The structure of mouse RIPK1 RHIM-containing domain as a homo-amyloid and in RIPK1/RIPK3 complex

Abstract

Receptor-interacting protein kinase 1 (RIPK1) is a therapeutic target in treating neurodegenerative diseases and cancers. RIPK1 has three distinct functional domains, with the center domain containing a receptor-interacting protein homotypic interaction motif (RHIM), which mediates amyloid formation. The functional amyloid formed by RIPK1 and/or RIPK3 is a crucial intermediate in regulating cell necroptosis. In this study, the amyloid structure of mouse RIPK1, formed by an 82-residue sequence centered at RHIM, is presented. It reveals the “N”-shaped folding of the protein subunit in the fibril with four β-strands. The folding pattern is shared by several amyloid structures formed by proteins with RHIM, with the central β-strand formed by the most conserved tetrad sequence I/VQI/VG. However, the solid-state NMR results indicate a structural difference between mouse RIPK1 and mouse RIPK3. A change in the structural rigidity is also suggested by the observation of weakened signals for mouse RIPK3 upon mixing with RIPK1 to form the RIPK1/RIPK3 complex fibrils. Our results provide vital information to understand the interactions between different proteins with RHIM, which will help us further comprehend the regulation mechanism in cell necroptosis.

Receptor-interacting protein kinase 1 (RIPK1) plays key roles in regulating multiple signalling pathways. Here, using solid state NMR, the authors report the amyloid structure of mouse RIPK1, formed by an 82-residue sequence centred at RHIM: it adopts an “N”-shaped folding of the protein subunit in the fibril.

Introduction

Receptor-interacting protein kinase 1 (RIPK1) is an essential upstream regulator of multiple signaling pathways, including inflammation, apoptosis, and necroptosis^1,2. Upon stimulation by tumor necrosis factor α (TNFα), RIPK1 is recruited to TNFα receptor 1 (TNFR1), followed by the recruitment of other proteins to form signaling complexes that activate different pathways downstream of TNFR1^1,3. One of the RIPK1’s functions in these complexes is to promote inflammation and immune system development by inducing nuclear factor-κB (NF-κB) activation^4,5. Additionally, RIPK1 can induce cell apoptosis or necroptosis, which requires its kinase activity. Consequently, RIPK1 has been an attractive target for developing new therapies for human diseases such as neurodegenerative diseases, cancers^6,7.

The multiple functions of RIPK1 are achieved through its various structural domains, including the N-terminal kinase domain, an intermediate domain, and a C-terminal death domain (Fig. 1a). RIPK1 belongs to the Ser/Thr kinase family. The kinase domain, with a canonical kinase fold, is essential to initiate the cascade of necroptosis and the caspase-dependent apoptosis^8,9. The C-terminal death domain is characterized as a six-helix bundle, mediating protein interactions for death receptor signaling and apoptosis¹⁰. While both the N-terminal and C-terminal domains of RIPK1 are well-folded, the intermediate domain is not. The intermediate domain is a sequence of more than 250 residues and contains a receptor-interacting protein homotypic interaction motif (RHIM) of about 20 residues. RHIM mediates RIPK1 interactions with other RHIM-containing proteins, such as receptor-interacting protein kinase 3 (RIPK3), forming a protein complex with amyloid structural properties^11,12 known as necrosome. In cells, forming such structures also requires RIPK1 kinase activity¹³. The hetero-amyloid of RIPK1/RIPK3 serves as a template, further inducing RIPK3 RHIM to form a homo-amyloid structure, activating RIPK3 kinase activity. In necroptosis, RIPK3 homo-amyloid acts as a central node, bridging multiple upstream signals to the downstream effector molecule mixed lineage kinase domain-like (MLKL) to induce cell membrane disruptions and cell death^14,15.

Fig. 1. Initial structural characterization of mRIPK1 amyloid fibrils.

a Schematic representation of mRIPK1 protein sequence with the kinase domain (blue), the RHIM-encompassing region (RHIM) (yellow), followed by the death domain (DD) (brown). The protein sequence of mouse RIPK1 RHIM domain (residues 485–566) used in this study is also shown. b Negative-stain transmission electron micrograph (TEM) of fibrils formed by mRIPK1 RHIM domain. The scale bar is 100 nm. This experiment was repeated at least 5 times independently with the same results. c 2D ¹H-¹⁵N correlation spectrum of uniformly ¹⁵N, ¹³C-labeled mRIPK1 fibrils, obtained at a field of 700 MHz with 100 kHz MAS frequency at 263 K.

Currently, the amyloid structures of the human RIPK1/RIPK3 RHIM complex¹⁶ and RIPK3 RHIM homo-oligomer (human hRIPK3 fibrils and mouse mRIPK3 fibrils) have been revealed by solid-state NMR (ssNMR)^17,18. In vitro, RIPK1 RHIM homo-oligomer (simplified as RIPK1 fibril) also adopts an amyloid structure. In this work, the structure of mouse RIPK1 (mRIPK1) fibril was determined using ssNMR, providing an essential piece in understanding the multiple-domain structure of RIPK1. Changes were also revealed upon mixing mRIPK1 and mRIPK3 to form the hetero-amyloid.

Results

Identification of the amyloid fibril core region using ¹H-detected and ¹³C-detected ssNMR

Mouse RIPK1 (residues 485–566), including the RHIM domain, was cloned and expressed using E. Coli. The TEM image of mRIPK1 fibrils, shown in Fig. 1b, depicts long, unbranched single-stranded fibrils. The 2D ¹H-¹⁵N correlation spectrum of uniformly ¹⁵N,¹³C-labeled mRIPK1 fibrils displays well-defined peaks (Fig. 1c) as well as some broad signals with lower resolutions (such as the area around ¹H 8.5 ppm and ¹⁵N 122 ppm).

3D ¹H-detected experiments, including hCANH, hcoCAcoNH, hCONH, and hCOcaNH, were carried out on mRIPK1 fibrils. These experiments provided sequential assignments (selected regions shown in Fig. 2), indicating the residues forming the fibril core structure were from K519 to V538. 2D ¹³C-¹³C DARR spectra (Fig. S1a) and 2D NCACX, NCOCX, and ¹⁵N-¹³C z-TEDOR (z-filter transferred echo double resonance, with mixing time of 6.4 ms) (Fig. S2) were also carried out to support the 3D sequential assignments.

