RNA therapeutics: current status and future directions

Abstract

RNA-based therapeutics have enabled drug discovery and design by introducing a fundamentally different perspective from conventional therapeutics. RNAs are not only the upstream molecules of proteins (coding) but also functional components of an extensive and intricate regulatory network (noncoding) that helps govern biological processes and mechanisms. Recent advances in functional genomics have accelerated the identification of functional RNAs with well-defined roles and their use in diagnostic and therapeutic approaches, which was previously challenging. RNA molecules can dynamically regulate gene function through their sequence and/or structure. Thus, RNA therapeutics offer distinct strategies for both loss- and gain-of-function interventions. This review focuses on druggable functional RNAs and discusses current and emerging technologies, as well as the scope for their potential clinical utility and targeted treatment. It also highlights the most studied optimization strategies and delivery systems designed to enhance the efficacy and tissue-specific delivery of RNA therapeutics to improve clinical outcomes. Finally, we offer our understanding and perspective on the field. Recent developments in enhancing the stability of these molecules in vivo have further improved the therapeutic potential of RNA-based approaches.

Introduction

Ribonucleic acid (RNA) has long been the “center” of the central dogma of molecular biology. Genes are transcribed into messenger RNAs (mRNAs) in the nucleus, and these mRNAs carry the message to be decoded by ribosomes for translating proteins in the cytoplasm, the end-effector products that execute cellular functions. There are several mechanisms that regulate the process and efficiency of transcription,¹ and translation,² ultimately controlling gene expression to the right levels for normal physiological function. The primary transcripts also undergo splicing,³ capping, polyadenylation tail processing,^4,5 RNA editing,⁶ and RNA modifications,⁷ before becoming mature and functional mRNAs, in which these processes also confer additional platforms for modulating gene expression. RNA, being a vital player in biology, did not stop with the discovery and characterization of mRNA in 1961.⁸ Noncoding RNAs include transfer RNAs (tRNAs) and ribosomal RNAs (rRNAs) that are fundamental to translation, small nuclear RNAs (snRNAs) as constituents of spliceosome for alternative splicing,^3,9 microRNAs (miRNAs) that are essential innate post-transcriptional regulators of mRNAs,^9,10 and also the advances in next-generation sequencing technologies have assisted in rapidly expanding the members of long noncoding RNAs (lncRNAs) and circular RNAs (circRNAs), the relatively newer classes of noncoding RNAs with diversified cellular functions^11–15 and many more waiting to be annotated.

As both coding and noncoding RNAs are substantial components of the complex network of gene expression regulation and function in cells, disruptions of RNA-dependent biological processes or faulty transcripts could be associated with many human diseases.^13,15–17 For example, metabolic reprogramming observed in cancers, which is orchestrated by intricate regulation of noncoding RNAs to meet energy demands, facilitate cancer cell proliferation, and further contribute to the immunosuppressive tumor microenvironment.¹⁸ Understanding how RNA controls or modulates gene function has been and will continue to be a promising approach for developing therapeutics, especially for hard-to-treat diseases.^19–22

Conventional small-molecule drugs are external entities designed to interact with biological targets, primarily proteins such as receptors and enzymes, through their chemical structure and properties to modulate cellular activities. However, they are limited in scope due to the difficulty of defining the desired targeting interface or active site of some proteins from their intrinsic structures, rendering those proteins undruggable.²³ On the other hand, being already innate functional molecules, RNA-based drugs offer unprecedented opportunities to modulate gene function directly at the nucleic acid level through universal yet specific Watson-Crick base pairing, or to interact with proteins through their secondary structure, thereby enabling targeting of the underlying disease-associated perturbations.²⁴ The FDA has approved several RNA-based therapeutics across various modalities that can restore lost gene function or mitigate toxicity arising from aberrant gene expression.²⁴ These modalities include antisense oligonucleotides (ASOs) for knocking down gene expression, directing specific splicing pattern or the first-in-class oligonucleotide enzyme inhibitor, RNA interference (RNAi)-based using siRNAs for degrading target RNAs, aptamers that fold into conformations complementary to their target proteins to inhibit activity, synthetic mRNA to serve as blueprint for antigen production in the coronavirus disease 2019 (COVID-19) vaccines, and the single guide RNA (sgRNA) crucial for providing CRISPR/Cas gene-editing system specificity. LncRNAs, circRNAs, and small activating RNAs (saRNAs) are other RNA modalities that are currently undergoing therapeutic development in clinical trials.

The prompt availability of the mRNA-based vaccines against COVID-19, caused by SARS-CoV-2 (Pfizer-BioNTech’s BNT162b2 and Moderna’s mRNA-1273) under emergency authorization was undoubtedly a major milestone in the history of RNA therapeutics. It boosted confidence in the potential of RNA therapeutics and demonstrated the integration of decades’ worth of concrete innovations from multiple disciplines, which brought about this success. Importantly, RNA-based therapeutic strategies have several advantages over small molecules: high specificity owing to complementary base pairing, programmability that allows for rapid adaptation to emerging diseases, and versatility that enables targeting of DNA, RNA, or proteins, as briefly mentioned above. Despite these advantages, RNA therapeutics also face several challenges, including susceptibility to nuclease degradation, which directly impacts their stability, inefficient drug delivery to intended targets, and immune-related problems. This review provides a snapshot of the timeline for some key events in RNA therapeutics development (Fig. 1), with a comprehensive overview of the design and mechanisms of RNA-based therapeutics (Fig. 2) by analyzing and presenting existing literature and progress made in preclinical and clinical studies, and how each RNA drug is commonly and/or uniquely optimized to alleviate its vulnerabilities. Additionally, this review outlines the major delivery systems developed and/or used for RNA therapeutics. Finally, we discuss the main hurdles RNA therapeutics still face and the emerging technologies that could help mitigate existing bottlenecks and move the field forward.

Fig. 1.

Timeline of the development of RNA therapeutics. Key background events and the first FDA-approved RNA drug of each category are indicated. Gray background represents research findings of different classes of RNA, green background represents the therapeutic development of RNA, and yellow background represents the ancillary development for RNA therapeutics. SNALP: stable nucleic acid lipid particles. Other abbreviations are indicated throughout the article. *: Under emergency use authorization. Created with BioRender.com

Fig. 2.

Overview of different RNA therapeutic agents in clinical and preclinical settings. RNA therapeutic agents (numbered in black circles) can modify gene expression relevant to diseases at several levels in cells, as indicated by the numbered open circles. This schematic figure summarizes, in general, the design and mechanism of each RNA therapeutic that we have covered in this review. Abbreviations are indicated throughout the article. Created with BioRender.com

Notably, the inclusive idea of RNA therapeutics could embrace RNA as therapeutic targets, rather than merely as tools, and target RNA-related processes, such as RNA modifications, which have also been implicated in gene expression regulation and disease.²⁵ Although disease-associated RNA molecules are strategically targeted by RNA therapeutics, those RNAs can also be targeted by small-molecule drugs, as exemplified by risdiplam-like compounds able to selectively bind to survival motor neuron 2 (SMN2) pre-mRNA at exon 7, 5′ splice site, and stabilize the interaction with U1 snRNA and promote splicing.²⁶ RNA modification systems can also be therapeutically targeted by small molecules, as reported in a study using STM2457 to inhibit the m⁶A (N6-methyladenosine) writer enzyme METTL3 (methyltransferase-like 3), which is linked to acute myeloid leukemia.²⁷ Nonetheless, we have focused on RNA therapeutic agents that act on their targets, including RNA, and on their application in RNA modification modulation, specifically with dCas13, in this review.

RNA therapeutics that have received FDA approval

Antisense oligonucleotide (ASO)

Brief background

ASOs are single-stranded oligo(deoxy)ribonucleotides typically ranging from 13 to 30 nucleotides in length that are designed to modulate gene expression by hybridizing to their RNA targets.²⁴ The discoveries of nucleic acid hybridization activity demonstrated in DNA double-helix structures, RNA duplexes, and DNA-RNA heteroduplexes via complementary base pairing have been foundational to molecular biology since the 1950s and 1960s.²⁸ The first experimental evidence of ASO-mediated inhibition was by Zamecnik and Stephenson in 1978; the replication of Rous sarcoma virus (RSV) in chick embryo fibroblast cells was inhibited with a complementary oligonucleotide designed to bind a region to disrupt circularization of RSV 35S RNA that appeared to facilitate viral replication.²⁹ A year later, Donis-Keller demonstrated that a DNA oligonucleotide hybridized to 5.8S rRNA from yeast or satellite tobacco necrosis virus RNA could direct RNase H-mediated cleavage of the target RNA, proposing a mechanism for ASO activity involving enzymatic degradation.³⁰ Throughout the 1980s, numerous studies investigated the use of ASOs in vitro to inhibit gene expression or elucidate their mechanisms.

In the 1990s, research into the therapeutic application of ASOs in vivo expanded significantly across various disease contexts, culminating in the initiation of clinical trials. Beyond the physiological studies confirming the effectiveness of ASOs and favorable phenotypic outcomes, research also focused on pharmacokinetics and pharmacodynamics of ASOs in various animal models. Preclinical pharmacokinetic studies of the antiviral ASO ISIS 2922 in monkeys,³¹ supported subsequent clinical trials that demonstrated safety and efficacy,³² resulting in the FDA approval of the first ASO and RNA drug fomivirsen in 1998.³² Currently, around a dozen ASO-based drugs have received FDA approval (Table 1).

Table 1.

FDA-approved RNA-based therapeutics^a

Therapeutic drug (brand name)	Modality	Known modification/ delivery	Route of administration	Target organ/tissue	Disease treated	Target gene/molecule	FDA-approved year	Reference
Fomivirsen (VITRAVENE)	ASO (RNase H)	PS/ naked	Intravitreal	Eye	Cytomegalovirus retinitis	CMV Immediate-Early 2 (IE2) mRNA	1998	28,32
Mipomersen (KYNAMRO)	ASO (RNase H)	2′-O-MOE, PS, m5C/ naked	Subcutaneous	Liver	Homozygous familial hypercholesterolemia (with hypercholesterolemia)	Apolipoprotein B-100 mRNA	2013	28,103
Eteplirsen (EXONDYS 51)	ASO (steric blocker - splicing)	PMO/ naked	Intravenous	Muscle	Duchenne muscular dystrophy (amenable to exon 51 skipping)	Exon 51 of dystrophin pre-mRNA	2016	19,28,99
Nusinersen (SPINRAZA)	ASO (steric blocker - splicing)	2′-O-MOE, PS, m5C/ naked	Intrathecal	CNS/motor neurons	Spinal muscular atrophy (amenable to exon 7 inclusion)	Survival motor neuron 2 (SMN2) pre-mRNA	2016	28,117
Inotersen (TEGSEDI)	ASO (RNase H)	2′-O-MOE, PS, m5C/ naked	Subcutaneous	Liver	Transthyretin amyloidosis (hereditary)	Transthyretin (TTR) mRNA	2018	28,105
Golodirsen (VYONDYS 53)	ASO (steric blocker - splicing)	PMO/ naked	Intravenous	Muscle	Duchenne muscular dystrophy (amenable to exon 53 skipping)	Exon 53 of dystrophin pre-mRNA	2019	28,114
Viltolarsen (VILTEPSO)	ASO (steric blocker - splicing)	PMO/ naked	Intravenous	Muscle	Duchenne muscular dystrophy (amenable to exon 53 skipping)	Exon 53 of dystrophin pre-mRNA	2020	28,115
Casimersen (AMONDYS 45)	ASO (steric blocker - splicing)	PMO/ naked	Intravenous	Muscle	Duchenne muscular dystrophy (amenable to exon 45 skipping)	Exon 45 of dystrophin pre-mRNA	2021	28,116
Tofersen (QALSODY)	ASO (RNase H)	2′-O-MOE, PS, m5C/ naked	Intrathecal	CNS	Amyotrophic lateral sclerosis	Superoxide dismutase 1 mRNA	2023	107,108,548
Eplontersen (WAINUA)	ASO (RNase H)	2′-O-MOE, PS, m5C/ GalNAc conjugate	Subcutaneous	Liver	Transthyretin amyloidosis (hereditary with polyneuropathy)	Transthyretin (TTR) mRNA	2023	106,548
Olezarsen (TRYNGOLZA)	ASO (RNase H)	2′-O-MOE, PS, m5C, m5U/ GalNAc-conjugated	Subcutaneous	Liver	Familial chylomicronemia syndrome, severe hypertriglyceridaemia	Apolipoprotein C3 mRNA	2024, 2026	109–111
Donidalorsen (DAWNZERA)	ASO (RNase H)	2′-O-MOE, PS / GalNAc-conjugated	Subcutaneous	Liver	Hereditary angioedema	Prekallikrein mRNA	2025	20,112
Imetelstat (RYTELO)	Oligonucleotide enzyme inhibitor	N3′ → P5′ thio-phosphoramidate oligonucleotide/ palmitoyl conjugate	Intravenous	Cancer cells	Myelodysplastic syndromes (MDS) with transfusion-dependent anemia	Human telomerase RNA subunit (hTR)	2024	119,120
Patisiran (ONPATTRO)	RNAi (siRNA)	2′-O-Me, 3′ end dT/ LNP (DLin-MC3-DMA, DSPC, cholesterol, PEG2000-C-DMG)	Intravenous	Liver	Transthyretin amyloidosis (hereditary with polyneuropathy)	Transthyretin (TTR) mRNA	2018	28,169
Givosiran (GIVLAARI)	RNAi (siRNA)	2′-F, 2′-O-Me, PS/ trivalent GalNAc conjugate on passenger strand	Subcutaneous	Liver	Acute hepatic porphyria	5′-aminolevulinate synthase 1 (ALAS1) mRNA	2019	28,172
Lumasiran (OXLUMO)	RNAi (siRNA)	2′-F, 2′-O-Me, PS/ trivalent GalNAc conjugate on passenger strand	Subcutaneous	Liver	Primary hyperoxaluria type 1	Glycolate oxidase mRNA	2020	28,174,175
Inclisiran (LEQVIO)	RNAi (siRNA)	2′-F, 2′-O-Me, 2′-H, PS/ trivalent GalNAc conjugate on passenger strand	Subcutaneous	Liver	Hypercholesterolemia and atherosclerotic cardiovascular disease	Proprotein convertase subtilisin/kexin 9 (PCSK9) mRNA	2021	28,178
Vutrisiran (AMVUTTRA)	RNAi (siRNA)	2′-F, 2′-O-Me, PS/ trivalent GalNAc conjugate on passenger strand	Subcutaneous	Liver	Transthyretin amyloidosis (hereditary with polyneuropathy, and hereditary/wild type with cardiomyopathy)	Transthyretin (TTR) mRNA	2022, 2025	28,144,170,171
Nedosiran (RIVFLOZA)	RNAi (siRNA)	2′-F, 2′-O-Me, 2′-O-methyl-4′-O-((methoxy)phosphoryl)methyluridine for 5′ end on guide strand, PS/ GalXC™ platform (GalNAc aminosugar conjugated adenosine or guanosine)	Subcutaneous	Liver	Primary hyperoxaluria type 1	Lactate dehydrogenase (LDH) mRNA	2023	176,549
Fitusiran (QFITLIA)	RNAi (siRNA)	2′-F, 2′-O-Me, PS/ GalNAc conjugate on passenger strand	Subcutaneous	Liver	Hemophilia A and B	Antithrombin (SERPINC1 gene) mRNA	2025	180,550
Plozasiran (REDEMPLO)	RNAi (siRNA)	GalNAc conjugate on passenger strand	Subcutaneous	Liver	Familial chylomicronemia syndrome	Apolipoprotein C3 mRNA	2025	181
Pegaptanib (MACUGEN)	Aptamer	2′-F, 2′-O-Me, and 3′-to-3′ linked dT terminal cap/ 5′ end 40 kDa PEG conjugate	Intravitreal	Eye	Neovascular (wet) age-related macular degeneration	VEGF-165 isoform	2004	28,191,232
Avacincaptad pegol (IZERVAY)	Aptamer	2′-F, 2′-O-Me, and 3′-to-3′ linked dT terminal cap/ 5′ end 40 kDa PEG conjugate	Intravitreal	Eye	Geographic atrophy secondary to age-related macular degeneration	Complement factor C5	2023	192,236
Tozinameran/ BNT162b2 (COMIRNATY, Pfizer-BioNTech)	mRNA (vaccine)	Cap 1 analog, 5′ UTR from human α-globin gene, 3′ UTR from human AES and MT-RNR1 gene, polyA tail (A30-linker10-A70), m1Ψ/ LNP (ALC-0315, DSPC, cholesterol, ALC-0159)	Intramuscular	Immune	Prevention of COVID-19	Encodes the SARS-CoV-2 spike protein	2021	281,284
Elasomeran/ mRNA-1273 (SPIKEVAX, Moderna)	mRNA (vaccine)	Cap 1, 5′ UTR contains GC-rich tract, 3′ UTR from human β-globin gene, polyA tail, m1Ψ/ LNP (SM-102, DSPC, cholesterol, PEG-DMG)	Intramuscular	Immune	Prevention of COVID-19	Encodes the SARS-CoV-2 spike protein	2022	280,284
mRNA-1345 (MRESVIA)	mRNA (vaccine)	Cap 1, 5′ UTR, 3′ UTR, polyA tail, m1Ψ/ LNP as in mRNA-1273	Intramuscular	Immune	RSV vaccine	Encode for RSV F protein	2024, 2025	285–288
Exagamglogene autotemcel (CASGEVY)	CRISPR (Cas9)	sgRNA is modified with PS and 2′-O-Me/ Ex vivo electroporation of RNP complex (gene editing)	Intravenous (edited cells)	Patient-derived CD34+ hematopoietic stem and progenitor cells, bone marrow	Sickle cell disease and transfusion-dependent β-thalassemia	Erythroid-specific enhancer of BCL11A	2023	345,346,350

^aAs of August 2026, including Casgevy as RNA-enabled CRISPR/Cas9 gene-editing therapy, which is setting the stepping stone for potential Cas13-based therapeutic applications in the future

Design and mechanisms

RNase H-dependent ASOs

RNase H is an endogenous nuclease that recognizes RNA-DNA hybrid duplexes as substrates and selectively cleaves the RNA strand.³³ This cleavage degrades the target mRNA and reduces the expression of its encoded protein. DNA-based ASOs exert their therapeutic effects by targeting mRNAs in a sequence-specific manner and recruiting RNase H for target degradation. These ASOs downregulate gene expression and are particularly useful for treating gain-of-function toxicities caused by genetic mutations or abnormal gene upregulation,³⁴ as well as for degrading exogenous viral RNA. Since RNase H is present in both nuclear and cytoplasmic compartments, ASOs can target many aberrantly upregulated RNA transcript involved in disease processes in these compartments.³³

Over the past four decades, numerous RNase H-dependent ASOs have been developed for various conditions. Recent preclinical examples include ASOs that reduce plasma triglycerides, total cholesterol, and non-HDL (high-density lipoprotein) cholesterol by targeting and decreasing mRNA levels of angiopoietin-like protein 3 (ANGPTL3) or ANGPTL4.³⁵ ASOs targeting the mRNA of hydrogen peroxide-inducible clone-5 (Hic-5), a gene associated with the progression from early-stage metabolic dysfunction-associated steatohepatitis (MASH) to advanced fibrosis, have demonstrated efficacy in reducing hepatic steatosis and fibrosis in mice.³⁶ ASOs designed to target α-synuclein mRNA have been shown to reduce α-synuclein protein levels and to be protective against the spread of fibril-induced pathology in mouse brain models of Parkinson’s disease.³⁷ ASOs targeting SARS-CoV-2 mRNA that were administered intranasally have demonstrated a reduction in viral load and infection severity in mouse models, with confirmed delivery to lung tissue.³⁸

Steric-blocker ASOs

ASOs can be designed to operate independently of RNase H through various mechanisms that provide steric hindrance at specific regions of the target transcript without inducing RNA degradation. Typically, this class of ASOs is composed of fully chemically modified oligonucleotides or analogs to avoid RNase H recruitment.³⁹ Steric-blocker ASOs offer greater flexibility than RNase H-dependent ASOs in modulating gene expression levels, either increasing or decreasing, via multiple strategies and targeting regions such as pre-mRNA splicing,⁴⁰ the 3′ polyA tail,⁴¹ translation,^42,43 and miRNA-binding sites.^44,45

One clinically successful strategy for therapeutic steric-blocker ASOs is to modulate the splicing of pre-mRNAs. A common genetic variant at the exon-intron junction of exon 6 in the ABCA4 gene results in a 35-nt elongation of exon 6, leading to premature termination of protein translation and a progressive retinal disorder. Screening 25 ASOs designed around this aberrant exon-elongation region identified a top candidate that restored approximately 50% of wild-type ABCA4 transcript and protein levels in patient-derived fibroblast-induced pluripotent stem cells differentiated into photoreceptor precursors and retinal organoids.⁴⁰ This strategy using ASO for modulating splicing was first demonstrated by Dominski and Kole in 1993.⁴⁶ A mutation in the first intron of β-globin contributing to β-thalassemia creates an additional 3′ splice site that uses the same normal branch point as the normal 3′ splice site uses. It competes and results in the preferred splicing event (90%). As one of the splicing corrections in the study, the authors designed an ASO to block the normal branch site so that the splicing machinery then selected a downstream cryptic branch point and restored normal splicing, while directly blocking the mutation did not restore splicing.⁴⁶ Splicing is generally affected by cis-acting elements in exons and introns that can function as splicing enhancers or silencers when interacting with the spliceosome.⁴⁷ For example, nusinersen ASO was designed to block an intronic splicing silencer in intron 7 to promote the inclusion of exon 7 of SMN2,^47,48 and eteplirsen ASO, which skips exon 51 of dystrophin, was initially designed around the basis of exonic splicing enhancers in exon 51 (Fig. 3a).^47,49 Scrutinizing and tiling targeting sequences are crucial owing to the sequence context and spatial orientation of both splicing elements and splicing machinery.

Fig. 3.

