Substrate-Selective Small-Molecule Modulators of Enzymes: Mechanisms and Opportunities

Abstract

Small-molecule inhibitors of enzymes are widely used tools in reverse chemical genetics to probe biology and explore therapeutic opportunities. They are often compared with genetic knockdown or knockout and are expected to produce phenotypes similar to the genetic perturbations. This review aims to highlight that small molecule inhibitors of enzymes and genetic perturbations may not necessarily produce the same phenotype due to the possibility of substrate-selective or substrate-dependent effects of the inhibitors. Examples of substrate-selective inhibitors and the mechanisms for the substrate-selective effects are discussed. Substrate-selective modulators of enzymes have distinct advantages and cannot be easily replaced with biologics. Thus, they present an exciting opportunity for chemical biologists and medicinal chemists.

Introduction

Reverse genetics and reverse chemical genetics.

The term chemical genetics refers to the use of libraries of small molecules to discover interesting biological observations and then discover the protein targets of the small molecules (Figure 1).[1] This is analogous to classical genetics, in which random mutagenesis is used to find interesting phenotypes followed by identifying the mutated genes (Figure 1). With the availability of sequenced genomes, reverse genetics becomes more popular, in which individual genes are mutated, knocked down, or knocked out, to examine phenotypes. Similarly, reverse chemical genetics, which develops small molecules targeting defined proteins and use them to probe biology, are also increasingly used in the chemical biology community (Figure 1). [1]

Figure 1.

Scheme showing the concepts of forward and reverse genetics as well as forward and reverse chemical genetics. The topic of this review is more related to reverse genetics and reverse chemical genetics. The juxtaposition of reverse genetics and reverse chemical genetics make it easy to equalize the effect of small molecule inhibitors to that of genetic knockdown or knockout. This review emphasizes that this is not necessarily the case and small molecule inhibitors of proteins may produce phenotypes that are different from that of genetic knockout/knockdown.

Because of the similarity between reverse genetics and reverse chemical genetics, it is easy to assume that small molecule inhibitors of a protein will produce effects equivalent to knockdown or knockout of the corresponding gene. This is reflected by the fact that we are almost always required during the peer review process to present genetic knockdown or knockout data to corroborate the effect of small molecule inhibition data. While in many cases, small molecule inhibitors and knockdown/knockout do produce similar phenotypes, this review aims to highlight that the two phenotypes do not have to match and rationalize an important reason behind it.

Substrate-dependent or substrate-selective inhibition.

Shokat and colleagues have previously pointed out that small molecule inhibition and genetic knockdown may not align, and proposed that the additional scaffolding roles of the proteins being inhibited as the explanation (small molecule inhibitors will not affect the scaffolding role but the knockdown will).[2] While this is an important possible explanation, here I will focus on another explanation, substrate-dependent or substrate-selective inhibition of enzymes, for the differential phenotypes produced by small molecule inhibition and genetic perturbation.

Many enzymes, especially those that catalyze protein post-translational modifications, work on many different substrates. Examples in the literature demonstrate that it is possible to develop small molecule inhibitors that only inhibit the activity of enzymes on some, but not all, of the substrates. This substrate-dependent or substrate-selective inhibition may explain the differential effects of small molecule inhibition and genetic perturbation, as genetic knockdown or knockout will affect all the substrates.

As will be seen later, some small molecules could inhibit the enzyme’s activity on certain substrates, but activate the activity on other substrates. Thus, I also use the terms “modulation” and “modulators” to more broadly include small molecules that could either activate or inhibit an enzyme.

Examples of substrate-dependent or substrate-selective modulation.

With this consideration, I searched PubMed for examples of substrate-dependent or substrate-selective inhibitors. There are actually many examples in the literature (Table 1) and these include small molecule modulators for PTM enzymes (kinases, proteases, and deacylase) as well as enzymes or transporters that work on organic compounds.

Table 1.

Examples of substrate-dependent modulator of enzymes/transporters.

