A real-time FRET ubiquitin transfer assay for quantitative characterization of ternary complexes in targeted protein degradation

Summary

Inhibiting or redirecting E3 ubiquitin ligase activity with drug-like small molecules is of growing therapeutic interest, yet broadly applicable and robust assays for quantifying E3-catalyzed ubiquitination remain limited. Here, we present a real-time, high-throughput fluorescence assay utilizing purified, FRET-active E2-Ub conjugates to monitor ubiquitin transfer. This assay is highly versatile, requiring no engineering of the target protein or ligase, thereby accelerating assay development and minimizing the risk of artifacts. The single-step, single-turnover nature of the reaction enables rigorous and quantitative analysis of ubiquitination kinetics. We show that this assay can be used to measure key degrader characteristics, including degrader affinity for the target protein, degrader affinity for the ligase, affinity of ternary complex assembly, and catalytic efficiency of the ternary complex. The high sensitivity and accuracy of this comprehensive, single-assay approach to ternary complex characterization will empower the discovery and optimization of heterobifunctional degraders and molecular glues.

Graphical abstract

Highlights

• •FRET-active E2-Ub conjugates provide a versatile readout of protein ubiquitination
• •FRET assay enables multiparameter characterization of ternary complexes
• •Quantitative analysis is implemented using open-source SciPy numerical algorithms
• •Method supports high-throughput characterization of PROTACs and molecular glues

Motivation

Targeted protein degradation (TPD) is a rapidly advancing therapeutic strategy that selectively eliminates disease-associated proteins by co-opting the cell’s E3 ubiquitin ligases. We sought to develop a sensitive, high-throughput assay for quantitative characterization of ternary complex assembly and ubiquitin transfer activity.

Yu et al. develop a real-time homogeneous FRET assay for quantitative analysis of ubiquitin transfer in targeted protein degradation. Purified FRET-active E2-Ub conjugates enable high-throughput multiparameter characterization of heterobifunctional degraders and molecular glues.

Introduction

Protein ubiquitination is a fundamental regulatory mechanism implicated in virtually every aspect of eukaryotic biology.¹ Its functional versatility arises largely from the intricate regulation and substrate specificity of E3 ubiquitin ligases, a diverse class of enzymes that catalyze ubiquitin transfer from high-energy E2-Ub thioester conjugates onto target proteins.² The growing interest in ubiquitin ligases as therapeutic targets has spurred efforts to develop improved tools for monitoring E3-catalyzed ubiquitin transfer,^3,4,5 yet broadly applicable and robust analytical methods remain limited. This need is particularly pressing in the emerging field of targeted protein degradation (TPD)—a promising therapeutic strategy that eliminates pathogenic proteins by using small molecules—heterobifunctional degraders (or proteolysis-targeting chimeras [PROTACs]) and molecular glues—to recruit these target proteins to cellular ubiquitin ligases, thereby promoting their ubiquitination and subsequent degradation.⁶

The thioester-linked covalent conjugate of ubiquitin with an E2 ubiquitin-conjugating enzyme is a conserved, high-energy intermediate in protein ubiquitination and related post-translational modifications.^7,8,9,10 The universality of this intermediate, combined with the stoichiometric consumption of one E2-Ub conjugate molecule per E3-catalyzed turnover, makes quantifying E2-Ub depletion an attractive strategy for monitoring E3 enzyme kinetics. To enable such an analytical approach, we synthesized and purified fluorescence resonance energy transfer (FRET)-active E2-Ub thioester conjugates for a selection of E2 enzymes.¹¹ FRET labeling of E2-Ub conjugates with cysteine-reactive dyes has previously been applied to investigate E2-Ub conformational states and ubiquitin transfer at the single-molecule level.¹² In contrast, we developed an assay for ensemble, homogeneous measurements that can be performed in standard plate readers, obviating the need for technically demanding single-molecule FRET detection. Efficient fluorescence labeling of E2 and ubiquitin via sortase-catalyzed transpeptidation,¹³ combined with rapid purification of the reactive thioester conjugates from E1-catalyzed reaction components, yields high-purity reagents with a strong bulk FRET signal. The decay of this signal can be monitored in real time as the conjugate is consumed during E3-catalyzed ubiquitin transfer.^11,14

FRET-active conjugates for many different E2 conjugating enzymes have now been synthesized and are commercially available (STAR Methods). In this study, we sought to evaluate the utility of these reagents for the discovery and optimization of targeted protein degraders. We demonstrate that the FRET-based assay enables quantitative characterization of E3 ligase kinetics in the presence of molecular glues and heterobifunctional degraders. Our findings establish a versatile, accessible tool for accelerating the development of targeted protein degraders and deepening the mechanistic understanding of ubiquitin ligase function.

Results

FRET assay of E3 activity

The performance of the FRET assay was first evaluated using a simple two-component ubiquitin transfer reaction catalyzed by an engineered TRIM5α ubiquitin ligase. TRIM5α undergoes initial N-terminal monoubiquitination, followed by the extension of a K63-linked polyubiquitin chain.^11,15 The extension step of this reaction, mediated by the heterodimeric E2, UBE2N/UBE2V2, was monitored by tracking the FRET signal over time (Figure 1A). Raw fluorescence intensity signals detected at the donor and acceptor emission wavelengths exhibited significant variability due to pipetting errors, air bubbles in microplate wells, and other sample-handling inconsistencies (Figure 1B). However, this variability was largely eliminated when the normalized FRET signal was calculated as the ratio of acceptor-to-donor signal intensities. These results demonstrate the robustness of the FRET assay against non-FRET sources of experimental noise.

Figure 1.

FRET-active E2-Ub conjugates enable real-time monitoring of E3-catalyzed ubiquitin transfer

(A) Schematic of the FRET-based assay using a thioester conjugate of AF488-labeled ubiquitin and AF594-labeled UBE2N.

