Locked Nucleic Acid: Modality, Diversity, and Drug Discovery

Home , Locked nucleic acid, Oligonucleotide

Downloaded from orbit.dtu.dk on: Sep 23, 2021

Locked nucleic acid: modality, diversity, and drug discovery

Hagedorn, Peter H.; Persson, Robert; Funder, Erik D.; Albæk, Nanna; Diemer, Sanna L.; Hansen, Dennis J.; Møller, Marianne R; Papargyri, Natalia; Christiansen, Helle; Hansen, Bo R. Total number of authors: 13

Published in: Drug Discovery Today

Link to article, DOI: 10.1016/j.drudis.2017.09.018

Publication date: 2018

Document Version Publisher's PDF, also known as Version of record

Link back to DTU Orbit

Citation (APA): Hagedorn, P. H., Persson, R., Funder, E. D., Albæk, N., Diemer, S. L., Hansen, D. J., Møller, M. R., Papargyri, N., Christiansen, H., Hansen, B. R., Hansen, H. F., Jensen, M. A., & Koch, T. (2018). Locked nucleic acid: modality, diversity, and drug discovery. Drug Discovery Today, 23(1), 101-114. https://doi.org/10.1016/j.drudis.2017.09.018

General rights Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.

 Users may download and print one copy of any publication from the public portal for the purpose of private study or research.  You may not further distribute the material or use it for any profit-making activity or commercial gain  You may freely distribute the URL identifying the publication in the public portal

If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim.

Drug Discovery Today Volume 23, Number 1 January 2018 REVIEWS

Teaser LNA-modified antisense oligonucleotides (LNAs) are widely used in RNA therapeu-

tics. The structural diversity of LNAs affects most drug properties. Exploiting this diversity

offers new opportunities for discovering LNA-based drugs.

KEYNOTE REVIEW

Locked nucleic acid: modality,

diversity, and drug discovery Reviews

1 2 1 Peter Hagedorn has a

Peter H. Hagedorn , Robert Persson , Erik D. Funder , MSc in theoretical physics

1 1 1 and biophysics from the

Nanna Albæk , Sanna L. Diemer , Dennis J. Hansen , Niels Bohr Institute at the

1 3 University of Copenhagen,

Marianne R. Møller , Natalia Papargyri , and a PhD in molecular

1 1 1 biology and bioinformatics

Helle Christiansen , Bo R. Hansen , Henrik F. Hansen , from Ris½ National

1 1 Laboratory, Denmark. For

Mads A. Jensen and Troels Koch the past 6 years, Peter has been exploring the

structural diversity of LNA-modiﬁed antisense

Therapeutic Modalities, Roche Pharma Research and Early Development, Roche Innovation Center Copenhagen, oligonucleotides and how this impacts measured

properties, with a particular focus on developing

2970 Hørsholm, Denmark

predictive mathematical models that capture this

Pharmaceutical Sciences, Roche Pharma Research and Early Development, Roche Innovation Center

diversity.

Copenhagen, 2970 Hørsholm, Denmark

Department of Biotechnology and Biomedicine, Technical University of Denmark, 2800 Kgs. Lyngby, Denmark Robert Persson has a

PhD in microbiology and

biochemistry from the

University of Lund,

Over the past 20 years, the ﬁeld of RNA-targeted therapeutics has

Sweden. He has worked

advanced based on discoveries of modiﬁed oligonucleotide chemistries, with the intracellular

transport of secretory

and an ever-increasing understanding of how to apply cellular assays to

proteins and viral

identify oligonucleotides with improved pharmacological properties membrane proteins and

with the pharmacokinetics of proteins and

in vivo. Locked nucleic acid (LNA), which exhibits high binding afﬁnity oligonucleotides for more than 25 years. During the

past 8 years, Robert has focused on the bioanalysis,

and potency, is widely used for this purpose. Our understanding of RNA

biodistribution, and pharmacokinetics of LNA

biology has also expanded tremendously, resulting in new approaches to oligonucleotides.

engage RNA as a therapeutic target. Recent observations indicate that each Troels Koch has a PhD in

bioorganic chemistry from

oligonucleotide is a unique entity, and small structural differences the University of

Copenhagen, Denmark. He

between oligonucleotides can often lead to substantial differences in their

has worked in the area of

pharmacological properties. Here, we outline new principles for drug nucleic acid chemistry and

biology for 20 years. He co-

discovery exploiting oligonucleotide diversity to identify rare molecules founded Santaris Pharma

A/S, which pioneered the

with unique pharmacological properties.

locked nucleic acid (LNA) platform and LNA

therapeutics. Santaris was acquired by Roche in

August 2014, and he is presently vice president and

head of research at the Roche Innovation Centre in

Introduction Copenhagen. His main responsibilities are to develop

the chemical and biological properties of LNA,

The ﬂow of genetic information is a highly complex process. Of the approximately 43 000

improve and upgrade the LNA platform, reﬁne LNA

human genes currently annotated by the Ensembl project (release 89), approximately 47% antisense drug discovery processes, and establish a

RNA therapeutics drug pipeline in Roche.

encodes proteins, and the rest are all transcribed into different types of noncoding RNA (ncRNA)

[1]. It is clear that these ncRNAs represent a long and growing list of different types of RNA with

various functional roles and regulatory interactions [2,3]. This knowledge has provided a

stronger basis for the speciﬁc annotation of the relationships between structure, type, and

function of RNAs and human disease [4]. This is one reason for the increased interest in RNA

Corresponding author: Koch, T. ([email protected])

1359-6446/ã 2017 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). https://doi.org/10.1016/j.drudis.2017.09.018 www.drugdiscoverytoday.com 101

REVIEWS Drug Discovery Today Volume 23, Number 1 January 2018

therapeutics, which focuses not only on classic protein-coding Chemistry sets boundaries for success

mRNA intervention, but also on how the manipulation of non- In the ﬁrst generation of single-stranded therapeutic oligonucleo-

coding regulatory RNA can provide novel drugs for previously tides, the phosphorothioate (PS) internucleoside linkage was the

untreatable diseases. only chemical modiﬁcation [8,21–24]. The PS modiﬁcation

A rational strategy to target RNA is by synthetic oligonucleo- replaces one of the nonbridging oxygen atoms with sulfur in

tides. Although RNA exists in a complex biological matrix, en- the internucleoside phosphate (Fig. 1ai). This modiﬁcation creates

Reviews dogenous enzymatic RNA-processing mechanisms can be a chiral center at phosphorous producing two isomers: Rp and Sp.

exploited and recruited by the heteroduplex formation between Thus, for every PS linkage introduced in an oligonucleotide, two

RNA and synthetic oligonucleotides. RNA exhibits high turnover diastereoisomers are formed. Given that conventional solid-phase

ENT REVIEW KEYNOTE

rates and is more labile than DNA [5]; thus, inhibitory effects PS synthesis is not stereoselective, an n-mer PS oligonucleotide

n–1

induced by the oligonucleotide binding complement can be contains random mixtures of 2 diastereoisomers [25–30]. How-

executed rapidly and effectively. Zamecnik and Stevenson [6] ever, recent developments have now made it possible to synthesize

were the ﬁrst to show that a single-stranded synthetic DNA individual stereoisomers with high stereoselectivity [31–34], and

oligonucleotide could inhibit the expression of RNA, resulting with good yields [35–37]. Diastereoisomers are different chemical

in the stimulation of substantial research and development ac- compounds and the diversity of properties for a given PS oligonu-

tivity [7–13]. cleotide will be a gradient spanning from a little to a very large

The past four decades of research and development in oligo- difference in property. Generally, oligonucleotides with a Sp

nucleotide-based RNA therapeutics has produced groundbreak- conﬁguration provided better exonuclease resistance compared

ing results. The recent successes are based on improvements in with oligonucleotides with a Rp conﬁguration, whereas Rp iso-

three essential parameters for antisense technologies: (i) medici- mers were better substrates for DNA-dependent RNA poly-

nal chemistry efforts have devised much-improved nucleic acid merases, RNase H, and stimulated immune responses [38–40].

modiﬁcations for antisense oligonucleotides (AONs), of which Compared with the random mixtures, Rp oligonucleotides exhib-

the high-afﬁnity LNA [14–17] is among the most widely used; (ii) ited higher melting temperature (Tm) against RNA, whereas Sp

the knowledge revolution of the functions and biological molec- analogs had lower Tm. In the ﬁrst-generation oligonucleotide

ular mechanisms of RNA; and (iii) the translational drug discov- drugs, the PS modiﬁcation was normally not stereocontrolled.

ery part (i.e., discovery processes and models used in drug Except for rare examples [41], this class of drugs has not provided

discovery). The latter relates to bioinformatics drug design, pre- an adequate TI for AONs [42]. Two reasons for this were that the

diction algorithms, in vitro/vivo assays, better pharmacokinetics/ PS modiﬁcation decreases the Tm of the hybrid duplex and,

pharmacodynamics (PK/PD) models, absorption, distribution, although it protects against nuclease degradation to some extent,

metabolism, and excretion (ADME) models, and toxicology study this protection was still not adequate on its own for most thera-

designs. peutic purposes [43]. Still, the PS modiﬁcation, stereocontrolled

Recent developments have also provided insights into the fun- or not, turned out to be a central modiﬁcation for later genera-

damental properties of the chemical modality. Small chemical tions. Combined with the increased nuclease stability, PSs also

modiﬁcations of AONs have been shown to produce large property contribute to improved PK and cellular uptake properties of

differences even among AONs belonging to the same chemical or AONs. Unless otherwise stipulated, all antisense data described

structural class. The old concept of the ‘portability of chemistry’ here come from fully PS-modiﬁed AONs.

for AONs, whereby nearly all properties are shared for a given A preferred strategy to improve the binding afﬁnity of AONs to

chemical class, has turned out not to be true for many important RNA was to introduce electronegative substituents in the 2 -posi-

drug properties (e.g., activity and toxicity). In other words, we now tion of the furanose (Fig. 1aii) [7,8,44–47]. For nucleic acid analogs

know that the structure–activity relationships (SARs) for AONs are of this structural class, melting temperatures increased by approx-

more subtle. The fact that AONs belonging to a given structural imately 0.5–1.5 C/modiﬁcation against RNA. The most signiﬁ-

0 0 0

class only on a few parameters share the same properties and act cant members of this class are the 2 -F, 2 -O-CH3 and 2 -O-

more as individual unique compounds diversiﬁes and complicates CH2CH2-O-CH3 (MOE) substituent groups [44,46]. This second

drug discovery [13,18–20]. However, the fact that small chemical generation of drugs shows clear improvements compared with

modifications, even the configuration of a single bond, can have first-generation PS drugs, and two drugs of this type, mipomersen

profound property impacts creates a basis for SAR studies compa- [48] and nusinersen [49], have now been approved by the US Food

rable to classic small-molecule drug discovery. Therefore, during and Drug Administration (FDA).

