The advances and challenges of mRNA-LNP application in fields of hematology and oncology therapeutics

Abstract

The rapidly global deployment of COVID-19 mRNA-LNP vaccines established the clinical feasibility of combining RNA medicine with LNP delivery systems, transforming a long-standing experimental concept into a scalable therapeutic modality, previously considered impossible. Currently, the value of mRNA-LNPs has been expanding from infectious disease vaccines to cancer immunotherapy, personalized neoantigen vaccination, in vivo immune-cell engineering, protein replacement therapy, mRNA-encoded biologics and genome editing, with particular relevance to hematology and oncology. However, in hematologic and oncologic settings, therapeutic success is constrained by biological barriers that less prominent in prophylactic vaccination. Systemically administered LNPs are strongly influenced by hepatic sequestration, protein corona formation and mononuclear-phagocyte clearance; solid tumors additionally present heterogeneous vasculature, variable enhanced permeability and retention, dense extracellular matrix, elevated interstitial pressure and limited intratumoral diffusion. Hematologic applications face a different set of constraints, including access to spleen and bone marrow, selective delivery to T cells, NK cells and hematopoietic stem and progenitor cells (HSPCs), and preservation of the normal marrow niche. In this Review, we therefore discuss the molecular and principles of formulation that underpin mRNA-LNP performance in hematology and oncology, including mRNA engineering, lipid nanoparticle architecture, intracellular trafficking and selective organ targeting(SORT) delivery strategies, oncology applications spanning neoantigen vaccines, tumor-microenvironment reprogramming, mRNA-encoded biologics, gene editing and combination therapy; and hematology applications including in vivo CAR-T, CAR-NK and HSPC engineering. We further consider how inefficient extrahepatic delivery, limited endosomal escape, context-dependent reactogenicity, repeat-dosing constraints and the manufacturing complexity of individualized products continue to limit broader clinical translation. Emerging strategies, including artificial-intelligence-guided RNA and lipid design, self-amplifying RNA, circular RNA and next-generation thermostable formulations, may help address these application limitations in the fields of hematology and oncology.

Introduction

The emergency authorization and global distribution of BNT162b2 and mRNA-1273 marked a watershed moment for biomedicine not merely for vaccinology but for the broader field of nucleic acid therapeutics [1, 2]. This demonstrated that mRNA-LNP technology could be developed, manufactured at scale, and deployed across diverse populations with remarkable efficacy and safety [3, 4]. The success solved the central paradox of mRNA therapy: how to deliver a large, negatively charged, highly degradable molecule into the cytoplasm of target cells where it can direct protein synthesis and proved three defining attributes of the mRNA-LNP platform, including rapid cell-free synthesis compressing antigen-design timelines from months to days, transient protein expression without genomic integration risk, and an ionizable lipid carrier system that enables efficient intracellular delivery. This insight, combined with decades of lipid chemistry research culminating in the development of ionizable lipid nanoparticles, transformed mRNA from a laboratory tool into a clinically viable therapeutic modality [5].

In the post-COVID era, mRNA-LNP technology has been expanding into the fields of hematology and oncology as a versatile and cutting-edge powerful tool rather than prevention of infectious diseases, with applications rapidly spanning in gene therapy, future potential in personalized medicine to produce therapeutic proteins, antigens, or gene-editing machinery [6]. This information-based medicine offers unprecedented flexibility, as the same manufacturing infrastructure can produce vaccines against different pathogens by simply changing the mRNA sequence, and the same LNP formulation can deliver mRNA encoding Cas9 nuclease, chimeric antigen receptors (CARs), or missing metabolic enzymes for diverse therapeutic applications in hematology and oncology. However, the biological requirements of cancer and blood-disease therapy differ substantially from those of prophylactic intramuscular vaccination. Solid tumors are spatially heterogeneous ecosystems in which abnormal vasculature, irregular perfusion, high interstitial pressure, stromal fibrosis and dense extracellular matrix can restrict nanoparticle extravasation and deep tissue penetration. The classical enhanced permeability and retention (EPR) paradigm remains relevant, but EPR is highly variable across tumor types and patients and is not sufficient to explain all nanoparticle entry and retention. Active endothelial transport and cell-mediated trafficking have therefore emerged as complementary mechanisms that should be considered when designing LNPs for cancer delivery. Hematologic targets pose a different challenge: circulating lymphocytes, splenic immune cells, bone-marrow-resident malignant cells and HSPCs occupy anatomically and functionally distinct compartments, and off-target delivery to hepatocytes or marrow-supporting cells may carry consequences that are not captured by vaccine reactogenicity studies [7–10].

At the same time, hematology and oncology are uniquely suited to the programmability of mRNA. A single delivery platform can encode tumor antigens, cytokines, co-stimulatory ligands, antibodies, bispecific engagers, chimeric antigen receptors or gene-editing enzymes, allowing the biological function of the drug to be changed primarily by altering the RNA sequence rather than rebuilding an entirely new manufacturing process [11–14]. Personalized neoantigen vaccines have generated durable T-cell responses and randomized evidence of clinical benefit in melanoma, whereas targeted mRNA-LNPs are beginning to reprogram immune cells directly in vivo and to deliver genome editors to hematopoietic stem cells [15–19]. These developments make mRNA-LNPs more than a delivery technology; they provide a framework for transient, controllable and potentially multiplexed therapeutic programming.

In this Review, we therefore reorganize the mRNA-LNP field around the biological and translational problems that matter most in hematology and oncology. We first discuss mRNA and LNP design principles in the context of tumor, immune-cell and HSPC delivery. We then examine systemic, solid-tumor and hematologic delivery barriers, including hepatic sequestration, tumor vascular and extracellular-matrix constraints, particle size, charge and elasticity, and access to bone marrow and lymphoid tissues. The therapeutic sections are divided into oncology and hematology, followed by engineering strategies for selective delivery, safety and repeat dosing, manufacturing and regulatory considerations, and the evolving clinical landscape. Figure 1 provides contextual background on the broader application space of mRNA-LNP technology; the discussion that follows is deliberately restricted to mechanisms, barriers and therapeutic strategies that are most relevant to hematology and oncology, while subsequent figures illustrate representative platform mechanisms and therapeutic workflows. Looking forward, we explore how AI-driven design, novel biodegradable lipids, and advanced imaging modalities are converging to create next-generation systems capable of cell-type-specific delivery. Finally, we summarize the major regulatory and translational challenges that will determine whether these emerging platforms can progress from experimental technologies to clinically deployable medicines.

Fig. 1.

Clinical application for mRNA-LNP, including vaccines against infectious disease, protein replacement, rare genetic disorder, tumors, and different administration routes

Design principles of mRNA-LNPs for hematologic and oncologic applications

The development of mRNA-LNP therapeutics represents the convergence of two distinct scientific trajectories, the recognition of mRNA as an informational molecule and the engineering of lipid-based delivery systems, as being illustrated in Fig. 2. This journey, spanning over sixty years, illustrates how fundamental discoveries can ultimately transform medicine. In previous decades, researchers recognized the therapeutic potential of exogenous mRNA to direct the production of any protein of interest. However, it faced seemingly insurmountable obstacles that naked mRNA is rapidly degraded by ubiquitous ribonucleases (RNases) in biological fluids, its large size and negative charge prevent passive diffusion across cell membranes, and the immune system has evolved sophisticated sensors to detect foreign RNA as a signal of viral infection. Encouragingly, Karikó and Weissman addressed the immunogenicity challenge [20]. They discovered that RNA synthesized in vitro triggered strong inflammatory responses through Toll-like receptors (TLRs), but that incorporating naturally occurring nucleoside modifications, particularly pseudouridine and N1-methylpseudouridine, abolished TLR recognition while preserving translation efficiency [20, 21]. This finding explained why endogenous mRNA, which contains modified nucleosides, is less immunogenic and provides the chemical basis for designing mRNA therapeutics that evade innate immune sensors.

Fig. 2.

Schematic representation of mRNA-LNP vaccine procedure. In vitro prepared RNA, in aqueous phase and components of lipid nanoparticles are mixed in microfluidic apparatus, including conventional mRNA, self-amplifying RNA and functional circular RNA. UTR, untranslational region; GoI, gene of interest; IRES, internal ribosome entry site

Molecular design of mRNA: stability, translation and innate sensing

mRNA synthesis is a fundamental step in the development of mRNA-LNP modality. Modern synthetic mRNA contains several key elements. The 5’ cap, typically cap1 or cap2 structure, protects mRNA from decapping enzymes and promotes translation initiation by binding eukaryotic translation initiation factors (ETIF) [22]. 5’ and 3’ untranslated regions (UTRs), usually optimized UTR sequences from highly expressed genes (e.g., human α-globin) enhance ribosome recruitment and mRNA stability [23–26]. The protein-coding region is codon-optimized for human expression and incorporates modified nucleosides to prevent recognition and activation of innate immunity which limits the translation efficiency. A poly(A) tail of 100–150 nucleotides at 3’ terminal enhances stability and translation, working synergistically with the 5’ cap to circularize the mRNA for efficient ribosome recycling [27, 28]. These variables should not be optimized in isolation. In oncology, stronger or more persistent antigen expression may improve priming but can also intensify inflammatory signaling; in cytokine therapy, excessive or prolonged expression may narrow the therapeutic window; and in CAR or genome-editing applications, transient expression can be an advantage because it limits long-lived receptor activity or nuclease exposure.

