NUPR1 is a protumoral multifunctional intrinsically disordered protein (IDP), which is activated during the acute phases of pancreatitis. It interacts with other IDPs such as prothymosin α, as well as with folded proteins such as the C-terminal region of RING1-B (C-RING1B) of the Polycomb complex; in all those interactions, residues around Ala33 and Thr68 (the ‘hot-spot’ region) of NUPR1 intervene. Its paralogue, NUPR1L, is also expressed in response to DNA damage, it is p53-regulated, and its expression down-regulates that of the NUPR1 gene. In this work, we characterized the conformational preferences of isolated NUPR1L and its possible interactions with the same molecular partners of NUPR1. Our results show that NUPR1L was an oligomeric IDP from pH 2.0 to 12.0, as judged by steady-state fluorescence, circular dichroism (CD), dynamic light scattering, 1D 1H-NMR (nuclear magnetic resonance), and as indicated by structural modelling. However, in contrast with NUPR1, there was evidence of local helical- or turn-like structures; these structures were not rigid, as judged by the lack of sigmoidal behaviour in the chemical and thermal denaturation curves obtained by CD and fluorescence. Interestingly enough, NUPR1L interacted with prothymosin α and C-RING1B, and with a similar affinity to that of NUPR1 (in the low micromolar range). Moreover, NUPR1L hetero-associated with NUPR1 with an affinity of 0.4 µM and interacted with the ‘hot-spot’ region of NUPR1. Thus, we suggest that the regulation of NUPR1 gene by NUPR1L does not only happen at the DNA level, but it could also involve direct interactions with NUPR1 natural partners.
NUPR1 is an 82-residue-long (8 kDa), highly basic, monomeric IDP that is overexpressed during the acute phase of pancreatitis [1,2]. It does not have a stable secondary and tertiary structure in any region of its sequence, as it occurs in other IDPs [3–5]. NUPR1 is involved in chromatin remodelling, in several protein–protein interactions (PPIs) with different partners during transcription, and it is an essential element in the stress-cell response and cell-cycle regulation, but its exact functions are yet unknown [6,7]. It plays key roles in pancreatic tumorigenesis, acting downstream of the KrasG12D oncogene mutation . Furthermore, NUPR1 is involved in apoptosis through the formation of a complex with another IDP, prothymosin α , and is implicated in DNA binding  and its repair  through interaction with the male-specific lethal protein (MSL1), and in the association with proteins of the Polycomb group . In these interactions, NUPR1 uses two hydrophobic regions, the so-called ‘hot-spot’, centred around residues Ala33 and Thr68 [9,12]. The expression of the NUPR1 gene is down-regulated by NUPR1L, a 97-residue-long NUPR1-like isoform; in turn, the expression of NUPR1L is p53-regulated . NUPR1L translocates to the cell nucleus and also binds to DNA ; furthermore, threading and homology-based modelling studies suggest that NUPR1L has properties similar to those of members of the HMG (high mobility group)-like family of chromatin regulators, and those studies indicate the presence of a helix-loop-helix fold upon DNA binding .
In this work, we characterized the conformational properties of isolated NUPR1L in solution, and we compared them with those of NUPR1. Our results showed that NUPR1L was an oligomeric IDP in a wide pH range; however, its circular dichroism (CD) spectrum showed evidence of turn- or helix-like structures, and the sole Trp residue of the protein is sequestered from solvent, as shown by fluorescence. Furthermore, the 1D 1H-NMR (nuclear magnetic resonance) spectrum shows that all the observable methyl protons of the protein were clustered ∼0.9 ppm, as it could be expected from a polypeptide chain mostly devoid of tertiary and secondary structures. Those local structures observed by CD and fluorescence were not stable, as judged by chemical and thermal denaturations followed by CD and fluorescence. In contrast, the far-UV (ultraviolet) CD spectrum of isolated NUPR1 does not show evidence of local structure . NUPRL1 also interacted with prothymosin α and C-RING1B, with affinities in the low micromolar range. Furthermore, NUPR1L hetero-oligomerized with NUPR1, and the latter was engaged in the interaction by using the same ‘hot-spot’ region involved in the association with both C-RING1B and prothymosin α [9,12]. These results suggest that the modulation of NUPR1 by NUPR1L did not only occur at the transcriptional level, but it also could take place at the protein level.
Deuterium oxide was obtained from Apollo Scientific (U.K.). Sodium trimethylsilyl [2,2,3,3-2H4] propionate (TSP), imidazole, DNase, Trizma base, ANS (8-anilino-1-naphthalene sulfonic acid), and His-Select HF nickel affinity gel were from Sigma–Aldrich (Spain). The β-mercaptoethanol (β-ME) was from Bio-Rad. Triton X-100 was from VWR (Barcelona, Spain). IPTG was from Apollo Scientific (Stockport, U.K.). Dialysis tubing, with a molecular mass cut-off of 3500 Da, was from Spectrapor (Spectrum Laboratories, Japan). Amicon centrifugal devices with a cut-off molecular mass of 3000 Da were from Millipore. Standard suppliers were used for all other chemicals. Water was deionized and purified on a Millipore system.
Expression and purification of proteins
NUPR1 was produced and purified in Luria-Bertani (LB) media as described [10–12] in C41 cells . For 15N-labelled wild-type NUPR1, the same protocol as in LB was used, with 15NH4Cl (1 g/l of media) as the sole source of nitrogen, supplemented with vitamins and oligoelements. The NUPR1 concentration was determined from the absorbance of the two Tyr residues in the amino acid sequence .
Expression of prothymosin α was carried out as described previously  in M15 strain. As the protein has not any aromatic residue, its concentration was determined by measuring the absorbance at 205 nm .
Expression and purification of C-RINGB in LB was carried out as described in C41 cells . The expression of the protein in minimal media was identical with that described in NUPR1.
Expression of NUPR1L was attained in a Genscript (New Jersey, U.S.A.) Escherichia coli expression vector, codon-optimized for its use in E. coli strains, and subcloned in a pET-30 vector with kanamycin resistance. The sequence of expressed NUPR1L in this vector was MHHHHHHMEA PAERALPRLQ ALARPPPPIS YEEELYDCLD YYYLRDFPAC GAGRSKGRTR REQALRTNWP APGGHERKVA QKLLNGQRKR RQRQLHPKMR TRLT. Expression was carried out in BL21(DE3) strains with a final kanamycin concentration of 100 mg/ml at 37°C. The cells were cultured in the 1 l flasks. They were induced with a final concentration of 0.8 mM IPTG when the absorbance at 600 nm was 0.4–0.9; after induction, the temperature was decreased to 15°C and the cells were grown for 15–16 h. Cells were harvested at 8000 rpm in a JA-10 rotor (Beckman Coulteck) for 15 min. The pellet from 5 l of culture was resuspended in 50 ml of buffer A [500 mM NaCl, 5 mM imidazole, 20 mM Tris (pH 8), 0.1% Triton X-100, and 1 mM β-ME], supplemented with a tablet of Sigma Protease Cocktail EDTA-free and 2 mg of DNase (per 5 l of culture). After being incubated with gentle agitation at 4°C for 10 min, cells were disrupted by sonication (Branson, 750 W), with 10 cycles of 45 s at 55% of maximal power output and an interval of 15 s between the cycles. All the sonication steps and the interval waits were carried out with the samples in ice. The lysate was clarified by centrifugation at 18 000 rpm for 40 min at 4°C in a Beckman JSI30 centrifuge with a JA-20 rotor.
The clarified lysate from such first centrifugation did not contain NUPR1L, which was present in the precipitate. Thus, the precipitate was treated with buffer A supplemented with 8 M urea and a tablet of Sigma Protease Cocktail EDTA-free. The resuspended sample was treated with another 10 cycles of sonication in ice, and the sample was clarified by centrifugation at 20 000 rpm for 30 min at 4°C. NUPR1L was in the supernatant and was purified by immobilized affinity chromatography (IMAC). The supernatant was added to 5 ml of Ni-resin previously equilibrated in buffer A supplemented with 8 M urea. The mixture was incubated for 20 min at 4°C, and afterwards, the lysate was separated from the resin by gravity. On-column refolding was carried out during the washing step with 20 ml of buffer B [20 mM Tris (pH 8.0), 500 mM NaCl, 1 mM β-ME, and 20 mM imidazole]; the protein was, then, eluted by gravity from the column with buffer C [20 mM Tris (pH 8.0), 500 mM NaCl, 1 mM β-ME, and 500 mM imidazole]. The eluted NUPR1L was extensively dialyzed against buffer D [100 mM sodium phosphate (pH 7.0) with 500 mM NaCl and 1 mM β-ME]. The final yield of protein was 1.0–1.5 mg/l of culture, and the protein was 85–90% pure as judged by SDS gels. This low yield precluded the expression of NUPR1L in the culture minimal medium to label the protein for NMR studies.
