Molecular Imaging of Hypoxia-Inducible Factor 1α and von Hippel-Lindau Interaction in Mice

Abstract

Tumor hypoxia plays a crucial role in tumorigenesis. Under hypoxia, hypoxia-inducible factor 1α (HIF-1α) regulates activation of genes promoting malignant progression. Under normoxia, HIF-1α is hydroxylated on prolines 402 and 564 and is targeted for ubiquitin-mediated degradation by interacting with the von Hippel-Lindau protein complex (pVHL). We have developed a novel method of studying the interaction between HIF-1α and pVHL using the split firefly luciferase complementation-based bioluminescence system in which HIF-1α and pVHL are fused to amino-terminal and carboxy-terminal fragments of the luciferase, respectively. We demonstrate that hydroxylation-dependent interaction between the HIF-1α and pVHL leads to complementation of the two luciferase fragments, resulting in bioluminescence in vitro and in vivo. Complementation-based bioluminescence is diminished when mutant pVHLs with decreased affinity for binding HIF-1α are used. This method represents a new approach for studying interaction of proteins involved in the regulation of protein degradation.

Tissue hypoxia plays a critical role in many pathophysiologic conditions, such as cancer. In solid tumors, hypoxia exists as a result of inadequate oxygen supply created by abnormal microcirculation and diffusion conditions. Clinically, hypoxic tumors are more resistant to therapy and are a poor prognostic factor.¹ To improve therapeutic options for cancer patients, we need a better understanding of the hypoxic tumor microenvironment and the molecular changes that tumor cells use for adaptation.

Central to hypoxic adaptation is the transcription factor hypoxia-inducible factor 1α (HIF-1α). HIF-1α is overexpressed in many human cancers in part owing to oncogenic mutations² and mitochondrial perturbations.³ HIF-1α overexpression is correlated with poor clinical outcomes.^1,4,5 HIF-1α levels are regulated in an oxygen-dependent manner. Under normoxic conditions, proline residues in the oxygen-dependent degradation (ODD) domain of HIF-1α (P402 and P564) are hydroxylated by prolyl hydroxylases (PHDs).^6–10 Following hydroxylation, von Hippel-Lindau protein (pVHL) complex adds ubiquitin onto HIF-1α, targeting it for degradation.¹¹ Under hypoxia, the PHDs are unable to hydroxylate HIF-1α efficiently. HIF-1α escapes degradation and forms a heterodimer with HIF-1β to regulate target genes, many of which are crucial in tumorigenesis and metastasis.¹² Genetic mutations in VHL disease cause an autosomal dominant, inherited human cancer syndrome with predispositions for hemangioblastoma, clear cell renal carcinoma, and pheochromocytoma.

As HIF-1α and VHL are important for malignant progression, studying the in vivo interaction between HIF-1α and pVHL is essential for a better understanding of the regulation of HIF-1α by pVHL and, moreover, the development of novel anticancer therapies. We have developed a novel method of studying this interaction using the firefly luciferase protein complementation-based bioluminescence. Using the split firefly luciferase complementation system,^13–18 we have generated split luciferase protein chimeras fused to either HIF-1α or pVHL. Upon hydroxylation, pVHL interacts with HIF-1α, leading to complementation of the split luciferase fragments and, more importantly, an active enzyme capable of bioluminescence. This unique system has allowed us to (1) study the interaction of HIF-1α and pVHL at the molecular level, (2) image HIF-1α and pVHL interaction in mice, and (3) evaluate how clinically derived pVHL mutations affect HIF interaction.

Materials and Methods

Cell Culture

293T and HCT116 cells were obtained from American Type Culture Collection (Manassas, VA) and grown in Dulbecco’s Modified Eagle’s Medium containing 10% fetal calf serum at 37°C.

Plasmids

Synthetic firefly luc2 reporter plasmid (pGL4) (Promega, Madison, WI) was used as the polymerase chain reaction (PCR) template. The human HIF-1α and VHL expression plasmid were gifts from Dr. H. Franklin Bunn (Harvard University) and Dr. Judith Frydman (Stanford University), respectively. HIF-1α P402A and P564G mutants were made as previously described.¹⁹ VHL V84L, F119S, P154L, R167Q, and L188V mutations were introduced by PCR-mediated mutagenesis and confirmed by deoxyribonucleic acid (DNA) sequencing.

