandem affinity chromatography, 2D-PAGE and mass spectrometry (14). A C-end cysteine-substituted Cc (E104C) tethered to a thiol-Sepharose 4B matrix allowed sifting extracts from Jurkat cells either untreated or treated for 6 h with 10 μM camptothecin (CPT), which primarily induces single strand breaks that can further be converted to DNA double strand breaks (DSBs) (20). 2D-PAGE experiments enabled us to detect up to 21 putative targets, accessible to Cc upon DNA damage but not under homeostasis. Mass spectrometry analysis identified the spot marked in Fig. 1b in red as NPM.
Bimolecular fluorescence complementation (BiFC) experiments (21) confirmed the interaction in cells. A construct of Cc fused to the C-end domain of the yellow fluorescent protein (YFP) was available at our lab (14), and we built another one encoding full length NPM (hereafter NPM(1-294)) ahead of the N-end fragment of the YFP (nYFP). After co-transfection, 10 μM CPT was added to the culture. The fluorescence pattern and DAPI overlay (Fig. 1c) clearly show that NPM interacts with Cc in both the nucleus and the cytoplasm. Highly fluorescent spots within nuclei pinpoint condensates of NPM molecules. Protein expression was confirmed by Western blot analysis (Fig. 1d). Control cells co-transfected with bFos and bJun chimeras displayed YFP reconstitution. They did not when using bFosΔZip construct lacking the Jun-binding domain (Fig. S1b).
Cc migrates into the cell nucleus upon DSBs accumulation solely (16). We confirmed the DNA damage after 4 h treatment by probing the accumulation of H2AX (γH2AX) (Fig 1e). Accordingly, phosphorylated histone H2AX (γH2AX), which quickly accrues after DNA damage, is detected in cell lysates 4 h after treatment (Fig 1e). Moreover, we detected endogenous NPM in Western blots of Cc immunoprecipitated cell extracts from HeLa cell line. Co-immunoprecipitation of proteins was observed from both cytoplasmic and nuclear fractions only upon DNA damage, 4 h after adding 20 μM CPT (Fig. 1f), but not from the lysates of untreated cells. When using IgG for IP, none of the NPM-featuring band was detected (Fig. 1f). Immunoblotting against α-tubulin, COX-IV (cytochrome c oxidase subunit IV) and histone H3 antibodies confirmed the purity of subcellular fractions (Fig. 1g).
Cytochrome c competes with p19ARF for NPM binding
NPM participates in DNA damage response triggered by radiation (10). Indeed, NPM modulates the interaction between two p53 regulators –HDM2 and ARF– (12, 13) (Fig. 2a). We hypothesize that Cc plays a role in ARF release form the NPM complex upon DNA damage by competition for the binding site in NPM. To understand the molecular details of both interactions and test the hypothesis we designed several NPM constructs and assessed their ability to bind p19ARF and Cc (Fig. 2b) in vitro. All NPM constructs folded correctly (Fig. S2a-d).
The 1H nuclear magnetic resonance (NMR) signal of the Met80-εCH3 of reduced Cc senses any interaction of this protein with binding partners (22). Notably, this signal broadened beyond detection limits upon addition of 6 μM full-length NPM (NPM(1-294)) onto a 13 μM Cc sample (Fig. 2c). Titration of the p19ARF(1-37) domain — responsible for NPM binding (23-25) — onto the Cc:NPM mixture gradually restored the Met80-εCH3 signal (Fig. 2c). Hence, the p19ARF competes with Cc for NPM binding. Likewise, the Met80-εCH3 signal of Cc disappeared upon adding the NPM oligomerization domain core (NPM (9-122)), known to bind p19ARF (12) showing both proteins, Cc and p19ARF, bind to the oligomerization domain core. Then, addition of p19ARF(1-37) to the mixture partially rescued the NMR signal (Fig. S3). At high p19ARF(1-37) concentrations, protein aggregation prevented further recovery.
