Keywords

1 Introduction

1.1 Discovery of Bioactive Organosilanes

Among the elements most closely associated with bioactive molecules – carbon, hydrogen, nitrogen, and oxygen – only oxygen and hydrogen are present in the Earth’s crust at substantial levels. Oxygen is the most abundant element by weight at 47 % and hydrogen tenth most abundant at 0.14 %. Carbon is 16th (0.03 %) and nitrogen 21st (0.005 %). Silicon is the second most abundant element (28 % of the Earth’s crust), and while silicon is the element most similar to carbon, virtually all naturally occurring silicon is in the form of silicates, bonded only to oxygen atoms (silicon carbide, which has no Si–O bonds, is naturally occurring but rare substance, discovered in the 1950s [1]). Following the emergence of organosilicon chemistry more than a century ago [2], a search for biologically active organosilanes emerged, a search that continues today ([3]; for reviews, see [4–13]). The absence of naturally occurring substances with bonds between silicon and carbon makes bioactive organosilanes all the more interesting.

There are two fundamental approaches for discovery of novel bioactive molecules: design and random screening. Random screening is an extraordinarily useful approach for discovering breakthrough bioactive molecules and uses for otherwise arbitrary compounds, such as the laundry-additive-turned-billion-dollar-herbicide Roundup [14]. Random screening will discover bioactive structures that could not be derived through a design process, but for the most part, intellectual involvement by chemists comes only after identification of a lead molecule.

Random screening presumably was responsible for discovery of bioactive organosilanes for which there are no carbon analogs, such as cisobitan (1) [15], SAN 58–112 (2) [16], and phenyl silatrane (3) (Fig. 1) [17].

Fig. 1
figure 1

Bioactive organosilanes without carbon analogy, identified by random screening

Full size image

1.2 Silicon as a Substitute for Carbon

In contrast, the design process requires, by its very nature, involvement of chemists from the outset. One of the most reliable ways to prepare a biologically active organosilane is replacement of a carbon atom by a silicon atom (see reference [4–13], especially [7] and [12]). This is arguably the most subtle change one can make in terms of size and electronegativity (outside of hydrogen–deuterium exchange), but not all carbons are replaceable by silicon. Silicon does not form stable double bonds [18, 19]. The Si–O, Si–N, and Si–halogen bonds are strong, but they hydrolyze readily [20]. Silicon–hydrogen bonds are readily oxidized in vivo [21]. Replacement of a quaternary carbon by silicon is therefore the most readily realized approach to bioactive organosilane discovery, yet even swapping quaternary carbon with silicon has limitations. These parameters are nicely illustrated by the naturally occurring pyrethrin insecticides and the evolution of the synthetic, commercial analogs of these natural products (Fig. 2).

Fig. 2
figure 2

Naturally occurring insecticide pyrethrin I and three carbon-based commercial analogs. Years of discovery are given. Only etofenprox is compatible with silicon substitution. Two organosilanes are produced commercially for their biological activity: fungicide flusilazole and insecticide silafluofen

Full size image

1.3 The Pyrethrin Pathway

Pyrethrin I is an insecticidal compound isolated from the common chrysanthemum flower (Fig. 2) [22]. The pyrethins, however, are not light stable and therefore unsuitable for field use. Cypermethrin is a potent, light-stable commercial analog that has been used in the field with great success [23]. In both of these molecules, the sole quaternary carbon (black dot) cannot be replaced by silicon because silicon in a three-membered ring is extremely unstable [24, 25]. In addition, this quaternary carbon is alpha to a carbonyl, and silicon alpha to a carbonyl is also unstable [26]. In the third pyrethroid, esfenvalerate [27], the dimethylcyclopropane has been replaced by an isopropyl group, but this structure remains unsuitable for a C to Si exchange (at the black dot) for two reasons: the resulting Si–H bond would be rapidly oxidized [21] and the silane remains adjacent to a carbonyl. In the fourth pyrethroid etofenprox, only the gem-dimethyl group and the central oxygen atom remain from the natural pyrethrin I [28]. Etofenprox does retain the quaternary carbon, not in a cyclopropane, and therefore was the first analog of the natural product in which silicon substitution was possible. Replacement of this quaternary carbon with silicon retained insecticidal properties, and the resulting silafluofen is a commercial insecticide [29, 30]. Silafluofen and the agricultural fungicide flusilazole [31] are currently the only organosilanes manufactured because of their biological activity.

