Differential Analysis of Cereblon Neosubstrates in Rabbit Embryos Using Targeted Proteomics

Abstract

Targeted protein degradation is the selective removal of a protein of interest through hijacking intracellular protein cleanup machinery. This rapidly growing field currently relies heavily on the use of the E3 ligase cereblon (CRBN) to target proteins for degradation, including the immunomodulatory drugs (IMiDs) thalidomide, lenalidomide, and pomalidomide which work through a molecular glue mechanism of action with CRBN. While CRBN recruitment can result in degradation of a specific protein of interest (e.g., efficacy), degradation of other proteins (called CRBN neosubstrates) also occurs. Degradation of one or more of these CRBN neosubstrates is believed to play an important role in thalidomide-related developmental toxicity observed in rabbits and primates. We identified a set of 25 proteins of interest associated with CRBN-related protein homeostasis and/or embryo/fetal development. We developed a targeted assay for these proteins combining peptide immunoaffinity enrichment and high-resolution mass spectrometry and successfully applied this assay to rabbit embryo samples from pregnant rabbits dosed with three IMiDs. We confirmed previously reported in vivo decreases in neosubstrates like SALL4, as well as provided evidence of neosubstrate changes for proteins only examined in vitro previously. While there were many proteins that were similarly decreased by all three IMiDs, no compound had the exact same neosubstrate degradation profile as another. We compared our data to previous literature reports of IMiD-induced degradation and known developmental biology associations. Based on our observations, we recommend monitoring at least a major subset of these neosubstrates in a developmental test system to improve CRBN-binding compound-specific risk assessment. A strength of our assay is that it is configurable, and the target list can be readily adapted to focus on only a subset of proteins of interest or expanded to incorporate new findings as additional information about CRBN biology is discovered.

Graphical Abstract

Highlights

• •Targeted assay developed for CRBN neosubstrates associated with development.
• •Assessed rabbit embryos for IMiD impact after maternal exposure.
• •Identified IMiD-specific in vivo decreases in neosubstrates in embryos.

In Brief

The E3 ligase Cereblon (CRBN) is commonly used for protein degraders but carries with it the risk of degradation of additional proteins (CRBN neosubstrates) via a molecular glue mechanism of action. Here, we curated a set of 25 proteins of interest associated with CRBN-related protein homeostasis and/or embryo/fetal development and developed a targeted mass spectrometry assay for monitoring them. We applied this to rabbit embryos that were exposed maternally to immunomodulatory drugs and report in vivo decreases in many neosubstrates.

Targeted protein degradation is a rapidly growing field centered on selective removal of a target protein of interest using one of a variety of related approaches to hijack cellular cleanup machinery via exogenous compound treatment (1, 2, 3). Of note, the two most advanced classes of protein degraders, proteolysis-targeting chimeras and molecular glues, rely on hijacking E3 ligases to ubiquitinate a target protein resulting in proteasomal degradation (1). In both cases, the most used E3 ligase currently is Cereblon (CRBN); however, this ligase carries with it the potential risk for additional substrate degradation beyond the target protein. These additional non-native substrates, or neosubstrates, are best exemplified by immunomodulatory drugs (IMiDs) like thalidomide, lenalidomide, and pomalidomide which degrade a diverse variety of neosubstrates (4) through a molecular glue mechanism of action.

IMiDs are known to cause birth defects in animals and humans, and it is believed that CRBN-mediated neosubstrate degradation is a primary mechanism. This is largely based on the degradation of the neosubstrate SALL4 in rabbit embryos exposed to thalidomide (5) coupled with birth defects observed in human heterozygote loss of function mutations in SALL4 (6) that are similar to those induced by thalidomide. While the data are convincing that SALL4 degradation is important to understanding the risk of birth defects with compounds that bind CRBN, many neosubstrates are affected by CRBN binding and it is unclear whether one or more other neosubstrates would also contribute to an increased risk of birth defects. Thus, identifying CRBN-mediated neosubstrates associated with developmental biology and being able to monitor them as part of a drug development program for any compound binding CRBN is an important aspect of reducing or managing the potential risk of birth defects in humans. A number of previous studies have identified many CRBN-mediated neosubstrates from a variety of IMiD compounds across different cell lines (4, 7, 8, 9) but the relevance of these changes in vivo, specifically in the developing embryo, is largely unknown beyond the limited data with SALL4. Additionally, most of this discovery work was done using mass spectrometry (MS)based discovery proteomics, which is a powerful technique for identifying new neosubstrates, but lacks the throughput for acting as a screening assay.

To handle the higher throughput needed to evaluate candidate compounds for CRBN-mediated neosubstrate degradation, we developed a targeted MS assay. Targeted MS assays offer robust relative quantitation with the option for absolute quantitation with external calibration curves, increased selectivity and specificity compared to antibody-based detection methods, and offer high sensitivity for detection, particularly when coupled with peptide enrichment methods (10, 11, 12, 13). Additionally, with careful peptide sequence selection, methods can usually be generated such that they will be applicable across multiple species, including humans and common preclinical animal models. Here, we describe the curation of a targeted neosubstrate panel and the creation of a targeted MS method to analyze these selected neosubstrates. We then apply the methodology to rabbit embryos exposed to several IMiDs as a proof-of-concept for the assay.

Experimental Procedures

Reagents

Triethylammonium bicarbonate (TEAB, 90114), HALT protease inhibitor cocktail (78446), DTT (A39255), iodoacetamide (A39271), TFA (85183), formic acid (FA, 85178), dimethyl pimelimidate (PI21667), LC-MS methanol (A456-4), HPLC chloroform (AA43685K2), LC-MS acetonitrile (ACN, A955-4), and LC-MS water (W6-4) were purchased from Thermo Fisher Scientific. Trypsin/Lys-C (V5072) was purchased from Promega. Ammonium formate (78314), triethanolamine (T1377), ethanolamine (E9508), sodium acetate (71196), and Trizma preset crystals, pH 8.3 (T8943) were purchased from Sigma-Aldrich.

For animal dosing, thalidomide (CAS 50-35-1; 99.3% activity) was purchased from TCI America (T25245G), and pomalidomide (CAS 19171019-8; 99.9% activity) and lenalidomide (CAS 191732-72-6; 99.3% activity) were purchased from Thermo Fisher Scientific (AC465700010 and AC461590010). Thalidomide, pomalidomide, and lenalidomide were all formulated for dosing separately in aqueous vehicle consisting of a suspension of 1% (w/v) carboxymethylcellulose sodium in purified water (pH 4.4 ± 0.1) and stored refrigerated.