(B) Reaction progression curves recorded by monitoring donor (read 1) and acceptor (read 2) fluorescence emissions. Normalized FRET is calculated as the ratio of acceptor-to-donor fluorescence intensity at each time point. This normalization suppresses non-FRET-related variability due to pipetting or optical artifacts.

(C and D) SDS-PAGE analysis of quenched reactions imaged for AF594 (C) and AF488 (D) fluorescence.

(E) Overlay of normalized FRET decay curves (lines) and quantification of intact E2–Ub from SDS-PAGE (squares), confirming that FRET signal decay accurately reflects E2-Ub consumption.

To confirm that the decay of the FRET signal results from the dissociation of the FRET-active E2-Ub thioester conjugate, the reaction was quenched at different time points, and its products were analyzed by SDS-PAGE (Figures 1C and 1D). Quantification of acceptor (AF594) fluorescence associated with E2-Ub and E2 bands on SDS-PAGE showed an excellent correlation with FRET data, confirming that the FRET readout accurately reflects the concentration of the intact E2-Ub conjugate remaining in the reaction (Figures 1C–1E).

Imaging of the same SDS-PAGE gel for donor fluorescence revealed the fate of the AF488-labeled ubiquitin transferred from the E2-Ub conjugate (Figure 1D). This analysis showed that when the concentration of E3 ligase exceeded that of the E2-Ub conjugate, the reaction proceeded predominantly in a single-turnover regime, with only one fluorescent ubiquitin transferred onto the Ub-R3-T5α construct. The ability to monitor protein ubiquitination kinetics using a single-step, single-turnover assay enables a rigorous and quantitative analysis of the reaction kinetics, as described in this study.

Degrader-dependent enhancement of E3-catalyzed ubiquitin transfer

To explore the utility of the FRET assay for studies of TPD, we selected a well-characterized model system of BRD4 degradation by dBET1 and related compounds that recruit BRD4 to the CRL4^CRBN E3 ubiquitin ligase complex^16,17 (Figure 2A). The DDB1-cereblon (CRBN) adapter module and the neddylated NEDD8-CUL4A-Rbx1 catalytic module of the CRL4^CRBN were generated separately, as described in STAR Methods, and mixed to reconstitute the active CRL4^CRBN ligase complex. The FRET-active E2-Ub thioester conjugate used in these experiments comprised UBE2D3 labeled with the AF594 acceptor fluorophore and ubiquitin labeled with fluorescein donor dye. The BD1-BD2 construct (amino acids [aa] 44–460) of human BRD4 was used as the ubiquitination substrate in these experiments.

Figure 2.

Monitoring degrader-dependent ubiquitin transfer using FRET detection and kinetic modeling of FRET decay

(A) Schematic of the FRET assay used to monitor dBET1-induced ubiquitination of BRD4 by the CRL4^CRBN ligase.

(B) Control reactions reveal background activity of neddylated Cul4A-Rbx1 that is independent of both substrate and degrader.

(C) Degrader-dependent enhancement of FRET decay exhibits the characteristic hook effect of heterobifunctional degraders.

(D) SDS-PAGE analysis of reaction products at the 1-h time point. Gel imaged for fluorescein fluorescence using a Typhoon scanner.

(E) E3 titration series showing FRET decay at increasing concentrations of CRL4^CRBN in the presence and absence of dBET1. Both degrader-dependent and degrader-independent ubiquitin transfer increase with E3 concentration, with maximal separation observed at intermediate E3 concentrations.

(F) FRET progress curves from 120 replicates for each of three sample compositions (I–III), dispensed into a 384-well plate using automated liquid handling.

(G) Z′ factors calculated over time for individual reads and cumulative datasets.

(H) FRET decay progress curves from a dBET1 titration series monitoring BRD4 ubiquitination by CRL4^CRBN.

(I) In Michaelis-Menten kinetics, substrate depletion below K_m is well approximated by a simple exponential decay.

(J) Enzyme titration series under Michaelis-Menten conditions yield a family of exponential decay curves bounded by the same F_max and F_min values.

(K) The differential equation describing time-dependent E3 ligase inactivation can be solved numerically to yield a theoretical substrate depletion curve.

(L and M) Global least-squares fitting across the titration series determines best-fit model parameters, yielding excellent agreement between theoretical and experimental values.

Control reactions performed without dBET1 revealed that a significant enhancement of FRET decay occurred in the presence of NEDD8-CUL4A-Rbx1 alone (Figure 2B). This background activity was only slightly increased by the addition of DDB1-CRBN and was independent of the BRD4 substrate. However, when dBET1 was titrated into the reaction mixture containing all components, a clear degrader-dependent acceleration of FRET decay was observed (Figure 2C). This enhancement increased as the dBET1 concentration approached that of the E3 ligase, then declined when dBET1 exceeded BRD4 substrate levels, exhibiting a characteristic hook effect.¹⁸ This dependence of FRET decay on degrader concentration suggests that the rate of ubiquitin transfer can serve as a quantitative readout of ternary complex assembly.

To further characterize the reaction, we analyzed products formed one hour after addition of the FRET-active E2-Ub conjugate to the assembled ternary complex, using SDS-PAGE (Figure 2D). Fluorescence imaging of the SDS-PAGE gel revealed that the background activity of NEDD8-CUL4A-Rbx1 alone is highly processive, as indicated by the high molecular weight of the predominant fluorescence band. In contrast, the primary product of the degrader-dependent reaction was monoubiquitylated BRD4 (Figure 2D).

