SPR-Measured Dissociation Kinetics of PROTAC Ternary Complexes Influence Target Degradation Rate

Bifunctional degrader molecules, known as proteolysis-targeting chimeras (PROTACs), function by recruiting a target to an E3 ligase, forming a target/PROTAC/ligase ternary complex. Despite the importance of this key intermediate species, no detailed validation of a method to directly determine binding parameters for ternary complex kinetics has been reported, and it remains to be addressed whether tuning the kinetics of PROTAC ternary complexes may be an effective strategy to improve the efficiency of targeted protein degradation. Here, we develop an SPR-based assay to quantify the stability of PROTAC-induced ternary complexes by measuring for the first time the kinetics of their formation and dissociation in vitro using purified proteins. We benchmark our assay using four PROTACs that target the bromodomains (BDs) of bromodomain and extraterminal domain proteins Brd2, Brd3, and Brd4 to the von Hippel–Lindau E3 ligase (VHL). We reveal marked differences in ternary complex off-rates for different PROTACs that exhibit either positive or negative cooperativity for ternary complex formation relative to binary binding. The positively cooperative degrader MZ1 forms comparatively stable and long-lived ternary complexes with either Brd4BD2 or Brd2BD2 and VHL. Equivalent complexes with Brd3BD2 are destabilized due to a single amino acid difference (Glu/Gly swap) present in the bromodomain. We observe that this difference in ternary complex dissociative half-life correlates to a greater initial rate of intracellular degradation of Brd2 and Brd4 relative to Brd3. These findings establish a novel assay to measure the kinetics of PROTAC ternary complexes and elucidate the important kinetic parameters that drive effective target degradation.


Table of Contents:
Supporting Tables   Table S1. Fitted SPR data for PROTACs (binary) and PROTAC: Brd4 BD2 complexes (ternary) binding to immobilized VHL and comparison to literature ITC data. Table S2. SPR and FP binding studies with isolated recombinant BET BDs and BET BD point mutants. Table S3. Fitted degradation time course data for HEK293 cells upon treatment with MZ1 (333 nM). Figure S1. Selection of PROTAC:target ratio for ternary binding experiments. SI8 Figure S2. No significant interaction between Brd4 BD2 and biotin-VHL is observed in the absence of PROTAC. Figure S3. Illustration of data treatment for ternary single-cycle kinetic (SCK) binding experiments using immobilized biotin-VHL. Figure S4. Representative SPR sensorgrams for PROTAC (binary) or PROTAC:Brd4 BD2 (ternary) binding to immobilized VHL (for PROTACs MZ1, AT1, MZP55, MZP61).  Figure S8. Representative SPR sensorgrams for AT1:BD (ternary) binding to immobilized VHL (varying the individual BET bromodomain). Figure S9. Representative SPR sensorgrams for PROTAC:BD (ternary) binding to immobilized VHL (for PROTACs MZ1 and AT1, and different BET bromodomain point-mutants).

SI33 -SI35
Figure S10. Fitted Fluorescence Polarization (FP) competition data for MZ1 (binary) and MZ1:BD (ternary) binding to VHL in solution (for individual BET bromodomains and bromodomain point mutants). Figure S11. Representative Western blot for degradation time course data using HEK293 cells.