the discovery phase, it is now possible to produce more unique A high point for synthetic nucleic acid analogs was reached in

compounds for nearly any RNA target with larger therapeutic 1997 with LNA [15,50] (Fig. 1aiii). LNA exhibited unprecedented

indexes (TI). high RNA-binding afﬁnity [51], providing melting temperature

In this review, we focus on three aspects of LNA: (i) state of the increases of 3–9 C/modiﬁcation depending on the oligonucleo-

art of the modality, including structure, fundamental properties, tide position and design complements [52]. The afﬁnity increase

preferred designs, mechanisms of action, and cellular uptake; (ii) per LNA nucleoside substitution is a reproducible (‘portable’)

illustrate how subtle structural modiﬁcations have profound property [14,51–55] and, combined with the increased nuclease

impacts and, in some cases, produce ‘all or none’ phenotypic stability, the driver behind the many therapeutic and diagnostic

effects; and (iii) how the insight or ‘Erkenntnis’ of property diver- life-science applications that have been demonstrated with, and

sity can be used to transform and improve the classic oligonucle- published on, LNAs [56–63]. Given these signiﬁcant improve-

otide drug discovery process and lead generation. ments, LNA is now widely used and acknowledged by many to

102 www.drugdiscoverytoday.com

Drug Discovery Today Volume 23, Number 1 January 2018 REVIEWS

(a) (b) 5'- - 3' Gapmer Rp Sp

5'- - 3' Mixmer ⇔

LNA DNA PS IIIIII IV V

(c)

Cytoplasm Nucleus miRNA gene

DNA KEYNOTE REVIEW Transcription and cropping Transcription pre-mRNA Post-transcriptional Reviews

modification

Pre-miRNA

5' cap 3' poly(A) tail 4 Inhibition 5 Inhibition of 2 Activation of 5' cap RNA splicing of RNase H Export, AGO loading, Mature mRNA AAAAA formation and unwinding

AGO Hybridization Gapmer Mixmer Mature miRNA

AAAAA AAAAA AAAAA mRNA1 Formation of AON-mRN A Mixmer Translation heteroduplex AAAAA mRNA2

Target protein 60S 60S ribosomal AAAAA AAAAA AAAAA subunit AAAAA AAAAA RNase H 40S ribosomal 40S

AAAAA AAAAA subunit

7 Derepression of protein 6 Repression of protein 1 Normal protein 2 Activation of RNase H 3 Steric hindrance of translation translation translation ribosomal subuni t binding

Drug Discovery Today

FIGURE 1

Chemistry of antisense oligonucleotides and mechanisms of action on RNA. (a) First-generation phosphorothioates (PS) (i), second-generation 2 -O-substituted

ribofuranoses (ii), third-generation locked nucleic acid (LNA) illustrated as non-stereodefined phosphorothioates (iii), the bicyclic structure of LNA (iv), and in (v)

the two stereodefined diastereoisomers of the internucleoside phosphorothioate (Rp and Sp). (b) Gapmer design (top) in which consecutive LNA segments (dark

orange) are flanking a central DNA segment (yellow), and mixmer design (bottom), where the nucleotides are ‘mixed’ throughout the compound. Nearly all

designs contain PS internucleoside linkages (red line). The gray line below each nucleoside indicates the potential for forming hydrogen bonds with

complementary nucleobases in other nucleosides via Watson–Crick base-pairing. (c) Mechanisms of action by which LNA oligonucleotides can intervene with

and inhibit RNA function. Adapted from Ref. [141].

belong to a third generation of chemical modiﬁcations used in The difference between LNA and 2 -OMe RNA, for example,

RNA therapeutics. can be observed in the helical structures that they form. Recent

‘LNA’ refers to the bicyclic structure in which the ﬂexibility of structural data indicate that LNA/RNA duplexes perturb the

the parent furanose has been locked [15] (Fig. 1aiv). The bicyclic A-type helix, whereas 2 -OMe RNA/RNA attain a more classic

structure of LNA nucleosides is the key to the improved binding A-type structure [76]. LNA/LNA duplexes result in an extreme

properties, and has served as scaffold for many other bicyclic ‘underwinded’ helix in which the major grove is increased by a

analogs also exhibiting improved binding properties [64–68]. dramatic decrease of the slide-and-twist values and a pitch value of

0 0

The 2 -O-CH2-4 linkage is transformational and converts a ﬂexible 39 A˚ compared with 29 A˚ for RNA/RNA. There is a consecutive

furanose into a rigid bicycle [69–74]. One thermodynamic conse- order of structural change of the classical A-type helix from

quence of the ‘locking’ compared with native deoxyribose is that RNA/RNA, RNA/LNA to LNA/LNA [76,77].

the free energy of solvation is 1–2 kcal/mol lower and, thus, more Transforming a ﬂexible furanose moiety into a rigid bicycle

0 0

favorable because of the lipophilic nature of the 2 -O-CH2-4 reduces the number of conformational degrees of freedom,

linkage [75]. including rotations around internal bonds [75]. Structurally, this

www.drugdiscoverytoday.com 103

REVIEWS Drug Discovery Today Volume 23, Number 1 January 2018

preorganizes the molecule and reduces the hybridization entropy produces a long half-life of the LNA/miRNA heteroduplex, leading

penalty. Although a net reduction in the entropy change for to long lasting derepression of the proteins that are under the

hybridization (DDS) has been reported as a main driver for the control of the miRNA.

high afﬁnity [78], the situation is more complex. Given that LNA

distorts the classic A-form helix, enthalpic contributions, such as Affinity, potency, and specificity of LNA oligonucleotides

hydrophobic interactions and P–P-stacking between the bases, It is well documented that the high RNA-binding afﬁnity of LNA is

Reviews are also increased [77]. linked to its potency [58,57,99]. The term ‘potency’ is used here for

the activity per administered dose and is not necessarily related to

LNA designs and RNA intervention mechanisms the concentration in cells or in tissues. Although high target

ENT REVIEW KEYNOTE

LNA designs can be divided in two main categories: mixmers and afﬁnity is a driver behind higher potency [14,18], many other

gapmers (Fig. 1b). In a mixmer, LNA and DNA nucleosides (dark- factors are also determinants, and higher afﬁnity can be balanced

orange and yellow boxes, respectively in Fig. 1b) are interspersed out by other potency-limiting parameters. Examples of such lim-

throughout the sequence of the oligonucleotide, whereas, in a iting factors are: design, RNase H recruitment [100], tissue/cellular

gapmer, two LNA segments at both ends of the oligonucleotide are uptake, and/or tissue distribution. For instance, it has been shown

separated by a central segment or gap of DNA nucleosides. At that a fully LNA-modiﬁed 14-mer targeting the coding region of a

present, LNA designs are usually made with fully modiﬁed non- transcript did not inhibit protein synthesis [84,101], because the

stereodeﬁned PS internucleoside linkages (red lines in Fig. 1b). ribosome ‘read through’ the binding site of the LNA, despite the

Unless otherwise speciﬁed, all data mentioned here are obtained high RNA afﬁnity. When some of the central LNA nt were substi-

from such LNA PS random mixtures. tuted with DNA-nt so that the 14-mer became a LNA gapmer, the

To inhibit mRNA expression and protein translation (Fig. 1c, gapmer recruited RNase H and mediated cleavage and degradation

mechanism 1), gapmer designs are the most potent. This is because of the message, despite the lower afﬁnity [84].

the central DNA/PS segment, which is longer than 7–8 DNA It has also been shown that afﬁnity increase by length

nucleotides (nt), recruits the RNA-cleaving enzyme RNase H when expansion of gapmers only improved the potency to a certain

the gapmer is hybridized to the mRNA [79–84]. Gapmers targeting compound-speciﬁc point/threshold, after which the potency de-

both exons and introns work equally efﬁciently and, thus, all creased as the length, and afﬁnity, increased [99]. This ‘threshold

stages of mRNA and also ncRNA processing can be addressed. afﬁnity’ was studied further by Pedersen et al. using a kinetic model

Nuclease recruitment by gapmers occurs in both the cytoplasm to simulate LNA gapmer recruitment by RNase H [100]. The model

and the nucleus [85,86], although predominantly in the nucleus showed very fast LNA–RNA hybridization on-rate and recruitment

[80] (Fig. 1c, mechanism 2). Mixmers can also be used to block of RNase H. IC50 curves were generated as a function of LNA

translation, but must be designed to bind close to, at, or upstream concentrations and, when these were plotted as a function of

of the translational start site to do so [83,84] (Fig. 1c, mechanism LNA–RNA afﬁnity, a parabola emerged. This illustrated in afﬁnity

3). The mixmer will inhibit translation by blocking the binding of terms that potency increased as the free energy of hybridization

the ribosomal subunits to the mRNA. Alternatively, mixmers decreased to approximately DG = 20–25 kcal/mol, but below

0 0

binding close to the 5 end of the pre-mRNA can prevent 5 -cap this point, and for compounds with even higher afﬁnity, the effect

formation and thereby inhibit translation (Fig. 1c, mechanism 4). reversed and potency decreased. The existence of such ‘optimal’

There is sometimes the need to intervene with RNA processing afﬁnity was explained to be a tradeoff of initial efﬁcient target

and information ﬂow without cleaving or degrading the target, or binding coupled with inefﬁcient deassociation of the cleaved

inhibiting protein translation. For instance, in the case of splice LNA–RNA duplex. Nonspeciﬁc sequestration of LNAs through

switching/redirection, or blocking natural antisense or other kinds protein binding might also be length dependent because of

of ncRNA, mixmers are often the preferred option [87–93] (Fig. 1c, the negative charge contributed by each PS linkage, and might

mechanism 5). A mixmer cannot support RNA cleavage but acts as also contribute to the observed length-dependent potency of LNAs

an efﬁcient steric block to mediate a phenotype without destroy- [99].

ing the target RNA. This intervention changes the expression It is a cornerstone in RNA-based drug discovery to ensure that

proﬁle, creating potentially both ‘loss-of-function’ and ‘gain-of oligonucleotides interact speciﬁcally with their intended RNA

function’ phenotypes [91–93]. targets. For LNA gapmers that bind with high afﬁnity to fully

Inhibition, or sequestering, of miRNA is a gain of function that complementary RNA target regions, extra care must be taken to

has attracted a special interest [94–97]. miRNAs are an important ensure that they do not bind efﬁciently to partially mismatched,

class of regulatory ncRNAs that can bind to partially complemen- unintended RNA targets [18]. Although the mismatched base pairs

tary sites located in the 3 untranslated regions (UTRs) of target in the unintended RNAs reduce the binding afﬁnity compared

mRNAs, promoting translational repression or deadenylation with the fully matched target, it has been reported that, in some

and degradation (Fig. 1c, mechanism 6). Sequestering miRNAs instances, afﬁnity of LNAs can be designed so high that duplexes

by LNA mixmer hybridization has wide therapeutic potential between mismatched unintended RNAs and gapmers persist long

[98] (Fig. 1c, mechanism 7). The miR-122-antagonizing LNA mix- enough, and in sufﬁcient amounts, to allow effective RNase H-

mer ‘miravirsen’ was progressed to Phase 2 clinical trials and cleavage [18,102–104]. However, character- and thermodynamics-

exhibited safe and potent HCV viral titer reductions [97]. The based bioinformatics algorithms allow for computational gapmer

high binding constant of LNA mixmers produces a slow off-rate, sequence analysis where the likelihood of binding to mismatched

which is the critical parameter for efﬁcient miRNA antagonism. unintended RNA can be predicted. Thus, bioinformatics enable

Combined with the long tissue half-life of LNA, the slow off-rate the selection of sequence-speciﬁc oligonucleotides designed to

104 www.drugdiscoverytoday.com

Drug Discovery Today Volume 23, Number 1 January 2018 REVIEWS

have either no or only a few mismatched hybridizations. The to GalNAc, the high turnover of ASGR will internalize recep-

sequence speciﬁcity of gapmers can also be evaluated experimen- tor-GalNAc-oligonucleotide efﬁciently via clathrin receptor-me-

tally using transcriptome-wide proﬁling approaches, such as diated endocytosis. GalNAc-conjugated oligonucleotides

microarrays or RNA-seq [18]. Taken together, these computational typically provide 10–20-fold reductions in the dose where half

and experimental methods are important tools to identify com- maximum responses were seen, and have revolutionized hepatic

pounds with high sequence speciﬁcity and low off-target potential RNA targeting [108].

