**Anti-Trypanosomal Proteasome Inhibitors Cure Hemolymphatic and Meningoencephalic Murine Infection Models of African Trypanosomiasis**

**Srinivasa P S Rao 1,\*, Suresh B Lakshminarayana 1, Jan Jiricek 1, Marcel Kaiser 2,3, Ryan Ritchie 4, Elmarie Myburgh 5, Frantisek Supek 6, Tove Tuntland 6, Advait Nagle 6, Valentina Molteni 6, Pascal Mäser 2,3, Jeremy C Mottram 7, Michael P Barrett <sup>4</sup> and Thierry T Diagana <sup>1</sup>**


Received: 9 January 2020; Accepted: 14 February 2020; Published: 17 February 2020

**Abstract:** Current anti-trypanosomal therapies suffer from problems of longer treatment duration, toxicity and inadequate efficacy, hence there is a need for safer, more efficacious and 'easy to use' oral drugs. Previously, we reported the discovery of the triazolopyrimidine (TP) class as selective kinetoplastid proteasome inhibitors with in vivo efficacy in mouse models of leishmaniasis, Chagas Disease and African trypanosomiasis (HAT). For the treatment of HAT, development compounds need to have excellent penetration to the brain to cure the meningoencephalic stage of the disease. Here we describe detailed biological and pharmacological characterization of triazolopyrimidine compounds in HAT specific assays. The TP class of compounds showed single digit nanomolar potency against *Trypanosoma brucei rhodesiense* and *Trypanosoma brucei gambiense* strains. These compounds are trypanocidal with concentration-time dependent kill and achieved relapse-free cure in vitro. Two compounds, GNF6702 and a new analog NITD689, showed favorable in vivo pharmacokinetics and significant brain penetration, which enabled oral dosing. They also achieved complete cure in both hemolymphatic (blood) and meningoencephalic (brain) infection of human African trypanosomiasis mouse models. Mode of action studies on this series confirmed the 20S proteasome as the target in *T. brucei*. These proteasome inhibitors have the potential for further development into promising new treatment for human African trypanosomiasis.

**Keywords:** sleeping sickness; drug discovery; *Trypanosoma* growth inhibitors

#### **1. Introduction**

Human African trypanosomiasis (HAT) is a neglected tropical disease caused by the protozoan parasites *Trypanosoma brucei gambiense* and *Trypanosoma brucei rhodesiense*. The disease is endemic to sub-Saharan Africa and transmitted by tsetse flies (*Glossina* spp.). Over the last decade, there has been a significant reduction in the number of new cases of HAT, reaching below ~1000 reported new cases per annum in 2018 [1]. HAT comprises hemolymphatic (stage 1) and meningoencephalic (stage 2) infections. The successful introduction of nifurtimox–eflornithine combination therapy (NECT) for the treatment of gambiense HAT significantly helped in achieving a cure in stage 2 HAT patients [2]. Although NECT is effective, it requires long infusions and continuous monitoring. The introduction of fexinidazole, as an oral drug capable of curing stage 1 and stage 2 disease, offers great potential, and a further orally available drug, acoziborole, is currently being evaluated in late-stage clinical trials [3]. For treatment of rhodesiense HAT, suramin and melarsoprol (a highly toxic arsenical) are still used. New drugs remain desirable if we are to assure no repeat of the historical re-emergence of HAT, following a successful campaign in the mid-twentieth century where cases had dropped to the low thousands, only to resurge to an estimated 300,000 cases by the turn of the century [4]. Recent publications showing that trypanosomes dwell in adipose tissue [5] and skin [6], along with several reports of possible animal reservoirs of gambiense trypanosomes and latent human infections, all point out potential threats to the elimination of HAT [7].

Novartis, in collaboration with academic partners, embarked to find novel, safe short-course therapies for treatment of all forms of HAT. Previously, we reported [8] the discovery of the triazolopyrimidine (TP) chemical class as growth inhibitors of *Leishmania donovani*, *Trypanosoma cruzi* and *T. b. brucei* and identified the 20S proteasome as the target responsible for the pharmacological activity. Furthermore, exemplar from this class (GNF6702) demonstrated efficacy in the murine models for the three indications [8]. An earlier study had shown in vitro growth inhibition activity of TP cpds against *T. b. brucei* and GNF6702's stage 2 efficacy at highest dosing regimen of 100 mg/kg once daily [8]. Here, we describe detailed biological, chemical and pharmacological characterization of the three TP class of inhibitors (GNF3849, GNF6702 and NITD689) in various HAT-specific assays. The TPs are active against disease causing *T. b. gambiense* and *T. b. rhodesiense* strains, as well as drug-resistant (melarsoprol and pentamidine) isolates. These compounds inhibit the chymotrypsin activity of the 20S proteasome and are trypanocidal showing concentration-time dependent kill. Stage 2 HAT treatment requires compounds to have unique properties that enable them to cross the blood–brain barrier in order to be efficacious against CNS-resident parasites. Two compounds, GNF6702 and the newer analog NITD689, had favorable physicochemical and pharmacokinetic properties amenable for oral dosing and achieving free brain concentrations required for stage 2 efficacy. They also achieve relapse-free cure in mouse models of stage 1 and 2 trypanosomiasis, in a dose-dependent manner, suggesting the potential to treat all forms of HAT.

#### **2. Materials and Methods**

#### *2.1. Parasites, Cell Culture and Growth Inhibition Assays*

The *T. b. brucei* strain Lister 427 (bloodstream form) parasites were continuously grown in complete HMI-9 medium supplemented with 10% Serum Plus and 10% heat-inactivated fetal bovine serum (FBS) [8]. All other parasite strains were cultured as described elsewhere [9].

For determination of 50% growth inhibition, all compounds were dissolved in DMSO, and 200 nL of threefold serially diluted compounds were added into solid-bottom 384-well white plates (Greiner Bio-One, Kremsmunster, Austria) by an Echo 555 acoustic liquid-handling system. Forty microliters of 10<sup>4</sup> cells/mL of *T. b. brucei* parasites was added into each well, and the plates were incubated in a 5% CO2 incubator at 37 ◦C for 48 h. Viability of parasites were determined by measuring intracellular ATP levels, using CellTiter-Glo (CTG) luminescent cell viability reagent (Promega, Madison, Wisconsin WI, USA). The EC50 values were determined by using GraphPad Prism software. Growth inhibition assays for all

clinical isolates were carried out as described earlier [9]. All experiments had two technical replicates and three biological replicates. Appropriate statistical tests were used for evaluating significance, and mean ± SEM is represented in the tables.

#### *2.2. HepG2 Cytotoxicity Assay*

The HepG2 (human hepatocellular carcinoma) cells were obtained from ATCC (American Type Culture Collection) and grown in RPMI media. Cytotoxicity assay was performed in 384-well format, and 25 <sup>μ</sup>L, approximately 1.6 <sup>×</sup> <sup>10</sup><sup>4</sup> cells/mL, suspension was added to clear-bottom 384-well Griener plates and incubated in 5% CO2 incubator, at 37 ◦C for 24 h. Once the cells adhered, 125 nL of threefold serially diluted compounds in DMSO was added. Plates were incubated for 96 h, at 37 ◦C, in a 5% CO2 incubator. Plates were read for viability by adding 10 μL of CCK-8 reagent (APExBIO) into each well, followed by3hincubation and absorbance reading at 450 nM, using an Envision plate reader. Absorbance values were used for determination of cytotoxic concentration (CC50) required to inhibit growth by 50%, using GraphPad Prism software. Puromycin was used as a positive control. All experiments had two technical replicates and three biological replicates. Appropriate statistical tests were used for evaluating significance, and mean ± SEM is represented in the tables.

#### *2.3. Determination of Solubility, PAMPA, Plasma Protein Binding, Brain Tissue Binding and Microsomal Clearance*

Test compounds' solubility was determined in a high-throughput thermodynamic solubility assay, as described previously [10]. The PAMPA (parallel artificial membrane) assays were carried out, using a standard protocol [11]. Plasma protein binding was determined by using mouse blood [12], whilst brain tissue binding was determined by using rat brain tissue homogenate. Intrinsic metabolic clearances were determined in mouse liver microsomes, using the compound depletion approach and LC–MS/MS quantification [13].

#### *2.4. Time to Kill and Reversibility Assays*

Time-to-kill experiments were carried out to determine the ability of compounds to kill bloodstream form of *T. b. brucei* Lister 427 at 6, 24 and 48 h post-compound treatment. Viability of parasites were assessed by measuring ATP content as a surrogate. The assay was conducted in 384-well format, in a similar manner as the growth inhibition assay stated above, but with minor modifications. Compound-containing plates were incubated with 40 <sup>μ</sup>L of approximately 1 <sup>×</sup> 105 parasites per mL, and at each time point, CTG reagent was added to lyse the parasites, and luminescence was measured after 30 min of incubation, using a Tecan M1000 plate reader.

Reversibility assessment to establish time and concentration required to achieve irreversible (relapse-free) growth inhibition in vitro was carried out as described elsewhere [14]. The ACcure is the absolute concentration required to achieve sterile cure under in vitro conditions with incubation of compound for 24, 48 and 72 h, respectively.

All experiments had two technical replicates and three biological replicates. Appropriate statistical tests were used for evaluating significance, and mean ± SEM was represented.

#### *2.5. In Vivo Pharmacokinetic (PK) Analysis*

Determination of intravenous (i.v.) and per oral (p.o.) pharmacokinetics (PK) were carried out by using male BALB/c mice. For i.v. PK studies, compounds were formulated at a concentration of 2.5 mg/mL in 75% PEG300 and 25% D5W (5% dextrose in distilled water). To avoid any granular material, the solution was filtered, using a 0.45 μm syringe filter. The filtered solutions were dosed intravenously to mice via the lateral tail vein, at 5 mg/kg, with a dosing volume of 2 mL/kg. For all the p.o. PK studies, compounds were resuspended in 0.5% *v*/*v* methyl cellulose in 0.5 % *v*/*v* tween 80 solution. For 20 mg/kg dosing, 200 μL of compounds was administered orally to mice. Six blood samples of 50 μL each were collected serially from each animal, up to 24 h after dosing. Blood samples were collected into heparin microtainers and centrifuged, and then plasma was separated and frozen until analysis. Plasma samples (20 μL) were extracted with acetonitrile:methanol (3:1) containing internal standard. The samples were vortexed and then centrifuged in an Eppendorf centrifuge 5810R, at a setting of 4000 rpm, for 5 min, at 4 ◦C. The supernatant was transferred to a 96-well analysis plate and analyzed, using optimized LC/MS/MS conditions. For every experiment, 3 mice were used per compound, per dose. The plasma concentration-time profile was obtained by plotting the mean value from the three animals at each time point. Various PK parameters, such as Cmax (maximum concentration), area under curve (AUC) and oral bioavailability (F) for compounds, were calculated by non-compartmental regression analysis, using an in-house fitting program developed at GNF [8]. All the in-life studies were carried out under protocols approved by the Animal Care and Use Committee (IACUC), following animal ethics guidelines of GNF.

For measurement of the brain-to-plasma drug concentration ratio, mice were dosed intravenously with 1 mg/kg compounds. Mice were euthanized at 5 and 60 min post-dosing, and blood and brains were collected. The compound concentrations in plasma and brains were measured, following the protocol described above.

#### *2.6. Hemolymphatic Mouse Model (Stage 1 HAT E*ffi*cacy Model)*

The *T. b. brucei* STIB795 acute mouse model mimics the hemolymphatic stage of the sleeping sickness disease. We used six female NMRI mice per experimental group, divided into two cages. Heparinized blood from a donor mouse with approximately 5 <sup>×</sup> 106 per mL parasitemia was suspended in phosphate saline glucose (PSG), to obtain a parasite suspension of 1 <sup>×</sup> 105 per mL. Each mouse was injected with 0.25 mL of parasite suspension (10<sup>4</sup> bloodstream forms of *T. b. brucei* STIB795) intraperitoneally. All compounds were formulated in 0.5% methylcellulose and 0.5% Tween80. Three days post-infection, test compounds were administered orally on four consecutive days, in a volume of 100 μL/10 g. Three mice served as infected–untreated controls. The control mice were not injected with the vehicle, because we have established in our labs that this vehicle does not affect parasitemia or the mice. Until 31 days post-infection, parasitemia was monitored microscopically by tail-blood examination twice a week. Mice were considered cured when there was no parasitemia detected in the tail blood. All the results from the individual experiments were reported as number of mice cured over total number of mice treated. A Kaplan–Meier plot was used for representing the number of mice cured at different treatment doses.

