Novel irreversible peptidic inhibitors of transglutaminase 2

Abstract

Transglutaminase 2 (TG2), also referred to as tissue transglutaminase, plays crucial roles in both protein crosslinking and cell signalling. It is capable of both catalysing transamidation and acting as a G-protein, these activities being conformation-dependent, mutually exclusive, and tightly regulated. The dysregulation of both activities has been implicated in numerous pathologies. TG2 is expressed ubiquitously in humans and is localized both intracellularly and extracellularly. Targeted TG2 therapies have been developed but have faced numerous hurdles including decreased efficacy in vivo. Our latest efforts in inhibitor optimization involve the modification of a previous lead compound's scaffold by insertion of various amino acid residues into the peptidomimetic backbone, and derivatization of the N-terminus with substituted phenylacetic acids, resulting in 28 novel irreversible inhibitors. These inhibitors were evaluated for their ability to inhibit TG2 in vitro and their pharmacokinetic properties, and the most promising candidate 35 (k_inact/K_I = 760 × 10³ M⁻¹ min⁻¹) was tested in a cancer stem cell model. Although these inhibitors display exceptional potency versus TG2, with k_inact/K_I ratios nearly ten-fold higher than their parent compound, their pharmacokinetic properties and cellular activity limit their therapeutic potential. However, they do serve as a scaffold for the development of potent research tools.

A novel peptidic scaffold was used to design a library of inhibitors that exhibit exceptional efficiency against tissue transglutaminase, providing a framework for the development of potent research tools.

Introduction

Transglutaminases are a family of calcium-dependent multifunctional enzymes whose primary function involves the crosslinking of proteins and peptides through formation of an N^ε(γ-glutaminyl)lysine bond.^1,2 Transglutaminase 2 (TG2) is the most widely studied isozyme of the transglutaminase family, with human TG2 being expressed and active throughout the body.^3,4 The activity of this protein is tightly controlled by both redox reactions and calcium ions.^5–7 This isozyme is especially relevant to human physiology as it not only catalyses the crosslinking of proteins, but it also acts as a G-protein vital to signalling cascades within cells.^8–10 The two major activities of TG2, G-protein and crosslinking, are mutually exclusive, wherein the binding of GTP is impossible when the enzyme adopts its open, catalytically active conformation, and when the enzyme adopts its closed conformation to form a GTP binding site, the catalytic binding site for transamidation is occluded.^11–14 The multifunctional role of TG2 has also resulted in it being implicated in a variety of pathologies including the epidermal to mesenchymal transition of cancer stem cells, celiac disease, and fibrosis.^15–23

Targeted TG2 therapies have been developed to primarily block its transamidation activity. Some covalent inhibitors have been designed to abolish both the catalytic transamidation activity and the G-protein signalling activities by reacting irreversibly with the enzyme active site, and locking the enzyme in its open conformation.^24,25 One irreversible inhibitor produced by Zedira GmbH has recently completed phase II clinical trials for the treatment of celiac disease and has now progressed into phase II clinical trials for liver fibrosis as well.^26,27

Our group has published numerous novel activity substrates, probes, and inhibitors for TG2.^{24,25,28–32} Our more recent medicinal chemistry efforts have produced highly potent and efficient inhibitors that abolish both transamidation and GTP binding activities.²⁵ Furthering our medicinal chemistry efforts, herein we disclose 28 irreversible covalent inhibitors based on a novel peptidic scaffold. The top inhibitors from this new series exhibit efficiencies that are nearly an order of magnitude higher than those of our previous libraries.^25,33–38

Results and discussion

Design

The design for this library of TG2 inhibitors stemmed directly from three previous lead compounds, NC9, AA9, and NM72 (see Fig. 1). Inhibitor AA9 was shown²⁴ to have been optimized with respect to two features. The first feature was the alkyl side-chain linker to the acrylamide warhead, which appears to offer low off-target reactivity and high isozyme selectivity. The second design feature is the large hydrophobic naphthoyl group at the C-terminus. The naphthoyl moiety is thought to be bound in the hydrophobic pocket of TG2, sometimes referred to as the ‘D-site’, greatly improving affinity. Molecular modelling suggested that the incorporation of a spacer into the backbone of AA9, particularly an amino acid residue, may be beneficial to affinity by pushing the hydrophobic bulk of the naphthoyl group further into the D-site.

