Observation of pH-Dependent Residual Structure in the Pmel17 Repeat Domain and the Implication for its Amyloid Formation

Abstract

The varying conformational states of amyloid-forming protein monomers can determine their fibrillation outcome. In this study we utilize solution NMR and the paramagnetic relaxation enhancement (PRE) effect to observe monomer properties of the repeat domain (RPT) from a human functional amyloid, premelanosomal protein, Pmel17. After excision from the full-length protein RPT can self-assemble into amyloid fibrils, functioning as a scaffold for melanin deposition. Here, we report possible conformational states of the short RPT (sRPT) isoform, which has been demonstrated to be a fibrillation nucleator. NMR experiments were performed to determine conformational differences in sRPT by comparing aggregation-prone vs. non-aggregating solution conditions. We observed significant chemical shift perturbations localized to residues near the C-terminus, demonstrating that the local chemical environment of the amyloid core region is highly sensitive to changes in pH. Next, we introduced cysteine point mutations for covalent attachment of PRE ligands to sRPT to facilitate observation of intramolecular interactions. We also utilized solvent PRE molecules with opposing charges to measure changes in the electrostatic potential of sRPT in different pH environments. These observed PRE effects offer insight into initial molecular events that might promote intermolecular interactions, which can go on to trigger the fibrillation. Taken together, our results show that sRPT monomers adopt a conformation inconsistent with fully random coil at neutral pH and undergo conformational changes at lower pH values. These observations highlight a regulatory mechanisms via organelle-associated pH condition that can affect fibrillation activity of proteins like RPT.

Graphical Abstract

Introduction

Many degenerative diseases are associated with accumulation of cross-β filamentous protein assemblies referred to as amyloid fibrils. Amyloid fibrils can accumulate in several organ and tissue types, and are correlated with cellular senescence.¹ Individual fibrils are comprised of repeating units of the misaggregating proteins that are stabilized into interlocking β-sheet conformations which spiral into the larger fibrillar structure with a regular periodicity. Macroaggregation of amyloid fibrils into plaques are markers for advanced stages of amyloidosis. Amyloid fibrils are incredibly stable and represent some of the most degradation-resistant complexes known to biology.² Initially, amyloid aggregations were believed to be permanent structures, with accumulation of plaques within afflicted organs akin to scar tissue that reported on previous molecular damage. However, a consensus has since emerged that certain amyloid fibrils can have specific beneficial pathologies, and dynamic fibril assembly might be a common biological phenomenon. Amyloid fibrils can also undergo disaggregation mechanisms, and turn-over is an active process in many amyloidosis diseases.³ These observations lead to the hypothesis that amyloid fibrils could be catalytic and non-toxic in some biological pathways while regulated by aggregation/disaggregation mechanisms.⁴

These “functional amyloids” have been characterized in organisms ranging from prokaryotes to homo sapiens and are involved in diverse pathologies.^{5, 6} Unicellular organisms use functional amyloids to modulate cell surface activities involving mechanosensitivity and in the production of biofilms. Multicellular organisms have evolved functional amyloids contributing to regular biological pathways including reproduction, immunity, hormone regulation, pigment biosynthesis, and neuronal plasticity. The reversible fibrillation qualities of these proteins allow researchers to probe intermediate steps along the fibrillation mechanisms. New information gained from studying these potentially tunable amyloid systems can provide valuable insight into the mechanisms of disease-related amyloids.⁷

Here we present new results from probing conformational states of premelanosomal protein (Pmel17), one of the most well-studied functional amyloids. Pmel17 is found in melanocytes, a type of specialized skin and eye tissue cells. These cells have acidic organelles akin to lysosomes called melanosomes, where the biosynthesis of melanin occurs. The compartmentalized pH can be as low as 3 within “heavily melanated” melanosomes.⁸ During the stages of melanosome maturation the pH drops to below 4, and Pmel17 is proteolytically cleaved until a relatively short, disordered repeat (RPT) domain is left.^9–11 Two isoforms of RPT, the short (sRPT) and long (lRPT) variants, combine to form regular amyloid structures. The resulting striated, fibrillar structures are easily visible using tissue staining as they become a platform catalyzing biosynthesis of melanin while sequestering toxic reaction intermediates.^12–14 The Pmel17 amyloid only forms under acidic pH conditions within the melanosome and rapidly dissociates under neutral cytosolic conditions.^{15, 16} This serves to diminish toxic effects by compartmentalizing the conditions in which this amyloid is stable.

RPT fibrillation rate has been characterized to be pH-dependent with shorter lag phases at lower pH. ^{16, 17} Furthermore, sRPT has been characterized to be a nucleator for the fibrillation reaction and by several metrics can fibrillate more quickly versus lRPT.¹⁸ This investigation characterizes interactions between human sRPT monomers as they begin to aggregate. NMR experiments were performed at a range of pH values (4–7) to determine conformational differences of sRPT in aggregation-prone (low pH) and non-aggregating (high pH) solution conditions. We observed significant chemical shift perturbations localized to residues near the C-terminus, demonstrating that the local chemical environment of the amyloid core region is sensitive to changes in pH even in the monomer form. Molecular events immediately preceding fibrillation can be detected by ligating amyloid-forming proteins with paramagnetic relaxation enhancement (PRE) ligands. Therefore, we introduced several cysteine point mutations to covalently attach PRE ligands to sRPT for observation of potential intra and intermolecular interactions.