Competition between p19ARF(1-37) and Cc for the NPM oligomerization domain (NPM(1-130)) was also observable in electrophoretic mobility shift assays (EMSA). Free NPM(1-130) ran rapidly across the gel, but lagged in presence of p19ARF(1-37) or Cc (Fig. 2d), the latter inducing the largest shift. Interestingly, progressive additions of Cc to the initial p19ARF(1-37):NPM(1-130) mixture resulted in retardation patterns indicating a change in complex composition from p19ARF(1-37):NPM(1-130) to Cc:NPM(1-130) (Fig. 2d). Hence, Cc and p19ARF compete for binding to the oligomerization domain of NPM.
NPM uses its oligomerization domain to recognize Cc and p19ARF
Isothermal titration calorimetry (ITC)-derived dissociation constant (KD) values (Table S1 and Fig. S4a-d) indicate Cc shows similar affinities towards the full-length protein (NPM(1-294)) and its oligomerization core (NPM(9-122)). The Cc:NPM stoichiometry was 2:1 in both cases, with no binding cooperativity indeed. Titration analysis using a general binding model based on the overall binding constants was performed according to ref. 26. Removal of the oligomerization domain in the NPM(123-294) construct, which keeps the histone binding region (27), increases the KD value to 13 μM. Further, Cc shows low affinity (KD ca. 200 μM) to the NPM(225-294) construct— responsible for nucleic acid binding (28) and nucleolar location (29). Hence, Cc preferentially binds NPM by its oligomerization domain and disordered histone binding region. When high ionic conditions were employed, a single binding site for Cc in NPM(1-294) or NPM(9-122) were observed (Table S1 and Fig. S5a-b). Noteworthy, the estimated values for enthalpy and entropy did not contain any extrinsic contribution from buffer de/ionization as a buffer with negligible ionization enthalpy was employed in the ITC experiments. Those parameters can thus be considered as intrinsic thermodynamic parameters for the interaction of Cc with its partners. Interestingly, ITC titrations using oxidized Cc resulted in similar binding affinities as those attained when using reduced Cc (Table S1 and Fig. S5c-d).
Notably, kinetic dissociation constant (koff) values obtained by surface plasmon resonance (SPR) indicate the half-life of Cc:NPM(1-294) complex is several fold that of Cc:NPM(9-122) (Fig. S4e). Indeed, electrostatic interactions between Cc and the A2 region of NPM (see below) would hinder dissociation. NPM tails may affect such kinetics by steric hindrance and/or because the entropic cost of restraining their dynamics.
The interactions between NPM constructs and p19ARF(1-37) yield rather complex thermograms (Fig. S4f-g), displaying three apparent phases when NPM(1-294) is the analyte. Fitting them considering three classes of independent binding sites would bias the analysis of the first titration phase. Restricting the fit to this one, yielded a KD value of 0.28 μM and a stoichiometry of 5.6 using a single class of independent binding sites (Fig. S4f, Table S1). NPM(9-122) binds p19ARF(1-37) with the same affinity as NPM(1-294) (Fig. S4g, Table S1). Despite their different affinities towards NPM, Cc and p19ARF(1-37) may compete during an eventual accrual of Cc in the nucleus after DNA damage (16). Notably, binding of the unstructured and flexible p19ARF(1-37) to the NPM constructs was entropy driven, as found in many complexes involving intrinsically disordered proteins (30-32).
Electrostatic interactions are key in cytochrome c and nucleophosmin complex formation
Because of its high pI, Cc may interact with negative NPM regions. According to KD values, the affinities of Cc towards NPM(1-294) and NPM(9-122) decrease by 12-fold and 3-fold, respectively, upon addition of 0.1 M KCl (Fig. S5a-b, Table S1). In agreement, the affinity loss correlates with increases in binding enthalpy. The larger ionic strength effects observed with NPM(1-294) may reflect electrostatic repulsion between NPM disordered regions aiding Cc access to its acidic binding site at the oligomerization ring and the nearby disordered acidic, histone binding region of NPM (Fig. 2b). At high ionic strength, the stoichiometry of the complex formed by Cc with both, NPM(1-294) and NPM(9-122), was 1:1 (Table S1). Hence, one of the binding sites observed at low ionic-strength values may correspond to unspecific electrostatic interactions.
