Structure-guided de novo design of DELE1 binders targeting the α1 oligomerization interface.

Schematic overview of the de novo protein design pipeline used to generate DELE1 binders. A monomeric DELE1 C-terminal domain (DELE1CTD) subunit (red) was used as a template to target the α1 helix that mediates oligomerization in the oligomeric DELE1CTD cryo-EM structure (PDB: 8D9X). RFdiffusion was applied to generate binder backbone scaffolds through iterative denoising steps, starting from Gaussian noise and converging on structured helical scaffolds positioned against the DELE1 α1 helix surface. Candidate backbones were subsequently subjected to sequence optimization using ProteinMPNN. The resulting DELE1–binder complexes were evaluated using AlphaFold2 and AlphaFold3 structure prediction, and designs were filtered based on prediction confidence metrics, including binder RMSD relative to the starting backbone, per-residue confidence (pLDDT), and interface pairwise aligned error (PAE). High-confidence designs were selected for experimental validation.

Designed binders directly associate with DELE1 and disrupt its oligomeric state in vitro.

(A, C, E, G, I) Size-exclusion chromatography (SEC) profiles of MBP-DELE1 co-expressed with individual 6His-tagged binders (binder1, binder2, binder3, binder5, and binder11). Elution profiles show the DELE1–binder complex (blue trace), oligomeric DELE1 species (gray trace), and excess unbound binder (black trace). Elution volumes are indicated on the x-axis. (B, D, F, H, J) Coomassie-stained SDS–PAGE analysis of SEC fractions corresponding to the major complex peaks shown in panels A, C, E, G, and I. Fractions from the dominant SEC peak contain both MBP-DELE1 and the corresponding binder, confirming stable complex formation. Lanes labeled “excess binder” contain unbound binder alone. Across all binders shown, co-expression with DELE1 resulted in formation of a homogeneous DELE1–binder complex that eluted at later volumes relative to oligomeric DELE1 alone, consistent with reduced higher-order assembly of DELE1 in vitro.

Designed binders inhibit mitochondrial stress-induced ATF4 activation.

(A) Experimental schematic illustrating the timeline for binder expression and stress treatment. HEK293T cells were transfected with GFP-tagged binders for 24 hours, followed by treatment with mitochondrial stressors for 16 hours prior to harvest. (B) Immunoblot analysis of ATF4 levels in HEK293T cells expressing GFP-tagged binders and treated with CCCP. Binder expression was monitored by GFP immunoblotting, and β-actin served as a loading control. (C) Quantification of ATF4 levels from CCCP-treated samples shown in panel B, normalized to β-actin and expressed relative to mock-transfected cells. Data represent mean ± s.d. from 2 independent experiments. (D) Immunoblot analysis of ATF4 levels in HEK293T cells expressing GFP-tagged binders and treated with oligomycin. Binder expression was detected by GFP immunoblotting, with β-actin as a loading control. (E) Quantification of ATF4 levels from oligomycin-treated samples shown in panel D, normalized to β-actin and expressed relative to mock-transfected cells. Red arrows indicate binders that produced marked suppression of ATF4 induction. DMSO represents cells treated with DMSO and serves as the blank control, rather than treatment with CCCP or oligomycin. Mock represents cells treated with CCCP or oligomycin as a positive control; however, no binders were transfected or overexpressed in these cells. Data represent mean ± s.d. from 2 independent experiments.

Delayed restoration of mitochondrial network morphology following transient stress in binder-expressing cells.

(A) Representative fluorescence microscopy images of mitochondrial morphology during recovery following transient CCCP treatment. U2OS cells expressing a mitochondria-targeted pre-CoxIV–BFP reporter alone (Veh) or together with binder1-GFP or binder5-GFP were treated with CCCP for 2 hours, followed by washout and recovery for the indicated times (4h, 8h, and 24h). Insets show 4× magnified views of the regions outlined by red boxes. Scale bars, 20 μm. (B) Time course quantification of mitochondrial morphology recovery following CCCP treatment. The percentage of cells exhibiting elongated mitochondrial networks was quantified at baseline (0 h), after 2h CCCP treatment, and at the indicated times (4h, 8h, 24h) following CCCP removal. Control cells (Veh) rapidly restored elongated mitochondrial morphology, whereas cells expressing binder1 or binder5 showed delayed recovery at early time points. Data represent mean ± s.d. from three independent experiments. (C) Quantification of the percentage of cells with elongated mitochondria 4h after CCCP removal for the conditions shown in panel A. Bars represent mean ± s.d. from three independent experiments. (D) Quantification of the percentage of cells with elongated mitochondria 8 h after CCCP removal for the conditions shown in panel A. Bars represent mean ± s.d. from three independent experiments. Statistical significance was assessed using one-way ANOVA, with significance levels indicated. Veh represents cells treated with as a positive control; however, no binders were transfected or overexpressed in these cells. **: <0.01, *:<0.05, ns: not significant.