LNP composition and internal architecture

The standard LNP carrier comprises four functional components, each contributing distinct properties to the delivery system [29]. Cationic lipids, which form stable complexes with nucleic acids through electrostatic interactions, emerged in the late 1980s and enabled much higher mRNA loading [30]. However, their permanent positive charge caused significant toxicity in vivo, as they interacted nonspecifically with negatively charged cell surfaces and serum proteins. Ionizable lipids are key innovation in LNP delivery: they become positively charged at acidic pH, facilitating mRNA complexation and endosomal release, while remaining largely neutral at physiological pH to reduce nonspecific interaction and toxicity [31]. At low pH during formulation, these lipids are positively charged, facilitating electrostatic complexation with negatively charged mRNA. At physiological pH 7.4, they remain neutral, minimizing toxicity during circulation. Upon endosomal uptake, the acidic endosomal environment pH ~ 5–6 protonates the ionizable lipids, triggering membrane disruption and mRNA release into the cytoplasm. The first clinically validated application came not with mRNA but with siRNA. In 2018, the FDA approved Patisiran, Onpattro^®, an LNP-formulated siRNA for hereditary transthyretin amyloidosis (ATTRv), definitively proving that LNPs could enable safe and effective nucleic acid therapy [32]. This pH-dependent behavior is exemplified furthermore by FDA-approved lipids, ALC-0315 (BNT162b2) and SM-102 (mRNA-1273) [33, 34]. Helper phospholipids, typically distearoylphosphatidylcholine (DSPC) or dioleoylphosphatidylethanolamine (DOPE), stabilize the LNP structure and promote non-bilayer phase transitions that facilitate endosomal membrane fusion. DOPE, with its cone-shaped molecular geometry, particularly enhances the formation of inverted hexagonal phases that destabilize endosomal membranes. Cholesterol modulates membrane fluidity and packing density, contributing to particle stability and facilitating lipid mixing with cellular membranes. Its presence is essential for maintaining LNP integrity in biological environments. PEG-lipids provide a steric barrier that prevents particle aggregation during storage and opsonization in vivo. The lipid anchor length, typically C14-C18, controls PEG-lipid dissociation rate, which in turn influences cellular uptake, rapid dissociation exposes the particle surface and promote cellular interaction, while slower dissociation prolongs circulation [35, 36].

Early models depicted LNPs as simple core-shell structures with an aqueous core [37]. Advanced characterization techniques, including cryo-electron microscopy and small-angle X-ray scattering, have revealed a more complex internal architecture. Modern mRNA-LNPs are best described as nanostructured core particles containing inverted micellar structures, where ionizable lipids form electron-dense domains interspersed with aqueous pockets containing mRNA complexed with lipids [38]. This structure arises from the self-assembly process during manufacturing and critically influences stability and intracellular trafficking.

The journey of mRNA-LNPs

The mRNA-LNP system starts by injecting LNPs that contain mRNA molecules and ends in degradation by RNase and lysozymes [39]. The process involves several key steps, as illustrated in Fig. 3. Following contact with target cells, LNPs are internalized predominantly through endocytic pathway. Components such as apolipoprotein E (ApoE) facilitate the uptake process by binding to the surface of LNPs [40]. After internalization by cells, LNPs become enclosed in endosomes which function as small cellular vesicles. When the LNP merges with the endosomal membrane, it enables the release of mRNA into the cytoplasm. Upon release into the cytoplasm, mRNA must immediately engage the cellular translation machinery. The 5’ cap is recognized by eukaryotic initiation factor 4E (eIF4E), which recruits the ribosome [41]. The mRNA is translated exactly like an endogenous transcript, producing the encoded protein. The duration of expression depends on mRNA stability and dilution through cell division. For vaccines, the translated antigen is processed and presented on major histocompatibility complex(MHC)molecules, triggering adaptive immune responses [42, 43]. For protein replacement therapies, the secreted protein enters the circulation and exerts its intended physiological function in the right physiological metabolism process [44, 45]. For gene editing applications, the translated Cas9 nuclease must locate and cleave its genomic target before being degraded [46].

Fig. 3.

The immune response mechanisms mRNA vaccines. Briefly, after encapsulated mRNA being released from endosomes into cytoplasm, the protein of interest is produced and lysed to generate antigenic peptides to be presented by the major histocompatibility complex (MHC) on the cell surface, recognized by receptors on the ancestral B cells or T cells, then eliciting humoral and cellular immune response. RNA molecules are recognized by endosomal TLRs such as TLR3, TLR7, and TLR8, as well as cytosolic sensors including RIG-I and MDA5, leading to the induction of type I interferons and other proinflammatory cytokines that support antigen presentation and T cell priming. LNP, lipid nanoparticle; TLR, toll-like receptor; RIG-I, retinoic acid-inducible gene I; MDA5, Melanoma differentiation–associated gene 5; MAVS, Mitochondrial antiviral signaling protein; IRF, interferon regulatory factor; TCR, T cell receptor; BCR, B cell receptor; APC, antigen presentation cell; MHC, major histocompatibility complex; IFAR, interferon α receptor

The pathway from intramuscular injection (for vaccines) or intravenous infusion (for systemic therapies) to functional protein expression is remarkably complex. Understanding each step helps illuminates both current limitations and opportunities for improvement. Following administration, LNPs encounter a hostile environment. Serum proteins rapidly adsorb onto the LNP surface, forming a “protein corona” that can alter particle properties, trigger clearance by the mononuclear phagocyte system, or promote degradation. The PEG corona reduces but does not eliminate protein adsorption [47]. Nucleases in biological fluids would rapidly degrade naked mRNA, but encapsulation within the LNP provides nearly complete protection [48].

LNPs do not actively target specific cells in their basic formulation. Instead, they rely on passive processes. Upon encountering a cell, LNPs bind to the cell surface through nonspecific interactions-electrostatic forces, hydrophobic interactions, or adsorption of opsonizing proteins that then bind cellular receptors. The predominant uptake mechanism is receptor-mediated endocytosis, particularly through low-density lipoprotein (LDL) receptor family members that recognize ApoE bound to the LNP surface [49]. It explains the natural tropism of many LNPs for hepatocytes, which express high levels of LDL receptors. This cell-context dependence is a central reason why hematology- and oncology-directed LNP development increasingly relies on disease-relevant primary cells, organoids, tumor spheroids and in vivo screening rather than on a single reporter cell line.

Biological and biophysical barriers in hematology and oncology

Hepatic sequestration and systemic biodistribution

Intravenously administered LNPs frequently show strong hepatic tropism. Fenestrated liver vasculature, rapid uptake by Kupffer cells and other cells of the mononuclear phagocyte system, and adsorption of ApoE that facilitates LDL-receptor-mediated hepatocyte uptake all contribute to this bias [49, 50]. For liver-directed protein replacement or genome editing, this behavior can be therapeutically advantageous, but for leukemia, lymphoma, myeloma, HSPC engineering or most solid tumors, it represents a pharmacologic sink that reduces the fraction of dose available to the intended target.

Biodistribution is dynamically shaped after LNPs enter blood. Adsorbed proteins form a corona that can alter surface charge, receptor engagement and clearance; PEG-lipid shedding progressively exposes the particle surface; and ionizable-lipid chemistry can change not only lipid pharmacokinetics but also the distribution of translated protein [35, 51]. Thus, organ-level fluorescence or total lipid measurements alone may overestimate therapeutically productive delivery. A rigorous evaluation should distinguish nanoparticle accumulation, intact RNA delivery, cytosolic release and final protein expression, because each step can diverge substantially across organs and cell populations. This distinction becomes particularly important when the desired target is a small compartment such as marrow HSPCs or a rare tumor-infiltrating immune subset.

Solid-tumor vascular and extracellular-matrix barriers

Solid tumors impose sequential barriers before an mRNA-LNP reaches a malignant cell. Abnormal and spatially heterogeneous vasculature produces irregular blood flow and variable permeability. The EPR effect-increased nanoparticle extravasation through abnormal tumor vessels combined with impaired lymphatic drainage has historically guided cancer nanomedicine design, but its magnitude is inconsistent across tumor types, lesion sizes, metastatic sites and patients. Recent work indicates that active endothelial transport, macrophage-mediated trafficking and other transcellular mechanisms can contribute substantially to nanoparticle entry and intratumoral redistribution [52]. Accordingly, passive accumulation should not be assumed to be uniform, and clinical translation may require biomarkers or imaging approaches capable of identifying tumors in which a particular delivery mechanism is operative.

After extravasation, the tumor interstitium becomes a second major barrier. Dense collagen and hyaluronan networks, cancer-associated fibroblasts, solid stress and elevated interstitial fluid pressure can reduce convective transport and physically constrain diffusion. These effects are especially pronounced in desmoplastic tumors, where nanoparticle accumulation at the tumor periphery may coexist with poor penetration into viable tumor regions. The result is spatially heterogeneous exposure: perivascular cells can receive high LNP concentrations while hypoxic or deeply embedded tumor cells remain undertreated [52]. For mRNA therapeutics, incomplete penetration is particularly consequential because functional activity depends on delivery to enough target cells to generate antigen, cytokine, antibody, CAR or gene-editing activity at biologically meaningful levels.

The tumor microenvironment also changes over time and with therapy. Vascular permeability, stromal density, macrophage composition and interstitial mechanics may differ between primary and metastatic lesions and may be altered by chemotherapy, radiotherapy, antiangiogenic therapy or immune checkpoint blockade. Therefore, the optimal LNP for a pretreatment biopsy is not necessarily optimal after several treatment cycles. An important future direction is to treat delivery phenotype as a dynamic tumor characteristic rather than a fixed property, integrating imaging, tissue biomarkers and adaptive formulation strategies to identify when passive accumulation, active targeting or locoregional administration is most appropriate.

Physicochemical determinants of tumor penetration and cellular uptake

Particle size, surface charge, shape, PEG density, lipid pKa and mechanical properties collectively determine how LNPs move through tumor vasculature and extracellular matrix. Smaller particles generally diffuse more readily through dense interstitium, whereas larger particles may show greater retention but poorer deep penetration; the optimal size therefore depends on the desired balance between circulation, extravasation, diffusion and cellular uptake [52, 53]. Surface charge is similarly context dependent. Strong positive charge not only increases membrane association but also promotes nonspecific protein adsorption, clearance and toxicity, whereas near-neutral particles may circulate longer but require additional mechanisms for efficient tumor-cell entry. PEG prolongs colloidal stability and circulation, yet excessive or persistent PEG shielding impairs cellular uptake-the classic ‘PEG dilemma’ [35, 51, 54].

Elasticity and deformability are emerging as additional design variables. Nanoparticle mechanics can influence vascular margination, barrier transit, tumor penetration and the preferred endocytic pathway. Recent LNP studies show that softer particles can penetrate some tumor models rapidly and can display cell-type-specific uptake patterns, while stiffer formulations may exhibit slower permeation but longer retention in mechanically rigid tumors [55]. These findings argue against treating LNP elasticity as an incidental consequence of composition. Instead, mechanical properties may need to be matched to the stiffness and stromal architecture of the target tumor, much as ionizable-lipid pKa is matched to endosomal acidification.

