Multi-inhibitor prodrug constructs for simultaneous delivery of anti-inflammatory agents to mustard-induced skin injury

Abstract

The molecular pathology of a sulfur mustard injury is complex, with at least nine inflammation-related enzymes and receptors upregulated in the zone of the insult. A new approach wherein inhibitors of these targets have been linked by hydrolyzable bonds, either one to one or via separate pre-attachment to a carrier molecule, has been shown to significantly enhance the therapeutic response compared with the individual agents. This article reviews the published work of the authors in this drug development domain over the last 8 years.

The pathophysiological chemistry of a sulfur or nitrogen mustard injury to skin is complex. Studies of sulfur mustard exposures to combatants and civilians during the Iran–Iraq War (1980–1988) have enabled the U.S. Army Medical Research Institute of Chemical Defense to chronicle the visual and experiential aspects of the injury.¹ Army experts reported that the exposed victims presented with burning sensations, desquamation, erythema, vesication/ulceration, necrosis, puritus, hyperesthesia, vitiligo, multiple angiomas, secondary infections, and xerosis over the next 45–50 days postexposure.

While a detailed gross visual pathology of the mustard-induced injury can be medically instructive about the formation and healing stages of the wound itself, it provides little guidance for rational new drug design. The molecular chemistry associated with the visual pathologies has advanced over the last 2 decades, and this illuminates opportunities for creative new drug development. At short times after contact between skin and the vesicant, inflammatory enzymes are upregulated, including inducible nitric oxide synthase (iNOS), transient receptor potential cation channel subfamily V member 1 (TRPV1), cyclooxygenase 1 (COX-1), and COX-2.^2–4 Subsequently peroxisome, fatty acid amide hydrolase (FAAH), esterlytic (acetylcholinesterase (AChE)), proteolytic (matrix metalloproteinase-2 (MMP-2), MMP-9) and other bond-cleaving enzymes further extend the injury.⁵ Our group and others have shown how inhibition of AChE can be a significant inhibitor of topical inflammation.^6,7

While the precise identifications of these vesicant-related targets have only recently been reported, U.S. Army researchers traditionally selected potential therapeutic candidates from long-established families of wound-healing agents.^8,9 These include the non-steroidal anti-inflammatories (NSAIDs), the antioxidants, the protease inhibitors, and the vanilloids.^8–10 Although the mechanism of anti-inflammatory activity is unknown for capsaicin mimics—except that they strongly associate at the vanilloid TRPV1 receptor—several groups have shown that candidate agents bearing the 4-hydroxy-3-methoxybenzyl (vanilloid) moiety can display considerable anti-inflammatory effects.^11–14 We have made use of this vanilloid effect in the design of our new generation of antimustard pharmaceuticals.

To permit quantitative comparisons of therapeutics, Casillas and colleagues developed the mouse-ear vesicant model (MEVM), in which the weight increase due to edema and inflammation caused by mustard (or phorbol ester/tetradecanoylphorbol-13-acetate (TPA)) in one mouse ear is compared to the suppression of that weight increase in the test ear. For the test ear, mustard and potential therapeutic are applied. The inflammation suppression is measured as a percentage, except for a few cases in which the candidate therapeutic augments and enhances the injury. Such irritants add to the edema induced by the mustard. Casillas’ MEVM has become the standard in vivo measurement of antimustard therapeutics.^15,16 We have used the Casillas MEVM to obtain all the measurements of inflammation suppression reported herein.^6,7,27

While Casillas found edema suppression following treatment with known COX inhibitors, established protease inhibitors, TRPV1 inhibitors, and several other categories of inhibitors, no evidence of summative effects was observed. More recently, other investigators have argued that bi- and trifunctional molecules in which constituent pieces targeted the individual enzymes could have impressive therapeutic benefits.^17–19 A bifunctional molecule in which two individually active components are tethered by readily hydrolyzable bonds such as esters, carbonates, and carbamates, may be an improved drug candidate. Thus, a single molecule can be both a prodrug and a facilitated, on-site controlled release platform for the individual therapeutic components providing a summative effect at the same pathological locus. In addition, if the bond linking the components is a suitably designed carbamate that can carbamoylate the active site of FAAH, then the bond itself can augment the overall activity of the conjugate.