Antibody Column Generation

Generation of rabbit polyclonal antibodies was performed as described previously (14) at BIOSYNTH and all procedures were performed in accordance with Institutional Animal Care and Use Committee regulations. Briefly, recombinant target peptides were conjugated to keyhole limpet hemocyanin and injected into New Zealand white rabbits along with an adjuvant to induce an immune response. After collecting the sera at the end of the immunization period, titers were assessed using ELISA and the sera was purified via ligand affinity using target peptides with stable isotopes other than lysine and arginine incorporated. The use of alternate heavy labels in the purification step was to minimize impact of any passenger peptides (15). Purified antibody concentration was quantified by A280 on a Nanodrop 2000 (Thermo Fisher Scientific).

Assembly of the antibody columns was done as previously described (10, 11, 12, 13, 16). Briefly, protein G agarose beads (Thermo Fisher Scientific, PI82082) were packed into columns (IDEX Health & Science #5050IP0502100320) utilizing a high-pressure packing system (Teledyne SSI) on a Shimadzu HPLC. Anti-peptide antibodies (140–200 μg per antibody) in 100 mM sodium acetate, pH 5.5 were flowed over the column overnight to allow complete binding of the antibodies, which was confirmed by A280 measurements before and after binding. After the overnight incubation, the columns were washed with PBS and then antibodies were crosslinked to the beads with a 30 min incubation of 7.78 mg/ml dimethyl pimelimidate in 100 mM triethanolamine. The crosslinker was quenched with 100 mM ethanolamine, pH 8.0 followed by washing the columns with 200 mM Tris, pH 8.3 for 30 min. Next, the columns were washed with 0.5% TFA for 30 min and 25 mM ammonium formate for another 30 min.

Embryo Harvest from IMiD Exposed Pregnant Rabbits

All animal procedures were conducted in the spirit of Good Laboratory Practice regulations but were not audited. The animal studies were conducted in a test facility that was accredited by the Association of Assessment and Accreditation of Laboratory Animal Care, and all procedures were reviewed and approved by an Institutional Animal Care and Use Committee (Animal use protocol number GTN-2017-01202). For both embryo-harvesting and modified developmental toxicity studies, time-mated female New Zealand white rabbits (approximately 4–6 months old and 3.0–3.5 kg at dose administration; Covance Research Products, Inc) were housed individually and were provided approximately 150 g/day of Certified Rabbit Diet 2030C (Envigo Teklad Global Diet) and locally sourced water (purified through reverse osmosis) ad libitum. Environmental conditions were set to maintain relative humidity at 30% to 70% and temperature of 61 to 72 °F. Room lighting was set to provide a 12-h light/dark cycle.

To explore potential changes with the targeted panel of neosubstrates, pregnant rabbits (n = 3/group) were administered thalidomide (180 mg/kg/day), pomalidomide (180 mg/kg/day), or lenalidomide (20 mg/kg/day) by oral gavage once daily from gestation day (GD) 7 to 11. These doses were selected based on literature summarized below to maximize potential for developmental toxicity while minimizing maternal toxicity in order to identify changes in embryo proteins on GD 12 (mid-organogenesis) (5, 17) (https://www.accessdata.fda.gov/drugsatfda_docs/nda/2005/021880s000_Revlimid_PharmR.pdf, https://www.accessdata.fda.gov/drugsatfda_docs/nda/2013/204026Orig1s000PharmR.pdf). Oral doses of 180 mg/kg/day of thalidomide once daily from GD 7 to 19 have been shown to result in classic thalidomide developmental toxicity in rabbits (17, 18). Thalidomide at 150 to 180 mg/kg/day for shorter periods (e.g., GD 7–8, 7–11, and 7–15) also results in classic thalidomide malformations but with less embryo-fetal lethality (18). The seminal publication demonstrating SALL4 protein degradation in GD 12 rabbit embryos was at 180 mg/kg/day of thalidomide starting on GD 7 (5). Pomalidomide exposure resulted in dose-dependent increases in fetal malformations, embryo-fetal lethality, and maternal toxicity increased at 100 and 250 mg/kg/day (https://www.accessdata.fda.gov/drugsatfda_docs/nda/2013/204026Orig1s000PharmR.pdf). Previous data with lenalidomide demonstrated that 20 mg/kg/day was the highest tolerated dose since mortality and abortions were observed at higher doses; lenalidomide-related developmental toxicity at this dose manifested as lower fetal body weight, higher embryo-fetal lethality, and higher developmental variations (17).

Clinical signs, body weight, and food consumption were monitored throughout the study. Euthanasia occurred on GD 12 via an intravenous injection of a barbiturate euthanasia solution followed by exsanguination. The gravid uterus was removed and the number of corpora lutea in each ovary were recorded. Segments of the uterus containing embryos were placed in a Petri dish with Dulbecco's phosphate-buffered saline. Each uterus was opened along the antimesometrial side of each implantation site to expose the embryo. At least four viable embryos were collected per litter, separated from the visceral yolk sac and extraembryonic membranes, and each embryo was placed into separately labeled microfuge tubes. Excess Dulbecco's phosphate-buffered saline was removed, and the embryo was flash-frozen in liquid nitrogen.

LC-MS/MS Sample Preparation

Frozen embryos were transferred to 1.5 ml Eppendorf tubes (022431081) preloaded with stainless steel beads and were homogenized in 300 μl of homogenization buffer (100 mM TEAB pH 8.0, 100 mM NaCl, 1X HALT). A Geno/Grinder 2010 instrument (SPEX SamplePrep) was used with a 2 min cycle of 1300 rpm shaking to disrupt the embryos. Following homogenization, the supernatant was transferred to a new tube with 300 μl of 10% SDS in it (final 5% SDS concentration). Samples were heated at 90 °C for 10 min and then sonicated in a cup-horn sonicator for 30 s with 1 s pulses at 90% amplitude. The heating and sonication steps were repeated once. Samples were frozen at −80 °C until additional processing was performed.

Samples were thawed and one additional round of heating and sonication was performed as above. Samples were then centrifuged at room temperature at 15,000g for 10 min. A pierce bicinchoninic acid assay (Thermo Fisher Scientific, 23225) was performed according to manufacturer instructions at a 1:5 dilution to determine protein concentration. Aliquots of 125 μg of protein in a final 95 μl volume were prepared in new 1.5 ml Eppendorf tubes using 50 mM TEAB, 50 mM NaCl, 5% SDS as additional diluent. To each sample, 2 μl of 250 mM DTT was added and samples were incubated at 56 °C for 45 min. Samples were cooled to room temperature and 3 μl of 500 mM iodoacetamide was added to each sample and incubated in the dark at room temperature for 30 min.