1.4 Silanols

Silicon ethers, the Si–O–C linkage, have stability that is related to the nature of the component alcohol and the silane components [32]. Some silane-derivatized alcohols and amines have been found to enhance the bioactivity of the parent compounds, presumably by hydrolysis in vivo to the alcohol/amine and the corresponding silanol (see reference [4–13], especially [7] and [12]). Some silanols have been found to have very intriguing biological properties, most notably the hypnotic properties of diphenylsilanediol 4 [33] and sila-haloperidol 5 [34, 35] (Fig. 3). The hydroxyls of silanols exchange rapidly in the presence of water, but they remain silanols (see Fig. 12). The primary instability of silanols is dehydration and formation of siloxanes, structures containing the Si–O–Si bonding arrangement (see Fig. 4). Siloxanes are generally more stable than their Si–O–C counterparts (see Fig. 1). The many useful properties of siloxanes form the basis of the large organosilane industry [36].

Fig. 3
figure 3

Bioactive organosilanols

Full size image
Fig. 4
figure 4

Silanediols can readily dehydrate to form siloxanes (silicones), but steric hindrance can slow or prevent oligomerization

Full size image

1.5 Silanols, Silanones, and Siloxanes

Silanediols (6) are best known as precursors of siloxanes 7 (aka silicones), the product of a condensation reaction that is catalyzed by both acid and base (Fig. 4). The polymerization of dimethylsilanediol 8 is reported to be spontaneous, with even very pure 8 being unstable [37]. Steric hindrance plays a significant role here; dimethylsilanediol 9 polymerizes at one tenth the rate of 8 [38, 39]. Diisobutylsilanediol 10 is a liquid crystal, illustrating both stability and the outstanding ability of silanols to be hydrogen bond donors and acceptors [40, 41]. With sufficient steric bulk, as in di-tert-butylsilanediol 11, even forcing conditions do not lead to polymerization. Solvation of silanols can also play a significant role as well. Tacke found that dimethylsilanol-substituted alanine 12 formed siloxane 13 when it was isolated, but dissolving 13 in water gave a spontaneous, quantitative hydrolysis to yield 12 [42, 43]. This finding suggests that the insolubility of most siloxanes in water adds to their hydrolytic stability.

The stability of simple silanediols like 10 provided encouragement for efforts to design biologically active analogs that would remain monomeric [44, 45]. An early example of a bioactive silanediol is 4 (Fig. 3) which has barbiturate-like properties [33].

Silicones (e.g., 7) were first prepared more than 100 years ago in an attempt to make silicon analogs of ketones, inadvertently illustrating the instability of double bonds to silicon (14) [46]. This contrasts with the proclivity of 1,1-diols 15 to dehydrate and form carbon–oxygen double bonds 16. The stability of carbonyls relative to their hydrates is particularly pronounced for carboxylic acid derivatives, especially amides such as 17. Peptides 17 are polyamides that can be hydrolyzed by protease enzymes to their respective acid 19 and amine 20 under mild physiological conditions. Proteases catalyze the addition of water (or alcohol or thiol) to the amide carbonyl, in part by stabilizing the tetrahedral intermediate 18 [47]. Tight binding, nonhydrolyzable mimics of this tetrahedral group can be effective inhibitors of these enzymes and therefore pharmaceuticals [48–50].