Fig. 2. ¹H-detected 3D experiments for mRIPK1 assignments at 100 kHz MAS.

Signals from 3D hCaNH (orange), hcoCAcoNH (blue), hCONH (magenta) and hCOcaNH (green) ssNMR experiments of uniformly ¹⁵N, ¹³C-labeled mRIPK1 fibrils were used for sequential and intra-residue assignments. Schematic representation of the coherence transfer steps was presented on the right for the four experiments. These spectra were recorded at a field of 16.4 T at 263 K.

¹³C-detected experiments displayed additional resonances. Matching the resonance positions to the residue type, these resonances are likely from residues A514, D515/516, L517, I518, L540, S542, and three proline residues just outside the identified sequence, suggesting these residues also contribute to the fibril structure. ¹³C-detected 3D NCACX and 3D NCOCX experiments display the signals from A514 to M536, providing additional support for the assignment by 3D ¹H-detected experiments. Strip plots of ^{1 3}C-detected 3D NCACX and 3D NCOCX spectra from A514 to I518 (Fig. S3) are shown. Strip plots from T521 to H533, corresponding to the region shown in Fig. 2, are also displayed in Fig. S4. It is interesting to notice that the connection between Q529 and I530, which is at the center of RHIM domain, is weak. Considering the spectra regions showing the sequential connectivity before Q529 and after I530 are in relatively good quality, and there are no other IQ or IG residue pair in the entire sequence, the assignment is with high confidence. A 2D ¹H-¹³C correlation spectrum of mRIPK1 fibrils is shown in Fig. S1b, using mRIPK1 expressed in ¹⁵N-NH₄Cl,¹³C-Glucose and D₂O media with later labile ¹H back-exchanged. All the assignments are displayed in Table S1.

The secondary structure predictions based on the chemical shift (Δδ¹³Cα-Δδ¹³Cβ) and TALOS-N generally agree with each other (Fig. 3). Two short β-strands from residue N524 to S526 and I528 to I530 are suggested by the secondary chemical shift calculation, which are predicted as part of a continuous β-strand by TALOS-N. According to the secondary chemical shift prediction, I530 is right at the edge of a β-strand with the value of Δδ¹³Cα-Δδ¹³Cβ close to 0 ppm. This could lead to a decreased rigidity of I530, which might explain that only a weak connection between Q529 and I530 was observed in the 3D sequential experiments. Therefore, the mRIPK1 fibril subunit structure contain about 3 or 4 β-strands. We covered residues from D516 – G539 for the later structural calculation.

Fig. 3. Secondary structure prediction of mRIPK1 fibrils from chemical shifts and TALOS-N.

The amino acid sequence was displayed at the top of the figure with secondary structure prediction using PSIPRED Asterisks indicated residues with ambiguous or missing NMR data. Predicted secondary structure of the mRIPK1 fibrils based on backbone secondary chemical shifts (Δδ¹³Cα–Δδ¹³Cβ) (ppm) (middle). The glycine Cα secondary shift was shown in gray. TALOS-N [https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3701756/] prediction based on ssNMR chemical shifts showing dihedral angles φ and ψ values (bottom). Data presented are mean values ± SD of 25 best structures based on TALOS-N. Source data are provided as a Source Data file.

J-coupling-based ¹H-¹³C INEPT and TOBSY experiments were also carried out, showing peaks only from the peptide mobile segment (Fig. S5). The number of peaks shown by TOBSY is less than expected, considering there are 82 residues, and fewer than 30 residues are seen by cross-polarization-based experiments (Fig. 2 and S1, S2). For example, there are six serine and nine threonine residues outside the fibril core region. We only observed fewer than ten overlapped peaks at the corresponding region. There are five proline residues, 3 methionine, 3 glutamine, 3 glutamic acid and seven histidine residues (including the six histidine residues in His tag) outside the fibril core. At the corresponding positions, only a few peaks can be seen. The results indicate that the motion of some of the residues outside the fibril core domain is in the intermediate time scale, or the peaks are just heavily overlapped.

3D Structure of mRIPK1 Fibril

The X-ray diffraction of mRIPK1 fibril in Fig. S6a indicates typical features of cross-β amyloid structure, with equatorial and meridional diffraction at about 9.6 Å and 4.7 Å, respectively. The 4.7 Å distance was used as a constraint in the structural calculation to define the distance between neighboring hydrogen-bonded β-strands. The 9.6 Å distance represents the typical distance between two β-sheets.

The mass-per-length (MPL) value, indicating the molecular weight of the fibril per unit length, provides insight into the number of subunits in the cross-β unit of a fibril. Dark-field Beam Tilted (BT)-TEM was used to obtain the MPL value for the fibril. Figure S6b shows a typical image containing a mixture of mRIPK1 fibrils and tobacco mosaic virus (TMV), with TMV serving as a molecular weight standard with an MPL value of 131 kDa/nm. The MPL value of mRIPK1 was determined to be 21.6 kDa/nm with a width of about 6 kDa/nm using a Gaussian fit (Fig. S6c) on a total of 403 counts. The theoretical value for a single subunit in the cross-β unit of the fibril is 21.3 kDa/nm (mRIPK1 subunit molecular weight is 10.01 kDa, and the distance between two hydrogen-bonded subunits is 4.7 Å, thus 10.01 kDa /0.47 nm = 21.3 kDa/nm). Therefore, mRIPK1 fibril contains only a single subunit in the cross-β unit.

Atomic force microscopy (AFM) was also used to obtain the information on the dimension of the fibril, indicating a fibril height of 3.1 ± 0.4 nm based on 209 readings (Fig. S6d, e, f). Residues not in the fibril structural core may distribute randomly around the fibril, adding error to the height readings. The height profile along the fibril axis could reveal the fibril’s periodic twist information. However, mRIPK1 fibril showed ambiguous results, with distances ranging from 16–35 nm for a complete 360° twist (Fig. S7), indicating that mRIPK1 fibril might have a heterogeneous twist state. Consequently, the twist was not considered in calculating the final fibril structure.