Enzymes	inhibitors/activators	substrates	effects	Ref
oxygenation enzymes
CYP2C9		(S)-Flurbiprofen; naproxen	Phenylsulfone activates reaction with flurbiprofen but inhibits reaction with naproxen; dapsone activates both.	[3]
CYP3A4		Testosterone, terfenadine, midazolam, and nifedipine	These substrates affect each other in a substrate-dependent matter. nifedipine inhibited testosterone 6β-hydroxylation in a concentration-dependent manner, testosterone did not inhibit nifedipine oxidation	[4]
CYP3A		Testosterone, midazolam, and nifedipine	Erlotinib inhibited testosterone and nifedipine oxidation, but stimulated midazolam oxidation	[5]
COX-2		arachidonic acid (AA) and the endocannabinoids 2-arachidonoylglycerol (2-AG) and arachidonoylethanolamide (AEA).	R-profens only inhibits the oxygenation of the endocannabinoids. COX-2 functions as a dimer with one serving as an allosteric subunit. The binding of R-profen in the allosteric subunit inhibits the catalytic subunit’s ability to process 2AG and AEA, but not AA. In contrast, those more potent (slow tight binding) COX-2 inhibitors occupy both subunits and thus inhibit both types of substrates.	[6]
COX-2		AA and 2-AG, AEA	AM8138 increases activity on 2-AG, but not AA.	[7]
COX-2		AA and 2-AG, AEA	Inhibits activity on 2-AG and AEA, but not on AA. reduced anxiety-like behaviors in mice via increased endogenous cannabinoid signaling	[8]
proteases
Thrombin		protein C and fibrinogen	LY254603 promoted protein C cleavage but inhibited fibrinogen cleavage, thus modulating thrombin’s anticoagulant and procoagulant functions.	[9]
ADAM10		glycosylated peptide substrate versus nonglycosylated peptide substrate in vitro; CXCL16 and Syndecan-4 in cells.	CID 3117694 inhibits glycosylated peptide substrate cleavage selectively. In cells it inhibits CXCL16 cleavage, but not syndecan-4 cleavage. It inhibits cell migration and wound closure.	[10]
MMP-2	P713 (a peptide that binds the CBD domain of MMP-2)	gelatin	P713 inhibits gelatin cleavage, but does not have much effect on a peptide substrate that does not require CBD binding.	[11]
ADAM17		heregulin and TGFα	Compound 17 inhibits shedding of heregulin but not TGFα in cells.	[12]
IDE		insulin and Aβ	Compound 5 inhibits Aβ degradation but not insulin degradation.	[13]
IDE		insulin and glucagon	Inhibit insulin cleavage preferentially over glucagon because they bind the exosite, not the catalytic site.	[14]
gamma-secretase		Amyloid beta precursor protein (APP) and Notch	ELN-475516 preferentially inhibits APP cleavage over Notch cleavage	[15]
gamma-secretase		APP and Notch	Preferentially inhibits APP cleavage over Notch cleavage. Interestingly, it occupies the active site similar to Semagacestat, which is not selective.	[16]
sirtuins
SIRT2		acetyl peptide and myristoyl peptide in vitro	Preferentially inhibits the hydrolysis of acetyl peptide over myristoyl peptide in vitro. In cells, it inhibits SIRT2-catalyzed tubulin deacetylation and ARF6 demyristoylation, but cannot inhibit SIRT2-catalyzed KRas4A depalmitoylation.	[17–19]
SIRT2		acetyl peptide and myristoyl peptide	Preferentially inhibits the hydrolysis of acetyl peptide.	[20]
SIRT2		acetyl, 4-oxononanoyl, decanoyl, myristoyl	Inhibits SIRT2’s activity on acetyl and myristoyl peptides, but activates on 4-oxononanoyl and not much effects on deacnoyl peptide.	[21]
SIRT2		acetyl, 4-oxononanoyl, decanoyl, myristoyl	Inhibits SIRT2’s activity on acetyl and myristoyl peptides, but activates on decanoyl and not much effects on 4-oxononanoy peptide.	[21]
SIRT6		acetyl peptide and myristoyl peptide	Fatty acids promote the hydrolysis of acetyl peptide but inhibits the hydrolysis of myristoyl peptide.	[22]
SIRT6		acetyl peptide and myristoyl peptide	Promote the hydrolysis of acetyl peptide but have no effect on the hydrolysis of myristoyl peptide. Shown to work in cells and mice.	[23,24]
SIRT6		acetyl peptide and myristoyl peptide	Promotes the hydrolysis of acetyl peptide but has no significant effect on the hydrolysis of myristoyl peptide. Tested in vitro only.	[25]
SIRT6		acetyl peptide and myristoyl peptide	CL-5D promotes the hydrolysis of acetyl peptide but inhibits the hydrolysis of myristoyl peptide. Used for in vitro studies only. CL-5D does not increase the binding affinity of acetyl-peptide to SIRT6. Instead, it may promote the chemistry step or confirmation change that is required for the chemistry.	[26]
SIRT6			Promote the hydrolysis of acetyl peptide but not myristoyl peptide. They react with Cys18 of SIRT6. Work at cellular level.	[27]
kinases
mTOR		RPS6 and 4EBP1	Rapamycin inhibits phosphorylation of RPS6, but not 4EBP1, unlike RapaLink-1, which inhibits both. RapaLink1 also inhibits mTORC2. Rapamycin and Rapalink-1 have opposite effects on starvation resistant.	[28–31]
PDK1		S6K and AKT	PS210 binds to the PIF hydrophobic pocket of PDK1, which is required for docking certain substrates, such as S6K. In cells, treatment with the prodrug form PS430 leads to inhibition of S6K phosphorylation, but not AKT phosphorylation, which does not require the PIF pocket.	[32]
p38a		STAT1, MK2, ATF2	Inhibition of STAT1 phosphorylation in cells is the most obvious, that of MK2 and ATF2 is minor. UM101 reduced LPS-/hyperthermia-induced acute lung injury. In HMVECL cells, it modulates the TNFα-induced gene expression differently from SB203580 (an inhibitor that is not substrate-selective) but there is significant overlap. UM101 is proposed to bind the glutamate-aspartate (ED) domain of p38a, which is important for binding to certain substrates.	[33]
CK2		CK2β-dependent and independent substrates	These compounds inhibit the interaction between α and β subunits, thus selectively inhibit the phosphorylation of substrates that depends on the CK2 α,β complex.	[34]