We attempted to re-purify the protein recovered from IMAC by using gel filtration chromatography in a Superdex 16/600, 75 pg column performed in an AKTA Basic system (GE Healthcare) by monitoring the absorbance at 280 nm; nevertheless, the protein was bound to the column and did not come out within the bed volume of the column. Further attempts with a Heparin column (5 ml HiTrap Heparin column, GE Healthcare) were also unsuccessful, leading to protein precipitation during dialysis to exchange the protein into the loading column buffer.
The eluted protein from IMAC showed absorbance at 260 nm, suggesting that NUPR1L was probably contaminated with oligonucleotides, the product of the DNAse cleavage reaction. We also tried to eliminate the DNA traces by using different concentrations of polyethylenimine [ranging from 0.2 to 1% (v/v)] , but all NUPR1L co-precipitated with DNA. We believe that this contamination was due to electrostatic interactions between the highly charged NUPR1L (theoretical pI is 10.50) and the nucleic acid. The total protein concentration, Pc (in mg/ml), was determined by using the expression : Pc = 1.55 A280 − 0.75 A260, where A280 and A260 are the absorbance of the dialyzed protein solution at 280 and 260 nm, respectively. However, it is important to note that the presence of di-, tri-, or tetra- deoxyoligonucleotides (after treatment with the DNase) did not affect the spectroscopic signals of both fluorescence and far-UV CD, because DNA is spectroscopically silent in fluorescence, and in the far-UV region of CD spectra, between 210 and 240 nm, short deoxyoligonucleotides do not absorb [20,21]. It is also interesting to note that deoxyoligonucleotides in aqueous solution (or intact DNA, in the unlikely case that DNase had been ineffective) show a small or slightly positive ellipticity ∼222 nm [20–24], where we have carried out the study of the CD biophysical properties of NUPR1L (see the Results section).
The sequence of the designed protein did not have any cleavage site to remove its His-tag, and we did not attempt to eliminate it due to the difficulties experienced with other tags in the preliminary studies of the protein.
Fluorescence spectra were collected on a Cary Varian spectrofluorimeter (Agilent, U.S.A.), interfaced with a Peltier, at 25°C. Experiments were carried out in buffer D. The samples were prepared the day before and left overnight at 5°C. Before experiments, the samples were incubated for 1 h at 25°C. NUPR1L concentration in the pH- or urea-denaturation experiments was 10 µM (in protomer units). For pH-denaturation experiments with NUPR1, we used a final concentration of 10 µM. For experiments in the presence of ANS with either protein, a final concentration of 100 µM of the probe was added with the same final protein concentration (10 µM, in protomer units for NUPR1L). A 1 cm pathlength quartz cell (Hellma) was used.
In the pH-induced unfolding curves, the pH was measured after completion of the experiments with an ultra-thin Aldrich electrode in a Radiometer pH meter (Copenhagen). The acids and salts used were: pH 2.0–3.0, phosphoric acid; pH 3.0–4.0, formic acid; pH 4.0–5.5, acetic acid; pH 6.0–7.0, NaH2PO4; pH 7.5–9.0, Tris acid; pH 9.5–11.0, Na2CO3; pH 11.5–13.0, Na3PO4. Chemical and pH denaturations were repeated three times with new samples at any of the concentrations assayed.
For urea-denaturation experiments, the same experimental set-up described above was used for both proteins. Samples were prepared the day before from an 8 M concentrated stock and left overnight to equilibrate. Before experiments, the samples were kept for 1 h at 25°C. The stock urea concentration was determined as described previously . The denaturation experiments were carried out in the phosphate buffer (pH 7.0, 50 mM) in 125 mM NaCl and 0.125 mM β-ME.
The emission intensity weighted average of the inverse wavelengths (also called the spectrum mass centre or the spectral average energy of emission), 〈1/λ〉, was calculated as described previously . Briefly, we define 〈1/λ〉 as , where Ii is the intensity at wavelength λi. From this definition, 〈1/λ〉 is an integral of the fluorescence spectrum, and thus, it allows us to obtain information over all the intensities acquired (instead of using a single wavelength). We shall report 〈1/λ〉 in units of µm−1. This parameter was used during the denaturation experiments and the titrations with other proteins (see below).
To follow the intrinsic fluorescence of Tyr and/or Trp, samples were excited at both 280 and 295 nm. The slit widths were 5 nm for both the excitation and emission lights. The spectra were recorded between 300 and 400 nm. The signal was acquired for 1 s and the increment of wavelength was set to 1 nm. Blank corrections were made in all spectra.
For experiments to elucidate whether there was binding between C-RING1B and NUPR1L, 5 µM of each protein (in protomer units) was used, with the experimental set-up described above.
The same experimental set-up was used for experiments in the presence of ANS, but the excitation wavelength was 370 nm and emission light was collected from 400 to 600 nm. Only experiments with the wild-type NUPR1 were carried out with this probe.
Thermal denaturations of isolated NUPR1L were carried out by following the fluorescence of NUPR1L at several wavelengths, after excitation at 280 or 295 nm. The scan rate was 60°C/h, data were collected every 0.2°C, and the average time was 1 s.
Thermal denaturations were also carried out in an equimolar mixture of both protein isoforms [10 µM (in protomer units) of NUPR1L and 10 µM of NUPR1] to find out whether any of them acquired a stable structure upon complex formation. Furthermore, thermal denaturations were also carried out in an equimolar mixture of NUPR1L (10 µM, in protomer units) and prothymosin α (10 µM). In the thermal denaturations with both complexes, the experimental set-up was as described above. Finally, the same experimental set-up was used for experiments on the complex formation between NUPR1L and C-RING1B (with a final concentration of 5 µM concentration for each protein, in protomer units).
Quenching by iodide and acrylamide was examined at different solution conditions at 10 µM (in protomer units) for NUPR1L and 10 µM for NUPR1: pH 3.5 (formic buffer, 50 mM), pH 6.5 (phosphate buffer, 50 mM), and pH 8.5 (Tris buffer, 50 mM); experiments were also carried out in the presence of 6 M urea at pH 6.0 (phosphate buffer). Excitation was carried out at both 280 and 295 nm (for NUPR1L) or 280 nm (for NUPR1); emission was measured from 300 to 400 nm. The ionic strength was kept constant by the addition of KCl; also, Na2S2O3 was added to obtain a final concentration of 0.1 M to avoid the formation of . The presence of KCl did not modify the structure of NUPR1L, as concluded from the absence of changes in the shape and ellipticity of the CD spectra in the presence of KCl (ranging from 0 to 0.5 M). The data were fitted to [27,28]:
where Ksv is the Stern–Volmer constant for collisional quenching; F0 is the fluorescence intensity in the absence of KI; and F is that at any KI concentration. The range of KI concentrations explored was 0–0.7 M. We also carried out experiments at different concentrations of NUPR1L, namely 4, 15, and 30 µM (in protomer units) in the phosphate buffer (pH 8.0), 500 mM NaCl, and 1 mM β-ME, to elucidate whether the Trp and/or Tyr residues were involved in the oligomeric interface.
where ν is the dynamic quenching constant. Fitting was carried out by using KaleidaGraph (Synergy software) working on a PC.
For the titration between NUPR1L and NUPR1, increasing amounts of NUPR1, in the range 0–10 µM, were added to a solution with a fixed concentration of NUPR1L (4 µM, in protomer units). Experiments were carried out in 50 mM buffer phosphate (pH 7.0), 150 mM NaCl ,and 0.2 mM β-ME. Other details of the experimental set-up have been described above, except that the excitation wavelength was 295 nm, as NUPR1 has no Trp residue. The samples were prepared the day before and left overnight at 5°C; before measurements, the samples were incubated 1 h at 25°C. The dissociation constant of the complex, Kd, was calculated by fitting the plot of the observed fluorescence change of NUPR1L versus added NUPR1 to [29,30]:
where F is the measured fluorescence at any particular concentration of NUPR1L after subtraction of the blank; ΔFmax is the maximal change in the fluorescence of NUPR1L when all of NUPR1 was forming the complex compared with the fluorescence of isolated NUPR1L; F0 is the fluorescence intensity when no NUPR1 was added; [NUPR1L]T is the constant, total concentration of NUPR1L (4 µM, in protomer units); and [NUPR1]T is that of NUPR1, which is varied during the titration. Kd was determined by following the changes in the fluorescence at selected wavelengths of the sole Trp62 of NUPR1L and fitting experimental data to eqn (3). The titration was repeated three times with new prepared samples. At all used concentrations, the absorbance of NUPR1 species was kept lower than 0.2 units of absorbance (at 280 nm) to avoid inner-filter effects, during fluorescence excitation . Fitting to eqn (3) was carried out by using KaleidaGraph (Synergy Software) working on a PC.
