Discovery and resupply of pharmacologically active plant-derived natural products: A review

Abstract

Medicinal plants have historically proven their value as a source of molecules with therapeutic potential, and nowadays still represent an important pool for the identification of novel drug leads. In the past decades, pharmaceutical industry focused mainly on libraries of synthetic compounds as drug discovery source. They are comparably easy to produce and resupply, and demonstrate good compatibility with established high throughput screening (HTS) platforms. However, at the same time there has been a declining trend in the number of new drugs reaching the market, raising renewed scientific interest in drug discovery from natural sources, despite of its known challenges. In this survey, a brief outline of historical development is provided together with a comprehensive overview of used approaches and recent developments relevant to plant-derived natural product drug discovery. Associated challenges and major strengths of natural product-based drug discovery are critically discussed. A snapshot of the advanced plant-derived natural products that are currently in actively recruiting clinical trials is also presented. Importantly, the transition of a natural compound from a "screening hit" through a "drug lead" to a "marketed drug" is associated with increasingly challenging demands for compound amount, which often cannot be met by re-isolation from the respective plant sources. In this regard, existing alternatives for resupply are also discussed, including different biotechnology approaches and total organic synthesis. While the intrinsic complexity of natural product-based drug discovery necessitates highly integrated interdisciplinary approaches, the reviewed scientific developments, recent technological advances, and research trends clearly indicate that natural products will be among the most important sources of new drugs also in the future.
Copyright © 2015. Published by Elsevier Inc.

Full text

Research review paper
Discovery and resupply of pharmacologically active plant-derived
natural products: A review
Atanas G. Atanasov
a,
⁎
,1
, Birgit Waltenberger
b,
⁎
,1
, Eva-Maria Pferschy-Wenzig
c,
⁎
,1
,ThomasLinder
d
,
Christoph Wawrosch
a
, Pavel Uhrin
e
,VeronikaTemml
f
, Limei Wang
a
, Stefan Schwaiger
b
,ElkeH.Heiss
a
,
Judith M. Rollinger
a,b
, Daniela Schuster
f
, Johannes M. Breuss
e
, Valery Bochkov
g
, Marko D. Mihovilovic
d
,
Brigitte Kopp
a
, Rudolf Bauer
c
,VerenaM.Dirsch
a
, Hermann Stuppner
b
a
Department of Pharmacognosy, University of Vienna, Althanstrasse 14, 1090 Vienna, Austria
b
Institute of Pharmacy/Pharmacognosy and Center for Molecular Biosciences Innsbruck (CMBI), University of Innsbruck, Innrain 80-82, 6020 Innsbruck, Austria
c
Institute of Pharmaceutical Sciences, Department of Pharmacognosy, University of Graz, Universitätsplatz 4/I, 8010 Graz, Austria
d
Institute of Applied Synthetic Chemistry, Vienna University of Technology, Getreidemarkt 9/163-OC, 1060 Vienna, Austria
e
Institute of Vascular Biology and Thrombosis Research, Center of Physiology and Pharmacology, Medical University of Vienna, 1090 Vienna, Austria
f
Institute of Pharmacy/Pharmaceutical Chemistry and Center for Molecular Biosciences Innsbruck (CMBI), University of Innsbruck, Innrain 80-82, 6020 Innsbruck, Aus tria
g
Institute of Pharmaceutical Sciences, Department of Pharmaceutical Chemistry, University of Graz, Humboldtstrasse 46/III, 8010 Graz, Austria
abstractarticle info
Article history:
Received 16 January 2015
Received in revised form 16 July 2015
Accepted 7 August 2015
Available online 15 August 2015
Keywords:
Natural products
Plants
Drug discovery
Phytochemistry
Pharmacology
Medicine
Ethnopharmacology
Computer modeling
Organic synthesis
Plant biotechnology
Medicinal plants have historically proven their value as a source of molecules with therapeutic potential, and now-
adays still represent an importa nt pool for the identification of novel drug leads. In the past decades, pharmaceutical
industry focused mainly on libraries of synthetic compounds as drug discovery source. They are comparably easy to
produce and resupply, and demonstrate good compatibility with established high throughput screening (HTS) plat-
forms.However,atthesametimetherehasbeenadeclining trend in the number of new drugs reaching the market,
raising renewed scientific interest in drug discovery from natural sources, despite of its known challenges. In this
survey, a brief outline of historical development is provided together with a comprehensive overview of used ap-
proaches and recent developments relevant to plant-derived natural product drug discovery. Associated challenges
and major strengths of natural product-based drug discovery are critically discussed. A snapshot of the advanced
plant-derived natural products that are currently in actively recruiting clinical trials is also presented. Importantly,
the transition of a natural compound from a “screening hit”through a “drug lead”to a “marketed drug”is associated
with increasingly challenging demands for compound amount, which often cannot be met by re-isolation from the
respective plant sources. In this regard, existing alternatives for resupply are also discussed, including different bio-
technology approaches and total organic synthesis.
While the intrinsic complexity of natural product-based drug discovery necessitates highly integrated interdisci-
plinary approaches, the reviewed scientific developments, recent technological advances, and research trends
clearly indicate that natural products will be among the most important sources of new drugs alsoin the future.
© 2015 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY license
(http://creativecommons.org/licenses/by/4.0/).
Biotechnology Advances 33 (2015) 1582–1614
Abbreviations: 4CL, 4-coumaroyl CoA ligase; ADME/T, absorption, distribution, metabolism, excretion (and toxicity); BIA, benzylisoquinoline alkaloid; C4H, cinnamate 4-hydroxylase;
CoA, coenzyme A; CRISPR/Cas9, clustered regulatory interspaced short palindromic repeat/CRISPR associated protein 9; DMPP, dimethylallyl-pyrophosphate; DNP, Dictionary of Natural
Products; DNTI, Drugs from Nature Targeting Inflammation; EMA, European Medicines Agency; FDA, US Food and Drug Administration; GGPP, geranylgeranyl diphosphate; GPCR, G-
protein coupled receptor; GC, gas chromatography; HPLC, high performance liquid chromatography; HTS, high-throughput screening; iNOS, inducible nitric oxide synthase; IPP,
isopentenyl-pyrophosphate; MEP, 2C-methyl-D-erythriol-4-phosphate; MS, mass spectrometry; NCI, (US) National Cancer Institute; NCT, National Clinical Trial; NME, new molecular entity;
NMR, nuclear magnetic resonance; NPD, Natural Product Database; OPLS-DA, orthogonal projection to latent structures discriminant analysis; PAL, phenylalanine ammonia lyase; PLS-DA,par-
tial least square regression modeling discriminant analysis; PPAR, peroxisome proliferator-activated receptor; QSAR, quantitative structure activity relationship; TAL, tyrosine ammonia lyase;
TALEN, transcription activator-like effector nuclease; TCM, traditional Chinese medicine; THC, tetrahydrocannabinol; SHE, safety, health, and environment.
⁎Corresponding authors.
E-mail addresses: atanas.atanasov@univie.ac.at (A.G. Atanasov), birgit.waltenberger@uibk.ac.at (B. Waltenberger), eva-maria.wenzig@uni-graz.at (E.-M. Pferschy-Wenzig).
1
These authors contributed equally to this work.
http://dx.doi.org/10.1016/j.biotechadv.2015.08.001
0734-9750/© 2015 The Authors. Published by Elsevier Inc. This is an open access article under the CCBY license (http://creativecommons.org/licenses/by/4.0/).
Contents lists available at ScienceDirect
Biotechnology Advances
journal homepage: www.elsevier.com/locate/biotechadv