These PRE effects can indicate potential initial molecular events to facilitate intermolecular interactions, which can go on to trigger the fibrillation. Thus, by raising or lowering solution pH, the relative population of this state would be modulated. We also utilized complementary pairs of charged solvent PRE (sPRE) molecules to measure the electrostatic potential of sRPT at pH 4 or 6. We observed that the negatively charged protein becomes charge neutralized at pH 4 with several positively charged regions that may be involved in the fibrillation mechanism. Taken together, these results show that sRPT monomers adopt a conformation inconsistent with fully random coil at neutral pH and undergo conformational changes at lower pH values. It is probable that residual structure exists that is more or less protected by the terminal regions as a function of pH. PRE contacts describe a residual structure with two turns that make several contacts over the amyloid core-forming region near residue W381 with increasing strength as a function of decreasing pH. These observations highlight a regulatory mechanism that can affect fibrillation activity of proteins like sRPT.

Materials and Methods

Protein Expression and Purification

A plasmid containing the WT sRPT construct in a pET21a vector was used as previously described.¹⁹ Mutants were generated using a QuikChange^® II XL Site-Directed Mutagenesis Kit (Agilent Technologies, Santa Clara, CA) and custom primers. A list of these primers is available for each mutant in Table S1. The WT sRPT and each of the mutants were expressed and purified as previously described. ^{16, 18, 19} Purification and storage buffers were supplemented with 2 mM of DTT for the cysteine mutants to inhibit dimerization. For ¹⁵N isotopic labeling, a standard minimal media expression protocol was followed.²⁰ Bacterial cultures were kept at 37°C with constant agitation and aeration during expression. Overnight starter cultures were inoculated from sRPT plasmids transformed into BL21 Gold DE3 competent cells (Agilent Technologies). The next morning the overnight culture was combined with the larger volume to form the expression culture. Growth was allowed to proceed for 30 min or until a 600 nm optical dispersion of 0.6 was observed. At this point expression was induced with addition of 1.5 mM IPTG. The expression culture was supplemented with 5 mL of 12.5% D-glucose at the time of induction and twice more at one and two hours post-induction. After 3.5 hours of induced growth cells were collected via centrifugation. From this point cell lysis and purification is the same as for non-isotopically labeled sRPT.

NMR Spectroscopy and Data Analysis

NMR experiments were recorded at 295K on a Bruker 600 MHz spectrometer equipped with a cryogenic probe. All samples contained about 100 μM of sRPT protein. HSQC experiments were collected with 8 scans and 128 complex points in the indirect dimension. ¹⁵N T₁ and T₂ relaxation time measurements were conducted using pulse sequences modified from those published previously.^{21, 22} T₁ experiments were collected in a non-interleaved dataset using 8 delay values ranging from 4 to 800 ms. T₂ experiments were collected using CPMG-based interleaved experiments with 8 delay values ranging from 4 to 400 ms. T₂ was calculated by fitting the intensity curve to an exponential decay function. Error values were extrapolated by Monte Carlo analysis with reference to the experimental signal to noise ratio as described previously.²³

Buffer systems representing lower pH aggregation prone conditions versus higher pH non-fibrillating conditions were used as previously described.^{18, 19} Briefly, 20 mM sodium acetate with 100 mM NaCl was used when buffering from pH 4 to 6. The higher pH buffering system was 20 mM MES with 100 mM NaCl, used when buffering from pH 5 to 7. NMR spectra were processed using NMRPipe.²⁴ Processed spectra were visualized and assigned using POKY.^{25, 26} Chemical shift perturbations for the pH titration experiment were calculated as described by Williamson with reference to the spectrum at pH 7.²⁷ Adjustments to pH were made by titrating ~0.5 μL of 3M HCl into the NMR samples. Peak intensity lists from relaxation measurement datasets were extracted using in-house Python scripts from peak lists assigned from reference spectra using POKY.

Resonance assignments for sRPT were referenced to those previously published for lRPT.¹⁵ Some major shifting was observed for residues near the isoform splice site, but otherwise most chemical shifts remained unchanged for the equivalent residues in sRPT versus lRPT. Assignments were verified by collecting ¹⁵N 3D NOESY experiments at pH 5 and 6 with 300 ms mixing time, and walking resonance through the backbone for uncertain assignments. This was done for the mutants as well.

Paramagnetic Ligation

The first paramagnetic tag used for chemical ligation of cysteine mutants was MTSL, a methanesulfonothioate with the paramagnetic moiety (a 1-Oxyl-2,2,5,5-tetramethylpyrroline group, commonly called PROXYL) on the opposite side of the disulfide bond. The PROXYL moiety contains the relaxation-enhancing nitroxide with its unpaired electron. MTSL reacts with cysteines to covalently ligate PROXYL to the amino acid side chain via a reducible disulfide bridge with methanesulfinate as the leaving group. After losing the leaving group, the covalently attached ligand is known as MTSL. Prior to ligation reactions, protein stocks were reduced with 10 mM of DTT for two hours at room temperature. The reductant was then removed via buffer exchange into the labeling buffer (20 mM Tris-HCl, pH 8 and 100 mM NaCl) using a PD-10 size exclusion column (Cytiva, Marlborough, MA). The eluted protein was reacted with 10-fold excess of MTSL (Caymen Chemical, Ann Arbor, MI) and allowed to incubate at room temperature overnight with rotation after bringing the final volume up to ~7 mL. The following day excess MTSL ligand was removed via several rounds of buffer exchange into the storage buffer or reaction buffer via spin filtration. Labeling efficiency was determined to be >98% via LC/MS.

