Main

Current strategies for small-molecule-mediated targeted protein degradation (TPD)1 rely on compound-mediated induction of proximity between a neosubstrate and an E3 ubiquitin ligase2. Such ‘degraders’ can either be heterobifunctional, harbouring distinct entities for the target and E3 (‘PROTACs’), or monovalent, binding to either the target or the ligase to adapt its surface and induce a cooperative tripartite assembly (molecular glue degraders, MGDs). Both degrader types have entered clinics3, and the discovery of alternative chemotypes that may unlock previously unexplored TPD strategies is highly desirable.

Natural products (NPs) and their analogues have yielded diverse inducers of protein degradation4, raising the possibility that new degrader chemotypes could be derived from NPs. Pseudo-natural products combine natural-product fragments in arrangements and combinations not observed in NPs. They retain the biological relevance of NPs but open new chemical space and therefore may have unexpected and novel targets5,6, so exploration of their bioactivity5,7 may identify novel small-molecule degrader chemotypes and E3 ligases.

The haem-binding enzyme indoleamine-2,3-dioxygenase 1 (IDO1) converts tryptophan (Trp) to kynurenine (Kyn), and Trp shortage and Kyn elevation are linked to reduced anti-tumour immunity, though by different mechanisms8,9,10,11,12. Moreover, IDO1 expression and Kyn levels are related to Epstein Barr virus (EBV)-associated lymphoma13 and to neurodegeneration14,15. In general, clinical exploration of different IDO1 inhibitors for cancer treatment has16,17,18,19,20,21,22, despite encouraging preclinical data23,24, met with limited success25,26. Plausible reasons include that IDO1 may have non-enzymatic functions27,28,29,30,31,32,33,34. Nevertheless, IDO1 inhibitors are in active clinical trials35. The limitations of enzymatic inhibitors36 may be overcome by IDO1 degradation, as initially supported by IDO1-directed proteolysis-targeting chimeras (PROTACs)37,38,39, but monovalent IDO1 degraders have yet to be identified.

We have identified a class of pseudo-natural products derived from (−)-myrtanol, termed iDegs, that both inhibit IDO1 and induce IDO1 degradation. iDegs induce structural changes in IDO1 that cause enhanced ubiquitination and augmented degradation by CRL2KLHDC3, a ligase we identified to also mediate the ubiquitination and native degradation of IDO1. Our work defines a unique mechanism of action, a previously not identified type of degrader, and reveals an E3-ligase previously not used for small-molecule-mediated protein degradation.

Results

Identification of IDO1 degraders

A library of 157,332 small molecules were screened in a cell-based assay (Kyn assay; Extended Data Fig. 1a) that quantified Kyn, the product of the IDO1 reaction. So as to screen under conditions that would be sensitive to IDO1 levels, we considered that IDO1 expression can be induced by interferon gamma (IFN-γ)40. We thus performed the screen in IFN-γ-stimulated BxPC3 cells21. The (−)-myrtanol-derived pseudo-natural product, hereafter termed iDeg-1 (see Supplementary Scheme 1 for the synthesis), inhibited Kyn formation with a half-maximal inhibitory concentration (IC50) of 0.83 ± 0.31 µM in the screening assay (Fig. 1a), which we confirmed using orthogonal Kyn assays in IFN-γ-stimulated BxPC3, SKOV-3 and HeLa cells (IC50 values of 1.1 ± 0.1 µM, 1.6 ± 0.3 µM and 1.7 ± 1.2 µM, respectively; Fig. 1b and Extended Data Fig. 1b,c). Further indication that iDeg-1 targets IDO1 came from a cell viability assay in two-dimensional (2D) and 3D cultures41,42: iDeg-1 reduced IDO1-dependent SKOV-3 cell death induced by IFN-γ (Extended Data Fig. 1d,e). iDeg-1 only slightly affected IDO1 enzymatic activity (Extended Data Fig. 1f), did not impair IDO1 transcription (Extended Data Fig. 1g,h), but dose-dependently reduced IDO1 protein levels up to 45 ± 15% at 10 µM after 24 h (Fig. 1c,d) without inhibiting in vitro translation of IDO1 or global protein translation (Extended Data Fig. 1i–n).

