3.2. Structure–Activity Relationship (SAR)
To identify promising NSGMs, we measured the IC
50 of 22 NSGMs that inhibited CatG by at least 50%.
Figure 4 shows sigmoidal dose–response relationships for a select group of NSGMs, suggesting nearly a 1000-fold range of potencies. While sulfated benzofuran dimers
4–
15 displayed IC
50s in the range of 5 to >50 µM (
Table 1), sulfated flavonoid dimers presented IC
50 in the range of 0.05–10 µM (
Table 2). At first glance, this could represent a preference for the flavonoid scaffold, but such a generalization would be inaccurate because there is a vast difference in the level of sulfation between the two scaffolds. While the benzofuran dimers had only one or two sulfate moieties, the flavonoid dimers had a minimum of four sulfates. More interestingly, the level of sulfation, although necessary and important for binding to electropositive CatG (
Figure 1), appears to contribute additional factors, which are described below.
Benzofuran-based NSGMs—Although sulfated benzofuran NSGMs displayed modest IC
50s (5.5 to >50 µM,
Table 1), interesting insights can be derived in terms of CatG recognition. CatG is a strongly basic protease with a predicted pI of 12 [
49] and a predominantly electropositive surface area (
Figure 1). A priori, this implies a high probability of binding to polyanionic species such as long chain GAGs. Nevertheless, mono- and di-sulfated benzofurans, consisting of hydrophobic and aromatic groups, were found to be modest inhibitors of CatG. In fact, the hydrophobic substituents, rather than the number of sulfate moieties, appeared to be more important for CatG inhibition in this series. This implies that the monosulfated benzofuran scaffold may serve as a useful fragment for conjugation with a promising hit from another screen.
This is not the first time that hydrophobic scaffolds and/or substituents of highly-sulfated NSGMs have been found to induce inhibition of serine proteases. In fact, an optimal combination of hydrophobic and anionic forces was proposed to be the basis for both affinity and selectivity of sulfated NSGMs, as described in a recent perspective [
42,
50,
51]. More specifically, whereas sulfated benzofuran dimer
5, having a phenethyl group at the R
2 position, displayed an IC
50 of 5.5 µM, dimer
15, with a sulfate group in the same position, was more than 10-fold less potent (
Table 1). However, the presence of a substituted aromatic ring in the same position, e.g., dimers
6–
8, resulted in at least a five-fold loss in potency. Introducing an alicyclic ring, as in dimer
4, at the same position (R
2) led to a nine-fold loss in potency. These results suggest that the phenethyl group at the R
2 position occupies a well-defined pocket that contributes to the selectivity of binding.
The presence of a bulky hydrophobic group at the R
1 position of sulfated benzofuran dimers appears to marginally favor CatG inhibitory potency as evidenced by NSGM
10 (IC
50 29 µM), which was almost two-fold more potent than NSGM
4 (
Table 1). Likewise, the benzyl group at the R
4 position (dimer
12) is slightly favored over phenyl (dimer
11) or substituted phenyl group (dimer
13). Of note, both of these observations convey the importance of the nature of hydrophobic groups in improving potency. Interestingly, enhancing sulfation level did not offer better inhibition potential over monosulfated benzofuran dimer
5, as evidenced by disulfated NSGMs
14 and
15. In fact,
14 having an additional sulfate at the R
3 position was nearly three-fold less potent than dimer
5, while NSGM
15 was essentially inactive (IC
50 > 50 µM,
Table 1).
Overall, these results for sulfated benzofuran dimers underscore the importance of the aromatic/hydrophobic groups at the R2 position, while also emphasizing the importance of the positions, rather than the number, of sulfate groups for CatG inhibition. The SAR observed for this series of inhibitor hits suggests that appropriate modification or conjugation at the R1 and R3 positions may significantly increase inhibition potency. From a drug design/discovery perspective, such modifications are much easier to introduce than altering the number and position of sulfate groups.
Flavonoid-based NSGMs—This sub-library of dimers is made up of di-quercetin, quercetin–apigenin, and di-apigenin dimers with 8, 6, and 4 sulfate groups, respectively. Alternatively, this class of NSGMs was completely different from the sulfated benzofuran dimers in terms of both the scaffold as well as the level of sulfation.
Table 2 lists the inhibition potencies of NSGMs
18–
29, which present a fairly wide range, from 0.05 to 10.3 μM. Interestingly, as a group, tetra-sulfated NSGMs, e.g.,
18–
23, were ~50-fold less potent than octasulfated NSGMs, e.g.,
24–
28. Alternatively, this group of NSGMs presents the conclusion that higher sulfation is better for CatG inhibition, an observation directly opposed to the results presented above for NSGMs
4–
15 (
Table 1). More specifically, comparing NSGMs with identical linkers, e.g.,
25 (IC
50 0.05 µM) vs.
19 (IC
50 3.3 µM), shows a 66-fold difference between octasulfated di-quercetin NSGM and its tetra-sulfated di-apigenin counterpart.