SI37
Supporting Experimental Methods

Cell biology and degradation studies SI44
Cell lines and culture SI44

Degradation time course assays SI44
Supporting References SI45

SI3
Supporting Tables:   Table S1. Fitted a Analysis where possible is using data from kinetic fitting using a 1:1 Langmuir model including a component for mass transfer effects (Kin), or otherwise using steady state affinity fitting (SSA). SPR binding analyses for binary complexes were performed in multi-cycle kinetic mode at 285.15 K; SPR binding analyses for ternary complexes were performed in single-cycle kinetic (SCK) mode at 298.15 K. For SPR data, listed values were calculated from fitted kinetic data as follows: dissociation constant (KD = koff / kon), dissociative half-life (t1/2 = ln2/koff), cooperativity (α = KD binary /KD ternary ), difference in standard Gibbs free energy of binding for ternary complex relative to binary (DDG = DG ternary -DG binary ) for which, in SI6 each case, DG = RTlnKD; where KD is the appropriate binary or ternary dissociation constant (in M, although in reality dimensionless), R is the ideal gas constant (R = 1.9872 cal.K -1 mol -1 ), T is the experimental temperature (in K). Note: the off rates for some complexes are too fast to be quantified using a Biacore T200 so are reported as above the upper limit of the instrument's typical working range (koff >1 s -1 ).  ) and the dissociation constant (KD) of the binary PROTAC-target interaction. Based on this relationship, for ternary binding experiments, we elected to pre-incubate the PROTAC with a near-saturating concentration of the target protein (corresponding to at least 20-fold in excess of the binary KD of the PROTAC/target protein interaction, and at all times in stoichiometric excess relative to the concentration of PROTAC), to ensure a minimum binary occupancy of greater than 95 %. This ensures that the concentration of free PROTAC remaining in the injected well solution, which will also compete with the PROTAC:target binary complex for binding to the immobilized E3 ligase, remains negligible. The binary affinities for the majority of the PROTAC/bromodomain pairs used in this study have previously been determined (listed in (b); shown to be 1:1 interactions as measured by ITC), 1, 2 .
Based on these values, for PROTAC ternary binding experiments using immobilized biotin-VHL, we elected to set the minimum concentration of free 'near-saturating' bromodomain in each well solution to be 2 µM; such that, for all PROTACs tested, the fraction of binary PROTAC:BD complex formed prior to injection would be expected to be in the range 95 to 98 %. We deemed this concentration an acceptable compromise between consumption of target protein and the expected accuracy of the measured binding response; this decision will almost certainly vary according to the nature of the experiment and the interacting partners. The equivalent calculation can also be made for other types of ternary interaction, or for PROTACs, if the reversed orientation is used (i.e. the target protein immobilized and excess E3 ligase is used in solution).

Figure S2. No significant interaction between Brd4 BD2 and VHL is observed in the absence of PROTAC.
Sensorgrams are shown for a representative ternary single-cycle kinetic (SCK) experiment to measure binding of MZ1-Brd4 BD2 to immobilized biotin-VHL. The first sensorgram (a) shows the doublereferenced binding data for sequential injection of increasing concentrations MZ1 (1.6 nM to 1000 nM) in the presence of near-saturating concentrations of Brd4 BD2 (2 to 25 µM) over immobilized biotin-VHL, resulting in a binding response due to ternary complex formation in the presence of PROTAC. The second sensorgram (b) is the corresponding binding response from a series of blank injections (2 µM Brd4 BD2 in SPR buffer) used for background subtraction. No significant binding of Brd4 BD2 to VHL is observed in the absence of PROTAC; as was also the case for all other purified bromodomains used in this study (data not shown).

Figure S3 (preceding page). Illustration of data treatment for ternary single-cycle kinetic (SCK) binding experiments using immobilized biotin-VHL.
Single representative experiments are shown for MZ1:Brd4 BD2 (a) and MZ1:Brd4 BD2 (b). For each ternary binding experimental repeat, three to four replicate titrations were performed over two to three flowcells with different immobilized surface densities of biotin-VHL. To facilitate comparison between ternary complexes and with binary multicycle data, as well as to make more efficient use of space in figures, sequential injections for ternary SCK experiments were overlaid in a format similar to that typically used for multi-cycle experiments. This was done for each replicate, by shifting the horizontal axis to align the injection time, as illustrated above. Double-referenced data were then globally fitted simultaneously over all flow cells using a 1:1 Langmuir interaction model, with a term for mass-transport included (processed using zip fitting in Scrubber) (BioLogic Software).