Overview of mechanisms of recent FDA-approved RNA drugs. a Steric-blocker ASOs designed to modulate alternative splicing. Nusinersen ASO is designed to block the intronic splicing silencer (ISS) element in intron 7 to promote the inclusion of exon 7 of survival motor neuron 2 (SMN2) and generate full-length SMN protein for treating spinal muscular atrophy (SMA) caused by loss-of-function mutations on SMN1. Eteplirsen ASO is designed to block the exonic splicing enhancer (ESE) element in exon 51 to promote the skipping of exon 51 of dystrophin and generate truncated but in-frame and (partially) functional dystrophin. This can treat Duchenne muscular dystrophy (DMD) caused by certain genetic frame-shifting mutations (depicted by the berry-colored line) on dystrophin that are amenable to exon skipping. Gray dashed lines indicate normal splicing, and red dashed lines represent drug-induced splicing. b Instead of normal conversion of glyoxylate to glycine by alanine-glyoxylate aminotransferase (AGT), genetic mutations of liver AGT cause primary hyperoxaluria type 1 (PH1) due to excessive accumulation of oxalate coming from unprocessed glyoxylate. The lumasiran siRNA drug is designed to knock down glycolate oxidase (GO) mRNA and its enzyme synthesis to lower the conversion of glycolate to glyoxylate, whereas the newer FDA-approved nedosiran siRNA drug is designed to target lactate dehydrogenase (LDH) mRNA, directly reducing the enzyme that catalyzes glyoxylate to oxalate. Both siRNAs are conjugated with GalNAc (depicted as the black hexagons) for liver targeting and can lead to a decrease in oxalate levels. c Geographic atrophy (GA) secondary to age-related macular degeneration has been linked to the contribution of complement cascade-mediated inflammation occurring in the eye. The most recent FDA-approved RNA aptamer, avacincaptad pegol, is selected for its interaction with complement protein C5 to prevent the activation and conversion of C5 into C5a and C5b by C5 convertase (C3 convertase + C3b) and to stop the complement cascade at this point. d Mechanisms of raising immunity against an antigen (here, RSV prefusion F membrane protein) that is encoded in an mRNA vaccine. Antigen encoding mRNA is often delivered via LNPs and is then translated using host machinery and can be translocated to the plasma membrane intrinsically and recognized by B lymphocytes through B cell receptors (BCRs), or its fragments from proteasomal degradation can be presented through MHC class I in any nucleated cells, more specifically, local muscle cells, for activating CD8⁺ cytotoxic T lymphocytes upon recognition. Fragments presented through MHC class II in professional antigen-presenting cells can activate CD4⁺ helper T lymphocytes for a coordinated immune response that includes promoting antibody production from B cells and the activities of cytotoxic T cells and macrophages. Immune memory or immunity is formed after challenge with the mRNA-encoded antigen. e CRISPR-edited autologous HSPCs can restore the expression of gamma-globin to form functional hemoglobin with alpha-globin in eligible patients and reduce symptoms from sickle cell disease and transfusion-dependent β-thalassemia. The general procedure entails mobilizing the patient’s HSPCs from bone marrow to blood for collection and isolation of CD34⁺ HSPCs. Editing thereby disrupts the erythroid enhancer region of BCL11A in HSPCs via Cas9 with the guidance of sgRNA. Myeloablation is conducted to eliminate original HSPCs prior to introducing the edited HSPCs (Casgevy) through intravenous infusion. Engraftment of these edited HSPCs by the patient’s bone marrow enables the proliferation of edited HSPCs with reduced BCL11A expression, leading to the differentiation of erythroid cells that have increased production of gamma-globin. Created with BioRender.com

Targeting and blocking the poly(A) signal site on the 3′ end of the long noncoding RNA nuclear paraspeckle assembly transcript 1 (NEAT1) using an ASO affects poly(A) tail processing and promotes transcriptional readthrough. This leads to increased levels of NEAT1_2, a longer isoform associated with differentiation pathways, counteracting the high NEAT1_1 levels linked to aggressive neuroblastoma phenotypes.⁴¹ Another strategy employs readthrough ASOs designed to bind downstream of premature stop codons, promoting translational readthrough of nonsense mutations and thereby restoring full-length protein synthesis.⁴² ASOs that target miRNA-binding sites in the 3′ untranslated region (UTR) can block endogenous miRNA interactions, leading to increased mRNA stability and translation. For example, an ASO targeting the MECP2 (methyl-CpG-binding protein 2) 3′ UTR blocked binding by miR-22-3p, miR-132-3p, and miR-483-5p and helped to alleviate MeCP2 deficiency in patient-derived fibroblasts of Rett syndrome.⁴⁴ Similarly, an ASO targeting the miR-29b binding site in the 3′ UTR of progranulin mRNA increased progranulin levels in a humanized mouse model. Since reduced progranulin is associated with frontotemporal dementia and Alzheimer’s disease, this approach may have therapeutic potential.⁴⁵

Besides the mode of masking miRNA-binding sites on mRNAs, another design involves ASOs directly binding to mature miRNA targets through complementary base pairing. These miRNA inhibitor ASOs, also known as anti-miRNA oligonucleotides (AMOs), neutralize the activity of endogenous miRNAs. Krützfeldt et al. pioneered in vivo therapeutic approaches by using cholesterol-conjugated single-stranded RNA AMOs, or antagomirs, to inhibit miRNAs systemically.⁵⁰ Given the involvement of miRNAs in numerous human diseases, these methodologies have significant therapeutic potential. For example, antagomir-122 was designed to silence the liver-abundant miR-122 and assess its effects on target gene expression and phenotypes, while antagomir-16 targeted the ubiquitously expressed miR-16 to evaluate its bioavailability in mice. The same antagomir-122 study also identified that cholesterol biosynthesis pathways were particularly affected, and treatment with antagomir-122 reduced plasma cholesterol levels in mice.⁵⁰ Notably, the first miRNA inhibitor ASO to enter clinical trials was miravirsen, designed to target miR-122 for the treatment of hepatitis C virus infection. Although miravirsen’s clinical development was later discontinued due to limited efficacy in reducing viral load, it successfully lowered miR-122 levels.⁵¹ There are a few other potential AMOs still in clinical development (NCT06979362, NCT06979375), such as the CDR132L AMO that targets miR-132 to treat patients with myocardial infarction,⁵² and RGLS8429 that targets the miR-17 family of miRNAs for treatment of autosomal dominant polycystic kidney disease (ADPKD) (NCT05521191),⁵³ which favorable tolerability was indicated.^52,53

Synthesis

The current standard method for oligonucleotide synthesis is the well-established phosphoramidite-based solid-phase synthesis.⁵⁴ In 1965, Letsinger and Mahadevan first described the use of polymer support for the stepwise synthesis of oligonucleotides,⁵⁵ a method known as solid-phase synthesis. Several key innovations had been developed to maximize synthesis efficiency, including the determination of attaching the support at the C3′-OH of the starting nucleoside and keeping the C5′-OH free for incorporating new nucleoside 3′-phosphate,⁵⁶ the use of phosphotriester chemistry to address side reactions seen with phosphodiester methods,⁵⁷ and the switch to phosphite triester chemistry for highly reactive P(III).⁵⁸ It was not until the advent of the phosphoramidite approach developed by Caruthers’s group in the early 1980,⁵⁹ that solid-phase synthesis of nucleic acids was finally fine-tuned.

The phosphoramidite synthesis starts with a nucleoside attached to a solid support at the C3′-OH, with the C5′-OH protected by a 4,4′-dimethoxytrityl (DMT) group. Detritylation exposes the C5′-OH, which then reacts with a phosphoramidite intermediate protected at C5′-OH and linked through the phosphorus atom to the oxygen of the nucleoside 3′-OH. A phosphorus-oxygen bond is formed with the C5′-OH via a coupling reaction catalyzed under specific conditions. This bond initially exists as a phosphite triester (P(III)), which is subsequently oxidized to a stable phosphate triester (P(V)). Unreacted 5′-hydroxyl groups are then capped to prevent further extension and to prepare for the next addition; detritylation of the incorporated nucleotide is repeated. Finally, the complete oligonucleotide is detached from the solid support and deprotected by removing the nucleobase and phosphate protecting groups. A key advantage of phosphoramidite chemistry is that these intermediates are stable and can be stored for long periods,^59,60 enabling automation of coupling.⁶¹

Although phosphoramidite synthesis has enabled large-scale and versatile oligonucleotide production for decades, it faces limitations in scaling up RNA therapeutics for broader patient populations, such as those with cardiovascular diseases or cancer, rather than rare genetic disorders. The increase in the use of greener methods is due to the use of non-environmentally friendly/harsh chemicals and increased energy consumption.⁶² Recent developments include enzymatic methods like employing ligases to assemble smaller but purer fragments into full antisense oligonucleotides,⁶³ the collaborative use of polymerase and endonuclease V to elongate self-priming hairpin templates that encode the ASO sequence and release it upon cleavage once complete,⁶⁴ and others. These innovative approaches are extensively reviewed in works by Obexer et al.⁵⁴ and Mohammed et al.⁶⁰

Optimization

The main goal of optimizing ASOs with chemical modifications is to enhance therapeutic potency by addressing the inherent vulnerabilities of oligonucleotides, especially RNA molecules, which have low chemical stability due to 2′-OH groups and are susceptible to endogenous endo- and exonucleases, affecting their metabolic stability. Additional common optimization strategies include enhancing binding affinity for targets,²⁸ improving cellular uptake through the cell membrane,⁶⁵ and reducing toxicities.⁶⁶ Numerous chemical modification options for nucleotides have been reported; however, here we focus on the most widely implemented ones used in ASO drug development and optimization (Fig. 4).

Fig. 4.

Common chemical modifications employed to stabilize RNA-based drugs. Modifications that are performed on the ribose ring of nucleotides are highlighted in green, on the nucleotide linkage backbone in purple, and on the bases in yellow. Common modifications on the ribose ring at C2 to mask the free 2′-OH of RNA are 2′-fluoro (2′-F), 2′-O-methyl (2′-O-Me), 2′-O-methoxyethyl (2′-O-MOE), 2′-amino (2′-NH₂), locked nucleic acid (LNA), and constrained ethyl (cEt) nucleotides. In glycol nucleic acid (GNA), the ribose ring is replaced by a glycol monomer unit. Unlocked nucleic acid (UNA) has an open ribose ring. Phosphorodiamidate morpholino oligonucleotides (PMOs) have a morpholino ring in place of the ribose, with a phosphodiester backbone replaced by a phosphorodiamidate linkage. Common backbone modifications of phosphodiester (PO) include phosphorothioate (PS), phosphorodithioate (PS2), and thiophosphonoacetate (thioPACE) or phosphonoacetate (PACE). An (E)-vinylphosphonate moiety at the 5′ end [5′-(E)-VP] of the siRNA guide strand enhances binding to Ago2, thus its selection and gene silencing. Cap 1 analog, 2S cap, and AvantCap are examples of 5′ cap (cap 1) analogs for synthetic mRNA to increase stability and translation. Pseudouridine (Ψ), N1-methylpseudouridine (m¹Ψ), and 5-methylcytidine (m⁵C) are modified bases that are commonly incorporated to reduce the immunogenicity of RNA drugs. An aromatic or aliphatic hydrophobic group is added through an amide linkage on deoxyuridine in the application of a slow-off-rate modified aptamer (SOMAmer) to introduce more functional groups aiding with target binding. N6-methyladenosine (m⁶A) is a common natural RNA modification. 7-methylguanosine (m⁷G) is the nucleobase on the 5′ cap. U in SOMAmer represents a modified deoxyuridine used in DNA aptamers. Created with BioRender.com

Phosphorothioate (PS) oligonucleotides are modified on their backbone linkage by replacing a non-bridging oxygen with sulfur. They are among the most widely used modifications in therapeutic ASOs.⁶⁷ PS modification provides ASOs with strong resistance to nuclease,⁶⁸ and enhances their ability to bind proteins, especially serum proteins. It has been reported that ASOs targeting intercellular adhesion molecule-1 (ICAM-1) mRNA with PS modification are strongly bound to plasma proteins, primarily albumin, with lesser binding to α2-macroglobulin.⁶⁹ Another study also outlined that PS ASOs frequently bind to albumin and histidine-rich glycoprotein.⁷⁰ Thus, PS modification significantly increases the metabolic stability of ASOs, resulting in longer half-lives in circulation and delayed clearance.^39,71 Although one drawback is that PS modification reduces pairing stability with the target,⁶⁸ this can be mitigated by other chemical modifications as described below.

Chemical modification at the C2 position of the ribose ring is a convenient site to optimize the stability of ASOs, as it can eliminate the 2′-OH-mediated strand cleavage to ameliorate chemical stability,⁷² hinder nuclease attacks on the nearby phosphate to improve metabolic stability,⁷³ and provide preorganization of the sugar ring to strengthen pairing stability.^74,75 The 2′-fluoro (2′-F),⁷⁶ and 2′-O-methyl (2′-O-Me),⁷⁷ are among the earliest effective chemical modifications on the C2 position that increase the overall stability of oligonucleotides. These modifications confer nuclease resistance and better pairing stability when binding to their complementary RNA targets via preorganizing the sugar pucker to form the C3′-endo conformation.⁷⁴ Another notable modification is the 2′-O-methoxyethyl (2′-O-MOE) modification, first reported in 1995.⁷⁸ This second-generation chemical modification has become widely used in the clinical development of ASO drugs due to several advantageous properties: improved hybridization affinity and greater specificity for RNA targets, increased stability against nucleases,^78,79 and higher tolerability in clinical applications.^80–82 Structure-based insights as revealed through the crystal structure of a fully 2′-O-MOE-modified RNA duplex confirmed the C3′-endo conformation of the sugar ring and showed extensive hydration mediated by MOE groups. These could explain the improved RNA-binding affinity and protection from nuclease attack on phosphodiester linkages.⁷⁵ In fact, a substantial portion of FDA-approved ASOs contain 2′-O-MOE modifications (Table 1).

The use of locked nucleic acid (LNA) is another popular modification due to its ability to enhance hybridization stability compared to other C2′-modifications, and its first application dates back to 1997.⁸³ Unlike the others that aid in preorganizing the ribose ring, LNA contains a 2′,4′-methylene bridge connecting the C2′-oxygen and the 4′ carbon of the ribose ring (Fig. 4) that further restricts the ribose to the C3′-endo sugar pucker conformation, which favors RNA-like duplex formation with complementary RNA targets.⁸³ This enhances binding affinity by reducing the entropic cost of hybridization, leading to increased melting temperatures of duplexes.⁸⁴ Consequently, LNA incorporation can allow for the design of shorter ASOs while maintaining target affinity. Obad et al. first demonstrated the functional use of fully LNA-modified 8-mer AMOs mainly targeting seed regions of oncogenic miR-21 and liver-specific miR-122.⁸⁵ CDR132L is an LNA-based AMO targeting miR-132, which is responsive to myocardial stress, and has been shown to be well-tolerated. Although the results from a phase II clinical trial treating left ventricular systolic dysfunction after myocardial infarction did not demonstrate statistical efficacy, more clinical assessments are still ongoing (NCT06979362, NCT06979375).⁵² LNAs are also frequently incorporated into the flanking regions of gapmer ASOs⁸⁶ to augment target-binding affinity and increase nuclease resistance.⁸⁷ Constrained ethyl (cEt) modification is a derivative of LNA, which displayed reduced hepatotoxicity in mice while preserving potency when compared to LNA.⁸⁸ Danvatirsen, or formerly AZD9150, is one example of a cEt ASO that targets STAT3 mRNA to reduce its protein and has been clinically evaluated in combination with anti-PD-L1 therapy.⁸⁹

In contrast to PS-modified DNA ASOs that retain the ability to engage RNase H action with a bound RNA target,⁶⁸ the C2 modifications on nucleotides make them resemble RNA more than DNA.³⁹ They can be suitably applied in steric-blocker ASOs, but it is not ideal for RNase H-dependent ASOs because a fully modified C2 position prevents the recruitment of RNase H.^41,90 To address this dilemma, gapmer ASOs (usually 20mer, 5-10-5) are designed with a DNA central region (the “gap,” usually 10mer) to recruit RNase H activity and are flanked by “RNA-like” segments on both sides (usually 5mer each side).⁹⁰ The RNA-like flanking segments could incorporate C2′ chemical modifications, including 2′-F, 2′-O-Me, 2′-O-MOE,⁹¹ or LNA,³⁶ for the abovementioned benefits. Gapmers have become the primary design approach for RNase H-dependent ASOs.^36,37,81,91 Bepirovirsen is a fully PS-modified gapmer ASO with 2′-O-MOE on the flanking regions, and it targets a conserved region that is present on all hepatitis B virus (HBV) transcripts.⁹² Recently, results from two phase III clinical trials showed that this antiviral drug candidate achieved functional cure as defined by sustained HBV DNA levels below the detection limit and HBV surface antigen (HBsAg) loss in around 20% of the participating patients, whereas none were achieved with placebo (NCT05630807, NCT05630820).⁹³ Adverse events (AEs) were recorded more frequently with bepirovirsen, with injection site reactions being the most common milder AEs and increased levels of alanine aminotransferase (ALT) being the most common higher-grade AEs. Nonetheless, increased ALT was associated with decreased HBsAg levels and was regarded as a therapeutic response to bepirovirsen.⁹³

Phosphorodiamidate morpholino oligonucleotides (PMOs) are a class of ASOs considered more of a synthetic analog of nucleic acids, with the first precursor reported in 1989.⁹⁴ In PMOs, the ribose ring is replaced by a morpholino ring, and the negatively charged phosphodiester linkage is replaced by the uncharged phosphorodiamidate linkage (Fig. 4). Attaching the same natural purine and pyrimidine bases, PMOs can be used to design therapeutic steric-blocker ASOs capable of complementary base pairing with their target RNAs.^43,95 They are, however, incapable of triggering RNase H recruitment to the target site. PMOs possess multiple advantageous features for drug development, including high metabolic stability due to resistance to nucleases and other lytic enzymes,^96,97 high solubility in aqueous solution,⁹⁷ and greater binding activity than equivalent DNA-based ASOs.⁹⁸ PMOs gained the first FDA approval in 2016 for the steric-blocker ASO eteplirsen.⁹⁹

Modifications on nucleobases are also common, in addition to changes on the backbone and ribose ring. One frequently used modification is the methylation of carbon 5 in cytosine, known as 5-methylcytidine (m⁵C) (Fig. 4), which has been reported to exhibit higher stability, increased RNA-binding affinity, and the ability to reduce immune stimulation against toll-like receptors (TLRs) 7 and 9 recognition and activation within CpG motifs.^100–102

FDA-approved treatments

Fomivirsen was the first ASO drug approved by the FDA in 1998, which is given intravitreally to treat CMV retinitis. Fomivirsen targets and leads to the degradation of the viral region 2 of the major immediate-early transcriptional unit (IE2) that is essential for viral replication, therefore inhibiting viral activity.³² Other FDA-approved ASOs that use an RNase H-mediated mechanism for degradation of target mRNA are as follows. The second approved ASO, mipomersen, is indicated for familial hypercholesterolemia characterized by elevated apolipoprotein B-100 (apoB-100) and low-density lipoprotein cholesterol (LDL-C).¹⁰³ Liver apoB-100 serves as the core for assembling LDL and very-low-density lipoprotein (VLDL) particles, while mipomersen targets apoB-100 mRNA and reduces levels of atherogenic apoB-100-containing lipoproteins.¹⁰³ Transthyretin (TTR) protein is predominantly synthesized in the liver and exists as a native tetrameric structure.¹⁰⁴ However, wild-type TTR can become destabilized over time or due to inherited mutations in the TTR gene, leading to misfolded proteins that form amyloid fibrils.¹⁰⁴ Inotersen is approved for treating hereditary transthyretin amyloidosis (hATTR) by targeting TTR mRNA, thereby inhibiting TTR protein production and circulation.¹⁰⁵ Eplontersen is another ASO approved later that also targets TTR mRNA for treating hATTR. The main distinction between eplontersen and inotersen is that the former uses a targeting delivery strategy with an N-acetylgalactosamine (GalNAc) conjugate to facilitate uptake by hepatocytes. Eplontersen is advantageous over inotersen because it requires a lower, less frequent dose.¹⁰⁶ Tofersen is an ASO given intrathecally that targets superoxide dismutase 1 (SOD1) mRNA, approved to treat amyotrophic lateral sclerosis patients who have likely gain-of-function mutations in the SOD1 gene.^107,108 Olezarsen is approved for familial chylomicronemia syndrome (FCS) and severe hypertriglyceridaemia.^109,110 Genetic variants causing insufficient lipoprotein lipolysis function underlie FCS, which results in dangerous levels of triglycerides and increased risk for acute pancreatitis. Apolipoprotein C-III (APOC3) became a therapeutic target with olezarsen because of its position in regulating triglyceride metabolism. Olezarsen has consistently demonstrated efficacy in lowering triglyceride levels in clinical trials.^110,111 For prophylaxis against hereditary angioedema (HAE) attacks, the latest approved ASO is donidalorsen.¹¹² Activated coagulation factor XII converts prekallikrein to active kallikrein, which in turn converts kininogen into proinflammatory vasodilator bradykinin. HAE is caused by loss of function of C1 esterase inhibitor (C1-INH), which negatively regulates kallikrein and the generation of bradykinin. Donidalorsen, by targeting prekallikrein mRNA, can compensate for the missing role of C1-INH in limiting bradykinin production.^112,113

Besides RNase H-dependent ASOs, there are five steric-blocker ASOs that all modulate splicing and have been approved by the FDA (Table 1), and additionally, one patient-customized ASO. These splicing-switching ASOs modulate alternative splicing to skip or include a particular exon to at least partially, if not fully, restore the gene product and protein activity to improve symptoms. These ASOs provide a targeted therapeutic approach to mitigate the consequences of genetic mutations in the management of rare genetic disorders, alongside gene editing. Duchenne muscular dystrophy (DMD), caused by genetic frameshift mutations on dystrophin, can sometimes be amenable to skipping a particular exon to restore the reading frame. Although truncated, dystrophin could still be practically functional. Eteplirsen, which skips exon 51,^19,99 golodirsen,¹¹⁴ and viltolarsen that skip exon 53,¹¹⁵ and casimersen that skips exon 45,¹¹⁶ are all PMO ASOs administered intravenously. In contrast, to treat spinal muscular atrophy (SMA) caused by genetic mutations in SMN1, 2′-O-MOE- and PS-modified nusinersen induces the inclusion of exon 7 of SMN2 pre-mRNA to essentially produce full-length SMN protein (Fig. 3a).¹¹⁷ This is because SMN2 gene is a nearly identical copy of the SMN1 gene, but has a single nucleotide substitution in exon 7 that causes the exclusion of exon 7 during splicing and not generating the full-length efficiently. Similarly, FDA authorized Milasen for investigational use and it was the first N-of-1 treatment customized for Mila, a patient with Batten′s disease. It conveniently modeled nusinersen to generate an ASO that induces exon 6-exon 7 normal splicing in MFSD8 (major facilitator superfamily domain-containing 8) pre-mRNA.¹¹⁸ Both these ASOs are/were given intrathecally to reach the CNS.

