The metastasis suppressor KISS1 is an intrinsically disordered protein slightly more extended than a random coil

The metastasis suppressor KISS1 is reported to be involved in the progression of several solid neoplasias, making it a promising molecular target for controlling their metastasis. The KISS1 sequence contains an N-terminal secretion signal and several dibasic sequences that are proposed to be the proteolytic cleavage sites. We present the first structural characterization of KISS1 by circular dichroism, multi-angle light scattering, small angle X-Ray scattering and NMR spectroscopy. An analysis of the KISS1 backbone NMR chemical shifts does not reveal any preferential conformation and deviation from a random coil ensemble. The backbone 15N transverse relaxation times indicate a mildly reduced mobility for two regions that are rich in bulky residues. The small angle X-ray scattering curve of KISS1 is likewise consistent with a predominantly random coil ensemble, although an ensemble optimization analysis indicates some preference for more extended conformations possibly due to positive charge repulsion between the abundant basic residues. Our results support the hypothesis that KISS1 mostly samples a random coil conformational space, which is consistent with its high susceptibility to proteolysis and the generation of Kisspeptin fragments.


Introduction
Metastasis suppressors are able to slow or block metastasis without preventing primary tumor development. The KISS1 gene was originally identified as a potent melanoma metastasis suppressor [1], where expression levels correlated inversely with the metastatic potential in a panel of melanoma cell lines [2]. The same inverse correlation with tumor stage and overall survival rate was also described for bladder cancer, where every tumor that had developed distant metastasis showed complete absence of KISS1 expression [3,4]. Similar observations on tumor progression, metastasis, and survival in different human cancers make KISS1 an interesting target for controlling metastasis in a therapeutic context [5]. Human KISS1 has an N-a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 transformed with the KISS1-pET11d construct. Cells were grown at 37˚C to an OD 600 of 0.6 and protein expression was induced with 1 mM IPTG for 3 h at 37˚C. For NMR studies, the cells from 3 L of LB were resuspended in 1 L of M9 minimal medium with isotopic enrichment (1 g/L 99% 15 NH 4 Cl and 2 g/L 99% 13 C 6 -glucose) and expression induced with 1 mM IPTG for 3 h at 37˚C [27]. Cells were harvested and resuspended in lysis buffer (20 mM Tris, pH 8.0, 1 mM DTT, plus one tablet of Complete protease inhibitor cocktail), and sonicated on ice. The lysate was clarified by ultracentrifugation at 4˚C and the supernatant was loaded on a Hiload 26/10 Q Sepharose column (GE Healthcare) with a column volume (CV) of 53 mL. The column was washed with 20 mM Tris, pH 8.0, 1 mM DTT, 1 mM EDTA (buffer A) and the protein was eluted with a linear gradient of salt up to 2 M NaCl in 3 CV). Fractions containing the protein (as seen in SDS-PAGE) were pooled and prepared for reverse phase chromatography by adding trifluoroacetic acid (TFA) to final concentrations of 0.1%. The solution was clarified by a 0.22 μm filter and loaded on a Phenomenex Jupiter C 18 250x10 mm column with 10 μm particles and 300 Å pore diameter previously equilibrated with 0.1% aqueous TFA. The protein was eluted with a 0-55% gradient of elution buffer (90% aqueous acetonitrile, 0.1% TFA) in 3 CV and freeze-dried. The purity of the sample was evaluated by overloaded SDS-PAGE analysis, and mass spectrometry measurements indicated that the initial methionine was processed by bacterial enzymes. For NMR analysis the lyophilized protein was dissolved in water with 2 mM DTT and the pH adjusted to 5.7. For other biophysical analysis the powder was dissolved in PBS (10 mM phosphate, 140 mM chloride, 153 mM sodium ion, 4.5 mM potassium ion) pH 7.0, 1 mM DTT, chromatographed on a HiLoad Superdex 200 16/60 column (GE Healthcare), and concentrated by ultrafiltration using Amicon devices with 3 kDa cut-off. Protein concentration was determined by UV absorbance at 280 nm using the extinction coefficient calculated from the amino acid composition (6990 M -1 Ácm -1 ).