For the di-apigenin-based NSGMs, the most potent molecule,
22 (IC
50 1.3 µM), had a 2,6-bis(methylene)pyridine linker (
Table 2). Substitution of the linker with a 1,3-bis(methylene)benzene in NSGM
21 resulted in an eight-fold decrease in potency, possibly indicating a role for the heteroatom in binding. A change from meta- to para-substitution in the linker, i.e.,
21 vs.
19, resulted in a two-fold loss. This implies that meta to para change probably impacts the relative spatial arrangement of each monomer as well as their sulfates.
For the di-quercetin-based NSGMs, the most potent NSGM, 25 (IC50 0.05 µM), had a 1,4-bis(methylene)benzene linker. Introducing methyl groups on a linker aryl ring (NSGM 26) did not affect CatG inhibitory potency; however, a three-fold decrease in potency was observed when the linker was changed to 1,3-bis(methylene)benzene (NSGM 27, IC50 0.15 µM). This is similar to what was observed for the di-apigenin-based NSGMs (above). Contrary to the observation with di-apigenin dimer 22, the presence of a heteroatom in the linker in di-quercetin dimer 28 had no impact on potency when compared with respective parent NSGMs (e.g., 21 and 27). Introducing a more flexible linker induced a 4-fold loss of potency (i.e., 25 vs. 24), and suggests a possibly-limited ability to maneuver around the linker structure.
Finally, the most convincing evidence of the importance of the linker is seen with NSGM 29, a hexa-sulfated quercetin–apigenin heterodimer carrying a 1,4-bis(methylene)benzene linker. NSGM 29 displayed an IC50 of 0.14 µM, which is several-fold lower than the di-apigenin-based NSGMs (1.3–10.3 µM), but only 2.8-fold higher than the most potent di-quercetin NSGM 25 (0.05 µM). In fact, heterodimer 29 is equipotent with most di-quercetin NSGMs, despite having two fewer sulfate groups. This implies that there is enough structural space available around the quercetin–apigenin heterodimeric scaffold for the discovery of more potent leads, if needed.
3.4. Salt-Dependence of CatG Inhibition in the Presence of NSGM 25
A fundamental point being advanced in NSGM-based mimicry is the presumed increase in hydrophobic forces contributing to binding affinity. As is well-recognized, heparin and other GAGs utilize primarily electrostatic forces in binding to proteins [
54]. Only when the contribution of nonionic forces, e.g., van der Waals, and/or directional ionic forces, e.g., hydrogen bonding (H-bonding), is high enough do GAGs exhibit a high level of selectivity. This is exemplified by the classic case of antithrombin binding to heparin, which exhibits nearly 60% nonionic binding energy [
55,
56]. The resolution of overall binding energy into ionic and nonionic contributions is typically achieved by performing affinity measurements as a function of the ionic strength of the buffer. According to the protein–polyelectrolyte theory [
57], the two contributions can be resolved from a double-log plot of binding affinity against the Na
+ concentration, as defined by the equation log K
I = log K
I,NONIONIC + Zψ × log [Na
+], where Z represents the number of salt interactions, and ψ is the proportion of Na
+ released per anion upon ligand binding and is equal to 0.8 for heparin [
55].
To resolve these two types of contributing forces, we measured the IC
50 of NSGM
25 at pH 7.4 and 37 °C as a function of the ionic strength of the buffer (
Figure 6A). The dose dependence profiles clearly show a loss in potency, as expected. In fact, the IC
50 values increased from 0.043 ± 0.01 µM to 3.42 ± 0.52 µM as NaCl concentration was increased from 50 mM to 200 mM (
Table 4). This represents a substantial loss of ~80-fold in inhibition potency, and demonstrates that ionic forces are important to the CatG–NSGM
25 system.
The measured IC
50s of our allosteric inhibitors are distinct from inhibition constants K
I, which are thermodynamic constants. Cheng and Prusoff provided a mathematical foundation for the transformation of IC
50s into K
Is for uncompetitive inhibitors. In their formulation, K
I is equal to IC
50 × [S]/(K
M + [S]) [
43], which is directly applicable to our study. Thus, it becomes possible to utilize the linear double-log analysis described above for our NSGM
25, which is an uncompetitive inhibitor.
Figure 6B shows the double-log plot of K
I versus Na
+ concentration, where the inhibition constants were calculated from the observed IC
50 values. Linear regression yielded a slope of 2.94 and an intercept of −4.32 (
Table 5). Whereas the former corresponds to an ionic binding energy of 4.88 kcal/mol at 37 °C and 100 mM salt, the latter yields a nonionic binding energy of 6.12 kcal/mol. These results show that the NSGM
25 interaction with CatG is driven by both electrostatic (~44%) and nonionic forces (~56%). Such important roles of two forces have not been observed earlier. More importantly, the higher nonionic component bodes well for the discovery of second-generation inhibitors with higher selectivity.