Contents
1. Introduction.............................................................. 1583
2. Plant-derived drug discovery: history, challenges, and significance ..................................... 1583
2.1. Naturalproductsasdrugcandidates:ahistoricalperspective .................................... 1583
2.2. Challengescontributingtothedeclineofplant-derivednaturalproductsasdrugdiscoverysource.................... 1584
2.3. Renewaloftheinterestinnaturalproduct-baseddrugdiscovery................................... 1586
3. Approachesforthediscoveryofnewpharmacologicallyactiveplantcompounds............................... 1586
3.1. Approachestoselectingstartingmaterial ............................................. 1586
3.2. Considerationsregardingthechoiceofbioassays.......................................... 1592
3.3. Approaches for the identificationofactiveplantconstituents .................................... 1595
4. Approachesforresupplyofpharmacologicallyactiveplantcompounds ................................... 1597
4.1. Plantcellandtissueculture ................................................... 1597
4.2. Heterologousproduction..................................................... 1599
4.3. Organicsynthesis ........................................................ 1602
5. Concludingremarks .......................................................... 1605
Acknowledgments ............................................................. 1606
References................................................................. 1607
1. Introduction
For millennia, medicinal plants have been a valuable source of ther-
apeutic agents, and still many of today's drugs are plant-derived natural
productsor their derivatives (Kinghorn et al., 2011;Newman and Cragg,
2012). However, since natural product-based drug discovery is associat-
ed with some intrinsic difficulties (discussed in more details in
Section 2.1), pharmaceutical industry has shifted its main focus toward
synthetic compound libraries and HTS for discovery of new drug leads
(Beutler, 2009; David et al., 2015). The obtained results, however, did
not meet the expectations as evident in a declining number of new
drugs reaching the market (David et al., 2015; Kingston, 2011;
Scannell et al., 2012). This circumstance revitalized the interest in
natural product-based drug discovery, despite its high complexity,
which in turn necessitates broad interdisciplinary research approaches
(Heinrich, 2010a). In line with this, the Austrian “Drugs from Nature
Targeting Inflammation (DNTI)”consortium was formed in 2008 unit-
ing scientists with expertise in multiple disciplines relevant for natural
product-based drug discovery. The DNTI program aimed at identifying
and characterizing natural products with anti-inflammatory activity
by the combined and synergistic use of computational techniques,
ethnopharmacological knowledge, phytochemical analysis and
isolation, organic synthesis, plant biotechnology, and a broad range of
in vitro, cell-based, and in vivo bioactivity models [e.g., (Atanasov et al.,
2013b; Fakhrudin et al., 2014; Schwaiberger et al., 2010); further details
for the DNTI consortium are available at http://www.uibk.ac.at/
pharmazie/pharmakognosie/dnti/]. Using their multidisciplinary
expertise and the gathered experience from the DNTI participation,
the authors of this review want to summarize here the currently
established strategies and recent developments in the discovery and
resupply of plant-derived bioactive natural products.
2. Plant-derived drug discovery: history, challenges, and significance
2.1. Natural products as drug candidates: a historical perspective
The first written records on medicinal applications of plants date
back to 2600 BC and report the existence of a sophisticated medicinal
system in Mesopotamia, comprising about 1000 plant-derived
medicines. Egyptian medicine dates back to about 2900 BC, but its most
useful preserved record is the “Ebers Papyrus”from about 1550 BC,
containing more than 700 drugs, mainly of plant origin (Borchardt,
2002; Cragg and Newman, 2013; Sneader, 2005). Traditional Chinese
medicine (TCM) has been extensively documented over thousands of
years (Unschuld, 1986), and the documentation of the Indian Ayurveda
system dates back to the 1st millennium BC (Patwardhan, 2005).
The knowledge on the medicinal application of plants in the Western
world is mainly based on the Greek and Roman culture. Of particular im-
portance are the compendia written by the Greek physician Dioscorides
(1st century AD), and by the Romans Pliny the Elder (1st century AD)
and Galen (2nd century AD) (Sneader, 2005). The Arabs preserved a
large amount of the Greco-Roman knowledge during the Dark and Mid-
dle ages (i.e., 5th to 12th centuries), and complemented it with their
own medicinal expertise, and with herbs from Chinese and Indian tradi-
tional medicines (Cragg and Newman, 2013). The invention of letterpress
by Johannes Gutenberg led to a resurrection of the Greco-Roman knowl-
edge in the 15th and 16th century, and to the compilation of several very
influential herbal books that were widely distributed in Europe, like The
Mainz Herbal (Herbarius Moguntinus, 1484) and The German Herbal
(1485), both edited by Gutenberg's partner Peter Schöffer, the Herbarium
Vivae Eicones (Otto Brunfels; 1530), the Kreütter Buch by Hieronymus
Bock (1546) that was written in German, and De Historia Stirpium by
Leonhart Fuchs that was published in Latin in 1542 and also in German
in the following year (Sneader, 2005).
During all that time, medicinal plants were only applied on an em-
pirical basis, without mechanistic knowledge on their pharmacological
activities or active constituents. It was only in the 18th century that
Anton von Störck, who investigated poisonous herbs such as aconite
and colchicum, and William Withering, who studied foxglove for the
treatment of edema, laid the basis for the rational clinical investigation
of medicinal herbs (Sneader, 2005).
Rational drug discovery from plants started at the beginning of the
19th century, when the German apothecary assistant Friedrich Sertürner
succeeded in isolating the analgesic and sleep-inducing agent from opium
which he named morphium (morphine) after the Greek god of dreams,
Morpheus. He published a comprehensive paper on its isolation, crystalli-
zation, crystal structure, and pharmacological properties, which he stud-
ied first in stray dogs and then in self-experiments (Sertürner, 1817).
This triggered the examination of other medicinal herbs, and during the
following decades of the 19th century, many bioactive natural products,
primarily alkaloids (e.g., quinine, caffeine, nicotine, codeine, atropine, col-
chicine, cocaine, capsaicin) could be isolated from their natural sources
(Corson and Crews, 2007; Felter and Lloyd, 1898; Hosztafi, 1997; Kaiser,
2008; Kruse, 2007; Sneader, 2005; Zenk and Juenger, 2007). Apothecaries
who specialized in the purification of these compounds were the progen-
itors of pharmaceutical companies. The first one was H.E. Merck in Darm-
stadt (Germany) who started extracting morphine and other alkaloids in
1826 (Kaiser, 2008). Subsequently, efforts were undertaken to produce
natural products by chemical synthesis in order to facilitate production
at higher quality and lower costs. Salicylic acid was the first natural com-
pound produced by chemical synthesis in 1853 (Kaiser, 2008).
After the discovery of penicillin (1928), an era of drug discovery
from microbial sources was initiated in the 1930s, that laid the scientific
1583A.G. Atanasov et al. / Biotechnology Advances 33 (2015) 1582–1614

and financial foundation of the modern pharmaceutical industry after
World War II. At that time, the therapeutic use of extracts and partly
purified natural products was increasingly replaced by the use of pure
compounds (Beutler, 2009; David et al., 2015). Despite the advent of
combinatorial chemistry and HTS campaigns during the last decades,
the impact of natural products for drug discovery is still very high. Of
the 1073 new chemical entities belonging to the group of small mole-
cules that had been approved between 1981 and 2010, only 36% were
purely synthetic, while more than the half were derived or inspired
from nature (Newman and Cragg, 2012). A substantial number of
these compounds have been discovered in higher plants (Kinghorn
et al., 2011). Particularly prominent examples of plant-derived natural
compounds that havebecome indispensable for modern pharmacother-
apy can be found in the field of anti-cancer agents, e.g., paclitaxel and its
derivatives from yew (Taxus) species, vincristine and vinblastine from
Madagascar periwinkle (Catharanthus roseus (L.) G. Don), and
camptothecin and its analogs initially discovered in the Chinese tree
Camptotheca acuminata Decne. (Cragg and Newman, 2013; Kinghorn
et al., 2011). Further notable examples include the cholinesterase inhibi-
tor galanthamine that has been approved for the treatment of Alzheimer's
disease and was initially discovered in Galanthus nivalis L. (Mashkovsky
and Kruglikova-Lvova, 1951), and the important antimalarial and poten-
tial anti-cancer agent artemisinin originally derived from the traditional
Chinese herb Artemisia annua L. (Klayman et al., 1984).
2.2. Challenges contributing to the decline of plant-derived natural products
as drug discovery source
Since very often plants are collected directly from their natural habitat,
the correct identification and nomenclature are essential and the basis for
all following steps. For an unambiguous identification, a combination of
methods might be necessary, such as genetic and chemical analysis in ad-
dition to morphological and anatomical characterization (Bucar et al.,
2013). Continuously ongoing modifications in plant taxonomy as well
as synonymy issues add to the difficulty of this challenging task (David
et al., 2015). Moreover, collection of the plant material and accurate doc-
umentation, botanical identification, as well as preparation of the herbar-
ium vouchers are tasks that cannot be automated (David et al., 2015)and
need specialists who become increasingly rare (Bucar et al., 2013).
Important challenges related with the use of plants as a source for
identification of bioactive compounds are related with the accessibility
of the starting material. Often the available amount of natural products
is low. Although many plant-derived natural products have already
been isolated and characterized, available compound quantities are
often insufficient for testing for a wide range of biological activities.
While small amounts of plant material are usually required for an initial
pharmacological evaluation, much larger quantities are needed for
through characterization of the pharmacological activity of its constitu-
ents. Furthermore, limited availability becomes even more problematic
when a bioactive plant-derived natural product is identified to have a
very promising bioactivity and becomes a pharmaceutical lead. Recol-
lections ofwild species may turn difficult, since plant habitats can rapid-
ly disappear under anthropic pressure. Moreover, the habitat of plants,
particularly of protected species, needs to be respected whencollecting
from the wild (David et al., 2015), and season-dependent chemical
composition of plant material may limit the time window for recollec-
tion (e.g., blossom collection necessitates collection during the
flowering season). In cases of imported plant material, also an entire
array of additional factors might affect its accessibility, for example
local wars, natural catastrophes, or changing legal regulations for
cross-border traveling and export of plant material. The importance of
plant material accessibility is illustrated by a recent study (Amirkia
and Heinrich, 2014) investigating the correlation between species
abundance of alkaloids occurrence and their use as pharmaceutical
drugs. Species distribution was assessed on the basis of Global Biodiver-
sity Information Facility (GBIF) data. The authors found that 93% of all
alkaloids in medical use have more than 50 occurrences in the GBIFda-
tabase, and only two have less than 10 occurrences. Therefore, the au-
thors conclude that natural products occurring in many different
species are more favorable for medicinal use, and that supply con-
straints are a major obstacle to the successful research, development
and commercialization of natural products.
In many cases, when a plant becomes commercialized as a herbal
medicine or when one of its constituents starts getting used as a pharma-
ceutical drug, its populations become threatened due to extensive
wildcrafting and unsustainable harvesting techniques (Cordell, 2011;
Vines, 2004). The classical example for this compound supply problem
was the so-called “taxol supply crisis”(Cragg et al., 1993; Kingston,
2011). When the compound turned out to possess remarkable clinical ef-
ficacy in ovarian cancer, suddenly the demand for taxol increased tremen-
dously. However, at that time, the compound was only accessible from
the bark of the western yew (Taxus brevifolia L.). This was on one hand
problematic because the whole production process including tedious
bark collection and drying, extraction, and purification was very time-
consuming. On the other hand, concerns on the ecological impact of in-
tensive bark collection were raised (Cragg et al., 1993; Kingston, 2011).
Although taxol is meanwhile accessible via alternative routes (see
Section 4.2), the problem of sustainable supply of herbal material still fre-
quently occurs. Cultivation would be a more sustainable alternative to
wildcrafting, nevertheless, approximately two thirds of the 50,000 medic-
inal plant species used world-wide are still wildcrafted (Canter et al.,
2005). Therefore, institutions like WHO (WHO, 2003)andEuropeanMed-
icines Agency (EMA) (EMA, 2006) developed guidelines on good agricul-
tural and collection practices (GACP) for medicinal plants in order to
promote sustainable plant collection techniques and to reduce the ecolog-
ical problems produced by wildcrafting of medicinal plants.
Apart from that, ecological and legal considerations also have an im-
portant influence on accessibility of plants as a source of drug discovery,
especially laws dealingwith plant access and sharing of benefits, as well
as patentability issues with local governments in the countries of origin.
The United Nation's Convention on Biological Diversity (CBD; http://
www.cbd.int/doc/legal/cbd-en.pdf), signed in 1992 by the international
community in Rio de Janeiro, Brazil, aims at: (1) conserving the biodi-
versity; (2) sustainably using its genetic resources; and (3) sharing
the benefits from their use in a fair and equitable manner (Cragg et al.,
2012; Kingston, 2011; Soejarto et al., 2004). Resulting from CBD, the
provider countries, their people and representatives also become im-
portant stakeholders that need to be considered in plant-based drug
discovery programs (Heinrich, 2010a). Although the CBD provided a
framework for countries to regulate and define bioprospecting, the trea-
ty left many open questions, particularly in the issue of access and ben-
efit sharing (Cragg et al., 2012; Kingston,2011). On one hand, CBD could
not always rebut the skepticism toward bioprospecting in many devel-
oping countries due to previous exploitation of their biodiversity; some
countries issued very stringent protective regulations, and some were
very slow in establishing the necessary legal framework, leadingto con-
fusion where to go for permissions and who had authority to grant
them. On the other hand, the expectations of biodiversity-rich countries
on the potential monetary benefits to achieve from drugs developed
from their genetic resources were highly exaggerated, if one considers
that from the 114,000 extracts derived from 12,000 species that the
US National Cancer Institute (NCI) investigated over decades, only
taxol and camptothecin are currently used as pharmaceutical drugs.
These issues frequently hampered the access to samples from
biodiversity-rich countries in the last two decades and thereby discour-
aged pharmaceutical companies from natural product-based drug dis-
covery (Cragg et al., 2012; David et al., 2015; Kingston, 2011). In order
to improve this unfavorable situation, the Nagoya Protocol on access
to genetic resources and the fair and equitable sharing of benefits aris-
ing from their utilization to the convention on biological diversity
(http://www.cbd.int/abs/doc/protocol/nagoya-protocol-en.pdf), has
been published in 2011 and has come into force in October 2014, after
1584 A.G. Atanasov et al. / Biotechnology Advances 33 (2015) 1582–1614

