Supplementary Information Supplementary Figures

Supplementary Figure 1: Determination of the distribution of RT on the T/P substrate from anisotropy (r) data. A) Predicted response of r and to hydrodynamic radius. The hypothetical r value for the RT-T/P complex based on molecular weight (MW) is indicated by arrow. The range for the experimentally observed r values from Figure 1a is highlighted in pink. B) Schematic illustrating the slow ( slow) and fast ( fast) tumbling of the fluorescence dye attached to the T/P substrate. The Perrin equation which includes contributions from  slow and fast is also shown. C) A simplified model of RT sliding (black oval) on the T/P substrate (blue lines). If RT moves toward the +10 position, it interacts with the fluorescein dye (green). A plot of  slow as a function of the position of RT on the T/P substrate is also shown. D) Relationship between r values, predicted from the model in (C), and the standard deviations (σ) for the distributions of RT on the T/P. The response to changes in is dependent on the constant where 1 < 10. Values of  derived were derived for the experimentally determined r values for the RT-T/P, RT-T/P-dNTP and RT-T/P-EFV complexes (dotted lines) are were used to plot Figure 1b).


Supplementary
: The cis isomer of Cy3 is stabilized through interaction with RT. A) Snapshots (100 ps increments) of the structure of Cy3 (green sticks) from a single 50 ns AMD simulation of the RT-T/P-dNTP ternary complex. The p66 and p51 subunits of RT are colored light and dark grey, respectively. The template and primer strands are colored red and orange, respectively. B,C) Representative  Cy3 trajectories of the 5'-Cy3 labeled T/P are show in the absence (B) and presence (C) of RT. In the  Cy3 histograms, the cis isomer region is highlighted in green.

Supplementary Figure 9: Rates or arrival and departure from the RT-T/P-dNTP complex in the absence and presence of EFV. A)
Representative single-molecule PIFE traces for the WT RT-T/P-dNTP complex. The red lines indicate fitted idealized states (see Methods). Definition of high ( high ) and low ( low ) dwell times are shown. B) Representative single-molecule PIFE traces for the WT RT-T/P-dNTP complex in the presence of EFV. The red lines indicate fitted idealized states (see Methods). C) Representative slow and fast rates of arrival and departure (k arrival and k departure ) for the WT RT-T/P-dNTP complex. The plot of the log of the survival function for the WT RT-T/P-dNTP complex, ln(S), versus values for arrival or departure is biphasic and is best-fit to a double-exponential curve (R 2 ≥ 0.99), whereas a linear dependence between these values would have indicated only one dominant rate. Because the slow rates dominated the S() distributions (~90% of total before EFV and ~75% of total after EFV) and were most affected by EFV, we reported these in Figure 6d. D) The fast rates of k arrival and k departure from the WT RT-T/P-dNTP complex in the absence (N = 630 traces ) and presence ( N= 573 traces) of EFV. e, The fast rates of k arrival and k departure from the K103N RT-T/P-dNTP complex in the absence (N = 926 traces) and presence (N = 725 traces) of EFV. For (D) and (E), * P<0.05, *** P<0.001, n.s., not significant (two tailed t-test, equal variance).

Supplementary Movie 1: 3D visualization of EFV-bound HIV-1 RT and the NNRTI binding pocket.
WT RT in complex with EFV (1FK9) is shown as a solvent-accessible surface; p51 is shown in brown, p66 in green. The movie begins with a view of the empty T/P binding groove flanked by the thumb (left) and fingers (right), and subsequently zooms into the NNRTI binding pocket containing EFV (yellow). K103N, K101, and E138 are indicated. Next, an overlay of EFV-bound K103N RT (1FKO) appears in lighter colors, highlighting the disruption of the E138-K101 salt bridge by K103N. Lastly, the camera zooms out to show the salt bridge in the context of the thumb and fingers subdomains.

Fluorescence anisotropy
Fluorescence anisotropy experiments were performed as previously described (5,6), using a Varian Cary Eclipse fluorescence spectrophotometer. The excitation and emission wavelength were set at 485 and 520 nm, respectively, and the excitation and emission slit widths were set at 5 and 10 nm respectively. The concentration of T/P in all experiments was 5 nM (in a 400 μL cuvette). Anisotropy (r) was calculated as: where I VV is the fluorescence intensity with vertically oriented excitation and emission polarizers and I VH is the fluorescence intensity with a vertically oriented excitation polarizer and a horizontally oriented emission polarizer. The G-factor, defined as G=I HV /I HH , was measured before each experiment to ensure a value of ~ 1.65. Anisotropy values were collected in triplicate using an integration time of 2.0 s. K d values were calculated by fitting curves to a standard single-site binding equation using Origin (OriginLab).

