Pharmaceuticals from marine sources: past, present and future

Abstract

A large proportion of the current pharmacopoeia can trace its origin back to nature. Biodiversity in the Earth’s largest habitat, the ocean, is greater than that on land, suggesting a huge potential for new bioactive chemistry. Compounds isolated from marine organisms have novel structures and modulate human disease targets with novel mechanisms of action. Around 15–20 marine-derived compounds have been approved for clinical use against cancer, pain, viral infection and heart disease. Some of the most potent compounds are directed towards cancer cells via linkage to an antibody. The approved marine-derived pharmaceuticals have been produced via (semi)-synthesis, while expression of the biosynthetic gene clusters in a microbial host is a future alternative for the production of those compounds. Technical improvements in the extraction, isolation and structural characterisation of marine-derived compounds, as well as the reduction in effort wasted in isolation of known compounds, have sped up the discovery process, meaning that industry is taking a renewed interest in this field as a source of pharmaceuticals with novel mechanisms of action to treat human disease with unmet need.

Key words: Marine natural products, drug discovery, aquaculture, semi-synthesis antibody-drug conjugates

Introduction

Compounds of natural origin constitute a large proportion of the current pharmacopoeia, with around 70% of currently used antibiotics and anticancer agents tracing back to natural sources, such as plants and microbes^​1​. While plant-based medicines have a recorded history of over 5,000 years and microbially-derived medicines have been in use for almost a century, compounds from marine sources have recently only been investigated as potential medicines comparatively recently^​1​. Recent clinical approvals started at the turn of the millennium and to date more than 15–20 marine-derived pharmaceuticals have been clinically approved, mainly for cancer treatment predominantly owing to funding by the US National Cancer Institute from the 1960s^​2​.

The exact number of pharmaceuticals of marine origin that have received approval is debated, as some are closely related structurally^​2​. This review outlines why marine natural products (MNPs) are an attractive prospect for future drug discovery, provides a history of the field, successes in the clinic and challenges associated with gaining clinical approval, including the need to obtain sufficient amounts of often very complex compounds. The pipeline of compounds currently in preclinical evaluation is promising and sophisticated technologies will improve the ability to discover, isolate, identify and modify marine-derived compounds with potent and selective activity, and new mechanisms of action against a range of human diseases that are currently without adequate treatment.

Marine biodiversity as sources of new pharmaceuticals

The driver behind natural product (NP) discovery for pharmaceutical applications is that the greater the biological diversity investigated, the greater the chemical diversity will be found. For this reason, the pioneers of MNP research investigated marine invertebrates from phyla with no terrestrial representatives, such as sponges, soft corals and sea mosses^​3​. Marine animal biodiversity at the level of phylum is greater than that on land; while there are around 20 marine phyla with no terrestrial representatives, there are few animal phyla that have only terrestrial representatives^​4​. 

Global distribution can be mapped to show that there are biodiversity hotspots and that the deep seas have only been sparsely investigated^​5​. This phyletic diversity at a very high level is complemented by a high diversity of marine microorganisms, the extent of which is not yet fully appreciated. It is hypothesised that conditions in the seas increase evolutionary pressure to produce unique bioactive molecules^​6​. These conditions include high hydrostatic pressure, high salt concentrations, low light and oxygen levels, as well as the fact that many invertebrates have no obvious physical defences and lack an immune system^​7​. Filter-feeding invertebrates should be subject to predation and overgrowth as well as infection by pathogenic microorganisms, but some have developed a chemical defence strategy that acts as an alternative to an immune system. The NPs produced as part of this chemical defence often have properties similar to drug molecules as they have to traverse many of the same barriers to reach the site of action, which may be an intracellular receptor^​6​.

Once screening of marine invertebrates started, it appeared that certain phyla had a higher occurrence of bioactivity against human disease screens than terrestrial NPs. Success rates are hard to quantify, but a simplistic estimation divides the number of known MNPs (around 40,000) by the number of approved marine-derived drugs (between 15 and 20), giving a success rate of around 1 in 2,000–2,700^​8​. MNPs appeared to have greater chemical diversity and a greater frequency of new skeletons than other natural sources, suggesting they could have novel bioactivities and modes of action against disease targets^​9​. A recent study demonstrated that three-quarters of marine microbial NPs are structurally closely related to those from terrestrial microbes^​10​. The occurrence of many of these unique microbial NPs in marine invertebrates suggests the potential microbial origin of some marine invertebrate NPs. Some of these unique skeletons have shown unique modes of action, such as ecteinascidin-743, isolated from a Caribbean sea squirt (see below), which binds to the DNA minor groove, thus bending the helix to the major groove. This triggers a cascade of events that affects several transcription factors, DNA binding proteins and DNA repair pathways, resulting in perturbation of the cell cycle^​11​. There are many other examples of novel modes of activity that will be discussed throughout this article. The current success of MNPs in the clinic and the pipeline appears to be linked to the evolutionary success of marine invertebrates and micro-organisms, as well as the structurally unique chemistry they produce.

Processes for marine natural product drug discovery

A typical workflow for MNP drug discovery starts by obtaining biomass; for instance, a marine invertebrate or a microbial culture obtained from a marine source (e.g. sediment, sponge etc), which is then subjected to large-scale fermentation (see Figure 1). 