1.6 Protease Enzymes and Their Inhibition

Protease enzymes mediate many important biological processes and are pharmaceutical targets for a broad range of diseases [48–50]. Protease inhibitors as pharmaceuticals were first introduced to treat hypertension nearly 40 years ago [51]. Figure 6 shows four examples of commercial protease inhibitors, with gray boxes highlighting the functional group that replaces and mimics the tetrahedral intermediate of amide hydrolysis 18. Fosinoprilate utilizes a phosphinic acid to replace the hydrated carbonyl and interact with the active site zinc ion of angiotensin-converting enzyme [52, 53]. Inhibiting this metalloprotease lowers blood pressure [54]. Atazanavir is an inhibitor of the HIV protease, an aspartic protease, and prevents reproduction of the HIV virus [55, 56]. In this case, a hydroxyethylene group replaces the hydrated amide and hydrogen bonds to the active site aspartic acids. The boronic acid bortezomib is a treatment for cancer; the electrophilic boron binds to the nucleophilic threonine hydroxyl group at the active site [57]. Sitagliptin binds to the active site of a serine protease (DPP-4) and is a treatment for diabetes [58].

2 Silanediols in Drug Design

Silanediols 21 seemed nearly ideal isosteres of the hydrated amide carbonyl 18, but at the beginning of our investigations, this idea had only been investigated sparingly (Fig. 7) [59]. The testing of silanediols as hydrated carbonyl mimics was largely suppressed by the general knowledge of the instability of silanediols toward polymerization (Fig. 5) and a dearth of methods for preparing silanediols with functionality and stereochemistry.

Fig. 5
figure 5

Silanones 14 are very unstable toward hydration, whereas ketones 16 are more stable than their hydrates. Amides such as 17, much less prone toward hydration than ketones, are substrates for protease enzymes

Full size image
Fig. 6
figure 6

Examples of commercial protease inhibitor drugs used for the treatment of hypertension (fosinoprilate), HIV infection (atazanavir), cancer (bortezomib), and diabetes (sitagliptin)

Full size image
Fig. 7
figure 7

Silanediol 21 as an isostere of hydrated amide 18. Compounds 8 and 22 had been found to be ineffective as hydrolase inhibitors. Redesigning protease inhibitor 23 (shown as the ketone hydrate) gave the silanediol inhibitor 24

Full size image

Prior to our studies, two silanols had been evaluated as inhibitors of hydrolase enzymes, 8 and 22. An aqueous solution of dimethylsilanediol 8 was found to be ineffective as an inhibitor of the metalloprotease angiotensin-converting enzyme at 300 mM [60]. In addition, silanetriol 22 was prepared as a potential inhibitor of a beta-lactamase and was inactive at the highest concentration tested, 0.1 mM [61].

Our first vehicle for testing structure 21 as a nonhydrolyzable analog of 18 and the centerpiece of a protease inhibitor was silanediol 24, by analogy with 23. Structure 23 is shown as the hydrate of a ketone that was tested as a replacement for an Ala-Phe dipeptide and was found to be a 1 nM inhibitor of angiotensin-converting enzyme [62]. The hydrate 23 is presumably the form that binds at the ACE active site [63]. The silanediol in 23 was anticipated to have a steric environment that would inhibit oligomerization, shielded on each side by branched alkyl substituents, not unlike diisobutylsilanediol 10 (Fig. 4).

2.1 Synthesis of Silanediol Peptidomimetics

2.1.1 First-Generation Silanediol Syntheses

A synthesis of 24 was needed, including a strategy for carrying the silanediol through the synthetic manipulations that would be required. Diethers of silanes (silyl acetals) are rather unstable toward hydrolysis and were rejected as a viable form of protected silanediol [64]. Acid-catalyzed hydrolysis, “protodesilylation,” appeared to be an interesting alternative [65]. Acidic cleavage of Si–C bonds of aryl silanes is a classic electrophilic aromatic substitution reaction and therefore electronically tunable if simple phenyl groups proved to be less than perfect (see Fig. 9) [66]. The precursor of silanediol 24 therefore was diphenylsilane 25 (Fig. 8).

