**1. Introduction**

Severe diseases including cancer are often associated with the aberrant or uncontrolled expression of pathologically relevant genes. The use of small interfering RNAs (siRNAs) for inducing gene knockdown through RNA interference (RNAi) has become a powerful and promising technology to specifically downregulate target genes by harnessing a highly conserved cellular protein machinery called RNA-induced silencing complex (RISC) [1,2]. Today, siRNAs for any target RNA sequence of interest can be easily designed and synthetically generated [3]. They are thus also able to target proteins which are otherwise undruggable by small molecules and offer new therapeutic strategies. This also includes the knockdown of non-protein encoding mRNAs such as "long noncoding RNAs" (lncRNA) which are crucial biological regulators [4]. However, siRNAs are large and negatively charged molecules, which impair their cellular delivery and uptake, and are very susceptible to nuclease degradation. To address these issues, siRNAs can be chemically modified, e.g., by introducing non-natural nucleotide analogues or the covalent coupling targeting moieties [5,6], and/or it is often required to formulate them into nanoparticles.

In the recent years, several lipid molecules [7–9], cationic polymers [10,11], inorganic nanoparticles and other systems have been investigated for nucleic acid delivery [12–14]. Cationic materials are able to form nano-sized complexes through electrostatic interactions and efficiently protect their payload. For liver-related diseases, suitable siRNA carriers have already been established and, more recently, the first siRNA-based therapeutics have been translated into the clinic. In 2018, Patisiran (ONPATTRO®) was approved by the FDA. Relying on a liposomal siRNA formulation, Patisiran downregulates transthyretin in the liver for the treatment of hereditary transthyretin-mediated amyloidosis (hATTR) [15]. The second siRNA therapeutic, Givosiran (GIVLAARI®), was approved one year later for targeting aminolevulinic acid synthase 1 (ALAS1) in the treatment of acute hepatic porphyrias (AHPs). Givosiran is an siRNA bearing extensive chemical modifications and it is coupled to GalNAc for the targeted delivery to hepatocytes [16,17]. Despite these impressive advances, most siRNA delivery systems are unable to efficiently deliver siRNAs into target sites other than the liver, thus showing serious limitations, for example, in anti-tumor therapies [18,19]. In this context, polymeric nanoparticles may be particularly promising due to their versatility based on the implementation of chemical modifications.

There are several rate-limiting steps which need to be addressed when using nanosized carriers for systemic application. On the extracellular level, nanocarriers bind serum-components which may lead to aggregation, decomposition and/or opsonization by phagocytic cells. Poor tissue penetration and low internalization rates, followed by insufficient endo-/lysosomal release and intracellular siRNA release into the correct compartment are further limiting steps on the cellular/intracellular level [20,21].

Cationic polymers are promising candidates for addressing these issues, and in particular polyethylenimines (PEIs) are widely explored in this regard. PEIs are commercially available in branched and linear structures over a wide range of molecular weights (0.8–100 kDa). Branched PEIs comprise of amines in every third position, at a 1:2:1 ratio of 1◦:2◦:3◦ amines. The linear form consists of only secondary and a very few primary amines [22–25]. Polypropylenimine dendrimers (PPI) are another class of commercially available cationic polymers. PPIs are highly branched molecules with well-defined structure and size. They are composed of a diaminobutane core molecule and propylene imine repeating units. PPI dendrimers contain only tertiary amines in the interior and primary amines at the surface (see Figure 1); the number of repeating units determines the so-called 'generation' of the dendrimer and the number of surface functionalities [26]. However, in contrast to (larger) pDNA molecules, PPI dendrimers have not been explored so extensively for delivering siRNAs [27]. Marked differences have been observed between the delivery of siRNA and pDNA when choosing the optimal PPI generation. The smaller G3 PPI dendrimer (1687 Da, 16 1◦ amines) has been identified as most efficient for pDNA complexation and gene expression [28]. For the delivery of the smaller and more rigid siRNA molecules or small oligonucleotides, the G4 PPI (3514 Da, 32 1◦ amines) showed the highest uptake rates and gene knockdown efficiencies, when compared G3 and G5 PPI dendrimers [29]. Still, another study used the PPI G3 for the siRNA mediated knockdown of the ubiquitin ligase ITCH in various pancreatic cancer cell lines, showing an ~ 50% downregulation only at high siRNA concentrations (5 μg/200 μL) [30].

When considering PEIs or PPIs for therapeutic nucleic acid delivery, there are major concerns impeding their clinical translation. While it is well known that the transfection efficacy of these polymers increases with the molecular weight or number of generations, this is usually associated in parallel with higher cytotoxicity. Furthermore, PEI and PPI complexes tend to aggregate over time in physiological buffers or upon contact with biological fluids, e.g., serum [28,31–33].