The Pharmaceutical Journal/Shutterstock.com

This biomass is then extracted using chemical solvents, after which the solvent is removed generating an extract potentially containing hundreds to thousands of compounds. Downstream steps are to fractionate these extracts to remove water-soluble compounds, such as salts, sugars and highly soluble metabolites and lipophilic compounds (e.g. fats). The fractions can be subjected to biological assays relevant to human disease followed by a cycle of chromatographic purification and assays to pinpoint the biologically active compounds^​12​. Often these steps are complemented by using liquid chromatography — mass spectrometry (LC-MS) to determine which of these extracts and fractions contain compounds that have not previously been discovered to target effort towards novel bioactives. Once a pure bioactive compound is identified, its structure is determined using MS in conjunction with nuclear magnetic resonance (NMR) data^​13​. The pure compound can either be used as is, or its structure modified. In all cases, the material will need to be obtained in higher quantities for the downstream mechanism of action (MoA) and animal testing before embarking on an investigational new drug application, potentially leading to human clinical trials, which can take many years before approval. More information can be found under ’Supplying enough material for clinical trials and clinical use’^​2​.

Investigating natural products from marine sources

NPs have been utilised as ‘drug entities’ for millennia, including morphine and, more recently, penicillin. Towards the end of the 20th century, there was a significant move towards “high technology systems” including the concept of diversity-oriented syntheses (DOS), whereby millions of compounds could be easily synthesised and then tested in advanced biological systems. As of October 2019, after testing tens of millions of compounds in countless screening systems, at best four approved drugs (meaning approved by the US Food and Drug Administration (FDA) and/or the corresponding entities globally), have been identified for any human disease from utilising such methodologies^​1​. This demonstrates that such systems are useful for developing an active agent once identified, but they fail in the discovery phase. 

With the advent of new analytical systems that permit the identification of relationships between NP isolates, coupled with the application of new genomic analytical capabilities, particularly with microbial sources, these capabilities have enabled identification of many potential new leads in disease-related screening systems (see below)^​1​.

Back to the 1950s and early marine investigations

Although there were chemists who liked to dive (mainly in warm waters) in the late 1950s, in the early 1960s and later, disparate opportunities came together. The first was the availability of reliable commercial scuba gear permitting non-military divers to explore up to roughly 40 metres deep using compressed air and then with mixed gases to greater depths. Early funding (1940s) possibly by the US (NIH) for basic studies on sponge metabolites led to some interesting compounds that can be considered the first marine bioactive agents that led to drugs used in humans. Funding from the US National Cancer Institute (NCI) in the 1960s to specific groups to collect, fractionate and then have in vitro/in vivo assays performed by NCI contractors. These groups worked on plants (USDA and Wall) and marine invertebrates (Pettit and Rinehart). Later funding (1986 onwards) for collections via NCI contractors and subsequent testing led to later drug entities (see below).

From the middle of the 1940s onwards, Bergmann and Feeney investigated the Caribbean sponge, (Tethya crypta) for novel terpenes, and from acetone extracts they discovered novel nucleosides with arabinose as the sugar moiety. The group reported the isolation and identification of spongothymidine (compound 1), spongouridine (compound 2) and spongosine (2-methoxyadenosine; compound 3), from the sponge (now renamed Cryptothethya crypta)^{​14–16​}. Although none were developed as drugs, they led to a significant number of both antiviral and antitumour leads/drugs over the next few years, such as acyclovir against herpes virus or 6-mercaptopurine, the nucleoside-based anticancer drug^​17​.

The early reports of potential bioactivity from compounds 1–3 concerned their potential as sources of uracil for E. coli auxotrophs^​18​. Research by synthetic chemists took the information from the first Bergmann paper and synthesised spongouridine (compound 2)^​19​; other arabino-nucleosides in 1957 by Fox et al.^​20​ and Walwick’s synthesis in 1959 of Ara-C (compound 5)^​21​ provided the necessary material for Lee et al. to uncover the antitumour potential of the semisynthetic derivative Ara-A in 1960 (compound 4)^​22​ and for Evans et al. subsequent discovery of the utility of Ara-C (compound 5) as an antitumour agent in 1961^​23​.

In 1969, microbiologists at Parke-Davis demonstrated the fermentative production, rather than semisynthesis, of Ara-A using the terrestrial microbe Streptomyces antibioticus. This was only published as a patent by Parke-Davis^​24​. Fifteen years later, the Cimino group in Italy reported the isolation of “biosynthesised Ara-A” together with spongouridine (compound 2) from the Mediterranean gorgonian coral Eunicella cavolini^​25​. In this case, synthetic chemists had made the compound before it was discovered in nature. These discoveries implied a microbial source, even though the “nominal sources” ranged from invertebrates to terrestrial microbes.

In 2015, the Gerwick group used a fresh collection of the Caribbean sponge Bergmann used and isolated a strain of Vibrio harveyi from it; on fermentation it yielded spongosine (compound 3). Therefore, it is a reasonable assumption that bacteria were responsible for the Bergmann compounds 50 or more years earlier^​26​.