Heteronuclear single-quantum correlation (HSQC) spectra of 15N labelled Cc samples, before and after the addition of NPM(1-294) (Fig. 3a) allowed us to measure averaged values (Δδavg) for chemical-shift perturbations (CSP). Notably, 11 amide resonances (residues G6, K7, K8, Q16, G24, H33, G77, T78, I81, K86 and E89) shifted substantially (Δδavg > 0.05 ppm) (Fig. 4b). Others underwent smaller yet significant perturbations. Most shifted signals mapped to fringes of the heme cleft (Fig. 3b), as in other interactions with Cc partners (14, 33, 34). Four lysine residues located at the interface. Addition of the N-terminal core domain of NPM (NPM(9-122)) to Cc samples induced significant but smaller changes in 21 amide Cc resonances, consistently with the higher KD value (Fig. S6a) spread across the same surface patch. Signals either shifted or broadened (Fig. S6a-b). Again, the interaction surface includes ionic residues: five lysines (5, 7, 8, 13 and 86) and two acidic residues (E89 and E93).
Conversely, we recorded TROSY (transverse relaxation optimized spectroscopy) spectra of a 2H-15N-labeled sample of NPM(1-130). Some NPM resonances shifted upon Cc addition at a molar Cc:NPM(1-130) ratio of 2:1 (Fig. S7a). Most residues showing significant CSPs clustered at the acidic “top patch” of the oligomerization domain (Fig. 4a,c). Likewise, p19ARF(1-37) protein binds the same region of NPM(1-130) (Fig. 4b), despite CSPs being smaller (Fig. S7b-d). Nevertheless, unlike Cc, p19ARF(1-37) binds the interfaces between the NPM monomers, also interacting with residues at the opposite face of NPM(1-130) (Fig. 4b, indicated by asterisks). This fuzziness, in line with our ITC data, may explain NPM(1-130) undergoing smaller CSPs in the presence of p19ARF(1-37) (Fig. S7c-d). The eventual impact of His-tag on the core structure of NPM can be discarded since no significant structural changes were observed (root-mean-square deviation of backbone atoms is 0.35 Å; Fig. S8) when the two pentameric NPM X-ray structures available at the Protein Data Bank—with (2P1B, 35) and without His-tag (4N8M, 25)—are compared. We thus corroborated that His-tagged NPM(1-130) is properly folded as inferred from NMR and CD (circular dichroism) experiments (Fig. S9a,b). These findings agree with the SDS-PAGE assays performed with His-tagged NPM(1-130) samples, which show an intense majority band corresponding to the pentameric form at any ionic strength (Fig. S9c).
Three-dimensional structures of NPM and the Cc:NPM complex
We resorted to X-ray diffraction to get further structural data of the Cc:NPM complex. Cc and NPM(9-122) —the oligomerization domain core— were co-crystallized using the microbatch technique in 25% PEG 400, 0.1 M MES, 0.1 M MgCl2 and 0.1 M TCEP as an additive. The refined model (PDB 5EHD) contained 20 monomers of NPM(9-122) in the asymmetric unit, arranged in four pentamers as previously observed (35). Residues 9-14, 119-122 and the C-terminal His-tag were invisible in the electron density map. Each NPM(9-122) monomer forms a compact domain comprising an eight-stranded β-barrel of jellyroll topology (Fig. 5a), as previously found (35) (PDB 2P1B). No electron density associated with Cc was noticeable in any of the four NPM pentamers, despite its presence in the co-crystal being supported by SDS-PAGE (Fig. S10).
One face of the NPM(9-122) pentamer contains the N- and C-termini (top view of Fig. 5a), and the flexible loops connecting β2 and β3 (highlighted by dotted lines) comprising the residues D34 to E39 of the acidic tract A1 (Fig. 2b). These loops face Cc in our NMR analysis (Fig. 4a) and are evident in the Cc:NPM(9-122) electron density map whereas being invisible in the published structure of free NPM(9-122) (35). The lack of Cc electron density in the maps suggests this protein assumes multiple orientations across the crystal.