Labeling using the lanthanide series X-DOTA cage complexes has been previously described.^{28, 29} These ligands can easily be attached to cysteine residues by modifying one of the DOTA arms with a pyridyl disulfide group. This enables conjugation to cysteines via a disulfide bond with 2-mercaptopyridine (SPy) as the leaving group.²⁸ The reaction buffer was the same as for MTSL, but reductant is not necessary. Mutants were reacted with 2-fold excess of the DOTA ligands overnight at room temperature. Labeling efficiency was determined to be >98% via LC/MS. Gd-DOTA-SPy was used for PRE experiments and Tm-DOTA-SPy was used for PCS experiments. Lu-DOTA-SPy was used as a diamagnetic reference for both. X-DOTA-SPy ligands were synthesized and purified by the NHLBI Chemistry and Synthesis Center.

Measuring Paramagnetic Relaxation Enhancement

The two-point delay method was used for measuring the ¹H transverse relaxation rates (R₂) and PRE for the backbone chain amide of each sRPT amino acid residue.^30–33 The first delay time (T_a) for each dataset was set to about 50 μs and the second delay time (T_b) was set on a per-dataset basis. T_b was determined by the delay needed to attenuate the signal in the 1D version of the experiment by ~75%. For most intramolecular PRE experiments T_b was set to about 74 ms. For intermolecular experiments T_b was set to about 145 ms. The diamagnetic reference experiments should use the same exact delay values. This ensures accurate calculation of the PRE values while negating other experimental factors. Additionally, a version of the experiment using a soft pulse for gradient water suppression will be required to accommodate a given disordered protein’s long relaxation times. PRE was calculated using Equation 1:

$$PRE=R_{2,para}-R_{2,dia}=\frac{1}{T_b-T_a}ln\frac{I_{dia,T_b}\times I_{para,T_a}}{I_{{dia,T}_a}\times I_{{para,T}_b}}$$ [1]

Where ${\mathrm{I}}_{\text{para}}$ and ${\mathrm{I}}_{\text{dia}}$ are resonance intensities from the paramagnetic and diamagnetic experiments, respectively, at each delay time. When calculating ${\mathrm{R}}_2$ from the paramagnetic or diamagnetic experiments separately, Equation 1 is reduced to Equation 2:

$$R_2=\frac{ln\left(\frac{I_{T_b}}{I_{T_a}}\right)}{-T_b+T_a}$$ [2]

Diamagnetic controls to obtain the reference ${\mathrm{R}}_2$ values were collected using the same sample with the nitroxide radical quenched using ascorbate. The difference in ${\mathrm{R}}_2$ values between the paramagnetic and diamagnetic experiments gives a unique PRE value for each amino acid residue.

Modeling of sRPT ensembles with PRE potentials

Ensemble calculations of sRPT with and without PRE restraints used Xplor-NIH version 3.4.^{34, 35} The input sRPT structure was in silico labeled with the CTSA tag at sites 317, 324, 371, and 401 and used as the input for 50 member ensembles. Xplor BOND, ANGL, IMPR, and torsionDB energy terms were used throughout. The Xplor-NIH repel potential was used at high temperature (3000 K) and the eefx implicit solvent potential was used at low temperature (298.15 K). When present, PRE restraints used the prePot energy term in “Sb” mode with square well potentials and a fixed correlation time (${\mathrm{\tau }}_{\mathrm{c}}$) of 5 ns, emulating prior work with $\mathrm{\alpha }$-synuclein.³⁶

Both cases followed the same general protocol, starting from an extended conformation, after which torsion angles were randomized. An initial high-temperature dynamics stage was run to 10,000 steps followed by 200 steps of high-temperature minimization. The temperature was then swapped to the lower value, followed by an additional 200 steps of low-temperature minimization. This was followed by the main 1000 step minimization in torsion-angle space and a final 500 step cartesian minimization. These final two minimization stages were either performed without the PRE energy terms for the control ensemble, or with them for the restrained ensemble. To protect against poor starting conformations, the minimization was rerun if the overall PRE correlation was not > 0.9.

In addition to Xplor-NIH for correlations, the analysis used PyMOL and Python. Distances were analyzed using the metric from Salmon et al. ³⁶

$${\Delta }_{ij}=\mathrm{l}\mathrm{o}\mathrm{g}\left(〈d_{ij}〉/〈d_{ij}^0〉\right)$$ [3]

where $d_{ij}$ refers to the interresidue distances in the ensemble with PRE restraints and $d_{ij}^0$ to the ensemble without them. Radii of gyration were computed in PyMOL.