Amino-Terminal Firefly Luciferase

The amino-terminal region of firefly luciferase (NLUC) (amino acids 1–398) was amplified using a forward primer designed with an NheI restriction site and the start codon and the reverse primer designed with a BamHI restriction site. This fragment was cloned into pcDNA3.1(+). NLUC-HIF-1α and NLUC-VHL constructs were made by PCR amplification using the forward primer designed with a BamHI restriction site and a linker sequence (GGGGSGGGGS [nucleotide sequence 5′-GGTGGAGGCGGTTCAGGCGGAGGTGGCAGC-3′]) and a reverse primer containing a stop codon and an XhoI restriction site and cloned into pcDNA3.1(+)-NLUC.

Carboxy-Terminal Firefly Luciferase

Firefly luciferase amino acids 398 to 550 were amplified using the forward primer designed with a BamHI restriction site and the reverse primer designed with a stop codon and an XhoI restriction site. Fragments were cloned into pcDNA3.1(+). HIF-1α–carboxy-terminal firefly luciferase (CLUC) and VHL-CLUC constructs were made by PCR amplification using the forward primer designed with an NheI restriction site and the start codon and a reverse primer containing a linker sequence (GGGGSGGGGS) and a BamHI restriction site and subcloned into pcDNA3.1(+)-CLUC.

PHD2 Knockdown

Short hairpin plasmids were made by inserting a 19-mer sequence of PHD2 into BamHI/EcoRI sites of pSIREN-RetroQ (Clontech, Mountain View, CA) (manuscript in preparation). Following retroviral infection of HCT116 cells, short hairpin RNA (shRNA) to PHD2 was stably selected using puromycin (1 μg/mL). PHD2 knockdown was verified by Western blot (Novus Biologicals, Littleton, CO).

In Vitro Firefly Luciferase Assays

293T or HCT116 cells were plated in 12-well culture plates to achieve 90 to 95% confluence on the day of transfection. HIF-1α and VHL split luciferase expression plasmids were transfected (1:1) using Lipofectamine 2000 (Invitrogen, Carlsbad, CA) following the manufacturer’s protocol. After 24 hours, cells were lysed in 200 μL of Reporter Lysis Buffer (Promega) using a single freeze-thaw cycle and centrifuged for 5 minutes at 10,000g at 4°C. In vitro split luciferase assays were performed by adding 20 μL of the supernatant to 100 μL of the luciferase assay reagent (Promega) followed by photon counting in the luminometer (BD Monolight 2010, Franklin Lakes, NJ) for 10 seconds. All reactions were done in triplicate. Relative light unit readings were normalized by cotransfecting with pCMV-β-galactosidase plasmid (Clontech) and using the Beta-galactosidase Enzyme Assay System (Promega).

Antibodies and Immunoblotting

293T cells were transfected with HIF-1α, and pVHL split luciferase expression plasmids were as described above. After 24 hours, cells were lysed in radioimmunoprecipitation buffer (50 mM Tris-HCl, pH 7.4, 1% NP-40, 0.25% sodium deoxycholate, 150 mM sodium chloride, 1 mM ethylene-diaminetetraacetic acid, and protease inhibitor cocktail [Sigma, St. Louis, MO]). Cells were sonicated for 20 seconds. Cell lysate was centrifuged at 16,000g for 10 minutes at 4°C. Protein concentration was determined using the Bio-Rad DC Protein Assay (Bio-Rad, Hercules, CA) using Bovine serum albumin (BSA) standards (Pierce, Rockford, IL). A total of 50 μg of protein was loaded onto 10% sodium dodecyl sulfate–polyacrylamide gel electrophoresis gels and transferred onto immunoblot polyvinyl difluoride membranes. pVHL and HIF-1α (WT and P564G mutant) split luciferase immunoblots were probed using anti-VHL (BD Biosciences Pharmigen, San Diego, CA) and antiluciferase pAb (Promega) antibody, respectively. Even protein loading was determined using α-tubulin or glyceraldehyde-3-phosphate dehydrogenase (Sigma).

