Designing nucleic acid-based therapeutics for cancer treatment: Updates on the state of the art

Abstract

As in art, the ability of nucleic acids to be designed and synthesized as a novel treatment modality is limited only by the imagination. Nucleic acids of virtually all sizes and forms can be synthesized on demand, from short antisense oligonucleotides to large mRNAs and to entire chromosomes. Given the genetic basis of cancer, nucleic acid-based therapy is a particularly promising avenue for anticancer therapeutic development. This has led to a profusion of studies exploring strategies to utilize nucleic acid-based drugs to treat cancer, with some approaches demonstrating great potential for clinical translation. In this review, we summarize the various nucleic acid-based strategies being developed for cancer therapy. We also provide a comprehensive overview of current efforts to enhance the potency and safety of nucleic acid-based drugs, exploring advances in nucleotide composition, design, and delivery strategies.

Graphical abstract

This review highlights how nucleic acid therapeutics are reshaping cancer treatment by advancing design principles, chemical modifications, and delivery strategies. It underscores both opportunities and obstacles for clinical translation, offering a roadmap for turning molecular innovation into effective, clinically viable cancer therapies.

Introduction

The origins of nucleic acid-based therapeutics can be traced back to the pioneering work by Friedmann and Roblin over 50 years ago.¹ They first conceptualized the use of functional gene copies to treat genetic disorders, sparking a revolution in our understanding of inherited conditions and laying the groundwork for novel therapeutic approaches. Since then, significant advances in nucleic acid-based therapeutics have propelled the field forward, offering promising solutions in the realm of disease treatment. One notable advance is the emergence of gene editing technologies, particularly CRISPR-Cas9.^2,3 This groundbreaking tool enables precise modifications of genetic material, providing unprecedented opportunities for correcting disease-causing mutations or disrupting disease-associated pathways, including those involved in certain types of cancer.³ RNA-based therapeutics has undergone similar major advancements. The development of messenger RNA (mRNA) vaccines, exemplified by those used in the COVID-19 pandemic, has demonstrated the potential of using RNA molecules to trigger the production of therapeutic proteins or antigens that can stimulate the immune system to target diseased cells.^4,5 RNA interference (RNAi) has also matured into a potent strategy for silencing disease-causing genes, offering promising prospects for precision medicine.⁶ Another noteworthy development involves the refinement of nucleic acid delivery systems. By utilizing advanced nanoparticle formulations, extracellular vesicles (EVs), and viral vectors, the barriers that hinder effective delivery can be overcome, allowing for improved uptake and action of nucleic acid-based therapeutics.^7,8,9,10

The unique properties and capabilities of nucleic acid-based therapeutics provide distinct advantages over conventional approaches. First, nucleic acid molecules can be engineered with high specificity, particularly evident in gene editing and RNAi techniques, allowing for selective modulation of gene expression while reducing off-target effects and potential side effects that are commonly associated with non-specific treatments.^1,6 Second, nucleic acid molecules, whether in the form of DNA, RNA, or oligonucleotides, possess broad versatility in their applications, from restoring gene function to silencing diseased genes or modulating immune responses.^1,6,11,12,13 Such versatility facilitates the development of customized treatment strategies tailored to individual patients, aligning with the principles of personalized medicine, where therapies are designed based on each patient’s unique characteristics. Moreover, nucleic acid therapeutics can be produced relatively rapidly due to advancements in manufacturing technologies, supporting rapid translation from research to clinical applications in urgent settings such as viral outbreaks.^4,5 Furthermore, recent advancements in chemical modification and carrier encapsulation techniques have improved their stability, durability, and safety.^11,14 This is a crucial aspect to consider when contemplating long-term treatment strategies or when targeting sensitive cell types, such as stem cells or germ cells.¹⁴

These attributes have spurred strong interest in cancer therapy, where precise and versatile targeting is critical. Since the initial application of gene transfer in cancer treatment, extensive research and advancements have been made in utilizing nucleic acids such as small interfering RNAs (siRNAs), antisense oligonucleotides (ASOs), aptamers, and CRISPR-Cas9 gene editing systems to silence oncogenic genes, disrupt key protein-protein interactions, or correct genetic mutations.^3,13,15,16 They are also powering new forms of cancer immunotherapy, such as nucleic acid vaccines and immunomodulators.¹² The first clinical trial of nucleic acid-based therapeutics for cancer treatment marked a significant milestone in the field. Conducted in 1993, the trial aimed to evaluate the safety and efficacy of a gene encoding a foreign major histocompatibility complex protein (HLA-B7) in five HLA-B7⁻ patients with stage IV melanoma by injection of DNA-liposome complexes.¹⁷ The groundbreaking trial has demonstrated tumor regression in one patient, providing valuable insights into the feasibility and potential of nucleic acid-based therapeutics as a viable treatment option for cancer. However, the clinical application of nucleic acid therapeutics has encountered distinctive challenges due to their unique physicochemical properties, pharmacokinetics, and toxicology profiles. Over the past few decades, numerous chemical modifications of nucleic acids have been developed to enhance their resistance to enzymatic degradation and improve their stability and specificity.^18,19 Moreover, innovative chemical conjugates and delivery systems have been developed to enhance the stability, cellular uptake, and intracellular delivery of nucleic acid therapeutics, ensuring their effective delivery to tumor sites.^7,11

This review aims to provide an up-to-date overview of the state of the art in designing nucleic acid-based therapeutics for cancer treatment. We explore recent advancements and updates in the rapidly evolving field of nucleic acid-based anticancer strategies. Additionally, we discuss advances in nucleotide composition, design, and delivery strategies of nucleic acid-based approaches, as well as the challenges and future directions in harnessing the full potential of these therapeutics. By gaining insights into the latest developments, we can foster the development of innovative strategies and contribute to the advancement of cancer therapeutics toward improved patient outcomes.

Nucleic acid-based strategies to suppress tumor progression

Numerous classes of nucleic acids have been explored for anticancer therapy, utilizing a multitude of strategies to induce antitumor effects. In this section, we cover four major approaches that employ nucleic acid-based therapeutics in cancer treatment.

Gene transfer

Gene transfer introduces exogenous DNA or RNA into target cells, typically using plasmids, mRNA, or viral vectors. The transgene can be endogenous, derived from another species, or synthetically constructed to incorporate specific modifications. Therapeutic purposes include overexpressing tumor-associated antigens (TAAs) for immune activation, delivering cytotoxic or regulatory transgenes, restoring mutated tumor suppressors, or inducing apoptosis in tumor cells.²⁰ The therapeutic effect depends not only on the target cell but also on the duration and level of expression. For example, cancer vaccines typically require short-term antigen production that reaches antigen-presenting cells to activate antigen-specific adaptive immune responses. Conversely, tumor suppressor genes generally require prolonged, widespread expression to suppress tumor growth. Thus, gene transfer requires fine control of gene expression in both spatial and temporal aspects.

There are a number of strategies that convey this control, including the choice of nucleic acid type (plasmid DNA/mRNA) and delivery method, and the addition of regulatory elements. mRNA-based gene transfer lasts for only a few days before the mRNA molecule is eventually degraded. Plasmid-based delivery, however, lasts slightly longer, although it is unable to induce sustained gene expression and will become diluted as cells divide. Integrating viral vectors can insert their genetic material into the host genome permanently, ensuring that each daughter cell carries the gene of interest and facilitating sustained gene expression.²⁰ Furthermore, regulatory elements such as tissue-specific promoters and enhancers or inducible gene expression systems provide greater spatial and temporal control. Depending on the gene being transferred and the protein expression patterns required to achieve antitumor effects, these factors can be modulated to fine-tune expression in the desired cell type. The following subsections elaborate on different anticancer approaches that utilize gene transfer.

Vaccination

Vaccination introduces antigens to elicit an antigen-specific immune response. In cancer, this is typically achieved via TAAs or tumor-specific neoantigens delivered as purified proteins, tumor lysates, or nucleic acid-based molecules such as mRNA and plasmids that encode the antigen for in situ expression.²¹

Early cancer vaccines involved the delivery of common TAAs identified to be aberrantly expressed or upregulated in certain cancers.²² However, this strategy is not ideal, as these targets may also be expressed at lower levels in healthy tissues. Since these abnormal expressed proteins are still self-proteins, antigen-specific T cells can undergo central and peripheral tolerance, preventing their effective application in cancer therapy. Thus, these early strategies based on TAA presentation suffered from poor specificity and limited onset of immune activation.²³

More recently, personalized neoantigen-based cancer vaccines have been investigated as a more targeted strategy. This involves identifying tumor-specific mutations in individual tumors that are expressed exclusively in tumor cells, thereby enabling an immune response against them.^22,24,25 Preclinical studies have demonstrated the safety and immunogenicity of this approach, and early-phase clinical trials have shown encouraging results.²⁴ One of the most relevant trials is the IVAC MUTANOME trial (NCT02035956), which involved an RNA vaccine encoding multiple TAAs and neoantigens based on the unique mutation signatures of patient tumors. Tested in stage III and IV melanoma patients, this mRNA vaccine induced antitumor CD4⁺ and CD8⁺ T cell responses and synergized with immune checkpoint inhibitors.²⁶

Despite this progress, cancer vaccines remain at an early stage. To date, only two prophylactic cancer vaccines, namely the hepatitis B virus vaccine and the human papillomavirus (HPV) vaccine, which prevent virus-associated cancers by protecting against oncogenic viral infection, and one therapeutic cancer vaccine (sipuleucel-T for prostate cancer), have been approved by the US Food and Drug Administration (FDA). Most others, including neoantigen-based cancer vaccines, are currently limited by challenges associated with identifying suitable and potent neoantigens.^27,28,29 Moreover, development of personalized therapies is time-consuming, requiring extensive screening processes to identify suitable targetable mutations. Recent studies have worked on improving neoantigen discovery and the identification of immunogenic neoepitopes recognized by effector cells.²⁴

Expression of genes for immunomodulation

Another strategy to induce antitumor immune responses against cancer involves the use of pro-inflammatory immunoregulatory elements, such as cytokines and chemokines. Unlike recombinant protein therapy, nucleic acid-based delivery via mRNA, plasmids, or viral vectors enables sustained, localized expression, reducing the need for repeated dosing and minimizing systemic toxicity. Given that immunomodulatory proteins secreted from one cell can have broad immunostimulatory effects within the local tumor microenvironment, it is unnecessary for nucleic acids encoding these proteins to be delivered to all tumor cells. Representative examples include interleukin-12 (IL-12), IL-27, IL-15, and IL-2.^30,31,32,33 Given the complex nature of tumors, many of these studies also use combination therapies, combining multiple immunomodulatory elements or immune checkpoint inhibitors simultaneously, or multiplexing their use with other anticancer therapeutics, such as a vaccine/antigen or chemotherapeutics.^25,31

Expression of pro-apoptotic genes, tumor suppressors, and antitumor proteins

Similar to the delivery of nucleic acids encoding immunomodulatory proteins, nucleic acids can also be used to deliver genes encoding other proteins that confer anticancer effects. Over the years, genes encoding a number of different classes of molecules capable of exerting anticancer effects have been developed, including but not limited to antitumor antibodies or antibody fragments, which are secreted, bind to tumor antigens, and activate immune cell-mediated killing,³⁴ pro-apoptotic genes that induce cancer cell death,^35,36 and tumor suppressor genes that attempt to restore mutated tumor suppressor genes.³⁷ Such strategies expand the toolkit of gene therapy beyond immunomodulation, allowing direct targeting of cancer cells through well-defined molecular mechanisms. Success depends on sustained expression, precise targeting, and compatibility with existing tumor biology, factors that continue to be refined in preclinical and translational studies.

