Structural investigation on the intrinsically disordered N-terminal region of HPV16 E7 protein

Abstract

Human papillomavirus (HPV) is the major cause of cervical cancer, a deadly threat to millions of females. The early oncogene product (E7) of the high-risk HPV16 is the primary agent associated with HPV-related cervical cancers. In order to understand how E7 contributes to the transforming activity, we investigated the structural features of the flexible N-terminal region (46 residues) of E7 by carrying out N-15 heteronuclear NMR experiments and replica exchange molecular dynamics simulations. Several NMR parameters as well as simulation ensemble structures indicate that this intrinsically disordered region of E7 contains two transient (10-20% populated) helical pre-structured motifs that overlap with important target binding moieties such as an E2F-mimic motif and a pRb-binding LXCXE segment. Presence of such target-binding motifs in HPV16 E7 provides a reasonable explanation for its promiscuous target-binding behavior associated with its transforming activity. [BMB Reports 2016; 49(8): 431-436]

INTRODUCTION

Human papillomaviruses (HPVs) are the primary agents causing several types of cancers including cervical cancer, one of the leading causes of female death in developing countries. Even though anti-HPV vaccines became available (1), effective pharmaceuticals are still needed for those who have already developed an infection with HPV. Approximately 200 types of HPVs are known, and many of them are classified into high- or low-risk types based on their clinical prognosis (2). HPV16 is a high-risk alpha HPV, one of the most prevalent HPVs found in cervix carcinomas. Its genome encodes for 8 proteins among which two oncoproteins, E6 and E7, are known to be essential factors for the development of cancer by debilitating p53 (by E6) and functioning as retinoblastoma (pRb) tumor suppressor (by E7), respectively (3). E7 is known to be primarily responsible for the transforming activity, and is known to interact with diverse cellular targets (4, 5). This sort of promiscuity is observed in other viral proteins, including the X protein of hepatitis B virus (6). The reason for such promiscuity is not clearly understood.

The E7 of HPV16 consists of 98 amino acid residues and is composed of three conserved regions (CRs), based on the homology with the adenovirus E1A protein (Fig. 1). Structurally, E7 has a flexible or disordered N-terminus and a well-structured C-terminal domain (7). The disordered N-terminal region contains two repeats, CR1 (residues 2-15) and CR2 (residues 16-41). The former interacts with several cellular targets such as p600, Skp2, p300, and IRF-1, and the latter with HPV E2, Rb1, FHL2, TBP, CKII, p300, HIF-1a, and p21 to name a few. The C-terminal structured domain contains CR3 (residues 54-95) which forms a zinc finger with a β₁β₂α₁β₃α₂ topology (8, 9). Interestingly, E7 may form oligomers, depending upon solution conditions (10-13). Among various oligomers, the dimers are known to be the most prominent at physiological condition (11). Dimerization of E7 is believed to occur via the structured C-terminal domain (50-98) involving the α-helices of each monomer and an intermolecular antiparallel β-sheet between the β2 strand of one monomer and the β3 strand of the other (11).

Fig. 1. The primary sequence of HPV16 E7 oncoprotein is drawn with two conserved regions (CR1, CR2), L₂₂XCXE₂₆ motif (sky blue) that binds with the B-box of pRb, CR3 and two conserved Zn-binding motifs (C₅₈XXC₆₁, C₉₁XXC₉₄, Cys marked in yellow) within CR3. The boxes in CR3 indicate the secondary structures.

Proteins often contain short flexible linkers or disordered loops that are typically composed of less than 10-20 amino acid residues. This has been known as for decades as the phenomenon of protein disorder. A special case of protein disorder that has emerged since 1990s is intrinsically disordered proteins (IDPs), which do not form a well-defined three-dimensional structure but are nonetheless functional (14). These unorthodox proteins are defined to contain at least one long (more than 40, and up to hundreds, of residues) intrinsically disordered region (IDR). An important point is that without relying on three dimensional structures, many IDPs and IDRs are responsible for specific biological functions such as transcription, translation, chaperoning, and cell cycle regulation (15), or causing diseases such as prion diseases, neurodegeneration, cancers and so on (16, 17). Several viral proteins in HIV, HBV, SARS virus, AI virus and MERS virus are IDPs (6, 18). The mechanism of how IDPs/IDRs perform biological functions or cause diseases is not well understood as yet.