Imetelstat is a first-in-class oligonucleotide therapeutic that employs sequence-specific, antisense-like hybridization to the RNA component of human telomerase (hTR/TERC), that is, the template region used for telomere synthesis, thereby competitively inhibiting telomerase activity catalyzed by the telomerase reverse transcriptase subunit (hTERT). However, the mechanism of action is as an enzyme competitive inhibitor, and the literature does not regard it as a conventional ASO-based therapeutic.^119,120 Imetelstat is unique in its chemical composition as an N3′ → P5′ thio-phosphoramidate oligonucleotide conjugated to a palmitoyl group at its 5′ end.¹²⁰ Increased telomerase activity due to TERT reactivation is often observed in cancers, which represents a therapeutic target. Imetelstat is currently approved for patients with lower-risk myelodysplastic syndromes and transfusion-dependent anemia.

RNA interference (RNAi)-based

Brief background

RNAi-based therapeutics are small double-stranded RNAs (dsRNAs) that target RNA transcripts for gene knockdown. The phenomenon known as post-transcriptional gene silencing (PTGS) or RNA interference (RNAi) was first documented in 1990, although the underlying mechanism was not understood at the time.¹²¹ In 1998, Fire and Mello published the first detailed report showing that hybridization between exogenous dsRNA and a homologous gene sequence led to high-specificity silencing of the endogenous mRNA transcript, followed by a corresponding decrease in protein levels in C. elegans.¹²² In parallel, an in-depth analysis of the endogenous PTGS mechanism in C. elegans began in the 1980s, and in 1993, Ambros and Ruvkun identified the first microRNA (miRNA), lin-4, as a non-protein-coding transcript regulating lin-14, thereby revealing a new class of noncoding RNA involved in cellular regulation.^123,124 By 2001, studies indicated that miRNA biogenesis and RNAi pathways intersect at the stage where Dicer processes RNA precursors for both mechanisms.¹²⁵

The utilization of RNAi for therapeutic purposes in different disease settings followed soon in the early 2000s,¹²⁶ with clinical trial reports showing efficacy and tolerability.^127,128 A phase I ascending dose trial targeting proprotein convertase subtilisin/kexin type 9 (PCSK9) with an siRNA drug showed an average 40% reduction in LDL-C levels.¹²⁸ The first clinical trial involving a miRNA mimic, MRX34, aimed to introduce endogenous tumor-suppressor miR-34a to treat patients with solid tumors. However, the trial was discontinued due to immune-related adverse events.¹²⁹ Patisiran was the first FDA-approved RNAi-based treatment, an siRNA drug authorized in 2018 to reduce TTR expression in transthyretin amyloidosis.¹³⁰ There are so far eight FDA-approved siRNA drugs (Table 1).

Design and mechanisms

MicroRNA (miRNA) mimic

Most miRNAs are transcribed by RNA polymerase II.¹³¹ In the nucleus, the primary miRNA transcripts (pri-miRNAs) are processed by the RNase III enzyme Drosha and the dsRNA-binding protein DGCR8,¹³² and cleaved into precursor miRNAs (pre-miRNAs), which harbor a stem-loop structure with mismatches. Exportin-5, along with Ran-GTPase, facilitates the exit of the pre-miRNAs from the nucleus to the cytoplasm,¹³³ where RNase III enzyme Dicer further cleaves pre-miRNAs into double-stranded miRNAs, which are around 22 nucleotides long, featuring a 3′ overhang of two nucleotides.¹⁰ The miRNA duplex is then loaded onto an Argonaute protein (Ago) with the guide (antisense) strand preferentially retained, containing a seed region spanning nucleotides 2 to 7 from the 5′ end, while nucleotide 8 is also critical for target interaction and forms the RNA-induced silencing complex (RISC) that is able to target RNAs.¹⁰ miRNAs generally silence gene expression via two main mechanisms: (1) Ago2-mediated cleavage of perfectly complementary mRNA targets to the guide strand, leading to degradation by cellular nucleases; and (2) translational repression and/or deadenylation, which can result in mRNA decapping and decay when there is incomplete complementarity outside the seed region.^134,135 A single miRNA can regulate multiple target genes due to imperfect complementarity outside the seed region; these targets are often associated with the same biological pathway. For example, miR-124 induces a brain-related expression profile, whereas miR-1 promotes a muscle-related profile in HeLa cells, with each targeting approximately 100 mRNAs.¹³⁶ Given the frequent, often imperfect, binding outside the seed region, Ago2-mediated cleavage is less common, and mRNA repression typically occurs through translational inhibition or deadenylation.¹³⁷

Aberrant expression of miRNAs is associated with many diseases, while AMOs can inhibit overexpressed oncogenic or disease-causing miRNAs; synthetic miRNA mimics aim to restore the function of downregulated tumor suppressors,^138,139 or other protective miRNAs. Notably, miR-124 exhibits neuroprotective effects by promoting an anti-inflammatory microglia phenotype (M2-like).¹⁴⁰ Delivery of miR-124 mimics via lipid nanoparticles to lipopolysaccharide-treated mice reduced neuroinflammation, as evidenced by decreased cytokine expression.¹⁴¹ Moreover, miR-125b-2-3p mimic showed potential cardioprotection in mice that underwent coronary occlusion and reperfusion. Treatment with miR-125b-2-3p mimic reduced the heart infarct size, although the molecular mechanism of the protective effect is yet to be investigated.²² Exosomal miR-574-3p derived from adipose tissue mesenchymal stem cells, or its mimics, exert chondroprotective effects in a rat model of knee osteoarthritis.¹⁶

Although miRNA mimics able to target multiple genes can effectively suppress specific pathways, this approach increases the risk of off-target effects. Therefore, careful design of the seed region and consideration of mismatches are essential. Designing miRNA mimics involves selecting the guide sequence, particularly the seed region, which can be identical to, shorter than, or longer than the endogenous sequence to modulate the number of targets. The passenger or sense strand can be modified to better match the guide strand to enhance duplex stability.¹⁴²

Small interfering RNA (siRNA)

Similar to endogenous miRNAs, siRNAs are exogenous small dsRNAs that mediate gene silencing through the Ago-associated RISC. In a therapeutic setting, exogenous siRNAs are administered and expected to incorporate into RISC, where the guide strand is selected to mediate sequence-specific gene knockdown and the passenger strand is cleaved and discarded. Hence, therapeutic siRNAs are developed to target upregulated RNA transcripts implicated in diseases.

Specifically, based on the rationale that chemokine CX3CL1 and its receptor CX3CR1 promote tumor progression by regulating multiple cancer-related pathways, delivering CX3CR1-targeting siRNA within extracellular vesicles (EVs) derived from M1-polarized macrophages induced CX3CR1 degradation and slowed the growth of pancreatic adenocarcinoma xenografts in mice.¹⁴³ Abnormal deposits of mutant or wild-type TTR protein cause ATTR, which can affect several organs and lead to polyneuropathy, cardiomyopathy, and vision impairment. Systemic delivery of TTR-targeting siRNA drugs (patisiran and vutrisiran) has been approved for treating ATTR by reducing TTR expression mainly in the liver (Table 1).^130,144 A novel lipid-conjugated siRNA designed to locally inhibit TTR in the eye has also been evaluated in mice as a potential ocular therapy.¹⁴⁵ siRNA targeting hepatic angiotensinogen effectively lowers arterial pressure and circulating angiotensinogen in rats,¹⁴⁶ with similar therapeutic effects observed in humans in a phase I study, without adverse events.¹⁴⁷

Unlike miRNA (mimics), which can have multiple targets with partial complementarity, siRNAs usually induce Ago2-mediated cleavage of a single fully complementary target RNA by design. This mechanism leads to target degradation and can reduce off-target effects.¹⁴⁸ The design of therapeutic siRNAs begins with selecting a guide sequence, typically 19–21 nucleotides long. Longer siRNAs, up to 30 nucleotides, can be more potent despite the risk of inducing interferon responses.¹⁴⁹ The siRNA duplex features two 3′ end overhangs, as in miRNAs, that facilitate recognition by the RNAi machinery.¹⁵⁰ Although siRNAs are designed to target specific RNA sequences based on the human genome, off-target effects can still occur, often due to the unintended incorporation of the passenger strand into RISC. The thermodynamic asymmetry of duplex ends affects strand selection, with the less stable 5′ end of one strand favoring its incorporation as the guide strand.¹⁵¹ Strategies for reducing off-target effects include chemical modifications, which are discussed in the following section.

Both miRNA mimics and siRNAs predominantly operate in the cytoplasm owing to the cellular localization of RISC assembly and can effectively silence cytoplasmic RNA targets. In contrast, RNase H-dependent ASOs have a broader scope of application against nuclear noncoding RNA targets.³³ However, the potential for nuclear RNAi could not be entirely excluded considering studies such as the first report of knockdown of small nuclear RNA 7SK and U6 via siRNA-programmed RISC,¹⁵² the functional existence of the RNAi factors in the nucleus for guided cleavage,¹⁵³ discovery of stress-induced response complex consisted of Ago as well as some transcription and splicing regulators can transport small oligonucleotides to nucleus for action,¹⁵⁴ to the recent demonstration that miRNA-mediated gene silencing can occur in nucleus, and with chromatin-associated RNA-like HMGA2.¹⁵⁵ Further research is needed to fully unveil the potential of nuclear RNAi mechanisms, including effects on gene transcription and splicing.¹⁵⁶

Synthesis

Small RNA drugs, including siRNAs and miRNA mimics, are primarily produced chemically through solid-phase synthesis, like ASOs. Additionally, they can also be synthesized enzymatically via in vitro transcription (IVT) or using plasmid or viral vectors.¹⁵⁷ Short hairpin RNAs (shRNAs) are encoded in vectors for host cells to transcribe and process into siRNA duplexes. They are effectively used to induce RNAi.¹⁵⁸ However, the choice of an appropriate delivery vehicle for in vivo shRNA-mediated gene silencing is crucial; lentiviral vectors have been shown to provide longer-lasting gene silencing.¹⁵⁹ A study compared chemically synthesized miRNA mimics with those produced via IVT, and demonstrated comparable efficacy between the two methods, with chemical synthesis offering a faster and more efficient production process.¹⁶⁰

Optimization

Although some chemical modifications used in ASOs to address innate susceptibilities are also available for optimizing dsRNA drugs, there are more complex considerations when optimizing siRNAs and miRNA mimics. These molecules consist of both guide and passenger strands and require key protein interactions with Ago and the guide strand selected to form RISC. Modification with 2′-F has been shown to be favorable on both passenger and guide strands of siRNA due to its small size, as well as enhanced duplex stability and greater knockdown efficiency.¹⁶¹ PS and 2′-O-Me modifications have also been widely incorporated onto siRNA therapeutics (Table 1).¹⁶² However, the bulkier 2′-O-MOE abolished siRNA activity and was not well tolerated on either strand.¹⁶¹

Ago2 interacts with the 5′ phosphate of the guide strand mostly through its Mid domain, which forms a binding pocket, with the PIWI domain contributing one side.¹⁶³ Modification at the 5′ end of the guide strand with (E)-vinylphosphonate (5′-E-VP) (Fig. 4) induces a subtle structural change at the binding pocket of Ago2 that increases its affinity for Ago2 interaction and enhances siRNA potency.¹⁶⁴ A recent study showed that the addition of 5′-E-VP allowed the incorporation of LNA at the 5′ end of the guide strand.¹⁶⁵ On the other hand, blocking the 5′-end phosphate of the passenger strand with a morpholino modification (5′-morpholino) promoted guide-strand selection, reducing off-target effects and enhancing overall RNAi activity.¹⁶⁶ Position 6 of the miRNA guide strand is the pivot to form a functional interaction between RISC and target, and there is a kink between positions 6 and 7 when bound to Ago2,¹⁶³ so destabilizing this region could diminish miRNA-like off-target effects for siRNAs.¹⁶⁷ Replacing with a single glycol nucleic acid (GNA) substitution, which lacks a cyclic sugar moiety and has the destabilizing attribute, at position 6 on an siRNA drug candidate reduced toxicity associated with off-target effects in rats and has reentered clinical development.¹⁶⁸

FDA-approved treatments

In 2018, patisiran became the first siRNA-based therapeutic approved by the FDA, administered intravenously for the treatment of ATTR (Table 1).^130,169 Patisiran siRNA drug contains a 2′-O-Me modification and is encapsulated in lipid nanoparticles for delivery to suppress TTR expression, thereby reducing circulating TTR and amyloid deposits in tissues.¹⁶⁹ In 2022, another siRNA drug, vutrisiran, was approved for treating ATTR with polyneuropathy,^144,170 and was approved for cardiomyopathy in 2025.¹⁷¹ Vutrisiran is chemically modified with PS linkages, conjugated to GalNAc for targeted hepatic delivery, administered subcutaneously, and does not require premedication, unlike patisiran.¹⁷⁰

Additionally, six other siRNA drugs have also received FDA approval. Deficiency in hepatic heme biosynthesis, caused by genetic mutations, upregulates the rate-limiting enzyme aminolevulinic acid synthase 1 (ALAS1), resulting in overproduction of neurotoxic heme intermediates δ-aminolevulinic acid and porphobilinogen, leading to acute hepatic porphyria (AHP). Givosiran, a GalNAc-conjugated siRNA targeting hepatic ALAS1, is administered subcutaneously for the treatment of AHP.¹⁷² Lumasiran is approved for primary hyperoxaluria type 1 (PH1), a genetic disorder caused by a deficiency of liver alanine-glyoxylate aminotransferase, which converts glyoxylate to glycine. In PH1, glyoxylate oxidation produces excess oxalate, leading to renal pathology.¹⁷³ This GalNAc-conjugated siRNA reduces hepatic mRNA encoding glycolate oxidase, thereby inhibiting glyoxylate and subsequent oxalate synthesis.^174,175 Nedosiran, another siRNA drug approved for PH1, targets hepatic lactate dehydrogenase to prevent glyoxylate conversion to oxalate (Fig. 3b).¹⁷⁶ Its developers utilize the GalXC™ platform for liver-specific targeting.¹⁷⁶ Blood LDL-C is taken up by hepatocytes via LDL receptor (LDLR)- mediated endocytosis for metabolism. PCSK9 mediates the degradation of LDLR, leading to elevated LDL-C levels and hypercholesterolemia. Inclisiran, a GalNAc-conjugated siRNA that targets hepatic PCSK9 mRNA, has been approved for treating hypercholesterolemia and atherosclerotic cardiovascular disease (ASCVD).^177,178

The latest two siRNA drugs approved by the FDA in 2025 are fitusiran, indicated for hemophilia A and B, and plozasiran for familial chylomicronemia syndrome (Table 1). They are also liver-targeting by bearing a GalNAc conjugate. Mutations in coagulation factors VIII (FVIII) and IX (FIX) cause functional deficiency and lead to Hemophilia A and B, respectively, which are bleeding disorders. Activated factors VIII (FVIIIa) and IX (FIXa) activate factor X, which in turn activates thrombin that promotes blood clotting by triggering the conversion of fibrinogen to fibrin.¹⁷⁹ As the liver produces the natural anticoagulant antithrombin, fitusiran alleviates hemophilia A and B through targeting SERPINC1 mRNA encoding antithrombin for degradation and lifting the inhibition of thrombin.^179,180 Plozasiran is based on the same rationale as that of RNase H-dependent ASO olezarsen, to degrade APOC3 mRNA, but through the siRNA-RISC mechanism to help combat high levels of triglycerides.¹⁸¹

Aptamers

Brief background

Aptamers are single-stranded oligonucleotides that selectively bind to molecular targets through conformational complementarity through their three-dimensional structures, in contrast to sequence complementarity. Systematic evolution of ligands by exponential enrichment (SELEX), first reported in 1990, is used to identify aptamers.^182,183 This process begins with synthesizing a library of oligonucleotides with random sequences flanked by fixed regions for downstream amplification. These random sequences typically can range from 20 to 100 nucleotides in length, providing a diverse range of structural candidates targeting specific molecules, primarily proteins, for clinical applications.¹⁸⁴ Bock et al. first reported DNA aptamers in 1992, identifying aptamers that bind and inhibit human thrombin activity in plasma.¹⁸⁵ Regarding RNA aptamers, the first breakthroughs date back to 1990, with independent laboratories isolating functional RNA aptamers. Tuerk and Gold identified RNA aptamers interacting with T7 RNA polymerase; one matched the sequence found in wild-type bacteriophage mRNA.¹⁸³ Ellington and Szostak identified RNA aptamers capable of binding various dye compounds that structurally mimic metabolic cofactors.¹⁸²

Aptamers quickly gained recognition as potential nucleic acid therapeutics.¹⁸⁶ Early studies in the 1990s focused on aptamers modulating protein activities involved in blood coagulation,¹⁸⁷ followed by applications in various diseases and conditions such as inflammation, infectious diseases, and neurodegenerative disorders.^188–190 Pegaptanib was the first FDA-approved aptamer, authorized in 2004, for the treatment of neovascular (wet) age-related macular degeneration,¹⁹¹ with the second RNA aptamer, avacincaptad pegol, approved in 2023 (Table 1).¹⁹²

Design and mechanisms

Aptamers function similarly to therapeutic monoclonal antibodies by binding their molecular targets with high affinity and specificity through conformational complementarity. The interactions between aptamers and their targets are primarily mediated by van der Waals forces, electrostatic interactions, hydrogen bonds, stacking interactions, and shape complementarity.¹⁹³ Their binding affinities typically range from low nanomolar to picomolar concentrations, comparable to those of monoclonal antibodies.¹⁹⁴ Aptamers possess several advantages over antibodies, which may confer a competitive edge. They are produced entirely through solid-phase chemical synthesis, eliminating the need for animal use and thereby enhancing manufacturing scalability, reducing batch-to-batch variability, and ensuring consistent efficacy. Additionally, aptamers are less sensitive to external factors such as temperature and can be chemically modified to improve stability and binding properties, similar to other RNA-based drugs.^194,195

Aptamers have garnered significant therapeutic interest due to their ability to specifically bind endogenous proteins through precise conformational interactions and modulate their activity. They can function as agonists, antagonists, or neutralizing agents. These molecules target a broad range of proteins, including proteases,¹⁹⁶ kinases,^197,198 cytokines,¹⁹⁹ cell-surface receptors,^197,198,200 and other types of proteins or molecules.²⁰¹

Antagonist

An aptamer that antagonizes TLR4 has been shown to improve symptoms in mouse models of experimental autoimmune encephalomyelitis (EAE) and cuprizone-induced demyelination. It also promotes the preservation and restoration of myelin and oligodendrocytes, which are essential for therapeutic efficacy. This TLR4-targeting aptamer presents promising potential as a treatment for multiple sclerosis, a neurodegenerative disease characterized by demyelination and autoimmune responses.²⁰⁰ Clinical trial reports have demonstrated the safety and pharmacokinetic profiles of this aptamer and have shown preliminary evidence of reducing mortality and disability in patients with acute ischemic stroke (NCT04734548).²⁰²

Agonist

EphA2 (erythropoietin-producing hepatocellular carcinoma A2) is a receptor tyrosine kinase that plays a role in tumor development and metastasis. Without its canonical ligand eA1, EphA2 interacts with E-cadherin and integrins. EphA2 is often overexpressed in tumors,¹⁹⁸ and is highly expressed in glioblastoma stem cells (GSCs), contributing to treatment resistance. RNA aptamers, along with their truncated derivatives, can bind EphA2 and inhibit cell growth in patient-derived GSCs. In silico modeling of the truncated RNA aptamer complexed with EphA2 indicated binding to the ligand-binding domain, suggesting that the aptamer mimics ligand binding and may act as an agonist to restore canonical signaling.¹⁹⁸ Additionally, the IR-A62 aptamer, which targets the extracellular domain of the insulin receptor, functions as both a positive and a negative allosteric modulator depending on insulin levels (low and high, respectively) and acts as an agonist to the insulin receptor.¹⁹⁷

Neutralizer

SARS-CoV-2 spike protein-expressing viral particles can be neutralized by aptamers selected against the receptor-binding domain (RBD) of their spike protein, thereby blocking the interaction with angiotensin-converting enzyme 2 (ACE2) receptors, which represent the primary entry point for infection.²⁰³ Additionally, aptamers can be engineered to target non-protein molecules such as small metabolites. Elevated glutamine levels have been associated with resistance to androgen receptor (AR) inhibitor therapy in prostate adenocarcinoma.²⁰⁴ A 56-mer RNA aptamer, selected and optimized for binding to L-glutamine, has been shown to restore castration sensitivity and limit proliferation of castration-resistant prostate adenocarcinoma in mouse models. Furthermore, functionalized gold nanoparticles carrying the aptamer demonstrated enhanced tumor-targeting capabilities. Glutamine sequestration also further reduced the activity of the transcription factor FOXM1 that activates resistance-related fibroblast growth factor 8, providing secondary therapeutic benefits.²⁰¹

Furthermore, advanced selection techniques, including cell SELEX and in vivo SELEX, facilitate the development of aptamers targeting complex biological molecules. These methods are particularly valuable when targets are difficult to purify in vitro, such as membrane proteins with heavy post-translational modifications like glycosylation, or when targeting proteins in their endogenous native state within their microenvironment.