Samples were then precipitated via methanol chloroform precipitation (19). To each 100 μl sample, 400 μl of methanol, 200 μl of chloroform, and 300 μl of water were added. Samples were vortexed briefly and then centrifuged at 3000g for 5 min. The supernatant above the resulting protein disk was discarded and samples were washed with 300 μl of methanol and centrifuged at 9000g for 2 min. The supernatant was again discarded, and the samples were washed with 500 μl of methanol and centrifuged at 9000g for 2 min. The supernatant was discarded, and samples were resuspended in 100 μl of resuspension buffer consisting of 50 mM TEAB plus stabile isotope labeled (SIL) peptides. Protein pellets were dissociated by striking across a tube rack five times and then sonicating for 5 to 7 1 s pulses at 70% amplitude in a cup-horn sonicator. This process was repeated 1 to 2 times as need to completely dissociate the protein pellet such that no visible chunks remained. Then, 2 μg of a trypsin/Lys-C mix was added to each sample (1:50 enzyme to protein ratio), and samples were placed in a 37 °C incubator overnight for proteolytic digestion.

Following overnight digestion, samples were split for immunoaffinity (IA)-enriched and nonenriched runs. For the unenriched samples, 20 μl of each digest was transferred to a 1 ml deep well 96-well plate (Eppendorf, 2231000919) which already contained 5 μl of 10% TFA and 25 μl of 0.1% TFA. For these samples, 10 μl (4 μg) of the resulting diluted samples were injected on the instrument for analysis. For the enriched samples, 80 μl of each sample were transferred into a 1 ml deep well 96-well plate and 30 μl (30 μg) of each sample was injected on the instrument for simultaneous enrichment and analysis.

Development and Analytical Validation of Targeted MS Assays

SIL Peptides

Peptide sequences for target proteins were selected to be proteotypic and conserved across as many of the following species as possible: human, rabbit, cynomolgus macaque, dog, pig, and mouse (Supplemental Table S1). For some targets, conservation across all species was limited and preference was given to rabbit in this dataset. Additional peptides covering nonconserved species can also be added to the assay for future work in other species. Preference was given for peptides shown to be detectable by MS in public databases (https://www.proteomicsdb.org/, https://massive.ucsd.edu/) or in-house data. For targets where antibody enrichment was employed, peptide sequences were also assessed for immunogenicity to prioritize sequence selection. SIL peptides (heavy lysine or arginine) were purchased (BIOSYNTH) with >99% isotopic purity and >95% peptide purity. When present in the sequence, cysteine residues were synthesized as carbamidomethylated versions. For most sequences, extended versions of the peptides were purchased such that the N and/or C termini of the peptides were extended by 2 to 3 amino acids based on the human sequence (or rabbit sequence where not conserved). Peptide concentration was determined by amino acid analysis performed by the vendor (BIOSYNTH), and 1 nmol aliquots were stored as lyophilized powder at −80 °C until needed. Peptides were resuspended in 0.1% FA, 30% ACN, and combined to create a mastermix at a 40X concentration. On column spike levels for each SIL are listed in Supplemental Table S1. Because all targeted peptides have a SIL and relative quantitation was employed, this assay is classified as tier 2 (20).

Targeted MS Methods

Digested peptides were analyzed by LC-MS/MS using a Dionex Ultimate 3000 UPLC (Thermo Fisher Scientific) coupled online to an EASYSpray ion source (Thermo Fisher Scientific) and Eclipse Tribrid mass spectrometer (Thermo Fisher Scientific). The IA enrichment method was performed as previously described (10, 11, 12). Briefly, peptides were first flowed over an anti-peptide antibody column. The enriched peptides were retained, and the column was washed with 25 mM ammonium formate followed by 300 mM ammonium formate. Bound peptides were eluted with 0.5% TFA onto a C18 trap column (Thermo Fisher Scientific, 174500) heated to 50 °C where they were desalted online. The trap column was then switched in-line with a 15 cm EASYSpray C18 reverse phase analytical column (Thermo Fisher Scientific, ES900) heated to 50 °C. Peptides were eluted with a linear gradient of solvents A (0.1% FA in water) and B (0.1% FA, 90% ACN) from 6 to 35% B over 10 min at a flow rate of 400 nl/min. Nonenriched runs were performed similarly except that peptides were loaded directly onto the C18 trap column for online desalting. The trap column was then switched in-line with the same analytical column as above. Peptides were eluted with a linear gradient of solvents A and B from 3 to 7% B over 2.5 min, followed by a ramp to 40% B over an additional 17.5 min all at a flow rate of 400 nl/min.

Peptides were ionized at 1.7 kV in positive mode and tMS2 scans were performed for targeted peptides within scheduled retention time windows using automatic gain control and maximum injection time settings customized for each peptide (Supplemental Tables S2 and S3). An isolation window of 0.7 m/z was used for both methods. Immunoenriched samples were acquired at a 50,000 MS2 resolution and the nonenriched samples were acquired at a 30,000 MS2 resolution, both in profile mode.

Informatics Methods

All data were analyzed in Skyline Daily (21) with the latest version of the software (https://skyline.ms/). Peak integration boundaries were adjusted as needed to integrate complete peaks and exclude interferences where present. For endogenous peptides that were not detected in a sample, missing values were recorded. For peak identification, 2 to 3 transitions were required to be present in both light and heavy peptides at the same retention time and with less than 10 ppm mass error. For final peak area analysis, the top transition with the highest abundance was used and the light/heavy ratio was calculated in Skyline. Data were exported to Excel for further analysis. Additional normalization to the global standard peptide from GAPDH was performed. To do this, the average light/heavy ratio for GAPDH over the whole dataset was determined and then each individual sample abundance of GAPDH was divided by the average to generate a ratio to average for each of the runs. All other peptides in each run were then divided by this scaling value to normalize any variability in sample prep.

For graphical visualization of individual proteins in GraphPad Prism (v9.5.1, https://www.graphpad.com/), normalized values were scaled by the average of the undosed controls and multiplied by 100. Average fold changes were visualized in a heatmap using the Morpheus tool (https://software.broadinstitute.org/morpheus/).

