Solid-state nanopore analysis on conformation change of p53TAD–MDM2 fusion protein induced by protein–protein interaction

 

Hongsik Chae, Dong-Kyu Kwak, Mi-Kyung Lee, Seung-Wook Chi and Ki-Bum Kim

A Department of Materials Science and Engineering, Seoul National University, Seoul 08826, Korea.

B Disease Target Structure Research Center, KRIBB, Daejeon 34141, Korea.

C Department of Proteome Structural Biology, KRIBB School of Bioscience, Korea University of Science and Technology, Daejeon 34113, Korea

D Research Institute of Advanced Materials (RIAM), Seoul National University, Seoul 08826, Korea

 

 

Although protein–protein interactions (PPIs) are emerging therapeutic targets for human diseases, devel- opment of high-throughput screening (HTS) technologies against PPI targets remains challenging. In this study, we propose a protein complex structure to effectively detect conformational changes of protein resulting from PPI using solid-state nanopore for a novel, widely-applicable drug screening method against various PPI targets. To effectively detect conformational changes resulting from PPI, we designed a fusion protein MLP (MDM2-linker-p53TAD), where p53TAD and MDM2 are connected by a 16 amino acid linker. The globular conformation of MLP exhibited a single-peak translocation event, whereas the dumbbell-like conformation of nutlin-3-bound MLP revealed as a double-peak signal. The proportion of double-peak to single-peak signals increased from 9.3% to 23.0% as nutlin-3 concentration increased. The translocation kinetics of the two different MLP conformations with varied applied voltage were ana- lyzed. Further, the fractional current of the intra-peak of the double-peak signal was analyzed, probing the structure of our designed protein complex. This approach of nanopore sensing may be extendedly employed in screening of PPI inhibitors and protein conformation studies.

 

 

Introduction

Protein–protein interactions (PPIs) are involved in crucial regulatory processes of cell signaling such as cell division, programmed cell death, and tumorigenesis. Therefore, the regulation of PPIs have been actively investigated as emerging therapeutic targets for human diseases.1 Among traditional drug targets, enzymes and therapeutic antibodies have been primarily used for drug development. However, the develop- ment of enzyme-targeted drugs has been hampered by off- target side-effects,2 and the development of antibodies has suffered from difficulties in developing an economically viable manufacturing process.3 Therefore, an expansion of druggable targets is urgently needed to accelerate drug development. In this respect, PPIs have attractive advantages compared to other methods to identify drug molecules owing to their broad protein–protein interfaces and the extremely high specificity of the interaction. In particular, the discovery of PPI inhibitors that are able to specifically interfere with the regulatory roles of PPIs is a promising strategy to fulfill unmet needs in pharmaceutical and medical industries.4

To efficiently discover proper PPI inhibitors, development of high-throughput screening (HTS) processes of the drug molecule is essential. However, development of HTS techno- logies against PPI targets remains challenging in spite of recent advances in drug screening processes. To date, several techniques such as nuclear magnetic resonance (NMR),5 surface plasmon resonance (SPR),6 fluorescence polarization (FP), and fluorescence resonance energy transfer (FRET)7 have been developed for PPI analysis and drug screening.8,9 However, NMR requires a large amount of sample and SPR exhibits low sensitivity for detection of small-molecule binding.10 For PPI detection using FP and FRET, the immobil- ization of proteins or labeling of fluorophores is indispensable for detection.7 Undoubtedly, the development of a fast and label-free detection technique is required for efficient discovery of drug molecules based on PPIs.

Recently, solid-state nanopore has been highlighted as an ultra-sensitive and native-state measurement technique at the single-molecule level11,12 and is applicable to characterize protein–protein interactions,13–17 folding/unfolding of biomolecules,18–21  and  conformational  changes.22  For instance, Freedman et al. detected the gp120–antibody inter- action from a larger current drop signal (ΔI) in solid-state nanopore  measurement  resulting  from  the  gp120–antibody interaction forming a complex with a larger volume.23 In addition, Yusko et al. reported the aggregation of β-amyloid protein monomers to form a fibrillar structure, detected by current drop and dwell time changes, as well as changes in event frequencies.24 Waduge et al. also recently reported the detection of two different structures of calmodulin complex induced by calcium ions using high-bandwidth nanopore measurement.25 The authors designed a system in which the two globular domain is connected by a flexible linker that forms either a single or double globular domains depending on Ca2+ addition, as these ions bind to calmodulin and inhibit its interaction. The authors noted that the extended form of the calmodulin shows a longer dwell time and larger current drop than the one forming a single globular structure.