Microbial involvement in generation of marine natural products 

Moving into the 21st century, it has become increasingly obvious that microbes, whether they are currently fermentable, are most likely the actual source(s) of bioactive materials isolated from the Porifera and other marine invertebrates. The Piel group at ETH Zurich and their collaborators working with a sponge originally from Palau, and with similar sponges from Okinawa, have demonstrated that the source of the majority (>20 of the 35 plus) of bioactive agents isolated from the Palauan sponge are produced by a yet unfermented bacterium^​27​.

The T. swinhoei sponges, collected in Palau and Japan, both produce an analogue of the Amazonian Paederus beetle toxin (itself produced by an as-yet uncultivated Pseudomonad on the female beetle), demonstrating that nature utilises similar genomes across the gamut of sources^​28​ (see Figure 2). In 2022, a review article by Hemmerling and Piel highlighted how the biosynthetic novelty in bacterial genomes can be accessed, and how to further develop access to genomes^​29​. To date, there has not been an unequivocal isolation of a completely sponge-sourced bioactive agent, despite searching for such an agent for over 60 years.

The Pharmaceutical Journal

Two other early bioactive agents in the first case (the dolastatins) are now definitively from a prokaryotic cyanobacterium, even though they were first isolated from the sea slug Dolabella auricularia (see Figure 3). The second, the multiple ecteinascidins, date from studies in the Rinehart laboratory in the early 1980s onwards originally from the marine tunicate Ecteinascidia turbinata (compound 6; see Figure 2). In 2011, the Sherman group demonstrated that the actual producer in the tunicate was a microbe (Candidatus Endoecteinascidia frumentensis) symbiotic with the tunicate^​30​. In 2015, it was shown that the bacterium was, in fact, a “stripped” genomic version that in exchange for protection from predators, the tunicate provided essential nutrients to the microbe^​31​. The fact the bacterial genome was “stripped” indicates that the symbiosis between it and the host had been established a long time ago.

Supplying enough material for clinical trials and clinical use

Marine (micro)organisms are characterised by an impressive chemical diversity and a thorough analysis of a single marine invertebrate can afford tens, if not hundreds, of secondary metabolites that can be used to build a “natural combinatorial library” for drug development^​32​. Often these metabolites are rare owing to being contained at low levels in the living organisms, which are, in turn, available only in a few specimens because their collection is limited by environmental concerns. Progress in structure elucidation techniques now allows full NMR and MS spectral analysis with less than a 1 milligram  sample, but the preliminary in vivo pharmacological screening and further drug development need larger amounts of test compounds^​13​.

The scientific literature is full of cases of marine metabolites that have shown potent and promising bioactivities during preliminary screening, but whose limited availability has hampered further development. Progress in aquaculture and chemical synthesis or semi-synthesis have provided valid alternatives to solve the supply bottleneck. Recent advances in molecular genetics allow the identification of biosynthetic genes in producing organisms and their cloning into microbes suitable for large-scale fermentation. This approach is promising, but has only proved successful at the research level and not yet at the industrial scale for marketable marine drugs. Developments in this area will be covered in more detail in a later section^​1​.

The following examples of aquaculture and chemical synthesis or semi-synthesis illustrate how the discovery of potential leads from diverse marine sources has been translated into drugs that have entered the market, by allowing large-scale production. These successful examples ensure that pharmaceutical companies now regard marine natural product-derived pharmaceuticals with renewed interest. 

Aquaculture

Many pharmaceutically interesting marine macro- or microorganisms are difficult to culture under artificial conditions^​33​. A detailed understanding of living conditions in the natural environment is necessary to develop the optimal cultivation methods and, in the case of microorganisms, co-cultivation of two or more different strains, also called mixed fermentation, has provided a significant enhancement in the production of constitutive compounds or accumulate cryptic ones^​34​.

In the case of macroorganisms, aquaculture, namely the controlled cultivation of aquatic organisms in the sea or in artificial environments, has proved to be a valid option for large-scale production in several cases. One of the best examples of successful aquaculture supply is that of bryostatin 1 (compound 8), a macrocyclic polyketide lactone from the bryozoan Bugula neritina, that acts as a potent modulator of protein kinase C and has been in clinical trials as an anti-cancer agent, an anti-AIDS/HIV agent and for treatment of patients with Alzheimer’s disease^​35,36​. The initial preclinical and clinical tests on bryostatin 1 required gram quantities of the compound obtained by collecting 13 metric tonnes of wet Bugula neritina^​35​. This strategy could not sustain further experiments, and possible clinical use of the compound, so scientists resorted to aquaculture. The producing bryozoan was successfully grown using both in-sea and on-land aquaculture and the levels of bryostatin were high enough to obtain the needed supply of the compound^​37​. The producing bryozoan was successfully grown using a combination of in-sea and on-land aquaculture that involved the settlement of “tadpole” larvae on artificial materials, followed by a pre-grow period in tanks, and then grow-out in the ocean on anchored PVC pipe structures. The levels of bryostatin were high enough to obtain the needed supply of the compound^​37​.

