Research into second generation DKA inhibitors shortly after the FDA approval of MK-0518 led to the design of a set of tricyclic hydroxypyrroles that mimicked the common DKA metal binding pharmacophore. Optimization of a derived set of 10-hydroxy-7,8-dihydropyrazinopyrrolopyrazine-1,9-dione compounds resulted in one of the first raltegravir me-too leads, MK-2048 (Figure 1). MK-2048 has exhibited an IC₉₅ of 40 nM in the presence of 50% NHS, favorable pharmacokinetics, and potent antiretroviral activity against four IN mutants displaying raltegravir resistance 5152.

Early modification of the DKA motif by Japan Tobacco resulted in the design of a group of 4-quinolone-3-glyoxylic acids 49 that retained the coplanarity of DKA functional groups. A potent compound from this original study contained only a β-ketone functional group and a carboxylic acid functional group, which were coplanar, and showed a 1.6 μM IC₅₀ value in a strand transfer assay. Derivatives of this parent compound exhibited up to a 7.2 nM IC₅₀ value in strand transfer assays and a 0.9 nM EC₅₀ in an antiviral assay. This activity proved that a monoketo motif could be an efficacious alternative to the accepted DKA. A 2005 license agreement between Japan Tobacco and Gilead Sciences led to the clinical development of GS-9137 (a.k.a. elvitegravir) [Figure 1, 43], a quinolone carboxylic acid strand-transfer specific inhibitor that displayed an IC₅₀ of 7 nM against IN and an antiviral EC₉₀ of 1.7 nM in the presence of NHS. In terms of pharmacokinetics (Additional file 1), in rat and dog elvitegravir displayed a 34% and 30% bioavailability, a 2.3 h and 5.2 h half-life, and a 8.3 mL/min/kg and 17 mL/min/kg clearance, respectively. Interestingly though, its half-life in human was shown to increase from 3 hours when dosed alone to 9 hours when boosted with the protease inhibitor, ritonavir 53. Similarly, its bioavailability increased 20-fold when administered in combination with ritonavir. These observations back a valid argument that elvitegravir may become a second-generation IN inhibitor, in that its significantly improved pharmacokinetic profile when boosted may increase patient compliance by allowing a simple once daily treatment (raltegravir is administered twice daily). Similar to raltegravir, though, elvitegravir has been shown to provoke T66I and E92Q viral resistance mutations, as well as substitutions of amino acids flanking raltegravir-induced substitution sites (Q146P and S147G) 54.

In studies to develop follow-on analogs of S-1360, the two involved groups jointly discovered a novel lead naphthyridinone, GSK-364745 (Figure 1). This compound contains a hydrophobic fluorobenzyl substituent flexibly linked to a chelatable quinolone region. GSK-364735 inhibited IN in an in vitro strand transfer assay with an IC₅₀ of 8 nM, and it showed an antiviral EC₉₀ value of 40 nM in MT-4 cells in the presence of 20% NHS. Acceptable pharmacokinetics were achieved, with bioavailabilities of 42%, 12%, and 32%; half-lives of 1.5 h, 1.6 h, and 3.9 h; and clearances of 3.2 mL/min/kg, 8.6 mL/min/kg, and 2 mL/min/kg in rat, dog, and rhesus monkey, respectively (Additional file 1). However, when tested against mutant viruses, the compound exhibited greatly decreased activity – 17-fold reduction against T66K, 210-fold reduction against Q148K, 73-fold reduction against Q148R, and 23-fold reduction against N155S 55.

A pyrimidine carboxamide similar in structure to raltegravir was recently propelled into Phase II clinical trials by a separate group. This compound was different from raltegravir in that raltegravir's 1,3,4-oxadiazole group was substituted with a cyclic sulfonamide moiety (Figure 1), but its in vitro potency was similar with an IC₅₀ value of 20 nM. However, multiple mutations were almost immediately observed to have occurred in viral response to treatment with BMS-707035, which included V75I, Q148R, V151I, and G163R 32. Unfortunately, the severity of resistance conferred by each of these mutations has not been disclosed, nor have pharmacokinetic properties of the drug. What is known, however, is that the drug did not last long in Phase II trials, and testing was abruptly terminated in early 2008 56. An explanation of the termination of the trial has not been publicly provided.

