*2.2. In Silico*/*Experimental Interaction of Single Tail Self-Assembling Dendrons and siRNA*

Mesoscale simulations were again exploited for predicting the interactions between the amphiphilic dendrons **1**–**8** and siRNA molecules, as illustrated in Figure 3.

**Figure 3.** *Cont*.

**Figure 3.** Morphologies of the systems formed at a dendrimer-to-siRNA charge ratio (N/P) = 10 by the siRNA molecules and the amphiphilic dendrons **1** (**a**), **2** (**b**), **3** (**c**), **4** (**d**), **5** (**e**), **6** (**f**), **7** (**g**) and **8** (**h**) as obtained from molecular simulations. Left panels are zoomed images of the right panels. Colors as in Figure 2, except for the siRNA molecules, shown as red sticks and the water and counterions, portrayed in the left panels as light blue field for clarity. Adapted from [7] with the permission of John Wiley and Sons, 2016.

The computer images showed that, at the same dendrimer-to-siRNA charge ratio (N/P) of 10, the nucleic acid fragments are well encased within the micellar network formed by all self-assembling amphiphilic dendrimers **1**–**6** and so are efficiently protected from the surrounding environment. For the non-self-assembling systems **7** and **8**, the siRNA molecules are interspersed within an unorganized solution of single dendrons and water, and therefore in principle more susceptible to ultimate degradation by e.g., RNAses.

The in silico quantitative characterization of the siRNA/amphiphilic dendron interactions was next carried out by atomistic molecular dynamics (MD) simulations performed in the framework of the so-called molecular mechanics/Poisson-Boltzmann surface area (MM/PBSA) methodology [9–15] (the methodology is described in detail in the Supporting Information in the companion paper [9]). The values of free energy of binding (ΔGbind) between the different self-assembling dendrons **1**–**6** towards the siRNA sequence directed against the mRNA coding for the heat shock protein 27 (Hsp27)—a small molecular chaperone which is a vital regulator of cell survival and a major player in drug resistance—were initially calculated, as listed in the first column of Table 2. However, in order to perform a rigorous discussion about these aspects, the concept of effective free energy of binding (ΔGbind,eff) was introduced, as follows. From an extensive analysis of each dendron micelle/siRNA complex MD trajectory, the number Neff of dendron branches (each bearing a positive charge) in permanent contact with the siRNA was determined, leading to the values listed in the second columns of Table 2. The precise identification of each dendrimer branch involved in the siRNA binding, and the corresponding individual contribution afforded to the overall binding energy, was then carried out by a per-residue free energy decomposition technique (described in the Supporting Information in the companion paper [9]). The cumulative results of this analysis are reported in the third column of Table 2, showing the effective contribution to the total free energy of binding (ΔGbind,eff). However, to be able to compare simulation data among themselves (and with experimental data, see below) the effective-charge-normalized values of ΔGbind,eff, i.e., ΔGbind,eff/Neff, were considered, as shown in the fifth column of Table 2.

**Table 2.** The free energy of micellization (ΔGbind, kcal/mol), number of effective charges (Neff), effective free energy of binding (ΔGbind,eff, kcal/mol), and effective-charge-normalized free energy of binding (ΔGbind,eff/Neff) between siRNA and dendrons **1**–**6** as derived from molecular simulations. The last column reports the siRNA/dendron binding data (C50, μM) obtained from ethidium bromide (EB) displacement assays. Adapted from [7] with the permission of John Wiley and Sons, 2016.


From the values reported in Table 2, the effective interaction between the micelles formed by the different dendrons with siRNA (ΔGbind,eff) increases with increasing alkyl chain length, the last three dendrons definitely being the strongest siRNA binders in the order: **4** < **5** < **6**. Limiting the discussion to these three self-assembling systems, the micelles formed by dendron **4** were characterized by a total number of charges equal to +64, the 48% of which (Neff = 31 are involved in productive siRNA binding. This translates into the corresponding ΔGbind,eff/Neff value of −0.41 kcal/mol. In contrast, the other two efficient self-assembled micelles generated by dendrons **5** and **6** were not only both able to exploit 61% of their positive charges (49/80 and 62/102 for **5** and **6**, respectively) but they also did it more efficiently (ΔGbind,eff/Neff = −0.59 kcal/mol for **5** and −0.56 kcal/mol for **6**, Table 1), ultimately characterizing the corresponding micelles as the best siRNA binders of the whole series.

