Methods for Delivering Genes to Specific Body Tissues

AAV is a naturally occurring member of the Dependoparvovirus genus.124 It is a replication‐deficient virus, relying on the presence of a helper virus (adenovirus) to infect humans, non‐human primates, and other mammalian species. AAV is a non‐enveloped virus comprised of a capsid protein shell that houses the genetic material within. The capsid is composed of three integral viral proteins (VPs) that assemble into an icosahedral symmetry structure. The AAV capsid proteins, VP1, VP2, and VP3, are present in a 1:1:10 ratio within the capsid, respectively.124, 125 VP1 is the largest among the capsid proteins, roughly 87 kDa in size, and plays a key role in the ability of AAV to enter the host cell. VP2 is approximately 73 kDa where prior work has shown this capsid subunit is dispensable and an ideal target for modifications with non‐native peptide (e.g., targeting moieties).126 VP3 is the smallest of the three capsid proteins, approximately 62 kDa, but is the most abundant of the three and is critical for viral capsid assembly. As such, VP3 is necessary to maintain the overall integrity of the capsid. These three capsid proteins assemble 60 protein subunits to form a highly ordered icosahedral capsid, forming a capsid that is roughly 20 nanometers in diameter.27, 127 More recent literature also suggests AAV capsid proteins are involved in episomal gene expression suggesting these elements facilitate both viral entry in target cells and persistence of gene expression.128, 129 As such, the functional properties of AAV capsids dictate many key features necessary for their therapeutic use and have been the focal point of AAV engineering for tissue‐specific gene delivery.

AAV contains a single‐stranded DNA genome of ~4.7 kilobases.26, 130 There are three main components to the AAV genome: the inverted terminal repeats (ITRs), the replication (rep), and capsid (cap) genes. ITRs are sequences that are crucial for the replication and packaging of the virus and are located at both ends of the genome itself.131 Once the virus transduces the host cell, the ITRs serve as a signal for intracellular production of AAV (e.g., rescue, replication, and packaging) and for most recombinant AAV are derived from AAV serotype 2. The rep gene encodes for the proteins utilized in AAV viral genome replication and integration into the host cell genome. The rep gene encodes proteins essential for viral propagation and host transduction. The cap gene encodes for the capsid proteins (VP1, VP2, and VP3). An alternate open reading frame in the cap gene encodes for the assembly‐activating protein (AAP) which aids in the structural integrity of the capsid during assembly.132 Recombinant forms of AAV used in gene therapy applications are generated by replacing rep and cap genes with a transgene of interest.

To initiate binding to the cell, many AAV interact with several cell‐surface associated glycans and proteoglycans terminated with galactose, sialic acid, and heparan sulfate.133, 134 Protein co‐receptors such as FGF receptors, 37 kDa/67 kDa LamR, or specific integrins also play a large role in cell binding, contingent upon the AAV serotype and tissue type.134 AAV entry into host cells is highly regulated and for many AAV serotypes requires interactions with the AAV receptor (AAVR)—the transmembrane protein KIAA0319L.135 Once AAV has bound to AAVR and/or other co‐receptors, the virus enters the cells via receptor‐mediated endocytosis. Upon entry, the acidic environment of the endosome triggers a conformational change in the capsid proteins that allows a facilitated release of the viral genome into the cytoplasm.136 The host cell machinery is then used to facilitate transcription, replication, and integration of the AAV genome into the host genome.

AAV immunogenicity, defined as the propensity of the gene vector to evoke an immune response by the host, has been studied extensively as this could render these therapies ineffective if not appropriately controlled. While generally considered less immunogenic compared to other viral vectors, this response is dependent on the AAV serotype used, dosage requirements, and tissue type being targeted.137, 138 AAV vectors can trigger innate immune responses through the recognition of the AAV genome via pattern recognition receptors. The pattern recognition receptors can trigger a mild immune response, including natural killer cell activation or interferon production. Zhu et al. demonstrated that the TLR9‐MyD88‐type I interferon (IFN) pathway played a major role in the immune response to AAV, and that inhibiting this pathway had the potential to improve the efficacy of AAV.139 In addition, Hösel et al. observed the innate immune response to AAV vectors in human liver cells.140 Their work found that the vectors were recognized by Toll‐like receptor 2, which in turn caused an upregulation of inflammatory cytokines, initiating a larger immune response. Recent research conducted by Bucher et al. found that the innate immune responses brought on by AAV in dendritic cells could be attributed to extra unpackaged viral DNA found in AAV vectors during preparation, revealing potential pathways to mitigate the immune response seen by improving the AAV vector purification process.141 AAV vectors can also induce an adaptive immune response over time, specifically with repeated serotype use or tissue targeting. These subsequent administrations can lead to the production of neutralizing antibodies against the capsid proteins, which in turn can neutralize the virus and reduce its effectiveness. In short, the immunogenicity of AAV is very dose dependent, where less frequent (e.g., staggered over months) and lower dosages result in a weaker immune response.142, 143 An epidemiology study conducted by Calcedo et al. found that neutralizing antibodies affected different AAV serotypes to varying degrees, and that the presence of neutralizing antibodies greatly hindered AAV‐mediated transduction.144 Further, it was found that AAV2 is more likely to provoke a strong immune response and the antibodies to AAV2 were most common, followed by antibodies to AAV1, in comparison to other AAV serotypes, such as AAV7 and AAV8.

Although AAV is considered non‐pathogenic and safe for clinical use, large doses can cause immunotoxicity (e.g., cytokine storm) in tissues such as the liver, lung, and heart which can lead to potentially severe and life‐threatening outcomes.145 Hinderer et al. conducted a study wherein non‐human primates and piglets displayed severe toxicity following IV injection of AAV vectors.146 Their study concluded high AAV doses lead to both systemic and sensory neuron toxicity, regardless of the transgene or serotype used. Another safety concern for insertional mutagenesis arises when the viral genome integrates into the host's genome, which can lead to oncogenesis or genetic abnormalities. Other gene therapies that involve vectors, such as lentiviruses or retroviruses, can fully integrate their genomes into the host DNA, while AAV generally does not readily integrate into the host genome. But in rare cases, often induced by high doses or certain serotypes, AAV can integrate into the host genome at specific chromosomal regions.136 An analysis of AAV integration sites within the human chromosome conducted by Miller et al. elucidated the key integration sites and a profile for insertional mutagenesis risks for AAV vectors.147 They found that insertional mutagenesis hotspots exist within the chromosome, but subsequent work has shown that AAV vector design can be further optimized to mitigate this risk.148

The key challenge in AAV‐based gene therapy is the pre‐existing immunity that exists among the patient population. As AAV infections are naturally occurring, it has been found ~80% of the human population possess anti‐AAV neutralizing antibodies where their presence may hinder broad deployment of AAV gene therapies.149 Considering this, patients receiving AAV gene therapies are often treated with corticosteroids to mitigate the pre‐existing antibodies and dampen the immune response. Varying the serotypes used can overcome this pre‐existing immunity, as some serotypes, such as AAV9, may not be as widely encountered in the general population, and thus have a lesser immune response.150 As for the long‐term safety of AAV‐based gene therapies, research is still actively being pursued to identify and define the factors that impact long‐term patient health.

AAV vectors exhibit a range of tissue‐dependent tropisms, determined primarily by the serotype of AAV and the ability of those serotypes to identify and bind to specific cell surface receptors. As such, endogenous and active targeting are not mutually exclusive concepts in the design of AAV for tissue‐specific gene delivery. There are 12 AAV serotypes available and we will highlight a few commonly employed in gene therapy applications. Table 2 summarizes previously identified co‐receptors and tissue tropisms of naturally occurring AAVs. We note while there is generally consensus on serotypes most amenable for transduction in specific tissues (e.g., liver‐tropic serotypes AAV3/AAV8, BBB‐penetrating variants of AAV9). However, comparisons are often difficult across studies given the differences in AAV manufacturing between laboratories as well as selection of reporter transgene and promoters utilized. Moreover, most AAV serotypes show broad activity in a range of tissues with a preference for certain tissues based on the presence and overall levels of their co‐receptors on permissive cell types. As such, we will highlight outcomes of a two key studies with head‐to‐head comparisons of biodistribution for multiple AAV serotypes following IV administration in mice. To the best of our knowledge, similar analyses have not been conducted in non‐human primates or other large animal models.