Reverse Response Curve

A pool of undosed rabbit embryo lysate was created and split out into 30 aliquots of 100 μg of protein per aliquot. A 10-point curve of the heavy SIL peptides spanning a range from 65.5 amol to 1000 fmol on column (Supplemental Table S5A) was created and spiked into the endogenous matrix in process triplicate. Samples were processed and analyzed via LC-MS/MS as described above. Data integration was performed in Skyline as described above, with two differences: first the heavy/light ratio was calculated manually in Excel, and second, the additional global standard normalization step was not performed. The heavy/light ratios were then graphed in Prism using a nonlinear least squares regression with 1/Y weighting in most cases. If the resulting curve did not sufficiently fit the data, a 1/X2 weighting was used. In one case (SLC16A1), a sigmoidal fit with 1/Y weighting was used to generate the standard curve. Graphs were plotted with log₁₀ transformed axes to assist visualization, but all calculations were done on untransformed data.

Modified Rabbit Developmental Toxicity Study With Thalidomide

Following an acclimation period, presumed pregnant female rabbits were administered vehicle (n = 5) or 180 mg/kg/day thalidomide (n = 10) by oral gavage once daily from GD 7 to 11. Clinical signs, body weight, and food consumption were monitored throughout the study. Euthanasia occurred on GD 29 via an intravenous injection of a barbiturate euthanasia solution followed by exsanguination. A gross examination of the abdominal, thoracic, and pelvic viscera was performed. The gravid uterus was removed and weighed and the number of corpora lutea in each ovary, and the number, type, and position of implantation sites were recorded. Uteri of apparently nonpregnant rabbits were placed in 10% ammonium sulfide solution to confirm absence of implantation sites (22). Viable fetuses were removed from the uteri and individually weighed. A detailed external examination of each fetus was conducted, including assessment of palatal closure. External, visceral, and skeletal findings were recorded as developmental variations, malformations, or unclassified (23). All rabbit fetuses were examined for visceral abnormalities using a modification of the method described by Stuckhardt and Poppe (24) and sexed internally during the visceral exam. Following visceral exams, all the fetuses were eviscerated, fixed in ethanol, macerated in potassium hydroxide, and stained with alizarin red for skeletal examinations.

Experimental Design and Statistical Rationale

The reverse response curve was conducted in process triplicate created from a common pooled lysate. The IMiD study utilized three pregnant rabbits per IMiD and from each rabbit, four embryos were analyzed. The control group consisted of two pregnant rabbits which were not exposed to any compound and seven embryos were examined from the first rabbit and five embryos from the second rabbit. Comparative statistical analysis of each compound compared to control was performed in Prism using a two-way ANOVA with compound and litter as factors to account for covariates. A multiple comparison adjusted p value was calculated in R using the p.adjust function with the false discovery rate (FDR) method (25). Log₂ transformations of the average fold changes of each compound over control were calculated in Excel. Comparisons with a FDR of 0.01 or less were considered significant. For the developmental toxicity study, five pregnant female rabbits were administered vehicle and 10 were administered thalidomide; all rabbit fetuses were examined for visceral abnormalities.

Results

Selection of CRBN Substrate and Neosubstrate Proteins for Assay Inclusion

With a goal of developing an assay for assessment of CRBN-mediated neosubstrate risk of potential developmental toxicity for CRBN-binding compounds, we began by first surveying the literature for existing information on previously identified neosubstrates. Much of this information came from discovery proteomic analysis of cells treated with different IMiD compounds plus one previously developed targeted MS assay which examined CRBN and eight neosubstrates (26). We further prioritized the reported CRBN neosubstrates based on known biology during development, specifically including those proteins with a role in embryo-fetal or early postnatal survival (e.g., decreased survival in homozygous KO mice), evidence of abnormal morphological development (e.g., malformations), or other impact (e.g., delayed growth) with altered protein expression.

Although KO mouse data have been shown to be highly useful for predicting malformations and/or embryo-fetal lethality in embryo-fetal development studies (Catlin 2020), the publications describing the phenotype for these models are often limited making it difficult to discern the more detailed biological effects. For example, it is easier to define fetal mortality as this would be manifested as altered mendelian ratios or overall decreased numbers of pups at birth; however, if mortality is only observed postnatally and no cause is described, it is difficult to determine if this was due to a developmental effect in utero or due to other factors such as lack of maternal care. In addition to individual literature references, there are on-line databases, such as MGI (www.informatics.jax.org/) that contain variable amounts of information on KO animals that can be used to provide some information on a target but can lack sufficient detail to discern causative effects on embryo-fetal development so caution is needed not to overinterpret confidence in this information. Based on variability in detail available from the literature, source references were provided in Table 1 for each target protein.

Table 1.

