Macrocyclic Peptides as a Novel Class of NNMT Inhibitors: A SAR Study Aimed at Inhibitory Activity in the Cell

Abstract

Nicotinamide N-methyltransferase (NNMT), which catalyzes the methylation of nicotinamide, is a cytosolic enzyme that has attracted much attention as a therapeutic target for a variety of diseases. However, despite the considerable interest in this target, reports of NNMT inhibitors have still been limited to date. In this work, utilizing in vitro translated macrocyclic peptide libraries, we identified peptide 1 as a novel class of NNMT inhibitors. Further exploration based on the X-ray cocrystal structures of the peptides with NNMT provided a dramatic improvement in inhibitory activity (peptide 23: IC₅₀ = 0.15 nM). Furthermore, by balance of the peptides’ lipophilicity and biological activity, inhibitory activity against NNMT in cell-based assay was successfully achieved (peptide 26: cell-based IC₅₀ = 770 nM). These findings illuminate the potential of cyclic peptides as a relatively new drug discovery modality even for intracellular targets.

Nicotinamide N-methyltransferase (NNMT) is a cytosolic enzyme that catalyzes methyl transfer from S-adenosyl-l-methionine (SAM) to nicotinamide (NA) to provide 1-methylnicotinamide (MNA).^1,2 The role of NNMT has traditionally been known as the nicotinamide clearance and xenobiotic detoxification.³ Recent research has elucidated other important functions of NNMT in various diseases by modulating levels of its products, substrates, and cofactors.⁴ In fact, it has been reported that NNMT is overexpressed in a variety of diseases such as cancer,⁵⁻⁷ metabolic disorders,⁸⁻¹⁰ cardiovascular disease,^11,12 and Parkinson’s disease.^13,14 Thus, it is fair to say that NNMT could become a potential therapeutic target due to its biological function in a wide range of disease areas. However, to date, although several NNMT inhibitors have been reported in response to the increasing interest mentioned above, these are limited to substrate mimetics or covalent inhibitors.¹⁵⁻¹⁹

Constrained peptides have attracted increasing attention as a new therapeutic modality in the recent years.²⁰ Among of them, cyclic peptides represent one of the major molecular classes, due to their conformational restraints, which can facilitate target engagement and resistance to proteolysis.²¹⁻²³ To meet this growing interest in cyclic peptides, affinity selection methods for cyclic peptides utilizing in vitro translated macrocyclic peptide libraries produced through flexizyme-mediated genetic code reprogramming have been reported.²⁴⁻²⁶ It is also known that cyclic peptides possess the potential for cell permeability by forming intramolecular hydrogen bonds, which can mask polar sites attributed to hydrogen bond donors and acceptors of the peptide amide moieties.²⁷ Although recent intensive studies identified many cell-penetrating cyclic peptides and revealed their cell entry mechanisms, intracellular proteins have remained challenging targets in the field of peptide drug discovery.²⁸ Herein, we describe structure–activity relationship (SAR) studies of cyclic peptides identified from affinity selections utilizing this in vitro translation system as a novel class inhibitors against NNMT and demonstrate a lipophilicity–potency balancing approach aimed at obtaining NNMT inhibitory activity in a cell-based assay.

An affinity selection campaign with in vitro translated macrocyclic peptide libraries against NNMT successfully yielded macrocyclic peptide 1, which exhibited desired NNMT inhibitory activity (IC₅₀) against both human and mouse NNMT (human IC₅₀ = 28 nM, mouse IC₅₀ = 25 nM). Values of IC₅₀ were calculated by the estimation of MNA production from the NNMT-mediated methylation of NA in the presence of SAM (see Supporting Information). Given this result, peptide 1 was defined as a hit compound for subsequent exploration, although it was initially viewed as a highly polar compound containing hydrophilic moieties on its side chains. In order to obtain inhibitory activity in cells, we decided to perform chemical modification of 1 to enhance the cell-free biological potency while simultaneously seeking to reduce the number of hydrogen bond donors (No. HBD) and increase AlogP as indicators of lipophilicity, which was thought to be key factors in terms of obtaining cell penetration (1: No. HBD = 14, AlogP = 0.380).

All of peptides described in this paper were synthesized by Fmoc solid-phase peptide synthesis using Sieber amide resin, Rink amide resin, or cysteamine 2-chlorotrityl resin.²⁹ After resin cleavage and global deprotection under acidic conditions, the resulting linear peptides underwent annulation by S_N2 reaction between the thiol and acetyl chloride moiety with an organic base under high dilution conditions to afford target cyclic peptides (Scheme 1). The cyclic peptides were all purified using reversed-phase high-performance liquid chromatography (HPLC) and confirmed to have >90% purity by HPLC analysis.³⁰

Scheme 1. Representative Synthetic Route.