[105]. It is possible that the major uptake observed from the endoso-

mal/MP pathways operates together with other uptake mecha-

Cellular uptake nisms. In this way, high uptake/low productivity pathways

Entry of oligonucleotides into cells is a complex process and a (endocytosis) might operate in parallel with low uptake/high

comprehensive report is beyond the scope of this review. Here, we productivity pathways. For example, it has been shown that

focus on central elements that are relevant for LNAs, but the nucleic acid channels exist in rat kidney and brain cells [125], KEYNOTE REVIEW

interested reader may consult recent reviews for more information although it has not been demonstrated that this is a general uptake

[106–109]. mechanism. Cellular uptake is a highly diverse and differentiated

When oligonucleotides are transfected with cationic lipids, the process that is dependent on many parameters, such as cell type,

Reviews

lipids fuse with the plasma membrane and deliver oligonucleo- proliferation/cell cycle state, extracellular composition (both in

tides directly into the cytoplasm. However, the oligonucleotides vitro and in vivo), oligonucleotide design/substitutions pattern,

concentrate quickly in the nucleus, which is advantageous for concentration, and phosphorothioate conﬁguration (see below).

antisense because RNase H is predominantly found in the nucleus

[80]. When oligonucleotides are incubated without added trans- Diversity: an inherent attribute for oligonucleotides

fection agents, ‘naked’ uptake also occurs [106,110–112]. It was Structural diversity

illustrated that ‘naked’ LNAs were taken up and exhibited potent It is normal practice in the ﬁeld of small-molecule drug discovery

activity in cell cultures [113]. This process was named ‘Gymnosis’ to produce collections or libraries of chemical entities. It is gener-

after the Greek ‘gymnos’ or ‘naked’, and activity was found for ally appreciated that library size is not a goal in itself, but rather the

nearly all cell types, including ‘hard-to-transfect’ cells, such as generation of structural, and thereby functional, diversity, and

suspension cell lines and T cells. Gymnotic silencing provides property reﬁnement [126]. For oligonucleotides, structural differ-

more AON-speciﬁc phenotypes. Given that transfection/delivery ences also deﬁne functional diversities, but because of the partic-

lipids are not used in gymnotic assays, a cellular response is only ular compound class, the diversity framework is derived from

related to, or is a consequence of, the added AON [113–115]. different property determinants. The diversity framework can be

Cellular uptake of naked oligonucleotides and LNA is predomi- divided in ﬁve components or determinants for structural diversi-

nantly driven by endocytosis. Many different pathways and cellu- ty: (i) nucleobase or sequence diversity: the variation in nucleo-

lar proteins are engaged in the uptake, such as clathrin, dynamin, base composition that constitutes oligonucleotide sequence; (ii)

and caveolar-dependent endocytosis [116,117]. However, cla- nucleoside diversity: the variation in chemical structure of the

thrin-independent endocytosis also exists [118], and it has been nucleoside building blocks used to produce a given oligonucleo-

shown that antisense activity in hepatocytes can be blocked by tide; (iii) backbone diversity: the variation in backbone internu-

adaptor protein AP2M1 but not by clathrin and caveolin [119]. cleoside linkage structure, including variations in the chiral

It was recently shown that LNA gapmers were taken up in orientation of any phosphorothioate linkages; (iv) architectural

gymnotic assays in human primary T cells via macropinocytosis diversity: the variation in design and combinatorial nucleotide

(MP) [120]. MP is a ﬂuid-phase endocytosis pathway driven by make-up of an oligonucleotide; and (v) macromolecular structure

actin–myosin cell protrusions creating larger vesicles. The fate of diversity: the variation in structure and topology of the entire

macropinosomes differs from that of other endosomes and they molecule.

are directed toward the late-stage endosomes/lysosomes. Once The ﬁrst four diversity determinants are classic oligonucleotide

taken up, LNAs were released from MPs/late endosomes and parameters. They represent the ‘digital’ Watson–Crick hybridiza-

showed potent and speciﬁc RNA target knockdown. tion properties. The antisense literature has covered these in great

The mechanisms by which oligonucleotides are released from detail over the past 30 years, but it was only recently appreciated

endosomes/macropinosomes are still unclear, but maturation of that the nucleobase sequence is not the only determinant, but that

the endosomes from early to late endosomes (multivesicular bod- the global properties are derived from the position of each struc-

ies) is important [85]. It was also shown the LNAs co-localize with tural unit in the context of the entire molecule. The term

various cellular bodies (GW and P-bodies) and that proteins, such ‘sequence-related properties’ has been used frequently and some

as Ago2, TPC-, Hsc-70, and PKC-a proteins, which bind to endo- properties can be related to a nucleobase sequence motif alone.

somes, are involved in the endosomal maturation process and However, this description is at best inadequate, and can be

probably also in their release [121]. Once oligonucleotides have meaningless without reference to the speciﬁc composition of

escaped from the endosomes, activity occurs in the cytoplasm [85] the molecule.

or, after rapid trafﬁcking, also in the nucleus [122,123]. Macromolecular structure diversity is related to the variation in

One of the most successful uptake enhancers is the triantenn- overall structure and topology of the oligonucleotide. This param-

ary N-acetylgalactosamine (GalNAc) ligand, which targets the eter relates more to the ‘analog’ or aptameric binding properties.

high-density asialoglycoprotein receptor (ASGR) expressed on When the ﬁrst four determinants are ﬁxed, a resulting single

hepatocytes [108,124]. When oligonucleotides are conjugated molecule will be able to attain many different shapes and overall

www.drugdiscoverytoday.com 105

REVIEWS Drug Discovery Today Volume 23, Number 1 January 2018

structures each representing different energy levels. This diversity 100

parameter can be illustrated by modeling, and quantum mechani-

cal (QM) calculations have demonstrated the 3D relations in

molecular structure, topology, and electrostatics of LNA PS oligo-

nucleotides [127,128]. Among all possible macromolecular struc- 40

mRNA (% control)

tures, QM or molecular dynamics calculations will be able to

Reviews identify a speciﬁc minimized structure that is most representative

HIF1A 0

for the molecule [75,127]. Below, we ﬁrst describe functional

Gapmers with same nucleobase sequence but different LNA patterns

diversity for the classic PS random mixtures, and then diversities

Drug Discovery Today

ENT REVIEW KEYNOTE

for stereo-deﬁned PS LNAs.

FIGURE 2

Functional diversity in vitro

Architectural diversity of locked nucleic acid (LNA) gapmers affects target

Human pre-mRNA sequences are typically between 10 000 and 65

knockdown in vitro. LNA gapmers with the same nucleobase sequence but

000 nt in length with a median length of approximately 25 000 nt different LNA design patterns were screened in a gymnotic HeLa cell assay at

[1]. Therefore, tens of thousands of possible sites on a pre-mRNA 5 M, and knockdown of target mRNA was measured after 3 days by qRT-

PCR. The gapmers were targeted to the same 13-nucleotide (nt) region on

can be targeted by gapmers. Libraries of gapmers that are posi-

the human HIF1A mRNA. In all, 32 design patterns were evaluated, written as

tioned at unique and nonoverlapping sites on a RNA target mani-

LxxxDDDDDDxxL (L: LNA, D: DNA, x: either LNA or DNA). All values are shown

fest diversity with respect to their nucleobase composition. It is

as averages across four independent biological replicates. Error bars indicate

widely recognized that gapmer potency is affected by the target the SD.

nucleobase sequence. This is usually attributed to the structure of

the RNA in the target region, where less-structured RNA leads to vitro, the 3–10–3 design had a twofold longer half-life in kidney

more-potent gapmer responses [129]. However, other factors that after intravenous (i.v.) administration to mice [131].

affect potency can also be attributed to the nucleobase sequence, An evaluation of both architectural and nucleoside diversity was

such as sequence-speciﬁc intracellular localization [109]. reported by Stanton et al. [132]. An initial screening campaign

Nucleobase diversity was illustrated in a study of 57 and 71 fully identiﬁed three potent PS LNA-gapmers targeted to the Glucocor-

phosporothioated gapmers (20mers, mixtures of 2 diastereoi- ticoid Receptor mRNA. Each of the three gapmers was of the 3–8–3

somers). The gapmers were designed with ﬁve 2 -O-methoxyethyl design but with completely different nucleobase sequences. The

(MOE) modiﬁcations in each ﬂank (5–10–5 design). A range of authors designed a library of 109 gapmers where they varied

activities was seen when screened in mouse primary hepatocytes lengths between 14 and 20 nt, and modiﬁed ﬂanks with various

0 0

and human T-24 cells [130]. Some gapmers were not active at all numbers of LNA, MOE, 2 -OMe, and 2 F nt. Some gapmers

0 0

and showed no knockdown of the target. Most compounds contained both LNA and 2 -OMe, or LNA and 2 -F in the ﬂanks.

showed some activity and a few in both libraries were judged to In vitro and in vivo potency were observed to be highly sensitive

be potent. to small changes in length, nucleoside composition, and number

In another study, more than 200-fold differences in potency in of modiﬁcations in the ﬂanks. Thus, the study demonstrated

vitro were reported for a library of 47 LNA-phosphorothioate an apparently unpredictable and wide property diversity of

gapmers targeted to various regions of the HCV RNA genome oligonucleotides.

[131]. Most of these nucleobase sequence-diverse gapmers were Another example of how architectural diversity can impact

16mers with four LNAs in each ﬂank (a 4–8–4 design, mixtures of potency in vitro is shown in Fig. 2. Here, 32 iso-sequential LNA-

2 diastereoisomers). Interestingly, when screening for general gapmers targeting HIF1A, were screened for target knockdown at

cytotoxic effects in a cell proliferation assay, the gapmers were 5 mM in HeLa cells. The LNA design for these gapmers was

observed to also demonstrate more than 200-fold differences in LxxxDDDDDDxxL, where L stands for LNA, D for DNA, and x

the concentrations at which half-maximal cytotoxicity was ob- for either LNA or DNA, giving 32 different combinations. Mea-

served. These properties were not closely correlated, and the sured knockdowns ranged from no effect to 80%, achieved by the

authors were able to select seven gapmers that were both well compound with the design LLLDDDDDDDDDL. The predicted Tm

tolerated and highly potent in vitro. Out of the seven ‘hits’, the [100] of the designs ranged from 40 C to 60 C, and did not

gapmer with the highest potency reduced antiviral activity to correlate with potency.

<5%. This target region was explored in detail by designing 21

additional LNA gapmers with a 4–8–4 design closely tiled across Functional diversity in vivo

this region. Upon screening of this library, a range of activities was Administered oligonucleotides will interact differentially at each

again observed. Some compounds did not affect HCV activity at of the steps they encounter on their way to the site of action. These

all, but three gapmers, consecutively spaced only 1 nt apart, were steps include absorption at the site of administration and plasma

even more potent than the original ‘lead’. The authors also brieﬂy protein binding in the circulation; the uptake in different tissues

explored architectural diversity by changing the LNA-modiﬁca- and protein binding at the cell surface; the uptake into cells; and

tion pattern of the gapmer with highest potency and designed an their intracellular transport to the site of action. Under in vivo

additional 47 gapmers. They observed that 4–8–4 and 3–10–3 conditions, the ﬁve diversity parameters described above will also

designs had similar potency, whereas 5–6–5 and 2–12–2 designs collectively govern the properties of the speciﬁc molecule. Func-

were three and ﬁve times less potent, respectively. Notably, al- tional diversity is very high for some properties and smaller for

though the 4–8–4 and 3–10–3 designs were of similar potency in others.