Figure S4 (preceding page). Representative SPR sensorgrams for PROTAC (binary) or PROTAC:Brd4 BD2
(ternary) binding to immobilized VHL (for PROTACs MZ1, AT1, MZP55, MZP61). SPR binding analyses for binary complexes were performed in multi-cycle kinetic mode at 285.15 K; SPR binding analyses for ternary complexes were performed in single-cycle kinetic (SCK) mode at 298.15 K and analysed as described ( Figure S3). For MZP55 and MZP61 nonspecific effects were observed during the second half of injection; hence these binary KD values may be considered estimates. For MZP61, only steady state fitting was performed. Figure S5. Effect of varying the PROTAC:target ratio (MZ1:Brd4 BD2 ). To further explore the anticipated effect of varying the ratio of the PROTAC and target protein on efficiency of ternary complex formation, a simple SPR study was undertaken whereby varying ratios of MZ1 and Brd4 BD2 (as depicted in (a)(i)), were mixed and allowed to equilibrate, then the binding response at 285.15 K measured by injecting over an SPR surface coated with immobilized biotin-VHL (~ 500 RU). For each injection, doublereferenced sensorgrams were then fitted to a 1:1 Langmuir interaction model using Biacore T200 Evaluation Software (GE Healthcare). For wells in which the concentration of MZ1 would exceed that of Brd4 BD2 (dark grey squares depicted in (a)(i)), the binding data were not fitted, as in these cases the binding response for the MZ1:Brd4 BD2 complex (ternary complex formation with VHL) is reduced due to competitive binding of excess free PROTAC to form binary PROTAC:VHL complexes (the so-called 'hook effect'). For all other wells, binding constants are tabulated in (a)(ii). For different ratios of MZ1:Brd4 BD2 (1:1, 1:5, 1:25, 1:125), the fitted SPR binding constants (KD, kon, koff, Rmax) were then plotted (as shown in (b)). The fitted sensorgrams measured for each ratio (MZ1:Brd4 BD2 ) are shown in (c) -(f).
Although a limited study, the fitted binding data are consistent with our general expectations regarding selection of optimal PROTAC:target ratio for ternary binding studies ( Figure S1). In particular, it is apparent that in this case use of a 1:1 ratio of PROTAC:target would seem likely to result in underestimation of the true binding affinity (KD) for the ternary complex (refer (a)(i)). The fitted on-rate is slowest for the 1:1 ratio (MZ1:Brd4 BD2 ) (plot of kon, in (a)(ii)) and the fitted off-rate also is fastest for this ratio (plot of koff, in (a)(i)). Similarly, the fitted Rmax appears to be lower for the 1:1 ratio (refer (b)(iv)). This is consistent with our other kinetic data (Table S2) indicating that MZ1 both binds more slowly to VHL and dissociates more quickly, as compared to the interaction of the MZ1:Brd4 BD2 complex SI22 with VHL. As the measured SPR binding response of a mixture of MZ1 and MZ1:Brd4 BD2 will be dominated by the significantly higher molecular weight of MZ1:Brd4 BD2 (relative to MZ1 alone), the most apparent effects of free MZ1 competing for binding might be expected to be a reduction in the overall binding response and a shift in the fitted kinetic parameters for MZ1:Brd4 BD2 binding towards those of MZ1. These data are consistent with this conclusion.
Notably, as the ratio of Brd4 BD2 relative to MZ1 increases, each of the fitted kinetic parameters appears to gradually coalesce (refer plots in (b)). Again, this is likely due the lower concentration of free MZ1 remaining in the injected well solution, such that its competitive effect on the measured binding response is gradually reduced, until essentially only ternary complex formation is measured. Figure S6. Reversed-format SPR binding experiments (immobilized Brd4 BD2 ). SPR sensorgrams are shown for preliminary experiments conducted in a reversed format, for PROTAC (binary) or PROTAC:VHL (ternary) binding to immobilized Brd4 BD2 (for PROTACs MZ1, AT1, MZP61), measured at 285.15 K. These data are shown for qualitative, rather than quantitative purposes. Whilst reasonable kinetic fits to a 1:1 Langmuir interaction model were obtained for binary binding data (in particular for MZ1 and AT1), this was not the case for ternary experiments. Although all ternary binding experiments show a significant increase in overall binding response consistent with binding of the anticipated PROTAC:VHL complex (relative to PROTAC alone), the resulting ternary sensorgrams could not be adequately fitted using a 1:1 Langmuir interaction model. Qualitatively, however, a number of observations are able to be made. Firstly, despite the relatively poor kinetic fitting, it is notable that the binary binding of MZP61 to immobilized Brd4 BD2 appears to be high-affinity, characterized by very slow dissociation kinetics (refer (e)) as compared to either MZ1 (refer (a)) or AT1 (refer (c)). This is indeed consistent with the higher expected binary affinity of the tetrahydroisoquinoline binder of MZP61 for Brd4 BD2 , relative to the triazolodiazepine binder of either MZ1 or AT1. 2 Secondly, in the presence of VHL, the rate of dissociation of the MZP61:VHL complex from immobilized Brd4 BD2 (refer (f)) can be observed to be qualitatively very much faster than MZP61 alone, reflecting the overall negative cooperativity expected for this ternary complex relative to the binary interaction. Figure S7. Representative SPR sensorgrams for MZ1:BD (ternary) binding to immobilized VHL (varying the individual BET bromodomain). SPR binding analyses for ternary complexes were performed in single-cycle kinetic (SCK) mode at 298.15 K and analysed as described ( Figure S3).  Cells were treated with 0.1 % v/v DMSO (vehicle) or 333 nM MZ1 over a range of time points (20 min -420 min) prior to lysis. Samples (40 µg total protein/well) were resolved by SDS-PAGE, transferred to nitrocellulose membrane and probed with anti-Brd2, anti-Brd3 or anti-Brd4 primary antibodies, followed by either goat anti-mouse or donkey anti-rabbit IRDye 800CW secondary antibodies. Bands were then detected using a ChemiDoc (BioRad) and quantified (Image Studio Lite, version 5.2) with normalization to β-actin and DMSO control per time point.
Individual BET bromodomains and bromodomain point mutants were expressed and purified as described previously, 1, 4 with the following modifications. The clarified cell lysate was affinity purified using His Trap HP (1mL) Ni sepharose columns (GE Healthcare) and eluted in 250 mM imidazole in 20 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES), 500 mM sodium chloride and 1 mM β-mercaptoethanol, pH 7.5. Eluted proteins were purified directly without cleavage of the His6 tag, by size exclusion chromatography (SEC) on a Superdex 75 16/60 Hiload gel filtration column on an ÄKTApure TM system (GE Healthcare) in the following buffer: 20 mM HEPES, 500 mM sodium chloride and 1 mM tris(2-carboxyethyl)phosphine (TCEP), pH 7.5. The mass and purity of the proteins were subsequently verified by mass spectrometry (FingerPrints Proteomics Facility, University of Dundee, Scotland).