Fig. 8
figure 8

ACE inhibitor 24 and its synthesis beginning with difluorodiphenylsilane 30

Full size image
Fig. 9
figure 9

Acid-catalyzed hydrolysis of the diphenylsilane involves participation of the adjacent amides

Full size image

In this initial exploration of silanediol synthesis, difluorodiphenylsilane 30 was coupled with metalated dithiane 29 followed by the reaction with enantiomerically pure lithium reagent 28 [67]. The resulting silane 32 contains all the carbons of dipeptide mimic 24 but with functional group elaboration required. Lithium reagent (R)-28 was prepared from commercially available (S)-ester 26.

The dithiane in 32 was hydrolyzed to the corresponding ketone and then reduced to the corresponding alcohol. The nitrogen of 33 was introduced as a phthalimide using Mitsunobu chemistry, followed by hydrolysis and conversion of the amine to benzamide 33.

Following cleavage of the benzyl ether of 33, the diastereomers were separated and the alcohol of the (S,R) stereoisomer was oxidized to acid 34. Condensation with the tert-butyl ester of (S)-proline gave silane 35 [67].

2.1.2 Silanediols from Diphenylsilanes

In the ultimate synthetic step, deprotection of 35 (Fig. 8) was accomplished by treatment with triflic acid, serving to remove the tert-butyl ester and both phenyl groups on silicon. The optimized sequence for isolation of silanediol 24 involved neutralization of the mixture with ammonium hydroxide and then conversion of the silane to a crystalline difluorosilane with aqueous hydrofluoric acid [68]. Hydrolysis of the difluorosilane to the silanediol was readily accomplished by treatment with aqueous base.

Direct conversion of diarylsilanes to silane ditriflates using triflic acid is well known [69]; however, acidic hydrolysis of the diphenylsilane 35 and related compounds is believed to involve participation of the flanking amide carbonyls (Fig. 9). Beginning with ipso-protonation of the aromatic ring (36), loss of benzene requires a nucleophile to attack silicon, allowing departure of the benzene, a role readily fulfilled by either of the flanking amides. A second protonation–nucleophilic attack leads to 38. Five-membered silyl ethers are strained, and therefore, these undergo a facile hydrolysis to the silanediol [70]. Participation of the amides is consistent with our observations of intermediates by 1H NMR [71] and the work of Nielsen and Skrydstrup with monoamide 39 [72]. Whereas triflic acid hydrolysis of diamide 35 leads to 38 within minutes, and thus the silanediol 24, treatment of monoamide 39 under similar conditions results in rapid loss of the first phenyl group and a very slow cleavage of the second [72]. Recent work by the Skrydstrup group has shown that activation of the aromatic rings by alkyl substitution, as in 40, allows for hydrolysis of both Si–aryl bonds with trifluoroacetic acid instead of triflic acid, presumably by lowering the activation energy for the ipso-protonation step (see also Fig. 20) [73].

2.1.3 Inhibition of Metallo- and Aspartic Proteases

Using chemistry related to the initial investigation (Fig. 8), a set of silanediols were prepared as analogs of other known protease inhibitors (Fig. 10). Silanediol 24, an isostere of hydrated 23a (see Fig. 7), was found to be an effective inhibitor of angiotensin-converting enzyme (ACE), a metalloprotease with a zinc ion at the active site [74]. Compound 23a inhibits ACE with an IC50 of 1 nM, and silanediol 24 was found to inhibit this enzyme with a K i of 3.8 nM [62]. Similarly, silanediol 42 was prepared as an analog of phosphinic acid 41, an inhibitor of thermolysin [75, 76]. Silane 42 and phosphinic acid 41 are constructed around second row elements. As such, the silane and phosphorus units are closer in size than silicon is to carbon, but are also dramatically different in their acidities. Silanediols have pK a values of approximately 12 [44, 77] and would be unionized at physiological pH, whereas phosphinic acids have a pK a in the range of 3–4 and are anionic at physiological pH. Ionized phosphinic acid 41 would be expected to benefit from a Coulombic attraction to the active site zinc cation and silanediol would not. Nevertheless, phosphinic acid 41 inhibits thermolysin with an IC50 of 10 nM, and silane 42 inhibits thermolysin with a very similar IC50 of 40 nM. A crystal structure of 42 bound to thermolysin has been obtained; see Fig. 11.