Modulators for oxygenation enzymes.

Some of the early examples of substrate-dependent modulators comes from the drug metabolism field, where it was observed that certain small molecules could affect the drug-metabolizing cytochrome P450 enzymes in a substrate-dependent manner. For example, phenylsulfone can activates the oxidation of (S)-flurbiprofen but inhibits the oxidation of naproxen by the P450 enzyme CYP2C9, while dapsone activates the oxidation of both substrates. [3] Similarly, Erlotinib, an inhibitor for the receptor tyrosine kinase EGFR, inhibits the oxidation of some substrates but stimulated the oxidation of others by another P450 enzyme CYP3A.[5] These substrate-dependent inhibitions are important factors that determine drug-drug interaction.

Cyclooxygenase 2 (COX-2) is an oxygenation enzyme that play important roles in inflammation and pain signaling. Substrate-selective inhibitors for COX-2 have been well demonstrated and may offer unique therapeutic opportunities.[8] COX-2 can convert arachidonic acid (AA) to prostaglandins, which have important roles in inflammation. COX-2 can also convert the 2-arachidonoylglycerol (2-AG) and arachidonoylethanolamide (AEA) to the corresponding prostaglandin derivatives. 2-AG and AEA are called endocannabinoids because they are endogenous ligands for the cannabinoids receptors, which produces certain beneficial effects (e.g. reducing neuropathic pain and increasing appetite) upon stimulation. The selective inhibition of COX-2’s ability to metabolize 2-AG and AEA thus could increase cannabinoids receptor signaling and produce desirable effects. Many small molecule inhibitors have been shown to selectively inhibit COX-2’s activity on 2-AG and AEA without affecting its activity on AA (Table 1).[6,7,35] The (S) enantiomers of naproxen, ibuprofen and flurbiprofen (commonly used over-counter pain and fever medication) inhibits COX-2’s activity on AA, but the (R) enantiomers typically lacks activity on AA, but interestingly can inhibit COX-2’ activity on 2-AG and AEA.[6] Additionally, converting the carboxylic acid moiety in indomethacin to a morpholino amide generates a compound LM-4131 that can selectively inhibit COX-2’s activity on 2-AG and AEA, but not on AA.[35] LM-4131 reduced anxiety-like behaviors in mice via increased endogenous cannabinoid signaling, suggesting that substrate-selective inhibitors of COX-2 could have interesting therapeutic applications. 13(S)-Methylarachidonic acid (AM-8138) is also shown to selectively activate COX-2’s activity on 2-AG without affecting its activity on AA.[7]

Protease inhibitors.