Measuring Electrostatic Potential

The sPRE-based ESP measurement technique was use as described by Iwahara et al.,^{37, 38} and as applied to disordered proteins by Kay et al.³⁹. Complementary datasets using a pair of sPRE molecules allows calculation of a disordered protein’s ESP by applying a prefactor to the ratio of PRE values induced by two different sPRE molecules as shown in Equation 4:

$$\mathrm{E}\mathrm{l}\mathrm{e}\mathrm{c}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{s}\mathrm{t}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{c}\mathrm{P}\mathrm{o}\mathrm{t}\mathrm{e}\mathrm{n}\mathrm{t}\mathrm{i}\mathrm{a}\mathrm{l}=-\frac{{\mathrm{k}}_{\mathrm{B}}\mathrm{T}}{∆\mathrm{e}}\left({\mathrm{P}\mathrm{R}\mathrm{E}}_{\mathrm{j}}/{\mathrm{P}\mathrm{R}\mathrm{E}}_{\mathrm{k}}\right)$$ [4]

Where ${\mathrm{P}\mathrm{R}\mathrm{E}}_{\mathrm{j}}$ is the PRE value calculated from the dataset with the higher charge state and ${\mathrm{P}\mathrm{R}\mathrm{E}}_{\mathrm{k}}$ is the PRE value calculated from the dataset with the lower charge state. For the prefactor, kB is the Boltzmann constant and $\mathrm{T}$ is the temperature in Kelvin. In the denominator, $\mathrm{\Delta }$ is the difference in charge between the selected ligand pairs and e is the fundamental charge value in Coulomb units. It is best practice to use at least three different sPRE molecules with the same paramagnetic moiety and three differently charged but similar functional groups. This way the three combinations of pairs can be used as an internal check for the accuracy of the ESP calculation. The PROXYL moiety is again employed. Now modified with an aminomethyl group for a cationic molecule, a carbamoyl group for a neutral derivative, or a carboxy group for the anionic version. The three molecules: aminomethyl-PROXYL, carbamoyl-PROXYL, and carboxy-PROXYL (all three obtained from MilliporeSigma, Burlington, MA) were measured for their PRE effect on WT sRPT at pH 4 or 6. Those PRE values were then used with Equation 4 to calculate a unique ESP value for each amino acid residue. We note that carboxy-PROXYL was not usable for pH 4 experiments as its carboxy group would be mostly in a protonated state at the lower pH, effectively turning it into a second neutral ligand. We also chose to double the ligand concentration from 5 mM at pH 6 to 10 mM at pH 4. The PRE values at pH 4 using 5 mM of the ligands were very small, all less than 2 Hz (data not shown). No clear differences between the PRE values at pH 4 were observable even with 10 mM of the ligands, demonstrating that sRPT really is neutralized at the lower pH and does not have a particular affinity for any charge over most of the residues.

Results

sRPT is pH-sensitive and has NMR parameters inconsistent with fully random coil

The wild-type (WT) sRPT protein was probed for varying conformational states as a function of pH. Chemical shift perturbation (CSP) analysis was used to describe local environmental changes that sRPT undergoes during the processes leading up to melanogenesis. A “dove-tail” titration using two samples was performed to accommodate the wide buffer range (pH 7 to 3) that sRPT can experience inside the melanosome. Chemical shifts were measured over a 15-point titration using the heteronuclear single quantum coherence (HSQC) NMR experiment. ²⁷ (Figure 1A and Figure S1) CSPs increase significantly below pH 5, but this is not experienced by all residues uniformly. (Figure 1B) Comparing chemical shifts at the start and end of the titration shows that the majority of significantly changing resonances are associated with residues localized around the amyloid-core forming region (Figure 1C).¹⁶ Other residues with significant changes in chemical shift are acidic residues that would become more protonated during the titration. However, the CSP plot also shows significant perturbations of neutral and basic residues in the amyloid core, implying that the localized chemical environment in this region must be undergoing dynamic changes not explainable by the pH-effect alone. Furthermore, the acidic residues in the core-forming region experience larger CSPs versus the more N-terminal acidic residues. For example, glutamic acid residues E318 or E331 at the N-terminus versus E383 or E393 in the core-forming region (Figure 1B). The most N-terminal glutamic acid shifts much less than the same residue type that is in the core-forming region.

Figure 1. pH-Dependent Chemical Shift Perturbation Analysis of sRPT.

(A) A selection of overlayed HSQC spectra from the 15-point pH titration. Spectra are colored according to measured pH value of the sample after adjusting pH with 0.5 M HCl: blue is 6.0, green is 5.4, yellow is 4.8, orange is 4.2, and red is 3.8. Some residues with significant CSP are labeled. (B) Four select residues are shown with their entire CSP plots. The two glutamic acid residues represent acidic residues near the N- or C-terminus. W381 represents an important residue in the mature fibril core that undergoes significant CSP. M366 represents a residue that does not undergo significant CSP. (C) The per-residue CSP values from pH 6.0 versus 3.8. Residues with titratable carboxylic acids are labeled. Spaces indicate the presence of unobservable proline residues. The black bar under the residue labels indicates the previously described core-forming region.¹⁶

NMR relaxation times for WT sRPT in the different pH conditions were also measured. The ¹⁵N spin-lattice (T₁) and spin-spin (T₂) relaxation times remained relatively constant over three pH values, suggesting no significant aggregation occurring at pH 5 within the duration of the NMR experiments. (Figure S2 and Table S2) The C-terminus of sRPT has a nearly 50 ms reduction in the average T₂ relaxation times across all residues versus the N-terminus. A minimum is observed over the amyloid-core forming region. This indicates there might be some slight order over this region compared to the rest of the protein which is more dynamic and typical of disordered proteins. The sharp increase in T₁ and T₂ relaxation times at the C-terminus is typically a feature of globular proteins, as their termini are completely disordered, and is inconsistent with a fully random coil conformation for sRPT. Finally, the N-terminus does not feature an increase in relaxation time like the C-terminus, indicating that it too does not behave as a pure random coil conformation. This is consistent with the CSP data showing some shifts at the N-terminal residues T316-A319. Furthermore, signals were not observed for residues H313 and R314 until the pH was very low; and G312 was never observed, suggesting solvent exchange might be prevalent for the N-terminal residues. (Figure S1).

