Modulation of assembly of TDP-43 low-complexity domain by heparin: From droplets to amyloid fibrils

Abstract

TAR DNA-binding protein 43 (TDP-43) is an RNA-regulating protein that carries out many cellular functions through liquid-liquid phase separation (LLPS). The LLPS of TDP-43 is mediated by its C-terminal low-complexity domain (TDP43-LCD) corresponding to the region 267–414. In neurodegenerative disorders amyotrophic lateral sclerosis and frontotemporal dementia, pathological inclusions of the TDP-43 are found that are rich in the C-terminal fragments of ∼25 and ∼35 kDa, of which TDP43-LCD is a part. Thus, understanding the assembly process of TDP43-LCD is essential, given its involvement in the formation of both functional liquid-like assemblies and solid- or gel-like pathological aggregates. Here, we show that the solution pH and salt modulate TDP43-LCD LLPS. A gradual reduction in the pH below its isoelectric point of 9.8 results in a monotonic decrease of TDP43-LCD LLPS due to charge-charge repulsion between monomers, while at pH 6 and below no LLPS was observed. The addition of heparin to TDP43-LCD solution at pH 6, at a 1:2 heparin-to-TDP43-LCD molar ratio, promotes TDP43-LCD LLPS, while at higher concentration, it disrupts LLPS through a reentrant phase transition. Upon incubation at pH 6, TDP43-LCD undergoes gelation without phase separation. However, in the reentrant regime in the presence of a high heparin concentration, it forms thick amyloid aggregates that are significantly more SDS resistant than the gel. The results indicate that the material nature of the TDP43-LCD assembly products can be modulated by heparin which is significant in the context of liquid-to-solid phase transition observed in TDP-43 proteinopathies. Our findings are also crucial in relation to similar transitions that could occur due to alteration in the molecular level interactions among various multivalent biomolecules involving other LCDs and RNAs.

Significance

In biological cells, dynamic membraneless compartments are formed by the condensation of multivalent biomolecules, such as RNA and proteins, through a process known as liquid-liquid phase separation (LLPS). Any misregulation in the dynamics of LLPS-induced compartments leads to the aberrant solidification of some of their molecular components and consequent accumulation of pathological aggregates. Here, we show that the assembly of TDP-43 low-complexity domain is modulated by heparin, a model polyanion, to generate fibrillar aggregates. The physicochemical properties of these aggregates are different from hydrogel formed by TDP-43 LCD alone. Our finding is relevant in the context of liquid-to-solid phase transition that is associated with macromolecular assemblies that form through LLPS.

Introduction

Biomolecular phase separation inside a biological cell is the organizing principle behind the formation of various membraneless organelles, including nucleoli, P granules, Cajal bodies, and stress granules (1,2). The prominent constituents of such membraneless organelles are multivalent biomolecules, such as proteins with low-complexity domains, and nucleic acids. The constituent molecules, through their numerous weakly interacting motifs, form temporary assemblies by a process known as liquid-liquid phase separation (LLPS) (3). Although membraneless organelles that form through LLPS possess liquid-like properties, their composition is not isotropic (4, 5, 6, 7). They harbor different subcompartments that are mutually immiscible, and their material properties lie in the continuum between liquid and solid. In other words, there exist multiple phases among membraneless organelles. Also, the aberrant solidification of many such constituent proteins in membraneless organelles gives rise to toxic aggregates, which are implicated in neurodegeneration (8,9). It is thus important to understand how LLPS and pathological solidification are related. Given that a multitude of interactions among low-complexity domains and nucleic acids give rise to LLPS, the molecular basis of the liquid-to-solid transformation of such assemblies is not clear. Therefore, minimalistic in vitro systems can help understand the underlying molecular level interactions that form the basis of diversity in the material nature of the phase-separated assemblies. In this study, we have utilized the low-complexity domain of TDP-43 protein (TDP43-LCD), corresponding to amino acid region 267–414, to understand its phase behavior at different pH values and in the presence of heparin. Heparin is a glycosaminoglycan that has been used as a good model system to mimic the effect of different classes of polyanions on protein aggregation (10, 11, 12). Given that cellular environment is rich in polyanions like polynucleotides, such as DNAs and RNAs, we used heparin as a model polyanion to understand its effect on TDP43-LCD assembly.

TDP-43 is a multifunctional protein (Fig. 1 A) and consists of a structured N-terminal domain (13), two RNA recognition motifs (14), and a disordered low-complexity domain (TDP43-LCD) (15,16). The cellular function of TDP-43 is to mediate RNA processing and the formation of various membraneless organelles (biomolecular condensates) through LLPS (15,17,18). The TDP43-LCD is rich in several polar and aromatic amino acid residues that form numerous weak interactions with other proteins and nucleic acids. In other words, macromolecules participating in the formation of liquid-like membraneless organelles are multivalent and harbor several weakly interacting motifs (3). It is believed that, due to factors not yet known, these temporary assemblies can also serve as a reservoir for pathological aggregation of TDP-43 (19, 20, 21, 22, 23). However, this notion is debatable, and it is still unclear whether aggregation of TDP-43 occurs due to the aberrant stabilization of membraneless assemblies (formed through LLPS) or whether aggregation and phase separation are two independent processes. It seems unlikely that underlying weak interactions that confer the fluidity to these membraneless organelles are enough to form a stable amyloid assembly. Given that TDP-43 participates in ribonucleoproteins formation, a membraneless assembly of RNA and proteins, its aggregation could also be initiated by other constituent macromolecules.

Figure 1.

TDP43-LCD purification and initial characterization. (A) Domain architecture of TDP-43. The low-complexity domain (TDP43-LCD) used in this study is highlighted in the dashed box. (B) Recovery of TDP43-LCD after removal of urea. TDP43-LCD was purified by solubilizing the inclusion bodies in 8 M urea, and subsequent urea removal through dialysis against different pH buffers. The data represent mean ± SE for n = 3 independent experiments. (C) Spectral signatures of 20 μM purified TDP43-LCD in 25 mM MES buffer (pH 6), along with 8 M urea-denatured control. The CD data for 8 M urea control could only be obtained until 210 nm, due to high spectral noise below 210 nm. (D) LLPS of purified TDP43-LCD was initiated by diluting concentrated proteins in LLPS buffer (25 mM Tris-HCl + 100 mM NaCl [pH 7.4]), to a final concentration of 20 μM. Aggregates and droplets are observed in the tube that was prepared by dilution from 300 μM stock. To see this figure in color, go online.