Expression and purification of GST-BirA.
A plasmid (pGEX6P-1, Amp r ) containing the BirA enzyme as an N-terminal GST-fusion protein with a TEV protease cleavage site (gift of the MRC PPU Reagents and Services, University of Dundee, Scotland; Genbank: M10123) was transformed into BL21 (DE3) cells and expressed and purified based on a modified literature procedure. 7 Briefly, a 10 mL starter culture of LB medium containing ampicillin (100 µg/mL) and D-glucose (0.4 % v/v) was inoculated from a single colony and grown overnight at 37 °C in a shaking incubator (200 rpm). The starter culture (8 mL) was added a 1 L culture of TB containing ampicillin (100 µg/mL) and D-glucose (0.8 % v/v) and grown at 37 °C for 3 h. At an optical density (A600) of approximately 1.1, the temperature was lowered to 23 °C and expression was induced using Isopropyl β-D-1-thiogalactopyranoside (IPTG) (0.4 mM) for approximately 16 h at 23 °C (180 rpm). Cells were harvested by centrifugation (20 min, 4200 rpm) in a JC-M6 centrifuge (Beckman Coulter). Cells were resuspended on ice in 50 mL of GST buffer consisting of 50 mM HEPES, 500 mM sodium chloride, 5 % v/v glycerol and 5 mM DTT, supplemented with Complete protease inhibitor (Roche) and lysed using a Stansted Cell Disruptor (Stansted Fluid Power). Lysate was centrifuged (20,000 rpm, 4 °C) in an Avanti J-25 centrifuge (Beckman Coulter), filtered (0.45 µM syringe filter) and passed twice over a Glutathione Sepharose 4B resin (5 mL bed volume) (GE Healthcare) pre-equilibrated in GST Buffer. The column was washed with GST Buffer (40 mL) and the protein eluted in GST Buffer containing 20 mM L-glutathione (Sigma Aldrich) (20 mL). The eluted GST-BirA protein was purified directly by SEC on a Superdex 75 16/60 Hiload gel filtration column on an ÄKTApureTM system (GE Healthcare) in the following buffer: 20 mM HEPES, 150 mM sodium chloride, pH 7.5 and the protein concentrated (0.4 mg/mL), flash-frozen in N2 (liq.) and stored at -80 °C.