<sup>ǂ</sup> <sup>ǂ</sup>

In silico data were experimentally confirmed by fluorescent ethidium bromide (EB) displacement assays (see Figure A1 in Appendix A). According to these tests, the value of C50 (i.e., the dendron concentration at which the fluorescence of siRNA-intercalated EB is reduced by 50%), is an indication of siRNA binding efficiency: The lower the C50 value, the stronger the siRNA/micelle binding. As seen in Figure A1, the negative-control (i.e., non-self-assembling) dendrons **7** and **8** were not able to displace EB from the nucleic acid. In line, also for the self-assembling dendron **1**, bearing the shortest alkyl tail (C14), the C50 value was not determined, due to the high CMC value required for its self-assembly. On the other hand, for the dendron series **2**–**6**, progressively decreasing C50 values were obtained (Figure A1 and Table 2, last column), with dendron **6** being the most effective siRNA binder, in full agreement with computer-based predictions.

#### *2.3. In Vitro siRNA Delivery Performance of Single-Tail Self-Assembling Dendrons*

Based on the results described above, dendrons **4**–**6** were selected for in vitro delivery of siRNA molecules targeting both the Hsp27 and the translationally controlled tumor protein (TCTP, a highly conserved protein present in all eukaryotic organisms that regulate cell survival in human tumors) in human castration-resistant prostate cancer (HCPC) PC-3 and C4-2 cell lines (Figure 4a,b). In these experiments, all other dendrons (**1**, **3**, **7** and **8**) were also tested for comparison and/or negative controls.

**Figure 4.** *Cont*.

**Figure 4.** Self-assembled dendron-mediated siRNA delivery and gene silencing of Hsp27 (**a**) and TCTP proteins (**b**) in human prostate cancer PC-3 (solid bars) and C4-2 (patterned bars) cell lines (N/P = 10). Control: Untreated cells. The results obtained from non-self-assembling dendrons **1**, **7** and **8** are also shown for negative-control purposes. (**c**) Expression of Hsp27 in PC-3 cells treated with 50 nM siRNA delivered by self-assembled dendron **4** (N/P = 10), by Lipofectamine, and by the high-generation covalent triethanolamine (TEA)-core PAMAM-based dendrimer G7. Control: Untreated cells. Negative controls: Self-assembled dendron **4** alone and a scrambled (i.e., non-silencing) siRNA sequence. The expression levels of Hsp27 and TCTP proteins were quantified by western blots 72 h post treatment using vinculin as the control. Adapted from [7] with the permission of John Wiley and Sons, 2016.

As shown in this Figure, the best siRNA delivery capacity was obtained with the nanovector based on the self-assembled form of dendron **4**, whose related gene silencing effect was even superior to that obtained both with the gold standard commercial vector Lipofectamine and with the high-generation (G7) covalent PAMAM-based dendrimer. This featured an extended triethanolamine (TEA) core developed by our group and discussed in detail in the companion paper [9]. As anticipated by in silico predictions, nanocarriers formed by dendrons bearing shorter aliphatic (**1**–**3**), hydrophilic (**7**) or no chains (**8**) fail to elicit meaningful gene silencing in all cases. Indeed, while **7** and **8** are intrinsically unable to self-assembly (negative controls), the former dendron set (**1**–**3**) likely cannot generate stable micelles and/or siRNA complexes at the experimental conditions employed for transfection, due to their high CMC and C50 values (Tables 1 and 2). On the other hand, nanomicelles from dendrons with longer alkyl tails (**5** and **6**) were also able to achieve significant siRNA delivery and silenced the target Hsp27 and TCTP genes, yet with performance lower than that obtained with dendron **4**.

These last results seem to be in slight contradiction with the computer-based prediction reported in Table 2, according to which the best (i.e., most favorable) values of ΔGbind,eff/Neff and C50 are indeed obtained for self-assembled dendrons **5** and **6**. This ranks these nanovectors as the tightest siRNA binders. To investigate whether this disagreement was real or only apparent, cellular uptake experiments based on flow cytometry were performed using siRNA in complex with micelles from **4**, **5**, and **6** (Figure 5a). These data clearly show that this step is not governing the differential behavior of these nanovector/siRNA performance along the in vitro gene silencing pathway, as all three systems are characterized by comparable cellular uptake efficiency.