Due to the large content of reactive amines in PEIs and PPIs, their chemical modification has been extensively explored, for further improvement of PEI- and PPI-based delivery systems. A general approach towards reducing cytotoxicity of cationic polymers is the shielding of the strong cationic charge by PEGylation. Although this strategy improved the biocompatibility and colloidal stability of the complexes, their cellular uptake and transfection efficacies were decreased [34,35]. To address this issue, targeting moieties have been introduced for cell specific internalization of pDNA- and siRNA complexes. This was shown for the coupling of the luteinizing hormone-releasing hormone (LHRH) peptide or folate onto the distal end of G4 or G5 PPI dendrimers [36,37]. In another

approach, G4 PPI was modified with maltose for improved biocompatibility, and the loss of cellular uptake was compensated by coupling of an anti-EGFRvIII single chain antibody [38]. Other chemical modifications focused only on pDNA delivery. Another widely explored method for modifying cationic polymers is the introduction of lipophilic molecules. Here, biological activities were improved; for example after alkylcarboxylation of G4 PPI with 10-bromodeconoic acid or fluorination of G3-G5 PPIs with heptafluorobutyric acid [39,40]. Other studies explored PPI copolymers like PPI-poly-L-lysine, PPI-oligoethylenimine or PPI-Pluronic® P123 [33,41,42]. Smaller molecules were found to be efficient as well. The coupling of amino acids arginine, leucine or lysine strongly enhanced gene expression efficiencies of G3 PPI-based complexes and even of the very small G2 PPI [43,44]. Similarly, the modification of the larger G4/G5 PPIs with heterocyclic amines (histidine, piperazine-2-carboxyilic acid and 3-pyridyl acetic acid) reduced cytotoxicity, while leading to higher gene expression levels [45].

In stark contrast, chemical modifications of PPI-based systems with the aim of improving siRNA delivery have been barely explored at all. For creating a library, G1 PPI was modified with various alkyl epoxides, prior to formulation with PEG-lipids and siRNA to form nanoparticles. The most efficient derivative was a C16-modified G1 PPI that was able to downregulate Tie2 in endothelial cells in vitro and in vivo [46]. Another approach anchored G3 PPI onto gold nanoparticles. The Au-G3/siRNA complexes achieved a 70% knockdown of the target gene BCl-2, with negligible nonspecific toxicity [47]. Alternatively, an optimized formulation rather than a covalent modification employed the polyphenol epigallocatechin-3-O-gallate (EGCG) for pre-mixing with siRNA, prior to complexation with a G2 PPI. This ternary complex reduced luciferase reporter gene activity by 70% [48]. These findings identify PPI dendrimers as attractive candidates for further improvement. Moreover, PPI dendrimers are available at large scales and more defined structures compared to "normal" polymers, which is of particular importance when it comes to medical applications [49].

Recently, we selected a range of branched low molecular weight PEIs (2, 5, and 10 kDa) which were modified with tyrosine at the primary amines, yielding the tyrosine-grafted PEIs P2Y, P5Y and P10Y, respectively. This modification strongly enhanced siRNA-mediated knockdown efficiencies of different target genes in various cell lines at very low polymer/siRNA ratios, with a 70–90% target gene downregulation. Notably, even in the case of the smallest 2 kDa PEI, which is otherwise biologically inactive, tyrosine modification achieved knockdown efficacies of 50%. In addition, the colloidal stability of the tyrosine-modified PEI complexes upon incubation with serum was markedly improved [50–52]. More recently, we extended this strategy towards the modification of linear PEIs. Linear PEIs are more biocompatible than their branched counterparts and they can be produced with narrow polydispersities [53]. However, in their unmodified form, linear PEIs cannot sufficiently deliver siRNA into cells. We selected 2.5, 5, 10 and 25 kDa linear PEIs for tyrosine modification. Again, all new polymer derivatives, called LP2.5Y, LP5Y, LP10Y and LP25Y, showed excellent knockdown efficacies of 80–90% in various cell lines, including hard-to-transfect cells. The most active candidates, the linear LP10Y as well as the branched P5Y and P10Y, have also been successfully evaluated in tumor xenograft mouse models for siRNA-mediated oncogene targeting [54].

Thus, as shown previously, and in this paper, the tyrosine modification of branched or linear, small PEIs markedly improves their efficacy and biocompatibility. Based on these findings, we have now extended this approach for the first time towards a PPI dendrimer and present the superior properties of a tyrosine-modified fourth generation PPI ("PPI-G4-Y"). This covers the physicochemical characterization of the PPI-G4-Y/siRNA complexes which show high complexation efficacy and complex stability, and little surface charge. High biological activity and low cytotoxicity were found and compared to various tyrosine-modified branched and linear PEI complexes in different cell lines. In combination with first in vivo biocompatibility studies, our data identify PPI-G4-Y dendrimers as versatile and particularly promising siRNA delivery platform.
