Thalidomide promotes degradation of SALL4, a transcription factor implicated in Duane Radial Ray syndrome

Abstract

In historical attempts to treat morning sickness, use of the drug thalidomide led to the birth of thousands of children with severe birth defects. Despite their teratogenicity, thalidomide and related IMiD drugs are now a mainstay of cancer treatment; however, the molecular basis underlying the pleiotropic biology and characteristic birth defects remains unknown. Here we show that IMiDs disrupt a broad transcriptional network through induced degradation of several C₂H₂ zinc finger transcription factors, including SALL4, a member of the spalt-like family of developmental transcription factors. Strikingly, heterozygous loss of function mutations in SALL4 result in a human developmental condition that phenocopies thalidomide-induced birth defects such as absence of thumbs, phocomelia, defects in ear and eye development, and congenital heart disease. We find that thalidomide induces degradation of SALL4 exclusively in humans, primates, and rabbits, but not in rodents or fish, providing a mechanistic link for the species-specific pathogenesis of thalidomide syndrome.

Keywords: biochemistry; chemical biology; developmental biology; human; teratogenicity; transcription factors; ubiquitin.

Figure 1.. Identification of SALL4 as an IMiD-dependent CRL4^CRBN substrate.

(A–C) Scatter plot depicting identification of IMiD-dependent substrate candidates. H9 human embryonic stem cells (hESC) were treated with 10 µM thalidomide (A), 5 µM lenalidomide (B), 1 µM pomalidomide (C), or DMSO control, and protein abundance was analyzed using TMT quantification mass spectrometry (see Materials and methods for details). Significant changes were assessed by a moderated t-test as implemented in the limma package (Ritchie et al., 2015), with the log2 fold change (log2 FC) shown on the y-axis, and negative log₁₀ p values on the x-axis (two independent biological replicates for each of the IMiDs, or three independent biological replicates for DMSO). (D) Heatmap displaying the mean log2 FC of the identified IMiD-dependent targets comparing treatment with thalidomide, lenalidomide, and pomalidomide. Mean log2 FC values were derived from averaging across proteomics experiments in four different cell lines (hESC, MM1s, Kelly, SK-N-DZ). The heatmap colors are scaled with blue indicating a decrease in protein abundance (−2 log2 FC) and red indicating no change (0 log2 FC) in protein abundance. Targets newly identified in this study are marked with a green dot, ZnF containing targets with a cyan dot, and previously characterized targets with a gray dot. Substrates are grouped according to their apparent IMiD selectivity in the mass spectrometry-based proteomics. It should be noted that this does not refer to absolute selectivity but rather relative selectivity.

Figure 1—figure supplement 1.. Mass spectrometry profiling of IMiDs.

(A) Schematic representation of the mass spectrometry-based proteomics workflow used for IMiD profiling (the raw output of mass spectrometry processing is provided in Figure 1—source data 1–14). (B) Chemical structures of compounds used in this study. (C) Scatter plots depicting identification of treatment-dependent substrate candidates. Kelly cells were treated with 10 µM thalidomide (3× biological replicates), 5 µM lenalidomide (3× biological replicates), 1 µM pomalidomide, or DMSO as a control (3× biological replicates) for 5 h (top row). MM1s cells were treated with 10 µM thalidomide (2× biological replicates), 5 µM lenalidomide (2× biological replicates), 1 µM pomalidomide (2× biological replicates), or DMSO as a control (3× biological replicates) for 5 h (middle row). SK-N-DZ cells were treated with 0.1 µM CC-220, 1 µM dBET57, 1 µM pomalidomide (3× biological replicates), or DMSO as a control (3× biological replicates) for 5 h (bottom row). Protein abundance from each experiment was analyzed using TMT quantification mass spectrometry (see Materials and methods for details). Significant changes were assessed by a moderated t-test as implemented in the limma package (Ritchie et al., 2015), with the log2 fold change shown on the y-axis, and negative log₁₀p values on the x-axis (the number of replicates is indicated above). Hits that met the set significance threshold (fold-change greater 1.5 and log10 p value below 0.001) are displayed with a red point (•).

