Cytochrome P450/ABC transporter inhibition simultaneously enhances ivermectin pharmacokinetics in the mammal host and pharmacodynamics in Anopheles gambiae

Mass administration of endectocides, drugs that kill blood-feeding arthropods, has been proposed as a complementary strategy to reduce malaria transmission. Ivermectin is one of the leading candidates given its excellent safety profile. Here we provide proof that the effect of ivermectin can be boosted at two different levels by drugs inhibiting the cytochrome or ABC transporter in the mammal host and the target mosquitoes. Using a mini-pig model, we show that drug-mediated cytochrome P450/ABC transporter inhibition results in a 3-fold increase in the time ivermectin remains above mosquito-killing concentrations. In contrast, P450/ABC transporter induction with rifampicin markedly impaired ivermectin absorption. The same ketoconazole-mediated cytochrome/ABC transporter inhibition also occurs outside the mammal host and enhances the mortality of Anopheles gambiae. This was proven by using the samples from the mini-pig experiments to conduct an ex-vivo synergistic bioassay by membrane-feeding Anopheles mosquitoes. Inhibiting the same cytochrome/xenobiotic pump complex in two different organisms to simultaneously boost the pharmacokinetic and pharmacodynamic activity of a drug is a novel concept that could be applied to other systems. Although the lack of a dose-response effect in the synergistic bioassay warrants further exploration, our study may have broad implications for the control of parasitic and vector-borne diseases.


Results
Design of drug-drug interaction study. The experiments were conducted following a randomised crossover design (Fig. 1). In phase I all mini-pigs received a single 800 mcg/kg dose of oral ivermectin and sampling was done as described below. After a washout period of 14 days, the animals were randomly assigned to pre-treatment with either ketoconazole (200 mg daily, orally, for 14 days), rifampicin (10 mg/kg daily, orally, for 14 days) or nothing in a 1:1:1 ratio and ivermectin dosage and sampling was repeated. The total wash-out period between ivermectin doses was 30 days. Ketoconazole is a dual inhibitor of the CYP and the P-glycoprotein (P-gp), which plays a key role as xenobiotic pump in the blood-brain barrier and other epithelial barriers 23,24 . Inhibiting the P-gp could theoretically lead to toxicity due to abnormal accumulation of the drug in the brain 30,31 . The dose of 200 mg/day was calculated to maximise the time ketoconazole remained above the 50% inhibitory concentration (IC 50 ) for the CYP 3A4 (estimated in 0.022-0.025 µM [1.6-13.2 ng/ml] 47 ) but remaining below the much higher IC 50 for the P-gp (estimated in 5.6 µM [2975 ng/ml] 48 ). Irrespectively of this, drug quantification in cerebrospinal fluid (CSF) was added as an additional safety point. Rifampicin a CYP/P-gp inducer 49 was included as we expected to see reduced bioavailability and penetration into CSF.
The main outcome measure was time above the ivermectin concentration that kills 50% Anopheles gambiae in 10 days [10-day lethal concentration 50 (LC 50 )], which has been estimated at 6 ng/ml 50 . Secondary outcome measures were the maximum concentration reached (Cmax), the total area under the curve (AUC) and the area under the curve above the LC 50 (AUC > LC 50 ).
Design and sample size of mosquito bioassay. Mosquitoes were membrane fed with reconstituted blood as described by Bousema et al. 51 . Feedings were done with samples drawn directly from the pigs at different intervals and pair-matched according to ivermectin concentration, thus the only different between the samples in a pair was the presence or absence of ketoconazole.
Since mosquitoes grouped in cups for feeding, we used a cluster design in which the unit of intervention was mosquito cups. Our primary outcome was mean 10-day mosquito mortality after membrane feeding. Sample size calculations were conducted according to the method described by Gangnon and Kosorok 52 . The minimum expected 10-day mortality in mosquitoes imbibing blood containing 9.9 ng/ml of ivermectin is 66%; this is based on the predictive function described by Ouedraogo et al. 50 . Based on previous observations and with all mosquitoes having the same colony origin and age range, we chose a conservative intracluster correlation of 0.15. The frailty variance of clusters was resembled by a gamma of 1. All experiments were performed in triplicate. With these parameters, at least 4 clusters of 44 mosquitoes were needed per arm to achieve 80% power at 5% significance level accounting for 10% non-feeders.