We then resorted to electron microscopy (EM) to get structural information on NPM(1-294), as well as the complexes Cc:NPM(1-294) and Cc:NPM(9-122). Because of their small size and flexibility, negative staining was used. This technique, because of its low-resolution, cannot describe the nature of the interaction between NPM and Cc but has nevertheless provided us with clear information on where Cc binds. The images of NPM(1-294) (162 kDa) revealed a major, end view often showing a clear five-fold symmetry (Fig. S11a) and a scant one revealing a base structure from which several, thin and flexible masses protrude. 28637 particles were used for 3D reconstruction. C5 symmetry was applied since NPM is a homopentamer. The volume obtained showed (Fig. 5b) a base of ∼5 nm width and 4.5 nm height from which five arms of 5 nm length protrude, with a ∼3 nm globular mass at the tip of each arm. Our crystal structure of NPM(9-122) fits well into this 3D reconstruction. Likewise, the atomic structure of the last 51 residues of NPM (36) (PDB 2VXD, residues 243-294) docks well into the globular arm-tip domains (Fig. 5e). The disordered middle region of NPM (residues 123-242) links the two structural domains (Fig. 2b). As expected for an EM reconstruction, the NPM His-tag cannot be visualized due to low resolution and its high mobility.
When staining a Cc:NPM(1-294) mixture, EM views were very similar to apo-NPM(1-294) ones, displaying the two orientations above (Fig. S11b). Outwardly, no obvious differences marked the Cc molecule bound to the NPM(1-294) pentamer, indicating the Cc molecule lodges within the cavity formed by the NPM(1-294) arms. The 3D reconstruction of the complex, given its small size (174 kDa) and flexibility, required 29500 particles. We imposed C5 symmetry throughout the 3D reconstruction procedure despite the inaccuracy for reconstructing a single Cc molecule in the complex. Nevertheless NPM(9-122) and NPM(1-294) variants bind Cc with similar affinities (Table 1), suggesting the N-terminal core domain of NPM is the principal Cc interaction site and has five identical, potential binding sites.
The resulting volume (Fig. 5c) resembled that observed for apo-NPM(1-294), displaying a central mass and five protruding arms ending in small globular masses. Noteworthy, there are two important differences between the Cc:NPM(1-294) complex and free NPM(1-294): first, a retraction in the arms (∼2 nm) closer to the core mass and second, a protrusion in the core partially filling the central cavity and touching the base of the arms. This bulge discloses the Cc molecule attached to the central cavity, in contact with the NPM core domain, and induces a contraction of the arms by interacting with their base. Such a contraction seems to induce the lateral protrusion of part of the flexible middle region of the chaperone. Only a single Cc molecule is bound to the NPM(1-294) pentamer in this EM structure (Fig. 5c,f) obtained at high ionic strength — ca. 150 mM before drying. Indeed, ITC data indicate a 1:1 stoichiometry for the interaction between Cc and the NPM(1-294) oligomer at high ionic strength (Table S1).
The atomic structures of the NPM(9-122) pentamer and the C-terminal globular domain dock well into the corresponding positions of the volume (Fig. 5f). NMR-restrained docking calculations enabled us to fit the Cc molecule in the EM protruding mass. CSPs obtained from NMR analysis of the NPM(9-122), complex were used as input data. High-score models placed Cc towards the highly acidic face of NPM(9-122), as shown for the best-scoring solution (Fig. S12). In this one, the heme cleft points to the acidic stretches of NPM(9-122). Docking results were consistent with Cc lodging in the EM protrusion contacting the inner side of the NPM core domain.
Since NPM(9-122) binds Cc with the same stoichiometry as NPM(1-294) (Table S1), we sought to confirm the location of Cc in the complex. Images of negatively stained Cc:NPM(9-122) complex (Fig. 5d) revealed a small, globular structure consistent with the molecular mass of the complex (84 kDa). 21215 particles were used for the 3D reconstruction, and C5 symmetry was again imposed. The volume obtained (Fig. 5d) displayed the base alike the chaperone core NPM(1-294) and the Cc:NPM(1-294) complex. Resembling the Cc:NPM(1-294) complex, a small mass protrudes from the base of the pentamer, which can be clearly attributed to bound Cc. This reading invigorates once NPM(9-122) and Cc dock in their corresponding places (Fig. 5g).
All these findings point to Cc binding with NPM(1-294) through the internal part of the NPM core (NPM(9-122), assisted by some regions at the base of the inner side of the chaperone arms, probably inducing their contraction. These results also explain why, even though the NPM pentamer contains five identical binding sites for Cc, only one molecule binds to the chaperone: once the first Cc molecule binds to one of the five potential binding sites, there is no room for the others.