Physicochemical optimization should therefore be evaluated across a complete delivery cascade rather than through a single potency assay. A formulation that maximizes uptake in two-dimensional cancer cells may fail in three-dimensional spheroids because of limited penetration, whereas a deeply penetrating particle may generate inadequate intracellular release. Rational cancer delivery requires simultaneous consideration of vascular access, interstitial transport, target-cell uptake, endosomal escape and off-target organ exposure. This multi-objective problem is one reason AI-guided formulation design and high-throughput in vivo barcoding are becoming increasingly valuable [56–58].

Hematologic tissue and cellular barriers

Hematologic delivery differs fundamentally from that to a solid tumor. Many targets are mobile or distributed across blood, spleen, lymph nodes and marrow rather than confined to a discrete mass. Resting T cells can be relatively resistant to transfection, NK cells require preservation of cytotoxic phenotype and viability, and HSPCs reside in specialized marrow niches that maintain quiescence, self-renewal and lineage potential [9, 59, 60]. The therapeutic objective is therefore not simply to increase total tissue accumulation but to achieve productive delivery to a particular immune or stem-cell subset while minimizing exposure of non-target hematopoietic and stromal cells.

The spleen is an attractive compartment for immune-cell programming because it contains abundant antigen-presenting cells and lymphocytes and can be accessed by altering lipid composition or surface chemistry. Spleen-tropic lipidoids and SORT-like strategies have been used to redirect mRNA away from the liver and toward splenic immune populations [61–65]. Bone marrow is more challenging. Sinusoidal endothelium and the complex niche permit access to circulating nanoparticles, but productive delivery to HSPCs competes with uptake by macrophages, endothelial cells, mature leukocytes and malignant cells. Bone-marrow-homing and CD38-targeted LNPs demonstrate that this barrier is tractable, yet they also highlight a safety issue: a delivery system potent enough to edit diseased HSCs or myeloma cells may also expose normal stem cells and marrow-supporting cells to RNA and ionizable lipids [8].

Biodistribution tracking and advanced imaging

Because successful mRNA-LNP therapy requires completion of multiple sequential steps, biodistribution should be measured at more than one biological level. Particle tracking can quantify where the carrier accumulates; RNA-sensitive methods can determine where intact or released transcripts are present; and reporter or therapeutic-protein measurements reveal where productive translation ultimately occurs. A recent framework explicitly separates particle-, transcript- and protein-level tracking, emphasizing that these readouts can diverge substantially [11]. This distinction is particularly important for oncology, where an LNP may accumulate in tumor-associated macrophages without reaching malignant cells, and for hematology, where organ-level spleen or marrow signal does not identify which leukocyte or progenitor subset has been transfected.

Current approaches include fluorescent and radiolabeled lipids, in situ hybridization, quantitative PCR, RNA barcoding, single-cell sequencing and reporter proteins. Multiplexed barcodes can compare dozens to hundreds of formulations in vivo, while single-cell methods can identify cell-type-specific tropism within an organ [66]. Imaging modalities can complement these molecular assays by visualizing whole-body trafficking and lesion-level heterogeneity. The goal is not merely descriptive pharmacokinetics: quantitative tracking can provide a mechanistic bridge between LNP design, delivered dose, cell-specific expression and therapeutic response, and may eventually support patient selection or adaptive dosing in cancer nanomedicine.

mRNA-LNP therapeutics in oncology

The flexibility, scalability, and safety of mRNA-LNP technology render it suitable for cancer therapy in patients. However, traditional mRNA-LNP modalities cause hepatic retention in recipient leading to inefficient translation of delivered mRNA and hepatic toxicity, particularly limiting application of systemic administration. Currently, although efforts are underway to enhance the efficacy and safety of oncology therapy, there is still no tumor-specific LNPs that facilitate the successful application for cancer therapy. Herein, we present representative developments in Table 1 and discuss the advancements of mRNA-LNP in the application of oncology. We analyze each approach in the context and highlight the obstacles that need to be addressed to achieve these applications in this field.

Table 1.

Representative mRNA-LNP applications and evidence in oncology

Examples	mRNA-LNP	Phase/Tumor type	Advantages	Refs
Personalized neoantigen vaccination; V940 (mRNA-4157)	Patient-specific neoantigen mRNA-LNP, with pembrolizumab.	Randomized phase IIb/Melanoma	The combination improved recurrence-free outcomes versus pembrolizumab alone in the phase IIb study.	[16], [74]
Local OX40L expression; mRNA-2416	OX40L mRNA-LNP,	Phase I/II/Solid tumors and ovarian cancer	Biopsies supported local OX40L expression and inflammatory pharmacodynamics; prespecified efficacy endpoints were not met.	[83]
Multiplexed local immune stimulation; mRNA-2752	OX40L, IL-23 and IL-36γ mRNAs in an LNP, intratumoral; alone or with durvalumab.	Phase I (NCT03739931)/Solid tumors	The study reported tumor-microenvironment immune changes and exploratory antitumor activity.	[84]
Intratumoral cytokine programs	IL-12/IL-27 or IL-21/IL-7/4-1BBL mRNA-LNP, local injection.	Preclinical/Melanoma model	Local immune activation and antitumor effects in selected models.	[88], [89]
mRNA-encoded bispecific engager	GPC3-directed T-cell engager mRNA in organ-directed LNP.	Preclinical/Carcinoma model	Antitumor activity after in vivo expression of the encoded engager.	[94]
Tumor genome editing	Cas9 mRNA and guide RNA in targeted LNP.	Preclinical/Glioblastoma models.	Gene disruption and tumor-growth effects in orthotopic models.	[96]
Combined cytotoxic and immune therapy	IL-12 mRNA plus docetaxel prodrug in lung-tropic LNP.	Preclinical/Lung-metastasis model.	Combination antitumor activity in the model.	[102]

Shared-antigen and personalized neoantigen vaccines

Cancer vaccination is among the most clinically advanced non-infectious applications of RNA therapeutics. Early RNA delivery studies established that RNA can target dendritic cells and exploit innate antiviral pathways to generate strong antitumor T-cell responses [67]. Shared-antigen vaccines aim to target recurrent tumor-associated antigens, whereas personalized neoantigen vaccines encode patient-specific mutations identified by tumor and matched-normal sequencing. The latter strategy is particularly well suited to mRNA because a multiepitope product can be synthesized rapidly without changing the fundamental manufacturing platform [68–72]. Fig. 4 illustrates an end-to-end personalized neoantigen workflow.

Fig. 4.

Workflow for personalized neoantigen-based liver cancer vaccine development. Tumor tissue and peripheral blood are collected for genomic analysis from clinical patients, followed by whole-exome and RNA sequencing to identify somatic mutations. Candidate neoantigens are predicted, prioritized, and selected using computational pipelines integrating parameters such as mutation calling, HLA typing, and peptide–MHC binding predictions. Validated targets are then manufactured into vaccine products and formulated with appropriate delivery systems or adjuvants. After being tested in specific HLA-integrated humanized mice, personalized neoantigen-based liver cancer vaccines are administered alone or alongside other therapies such as ICIs or chemotherapies. Ex vivo validation of neoantigen-specific immune responses is often done through flow cytometric analysis or IFN-γ ELISpot using patient peripheral blood mononuclear cells. HCC, Hepatocellular carcinoma; HLA, human leukocyte antigen

Clinical evidence has strengthened the rationale for this approach. Individualized RNA vaccines can generate durable polyclonal T-cell responses, and the randomized phase 2b KEYNOTE-942 study demonstrated improved recurrence-free outcomes when mRNA-4157 (V940) encoding 34 patient-specific neoantigens was combined with pembrolizumab in resected melanoma [15, 16, 73, 74]. Otherwise, BioNTech announced to terminate the BNT122 CRC Phase 2 trial of autogene cevumeran with unmodified mRNA encoding 20 patient-specific neoantigens, as adjuvant monotherapy in ctDNA-positive, resected Stage II high-risk or Stage III colorectal cancer. These results shift the field from proof of immunogenicity toward questions of clinical optimization: how many neoantigens should be encoded, which HLA-binding and clonality features best predict effective targets, how should vaccines be integrated with checkpoint blockade, and how can manufacturing timelines be shortened for patients with rapidly progressive disease? The answers will depend not only on antigen prediction but also on efficient delivery to antigen-presenting cells and on the immune state of the tumor microenvironment.

For hematologic malignancies, vaccination may be particularly attractive in settings of low disease burden or measurable residual disease, where a relatively small malignant population must be controlled by durable immune memory. However, immune escape, HLA loss, antigen heterogeneity and treatment-induced immunosuppression remain important barriers. Rather than treating the vaccine as an isolated product, current development increasingly views mRNA vaccination as one component of combination immunotherapy, paired with checkpoint blockade, cytokine modulation or cell-based approaches [75–78].

Reprogramming the tumor microenvironment

Many solid tumors are immunologically ‘cold’, poor vascular organization, lack of lymphatic drainage and high extracellular matrix density. These characters lead to limited effector-cell infiltration, dysfunctional antigen presentation and suppressive myeloid and stromal networks, thus limiting the diffusion of LNPs from injection sites to tumors. Intratumoral injection with mRNA-LNPs provide a means to reprogram this environment by producing potent immunomodulators locally rather than administering them systemically. Intratumoral or tumor-directed LNPs encoding IL-12, IL-27, IL-23, IL-36γ, OX40L, IL-7, IL-21, 4-1BBL and other combinations have produced local immune activation and systemic antitumor responses in preclinical models or clinical trials in patients combining with PD-L1 monoclonal antibody [79–87]. The central advantage is pharmacologic compartmentalization: cytokines with unacceptable systemic toxicity can be generated transiently within or near the tumor, increasing local exposure while limiting circulating concentrations.

The same principle can be extended to innate immune activation and other cytokine programs. Incorporating or co-delivering TLR agonists can enhance dendritic-cell maturation and antigen presentation. Targeted LNP delivery of IL-15 super-agonist mRNA has shown that spatial control can increase intratumoral cytokine exposure while limiting systemic toxicity, and co-delivery of IL-33 mRNA with a STING agonist has been used to enhance dendritic-cell maturation, type I interferon signaling, cytotoxic T-cell activity and checkpoint-blockade responsiveness in a poorly immunogenic breast-cancer model [88–90]. These approaches support a broader cold-to-hot tumor strategy in which LNPs coordinate innate activation, antigen presentation and effector-cell recruitment rather than merely delivering a single inflammatory mediator. mRNA can also encode co-stimulatory ligands or antibody-like molecules that locally block inhibitory pathways. The therapeutic objective is therefore not simply to ‘add inflammation’ but to coordinate antigen release, dendritic-cell activation, T-cell priming, effector-cell recruitment and reversal of checkpoint-mediated dysfunction.