Previously, we reported the interaction between the p53 transactivation domain ( p53TAD) and mouse double minute 2 (MDM2) as well as its inhibition by nutlin-3 using solid- state nanopores.26 p53 is a tumor suppressor protein that induces cell cycle arrest and apoptosis in response to stress signals such as DNA damage. MDM2 is known to negatively regulate the tumor suppressive activity of p53 via a direct interaction with p53TAD (Kd = ∼600 nM).27,28 It is well known that nutlin-3 binds to MDM2 by mimicking the α-helical conformation of the 15-residue p53TAD peptide (residues 15–29), thus preventing the binding of p53TAD and MDM2.29,30 In our previous reports, we detected the interaction between p53TAD and MDM2 by changes in translocation event frequency. Namely, positively charged MDM2 protein (isoelectric point; pI = 9.0) is driven through a nanopore by the applied negative potential. However, when a MDM2/p53TAD complex forms, the overall charge state of the complex changes to negative as a result of charge masking by the negatively charged p53TAD ( pI = 3.6). Therefore, this protein complex is not driven through a nanopore by the applied negative potential. The addition of nutlin-3 liberates MDM2 from the MDM2/p53TAD complex, allowing the translocation of MDM2 through a nanopore under the negative potential.

In our previous report, we demonstrated that solid-state nanopore can be a valuable platform for screening drug mole- cules by monitoring the changes of translocation event fre- quency resulted from charge reversal as a result of PPI. However, this concept has a serious limitation to apply to a variety of PPI target since proteins involving the interaction should have different charge types ( positive or negative). Here, we aimed to develop a novel screening method for detecting PPI inhibitors using solid-state nanopores that can be applied to various PPI targets, without the need of charge reversal of the interacting proteins.

 

 

Results and discussion

We first designed a fusion protein complex MLP (MDM2- linker-p53TAD, 48.5 kDa, pI = 5.18), which includes a relatively large glutathione-S-transferase (GST-p53TAD, pI = 4.8) tag at the C-terminus of p53, where MDM2 and p53TAD are con- nected by a 16 amino acid long linker ((GGGS)4) to induce a conformational change between MLP and MLP-nutlin-3 complex. MLP shows a closed form, globular shape confor- mation as shown in Fig. 1a (left). The addition of nutlin-3 dis- rupts the interaction between p53TAD and MDM2 owing to its relatively high binding affinity with MDM2 (Kd = ∼100 nM).30

Thus, GST-p53TAD is released from MDM2, which changes MLP structure from a closed globular form to a dumbbell-like open form (tMLP) as shown in Fig. 1a (right).

In these experiments, we utilized SiNx nanopores with a highly insulating dielectric substrate31 coupled with a high fre- quency amplifier (4.16 MHz) to improve the signal-to-noise ratio and temporal resolution of our detection system. We expected two types of translocation signals as is schematically shown in Fig. 1b. Because the translocation signature of a single molecule event reflects the detail structure of protein complex, the nanopore signal of MLP is expected to be a single peak (type I), whereas tMLP is expected to have a double-peak signal (type II) even though their overall volume is the same.

 

Nuclear magnetic resonance (NMR) result of p53TAD and MDM2 interaction

Prior to nanopore detection, we examined the PPI between MDM2 and p53TAD in a single fusion protein frame of MLP using NMR spectroscopy (Fig. 2). The 2D 15N–1H heteronuclear single quantum correlation (HSQC) spectra of 15N-labeled MLP were acquired in the absence or presence of nutlin-3. In the absence of nutlin-3, the 2D HSQC spectrum showed the 15N–1H crosspeaks mainly from the p53TAD region of MLP. Unlike in free p53TAD,32 15N–1H crosspeaks from the MDM2- binding residues in p53TAD of MLP ( p53TAD residues 15–29, colored green in Fig. 1a) were not observed owing to severe line-broadening by MDM2 binding. This line broadening of crosspeaks indicates that MLP retains the PPI between MDM2 and p53TAD. In contrast, we found a noticeable recovery of several NMR crosspeaks of MLP after the addition of nutlin-3. The NMR resonance assignments revealed that the restored residues (residues 18, 20, 23, 28, and 29) are predominantly located in the MDM2-binding site of p53TAD in MLP, indicat- ing that nutlin-3 disrupts the PPI between MDM2 and p53TAD as expected, thereby inducing a conformational change in MLP.