TDP43-LCD serves as a good model system for LLPS studies because, 1) it can undergo LLPS in vitro, 2) the pathological inclusions found in ALS/FTLD are rich in mainly truncated fragments of 25 and 35 kDa C-terminal regions of TDP-43, which also includes the TDP43-LCD region (24,25), and 3) nearly 95% of pathological mutations are reported in this region (26). In earlier reports, various short peptides of TDP-43 were studied to understand its aggregation mechanism (27, 28, 29, 30). These short peptide fragments, however, lack LLPS ability and, therefore, reports attempting to study the aggregation of TDP43-LCD in relation to LLPS are scarce. Hence, TDP43-LCD serves as minimalistic aggregation-prone fragment that has the ability to undergo LLPS.

Given the dynamic nature of TDP43-LCD, it is essential to ensure the quality during its purification since a slight leftover impurity during its purification and storage could alter its assembly properties. Previously reported works have been mostly carried out on TDP43-LCD diluted from 8 M urea containing storage buffer, which was purified through acetonitrile-based reverse-phase chromatography (17,31,32). A residual acetonitrile or urea concentration in TDP43-LCD reaction buffer could affect its solution state behavior. We designed a different purification protocol for the recombinantly expressed protein from E. coli to ensure that no residual impurities are present after the purification. We first studied the LLPS propensity of TDP43-LCD at different pH conditions and in the presence of salt. It was observed that initial conformation and assembly state of TDP43-LCD is highly sensitive to the solution conditions (pH, salt, and other cosolutes). Based on this observation, we designed appropriate buffer conditions to study its assembly process in the dispersed state, with and without heparin, a polyanion. We observed a different aggregation pattern in the absence and presence of heparin. In the absence of heparin, TDP43-LCD forms hydrogel through a “gelation without phase separation” process, which consists of a 3D supramolecular network of thin fibers, whereas in its presence it formed thick fibrils that were more SDS stable than the hydrogel. This observation hints that the assembly pathway and final assembly product of TDP43-LCD is highly dependent on its physicochemical environment. The findings have implications in aberrant-phase transition that has been observed in many neurodegenerative diseases, such as ALS, FTD, Alzheimer’s disease, etc.

Materials and methods

Reagents

All chemicals and reagents were purchased from Sigma-Aldrich (St. Louis, MO, USA) unless otherwise stated. Bacterial culture reagents such as LB medium and kanamycin sulfate were purchased from HiMedia (Bangalore). Buffer stocks (250 mM) of different pH ranges were prepared as follows: sodium acetate/acetic acid for pH 4 and 5; MES for pH 6; Tris-HCl for pH 7 and 8; glycine-NaOH for pH 9 and 10; sodium bicarbonate-NaOH for pH 11; and potassium chloride-NaOH for pH 12. The thioflavin-T (ThT) stock was prepared by dissolving the dry powder in doubled distilled water, filtering it through a 0.22-μm syringe filter, and measuring its concentration at 420 nm using a molar extinction coefficient of 24,240 M⁻¹ cm⁻¹. The ANS stock was prepared similarly, and its concentration was determined at 350 nm using a molar extinction coefficient of 5000 M⁻¹ cm⁻¹.

Recombinant TDP43-LCD production and fluorescence tagging

TDP43-LCD plasmid construct was a kind gift from Prof. Nicolas Fawzi (Brown University). The gene sequence encoding TDP-43 region 267–414 was cloned in NdeI and XhoI restriction sites of the pj414 vector backbone (DNA2.0). The plasmid was transformed into BL21(DE3) E. coli expression strain, and the colony was transferred in an LB medium containing 50 μg mL⁻¹ kanamycin. The culture medium was grown at 37°C at 200 rpm until OD₆₀₀ reached 0.6. Subsequently, the culture was induced with 1 mM IPTG and grown further for 8 h at 37°C and 200 rpm. In the subsequent step, the culture was harvested by centrifugation at 5000 × g for 20 min and stored at −20°C until further use. Before purification, the culture pellet was thawed at 4°C and the pellet was resuspended in the lysis buffer containing 25 mM Tris-HCl (pH 8.0), 100 mM NaCl, 1× concentration of protease inhibitor, 10 units of DNase and RNase (Thermo Scientific, Waltham, MA, USA), and hen egg white lysozyme to a final concentration of 0.5 mg mL⁻¹. Around 20 mL lysis buffer was added per gram of the culture pellet. The resuspended pellet was incubated on ice for 30 min and then subjected to sonication for 20 min at on and off cycles of 15 s each at 35% amplitude and 4°C temperature. The lysate was centrifuged at 10,000 × g for 60 min at 4°C using a tabletop centrifuge (Thermo Scientific). The pellet fraction containing TDP43-LCD inclusion bodies was saved and supernatant was discarded.

The pellet was resuspended in inclusion body solubilization buffer (20 mL per gram of pellet), containing 20 mM Tris-HCl (pH 8.0), 100 mM NaCl, 2 M urea, and 1% Triton X-100, maintained at 25°C. The resuspended pellet was centrifuged at 10,000 × g for 30 min at 25°C. The supernatant was discarded and the resuspension step was repeated. Subsequently, the pellet was resuspended in 20 mM Tris-HCl buffer and 100 mM NaCl to wash residual urea and Triton X-100. At this stage, the washed inclusion bodies were either directly proceeded for purification or stored at −20°C until further use. The inclusion bodies at this stage were ∼90% pure.

