Immunomodulatory drugs bind to cereblon (CRBN) to confer differentiated substrate specificity on the CRL4CRBN E3 ubiquitin ligase. Here we report the identification of a new cereblon modulator, CC-885, with potent anti-tumour activity. The anti-tumour activity of CC-885 is mediated through the cereblon-dependent ubiquitination and degradation of the translation termination factor GSPT1. Patient-derived acute myeloid leukaemia tumour cells exhibit high sensitivity to CC-885, indicating the clinical potential of this mechanism. Crystallographic studies of the CRBN–DDB1–CC-885–GSPT1 complex reveal that GSPT1 binds to cereblon through a surface turn containing a glycine residue at a key position, interacting with both CC-885 and a ‘hotspot’ on the cereblon surface. Although GSPT1 possesses no obvious structural, sequence or functional homology to previously known cereblon substrates, mutational analysis and modelling indicate that the cereblon substrate Ikaros uses a similar structural feature to bind cereblon, suggesting a common motif for substrate recruitment. These findings define a structural degron underlying cereblon ‘neosubstrate’ selectivity, and identify an anti-tumour target rendered druggable by cereblon modulation.
This is a preview of subscription content, access via your institution
Open Access articles citing this article.
Molecular Cancer Open Access 11 April 2022
Nature Communications Open Access 10 February 2022
A proximity biotinylation-based approach to identify protein-E3 ligase interactions induced by PROTACs and molecular glues
Nature Communications Open Access 10 January 2022
Subscribe to Journal
Get full journal access for 1 year
only $3.90 per issue
All prices are NET prices.
VAT will be added later in the checkout.
Tax calculation will be finalised during checkout.
Get time limited or full article access on ReadCube.
All prices are NET prices.
Ito, T. et al. Identification of a primary target of thalidomide teratogenicity. Science 327, 1345–1350 (2010)
Dimopoulos, M. A., Richardson, P. G., Moreau, P. & Anderson, K. C. Current treatment landscape for relapsed and/or refractory multiple myeloma. Nat. Rev. Clin. Oncol. 12, 42–54 (2015)
Lopez-Girona, A. et al. Cereblon is a direct protein target for immunomodulatory and antiproliferative activities of lenalidomide and pomalidomide. Leukemia 26, 2326–2335 (2012)
Zhu, Y. X. et al. Cereblon expression is required for the antimyeloma activity of lenalidomide and pomalidomide. Blood 118, 4771–4779 (2011)
Krönke, J. et al. Lenalidomide induces ubiquitination and degradation of CK1α in del(5q) MDS. Nature 523, 183–188 (2015)
Angers, S. et al. Molecular architecture and assembly of the DDB1–CUL4A ubiquitin ligase machinery. Nature 443, 590–593 (2006)
Higa, L. A. et al. CUL4–DDB1 ubiquitin ligase interacts with multiple WD40-repeat proteins and regulates histone methylation. Nat. Cell Biol. 8, 1277–1283 (2006)
Jin, J., Arias, E. E., Chen, J., Harper, J. W. & Walter, J. C. A family of diverse Cul4-Ddb1-interacting proteins includes Cdt2, which is required for S phase destruction of the replication factor Cdt1. Mol. Cell 23, 709–721 (2006)
He, Y. J., McCall, C. M., Hu, J., Zeng, Y. & Xiong, Y. DDB1 functions as a linker to recruit receptor WD40 proteins to CUL4–ROC1 ubiquitin ligases. Genes Dev. 20, 2949–2954 (2006)
Chamberlain, P. P. et al. Structure of the human Cereblon–DDB1–lenalidomide complex reveals basis for responsiveness to thalidomide analogs. Nat. Struct. Mol. Biol. 21, 803–809 (2014)
Fischer, E. S. et al. Structure of the DDB1–CRBN E3 ubiquitin ligase in complex with thalidomide. Nature 512, 49–53 (2014). 10.1038/nature13527
John, L. B. & Ward, A. C. The Ikaros gene family: transcriptional regulators of hematopoiesis and immunity. Mol. Immunol. 48, 1272–1278 (2011)
Dijon, M. et al. The role of Ikaros in human erythroid differentiation. Blood 111, 1138–1146 (2008)
Gandhi, A. K. et al. Immunomodulatory agents lenalidomide and pomalidomide co-stimulate T cells by inducing degradation of T cell repressors Ikaros and Aiolos via modulation of the E3 ubiquitin ligase complex CRL4(CRBN.). Br. J. Haematol. 164, 811–821 (2014)