Fig. 1: iDeg-1 reduces IDO1 protein levels via the UPS.

a, Structure of screening hit iDeg-1 and IC50 value in the Kyn assay in BxPC3 cells. Data are presented as mean value ± s.d.; n = 3 biological replicates. b, Kyn assay in BxPC3 and SKOV-3 cells after treatment with iDeg-1 and 50 or 5 ng ml−1 IFN-γ, respectively, for 48 h before detection of Kyn levels utilizing p-DMAB. c,d, IDO1 protein levels in BxPC3 cells upon treatment with IFN-γ and iDeg-1 for 24 h. Representative immunoblots (c) and quantified band intensities from c (d). Data are presented as mean values ± s.d.; n = 4 biological replicates. VCL, vinculin. e,f, Detection of ubiquitinated IDO1 by means of a TUBE pulldown after treatment of IFN-γ-stimulated BxPC3 cells with iDeg-1 or DMSO. e, Cells were treated for 6 h with 50 µM iDeg-1 before the TUBE pulldown. Representative immunoblots of n = 3 biological replicates are shown. f, Cells were treated with 450 nM carfilzomib (CFZ) 60 min before the addition of iDeg-1 or DMSO for 2 h, followed by TUBE pulldown. Representative immunoblots of n = 3 biological replicates for IDO1 are shown. UB, ubiquitin. The dashed line indicates where lanes of the same blot were spliced together. g,h, HEK239T cells were electroporated with rhIDO1 protein. Cells were treated with CFZ for 30 min before the addition of 3.33 µM iDeg-1 and further incubation for 6 h. g, Representative immunoblot of n = 3 biological replicates. h, Quantified band intensities from g, representing samples treated with compound relative to DMSO (set to 100%). Data are presented as mean values ± s.d.; n = 3 biological replicates. i, Volcano plot of iDeg-1-induced changes in the global proteome. HEK239T cells were electroporated with rhIDO1 protein followed by treatment with 10 µM iDeg-1 or DMSO for 6 h and MS analysis. In total, 7,541 proteins were detected. Plot generated using Perseus. Statistical significance was assessed using both-sided t-test with error-corrected P values (S0 = 0.5 (non-linear cutoff), FDR = 0.01). j,k, IDO1 immunoprecipitation (IP) and identified peptide of IDO1 (j) or ubiquitin (k) with diGly modification. IFN-γ-stimulated BxPC3 cells were treated with 20 µM iDeg-1 or DMSO for 6 h before the IP. Data are presented as mean values ± s.d.; n = 3 biological replicates.

Source data

To determine whether iDeg-1 induces degradation via the ubiquitin proteasome system (UPS), IFN-γ-stimulated BxPC3 cells were treated with iDeg-1 for 6 h (Fig. 1e), or for 2 and 4 h after pretreatment with the proteasome inhibitor carfilzomib (CFZ; Fig. 1f and Extended Data Fig. 2a). Increased polyubiquitination was detected with a tandem ubiquitin binding entity (TUBE) pulldown from cell lysates43. Ubiquitination was induced shortly after compound addition, but protein reduction was only detectable after 24 h in IFN-γ-stimulated BxPC3 cells (Fig. 1c,d). Because the capacity to induce degradation may be masked by the ongoing IFN-γ-stimulated production of IDO1, we directly introduced enzymatically active recombinant human IDO1 protein (rhIDO1) into HEK293T cells by electroporation (HEKrhIDO1 cells). iDeg-1 dose-dependently reduced Kyn amounts with an IC50 value of 0.45 ± 0.1 µM (Extended Data Fig. 2b,c) and lowered IDO1 protein after 6 h by 46% at a concentration of 10 µM (Extended Data Fig. 2d,e). CFZ inhibited degradation (Fig. 1g,h and Extended Data Fig. 2f), indicating involvement of the UPS. iDeg-1 also inhibited Kyn production in HEK293T cells, which transiently express IDO1, with an IC50 of 0.42 ± 0.2 µM (Extended Data Fig. 2g). Therefore, iDeg-1 depletes IDO1 independently of the effects of IFN-γ.