After determining of the ivermectin levels in all samples in the drug-drug interaction study, we selected 4 paired samples with matching ivermectin concentrations (+/−1 ng/ml) from naïve and ketoconazole pre-treated groups. Baseline serum samples collected before ivermectin treatment in ketoconazole and naïve pigs were used as controls. These samples were shipped frozen to the insectary facilities of the Ifakara Health Institute in Bagamoyo, Tanzania for membrane feeding assays. It was hypothesised that induction of the mosquito CYP/P-gp by rifampicin could improve survival in presence of ivermectin, but this could not be tested due to a reduced absorption of ivermectin in the mini-pigs pre-treated with rifampicin (see below).
Pharmacoenhancement. Ivermectin in both phases and pre-treatments in phase II were administered uneventfully to all subjects. Blood samples were obtained according to the planned schedule. Lumbar punctures were performed with mixed success. It was possible to insert a tunnelled intrathecal catheter and reservoir in 5 out of 6 pigs. Two reservoirs were removed within 48 hours due to motor impairment with full recovery after Figure 1. Illustration of the randomized-crossover design used in the study. In phase I all six pigs received a single 800 mcg/kg dose of oral ivermectin. After a washout period of 14 days, the animals were randomly assigned to pre-treatment with ketoconazole, rifampicin or nothing in a 1:1:1 ratio and ivermectin dosage and sampling was repeated. The time between ivermectin doses was 30 days. Figure by Juliane Chaccour.
SCIentIfIC REPORts | 7: 8535 | DOI:10.1038/s41598-017-08906-x removal. One pig (#72) was euthanised 72 hours post procedure due to non-reversible motor impairment and was replaced by a seventh animal (#75). At least one CSF sampling point was obtained from 5 out of 6 pigs in phase I. The success rate was considerably higher during phase II with CSF samples obtained from 6 out of 6 pigs.
Ketoconazole enhances ivermectin's PK. The main plasma PK parameters in all groups are presented in Table 1. The plasma PK parameters of ivermectin-naïve pigs were compatible with previously published data 53,54 .
Firstly, we compared the PK parameters between ivermectin-naïve pigs and those who received it in phase II without pre-treatment by means of a Mann-Whitney test and found no difference in time above LC 50 , Cmax, AUC or AUC > LC 50 (P values of 0.12, 0.61, 0.16 and 0.32 respectively). Given these results, the small sample size and relative large variations between subjects (as expected in a mini-pig model), these two groups (excluding one animal for which lamda could not be calculated) were combined for comparisons between ivermectin alone and ketoconazole/rifampicin pre-treated pigs. In order to respect sample independence, only the values of the second treatment of mini-pig 73 were used, this penalized the main outcome by increasing the time above LC 50 of the combined ivermectin.
The PK curves of all three groups (ivermectin alone, ketoconazole + ivermectin and rifampicin + ivermectin) are shown in Fig. 2.
Pre-treatment with ketoconazole significantly increased the time above LC 50 3-fold from a median of 36 to 111 hours (P = 0.033). The Cmax was also increased 2-fold (median 14 vs 28 ng/ml, P = 0.033), along with a more than 3-fold increase in the AUC > LC 50 (median 212 vs 711, P = 0.033). The total AUC was also increased more than 2-fold (862 vs 2092 h*ng/ml) but this difference was not statistically significant (P = 0.067).
Pre-treatment with rifampicin seriously hampered ivermectin absorption making the AUC and AUC > LC 50 calculations uninterpretable. Only one pig reached ivermectin plasma concentration above 6 ng/ml, which was maintained for less than 5 hours in total (Fig. 2).

Pre-treatment with ketoconazole did not increase ivermectin in the CSF.
No clinical adverse events were observed. Although detectable in some, ivermectin was below quantification level (0.5 ng/ml) in all CSF samples obtained ( Table 2). Of 24 CSF samples available from phase I, ivermectin was detectable in 4. Of 37 CSF samples available from phase II, ivermectin was detectable in 4 samples (2 in the ketoconazole group and 2 from rifampicin group).

Mosquito bioassay.
Out of a total of 169 samples available we selected four pairs with matching ivermectin concentrations (+/−1 ng/ml) that only differed in the presence of ketoconazole, plus two controls; one from a ketoconazole-only and one from a fully untreated pig (Table 3). In the rifampicin group, there was only one sample with matching ivermectin concentration in naïve pigs.
