Optimization of Dipeptidic Inhibitors of Cathepsin L for Improved Toxoplasma gondii Selectivity and CNS Permeability

Abstract

The neurotropic protozoan Toxoplasma gondii is the 2^nd leading cause of death due to foodborne illness in the US, and has been designated as one of five neglected parasitic infections by the Center for Disease Control and Prevention. Currently, no treatment options exist for the chronic dormant-phase Toxoplasma infection in the central nervous system (CNS). T. gondii cathepsin L (TgCPL) has recently been implicated as a novel viable target for the treatment of chronic toxoplasmosis. In this study, we report the first body of SAR work aimed at developing potent inhibitors of TgCPL with selectivity vs the human cathepsin L. Starting from a known inhibitor of human cathepsin L, and guided by structure-based design, we were able to modulate the selectivity for Toxoplasma vs human CPL by nearly 50-fold while modifying physiochemical properties to be more favorable for metabolic stability and CNS penetrance. The overall potency of our inhibitors towards TgCPL was improved from 2 μM to as low as 110 nM and we successfully demonstrated that an optimized analog 18b is capable of crossing the BBB (0.5 brain/plasma). This work is an important first step toward development of a CNS-penetrant probe to validate TgCPL as a feasible target for the treatment of chronic toxoplasmosis.

Graphical Abstract

With an estimated two billion people chronically infected, Toxoplasma gondii is one of the most pervasive infectious agents in the world.^{1, 2} This neurotropic protozoan is the 2^nd leading cause of death due to foodborne illness in the US, and has been designated as one of five neglected parasitic infections targeted by the Center for Disease Control and Prevention for public health action.^3–5 In its mammalian hosts, T. gondii exists in two principal forms, the tachyzoite and bradyzoite. The initial disease state is caused by the rapidly replicating tachyzoite form. This stage typically lasts for a several days to weeks during which the parasite disseminates to peripheral sites including the central nervous system (CNS). T. gondii then transitions into a chronic bradyzoite cyst form within the CNS and muscle tissue to establish life-long infection.^{6, 7} This chronic stage infection was once considered to be benign, but recent studies have identified associations with variety of host behavioral changes and mental illnesses including schizophrenia, bipolar disorder, and obsessive compulsive disorder.^8–12 In the case of immunocompromised hosts (e.g. chemotherapy patients and HIV-positive individuals), the parasite cysts can reactivate into the active tachyzoite form, leading to more serious complications such as, encephalitis, blindness, and death.^{1, 13, 14} Currently, there is no efficacious T. gondii therapeutic capable of treating the chronic-phase infection. Given the limitations and current inadequacies of antiparasitic therapies, there is a serious need for the development of novel and alternative strategies for the treatment of chronic T. gondii infection.

Inhibition of vital cysteine proteases is an increasingly attractive approach for the treatment of various disease states caused by protozoans such as Trypanosoma cruzi, Plasmodium falciparum, and others.^15–17 Recently, we demonstrated that T. gondii cathepsin protease L (TgCPL) is critical to parasite survival during the chronic phase of infection.^{18, 19} Inhibition of TgCPL in tissue culture by the irreversible inhibitor morpholinurea-leucine-homophenylalaninevinyl phenyl sulfone 1 (LHVS) (Figure 1) kills T. gondii cysts, further implicating TgCPL as a viable target for treatment of the chronic stage infection. Unfortunately, LHVS is unable to cross the blood brain barrier precluding its use as a proof-of-concept probe in infected mice.²⁰ Additionally, LHVS is a nonselective and irreversible inhibitor, increasing the possibility of off-target effects and poor selectivity across cathepsin isoforms.

Figure 1.

Literature Lead Cathepsin L Inhibitors.

An enormous amount of drug discovery and development has been done in regard to both human and parasitic cathepsins isoforms, and extensive mechanistic and structural information is available.^21–28 Given the close homology between the human and parasitic isoforms of cathepsin L, we considered the vast libraries of inhibitors already developed against human isoforms to be a promising source of lead anti-parasitic compounds.²⁹ To identify a better potential lead compound than LHVS, we conducted a broad literature search of structurally diverse CPL inhibitors emphasizing physicochemical properties predictive of CNS permeability, selectivity over other human cathepsins, and potency (IC₅₀ < 1 μM).