This result highlights a key distinction between our method, which utilizes purified UBE2D3-Ub conjugates, and traditional in vitro ubiquitination assays, where E1-catalyzed E2-Ub synthesis and E3-catalyzed E2-Ub consumption occur concurrently within the same reaction.¹⁹ In conventional assays with UBE2D3, polyubiquitinated substrate species are typically observed. However, in reactions with preformed UBE2D3-Ub conjugates, the monoubiquitylated product predominates. This observation aligns with previous findings that Cullin-RING ligases (CRLs) utilize UBE2D isoforms for initial monoubiquitination (priming) of the substrate, whereas polyubiquitin chain extension is primarily mediated by UBE2R.²⁰

The underlying mechanism of the highly processive background activity observed with NEDD8-CUL4A-Rbx1 and UBE2D3, which contrasts with the degrader-mediated BRD4 monoubiquitination, remains to be elucidated. One possibility is that additional ubiquitination factors or substrates co-purify with the CUL4A-Rbx1 complex, which was affinity purified after co-expression of CUL4A and Rbx1 in insect cells (STAR Methods).

Assay performance in the high-throughput screening format

To evaluate the performance of the FRET assay in a high-throughput (HTS) format, we measured the Z′ factor, a widely accepted metric for assessing assay robustness and reliability.²¹ Because both degrader-dependent and degrader-independent ubiquitin transfer increase with E3 concentration, we performed an E3 titration to evaluate the dependence of assay performance on ligase concentration (Figure 2E). These data show that the difference between degrader-dependent and background FRET decay reaches a maximum at [E3] ≈ 40–50 nM. We then carried out Z′-factor measurements in a 384-well plate, using automated reagent dispensing, as described in STAR Methods. The plate contained 120 replicates for each of three distinct sample compositions (I, II, and III), with a final volume of 6 μL per well. Sample III, which contained all components necessary for degrader-dependent substrate ubiquitination, served as the positive control for Z′ factor calculations. Samples I and II were used as negative controls, monitoring E3-independent and degrader-independent consumption of the E2-Ub conjugate, respectively (Figure 2F).

The E2-Ub conjugate was the final component dispensed into the plate, initiating the ubiquitination reaction. The reaction was monitored for 4 h, with FRET measurements taken every 75 s across the entire 384-well plate. When the Z′ factor was calculated for each time point—using sample II as the negative control and sample III as the positive control—it peaked at 0.73, approximately 30 min after reaction initiation. The Z′ factor gradually declined after the initial 30 min but remained above 0.6 for the entire 4-h reaction duration (Figure 2G). These results strongly support the assay’s suitability for high-throughput discovery and characterization of targeted protein degraders. Moreover, the assay may be even more effective for HTS identification of E3 inhibitors, as Z′ values calculated between samples I and III exceeded 0.8.

In addition to single-read Z′ values, cumulative Z′ can be calculated by incorporating data from all reads between t = 0 and a given time point for each sample (STAR Methods). This approach greatly reduces the contribution of signal detection noise, while maintaining the contribution of sample preparation variability. The cumulative Z′ factor closely matched individual read Z′ values, indicating that sample preparation variability (e.g., E3 dispensing errors) is the primary limiting factor for assay performance. This observation suggests that the reagent concentrations, sample volumes, or data acquisition times could be further reduced without compromising assay performance.

Time-dependent inactivation model of E3 activity

The accuracy and sensitivity of the FRET assay present an opportunity for rigorous quantitative analysis of E3 ubiquitin ligase kinetics, addressing a long-standing challenge in the field. For example, the strong dependence of FRET decay on the degrader concentration suggests that both the concentration and activity of the ternary complex can, in principle, be derived from reaction rate measurements (Figure 2H).

Under Michaelis-Menten kinetics, the rate of substrate depletion is proportional to the concentration of the Michaelis complex [S·E]. When the substrate concentration is significantly lower than K_m—an assumption that holds in our system due to the relatively weak binding of E2-Ub conjugates to RING domains¹⁰—substrate depletion is well approximated by a simple exponential decay function (Figure 2I). Consequently, experiments initiated at the same initial substrate concentration but conducted at different enzyme concentrations would produce a series of exponential decay curves with identical upper and lower boundaries but distinct decay rates (Figure 2J). However, experimental data collected at varying degrader concentrations revealed clear deviations from the simple Michaelis-Menten model, suggesting that additional factors influence substrate depletion by the ternary complex (Figures 2H–2J).

In the degrader titration series, the FRET decay associated with ternary complex activity was evident only during the first 2–3 h of the reaction, after which all samples exhibited an exponential decay toward a common final FRET value (F_min) at a shared rate (k₀), independent of degrader concentration (Figure 2H). This late-phase decay reflects passive, E3-independent hydrolysis of the E2-Ub conjugate. The apparent loss of ternary complex enzymatic activity within the first 2–3 h of the reaction is reminiscent of time-dependent inactivation (TDI), a concept used in pharmacology to describe the gradual inactivation of cytochrome P450 enzymes.²²

To account for this gradual loss of E3 activity, we incorporated an inactivation rate parameter (k_inact) into the differential equation governing substrate depletion (Figure 2K). The theoretical substrate depletion curve was then obtained by numerically solving this equation, and model parameters were determined by performing nonlinear least-squares fitting of the theoretical curves to the data. This analysis yielded excellent agreement between the model and experiment throughout the course of the reaction for all degrader concentrations (Figure 2L).

Although our TDI model contains five free parameters, several factors support the robustness of fitting procedure (STAR Methods). Two of the five parameters were globally minimized across the entire degrader titration series. The three remaining parameters were strongly overdetermined by the dataset and exhibited low covariance, ensuring reliable and independent estimation (Figure 2M). While the precise mechanism underlying gradual E3 inactivation remains unknown, the k_E3 values extracted from the data using the TDI model provide a quantitative measure of the active ternary complex present at the start of each reaction.

