Soft drugs: design principles, success stories, and future perspectives

KEYWORDS:

• Antedrug
• drug metabolism
• esterases
• methylphenidate
• prodrug
• sofpironium
• therapeutic index
• toxic metabolite

1. Introduction

Soft drugs (SDs) are therapeutically active compounds that undergo a predicted fast metabolism into inactive metabolites after exerting their desired therapeutic effects. The goal of SD design is to control and direct metabolism, typically by incorporation of a metabolically sensitive moiety into the structure. The soft drug approach was explicitly introduced 40 years ago, when it was clearly highlighted in the title of a series of articles [1]. The terminology was selected to contrast that of Ariëns referring to nonmetabolizable hard drugs, i.e. drugs designed to not be metabolized at all and thus avoid the problems caused by reactive intermediates or active metabolites. However, there are very few drugs that form no metabolites whatsoever; possibly, some highly water-soluble drugs that are not metabolized in vivo can be considered as examples of metabolically hard drugs (e.g. enalaprilat, lisinopril, cromolyn, and alendronate). Hence, soft versus hard here represents the propensity of the drug toward being metabolized and not toward causing physical addiction and societal harm – the more common, albeit not very rigorously defined everyday usage of soft versus hard drugs [2].

2. Soft drug design – basic principles

The SD concept [3] is part of the more general recognition that drug design needs to (1) fully integrate metabolic considerations from the very beginning as metabolites contribute significantly to the overall activity and toxicity profile of the original drug Figure 1 (2) focus not on improving activity alone, but on improving the activity/toxicity ratio. This is usually characterized by the therapeutic index, typically defined as the ratio between the half-maximal toxic and effective doses: TI = TD₅₀/ED₅₀. These ideas are the main underlying principles of retrometabolic drug design, which incorporates both SD and chemical delivery system (CDS) design [4].

Figure 1. Illustration of the effects that metabolites have on the overall effect and toxicity of therapeutic agents. (a) Case of a hypothetical drug (D) that is metabolized into active, toxic, and inactive metabolites (M_a, M_t, and M_i, respectively). Corresponding time-profiles for concentrations, effects, and toxicities (arbitrary units) are shown for all components assuming that D is metabolized with different rates into these metabolites and they all have different elimination rates as well as activities and toxicities. For the case shown here, most of the effect is contributed by the original drug D and its active metabolite M_a, but most of the toxicity results from the toxic metabolite M_t. (b) Same for a hypothetical soft drug (SD) that is metabolized rapidly into a single inactive metabolite M_i of negligible activity and toxicity. Contrary to A, both activity and toxicity are driven only by those of the original SD even if M_i concentrations are quite high. Hence, complex PK profiles and toxicities due to reactive metabolites are avoided, and activity and toxicity can be better controlled. Time-profiles were obtained using single-compartment models with absorption, metabolism, and elimination rate constants (k_abs, k_m,a, k_el, etc.) and corresponding differential equations as indicated. All effects and toxicities were assumed to be directly proportional with concentrations (i.e. no effect compartment).

For most drugs, several metabolites are formed following administration, and they can contribute significantly not just to the overall activity, but also to toxicity and side effects. This can lead to complex time-profiles as illustrated in Figure 1(a), which shows the case of a hypothetical drug D that is present together with its active, toxic, and inactive metabolites. In about 70% of drugs associated with toxicity, reactive metabolite formation was a cause of toxicity [5]. A well-designed SD can lead to a much more simplified scenario (Figure 1(b)).

For SDs, inactivation should be relatively fast and free of interference from possible drug-drug interactions. Because cytochrome P450 enzymes, such as CYP3A, CYP2C9, and CYP2C19 oxygenases that are responsible for the metabolism of most drugs, are saturable and subject to inhibition and induction, inactivation by hydrolytic enzymes could be a better choice. This metabolism (e.g. cleavage of an ester bond) can be carried out rapidly and even extrahepatically by ubiquitously distributed esterases, which may not be entirely free of inhibition or polymorphism issues [6], but far less than CYPs (e.g. there are no known carboxylesterase-mediated drug-drug interactions in the clinic so far except for some reported interactions with ethanol [7]). Hydrolytic degradation also has the advantage that it forms an acidic metabolite and introduces a negative charge that is likely to not fit adequately in the ligand binding site, thus considerably diminishing receptor binding compared to the original ester-containing drug and increasing the likelihood that the metabolite is indeed inactive.