Inter-molecular residue alignment of the neighboring subunits was first investigated with ¹³C-¹³C correlation experiments on sparsely ¹³C-labeled mRIPK1 fibrils using [2-¹³C]-labeled glycerol as a carbon source. The DARR experiments with different mixing times were compared (Fig. 4). For short DARR mixing time (50 ms), the spectrum showed only intra-residue Cα-Cβ peaks for residues I, V, and S, as the Cα and Cβ carbons for other residue types were mostly not simultaneously ¹³C-labeled. However, for longer DARR mixing time (400 ms), the ¹³C-¹³C 2D spectrum displayed cross-peaks for inter-residue correlations with longer distances, showing Cα-Cβ peaks for D516, T521, N524, Q529, N532, N534, and M536. The result indicated that mRIPK1 fibrils were in parallel in-register conformation^19,20. Additional evidence came from NHHC experiments (Fig. S8a) on mRIPK1 fibrils with ¹³C-labeled and ¹⁵N-labeled mRIPK1 at 1:1 ratio. The NHHC spectrum, which only captures the inter-molecular correlations, displayed a similar spectrum to the NCA spectrum of ¹³C, ¹⁵N-uniformly labeled fibrils obtained at the same temperature, with no indication of cross peaks from an anti-parallel conformation. CHHC experiments conducted on ¹³C, ¹⁵N-uniformly labeled fibrils to look for ¹³Cα-¹³Cα peaks from the aligned inter-strand residue pairs further confirmed the parallel in-register conformation. Four conditions were tested for CHHC experiments, including t_HH-spindiff = 100 μs t_HC,short = 150 μs (Fig. S8b), t_HH-spindiff = 200 μs t_HC,short = 75 μs or t_HC,short = 150 μs (Fig. S8c) and t_HH-spindiff = 200 μs t_HC,short = 400 μs (Fig. S8d). Figure S8b did not show strong ¹³Cα-¹³Cα cross peaks, while more ¹³Cα-¹³Cα cross peaks were displayed with longer t_HH-spindiff (Fig. S8c, d). Most of the ¹³Cα-¹³Cα cross peaks were assigned as sequential correlations (shown in bold in Fig. S8d). Since no correlation peaks suggested an anti-parallel conformation in Fig. S8b, c, we confirmed that mRIPK1 fibril was in parallel in-register conformation.

Fig. 4. 2D ¹³C-¹³C DARR spectra of mRIPK1 fibrils using [2-¹³C]-glycerol labeling indicating an in-register parallel β-sheets architecture of the fibrils.

2D ¹³C-¹³C correlation spectrum with 50 ms (left) and 400 ms (right) DARR mixing time, showing that D516Cα-Cβ, T521Cα-Cβ, N524Cα-Cβ, Q529Cα-Cβ, N532Cα-Cβ, N534Cα-Cβ and M536Cα-Cβ can only be detected with the long mixing time (400 ms).

More inter-residue interactions were obtained using ¹³C-¹³C correlation experiments with DARR mixing times varying from 25 ms to 500 ms on the ¹³C-uniformly labeled sample. Figure S9 displayed a spectrum with 500 ms DARR mixing, where correlation peaks separated by three or more residues on the protein sequence were marked in magenta and those spaced within two residues apart were marked in black. For example, G527Cα displayed correlation peaks with M536Cβ/Cγ/Cε and V538Cγ1/Cγ2, indicating the last β-strand has close contact with the center β-strand. G527Cα also exhibited an ambiguous correlation peak with I530Cδ1 or I522Cδ1. This peak did not appear in ¹³C-¹³C correlation experiments with 200 ms DARR mixing, indicating a relatively long distance. To distinguish intra-molecular and inter-molecular interactions, mRIPK1 fibrils were prepared by diluting ¹³C-uniformly labeled peptides with non-labeled peptides in a ratio of 1:3 and 1:4. The intensity of ¹³C-¹³C correlation spectrum with 500 ms DARR was significantly attenuated; however, many correlation peaks indicating close interactions between the last β-strand and the center β-strand remained, such as the peaks between G527Cα and M536 Cε, M536 Cβ and Q529 Cγ/Cβ (labeled in bold in Fig. S10). This suggests these correlations are intra-molecular. The correlation peaks between the first β-strand and the center β-strand were mostly found in the aromatic region and at the isoleucine sites. However, the peaks in the aromatic region were too weak, mostly showing the correlations to the residue Cα. The peaks at the isoleucine sites could be observed upon dilution, such as peaks between I530/I522Cδ1 and I528Cγ1, and between I528Cδ1 and I530/I522Cγ1. However, those peaks were ambiguous.

Further experiments were carried out on fibril samples containing a 1:1 mixture of mRIPK1 generated using [2-¹³C]-labeled glycerol and [1,3-¹³C]-labeled glycerol. The ¹³C-¹³C correlation spectrum of the mixed labeling sample was compared to mRIPK1 fibrils generated using only [2-¹³C]-labeled glycerol or [1,3-¹³C]-labeled glycerol (Fig. S11). The comparison revealed numerous long-range inter-residue correlations present in the fibrils generated using [2-¹³C]-labeled glycerol or [1,3-¹³C]-labeled glycerol remained in the fibrils prepared with mixed labeling (labeled in black in Fig. S11), confirming they are intra-molecular correlations. For example, the peak between M536Cε and Q529Cδ observed for fibrils using [2-¹³C]-glycerol labeling scheme remained in the fibrils prepared with mixed labeling. Based on the evidence and the fact that there is only one subunit within the cross-β unit of the fibril, the observed correlations were all considered intra-molecular interactions. The residue correlations used for fibril structure calculation were summarized in Table S2.