Ketoconazole is not lethal to mosquitoes at the dose used. Firstly we assessed whether the ketoconazole concentration present in the samples could have a lethal effect per se. For this we compared the 10-day mortality of mosquitoes fed on samples containing only ketoconazole (baseline samples from ketoconazole pre-treated pigs) with the mortality of mosquitoes feeding on fully drug-free samples (baseline of control pigs). Given that all pigs received ivermectin in the first phase, this experiment served as well to assess whether there could be active metabolites increasing mortality in spite of a washing period of 30 days. Ketoconazole alone did not increase the mortality of Anopheles gambiae, in fact it provided a slight, protective effect (Long-rank P = 0.048, n = 222), as previously seen with other antimicrobials 55 (Fig. 3).   Ketoconazole synergises the mosquito mortality caused by ivermectin. Ketoconazole significantly increased the mortality of Anopheles gambiae fed in all samples containing ivermectin irrespectively of the ivermectin concentration (Fig. 4). The effect was predominantly observed at the expense of early mortality as seen by a reduction in the mean survival and the time to median mortality in all ketoconazole groups ( Table 3).
The mortality of mosquitoes fed on the only paired sample with rifampicin-ivermectin did not differ from that of ivermectin or rifampicin alone groups (data not shown).

Discussion
Our study shows that inhibition of the CYP 3A4 could safely increase ivermectin's time above target insecticidal concentration in the mammal host, increasing mosquito exposure. This finding could be harnessed to simplify the dosing regime and spare active pharmaceutical ingredient. Additionally, once outside the mammal host,  pharmacological CYP inhibition increases the mortality of exposed Anopheles gambiae mosquitoes, which could further improve efficacy. A natural next step would be to assess the potential implications of these findings in the field by means of modelling studies.
CYP inhibitors that also target the P-gp could theoretically increase penetration of the drug into the CNS and facilitate interaction with the GABA receptors, to which ivermectin has weak affinity. Thus, our main concern was that ketoconazole mediated CYP inhibition might reduce ivermectin's excellent safety profile. Our results suggest that the large therapeutic window of ivermectin and the difference in IC 50 for CYP and P-gp could safely allow CYP inhibition without a measurable increase of the drug in the CSF or clinically observed adverse events. Currently there are some "selective" CYP3A4 or P-gp inhibitors on the market but most drugs that affect one system typically have at least a partial effect on the other. Emerging more selective molecules 56 or even combination strategies 34 could overcome this issue.
Ivermectin resistance resulting from scale up of its veterinary use is well known. It was observed in nematodes as early as 1985 57 , four years after licensure of the drug. It was soon proven that resistance could be selected in the lab after only 8 generations of Haemonchus contortus exposed to sub-optimal dosing 58 . In arthropods highly ivermectin-resistant Musca domestica could be selected in the lab after only seven generations in the early 1990s 59 . Field reports in economically or public health relevant arthropods followed shortly 60,61 . The relative delay in global resistance reports was the result of ivermectin's broad spectrum. Most intestinal parasites in this spectrum could be removed with doses as low as 20 mcg/kg, but the 200 mcg/kg dose was selected based on the less susceptible organisms, the dose-defining species. The result is that any resistance arising in the more susceptible parasites could not be detected until their susceptibility was 10-fold higher 62 . This phenomenon has been called the window of escalation 62 and could be an important concept it if ivermectin is scaled up for vector control given that malaria vectors also differ in their ivermectin susceptibility 16,63,64 . While Anopheles gambiae 50 and Anopheles minimus seems to be remarkably susceptible based on their low LC 50 , others like Anopheles aquasalis or Anopheles dirus promise to become the dose-defining mosquito species 16,63,64 for the novel vector control use of the drug. Taking advantage of a synergist could potentially delay the emergence of ivermectin resistance in malaria vectors and extend the spectrum of ivermectin towards less susceptible vectors.
The findings from our study suggest that although metabolic pathways play a role in the mosquito defence from ivermectin, the activity of the detoxification mechanisms or efflux pumps could be modulated to synergise ivermectin and boost its efficacy. These findings warrant surveillance of early signs of metabolic resistance if ivermectin is used to reduce malaria transmission. At the same time, these results offer a simple potential tool to address metabolic resistance before it challenges efficacy.