Cell SELEX

Cell SELEX involves selecting aptamers against whole cells, targeting either surface or intracellular epitopes. For example, direct incubation of library sequences with the whole Mycobacterium tuberculosis H37Rv strain identified an aptamer that increased interferon-gamma (IFNγ) levels, while reducing bacterial load in the spleen of a TB-infected mouse model.²⁰⁵ Another example of cell-based SELEX using human U373-Magi-CCR5E cells identified an aptamer capable of binding to and inducing internalization of C-C chemokine receptor type 5 (CCR5). This aptamer further protected primary peripheral blood mononuclear cells (PBMCs) by neutralizing the infectivity of the HIV-1 R5 strain.²⁰⁶ Recently, the CW06 aptamer was identified by cell SELEX in the MCF7 breast cancer cell line, with its native target being glycosylated SLC25A24, a canonical mitochondrial carrier protein that was detected on the cell surface. CW06 treatment in xenograft-bearing mice showed reduced tumor growth accompanied by aptamer-mediated upregulation of SLC25A24, which could potentially play a regulatory role in the metabolism-senescence axis.²⁰⁷

In vivo SELEX

In vivo SELEX refers to the process of selecting aptamers within a living organism, where the target maintains its native conformation in physiological conditions and microenvironment. In the context of cancer therapeutics, a sophisticated screening of an RNA aptamer library was performed in a mouse model of intrahepatic colorectal cancer metastases via intravenous injections. RNA molecules were isolated from liver tumors after the first round, and this process was repeated for a total of 14 rounds to enrich for target-specific sequences. The final pool predominantly contained two RNA motif families, with the top candidates from each family selected for further characterization, and this study ultimately identified an aptamer that accumulated in tumors and later identified that its target is p68, an RNA helicase, within an in situ environment.²⁰⁸ Similarly, a bone and endothelial cell-targeting aptamer was successfully selected using a thio-modified aptamer library in a mouse model of prostate cancer bone metastasis.²⁰⁹

Bispecific aptamers

Proto-oncogene c-MET (c-mesenchymal-epithelial transition factor) is a transmembrane receptor tyrosine kinase that activates diverse biological responses through ERK and PI3K/AKT signaling pathways upon binding of hepatocyte growth factor cognate ligand. One approach involved generating a bispecific aptamer by linking an aptamer targeting c-MET with a tumor-specific nucleolin-binding aptamer. This conjugate can recruit MDM2 E3 ubiquitin ligase via nucleolin and promote ubiquitination and degradation of c-MET, which consequently reduces tumor size in xenograft models.²¹⁰ Another example employs a bispecific aptamer designed to simultaneously target PD-1 on T cells and PD-L1-expressing tumor cells to block this immune checkpoint inhibitory axis. Authors identified an aptamer against PD-L1 through cell SELEX using H460 lung cancer cells and linked it with a PD-1 aptamer.²¹¹

Synthesis

Functional RNA aptamers are generated through the SELEX process, which starts with libraries of random sequences. The construction of RNA libraries begins with solid-phase-synthesized DNA molecules. These DNA oligonucleotides are first amplified by large-scale PCR to generate enough material for multiple SELEX rounds, creating the DNA library used in DNA SELEX. Next, they are transcribed in vitro with RNA polymerase (e.g., T7 polymerase) to produce RNA libraries for RNA SELEX.¹⁸² Although large-scale PCR can sometimes decrease library complexity, especially in GC-rich regions, most libraries maintain a diversity of about 10¹³–10¹⁵ candidates, which has been sufficient for successful aptamer selection.^184,186

The entire library is incubated with the target molecule to separate bound from unbound sequences in washing steps. Bound oligonucleotides are then eluted and amplified via PCR-based techniques utilizing the fixed regions. This cycle of incubation and amplification is typically repeated for 8–12 rounds,¹⁸⁶ until aptamers are enriched, as indicated by a plateau in the elimination of library sequences. The final aptamers are individually characterized to further select for highest affinity and specificity. After identifying the best aptamer sequence, it can be scaled up again using standard solid-phase phosphoramidite synthesis.¹⁸⁴ Lately, UltraSELEX has been described as a non-iterative aptamer discovery method that applies gradient selection pressure to elute binders into fractions based on varying affinities or unbound states in a single cycle. The process incorporates high-throughput sequencing and the systematic gradient reproduction of enriching ligands (SGREELI) algorithm, which facilitates the scoring and ranking of aptamer candidates. This approach has been used as proof of concept to identify RNA aptamers targeting the SARS-CoV-2 RNA polymerase and HIV-1 reverse transcriptase.²¹²

Optimization

To improve chemical and metabolic stability, aptamers are often modified on the backbone with PS,²¹³ or phosphorodithioate (PS2),²¹⁴ and/or at the C2′ of the sugar ring, including 2′-F, 2′-O-Me, and 2′-amino (2′-NH₂) on pyrimidines, to increase resistance to nuclease degradation in serum and improve binding affinity.^215,216 Many of the aptamers discussed above are indeed modified with 2′-F at their pyrimidines.^198,208,217 A fully 2′-O-methylated RNA aptamer, Rondaptivon pegol (BT200), binds to the A1 domain of von Willebrand Factor (VWF). It underwent a phase II clinical trial for bleeding disorder hemophilia A (NCT04677803) and showed clinical efficacy in raising circulating endogenous FVIII, which normally complexes with VWF, or the administered FVIII protein therapeutics.²¹⁸ Spiegelmers, enantiomeric aptamers composed of L-ribose nucleotides (L-aptamer), are not recognized as substrates by nucleases.^184,219 Selection can begin by using mirror-image target proteins (composed of D-amino acids) to identify natural aptamers, after which the spiegelmer will recognize the natural protein target,²¹⁹ or directly using L-ribose nucleotides with mirror-image DNA polymerase.²²⁰ Olaptesed pegol is a chemokine CXCL12-neutralizing L-aptamer that has recently demonstrated safety and promising efficacy in a phase I/II clinical trial for the treatment of glioblastoma in combination with radiotherapy.²²¹ Other strategies to confer stability include capping at the 3′ termini with inverted (d)T to create 3′-to-3′ linkages, providing protection against 3′ to 5′ exonucleases,²²² or incorporating strong nuclease-resistant LNA.²²³

To expand the structural and conformational diversity of sequences in RNA libraries for selecting aptamers, other heavier modified nucleotides are incorporated, such as aminoallyl-deoxyuridine triphosphate (dUTP) derivatives²²⁴ and SOMAmers, or artificial nucleotides.²²⁵ SOMAmer, a slow-off-rate modified aptamer, features the use of dUTP derivatives that have additional functional groups to enhance target binding, similar to amino acid side chains.^222,226 To increase hydrophobic content, a range of aromatic and aliphatic hydrophobic groups are added through an amide linkage to the 5-position of uridine as dUTP derivatives for aptamer selection (Fig. 4).²²⁶ Libraries containing the four natural nucleotides and the artificial nucleotide 7-(2-thienyl)imidazo[4,5-b]pyridine (Ds), which has an unnatural base, have been used to select aptamers against VEGF-165 and IFNγ with more than a 100-fold improved affinity compared to using only the four natural nucleotides.²²⁵

Truncation is a common empirical step after identifying the best aptamer. Reducing the size of aptamers can optimize chemical synthesis for purer products and cost reduction and minimize unexpected interactions by retaining only the critical target-binding portion. Most aptamers end up being around 5–15 kDa after truncation.^184,186 To prevent rapid renal filtration, a highly effective method is to increase aptamer size without compromising activity, often through polyethylene glycol (PEG) conjugation,²²⁷ a technique also widely used in protein drugs to improve stability and solubility. Notably, the abovementioned rondaptivon pegol is a PEGylated aptamer, and because FVIII circulates in complex with VWF, the possibility of the 40 kDa PEG moiety indirectly contributing to the prolonged persistence of the VWF/FVIII complex by increasing hydrodynamic size and interfering with receptor-mediated clearance has been discussed, but remains to be validated.^218,228 However, sometimes the immune system reacts against PEG. Pre-existing antibodies against PEG have been observed in clinical trials,²²⁹ and anti-PEG antibodies could be diminishing the therapeutic potency of PEGylated aptamers or any other PEGylated drugs.^229,230

FDA-approved treatments

Pegaptanib was approved by the FDA in 2004 as the first RNA aptamer therapeutic for the treatment of age-related macular degeneration (AMD) through intravitreal injection.¹⁹¹ While normal vascular endothelial growth factor (VEGF) signaling is essential for regular blood circulation and wound healing, excessive VEGF activity is a primary factor in the pathogenesis of wet AMD, with VEGF-165 identified as a major isoform involved in abnormal neovascularization in AMD.²³¹ Pegaptanib is a 28-mer RNA aptamer conjugated with a 40 kDa PEG molecule at its 5′ end. This aptamer specifically binds to the VEGF-165 isoform, neutralizing it and thereby reducing pathological VEGF signaling.²³² However, pegaptanib was eventually discontinued due to intense market competition from anti-VEGF agents, such as monoclonal antibodies, which target multiple VEGF isoforms and demonstrate greater efficacy.²³³

Avacincaptad pegol received FDA approval in 2023 as the second therapeutic RNA aptamer indicated for geographic atrophy (GA) secondary to age-related macular degeneration, administered intravitreally.¹⁹² The complement cascade primarily functions in the detection and removal of pathogens as part of innate immunity.²³⁴ Cleavage of complement factor C5 is a key step in the cascade that induces inflammation, phagocytosis, and formation of the membrane attack complex, leading to cell death.²³⁴ However, complement cascade-mediated inflammation has been identified as a significant contributor to GA pathogenesis.²³⁵ Avacincaptad pegol is a 39-mer RNA conjugated with PEG that binds to complement protein C5 and inhibits its cleavage to prevent the formation of C5a and C5b (Fig. 3c), thereby disrupting downstream inflammatory processes.^192,236 Both these aptamers are chemically modified with 2′-F pyrimidines and 2′-O-Me purines and feature a 3′-to-3′ linked dT terminal cap.

Synthetic messenger RNA (mRNA)

Brief background

Single-stranded polynucleotide messenger RNA (mRNA) is the crucial mediator in the central dogma of molecular biology. The concept of “mRNA” was established in 1961 when the role of mRNA in transmitting genetic information became clearer.⁸ Later, in the 1970s, mRNA was characterized as comprising several components, including a 5′ cap, UTRs, a coding sequence (CDS), and a 3′ poly(A) tail.^5,237,238 These features contribute to mRNA stability, regulation, and translation efficiency. In parallel, Krieg and Melton established the method of in vitro transcription (IVT) to synthesize mRNA in microgram quantities from DNA templates.²³⁹

mRNA-based therapeutic concept was later supported by the first report of direct mRNA injection into mouse skeletal muscle and its expression in 1990.²⁴⁰ Subsequently, in 1992, Jirikowski et al. successfully reversed diabetes insipidus in rats via intrahypothalamic injection of vasopressin mRNA, demonstrating that the mRNA could remain active long enough to translate into functional vasopressin, thereby temporarily reversing phenotypic symptoms.²⁴¹ Together with the work of Wolff et al., this study established mRNA as a therapeutic platform. Studies have also indicated the potential of mRNA to serve as a vaccine that induces immune responses against the encoded antigen. In 1995, Conry et al. made the first attempt to use mRNA as a cancer vaccine by injecting human carcinoembryonic antigen (CEA) mRNA into mouse muscle, eliciting anti-CEA antibody responses after tumor challenge.²⁴² One phase I/II study among early clinical trials utilized synthetic mRNA encoding prostate-specific antigen (PSA) for cancer vaccine development, and evaluated both safety and effectiveness among the cohort (NCT00004211).²⁴³ Currently, although numerous mRNA-based therapeutics are still under clinical evaluation (NCT04917861, NCT06307431, NCT04503278, NCT06488313, NCT05295433, etc.), extensive research and development efforts on mRNA therapeutics culminated in the rapid availability of FDA-authorized mRNA vaccines for COVID-19 under an Emergency Use Authorization in 2020 during the pandemic.

Design and mechanisms

Protein replacement therapy

Synthetic mRNA can be used directly as a protein replacement therapy to ensure that cells express functional proteins essential for correcting the underlying deficiency in certain diseases. Many genetic mutations result in absent, nonfunctional, or abnormal proteins, disrupting normal biological functions. For example, an inherited metabolic disorder caused by mutations in the G6PC or SLC37A4 genes leads to glycogen storage disease type I, also known as glucose-6-phosphatase (G6Pase) deficiency. These genes encode the catalytic subunit G6Pase-α and the glucose-6-phosphate transporter (G6PT), respectively.²⁴⁴ Delivery of G6PC mRNA encapsulated in lipid nanoparticles to the liver has been shown to express functional human G6Pase, which successfully reduces hepatic abnormalities and improves fasting blood glucose levels in mouse models.²¹ Also, preclinical experiments in mice demonstrated that mRNA-based enzyme replacement therapy with two synthetic mRNAs encoding the human PCCA and PCCB genes, responsible for producing the α and β subunits of propionyl-CoA carboxylase (PCC), is well-tolerated and significantly enhances PCC activity. This approach shows potential as a therapeutic strategy for propionic acidemia/aciduria, an inherited pediatric metabolic disorder caused by PCC enzyme deficiency.²⁴⁵ Additionally, there are a few ongoing clinical trials for protein replacement therapy. ARCT-810 is an ornithine transcarbamylase (OTC)-encoding mRNA drug being developed by Arcturus Therapeutics, Inc. It is currently being evaluated in a phase II clinical trial for patients with hyperammonemia due to OTC deficiency (NCT06488313). ModernaTX, Inc. has also been assessing the long-term safety and clinical efficacy of mRNA-3705, which encodes methylmalonyl-coenzyme A mutase, for patients with methylmalonic acidemia due to deficiency of this enzyme (NCT05295433).

In certain cases, pathogenic conditions involve reduced protein activity resulting from gene downregulation. Cancer is characterized by loss of function in tumor-suppressor genes, such as the p53 gene, which plays a crucial role in regulating various cellular processes, including cell cycle control and DNA repair, and acts as a transcription factor. Utilizing a redox-responsive nanoparticle platform to deliver synthetic mRNA encoding p53 has been shown to induce cell cycle arrest and apoptosis in mouse models of hepatocellular carcinoma (HCC) and non-small cell lung cancer. Additionally, restoring p53 function enhances the efficacy of mTOR inhibitors in advanced stages of cancer.²⁴⁶

Vaccine application

Unlike replacement therapy, which involves encoding self-proteins to correct loss-of-function mutations, vaccine technology uses mRNAs to encode pathogen-derived or disease-associated antigens that stimulate the immune system (Fig. 3d). A well-known example is the COVID-19 mRNA vaccines developed during the pandemic that began in 2020, in which the mRNA encodes the viral spike glycoprotein. For other infectious diseases, an mRNA vaccine encoding the outer surface protein A (OspA) of the pathogen Borrelia burgdorferi has demonstrated a superior immune response in mice against Lyme disease after a single dose, compared to alum-adjuvanted recombinant OspA protein.²⁴⁷ Another candidate, BNT165b1, a malaria vaccine developed by BioNTech, has been evaluated in a phase I clinical trial (NCT05581641). This vaccine encodes the circumsporozoite protein present in Plasmodium falciparum, the malaria-causing parasite. Additionally, the Zika vaccine mRNA-1893 developed by Moderna has been evaluated in a phase II clinical trial (NCT04917861).

Synthetic mRNA vaccines can also be employed in immunotherapy, particularly for cancer treatment, to induce antibody or cytotoxic responses against tumor-associated antigens, such as cancer-specific antigens. Additionally, mRNA can encode and strategically induce self-proteins to modulate immune responses against disease. Specifically, mRNA-based DC vaccines leverage the antigen-presenting capabilities of DCs to initiate primary cytotoxic T lymphocyte (CTL) responses upon introduction of tumor antigens, leading to tumor rejection by CTLs.^243,248,249 For example, the clinical trial involving patient-derived DCs transfected ex vivo with mRNA encoding PSA successfully elicited T cell responses against PSA in patients with metastatic prostate cancer, with minimal toxicity or adverse effects.²⁴³ Type I interferon (IFN-I) signaling plays a crucial role in modulating cancer immunogenicity and responsiveness to checkpoint inhibitors. A recent study demonstrated that mRNA-loaded liposomes encoding the non-tumor-specific IFN-I driver antigen pp65 enhance antigen recognition and improve responses to immune checkpoint blockade in mice. This approach is based on the premise that lack of immune response contributes to tumor resistance; artificially inducing an early IFN-I response promotes epitope spreading and self-amplifying immune responses, thereby converting treatment-refractory tumors into responsive ones.²⁵⁰ T lymphocytes are another key cell type manipulated in immunotherapy, exemplified by chimeric antigen receptor (CAR) T cells. For instance, in a mouse model of heart failure, a study demonstrated the therapeutic efficacy of delivering modified mRNA via CD5-targeted lipid nanoparticles in vivo, leading to the generation of transient anti-fibrotic CAR T cells. The expression of anti-fibrotic CARs directed against fibroblast activation protein (FAP) enabled targeting of activated fibroblasts in vivo, thereby reducing fibrosis and restoring cardiac function after injury.²⁵¹ Additionally, a strategy employing CD8-conjugated lipid nanoparticles that deliver anti-CD19 CAR mRNA reprograms CD8⁺ T cells in rats and monkeys, providing a targeted delivery approach.²⁵² In vivo generation of CAR T cells through targeted mRNA delivery provides a more accessible therapeutic platform that could serve as an effective alternative to the complex ex vivo manufacturing process. Another innovative strategy is the employment of an mRNA vaccine encoding the oncofetal protein claudin 6 (CLDN6) to increase the antitumor activity of CLDN6-targeting CAR T cells against solid tumors with high CLDN6 expression. The potency of this vaccine is currently being assessed under a phase I trial (NCT04503278).²⁵³ Moreover, in a phase II clinical trial (NCT03897881), individualized neoantigen mRNA vaccine mRNA-4157 was evaluated in combination with pembrolizumab (antibody against PD-1) in patients with melanoma, and the vaccine group in combination with pembrolizumab demonstrated prolonged recurrence-free survival compared to pembrolizumab alone.²⁵⁴ As of the August 2026 news release, mRNA-4157 (intismeran autogene) showed promising clinical outcomes from a phase III trial for high-risk melanoma (NCT05933577).

Synthesis

The synthesis of polynucleotide mRNA (typically ranging from 1 to 15 kb),²⁵⁵ differs from the solid-phase synthesis method, which is more suitable for much smaller oligonucleotides up to around 100 bases. Instead, synthetic mRNA is produced via IVT. A linear DNA template encoding the target protein, such as linearized plasmid DNA or PCR products, is generated for an RNA polymerase (T7, T3, or SP6) to transcribe the template into RNA in the presence of ribonucleoside triphosphates as building blocks. The immobilized co-tethered T7 RNA polymerase-linear DNA complex, being also reusable, has recently been demonstrated to efficiently produce synthetic RNA transcripts that have robust expression in mice.²⁵⁶ In addition, the 5′ cap structure can be added either cotranscriptionally by supplementing the reaction with cap analogs or post-transcriptionally using capping enzymes.²⁵⁷ Similarly, the poly(A) tail can be encoded directly into the DNA template for a controlled length or appended afterward with polyadenylation enzymes but could result in a range of lengths.²⁵⁷

Purification of IVT mRNA products removes reaction-derived contaminants like DNA templates as well as short and/or incomplete mRNA or dsRNA byproducts to ensure the highest quality for clinical applications and prioritize safety and efficacy. Besides DNase treatment and mRNA precipitation, liquid chromatography (LC)-based methods are frequently employed for enhancing the purity of mRNA,²⁵⁸ such as size-exclusion chromatography,²⁵⁹ and affinity chromatography with oligo(dT).²⁶⁰ Reverse-phase high-performance LC (HPLC) has been shown to further rid residual dsRNA and other contaminants from IVT mRNA and mitigate potential immune activation responses.²⁶¹ Additionally, mesoporous silica-based liquid chromatography employing spermidine-mediated RNA adsorption has emerged as another purification approach for IVT mRNA, enabling efficient removal of dsRNA contaminants.²⁶² The immunostimulatory dsRNA byproducts could mainly arise from the loopback at the 3′ end of the IVT product and the incidental RNA-dependent RNA polymerase activity displayed by T7.^261,263 Apart from optimizing the purification strategies, engineering T7 variants directly to reduce the ability of byproduct formation offers another perspective.²⁶³

Optimization

Translation efficiency and mRNA turnover rate are two critical factors that fundamentally determine the productivity of mRNA therapeutics. These features are determined by all components of a mature mRNA, that is, the 5′ cap, 5′ UTR, CDS, 3′ UTR, and 3′ poly(A) tail. Stability dictates the availability of intact molecules that can reach target cells; partially degraded mRNAs are translated inefficiently or are likely to produce faulty proteins. Once inside the cell, stability and translation efficiency remain vital to ensure accurate protein synthesis. Therefore, the design principles of mRNA therapeutics focus on optimizing these components to produce stable, efficient, and functional mRNAs.