Selection of CRBN substrate and neosubstrate proteins for assay inclusion

Protein	Change with IMIDs (literature)	Developmental biology association (literature)
BSG (CD147)	Protein lowered (destabilized though nonubiquitin function) upon binding of IMiDs (Thal, Pom, Len) to CRBN (39)	Mouse knockout embryolethal, small, eye abnormalities (48, 49); Zebrafish knockdown malformations similar to thalidomide (head, fin, eyes) (39)
CSNK1A1 (CK1alpha)	Degraded by Pom (4, 35) and Len (4, 26, 35)	Mouse knockout embryolethal (50); mouse knockdown (>50%) decreases hematopoetic stem cell expansion and survival (51)
DTWD1	Degraded by Pom (4)	No apparent phenotype in mouse knockout (52)
ESCO2	RNA decreased by Thal (42)	Mouse knockout embryolethal (53), human mutations associated with phocomelia, craniofacial anomalies (54)
GLUL	Endogenous CRBN substrate, no impact from IMiDs (44)	Mouse knockout embryolethal, human deficiency associated with brain malformations and neonatal death (55)
GSPT1	Degraded by CC-885 (45, 46)	Mouse knockout viable at birth but die postnatally (52)
GZF1	Degraded by Pom (4)	Mouse knockout viable at birth but die postnatally (52)
IKZF1 (ikaros)	Degraded by Pom (8), Len (45), and CC-885 (45)	Mouse knockout viable at birth but die postnatally; lacks lymphoid cells (56)
IKZF2 (helios)	Degraded by CC-220 (41), a CRBN PROTAC (57), and NVP-DKY709 (58)	Mouse knockout viable at birth but die postnatally (59)
IKZF3 (aiolos)	Degraded by Pom (8) and Len (4, 8, 9, 26, 41)	Mouse knockout viable but with activated B cell phenotype (60)
IKZF4 (eos)	Degraded by CC-220 (40, 41)	Mouse knockout viable (61)
MEIS2	Endogenous CRBN substrate, competed by IMiDs (43)	Mouse knockout embryolethal (62); Human mutation heart defects (63)
PATZ1	Degraded by Pom (37)	Mouse knockout embryolethal, malformations (cranial, nervous system, cardiac), growth retardation (64)
RAB28	Degraded by Pom (4) and Len (4, 8)	Mouse knockout, eye abnormalities (65); Human mutation retinal dystrophy (66)
SALL4	Degraded by Thal (4), Pom (4, 38, 67), and Len (4, 38); Rabbit embryo SALL4 levels lower after exposed to Thal (5)	Human mutations in syndromes with overlapping clinical presentations to thalidomide embryopathy, including phocomelia [6]; Mouse knockout embryolethal, mouse heterozygote malformations (anorectal, heart, brain, kidney) (68, 69); mouse conditional knockdown in mesoderm induces limb defects (70); rabbit IMiD-induced lower expression in malformed embryos [5]
SLC16A1 (MCT1)	Protein lowered (destabilized though non-ubiquitin function) upon binding of IMiDs (Thal, Pom, Len) to CRBN (39)	Mouse knockout embryolethal (71)
TBX5	Direct Thal-TBX5 binding (no degradation) (29); Decreased expression by Thal in chick embryos and human embryonic fibroblasts (30)	Mouse knockout embryolethal, cardiac malformations (31); mouse heterozygote embryolethal, cardiac, and limb malformations (31); mouse conditional knockout in limb buds, perinatal lethal, absent forelimbs and sternum (32); zebrafish knockdown absent or malformed fins (33); chick embryo knockdown (53%) limb truncations and malformations (33); human mutation limb and heart defects (34)
TP63	Degraded by Thal (28)	Mouse knockout viable at birth but die postnatally, skin defects, absent limb, and limb malformations (72); human mutation hand and foot malformations (73)
WIZ	Degraded by Pom (4)	Mouse knockout late embryonic/perinatal lethal, craniofacial malformations, growth delay (74)
ZBTB16 (PLZF; ZFP145)	Degraded by Thal (38), Pom (38, 75), Len (38), CC-3060 (76), and CC-647 (76)	Mouse knockout limb and axial skeleton defects (77)
ZBTB39	Degraded by Pom (4)	No information found
ZFP91	Degraded by Thal (4), Pom (7, 41), Len (7), and CC-220 (4); degraded by Thal in rabbit testis (5)	Mouse knockout viable at birth but die postnatally, abnormal embryo size, skin defects (52)
ZNF276	Degraded by Pom (4, 41) and Len (41)	No information found
ZNF692	Degraded by Thal (4), Pom (41), and Len (4, 26, 41)	No information found
ZNF827	Degraded by Pom (41) and Len (4, 41)	Mouse knockout perinatal lethality (52)

The association of each target protein with developmental biology and any known impact on abundance level from IMiD exposure is noted. For IMiD associated abundance changes, the significance criteria in the original cited paper were used to determine inclusion in the table.

CRBN, Cereblon; ImiD, immunomodulatory drug; Len, lenalidomide; Pom, pomalidomide; Thal, thalidomide.

In addition to neosubstrates, we also included a few known substrates of CRBN where protein levels had been shown to be perturbed by IMiD binding. Overall, we selected a panel of 25 target proteins (Table 1) in addition to CRBN. This panel includes proteins that have been shown to change following exposure to one or more IMiDs based on the criteria used in the original papers cited (24 proteins) and one endogenous substrate which has no evidence of IMiD impact (GLUL). Furthermore, these targets include proteins that demonstrate roles in developmental biology either prenatally (13 proteins) or postnatally (six proteins) as well as those that have no identified role in development (six proteins). While this is not an exhaustive list of all potential CRBN neosubstrates reported in the literature, our target list includes the most well supported CRBN neosubstrates, particularly those currently known to have links to developmental toxicity.

Development of Targeted MS Assay

To enable robust quantitation of the protein targets selected, we chose to develop a parallel reaction monitoring targeted MS assay on a high-resolution, accurate mass instrument (Fig. 1A). We selected representative peptides from each target through a multifactorial approach considering proteotypicity, cross-species conservation, and preexisting data as discussed in the Experimental Procedures section and purchased heavy labeled synthetic standards for each of the target peptides. Because some of the targets are low-abundant proteins, we utilized a peptide IA enrichment step to improve detection and quantitation. This IA step was implemented as an online enrichment during the MS run to improve assay throughput via creating an antibody column that was switched inline during the targeted MS runs. To simplify sample handling, the assay was designed so that each sample can be split postdigestion and analyzed via two injections on the instrument: one with IA enrichment and one without enrichment in order to monitor all targets (Fig. 1).

Fig. 1.

Schematic diagram of experimental workflow.A, targeted MS workflow. B, timeline of animal dosing and collections for both targeted MS and developmental toxicity assessment. Figure created with BioRender.com. MS, mass spectrometry.

We first optimized the amount of material to analyze and selected 30 μg as the input for the IA step (Supplemental Fig. S1A, Supplemental Table S4A). We then ran the SIL peptides in both the enriched and nonenriched assays and assessed IA enrichment as the percent of nonenriched signal present in the IA runs for each peptide and found that all antibodies performed well with an enriched/unenriched percentage of 50% or higher (Supplemental Fig. S1B and Supplemental Table S4B) which gave us confidence in the method. Enrichment above 100% likely stems from differences in how the peptides behave in each assay without a complex matrix present, as was the case for this assessment.

Characterization of Targeted MS Assay Performance

We next characterized the performance of the whole assay by conducting a reverse response curve experiment (27). A 10-point curve of the heavy SIL peptides was created and spiked into the endogenous matrix in process triplicate. The IA and nonenriched methods were run on all samples (Fig. 2A and Supplemental Table S5) and response curves of the heavy SIL peptides compared to the endogenous signals were generated. The goal of this response curve experiment was not to determine limits of acceptance criteria, as this assay was designed as a tier two fit-for-purpose relative quantitation assay (20), but to better understand the robustness of the assay and its application to the rabbit embryo matrix.

Fig. 2.

Characterization of targeted MS assay performance. Reverse response curves were performed with heavy isotope-labeled peptides spiked into a common matrix sample and normalized to the endogenous peptide. A, in total, multiplexed acquisition for 16 peptides in IA and 20 peptides in the nonenriched runs were performed. Shown are example response curves for (B) SALL4 in the immunoenriched runs and (C) DTWD1 in the nonenriched runs. D, the number of points per peak was assessed for each endogenous peptide across 30 runs to ensure that sufficient data was collected to describe the peak (>10 points, dotted line on graph). MS, mass spectrometry.