Site-specific biotinylation of VCB-AviTag using GST-BirA.
Site-specific biotinylation of the VCB-AviTag was carried out using GST-BirA as described. 7 Briefly, the VCB-AviTag complex was first dialysed into a low salt buffer consisting of 20 mM HEPES, 20 mM sodium chloride, 1 mM TCEP, pH 7.5. The protein complex (100 µM in 952 µL of low salt buffer) was then mixed SI40 with magnesium chloride (5 µL of 1M solution) (Sigma Aldrich), adenosine triphosphate (20 µL of 100 mM solution) (Sigma Aldrich), thawed GST-BirA enzyme (20 µL of 50 µM solution) and D-Biotin (3 µL of 50 mM solution in 100 % DMSO) (Sigma Aldrich) and incubated for 1 h at 30 °C in an incubator with gentle shaking (90 rpm). After this time, an additional equivalent of GST-BirA and D-biotin were added and the complex incubated for a further 1 h at 30 °C. To the complex was added 100 µL of a 50 % slurry of Glutathione Sepharose 4B resin (pre-equilibrated into low salt buffer) (GE Healthcare) and incubated for 30 minutes at room temperature to capture the GST-BirA. Glutathione Sepharose 4B resin and unreacted D-biotin were subsequently removed by passing the sample over a PD Minitrap G 25 desalting column (GE Healthcare) into 20 mM HEPES, 150 mM sodium chloride, pH 7.5. The extent of biotinylation was evaluated by gel-shift assay with streptavidin, 7 and found to be essentially complete. The final biotinylated complex ('VHL-biotin') was concentrated to 100 µM, snap frozen in N2 (liq.) and stored at -80 °C.

SPR binding studies
SPR experiments were performed on a Biacore T200 instrument (GE Healthcare).

Immobilization of biotinylated VHL.
Immobilization of VHL-biotin was carried out at 25°C using either a Series S CM5 chip to which streptavidin had first been amine-coupled, or using a pre-coupled Series S SA chip. Where not expressly noted, experiments were performed using a CM5/streptavidin sensor chip. For CM5 chips, the surface was pre-equilibrated in HBS-P+ running buffer, containing 2mM tris(2-carboxyethyl) phosphine hydrochloride (TCEP), pH 7.4. Then, following activation of the surface with EDC/NHS (GE Healthcare or XANTEC) (contact time 420 sec or 600 sec, flow rate 10 μL/min), streptavidin (Sigma Aldrich) (prepared at 1 mg/mL in 10mM sodium acetate coupling buffer, pH 5.0) was immobilized by amine coupling to a density of 500-2000 RU, followed by deactivation using 1M ethanolamine. Next, the sensor chip was equilibrated in VHL running buffer consisting 20 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES), 150 mM Sodium chloride, 1 mM TCEP, 0.005 % TWEEN 20, pH 7.0, 1 % dimethyl sulfoxide (DMSO). Biotinylated VHL (100 nM biotinylated VHL, in VHL running buffer) was then streptavidin captured to the required surface density, using either manual injection (flow rate 10 μL/min) (CM5/streptavidin chip) or the automated wizard in the Biacore T200 control software (GE Healthcare) (SA chip, following surface preconditioning with three consecutive injections of 1M Sodium chloride in 50 mM Sodium hydroxide). For binary studies (binding of PROTAC only) the final surface density of biotinylated VHL was approximately 2000 RU; for ternary studies (binding of pre-formed PROTAC:target protein complex), multiple lower surface densities of biotinylated VHL were used (40, 80 and 120 RU) to minimize mass transfer effects. The reference surface consisted of an EDC/NHS-treated surface deactivated with 1M ethanolamine (CM5 chips) or unmodified preconditioned streptavidin surface (SA chips).
All interaction experiments (unless otherwise noted) were performed at 12 °C (binary) or 25 °C (ternary) in VHL running buffer. Sensorgrams were recorded at different concentrations of PROTAC (multi-cycle binary experiments) or PROTAC/target protein complex in the presence of near-saturating concentrations of target protein (single-cycle ternary experiments). For ternary experiments, the minimum concentration of target protein was selected to be approximately 20 to 50-fold in excess of the binary KD of the PROTAC/target protein interaction, to ensure a minimum binary occupancy of approximately 95 to 98 %.
All ternary experiments using immobilized biotin-VHL (unless otherwise noted) were run in single-cycle kinetic mode (rather than multi-cycle mode) due to the slower dissociation kinetics of many of these complexes. This enabled reduced overall experimental run times without the need for additional surface regeneration. For binary PROTAC binding experiments using immobilized biotin-VHL the dissociation kinetics were sufficiently fast, such that these were run in multi-cycle mode.