Krönke, J. et al. Lenalidomide causes selective degradation of IKZF1 and IKZF3 in multiple myeloma cells. Science 343, 301–305 (2014)
Lu, G. et al. The myeloma drug lenalidomide promotes the cereblon-dependent destruction of Ikaros proteins. Science 343, 305–309 (2014)
Cheng, Z. et al. Structural insights into eRF3 and stop codon recognition by eRF1. Genes Dev. 23, 1106–1118 (2009)
Preis, A. et al. Cryoelectron microscopic structures of eukaryotic translation termination complexes containing eRF1-eRF3 or eRF1-ABCE1. Cell Reports 8, 59–65 (2014)
Zhouravleva, G. et al. Termination of translation in eukaryotes is governed by two interacting polypeptide chain release factors, eRF1 and eRF3. EMBO J. 14, 4065–4072 (1995)
Hon, W. C. et al. Structural basis for the recognition of hydroxyproline in HIF-1 alpha by pVHL. Nature 417, 975–978 (2002)
Min, J. H. et al. Structure of an HIF-1α-pVHL complex: hydroxyproline recognition in signaling. Science 296, 1886–1889 (2002)
Glotzer, M., Murray, A. W. & Kirschner, M. W. Cyclin is degraded by the ubiquitin pathway. Nature 349, 132–138 (1991)
Sheard, L. B. et al. Jasmonate perception by inositol-phosphate-potentiated COI1–JAZ co-receptor. Nature 468, 400–405 (2010)
Tan, X. et al. Mechanism of auxin perception by the TIR1 ubiquitin ligase. Nature 446, 640–645 (2007)
Kikuchi, Y., Shimatake, H. & Kikuchi, A. A yeast gene required for the G1-to-S transition encodes a protein containing an A-kinase target site and GTPase domain. EMBO J. 7, 1175–1182 (1988)
Chauvin, C., Salhi, S. & Jean-Jean, O. Human eukaryotic release factor 3a depletion causes cell cycle arrest at G1 phase through inhibition of the mTOR pathway. Mol. Cell. Biol. 27, 5619–5629 (2007)
Deshaies, R. J. Protein degradation: prime time for PROTACs. Nat. Chem. Biol. 11, 634–635 (2015)
Winter, G. E. et al. Drug development. Phthalimide conjugation as a strategy for in vivo target protein degradation. Science 348, 1376–1381 (2015)
Lu, J. et al. Hijacking the E3 ubiquitin ligase cereblon to efficiently target BRD4. Chem. Biol. 22, 755–763 (2015)
Petzold, G., Fischer, E. S. & Thomä, N. H. Structural basis of lenalidomide-induced CK1α degradation by the CRL4(CRBN) ubiquitin ligase. Nature 532, 127–130 (2016)
McCoy, A. J. et al. Phaser crystallographic software. J. Appl. Crystallogr. 40, 658–674 (2007)
Emsley, P., Lohkamp, B., Scott, W. G. & Cowtan, K. Features and development of Coot. Acta Crystallogr. D 66, 486–501 (2010)
Murshudov, G. N. et al. REFMAC5 for the refinement of macromolecular crystal structures. Acta Crystallogr. D 67, 355–367 (2011)
Bennett, T. A. et al. Pharmacological profiles of acute myeloid leukemia treatments in patient samples by automated flow cytometry: a bridge to individualized medicine. Clin. Lymphoma Myeloma Leuk. 14, 305–318 (2014)
Thanks to G. Reyes, C. Havens, P. Jackson and H. Hadjivassiliou for discussions relating to this manuscript, and P. Jackson, J. Hansen, M. Correa, B. Fahr, M. Abbasian, E. Ambing, E. Rychak, D. Mendy and K. Hughes for technical assistance. We thank K. Motamedchaboki (Proteomics Core, Sanford Burnham Prebys Medical Discovery Institute) for mass spectrometry-based proteomic analysis. Thanks to J. Ballesteros and P. Hernandez at Vivia for patient sample testing. Parts of this work were conducted at the Advanced Light Source. The Berkeley Center for Structural Biology is supported in part by the National Institutes of Health, National Institute of General Medical Sciences, and the Howard Hughes Medical Institute. The Advanced Light Source is supported by the Director, Office of Science, Office of Basic Energy Sciences, of the US Department of Energy under contract No. DE-AC02-05CH11231. This work was also supported by Grant-in-Aid for Scientific Research on Innovative Areas “Chemical Biology of Natural Products” from the Ministry of Education, Culture, Sports, Science and Technology (MEXT) (23102002 to H.H.), by JST, PRESTO (T.I.) and by Grant-in-Aid for Young Scientists (B) from MEXT (26750374 to T.I.).
