Biophysical Characterization of the Complex between Human Papillomavirus E6 Protein and Synapse-associated Protein 97*♦

The E6 protein of human papillomavirus (HPV) exhibits complex interaction patterns with several host proteins, and their roles in HPV-mediated oncogenesis have proved challenging to study. Here we use several biophysical techniques to explore the binding of E6 to the three PDZ domains of the tumor suppressor protein synapse-associated protein 97 (SAP97). All of the potential binding sites in SAP97 bind E6 with micromolar affinity. The dissociation rate constants govern the different affinities of HPV16 and HPV18 E6 for SAP97. Unexpectedly, binding is not mutually exclusive, and all three PDZ domains can simultaneously bind E6. Intriguingly, this quaternary complex has the same apparent hydrodynamic volume as the unliganded PDZ region, suggesting that a conformational change occurs in the PDZ region upon binding, a conclusion supported by kinetic experiments. Using NMR, we discovered a new mode of interaction between E6 and PDZ: a subset of residues distal to the canonical binding pocket in the PDZ2 domain exhibited noncanonical interactions with the E6 protein. This is consistent with a larger proportion of the protein surface defining binding specificity, as compared with that reported previously.

The E6 protein of certain human papillomaviruses (HPVs) 4 (1-3) plays an active role in the development and pathogenesis of several types of cancers, with cervical cancer being the most prevalent type (4,5). HPV uses its E6 and E7 proteins, which interact with and inhibit several key proteins, to hijack the erstwhile highly controlled cellular environment (4, 6 -8). HPVs are broadly divided into two main groups: "high risk" and "low risk," based on their occurrence in cervical cancer (9), and HPV16 and 18 are the two most common high risk types of HPV. One of the hallmarks of HPV pathogenesis is the E6-mediated inactivation of the p53 tumor suppressor protein. However, a number of studies have shown the existence of a p53-independent mechanism that leads to uncontrolled cell growth (10 -12). For example, the E6 protein of the high risk (but not the low risk type) HPV interacts with the PDZ (PSD-95/Discs large/ZO-1) domains of several proteins such as SAP97 (synapse-associated protein 97), hScrib, MAGI, and MUPP1, through a conserved C-terminal motif (13)(14)(15)(16)(17). These PDZ-containing proteins participate in the maintenance of cell-cell contacts and cell polarity and are often found at tight and adherent junctions (18). The tumor suppressor protein SAP97 contains a series of three consecutive PDZ domains (PDZ 1 , PDZ 2 , and PDZ 3 ), one Src homology 3 domain, and one guanylate kinase-like domain. The mechanism by which the E6 protein interacts with these three PDZ domains is not well understood. It has been shown through cell pulldown assays that only the PDZ 2 domain of SAP97 interacts with the E6 protein of HPV16, whereas the E6 protein of HPV18 interacts with all three PDZ domains (13,15). Biochemical characterization of the PDZ-E6 interaction have shown that discrimination between the E6 proteins of HPV16 and HPV18 is due to the C-terminal amino acid, which is Leu-151 in HPV16 and Val-158 in HPV18 (19). Because PDZ domains are often organized in arrays, as in SAP97 (20), this poses the question: how do such repeats affect the binding to the HPV E6 protein, if at all? For example, steric occlusion, allostery, or cooperativity in binding could regulate SAP97-E6 interactions and further contribute to differential specificity toward HPV E6 variants.
To dissect these issues, we analyzed the interaction between HPV E6 and six SAP97 PDZ domain constructs: PDZ 1 ; PDZ 2 ; PDZ 3 ; the tandem constructs PDZ 12 and PDZ 23 ; and PDZ 123, which contains all three concatenated PDZ domains. We used the complete C-terminal domain of E6 and the entire PDZ region of SAP97 and combined both equilibrium and kinetic experiments in solution (21,22).
