De novo designed Hsp70 activator dissolves intracellular condensates

Abstract

Protein quality control (PQC) is carried out in part by the chaperone Hsp70, in concert with adapters of the J-domain protein (JDP) family. The JDPs, also called Hsp40s, are thought to recruit Hsp70 into complexes with specific client proteins. However, the molecular principles regulating this process are not well understood. We describe the de novo design of a set of Hsp70 binding proteins that either inhibited or stimulated Hsp70's ATPase activity; a stimulating design promoted the refolding of denatured luciferase in vitro, similar to native JDPs. Targeting of this design to intracellular condensates resulted in their nearly complete dissolution. The designs inform our understanding of chaperone structure-function relationships and provide a general and modular way to target PQC systems to condensates and other cellular targets.

Extended Data Figure 1:. Computational and experimental pipeline for designing JDMs.

a, Schematic of the exact computational workflow for designing fully de novo and partially de novo JDMs. b, Summary of the experimental (yeast display) pipeline for high throughput testing computationally designed JDMs. Round 1: computer generated designs were tested through three rounds of yeast display sorting (see Methods). Round 2: The 4 fully de novo JDMs were enriched and each of these JDMs underwent an SSM with three rounds of yeast display sorting. Round 3: Beneficial mutations identified through SSM were incorporated into combinatorial mutant libraries and underwent three rounds of yeast display sorting. At the end, the collected yeast cells were sequenced and the top 41 JDM designs were expressed, purified, and tested biochemically. c, The top 41 JDM designs come from two parents originally designed in Round 1. Shown are the two parent’s SSM maps. Each cell is a particular point mutant. Red indicates tight affinity to Hsc70 and blue indicates weak affinity to Hsc70. d, The top 41 JDM designs were transformed into yeast and were tested for specificity to ATP-loaded state of Hsc70 and binding to same site as native J-domains through DnaJB1 competition assays. Shown are the flow cytometry plots.

Extended Data Figure 2:. Sequences of top 41 JDM designs.

Sequence alignment of the top 41 JDM designs after yeast display.

Extended Data Figure 3:. Production and purification of top 41 JDM designs.

The top 41 JDM designs were transformed into E. coli., induced for expression, and purified. SEC traces of the top 41 JDM designs for further purification and characterization for their size and monodispersity.

Extended Data Figure 4:. Measuring the binding of JDMs to Hsc70.

a-b, A subset of the top 41 JDM designs (high protein yield and monodispersity) were tested in Hsc70 ATP turnover were also tested for their binding to ATP-loaded Hsc70 via BLI. Shown is the average steady-state binding titration of all the binders tested (n=3 experiments, line is EC50 curve fitting) (a) or a representative raw kinetic binding data (b).

Extended Data Figure 5:. Testing JDMs effect on Hsc70 ATP turnover and refolding.

a, A subset of the top 41 JDM designs with sufficient protein after purification and monodisperse SEC traces were tested in Hsc70 ATP turnover assays. Native DnaJA2 (Hsp40) was tested as a positive control. Lines represent average of fluorescence quenching (n=2 experiments for JDM1–5, the rest n=3 experiments). b, JDM37, the only JDM that activated Hsc70 ATP turnover, was tested for luciferase refolding. Native DnaJA2 (Hsp40) was tested as a positive control. Lines represent average luminescence recorded (n=3 experiments).

Extended Data Figure 6:. Measuring the effect JDMs have on E. coli. cell growth.

a-b, The top 41 final JDM designs were transformed into E. coli., induced for expression, grown at 42°C, and their cell growth was tracked. Shown is a representative cell growth curve (a) or the average maximum growth velocity (b) (n=5 experiments). Dotted line is for empty vector expression as control. (c) The JDMs with the most consistent and dramatic inhibition of E. coli growth at 42°C were tested for inhibition of DnaJA2 (100nM) activation of Hsc70. IC50s are shown on the bottom with the units of μM.

Extended Data Figure 7:. Measuring the binding of JDM-based condensate perturbators to Hsc70.

Representative raw kinetic BLI data of JDM37-based or JDM16-based condensate perturbators binding to ATP-loaded Hsc70. JDM37 and JDM16 alone are displayed as comparisons. Concentration used for each protein displayed here is 729nM.