**Figure 5.** (**a**) The mean fluorescence intensity of PC-3 cells treated with siRNA labeled with the fluorescent dye Cy-3 in complex with self-assembled nanovectors **4**, **5**, and **6** (50 nM siRNA, N/P = 10). Non-treated cells and cells treated with labeled siRNA alone were used as control. (**b**) siRNA release from their complexes with micelles of dendrons **4** (red), **5** (black), and **6** (yellow) as assessed using EB displacement assay performed in the presence of heparin. Adapted from [7] with the permission of John Wiley and Sons, 2016.

With this aspect ruled out, it was reasoned that the siRNA release of its nanovector could be the factor underlying the best performance in gene silencing of the **4**/siRNA complex. Thus, it was determined that it was the dissociation of the nucleic acid fragments from their **4**, **5**, and **6** self-assembled nanovectors via the EB displacement assay in the presence of heparin, a highly negatively charged polysaccharide that can compete with siRNA in nanocarrier binding. As can be seen in Figure 5b, by increasing heparin concentration the siRNA molecules were more efficiently displaced from their complex with **4** than from the two alternative dendron micelles. This led us to the conclusion that the siRNA release process might be more effective from the self-assembled dendron **4** than from the **5** and **6** complexes, respectively.

Collectively, these results led to the conclusion that, in agreement with computer predictions, the supermolecular assemblies formed by **5** and **6** were actually too stable to dissemble in physiological conditions. This enhanced stability negatively affects the siRNA release in the cellular environment, thereby impairing the relevant gene silencing effect. At the same time, the complexes of micelles originated by dendron **4** and siRNA represent the best compromise in term of nucleic acid fragment binding and release efficiency, and these properties ultimately translate into the most potent biological performance.

The final step of the in vitro characterization concerned the investigation of the cellular toxicity of our single-tail self-assembled dendrons. As shown in Figure 6 for the PC-3 cell line as an example, none of the self- and non-self-assembling dendrons were endowed with cytotoxic properties. Analogous results were obtained with the alternative prostate cancer cell line, C4-2.

**Figure 6.** Metabolite toxicity (**a**) and membrane damaging activity (**b**) of self- and non-self-assembling dendrons **1**–**8** and their complexes with a scrambled (i.e., non-silencing) siRNA sequence (50 nM siRNA, N/P = 10) in PC-3 cells using the 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl tetrazolium bromide (MTT) assay and the lactate dehydrogenase (LDH) assay, respectively. Untreated cells were used for control. Adapted from [7] with the permission of John Wiley and Sons, 2016.

In essence, all in silico and in vitro results concurred to indicate **4** as the amphiphilic dendron endowed with the best features for efficient self-assembly, and easy and stable formation of complexes with the siRNA molecules for delivery. Once inside the cell, there is effective disassembly during endosomal release for subsequent potent gene silencing.

#### *2.4. In Vivo siRNA Delivery Performance of Single-Tail Self-Assembling Dendron 4*

The in vitro best performing nanomicelles generated by the dendron **4** were finally challenged for in vivo gene silencing in a prostate cancer xenografted nude mouse model. The results obtained are gathered in Figure 7.

**Figure 7.** (**a**) In vivo gene silencing achieved by treating nude mice bearing prostate cancer (PC-3) tumors (30–50 mm3) with Hsp27 siRNA delivered by the self-assembling dendron **4**, buffer solution (control), the dendron **4** alone, and a scrambled siRNA/**4** complex (negative controls). Treatments (3 mg/kg siRNA, N/P = 5) were administered for a period of 4 weeks via intraperitoneal injections (twice per week). (**b**) In vivo inhibition of tumor growth quantified by tumor size and (**c**) antiproliferation activity in tumors assessed by immunohistochemistry in mice treated as described for panel (**a**). Panel (**c**), from left to right: control, **4** alone, **4**/siRNA, **4**/scrambled siRNA. Adapted from [7] with the permission of John Wiley and Sons, 2016.

Images in Figure 7 support the effective and specific in vivo gene silencing induced upon administration of the Hsp27-directed siRNA delivered by the self-assembled nanovector **4**. In particular, an impressive reduction of ~80% in Hsp27 expression and of ~65% in tumor size after 4 weeks of treatment were achieved (Figure 7a,b).

These data, coupled with no discernable signs of in vivo toxicity (Figure 8) and the absence of mice weight loss, support the potential utilization of the amphiphilic compound **4** as a siRNA vector for future therapeutic implementations.