We ultimately selected our initial lead compound 2 (Figure 1) based on its calculated properties: molecular weight (MW), topological polar surface area (tPSA), numbers of hydrogen bond acceptors and donors (HBA, HBD), and rotatable bonds (RotB), which are collectively closer to marketed CNS drugs vs non-CNS drugs than are those of LHVS and most other literature leads we considered (Table 1)³¹. In addition, we considered the ease with which we could expect to improve predicted BBB access by further lowering the tPSA.

Table 1.

Comparison of calculated properties of lead inhibitors vs CNS and non-CNS drugs

Compound	MW	SlogP	tPSA	HBA	HBD	RotB
CNS drug	288 ± 88	2.8 ± 1.4	46 ± 26	2.6 ± 1.3	1.0 ± 0.9	4.0 ± 2.6
non-CNS drug	383 ± 210	2.1 ± 2.9	98 ± 84	4.8 ± 4.1	2.8 ± 3.3	6.2 ± 5.5
1	527.69	3.6	104.8	5	2	14
2	381.48	2.4	99.8	4	2	10

Calculated using The Molecular Operating Environment (MOE), version 2008.10, Chemical Computing Group Inc., Montreal, Quebec, Canada.³⁰ Mean values and standard deviations derived from 198 approved CNS drugs and 1015 approved non-CNS drugs (DrugBank).

As the first ever SAR campaign targeting TgCPL, we elected to use a dipeptide nitrile scaffold as our lead, based on the extensive amount of dipeptide-type cathepsin inhibitors previously developed for related human isoforms.^32–34 We believed the use of dipeptide probes would enable rapid elucidation of the pharmacophore necessary for the selective inhibition of TgCPL due to their synthetic accessibility and ease of diversity introduction. Another key rationale for selecting 2 as our lead was the aminoacetonitrile “warhead”. The majority of small molecule protease inhibitors incorporate a functional group that can react with the active site cysteine or serine in a reversibly covalent manner that imparts high affinity with relatively low off-target reactivity. ^{32, 35, 36} The aminoacetonitrile group has been extensively studied and is present in a number of FDA-approved drugs and advanced clinical candidates, including vildagliptin, saxagliptin, and odanacatib.³⁷ Furthermore, its low molecular weight and relatively low tPSA were considered advantageous for achieving CNS permeability.

Finally, we selected 2 as a lead because of the availability of an X-ray co-crystal structure of it bound to human cathepsin L (HsCPL, (PDB: 3HHA).³¹ Opportunely, a crystal structure of TgCPL was also available from our previous work, which we expected would greatly facilitate the design of TgCPL-selective inhibitors.^{31, 38–41} The crystal structure of 2 bound to HsCPL shows that the active site Cys25 is covalently added to the nitrile, consistent with other reversibly-covalent nitrile based inhibitors of cysteine proteases, and that the ligand makes hydrogen-bonding interactions with Asp162, Gly68, and Gln19 (Figure 2).^{32, 35} Importantly, the bound inhibitor 2 can be overlaid on our crystal structure of TgCPL to identify proximal structural differences between the cathepsins that can be exploited to achieve selective inhibition of the T. gondii enzyme (Figure 3). ¹⁸ In particular this overlay reveals a smaller S2 pocket for the Toxoplasma cathepsin, as well as four non-conserved residues (colored in Figures 3A,B). Further comparison of the protein sequence alignment of TgCPL and the human cathepsin isoforms A-X, revealed that these key residues are either unique to TgCPL, or generally nonconserved across human isoforms. The S2 pocket across the human cathepsins tends to be more uniformly lipophilic (colored green in Figures 3C,D) than TgCPL, and the inclusion of a basic residue in the P2 position, which we predicted might interact strongly with non-conserved Asp-218 in TgCPL, is generally not tolerated. With the exception of human cathepsin B, which bears a Glu245 in this respective position, the overall topology in the S2 subsite is thus substantially different than that of TgCPL.⁴² This provided excellent supporting evidence that optimization in the P2 position would be paramount in an effort to gain selectivity over the bulk of the human isoforms.