Expression and Purification-Expression of PDZ domains was as described previously (24,25). For NMR experiments, cells containing the plasmid of PDZ 2 were grown in M9 minimal medium with 15 NH 4 Cl and/or [ 13 C]D-glucose as the sole source for nitrogen and carbon, respectively. The E6 WT and E6 L151V proteins were expressed as lipoyl domain fusion protein (see "cDNA Constructs"). Expression was done overnight at 25°C. The cells were harvested by spinning at 7,000 ϫ g for 10 min and resuspended in purification buffer (50 mM Tris/ HCl, 400 mM NaCl for PDZ domains and 50 mM Tris/HCl, 400 mM NaCl, 0.2% 2-mercaptoethanol for E6 proteins). The cells were lysed by sonication and thereafter centrifuged at 35,000 ϫ g for 1 h. The supernatant was filtered and loaded onto nickel (II)-charged chelating Sepharose FF column (Amersham Biosciences), equilibrated with purification buffer as above, and washed with 400 ml of the same buffer. The bound protein was eluted with 250 mM imidazole, pH 7.9, in 0.2% 2-mercaptoethanol in aliquots of 8 ml. Fractions containing proteins were pooled and further purified as follows: PDZ proteins were concentrated and purified on G-50 Sephadex (GE Healthcare) (PDZ 1 , PDZ 2 , and PDZ 3 ) or G-200 Superdex (GE Healthcare) (PDZ 12 , PSZ 23 , and PDZ 123 ) gel filtration chromatography column equilibrated with 100 mM potassium phosphate, pH 7.0. E6 proteins were first purified on an anion exchange column equilibrated with 50 mM Tris/HCl, pH 8.5, in 0.2% 2-mercaptoenthanol. Pure lipo-E6 was eluted at a gradient of 0 -500 mM NaCl, 50 mM Tris/HCl, pH 8.5, in 0.2% 2-mercaptoethanol. The lipoyl domain from the fractions containing purified lipo-E6 was then digested out with thrombin for 3 h at 37°C, filtered, and loaded on a cation exchange column equilibrated with 50 mM Tris/HCl, pH 7.5, in 0.2% 2-mercaptoethanol. Pure E6 was eluted by a gradient of 0 -500 mM NaCl in 50 mM Tris/HCl, pH 8.5, 0.2% 2-mercaptoethanol. The purity of the PDZ and E6 proteins were checked on SDS-PAGE stained with Coomassie Brilliant Blue, and their identity was confirmed by MALDI-TOF mass spectrometry. The amount of zinc bound to E6 was determined by a commercially available inactively coupled plasma atomic emission spectroscopy platform (ALS Scandinavia AB). Zinc was present in approximately a 1:1 molar ratio with the E6 protein, as reported previously (26). The protein concentrations were determined by amino acid analysis. The stability of and secondary structure of the expressed proteins were assessed using far UV CD on a Jasco J-810 Spectropolarimeter. CD spectra were recoded between 200 and 260 nm at 25°C with 20 -40 M protein in 50 mM potassium phosphate, pH 7.5.
Isothermal Titration Calorimetry (ITC), Fluorometric, and Stopped Flow Experiments-Calorimetric and fluorometric experiments were performed at 10°C in 50 mM potassium phosphate buffer, pH 7.5. E6 protein (twenty 2-l injections at 180-s intervals; stirring speed of 1000 rpm) was titrated into the PDZ solution using a microcalorimeter (ITC200; Microcal). The experiments were designed so that the c values were within 1-1000 (c value ϭ K a ϫ [Protein] ϫ N, where K a is the equilibrium association constant, [Protein] is the protein concentration, and N is the stoichiometry of the binding event). Heats of dilution were initially determined by titrating the E6 into buffer and buffer into PDZ protein, respectively. ORIGIN 7.0 (Microcal) was used to determine the thermodynamic properties of ligand binding using nonlinear least squares fitting assuming a single-site model. All of the values were the averages of two to five individual experiments. Equilibrium fluorometric measurements were performed by measuring the increase in tryptophan fluorescence upon binding in an SLM 4800 spectrofluorimeter (SLM Instruments). Excitation was at 290 nm, and emission was at 320 -360 nm. To determine the equilibrium constants for the E6-PDZ interaction, PDZ concentration was varied while keeping the concentration of E6 protein constant at 3 M. The data were then fitted to the standard equation for equilibrium binding to obtain the K d . All of the stopped flow binding experiments were done on an SX-20MV stopped flow spectrometer (Applied Photophysics, Leatherhead, UK). Fluorescence was monitored using the increase in tryptophan emission (excitation, 295 nm; emission, 330 Ϯ 30 nm). To determine the rate constants for the E6-PDZ interaction, the PDZ concentration was varied at a constant concentration of E6 (3 M). When estimating the amount of binding sites in concatenated PDZ domains, the E6 concentration was varied at a constant concentration of PDZ (3 M). Kinetic traces from time-resolved E6-PDZ binding experiments were fitted to single and double exponential functions, where A is the signal recorded with time t, ⌬A EQ and ⌬B EQ , are the amplitudes of the respective phase, and k obs is the observed rate constants. The k obs values were plotted versus PDZ or E6 concentration and fitted to the general equation for reversible association of two molecules (27,28).
k on is the association or on rate constant, k off is the dissociation or off rate constant, and [A] 0 and n are the initial concentrations of the varied and constants species, respectively.