Global proteome profiling of HEKrhIDO1 cells after 6 h of treatment with iDeg-1 showed that, besides IDO1, reduced levels were detected only for DOCK-8 and RHOBTB3. However, these effects could not be validated (Fig. 1i, Extended Data Fig. 3a,b and Supplementary Table 1). Analysis of ubiquitinated proteins for a diglycine (diGly) attached to lysine residues (K-ε-diglycine) that were modified with ubiquitin44, after iDeg-1 treatment and IDO1 immunoprecipitation uncovered K389 of IDO1 as a site of ubiquitination (Fig. 1j and Extended Data Fig. 3c). In addition, we observed a 3.5-fold increase in ubiquitin levels in the IDO1 immunoprecipitate after treatment with iDeg-1, confirming iDeg-1-induced IDO1 ubiquitination (Extended Data Fig. 3d). DiGly analysis further revealed K48 linkages in the ubiquitin chains (Fig. 1k and Extended Data Fig. 3e), which are associated with protein degradation via the UPS45.

Direct engagement of IDO1 in cells was proven by a cellular thermal shift assay (CETSA)46, which showed that iDeg-1 stabilizes IDO1 with a shift in the melting temperature (ΔTm) of 3.5 ± 0.4 °C (Fig. 2a,b). An isothermal CETSA experiment at 50 °C (Extended Data Fig. 4a,b) proved that stabilization was dose-dependent.

Fig. 2: iDegs interact with IDO1.

a,b, CETSA in intact SKOV-3 cells treated with 50 µM iDeg-1 or DMSO for 1 h followed by heat treatment and immunoblotting. a, Representative immunoblot for IDO1. b, Quantification of band intensities from a. Data are presented as mean values ± s.d.; n = 3. c, Structures of iDeg-2 and iDeg-3 and IC50 values in the Kyn assay in BxPC3 cells. Data are presented as mean values ± s.d.; n = 3 biological replicates. d,e, Influence of iDeg-1, iDeg-2 and iDeg-3 on IDO1 protein levels in BxPC3 cells. Cells were treated with IFN-γ and the compounds (3.33 µM) for 24 h before immunoblotting (d). Quantification of the band intensities is shown in e. Data are presented as mean values ± s.d.; n = 3 biological replicates. Dashed lines indicate where lanes of the same blot were spliced together. f, Influence on in vitro rhIDO1 activity. rhIDO1 was pre-incubated with the compounds at 37 °C for 90 min before the detection of Kyn levels using p-DMAB. Data are presented as mean values ± s.d.; n = 3 independent experiments. g, rhIDO1 thermal stability in the presence of 50 µM iDeg-1, iDeg-2 or iDeg-3 or DMSO and the apo-IDO1 inhibitor linrodostat (50 µM) using nanoDSF. rhIDO1 and the compounds were pre-incubated for 3 h at 37 °C before measurement. Representative results are shown (n = 3 independent experiments). h, Detection of haem-bound IDO1 by means of UV–vis spectroscopy in the presence of iDeg-1, iDeg-2 or iDeg-3 (100 µM), DMSO or linrodostat (100 µM). Incubation was performed at 37 °C for 3 h. Representative data are presented for n = 3 independent experiments.

Source data

Initial structure–activity exploration identified iDeg-2 (IC50 = 138 ± 23 nM) and iDeg-3 (IC50 = 46 ± 21 nM), which embody an iodine or an alkyne substituent, respectively, in the para position of the phenyl carbamate (Fig. 2c and Extended Data Fig. 4c,d). In BxPC3 cells exposed continuously to IFN-γ, iDeg-2 and iDeg-3 reduced IDO1 protein levels by 55 ± 3% and 62 ± 11%, respectively, as compared to 42 ± 2% by iDeg-1 at 3.33 µM (Fig. 2d,e). iDeg-2 and iDeg-3 also partially inhibited the enzyme in vitro (Fig. 2f) and stabilized rhIDO1, as determined by nano differential scanning fluorimetry (nanoDSF; Fig. 2g).