Chemical synthesis or semi-synthesis

Modern synthetic chemists can build extremely complex molecules with clinical potential that are identical to or based on the NP structure, even in the case of stereochemically dense molecules. For example, halichondrin B (compound 9) is a complex polycyclic polyketide, first reported from the sponge Halichondria okadai, which was collected in Japan, that acts as antimitotic by inhibiting microtubule dynamics, through binding to high-affinity sites at the positive ends of existing microtubules^​38​. Similarly to bryostatin 1 (compound 8), early preclinical investigations on halichondrin B (compound 9) were made possible only by a large campaign of sponge collection conducted after an environmental impact assessment. In parallel to some successful attempts of aquaculture of the sponge, Kishi at Harvard University and the Japanese company Eisai were involved in the total synthesis of the natural halichondrin B (compound 9) reported in 2019 and earlier in structural simplifications, with replacement of the macrocycle with a ketone group and truncation of the tail region of the molecule^​39,40​. It was discovered that the synthetically manipulated molecule, named eribulin (compound 10), was even more active in vivo and had a better therapeutic index than halichondrin B. Eribulin mesylate is an anticancer medication used to treat refractory metastatic breast cancer and liposarcoma^​41​.

Another way to conveniently apply advanced organic synthesis principles to the production of drugs is semisynthetic transformations on MNPs. In this case, easily available compounds are converted by chemical processes into the desired product. An example of this is ecteinascidin 743 (compound 6), a complex alkaloid isolated from the tunicate Ecteinascidia turbinata that interferes with transcription^​42​. Ecteinascidin’s preclinical and clinical trials to Phase II were carried out using bulk collection of the tunicate (around 100 tonnes of biomass). In this case, the solution toward large-scale production of the compound could not be total synthesis, as none of the published syntheses was amenable to this. The structural similarity of ecteinascidins with saframycins, known terrestrial microbial tetrahydroisoquinoline antitumour antibiotics, suggested the application of a semisynthetic strategy^​43​. As a result, all the current supply of trabectedin (ecteinascidin 743), for the treatment of advanced soft-tissue sarcoma and ovarian cancer, is obtained by semi-synthesis from the microbial product cyanosafracin B, produced by large-scale fermentation of Pseudomonas fluorescens^​44​.

Success in the clinic 

As previously described, the first marine-derived leads to drugs for human use were the arabinose-linked nucleosides. The second compound family to receive serious attention was a series of linear peptides isolated by the Pettit group from the sea hare Dolabella auricularia. To contextualise the scale necessary at that time, the first three collections in the Indian Ocean beginning in 1965 were at the 2–100kg level and resulted in the discovery of several cytotoxic agents following bioactivity-driven isolation against murine P388 leukaemia^{​45–47​}. Finally, in 1987, the Pettit group published the formal structure of dolastatin 10 (compound 7), following the collection of 1,600kg of material and subsequent isolation of 28.7mg of the product^​45​. Owing to the lack of isolable material, dolastatin 10 was made available by total synthesis, with a US patent issued in 1990^​46,47​.

Subsequently, the “true source” of the dolastatins was discovered by natural product chemists associated with the Moore group at the University of Hawaii. Initially, symplostatin 1 (compound 11) isolated from Symploca hydnoides (a cyanobacterium) from Guam was a close chemical analogue of dolastatin 10 (compound 7)^​48​. Three years later, Luesch et al. discovered dolastatin 10 from another collection of Symploca collected in Palau, later reclassifified as Caldora penicillata^​49,50​.

The isolation of dolastatin 10 in 10,000-times higher yield than from the sea hare, coupled with ecological evidence with Dolabella auricularia being a generalist herbivore, indicated that cyanobacteria are the producer of this compound class and its source in the sea hare. Following the discovery from cyanobacteria from the Pacific Ocean, the Luesch and Paul groups found these compounds in marine cyanobacteria from Florida and the Caribbean, suggesting their ubiquitous presence in reefs across the globe and providing a robust and reliable compound source by recognising ecological relationships and biosynthetic signatures, as these modified peptides possess “classical” cyanobacterial structural motifs^​51​. Recently, the Luesch group working with the Ding group at the University of Florida, published data on the biosynthetic gene cluster for dolastatin 10 from the producing cyanobacterium, thus effectively “completing the biosynthetic circle” for this compound^​52​.

Dolastatin 10 (compound 7) possesses extraordinary broad-spectrum activity in the picomolar range against various cancer cell types (e.g. melanoma, colorectal, ovarian), it targets the protein tubulin, which is critically involved in cell proliferation^​52​. Synthetic access to dolastatin 10 (compound 7) has provided an opportunity to synthesise strategic analogues, including monomethyl auristatins E and F (MMAE [compound 12] and MMAF [compound 13]), which possess altered permeability and diffusion properties^​53​. Although dolastatin 10 (compound 7) and derivatives went into clinical trials as potential anticancer drugs, none of these “simple single compounds” went beyond phase II clinical trials owing to their innate toxicities^​54​. 

The advent of the antibody drug conjugate (ADC) platform as an enabling technology has been a game changer for MNP development; it has provided the opportunity to use antibodies to guide the cytotoxin to the cancer cell, circumventing toxicity owing to systemic exposure (see Figure 4). The toxic payload is then released to target the cancer cells specifically^​54​. The finding that MMAE (compound 12) and MMAF (compound 13), which possess an N-monomethyl valine instead of the N,N-dimethyl valine in dolastatin 10 (compound 7) as the N-terminal residue, retained biological activity^​54​, opened up opportunities for conjugation at this residue via the amine. There are now at least 15 ADCs approved by the FDA and/or comparable governmental bodies outside of the United States, including five to seven that carry MMAE or MMAF as their cytotoxic payload^​2​. 