Figure 1—figure supplement 2.. Extended validation of IMiD-dependent targets.

(A) Heatmap summarizing the protein abundance of IMiD-dependent targets identified from proteomics data across four different cell lines (Kelly, MM1s, hES, and SK-N-DZ cells) and five different compounds (thalidomide, lenalidomide, pomalidomide, CC-220, and dBET57). The color scale displays a 2.5-fold decrease in protein abundance in blue and no change is displayed in white. NA indicates that the protein was not identified/quantified in the experiment. (B) Mass spectrometry scatter plot validation of IMiD-dependent targets. SK-N-DZ cells were treated with 1 µM pomalidomide to induce degradation of IMiD-dependent targets (left), degradation was rescued by co-treatment with 1 µM pomalidomide +5 µM MLN4924 (right), or treated with DMSO as a control for 5 h. Protein abundance from each experiment was analyzed using TMT quantification mass spectrometry (see Materials and methods for details). Significant changes were assessed by a moderated t-test as implemented in the limma package (Ritchie et al., 2015), with the log2 FC shown on the y-axis, and -log₁₀p values on the x-axis (for three biological replicates). Hits that met the significance threshold (fold-change greater 1.5 and log₁₀ p value below 0.001) are displayed with a red point (•). (C) Reporter ion ratios from (B) were normalized and scaled (see Materials and methods) and are depicted as a bar graph for the IMiD-dependent targets. Co-treatment with the neddylation inhibitor MLN4924 abrogated the degradation of all IMiD-dependent targets in accordance with a Cullin-RING ligase-dependent mechanism of degradation. Data are presented as means ± s.d. (n = 3 biological replicates). (D) Western blot validation of MLN4924 rescue experiment: SK-N-DZ or Kelly cells were treated with increasing concentrations of thalidomide or DMSO as a control. Following 24 h incubation, GZF1 (left) and DTWD1 (right) as well as GAPDH protein levels were assessed by western blot analysis (one representative out of three replicates is shown). (E) Multiple sequence alignment of the CxxCG-containing zinc finger domains from each of the IMiD-dependent targets identified by mass spectrometry in this study.

Figure 2.. Validation of SALL4 as a bona fide IMiD-dependent CRL4^CRBN substrate.

(A) H9 hESC were treated with increasing concentrations of thalidomide, lenalidomide, pomalidomide, or DMSO as a control. Following 24 h of incubation, SALL4 and GAPDH protein levels were assessed by western blot analysis. (B) As in (A), but treatment was done in Kelly cells. (C) Kelly cells were treated with increasing concentrations of thalidomide and co-treated with 5 µM bortezomib, 5 µM MLN4924, 0.5 µM MLN7243, or DMSO as a control. Following 24 h incubation, SALL4 and GAPDH protein levels were assessed by western blot analysis. (D) Parental Kelly cells or two independent pools of CRBN^-/- Kelly cells were treated with increasing concentrations of thalidomide. Following 24 h incubation, SALL4, CRBN, and GAPDH protein levels were assessed by western blot analysis. (E) Kelly cells were treated with 5 µM pomalidomide or DMSO as a control for 8 h, at which point the compound was washed out. Cells were harvested at 1, 2, 4, 8, 24, and 48 h post-washout and SALL4 and GAPDH protein levels were assessed by western blot analysis. (F) Kelly cells were treated with 5 µM pomalidomide for 1, 2, 4, 8, and 24 h, or with DMSO as a control. Following time course treatment, SALL4 and GAPDH protein levels were assessed by western blot analysis. One representative experiment is shown in this figure, from three replicates for each of the western blots.

Figure 2—figure supplement 1.. Extended validation of SALL4.