In vivo efficacy studies in mice were conducted at the Swiss Tropical and Public Health Institute (Basel) (License number 2813), according to the rules and regulations for the protection of animal rights ("Tierschutzverordnung") of the Swiss "Bundesamt für Veterinärwesen". They were approved by the veterinary office of Canton Basel-Stadt, Switzerland.

#### *2.7. Meningocephalic Mouse Model (Stage 2 HAT E*ffi*cacy Model)*

The GVR35 mouse CNS model mimics the second (meningoencephalic) stage of African trypanosomiasis. Female CD1 mice (~8 weeks old, from Charles River) were injected intraperitoneally (i.p.) with 3 <sup>×</sup> <sup>10</sup><sup>4</sup> *T. b. brucei* (GVR35-VSL2) bloodstream-form parasites [15]. As controls, a group of three untreated mice and another group of three mice treated with diminazene aceturate (DA) were included. The DA is a known anti-trypanosomal compound, which lacks brain penetration; hence, they clear only the blood parasitemia, leaving behind the parasites in brain. Mice treated with DA usually relapse after 42 days post-infection. Groups of six infected mice were dosed with TP compounds by oral gavage, once daily, from day 21 or 22 post-infection, for seven days.

Blood parasitemia was quantified weekly from day 21/22, by microscopy of blood from the tail vein. Mice were imaged for bioluminescence using an in vivo imaging system (IVIS) prior to treatment on day 21/22, and weekly after treatment, as described previously [8,15]. Briefly, mice, in groups of three, were injected i.p. with 150 mg of D-luciferin (Promega) per kilogram body weight in PBS and imaged after 10 min, using an IVIS Spectrum (PerkinElmer, Waltham, Massachusetts MA, USA). Living Image Software (PerkinElmer) was used for acquisition and analysis of bioluminescence imaging data. Bioluminescence in the same rectangular region of interest (ROI) on whole-body mouse images was quantified and is shown in total flux (photons per second). Images were cropped to the ROIs and composites of images from representative mice are shown. Uncured mice were euthanized within 1 or 2 days of parasite detection in the blood. Cured mice were euthanized between day 92 and 101 post-infection. All the results from the individual experiments were reported as number of mice cured over total number of mice treated. A Kaplan–Meier plot was used for representing the number of mice cured at different treatment doses.

All animal protocols and procedures were reviewed and approved by the UK Home Office (Project License PPL60/4442 entitled "Molecular Genetics of Trypanosomes and Leishmania") and University of York and University of Glasgow Ethics Committees, and was done in accordance with the Animals (Scientific Procedures) Act 1986 (ASPA).

#### **3. Results**

#### *3.1. Extended Characterization of TP Series of Kinetoplastid Proteasome Inhibitors for Treatment of Human African Trypanosomiasis*

Previously, we described our efforts to identify novel chemotypes for the treatment of HAT. These efforts led the identification of GNF6702, a prototypical pan-kinetoplastid inhibitor that was efficacious in the mouse models for all three kinetoplastid diseases [8].

All the three TP compounds (GNF3849, GNF6702 and NITD689) showed potent growth inhibition (EC50 < 70 nM) against *T. b. brucei* with varying cytotoxicity profile (Figure 1). Further evaluation of triazolopyrimidine analogues for the optimal blood–brain barrier penetration properties, such as lipophilicity (cLogP) and polar surface area, led to prioritization of NITD689, in addition to GNF3849 and GNF6702. All three compounds exhibited good membrane permeability, low mouse liver microsomal clearance and moderate lipophilicity (Figure 1). While GNF3849 had favorable potency against *T. brucei*, it was cytotoxic against HepG2 cells. Moreover, it also suffered from poor solubility and high plasma protein binding, indicating that higher total exposure might be needed to observe efficacy in in vivo. As reported previously [8], replacement of a phenyl group of GNF3849 with a pyridine moiety (GNF6702) improved solubility, lowered PPB and also created a molecule with a better cytotoxicity. Medicinal chemistry optimization by increasing SP3 fraction within the molecule by replacing pyridine group with a tertiary butyl group led to the identification of NITD689. The NITD689 showed improved solubility, reduced plasma protein binding and retained non-cytotoxic profile, with 30 nM potency against *T. b. brucei* (Figure 1 and Table 1).


**Table 1.** Biological, physicochemical and in vitro pharmacokinetic properties of the triazolopyrimidine inhibitors.

Note: All EC50 and CC50 values correspond to mean ± SEM (n = 4 biological replicates); \* LogD: measured octanol water co-efficient; Mol wt: molecular weight; PSA: polar surface area; MPO: multi-parametric optimization; PAMPA: parallel artificial membrane permeability assay; FA: fraction absorbed. **#** Khare et al., Nature 2016 [8]; \$ patent US 2019/0000852 A1.

**Figure 1.** Chemical structure of the triazolopyrimidine class of inhibitors.

#### *3.2. Biological Profiling of the TP Class of Inhibitors*

Triazolopyrimidines were evaluated for their potential to kill parasites under in vitro conditions. Time-to-kill assessment showed that all the three compounds were trypanocidal. At an early time point (6 h post exposure), no significant cidality was seen compared to the untreated control, even at the highest concentration tested (16.7 μM). However, at 24 and 48 h post-incubation, with compounds, significant kill was evident, suggesting this class of inhibitors requires 24 h in order to show cidality. All three compounds showed concentration and time-dependent lethality (Figure 2A).

**Figure 2.** Biological characterization of the TP class of compounds. (**A**) Time-to-kill profile indicating concentration-time dependent kill. (**B**) To achieve sterile cure without relapse under in vitro conditions. All experiments were carried out three times, independently, and mean ± SEM was plotted.

In order to evaluate the ability of TP compounds to kill all parasites without recrudescence (sterile cidality), we carried out drug wash-out assays. All the three TP compounds achieved sterile cure with ACcure (Absolute concentration of drug required to kill parasites without relapse) values of <100 nM, after a 72-h compound treatment and subsequent wash-out. The ACcure values also decreased as the time of exposure increased, indicating exposure-driven sterile cure (Figure 2B).

#### *3.3. The TP Class of Inhibitors Are Active against Clinical Isolates*

GNF3849 and GNF6702 were also tested for their ability to inhibit growth of clinical isolates of *T. b. gambiense* and *T. b. rhodesiense*. Our compounds showed single-digit nM potency against all strains tested. The compounds were also active against pentamidine- and melarsoprol-resistant strains of *T. b. rhodesiense*, as well as *T. b. brucei* (Table 2), suggesting a distinct mechanism of action, at least in terms of drug uptake, since resistance to these drugs relates to loss of transporters specific for their uptake [16,17].


**Table 2.** Growth inhibitory profile of the triazolopyrimidine class of compounds against various parasite strains.

Note: All EC50 values correspond to mean ± SD (n = 3 biological replicates).

#### *3.4. The TP Class of Compounds Are Proteasome Inhibitors*

The TP class of inhibitors is active against both *T. cruzi* and *T. brucei.* Resistance selection in *T. brucei* proved difficult; hence, we generated *T. cruzi* resistant isolates, using a classical resistant mutant generation approach. Whole-genome sequencing identified single nucleotide polymorphisms in F24L and I29M of β4 subunit of 20S proteasome. Further validation of the proteasome as a target for the TP class of compounds has been described elsewhere [8]. *T. brucei* parasites modified to express an F24L mutant β4 subunit of the 20S proteasome were previously shown to be resistant to the parent proteasome inhibitor. Here, we used the same strain to demonstrate its resistance to the new TP compounds, confirming that they, too, target the proteasome (Table 3). Meanwhile, bortezomib, a known chymotrypsin proteasome inhibitor, was equipotent against both wild-type and F24L mutant of *T. brucei*, suggesting that bortezomib and TP compounds interact differently by binding into different pockets within 20S proteasome.

**Table 3.** Mutations in F24L in the 20S proteasome β4 subunit confers resistance to TP class of compounds.


Note: All EC50 values correspond to mean ± SEM (n = 3 biological replicates); PSMBWT: *T. brucei* ectopically expressing wild-type copy of 20S proteasome β4 subunit; PSMBF24L: *T. brucei* ectopically expressing F24L mutant copy of 20S proteasome β4 subunit.

#### *3.5. In Vivo Pharmacokinetic Properties of TP Class of Compounds*

In vivo mice PK profiling was carried out, using intravenous and oral routes (Table 4). Following intravenous administration, all three TP compounds displayed a moderate volume of distribution (Vss: 1.2 to 1.5 L/kg), low total systemic clearance (2.5 to 15% of hepatic blood flow) and moderate-to-long elimination half-life (1.6 to 7 h). Following oral administration at 20 mg/kg, all three TP compounds showed good oral exposure, with bioavailability ranging from 34% to 100%. The dose-normalized total exposure was highest for GNF6702 in terms of AUC, followed by GNF3849 and NITD689. Although

NITD689 had the lowest exposure in terms of total concentration, it had the highest free Cmax and better free AUC compared to GNF3849, due to low plasma protein binding (Figure 1 and Table 4).


**Table 4.** In vivo pharmacokinetics properties of the TP class of compounds.

Note: I.V. PK: intravenous pharmacokinetics in mouse; P.O. PK: per oral pharmacokinetics in mouse; Vss: Volume of distribution at steady state; CL: total systemic clearance; T1/2: Elimination half-life; Cmax: maximum concentration reached in blood, values in parenthesis represent free fraction; Tmax: time to reach maximum concentration; AUC: exposure between 0 to infinity, values in parenthesis represent free fraction; F: oral bioavailability.

*3.6. TP Class of Compounds Is E*ffi*cacious against Hemolymphatic Infection in a Stage 1 HAT Mouse Model*

Since all three TP compounds showed promising in vivo PK properties in mice, they were evaluated for their ability to achieve cure in a mouse model of stage 1 (hemolymphatic) disease. GNF3849 achieved complete cure at all doses (7.5, 25 and 75 mg/kg, once daily, for four days). The other two compounds, GNF6702 and NITD689, showed dose-dependent cure, with increasing doses showing better cure (Figure 3). Minimum efficacious doses for GNF6702 and NITD689 were 1 and 10 mg/kg QD dosing for four days, respectively (Table 5).


**Table 5.** In vivo efficacy of GNF3849, GNF6702 and NITD689 in a HAT hemolymphatic mouse model.

NOTE: Mean day of relapse refers to days post infection; QD = once daily; BID = twice daily.

**Figure 3.** In vivo efficacy of GNF3849, GNF6702 and NITD689 in a HAT hemolymphatic mouse model. Six mice each were orally treated for four days, with varying doses of compounds, three days post-infection. Mice were monitored for 30 days post-infection, and cure plot (Kaplan–Meier plot) showing percentage of animals cured over time are shown.

#### *3.7. Assessment of Brain Permeability, Tissue Binding and Partitioning of TP Compounds*

In order to evaluate potential of brain permeability, all three compounds were tested in an MDR1- MDCK permeability assay. Compounds with better A to B permeability and reduced efflux ratio have good potential to reach high concentrations in the brain. Those that have a high efflux ratio are generally substrates for Pgp transporters, thereby having a higher propensity to be excluded from the brain. All three compounds had reasonable permeability and an efflux ratio of <3.5 (Table 6). To assess directly the TP compounds' ability to penetrate brain, mice were dosed with compound, and plasma and brains were collected after 5 and 60 min following intravenous dosing. GNF3849 had the highest brain-to-plasma ratio (B/P ratio), followed by GNF6702 and NITD689 (Table 6). Although GNF3849 had a better B/P ratio based on total concentration, it had high brain tissue protein binding (>99%), whereas GNF6702 and NITD689 had lower brain tissue binding (94.7% and 96.5%, respectively). This implies that the free concentration available at the site of infection could be lower for GNF3849 compared to the other two TP compounds.

**Table 6.** Assessment of brain permeability, tissue binding and partitioning of TP compounds.


Note: BTB: rat brain tissue binding; MDR1-MDCK = Multi Drug Resistant 1 overexpressing Madin–Darby canine kidney cells; A–B = Apical to Basolateral; B–A = Basolateral to Apical; B/P ratio = brain-to-plasma ratio.