Fig. 1. Previous TG2 inhibitors NC9, AA9, and NM72 that represent starting points for this work.

Recently we published²⁵ our investigation of N-terminal variants of AA9, resulting in the identification of lead compound NM72. In the present work, we built upon what we learned from that N-terminal optimization, incorporating selective phenylacetyl groups on the N-terminus, exploring the synergistic relationship between an extended backbone scaffold and a high affinity N-terminus.

Synthesis

To generate the first library of N-terminal Cbz-spaced scaffolds, we envisioned a convergent synthetic strategy, where the varied amino acids would be incorporated in the C-terminal naphthoyl moiety (Scheme 1). First, amino acid spacers 1–10 were generated by acylation of various amino acids with 1-naphthoyl chloride. Subsequent amide couplings with N-Boc piperazine gave Boc protected intermediates 11–18. Acid mediated N-Boc deprotections yielded the corresponding amines 19–26. Finally, coupling with acrylated NHS ester 27 produced the peptidic inhibitors 28–35. Inhibitors 36 and 37 were prepared via a modified route which began with the acrylation of Cbz-Lys-OH followed by an EDC·HCl mediated amide coupling with N-Boc piperazine to yield 38. N-Boc deprotection to give 39 was followed by a second EDC·HCl mediated amide coupling to yield inhibitors 36 and 37.

Scheme 1. Synthesis of spacer-piperazine-naphthoyl inhibitors 28–37 a) 1-naphthoyl chloride, NaOH_(aq), THF, rt, 2–3 h, 23–75%; b) N-Boc piperazine, EDC·HCl, HOBt, DIPEA, DMF, rt, o/n, 28–81%; c) TFA, DCM, rt, 4 h; d) 27, Et₃N, DCM, rt, o/n, 24–46%; e) acryloyl chloride, NaOH_(aq), THF, 0 °C, 1 h then N-hydroxysuccinimide, EDC·HCl, MeCN, rt, o/n; f) acryloyl chloride, NaOH_(aq), THF, 0 °C, 0.5 h; g) N-Boc piperazine, EDC·HCl, HOBt, DMF, rt, o/n, 13% (over two steps); h) 10% TFA, DCM, rt, 2 h; i) 9/10, EDC·HCl, HOBt, DMF, rt, o/n, 20–28%.

To access ligands with an amino acid spacer between the acryloyl-lysine and the backbone piperazine, an alternative synthetic route was used (Scheme 2). Acylation of N-Boc piperazine with naphthoyl chloride followed by subsequent N-Boc removal gave key starting amine 40. Amine 40 was subjected to amide couplings with N-Boc protected amino acids to yield 41–43 – subsequent N-Boc deprotection gave amines 44–46. Finally, inhibitors 47–49 were accessed via reaction of amines 44–46 with activated NHS ester 27.

Scheme 2. Synthesis of spacer-piperazine-naphthoyl inhibitors 47–49. a) N-Boc amino acid, EDC·HCl, HOBt, DMF, rt, o/n 52–61%; b) TFA, DCM, rt, 2.5 h; c) 27, Et₃N, DCM, rt, o/n, 48–70%.

Finally, in order to probe the synergistic effects of combining backbone modification (Scheme 3) with results from a recent study on the N-terminal moiety effects,²⁵ we designed a synthetic scheme to incorporate select amino acid spacers between the piperazine and naphthoyl groups, as well as select substituted phenylacetyl groups at the N-terminus. Building on the promising structure–activity relationships (SAR) observed with l-/d-Ala and l-/d-Phe-containing inhibitors, 2, 3, 7, and 8 were subjected to HBTU-mediated amide couplings with N-Boc piperazine to yield N-Boc amines 50–53. Boc deprotection of 50–53 yielded TFA salts 54–57 which were subjected to HBTU-mediated amide couplings to give 58–61. Installation of the acrylate warhead group was achieved through Pd/C-catalysed hydrogenolysis of the Cbz protecting group from the lysine side chain to yield amines 62–65 – acrylation of amines 62–65 then gave access to acrylamides 66–69. N-Boc deprotection of acrylamides 66–69 gave TFA salts 70–73 that were subjected to HBTU-mediated amide couplings with various substituted phenylacetyl acids to yield inhibitors 74–88.