The Pharmaceutical Journal/Sources within image

The breakthrough was the development of brentuximab vedotin, which was first approved by the FDA in 2011 for Hodgkin’s lymphoma and anaplastic large-cell lymphoma^​55​. Brentuximab vedotin was approved by NICE for the treatment of CD30-positive Hodgkin lymphoma in 2017^​56​. Other MMAE- and MMAF-containing ADCs have been FDA approved, in 2019 polatuzumab vedotin for B-cell lymphoma and enfortumab vedotin for refractory bladder cancer^​57,58​. Polatuzumab vedotin was also approved by NICE in 2020 for the treatment of B-cell lymphoma^​59​. However, enfortumab vedotin does not have UK authorisation^​60​. In 2020 belantamab mafadotin was approved for relapsed/refractory multiple myeloma; this was withdrawn in November of 2022 owing to not meeting the required level of PFS^​54​. In 2021, tisotumab vedotin was approved for recurrent or metastatic cervical cancer^​54​. Tisotumab vedotinis is not yet approved by the MHRA, but a NICE technology appraisal is expected^​61​. Disitamab vedotin was approved by the Chinese Center for Drug Evaluation (CDE) of the National Medical Products Administration (NMPA) for HER2-positive late-stage breast cancer patients in 2021^​54​. With the proven success of ADCs, the “flood gate” is now wide open, with more than 100 ADCs containing dolastatin 10-based payload undergoing preclinical and clinical development for many different cancers; the favourable toxicity profiles directly correlated with clinical impact. This example highlights that one potent marine compound can provide the basis for more cancer drugs with tremendous clinical impact, benefitting a large patient population and saving thousands of lives^​54​.

The Pharmaceutical Journal

There have been other examples of success before the vedotins, including for non-oncology indications. Ziconotide was approved in 2006 for chronic pain treatment and as a primary alternative to morphine, displaying up to 1,000 times greater analgesic potency as a calcium ion channel blocker^​62​. However, its requirement for a complex method of delivery to the spine has limited its utility. Approved marine hypertriglyceridemia drugs include omega-3 fatty acid and esters (2001: omega-3-acid ethyl esters, 2012: 2014: omega-3-carboxylic acids), and ethyl eicosapentaenoic acid (2012), all derived from marine sources, such as fish or krill^{​63–65​}. 

The biggest impact of marine drugs has been seen in the oncology space, where the investments were highest. Trabectedin/ET-743 was first approved in Europe in 2007 for the treatment of advanced soft-tissue sarcoma and has a unique mechanism of action based on interaction with the minor groove of DNA, bending the helix to the major groove^​66​. This binding to DNA triggers a cascade of events affecting several transcription factors, DNA binding proteins, and DNA repair pathways, resulting in perturbation of the cell cycle. Trabectedin was approved by NICE for the same indication in 2009^​67​. Eribulin mesylate, targeting tubulin in a manner distinct from that of dolastatin 10, was approved by the FDA in 2010 and NICE in 2011, as a third-line treatment for metastatic breast cancer^​68​. Linking eribulin to antibodies is emerging as a successful strategy to improve the pharmacological and toxicity profile^​69​. Plitidepsin (Aplidin) (compound 14) discovered from the sea squirt Aplidium albicans was approved in Australia in late 2018 for the treatment of relapsed and refractory multiple myeloma patients in combination with dexamethasone (see Figure 2)^​70​. By targeting eukaryotic elongation factor 1A (eEF1A), it also possesses antiviral activity and underwent clinical trials as a host-directed antiviral for the treatment of SARS-CoV-2 infections^​71​.

Feeding the marine pharmaceutical pipeline 

As most marine-derived compounds in the clinic or under clinical development target cancer, it is important to highlight that ongoing global MNP preclinical pharmacological research encompasses diverse clinically relevant pharmacological activities. A series of comprehensive and systematic reviews from 1998 to 2021 (with the latest of these reviews covering 2019 to 2021 is in preparation with details below and in Table 1) offer insight into numerous MNP that have demonstrated other activities beyond cancer^{​72–80​}. A group at Midwestern University curated a website that covers marine sources and their potential, which can be found here^​2​. A novel approach that uses warheads based on auristatin is also being developed where in place of a standard mAb targeting moiety, several novel, synthetic, structurally constrained bicyclic peptides (“Bicycle”) have been created that target EphA2 (BT5528) and Nectin-4 (BT8009)^​53,81​. Unlike the limitations of current mAb-containing ADCs, bicycle toxin conjugates (BTCs) have been designed to have short plasma half-lives where effects are balanced by rapid and efficient tumour penetration, coupled with high C_max levels allowing these BTCs to release the cell penetrant payload (MMAE) killing the target cell and producing efficient bystander killing^{​53,81,82​}. Since this novel warhead/targeting modality is in early clinical development, there are still substantial hurdles to overcome, including human pharmokinetics and pharmacodynamics, the unknown potential adverse events that may be associated with this novel targeting modality, and the eventual proof of efficacy in humans^​82​.