Zincarelli et al. performed a comprehensive assessment of in vivo gene expression and biodistribution of AAV vector serotypes 1–9 in mice162 (Figure 2). The study was one of the first in the field to perform direct comparisons of tissue tropism profile for multiple AAV serotypes, produced in the same manner, carrying the same transgene (luciferase) under the control of the same promoter, and across a broad range of tissues following tail vein injection. They found that each serotype had varying levels of expression, and defined these levels as low, moderate, and high expression groups by monitoring bioluminescence via IVIS in live animals. AAV2, AAV3, AAV4, and AAV5 were of the low expression group (Figure 2a). AAV1, AAV6, and AAV8 were of the moderate expression group (Figure 2b). AAV7 and AAV9 were of the high expression group with bioluminescent signal detected primarily in the liver (Figure 2c). In all cases, gene expression persisted for up to 9 months (Figure 2d) highlighting the durability of expression using AAV gene vectors. As for distribution, it was observed that the liver and the hindlimbs were the main location of expression for AAV1, 2, 5, 6, 7, and 9. AAV8 and AAV9 were expressed more ubiquitously throughout tissues with AAV9 showing the strongest transduction overall. Interestingly, they also found that AAV4 had the highest expression in the lung and kidney, with additional expression seen in the heart. AAV6 was also seen to have expression in the heart, as well as in the liver and skeletal muscle. This study demonstrates how differences in AAV capsids across serotypes can significantly affect transduction in tissues throughout the body.

A recent study published by Walkey et al. further examined the tissue and cellular tropisms of multiple AAV serotypes following IV injection in mice.170 AAV3B (a subtype of AAV3), AAV4, AAV5, AAV6, AAV7, AAV8, AAV9, AAVrh8 (rhesus macaque‐derived AAV8 variant), AAVrh10 (rhesus macaque‐derived AAV10 variant), and AAVrh74 (variant derived from rhesus macaque muscle tissue) were delivered systemically to male and female mice to observe transgene expression across tissues. The majority of AAV serotypes, except AAV3B and AAV4, displayed substantial liver transduction further highlighting the challenges in achieving extrahepatic gene delivery. They also found that AAV4 was observed in endothelial cells in most tissues, as well as having transduced beta cells in the pancreas. Interestingly, transduction with several AAV serotypes was also observed in the adrenal, testicular, and ovarian tissues which, to the best of our knowledge, have not been sites of transduction documented in prior work.

Researchers often employ directed evolution to optimize AAV capsid properties, enhancing and altering capsid proteins' characteristics to influence the vector's behavior.27 Directed evolution is the process of generating a new, distinct amino acid sequence in the capsid and selecting for capsids possessing desirable properties. In the scope of AAV capsid design, directed evolution allows researchers to specifically engineer capsid proteins (VP1, VP2, and VP3) to enhance existing or grant novel capsid characteristics promoting higher transduction efficiency, increased tissue specificity, and increased immune evasion. An efficient strategy to generate a library of novel AAV capsids is error‐prone polymerase chain reaction (PCR), which relies on random mutations' introduction into the gene encoding the AAV capsid proteins during the amplification process. This process aims to introduce such mutations at a high rate, creating a vast library of genetic variants of the capsid. This library is then tested for favorable characteristics that will enhance the transduction efficiency of the AAV capsids. For example, Qian et al. employed error‐prone PCR to engineer AAV5 for applications in hepatocytes.171 They could improve the AAV5 transduction efficiency while simultaneously maintaining the vector's uniquely low seroreactivity. The capsids resulting from this study allow for optimized hepatocyte transduction and are a promising option for liver‐targeted gene therapies. Characteristics such as cell binding, capsid formation, and genome packaging are tested and screened to filter for those that show optimal characteristics for the specific application.

Another common method used for AAV library generation is phage display. This method employs genetically modified phages, which are viruses that specifically infect bacterial cells to establish and optimize a capsid library. The phages are genetically engineered to express a broad range of varying capsid proteins on their surface; libraries of these phages are then screened for their affinity to a specific ligand or cell type in a high‐throughput fashion. Capsid variants or proteins that display a high affinity to targets of interest are retained and amplified to improve the desired behavioral components of the capsid. Work et al. employed this method in their early work to develop AAV2 viral vector capsids that were selective to and had enhanced transduction of vascular smooth muscle cells.172 By employing specific ligands from the smooth muscle cells that were identified via phage display, the AAV2 vectors were amenable to being engineered to specifically recognize and successfully bind to vascular smooth muscle cells more efficiently. Phage display proves very useful in identifying AAV capsid variants that have superior tropisms in varying tissue, as those capsid variants and proteins that are more favorable for a certain tissue type will display higher affinities to those target cells and target receptors. Additionally, researchers can narrow down AAV capsids based on their immunogenicity and protein interactions by exposing capsid variants to neutralizing antibodies or peptides, then observing those that are able to resist and avoid their immunogenetic effect.

Rational design approaches have also been used to optimize AAV vectors for enhanced tissue specific gene delivery. One approach is to create chimeric or hybrid AAV vectors. Chimeric vectors are formed by combining capsid proteins from varying AAV serotypes to create novel vectors with enhanced characteristics such as reduced immunogenicity or increased tissue tropism.27 Early work by Hauck et al. demonstrated the concept in a very functional manner, as they developed a chimeric AAV vector using capsid components from both AAV1 and AAV2 to combine the transduction characteristics of both vectors.173 In vivo data demonstrated that their newly formed chimeric vectors displayed liver and muscle transduction that was similar to that of the original parent vectors, hence demonstrating a successful implementation of the concept. Varying combinations can be employed to create new hybrid capsid formations and examine their benefits on gene delivery in a range of specific gene therapies. However, it should be noted that many AAV are closely related and may possess as few as six divergent amino acids in the entirety of their capsid (e.g., AAV1 and AAV6). As such, chimeric vector engineering approaches are best suited for AAV serotypes which are phylogenetically distinct and/or possess unique receptor binding preferences. Further mapping of the AAV capsid and receptor binding domains has also allowed for researchers to rationally design AAVs with enhanced binding affinities for their target cells and/or optimize AAV vector unpackaging once internalized.174, 175

In vitro screening can be used to efficiently evaluate a library of AAV capsid variants based on their transduction efficiency. This provides researchers the means to readily identify capsid variants with improved delivery capabilities, as well as identifying structural and/or characteristic features useful in transducing specific tissues, binding to specific receptors, crossing certain delivery barriers, or even navigating the immune system.176 For example, Schröder et al. generated an AAV library via error‐prone PCR and screened this library in vitro to identify aorta‐tropic AAV variant using vascular smooth muscle cells.177 They found that a novel peptide motif (RFTEKPA) improved in vitro transduction efficiency compared to previous motifs and further validated these engineered vectors in an ex vivo aorta model. In vivo screening is another step used to screen libraries of AAV capsid variants, where the vectors are tested directly in animal models to screen for enhancements in tissue delivery and gene expression. This can be accomplished by delivering a library of barcoded AAV to mice such that highly efficient AAV can be subsequently identified in specific tissues via next‐generation sequencing.178, 179 Screening for vectors that have high transduction efficiency in organs of interest while minimizing off‐target transduction is a key aspect to gene therapy that can help mitigate side effects as well as unwanted immune responses. An important consideration for the in vivo screening is these efforts may yield vectors highly efficient in animal models from which they were derived but less efficient in human tissues due to species‐specific differences in host receptor expression.180, 181 AAV vectors obtained from screening in non‐human primates typically yield vectors that are effective in humans182 but broad screening of AAV in non‐human primates requires significant expense. Alternatively, these concerns can be mitigated, at least in part, with a mixture of screening in both human in vitro models and in vivo animal models.

