Three classes of GOAT inhibitors have been described so far: product acyl-peptide analogs, a small molecule detected in a high-throughput screen, and a rationally designed bisubstrate analog.
When Ser3 in ghrelin (, Compound 2) was replaced with DAP ((S)-2,3-diaminopropionic acid) creating an octanoyl-amide in place of ester (, Compounds 3 and 4), both the 28-mer and 5-mer acyl ghrelins were potent GOAT inhibitors with IC50 values of 0.2 and 1 μM, respectively. It is likely that these compounds correspond to the strong product inhibition of GOAT, but the lack of hydrolytic sensitivity of the amide linkage confers greater stability.
Figure 13.2 GOAT inhibitors. (A) Chemical structures of ghrelin and GOAT inhibitors. 1: Des-acyl ghrelin. 2: Acyl ghrelin. 3: Amide-linked octanoyl ghrelin. 4: Amide-linked 5-mer octanoyl ghrelin with C-terminus amidated. 5: Inhibitors discovered by Garner and Janda (more ...)
While showing high potency, product analogs have pharmacologic challenges for in vivo
applications. As peptide compounds, their ability to penetrate cell membranes may be limited. Perhaps more importantly, they are likely potent agonists of GHS-R1a. Four residues of ghrelin functionally activate GHS-R1a about as efficiently as full-length ghrelin (Bednarek et al., 2000
). We have also found that a Tat-conjugated 10mer-amide is also a potent GHS-R1a agonist (see below).
Garner and Janda (2011)
carried out compound screening using their click assay. The assay's Z′ factor was determined to be 0.63, indicating high assay quality (Zhang et al., 1999
). A small “credit card” library of drug-like small molecules was then screened for inhibition of GOAT, and two related small molecule inhibitors were discovered (IC50
= 7.5 and 13 μM
, respectively, see , Compound 5). Interestingly, these compounds contain six- and eight-carbon alkyl chains, suggesting that they possibly compete for the octanoic acid binding site on GOAT. Although these compounds have not yet been explored in depth pharmacologically, they appear to represent attractive leads.
3.1. Bisubstrate analogs
It is now well established that mimics of the transition state of an enzyme-catalyzed reaction can serve as high-affinity inhibitors based on the premise that most enzymes have evolved to bind tightly to the transition state. For enzymes that use two substrates in a ternary complex mechanism, an attractive approach to rational inhibitor design involves covalent linkage of the two substrates to generate a bisubstrate analog, as shown schematically in . Such compounds can show energetically favorable interactions with enzymes because dual occupancy of the substrate binding pockets is facilitated without the entropic penalty incurred with random collision of the individual substrate molecules. To be most effective, it is understood that a tether for the linkage must be able to approximate a mechanistically relevant orientation of the two substrates, ideally capturing elements of the transition state. In the best cases, bisubstrate analogs can show binding free energies to an enzyme that are equal to or greater than the sum of the binding energies of the individual substrate components to the same protein. Successful examples have been recorded of bisubstrate analogs for protein kinases and protein acetyltransferases inspired by enzyme mechanism considerations. By placing an acetyl bridge between ATP and peptide substrate sequences for kinases, compounds that show low micromolar to subnanomolar affinities have been achieved for the insulin receptor kinase, protein kinase A, Csk tyrosine kinase, cyclin-dependent kinase, Abl tyrosine kinase, and the epidermal growth factor tyrosine kinase (Bose et al., 2006
; Cheng et al., 2006
; Hines and Cole, 2004
; Hines et al., 2005
; Jencks, 1981
; Levinson et al., 2006
; Medzihradszky et al., 1994
; Parang et al., 2001
; Shen and Cole, 2003
A related linker worked effectively for the histone acetyltransferase enzymes PCAF/GCN5 and p300/CBP which contain a CoA and peptide substrate fragments bridged by an acetyl spacer (Lau et al., 2000
; Sagar et al., 2004
). Several of these bisubstrate analogs have been useful in structural analysis of the enzyme reaction mechanism and substrate binding features (Liu et al., 2008b
). On the other hand, these analogs have suffered from limited pharmacologic utility because of their large size, polarity, and the challenges of cell membrane penetration. However, the discovery of cell-penetrating peptide sequences derived from the HIV Tat protein have allowed for cell and in vivo
applications for the bisubstrate analog HAT inhibitors (Bricambert et al., 2010
; Cerchietti et al., 2010
; Cleary et al., 2005
; Guidez et al., 2005
; Liu et al., 2008c
; Marek et al., 2011
; Oussaief et al., 2009
; Spin et al., 2010
; Wang et al., 2011
; Zheng et al., 2005
3.2. Development of GO-CoA-Tat, a potent and selective bisubstrate inhibitor of GOAT
Following the bisubsrate analog approach described above, we reported the development of GO-CoA-Tat (Barnett et al., 2010
). GO-CoA-Tat (, Compound 6) uses nonhydrolyzable amide and thioether linkages to combine octanoyl-CoA with the first 10 amino acids of ghrelin, which are 100% conserved in mammals. An HIV Tat-derived peptide sequence was attached to the C-terminus using a flexible linker to allow cell penetration. GO-CoA-Tat and a set of related analogs and control compounds (, Compounds 6–11) were synthesized by a solid-phase strategy.
We then tested these compounds in HEK and HeLa cells expressing ghrelin and GOAT stably transfected with our phPPG-mGOAT vector. Cells were maintained in a medium supplemented with octanoic acid, pre-incubated with the compound for 24 h, and then lysed. Intracellular acyl and des-acyl proghrelin were measured by ELISA, with values validated using kits from two manufacturers and standards made inhouse. We first tested GO-CoA-Tat in these models, and the mean inhibitory concentration was ~5 μM; control compound D4-Tat had no effect. Interestingly, maximum inhibition was achieved only after 24 h of incubation with GO-CoA-Tat. This could be due to the atypical behavior of the enzyme or inhibitor or due to preexisting intracellular stores of acyl ghrelin. To test this, we used the radioassay described in Section 2.3 and found substantial inhibition occurred within 5 min with 100 nM GO-CoA-Tat. This too suggests that there are intracellular stores of acyl ghrelin in these cells.