Extended Data Figure 8:. Fluorescence metrics of condensate perturbators expressed in mammalian cells.

a-b, Average mCherry (a) or TSapphire (b) fluorescence intensity of different constructs tested in HEK293T cells. Each point represents a single cell (n=20 cells). c, Scatterplot comparing Tsapphire fluorescence intensity (condensate perturbator expression) to number of RIα puncta per cell. d-e, Time-course imaging of HEK293T cells expressing mCherry-tagged RIα, either cAMP sensor ICUE3 (d) or PKA sensor AKAR4 (e), and either JDM37-based or JDM16-based condensate perturbators. Also no transfection and mCherry tagged RIα were done as controls. In each condition, 10nM isoproterenol was added. Solid lines indicate representative average time with error bars representing standard error mean (SEM) (n=at least 15 cells per curve). f-g, Average mCherry (f) or TSapphire (g) fluorescence intensity of different constructs tested in Beas2B cells. Each point represents a single cell (n=20 cells). h, Scatterplot comparing Tsapphire fluorescence intensity (condensate perturbator expression) to number of EML4-Alk puncta per cell.

Extended Data Figure 9:. Exploring the sequence and structural determinants for activating Hsp70.

a, AlphaFold2 predictions of the top 41 JDMs in complex with Hsp70. JDM37 is solid green while the remaining JDMs are transparent. b, Multiple viewpoints of either native J-domain:Hsp70 (cyan:yellow) (PDB: 5NRO) or JDM37:Hsp70 complex (blue:green). c, Sequence alignment of native J-domain from DnaJ versus JDM37. d, Clustal alignment of all human native J-domains. e, Various maps of either native J-domain:Hsp70 (top) or JDM37:Hsp70 (bottom) complex. Left: ribbon representation, Middle: electrostatic map of native J-domain or JDM37, Right: electrostatic map of Hsp70.

Figure 1:. De novo design of Hsp70 binders/activators.

a, Structures of Hsp70 with either native or designed proteins. Left: Crystal structure of DnaK (Hsp70) with DnaJ (Hsc40) (PDB: 5NRO). Middle: AlphaFold structure prediction of DnaK with fully de novo protein designed to bind to the same region as DnaJ. Right: AlphaFold structure prediction of DnaK with partially redesigned DnaJ. b, Design and biochemical characterization of fully de novo designed J-domain mimics (JDMs). Left: AlphaFold complex prediction of selected JDM with DnaK. Middle: Representative trace of biolayer interferometry (BLI) measurements of JDM binding to either ATP or ADP-loaded Hsc70. Right: In vitro assays measuring ATP turnover by Hsc70 with either native DnaJA2 (Hsp40) or JDMs. Lines represent average of fluorescence quenching (n=3 experiments). Bottom: Sequence alignment of the two JDMs presented here. c, In vitro assays measuring the refolding of denatured luciferase where Hsc70 and either native DnaJA2 (Hsp40) or JDM37 fused to DnaJ without its native J-domain (JDM37-DnaJ) was added. Lines represent average luminescence recorded (n=3 experiments).

Figure 2:. Fusing de novo designed JDMs with substrate binding domains dissolves intracellular condensates.

a, Schematic of strategy to dissolve RIα condensates. Protein that drives condensation (here, RIα) is tethered with mCherry. JDM37 is fused to an mCherry nanobody and mT-Sapphire, and this tool is called condensate perturbator. b, Expression of engineered condensate perturbators dissolves mCherry-tagged RIα puncta in HEK293T cells (see Methods for details). Top: Representative epifluorescence images of the various conditions tested. Scale bar = 10μm. Bottom: Quantification of number of RIα puncta per cell. Each point represents a single cell (n=20 cells). c-d, Time-course imaging of HEK293T cells expressing mCherry-tagged RIα, either cAMP sensor ICUE3 (c) or PKA sensor AKAR4 (d), and either JDM37-based or JDM16-based condensate perturbators. In each condition, 10nM isoproterenol was added. Solid lines indicate representative average time with error bars representing standard error mean (SEM) (n=at least 15 cells per curve). e, Schematic of strategy to dissolve EML4-Alk oncogenic condensates. Protein that drives condensation (here, EML4-Alk) is tethered with mCherry. f, Expression of engineered condensate perturbators dissolves mCherry-tagged EML4-Alk puncta in Beas2B cells (see Methods for details). Top: Representative epifluorescence images of the various conditions tested. Scale bar = 10μm. Bottom: Quantification of number of EML4-Alk puncta per cell. Each point represents a single cell (n=20 cells). g, Raw FRET ratios of HEK293T cells expressing mCherry-tagged RIα, Ras sensor Ras-LOCKR-S, and either JDM37-based or JDM16-based condensate perturbators. Each point represents a single cell (n=18 cells). h, Cell growth curves of Beas2B cells expressing mCherry-tagged EML4-Alk with or without condensate perturbators (n=3 experiments). Line represents average from all 3 experiments. For the quantification of number of puncta per cell in b and f, cells only with sufficient expression of the condensate perturbator were chosen for analysis (see Methods for details).