Gene expression regulation

The onset of genetic mutations and dysregulation of gene expression are the underlying causes leading to tumorigenesis. Indeed, the development of cancer often involves the accumulation of multiple mutations in genes associated with essential cellular processes, such as cell division, apoptosis regulation, and cellular migration or metastasis. This provides an opportunity for cancer therapy, as the expression of these dysregulated or mutated genes can be directly controlled at the genetic level. However, it should be taken into consideration that the selection of suitable targets for cancer therapy is a vital factor that can affect the success of gene expression regulation. For efficient tumor suppression, it is important to target oncogenes that are vital for maintaining tumorigenicity. Often, tumor cells develop multiple mutations that confer oncogenicity through diverse pathways. In these cases, targeting a single pathway may have little if any effect on suppressing tumor progression. Thus, most current approaches in this area look at targeting commonly upregulated key oncogenic drivers, such as KRAS, or restoring the expression of crucial tumor suppressor genes, such as p53, which are more likely to be effective across a broad range of patients.

Strategies for cancer treatment relying on gene expression regulation use a variety of tools to correct aberrant gene expression, silence mutated proto-oncogenes, or recover the expression of genes essential for preventing tumorigenesis. This section outlines the utility of these tools in cancer therapy, focusing on strategies that function at the RNA level by altering mRNA abundance or translation (Figure 1).

Figure 1.

Approaches for nucleic acid-based gene expression regulation in cancer therapy

(1) Genome editing via the CRISPR-Cas9 system facilitates the permanent alteration in target genes, allowing knock out of oncogenes or knock in of antitumor genes (a). Alternatively, prime editing using a longer pegRNA in conjunction with a dCas9 fused to a reverse transcriptase can be applied to rectify mutations in genes without the need for double-stranded breaks (b). (2) Transcriptional regulation relies on the use of epigenetic modifications and transcriptional regulators to control the rate of production of mRNA. Epigenome editing using dCas9 fused to epigenetic regulatory enzymes can regulate transcription of target genes, either restoring the expression of hypermethylated tumor suppressor genes or repressing the expression of oncogenic genes via hypermethylation (a). lncRNAs and saRNAs can interact with RNA polymerases and transcription factors to control the transcription of target genes. Alternatively, CRISPR-mediated transcriptional regulation allows transcriptional activation or suppression of specific targeted genes by making use of a dCas9 fused to a transcriptional activator or repressor (b). (3) RNA modulation, such as RNA base editing, allow non-permanent alteration to transcripts by modifying individual nucleotides on target mRNA species. The use of the RNA-binding dCas13 fused to ADAR allows deamination of specific mRNA loci, changing adenosine to inosine, which is subsequently interpreted as guanine (a). Exon skipping utilizes ASOs, which bind to specific regions of the pre-mRNA and induces its exclusion from the mature miRNA transcript following processing by the spliceosome (b). In either case, RNA modulation can be used to correct mutated genes or to prevent the oncogenic effects mediated by oncogene encoding mRNAs. (4) Gene silencing involves the use of siRNAs/miRNA mimics (a), ASOs/gapmer/antagomirs (b), and DNAzymes/ribozymes (c) to induce degradation of target RNAs. Double-stranded siRNA and miRNA mimics are processed by Dicer before being loaded onto RISC, wherein they can bind to and inhibit target mRNA, inducing their degradation. Single-stranded ASOs and gapmers act in a similar manner by directly binding to target RNAs and inducing their degradation, while antagomirs can inhibit the action of target miRNAs. DNAzymes are catalytic DNA oligonucleotides that act by cleaving target mRNA strands. Directly targeting oncogenic mRNA provides an effective strategy to achieve tumor suppression. Alternatively, by targeting the endogenous RNA-based regulatory elements in the cell, such as ceRNAs, lncRNAs, and cellular miRNAs, it is possible to modulate the translation of oncogenic mRNAs without having to directly target oncogenic mRNAs. ADAR, adenosine deaminase acting on RNA; ASO, antisense oligonucleotide; ceRNA, competing endogenous RNA; CRIPSR, clustered regularly interspaced short palindromic repeats; dCas13, endonuclease-deficient Cas13; lncRNA, long non-coding RNA; miRNA, microRNA; pegRNA, prime editing guide RNA; RISC, RNA-induced silencing complex; saRNA, small activating RNA; sgRNA, single-guide RNA; siRNA, short interfering RNA; TSS, transcription start site.

Gene silencing

Gene silencing is a core strategy in cancer therapy that targets oncogenic RNAs, including mRNAs, oncogenic microRNAs (oncomiRs), or long non-coding RNAs (lncRNAs), using nucleic acid-based tools such as ASOs, siRNAs, or miRNA mimics.^38,39 This prevents the target oncogenic RNA from carrying out its function or being translated and can induce its degradation.³⁹ Some of these targets are simply mutated proto-oncogenes, where the mutated form induces oncogenic effects. Alternatively, certain cancers exploit dysregulated expression of certain RNA species, which are not harmful at normal physiological levels but become oncogenic upon their upregulation. The most obvious targets for cancer therapy are mRNAs encoding oncogenic proteins that confer enhanced proliferative signaling, resistance to cell death, immune evasion, replicative immortality, sustained angiogenesis, invasion, and metastasis. Frequently targeted examples include mutated forms of constitutively active KRAS, upregulated PD-L1, and constitutively active or overexpressed growth factor receptors such as epidermal growth factor receptor (EGFR) and human epidermal growth factor receptor 2 (HER2).^34,40,41,42

miRNAs, particularly oncomiRs, are another class of important targets.⁴³ Key examples include miR-125b, which promotes proliferation, invasion, and migration, and the miR-17-92 cluster, which regulates the cell cycle, proliferation, and apoptosis.^43,44,45 lncRNAs also contribute to tumorigenesis through diverse mechanisms. Oncogenic lncRNAs include MALAT1, HOTAIR, and NEAT1.^46,47,48 They exert effects by acting as miRNA sponges, influencing transcription of genes via epigenetic regulation, or serving as scaffolds for multiple regulatory factors. For example, HOTAIR acts as a miRNA sponge for tumor-suppressive miR-149-5p. Studies have shown that silencing of HOTAIR can restore miR-149-5p expression and induce tumor suppression.⁴⁷

Silencing or downregulation of these targets can present a feasible approach for suppressing cancer progression. However, as mentioned previously, selecting a potential target is a key factor that can affect the efficacy of gene silencing on tumor suppression and often varies between individuals and different cancer types. Certain commonly mutated proto-oncogenes serve as major oncogenic drivers in certain cancers. For example, KRAS mutations are found in ∼85% of pancreatic ductal adenocarcinomas, ∼45% of colorectal adenocarcinomas, and ∼30% of lung adenocarcinomas.⁴⁹ Identifying specific mutations is often necessary, as the wild-type gene may be essential for the normal function of healthy cells, and simply silencing the gene in all cells can lead to adverse effects. In the case of KRAS, genetic sequencing has revealed that KRAS G12D and KRAS G12V are the most common mutations found in pancreatic cancer, which in turn has led to a race to develop mutation-specific RNAi approaches that can be safely administered in vivo without inducing off-target knockdown in healthy cells.⁵⁰

The go-to approach for gene silencing involves siRNAs or ASOs designed to complement the target gene for effective binding and engagement. While straightforward, these approaches are complemented by alternate strategies inspired by endogenous RNA regulatory elements, such as competing endogenous RNAs (ceRNAs) and miRNA mimics. For example, introducing specific miRNA mimics into cells can broadly induce changes in mRNA levels of multiple downstream genes.⁵¹ These miRNA-based anticancer therapies function by suppressing gene expression in numerous oncogenic factors with varying efficacy, and their development has progressed significantly in recent years.^52,53,54 Similarly, miRNA sponges, including circular RNAs (circRNAs), can modulate miRNA content in cells, which in turn can lead to changes in mRNA expression.⁵⁵ For instance, circRNA sponges targeting oncogenic miR-21 suppress lung cancer.⁵⁶ circRNAs have also been implicated in liver, gastric, bladder, colorectal, breast, cervical, and brain cancers, indicating their potential broader clinical application in cancer therapy.^57,58,59,60

Lastly, a class of nucleic acid-based catalytic molecules, termed DNAzymes and ribozymes, has been reported to be capable of inducing gene silencing. DNAzymes bind to target mRNA and induce cleavage at the binding site, reducing the copy number of transcripts.⁶¹ More recently, and of greater relevance, DNAzymes against multiple oncogenic mRNAs, including MMP-9, VEGFR-1/2, and c-jun, have shown significant therapeutic effects in in vivo models. DNAzymes against two targets have moved into phase 1 clinical trials. An in-depth review of the use of DNAzymes in cancer therapy has been reviewed elsewhere by Thomas et al.⁶² Similar studies have been conducted for ribozymes. Angiozyme is a stabilized ribozyme cleaving VEGFR-1 mRNA, leading to delayed tumor growth and decreased vascularization.⁶³ Ribozymes for other targets, including EGFR mRNA, have been studied, but they suffer from poor delivery efficiency and limited cytoplasmic accumulation.⁶⁴ Despite their rapid evolution in recent years, this class of molecules is still in its infancy and requires significant advances in delivery efficiency, silencing efficacy, and ease of design, all of which are currently inferior to what is achievable using antisense strategies.

RNA modulation

An alternate approach for controlling gene expression involves direct manipulation of RNA molecules themselves. While there are a multitude of different RNA modulation mediated by various molecules and enzymes, this section focuses on two principal nucleic acid-based approaches: RNA base editing and exon skipping. Both approaches act on pre-mRNA in the nucleus to modulate gene expression post-transcriptionally, offering reversible and programmable means of therapeutic intervention.

Exon skipping, mediated by ASOs, involves the use of antisense strands that bind to pre-mRNA, inducing the removal of targeted exons from the final mRNA transcript. This technique can correct a number of different genetic mutations either by skipping out-of-frame exons or mutated in-frame/out-of-frame exons to restore gene function, or by skipping out-of-frame exons to disrupt gene function. One example of this involved the use of an ASO designed to induce skipping of the out-of-frame exon of the E-26 transformation-specific-related gene (ERG). The ASO suppressed ERG protein levels and inhibited its function in prostate cancer cells, resulting in decreased cell proliferation.⁶⁵ Interestingly, some cancers involve mutations that activate pseudoexons.⁶⁶ Moreover, many genes associated with cancer progression have been shown to contain pseudoexons, including tumor-suppressive genes ATM and BRCA1. Skipping mutated pseudoexons to restore normal gene function holds promise as a strategy in this respect for cancer therapy.

RNA base editing is a potent tool for manipulating RNA in an effort to correct mutations and regulate gene expression. Given its transient, non-permanent effect, it is safer than gene editing approaches that result in permanent genomic alterations. RNA editing employs enzymes such as adenosine deaminase acting on RNA (ADAR), which convert adenosine to inosine via deamination, and the resulting inosine is interpreted as a guanine residue by the translational machinery. However, ADARs are generally promiscuous and lack specificity. Guide RNA (gRNA)-directed RNA editing in conjunction with an RNA binder, such as catalytically inactive Cas13 (dCas13), can mitigate potential off-target editing and allow specific RNA editing of target mRNAs. Termed RNA editing for programmable A-to-I replacement (REPAIR), this approach has been applied in preclinical studies to correct missense and nonsense mutations in genetic disorders such as X-linked nephrogenic diabetes insipidus.⁶⁷ More recently, Abudayyeh et al. have demonstrated the ability of LwaCas13a to target and degrade oncogenic targets, including KRAS and CXCR4, with efficiency comparable to RNAi and greater specificity.⁶⁸ These studies lay the basis for potential applications of REPAIR in cancer therapy. An alternate strategy, termed recruiting endogenous ADAR to specific transcripts for oligonucleotide-mediated RNA editing (RESTORE), utilizes ASOs to recruit ADARs for editing specific transcripts. RESTORE has been applied to treat genetic disorders such as alpha-1 antitrypsin deficiency by targeting the SERPINA1 gene and has shown promising results. Here, too, the authors demonstrated recovery of a clinically relevant PiZZ mutation with minimal off-target effects.^69,70 However, the practical application of RNA-editing strategies is currently constrained by challenges associated with achieving efficient gene delivery.⁷¹

Transcriptional regulation

Transcription regulation modulates gene expression by influencing the production of specific transcripts in the nucleus. While traditional protein-based transcriptional regulatory elements, including promoters, enhancers, silencers, and insulators, play key roles, this section focuses on nucleic acid-based regulatory mechanisms.