Where the next generation of MNP pharmaceuticals will come from 

Developments in isolation procedures, including rapid high-content screening methods to strengthen current bioassays, and more precise structural elucidation methods along with a reduction in wasted effort via the use of bioinformatic tools and ‘omics’ approaches will likely yield the next generation of MNP pharmaceuticals^​83​. Future drug candidates with improved properties will be developed by solving sustainable supply issues, discovering their bioactivity, mode of action (MoA), and optimising their pharmacological properties by the generation of analogues^​84​. 

Exploration of new natural sources such as those coming from extreme or deep environments or bringing uncultivable microorganisms into culture will generate new MNPs as future drug candidates. This will require development in engineering to sample at great ocean depths and an improved understanding of microbial physiology to enable the unculturable to be cultured^​85,86​. 

As less than 1% of microbes can be cultivated, partly owing to the fact that conventional methods may enrich the growth of specific microorganisms while hindering the cultivation of others, extensive efforts have been made to solve this problem. New innovative culturing methods have been developed in the last few years^​87,88​. One example is the isolation chip (iChip) method, where microbes are localised to small wells at the single-cell level before being inserted into in situ environments. This has been shown to be a promise-culturing strategy to cultivate bacterial species that are not able to be cultured as they do not grow in laboratory conditions^​89​. An improvement of this method has been recently published where the iChip and petri dish plating methods to cultivate bacteria from soil contaminated with or without crude oil were used. iChip improved the richness and evenness of bacterial isolates in both the natural and contaminated soil samples^​90​. These systems work well on land, though would be restricted to perhaps mud flats or equivalent with marine-derived microbes. This is because it is a terrestrial technology that needs to be buried in the soil or sediment.

The bioassay-guided isolation approach to obtain pure bioactive compounds from active marine organism extracts is still the most successful approach^​91​ (see above and Figure 1). However, the major drawback of this strategy is the time spent on repeated isolation of known structures. The development of rapid dereplication processes aimed at the early identification of known compounds is a priority issue in MNP drug discovery. Subsequently, several dereplication approaches have been developed^​92​. The most used strategy uses hyphenated analytical methodologies, which combine chromatographic separation techniques, predominantly LC, and spectrometric data, mainly MS in the high resolution (HRMS) mode and/or with ion fragmentation (MS/MS) along with ultraviolet (UV) and/or nuclear magnetic resonance (NMR) spectroscopic techniques. The obtained data are analysed using databases of natural products (e.g. MarinLit for MNP, Natural Product Atlas for microbial NP, CyanoMetDB for cyanobacteria NP) and in the case of LC/MS can be analysed using molecular networking (MN)^​93​ (see Figure 5). This computational strategy was introduced in 2012 and organises the mass spectral data based on the similarity between precursor ions’ mass fragmentations allowing the organisation of extensive tandem mass data sets and the mapping of the chemical diversity in the sample. In this way, it is possible to discover molecular families and substructures in MS/MS data. The combination of molecular networking with Global Natural Products Social Networking (GNPS), an open platform that compiles many metabolomic datasets is an effective approach for early identification of NP and annotation of their analogues without the need for isotopic labelling^​93,94​. 

More recently, feature-based molecular networking (FBMN) was created, showing more representative results by pre-processing the MS raw data with different computational tools^​95​. The software detects the features of the precursor ions and filters the data by isotope patterns, retention time, or even ion mobility separation. These tools generate feature tables that can be submitted to GNPS obtaining much cleaner and more sensible results. GNPS includes more than 80,000 MS/MS spectra and is the most powerful informatics tool in NP dereplication and the most utilised among the NP research community^​94​. The reference dataset is not complete but is improving continuously via the support of the natural products community (see Figure 5). 

NMR based dereplication methods such as MixONat algorithm in Python for ¹³C NMR and computer-aided ¹³C NMR, represent a good alternative to the MS dereplication^​96​. Although methods for dereplication using MS are more developed than for NMR, further research on NMR methods, especially when integrated with MS/MS methods, would increase dereplication efficiency and accuracy.

The Pharmaceutical Journal/Shutterstock.com

High throughput screening (HTS) continues to be a key strategy for identifying chemical compounds capable of modulating specific disease-related targets. New assays are constantly being developed to support drug discovery efforts and miniaturisation of assays allows biological screening to be performed with very small amounts of the pure NP or extracts. For example, a HTS was reported to identify compounds that inhibit biofilm formation using a Salmonella strain from 1,120 extracts^​97​.

The reduction in the quantity needed for structural elucidation of complex natural products (i.e. the use microcryoprobe NMR) means only nanomoles of material are needed to solve a complete structure^​98​.

Metabolomics can be defined as the qualitative and/or quantitative analysis of the whole set of low molecular weight metabolites present in a biological sample. Targeted metabolomics is used to analyse a set of defined metabolites and is commonly used to study human metabolism, whereas ‘untargeted metabolomics’ detects as many metabolites as possible, often without a complete identification of each, but allows clustering into groups^​99​.