Figure 2.

LIGPLOT (right) diagram generated in MOE, shows key interactions between 2 and HsCPL. Hydrogen bonds: blue arrows, lipophilic interactions: blue spheres. Derived from PDB: 3HHA

Figure 3.

Comparison of 2 bound to HsCPL (A, C) and a superposition of 2 modeled into the active site of TgCPL (B, D) depicting key non-conserved residues. A) HsCPL with key residues colored. B) TgCPL with key residues colored C) Lipophilicity mapping of HsCPL (green: lipophilic; magenta: polar); D) Lipophilicity mapping of TgCPL. (Derived from PDB codes: 3HHA and 3F75)

Our primary objectives in this preliminary work were: 1) to develop analogs with better predicted CNS permeability without losing potency; and 2) to determine if we can begin to shift the selectivity of 2 away from HsCPL and towards TgCPL. Herein we report the development of a series of dipeptide nitrile probes that help to define pharmacophoric features key to improving both potency and selectivity for TgCPL, including one analog with CNS permeability in vivo.

Synthesis of analogs 7a–r (Scheme 1) was performed analogous to the literature precedent for the peptide-nitrile cathepsin inhibitors.^{31, 32, 34, 41} We began with the esterification of various commercially available L-amino acids 3a–h to afford their respective methyl esters 4a–h. Subsequent amide coupling to the appropriate aryl acids gave 5a–r. The methyl ester was then saponified under mild conditions with LiOH in a THF:MeOH system to afford the desired free acid 6a–r. Initial coupling with 2-aminoacetonitrile gave several unwanted side reactions and generally low yields when using DMF or THF as the solvent, regardless of the temperature or coupling reagent used. Upon changing to dichloromethane we found that, while solubility was not always ideal, the desired products 7a–r were formed cleanly and in moderate yields.

Scheme 1. General synthesis of dipeptide nitriles.

Reagents and conditions: a) H₂SO₄, MeOH, reflux; b) HATU, DIPEA, DMF; c) 1N LiOH_(aq), THF, MeOH; d) 2-aminoacetonitrile bisulfate, HATU, DIPEA, DCM, 0°C-rt

Synthesis of 13 was achieved as shown in Scheme 2. Nε-Boc-L-lysine methyl ester hydrochloride 8 was coupled to 4-fluorobenzoic acid to provide intermediate 9. Boc removal from 9 afforded the HCl salt of 10 which was immediately followed by dimethylation of the sidechain to provide compound 11. Saponification of 11 with lithium hydroxide provided 12, and subsequent amide coupling using HATU gave the desired product 13.

Scheme 2. Synthesis of dimethyl-lysine analog.

Reagents and conditions: a) 4-Fluorobenzoic acid, HATU, DIPEA, DMF, 98%; b) 4N HCl/dioxane, 99%; c) formaldehyde, sodium triacetoxyborohydride, MeOH, 32%; d) 1N LiOH_(aq), THF, MeOH, 100%; e) 2-aminoacetonitrile bisulfate, HATU, DIPEA, DCM, 0°C-rt, 11%.

Trifluoroethyl amine analogs (18a–b, 19) were generally synthesized as shown in Scheme 3 following previously established methods in the development of odanacatib.^{43, 44} Synthesis began with the condensation of the trifluoroacetophenone 14 and L-Leucine-methylester under basic conditions to afford the imine intermediate 15. The R,S diastereomer 17, was obtained via dropwise addition of NaBH₄ dissolved in THF at −40°C in a 10:1 dr. Alternatively the (S,S) diastereomer could be obtained through a chelation controlled reduction of the imine with zinc borohydride in acetonitrile/dimethoxyethane at −40°C, providing 16 in an 11:1 dr. The (S,S) and (R,S) compounds 18a, 18b, and 19 were obtained in 45–94% yield through the previously described amide coupling conditions (Scheme 1).

