How can we improve peptide drug discovery? Learning from the past

KEYWORDS:

• cyclic peptides
• drug design
• peptides

1. Introduction

With more than 80 peptides now FDA approved and hundreds more in preclinical or clinical trials there is no doubt that peptides are having an impact in the pharmaceutical industry [1–4]. That this is occurring reflects the undoubted advantages of peptides, including high potency and selectivity for their therapeutic targets relative to traditional small molecule therapeutics. The potency and selectivity in part derives from the larger size of peptides relative to small molecule drugs, allowing them to capture more binding interactions with target receptors, thus having high affinity (and hence potency) for those receptors but not with others (i.e. driving selectivity). Additionally, peptides have the advantage of filling a gap in molecular size space between small molecule therapeutics (<500 Da) and larger (>5000 Da) protein-based biologics [3]. Biologics generally do have excellent specificity and potency but are typically not orally available like their small molecule counterparts and are generally more expensive than small molecule drugs. Peptides thus have the potency advantage of biologics and the cost advantage of small molecule drugs. Furthermore, peptides have the potential to address classes of targets that previously have been regarded as ‘undruggable,’ including protein–protein interactions in the intracellular space [5].

These advantages of peptides are reflected in the current market estimate for therapeutic peptides of US $15.5 billion [6], and by the significant number of recent approvals for peptide-based drugs. For example, more than 10 peptide-based drugs have entered the market since 2017, including Zegalogue®, Lupkynis^TM, Mycappsa®, Scenesse®, Rybelsus® and Vyleesi®. The therapeutic indications addressed by peptides are diverse and include metabolic disease, oncology, infectious disease, CNS diseases, autoimmune diseases and a wide range of others [7]. Metabolic disease accounts for the majority of approved therapeutic peptides to date, closely followed by oncology applications.

Despite the current market successes, in the past the enthusiasm for peptides by the pharmaceutical industry has been somewhat muted by the difficulty in achieving oral bioavailability for peptides [8]. This lack of oral bioavailability is clearly apparent from the fact that the administration routes of currently approved peptide therapeutics are dominated by the subcutaneous (78%) and intravenous (13%) routes [6]. Other factors that have limited peptide market penetration include concerns about their potential manufacturing costs [9,10], short biological half-lives, and poor tissue penetration [11]. For clarity, we note that in this editorial we define peptides as comprising ~50 or fewer amino acids, and that we specifically focus on structurally constrained peptides for the rest of this article.

Over the last two decades multiple approaches have been tried to ameliorate the limitations of peptides, including the use of cyclic [12,13] or otherwise constrained peptides, including disulfide-rich peptides [14], natural ultra-stable scaffolds such as cyclotides onto which a desired bioactivity can be grafted [15], as well as artificially constrained peptides including stapled peptides [16] or peptides built on ‘bicycle’ molecular templates [17]. Most, or indeed all, these approaches have shown promise across a range of applications, with many lead peptides still working their way through the development pipeline and we expect positive outcomes from many of these in the next few years.

Before constrained macrocycle or template-based approaches became popular, peptide leads were typically discovered by observing the pharmacological activities of endogenous peptides in humans or animals and then deploying that peptide, or a mimic, to deliver those activities in a therapeutic setting. Insulin was the first and still most prominent example of that approach. In another early approach, peptidic extracts from bacteria, plants, or animals were screened for a desired activity that could be developed into a therapeutic. The conotoxins [18] from venomous marine snails being screened for ion channel activity are an example of that approach, which led to the development of the analgesic peptide ziconotide which works by blocking calcium channels in the spinal cord, as is exenatide, a synthetic version of a peptide derived from the saliva of the Gila monster which is used for the treatment of type 2 diabetes [19]. These two examples illustrate one of the challenges of peptide drug development in that neither is orally administered. Ziconotide is delivered via a surgically implanted pump and has achieved only moderate market success whereas exenatide delivered via subcutaneous injection achieved much greater market success. Space limitations in this editorial prevent a detailed discussion of the approaches used for the extraction, isolation and purification of such natural product peptides but these are covered in recent reviews and references therein [15,18,19]. However, we do note that one of the advantages of peptides over other natural products is that the identification of peptides is nowadays greatly facilitated by the availability of transcriptomic or genomic information [20].