reaching ratification by 50 countries. The protocol islegally binding and
particularly aims at bringing more clarity into questions of access and
benefit sharing (Burton and Evans-Illidge, 2014; Oliva, 2011). Although
some researchers are worried that the protocol might lead to stricter
regulations that could hamper drug discovery and even be counterpro-
ductive for biodiversity conservation (Gilbert, 2010), others expect that
–provided that the protocol is implemented into national laws in a
sensitive way –the higher legal certainty will revitalize the interest to
investigate plants from biodiversity-rich countries, and thereby provide
incentives to conserve biodiversity (Burton and Evans-Illidge, 2014;
Cragg et al., 2012).
Besides the accessibility of the plant material, also its quality is of
great importance. Available plant material often varies on quality and
composition and this can hamper the assessment of its therapeutic
claims. The chemical composition is not only dependent on species
identity and harvest time, but also on soil composition, altitude, actual
climate, processing, and storage conditions. Moreover, during extrac-
tion, as well as during the isolation processes, transformation and
degradation of compounds can occur (Bucar et al., 2013; Jones and
Kinghorn, 2012). Another aspect determiningthe chemical composition
of the startingplant material is that endophytic organisms, suchas fungi
and bacteria, might inhabit plants. As a result, natural products present
in the collected plant material might be in some occasions metabolites
of the endophytic organism, or plant products induced as a result of
the interaction with this organism (David et al., 2015).
Further complications related to the resupply of bioactive natural
products arise from the fact that natural products are more likely to
have complex chemical structures with numerous oxygen-containing
substituents and chiral centers, which hampers the development of
methods for total synthesis or derivatization that might be needed for
property optimization of drug candidates. In contrast, pharmaceutical
leads originating from synthetic libraries are usually comparably easy
to generate and modify using simple chemical approaches (Butler,
2004; Henrich and Beutler, 2013; Li and Vederas, 2009).
Another major challenge for natural product drug discovery
programs is common incompatibility of natural products with HTS
(Koehn and Carter, 2005). Investigation of a large number of plant
extracts by HTS, followed by the identification and characterization of
bioactive constituents is highly challenging. Adaption and changes of
sample preparation and assay designs are necessary in order to apply
HTS for bioactivity detection of plant extracts and to identify potent
pure compounds thereof. In general, HTS can be conducted using
cell-free or cell-based assays. It requires high reproducibility, accuracy,
robustness, and reliable liquid handling systems. Test compounds
should not decompose or precipitate, should not interfere with assay
reagents nor show non-specific effects. Especially natural products
often fail in fulfilling these requirements. The maintenance of plant
extracts integrity might be very problematic due to their complexity.
Extracts often show high viscosity, tend to aggregate or precipitate, or
contain components that non-specifically bind proteins, which can re-
sult in misleading assay outcomes, therefore necessitating sophisticated
sample preparation or fractionation of thecrude extracts prior to testing
(Coan et al., 2011; Johnson et al., 2011; Maes et al., 2012; Schmid et al.,
1999; Tu et al., 2010). Natural product extracts are also likely to contain
fluorescent or fluorescence quenching compounds, which interfere
with the fluorescent HTS endpoint measurements, whereby the pres-
ence of colored compounds might also interfere with colorimetric HTS
endpoints (Gul and Gribbon, 2010; Henrich and Beutler, 2013; Zou
et al., 2002). Moreover, plant extracts may contain compound classes
that are obstructive for certain assay types and might lead to false
positive or false negative results. Particularly common highly apolar
compounds, such as fatty acids (Balunas et al., 2006), common polar
compounds, such as polyphenols and flavonoids (Zhu et al., 1997,
2013; Zou et al., 2002), as well as chlorophyll (Henrich et al., 2006)
might be especially problematic since they can interfere with a range
of different assays. Much effort is necessary to remove such constituents
from samples prior to testing (Cardellina et al., 1993; Collins et al., 1998;
Picker et al., 2014) or to modify the assay system in order to avoid their
detection (Sasiela et al., 2008). Next to organic molecules, also some
inorganic constituents, such as metals, can lead to false positive results
in HTS (Hermann et al., 2013). This might be especially problematic
for HTS of plant extracts, since many plants concentrate metals from
their environment (Fernando et al., 2013), and metal impurities could
be present in commercially available plant samples (Eisenberg et al.,
2011). Furthermore, cytotoxic constituents might cause problems in
cell-based assays, since they can mask the detection of other bioactiv-
ities or the presence of other compounds with the desired efficacy.
Saponins, for example, which possess detergent effects, can lead to the
lysis of cells and therefore interfere with theresults of cell-based assays
(Henrich and Beutler, 2013). For further details on the application of
HTS to natural product samples and its challenges,the reader is referred
to the excellent and comprehensive recent review by Henrich and
Beutler (Henrich and Beutler, 2013).
Further difficulty is set by the fact that determination of the precise
molecular mechanism of action of natural products is a challenging
task [e.g., curcumin, triptolide; (Corson and Crews, 2007)].However,a
detailed knowledge of the interaction of a drug candidate compound
with its molecular target is very advantageous for the drug develop-
ment process, because it allows property optimization by medicinal
chemistry approaches, and on some occasions a more appropriate
clinical trial design.
The conduction of rigorous clinical trials needed for approval of
natural products as drugs represents another major difficulty. While
such clinical trials are often feasible just with industrial support due to
the high costs, at the same time the interest of pharmaceutical compa-
nies in natural products that are not synthetically modified is limited
due to controversies with their patentability [e.g., curcumin; (Corson
and Crews, 2007)]. In this line, the recent situation concerning patent-
ability of natural products got even more difficult after new guidelines
were issued on 4th of March 2014 by the United States Patent and
Trademark Office (“Guidance for Determining Subject Matter Eligibility
of Claims Reciting or Involving Laws of Nature, Natural Phenomena, &
Natural Products”). The new guidelines state that a patent claim must
demonstrate a “marked difference”from a known natural law, material,
or phenomenon, and their issuing came after two relevant high-profile
Supreme Court decisions: The Association for Molecular Pathology versus
Myriad, which ruled that isolated and purified DNA is not patentable,
and Mayo versus Prometheus, which ruled that methods for determina-
tion of optimal drug doses based on levels of a naturally occurring me-
tabolite is not eligible for patenting (Harrison, 2014). Aside of issues
related to patentability of natural products, it should be also noted
that there has been some general shift in the pharmaceutical industry
from small molecule-based drug discovery toward biological big
molecules (biologicals; e.g., therapeutic proteins or nucleic acids)
(Appendino et al., 2010). However, the patient costs for biologicals are
much higher than the costs for small molecule drugs, and their increas-
ing use is putting a rising pressure on national health insurances.
Furthermore, the high prices of this type of pharmaceuticals is not
expected to quickly drop in the near future after patents expiration,
because biological generics (biosimilars) require additional clinical bio-
equivalence studies to be approved for useand overall have significantly
higher development and manufacturing costs in comparison to small
molecule generics. Therefore, to relive the increasing financial burden,
turning back to some “old”small molecule-based approaches is
conceivable (Appendino et al., 2010).
Due to the challenges described above, the interest in natural
product-based drug discovery has been gradually declining. Even in
the last decade, many big and medium-sized pharmaceutical compa-
nies, which were still active in the area in the 1990s, terminated their
natural product programs, leaving natural product research to a major
extent to academic universities and start-up companies (Beutler,
2009; David et al., 2015; Ortholand and Ganesan, 2004).
1585A.G. Atanasov et al. / Biotechnology Advances 33 (2015) 1582–1614