The purified inclusion bodies were resuspended in affinity buffer A (20 mL per gram of pellet), containing 20 mM Tris-HCl (pH 8.0), 50 mM NaCl, 10 mM imidazole, and 8 M urea. The resuspended mixture was incubated at room temperature overnight with constant stirring. The dissolved inclusion bodies mixture was centrifuged at 10,000 × g for 60 min to remove any insoluble debris. The solubilized TDP43-LCD was then incubated with Ni-NTA agarose resin (QIAGEN, Hilden, Germany), which was pre-equilibrated with affinity buffer A. The Ni-NTA resin and protein were incubated for 30 min on an end-to-end rotatory shaker. The mixture was poured into a polypropylene gravity flow column (QIAGEN, Hilden, Germany), and flow-through was collected. The Ni-NTA resin was washed with 2 column volumes of wash buffer (buffer A with 20 mM imidazole). The protein was eluted with elution buffer (buffer A with 100 mM imidazole). All fractions collected were analyzed through SDS-PAGE using a 15% resolving gel, and visualized through Coomassie brilliant blue R-250 staining. The fractions containing purified TDP43-LCD were pooled, filled in a dialysis tube of 10 kDa cutoff, and subjected to dialysis against 50 mM sodium acetate buffer (pH 5). The dialysis against pH 6 and 7 buffer resulted in precipitation, hence subsequent dialysis was carried out at pH 5. After overnight dialysis, subsequent dialysis was carried out against 2.5 mM pH 5 buffer. The protein was filtered through a 0.22-μm filter, and the concentration was measured at 280 nm using Beer-Lambert’s method using a molar extinction coefficient of 17,990 M⁻¹ cm⁻¹. The protein was concentrated using an Amicon centrifugal membrane concentrator (Millipore, Burlington, MA, USA) and stored at −20°C.

Atto-488-tagged protein was prepared by incubating Atto-488 NHS ester dye (Sigma-Aldrich) with 100 μM protein at room temperature for 1 h at constant end-to-end mixing. Subsequently, the sample was placed in a dialysis bag of 10 kDa cutoff size and dialyzed overnight against 2 mM sodium acetate buffer (pH 5) to remove unreacted dye. The labeling efficiency was checked using the manufacturer’s protocol. The dye-conjugated protein was aliquoted and stored at −20°C.

Spectroscopy

The turbidity of the protein sample was measured at 600 nm on a Cary 4000 UV-visible spectrometer (Agilent, Santa Clara, CA) equipped with a thermostated water bath. Fluorescence experiments were carried out on Cary Eclipse Spectrofluorimeter (Varian, Palo Alto, CA). Intrinsic tryptophan fluorescence was recorded over a wavelength range of 310–550 nm, at 295 nm excitation wavelength. The ThT was added to the reaction mixture to a final concentration of 20 μM, incubated for 2 min, and fluorescence spectra were collected over a wavelength range of 470–550 nm, with an excitation wavelength of 450 nm. The excitation and emission slits were 5 nm each. The surface hydrophobicity was monitored by adding ANS to a final concentration of 40 μM, and fluorescence scan was monitored from 400 to 600 nm, with 350 nm excitation. For CD spectroscopy, the sample was taken in a quartz cuvette of 0.1 cm pathlength (Hellma, Müllheim, Germany). The sample were analyzed using a J-815 CD spectropolarimeter (Jasco, Tokyo, Japan) equipped with a peltier temperature controller under a continuous flow of nitrogen. The wavelength range was 195–250 nm, the scan rate was 50 nm min⁻¹, and the data pitch was 0.5 nm. The data were an average of three scans.

Fluorescence and differential interference contrast microscopy

All fluorescence microscopy images were acquired at a 1:50 molar ratio of labeled and unlabeled TDP43-LCD. For the gel and fibrils, preformed assemblies were mixed with the labeled TDP43-LCD just before the imaging. The images were acquired by placing the demixed sample on a glass slide, and samples were immediately covered with a coverslip. The boundary of the coverslip was sealed with transparent nail paint so that no evaporation could take place. The images were acquired on a Leica SP8 confocal microscope under a 63× oil immersion objective, using an Atto-488 filter (excitation 500 nm; emission 520 nm).

Fluorescence recovery after photobleaching (FRAP) was performed using a Leica SP8 confocal microscope with 2.4-mW laser intensity for bleaching, 63×/1.4 oil immersion objective, and a photomultiplier tube detector. For each assembly, the FRAP analysis was carried out on three independent samples, and recovery curves were averaged. Five prebleach frames were selected, which was followed by 10 bleach frames, and the recovery was monitored for over 50 s. Constant regions of interest (ROIs) were photobleached on droplets/aggregates, and fluorescence recovery for the ROIs was normalized considering the fluorescence intensity of the prebleach region as 1.

The average fluorescence intensity of each ROI was collected over time and corrected for bleaching using the equation:

$${\text{I}}_{\text{cor}}\left(\text{t}\right)=\left[\text{I}\left(\text{t}\right)-\text{B}\right]/\left[{\text{I}}_{\text{ref}}\left(\text{t}\right)-\text{B}\right],$$

where B is the background, I_ref (t) is the intensity of unbleached droplet at any given time, and I (t) is the intensity of bleached ROI.

The corrected intensity was then normalized as follows:

$${\text{I}}_{\text{norm}}\left(\text{t}\right)=\left[{\text{I}}_{\text{cor}}\left(\text{t}\right)-{\text{I}}_{\text{cor}}\left(\text{bleached}\right)\right]/\left[{\text{I}}_{\text{cor}}\left(\text{prebleached}\right)-{\text{I}}_{\text{cor}}\left(\text{bleached}\right)\right].$$

The relative intensity was fit to a single exponential curve as follows:

$${\text{I}}_{\text{norm}}\left(\text{t}\right)={\text{I}}_{\text{recovered}}(1-{\text{e}}^{-\text{t}/\tau }),$$

where τ is the apparent recovery time and I_recovered is the maximum recovery.

The half time of recovery (t_1/2) was calculated using the equation:

$${\text{t}}_{1/2}=\tau \ln \left(2\right).$$

The percent mobile fraction was calculated using following equation:

$$\%{\text{M}}_{\text{f}}=\left[{\text{I}}_{\text{recovered}}-{\text{I}}_{\text{cor}}\left(\text{bleached}\right)\right]/\left[{\text{I}}_{\text{cor}}\left(\text{prebleached}\right)-{\text{I}}_{\text{cor}}\left(\text{bleached}\right)\right].$$

To visualize protein droplets through differential interference contrast (DIC) microscopy, 10 μL phase-separated protein solution was aliquoted on a clean glass slide and was covered using a glass coverslip. The slide was analyzed using a DIC microscope (Nikon), and images were acquired.

Attenuated total reflectance FTIR

The aggregated protein solutions were centrifuged for 30 min at 12,000 × g using an Eppendorf tabletop centrifuge. The supernatant was discarded, and pelleted fractions (aggregate/hydrogel) were resuspended in D₂O and centrifuged again to remove residual H₂O. The pellet was taken out from the tube using a clean spatula and deposited on a diamond crystal mounted on an attenuated total reflectance (ATR) accessory of a Bruker Tensor27 FTIR instrument. An average of 64 scans were taken in percent transmittance mode. The percent transmittance data were converted into absorbance, and a second derivative operation was applied to the absorbance data to deconvolute the spectra and identify the peaks.