Optical Charge-Coupled Device Imaging in Living Mice

All of the animal handling was performed in accordance with Stanford University Animal Research Committee guidelines. Nude mice (nu/nu) were purchased from Charles River (Wilmington, MA). 293T cells were cotransfected with NLUC-HIF-1α (amino acids 556–603 WT or P564G mutant) and VHL-CLUC DNA using Lipofectamine 2000 following the manufacturer’s protocol. Twenty-four hours after transfection, cells were harvested; 10⁷ cells were injected subcutaneously while the mice were anesthetized with isoflurane. Animals were imaged using an optical charge-coupled device (CCD) camera (Xenogen IVIS, Xenogen Corp.) 15 minutes after intraperitoneal injection of D-luciferin dissolved in phosphate-buffered saline (150 mg/kg). Bioluminescence was analyzed by Living Image software (Xenogen Corporation) and Igor Image Analysis software (Wavemetrics, Seattle, WA). Statistical analyses were performed using the Student t-test (JMP statistical software, SAS Institute Inc, Cary, NC).

Results

NLUC-HIF-1α and VHL-CLUC Split Luciferase Result in Complementation-Based Bioluminescence

Our goal was to develop a noninvasive method of studying the interaction between HIF-1α and pVHL in vivo. The split firefly luciferase complementation system requires that the HIF-1α and pVHL portions of the chimeras interact in a manner aligning the luciferase fragments in the proper conformation to create an active enzyme. To determine the optimal orientation of the split luciferases, several different HIF-1α and pVHL constructs were made (Figure 1A). The combination of NLUC-HIF-1α and VHL-CLUC yielded complementation-based bioluminescence, whereas the reciprocal combination of NLUC-VHL and HIF-1α-CLUC did not (Figure 1B), likely because the luciferase fragments are unable to assemble in the appropriate conformation for activity.

Figure 1.

In vitro hypoxia-inducible factor 1α (HIF-1α) and von Hippel-Linda protein (pVHL) complementation-based split luciferase activity. All assays were done in triplicate. Error bars indicate ± 1 SD. A, Diagram of the HIF-1α and VHL split luciferase constructs: (A) amino-terminal firefly luciferase fragment (NLUC)-HIF-1α (amino acids 344–603); (B) NLUC-VHL; (C) HIF-1α (amino acids 344–603)–carboxy-terminal firefly luciferase fragment (CLUC); (D) VHL-CLUC. B, Luciferase assays were performed on 293T cells transiently transfected with (A) NLUC or CLUC alone, (B) NLUC-HIF-1α (amino acids 344–603) and VHL-CLUC, or (C) NLUC-VHL and HIF-1α (amino acids 344–603)-CLUC. NLUC-HIF-1α (amino acids 344–603) and VHL-CLUC coexpression results in complementation-based bioluminescence. C, Dissecting the oxygenation-dependent degradation domain. Diagram of the HIF-1α split luciferase constructs: NLUC chimeras expressing HIF-1α amino acids: (B) 344 to 603; (C) 531 to 650; (D) 531 to 603; (E) 531 to 575; (F) 556 to 603; (G) 556 to 650; (H) 556 to 575. D, Luciferase assays were performed on 293T cells transiently transfected with (A) VHL-CLUC alone or VHL-CLUC cotransfected with (B) HIF-1α (amino acids 344–603), (C) HIF-1α (amino acids 531–650), (D) HIF-1α (amino acids 531–603), (E) HIF-1α (amino acids 531–575), (F) HIF-1α (amino acids 556–603), (G) HIF-1α (amino acids 556–650), (H) HIF-1α (amino acids 556–575).

The HIF-1α ODD domain was further dissected to identify the regions within the carboxy-terminus capable of complementation-based split luciferase activity (Figure 1C). Six additional NLUC-HIF-1α ODD domain constructs were made (Figure 1D). Although all of the constructs resulted in high levels of luciferase complementation, variable levels of split luciferase activity were observed depending on the size of the HIF-1α construct used. Amino acids 556 to 575 of HIF-1α are able to complement VHL-CLUC, but the highest luciferase activity is achieved using the two constructs composed of 45 and 48 amino acids (556–603 and 531–575). These findings suggest that the sizes of the interacting protein fragments influence both the binding of the two proteins of interest and the ability of luciferase fragments to complement.