#### *3.8. TP Class Compounds Are E*ffi*cacious against Meningoencephalic Infection in HAT Mouse Model*

Having demonstrated brain permeation, all three compounds were evaluated in a stage II (meningoencephalic) mouse model of HAT, with various doses (Figure 4). The recrudescence of bioluminescent parasites was monitored by using an in vivo imaging system over a period of three months (Figure 5). GNF6702, which had the best exposure and longest half-life, was dosed by oral gavage at 3, 10, 30 and 100 mg/kg, once daily. A dose-dependent increase in cure rate was noticed, with 30 and 100 mg/kg showing complete cure, and 10 mg/kg showing partial cure, with four out of six mice achieving relapse-free cure (Table 7). GNF3849 was the next best molecule, with reasonable exposure (~2.5 fold less than GNF6702) and a long half-life of 6.5 hours. Hence, they were dosed at 7.5 and 75 mg/kg, once daily. GNF3849 failed to achieve complete cure at either dose with 75 mg/kg, showing only 50% cure. GNF3849 had high plasma protein (>99%) and brain tissue binding (>99%), which could have resulted in significantly lower free concentration of compounds required to achieve 100% relapse-free cure.

NITD689 had promising physicochemical properties, better permeability and the lowest efflux ratio, although the exposure was lower than GNF6702 (Tables 2, 4 and 6). NITD689 also had a moderate half-life of 1.6 h; hence, we adopted both a once-daily and twice-daily dosing schedule. Both doses of 15 mg/kg twice daily and 30 mg/kg once daily failed to cure, suggesting that the time and concentrations reached were not sufficient to kill all parasites in the brain. However, 30 mg/kg twice daily and 60 mg/kg once daily of NITD689 achieved complete cure, without relapse.

**Figure 4.** In vivo efficacy of GNF3849, GNF6702 and NITD689 in HAT meningoencephalic mouse model. Six mice each were orally treated for seven days, with varying doses compounds, 21or 22 days post-infection. Mice were monitored for 92–94 days post-infection, and cure plots (Kaplan–Meier plots) showing percentage of animals cured over time are shown. Three mice each were dosed with vehicle control and diminazene aceturate. Note the early parasite recrudescence in mice treated with diminazene aceturate.

**Figure 5.** Bioluminescence imaging of mice infected with *T. b. brucei*. Dose-dependent clearance of parasites from triazolopyrimidine class of inhibitors. In vivo quantification of bioluminescent *T. b. brucei* (GVR35–VSL2) in infected mice before and after treatment; day 21/22, start of treatment; day 28/29, 24 h after last dose; day 44/50 and day 92/94, parasite recrudescence or cure in mice treated with GNF3849, GNF6702 and NITD689 (images of two representative mice from a total of six mice are shown). Blood parasitemia (in parasites/mL, red font below image) and whole mouse total flux (in photons per second, black font above image) values of each animal are shown; QD, once daily; BID, twice daily; N.D., not detectable; Tx, treatment. The same two representative mice are shown for all time points. Mice with detectable parasites were euthanized and are therefore not shown at day 92/94. Images from uninfected mice, aged-matched for day 21, that were collected independently, using the same acquisition settings, are shown in the gray box (two of three mice are shown).


**Table 7.** In vivo efficacy of GNF3849, GNF6702 and NITD689 in a HAT meningoencephalic mice model.

NOTE: Mean day of relapse refers to days post infection; \* mean values shown are for the mouse which relapsed; values in parenthesis are for mice which did not relapse; QD = once daily; BID = twice daily.

#### **4. Discussion**

Several research groups have been working on the development of anti-trypanosomal compounds with a potential to treat HAT. The most significant challenge has been achieving brain penetration for chemical molecules, which is critical for killing brain stage parasites to achieve complete cure in stage 2 infection. For example, N-myristoyltransferase inhibitors have been shown to be potent trypanocides and curative of stage 1 models of the disease, but failed to achieve reasonable brain concentrations, thereby leading to failure to cure stage 2 models of HAT [18]. Attempts to screen compound libraries specifically inhibiting kinases [19] and proteases [20] also identified potent trypanocides which failed to achieve complete cure in mouse models, due to lack of adequate brain penetration. Our attempts to find novel trypanocides led to the identification of potent growth inhibitors. Further medicinal chemistry optimization of compounds led to brain-penetrant derivatives belonging to TP class (GNF6702 and NITD689) which completely cured both stages of infection.

We had previously reported that the TP class of molecules are chymotrypsin proteasome inhibitors with pan-kinetoplastid activity [8]. All the compounds described in the current manuscript also inhibited chymotrypsin activity of the 20S proteasome in *T. brucei*. The *T. b. brucei* strain overexpressing F24L mutation in β4 subunit of 20S proteasome showed greater than 60-fold shift in growth inhibition concentration for the TP compounds, confirming on-target activity. Recently, Wyllie and co-workers also described proteasome inhibitors with the potential to treat visceral leishmaniasis [21]. The proteasome inhibitors described here demonstrate great potency against clinical isolates of *T. b. gambiense* (EC50 = < 10 nM), compared to fexinidazole (EC50 = 0.95–3.3 μM) [9] and acoziborole (EC50 = 0.18–1 μM) [14]. Our compounds also showed concentration and time-dependent cidality and relapse-free kill of all parasites in wash-out assays in vitro, at concentrations below 100 nM. These cidality properties are essential for achieving cure in HAT mouse models [22]. Extensive medicinal chemistry optimization helped in improving PK properties required to achieve moderate brain penetration, which proved essential to cure brain infection. Both GNF6702 and NITD689 completely cured a stage 2 infection in a HAT mouse model, at 30 and 60 mg/kg dose, respectively. In addition, the exposures reached in the animal models were higher than their respective ACcure concentrations required for sterilization in vitro. Furthermore, these doses were lower than the curative dose of fexinidazole [9] (200 mg/kg) and comparable to that of acoziborole (25 mg/kg) [14]. A detailed structure activity relationship of TP class of compounds against both *T. brucei* and *T. cruzi* has been described by Nagendar and co-workers [23]. Although, one of their compounds, compound **20**, had a brain-to-plasma ratio of 0.23, it was not profiled in the stage 2 HAT model, due to higher plasma protein binding (98.5%), which might affect free concentration in brain required for achieving stage 2 efficacy.

While other developments in the treatment of HAT have been very promising, the TP class of proteasome inhibitors has significant potential for further progress. Other than acoziborole, which is in phase II studies for HAT, the proteasome inhibitors described here are the most advanced compounds with drug-like properties. They not only have promising in vitro and in vivo potency in disease relevant HAT models, but also have favorable pharmacokinetic properties, with potential for further development.

**Author Contributions:** Conceptualization: S.P.S.R., F.S., A.N., P.M., J.C.M., M.P.B. and T.T.D.; data curation: S.P.S.R. S.B.L., E.M. and T.T.; formal analysis, S.P.S.R., S.B.L., J.J., R.R. and E.M.; funding acquisition, S.P.S.R., P.M., J.C.M., M.P.B. and T.T.D.; methodology, S.P.S.R., S.B.L., A.N., T.T., R.R., E.M.; project administration, S.P.S.R., J.J., V.M., P.M., J.C.M., M.P.B. and T.T.D.; resources, J.J., M.K., F.S., T.T., E.M.; supervision, S.P.S.R., S.B.L, J.J., M.K., E.M., F.S., A.N., V.M., P.M., J.C.M., M.P.B. and T.T.D.; writing—original draft, S.P.S.R. and M.P.B.; Writing—review and editing, S.P.S.R., S.B.L., E.M., A.N., P.M., J.C.M., M.P.B. and T.T.D. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by Wellcome Trust, United Kingdom, grant number WT-103024MA and WT-104976.

**Acknowledgments:** Authors would like to thank Manoharan V., Wong J., Koh H., Vachaspati V., Karuna R., Wan K.H., Patra D., Biggart A., Lai Y.H., Liang F., Davis L.C., Mathison C.J., Liu X., Ballard J., Yeh V., Groessl T., Shapiro M., Smith P., Beer D. from Novartis and Cal M., Rocchetti R., Keller-Märki S., Riccio G. and Braghiroli C., from Swiss TPH for their technical assistance.

**Conflicts of Interest:** The authors declare no conflicts of interest. Some of the authors (S.P.S.R., S.B.L., J.J., F.S., T.T., A.N., V.M. and T.T.D.) are Novartis Employees.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **Phenotypic Drug Discovery for Human African Trypanosomiasis: A Powerful Approach**

#### **Frederick S. Buckner 1,\*, Andriy Buchynskyy 2, Pendem Nagendar 2, Donald A. Patrick 3, J. Robert Gillespie 1, Zackary Herbst 1, Richard R. Tidwell <sup>3</sup> and Michael H. Gelb <sup>2</sup>**


Received: 31 December 2019; Accepted: 2 February 2020; Published: 5 February 2020

**Abstract:** The work began with the screening of a library of 700,000 small molecules for inhibitors of *Trypanosoma brucei* growth (a phenotypic screen). The resulting set of 1035 hit compounds was reviewed by a team of medicinal chemists, leading to the nomination of 17 chemically distinct scaffolds for further investigation. The first triage step was the assessment for brain permeability (looking for brain levels at least 20% of plasma levels) in order to optimize the chances of developing candidates for treating late-stage human African trypanosomiasis. Eleven scaffolds subsequently underwent hit-to-lead optimization using standard medicinal chemistry approaches. Over a period of six years in an academic setting, 1539 analogs to the 11 scaffolds were synthesized. Eight scaffolds were discontinued either due to insufficient improvement in antiparasitic activity (5), poor pharmacokinetic properties (2), or a slow (static) antiparasitic activity (1). Three scaffolds were optimized to the point of curing the acute and/or chronic *T. brucei* infection model in mice. The progress was accomplished without knowledge of the mechanism of action (MOA) for the compounds, although the MOA has been discovered in the interim for one compound series. Studies on the safety and toxicity of the compounds are planned to help select candidates for potential clinical development. This research demonstrates the power of the phenotypic drug discovery approach for neglected tropical diseases.

**Keywords:** *Trypanosoma brucei*; human African trypanosomiasis; drug discovery; high-throughput screening; blood–brain barrier; brain permeability; pharmacology; phenotypic drug screening

#### **1. Introduction**

Two *Trypanosoma brucei* subspecies, *gambiense* and *rhodesiense*, cause human African trypanosomiasis (HAT). Natural transmission occurs in 36 countries of sub-Saharan Africa via the bite of infected tse-tse flies. After the initial cutaneous inoculation, early stage (hemolymphatic) infection occurs. In the Central/West-African form (Gambian HAT), the early stage can last for months to years before progressing to late-stage infection in the central nervous system (CNS). In the East-African form (Rhodesian HAT), the early stage lasts only a few weeks to months before causing late-stage disease. Once the parasites enter the CNS, patients suffer neuropsychiatric effects that culminate in coma and death if untreated (hence the name "sleeping sickness"). Optimal treatments for HAT will address the infection in the CNS, necessitating that drugs have the ability to penetrate the blood–brain barrier (BBB).

The therapeutic landscape for HAT has recently been upgraded with the approval of fexinidazole for the treatment of both the first-stage and second-stage of HAT due to *T. b. gambiense* in adults and children aged ≥6 years [1]. Fexinidazole represents the first all oral therapy for this disease and will likely be a major advancement over nifurtimox-eflornithine combination therapy (NECT), which requires parenteral administration (usually in a hospital setting) for the eflornithine component [2]. As much as fexinidazole is a welcome advancement, there continues to be a need for drug discovery for HAT in order to strengthen the therapeutic armamentarium: (1) for patients that cannot tolerate fexinidazole; (2) to address the inevitable risk of drug resistance; (3) to eliminate the need for a staging spinal tap before initiating therapy [3]; and (4) to respond to the unmet need of safe and effective drugs for Rhodesian HAT. Toward this end, our research group is conducting a drug discovery campaign to identify compounds that are distinct from nitroheterocycle drugs such as nifurtimox and fexinidazole, and that will meet the target product profile of an oral drug with activity for late stage HAT [4].

A high throughput phenotypic screen, performed at the Genomics Institute of the Novartis Research Foundation (GNF), formed the basis of this drug discovery campaign [5]. The compound library of 700,000 compounds consisted of a collection of small-molecules with druglike properties and structural diversity. The library had been previously profiled in more than 60 high throughput screens (both biochemical and cell-based), allowing for the identification and elimination of compounds with a "frequent hitter profile". The phenotypic assay measured the inhibition of bloodstream-form *T. brucei* cultures at a single compound concentration of 3.6 μM. The screen resulted in 3889 primary hits with an inhibition of greater than 50% for a 0.6% hit rate (Z' score > 0.6). The primary hits were further tested in dose-response assays in order to measure EC50 values. In parallel, the cytotoxicity (CC50) of the hit compounds was measured against cultures of the human hepatoma cell line (Huh7). A final set of compounds was compiled with *T. brucei* EC50 < 3.6 μM and CC50 > 10 μM, consisting of 1035 molecules that could be grouped into 115 distinct scaffolds [5]. This paper summarizes the progress to date of our drug discovery campaign to identify preclinical candidates for HAT, starting from this phenotypic screen.