All authors are or have been employees or collaborators of Celgene.
Extended data figures and tables
a, The growth inhibitory IC50 value of CC-885 in human cancer cell lines. The cell proliferation of 132 human cancer cell lines treated with varying concentrations of CC-885 for 3 days was assessed by Cell Titre Glo assay. Cancer types as indicated at the bottom are labelled with different colours. The IC50 value of growth inhibition was determined using ActivityBase (IDBS). Data are the mean of two or more biological replicates. The n number for each cell line is shown on the right of the panel. b, The effect of CC-885 on AML samples taken from patients. The top panel shows the effects on leukaemic cells; the bottom panel shows the effects on normal lymphoid cells. The sample from patient B did not contain sufficient normal lymphoid cells for analysis. c, The effect of CC-885 on cell proliferation in parental and CRBN−/− 293FT HEK cells. Result is representative of three biological replicates. Data are shown as mean ± s.d., n = 3 technical replicates. d, The effect of lenalidomide, pomalidomide and CC-885 on cell proliferation in AML cell lines, THLE-2 and human PBMC. Cells were treated with varying concentrations of lenalidomide, pomalidomide or CC-885. At day 3, cell proliferation was assessed using CTG assay. Numbers shown are the growth inhibitory IC50 values of lenalidomide, pomalidomide and CC-885, from biological replicates with n = 5, 3, and 20, respectively.
Extended Data Figure 2 CC-885, but not lenalidomide, promotes the interaction between GSPT1 and CRBN in vitro.
a, b, Immunoblot analysis of anti-HA immunoprecipitates (a) and whole-cell lysates (WCL) (b). HA-tagged GSPT1 or IKZF1 produced in 293FT CRBN−/− HEK cells were used to capture CRBN from 293FT HEK cells expressing GSPT1-specific shRNA, shGPST1-1. DMSO, 10 μM lenalidomide or 10 μM CC-885 were included in the binding assay. The IKZF1 Q146H mutation results in a reduction of cereblon binding mediated by either lenalidomide or CC-885. Some residual binding is observed with CC-885, consistent with Fig. 1d, which shows that CC-885 is more potent than lenalidomide against IKZF1. c, Coomassie stain of CRBN–DDB1 pull-down with GSPT1 using purified components. Purified GST or GST–GSPT1 domains 2 and 3 (amino acids 437–633) bound to magnetic glutathione beads was incubated with purified CRBN–DDB1 in the presences of CC-885, lenalidomide (len), pomolidamide (pom), or DMSO (vehicle) for 1 h at room temperature before three rapid washes. CC-885, but not lenalidomide, pomalidomide or DMSO, mediated the binding of GST–GSPT1 to CRBN–DDB1. This experiment was performed three times. For gel source data, see Supplementary Information Fig. 1.
a, Ubiquitinated protein products, as shown in Fig. 1e, were blotted for GSPT1 with mouse anti-GSPT1 monoclonal antibody 8A2, CRBN, and His-tagged Ubiquitin. Arrowheads mark nonspecific bands. b, 293FT parental and CRBN−/− cells transiently expressing His-tagged ubiquitin was treated with CC-885 and MG132 as indicated for 8 h. Ubiquitinated protein products were pulled down with Ni-NTA agarose under denaturing conditions, followed by elution via boiling in SDS loading buffer. Eluates were diluted with immunoprecipiation buffer and GSPT1 was immunoprecipitated with a mouse anti-GSPT1 monoclonal antibody 17G9. Immunoprecipitates were then subjected to immunoblot analysis for GSPT1 and ubiquitin with two different rabbit anti-ubiquitin antibodies. c, Immunoblot analysis of 293FT parental and CRBN−/− cells treated with CC-885 for 4 h. Cells were pretreated with DMSO, CC-885, MLN4924 or MG132 as indicated. d, 293FT parental or CRBN−/− cells treated with DMSO or CC-885 for 24 h were subjected to real-time qPCR (top) and immunoblot analysis (bottom). Result is representative of two biological replicates. Data are mean ± s.d., n = 3 technical replicates. For gel source data, see Supplementary Information Fig. 1.