For the titration between NUPR1L and NUPR1 with prothymosin α, the experimental set-up was the same as described above, except that excitation wavelengths were 280 and 295 nm for NUPR1L and only 280 nm for NUPR1. The concentration of prothymosin α was varied from 0 to 5 µM, and the concentrations of either NUPR1L (2 µM, in protomer concentration) or NUPR1 (2 µM) were kept constant during the titration. Kd was determined by following the changes in the value of 〈1/λ〉 or fluorescence intensity of each isoform, and fitting experimental data to an equation similar to eqn (3) (with [NUPR1]T swapped by [prothymosin α]T), using 〈1/λ〉 instead of fluorescence intensities at selected wavelengths.
Lifetimes of Trp62 of NUPR1L were measured on an EasyLife V™ lifetime fluorometer (U.S.A.) with the stroboscopic technique, by using as excitation source a pulsed light of a diode LED of 295 nm and an emission filter of 325 nm to avoid light scattering. By exciting at 295 nm, we are only monitoring the fluorescence lifetimes of Trp62 of NUPR1L. The number of channels used for every scan was 500, and the time of integration (i.e. the time over which the signal was averaged for every point of each scan) was 1 s. Three scans were averaged in each experiment, and they were repeated two times at 25°C in the 50 mM phosphate buffer (pH 7.0), 500 mM NaCl, and 1 mM β-ME.
The experimental fluorescence decays (D(t)) were fitted to a sum of exponential functions:
where τi is the corresponding lifetime of the species present in solution, and ai the pre-exponential factor for that species. The pre-exponential factors can be interpreted not only in terms of the populations of the corresponding species, but also in terms of the radiative probability constants and the molar extinction coefficients of the Trp. We also determined the mean lifetime, 〈τ〉, as
where the factors, fi, are defined as
The fitting procedure of the experimental fluorescence lifetime curves uses an iterative method based on the Levenberg–Marquardt algorithm . The temporal width of the excitation pulse, which distorted the observed decay, was taken into account through the instrument response function (IRF); the IRF was determined experimentally by using a scattered solution of ludox. To test the goodness of the fittings, we used the reduced χ2 statistic parameter that was calculated by measuring the spectral noise at time t, and then we applied a method developed for the stroboscopic optical boxcar to determine the measurement errors .
The far-UV CD spectra were collected on a Jasco J815 spectropolarimeter (Jasco, Japan) fitted with a thermostated cell holder and interfaced with a Peltier unit. The instrument was periodically calibrated with (+)10-camphorsulphonic acid. The concentration of NUPR1L was 17 µM (in protomer units) and that of NUPR1 was 15 µM. Experiments were acquired at 25°C unless stated otherwise. Molar ellipticity was calculated as described elsewhere [10,11,17].
Experiments were acquired at the corresponding pH with a scan speed of 50 nm/min, a response time of 2 s, a band width of 1 nm, and averaged over six scans. All spectra were corrected by subtracting the corresponding baseline. Samples were prepared the day before and left overnight at 5°C; before the experiments, the samples were incubated for 1 h at 25°C.
For the experiments, testing complex formation 20 µM of NUPR1L and 30 µM of C-RING1B (in protomer units) were mixed in the phosphate buffer, 50 mM (pH 7.0), in 0.125 mM β-ME and 125 mM NaCl. The experimental set-up was the same as that described above.
Thermal denaturations were carried out with a band width of 1 nm, a response of 8 s, and a scan rate of 60°C/h. Data were collected every 0.2°C, following the raw ellipticity at 222 nm. The experimental set-up for the complex between C-RING1B and NUPR1L was the same.
Analysis of pH-denaturation curves
The pH-denaturation experiments of both proteins were analyzed, assuming that both species, protonated and deprotonated, contribute to the fluorescence spectrum:
where X is the physical property being observed [either ellipticity or fluorescence (intensity at any particular wavelength or 〈1/λ〉)], Xa is the physical property being observed at low pH values (that is, the fluorescence or ellipticity of the acidic form), Xb is the physical property observed at high pH values, and pKa is the apparent midpoint of the titrating group. The apparent pKa reported was obtained from three different measurements, each prepared with a new sample of each protein.
Dynamic light scattering
The oligomerization state in the solution of NUPR1L was assessed by DLS measurements performed at 25°C in 100 mM phosphate (pH 8.0) and 500 mM sodium chloride, both in the presence and absence of 1 mM β-ME. Protein concentrations explored were 1.3 mg/ml (113 µM) and 2.0 mg/ml (170 µM). Experiments were performed at fixed angle (Θ = 173°) in a Zetasizer nano instrument (Malvern Instrument Ltd, U.K.) equipped with a 10 mW helium–neon laser (λ = 632.8 nm) and a thermoelectric temperature controller. Experiments were analyzed with Zetasizer software (Malvern Instrument Ltd, U.K.). Briefly, the Z-average size was obtained by fitting the autocorrelation function with the cumulants method. The hydrodynamic radius, Rh, of the solution is calculated by applying the Stokes–Einstein equation: D = kT/6πηRh, where k is the Boltzmann constant, T is the temperature, and η is the solution viscosity. Before the measurements, all the samples were centrifuged for 30 min at 14 000×g and filtered in a 0.2 µm cut-off Millex filter to remove large aggregates and dust. Samples were sonicated 1 min to remove bubbles once they had been introduced in the cuvette. After filtration and centrifugation, the samples still contained some big aggregates that did not allow the measurements to be conducted in an automatic set-up. The attenuator was set up manually and each sample was measured 10 times with 10 runs of 150 s each.
The NMR experiments were acquired on a Bruker Avance DRX-500 spectrometer equipped with a triple resonance probe and z-pulse field gradients. All spectra were calibrated with external TSP for 1H, and for the indirect dimensions as described previously . Experiments were acquired at 25°C, and the probe temperature was calibrated with a methanol NMR standard .
1D 1H NMR
The spectrum was acquired with a sample concentration of 9 µM of NUPR1L (in protomer units) in the phosphate buffer (pH 8.0), 500 mM NaCl, and 1 mM β-ME, in the presence of 10% D2O. The spectral width was 12 ppm, acquired with 16 K data points; the relaxation delay was set to 1.5 s and the number of scans was 36 K. The carrier frequency was set to the water signal in the centre of the spectrum, and the water signal was suppressed with the WATERGATE sequence . Spectrum was zero-filled to 64 K data points, and an exponential apodization function was applied, prior to Fourier transformation. The spectrum was processed, baseline-corrected, and analyzed using Topsin 1.3 (Bruker).
Experiments were acquired at pH 7.0 (phosphate buffer) in the presence of 0.125 mM β-ME and 150 mM NaCl. The peaks in the 2D 1H,15N-HSQC NMR spectra  of NUPR1 were identified using previously determined assignments at pH 4.5 (BMRB number 19 364) . Spectra were acquired in the TPPI, with the carrier frequency at the water resonance, which was removed by using the WATERGATE sequence . For experiments using 15N-labelled NUPR1, a [NUPR1] of 120 µM was used, and increasing concentrations of unlabelled NUPR1L (14.7, 36.8, 58.8, 80.9, 110.4, and 139.8 µM, in protomer units) were added. At each step, the final volume of the resulting sample was reduced to 500 µl by using Amicon centrifugal devices. For experiments with 15N-labelled C-RING1B, a [C-RING1B] of 125 µM (in protomer units) was used in all experiments, and increasing concentrations of unlabelled NUPR1L (14.7, 29.4, 36.8, 55.2, 73.6, 95.6, and 125.12 µM) were added. At each step, the final volume of the resulting sample was reduced to 500 µl as described above.