Computational modeling can also be used to further refine capsid structure variation on its function.183, 184 For example, Ojala and colleagues established a computational model called SCHEMA that mapped protein structure and optimal crossover points of DNA shuffling and used it to create a hybrid AAV capsid library185 (Figure 3a). Their SCHEMA‐generated library was functionally diverse and could be analyzed for optimal in vitro properties. SCHEMA was eventually able to generate a novel AAV capsid that efficiently transduced neural stem cells in an in vivo mouse model (Figure 3b,c). Thus, this method allows for a more in‐depth analysis and understanding of the genetic components responsible for the AAV capsid's performance and can help predict the impact of novel genetic sequences or variants before they are synthesized and applied. Through a combination of computational modeling, rational design, and in vivo screening, engineered variants of AAV have also been identified capable of efficiently bypassing the BBB and achieving robust transduction within the CNS following IV administration186 (Figure 3d,e). Structural AAV analysis can be conducted to establish links between directed evolution induced structural changes to the AAV capsid and their therapeutic and immunological effects. Cryo‐electron microscopy, x‐ray crystallography, and nuclear magnetic resonance are all methods employed to assess detailed changes in the AAV capsid allowing for further optimization of protein interactions with target cells and the immune system through further rounds of directed evolution.187, 188

A way to mitigate the immune response is by conducting a final screening step focused on selecting capsid variants against neutralizing antibodies that are found in human sera.177 As previously mentioned, the presence of pre‐existing neutralizing antibodies is a major barrier to effective AAV‐mediated gene delivery. A method currently used to select AAV capsid variants resistant to these neutralizing antibodies is screening the AAV library of capsid variants against plasma/serum with those antibodies and assessing and quantifying the immune response.189, 190 Variants that are least affected and can deliver their cargo are favorable and are selected for, while those that are hindered greatly are selected against. Recently, Chhabra et al. conducted a broad study examining the activity of neutralizing antibodies against six AAV serotypes in a sample of over 500 adults from 10 countries.191 They conducted a transduction inhibition assay and established an immune response profile elucidating the relationship between neutralizing antibodies and AAV serotypes, which can prove key in future work in vector design.

LNPs possess many advantages for nucleic acid delivery, including relatively simple and scalable composition and manufacturing,192 high encapsulation efficiency, efficient cargo shielding within their aqueous core, and strong transfection capabilities in diverse cell types. The four main components of conventional LNPs are ionizable tertiary amine‐containing lipids, cholesterol, helper lipids, and PEGylated lipids.193 The ionizable lipid components are nearly uncharged at physiological pH but progressively acquire a cationic charge as the environmental pH drops.194 This feature allows their spontaneous electrostatic coacervation with the anionic RNA during LNP assembly and later following LNP endocytosis can drive efficient endosomal membrane disruption and hence endosomal escape of the intact RNA cargo as the pH of the endolysosomal compartment begins to decrease.195

Maintaining colloidal, physicochemical, and functional stability of LNPs in physiological media has proven a significant challenge. For example, it has been shown plasma exposure can accelerate lipid desorption from LNPs as a result of clotting factors which alter their degradation kinetics.196 PEG desorption rate is another factor that affects both particle stability in physiological media and organ localization. Rapid desorption of PEGylated lipids leads to greater localization in the lung, liver, or spleen, while slow desorption negatively impacts cellular uptake and endosomal escape.197 Meaningful work from Voke et al. introduced a novel method to separate hard protein corona‐associated LNPs from other nanoparticulate plasma components, such as exosomes and lipoproteins, using a combination of density gradient‐based separation and mass spectrometry‐based proteomics.198 They later studied the impact of the most abundant proteins in the hard corona of the LNP, either individually or in a mixture, on cellular uptake, endosomal escape, and mRNA transfection in HepG2 cells. They demonstrated that ApoE, despite promoting cellular uptake of LNPs, reduced endosomal escape and compromised mRNA transfection. Similarly, vitronectin, despite increasing cellular uptake, has been shown to reduce transfection efficiency of LNPs.198 It should be noted this study primarily focused on uptake in a hepatic cell line where it would be valuable to build on these important findings and further study vitronectin‐dependent uptake in other cell types. Cholesterol content has also been shown to affect the colloidal stability and degradation rate of LNPs upon systemic circulation. Specifically in prior work, increases in cholesterol molar ratio led to improved stability and reduced degradation rate, as well as reduced RNA cargo leakage from the particle.199, 200

Conventional LNP formulations have been found to accumulate predominantly in the liver through an endogenous targeting mechanism mediated by apolipoprotein E (ApoE). The ionizable lipid component inherently dictates the natural tropism to the liver upon IV administration via ApoE‐mediated endogenous targeting.201 In addition to the ApoE‐mediated endogenous targeting,62 the natural hepatic tropism of LNPs results from a combination of the structural and functional features of the liver previously described in Section 2.3. Yet, the potential applications of mRNA replacement therapies beyond the liver are vast, and manipulating LNP's structure to shift their natural hepatic tropism toward extrahepatic organs has been widely sought after. An emerging concept is that of selective organ‐targeted LNPs (SORT LNPs), which draws inspiration from the ApoE‐mediated hepatic targeting of currently commercialized LNPs. Siegwart and colleagues have reported that adding a fifth component coined “sorting lipid” to the LNP can influence its protein corona composition, adding novel targeting entities to the LNP surface in situ following IV administration.202 This SORT lipid redirects LNP toward specific receptors on specific organs that would not be accessible otherwise.

As discussed in Section 2, particle size can greatly influence a nanocarrier's targeting to and within different organs. Particle size also impacts the glomerular filtration and hence the clearance rate of the particle from the circulation which further impacts the potential for extrahepatic accumulation which increases as a function of circulation time. LNPs are most commonly prepared by solvent injection using microfluidic devices. In addition to being an easily scalable and highly reproducible technique, it also provides better control over the resulting LNPs' size and polydispersity through the careful control of different mixing parameters.203 Belliveau et al. previously demonstrated that using staggered herringbone micromixers, LNPs' size and polydispersity were inversely proportional to the total flow rate (TFR) at which fluids were introduced into the device.204 They also showed that using higher molar ratios of PEGylated lipids has been shown to reduce the particle size of LNPs. A comprehensive study by Roces et al. also demonstrated similar results regarding the relation between particle size, polydispersity, and TFR, though they demonstrated that the size stabilized at 10–20 mL/min.205 In their study, they also demonstrated that both size and polydispersity decreased with increasing aqueous to organic phase flow rate ratio (FRR), with both parameters starting to plateau at an FRR of 3:1. Additionally, they showed that increasing cholesterol while reducing cationic lipid contribution resulted in smaller, more monodisperse particles that showed higher stability following purification and dilution steps, yet such an effect was not observed for ionizable lipids. Okuda et al. showed that adding NaCl to the aqueous buffer led to the formation of larger LNPs in a concentration‐dependent fashion, and that such larger LNPs led to higher transgene expression in DCs in vivo.206

Apparent pKa is a highly influential parameter for LNP trafficking throughout the body and within the liver. Apparent pKa is dictated by the chemical identity as well as the molar contribution of each lipid component of the LNP in question. pKa has been shown to govern particle uptake by specific cell populations in the liver. Lower pKa (6–6.5), when combined with a small particle size, predominantly favors hepatocyte accumulation as the particles carry a predominantly neutral charge in the vicinity of Kupffer cells. Whereas LNPs with higher pKa (7–7.15) show preferential uptake by LSECs. Elevating the pKa above 7.3 leads to a shift in particle accumulation toward Kupffer cells as the particles become increasingly cationic.62, 207 A detailed study by Dilliard et al. investigated the impact of pKa on organ‐specific delivery of LNPs and demonstrated that the underlying mechanisms could overlap passive and endogenous targeting modalities (Figure 4).207 Their study showed that LNPs with apparent pKa greater than 9.2 favored lung accumulation. As the pKa dropped below 9, LNPs started shifting toward the spleen, with maximum splenic targeting achieved by LNPs with apparent pKa below 6. LNPs with apparent pKa between 6 and 7 showed preferential liver deposition. Changes in pKa change the particle charge at physiological pH (Figure 4a), which also changes the biomolecular corona composition (Figure 4b–d). This ultimately affects the main organ of accumulation for the LNPs.