Reagents and conditions: (a) 5% piperazine, 0.1 M HOBt in DMF, rt; (b) 3 equiv of Fmoc amino acid, 3 equiv of HATU, 6 equiv of DIEA, 75 °C (rt for Cys); then 5% piperazine, 0.1 M HOBt in DMF, rt (repeated for elongation); (c) (i) 3 equiv of chloroacetic acid, 3 equiv of HCTU, 3 equiv of DIEA, DMF, rt; (ii) TFA/TIS/H₂O/DODT (92.5/2.5/2.5/2.5), rt; (iii) 6 equiv of Et₃N, DMSO (5 mM).

First, to identify removable hydrogen bond donors in the backbone, N-methylation scanning of peptide 1 was carried out. Although N-methylation can potentially change the peptide conformation, thus drastically influencing the biological activity,³¹ this scan revealed that the N–H group of 1-Phe (highlighted in red) could be methylated with a modest loss in activity and slight gain in lipophilicity (peptide 2 in Table 1, AlogP = 0.586).

Table 1. Results of N-Methylation Scanning.

^aValues represent an average of at least two experiments.

Since cyclic peptide drugs are typically are much larger in size than small molecule therapeutics (in this case, macrocyclic peptide 1 consists of eight amino acids and has a molecular weight of over 1100 Da), an X-ray cocrystal structure was considered highly desirable to provide valuable structural information to guide peptide design studies. Hence, for peptides 1 and 2, cocrystallization conditions were broadly screened, and to our delight, the cocrystal structure of 2 bound to NNMT was successfully obtained. It should be noted that the cocrystal structure of NNMT and peptide 2 was found to consist of a protein dimer, with the dimer interface located close to the protein–peptide binding region (Figure S1). Although interactions at the dimer interface could affect NNMT inhibition, we assumed that the dimerization was an artifact of crystal formation and decided to utilize the monomeric structure for subsequent ligand design. Interestingly, the X-ray structure analysis indicated that peptide 2 bound to the same pocket as the natural S-(5′-adenosyl)-l-homocysteine (SAH) substrate, with two conspicuous conformational changes compared with the reported cocrystal structure of NNMT-SAH-NA: absence of the N-terminal helix and opening of the flap covering the binding pocket of SAH and MNA (Figure 1).

Figure 1.

Comparison between cocrystal structure of NNMT-SAH-NA and NNMT-peptide 2. (A) X-ray crystal structure of NNMT complexed with SAH and NA (PDB code 3ROD). Orange CPK molecules represent SAH and nicotinamide. Gray ribbons represent NNMT. The purple and cyan domains represent N-terminal helix (res 1–13) and flap (res 200–214), respectively. (B) X-ray crystal structure of NNMT complexed with peptide 2 (PDB code 7EHZ). A green CPK molecule and gray ribbons represent peptide 2 and NNMT, respectively.

Peptide 2 binds to NNMT, with multiple interactions including a perpendicular edge-to-face π–π interaction between 1-MePhe and Tyr25, hydrogen bonds of the backbone amide carbonyls of 2 with Tyr20 and Tyr25 (via a water molecule) and Asn90, as well as the interaction of 4-Arg with Thr67 and Thr163 (Figure 2). Additionally, three intramolecular hydrogen bonds were formed in the bound conformation. Furthermore, the terminal carbamoyl (CONH₂) of peptide 8-Cys was found to be a solvent-exposed moiety, which could be removed with a 6-fold loss in activity (CONH₂-removed 2: human IC₅₀ = 800 nM) reducing two HBDs. Note that the position of the CONH₂ nitrogen atom could not be assigned in the X-ray structure analysis due to its poor electron density.

Figure 2.

Interactions between NNMT and peptide 2. The backbone of NNMT and peptide 2 are represented with gray ribbon and green sticks, respectively. The residues of NNMT interacting with peptide 2 are highlighted as pink sticks. Yellow and cyan dashed lines indicate hydrogen bonds and π–π interaction, respectively. Intramolecular hydrogen bonding in peptide 2 is also depicted as orange dashed lines.

As no interaction of 3-Tyr with NNMT was observed in the X-ray structure, SAR studies at R(3) were conducted in anticipation of excluding HBDs (Table 2). The hydroxyl group on the phenyl ring of 3-Tyr could be removed by preparing the 3-Phe derivative to reduce one HBD while essentially maintaining the potency of peptide 2. Moreover, further SAR exploration indicated that the introduction of Leu, a lipophilic alkyl side chain, actually enhanced the inhibitory activity while excluding the unessential HBD.

Table 2. Replacement of 3-Tyr.

^aValues represent an average of at least two experiments.