Certain RNA species, such as small activating RNAs (saRNAs) and some types of lncRNAs, can modulate the transcription of specific genes via interaction with protein co-factors and RNA polymerases. saRNAs have a structure similar to that of siRNAs and perform RNA activation to increase the production of transcripts. saRNA-based therapeutics have shown promising effects in preclinical studies for cancer therapy, using models of liver cancer,^72,73 bladder cancer,⁷⁴ and gastric cancer.⁷⁵ lncRNAs, however, can have varying effects and may enhance the transcription of genes by modulating chromatin structure or inhibit transcription by associating with polymerases.⁷⁶ However, their therapeutic use in cancer therapy is limited due to their larger size and complex nature. lncRNAs are more often targeted by RNAi to exert regulatory effects in cancer, as described previously.

CRISPR-mediated transcriptional regulation is another strategy that utilizes a fusion of dCas9 with transcriptional activators or repressors to localize these regulators to specified genomic loci. This enables targeted activation or suppression of transcription of target mRNA, facilitating the modulation of transcriptional levels of desired genes. In the context of cancer therapy, many cancers display hypermethylated regions of genomic loci that prevent the transcription of tumor suppressor genes. CRISPR-dCas9-based transcriptional activators have been explored to restore the expression of such epigenetically silenced tumor suppressor genes. A common construct used involves the fusion of dCas9 with the transcriptional activator protein domain VP64, which leads to rapid chromatin remodeling and gene activation when targeted to specified genomic regions. Braun et al. used this system to identify Chek2 as a suitable tumor suppressor gene target for the suppression of B cell acute lymphoblastic leukemia (B-ALL). The authors demonstrated that transcriptional activation of Chek2 was capable of slowing down tumor progression and increasing the sensitivity of B-ALL to chemotherapy.⁷⁷ Similar strategies have been utilized by other groups to rescue the expression of tumor suppressor genes that are aberrantly silenced in cancer, with some studies making use of alternate tools such as the telomerase-activating gene expression system in addition to CRISPR-dCas9 systems.⁷⁸

Epigenome editing uses dCas9 fused with enzymes such as histone acetyltransferases, DNA methyltransferases, or DNA demethylases, directed by single-guide RNAs to specific loci for precise chromatin remodeling.^79,80,81 Studies have reported the use of these strategies to recover the expression of key tumor suppressor genes, including BRCA1 and p16, resulting in decreased oncogenicity of tumor cells.^81,82 More recently, Liu et al. reported the development of CRISPR-DNMT1-interacting RNA, which was used to target epigenetic modifications to the promoter-exon 1-intron 1 region and induce a wave of local chromatin remodeling, resulting in locus-specific gene activation of key tumor suppressor genes.⁸²

Genome editing

Gene editing via the CRISPR-Cas9 system is another approach that harnesses its capability to permanently knock in or knock out a gene, thereby modulating the expression of cancer-associated genes. Most anticancer therapies using this approach involve the knock out of oncogenic genes. Over the years, multiple studies have reported the utility of this strategy in silencing key oncogenes to achieve cancer suppression. Rosenblum et al. demonstrated the use of a lipid nanoparticle (LNP)-delivered CRISPR construct to knock out PLK1, a gene that encodes a protein essential for mitosis. They achieved up to 98% of PLK1 editing, resulting in cell-cycle arrest and apoptosis of target cells.⁸³ Other targets for cancer therapy using CRISPR include Cdk5 (leading to decreased PD-L1 expression), NESTIN, and mutated forms of constitutively active EGFR and FAK.⁸⁴

Prime editing offers precise base substitutions and small insertions or deletions without double-stranded breaks, using a Cas9 nickase, reverse transcriptase complex guided by prime editing gRNA.⁸⁵ The prime editing approach for translational use is still in its infancy, and even the latest generations suffer from relatively low editing efficiencies. However, a few studies have investigated its potential application in cancer therapy. Abuhamad et al. used prime editing to restore a mutated form of the TP53 tumor suppressor gene in breast cancer. The authors demonstrated the ability to correct the TP53 missense C > T mutation (L194F) in a breast cancer cell line as detected using amplicon target sequencing.⁸⁶ However, they reported low editing efficiency and suggested further improvements before therapeutic use.

Despite the advent of these CRISPR-directed strategies for developing anticancer therapeutic approaches, their translatability into clinical use remains challenging. While CRISPR-Cas systems allow targeted delivery of otherwise non-specific transcriptional and gene editing tools to specific genomic loci, their large size, low efficiency, and potential for off-target effects remain major obstacles. Significant strides have been made in recent years to resolve these issues, with the development of smaller Cas9 systems, demonstrating greater editing efficiency and higher specificity. However, particularly in the context of cancer therapy, effective treatment usually requires editing nearly 100% of tumor cells to prevent relapse. For example, transcriptional activation of tumor suppressor genes or the knock out of oncogenic factors requires comprehensive editing for effective tumor suppression. Thus, even with recent advances in gene editing technology, its feasible application in cancer therapy remains uncertain at present.

Ethical considerations remain another significant hurdle. Most current CRISPR-based strategies focus on somatic cell editing, where changes are confined to the treated individual. However, germline modification introduces heritable changes with far-reaching implications. In theory, correcting inherited cancer mutations such as BRCA1/2 or TP53 in embryos could prevent hereditary cancer syndromes entirely. However, such interventions would permanently alter the human germline, raising concerns about consent for future generations, long-term safety, and potential inequities in access. International consensus currently discourages germline editing outside of tightly regulated basic research, given these unresolved issues. By contrast, somatic applications, such as engineering T cells or directly targeting oncogenic pathways, are considered ethically permissible and clinically relevant since their effects are not inheritable. Thus, while germline editing is often discussed in parallel with the potential of CRISPR-directed strategies, its application in cancer therapy remains ethically and legally constrained, with ongoing debate about whether its risks could ever be justified.

Direct protein regulation

Regulating protein function using nucleic acid-based therapeutics represents a versatile approach in cancer treatment. Proteins, as central effectors of cellular signaling, metabolism, and structural integrity, often serve as critical drivers of tumor initiation and progression when aberrantly expressed or activated. Therapeutic strategies therefore aim to directly (via aptamers or decoy oligonucleotides) or indirectly (via ASOs or siRNAs) modulate protein levels, activity, or interactions, either by disrupting oncogenic pathways and restoring normal cellular homeostasis, or by harnessing the immune system. This can lead to the suppression of tumor growth, reduced angiogenesis, inhibition of metastasis, and induction of cancer cell death.^13,87,88

Inhibition of transcription factors, regulatory proteins, or oncogenic proteins

Decoy oligonucleotides (also known as intramers) represent an emerging strategy for protein regulation. These synthetic oligonucleotides mimic natural DNA-binding sites to competitively sequester transcription factors or other regulatory proteins, thereby preventing them from engaging their genomic targets. Such decoy strategies broaden the scope of nucleic acid-based therapeutics by enabling direct modulation of nuclear proteins that are otherwise difficult to target with small molecules or antibodies. For example, a cyclic STAT3 decoy binds to pSTAT3 protein, suppressing tumor growth and inducing apoptosis in non-small cell lung cancer (NSCLC).⁸⁹ A phase 0 clinical trial in head and neck squamous cell carcinoma patients demonstrated that intratumoral injection of the STAT3 decoy downregulated key STAT3 target genes, such as Bcl-XL and cyclin D1, with no detectable toxicity.⁹⁰ Ongoing investigations aim to advance this into a phase 1 trial. While the nuclear factor κB (NF-κB) decoy has also demonstrated preclinical anticancer activity, including inhibition of tumor progression in NSCLC and suppression of pulmonary metastasis in osteosarcoma models, it has not yet progressed to cancer trials.^91,92

Aptamers can be designed to interfere with the protein’s function through various mechanisms, such as blocking protein-protein interactions, promoting protein degradation, inhibiting enzymatic activity, or modulating protein conformational changes. Among the earliest and most extensively studied inhibitory aptamers is the thrombin-binding aptamer (TBA), a guanine-rich DNA oligonucleotide originally identified as a high-affinity binder of human α-thrombin through specific interactions with thrombin’s exosite I, thereby competitively inhibiting fibrinogen recognition.⁹³ Although TBA did not advance after phase 1 clinical trials as an anticoagulant, it has since demonstrated notable antiproliferative effects against breast and prostate cancers.⁹⁴ Over 20% of aptamers derived from the systematic evolution of ligands by exponential enrichment process are predicted to adopt G-quadruplex (G4) structures.⁹⁵ TBA itself forms an antiparallel G4 stabilized by Hoogsteen hydrogen bonding and monovalent cations, underscoring its structural uniqueness and highlighting the therapeutic potential of leveraging G4 motifs.^93,96 While TBA has been indispensable as a structural model for understanding G4 folding and stability, it has more recently attracted renewed interest in non-cancer indications, including heart failure and long COVID. Building on insights from TBA, AS1411, a G4 aptamer targeting the external domain of nucleolin, has progressed into clinical trials for multiple cancer-targeted therapies, exemplifying the translational potential of this structural class.^97,98 Other examples include an RNA aptamer inhibiting cytoskeleton-associated protein 4 (CKAP4), a transmembrane receptor, which impedes phosphatidylinositol 3-kinase/AKT signaling pathway and bladder cancer metastasis, and NOX-A12, a Spiegelmer acting as a CXCL12 antagonist, which disrupts homing and accumulation of chronic lymphocytic leukemia (CLL) cells in the bone marrow and delays glioblastoma recurrence.^{99,100,101,102} These encouraging findings have facilitated the advancement of NOX-A12 through various clinical trials targeting multiple cancer, including CLL, multiple myeloma, glioblastoma, and metastatic colorectal and pancreatic cancers (NCT01521533, NCT04121455, NCT01486797, and NCT03168139). Recent interim clinical results from NCT04121455 indicated that patients with newly diagnosed glioblastoma who received NOX-A12 in combination with radiotherapy and bevacizumab experienced consistent tumor reduction, more profound and durable therapeutic responses, and extended survival.¹⁰³ However, the clinical pipeline necessitates further investigation.

Modulation of immunoreceptors

Aptamers can act as immunomodulators, enhancing immune responses by either activating co-stimulatory receptors or obstructing immunosuppressive signals, thereby triggering specific antitumor immune responses. As a result, aptamers have been selected based on their affinity toward immuno-stimulatory receptors, such as STING,¹⁰⁴ CD28,¹⁰⁵ and 4-1BB (also known as CD137),¹⁰⁶ as well as immune checkpoint targets, including CTLA-4,^107,108 NKG2A,¹⁰⁸ PD-L1,^107,109 and Siglec-15.¹¹⁰ It is worth noting that the agonistic or antagonistic effects of aptamers on their respective targets are highly dependent on their form. For example, 4-1BB aptamers require multimerization to induce co-stimulation, and CTLA-4 aptamers, in their tetrameric form, show enhanced affinity and specificity, triggering significant antitumor immunity in melanoma models.^111,112 CD28 aptamers act as antagonists of the co-stimulatory receptor in monomeric form but function as agonists when dimerized, similar to the previously described 4-1BB aptamers.^106,113 Beyond monospecific aptamers, bispecific aptamers have been designed to bind to both immunoreceptors and tumor-specific targets, whether on tumor cells or in tumor stroma. These bispecific aptamers serve as a bridge, physically shortening the distance between immune cells and tumor cells, thereby facilitating antitumor immune responses. For instance, 4-1BB aptamers have been conjugated to aptamers that target vascular endothelial growth factor (VEGF),¹¹⁴ osteopontin,^114,115 or prostate-specific membrane antigen.¹¹⁶ Several more multivalent bispecific aptamers have been comprehensively reviewed elsewhere.¹¹⁷ However, further evaluation is necessary to establish the mechanistic basis, potential new functionalities, and biological outcomes of multispecific/multivalent aptamers.