Extracting ternary complex characteristics from the ubiquitin transfer rates

Ternary complex formation in TPD is fundamentally a three-component binding equilibrium problem, where the binding affinities and cooperativity between the degrader, target protein, and E3 ligase dictate the overall assembly (Figure 3A). Mathematically, this problem reduces to solving a system of three equations with three unknowns (Figure 3B). However, the general solution is given by the root of a fifth-degree polynomial, for which no closed-form analytical solution exists.²³ While various approximations have been developed for specific limiting cases,^23,24,25,26 none are universally applicable. Nevertheless, ternary complex concentration [EDS] can be computed to arbitrary precision by using a numerical equation solver, given known total concentrations [E]_t, [D]_t, [S]_t and three equilibrium constants $K_d^{ED}$, $K_d^{SD}$, and $K_d^{EDS}$.

Figure 3.

Quantitative analysis of ternary complex assembly using FRET-based ubiquitin transfer assay

(A) Schematic of the reaction and binding equilibria underlying ternary complex formation by heterobifunctional degraders.

(B) System of three mass conservation equations describing the equilibria depicted in (A).

(C) Ternary complex concentration [EDS] and its associated FRET decay rate k^EDS are calculated by numerically solving the system of equations in (B).

(D) Experimental k^EDS values obtained from degrader titration series at multiple substrate concentrations. The BD2 (aa 333–460) construct of BRD4 was used as ubiquitination substrate.

(E and F) Nonlinear least-squares fitting of experimental (D) and modeled (C) k^EDS values yields four parameters describing ternary complex assembly and activity: k_cat/K_m, $K_d^{ED}$, $K_d^{SD}$, and $K_d^{EDS}$.

According to Michaelis-Menten kinetics (Figure 2I), the rate of E2-Ub consumption by the ternary complex (k^EDS in Figures 3A–3C) is approximately proportional to ternary complex concentration when [E2-Ub] ≪K_m, with the proportionality coefficient corresponding to the catalytic efficiency of the ternary complex (k_cat/K_m). Thus, the rate of E2-Ub depletion can be numerically calculated based on three known total concentrations ([E]_t, [D]_t, and [S]_t) and four model parameters: k_cat/K_m, $K_d^{ED}$, $K_d^{SD}$, and $K_d^{EDS}$. A least-squares minimization algorithm can then be applied in conjunction with this numerical solution to determine best-fit values for these four parameters from a well-sampled experimental dataset containing E2-Ub consumption rate measurements across varying ligase, substrate protein, and degrader concentrations (STAR Methods).

The rate of E2-Ub depletion by the ternary complex k^EDS is determined experimentally by subtracting the degrader-independent background depletion rate from the k_E3 values derived in the previous section $k^{EDS}\left({\left[D\right]}_t,{\left[S\right]}_t\right)=k_{E3}\left({\left[D\right]}_t,{\left[S\right]}_t\right)-\overline{k_{E3}}\left({\left[D\right]}_t=0,{\left[S\right]}_t\right)$, where $\overline{k_{E3}}\left({\left[D\right]}_t=0,{\left[S\right]}_t\right)$ denotes the average of k_E3 values measured at [D]_t = 0 for all [S]_t concentrations (Figure 3D). To ensure that the four free parameters of the model are well constrained in the dataset, we performed k^EDS measurements at twelve different [D]_t concentrations for each of four distinct substrate protein concentrations [S]_t (Figure 3D). Two replicate measurements were performed for each sample to evaluate the magnitude of the experimental rate measurement error. Experiments shown in Figure 3 were performed with the BD2 (aa 333–460) construct of BRD4.

At this level of precision, the rich informational content of the dataset is evident from the distinct patterns of k^EDS values, which rise and fall as [D]_t increases, reaching maxima of varying magnitudes at different [D]_t values for each [S]_t concentration. Indeed, the standard errors of the least-squares fitting procedure, as computed from the covariance matrix, indicate strong convergence and precise parameter estimation. Further validation of the model’s stability was performed by introducing increasing levels of artificial noise into the simulated datasets, confirming that the fitting procedure remains robust under realistic levels of experimental variability (STAR Methods).

Multiparametric view of TPD

The four degrader characteristics determined using the FRET assay provide a detailed, multiparametric view of ternary complex assembly and activity. As an illustrative example, we compared two well-characterized degraders, dBET1 and dBET6, both of which recruit the BRD4 substrate to CRL4^CRBN. These heterobifunctional degraders differ only in the length of the linker connecting their two warheads: a CRBN-binding thalidomide moiety and the BRD4-binding ligand, JQ1.

The BRD4 BD2 binding affinities of dBET1 and dBET6 ($K_d^{SD}$ ≈ 50 nM), measured using our assay (Figures 3E and 3F), are similar to each other and to the affinity reported for isolated JQ1 by isothermal titration calorimetry (ITC).²⁷ In contrast, while dBET6 shows CRBN-binding affinity ($K_d^{ED}$ ≈ 150 nM) comparable to the values reported for thalidomide analogs measured by ITC,²⁸ dBET1 exhibits nearly 10-fold tighter binding to CRBN. This suggests that the shorter linker of dBET1 may facilitate additional favorable interactions between CRBN and the degrader. For example, the formation of even a single well-oriented hydrogen bond between the JQ1 warhead and CRBN could account for a 10-fold increase in binding affinity.

Analysis of the ubiquitination rates also yielded the affinity of ternary complex assembly $K_d^{EDS}$ (Figures 3A and 3B), which can be used to calculate assembly cooperativity defined as the ratio $\alpha =K_d^{SD}/K_d^{EDS}$.²⁹ For dBET6, no significant positive cooperativity was observed (α ≈ 1), consistent with crystal structures showing no direct favorable interactions between BRD4 and CRBN in the ternary complex.³⁰ For dBET1, the cooperativity was negative (α < 1), likely due to its shorter linker imposing steric clashes and reducing conformational entropy, thereby contributing unfavorably to ternary complex stability.