Size Exclusion Chromatography-Multi-angle Laser Light Scattering (SEC-MALS)-MALS measurements were performed using 50 mM sodium phosphate buffer, pH 7.5 (ionic strength corrected to 150 mM using NaCl). SEC-MALS measurements were executed using a Shimadzu HPLC (to facilitate analytical SEC) daisy-chained with a Wyatt Technologies Dawn HeleosII multi-angle light scattering detector and Optilab dRX refractometer (Wyatt Tech). A column oven combined with the use of Peltier-controlled autosampler, light scattering detector and refractometer effected temperature thermostatting to 20°C (Ϯ 0.1°C) throughout the set-up. Test injections of 1 mg/ml BSA solutions were used to determine the delay volumes between instruments and the effects of band-broadening therein, using the Astra software as per the manufacturer's recommendations (Wyatt Tech).
Two different analytical size exclusion chromatography columns were used in these studies: (i) a 8 ϫ 300-mm silicabased KW802.5 size exclusion chromatography column (Shodex) for studies of PDZ 12 , PDZ 23 , PDZ 123 , and complexes with E6 L151V and (ii) a polymer-based 10 ϫ 300-mm Superdex 75 column (GE Healthcare) to study the apo form of E6. SEC-MALS measurements typically involved a 25-50-l injection of a given protein at a flow rate of 0.5 ml/min. Ligand-free forms of PDZ 12 , PDZ 23 , and PDZ 123 were loaded onto SEC columns at a final protein concentration of 233 M, whereas free E6 L151V was studied over a wide concentration range of 500 -1340 M. PDZ 123 -E6 L151V complexes were formed by incubating ϳ120 M PDZ 123 with either a 2-or 3-fold molar excess of apo-E6 L151V , for at least 2 h at 20°C prior to SEC-MALS measurements.
Molar masses were determined by measuring the intensity of scattered light at 18 different scattering angles, as a function of protein concentration and elution volume (from the SEC column). Thereafter, the intrinsic instrumental base line for each data channel was subtracted, and the molar mass across a given two-dimensional slice of the elution profile was determined using the ASTRA software (Wyatt Tech). To ensure robust data fitting, for each elution peak the apparent molar mass was determined as a function of the width of the fitted window, and also for the front end, the center and trailing edge of the peaks were compared. In general, the studied materials were highly homogeneous and mono-disperse, with the measured molar masses being highly reproducible and independent of the data window used for curve fitting. The exception to this was the unbound form of E6 L151V , which tended to form heterogeneous, higher order oligomers, and aggregates, as reported previously (29,30).
NMR Titration-15 N-1 H titration experiments were acquired on Varian INOVA 600 and 800 MHz spectrometers equipped with cryogenically cooled and room temperature probes, respectively, at 310 K in 50 mM potassium phosphate pH 6.9. Protein samples were dissolved in 10% D 2 O, and twodimensional 15 N HSQC spectra for PDZ 2 (300 M) were recorded with increasing concentrations of E6 L151V or E6 WT (0 -290 M). For assignment purposes 15 N NOESY-HSQC, 15 N TOCSY-HSQC, HNCACB, CBCA(CO)NH, HN(CA)CO, and HNCO experiments were recorded for the unbound PDZ 2 using pulse field gradient enhanced NMR spectroscopy.
The data processing and analysis were done with NMRpipe (31) and Sparky NMR assignment and integration software, respectively. Backbone assignments for the free state were essentially complete. Residues Val-313, Leu-329, and Asn-376 could, however, not be assigned, presumably because of severe line broadening. 15 N and 1 H N assignments for the bound states were obtained by following the peaks in the titration experiments. Analysis and calculation of chemical shift changes were performed using Equation 4 and as described (33).
CSP is the combined chemical shift change, ⌬ H is the change for 1 H, and ⌬ N is the change for 15 N in units of ppm. NMR assignment data has been deposited in the BioMagResBank (accession number 17373).

RESULTS
E6 Binds to All Three PDZ Domains of SAP97-First we investigated whether each PDZ domain, PDZ 1 , PDZ 2 , and PDZ 3 , individually binds to the C-terminal domain of HPV16 E6. Standard equilibrium binding and time-resolved stopped flow experiments were used to determine affinity constants for these interactions. The E6 protein contains a tryptophan at position 132, which serves as a convenient fluorescent reporter group for binding studies. An increase in tryptophan fluorescence was recorded when either E6 WT or E6 L151V was mixed with PDZ 1 , PDZ 2 , and PDZ 3 , respectively. Experiments with increasing concentrations of PDZ resulted in saturation of the fluorescence signal at high concentration of PDZ, showing that each of PDZ 1 , PDZ 2 , and PDZ 3 interacts with E6 WT in vitro (not shown). The E6 L151V mutant has a higher affinity toward the PDZ domains as compared with the E6 WT and has similar immortalization properties as the HPV18 derived E6 protein (19). Thus, we decided to use this mutant as a mimic for HPV18 E6 for subsequent experiments. We performed ITC binding experiments to independently characterize E6 L151V -PDZ interactions and to probe the thermodynamics of binding (Table 1 and Fig. 1). As shown in Table 1, the binding was enthalpy-driven for PDZ 2 and PDZ 3 , whereas for PDZ 1 , a favorable entropy was observed. Among the three PDZ domains of SAP97, PDZ 3 had the highest affinity for the E6 L151V . Importantly, when performing ITC experiments on the tandem constructs PDZ 12 , PDZ 23 , and PDZ 123 , the binding stoichiometries agreed well with a model where each PDZ domain can bind one E6 L151V molecule ( Table 1).