Initially, IDPs/IDRs were thought to be completely unstructured with no observable secondary structures (19, 20). However, during the last two decades, useful information was obtained function from several NMR structural investigations that helps us understand how IDPs/IDRs. In contrast to the initial conjecture, these studies demonstrated that as many as three dozen IDPs/IDRs contain transient local structural elements or pre-structured motifs (PreSMos) in their target-unbound state, which mediates the binding of IDPs/IDRs with targets (15). Based upon the PreSMo concept, IDPs/IDRs can be classified into two types: the mostly unstructured (MU) type that contain PreSMos, and the completely unstructured (CU) type with absolutely no transient secondary structures that may represent true random coils. Currently, ∼80% of IDPs/IDRs are known to be the MU type. Since IDPs/IDRs cannot form a spatially disposed target-binding pocket found in globular proteins, they seem to utilize transiently-structured PreSMos that are primed for target binding in order to interact with targets (proteins, nucleic acids, lipids, metals, etc.).

One well-known characteristic of IDPs/IDRs is their promiscuity; IDPs can interact with many partners, in contrast to globular proteins. For example, the 73-residue intrinsically disordered N-terminal transactivation domain (TAD) of p53 contains three PreSMos which enable binding of p53 TAD to different targets (21-28). Since the N-terminal region of E7 is known to be disordered and interacts with many targets (4) we hypothesized that this IDR should contain at least one PreSMo. To test this hypothesis, we investigated the structural properties of the N-terminal region (residues 1-46) of HPV16 E7 oncoprotein (N-E7) by high resolution NMR spectroscopy combined with replica exchange molecular dynamics (REMD) simulation. Results indicate that N-E7 has two clear helical PreSMos and additional non-helical transient structures. Here, we discuss the ramifications of the presence of these PreSMos in terms of E7 function.

RESULTS

A fingerprint region of a ¹⁵N-¹H heteronuclear single quantum coherence (HSQC) spectrum of N-E7 with assigned residues is shown in Fig. 2. The spectrum has narrow chemical-shift dispersion in both ¹⁵N and ¹H dimensions, indicating that N-E7 does not form a globular structure under the non-denaturing experimental condition used (aqueous buffer at pH 6.5). Such an overall disordered nature of N-E7 is consistent with the results of the previous metal-free CD spectrum (11) and NMR experiments on N-E7 (29), and also with bioinformatics predictions (30). We achieved a full NMR resonance assignment for backbone ¹⁵N and amide protons of N-E7 by following the standard triple-resonance assignment procedure, except for 3 prolines which do not have backbone amide NH nitrogens and protons. Resonance assignment for N-E7 is summarized Table S1 (See Supplementary Information). The first two N-terminal residues of the recombinant N-E7, Met, and His originated from the N-terminal glutathione-S-transferase fusion linker. The level of achieved resonance assignment was sufficient for subsequent structural characterization of N-E7, in particular, for delineation of residues that form PreSMos.

Fig. 2. A fingerprint region in a ¹⁵N-¹H HSQC spectrum of the N-terminal region of HPV16 E7 oncoprotein (N-E7) obtained at 10℃ and pH 6.5 on 90% H2O/10% D₂O. The backbone ¹⁵N and amide of 43 residues out of 46 residues were assigned (3 proline residues without amide NH are not visible in this spectrum).

