Several different rounds of docking experiments were performed, in which the charges on the DDE + 2 Mg motif, the charge on the central oxyanion of raltegravir, and the location and identity of the “steric wall” mimicking the viral cDNA’s cytosine-adenosine (CA) overhang were modified (see Materials and Methods). In all of these different rounds of docking experiments, the same conclusions were obtained: raltegravir displayed both the “primary mode” and the “flipped mode” against only the wild type ensemble of conformations of the catalytic core domain. The primary mode was much less accessible in the G140S/Q148H mutant’s ensemble of conformations. When the primary mode can be formed against this mutant, raltegravir seems less able to interfere with the putative catalytic role of H67, since the side-chain of H67 has flipped in the mutant, which caused its NH atom to be much farther away from the oxygen atoms of Raltegravir.
In the active site of integrase, either Y143 or H67 have the required properties to activate a catalytic water capable of hydrolyzing the phosphate backbone of DNA bound in the active site. In patients failing therapy with either raltegravir or elvitegravir, the mutations Y143R/C/H have been detected, but no mutations of H67 have been observed in patients.36,37
In addition, mutant integrase enzymes with Y143F are competent at replication, and mutants with Y143N are viable but have a delayed replication phenotype.38,39
The Y143G mutant is also infective, but H67E and H67Q/K71E are not infective, with the latter double mutant being completely defective.34
Although Y143 has been shown to cross-link to DNA, Y143 does not play a significant role in the ability of HIV to replicate.34
In an in vitro
study using site-directed mutagenesis, Y143F was more than twice as active as the H67F mutant.12
However, that assay only detected ligation of target and donor DNA. Since gene expression was not part of the assay, it did not differentiate between productive and defective integration events, which can reduce the clinical relevance of that result. In another study the H67S mutant displayed integrase activity similar to a F185K “wild type model,” but this assay was performed with manganese instead of magnesium, which is known to significantly affect the sequence specificity of the interactions between HIV integrase and the viral cDNA, at least.40,41
In addition, that result may well be a consequence of serine’s ability to act as a nucleophile in a manner similar to Y or H. Since the Y143R/G/F mutants of integrase are viable and infective, and since no mutants of H67 have yet been encountered in patients, the sum of these data suggests that H67 is more likely to play a catalytic role than Y143. This notion underscores the significance of the presented observations regarding the dynamic display patterns of H67 and the ability of specific rotamers of H67 to interact strongly with raltegravir.
While the predicted binding modes of raltegravir presented herein are consistent with the main SAR trend governing the potency of advanced HIV integrase inhibitors,1,5
they are significantly different than the binding modes predicted in a previously-published model by Chen and co-workers.12
This difference may be a consequence of their published model containing improper bidentate interactions between D116 and a Mg.12
In addition, when generating the coordinates for the 140s loop that were missing in their starting crystal structure, they used a loop-building tool which constructed a model with an open conformation of the 140s loop. A bacterial transposase:DNA complex was then used as the source for the position of the DNA in their HIV integrase complex. In Chen’s model the backbone of the integrase and the entire DNA molecule were treated as rigid during the initial energy minimization calculations, which could have trapped the system in an artificial energy well. This led to a fixed open conformation of the 140s loop, when the closed conformation is more likely to be the active, DNA-bound conformation. In our approach we spliced in the coordinates of the closed 140s loop from another crystal structure of HIV integrase when we created our models. MD simulations were then used to generate many different open and closed conformations of the 140s loop, which were included in our docking studies against targets that all displayed the proper coordination geometry between the DDE motif and the two magnesium ions. The aforementioned flaws in the approach described by Chen et al
. may explain their surprising conclusion that HIV integrase inhibitors only interact strongly with a single magnesium ion in the active site,12
which is at odds with the widely-known SAR trends discussed previously.
In our presented models, the wild type system displayed oscillations between open and closed states of the 140s loop throughout the entire MD simulation. The E92Q/N155H mutant’s MD exhibited a higher amplitude and frequency of oscillations between open and closed states. The G140S/Q148H mutant’s MD showed more restricted motion around a distorted, closed conformation of the 140s loop. However, these observed differences in dynamic behavior should be validated with NMR or other experimental techniques. These differences in conformational preferences and dynamic flexibility displayed by the 140s loop, combined with the significant differences in the dynamic display pattern of the critical H67 residue, contribute to the fact that the G140S/Q148H mutant’s ensemble contained many fewer conformations against which raltegravir could dock well, relative to the wild type. The G140S/Q148H mutant’s ensemble of snapshots was much less accessible to the predicted primary binding mode of raltegravir, and the flipped binding mode was never observed against this mutant. The trend in accessibility indicates that “kinetic gating” could contribute to the drug-resistance profile of the G140S/Q148H mutant. In addition, if raltegravir induces any significant structural changes in the catalytic domain to achieve its high affinity and inhibitory activity, then its binding would likely pay a larger enthalpic penalty to induce such changes in this more rigid G140S/Q148H mutant. Thus, the results presented indicate that kinetic gating and/or induced fit effects are plausible mechanisms for raltegravir resistance of the G140S/Q148H mutant.
Should this hypothesis of the mechanisms of drug-resistance for the G140S/Q148H mutant be correct, then the following strategy would be useful in guiding the design and evaluation of integrase inhibitors with resistance profiles superior to raltegravir: create fairly rigid compounds with structures that are pre-optimized to interact well with the closed conformations of this double mutant and the wild type integrase. Differences in the dynamic display pattern of His67 must also be taken into account when optimizing inhibitors against this mutant.
The single mutation N155H is a primary/signature mutation that confers raltegravir-resistance in the clinic.6
E92Q is linked with N155H to make a double mutant that is far more raltegravir-resistant than either single mutant.21
In the primary binding mode of raltegravir against the wild type catalytic domain, the fluoro-benzyl group of raltegravir forms a favorable electrostatic interaction with N155. This binding mode has a much more favorable estimated free energy of binding than the “flipped” mode, which interacts well with E92. The fact that the primary mode interacts well with N155 and displays a better binding energy than the flipped mode is in good agreement with the known trends in resistance profiles for the N155H and E92Q single mutants. Additional docking studies need to be performed before predicting raltegravir’s binding mode against this double mutant. But the two binding modes that raltegravir is predicted to display against the wild type appear to explain why the E92Q/N155H double mutant is highly raltegravir-resistant.
If the current preliminary hypothesis of the mechanism of drug resistance for the E92Q/N155H mutant is correct, then a very different strategy should be useful when designing inhibitors with enhanced efficacy against this double mutant. Finding a new class of inhibitors that prevents this mutant from sampling the active conformations of the 140s loop could be quite useful. To defeat a mechanism of drug resistance that involves enhanced flexibility of the critical 140s loop and changes to the surface structure of the active site that affect both binding modes that raltegravir displayed against the wild type, future studies will also involve searching for an allosteric binding site where inhibitors can potentially stabilize the inactive conformations of this critical loop near the active site.