106 www.drugdiscoverytoday.com

Drug Discovery Today Volume 23, Number 1 January 2018 REVIEWS

Until recently, the least diversity had been observed in the Taken together, tissue uptake of LNAs shows a high degree of

systemic exposure to LNAs after subcutaneous administration. diversity. The question is whether the differences in tissue uptake

In circulation, LNAs are bound to plasma proteins to a high degree, among a series of compounds are correlated with target activity

primarily serum albumin, although the exposure is species depen- and adverse effect of the oligonucleotides. The short answer is no:

dent. Mice need doses ten times higher than humans per kg to there is no direct correlation between the tissue uptake across a

reach the same exposure in plasma [133]. Such comparisons have panel of different compounds and activity. In the study illustrated

been discussed for 2 -O-(2-methoxyethyl)-modiﬁed oligonucleo- in Fig. 3, the target knockdown (blue bars) and tissue uptake (green

tides [133] and are important for our understanding of species- bars) varied ﬁvefold. Again, there was no direct correlation be-

speciﬁc effects. At therapeutically relevant plasma concentrations tween the liver uptake and target knockdown across the com-

(up to 10 mg/ml), 95–99% of LNAs are bound to plasma proteins in pounds. Thus, an oligonucleotide with a very low liver uptake

the rat. In monkey and humans, a lower fraction is bound to could be very active.

plasma proteins (90–95%) and, in the mouse, even a lower frac- The higher afﬁnity of LNAs and LNA analogs has been reported KEYNOTE REVIEW

tions [134]. The free (unbound) fraction of oligonucleotide is to result in hepatotoxicity [135]. By contrast, it has also been

secreted via the urine. Within the ﬁrst 24 h post dose, the plasma reported that hepatotoxicity was signiﬁcantly reduced by using

concentration is reduced by two to three orders of magnitude. This shorter version of the LNAs (from 20mer to 14mer) [135]. One

Reviews

reduction results from the rapid redistribution of the oligonucleo- should be careful to calculate any property correlations in relation

tide to different tissues, predominantly liver and kidney, followed to a single determinant, such as length. In other words, does a

by skin, bone marrow, spleen, uterus, lymph node, and lung property difference result from a length change or because differ-

[56,101]. ent compounds are being compared? For instance, it was shown

Using a dosing schedule of 15 mg/kg on days 0, 3, 7, 10, and 14 that a 13mer LNA was toxic, but, by adding just one LNA nt, the

of nine LNA oligonucleotides in mice (dosed, i.v.), a different corresponding 14mer was nontoxic [19]. The speciﬁc composition

uptake and accumulation was revealed in the liver (green bars in and the nucleotide positioning (including LNA nt) is also a deter-

Fig. 3). The 16mer gapmer (3–10–3) oligonucleotides with differ- minant because LNAs with the same nucleobase sequence and

ent nucleobase sequences were designed against the same RNA with different positions of the LNA in the ﬂanks can have different

target. At sacriﬁce on day 16, the liver concentration of some of hepatotoxic proﬁles [19,136]. These studies indicate that some

these oligonucleotides was as low as 15 mg/g liver tissue, whereas sequence motifs inﬂuence the hepatotoxic potential. Recent

others showed values of almost 80 mg/g liver tissue. These results experiments in mice suggest that the number of unintended

are by no means unique. Other studies showed differences in the RNA targets that are effectively reduced in the liver after systemic

liver uptake of up to 25-fold. Uptake of LNA oligonucleotides in administration of high-afﬁnity gapmers, such as LNA gapmers, are

kidney and other tissues show similar diversities. A kidney:liver also correlated with the hepatotoxic potential of the gapmers [96–

disposition ratio spanning over more than two orders of magni- 98]. High-afﬁnity gapmers can also affect unintended partially

tude has been observed in mice (H.F. Hansen, pers. commun.). mismatched RNA targets [137]. These results are consistent with a

model relating increased unintended targeting with increased risk

of some of these targets being involved in critical cellular functions

leading to hepatotoxicity. Alternatively, it could be that the

concentration of cleaved RNA in the hepatocytes can reach a level

100 2800

that is in itself toxic to the cells. That hepatotoxicity can be a

ALT concentration

75 2100 consequence of off-target effects also helps explain why structural

diversity affects the hepatotoxic potential. As discussed above,

50 1400 structural diversity clearly affects activity on the intended target

and, therefore, it is reasonable to expect that structural diversity

or mRNA level 25 700

also can affect activity on off-targets.

Using the dosing schedule described above, ﬁve administrations Oligonucleotide concentration

0 0

SalineA B C D E F G H I over 15 days (every third to fourth day) of 15 mg/kg, and measur-

Oligonucleotide ing alanine aminotransferase (ALT) in plasma/serum as a measure

Drug Discovery Today for hepatotoxicity, high hepatotoxicity was found for both oligo-

nucleotides taken up at high and very low tissue concentrations

FIGURE 3

(red bars in Fig. 3). Likewise, there are oligonucleotides that are

Nucleobase diversity affects uptake, target knockdown, and hepatotoxic

taken up in high concentrations without being toxic. Taken

potential in vivo. Nine LNA gapmers (labeled A–I) of the design 3–10–3, with

together, these results suggest that the liver uptake for different

different nucleobase sequences but all targeted to the same target RNA,

oligonucleotides does not necessarily relate to activity or toxicity

were administered intravenously at 15 mg/kg on days 0, 3, 7, 10, and 14 in

mice (n = 5 mice per treatment group). On day 16, the livers were isolated even for oligonucleotides that are seen as ‘similar’ (i.e., with same

and homogenized, and the oligonucleotide concentration (green bars, g/g length, design, and chemistry).

tissue) and target mRNA levels (blue bars, % saline control) were determined

It has been reported that some LNAs are toxic [19,102–

by hybridization-ELISA and qPCR, respectively. The alanine aminotransferase

104,131,132,135]. The toxicity cannot be caused by the LNA

(ALT) level (red bars, IU/L) was measured in corresponding serum samples by

nucleosides per se, because there are also many reports of LNAs

ELISA. For comparison, mRNA and ALT levels for mice in the study

administered with saline are shown to the left. All values are shown as group that are not toxic [19,97,99,102,103,131]. It has also been clearly

averages. Error bars indicate the SD. demonstrated that toxicity can be mitigated by either increasing or

www.drugdiscoverytoday.com 107

REVIEWS Drug Discovery Today Volume 23, Number 1 January 2018 (a) (b)

120 50 100 40 80 30 60 20 40 mRNA (% control)

Reviews 10 20 0 0

HIF1A

Number of diastereoisomers

Diastereoisomeric gapmers with different backbone configurations 0 40 80 120 ENT REVIEW KEYNOTE HIF1A mRNA (% control)

Drug Discovery Today

FIGURE 4

Backbone diversity of diastereoisomers affects target knockdown in vitro. (a) Selected stereoisomers (n = 236) of a 13mer human HIF1A targeting LNA gapmer,

screened in HeLa cells at 5 mM by gymnosis. The phosphorothioates (PS) configuration (Rp/Sp) for at least one phosphate is varied in each stereoisomer. After 3

days, knockdown of the target mRNA was measured by qPCR (and compared with a saline control). (b) The measured target knockdowns of the diastereoisomers

approximately follow a normal distribution (P = 0.32 by the Shapiro–Wilk test of normality).

decreasing the number of LNA nucleotides or, for iso-sequential In another experiment, six 13mer stereodeﬁned ApoB-targeting

AONs, by changing the nucleoside composition (architectural LNAs were tested in mice. The LNAs were dosed i.v. at 1.0 mg/kg

diversity), or, as observed in Fig. 3, by changing the nucleobase (Fig. 5). The stereodeﬁned LNAs (A–F) downregulated ApoB

sequence (nucleobase diversity). Irrespective of the underlying mRNA from 87% to 20%, with the random-mixture LNA,

mechanism affected by these structural changes, it is apparent RTR3833, reducing mRNA to approximately 50% (blue bars in

that diversity can be exploited to mitigate toxicity as well as Fig. 5). The uptake in liver spanned one order of magnitude (1.34–

optimizing other properties. 0.135 mg/g, green bars in Fig. 5). It was interesting to note that

high uptake did not relate to the highest activity for the six

Stereodefined LNA oligonucleotides stereodeﬁned compounds (Pearson’s correlation r = 0.35 and

Nearly all AONs are synthesized without stereodeﬁnition of the PS P = 0.44; Fig. 5). Only small differences in ALT concentrations

internucleoside linkage. Thus, for the LNA PSs described above,

each ‘compound’ exists as a random mixture of diastereoisomers.

Fundamentally, diastereoisomers are not identical and do have

140 700

different properties. Moving away from the random mixtures and

making single stereodeﬁned PS AONs offers a new dimension to 120 600

RNA therapeutics [32]. One aspect of this is that macromolecular 100 500

ALT concentration

diversity can now be studied, such as by computer modeling (see

80 400

above). An interesting outcome of this was QM calculations illus-

trating that changing a single PS bond conﬁguration can produce a 60 300

macromolecular structure diversity to the same extent as, for or mRNA level

40 200

example, nucleobase or nucleotide changes [127,128].

20 100

In an in vitro experiment, 236 stereodeﬁned LNAs, randomly Oligonucleotide concentration

selected from 4096 possible isomers of a 13mer targeting human 0 0

HIF1A, were tested. A difference in knockdown ranging from 0% to

77% was observed (Fig. 4a). The distribution of HIF1A mRNA levels Saline Stereo F Stereo A Stereo B Stereo E Stereo C Stereo D

RTR3833

after treatment with gapmer is represented in a histogram in

Fig. 4b (gray bars). The solid line in Fig. 4b represents a normal Oligonucleotide

distribution with mean and standard deviation estimated as the Drug Discovery Today

median (70%) and the median absolute deviation (22%) of the

FIGURE 5

knockdown data. From this, it can be seen that the distribution of

Backbone diversity of diastereoisomers affects target knockdown in vivo. A

the knockdown data as represented in the histogram is well

random-mixture locked nucleic acid (LNA) gapmer, RTR3833, targeted to the

approximated by the normal distribution (Fig. 4b). The original

mouse APOB mRNA, as well as six different fully stereodefined gapmers with

HIF1A

random-mixture LNA gapmer reduced mRNA to approxi- the same nucleobase sequence as RTR3833, labeled Stereo A–F, were

mately 50%. If it can be assumed that the extent of the knockdown administered at 1 mg/kg in mice (n = 5 mice per treatment group) on day 0.

On day 3, the livers were isolated and homogenized, and the oligonucleotide

achieved by each of the 4096 isomers follows a normal distribu-

concentration (green bars, mg/g tissue) and target mRNA levels (blue bars, %

tion, as suggested from the 236 isomers measured and represented

saline control) were determined by hybridization-ELISA and qPCR,

in Fig. 4b, we can infer that approximately 20% of all 4096 isomers

respectively. The alanine aminotransferase (ALT) level (red bars, IU/L) was

HIF1A

will reduce mRNA to a larger extent than the original measured in corresponding serum samples by ELISA. All values are shown as

random-mixture LNA gapmer. group averages. Error bars indicate the SD.