HIF-1α and pVHL Interaction Requires Proline Hydroxylation

The key step in the interaction of HIF-1α and pVHL is the hydroxylation of either proline 402 or proline 564.⁷ Thus, we examined whether luciferase activity is dependent on hydroxylation. We hypothesized that interference of hydroxylation by (1) mutation of hydroxylation targets P402 and P564, (2) small molecule inhibitors of prolyl hydroxylase, and (3) genetic silencing of one of the prolyl hydroxylases would inhibit HIF-1α and pVHL-mediated split luciferase activity.

First, P402 and P564 were mutated to amino acids that cannot undergo hydroxylation (Ala and Gly, respectively). The ability of mutant HIF-1α to complement VHL-CLUC was compared with wild-type (WT) HIF. Mutations of P402 and P564 led to a marked decrease in luciferase activity in all of the HIF-1α constructs tested (Figure 2A), indicating that HIF-1α and pVHL complementation is dependent on proline hydroxylation. To ensure that decreased complementation-mediated luciferase activity seen with P564G mutant is not due to a decreased level of protein, Western blot analysis was done with 293T cells transiently cotransfected with pVHL split luciferase in the presence of HIF-1α WT or P564G split luciferase. As seen in Figure 2F, the level of P564G split luciferase is actually higher than that of the WT. This may be due to the fact that P564G cannot be hydroxylated and therefore cannot undergo pVHL-mediated degradation. Despite its higher protein level, only the background level of bioluminescence is detected with the P564G mutant split luciferase.

Figure 2.

Effects of proline hydroxylation inhibition on hypoxia-inducible factor 1α (HIF-1α) and von Hippel-Lindau (pVHL) complementation-based split firefly luciferase activity. All assays were done in triplicate. Error bars indicate ± 1 SD. A, Effects of HIF-1α P402 and P564 mutation on the HIF-1α and pVHL split luciferase activity. Luciferase assays were performed on 293T cells transiently transfected with VHL–carboxy-terminal firefly luciferase fragment (CLUC) cotransfected with HIF-1α (amino acids 344–603), HIF-1α (amino acids 531–650), HIF-1α (amino acids 531–603), HIF-1α (amino acids 531–575), HIF-1α (amino acids 556–603), HIF-1α (amino acids 556–650), HIF-1α (amino acids 556–575). Wild-type (WT) HIF-1α is represented in gray; *P564G (or **P402A and P564G) HIF-1α mutants are shown in black. B, Effects of small molecule prolyl hydroxylase inhibitors on the HIF-1α and pVHL split luciferase activity. From left to right: 293T cells cotransfected with HIF-1α (amino acids 556–603) and VHL-CLUC (far left bar); addition of desferrioxamine (DFO) (100 μM for 12 hours); addition of CoCl₂ (100 μM for 12 hours); HIF-1α (amino acids 556–603) P564G and VHL-CLUC. C, Western blot demonstrating a > 90% decrease in the levels of PHD2 protein in HCT116 cells stably transfected with short hairpin RNA (shRNA) against PHD2 (right) compared with the HCT116 parental line (left). α-Tubulin was used as the loading control. D, silencing PHD2 on the HIF-1α and pVHL split luciferase activity. Amino-terminal firefly luciferase fragment (NLUC)-HIF-1α (amino acids 556–603) WT or P564G mutant and VHL-CLUC were cotransfected in HCT116 parental cell line (black bars) or HCT116 PHD2 knockdown cells (white bars). E, Western blot demonstrating overproduction of HIF-1α and pVHL split luciferases in transiently transfected 293T cells. Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) was used as the loading control. F, Western blot demonstrating HIF-1α WT and P564G mutant split luciferases in transiently transfected 293T cells. GAPDH was used as the loading control.

Second, inhibition of prolyl hydroxylase by the addition of small molecule inhibitors, desferrioxamine or cobalt chloride, caused a dramatic decrease in the luciferase activity (Figure 2, B and E). Third, silencing of PHD2 resulted in approximately 50% in luciferase activity. As expected, PHD2 knockdown had no effect on HIF-1α P564G and pVHL complementation. Despite a greater than 90% decrease in PHD2 protein (Figure 2, C and D),^20–22 the complementation-mediated split luciferase activity did not drop to background levels, probably because PHD2 silencing is not complete and PHD1 and PHD3 are still functional. Taken together, these results demonstrate that the interaction between HIF-1α and pVHL requires proline hydroxylation by the PHDs, and interference of proline hydroxylation leads to inhibition of the HIF-1α and pVHL split luciferase activity.