Particle deformability can impact its extravasation through the endothelium. Abundant data have explored liposome deformability by incorporating edge activators to promote particle elasticity and deformability, which are described in this previous review.208 Yet, limited data exist on the effects of deformability in LNPs on their endothelial extravasation. LNPs formulated using an unsaturated helper lipid (DOPE) exhibit higher membrane fluidity and result in more deformable LNPs, which showed higher liver accumulation in regions with smaller endothelial fenestrations. LNPs containing the saturated lipid 1,2‐distearoyl‐sn‐glycero‐3‐phosphocholine (DSPC) had more rigid membranes and showed a preference for the spleen, which has larger endothelial fenestrations.209

For LNP, passive and endogenous targeting are not mutually exclusive targeted delivery mechanisms as some changes to LNP composition can change its physical properties, such as size and charge, while simultaneously transforming the biomolecular corona composition of these particles. For example, shortening the ionizable lipid tail redirected LNPs from the liver to the spleen (Figure 5a–c) via uptake in splenic dendritic cells, macrophages, and neutrophils (Figure 5d–i), regardless of the head group composition.210 A potential explanation for this observation is that shorter tail groups led to the formation of larger LNPs, reducing the specific surface area of the system and exposing more of its helper lipid component (DSPC) on its surface. DSPC has been reported to less readily interact with ApoE than its unsaturated analogs. Both the increased LNP size and reduced ApoE content in its biomolecular corona lead to LNPs' targeting toward the spleen rather than the liver. ApoE binding to the surface of LNPs has been shown to drive ionizable lipid components to the core and push cholesterol to the surface of the LNP.211 Alternatively, the binding of vitronectin or fibrinogen to the NP surface has been shown to reveal cryptic binding sites on these proteins.212 Receptor‐mediated uptake is coordinated by the cell membrane receptors for the components in the LNP's biomolecular corona.

High throughput screening methods have enabled investigators to understand the impact of structural variations of LNP on differential organ/tissue/cell targeting and transfection efficiency at target sites. To that end, the DNA barcoding system developed by James Dahlman and colleagues has proven an invaluable tool to broaden the design space in LNP based therapeutics.213 The system allows the administration of broad arrays of LNP to a single animal and the subsequent generation of transfection data of every single particle within the array on a cellular level by co‐loading a unique DNA barcode to each particle along with the reporter mRNA cargo. The unique DNA barcode can later be sequenced in collected tissue to locate the preferred target cell of each LNP design in the array and how that LNP performed in terms of transfection or gene silencing using reverse transcription quantitative polymerase chain reaction (RT‐qPCR). This method can significantly reduce the number of animals required per study and avoids complications with in vitro–in vivo correlation. Yet, some concerns regarding the potential for simultaneously administered LNP within the array to impact each other's in vivo behavior have been raised; more data is still required to validate or invalidate such concerns.

The incorporation of PEGylated lipids, as well as PEG desorption kinetics play a central role in the endogenous targeting process of LNPs. The initial presence of PEG on LNPs upon entry into the bloodstream is crucial to minimize the adsorption of opsonins, which when found in high percentages in the biomolecular corona, can lead to rapid clearance by the MPS system. Although subsequent desorption of PEG is essential to permit the subsequent steps of endogenous targeting and allow efficient particle uptake and endosomal escape, slowly desorbing PEG can also prolong the circulation time of the LNP, leading to an enhancement in its extrahepatic accumulation. Generally, PEGylated lipids with shorter hydrophobic tails (e.g., 1,2‐dimyristoyl‐rac‐glycero‐3‐methoxypolyethylene glycol, DMG‐PEG) can desorb more rapidly than their counterparts with longer hydrophobic tails (e.g., 1,2‐distearoyl‐sn‐glycero‐3‐N‐amino(polyethylene glycol), DSPE‐PEG).214 It should be noted that as the molar contribution of PEGylated lipids increases, the impact of tail length on the desorption rate increases.215 It has also been demonstrated that fast‐desorbing PEG with short tails generates less anti‐PEG IgM in mice compared to slower‐desorbing PEG with longer acyl chains.216 Hence, it could be valuable to tailor the kinetics of de‐PEGylation by combining PEGylated lipids with varying acyl chain lengths on a single particle to minimize potential ABC of PEGylated LNPs. ABC is an immunogenic response that has been observed upon repeated administration of pegylated LNPs mediated by MZ B‐cells in the spleen resulting in anti‐PEG IgM generation. As previously noted, much higher doses of LNPs (1000‐fold increase in protein expression) may be required for therapeutic applications (e.g., protein replacement) relative to mRNA vaccine applications, which may dictate the incorporation of much higher doses of PEG.217 The ABC effect at this point may become an obstacle, and careful manipulation of PEG desorption kinetics or its full replacement with less immunogenic alternatives such as poly(carboxybetaine)s and poly(sulfobetaine)s, polysarcosines, poly(phosphoester)s, and poly(oxazoline)s may be necessary.218, 219 Another alternative may be resorting to cell‐membrane‐based coatings or functionalization with anti‐phagocytic signals like CD47.

Seminal work from the lab of Daniel Siegwart demonstrated that modifying the charge of LNPs could effectively redirect them to different organs, showcasing the adaptability of LNPs in targeted drug delivery.202, 207, 220 These SORT LNPs incorporate a sorting lipid outside the four key lipid components of classical LNPs, while maintaining the key lipids' molar ratios constant could redirect the LNP's natural tropism away from the liver. This effect occurs independently of the sorting lipid structure and is primarily a function of the sorting lipid's net charge and molar percentage of the SORT lipid in the overall LNP composition. Their studies showed that incorporating a permanently cationic lipid such as 1,2‐dioleoyl‐3‐trimethylammoniumpropane (DOTAP), didodecyldimethylammonium bromide (DDAB), or 1,2‐dioeoyl‐sn‐glycero‐3‐ethylphosphocholine (EPC) could shift particles from the liver to the spleen (at 10%–15% molar contribution) and eventually to the lung as well as the kidney (at a molar contribution greater than or equal to 40%). On the other hand, incorporating permanently anionic lipids (e.g., 18pA, 14pA, or 18 BMP) resulted in particle tropism exclusively to the spleen, with no detectable expression in other organs. These tropisms were maintained independent of structural variations to the linker and tail groups of the sorting lipids candidates. Meanwhile, ionizable cationic lipids, such as DODAP, and C12‐200 demonstrated a maintained tendency to direct LNPs to the liver. Moreover, Cheng et al. demonstrated this trend not only in an experimental ionizable cationic lipid, but also in the clinically established Dlin‐DMa‐MC3, a two‐tailed single head ionizable lipid.202 Dlin‐DMa‐MC3 has been previously reported to deliver siRNA and mRNA exclusively to the liver upon IV administration, but addition of an ionizable cationic lipid altered this tropism. In terms of toxicity, all SORT LNPs showed excellent liver and kidney tolerability, as well as no rise in inflammatory cytokines following in vivo administration.

Consistent preclinical data have been obtained in mice by several research teams that support the charge‐based targeting of LNPs to various organs. LNPs with a permanently positive charge under physiological pH following PEG desorption to target the lungs, their negatively charged counterparts can achieve spleen targeting and whereas generally neutral LNPs favored liver accumulation. The method by which LNP charge was induced, whether through introduction of a sorting lipid or manipulation of existing lipid components, did not appear to affect these tropisms. For example, an alternative and more simplistic approach to adding a sorting lipid to the four original lipid components of an LNP was proposed by LoPresti et al.221 Their study demonstrated that replacing the helper lipid component with either a cationic, anionic or neutral helper lipid led to a significant redirection of LNPs to the lungs, spleen, and liver, respectively, regardless of the ionizable lipidoid's chemical identity. Replacing the neutral DOPE with cationic DOTAP shifted liver to lung reporter protein expression from 36:1 to 1:56, while replacing DOPE with the anionic phosphatidylserine shifted liver to spleen reporter protein expression from 8:1 to 1:3. The underlying mechanism of this organ targeting mechanism could be related to alterations in the composition of the biomolecular corona as well as the potential that some LNP components could alter immune cell trafficking to specific organs. Using a barcoding high throughput in vivo screening approach, lymph node targeting LNPs have also been identified where lead formulations possessed neutral helper lipid (e.g., DOPE) at an optimal molar ratio and generally lacked cationic lipid species.222 Cationic lipids, such as DOTAP, have been shown to alter the composition of the biomolecular corona of LNPs to be predominated by vitronectin. Vitronectin is a glycoprotein involved in cell adhesion and coagulation that can bind to αVβ3 integrin receptors highly expressed in the lungs.223 Vitronectin tends to bind the glycosaminoglycan heparin, which is abundant in serum.224 It is worth noting that such cationic LNP‐polyanion (heparin) interaction could lead to formation of LNP aggregates small enough to be entrapped in the narrow capillaries of the first organ it encounters (e.g., lung) following injection.