An example of a response curve for the IA assay is shown in Figure 2B for SALL4 and an example for the nonenriched assay is shown in Figure 2C for DTWD1. Both show good linear response with detectable signal from the heavy peptide at 0.4 fmol and 0.065 fmol on column, respectively. Response curves for the other peptides in the assay which were detectable endogenously are shown in Supplemental Figs. S2 and S3. Not all peptides targeted could be detected in the rabbit embryo lysates, but most (29/34) were. In addition to assay linearity, we assessed the quantitative robustness of the assay by determining the number of points across the peak for each endogenous peptide detected in the response curve experiments. We set a minimum cut-off of 10 points per peak which was achieved by all the peptides as shown in the boxplot in Figure 2D.

Abundance Changes in CRBN Neosubstrate Proteins Following IMiD Exposure

We next applied this characterized assay to a set of embryonic samples exposed indirectly following maternal IMiD administration (Fig. 1B). Rabbit embryos for analysis via the targeted MS assay were removed on GD 12 from undosed rabbits or rabbits dosed with thalidomide, pomalidomide, or lenalidomide once daily from GD 7 through 11. In all, we were able to quantify 20 of the protein targets plus CRBN in these embryos (Supplemental Table S6). SALL4 was the first of these proteins that we examined in detail as it is a known target of these IMiDs in vivo and as demonstrated in Figure 3A, loss of this protein was observed in all dosed animals compared to the controls. Tight grouping of embryos from the same mother was observed and mother-dependent effects could also be seen, particularly in rabbit three of pomalidomide where the impact of dosing appeared to be less than the other rabbits (Fig. 3A). Ikaros (IKZF1) is likewise a known IMiD target and was observed here to decrease significantly in response to pomalidomide (Figs. 3B and 4).

Fig. 3.

Abundance changes in selected CRBN neosubstrate proteins following IMiD dosing. Pregnant rabbits were dosed once daily from GD7-GD11 with thalidomide (180 mg/kg/day), pomalidomide (180 mg/kg/day), lenalidomide (20 mg/kg/day), or left undosed. Abundances of each protein target in each embryo were assessed approximately 24 h post final dose on GD11 and were grouped by mother (denoted as R1, R2, and R3) and graphed as percent of the average signal of the undosed embryos. Four embyros were analyzed for each mother except for the untreated controls where seven and five embryos were analyzed for R1 and R2, respectively. Individual embryo values are depicted as diamonds and the mean value per pregnant mother is represented by the bar graph. A, SALL4 shows differing responses to each IMiD with thalidomide having the greatest overall decrease in abundance levels. B, IKZF1 was decreased in rabbits dosed with pomalidomide. C, TP63 was observed to have a modest decrease following thalidomide exposure. D, DTWD1 showed the most robust decrease of all targets and was most impacted by pomalidomide. CRBN, Cereblon; ImiD, immunomodulatory drug.

Fig. 4.

Heatmap summary of all targets following IMiD dosing. Shown are the average log₂ fold changes of each compound compared to the undosed embryos (n = 12 for all groups). Proteins marked with an asterisk are considered significant based on an FDR of 0.01 or less from a two-way ANOVA using litter as a covariate. FDR, false discovery rate; ImiD, immunomodulatory drug.

TP63 exhibited a small decrease with thalidomide exposure (Figs. 3C and 4), consistent with a previous report (28). Similar results were seen for DTWD1 where rabbit three of pomalidomide was again an outlier (Figs. 3D and 4), with otherwise relatively uniform decreases in protein abundance across replicates. Interestingly, DTWD1 was the protein with the largest decrease in abundance among all neosubstrates with many other proteins also exhibiting a significant (FDR < 0.01) impact of IMiD exposure on abundance (Fig. 4). In total, 14 proteins in the panel had a significantly decreased abundance following thalidomide exposure with the other compounds having slightly fewer targets decrease (Fig. 4). However, pomalidomide dosing resulted in two proteins significantly increasing in abundance (including CRBN), resulting in a similar number of differential proteins as thalidomide (Fig. 4). These increased abundances may represent a change in protein homeostasis unique to pomalidomide, though additional work would be needed to confirm this possibility. While there were many proteins that were similarly decreased by all three IMiDs, no compound had the exact same neosubstrate degradation profile as another. Individual graphs of all quantified proteins are provided in Supplemental Fig. S4.

Developmental Toxicity of Thalidomide in Rabbit

To provide physiological context to the neosubstrate data, an in vivo developmental toxicity study with thalidomide was conducted in pregnant rabbits to confirm observations in the literature at this dose and duration. Thalidomide was selected based on preexisting data in the literature and because it gave the strongest signal in the neosubstrate analysis among the three compounds. A similar 5-day dosing period from GD 7 to 11 was followed by a nondosing period out to GD29, after which fetal development impacts were assessed (Fig. 1B). There were no thalidomide-related clinical signs, but lower maternal body weight change (0.14x control from GD 12–16 and 0.4x control from GD 16–21) and food consumption (0.77x control from GD 16–21) were observed after the end of the dosing phase. These differences from control did not persist to the end of the study. One thalidomide-dosed doe aborted on GD 29, but this was not considered to be related to thalidomide due to the single incidence and lack of literature evidence demonstrating an association between thalidomide and abortions in rabbits at this or higher doses. Thalidomide-related higher postimplantation loss was observed with concomitant lower mean number of viable fetuses per litter (7) compared with control (9). Thalidomide exposure also resulted in lower fetal body weights (0.89x control).

Classic thalidomide-related fetal dysmorphology was observed in all dosed litters and nearly all fetuses in each litter (Supplemental Table S7). Across external, visceral, and skeletal fetal evaluations, dysmorphology was primarily observed in limbs/long bones including digits, the head (e.g., small brain, dilated cerebrum ventricles, cleft palate, cranioschisis, and craniomeningocele), the eye, major organs (primarily kidney, heart, and vessels), abdominal region (diaphragmatic hernia, omphalocele), and the tail region (Fig. 5).

Fig. 5.

Thalidomide-induced rabbit fetal dysmorphology. Representative images are shown of rabbit fetal external and skeletal abnormalities induced following in utero thalidomide exposure. A, arrow indicates absent forepaw digit with only small skin tag present. B, encircled region shows absent forepaw digits with only two of the normal five present. C, arrow indicates fusion of sacral centrum to caudal centrum. D, arrow denotes a short ulna. E, arrows point to bent long bones (bilateral femurs and fibula). Area marked with an asterisk is to note that hindpaw was present in animal but removed for image generation.