(A) HEK293T cells were treated with increasing concentrations of thalidomide, lenalidomide, pomalidomide, or DMSO as a control. Following 24 h incubation, SALL4 and GAPDH protein levels were assessed by western blot analysis. (B) As in (A), but with H661 cells. (C) As in (A), but with SK-N-DZ cells. (D) HEK293T cells were treated with increasing concentrations of thalidomide and co-treated with 5 µM bortezomib, 5 µM MLN4924, 0.5 µM MLN7243, or DMSO as a control. Following 24 h incubation, SALL4 and GAPDH protein levels were assessed by western blot analysis. (E) As in (D), but with SK-N-DZ cells. (F) Parental HEK293T cells or two independent pools of CRBN^-/- HEK293T cells were treated with increasing concentrations of thalidomide. Following 24 h incubation, SALL4, CRBN, and GAPDH protein levels were assessed by western blot analysis. (G) Kelly cells were treated with 1 µM pomalidomide or DMSO as a control for 8 h, at which point the compound was washed out. Cells were harvested at 1, 2, 4, 8, 24, and 48 h post-washout, and SALL4 and GAPDH protein levels were assessed by western blot analysis. (H) Kelly cells were treated with 1 µM pomalidomide for 1, 2, 4, 8, and 24 h, or with DMSO as a control. Following time course treatment, SALL4 and GAPDH protein levels were assessed by western blot analysis. (I) Thalidomide treatment did not influence the expression of SALL4 mRNA. hES cells treated with 10 µM thalidomide or DMSO as a control for 24 h were subjected to quantitative RT-PCR to assess the levels of total SALL4 mRNA. The mRNA levels were normalized to those of GAPDH (housekeeping gene) mRNA. The SALL4 mRNA level remained stable or increased, which is in contrast to the decrease in protein abundance observed in proteomics and western blot analysis. mRNA fold change was determined from n = 2 with three technical replicates. (J) To validate the specificity of the antibody used, we transfected Kelly or HEK293T cells with a plasmid expressing mCherry, Cas9, and one of three guide RNAs (sgRNA1, sgRNA2, sgRNA3) targeting the SALL4 gene, or a mock control. Following 48 h incubation, SALL4 and GAPDH protein levels were assessed by western blot analysis and for sgRNA1 and sgRNA2 a loss of the specific bands for SALL4 was observed in accordance with the antibody being specific for SALL4. sgRNA3 had no effect, which is likely to be because of an ineffective sgRNA. (K) To validate the specificity of the antibody used, we transfected Kelly cells with a plasmid overexpressing Flag-mmSALL4, Flag-hsSALL4, or no transfection. Following 48 h incubation, SALL4 and GAPDH protein levels were assessed by western blot analysis. One representative experiment is shown in this figure, from three replicates for each of the western blots.

Figure 3.. SALL4 ZnF2 is the zinc finger responsible for IMiD-dependent binding to CRL4^CRBN.

(A) Multiple sequence alignment of the validated ‘degrons’ from known IMiD-dependent zinc finger substrates, along with the two candidate zinc finger degrons from SALL4. (B) TR-FRET: titration of IMiD (thalidomide, lenalidomide and pomalidomide) to DDB1∆B-CRBN_Spy-BodipyFL at 200 nM, hsSALL4_ZnF2 at 100 nM, and terbium-streptavidin at 4 nM. (C) As in (B), but with hsSALL4_ZnF4 with DDB1∆B-CRBN_Spy-BodipyFL at 1 µM. (D) TR-FRET: titration of DDB1∆B-CRBN_Spy-BodipyFL to biotinylated hsSALL4_ZnF2, hsSALL4_ZnF1-2, or hsSALL4_ZnF4 at 100 nM and terbium-streptavidin at 4 nM in the presence of 50 µM thalidomide. (E) As in (B), but with hsSALL4_ZnF1-2. (F) As in (B), but with hsSALL4_ZnF2 and hsSALL4_ZnF2^G416A mutant as thalidomide titration. (G) Kelly cells transiently transfected with Flag-hsSALL4^WT, Flag-hsSALL4^G600A, or hsSALL4^G600N were treated with increasing concentrations of thalidomide or DMSO as a control. Following 24 h of incubation, SALL4 (α-Flag) and GAPDH protein levels were assessed by western blot analysis (one representative experiment is shown out of three replicates. (H) As in (G), but with Flag-hsSALL4^WT, Flag-hsSALL4^G416A, or Flag-hsSALL4^G416N. (I) In vitro ubiquitination of biotinylated hsSALL4_ZnF1-2 by CRL4^CRBN in the presence of thalidomide (10 µM), lenalidomide (10 µM), or pomalidomide (0.1, 1 and 10 µM), or DMSO as a control.