The hypothesis regarding immune cell trafficking is more related to the proposed mechanism of action of anionic lipids. Anionic lipids have been shown to enhance the immune cell infiltration into the spleen by LoPresti et al.,221 and further work from Alvarez Benedicto et al. who showcased spleen‐targeted SORT LNP by incorporating anionic 18:1 phosphatidic acid (PA) as a sorting lipid in their LNP.225 These LNPs could produce Chimeric Antigen Receptor (CAR) T‐cells in vivo in wild type mice by delivering CAR encoding mRNA to splenic CD3+, CD8+, and CD4+ T‐cells with high specificity. The same was also shown when exchanging cholesterol with an anionic analog, the naturally occurring bile acid, resulting in improved splenic tropism following IV or intraperitoneal injection.226

Beyond the SORT LNP approach, extrahepatic delivery can be enhanced by prolonging LNP circulation time. A relevant approach was proposed by Chander et al. who reformulated LNPs to contain a molar ratio of 40% helper lipids rather than 10%.227 Using either DSPC or egg sphingomyelin (ESM) as the helper lipid in such a high molar concentration altered the morphology of the LNP to feature a small solid core surrounded by a lipid bilayer. This altered conformation led to better RNA shielding in the core, higher serum stability of the particle, and prolonged circulation. Results also showed LNPs with 40% ESM significantly outperformed the Onpattro formulation in in vitro transfection (10‐fold enhancement in luminescence) as well as splenic (three‐fold increase) and bone marrow (two‐fold increase) accumulation in vivo. Isaac et al. identified an endogenous targeting strategy with vitamin D3 containing LNPs which enables targeting of the vitamin D receptor leading to highly selective delivery to the pancreas.228 More recently, Su et al. demonstrated that, using their own library of a degradable ester‐core ionizable lipid, they could achieve more exclusive lung targeting when they also added a permanently cationic lipid and entirely removed cholesterol from their formulation. They showed similar exclusive targeting of the liver upon re‐incorporation of cholesterol and a phospholipid into the formulation.229

With these established and newly discovered principles of endogenous targeting for LNPs, engineering LNPs that are enriched with endogenous proteins known to shift their biodistribution to desired organs has also been explored. To accomplish this, LNPs can be engineered to preferentially accumulate specific species such as ApoE and vitronectin to influence their fate in vivo by varying their net charge and/or lipid content as described above. For example, LNPs can be designed as a function of ionizable lipid chemistry to increase levels of ApoE in the protein corona leading to enhanced delivery to hepatocytes via LDL receptors.207, 230, 231, 232 Similarly, LNPs have also been designed in an ionizable lipid‐dependent manner which acquire a protein corona enriched in vitronectin to enable targeting the lung, kidneys, and the tumor microenvironment via uptake in integrin overexpressing cancer cells.233, 234, 235 In recent work, albumin‐binding LNPs were generated by substituting PEGylated lipids with an Evans blue‐modified lipid to stabilize the formulation.236 This LNP has shown promise for mRNA vaccine applications by minimizing hepatic accumulation while promoting albumin‐mediated lymphatic transport and receptor‐mediated uptake in dendritic cells. Another related biophysical approach is altering membrane fluidity through the incorporation of more unsaturated lipid components into the LNP. This can alter or enhance the protein corona binding to the LNP's surface. Zhang et al. demonstrated membrane fluidity alterations by exchanging the helper lipid DSPC with its unsaturated analog DOPE, which led to enhanced ApoE binding on DOPE‐containing LNPs and enhanced liver rather than spleen targeting.209

Lipid NP components have also been leveraged to address the inherent limitations of calcium phosphate (CaP) gene delivery technologies. CaP is an inorganic material that has been used for several decades to condense and deliver RNA as well as facilitate viral transduction, yet its physical instability limits its in vivo application. Hence, several approaches have been explored to enhance CaP's efficacy, such as coating it with polymers or lipids.237 Lipid coating could address the inherent instability of CaP‐siRNA complexes.237 For instance, a CaP‐siRNA core coated with cationic lipids, then functionalized with PEG and anisamide, resulted in a nanocarrier that effectively targeted sigma receptors in B16F10 melanoma cells. Following a single IV injection, antiluciferase siRNA led to a 78% reduction in luciferase expression in metastasized B16F10 in vivo.238 Another approach relied on multilayer lipid deposition on CaP cores, with an inner layer of the anionic lipid DOPA followed by a cationic or neutral lipid.239 This setting significantly improved the gene silencing effect and allowed the co‐loading of a chemotherapeutic agent. The lipids used in these studies are similar in composition to those used for classical LNP preparation.

While many new non‐viral gene delivery approaches have emerged in the past approximately three decades following its discovery, PEI is one of the most robustly studied systems for polymer‐mediated encapsulation of nucleic acids for gene delivery.242 PEI is a cationic polymer consisting of monomers of ethylenimine which are covalently linked together to form the PEI backbone and can be arranged in linear or branched architecture with varying overall polymer MW. PEI polymers with MW of 20–25 kDa in linear or branched form have been shown to be highly efficient at transfection of mammalian cells.242 Previous work has shown that the cationic PEI is able to effectively and spontaneously complex with negatively charged DNA or other nucleic acids (e.g., siRNA and mRNA). In more recent work, Binder et al. utilized molecular dynamics simulations to determine the thermodynamics of PEI and siRNA complexation.243 They simulated complexes with PEI of varying MWs and degrees of branching where they validated that the complexation of PEI and siRNA is thermodynamically favorable. In addition, they showed that the reaction was even more favorable at pH similar to endolysosomal conditions than at neutral pH. However, significant concerns remain on the associated toxicity and proinflammatory responses associated with PEI‐based PNP. To address this, lower MW (e.g., 2 kDa) PEI can be crosslinked using bioreducible linkers (e.g., disulfide‐linked) to achieve optimal DNA/RNA packaging efficiency while reducing the cytotoxicity of the resulting PNP.244, 245

PBAEs are a unique class of biodegradable polymers that have been developed for gene delivery applications. PBAE structure is an amine‐containing polyester, with a highly modifiable structure.246 Lynn and Langer detailed their rationale for the synthesis of PBAEs, basing it upon a poly(ester amine) approach by combining esters and amines via Michael Addition reaction enabling production of a diverse library of polymers.247 Importantly, PBAEs are biodegradable which addresses one of the major drawbacks of conventional, nondegradable PEI‐based PNPs. In regards to their performance, PBAE is highly tunable with different amine/acrylate combinations producing PNP with unique DNA compaction and endosomal escape capabilities. Moreover, PBAE gene delivery efficiency and targeting to specific tissues can be modified by altering the terminal end‐capping groups which is highly advantageous for application‐specific design and optimization. We will further elaborate on specific design modifications for organ‐selective targeting of PBAE PNP in Section 3.3.3. In initial work to establish the clinical relevance of this system, Green et al. illustrated an optimized PBAE PNP‐mediated gene delivery comparable to adenoviral gene vectors in vitro and it has gone on to be successful in a wide range of preclinical in vivo models.248

PLGA is a biocompatible and biodegradable polymer used to fabricate PNP for both drug and gene delivery. PLGA is a block copolymer consisting of lactic and glycolic acid subunits, which allows for a uniquely tailorable polymer for NP formulation. The polymer MW, monomer ratios, and monomer MWs can all be altered through polymer synthesis. Cun et al. described a method of encapsulating siRNA into PLGA NPs using double‐emulsification.249 They achieved >50% encapsulation efficiency and a particle diameter of 250 nm, demonstrating that PLGA NPs provide an avenue for delivery of siRNA and other nucleic acid cargoes. One advantage of PLGA over other PNP is that PLGA can be readily used for combined delivery of small molecules and nucleic acids. Heo and Lim showed that PLGA NPs could be used for simultaneous delivery of siRNA and imiquimod for the immunomodulation of dendritic cells.250 This potent combination therapy outperformed the commercial control in silencing the immunosuppressive STAT3 gene.