Discussion

Here, we describe the development of a targeted MS method for monitoring CRBN and some known neosubstrates/substrates and how it was used to quantify relative abundance in embryos exposed following maternal IMiD administration to pregnant rabbits, a species sensitive to thalidomide-induced malformations. We conducted a fit-for-purpose assessment to evaluate assay performance and then applied the assay to an in vivo study of pregnant rabbits that were administered IMiDs once daily from GD 7 to 11. The resulting data showed decreases in several different neosubstrate proteins in GD 12 rabbit embryos approximately 24 h following maternal IMiD dosing, including the previously demonstrated decrease in SALL4 by immunohistochemistry (5). Consistent with previous data, maternal thalidomide administration from GD 7 to 11 resulted in expected fetal anomalies on GD 29, providing a potential phenotypic link to the neosubstrate changes observed on GD 12 following GD 7 to 11 administration of 180 mg/kg/day thalidomide.

In examining our data, multiple patterns relative to IMiD changes and developmental biology emerged and thus the results were divided into four groups based on these patterns. The groups were defined based on literature associations with prenatal development and observed patterns in our rabbit embryo data. Thus, groups one and two consist of proteins that were observed to change in response to IMiD dosing in the rabbit embryos and which did (group one) and did not (group two) have an association with prenatal development. Conversely, proteins in groups three and four were not observed to change in the rabbit embryo data and did (group three) or did not (group four) have associations with prenatal development in the literature.

In group one, we noted that nine proteins (SALL4, ZBTB16, PATZ1, RAB28, CSNK1A1, WIZ, TP63, SLC16A1, and TBX5) decreased in embryos following maternal administration of one or more IMIDs and appear to be important in developmental biology; for all there is evidence of decrease by one or more IMiDs in the literature (Table 2). As expected, SALL4 was degraded by all IMiD treatments, and we also observed a similar ranking of protein degradation as previously reported in vitro with thalidomide > pomalidomide > lenalidomide (4).

Table 2.

Neosubstrate changes observed in rabbit embryos contextualized with existing knowledge of changes with IMiDs and role in developmental biology

Group	Protein	Change with IMiDs (literature)	Association with prenatal development	Change with IMiDs in rabbit embryo following maternal exposure
1	SALL4	↓ (Thal, Pom, Len)	Yes	↓ (Thal, Pom, Len)
	ZBTB16 (PLZF; ZFP145)	↓ (Thal, Pom, Len, CC-3060, CC-647)	Yes	↓ (Thal, Pom, Len)
	PATZ1	↓ (Pom)	Yes	↓ (Thal, Pom, Len)
	RAB28	↓ (Pom, Len)	Yes	↓ (Thal, Pom, Len)
	CSNK1A1 (CK1alpha)	↓ (Pom, Len)	Yes	↓ (Thal, Pom, Len)
	WIZ	↓ (Pom)	Yes	↓ (Thal, Pom)
	TP63	↓ (Thal)	Yes	↓ (Thal)
	SLC16A1 (MCT1)	↓ (Thal, Pom, Len)	Yes	↓ (Thal, Len)
	TBX5	↓ (Thal)	Yes	↓ (Thal)
2	GZF1	↓ (Pom)	Noa	↓ (Thal, Pom, Len)
	ZFP91	↓ (Thal, Pom, Len, CC-220)	Noa	↓ (Thal, Pom, Len)
	IKZF1 (ikaros)	↓ (Pom, Len, CC-885)	Noa	↓ (Pom)
	IKZF4 (eos)	↓ (CC-220)	No	↑ (Pom, Len)
	ZNF276	↓ (Pom, Len)	NA	↓ (Thal, Pom, Len)
	ZBTB39	↓ (Pom)	NA	↓ (Thal, Pom, Len)
	DTWD1	↓ (Pom)	No	↓ (Thal, Pom, Len)
3	MEIS2	↑ (Competed by IMiDs)	Yes	No
	BSG (CD147)	↓ (Thal, Pom, Len)	Yes	No
	ESCO2	↓ (Thal)	Yes	N/D
	GLUL	No	Yes	No
4	GSPT1	↓ (CC-885)	Noa	No
	ZNF827	↓ (Pom, Len)	Noa	N/D
	IKZF2 (helios)	↓ (CC-220, TPD)	Noa	N/D
	IKZF3 (aiolos)	↓ (Pom, Len)	No	N/D
	ZNF692	↓ (Thal, Pom, Len)	NA	N/D

ImiD, immunomodulatory drug; Len, lenalidomide; NA, no information found; N/D, not detected; Pom, pomalidomide; Thal, thalidomide.

^aViable at birth but developmental effects postnatally.

Previously, TBX5 has been shown to bind directly to thalidomide (29) and have decreased RNA levels following thalidomide exposure in chick embryos (30). Additionally, mutations in this gene are associated with a strong developmental toxicity signal (embryolethality and cardiac/limb malformations) in several species (mice, zebrafish, chick embryos) (31, 32, 33), as well as with Holt-Oram syndrome in humans which phenocopies thalidomide embryopathy (34). Our data show rabbit embryos exposed to thalidomide have a decrease in TBX5 abundance, though it is unclear if this is a direct effect of TBX5-thalidomide binding, a result of CRBN E3 ligase activity, or some other upstream effect resulting in decreased gene expression.

Another group one protein, CSNK1A1, has been previously reported as a specific neosubstrate for lenalidomide (4, 35, 36), but here we also observed degradation of this protein by all three IMiDs with lenalidomide displaying the greatest degree of decrease. Additionally, PATZ1 and RAB28 were also degraded by all three IMiDs in our data, while previous literature has demonstrated PATZ1 degradation by pomalidomide in a zinc finger library screen in U2OS cells (37) and RAB28 degradation by lenalidomide in MM1.s cells and human embryonic stem cells, and pomalidomide in those same cells plus Kelly and SK-N-DZ cells (4, 8). Likewise, WIZ was degraded by two IMiDs in our samples, only one of which was previously reported (Table 2). ZBTB16 was also degraded by all three IMiDs in our samples, consistent with reports of degradation by these same IMiDs in HuH7 and HEK293T cells (38).

SLC16A1 was previously shown to be stabilized along with BSG by CRBN through a non-E3 function of the protein that was inhibited by IMiD binding (39). Consistent with this, in our data SLC16A1 was decreased by two of the IMiDs dosed. The two main isoforms of TP63 (ΔNp63 and TAp63) have both been shown to be degraded by thalidomide (28) and our rabbit embryo data supports this with a modest but significant decrease in abundance. The peptide used in our assay is shared between both major isoforms of TP63 and therefore our data represents the average change in abundance for both.