The spectra were typically acquired with 2 K complex points in the 1H dimension, 60 complex points in the 15N dimension, with 32 or 64 scans. Typical spectral widths were 6000 (1H) and 1500 (15N) Hz. The resulting matrix of each experiment was zero-filled to double the number of original points in all dimensions, and shifted squared sine-bell apodization functions were applied, prior to Fourier transformation. NMR data were processed and analyzed using Topsin 1.3 (Bruker). Signal intensities in the NMR titration experiments were measured by using the same programme, and in each spectrum, the intensities were corrected by the corresponding value of the receiver gain at each particular concentration of the titrating protein.
Predictions of the disorder propensity along the primary structure of NUPR1L were carried out through the use of scores determined through several independent algorithms [37–43] and obtained by submitting the protein sequence to the respective web-based servers with default computational parameters. A 3D model of NUPR1L was also obtained in a molecular dynamics simulation performed with the GROMACS  package, by using the AMBER ff99SB-ILDN force field  for the protein and TIP3P model  for water, and following a protocol previously adopted for other IDPs [47,48].
NUPR1L was disordered in a wide pH range
NUPR1 and NUPR1L are highly basic proteins with pI values of 10.5 and 9.98, respectively. Thus, to uncover and probe the possible presence of a fraction of biased folded populations, we analyzed the structure of both proteins at varying pH. To this end, we used several biophysical techniques, namely intrinsic and ANS fluorescence, CD, and DLS. In particular, we used intrinsic fluorescence to monitor changes in protein tertiary structure and specifically around its fluorescent residues. We also measured the fluorescence lifetimes of the sole Trp62 of NUPR1L to elucidate its mobility, and whether this was related to the overall tumbling of the molecule. Moreover, ANS fluorescence was used to monitor the burial of solvent-exposed hydrophobic patches. We acquired far-UV CD spectra to monitor the changes in the secondary structure. A 1D 1H NMR spectrum was acquired at pH 8.0 at low NUPR1L concentration to monitor the presence of the secondary and tertiary structures. And finally, DLS was used to monitor the size of NUPR1L at physiological pH, whereas we have previously tested that NUPR1 is a monomer at any pH within a large concentration range (from 20 µM to 2 mM) [10–12].
Steady-state fluorescence and thermal denaturations
NUPR1L has a single Trp and five Tyr residues. The fluorescence spectrum of NUPR1L at physiological pH showed a maximum at 330 nm, suggesting that (i) the sole Trp62 dominated the emission spectrum and (ii) it was mostly buried within the structure (Supplementary Figure S1A). The pH-dependence of 〈1/λ〉 [after excitation either at 280 (Supplementary Figure S1A, inset) or 295 nm] and fluorescence intensity at 350 nm, after excitation at 295 nm, showed two transitions (Figure 1A, right axis, blue squares). The first transition finished at pH 6.0, and the apparent pKa of the transition was 5.0 ± 0.3 (eqn 1). This value was due to the titration of some of the seven Glu or three Asp residues present in NUPR1L, with a high pKa due to a highly hydrophobic environment [49,50] (see Discussion). However, we cannot rule out, at this stage, the titration of any of the two free Cys residues (experiments were carried out in 1 mM β-ME) or even some of the two naturally occurring His residues in the NUPR1L sequence, with an extremely low pKa. The second transition, observed after excitation at both wavelengths, occurred at basic pH, starting at pH >10, but we could not determine the titration midpoint due to the absence of a baseline at the highest pH values. This transition was due to the titration of the Tyr in the sequence. On the other hand, the fluorescence intensity at 350 nm, after excitation at 280 nm, only showed a single transition at basic pH (Figure 1A, left axis, black circles). Conversely, the pH titrations of NUPR1 only showed a basic titration at pH >10.0 (either followed by fluorescence intensity at 308 nm or by 〈1/λ〉), due to the sole two Tyr residues in the sequence (Supplementary Figure S2A).
Spectroscopic features of NUPR1L upon pH changes.
Thermal denaturations of NUPR1L at several pH values (3.0, 5.5, 7.0, 8.4, and 13.0) were carried out by following the changes in the intrinsic fluorescence. At all pH values explored, we did not observe any sigmoidal transition (Figure 1A, inset), and protein precipitation was observed at acidic pH. Thermal denaturations of isolated NUPR1 do not show any sigmoidal transition .
We also tried to follow the denaturation of NUPR1L in the presence of urea by excitation at either 280 or 295 nm. However, as it happened with the thermal denaturations, no sigmoidal curves were observed at any of the two excitation wavelengths (Supplementary Figure S1B). A similar behaviour has been reported in the chemical denaturations of isolated NUPR1 .
At low pH, the ANS fluorescence intensity at 480 nm was relatively large and decreased as the pH was raised (Figure 1B, right axis, blue squares), suggesting that NUPR1L had solvent-exposed hydrophobic regions under acidic conditions. We could not determine the pKa of this titration due to the absence of an acidic baseline (a similar behaviour was observed for the 〈1/λ〉 of ANS as the pH was varied, data not shown). The burial of solvent-exposed hydrophobic residues was complete at pH 6.0, as the transition observed by following the intrinsic fluorescence showed (Figure 1A).
On the other hand, the changes in the fluorescence intensity at 480 nm and 〈1/λ〉 for NUPR1 were small (Supplementary Figure S2B) at acidic pH values; this finding suggests the lack of a large amount of hydrophobic surface becoming solvent-exposed at low pH. In contrast, we observed a slight increase in intensity at 480 nm and 〈1/λ〉 at basic pH values, which occurred concomitantly to the changes in the intrinsic fluorescence (Supplementary Figure S2A); they were also due to the solvent-exposure of hydrophobic regions when the two Tyr residues are titrating. Interestingly enough, this region is highly hydrophobic  and it has been involved in NUPR1 binding to other molecules [9,11,12].
Then, our results indicate that, at low pH, NUPR1L exposed to the solvent a large amount of hydrophobic regions, whereas there was no evidence in NUPR1 of any hydrophobic patches exposed to the solvent at low pH.
Solvent-exposure of NUPR1L fluorescent residues monitored by iodide and acrylamide quenching
To further examine the tertiary structure around fluorescent residues and their solvent-exposure for both NUPR1L and NUPR1, we carried out iodide and acrylamide quenching studies (Table 1). The Ksv values obtained for NUPR1L in the absence of any denaturant were similar to those observed for other proteins (see ref.  and references therein). In general, under non-denaturing conditions, the Ksv values in proteins are larger at 280 than at 295 nm, because at 280 nm both Trp and Tyr residues are being excited. As a general trend, the Ksv values in NUPR1L in the presence of either acrylamide or KI were smaller at acidic pH than at physiological values; therefore, at low pH, the protein acquired conformations that hampered quencher accessibility. These results indicate that the structure of NUPR1L underwent some modifications at acidic pH (in agreement with results from intrinsic and ANS fluorescence, Figure 1). In the presence of urea, the Ksv values at 280 and 295 nm were larger than those in the absence of denaturant (Table 1), indicating that the fluorescent residues were more solvent-exposed.
|280 nm||295 nm||280 nm||295 nm||280 nm||280 nm|
|Conditions||Ksv (M−1) (ν)||Ksv (M−1)2||Ksv (M−1)||Ksv (M−1)||Ksv (M−1) (ν)||Ksv (M−1)3|
|pH 3.6||56 ± 4 (2.4 ± 0.2)||11.5 ± 0.5||1.5 ± 0.2||0.54 ± 0.05||9.1 ± 0.6 (1.2 ± 0.1)||1.1 ± 0.1|
|pH 6.5||61 ± 9 (1.2 ± 0.2)||14.2 ± 0.7||2.9 ± 0.2||2.0 ± 0.1||6.5 ± 0.9 (1.5 ± 0.2)||1.5 ± 0.1|
|pH 8.5||85 ± 10 (0.5 ± 0.2)||16.4 ± 0.5||2.9 ± 0.2||2.1 ± 0.1||4||4|
|6 M Urea3||120 ± 20 (1.4 ± 0.4)||17.7 ± 0.6||3.8 ± 0.1||3.6 ± 0.3||10 ± 1 (0.6 ± 0.1)||1.7 ± 0.1|
|280 nm||295 nm||280 nm||295 nm||280 nm||280 nm|
|Conditions||Ksv (M−1) (ν)||Ksv (M−1)2||Ksv (M−1)||Ksv (M−1)||Ksv (M−1) (ν)||Ksv (M−1)3|
|pH 3.6||56 ± 4 (2.4 ± 0.2)||11.5 ± 0.5||1.5 ± 0.2||0.54 ± 0.05||9.1 ± 0.6 (1.2 ± 0.1)||1.1 ± 0.1|
|pH 6.5||61 ± 9 (1.2 ± 0.2)||14.2 ± 0.7||2.9 ± 0.2||2.0 ± 0.1||6.5 ± 0.9 (1.5 ± 0.2)||1.5 ± 0.1|
|pH 8.5||85 ± 10 (0.5 ± 0.2)||16.4 ± 0.5||2.9 ± 0.2||2.1 ± 0.1||4||4|
|6 M Urea3||120 ± 20 (1.4 ± 0.4)||17.7 ± 0.6||3.8 ± 0.1||3.6 ± 0.3||10 ± 1 (0.6 ± 0.1)||1.7 ± 0.1|
Errors are from fitting to eqns (1) and (2). The Ksvs values were obtained by fitting the fluorescence intensity at 330 nm versus [KI] or [Acrylamide]. Values within parentheses in the acrylamide section correspond to ν, the dynamic quenching constant. In NUPR1, excitation was carried out at 280 nm, since the sequence only contains two Tyr residues.