Another biodegradable polymer system receiving considerable attention for nucleic acid delivery is poly(amine‐co‐ester) (PACE). Zhu et al. were able to establish relationships between the MW of the PACE polymer and its ability for effective complexation with mRNA.139 In addition, they showed that PACE polymers with a MW of 5–10 kDa were most effective at promoting endosomal escape for efficient mRNA delivery. Piotrowski‐Daspit et al. screened a library of PACE polymers with varied MW and functionalization to generate PNPs to evaluate their blood concentration over time and tissue biodistribution within mice after systemic administration.251 Overall, they noted that it was very difficult to reduce liver and spleen accumulation of PACE PNPs, regardless of circulation time changes. However, an increase in retention time in the blood may allow for additional PNPs to reach more difficult‐to‐reach sites such as the bone marrow, heart, or lungs. Overall, they determined that the polymer chemistry and NP characteristics strongly impacted the fate of PNPs and delivery ability of nucleic acid therapeutics due to their impact on retention time and tissue distribution.

An emerging body of work focuses on zwitterionic PNPs. Zwitterions feature positively and negatively charged regions within their structures yet are overall neutral in charge.252 This property can allow them to facilitate nucleic acid encapsulation efficiency rivaling cationic polymer yet at much lower cytotoxicity. Shen et al. developed a zwitterionic polymer‐based polymeric micelle.253 This system switched surface charge in acidic environments such as the endosome. When delivering plasmid DNA, the zwitterionic polymer micelles showed effective gene transfection and low cytotoxicity in acidic pH. Zwitterionic modifications can also be made to existing polymers used for nucleic acid delivery. For instance, Liu et al. modified PEI with zwitterionic molecules in an attempt to reduce the cytotoxicity associated with PEI.254 These NPs showed a significant reduction in cytotoxicity when tested in vitro, although with slightly less effective gene silencing for encapsulated siRNA cargo. Zwitterionic polymers could yield a promising platform for gene therapies, especially for navigating acidic environments such as the tumor microenvironment and the endo/lysosome.

Similarly to zwitterions, charge‐altering releasable transporters (CARTs) are an emerging class of polymers used for nucleic acid delivery due to their ability to change their charge depending on their environment. McKinlay et al. reported on the use of CARTs to deliver mRNA, in which the polymeric material initially conveys a positive charge in order to complex with the negatively charged mRNA.255 Upon delivery into the cell, the CARTs undergo a charge‐altering mechanism that converts the positively charged PNPs into neutral small molecules, releasing the mRNA contained within NP. It is believed that the CARTs enable more efficient endosomal escape, owing to their changing charge. In addition, CARTs have been shown to exhibit organ‐specific targeting of nucleic acids with specific modifications to their polymer backbone.256 Several past studies have reported targeted delivery to the lungs or spleen, depending on the CART properties. While glycine‐based CARTs show enhanced retention and transfection in the spleen when delivered systemically, lysine‐based CARTs exhibit higher expression levels in the lungs.255, 257 Other alterations, such as guanidinylated serinol CARTs, showed selectivity for either the spleen or the lungs, depending on specific properties of the CART, such as block length, charge ratio, and degree of polymerization.256

CaP‐based gene vectors, previously discussed in the context of LNPs, have also shown promise when used in combination with PNP‐based therapeutic approaches. For example, Giger et al. explored the incorporation of PEG‐conjugated alendronate to coat CaP‐siRNA NPs and could demonstrate that, despite being taken up predominantly via clathrin‐mediated pathways, these particles could still exert potent gene silencing in vitro.258 These CaP‐siRNA NPs achieved ~80% VEGF mRNA knockdown in vitro in pancreatic cancer cells (PanC‐1). Pitella et al. demonstrated that a CaP‐siRNA system co‐assembled with a pegylated charge‐conversional polymer (CCP) showed high siRNA loading and rapid endosomal escape.259 These particles enabled significant VEGF silencing (68% knockdown) in subcutaneous BxPC3 tumors following systemic administration in vivo.260 PEI modification is another approach to stabilize CaP‐based RNA delivery systems further and enhance their transfection performance. This was demonstrated by Devarasu et al., who reported the fabrication of a CaP‐based system coated with alternating layers of siRNA and PEI.261 This system showed potent 95% luciferase silencing in vitro and in a murine tumor xenograft model in vivo.

While synthetic polymers have remained popular for gene delivery applications given their ability to be manipulated in a controlled and systematic fashion, many naturally derived polymers have garnered attention for use in PNP formulation. PLL is a cationic polypeptide that has been widely studied for gene delivery purposes.262 Due to the positively charged nature of PLL, it easily complexes with DNA and RNA to form NPs that can be delivered intravenously.263 It has been determined in prior work that PLL containing >20 amino residues in length can efficiently condense plasmid DNA in PNP with diameters below 100 nm at physiological ionic strength and pH. To improve the stability and biocompatibility of PLL‐based PNP, PLL‐PEG copolymers are often used to reduce their net surface charge and prevent their aggregation. One of the most clinically advanced PNP consists of a 30‐mer PLL with an N‐terminal cysteine conjugated to a 10 kDa PEG (CK30) which has been tested in the context of inhaled gene delivery264 and gene delivery to the CNS.265

Chitosan is a naturally occurring polysaccharide that can be found in crustaceans, fungi, and insects.266 Chitosan remains a popular natural polymer for the encapsulation and delivery of gene therapies. An early study by Erbacher et al. demonstrated that chitosan could condense and complex DNA into NPs in the range of 50–100 nm, proving chitosan's promise for gene delivery.267 These nano‐complexes effectively transfected HeLa cells, maintaining transgene expression for 96 h. Furthermore, Rudzinski et al. evaluated chitosan NPs for their ability to deliver siRNA to HCT‐116 colon cancer cells.268 They showed that chitosan PNP had comparable transfection efficiency to Lipofectamine 2000 and were able to effectively reduce the expression of β‐catenin, a protein associated with colon cancer spread. In a similar comparison, Redhwan et al. utilized chitosan to encapsulate siRNA with the ultimate goal of reducing apoptosis caused by hyperglycemia.269 The siRNA encapsulated within chitosan PNP exhibited a higher transfection efficiency than the naked siRNA in Neuro2A cells. When compared to lipofectamine, the chitosan NPs were able to achieve comparable knockdown of target genes and reduced apoptotic markers levels observed in the untreated control group. Zhou et al. prepared chitosan NPs with double stranded DNA (dsDNA) cargo.270 In their work, they determined that the dsDNA‐containing chitosan NPs activated clathrin‐mediated endocytosis through the upregulation of the clathrin heavy chain gene. This study was the first of its kind to determine the mechanism by which dsDNA is delivered into cells.

Other polypeptide systems have also shown promise as carriers with several advancing toward clinical impact. Peptide‐mediated gene delivery can be sub‐classified into cationic peptides, amphipathic peptides, RNA covalent conjugates, and phase‐separation systems. The most prominent cationic peptide in clinical trials is protamine, which has been evaluated for mRNA delivery in non‐small cell lung cancer (NCT00923312↗ and NCT01915524↗), metastatic melanoma (T00204607), prostate cancer (EudraCT 2008‐003967‐37 and NCT0187738), and rabies (NCT02241135↗).271, 272 Synthetic cationic peptides that tried to improve the overall performance featured branching and histidine residues (e.g., H3K(+H)4b), which improved particle assembly as well as endosomal escape simultaneously.273 Overall, many of these cationic peptides exhibit lower transfection performance than other systems, potentially due to limited cytosolic mRNA release.274 In contrast to predominantly cationic peptides, amphipathic peptides feature both cationic and hydrophobic regions, with the hydrophobic regions playing a stronger fusogenic role at the plasma and endosomal membranes through hydrophobic interactions.275, 276, 277

Sun et al. reported the development of a conjugate peptide based on the pH‐responsive, self‐coacervating, histidine‐rich HBpep peptide, modified by lysine insertion at position 16, shifting its isoelectric point from 7.5 to 9, thereby rendering the system monomeric at physiological pH.278 They then further conjugated the lysine head with a disulfide bond containing a hydrophobic group that induces self‐coacervation of the system at near neutral pH, where it aggregates by liquid–liquid phase separation to form micrometer‐sized coacervates at physiologic pH. Yet the disulfide modification renders the system self‐immolative under the reducing glutathione‐rich intracellular environment. Their system could affect energy‐independent endocytosis of peptide‐mRNA coacervates, followed by glutathione‐mediated release directly into the cytosol. Miliotou et al. presented direct conjugation of in vitro transcribed RNA to protein transduction domain (PTD, peptide sequence PFVYLI).279 They argued that the covalent conjugation rather than mere electrostatic assembly approach of the peptide carrier and the RNA cargo, although more chemically limiting, could lead to reduced cargo loss to RNase degradation. Their system showed promise in an in vitro model of β‐thalassemia.