Soon after promising clinical data regarding the progress of MK-0518 became available, a novel DKA-related class of IN inhibitory compounds (Figure 3, Additional file 1) was developed through screening of inhibitors of HCV polymerase, which demonstrates a high degree of structural similarity to IN 31. Specifically, IN and HCV polymerase possess a similar active site amino acid geometry, and both utilize two magnesium ions in their catalysis. A class of dihydroxypyrimidine carboxamides was derived as HCV polymerase inhibitors from DKAs, and they were found to exhibit improved drug-like properties and correct Mg²⁺ binding geometry. Most of these compounds were inactive against IN, but a substitution of the free carboxylic acid with a benzyl amide yielded compound 1, with nanomolar IN inhibitory activity in enzymatic assays. Compound 1 showed a decent pharmacokinetic profile, with a bioavailability of 15%, plasma clearance of 5 mL/min/kg, and a half-life of 3 hours. Further structure activity relationship (SAR) studies upon the amide moiety of 1 led to the identification of a superior para-fluorobenzyl substituent (compound 2). Compound 2 exhibited an IC₅₀ of 10 nM in the enzymatic assay, as well as an improved oral bioavailability in rats of 29%. However, both compounds 1 and 2 were inactive in cell-based assays, due to poor solubility, poor cell permeability, and significant plasma protein binding 31.

This group pushed on in their search for raltegravir me-too drugs with further SAR studies upon the above N-alkyl hydroxypyrimidinone lead compounds (Figure 3). As a benzyl amide substitution of a free carboxyl instilled nanomolar activity upon said compounds, a library of over 200 different amide modifications was synthesized and screened for inhibitory potency 57. A 4-fluoro-substituted benzene was shown to be optimal for IN inhibition, with an IC₅₀ value in enzymatic assays of 10 nM. However, though compounds optimized in this fashion were active in the enzymatic assay, they lacked potency in cell based assays. The thiophene ring in the 2-position of the pyrimidine core was shown to have little effect upon the interaction of the compound with IN, and so this position was chosen for more dramatic changes influencing physiochemical properties of inhibitors. Introduction of a basic group to a 2-benzyl derivative resulted in increased cell permeability and inhibition of viral replication in the presence of fetal bovine serum (FBS) with a CIC₉₅ of 300 nM (compound 3). This compound showed an oral bioavailability of 59% and 93%, a half-life of 1.73 h and 6.78 h, and a plasma clearance of 14 mL/min/kg and 0.5 mL/min/kg in rats and dogs, respectively. However, weak activity in the presence of 50% NHS exposed the mobile nature of chosen 2-position substituents. In response the phenyl group at this position was removed and the NH methylated, to confer reduced lipophilicity (and reduced plasma protein binding) but maintain the presence of the mandatory amino group. Compound 4 was thus born, exhibiting a 95% human plasma protein binding and a 400 nM CIC₉₅ in the presence of 50% NHS. Pharmacokinetics of compound 4 included an oral bioavailability of 27% and 90%, a half-life of 0.43 h and 6.0 h, and a plasma clearance of 75 mL/min/kg and 2 mL/min/kg in rats and dogs, respectively. Separately, smaller acyclic amines were substituted into the 2 position and similarly assayed for activity 57. It was found that a dimethylaminomethyl substituent separated by an sp³-carbon spacer bestowed significant cell based potency, at a CIC₉₅ of 78 nM in 50% NHS (compound 5). In rats, dogs, and monkeys, compound 5 had a prolonged plasma half-life (2.1, 4.8, and 1.9 h, respectively), moderate to low clearance (16, 1.9, and 15 mL/min/kg, respectively) and moderate to excellent oral bioavailability (28%, 100%, and 61%, respectively) 57.