Fig. 10
figure 10

Silanediol inhibitors by analogy: known inhibitors (left) and their silane analogs (right)

Full size image
Fig. 11
figure 11

Crystal structure of silanediol 42 bound to thermolysin (PDB #1Y3G) and its overlay with 41b also bound to thermolysin (PDB #5TMN). Silanediol 42 in yellow and 41b in light gray. The overlay also shows the active site zinc in dark gray illustrating that they each touch the zinc in the same way

Full size image

A single example of a silanediol aspartic protease inhibitor has been studied, compound 44, prepared by analogy to HIV protease inhibitor 43 [78]. Carbinol 43 inhibits the HIV protease with an IC50 of 0.38 nM, and silanediol 44 is sevenfold less potent, 2.7 nM. Side-by-side testing of 43 and 44 and the similarly potent and commercial HIV inhibitor indinavir [79] found that they all cross cell membranes with equal efficiency, as determined by their ability to prevent infection of cells by the HIV virus. These studies demonstrated that silanediol 44 has drug-like properties [78].

Syntheses of silanediols 42 and 44 followed paths similar to the preparation of 24 (Fig. 8), using enantiomerically pure lithium reagents 45 and 46. While effective, this strategy was not efficient nor was it readily scalable. More efficient and catalytic methods have been developed recently and have been utilized for the silanediols that have followed the examples described below.

A crystal structure of silanediol 42 (PDB #1Y3G) bound to the active site of thermolysin found that the silanol oxygens are arrayed about the active site zinc in precisely the same way that the phosphorous oxygens are for the related phosphinamide 41a (PDB #5TMN) (Fig. 11) [80]. Phosphinic acid 41 and phosphinamide 41a are both 10 nM inhibitors of thermolysin, but only the latter has been crystallized at the active site of thermolysin [75, 76, 81].

2.1.4 Inhibition of Serine Proteases

Two of the more recent subjects of investigation are 48 and 50 (Fig. 12). Both of these structures target serine proteases, which are fundamentally different from the metallo- and aspartic proteases targeted initially. Metalloproteases have an active site zinc ion to coordinate with and activate an amide carbonyl. This zinc ion also delivers the water nucleophile (see inhibitors 23, 24, 41, and 42). Aspartic proteases have two aspartic acid residues at the active site, delivering the nucleophilic water molecule and stabilizing the hydrated carbonyl by hydrogen bonding (inhibitors 43 and 44).

Fig. 12
figure 12

Protease substrate residue definitions and silanediol inhibitors by substrate analogy

Full size image

In contrast to the metallo- and aspartic proteases, serine proteases use a serine residue of the enzyme as the nucleophile. This reaction leads to departure of the amine portion of the peptide substrate, leaving the carbonyl esterified to the serine residue. In a second step, water hydrolyzes the ester, freeing the acid and regenerating the active site. Many serine protease inhibitors have electrophilic functionality that interact with the serine alcohol, but only carry recognition groups on one side of the active site nucleophile [82]. Using the Schechter and Berger descriptors [83], only the non-prime residues are represented in many serine protease inhibitors (Fig. 12). To adequately protect the silanediol moiety from oligomerization, we desired substitution at both P1 and P1′ sites to provide steric shielding. From the perspective of protease inhibitor selectivity – a major consideration in protease inhibitor design [84] – accessing recognition centers on both sides of the enzyme active site can be advantageous.

For the classic serine protease chymotrypsin, an enzyme that requires an aromatic side chain at P1 for substrate recognition, we modeled the silanediol inhibitor after the chromogenic substrate 47 and utilized a methyl group at P1′ (Fig. 12) [85]. Silanediol 48 was found to inhibit chymotrypsin with an IC50 of 107 nM [85].

A medically important serine protease, Factor XIa, was chosen as a second test case for study. Silanediol 50 was prepared as a potential inhibitor of Factor XIa, which is a key component of the coagulation cascade [86]. Silanediol 50 is a mimic of Factor IX, the substrate for Factor XIa, which is cleaved between Ala146 and Arg145 (49). The P1 guanidinium ion of 49 is required for recognition by Factor XIa [87].