Size Exclusion Chromatography-Multi Angle Light Scattering (SEC-MALS)
These experiments were performed using a Superdex 200 10/300 GL column (GE HealthCare) connected to a DAWN-HELEOS light scattering detector and an Optilab rEX differential refractive index detector (Wyatt Technology) at 25˚C. The column was equilibrated with running buffer (PBS pH 7.0, 1 mM TCEP, 0.03% NaN 3 ) and the SEC-MALS system was calibrated with a sample of Bovine Serum Albumin (BSA) at 1 g/L in the same buffer. A sample of 100 μL of Kiss1 protein at 0.8 g/L was injected at 0.5 mL/min. Data acquisition and analysis employed ASTRA software (Wyatt). Based on numerous measurements on BSA samples under the same or similar conditions, we estimate that the experimental error in molar mass is around 5%.

CD spectroscopy
The spectrum of a 52 μM sample of Kiss1 in PBS pH 7.0, 0.1 mM DTT was recorded on a Jasco-815 spectropolarimeter using a quartz cuvette (0.1 mm path length). Thermal denaturation was induced by heating (at a rate of 1˚C/min) a sample of 8.0 μM Kiss1 in the same buffer in a stoppered cuvette with 2 mm path length.
for spectral assignment were obtained using a 112 μM [U-13 C, 15 N] KISS1 sample in water with 5% (v/v) 2 H 2 O, 2 mM DTT and pH 5.7 adjusted with concentrated HCl and NaOH. A pH lower than 7.0 was chosen to reduce the solvent exchange of the amide protons that made not visible several of the Kiss1 signals in the 1 H, 15 N-HSQC spectrum at pH 7.0 (data not shown). The value of 5.7 was considered appropriate as only the two histidine residues of KISS1 will be affected (becoming predominantly protonated in the NMR sample). 1 H N , 15 N, 13 C 0 , 13 C α , 13 C β and 1 H α assignments were obtained from the analysis of two dimensional 1 H, 15 N-HSQC, 1 H, 13 C-HSQC and three dimensional HNCO, HN(CA)CO, HNCACB, HN (CO)CACB, HNCA, HN(CO)CA, HN(CA)HA, and HN(COCA)HA experiments. A partial automatic assignment of the backbone and 13 C β resonances was obtained with the program MARS [30] and then completed manually. Several 1 H α assignments were obtained or confirmed in the 1 H, 13 C HSQC spectrum. The assignment has been deposited in the BioMagRes-Bank (accession number 26935). Secondary chemical shift values were calculated as the difference between the measured chemical shifts and their amino-acid specific random-coil values [31] using nearest-neighbor amino acid sequence corrections [32]. The latter were obtained from the web server http://www1.bio.ku.dk/english/research/bms/research/sbinlab/ groups/mak/randomcoil/script/, which also includes corrections for temperature and pH effect [33]. Backbone amide 15 N transverse (T 2 ) relaxation times were measured at 25˚C and 81.1 MHz 15 N Larmor frequency. The pulse sequence for 1 H detected 15 N T 2 relaxation measurement [34] was implemented in a series of two dimensional 1 H-15 N HSQC spectra with different 15 N T 2 relaxation delays (48.05, 64.06, 80.08, 128. 13, 192.19, 256.26, and 400.4 ms). The HSQC experiments were acquired in an interleaved mode and the intensity of the backbone amide signals was fitted to a single exponential decay. Signal overlap prevented the reliable measurement of the signal intensity for a number of residues, while the C-terminus yielded an anomalously high T 2 value with a very large error and was excluded from the analysis.
Small-Angle X-ray Scattering (SAXS) measurement and analysis SAXS measurements were carried out at the BM-29 BioSAXS beamline at the ESRF Storage Ring (Grenoble, France) using a sample-to-detector distance of 2.87 m [35]. KISS1 samples were measured at 10˚C (to minimize radiation damage) and concentrations of 4.0 and 1.0 g/L in PBS, pH 7.0, 1 mM DTT. The scattering pattern of the buffer solution was recorded before and after measuring each protein sample. The obtained scattering profiles covered a momentum transfer range of 0.003 < s < 0.49 Å -1 . A comparison of ten consecutive X-ray exposures indicated the absence of radiation damage. The final curves at each concentration were derived after subtracting the averaged buffer from the protein patterns using standard protocols within PRIMUS [36]. SAXS curves at both concentrations were merged to derive a final profile that was used for the subsequent structural analysis. The forward scattering, I(0), and the radius of gyration, R g , were evaluated using Guinier's approximation [37], assuming that at very small angles (s < 1.3/R g ), the intensity can be described as I(s) = I(0) exp(−(sR g ) 2 /3). The pairwise distance distribution function, p(r), and the maximum particle diameter, D max , were computed with GNOM [38] using a momentum transfer range of 0.013 < s < 0.29 Å -1 .
The ensemble optimization method (EOM) was used to describe the overall properties of KISS1 in solution [24]. To this aim a pool of 10,000 conformations was built with Flexible-Meccano [39,40] and side-chains were added with Sccomp [41]. For each conformation the theoretical SAXS curve was computed with Crysol [42] and used for the EOM sub-ensemble optimization. Two hundred independent EOM runs with 50 selected conformers were performed using the χ 2 statistical test as the selection criterion. The resulting ensemble composed of 10,000 conformations was structurally evaluated using its R g distribution.