Scheme 3. Synthetic scheme to generate N-terminal substituted phenylacetyl inhibitors 101–115. a) N-Boc piperazine, HBTU, Et₃N, DMF, rt, o/n, 39–82%; b) TFA, DCM, rt, 4 h; c) Boc-Lys(Z)-OH, HBTU, DIPEA, DMF, rt, o/n, 61–82%; d) Pd/C, MeOH, rt, o/n; e) acryloyl chloride, DMAP, Et₃N 0 °C-rt, 2–4 h, 46–71%; f) TFA, DCM, rt, 4 h; g) substituted phenylacetic acid, HBTU, DIPEA, DMF, rt, o/n, 22–92%.

Kinetic evaluation

To measure the efficiency of the inhibitors synthesized herein, the irreversible inhibition of recombinant TG2 was monitored by a continuous chromogenic assay based on the use of substrate AL5 as detailed in the experimental section (Fig. 2A). Under Kitz and Wilson conditions, observed first-order rate constants of inactivation were measured and fitted to a saturation model to calculate k_inact and K_I (Fig. 2B).^39,40 As the efficiency of our inhibitor libraries has increased, we have approached the sensitivity limit of our assay. The inhibition of the enzyme at high [I] occurs at such a fast rate that calculating first order rate constants becomes unreliable. In these cases, we used linear regression to fit the rate constants measured at the lowest [I] to obtain a slope that corresponds to the k_inact/K_I efficiency ratio.

Fig. 2. Representative raw kinetic curves and fitting for inhibitor 35 (JA38) with activity assay substrate AL5 (α = 1 + [S]/K_M). A) Raw blank subtracted kinetic curves of absorbance versus time truncated to 3 half-lives. B) Saturation fitting of the observed first-order rate constants of TG2 inactivation versus the inhibitor concentration divided by alpha to acquire the inhibition parameters k_inact and K_I.

The first series of inhibitors tested herein incorporated an amino acid spacer residue between the piperazine and naphthoyl moieties (Table 1). These inhibitors all displayed inhibitor efficiency ratios greater than that of the parent AA9.²⁴ The lead compound from this first library, inhibitor 35 (aka JA38), displayed a k_inact/K_I ratio of 760 ± 200 × 10³ M⁻¹ min⁻¹ and featured a hydrophobic d-phenylalanine residue. Interestingly, another hydrophobic residue, l-alanine, was the second-best inhibitor from this series, with a k_inact/K_I ratio of 610 ± 120 × 10³ M⁻¹ min⁻¹, indicating that hydrophobic residues are well tolerated relative to glycine. Interestingly, the most efficient inhibitors from this series had the lowest K_I values but not the highest k_inact values. In contrast, compounds 31 and 32 have the highest k_inact values but also relatively high K_I values. In the context of the two-step inactivation mechanism, it appears that 35 is bound with high affinity in a pose that allows for moderate subsequent reaction, whereas 32 is bound more weakly, but in a pose that allows for more rapid subsequent reaction. This also illustrates the importance of considering the k_inact/K_I ratio as a measure of overall efficiency, reflecting binding affinity and the reactivity of the enzyme-inhibitor complex.

TG2 inhibition data corresponding to piperazine-spacer-naphthoyl inhibitors 28–37.

Inhibitor	A.A. spacer	K I (μM)	k inact (min−1)	k inact/KI (×103 M−1 min−1)
AA9 (ref. 25)	—	4.8 ± 1.2	1.2 ± 0.2	244 ± 75
28	Gly	0.3 ± 0.1	0.16 ± 0.02	513 ± 157
29	l-Ala	2.3 ± 0.4	1.40 ± 0.15	610 ± 120
30	d-Ala	1.3 ± 0.4	0.75 ± 0.09	590 ± 180
31	l-Ile	8.1 ± 2.4	2.40 ± 0.60	300 ± 120
32	l-Leu	12.8 ± 2.8	3.10 ± 0.60	240 ± 70
33	d-Leu	4.4 ± 1.6	1.40 ± 0.46	330 ± 160
34	l-Phe	2.1 ± 0.5	0.99 ± 0.15	470 ± 130
35(akaJA38)	d-Phe	1.0 ± 0.2	0.80 ± 0.09	760 ± 200
36	l-Val	2.9 ± 1.0	0.79 ± 0.20	272 ± 117
37	d-Val	1.3 ± 0.5	0.75 ± 0.15	595 ± 256