The application of ‘omics’ approaches, with the decrease in costs of DNA sequencing, the availability of MS (metabolomics, lipidomics, and proteomics) and the development of bioinformatics tools, have improved the analysis of biological systems and increased the possibilities of discovering new bioactive MNPs^​83​. 

MNPs are generated by biosynthetic enzymes that construct core scaffolds, introduce pharmacophores and create metabolic diversity through a complex series of enzymatic reactions and regulatory switches. These enzymes are encoded by a group of genes in close genomic proximity that are organised into biosynthetic gene clusters (BGC)^​100​. As the structures of these enzymes are highly conserved, the BGCs can be used to computationally predict potential metabolites by genome mining (GM) which comprises computational methods for the automatic detection and annotation of BGCs from genomic data (see Figure 6). This approach allows for the prediction of the structure of the NP based on genetic information before its isolation and the structural elucidation along with their possible functional and chemical interactions. The low-cost and rapid development of genome sequencing methods with the development of bioinformatic analysis tools for BGC identification, and genome mining (GM) has become a key technology to exploit and explore NP diversity and is revolutionising the search for new MNPs^​101​. Powerful computer analysis of genomes is allowing the identification of millions of putative biosynthetic pathways for novel molecules of possible avenues for creating entirely new drugs^​102​. A plethora of new platforms and databases were developed to computationally mine genetic data such as the open-source platforms Antibiotics and Secondary Metabolite Analysis Shell (AntiSMASH), PRISM, BAGEL, TOUCAN, and others^​103​. Once the BGC information is obtained, the next step is the prediction of NP structures. DeepRiPP, RiPPMiner, and RODEO (Rapid ORF Description and Evaluation Online) are some tools that allow the prediction of NP structures (see Figure 6)^​83​.

The Pharmaceutical Journal/PharmaMar

Another interesting tool is metagenomics, a process of extracting microbial genomes directly from environmental samples, regardless of sample type or microbial abundance. Metagenomics is a good alternative to overcome some of the difficulties in NP discovery by profiling microbial communities, such as the study of uncultivated species^​104​. The metagenomic sequence data assembly of the endosymbiotic genus Entotheonella isolated from the marine sponge Theonella swinhoei allowed the identification of their ‘metabolic talent’ and led to the discovery of many unprecedented BGCs^​27​.

Genetic engineering of different biosynthetic pathways requires different levels of engineering to manipulate them. The RiPP metabolites are the most straightforward and the manipulation of the BGCs or modifying enzymes can be achieved to generate new products either in host cells or via the application of modified enzymes on synthetic substrates^​105​. Other BGCs such as non-ribosomal peptide synthetase (NRPS) or polyketide synthase (PKS) require more involved manipulation to generate new analogues of known compounds, although the use of mutations to allow non-natural substructures to be incorporated is very successful^​106​. Future development in understanding of BGCs and genetic engineering will likely allow at first the generation of new analogues with improved properties and subsequently the generation of new chemical diversity to be tested against human disease. 

The application of machine learning (ML) methods, defined as the study and application of algorithms that perform classification and prediction tasks through pattern recognition, to natural product research is accelerating the identification of new molecular entities for drug discovery. Prihoda et al. conducted a comprehensive review on ML methods in NP discovery with applications in genome mining, predictions of structures, and types of activities expected from those molecules^​107​. More recently, a biosynthetic-based ML bioinformatic method to predict natural product biological activity from BGC sequences was published in 2021^​108​. In this case, machine learning algorithms have been used to predict biological activity from biosynthetic gene clusters, enabling the identification of potential drug candidates from natural products. By analysing gene clusters, these methods help in understanding the biological activity of natural compounds and their potential as drug leads. Furthermore, a summary of the recent advancements in ML-assisted MS and NMR data analysis to establish the chemical structures of NPs was recently reported^​109​. One example of machine learning-assisted structure annotation of natural products based on mass spectrometry (MS) and nuclear magnetic resonance (NMR) data is the use of tools like SMART 2.0, SMART-Miner, and COLMAR^{​110–112​}. These tools utilise machine learning algorithms, such as convolutional neural networks and Bayesian approaches, to analyse complex NMR spectra and identify and annotate primary metabolites in natural product mixtures^{​110–112​}. By leveraging AI techniques, researchers can enhance the efficiency and accuracy of structure elucidation and annotation of natural products, ultimately aiding in the discovery and characterisation of novel compounds from complex biological samples.

The output from GNPS can also be imported into other analysis programs such as Cytoscape, Metaboanalyst, QIIME 2 which offer ML tools. XGBoost classifier is an 1D ¹³C NMR data and machine learning method that predicts the chemical class of a NP with higher accuracy, outperforming other algorithms of the same type^​113​. Computational methods to predict the MS of small molecules can be achieved with the use of machine learning (ISIS, CSI:FingerID) and molecular properties of unknown molecules are predicted by comparison with fragmentation trees using machine learning techniques^​83​.