Sequence analysis and disorder prediction
The amino acid composition of KISS1 shows a high content of polar residues (more than 60%, a feature that is typical for intrinsically disordered proteins [43]. The degree of disorder in KISS1 was predicted with the metaPrDOS web server that integrates the results of eight different methods [44]. Regions with a disorder tendency smaller than 0.5 comprise residues 41-53 and 91-123 (Fig 1). The latter region with reduced disorder tendency is largely contained within Kisspeptin54, the major fragment of KISS1 that encompasses residues 68-121. A prediction of protein binding regions with the web server ANCHOR [45] identified two long segments (18-71 and 84-127) that contain both regions with reduced disorder tendency.

Protein sample preparation
The polypeptide studied here corresponds to the full length KISS1 amino acid sequence (as described in Uniprot entry Q15726) without the secretion signal peptide, and encompass KISS1 residues 19-138 (with a theoretical molecular weight of 12.6 kDa). No extraneous residues that could perturb the native structural features of the protein were introduced. Since no affinity tag was used, high expression levels of the recombinant gene were necessary for efficient protein purification. Considering that the human cDNA sequence coded two leucine residues with codons of low abundance in E. coli, a synthetic gene with codons optimized for bacterial expression was used. KISS1 was then purified by anion exchange, reverse phase, and gel filtration chromatographic separations. Still the yield of pure protein was very low, between 0.1 and 0.6 mg/L of bacterial culture.

Structural analysis by SEC-MALS and CD
The protein eluted from the SEC column at a volume corresponding to an apparent molar mass of 28 kDa, according to calibration with molecular weight standards. The mass derived from MALS data is 11.7 kDa (Fig 2A), however, demonstrating that Kiss1 is monomeric. These inconsistent results indicate that Kiss1 is either folded with an elongated shape, or flexible and disordered [46]. In either case it would elute at a smaller volume than a globular protein of the same mass.
The far-UV CD spectrum of Kiss1 shows a minimum at 198 nm and a shoulder at 226 nm, consistent with a predominantly random-coil protein with little secondary structure (Fig 2B). The thermal denaturation curve of Kiss1 followed by the changes in the CD signal at 222 nm ( Fig 2C) does not show any cooperative folding-unfolding transition, indicating that KISS1 lacks a defined tertiary structure.