One way to address multiple targets is with multiple drugs, a principle well illustrated by combination packages used in cancer chemotherapeutics.^20,21 Since our concern in this study has been for topical therapeutics for skin exposed to sulfur mustard or other external chemical insult, there is often a hidden benefit in considering a two- or even three-drug conjugate. Many anti-inflammatory classes are highly hydrophilic and hence poorly absorbed in topical dosage forms. As examples, consider the iNOS inhibitors aminoguanidine (ClogP = −2.75), 1400W (free base, ClogP = 0.49), and the nitro-guanidines [such as F₃CCH₂-NH(C=NH)-NH-NO₂, ClogP = −1.85]. Many AChE inhibitors, such as pyridostigmine (ClogP = −4.26), hexonium (ClogP = −9.09), Reminyl (free base ClogP = 1.03) and neostigmine (logP = −2.2), are poorly absorbed through skin. Similarly, while the common COX inhibitors, such as ibuprofen (ClogP = 3.68), diclofenac (LogP = 4.51), S-naproxen (ClogP = 2.82) and indomethacin (logP = 4.18), appear from their logP values to be sufficiently lipophilic for topic use, they are actually not. The NSAID carboxylic acid head groups often associate as dimers. They are ineffective for most topical applications, and they are strongly contraindicated on broken or wounded skin^22,23 (all ClogPs were determined using ChemBioDraw Ultra 14.0 and all logPs are derived from the published literature).

Thus, the design of covalent conjugates linking inhibitor types and including judicious use of nonpolar organic architecture—preferably with hydrolytic or enzymatic cleavage points (Fig. 1)—might dually provide a package of inflammation inhibitors on a single lipophilic prodrug-like platform. This approach has been successful for us in the design of mustard therapeutics, and we have revealed the results in our recent publications. Herein we review the range of inhibitor types for which we have prepared and tested topically useful anti-inflammatory agents. This review shows how the bifunctional approach to drug design can be extraordinarily useful in new interventional therapies for mustard injury on skin.

Figure 1.

General platform design of bifunctional drugs.

Linked COX and AChE inhibitors (type 1)

As noted above, the simultaneous inhibition of both COX and AChE is an attractive concept for anti-inflammation–related drug development. For that purpose, we have linked NSAIDs to the 3,3-dimethylbutyl acetyl moiety, the latter constituting a well-known lipophilic anticholinergic and bioisosteric mimic of acetylcholine.^24,25 A spacer group, the 4-hydroxybenzyl alcohol unit, was spliced between the 3,3-dimethylbutyl acetyl portion and the NSAID to enhance the lipophilicity The chemistry to accomplish the covalent conjugation of these components has been described elsewhere.⁷ Three examples of that class are displayed in Table 1.

Table 1.

Type 1 therapeutics: derivatives of NSAIDs linked to a 3,3-dimethylbutyl acetyl moiety

(1)
(2)
(3)

By themselves, tested singly in the in vivo model, the respective NSAIDs (COX inhibitors) incorporated herein as the left arm of the drug conjugate (the so-called inhibitor 1 in the general platform), were ineffective in suppressing irritation and edema triggered by topical application of the half-mustard chloroethylethylsulfide (CEES). Similarly, the right-hand arm designated inhibitor 2 in the general platform, the anticholinergic 3,3-dimethylbutyl acetate (IC₅₀ AChE 570 μM), has negligible anti-inflammatory activity by itself.²⁴ We previously reported these results in the MEVM: ibuprofen was an irritant and augmented the CEES-induced inflammation, S-naproxen measured 11% suppression, diclofenac yielded a 17% suppression, and indomethacin displayed a 46% suppression.⁶