Combination with immune checkpoint blockade is especially compelling because the two modalities address different bottlenecks. mRNA vaccination or cytokine expression can expand and recruit tumor-reactive lymphocytes, while PD-1/PD-L1 blockade can restore function in exhausted cells [75–78]. Chemotherapy and radiotherapy may further contribute through immunogenic cell death and antigen release. The emerging design principle is temporal and spatial orchestration: mRNA-LNPs can be engineered to generate immune signals in the tumor at the stage when they are most likely to synergize with checkpoint blockade, cytotoxic therapy or adoptive-cell treatment. This strategy supports the activation of CD8⁺ and T_H1 cells, recruitment of natural killer T cells and activation of CD103⁺ and CD8⁺, dendritic cells, which are necessary for antigen cross-presentation and activation of CD8⁺ T cell mediated cellular immune response.

mRNA-encoded antibodies, bispecific engagers and checkpoint modulators

mRNA-LNPs can turn transfected cells into temporary bioreactors for therapeutic proteins, including monoclonal antibodies and bispecific engagers. This strategy avoids repeated manufacture and infusion of recombinant protein and can produce sustained local or systemic exposure after a single administration, depending on tissue tropism and the secretion properties of the encoded molecule [12, 91–93]. For oncology, the concept is attractive when the therapeutic protein has a short serum half-life, is difficult to manufacture, or would benefit from preferential expression within the tumor or liver.

Bispecific T-cell engagers illustrate the potential of this approach. mRNA can encode molecules that connect a tumor antigen, such as CLDN6 expressed on surface of multiple tumor cells and GPC3 expressed on surface of hepatoma cells, to CD3 or combine tumor targeting with co-stimulatory signaling, and organ-selective LNPs can further control where the biologic is produced [93, 94]. In principle, the same framework can encode checkpoint inhibitors, cytokine-antibody fusions or multi-specific constructs with programmable half-lives. However, the pharmacology differs from conventional biologics: dose is determined by the number and identity of transfected cells, RNA translation kinetics and protein secretion, making biodistribution and expression control critical determinants of safety.

Gene editing of oncogenic dependencies and therapy-resistance pathways

Genome editing extends mRNA-LNP therapy from transient protein expression to durable modification of tumor or host cells. The use of LNPs in delivering CRISPR-Cas9 components for gene editing presents another exciting avenue, potentially enabling the correction of genetic disorders at their source. Delivering Cas nuclease or base-editor mRNA together with guide RNA avoids prolonged nuclease expression associated with viral vectors and can reduce the duration of off-target editing risk [14, 46, 95, 96]. In cancer, this approach can be used to disrupt oncogenic dependencies, alter immune checkpoints, sensitize tumors to therapy or reprogram immune cells. The major challenge is not editor potency alone but delivery to a sufficient fraction of malignant cells within a heterogeneous lesion.

Glioblastoma provides a stringent example. Targeted LNPs carrying CRISPR components have achieved therapeutically meaningful editing in orthotopic models, including disruption of PLK1 and editing strategies directed at miR-10b-associated pathways, with effects on tumor growth and survival [97, 98]. These studies demonstrate that local or regionally targeted LNP delivery can overcome some barriers of the central nervous system and provide functional gene editing within an aggressive solid tumor. They also illustrate why CNS material should remain in this Review only when it directly addresses an oncologic problem such as glioblastoma rather than serving as a general discussion of neurodegenerative delivery.

The translational threshold for tumor editing is high. Mosaic editing may leave resistant clones, while nonspecific delivery to normal proliferating cells can create unacceptable toxicity. Future strategies will likely combine tumor-selective LNP chemistry, ligand targeting, microRNA-based detargeting and editors engineered for higher specificity [99–101]. In many cancers, transient mRNA-based modulation of a pathway may ultimately prove more practical than permanent editing; therefore, the field should evaluate editing only where durable genomic change offers a clear biological advantage over reversible RNA expression. Looking forward, the application of LNP-CRISPR in the field of solid tumors represents a promising breakthrough. The continuous optimization and improvement of CRISPR/Cas9 demonstrate its tremendous research and application potential in the future of tumor therapy.

Co-delivery with conventional anticancer agents and multidrug resistance

A distinctive opportunity for LNPs is the coordinated delivery of mechanistically different payloads. In oncology, this includes pairing mRNA with hydrophobic chemotherapeutics or prodrugs so that cytotoxic therapy and immune modulation reach the same organ or lesion with matched pharmacokinetics. Chemotherapy can debulk tumor, induce immunogenic cell death and release antigens, while an mRNA cargo can promote dendritic-cell activation, cytokine expression or reversal of resistance pathways. A recent chemotherapeutic-derived LNP strategy combined immunogenic cell death with CD40L/Flt3L mRNA-mediated dendritic-cell activation, illustrating how the carrier itself can be incorporated into a chemo-immunotherapy design [102].

Direct co-encapsulation is also feasible. Lung-tropic LNPs have been engineered to co-deliver IL-12 mRNA and a docetaxel prodrug, producing synergistic antitumor activity in a preclinical lung-metastasis model while limiting systemic toxicity [103]. Such approaches are relevant to multidrug resistance because synchronized delivery can couple a cytotoxic agent with an RNA that changes the tumor microenvironment, restores immune recognition or suppresses a resistance determinant. The design challenge is substantial: hydrophobic small molecules and large anionic RNAs have different loading chemistries and release requirements, and a formulation optimized for one cargo can compromise the other. Nevertheless, combination LNPs provide a conceptual route to move beyond single-payload nanomedicine toward programmable, multi-mechanism cancer therapy. Whereas these strategies primarily address the spatial and immunologic complexity of solid tumors, hematologic application impose a different challenge.

mRNA-LNP therapeutics in hematology

Although clinical researches are currently exploring for hematologic treatment with mRNA-LNP technology and the landscape of hematology is changing rapidly, most of application in this field is still under preclinical settings. Here, we also list the representative mRNA-LNP application in hematology and discuss the developments in this field. We provide insights from preclinical and clinical studies that leading to this powerful platform being considered the next frontier in hematology.

Table 2.

Representative mRNA-LNP applications and evidence in hematology

Application	LNP and RNA strategy	Phase	Advantages	Refs
In vivo CD19 CAR-T	Antibody-displayed LNP delivers CAR mRNA to circulating T cells.	Phase I;	Initial administration and early clinical observations were reported.	[104]
In vivo CD22 CAR-T	Anti-CD8 nanobody-functionalized LNP delivers CD22 CAR mRNA.	Preclinical	Transient CAR expression and inhibition of tumor growth after in vivo T-cell programming.	[18]
CAR-NK for B-cell malignancies and AML	CAR mRNA-LNP transfects expanded NK cells ex vivo, followed by cell transfer.	Preclinical	CD19/BCMA or CLL-1 CAR-NK cells showed antigen-specific activity in these models.	[111]
In vivo HSPC editing for hemoglobinopathies	Optimized LNP carries base-editor mRNA and guide RNA to HSPCs.	Preclinical	HBG1/2 regulatory editing and improved globin balance in derived erythroid cells.	[113], [114]
Marrow-homing editing platform	Compositionally redirected LNP delivers mRNA and editing components to marrow cells.	Preclinical	Delivery to multiple marrow cell types; CRISPR/base-editing readouts in relevant models.	[8], [117]
Coagulation-factor replacement	Systemic LNP delivery of factor VIII or factor IX mRNA.	Preclinical	Functional factor expression and hemostatic activity were reported in animal studies.	[123–125]

In vivo CAR-T engineering

CAR-T therapy has transformed the treatment of B-cell malignancies, but current ex vivo manufacturing requires leukapheresis, viral or nonviral engineering, prolonged cell expansion, release testing and reinfusion. Manufacturing delays, cost, batch failure and the need for specialized centers limit access [104]. mRNA-LNPs have shown impressive early success for the treatment of cancer disease and offer a different paradigm: deliver a CAR transcript directly to T cells in vivo and allow transient receptor expression within the patient. Because mRNA does not integrate into the genome, CAR expression is self-limited, which may reduce the duration of on-target/off-tumor toxicity, the risk of severe toxicities such as cytokine release syndrome and provides a built-in mechanism for dose control.

The central technical requirement is selective T-cell delivery. Antibody- and ligand-functionalized LNPs, antibody-capture systems and chemically redirected formulations can enrich delivery to CD8⁺ or other T-cell subsets [105–108]. Recent studies have generated CAR-T cells in vivo and demonstrated antitumor activity, including CD8-targeted mRNA-LNPs that reprogram CD22 CAR-T cells for hematologic malignancies [18, 104]. Multiplexed RNA delivery can also combine CAR expression with gene editing or additional signaling modules, creating a path toward transient, multi-gene immune-cell engineering [109]. Fig. 5 illustrates the concept of in vivo CAR-T generation.

Fig. 5.

Schematic of the procedure for in vivo preparing CAR-T by using mRNA-LNP reprogramming technology, resulting in tumor targeting and killing. CAR-T, chimeric antigen receptor

Transient CAR expression is both an advantage and a limitation. It improves controllability but may require repeated dosing to sustain antitumor pressure, particularly in diseases with high burden or persistent antigen. Repeated LNP administration introduces questions of anti-PEG immunity, innate activation, changing biodistribution and cumulative lipid exposure [110]. The optimal clinical role may therefore differ from that of conventional permanent CAR-T therapy: mRNA-LNP CAR-T could serve as a rapidly deployable bridge, a controllable strategy for antigens associated with higher toxicity, or a platform for repeated immune-cell programming in diseases in which transient activity is sufficient.

Toward in vivo CAR-NK engineering

NK cells are attractive effectors for transient CAR engineering because they possess intrinsic cytotoxicity, can recognize stressed cells without HLA-restricted antigen presentation and may offer a different safety profile from conventional CAR-T cells. CAR-NK generation still relies predominantly on ex vivo manipulation, but mRNA-LNP transfection offers a transient, non-integrating way to introduce CAR specificity and provides a technological bridge toward eventual direct in vivo NK-cell programming.