Xplor-NIH was used to calculate the structural model for mRIPK1 fibril using two rounds of calculations, and the final structure was shown in Fig. 5 with the sequence from D516 – G539. The structure was deposited at the PDB data bank with code: 8IB0. Other segments with disorders or intermediate dynamics were not included in the calculation. The overlay of 10 structures with the minimum energies is shown in Fig. 5a, with the intra-molecular constraints obtained from ssNMR marked. The fibril subunit displays four β-strands with a dimension of about 3.1 × 2.8 nm, matching the AFM result which shows a fibril height of 3.1 ± 0.4 nm (Fig. 5b). Three β-strands (β1, β3, β4) are sandwiched together, but the second β-strand (β2) adopts an orientation nearly perpendicular to β1 and β3. β3 (I528–I530) is the shortest β-strand. Figure 5c shows the fibril structure with 10 identical subunits. The structural statistics for mouse RIPK1 fibril are summarized in Table S3.

Fig. 5. 3D structure of mRIPK1 fibril.

a An overlay of 10 lowest-energy NMR structures calculated by xplor-NIH with NMR restraints indicated. The red solid line indicated the unambiguous assignments, and blue dashed line were from the ambiguous assignments. b Top view of a central monomer from residues D516 – G539, showing the fibril core of the RHIM domain. The dimension of the fibril core (3.1 × 2.8 nm) and the distance between β-sheets (9.6 Å) were indicated. The cartoon representation of mRIPK1 fibril subunit showed four β-strands with an “N” shape. c Sticks representation of mRIPK1 fibril looking down the fibril axis, indicating the in-register parallel alignment between each subunit. All figures were prepared using PyMOL (version 2.5.7).

Structure changes of amyloid protein subunit upon mRIPK1 mixing with mRIPK3

Since mRIPK1 interacts with mRIPK3 in cells to form hetero-amyloids, their hetero-amyloid structure was also investigated using ssNMR. mRIPK1 and mRIPK3 proteins were expressed and purified separately using HPLC. To ensure thorough mixing, the proteins were first dissolved in 30% acetic acid at 80 °C and then mixed at an approximate molar ratio of 1:1 based on nanodrop UV absorption readings and SDS-PAGE analysis (Fig. S12a). TEM images indicated no fibril formation before dialysis (Fig. S12b), which also showed the image of hetero-amyloid fibrils formed after dialysis. Although the preparation method differs from how homo-amyloid is made, a test on mRIPK1 homo-amyloid using both methods indicated no difference in the fibril structures (Fig. S13).

The hetero-amyloid was later dissolved in 8 M urea, and SDS-PAGE was used to determine the molecular ratio of mRIPK1 to mRIPK3 in the hetero-amyloid (Fig. S12a, lane 8-11, 15-16), indicating a ratio about 1.7:1 based on band darkness integration. The supernatant after collecting the hetero-amyloid was also obtained and concentrated in urea, showing a dominant mRIPK3 band in the gel (lane 14). The result is consistent with a higher concentration of mRIPK1 in the hetero-amyloid pellet.

The fluorescence intensity induced by Thioflavin (ThT) binding on the fibril was measured (Fig. S12c). Interestingly, mRIPK1 homo-amyloid displayed significant higher fluorescence intensity than mRIPK1/mRIPK3 hetero-amyloid and mRIPK3 homo-amyloid, with mRIPK3 homo-amyloid displayed the lowest fluorescence intensity among the three fibrils. A post-assembly mixture of mRIPK1 fibril and mRIPK3 fibril in 1:1 ratio was also prepared, which displayed a ThT fluorescence intensity slightly higher than mRIPK1 homo-amyloid, consistent with a combination of two types of fibrils after the assembly. This indicates a successful mixing of mRIPK1 with mRIPK3, with the hetero-amyloid adopting a different surface profile from mRIPK3 or mRIPK1 homo-amyloids, affecting ThT binding on the fibril.

The ¹³C-¹³C correlation spectra of mRIPK1/mRIPK3 hetero-amyloid using 50 ms DARR mixing were compared to mRIPK1 homo-amyloid and mRIPK3 homo-amyloid for both ¹³C uniformly-labeled (Fig. S14) and selectively-labeled samples using [2-¹³C]-glycerol (Fig. S15). The secondary chemical shifts for mRIPK1 and mRIPK3 were compared between hetero- and homo-amyloids in Fig. S16 (left and middle), revealing no significant chemical shift changes upon mixing the two proteins. Furthermore, the same type of residues in corresponding positions in mRIPK1 and mRIPK3 homo-amyloids gave almost identical chemical shifts (Fig. S16 right), such as S526/S446, Q529/Q449. This supports a conserved secondary conformation for mRIPK1 and mRIPK3 in both homo-amyloid and hetero-amyloid. A complete summary of chemical shifts observed in [2-¹³C]-glycerol labeled homo-amyloid and hetero-amyloid is displayed in Table S5.

Figure S15 showed that the cross-peaks of mRIPK1 in hetero-amyloid displayed intensities slightly <60% compared to mRIPK1 homo-amyloids. Since the amount of mRIPK1 was also reduced to about 60% in the hetero-amyloid (1.7/(1.7 + 1) = 0.63), it suggests that the signal of mRIPK1 generally remained in the hetero-amyloid. However, mRIPK3 displayed significant intensity attenuations. Figure S15 quantified the intensity decrease on several mRIPK3 peaks upon mixing with mRIPK1, showing only <10% signal intensity compared to mRIPK3 homo-amyloids. Only those peaks that were not overlapping with others were shown.

Furthermore, the peak intensity changes on inter-residue cross-peaks were investigated using DARR with a long mixing time. Figure 6 shows a ¹³C-¹³C correlation spectra comparison between homo- and hetero-amyloids with 400/500 ms mixing time. For mRIPK1, not many changes were observed upon forming the hetero-amyloid, with key cross-peaks defining the close association of the β-strands remaining. For mRIPK3, cross-peaks such as L456Cγ-Q449Cδ, G451Cα/L456Cβ-Q449Cδ, Q449/V457Cβ-L456Cγ, Q449/V457Cβ-G451Cα/L456Cβ, and Q449/N454Cα-G451Cα/L456Cβ, which define close interactions between L456 and Q449, disappeared or significantly attenuated in the hetero-amyloid. Although the general signal intensity decrease for mRIPK3 may cause the disappearance of inter-residue cross-peaks, the results together suggest that mRIPK3 in the hetero-amyloid is diluted by the more abundant mRIPK1 molecules and adopts a less rigid structure compared to its homo-amyloid form.