Study of the pVHL Mutants Using the Split Luciferase Complementation System

The HIF/pVHL split luciferase system was used to study clinically derived pVHL mutations. VHL V84L, F119S, P154L, R167Q, and L188V mutants (Figure 3A²³) were chosen because they exhibit a variety of clinical phenotypes and have been evaluated for their ability to interact with HIF-1α. VHL V84L and L188V appear to have binding affinities for HIF-1α that are roughly similar to the WT VHL, whereas F119S has a markedly diminished binding to HIF-1α.^24,25 P154L mutation demonstrates a modest level of binding to HIF-1α.²⁶ R167Q is able to bind HIF-1α but with a slightly weaker affinity compared with WT.²⁵ In descending order, WT VHL ≈L188V > V84L ≈R167Q > P154L > F119S. Of the pVHL mutants studied, P154L and F119S exhibited the two lowest complementation-based bioluminescences, suggesting that these mutants are significantly impaired in their ability to bind HIF-1α compared with the WT VHL (Figure 3B). Unlike F119S, L188V mutant retained the ability to complement HIF-1α level similar to WT VHL, whereas V84L and R167Q exhibited slightly lower complementation-based luciferase activities. These results are in agreement with previous reports showing similar findings using pVHL capture studies.^24–26

Figure 3.

von Hippel-Lindau (pVHL) mutant studies. A, Crystal structure of the hypoxia-inducible factor 1α (HIF-1α) (stick figure) binding to VHL (green ribbon).²⁷ HIF-1α in red; VHL in gray; elongin C in light tan; elongin B in bronze. pVHL residues as follows: V84, magenta; F119, light yellow; P154, white; R167, blue; L188, orange. B, In vitro split luciferase assays using pVHL mutants. Luciferase assays were performed on 293T cells transiently transfected with amino-terminal firefly luciferase fragment (NLUC)-HIF-1α (amino acids 344–603) cotransfected with carboxy-terminal firefly luciferase fragment (CLUC) chimeras expressing wild-type VHL, V84L, F119S, P154L, R167Q, and L188V. All assays were done in triplicate, each with 10-second photon counting. Error bars indicate ± 1 SD.

Interaction between HIF-1α and pVHL Can Be Imaged in Living Mice

293T cells transfected with VHL-CLUC and NLUC-HIF-1α WT or P564G were implanted subcutaneously in the right and left flanks, respectively, of nude mice. After intraperitoneal injection of D-luciferin, mice were imaged by CCD camera, demonstrating that the interaction between HIF-1α and VHL can be imaged in living mice using noninvasive methods. Moreover, these results confirm that in vivo HIF-1α and pVHL split luciferase interaction also requires HIF-1α hydroxylation, as shown by the lower bioluminescence signal emitted by the P564G mutant cells (right) when compared with the WT and VHL-CLUC-expressing cells (left) (p < .001) (Figure 4, A and B).

Figure 4.

Studying hypoxia-inducible factor 1α (HIF-1α) and von Hippel-Lindau (pVHL) interaction in vivo using bioluminescent optical imaging. A, 293T cells cotransfected with VHL–carboxy-terminal firefly luciferase fragment (CLUC) and amino-terminal firefly luciferase fragment (NLUC)-HIF-1α wild-type (WT) (amino acids 556–603) (left flank) or NLUC-HIF-1α P564G (right flank) were implanted subcutaneously into nude mice. B, In vivo bioluminescence signals were quantified (photons/s/cm²/steradian), and the means were plotted. Error bars indicate ± 1 SD (n = 9 mice); p < .001. C, 293T cells cotransfected with NLUC- HIF-1α (amino acids 556–603) and WT VHL-CLUC (right abdomen) or VHL F119S-CLUC (left abdomen) were implanted subcutaneously into nude mice. D, In vivo bioluminescence signals were quantified (photons/s/cm²/steradian), and the means were plotted. Error bars indicate ± 1 SD (n = 8 mice); p < .001.

We also imaged the interaction between HIF-1α and WT VHL or VHL F119S mutant. NLUC-HIF-1α was transfected with either WT or the F119S VHL-CLUC and implanted in the right and left abdomens of immunodeficient mice, respectively. The bioluminescence signal produced by NLUC-HIF-1α and WT VHL-CLUC expression (right) is significantly greater than with F119S VHL (left) (p < .001) (Figure 4, C and D), indicating that VHL F119S mutant has a decreased affinity for HIF-1α. Furthermore, this difference is readily appreciated visually with optical imaging.