(**g**)

**Figure 8.** In vivo levels of inflammatory cytokines interleukine-6 (IL-6, **a**), interferon-gamma (INF-γ, **b**), and tumor necrosis factor-alpha (TNF-α, **c**), observed in male C57BL/6 mice treated with **4** and its siRNA complex (3.0 mg/kg siRNA, N/P = 5) by intravenous injection. Non-treated mice and animals treated with buffer solutions were used for control. Mice were sacrificed 24 h after treatment and the ELISA assay was performed on the collected serum. (**d**) Hepatic enzyme levels (aspartic aminotransferase AST (solid bars), alanine transferase ALT (patterned bars), gamma-glutamine transferase γ-GLT (empty bars), (**e**) kidney functions (creatinine CRE (mmol/L, solid bars), blood urea nitrogen BUN (μmol/L, patterned bars), and (**f**) cholesterol level in the blood measured in the treated mice after sacrifice. (**g**) Pathology of major organs from treated mice (HES staining). Columns (from left to right): Control, **4**, **4**/siRNA, and **4**/scrambled siRNA. Rows (from top to bottom): Heart, kidney, liver, lung, spleen. Adapted from [7] with the permission of John Wiley and Sons, 2016.

#### **3. siRNA Delivery by Double-Tail Self-Assembling Amphiphilic Dendrons**

#### *3.1. Design, Optimization and Chemico-Physical Charactrization of Double-Tail Self-Assembling Amphiphilic Dendrons*

With the haste of further improving the siRNA delivery and related gene silencing performance of the lead self-assembling amphiphilic dendron **4**, it was questioned whether increasing its hydrophobic portion would render this molecule even more efficient and effective. According to the devised strategy, it was ultimately decided to decorate **4** with a second C18-long alkyl chain. In order to do so, a series of computer-based molecular design and optimization studies were carried out, finally leading to the new double-tail amphiphilic dendron structure **AD** (where **AD** stands for Amphiphilic Dendron) shown in Figure 9a [6]. In buffer solution per se, molecular simulations predicted **AD** to self-assemble into large (~200 nm) vesicle-like structures (Figure 9b) called dendrimersomes [16]. After synthesis of **AD** by click-chemistry, dynamic light scattering (DLS) and transmission electron microscopy (TEM) indeed confirmed both the size and the shape of the supermolecular entities formed by **AD** upon self-assembly (Figure 9c). Notably, the vesicle corona structure closely matched a bilayer, characterized by a thickness of 7 nm—approximately equal to twice the molecular length of **AD** (~3.5 nm).

**Figure 9.** (**a**) The structure of the double-tail amphiphilic dendron simulated (**b**) and experimental TEM images (**c**) of the vesicle-like dendrimersomes formed by **AD** alone upon self-assembling in buffer solution. In panel b, the hydrophilic units of **AD** are portrayed in light green, the hydrophobic units are depicted in dark green, while light grey spheres are used to show some representative water molecules. Simulated (**d**) and experimental TEM images (**e**) of the nanosized spherical micelles formed by **AD** in the presence of siRNA molecules in solution. In panel (**d**), colors as in panel (**b**), while the nucleic acid fragments are portrayed as orange sticks. Adapted from [6] with permission of John Wiley and Sons, 2014.

Quite surprisingly, however, when simulations of the self-assembly of **AD** were performed in the presence of siRNA many highly ordered smaller spherical micelles of 6–8 nm diameter were obtained, with the siRNA fragments nicely interspersed within them (Figure 9d). The explanation for this vesicular to micellar structural transition likely resides in the greater positively charged surface area exposed by the micelles with respect to the dendrimersomes, which provides more efficient electrostatic interactions with the negatively charged siRNA. This, in turn, leads to the formation of more stable and compact complexes between the self-assembled nanocarriers and the nucleic acids molecules. The computer predictions were supported by TEM imaging (Figure 9e), confirming that **AD** is able to dynamically self-assemble into responsive and adaptive supramolecular assemblies in the presence of external stimuli, i.e., negatively-charged small polyanions.