Proof-of-concept studies have shown that mRNA-LNP transfection can generate functional CD19- and BCMA-directed CAR-NK cells with high CAR expression and antigen-specific killing; the engineered cells subsequently suppress tumor growth after adoptive transfer in vivo [111]. The hematologic relevance is immediate: CD19 is shared across multiple B-cell malignancies, whereas BCMA is a validated target in multiple myeloma. More recent work in acute myeloid leukemia similarly used mRNA-LNPs to generate CLL-1 CAR-NK cells ex vivo, followed by repeated CAR-NK infusions in xenograft models; this study also identified the NKG2A/HLA-E axis as an adaptive resistance mechanism that can be addressed by checkpoint modulation [112]. These studies establish the compatibility of LNP-mediated mRNA transfer with NK-cell engineering, but they should not be conflated with direct systemic generation of CAR-NK cells inside the patient. Achieving selective in vivo delivery of CAR mRNA to endogenous NK cells remains a key next step.

Several barriers therefore separate current mRNA-LNP CAR-NK manufacturing from true in vivo CAR-NK generation. A systemic formulation would need to target endogenous NK cells efficiently without excessive uptake by liver or myeloid cells, preserve NK-cell viability and phenotype, avoid unwanted transfection of other leukocytes, and achieve sufficient CAR expression without provoking systemic inflammation. Transient CAR expression may also necessitate repeated administration. A major opportunity is combinatorial programming: CAR mRNA could eventually be paired with cytokines, checkpoint modulators or homing receptors, provided that additional cargos do not compromise NK-cell function. Framing CAR-NK in this way highlights both the promise of the approach and the current evidence gap that must be closed before direct in vivo reprogramming can become a clinical strategy.

HSPC targeting and in vivo genome editing

HSPC engineering is one of the most consequential hematologic applications of mRNA-LNPs. mRNA-LNP delivery systems that might enable efficient and selective in vivo genetic manipulation of HSPCs by passing the precondition protocols. Current autologous gene therapy requires stem-cell mobilization or harvest, ex vivo editing, myeloablative or reduced-intensity conditioning and reinfusion. These steps are clinically burdensome and can limit access, particularly for patients with fragile marrow or inherited bone-marrow-failure syndromes. Targeted in vivo delivery of editor mRNA could, in principle, convert a complex cell-manufacturing procedure into a directly administered medicine [113–115].

Hematopoietic stem and progenitor cells

Antibody-targeted and compositionally optimized LNPs have delivered mRNA to hematopoietic stem cells (HSCs ) and HSPCs in vivo, and recent work has extended this concept to human CD34⁺ cells and non-human primates [116]. In vivo base editing of HBG1/2 regulatory sequences has been used to reactivate fetal hemoglobin in human HSC models, providing a potential route toward treatment of sickle cell disease and transfusion-dependent β-thalassemia [19]. These approaches avoid permanent vector integration and limit editor exposure to the period during which mRNA is translated, a desirable feature for genome editing in long-lived stem cells.

The key biological requirement is preservation of stem-cell function. Editing efficiency is insufficient if treated HSPCs lose quiescence, self-renewal, multilineage potential or long-term repopulating capacity. Moreover, apparent marrow delivery can mask broad exposure of mature leukocytes, macrophages and stromal cells. For this reason, future HSPC studies should pair editing readouts with serial transplantation, clonal tracking, lineage output and long-term marrow safety. The clinical value of in vivo HSPC editing will depend as much on durable hematopoietic integrity as on the percentage of alleles edited at early time points.

Bone-marrow-directed therapy for hematologic malignancies

Bone marrow is both a therapeutic target and a protected microenvironment for many hematologic malignancies. Myeloma cells, leukemic stem cells and other malignant populations interact with stromal, endothelial and immune cells that can support survival and drug resistance. LNPs that home to marrow therefore create opportunities not only for direct tumor-cell killing but also for remodeling the niche that sustains malignant cells. Coating the LNP with ligands and antibodies, or covalent modification of the LNPs with application-specific peptide epitopes could enhance preferencial tissue uptaking and cell-specific targeting delivery [91, 117].

CD38-targeted LNPs have delivered therapeutic RNA to the marrow in multiple myeloma models, demonstrating that a disease-associated surface antigen can be used to redirect nanoparticle uptake toward malignant plasma cells.⁸ Bone-marrow-homing LNPs have also achieved mRNA delivery to a broad range of marrow cell types, including healthy and diseased HSCs, leukemic stem cells, lymphocytes, macrophages and leukemia cells, and have enabled CRISPR or base-editing activity in relevant models as illustrated in Fig. 6 [9, 118, 119]. This breadth is therapeutically powerful but underscores the need for cell-level specificity: a formulation that enters many marrow populations may be appropriate for niche-wide immune remodeling but less suitable for a highly potent nuclease or pro-apoptotic cargo.

Fig. 6.

mRNA-based genome modification. mRNA encoding Cas9 protein and sgRNA targeting gene of interest were encapsulated in LNP and administered intravenously. After entry of targeting cell, Cas9 mRNA was translated and Cas9-sgRNA ribonucleotide protein complex was formed, transported into nucleus. In the nucleus, the Cas9 follows the sgRNA to the target DNA sequence and make a cut cross both strands of the DNA. Then cell recognizes and damaged DNA and fixes the cut under the error-prone mechanism. sgRNA, small guide RNA

Future marrow-directed therapy will likely combine three levels of selectivity: preferential access to the marrow, binding to a disease-associated target and context-dependent RNA expression. For myeloma, lymphoma with marrow involvement or Acute myeloid leukemia(AML), such designs could deliver pro-apoptotic proteins, immune agonists, CARs or genome editors directly into the disease compartment while reducing hepatic exposure. The marrow therefore provides a model of why organ targeting alone is not enough; clinically useful LNPs will need cell-type and payload-specific control within the target organ.

Protein replacement in hematologic disorders

mRNA-LNP protein replacement is especially relevant to hematologic disorders caused by deficiency of a secreted or circulating protein. Rather than repeatedly infusing recombinant protein, mRNA can transiently turn host cells into a source of the missing factor. Hemophilia provides a clear example: LNP-delivered factor VIII or IX mRNA has produced functional coagulation-factor expression in preclinical models, and galactosylated formulations have been explored to enhance liver-directed factor VIII production [120–122]. mRNA can also encode MG1113, a humanized monoclonal antibody products targeting tissue factor pathway inhibitor (TFPI) that modify coagulation, providing additional strategies for hemophilia management, enabling sustained in vivo antibody production as a factor-unrelated therapeutic strategy [123].

The principal limitation is duration. Unlike genome editing, protein-replacement mRNA generally requires repeat dosing, and the therapeutic window is shaped by expression kinetics, immune responses to the encoded protein and cumulative exposure to the LNP. This makes biodegradable lipids, low-reactogenicity formulations and predictable dose-expression relationships particularly important [110, 124–126]. In a hematology-focused framework, broader examples of unrelated metabolic disorders are less informative than the specific question of whether repeated mRNA administration can safely maintain clinically meaningful levels of blood proteins over months or years.

Optimization strategies and future directions for hematology and oncology

Optimizing mRNA-LNPs for hematologic and oncologic applications requires matching particle design to the target tissue, cell population, route of administration, and anticipated dosing schedule. Here, we summarize the conventional and current lipid nanoparticle design strategies for hematologic and oncologic applications as shown in Table 3, highlighting the advantages and challenges of each approach in terms of LNP design strategy and selective cellular expression, discussing the potential improvements as active therapeutic opportunities in hematology and oncology.

Table 3.

Lipid nanoparticle design strategies for hematologic and oncologic mRNA delivery

LNP strategy	Design principle	Intended target or use	Potential advantages	Key limitations	Refs
Conventional ionizable four-lipid LNP	Ionizable lipid, helper phospholipid, cholesterol and PEG-lipid	Antigen expression after intramuscular vaccination	Clinically established mRNA delivery architecture; scalable formulation	Inefficient extrahepatic or cell-selective therapeutic delivery. IV dosing can favor hepatic uptake.	[29, 33, 34]
Compositionally redirected / SORT-like LNP	Additional charged or ionizable lipid and altered lipid ratios shift functional organ distribution.	Spleen, lung or other extrahepatic compartments; immune-cell access.	May reduce dependence on intrinsic liver tropism without attaching a receptor ligand.	Organ accumulation is not target-cell expression. Protein corona, route, dose and species can alter tropism.	[127, 129]
Ligand-functionalized LNP	Antibody, nanobody, peptide, aptamer displayed or recruited at the particle surface.	CD8 T cells for in vivo CAR programming; other receptor-defined cells.	Enhanced uptake by cells expressing a selected surface receptor.	Receptor specificity, ligand density, binding-site effects, immunogenicity and off-target translation require testing.	[18, 95, 131]
Bone-marrow-homing LNP	Lipid chemistry or surface properties favor marrow access after systemic administration.	HSPCs, leukemic cells and marrow immune populations.	Access to a compartment poorly served by conventional liver-tropic LNPs.	Marrow homing may expose many cell types; editor or cytotoxic payloads require cell-level specificity and long-term safety.	[9, 19]
Stimuli-responsive LNP system or locoregional delivery	Release or local activation coupled to pH, redox state, proteases or an external trigger such as ultrasound.	Accessible tumors and locoregional immunotherapy.	May confine activity in space or time and reduce systemic exposure.	Trigger penetration and uniformity across heterogeneous lesions remain uncertain; evidence is mainly preclinical.	[135, 83]
Biodegradable or lower-reactogenicity LNP	Labile ionizable lipids, alternative PEG shielding or zwitterionic components.	Repeat-dose CAR programming, protein replacement and vaccination.	May reduce lipid persistence or inflammatory signaling while retaining delivery.	Improved acute tolerability in a model does not establish repeat-dose safety, potency or consistent biodistribution in patients.	[136, 137 138]

Organ- and cell-selective lipid chemistry

One of the most important advances after the first generation of liver-tropic LNPs has been the realization that organ distribution can be altered by lipid composition itself. SORT formulations use additional charged or ionizable components to redirect mRNA expression toward liver, lung or spleen, demonstrating that biodistribution is an engineerable property rather than a fixed consequence of systemic administration [61, 127, 128]. Related approaches substitute helper lipids, modify steric lipids or alter ionizable-lipid structure to favor spleen, lymph node or other extrahepatic sites [129, 130].