108 www.drugdiscoverytoday.com

Drug Discovery Today Volume 23, Number 1 January 2018 REVIEWS

were seen for the stereodeﬁned LNAs (A–F) and none of them Diversity offers new opportunities for RNA

were above twofold of the ALT concentration observed in saline- therapeutics

treated controls (red bars in Fig. 5). Consequently, none of the The standard, target-centric, model of RNA drug discovery begins

stereodeﬁned LNAs displayed markedly increased hepatotoxicity. with the identiﬁcation of a target known to be causally involved in

Thus, in line with the QM observations, in vitro knockdown and a particular disease (Fig. 6a). From the sequence of the RNA target,

in vivo uptake and knockdown are highly diverse among diaster- libraries of complementary AONs can then be designed and syn-

eoisomers. thesized. The functional properties of the synthesized AONs are

The observation that individual stereoisomers can have sub- evaluated in a screening cascade of cell-based assays and animal

stantially improved pharmacological properties compared with a models, to identify active and well-tolerated compounds (Fig. 6b).

random mixture was also recently reported for the 5-10-5 MOE- In this approach, medicinal chemists are confronted with a

gapmer mipomersen [33]. Mipomersen comprises 524 288 stereo- large chemical space to explore. For gapmers designed to knock

isomers and targets APOB RNA. The authors demonstrated for a set down the target RNA, as well as mixmers inhibiting miRNAs or KEYNOTE REVIEW

of six different fully stereodeﬁned versions of mipomersen that modulating splice switching on pre-mRNA, the number of chemi-

three were more stable than the random mixture in both rat liver cally feasible compounds against any target is on the order of 10

homogenate and in rat serum, and two were cleaved faster by (Box 1). Despite great advances in high-throughput synthesis and

Reviews

RNase H in vitro [33]. The stereoisomer that was both more stable screening technology, such a large space prohibits exhaustive

and cleaved signiﬁcantly faster was also tested in mice, and evaluation.

showed signiﬁcantly longer-lasting lowering of serum ApoB pro- The current target-centric discovery approach relies on the

tein levels compared with the random mixture when dosed intra- principle of ‘evaluation in-parallel and local optimization’. That

peritoneally eight times at 10 mg/kg [33]. Interestingly, this is, libraries of AONs that predominantly explore one parameter (

stereoisomer had a triplet stereochemical motif, SSR, within the for example potency) are designed, synthesized, and screened in

gap, which the authors argued could be particularly suited to parallel, and, from these, sets of AON hits are selected that are

promote target RNA cleavage by RNase H. The authors tested optimal with respect to that parameter. Given that synthesis and

the code in stereodeﬁned gapmers targeted to FOXO1 RNA and screening of AONs takes on the order of weeks for the in vitro part of

APOC3 RNA, and demonstrated an improvement compared with the screening cascade, and on the order of months for the in vivo

the random mixtures. However, this ﬁnding is in contrast to an part, parallel evaluation of libraries of AONs is necessary to keep

earlier study, where, among a set of 31 stereoisomers, no clear the timespan of the discovery project on the order of 1–2 years.

differences in in vitro potency was observed between the stereo- Using the analogy of chemical space, the map guiding the search

isomers containing the SSR code and those that did not [34]. More- for active and tolerated AONs is as such constructed ‘along the

comprehensive analyses with larger sets of stereoisomers are need- way’, and the path forward determined only by what is known in

ed to establish whether such a preferred triplet code is a general the local regions explored up until that point. This approach does

phenomenon. not necessarily lead to the identiﬁcation of the globally most-

The diversity data observed for stereodeﬁned AONs also point active and best-tolerated AON, but historically have nevertheless

the way to a novel understanding of how AON chemistry affects performed well and produced numerous drug candidates that have

function. In charged AONs, the PS linkage is required for optimal entered clinical testing in human trials (see above).

antisense activity [23]. This is also true for AONs in which the PS To make these concepts more concrete, we illustrate in Fig. 6c,d

linkage is not required for making them nuclease resistant. Mech- a recent drug discovery project in which local regions of chemical

anistically, the importance of the PS linkage is often assigned to space were explored with respect to nucleobase as well as archi-

the heavy atom effect of sulfur, which alters the sphere of hydra- tectural diversity (Fig. 6c). The gapmers from the ﬁrst scanning

tion surrounding the molecule relative to the PO linkage. This, library (green dots in Fig. 6c,d) were screened for activity on the

plus the increased polarizability of the PS linkage and its ability to intended target as well as cellular toxicity. As seen in Fig. 6d, some

form salt bridges, alters protein binding in addition to increasing were active and well tolerated, but many were not. The scanning

AON lipophilicity. However, because all possible combinations of library primarily explores the nucleobase determinant, a classic

Rp and Sp diastereomers are present in the classic nonstereode- ‘gene-walk’ (Fig. 6c). Next, the nucleobase sequences for a handful

ﬁned AON random mixtures (see above), the effects of each of interesting compounds were identiﬁed and used as a basis for

individual Rp and Sp are so diminutive that the PSs collectively designing optimization libraries exploring nucleoside and/or ar-

appears ‘pseudosymmetric’, essentially neutralizing the chiral chitectural diversity (blue dots in Fig. 6c,d). As seen in Fig. 6d, the

contributions of a PS linkage. However, the nature of a classic optimization libraries identiﬁed a higher fraction of leads that

random mixture, essentially a library of diastereoisomers, might in were potent and well tolerated compared with the hits in ﬁrst

itself have a positive role. Given that random mixtures exhibit all scanning libraries (upper right quadrant in Fig. 6d).

possible isomers, including isomers with extreme and/or desired Furthermore, having identiﬁed leads where both nucleobase

properties, such as high potency, they might outbalance less sequence and architectural design had been locally optimized,

active, or inactive, isomers [34,138]. It is possible that such a additional improvements in functional properties might be possi-

‘library effect’ and the heavy atom effect work together to promote ble by synthesizing new libraries of fully stereodeﬁned versions of

AON PS activity. By contrast, advantages with AON technology these gapmers (Fig. 6a). One such example was presented above for

might occur when both the heavy atom effects of sulfur combined fully stereodeﬁned gapmers targeting HIF1A mRNA (Fig. 4). The

with pure phosphorus chirality are used (see below). Naturally, this high property diversity observed for PS diastereoisomers offers a

can only occur for stereodeﬁned PSs. new macromolecular diversity parameter for AON optimization.

www.drugdiscoverytoday.com 109

REVIEWS Drug Discovery Today Volume 23, Number 1 January 2018

(a) (b) Disease RNA target Cellular activity assays knowledge identification evaluating target RNA and protein after treatment with gapmer

Scanning library (n = ~1000) Cellular toxicity assays

Reviews spread out across target for relevant tissues where the using a few standard designs gapmers will accumulate to identify hits ENT REVIEW KEYNOTE Animal models Optimization libraries (n = ~500) to evaluate safety and PK/PD tiled closely in a narrow region using many different designs to identify leads/candidate drugs

Clinical development in humans Optimization libraries (n = ~500) of fully stereodefined gapmers to identify candidate drugs

(d) Scanning Optimization (c) High Low or no

Cellular toxicity assay Architectural diversity

(ranked by prevalence) 1 140 000 Low or no High Nucleobase diversity (start position on RNA) Cellular activity assay

Drug Discovery Today

FIGURE 6

RNA drug discovery and diversity. (a) The standard, target-centric, model of RNA drug discovery is a linear process in which new RNA targets are identified

through knowledge of a particular disease, and libraries of gapmers complementary to the RNA target are synthesized. The hits from one library are used as a

basis for designing gapmers in the next library. (b) Simplified screening cascade starting with high-throughput screening in cellular assays and ending with more

low-throughput evaluation in animal models. (c) An example where a region of chemical space was explored using 3661 LNA gapmers based on the principle of

‘evaluation in-parallel and local optimization’. Coordinates for nucleobase diversity were approximated by the start position on the RNA target (1706 unique

positions evaluated). For architectural diversity, all 1186 different designs used in the discovery project were ranked by the number of gapmers in which they

were used, and this ranking was used as the coordinate. The most often-used design is at the top of the graph. (d) Measured activity and toxicity in vitro for the

same gapmers as in (c). Based on the measured values, gapmers were divided into four regions, and the relative fraction of gapmers from scanning and

optimization libraries in each region are shown using pie charts. Gapmers selected for evaluation in animal models came from the upper-right region with high

activity and low or no toxicity. Abbreviation: PK/PD, pharmacokinetics and pharmacodynamics.

Also, in contrast to classic random mixture AONs, PS stereodeﬁned region (deep strategy)? As calculated in Box 1, thousands of

oligonucleotides offer a unique opportunity: it is now possible to different modiﬁed gapmers can be designed against any target

study the complex interactions between AONs and, for example, region, so each of these strategies is clearly possible. The answer

proteins by computational modeling. depends on whether nucleobase or architectural diversity affects

The understanding of how structural diversity determinants the therapeutic index the most, and what balance between them

affect functional properties is essential to guide heuristic explores chemical space most efﬁciently. Traditionally, nucleo-

approaches in drug discovery. For example, what is the most base diversity has been explored the most, with the aim of

efﬁcient path to highly potent gapmers? Is it scanning the target searching for the regions on the transcript most accessible to

RNA comprehensively in 1000 regions with one selected stan- the AONs. However, as the examples presented in the sections

dard design (broad strategy)? Is it scanning the target in only above show, we are observing considerable impact on potency

100 different regions but testing ten different gapmer designs in from other diversity determinants. Consequently, we are mov-

each (narrow strategy)? Or is it evaluating only ten different ing away from pure broad strategies with only a single ﬁxed

target regions but with 100 different gapmer designs in each design, towards more balanced and deeper strategies where

110 www.drugdiscoverytoday.com

Drug Discovery Today Volume 23, Number 1 January 2018 REVIEWS

BOX 1 gy). However, the pronounced functional effects that small

The chemical space of RNA therapeutics changes in chemical structure have (see above) offer new oppor-

tunities for AON drug discovery. Diversity has now led us to search

Chemical space can be thought of as an analogy to the

cosmological universe, with chemical compounds populating this much deeper and in a narrower target sequence space (narrow or

space instead of stars [139]. How many molecules populate this deep strategy). If we on top of this add the diversity of diastereoi-

space? For small organic molecules with up to 17 atoms of C, N, O, somers, then a large diversity/property spectrum will unfold even

S, and halogens, and taking chemical stability and synthetic

from just a single classic AON PS compound mixture (ultra-deep

feasibility in account, the number has been estimated to be more

11 strategy). Deep or ultra-deep strategies offer not only an introduc-

than 10 compounds [140]. Here, we estimate the number of

tion of a much larger structural space to search in, but more

possible gapmers that can be designed against an RNA target. For

importantly, also a possibility to begin with much ‘tougher’ selec-

an RNA that is n nucleotides long, and where the gapmer is l

nucleotides long, in total there will then be n l + 1 possible tion criteria. It is possible to begin with higher sequence restric-

target regions of length l. Considering next the backbone tions and/or requirements. If, for instance, many and strict KEYNOTE REVIEW

chemistry, for a gapmer of length l, there will be l 1 linkages requirements are applied, such as species crossreactivity, sequence

between nucleotides, each of which can be in either R - or S -

p p speciﬁcity to only the intended target [15,18], deselection of

l–1

form. This gives 2 different stereoisomers. Finally, any type of m

predicted sequence motifs associated with increased toxic poten-

Reviews

possible modified nucleosides can in principle be used in the

tial [18,19,136], deselection of unwanted sequence motifs from a

regions flanking the DNA gap. Therefore, for a gapmer of length l,

chemistry, manufacturing, and control perspective, and so on, only

with a gap of length g, the number of different modified gapmers

l–g

that can be synthesized equals m . Thus, considering all gapmers a very few sequences will ﬁnally comply. In the case where a few

of lengths between lmin and lmax, taken together the number of 16mer compound mixtures have been identiﬁed complying with

different, fully stereodefined, gapmers is equivalent to Eq. (I): all selection criteria, it should then be possible to ﬁnd compounds

with even better properties among the 32 768 structurally different

Xlmax diastereoisomers of each 16mer. Deep or ultra-deep strategies will

l1 lg

ð þ Þ : ð Þ

n l 1 2 m I also be advantageous for discovery programs focused on, for exam-

l¼lmin

ple, splice-switching or miRNA inhibition, because here the rele-

vant RNA target regions sequence wise are very narrow.