Atomic force microscopy

From the tubes containing aggregated samples, a 10 μL sample was deposited on a freshly cleaved mica sheet, and incubated for 20 min at room temperature. Subsequently, the mica sheet was gently washed with 1 mL double-distilled water and kept for a minimum of 2 h to allow the evaporation of residual water. The sheet was then analyzed in tapping mode using a cantilever mounted on an atomic force microscopy (AFM) instrument (WITec, Ulm, Germany). Each image was flattened and analyzed using Project FOUR software.

Results

Preparation of TDP43-LCD

In this study, TDP43-LCD was purified from inclusion bodies solubilized in 8 M urea (Fig. S1). During initial optimization, after purification, the urea was removed by dialysis against 25 mM Tris-HCl buffer (pH 7.4). We observed major precipitation of TDP43-LCD during dialysis, with only 20% remaining in a soluble state (Fig. 1 B). In the following experiment, urea was removed by dialysis against pH 6 and 5 buffers, and nearly 70 and 90% recovery was observed (Fig. 1 B). TDP43-LCD in our study is hexahistidine tagged, and its isoelectric point (pI) is 9.8. We optimized dialysis conditions by setting the pH of the dialysis buffer far below its pI, as a result of which electrostatic repulsions at lower pH reduced TDP43-LCD aggregation. We further examined the effect of buffer strength on the aggregation of TDP43-LCD by dialyzing it against 25 and 50 mM sodium acetate (pH 5) and found that recovery in 50 mM buffer was nearly 5% more than in 25 mM buffer (data not shown) suggesting that buffer strength also plays a role in the protein recovery. After initial dialysis in 50 mM pH 5 buffer, the final round of dialysis was carried out in a very low buffer strength (2 mM), because final dialysis against water resulted in slight precipitation of the protein. Thereafter, the protein was concentrated to 100 μM, nearly five times the working concentration and stored at −20°C. The TDP43-LCD used in our study was hexahistidine tagged and we did not remove the tag as TDP43-LCD readily precipitates at the pH required to cleave the tag through thrombin, which results in the reduction of its yield considerably. Moreover, a recent study demonstrates that the presence of hexahistidine tag did not affect the LLPS of TDP43-LCD (33). Next, we carried out the spectroscopic analysis of purified TDP43-LCD (Fig. 1 C). The far-UV CD signature of TDP43-LCD displayed a single negative peak centered at 200 nm, indicating its disordered conformation. Intrinsic tryptophan emission spectrum showed λ_max at 357 nm, which is the signature of completely exposed tryptophan residues in water, further confirming the disordered state of TDP43-LCD. Compared with the urea-denatured control, the ANS fluorescence spectrum of purified TDP43-LCD showed a low-intensity peak with λ_max around 500 nm. Interestingly, the ANS fluorescence intensity of the purified TDP43-LCD was 65% more than that of the urea-denatured control. The higher ANS intensity may be due to the enhanced hydrophobic contacts that are reported to be augmented by the helix-helix dimerization of TDP43-LCD in the region 331–343 (34).

A recent report suggested that low-complexity domains can be stored at extreme pH at high concentrations (35). In our case, however, concentrating the protein beyond 300 μM resulted in its gelation. In such a case, direct dilution of TDP43-LCD in the phase separation buffer (25 mM Tris-HCl [pH 7.4] and 100 mM NaCl) resulted in its immediate aggregation visible in the test tube (Fig. 1 D). On the other hand, when 300 μM stock was first diluted to 100 μM in 2 mM pH 5 buffer and incubated for 15 min on ice, and then finally diluted in the phase separation buffer, we observed droplets without any trace of irregularly shaped aggregates (Fig. 1 D). The DLS examination of TDP43-LCD at different concentrations shows the prevalence of large size species at 300 μM, the dimension of which significantly reduced upon dilution to 100 μM, and subsequently to the working concentration of 20 μM (Fig. S2). This is probably the reason why dilution from a 300 μM stock of TDP43-LCD resulted in aggregation.

Thus, with the rational manipulation of the physicochemical conditions of the solution, we were able to purify TDP43-LCD with a simple protocol. This purification method obviates the need for HPLC purification and lyophilization, as discussed in the previous studies, and helps to get a protein preparation free from the residual impurities.

Optimizing LLPS conditions

It is reported that the LLPS propensities of low-complexity domains are highly dependent on temperature. Therefore, any change in the solution temperature during experiments might cause nonreproducibility (36,37). We studied the LLPS of TDP43-LCD at 23°C, and all its subsequent incubations were done at this temperature. It allowed us to reproduce our results as it did not entail shifting the samples from ice to room temperature during imaging and spectroscopic examinations, etc. We first determined the saturation concentration (C_sat) of TDP43-LCD at pH 7.4 in the presence of 100 mM NaCl. In a given solution condition, C_sat is the minimum concentration of a protein at which it undergoes LLPS. As measured through scattering at 600 nm, TDP43-LCD exhibited LLPS at 6.6 μM concentration in the presence of 100 mM NaCl (Fig. 2 A).

Figure 2.

LLPS of TDP43-LCD. (A) LLPS propensity of TDP43-LCD at different monomer concentrations was quantified by measuring turbidity at 600 nm. The buffer was 25 mM Tris-HCl with 100 mM NaCl. The C_sat of TDP43-LCD was calculated by observing intersection points of intercepts of mixed and demixed baselines (inset). (B) pH-Dependent LLPS of 20 μM TDP43-LCD. Different buffers were used at a final concentration of 25 mM to obtain the desired pH (indicated in the reagents section of materials and methods) with no NaCl added. (C) The left picture indicates that TDP43-LCD LLPS converts a clear solution into a turbid solution. (D) Fluorescence and corresponding DIC images of Atto-488-tagged TDP43-LCD droplets that formed at pH 7.4. (E) Fluorescence images of droplet coalescence indicates the liquid nature of TDP43-LCD droplets. Scale bar, 10 μm. The error bars represent mean ± SE for n = 3 independent experiments. To see this figure in color, go online.