Despite being conceptual opposites, SDs are sometimes still confused with prodrugs, mainly because (1) both undergo predicted metabolic changes and (2) both rely primarily on enzymatic hydrolysis. SDs, however, are active per se and are inactivated by a built-in mechanism, whereas, prodrugs are inactive and must be activated. A considerable portion of existing small-molecule drugs (10–15%) are, in fact, prodrugs [8,9]. Most, however, are accidental prodrugs that require metabolic activation, but were not intentionally designed to do so. Nevertheless, rational prodrug design is increasingly successful and resulted in several clinically approved products accounting for ~10% of new drugs approved in the last decade [9]. Unfortunately, the SD/prodrug confusion is only deepened by some authors who use the antedrug term to designate SDs [10]. This is misleading and should be abandoned as the Latin ante- prefix is similar in meaning to the Greek pro- (e.g. prior to, precedent); hence, antedrug suggests the same as prodrug (i.e. need for metabolic activation and not inactivation) and, thus, the opposite of what it is supposed to stand for.

3. Clinical success stories

During the four decades since the introduction of the concept, the SD approach has been explored in almost all therapeutic areas by various academic and industrial research centers including several major pharmaceutical companies. A number of these projects led to clinically approved products. Rationally designed SDs that reached marketing approval include, for example (Table 1):

• the ultra-short acting soft β-blockers esmolol and landiolol

• the soft glucocorticoid loteprednol etabonate approved for inflammatory and allergy-related ophthalmic disorders

• the ultra-short acting soft opioid analgesic remifentanil

• remimazolam, a soft benzodiazepine approved for general anesthesia

• the ultrashort-acting soft calcium-channel blocker clevidipine approved for i.v. use in the reduction and control of blood pressure in cardiac surgical procedures

• sofpironium bromide, a soft anticholinergic glycopyrrolate analog likely to receive approval for hyperhidrosis treatment (NDA submitted in Japan).

Table 1. Rationally designed or accidental soft drugs that are approved for clinical use (trademark provided) or have reached clinical development. For each compound, the entity responsible for its development, the year of the first publication, and the approved trademark (if available) are also included.

Class	SD Compound	Trademark (year of approval)	Developed at	First publication
Beta-blockers	Adaprolol		Xenon Vision	1982
	Esmolol	Breviblock (1986)	American Critical Care	1982
	Landiolol	Onoact (Japan 2002); Rapiblock (Europe 2016)	Ono Pharmaceuticals	1992
Opioid analgesics	Remifentanil	Ultiva (1996)	Glaxo Wellcome	1991
Benzodiazepines	Remimazolam (CNS 7056)	(Japan 2020)	GlaxoSmithKline/PAION	2002
Anesthetics	Etomidate [accid.]	Amidate	(accidental, Janssen Pharm.)	(1965)
	Articaine (local) [accid.]	Ultracaine, Septocaine	(accidental, Hoechst)	(1969)
	AZD3043		AstraZeneca	2004
	ABP-700 (CPMM)		Harvard/Medicines Comp.	2009
Corticosteroids	Loteprednol etabonate	Lotemax, Alrex, Zylet (1998)	Pharmos/Bausch & Lomb	1981
	Etiprednol dicloacetate		IVAX Res.	2002
	Fluocortin butyl	Vaspit, Novoderm (1977)	Schering AG	1975
	Itrocinonide		Astra	2002
Ca channel blockers	Clevidipine	Cleviprex (2008)	Astra Hässle	1999
Anticholinergics	Sofpironium (BBI-4000)	(NDA in Japan 2020)	Bodor Lab/Brickell Biotech.	2002
Psychostimulants	Methylphenidate [accid.]	Ritalin, Concerta, Daytrana	(accidental, Ciba)	(1944)
Antiarrhythmics	Budiodarone (ATI-2042)		ARYx	1996
	Celivarone (SSR149744C)		Sanofi-Aventis	2011
Gastro-prokinetics	Naronapride (ATI-7505)		ARYx	2007
Anticoagulants	Tecarfarin (ATI-7505)		ARYx	2009
PDE4 inhibitor	Lotamilast (E6005, RVT-501)		Eisai	2013
	LEO-29102		Leo Pharma	2014

Structures are shown in Figure 2 together with corresponding non-soft analogs that served as the starting points of their designs. Crisaborole, a PDE4 inhibitor approved for the treatment of atopic dermatitis that has a benzoxaborole and not an ester moiety, has been suggested as an example of non-canonical SD because it is extensively metabolized via oxidative deboronation and subsequent hydrolysis into inactive metabolites [11]. However, it shows some systemic accumulation and an elimination half-life >10 h. Since other similar compounds (e.g. the antifungal tavaborole) show good metabolic stability and long elimination half-lives, one cannot consider this function as a soft spot in general. Even before the formal introduction of the SD concept, there were therapeutic agents with metabolically labile ester moieties in their structure [4]. Just as there are accidental prodrugs, there are also accidental SDs, i.e. approved drugs that are, in fact, SDs even though they were not intentionally designed as such, for example (Table 1, Figure 2) [12]:

Figure 2. Clinically approved SDs including both rationally designed ones (top section, shown in parallel with the original drug structures that served as lead for their design in the right column) and accidental ones (bottom row). The metabolically labile soft spots are highlighted in magenta.