Scheme 3.

Reagents and conditions: a) Leucine hydrochloride, K₂CO₃, MeOH, 50°C, 18h; b) Zn(BH₄)₂, ACN/DME, −40°C, 5h, 87%; c) NaBH₄, THF, 93%; d) 2-aminoacetonitrile bisulfate, or 1-aminocyclopropane-1-carbonitrile HCl, HATU, DIPEA, DCM, 0°C-rt, 45–89%

Fluoroleucine analog 24 was prepared as shown in Scheme 4 through a sequence derived from previously reported methods for analogous compounds.^44–47 Dehydroleucine 20, as the CSA salt, was coupled to 4-fluorobenzoic acid to provide intermediate amide 21. Initially, we tried to achieve fluorination of 21 using a variety of the previously reported conditions, predominantly using reagents that are a source of nucleophilic fluorine such as HF:Urea or HF:Pyridine. However, these conditions generally suffered from poor yields and unwanted side reactions such as cyclization or elimination. We then opted to try an electrophilic source of fluorine as an alternative approach, derived from that reported by Fanelli et. al..⁴⁸ Activation of the olefin with Fe(III), followed by the sequential addition of Selectfluor and NaBH₄ provided the desired product 22 in moderate which, was then saponified to 23 and directly coupled to 2-aminoacetonitrile, affording the desired compound 24.

Scheme 4.

Reagents and conditions: a) 4-fluorobenzoic acid, HATU, DIPEA, DMF, 34%; b) Selectfluor, Fe₂(SO₄)₃, NaBH₄, acetonitrile:water, 0°C-rt, 27%; c) 1N LiOH_(aq), THF:EtOH, 0°C-rt, 100%; d) 2-aminoacetonitrile bisulfate, HATU, DIPEA, DCM, 0°C-rt, 18%

All new compounds were evaluated in vitro for inhibition of both TgCPL and HsCPL activity by monitoring the rates of hydrolysis of Z-Leu-Arg-7-amino-4-methylcoumarin (Z-L-R-AMC). Results are summarized in Tables 2–5. We first re-synthesized literature lead 2 to benchmark its relative potency against TgCPL versus HsCPL. In vitro testing revealed low micromolar potency against TgCPL, and 46-fold more potent inhibition against HsCPL (Table 2).

Table 2.

Cathepsin L inhibitory activity of initial P2 and P3 analogs.

Name	Ar	R	TgCPLa IC50 (μM)	HsCPLb IC50 (μM)	Selectivity for HsCPL c
1	N/A	N/A	0.002	0.003	0.7
2			2.0	0.044	46.4
7a			2.4	0.11	22.5
7b			1.4	0.13	11.3
7c			1.4	0.042	33.5
7d			1.2	0.092	13.3
7e			6.1	0.086	71
7f			4.6	0.24	18.8
7g			0.12	0.077	1.9
7h			0.12	0.32	0.4

^a,bIC₅₀ for Toxoplasma gondii Cathepsin L (TgCPL) or human Cathepsin L (HsCPL). Values are mean of at least 3 independent experiments.

^cSelectivity ratio for human Cathepsin L as defined by TgCPL IC₅₀/HsCPL IC₅₀.

Table 5.

Inhibitory activity of trifluoroethylamine and cyclopropyl-nitrile analogs.

Name		R	TgCPLa IC50 (μM)	HsCPLb IC50 (μM)	Selectivity for HsCPL c
7n			0.25	0.28	0.9
19			0.66	0.59	1.1
18a			0.11	0.093	1.2
26			0.40	0.14	2.9
18b			0.16	0.084	1.9

See Table 2 for heading definitions.