Electrophilic coordination to an active site serine nucleophile by a silanediol is inherently different from hydrogen bonding of the silanediol group in an aspartic protease active site or zinc ion coordination in metalloproteases. There was, however, ample experimental evidence for this kind of electrophilic role for silanols. We had observed a rapid exchange of silanediol hydroxyls by two equivalents of methanol-d 4 when 44 was dissolved in this solvent (Fig. 13) [71]. Additional evidence for rapid exchange of hydroxyls in silanols came from the work of Tacke. For example, enantiomerically pure silanol 52 was found to be stable in aprotic solvents, but when exposed to aqueous solutions 52 rapidly racemized [88].

Fig. 13
figure 13

Hydroxyls on silicon rapidly exchange with water and alcohols as evidenced by substitution by deutero-methanol in 44 and the racemization of 52 in the presence of water

Full size image

2.1.5 Second- and Third-Generation Silanediol Syntheses

Initially, syntheses of 48 and 50 used an approach similar to that used earlier (Fig. 8), but these syntheses were superseded by better techniques for assembly and control of the two stereogenic centers flanking the silanediol [89, 90]. In all cases, a diphenylsilane was carried through the synthesis sequence and hydrolyzed to the silanediol by strong acid as the last step.

The synthesis of 59, precursor of inhibitor 48, began with cycloaddition of dichlorodiphenylsilane and isoprene to yield dihydrosilole 53 (Fig. 14). This reaction can easily produce hundreds of grams in a single run [91]. Asymmetric hydroboration of 53 with pinene-derived isopinocamphylborane gave 54 in >90% ee [85, 92]. Warming 54 with aqueous HF leads to Peterson-like fragmentation and formation of fluorosilane 55 with the methyl stereogenic center in place. Fluorosilanes are relatively moisture insensitive and also react efficiently with nucleophiles [93]. Conversion of fluorosilane 55 to ketone 56 was followed by asymmetric reduction of the ketone using stoichiometric CBS-borane reagent to give 57 with >90% diastereoselectivity [94]. Mitsunobu inversion of the alcohol by phthalimide installed the amine in 58 with the correct stereochemistry. Oxidative cleavage of the alkene then gave the amino acid 59 full control of both stereocenters. Conversion of 59 to 48 then followed standard protocols [85].

Fig. 14
figure 14

The stereogenic centers of 59 were installed using asymmetric hydroboration reagents

Full size image

A direct and efficient method for preparation of the α-aminosilane group with control of enantioselectivity was introduced by Nielsen and Skrydstrup using silyllithium reagents and sulfinimines [72]. This method was applied to the synthesis of silanediol precursor 62 (Fig. 15). Silyllithium reagent 61 added to the enantiomerically pure 60 to give 62 with excellent diastereocontrol [90]. Use of racemic 61 demonstrated that the methyl stereogenic center does not influence the diastereoselectivity of the reaction with 60.

Fig. 15
figure 15

Using the Skrydstrup protocol, silyllithium reagent 61 added to sulfinimine 60 to give 62 with complete stereocontrol

Full size image

The coupling of Davis–Ellman sulfinimines [95] with silicon nucleophiles by Skrydstrup was an excellent solution for stereocontrol to the α-aminosilane stereogenic center, but the stereochemistry of the β-silyl acid portion of these inhibitors (see 61 in Fig. 15) was in need of a better solution as well. The asymmetric hydroboration approach for control of this stereocenter, illustrated in Fig. 14 via 53, was limited by the lack of readily available 2-substituted-1,3-dienes [96]. In the original syntheses of the inhibitors shown in Fig. 8, an enantiomerically pure lithium reagent was used, either prepared from the Roche ester 26 or using Evans chiral auxiliary technology 63 (Fig. 16).