The 7-methylguanosine (m⁷G cap) is linked to the 5′ end of the mRNA via a triphosphate bridge (ppp) to the first nucleotide (N) of mRNA (cap 0, m⁷GpppN), where cap 1 and cap 2 structures that have 2′-O-Me modification on the first or first two nucleotides, respectively, are common in higher eukaryotes.²⁶⁴ Decapping enzymes, such as Dcp1/2 and DcpS, hydrolyze the triphosphate linkages between the α- and β-phosphates and between the β- and γ-phosphates, respectively, thereby removing the m⁷G cap.²⁶⁵ Researchers have been developing novel cap analogs (Fig. 4) to enhance their affinity for cap-binding protein eIF4E (eukaryotic initiation factor 4E) to increase mRNA’s translational efficiency and resistance to decapping enzymes, which delays mRNA decay.^4,266 For instance, 2S cap analogs containing the 1,2-dithiodiphosphate moiety at the α,β positions of the triphosphate bridge or at the α,β and β,γ positions of the tetraphosphate bridge contribute to mRNA stability and translation efficiency.⁴ Additionally, AvantCap is a trinucleotide cap analog featuring an N6-benzylated adenine base and a 2′-O-Me modification on the ribose of the second nucleotide (m⁷Gppp^Bn6A_mpG).²⁶⁶ The poly(A) tail is key to translation initiation and poly(A)-binding protein (PABP). It also protects mRNA from 3′ to 5′ exonucleases, ensuring its stability. The poly(A) tail typically averages around 200 adenine residues in mammals, and a length of 30 nt is generally required for stability and translation.⁵ Segmentation of the poly(A) tail with heteronucleotide spacers is a newer strategy to mitigate DNA plasmid instability during amplification in bacteria that results from bearing long homopolymeric As.²⁶⁷

The 5′ and 3′ UTRs are also essential for modulating/determining translation efficiency and mRNA stability. The 5′ UTR may contain elements such as internal ribosome entry sites (IRESs) and Kozak sequences flanking the AUG start codon that facilitate translation initiation. At the same time, the 3′ UTR typically affects mRNA stability and contains miRNA-binding sites.²³⁸ These regions can be optimized by designing or selecting specific UTR sequences known to enhance mRNA stability, ribosome binding, and translation initiation,²⁶⁸ such as 5′ UTRs of ribosomal proteins,²⁶⁹ or UTRs from complement factor 3 and cytochrome P450 2E1 (CYP2E1) genes,²⁷⁰ which have been shown to yield high expression levels. Screening of natural 3′ UTRs identified highly effective sequences, including the 3′ UTR of human β-globin, mitochondrial 12S rRNA (MT-RNR1), and the amino-terminal enhancer of split (AES).²⁷¹

In the coding region, key design considerations include codon optimization, nucleotide modifications, and the addition of signal peptides as needed, such as secretion signals.²⁷² Preferential codon usage among individual hosts is mainly attributed to differences in tRNA composition, which can make protein expression difficult in heterologous hosts.²⁷³ It has been reported that besides codons themselves, RNA secondary structure synergizes with codon usage to affect mRNA stability and ultimately protein translation.²⁷⁴ Ward et al. have recently reviewed algorithms that assist in analyzing codon usage and RNA structure for sequence optimization.²⁷⁵ It is important to balance codon optimization, but not to remove key sequences that could modify RNA modification/editing spots or secondary structure, which could affect stability and translation efficiency.^274–276 Exogenous mRNA is inherently immunogenic because it is recognized as a viral infection signal; however, this immune response can be mitigated by nucleotide modifications such as replacing uridine with pseudouridine (Ψ),²⁷⁷ which helps mRNA evade innate immune sensors.^102,277 Notably, N¹-methylpseudouridine (m¹Ψ), used alone or in combination with m⁵C in mRNA, has been reported to outperform Ψ in terms of gene expression and immunogenicity.^278,279 RNA structure and stability can also be improved with m¹Ψ modification.²⁷⁴

FDA-approved treatments

In 2020, FDA authorized the emergency use of two mRNA COVID-19 vaccines: Comirnaty (Pfizer-BioNTech, BNT162b2) and Spikevax (Moderna, mRNA-1273), administered intramuscularly to induce immunity against SARS-CoV-2.^280,281 Both vaccines received full FDA approval in 2021 and 2022, respectively.²⁸² They contain codon-optimized mRNA encoding the prefusion-stabilized SARS-CoV-2 spike protein as the antigen. This prefusion spike protein, which facilitates viral entry into host cells via the ACE2 receptor, is stabilized with two proline substitutions.²⁸³ BNT162b2 is delivered in LNP composed of ALC-0315 (ionizable lipid), DSPC, cholesterol, ALC-0159 and having optimizations including incorporation of m¹Ψ for U, cap 1 analog for 5′ cap, 5′ UTR derived from human α-globin RNA, 3′ UTR composed of segments from the human AES and MT-RNR1 genes, and a poly(A) tail of 100 adenosine residues segmented with a 10-nt linker.^281,284 mRNA-1273 is also delivered in an LNP and is composed of SM-102 (an ionizable lipid), DSPC, cholesterol, and PEG-DMG. It features incorporation of m¹Ψ as well but has a natural mammalian cap 1 and 3′ UTR from human β-globin. The 5′ UTR and poly(A) tail composition of mRNA-1273 have not yet been fully disclosed, with only the presence of a GC-rich tract in the 5′ UTR known.^280,284

Based on favorable clinical trial outcomes,²⁸⁵ the FDA also approved mRNA-1345 (MRESVIA) in 2024 for intramuscular injection to protect adults aged 60 and older against respiratory syncytial virus (RSV),²⁸⁶ and in 2025 expanded to adults between the ages of 18 and 59 who are at higher risk.²⁸⁷ This LNP-encapsulated mRNA vaccine encodes the RSV F membrane glycoprotein stabilized in the prefusion conformation (preF), capable of eliciting neutralizing antibodies against both RSV A and B strains.²⁸⁵ mRNA-1345 was optimized from a prior candidate, mRNA-1172, which was already showing promising clinical effectiveness, to further enhance its immunogenicity. Codons were redesigned, having structural influences from mRNA in consideration,²⁷⁴ and the cytoplasmic tail of preF was deleted. This vaccine contains m1Ψ modification, standard 5′ and 3′ UTR elements, an encoded poly(A) tail, and is capped with cap 1. Current literature indicates that the LNP formulation appears to be the same as that of mRNA-1273.²⁸⁸

CRISPR/Cas

Brief background

In 1987, Ishino et al. first reported an unusual sequence motif in the 3′ flanking region of the IAP gene of E. coli, comprising repetitive 29-nucleotide sequences separated by 32-nucleotide spacers.²⁸⁹ These repetitive sequences with spacers were subsequently observed across various bacterial and archaeal species,²⁹⁰ and were formally designated as clustered regularly interspaced short palindromic repeats, “CRISPR,” in 2002.²⁹¹ By 2007, it was recognized that CRISPR systems function as prokaryotic adaptive immune mechanisms, whereby organisms incorporate fragments of exogenous DNA as spacers into their CRISPR arrays, enabling sequence-specific recognition and defense against invading genetic elements.²⁹² The entire CRISPR locus with integrated foreign DNA is transcribed into CRISPR RNA (crRNA) and processed by particular CRISPR-associated proteins (Cas) with endonuclease activity to produce mature crRNAs, approximately 20 nucleotides long, capable of recognizing target DNA.²⁹³ Some Cas proteins are involved not only in crRNA maturation but also serve as the effector molecules that cleave foreign DNA via double-strand breaks (DSBs).²⁹⁴ Different bacterial and archaeal species may harbor multiple CRISPR/Cas loci encoding distinct molecular machineries that execute slightly different mechanisms, as summarized by Xu and Li.²⁹⁵ The discovery of trans-activating crRNA (tracrRNA) outlined a novel pathway wherein tracrRNA guides the maturation of crRNA with the assistance of endogenous RNase III, obviating the need for other Cas proteins besides the effector Cas.²⁹⁶ Effective targeting also requires the presence of a protospacer adjacent motif (PAM), typically 2–5 nucleotides, located adjacent to the target sequence, which activates Cas nuclease activity.²⁹⁷

In 2013, the successful validation of a prokaryotic gene-editing system in mammalian cells marked a pivotal advancement in human genetic engineering.²⁹⁸ The Zhang laboratory transfected human and mouse cell lines with a CRISPR/Cas9 system comprising tracrRNA, pre-crRNA, and Cas9, resulting in targeted DNA cleavage.²⁹⁸ The therapeutic potential was also quickly realized within the same year. A Mutated cystic fibrosis transmembrane conductance regulator (CFTR) gene within CF patient-derived intestinal stem cells was corrected by employing the CRISPR system, which also showcased the potential for personalized gene therapy.²⁹⁹ Shortly, the first clinical trial was initiated in 2016. The trial aimed to disrupt the PDCD1 gene (programmed cell death protein 1), which encodes the PD-1 immune checkpoint protein, in autologous T cells isolated from patients with refractory non-small-cell lung cancer.³⁰⁰ Although the authors recommended employing more advanced gene-editing methods to enhance therapeutic efficacy, the study established the overall safety and feasibility of CRISPR/Cas9-mediated gene editing in this context.³⁰⁰ The first FDA-approved CRISPR treatment was recently achieved in 2023 (Table 1).

Design and mechanisms

Cas9

In gene-editing applications, crRNA and tracrRNA together form the single-guide RNA (sgRNA),²⁹⁷ which is the crucial component of the CRISPR/Cas system that renders specificity. The sgRNA is designed to guide the effector Cas9 nuclease to the targeted genomic locus based on complementary base pairing. Following DSB cleavage, the cell’s DNA repair mechanisms are activated. When a repair or donor template of interest is supplied, cells can use homology-directed repair (HDR) to insert the desired sequence at the DSB site, resulting in a gene knock-in. For example, researchers successfully integrated a functional α-globin gene and a truncated erythropoietin receptor into the β-globin locus of hematopoietic stem and progenitor cells (HSPCs) derived from patients with α-thalassemia major.³⁰¹ In the absence of a repair template, cells typically repair DSBs via non-homologous end joining (NHEJ), which can lead to a higher degree of insertions or deletions (indels) and gene disruption, effectively resulting in gene knockout. An example is the disruption of oncogenic receptor-interacting protein kinase 4, which enhanced the sensitivity to cisplatin- or doxorubicin-induced apoptosis in melanoma cell models.³⁰²

The blunt-end DSB produced by wild-type Cas9 results from its HNH catalytic domain cleaving the DNA strand complementary to the sgRNA, while its RuvC catalytic domain cleaves the non-complementary strand.²⁹⁷ Mutations in either of these domains eliminate cleavage activity on that respective strand, creating a Cas9 nickase (Cas9n) that only cuts one DNA strand without causing a DSB. The main advantage of using Cas9n is that the single-strand break (SSB) it causes is repaired predominantly through the cellular SSB repair machinery, consisting of components similar to the base excision repair pathway,³⁰³ thereby reducing the levels of indels.³⁰⁴ One famous therapeutic application of Cas9n is as a base editor (BE),^305,306 which involves fusing it to cytidine deaminase or adenosine deaminase to mediate C to U conversion (ultimately T) or A to inosine (I) conversion (ultimately read as G), respectively. Liu’s team first reported the use of Cas9n fused to rat cytidine deaminase APOBEC1 guided by sgRNA for base editing. They also published a BE that fuses Cas9n with an engineered tRNA adenosine deaminase capable of A-to-I conversion in DNA.^307,308 Prime editing (PE) is another promising therapeutic gene-editing strategy that uses Cas9n fused to reverse transcriptase (RTase), guided by a PE guide RNA (pegRNA) to perform all types of DNA editing, including insertions, deletions, and point mutations, without causing DSBs.³⁰⁹ In PE, the design of pegRNA combines the functions of sgRNA to guide Cas9 to the target DNA, a sequence serving as a primer binding site, and a repair template for RTase. Cas9n nicks one DNA strand at the target site, and the resulting free strand hybridizes with the pegRNA’s primer binding site, creating a primer for RTase to extend this strand based on the repair template via reverse transcription. The nicked DNA strand now contains an unedited 5′ flap and an edited 3′ flap, which can later anneal with the non-nicked strand through thermodynamic equilibration, with assistance from endogenous endonuclease FEN1.³⁰⁹

Catalytically dead Cas9 (dCas9) completely loses its endonuclease activity due to engineered mutations in both catalytic domains.²⁹⁷ Since dCas9 retains DNA-binding ability, it was repurposed for regulating gene expression. Fusion of ten herpes simplex virus VP16 domains to dCas9, along with multiple sgRNAs targeting within 300 bp upstream of the transcription start site, successfully activated the expression of IL1RN, SOX2, and OCT4, either individually or in multiplexing in HEK293T cells.³¹⁰ Thus, a new therapeutic role emerged for dCas9 in delivering various transcriptional regulators for potential therapeutic purposes, aiming to correct gene dysregulation associated with disease. The PSA gene, often abnormally upregulated in prostate cancer due to high activity of the PSA promoter, was one therapeutic target. Fusion of dCas9 with the Krüppel-associated box (KRAB) domain of the transcriptional repressor Kox1 suppressed PSA gene expression at the transcriptional level, resulting in protective effects, including reduced growth and migration and increased apoptosis, in LNCap and PC3 prostate cancer cell lines.³¹¹ These strategies highlight the role of dCas9 in regulating gene expression both epigenetically and transcriptionally. dCas9 can also be applied in sgRNA-guided BEs. A new approach demonstrated that the use of dCas9 fused with chromatin-modulating peptides, in addition to cytidine or adenosine deaminases, helped resolve closed chromatin structures and resulted in comparable editing efficiency to that of Cas9n but greatly reduced unintended indels at target sites and off-target effects in cells.³¹²

Cas12

The discovery of multiple Cas12 subtypes as single effectors provides alternative options to Cas9 for gene editing. Target DNA recognition by Cas12/crRNA involves a thymidine-rich PAM, such as 5′-TTTV for Cas12a (formerly Cpf1),³¹³ which expands the range of target genes that can be edited or the design of target sites in genes, in contrast to Cas9, which recognizes a guanosine-rich motif.³¹⁴ Most Cas12 subtypes can cleave dsDNA, producing sticky-end breaks, and some have nucleolytic activity toward ssDNA as well.³¹⁵ An additional advantage of Cas12-based systems over Cas9 is their smaller size, especially subtypes like Cas12f, which range between 400 and 700 amino acids, making it easier to package these enzymes into popular but size-limited adeno-associated virus vectors.³¹⁶ Equipped with these features, Cas12s are involved not only in traditional gene editing but also in BE, acting as carriers of transcriptional regulators fused with dCas12,³¹⁷ and in PE with Cas12a.³¹⁸ Kleinstiver et al. engineered AsCas12a based on its structural insights to improve its editing efficiency and evaluated the abilities of multiplex gene editing, epigenetic editing with a catalytically inactive version fused to the synthetic tripartite activator VPR domain (VP64-p65-Rta) to upregulate gene expression, and base editing with fusions to the rat APOBEC1.³¹⁷ Liang et al. developed a series of intact or nickase Cas12a-mediated PE configurations, fused or not fused with RTase, and using circular RNA encoding the same information as pegRNA.³¹⁸

Cas13

Prokaryotes employ the CRISPR-Cas system to also defend against RNA invasion by cleaving the target ssRNA.³¹⁹ Molecular and crystal structure analyses of Cas13 have revealed that it possesses two distinct RNase catalytic activities: one responsible for crRNA maturation and the other for target RNA cleavage and degradation.³²⁰ This characterizes Cas13 as a single-component effector Cas that is programmable by a crRNA to recognize and cleave ssRNA targets. This unique tool achieves RNA silencing like other discussed RNA therapeutic approaches aimed at target degradation, making the Cas13 system even more relevant to RNA therapeutics. Codon-optimized Cas13a (formerly known as C2c2) from L. wadei achieved efficient gene knockdown in HEK293FT cells, comparable to RNAi, without detecting off-target effects. Authors also reported using nuclear localization or export signals to control the cellular localization of Cas13a’s activity, unlike RNAi-mediated knockdown, which typically occurs in the cytoplasm. Furthermore, multiplex targeting is possible due to Cas13’s ability to process pre-crRNA.³²¹ Another demonstration is the use of Cas13b crRNAs designed to target multiple conserved regions of HBV RNA intermediates coding for antigens HBeAg (core) and HBsAg (surface), leading to decreased antigen expression in an HBV-infected HepG2 cell model.³²² The use of high-fidelity Cas13d from R. flavefaciens targeting the transcript bearing an abnormal expansion of hexanucleotide repeat in chromosome 9 open-reading frame 72 (C9ORF72) gene, which is a common genetic cause for amyotrophic lateral sclerosis and frontotemporal dementia, alleviated pathology in mice.³²³

dCas13 is engineered to lose catalytic activity but retains RNA-binding ability.³²¹ It can affect the structure, function, or translation of RNA when fused with different RNA-editing effector enzymes. For example, dCas13b fused to a deaminase domain of adenosine deaminase acting on RNA (ADAR) has demonstrated A-to-I editing, which is read as G by ribosomes during translation.³²⁴ It has been used to correct disease-related G-to-A mutations, such as AVPR2 878G>A in X-linked nephrogenic diabetes insipidus.³²⁴ Brito et al. designed dCas13b fused to the m⁶A RNA demethylase ALKBH5 to study the relationship between the levels of the pathogenic Htt1a transcript variant in Huntington’s disease and its m⁶A modification in intron 1. Along with other methods, they demonstrated that m⁶A modification affects the expression of this pathogenic variant; however, demethylation via dCas13b-ALKBH5 reduced its production,³²⁵ suggesting the potential for demethylation as a therapeutic approach for abnormal m⁶A modification in diseases. Additionally, dCas13 fused with translational suppressor 4EHP (eIF4E-homologous protein) silenced the translation of endogenous mRNAs that were targeted by crRNA at start codons in HEK293 cells.³²⁶ Overall, crRNA-guided (d)Cas13a systems are a distinguished, versatile tool for regulating gene expression at several RNA levels with specificity, i.e., at RNA transcript, RNA editing, and RNA modification.

Synthesis

sgRNAs, usually about 100 nucleotides long,³²⁷ can be synthesized chemically through solid-phase synthesis, enzymatically via in vitro transcription (IVT),³²⁸ or biologically by encoding sgRNAs in DNA vectors for the recipient cells to produce them intracellularly using their transcription machinery.³²⁹ Although this method offers constant expression, viral vectors, particularly lentiviral vectors upon integration, might introduce mutagenesis at important regulatory and gene regions, potentially leading to oncogenesis.³³⁰ Besides linearizing DNA templates for IVT from sgRNA-encoded plasmids, generating DNA templates for sgRNAs and IVT can be simplified through an overlap extension PCR-based technique.³³¹ However, the 5′ triphosphate produced by IVT has been shown to cause inflammation.³³² Chemical synthesis may be a preferred approach for producing sgRNAs, as it tends to be less immunostimulatory with higher quality for large-scale production.³²⁷ Additionally, chemical modifications that enhance sgRNA stability and immunogenicity (discussed below) or confer other chemical functionalities can be easily incorporated during chemical synthesis.^333,334 Since the yield decreases as the oligonucleotide lengthens during synthesis, methods that attach smaller fragments to form the full sgRNA are advantageous.³³⁴ Tijaro-Bulla et al. describe an approach for sgRNA synthesis where each of the three initial fragments is synthesized via automated solid-phase synthesis, incorporating 3′-propargyl-5-methylcytosine and/or 5′-azidothymine for click chemistry, which are then assembled through a template strand for final covalent linkage.³³⁴

Optimization

Sequence selection is the initial step for optimizing sgRNA design to achieve high sensitivity and high specificity. CRISPR/Cas cleavage is mediated by the pairing of a 20-nt sequence at the 5′ end of the sgRNA and the target DNA, together with recognition of an adjacent PAM.³³⁵ However, off-target effects occur due to sequence-dependent factors, such as the number, position, and distribution of mismatches between the sgRNA and target DNA being tolerated.³³⁶ Researchers have developed various sgRNA design tools specifically for on-target identification and off-target prediction. DeepCRISPR is a computational platform that combines predictions for both on-target and off-target sites into a single framework, utilizing deep learning to analyze all possible genome-wide, unlabeled 20-nt sgRNAs with NGG PAM and integrated epigenetic features, followed by data-driven fine-tuning with existing labeled sgRNAs.³³⁷

After selecting the optimal sequence for sgRNAs, additional downstream chemical modifications are often introduced to address the innate instability and immunogenicity of RNA molecules. Chemical modifications at both the 5′ and 3′ ends of sgRNAs, such as 2′-O-Me on the ribose ring combined with PS or thioPACE backbones,³³⁸ improve genome editing efficiency in primary T cells and CD34⁺ HSPCs.^339,340 A structure-based approach to Cas9-sgRNA emphasizes the importance of 2′-OH and phosphate groups in the region of sgRNA that interact with the Cas9 protein via hydrogen bonds, while approximately 70% of the remaining sgRNA can be chemically modified with 2′-O-Me and 2′-F on the ribose, along with PS, to further enhance editing ability compared to modifications limited to the 5′ and 3′ ends, as demonstrated in mice.³⁴¹ Additionally, chemical modifications with cEt in the tracrRNA-binding region further increase gene disruption activity.³⁴² These modifications are known to enhance nuclease resistance, stability, and binding affinity to the target.³²⁷

One possible reason some sequences are resistant to editing or off-target effects is sgRNA misfolding. Creating a highly stable hairpin structure in the sgRNA can promote proper folding and boost genome editing efficiency, along with PS and 2′-O-Me modifications.³⁴³ Furthermore, unmodified sgRNAs have been shown to induce an IFN-I response in human PBMCs in vitro.³⁴⁴ These are recognized by the RIG-I (retinoic acid-inducible gene I) pathway of pattern recognition in the cytosol.³³² However, inflammatory signaling is mitigated by 2′-O-Me and PS modifications or by removing the 5′ triphosphate created during IVT synthesis of sgRNA.^327,332,344

FDA-approved treatments

Casgevy (exagamglogene autotemcel) became the first FDA-approved CRISPR/Cas therapy designed to treat sickle cell disease and transfusion-dependent β-thalassemia caused by insufficient functional β-globin levels.^345,346 Fetal γ-globin (HbF) production decreases after birth through repression by the BCL11A transcription factor,³⁴⁷ as part of normal gene regulation, while β-globin expression takes place. However, in disease settings without functional adult β-globin, this inhibitory mechanism is leveraged as a therapeutic target. Casgevy therapy utilizes Cas9 effector guided by sgRNA to edit the erythroid-specific GATA1 TF-binding site (master regulator of erythropoiesis³⁴⁸) on the enhancer region of BCL11A. The targeted GATA1 site is disrupted after Cas9 cleavage due to being predominantly repaired via NHEJ and consequently reduces the expression of BCL11A. HbF levels increase again after the BCL11A inhibitory mechanism is interrupted and can compensate for the loss of adult β-globin in patient-derived CD34⁺ HSPCs.³⁴⁹ The delivery of Cas9-sgRNA ribonucleoprotein (RNP) complex into patient HSPCs is achieved by ex vivo electroporation. After ablation of the original HSPCs, the edited autologous HSPCs (Casgevy) are transfused back into the patient for engraftment in the bone marrow so that they can proliferate (Fig. 3e). The sgRNA portion is synthesized using solid-phase synthesis with three 2′-O-Me and PS chemical modifications at each end.³⁵⁰

Delivery systems for RNA therapeutics

Cellular delivery of small RNA drugs in their “naked” form without a carrier or gymnotic delivery is possible and is facilitated by specific chemical modifications.^351,352 Evaluation of cellular uptake efficiency of various naked ASOs across different modification patterns in human cell lines confirmed that PS modification plays a crucial role in enhancing cellular uptake, likely due to its increased affinity for cell-surface proteins.³⁵¹ One primary mechanism for ASO internalization is via clathrin- or caveolin-mediated endocytosis, which involves interactions with cell-surface proteins.⁶⁵ In human cells, PS ASOs can interact with epidermal growth factor (EGF) and the EGF receptor to facilitate clathrin-mediated endocytosis.³⁵³ Additionally, productive uptake of PS/2′-O-MOE-modified ASOs has been demonstrated in mouse liver via clathrin- or caveolin-independent pathways, involving the adapter protein complex-2μ subunit (AP2M1).³⁵⁴ Delivering much larger-sized mRNA therapeutics in naked form presents significant challenges. Chemical modifications via the incorporation of Ψ or m¹Ψ can enhance mRNA’s intracellular stability, translation, and reduce immune activation but are usually insufficient to ensure efficient cellular uptake. However, some approaches have proposed this possibility. Abassi et al. evaluated a carrier-free mRNA vaccine designed to boost immunity against SARS-CoV-2. This vaccine targeted antigen-presenting cells located in the skin and was delivered via jet injection. Their findings demonstrated proof of concept delivery of naked mRNA, diminished local immune responses and tissue damage at the injection site, yet remained effective in reducing viral load in mice, as well as induced neutralizing antibodies in non-human primates.³⁵⁵

Incorporating functional conjugation and/or delivery vehicles can further enhance cellular uptake and delivery efficiency, especially for drugs other than ASOs. RNA drugs exhibit low permeability due to their negative charge and hydrophilicity, and they may elicit immunogenic responses. These challenges justify the development of specialized delivery systems for RNA-based therapeutics. Common delivery platforms include viral vectors, which are among the earliest gene delivery methods, as well as nonviral systems such as conjugates and carriers that include lipid-based nanoparticles, polymer-based nanoparticles, inorganic nanoparticles, and exosomes (Fig. 5). While viral vectors have significantly contributed to the progress of RNA delivery,^356,357 especially in the form of gene therapy, concerns remain regarding integrating viral vectors’ safety in humans, primarily due to their genome-integrating capacity.³³⁰ We limit our discussion here to the most studied nonviral carrier systems.