Since the inclusion of an Arg residue would be expected to bring a substantial reduction in passive cell permeability due to the presence of multiple HBDs, the replacement of 4-Arg was thought to be crucial to pursue intracellular inhibitory activity. As shown in Table 3, introduction of Lys, a basic side chain as Arg, resulted in retention of inhibitory activity while reducing the number of HBDs (peptide 8). On the other hand, N,N-dimethylation of 4-Lys substantially increased the compound lipophilicity (AlogP: 1.12 to 2.14), while slightly diminishing potency (peptide 9). Isosteric replacement with heteroaromatic rings (peptides 10 and 11) was not tolerated nor was the introduction of other neutral lipophilic or hydrophilic side chains such as Leu, Phe, and N-acetylated Lys (data not shown). Lys analogue 12 with Leu at R(3) showed slightly higher inhibitory activity than 8, consistent with the SAR at R(4) described in Table 2. In addition, expanding the bulkiness of Leu provided further improvement in IC₅₀ along with increase in AlogP value (peptide 13). The X-ray structure of 8 revealed a water molecule tucked between the terminal amino group of 4-Lys and the carbonyl of Thr163 (Thr163 C=O···H₂O: 1.8 Å) (Figure S2). To release the water and gain a direct interaction of the amino group with Thr163, homo-Lys was employed as an elongated-Lys analogue. As a result, 4-homo-Lys derivative 14 showed significantly improved inhibitory activity and provided an approximately 50-fold potent IC₅₀ compared with 4-Lys analogue 12.

Table 3. SAR Exploration of R(4).

^aValues represent an average of at least two experiments.

Next, in order to fill the lipophilic spaces around the two proline residues suggested by the cocrystal X-ray structure, substitution effects in the prolines were examined based on peptide 12′ (Table 4). For 2-Pro, introducing a trifluoromethyl group with either stereochemistry at the 4-position of the proline ring indicated more potent biological activity with 4R-substitution (peptide 17). Further dramatic improvement in potency was accomplished by attaching a phenyl group to the 2-Pro residue with the same stereochemistry (peptide 18). For 5-Pro, while difluoro substitution resulted in enhanced potency, the bulkier trifluoromethyl group lowered the inhibitory activity for both stereochemistries (peptides 19, 20, and 21). This observation is consistent with that the limited space was apparently available in the vicinity of the 5-Pro residue, as indicated by the X-ray cocrystal structure (Figure S2).

Table 4. Substitutions on Prolines with Lipophilic Groups.

^aValues represent an average of at least two experiments.

Peptide X shown in Figure 3A, another peptide identified from the selection studies of in vitro translated macrocyclic peptide libraries, did not proceed to further SAR exploration because it appeared to be a less favorable starting point for obtaining cell penetration, due to its high molecular weight (1274 Da) and number of HBDs (17). However, the X-ray structure of peptide X bound to NNMT, having almost the same protein backbone conformation as NNMT bound to peptide 2, provided useful information for our SAR studies. The terminal CONH₂ group of peptide X was found to be engaged in a close interaction with Val143, confirmed by the significant decrease in inhibitory activity found upon removal of the C-terminal Gly (Figure 3A). On the other hand, superposition of cocrystal structure of 2 with NNMT and X with NNMT proposed that extension of the side chain at position 7 (7-F4CON) in peptide 2 could approach Val143 (Figure 3B).

Figure 3.

SAR of peptide X (A) and superimposed structure of NNMT cocrystal structures of peptide X (PDB code 7EGU) and 2 (B). Peptide X and 2 are highlighted in blue and green, respectively. Sky blue and white ribbons represent NNMT protein for peptide X and 2, respectively. Val143 is highlighted in magenta. Interaction between Val143 and peptide X is shown as an orange dashed line.

Therefore, to incorporate this promising interaction with Val143 observed in peptide X into our SAR exploration, elongated side chains with the potential to replace 7-F4CON were investigated (Table 5). Whereas replacement with a nonsubstituted biphenyl group resulted in reduced inhibitory activity (peptide 22), the introduction of a hydrogen bond acceptor that could interact with Val143 into the terminal phenyl ring considerably increased the potency of NNMT inhibition to furnish the most potent peptide found in these studies (peptide 23, human IC₅₀ = 0.15 nM).

Table 5. SAR Exploration of R(7).

^aValues represent an average of at least two experiments.