Next, we fixed the protein concentration at 20 μM and checked its LLPS at different pH values at 23°C (Fig. 2 B). Near its pI (9.8), TDP43-LCD underwent LLPS, with maximum OD₆₀₀ observed at pH 8 and 9. The extent of LLPS decreased below pH 8 and diminished at pH 6 and below. Similar observations were made at pH 10 and above, with LLPS diminishing at pH 12. The demixed solution at pH 7.4 turned turbid and showed droplet-like structures when observed through fluorescence and DIC microscopy (Fig. 2, C and D). Some of the droplets show coalescence, which points toward their liquid nature (Figs. 2 E and S3).

LLPS propensity of TDP43-LCD in the presence of salt and heparin

We identified that, at pH 6, 20 μM TDP43-LCD did not form droplets due to electrostatic repulsion. We intended to screen the electrostatic repulsion using NaCl, a monovalent counterion, and heparin, a polyanion. We prepared different sample tubes containing 20 μM TDP43-LCD, and a measured amount of NaCl or heparin was added to each tube, followed by OD₆₀₀ measurement. With the addition of NaCl, we observed a gradual increase in LLPS due to charge screening, as evident from OD₆₀₀, until it became saturated at 200 mM (Fig. 3 A). With a gradual addition of heparin, a polyanion, we observed a window-like phase behavior, wherein, with the initial addition of heparin up to 10 μM, an increase in OD₆₀₀ was observed (Figs. 3 B and S4). With a further increase in heparin concentration, we observed a decrease in OD₆₀₀, which completely diminished at 20 μM. This process is known as reentrant phase transition, wherein a monotonic addition of a counterion (heparin in this case) results in an initial condensation, and subsequent decondensation. This transition is characterized by a lower and upper critical concentration of heparin. The heparin and TDP43-LCD systems exist as a single decondensed phase below lower critical and above upper critical concentration of heparin. This observation was consistent with previous reports wherein the reentrant phase transition of TDP43-LCD has been observed in the presence of many small molecules and heparin (33,38). Notably, heparin-induced droplets were smaller in size than RNA-induced droplets (Fig. S5). Heparin is a highly negatively charged multivalent polymer and it could be hypothesized that its initial addition resulted in charge neutralization of TDP43-LCD, and consequent demixing of the system. With the subsequent addition of heparin, charge inversion of the heparin + protein system occurs, which results in the electrostatic repulsion and consequent remixing of the overall system.

Figure 3.

Effect of monovalent and polyvalent anions on TDP43-LCD phase behavior. TDP43-LCD (20 μM) was incubated in 25 mM MES buffer (pH 6), and the effect of NaCl and heparin was tested at 23°C. (A) In the presence of an increasing concentration of NaCl, an increase in OD₆₀₀ was observed with a peak at 200 mM NaCl. Corresponding DIC images at 200 mM NaCl reveal droplets (lower panel). (B) With heparin, reentrant phase behavior was observed with an abundance of irregular structures (black arrows) and droplets in demixed regime (10 μM heparin). Scale bar, 10 μm. The error bars represent mean ± SE for n = 3 independent experiments.

Interestingly, we noted that, in the demixed regime of the heparin-protein mix, apart from regular turbidity, which is being observed due to droplet scattering, we also observed clump-like structures through visual examination. The microscopic examination of this solution revealed the presence of irregularly aggregated structures coexisting with the regular droplets. In multicomponent systems, such as in this case, the homotypic interactions (protein-protein) compete with heterotypic interactions (protein-heparin). At low heparin concentration, probably, some of the TDP43-LCD droplets further coalesce into irregular aggregates, whereas, in the case of higher heparin concentration, due to high negative charge density, droplet formation and hence coalition is not observed. It must also be noted that the OD₆₀₀ values of heparin-incubated samples is the sum of scattering from both droplets and aggregates, which is the reason why there is a difference in the apparent droplet number and size among the 200 mM NaCl-containing sample and the 10 μM heparin-incubated sample. The heparin-mediated transition occurred in a very narrow heparin concentration regime, which was substoichiometric to the protein concentration employed (20 μM). A Similar transition was observed for Arg-rich motifs, where they undergo demixing and remixing transition in the presence of substoichiometric RNA concentration (39).

Heparin modulates the assembly of TDP43-LCD

We obtained two different conditions under which TDP43-LCD did not undergo LLPS. At pH 6, due to electrostatic repulsion, and in the presence of 20 μM heparin, the solution mixture displays reentrant transition. We observed no visual turbidity under both conditions even after 3 days of incubation. After 3 days, we examined TDP43-LCD spectroscopically. The far-UV CD spectra of TDP43-LCD incubated under both conditions showed minima centered around 218 nm, indicating the prevalence of β-sheet structures (Fig. 4 A). This was significantly different from the control sample (TDP43-LCD at 0 time point at pH 6), which showed a negative peak at 200 nm, a characteristic of disordered polypeptides. The corresponding intrinsic tryptophan fluorescence spectra of incubated proteins showed a reduction in the peak intensity compared with that of control, along with a considerable blue shift, with λ_max shifting from 357 nm for the control to 348 nm for the incubated samples (Fig. 4 B). This blue shift transition indicated the burial of tryptophan residues in the apolar environment, either due to significant structure formation in the TDP43-LCD chain, or intermolecular association.

Figure 4.

Effect of monovalent and polyvalent anions on TDP43-LCD assembly. Protein (20 μM) was incubated for 3 days at 23°C in 25 mM MES buffer (pH 6) with or without 20 μM heparin. Control is freshly reconstituted 20 μM TDP43-LCD at pH 6. (A) Far-UV CD spectra of pH 6 incubated TDP43-LCD with or without 20 μM heparin. (B) Intrinsic tryptophan fluorescence spectra show a blue shift with incubation from 357 to 348 nm. (C) ThT fluorescence spectra. (D) AFM imaging was carried out by spreading the sample on a mica sheet, and a 10 × 10 μm area was scanned. The pH 6 (without heparin) sample shows a network-like morphology indicative of amyloid hydrogel formation. For heparin-incubated samples, thick fibrils with lateral stacking at some points are discernible. Line profiling of a cross section of AFM data (white line) using project V software shows a difference in the fibril thickness. (E) Proteinase K digestion of aggregates for 30 min reveals different degradation profiles. Detailed optimization is included in Fig. S6, and the original uncropped gel is shown in Fig. S7. (F) ATR-FTIR spectra of incubated samples. The red trace is the pH 6-incubated sample (without heparin), the blue trace is with heparin, and the control is shown in black. For FTIR, the control sample was prepared by allowing droplets that formed at pH 7.4 to aggregate and settle for 30 min, and centrifuging the aggregates. To see this figure in color, go online.