Cationic polymers are largely associated with effective gene delivery. El‐Kharrag et al. showed that there was a positive correlation between NP charge and transfection efficiency. All NPs tested were positively charged, and as the magnitude of the positive charge increased, the percent of transfected human hematopoietic stem and progenitor cells increased.280 However, one of the main issues with cationic polymers is cytotoxicity.281 While this is observed with cationic PNP regardless of their therapeutic cargo, this can prove especially problematic for PNP‐based gene therapies. A large body of research has looked at the cytotoxicity of PEI specifically in the context of gene delivery. This is due to the positively charged polymer associating with and compromising the negatively charged cell membrane.282 At neutral pH, PLGA does not have the positive charge required to complex the anionic RNA phosphodiester backbone. Thus, to utilize PLGA as an RNA delivery system, scientists have added cationic chemical groups such as chitosan to deliver siRNA in mice.4

There are conflicting reports on the impact of NP size on gene delivery abilities. Ogris et al. generated DNA/transferrin‐PEI complexes of varying sizes, from 40 to >1000 nm.283 Using these complexes to reduce luciferase gene expression, the small particles showed at least a 10‐fold reduction in multiple cell lines. This suggests that smaller NPs have better transfection abilities than large NPs. In a separate study, Ogris et al. studied DNA/PEI particle size and found conflicting results.284 They showed that large, aggregated complexes showed higher levels of gene expression. As such, the impact of NP size on transfection efficiency will be largely context dependent, relying on polymer material and target cell type. Moreover, the differences in cargo loading for larger versus smaller polyplexes make direct comparisons of transduction efficiency in specific cell types challenging.

The MW of the polymers used in NPs has been shown to impact cytotoxicity and delivery. When looking at gene delivery specifically, the differences due to MW can greatly affect transfection abilities. For instance, Choi et al. investigated chitosan and RNA polyplexes with varying MWs of chitosan as an immunotherapy for cancer. They found that the highest MW chitosan had the greatest effect, resulting in the largest immune response among the polyplexes tested.285 While a large immune response can be beneficial for immunomodulatory therapeutics, a dramatic increase in immune cell infiltration can prove harmful in other situations. In fact, while high MW polymers are associated with higher transfection efficiency, they can also lead to high cytotoxicity and poor nucleic acid cargo release at the target site.286 Breunig et al. found that incorporating disulfide bonds within the polymer structure resulted in lower toxicity.287 In the reductive environment within the cell, the high MW polymer NPs were broken down to lower MW polymers which are less cytotoxic and better tolerated. Alameh et al. found that an increase in the MW of chitosan resulted in higher NP internalization within cells and more knockdown by the siRNA cargo.288

Similar to LNPs, pKa, or acid dissociation constant, significantly impacts the ability of PNPs to deliver nucleic acid payloads.289 During endosomal maturation, the pH of the endosome becomes extremely low. This presents a threat to nucleic acid stability, and the pKa of the polymer allows for buffering of the endosomal pH to a range that is more suitable for nucleic acids. A study by Du et al. studied a range of triblock copolymers with pKas ranging from 5.2 to 7.0. This study found that polymers with intermediate pKa values of 5.8–6.2 provide the best gene silencing via siRNA delivery and validated that this was due to more efficient endosomal escape.290 However, the exact pKa range that enables the most efficient endosomal escape is likely to depend on the carrier, as well as the tissue in which the nucleic acid is being delivered.291 In addition to pKa, other polymer properties have been altered to increase endosomal escape. Chiper et al. modified low MW PEI polymers to create polyplexes that dissolved at low endosomal pH, while buffering the pH within the endosome to protect nucleic acids from degradation.292 They went on to use this PEI PNP platform as a treatment to deliver nucleic acids to the muscle to treat Duchenne's muscular dystrophy in a mouse model.

One popular method for improving the efficiency of polymer nanocarriers is to increase the branching of the polymer. This is achieved by creating more end groups to the polymer, as opposed to the traditional linear polymer with only two end groups. Much work has been done to investigate the effects of branching of PEI and PBAE, with this structural modification also being studied in other polymer systems. Hyperbranched PBAE (HPBAE) has been studied at length for gene delivery in a range of cell and tissue systems. Li et al. found that removing the low MW (<15 kDa) components when formulating HPBAE PNP increased transfection efficiency.293 Li et al. also found that branch unit distribution, a structural property of HPBAE, plays an important role in transfection efficiency.294 A more uniform branch unit distribution of HPBAE resulted in a higher transfection efficiency. HPBAE PNPs have also been investigated and functionally tested as a carrier for siRNA gene silencing. Ooi et al. showed that HPBAE provides a better alternative to linear PBAE for siRNA delivery, utilizing a lower siRNA dose and lower polymer/siRNA ratio.295

Building upon this work, Liu et al. utilized the HPBAE polyplexes encapsulating minicircle DNA to establish feasibility of these polyplexes to deliver genetic material to the brain.296 The HPBAE was modified with disulfide bonds to promote rapid degradability, which did not negatively impact the cytotoxicity of the polyplexes. The minicircle DNA was delivered via polyplex to multipotent adipose‐derived stem cells and astrocytes, and high transfection efficiency was observed of >70% and >50% transfection efficiency, respectively. Rui et al. also demonstrated that HPBAEs can be modified for further optimization and expand its utility even beyond nucleic acid delivery.297 Utilizing a carboxylated PBAE system, they encapsulated a protein for cytosolic delivery and CRISPR/Cas9 for genetic engineering within the nucleus. They achieved over 75% gene knockout across various cell types, indicating that further modification of HPBAEs could lead to more effective protein delivery, which has proven difficult in the past. HPBAEs also show promise for large‐scale manufacturing. Zeng et al. found that lyophilized and frozen HPBAE particles still showed potent transfection abilities up to 1 year after formation.298

High throughput screening has proven essential in the design and testing of PNP as gene delivery vehicles. Ulkoski et al. developed a robotic‐assisted automated high throughput screening method to investigate several important characteristics of PNPs for delivery of mRNA including cytotoxicity, cellular uptake, transfection efficiency, nucleic acid encapsulation efficiency, and particle size.299 Through this screening, they uncovered structure–function relationships where copolymers with increasing charge density and optimal hydrophobicity by controlling alkyl side chain length led to enhancements in mRNA encapsulation efficiency. Rodrigues et al. evaluated various polymer NP formulations for effective mRNA delivery to fibroblasts.300 They tested 152 formulations using their high throughput screening method to determine their effectiveness for delivering a Cre recombinase mRNA in a mouse fibroblast Cre‐loxP reporter cell line. From these 152 formulations, they were able to rapidly eliminate cytotoxic compounds and compare the effectiveness of lead formulations to commercial controls (Lipofectamine) based on Cre‐mediated GFP expression. They identified a lead candidate which mediated significant mRNA delivery to fibroblasts in vitro based on its high buffering capacity to promote endosomal escape and cytosolic mRNA release. Using this lead PNP, their group went on to demonstrate significant mRNA delivery to activated skin fibroblasts following intradermal injection in tdTomato mice. Rui et al. developed a new high throughput screening assay for simultaneous determination of PBAE PNP cellular uptake and endosomal escape in vitro.301 This assay was highly predictive of mRNA delivery abilities of PBAE variations, and they were able to screen for top PNP formulations. Following identification of lead candidates, they found PBAE NP formulations with greater polymer backbone hydrophobicity mediated higher levels of mRNA‐mediated Cre expression as indicated by tdTomato expression in Ai9 mice following tail vein injection.