In order to determine if the amino acid spacer location was critical to inhibitor efficiency, three hydrophobic spacer derivatives were then inserted in between the lysine(acrylamide) and piperazine moieties (Table 2). Notable decreases in efficiency were observed for all three inhibitors 47–49. For derivative 47, presenting a glycine adjacent to the lysine(warhead) residue, a k_inact/K_I ratio of 330 ± 30 × 10³ M⁻¹ min⁻¹ was obtained, compared to the analogous inhibitor 28, which had the glycine spacer in the piperazine-spacer-naphthoyl position and showed a k_inact/K_I ratio of 513 ± 157 × 10³ M⁻¹ min⁻¹. Further, relative to l-Val inhibitor 36, interchanging the piperazine and valine residue resulted in a ∼5-fold drop in efficiency. And finally, displacing the d-Val residue of the highly efficient inhibitor 37 resulted in a decrease in efficiency of nearly an order of magnitude in inhibitor 49. These comparisons confirm that the piperazine-spacer-naphthoyl scaffold was clearly superior to the spacer-piperazine-naphthoyl alternative, and must be conserved in the next series.

Inhibition of TG2 data corresponding to spacer-piperazine-naphthoyl inhibitors 47–49.

Inhibitor	A.A. spacer	K I (μM)	k inact (min−1)	k inact/KI (×103 M−1 min−1)
47	Gly	2.3 ± 0.4	0.77 ± 0.07	330 ± 30
48	l-Val	5.1 ± 3.2	0.29 ± 0.12	56 ± 42
49	d-Val	3.4 ± 2.8	0.22 ± 0.11	65 ± 63

The final series of inhibitors examined whether a synergistic relationship could be obtained by combining the results from the first half of this SAR study with the results from our previous N-terminal SAR (Table 3).²⁵ Although this series did have efficiency ratios all above 100 × 10³ M⁻¹ min⁻¹, the lead inhibitor 87, bearing an N-terminal ortho-fluorophenylacetyl group and an l-alanine spacer, only reached an efficiency value of 502 ± 83 × 10³ M⁻¹ min⁻¹, which is significantly less than that of inhibitor 35 (JA38). It was noted, however, that the lead compound from this series did incorporate an ortho-fluorophenylacetyl terminus, mimicking that of the previous N-terminal SAR lead inhibitor NM72 (Fig. 1),²⁵ hinting that the ortho-fluorophenylacetyl derivative may be one of the more beneficial N-terminal groups for peptidic scaffolds of TG2 covalent inhibitors.

Inhibition of TG2 data corresponding to substituted N-terminal phenylacetyl inhibitors 74–88.

Inhibitor	N-terminal		A.A. spacer	K I (μM)	k inact (min−1)	k inact/KI (×103 M−1 min−1)
	X	Y
74	Me	H	l-Ala	9.5 ± 2.2	3.95 ± 0.77	418 ± 128
75	Me	H	d-Ala	14.9 ± 3.9	5.59 ± 1.27	374 ± 129
76	Me	H	l-Phe	8.8 ± 5.1	1.51 ± 0.71	171 ± 128
77	Me	H	d-Phe	n.d.	n.d.	238 ± 7a
78	Cl	H	l-Ala	9.3 ± 2.1	1.97 ± 0.36	211 ± 61
79	Cl	H	d-Ala	n.d.	n.d.	170 ± 3a
80	Cl	H	l-Phe	1.6 ± 0.9	0.25 ± 0.07	155 ± 102
81	Cl	H	d-Phe	n.d.	n.d.	127 ± 12a
82	Br	H	l-Ala	9.8 ± 2.1	2.69 ± 0.47	275 ± 75
83	Br	H	d-Ala	n.d.	n.d.	209 ± 5
84	Br	H	l-Phe	1.3 ± 0.4	0.24 ± 0.03	182 ± 56
85	Br	H	d-Phe	n.d.	n.d.	132 ± 8a
86	OMe	H	l-Ala	6.0 ± 1.1	1.52 ± 0.21	252 ± 59
87	F	H	l-Ala	3.3 ± 0.5	1.68 ± 0.15	502 ± 83
88	F	F	l-Ala	5.1 ± 1.2	1.16 ± 0.20	225 ± 67

^aData obtained by linear regression analysis.