To explore the possibility of intermolecular interaction, ThT fluorescence spectra of incubated samples were recorded (Fig. 4 C). Unlike CD and fluorescence profiles, the ThT fluorescence intensity of the samples incubated in both conditions varied significantly. The ThT fluorescence intensity of the heparin-incubated sample was four times higher than that of the pH 6-incubated sample. This observation hints that, despite the presence of cross-β-structure in both the incubated samples, there might exist a difference in the amyloid quaternary fold. We performed a nanoscale examination of both the assembly products using AFM and found that, at pH 6, TDP43-LCD assembled into a system-spanning fibrillar structure with dense network-like morphology, and the heparin-incubated sample showed fibrillar structure stacking laterally at certain points of contact (Fig. 4 D). The fibrils also differed in height, with an average fibril thickness of 2.5 and 8 nm in the case of pH 6- and heparin-incubated samples, respectively. To assess whether aggregates formed under the two conditions share a similar structural core, we carried out proteinase K digestion of the mature aggregates (Figs. 4 E and S6). The time-dependent digestion profile of both the aggregates revealed different bands on SDS-PAGE, suggesting a dissimilar proteinase K-resistant core. To examine whether both the aggregates share similar structural folds, we carried out an ATR-FTIR analysis of the protein assemblies and performed second-derivative spectral deconvolution to find the peaks (Fig. 4 F). Amyloid aggregates formed under both the conditions showed two peaks centered at 1627 and 1654 cm⁻¹. The 1627 cm⁻¹ peak corresponds to a cross-β fold, a signature peak for amyloid folds (40,41). The 1654 cm⁻¹ peak corresponds to the α-helical fold, which has been found in many amyloid assemblies and intermediates (42,43). As a control, we also examined phase-separated droplets, which form at pH 7.4, that were left to mature for 30 min at room temperature. After maturation, most droplets coalesced into irregularly shaped clumps and settled down in the tube. We centrifuged these clumps at 13,000 × g for 30 min, washed them with D₂O, and analyzed them through ATR-FTIR. The control samples prepared in such a manner displayed a single peak centered at 1640 cm⁻¹, indicative of a disordered motif with slight β-sheet content. From these observations, we can conclude that TDP43-LCD can form a fibrillar assembly even without an intermediate LLPS stage that could be modulated by heparin.

TDP43-LCD forms amyloid hydrogel and solid amyloid fibrils under different conditions

The appearance of a system-spanning 3D cross-linked structure by pH 6-incubated TDP43-LCD (Fig. 4 D) prompted us to investigate whether it forms a hydrogel-like structure. In the previous studies, various low-complexity domains have been reported to form hydrogels by virtue of loose intermolecular associations (29,44). A closer examination of AFM images recorded at a larger area (75 × 75 μm) revealed the presence of many thick and dense clusters of samples scattered at different positions on the mica sheet (Fig. 5 A). In the case of heparin also, we found clusters of aggregates, indicating lateral stacking of fibrils at some points (Fig. 5 B). To decipher whether heparin had any effect on the initial conformation of TDP43-LCD, we added 20 μM heparin to the equimolar concentration of protein. The 1:1 molar ratio of the heparin-protein mixture remained in dispersed phase due to the reentrant transition (Fig. 3 B). The heparin addition resulted in a blue shift of the tryptophan fluorescence spectra, with λ_max shifting from 357 to 353 nm. We propose that heparin either induces some structural change in TDP43-LCD or acts as a nucleus that recruits protein monomer on its surface.

Figure 5.

Material state characterization of aggregates formed by TDP43-LCD at pH 6, with and without 20 μM heparin. (A) Protein sample that was allowed to aggregate for 3 days was applied on a mica sheet, and a larger area 75 × 75 μm was selected to obtain a broader view of the aggregate morphology. We found two different aggregation patterns on a mica sheet on which the pH 6-incubated sample was applied, with an evenly spread sample (region 1) and uneven clumps (region 2). (B) AFM image of the heparin-incubated sample shows discreet clump-like aggregates. The lower panel indicates intrinsic tryptophan fluorescence spectra of freshly incubated TDP43-LCD in 25 mM MES (pH 6), with or without 20 μM heparin. Immediately after heparin addition, a blue shift is observed. (C) To check the flow of hydrogel under gravity, tubes were inverted, and the flow of the sample was monitored. In parallel, DIC microscopy of the samples was done. The stability of aggregates was measured by incubating the samples with different concentrations of SDS. The error bars represent the mean ± SE of n = 5 independent experiments. (D) FRAP analysis of TDP43-LCD assembly products. The left panel is the confocal imaging of the Atto-488-tagged TDP43-LCD assembly products, and the right panel corresponds to recovery kinetics after photobleaching a region of interest (ROI). Scale bar, 1 μm. The recovery was highest for the liquid phase, followed by the gel phase, and recovery was not observed in the case of solid assembly. The error bars represent the standard errors of the mean for three independent measurements. To see this figure in color, go online.

To test whether TDP43-LCD formed a hydrogel, we concentrated TDP43-LCD to 300 μM and incubated it at room temperature. Within 12 h, the protein polymerized into a gel-like structure, which did not flow under the influence of gravity when the tube was inverted (Fig. 5 C). On the other hand, the heparin-mediated aggregate solution flowed down under the influence of gravity. To further analyze the hydrogel morphology of the pH 6-aggregated protein, we vortexed the hydrogel vigorously. The fragmented particles were placed on a microscopic slide and analyzed under a DIC microscope. We observed transparent gel-like fragments in the case of the pH 6-incubated protein, compared with heparin-mediated aggregates, which displayed some opaque structures on a clear background. We further assessed the SDS stability of the hydrogel by first diluting it to 20 μM in the presence of a different concentration of SDS and recording its ThT fluorescence. The ThT fluorescence of the hydrogel reduced to a basal level in 0.2% SDS, whereas heparin-mediated aggregates were resistant to SDS-mediated dissociation up to 1% SDS concentration. Previous reports have highlighted the SDS and heat sensitivities of amyloid hydrogels formed by LCDs or their fragments (29,45). Next, we performed FRAP analysis of three types of assembly products (liquid, gel, and solid) and measured their recovery kinetics after photobleaching (Fig. 5 D and Table 1). The recovery was highest in the case of the liquid phase (91%), followed by the gel phase, for which recovery was nearly 82%. The recovery half time for the gel phase was nearly double that of liquid assemblies. For solid aggregates, however, we did not observe any recovery. The limitations of comparing the FRAP recovery kinetics of the gel and solid assemblies with liquid-like assemblies are their nonuniform 3D structures. Although the chosen ROI in a 2D plane was similar for droplets, gels, and solids, the nonuniform thickness of their gels and solid assemblies complicates the analysis. With these findings, however, it is evident that, in the presence of heparin, TDP43-LCD formed amyloid aggregates, the physical properties of which are different from the spontaneous amyloid-like hydrogels that formed at pH 6.