Figure 3—figure supplement 1.. Biochemical characterization of SALL4 binding to CRBN.

(A) TR-FRET: titration of DDB1∆B-CRBN_Spy-BodipyFL to biotinylated hsSALL4_ZnF2, hsSALL4_ZnF1-2, and hsSALL4_ZnF4 at 100 nM and terbium-streptavidin at 4 nM in the presence of lenalidomide at 50 µM. (B) As in (A), but in the presence of pomalidomide at 50 µM. (C) TR-FRET: titration of lenalidomide to DDB1∆B-CRBN_Spy-BodipyFL at 200 nM, hsSALL4_ZnF2^WT, hsSALL4 _ZnF2^G416A at 100 nM, and terbium-streptavidin at 4 nM. (D) As in (C), but titrating with pomalidomide. (E) TR-FRET: titration of thalidomide to DDB1∆B-CRBN_Spy-BodipyFL at 1 µM, hsSALL4_ZnF4 or hsSALL4_ZnF4^Q595H mutant at 100 nM, and terbium-streptavidin at 4 nM. (F) TR-FRET: titration of thalidomide to DDB1∆B-CRBN_Spy-BodipyFL at 200 nM, hsSALL4_ZnF1-2^WT, hsSALL4_ZnF1-2^G416N, and hsSALL4_ZnF1-2^S388N at 100 nM, and terbium-streptavidin at 4 nM. (G) As in (F), but titrating with lenalidomide. (H) As in (F), but titrating with pomalidomide. (I) TR-FRET: titration of DDB1∆B-CRBN_Spy-BodipyFL to biotinylated hsSALL4_ZnF1-2^WT, hsSALL4_ZnF1-2^G416N, and hsSALL4_ZnF1-2^S388N at 100 nM, and terbium-streptavidin at 4 nM in the presence of thalidomide at 50 µM. (J) TR-FRET: titration of DDB1∆B-mmCRBN_Spy-BodipyFL to biotinylated hsSALL4_ZnF2,hsSALL4_ZnF1-2, and IKZF1∆ (Petzold et al., 2016) at 100 nM, and terbium-streptavidin at 4 nM in the presence of thalidomide at 50 µM. (K) TR-FRET: titration of lenalidomide to DDB1∆B-CRBN_Spy-BodipyFL at 200 nM, hsSALL4_ZnF2, mmSALL4 _ZnF2, and drSALL4 _ZnF2 at 100 nM, and terbium-streptavidin at 4 nM. (L) As in (K), but titrating pomalidomide. All TR-FRET data in this figure are presented as means ± s.d. (n = 3).

Figure 4.. Identification of the sequence differences in the IMiD-dependent binding region of both CRBN and SALL4 in specific species.