a, Controls for the in vitro ubiquitination assay shown in Fig. 1f. Recombinant protein products as indicated were incubated with or without ATP (10 mM) and CC-885 (100 μM) in the ubiquitination assay buffer at 30 °C for 2 h, and then separated by SDS–PAGE followed by Coomassie staining or immunoblot analysis with anti-GSPT1, anti-cereblon, and anti-ubiquitin antibodies. The anti-GSPT1 blot shown is the same one shown in Fig. 1f. The additional blots and Coomassie staining analysis was performed on samples from the same in vitro ubiquitination reactions. Anti-GSPT1 blotting and Coomassie staining was performed three times, anti-cereblon and anti-ubiquitin blotting was performed twice. b–d, To demonstrate the CRBN-dependence of GSPT1 ubiquitination by CC-885-CRL4, we reconstituted DDB1 with an alternative DCAF, SV5-V, and showed that SV5-V is not capable of recruiting GSPT1 for polyubiquitination. b, DDB1 binds to GST–SV5-V, but not GST alone. Coomassie stain with lanes 1, 2, and 3 showing individual proteins, and lanes 4 and 5 showing pull-down of DDB1 incubated with GST or GST–SV5-V bound to glutathione magnetic beads and washed three times. This experiment was performed twice. c, GST–SV5-V–DDB1 protein complex generated by pull-down in b was used to show binding of CUL4–RBX1 to the GST–SV5-V–DDB1 protein complex but not GST alone. Coomassie stain with lanes 1, 2, and 3 showing individual proteins or protein complexes used in the pull-down, and lanes 4 and 5 showing pull-down of CUL4–RBX1 incubated with GST or GST–SV5-V–DDB1 bound to glutathione magnetic beads and washed three times. This experiment was performed once. d, In vitro ubiquitination of GSPT1 by CRBN–DDB1 but not SV5-V–DDB1. SV5-V–DDB1 complex was formed by incubation of individually purified proteins and purification over size-exclusion chromatography. Recombinant protein products as indicated were incubated with either CRBN–DDB1, SV5-V–DDB1, DDB1 alone, or the absence of any DDB1 complex, with and without ATP (10 mM) or CC-885 (100 μM) in ubiquitination assay buffer at 30 °C for 2 h, and then separated by SDS–PAGE followed by immunoblot analysis with anti-GSPT1 and anti-SV5-V antibodies. Asterisk indicates background bands present from the SV5-V protein purification. The anti-SV5-V western blot indicates that SV5-V is auto-ubiquitinated, as expected for a functional DCAF that is bound to the DDB1–CUL4–RBX1 complex. e, Immunoblot analysis of 293FT parental and CRBN−/− cells treated with 100 μg ml−1 cyclohexamide (CHX) for the indicated periods. Where indicated, cells were pretreated with DMSO or 10 μM CC-885 for 30 min. This experiment was performed once. f, Immunoblot analysis of 293FT parental cells, CRBN−/− cells, and CRBN−/− cells stably expressing CRBN wild-type or variants as indicated. CRBNiso2 (CRBN isoform 2) showed similar E3 ligase activity towards GSPT1 as compared to CRBN isoform 1 (data not shown). CRBN isoform 2 lacks an alanine residue at position 23 of CRBN isoform 1, and as such the numbering is shifted compared to isoform 1. Note that the CRBN(W385A) mutant showed diminished activity in cells treated with 0.1 μM CC-885, and similar activity in cells treated with 1 μM CC-885, compared to wild-type CRBN. CRBN(E376V) mutant had no activity at both concentrations. Immunoblot analysis of 293FT parental and CRBN−/− cells treated with 100 μg ml−1 cyclohexamide for the indicated periods. Where indicated, cells were pretreated with DMSO or 10 μM CC-885 for 30 min. For gel source data, see Supplementary Information Fig. 1.