Initially we replaced the pyrazole heterocycle in the P3 position of 2 with phenyl in order to remove two nitrogen atoms (7b) and rapidly lower the tPSA of 2, thereby increasing the potential for CNS permeability. This exchange resulted in a 2–3-fold drop in potency against HsCPL while enhancing activity against TgCPL, improving the TgCPL-selectivity about 4-fold. The t-butyl group of 7b was then replaced by trifluoromethyl to stabilize the now electron-rich P3 aromatic ring to potential oxidation by CYP450 enzymes, giving 7a. This analog lost potency against TgCPL, but maintained effectiveness versus HsCPL. Since both of these analogs had reduced selectivity for HsCPL compared to lead 2 we retained both of the P3 groups during our initial survey of P2.

The vast majority of selectivity between human cathepsin isoforms comes from the P2-S2 interaction.⁴⁹ The S3 and S1 subsites also play significant roles in inhibitor selectivity, while the S′ subsites (toward the C-terminus of the scissile bond) are highly conserved between isoforms and generally difficult to exploit.²² Inspection of the cathepsin crystal structures revealed several active site residues unique to the T. gondii isoform vs the human isoform that could potentially be exploited to gain selectivity and potency. As noted above, the TgCPL S2 site contains a unique aspartate residue (Asp218), as opposed to the human isoform, which bears a hydrophobic alanine in this position (colored green in Figures 3A, B). Overlaying TgCPL with the co-crystal structure of our lead compound bound to HsCPL indicates that the 3-methylphenylalanine (3-Me-Phe) core of 2 sits approximately 3–4 Å from this key aspartate (Figure 3B). Due to the proximity of the core to Asp218 we hypothesized we might be able to build in a potential H-bond to improve potency and selectivity.

The P2 3-methylphenylalanine residue of analogs 7a and 7b was thus replaced with 3-hydroxyphenyalanine or 3-methoxyphenylalanine (7c–7f) in an attempt to pick up a selective hydrogen bonding interaction with the non-conserved Asp218 of TgCPL. We had hoped this would provide some initial gain in selectivity toward the parasite cathepsin, but the results were inconsequential changes in potency. This might be due to the rigidity of the Asp218 backbone orienting the residue away from the S2 pocket and preventing a beneficial interaction.

Molecular modeling also identified several predicted clashes with HsCPL when the larger phenylalanine side chains are in the P2 position (2 in Figure 4). We were pleased to discover that replacement of the entire aromatic P2 residue with a less bulky leucine (7g,7h) significantly reduced selectivity for HsCPL, consistent with the smaller S2 subsite in the TgCPL as compared to the human isoform (Table 2). With this single change, we reduced the selectivity of 2 for HsCPL from 10–20 fold to <2 fold. This is an excellent validation that the differences in the S2 pockets can be exploited for improving TgCPL selectivity.

Figure 4.

Overlay of 2 (gray) and 18b (orange) into the crystal structure of TgCPL (PDB code: 3F75) depicting clashes with 2 in the S2 pocket. Hydrogen bonds and clashes are represented as blue and orange dashed lines, respectively. Pocket surface coloration represents polar (purple), lipophilic (green), and solvent exposed (red) areas.

We then returned to optimization of the P3 position in an effort to reduce the molecular weight of 7h while retaining a halogenated pendant for metabolic stability (Table 3). Removal of the trifluoromethyl from the P3 pendant to afford the simple phenyl (7i) unfortunately reversed the selectivity back to HsCPL over TgCPL. A similar trend was observed with analogs 7j, 7l, and 7m, which bear a 3-methyl, 2-chloro-4-fluoro, or 2,4-dichloro substitution pattern, respectively, on the P3 pendant. Replacement of the 3-trifluoromethyl of 7h with a fluorine (7k) also resulted in a large drop in potency for TgCPL and an increase for the human isoform. The 4-fluoro analog 7n, however, was equipotent at TgCPL and HsCPL, and lowered the MW below 300, making this a more promising candidate for CNS permeability as compared to the somewhat more potent and TgCPL-selective 4-trifluoromethyl analog 7h.

Table 3.

Inhibitory activity of additional aryl P3 substitution analogs.

Name	Ar	TgCPLa IC50 (μM)	HsCPLb IC50 (μM)	Selectivity for HsCPLc
7i		0.35	0.11	3.3
7j		0.18	0.015	11.7
7k		0.36	0.048	7.6
7l		0.54	0.28	1.9
7m		0.53	0.14	3.8
7n		0.25	0.28	0.9

^a,b,cSee Table 2 for heading definitions.