 

Nanopore detection of the p53TAD and MDM2 interaction with and without nutlin-3 drug molecule

For the nanopore experiments, 1 M KCl in 1× PBS buffer solu- tion ( pH = 7.4) was filled in both the cis and trans chambers and 200 mV of electric potential was applied across the mem- brane to drag negatively charged MLP toward the nanopores. The protein samples were prepared by incubating MLP protein with increasing molar ratios of nutlin-3 to MLP of zero-fold (free), 1-fold (1 : 1), 5-fold (1 : 5), and 10-fold (1 : 10) for 2 h. A typical I–V curve of the nanopore is presented in Fig. 3a. The schematic structure and electron micrograph of our low-noise device are shown in Fig. 3a inset. After addition of 100 nM of protein sample, translocation signals were detected (Fig. 3b) way of collecting statistical information is very different from other methods that rely on an ensemble approach. The fraction of type II events among all translocation events was 9.3% (1 : 1), 12.0% (1 : 5), and 23.0% (1 : 10) as shown in Fig. 3c.

To check the validity of the experimental results, we esti- mated the fraction of type I and type II events for each molar ratio of MLP and nutlin-3 using calculation model for protein– ligand binding affinity from the dissociation equilibrium constant equation.33 Employing the dissociation constant of p53TAD and MDM2 (Kd = ∼600 nM) and that of nutlin-3 and MDM2 (Kd = ∼100 nM) reported in the literatures,27–30 the estimated fraction of type II events for each molar ratio of MLP and nutlin-3 was 38.2% (1 : 1), 80.7% (1 : 5), and 90.0% (1 : 10), which shows higher than the one detected from nanopore (Fig. 3c). One of the reason is that the sample is not property handled to guarantee all the possible interaction between nutlin-3 and MDM2. For instance, for the nutlin-3 to interact with MDM2, the original interaction between p53TAD and MDM2 should first be dissociated. This dissociation process may be a bottle-neck process for the full reaction between nutlin-3 and MDM2. Another reason is that the linked struc- ture of p53TAD and MDM2 may exhibit a lower Kd than the reported value owing spatial proximity between p53TAD and MDM2.34,35 Also, the different diffusivity of the proteins might lead to the protein molecule to approach the pore entrance in different rate. In the solid-state nanopore experiments, the bio- molecules are driven through the nanopore by the applied electric field. Because most of the electrical field is applied near the entrance of the nanopore, the biomolecules should approach the pore entrance by the random motion, where the electrophoretic velocity overwhelms the diffusive velocity of the proteins, thereby electrophoretically driven through the nano- pore.36 The conformational change of the protein can result in large differences in diffusivity. The estimated bulk diffusion coefficient (D0) for MLP yields D0,MLP = 68.05 nm2 μs−1.37 D0 for tMLP was estimated to be lower than MLP (D0,tMLP = 45.40 nm2 μs−1), where tMLP is assumed to be a dumbbell.38 This lower D0 for tMLP might affect the translocation event frequency as explained above. Another possibility is that some type II signal events appeared as type I events owing to an insufficient time resolution of the detection system.

 

Protein transport kinetics

Dwell time histograms were derived from the scatter plots and are presented in Fig. 4e–h. The distribution of type II events in all nutlin-3 concentrations clearly shows that tMLP exhibited a slower translocation through the pore. To quantitatively evaluate the translocation behavior of MLP and tMLP, we used the 1D diffusion-drift model proposed by Ling et al.;39 where F(t ) is the probability density function, heff is the effective pore thickness, Dp is the protein diffusion coefficient inside the pore, and vd is the drift velocity of the protein, respectively. We extracted the free parameters Dp and vd for type I and II events by fitting the dwell time histograms to eqn (1) (R2 > 0.96). The obtained Dp for type I events were 4.92 ± 0.10, 5.05 ± 0.13, 4.57 ± 0.10, and 5.19 ± 0.09 nm2 μs−1 for free, 1: 1, 1: 5, and 1 : 10, respectively. In contrast, the Dp values for type II events were much lower: 0.29 ± 0.02, 0.33 ± 0.01, and 0.37 ± 0.02 nm2 μs−1 for 1 : 1, 1 : 5, and 1 : 10, respectively. In addition, lowervd values for type II events were acquired as 0.71 ± 0.01, 0.66 ± 0.01, and 0.79 ± 0.01 nm μs−1 for 1 : 1, 1 : 5, and 1 : 10, respect- ively, whereas for type I events the values were 1.48 ±  0.04, 1.46 ± 0.02, and 1.44 ± 0.04 nm μs−1, respectively (1.46 ± 0.02 nm μs−1for free). The similar Dp and vd values for each nutlin-3 concen- tration reveals the identical structure of MLP and tMLP. On the other hand, lower Dp and vd values were obtained for type II events, suggesting that the MLP structure changes into a slower and less diffusive state in the nanopore. Comparing the Dp values to the bulk diffusion coefficient shows that Dp for MLP was ∼14 times smaller than D0, indicating that the diffusive motion of MLP is severely retarded inside the nanopore. On the other hand, the Dp value for tMLP was ∼137 smaller compared to D0,tMLP, which is considered to reflect not only the spatial confinement in the nanopore but also a protein–pore interaction.25,36,40,41