Pharmacokinetic properties

In light of its excellent efficiency, inhibitor 35 (JA38) was selected for further evaluation with respect to certain in vitro pharmacokinetic properties (at Pharmaron, Inc.), in order to provide information relevant to predicting its therapeutic potential. The original scaffold inhibitor 28 was selected for the same evaluation, to provide a point of reference. As shown in Table 4, both inhibitors were found to have log P values just below 3, which bodes well for their bioavailability as well as their solubility. The passive diffusion of 28 across an artificial membrane (PAMPA) was good, and the permeability of 35 (JA38) was even better, suggesting that the structural changes guided by SAR would also improve cellular permeability. This permeability was confirmed directly in MDCK-MDR1 cells. In the presence of zosuquidar as a Pgp inhibitor, bilateral diffusion rates of both inhibitors were excellent. However, in the absence of this inhibitor, the efflux ratio for initial inhibitor 28 was high (15.5), and for optimised inhibitor 35 (JA38) it was even higher (96.7). Taken together, these results indicate that these inhibitors are capable of reaching intracellular targets; however, further optimisation may be required to decrease their affinity for the Pgp transporter in order to increase their bioavailability.

Pharmacokinetic data^a determined for inhibitors 28 and 35 (JA38).

Inhibitor	28	35 (JA38)
Log P	2.99	2.73
−Log Pe	5.33	4.98
P app(A-B) (×10−6, cm s−1)	0.55	0.06
P app(B-A) (×10−6, cm s−1)	8.43	5.46
Efflux ratio	15.5	96.7
P app(A-B) (×10−6, cm s−1) + Pgp Inh.	1.27	4.86
P app(B-A) (×10−6, cm s−1) + Pgp Inh.	1.59	4.90
Efflux ratio + Pgp Inh.	1.25	1.00

^aPharmacokinetic data measured by Pharmaron Inc.

Cellular evaluation

Inhibitor 35 (aka JA38), having shown the highest efficiency in our biochemical activity assay (Table 1), was also evaluated for efficacy in cellular assays against SCC-13 cells. These SCC-13 cells were originally derived from skin squamous cell carcinoma⁴¹ and obtained from ATCC. In these assays, the benchmark compound NC9 was included as a reference.^42,43 As shown in Fig. 3, inhibitor 35 (aka JA38) does inhibit the proliferation of SCC-13 cells in a dose-dependent fashion; however, it does not appear to be as potent as NC9 (structure shown in Fig. 1). Inhibitor 35 (aka JA38) was also found to inhibit SCC-13 invasion, as measured in a Matrigel migration assay (Fig. 4). It is interesting to note that while inhibitor 35 (aka JA38) is a much more efficient inhibitor than NC9, and its cellular permeability is higher, it is somewhat less potent at inhibiting cell growth (Fig. 3) but more potent at blocking invasion. Additional mechanism of action studies are underway to investigate the potential discrepancy between the highly enhanced biochemical efficiency of JA38 and its moderately improved biological activity.

Fig. 3. Dose-dependent inhibition of SCC-13 cell proliferation with NC9 and JA38 (see ESI†). Scale bars represent 50 μm.

Fig. 4. Dose-dependent inhibition of SCC-13 cell invasion with NC9 and JA38, measured by fluorescent detection according to the Matrigel invasion assay (see ESI†).

Conclusion

In summary, we have designed, synthesized and evaluated a new series of 28 peptidomimetic irreversible inhibitors of tissue transglutaminase. Among these, several stood out as having some of the highest efficiencies (k_inact/K_I ratios) of any in this class. Notably, inhibitor 35 (aka JA38) was more efficient than NM72, the top inhibitor from the last generation.²⁵ We confirmed the permeability of JA38, but also its susceptibility to Pgp-mediated efflux. We also demonstrated the ability of JA38 to inhibit the growth and invasion of SCC-13 cells. However, the biological activity of JA38, relative to hit compound NC9, is not as improved as one would expect, based on its efficiency in our biochemical assay with purified TG2. This suggests future mechanism of action studies may be necessary to shed light on this. Aside from the context of potential therapeutic application, we note the utility of this scaffold for the creation of future research tools for application in chemical biology. The tolerance of many different amino acid residues in the ‘piperazine-Xaa’ backbone is indicative of a synthetically accessible site that may be amenable to further modification in the development of future probes as powerful research tools.

Conflicts of interest

There are no conflicts to declare.