#### **2. Materials and Methods**

The methods employed in this paper have been described in detail in previous publications as follows. All murine experiments were approved by the University of Washington Institutional Animal Care and Use Committee, IACUC approval code 4248-01 (animal welfare approval number A3464-01).

#### *2.1. Phenotypic Screen for T. brucei Growth Arrest*

Compound libraries were screened against the bloodstream form of the *T. brucei* isolate, Lister 427 [5]. Parasites were grown in 1536-well plates in 5.5 μL of HMI-9 medium in the presence of library compounds. All wells including negative controls contained a final of 0.4% dimethyl sulfoxide. Plates were incubated at 37 ◦C for 48 h and the parasite density was determined using the CellTiter-Glo reagent (Promega, Madison, Wisconsin WI, USA), a firefly luciferase assay system that measures the amount of cellular adenosine triphosphate present in plate wells.

#### *2.2. In Vitro Parasite Growth Arrest Assay*

Follow up compounds were tested for antiparasitic activity on *T. brucei brucei* (strain BF427) [5]. Parasites were tested in triplicate in the presence of serial dilutions of compound, and growth was quantified with AlamarBlue. Pentamidine isothionate (Aventis, Dagenham, U.K.) was included as a positive control in each assay (EC50 = 1.2 ± 0.3 nM).

#### *2.3. Mammalian Cell Cytotoxicity Assay*

Compounds were tested for cytotoxicity against CRL-8155 cells (human lymphoblasts) [5]. Cells were grown in culture with serial dilutions of compounds for 48 h and cytotoxicity was assayed using AlamarBlue (Life Technologies, Carlsbad, California CA, USA). Each dilution was assayed in quadruplicate with the standard error of the mean values averaging < 15%. Concentrations causing 50% growth inhibition (CC50) were calculated by nonlinear regression using GraphPad Prism software (San Diego, CA, USA).

#### *2.4. Solubility Measurement*

Solubility was measured in pH 7.4, pH 6.5, and pH 2.0 aqueous buffers in a two-tier system via LC-MSMS. In Tier 1 testing, 1 μL of DMSO stock (20 mM) was measured with a Hamilton syringe and diluted to 400 μL with the respective buffer, giving a final concentration of 50 μM test compound with 0.25% DMSO. The buffer solutions were capped and incubated while shaking at 37 ◦C for 24 h until equilibrium was reached. Buffer solutions were centrifuged at 14,000× *g* for 15 min and two aliquots are taken from the supernatant. The concentration of the test compound in each aliquot was determined by liquid chromatography-mass spectrometry/mass-spectrometry analysis and by calculations using a linear regression of the test compound standards made over a range of known concentrations. Solubility was reported as the final concentration in the supernatant. If the concentration in the supernatant was determined to be 50 μM (maximum solubility for Tier 1), then a Tier 2 test was carried out. In Tier 2 testing, 5 μL of the test compound's 20 mM DMSO stock was transferred to a microcentrifuge tube with a Hamilton syringe. The DMSO was then removed in a Speed-Vac concentrator and the test compound was diluted with 100 μL of the respective buffer, giving a final concentration of 1 mM test compound with negligible DMSO. The sample was heated and agitated by vortexing and by pipetting up and down to ensure the test compound was completely exposed to the buffer. The sample was then capped and incubated for 24 h while shaking at 37 ◦C. Buffer solutions were centrifuged at 14,000× *g* for 15 min and two aliquots were taken from the supernatant. The concentration of the test compound in the aliquots was determined by LC-MSMS, as described above.

#### *2.5. Permeability Across Monolayers of MDCKII-MDR1 Cells*

This assay utilizes Madin−Darby canine kidney cells that were transfected with the human MDR1 (P-gp) gene [5,6]. Permeability across these monolayers was measured in triplicate. The assay was performed with and without the addition of GF-120918, an inhibitor of the MDR1 efflux pump, to determine if the compound was a pump substrate. Propranolol was used as a permeable, non-MDR1 substrate control, and amprenavir was used as a permeable, MDR1 substrate control.

#### *2.6. Pharmacokinetic Studies in Mice*

Test compounds were administered to mice by oral gavage followed by blood sampling at intervals of 30, 60, 120, 240, 360, 480, and 1440 min [5,7]. Compounds were dosed orally at 50 mg/kg in 0.2 mL of dosing solution (7% Tween 80, 3% ethanol, 5% DMSO, 0.9% saline). Experiments were performed with groups of three mice per compound as published. Plasma was separated and extracted with acetonitrile for measurements of the compound concentrations by liquid chromatography/tandem mass spectrometry.

#### *2.7. Brain Permeability Studies*

Test compounds were injected (ip) at 5 mg/kg to three mice in a vehicle consisting of DMSO (5%), Tween 80 (7%), and EtOH (3%) in physiological saline (0.9%) solution [8]. At 1 h post injection, blood was collected, and plasma was separated by centrifugation. Simultaneously, the brain was removed and homogenized in acetonitrile. Concentrations of compound in the plasma and brain were determined via liquid chromatography/tandem mass spectrometry. Calculations of brain levels accounted for 3% volume/weight of blood in the brain.

#### *2.8. Anti-Parasite E*ffi*cacy Studies in Mice (Acute Model)*

Female Swiss-Webster mice age 6−8 weeks (group size = 5) were infected with 2 <sup>×</sup> 10<sup>4</sup> *T. brucei rhodesiense* STIB 900 strain on day 0, then administered the compound or vehicle for five days [5,8]. Treatments were administered orally in the same vehicle described above at a dose and schedule anticipated to maintain plasma concentrations well above the EC50. The first dose was 48 h after parasite injection, and dosing was 12 or 24 h apart. Parasitemia was monitored for 60 days by microscopic

analysis of blood collected from tail bleeds. Cures were defined by sustained clearance of microscopic parasitemia through the end of the 60-day observation period. Mice were euthanized when parasitemia was evident on the peripheral blood slides.

#### *2.9. Anti-Parasite E*ffi*cacy Studies in Mice (Chronic Model)*

Groups of six to eight mice were infected with 1 <sup>×</sup> 104 *T. b. brucei* (strain TREU667) to establish a chronic infection [8]. Treatment began on day 21 post-infection, and mice received 50 mg/kg test compound orally twice per day for 10 days (total of 20 doses) in a 200 μL volume of vehicle. A control group received the vehicle with no compound and another control group received a single intraperitoneal dose of diminazene aceturate at 10 mg/kg on day 21. The diminazene aceturate clears parasites from the blood, but because it does not cross the BBB, the blood is later repopulated from parasites in the CNS. After dosing, parasitemia was monitored via microscopic examination of tail blood slides until 180 days post-infection. Mice were removed from the experiment when parasites were detected in the blood.

#### *2.10. Chemical Synthesis Procedures*

For compound series 1, 2, 4, and 9, the synthesis methods are in the references. The methods for the synthesis of the initial hit compounds for compound series 3, 5, 6, 7, 8, 10, and 11 are provided in the SupplementaryMaterials (Figures S1–S7). Compounds were purified to>95% purity by high performance liquid chromatography (Varian Prep star system) using a reverse phase C18 semi-preparative column (YMC S5 ODS-A 20 × 100 mm column) and a solvent program of methanol/water or acetonitrile/water with 0.1% trifluoroacetic acid.

#### *2.11. Metabolite Identification*

A 90-min incubation in 0.5 mg/mL mouse liver microsomes preceded liquid chromatography-mass spectrometry/mass-spectrometry analysis for each compound of interest. After incubation, aliquots drawn at various time points in the incubation period were prepared for LC-MSMS analysis alongside a microsome mixture blank containing no compound. A Thermo LTQ Orbitrap Tandem Hybrid Mass Spectrometer combined with an Acquity UPLC system was used for analysis and the MS/MS data were analyzed using MZMine 2.30 software. Peaks from the blank were compared with peaks from the microsome incubation mixture aliquots taken at different timepoints in order to determine whether the detected analyte was a product of the background components or a metabolic product of the compound in question.

#### **3. Results**

#### *3.1. Selection of Hit Compounds*

The set of 1035 active/selective compounds was further curated by a committee of three medicinal chemists involved in the project. The following filters were applied, resulting in 17 selected compounds of different chemical scaffolds: (1) compliance with the Lipinski's rule of 5 (MW < 500, Log P not > 5, not more than 5 H-bond donors, not more than 10 H-bond acceptors) [9]; (2) avoidance of compounds with structural alerts for toxicity (e.g., avoiding alkylating agents, etc.) [10]; (3) avoidance of molecules with > 1 chiral center (to help simplify synthesis and control costs of goods); (4) avoidance of singletons, which meant excluding compounds for which analogs of the same scaffold in the library were inactive as this suggested that further optimization would be difficult to accomplish; and (5) emphasis on chemical tractability as judged by the expertise of the medicinal chemists. The structures, chemical properties, and screening results of the 16 selected hit compounds are shown in Table 1a,b.


*Trop. Med. Infect. Dis.* **2020**

, *5*, 23


\* had poor ionization in MS. B: P ratio data are for close analog (see [11]).

 =

#### *3.2. Screening for Brain Permeability*

The long-term goal of the project was to develop a drug for treatment of HAT including late-stage disease involving the central nervous system. For this reason, we included an early screen in our compound triage process to identify hit compounds that demonstrated at least moderate brain permeability in mice. To assess this, the sixteen hit compounds were injected subcutaneously and mice were sacrificed at 60 min (n = 3 mice) for the simultaneous collection of plasma and whole brains for the quantification of compound concentrations. The ratios of brain to plasma levels are shown in Table 1. Eleven compounds (**1–10, 12**) had brain levels of at least 25% of plasma levels and were considered candidates for development. One compound (**11**) had modest brain permeability of 9%, but was included for further chemistry despite this marginal activity. Four compounds (**13–15, 17**) had minimal concentrations in the brain (< 2% of plasma levels) and were no longer pursued. One compound (**16**) had nearly undetectable plasma and brain levels at the 60-min time point and was also considered unsuitable for further development.

#### *3.3. Hit-to-Lead Optimization*

The program supported three medicinal chemists through a six-year period. Eleven hit compounds representing different scaffolds were pursued with a total of 1539 compounds synthesized. Lead optimization was generally performed on 2–3 scaffolds at a time due to the available manpower. Since the molecular targets or mechanisms of action were unknown for all the hit compounds at the start of the project, new compounds were designed using standard medicinal chemistry principles [12]. For illustration, compounds such as **2** were divided into regions where small changes were discretely introduced by synthetic methods (Figure 1). New compounds were made on a 3–5 milligram scale at >95% purity by HPLC, nuclear magnetic resonance, and mass spectrometry analysis.

**Figure 1.** Medicinal chemistry strategy for optimization of Scaffold 2.

New compounds had *T. brucei* EC50, mammalian host cell cytotoxicity (CC50), and aqueous solubility assayed per our screening cascade (Figure 2). The cut-off values for further advancement are shown. For most compound series, an initial major focus was to improve activity against *T. brucei* cells to an EC50 < 200 nM before concentrating on pharmacokinetic (PK) activity. The screening results were continuously reviewed by the chemistry group to define structure activity relationships (SAR). Changes in different regions of the molecule that resulted in improved activities were combined in subsequent molecules. As compounds were identified with substantially improved potency against *T. brucei* (and retained selectivity compared to mammalian cells), they were tested in a single dose PK assay in mice. Mice were administered compounds by oral gavage at 50 mg/kg and whole blood samples were collected on blotting cards at serial time points. The parameters of maximum blood concentration (Cmax) and the concentration of compound integrated over time (area under curve, AUC) were of primary interest. The general rule was to attain plasma concentrations 10 times above the *T. brucei* EC50 value for at least eight hours. For some series, we performed in vitro microsome stability assays prior to the PK experiments (Figure 3), although our experience was that microsome studies were just as expensive and labor intensive as "shotgun" PK experiments and provided less information, particularly about oral bioavailability. In order to improve the PK profile of compounds, substitutions were introduced that were associated with improved metabolic stability such as introducing fluorine groups to protect possible oxidation sites or introduction of N-cycloalkyl groups (pyrrolidine, piperidine) to decrease oxidative N-demethylation [12,13]. Selected compounds were subjected to metabolite identification studies to define metabolic weak points to inform the design of the next round of compounds to improve the metabolic profile (see Methods). Compounds that matched the selection criteria for antiparasitic activity and PK properties were next subjected to brain permeability studies in mice according to the flow chart (Figure 2). A brain-to-plasma ratio of 0.3 was the minimum value as a go/no-go requirement for further advancement. The compounds passing all of the above testing criteria were upscaled (75 mg synthesis) for efficacy studies in the *T. brucei* acute infection model in mice. Compounds showing cures in the acute infection model were then tested in the more challenging chronic infection model that requires clearing parasites from the CNS. Three compound series remained active in our program, having passed the different levels of our screening cascade (discussed below).