The ν was 0.0 ± 0.1 under the four conditions.
The quenching experiments in the presence of urea were carried out at pH 6.5, 100 mM phosphate buffer for both proteins.
In NUPR1, we only explored the quenching at acidic and physiological pH values. In contrast with NUPR1L, there were little differences between the Ksv values at acidic and physiological pH. Furthermore, the differences with the value at 6 M urea were not very large either, indicating that the two Tyr residues were more solvent-exposed in NUPR1 than the fluorescent residues in NUPR1L.
To sum up, our quenching results indicate that NUPR1L had its fluorescent residues buried (as further pinpointed by the maximum wavelength at physiological pH, Supplementary Figure S1A) and such burial was more distinct at acidic pH than under physiological conditions.
Solvent-exposure of NUPR1L fluorescent residues at different protein concentrations monitored by iodide quenching
DLS (see below) showed the presence of oligomeric species under our experimental conditions. Then, we decided to monitor the quenching of the fluorescent residues at different NUPR1L concentrations to: (i) provide an alternative probe to monitor protein self-association and (ii) elucidate whether some of the fluorescent residues were directly involved in the binding (i.e. oligomerization interface). We observed that, as the protein concentration was raised, the Ksv parameter decreased: 3.2 ± 0.1 M−1 (4 µM), 2.4 ± 0.2 M−1 (15 µM), and 1.7 ± 0.3 M−1 (30 µM, in protomer units), for an excitation wavelength of 280 nm; and 2.8 ± 0.3 M−1 (4 µM), 1.7 ± 0.3 M−1 (15 µM), and 1.2 ± 0.2 M−1 (30 µM), for an excitation wavelength of 295 nm. These results are in very good agreement with those measured at slightly different conditions for a protein concentration of 10 µM (in protomer units) (Table 1). Therefore, NUPR1L was involved in a protein concentration-dependent equilibrium and Trp62 intervened in such processes.
We also determined the fluorescence lifetimes of Trp62 of NUPR1L to see if there were differences among the values measured for Trp62 and those for other folded proteins in the literature [28,53–55]. The reduced χ2 value, characterizing the goodness of the fitting, was always lower than 1.20, and the weighted residuals were randomly distributed around zero, being less than 3. The lifetime behaviour of the sole Trp62 could be described with two exponentials; attempts to fit the experimental data to more than two exponentials led to increase in the χ2. The presence of several lifetimes can be due to : (i) internal protein motions; (ii) different non-relaxing states of the tryptophan; and (iii) a ground-state heterogeneity due to conformational equilibria. The first and second lifetime values of Trp62 were (with a χ2 = 0.98): τ1 = 3.0 ± 0.1 ns and τ2 = 0.38 ± 0.08 ns (with pre-exponential factors of 0.138 ± 0.009 and 2.98 ± 0.60, respectively). The 〈τ〉 value was 1.11 ns. The individual τi values were similar to those observed for other well-folded, monomeric, single-tryptophan proteins [29,53–55], but the 〈τ〉 was slightly smaller than the lifetime measured for the Trp in such cases (e.g. 2.2 ns for the histidine phosphocarrier protein , 4.9 ns for the G protein , and 1.6 ns for the SAM domain of p73 ). Therefore, our results indicate that the properties of the electronic excited state of Trp62 of NUPR1L are different from those of Trp residues in other monomeric polypeptides.
Taken together, all the fluorescence results indicate that NUPR1L had residual structure around the sole Trp62, which was partially buried from the solvent; this structure was pH-dependent and it was not rigid [as indicated by the absence of sigmoidal transitions in the thermal (Figure 1) and chemical denaturations (Supplementary Figure S1B)]. Conversely, NUPR1 did not show any evidence of pH-dependent conformational changes up to basic pH, when the two Tyr residues were titrated.
The far-UV CD spectrum of NUPR1L at pH 7.5 had minima at 210 nm and a wide shoulder ∼222 nm (Supplementary Figure S1C), which suggest the presence of helix- or turn-like conformations. This spectrum is different from that observed for NUPR1, with a minimum at ∼205 nm , which was a characteristic of random-coil conformations. However, as NUPR1L has one Trp, five Tyr, one Phe, and two His, we cannot rule out the absorbance of aromatic residues, which also occur at these wavelengths [56–58]. Decomposition of the far-UV CD spectrum at the same pH, by using the k2d algorithm available online at the DICHROWEB site [59,60], indicates that the protein had 13% of helical structure.
In NUPR1L, at low pH, the intensity of the far-UV CD at 222 nm decreased, in absolute value, compared with that at physiological pH (Δ[Θ] = 2500 deg cm2 dmol−1), and therefore, the population of helix- or turn-like structures also decreased (Figure 1B, left axis, black circles). This variation occurred concomitantly to the changes observed in ANS intrinsic fluorescence (Figure 1B, right axis, blue squares), suggesting that there were pH-dependent conformational changes associated with solvent-exposure of hydrophobic regions involved in the formation of those helix- or turn-like structures. We could not determine the pKa of this titration due to the absence of an acidic baseline. As it happened with the thermal denaturations followed by fluorescence, the transitions followed by the ellipticity at 222 nm did not show any sigmoidal behaviour, suggesting that possible helical- or turn-like conformations were not rigid (Figure 1B, inset).
On the other hand, the ellipticity at 222 nm in NUPR1 also decreased, in absolute value at low pH (Supplementary Figure S2C), but the change was smaller than for NUPR1L (Δ[Θ] = 700 deg cm2 dmol−1), and it did not occur concomitantly with the ANS fluorescence (Supplementary Figure S2B).
To sum up, for NUPR1L, the changes with the pH in CD mirrored those of both intrinsic and ANS fluorescence, indicating the disruption of hydrophobic local secondary structure around Trp62 at acidic pH. In addition, the far-UV CD spectrum at physiological pH indicates the presence of turn- or α-helix-like conformations, which are not rigid (as indicated by the absence of sigmoidal transitions in the thermal denaturation experiments). The behaviour of NUPR1 was the opposite, with almost no changes in the secondary structure at acidic pH, and the absence of structured conformations observable by far-UV CD .
Nuclear magnetic resonance
We acquired a 1D 1H NMR spectrum at very low NUPR1L concentration (9 µM, in protomer units) at pH 8.0, in the presence of 500 mM NaCl and 1 mM β-ME (Supplementary Figure S3). In the up-field region of the spectrum, all the methyl protons appeared clustered ∼0.9 ppm, as they are expected in a polypeptide chain devoid of the rigid secondary and tertiary structures [34,61]. On the other hand, in the amide region, a very low signal intensity was observed and very broad peaks were detected (Supplementary Figure S3B) [the sharp peaks are due to the aromatic protons of the His-tail (∼8.5 ppm) and the presence of protons of the bases of deoxyoligonucleotides]; furthermore, no signal, ∼10.20 ppm, was observed for the indole moiety of Trp62, where it should be observed for a disordered polypeptide chain . These results can be explained as due to: (i) the presence of slow-to-intermediate conformational exchange in NUPR1L involving such aromatic ring, which broadens the signal; or alternatively, (ii) signal broadening due to presence of large molecular mass species. The latter explanation agrees with the experimental observation that Trp62 was involved in the self-association equilibrium of NUPR1L as shown by the quenching results (see above).
Then, the NMR indicates that NUPR1L was a disordered protein.