 

Voltage dependence of protein translocation

In order to investigate the voltage-dependent translocation of the proteins, additional nanopore experiment were performed by varying applied voltage. We applied the electric potential of 100 mV, 150 mV, and 200 mV to protein mixture of 1 : 10 (MLP to nutlin-3 ratio) and detected both type I and type II event translocation events as shown in Fig. 5a. We analyzed dwell time histogram of both type I and type II event from the scatter plots (Fig. 5b) and the 1D diffusion-drift func- tion (eqn (1)) was fitted to this data (R2 > 0.93). From the fitted graph, the most probable value of dwell time of type II event was estimated as 16.00, 12.59, and 10.85 μs, for 100, 150, and 200 mV, respectively. The extracted drift velocity of both type I and type II event exhibit linear dependence to the applied voltage (Fig. 5c). The mean electrophoretic mobility was estimated as 74.2 ± 2.31 and 31.5 ± 3.53 nm2 ms−1 V−1 for type I and type II, respectively. The normalized capture rate (Rc/C0; Rc is the measured capture rate, and C0 is the bulk protein concentration, respectively) of type I and type II was shown in the Fig. 5d. We observed that the normalized capture rate of type I event increases with a larger slope (0.20 ± 0.01 s−1 nM−1 mV−1) than that of type II (0.04 ± 0.01 s−1 nM−1 mV−1), but both type of event shows linear dependency to the applied voltage, reflecting that the protein captures were driven by electric field and limited by diffusion process.42

 

Protein size analysis

Fig. 6 shows histogram of fractional current drop for type II intra-peak. Since GST-p53TAD, MDM2, and amino acid linker exhibit different volume, three different current drop inside the type II event was measured. The intra-peak information of the signal was defined to IH, IM, and IL for high-, mid-, and low-level of fractional current drop, respectively (Fig. 6a inset).

The mean fractional current drop, 〈IH〉, shows ∼0.071, 0.077, and 0.068 for 1 : 1, 1 : 5, and 1 : 10, respectively. The 〈IM〉 value exhibits lower value than the 〈IH〉 showing 〈IM〉 ∼ 0.044, 0.045, 0.044, for 1 : 1, 1 : 5, and 1 : 10, respectively. We found IL, which structurally reflected by amino acid linker, with mean value 〈IL〉 as 0.020, 0.019, and 0.022, for 1 : 1, 1 : 5, and 1 : 10, respectively. Each 〈IH〉, 〈IM〉 and 〈IL〉 values show consistent in all nutlin-3 concentration condition, indicating identical signal (Fig. 6). Since the structure of the protein complex used in the experiment can be divided into three parts, we assumed that the value IH, IM, and IL represents each GST-p53TAD, MDM2, and amino acid linker, respectively. On the basis of the value, hydrodynamic diameter of the protein was estimated by using analytical solution25,36,43

Whereas GST-p53TAD and MDM2 are assumed as globular structures, we assumed that the amino acid linker exhibits cylindrical structures. By taking the amino acids linker as membrane-thick long cylinder, the estimated diameter of the cylinder yields 2.72 ± 1.47 nm. This result is ∼5 times larger than the reported results.44 It is not clearly known but prob- ably the linker is not fully stretched when passing through the nanopore. It is also possible that due to fast translocation speed, the IL, which is located in-between GST-p53TAD and MDM2 intra-peaks, was not sufficiently resolved.