We then explored a dimethyl lysine at P2 as a basic group to establish a salt bridge with the non-conserved Asp218 of TgCPL (13 in Table 4). While this was predicted by modeling to be the optimal length to interact with the S2 aspartic acid, it nevertheless resulted in a major loss in potency against both isoforms. However, it should be noted that the loss in potency was significantly greater against the human isoform, demonstrating that despite the S2 pocket of these two enzymes being very similar, subtle differences at the P2 residue can offer drastic changes to the selectivity profile. Further exploration of basic residues at P2 is therefore warranted in future SAR studies.

Table 4.

Inhibitory activity of additional aliphatic P2 analogs

Name	R	TgCPLa IC50 (μM)	HsCPLb IC50 (μM)	Selectivity for HsCPL c
7o		0.42	0.19	2.3
7p		6.6	1.2	5.7
13		14	816	0.02
24		4.0	2.5	1.6
25		2.0	1.2	1.6
7q		1.2	3.5	0.3
7r		9.8	4.1	2.4

^a,b,cSee Table 2 for heading definitions.

Cyclohexyl P2 analogs have been reported as inhibitors of human protease inhibitors that offered some improvement in either binding or metabolic stability.⁵⁰ To further probe aliphatic SAR at the TgCPL P2 position, both were made (7o and 7p, Table 4); however, neither offered any improvement. Rather, a decrease in potency and TgCPL selectivity was observed for both, more drastically in the case of 7p, which is a closer mimic of valine than leucine. Substitutions much larger than leucine were thus apparently not well tolerated in TgCPL, consistent with the poor activity we had observed earlier with our phenylalanine analogs, so we attempted to install isosteres that were closer in size to leucine. While the use of fluoroleucine to improve metabolic issues has been successfully demonstrated in the development of odanacatib⁵¹, we were surprised to see a drastic drop in potency against both TgCPL (IC₅₀= 4.05 μM) and HsCPL (IC₅₀= 2.51 μM) with compound 24. Similarly, exchange of the P2 leucine with either a neopentylglycine, or dehydroleucine (7q, 25) resulted in a significant loss of potency (IC₅₀= 1.19 μM and 2.00 μM respectively). Surprisingly, even the cyclopropyl analog 7r showed a large drop in potency as compared to 7n. While these structural changes are minor, the substantial changes in potency clearly demonstrate the high level of sensitivity of the P2 position to steric modification. Within this initial set of dipeptide nitrile TgCPL inhibitors, leucine thus proved to be the most optimal residue in the P2 position for both potency and reduced selectivity for HsCPL.

Most CNS drugs do not tolerate more than 1 or 2 hydrogen bond donors. As shown in the LIGPLOT diagram from the co-crystal structure of HsCPL and compound 2 (Figure 2), the carbonyl of the P2–P3 amide is not actively engaged by the enzyme and therefore may be unnecessary. Replacement of this amide with the well-established trifluoroethlyamine bioisostere offered an effective way to reduce our overall tPSA, greatly enhancing our chances for CNS penetration.^{43, 51–53} We found that the (R,S) diastereomer 19 gave a substantial drop in overall potency for both cathepsins (Table 5). The (S,S) diastereomer 18a, however, doubled the potency against both the Toxoplasma and human enzymes, providing the most potent nitrile-based inhibitor to date against TgCPL with an IC₅₀= 110 nM. This diastereomeric preference is consistent with literature reports of other dipeptide cathepsin inhibitors incorporating the trifluoroethyl amine bioisostere.⁴³