In addition to the three equilibrium constants describing ternary complex assembly, our assay also measured the value of k_cat/K_m, which quantifies the catalytic efficiency of the assembled ternary complex. The magnitude of this parameter can be used to evaluate whether the relative orientation of components within the ternary complex favors productive ubiquitination by positioning of the substrate lysine near the E2-Ub thioester bond.

Finally, we investigated the utility of the assay for studies of molecular glues—small molecules that induce ternary complex assembly by promoting direct favorable interactions between the ligase and the substrate. For this purpose, we generated a construct of the human protein SALL4 comprising zinc fingers ZF1 and ZF2 (aa 378–453) and examined its ubiquitination by the CRL4^CRBN ligase in the presence of pomalidomide (Figure 4A). SALL4 is a developmental transcription factor implicated in thalidomide embryopathy, and the structural basis of its recruitment to CRBN by thalidomide-related molecular glues is well characterized.^31,32,33,34

Figure 4.

Quantitative modeling of molecular glue activity in systems with non-negligible E3-substrate affinity

(A) Recruitment of SALL4 to CRL4^CRBN by thalidomide and its analogs is implicated in thalidomide embryopathy.

(B–D) FRET progress curves for ubiquitination of SALL4_378–453 at varying substrate concentrations. (B) Control reactions lacking CRL4^CRBN show negligible ligase-independent enhancement of FRET decay by SALL4 alone. (C) Reactions containing CRL4^CRBN in the absence of pomalidomide reveal degrader-independent enhancement of ubiquitin transfer. (D) Addition of pomalidomide further accelerates FRET decay.

(E) Schematic of binding equilibria contributing to both molecular glue-dependent and -independent ubiquitin transfer.

(F) System of mass conservation equations describing the equilibrium distribution of ED, ES, and EDS complexes.

(G) Ubiquitin transfer rates k^Tot are experimentally determined by subtracting SALL4-independent background from measured k_E3 values. Theoretical values are obtained by numerically solving the system in (E) to determine [ES] and [EDS], and model parameters are obtained by least-squares fitting of the experimental and theoretical k^Tot values.

(H) Experimental k^Tot values (circles) and model predictions (lines) across substrate and degrader titration series. Best-fit parameter values are shown.

(I) Dependence of k^Toton SALL4 concentration in the absence and presence of pomalidomide, demonstrating enhanced ligase-substrate interaction upon molecular glue addition.

We performed pomalidomide titrations at five different SALL4 concentrations, and FRET measurements for all samples, including controls, were recorded in parallel in a 384-well plate over a 4-h period. Control reactions lacking CRL4^CRBN were included for each SALL4 concentration and revealed that ligase-independent enhancement of FRET decay is negligible for SALL4 concentrations below 1 μM (Figure 4B). For substrate proteins with weaker ternary complex binding, however, higher concentrations may be required, in which case the ligase-independent contribution should be explicitly subtracted from the measured FRET decay rates.

In contrast to the BRD4 system, we observed that the addition of SALL4 accelerated the depletion of the FRET-active UBE2D3-Ub conjugate by CRL4^CRBN, even in the absence of pomalidomide (Figure 4C), whereas the addition of pomalidomide further increased FRET decay rates (Figure 4D). These observations indicate that, unlike BRD4, the degrader-independent interaction between CRL4^CRBN and SALL4, described by the dissociation constant $K_d^{ES}$, must be explicitly included in the binding equilibria (Figure 4E). In contrast, the affinity of pomalidomide for SALL4 alone was negligible $K_d^{SD}\approx 0$ and can be omitted from the model. Accordingly, the molecular glue system is described by a modified set of coupled equilibrium equations (Figure 4F).

Another key distinction from the BRD4 system is that both the ternary complex [EDS] and the binary CRL4^CRBN-SALL4 complex [ES] are catalytically competent and contribute to ubiquitin transfer. Consequently, the total substrate-dependent FRET decay rate is given by k^Tot = k^EDS+k^ES (Figure 4E). Experimentally, k^Tot is determined by subtracting SALL4-independent background rate from all measured k_E3 values: $k^{Tot}=k_{E3}\left({\left[D\right]}_t,{\left[S\right]}_t\right)-\overline{k_{E3}}\left({\left[D\right]}_t,{\left[S\right]}_t=0\right)$, where $\overline{k_{E3}}\left({\left[D\right]}_t,{\left[S\right]}_t=0\right)$ denotes the average of k_E3 values measured at [S]_t = 0 for all [D]_t concentrations (Figure 4G).

Using a computational strategy analogous to the one described in the previous section, best-fit model parameters can be determined from a well-sampled experimental dataset (Figures 4G and 4H; STAR Methods). In this analysis, we assume that the catalytic efficiencies of the [EDS] and [ES] complexes are identical, thereby limiting the number of free parameters in the least-squares minimization to four: k_cat/K_m, $K_d^{ED}$, $K_d^{ES}$, and $K_d^{EDS}$. The resulting theoretical curves are in good agreement with the experimental data, although some systematic deviations remain (Figure 4H).

Although model parsimony remains an inherent limitation when describing a complex biochemical system with multiple fitted parameters, the more than 10-fold enhancement of SALL4 recruitment by CRL4^CRBN in the presence of pomalidomide (Figure 4I) was clearly evident and illustrates the power of the assay for quantitative comparison of molecular glue candidates.

Collectively, these findings demonstrate that the sensitivity and precision of the FRET ubiquitin transfer assay enable the quantitative analysis of ternary complex assembly and activity, including systems in which both degrader-dependent and degrader-independent interactions contribute to catalysis.