IDO1 binds haem as a cofactor in the active site, and exists in haem-free (apo) and haem-bound (holo) forms in cells19. UV–vis spectroscopic analysis revealed that in the presence of iDeg-1, iDeg-2 or iDeg-3, the specific Soret absorbance peak of haem-bound IDO1 (holo-IDO1; Fig. 2h) is reduced, showing that iDegs displace haem and, with different potencies, bind to apo-IDO1. Accordingly, addition of hemin reduced the potency of iDeg-1, iDeg-2 and iDeg-3 in the Kyn assay (Extended Data Fig. 4e–g) and dose-dependently elevated Kyn levels in the presence of iDeg-2 (Extended Data Fig. 4h). As we detected both enzymatic inhibition and degradation of IDO1 by iDeg-3, the compound class was termed iDeg.

Crystal structure shows iDegs alter conformation of the IDO1 C-terminal region

The co-crystal structure of IDO1 in complex with iDeg-1 (PDB 9RIS) and iDeg-2 (PDB 9FOH) at 2.1 Å and 1.6 Å resolution, respectively (Supplementary Table 2 and Extended Data Fig. 5a), revealed that iDegs reside in the haem binding site in IDO1. The phenyl carbamate occupies lipophilic pocket A47 in the distal haem site, and the pyrrolidine and the sulfonyl group occupy the haem binding pocket. The monoterpene scaffold only slightly protrudes into the D-pocket, and the tert-butyl phenyl group is located in the solvent-exposed B-pocket (Fig. 3a,b). Interestingly, binding to this latter pocket has previously been observed mainly for holo-IDO1 inhibitors47 (Extended Data Fig. 5b). iDegs binding occurs through numerous hydrophobic interactions, a water-bridged hydrogen bond between the carbamate nitrogen and the hydroxyl group of S167 and a hydrogen bond between the sulfonyl oxygen of iDeg-2 and H346 (Fig. 3c). As many of the same IDO1 residues are responsible for haem coordination, this binding mode is mutually exclusive with haem and ensures that iDegs can only bind apo-IDO1. The iodine moiety of iDeg-2 establishes additional hydrophobic contacts within the distal region of the A-pocket, specifically with the main chain of L124 and V125 (Extended Data Fig. 5c). These interactions probably contribute to the enhanced activity of iDeg-2 compared to iDeg-1 (Fig. 2). Comparisons to all IDO1 structures revealed that iDeg-1 and iDeg-2 induce a striking conformational rearrangement (Fig. 3d,e). In the presence of iDeg-1, the electron density map allowed the building of a shorter C-terminal K-helix although with comparatively high B-factors. In contrast, in the iDeg-2-bound structure, no electron density for the K-helix was detected (Fig. 3d and Extended Data Fig. 5d). This absence is not attributed to crystal packing forces, as the K-helix is present in the crystal structure of IDO1 with the inhibitor apoxidole with identical space group, unit cell and similar crystallization conditions (PDB 8ABX; Extended Data Fig. 5e)48. An overlay of the structures of IDO1 bound to iDeg-2 or with the apo-IDO1 inhibitor linrodostat revealed that helices B, C, F, H and J are all reoriented and the J-helix is also remodelled in the iDeg complex (Fig. 3e). In all previously reported IDO1 structures, the K-helix is embraced through F, J and H helices and the E–F loop (Extended Data Fig. 5f). In contrast, in the iDeg-1- and iDeg-2-bound complexes, tethers to the K-helix are weakened due to cumulative movements of several amino acids within the J-helix. Specifically, a substantial rotation and translation of H346, which coordinates the central iron of the haem group in holo-IDO1, shifts the J-helix C terminus towards the iDeg-2 binding pocket. Residues R343, F270 and T395 also adopt an alternative conformation compared to all other published IDO1 structures (Fig. 3f,g and Extended Data Fig. 5f). Weakened interactions between the K-helix and surrounding helices F, J and H and the E–F loop due to their reorientation presumably increase the flexibility of the K-helix, accounting for its invisibility in the electron density map. As previously reported IDO1 inhibitors do not substantially alter the overall structure of IDO1 compared to holo-IDO147 (Extended Data Fig. 5f), the observed conformational rearrangements represent a novel and unique binding mode of the iDegs.