The metabolic stability of our dipeptide nitrile series was initially evaluated in mouse liver microsomes (MLM) to assess suitability for future in vivo studies and to further guide our design. Compound 7h was evaluated to benchmark the dipeptide nitrile series (Table 6). We observed that compound 7h had a reasonably long half-life (37 min). This parameter was further improved to >60 min with less lipophilic compound 7n, despite several metabolic liabilities predicted by SMARTCyp (Figure 5).⁵⁴ Interestingly, while replacement of the P2 amide with a trifluoroethylamine (18a) removed an amide that could be a metabolic liability, the half-life in mouse liver microsome decreased to 4.4 minutes. It is likely that the increased lipophilicity simply enhanced binding to metabolizing enzymes. Two of the most likely positions for metabolism of 18a are oxidation alpha to the nitrile warhead and hydrolysis of the P1 amide bond.⁵⁴ The incorporation of a cyclopropyl at the P1 position serves to block this site from oxidation and to slow hydrolysis of the P1 amide, a strategy successfully used in the development of odanacatib and other inhibitors of human cathepsins.^{51, 53, 55} Cyclopropyl analog 18b indeed provided a significantly improved MLM half-life of 22 min, affording a compound more suitable for further in vivo analysis. It also retained potency for TgCPL, which is consistent with the predicted projection of the cyclopropyl group at P1 out into solvent (Figure 4).

Table 6.

Metabolic stability of select analogs in mouse liver microsomes.

Compound	R1		R2	MLM t1/2 (min)
7h	3-CF3			37
7n	4-F			>60
18a	4-F			4.4
18b	4-F			22

Figure 5.

SMARTCyp prediction of the top three sites of metabolism.

Compound 18b was advanced to an abbreviated in vivo pharmacokinetic study in mice (Table 7). Our main objective was to determine if this chemotype could penetrate the CNS. Significantly, we were able to detect compound in the brain, with a brain/plasma ratio of 0.5 at 30 minutes. Unfortunately, the overall levels in both brain and plasma were quite low, despite the fact that 18b was shown to have good aqueous solubility (630 μM) and excellent plasma stability (~100% parent remaining after 24h). The plasma levels and rapid disappearance indicates that this compound has a high volume of distribution, high clearance or both. Studies are currently underway to address this issue and to develop new analogs with improved pharmacokinetic profiles.

Table 7.

In Vivo Exposure of Compound 18b Following IP Administration to Mice^a

Time Post-Dose	30 min	2 h	4 h	7 h
Plasma Conc. (ng/mL)	110	138	10	BLQ
Brain Conc. (ng/g)	54	BLQ	BLQ	BLQ

^aCD-1 mice were injected intraperitoneally with a single 10 mg/kg dose. The data shown are mean values from 2 mice at each time point. BLQ = Below Limit of Quantification.

In summary, we have demonstrated for the first time that it is possible to modify a well established chemotype for inhibition of human cathepsins (dipeptide nitrile) and shift selectivity away from the human isoform and towards the Toxoplasma gondii isoform, a novel target for toxoplasmosis. Furthermore, we have demonstrated that the physical properties of the dipeptide can be adjusted to be more favorable for CNS penetration without loss of activity, particularly through reduction in tPSA and number of HBA. Key analogs in this campaign are summarized in Figure 6. Initially, our lead compound 2 was approximately 46-fold selective for HsCPL over TgCPL. We developed several compounds that either displayed equipotency for HsCPL and TgCPL (7n), or exhibited slight selectivity (up to 3-fold) for TgCPL over HsCPL (7h,7q). Notably, compound 13 (Table 4) demonstrates the potential for achieving outright selectivity for TgCPL through the incorporation of a basic residue in the P2 position, an avenue we are exploring in future studies. Furthermore, we improved the potency of our inhibitors for the T. gondii enzyme from 2 μM to as low as 115 nM. Although we successfully demonstrated that compound 18b can penetrate the blood brain barrier, it also exhibits rapid clearance in vivo. Current endeavors in our lab are aimed at further defining the pharmacophore for TgCPL via optimization of the P3 and P2 positions. Additionally, we are improving the pharmacokinetic profile for these inhibitors, so they may be advanced to in vivo murine infection studies in order to establish proof-of-concept that TgCPL is a viable target for treating the chronic form of toxoplasmosis.

Figure 6.

Key compounds in the optimization of 2 for TgCPL potency, metabolic stability, and a CNS-penetrant profile.