Fig. 16
figure 16

Asymmetric hydrosilylation, reductive opening of 66, and addition to sulfinimines yield advanced silanediol intermediates rapidly and with complete control of stereochemistry

Full size image

One of the most efficient methods for forming Si–C bonds is hydrosilylation [97–100]. Intramolecular hydrosilylation has been extensively studied, but the asymmetric transformation of 65 into 66, needed for our efforts, had only been achieved with low enantioselectivity [101, 102]. In a survey of commercially available rhodium ligands, it was found that (S,S)-diethylferrotane gave >90 % ee for this transformation and a similarly high enantioselectivity for homologs of 65 carrying groups other than methyl. A second breakthrough was the discovery that treatment of 66 with lithium directly and quantitatively generated the dianion 67. In a reaction similar to that shown in Fig. 15, addition of 67 to sulfinimine 60 gave 62 in good yield and with complete control of stereochemistry (Fig. 16). Coupling of 67 and 68 gave 69. Hydrolysis of the sulfonamide 69, protection of the amine with a Boc group, and oxidation of the alcohol to an acid completed the preparation of key intermediate 70 in short order [103].

The combination of intramolecular hydrosilylation, direct reductive opening of the silyl ether, and coupling of the resulting silyl anion with a sulfinimine (Fig. 16) has proven to be a very efficient way to assemble the core components of the silanediol precursors such as 62 and 70. Compound 70 can be directly coupled with amino acids and hydrolyzed to the silanediol, providing protease inhibitor candidates. Challenges remain, however. The key intermediate for assembly of Factor XIa inhibitor 50 (Fig. 12) is the readily prepared 62 (Fig. 16). The conversion of 62 to 50, however, was a hard slog, requiring extensive functional group manipulation and illustrating the opportunities for improvement that remain [90].

3 Summary and Future Prospects

Silanediols embedded in a peptide-like structure were conceived as mimics of the tetrahedral intermediate of amide hydrolysis, with the potential for inhibiting protease enzymes. Starting with a protease substrate, 71 (Fig. 17), a first-generation inhibitor can be designed by replacing the scissile amide bond with a methylene-silanediol unit, e.g., 72. This concept has been tested for three of the four classes of protease enzymes and found to be very effective.

Fig. 17
figure 17

First-generation silanediol protease inhibitors by substitution of the hydrated amide with a methylene silane group and the active sites of three protease classes for which silanediols are effective inhibitors

Full size image

3.1 Mechanistic Considerations and Future Challenges

Aspartic proteases, enzymes that contain two aspartic acids at the active site, mediate proteolysis through hydrogen bonding, 73 (Fig. 17). The aspartic acid units are generally considered to have one ionized and one unionized at the active site, and these stabilize the amide hydrate [104]. A silanediol replacement of this hydrated peptide is expected to readily interact with the active site carboxylic acids because silanols are outstanding hydrogen bond donors and acceptors [44, 45, 105, 106]. This potential has been evaluated using the HIV protease, taking advantage of the C 2 symmetry of the enzyme with a C 2 symmetric inhibitor 44 (Fig. 10). Inhibitor 44 was prepared by analogy to a published inhibitor (43) that contained a hydroxyethylene replacement for the amide bond. This silanediol-for-carbinol switch, without reoptimization, gave an effective inhibitor with an IC50 under 3 nM. More importantly, this study found that the silanediol was as effective as the commercial drug indinavir at crossing cell membranes and protecting cells from infection [78]. Additional examples of silanediol-based inhibitors of aspartic proteases would be desirable.

Two metalloprotease enzymes have been potently inhibited by silanediols. Metalloproteases have a zinc ion, held at the active site by a glutamate and two histidine ligands, 74. This zinc ion stabilizes the amide hydrate by coordination to the geminal hydroxyl groups of the hydrated carbonyl. Silanols can ligate and chelate metal ions [107–109]. Silanediol 42 (Fig. 10) was prepared as an analog of a phosphinic acid 41 and phosphinamide, 41a, and was found to inhibit the metalloprotease thermolysin with an IC50 of 40 nM. Moreover, both 41a and 42 were examined by X-ray spectroscopy and found to coordinate to the zinc active site with precisely the same arrangement of their two oxygen atoms. An additional example of metalloprotease inhibition by a silanediol is the 4 nM inhibition of ACE by silanediol 24 (Fig. 10). This silanediol was patterned after the ketone inhibitor 23a and its presumed inhibitor form, the ketone hydrate 23 (Fig. 7).