(A) Close-up view of the beta-hairpin loop region of Ck1a (CSNK1A1) interacting with CRBN and lenalidomide (PDB: 5fqd) highlighting the additional bulkiness of the V388I mutation (PDB: 4ci1) present in mouse and rat CRBN. CSNK1A1 and lenalidomide are depicted as stick representations in magenta and yellow, respectively, the Ile391 of mouse CRBN corresponding to human Val388 is depicted as a stick representation in cyan, and CRBN is depicted as a surface representation. (B) TR-FRET: titration of DDB1∆B-hsCRBN_Spy-BodipyFL, or DDB1∆B-hsCRBN^V388I_Spy-BodipyFL to biotinylated hsSALL4_ZnF1-2 at 100 nM, and terbium-streptavidin at 4 nM in the presence of 50 µM pomalidomide or DMSO. (C) mES cells were treated with increasing concentrations of thalidomide and pomalidomide or DMSO as a control. Following 24 h of incubation, SALL4 and GAPDH protein levels were assessed by western blot analysis. (D) mES cells constitutively expressing Flag-hsCRBN were treated with increasing concentrations of thalidomide. Following 24 h of incubation, ZFP91 and GAPDH protein levels were assessed by western blot analysis. (E) As in (C), but measuring GZF1 and GAPDH protein levels. (F) As in (C), but measuring SALL4, hsCRBN (α-Flag), and GAPDH protein levels. (G) Kelly cells were transiently transfected with Flag-hsSALL4, Flag-mmSALL4, or Flag-mmSALL4 containing a humanized ZnF2 (Y415F, P418S, I419V, L430F, Q435H), and treated with increasing concentrations of thalidomide. Following 24 h of incubation, hsSALL4, mmSALL4, humanized mmSALL4 (α-Flag), and GAPDH protein levels were assessed by western blot analysis. (H) TR-FRET: titration of thalidomide to DDB1∆B-CRBN_Spy-BodipyFL at 200 nM, hsSALL4_ZnF2, mmSALL4_ZnF2, or drSALL4_ZnF2 all at 100 nM, and terbium-streptavidin at 4 nM. Data are presented as means ± s.d. (n = 3). (I) As in (G), but with Flag-drSALL4.

Figure 4—figure supplement 1.. Species-specific effects.

(A) mES cells were treated with increasing doses up to 100 µM of thalidomide. Following 24 h incubation, SALL4 and GAPDH protein levels were assessed by western blot analysis. (B) Kelly cells were transiently transfected with Flag-hsSALL4 and treated with increasing concentrations of thalidomide. Following 24 h of incubation, SALL4 and GAPDH protein levels were assessed by western blot analysis. (C) As in (B), but with Flag-mmSALL4. (D) Kelly cells were transiently transfected with Flag-mmSALL4 and treated with increasing doses of CC-885, with no transfection as a negative control. Following 24 h incubation, mmSALL4 (α-Flag) protein levels were assessed by western blot analysis. One representative is shown out of three replicates for each western blot. (E) Gene expression profiles for IMiD-dependent substrates were derived from the genotype-tissue expression (GTex) dataset and are presented as a heatmap.

Figure 5.. Sequence differences in the IMiD-dependent binding region of both CRBN and SALL4 interfere with ternary complex formation in specific species.

(A) A multiple sequence alignment of the region of CRBN critical for IMiD mediated ZnF binding from human, bush baby mouse, rat, macaque, marmoset, and rabbit is shown, highlighting the V388I polymorphism. (B) A multiple sequence alignment of SALL4_ZnF2 from human, macaque, marmoset, bush baby, rabbit, mouse, rat, zebrafish, and chicken, highlighting the differences in sequence across species. (C) Schematic summary of species-specific effects of IMiD treatment on ZnF degradation and relationship to thalidomide syndrome phenotype. The top panel depicts sensitive species: hsCRBN^V388 is capable of IMiD-dependent binding, ubiquitination, and subsequent degradation of hsSALL4 and hsZnF targets, and thalidomide embryopathy is observed. The middle panel depicts insensitive species: mmCRBN^I391 is capable of binding IMiDs, but not binding mmSALL4 and mmZnF targets, and no embryopathy is observed. The bottom panel depicts humanizing CRBN as ineffective for inducing the phenotype: hsCRBN^V388 is capable of IMiD-dependent ubiquitination and subsequent degradation of mmZnF proteins, but not mmSALL4, and the embryopathy is not observed. These data are consistent with a ‘double protection’ mechanism caused by mutations in both CRBN and SALL4 preventing IMiD-dependent binding and subsequent degradation in insensitive species. (D) Heatmap comparing the sequence conservation of IMiD-dependent targets across 30 different species. High conservation is displayed as blue and low conservation is displayed as white.