For a standard pre-mRNA that is n = 25 000 nt in length, and There is no doubt that harnessing the full potential offered by

considering gapmer lengths between l = 12 nt and l = 20 nt,

min max diversity drives AON drug discovery in the direction of small-

with a minimal gap of g = 6 nt, and where m = 5 types of

0 0 molecule concepts. This is with respect to ‘higher’-throughput

nucleosides (DNA, LNA, MOE, 2 OMe, and 2 F) are considered,

19 20 screenings and working with single molecules (fully stereodeﬁned

there are 8.89 10 10 different possible gapmers.

AONs). However, AON drug discovery exhibits inherent advan-

Equation I can also be used when estimating the number of

mixmers designed to target a narrow region of RNA, such as a tages over small-molecule principles. AON drug discovery con-

miRNA or a splice-site. In this case, with the same parameters as for cepts are more ‘rational’. Watson–Crick base-pairing rules govern

gapmers but setting n = 20 nt and g = 0 nt, there are 6.17 10 target RNA binding, whereas, for small molecules, complex

10 different possible mixmers, about the same number as for

computational docking studies must be undertaken. Also, impor-

gapmers. The smaller region of RNA that can be targeted is

tantly, oligonucleotide production by solid-phase synthesis fol-

balanced out by having the entire oligonucleotide available for

lows shared principles for all compounds. Furthermore, as we

modifications, where, for gapmers, there is a gap where only DNA

develop our understanding of these new SARs, we expect to apply

can be used.

an even more ‘rational’ approach to AON drug discovery, such as

by developing in silico algorithms that are more predictive for AON

multiple designs are tested in highly overlapping regions on the screening outcomes taking diversity into account. This has already

target RNA. been done for hepatotoxic potential [19]. In this way, bioinfor-

Another point to consider is the multidimensional nature of the matics and computational modeling can enable faster discovery

local optimization in chemical space, where several properties of processes and, throughout the process, more informative deci-

gapmers have to be acceptable and drug-like at the same time. sions, thereby creating stronger lead candidates.

Should we then measure all functional properties for the entire

library of gapmers? Or are there some properties that are not Concluding remarks

necessary to consider early on, because diversity can be introduced The fact that structural differences also deﬁne functional diversi-

at a later stage to ensure that the wanted functional properties can ties is a given for all compound classes. We have shown that not

be achieved? Say, for example, that exploring nucleobase diversity only sequence, but also the design of an AON is a strong property

has identiﬁed potent but not well-tolerated gapmer hits. When determinant (Fig. 2). A clear SAR exists within a deﬁned target

then nucleoside, architectural, or backbone diversity is explored, nucleobase sequence simply by varying the sugar and phosphate

will the chance of ﬁnding leads that are both potent and well backbone within that sequence (Fig. 4). This creates AONs with

tolerated be higher if we had not begun with hits only optimized different pharmacological properties (Figs 3 and 5). Exploration of

for potency? the large property space is the reason why drug discovery libraries

If AONs shared the same properties simply by membership of in our laboratories have expanded greatly. Initially, we used

the same chemical class (‘portability of chemistry’), only sequence mainly a broad screening strategy (gene walk), but now we are

would differentiate properties and broad gene walks would be also exploring compound diversities within selected nucleobase

required to have enough compounds to select from (broad strate- sequences (deep strategy) (Fig. 6).

www.drugdiscoverytoday.com 111

REVIEWS Drug Discovery Today Volume 23, Number 1 January 2018

The diversity we experienced for stereodeﬁned LNAs (with the discovery strategies are producing increasingly better RNA thera-

same sequence and nucleoside design) represents an extra dimen- peutics. We have realized that the old concept in antisense of

sion by which even more unique leads might be identiﬁed. The ‘portability of chemistry’ only provides a starting point for RNA

basis for their uniqueness might, in part, be created by the positive drug discovery and, as shown here, structural diversity at all levels

heavy atom effects of sulfur and the introduction of chirality on (nucleobase, nucleoside, backbone, and architectural) can be used

phosphorous, which we think act as structural diversity parame- to produce AONs with improved drug properties.

Reviews ters in themselves. A chiral phosphorous scaffold might be ideal

for designing drugs interacting with the chiral world of biology. Conflict of interest

Given that AONs are based on something as predictable as All authors except N.P. are full time employees at Roche Innova-

ENT REVIEW KEYNOTE

Watson–Crick base-pairing to RNA, binding to intended and tion Centre Copenhagen A/S.

unintended RNA targets can be predicted computationally and

measured experimentally [18]. This provides a means for assessing Acknowledgments

unintended hybridization-induced biological effects of AONs We would like to thank Cy A. Stein for helpful comments on the

[105]. Still, they are, perhaps unsurprisingly, unique molecules manuscript, and Mikkel Adriansen for in-depth analysis of the

that, in a therapeutic setting, provide unique and yet ‘hard-to- mass spectra of stereo-deﬁned oligonucleotides. For their excellent

predict’ pharmacological properties. Clearly, more work is needed technical assistance, we would like to thank Helle Nymark and

to build understanding of the underlying mechanism of AON Annette Glidov for qPCR high-throughput screening, Charlotte

diversity, and to determine how much the TI can be expanded Øverup for mouse liver and serum measurements, Rikke Sølberg,

by these new strategies. Camilla T. Haugaard, and Lisbeth Bang for handling of mice in the

However, what is clear at this point is that antisense has devel- animal facilities, and Valesi Mukaka for the synthesis and

oped immensely over the past four decades, and the new drug formulation of oligonucleotides for mouse studies.

References

1 Curwen, V. et al. (2004) The Ensembl automatic gene annotation system. Genome 20 Crooke, S.T. et al. (2017) The effects of 2 -O-methoxyethyl containing

Res. 14, 942–950 antisense oligonucleotides on platelets in human clinical trials. Nucleic Acid Ther.

2 Morris, K.V. and Mattick, J.S. (2014) The rise of regulatory RNA. Nat. Rev. Genet. 15, 121–129

423–437 21 Beaucage, S.L. et al. (1999) Current Protocols in Nucleic Acid Chemistry. John Wiley &

3 Lin, M.F. et al. (2011) Locating protein-coding sequences under selection for Sons, Inc.

additional, overlapping functions in 29 mammalian genomes. Genome Res. 21, 22 Brown, T. and Brown, D.J.S. (1991) Modern machine-aided methods of

1916–1928 oligodeoxiribonucleotide synthesis. In Ligonucleotides and Analogues: A Practical

4 Esteller, M. (2011) Epigenetic changes in cancer. F1000 Biol. Rep. 3, 9 Approach (Eckstein, F., ed.), pp. 13–14, IRL Press

5 Alexa Dickson, M. and Jeffrey Wilusz, C.J.W. (2009) mRNA turnover. In 23 Cohen, J.S. (1993) Phosphorothioate oligonucleotides. In Antisense Research and

Encyclopedia of Life Sciences. John Wiley & Sons Ltd. Applications (Crooke, S.T. and Lebleu, B., eds), pp. 205–223, CRC Press

6 Zamecnik, P.C. and Stephenson, M.L. (1978) Inhibition of Rous sarcoma virus 24 Wengel, J. (2001) LNA (locked nucleic acid). In Antisense Drug Technology;

replication and cell transformation by a speciﬁc oligonucleotide. Proc. Natl. Acad. Principles, Strategies, and Applications (Crooke, S.T., ed.), pp. 339–357, Marcel

Sci. U. S. A. 75, 280–284 Dekker

7 Cook, P.D. (1999) Making drugs out of oligonucleotides: a brief review and 25 Burgers, P.M. and Eckstein, F. (1978) Absolute conﬁguration of the diastereomers

perspective. Nucleosides Nucleotides 18, 1141–1162 of adenosine 5 -O-(1-thiotriphosphate): consequences for the stereochemistry of

8 Uhlmann, E. and Peyman, A. (1990) Antisense oligonucleotides: a new therapeutic polymerization by DNA-dependent RNA polymerase from Escherichia coli. Proc.

principle. Chem. Rev. 90, 544–579 Natl. Acad. Sci. U. S. A 75, 4798–4800

9 Uhlmann, E. (2000) Recent advances in the medicinal chemistry of antisense 26 Eckstein, F. et al. (1976) Stereochemistry of polymerization by DNA-dependent

oligonucleotides. Curr. Opin. Drug Discov. Dev. 3, 203–213 RNA-polymerase from Escherichia coli: an investigation with a diastereomeric ATP-

10 Eckstein, F. (2007) The versatility of oligonucleotides as potential therapeutics. analogue. Proc. Natl. Acad. Sci. U. S. A. 73, 2987–2990

0 0

Expert Opin. Biol. Ther. 7, 1021–1034 27 Eckstein, F. and Gindl, H. (1968) Uridin-2 ,3 -O,O-cyclothiophosphate. Chem. Ber.