These metallo- and aspartic protease inhibitors bind to the active site of the enzyme through noncovalent interactions. In contrast, most inhibitors of serine proteases incorporate an electrophilic group that covalently engages the active site serine. In the case of silanediol inhibitors 48 and 50, the electrophilic group is a silanediol. A crystal structure of these inhibitors bound to the active sites and demonstrating the covalent attachment of the silane to the serine nucleophile has not yet been achieved.

The fourth class of proteases is the cysteine proteases [110], for which inhibition by a silanediol has not been examined (Fig. 18). The use of silanediols for these important targets implies the formation of a silicon–sulfur bond, e.g., 77 [111]. Should the strength of the bond of the enzyme thiol to silicon be important, the lower bond strength for Si–S relative to Si–O (see 79) may limit the use of silanediols as inhibitors of these enzymes.

Fig. 18
figure 18

Cysteine proteases utilize a nucleophilic sulfur to hydrolyze amides. Inhibition of this important class of proteases by silanediols has not been demonstrated

Full size image

For all of the potential inhibitors discussed, the specificity and potency of the inhibition may, in large part, be determined by the side chain substituents that flank the central silicon atom (see Fig. 12). On the other hand, the silanediol imparts a polarity that lies somewhere between a phosphinic acid and a carbinol. This intermediate polarity has the potential to produce bioactive substances with useful log P values [112].

Preparation of silanediols as protease inhibitors, assembling the functionality and stereogenic centers in a general and efficient way, can now be routinely accomplished, but some targets remain a challenge. The original path to a silanediol required many functional group transformations, from the Roche ester 26 to the penultimate product 34 (Fig. 19). In the currently most efficient path, a 2-substituted allyl alcohol is converted to an intermediate like 69 in three steps, requiring only exchange of nitrogen substituents and oxidation of the primary alcohol. This route enjoys full control of stereochemistry using asymmetric hydrosilylation and a Davis–Ellman sulfinimine. Despite these advances, incorporation of the guanidine functional group for Factor XIa inhibitor 50 (Fig. 12) required many functional group manipulations [90].

Fig. 19
figure 19

Comparison of two routes to advanced silanediol precursors 34 and 70

Full size image

An additional synthesis challenge is optimization of the ultimate deprotection step. There have been contributions from several labs [113], modifying the phenyl groups on silicon with a goal of avoiding the use of triflic acid in the deprotection. A variety of electrophiles can be utilized for this transformation, including bromine and mercury (Fig. 20) [114]. Protons are the most desirable of these; bromine can add to alkenes and activated aromatics, and mercury is a toxic heavy metal that has the potential to complicate enzyme assays if it is not fully removed from the inhibitor product.

Fig. 20
figure 20

The challenges of silanediol deprotection and the use of electrophiles other than protons

Full size image

The most successful of the efforts to optimize the acid deprotection procedure has been from the Skrydstrup lab and the addition of methyl substituents to the phenyl groups (Fig. 21) ([73]; for the use of methoxy substitution to promote ease of protodesilylation reactions, see [115] and references therein). This work beautifully illustrates the utility of having electron-donating substitution appropriately placed. Methyl groups at the meta position (83) do not adequately facilitate aryl hydrolysis (see Fig. 9), but para substitution allows for removal of both toluene groups using only trifluoroacetic acid, which is far more mild than triflic acid.

Fig. 21
figure 21

Skyrdstrup’s substituted phenyl silanes allow the use of milder acid for hydrolysis, particularly for para substitution

Full size image

The future of silanediols as pharmaceutical agents will also require a detailed understanding of the pharmacokinetics of these silanediols in peptide-like environments. Much is left to explore!