2.3. Renewal of the interest in natural product-based drug discovery
The results obtained by HTS of large synthetic compound libraries,
which were introduced in the 1990s (Scannell et al., 2012), did not
meet the expectations. Instead of introducing more drugs to the market,
the approval rates of new drugs declined. Whereas 45 new drugs were
approved by the US Food and Drug Administration (FDA) in 1990, only
21 were approved in 2010 (David et al., 2015; Kingston, 2011). While
the reasons for this declining trend are complex (Kola and Landis,
2004), one important aspect is that synthetic compound libraries usually
cover only a small range of the chemical diversity. Moreover, due to sim-
ilar generation strategies, the HTS-compound libraries of different compa-
nies often overlap. Due to high sample numbers in such libraries,
compounds to be investigated further are often selected quickly, mainly
based on potency values (Scannell et al., 2012), although a negative corre-
lation of cell-free in vitro potency and favorable ADME/T (absorption, dis-
tribution, metabolism, excretion/toxicity) is likely (Gleeson et al., 2011).
The plant kingdom includes a high number of species, producing a
diversity of bioactive compounds with different chemical scaffolds.
According to previous estimations only 6% of existing plant species
have been systematically investigated pharmacologically, and only
around 15% phytochemically (Cragg and Newman, 2013; Fabricant
and Farnsworth, 2001; Verpoorte, 1998, 2000). Although today the
percentage of better characterized species is likely higher due to ongo-
ing research efforts, it is still conceivable that there is a huge number
of plant compounds that are not well investigated pharmacologically
in the approximately 310,000 plant species described so far (IUCN,
2015). Unfortunately, as a result of ongoing climate changes and
anthropogenic factors, a significant decrease in global vegetative species
in the next years is predicted (Maclean and Wilson, 2011; Thomas et al.,
2004), endangeringthe sources of potential new drugs from nature and
prompting urgent actions.
Since natural products are made from living organisms, they possess
properties that are evolutionary optimized for serving different biolog-
ical functions (e.g., binding to specific target proteins or other biomole-
cules) (Appendino et al., 2010; Hunter, 2008). Detailed analyses of
structural differences between natural products and molecules generat-
ed by combinatorial synthesis found that major differences originate
from the introduction of properties making combinatorial synthesis
more efficient. For example, chiral separation is challenging and expen-
sive. Therefore, creating molecules with a low number of chiral centers
is favorable. Besides a much lower number of chiral centers, synthetic
compounds tend to have a lower molecular weight, a higher number
of freely rotatable bonds, higher chain lengths, a lower number of
rings, less oxygen but more nitrogen, sulfur, and halogen atoms, a
lower number of Lipinski-type H-bond acceptors and donors, and
higher calculated octanol-water partition coefficients (cLogP values).
Other differences are the complexity of ring systems and the degree of
saturation (Feher and Schmidt, 2003; Koehn and Carter, 2005; Lee and
Schneider, 2001; Stahura et al., 2000). These structural differences,
especially the significantly lower number of chiral centers, the lower
size, and the higher flexibility result in weaker and less specific activity
of synthetic compounds (Feher andSchmidt, 2003; Klebe,2009). On the
contrary, natural products often have selective biological actions due to
binding affinities for specific proteins relevant for their biological
functions, possess superior chemical diversity and complexity
developed during biosynthesis (Clardy and Walsh, 2004; Koehn and
Carter, 2005), and often have more advantageous ADME/T properties.
Particularly in the context of drug discovery from medicinal plants, a
big advantage is that sometimes well documented ethnopharmacological
information about the traditional use is available, which can provide hints
for compounds therapeutically effective in humans (Corson and Crews,
2007; Heinrich, 2010a; Heinrich and Gibbons, 2001; Kinghorn et al.,
2011). In line with this notion, analysis of 122 plant-derived compounds
identified to be globally used as drugs revealed that 80% of them originate
from plants that have ethnomedical use identical or related to the
indications for which the respective pure compounds are prescribed
(Fabricant and Farnsworth, 2001; Farnsworth et al., 1985). Furthermore,
it was even demonstrated that natural products used for the develop-
ment of medicines are highly likely to be used traditionally, even if this
was not known at the stage of drug development (e.g., the discovery of
the anti-cancer agent taxol from T. brevifolia was done with a random
screening approach, but later on it came to light that the plant has been
used by western Indian cultures as a medicine) (Heinrich, 2010a;
Moerman, 1998). Importantly, knowledge about traditionally used med-
icines is disappearing even faster than the biodiversity of plant species
and with the current tempo of globalization, much valuable information
is on the risk of getting lost forever (Appendino et al., 2010).
Resulting from the above discussed advantages of natural products,
in spite of the predominant industrial focus on HTS approaches with
synthetic compound libraries, natural products still represent a valuable
source for drug discovery (Newman and Cragg, 2012). Plant-derived
natural products approved for therapeutic use in the last thirty years
(1984–2014) are summarized in Table 1. As evident from the table,
these plant-derived natural products are modulating a diverse range
of molecular targets and are used for the treatment of various disease
conditions. As a snapshot of thecurrent state and perspective for future
developments, overview of advanced plant-derived small chemical en-
tities that are in actively recruiting clinical trials are presented in Table 2.
Reflecting better appreciation of the advantages of natural products
and the growing interest in plant-derived natural product-based drug
discovery, a rapid increase in the number of scientific studies targeting
this research area is observed upon analysis of the recent PubMed pub-
lication trends (Fig. 1; data retrieved with MEDSUM, http://webtools.
mf.uni-lj.si/public/medsum.html). The revived scientificinterestin
plant-derived natural product-based drug discovery is paralleled with
major scientific and technological advances in the relevant research
fields, including better understanding of diseases and their underlying
mechanisms, advances in screening methods and analytical equipment,
increasing number of targets available for testing, and improved possi-
bilities for optimization of natural leads using synthetic modification
strategies [reviewed in (Henrich and Beutler, 2013; Koehn and Carter,
2005; Li and Vederas, 2009; Paterson and Anderson, 2005)].
3. Approaches for the discovery of new pharmacologically active
plant compounds
3.1. Approaches to selecting starting material
In the random screening approach (Table 3), plant extracts, enriched
fractions, or isolated compounds are randomly selected on the basis of
their availability. In the context of plant-based drug discovery, this ap-
proach might be highly advantageous when applied with samples origi-
nating from regions of high biodiversity and endemism, as the chemical
diversity of natural products can reflect the biodiversity of their source
organisms (Barbosa et al., 2012; Henrich and Beutler, 2013). The random
selection of test material has the potential to result in the identification of
unexpected bioactivities that could not have been predicted based on the
existing knowledge. However, the used pharmacological assays often
have a small- or medium-throughput, and the starting test samples
(extracts, fractions, or pure compounds) are often available only in
small amounts, limiting the number of bioassays in which they can be
tested. Therefore, as alternative to random testing intrinsically suffering
from a low hit-rate, a variety of knowledge-based strategies can be ap-
plied to increase the probability for the identification of relevant bioactive
compounds out of a smaller number of test samples with the use of a
limited number of carefully selected pharmacological assays.
The classical knowledge-based approach is the ethnopharmacological
approach (Table 3), where the traditional medicinal use of plants consti-
tutes the basis for the selection of the test material and the pharmacolog-
ical assay. Ethnopharmacology involves the observation, description, and
experimental investigation of traditionally used drugs and their
1586 A.G. Atanasov et al. / Biotechnology Advances 33 (2015) 1582–1614