Dynamic light scattering
In NUPR1L, a scattering signal accounting for 99.8% of the intensity appeared in a broad peak with a hydrodynamic radius, Rh, of 3.3 ± 1.5 nm (from the Stokes–Einstein equation) (Figure 2), which corresponded to a molecular mass of 54 kDa by using an empirical mass versus size calibration in the software of the instrument. A similar signal was observed for both protein concentrations assayed: 2 and 1.3 mg/ml, indicating that the order of the oligomeric state did not change. Therefore, NUPR1L was mainly a tetramer in solution under the conditions studied; however, it must be pinpointed that the high polydispersity of the peak (46%) indicated the presence of other species in solution, and then, other self-associated species must be present in equilibrium. Then, NUPR1L exists in a monomer/multimer equilibrium. Samples treated with DNAse and either in the presence or in the absence of 1 mM β-ME yielded similar results. Therefore, the DLS measurements indicate that NUPR1L was an oligomeric protein.
DLS measurement of the Rh of NUPR1L.
A basic bioinformatic analysis of NUPR1L sequence was performed to gain further insights into its structural properties. To this aim, many well-assessed and independent algorithms [38–44] for predicting the disorder propensity in protein sequences were used. The findings obtained are qualitatively summarized in Figure 3 (the corresponding quantitative analysis is reported in Supplementary Figure S4). The results indicate that regions prone to be structurally ordered in NUPR1L found in the two segments of the sequence. The first is the one around residues 25–40, which is rich in Leu and Tyr residues and contains the only Phe in NUPR1L, with these three types of amino acids encompassing 50% of the local residues. The other region shows a lower agreement among the predictors and is much less defined, roughly including residues 60–80. It is interesting to note that Trp62 lies at the edge of such region, and this could explain the tendency of this residue to be partially buried from the solvent (as shown by the value of the maximum wavelength in the fluorescence spectra, Supplementary Figure S1A, and the quenching results, Table 1), and probably involved in the local structure. Furthermore, this local structure, which was not rigid (Figure 1), could explain the presence of turn- or helix-like conformations detected by far-UV CD (Supplementary Figure S1C).
Predicted disordered regions in the NUPR1L sequence.
We also tried to build a model for the ensemble of structures of NUPR1L through an unbiased, brute force molecular dynamics simulation carried out with explicit water molecules, starting from the protein in an extended conformation. The relatively large size and flexibility of NUPR1L precluded an adequate sampling  of the statistical ensemble. However, the fraction of residues involved in helical structures (16.0 ± 1.5%) was in agreement with the CD decovolution results (13%, see above), and Trp62 was mostly partially sheltered from the solvent in spite of being in a region almost uncollapsed (Supplementary Figure S5), consistently also with our quenching experimental findings. The presence of such uncollapsed region would explain why the Ksv value in 6 M urea at 295 nm was slightly higher than that in aqueous solution at a similar pH (Table 1).
NUPR1L did hetero-associate with NUPR1
As NUPR1L down-regulates the expression of NUPR1 gene, we wondered whether such modulation would involve binding to this protein. To that end, we carried out fluorescence titrations by keeping fixed the concentration of NUPR1L, while NUPR1 was the titrating molecule. We also tried to carry out isothermal titration calorimetry experiments, but at the required NUPR1L concentrations, the protein always precipitated. Our fluorescence results show that the value of Kd was 0.8 ± 0.3 µM (Figure 4A), as measured by excitation at either 280 or 295 nm. As NUPR1 has no Trp, the fact that the fluorescence corresponding to the excitation of this amino acid was affected indicates that the NUPR1L region involved in the binding must include Trp62.
Interaction of NUPR1L with its isoform NUPR1 and prothymosin α as measured by spectroscopic techniques.
We also tried to monitor the changes in the intensity of 2D 1H, 15N-HSQC in NUPR1, since we have previously observed that the addition of a protein to NUPR1 led to broadening of cross-peaks in such NMR spectra . Although by titrating NUPR1L over NUPR1 we were altering the population of oligomeric NUPR1L species, this experiment was performed to have another independent estimation of the apparent Kd. Under these conditions [pH 7.0 (phosphate buffer) in the presence of 0.125 mM β-ME and 150 mM NaCl], the only residues of NUPR1 whose cross-peaks could be observed were Gly16, Glu18, Asp19, Ser22 (Ser9), Ser23 (Asn53), Ser27, Leu29, Tyr30, Ser31, and Arg82 (by using the assignment of NUPR1 at pH 4.5, BMRB number 19 364 ), with residues within the parentheses indicating cross-peak overlapping (Supplementary Figure S6A). A disordered protein has sharp NMR resonances due to short-effective correlation times because of fast internal motions. Upon the addition of NUPR1L, there were not changes in the chemical shifts of any of the cross-peaks (as also previously observed for C-RING1B ), indicating that NUPR1 remained disordered when in complex with its paralogue . However, the behaviour of the variation of the NMR cross-peak intensities upon NUPR1L addition was not a simple sigmoidal one, but rather the intensity of the cross-peaks increased up to 20 µM (in protomer units) of NUPR1L, and at higher NUPR1L concentrations, the intensity started decreasing in a sigmoidal manner (Figure 4B). As the cross-peak intensities are proportional to the amount of NUPR1 (the observed protein), the behaviour of the curve must be related to multi-state-binding equilibria involving NUPR1 and the oligomeric NUPR1L in solution, which are slow-to-intermediate within the NMR time scale . In fact, similar titrating curves have been observed in macromolecules that partition differently between a solution and a membrane, and where the macromolecule aggregates in the presence of the membrane (see ref.  and references therein).
Although NUPR1L and NUPR1are associated, the thermal denaturations followed by fluorescence of an equimolar mixture (in protomer units) of both proteins did not yield any sigmoidal transition (data not shown), indicating that the complex was still disordered (as further pinpointing by the absence of changes in the chemical shifts in the spectrum of NUPR1).
NUPR1L was bound to prothymosin α with a micromolar affinity
We have previously shown that NUPR1 interacted with prothymosin α through the region close to the 30 s along its sequence, since fluorescence-binding studies show the involvement of the two Tyr residues  in that region. We wondered whether NUPR1L also interacted with prothymosin α. We first tried to carry out isothermal titration calorimetry experiments, but under all the conditions explored, at the concentrations of NUPR1L required for the titration, the protein was precipitated. Then, we carried out fluorescence titrations keeping constant the amount of NUPR1L in the cuvette. These titrations yielded a Kd of 0.6 ± 0.3 µM (Figure 4C) (by excitation at either 280 or 295 nm); control experiments under the same conditions (with λex = 280 nm) led to a Kd of 0.34 ± 0.03 µM for NUPR1 (Figure 4C, inset). Therefore, the affinities of both protein isoforms for prothymosin α were similar within their experimental uncertainties. Thermal denaturations followed by fluorescence of both complexes (in an equimolar mixture, in protomer units for NUPR1L) did not lead to a sigmoidal transition, suggesting that both proteins remained disordered in the hetero-oligomer (data not shown). The fact that for NUPR1L an estimation of Kd was obtained after excitation at 295 nm indicates that the region around Trp62 was also involved in binding to prothymosin α.
NUPR1L was bound to C-RING1B with a low micromolar affinity
After assessing that NUPR1L was bound to prothymosin α, similarly to NUPR1, we wondered whether the same happened for C-RING1B. For this domain, we have measured the Kd of NUPR1 to be ∼12 µM . To test the interaction between NUPR1L and C-RING1B, we used steady-state far-UV CD and fluorescence. Both experiments (Figure 5A and Supplementary Figure S7A) showed that there was binding between the two molecules, since the fluorescence and CD spectra of the equimolar complex were different from those obtained by the addition of the spectra of the two isolated proteins. Therefore, we reasoned that the binding of NUPR1L should increase the thermal stability of C-RING1B if the complex dissociates prior to the rate-limiting step of the thermal denaturation. As expected, the presence of NUPR1L increased the apparent thermal denaturation midpoint as measured by both far-UV CD (Figure 5B) and fluorescence (Supplementary Figure S7B). We did not determine Kd from changes in the denaturation midpoints due to irreversibility (we observed precipitation in all complexes formed with the C-RING1B domain). Furthermore, we did not attempt to measure the affinity constant by fluorescence, since both proteins are oligomers (the self-association constant of C-RING1B is ∼200 µM ). We also tried to determine the association between the two proteins by isothermal titration calorimetry, but in all attempts the NUPR1L sample was precipitated at the required concentrations.
Interaction of NUPR1L with C-RING1B as measured by spectroscopic techniques.