To improve cell-based potency and bioavailability of the above molecules, this group began to study the effect of methylation of their N-1 pyrimidine nitrogens (Figure 4, Additional file 1). The rationale for this decision was based upon their discovery that the amine contained in the ring must occupy the benzylic position with respect to the pyrimidine and that small alkyl groups are preferred on the nitrogen of the saturated heterocycle 57. A methyl group was initially scanned on the pyrrolidine ring, and substitution on position 4 gave the best enzymatic activity. Substitution of the free hydroxyl group of a resulting trans-4-hydroxy pyrrolidine with a methoxy substituent produced potent activity (compound 6) in both in vitro (IC₅₀ = 180 nM) and cell-based assays (CIC₉₅ = 170 nM in 50% NHS) 58. From here the group tested other substitutions, of which a fluorine (compound 7 – CIC₉₅ = 250 nM) or a difluoro derivative (compound 8 – CIC₉₅ = 170 nM) were well accepted. Activity was found to be further augmented by substituting a six-membered derivative in position 2 of the pyrimidine, and the morpholine derivative 9 and piperidine derivative 10 displayed slightly improved cell-based potencies (100 nM and 190 nM CIC₉₅ in 50% NHS, respectively). In terms of pharmacokinetics, the morpholine derivative 9 was the most ideal candidate for further testing, with bioavailabilities of 92%, 100%, and 53%; half-lives of 1.5 h, 10 h, and 1.4 h; and plasma clearance rates of 22 mL/min/kg, 3 mL/min/kg, and 14 mL/min/kg in rat, dog, and rhesus monkey, respectively 58.

A further optimization study analyzed the enzymatic and pharmacokinetic implications of a different, ^tbutyl substitution at the C-2 position of the pyrimidine scaffold of the above compounds [Figure 4, 59]. Further introduction of a benzylamide to the right side of the scaffold proved necessary for activity in serum conditions. Multiple derivatives were designed using the N-methyl pyrimidone scaffold, including a sulfone (compound 11) and an N-methyl amide (compound 12) that showed CIC₉₅s of 20 nM and 10 nM in 50% NHS, respectively. This encouraging data inspired further substitutions of the 2-N-methyl carboxamide, for optimization of pharmacokinetic behavior. An unsubstituted amide 13 exhibited a promising inhibitory profile (IC₅₀ = 20 nM in enzymatic assay, CIC₉₅ = 10 nM in 50% NHS), prompting multiple further substitutions of the N-methyl residue with an N-ethyl (compound 14) and an ⁱN-propyl (compound 15). The pharmacokinetic profiles of 11, 12, and 13 were not optimal (Additional file 1), and none of these substitutions were beneficial in this respect. Bioavailability was 17%, 18%, and 23%; half-life was 1.8 h, 1.6 h, and 3.6 h; and plasma clearance was 37 mL/min/kg, 24 mL/min/kg, and 55 mL/min/kg in rat for 11, 12, and 13, respectively 59.

Parallel to the above N-methylpyrimidone studies, the same group was working toward optimization and cyclic constraint of the dihydroxypyrimidine-4-carboxamide amide side chain, yielding a novel class of dihydroxypyridopyrazine-1,6-dione compounds [Figure 5, 60]. Coplanarity of the amide carbonyl group in the constrained ring with respect to the dihydroxypyridinone core and a resulting limitation of flexibility of the 4-fluorobenzyl side chain (compound 16) were shown through molecular modeling to be essential for inhibitory activity. Compound 16 inhibited IN strand transfer in vitro at an IC₅₀ of 100 nM and HIV replication in cell culture at a CIC₉₅ of 310 nM, with little cytotoxicity. Limited pharmacokinetic data has been provided for this class of compounds, but compound 16 was shown to have a 69% oral bioavailability in rats, and plasma concentrations were maintained between 0.64 and 0.50 μM from the second to the twenty-fourth hour (Additional file 1). There was concern about the dihydroxypyrimidone core and its metabolites irreversibly associating with liver microsomal proteins, but only a non-significant level (<50 pmol equiv/mg/60 min) of interaction was observed 60.