Many proteases have substrate-selective or substrate-dependent inhibitors reported (Table 1). One of the earlier examples is a thrombin modulator. Thrombin is a serine protease that is important for the blood coagulation process via cleaving fibrinogen. However, it also has anti-coagulation activity via proteolytic activation of protein C. Lilly Research Laboratory developed LY254063, which inhibits thrombin’s activity on fibrinogen but promotes its activity on protein C.[9] Such substrate-selective thrombin modulators may be therapeutically useful as anticoagulation agents.

The gamma-secretase cleaves amyloid precursor protein (APP) and promotes the generation of Aβ. Thus, small molecule inhibitors of gamma-secretase have been the targets of many medicinal chemists. However, gamma-secretase also cleaves other substrates, notably Notch, which has important roles in cellular development. Gamma-secretase inhibitors that preferentially inhibit the cleavage of APP but spares Notch are thus highly desirable. Two compounds ELN475516 and Avagacestat have been report to inhibit APP cleavage but not Notch.[15,16] Avagacestat was tested in clinical trials for treating Alzheimer’s diseases, but unfortunately did not produce beneficial effects.[36]

A disintegrin and metalloproteinase domain-containing protein 10 (ADAM10) and ADAM17 are integral membrane metalloproteases that are responsible for the cleavage of many cell surface proteins, including those important for immune responses, such as TNFα. The substrate-selective targeting is thus of considerable interest for treating cancer, inflammation and neurodegeneration. An ADAM10 inhibitor, CID 3117694, is shown to inhibits glycosylated peptide substrate cleavage selectively in vitro.[10] In cells it inhibits CXCL16 cleavage, but not syndecan-4 cleavage, and inhibits cell migration and wound closure. An ADAM17 inhibitor, compound 17, can selectively inhibits the cleavage of heregulin over TGFα in cells.[12]

Insulin degrading enzyme (IDE) is another protease with multiple physiological substrates, including insulin, glucagon, and APP-derived amyloidogenic peptides. Selectively inhibiting IDE’s activity on insulin but not glucagon could potentially be a useful strategy to treat type II diabetes. Using high-throughput screening and further lead development, several small molecule inhibitors (i.e. compound 37 and 63) of IDE were obtained that shown selective inhibition of insulin cleavage over glucagon cleavage.[14] Another IDE inhibitor, compound 5, inhibited the cleavage of APP peptide but promoted the hydrolysis of insulin.[13] These examples raise the possibility of developing substrate-selective modulators of IDE for different diseases.

Sirtuin modulators.

The sirtuin family of enzymes are NAD⁺-dependent protein lysine deacylases. They can remove different acyl lysine modifications, such as acetyl, succinyl, and long-chain fatty acyl groups, and regulate numerous proteins via their lysine deacylation activity.[37] In terms of substrate-dependent modulators, two sirtuin family members, SIRT2 and SIRT6, are particularly interesting. They both can remove multiple acyl lysine modifications, including the short acetyl and the long-chain fatty acyl groups (e.g. decanoyl, decanoyl, myristoyl and palmitoyl).[37] Their ability to work on both the very short and long acyl groups made it possible to have some interesting substrate-dependent modulators.

Several SIRT2 inhibitors have been found to preferentially inhibit the deacetylation activity of SIRT2, with no or very weak effects on the demyristoylation activity of SIRT2. A mechanism-based thiomyristoyl lysine compound (TM) can potently inhibit the deacetylation activity of SIRT2 in vitro, but the inhibition potency on the demyristoylation activity is much weaker.[17] NPD-11033, a small molecule SIRT2 inhibitor discovered through high-throughput screening, was shown to selectively inhibit SIRT2’s deacetylation activity.[20] Several other SIRT2 inhibitors, such as AGK2 and SirReal2, were also found to preferentially inhibit the deacetylatin activity of SIRT2, both in vitro and in cells.[18]

SIRT2 has also been reported to remove some other acyl lysine modifications, such as 4-oxononanoyl. It is reported that in vitro 1-aminoanthracene can inhibit the deacetylation and demyristoylation activity of SIRT2, but interestingly can activate its activity on 4-oxononanyl lysine.[21] Another compound, propofol, inhibits SIRT2’s activity on acetyl and myristoyl peptides, but activates SIRT2 on decanoyl and not much effects on 4-oxononanoy peptide.[21] However, whether these compounds exhibit such effects in cells is not clear.