Fig. 6. 2D ¹³C-¹³C correlation spectra of [2-¹³C]-glycerol labeled protein fibrils with 400 or 500 ms DARR mixing, indicating different changes for mRIPK1 and mRIPK3 in the hetero-amyloid.

Recorded on the mRIPK1 fibrils (red), the mRIPK3 fibrils (purple), and the mRIPK1/mRIPK3 fibrils (dark blue). a The correlation peaks of mRIPK1 generally matched with those in the hetero-amyloid. b Many correlation peaks of mRIPK3 disappeared or attenuated in the mRIPK1/mRIPK3 spectrum, as indicated by the bold labeling for the mRIPK3 fibrils. The peaks for L456Cγ-Q449Cδ, G451Cα/L456Cβ-Q449Cδ, Q449/V457Cβ-L456Cγ,Q449/V457Cβ-G451Cα/L456Cβ, and Q449/N454Cα-G451Cα/L456Cβ were highlighted in (b) and (c) which define a close interaction between the center and the last β-strand.

Discussion

The structure of mRIPK1 closely resembles the published hRIPK3 fibril structure with an ‘N’ shape (Fig. 7). However, the structure of mRIPK1 fibril exhibits a noticeable difference compared to the mRIPK3 fibril, where the last β-sheet of mRIPK3 fibril extends away from the center β-sheet. This is intriguing, considering they all share highly conserved sequences. The center β-strand for all the fibril structures here is (I/VQI/V), followed by a G residue. Both hRIPK3 and mRIPK1 start the center β-strand with a G residue (GI/VQI/VG). But in mRIPK3, a negatively charged E residue with a longer side-chain replaces the beginning G residue (EVQIG) (Fig. 7c). We then compared the sequences of other proteins with the RHIM domain (Fig. S17) and found the residue immediately before the conserved segment (I/V/LQI/V/L) varied a lot from G to a hydrophobic residue with a long side chain (M) or from a negative charged E to a positive charged R. Other differences in the sequence could be found at the end of the last β-strand. For example, two P residues are at residue 459, 460 for mRIPK3 whereas at the corresponding position for mRIPK1, hRIPK3, they are GL, QQ respectively. It is plausible that these residue differences collectively play a role in modulating RHIM core fibril structures.

Fig. 7. Comparison of mRIPK1 and hRIPK3, mRIPK3 fibril structures.

a ssNMR structure of mouse RIPK3 fibrils (blue, PDB code: 6JPD) and (b) ssNMR structure of human RIPK3 fibrils (pink, PDB code: 7DA4). c The overlay of subunit structures of mRIPK1 (green), mRIPK3 (blue) and hRIPK3 (pink).

The structure of mRIPK3 homo-amyloid is less compact than mRIPK1 homo-amyloid. Considering that mRIPK1 can interact with mRIPK3 in mouse cells, the difference in the fibril structure of mRIPK1 and mRIPK3 may affect the interaction and structure of the hetero-amyloid formed by mRIPK1 and mRIPK3. Indeed, our SDS-PAGE study on hetero-amyloids with an mRIPK1: mRIPK3 ratio of 1.7:1 indicated a slightly higher preference of mRIPK1 to interact with itself than with mRIPK3. NMR correlation spectra on the hetero-amyloid displayed a decrease in the intensity of mRIPK3 correlation peaks, disproportionate to its content, indicating that mRIPK3 was less rigid and likely had fewer interactions between the center and the last β-strand compared to its homo-amyloid form.

Recently, an amyloidase (HSPA8) was discovered that could suppress necroptosis by inhibiting and reversing amyloid formation of RHIM-containing proteins. This suggests that HSPA8 specifically recognizes RHIM-containing proteins such as mRIPK3 through the last β-strand, which they summarized as NX1φX3 (Fig. S17b, φ is a conserved hydrophobic residue)²¹. A relative flexibility of this segment for mRIPK3 in both homo- and hetero-amyloids would allow possible interactions between HSPA8 and the functional amyloid.

Our experiments showed an mRIPK1:mRIPK3 ratio of 1.7:1 for mRIPK1/mRIPK3 hetero-amyloid, even though mRIPK1 and mRIPK3 were mixed in an ~1:1 ratio. This differs from previous research using an hRIPK1:hRIPK3 ratio of 1:1 for hRIPK1/hRIPK3 hetero-amyloid. The structure for hRIPK1/hRIPK3 hetero-amyloid was different as well. In the previous research, hRIPK1 and hRIPK3 were polycistronically subcloned and co-expressed, and the hetero-amyloid was purified directly from the cell lysate¹⁶. Therefore, the preparation protocol differed from ours.

Our SDS-PAGE study contains several potential error sources, such as the integration method in the gel analysis and how thoroughly the amyloid can be dissolved by urea, etc. However, the errors are not expected to be significant. Furthermore, it is also possible that the interaction of RIPK1/RIPK3 could vary between human, mouse and other species, resulting in different molecular ratios. It is recognized that mouse and human cell necroptosis regulation have some difference^22,23. Further investigation into this issue is required.

Methods

Protein expression and purification

Mouse RIPK1 protein (485-566) and mouse RIPK3 protein (409-486) genes with an N-terminal His×6 tag were subcloned into the pET32a vector and expressed in E. coli Rosseta (DE3) cells (Transgene, China). Unlabeled M9 medium contained Na₂HPO₄ 3 g, K₂HPO₄ 1.5 g, NaCl 0.25 g, 2 M MgSO₄ 1 mL, 2 M CaCl₂ 0.1 mL, Thiamine 50 mg, Glucose 4 g, NH₄Cl 1.5 g per liter of solution. For ¹⁵N and/or ¹³C labeling, mRIPK1 was prepared by expression using labeled M9 medium containing ¹⁵NH₄Cl and ¹³C₆-glucose or [1,3-¹³C]-glycerol, [2-¹³C]-glycerol (Cambridge Isotope Laboratories) as the sole nitrogen and carbon sources. All media were supplemented with 100 mg/mL ampicillin (Sangon Biotech, Shanghai, China). For protein expression, 5 mL LB (Sangon Biotech, Shanghai, China) were inoculated from frozen glycerol stocks of E. coli Rosseta (DE3) containing the target gene and incubated for 7 h (220 rpm/37 °C). The cells were then transferred to 100 mL of unlabeled M9 medium and incubated overnight. The overnight cultures (100 mL) were inoculated into 400 mL of M9 medium and then incubated for 7 h at 37 °C until the OD600 reached 0.8–1.0. After centrifugation (Sorvall Lynx 6000; Thermo Fisher Scientific, USA) at 7808× g for 10 min, the cell pellets were collected and resuspended in 500 ml of either unlabeled or ¹³C, ¹⁵N-labeled M9 medium and resuscitated for 20 min. Protein expression was induced with 1 mM IPTG (Isopropyl β-d-1-thiogalactopyranoside) (Sangon Biotech, Shanghai, China) at 22 °C for 10 h. The expression and purification of mouse RIPK3 protein were performed according to previously published methods¹⁷.