Paramagnetic Tags Indicate Long-Range Intramolecular Contacts

Four sRPT cysteine mutations were made for labeling with PRE tags. Mutation sites were selected based on similarity to cysteines and a low likelihood to interfere with fibrillation activity. Fortunately, sRPT does not contain any native cysteines. Two alanine residues near the N-terminus, A317 and A324, were selected for mutation along with two serine residues near the C-terminus, S371 and S401. The four cysteine mutants: A317C, A324C, S371C, and S401C were recombinantly expressed and covalently labeled with a paramagnetic nitroxide radical, MTSL, a commonly used tag for introducing a non-native paramagnetic center to proteins.^{32, 36, 40–42}

These paramagnetically labeled sRPT mutants were further utilized for PRE experiments detecting nuclear spins in proximity to the nitroxide radical. The spins of atomic nuclei show varying enhanced relaxation rates (R₂) which depend on their distance to the unpaired electron due to the PRE effect. These types of PRE experiments have been well-established for measuring long-range contacts in disordered proteins.^{40, 43–48} We initially probed for intermolecular interactions between sRPT monomers using paramagnetically-tagged ¹⁴N protein in solution with ¹⁵N WT sRPT. In this experiment only signal from the WT protein is observable via NMR. Surprisingly, no specific intermolecular interactions were detected even at pH 5. Significant PRE with MTSL-labeled S371C was not detected (Figure S3A). All of the PRE values are around 0±1 Hz; and is interpreted as no significant dimerization with long associated lifetimes. An alternative PRE probe with a gadolinium-chelating DOTA cage (Gd-DOTA-M8-SPy), which should induce significantly larger PREs, produced similar negative results when attached to the S401C mutant (Figure S3B). Finally, MTSL-labeled A317C at pH 4 did produce an enhanced PRE with the lone aromatic W381 (Figure S3C). However the magnitude of the PRE data is overall not significant with all values still less than 1 Hz. It is important to point out that our experimental conditions, relatively low temperature and low protein concentration, were chosen to favor the monomer state of the protein to allow NMR measurements to be carried out. A significant and rapid loss of signal is expected once fibrillation occurs.

Given the lack of intermolecular contacts, we also used paramagnetically-tagged ¹⁵N mutants to detect intramolecular interactions within spin labeled sRPT (Figure S4). So, it is proposed that all of the PRE contacts described heretofore are a result of variation of the monomer conformations under different buffer conditions. Experiments were conducted at 295K (22° C) rather than 37° C to slow the fibrillation rate over the data collection period of more than 30 hours. Less than 10% signal loss was observed over the course of the experiments, indicating the majority of sRPT remains as soluble monomer under these conditions. If dimerization occurs the association is short-lived or they proceed to NMR-invisible higher-order oligomers, since no inter-molecular PRE was observed.

PRE values detecting intramolecular interactions near the labeling sites were obtained for each mutant at three acidic pH values: 6, 5, and 4. Generally, the observed PRE values tend to increase with decreasing pH (Figure S4). The fibrillation of sRPT in the cell requires that the pH within the melanosome lowers below pH 4, so it is reasonable to propose that the intramolecular contacts described by these PRE interactions are characterizing a pro-fibrillar conformation of the monomer as it becomes protonated under these conditions. Some long-range intramolecular interactions were detected for each mutant with more favorable contacts detected as a function of decreasing pH. N- to C-terminal contacts are clearly observed that increase when pH is lowered to 4.

To highlight only conformational changes that occur as pH is lowered, it makes sense to plot the difference in PRE values by subtracting values at pH 6 from those measured at pH 4 to determine which contacts are modulated by pH (Figure 2). This context more clearly shows the unique intramolecular contacts forming at low pH that comprise the presumed pro-fibrillar monomer conformations. The PRE technique is sensitive even to brief contacts that might be present at pH 4 but are less likely at pH 6. Residues that do not experience significant changes as a function of pH would have delta PRE values closer to the baseline. This data points to several segments of sRPT that come together, albeit transiently, over the amyloid-core forming region.

Figure 2. Delta PRE Values for MTSL-Ligated sRPT Mutants.

The experimental per-residue delta PRE values calculated from the difference between PRE values obtained at pH 6 and 4 for each MTSL-labeled cysteine mutant. (A)A317C, (B)A324C, (C)S371C, and (D)S401C all have different profiles that describe long-range PRE contacts to the labeling sites. Error bars are generated from duplicate experiments.

Overall, the two N-terminal mutants appear to reflect the same interaction (Figure 2A & B). They seem to be describing a residual turn where the N-terminus folds over to make contacts around residue T342. This turn additionally contacts the amyloid core-forming region around residues E380 and I391. No significant change in PRE values from residues adjacent to the N-terminal labeling sites was observed. They remain at the maximum range of around 20 Hz at each pH value (Figure S4 A & B). Elevated PRE values could be observed for residues forming contacts around residue T342 as well as to the core-forming region (E380 and I391) at all pH values, with the largest effect at pH 4. This suggests that this turn propensity and contact to the core-forming region exist immediately upon the complete proteolytic cleavage of Pmel17, and are stabilized at lower pH values.