PEGylation strategies are often utilized with PNP to improve their stability in vivo and enhance circulation time (to be discussed further in Section 3.3.4). However, PEG‐coated PNP have been found to have lower cellular uptake efficiency in certain instances and these tradeoffs are important to consider when developing a PNP gene delivery system. Glodde et al. observed no difference in PEI‐PEG polyplex stability between polyplexes formed with 2 and 25 kDa PEI, but the stability of the polyplex with high PEI MW was greatly reduced with increased PEG grafting.302 Iqbal et al. found that PEGylating PBAE NPs did not significantly decrease their uptake in three different cell lines.303 PEGylated PBAE showed a higher DNA transfection capacity, as well as a lower cytotoxicity when compared to PEI. These results did not include in vivo work, which would be useful to capture the whole picture of systemic delivery of PEGylated NPs for gene delivery. Indeed, Grun et al. found that in vitro results did not translate in vivo when investigating the incorporation of PEG for PACE polyplexes.304 They were able to conclude that there was an optimal amount of PEG grafting to improve polyplex stability while not drastically decreasing transfection abilities, between 0.25% and 5% wt% PACE‐PEG, and the impact of PEGylation on transfection was also affected by the mode of delivery of the NPs. As determining the optimal PEG amount can vary due to a variety of factors, Dogan et al. developed an artificial neural network to model the transfection efficiency depending on PEG MW and concentration, as well as NP concentration.305

Polymers are easily modified, resulting in large numbers of potential PNPs for gene therapies with specific tropism to disease‐relevant organs in vivo. In particular, PBAE PNPs are notable for organ‐specific delivery to the liver, lung, brain, spleen, and other organs.306 While the structure is consistent across PBAE molecules, many of the other aspects of the structure can be altered such as hydrophobicity, charge density, and functional terminal end groups. Recently, Kim et al. created a library of PBAE polymers that combined different monomer variations and screened a library of PBAE PNPs for luciferase mRNA delivery efficiency following intramuscular injection in the hind limbs of mice.307 Through this screen, PBAE PNPs containing amphiphilic amines were identified which achieved up to 2 weeks of mRNA expression and limited off‐target delivery to the spleen or liver, significantly outperforming a conventional LNP formulation in both regards. Kaczmarek et al. utilized a library of 30 PBAEs formulated with different end‐capping groups, diacrylate:amine ratio, and alkylamine hydrocarbon length to investigate their impact on pulmonary delivery of mRNA and DNA following IV injection.308 Two PBAEs PNP formulations, incorporating only slight variations in the diacrylate‐amine backbone chemistry, were identified as the most efficacious for DNA and mRNA delivery in vivo illustrating how PBAE polymer chemistry dictates PNP gene delivery. Interestingly, overall luciferase expression was significantly higher for mRNA as compared to DNA independent of the specific PBAE PNP carrier utilized. Kavanaugh et al. recently determined the end‐capping of PBAE with diethylenetriamine enables PBAE PNPs to target the lung following systemic administration and with three repeated doses, achieved successful gene editing of ≥10% disease‐relevant cell types including bronchial and alveolar epithelial cells (Figure 6).309

In a similar effort to determine how the chemical composition and synthesis of HPBAEs impacts organ targeting, Yong et al. created a library of over 200 HPBAEs containing varying structures synthesized from different components.310 With this library, they efficiently screened NPs for their ability to target specific organs in vivo. They found that the most important factor in organ targeting with HPBAEs was the terminal amine structure. Based on this structure, PBAE PNPs containing ethylenediamine and morpholine terminal amines were able to target mRNA to the spleen or liver, respectively. PBAE PNPs containing methylpentane terminal amines showed a much broader activity and efficient mRNA delivery to the liver, spleen, and lungs. While the majority of investigation into polymer branching has been done with PBAE, the findings from studying HPBAE suggest that branching of other polymers used for gene delivery could be useful in increasing transfection efficiency and specific organ targeting.

Similarly to PBAE, the structure of PEI has been manipulated to investigate the impact of branching on transfection efficiency. However, branched PEI has been associated with higher levels of cytotoxicity than other branched polymer systems. This can be overcome using targeting moieties (see Section 4 for further details). Zou et al. successfully delivered DNA to the lungs using a linear PEI complex that was systemically administered via tail vein injection.311 These studies and many others paved the way for a large body of work utilizing PEI as a basis for gene therapy. For instance, Wong et al. furthered PEI research with their investigation of transfection efficiency in mouse neonates.312 Many genetic disorders are present at birth and providing gene therapies to the immature immune system of infants could reduce the immune response to these foreign materials. Similar to conventionally used adult mice, systemically delivered plasmid DNA carrying PEI complexes showed high levels of transfection in the liver in mouse pups, and transfection was also found in various other organs of interest, including the spleen, lungs, and brain, although to a lesser extent than in the liver.

PEG‐coated PNP are not as easily taken up by immune cells, helping the NPs to circulate in the bloodstream for longer periods of time and giving a higher chance for extravasation and delivery of their cargo to target sites.313 In addition, the PEG chains on the NP surface reduce protein adsorption, also increasing the stealth properties of PNP. PEG can be used to neutralize the absolute zeta potential of NPs, as it is a hydrophilic polymer with net zero charge. When PEG was added to polyion complexes, there was lower cytotoxicity.314 Notably, the coating of PEG on NPs has been found to alter the protein corona that forms on NPs, which impacts where particles ultimately end up.315 A study by Perry et al. investigated how PEG grafting density impacted biodistribution of polymeric nanogels.316 At low density, PEG forms a mushroom conformation on NP surfaces, and at high density forms a brush conformation. Interestingly, there did not appear to be a significant difference in blood retention and biodistribution between low‐ and high‐density PEG‐grafted NPs. In general, PEGylated NPs circulated in the blood for much longer than non‐PEGylated NPs, remaining for greater than 24 h as compared to non‐PEGylated particles, which were not detected in the blood after 3 h post‐administration. The biodistribution was also altered at 24 h, with nearly all non‐PEGylated particles accumulating in the liver. While a large majority of the PEGylated particles also accumulated in the liver, NPs were also observed in the blood, spleen, lung, and kidney at higher rates than non‐PEGylated NPs. Recent work from Kwak et al. has also shown densely PEGylated PBAE PNP are retained in circulation over a significantly longer timeframe than PBAE PNP without a PEG coating as well as conventional LNPs317 (Figure 7). Notably, ~30% of injected dose was retained in the bloodstream after 4 h using PEGylated PBAE PNP whereas only <1% remained for unPEGylated PBAE PNPs and the conventional LNPs. A study by Mosquiera et al. found similar results in mice, with PEGylated nanocapsules (NCs) showing higher blood circulation at 24 h than poloxamine‐coated NCs. The PEGylated NCs also exhibited less liver accumulation, a common site of clearance from the blood.318 We note the findings in this study and others will be dependent on the NP onto which the PEG is conjugated, as well as the MW of the PEG used.

Azzam et al. found that chitosan NPs have a natural propensity for accumulating in collagen.319 They delivered siRNA to the liver using chitosan NPs but found that modification of the NP with a peptide increased deposition within the cell. Ward et al. evaluated the effect of PLL MW on DNA delivery in mice and rats.320 Specifically, 20 and 211 kDa MW PLL were complexed with three different types of DNA. They then evaluated how PLL MW and DNA type impacted delivery to various organs after systemic administration. When compared to free DNA, all complexes with the 211 kDa PLL showed increased deposition in the carcass (the body mass after main organs were removed) and decreased deposition in the liver. As previously discussed, extrahepatic delivery proves to be a challenge for gene therapies. These results indicate that PLL could be a promising avenue for gene therapeutics particularly given its past testing in clinical trials. As previously noted, Shen et al. compared cationic and zwitterionic polymers for the delivery of mRNA to the spleen and lymph nodes.253 When evaluated in vitro in comparison to cationic NPs, the zwitterionic polymer NPs showed a 39,500‐fold increase in protein expression and selective delivery to the spleen and lymph nodes in vivo. These findings suggest that capitalizing on the potential of zwitterionic polymers could prove highly beneficial for gene delivery.