• etomidate, an ester-containing sedative hypnotic, which was one of the most frequently used agents in emergency settings until it was replaced by propofol

• articaine, a local anesthetic widely used in dentistry

• methylphenidate, an amphetamine-related methyl ester used for the treatment of attention-deficit-hyperactivity disorder (ADHD) that can be considered a soft psychostimulant.

4. Future perspectives

When evaluating the potential of drug design and development approaches, one needs to remember that the field faces increasing challenges for a number of reasons including, but not limited to the ever-expanding regulatory burden, the shrinking pool of remaining new druggable therapeutic targets, and the increasing but unrealistic public expectation that there should be no side effects or abuse potential – not to mention the problems of managing the needed highly multidisciplinary and collaborative research and development (R&D) work. Not surprisingly, introducing new drugs to the market is increasingly difficult, and the number of FDA-approved drugs the pharma industry could develop for the same amount of R&D spending (e.g. inflation-adjusted 1 USDB) has been decreasing exponentially. Since 1950, it has been halved quite consistently every nine years – a sort-of reverse Moore’s law and the exact opposite of the exponential growth in computational power [13].

In light of these, the SD approach, which provides general drug design strategies, has particular potential. Because it often starts from a known active structure and focuses on designing safer drugs by decreasing side effects and toxicity, the likelihood of success is increased, especially considering the perspective highlighted by Sir James Black: ‘the most fruitful basis for the discovery of a new drug is to start with an old drug’. SDs are especially likely to succeed in areas where the desired activity is localized, relatively short-lived, or susceptible to easy titration. Along these line, anesthesiology, where adjustable control during the surgical procedure and quick recovery at the end are desirable, is one area particularly well-suited for SD drug applications [14,15]. Another one is dermatology, where topical application is convenient, as the skin is easily accessible, and localized activity is particularly desirable [11]. Similarly, ophthalmology and inhalation therapies have considerable yet unexplored potential for SD design [16]. Computational tools are now available to assist such future applications including a computer-aided SD expert system that integrates SD-specific structure-generating rules with a calculated properties-based ranking algorithm [17]. This can generate new virtual libraries of SD structures and assist in selecting the most promising new candidates.

Among challenges, one needs to mention the importance of achieving an adequate balance between maximizing the desired local activity and minimizing the undesired systemic toxicity: SDs have to be sufficiently stable to reach their intended targets/receptors and produce their desired effects while remaining sufficiently fragile to not cause unwanted systemic side effects. Several SD designs failed in the end because the metabolic degradation was too fast and acceptable activity could not be achieved. For ester-containing drugs, including SDs and prodrugs, a further challenge is that esterase activities vary strongly among species as well as among organs and tissues. This must be considered during preclinical evaluations, especially because rodents tend to hydrolyze much faster than humans (at least for aliphatic esters) [4]. Fortunately, considerable recent progress has been made in characterizing esterase activities and their impact on drug metabolism [7]. Differential distribution of such enzymes, where it exists (e.g. paraoxonase 1, which is active in plasma, but not in most tissues), can be exploited to provide further targeting [11,18]. Regarding the future of SDs, it is encouraging that there are many new projects initiated relatively recently, and more than a few will certainly lead to new clinically approved drugs. Some of the more intriguing one include soft JAK inhibitors, ROCK inhibitors, phosphodiesterase 4 (PDE4) inhibitors, TRPV1 modulators (soft capsaicin analogs), HDAC inhibitors, immune-modulators (e.g. soft cyclosporine and tacrolimus analogs), cytokine modulators (e.g. IL-5 inhibitors and TLR7 agonists), S1PR1 inhibitors (soft fingolimod analogs), cannabinoids, and many others [11,12].

5. Expert opinion

Soft drug design is part of the more general retrometabolic drug design concept. It is very general and can be applied in a wide range of therapeutic areas to generate innovative, new chemical entities. The goal is to control and direct metabolism by incorporation of a metabolically, preferentially hydrolytically labile moiety into the structure. In most cases, this is achieved via SDs that are close analogs of known successful lead compounds and by using an ester as the metabolically sensitive soft spot. However, both de novo design (i.e. novel structural frameworks) and non-ester type soft spots are possible. All SD projects need to include detailed metabolic characterization and confirmation of the inactivity of the metabolite. SDs are particularly well-suited for therapeutic applications where the desired activity is localized, short or ultrashort, or susceptible to easy titration. Along these lines, anesthesiology, dermatology, ophthalmology, and inhalation therapeutics represent areas of particular promise. During the four decades since the introduction of the concept, several rationally designed SDs received marketing approval (Table 1, Figure 2), and there are a variety of interesting applications and projects initiated by academic and/or industrial researchers in development.

Declaration of interest

The authors have no relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.

Reviewer disclosures

Peer reviewers on this manuscript have no relevant financial or other relationships to disclose.

Additional information

Funding

This paper was not funded.