Fig. 3: Co-crystal structures reveal that iDeg-1 and iDeg-2 bind to apo-IDO1 and induce conformational changes in its C-terminal region.

a, Cartoon diagram of the IDO1-iDeg-1 and IDO1-iDeg-2 structures. Dashed lines indicate regions lacking electron density, including the E–F loop, J–K loop and K-helix, with parentheses denoting their proposed flexibility. The positions of ubiquitinated K389 within the K-helix and H346 are highlighted. The α-helices of the large domain (residues 121–403) are labelled A to K. b, Surface model (cut-in side view) showing iDeg-1 and iDeg-2 in the active site of the IDO1–iDeg-2 structure. Hydrophobic pockets A, B and D and the haem binding site are indicated. c, Zoom-in view of the amino acids involved in iDeg-2 binding. Hydrogen bonds are indicated by black dashed lines. d, Comparison of the IDO1 structures with linrodostat (PDB 6DPR-B), iDeg-1 and iDeg-2, illustrating the transition of the K-helix from a rigid to a dynamic conformation upon iDeg binding. Surface regions interacting with the K-helix are highlighted. e, Overlay of IDO1-iDeg-2 (violet) and IDO1–linrodostat (grey; PDB 6DPR-B) structures. iDeg binding triggers local and long-distance structural perturbations (red dashed arrows). f, iDeg-2 binding induces a conformational shift in H346, moving the J-helix towards the iDeg-2 binding pocket, including R343. F270 also adopts a shifted conformation. g, Residues T395, Q281 and R343, which are involved in stabilizing the K-helix through hydrogen bonding, similarly adopt shifted positions in the iDeg-bound state.

Mechanistically, because iDegs bind deep in the active site, it is unlikely that they directly contact an E3 ligase. Instead, iDeg-induced degradation possibly involves increased accessibility of the C-terminal K-helix bearing the ubiquitinated K389, as compared to the conformations observed for all published IDO1 inhibitors.

To explore the contribution to IDO1-induced degradation of several amino acids that adopt substantially shifted positions in the iDeg-bound structures, we generated an IDO1 stability reporter in KBM7 cells harbouring an inducible Cas9 cassette49 (Fig. 4a). Supporting a role of the C-terminal region of IDO1 in degradation, only N-terminally fused BFP enabled iDeg-induced reporter degradation (Fig. 4b). Alanine substitutions were introduced at H346, R343 and F270, which interact with haem in the holo-IDO1 structure50 and adopt different conformations in the iDeg-bound state compared to all other published IDO1 structures. In addition, we tested the T395M mutation, as T395 located in the K-helix contributes to its stabilization through hydrogen bonding with R343 (J-helix) and Q281 (E–F loop). All four mutants displayed lower protein stability in cells, indicating their role in maintaining a stable IDO1 conformation (Fig. 4d). Degradation induced by iDeg-1, iDeg-2 or iDeg-3 in cells expressing these IDO1 mutants was rescued only by the H346A mutation, highlighting their interaction with H346 as being crucial for degradation (Fig. 4e).