Compaction of SAP97 PDZ 123 Occurs upon Complex Formation with E6-To investigate whether any conformational changes were reflected in changes in hydrodynamic volume on E6 L151V -PDZ complex formation and to directly measure the stoichiometry of binding, we combined size exclusion chromatography with SEC-MALS. SEC-MALS combines the resolving power of SEC, with the ability of MALS to provide accurate, absolute measurements of molecular mass as different biomolecules elute from the SEC column. Purified PDZ 12 , PDZ 23 , and PDZ 123 preparations each eluted as single peaks from analytical SEC columns (see "Materials and Methods" and Fig. 1d). In each case, the measured molar mass using SEC-MALS was very close to the theoretical mass for a monomer, and there was no evidence for mass heterogeneity across the eluted peak (Table 2). For these concatenated PDZ domains, the relationship between elution volume and molar mass was that typically expected for globular proteins, with the larger proteins eluting before smaller ones: PDZ 123 (measured mass, ϳ36.8 kDa) eluted first followed by PDZ 23 (measured mass, ϳ27.0 kDa) then PDZ 12 (measured mass, ϳ21.6 kDa). Thus, each PDZ concatenamer behaves as a homogene-ous, stable monomer, even at relatively high protein concentrations (here between 5 and 8.6 mg/ml).
The situation for E6 L151V was, however, more complicated. Small monomeric proteins, for example the size of an E6 monomer, inherently scatter light quite weakly, thus requiring the use of higher protein concentrations to obtain measurable scattering signals. Thus, it is important to bear in mind that the use of high protein concentrations for SEC-MALS may affect the oligomeric state of the protein under study. Recombinant E6 L151V preparations, although pure and unmodified  Table 1 for fitted parameters. d and e, SEC-MALS analysis of PDZ 123 in the presence and absence of a 3-fold molar excess of HPV E6. d, the elution volumes from an analytical Superdex 75 10/30 column were very similar for the apo-and saturated PZD 123 -E6 complex (red and black traces, respectively). For clarity, the UV signals of PDZ 123 peaks are normalized to have similar intensities. The additional peak at ϳ19.4 ml corresponds to unbound HPV E6 (and reductants in the E6 samples used to prevent thiol-mediated oligomerization). e, SEC-MALS was used to determine the molar mass across the peaks for the apo-and saturated PZD 123 -E6 complex (trace colors are as in d). PDZ 123 exists as a discrete monomer, even at high protein concentrations (Ͼ8 mg/ml). Despite having a similar elution volume to apo-PDZ 123 , the saturated PDZ 123 -E6 complex has a significantly increased molar mass. This mass difference was consistent with each PDZ 123 binding three E6 molecules, which agrees with the number of binding sites available and independent ITC and kinetic measurements (Table 1).