**Figure 2.** Screening cascade.

**Figure 3.** Optimization (**A**) and further optimization (**B**) of Scaffold 1.

When roadblocks prevented progress for specific scaffolds, they were discontinued and replaced with new scaffolds from the list of 17 candidates in Table 1. The reasons for discontinuing various compound series are summarized in Table 2. One candidate scaffold (12) has yet to be pursued.


**Table 2.** Summary of progress in hit-to-lead campaigns for the 12 compound series.


#### *3.4. Active Compound Series (Highlights)*

Three scaffolds (1, 2, and 9) have been developed to the level of lead compounds ready for late preclinical studies [5,8,11,14–16]. The optimization of hit compound 1 is shown in Figure 3. The different regions of the molecule (I-V) are indicated in the center structure. The number of variants made and tested, and the optimal substitution at each region are indicated in the surrounding structures. At the bottom left (Figure 3), the partially optimized compound (HB-175) is shown, which combines the best substituents of regions I, II, and III [5]. The changes included efforts to improve metabolic stability by making the following modifications: (a) replacing the furan amide with mono or di-substituted fluoro pyrrolidine ureas, or with dimethyloxazole amide; (b) by substitution at C6 of the pyridine/pyrimidine ring; and (c) replacing azabenzofurans with imidazopyridines. Subsequent work was dedicated to further improve metabolic stability and brain permeability, leading to the current lead compound PN-302. The changes included altering the core ring system from an imidazopyridine to a triazolopyrimidine [14].

The strategy for optimizing Scaffold 2 is illustrated in Figure 1. Changes that improved antiparasitic activity, metabolic stability, and brain permeability included: (1) substitution of the phenyl group on the left side with 3-fluoropyrrolidine; (2) rigidifying the linker with a benzthiazole as opposed to an alkyl-linked thiazole; and (3) fluorination of the right-sided phenyl group. Of note, the stereochemistry of the fluorine substituent on the pyrrolidine was critical for activity [8]. The illustrated lead compound, 45DAP076, has excellent metabolic stability and excellent brain permeability properties (Figure 4). Importantly, it was shown to have *curative* activity in the chronic *T. brucei* infection model [8], putting it in a category of very few compounds with such high potential for development for HAT.

**Figure 4.** Scaffold 2: Optimization of hit "thiazole" compound to lead compound 45DAP076.

The third compound series that remained active in the program was Scaffold 9, the thiohydantoins (Figure 5). Changes to the central thiohydantoin moiety itself abrogated antiparasitic activity, so this portion of the molecule was held constant. However, through making systematic substitutions in the two terminal rings systems, we identified highly potent inhibitors (EC50 as low as 2 nM) and excellent brain permeability (brain:plasma = 1.68). Compound BA-738 cured mice with acute *T. brucei* infection [11], but only gave partial cures (20% of mice) in the chronic infection model (unpublished). Further optimization will be necessary before advancing this series for late-preclinical studies such as the safety screens shown in Figure 2.

**Figure 5.** Scaffold 9: Optimization of hit "thiohydantoin" compound to lead compound BA-738.

#### **4. Discussion**

This paper summarizes the results of a drug discovery campaign to identify preclinical candidates for HAT starting from a phenotypic screen. Detailed results relating to four of the scaffolds have been published (see references in Table 2), but an overview of the general strategy and complete results has not been previously reported. Some helpful points can be learned by studying the failed scaffolds as well as by studying the successful ones. As indicated, 17 compounds representing distinct scaffolds (Table 1) were selected from the original hit list of 1035 compounds. By definition, these compounds had activity against *T. brucei* cells and thus demonstrated sufficient cell permeability to reach intracellular targets. This feature of the cell-based screen provided a theoretical advantage over biochemical (acellular) screens where hit compounds subsequently have to be tested for (and perhaps optimized for) cell permeability properties. Similarly, the screening protocol included a counter screen against mammalian cells that eliminated compounds with cytotoxicity. A whole-cell cytotoxicity assay identified any type of cellular toxicity and thus was broader than a counter screen against, for example, a mammalian homolog in an enzymatic screen. Thus, these features illustrate the potential advantages of phenotypic screens over target-based screening with the acknowledged disadvantage that the target of activity is unknown. As a result, the hit-to-lead optimization was done agnostically to the target. For most of the project, we chose not to divert time and resources to the effort of target identification. Hit-to-lead chemical optimization was done using standard medicinal chemistry approaches without guidance from protein crystal structures. The target of compound series **1** (the trypanosome proteasome) has subsequently been identified [18], but this did not contribute to designing or synthesizing the current lead compounds (Figure 3). The results of the program to-date are that three of 11 scaffolds (27%) have been optimized to the point of giving cures in the murine model of *T. brucei* infection. More work needs to be done before compounds are brought to clinical trials, but the output of this campaign shows strong promise for delivering clinical candidates, particularly when compared to the results of target-based screening efforts for other microbial pathogens. For example, GlaxoSmithKline reported the outcome of 70 high-throughput screening campaigns (67 target based and three whole cell) for antibiotic development with only five leads delivered, translating to a 7% success rate [19].

As mentioned, the target product profile for HAT dictates that the final drug be administered orally [4], thus we filtered the hit list for compounds that were compliant with Lipinski's rule of five [9]. There were some examples (e.g., compounds 14 and 15) for which the rules were slightly relaxed, although this proved to be disadvantageous as those compounds were terminated due to poor brain permeability. The decision to triage compounds early in the campaign based on brain permeability was done for the following reasons. First, it has been reported that 98% of small molecule drugs do not cross through the BBB [20], meaning the BBB is a major obstacle for developing drugs intended for CNS diseases. Furthermore, the predictive tools to design changes in molecules to improve brain permeability are unreliable when applied to diverse sets of compounds. Thus, we reasoned that it would be helpful to identify compounds with at least moderate brain permeability properties at the start, rather than struggling to try to build in this property later in the process. In order to improve the probability of brain penetrant compounds, we favored molecules with MW < 450 as this has been

shown to be an approximate cut off for BBB permeability [21]. A caveat to the brain permeability studies was that the measurements were of total, rather than free, concentrations of the compounds. Thus, it is possible that some compounds were concentrated in the brain due to high tissue binding (e.g., from lipophilicity), and could be inaccessible to bind targets in the trypanosomes. With only one exception (compound **7** with MW of 469 g/mol), all the compounds that passed the permeability test of a brain to plasma ratio of > 0.25 had MWs < 450 g/mol (Table 1). In contrast, the remaining compounds that failed the permeability test had a MW > 450 g/mol with the exception of **12**, which had a MW of 391 g/mol. Amongst the 11 scaffolds that were further developed, only one (compound series **7**, which had the highest starting MW) failed to advance due to the inability to maintain or improve adequate brain permeability. The results indicate that the strategy to select compounds for good brain permeability early in the process was effective.

As noted above, the synthesis of new compounds was not guided by structure-based drug design. Rather, design and synthesis were guided by standard medicinal chemistry approaches [12]. Specifically, hit molecules were divided into specific regions and substitutions were systematically introduced (Figure 1) to provide analogs for biological testing as per our screening cascade (Figure 2). The results of the biological testing were returned to our chemistry group to generate SAR that informed iterative rounds of synthesis and optimization. As regions of the scaffolds were improved, the various substitutions were combined, often leading to additive or multiplicative improvements. The pharmacological properties of the compounds were evaluated early in the screening process given the importance of optimizing this parameter. The "shotgun PK" experiments provided Cmax and AUC values for initial insights into the absorption, distribution, metabolism, and elimination (ADME) of the compounds. Some scaffolds were also analyzed in microsome stability assays to help track the rates of metabolic degradation by CYP450 enzymes [8,11]. When typical methods such as fluorination did not adequately help with metabolic stability, we performed metabolite identification studies with qualitative mass spectrometry (using incubations of compound in liver microsomes) to understand the molecular targets of degradation so that new analogs could be designed to overcome the weaknesses. In rapid succession, the compounds matching our "go" criteria were then tested for in vivo CNS permeability in mice. At an early stage in the program, we utilized an in vitro trans-well methodology using MDR1-MDCK cells [5] to model permeability across the BBB [6]. Although this method has been widely used, we had specific examples in which the results of the MDR1-MDCK assay were not consistent with the in vivo brain-permeability data (not shown). We also determined that in vivo brain-permeability studies could be efficiently performed with three mice per compound at a single harvest time of 60 min post-dose (5 mg/kg IP). The combination of fewer specimens for mass spectrometry analysis plus greater predictive accuracy made the in vivo experiments preferable to the MDR1-MDCK model in our view.

The reasons for scaffold failures are indicated in Table 2. The most common reason for discontinuation (five scaffolds) was failure to make significant improvement in anti-*T. brucei* activity (EC50). There were no absolute criteria, but if EC50 values of < 200 nM could not be achieved, a compound series was stopped. The number of compounds made in these failed series were 41, 47, 66, 91, and 131, respectively. Obviously, it was a judgement decision as to when to no longer expend resources on a scaffold due to failure to achieve the sufficient target efficacy, but the listed scaffolds that failed typically involved the work of one chemist over a period of approximately one year. The next most common reason for failure (Scaffolds 5 and 7) involved difficulties achieving the desired PK endpoints. This primarily involved the failure to achieve robust plasma exposure of the compounds so that in vivo antiparasitic activity was unlikely to be achievable. The underlying problem with Scaffold 5 was presumably related to poor aqueous solubility (< 1 μM at pH 7.4 and pH 2.0). For Scaffold 7, it was poor metabolic stability that precluded its further development. Finally, Scaffold 4 was discontinued because of a "static" killing mechanism [17]. This became apparent during in vivo efficacy experiments when administration of the compound resulted in temporary suppression of parasites followed by rebound. The "static mechanism

could be recapitulated in vitro in "washout" experiments [17], which was latter incorporated into our screening routine to avoid repeating the problem.

As mentioned, the biochemical targets of the compounds were unknown at the start of the program. Work through collaborators at Genomics Institute of the Novartis Research Foundation (with contributions from the University of Washington group) led to the discovery of the target of Scaffold 1 as the trypanosome proteasome [18]. Research against the closely related *Leishmania* parasites further confirmed that proteasome inhibition was the mechanism of action for this compound series [22]. At this more advanced stage in the program, more resources will now be committed for target identification of the remaining two scaffolds. This work could help with further optimization toward inhibiting the parasite target, for example, by allowing us to develop an enzyme assay or to obtain a crystal structure, but more importantly, the information may be useful for guiding future safety studies on the compound. If the parasite target is identified and it has human orthologs, then directed efforts can be made to optimize compounds that avoid or minimize activity on the human orthologs. The information could also guide future animal studies (and clinical trials) to help predict potential toxicities in mammalian hosts.

#### **5. Conclusions**

Three compound series stemming from a high-throughput phenotypic screen remained viable in this program for HAT drug development. The lead compounds demonstrated curative activity in murine models of *T. brucei* infection. The compounds are now undergoing safety studies as indicated in the bottom of the screening cascade (Figure 2). Dose/response studies in mice are also underway to establish the optimal doses and dosing schedules that will define the pharmacodynamic parameters of the leads. Subsequent investigations will include rat toxicity studies to determine the toxicities resulting from high doses of compounds. For all three series, back up compounds are available in case we encounter significant problems with toxicity or other setbacks. The described drug discovery campaign, conducted in academic centers, remains on track for producing at least one or more late preclinical candidates for HAT.

**SupplementaryMaterials:**The following are available online athttp://www.mdpi.com/2414-6366/5/1/23/s1, Synthetic Schemes and Procedures for the Synthesis of Compound Series 3, 5, 6, 7, 8, 10, and 11. Figure S1: Synthetic scheme for compound series 3. Figure S2: Synthetic scheme for compound series 5. Figure S3: Synthetic scheme for compound series 6. Figure S4: Synthetic scheme for compound series 7. Figure S5: Synthetic scheme for compound series 8. Figure S6: Synthetic scheme for compound series 10. Figure S7: Synthetic scheme for compound series 11.