The two C-terminal mutants each describe unique contacts. S371C indicates another residual turn potentially present in the C-terminal half of sRPT (Figure 2C), while S401C can be influenced by the core-forming region immediately preceding it (Figure 2D). Near S371C are an unusual number of amino acid signals broadened away C-terminal to the labeling site. This indicates some residues around E380-I390 are collapsing back to make contacts directly over the labeling site at C371. This can be seen for all three pH values (Figure S4C). Thus, this conformation is not pH triggered but must exist immediately upon proteolytic cleavage. The complete attenuation of signal greater than 10 residues away might point to a more stable conformation in this region relative to the N-terminus, consistent with the ¹⁵N relaxation data highlighting a less dynamic portion of sRPT corresponding to the amyloid core-forming region. This kind of signal attenuation is not observed with the N-terminal mutants because the labeling site is not in the middle of the expected first turn. Some PRE to residues T330-A340 are also detected in support of the PRE contacts indicated by the N-terminal mutants.

The S401C mutant describes contacts that are less direct. As indicated by the ¹⁵N relaxation dataset, this labeling site is in a region of sRPT that remains significantly more dynamic versus the rest of the protein. Unlike other mutants, however, the PRE values close to the S401C mutant site show strong pH dependence, with maximum PRE of 20, 12, and 8 Hz at pH 4, 5, and 6, respectively (Figure S4D). This is also reflected in the largest ΔPRE observed among all the mutants (Figure 2D). Interestingly the elevated ΔPRE values can be observed all the way to about residue S360, including the amyloid core-forming region. Perhaps the PRE data of this mutant is reflecting changes in the conformational ensemble involving the core region as the pH is lowered. In addition, a few PREs to the residues around T330-T335 are slightly enhanced. The same region is also highlighted by the N-terminal PRE tags, suggesting there is a weak interaction between the first and second turn.

Taken together, these PRE datasets describe a conformational transition of the sRPT monomer state that includes at least two residual turns which come together as the pH within the melanosome is lowered. Keep in mind, however, that these turns exist transiently or in low population. Since these turn interactions seem to correlate strongly with the pH, they might drive formation of higher order oligomers that go on to fibrillate. Fluorescence quenching of W381 has previously been monitored to quantify higher order oligomerization and fibrillation rates.^{16, 17} It is likely that residual contacts forming between the two turns are restricting the bond rotation of the W381 sidechain. This slow stabilization over time leads to measurable fluorescence quenching as RPT fibrillation proceeds. It is tantalizing to suspect that stabilizing these weak contacts is a requirement for fibrillation and this might be related to the relatively long lag phase described before exponential fibril growth. This might also explain why a larger percentage of the sRPT peptide backbone chain is more protected from enzymatic digestion relative to lRPT after fibrillation.¹⁶ We postulate that the contacts described by these PRE datasets form a low population conformation that is productive for higher order oligomerization.

Application of PRE Datasets as Restraints for Ensemble Calculation

To visualize potential conformers that satisfy our experimental PRE data, we modeled sRPT using Xplor-NIH, where the protein was in silico labeled with the MTSL at positions 317, 324, 371, and 401. We used a similar strategy to that used by Salmon et al. for α-synuclein,³⁶ where a target ensemble restrained by the PRE data was compared against a control ensemble without the experimental restraints. A selection of models rendered in Figure 3 highlight the variability of conformations that would satisfy the PRE restraints. The two residual turns indicated in magenta and cyan were estimated from the PRE profiles in Figure 2. The models usually show formation of the two residual turns with varying distances between each other.

Figure 3. PRE-based ensemble modeling.

Selected models of sRPT that illustrate how the two residual turns, in the cases where they are realized, might come together to form pro-fibrillar conformations. Each of these satisfy the PRE restraints. A key visualizes the first residual turn (magenta) and the second residual turn (cyan) with W381 marked in orange. Beginning and end residues of the residual turns are approximated based on the PRE profiles in Figure 2.

The resulting distance map for representative 50-membered ensembles with and without the four PRE potentials is shown in Figure 4A, where the overall correlation of the PRE-restrained ensemble was 0.97 with an overall quality factor of 0.11. The distance map highlights proximity (red) between the N-terminus and the 330–350 region, consistent with the formation of the proposed first turn. Additionally, there is a broad increase in proximity between the 340–360 region and the amyloid core-forming region, indicative of a second, broader turn in this region. The radii of gyration (Figure 4B) for the ensembles with the PRE restraints (22.88 ± 5.591698 Å) and without (23.45 ± 4.649254 Å) were similar to those seen previously for α-synuclein, with a visible skew towards smaller values when restraints were present. The proximity map reveals that the PRE profiles underestimate the length of the second residual turn and that it may include a larger N-terminal portion than expected.

Figure 4. Impact of PRE restraints on mutant sRPT ensembles.

(A) Proximity maps for 50 member sRPT ensembles incorporating PRE restraints for residues A317C, A324C, S371C, and S401C. Distances are plotted according to the metric ${\Delta }_{ij}=\mathrm{l}\mathrm{o}\mathrm{g}\left(〈d_{ij}〉/〈d_{ij}^0〉\right)$,³⁶ where $d_{ij}$ refers to the interresidue distances in the ensemble with PRE restraints and $d_{ij}^0$ to the ensemble without them, with shorter distances for the PRE ensemble shown in red and longer distances in blue. The addition of PRE restraints overrepresents proximity for the N-terminus (315–325) to the 330–350 region, and between the 340–360 region with the putative amyloid forming core(360–396). (B) Distribution of the radii of gyration for the ensembles with PRE restraints (blue) and without (red), highlighting the shorter skew in the presence of PRE restraints.