 

 

Conclusion

In this study, we have demonstrated the detection of nutlin-3- induced conformational change in the MLP using solid-state nanopores. The characteristic double-peak was successfully detected using a high-frequency amplifier (4.16 MHz), com- bined with our low-noise solid-state nanopore device. Increased concentration of nutlin-3 resulted in an increased appearance of double-peaks. The in-pore diffusion coefficient determined for MLP by applying the 1D diffusion-drift model was ∼14 times smaller than the bulk diffusion coefficient, and ∼137 times smaller than that observed for tMLP. Our nano- pore experiments suggest that drug-induced conformational changes in proteins can be monitored at the single molecule level and our system enables the efficient prediction of drug where dm is the protein diameter, heff is the effective thickness of the pore (7 nm), and dp is the pore diameter (8 nm), respect- ively. The data yield dm = 4.07 ± 0.81, 3.49 ± 1.10 nm for IH and IM, respectively. In comparison to the experimental results, X-ray crystallographic structure of GST-p53TAD and MDM2 were referred from Protein Data Bank (PDB code: 1BG5; GST- p53TAD, 1YCR; MDM2). The estimated physical dimension of crystallographic structure for GST-p53TAD and MDM2 yields 4.7 × 5.3 × 6.0 and 2.4 × 2.6 × 4.1 nm3, respectively. Considering the crystallized protein sample in X-ray diffraction measurement undergoes quite different surrounding environ- ment to the aqueous solution, our nanopore result well reflects the different size of the proteins.

 

 

Materials and methods

Nanopore experiment

100 nm-thick low-stress LPCVD SiNx was deposited on 500 µm thick Si substrate. To fabricate free-standing 2 × 2 mm2 window of SiNx membrane, photolithography and reactive ion etching process was performed, followed by KOH wet etching on 5 × 5 mm2 Si chip. The SiNx membrane was transferred onto separately prepared quartz substrate, which is fabricated as previously described,25 then the membrane was etched by CF4 plasma to a desirable thickness (CF4 40 sccm, 0.05 Torr, 50 W, etch rate ∼30 nm min−1). The nanopore was drilled into the SiNx membrane using focused electron beam using TEM. Prior to use, the fabricated nanopore device was treated with oxygen plasma (15 mA, 0.20 mbar) for 2 minutes before assem- ble into the custom made Teflon cell with PDMS gaskets, which prevents solution from leaking. 1 M KCl in 1× PBS buffer ( pH = 7.4) was filled the cis and trans chambers. After that, the assembled cell was placed in a Faraday cage. In all experi- ments, MLP concentration was fixed to 100 nM and introduced in the cis chamber. Translocation experiments were conducted under 200 mV potential across the membrane using Ag/AgCl electrode. Data were collected using Chimera VC 100 (Chimera Instruments, USA) with sampling rate of 4.16 MHz (Data fil- tered at 500 kHz low-pass filter). Translocation event were ana- lyzed by using python 2.7 based program “pythion” (Developed in Wanunu group), and Clampfit 10.4 and described using OriginPro 8 software.

 

Expression and purification of proteins

The DNA construct encoding MLP fusion protein [MDM2 (residues 3-109)-linker-p53TAD (residues 1-73)-GST] was cloned into the pET-21a vector. The MLP construct includes a linker composed  of  16  residue-long  ((GGGS)4)  between  the C-terminus of MDM2 and the N-terminus of p53TAD. The MLP protein was overexpressed in Escherchia coli Rosetta™ 2 by induction with 0.5 mM isopropyl-β-D-thiogalactoside (IPTG) at the OD600 of 0.7. After IPTG induction, cells were grown at 20 °C for 12 hours in LB or M9 minimal media. The MLP protein was purified using GST affinity chromatography (GSTrap™, GE healthcare), anion exchange chromatography (Hitrap™ Q, GE healthcare), and gel-filtration chromatography (HiLoad® 16/600 Superdex® 75pg, GE healthcare). For NMR experiments, 15N-labeled MLP protein was expressed in M9 minimal media containing 15N-NH4Cl and purified as described above. The nutlin-3 was purchased from Cayman Chemical Inc.

 

NMR spectroscopy

All NMR spectra were acquired in a Bruker 900 MHz spectro- meter equipped with a cryogenic probe at KBSI (Ochang, Republic of Korea). The 2D 15N–1H HSQC spectra of MLP were obtained at 298 K in the absence or presence of nutlin-3. The MLP protein of 100 μM was treated with Nutlin-3a at a molar ratio of 1 : 5. All the NMR data were processed using the NMRPipe/NMRDraw45 and SPARKY softwares.