Table 1
Plant-derived natural products approved for therapeutic use in the last thirty years (1984–2014)
a
.
Generic name and chemical structure Plant species (literature reference) Trade name (year of
introduction)
Indication (mechanism of action)
Artemisinin Artemisia annua L. (Klayman et al., 1984) Artemisin (1987) Malaria treatment (radical formation)
Arglabin Artemisia glabella Kar. et Kir. replaced by Artemisia
obtusiloba var. glabra Ledeb. (Adekenov et al., 1982)
Arglabin (1999) Cancer chemotherapy (farnesyl transferase
inhibition)
Capsaicin Capsicum annum L., or C. minimum Mill.
(Toh et al., 1955)
Qutenza (2010) Postherpetic neuralgia (TRPV1 activator)
Colchicine Colchicum spp. (Leete and Nemeth, 1960) Colcrys (2009) Gout (tubulin binding)
Dronabinol / Cannabidol
Dronabinol
Cannabidol
Cannabis sativa L. (Vree et al., 1972) Sativex
b
(2005) Chronic neuropathic pain (CB1 and CB2
receptor activation)
Galanthamine Galanthus caucasicus (Baker) Grossh.
(Tsakadze et al., 1969)
Razadyne (2001) Dementia associated with Alzheimer's disease
(ligand of human nicotinic acetylcholine
receptors (nAChRs))
Ingenol mebutate Euphorbia peplus L. (Hohmann et al., 2000) Picato (2012) Actinic keratosis (inducer of cell death)
Masoprocol Larrea tridentata (Sessé & Moc. ex DC.) Coville
(Luo et al., 1998)
Actinex (1992) Cancer chemotherapy (lipoxygenase
inhibitor)
Omacetaxine mepesuccinate
(Homoharringtonine)
Cephalotaxus harringtonia (Knight ex Forbes) K. Koch
(Powell et al., 1974)
Synribo (2012) Oncology (protein translation inhibitor)
(continued on next page)
1587A.G. Atanasov et al. / Biotechnology Advances 33 (2015) 1582–1614

bioactivities. It represents a transdisciplinary concept based on botany,
chemistry, biochemistry, and pharmacology, that involves many disci-
plines beyond natural science, such as anthropology, archeology, history,
and linguistics (Fabricant and Farnsworth, 2001; Heinrich, 2010a; Leonti,
2011). Some prominent examples of approved drugs that were initially
discovered by the use of ethnopharmacological data are: khellin from
Ammi visnaga (L.) Lam. that served as lead compound for the develop-
ment of chromoglicic acid, the sodium salt of which is used as mast cell
stabilizer in allergy and asthma; galegine from Galega officinalis L. that
was the template for the synthesis of metformin and triggered the subse-
quent development of biguanidine-type antidiabetic drugs; papaverine
from Papaver somniferum L. which was the basis for the development of
the antihypertensive drug verapamil; quinine from the bark of Peruvian
Cinchona species that was used to treat malaria and inspired the synthesis
of chloroquine and mefloquine which largely replaced quinine in the mid
of the 20th century (Cragg and Newman, 2013; Fabricant and
Farnsworth, 2001); the antimalarial drug artemisinin that has been iso-
lated from the TCM herb A. annua L. in 1971 and led to the development
of derivatives, such as sodium artensunate or artemether, that are nowa-
days widely used to treat malaria (Klayman, 1985; White, 2008).
In the well-established traditional medical systems, such as TCM or
Ayurveda, ethnopharmacological knowledge is comparably easily
accessible, as these systems possess an established body of written
knowledge and theory that has often been revised throughout the
centuries and is still in use today. In medical systems based on herbal-
ism, folklore, or shamanism, however, no written documents exist,
and the herbal formulations used are often kept secret by the practi-
tioners, making the information more difficult to access (Brusotti
et al., 2014; Fabricant and Farnsworth, 2001). Depending on the herbs
to be studied, information can be acquired from different sources,
including books on medical botany (e.g., Lewis, 2003), herbals
(e.g., Adams et al., 2012), review articles on medicinal plants used in a
certain geographic region or by an ethnic culture (e.g., Gairola et al.,
2014), field work (e.g., Kunwar et al., 2009), and computer databases
(Leonti, 2011; Ningthoujam et al., 2012).
A system particularly amenable to ethnopharmacological studies is
TCM. In contrast to other traditional medical systems, TCM has always
been regarded as a science, and it has been taught at medical schools
for more than 2000 years. The first textbook fully devoted to the
description of herbal drugs is the Shen-nung-pen-ts'ao ching (Shen
Nung's Classic of Pharmaceutics). Beginning with this compendium,
the first version of which was probably written down during the later
Han period (25–220 AD), the literature of Chinese Materia Medica de-
veloped by continuous addition of new drugs as well as re-evaluation
and addition of new indications for existing herbs during the centuries
(Bauer, 1994; Unschuld, 1986; Zhu, 1998). The fact that TCM always
possessed a scientific status makes it an extremely valuable source for
acquisition of ethnopharmacological data, as the development in the
use of medicinal plants can be readily traced back in history by studying
the ancient textbooks.
The use of the ethnopharmacology-based approach, however, is
associated with multiple challenges:
(1) Herbs that have been selected as study candidates based on
ethnopharmacological data require not just detailed knowledge
about their habitat, abundance, correct botanical authentication,
whether they are threatened or endangered, and which permits
are necessary in order to collect and investigate them (David
et al., 2015; Fabricant and Farnsworth, 2001), but might also pro-
voke occasions of legal right-claims from the country of origin or
from ethnical groups in which the traditional knowledge was
originally generated. In this context, also access and benefit
sharing issues determined in the United Nation's Convention
on Biological Diversity, and in the Nagoya Protocol need to be
respected (see Section 2.2). These restrictions make collection
of plants on an ethnopharmacological basis more tedious and
time-consuming than a mere random collection, which is
regarded more feasible for the common practices of pharmaceu-
tical industry (Fabricant and Farnsworth, 2001).
(2) Some traditional systems, suchas TCM and Ayurveda, involve the
application of sophisticated multicomponent mixtures. The
complexity of these formulations and possible synergistic effects
heavily complicate the identification of active principles. On the
other hand, this combinatorial approach might provide new
perspectives in the treatment of multifactorial diseases, such as
dementia, that might be better addressed by a multitarget-
oriented, combinatorial approach (Kong et al., 2009).
(3) The definitions of health and disease in traditional medicine often
widely deviate from the Western reductionist approach that is
mainly based on anatomy, physiology, and cell and molecular
biology. For example, the theory of TCM has been strongly
influenced by Chinese philosophy, like the theory of Yin and
Table 1 (continued)
Generic name and chemical structure Plant species (literature reference) Trade name (year of
introduction)
Indication (mechanism of action)
Paclitaxel Taxus brevifolia Nutt. (Wani et al., 1971) Taxol (1993), Abraxane
c
(2005), Nanoxel
c
(2007)
Cancer chemotherapy (mitotic inhibitor)
Solamargine Solanum spp. (Hsu and Tien, 1974; Liljegren, 1971) Curaderm
d
(1989) Cancer chemotherapy (apoptosis triggering)
Notes:
a
Resources: (Butler, 2005, 2008; Butler et al., 2014; Fürst and Zündorf, 2014; Newman and Cragg, 2012)andwww.drugs.com.
b
Mixture of the two compounds.
c
Paclitaxel nanoparticles.
d
Containing not just solamargine but also other solasodine glycosides.
1588 A.G. Atanasov et al. / Biotechnology Advances 33 (2015) 1582–1614

Yang, emphasizing the balance of functional systems, and the
theory of the Five Phases (Wu Shing)(fire, water, metal, wood,
and earth), that are connected to five functional areas in the
body (liver, heart, lung, kidney, and spleen and stomach)
(Cheng, 2000; Uzuner et al., 2012). Such discrepancies to Western
terminology often complicate the correct interpretation of
Table 2
Plant-derived natural products in clinical trials
a
.
Generic name and chemical structure Plant species (literature reference) Number of recruiting clinical trials
b
: indications (potential mechanism of
action)
β-Lapachone Haplophragma adenophyllum (Wall. ex G. Don)
Dop (Joshi et al., 1979)
1 trial: Solid tumors (E2F1 pathway activator)
Curcumin Curcuma longa L. (Turmeric)(Janaki and Bose,
1967)
26 trials: Cognitive impairment, different types of cancer, familial
adenomatous polyposis, schizophrenia, cognition, psychosis, prostate
cancer, radiation therapy, acute kidney injury, abdominal aortic aneurysm,
inflammation, vascular aging, bipolar disorder, irritable bowel syndrome,
neuropathic pain, depression, somatic neuropathy, autonomic dysfunction,
Alzheimer's disease, plaque psoriasis, fibromyalgia, cardiovascular disease
(NF-κB inhibition)
Epigallocatechin-3-O-gallate Camellia sinensis (L.) Kuntze (Green tea)
(Kada et al., 1985)
14 trials: Epstein-Barr virus reactivation in remission patients with
nasopharyngeal carcinoma, multiple system atrophy, Alzheimer's disease,
cardiac amyloid light-chain amyloidosis, Duchenne muscular dystrophy,
cystic fibrosis, diabetic nephropathy, hypertension, fragile X syndrome,
different types of cancer, obesity, influenza infection (cell growth arrest and
apoptosis induction)
Genistein Genista tinctoria L. (Perkin and Newbury, 1899) 5 trials: Colon cancer, rectal cancer, colorectal cancer, Alzheimer's disease,
non-small cell lung cancer, adenocarcinoma, osteopenia, osteoporosis
(protein-tyrosine kinase inhibitor, antioxidant)
Gossypol Gossypium hirsutum L. (Heinstein and El-Shagi,
1981)
2 trials: B-cell chronic lymphocytic leukemia, refractory chronic
lymphocytic leukemia, stage III chronic lymphocytic leukemia, stage IV
chronic lymphocytic leukemia, non-small cell lung cancer (Bcl-2 inhibitor)
Picropodophyllotoxin Podophyllum hexandrum Royle, replaced by
Sinopodophyllum hexandrum (Royle) T.S. Ying
(Jackson and Dewick, 1984)
1 trial: Glioblastoma, gliosarcoma, anaplastic astrocytoma, anaplastic
oligodendroglioma, anaplastic oligoastrocytoma, anaplastic ependymoma
(tubulin binding/IGF-1R Inhibitor)
Quercetin Allium cepa L. (Bilyk et al., 1984) 9 trials: Chronic obstructive pulmonary disease, Fanconi anemia, different
types of prostate cancer, diabetes mellitus, obesity, diastolic heart failure,
hypertensive heart disease, heart failure with preserved ejection fraction,
hypertension, oxidative stress, Alzheimer's disease, pancreatic ductal
adenocarcinoma, plaque psoriasis (NF-κB inhibition)
Resveratrol Vitis vinifera L. (Langcake and Pryce, 1976) 22 trials: Pre-diabetes, vascular system injuries, lipid metabolism disorders
(including non-alcoholic fatty liver disease), endothelial dysfunction,
gestational diabetes, cardiovascular disease, type 2 diabetes mellitus,
inflammation, insulin resistance, disorders of bone density and structure,
metabolic syndrome, coronary artery disease, obesity, memory impairment,
mild cognitive impairment, diastolic heart failure, hypertensive heart
disease, heart failure with preserved ejection fraction, hypertension,
oxidative stress, polycystic ovary syndrome, Alzheimer's disease (NF-κB
inhibition)
Notes:
a
Resources: (Butler, 2005, 2008; Butler et al., 2014; Fürst and Zündorf, 2014; Newman and Cragg, 2012), www.clinicaltrials.gov,andwww.drugs.com.
b
Determined from www.clinicaltrials.gov on 22nd of October, 2014, including trials in which the respective natural product is applied alone or as a mixture with other constituents,
excluding studies with unknown status.
1589A.G. Atanasov et al. / Biotechnology Advances 33 (2015) 1582–1614