We have previously described that upon the addition of NUPR1 to C-RING1B, the cross-peaks of the 2D 1H,15N-HSQC spectrum of C-RING1B become broadened , due to an intermediate exchange within the NMR time scale. To test whether this behaviour was also observed in NUPR1L, we titrated the protein over 15N-labelled C-RING1B (at a concentration of 125 µM, in protomer units). The cross-peaks of the spectrum of C-RING1B also broadened as [NUPR1L] was increased (Supplementary Figure S6B). However, in the NUPR1L-C-RING1B complex, the signal broadening could be due to: (i) a slow-to-intermediate conformational exchange within the NMR time scale (as it seems to happen in the NUPR1–C-RING1B complex ) or (ii) complex formation between the two molecules, due to the large molecular mass of oligomeric NUPR1L (Figure 2). In contrast with what one might have expected from our previous titrating experiments between NUPR1L and NUPR1 (Figure 4B), the decrease in intensity of the cross-peaks of C-RING1B (Supplementary Figure S6) followed a sigmoidal behaviour, and we were able to fit those changes in intensity to eqn (4) (Figure 5C). All the residues led to a similar value of the apparent affinity constant, Kd, of 57 ± 18 µM. However, it is important to pinpoint that this is an apparent value, as (i) we were affecting the population of oligomeric species of NUPR1L during the titration and (ii) fitting of intensity decreases in NMR cross-peaks from slow-to-intermediate exchange equilibria does not yield reliable dissociation constants . The difference in the behaviour of the NUPR1L–C-RING1B titration, when compared with that of NUPR1 (Figure 4B), must be due to the fact that C-RING1B does not show different multi-state equilibria involving the oligomeric NUPR1L in solution.
NUPR1L was an IDP with local residual structure
Structural features of IDPs are difficult to characterize, due to their inherent flexibility. Interestingly enough, when there are two paralogues, which are both IDPs, studies of their conformational properties under different conditions, by using several biophysical probes, can provide insights into at what level of exquisite detail sequences modulate structure and function in IDPs.
We conclude that NUPR1L was an IDP on the basis of two pieces of evidence. The absence of sigmoidal transitions in fluorescence and CD thermal (Figure 1) and in the fluorescence chemical denaturations (Supplementary Figure S1B) indicates the lack of secondary and tertiary structures. Although such absence is one of the features to assess protein disorder [66–68], it could be argued that non-sigmoidal denaturation curves are also due to non-cooperative transitions. However, it would be highly unlikely that non-cooperative transitions for the same protein were observed in chemical (Supplementary Figure S1) and thermal denaturations (Figure 1). To provide an additional probe to monitor the conformational features of NUPR1L, we carried out the acquisition of an 1D 1H NMR spectrum at very low protein concentration (Supplementary Figure S3) (it was not feasible to isotope-label NUPR1L for NMR purposes due to the low protein yield in rich media, see the Materials and Methods section). The amide signals had very low intensity and they were very broad; however, the up-field shifted region showed that all the observable methyl protons appear at the expected chemical shifts for polypeptide chains devoid of the secondary and tertiary structures: 0.9 ppm .
However, even though NUPR1L was mainly disordered, there was evidence of flickering, residual structure (from both intrinsic and ANS fluorescence and CD). This structure disappeared at low pH (as judged by the decrease of ellipticity at 222 nm, Figure 1B), when hydrophobic patches were also solvent-exposed (as judged by ANS titrations). Therefore, the residual structure in NUPR1L was sustained by the presence of hydrophobic interactions, which create a folding nucleus [69,70]. The fact that the maximum wavelength of the fluorescence spectrum was 330 nm (Supplementary Figure S1A) indicates the burial of the side-chain of Trp62, in agreement with our molecular modelling. Apart from the pH-dependent changes of ANS fluorescence and in the ellipticity at 222 nm (Figure 1B), and the small value of the λmax of the fluorescence spectrum at physiological pH (Supplementary Figure S1A), there is another piece of evidence that indicates the formation of a hydrophobic cluster: the measured pKa value in the conformational changes followed by fluorescence at 295 nm (Figure 1A). This value, which could be attributed to Asp and/or Glu residues, is higher than the values reported for random-coil models (4.0 and 4.3) [49,50]. This difference can only be explained as due to a hydrophobic environment around some of the Asp and/or Glu residues. The local hydrophobic patch detected by the analysis of sequence around residues 25–40 comprises two Glu and two Asp residues, but there are also two Glu close to Trp62 in the NUPR1L sequence: Glu55 and Glu69. Thus, the most probable candidates responsible for those fluorescence pH-titrating changes are the latter glutamic residues.
Changes in the secondary (far-UV CD) and tertiary structure (intrinsic fluorescence), and in the burial of hydrophobic residues (ANS fluorescence) occurred concomitantly at low pH in NUPR1L. Under acidic conditions, species showing smaller ellipticity (and, therefore, secondary structure) were more populated; these NUPR1L species had hindered solvent-accessibility towards I− and acrylamide quenchers of Trp and Tyr residues. The increase of ANS fluorescence (larger solvent-accessible hydrophobic surface area) at low pH may appear difficult to reconcile with the results of I− or acrylamide quenching (smaller solvent accessibility towards Trp/Tyr at low pH, Table 1). However, the larger amount of hydrophobic surface area (monitored by ANS), which became solvent-exposed at acidic pH values, could involve many Leu residues, which are highly abundant in NUPR1L (11 out of 97 amino acids). It should be noted that the only region that theoretical predictions agree to indicate as ordered in NUPR1L includes residues 25–40 (Figure 3) and corresponds to a small and moderately hydrophobic patch (formed by four Tyr, one Phe, and two Leu).
On the other hand, the different biophysical probes suggest that, apart from the changes in ellipticity at 222 nm (secondary structure) (Supplementary Figure S2C), there were no pH-dependent conformational changes in NUPR1. The conformational changes observed by CD must be due to the titration of the Asp and Glu residues of the protein, which appear to be involved in local structure around its Pro residues . Furthermore, the far-UV CD spectra of both proteins (Supplementary Figure S1C and our previous results ) suggest that NUPR1L was more ordered than its isoform, although local structure was not stable, as concluded from the thermal denaturations (Figure 1). Therefore, the two isoforms showed different pH-dependent conformational changes that must be directly related to their amino acid composition.
Considering the 60% of homology of sequence between NUPR1L and NUPR1, we wondered what is the main difference between the two isoforms responsible of their unlike conformational preferences. Comparison of the sequences of the two proteins  indicates that a key difference is the number of Leu (7.3% in NUPR1 versus 11.3% in NUPR1L), Gly (11.0% in NUPR1 versus 6.2% in NUPR1L), and Ser residues (9.8% in NUPR1 versus 2.1% in NUPR1L); thus, NUPR1 has a larger number of disorder-promoting residues than NUPR1L [67,68]. The larger faction of order-promoting residues is enough to promote local residual structure, involving turn- or helix-like conformations (as judged from the shape of the far-UV CD spectrum, Supplementary Figure S1C). Therefore, non-conserved residues of NUPR1L are important for attaining a local structure (i.e. a folding nucleus); this local structure is also reflected in our simulations (Supplementary Figure S5) and is well populated in our simulations of the DNA–NUPR1L complex , where a helix-turn-helix motif was observed. We cannot unambiguously rule out that the presence of: (i) the turn- or helix-like structures, as detected by far-UV CD (Supplementary Figure S1C) or (ii) the hydrophobic clusters (as detected by fluorescence) could be due to the presence of di or tri- deoxyoligonucleotides. However, the use of DNase during purification (see the Materials and Methods section) ensures that the presence of such deoxyoligonucleotides is negligible (as indicated by the NMR spectrum, Supplementary Figure S3). In turn, our results also indicate that IDPs may be as sensitive to mutations as, at least, their globular counterparts. Other isoforms of IDPs, or disordered regions in well-folded proteins, have been recently described to be highly sensitive to specific mutations, in terms of the residual promoted structure [71,72]; even, it has been shown that for some families of IDPs, the structure present in the transition state of the binding reaction to other partners is encoded within the particular IDP sequence .
NUPR1L was an oligomeric IDP
Apart from the presence of residual structure, there is another crucial difference between NUPR1 and NUPR1L: only the latter self-associates, as indicated by: (i) DLS (Figure 2); (ii) the quenching experiments at different NUPR1L concentrations; and probably, (iii) the absence of indole signal in the 1H NMR spectrum (Supplementary Figure S3). This self-association also is likely due to the larger number of Leu residues in NUPR1L, and thus, their order-promoting action would concomitantly trigger the presence of local secondary structure and favour quaternary rearrangements.