Fig. 3 shows the NMR parameters measured for N-E7. The left panel in Fig. 3 is the summary of chemical shifts and SSP (secondary structure propensity) scores. The SSP scores are obtained by combining various chemical shifts (Hα, Cα, Cβ) (31) and hence are often more informative in detecting PreSMos than individual chemical shifts. Positive SSP scores over 4 residues or more indicate the formation of a helix, whereas negative values suggest non-helical (β-type) secondary structures. Fig. 3D shows that N-E7 contains two helical PreSMos, one formed by residues 7-14 and the other by residues 20-26. The first is ∼10% pre-populated and the second ∼20%. Note that the first is an E2F mimic and the second PreSMo corresponds to the well-known Rb-binding LXCXE segment. The C-terminal portion (residues 30-46 in CR2) of N-E7 has a potential to form additional non-helical (β-type) PreSMos (see Fig. 1 and Fig. 3D). A recent prediction study on full E7 using the ANCHOR program also suggested similar helical PreSMos as potential target-binding segments (30). The location of two helical PreSMos in N-E7 is also visible by ¹H–¹⁵N heteronuclear NOEs (Fig. 3E). Small positive ¹H–¹⁵N heteronuclear NOEs (typically 0-0.5) are indicative of local pre-structuring (15). ¹H–¹⁵N heteronuclear NOEs for stable secondary structures in globular proteins are 0.8-1.0, and large negative values indicate a highly flexible region. Small (< 5 ppb/K) backbone NH temperature coefficients (Fig. 3H) are evidence for formation of hydrogen bonds. Fig. 3H shows that helical PreSMos have 2 residues whose temperature coefficients are very small (down to 3 ppb/K), and additional 2-3 residues have values of ∼5.

Fig. 3. Left panel: deviation of (A) ¹Hα, (B) ¹³Cα, (C) amide carbonyl chemical shifts from random coil values and (D) the SSP (secondary structure propensity) scores. In (D) positive scores indicate helical propensity while negative values suggest formation of non-helical type PreSMos. Right panel: ¹H–¹⁵N heteronuclear NOEs (E), backbone ¹⁵N relaxation times, T₁ (F), T₂ (G), and temperature coefficients of the backbone amide hydrogens (H). The horizontal lines in (F) and (G) indicate an average value.

Note that the residues 33-34, potentially forming non-helical transient structures suggested by the SSP scores, also have small temperature coefficients. Presence of non-helical transient structures was noted in a previous ¹³C-based NMR study which used a similar-sized N-terminal fragment (residues 1-40) of E7 (29). This ¹³C-HMQC based NMR study concluded that two helical PreSMos were formed by residues 8-13 and 17-29, which was slightly different from our results. Also, the degree of pre-population for the two helical PreSMos was reported to be less than 5%, which is in clear contrast to what we observe. Furthermore, this study used as much as 50% of a helix-inducing solvent (trifluoroethanol, TFE) in order to enhance the visibility of helical segments. By definition, PreSMos are the transient structures detected in aqueous solution (15). Almost all NMR studies on IDPs/IDRs used N-15 isotope labeled proteins under aqueous solutions. In the early days of IDP research, some investigators employed hydrophobic solvents in order to make “expected” helices clearly visible in NMR measurements (32, 33); however, the helices observed in such a manner are not accepted as PreSMos (15). Hence, it is not clear how reliable the previous C-13 NMR results are in terms of characterizing PreSMos.

In order to augment the results of NMR measurements, we undertook REMD simulations on N-E7. After performing clustering analysis for the temperature trajectories, ensemble structures of highly-populated 10 clusters in each temperature trajectory were collected. In order to check the extent of agreement between the computed ensemble structures and the structures suggested by NMR data, the Hα chemical shifts calculated from the ensemble structures by Shift-X program (34) were compared with the experimentally measured NMR chemical shifts. Using this consistency check procedure, we obtained the structural ensembles of N-E7 that showed a high Pearson’s correlation (PC) with experimental results (PC coefficient > 0.65). The final ensemble structures of N-E7 possess two helical PreSMos; one is pre-populated at ∼15% and is formed by residues 3-15 which covers the most part of CR1, and the other is formed by residues 17-30 and is ∼10% pre-populated and encompasses the N-terminal half of CR2. Thus, the two calculated helical PreSMos by REMD are slightly longer than the ones detected by NMR, i.e., one with residues 7-14 and the other with 20-26. Combining NMR and REMD results we conclude that two helical PreSMos in N-E7 are formed by the residues 7-14 and by 20-26, which are common denominators suggested by both techniques. Fig. 4 shows two ensemble structures of N-E7. The left panel is an ensemble superimposed along the first helical PreSMo (residues 7-14), whereas the left panel displays an ensemble superimposed along the second helical PreSMo (residues 20-26). The residues 7-14 and 20-26 in the REMD ensembles are pre-populated at the level of ∼20% and ∼11%, respectively.