Structural and mutational validation of the DELE1–binder interaction interface.

(A) Crystal structure of binder5 determined at 2.6 Å resolution, shown in two orientations related by a 180° rotation. The structure adopts a compact four-helix bundle fold, with N- and C-termini indicated. (B) Superposition of the binder5 crystal structure (teal) with the design model (purple), showing close agreement between the experimental structure and the predicted fold (RMSD = 1.5 Å). (C) Designed model of binder1 (purple) bound to the DELE1 α1 helix (blue). Side chains of key hydrophobic residues on DELE1 (L229, F240, and F250) and interacting residues on binder1 are shown and labeled. Dashed lines indicate predicted side-chain interactions. (D) Designed model of binder5 (purple) bound to the DELE1 α1 helix (blue). Residues forming the predicted hydrophobic interaction interface on both DELE1 and binder5 are shown and labeled. (E) Affinity pulldown analysis of MBP-DELE1 (wild-type or L229E/F240E/F250E triple mutant, “3E”) co-expressed with 6His-tagged binder1. Ni–NTA pulldown via the binder tag (left) and reciprocal amylose pulldown via MBP-DELE1 (right) show robust co-purification with wild-type DELE1 but loss of interaction with the 3E mutant. (F) Affinity pulldown analysis of MBP-DELE1 (wild-type or 3E) co-expressed with 6His-tagged binder5, performed as in panel E. Binder5 co-purifies with wild-type DELE1 but not with the 3E mutant.

Designed binders do not disrupt the DELE1-HRI interaction.

(A) SEC analysis of MBP–TEV–HRINTD alone (gray trace), the purified DELE1–binder1 complex (black trace), and the mixture of DELE1–binder1 with MBP–TEV–HRINTD (blue trace). Upon mixing, the DELE1–binder1 peak is reduced and replaced by earlier-eluting high–molecular weight species corresponding to DELE1–HRI complexes, while free binder1 appears as a later-eluting peak. (B) Coomassie-stained SDS–PAGE analysis of SEC fractions from the mixture (blue trace) in panel A. Fractions corresponding to the early-eluting peaks contain both MBP–DELE1-CTD and MBP–TEV-HRINTD, confirming formation of DELE1–HRI complexes. Binder1 is absent from these fractions and detected only in later fractions corresponding to displaced binder. (C) SEC analysis performed as in panel A using the DELE1–binder5 complex. Mixing with MBP–TEV–HRINTD similarly yields early-eluting DELE1–HRI species and displacement of binder5. (D) SDS–PAGE analysis of SEC fractions from the mixture (blue trace) in panel C confirming the formation of DELE1-HRI complexes and the displaced binder5.

Crystallography data collection and refinement statistics.

Values in parentheses are data for the highest resolution shell.

Structural rationale for targeting the DELE1 oligomerization interface using de novo protein design.

(A) Surface representation of the oligomeric DELE1CTD (PDB: 8D9X) assembly highlighting the oligomerization interface. Individual DELE1 subunits are colored distinctly to illustrate the multimeric architecture. Designed binders were engineered to engage the α1 helix at the oligomerization interface, thereby stabilizing a monomeric DELE1CTD species. (B) Ribbon representation of a monomeric DELE1CTD subunit highlighting hydrophobic hotspot residues on the α1 helix (F240, L242, F250, and L251) selected for binder targeting. These residues form a contiguous interaction surface that is buried in the oligomeric assembly and were used to guide de novo binder design. (C–D) AlphaFold3-predicted structures of the DELE1–binder1 (C) and DELE1–binder5 (D) complexes. Left, structural models are shown in two orientations related by a 180° rotation. Structures are colored according to per-residue pLDDT confidence scores (scale shown below; red, low confidence; teal, high confidence). Right, PAE matrices for each complex. Low PAE values (blue) indicate high confidence in the relative positioning between residues, whereas higher PAE values (red) indicate increased uncertainty.

Co-expression and Ni–NTA pulldown of DELE1 with designed binders.

(A) Schematic of the bacterial co-expression constructs. MBP-tagged DELE1CTD was co-expressed with individual 6His-tagged binders, enabling affinity purification through the binder tag. (B) Coomassie-stained SDS–PAGE analysis of Ni–NTA pulldown experiments for representative binders (binder4, binder5, and binder7). Lanes show soluble lysate (Sup), insoluble fraction (Pellet), unbound material (Unbound), and eluted fractions (Elution). MBP-DELE1CTD co-purifies with the corresponding 6His-tagged binders, indicating stable complex formation.

Additional designed binders form stable complexes with DELE1 in vitro.