Discussion

We have established a novel method of studying the E3 ubiquitin ligase and substrate protein–protein interaction using the split firefly luciferase complementation-based bioluminescence system. NLUC and CLUC firefly luciferase fragments have been fused to two proteins of interest, and the two luciferase fragments are brought together to generate a catalytically active enzyme when the two proteins of interest interact. Firefly luciferase complementation has several advantages over other complementation techniques. Unlike the split ubiquitin, dihydrofolate reductase, and β-galactosidase complementation systems, the split firefly luciferase system is highly sensitive and can be used to image living subjects. Compared with fluorescence resonance energy transfer (FRET), firefly luciferase has a high signal to noise ratio, a more stable signal, and can be used in vivo. Compared with the Renilla split luciferase system, firefly luciferase produces more stable luminescence with a longer wavelength emission spectrum and, hence, less overlap with hemoglobin and is more amenable for imaging in live animals.

We have used the split firefly luciferase complementation-based bioluminescence system to study two proteins of great importance in cancer biology: HIF-1α and pVHL. As a master regulator of the cellular functions under hypoxia, HIF-1α targets genes that promote malignant progression, metastasis, and resistance to therapy.¹¹ The VHL protein is required for degradation of HIF-1α, and its absence predisposes patients to cancer. Therefore, HIF-1α appears to be an attractive target for anticancer therapy, and manipulation of pVHL regulation of HIF-1α may be beneficial for developing novel chemoprevention strategies in pVHL patients.

Current understanding of the interaction between HIF-1α and pVHL has thus far been gained using methods such as reporter assays, immunoblotting, and pVHL capture studies. The limitations of such techniques are that they are indirect measures of the interaction, represent static snapshots, cannot be used in live animals, and are not adaptable for high-throughput screening. The HIF-1α and pVHL split luciferase complementation system circumvents these problems. Using the split luciferase chimeras expressing HIF-1α and pVHL, we have engineered a system that allows in vivo visualization of the proline hydroxylation–dependent interaction between HIF-1α and pVHL. In addition, the HIF-1α/pVHL split luciferase system can be used to study mutations found in VHL families and how such mutations may affect the interaction between HIF-1α and pVHL.

The split firefly luciferase system has its shortcomings. As is the case with most complementation techniques, complemented luciferase activity is lower than intact reporter activity. In addition, not only do the two proteins of interest have to interact, but they must also do so in such a way as to allow the two split luciferase fragments to come together for activity. For this reason, orientation of the split luciferases matters. Whereas NLUC-HIF-1α and VHL-CLUC are able to complement luciferase activity, NLUC-VHL and HIF-1α-CLUC cannot (see Figure 1B). This may be because NLUC and CLUC cannot align in proper conformation in the latter case. In addition, the sizes of the fusion proteins affect complementation-based luciferase activity. Although the entire ODD domain is capable of complementing pVHL luciferase, higher bioluminescence signal was observed when smaller ODD fragments were used (see Figure 1D). This may be because smaller ODD fragments provide less steric hindrance and allow better complementation between NLUC and CLUC. Finally, because firefly luciferase is a bigger protein than Renilla luciferase, the split firefly luciferase system may be more likely to suffer from potential steric hindrance problems than the split Renilla luciferase system.

This system represents a novel technique for studying the interaction between HIF-1α and pVHL in vivo. We have established a new technique for studying and visualizing the interaction between an E3 ubiquitin ligase and its substrate and investigating the importance of post-translational modifications in protein–protein interaction. In addition, the simplicity of the system will allow high-throughput drug screening to identify compounds that modulate the interaction between HIF-1α and pVHL. Once such compounds are identified, imaging studies will allow for intratumor target validation assessment in animals. Finally, the utility of this system is not limited to HIF-1α and pVHL. One can imagine how this system can be used to study (1) proteins involved in the HIF-1α signaling pathway (eg, HIF-1β and factor inhibiting HIF), (2) E3 ubiquitin ligase and its substrate (eg, MDM2 and p53), and (3) post-translational modifications in protein–protein interaction (e.g., protein phosphorylation and acetylation).