Pseudocontact Shifts Confirm N-terminal Turn and C-terminal Contacts

To validate the potential contacts identified by the PRE data above, we chose to use pseudocontact shifts (PCS), a paramagnetic phenomenon where (instead of relaxation enhancement) NMR signals will shift to a new ppm value. Lanthanide series metals chelated by rigid DOTA (1,4,7,10-tetraazacyclododecane-1,4,7,10-tetraacetic acid) cages have previously been validated for increasing chemical shift dispersion in disordered proteins when the PRE tag is covalently ligated.^{29, 49} The amount of shift observed is directly proportional to the observed nucleus’ distance and orientation relative to the lanthanide metal center. Therefore, they are sensitive to long-range contacts that can form in disordered protein. In addition, they can act as a molecular ruler to resolve signals from residues that are directly adjacent to the labeling site and thus can resolve signal overlap to confirm assignments.⁴⁹

To induce PCS in sRPT, we used a DOTA tag chelated to Thulium (Tm-M8-SPy) to label the A317C mutant and collect HSQC experiments. A diamagnetic version of the tag using lutetium as the chelated lanthanide (Lu-M8-SPy) was used to collect a reference spectrum (Figure S5). The PCS effect is observed here simply as the change in proton ppm shift value of the amino acid resonances between each HSQC spectrum. Amino acid residues of sRPT that are near the labeling site experience the largest shifts and other outlier shifts indicate potential long-range contacts. Note that values of the PCS are quite small, consistent with the PRE data, reflecting the low population of the residual structure.

The major contacts from the MTSL labeling experiments are also revealed in this context. Residue T342 has a relatively high PCS along with its surrounding residues. This might be near the “clinch” contact that forms the N-terminal residual turn. Similarly, residues T330, E331, and V332 also have large PCS values (Figure S6). This, however, may not reflect contact formation to the labeling site, as PCS affect is attenuated for residues adjacent to prolines (P328 and P337), which has been previously described with the FUS protein.⁴⁹ Finally, the contact to the residue W381 can be observed, albeit with very small PCS value. In addition to the above observations, the experiment was able to resolve many of the many of the overlapped signals from threonine and alanine residues with high sequence repeat. This allowed more accurate reporting of the resonance assignments for each amino acid residue.

Measuring Electrostatic Potential Visualizes Effects from Protonation at Low pH

The PRE effect has recently been applied to measure the electrostatic surface potential (ESP) of proteins.^{37, 38, 50} This can be done even without knowing any structural information, as is the case with disordered proteins.³⁹ Oppositely charged cosolutes containing a conserved moiety with an unpaired electron can report different PRE values depending on an individual residue’s charge sensitivity. (Figure S7) This type of PRE where the paramagnetic center is not covalently ligated to the protein is called solvent PRE (sPRE). This technique for measuring ESP is extremely valuable for disordered proteins because it can report information on relative charge across the protein sequence. These charge distributions might be able to confirm the possibility of observing residual contacts within the protein.

The ESP profiles for WT sRPT at pH 6 or pH 4 vary significantly (Figure 5). Residues with ESP values above or below the standard deviation from the average are indicated. At pH 6 sRPT is uniformly negatively charged. Residues that are the most negatively charged are represented by several serine and threonine residues. This is unexpected as the polarity of these uncharged hydroxyl groups should not have a greater contribution to negative charge over the deprotonated carboxylic acids. One exception is a glycine residue, G347. It has one of the lowest negative charge. All of these residues have PRE values lower than 0.5Hz in the presence of anionic solvent and they are one to three residues C-terminal to an acidic residue. Residues that are less negatively charged are overrepresented for valine and isoleucine amino acids as well as the immediate N-terminal residues. The hydrophobic residues are resistant to extreme charge variability, and the N-terminal residues may be influenced by a positively charged arginine at residue 314. The only other positively charged residue, another arginine at position 387, may have its apparent charge suppressed by the carboxylic acid of neighboring E388.

Figure 5. Measuring sRPT Per-Reside Electrostatic Potential.

Experimental values of per-residue electrostatic potentials for 100 μM of sRPT at pH 6 (A) and pH 4 (B). At pH 6 sRPT is completely negatively charged. When sRPT becomes more protonated at pH 4 (B), the protein becomes neutralized. The immediate N-terminus becomes positively charged while the C-terminus remains negatively charged. Note the y-axis ranges in each panel are not equal. Amino acid residue datapoints are colored if their electrostatic potenital values are above (blue) or below (red) the standard deviation from the average value in each calculation. At pH 6 residues above the standard deviation and more positively charged are T316, A317, E318, A319, V327, V332, V333, V353, V358, I359, and V364. At pH 4 these residues include T316, A317, E318, A319, G325, V332, S336, E344, V353, T378, and W381. At pH 6 residues below the standard deviation and more negatively charged are N321, T336, S346, G347, T349, S360, T361, S371, T372, T378, T385, A399, and S400. At pH 4 these residues include E370, E380, E383, I391, G396, D398, A399, S401, and S402. Error bars are generated from duplicate experiments.