**Author Contributions:** Conceptualization, M.H.G., F.S.B., and R.R.T.; Methodology, M.H.G., F.S.B., R.R.T., A.B., D.A.P., P.N., Z.H., and J.R.G.; Writing—original draft preparation, F.S.B.; Writing—review and editing, F.S.B., M.H.G., R.R.T., A.B., D.A.P., P.N., Z.H., and J.R.G.; Project administration, M.H.G., F.S.B., and R.R.T.; Funding acquisition, M.H.G., F.S.B., and R.R.T. All authors have read and agreed to the published version of the manuscript.

**Funding:** The research was supported by the National Institutes of Health (grant R01AI106850 and R01AI147504).

**Acknowledgments:** For the chemical synthesis, we are thankful for the contributions by Hari Babu Tatipaka, Naveen K. Chennamaneni, Praveen K. Suryadevara, Elizabeth Raux, Moloy Banerjee, Amit Thakkar, Neil R. Norcross, Patrick Weiser, Kishore Kumar, GV Reddy, Stanislav Bakunov, Svetlana Bakunova, and Daniel G. Silva. For the analytical chemistry, we thank Joshua McQueen. For contributions to in vitro and in vivo biological experiments we thank Matthew Hulverson, Ranae Ranade, Uyen Nguyen, Sharon Creason, Nicole Duster, Jennifer Arif, Nora Molasky, and Aisha Mushtaq. Finally, we acknowledge contributions from David Boykin.

**Conflicts of Interest:** The authors declare no conflicts of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Review* **The Trypanosomal Transferrin Receptor of Trypanosoma Brucei—A Review**

#### **Christopher K. Kariuki 1,2,\*, Benoit Stijlemans 1,3 and Stefan Magez 1,4,\***


Received: 9 September 2019; Accepted: 25 September 2019; Published: 1 October 2019

**Abstract:** Iron is an essential element for life. Its uptake and utility requires a careful balancing with its toxic capacity, with mammals evolving a safe and bio-viable means of its transport and storage. This transport and storage is also utilized as part of the iron-sequestration arsenal employed by the mammalian hosts' 'nutritional immunity' against parasites. Interestingly, a key element of iron transport, i.e., serum transferrin (Tf), is an essential growth factor for parasitic haemo-protozoans of the genus *Trypanosoma*. These are major mammalian parasites causing the diseases human African trypanosomosis (HAT) and animal trypanosomosis (AT). Using components of their well-characterized immune evasion system, bloodstream *Trypanosoma brucei* parasites adapt and scavenge for the mammalian host serum transferrin within their broad host range. The expression site associated genes (ESAG6 and 7) are utilized to construct a heterodimeric serum Tf binding complex which, within its niche in the flagellar pocket, and coupled to the trypanosomes' fast endocytic rate, allows receptor-mediated acquisition of essential iron from their environment. This review summarizes current knowledge of the trypanosomal transferrin receptor (TfR), with emphasis on the structure and function of the receptor, both in physiological conditions as well as in conditions where the iron supply to parasites is being limited. Potential applications using current knowledge of the parasite receptor are also briefly discussed, primarily focused on potential therapeutic interventions.

**Keywords:** trypanosomosis; iron; transferrin; transferrin receptor; nutritional immunity; flagellar pocket

#### **1. Trypanosomes and Their Need for Iron during a Mammalian Infection**

Iron is an essential requirement for life [1–3]. Iron's biological utility lies in its cycling between two oxidation states, namely ferrous (Fe2<sup>+</sup>) and ferric (Fe3<sup>+</sup>) [4]. Thus, iron can serve as redox catalyst, which accepts and donates electrons [4]. As a result of this redox capability, once absorbed from the environment via the mammalian duodenum, iron is not circulated freely in mammalian tissues, as it readily catalyzes conversion of H2O2 into toxic free radicals via the Fenton reaction [1,5].

Under conditions of neutral pH and high oxygen tension of most physiological fluids, such as mammalian serum, iron exists predominantly in its ferric (Fe3<sup>+</sup>) form [3]. Given its high hydrolytic propensity, under these conditions, ferric iron (Fe3+) in excess of 2.5 <sup>×</sup> 10−<sup>18</sup> M readily polymerizes, resulting in an insoluble and bio-inert form [6]. Therefore, to keep ferric iron (Fe3<sup>+</sup>) soluble, bio-available, and render it non-toxic, vertebrates such as mammals store and transport iron via specific iron sequestering molecules [3,6,7]. Storage of iron is achieved using two formats; in an soluble form as a mobilizable reserve by ferritin and an insolubly form as hemosiderin [3,8]. Transport of ferric iron

(Fe3<sup>+</sup>) in the mammalian serum is usually as a tight, but reversible association, with an abundant serum carrier protein family, the transferrins [3,7,9,10].

Serum transferrin/serotransferrin (Tf) is the transporter of ferric iron (Fe3<sup>+</sup>) in the blood serum of vertebrates, acting as the connection between the ferritin storing hepatocytes to the diverse cellular population of the vertebrate body [9,10]. Serum transferrin, i.e., a 80 kDa glycoprotein, has been structurally resolved, indicating a bi-lobed tertiary structure (N- and C-lobes) with a short connecting loop between them as well as possessing two domains per lobe, with the ferric iron (Fe3+) binding sites located within the inter-domain clefts of each lobe [11]. In the presence of an anion, e.g., bicarbonate or carbonate, and at a physiological serum pH range, serum transferrin can bind either mono- or di-ferric iron atoms, transforming from apo- (iron-free) to holo- (iron-laden) transferrin [3,11]. Though both lobes bind ferric iron (Fe3<sup>+</sup>), there is a difference in binding capability as the C-terminal lobe binds Fe3<sup>+</sup> more tightly and releases it more slowly [12]. Mammalian serum Tf is expressed in the liver, central nervous system (CNS), reproductive organs, spleen, and kidneys [13].

The rich resource of bio-available iron is a prized target for parasites, particularly those that reside in the bloodstream, such as the trypanosome species, *T. brucei*, thus leading to the diseases, human African trypanosomosis (HAT) and animal trypanosomosis (AT).

Animal trypanosomosis (also known as animal trypanosomiasis) is a parasitic disease of veterinary importance in the tropical world [14]. From its cradle in Africa, through the steppes of Asia to the far ends of the Americas, the disease has devastatingly negative economic and societal impacts [15,16].

In domestic livestock, AT is a wasting disease assigned by various names such as '*Nagana*', '*Derrengadera*', '*Dourine*', '*Mals de coits*', or '*Surra*', depending on the causative *Trypanosoma* species [17]. Five of the most important species, namely, *T. vivax*, *T. evansi*, *T. congolense*, *T. equiperdum*, and *T. brucei* cause AT in all domestic animals [16]. These trypanosome species are considered heteroxenous (Figure 1) [18]. Of note, the *T. brucei* (sub-genus *Trypanozoon*) clade can be further divided into strictly animal infecting parasites (*T. brucei brucei*) and parasites able to infect also humans and higher primates, namely the zoonotic *T. brucei rhodesiense* and the anthroponotic *T. brucei gambiense* [19,20]. This is attributed to the fact that these latter parasites acquired the ability to resist trypanolytic molecular complexes, expressed by humans and higher primates as part of the innate immune system [20–24]. Due to this reason, and its implications for human medicine, the *T. brucei* clade has received greater attention and is better characterized, despite being it having a lower, though still potent, worldwide impact on veterinary economy, than for example *T. congolense* (Sub-genus *Nannomonas*) or *T vivax* (Sub-genus *Dutonella*) [25].

**Figure 1.** The trypanosomes' lifecycle. (**A**) Cyclical transmission of *T. brucei* and *T. congolense* (causing 'Nagana') occurs in the Tsetse fly (*Glossina* species), in grey is a representation of the procyclic form and in red is the representation of the bloodstream form (BSF); (**B**) mechanical transmission via tabanids

leading to trypanosomosis caused by *T. vivax* and *T. evansi*; (**C**) sexual transmission occurs in equines for *T. equiperdum* during the course of dourine [26–28].

By parasitizing the haemo-lymphatic environment, trypanosomes must reconcile two seemingly conflicting requirements, namely, to avoid the immune responses of the mammalian host by rapid variation of their plasma membrane as well as efficiently acquire potentially scarce nutritive resources from their environment via the same plasma membrane [29].

#### **2. Expression Sites; the Trypanosomal Swiss Army Knife for Host Adaptation**

Trypanosomes exemplify the general survival strategy of phenotypic variation, a mechanism by which diverse parasitic organisms, from viruses to eukaryotes, contain a subset of contingency genes hypermutating as a rapid adaptation to hostile or changing environments [30]. This is the classically described paradigm of antigenic variation of the major trypanosome surface glycoproteins, the variable surface glycoprotein (VSG) [31,32]. Antigenic variation in trypanosomes occurs when a subset of a trypanosome population randomly switches the expression of a particular set of VSG genes, thus displaying a different surface coat of VSGs [29,30,33]. Contrary to avoiding the mammalian hosts' immune system, VSGs are highly immunogenic, with surface epitopes that are highly recognizable by the mammalian hosts' immune system [29]. When the immune system clears the parasite population bearing the recognized VSG by the humoral response, there is the emergence of sub-population of the haemo-protozoans with a different surface coat of VSG homodimers, against which the immune system has to prepare another humoral response [30]. The ebb and rise of different populations of trypanosomes is reflected on the parasitemia pattern observed microscopically and confers to the parasite an advantage of establishing a controlled but chronic host infection [29].

The active *T. brucei* spp. bloodstream form (BSF) VSG gene is obtained from an arsenal of more than 1500 VSG genes, most of which are pseudogenes in sub-telomeric silent arrays [34–36]. For expression, the VSG gene has to be relocated to a devoted genomic environment, aptly termed the expression site (ES), at the telomeric regions of one of the large chromosomes [37–40]. In the ES, specifically the bloodstream form expression site (BES), transcription occurs in a polycistronic manner, with VSG genes always the last gene of the unit, separated from the other genes, by 70 bp repeats (Figure 2A) [39,41,42]. The polycistronic nature of transcription allows a tightly regulated expression of the VSG gene from only 1 of the 20 telomeric expression sites (ESs), in conjunction with associated proteins within the ES, namely the expression site associated genes (ESAGs) [35,43]. The polycistronic mRNA is transcribed from a highly conserved BES promoter, that has also been shown to be sensitive to temperature changes, and which is considered a specificity signal that triggers the activation of the BES upon encountering the bloodstream of a mammalian host [42].

With only one type of VSG being expressed at a time, a portion of the trypanosome population is guaranteed invisibility from the host's immune system at any given time [35,43–47]. Variation within the VSG is achieved by homologous DNA recombination such as gene conversion, targeted to the active VSG ESs (Figure 2B,C) or by either transcriptional switching (in-situ activation and in-activation (Figure 2D) between the VSG ESs) [40].

**Figure 2.** Modes of variable surface glycoprotein (VSG) gene switching. (**A**) The expression site (ES) of VSG including the proximal promoter, expression site-associated genes (ESAGs), the VSG, and their associated repeats. Transcription occurs in a polycistronic manner, with the ESAGs 7 and 6 being most proximally located and the VSG most distally located on the expression sites. (**B**) Mechanism of gene conversion: A VSG gene conversion event occurs when a VSG gene from a silent ES is copied (via homologous recombination) into an active ES where it is expressed. (**C**) Mechanism of segmental gene conversion: Various VSG gene recombination events occurring in the silent ES, leading to formation of a novel and mosaic VSG gene, which is copied via homologous recombination into the active ES. (**D**) Mechanism of transcriptional switching (in-situ (in)-activation): A non-recombination event occurs that activates a new (previously silent) ES while inactivating a previously active ES [48].