²H, ¹³C, ¹⁵N-labeled mRIPK1 was expressed in M9 medium prepared in 99.8% D₂O (Sigma-Aldrich 617385, Oakville, Canada) and supplemented with U-¹³C₆ glucose, ¹⁵NH₄Cl. E. coli Rosseta (DE3) cells were grown in 99.8% D₂O LB medium containing 100 mg/mL ampicillin for 8 h at 37 °C before the culture was centrifuged for 5 min at 6000 × g. The cell pellet was resuspended in 50 mL 99.8% D₂O M9 medium and then grown at 37 °C overnight. The overnight culture was transferred into 150 mL of 99.8% D₂O M9 medium until the OD600 reached 0.8–1.0. Finally, the cells were collected by centrifugation (8327 × g, 10 min) and resuspended in 200 mL of 99.8% D₂O M9 medium. The cells were induced with 1 mM IPTG after resuscitation and grown at 22 °C for 18 h for protein expression. The labile ¹H was back-exchanged during the protein purification and fibril preparation.

For lysis, the cells were centrifuged at 10628 × g for 10 min and resuspended in lysis buffer containing 50 mM Tris-HCl (pH 8.0), 300 mM NaCl and 1 mM β-mercaptoethanol. The cells were lysed using a high-pressure nano homogenizer (FB-110X, Shanghai Litu Ins, China) at 800-900 bar and centrifuged at 13521 × g for 60 min (Hermle Z446K, Hermle Labortechnik GmbH, Germany). The pellet was dissolved in 20 mL of dissolving buffer containing 6 M guanidine hydrochloride, 50 mM Tris-HCl (pH 8.0) and 300 mM NaCl, by rotating the sample at 4 °C for 60 min. The solution was centrifuged at 13521 × g at 4 °C for 10 min, and the supernatant was collected and incubated with Ni-NTA beads 6 FF (Smart-Life Science, China) at 4 °C for 60 min. After elution with wash solution (6 M guanidine hydrochloride, 50 mM Tris-HCl (pH 8.0), 300 mM NaCl, 20 mM imidazole) to remove host proteins, the target protein was eluted with elution solution (6 M guanidine hydrochloride, 50 mM Tris-HCl (pH 8.0), 300 mM NaCl, 400 mM imidazole). The protein elution solution was stored at -80 °C if not directly used for fibril preparation.

Assembly of mRIPK1 fibrils in vitro for solid-state NMR studies

1 mg/mL protein elution solution was dialyzed for 4 days using 3.5-kDa dialysis membranes (MWCO 3.5 K, W.45 mm, Diameter 29 mm; BBI Life Science, Shanghai, China) in Milli-Q water (pH 7.5) at room temperature, with the water replaced twice every 24 h. During this process, the imidazole and other salts were exchanged into water, and white precipitate gradually formed, indicating successful production of the mRIPK1 fibrils. This procedure was applied to uniformly ¹³C,¹⁵N-labeled samples, and deuterated samples.

For mRIPK1/mRIPK3 hetero-amyloid, and other mRIPK1 homo-amyloid samples that required mixing two different isotopically labeled peptides (such as mRIPK1 homo-amyloid with ¹⁵N- and ¹³C-labeled peptides in a 1:1 ratio, with ¹³C-labeled peptide diluted into non-isotopically labeled peptide in 1:3 or 1:4 ratio, or 1:1 mixing of mRIPK1 prepared using [2-¹³C]-glycerol and [1,3-¹³C]-glycerol labeling scheme), each protein was further purified by HPLC, lyophilized, and dissolved in 30% acetic acid. The protein samples were submerged in a water bath heated to 80 °C until complete dissolution, and the protein concentration was quantified via spectrophotometry (Thermo Scientific NanoDrop 1000). The two proteins were then mixed in a specified molar ratio and the mixture was dialyzed at room temperature for 4 days against pure water with daily water changes.

The assembly of amyloid fibrils during dialysis was confirmed using transmission electron microscopy (TEM). Subsequently, the fibril samples were centrifuged at 225,000× g in a fixed-angle rotor (Type 70 Ti, Beckman Optima XPN-100, USA) for 1 h at 4 °C and transferred into Bruker 3.2 mm using a house-made tool, or centrifuged at 50,000 × g using a swing-bucket rotor (SW 41 Ti, Beckman Optima XPN-100, USA) and transferred into 0 .7 mm rotors using the Bruker packing device. Extra water was removed during packing. Usually, 20 mg of wet sample can be packed into a 3.2 mm standard-wall rotor, 30 mg wet sample can be packed into a 3.2 mm thin-wall rotor, and 1.2–1.6 mg wet sample can be packed into a 0.7 mm rotor.