Table 1.

Percent mobile fraction (%M_f) and recovery half time (t_1/2) of droplets and gel from the FRAP data

Sample	%Mf	t1/2 (s)
Droplets	91 ± 5.6	8.5 ± 1.1
Gel	82 ± 6.1	16.8 ± 1.7
Solid	ND	ND

The solid assemblies did not show any recovery after photobleaching, hence the %M_f and t_1/2 values remained non-determinable (ND). The data represent mean ± SE for n = 3 droplets/gel samples.

TDP43-LCD hydrogel formation is kinetically controlled

We found that, while decreasing the solution pH of TDP43-LCD to 6 or below results in the abrogation of its LLPS by electrostatic repulsion (Fig. 2 B), it forms supramolecular 3D hydrogel assembly at pH 6 (Fig. 5). It is thus evident that, when LLPS is discouraged and the kinetics of association of TDP43-LCD is slowed down considerably, its assembly behavior changes. To investigate whether the process of hydrogel formation was kinetically controlled, we incubated the protein in lower pH buffers, i.e., pH 4 and 5. The tryptophan fluorescence maximum of freshly reconstituted protein is 357 nm, which, during aggregation, gradually shifts to 348 nm (Fig. 4 B). We, therefore, calculated the time-dependent values of λ_max of the samples as a measure of its aggregation. We found that the onset and saturation of the aggregation of TDP43-LCD was slowest for the pH 4-incubated sample and fastest for the pH 6-incubated sample (Fig. 6 A).

Figure 6.

pH-Dependent hydrogelation kinetics of TDP43-LCD. Protein (20 μM) was incubated at varying pH ranges (A) and pH 4 and 5 at varying salt concentrations (B). The time-dependent wavelength maximum of intrinsic tryptophan fluorescence was recorded as a measure of hydrogel formation. The solid line is for eye guidance only. To see this figure in color, go online.

The addition of NaCl screens the net positive charge on TDP43-LCD and, in the next experiment, at pH 4 and 5, a measured amount of NaCl was added to the solution. The gradual addition of NaCl at lower pH screens the net positive charge on TDP43-LCD, which eventually induces the droplet formation through LLPS (31,34). The maximum concentration of NaCl at pH 4 and 5 was maintained below the critical concentration at which TDP43-LCD undergoes LLPS. Both at pH 4 and 5, the addition of NaCl resulted in faster kinetics of aggregation, as discernible from the onset and saturation of the λ_max shift (Fig. 6 B). We did not consider the pH 6-incubated samples because they undergo LLPS even at a low concentration of NaCl. For the same reason, we did not monitor kinetics at 100 mM NaCl at pH 5. Altogether, these experiments suggest that hydrogel formation by the TDP43-LCD is a kinetically controlled process, which is modulated by the solution conditions dictated by the net charge on the protein.

TDP43-LCD forms a hydrogel via submicroscopic solid oligomeric intermediates

TDP43-LCD in this study readily formed liquid droplets even at 10 μM protein concentration at physiological pH, which is noticeable through visual (turbidity) and microscopic examination (Figs. 1, 2, and 3). However, at pH 6 and below, we did not observe demixing even at protein concentrations as high as 350 μM. The formation of TDP43-LCD hydrogel in these conditions raises the question of whether LLPS is an obligatory intermediate in this case or whether hydrogelation proceeds through a different pathway. To answer this question, a 290 μM (5 mg mL⁻¹) stock of TDP43-LCD was incubated at 23°C at pH 6, and samples were subjected to both microscopic and nanoscopic examination through DIC microscopy and AFM, respectively, at three distinct stages of hydrogelation (Fig. 7). We did not observe any droplets in DIC microscopy at 0 h (initial phase) and 10 h (intermediate phase). The corresponding AFM examination revealed the presence of nearly 10-nm oligomers and interspersed fiber-like structures (Fig. 7 B). The end-stage assembly product (24 h) was visible in the test tube as a thick hydrogel that did not flow under gravity and trapped an introduced air bubble. This observation hints at the possibility of distinct aggregation pathways for proteins with low-complexity domains that undergo LLPS even at low concentrations under physiological conditions. The end-stage product appeared as a transparent and translucent structure under DIC microscopy and a cross-linked fibrillar structure under AFM.

Figure 7.

Time-dependent examination of TDP43-LCD hydrogelation at (A) microscopic (DIC) and (B) nanoscopic (AFM) scales. TDP43-LCD was incubated at 23°C at pH 6 at 5 mg mL⁻¹ monomer concentration. Samples were aliquoted at three different time points and examined through a DIC microscope and AFM. Post 24 h, an air bubble was introduced through a micropipette in the hydrogel, which appeared trapped and provided a visual confirmation for hydrogelation. To see this figure in color, go online.

Discussion

This work demonstrates how a slight change in the solution condition can modulate the assembly landscape of TDP43-LCD and give rise to products of different material nature. Solution pH-induced net charge on TDP43-LCD monomers greatly influence its assembly process, which in turn can be modulated by salt (Fig. 2). In the cellular context, an analogous change in the net charge of a polypeptide can occur either due to posttranslational modifications or mutations, which alters its isoelectric point. Alternatively, any change in the constituent molecular components in the temporary cellular compartments, such as ribonucleoprotein granules, of which these phase-separated assemblies are part, can modify the physicochemical environment of a polypeptide and promote or discourage its assembly (33,46).