NPM undergoes phase separation with ARF or Cc
NPM is known to undergo liquid-liquid phase separation via heterotypic interactions with nucleolar components (e.g., proteins displaying arginine-rich (R-rich) motifs) able to interact with the NPM acidic tracts (37,38). Since ARF – which contains R-rich motifs – interacts with NPM, we addressed its ability to promote phase separation of NPM in vitro. Titration of NPM(1-294) with p19ARF(1-37) resulted in formation of phase separated droplets (Fig. 6a, left), as does NPM with human p14ARF to form gel-like condensates (39). This process was observed when the p19ARF(1-37):NPM(1-294) ratio increased up to 5:1, being clearly evident at 10:1 (Fig. 6a, left). NPM(1-294) phase separated with p19ARF(1-37) or Cc also in the presence of 100 mM KCl (Fig. S13). Phase separation occurs after significant saturation in the p19ARF(1-37):NPM(1-294) complex. This could explain the rather complex thermograms showed by NPM in complex with p19ARF(1-37), as discussed above (Fig. S4f-g). We likewise observed phase separation when using Oregon Green 488-labeled NPM(1-294) and p19ARF(1-37) labeled with rhodamine (Fig. 6b, upper). Remarkably, Cc does not contain R-rich motifs but likewise mediated phase separation of NPM(1-294) starting at the 5:1 Cc:NPM(1-294) ratio (Fig. 6a, right), and labeled NPM(1-294) showed phase separation in complex with Alexa Fluor 647-labeled Cc, which exhibits signs of spinodal decomposition (Fig. 6b, lower). The fact that ARF and NPM form heterotypic phase separated droplets is in agreement with ARF sequestration by NPM in the nucleolus (3), whereas the herein reported finding that Cc is able to form phase separated droplets with NPM is consistent with the novel interaction detected between the hemeprotein and NPM in the nucleoli (Fig. 1c).
Cc liberates ARF from the nucleolus
To test the relevance of such in vitro findings under physiological conditions, we monitored how Alexa Fluor 647-Cc could interfere with Oregon Green 488-NPM(1-294) and Rhodamine-p19ARF(1-37) accumulation in isolated mammalian nucleoli. As observed in Fig. 6c (upper), both NPM and p19ARF(1-37) are incorporated within nucleoli in the absence of Cc. However, in the presence of increasing concentrations of Alexa Fluor 647-Cc, the fluorescence of p19ARF(1-37) gradually diminished up to disappear while that of NPM(1-294) remains unalterable (Fig. 6c). Addition of Cc to pre-incubated p19ARF:NPM nucleoli also resulted in a decrease in the p19ARF(1-37) fluorescence over time (Fig. 6d). These results clearly indicate that p19ARF is in fact released from the nucleoli when Cc is incorporated to the membrane-less organelle (Fig. 6e).
Discussion
Nucleoli are key organelles in the cell stress response and different stimuli cause the loss of function of such essential membrane-less structures (40,41). A crucial element in this process is NPM, which is responsive to stress of different kinds (42, 43). NPM regulates the ARF/p53 pathway by keeping ARF sequestered in the nucleoli but releasing it upon DNA damage (12). We herein demonstrate that endogenous Cc is translocated into the nucleus in both cancer and non-cancer cells, thereby interacting with endogenous NPM, upon DNA insults.
At physiological ionic strength, a single Cc molecule binds to the oligomerization domain of NPM. The middle-disordered region of NPM provides a kinetic barrier for the interaction with Cc, without affecting the affinity. Our crystal (Cc:NPM(9-122)) and EM structures (Cc:NPM(9-122) and Cc:NPM(1-294)) disclose that binding of Cc to the NPM oligomerization core strongly affects its loops and the intrinsically disordered middle region of full-length NPM. This region, essential for histone chaperone activity and heterotypic transitions of NPM (37), shrinks and approaches to the core domain to which Cc is bound. The 3D structure herein displays Cc hinders the access to the acidic tracts of the disordered binding region of NPM, to which several partners bind (38). Our biophysical data clearly indicate that Cc can oust p19ARF(1-37) from NPM. The CSPs maps of NPM in the Cc:NPM and ARF:NPM complexes visibly display an overlap between their respective binding sites on NPM.