Unlike thalidomide, lenalidomide administration to pregnant rabbits did not result in fetal malformations (17). No neosubstrate was only decreased by lenalidomide, this may be expected since both thalidomide and pomalidomide exhibit similar developmental toxicity effects of lenalidomide (reduced fetal growth, increased fetal developmental variations, and increased embryo-fetal lethality) in addition to fetal malformations. However, some of the developmentally important group one neosubstrates are degraded in the thalidomide and/or pomalidomide exposed rabbit embryos but not lenalidomide (WIZ, TBX5, and TP63). This may highlight potentially important differences in CRBN neosubstrate degradation patterns in the rabbit embryo with different CRBN-binding compounds, including the three IMiDs evaluated here. These differences may help explain varied developmental phenotypes and potentially may help identify and explain differences in developmental toxicity of other CRBN-binding compounds or specific doses/exposures of some compounds.

Group two consisted of seven proteins (GZF1, ZFP91, IKZF1, IKZF4, ZNF276, ZBTB39, and DTWD1) which were modulated by at least one IMiD in both rabbit embryos and literature reports (Table 2) but have unclear links to prenatal development based on KO mouse data. We did note that while KO mice for IKZF1, GZF1, and ZFP91 all had viable offspring, preweaning lethality was observed for each of these, indicating a potential link to prenatal developmental that manifests after birth. However, the other four proteins had either no information or no abnormal phenotype following KO in mice (Table 1). Among this group of proteins, IKZF4 is notable as its abundance increased upon treatment with two IMiDs. Available literature has shown a decrease in response to CC-220 (40, 41) but no changes with the IMiDs used here.

Also, of note in this group is DTWD1, which showed the largest decreases in abundance levels following all IMiD treatments (Fig. 4). This is in contrast to the pattern observed in vitro when it was first discovered to be a neosubstrate and was only degraded by pomalidomide (4). Consistent with this finding, we observed the greatest impact on DTWD1 protein levels in the pomalidomide dosed embryos. Along with DTWD1, all other proteins in this group aside from IKZF1 and IKZF4 were decreased following dosing with all three IMiDs, suggesting that these targets are also a robust response group for this class of compound, even if an association with development is unclear.

Group three contained four proteins that had an association with developmental biology and evidence of change by IMiDs in the literature (MEIS2, BSG, ESCO2, and GLUL) but were either not impacted or not detected in the rabbit embryos under the conditions of our study (Table 2). ESCO2 has been shown to be transcriptionally impacted by thalidomide treatment in human pluripotent stem cells (42) but was not detected in our assay. MEIS2 was previously identified as an endogenous substrate of CRBN that can be competed by IMiDs (43), but in our data levels were not impacted following IMiD dosing. GLUL was also previously shown to be an endogenous substrate of CRBN, with increased degradation in the context of high glutamine levels (44). In our study, no significant changes in GLUL levels were observed, possibly because glutamine was not perturbed. BSG is stabilized by CRBN binding through a non-E3 mechanism as discussed above. However, unlike SLC16A1, BSG remained constant in the embryos following IMiD exposure for unknown reasons.

Finally, group four consisted of five proteins (GSPT1, ZNF827, IKZF2, IKZF3, and ZNF692) which did not change or were not observed in rabbit embryos. The lack of change in GSPT1 is consistent with having minimal association in developmental biology, however there is evidence of decrease by a different IMiD (CC-885) in myeloid leukemia and multiple myeloma cells in the literature (45, 46). The other proteins in this group also had literature evidence of change in response to one or more IMiDs but no or limited known developmental biology association and were not detected in embryos, indicating that abundance levels of these targets were below our limit of detection. IKZF2 and IKZF3 had IA enrichment reagents and the lack of detection of these proteins indicate that the levels are likely very low, while ZNF827 and ZNF692 did not employ IA enrichment and could benefit from that for detection in this matrix.

As previously noted by in vitro work (4), we observed differences in the neosubstrate degradation patterns between the three IMiDs assessed. The differences in neosubstrate degradation in rabbit embryos with different IMiDs may play a role in apparent phenotype responses following exposure, such as the limited fetal morphological effects of lenalidomide in rabbit (potentially limited by maternal toxicity) compared with thalidomide. Unlike the previous reports however, in our dataset animals dosed with thalidomide had the greatest number of neosubstrate decreases in embryos. This difference may be related to the dosing levels used in our experimental design, timing of sample collection relative to last dose, species and tissue differences, or may be related to the maternal route of exposure of the IMiDs. Of particular note, our dataset is the first to show decreases in abundance in embryos following maternal exposure for many of these targets which, together with the consistent identification of many of these proteins as in vitro targets of IMiD-induced degradation and their known developmental toxicity risks, highlights the potential value of monitoring them as part of a CRBN binding drug discovery program. Because neosubstrate changes can be cell-, tissue-, and species-specific, we believe evaluating neosubstrate changes in vivo during a critical window during embryo-development in the same experimental system that demonstrates developmental toxicity provides more relevant information on relationship between neosubstrate changes and developmental toxicity compared with simply evaluating neosubstrate changes in vitro.

An important observation of our data is that several of these neosubstrates with known developmental biology implications were concurrently decreased in the rabbit embryos following IMiD exposure. This may indicate that SALL4 degradation, while important, may not fully explain previously observed IMiD-induced developmental toxicities. With a better understanding of how or if various neosubstrates are important in developmental toxicity it may be possible to improve CRBN-binding compound, species, and tissue-specific risk assessment. Because the complete mechanism of thalidomide embryopathy is not yet understood, it would likely be beneficial to monitor a larger panel of targets as we have done here rather than just focusing on one or a few neosubstrates during drug development to better contextualize any preclinical toxicity observed and any potential developmental toxicity impacts. While more research is needed to use the existing information for risk assessment purposes, these data may be useful to screen for compounds without these neosubstrate changes. The exact set of proteins to be monitored will likely be context dependent based on the compound(s), experimental system(s) employed, and the goal of the assay and including all targets from all groups is likely to provide greater confidence in safety. However, if a reduced size assay is desired, then from a developmental toxicity perspective using a rabbit embryo model, we recommend minimally monitoring the proteins in group one and optionally the three proteins in group two with known postnatal impacts (GZF1, ZFP91, and IKZF1). An advantage of our assay in this regard is that it is readily configurable and can be easily expanded to include new targets of interest as they are identified and could also be multiplexed with the protein target of interest in a degrader program.

Data Availability

All RAW and Skyline analysis files have been uploaded to the Panorama Public (47) repository and can be accessed via https://panoramaweb.org/CRBN_neosubstrates.url and RAW files can also be downloaded from the ProteomeXchange Consortium with following identifier PXD050568.

Supplemental data

This article contains supplemental data.

Conflict of interest

All authors are employees of Pfizer, Inc.