Membraneless organelles that form through LLPS are multicomponent, and are composed of many layers or subregions, whose material properties can vary from a continuum of states ranging from liquid to gel to solid (4,20,47). The cell can regulate the material states not only by controlling the protein expression level but also by regulating the production of other components, such as RNA, poly ADP ribose, and other multivalent components, e.g., proteins with low-complexity domains (48, 49, 50). The multivalent homopolymeric ligands, such as RNA, provide additional interaction motifs to the protein undergoing LLPS (51). The reentrant transition of TDP43-LCD in the presence of one such multivalent homopolymer heparin occurs in a very narrow range, which culminates in the generation of solid fibrils without an intermediate LLPS stage (Fig. 3 B). The reentrant phenomenon, which is observed in phase-separating systems, has been proposed to have a regulatory role, wherein the cell can tune the formation and dissolution of liquid condensates through the expression of multivalent ligand such as RNA (33,49).

The connection between LLPS and fibrillar aggregation is rather poorly established, with an assumption that the concentrated environment of protein droplets gives rise to aggregates. This hypothesis, however, has been brought under scrutiny in recent times because the interactions that give rise to LLPS are loose and are “slithering” in nature and, intuitively, least likely to give rise to a stable amyloid assembly (52,53). One assumption to support transition of droplets to amyloid could be that the initial collapsed phase-separated state may slowly allow amyloid-forming interactions, which are kinetically inaccessible otherwise. In fact, among a multitude of interacting motifs that a polypeptide harbors, all of them may not be available for interaction at a particular solution condition. Several proteins have been shown to undergo LLPS at both low and high salt concentrations, with electrostatic interactions driving LLPS at low salt regime and hydrophobic and nonionic interactions causing LLPS at high salt concentration (54). Thus, owing to competing sets of interactions among different motifs in a macromolecule, even homotypic LLPS can be modulated by solution environment. In a multicomponent system, large sets of both homotypic as well as heterotypic competing interactions exist, which complicate the phase behavior of the system. We observed a different assembly pattern for TDP43-LCD in conditions discouraging LLPS (pH 6 or below), both in the presence and absence of heparin.

As widely posited for the proteins undergoing LLPS, the soluble dispersed protein gives rise to liquid droplets, which might eventually mature into gel or solid state (21,22,55,56). In our case, however, at pH 6 or below, TDP43-LCD undergoes liquid to gel transition without an accompanying LLPS stage (gelation without phase separation). In fact, at pH 7 and above, where TDP43-LCD undergoes LLPS, droplets rapidly matured into irregularly shaped clumps that settled down in the tube (Fig. S8). Many LCDs containing proteins have been proposed to undergo gelation through an intermediate LLPS state and, to date, there is no report on gelation without phase separation behavior for LCDs (53). From the “stickers” and “spacers” framework of the associative polymers, gelation without phase separation is theoretically possible (57). Stickers are the residues that take part in the biomolecular interactions, and spacers or linkers are the residues that separate stickers and do not directly take part in the interaction (58). Linkers, however, modulate the overall assembly process of such proteins by virtue of their length and tendency to interact with the solvent. It is proposed that linkers of longer length provide flexibility to the sticker residues to form a system-spanning network (57). The percolation threshold, the concentration at which sol-gel transition takes place, for such proteins is below their C_sat, (concentration threshold for LLPS). In this condition, the protein undergoes gelation without LLPS. In this study, it could be possible that, at pH 6 or below, the electrostatic repulsions reduced the overall number of interactions, thereby reducing the effective valency of TDP43-LCD. The C_sat of proteins with low valency is higher and, hence, more number of molecules are required to induce LLPS. In this study, however, the percolation threshold of TDP43-LCD at pH 6 was lower than its C_sat at pH 6, as a result of which we never observed LLPS even at 500 μM concentration (data not shown). Recent reports suggest that the phase separation of LCDs may be hierarchical and that nanoscale clusters of protein might eventually give rise to mesoscale droplets (59, 60, 61, 62). For pH 6-incubated TDP43-LCD, the initial appearance of solid oligomers of the nanoscale size that finally give rise to gel affirm that these oligomers cannot be the precursor for the phase separation that occurs at pH 7.4 and, therefore, pathways for gelation and phase separation are different for TDP43-LCD (Fig. 7).

Multivalent charged polymers, such as RNA and DNA, are considered as “superscaffolds” as, due to their high charge density, they can recruit protein monomers/oligomers and decrease kinetic and thermodynamic barriers for nucleation of aggregation (63,64). In the presence of heparin, TDP43-LCD formed solid amyloid fibrils, whose SDS stability and ThT fluorescence was much higher than that of amyloid hydrogel, which formed at pH 6, without heparin. It could be postulated that interactions stabilizing heparin-induced amyloids are different and stronger than those stabilizing hydrogels. We also observed lateral stacking of fibrils, which is also observed for the aggregates of other proteins, and is reported to be catalyzed by heparin (65, 66, 67). The lateral association may further enhance the stability of such fibrils.

Given that different sets of interactions (homotypic and heterotypic) regulate the assembly process of phase-separating systems, it could be hypothesized that the association of constituent macromolecules might be under kinetic control. In this study, the kinetics of assembly formation was different in all three conditions, with faster kinetics of association at pH 7 and above, and in the presence of heparin; and slower kinetics of association at pH 6 and below. The sets of interactions that were kinetically accessible under each condition determined the material nature of each assembly product.

Overall, this study has shown how a change in its physicochemical environment can modulate the material state of the assembly product of the TDP-43 low-complexity domain (Fig. 8). Our study is relevant in the cellular context, wherein different material states are observed in temporary cellular compartments that form through LLPS. Contrary to a common paradigm, where a protein’s liquid state is considered a precursor for all phase transitions (solid or gel), our report demonstrates that such transitions can also happen without an accompanying LLPS. In recent reports, the hierarchy of the phase transition has been followed through a battery of nanoscopic methods that have uncovered important on-pathway nanoscale intermediates that eventually lead to mesoscale assemblies. In light of these recent developments, our findings pave the way for studying the link between different material states that are formed through phase transition in a broader context.

Figure 8.

Schematic of TDP43-LCD assembly under three different solution conditions. To see this figure in color, go online.

Author contributions

D.K.G. and R.B. conceptualized the study. D.K.G. performed the experiments. D.K.G. and R.B. analyzed the data. D.K.G. and R.B. wrote the manuscript.

Editor: Samrat Mukhopadhyay.