For hematology and oncology, the value of organ-selective chemistry lies in matching biodistribution to disease biology. Spleen-tropic formulations can increase access to antigen-presenting cells and lymphocytes; lung-tropic LNPs may be useful for pulmonary metastases; and liver-detargeted formulations can increase the fraction of dose available to extrahepatic targets. The next step is to combine organ selectivity with cell selectivity so that an LNP enriched in marrow, spleen or tumor is also preferentially internalized by the malignant or immune population of interest.

Surface functionalization and receptor-directed targeting

Surface functionalization provides a direct route to cell-specific delivery. Antibodies, nanobodies, peptides and aptamers can be displayed on or recruited to the LNP surface to recognize receptors on T cells, HSPCs, myeloma cells, tumor endothelium or malignant cells [131–134]. Antibody-capture systems and modular conjugation strategies are particularly attractive because they separate the core LNP formulation from the targeting ligand, potentially enabling a single platform to be redirected toward multiple cell types [9, 19].

Targeting ligands nevertheless introduce trade-offs. Increased surface complexity can alter particle size, PEG organization and protein-corona formation; high-affinity binding near vessels may reduce deep tumor penetration through a ‘binding-site barrier’; and antibody-conjugated products can complicate manufacturing, stability and comparability. Consequently, the most effective ligand is not necessarily the one with the highest affinity. The desired receptor density, internalization pathway, target-cell abundance and anatomical accessibility should all be considered when optimizing ligand valency and binding strength.

Stimuli-responsive and locoregional delivery

Stimuli-responsive systems add spatial or temporal control by coupling cargo release or LNP activation to features of the tumor microenvironment or to an external trigger. pH, redox state, protease activity, ultrasound, light or other stimuli can be used to expose targeting groups, remove PEG shielding or destabilize the carrier preferentially at the disease site. Ultrasound-activatable lipid platforms, for example, can confine innate immune stimulation and mRNA vaccination to a defined region [135]. Such approaches may be especially useful for solid tumors in which systemic delivery is limited by vascular or stromal barriers.

Locoregional administration offers a simpler form of spatial control. Intratumoral injection can generate high local concentrations of cytokine or immunomodulatory mRNA while limiting systemic exposure [83]. The trade-off is that it is most applicable to accessible lesions and may not address disseminated disease. In practice, future oncology regimens may use a spectrum of delivery routes-systemic LNPs for metastatic or hematologic targets and locoregional LNPs for immune priming or tumor-microenvironment remodeling.

Biodegradable and lower-reactogenicity lipids

Repeat dosing makes lipid clearance and inflammatory potential central design constraints. Biodegradable ionizable lipids incorporate ester, ketal or other labile bonds that allow rapid decomposition after cargo delivery, reducing tissue persistence and cumulative exposure [136]. Some degradable formulations also improve endosomal escape or alter organ tropism, illustrating that safety and potency do not need to be opposing objectives [137]. PEG antibodies was a risky factor of mRNA-LNP based vaccines and therapeutics for biosafety concerns. PEG alternatives and zwitterionic components are being explored to reduce anti-PEG responses and inflammatory signaling while maintaining colloidal stability [138].

These strategies are particularly important for hematology. Transient CAR-T or CAR-NK programming may require repeated infusions, protein replacement may require chronic dosing, and HSPC-directed therapy must minimize damage to a long-lived stem-cell compartment. A lipid that is acceptable for one or two vaccine doses may be suboptimal for these settings. Safety-by-design should therefore include not only acute cytokine release but also metabolic clearance, marrow exposure, immune memory to the carrier and cumulative organ toxicity.

AI-guided RNA design

As illustrated in Fig. 7, the development of mRNA-LNP therapeutics begins with optimization of the RNA sequence for stability, translation efficiency and controlled innate immune sensing, including a promoter sequence, an open reading frame encoding the gene of interest, 5′ and 3′ untranslated regions (UTRs), and a poly(A) tail [27, 139, 140]. A central technical bottleneck is an incomplete, quantitative understanding of how element sequences, RNA structure and modifications determine stability, translation and innate immune sensing. Artificial intelligence (AI), including machine- learning approaches and shared bioinformatics resources could accelerate optimization of sequence design and lipid nanoparticles accelerating the mRNA-LNP vaccine development [141–151]. By facilitating AI technology, the Pfizer research team recently developed a platform to synthesize circular DNA as the starting template, achieving isothermal enzymatic DNA amplification through phi29-driven rolling circle amplification (RCA), and then directly performing IVT after linearization with TypeIIS restriction enzymes, all in a single reaction tube, named Unified Sequential Template Amplification and Transcription (USTAT) [152]. The entire process from circular template to mRNA takes only two days, completely eliminating the bacterial fermentation step and opening a new chapter in modular, rapid mRNA manufacturing. AI programs not only recognize and optimize internal ribosome entry site (IRES), but also can be employed to design a large number of IRES candidates de novo for mRNA translation in specific organs or cells expanding the molecular toolkit for scaling up next-generation biomedical discovery and RNA-based therapeutics [153–155]. For example, Zhang and colleagues utilized deep generative models to design mRNA (linear or circular) sequences with enhanced translational capacity and stability [156]. These advancements highlight the vast potential of deep generative artificial intelligence for mRNA therapeutics.

Fig. 7.

Schematic illustration of AI driven mRNA design with enhanced stability and translation, LNP screen and generation of optimized mRNA-LNP. AI, artificial intelligence

AI-guided lipid discovery

Furthermore, AI and machine learning methods can be adapted to screen and optimize LNP formulation [157–159]. The chemical space of potential ionizable lipids is vast-estimated at > 10^{^60} possible structures. Traditional synthesis and screening explore only a tiny fraction. Machine learning approaches are beginning to transform discovery by predicting lipid performance from structure to target delivery [160–163]. The AI-guided ionizable lipid engineering (AGILE) platform integrates deep learning with combinatorial chemistry to prioritize candidate lipid before experimental validation, thereby accelerating the design-build-test cycle [164]. AI and machine learning can also support personalized neoantigen selection to improve tumor targeting and accelerate the design and effective personalized cancer immunotherapy [165, 166].

Training sets comprising thousands of synthesized lipids with measured delivery efficiency enable models that identify promising scaffolds, predict optimal headgroup-tail combinations, and suggest novel chemistries [163]. Early reports suggest AI-guided discovery can identify lipids with enhanced potency and novel tropism profiles, accelerating the design-build-test cycle from months to weeks [167, 168]. The study leverages AI and virtual screening to facilitate the rational design of ionizable lipids by predicting two key properties of LNPs, apparent pKa and mRNA delivery efficiency, and nearly 20 million ionizable lipids were evaluated through two iterations of AI-driven generation and screening.

Machine learning models are being trained on large datasets of LNP compositions and their resulting delivery outcomes. These models can predict the performance of novel formulations, suggest promising lipid structures, and guide experimental efforts. For example, the LiON (Lipid Optimization Network) deep learning platform can design and prioritize ionizable lipids with desired pKa, biodegradability, and predicted in vivo performance. Using this model, the investigators identified FO-32, which achieved state-of-the-art nebulized mRNA delivery in the mouse lung [169]. Su et al. reported that head and tail arrangements of LNPS are key determinants of delivery efficiency and organ specificity. In this study, an AI-guided candidates, notably lipid P1, adopted stable three-tail cone-shaped conformations that promoted IgM protein corona formation and enabled spleen-targeted mRNA delivery. In preclinical models, P1-based mRNA vaccines triggered strong antibody and T-cell responses, leading to marked tumor suppression. These results highlight the pivotal role of lipid spatial conformation and the potential of AI-driven strategies to optimize lipid nanoparticles for organ-specific mRNA delivery and translation, supporting the development of personalized mRNA nanomedicines [170, 171]. This integrated approach offers a framework for designing advanced LNPs and has the potential to unlock the full potential of mRNA therapeutics.

Self-amplifying RNA and circular RNA formats

Self-amplifying RNA is similar to linear RNA with additional sequence coding RNA-dependent RNA ploymerase complex which amplifies RNA by self-amplifying RNA as template [172]. These constructs incorporate replicase genes from alphaviruses enabling the RNA intracellular amplification, achieving equivalent protein expression at 10-100-fold lower doses than conventional mRNA. This dose-sparing could reduce LNP requirements and improve tolerability [173, 174]. The trade-off is larger RNA size (9–12 kb vs. 2–4 kb), potentially making manufacturing more challenging and increasing immunogenicity risk [27]. The first saRNA-LNP vaccine for COVID-19 was approved in Japan in 2024, and saRNA-LNP vaccines for other pathogens are in clinical development, with applications extending to cancer and protein replacement as the technology matures [175–179]. Furthermore, saRNA-LNP could be adapted for immunotherapeutic platforms for breast cancer to elicit antigen-specific immune response to tumor-associated antigens and neoantigens or immunostimulator for glioblastoma [180–183].

Circular RNA represents another structural innovation with significant potential which substantially improves the stability of RNA. Unlike linear mRNA, circular RNAs are characterized by lacking free ends, making it resistant to exonuclease-mediated degradation and potentially extending protein expression duration without the immunogenic concerns associated with modified nucleotides [184]. The circular configuration eliminates the need for 5’ caps and poly-A tails, simplifying manufacturing while improving stability [185–187]. Unlike linear mRNA, translation initiation of circular RNA needs an internal ribosome entry site or incorporation of N⁶-methyladenosine (m⁶A) modification in their UTRs [188–192]. Additionally, circular RNAs induce weaker undesirable immune response without activating innate immunity [193]. Early clinical candidates are exploring circular RNA for vaccines and therapeutic protein expression, though efficient circularization methods and purification remain active areas of process development [194–197].