Ternary interaction experiments (immobilized VHL).
PROTACs (10 mM in 100 % DMSO) were initially prepared at 1 μM or 200 nM in VHL running buffer with a concentration of 2 % DMSO. This solution was mixed 1:1 with a solution of 50 µM of the corresponding bromodomain target protein in VHL running buffer without DMSO, to prepare a final solution (300 μL) of 500nM or 100nM PROTAC and 25 µM bromodomain in VHL running buffer containing 1 % DMSO. This complex was then serially diluted in VHL running buffer containing 2 µM bromodomain and 1 % DMSO (5-point five-fold serial dilution, 500 nM -800 pM or 100 nM -160 pM final concentration of PROTAC, 25 µM -2 µM final concentration of bromodomain). For ternary experiments, solutions were injected sequentially in single-cycle kinetic format without regeneration (four replicate series per experimental repeat, contact time 100 sec, flow rate 100 μL/min, dissociation time 720 sec) using a stabilization period of 30 sec and syringe wash (50 % DMSO) between injections. High flow rates and multiple surface densities were used to minimize mass transfer effects. At least two series of blank injections (VHL running buffer containing 2 µM bromodomain and 1 % DMSO) were performed for all single cycle experiments to be used for blank subtraction.

Ternary interaction experiments varying the PROTAC:target ratio (MZ1:Brd4 BD2 ) (immobilized VHL)
These were run similarly to other ternary interaction experiments, with the modifications that these experiments were recorded in multi-cycle format at 12 °C using a pre-coupled Series S SA chip to which approximately 500 RU of biotin-VHL had been immobilized. MZ1 (10 mM stock in 100 % DMSO) was serially diluted to prepare a 5-point concentration series (2 μM, 400 nM, 80 nM, 16 nM, or DMSO vehicle) in VHL running buffer with a final concentration of 2 % DMSO. This solution was mixed 1:1 in a plate format with a corresponding 5-point concentration series of Brd4 BD2 (2 μM, 400 nM, 80 nM, 16 nM, or running buffer vehicle) in VHL running buffer without DMSO, to ultimately prepare final solutions (300 μL) of PROTAC:BD in varying ratios (1:1, 1:5, 1:25, 1:125; as depicted in Figure S5(a)(i)) in VHL running buffer containing 1 % DMSO. Solutions were injected sequentially in multi-cycle kinetic format without regeneration (single injection per concentration, contact time 120 sec, flow rate 80 μL/min, dissociation time variable, 2500 to 4000 sec) using a stabilization period of 30 sec and syringe wash (50 % DMSO) between injections. A single blank injection (DMSO vehicle) at each concentration of Brd4 BD2 was measured and used for background subtraction of all sensorgrams with the same Brd4 BD2 concentration.