Fig. 4: KLHDC3 is involved in iDeg-mediated IDO1 degradation.

a, IDO1 stability reporter design. b, IDO1 levels detected by IDO1 stability reporters. KBM7 IDO1 reporter cells were treated with iDeg-1, iDeg-2 or iDeg-3 (1 µM) for 24 h before detection of IDO1 levels using flow cytometry. Normalized BFP-to-mCherry ratios (norm. BFP) were calculated per genotype, respectively. Data are presented as mean values ± s.d.; n = 3 biological replicates. c, Representative histogram for the iDeg-1-mediated depletion of BFP-IDO1 (24 h, 10 µM). d, Influence of IDO1 mutations on BFP-IDO1 protein levels. Data are presented as mean values ± s.d.; n = 3 biological replicates. e, Influence of IDO1 mutants on the degradation efficiency of iDegs (24 h, 10 µM). Data are presented as mean values ± s.d.; n = 3 biological replicates. f, Identification of genes required for native IDO1 degradation. g, Identification of genes required for iDeg-1-mediated IDO1 degradation. CSN, COP9 signalosome. In f and g, genes are highlighted for P < 0.05 (one-sided MAGeCK) and log2(fold-change, FC) > 1.585 (n = 2 biological replicates). h, IDO1 depletion is rescued by 10 h of co-treatment with either CFZ, TAK243 or MLN4924 (1 µM each). Data are presented as mean values ± s.d.; n = 3 biological replicates. i, Proximity labelling and enrichment of biotinylated KLHDC3 in HEK293T cells expressing IDO1–TurboID biotin ligase. The treatment time with iDeg-3 or DMSO was 2.5 h. A representative immunoblot is shown for n = 2 biological replicates, and 2% of the total protein input used for the pulldown was loaded in the input lane. j, KLHDC3 affinity towards WT (IDO1C-deg) and mutant (IDO1C-deg-G403K and IDO1C-deg-K389R) IDO1 C-terminal peptides determined using ITC. NB, no binding. k, Fluorescent ubiquitin transfer from neddylated CRL2KLHDC3-activated UBE2R2 to the indicated C-terminal IDO1 peptides over time. l, IDO1 stability reporter variants in KBM7 cells measured by flow cytometry and depicted normalized to the WT (that is, –EG) reporter. Data are presented as mean values ± s.d.; n = 3 biological replicates. See also Supplementary Fig. 1.

Source data

iDegs promote IDO1 ubiquitination by KLHDC3

To identify the E3 responsible for iDeg-mediated IDO1 degradation, we conducted a fluorescence-activated cell sorting (FACS)-based CRISPR–Cas9 screen using an sgRNA library targeting 1,301 ubiquitin-associated genes (six sgRNAs per gene)51. Cas9 expression was induced for 72 h before 14 h of compound treatment, then the cells were enriched for increased or decreased BFP levels using FACS and the corresponding sgRNAs quantified by deep sequencing (Fig. 4f,g, Extended Data Fig. 6b and Supplementary Fig. 1), revealing the genes functionally required for iDeg-induced degradation.

As expected, knockout of proteasome subunits or of genes involved in neddylation counteracted iDeg activity (Fig. 4f,g and Extended Data Fig. 6b,c), thus phenocopying the IDO1 stability reporter behaviour upon chemical perturbation of the UPS (Fig. 4h and Extended Data Fig. 6d). Importantly, we further identified the cullin-RING ligase (CRL) complex, including cullin2 (CUL2), RBX1, elongin B/C (EloB and EloC) and the Kelch domain containing protein 3 (KLHDC3), as required for IDO1 degradation.

Unexpectedly, genetic disruption of the CRL2KLHDC3 complex also affected baseline IDO1 turnover under vehicle (dimethyl sulfoxide, DMSO) treatment. In the presence of iDegs, however, knockout of these genes led to an even higher enrichment of the corresponding sgRNAs (Fig. 4g and Extended Data Fig. 6b,c). This indicated that the compounds could function by enhancing the efficiency of IDO1 degradation. To validate these findings, we employed the TurboID approach and expressed IDO1 fused to the TurboID biotin ligase in HEK293T cells followed by enrichment of biotinylated proteins. Indeed, KLHDC3 was slightly biotinylated in the control condition, which increased upon treatment of cells with iDeg-3 (Fig. 4i), thus confirming that IDO1 and KLHDC3 interact in cells, and that this interaction is enhanced in the presence of the degrader. Contrary to classical degrader modalities such as PROTACs or molecular glue degraders, which typically function by inducing proximity between an E3 and a target that is functionally inconsequential in the absence of the small molecule, iDegs thus appear to promote a native route for IDO1 turnover.