The utilization of aptamers for protein targeting in cancer therapy shows tremendous potential. Their versatility, small size, and structural flexibility make them appealing for precise and efficient targeting of cancer cells. Nonetheless, their clinical translation has been slowed by the inherent limitations, particularly their short half-life, structural fragility, and lack of standardized delivery platforms compared to siRNAs or ASOs. High affinity does not always translate into robust pathway inhibition or therapeutic efficacy in vivo, as target redundancy or compensation can diminish the biological impact of target inhibition. Structural polymorphism, particularly of G4-based aptamers, also introduces instability and variability under physiological ionic conditions, limiting reproducibility. Moreover, rapid renal clearance, even after chemical modification, prevents sustained target engagement. Beyond these intrinsic barriers, aptamers must also compete with established modalities such as monoclonal antibodies and checkpoint inhibitors that offer longer half-lives and Fc-mediated effector functions. Together, these factors have contributed to the limited clinical survival of otherwise promising candidates. Ongoing research aims to address these limitations through strategies such as multivalent aptamers, aptamer-drug conjugates, and nanoparticle-based delivery to extend circulation and enhance potency. In parallel, improving the efficiency and reliability of the aptamer selection process remains critical, as identifying candidates with high affinity and specificity for target proteins can be a complex and time-consuming process. Advances in these areas will be key to unlocking the full therapeutic promise of aptamers.

Innate immune activation

A vastly different yet increasingly popular approach toward nucleic acid-based anticancer therapy revolves around the utilization of immunomodulators to bolster innate immunity. Unlike strategies that alter tumor-intrinsic gene expression, this approach exploits the immune system’s ability to sense and eliminate malignant cells. At its core are pattern recognition receptors (PRRs), which recognize nucleic acids released from stressed or dying cancer cells as damage-associated molecular patterns. Engagement of PRRs activates signaling cascades that drive the production of type I interferons (IFN-I), proinflammatory cytokines, and chemokines, thereby recruiting and priming immune effector cells. This immunostimulatory environment can convert “cold” tumors to “hot” tumors that are more susceptible to immune-mediated clearance.¹² To therapeutically harness this mechanism, synthetic nucleic acid ligands have been developed to engage cytosolic and endosomal PRRs, such as those targeting retinoic acid-inducible gene I (RIG-I)-like receptors (RLRs), cyclic guanosine monophosphate (GMP)-adenosine monophosphate (AMP) synthase (cGAS), absent in melanoma 2 (AIM2)-like receptors, and Toll-like receptors (TLRs) (Figure 2). These ligands have shown promise in preclinical settings, and several are in clinical development.^12,118 This section delves into the translational advances of these immunomodulatory approaches, emphasizing representative preclinical studies and ongoing clinical trials.

Figure 2.

Nucleic acid-sensing receptors of the innate immune system

Nucleic acid sensors of the innate immune system, including AIM2, cGAS, TLRs, RIG-I, and MDA5, recognize specific DNA and RNA molecules. AIM2, cGAS, RIG-I, and MDA5 recognize cytoplasmic nucleic acids. AIM2 forms inflammasomes with ligands and ASCs, leading to the activation of caspase-1 and production of IL-1β and IL-18. The ER-resident protein STING receives cGAMP produced by cGAS. It interacts with TRAF3 to induce IRF3 phosphorylation via TBK1, or associates with TRAF6 to activate NF-κB through IKKα/β. RIG-I and MDA5 associate with the mitochondrial MAVS adaptor protein, triggering the activation of IRF3, IRF7, and NF-κB through TRAF3- and TRAF6-dependent pathways. TLR3, TLR7, and TLR9 sense nucleic acids in endosomes. TLR3 signals through TRIF, which is associated with TRAF3- and TRAF6-dependent pathways. TLR7 and TLR9 signal through the MyD88 adaptor protein, which is associated with TRAF3 and TRAF6 to activate IRF3, IRF7, and NF-κB. The resulting intracellular signaling cascades transcriptionally induce the expression of genes encoding pro-inflammatory cytokines and type I interferons. AIM2, absent in melanoma 2; ASC, apoptosis-associated speck-like protein containing a CARD; ATP, adenosine 5′-triphosphate; CARD, caspase recruitment domain; cGAMP, cyclic guanosine monophosphate-adenosine monophosphate; cGAS, cGAMP synthase; dsDNA/RNA, double-stranded DNA/RNA; ER, endoplasmic reticulum; GTP, guanosine 5′-triphosphate; IKKα/β, inhibitor of nuclear factor κ kinase subunit α/β; IL-1β/18, interleukin-1β/18; IRF3/7, interferon regulatory factor 3/7; MAVS, mitochondrial antiviral-signaling protein; MDA5, melanoma differentiation-associated protein 5; MyD88, myeloid differentiation primary response 88; NF-κB, nuclear factor kappa light chain enhancer of activated B cells; PYD, pyrin domain; RIG-I, retinoic acid-inducible gene I; ssDNA/RNA, single-stranded DNA/RNA; STING, stimulation of interferon genes protein; TLR3/7/9, Toll-like receptor 3/7/9; TBK1, TANK-binding kinase 1; TRAF3/6, tumor necrosis factor receptor-associated factor 3/6; TRIF, Toll-interleukin-1 receptor domain-containing adaptor-inducing interferon-β.

Short 5′-triphosphate (5′ppp) or 5′-diphosphate (5′pp) RNAs represent potent agonists of cytosolic RLR sensors.^119,120,121 Classic examples are polyinosinic-polycytidylic acid (poly(I:C)) and its derivatives, which trigger IFN-I production, dendritic cell (DC) maturation, natural killer cell activation, and macrophage repolarization toward a tumoricidal phenotype, and exhibit direct cytotoxicity against certain tumor types.^{122,123,124,125,126,127,128,129,130} To improve selectivity and reduce toxicity associated with broad-acting ligands like poly(I:C), newer chemically synthesized 5′ppp duplex RNAs such as M8, SLR14, and immunomodulatory RNA (immRNA) have been developed.^131,132,133 Rational design, including sequence extension and structural tuning, can modulate the inflammatory response. For instance, a modified structure of M8 with a uridine-rich stem loop spanning 99 nt induced a more extensive and robust production of IFN-β and innate inflammatory cytokines through RIG-I activation while sparing MDA5.¹³⁴ Similarly, a small hairpin RNA SLR14, as short as 14 bp and containing a stable tetraloop, constrains RIG-I into a defined orientation, inducing a robust IFN-I response with reduced off-target effects.¹³⁵ More recently, the Luo group developed a series of minimal hairpin immRNAs, each containing 1 purine base insertion along the 10-bp RNA hairpin backbone.¹³⁶ Interestingly, substituting the guanosine at position 9 with adenine (3p10LA9) enhances the potency of immRNA in breast cancer models.^131,136 Early clinical testing underscores both the promise and limitations of RIG-I agonists. In a phase 1 trial of MK-4621 (a 5′ppp RNA formulated for intratumoral injection), patients tolerated the drug and showed biomarker evidence of RIG-I activation, but no objective responses were observed at tested doses (NCT03065023). These results suggest that RIG-I agonists may be most effective in combination regimens. Beyond linear RNAs, circular RNAs, lacking a 5′ end, have been discovered to activate RIG-I and MDA5 in the presence of lysine-63-linked polyubiquitin chains, with evidence of potent adjuvant activity in preclinical tumor models.^137,138 However, an N6-methyladenosine modification on circular RNAs abrogates immune gene activation and adjuvant activity.¹³⁸ Debate remains regarding whether RIG-I activation is mediated directly by the circular RNAs themselves or by other linear RNA contaminants.^138,139

Parallel progress has been made in the cGAS-STING axis, which senses cytosolic DNA to drive strong IFN-I responses. One notable predecessor in this endeavor is cyclic dinucleotides (CDNs). Several types of CDNs, such as c-di-GMP, c-di-AMP (CDA), and 3′-3′- cyclic guanosine monophosphate-adenosine monophosphate (cGAMP), have demonstrated potential in triggering the cGAS-STING signaling cascade in cancer therapy.^{140,141,142,143} However, despite initial enthusiasm, early clinical trials with Merck’s ulevostinag (also known as MK-1454) (NCT03010176 and NCT04220866) and Novartis’s ADU-S100 (MIW815) (NCT03172936, NCT02675439, and NCT03937141) have been terminated due to the lack of impressive anticancer activity as stand-alone treatments in patients with solid tumors or lymphomas. This outcome leads to the realization of the hurdles in CDN development, necessitating a timely re-evaluation. A novel class of CDN candidates, namely dazostinag (TAK-676), incorporates a carbocyclic pyrimidine scaffold to bolster passive permeability and resistance against plasma clearance, enabling intravenous administration and better tumor access. Preclinical studies have demonstrated that dazostinag exhibits robust antitumor activity, coupled with long-lasting immune memory response in mouse models of colon carcinoma and melanoma.^144,145 Building on these promising results, dazostinag has advanced to clinical assessment in cancer patients with advanced solid tumors (NCT04879849 and NCT04420884). Related advances include TAK-500, a dazostinag-ADC, targeting CCR2⁺ cells, currently under investigation both as a monotherapy and in conjugation with pembrolizumab (NCT05070247), as well as KL340399, an intratumoral STING agonist (NCT04096638).

Another cytosolic DNA sensor, AIM2, assembles inflammasomes that promote IL-1β and IL-18 release and pyroptosis.^146,147 One prominent example of a classical AIM2 agonist is poly(deoxyadenylic-deoxythymidylic) acid (poly(dA:dT)), a synthetic construct consisting of repetitive double-stranded DNA (dsDNA) sequences comprising poly(dA-dT):poly(dT-dA) motifs.¹⁴⁸ In a notable investigation on the role of the AIM2 inflammasome in bladder cancer, researchers demonstrate that introducing poly(dA:dT) into tumors can activate AIM2, resulting in amplified therapeutic effectiveness of the Bacillus Calmette-Guérin vaccine.¹⁴⁹ However, its non-specific activation of cGAS and RIG-I emphasizes the need for more selective AIM2 agonists.^150,151,152 Currently, AIM2 modulation remains at the preclinical stage.

TLRs constitute a crucial family of PRRs.¹⁵³ Among them, TLR3 recognizes dsRNA and exhibits particular sensitivity to activation elicited by the aforementioned poly(I:C) and derivatives within the endosomes.¹⁵⁴ One derivative, poly-IC12U (Ampligen [rintatolimod], Hemispherx), selectively stimulates TLR3 while minimizing toxicity by avoiding MDA5 activation.¹⁵⁵ The adjuvant potential of poly-IC12U is currently being evaluated in the context of cancer vaccines, particularly in combination with intraperitoneal release alongside DC-based vaccines in patients afflicted with ovarian cancer (NCT02432378).^{156,157,158,159} TLR9 recognizes bacterial DNA motifs containing unmethylated cytosine-phosphate-guanine (CpG) dinucleotide, activating T helper 1 cell immunity through the MyD88-dependent pathway.^153,160,161 Several CpG oligodeoxynucleotides (ODNs) have been explored in clinical trials. Notable examples include CpG 1018 (Dynavax), which has shown promise as an adjuvant with anti-CD20 antibodies for non-Hodgkin lymphoma, and MGN1703 (lefitolimod), a DNA molecule that forms a distinctive dumbbell-shaped structure and has demonstrated efficacy in metastatic tumors.^{160,162,163,164,165,166,167} Other CpG ODNs such as CpG 7909, SD-101, and CpG K3 have also been studied as adjuvants in cancer immunotherapy.^161,168

While transitioning from proof-of-principle studies to clinical testing, nucleic acid immunomodulators face several challenges. Systemic administration is constrained by dose-limiting toxicities such as cytokine storm and widespread inflammation, thus making tumor-selective delivery a critical priority. Tumor heterogeneity further limits the uniformity of response, particularly in tumors with insufficient immune infiltrates or suppressive microenvironments that dampen PRR signaling. Given the genetic and epigenetic adaptability exhibited by tumor cells, they tend to evade therapies that target a single factor, such as specific oncogene/kinase inhibitors and immunotherapies aimed at tumor antigens.¹⁶⁹ Indeed, early trials confirm that monotherapy rarely induces durable tumor regression, indicating that most PRR agonists likely require integration into rational combination regimens with checkpoint blockades, vaccines, or conventional therapies. To overcome these limitations, innovative strategies are pursuing bifunctional RNAs that combine RIG-I activation via 5′ppp with gene silencing (e.g., targeting BCL-2, TGF-β, miR-125b), demonstrating efficacy in preclinical models of melanoma and pancreatic and breast cancers.^131,170,171 Nonetheless, this field is still in its infancy. Species-specific receptor expression and signaling differences further complicate clinical predictability. For example, TLR9 is situated on the endosomal membrane, necessitating uptake and endosomal maturation of CpG DNA to exhibit immunostimulatory effects. However, TLR9 expression pattern differs between species. Murine TLR9 is expressed in both myeloid and plasmacytoid DCs, while human TLR9 is exclusively expressed in plasmacytoid DCs. This difference is likely to influence the translation of CpG-based therapies from preclinical mouse models to clinical studies in humans.