SIRT6 modulators also have very interesting substrate-dependent effect. Long-chain fatty acids, such as palmitic acid, can activate the deacetylation activity of SIRT6 but inhibits the demyritoylation activity in vitro.[22] Stimulated by this observation, many SIRT6 modulators have been developed (Table 1) that can promote deacetylation but not demyristoylation. Two types of these SIRT6 modulators have been shown to work at cellular levels, MDL-800/801/811 and nitro fatty acids.[23,24] They can increase histone H3 acetylation and MDL-800/801/811 have been shown to have anticancer and anti-inflammation properties.[24,38]

Protein kinase inhibitors.

While most protein kinase inhibitors are active site-directed and do not have substrate-selective inhibition, a few inhibitors are reported to have substrate-dependent effect. Rapamycin, an important molecule that catalyzed the field of chemical biology and is clinically used as an immunosuppressant,[39] is in fact a substrate selective inhibitor.[28] Rapamycin inhibits the mammalian target of rapamycin (mTOR), a protein kinase that plays important roles in regulating cell growth. mTOR exists in two complexes, mTORC1 and mTORC2. Rapamycin selectively inhibits mTORC1. mTORC1 has multiple protein substrates, and rapamycin only inhibits the phosphorylation of some of these substrates. For example, rapamycin reduced S6K1 T389 phosphorylation, but not 4EBP1 T37/46 phosphorylation.[28] In contrast, a third-generation of mTOR inhibitor, Rapalink-1 (rapamycin linked to a kinase active site binder MLN0128), inhibits all the substrates of mTORC1 and also inhibits mTORC2.[29,30] Thus, rapamycin is a substrate-selective inhibitor of mTORC1. This substrate-selective inhibition of mTORC1 may underlies the beneficial effect of rapamycin. For example, rapamycin treatment makes flies more resistant to starvation while rapalink-1 have the opposite effect.[30]

PDK1 is a kinase that is upstream of mTORC1 and also plays important role in regulating cell growth and proliferation. PS210 is a substrate-selective inhibitor of PDK1.[32] PS210 binds to the PIF hydrophobic pocket of PDK1, which is required for docking certain substrates, such as S6K, but not others. In cells, treatment with the prodrug form of PS210, PS430, lead to inhibition of S6K phosphorylation, but not AKT phosphorylation, which does not require the PIF pocket.

Casein kinase 2 (CK2) is a serine/threonine protein kinase that phosphorylate and regulate numerous proteins and cellular pathways. It can either exist as a standalone catalytic subunit (CK2α or α’) dimer, or in a tetrameric complex with two regulatory CK2β subunits.[34] The complexation with the CK2β subunits affects its substrate profile. Thus, small molecules that disrupts the CK2α-CK2β protein-protein interaction may selectively inhibit the phosphorylation of substrate proteins that depends on the CK2α-CK2β complex. Two CK2 inhibitors, compound 1 and 6 (Table 1), are reported to disrupt the CK2α-CK2β complex, although the substrate-dependent inhibition of CK2 by these compounds is not been well demonstrated.[34]

Mechanisms of substrate-dependent inhibition

Learning from the examples listed in Table 1, I summarize a few mechanisms that can lead to substrate-selective or substrate-dependent modulation of enzymes. Some of the mechanisms are complex and not well understood, but the generalization below may provide useful guiding principles for future development of substrate-selective enzyme modulators.

Targeting exosites.

Exosites of enzymes refer to the substrate binding sites that are outside of the active site. Exosites provide binding affinity for substrates and thus are important for recruiting substrates to the enzyme active site. The reason why targeting exosites could provide substrate-selective inhibition is that different substrates may have different requirements for the exosites (Figure 2). A very nice example of this is the IDE substrate-selective inhibitors 37 and 63.[14] An X-ray crystal structure of 63 in complex with IDE and the glucagon substrate shows that inhibitor 63 and glucagon bind IDE at the same time, at the exosite and active site, respectively, which explains why 63 does not inhibit glucagon cleavage by IDE. In contrast, comparison with the IDE structure with insulin bound reveals that 63 binding would negatively affect insulin binding, explaining why 63 would inhibit insulin cleavage. The inhibitors for ADAM10 (CID 3117694), ADAM17 (compound 17), MMP-2 (P713), and IDE (compound 5) also fall into the exosite targeting category. The kinase inhibitors for PDK1 (PS210) and p38 (UM101) also works by targeting exosites of the corresponding kinases.