This immune evasion tactic also serves a rather utilitarian purpose for the parasite. As stated previously, in addition to successfully evading the host's immune system, the extracellular parasite must somehow combine the antigenic variation of its homogenous and over-arching surface coat, with the uptake of different substrates that it requires for its survival [49]. The parasite approaches this possible problem using a specific set of genes, namely the expression site associated genes (ESAGs). Within the *T. brucei* ES's, the BSF trypanosome also has a more limited repertoire of ESAGs (Figure 3) [43]. Most of the ESAGs encode predicted proteins that contain N-terminal signal sequences as well as putative hydrophobic membrane spanning segments, indicative of surface exposed proteins, such as the integral membrane proteins such as the receptor-like transmembrane adenylyl cyclase (ESAG4), a surface transporter (ESAG10), and the glycosyl phosphatidyl inositol (GPI)-anchored heterodimeric trypanosomal transferrin receptor (ESAG6/7) and serum-associated resistance antigen (SRA) [50–52]. Several ESAGs have been found to belong to multigene families including pseudogenes and members that are not transcribed within the ESs, aptly named "genes related to ESAGs" (GRESAGs) [51]. In total, within the *T. brucei* clade, sequencing has revealed about 12 polymorphic genes comprising the ESAGs, and approximately 20 different variants of each ESAG [43,45,46,51,53]. In comparison, there are relatively fewer homologs or orthologs in the closely related *T. congolense* or *T. vivax* [54]. An analysis of the cell-surface phylome for the trypanosomes revealed that for some ESAGs, e.g., ESAG1, there are neither homologs nor orthologs, indicating recent innovation by the *T. brucei* clade [54]. Other ESAGs, for example the ESAG6 and 7, encoding the trypanosomal transferrin receptor and that are considered essential for bloodstream iron scavenging in the *T. brucei* spp., are missing from *T. vivax* but are present

in *T. congolense*, which is indicative of a more recently shared ancestry between *T. congolense* and *T. brucei* [54].

**Figure 3.** The ESAG 6 and 7 genes (forming the pESAG 6 and 7 heterodimeric transferrin receptor) are transcribed as part of the polycistronic VSG mRNA. The polycistronic VSG mRNA is transcribed from the active subtelomeric expression site (ES) (only 1 of the 20 available ES is active at a time). The gene products appear to have a similar structure except for the glycosyl phosphatidyl inositol (GPI) peptide on ESAG 6 [35,45–47].

Some genes in the expression site, i.e., ESAG 6, 7, and SRA, share an evolutionary origin with the VSGs, and may thus confer an increased capacity for the parasite to adapt to various mammalian hosts [51,52,55]. Given that some of the ESAGs are involved in substrate capture, it therefore seems plausible that the transcriptional switching between multiple expression sites would offer the parasite antigenic variation for these minor surface proteins [40,50].

More importantly, the transcriptional switching of BESs would also allow the selection and expression of the appropriate ESAG 6 and 7 genes for efficient capture of the requisite host transferrin molecule, which would enable the trypanosome to adapt not only to the different mammalian host range that is available to the *T. brucei* spp., but as well to the mammalian hosts' "nutritional immunity" [56,57].

#### **3. The Trypanosomal Transferrin Receptor: A Structural Review**

As indicated previously, ESAG 6 and 7 genes (encoding the trypanosomal heterodimeric transferrin receptor) are transcribed in a polycistronic mRNA together with the current VSG from an upstream promoter (Figure 3) [58,59].

However, given that both ESAG 6 and 7 genes are situated nearest to the ES promoter site with the end of ESAG 6 being approximately +5.3 kB from this promoter, there is a low but detectable transcription occurring from 'inactive' ESs in bloodstream form (BSF) trypanosomes [60,61]. It has been estimated that up to 20% of ESAG 6 mRNA originates from 'inactive' ESs [60]. This promoter-proximal position of the two genes provides the parasite with a flexibility in the regulation of the genes, providing a competitive edge, especially during periods of limited transferrin uptake, e.g., during the switch to another host.

Different but homologous pESAG6-7 heterodimers encoded by the different ESs are present in the BSF trypanosomes (Figure 4), differing in sequence identity by only 1–10% [60]. The proteins pESAG6 and 7 have been shown to be synthesized individually, containing N-terminal signal sequences that are not present in the matured protein forms [62].

In *T. brucei*, the two genes (ESAG 6 and 7) give rise to heterogeneously glycosylated proteins between 50–60 kDa and 40–42 kDa, respectively [63,64]. Only heterodimeric complexes (with a 1:1 stoichiometry) of the products from the two genes form a functional *T. brucei* trypanosomal TfR, indicating that there is a combination of elements, specific to each subunit, that are required for the

transferrin binding site [65,66]. Though the protein subunits are glycosylated, the pESAG 6/pESAG 7 heterodimer can function just as well without this post-translational modification [64]. Small amino acid switches in the surface exposed loops of the pESAG 6-pESAG 7 complex, forming the transferrin binding site, brings about differences in affinity of the various TfRs for their respective Tf ligands in different hosts (ranging from 2 nM to 1 μM), all lying within the reported physiological range of Tf (30–40 μM) in the mammalian serum [67]. This allows rapid adaptation of the parasites' Tf scavenging capacity in different hosts, particularly in the presence of host antibodies [68–70]. This high affinity for Tf combined with the rapid recycling of the TbTfR enables the bloodstream parasites to actively compete, despite efforts by the mammalian host immune system, for the limited substrate until a novel higher affinity TfR is expressed [61].

**Figure 4.** The products of ESAG 6 gene (pESAG6) and ESAG 7 (pESAG7) in *T. b. brucei* Strain 427 (UniProt ID Q8WPU1\_9TRYP and Q8WPU2\_9TRYP respectively) are similar along their N-terminus, differing only at their C-terminal end (GPI anchor peptide).

The ESAG 6 and 7 products (pESAG6 and pESAG7, respectively) are nearly identical in sequence, especially in their N-termini, which also contain the ligand binding sites, differing only in their C-termini [65]. The two proteins also share significant homology (20% identity and 60% similarity) with an A domain type VSG N-terminus, indicating a possible structural design requirement for accommodation within the dense VSG protein coat as well as an evolutionary relationship between the two ESAGs and the VSG N-terminal domain [51,65,66,71]. The pESAG6 and 7 appear to have a number of the structurally conserved features of the N-terminus of the VSG class A, whereby, especially, the ESAG 7 gene appears to be a VSG gene conversion domain [66]. Salmon et al. [63] showed that the ESAG 7/6 can be aligned to VSG (sharing significant similarity) and hypothesized that the binding sites are most likely to occur on the surface exposed loops of the heterodimeric protein (Figure 4), i.e., in the dashed boxes (where they align with the surface exposed regions of VSG gene products). Sequence secondary structure predictions e.g., JPRED indicates that these regions are most likely in loops. Prediction of ligand binding regions using the Kolaskar and Tongaonkar method [64] indicates their surface accessibility (again confirming the VSG linkage). In fact, a VSG-based chimeric TfR has been constructed and shown to effectively bind Tf [66]. This was achieved by grafting the C-termini of either ESAG 6 or 7 with the N-terminus of a MiTat 1.5 VSG [66]. The heterogeneously expressed chimeric VSG-ESAG 6 and VSG-ESAG 7 gave a heterodimeric receptor that bound Tf equally well

as the native pESAG6/7 heterodimer (1.2 ± 0.27 μM vs. 0.97 ± 0.36 μM, respectively) in *Xenopus* oocytes [65,66].

The pESAG6 has a hydrophobic C-terminus that is eventually replaced by a GPI-anchor, making it the plasma-membrane bound member for the heterodimeric TfR [63,65,66,72]. An alignment of publicly available ESAG 6 gene sequences reveals the gene's homology within the *T. brucei* subspecies (Figure 5A), particularly indicative of the reported close phylogenetic relation between *T. brucei* and *T. evansi* [25,73]. In contrast, the pESAG7 has no such modification, therefore it is hypothesized to bind non-covalently with the ESAG6 [63]. Between the two proteins, there are differences in residues within the stretches forming the ligand binding site, i.e., positions 205–215 and 223–238 of pESAG6 and 7, respectively [65,66]. These amino acid stretches, especially on pESAG6, have been mapped to surface exposed loops by modeling on the resolved N-terminal VSG crystal structure (Figure 5B) [66]. Switching, by site-directed mutagenesis, of key amino acid within these stretches of pESAG6 and 7, resulted in a predicted enhancement of the Tf binding capacity [66]. Further proof was obtained from the site-directed mutagenesis of residues immediately outside each of the four domains, which resulted in loss of the Tf binding capacity, predictably due to loss of the surface exposed loop structure [66]. Despite the endeavors to model the TfR by various groups [66,74], there has been no actual structure (via X-ray crystallography or NMR) resolved yet, not for the pESAG6 neither the pESAG 7 nor the pESAG6/7 heterodimer.

**Figure 5.** (**A**) The ESAG 6 gene is homologous within the *T. brucei* subspecies. A protein BLAST query for ESAG6 (which is the GPI-anchored partner of the ESAG7/6 heterodimer) revealed approximately 100 sequences from *T. brucei*, *T. evansi*, and *T. congolense*. The results were then assembled to make an alignment and a phylogenetic tree. Alignment was done using the program MUSCLE, and the phylogenetic tree was

constructed via Neighbor joining algorithm with a JTT (Jones-Taylor-Thornton) protein substitution model: (**B**) Modelled ESAG6 is similarly structured between the different *T. brucei* subspecies i.e., I. *T. b. rhodesiense*, II. *T. b. gambiense*, III. *T. b. brucei*, IV. *T. evansi*, and V. *T. equiperdum*. (VI.) An overlay model built from a structural alignment of all the models indicates that the predicted binding site is on surface exposed loops (red ribbons) and occurs similarly in all the proteins. Helices are denoted by the blue ribbons, while the brown ribbons denote loops and the magenta ribbons denote beta sheets. Modelling was done on the *T. b. brucei* VSG ILTat 1.24 (PDB ID: 2VSG) using the SWISS-MODEL homology modelling server (https://swissmodel.expasy.org/) [75–79] with the 2.7 Å X-ray diffraction structure of ILTat 1.24 (2VSG.pdb) as a template [80].

#### **4. Fishing from a Hole; the Flagellar Pocket and the Quest for Iron**

Iron is already a tightly controlled resource within the mammalian body fluids, with iron chelation molecules, i.e., serum transferrin (in blood and lymph) and lactoferrin (in external secretions), restricting the amount of bio-available ferric iron (Fe3<sup>+</sup>) in body fluids to about 10−<sup>18</sup> M [81,82].

Iron availability is a key component employed by mammals to minimize the parasite burden and increase the hosts' fitness [83]. By coupling of the mammalian immunosurveillance apparatus to iron metabolism, immunocompetence is associated with iron regulation [56,83]. Thus, the presence of parasites, indicated by their concomitant biochemical signals, signals a hazard to the mammalian system leading to the triggering of the acute-phase immune response [56]. A consequence of this is the sequestration of iron, thus limiting the bio-availability of this essential nutrient for circulating pathogens, a host-defense strategy known as 'nutritional immunity' [56]. Additionally, this also serves in strengthening specific immune effector mechanisms including the proliferation and functionality of immune cells, activation of cytokines, nitric oxide (NO) formation, activation of cellular proteins/peptides, and hormones that are dependent on iron availability [84,85].

Therefore, to sidestep nutritional immunity and obtain ferric iron (Fe3<sup>+</sup>) from their hosts body fluids, parasites have to either compete against these chelates by devising their own iron chelation molecules, e.g., bacterial siderophores, or cleave the mammalian iron chelates by releasing proteases, e.g., bacterial reductases, or scavenge for these chelates by using specific receptors, e.g., trypanosomal transferrin receptors [63].

Transferrin (Tf) acquisition is the main route of iron uptake for BSF trypanosomes, particularly *T. brucei* spp., which are exclusively extracellular within the bloodstream and which has been the model organism for studying Tf uptake [13,57,65,86–88]. Tf uptake has been shown to be saturable, indicative of receptor mediated endocytosis (RME) with the ligand in this case, holo-/apo-Tf, being specifically competed out from its receptor, the trypanosomal TfR [57,65].

Binding of Tf and recycling of the trypanosomal TfR occurs in a process quite different to that observed in mammalian cells (Table 1) [7,89]. The main route of iron uptake in trypanosomes is localized within the trypanosomal flagellar pocket (FP) [57,90].