11 Khvorova, A. and Watts, J.K. (2017) The chemical evolution of oligonucleotide 101, 1670–1673

therapies of clinical utility. Nat. Biotechnol. 35, 238–248 28 Eckstein, F. and Gindl, H. (1969) Polyribonucleotides containing a thiophosphate

12 McClorey, G. and Wood, M.J. (2015) An overview of the clinical application of backbone. FEBS Lett. 2, 262–264

antisense oligonucleotides for RNA-targeting therapies. Curr. Opin. Pharmacol. 24, 29 Saenger, W. and Eckstein, F. (1970) Stereochemistry of a substrate for pancreatic

52–58 ribonuclease Crystal and molecular structure of the triethylammonium salt of

0 0

13 Koch, T. (2013) LNA antisense: a review. Curr. Phys. Chem. 3, 55–68 uridine 2 ,3 -0,0-cyclophosphorothioate. J. Am. Chem. Soc. 92, 4712–4718

14 Obika, S. et al. (1998) Stability and structural features of the deplexes containing 30 Wilk, A. and Stec, W.J. (1995) Analysis of oligo(deoxynucleoside

0 0

nucleoside analogues with a ﬁxed N-type conformation, 2 -0,4 -C- phosphorothioates)s and their diastereomeric composition. Nucleic Acid Res. 23,

methyleneribonucleosides. Tetrahedron Lett. 39, 5401–5404 530–534

15 Singh, S.K. et al. (1998) LNA (locked nucleic acids): synthesis and high-afﬁnity 31 Iyer, R.P. et al. (1998) Solid-phase stereoselective synthesis of oligonucleoside

nucleic acid recognition. Chem. Commun. 0, 455–456 phosphorothioates: The nucleoside bicyclic oxazaphospholidines as novel

16 Wahlestedt, C. et al. (2000) Potent and nontoxic antisense oligonucleotides synthons. Tetrahedron Lett. 39, 2491–2494

containing locked nucleic acids. Proc. Natl. Acad. Sci. U. S. A. 97, 5633–5638 32 Wang, J.-C. and Just, G. (1997) A stereoselective synthesis of dinucleotide

17 Swayse, E.E. et al. (2007) Antisense oligonucleotides containing locked nucleic acid phosphorothioates, using chiral indol-oxazaphosphorine intermediates.

improve potency but cause signiﬁcant hepatotoxicity in animals. Nucleic Acids Res. Tetrahedron Lett. 38, 3797–3800

35, 687–700 33 Iwamoto, N. et al. (2017) Control of phosphorothioate stereochemistry

18 Hagedorn, H.P. et al. (2017) Managing the sequence-speciﬁcity of antisense substantially increases the efﬁcacy of antisense oligonucleotides. Nat. Biotech. 845–

oligonucleotides in drug discovery. Nucleic Acid Res. 2262–2282 851

19 Hagedorn, P.H. et al. (2013) Hepatotoxic potential of therapeutic oligonucleotides 34 Wan, W.B. et al. (2014) Synthesis, biophysical properties and biological activity of

can be predicted from their sequence and modiﬁcation pattern. Nucleic Acid Ther. second generation antisense oligonucleotides containing chiral phosphorothioate

23, 302–310 linkages. Nucleic Acids Res. 42, 13456–13468

112 www.drugdiscoverytoday.com

Drug Discovery Today Volume 23, Number 1 January 2018 REVIEWS

35 Oka, N. et al. (2008) Solid-phase synthesis of stereoregular 64 Rajwanshi, V.K. et al. (2000) The eight stereoisomers of LNA (locked nucleic acid): a

oligodeoxyribonucleoside phosphorothioates using bicyclic oxazaphospholidine remarkable family of strong RNA binding molecules. Angew. Chem. Int. Ed. 39,

derivatives as monomer units. J. Am. Chem. Soc. 130 (47), 16031–16037 1656–1659

0 0

36 Stec, W.J. et al. (1998) Deoxyribonucleoside 3 -O-(2-Thio- and 2-Oxo-’spiro’-4,4- 65 Rosenbohm, C. et al. (2003) Synthesis of 2 -amino-LNA: a new strategy. Org.

pentamethylene-1,3,2-oxathiaphospholane)s: monomers for stereocontrolled Biomol. Chem. 1, 655–663

synthesis of oligo(deoxyribonucleoside phosphorothioate)s and chimeric PS/PO 66 Singh, S.K. et al. (1998) Synthesis of novel bicyclo [2.2.1] ribonucleosides:

0 0

oligonucleotides. J. Am. Chem. Soc. 120, 7156–7167 2 -amino-and 2 -thio-LNA monomeric nucleosides. J. Org. Chem. 63, 6078–6079

37 Oka, N. et al. (2002) Diastereocontrolled synthesis of dinucleoside 67 Singh, S.K. et al. (1998) Synthesis of 2 -amino-LNA: a novel conformationally

phosphorothioates using a novel class of activators, dialkyl(cyanomethyl) restricted high-afﬁnity oligonucleotide analogue with a handle. J. Org. Chem. 63,

ammonium tetraﬂuoroborates. J. Am. Chem. Soc. 124, 4962–4963 10035–10039

38 Koziolkiewicz, M. et al. (1995) Stereodifferentiation-the effect of P chirality of oligo 68 Soerensen, M.D. et al. (2002) Alpha-L-ribo-conﬁgures locked nucleic acid (alpha-L-

(nucleoside phosphorothioates) on the activity of bacterial RNase H. Nucleic Acids LNA): synthesis and properties. J. Am. Chem. Soc. 124, 2164–2176

Res. 23, 5000–5005 69 Egli, M. et al. (2001) X-ray crystal structure of a locked nucleic acid (LNA) duplex

39 Koziolkiewicz, M. et al. (1997) Stability of stereoregular oligo(nucleoside composed of a palindromic 10-mer DNA strand containing one LNA thymine

phosphorothioate)s in human plasma: diastereoselectivity of plasma 3 - monomer. Chem. Commun. 0, 651–652

KEYNOTE REVIEW

exonuclease. Antisense Nucleic Acid Drug Dev. 7, 43–48 70 Nielsen, K.E. et al. (2000) Solution structure of an LNA hybridized to DNA: NMR

40 Krieg, A.M. et al. (2003) P-chirality-dependent immune activation by study of the d(CT(L)GCT(L)T(L)CT(L)GC):d(GCAGAAGCAG) duplex containing

phosphorothioate CpG oligodeoxynucleotides. Oligonucleotides 13, 491–499 four locked nucleotides. Bioconjug. Chem. 11, 228–238

41 Monteleone, G. et al. (2015) Mongersen, an oral SMAD7 antisense oligonucleotide, 71 Nielsen, K.E. et al. (2004) NMR studies of fully modiﬁed locked nucleic acid (LNA)

Reviews

and Crohn’s disease. N. Engl. J. Med. 372, 1104–1113 hybrids: solution structure of an LNA:RNA hybrid and characterization of an LNA:

42 Gura, T. (1995) Antisense has growing pains. Science 270, 575–577 DNA hybrid. Bioconjug. Chem. 15, 449–457

43 Eckstein, F. (2014) Phosphorothioates, essential components of therapeutic 72 Petersen, M. et al. (2000) The conformations of locked nucleic acids (LNA). J. Mol.

oligonucleotides. Nucleic Acid Ther. 24, 374–387 Recognit. 13, 44–53

44 Freier, S.M. and Altmann, K.-H. (1997) The ups and downs of nucleic acid duplex 73 Petersen, M. et al. (2002) Locked nucleic acid (LNA) recognition of RNA: NMR

stability: structure-stability studies on chemically-modiﬁed DNA:RNA duplexes. solution structures of LNA:RNA hybrids. J. Am. Chem. Soc. 124, 5974–5982

Nucleic Acid Res. 25, 4429–4443 74 Petersen, M. et al. (2003) Structural characterization of LNA and alpha-L-LNA

45 Galderisi, U. et al. (2001) Clinical trials of a new class of therapeutic agents: hybridized to RNA. Nucleosides Nucleotides Nucleic Acids 22, 1691–1693

antisense oligonucleotides. Emerg. Drugs 6, 69–79 75 Xu, Y. et al. (2017) The free energy of locking a ring: changing a

46 Martin, P. (1995) Ein neuer Zugang zu 2 -O-Alkylribonucleosiden und deoxyribonucleoside to a locked nucleic acid. J. Comput. Chem. 38, 1147–1157

Eigenschaften deren Oligonucleotiden. Helv. Chim. Acta 78, 486–504 76 Yildirim, I. et al. (2014) Interplay of LNA and 2 -O-methyl RNA in the structure and

47 Thuong, N.T. and He´le`ne, C. (1993) Sequenzspeziﬁsche Erkennung und thermodynamics of RNA hybrid systems: a molecular dynamics study using the

Modiﬁkation von Doppelhelix-DNA durch Oligonucleotide. Angw. Chem. 105, revised AMBER force ﬁeld and comparison with experimental results. J. Phys.

697–723 Chem. B 118, 14177–14187

48 Raal, F.J. et al. (2010) Mipomersen, an apolipoprotein B synthesis inhibitor, for 77 Eichert, A. et al. (2010) The crystal structure of an ‘All Locked’ nucleic acid duplex.

lowering of LDL cholesterol concentrations in patients with homozygous familial Nucleic Acids Res. 38, 6729–6736

hypercholesterolaemia: a randomised, double-blind, placebo-controlled trial. 78 Hughesman, C.B. et al. (2011) Role of the heat capacity change in understanding

Lancet 375, 998–1006 and modeling melting thermodynamics of complementary duplexes containing

49 Chiriboga, C.A. et al. (2016) Results from a phase 1 study of nusinersen (ISIS-SMN standard and nucleobase-modiﬁed LNA. Biochemistry 50, 5354–5368

(Rx)) in children with spinal muscular atrophy. Neurology 86, 890–897 79 Frieden, M. et al. (2003) Expanding the design horizon of antisense

0 0

50 Obika, S. et al. (1997) Synthesis and conformation of 3 -0,4 -C- oligonucleotides with alpha-L-LNA. Nucleic Acids Res. 31, 6365–6372

0 0

methyleneribonucleosides, novel bicyclic nucleoside analogues for 2 ,5 -linked 80 Cerritelli, S.M. and Crouch, R.J. (2009) Ribonuclease H: the enzymes in

oligonucleotide modiﬁcation. Chem. Commun. 1643–1644 Eukaryotes. FEBS J. 276, 1494–1505

51 Wengel, J. et al. (1999) LNA (locked nucleic acid). Nucleosides Nucleotides 18, 1365– 81 Crooke, S.T. (1998) Molecular mechanisms of antisense drugs: RNase H. Antisense

1370 Nucleic Acid Drug Dev. 8, 133–134

52 Singh, S.K. and Wengel, J. (1998) Universality of LNA-mediated high-afﬁnity 82 Wu, H. et al. (2004) Determination of the role of the human RNase H1

nucleic acid recognition. Chem. Commun. 0, 1247–1248 in the pharmacology of DNA-like antisense drugs. J. Biol. Chem. 279,

53 Koshkin, A. et al. (1998) LNA (locked nucleic acids): synthesis of the adenine, 17181–17189

cytosine, guanine, 5-methylcytosine, thymine and uracil bicyclonucleoside 83 Braasch, D.A. and Corey, R. (2000) Locked nucleic acid (LNA): ﬁne-tuning the

monomers, oligomerisation, and unprecedented nucleic acid recognition. recognition of DNA and RNA. Chem. Biol. 55, 1–7

Tetrahedron 54, 3607–3630 84 Braasch, D.A. et al. (2002) Antisense inhibition of gene expression in cells by

54 Wengel, J. et al. (2001) LNA (locked nucleic acid) and the diastereoisomeric alpha- oligonucleotides incorporating locked nucleic acids: effect of mRNA target

L-LNA: conformational tuning and high-afﬁnity recognition of DNA/RNA targets. sequence and chimera design. Nucleic Acids Res. 30, 5160–5167

Nucleosides Nucleotides Nucleic Acids 20, 389–396 85 Castanotto, D. et al. (2015) A cytoplasmic pathway for gapmer antisense

55 Fluiter, K. et al. (2005) On the in vitro and in vivo properties of four locked nucleic oligonucleotide-mediated gene silencing in mammalian cells. Nucleic Acids Res. 43,

acid nucleotides incorporated into an anti-H-Ras antisense oligonucleotide. 9350–9361

Chembiochem 6, 1104–1109 86 Liang, X.H. et al. (2017) RNase H1-dependent antisense oligonucleotides are

56 Koch, T. and Ørum, H. (2007) Locked nucleic acid. In Antisense Drug Technology robustly active in directing RNA cleavage in both the cytoplasm and the nucleus.