Advances in nucleic acid drug design and delivery to improve stability, biodistribution, efficacy, and safety

Nucleic acid-based therapeutics for anticancer therapy face several challenges that hinder their successful translation from the laboratory to clinical applications. Their large size and negative charge hinder cellular uptake, while enzymatic degradation, rapid clearance, and poor tumor penetration limit pharmacokinetics and biodistribution. Off-target effects, endosomal trapping, and unintended immune activation further complicate clinical use.^12,172 Addressing these barriers requires two complementary strategies: chemical modification of nucleic acids and specialized delivery systems that stabilize, protect, and guide therapeutic constructs to their desired sites of action.

Chemical modification

Nucleic acids can be chemically modified to confer favorable therapeutic properties, such as greater nuclease resistance, higher target binding affinity, and reduced toxicity and immunogenicity. The choice of modifications depends on factors such as target genes, administration route, and delivery approach. Some modifications act effectively on their own.¹⁷³ Most approved oligonucleotide drugs incorporate multiple, complementary modifications rather than relying on a single change, as combinations tend to improve stability and efficacy without compromising function.^174,175,176 However, it should be taken into consideration that certain modifications intended to achieve a specific purpose may have unintended detrimental consequences. For example, modifications aimed at enhancing binding specificity might obstruct RNase H-mediated cleavage, while others could induce hepatotoxicity. Thus, care should be taken when incorporating modifications into nucleic acids. This section focuses on the chemical modifications of oligonucleotides, including ASOs and siRNAs, which have been widely applied and approved for clinical use, and to some extent, particular classes of nucleic acids such as immune agonists and aptamers (Figure 3).

Figure 3.

Structures of common chemical modifications used in nucleic acid therapeutics

(A) First-generation modifications primarily target the phosphate backbone to enhance nuclease resistance. (B) Second-generation modifications focus on the 2′ position of the ribose ring to improve stability and binding affinity. (C) Third-generation modifications involve nucleic acid analogs to further enhance specificity and therapeutic efficacy. Among these analogs, peptide nucleic acids (PNAs) can be modified at the α-, β-, or γ-position with side chains such as cationic quaternary ammonium groups, chiral amino acids, and PEGylated or miniPEG moieties to improve solubility, cellular uptake, and hybridization properties.

Modifications to the backbone

The first generation of chemical modifications focused on enhancing nuclease resistance. Phosphate backbone modifications, such as phosphorothioate (PS), are widely utilized to enhance bioavailability through reduced renal clearance and improved binding affinity to plasma proteins, primarily albumin and α2-macroglobulin. PS linkages replace a non-bridging oxygen with sulfur, generating Sp and Rp stereoisomers with distinct nuclease sensitivity and RNase H support. However, both stereoisomers may alter oligonucleotide function, making excessive PS modification inadvisable. One major limitation of PS chemistry lies in the exponential increase in stereoisomeric diversity as the number of modified linkages grows. Each sulfur substitution generates two possible chiral configurations, and therapeutic oligonucleotides with multiple PS linkages produce highly complex mixtures containing hundreds of thousands of unique stereoisomers. For example, mipomersen, which contains 19 PS linkages, can theoretically yield more than 500,000 distinct stereoisomeric forms. This heterogeneity affects biochemical behavior, including nuclease resistance, protein binding, tissue distribution, and even toxicity. Consequently, a single antisense drug in clinical use may actually represent a heterogeneous population of molecules with variable biological properties, complicating both pharmacological predictability and regulatory assessment. Considerable efforts have been directed toward the development of synthesis techniques that allow preparation of stereochemically pure PS linkages. These methods enable the design of oligonucleotides with defined chiral configurations, offering the potential to reduce heterogeneity, enhance reproducibility, and improve the therapeutic index, representing an important next step in antisense drug development.^177,178 Fomivirsen (Vitravene), a 21-mer PS oligo-2′-deoxynucleotide, exemplifies the clinical potential of PS chemistry. It was the first approved antisense drug for treating cytomegalovirus (CMV) retinitis in immunocompromised patients, particularly those afflicted with acquired immunodeficiency syndrome (Table 1).^{173,179,180,181}

Table 1.

Modifications of nucleic acid therapeutics approved by the FDA

Drugs	Modifications	Molecular targets	Indications	Delivery system/dose	Administration route/target tissue	FDA approval year	References
ASOs
Fomivirsen	PS	CMV IE-2	CMV retinitis	naked/330 μg per eye every 4 weeks	IVI/eye	1998	Jabs and Griffiths173; Perry and Balfour179; de Smet et al.180; Geary et al.181
Mipomersen	PS, 2′-MOE	ApoB-100	HoFH	naked/200 mg once weekly	SC/liver	2013	Hair et al.182; Duell and Jialal183; Santos et al.184; Raal et al.185
Nusinersen	PS, 2′-MOE	exon 7 of SMN2	SMA	naked/12 mg every 4 months	ITT/CNS	2016	Haché et al.186; Finkel et al.187; Chiriboga et al.188
Eteplirsen	PMO	exon 51 of DMD	DMD	plasma proteins/30 mg kg−1 once weekly	IV/muscle	2016	Mendell et al.189; Lim et al.190; Khan et al.191
Inotersen	PS, 2′-MOE	TTR	hereditary ATTR	naked/284 mg once weekly	SC/liver	2018	Benson et al.192; Crooke et al.193
Volanesorsen	PS, 2′-MOE	ApoC-3	FCS	naked/285 mg once weekly	SC/liver	2019	Post et al.174; Witztum et al.194
Golodirsen	PMO	exon 53 of DMD	DMD	naked/30 mg kg−1 once weekly	IV/muscle	2019	Frank et al.195
Viltolarsen	PMO	exon 52 of DMD	DMD	naked/80 mg kg−1 once weekly	IV/muscle	2020	Dhillon196; Clemens et al.197
Casimersen	PMO	exon 45 of DMD	DMD	naked/30 mg kg−1 once weekly	IV/muscle	2021	Zakeri et al.198; Shirley199
Tofersen	PS, 2′-MOE	SOD1	ALS	naked/100 mg once biweekly (initial) or 100 mg once monthly (maintenance)	ITT/CNS	2023	Meyer et al.200; Blair201
Eplontersen	PS, 2′-MOE	TTR	hereditary ATTR	GalNAc/45 mg once monthly	SC/liver	2023	Viney et al.202; Coelho et al.203
siRNAs
Patisiran	PS, 2′-OMe, 2′-F	TTR	hereditary ATTR	LNP/0.3 mg kg−1 every 3 weeks	IV/liver	2018	Adams et al.204; Urits et al.205
Givosiran	PS, 2′-OMe, 2′-F	ALAS1	AHP	GalNAc/2.5 mg kg−1 once monthly	SC/liver	2019	Sardh et al.206; Syed207
Lumasiran	PS, 2′-OMe, 2′-F	HAO1	PH1	GalNAc/3 mg kg−1 every 3 months	SC/liver	2020	Garrelfs et al.208
Inclisiran	PS, 2′-OMe, 2′-F	PCSK9	HeFH	GalNAc/284 mg every 6 months	SC/liver	2021	Ray et al.209; Merćep et al.210
Vutrisiran	PS, 2′-OMe, 2′-F	TTR	hereditary ATTR	GalNAc/25 mg every 3 months	SC/liver	2022	Habtemariam et al.211; Adams et al.212
Nedosiran	PS, 2′-OMe, 2′-F	LDH	PH1	GalNAc/160 mg once monthly	SC/liver	2023	Syed213; Hoppe et al.214
Aptamers
Pegaptanib	PS, 2′-OMe, 2′-F	VEGF165	AMD	PEG/0.3 mg every 6 weeks	IVI/eye	2004	Bakri et al.215; Fine et al.216
Izervay	PS, 2′-OMe, 2′-F	C5 complement protein	geographic atrophy	PEG/2 mg once monthly	IVI/eye	2023	Danzig et al.217; Al Shaer et al.218

AHP, acute hepatic porphyria; ALAS1, 5′-aminolevulinate synthase 1; ALS, amyotrophic lateral sclerosis; AMD, age-related macular degeneration; ApoB-100, apolipoprotein B-100; ApoC-3, apolipoprotein C3; ATTR, transthyrein amyloidosis; CMV, cytomegalovirus; CNS, central nervous system; DMD, Duchenne muscular dystrophy; 2′-F, 2′-fluoro; FCS, familial chylomicronemia syndrome; GalNAc, N-acetylgalactosamine; HAO1, hydroxyacid oxidase 1; HeFH, heterozygous familial hypercholesterolemia; HoFH, homozygous familial hypercholesterolemia; IE-2, immediate-early 2; ITT, intrathecal injection; IV, intravenous injection; IVI, intravitreal injection; LDH, hepatic lactate dehydrogenase; LNP, lipid nanoparticle; 2′-MOE, 2′-O-methoxy-ethyl; PCSK9, proprotein convertase bubtilisin/kexin type 9; PH1, primary hyperoxaluria type 1; PMO, phosphorodiamidate morpholino oligomer; PS, phosphorothioate; SC, subcutaneous injection; SMA, spinal muscular atrophy; SMN2, survival motor neuron 2; SOD1, superoxide dismutase 1; TTR, transthyretin; 2′-OMe, 2′-O-methylation; VEGF, vascular endothelial growth factor.

Beyond standard oligonucleotides, backbone modifications have been implemented in the aforementioned CDNs and ODNs. Natural CDNs are susceptible to rapid degradation by host phosphodiesterases, limiting their stability and efficacy. To overcome this, researchers have developed a series of synthetic CDN derivatives designed to resist enzymatic cleavage and enhance STING affinity.²¹⁹ For instance, the Rp, Rp (R,R) dithio-substituted CDA incorporates a dithio diastereomer with both 2′-5′ and 3′-5′ phosphate linkages, significantly improving resistance to phosphodiesterases and antitumor activity across diverse cancer models, including melanoma, colon cancer, breast cancer, pancreatic adenocarcinoma, lung adenocarcinoma, and squamous cell carcinoma.^140,219 Similarly, synthetic ODNs with PS-linked CpG motifs showcase enhanced nuclease resistance, prolonged half-life, and optimized immunostimulatory potential for 5′-purine-purine-CpG-pyrimidine-pyrimidine-3′ structures.^220,221

Modifications to the ribose

Second-generation chemistry targets the ribose 2′-O position to improve nuclease resistance and binding affinity and reduce immune stimulation. To date, the most often used modifications in clinical practice are 2′-fluoro (2′-F), 2′-O-methyl (2′-OMe), and 2′-O-methoxyethyl (2′-MOE). In the 2′-OMe ribose modification, a methyl group is added to the 2′ hydroxyl of the ribose molecule. It is naturally found in RNAs and enhances thermal stability. The 2′-F modification is small in size and highly electronegative, allowing the nucleotide to acquire a C3′-endo conformation and provide considerable gains in binding affinity. 2′-F also enhances nuclease stability and dramatically decreases immunological stimulation in in vitro models.²²² One disadvantage of the 2′-F modification is that it offers less protection against nuclease degradation than 2′-OMe and 2′-MOE. A combination of 2′-F- and 2′-OMe-modified nucleotides can overcome this limitation. In practice, 2′-F in combination with 2′-OMe modifications is used in most approved therapeutic siRNA drugs (Table 1).^{204,205,206,207,208,209,210,211,212,213,214}

The 2′-MOE modification is even bulkier. It locks ribose in the C3′-endo sugar pucker, stabilizing its conformation and increasing melting temperatures. The 2′-MOE modification can discriminate the target sequence from a similar, non-targeting sequence that includes mismatched base pairs, reducing the probability of off-target effects.²²³ Nearly 50% of FDA-approved ASO drugs contain 2′-MOE modification (often with PS backbones), and several drug candidates are under clinical trials in phase 3. For example, mipomersen (Kynamro, ApoB ASO)^{182,183,184,185} and nusinersen (Spinraza, SMN2 splice-switching ASO)^186,187,188 are fully 2′-MOE modified. Other FDA-approved 2′-MOE drugs include inotersen,^192,193 volanesorsen,^174,194 tofersen,^200,201 and eplontersen (Table 1).^202,203

Third-generation chemistry employs locked nucleic acids (LNAs) that covalently tie the 2′-oxygen to the 4′-carbon, yielding very high affinity and melting temperatures. Fully modified LNAs are unable to exert RNase H cleavage. LNA gapmers consist of short DNA oligonucleotide sequences in the middle with modified RNA sequences on either side, leveraging the high affinity of LNA while restoring RNase H activity. LNA and related bicyclic sugars (e.g., cEt, constrained ethyl) greatly increase binding, but they were associated with hepatotoxicity if overused.²²⁴ Modern gapmer designs use only a few LNA/cEt residues to boost potency while avoiding off-target cleavage and are currently being used for the treatment of a number of different conditions, including many cancers.

