Strain-specific Effect on Biphasic DNA Binding by HIV-1 Integrase

Abstract

The oligomerization of HIV-1 integrase onto DNA is not well understood. Here we show that HIV-1 integrase binds the DNA in a biphasic (high- and low-affinity) modes. For HIV-1 subtype B, the high-affinity mode is ~100-fold greater than low-affinity mode (K_d.DNA=37 and 3400 nM, respectively). The Kd.DNA values of patient-derived integrases containing subtype-specific polymorphisms were affected 2–4–fold, suggesting that polymorphisms may influence on effective-concentrations of inhibitors, since these inhibitors preferably bind to integrase-DNA complex.

Subtype-specific differences in sensitivity and the impact of naturally occurring polymorphisms (PMs) are major emerging threats to the success of combination antiretroviral therapy (cART). In support of such subtype-specific differences, we recently reported that susceptibility to the non-nucleoside reverse transcriptase inhibitors rilpivirine and etravirine varies among subtypes [1]. Our real-life study also indicated subtype-specific differences in first- and second-line treatment failure with protease inhibitors [2]. Furthermore, our study of a South African cohort identified M50I as a naturally occurring PM [3], which was the first mutation to emerge under bictegravir pressure [4]. In an Ethiopian cohort, we found primary integrase strand transfer inhibitor (INSTI) mutations in minor viral variants of untreated patients [5]. Collectively, these and several published reports indicate subtype-dependent responses to cART and suggest that natural PMs can impact treatment outcome [6–9].

Integrase strand transfer inhibitors (INSTIs) are the latest addition to the cART armamentarium. Following World Health Organization (WHO) recommendations, the INSTI dolutegravir (DTG) has been rolled out on a large scale in low- to middle-income countries (LMICs), which predominately have non-B HIV-1.

INSTIs preferably bind to the IN/DNA complex [10], suggesting that INSTI efficacy may depend upon DNA binding to IN. In the absence of structures for HIV-1 IN/DNA/INSTI complexes, the structures of prototype foamy virus IN/DNA/INSTI complexes [11–13] have served as guides to infer the mechanisms of INSTI inhibition/resistance. DNA binding by HIV-1 IN has been investigated by covalent crosslinking [14–16], nuclease protection [16], and fluorescence-based assays [17]. These reports conclusively revealed oligomeric assembly of HIV-1 IN onto the substrate DNA. Some studies [14–16] also show cooperative binding of two IN monomers to DNA, and a biphasic dissociation was noted when preassembled IN/DNA complex was challenged with ssDNA [16]. In this indirect approach, the two DNA binding states were referred to as (i) ssDNA-resistant and (ii) an unstable ternary (integrase−LTR−polynucleotide, IN·S·C) complex state. While these and other [18–20] reports have demonstrated multimeric assembly of IN molecules onto the DNA substrate, the sequence of events in IN/DNA complex formation remains unknown. Here, we present the first direct measurement showing a biphasic DNA binding by HIV-1 IN. Our data show that HIV-1 IN binds DNA first in a high-affinity mode followed by a low-affinity mode, and that this binding depends upon a specific combination of PMs in the context of other substitutions.

Microscale thermophoresis (MST) is a highly sensitive biophysical technique for characterizing intermolecular interactions; it is based on directed molecular movement in a temperature gradient [21]. We used MST to determine the DNA binding affinity (K_d.DNA) of IN derived from laboratory strain pNL4–3 (representing HIV-1B). Normalized thermophoresis time traces with increasing IN concentrations (0.007–13,500 nM) and fixed 5´-Cy5-labeled DNA (20 nM) are shown in Fig.1a. The binding isotherm obtained by plotting the difference in normalized fluorescence against increasing IN concentration (Fig.1b) showed the binding of IN to DNA to be a biphasic phenomenon. Fitting the data points to a biphasic model yielded K_d.DNA values of 37±2 nM and 3400±500 nM for the high- and low-affinity binding modes of HIV-1B IN, respectively. To test whether the low-affinity binding is associated with anomalous binding to the 13nt ssDNA (single-stranded) region of the 31/18-mer DNA used here, we repeated MST assays with a DNA substrate containing 19-nucleotides from the 5’-UTR of HIV-1C annealed to a 21-mer or 24-mer DNA oligonucleotides, leaving only 2- or 5-nucleotides ssDNA 5’-overhangs, respectively. The data exhibited similar K_d.DNA values for each of the two phases (data not shown), ruling out the possibility that anomalous IN binding to ssDNA is responsible for the observed low-affinity mode.

Figure 1: Biphasic modes of DNA binding by HIV-1 IN and the effect of natural polymorphism M50I on INSTI efficacy.