TABLE 1 Isothermal titration calorimetric data for the interaction between E6 L151V and the respective PDZ domains
Stoichiometry from stopped flow is also included (see "Results").   (21,29,30). Whereas the oligomeric state of E6 was not consistent at the protein concentrations required for SEC-MALS, we did observe, with one preparation, a single clear E6 L151V peak eluting from the SEC column that had a molecular mass consistent with a 50 -50% mixture of E6 monomers and dimers (Table 2). Having demonstrated that each PDZ construct was monomeric and that E6 L151V tends to form higher order oligomers, we then used SEC-MALS to try and independently determine the apparent binding stoichiometry. Given the complex solution behavior of E6 L151V , the measured molecular masses for PDZ 123 complexes were surprisingly straightforward. At a molar excess of just over 2.25 molecules of E6 L151V to one PDZ 123 , we found no evidence for any complex larger than a 2:1 complex (Table 2). Similarly, at a molar excess of three molecules of E6 to one PDZ 123 , there was no species larger than a 3:1 complex (Table 2 and Fig. 1, d and e). Thus, it appears that PDZ 123 , which contains the three PDZ domains of SAP97, can bind three molecules of E6 L151V . We cannot exclude the possibility that if E6 L151V can form stable dimers (as per the singular observation for apo-E6 L151V ), what appears to be a 3:1 complex comprises only two occupied binding sites, one of which contains a bound dimer. However, the fact that we measured a 3:1 complex from ITC and, as discussed below, from kinetics, corroborates that PDZ 123 has three binding sites for the E6 protein. Thus, the 3:1 complex observed from SEC-MALS appears to be an accurate observation and not an artifact from high protein concentrations. Interestingly, the elution volumes for this E6 L151V -PDZ 123 3:1 complex were extremely similar to that of the apo-PDZ 123 (15.65 versus 15.5 ml, respectively). This similar elution volume suggests that the hydrodynamic shapes of apo-and saturated PDZ 123 are very similar despite their large difference in molecular mass (36,922 versus 63,198 Da). One likely interpretation of this observation is that the complex has undergone a conformational change upon binding of E6. The data in Fig. 1 (d and e) were interpreted with the utmost caution. In reproducible 3:1 SEC-MALS experiments, we obtained three main "peaks": (i) the front end of the first peak; (ii) the tail of the first peak; and (iii) unbound E6, presumably in a mixed oligomeric state, as per the apo-form of E6. Despite repeated attempts with the highest resolution silica SEC columns, these peaks could not be further resolved. Thus, any mass determination across the peak is a weighted average of the masses for all species present. For this reason, we fitted the mass for the peak (i), the front end of the main peak that was minimally "contaminated" by other species. The fact that the mass of this species was virtually identical to that expected for the saturated complex strongly suggests that it corresponds to the 3:1 complex. Peak (ii) is likely a mixture of E6 oligomers and perhaps partially saturated PDZ 123 , whereas the third peak is unbound E6.

Protein
The Dissociation Rate from SAP97 Distinguishes the Binding of HPV16 and HPV18 E6 Proteins-Having established that all three PDZ domains of SAP97 bind E6 and that PDZ 123 can form a 1:3 complex with the E6, we wanted to learn more about the mechanism of the SAP97 and E6 interaction, by studying binding kinetics. Sorting out the mechanism of interaction of ligands, such as E6, that bind to multiple binding sites on the same protein is a complex task. However, a useful approach is to investigate binding at the respective binding site individually and then in concert. Initially we measured the on and off rate constants for the PDZ-E6 WT and PDZ-E6 L151V interactions by rapidly mixing the proteins in a stopped flow spectrometer and monitoring the change in intrinsic fluorescence of the E6 protein upon binding. The on-off rate constants (Table 3) were estimated from a plot of the observed  (Table 3). Therefore, the higher off rate constants of HPV16 E6, as compared with those of the HPV18 E6 mimic E6 L151V , account for its lower affinity toward the SAP97 PDZ domains, in particular for PDZ 2 and PDZ 3 .

Concatenation of PDZ Domains Modulates
Binding-To investigate whether the E6/PDZ interaction in SAP97 is influenced by neighboring PDZ domains, we performed stopped flow binding experiments using the concatenated PDZ constructs PDZ 12 , PDZ 23 , and PDZ 123 . The binding reaction was monophasic for E6-PDZ 12 (Equation 1). Interestingly, E6-PDZ 23 and E6-PDZ 123 displayed biphasic binding kinetics (Equation 2) including a slow phase, which appeared to decrease (within error) with increasing concentrations of PDZ 23 or PDZ 123 (Fig. 2b). Such a phase is fully consistent with an initial binding to both domains followed by a slow equilibration governed by the respective off rate constants of the two domains. The on-off rate constants were estimated from the slopes and intercept of the fast phase (Fig. 2b). The on rate constant for the PDZ 12 construct was similar or slightly higher as compared with PDZ 1 or PDZ 2 alone (8.7 M Ϫ1 s Ϫ1 compared with 6.9 and 6.7 M Ϫ1 s Ϫ1 , respectively), whereas the on rate constants for PDZ 23 and PDZ 123 were significantly higher (25 and 27 M Ϫ1 s Ϫ1 , respectively) compared with those of the individual domains. The kinetics for multivalent binding is complex and has been discussed in more detail elsewhere (34). In theory, any of the individual on rate constants or their sums may appear as a phase. Here, it is clear that the k on of PDZ 12 is different from the sum of those for PDZ 1 and PDZ 2 . For PDZ 23 , the k on appears to be higher than the sum of PDZ 2 and PDZ 3 , whereas the observed k on for PDZ 123 is close to the expected or slightly higher. In fact, the on rate constants of PDZ 23 and PDZ 123 appear to be similar, which probably is a reflection of the low k on of the PDZ 12 tandem, which results in a small difference between PDZ 23 and PDZ 123 . We conclude, based on the nonadditive . c and d, observed rate constants for PDZ 3 /E6 L151V (c) and PDZ 123 /E6 L151V displacement (d) plotted against increasing concentration of the peptide used to trap the PDZ proteins subsequent to the dissociation from E6 L151V (SRTRRETQV, corresponding to the C terminus of E6 L151V ). e, experiment to estimate the number of binding sites in PDZ 12 . The concentration of E6 was increased at a constant concentration of PDZ 12 (6 M), and Equation 3 was fitted to the data to obtain n (Table 1). f, the observed rate constant k obs as a function of E6 L151V at a constant concentration of either PDZ 2 or PDZ 3 (1 M). The solid line is the expected dependence of k obs for E6 L151V under pseudo-first order conditions according to the parameters in Table 3, determined by varying PDZ 2 . The kink in the data between 5 and 10 M may reflect dissociation of E6 oligomers prior to the association with PDZ 2 or PDZ 3 . See Table 3 for fitted parameters. on rate constants for PDZ 12 and on the similar on rate constants for PDZ 23 and PDZ 123 , that some adverse steric effect modulates the binding of E6 to PDZ 12 , which decreases the on rate constant to one or both of the domains.