Figure 2.

Exosite binding leading to substrate-selective inhibition. (A) Cartoon showing how targeting an exosite could lead to substrate-selective inhibition of an enzyme. An inhibitor (blue triangle) targeting the exosite of the shown enzyme selectively blocks substrate 2 (gold) from binding, leading to the selective inhibition of the enzyme on substrate 2. (B) Comparison of human IDE structures with inhibitor 63 (cyan) and glucagon (purple) bound (PDB 6EDS) and IDE with insulin (gold) bound (PDB 2WBY) shows that inhibitor 63 binds to a site that does not clash with glucagon, but would clash with insulin. Inhibitor 63 in the IDE-insulin structure on the right is from the superimposed PDB 6EDS structure.

Targeting allosteric sites.

Allosteric regulation refers to the regulation of an enzyme by binding of an effector molecule at a site that is different from the active site. Thus, allosteric sites are similar to exosites, but allosteric sites are not bound by substrates. However, there are cases where allosteric sites and exosites are hard to differentiate (see mTOR and COX-2 discussion below, as well as the SIRT6 discussion further below). Allosteric regulation can tune the binding or activity of different substrates, thus achieve substrate-selective modulation (Figure 3).

Figure 3.

Allosteric site binding leading to substrate-selective inhibition. (A) Cartoon showing how targeting an allosteric site could lead to substrate-selective inhibition of an enzyme. An inhibitor binding to the allosteric site of the shown enzyme lead to slight changes in the active site and this structural change selectively blocks substrate 1 from binding, while promotes or does not inhibit substrate 2 from binding/reacting. (B) Superimposed mTOR structures from PDB 5FLC (mTORC1 complex with rapamycin and FKBP12) and 4JSX (truncated mTOR with mLST8 and Torin-2) showing that rapamycin binds to an allosteric site that is distinct from the mTOR kinase active site that is occupied by Torin-2. (C) Superimposed SIRT6 structures from PDB 3ZG6 (SIRT6 with ADP-ribose and myritoyl-lysine substrate) and 6XVG (SIRT6 with MDL-801) showing that MDL-801 occupies the myristoyl binding pocket, which inhibits myritoyl-lysine substrate from binding to SIRT6, but promote acetyl-lysine substrate binding to SIRT6. The MDL-801 binding site can thus be considered an allosteric site for the acetyl-lysine substrate and an exosite for myristoyl-lysine. ADPR: ADP-ribose. Myr-Lys: myristoyl-lysine.

The thrombin inhibitor, LY254603, likely works by binding to an allosteric site. It is developed based on the rationale that thrombin is regulated by allosteric mechanism that tunes its activity on different substrates.[9]

The mTOR inhibitor rapamycin is the one of most well-known allosteric inhibitors. It binds FKBP12 and the rapamycin-FKBP12 complex then binds to mTORC1 at the location that is distant from the active site.[40,41] How this allosteric site binding achieves substrate-selective effects is not entirely clear. It is possible that the binding of some substrates directly conflicts with the rapamycin-FKBP12 binding. In this case, it could be an exosite binding. It could also be a truly allosteric mechanism if the rapamycin-FKBP12 binding subtly changes the active site of mTORC1, which weakens the binding/activity of certain substrate but not others.

COX-2 substrate-selective inhibitors mentioned in Table 1 is thought to work via an allosteric mechanism. COX-2 works as a homodimer and it is believed that one subunit of the dimer serves as an allosteric regulatory subunit for the other catalytic one. The binding of the substrate selective inhibitors to the substrate binding pocket of the allosteric subunit produces subtle changes in the active site of the catalytic subunit, making it unable to process 2-AG and AEA but still able to process AA.[6] To fully inhibit COX-2 and its ability to process AA, a more potent inhibitor is needed to occupy the substrate binding pocket on the catalytic subunit. In the COX-2 case, the allosteric site also happens to be a substrate-binding active site, but it works in trans to regulate the activity of another subunit in the dimer.

Targeting subtle differences in the binding pockets of substrates (a hybrid of exosite and allosteric site targeting).