(A, C, E, G, I, K) Size-exclusion chromatography profiles of MBP-DELE1 co-expressed with additional 6His-tagged binders (binder4, binder6, binder7, binder8, binder9, and binder10). Elution profiles show the DELE1–binder complex (blue trace), oligomeric DELE1 species (gray trace), and excess unbound binder (black trace). Elution volumes are indicated on the x-axis. (B, D, F, H, J, L) Coomassie-stained SDS–PAGE analysis of SEC fractions corresponding to the major SEC peaks shown in panels A, C, E, G, I, and K. Fractions from the dominant peak contain both MBP-DELE1 and the corresponding binder, indicating complex formation. For binder10, co-elution with DELE1 was reduced relative to other binders, indicating a weaker complex formation. Lanes labeled “excess binder” contain unbound binder carried through affinity purification. These data extend the results shown in Figure 2 and demonstrate that multiple independently designed binders associate with DELE1 and form homogeneous complexes in vitro.

Designed binders directly interact with DELE1 in cells and do not suppress endoplasmic reticulum stress induced ATF4.

(A) Co-immunoprecipitation analysis of binder association with DELE1 in HEK293T cells. FLAG-tagged DELE1-GFP was co-expressed with GFP, binder1-GFP, or binder5-GFP. Whole-cell lysates (Input) and anti-FLAG immunoprecipitants (IP: FLAG) were analyzed by immunoblotting for DELE1 and GFP-tagged binders. (B) Immunoblot analysis of ATF4 levels in HEK293T cells expressing GFP-tagged binders and treated with thapsigargin for 6 hrs. Binder expression was monitored by GFP immunoblotting, and β-actin served as a loading control. (C) Quantification of ATF4 levels from CCCP-treated samples shown in panel B, normalized to β-actin and expressed relative to mock-transfected cells. Data represent mean ± s.d. from 2 independent experiments. DMSO represents cells treated with DMSO and serves as the blank control, rather than treatment with CCCP or oligomycin. Mock represents cells treated with CCCP or oligomycin as a positive control; however, no binders were transfected or overexpressed in these cells.

Mitochondrial stress triggers fragmented network morphology.

Representative fluorescence microscopy images of mitochondrial morphology in U2OS cells before and after CCCP treatment. Cells were transiently transfected with BFP-tagged pre-CoxIV to label mitochondria and, where indicated, co-expressed with binder1–GFP or binder5–GFP. (A) Under untreated conditions, control cells and cells expressing binder1 or binder5 display an interconnected, elongated mitochondrial network. (B) Following 2 hours of CCCP treatment, mitochondria undergo robust fragmentation in all conditions, regardless of binder expression. For each condition, a higher-magnification view of the boxed region (red) is shown to the right. Scale bars, 20 µm. Veh represents cells treated with as a positive control; however, no binders were transfected or overexpressed in these cells. **: <0.01, *:<0.05, ns: not significant.

Structural validation of binder5 and biochemical characterization of the DELE1 3E mutant.

(A) Representative electron density map (2Fo–Fc) for binder5 contoured around the refined atomic model, shown in two orientations related by a 90° rotation. Side chains of selected residues are labeled to illustrate the quality of density. Side-chain conformations from the crystal structure (teal) are overlaid with those from the design model (purple) to highlight local differences between the experimental structure and the predicted model. (B) SEC profiles of MBP-DELE1 wild-type (gray trace) and the L229E/F240E/F250E triple mutant (“3E”, blue trace). The 3E mutant elutes at a later volume relative to wild-type DELE1, indicating a shift toward a smaller apparent molecular weight species. (C) Coomassie-stained SDS–PAGE analysis of SEC peak fractions for MBP-DELE1 wild-type. (D) Coomassie-stained SDS–PAGE analysis of SEC peak fractions for the MBP-DELE1 3E mutant.

Direct interaction between DELE1 and HRI-NTD is preserved in the presence of designed binders.

(A) SEC profile of purified MBP–TEV–HRINTD, showing a homogeneous, early-eluting peak. (B) Coomassie-stained SDS–PAGE analysis of SEC fractions from panel A, confirming purity of MBP–TEV-HRINTD. (C) SEC profile of the DELE1–HRI complex following TEV protease treatment. Two major high–molecular weight species are observed, corresponding to complexes containing MBP–DELE1 and cleaved HRINTD. (D) SDS–PAGE analysis of SEC fractions from panel C, showing co-elution of MBP–DELE1 and cleaved HRINTD. The cleaved MBP was eluted in later fractions. (E) SEC analysis of the pre-formed DELE1–HRI complex incubated with excess binder1. The DELE1–HRI complex remains intact, while excess binder1 elutes as a separate low–molecular weight peak. (F) SDS–PAGE analysis of SEC fractions from panel E, confirming retention of the DELE1–HRI complex and absence of binder1 in the complex fractions. (G) SEC analysis of the pre-formed DELE1–HRI complex incubated with excess binder5, showing a stable DELE1–HRI complex and a separate low–molecular weight peak corresponding to excess binder5. (H) SDS–PAGE analysis of SEC fractions from panel G, confirming that binder5 does not disrupt the DELE1–HRI complex.