Discussion

In this study, we demonstrate the utility of purified, FRET-active E2-Ub conjugates for quantitative analysis of ubiquitin transfer in TPD. With purified E2-Ub reagents, the E3 ligase activity can be reconstituted independently from E1-catalyzed E2-Ub synthesis, which greatly simplifies data interpretation. The high sensitivity of FRET detection allows monitoring of enzyme kinetics at physiologically relevant concentrations and in the single-step, single-turnover regime, providing a foundation for rigorous modeling of enzyme kinetics. Here, we illustrate the power of this analytical approach by measuring several key characteristics of targeted protein degraders: affinity for the target protein, affinity for the ligase, affinity of ternary complex assembly, and catalytic efficiency of the ternary complex. These parameters, measured in a single experiment, are in good agreement with values previously obtained using other biophysical techniques.

The methodology is highly resistant to common sources of experimental noise, as shown by the robustness of normalized FRET measurements across replicate wells and independent experiments. In addition, the assay performs well in the high-throughput format, which will facilitate screening of compound libraries in reconstituted systems containing catalytically active ligases and unmodified target proteins. Its capacity to deliver side-by-side, multiparameter evaluation of compound libraries enables efficient structure-activity relationship (SAR) exploration. Acquisition of large, high-precision datasets containing internally consistent affinity and activity parameters may facilitate the application of artificial intelligence (AI) and machine learning tools for predictive modeling. Such models could aid in correlating the degrader’s chemical structure with ternary complex properties, an area of active interest in modern drug discovery.

One of the key strengths of our approach lies in its broad applicability to diverse ubiquitin ligases and substrates. The assay requires no engineering or modification of the E3 ligase or substrate protein, instead relying on preformed, FRET-active E2-Ub conjugates. This design avoids potential artifacts that may arise from fusion constructs or epitope tags and allows the assay to be adapted readily across different E3-substrate systems. While the assay is demonstrated here in the context of the CRL4^CRBN system, its modular design allows straightforward extension to other E3 ligases and ubiquitination pathways. Beyond TPD, the approach could be adapted to study natural substrate recognition, E3-substrate selectivity, and the effects of post-translational modifications or mutations on ubiquitin transfer kinetics.

In summary, the FRET-active E2-Ub conjugates described here provide a sensitive, quantitative, high-throughput, and versatile analytical platform that enables both fundamental studies of E3 ligase function and regulation as well as translational applications in TPD and beyond.

Limitations of the study

The FRET assay described here is implemented using purified recombinant components and, therefore, does not fully capture the complexity of intracellular pathways. In addition, recombinant E3 ligase preparations may contain co-purifying cofactors or substrates that contribute to background ubiquitin transfer activity, necessitating appropriate experimental controls and optimization of assay conditions.

The quantitative framework relies on simplified equilibrium and kinetic models that may not fully capture more complex mechanistic features of ubiquitin transfer reactions. Analysis of weak ternary complex systems may require elevated concentrations of ternary complex components, which can increase contributions from degrader-independent ubiquitin transfer and necessitate explicit inclusion of this background activity in the numerical models.

Resource availability

Lead contact

Requests for further information, resources, and reagents should be directed to and will be fulfilled by the lead contact, Dmitri N. Ivanov (ivanov@uthscsa.edu).

Materials availability

Plasmids generated in this study are available upon request to the lead contact.

Data and code availability

• •Processed experimental datasets generated in this study are publicly available at Zenodo (https://doi.org/10.5281/zenodo.20101291) and from GitHub (https://github.com/ivanov-laboratory/TPD) as of the date of publication.
• •Original Python scripts used for numerical analysis and figure generation are publicly available at Zenodo (https://doi.org/10.5281/zenodo.20101291) and from GitHub (https://github.com/ivanov-laboratory/TPD) as of the date of publication.
• •Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Acknowledgments

This work was supported in part by NIH R37AI136697 (D.N.I.), The Welch Foundation Research Grant AQ-1996-20230405 (D.N.I.), NIH NCI R21CA286307 (D.L.), NIH R01GM115568 (S.K.O.), R01GM128731 (S.K.O.), and CPRIT RR200030 (S.K.O.). S.K.O. holds the MCC 40th Anniversary Endowed Distinguished Professorship in Oncology. We also appreciate the technical assistance provided by Dr. Daifeng Jiang at the Target Discovery Core and the Mays Cancer Center Drug Discovery Shared Resource at University of Texas Health San Antonio, supported by CPRIT Core Facility Award RP250601 and National Cancer Institute Cancer Center support grant P30 CA054174, respectively.

Author contributions

D.N.I. conceived the project idea and developed the methodology; D.N.I., C.H.Y., and V.D. performed the experiments and analyzed the data; D.N.I. wrote the data analysis scripts; C.H.Y., V.D., D.L., P.E., S.D., D.N., and A.N. generated reagents for the study; D.N.I., D.L., D.Z., and S.K.O. provided funding and supervised the research; D.N.I. wrote the original manuscript draft. All authors reviewed and edited the final version of the manuscript.

Declaration of interests

Pending patent applications PCT/US2022/048268, US 18/712304, and EU 22899266.5 list the Board of Regents, The University of Texas System, as the applicant and D.N.I. as an inventor. D.N.I. is a co-founder and a shareholder of E3 Bioscience LLC, a commercial entity that manufactures reagents described in this study.