SDS-PAGE experiments

SDS-PAGE was utilized to analyze the hetero-amyloid formed by mRIPK1 and mRIPK3 proteins. Both proteins were adjusted to the same starting concentrations, as determined by NanoDrop. Each protein or fibril sample was dissolved in 8 M urea and heated to improve solubility. For the hetero-amyloid prepared using the aforementioned protocol, the fibril was collected after dialysis by centrifugation at 225,000 × g (Beckman Optima XPN-100, USA) for 1 h at 4 °C, while the supernatant was also collected and concentrated with urea. All samples were then spun at 13682 × g for 5–10 min, and the supernatant was filtered using a 0.22 μm filter head. Samples were boiled in SDS-loading buffer and separated using 15% SDS-PAGE for the experiments. Urea was chosen due to its superior performance in SDS-PAGE; however, fibril dissolved more thoroughly in 30% acetic acid. The SDS-PAGE gel images were acquired with Amersham Imager 680 Chemiluminescent Imaging System (GE Healthcare), and analyzed using the Image J (NIH, version 1.54).

Thioflavin T (ThT) assay

Amyloid formation was monitored using ThT binding assays. ThT was added to a Costar 96-well plate (Corning) at a final concentration of 50 μM. The protein was first dissolved in 30% acetic acid, and the concentration was adjusted to a total of 20 μM for the hetero-amyloid or post-assembly mixture of mRIPK1 and mRIPK3 homo-amyloids in a 1:1 ratio of mRIPK1 and mRIPK3 in a 200 µL volume using 10 mM PB buffer (pH 7.4). For the homo-amyloids, the protein concentration was adjusted to 10 μM. The supernatant from the hetero-amyloid preparation was also incubated, and its ThT binding was monitored. Fluorescence intensity was measured in triplicates using a Perkin-Elmer EnSight Multimode Plate Reader, with excitation at 430 nm and emission recorded between 450 and 600 nm. Data were collected continuously at 25°C. Statistical analysis was performed using Origin2020 (Northampton, Massachusetts, USA).

Negative-stain transmission electron microscopy

Mouse RIPK1 was characterized by TEM. A 10 μL sample was placed on a 300 mesh copper grid with a carbon-plated supporting film (Beijing Zhongjingkeyi Technology Co., Ltd., Beijing, China) and stained with 2% (w/v) uranyl acetate for 45 s. Excess liquid was blotted off, and the grid was air-dried for 60 min. TEM analysis was conducted using a Tecnai G2 Spirit Transmission Electron Microscope (FEI, Czech) operating at 120 kV.

Atomic force microscopy

The surface morphology of the fibrils on a mica surface was characterized by atomic force microscopy (AFM). A 1 mg/ml fibril solution was diluted 1:100, 20 μL of the sample was placed on clean mica slices and incubated for 3 min at room temperature. The mica slices were then washed with Milli-Q water to remove residual protein molecules and buffer. Excess water was poured off, and the samples were allowed to dry for >10 h. AFM images were obtained using a Dimension Icon AFM with Bruker Nano scope V controller in tapping mode in air (Bruker, USA). SCANASYST-Air probes with a spring constant of 0.4 N/m and a resonance frequency of 70 kHz (Bruker, USA) were used for the measurements. The images were scanned using 528 pixels per line (528 lines) over a 1.5 µm² area at a frequency of 0.986 Hz. All AFM images were analyzed by NanoScope Analysis (version 1.8).

X-ray diffraction measurement

The sample solution was ultracentrifuged at 225,000 × g for 1 h at 4 °C to collect the fibrils. X-ray diffraction measurements were performed on a Bruker X-ray diffractometer using Cu κα radiation at 0.154184 nm wavelength (Bruker D8 Focus, Bruker, Karlsruhe, Germany). The distance between the sample and the detector was 50 mm. The fibril pellet was mounted in a loop and images were collected at room temperature. Images were processed with the adxv.x86_64RHEL6 program (Scripps Research Institute, La Jolla, CA, United States). The diffraction of each fibril sample was replicated at least twices.

MPL measurement by BT-TEM

The number of protofibrils in the mature mRIPK1 fibril was determined by beam tilted (BT) TEM. Images were collected using Tecnai G2 Spirit Transmission Electron Microscope operating at 120 kV (FEI, Czech), and the intensities of the fibril images in pixel units were analyzed using ImageJ software (NIH, version 1.54). A total of 403 fibril counts were obtained from 50 images.

Tobacco mosaic virus (TMV) was generously provided by Dr. Wenbing Zhou (Yuxi Tobacco Company, China) and used as a standard to calculate the molecular weight per unit length (MPL) for the fibrils. TMV has a theoretical MPL value of 131 kDa/Ų⁴. The virus was extracted from fresh tobacco infected with TMV and stored in 0.01 M PBS (pH 7.0).

For sample preparation, 5 μL of diluted TMV solution and 10 μL of fibril sample were mixed and deposited on the surface of a grid with an ultrathin carbon film (Beijing Zhongjingkeyi Technology Co., Ltd., Beijing, China). After 20 s, excess solution was removed by blotting with a filter paper. The grid was then washed ten times with ultrapure water. Excess water was removed, and the grid was allowed to dry at room temperature.

The MPL value for each fibril segment (units of kDa/nm) was calculated using the equation:

$$\mathrm{MPL}=\frac{131}{I_{\mathrm{F}}}\times I_{\mathrm{TMV}}$$ 1

where:

• I_F  is the integrated image intensity for one fibril segment after background correction, calculated as F- (I_B1 + I_B2)/2, with F being the integrated image intensity for the fibril segment and I_B1 and I_B2 being the background intensities on both sides of the fibril segment.

• I_TMV is the integrated image intensity for TMV after background correction, calculated as B- (I_B1  +  I_B2)/2, with B being the integrated image intensity for TMV and I_B1 and I_B2 being the background intensities on both sides of the TMV rods²⁵.

This methodology allows for precise determination of the MPL values of mRIPK1 fibrils by comparing them against the known standard of TMV.