Fig. 5.

Common delivery systems for RNA therapeutics. a RNA drugs can be conjugated with PEG, lipids, antibodies, peptides, sugars, and/or aptamer moieties to enhance delivery efficiency. b Lipid nanoparticles (LNP) encapsulate RNA drugs within and are composed of four components: ionizable/cationic lipid, helper lipid, cholesterol, and PEGylated lipid. c RNA drugs can also be complexed with and loaded onto polymeric nanoparticles (PNPs). PNPs can also incorporate PEG onto the surface, and, like in LNPs, functionalization can be achieved through the PEG moiety. d Gold nanoparticles are commonly used to serve as an inner core that RNA drugs can complex laterally onto it, also known as spherical nucleic acid, via covalent bonds or electrostatic interactions with various types of linkers (as depicted by “ – ” or “ • ”, respectively). e Exosomes are naturally occurring EVs that carry a variety of cargoes between cells and have been conveniently adopted for RNA drug delivery. For example, miRNA mimics can be delivered to lung macrophages using mannose-decorated EVs that dock to mannose receptor C-type 1 (MRC1) on lung macrophages.⁴⁹⁹ Created with BioRender.com

Conjugates

Neutral, biocompatible, and hydrophilic polymer polyethylene glycol (PEG) has been extensively applied in various (macro)molecules, especially as conjugates for protein or peptide drugs to enhance solubility and stability in the bloodstream.³⁵⁸ PEGylation on RNA drugs provides protection against nucleases and renal filtration, thereby improving stability.³⁵⁹ PEGylation is commonly applied to aptamers for their optimization and delivery strategy. Both FDA-approved RNA aptamers are conjugated with PEG. A study reported an augmented half-life of up to 10 h with the use of a high-molecular-weight PEG conjugated to the aptamer compared to the non-functionalized aptamer, with a half-life of less than an hour.³⁶⁰ Lu et al. developed polymer-assisted compaction of DNA (pacDNA), where a small number of ASOs are linked to a brush-shaped PEG that shields ASOs from nonspecific protein interactions without impairing hybridization to their target RNA.³⁶¹ This brush-shaped PEG conjugation to ASOs targeting oncogenic KRAS mRNA enhanced tumor uptake and antitumor activity in mouse models bearing human non-small-cell lung carcinoma xenografts.³⁶²

Lipids such as fatty acids or cholesterol are another often employed conjugate to improve cellular delivery. Hydrophobic lipid moieties can facilitate interactions with plasma proteins and lipoproteins.³⁶³ For example, ASO conjugated with palmitic acid exhibits increased affinity for plasma proteins, including albumin, HDL, LDL, and IgG, which slows plasma clearance and urinary excretion in mice and increases tissue accumulation in the liver, quadriceps, and heart.³⁶³ Conjugation of cholesterol to a PS ASO targeting ICAM-1 facilitates cellular uptake into the liver, especially into Kupffer cells (liver macrophages) in rats.³⁶⁴ Lipid modification with a palmitoyl group at the 5′ end of a then oligonucleotide drug candidate GRN163 showed improved uptake into the xenograft tumors in mice and greater target inhibition.³⁶⁵ It is now the approved telomerase inhibitor imetelstat.

Peptide conjugates serve two main purposes: promoting cell penetration via cell-penetrating peptides (CPPs) or enabling interactions with cell-surface receptors to enhance endocytosis. CPPs are typically cationic and/or amphipathic peptides derived from natural proteins with translocating properties, aiding cargo translocation across membranes. They are particularly effective with neutrally charged ASOs such as phosphorodiamidate morpholino oligomers (PMO) and peptide nucleic acids (PNA).^95,366,367 For example, conjugation with the CPP Pip6a enabled the delivery of splicing modifier PMO ASO to reach the mouse brain and spinal cord via tail vein injection and induce exon 7 inclusion in the SMN2 transcript.³⁶⁶ Receptor-targeted conjugates include ASOs conjugated with peptides like bombesin to target G protein-coupled receptors such as gastrin-releasing peptide receptor, promoting receptor-mediated endocytosis.³⁶⁸ A functionalized lipid nanoparticle with an arginylglycylaspartic acid (RGD) peptide to target αvβ3 integrins improved the delivery of granulocyte-macrophage colony-stimulating factor (GM-CSF) mRNA to mouse endometrium.³⁶⁹

Antibody or ligand conjugates enable tissue-specific delivery of RNA therapeutics. Conjugating PMO to an anti-mouse transferrin receptor (TfR) antibody allows crossing of the blood-brain barrier.³⁷⁰ Similarly, an anti-CD71 (or TfR) antigen-binding fragment conjugated with siRNA targets cardiac and skeletal muscle tissues, demonstrating sustained gene silencing in these tissues for up to a month in treated mice.³⁷¹ Delivery of a tumor-suppressor miR-34a mimic conjugated with anti-CD47 antibody, an immune checkpoint protein CD47 that is overexpressed in triple-negative breast cancer (TNBC), successfully restored miR-34a’s therapeutic function in a mouse model by downregulating PD-L1 and promoting apoptosis.³⁷² GalNAc is an extensively studied ligand used for delivery of ASOs and siRNAs to the liver because of its high affinity for the asialoglycoprotein receptor (ASGPR) present on hepatocytes.^373,374 These conjugates enter hepatocytes via receptor-mediated, clathrin-dependent endocytosis and are metabolized to release the active RNA drug inside the cell. Triantennary GalNAc conjugation greatly improved hepatic delivery of ASOs and increased efficacy 7- to 10-fold for mRNA targets involved in liver diseases.³⁷³ Likewise, GalNAc-conjugated siRNAs on the passenger strand are internalized into endosomes through receptor-mediated endocytosis, and the acidic environment within facilitates the dissociation of the GalNAc-conjugated siRNA from ASGPR. Endosomal glycosidases cleave the GalNAc moiety at the β-O-glycosidic linkage to release siRNA drug.³⁷⁴ Recent efforts on chemically stabilizing this linkage suggested that the interaction between the antisense strand and Ago2/RISC, other than the stability of the GalNAc conjugate to the drug, is potentially more critical to enhance siRNA knockdown duration.³⁷⁴ As a matter of fact, seven currently FDA-approved siRNA drugs employ a GalNAc conjugate for liver-specific treatment, which are mostly for liver-associated metabolic diseases (Table 1).

Aptamers, being RNA therapeutics themselves, are often adopted for functionalizing other RNA drugs or vehicles for targeted delivery due to their high specificity to their ligands. Conjugating an aptamer that recognizes prostate-specific membrane antigen (PSMA) enables targeted delivery of siRNA to PSMA-expressing prostate cancer xenografts in mice.³⁷⁵ One recent study utilized two aptamers, TuTu22 and SYL3C, and one ligand, hyaluronic acid (HA), which targets EGFR, EpCAM, and CD44, respectively, to functionalize the carrier vehicle for CRISPR plasmids to knock out the c-Met gene in cancer cells.³⁷⁶

Carrier systems

Lipid-based and lipid nanoparticles (LNPs)

LNPs are the most common vectors used for mRNA vaccines delivery and can also deliver small RNA drug as done with siRNA therapeutic patisiran. LNPs are composed of four primary components: ionizable/cationic lipids, cholesterol, PEGylated lipids (PEG-lipids), and helper phospholipids (Fig. 5b).³⁷⁷ Ionizable/cationic lipids play a vital role in the encapsulation of negatively charged nucleic acids through electrostatic interactions and facilitate their delivery into the cytoplasm.³⁷⁸ Commonly utilized ionizable or cationic lipids include DLin-MC3-DMA, SM-102, ALC-0315, DOTMA, and DOTAP.^379,380 DLin-MC3-DMA is an ionizable lipid containing two unsaturated linoleyl-derived chains and a dimethylamino head group that becomes protonated at acidic pH, facilitating the encapsulation of negatively charged nucleic acids.³⁸¹ This ionizable lipid is clinically approved for formulating the LNP to deliver siRNA patisiran, whereas ALC-0315 and SM-102 are used for the formulations of LNPs to deliver the first two COVID-19 vaccines, BNT162b2 and mRNA-1273, respectively. Ionizable lipids are also believed to have chemical properties that may help with endosomal release.³⁸²

Cholesterol plays a critical role in the formulation of LNPs. Besides natural cholesterol, various cholesterol derivatives, including oxidized and esterified forms, have been studied for their effects on LNP structure and functionality. Cholesterol analog β-sitosterol has been shown to alter LNP morphology, resulting in faceted nanoparticles that enhance endosomal escape.³⁸³ Cholesteryl hemisuccinate, an ionic derivative, is frequently employed in pH-sensitive formulations to facilitate the release of nucleic acids.³⁸⁴ Oxidized cholesterol derivatives such as 7-ketocholesterol have been investigated for their ability to modify membrane packing and alter membrane-associated cellular processes.³⁸⁵ Overall, the choice of cholesterol derivative affects particle morphology, biodistribution, and transfection efficiency.

PEG-lipids are PEG moieties commonly conjugated through stable ester or amide bonds to lipids, ensuring robust incorporation into the lipid layer via the lipid part while extending the PEG chains outward to minimize immune recognition and enhance solubility and stability by creating an aqueous barrier.³⁸⁶ Typical PEG-lipids include DMG-PEG2000 (1,2-dimyristoyl-rac-glycerol-methoxypolyethylene glycol 2000) and DSG-PEG2000 (1,2-distearoyl-sn-glycero-3-phosphoethanolamine-PEG2000).³⁸⁷ The PEG moiety also serves as a site for conjugating ligands to target specific cells or tissues. A study involving encapsulated mRNA demonstrated that LNPs functionalized with PEG-folate and PEG-maleimide exhibited increased association and interaction with cancer cells compared to non-functionalized LNPs.³⁸⁸

Helper lipids are natural phospholipids like DSPC (1,2-distearoyl-sn-glycero-3-phosphocholine), which is used in the patisiran siRNA drug and the COVID-19 mRNA vaccines, and DOPE (1,2-dioleoyl-sn-glycero-3-phosphoethanolamine), which has been shown to increase luciferase mRNA delivery in vivo in substitution of DSPC.³⁸⁹ Their function has been less clear but is typically thought to provide structural scaffolding to assist LNP assembly and stability. However, they have started gaining focus as a point of development to enhance membrane fusion and facilitate endosomal escape.^389,390 Phospholipids have also been the sole component of liposomes, a subset of lipid-based nanoparticles characterized by their simple, uniform composition that mimics the structure of natural lipid bilayers. One example is the encapsulation of an EphA2-targeting siRNA therapeutic cargo within the aqueous core of a formulation using DOPC (1,2-dioleoyl-sn-glycero-3-phosphatidylcholine) natural liposomes, which is currently undergoing a phase I clinical trial (NCT01591356).³⁹¹

Polymer-based nanoparticles (PNPs)

Polymeric nanoparticles are a versatile class of carriers for gene and drug delivery due to their tunable physicochemical properties, biocompatibility, and potential for sustained release.³⁹² Cationic polymers such as polyethylenimine (PEI) or poly(β-amino esters) (PBAEs) are often used for nucleic acid delivery because of their positive charges, which allow for encapsulation of negatively charged nucleic acids through electrostatic interactions.³⁹³ Low-molecular-weight PEI has been modified with hydrophobic polycaprolactone to minimize cytotoxicity and maintain transfection efficiency when delivering siRNA to cells.³⁹⁴ Also, PBAE-PEG has been incorporated into the LNP platform for lung-targeting delivery of mRNA in mice.³⁹⁵ Biodegradable and biocompatible polymers, such as poly(lactic-co-glycolic acid) (PLGA), are also frequently employed as drug delivery carriers due to their ability to provide sustained release profiles and their extensive use in FDA-approved biomedical products.³⁹⁶ HA is a functional ligand for cell-surface CD44 receptors that are expressed on tumor cells. A hybrid nanoparticle composed of PLGA and functionalized with HA was utilized to encapsulate paclitaxel and focal adhesion kinase-targeting siRNA, which aimed to target CD44-positive chemoresistant ovarian cancer cells to enhance therapeutic efficacy.³⁹⁷ PLGA nanoparticles have been studied to deliver ASOs targeting MALAT1, where investigators demonstrated significantly greater ASO accumulation in multiple organs in mice compared to naked ASOs, highlighting the potential of PLGA-based nanoparticles to enhance delivery as well as offer a distinct biodistribution.³⁹⁸ Natural polymers inherently possess biocompatibility and biodegradability, which make them effective for use in nanoparticle-based delivery. However, natural polymers tend to aggregate and interact with biological components, leading to suboptimal delivery and reduced transfection efficiency. Several functionalization strategies, such as those applied to chitosan, can improve these properties and enhance transfection efficiency.³⁹⁹

Inorganic nanoparticles

Inorganic nanoparticles, such as gold (Au), silver (Ag), silica (SiO₂), and iron oxide (Fe₃O₄), are emerging as promising carriers for RNA-based therapeutics owing to their advantageous physicochemical properties, including high surface area, stability, and ease of functionalization. Although these nanoparticles are not yet approved for clinical use in RNA therapeutics, several preclinical studies are currently underway.⁴⁰⁰ These nanoparticles usually serve as an inner core where RNA drugs can complex laterally onto it (Fig. 5d). A recent innovative approach describes a double-layer carrier system comprising LNP encapsulating mRNAs that are first conjugated to gold nanoparticles (AuNPs) via the functionalized thiolated antisense leader sequence. This approach showed better delivery and expression of mRNAs in several human cell lines compared with delivery by AuNPs alone.⁴⁰¹ Mesoporous silica nanoparticles (MSNs) have also been developed to accommodate siRNA for delivery. Conventional MSNs have single-digit pore sizes in nm used for small molecules, but a study demonstrated that modifying the pore size of silica nanoparticles to ~23 nm can allow effective loading and release of small RNA therapeutics for gene silencing applications.⁴⁰²

Exosomes

Exosomes are cell-derived lipid-bilayer vesicles released into the extracellular space to facilitate communication between cells. Exosomes deliver proteins, lipids, and various nucleic acids, including RNA.⁴⁰³ Recipient cells take up exosomes through several mechanisms, including clathrin-mediated and caveolin-dependent endocytosis, pinocytosis, and, in some cases, direct fusion with the plasma membrane.^404–406 Exosomes are promising carriers for RNA drug delivery because they are naturally occurring, which may reduce immune response and cytotoxicity.^404,407 The lipid bilayer can protect RNA payload from degradation and provide stability in circulation to facilitate transport to recipient cells. Mesenchymal stromal cell (MSC)-derived exosomes enriched with a multi-functional miR-17-92 cluster, given through intravenous injection, improved both tissue and functional recovery for rats with traumatic brain injury. They observed neurogenesis, angiogenesis, neuroprotection, and anti-neuroinflammation compared to using exosomes alone. These findings indicate that exosomes can transport RNA cargo to injury sites and contribute to therapeutic efficacy.⁴⁰⁸ In some cases, exosomes exhibit inherent tissue specificity due to their affinity for the tissue from which they originated, enabling the targeted delivery of their payloads. For example, exosomes derived from M2 macrophages showed inflammation-tropism and anti-inflammatory capabilities after systemic administration.⁴⁰⁹ Similarly, exosomes isolated from hepatic stellate cells effectively delivered Cas9-sgRNA RNP and showed liver accumulation in mice.⁴¹⁰ Cancer (SKOV3 cell line)-derived exosomes delivering PARP-1-targeting Cas9-sgRNA-expressing plasmids also showed selective accumulation in mouse xenografts grown from ovarian cancer SKOV3 cells after intravenous administration and reduced tumor growth.⁴¹¹

As a matter of fact, there are a few early clinical trials assessing the delivery of therapeutic RNA cargo in exosomes,⁴⁰⁷ as follows. mRNA encoding for LDLR is packed in exosomes produced from normal donor bone marrow-derived mesenchymal stem cells (MSCs) for treating familial hypercholesterolemia as protein replacement for LDLR deficiency (NCT05043181),⁴¹² MSCs-derived exosomes encapsulating KRAS G12D-targeting siRNA for pancreatic cancer patients with KRAS G12D mutation (NCT03608631),⁴¹³ and MSCs-derived exosomes enriched with neuroprotective miR-124 for treating patients with acute ischemic stroke (NCT03384433).⁴¹⁴

Other developing RNA therapeutics

Long noncoding RNAs (lncRNAs)

Brief background

Expanding research on annotating and characterizing long noncoding RNAs (lncRNAs) has broadened the spectrum of RNA therapeutic agents as well as potential RNA targets implicated in diseases. LncRNAs are a diverse class of noncoding transcripts, typically exceeding 200 nucleotides in length, that participate in various functions and mechanisms of action. One of the first characterized lncRNAs, H19, was identified in 1984 and initially characterized as an oncofetal transcript due to its epigenetic regulation via genomic imprinting.⁴¹⁵ The discovery of Xist/XIST also revealed its epigenetic regulator role in orchestrating chromosome X-inactivation.⁴¹⁶ Subsequently, lncRNAs have been found to be dysregulated across various diseases, including multiple cancer types. For instance, by using tumor microarrays, it was found that in non-small cell lung cancer patients, metastasis-associated lung adenocarcinoma transcript 1 (MALAT1) was deregulated.¹³ Demonstration of targeting MALAT1 using LNA gapmer ASOs for degradation succeeded in reducing cell proliferation and inducing apoptosis in multiple myeloma models.⁴¹⁷ On the other hand, the effects of overexpressing the tumor-suppressor lncRNA growth arrest-specific 5 (GAS5) were evaluated in human melanoma cells using lentiviral delivery and resulted in inhibition of cell migration and invasion, primarily via downregulation of matrix metalloproteinase 2 (MMP2) expression.⁴¹⁸ These findings support the therapeutic potential of lncRNAs and suggest they could be developed as RNA replacement or overexpression therapeutics. Though lncRNA-based clinical trials are still a relatively unexplored area compared to other classes of RNA therapeutics, there have been a few studies conducted in recent years, as discussed in the following section.

Design and mechanisms

miRNA sponging

LncRNAs participate in competing endogenous RNA (ceRNA) interactions by containing miRNA-binding sites, which enable lncRNAs to act as molecular sponges to sequester miRNAs and thereby prevent miRNA-dependent inhibition of target mRNAs.⁴¹⁹ Ye et al. showed the oncogenic function of H19 in HCC, where elevated H19 levels promote the sponge-like function against tumor-suppressor miR-193b, thereby protecting MAPK1 and other oncogenes from miRNA-mediated inhibition. Elevated H19 correlates with increased tumor aggressiveness and a poorer prognosis in HCC patients.¹⁵ These findings suggest that H19 can be targeted by ASO or RNAi, which could restore the normal inhibitory regulation of MAPK1 by reducing miR-193b sponging. An observational clinical study has also explored the relationship between H19 and insulin-like growth factor 1 receptor (IGF-1R) in the context of HCC and diabetes mellitus (DM) (NCT04767750). Although the results are pending publication, this research could substantiate a pathophysiological link between HCC and DM.

Gene expression regulation

LncRNAs are capable of regulating transcription through either cis- or trans-regulatory functions, depending on the extent of their influence. DNA damage-induced PANDA (p21-associated ncRNA, DNA damage-activated) lncRNA is regulated by p53 and transcribed from the CDKN1A (which encodes p21) gene’s promoter. Upon induction, PANDA bound to the transcription factor NF-YA and inhibited its ability to activate pro-apoptotic genes.⁴²⁰ Hung et al. demonstrated that siRNA-mediated depletion of PANDA sensitized fibroblasts to doxorubicin-induced apoptosis, suggesting that targeting PANDA could enhance cancer cell susceptibility to chemotherapy.⁴²⁰ Additionally, a phase I/IIa trial involved H19-DT-A, also known as BC-819, a DNA plasmid utilizing the promoter of the extensively studied lncRNA H19 to drive the expression of diphtheria toxin A (DT-A) in tumor cells overexpressing H19. The activation of DT-A within tumor cells inhibits protein synthesis, leading to decreased tumor growth in ovarian and peritoneal cancers (NCT00826150).⁴²¹ BC-819 has also been tested in patients with recurrent nonmuscle invasive bladder tumors in a phase II trial (NCT00595088),⁴²² where it inhibited tumor growth in approximately two-thirds of cases and delayed recurrence, indicating its potential as a therapeutic option for bladder cancer.