The binding pockets of different substrates at the active site may differ, and in that case, if a small molecule occupies the differential part of the pockets, it may prevent one substrate from binding, but at the same time promote another substrate to bind or react. This is likely the case for the SIRT6 modulators (such as palmitic acid and CL-5D, Table 1) that can activate deacetylation but inhibits demyristoylation. SIRT6 can bind acetyl lysine substrate or myristoyl lysine substrate at the same active site, but the longer myristoyl group occupies a hydrophobic pocket that is not occupied by the acetyl group. The substrate-selective activators of SIRT6 likely binds to this pocket, which has been demonstrated for the MDL series and UBCS039 (Table 1).[23,25,42] The binding promotes the deacetylation reaction, leading to the selectively activation of the deacetylation activity of SIRT6. At the same time, the binding could inhibit myristoyl peptide from binding to the SIRT6 active site, leading to the inhibition of the demyristoylation activity, which has been shown for palmitic acid and CL-5D.[22,26] The SIRT6 modulator case is in principle is a hybrid of exosite and allosteric targeting, as the site being targeted is both a myristoyl substrate-binding exosite and also an allosteric site for the acetyl substrate. The SIRT2 modulators, 1-aminoanthracene and propofol (Table 1), likely works in a similar fashion.

Binding affinity differences in substrates.

Two substrates may have very different binding affinities for the same enzyme. If a small molecule inhibitor’s binding affinity falls between these two substrates’ binding affinities, then the inhibitor may preferentially inhibit the substrate with the lower binding affinity (Figure 4). For the SIRT2 inhibitor TM,[17] NPD-11033,[20] AGK2,[43] and SirReal2,[44] this is likely the case. SIRT2 can hydrolyze both acetyl lysine peptides and myristoyl lysine peptides and these inhibitors are much better at inhibiting SIRT2’s deacetylation activity. This is because acetyl peptide binds to SIRT2 less tightly compared to myristoyl peptides.[45] Thus, acetyl peptide can be more easily displaced by the inhibitors. In this case, the inhibitors should be competitive with the substrates (substrate-competitive inhibitors).

Figure 4.

Cartoon showing how active-site targeting inhibitors could lead to substrate-selective inhibition of an enzyme. An inhibitor binding to the active site of the shown enzyme prevent the low-affinity substrate from binding, but does not prevent the high-affinity substrate from binding to the enzyme, leading to selective inhibition of the low affinity substrate.

Surprisingly, this may also apply to rapamycin’s substrate-dependent inhibition of mTOR. It is reported that substrates that are more efficiently phosphorylated by mTORC1 are also more resistant to rapamycin inhibition.[28] In other words, a better substrate will be more resistant to rapamycin inhibition while a worse substrate will be more sensitive to inhibition by rapamycin. It is possible that protein substrate binding and rapamycin-FKBP12 binding to mTORC1 are in conflict and a better protein substrate can outcompete the binding of rapamycin-FKBP12. In contrast, the active site directed inhibitors like Torin2 competes with ATP, and thus has no substrate protein selectivity.

Interaction of inhibitor and substrates at the enzyme active site.

Another mechanism for the substrate-selective modulation of enzymes is that the modulators may have cooperative binding to the enzyme active site with some substrates, but have competitive or induce unproductive binding with other substrates (Figure 5). This is likely the case for the substrate-dependent effects for the modulators of drug metabolizing P450 enzymes. Early on it was proposed that the P450 enzyme active site can be bound by multiple substrates/inhibitors at the same time.[46] This would then allow interesting substrate-substrate or substrate-inhibitor interactions to occur, producing substrate dependent inhibition or activation effects. This is highly possible as most of the substrates/inhibitors are hydrophobic molecules and they may interact with each other via hydrophobic interactions in the enzyme active site. Later, X-ray crystal structures of CYP3A4 indeed suggest that this is possible. For example, a structure of CYP3A4 in complex with a testosterone covalent dimer shows that the active site is large enough to accommodate multiple substrates.[47]

Figure 5.

Substrate-selective inhibition due to interaction of inhibitor and substrates at the enzyme active site. (A) Cartoon showing how active-site targeting modulators could have different interactions with substrates, leading to substrate-selective inhibition of an enzyme. The modulator competes with substrate 1 and promote substrate 2. (B) A structure of CYP3A4 in complex with a testosterone covalent dimer (PDB 7LXL) showing that the active site is large, which could allow substrate and inhibitor binding at the same time and thus enable substrate-inhibitor interaction that could affect catalysis in a substrate-dependent manner.

Experimental techniques for mechanistic investigations.