(Crooke, S.T., ed.), pp. 519–565, CRC Press Mol Ther. 2075–2092

57 Koch, T. et al. (2008) Locked nucleic acid: properties and therapeutic aspects. In 87 Dominski, Z. and Kole, R. (1993) Restoration of correct splicing in thalassemic pre-

Therapeutic Oligonucleotides (Kurreck, J., ed.), pp. 103–141, RSC Publishing mRNA by antisense oligonucleotides. Proc. Natl. Acad. Sci. U. S. A. 90, 8673–8677

58 Yamamoto, T. et al. (2011) Antisense drug discovery and development. Future Med. 88 Rigo, F. et al. (2014) Antisense oligonucleotide-based therapies for diseases caused

Chem. 3, 339–365 by pre-mRNA processing defects. Adv. Exp. Med. Biol. 825, 303–352

59 Oerum, H. and Wengel, J. (2001) Locked nucleic acids: a promising molecular 89 Sierakowska, H. et al. (1996) Repair of thalassemic human beta-globin mRNA in

family for gene-function analysis and antisense drug development. Curr. Opin. mammalian cells by antisense oligonucleotides. Proc. Natl. Acad. Sci. U. S. A. 93,

Mol. Ther. 3, 239–243 12840–12844

60 Oerum, H. et al. (2003) Locked nucleic acids (LNA) and medical applications. Lett. 90 Passini, M.A. et al. (2011) Antisense oligonucleotides delivered to the mouse CNS

Pep. Sci. 10, 325–334 ameliorate symptoms of severe spinal muscular atrophy. Sci. Transl. Med. 3, 72ra18

61 Kloosterman, W.P. et al. (2006) In situ detection of miRNAs in animal embryos 91 Kole, R. and Krieg, A.M. (2015) Exon skipping therapy for Duchenne muscular

using LNA-modiﬁed oligonucleotide probes. Nat. Methods 3, 27–29 dystrophy. Adv. Drug Deliv. Rev. 87, 104–107

62 Obika, S. (2004) Development of bridged nucleic acid analogues for antigene 92 Li, L. et al. (2016) Activating frataxin expression by repeat-targeted nucleic acids.

technology. Chem. Pharm. Bull. (Tokyo) 52, 1399–1404 Nat. Commun. 7, 10606

63 Orom, U.A. et al. (2006) LNA-modiﬁed oligonucleotides mediate speciﬁc 93 Wahlestedt, C. (2013) Targeting long non-coding RNA to therapeutically

inhibition of microRNA function. Gene 372, 137–141 upregulate gene expression. Nat. Rev. Drug Discov. 12, 433–446

www.drugdiscoverytoday.com 113

REVIEWS Drug Discovery Today Volume 23, Number 1 January 2018

94 Jonas, S. and Izaurralde, E. (2015) Towards a molecular understanding of 119 Koller, E. et al. (2011) Mechanisms of single-stranded phosphorothioate modiﬁed

microRNA-mediated gene silencing. Nat. Rev. Genet. 16, 421–433 antisense oligonucleotide accumulation in hepatocytes. Nucleic Acids Res. 39,

95 Li, Z. and Rana, T.M. (2014) Therapeutic targeting of microRNAs: current status 4795–4807

and future challenges. Nat. Rev. Drug Discov. 13, 622–638 120 Fazil, M.H. et al. (2016) GapmeR cellular internalization by macropinocytosis

96 Lanford, R. et al. (2010) Therapeutic silencing of microRNA-122 in primates with induces sequence-speciﬁc gene silencing in human primary T-cells. Sci. Rep. 6,

chronic hepatitis C virus infection. Science 327, 198–201 37721

97 Janssen, H.L. et al. (2013) Treatment of HCV infection by targeting microRNA. N. 121 Castanotto, D. et al. (2016) Protein kinase C-alpha is a critical protein for antisense

Engl. J. Med. 368, 1685–1694 oligonucleotide-mediated silencing in mammalian cells. Mol. Ther. 24, 1117–1125

Reviews 98 van Rooij, E. and Kauppinen, S. (2014) Development of microRNA therapeutics is 122 Mechti, N. et al. (1991) Nuclear location of synthetic oligonucleotides

coming of age. EMBO Mol. Med. 6, 851–864 microinjected somatic cells: its implication in an antisense strategy. Nucleic Acids

99 Straarup, E.M. et al. (2010) Short locked nucleic acid antisense oligonucleotides Symp. Ser. 24, 147–150

potently reduce apolipoprotein B mRNA and serum cholesterol in mice and non- 123 Leonetti, J.P. et al. (1991) Intracellular distribution of microinjected antisense

ENT REVIEW KEYNOTE

human primates. Nucleic Acids Res. 38, 7100–7111 oligonucleotides. Proc. Natl. Acad. Sci. U. S. A. 88, 2702–2706

100 Pedersen, L. et al. (2014) A kinetic model explains why shorter and less afﬁne 124 D’Souza, A.A. and Devarajan, P.V. (2015) Asialoglycoprotein receptor mediated

enzyme-recruiting oligonucleotides can be more potent. Mol. Ther. Nucleic Acids 3, hepatocyte targeting strategies and applications. J. Control. Release 203, 126–139

e149 125 Shi, F. et al. (2007) In situ entry of oligonucleotides into brain cells can occur

101 Obad, S. et al. (2011) Silencing of microRNA families by seed-targeting tiny LNAs. through a nucleic acid channel. Oligonucleotides 17, 122–133

Nat. Genet. 43, 371–378 126 Galloway, W.R. et al. (2010) Diversity-oriented synthesis as a tool for the discovery

102 Kakiuchi-Kiyota, S. et al. (2014) Comparison of hepatic transcription proﬁles of of novel biologically active small molecules. Nat. Commun. 1, 80

locked ribonucleic acid antisense oligonucleotides: evidence of distinct pathways 127 Koch, T. et al. (2014) Quantum mechanical studies of DNA and LNA. Nucleic Acid

contributing to non-target mediated toxicity in mice. Toxicol. Sci. 138, 234–248 Ther. 24, 139–148

103 Burel, S.A. et al. (2016) Hepatotoxicity of high afﬁnity gapmer antisense 128 Bohr, H.G. et al. (2017) Electronic structures of LNA phosphorothioate

oligonucleotides is mediated by RNase H1 dependent promiscuous reduction of oligonucleotides. Mol. Ther. Nucleic Acids 428–441

very long pre-mRNA transcripts. Nucleic Acids Res. 44, 2093–2109 129 Vickers, T.A. et al. (2000) Effects of RNA secondary structure on cellular antisense

104 Kasuya, T. et al. (2016) Ribonuclease H1-dependent hepatotoxicity caused by activity. Nucleic Acids Res. 28, 1340–1347

locked nucleic acid-modiﬁed gapmer antisense oligonucleotides. Sci. Rep. 6, 30377 130 Watt, A. and Freier, S. (2007) Basic principles of antisense drug discovery. In

105 Lindow, M. et al. (2012) Assessing unintended hybridization-induced biological Antisense Drug Technology (Crooke, S.T., ed.), pp. 117–141, CRC Press

effects of oligonucleotides. Nat. Biotechnol. 30, 920–923 131 Laxton, C. et al. (2011) Selection, optimization, and pharmacokinetic properties of

106 Lebedeva, I. et al. (2000) Cellular delivery of antisense oligonucleotides. Eur. J. a novel, potent antiviral locked nucleic acid-based antisense oligomer targeting

Pharm. Biopharm. 50, 101–119 hepatitis C virus internal ribosome entry site. Antimicrob. Agents Chemother. 55,

107 Akhtar, S. et al. (1991) Interactions of antisense DNA oligonucleotide analogs with 3105–3114

phospholipid membranes (liposomes). Nucleic Acids Res. 19, 5551–5559 132 Stanton, R. et al. (2012) Chemical modiﬁcation study of antisense gapmers. Nucleic

108 Juliano, R.L. (2016) The delivery of therapeutic oligonucleotides. Nucleic Acids Res. Acid Ther. 22, 344–359

44, 6518–6548 133 Yu, R.Z. et al. (2015) Predictive dose-based estimation of systemic exposure

109 Crooke, S.T. et al. (2017) Cellular uptake and trafﬁcking of antisense multiples in mouse and monkey relative to human for antisense oligonucleotides

oligonucleotides. Nat. Biotechnol. 35, 230–237 with 2 -o-(2–methoxyethyl) modiﬁcations. Mol. Ther. Nucleic Acids 4, e218

110 Manoharan, M. (2002) Oligonucleotide conjugates as potential antisense drugs 134 Watanabe, T.A. et al. (2006) Plasma protein binding of an antisense

with improved uptake, biodistribution, targeted delivery, and mechanism of oligonucleotide targeting human ICAM-1 (ISIS 2302). Oligonucleotides 16,

action. Antisense Nucleic Acid Drug Dev. 12, 103–128 169–180

0 0

111 Debart, F. et al. (2007) Chemical modiﬁcations to improve the cellular uptake of 135 Seth, P.P. et al. (2009) Short antisense oligonucleotides with novel 2 -4

oligonucleotides. Curr. Top. Med. Chem. 7, 727–737 conformationaly restricted nucleoside analogues show improved potency without

112 Eckstein, F. (2007) The versatility of oligonucleotides as potential therapeutics. increased toxicity in animals. J. Med. Chem. 52, 10–13

Expert Opin. Biol. Ther. 7, 1021–1034 136 Burdick, A.D. et al. (2014) Sequence motifs associated with hepatotoxicity of

113 Stein, C.A. et al. (2010) Efﬁcient gene silencing by delivery of locked nucleic acid locked nucleic acid-modiﬁed antisense oligonucleotides. Nucleic Acids Res. 42,

antisense oligonucleotides, unassisted by transfection reagents. Nucleic Acids Res. 4882–4891

38, e3 137 Kamola, P.J. et al. (2015) In silico and in vitro evaluation of exonic and intronic off-

114 Moghimi, S.M. et al. (2005) A two-stage poly(ethylenimine)-mediated target effects form a critical element of therapeutic ASO gapmer optimization.

cytotoxicity: implications for gene transfer/therapy. Mol. Ther. 11, 990–995 Nucleic Acids Res. 43, 8638–8650

115 Symonds, P. et al. (2005) Low and high molecular weight poly(L-lysine)s/poly(L- 138 Li, M. et al. (2017) Synthesis and cellular activity of stereochemically-pure 2[prime

lysine)-DNA complexes initiate mitochondrial-mediated apoptosis differently. or minute]-O-(2-methoxyethyl)-phosphorothioate oligonucleotides. Chem.

FEBS Lett. 579, 6191–6198 Commun. 53, 541–544

116 Mettlen, M. et al. (2009) Dissecting dynamin’s role in clathrin-mediated 139 Lipinski, C. and Hopkins, A. (2004) Navigating chemical space for biology and

endocytosis. Biochem. Soc. Trans. 37, 1022–1026 medicine. Nature 432, 855–861

117 Lajoie, P. and Nabi, I.R. (2010) Lipid rafts, caveolae, and their endocytosis. Int. Rev. 140 Reymond, J.L. (2015) The chemical space project. ACC Chem. Res. 48, 722–730

Cell. Mol. Biol. 282, 135–163 141 Chan, J.H. et al. (2006) Antisense oligonucleotides: from design to therapeutic

118 Howes, M.T. et al. (2010) Molecules, mechanisms, and cellular roles of clathrin- application. Clin. Exp. Pharmacol. Physiol. 33, 533–540

independent endocytosis. Curr. Opin. Cell Biol. 22, 519–527

114 www.drugdiscoverytoday.com