NMR assignment and analysis
The resonances of the KISS1 backbone and side chain C β nuclei were assigned from the standard suite of 3D triple resonance spectra. For the N-terminal residues G19 and E20 no signals could be observed in the 1 H-15 N HSQC spectrum. For the remaining 103 non-proline residues the 1 H-15 N HSQC signals could be unambiguously assigned, as shown in Fig 3. This spectrum also shows a few minor signals with about 5% of the major signals' intensity. As some of them appear to be duplicate signals of residues preceded by a proline we suggest that they arise from small populations of KISS1 with cis isomeric prolyl peptide bonds. This suggestion is further supported by the 13 C β chemical shifts of the corresponding proline residues, which are close to the average value measured for cis prolines in proteins [47].
The low dispersion of the backbone 1 H N chemical shifts observed in the HSQC spectrum indicates that KISS1 is largely disordered and flexible under native conditions. Yet, two regions of KISS1 around residues 40 and 115 show relatively short transverse relaxation time T 2 values (Fig 1) and correspond to sequences with reduced predicted disorder tendency. A shorter than average 15 N T 2 relaxation time results from restricted local flexibility on a fast (ns to ps) timescale, pointing to a local deviation from the otherwise unrestricted and uncorrelated mobility characteristic for IDPs. There is, however, some inverse correlation between local residue bulkiness and 15 N T 2 relaxation times (Fig 1), suggesting that local restrictions in polypeptide chain dynamics could also be explained by steric hindrance between side chains and the backbone, as found for other IDPs [48].
The NMR chemical shift deviations were computed with respect to random coil values obtained from small disordered peptides [31]. Large deviations were observed only for the nuclei of the C-terminal residues and for the C β of C53 (Fig 4). These deviations may derive from imperfection in the reference data set of random coil values, as noted before [49], rather than to local conformational preferences. For all other residues the chemical shift differences are within the RMSD measured in IDPs [50]. While the H α , C 0 , C α shift deviations of residues 34 to 42 might suggest a local tendency for helicity, the values are very small and the tendency is not confirmed by the C β shift deviations. In summary, our analysis of chemical shift deviations and 15 N T 2 relaxation times does not provide clear evidence for any conformational preference (Fig 4), suggesting that KISS1 behaves as a random coil polypeptide. However the ensemble of local conformations at the protonated histidine side chains might be different at pH 5.7 than at 7.0, used for the other measurements.

SAXS analysis
We furthermore collected SAXS data in order to probe the overall properties of KISS1 in solution [52]. Yet, the SAXS curve shown in Fig 5A presents no distinct features, and the Kratky representation is typical for a disordered protein, with a monotonic increase of I(s)s 2 with the momentum transfer s. We observed a concentration effect on the SAXS data at the highest concentration used, and we merged curves measured at 1 and 4 mg/mL to reduce interparticle interactions while preserving a good signal to noise ratio. We used the Scåtter program [53] to calculate the molecular weight of the particles with a result of 9 kDa, indicating that the SAXS data correspond with a monomeric protein. An analysis of the smallest angle data by Guinier's approach for a momentum transfer range sÁR g < 1.3 (where R g is the radius of gyration) indicates that KISS1 has an average radius of gyration of 34.7 ± 0.5 Å. This value is slightly larger than expected for an IDP of 120 residues (R g RC = 30.1 Å) suggesting that KISS1 could transiently adopt more extended structures [54].
In order to glean information on the conformational space sampled by KISS1 in solution from our SAXS data we modeled an atomistic ensemble of 10,000 KISS1 conformers using Flexible-Meccano [39], where torsion angle pairs were selected randomly from a database of amino acid specific conformations in loop regions of high-resolution X-ray structures [40]. After adding side-chains, the theoretical SAXS profile for each generated conformation was computed and then averaged over the entire ensemble. The resulting curve (Fig 5B), which represents a canonical random coil, agrees well with the experimental SAXS curve for KISS1 (χ i 2 = 0.87). Yet, some systematic deviations are observed for small momentum transfer s 0.2 Å -1 . A more detailed picture of the molecular sizes and shapes adopted by KISS1 in solution can be obtained via the Ensemble Optimization Method (EOM) [24] where a genetic algorithm selects a sub-ensemble of conformations that describes the experimental curve better than a random coil model (Fig 5B). EOM application to the SAXS curve of KISS1 yields a set of conformations that cover a broad range of molecular sizes, with a radius of gyration between 20 and 60 Å, consistent with a high level of disorder and flexibility (Fig 5C). Compared to the R g distribution within the initial pool of conformations, which represents the canonical disorder of KISS1, the subset selected by EOM contains more conformations with larger R g values. This result indicates a preference for extended conformations of KISS1 in solution.