Extended Data Figure 5 Identification of the region of GSPT1 indispensable for CC-885-dependent destruction.
a, Schematic of human GSPT1 and Saccharomyces cerevisiae SUP35 chimaeras. Truncation analysis of GSPT1 revealed that domains 2 and 3 of GSPT1 contain the CC-885-dependent CRBN-binding motif (data not shown). In this region, GSPT1 and SUP35 share 53% sequence identity. Chimaeric fusion sites were selected from several stretches of identical regions to ensure proper folding of the resultant fusion product. Changes in protein level of these fusion proteins in 293T HEK cells in response to CC-885 treatment, as shown in b–d, is summarized on the right. +, protein degraded; −, no change of protein level. b–d, Immunoblot analysis of 293T HEK parental cells or cells stably expressing HA-tagged GSPT1, SUP35 or GSPT1–SUP35 chimaeric proteins. Note that SUP35 expression level is relatively low compared to GSPT1, possibly owing to its intrinsic instability when expressed in human cell lines. e, Immunoblot analysis of 293T HEK cells stably producing HA-tagged GSPT1 wild-type and variants. Non-conserved amino acids in GSPT1 as shown in Fig. 2a were replaced with corresponding residues in SUP35. f, g, Immunoblot analysis of anti-HA immunoprecipitates (f) and whole-cell extracts (g) of 293T HEK cells expressing GSPT1 wild-type and mutants. DMSO or CC-885 was added into the whole-cell extract after lysis for 6 h. For gel source data, see Supplementary Information Fig. 1.
a, Immunoblot analysis of MOLM-13 and OCI-AML2 parental cells or cells stably expressing HA-tagged GSPT1(G575N). Cells were treated with CC-885. b, Cells shown in a were incubated with CC-885 at the indicated concentration for 3 days. Cell proliferation was determined by CellTiter-Glo (CTG). Result is representative of three biological replicates. Data are presented as mean ± s.d., n = 3 technical replicates. c, 293T HEK cells were infected with an increased amount of empty lentiviral vector (EV) or vectors expressing control shRNA or any of the four GSPT1-specific shRNAs. Seven days after infection, cells were imaged using phase-contrast microscope (bottom) and collected for immunoblot analysis (top). Scale bars measure 0.5 mm. Images shown are representative of three captured images. d, e, Cell growth curve of MOLM-13 parental cells or cells transduced with lentiviral vectors expressing a control shRNA (shCNTL) or GSPT1 specific shRNA (shGSPT1-1 to 4). Immunoblot analyses of whole-cell extracts (d) were carried out at day 4 after transduction with lentiviral vectors. Cell growth (e) was quantified with CTG at day 4, 6, and 8 after transduction. Result is representative of three biological replicates. Data are mean ± s.d., n = 3 technical replicates. For gel source data, see Supplementary Information Fig. 1.
Extended Data Figure 7 Kinetic parameters for cereblon–DDB1–CC-885 binding to GSPT1, and sample electron density from the crystal structure of the complex.
a, Reference-corrected surface plasmon resonance binding curves for various concentrations of cereblon–DDB1 (threefold dilutions from 1 μM, coloured traces) flowed over a surface of covalently immobilized anti-GST antibody bound to GST–GSPT1 domains 2 and 3 at 10 °C in the presence of saturating levels of CC-885 or control compound glutarimide. Kinetic parameters shown were determined by fitting with a 1:1 kinetic binding model (black lines) using the Bioacore T200 kinetic analysis software package. Binding in the presence of gluatrimide could not be quantified. We show a representative set of curves from three independent experiments. For a 1:1 binding model where A + B ⇆ AB, the net rate of complex formation is given by the equation d[AB]/dt = ka[A][B] – kd[AB] and the rate of complex disassociation is given by kd[AB], where ka is the association rate constant (M−1s−1) and kd is the dissociation rate constant (s−1). KD, the equilibrium dissociation constant, is defined by KD = kd/ka. Rmax is a measurement of the analyte binding capacity, or maximum response. b, Analysis of the steady-state response versus the concentration of analyte in the presence of CC-885, and the determined affinity constant. c, GST–GSPT1 interacts with endogenous binding partner eRF1. We show that the purified GST–GSPT1 domains 2 and 3 (amino acids 437–633) protein used in this SPR binding assay and the electron miscroscopy experiments (Extended Data Fig. 8) is competent to bind purified eRF1 as reported in ref. 17. Coomassie stain, with lanes 1, 2 and 3 showing individual proteins. Lanes 4 and 5 show pull-down of GST; lanes 6 and 7 show pull-down of GST–GSPT1 domains 2 and 3, all were bound to magnetic glutathione beads incubated with eRF1 ± CC-885 for 1 h and washed three times. Lanes 8 and 9 show pull-down of GST-GSPT1 bound to magnetic glutathione beads incubated with CRBN–DDB1 ± CC-885 for 1 h and washed three times. This experiment was performed twice. For gel source data, see Supplementary Information Fig. 1. d, Sample electron density from the cereblon–DDB1–CC-885–GSPT1 crystal structure with cereblon residues shown in purple, GSPT1 residues shown in grey, and CC-885 shown in green. Refined 2Fo − Fc density is shown as a blue mesh contoured at 1.4σ. Fo − Fc difference density, shown as a green mesh contoured at 4σ, was generated by a single round of Refmac5 refinement calculated in the absence of GSPT1 residues 570–577.