The trypanosomal FP, a membrane invagination surrounding the base of the flagellum, is a specialized organelle with multiple roles in the trypanosome [91,92]. This region is uniquely excluded from the sub-pellicular microtubule array under the parasites' plasma membrane [93,94]. The FP is also delineated from the rest of the plasma membrane by the FP collar, an electron-dense annulus, without which the FP is lost [93,95]. This collar encloses the FP lumen, a space filled with a carbohydrate-rich matrix with a poorly defined composition and unknown function [91]. The FP is the main turnover point for parasite nutrition [91]. As the only site of exo- and endocytosis by the trypanosome, the FP is part of a multi-organelle intracellular complex comprised of the Golgi complex, the endoplasmic reticulum as well as secretory and endocytic organelles, making it an important cog in the trypanosome's virulence and protein trafficking [92,95]. The efficiency of the FP protein trafficking is comparable to that of mammalian cells, which is quite remarkable, given that it covers about 2–5% of the total surface of the trypanosome [88]. In BSF trypanosomes, the FP is a site of high protein trafficking, with infectivity tied closely to a high rate of endocytosis [95]. Efficient nutrient scavenging occurs in the FP via selective retention of many of the invariant or variant host-associated nutrient receptors within its lumen by yet-unknown mechanisms but, mostly postulated to be the dynamic result of the high endocytic rate [95]. It is within the FP that the trypanosomal TfR binds iron laden transferrin (holo-Tf) as well as iron free transferrin (apo-Tf) as a GPI-anchored heterodimeric complex (Figure 6) [57,86].


**Table 1.** Features of the transferrin receptors (TfR) of *T. brucei* and human cells. [86].

Reprinted from Steverding, D. The transferrin receptor of Trypanosoma brucei. Parasitol. Int. 2000, 48, 191–198 doi:10.1016/S1383-5769(99)00018-5, with permission from Elsevier.

**Figure 6.** The trypanosomal transferrin receptor is located in the flagellar pocket and is internalized via receptor mediated endocytosis. The transferrin receptor of T. brucei (TbTfR) binds one molecule of Tf [86]. The parasite internalizes host transferrin by transferrin receptor-mediated endocytosis, facilitated by TbRab5. Holo- and apo-transferrin are bound at pH 7 and released at pH 5. Iron is incorporated into the cell cytosol and Tf is degraded by the lysozyme [19]. **Key: E7**/**E6**, heterodimeric transferrin receptor; **ellipse with Fe**, holo-transferrin; **empty ellipse**, apo-transferrin; **FP**, flagellar pocket; **FL**, flagellum; **V**, endo- and exocytotic vesicles; **EN**, endosome; **LY**, lysosome; **CY**, cytosol. Reprinted from Steverding, D. The transferrin receptor of Trypanosoma brucei. Parasitol. Int. 2000, 48, 191–198 doi:10.1016/S1383-5769(99)00018-5, with permission from Elsevier.

The TfR–ligand complex is endocytosed via a clathrin-dependent pathway [12,96]. Invagination of clathrin-coated vesicles leads to internalization of the receptor–ligand complex and subsequent discharge into the intracellular tubular system [95,96]. The endocytosis process has been hypothesized to involve the cleavage of the intracellular GPIs by the GPI-phospholipase leading to production of DAG and inositol-phosphoglycan [97–99]. DAG is an intracellular second messenger for signaling in eukaryotes [97]. Its role in stimulating endocytosis of Tf in the BSF trypanosome is proposed to be an adaptation of *T. brucei* to compete effectively with the mammalian host cells for Tf, as it does not have the same effect in mammalian Tf endocytosis [97]. Binding of DAG to its cognate receptors leads to their activation with the subsequent downstream activation of the protein tyrosine kinase (PTK)-dependent DAG signaling pathway [97]. The PTK is responsible for the phosphorylation and activation of the other proteins of the endocytic system including clathrin, actin, adaptins, and other components [97].

Once in the endosome, the acidic pH (6.5–4.5) enhances the release of the iron bound to the holo-Tf:TfR complex leading to formation of apo-Tf:TfR complex (Table 2) [57,86]. However, at the low (acidic) pH, the trypanosomal TfR, in contrast to the mammalian TfR, loses affinity for apo-Tf [19]. The apo-Tf is in turn delivered to the lysosomes for proteolytic degradation by the *T. brucei* cathepsin B-like protease (TbcatB) [100,101]. The resulting fragments are transported out of the cell via the TbRAb11 positive recycling vesicles [102]. TbRAB11 is specifically present in endosomal structures with recycling cargo molecules [102].


**Table 2.** Kd-values of ligand–receptor complexes for apo- and holo-transferrin at pH 7 and pH 5 [86].

Reprinted from Steverding, D. The transferrin receptor of Trypanosoma brucei. Parasitol. Int. 2000, 48, 191–198 doi:10.1016/S1383-5769(99)00018-5, with permission from Elsevier.

The iron released from holo-Tf is initially converted from Fe3<sup>+</sup> (insoluble) to Fe2<sup>+</sup> (soluble) via two ferric reductases, i.e., a cytochrome b561-type (Tb927.6.3320) and a NADPH-dependent flavoprotein (Tb11.02.1990), before being imported into the cytoplasm in cooperation with the divalent cation transporter, *T. brucei* Mucolipin-like protein (TbMLP) [105]. The TbMLP, a protein of the endocytic system, is expressed both in the bloodstream and insect stages of the parasite, with high expression in the lysosomes [105]. However, other iron transport mechanisms such as ferric reductase and putative divalent metal transporters containing ZIP domains might also be involved [105,106]. Once in the cytosol, it is presumed that excess iron is stored in a storage compartment and released when cytosolic iron levels decline, which in turn is a possible signal for the TfR upregulation [107].

During periods of acute iron scarcity (which has only been documented using in vitro cultivated trypanosomes), mirroring those expected during the switch from one host to another or during chronic infection in the mammalian host, it has been shown that the TfR can be found outside of the FP [107]. The receptor then exists as islands within the VSG coat and was initially hypothesized to allow the parasite to utilize a wider surface area to capture any Tf [61,71]. This spillover is precipitated by an approximately 3- to 5-fold upregulation of the transcription of the receptor [61]. The upregulation of the *TfR* gene is not triggered by an increase in serum concentrations of the apo-Tf, as the receptor does not discriminate between holo- and apo-Tf [68,108]. Rather, upregulation has recently been shown to be mediated via the 3' untranslated region (UTR) of the *TfR* gene that gets activated upon a reduction in the cytosolic iron concentration [107,109]. This signal triggers the upregulation long before the depletion of the iron stores in the parasite, thus allowing division of the parasite with the subsequent cycling and selection of a suitable high affinity TfR from the ESAGs [107]. The TfR has also been modelled to spread out from the surrounding VSG molecules, which previously was presumed to allow adequate contact and capture of Tf [74]. However, this hypothesis has recently been contradicted by evidence that the TfR outside of the FP is not functional, as it is composed of an ESAG6 homodimer, rather than an ESAG6/7 heterodimer [110]. The GPI valence in trypanosomes has been shown to be a critical determinant of intracellular sorting, with molecular complexes with two GPIs (GPI2) being

trafficked out of the FP, one GPI (GPI2) being retained within the FP, and non-GPI-anchored complexes being degraded in the lysosome [110]. Given that ESAG6 is the GPI-anchored partner in the complex, when over-expressed as homodimer during iron starvation, its GPI valence allows its escape from the FP [110].

#### **5. The Trypanosomal Transferrin Receptor as A Target for Chemotherapeutic Purposes**

Given the differences between the mammalian and the trypanosomal iron uptake in the form of Tf, it seems feasible to selectively target this pathway for chemotherapeutic purposes [106]. The mammalian host-trypanosome interaction is characterized by a macrophage hyper-activation, which through enhanced erythrophagocytosis cascades to anemia [111,112]. This iron deficiency represents a key challenge to the BSF trypanosomes, a situation further exacerbated by the release of various cytokines and hormones, such as the hepatocyte-derived hepcidin [113]. Hepcidin helps to further accentuate iron deficiency, by down-regulating the iron-exporting ferroportin-1, limiting the contribution of cellular iron to the blood [56,113]. As the parasite has to survive in already limiting conditions, interfering with its otherwise efficient scavenging of iron from the host, may represent a new strategy for treatment of AT [84].

The use of iron chelators to deprive parasites of iron and therefore limit the parasites' growth within the mammalian host represents an interesting chokepoint [56,63,106,114]. In another haemo-protozoan parasite, *Plasmodium*, the iron metabolism has been a successful in vivo target for many compounds, with iron chelation being a consistently applied therapy [115–117]. One iron chelator in particular, a desferrithiocin analogue, has been already applied in human trials for mitigation of iron mediated damage in transfusional iron overload [118,119]. While not curative in nature, the chelation of iron from the incorporation into apolipoproteins or even into Fe3+-containing enzymes, e.g., ribonucleotide reductase, by chelators such as deferoxamine, has been shown to limit in vitro the growth of trypanosomes [104,120]. Given that the BSF parasite population needs to rapidly divide to keep up with the high rate of clearance by the immune system, this approach seems useful [121]. However, the use of these chelators is limited by both their water solubility, as well as their cytotoxicity, when applied in vitro on mammalian cells [120]. The potential for these drugs, however, is in the reduction of these unsuitable traits, e.g., by making lipophilic iron-chelating agents with reduced toxicity and in using these in conjunction with currently available anti-trypanosomal drugs [84,120].

Blocking the uptake of Tf by selectively targeting the trypanosomal TfR has been proposed severally as an alternative to controlling parasitemia in mammals [58,84]. It has already been shown that the TfR can be targeted by antibodies, with the main problem being delivery of the antibodies to the FP in sufficient quantities to achieve a therapeutic effect [57]. The question then arises whether a smaller targeting molecule, such as a single-domain antibody (sdAb) could work to target the receptor and block it. Such sdAbs do exist as nanobodies (Nbs), i.e., nanometer-sized camelid derived single-domain antibody fragments, which have been used for such cryptic targets as beta-lactamase enzyme active sites amongst other targets [122–124]. In addition, the specificity of nanobodies for the trypanosomes' unique and cryptic sites on their surface proteins has been previously applied to deliver toxic molecules into trypanosomes [125]. Hereby, a nanobody against the conserved region of VSG and conjugated to a trypanolytic molecule, i.e., human ApoL1, allowed to target and kill *ˆT. b. brucei* and *T. b. rhodesiense* parasites [125]. In addition, nanobodies against the variable part of VSG have also been successfully raised that block the endocytic machinery or target drugs to the trypanosome thereby causing its lysis and death [126,127]. Despite these properties, no experiments with nanobodies have been published with regards to the Tf uptake. Though Tf uptake is a potentially lucrative target, the BSF parasite in vivo would be surrounded by adequate ligand, even in the anemic state, thus requiring a considerable dose of a very high affinity nanobody to block, appreciably, the trypanosomal TfR. In addition, the rapid in vivo half-life of nanobodies, due to their size (15 kDa), which is below the renal cut-off (>50 kDa), might further hamper their applicability [122]. Yet, this might be circumvented by increasing their in vivo retention time by generating half-life extended constructs [128–130].

#### **6. Conclusions**

Understanding the trypanosomal transferrin receptor promises to provide unique insights into the trypanosome physiology. As an adaptation to the iron-scavenging lifestyle of the parasite, this molecular complex represents an interesting interface between the host and parasite [106]. Though there has been great progress made in unravelling the working mechanisms of this molecule, there has not been a diagnostic application based on the receptor, nor has there been development of chemotherapeutic agents targeting this essential parasite molecule. There are also gaps in understanding the function of this receptor particularly in the in vivo disease state. This is mainly attributed to the lack of suitable models to fit this context. It is of interest to unravel, in particular, the relevance of the TfR for the BSF trypanosome in this context (in vivo disease state), especially when faced by the fact that the *T. vivax* parasite survives just as well in the host bloodstream without any homologue of the *T. brucei* or *T. congolense* TfR.

Nonetheless, the TfR is a necessary element for the successful *T. brucei* parasitization of the mammalian host. It is also a relatively invariant molecule in comparison to its homologue, the VSG. Specifically targeting this heterodimeric molecule, at least for chemotherapeutic purposes, provides a novel way to deliver trypanocides or even to slow down parasite growth. This would allow parasite clearance by the infected mammalian host's immune system, thus controlling parasitemia as well as the inflammation that ensues.

**Author Contributions:** All authors contributed equally to the manuscript.

**Funding:** The authors acknowledge the financial support of the Interuniversity Attraction Pole Program (PAI-IAP N. P7/41, http://www.belspo.be/belspo/iap/index\_en.stm), and the FWO G015016N (Characterization of the cellular and molecular mechanisms leading to the development of inflammation-driven immunopathologies in African Trypanosome infection). BS is a research fellow supported by the Strategic Research Programs: targeting inflammation linked to infectious diseases and cancer (Nanobodies for Health, SRP3) and Molecular Imaging and TARgeting of Macrophages in Inflammation (ITARMI, SRP47).

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).