ethnopharmacological data. Moreover, the holistic, personalized
approach of these medical systems is difficult to assess by many
of the bioassay systems currently used to prove pharmacological
activity. However, the emerging omics- and systems biology-
based technologies might be better suited to address these issues,
due to their more holistic orientation (Guo et al., 2014; Ngo et al.,
2013).
Next to the ethnopharmacological approach, another possibility for
selection of plant material for the pharmacological testing is the
chemosystematic or phylogenetic approach, making use of chemotaxo-
nomic knowledge and molecular phylogenetic data in order to select
plant species from genera or families known to produce compounds
or compound classes associated with a certain bioactivity or therapeutic
potential in a more targeted manner (Barbosa et al., 2012). Combined
phylogenetic and phytochemical studies have shown that there is a
strong phylogenetic signal in the distribution of secondary metabolites
in the plantkingdom that can be exploited in the search for novel natu-
ral products (Saslis-Lagoudakis et al., 2011, 2012). As an example,
Rønsted et al. and Larsen et al. used a phylogenetic approach to select
the most promising target plants from the genus Narcissus and from
the Amaryllidaceae tribe Galantheae for discovery of further acetylcho-
linesterase inhibitory alkaloids (Larsen et al., 2010; Rønsted et al.,
2008). The combination of phylogenetic information with traditional
ethnobotanical knowledge constitutes the emerging field of “phyloge-
netic ethnobotany”(Saslis-Lagoudakis et al., 2011). The basic assump-
tion of this approach is that medicinal properties are not randomly
distributed throughout the plant kingdom, but that some plant taxa
are represented by more medicinal plants than others, and that
selection of species from these “hot”taxa will lead to higher success
rates in drug discovery. Particularly the exploration of cross-cultural
ethnomedical patterns within a phylogenetic framework is regarded
as a very powerful tool for identification of highly promising plant
groups, when phylogenetically related plant species from very distant
regions are found to be used for medical conditions in the same
therapeutic areas (Saslis-Lagoudakis et al., 2011, 2012).
The ecological approach to select plant material is based on the ob-
servation of interactions between organisms and their environment
that might lead to the production of bioactive natural compounds. The
hypothesis underlying this approach is that secondary metabolites,
e.g., in plant species, possess ecological functions that may have also
therapeutic potential for humans. For example, metabolites involved
in plant defense against microbial pathogens may be useful as antimi-
crobials in humans, or secondary products defending a plant against
herbivores through neurotoxic activity could have beneficial effects in
humans due to a putative central nervous system activity (Barbosa
et al., 2012). This hypothesis might be justified due to the fact that a
high proportion of biochemical architecture is common to all living or-
ganisms; considering this, it seems reasonable that secondary metabo-
lites from organisms as distant as plants, fungi and bacteria are all able
to interact with the macromolecules of the human body (Caporale,
1995). In a subtype of this approach, those plants are selected as poten-
tially active candidates that are ingested by animals for putative self-
medication purposes and for reconstitution of physiological homeosta-
sis, e.g. in order to alleviate microbial or parasitic infection, to enhance
reproduction rates, to moderate thermoregulation, to avoid predation,
and to increase alertness. This concept is also referred to as
zoopharmacognosy (Barbosa et al., 2012; Forbey et al., 2009; Huffman,
2003). For example, compounds with antimalarial and antiprotozoal ac-
tivity could be isolated from plant species that were ingested by chim-
panzees and baboons in the wild in unusual feeding behavior,
supposedly in order to control intestinal parasite infection (Krief et al.,
2004; Obbo et al., 2013).
Computational methods are another very powerful knowledge-
based approach that helps to select plant material or natural products
with a high likelihood for biological activity. These methods can also
aid with the rationalization of biological activity of natural products. In
silico simulations can be used to propose protein ligand binding charac-
teristics for molecular structures, e.g., known constituents of a plant
material. Compounds that perform well in in silico predictions can be
used as promising starting materials for experimental work. Activity
predictions using virtual screening have intriguing success rates (Hein
et al., 2010), and can be conducted with a wide variety of different
Fig. 1. PubMed publication trend analysis, demonstrating increased scientific interest in plant-derived natural product pharmacology, chemistry, and drug discovery. Thedata were re-
trieved with MEDSUM (http://webtools.mf.uni-lj.si/public/medsum.html) on 15th of June 2015, and cover the time period 1982–2012 (newer data are not included because of the
lack of coverage). As indicated, the used search keywords were plant chemistry,plant pharmacology,plant natural product,plant compound,plant drug discovery,plant bioactivity,andthe
total number of PubMed publications per year was retrievedby search with the symbol *. The trend analysis reveals that the increase of PubMed citations in the target areas is faster
than the increase in the total number of annual PubMed citations (indicated by the steeper slopes of the respective trend lines).
1590 A.G. Atanasov et al. / Biotechnology Advances 33 (2015) 1582–1614