In the linked conjugates (1) to (3), a different pattern emerges. In this lipophilic prodrug format, hydrolysis in 80% human plasma in phosphate buffered saline (PBS) at 37 °C liberated the inhibitors with half-lives of 2–7 h, depending on structure. Although these pharmaceuticals are intended for topical use, it is well known that plasma contains the same esterase, protease, oxidase, and other enzymes found in skin. We and others have used plasma as a testable surrogate for skin enzymes. Other than the observation that lability of the hydrolyzable bonds in these conjugates is necessary for prodrug effects to be manifest, we have not yet established what an ideal cleavage rate should be. Even in (1), in which the conjugated ibuprofen masked the usual irritant effect of NSAIDs, a modest 20% inflammation suppression was observed. The diclofenac conjugate (2) displayed a98% suppression and the indomethacin conjugate (3) a 91% suppression of CEES-induced mouse ear edema. Perhaps more amazing was that the anti-COX and anti-AChE effects of the original molecules, inhibitors 1 and 2, were retained in the conjugates. For (1), however, inhibition of COX-1 and COX-2 was barely detectable at 90 μM, although AChE inhibition was robust at an IC₅₀ of 1.93 ± 0.64 μM. For (2), COX-1 IC₅₀ was 0.88 ± 0.12 μM, COX-2 IC₅₀ was 18.0 ± 5.0 μM, and AChE IC₅₀ was 0.51 ± 0.02 μM, and for the indomethacin conjugate (3)m COX-1 IC₅₀ was 49 ± 0.1 μM, COX-2 IC₅₀ was 3.8 ± 0.3 μM, and AChE IC₅₀ was 2.3 ± 0.9 μM. The SAR design conclusion in these linked COX and AChE dual-function inhibitors is that AChE inhibition in the conjugate is more predictive of in vivo suppression of inflammation than is COX inhibition.

It should be noted that the lipophilicity of all these conjugates has been improved significantly into a range suitable for topical dosage forms: for (1), ClogP = 7.37; for (2). ClogP = 8.26; and for (3), ClogP = 7.87. In this class, as in the others, the component inhibitors are freed by in situ hydrolysis.

Linked COX and AChE inhibitors (type 2)

We have described a second type of anti-COX/anti-AChE conjugate (NSAID linked to galantamine) that provides considerable augmentation in anti-inflammatory activity over the individual components (Table 2).⁶ In this set, we studied the well-known anticholinergic AChE inhibitor galantamine (IC₅₀ = 1.12 ± 0.31 μM approved in the United States for the treatment of Alzheimer’s disease. As normally supplied in its hydrobromide salt form (Reminyl), the compound shows poor skin penetration, and in our MEVM we measured only 5% suppression of CEES-induced inflammation and 29% suppression of TPA-induced injury. The free base is somewhat better and shows an MEVM suppression of 69% for CEES and 72% for TPA. Compound (4)—the ester of galantamine and ibuprofen—proved to be very insoluble, and no useful MEVM data could be measured. Since in (4), the AChE IC₅₀—which has proven to be the best predictor of conjugate efficacy in the rodent model—was not impressive (> 40 μM the compound was not studied further. Additionally, in our hands ibuprofen itself has proven to be a topical irritant, and lipophilic conjugates of it show only modest improvement in the ear inflammation-suppression assay. The linked inhibitors of galantamine with indomethacin and naproxen (5) and (6), however, had considerably improved properties.

Table 2.

Type 2 therapeutics: NSAID–galantamine ester derivatives.

(4)
(5)
(6)

Compound (5), galantamine linked to indomethacin, gave one of the most significant AChE inhibitions (IC₅₀ = 0.49 ± 0.02 μM seen in any of our studies. This was 50% less than that of free galantamine. In addition, its COX-1 and COX-2 inhibitions were similarly impressive at IC₅₀ = 0.032 ± 0.002 μM and IC₅₀ = 1.70 ± 0.12 μM respectively. The in vivo MEVM screen showed 100% suppression of TPA-induced injury and 75% inhibition of CEES-induced inflammation.

Compound (6), galantamine linked to S-naproxen, gave significant AChE inhibition (IC₅₀ = 37.3 ± 1.2 μM as well as COX-1 and COX-2 inhibitions of IC₅₀ = 3.35 ± 0.16 μM and IC₅₀ = 18.0 ± 1.2 μM respectively. The MEVM results bore out the predictive quality of the in vitro markers, displaying 100% suppression of TPA-induced inflammation and 88% suppression of CEES-induced injury.

Here, again, the conjoined versions (5) and (6) were more lipophilic (e.g., ClogP = 6.38 and 5.01) for the two respective conjugates than for any of the individual inhibitors being transported. The inflammation suppression displayed by (6), especially in the cases of TPA inductions of the wound, were greater than the sum of those observed for individual COX (11% suppression for S-naproxen) and AChE (69% suppression for free base galantamine) inhibitors.