Additionally, emerging data indicates that active ivermectin metabolites could cause mosquito mortality even when the mother drug is no longer detectable in plasma 65,66 . Increasing the time above mosquitocidal levels of ivermectin at the expense of reducing active metabolites could leave overall mosquito mortality unchanged 20 . Previous animal studies suggest dual CYP3A4/P-gp inhibitors do not significantly modify the PK of one of the major ivermectin metabolites, 3 O-desmethyl ivermectin 35,36 potentially suggesting an increase in systemic exposure mostly due to reduced excretion rather than reduced metabolism. Nonetheless, the evaluation of ivermectin metabolites is a complex process 22,67 and their potential role in the insecticidal effect of the drug would require additional investments.
Ketoconazole was used purely to test the concept of dual CYP/P-gp inhibition enhancing the PK and the ivermectin-related mosquito mortality, but it is in no way a molecule suitable for combination in MDA campaigns. Some additional limitations of this study include: a small number of mini-pigs was used for the pharmacoenhancement experiment, which explains the relative large concentration ranges; this pharmacokinetic effect, however, has previously been seen in other animal models 35,36 . We included rifampicin, a dual CYP3A4/P-gp inducer 49 to validate the results, evaluate changes in the CSF concentration and better understand the potential routes of manipulation of detoxification enzymes/transporters in the mosquito. Yet, Rifampicin mediated CYP 3A4/P-gp induction led to a marked reduction in systemic ivermectin exposure, possibly due to reduced intestinal absorption and/or pre-systemic metabolism; Although there is no evidence that ivermectin is a substrate of Organic Anion-Transporting Polypeptides (OATPs) in hepatic tissue or the intestinal epithelium, decreased bioavailability when co-administered with orange juice suggests this could occur 68,69 . If this was the case, inhibition of OATPs mediated uptake by rifampicin 70 could have also contributed to the reduced ivermectin absorption seen in our rifampicin pre-treated mini-pigs. Rifampicin-ivermectin samples were insufficient for the mosquito experiment. Nonetheless, the role of MDR1 polymorphism on the pharmacokinetics and pharmacodynamics of ivermectin in humans requires further exploration 32,33 , particularly in the African setting 71,72 . Finally, we acknowledge the lack of a dose-response effect in the membrane feeding assays. These assays were performed with reconstituted blood which can lead to varying final drug concentrations given to mosquitoes caused by the manual mixture; this could explain the apparent absence of a dose-response effect in our experiments but does not affect our conclusion about a pharmacodynamic synergism. Future experiments assessing drug concentrations inside the mosquito could help clarify this potential source of bias.
When used as anthelmintics, dual pharmacokinetic and pharmacodynamic enhancement of macrocyclic lactones such as ivermectin can occur by modulation of the ABC transporter. This results in higher drug bioavailability in the host and higher drug penetration in the nematode 73,74 , which is the main driver of efficacy. However, this dual in vivo enhancement by inhibiting the CYP has not been reported, even though it could have important implications for the treatment of certain helminths [75][76][77][78] . Particularly in the context of wider use of ivermectin mass-treatment as resistance in endo-and ectoparasites can be induced simultaneously 79 .
Provided an appropriate regimen of a selective molecule is defined and the appropriate cost-effectiveness evaluations are conducted, our findings could be applied to other systems and have important implications for the control of different vector-borne diseases.

Materials and Methods
Mini-pig procedures. The pre-medication drugs were administered by crushing the tablets and mixing them with canned meat products used for enrichment. For ivermectin administration and sampling during the first 8 hours, the animals were sedated with a single intramuscular dose of tiletamine-zolazepan (4 mg/kg) and maintained with inhalatory isoflurane (1.5-2%). In pigs, there is no evidence of a significant interaction (other than being a substrate) of these drug classes with the CYP3A4 or the P-gp 80 . Ivermectin was administered via nasogastric tube.
For blood sampling during the first eight hours, a femoral line was left in place and later removed during recovery from anaesthesia. Blood sampling thereafter was done by jugular puncture under restrain. Sampling points for blood were: pre-treatment, at 0.5, 1, 2, 3, 4, 5, 6, 7, 8, 24, 48/72 hours and at days 6/7, 9 and 14. For the mosquito bioassay, additional serum samples were obtained at every time point, and frozen at −20 °C.