Fig. 4. Two REMD ensembles of N-E7. The left (A) is generated by aligning the residues 7-14 that form the first helical PreSMo. The right (B) is an ensemble generated by superimposing the second helical PreSMo. All ensemble structures showed a high correlation with the SSP scores from NMR experiments (PCC > 0.65, See Materials and Methods Section). Shown at the bottom (C) is a schematic diagram showing the location of two helical PreSMos suggested by both NMR and REMD simulations. The structures are color-coded from the N- to the C-terminus (blue→green→yellow→brown).

DISCUSSION

Currently, the high resolution heteronuclear multidimensional NMR spectroscopy is the best tool providing the exact location of PreSMos in IDPs/IDRs. The PreSMos are transient secondary structures detected by NMR in aqueous solutions, without using secondary structure-inducing hydrophobic solvents such as TFE. They are increasingly being acknowledged as the putative target-binding active sites in IDPs/IDRs (35, 36). Presence of PreSMos in IDPs is accepted well in recent years, especially after it was succinctly demonstrated by our laboratory three years ago that PreSMos are present in 3 dozens of MU type IDPs/IDRs (16). However, their existence in the target-free states of IDPs/IDRs was mostly ignored in the 1990s when the concept of IDPs was barely being conceived, and experimental techniques used for characterization of IDPs/IDRs, e.g., CD spectropolarimetry, gel electrophoresis, did not provide detailed per-residue structural information on IDPs/IDRs. Some early NMR reports failed to acknowledge the existence of PreSMos due to the fleeting or transient nature of PreSMos, misleading the IDP field (19, 20, 32). In fact, none of the PreSMos discovered so far are 100% pre-populated. An average pre-population of PreSMos is ∼30% (± 10). Recently, Flexible-Meccano calculations in combination with NMR parameters are possible, and show explicit distribution of different conformer populations within a structural ensemble of IDPs/IDRs (37). In using NMR parameters to characterize PreSMos, it is important to combine several parameters before a conclusion is drawn on the existence of PreSMos since employing one parameter often fails to unambiguously point to the presence of a PreSMo due to their transient nature. In some MU type IDPs/IDRs such as 4EBP1, p53TAD, and HIV-1 Nef, the existence of PreSMos is very clearly recognized. In other cases, detecting PreSMos in synucleins and preS1 of hepatitis B virus was not so straightforward (15). The N-E7 seems to belong to a latter type.

Due to its structural flexibility, it has not been straightforward to obtain the detailed structural picture of the N-terminal disordered region of E7. Attempts to understand the role of the disordered N-terminal region of E7 were made by bioinformatics analyses using various independent disorder prediction programs; it suggested that the N-terminal region of HPV16 E7 was not entirely disordered (30). In consistent with such prediction results, our investigation, together with the previous CD and NMR studies, show that the N-terminal half of E7, either N-E7 (this study) or the 40-residue N-terminal fragment (29), is disordered and contains two helical PreSMos that mediate binding of N-E7 with partners; the first PreSMo is an E2F-mimic motif and the second encompasses the LXCXE motif allowing the binding of E7 with the retinoblastoma tumor suppressor protein, pRB (38). We conclude that the N-terminal half of E7 displays promiscuity using its multiple PreSMos. The full E7 oncoprotein is therefore a hybrid type IDP resembling a prion protein, where an N-terminal disordered region coexists with a C-terminal globular domain. An abnormal behavior in the electrophoretic behavior of E7 was noticed in an earlier study and was explained in terms of general charge effects (11). With the knowledge that the full-length E7 is a hybrid type IDP, we propose that the abnormal electrophoretic behavior of E7 should be ascribed to its disordered N-terminus since IDPs are known to behave atypically in gel electrophoresis or in gel permeation chromatography eluting at a position corresponding to a larger protein (39).

MATERIALS AND METHODS

Detailed experimental and computational procedures are described in Supplementary Information.