To get an independent measure of the off rate constants, we performed displacement experiments as described previously (24). Two observed off rate constants were obtained for each of PDZ 12 , PDZ 23 , and PDZ 123 (Fig. 2 and Table 3). For PDZ 123 one might expect three observed off rate constants, but because all three k off values for the single domains are within 1 order of magnitude, a triple exponential is difficult to resolve and may well appear as a double exponential in the experiment. The off rate constants of PDZ 12 and PDZ 23 agreed well with those measured for the respective single domains in similar experiments ( Table 3), suggesting that the tandem PDZ domains bind E6 simultaneously and thus further corroborating the stoichiometries determined by SEC-MALS and ITC. However, because the precision and accuracy is very high in displacement experiments, the 2-fold lower off rate constants of these two tandems, as compared with the single domains, indicate additional interactions that are not present in the single PDZ-E6 complexes. Such additional interactions may include a conformational change upon binding.
To further confirm the number of E6 interaction sites in the PDZ constructs, the E6 binding was assessed by stopped flow in the region of second order kinetics (Equation 3, Fig.  2e, and Table 1). These data agreed well with the stoichiometry observed in ITC and SEC-MALS.
Dissociation of E6 Oligomers Appears Rate-limiting for Binding at High Micromolar Concentration of E6-The fact that full-length E6 forms oligomers has been demonstrated previously (21,29). Here, we showed by SEC-MALS that this oligomerization occurs also for the E6 C-terminal domain. This behavior would explain the observed rate constant for binding of E6 to PDZ domains, when the E6 concentration was increased (Fig. 2f). With ϳ10 M of E6 (i.e. 20 M before the 1:1 mixing in the stopped flow), the linear increase in k obs was abruptly stopped, and the dependence was apparently "saturated." However, this rather distinct kink in the dependence of k obs on E6 concentration is likely the result of an aggregation/oligomerization event rather than of a change in rate-limiting step in the binding reaction. The observed behavior is consistent with a mechanism where the E6 forms oligomeric species that need to separate before associating with the PDZ domain. At lower concentrations of E6, the kinetic data were consistent with E6 being a monomer, and its C terminus was thus accessible for binding. First, dilution of E6 into buffer (going from 6 to 3 M) resulted in a trace, which was a perfect straight line. Second, when PDZ 1 , PDZ 2 , and PDZ 3 concentrations were varied (Fig. 2a), there was no evidence for a rate-limiting dissociation of E6. Instead, the traces were perfectly monophasic (Equation 1), and k obs was increasing linearly with PDZ concentration, showing that E6 did not form (rate-limiting) oligomers at the initial concentration of the experiment (6 M). It is also worth noting that the apparent rate of the proposed dissociation at higher E6 concen-trations appears to differ between those of PDZ 2 and PDZ 3 in Fig. 2f (ϳ20 s Ϫ1 ) and PDZ 12 in Fig. 2e (40 -50 s Ϫ1 ). The discrepancy in rate constants is, however, consistent with the heterogeneity of the oligomeric states that E6 may populate, as suggested by the SEC-MALS experiments. In conclusion, the C-terminal domain of E6 appears to oligomerize at ϳ10 -20 M, and the rate of dissociation is 20 -50 s Ϫ1 . Note that the on-off rate constants and equilibrium constants reported in Table 3 are not affected by this E6 oligomerization, because the E6 concentration was below 10 M in these experiments. Further, the on rate constant for the association of a short peptide corresponding to the C terminus of HPV18 E6 and SAP97 PDZ 2 (24) is virtually identical to that for the E6 L151V -PDZ2 in the present study. In the ITC experiments, the data points at higher E6 concentrations might be affected by the equilibrium between free and oligomerized E6 molecules, but the general agreement between K d from kinetics and ITC suggests that there is a minor effect on the fitted parameters.