IDO1 is a natural substrate of CRL2KLHDC3

The CRL substrate receptor KLHDC3 has not yet been employed for small-molecule-induced protein degradation. KLHDC3 recognizes C-degrons52,53, which explains why the C-terminally fused IDO1-BFP reporter was not degraded, but the BPF-IDO1 reporter was (Fig. 4b). The C-terminal EG sequence of human IDO1 is consistent with the distinguishing feature of a KLHDC3 C-degron, which is a C-terminal glycine52,53,54. In agreement, a peptide IDO1C-deg corresponding to the C-terminal region (residues 381–403) bound KLHDC3 with a KD of 103 nM (Fig. 4j and Supplementary Fig. 2a). Substituting the C-terminal glycine with lysine (peptide IDO1C-deg-G403K) abolished binding to KLHDC3. Replacing the ubiquitination site (peptide IDO1C-deg-K389R) did not impact E3-degron complex formation (Fig. 4j and Supplementary Fig. 2b).

To test whether IDO1 is a direct substrate of CRL2KLHDC3, we biochemically reconstituted ubiquitination. Because KLHDC3-EloB/C can form an autoinhibited self-assembly55, we used a monomeric version of KLHDC3 (C-terminal Gly-to-Lys mutant). Assays were performed in the ‘pulse-chase’ format. The pulse reaction generates a thioester-linked E2 conjugate with fluorescently labelled ubiquitin. Next, E3 and substrate are added, and fluorescent ubiquitin transfer from E2 (UBE2R2) to the substrate and subsequently to a substrate-linked ubiquitin is observed over time. The IDO1C-deg peptide was ubiquitinated in vitro in a CRL2KLHDC3-dependent manner, whereas the mutant versions were not (Fig. 4k and Supplementary Fig. 2b). In cells, mutation of the IDO1 C-terminal glycine increased abundance, whereas mutation of the non-optimal –EG degron to an optimal –RG C terminus reduced IDO1 abundance (Fig. 4l). These findings demonstrate that IDO1 is a substrate of KLHDC3 and that the C-terminal glycine-based degron is essential for interaction with the E3 ligase.

To benchmark the degradation potency of iDeg-1, iDeg-2 and iDeg-3 and subsequently the identified inhibitor and degrader iDeg-6 (IC50 of 16 ± 5 nM in the Kyn assay; Fig. 5a and Extended Data Fig. 7a), IDO1 levels were detected after stimulation of cells for 24 h with IFN-γ followed by a washout (to avoid IDO1 expression) and compound addition. iDeg-6 most potently depleted IDO1 in the cells, with a maximal achievable degradation (Dmax) of 70% at 100 nM and a half-maximal degradation concentration (DC50) for IDO1 degradation of 6.5 ± 3 nM (Fig. 5a–c and Extended Data Fig. 7b,c). Also, in the BFP-IDO1 reporter cell lines iDeg-1 was the least and iDeg-6 the most potent compound (Extended Data Fig. 7d). In vitro iDeg-6 inhibited IDO1 activity nearly completely and more potently than iDeg-1 to 3 with an IC50 of 1.6 ± 0.3 µM (Extended Data Fig. 7e). Direct binding of iDeg-6 to IDO1 was detected using isothermal titration calorimetry (ITC; Supplementary Fig. 3). The compound induced thermal stabilization of the protein and haem displacement to a higher extent as compared to iDeg-1, iDeg-2 and iDeg-3 (Extended Data Fig. 7f,g). We therefore used iDeg-6 for further validation.