Declaration of generative AI and AI-assisted technologies in the writing process

During the preparation of this work, the authors used ChatGPT 5.1 to correct typographical errors and improve grammar. After using this tool/service, the authors reviewed and edited the final version of the manuscript as needed and take full responsibility for the content of the published article.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Bacterial and virus strains
E. coli BL21 DE3	New England Biolabs	Cat# C2527H
E. coli DH5α	New England Biolabs	Cat# C2987H
Chemicals, peptides, and recombinant proteins
Fluor.Az.Ub-AF594.UBE2D3	E3 Bioscience	Lot# 0144
AF488.Ub-AF594.UBE2D3	E3 Bioscience	Lot# 0124
Fluor.Az.Ub-AF594.UBE2N	E3 Bioscience	Lot# 0112
Ub-R3-TRIM5α	Herkules et al.11	N/A
UBE2V2	Herkules et al.11	N/A
Human Cullin4A-Rbx1	Nayak et al.35	N/A
Human DDB1-CRBN	Fischer et al.28	N/A
dBET1	Sigma-Alrich	Cat# SML2687
dBET6	Sigma-Alrich	Cat# SML2683
Pomalidomide	Sigma-Alrich	Cat# P0018
Bovine Serum Albumin	Sigma-Alrich	Cat# A4503-100G
DMSO	Fisher Scientific	Cat# AAJ66650AD
Desthiobiotin	IBA Lifesciences	Cat# 2-1000-002
Strep-Tactin 4Flow Resin	IBA Lifesciences	Cat# 2-1250-025
Isopropyl-β-thiogalactoside (IPTG)	Gold Biotechnology	Cat# I2481C50
ATP solution (100 mM)	ThermoFisher Scientific	N/A
Novex Tris-Glycine Mini Protein Gels (4–20%)	ThermoFisher Scientific	Cat# XP04205BOX
Novex Tris-Glycine SDS Running Buffer (10x)	ThermoFisher Scientific	Cat# LC26755
BenchMark Fluorescent Protein Standard	ThermoFisher Scientific	Cat# LC5928
Gibson Assembly Master Mix	New England Biolabs	Cat# E2611L
His10-human-BRD4	This paper	N/A
LB broth	Fisher Scientific	Cat# BP1426-2
2-mercaptoethanol	Fisher Scientific	Cat# AC125470100
Pierce Protease Inhibitor Mini Tablets, EDTA-free	ThermoFisher Scientific	Cat# A32955
HisPur Ni-NTA resin	ThermoFisher Scientific	Cat# 88223
Imidazole	Fisher Scientific	Cat# 03196-500
Critical commercial assays
QIAprep Spin Miniprep Kit	Qiagen	Cat# 27104
Deposited data
FRET decay datasets	This paper	https://doi.org/10.5281/zenodo.20101291
Recombinant DNA
pET21-	Ivanov Lab	N/A
pET21-His10-NusA-SALL4-aa378-453	Ivanov Lab	N/A
Software and algorithms
Original python scripts	This paper	https://doi.org/10.5281/zenodo.20101291
Adobe Illustrator 2024	Adobe	www.adobe.com
Excel Office	Microsoft	www.microsoft.com
Python 3.8	Python Software Foundation	www.python.org
Image Studio 6.1	LICORbio	www.licorbio.com
Gen5 v3.10	Biotek	www.agilent.com
MATLAB R2023b	MathWorks	www.mathworks.com
Other
Nanodrop 2000c spectrophotometer	ThermoFisher Scientific	Cat# ND-2000C
AKTA purification system	Cytiva	N/A
384-well microplates	Corning	Cat# 3575
384-well shallow well microplates	Revvity	Cat# 6059260
Echo acoustic liquid handler	Beckman Coulter	N/A
MANTIS liquid dispenser	Formulatrix	N/A
BioTek Synergy Neo2 multi-mode reader	Agilent	N/A
Amersham Typhoon Trio variable mode imager	Cytiva	N/A
HiLoad 16/600 Superdex 75 pg SEC column	Cytiva	Cat# 28989333
TopSeal-A PLUS transparent film	Revvity	Cat# 6050185
Amicon Ultra-15 centrifugal filter units	Millipore Sigma	Cat# UFC900308

Experimental model and study participant details

Bacterial cell lines

BL21 (DE3) chemically competent E. coli cells were purchased from New England Biolabs (#C2527H). DH5α competent E. coli cells were purchased from New England Biolabs (#C2987H). Cells were grown in LB broth (Fisher Scientific, # BP1426-2) at 37°C with shaking at 220 rpm.

Method details

Protein reagents and chemicals

Purified FRET-active E2-Ub thioester conjugates were obtained from E3 Bioscience (e3bioscience.com). Ub-R3-TRIM5α and UBE2V2 proteins were prepared as previously described.¹¹ All recombinant protein expression vectors were generated by Gibson assembly using the Gibson Assembly Master Mix (New England Biolabs, #E2611L). Human BRD4 and SALL4 constructs were generated using standard E.coli expression and affinity purification protocols.¹¹ The BRD4 constructs, BD1-BD2 (aa 44–460) and BD2 (aa 333–460) contained an N-terminal His₆ tag. The SALL4 ZF1-ZF2 construct (aa 378–453) contained an N-terminal His₁₀-NusA solubility enhancement tag.³⁶ Expression plasmid DNA was transformed into and expressed in E. coli BL21(DE3) cells (New England Biolabs, #C2527H) and grown in LB media at 37°C until the OD₆₀₀ value reached ∼0.6. Protein expression was induced with 1 mM IPTG (Gold Biotechnology, #I2481C50) at 20°C for 12–16 h. Bacteria cells were harvested, lysed by sonication in lysis buffer (50 mM Tris pH 8.0, 1 M NaCl, 15% glycerol, and 20 mM 2-mercaptoethanol) supplemented with Pierce Protease Inhibitor Mini Tablets, EDTA-free (ThermoFisher Scientific, # A32955), and centrifuged at 27,000 x g at 4°C for 45 min. Strep-tagged recombinant proteins were isolated by strep-tactin affinity chromatography according to the manufacturer’s protocol (IBA Lifesciences, # 2-1250-025) and the proteins were eluted with 2.5 mM desthiobiotin (IBA Lifesciences, # 2-1000-002). His-tagged recombinant proteins were isolated by Ni-NTA affinity chromatography according to the manufacturer’s protocol (ThermoFisher Scientific, # 88223) and the proteins were eluted with 500 mM imidazole (Fisher Scientific, # 03196-500). All the eluted proteins were further purified by size-exclusion chromatography on a HiLoad 16/600 Superdex 75 pg column (Cytiva, # 28989333) using an ÄKTA purification system. Fractions containing recombinant proteins were combined, concentrated using Amicon Ultra-15 centrifugal filters (Millipore Sigma, # UFC900308), and stored at −80°C.