a) Example time traces of a typical MST assay containing 20 nM 5’-end-Cy5-leveled DNA with increasing concentrations of HIV-1 IN. The change in thermophoresis is expressed as the change in the normalized fluorescence (ΔF_norm= F_hot/F_cold), where F_hot and F_cold correspond to average fluorescence values between defined areas marked by red and blue bars, respectively [25]. All traces are normalized to the same initial relative fluorescence value of 1. All MST assays reported here were conducted in a buffer containing 50 mM Tris-HCl, pH 7.8, 100 mM KCl and 0.1% pluronic F-127. Many different buffer compositions have been used to study biochemical properties of HIV-1 IN suggesting the versatility of enzyme with respect to conditions [14, 26–32]. Here we show the MST data under assay conditions that did not show IN aggregation. (b-e) Final binding isotherm obtained by plotting the difference in normalized fluorescence against increasing concentration of HXB2-IN (b), HIV-1C IN_L101I (c), and HIV-1C IN_M50I/L101I (d) and (e) HXB2-IN with in presence of 10 mM MgCl2 and 21/19-mer (mimicking a substrate after 3’-end processing) DNA substrate. To obtain K_d.DNA values in high- and low-affinity modes, we analyzed MST data independently with three programs: OriginLab (version 18; OriginLab Corp. Northampton, MA, USA), Prism (version 6.0; GraphPad Inc. La Jolla, CA, USA) and MO Affinity Analysis (version 2.2.7, NanoTemper Technologies Inc, Cambridge, MA, USA). We wrote individual equations (available upon request) for fitting the data points in high-affinity, high-affinity and biphasic modes for OriginLab (version 18; OriginLab Corp. Northampton, MA, USA), and Prism (version 6.0; GraphPad Inc. La Jolla, CA, USA) using the quadratic equation provided by MO Affinity Analysis software. The K_d.DNA values for high- and low-affinity were obtained by fitting the data points belonging to high- and low-affinity modes by non-linear regression. Since MO Affinity Analysis (version 2.2.7, NanoTemper Technologies Inc, Cambridge, MA, USA) software does not have in-built biphasic equation, individual low- and high-affinity K_d.DNA values were determined by excluding the data points corresponding to complementary modes. The figures presented here were generated by Prism. (f) 3’-endonuclease activities of INs used in this study. (g) Clustering analysis of inhibiting virus replication by 2-fold change of the patient-derived recombinant viruses (HIV-1B, n = 6 and HIV-1C, n = 14)) relative to the control virus NL4–3 by INSTIs raltegravir (RAL), elvitegravir (EVG), dolutegravir (DTG) and cabotegravir (CAB) as described previously [23].

Very recently, we showed that IN PMs likely mediate IN/DNA or protein/protein interactions [22]. To assess whether subtype-specific PMs affect IN binding to DNA, we cloned and purified INs from viruses isolated from two treatment-naïve HIV-1C infected patients, as recently described [23]. The IN from one isolate contained the L101I PM, whereas the other contained a dual PM M50I/L101I. Our focus on M50I and L101I was based upon (i) the position of M50 in the cryoEM structure of the intasome [24], (ii) the close proximity of L101 to the IN active site, (iii) the presence of the M50I PM in HIV-1C patients from South Africa [3], and (iv) emergence of M50I as the first mutation under bictegravir pressure [4]. The patient-derived INs were then used in MST assays to determine K_d.DNA. Normalized fluorescence differences fit well to a biphasic DNA binding model for HIV-1C_L101I and HIV-1C_M50I/L101I INs (Fig.1c and 1d, respectively). The K_d.DNA values of HIV-1C_L101I in high- and low-affinity binding modes were 167±6 nM and 2160±120 nM, respectively, indicating that the L101I PM decreased DNA binding affinity by ~4.5-fold in high-affinity mode, while moderately increasing affinity (~1.6-fold) in the low-affinity binding mode when compared to HIV-1B IN. The K_d.DNA values of HIV-1C_M50I/L101I IN in high- and low-affinity binding modes were 80±4 nM and 1070±100 nM, respectively, suggesting that the dual PM reduced IN DNA binding affinity in high-affinity mode by ~2-fold and increased it by ~3.2-fold in the low-affinity mode relative to pNL4–3-derived IN.

Notably, the patient-derived (HIV-1C) INs contained additional changes compared to HXB2 IN. To evaluate whether changes in the K_d.DNA of PM-containing INs were the result of the PMs themselves or in the context of other residue differences, we generated M50I and L101I mutations in HXB2 IN. The results indicated that WT, M50I, and L101I had comparable K_d.DNA (ranging between 50 to 100 nM for high- and 3000 to 4000 nM for low-affinity modes), suggesting that PMs M50I and L101I alter DNA binding affinity only in the context of additional substitutions present in patient derived INs. The presence of 10 mM Mg²⁺ did not change K_d.DNA (48±2 nM and 3200±380 nM for high- and low-affinity modes, respectively) of HXB2 IN (Fig.1e). This result is not surprising since no coordination of Mg²⁺ with DNA (PDB file 3L2R, [12]) was noted in pre-integration PFV IN/DNA complex. The structure of the HIV-1 IN/DNA pre-integration complex is yet to be determined. Moreover, a comparison of Mg²⁺ and Mn²⁺ containing DNA-bound PFV IN structures did not show changes in the organization of either the active site or the positioning of the 3′ end of viral DNA, suggesting that DNA binding is independent of divalent cations [12]. All INs demonstrated catalytic activity (Fig.1f).

To assess if the presence of specific PMs affect INSTI response, we performed INSTI drug resistance assays using patient-derived infectious molecular clones of HIV-1B (n=6) and HIV-1C (n=14) from treatment-naïve patients to evaluate subtype specificity and the effect of M50I. A wide variability in EC₅₀ (0.07 nM to 12.74 nM) of four INSTIs was observed. Cluster analyses using EC₅₀ showed one cluster of three M50I (both from HIV-1B and HIV-1C) viruses with two non-M50I viruses, and another cluster with two HIV-1C M50I and one HIV-1B virus (Fig.1g), indicating no subtype-specific differences, and suggesting that specific PMs may exhibit inhibitory effects only when present with other mutations.

In summary, our results suggest that (i) IN/DNA assembly occurs in a biphasic manner, (ii) subtype-specific PMs affect IN DNA binding affinity in the context of additional mutations, and (iii) PMs may contribute to subtype-specific INSTI efficacy, and will be further explored in future analyses.