Noncanonical Interactions Are Involved in the Binding between E6 and PDZ 2 -To investigate the changes that occur upon binding of E6 to a PDZ domain at the molecular level on a residue per residue basis, we performed NMR titration experiments on the interactions between the single PDZ 2 domain and E6 WT and E6 L151V , respectively. PDZ 2 was chosen for the NMR studies because extensive biophysical characterization of the interaction between PDZ 2 and peptides corresponding to the C terminus of the E6 protein has been performed (24,33,35). 15 N-1 H correlation maps were recorded for the free and bound forms of the PDZ 2 protein (Fig. 3), thus allowing for subtle effects in the E6 protein-PDZ interaction to be sorted out. The combined chemical shift changes of amides were calculated using Equation 4. Structural changes were mapped based on the extent each peak moved in the bound state relative to the free state and were divided into four different classes: large, medium, small, and no change (see legend to Fig.  3). Similar structural changes were seen for the E6 WT -PDZ 2 and E6 L151V -PDZ 2 interactions ( Fig. 3 and supplemental materials). Those residues that underwent large and medium chemical shift changes were divided into two groups based on their proximity to the binding site. Group 1 comprised those residues close to the canonical peptidebinding site of the PDZ domain (36), which are expected to make a direct or indirect interaction with the E6 (Lys-324, Gly-328, Gly-330, Gly-335, His-341, and Asn-393), whereas group 2 comprised those distant from the ligand binding site (Thr-351, Ile-354, Glu-355, Glu-356, Glu-357, His-360, Leu-371, Glu-380, Glu-385, and Phe-397) (Fig. 3). Interestingly, residues Glu-355, Gly-357, Leu-371, and Phe-397 (distance to side-chains of the bound peptide in an NMR structure was 8.7-13.5 Å; Fig. 3, b and c) that underwent large or medium change in the E6-PDZ 2 interaction exhibited small or no change when the C-terminal peptide of E6 was used (4, 6 -8, 33). Therefore, residues Glu-355, Gly-357, Leu-371, and Phe-397 constitute a subset of residues that participate in noncanonical interactions either via in-

DISCUSSION
Cervical cancer is caused by certain strains of HPV through expression of the oncogenic proteins E6 and E7. The E6 protein has evolved to bind to several important host regulatory proteins, such as the tumor suppressor proteins p53 and SAP97, and thus mark these proteins for destruction to keep HPV-infected cells alive. The exact mechanisms by which host proteins are targeted by the viral proteins are very complex (4, 6 -8), and also it is not known how the viral proteins E6 and E7 can be expressed for many years without being detected and destroyed by the immune system (37). Here we have looked at mechanistic aspects of the complex formation between E6 and the PDZ domains of SAP97 and revealed several novel facets of this interaction that are recapitulated in Fig. 4.
One major finding is that the E6 proteins of both HPV16 and HPV18 bind to all three PDZ domains of SAP97 in vitro. By using a cell pulldown assay, it was previously suggested that only the PDZ 2 domain of SAP97 could bind the E6 protein of HPV16 (13), whereas E6 from HPV18 was shown to bind to all three PDZ domains from SAP97 (15). The basis of this difference is not clear. Furthermore, a swap of the last amino acids (Leu and Val) completely reversed the immortalization and binding affinities of the two E6 proteins (19,38). In this study we have measured the affinities of the respective PDZ domains with the E6 from HPV16 and a "pseudo" HPV18 E6 protein (E6 L151V ). All three PDZ domains have similar association or on rate constants for the two different E6 constructs. However, the dissociation (off rate) constants differed and may explain the observed difference in virulence between the two HPV types (19,38). Although Val and Leu residues both have a hydrophobic side chain, the Val residue of HPV18 E6 is smaller than the Leu of HPV16 E6, which may allow for a more snug fit in the hydrophobic binding pockets of the PDZ domains.