Anisotropy calculations
For all theoretical calculations, r was calculated using the Perrin equation: where r o is the fundamental anisotropy,  the lifetime and  the rotational correlation time (7).
For fluorescein, empirically determined values of r o = 0.39 ns andnswere used (8,9).was calculated using the equation: where  is the viscosity of water in poise (0.890 cP for all calculations), V is the volume of the molecular complex, R the gas constant and T temperature (298.15 K for all calculations). For calculations of hydrodynamic radius, the radius was derived directly from the volume term assuming a spherical RT-T/P complex. To determine  by molecular weight (MW), V was further defined as : where MW is the molecular weight of the RT-T/P complex (130 kD),  the specific volume

Single-molecule TIRF microscopy
Single-molecule TIRF microscopy was performed as described previously (10). We used an inverted fluorescence microscope (Olympus IX71) that was modified for prism-based TIRF and coupled to a 532 nm diode laser (CrystaLaser) and a 647 nm diode laser (Melles Griot). Fused silica slides (G. Finkenbeiner, Inc.) were surface-passivated with PEG and biotin-PEG (Laysan Bio., Inc.), as described previously (11). Fluorescence from individual tethered complexes was sent through a 550 longpass filter and, using a DualView apparatus (Photometrics), was split into Cy3 or Cy5 emission pathways by a 610 nm dichroic mirror, respectively filtered by either a 580/40 nm bandpass filter or a 660 nm longpass filter (Chroma Technology Corp.), and subsequently imaged onto two halves of an electron-multiplying CCD (EMCCD) camera (Andor Technologies). Signals were identified by intensity thresholding and by goodness of fit of 7 × 7 pixel peaks to a 2D Gaussian. Time-dependent intensity traces were extracted and corrected for local background. All analyses of single-molecule data were performed using custom-written software in MATLAB (The Mathworks, Inc.).
The buffer used in TIRF microscopy experiments also contained 0.1 mg/ml glucose oxidase, 0.02 mg/ml catalase, 0.4% wt/v β-d-glucose, and 2 mM Trolox. Unless otherwise indicated, the TTP, EFV and NVP concentrations were held at 50 μM, 500 nM and 1 μM, respectively, ensuring saturating binding conditions for all ligands.

Single pair FRET
Unlabeled T/P was surface tethered by introducing 2 M T/P into the flow cell to allow  Table 1 were derived from the fitted E app values using the Förster equation: where r is the distance between the dipoles of Cy3 and Cy5, and R o is the empirically calculated

Single-molecule PIFE
All PIFE data was acquired on a 512 × 256 pixel region of the EMCCD at 33 Hz. T/P molecules were surface tethered to the PEGylated flow cell via a biotin:streptavidin:biotin-PEG linkage.
The concentrations of T/P was typically 20 pM, which provided an optimal surface density of

Model for the distribution of RT on T/P substrate based on anisotropy data
To account for the observed changes in r for the RT-T/P complex upon binding TTP and/or EFV, we considered a model in which  Figure 1B).

Cy3 parameters used in AMD simulations
The unmodified Cy3 linker dihedral parameters (see Supplementary Methods)  In regard to the role of the salt bridge in K103N mechanism, our FRET data agrees with the crystallographic observation that K103N-EFV retains the E138-K101R salt bridge (24), since the thumb/fingers distance of K103N RT was essentially the same as that of WT RT. Our data additionally indicates a closed thumb/fingers conformation only in the presence of dNTP, indicating that, in some scenarios (such as K103N) dNTP binding can counteract the allosteric effects of NNRTI on the thumb/fingers distance. If our model of EFV resistance by K103N is correct, we would therefore predict that a crystal structure of the K103N RT-T/P-EFV-dNTP complex would show a broken E138-K101 salt bridge. Further, it is interesting that NVP induced less RT mobility than EFV, and the E138D mutation had little effect on NVP susceptibility even though the E138-K101 salt bridge appears in the WT-NVP structure (25). Combined with the observation that RT displays a closed thumb/fingers conformation in the presence of both NVP and dNTP, suggesting that NVP cannot bind the ternary RT-T/P-dNTP complex (consistent with recent ITC data (26)), we hypothesize that the degree to which NVP forces the E138-K101 salt bridge is less than that of EFV, accounting for its relatively lower NNRTI activity, perhaps owed in part to its molecular structure in the binding pocket which is unable to hold E138 in place for optimal salt bridge geometry. Since our data indicate an allosteric 'tug-of-war' between NNRTI and dNTPs which directs the thumb/fingers grip of RT, even a small increase in electrostatic energy in the salt bridge could account for an increased tendency to bind dNTP. Together with the observation that K103N allows an EFV-bound enzyme to bind dNTP, we therefore suggest that an NNRTI which affords additional energetic stabilization near the p51/p66 hinge domain (including additional stabilization of the salt bridge and surrounding residues) would potentially inhibit RT with increased potency over current NNRTIs.