For obtaining CSF samples during the first 8 hours, an intrathecal catheter was placed by means of fluoroscopy-guided lateral lumbar puncture 81 (Fig. 5). Given the challenges posed by the particular anatomy of the pig, efforts were done to tunnel the intrathecal catheter and connect it to a subcutaneous reservoir in order to avoid repeating the technically demanding lumbar puncture during phase II of the study (Fig. 5). Given the risk of damaging nerve roots when accessing the reduced thecal space, the pigs received at least 20 ml/kg of isotonic IV fluids during the first 8 hours and motor impairment lasting more than 72 hours was included as a humane endpoint criterion. Sampling points for CSF were: pre-treatment and at 0.5, 1, 2, 3, 4, 5, 6, 7 and 8 hours.

Mosquito procedures.
For all experiments, we used the fully susceptible Anopheles gambiae s.s. colony maintained at the insectary in Bagamoyo. Mosquitoes were reared at 28 °C, 80% humidity and with a photoperiod of 12:12. Larvae were kept in large plastic bowls covered with netting containing approximately 2 litres of water with larval densities not exceeding one 4 th instar larvae per ml. Larvae were fed Tetramine fish flakes ad libitum. Once they developed into pupae these were collected using disposable plastic Pasteur pipettes and transferred to small bowls containing approximately 200 ml of clean water. The pupae bowls were then placed inside fully screened cages (30 × 30 × 30 cm) where adult mosquitoes were allowed to emerge and given 10% glucose as a sugar source ad libitum.
Only mosquitoes of 2-5 days of age were used for the experiments. Before the experiments, mosquitoes were starved from sugar for 12-18 hours and from water for one hour. Hungry females were selected by applying a bottle with heated water (37-40 °C) to the cage. 40-50 hungry females were transferred to each paper cup using a mouth aspirator.
We performed membrane feeding assays with serum replacement as previously described 51 . For each experiment, the selected pig sample was thawed and 1 ml of whole blood was centrifuged at 2100 rpm for 10 minutes. The serum volume was noted and replaced with the same amount of the sample pig-serum being tested. After this, tubes were inverted 10 times. Each sample was tested in triplicate.
In order to reduce potential confounding arising from variations in bloodmeal size, partially fed mosquitoes were discarded and only fully engorged females followed up for the mortality assessment. After each feeding experiment, the cups were placed on ice for 1-2 minutes to select and transfer fully engorged females.
Cups with fully engorged mosquitoes were kept at 24-26 °C and 65-70% relative humidity inside climate controlled incubators. Mosquitoes were maintained with 10% glucose solution using impregnated cotton. Mortality was recorded daily for 9 days by a person blinded to the allocation of the cups. Dead mosquitoes were removed from the cups daily.
Analytics and PK calculations. Blood samples were drawn in 3 ml EDTA tubes and centrifuged. Plasma was separated and frozen at −20 °C until analysis. CSF samples were frozen at −20 °C.
Ivermectin levels in plasma and CSF were determined using a validated adaptation of a previously described HPLC-FLD 82 . The detection limit in plasma and CSF was 0.1 ng/ml, the quantification limit was 0.5 ng/ml.
We used the lethal concentration 50 of Anopheles gambiae described by Ouedraogo 50 (6 ng/ml) as target concentration and particularly aimed at increasing the time ivermectin was present in plasma above this level. Pharmacokinetic analyses were performed using Phoenix WinNonlin 6.4 (Certara, Princeton, NJ, USA). Pharmacokinetic parameters were derived from non-compartmental analyses.
Animals. Six hybrid mini-pigs (2 male, 4 female) weighing between 46 and 70 kg were procured from our accredited breeding centre at the University of Navarra. They were housed individually throughout the study at the animal research facilities in the University of Navarra. From arrival onward, the animals were assessed daily for general wellbeing and after the intervention, also for specific ivermectin toxicity signs.
European pharmacopeia standard ivermectin for HPLC calibration was procured from Sigma Aldrich.

Statistics. The comparison of PK parameters between ivermectin groups was done by means of a two-sided
Mann-Whitney test. Comparison between combined ivermectin and ketoconazole plus ivermectin groups was done with a one sided Mann-Whitney test.
Survival Kaplan-Meier analysis was performed in Addinsoft's XLSTAT ® Version 2016.04.32229 (New York, NY, USA). Comparisons of survival patterns were done with Log-rank test using a 5% significance level adjusted for cluster design.
Ethics. All procedures were reviewed and approved by the animal experimentation ethics committee of the Universidad de Navarra (Registry number E27-15(135-12E4)). All procedures were performed in accordance with the relevant guidelines and regulations.
Data availability. The datasets generated during the current study are available from the corresponding author on reasonable request.