Contrary to the previous results, we find that PDZ 1 displays the highest affinity for HPV16 E6, whereas PDZ 3 binds strongest to E6 L151V (i.e. to HPV18 E6). However, the affinities are within the same order of magnitude (Tables 1 and 3), suggesting a high degree of promiscuity in E6-PDZ interactions. Furthermore, SEC-MALS, ITC, and kinetic studies show that all three putative binding sites in SAP97 are occupied by E6 proteins in a quaternary complex. Previous experiments using electron microscopy on full-length SAP97 showed that the molecule is present in a monomer-dimer equilibrium and that the dimerization occurs via its N-terminal L27 domain (39). Furthermore, it was demonstrated that monomeric SAP97 is a relatively dynamic protein that exists either in an extended conformation or as a more compact ring-like structure. In another study on the PDZ region of SAP97, it was demonstrated by small angle x-ray scattering that PDZ 123 is indeed flexible, in particular the region between PDZ 2 and PDZ 3 (40). It was further shown that the PDZ 12 part was more conformationally restricted and displayed a dumbbell-like shape, in agreement with our kinetic FIGURE 3. NMR data of the E6-PDZ interaction. a, an overlay of HSQC spectra of free PDZ 2 (red) and E6 L151V -PDZ 2 complex (blue). The arrows indicate residues that moved substantially and are discussed in the text. b and c, residues in PDZ 2 undergoing structural change upon binding to the E6 protein. Color codes for chemical shift changes of residues in PDZ 2 (Protein Data Bank code 2IOL) are: red, large change, ⌬ Ն 0.15 ppm; orange, medium change, 0.10 Յ ⌬ Ͻ 0.15 ppm; yellow, small change, 0.05 Յ ⌬ Ͻ 0.10 ppm; and gray, no change, 0.00 Յ ⌬ Ͻ 0.05 ppm. The PDZ 2 domain was titrated with E6 WT (b) and E6 L151V (c). The four residues that display distinct chemical shift changes as compared with binding of a C-terminal peptide (33) are labeled. The canonical binding pocket is indicated by the solid line. This picture was drawn with PyMol (32). FIGURE 4. Scheme for the interaction between HPV E6 and the PDZ domains of SAP97. E6 forms oligomers, which dissociate upon binding to SAP97 (k ϭ ϳ20 -50 s Ϫ1 ). All three PDZ domain of SAP97 may bind one E6 molecule each, and a conformational change of the quaternary complex gives a similar hydrodynamic radius to that of apo PDZ 123 . Dissociation rates govern the affinities for HPV16 and HPV18 E6 proteins for the PDZ domains of SAP97. Residues outside the canonical binding site of the second PDZ domain are affected by the E6 protein in the binding reaction.
experiments where the on rate constant for PDZ 12 is not the sum of those for PDZ 1 and PDZ 2 . The SEC-MALS experiments presented here are consistent with the PDZ region of SAP97 undergoing a structural change, perhaps a collapse or a compaction, upon interaction with the E6 proteins. These data are supported by rate constants of dissociation from stopped flow fluorescence, in particular the slower dissociations from PDZ 12 , PDZ 23 , and PDZ 123 , as compared with those of the single domains (Fig. 2, c and d, and Table 3). Whether the quaternary complex occurs in vivo depends on the expression levels of the E6 protein and of SAP97. Likewise, the reported oligomerization of E6 (21,29,30), which is also observed here, would also be dependent on E6 concentration, i.e. the expression level. Immunoblots suggest that the expression of E6 in vivo is low (41,42), but actual concentrations have not been reported and are very difficult to estimate. Expression of E6 may also vary both temporally and spatially, and the effects of oligomerization and dissociation of E6 in the infected cell remain to be investigated.
PDZ domains bind ligands like the E6 protein through the C terminus of the ligand, which becomes a strand in an extended ␤-sheet (43). Nevertheless, the potential of allosteric interactions in PDZ domains has been discussed and experimentally substantiated (36, 44 -47). Here, by NMR experiments, we found residues in PDZ 2 that are not part of the peptide-binding pocket but nevertheless experience chemical shift changes upon binding of E6. Importantly, four of these residues (Fig. 3) were unaffected by binding of an E6 C-terminal peptide (33). This subset of four residues, distal from the binding pocket, may change chemical shifts either through an intradomain allosteric effect (45) or by direct interactions between other parts of the E6 than its C terminus. In either case these noncanonical interactions may be employed by the E6 protein to increase affinity or even modulate binding to a third partner. Low affinity nonspecific interactions between E6 and PDZ 2 are less likely to cause these changes in chemical shifts because these four residues saturate at similar concentration as residues in the canonical binding pocket, upon titration with E6.
In conclusion, we report affinities, stoichiometries, rate constants, and conformational change(s) for the complex between E6 and SAP97. As a step toward better understanding of E6-mediated oncogenesis, our results highlight the dynamic nature of the E6-SAP97 binding and reveal mechanistic and molecular details of the interaction.