At pH 4 sRPT is charge neutralized as the carboxylic acid residues are protonated. The most positively charged residues again include the immediate N-terminus along with a few others. Notably, one residue, T378, switches from being significantly negatively charged at pH 6 to significantly positively charged at pH 4. This is surely due to its proximity with W381 which becomes significantly positively charged as well. At pH 4 the hydrophobic interactions that are driving contacts around W381 and bring the two residual turns together can overcome the polarity of T378. Residues that retain a significant negative charge are clustered around the C-terminus, including all of the carboxylic acid-containing residues from the second turn to the end of the protein. These residues are more resistant to protonation and so must be retaining some residual negative charge as parts of the N-terminus become positively charged. These opposite charges can become an intramolecular attraction that switches on at low pH, which explains the increased PRE effect exhibited by the labeled mutants at lower pH values.

Overall, these results confirm previous investigations implicating the protonation state of these carboxylic acid-containing residues as integral to the fibrillation mechanism.^15–18 Measuring the structure-independent ESP like this provides additional context and indicates the charge state of individual carboxylic acids along the protein chain. These plots clearly show that acidic residues at the N-terminus behave differently and independently from acidic residues at the C-terminus. We propose that these distinct activities represent opposite charge states developing between the two detected residual turns. These opposite charges may contribute to a population shift toward residual ensembles productive for fibrillation.

Discussion

Beyond functional amyloids, disordered proteins in general have been observed to be an integral component to biological pathologies.⁵¹ No longer dismissed as nonsense proteins, their highly dynamic nature confers tunable properties where conformational states can be regulated in unexpected ways. Disordered proteins and their residual structure ensembles have become significant targets for drug discovery.⁵² NMR and PRE techniques in particular have proven to be a valuable tools toward probing of disordered proteins to identify residual structure and how that structure drives function.^{36, 53–59} Residual structure is an advantageous property of disordered proteins. They can easily be modulated by solution condition to achieve a level of regulation. By simply shifting the equilibrium between order and disorder it can drive up probability of events correlated with function. This more simplistic regulation mechanism may be unappreciated in different biological systems that rely on disordered proteins and functional amyloids. Other functional amyloids likely have residual structures in their monomer forms that are regulated indirectly like sRPT. This ensemble of ultra-low-population states containing residual structure may well be the key important determinant for regulating fibrillation and can be targeted for therapeutics.

The RPT isoforms give the regulation mechanism examples of compartmentalization and neutralization via extreme protonation. After sRPT undergoes proteolytic cleavage at the onset of melanogenesis, propensity for residual turn formation begins immediately as indicated by the PRE profiles of covalently tagged mutants. This explains the observation that RPTs can fibrillate at any acidic pH value with a range of kinetic rates. A small population of pro-fibrillar residual conformations will eventually oligomerize and nucleate fibrillation at any acidic pH. It is just a matter of regulating the residual ensemble formation rate via pH change. As pH is gradually lowered during the further stages of melanogenesis, stabilization of the residual turns occurs while interactions between them become stronger and drive ensemble populations toward pro-fibrillar conformations. The CSP results indicating the majority of conformational changes occurring over the amyloid core-forming region at lower pH values are consistent with the detected long-range intramolecular contacts from the PRE experiments. This residual structure drives conformations which collapse around the amyloid core-forming region and will then go on to oligomerize. This is also supported by the observed relaxation data showing a reduction in dynamics over the amyloid core-forming region. We propose that this reduced dynamic property of the core-forming region is due to this region’s proclivity to forming a large residual turn that other parts of sRPT can condense around. The condensed ensemble is necessary for the events that precede nucleation and fibrillation. There must be a transition from a relatively closed ensemble to open conformation primed for oligomerization. We aim to show in the future that this closed conformation regulated by the two residual turns can provide the optimal conditions for core-forming regions to align between monomers. Then a relatively more open conformation for the ensemble occurs at the start of nucleation as intermolecular interactions increase. Fine tuning the experimental conditions will be required to observe the nucleation phase of the fibrillation by NMR. Interestingly, those residues highlighted by our PRE experiments that are involved in forming long range intramolecular contacts (T330-T335, E370, and T372) seem to overlap with the recently found mutant hotspots (G325, V332, A340, E370, and S371) associated with inheritable ocular pigment dispersion leading to pigmentary glaucoma.⁶⁰ Note that these sites are outside of the core-forming region of sRPT.

The results of the charged sPRE experiments also corroborate with the hypothesis of residual turn formation. The CSP plots correlate with activity over the amyloid core-forming region and development of the positively charged regions as a function of decreasing pH. The sPRE experiment does not report on conformational changes of disordered proteins as relative PRE values were obtained from experiments using similar cosolutes with different charge. The detected sPRE therefore only reports on the relative charged state of each residue independent of any residual conformations. At pH 4 the ESP profile of sRPT indicates the N-terminus and some area of the core-forming region including W381 become positively charged. Taken together, we believe the correlation of positive charge with CSP indicates these regions of sRPT tend to make contacts within the residual structure ensemble. As the carboxylic acids become protonated at different rates, the switch in charge state of the first turn increases its likelihood to interact with the second turn. Opposite charges attract and more accessible hydrophobic contacts can drive a higher population of residual conformational ensembles that can promote fibrillation.

Conclusions

We have observed residual structure containing long-range interactions in the short isoform of the Pmel17 RPT functional amyloid. The interactions are not solely driven by charge-charge contacts and yet can be modulated by pH change in correlation with the function of sRPT in the melanosome. This work illustrates a novel coupling of pH to dynamics for an intrinsically disordered protein in the context of prefibrillation.