Combinatorial chemistry approaches in which diversity in peptide sequences is achieved by ‘brute force’ combinations of amino acids was the next approach, which was then followed by library-based methods using biology to play the numbers game, i.e. to increase the number of peptides that can be screened rapidly against a particular therapeutic target. Such libraries ranging in size from 10⁶ to 10¹² members and various display modalities have been applied, including phage [13] bacterial [21] and mRNA-based approaches [22]. These approaches are often very successful in finding hits but if there has been a limitation in them, it is that the hits often do not have the right biopharmaceutical properties to turn them readily into drugs. The process of peptide-based drug development thus needs improvement. What can be done and how can lessons from the past help?

2. Expert opinion

One of the lessons from the past is that even if nature provides inspiration for a peptide-based drug lead, it is rare for a natural peptide to be the final product itself. Some type of medicinal chemistry optimization is typically needed to improve biopharmaceutical properties such as solubility, stability or biological half-life. For example, the development of the anti-diabetic peptides liraglutide and semaglutide which, like exenatide, are mimics of the human incretin GLP-1, required amino acid substitutions to avoid in vivo degradation and lipidation to increase circulating half-life [23]. The need for such modifications arises because in their natural endogenous setting peptides are tightly regulated by the innate physiology of the organism for their production and programmed degradation, but these controls are not present when delivered exogenously as isolated molecules, whether by oral, injection, or other delivery routes. Thus, another lesson from nature is that in future we might consider delivering peptides in more complex formulations that include such regulatory controls, for example driven by biological auxiliaries, such as processing enzymes for dose sensing, activation, and/or deactivation. This would be an extension from the current use of passive delivery control methods, including slow-release formulations common in small molecule drugs, or mechanical control devices such as implanted pumps used in the delivery of ziconotide or insulin. It would also build on successes already achieved in the local activation of drugs at the target site, for example including the promising results seen with the activation of peptide-based prodrugs by tumor microenvironment proteases, oxidative potential, or pH [24].

Expanding on the issue of peptide delivery, we note that while oral delivery is certainly a holy grail of peptide-based drug design, oral delivery is not necessary or even appropriate in all cases. Thus, another lesson that can be learned from nature is that the delivery route should be fit for purpose. For example, many plants produce peptides to protect themselves from foraging insects, so target oral ingestion as the delivery route. This requires protection from digestive proteases, explaining why plants are so good at producing stable peptide scaffolds such as cyclotides [15]. By contrast, many animals use venom injections for prey capture, as is common in snakes, spiders or fish hunting marine cone snails, the latter being where the natural peptide corresponding to ziconotide comes from. In those cases, direct injection into the target organism is the optimal delivery route. That leads to the next lesson from the past, albeit one that is well known, i.e. that peptide delivery will depend on the nature of the target and/or therapeutic setting.

Regarding peptide delivery, we stress that oral activity is not the same as oral bioavailability, and that there are cases where oral delivery is desired, but systemic oral bioavailability is not. The most obvious case here is for gut disorders where the receptor sites are in or near the lumen of the digestive tract. Linaclotide, a 14 amino peptide used in the treatment of IBS is a good example here. The stabilization of this peptide by its three disulfide bonds is sufficient to protect it from digestive proteases. These types of disorders, where oral delivery is desired, but systemic exposure is not, are being explored by multiple pharmaceutical companies, including Protagonist in the development of a gut-restricted, interleukin-23 receptor antagonist peptide, currently in a Phase 2 clinical trial for Crohn’s disease. These applications aside, it is still the case that oral delivery remains a major challenge for peptide-based drug design, and approaches ranging from cyclization, N-methylation, stapling, and scaffold grafting are being pursued.

A final lesson from nature is that bacteria, plants, and animals have developed stable scaffolds for delivery of their bioactive peptides, and as we noted above, chemists are now using some of these scaffolds as templates in drug design. While this approach has met with success in producing lead molecules, in our opinion more effort needs to be directed to choosing or designing scaffolds and approaches that will lead not only to binding interactions with the target receptor but also in scaffolds having the right biopharmaceutical properties that will allow the peptide to reach the receptor, which in some cases will involve crossing biological membranes either in the epithelial lining of the gut or at the cell surface for intracellular targets. In other words, effort needs to be directed at building in membrane transport and/or bioavailability. So far, there has been a high degree of success in building in stability to peptides, notably for macrocyclic peptides [3], stapled peptides [16], and bicyclic peptides [17]. Recent developments suggest that computational approaches also have the potential to play a role in designing peptide scaffolds that will have high stability [25].