Alternative sugar modifications are under evaluation in cancer treatment. Arabinose nucleic acid (ANA) modifies the ribose stereochemistry, forming highly stable duplexes effective in antisense gapmers.²²⁵ Fluoroarabinose nucleic acid combines the benefits of ANA and 2′-F, enhancing gene-silencing capacity with activity reported against BCL2 in cancer cells.²²⁶ Threose nucleic acid (TNA), with a four-carbon sugar backbone, offers high chemical stability and strong hybridization to RNA. Although less developed clinically, TNA has demonstrated efficient strand invasion and high biostability, raising interest for antitumor approaches targeting oncogenes such as c-Myc and Akt.^227,228 Recent studies have explored dideoxy-2′,3′-cGAMP analogs (e.g., dideoxy-2′,3′-cAAMP), which enhance IFN-stimulatory activity and antitumor efficacy in mouse models compared to natural 2′,3′-cGAMP.²²⁹

Modifications to bases

Base modifications involve the use of modified nitrogenous bases that can increase specificity and reduce immune activation. However, they are also commonly associated with increased toxicity. 5-Methyl dC prevents CpG motif recognition by TLR9, while pseudouridine or methylated bases blunt the innate immune response.²³⁰ Recent studies with LNA gapmers identified base derivatives such as 2-thiothymine, 5-hydroxycytosine, and 8-bromoguanine that decrease hepatotoxicity without substantially altering target binding,²³¹ Thus, despite the advances, a great deal of work is required before these strategies can be safely used in clinical practice.

Nucleic acid analogs

The two most common nucleic acid analogs used for oligonucleotide therapeutics are phosphorodiamidate morpholino oligomers (PMOs) and peptide nucleic acids (PNAs). PMOs replace the ribose ring with a morpholine moiety and use a neutral methylene phosphorodiamidate linker. This non-natural backbone is completely resistant to nucleases and proteases. PMOs confer stronger hybridization to the target RNA, thereby conferring higher silencing efficiency. However, PMOs are rapidly cleared due to their neutrality. Several PMO drugs have been approved for exon skipping in Duchenne muscular dystrophy, such as eteplirsen,^189,190,191 golodirsen,¹⁹⁵ viltolarsen,^196,197 and casimersen (Table 1).^198,199

PNAs use a neutral peptide-like backbone (N-(2-aminoethyl)-glycine) in place of the sugar-phosphate backbone. PNAs bind complementary DNA or RNA with exceptionally high affinity and specificity because they lack the electrostatic repulsion of natural backbones.²³² This neutrality also makes PNAs stable in biological fluids and resistant to nucleases and proteases. However, like PMOs, the neutral charge of PNAs causes rapid renal clearance unless delivered by special means. A limitation of unmodified PNAs is their tendency to form both parallel and antiparallel duplexes with DNA or RNA, complicating hybridization and reducing selectivity. To overcome this, various backbone-modified PNAs have been developed, including α-, β-, and γ-modified PNAs bearing side chains. For example, modifications at the α-position of the backbone often incorporate cationic or hydrophilic groups, improving hybridization fidelity and reducing non-specific interactions; modifications at the β-position can introduce PEGylated chains or miniPEG (PEG, polyethylene glycol) groups, which improve water solubility, reduce aggregation, and enforce a preferred helical conformation that prevents parallel binding; modifications at the γ-position can introduce positively charged lysine-like side chains, which enhance solubility and improve cellular uptake.^232,233 Analogous to ribose-modified oligonucleotides, sugar-like modifications can be incorporated into PNA backbones to further tune binding, stability, and nuclease resistance—for instance, using fluorinated or arabinose-like residues in the backbone or at terminal positions. Such derivatives are being actively explored as next-generation antisense or antigene agents. To date, however, no PNA therapeutic has advanced to the clinic.

Therapeutic aptamers likewise benefit from extensive chemical modifications. 3′ End alterations such as inverted thymidine or biotin can enhance stability by resisting 3′-exonuclease degradation in serum,^234,235 while 2′-ribose modifications (e.g., 2′-F, 2′-OMe, 2′-aminopyrimidine, LNA) protect RNA aptamers from alkaline hydrolysis. However, 2′-aminopyrimidine is less favored due to pairing instability and synthetic challenges.^{236,237,238,239,240,241,242} Engineering the aptamer backbone can increase nuclease resistance and facilitate cellular penetration by reducing the negative charge of the phosphodiester backbone, thus weakening electrostatic repulsion.^243,244,245 Base modifications, such as five-position uridine derivatives, further protect aptamers against nuclease degradation and thermal denaturation.²⁴⁶ For G4-based aptamers such as TBA, structural fine-tuning can strongly influence functional outcomes. Terminal modifications with aromatic moieties (e.g., naphthalene diimides, dialkoxynaphthalenes) at the 5′ and 3′ ends of TBA enhance anticoagulant properties by altering structural dynamics and strengthening thrombin affinity.⁹³ Site-specific substitutions at Thy3 and Thy12 with amino acid derivatives, aromatic carboxylic acids, or carbohydrate azides improve binding stability. Stabilization strategies, such as the light-triggered stapling, which incorporates photo-crosslinkers (e.g., xanthotoxin with a 3-cyanovinylcarbazole moiety) or a phenyl-furan moiety at the 5′ end, lock the G4 structure, extend metabolic stability, and allow precise control of conformational states, making TBA more reliable in therapeutic settings.^96,247 In anticancer applications, adding additional G-tetrads into TBA has been explored to preserve its antiproliferative effects while minimizing anticoagulant activity.⁹⁴ Beyond sequence-level modifications, higher-order engineering approaches such as circularization and nanoengineering strategies, including spherical nucleic acids, have proven valuable in overcoming the sensitivity of therapeutic aptamers, particularly RNA aptamers, to nuclease, thereby expanding their potential application in cancer therapy.^248,249 These diverse strategies demonstrate how chemical and structural modifications are central to advancing therapeutic aptamers. Notably, two aptamer drugs featuring diverse modifications have been granted FDA approval (Table 1).^{215,216,217,218}

Delivery strategies

Despite the advent of promising nucleic acid-based drugs and advances in their structural and molecular designs, only a small percentage have proven efficacious in animal models, and fewer still have entered clinical development. The major bottleneck in realizing the potential of nucleic acid therapeutics is their delivery. Naked nucleic acids are susceptible to degradation by nucleases present in body fluids and may also trigger immune responses.^{250,251,252,253} Furthermore, such nucleic acids lack specific targeting capability, thus necessitating high doses and causing off-target effects. Most important, naked nucleic acids are unable to bypass cellular barriers efficiently. Nucleic acids cannot easily cross the plasma membrane or, for DNA, the nuclear membrane.^254,255 Nucleic acids that are endocytosed may become entrapped and degraded in endosomal compartments.²⁵⁶ As described in the previous section, molecular and construct modifications of nucleic acids may contribute to overcoming some of these challenges, but they are only sufficient by themselves in a limited number of settings. Developing specialized drug delivery systems represents a more generalizable and modular solution for enhancing the delivery of therapeutic nucleic acids to desired subcellular compartments. This section describes five major classes of delivery systems for nucleic acid therapeutics, discusses their current stage of development, and highlights the advantages and drawbacks of each approach (Figure 4).

Figure 4.

Comparison of current delivery systems for nucleic acid-based therapies

This figure compares various nucleic acid delivery systems across several key parameters, including transfer efficiency, safety/biocompatibility, current scope of application, compositional complexity, and ease of manufacturing. This comparison highlights the strengths and limitations of each system, helping to guide decisions on the most suitable delivery approach for different nucleic acid-based therapeutics.

Nucleic acid conjugates

For small nucleic acid molecules, a ligand may be directly attached via a covalent linkage to grant them specific targeting, improve their pharmacokinetics, and/or enhance cellular uptake. A wide variety of ligands have been tested for these purposes, ranging from small molecules such as cholesterol, folate, and monosaccharides to peptides, antibodies, protein scaffolds, and even other nucleic acids. The conjugation of ligands via solid-phase (high efficiency) or solution-phase (broader applicability) approaches yields conjugates that enhance cellular uptake and specific targeting of tissues.²⁵⁷ Carbohydrate N-acetyl galactosamine (GalNAc)-conjugated siRNA has shown commercial success in liver targeting, with ongoing clinical trials in hepatic diseases.²⁵⁸ In hepatocellular carcinoma, however, off-target risks and reduced efficacy limit GalNAc usage. Alternatively, cholesterol or other lipophilic groups can direct nucleotides to the liver or other organs via lipoprotein binding.²⁵⁹

While conjugates perform well in some scenarios, their application scope is limited. Naked mRNA and plasmid DNA remain vulnerable to nuclease degradation even when conjugated. Antibody-oligonucleotide conjugates are possible but technically challenging, as controlling the sites and number of oligonucleotides per antibody is difficult. Many conjugated oligonucleotides remain small enough for rapid renal clearance.²⁶⁰ Unless ligands associate with long-circulating factors, these oligonucleotides rely on first-pass cellular uptake and are unable to persist for prolonged effects. Unlike carrier systems such as viruses or some LNPs, conjugated nucleic acids lack intrinsic endosomal escape mechanisms, with only a small fraction reaching the cytosol.²⁵⁸

Viruses

Viruses are naturally adept at delivering genetic material into host cells, making them valuable vector for gene therapy.¹⁰ Adenoviruses, adeno-associated viruses, lentiviruses, and herpes simplex viruses (HSVs) are currently the most clinically relevant.¹⁰ One key reason for their widespread use is the ability to transduce both dividing and non-dividing cells.²⁶¹ These viruses also have broad tropism, although serotype-specific variations exist.^262,263,264 For example, lentiviruses can permanently integrate their genome into the host genome via reverse transcriptase and integrase enzymes. This feature is advantageous for treatment contexts where the prolonged expression of a therapeutic transgene is desired, such as for chimeric antigen receptor T cell therapy or repeated drug administration to destroy cancer cells expressing a drug-responsive transgene products.^{265,266,267,268,269,270} Other general features of these viral vectors are contrasted in Table 2. The advantages and limitations listed for each vector are relative to other viral vectors. Viral vectors have been widely explored in cancer immunotherapy, both in vivo and ex vivo, exemplified in the treatment of leukemia, lymphoma, melanoma, prostate cancer, HPV-associated cancers, colorectal cancer, and pancreatic cancer.^{271,272,273,274,275} Beyond protein-encoding transgenes, viral vectors have shown potential for conveying genes encoding small silencing RNAs, providing more robust and persistent silencing effects compared to direct RNA transfer.^276,277,278

Table 2.