computational methods (Rollinger et al., 2008). In silico studies can
focus on the main constituents of herbal remedies (Rollinger et al.,
2009) as well as on any natural compounds with relevant biological
effects directly retrieved from the literature. The required availability
of structurally and stereochemically well-defined compounds and the
approach-inherent incapability to find novel compounds, constitute
the limitations of virtual screening studies. Molecular descriptors for a
compound can be calculated from its 2D or 3D structure. Simple exam-
ples for such descriptors are thenumber of rotatable bonds or hydrogen
bond acceptors. These descriptors can then be compared to datasets of
active compounds to recognize correlations and establish quantitative
structure activity relationship (QSAR) models to preselect compounds
with a higher likelihood of activity on a specific target. In a study by
Gavernet et al., for example, QSAR models were used to screen for
new potential anticonvulsantsin a naturalproducts databasecontaining
10,900 molecules. Using a series of computational filters, they proposed
four hit compounds, one of which was experimentally evaluated and
confirmed as active (Gavernet et al., 2008). When applying this method
to plant constituents, however, it is important to stress that natural
products often differ from typical synthetic pharmacologically active
molecules (e.g., size, number of aromatic rings, flexibility), that are
mainly found in bioactivity databases, such as CHEMBL (Gaulton et al.,
2012) or PubChem (Li et al., 2010). Consequently, it is necessary to
carefully evaluate datasets for compatibility (Feher and Schmidt, 2003).
Pharmacophore-based virtual screening constitutes another highly
successful computational method. A pharmacophore model is a 3D
arrangement of physicochemical features (e.g., hydrogen bond donor/
acceptor, hydrophobic area, aromatic ring) that represents the key
interactions between a ligand molecule and its target protein. As an
example, the chemical interaction pattern that defines the interaction of
magnolol with the binding site of PPARγ(PDB 3R5N) is presented in
Fig. 2A [for more details about the significance of this example the reader
is referred to (Zhang et al., 2011) and (Fakhrudin et al., 2010)].
3D-multiconformational compound libraries can be screened against a
pharmacophore model to retrieve molecules that map the
pharmacophore features and consequently have a high likelihood of
being active on the target. Depending on the target, this method can
achieve success rates between 2 and 30% (Hein et al., 2010). If
pharmacophore models are available for a range of targets, parallel virtual
screening can be used for so-called “target fishing”(Schuster, 2010). This
method can be highly valuable for target identification if a general activity
of an extract or pure compound is known. The structure can be screened
against a set of models for multiple targets to reduce the experimental
work identifying the molecular target(s) related to the bioactivity
(Duwensee et al., 2011; Steindl et al., 2006; Wolber and Rollinger, 2013).
A third computational method, moleculardocking, is widely used to
elucidate the mechanism of action and rationalize structure activity
relationships of natural products. The aim of docking is to accurately
predict the positioning of a ligand within a protein binding pocket
and to estimate the strength of the binding with a docking score
(Waszkowycz et al., 2011). As an example, Fig. 2B visualizes the
empty binding pocket of PPARγ, which can be used in docking
simulations to place new molecules into the binding site and to
calculate the binding free energy of the ligand. If the 3D structure of a
protein is available, either from X-ray crystallography, NMR data, or
through homology modeling, then ligand molecules can be computa-
tionally positioned directly in the binding pocket to analyze their puta-
tive target-ligand interactions and thus identify the crucial binding
features of the molecule. Docking can also be employed in large scale
virtual screenings, where a molecule is docked into a series of targets
and a suitable docking score is used to compare the results to identify
the best ranked matches. Docking has been widely employed to ratio-
nalize the structure–activity relationship of natural products. This was
demonstrated for example recently with constituents of Carthamus
tinctorius L., which showed different activities on indoxyl indoleamine
2,3-dioxygenase (Temml et al., 2013).
Computational methods also provide the means to discover
previously undescribed binding sites on known protein structures.
Pocket finders detect solvent-accessible cavities in the protein surface
that can indicate potential ligand binding sites. These sites can then be
analyzed computationally. In a study conducted by Hanke et al., for ex-
ample, this approach was used to identify binding sites for a series of
aminothiazole-featured pirinixic acid derivatives, which showed dual
activity on 5-lipoxygenase and microsomal prostaglandin E
2
synthase-
1, and new potential binding sites for both enzymes were identified
(Hanke et al., 2013).
Whereby in silico methods represent valuable filter tools in the
search of new activities for natural products, they can also be employed
to predict ADME/T properties (Kaserer et al., 2014)ortofind new
activities for already approved drugs (drug repurposing) (Steindl
et al., 2007).
Several virtual libraries comprising collections of natural products
have been reported. For example, the DIOS database is a collection of
9676 literature-derived compounds from plants described by Pedanius
Dioscorides in his fundamental encyclopedia De Materia Medica (1st
century AD), which is widely regarded as the precursor to all modern
pharmacopeias. The Natural Product Database (NPD) is a more general
database, that collects more than 122,700 compounds from natural
sources (Rollinger et al., 2004). In the CHM database, 10,216 compounds
from traditional medicines are collected for virtual screening
(Fakhrudin et al., 2010). The comprehensive Dictionary of Natural Prod-
ucts (DNP) is a commercially available subset library of the Chapman &
Hall/CRC Chemical Database. In 2012, ZINC announced the availability of
the ZINC Natural Product Like database. In the ZINC, also commercially
Table 3
Approaches to select plant material for natural product drug discovery.
Approach Characteristics Examples
Random approach Random selection of extracts from different plant species, enriched
fractions, or isolated natural products, mainly on the basis of their
availability.
Gyllenhaal et al. (2012),Khafagi and Dewedar (2000),Nielsen et al.
(2012),Oliveira et al. (2011),Shaneyfelt et al. (2006),Spjut (1985),
Yuan et al. (1991)
Ethnopharmacological
approach
Selection of the test samples based on traditional medicinal applications
of the plant species.
Atanasov et al. (2013a),Ekuadzi et al. (2014),Fakhrudin et al. (2014),
Noreen et al. (1998),Siriwatanametanon and Heinrich (2011)
Chemosystematic
approach
Selection of the test samples based on chemotaxonomy and phylogeny
taking into account that plant species from some genera or families are
known to produce compounds or compound classes associated with a
certain bioactivity or therapeutic potential.
Alali et al. (2005),Alali et al. (2008),Cook et al. (2014),Gunawardana
et al. (1992),Harinantenaina et al. (2008),Prasain et al. (2001),
Rønsted et al. (2008)
Ecological approach Selection of test samples based on the interactions between organisms
and their environment, considering that plant secondary metabolites
possess ecological functions from which a potential therapeutic use for
humans can be derived.
Brantner et al. (2003),Coley et al. (2003),Egan and van der Kooy (2012),
Krief et al. (2006),Mans et al. (2000),Nacoulma et al. (2013),Obbo et al.
(2013),Thoppil et al. (2013)
Computational
approach
Selection of test samples relying on in silico bioactivity predictions for
constituents of certain plant species.
Atanasov et al. (2013b),Fakhrudin et al. (2010),Grienke et al. (2014),
Grienke et al. (2011),Guasch et al. (2012),Pfisterer et al. (2011),
Sathishkumar et al. (2013),Waltenberger et al. (2011),
Zhao and Brinton (2005)
1591A.G. Atanasov et al. / Biotechnology Advances 33 (2015) 1582–1614

available and literature-derived natural compound databases are pro-
vided for download and virtual screening (Znplike; http://zinc.
docking.org/browse/subsets/special).
3.2. Considerations regarding the choice of bioassays
Drug discovery from plants requires a multidisciplinary approach in
which the success is largely dependent on a well-chosen set of in vitro
and in vivo assays. The choice of the bioassays, first and foremost deter-
mined by the study objectives, should optimally combine simplicity
with good sensitivity and reproducibility.
Historically, the investigation of plant-derived substances was based
on a forward pharmacology approach using in vivo animal tests, organ
or tissue models, or bacterial preparations, followed by in vitro investi-
gation of mechanistic underpinnings. In the more recent past, the ap-
proach of investigating plant-derived substances changed and is now
usually starting with screening of large collections of plant-derived
compounds (“libraries”) against pre-characterized disease-relevant
protein targets, with the aim to identify “hits”, compounds with the
desired activity that are then further studied in relevant in vivo models
with the aim to validate them (a reverse pharmacology approach). Both
the forward and the reverse pharmacology approaches use an overlap-
ping selection of bioassays but differ in the stage when the assays are
applied (Fig. 3;(Lee et al., 2012; Schenone et al., 2013; Takenaka,
2001; Zheng et al., 2013)).
The forward pharmacology, also known as phenotypic drug discov-
ery, first determines functional activity by detecting phenotypic chang-
es in complex biological systems and then characterizes the molecular
target of the active compounds. This traditional way of drug discovery
was carried out mainly in the era before the Human Genome Project,
and especially before the development of many of the modern molecu-
lar biology techniques.
The reverse pharmacology, also known as target-directed drug
discovery, starts by identifying a promising pharmacological target
against which compounds are screened and then obtained promising
hit compounds are validated in vivo. Both unbiased (random) com-
pound libraries as well as knowledge-based libraries (see Section 3.1)
can be used with this approach. While the reverse pharmacology ap-
proach has the advantage of reduced animal testing, its disadvantage
is that it often requires a huge amount of time and effort for the initial
stages, without a guarantee for in vivo efficacy.
It must be noted that in the medicinal plant research community in
the last decade the term “reverse pharmacology”,isalsousedsometimes
to designate “bedside-to-bench”or “field to pharmacy”strategies starting
with clinical efficacy data followed by in vivo and in vitro mechanistic
studies (Aggarwal et al., 2011; Graz, 2013; Patwardhan and Vaidya,
2010; Vaidya, 2006). However, this terminology use is conflicting with
the mainstream understanding existing in the broad drug discovery
scientific community, which would see the “bedside-to-bench”strategy
as a classical forward pharmacology (phenotypic drug discovery)
example starting with observation of phenotypic changes at organismal
level [e.g., reversal of disease symptoms in patients; (Lee et al., 2012;
Schenone et al., 2013; Takenaka, 2001; Zheng et al., 2013)]. In
this sense, terminology standardization is needed in order to avoid
miscommunication between researchers from different scientific
disciplines.
The importance of a proper selection of the initially used pharmaco-
logical assay is underlined by the fact that lack of clinical efficacy
(indicative for inappropriate pre-clinical models) is among the most im-
portant reasons for failure of novel drugs during development (Kola and
Landis, 2004; Schuster et al., 2005). In this section we present a brief over-
view of assays used to determine bioactivities of phytochemicals on the
protein-, cell-, or organism level. For further information on the topic,
we refer the reader to the following excellent reviews: (Agarwal et al.,
2014; Butterweck and Nahrstedt, 2012; Fallarero et al., 2014; Wang
et al., 2011a). Methods for activity assaying that are based on simple
chemical reactions, such as some widely used methods to determine in
vitro antioxidant properties (e.g., the 2,2-diphenyl-1-picrylhydrazyl
(DPPH) radical scavenging assay), have been reviewed elsewhere
(Gulcin, 2012; Moon and Shibamoto, 2009) and will not be discussed in
detail here.
Assays to detect bioactivities of natural products comprise in vitro
models with purified proteins, cell-based target-oriented or phenotypic
assays, models with isolated tissues or organs, and in vivo preclinical
Fig. 2. (A) Two molecules of magnolol concomitantly occupying the binding site of PPARγ(PDB 3R5N) are shown, with the chemical interaction pattern that defines the activity of the
moleculesdepicted. Yellow spheres represent hydrophobic interactions, red andgreen arrows mark hydrogen bond acceptor and donor atoms. This interaction pattern may be converted
into a structure-based pharmacophore model and usedfor virtual screening. (B) The empty binding pocket of PPARγis shown, which can be used in docking simulations to place new
molecules into the binding site and to calculate the binding free energy of the ligand.
1592 A.G. Atanasov et al. / Biotechnology Advances 33 (2015) 1582–1614