An LNP-encapsulated circular RNA (circRNA) vaccine encoding tumor antigens produces durable antigen expression, robust antitumor immunity and improved therapeutic efficacy, demonstrating its potential for cancer immunotherapy in a mouse model [198]. Results showed that small circRNA vaccines encoding tumor-associated antigens, neoantigens and oncoviral or viral antigens elicited substantial CD⁸⁺ and CD4⁺ T cell responses in young adult mice and in immunosenescent aged mice [199]. It demonstrates that antigen-encoding small circRNAs loaded in lipid nanoparticles elicit potent and durable T cell responses for robust tumor immunotherapy after subcutaneous injection in mice, particularly when combined with immune checkpoint inhibition. Intranasal vaccination with circular RNA encapsulated in lipid nanoparticles can induce localized mucosal immune responses for lung cancer therapy, representing a novel and promising approach for developing RNA vaccines targeting mucosal malignancies [200]. Circular RNA also can be employed to enhance mRNA–LNP technology in CAR-T cell therapy, gene editing and autoimmune disease [201–204].

Thermostable formulation, lyophilization and manufacturing considerations

mRNA-LNP thermostability remains a distribution bottleneck, particularly for low-resource settings. Current formulations require − 20 °C to -70 °C storage, though lyophilization and alternative lipid compositions or lipid-free are improving stability [205, 206]. However, research results showed that freezing and lyophilization could result in loss of internal structural order and reduced activity of mRNA-LNP [207]. Lyophilized mRNA-LNPs retaining activity after room-temperature storage would transform global access, particularly for cancer vaccines requiring distributed manufacturing [208]. Studies have shown that the use of cryoprotectants protects nanoparticles against loss of function and degradation in all storage conditions. Moreover, it is shown that the addition of sucrose enables all nanoparticles to remain stable and maintain their efficacy for up to a month when stored at -80 °C, regardless of cargo or type of nanoparticle [209].

Safety, toxicity and repeat-dosing challenges

mRNA-LNPs demonstrate high encapsulation efficiency (greater than 90%), which enables substantial delivery of therapeutic RNA to its target cells. A proven scalable manufacturing process exists for mRNA-LNPs which enables large-scale production appropriate for broad application. The short lifespan of mRNA, which breaks down within days to weeks after delivery prevents insertional mutagenesis and unregulated protein production. The mRNA-LNP platform avoids triggering anti-vector immune responses which enables multiple vaccine administrations. The biodegradable and biocompatible nature of LNPs helps minimize potential long-term toxic effects [210].

However, the main accumulation site of mRNA-LNPs is the liver which restricts their ability to target other tissues. Administration of PEGylated lipids has the potential to cause hypersensitivity reactions which must be managed with care. The need for cryogenic storage and transportation of mRNA-LNPs arises from their susceptibility to degradation at room temperature which presents logistical challenges. The existing versions of these formulations face significant challenges when attempting to target tissues beyond the liver yet research continues to work on solving this problem [211].

In the development of vaccines, the role of mRNA is relatively clear, to transiently express an antigen, activate the immune system, and generate a protective response. However, in therapeutic scenarios, mRNA faces entirely different requirements [212]. It may need to silence an abnormal gene, supplement a missing protein, reprogram a cellular state, or even directly generate a functional therapeutic system in the body. Therefore, mRNA is no longer just a tool to trigger immunity, but must become a controllable, reproducible, and verifiable therapeutic modality [213]. This implies that the future of cell therapy may not necessarily rely on an extremely complex centralized production chain. It could become a model of in vivo delivery, in vivo programming, and in vivo generation of effector cells. What should be considered is not just the cost, but the entire treatment framework, including patient accessibility, hospital deployment capability, business models, and global expansion paths [214–216]. The innovation path all will be rewritten. mRNA-LNP has already become a real clinical tool in some areas. But the next stage issue is whether mRNA truly cross the therapeutic threshold and become a sustainable, reproducible, regulated, and payable drug system? From this perspective, the most important questions are how mRNA can be more precisely controlled in the body, how mRNA support repeated dosing, how mRNA define new clinical endpoints and evidence paths, how mRNA transform from a bespoke miracle into a widely accessible therapeutic solution, how can mRNA turn platform capability into disease solutions [217]. This is true at the beginning of mRNA from a platform to clinical reality.

mRNA therapy truly opens not just mRNA-LNP vaccines, but programmable treatments within the body. Nowadays, when many people talk about mRNA therapy, the first thing that comes to mind is personalized cancer vaccines. This is of course important, and it indeed represents a key path from mRNA being a vaccine to becoming a treatment. For example, Moderna and Merck are jointly advancing a personalized cancer vaccine project, intismeran autogene (mRNA-4157/V940) [218–220]. As of publicly available information up to 2026, the Phase III trial for adjuvant treatment of melanoma has completed enrollment, as well as the Phase III project for lung cancer is underway, and the company has disclosed longer-term follow-up data. However, if mRNA therapy is only understood as a cancer vaccine, its true potential is still underestimated. The deeper value of mRNA lies in its potential to become an in vivo programmable therapeutic architecture. It can be used to supplement missing proteins, to silence harmful expressions, to temporally express therapeutic molecules, to directly generate cellular therapy functional modules in the body, and to construct dynamic therapeutic circuits from sensing to execution. In other words, the future of mRNA is not just another new drug category, but it could transform drugs from a fixed molecule into a designable, updatable, and iterative functional system [221].

The future clinical paradigm of mRNA-LNP therapy is very likely to rely more on mechanism-driven biomarkers, more flexible small-sample designs, model-based dose optimization, dynamic endpoints and surrogate endpoints, integration of real-world data and long-term follow-up [66]. This is also why once mRNA-LNP truly enters therapy, it is inevitably linked to a reconstruction of clinical trials. In the therapeutic adaptation, we face the challenges, including whether efficacy is stable, whether toxicity is controllable, whether repeated dosing is feasible, whether CMC can be consistently defined, whether clinical endpoints are accepted, whether there is a reference pathway in regulation, whether payers are willing to cover it, whether doctors know whom to treat, when to treat, and how to monitor.

Finally, we propose design rules and analytical standards to address outstanding knowledge gaps in expression durability and age-related blood-brain barrier alterations. Together, these insights chart a path for engineering mRNA-LNP therapeutics capable of meeting the rising burden of neurodegenerative disease in an ageing global population. Manufacturing innovations focus on automated platforms, reducing production timelines from nine weeks to under four weeks for personalized vaccines, while costs remain challenging at over $ 100,000 per patient [222].

In vaccine design, short-term expression is often an advantage, but in therapy, short-term expression may mean the need for repeated dosing. Once it enters a repeated dosing scenario, the complexity of the entire system immediately increases. Will innate immune activation reduce translation efficiency? Will adaptive immunity change the performance of subsequent doses? Will the delivery system, especially LNPs, pose new risks with repeated use? Do different tissues, different diseases, and different patient subgroups require completely different dosing and frequency strategies༟.

Delivery and translation efficiency

The 1–4% endosomal escape efficiency represents the fundamental biophysical limitation of current technology. Even highly potent LNPs lose the majority of their cargo to lysosomal degradation. Improving escape efficiency through better understanding of mechanism and rational design of lipids that more effectively destabilize endosomal membranes could transform potency and reduce required doses [223]. Future research should focus on key areas include developing novel LNP components and structures for precise organ and cell targeting, reducing non-specific accumulation and toxicity; incorporating biodegradable materials and new lipid alternatives (e.g., PEG alternatives) to lower LNP immunogenicity and toxicity, optimizing administration routes and dosing strategies to enhance mRNA stability and delivery efficiency, and leveraging AI and high-throughput screening to expedite LNP optimization and personalized therapy development [224].

Furthermore, cell-specific expression is expected by designing cell-specific RNA elements to incorporate regulatory recognition sites. For example, specific microRNAs were screened to prevent mRNA expression in non-target cells [225]. Marks and colleagues incorporated synthetic microRNA target sites mediated hepatocyte-silenced TAA mRNA vaccine enhanced immune response and reduced tumor burden [226]. It is reported synthetic super-enhancers also can be employed to precisely control transgene expression in specific cells [227, 228]. Emerging needle-free administration technologies is alternative platform to enhance the efficiency of RNA-LNP medicine, offering significant advantages for large-scale vaccination campaigns by addressing challenges associated with conventional needle-based injections, including needle-stick injuries, injection pain, and vaccine hesitancy [229].

Reactogenicity and toxicity

While COVID-19 vaccines established acceptable safety at scale, mRNA-LNPs are not immunologically inert [230]. Actually, reactogenicity still represents the manifestation of inflammatory response to vaccination. Balancing therapeutic immune activation against treatment-related reactogenicity remains a central design challenge [231–234]. Optimization of lipid composition, including the reduction of cationic lipids, as well as additional immunomodulatory strategies, have been reported to suppress inflammatory responses without compromising immunogenicity [235, 236]. Recently, Moderna released phase III clinical trial results of mRNA1010 which is comparable to a licensed standard influenza vaccine [237]. The solicited adverse reactions were shown more frequent with it. The LNP itself activates innate immunity through multiple pathways: ionizable lipids can trigger Toll-like receptor responses, and the nanoparticle structure promotes complement activation and cytokine release [238]. An in vitro whole blood assay was developed to assess the impact of mRNA-LNPs on immune cell function, cytokine release, and complement activation. Results showed that mRNA-LNPs significantly increased CD69 expression on T cells and natural killer cells, and CD80/CD86 on myeloid subsets, in a dose-dependent fashion. Furthermore, mRNA-LNPs elicited a robust release of pro-inflammatory cytokines, including tumor necrosis factor-α, interleukin (IL)-1β, monocyte chemoattractant protein-1, IL-6, and IP-10, indicating a potent immune response and a need for tailoring LNPs for specific target populations [239] By contrast, Zhao et al. reported membrane-destabilizing zwitterionic lipid-containing nanoparticles show reduced inflammation and neutrophil infiltration at the injection site due to their zwitterionic property, being compatible with existing targeted nanoparticle formulations, further improving mRNA delivery [138]. mRNA-LNPs provide excellent lung-targeting performance. However, they can induce clotting by activating the intrinsic coagulation pathway and platelets [240, 241]. Li et al. reported that intranasal prime-boost mRNA-LNP vaccination elicits potent T cell response for lung cancer therapy mitigating the systemic adverse effects commonly associated with intravenous mRNA-LNP vaccination [200].