Determine the mechanism for any given substrate-selective modulator of an enzyme will benefit greatly from X-ray crystal or cryo-EM structure of the enzyme with the modulator bound. In comparison with the structures of the enzyme with substrates or other inhibitors bound, we can gain important information about how the modulator achieves substrate-selective effects. Enzyme kinetics and binding studies will also be important. For example, binding affinity or catalytic efficiency may help to determine whether if substrate-selective inhibition is due to differences in substrate binding affinities. Noncompetitive or uncompetitive inhibition mode, which can be obtained from enzyme kinetic studies, may help to point to exosite or allosteric mechanisms of substrate-dependent inhibition.

Implications of substrate-selective inhibitors

The above summary and analysis indicate that it is feasible to develop substrate-selective or substrate-dependent small molecule inhibitors for enzymes. The existence of substrate-selective inhibitors is likely more prevalent than we previously realized. These thinking have important implications to the chemical biology, medicinal chemistry, and biology community.

Because of the substrate-selective inhibition, it is logical that pharmacological inhibition and genetic knockdown/knockout may produce different phenotypes. With this thinking, we should be open-minded when we review manuscripts and refrain from requiring that small molecule inhibitor data to completely match that of genetic perturbation.

An important question, however, is how to rule out off-target effects of small molecules if we do not require equivalent phenotype from genetic perturbation. If we do not rule out off-target effects of small molecule inhibitors, we may make wrong conclusions about the mechanism of action, which will in turn negatively impact our understanding of biology and the development of new therapeutics. For substrate-selective small molecule inhibitors, the lack of equivalent knockdown/knockout phenotype makes it more challenging to rule out off-target effects. However, there are still ways to gain confidence in the mechanism of action:

1. Use structurally and mechanistically different small molecule inhibitors. If there are many small molecule inhibitors within the same class, we can compare their ability to inhibit the transformation of specific substrates by the enzyme and see if their ability on the specific substrates correlate with the cellular phenotypes. Using SIRT2 inhibitors as an example, TM, AGK2, SiReal2, and NPD-11033 all have similar inhibition profiles (inhibits deacetyaltion better than demyristorylation),[18] thus we would expect that they would have similar phenotypes. Given this consideration, it is very beneficial to have a diverse set of well-characterized inhibitors for any given enzyme.

2. We can use negative control compounds that are structurally very similar to the active compounds but do not inhibit the target proteins and make sure that the negative control compounds do not produce the same phenotype as the active compounds. Again using SIRT2 inhibitors as an example, the TM compound is a mechanism-based inhibitor with a thiomyristoyl group. Changing the thiomyristoyl group to a myristoyl group (S to O) leads to a compound that is very similar to TM, but has no inhibition on SIRT2.[17] If TM’s phenotypic effects is through SIRT2 inhibition, then we expect that the highly negative control compounds to have no such phenotypic effects. If it is through off-target effect, then the negative control compounds will likely still have the effects because it is structurally highly similar to TM.

3. We can convert the small molecule inhibitor to the corresponding Proteolysis Targeting Chimera (PROTAC) degrader and examine the phenotype produced by the PROTAC degrader. PROTAC uses a bifunctional molecule, with one end of the molecule binding to the target protein and the other end recruiting a E3 ubiquitin ligase, leading to the ubiquitination and degradation of the target protein.[48] Given that the PROTAC degrader leads to the destruction of the enzyme, it should produce effects similar to that of knockdown/knockout. This again has been demonstrated with SIRT2 inhibitors. By changing the substrate-selective SIRT2 inhibitor TM to a PROTAC version, SIRT2’s activity in removing long-chain fatty acyl groups is also inhibited.[49]

While these practices unavoidably complicate the research investigation and will require additional efforts, the bright side to chemical biologists and medicinal chemists is that the prospect of small molecule inhibitor development is not limited by the knockout phenotype. In other words, an undesirable phenotype produced by a knockout does not preclude that small molecule inhibitors of the enzyme would produce a desirable phenotype. This consideration also provides important justifications for development of diverse small molecule inhibitors for the proteins of interest. The substrate-selective inhibition by small molecules is particularly important in the context of the increasing competition from biologics. Aside from to the ease of production, storage, and administration, the ability to achieve substrate-selective inhibition by small molecules is a distinct advantage that may be difficult to achieve by RNA and protein therapeutics. Thus, chemical biologists and medicinal chemists should embrace this complexity and opportunity to actively develop more substrate-selective modulators of enzymes.