The focus of the WRAIR discovery program with next generation quinoline methanols has been to reduce partitioning into the CNS since the relevant clinical target or targets of mefloquine adverse events are not known . The incidence of the most common adverse CNS effects of mefloquine is reduced substantially when the dose is lowered five-fold from the treatment to the weekly prophylaxis level. Presumably this is also associated with a five-fold decline in CNS levels as well. Consequently, at least a five-fold reduction in whole and free brain levels relative to mefloquine at the efficacious dose is probably the minimum requirement. In this study we have shown that at least a ten-fold reduction in both parameters is feasible through replacement of the piperidine ring of mefloquine with various other four-position substituents. The prospect of such a compound being better tolerated than mefloquine is, therefore, a possibility.
While a reduction in maximum brain concentration should reduce the incidence of CNS associated adverse events, optimizing this property alone will not be sufficient for a new quinoline methanol anti-malarial. The putative late lead compound must also have a sufficiently long half-life and intrinsic efficacy to be useful if administered as a single dose. Therefore, the goal is to balance these three potentially competing pharmacological properties in the same molecule. There is an opportunity to evaluate a subset of the larger 4-position library using surrogates of these endpoints. Maximum free and whole brain concentrations represent surrogates of CNS safety if they are sufficiently reduced relative to mefloquine at the therapeutic dose. Whilst the plasma concentration-time data generated are insufficient to calculate pharmacokinetic parameters directly in most instances, the proportion of drug remaining at 24 h relative to the 5 minute time point is probably a reasonable surrogate of half-life, or at least persistence in plasma. In vitro activity (IC90) can be utilized as a surrogate for intrinsic activity.
There are no compounds amongst this sub-library that meet all the requirements in terms of potency, low brain levels and residual plasma concentrations at 24 h (see Additional file 1) suggesting that further optimization is required. The ideal chemotype around which to anchor a future lead optimization program would be one in which a balance between all these three properties is achievable. The most active compounds were generally those with aliphatic, ether or thio ether side chains (WR308245, WR308246, WR308257, WR308265, WR177000, WR308387, WR308278 and WR308412). However, these compounds exhibited much higher brain concentrations (free and whole) than mefloquine, and in most instances, 24 h residual plasma concentrations were < 1% of those at 5 min, indicating a likely short half-life relative to mefloquine. The exception to this general trend was the diamine WR308396. As might be expected, all compounds with lower whole and free brain concentrations than mefloquine had a greater number of H-bond donors and/or acceptors than mefloquine. These compounds comprised a diverse array of structures including alcohols, acids, triazines, substituted cyclic imidazoles and diamines. However, with the exceptions of the diamines WR319581 and WR318746, these compounds exhibited a lack of potency (IC90 > 300 ng/ml) or low residual plasma concentrations (< 1% at 24 h versus 5 min). Of the 25 compounds that were not mefloquine or its enantiomers, only nine had residual plasma concentrations > 1% at 24 h, and five of these were diamines. Consequently, a balance between these properties may be most likely obtained amongst diamine quinoline methanols.
Identifying a new lead compound with the requisite balance of potency, brain partitioning and half-life might not be feasible if it was true that positive trends in these characteristics were mutually exclusive, however this does not appear to be the case. Maximum free brain concentrations, IC90s, and residual plasma concentrations amongst quinoline methanols did not appear to be correlated (Figure 3A-C). Therefore, at the outset there is no reason to suspect that a balance between these properties is not achievable in a diamine quinoline methanol. There are no generic algorithms that can be used to objectively judge the probability of success of a specific lead optimization programme a priori. Thus one must extrapolate on a project-by-project basis with the available data. The lack of a general correlation between maximum free brain concentrations, residual plasma concentrations and IC90s suggests that their probability distributions may be independent. Thus, we estimated the probability of a diamine quinoline methanol possessing all these properties to be around 4%. This value was obtained by multiplying the marginal probabilities of success for the individual parameters of interest (Figure 5). The confidence intervals are wide (0-21%) which is not surprising given the small sample sizes.
The probability estimate discussed above was based on free brain concentrations being representative of CNS exposure. We have not made the same calculations for maximum whole brain concentrations, because the assumption of independence between this variable and IC90 may arguably be less justifiable. However, two lines of evidence support the notion that a compound with an acceptable maximum free brain concentrations will also have an acceptable maximum whole brain concentrations. First, whole and free brain concentrations in the broader group of quinoline methanols were strongly correlated. This suggests that they are also not independent, or put another way, that there is a strong conditional probability whole brain concentrations will be lower than mefloquine if this is the case for free brain concentrations. Second, in the same broader group of quinoline methanols it is also true that the proportion of compounds exhibiting the desired whole brain concentrations (28%) was higher than the proportion of compounds exhibiting the desired free brain concentrations (8%). Additional studies are required to determine whether these observations will hold for a larger sample of diamine quinoline methanols.
It is appropriate to execute a lead optimization programme if it can be done with a reasonable probability of success in a time frame of up to two years . For planning purposes, it has been assumed the true probability of success for an individual compound lies somewhere between the lower confidence interval and 4%. Since the synthetic route is amenable to the synthesis of 100 new compounds in 12-18 months, it is anticipated that up to four potential late lead molecules will be identified in that time frame. The assumed low probability of success with any individual compound requires an aggressive method of early triage. The permeability of a compound across MDCK cell monolayers appears to correctly categorize it relative to mefloquine in terms of its maximum brain concentration in most instances. It is, therefore, reasonable to use this assay together with more routine P. falciparum susceptibility assays to rapidly prioritize compounds for in vivo studies. It may be necessary to conduct in vivo studies on a selection of active, but permeable compounds triaged using this technique in order to control the type II error rate. A lead optimization campaign conducted broadly along these lines is now underway.