Off-target effects of mRNA-LNP vaccines or therapeutics lead to undesired translation [242]. Zou and colleagues reported a system of multi-objective LNP engineering with artificial intelligence (MOLEA) that integrates high-dimensional lipid representations, cell-type-resolved transfection data and multitask optimization to design ionizable lipids with both high potency and biological selectivity. In this study, MOLEA learns structure-function relationships across diverse cellular contexts to identify lipids that preferentially deliver mRNA to target tissue while minimizing hepatocyte transfection [58]. Kim et al. found that co-delivery of antigen-encoding mRNA and toll-like receptor (TLR) agonists significantly enhanced antigen presentation in DC and tumor cell types, leading to robust CD8⁺ T cell responses [243]. This dual-targeting strategy elicited potent antitumor effects without systemic toxicity, demonstrating its potential in overcoming immune escape and improving mRNA vaccine efficacy. Parrett and colleagues inserted one to five miR-122 binding sites into the 5’ or 3’ untranslated regions (UTRs) of gene of interest to reduce the off-target effect in liver [99]. This research reinforces the utility of miR-specific binding sites to aid the future development of mRNA-LNP therapies for hepatic and non-hepatic tissues.

For prophylactic vaccines, this adjuvant effect may be beneficial. For chronic replacement therapy, it may pose substantial challenges [244]. Repeat dosing can elicit adaptive immune responses against both the payload (if neoantigenic) and the delivery system, potentially reducing efficacy over time. However, recent results of Kahwaji and colleagues indicated that T cells do not exhibit adverse immune activation in response to unmodified mRNA. These results provides a strong justification for reconsidering the routine use of modified mRNA in CAR T cell engineering, with the potential to substantially reduce manufacturing costs and expand access to CAR T cell therapies, gene editing and transient immunomodulation [245]. In a study by Hu and colleagues, they designed scarless circular mRNA-based CAR-T cell therapy which elicits superior antitumor efficacy in a mouse model [246]. This design avoids extraneous junction sequences, streamlining manufacturability and potentially reducing innate immune sensing and significantly increased both the magnitude and duration of anti-CD19 CAR and anti-GPRC5D CAR expression in primary human T cells. These findings suggest that innate immunity may play crucial role in CAR T cell therapies.

Manufacturing complexity for personalized products

Personalized cancer vaccines require manufacturing timelines compatible with patient survival, ideally 4–6 weeks from biopsy to infusion. This demands integrated workflows, including sequencing, neoantigen prediction, mRNA synthesis, LNP formulation, quality control, and release testing [72]. Scalable microfluidic platforms (e.g., SCALAR) address formulation throughput, enabling production from screening-scale (0.07 L/hr) to clinical-scale (17 L/hr). However, quality control for personalized products remains challenging because each batch represents a unique drug product requiring release testing without established comparability to prior batches [247]. Importantly, regulatory frameworks adapted from the pandemic experience will need to evolve to accommodate individualized therapeutic products. COVID-19 vaccines have proven that mRNA can be quickly designed, produced, and scaled up. But a drug is not defined this way. A real drug must confront another set of questions: Does it need to be administered repeatedly? Which tissues must it reach? When should it be expressed, for how long, and how strongly? How is its efficacy verified clinically? How does it navigate regulatory, payment, and commercialization systems to become a therapy that can be used continually? Therefore, mRNA’s next step is beginning to enter the most challenging areas of the drug world.

Regulatory pathways and CMC consideration

In 2024, the FDA released comprehensive guidance for therapeutic cancer vaccines providing sponsors with detailed recommendations for Investigational New Drug applications, addressing critical considerations for both early-phase and late-phase clinical trials specific to cancer vaccine development [248]. It is anticipated that the first commercial mRNA cancer vaccine could receive regulatory approval by 2029, marking a significant achievement in oncology and paving the way for broader applications of mRNA technology in cancer treatment. International regulatory harmonization frameworks are also being established to facilitate global access to approved mRNA cancer vaccines [249].

Traditional chemistry, manufacturing, and controls requirements assume fixed product specifications; personalized products with batch-to-batch variation challenge the regulatory workflow. Potency assays that predict clinical activity for each patient’s unique neoantigen set are technically demanding. Experience with COVID-19 vaccines accelerated regulatory science for mRNA platforms, but personalized therapeutics raise distinct questions requiring novel approaches to lot release, stability testing, and post-marketing surveillance [250, 251].

Instead, it is necessary to enter long-term management, chronic disease control, risk-benefit balance, and home self-medication systems. From this perspective, mRNA-LNP is no longer ‘like a drug,’ but in certain areas has already become a drug. But this is also where the problem becomes sharp. mRNA-LNP can become a drug, but that does not mean that all mRNA-LNP platforms can automatically become drugs. However, both regulators and industry are increasingly clear that things are not that simple [252].

In discussions on the regulation of RNA drugs, the EMA explicitly pointed out that platform knowledge can be transferred, but the ‘platform’ itself must be strictly defined; if all mRNA or all RNA drugs are considered the same platform, this concept is too broad. For products with different indications, different routes of administration, different delivery systems, or different mechanisms of action, the extrapolation of platform experience is still very limited. This statement actually exposes the core reality of RNA therapy. Vaccines have proven that mRNA-LNP can be industrialized, but treatments require mRNA-LNP to be druggable [253]. These are not the same thing. Industrialization emphasizes speed, manufacturing, and scaling, whereas drugability emphasizes dosage, tissue distribution, repeated administration, safety margins, efficacy endpoints, and long-term management.

In other words, what the mRNA platform truly needs to overcome is not whether a candidate can be made, but whether this candidate can be valid in clinical and regulatory terms. Existing regulatory frameworks may therefore be insufficient for the full range of emerging mRNA-LNP therapeutics. It is imperative to establish regulatory standard and cooperation to accelerate the marketing and accessibility of mRNA-LNP therapies [254].

Clinical and translational landscape in hematology and oncology

The number of mRNA-LNP vaccine clinical trials has experienced explosive growth after the pandemic, presenting obvious pandemic-driven characteristics and geographical differences. There are significant differences in registration characteristics and trial design among China, the United States, and Europe. Indications have rapidly expanded from infectious diseases to multiple fields such as tumors, autoimmune diseases, and metabolic diseases, indicating that mRNA technology is transforming from an infectious disease prevention tool into a platform technology with broad therapeutic potential [255].

However, the most relevant signal for hematology and oncology is the transition from vaccines toward therapeutic programming. Personalized cancer vaccines now have randomized clinical evidence, and multiple trials are testing combinations with immune checkpoint blockade or other therapies [222]. At the same time, in vivo immune-cell engineering, mRNA-encoded biologics and HSPC-directed editing are moving from conceptual demonstrations toward translational development. Within that landscape, hematology and oncology increasingly account for the applications in which cell specificity, repeat dosing and disease-compartment targeting are most critical.

A useful way to evaluate this pipeline is by therapeutic function rather than by platform novelty. Cancer vaccines are clinically most mature; encoded cytokines and bispecific molecules remain largely preclinical or early translational; in vivo CAR-T and CAR-NK are emerging as potential alternatives to centralized cell manufacturing; and HSPC-directed editing aims to replace ex vivo stem-cell manipulation with direct in vivo treatment. Each category faces a distinct translational bottleneck. Vaccines require reliable antigen selection and immune priming, cytokine therapies require spatial control, CAR strategies require immune-cell specificity and repeatability, and HSPC editing requires durable efficacy without compromising long-term hematopoiesis.

The field should therefore resist using the number of mRNA trials as a surrogate for therapeutic maturity. More informative milestones include reproducible extrahepatic delivery in large animals, target-cell transfection at clinically tolerable doses, evidence that repeat dosing does not alter biodistribution or provoke prohibitive immunity, and demonstration that transient expression produces durable clinical benefit when the biology requires it. Hematology and oncology are likely to remain the proving ground for these questions because both fields combine high unmet need with well-defined cellular targets and increasingly precise molecular biomarkers.

Future potential and conclusion

mRNA-LNP technology has progressed from a pandemic vaccine platform to a programmable therapeutic system with particular promise in hematology and oncology. Its core advantages-rapid sequence-level programmability, transient non-integrating expression, compatibility with multiplexed payloads and the ability to engineer tissue or cell tropism-align closely with the needs of personalized cancer vaccination, immune-cell reprogramming, hematopoietic gene editing and protein replacement [11, 13]. Yet these same applications expose limitations that were less apparent in intramuscular vaccination: hepatic sequestration, heterogeneous solid-tumor access, poor penetration through dense stroma, inefficient endosomal escape, restricted delivery to resting lymphocytes and HSPCs, repeat-dose immunity and the need to protect the normal marrow niche.

The next phase of development should therefore be barrier driven rather than application driven. In solid tumors, formulation design should be matched to vascular transport, stromal architecture, particle size, charge and elasticity rather than relying on EPR as a universal mechanism [55]. In hematology, organ-level targeting must be refined to cell-level specificity within spleen and marrow. In vivo CAR-T and CAR-NK strategies will require reliable repeated immune-cell programming, whereas HSPC editing will require long-term evidence that stem-cell self-renewal and lineage output are preserved [111, 112]. Across indications, biodistribution tracking at the particle, transcript and protein levels should become an integral part of translational development rather than an exploratory afterthought.

Engineering advances are beginning to address these needs. Organ-selective lipid chemistry, modular targeting ligands, stimuli-responsive systems, biodegradable lipids, RNA-based detargeting and AI-guided optimization can be combined to improve specificity and reduce systemic exposure [135]. The most successful mRNA-LNP medicines are unlikely to arise from a single universally optimal nanoparticle. Instead, the field is moving toward a family of indication-specific platforms in which RNA design, LNP chemistry, route of administration and target-cell biology are co-optimized.

For hematology and oncology, this shift changes the defining question. The proof that mRNA can function as a medicine has already been established. The remaining challenge is to deliver the right RNA to the right malignant, immune or stem cell, at the right time and dose, while preserving the tissues that support normal immunity and hematopoiesis. Achieving that level of precision would position mRNA-LNPs not merely as a broad delivery technology, but as a controllable therapeutic architecture for personalized cancer therapy, in vivo cell engineering and potentially curative treatment of blood disorders.

Author contributions

J.L. and Y.Y. conceived the manuscript and share senior authorship. Y.W. and J.L contributed equally to the literature review and writing. Y.C contributed to rephrasing and proofreading. Y.W. and Y.G. designed the figures. All authors discussed and edited the manuscript and approved the final version.

Funding

This work received no specific funding.

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics consent to participate

Not applicable.

Competing interests

The authors declare no competing interests.