Ubiquitination assays were performed on the purified fusion proteins without removing fusion tags. Human Cullin4A and Rbx1 proteins were co-expressed in insect cells, the heterodimer purified and subsequently neddylated following published protocols.^35,37 Human DDB1-CRBN heterodimer was similarly generated using insect-cell co-expression.²⁸ dBET1 (cat. #: SML2687), dBET6 (cat. #: SML2683) and pomalidomide (cat. #: P0018) were purchased from Sigma-Aldrich.

FRET assays

All FRET assays were performed in the buffer containing 50 mM sodium phosphate pH 7.4, 100 mM NaCl and 2 mg/mL bovine serum albumin (Sigma-Alrich, # A4503-100G). Manually-dispensed FRET assays were performed in 384-well black low-protein-binding assay microplates (Corning, #3575) at 40 μL final volume. Low-volume, high-throughput and quantitative assays were performed at 6 μL final volume using automated dispensing of reagents into low-volume ProxiPlates (Revvity, #6059260). In the 6 μL assays, degrader compounds were first dispensed as DMSO stocks at 60 nL per well using an Echo acoustic liquid hander (Beckman Coulter). The remaining components were dispensed as aqueous solutions using a Mantis automated liquid dispenser (Formulatrix). First, substrate protein stock and buffer were dispensed to 4 μL total volume per well to yield desired substrate protein concentration. Reconstituted E3 ligase stock was then dispensed at 1 μL per well. E3, substrate and degrader were incubated for 20 min prior to dispensing E2-Ub conjugates. E2-Ub conjugate stock was dispensed at 1 μL per well and the plates were sealed with TopSeal-A Plus transparent film (Revvity, #6050185) to prevent evaporation and fluorescence. Fluorescence was monitored over time using a Synergy Neo2 plate reader (Biotek) equipped with dual PMT detectors.

Table: Parameters for FRET assay

Reader	BioTek Synergy Neo2 multi-mode reader
Filter cube	255 Filter Cubes
Read interval	30 s
Total read time	4 h
Protocol	25°C, 5 s linear shake; read
Excitation/Emission 1	485/20; 528/20
Excitation/Emission 2	485/20; 635/32
Gain (voltage)	Side PMT = 100; Top PMT = 100
Read height	9.25 mm
Measurements per data point	100
Assay buffer	50 mM sodium phosphate pH 7.4, 100 mM NaCl, 2 mg/mL BSA

Computational data analysis

Extensively commented Python scripts used for data analysis and experimental datasets are available from GitHub: https://github.com/ivanov-laboratory/TPD. Briefly, data analysis was performed in two steps: first, catalytic rates were extracted from experimental FRET decay datasets (get_rates.py), and second, these rates were analyzed as a function of total reagent concentrations to derive ternary complex characteristics (glue_fit_rates.py or protac_fit_rates.py). Each of the two steps used a solver-in-the-loop strategy: a numerical solver produced a model prediction for each trial set of parameters, and those predictions were compared to the experimental data via nonlinear least squares (scipy.optimize.least_squares). Parameter values were optimized iteratively within least_squares to minimize the sum of squared residuals between model and data.

FRET decay fitting. Donor and acceptor intensities were parsed from plate-reader text files; FRET was computed as the ratio of acceptor to donor fluorescence intensity signals. For each sample, the FRET decay was modeled by a time-dependent inactivation (TDI) ODE (Figure 2K), which was numerically integrated with scipy.integrate.solve_ivp. The resulting timecourse served as the theoretical model inside least_squares. Two fitting tiers were used (Fit 1 and Fit 2). Fit 1 (5-parameter):

k_E3, F_max and k_inact were fitted per sample; k₀ and F_min were fitted globally. This was typically applied to long acquisitions (4–6 h) on a subset of wells (or column-by-column for the whole plate). Fit 2 (reduced): For shorter traces (1–2 h) or large datasets, globally determined parameters from Fit 1 were fixed and the remaining 2 or 3 parameters were refit per sample. Options included 2-parameter, 2-plus-1 (with k_inact constrained linearly vs. k_E3), or 3-parameter variants.

Equilibrium model fitting. Catalytic rates were fit to mass-action equilibrium models of ternary complex formation mediated by either heterobifunctional degraders (PROTACs) (Figure 3B) or molecular glues (Figure 4F). The algebraic systems enforcing mass conservation and binding equilibria were solved numerically with scipy.optimize.fsolve to obtain concentrations of the catalytically active species. These numeric predictions populated a model rate matrix that was compared to the experimental catalytic rate matrix within least_squares. Convergence of the fitting procedure and its sensitivity to experimental noise was investigated for different combinations of equilibrium constant values and reagent concentrations using scripts in the/simulate_and_analyze_rate_data folder of the GitHub repository.

Quantification and statistical analysis

Data were analyzed using Python 3.8 and MATLAB 2023b. Figures were made with Adobe Illustrator 2024. All the quantification and statistical details of the experiments are described in the figure legends and in the methods section.

Published: June 23, 2026