Solid-state NMR spectroscopy

Solid-state NMR spectra were performed on a Bruker AVANCE NEO 700 MHz (¹H Larmor frequency) spectrometer equipped with a 0.7 mm triple resonance (¹H, ¹³C, ¹⁵N) Magic Angle Spinning (MAS) probe for all ¹H-detected experiments at 100 kHz MAS, and a 3.2-mm triple resonance (¹H, ¹³C, ¹⁵N) MAS probe for ¹³C-detected experiments at 15 kHz MAS. Spectra were indirectly referenced to DSS. For ¹H-detected ssNMR experiments, 2D hNH and hCH and 3D hCONH, hCANH, hcoCAcoNH, hCOcaNH spectra²⁶ were recorded on uniformly ¹³C,¹⁵N-labeled sample. And 2D hNH and hCH spectra were recorded for ²H, ¹³C, ¹⁵N-labeled sample. The sample temperature was maintained at about 279 K. Typical 90° pulse lengths were 0.8 μs for ¹H, 2.0 μs for ¹³C and 2.5 μs for ¹⁵N. The ¹H-¹⁵N and ¹H-¹³C Hartman–Hahn cross-polarization (CP) employed a linear amplitude ramp of 90–100% on ¹H with the effective rf fields of 130 kHz, and 30 kHz on ¹⁵N or ¹³C, respectively, the optimized contact times were 0.6–3 ms. 20 kHz MISSISSIPPI scheme was used for water suppression (100 – 200 ms), and 10 kHz WALTZ-16 heteronuclear decoupling was employed during acquisition.

¹³C-detected ssNMR spectra using 2D dipolar-assisted rotation resonance (DARR)²⁷ with various mixing time (50 ms, 200 ms, 500 ms), and z-filter transferred echo double resonance (zTEDOR)^28,29 with a mixing of 6.4 ms were recorded on both uniformly ¹³C,¹⁵N-labeled samples and ¹³C selective-site labeled samples prepared using [1,3-¹³C] glycerol or [2-¹³C] glycerol as the carbon source. For 2D ¹³C–¹³C correlation experiments, the Hartman–Hahn CP was done with a ¹³C field strength of 51.2 kHz and the ¹H field strength adjusted to 68.2 kHz. For ¹³C-¹⁵N correlation experiments, the ¹³C frequency was set to 54 ppm during SPECIFIC-CP for NCACX and 172 ppm for NCOCX. For all the ¹³C-detected experiments, SPINAL 64 decoupling scheme was applied at a ¹H rf field of 86-90 kHz. 2D NHHC spectra were recorded on 1:1 mixture of ¹³C and ¹⁵N-labeled sample. 2D NCA, NCO, NCACX, NCOCX³⁰, ¹H-¹³C insensitive nuclei enhancement by polarization transfer (INEPT)³¹, INEPT-¹³C-¹³C-total through-bond-correlation spectroscopy (TOBSY)³², CHHC³³ and 3D NCACX, NCOCX, CONCA experiments were recorded on uniformly ¹³C,¹⁵N-labeled samples. Further details on the experimental acquisition parameters are given in Table S4. All spectra were processed in Topspin 4.0 and analyzed using the program Sparky (D. Goddard and D. G. Kneller, SPARKY 3, University of California, San Francisco). The ¹³C chemical shifts of mouse RIPK3 protein was taken directly from BMRB: 36243.

Structure calculation

The core structure of the mRIPK1 amyloid from residue D516 to G539 was calculated using Xplor-NIH (version 2.45.5). Two rounds of calculations were carried out. In the initial calculation, 192 structures were calculated from an extended conformation using 30 dihedral angle restraints (Table S1), 87 unambiguous intramolecular distance restraints, 23 ambiguous intramolecular restraints and 16 parallel intermolecular constraints (Table S2). The backbone ϕ and φ dihedral angle restraints were derived from TALOS-N³⁴ prediction based on the chemical shifts. The dihedral angles from either the N-terminal residues (D516, L517), C-terminal residues (D537, V538, G539) and predicted turn regions (residue F523, D524, G527, H533) were not used in the calculation due to uncertainty.

In the calculation, the strict symmetry module (symSimulation) was utilized, where only a single copy of protomer coordinates was maintained. 5 copies of the monomer subunit were used to generate the fibril segment, where 4 subunits were generated from the protomer using rigid body translations; twist angle was assumed to be zero in the calculation since AFM showed inconsistent twist angles with a distribution. The protocol started with torsion-angle dynamics for a duration of 10 ps or 5000 timesteps at 4000 K, followed by annealing to 25 K in decrements of 12.5 K for 20 ps or 2000 timesteps of torsion-angle dynamics at each temperature and finally 500 steps of energy minimizations in torsion angle and Cartesian coordinates. The calculation was done on the high-performance calculation platform at ShanghaiTech University.

6 structures with the lowest energy were selected as the starting structures for the second-round calculation of 192 structures. In the refinement, EEFx implicit solvent force field was applied to replace RepelPot term used in the first round of calculation. The side-chain chi1 angles from TALOS-N prediction (L517, -63.7°; T521, -62.4°; I522, -55.7°; N524, -60.6°; S525, 58.5°; S526, 66.9°; I528, -60.9°; I530, -60.3°; H533, -63.1°; N534, -62.9°; Y535, 180.0°; D537, -66.7°) were also used with the error of ±15° to improve the side-chain conformation. Finally, the best 10 structures were selected from the refinement and validated at https://validate-rcsb-2.wwpdb.org/. The structures were deposited to PDB: 8IB0. The statistical summary for the protein structure was shown in Table S3. And NMR chemical shifts were deposited into BMRB: 36547.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

J.L. prepared all samples for the structural studies, took AFM images, captured the EM and BT-TEM dark-field images; J.X.L. and J.L. carried out the structure calculations; J.W. helped in SSNMR experiments; J.Z. for the help in the XRD data collection; J.X.L., H.Y.W., J.W. and J.L., Original manuscript; J.L., and J.X.L., figures and tables; J.X.L. and J.L., Revision. All authors have read and agreed to the published version of the manuscript.

Peer review

Peer review information

Nature Communications thanks Margaret Sunde, and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

NMR assignments for the RHIM domains of mRIPK1 have been deposited in the BMRB database under accession number 36547. The structure data used in this study are available in the Protein Data Bank under accession code 8IB0. Previously reported structures retrieved from PDB include mouse RIPK3 (6JPD) and human RIPK3 (7DA4). The protein sequence used in this study is available from the UniProt database under accession code Q60855 and Q9QZL0. Additional data related to this paper are available from the corresponding author upon request. Source data are provided with this paper.

Code availability

The Python scripts used in this study along with the relevant data are available on request from Jun-xia Lu.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-024-51303-y.