Immobilization of biotinylated Brd4 BD2 .
Immobilization of biotinylated Brd4 BD2 was carried out at 25 °C using a pre-coupled Series S SA chip. The sensor chip was equilibrated in VHL running buffer consisting 20 mM 4-(2-hydroxyethyl)-1piperazineethanesulfonic acid (HEPES), 150 mM Sodium chloride, 1 mM TCEP, 0.005 % TWEEN 20, pH 7.0, 1 % dimethyl sulfoxide (DMSO). Biotinylated Brd4 BD2 (100 nM biotinylated Brd4 BD2 , in VHL running buffer) was then streptavidin captured to the required surface density using the automated wizard in the Biacore T200 control software (GE Healthcare), following surface preconditioning with three consecutive injections of 1M sodium chloride in 50 mM sodium hydroxide. For binary studies (binding of PROTAC only) the final surface density of biotinylated Brd4 BD2 was approximately 2800 RU; for ternary studies (binding of pre-formed PROTAC:target protein complex), the final surface density of biotinylated Brd4 BD2 was ~400 RU. The reference surface consisted of an unmodified preconditioned streptavidin surface. Interaction experiments were performed at 12 °C in VHL running buffer, as described for immobilized VHL.

Preliminary binary interaction experiments (immobilized Brd4 BD2 ).
PROTACs (10 mM stocks in 100 % DMSO) were prepared at 250 nM (300 μL) in VHL running buffer (20 mM HEPES, 150 mM Sodium chloride, 1 mM TCEP, 0.005 % TWEEN 20, pH 7.0) containing 1 % DMSO. This stock solution was then serially diluted in VHL running buffer containing final 1 % DMSO (either 5-point five-fold serial dilution, 250 nM -0.4 nM final concentration of PROTAC; 240 μL sample volume). Solutions were injected individually (duplicate wells) in single-cycle kinetic format without regeneration (contact time 100 sec, flow rate 100 μL/min, dissociation time variable 900 -3000 sec) using a stabilization period of 30 sec and syringe wash (50 % DMSO) between injections. In the case of MZP61 (binary), a regeneration step was used consisting of a single injection of 2 μM VHL in running buffer containing 1 % DMSO (contact time 100 sec, flow rate 100 μL/min, dissociation time 900 sec), which caused rapid dissociation of the formed ternary complex. Two series of blank injections were performed for all single cycle experiments.

Preliminary ternary interaction experiments (immobilized Brd4 BD2 ).
PROTACs (10 mM in 100 % DMSO) were initially prepared at 500 nM in VHL running buffer with a concentration of 2 % DMSO. This solution was mixed 1:1 with a solution of 50 µM of VHL target protein in VHL running buffer without DMSO, to prepare a final solution (300 μL) of 250nM PROTAC and 25 µM VHL in VHL running buffer containing 1 % DMSO. This complex was then serially diluted in VHL running buffer containing 2 µM VHL and 1 % DMSO (5-point five-fold serial dilution, 250 nM -400 pM final concentration of PROTAC, 25 µM -2 µM final concentration of VHL). For preliminary ternary experiments, solutions were injected sequentially in single-cycle kinetic format without regeneration (one series per experimental repeat, contact time 100 sec, flow rate 100 μL/min, dissociation time variable 900 -3000 sec) using a stabilization period of 30 sec and syringe wash (50 % DMSO) between injections. Two series of blank injections were performed for all single cycle experiments.

SPR data analysis.
Data were analysed using either or Scrubber (BioLogic Software) or Biacore T200 Evaluation Software (GE Healthcare). Sensorgrams from reference surfaces and blank injections were subtracted from the raw data (double-referencing) and the data was solvent-corrected prior to analysis. To calculate the association rate (kon), dissociation rate (koff), and dissociation constant (KD), data from all binary (multi-cycle) and ternary (single cycle) experiments were fitted using a 1:1 Langmuir interaction model, with a term for mass-transport included. For experiments conducted at multiple surface densities, this was performed using global fitting of data from all surface densities simultaneously (zip fitting in Scrubber) (BioLogic Software) (as described in Figure S3). For some ternary experiments, nonspecific effects were observed during the association phase at the top concentration only; in these cases, the fifth injection only was not included in global fitting.