Linked fatty vanilloid carbamates (type 3)

O-Phenyl and O-benzylcarbamates with a proximal lipophilic domain are well known as desirable molecular architecture in slowly reversible inhibitors of FAAH.²⁶ Inhibition of FAAH—often accomplished with a carbamate candidate drug—has turned out to be one of the more successful concepts used to design novel anti-inflammatories.¹² In 2013, our group demonstrated how using an O-benzylic vanilloid (4-hydroxy-3-methoxybenzyl) attached to a fatty (lipophilic) domain—in lieu of simple unsubstituted benzyl moieties often reported in the literature—can lead to candidates that simultaneously antagonize both FAAH and TRPV1, with impressive anti-inflammatory effects.²⁷ In 2016, others also recognized the merit in linking traditional COX-inhibitory NSAIDs to FAAH inhibitors by a carbamate bond as a fruitful approach to the design of new anti-inflammatories.^28,29 Bifunctional and trifunctional drug design has become an exciting pathway to novel medicines, but our approach remains in uncovering novel therapies for mustard-related injury.

Compounds (7) through (11) (Table 3) exemplify the class of fatty vanilloid carbamate. We reported the syntheses (from a phenol-protected vanillyl alcohol) and product characterizations.²⁷ There are apparently two highly lipophilic pockets in FAAH whose precise structures control inhibitor access to an active-site serine. The aryl ring of a phenylalanine at the junction of the channels appears to act as a “dynamic paddle.” There is a “membrane-access channel” and a “chain-binding channel,” with the latter designed to position the flexible fatty chain of the substrate/inhibitor in an optimum location to allow for hydrolysis to occur.³⁰ The ultimate biological activity is highly dependent on chain length in the drug, as it reflects on the total molecular lipophilicity. Because carbinol-bond esters of p-hydroxybenzyl alcohol have—in some circumstances—proved unstable on storage or in polar media (decomposing to a quinonemethide)—we have elected to cap the phenolic hydroxyl with an acetate that is readily solubilized in the screening assays.²⁷

Table 3.

Type 3 therapeutics: vanillyl alcohol–derived FAAH inhibitors.

(7)
(8)
(9)
(10)
(11)

The octyl chain in (7) and the six-carbon cyclohexyl ring in (9), along with the phenylethyl moiety in (10), impart just the right amount of lipophilicity. ClogP is 4.4 for (7), for 2.73 for (9), and 2.80 for (10). All three of these vanilloid–carbamate–fatty chain constructs had excellent inflammation suppression in the MEVM at 72%, 91%, and 80%, respectively, but compound (7), at ClogP = 4.4, is slightly too lipophilic and scores the lowest in in vivo performance in the MEVM.

The role of relative lipophilicity in the design of FAAH inhibitors is nicely illustrated by the decyl analog (8), whose ClogP is 5.46, and the morpholino analog (11), whose ClogP is 1.04. The too-lipophilic analog scored 47% in the MEVM, while the highly hydrophilic analog (11) did equally poorly at 4% suppression. These results are mirrored in the in vitro FAAH inhibition measurements. Here, (8) at 104 μM and (11) at 410 μM are clearly in a different class from (7), (9), and (10), whose IC₅₀s fall in the 8.0–15 μM range.

In the fatty vanilloid carbamates (7) to (11), we combine the weak anti-inflammatory activity of 4-hydroxy-3-methoxybenzyl alcohol (vanillyl alcohol), used in folklore medicine, found in ginger extracts, and presumed to act at the TRPV1 receptor, with a fatty amino moiety and a carbamate bond, which are known components of FAAH inhibitors. There are literature reports on the anti-inflammatory effects of vanillyl alcohol, but none give quantitative information.³¹ We measured vanillyl alcohol in the MEVM and observed a 42% suppression of inflammation. By themselves, all the amines, (i.e., octyl, decyl, cyclohexyl, phenethyl, and aminoethylmorpholine) used to provide a lipophilic domain in our bifunctional drugs are topical irritants. It may be concluded that, except for (11), the resulting conjugates show the combined effects of their vanilloid and their carbamate features, since both the MEVM and the FAAH testing results demonstrate anti-inflammatory activity.

Conclusions

Using three different therapeutic families, we have demonstrated that the synthetic combination of two candidate inhibitors of inflammatory targets linked by hydrolyzable bonds can generate new molecules whose therapeutic properties in a mouse-ear vesicant screen exceed those of their component parts. One can tailor the new candidates so that they possess enhanced lipophilicity for topical application, suppressed dermal irritation, and greater efficacy in addressing mustard-generated injuries. This paper reviews and summarizes the work of the authors in anti-inflammatory new agent development.^3–7,27,32