Molecular scaffolding

Many lncRNAs exert their function as molecular scaffolds, providing structural platforms for the assembly of protein complexes, namely RNP complexes, and facilitating RNA-protein interactions within the nucleus, for instance, scaffolding transcriptional factors for transcription modulation,⁴²³ and bringing transcriptional activator and deubiquitinase together to lessen proteasomal degradation of the former in the cytoplasm.⁴²⁴ Chen et al. examined the cytoplasmic EV-mediated lncRNA-associated transcript 1 (ELNAT1) and found that it promotes SUMOylation and EV packaging in bladder cancer.¹¹ Their study revealed that ELNAT1 functions as a molecular scaffold by binding to hnRNPA1 and the SUMOylation enzyme UBC9, which directs its own incorporation into EVs. These ELNAT1-containing EVs were then internalized by lymphatic endothelial cells, where ELNAT1 enhanced SOX18 transcription, thereby promoting lymphangiogenesis and lymph node metastasis. These findings underscore the role of lncRNAs in regulating intracellular transport and signal transduction by acting as a scaffold. An observational clinical study was planned to investigate the potential of ELNAT1 as a preoperative biomarker for lymph node metastasis in bladder cancer patients, which may help reduce the need for invasive diagnostic procedures and could also become a therapeutic target in the future (NCT05270174).

Synthesis, optimization, and delivery

Although the recent lncRNA-based clinical trials that we discussed above are still mostly observational to confirm their association with diseases or as a bioindicator in nature, their crucial role in regulating gene expression through diverse mechanisms warrants their seat in the expanding RNA therapeutics field. Depending on which lncRNAs end up as therapeutic targets, ASOs and siRNAs can be employed for their knockdown. In contrast, synthetic lncRNAs can also be administered for their protective functions as therapeutic agents. Tailoring will be inevitable to match their unique characteristics for their deployment as RNA therapeutics; nonetheless, many clinically established methods can fast-track their synthesis, optimization, and delivery. Insights could be particularly informed by the most similar synthetic mRNA therapeutics in size context and structural considerations from aptamers. There is one study that demonstrates the delivery of antitumor lncRNA MEG3 (maternally expressed gene 3) using exosomes conjugated with DSPE-PEG2000-cyclo-RGDyK peptide to target αvβ3 integrin receptors on tumor cells. Mice treated with lncRNA MEG3-loaded exosomes showed a therapeutic effect on reduced tumor growth in osteosarcoma xenograft models.⁴²⁵

Circular RNAs (circRNAs)

Brief background

The observations of circular RNAs (circRNAs) in eukaryotic cells were first reported in 1979,⁴²⁶ in a study that showed the presence of circRNAs predominantly in the cytoplasmic fraction of eukaryotic cells under an electron microscope.⁴²⁶ Not until later in 1991 did the “serendipitous discovery” of abnormal splicing that resulted in misordered spliced exons help propose the possible mechanism that would generate circRNAs endogenously.⁴²⁷ The arrival of high-throughput RNA sequencing,⁴²⁸ greatly facilitated the research on annotating novel circRNAs. CircRNAs have now been well-demonstrated to be involved in gene expression regulation and influence various biological pathways in different disorder settings,⁴²⁹ and have brought forth their therapeutic potential.^430,431 One of the early phase I/II clinical trial registered to assess the tolerability and efficacy of a potential circRNA drug, RXRG001. It encodes aquaporin 1 and is intended to treat patients with radiation-induced xerostomia and hyposalivation (NCT06714253).

Endogenous circRNAs are typically generated from backsplicing events of precursor transcripts or from lariat precursor structures that escape debranching.⁴³² Similar to lncRNAs, circRNAs can execute their biological function as noncoding entities through varying mechanisms. They can modulate gene expression at multiple levels,^14,433,434 or modulate protein and enzyme activity as well as pathways,⁴³¹ by serving as a molecular sponge,⁴³³ or scaffold,¹² or through direct interaction.¹⁴ In addition, some circRNAs have the potential to code for peptides or proteins,^435,436 which may also affect gene expression,⁴³⁶ or serve as scaffolds.⁴³⁵ Thus, circRNAs could become promising therapeutic targets when aberrantly overexpressed or dysregulated,⁴³⁶ together with adopting them as therapeutic agents, either for normalizing the dysregulated levels of beneficial circRNAs or exploiting them to serve as a blueprint for protein expression like mRNAs.

Design and mechanisms

Protein-coding function

The leading RNA therapeutic development in circRNAs is rivaling mRNA therapeutics for encoding proteins, predominantly in vaccine applications. Highly purified circRNAs can offer greater metabolic stability and potentially less immunogenicity compared to linear mRNA counterparts because of their covalently closed structure, that is, lacking single-stranded free ends for exonucleases or particular pattern recognition sensors.⁴³⁷ Without a 5′ cap or a 3′ poly (A) tail, protein translation from circRNAs initiates through the recognition of IRES,⁴³⁸ and m⁶A-mediated mechanisms also play a role in promoting translation.⁴³⁹

Liu and Feng’s group developed a circRNA vaccine against Zika virus infection in mice. To avoid the common antibody-dependent enhancement adverse event seen with intensified dengue virus infection due to immunity to Zika virus, envelope protein domain III (EDIII) and nonstructural protein NS1 were rationally chosen for antigen encoding in circRNAs. The optimized vaccine, encapsulating the two antigen circRNAs separately in LNPs, conferred complete protection against Zika virus in both active and passive maternal immunization settings.⁴⁴⁰ RNA-binding protein MSI2 was implicated in chondrocyte metabolism and function but was downregulated in cartilage samples from patients with osteoarthritis, as was its downstream transcription factor SOX5. CircRNA coding for MSI2 or SOX5 and delivered in LNP intraarticularly for chondrocyte-specific targeting alleviated osteoarthritis in the mouse model as judged by several measurements and exemplified by the observation of lowered volume of calcification of the meniscus and extra bone.⁴⁴¹ This demonstrates the utilization of circRNA modality for protein replacement therapy. Furthermore, recent research led by Huang et al. discovered natural tumor-specific circRNAs that encode antigenic peptides and can be presented to induce antitumor T cell immunity. CircFAM53B was found to be expressed in breast tumors, and the detection of circFAM53B-specific CTLs in patients’ blood was associated with a better prognosis. A vaccine with this type of circRNA in mice reduced tumor growth and metastasis, in addition to enrichment of IFNγ-releasing CD4⁺ and CD8⁺ T cells.⁴⁴²

Gene expression regulation

Besides coding for proteins, therapeutic effects can be made use of the role of circRNAs in regulating gene expression. CircSMARCA5 formed an R-loop with its host gene SMARCA5 DNA through direct binding at exons 15 and 16, which resulted in transcriptional pausing of SMARCA5, thereby decreasing its full-length expression but increasing the short-lived truncated form. This gene expression regulation affected the integrity of the DNA repair role of SMARCA5, but overexpression of circSMARCA5 sensitized breast cancer mouse xenografts to cisplatin.¹⁴ CircRNAs can also modulate translation, as shown in a mechanistic study where circCLASP2 scaffolded DHX9 (DExH-Box helicase 9) and PCMT1 mRNA (protein-L-isoaspartate (D-aspartate) O-methyltransferase 1) together to enhance translation of the latter, owing to DHX9 helping resolve the RNA secondary structure on the PCMT1 mRNA translation initiation site.⁴³⁴ Additionally, circRNAs can regulate gene expression post-transcriptionally, not directly, but by sponging miRNAs. miR-340-5p was identified to negatively affect neuron regeneration-associated genes (RAGs), and an exogenous circRNA was designed and evaluated to sponge miR-340-5p. This synthetic Circ-340-5p increased the protein levels of miR-340-5p-targeted RAGs in injured mouse retinas compared with a control circRNA.⁴⁴³ Similarly, endogenous circ_0003251 was found to be downregulated in degenerated nucleus pulposus cells from patients. Complexed with PLGA, circ_0003251 was injected into the surgical intervertebral space at the tail of the rat model to evaluate its potential toxicity and efficacy. Results showed no significant pathological damage in various rat tissues, but sponging miR-637 elevated AKT1 protein levels, thereby promoting cell proliferation and exerting protective effects in vivo.⁴³³

Protein and pathway activity modulation

CircRNAs could also have therapeutic potential in modulating the activity of proteins or biological pathways. CircUTRN is an AR-inhibited circRNA that can bind to and weaken the activity of acetyl-CoA carboxylase 1, consequently impairing de novo fatty acid synthesis, which then creates vulnerability in both castration-sensitive and castration-resistant prostate cancer cells. Administration of both circUTRN and AR signaling inhibitors restricted the growth of tumor xenografts from castration-resistant prostate cancer cells.⁴³¹ A liver readthrough circRNA RB1-ITM2B (RCRIN) can bind to RPL8 and promote its degradation by also binding to and recruiting E3 ubiquitin ligase RNF2. RPL8 deficiency activated insulin signaling and reduced hepatosteatosis, whereas overexpression increased liver triglyceride levels and caused hepatomegaly in mice on a high-fat diet. RCRIN is found downregulated in patients with MASLD and is correlated with higher levels of RPL8. Direct injection of synthetic RCRIN into the mouse liver showed therapeutic effects on reduced liver lipid accumulation and damage.⁴⁴⁴

Synthesis, optimization, and delivery

In general, circRNA synthesis can be accomplished chemically or by methods involving enzymes or ribozymes.⁴⁴⁵ The latter two both start with in vitro transcribed RNA precursors, but the circularization is achieved using ligases like T4 RNA ligase,⁴⁴⁶ or a self-splicing reaction like that of group I introns, respectively.⁴⁴⁷ A ribozyme-based approach could be more practical in accommodating longer sequences for eventual circRNAs. Wesselhoeft et al. developed a platform to generate exogenous circRNAs up to 5 kb for protein expression based on a permuted group I catalytic intron mechanism and utilized strategies such as inclusion of homology arms and spacer sequences to appropriately orient catalytic sites and maximize the productivity of the self-splicing reaction.⁴⁴⁸ However, this permuted intron-exon approach leaves the exons, where the circRNA sequence is inserted, in the final product. Addressing this issue, a more recent system by Cui et al. described introducing the sequence to be circularized in a rearranged fashion downstream of the optimized ribozyme and adding an antisense region upstream of the ribozyme that is complementary to the backbone plasmid after the 3′ end of the circRNA sequence to stabilize the folding of the ribozyme.⁴⁴⁹ Human prostate cancer cell lines transfected with this construct to overexpress circFOXO3 demonstrated the native noncoding function of the circRNA through changes in the outcome of several phenotypic assays.⁴⁴⁹

Optimization of synthetic circRNAs often entails the selection of IRESs. IRESs can affect both translational efficiency of circRNAs and ribozyme activity of precursors due to being highly structured, and could interfere with overall folding and splicing sites.⁴⁴⁸ Encephalomyocarditis virus (EMCV) and coxsackievirus B3 (CVB3) IRESs are among the most utilized IRESs.^448,450 The abovementioned vaccine against Zika virus was optimized by switching to the IRES of human rhinovirus B3 and including sequences of an RNA-binding motif for the 5′ UTR, an eIF4G-recruiting aptamer at the IRES, and the 3′ UTR of human α-globin 1 to enhance translation.⁴⁴⁰ Delivery practices for circRNAs are mainly adopted from platforms used for mRNA therapeutics, especially LNPs.^432,451 A recent study dedicated to customizing LNP formulation specifically for circRNA cargo. In comparison with clinically approved LNP compositions of mRNA carriers, the authors showed a fine-tuned formulation using 12D6.2 ionizable lipid, SOPC helper lipid, and varied molar ratios between them and cholesterol for encapsulating and delivering circRNAs into immune cells in mice.⁴⁵¹

RNA activation (RNAa)

Brief background

miRNAs and siRNAs have been well known to mediate post-transcriptional gene silencing of their target RNAs through the RNAi mechanism. Nonetheless, in 2006, Li et al. reported the use of synthetic 21-nt small dsRNAs in human cell lines that were designed to target DNA at gene promoter regions and it activated genes at the transcriptional level in a sequence-specific manner, thus termed RNA activation (RNAa).⁴⁵² RNAa was soon validated independently by another study using small dsRNAs that targeted the promoter region of progesterone receptor and observed functional gene activation that had effects on cognate downstream targets.⁴⁵³ These small dsRNAs are known as small activating RNAs (saRNAs) and have been appropriated as a new therapeutic modality to activate gene expression in disease-associated deficiencies. In 2009, a study investigated the therapeutic effects of promoter-targeting small dsRNAs delivered in the form of shRNAs by lentiviral vector and showed gene activation of VEGFA that translated into improved blood flow and vascularity in ischemic hindlimbs of mouse models.⁴⁵⁴ MTL-CEBPA was the first saRNA drug candidate to enter clinical trials in 2016 (NCT02716012, NCT04105335), followed by RAG-01 in 2024 (NCT06351904) and the latest early-phase I study with RAG-18 in 2025 that assesses tolerability in pediatric patients with DMD (NCT07282652). As research continues to dissect RNAa mechanisms deeper, saRNAs are on track for therapeutic development at the same time.

Design and mechanisms

Small activating RNAs (saRNAs)

Exogenous saRNAs mediating RNAa are generally 19- to 21-bp dsRNAs with a two-nucleotide overhang, like siRNAs and miRNAs. What the literature has shown so far for gene activation at the promoter region through saRNAs entails that the antisense or guide strand of saRNA (saRNA-AS) is preferentially loaded by Ago2,⁴⁵⁵ because knockdown of other Ago proteins did not severely impair RNAa activity.^452,456 This saRNA-AS-Ago2 complex then generally localizes to the promoter region,^455,457,458 interacts with RNA polymerase II, and recruits other key members to form the RNA-induced transcriptional activation (RITA) complex.⁴⁵⁸ The core members of RITA include RHA, a nuclear DNA helicase II also known as DHX9, and CTR9, a component of the PAF1 complex.^455,458 RNAa mechanism further carries on with changes on epigenetic markers that encourage transcription, such as the monoubiquitination of histone 2B,⁴⁵⁸ and dimethylation and trimethylation of H3K4.⁴⁵³ Other studies have also reported the potential involvement of a noncoding antisense transcript of the target gene that overlaps the promoter in RNAa.^456,459 However, how Ago2, being predominantly a cytoplasmic protein, is transported into the nucleus to mediate RNAa, with or without saRNA-AS loaded, remains mostly unverified. In an RNAi study, RISC was shown to be loaded in the cytoplasm and imported into the nucleus.⁴⁶⁰ Additionally, importin 8 has been identified to aid the transport of miRNAs associated with Ago2 into the nucleus.⁴⁶¹ Notably, endogenous RNAa activity mediated by miRNAs has also been established.^462,463 Well-annotated gene architecture and algorithms for identifying promoter regions and hotspots are crucial when designing targeting locations for saRNA. It could land closely upstream of the transcription start site,⁴⁵³ or even as distal as 2–3 kb up or downstream.^455,457 Avoiding low sequence complexity, such as CpG islands or Alu repeats, pre-established methylation, or single-nucleotide polymorphisms (SNPs), is important and advised as well.^452,457,464

Muscleblind-like (MBNL) proteins are RNA-binding proteins that control alternative splicing. Their functional insufficiency underlies myotonic dystrophy type 1 (DM1) due to incorrect splicing events overall, but treatment using saRNA to activate the MBNL1 gene in patient-derived primary fibroblasts shifted the splicing patterns closer to the ones seen in healthy controls.⁴⁶⁴ Metabolic protective regulator sirtuin 1 (SIRT1) is a histone deacetylase that governs expression of genes related to inflammation. Therapeutic effects of activating SIRT1 expression by saRNA on metabolic syndrome were shown in rats under a high-fat diet, which had a favorable lipid and glucose metabolism upon receiving saRNA treatment, as evidenced by various metabolic parameters.⁴⁶⁵ Myeloid-derived suppressor cells (MDSCs) are a type of undifferentiated and immunosuppressive cells that arise in the tumor microenvironment. On the other hand, transcription factor CCAAT/enhancer-binding protein alpha (C/EBPα) modulates myeloid precursor differentiation.⁴⁶⁶ Besides, downregulation of C/EBPα has been observed in many types of tumors.⁴⁶⁷ In clinical advancements, the Habib group has been developing a saRNA drug MTL-CEBPA,⁴⁵⁵ which has entered clinical trials (NCT02716012, NCT04105335). Initial clinical trial reports established the tolerability of MTL-CEBPA and showed some potential therapeutic effects. Induction of C/EBPα by MTL-CEBPA downregulated proteins associated with immunosuppressive activity in MDSCs, although efficacy might seem inconclusive among a small number of patients in four groups (received sorafenib or not, being viral HCC or not), experiencing different responses at study endpoint, from objective response to stable or progressive HCC (NCT02716012).⁴⁶⁸ RAG-01 is another saRNA drug candidate in clinical development led by the Li group. It activates the antiproliferative p21 gene, which is a downstream effector of the p53 pathway,⁴⁶⁹ and it is currently being evaluated in patients with nonmuscle-invasive bladder cancer (NCT06351904).

Synthesis, optimization, and delivery

The synthesis of small saRNAs is most convenient by chemical synthesis, as for ASOs and siRNAs. Due to relatively limited reports of chemical modifications for optimizing saRNAs compared to siRNAs, a recent study by Desaulniers et al. designed a library of STING (stimulator of interferon genes)-activating saRNAs containing assorted chemical modifications between 2′-O-Me, 2′-F, LNA, unlocked nucleic acid (UNA), abasic carbon spacer of different number of carbons, inverted dT, and mismatch. The best performer in balancing gene activation in cells and improved nuclease resistance was modifications with 2′-F adenosines and a C₃-propyl spacer (three carbons) replacing a nucleobase in the central region of the sense strand in combination with a single LNA modification at position 4 or 8 from the 5′ end on the antisense strand.⁴⁷⁰ More insights for optimization are among the clinical studies we’ve covered; 2′-O-Me was mainly incorporated on the sense strand to reduce immunogenicity for MTL-CEBPA,⁴⁵⁵ whereas 2′-F to the pyrimidines in the antisense strand retained RNAa activity as well as improved metabolic stability of the p21-activating saRNA duplex.⁴⁶⁹ Furthermore, 5′-biotin on the sense strand can eliminate potential off-target effects from sense-strand selection by Ago2.^458,464

Being chemically similar to siRNAs and miRNAs, carrier systems used for small dsRNAs can be easily adapted for delivering saRNAs. MTL-CEBPA saRNA drug is packed in SMARTICLES liposomes for delivery.⁴⁶⁸ RAG-01 (formerly dsP21-322) was initially encapsulated in an LNP consisting of ionizable DLin-KC2-DMA mixed with DSPC, cholesterol, and PEG-DMG in mouse preclinical studies,⁴⁶⁹ while the delivery strategy later switched to a proprietary lipid-conjugate system according to information released by Ractigen Therapeutics. The SIRT1-activating saRNA in a preclinical in vivo study was conjugated with a transferrin receptor aptamer for liver targeting.⁴⁶⁵ Some other potential carrier options used for other RNA therapeutics have also been investigated with saRNAs. Nanoparticles of recombinant apoferritin that have four added lysine residues on their N-terminus, for leveraging their proton sponge effect at acidic pH to destabilize the carrier itself and facilitate drug release, carried SIRT1-activating saRNA. The delivery design additionally incorporated a coating with a hydrogel consisting of chitosan, oxidized chondroitin sulfate, and sodium β-glycerophosphate for sustained release into the mouse knee joint cavity.⁴⁷¹ Also, loading p21-activating saRNA onto a tetrahedral DNA nanostructure with a sticky-end approach so that saRNA cargo is released upon cellular RNase H cleavage when administered to mice bearing cancer cell xenografts.⁴⁷²

Our understanding of the field and discussion

As of August 2026, 27 RNA-based therapeutics have been FDA-approved for treatment or prevention of various diseases (Table 1) and numerous other RNA drug candidates are in clinical trials. These reflect the ongoing development and investment in RNA therapeutics. For clinical effectiveness and tolerability, an RNA drug must be stable enough to reach the intended body or cellular location and perform its intended function without causing undesired effects, whether related to the target or off-target, and whether the drug induces toxicity from immune responses. This involves several interconnected aspects: (1) resistance to endogenous nucleases; (2) the apparent half-life or bioavailability of the RNA molecules in the system; (3) delivery to specific body locations to increase local dosage for on-target effect and reduce off-target effects; (4) maintaining RNA chemistry and structure at the target site consistent with previous testing to ensure effective function; (5) acting on the intended molecular target(s) based on the drug’s mechanism without off-target effects; (6) enabling endosomal escape if the RNA enters target cells via endocytosis; and (7) avoiding immune responses similar to those triggered by foreign infectious RNAs.

Failed clinical trials are mainly due to a lack of clinical effectiveness, followed by immune-related adverse events, delivery issues, and stability issues.^473,474 Despite the growing knowledge and continuing discoveries in every discipline of RNA therapeutics, many mechanisms still need clarification from a basic molecular research perspective. Knowing how dynamic and complex the cellular functional network is, understanding biological and cellular processes in response to RNA therapeutics and carriers is not a straightforward story. As new RNA chemical modifications, delivery strategies, or innovative carrier materials emerge, they prompt independent research to characterize these new interactions between the drug and the cell.