Finally, a combination of the most promising single-point mutations found in the above SAR studies was conducted to arrive at the compounds shown in Table 6. These compounds were tested in a cell-based assay designed to measure the MNA quantity after culturing of NNMT-overexpressing HEK293 cells (see Supporting Information). In this assay, our most potent peptide 23 did not demonstrate measurable cell-based inhibitory activity, ascribed to its high polarity (AlogP = 2.41, No. HBD = 11) preventing effective cell penetration. Therefore, more lipophilic peptides with fewer HBDs were designed by leveraging the SAR results described above. Peptide 25, with the above favorable lipophilic substituents combined and the C-terminal carboxamide group removed (HBDs were reduced from 11 to 6), showed slightly lower biological potency than compound 23 but exhibited a micromolar-level cell-based IC₅₀ that could be attributed to increased lipophilicity (AlogP = 6.25). Furthermore, while maintaining the most potent polar side chains, we reinvestigated the possibility of removing the basic side chain (N,N-dimethyl Lys) of 25, such as by replacement with the homoproline residue in peptide 26. Despite the slightly reduced potency, peptide 26 demonstrated a substantial improvement in cell-based activity, resulting in a submicromolar cell-based IC₅₀ value. The difference between cell-based IC₅₀ and cell-free IC₅₀ values was diminished in compound 26 (cell-based IC₅₀/cell-free IC₅₀ = 405) in comparison to 25 (cell-based IC₅₀/IC₅₀ = 9250). This result might be attributed to removal of the basic character of 25, which should facilitate cell penetration but is not reflected in the AlogP value, in addition to one HBD-reduction. It was noteworthy that cytotoxicity (up to 100 μM) was not observed in these peptides (25 and 26). Moreover, the biochemical selectivity of peptides 25 and 26 toward NNMT was confirmed by performing a counter assay against indolethylamine N-methyl transferase (INMT), which is known as a structurally similar transmethylation enzyme to NNMT.^32,33

Table 6. HBD-Reduced Lipophilic Peptides.

peptide	IC50 (h/m), nMa	cell-based IC50, μMb	CC20, μMc	No. HBD	AlogP
1	28/25	>100	>100	14	0.38
23	0.15/0.15	>100	>100	11	2.41
25	0.40/0.62	3.7	>100	6	6.25
26	1.9/2.0	0.77	>100	5	6.14

^aValues represent an average of at least two experiments.

^bValues were determined using NNMT-overexpressed HEK293 cells and represent an average of at least two experiments.

^cValues represent the concentration of peptides to indicate 20% ATP decline in the cells.

In summary, a novel class of macrocyclic peptide NNMT inhibitors containing noncanonical amino acids was discovered from in vitro translated macrocyclic peptide libraries and further chemical modification. Use of a structure-based drug design (SBDD) approach enabled the efficient identification of structural features amenable to modification for the improvement of biological activity and provided multiple highly potent NNMT inhibitors with subnanomolar IC₅₀ values. By balance of the increase of peptides’ lipophilicity and the potency enhancement, peptide 26 was discovered and found to demonstrate submicromolar cell-based inhibitory activity. We envision that this research will help further drug discovery projects related to NNMT inhibitors. In addition, it is notable that macrocyclic peptides with a molecular weight of over 1200 Da successfully provided potent inhibitory activity in a cellular assay. These results strongly suggest the broader potential for the use of macrocyclic peptides in targeting intracellular proteins by ligand design approaches to balance lipophilicity and biological activity.

Glossary

Abbreviations

NNMT

nicotinamide N-methyltransferase

SAM

S-adenosyl-l-methionine

NA

nicotinamide

MNA

1-methylnicotinamide

SAR

structure–activity relationship

HBD

hydrogen bond donor

No. HBD

number of hydrogen bond donors

SAH

(S-(5′-adenosyl)-l-homocysteine)

CONH₂

carbamoyl

INMT

indolethylamine N-methyltransferase

SBDD

structure-based drug design

HPLC

high-performance liquid chromatography

HOBt

hydroxybenzotriazole

HATU

1-[bis(dimethylamino)methylene]-1H-1,2,3-triazolo[4,5-b]pyridinium 3-oxide hexafluorophosphate

DIEA

N,N-diisopropylethylamine

HCTU

2-(6-chloro-1-H-benzotriazole-1-yl)-1,1,3,3-tetramethylaminium hexafluorophosphate

DODT

2,2′-(ethylendioxy)diethanethiol

TIPS

triisopropylsilane

DMF

N,N-dimethylformamide

DMSO

dimethyl sulfoxide

TFA

trifluoroacetic acid

Et3N

triethylamine

Trt

trityl

Pbf

2,2,4,6,7-pentamethyldihydrobenzofuran-5-sulfonyl

MePhe

N-methyl-l-phenylalanine

Tyr

l-tyrosine

Asn

l-asparagine

Arg

l-arginine

Thr

l-threonine

Cys

l-cysteine

Phe

l-phenylalanine

Leu

l-leucine

Pro

l-proline

F4CON

l-4-carbamoylphenylalanine

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsmedchemlett.1c00134.

• Synthetic procedure for peptides, characterization of peptides, protocols of biological assays, and supporting figures of X-ray structures (PDF)

The authors declare no competing financial interest.