Fig. 5: Validation of KLHDC3 as an E3 ligase regulating IDO1.

a, Structure of iDeg-6 and the IC50 value in a Kyn assay in BxPC3 cells. Data are presented as mean value ± s.d.; n = 3 biological replicates. b,c, Reduction of IDO1 protein levels by iDeg-6. Degradation efficiency was assessed in a modified set-up including IFN-γ washout before compound addition. BxPC3 cells were treated with 50 ng ml−1 IFN-γ for 24 h before washout, addition of iDeg-6 for 24 h, then immunoblotting (b) and quantification of the IDO1 protein levels (c) from b. Data are presented as mean values ± s.d.; n = 4 biological replicates, except for 50 and 500 nM (n = 3 biological replicates). d–f, IDO1 levels in SKOV-3 (d), SKOV-3 cells stimulated with IFN-γ for 24 h followed by a washout (e) or BT549 (f) cells. Cells were treated with iDeg-6 for 24 h before immunoblotting. Representative immunoblots are shown for n = 3 biological replicates for SKOV-3 and SKOV-3 with IFN-γ or n = 4 biological replicates for BT549. g,h, IDO1 protein levels in wild-type (WT) U2OS or KLHDC3 knockout (KO) U2OS cells. Cells were stimulated with 5 ng ml−1 IFN-γ for 24 h before washout, treatment with iDeg-6 or DMSO for 24 h, and immunoblotting (g). Representative data of n = 4 biological replicates (U2OS WT) or n = 2 biological replicates (KO cells) are shown. h, Quantification of the band intensities from g and Extended Data Fig. 8h. Data are presented as mean values ± s.d.; n = 3 biological replicates (U2OS WT) or n = 2 biological replicates (U2OS-KLHDC3 KO). i,j, IDO1 protein levels in KBM7-BFP-IDO1 (CTRL) or KBM7-BFP-IDO1 KLHDC3 KO1 or KO2 cells in the absence (h, normalized to CTRL) or presence of iDeg-6 (i, normalized to the respective genotype). The treatment time was 24 h. Data are presented as mean values ± s.d.; n = 3 biological replicates. k, Schematic representation of the IDO1 competition ubiquitination assay. EB, elongin B; EC, elongin C; N8, neddylation; UB, ubiquitin. l, Quantification of KLHDC3-dependent IDO1 C-terminal peptide ubiquitination (IDO1C-deg) upon titrating increasing concentrations of competing full-length apo-IDO1, iDeg-6-IDO1 or linrodostat-bound IDO1 to measure IC50 values. n = 2 independent experiments. m, Apo-IDO1 ubiquitination after two sequential 20-min incubations (first and second) with 42 µM of each indicated compound, followed by the addition of NEDD8-CUL2KLHDC3 to initiate ubiquitination. Data are presented as mean values; n = 2 independent experiments. See also Supplementary Fig. 4.

Source data

We estimated the contribution of IDO1 inhibition and IDO1 degradation by iDeg-6 to the total reduction of Kyn levels in the presence or absence of the proteasome inhibitor CFZ (Extended Data Fig. 7h–j and Supplementary Table 3). IDO1 degradation accounted for 81% or 59% of the achieved Kyn level decrease at 100 nM or 1 µM, respectively. However, the cellular response to proteasomal inhibition is rather complex and can induce compensatory degradation mechanisms that can impact total IDO1 levels.

iDeg-6 also depleted IDO1 protein in SKOV-3 and BT549 cells in the absence of or upon stimulation with IFN-γ in a dose- and time-dependent manner (Fig. 5d–f and Extended Data Fig. 8a–e). The neddylation inhibitor MLN4924 rescued iDeg-6-induced IDO1 depletion (Extended Data Fig. 8f) and increased IDO1 in the absence of iDegs, demonstrating that neddylation is required for both native and iDeg-induced IDO1 degradation. KLHDC3 knockout in U2OS or KBM7-BFP-IDO1 cells increased IDO1 amounts and rescued iDeg-6-dependent degradation (Fig. 5g–j and Extended Data Fig. 8g,h; also Extended Data Fig. [6e](https://www.nature.com/articles/s415