#### *3.2. In Vitro siRNA Delivery Performance of the Double-Tail Self-Assembling Dendron AD*

The nanosized micelles formed by self-assembly of the double-tail amphiphilic dendron **AD** were able to effectively protect siRNA from degradation (e.g., by RNAses) and to promote fast and quantitative uptake by castrate-resistant prostate cancer PC-3 cells—almost 100% internalization of the nanovector/siRNA complexes was reached within 30 min. As concerns the mechanism of cell internalization of the **AD**/siRNA complexes, macropinocytosis was found to be the leading cellular entry mechanism for this nanovector/siRNA system, in utter analogy with the G5 covalent PAMAM dendrimer counterpart discussed in the companion paper (Figure 15a in [9]). Indeed, neither chlorpromazine (an inhibitor of clathrin-mediated endocytosis) nor genistein (a caveolae-mediated uptake specific blocker) were effective while cytochalasin D (a specific macropinocytosis inhibitor that, by binding to actin filaments, interferes with actin polymerization and assembly) inhibited cellular uptake in a dose-dependent manner, as presented in Figure 10b. Again in analogy with what observed for the G5 TEA-core PAMAM covalent dendrimer (Figure 15b in [9]), confocal microscopy provided images showing significant co-localization of the siRNA/**AD** complexes with the macropinocytosis biomarker dextran. Whereas, negligible co-localization was observed in cells treated with the two alternative biomarkers transferrin (clathrin-mediated endocytosis) and CTX-B (caveolae-mediated endocytosis). Thus, these experiments further corroborate the concept that these self-assembling PAMAM-based amphiphilic dendrons behave very similarly to their covalent high generation counterpart.

The gene silencing effect induced by Hsp-27 targeting siRNA delivered by **AD** nanomicelles was next investigated both at the mRNA and protein levels using prostate (PC-3) and breast (MCF-7 and MDA-MB231) cancer cell lines. As seen in Figure 10, a remarkable attenuation of the Hsp27 expression was achieved in all cases. This effect was comparable, if not greater, than that achieved with Lipofectamine, the reference standard reagent for siRNA/DNA transfection (e.g., **AD** was 6 times more efficient than Lipofectamine in downregulating Hsp27 protein expression in PC-3 cells, Figure 10b).

**Figure 10.** *Cont*.

**Figure 10.** Downregulation of Hsp27 at the mRNA (**a**) and protein (b) level in prostate cancer PC-3 cells treated with siRNA (50 nM) delivered by the self-assembled double-tail dendron **AD** (N/P = 5). Inhibition of Hsp27 expression by siRNA/**AD** complexes (50 nM siRNA, N/P = 5) in breast cancer MCF-7 (**c**) and MDA-MB231 (**d**). Adapted from [6] with permission of John Wiley and Sons, 2014.

The knockdown of Hsp27 in these cancer cells by siRNA delivered by **AD** micelles was paralleled by significant anti-proliferative effects (~75% of cancer cell growth inhibition), in agreement with the evidence that Hsp27 silencing negatively affects PC-3 cell survival [9,17]. Similarly to what was observed for the high generation covalent PAMAM dendrimers [9], the mechanism leading to cell death relied on caspase-dependent induced apoptosis, that is, the programmed cell death was promoted by apoptotic caspases, a family of endoproteases that provide critical links in cell regulatory networks controlling cell death. Accordingly, a 3-fold increase in caspase-3/7 activation was observed on average, confirming that siRNA delivered by the nanovectors based on self-assembled **AD** is very effective in silencing Hsp27 and thereby inducing caspase-dependent anticancer activity in vitro. A remarkable additional feature of this biological outcome is that the silencing effect was completely maintained for one week, and was also effective when transfection was performed in the presence of serum, a prototypical Achilles' heel of cationic nanovectors in siRNA delivery. In fact, the negatively charged serum proteins can not only compete with siRNA binding to the nanocarrier, but also may, by virtue of strong electrostatic forces, lead to the the disintegration of the nanoassemblies with consequent premature nucleic acid release and degradation [9].