animal models. These methods mainly differ in their complexity and
throughput capacity and exhibit some advantages and disadvantages,
as covered in Table 4.
Classical protein-based in vitro assays rely on the measurement of
the functional activity of the investigated target protein in the presence
of a test compound, or of the physical interaction of the test compound
with the target protein. This class of assays can be usually performed in
any general-purpose laboratory without the need for cell culture or an-
imal facilities and are well suited for HTS. A recent evaluation of the
drugs that were approved by the FDA during the past three decades
(Rask-Andersen et al., 2011) revealed that the biggest group of protein
targets of approved drugs so far are receptors (193 proteins targeted
by 563 approved drugs), followed by enzymes (124 proteins targeted
by 234 drugs), transporter proteins (67 proteins targeted by 181
drugs), and other types of targets (51 types, targeted by 84 drugs).
The best represented target classes were the G-protein coupled recep-
tors (GPCRs) in the receptor group, the hydrolases in the enzyme
group, and the voltage-gated ion channels in the transporter protein
group (Imming et al., 2006; Rask-Andersen et al., 2011).
Observed inhibition in in vitro protein assays often reflects binding of
the test compound to the active center of the target protein, thereby
blocking it. In other cases, the assay might be designed to reflect the
inhibition of a protein–protein interaction required for the functional
activity, or protein activation induced upon compound binding. While
these assays inherently provide a “mechanism of activity”of the identi-
fied inhibitor or agonist, they cannot guarantee its functionality in more
complex biological systems such as cells and multicellular organisms. As
a result, many promising “hits”fail when further tested in cell-based in
vitro assays or in vivo animal experiments. Peculiarly, in spite of the cur-
rent focus of the pharmaceutical industry on target-oriented in vitro
screening, analysis of the first-in-class new molecular entities (NMEs)
approved by the FDA between 1999 and 2008 revealed that the majority
of them was first discovered using phenotypic assays (28 compounds
compared to 17 identified by target-based approaches). The surprising
result of this analysis led the authors to postulate that a target-centric
approach for first-in-class drugs may significantly contribute to the cur-
rent high attrition rates and low productivity in pharmaceutical re-
search and development (Swinney and Anthony, 2011). In spite of
this, it should be noted that assays with purified proteins have been ex-
tensively and on many occasions very successfully employed in many
different drug discovery programs e.g., in the identification of selective
inhibitors of various GPCRs and kinases (Alkhalfioui et al., 2009;
Cohen and Alessi, 2013; Zaman et al., 2003).
Assays with cultured mammalian cells are a relatively simple and
inexpensive alternative to in vivo assays for the initial assessment of phar-
macological activity. They are also applicable for bioactivity-guided isola-
tion of natural products from plant extracts. Compared to protein-based
in vitro assays, they provide bioactivity information at the cellular level.
Assays with cultured mammalian cells are mostly used in academic insti-
tutions for low to medium throughput screening. However, this type of
assays is also amendable for up-scaling to screen very large compound li-
braries, and in this form is still widely used primarily, but not exclusively
(Larsen et al., 2014), by the pharmaceutical industry.
The broad spectrum of available cell-based assays requires a careful
and meaningful selection of the assay design, depending on the aim of
the investigation and the intended throughput. The type of process
that is studied often dictates the selection of the cell type that needs
to be used. For example, endothelial cells are selected to study
angiogenesis; epithelial cells might be used for dermatological research.
Furthermore, cell-based assays might be target-oriented or phenotypic
(Table 4). Target-oriented assays are providing information on the in-
terference of investigated compounds with the function of a specific
protein or pathway [e.g., (Atanasov et al., 2013b; Orlikova et al., 2013;
Reuter et al., 2009)], whereby phenotypic models are providing infor-
mation on a changed cellular phenotype with a complex regulation
[e.g., cell proliferation (Kurin et al., 2012; Schwaiberger et al., 2010)].
Fig. 3. The forward pharmacology and reverse pharmacology approaches in natural product-based drug discovery.
1593A.G. Atanasov et al. / Biotechnology Advances 33 (2015) 1582–1614

Cell-based assays might be performed using primary mammalian cells
or cell lines. Although immortalized cell lines are easier to maintain in
culture, often results obtained with them are less relevant, because
immortalization and prolonged in vitro maintenance allows accumula-
tion of mutations and phenotypic changes. Used cells or cell lines can
also be derived from genetically modified transgenic animals. Other
approaches resort to engineering of cells in vitro to stably overexpress,
stably down-regulate, or knockout a gene product of interest (Ketting,
2011; Wang et al., 2014c). The spectrum of cell-based assays has
been widened by engineering of cells enabled to act as reporters
(e.g., luciferase- or fluorescence-based) of a specific intracellular response
influenced by the tested substance (Baird et al., 2014; Fakhrudin et al.,
2014). Aside of the use of reporter systems, a variety of phenotypic
parameters have been successfully selected as readout, e.g., changes in
cell morphology, cell adhesion, cell proliferation, migration, differentia-
tion status, metabolic status, redox status, cell apoptosis or senescence
(Atanasov et al., 2013b; Blazevic et al., 2014; Gaascht et al., 2014; Zheng
et al., 2013). Phenotypic cell-based assays can be used to verify the
activity of compounds identified by protein-based in vitro assays at the
cellular level. Moreover, cell-based phenotypic models might also be
used to study the underlying molecular mechanisms of certain biological
effects, possibly leading to the discovery of new target molecules or
pathways affecting the respective phenotype. For this purpose, changes
in intracellular signaling or gene expression are often characterized by
techniques such as immuno-cytochemistry, FACS-analysis, real-time
PCR, Western blotting, immunoprecipitation, or “omics”-techniques
(e.g., genomics-, transcriptomics-, proteomics- and metabolomics-
techniques) for characterization of global changes in gene expression or
metabolite quantities (Ladurner et al., 2012; Lee and Bogyo, 2013;
Schenone et al., 2013; Schreiner et al., 2011; Ziegler et al., 2013).
Besides mammalian cells, also yeasts have been employed for the
establishment of whole-cell phenotypic assays with applications in
drug discovery, e.g., for high throughput functional screening based on
the activation of caspases and other proteases involved in cell death
and inflammation (Hayashi et al., 2009), as a screening tool for pharma-
cological modulation of GPCRs (Minic et al., 2005), but also to under-
stand the mechanism of action of drugs or to identify novel drug
targets and target pathways (Hoon et al., 2008).
A special set of assays that stay on the interface of in vitro and in vivo
models encompasses in situ and ex vivo methods with isolated tissues or
organs (Luo et al., 2013; Teicher, 2009). They show the advantage of
usually resembling the in vivo situation more closely than the in vitro
tests, but also the disadvantages of lower throughput, more difficult re-
supply, ethical concerns related to the use of animals, andthe short half-
life of the isolated tissues and organs. Into this group of methods also
falls the recently described technology Ex Vivo Metrics™that uses intact
human organs ethically donated for research (Curtis et al., 2008). While
most of the mentioned limitations of ex vivo models are applicable also
to the Ex Vivo Metrics™technology, its strong advantage is that it elim-
inates potential species differences andoffers theclosest applicable bio-
logical system in terms of emulating human exposure to drugs prior to
clinical trials (Curtis et al., 2008).
In summary, even though the relevance of in vitro assays is limited
due to the inability to provide information on factors influencing the
activity of compounds in vivo (such as ADME/T), in vitro assays still
represent very important tools to identify and characterize bioactive
compounds.
Following the reverse pharmacology approach (Fig. 3), compounds
identified with a good activity in vitro need to be tested in vivo in
suitable animal models that can provide basic pharmacological and
toxicological data prior to subsequent human clinical trials.
Traditionally, in part due to the reasonably high homology and
similarity between mammalian genomes and physiology, and to the
relatively short reproductive cycle, mouse and rat models were used
for the assessment of the activity of plant extracts or isolated natural
products (Butterweck and Nahrstedt, 2012). Such animal models are
still crucial for drug evaluation and validation, as they providethe inves-
tigator with an integrated response encompassing efficacy, bioavailabil-
ity, side effects, and toxicity (ADME/T parameters) of the drug in a
whole organism, and are standardly used for pharmacokinetic and safe-
ty studies that are a pre-requisite for human clinical trials (Alqahtani
et al., 2013; Eddershaw et al., 2000; Parasuraman, 2011; Zhang et al.,
2012). Non-rodent mammalian species such as rabbits, dogs, swine,
and monkeys are also widely used in pharmacological safety and phar-
macokinetic studies (Parasuraman, 2011; Pellegatti, 2013; Swindle
et al., 2012; van der Laan et al., 2010; Zhang et al., 2012). Although
such large animal models are associated with limitations, such as higher
price and more pronounced ethical considerations, their use is still
widely practiced, especially by the pharmaceutical industry, since the
regulatory guidelines of FDA, European EMA, and other similar interna-
tional and regional authorities usually require safety testing in at least
two mammalian species, including one non-rodent species, prior to
human trials authorization (Parasuraman, 2011; van der Laan et al.,
2010). Indeed, rodent and human proteins are sometimes not similarly
Table 4
Basic types of bioassays for testing of phytochemicals and their comparative advantages and disadvantages.
Type of bioassay Advantages Disadvantages
In vitro assays with
purified proteins
High throughput; no cell culture or animal facilities necessary. Prone to irrelevant hits (compounds with low bioavailability unable to
reach the respective target in intact cells or in vivo</