We examined structure–activity relationships required for inhibition with compounds used at 6 μM. Consistent with what was seen in other assays, five residues of ghrelin were sufficient for inhibition but three residues were not. Inclusion of 10 residues increased potency, with a maximum of ~75% inhibition of acylation seen. CoA was also required for inhibition; this finding is discussed further in Section 4.3, and a version of the bisubstrate compound with a truncated two-carbon acyl group still showed some inhibition. Tat was required for inhibition, consistent with its role in entry into the cells and ruling out action on a surface receptor. None of the compounds was toxic to the cells in the low micromolar concentration range. GO-CoA-Tat's specificity is also reflected in its lack of inhibition of three acetyl-CoA-utilizing enzymes.
To further analyze GO-CoA-Tat's inhibition of GOAT, we developed a direct binding assay for GOAT, taking advantage of photocrosslinking technology. We first synthesized two chemically modified versions of our bisubstrate inhibitor, namely, GO-CoA-Tat-F4BP and GO-CoA-Tat-L5BP, in which Phe4 or Leu5, respectively, is replaced with a photoreactive amino acid benzoyl-phenylalanine and each is tagged with a biotin group (Barnett et al., 2010
). We showed that this compound could covalently crosslink to recombinant solubilized or microsomal GOAT. This crosslinking could be blocked by an excess of unlabeled GO-CoA-Tat, providing evidence for specificity and demonstrating direct binding of GO-CoA-Tat to GOAT.
For these experiments, we used GOAT with a C-terminal 3xFlag tag, produced in SF9 cells using baculovirus. Microsomes were prepared as above, and the reaction was performed either in the microsome membranes or with GOAT purified to homogeneity using anti-Flag affinity chromatography and the Fos-Choline-16 detergent (Anatrace). This detergent was chosen because of its high ability to solubilize GOAT and because we reasoned that the long alkyl chain is less likely to interfere with the octanoic acid binding site on GOAT.
Photocrosslinking reactions were performed in a small water-jacketed quartz cuvette, custom made for this purpose by Quark Glass. The cuvette was connected to the water line and suspended above a magnetic stir plate. A small teflon stir bar was added, with medium agitation. A mercury UV lamp with a ~360-nm long-wave filter, such as UVP #B-100AP, was positioned with the center of the lamp approximately 2 cm from the cuvette, positioning the sample at the position of peak intensity. A time course experiment (not shown) demonstrated that the reaction had neared completion by 30 min. Crosslinked membranes were then solubilized and immunoprecipitated. Biotinylation was visualized using SDS-PAGE and streptavidin–HRP or, for more sensitivity, streptavidin followed by polyclonal anti-streptavidin.
3.3. Glucose and weight control in mice with GO-CoA-Tat
Treatment of C57BL6 mice on medium-chain triglyceride (MCT) diets (Kirchner et al., 2009
) with GO-CoA-Tat at 40 mg/kg dose, but not with the control compound D4-Tat or vehicle, decreased plasma acyl ghrelin levels without changing the des-acyl ghrelin levels. Maximum inhibition was seen after 6 h, but some acyl ghrelin suppression was still detectable 24 h after GO-CoA-Tat treatment. Because of the daily fluctuations between animals and ad lib
feeding, we found that the acyl to des-acyl ghrelin ratio was a more sensitive and specific measure of inhibition.
We explored the effect on weight gain over a 1-month period in mice placed on an MCT diet. Daily IP injections of GO-CoA-Tat as above reduced the weight gain seen in vehicle-treated mice. As measured by QMR spectroscopy, the difference in weight was due to significantly reduced fat mass in the GO-CoA-Tat-treated animals. In contrast, GO-CoA-Tat- versus vehicle-treated ghrelin-knockout mice showed no statistically significant difference in weight or body composition. To investigate the potential for GO-CoA-Tat toxicity, we examined the blood chemistries and cell counts in the mice after 1 month of treatment with the agent. There was no apparent untoward effect on normal blood chemistries or cell counts under these conditions. Interestingly, WT mice treated with GO-CoA-Tat showed reduced IGF-1 and lower blood glucose, consistent with suppression of ghrelin-mediated somatotroph signaling.
To investigate the role of acute pharmacologic inhibition of acyl ghrelin in insulin signaling and glucose homeostastis, we pretreated with GO-CoA-Tat and then measured the response to a glucose challenge, first in isolated pancreatic islets and then in mice. The insulin response was increased in islets and mice, where the response was accompanied by reduced blood glucose. In contrast, there was no effect when the studies were repeated in ghrelin-knockout animals, suggesting that GO-CoA-Tat's effects on insulin are due to the inhibition of ghrelin acylation. Finally, we showed by quantitative PCR that islets isolated from mice pretreated with GO-CoA-Tat had a 20-fold reduction in expression of uncoupling protein 2 mRNA (UCP2
, which suppresses insulin secretion), but there was no change in UCP2
expression in the gastric fundus. Together, these data show a tissue-specific role for GOAT inhibition in augmentation of insulin secretion. Regulation of UCP2
also highlights the importance of ghrelin acylation in obesity and type 2 diabetes, underscoring the need for more drug-like GOAT inhibitors (Andrews et al., 2008
; Dezaki et al., 2008
; Joseph et al., 2002
; Sun et al., 2006
; Tong et al., 2010
; Zhang et al., 2001