At a first stage, the presence of such a quaternary structural organization in an IDP can be explained as due to multiple transient interactions, favoured by those local ordered structures, increasing the disorder in the oligomer [63,74], and yielding a fuzzy self-associated species [75,76]. In such a scenario, it is tempting to suggest that the region mainly responsible for the formation of such quaternary structure in NUPR1L is the local hydrophobic patch detected by the analysis of sequence around residues 25–40 together with the region ∼60–70 (comprising the Trp); therefore, about a quarter of protein residues could be variously implicated in local structures involved in the oligomerization interface. Alternatively, the formation of such quaternary structure could be triggered by long-range interactions among the polypeptide changes, without the need of residual local structure, as it has been shown by theoretical and experimental approaches in several models [77,78]. Formation of these high-order oligomers has been suggested to modulate spatial organization in the cell and signalling events [79,80], favouring functional versatility in the case of NUPR1L.
IDPs are involved in protein–protein contacts [3–5,81], but only a few are reported as oligomers in solution (that is, having a quaternary structure), while keeping their disordered nature in their corresponding monomers. One of the first examples was the ‘fuzzy’ dimer formed by the cytoplasmic domain of the T-cell receptor zeta subunit [82,83], although this putative dimer was later shown to be a monomer under a wide range of conditions, and its oligomerization was attributed to non-ideal protein–column–resin interactions . However, other studies have used a plethora of biophysical (spectroscopic and hydrodynamic) and biochemical techniques to unambiguously show the existence of intact self-associated IDPs in: oligomeric plant proteins ; dimeric proteins in the disordered umuD gene products ; oligomeric mitochondrial proteins ; oligomeric proteins associated with the Polycomb complex ; dimeric inhibitors of ribonucleotide reductase of Saccharomyces cerevisiae ; the hetero-oligomeric complex of tau protein and α-synuclein ; oligomeric acid-rich proteins of rod photoreceptors ; and (homo- or hetero-) oligomeric assemblies detected in liquid–liquid-phase separation [79,80]. There are even studies where specific regions, which are intrinsically disordered in otherwise folded proteins, are shown to be specifically involved in promoting self-association of the whole protein, maintaining their flexibility in the associated state [91,92]. It is important to indicate that in all those proteins, the regions involved in the oligomeric interface are those predicted (or experimentally tested) to have a certain propensity to be folded (as it happens in NUPR1L).
NUPR1L interacted with the same partners as NUPR1 but with an apparent different affinity
We have shown that NUPR1L was capable of interacting with the same partners of NUPR1, namely prothymosin α  and C-RING1B . However, while the affinity for prothymosin α was the same, within the uncertainty, for both protein isoforms, in the case of C-RING1B the apparent affinity of NUPR1L was smaller (i.e. larger Kd) compared with that of NUPR1: 60 versus 10 µM . This difference is due to the self-association of NUPR1L, since the reported constants should include the dissociation of both C-RING1B and NUPR1L, unless the complex involves the formation of a higher-order state between the oligomeric species of both proteins, in which case the real affinity between the two proteins must be lower (see below).
The apparent affinity of either protein isoform for prothymosin α was similar (∼0.5 µM), and thus, assuming that the stoichiometry of binding was 1 : 1, the affinity of NUPR1L for prothymosin α should be lower because the value measured must also take into account the dissociation of NUPR1L. This smaller value of the apparent dissociation constant can be rationalized as follows: assuming from the DLS measurements, the presence of mainly a tetrameric species for NUPR1L (given by the equilibrium constant: Ksd = [NUPR1L]4/[NUPR1L4]). The apparent dissociation constant for prothymosin α (assuming that the monomeric NUPRL1 is the competent-binding species):
From this expression, we can conclude that if the NUPR1L self-association equilibrium is shifted towards the monomer (Ksd very large and 4[NUPR1L4]/[NUPR1L] << 1), then Kdapp is similar to Kd (the observed affinity is more or less equal to the intrinsic affinity for the monomeric NUPR1L). On the other hand, if NUPR1L self-association equilibrium is displaced towards the tetramer (Ksd, very small and 4[NUPR1L4]/[NUPR1L] >> 1), then Kdapp is much larger than Kd (i.e. the observed affinity is much lower than that of the monomeric NUPR1L). It is important, however, to indicate that, as shown by the large polydispersity of the DLS experiments (see above), there must be another self-associated species present in solution, and therefore, the order of the other associated species will affect the expression given above.
Interestingly enough, the region of NUPR1L involved in binding to prothymosin α was that around Trp62 (as shown by the changes in fluorescence spectra after excitation at 295 nm, Figure 4C), which is a non-conserved residue between the two isoforms (it corresponds to an Arg residue in NUPR1) . Since we have shown above that this residue is important in developing a locally folded region, our binding results would indicate that in NUPRL1, folding and function are intertwined.
On the other hand, residues of NUPR1 involved in the binding of prothymosin α were those around Tyr30 and Tyr36 (as we monitored the binding by following the fluorescence of both aromatic residues); Tyr30 is a conserved residue in both isoforms, but residue Tyr36 is a Pro in NUPR1L . We have also observed that NUPR1 residues Ala33 and Thr68 are crucial in the binding to other molecules , but both residues were not conserved in NUPR1L: residue 33 corresponds to an Arg, and residue 68 to a Gln . These findings indicate that non-conserved residues between two isoforms, apart from being important in helping to create and develop residual secondary structure in NUPR1L (see above), are key in binding to other molecules. As all of the protein partners we have identified so far bound to both protein isoforms, the non-conserved residues must confer some specificity to bind other type of partners in both paralogues. Alternatively, the non-conserved residues might act as anchoring points for the protein partners and the conserved residues in each protein modulate the function according to some other, yet unknown, mechanism.
Binding to NUPR1 also involves the surrounding region around the sole Trp62 of NUPR1L (Figure 4A), as concluded from fluorescence results. Although at pH 7.0 we could only monitor the cross-peaks of residues around the 30s region in NUPR1 (see the Results section), their signals became broadened upon the addition of NUPR1L. Thus, the same NUPR1 region involved in binding to its molecular partners was also involved in association with its isoform. However, the apparent dissociation constant (∼0.8 µM) was smaller than that of NUPR1 when bound to C-RING1B (∼12 µM) . These findings indicate that NUPR1L could modulate NUPR1 gene, not only down-stream at DNA level, but also at the protein level by an efficient inhibition of the interactions with its partners, providing an alternative route to regulation of NUPR1 functions.
Our results show that NUPR1L is an IDP as its paralogue, NUPR1. The differences in sequences among isoform IDPs are critical to determine the presence of local residual secondary structure and quaternary rearrangements. Furthermore, we showed that NUPR1L and NUPR1 can interact with the same molecular partners, as well as they can interact directly with each other. Thus, our results indicate that NUPR1L could not only act as a modulator of NUPR1 functions at DNA level, but also interact with NUPR1 partners.
8-anilino-1-naphthalene sulfonic acid
C-terminal region of the Polycomb RING protein 1
dynamic light scattering
intrinsically disordered protein
immobilized affinity chromatography
instrument response function
sodium trimethylsilyl [2,2,3,3-2H4] propionate
J.L.N., J.L.I., A.C.-A., P.S., and B.R. designed the experiments and the research methodology. M.B.L., A.C.-A., B.R., P.S., and J.L.N. carried out the experiments and analyzed data. M.V. provided materials. J.L.N., J.L.I., A.C.A., P.S., and B.R. wrote the paper.
This work was supported by the Spanish Ministry of Economy and Competitiveness [CTQ 2015-64445-R (to J.L.N.), BIO2016-78020-R to A.C.-A., and FIS2014-52212-R to P.S.]; by the French La Ligue Contre le Cancer, INCa, Canceropole PACA, DGOS (labellization SIRIC), and INSERM (to J.L.I.). B.R. acknowledges kind hospitality and use of computational resources in the Magnetic Resonance Center (CERM), Sesto Fiorentino (Florence), Italy.
The Authors declare that there are no competing interests associated with the manuscript.
Present address: Instituto de Nutrición y Tecnología de los Alimentos (INYTA), Centro de Investigaciones Biomédicas (CIBM), Universidad de Granada, Avda. del Conocimiento, s/n, CP 18100 Armilla, Granada, Spain.