NPM is a well-known nucleolar marker, which is able to undergo liquid-liquid phase separation (LLPS) by homo- and heterotypic interactions and greatly contributes to the liquid-like features of nucleolus (37,38). Recent data has highlighted the key role of linear R-rich elements in the proteins exhibiting heterotypic interactions with NPM (37), yet K-rich stretches can also be involved in the assembly of other membrane-less organelles (e.g. stress granules) (44). According to Kriwacki and co-workers (37), heterotypic interactions require extended NPM conformations, which do occur at physiological ionic strength. Such interactions are crucial to enable recognition by RNA and assembly of pre-ribosomal particles (37). On the other hand, homotypic interactions have been detected under non-physiological, low-salt conditions, in which a compact conformation of NPM is populated. The balance between both types of interactions is proposed to be essential for keeping the “liquid-like scaffold of the granular component of nucleoli” (37). Our data herein provides a mechanism by which a small, K-rich globular protein such as Cc triggers the transition from an extended state to a compact conformation of NPM under physiological, high salt conditions. In this regard, it is worth noting the differences in shape and physical features between NPM droplets induced by ARF and Cc (Fig. 6a,b). We thus describe Cc as a cofactor modulating the conformational equilibrium of NPM, along with its multi-valence and LLPS mechanism. Cc-induced homotypic condensation of NPM could be a reversible switch for regulation of ribosomal biosynthesis under DNA damage and release of tumor suppressor factors like ARF.
In agreement with the NPM ability to form droplets not only with ARF but also with Cc, the hemeprotein is incorporated into the nucleoli so as to shift ARF out from the preformed NPM-ARF ensembles and from the nucleoli themselves, thereby explaining the molecular mechanism underlying the ARF release and its later activation upon DNA damage.
On the basis of our data, we propose that the Cc:NPM nucleolar interaction, involving the extended-to-compact transition of NPM and release of NPM-sequestered ARF, can be expanded to a general mechanism for DNA damage response in which the K-rich regions of Cc act in a pleiotropic, moonlighting manner on histone chaperones.
Author contributions
K.G.A., A.D.Q., I.D.M. and M.A.R. designed the research; K.G.A., M.A.C.C. carried out the proteomics and cellular experiments. K.G.A., C.A.E.R. and A.V.-Cruz cloned, expressed and purified proteins. K.G.A., A.D.Q. and A.V.-Campoy performed and analyzed ITC measurements. K.G.A., S.G.C., I.A. and I.D.M. performed and analyzed NMR data; K.G.A. run docking calculations. N.B.G. and J.A.H. acquired and analyzed the crystallographic data. E.S., X.S., A.D.Q. and K.G.A. implemented the phase separation assays. R.A and J.M.V. carried out the electron microscopy and the single-particle 3D reconstruction. K.G.A., A.D.Q., I.D.M. and M.A.R. wrote the paper.
Competing interest statements
The authors declare no conflict of interest. Data availability: Additional data are available from the corresponding author upon request.
Supplementary Materials
Materials and Methods
Figures S1 to S13
Tables S1 and S2
Acknowledgements
The authors thank Dr. J. Martínez-Fábregas (Scotland) for kind help with proteomic analyses. Funding: The authors are grateful to the Spanish Ministry of Economy and Competitiveness (BIO2015-70092-R, BFU2015-71017/BMC, BFU2017-90030-P, BFU2016-75984/BMC, PGC2018-096049-B), the European Commission (European Regional Development Fund and European Research Council (CONCERT, contract number 648201), the Regional Government of Andalusia (BIO198), the Ramon Areces Foundation, the NMR Facility at CITIUS (University of Seville) and Waters-TA Instruments. This work has been supported by iNEXT, grant number PID 3407, funded by the Horizon 2020 programme of the European Commission. C.A.E.R., M.A.C.C., A.V.-Cruz and E.S thank Cámara Foundation, the Spanish Ministry of Education, Culture and Sports (FPU18/06577 and FPU16/01513) and La Caixa foundation for their respective PhD fellowships. IRB Barcelona is recipient of a Severo Ochoa Award of Excellence from MINECO (Government of Spain).