Extended Data Figure 8 Imaging of GSPT1 binding to cereblon–DDB1–CC-885 by negative stain electron microscopy.
a, Negative stain class averages of N-terminal-tagged GST–GSPT1 domain 2 and 3 (amino acids 437–633) bound to purified cereblon–DDB1 in the presence of CC-885. As the GST-tag was fused to the N terminus of domain 2 of GSPT1, domain 3 can be identified as mediating the interaction with cereblon. Whereas the GST tag appears flexible in position, the two domains of GSPT1 are consistent in their orientation with the cereblon–DDB1 complex. GST-dimerization mediates the binding of a second substrate in the majority of the classes. Each class average is composed of between 40 and 70 individual particles. This experiment was performed three times. b, Negative stain class averages of cereblon–DDB1 and GST–GSPT1 domains 2 and 3 (amino acids 437–633) in the presence of DMSO instead of CC-885. No classes containing bound substrate were observed in the absence of CC-885. Each class average is composed of between 40 and 70 individual particles. This experiment was performed once. c, A negative stain class average of GST–GSPT1 domains 2 and 3 bound to cereblon–DDB1–CC-885, with DDB1 shaded purple, cereblon shaded green, GSPT1 domains 2 and 3 shaded blue, and the second dimerized GST tag and second GSPT1 left uncoloured. d, For comparison, the crystal structure of GSPT1 domains 2 and 3 bound to cereblon–DDB1–CC-885 with DDB1 in purple, cereblon in green, and GSPT1 in blue. The electron micrographs revealed a consistent configuration of GSPT1 with cereblon in all complex class averages and confirmed that domain 3 mediates cereblon interactions on the basis of the orientation of the GST-tag.
Cereblon and substrate proteins were co-expressed in 293FT CRBN−/− cells, co-immunoprecipitated in the presence or absence of CC-885 and lenalidomide, and analysed by western blot. a, Co-immunoprecipitation of GSPT1 with wild-type and mutant cereblon. veh, vehicle, DMSO; 885, 10 μM CC-885. b, Co-immunoprecipitation of Ikaros with wild-type and mutant cereblon. veh, vehicle, DMSO; LEN, 10 μM lenalidomide. c, Co-immunoprecipitation of DDB1 with wild-type and mutant cereblon. Results are representative of three biological replicates. d, e, Western blots showing the effect of lenalidomide or CC-885 on Ikaros and GSPT1 degradation; d, effect with human cereblon, e, effect with mouse cereblon. For convenience, the human amino acid numbering is used to discuss the corresponding residues in mouse. GFP and actin are shown as transfection and loading controls, respectively. This is a representative experiment of three biological replicates. For gel source data, see Supplementary Information Fig. 1.
About this article
Cite this article
Matyskiela, M., Lu, G., Ito, T. et al. A novel cereblon modulator recruits GSPT1 to the CRL4CRBN ubiquitin ligase. Nature 535, 252–257 (2016). https://doi.org/10.1038/nature18611
This article is cited by
Molecular Cancer (2022)
A proximity biotinylation-based approach to identify protein-E3 ligase interactions induced by PROTACs and molecular glues
Nature Communications (2022)
Nature Communications (2022)
Nature Reviews Drug Discovery (2022)