General features of common viral vectors for nucleic acid therapeutic delivery

Viral vector	Basic characteristics	Advantages	Limitations
Adenoviruses	•∼26- to 48-kb double-stranded DNA genome •non-enveloped •non-integrating	•high capacity •transduction of both dividing and non-dividing cells •broad tropism	•high immunogenicity (especially earlier generations) •preexisting immunity prevalent in general population
Lentiviruses	•∼9.7-kb single-stranded RNA genome •enveloped •integrating in nature, non-integrating versions available	•low immunogenicity •integration allows stable gene expression •transduction of both dividing and non-dividing cells •broad tropism	•risk of insertional mutagenesis •moderate capacity
Adeno-associated viruses	•∼4.7-kb single-stranded DNA genome •non-enveloped •non-integrating	•low immunogenicity •minimal health risk as the wild-type virus is not known to cause disease •ease of cloning due to small genome •diffusion through dense tumor mass due to small size •transduction of both dividing and non-dividing cells •broad tropism	•low capacity •requires complementary strand synthesis in host, hence delayed transgene expression, or double-stranded versions need to be generated
HSVs	•∼152-kb double-stranded DNA genome •enveloped •non-integrating	•high capacity •transduction of both dividing and non-dividing cells •broad tropism, including neuronal tropism	•health risk as the wild-type virus is highly dangerous

Despite their superior nucleic acid transfer efficiency and versatility, viral vectors harbor safety concerns. Components of viral origin may activate both innate and adaptive immune responses, causing acute immunotoxicity or immune memory that makes repeated dosing infeasible.²⁷⁹ Preexisting immunity to wild-type viruses also restricts usable serotypes or necessitates isolation from non-human animals.^280,281 Integrating viruses, such as retroviruses, may cause mutagenesis via random insertion.²⁸² While difficult to entirely eliminate, these risks have been reduced in newer viral generations.^{283,284,285,286} On the flip side, it is worth noting that the immunogenicity of viral vectors can be beneficial for mobilizing immune responses in vaccine applications, including cancer vaccines, although this advantage does not negate patient safety risks.²⁸⁷ Ex vivo delivery largely avoids these immunogenicity concerns.

Another major hurdle in viral vector translation is the high cost of manufacturing. Several production systems exist, and none is without limitations.²⁸⁸ Transiently transfected cells consume significant time, labor, and materials per virus unit. Stable cell lines allow scalable production but demands high initial investment and are inflexible, in that any vector modification requires the generation of a new cell line. In addition, deletion of viral replication and packaging genes necessitates helper viruses and/or complementing producer cells. This introduces contaminants, complicating purification and quality control.

LNPs

LNPs are a widely used class of nucleic acid carriers. Initially conceived as liposomes to encapsulate nucleic acids, LNPs have evolved into ionizable cationic lipids assembled with helper lipids, cholesterol, and PEG that minimize toxicity while improving endosomal escape and circulation half-life.^{289,290,291,292} For any of these lipid components, variations in molecular size, relative abundance, degree of saturation, and the specific chemical groups could have a considerable impact on the overall efficacy of the system. Today, both the more recent LNP designs and simpler lipoplexes as well as neutral liposomes, which have enjoyed a longer history of optimization and validation, are being developed for nucleic acid therapeutics.

LNPs are now widely used in the formulation of mRNA therapeutics, such as vaccines, and are under investigation in various cancer immunotherapy trials. FixVac (phase 2 for melanoma) uses cationic liposomes complexed with mRNA encoding four tumor-associated antigens administered via intravenous injection. Interim analysis showed stimulation of cytokine and CD4⁺ and CD8⁺ T cell responses with or without co-administration of the PD1 checkpoint inhibitor.²⁶ Another candidate in trials, mRNA-4650, employs intramuscularly injected LNPs containing the ionizable lipid MC3 to transport mRNA for antigens expressed by the autologous cancer in patients with gastrointestinal cancer, eliciting antigen-specific T cell responses.²⁹³ LNPs have also been used to deliver mRNA encoding anticancer proteins such as OX40L and proinflammatory cytokines IL-12 and IL-27.^294,295

The trade-off between safety and potency remains a problem for in vivo delivery using LNPs. As exemplified above, constitutively cationic lipids are toxic in both in vivo and ex vivo settings, and many formulations include synthetic, poorly biodegradable lipids. PEGylated lipids, widely used to extend circulation times, also raise PEG-specific antibodies.²⁹⁶ Studies tracking the fate and effects of such lipid components in biological systems are essential to avoid unexpected adverse responses in patients. Furthermore, LNPs tend to acquire a protein corona in body fluids, which alters their surface properties and thus their biodistribution and pharmacokinetics.^297,298 Being coated with lipoproteins in the serum, for example, might direct LNPs toward hepatocytes instead of their intended targets. The protein corona might also obscure “stealth” moieties on LNP surfaces that are meant to prevent the systemic clearance.²⁹⁸ It is worth noting, however, that deepening knowledge of how the protein corona gathers on LNPs has initiated efforts to design LNPs that preferentially adsorb proteins that facilitate distribution to tumors.²⁹⁹ Where possible, local administration, rather than systemic, can also mitigate unwanted immune responses and problems associated with the protein corona.

Nucleic acid polyplexes

Polyplexes are formed through electrostatic condensation when anionic nucleic acids and cationic polymers are mixed. The resultant complex is stable, compact, and protected from degradation. Early polymers such as diethylaminoethyl-dextran and poly-l-lysine achieved high cellular uptake but low functional nucleic acid transfer, likely due to endosomal entrapment of the cargo.³⁰⁰ This issue can be overcome by using polymers such as polyethylenimine (PEI) and dendritic polyamidoamine (PAMAM) dendrimers. These polymers are only partially protonated at neutral pH, and their continued protonation in acidic endosomal compartments possibly disrupts the compartment membrane, thus releasing the polyplexes, although the precise mechanisms are debated.³⁰¹ PEI and PAMAM, however, have limited biodegradability and may be toxic.^302,303 Such problems have motivated exploration of natural polymers, notably chitosan, cyclodextrin, and atelocollagen, as well as the design of more biodegradable PEI and PAMAM dendrimers.^{304,305,306,307} More complex designs have emerged in response to different challenges, such as polyplexes containing copolymers made from two or more species of monomers. Block copolymers between PEG and the cationic polymer are especially common due to the favorable “stealth” properties of PEG.³⁰⁸

Polyplexes have been used to deliver many different types of nucleic acids, although most formulations undergoing clinical trials involve plasmid DNA. These include a completed phase 2 trial for bladder cancer, where PEI polyplexes were used to transfer a plasmid containing the diphtheria toxin-A gene under the control of the H19 promoter via intravesical injection.³⁰⁹ In ovarian cancer, a cholesterol- and PEG-conjugated PEI lipopolymer was tested in a phase 1 trial to deliver a plasmid for expressing IL-12 via intraperitoneal administration.³¹⁰ As with the preclinical mouse models, the treatment elicited local IFN-γ production but no systemic toxicity. In the siRNA space, a phase 1 trial used a β-cyclodextrin polymer complexed with PEG-transferrin to target solid tumors.³¹¹ Despite promising preclinical results, this trial was terminated due to dose toxicity-limiting events in enrolled patients.

Nucleic acid polyplexes share many challenges with lipoplexes, including PEG-related issues, cationic polymer toxicity, and protein corona complications. To circumvent the problems associated with cationic polymers, decationized polyplexes have emerged as an interesting alternative that strikes a balance between nucleic acid encapsulation efficiency and treatment safety.³¹² Such polyplexes have shown promise in delivering plasmid DNA and siRNA, although functional delivery in vivo is not yet well studied.^312,313

EVs

EVs are small, membrane-bound particles released by cells that carry active biological agents, including proteins, nucleic acids, lipids, and carbohydrates, which function in physiological and pathological processes.^314,315 Their natural role in shuttling nucleic acids to and from cells has made them attractive as a new type of delivery vehicle for nucleic acid therapeutics. EVs can be loaded with nucleic acids either post-isolation using chemical reagents (e.g., cationic transfection reagents) or physical methods (e.g., electroporation, sonication), or they can be pre-loaded by transfecting the EV-producing cells with the desired nucleic acid or a genetic construct encoding it.³¹⁶ More recently, elaborate approaches have been developed to enhance the loading efficiency of endogenously produced nucleic acids into EVs via the incorporation of RNA-binding proteins fused to EV-enriched membrane proteins. The addition of cleavable linkers, internal ribosome entry sites, or inteins allows the subsequent release of the RNA to the EV lumen, where they can be delivered to the recipient cell.³¹⁷

As opposed to synthetic and viral materials, EVs are derived from cells, making them biodegradable and biocompatible. Various nucleic acids, mostly ASOs, siRNAs, and even CRISPR-Cas9 components, have been successfully delivered by EVs in animal models for various cancers, including acute myeloid leukemia and breast, pancreatic, liver, and gastric cancers.^{8,9,318,319,320,321,322} In in vivo applications, EVs from mesenchymal stem cells (MSCs) are a popular candidate, due to their demonstrated safety, relative ease of production, and inherent antitumor and regenerative properties.³²³ HEK293 cells are also commonly used to produce EVs due to the ease of culturing and transfecting them.^324,325 Human EVs from plasma and red blood cells (RBCs) have likewise proven effective at delivering functional silencing RNAs to suppress tumors.^9,326,327 Non-human sources, such as plants and bovine milk, are also being explored for large-scale EV production for therapeutic uses.^318,328

Importantly, the presence of endogenous biomolecules in EVs may confer advantageous properties for EV-facilitated drug delivery. CD47 naturally present on RBC-derived EVs, for example, acts as an anti-phagocytosis signal and can thus improve the blood retention of the EVs.³²⁹ As mentioned, MSC-derived EVs contain proteins and miRNAs that display anticancer properties.³²³ Moreover, EV-producing cells can be genetically engineered to express targeting molecules fused to EV membrane proteins, as demonstrated with artificial EGFR-targeting peptides and EGFR-targeting nanobodies.^330,331 Alternatively, targeting ligands can be attached to EVs post-isolation using systems such as streptavidin-biotin tagging, click chemistry, or enzymatic ligation. For instance, EGFR-binding nanobodies have been enzymatically ligated to RBC-derived EVs for delivery of an ASO against miR-125b and an immRNA that activates the RIG-I pathway in breast cancer.¹³¹

Despite their attractive features as therapeutic carriers, EVs possess the highest level of compositional complexity and population heterogenicity out of the systems described here. While some endogenous EV components may improve the delivery performance of EVs, other components could instead impair it. For instance, plasma membrane translocase activity during microvesicle biogenesis tends to produce EVs with increased phosphatidylserine exposure, which might accelerate their phagocytic clearance or alter cell-targeting properties.³¹⁴ Studies have furthermore demonstrated that EV samples from the same source consist of multiple subsets that diverge in size, protein and nucleic acid profiles, and lipid composition or arrangement.^332,333 Single-vesicle analysis techniques developed in the coming years will be valuable for untangling this heterogenicity for each EV source, while more standardized and sophisticated isolation techniques will enable efficient separation of specific EV subsets. There are also concerns regarding the scalability of EV manufacturing, especially from primary cell sources. Nonetheless, feasible methods have been described for isolating decent amounts of EVs from MSCs, RBCs, and plasma.^9,323,334 Improving three-dimensional culturing, genetic engineering, and physical and chemical stimulation techniques for EV-producing cells is expected to lead to continued improvements in EV yields.^335,336 Another hurdle is the common presence of contaminants, such as lipoproteins, protein aggregates, and exomeres, in EV samples, which further increases the complexity and compromises the reproducibility of the EV isolation process. It is hence imperative that researchers confirm that their EV samples are of sufficient purity for their application.³³⁷

Future look

Nucleic acid-based therapeutics present a versatile option for anticancer therapy that enables effective modulation of gene expression at multiple levels. Given the genetic basis of cancer, the ability to modulate the content and composition of gene expression is vital for effective anticancer effects. This review provides insights into the current landscape of research and development of nucleic acid-based therapies, their application in cancer, and advancements in delivery strategies with clinical potential. We have covered nucleic acid-based therapeutics that can (1) introduce genes into cells to exert a range of anticancer effects via modulation of oncogenic signaling pathways, immunomodulation, and induction of apoptosis; (2) induce genome editing via CRIPSR-Cas9 systems to permanently modify, delete, or correct oncogenic genes; (3) control transcriptional regulation of cancer-associated genes using transcriptional regulators with or without CRISPR-assisted targeting; (4) modify mutated oncogenic RNAs to restore function or inhibit translation of oncogenic transcripts through exon skipping and RNA base editing; (5) gene silencing via siRNAs, ASOs, or miRNAs to directly degrade oncogenic RNA species or other endogenous RNAs involved in cancer progression; (6) activate the innate immune system to achieve pro-inflammatory antitumor responses; and (7) interfere with the functions of cancer-associated proteins, activate immunostimulatory receptors, or antagonize immune checkpoints via the use of aptamers.