There have been numerous advances over recent years in increasing the oral bioavailability of certain classes of peptides via approaches involving N-methylation [26–28] or other modifications to amino acids that are now accessible via advances in synthetic chemistry [29]. Formulation approaches are also critical to oral delivery [28]. Furthermore, the factors that enhance cell penetration for specific classes of peptides are now beginning to be better understood [30]. However, the standard drug design approach of first generating hits with a desired biological activity and then using medicinal chemistry approaches to optimize biopharmaceutical properties may compromise the activity. Likewise, the approach of starting with a scaffold that has intrinsically beneficial biopharmaceutical properties (such as the stability and cell penetrating ability of cyclotides) and grafting a bioactive payload to it has the limitation that the final molecule might lose those beneficial biopharmaceutical properties when the payload is attached. Future studies are therefore likely to benefit from a more holistic approach where both the payload and delivery scaffold are considered early in the design phase. Screening approaches that select for both affinity and proteolytic stability are already in use, for example. In our opinion, computational approaches to the design of hyper-stable scaffolds [25], possibly combined with computerized docking [31] and machine learning [32] could play a major role. For example, a recent collaboration between our group and that of David Baker at the University of Washington, USA, demonstrated the ability to computationally generate a peptide sequence having a desired pre-defined structure. That work is a precursor to the ability to design ligands (and hence potential drug leads) that interact with a chosen molecular target. In that regard computational design offers the potential to go beyond nature and thus access a greater diversity of chemical space. One of our designs involved a heterochiral molecule where it was desired to have one left-handed helix and one right-handed helix, by using a mixture of L- and D-amino acids [25]. Massively parallel computational arrays combined with high-throughput production capabilities already exist to facilitate computational design approaches [33].

Another issue that needs to be addressed is cost-of-goods and environmental impacts. Cost of manufacture will be particularly critical for cases where oral delivery is desired, as the typically poor oral bioavailability of peptides means that larger doses, and hence larger manufacturing capability will be required than would be the case for injectables. This is relevant for example with the recent development of Rybelsus^TM, an orally administered version of semaglutide, for which a substantial investment ($80 million) in manufacturing facilities was required to accommodate demand. Most approved peptides are commercially manufactured using chemical (solid phase) synthesis, but recombinant approaches become more cost effective as the peptide size increases, with insulins being typically being produced using the recombinant route. While the recombinant production of proteins in Escherichia coli is common, it is not always successful for peptides and especially for small disulfide-rich peptides. To overcome that limitation, a yeast-based expression system tailored to cyclic peptides was recently reported [34]. This approach has the advantage of secreting the desired product, making purification easier.

Another approach being developed in parallel is the use of plants for peptide production. Specifically, many cyclic peptides such as cyclotides are naturally produced in plants and in future modified ‘designer’ cyclic peptides might be produced in such plant biofactories by the inclusion of the appropriate precursor genes. For example, a potent plasmin inhibitor was recently produced in a rapid and scalable approach in the plant Nicotiana benthamiana [35]. This host plant is readily able to be genetically transformed and is amenable to transient expression. It was used for example to express an antibody treatment during the Ebola outbreak in 2014. We note that plant-based expression is not likely to become the main route of manufacture of peptides but that it may be well suited to certain applications. In our own work, for example, we are exploring options for selectively producing therapeutic peptides in plant seeds, whereby the seed could effectively become a ‘biopill.’

Finally, while the current performance of peptides has been promising, improvement is needed across all types of drug modalities. A take home message from a study of 2018 FDA approvals [36] is that the cost of new drugs is rising and in many cases becoming unaffordable for a large proportion of patients, even in countries with well-developed health care systems. Peptides don’t necessarily provide a solution to this problem based on relatively expensive solid-phase synthesis approaches, but the ‘plant biofactory’ approach offers a promising alternative that might be attractive for some classes of peptides. Such an approach also promises to be more environmentally friendly in terms of production and potentially opens the possibility of making potent and specific peptide-based medicines more accessible to patients in developing countries. It would metaphorically close a loop whereby we will have gone from discovering bioactive peptides in plants by random screening, to making ‘designer’ peptides in plants, thus not only learning form nature, but improving on nature.

Declaration of interest

DJ Craik is an ARC Australian Laureate Fellow (FL150100146). The authors have no other 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 apart from those disclosed.

Reviewer disclosures

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

Additional information

Funding

Work in the authors’ laboratory on peptides is funded by a grant from the Australian Research Council (ARC) through the ARC Centre of Excellence for Innovations in Peptide and Protein Science (CE200100012).