RNA molecules are more vulnerable to the cellular environment than proteins and DNA because of the presence of free OH groups. Multiple robust chemical modification strategies are available to protect the C2′-OH group and RNA molecules overall from chemical instability and nuclease degradation, as discussed in the Optimization subsection for each class of RNA drugs. Many RNA therapeutic strategies and normal RNA functions depend on secondary structures that are thermodynamically formed and essential for activity.^475,476 A critical, often overlooked aspect is that most of the time these secondary structures form during transcription.⁴⁷⁷ The ability of RNA to adopt these structures is modulated by various factors, including sequence composition,⁴⁷⁸ ionic strength,⁴⁷⁹ and pH levels⁴⁸⁰ within the cellular environment. Additionally, RNA undergoes editing and sequence changes more frequently than DNA through natural editing processes, where even a single base modification could alter secondary structures,⁴⁸¹ and potentially impair therapeutic binding efficacy. Consequently, RNA structure can differ between physiological and disease conditions due to factors such as pH, ionic strength, and the chemical microenvironment, which vary, especially in disease contexts like inflammation, metabolic disorders, neurological diseases, or cancer.⁴⁸² Disease-associated alterations, such as mutations or modifications, may also affect RNA structure.^483–485 For example, mutant mRNA transcripts from the DMPK gene (dystrophia myotonica protein kinase), which contain expanded CUG repeats associated with myotonic dystrophy, tend to form hairpin structures within the repeats and accumulate as nuclear foci in neurons.^486,487 Various RNA-based therapies, such as ASO, RNAi-based, and dCas9 systems, have been employed to target the CUG repeat region in treating DM1.⁴⁸⁷ Furthermore, disease-associated SNPs can significantly affect mRNA and regulatory noncoding RNA structures, as demonstrated by the SNP fold algorithm; for instance, mutations in the 5′ UTR of the CPB2 gene, encoding thrombin-activatable fibrinolysis inhibitor, are linked to thrombotic disorders due to reduced RNA stability caused by conformational changes.⁴⁸⁸

RNA molecules are highly conformationally dynamic, making it more accurate to describe them as structural ensembles rather than adopting a specific, fixed structure.⁴⁸⁹ The heterogeneity of their secondary structures is affected by their physiological state (such as during nascent transcription or after post-transcriptional modification) or by the cellular environment (for example, different stages of the cell cycle), which can affect their biological functions (like one structure having a higher affinity for ligand binding than others).⁴⁷⁶ Echoing the concept behind in vitro and in vivo SELEX, researching how external and internal physicochemical factors modulate the structures of both RNA drugs and RNA targets under mimicking conditions could improve confidence in evaluating drug-target interactions and their clinical translation. After all, both the RNA drug and target ultimately coexist in the same in vivo environment where they bind and function.

RNA drugs, whether delivered as payloads in carriers like LNPs, are initially trapped in endosomes after cellular internalization via endocytosis. These drugs must be released before the endosomes mature into lysosomes, where they can be degraded, or they may be recycled outside the cell.⁴⁹⁰ Therefore, the ability of the RNA drug to escape from endosomes is a critical determinant of its intracellular availability and overall efficacy. Furthermore, the mechanism of payload release depends on the class of RNA drug enclosed.⁴⁹¹ In addition to influencing RNA structure and stability, pH changes have been believed to be crucial triggers for endosomal escape and for RNA release from carriers. The amine groups of polycations, such as PEI, which is commonly used in PNPs, increase protonation as the pH shifts from neutral to acidic environments, like in endosomes. Protonated PEI is believed to disrupt the endosomal membrane and facilitate escape, thereby delivering the enclosed or complexed drug into the cytoplasm.^492,493 Similarly, ionizable lipids in LNPs become cationic at acidic pH within endosomes. They can interact electrostatically with negatively charged phospholipids in the membrane, leading to membrane destabilization through mechanisms such as transitioning from a cylindrical to a conical shape.^493,494 Facilitating the release of cholesterol-conjugated siRNAs from endosomes and improving knockdown efficiency are achieved in tumor cell spheroid models in vitro by treating with various membrane-disrupting, cationic amphiphilic small molecules. However, these results are cell-type specific.⁴⁹⁵ This strategy requires in vivo validation but highlights the importance of endosomal escape for RNA therapeutics.

Emerging research on phase separation suggests that changes in RNA-protein complexes may influence RNA-based therapeutic strategies. Most targeting approaches have concentrated on the stability of RNA molecules and their delivery. However, the biggest challenge is achieving specificity and efficiency for the intended purpose, especially in a tissue-specific manner. In this context, antibody-based or aptamer-based therapies are being studied. While peptides and lipid conjugates can increase the affinity of RNA drugs for certain cell surface or plasma proteins, antibodies and aptamers offer high specificity for their targets and can be used either directly on the drug or functionalized on the carrier system to improve docking to recipient cells in a tissue-specific manner. Anti-CD44 and anti-PD-L1 aptamers were leveraged on cationic liposomes to target cancer cell-surface markers for delivering doxorubicin and siRNA disrupting immunosuppressive enzyme indoleamine 2,3-dioxygenase-1 in tumor-bearing mice.⁴⁹⁶ Anti-EGFR and anti-TROP2 (trophoblast cell-surface antigen 2) antibodies are attached to hybrid lipid-polymer nanoparticles via PEG lipid moieties designed to specifically target cancer cells overexpressing EGFR and TROP2 on their surfaces.⁴⁹⁷ These nanoparticles simultaneously deliver siRNAs targeting mutant p53 and TNFα in TNBC cell models. Besides recognizing tumor-associated antigens and synergistically causing detrimental effects in tumor cells with two collaborative siRNAs, this platform also acts as a multi-checkpoint to reduce off-target effects because the likelihood of an off-target cell expressing both surface antigens is low.⁴⁹⁷

Additionally, the specific interaction between glycan-recognizing receptors and their ligands can enable tissue-specific delivery. GalNAc is a well-known and well-characterized example, serving as a ligand for the liver-specific receptor ASGPR, and is frequently used with siRNA therapeutics.³⁷⁴ Mannose could be used to target receptors on macrophages present in the airways and lungs. MRC 1 (mannose receptor C-type 1, or CD206), expressed on macrophages, recognizes glycan structures, and together with its intronic-encoded miR-511-3p, they regulate and protect against allergen-induced lung inflammation.⁴⁹⁸ By anchoring mannose on the surface of cell-derived exosomes containing miR-511-3p mimics (Fig. 5e), the delivery of the RNA drug was explicitly directed to lung macrophages through interaction with the MRC1 receptor, thereby effectively regulating the local allergic inflammatory response.⁴⁹⁹ Modifying the components of the carrier systems offers another approach for adjusting the drug delivery profile. A library screening of six hydroxycholesterol compounds with hydroxyl groups modified at different locations along the cholesterol backbone and at four substitution levels showed that LNPs incorporating 7α-hydroxycholesterol at 25% or 50% enhanced the delivery of mRNA to human primary T cells ex vivo by approximately twofold.⁵⁰⁰ These examples emphasize the importance of available tissue-specific antigens, cell-surface receptors, and properties as markers.

The impact of RNA therapeutics is expected to be significant, given the abundance of unique sequences that are exclusively expressed. Although studying them preclinically in non-primate models could be challenging, many noncoding RNAs show promise because they are specific to humans and non-human primates, indicating their recent evolutionary development and potential functionality in complex biological systems. Understanding the diverse functions of noncoding RNAs can help piece together how biological processes and networks coordinate cellular functions, increasing certainty in selecting the right RNA targets or modalities. Before recognizing the importance of noncoding RNAs, addressing the C-value paradox was difficult, as the number of protein-coding genes could not explain the genomic disparity observed in higher eukaryotes. Different isoform transcripts, arising from alternative splicing, promoter usage, or polyadenylation,^501,502 can have varied molecular functions, including whether they are translated into proteins or act as noncoding RNAs (e.g., MAPT),⁵⁰³ or each noncoding transcript isoform potentially regulates different biological pathways (e.g., LINC01016).⁵⁰⁴ Proteins translated from each isoform can also have distinct functions,⁵⁰⁵ or be induced in different tissues or under various stress conditions like hypoxia and hyperglycemia.^506,507 Isoforms may inherently carry additional specificity, efficacy, and cellular localization signals necessary for drug development. For example, alternative splicing shifts HDAC6 and TP53BP1 from coding to intron-retaining noncoding isoforms, reducing their protein levels in colorectal carcinoma cell lines under forced hypoxia, as shown by novel, de novo sample-specific annotation strategies using RNA sequencing. Data from a cohort of colorectal samples in TCGA also support this observation of isoform switching.⁵⁰⁶

The innate immune response is activated against foreign nucleic acids as part of our defense system. This mechanism involves pattern recognition receptors (PRRs) that recognize foreign RNAs and can trigger the production of inflammatory cytokines, chemokines, and IFN-I.⁵⁰⁸ Several types of PRRs are located at different cellular and/or bodily sites. Exogenous RNA drugs can be recognized in circulation by soluble immune components, within endosomes by Toll-like receptors, and in the cytoplasm by intracellular PRRs.^474,509 Therefore, one approach to mitigate toxicity resulting from immunogenicity and tolerability issues is to design RNA drugs that resemble self-RNA more closely. Natural RNA modifications play a key role in the tight regulation of gene expression. These modifications can modulate RNA secondary and tertiary structures, serving specific functions in modulating gene expression outcomes.^510–512 Incorporating natural RNA modifications into RNA drugs could help decrease immunogenicity and also enhance specificity and effectiveness.¹⁰² The common chemical modifications already used in optimizing RNA drugs, such as m⁵C, Ψ, and ribose 2′-O-Me, are part of the prevalent natural RNA modifications,⁵¹² and have been shown to reduce immunogenicity in some studies.^277,513,514 Another frequently occurring natural RNA modification is m⁶A. In a study, recognized as a potent adjuvant, RIG-I-mediated innate immune activation was attenuated when exogenous circular RNAs were modified with m⁶A, similar to endogenous RNAs. It is proposed that the m⁶A reader YTHDF2 binds m⁶A-modified circular RNAs, preventing recognition and activation of the pattern recognition receptor RIG-I.⁵¹⁵ The immunogenicity of exogenous (synthetic) circRNAs remains inconclusive in the context of their therapeutic application.^437,515,516 Because circRNAs lack a free 5′-triphosphate end, they may be less readily recognized by RIG-I than their linear RNA counterparts. However, this alone may not guarantee immune evasion. Tai and Chen lately discussed the possible contributing factors to immunogenicity as the synthesis and purification methods, and the primary sequence, secondary structure, and RNA modifications of circRNAs. Also, the different approaches selected to assess immunogenicity, such as the panel of immune activation-related genes.⁵¹⁶ Deciphering how nature marks RNA as “self” in a context-dependent fashion remains a central challenge and may offer valuable insights for minimizing the immunogenicity of RNA therapeutics.

Toxicity and/or immunogenicity can also originate from external components of RNA drugs. While making RNA drugs more stable and better at binding to their targets, some toxicity arises when the body reacts against modified nucleotides or analogs (e.g., LNA), conjugations (e.g., PEGylation), or delivery carriers and their parts (such as viral vectors or cationic lipids in LNPs).^474,517,518 COVID-19 vaccines are delivered via PEGylated LNPs that have been shown to trigger the formation of anti-PEG antibodies (Abs). Although these do not interfere with the production of anti-spike protein Abs, it is suggested that anti-PEG Abs may lead to sequestration of PEGylated drugs through complement opsonization, potentially reducing drug availability.⁵¹⁸ Hypersensitivity reactions, as an acute immunogenic response and toxicity, have also been linked to pre-existing anti-PEG Abs, as seen in the clinical trial for PEGylated RNA aptamer, pegnivacogin,²²⁹ and there are speculations that PEGylated lipids, such as ALC-0159 in the COVID-19 vaccine, could trigger anaphylaxis.⁵¹⁷ Besides PEG, some ionizable cationic lipids in LNPs have also raised concerns about inflammation, as they have been shown to induce neutrophil infiltration and cytokine/chemokine production in mice.^519,520 Increased circulating levels of cytokine IL-1β were observed in patients after receiving the lipid-based mRNA cancer vaccine (autogene cevumeran) in a clinical study (NCT03289962).⁵²¹ Additionally, empty LNPs (SM-102) without RNA were sufficient to induce IL-1β release in human peripheral blood mononuclear cells. However, using m¹Ψ/m⁵C-modified RNA complexed with cationic lipids (DOTMA/DOPE) as a lipoplex instead of LNPs reduced the production of various cytokines.⁵²¹

The literature extensively shows the complexity of RNA drug development, including immunogenicity and other aspects we have discussed or overlooked. Immunogenicity seems to result from a combination of RNA sequence selection (whether it contains specific motifs in the sequence or structure, plus chemical modifications) and its interaction with delivery methods. These factors influence each other and the interaction with various biological systems in the body, reflecting a case-by-case nature shaped by multiple functionalities. Understanding how the chemical properties of each compound and its modifications modulate immunogenicity and toxicity is crucial for advancing the design of RNA drugs and carriers.^520,522 Recognizing differences across species in reactogenicity, immunogenicity, and tolerability to RNA therapeutics raises questions about how well preclinical results translate to clinical settings. However, it also suggests the need to investigate more realistic platforms for early testing. Studies recommend using IL-1ra-deficient mice for better prediction of patient responses to innate immune challenges.⁵²¹ Alternatively, engineered tissue multiorgan-on-chip platforms offer more authentic models of human diseases and microenvironments.^523,524 The development of a human tissue chip involving mature heart, liver, bone, and skin tissues, separated by a selectively permeable endothelial barrier and linked by recirculating vascular flow with CD14⁺ monocytes, mimicked the independent functions and interactions of these organs to an extent and modeled the pharmacokinetics and dynamics of doxorubicin.⁵²⁵ Developing a more advanced immune system-on-chip in the future could improve the prediction of toxicities caused by immunogenicity. These tissues could also be analyzed in greater detail.

Conclusions and perspectives

Tremendous work and effort from multiple disciplines have turned the RNA therapeutics concept into reality. This includes basic scientific research identifying targetable RNAs or developing RNA tools, finding strategies to modify vulnerable RNA, especially the free 2′-hydroxyl group, to significantly boost stability and resistance to nucleases in chemistry, exploring nanotechnology for unconventional RNA drug delivery systems, investigating the potential immunogenicity of RNA drugs, and fostering collaboration with the pharmacological industry for manufacturing, conducting clinical trials, assessing pharmacokinetic and pharmacodynamic studies, evaluating adverse events, and ultimately determining if the RNA drug is both safe and effective for clinical use by the FDA.

Engineered biomaterials that mimic the biochemical properties and functions of the extracellular matrix can provide an in vitro microenvironment for cell-cell and cell-matrix communication to investigate fundamental biology. Harnessing the biochemical features of the microenvironment,⁵²⁶ is possible owing to advances in spatial transcriptomics at subcellular-level resolution, which can also aid in designing specific, responsive RNA delivery systems. Such as, understanding the biochemical properties of the tumor microenvironment (TME) enables designing systems that are oxidation-responsive with reducing groups,⁵²⁷ pH-responsive using acid-sensitive and protonation of ionizable groups,⁵²⁸ enzyme-responsive to cathepsin,⁵²⁹ or potentially to other TME enzymes like matrix metalloproteinases, hypoxia-responsive using nitro-aromatic functional groups or azobenzene derivatives,⁵³⁰ and ATP-responsive delivery systems for RNA therapeutics.⁵³¹

Beyond finding methods to escape endosomal entrapment, directing the delivery of RNA drugs toward membrane fusion might be a promising alternative.⁵³² Membrane fusion bypasses endosomal involvement and directly delivers the payload into the cytoplasm, although it remains relatively rare.⁴⁹⁰ Both chemically formulated liposomes containing RNA drugs in their hydrophilic core and naturally occurring EVs secreted by cells possess lipid bilayers that can facilitate membrane fusion with recipient cells. Recent research has focused on engineering approaches to promote membrane fusion in addition to endocytosis.^533,534 Positively charged liposomes interact strongly with the recipient plasma membrane through electrostatic attraction. One study showed liposomes made of cationic lipids DOTAP and MVL5, and optimized with helper lipids DOPC or DOPE, efficiently fused with plasma membrane models and delivered water-soluble dyes. However, when considering the presence of serum and protein coating around these charge-dense liposomes, fusion shifted toward endocytosis.⁵³⁴ Peptides K and E were developed based on the idea of coiled-coil formation between SNARE protein subunits during vesicle docking, which promotes fusion.⁵³⁵ Enhanced versions of peptides K and E have been employed in liposomes encapsulating doxorubicin, resulting in improved drug delivery in vitro. The study showed that coiled-coil interaction between the peptide E4 and peptide K4 facilitated membrane fusion and drug release, under the effect of endocytosis inhibitors.⁵³³ Also, folate (FA) conjugates enhance targeting to tumor cells because the FA receptor is a surface glycoprotein overexpressed on many cancer cells. Confocal imaging revealed that membranes of FA-decorated exosomes containing siRNA localize on the plasma membrane of recipient cells, while the cargo is uniformly distributed in the cytosol.⁵³⁶ Although this study suggested exosome uptake could be through direct membrane fusion, it is not the sole manner and appears to be cell-type specific for both exosome-producing and receiving cells.^537,538 Understanding these differential mechanisms could help engineer a drug delivery system to harness this preferred feature.

Current targeted delivery of RNA therapeutics heavily relies on GalNAc for liver specificity. Characterizing cell- and tissue-specific surface markers can aid in designing suitable docking moieties for various RNA delivery modalities. Combining mass spectrometry-based proteomics, RNA sequencing technologies (particularly single-cell and spatial transcriptomics), and data science could lead to a platform for cataloging cell, tissue, and organ-specific surface marker profiles, especially those associated with disease. These markers may include membrane proteins, cell adhesion molecules, phospholipid head groups, and glycosylation or other modifications on cell-surface proteins or lipids. The antibody-drug conjugates (ADC) database (ADCdb) documents pharmaceutical information and biological activities for each FDA-approved ADC, as well as those in clinical trials and in preclinical status, including data from in vivo and cell line testing, many of which are yet to undergo biological testing. There were 6572 ADCs as of 2024.⁵³⁹ A thorough understanding of the biological mechanisms activated after docking moieties bind to surface markers, such as receptor-mediated endocytosis, is crucial for developing more effective and targeted drug delivery systems. Another approach to achieve cell or tissue specificity could involve using DNA-based vectors that encode RNA modalities controlled by tissue-specific promoters.⁵⁴⁰ Subretinal injection of a lentiviral vector encoding both intron-embedded miRNAs to suppress VEGF expression and produce the neurotrophic pigment epithelium-derived factor (PEDF), driven by the retinal pigment epithelium-specific VMD2 (vitelliform macular dystrophy 2) promoter, exhibited antiangiogenic activity to help alleviate age-related macular degeneration.⁵⁴¹ Although this approach may reduce off-target effects from RNA drugs binding to unintended molecular targets in other tissues or organs, the key challenge remains directing most of the dose to the target tissue to achieve effective treatment.

Rigorous research is necessary to annotate each isoform that arises under diseased microenvironments. RNA sequencing technologies combined with computational analysis can help build databases for new insights into isoforms. Inspiration could also come from the strategy used to study isoform-specific function in Drosophila cells.⁵⁴² Aptamers could be developed to target particular disease-related protein isoforms, and ASOs and RNAi-based drugs could be designed to be isoform-specific. Additionally, specific mRNA or lncRNA isoforms can be introduced to address disease-associated loss of function, increasing specificity and effectiveness. The area of RNA-based therapeutics has grown significantly and holds the potential to revolutionize the treatment of various diseases.

With the rise of artificial intelligence (AI) techniques, it is crucial to implement AI to address previously unthinkable scenarios that could threaten the success of RNA-based drugs. AI can simulate numerous scenarios that help humans analyze and develop effective RNA-based therapies. Recognizing that population-specific genetic variation exists and can affect disease development, progression, drug response, or prognosis,^543,544 we can apply machine learning to identify disease-relevant SNPs or other types of polymorphisms across populations to ensure the drug candidate is inclusive or to customize RNA drug sequences according to those polymorphisms, which could advance the concept of personalized medicine. Patient-customized treatment has already been demonstrated with a case report that describes the use of a patient-specific splicing modulator ASO. The patient was diagnosed with a fatal neurodegenerative disorder called Batten’s disease, caused by an insertion of a retrotransposon into intron 6 that disrupts the normal splicing between exon 6 and exon 7 of lysosomal MFSD8 RNA (major facilitator superfamily domain-containing 8).¹¹⁸ The authors modeled nusinersen, the FDA-approved ASO drug that modulates the splicing pattern of SMN2 RNA for treating spinal muscular atrophy, to design and develop Milasen, which improved the patient’s seizure symptoms.¹¹⁸ Recognizing the potential and need for personalized RNA medicine, Means et al. have developed a practical pipeline for creating patient-derived induced pluripotent stem (iPS) cells and differentiated organoid platforms to generate and evaluate patient-specific ASO therapeutics based on the patient’s genetic vulnerabilities.⁵⁴⁵

The combination of different classes of RNA therapeutics with other modes of therapy to achieve synergy has become popular and should be the expected approach to treating multifactorial and evolving diseases such as cancers, neurodegenerative disorders, and more. In seeking a potential multidrug regimen to alleviate symptoms and slow the progression of Parkinson’s disease (PD), a functional liposome carrier was created and tested in vitro to co-deliver levodopa and ibuprofen to neuronal cells, aided by an α-synuclein-recognizing aptamer. Levodopa, as a precursor of dopamine, aims to compensate for the deteriorating dopaminergic neurons, while the anti-inflammatory ibuprofen is intended for neuroprotection. The liposome-conjugated aptamer can direct this combined treatment to cells expressing α-synuclein, whose aggregated form is the pathological hallmark of PD.⁵⁴⁶ Dual-functional Janus nanoparticles made of silver and mesoporous silica were designed to load doxorubicin within the mesoporous structure and were coated with positively charged chitosan to help anchor MDR1 (multidrug resistance)-targeting ASO, as well as tumor marker mucin-1-recognizing aptamer on the surface via electrostatic interactions. Besides delivering chemotherapy drugs, drug-sensitizing ASO, and cancer-targeting aptamer, this multifaceted nanoparticle can also provide photothermal therapy, thanks to its silver component, which can convert near-infrared light into heat. The cytotoxicity of drug-resistant MCF7 breast cancer cells (MCF7/ADR) was effectively demonstrated with this combined treatment.⁵⁴⁷

We have come a long way from traditionally viewing RNA as just a genetic information transmitter to recognizing it as a therapeutic target and tool, which has opened up therapeutic avenues that were previously not possible.

Author contributions

Conceptualization: S.S.G. and B.Y.; writing, original draft preparation: B.Y., M.J.S., K.P., M.V., V.R. and E.I.R.; writing, review and editing: B.Y., M.J.S., K.P., M.V., and V.R., E.I.R.; M.Y., S.C., S.D. and S.S.G.; visualization: B.Y.; supervision, S.S.G.; project administration: S.S.G.; funding acquisition: S.S.G. All authors have read and agreed to the published version of the manuscript.

Competing interests

The authors declare no competing interests.