From another perspective, when confronted with its single-tail analogue **4,** the efficiency of in vitro silencing induced by the double-tailed **AD** dendron/siRNA complexes was practically identical (e.g., 97% for **AD** and 96% for **4** in PC-3 cells, Figure 10b,c, respectively). Most importantly, however, remarkably good results were obtained with **AD** when challenged against cancer stem cells (notoriously the cancer cell population most refractory to treatments), and also stem and primary cells that constitute the target of deadly viruses such as HIV-1. As shown in Figure 11, **AD** was indeed able to elicit significant RNAi interference (RNAi) in glioblastoma stem cells (PBT003), and in HIV-1 infected human primary peripheral blood mononuclear cells (PBMC-CD4+), and hematopoietic stem cells (HSC-CD34+). Specifically, upon treatment of the hardly tractable PBT003 cells with **AD** carrying the siRNA was directed against the signal transducer and activator of transcription 3 (STAT3), which is a key player protein in glioma-initiating cells, thought to be responsible for glioblastoma induction, progression and recurrence. A 35% reduction in the corresponding mRNA level was observed, while only 24% mRNA reduction was achieved upon transfection with Lipofectamine (Figure 11a). The data relative to the two HIV-infected cell lines (PBMC-CD4<sup>+</sup> and HSC-CD34+) transfected with the siRNA targeting the HIV-1 Tat/Rev gene are more striking. The Tat and Rev HIV-1 proteins are essential positive regulators of gene expression through interaction with RNA target elements present within the 5 translated leader sequence and envelope gene, respectively. Importantly, the genes encode the Rev and Tat overlap, with each being produced from a different reading frame. As seen from Figure 11b,c, in both cases a remarkable 50% reduction in the Tat/Rev mRNA expression was observed, while the prototypical delivery vector Lipofectamine (aka RNAiMAX) failed to induce any gene silencing in these cells

infected by the acquired immunodeficiency syndrome (AIDS) virus. Fundamentally, this translated into a 50% reduction of the viral infection (Figure 11d,e) and, to the best of the authors' knowledge, constitutes the first report on the successful and safe delivery of siRNA into primary and stem cells.

**Figure 11.** (**a**) STAT3 knockdown in glioblastoma stem cells (PBT003) treated with siRNA/**AD** complexes (50 nM siRNA, N/P = 5). The knockdown of Tat/Rev in primary human peripheral blood mononuclear cells (PBMC-CD4+) (**b**) and hematopoietic CD34<sup>+</sup> stem cells (HSC-CD4+) (**c**) treated as in (**a**). Effective inhibition of HIV replication in PBMC-CD4<sup>+</sup> (**d**) and HSC-CD4<sup>+</sup> (**e**) after treatment with the anti-Tat/Rev siRNA delivered by **AD** (50 nM siRNA, N/P = 5). Viral loading was assessed using HIV-1 p24 antigen ELISA 3 days post-treatment. PBMC-CD4<sup>+</sup> cells and HSC-CD34<sup>+</sup> cells were infected by NL4-3 virus at a multiplicity of infection (MOI) = 0.001 and by JR-FL virus (MOI = 0.005), respectively, for 5 days before transfection. (**f**) Proton-sponge effect in PC-3 cells transfected with siRNA/**AD** in the absence (solid bars) and presence (patterned bars) of the proton pump inhibitor Bafilomycin A1. Untreated cells were used as the control. Adapted from [6] with permission of John Wiley and Sons, 2014.

As for the high generation covalent PAMAM dendrimer analogues presented in the companion paper [9], it was reasoned that also in the case of the **AD** micelles, the so-called proton sponge effect [18] could be invoked to explain nucleic acid release from the self-assembled double-tail nanovectors and, therefore, potent siRNA delivery within the cell cytoplasm. According to the debated proton sponge concept, when cationic nanoparticles such as those formed by **AD** and its siRNA cargo enter acidic endosomal (and lysosomal) vesicles, unsaturated amino groups sequester protons supplied by the proton pump v-ATPase. These sequestered protons cause the pump to continue functioning, leading to the retention of chloride ions and water molecules. Eventually, osmotic swelling causes the rupture of the vesicle, allowing the cationic nanoparticles to enter the cytoplasm. As the **AD** amphiphiles feature tertiary amines in the interior of their PAMAM heads which become protonated at the endosomal acidic pH (5.5), the authors verified whether the proton sponge effect was effectively at play also during their mediated siRNA delivery. Indeed, as illustrated in Figure 11f, Hsp27 silencing in PC-3 cells was significantly reduced in the presence of Bafilomycin A1 (a proton pump inhibitor that prevents endosome acidification). This implies that the **AD**-mediated siRNA delivery was indeed dependent on the endosomal acidification process and that the proton sponge effect played a role in the endosomal escape and cytoplasmic release of the nucleic acid.

Finally, the same protocol adopted for investigating the cellular toxicity of the single-tail self-assembled dendrons described in Section 2.3 was applied to **AD**. As expected, no discernible toxicity was observed in all cell lines (PC-3, PBMC-CD4+, HSC-CD34+, and PBT003 transfected with 50 nM siRNA and **AD** at N/P = 5 and 10) using both the MTT and LDH assays, supporting further in vivo gene silencing experiments with this nanocarrier.