While this review focuses on therapeutic strategies, many non-nucleic acid-based approaches also exist, often in combination with those discussed here. With respect to the selection of optimal therapeutics for anticancer applications, it is important to take into consideration multiple factors such as the level of regulation and the therapeutic target. For instance, gene silencing of oncogenic RNAs via ASOs and other RNAi strategies is superior in efficacy and versatility to most other methods available and has few detrimental effects. Conversely, genome editing and transcriptional/epigenetic regulation via nucleic acid-based strategies or CRISPR-directed strategies remain insufficiently advanced for effective therapeutic use in in vivo clinical applications. Indeed, other non-nucleic acid-based strategies such as small-molecule histone deacetylase inhibitors and transcriptional regulatory elements could present themselves as more potent therapeutic tools at the current time, given their enhanced efficacy and ease of delivery.³³⁸ However, it is important to note that these therapeutics tend to be non-specific and exert broad regulatory effects that can lead to unintended side effects. Advances in the delivery and efficiency of targeted approaches such as CRISPR-directed strategies negate this issue and could be used to replace these non-specific therapeutics in the future.

The development of these intricate nucleic acid-based strategies has also been accompanied by advances in the structure and delivery of these therapeutics. At present, delivery strategies are one of the major factors limiting the application of nucleic acid-based therapeutics. As mentioned before, nucleic acid-based drugs often suffer from cellular permeability, susceptibility to degradation, and lack of delivery specificity. There are a large number of nucleic acid-based therapeutics that show impressive therapeutic effects in controlled in vitro settings but show limited or no effects in vivo. The strategies outlined in this review sum up common chemical modifications and delivery strategies that are being used to tackle these issues. While they have shown satisfactory effects in the delivery of smaller molecules such as siRNAs and ASOs, the delivery of larger molecules remains challenging. Thus, the advancement of new and improved delivery strategies is an essential aspect that needs to be addressed in future research if more complex nucleic acid-based therapeutics are to be translated for clinical use.

Immunogenicity represents another critical challenge across all nucleic acid-based therapeutic approaches, as unintended immune activation can compromise safety, efficacy, and durability of treatment. Gene regulation approaches, including CRISPR-Cas9-mediated editing or transcriptional modulation, carry risks of immune recognition of both nucleic acid components (e.g., gRNAs) and exogenous proteins (e.g., Cas nucleases), potentially provoking antibody formation or cytotoxic T cell responses. In strategies aimed at immune activation, such as PRR agonists (RIG-I, cGAS-STING, TLR ligands) or immunomodulatory aptamers, excessive or systemic activation can lead to cytokine storms or autoimmunity, necessitating the careful tuning of potency. TAA vaccines should also be carefully designed to avoid expression in normal tissues to avoid off-target immune attacks. Approaches that target protein function or receptor modulation may also elicit unintended immune responses if the therapeutic oligonucleotide alters receptor conformation or exposes neoepitopes. The composition and chemical modifications of the nucleic acids themselves can either reduce or exacerbate immunogenicity depending on their length, structure, and context. However, excessive modification may impair target binding or intracellular activity, requiring careful design to balance immunogenicity and efficacy. Finally, delivery systems add a further layer of complexity. LNPs can trigger complement activation and cytokine release and induce anti-PEG antibodies in PEGylated formulations. Nucleic acid polyplexes may stimulate inflammasomes or complement-mediated clearance, while viral vectors are prone to neutralizing antibodies, preexisting immunity, and T cell-mediated elimination. EVs are generally less immunogenic but can vary depending on donor cell source, surface protein composition, and payload. The route of administration and dosing regimen further modulate immune recognition and systemic exposure. These considerations highlight that immunogenicity must be carefully evaluated and mitigated at every stage, from molecular design to delivery formulation, to balance optimal antitumor efficacy with safety.

Scalability and manufacturing complexity also present major barriers. Large-scale synthesis and purification of nucleic acids with precise chemical modifications require sophisticated infrastructure and remain costly compared to small molecules or protein-based therapeutics. Additionally, regulatory pathways for approval of these relatively new classes of drugs are still evolving, contributing uncertainty to clinical development. Addressing these hurdles is crucial to ensure that nucleic acid-based anticancer therapies are not only feasible but also accessible and affordable.

Despite these challenges, there are a large number of clinical trials in progress that make use of nucleic acid-based therapeutics for anticancer therapy. Table 3 showcases ongoing clinical trials involving nucleic acid-based drugs for cancer treatment. While some of these clinical trials have shown favorable tolerability and safety profiles with varying levels of therapeutic effects, it is worth noting that there are currently no nucleic acid-based drugs approved by the FDA for cancer treatment (Table 1). However, recent advances in nucleic acid-based therapeutics and delivery strategies show great potential and could lead to the successful approval of nucleic acid-based anticancer agents in the coming years.

Table 3.

Nucleic acid-based clinical trials in progress for anticancer therapy

Drugs	Molecular target(s)	Cancer type(s)	Stand-alone/in combination	Delivery system	Route(s) of administration	Phase	Clinical trial identifier(s)
Immunomodulators
SB11285	STING	advanced solid tumors	atezolizumab	naked	IV	1	NCT04096638
CRD3874-SI	STING	advanced solid tumors	single agent	naked	IV	1	NCT06021626
Dazostinag	STING	advanced solid tumors	single agent/with pembrolizumab	naked	IV	1	NCT04420884, NCT04879849
TAK-500	STING	advanced solid tumors	single agent/with pembrolizumab	anti-CCR2 antibody conjugation	IV	1	NCT05070247
KL340399	STING	advanced solid tumors	single agent	naked	ITU	1	NCT05549804
DEC-C	RIG-I	melanoma	nivolumab	naked	oral	1b/2	NCT05089370
Poly-IC12U	TLR-3	pancreatic cancer/melanoma/TNBC/ovarian cancer	single agent/with celecoxib, various immunotherapies	naked	IV/IP	1/2	NCT05494697, NCT04093323, NCT05927142, NCT05756166, NCT02432378
BDC-1001	TLR-7, -8	advanced HER2-expressing solid tumors	single agent/with nivolumab	anti-HER2 antibody conjugation	not provided	1	NCT04278144
BDB018	TLR-7, -8	advanced solid tumors	single agent/with pembrolizumab	naked	IV	1	NCT04840394
EIK1001	TLR-7, -8	NSCLC	single agent/with pembrolizumab, various chemotherapies	naked	not provided	2	NCT06246110
SD-101	TLR-9	metastatic uveal melanoma in the liver/pancreatic cancer	single agent/with various immunotherapies	naked	hepatic artery infusion/pancreatic retrograde venous infusion	1/1b	NCT04935229, NCT05607953
CpG-ODN	TLR-9	advanced solid tumors	CAR-T cells secreting scFv against OX40	naked	ITU	1	NCT04952272
Aptamers
AS1411	nucleolin	metastatic renal cell carcinoma	single agent	naked	IV	2	NCT00740441
AM003	personalized	advanced solid tumors	single agent	naked	ITU	1	NCT06258330
NOX-A12	CXCL12	glioblastoma/metastatic pancreatic cancer	radiotherapy, bevacizumab, pembrolizumab, various chemotherapies	naked	IV	1/2	NCT04121455, NCT04901741
siRNAs
NUDT21 siRNA	NUDT21	retinoblastoma	single agent	naked	intravitreal	1	NCT06424301
NBF-006	KRAS mutation	NSCLC, pancreatic cancer and colorectal cancer	single agent	naked	IV	1	NCT03819387
EphA2 siRNA	EphA2	advanced or recurrent solid tumors	single agent	DOPC	IV	1	NCT01591356
Cbl-b siRNA	Cbl-b	advanced solid tumors	single agent	PBMCs	IV	1b	NCT06172894
siG12D LODER	KRAS G12D	pancreatic cancer	various chemotherapies	polymeric matrix	ITU	2	NCT01676259
iExosomes	KRAS G12D	metastatic pancreatic cancer	single agent	MSC exosomes	IV	1	NCT03608631
ASOs
Danvatirsen	STAT3	pancreatic cancer, NSCLC, and colorectal cancer	single agent/with durvalumab	naked	IV	2	NCT02983578
BP1001-A	Grb2	advanced or recurrent solid tumors	single agent/with paclitaxel	liposomes	IV	1	NCT04196257
BP1001	Grb2	acute myeloid leukemia	ventoclax and decitabine	liposomes	IV	2	NCT02781883
WGI-0301	Akt-1	advanced solid tumors	single agent	LNP	not provided	1	NCT05267899
AZD8701	FOXP3	advanced solid tumors	single agent/with durvalumab	naked	IV	1	NCT04504669
IMV-001	IGF-1R	glioblastoma	radiation therapy and temozolomide	personalized whole-tumor-derived cells	implantation with biodiffusion chambers	2b	NCT04485949
BP1002	BCL-2	advanced lymphoid malignancies/acute myeloid leukemia	single agent/with decitabine	liposomes	IV	1/1b	NCT04072458, NCT05190471
OT-101	TGF-β2	NSCLC/pancreatic cancer	single agent/with pembrolizumab, various chemotherapies	naked	IV	1/2/3	NCT06579196, NCT06079346
TASO-001	TGF-β2	advanced or metastatic solid tumors	aldesleukin	naked	IV	1	NCT04862767

BCL-2, B cell lymphoma 2; CAR-T, chimeric antigen receptor T cell; Cbl-b, casitas B lymphoma-b; CCR2, C-C motif chemokine receptor 2; DOPC, 1,2-dioleoyl-sn-glycero-3-phosphocholine; EphA2, ephrin type-A receptor 2; FOXP3, forkhead box P3; Grb2, growth factor receptor bound protein 2; IGF-1R, type 1 insulin-like growth factor receptor; IP, intraperitoneal infusion; ITU, intratumoral injection; IV, intravenous infusion; LNP, lipid nanoparticle; NSCLC, non-small cell lung cancer; NUDT21, nudix hydrolase 21; PBMCs, peripheral blood mononuclear cells; STAT3, signal transducer and activator of transcription 3; TGF-β2, transforming growth factor β; TNBC, triple-negative breast cancer.

Acknowledgments

We acknowledge assistance from Ruo Lin (Sun Yat-sen University) and Clarissa Sastrawidjaya (National University of Singapore). The figures were created with BioRender.com and ChemDraw. B.P. is funded by the National Natural Science Foundation of China (grant no. 82403821), the Basic Research Fund - Shenzhen Natural Science Foundation (grant no. JCYJ20250604142701002), the Futian Healthcare Research Project (grant no. FTWS2025013), and a start-up grant from The Eighth Affiliated Hospital, Sun Yat-sen University (grant no. GCCRCYJ071). M.K.J. is funded by the A∗STAR Industry Alignment Fund - Pre-Positioning Programme (grant no. EVANTICA IAF-PP H23J2a0097).

Author contributions

B.P., conceptualization, writing – original draft, writing – review & editing, project administration, supervision, and funding acquisition; M.K.J., conceptualization, writing – original draft, writing – review & editing, and project administration; A.H.L., conceptualization, writing – original draft, and writing – review & editing; N.M.T., writing – original draft and writing – review & editing; M.T.N.L., conceptualization, writing – review & editing, and supervision.

Declaration of interests

The authors declare no competing interests.