Artificial Intelligence (AI) is becoming a powerful tool for drug discovery, allowing scientists to predict how a natural product’s chemical structure affects its activity and which macromolecule might be its biological target. The applications of AI in natural product research, including the issues mentioned with ML, the pitfalls in training algorithms and approaches to avoid them were recently reviewed^​114​. An example where AI was applied to drug discovery is the use of AlphaFold for predicting the 3D structure of proteins. AlphaFold is a deep learning system developed by DeepMind that can predict the 3D structure of proteins from their primary amino acid sequence by learning from a vast amount of protein structure data in the Protein Data Bank^​115​. Another example is the deep learning techniques that have been applied to predict drug-target binding affinity, de novo drug-like molecule generation, property prediction, and protein pocket-conditioned molecular representations. These approaches leverage deep learning models to analyse molecular structures and properties, enhancing drug discovery processes^​115​.

Application of these approaches to data derived from complex microbiomes, predicting structures from genomic signatures, and the types of activity we might expect from those molecules, enables us to better understand their “genomic language”^​102​. For example, dereplication using integrated genome mining, MS/MS analysis along with in silico GNPS tools of the nearly complete genome of Streptomyces tendae VITAKN, which was isolated from the southern coast of India, yielded the discovery of columbamides, a new class of di- and trichlorinated acyl amides with cannabino-mimetic activity^​116​. New peptides as members of the viequeamide class were discovered from a marine cyanobacterium by application of the Small Molecule Accurate Recognition Technology (SMART), which employs a convolutional neural network to classify HSQC spectra of organic molecules using pattern recognition principles^​117​.

Future perspectives and conclusions 

NPs are specialised metabolites, whilst most conform to small molecule drug parameters, many are larger and more complex than these ‘standard’ small molecules synthesised and screened by pharmaceutical companies against druggable targets. They represent privileged scaffolds as they have evolved to interact with biological targets. While MNPs have been successful in targeting validated disease-relevant proteins, one strength of MNPs includes the potential for discovering new mechanisms of action through phenotypic screening and subsequent validation of proteins that have not been successfully drugged previously or even considered to be targeted. MNPs are genetically encoded, providing opportunities for genomics-driven discovery approaches and synthetic biology to identify and optimise bioactive natural products^​50​.

In that sense, MNPs will remain pioneering, lead to first-in-class pharmacological agents, and provide new starting points for the development of new pharmaceuticals. As demonstrated by dolastatin 10, while the target was not novel, the compound can be considered most-potent-in-class. Potency is a common theme among MNPs, which is one reason for their suitability as ADC payloads. We expect that the relevance of MNPs to the ADC field will further increase and that the ADC field will be dependent on MNPs to provide ultra-potent molecules that may be conjugated to cell- and target-seeking antibodies and other targeting agents such as the “bicyclic peptides”^​53,81​.

Because of their larger size and complexity, MNPs occupy different chemical space than other compounds, which collectively allows for unusual interactions with protein complexes and for targeting protein-protein interactions, which are emerging as a new “target” class. Druggable space is evolving and previously “undruggable” targets or protein interfaces will increasingly be targeted by MNPs, therefore extending the druggable genome^​118​. This can be extrapolated further to RNA-protein interactions, as shown for pateamine A by targeting eukaryotic initiation factor (eIF) 4A^​119​. Therefore, it is conceivable that MNPs will target various disease-associated RNAs, which have been emerging as therapeutic targets and will lead to new classes of small-molecule pharmaceuticals, with RNA playing a major role beyond vaccines^​120​. 

In addition to acting at the level of nucleic acids (DNA, RNA), MNPs interact with membrane components or metabolites, such as glycans or lipids, and we will likely witness new target interactions to be deciphered beyond those in the standard repertoire. The field of marine natural products will continue to be characterised by high-level innovation through synergy with enabling technologies, such as ADC and cryo-electron microscopy.

The translation of NPs into drugs requires identification of the molecular targets, mechanism of action and even synthetic lethal vulnerabilities or pleiotropic effects where one gene influences two or more seemingly unrelated phenotypic characteristics of organisms. This is demonstrated by didemnin B, which inhibits protein elongation at the level of eukaryotic initiation factor eEF1A. It was shown to have potentially beneficial polypharmacology that can serve as a predictive pharmacogenomics response marker^​121​. Identifying the most suitable therapeutic setting will be important for success. We expect to increasingly see repurposing of active compounds with the advance of new technologies and rigorous pharmacological assessment. New sensitive technologies to interrogate bioactivities at the molecular and cellular level will continue to be developed, increasing the value of these compounds with respect to their biomedical utility^​122​. 

The identification of mechanisms of action early in the discovery process will be key in accelerating their translation. A new target or mechanism will prompt the community to develop synthetic methods to solve the supply problem, which remain the number one hurdle for MNP drug development. In parallel, genomics-driven discovery will provide us with currently cryptic chemical diversity, and even assist with target and mechanism discovery. Synthetic biology methods will slowly but surely improve for the heterologous production of MNP, addressing the supply issue through biotechnological means. AI is expected to contribute and synergise with the field of MNP at all these levels, from metagenomics to structure and target identification, as well as production.

“Special” MNPs have intrinsic functions engineered by nature and evolution. They will continue to pave the way to new targets and be superior to unnatural approaches. As we coexist with nature, we learn more about the best approaches to deal with problems on this planet, including health solutions. We expect that marine sources will provide a rich supply of promising first-in-class therapeutic agents, assuming we can preserve the life-sustaining oceans and biological treasures contained in them, with new structural and biological characteristics^​123​.