Recently, the aforementioned importance of a β-amino substituent in the 2-position of the pyrimidine scaffold and the beneficial effect of the 1N-methylation were exploited in a systematic constraint of the 1N-methyl on the 1N-methylpyrimidinone scaffold (Figure 6, Additional file 1). With unsubstituted benzylmethylamine derivatives showing nanomolar enzymatic inhibition profiles similar to those of derivatives with saturated ring side chains (though little inhibition of viral replication in cell culture), it was decided that the 2-β-nitrogen would be modified to optimize physiochemical properties of pyrimidone compounds 61. For example, introduction of a sulfonamide (compound 17) resulted in a low shift in activity in serum conditions, suggesting an increased level of cell permeability. The (R)-17 enantiomer displayed a 7 nM enzymatic IC₅₀ value, a 31 nM CIC₉₅ in 50% NHS (two-fold more potent than its (S)-17 enantiomer contemporary), and acceptable pharmacokinetics including a 17% bioavailability and 55 mL/min/kg plasma clearance in rat. Sulfonamide derivatives showed similarly decent profiles (compound 18 = 12 nM IC₅₀ against strand transfer, 86 nM CIC₉₅ in cells in 50% NHS, and a 47% bioavailability and 48 mL/min/kg plasma clearance in rats). However, an even more significant improvement in potency occurred upon changing the sulfonamide moiety to a tetrasubstituted sulfamide (compound 19). The (R)-19 enantiomer inhibited IN with an IC₅₀ value and a CIC₉₅ value of 7 nM and 44 nM, respectively, but pharmacokinetics (9% bioavailability in rhesus monkey) were inadequate. Introduction of a more polar N-methylpiperazine (compound 20), however, produced a compound whose (S)-20 enantiomer inhibited IN at a CIC₉₅ of 6 nM in cell culture in the presence of 50% NHS. This compound was much more stable toward glucuronidation than its sulfamide counterpart, but low bioavailability and high plasma clearance in rats and dogs neutralized its promise. It was hence necessary to make use of other nitrogen functionalizations in order to optimize these pharmacological properties. The substitution of ketoamides and enlarged rings (compounds 21 and 22, respectively) resulted in potent inhibition of IN in cell based assays and much improved pharmacokinetics. The (S)-enantiomers of both compounds achieved CIC₉₅s of 43 nM and 13 nM in cell culture, respectively, as well as moderate pharmacologic properties in rats, dogs, and (compound (S)-22 only) monkeys 61.

A different group has recently built upon their prior optimization of the clinically efficacious L870,810 6263 by varying C5 substituents within their compounds' tricyclic scaffolds (Figure 7, Additional file 1). They originally developed the tricyclic scaffold to provide a pre-organized, energetic improvement to L870,810's unfavorable energy consumption upon rotational conversion from free state to bound state, leading to a more soluble and potent compound 23 62. In their recent work, C5-amino derivatives were prepared and assayed for improvement in strand transfer inhibitory potency and pharmacokinetics, due to their projected higher stability against hydrolysis than analogous carbamates or sulfamates 64. The most promising leads turned out to be a C5 sulfonamide (compound 24), a C5 sulfonylurea (compound 25), and a C5 sultam (compound 26). Compounds 24 and 25 retained potency in the presence of serum albumin and α-1 acidic glycoproteins, while 26 was negatively affected. Though the sultam 26 showed a lower IC₅₀ than the sulfonamide 24 and sulfonylurea 25 in enzymatic assays (13 nM as opposed to 28 nM and 62 nM, respectively), it lacked potency in cell culture in 50% NHS (EC₅₀ 49 nM as opposed to 11.4 nM and 8.4 nM, respectively). It is important to note that raltegravir showed an EC₅₀ value of 16 nM in cell culture in the presence of 50% NHS. Compound 26 was additionally lacking in bioavailability in both rat (4%) and dog (8%). However, compounds 24 and 25 showed slightly more promising profiles, with bioavailabilities of 15%/13% and 45%/16% and half-lives of 1.1 h/0.9 h and 4.9 h/4.5 h in rat and dog, respectively 64. This study exemplified the importance of rigidifying inhibitor pharmacophores in terms of conferring favorable potency and pharmacokinetic properties.