Discussion
The human tumor suppressor protein KISS1 is monomeric in solution, as shown by our SEC--MALS data, and CD measurements indicate no tertiary structure and little, if any, secondary structure. Thus, as found for many human cancer related proteins and as suggested also by its amino acid sequence, human KISS1 is an intrinsically disordered protein. The vast conformational heterogeneity of the chain makes it challenging to probe for local conformational preferences since all experimental data is ensemble averaged [55]. Nevertheless, we have been able to glean insight at the residue level by NMR. An analysis of chemical shifts did not consistently indicate regions with preferential conformations since differences with random coil values are small and within the range of the RMSD observed in IDPs [50]. It has been reported that Kisspeptin13 (residues 109-121) adopts helical structure in SDS micelles; this KISS1 fragment is a pharmacophore that served to identify hits with submicromolar affinity for the metastin receptor [56]. For full length KISS1 in aqueous solution, however, we do not find evidence for any preferential helical conformation in the same region. The same holds for the region corresponding to Kisspeptin54, which largely shows a random coil behavior as for the rest of the molecule. This is consistent with previously reported data on the isolated Kisspeptin54 fragment, for which a 3.7% total helical content was observed by CD but no distinguishable structural elements were detected by NMR [57]. The other sampled NMR parameter, 15 N T 2 relaxation times, does show some local deviations; however, the local reduction of flexibility appears to be due to sequence specific steric obstruction, rather than conformational rigidity. Overall, our NMR analysis of KISS1 indicates random coil behavior without local conformational preferences. A global tendency to populate more extended conformations, with a larger radius of gyration than expected for a random coil is suggested by our analysis of the SAXS data analysis. This tendency may be favored to mitigate positive charge repulsion between the abundant basic residues within the KISS1 sequence. The functional role of IDPs presumably derives from their conformational versatility, allowing for interactions with different ligands, and for regulation through post-translational modifications at accessible sites [58]. In the case of KISS1, apart from phosphorylation at residue Y110 [59], the major modification is proteolysis by furin and C-terminal amidation by peptidyl-glycine-α-amidating monooxygenase [7]. KISS1 and its fragments interact with KISS1R, their GPCR receptor [9], but the mode of binding is unknown. Intrinsically disordered proteins may adopt folded or partially folded conformations upon binding to their partners [60]. This may occur via i) "induced fit" if the binding partner induces the disordered protein to adopt the conformation of the bound state, ii) "conformational selection" if the binding partner selects a favorable conformation from the ensemble present in the free state, or iii) a mixture of both mechanisms [61][62][63]. In the case of KISS1, the induced fit mechanism might be the most likely one for binding to KISS1R since there appears to be no preferred conformation in the solution state ensemble, although the situation might be different in the vicinity of the cell membrane. KISS1 has also been implicated in enhanced mitochondrial biogenesis by increasing the levels of peroxisome proliferator-activated receptor-γ co-activator 1α (PGC1α), a transcriptional activator for many genes regulating mitochondrial mass and metabolism. The link between KISS1 and PGC1α might be the KISS1 interaction with ubiquilin-1 that protects PGC1α from degradation. However, the structural basis of the KISS1 interaction with ubiquilin-1 is unknown. Our work on full-length KISS1 lays the basis for further structural studies on its interactions with physiological partners.