Though there is minor variation in the in vitro activity of the above me-too IN inhibitors, their structures, mechanisms of action, and pharmacokinetics are highly similar. We believe that the development of me-too compounds may yield a relatively low amount of clinical success due to their similarities, and also due to the fact that nearly identical resistance profiles will be evoked by their application. However, we would like to note that it is definitely possible for a raltegravir me-too analog to evolve into a second-generation IN inhibitor. To further elucidate our viewpoint, we utilized the molecular docking program GOLD version 3.2 to conduct a docking study, using both the X-ray determined structure of 1BL3 IN complexed with an Mg²⁺ ion, and a collection of significant, above-described me-too compounds (Figure 8); for a detailed procedure, see 65. We propose that residues essential to the compounds' interaction with IN will obviously be prime candidates for resistance mutation. Furthermore, we hypothesize that the test of time will show that all of these me-too inhibitors will probably exhibit highly similar resistance profiles. As raltegravir has undergone extensive resistance profiling since the inception of its clinical employment (Table 1), we first compared our predicted interaction residues (Figure 8) to these experimental profiles, as a validation of the reliability of our technique. We found that five of our predicted interaction residues (T66, E92, Y143, Q148, and N155) have been already observed to confer a range of anywhere from 5- to 35-fold resistances to raltegravir inhibition of viral replication, respectively 66676869. We also saw that raltegravir makes direct interactions with the three residues encompassing the IN catalytic DDE motif (D64, D116, and E152), including a hydrogen bond with the glutamate. With this technique corroboration in hand, we decided to similarly predict the interaction residues of raltegravir's progenitors and a few me-too analogues, in order to provide evidence for our assertion that these compounds will ultimately experience a low probability of success in viral eradication, due to their generation of identical resistance profiles. As S-1360 was the first clinical IN inhibitor candidate, we thought it would be interesting to evaluate the similarity between its predicted interaction profile with 1BL3 (Figure 8) and that of raltegravir. We found that an identical interaction occurs between the two drugs and IN (D64, T66, D116, Y143, Q148, E152, and N155), but predicted an additional interaction of raltegravir with E92. This observation has been verified in clinical experimental resistance profiling, as mutation of E92 has not been observed for S-1360, but the E92Q mutation has conferred up to a 7-fold viral resistance to raltegravir 252670. We next observed the interaction profile of 1BL3 with L870,810 (Figure 8), as this is the naphthyridine carboxamide compound that directly led to the development of pyrimidinone carboxamides. We found that L870,810 and raltegravir similarly interacted with D64, T66, D116, Q148, E152, and N155. However, we saw here that only raltegravir interacted with E92. Though this residue has been observed to be mutated to a glutamine in response to L870,810 treatment, the mutation has conferred at most only a 2-fold resistance to the drug, while the same mutation confers up to a 7-fold resistance to raltegravir (Table 1) 2971. The fact that we did not observe a significant interaction between L870,810 and E92 in our docking study further confirms the relatively decreased importance of this residue in viral resistance to the compound. Along the same lines, we did see an interaction of L870,810 with V151, an interaction that was not present in our docking of raltegravir. In clinical experimental resistance profiling, the V151I mutation has been observed to confer up to an 18-fold resistance to L870,810, while the same mutation had a negligible effect on viral resistance to raltegravir (Table 1) 2971. The highly homologous naphthyridine carboxamide candidate, L870,812, has shown an interaction profile virtually identical to that of L870,810 in our docking study, and experimental resistances obtained in clinical observation have been identical as well 2971. As elvitegravir (GS-9137) and GSK-364735 have already been shown to exhibit near identical resistance profiles to raltegravir (Table 1) 67717273, we next used our docking technique to attempt to effectively predict these interactions (Figure 8). For GSK-364735, we were able to predict the interaction with IN residues Y143 and Q148, as well as the three members of the DDE motif. We then predicted that, similar to raltegravir, elvitegravir interacts with T66, E92, Y143, Q148, and the D116 and E152 of the DDE motif. We also saw that elvitegravir interacts with G140, and the G140S mutation has been shown to be associated with a 4-fold viral resistance to the drug, while the same mutation confers only a 1.6-fold resistance to raltegravir (Table 1). Again, the fact that we did not observe a significant interaction between raltegravir and G140 in our docking study further confirms the relatively decreased importance of this residue in viral resistance to raltegravir, but rather its nature of compensation for more meaningful mutations, such as Q148H.