Non-Cationic RGD-Containing Protein Nanocarrier for Tumor-Targeted siRNA Delivery

Despite the recent successes in siRNA therapeutics, targeted delivery beyond the liver remains the major hurdle for the widespread application of siRNA in vivo. Current cationic liposome or polymer-based delivery agents are restricted to the liver and suffer from off-target effects, poor clearance, low serum stability, and high toxicity. In this study, we genetically engineered a non-cationic non-viral tumor-targeted universal siRNA nanocarrier (MW 26 KDa). This protein nanocarrier consists of three function domains: a dsRNA binding domain (dsRBD) (from human protein kinase R) for any siRNA binding, 18-histidine for endosome escape, and two RGD peptides at the N- and C-termini for targeting tumor and tumor neovasculature. We showed that cloned dual-RGD-dsRBD-18his (dual-RGD) protein protects siRNA against RNases, induces effective siRNA endosomal escape, specifically targets integrin αvβ3 expressing cells in vitro, and homes siRNA to tumors in vivo. The delivered siRNA leads to target gene knockdown in the cell lines and tumor xenografts with low toxicity. This multifunctional and biomimetic siRNA carrier is biodegradable, has low toxicity, is suitable for mass production by fermentation, and is serum stable, holding great potential to provide a widely applicable siRNA carrier for tumor-targeted siRNA delivery.


Introduction
siRNA has emerged as an invaluable tool for studying gene functions and developing treatments for intractable diseases, such as cancer [1][2][3], viral infection [4,5], and genetic disorders [6,7]. Generally, siRNA mediates posttranscriptional gene silencing by targeting and cleaving complementary mRNA [8][9][10]. The advantage of siRNA over small chemical drugs is that siRNA sequences can be rapidly designed for highly specific inhibition of the target protein expression [11]. siRNA also has advantages over antisense oligonucleotides. In one head-to-head comparison, siRNA knockdown of gene expression was about 100-to 1000-fold more efficient than antisense oligonucleotides (ODNs) [12]. In 2018, the FDA approved the first siRNA drug (patisiran), and in 2019, they approved the second siRNA drug (Givlaari) [13], which suggests that siRNA drugs are emerging as the third class of medicine after chemical drugs and antibodies for treating diseases by targeting the root cause. However, most siRNA drugs so far have focused on targeting the liver. Therefore, targeting delivery systems that can deliver siRNA to various target tissues other than the liver are needed.
To successfully deliver RNA, vectors should be designed to overcome barriers such as toxicity, off-target effects, endosomal entrapment, and ease of mass production. However,

Construction of Fusion Protein Dual-RGD (Dual-RGD-dsRBD-18his) Expression Plasmid
The dsRBD-18His-containing plasmid was used as a PCR template. For amplification by PCR, the following primers were used: forward, 5 -AAAGGATCC ATG TGT GAT  TGT CGT GGA GAT TGT TTC TGT GGTAGT GCTGGTGAT CTTTCAGCAG-3 ; reverse, 5 -TTCTCGAGGCTGTCTCCACGGCCGTGGTGGTGGTGGTGGTGCACTGAGGTTTCTTCT-3 . BamH1 and Xho1 recognition sites were introduced in the primers and bolded. The PCR product and PET28a plasmid were digested with BamH1 and Xho1. After digestion, the PCR fragment was inserted into PET28a by T4 DNA ligase. Ligates were transformed into E. coli BL21 (DE3) competent cells. The positive clones were confirmed by gene sequencing.

Expression and Purification of Dual-RGD Recombinant Protein
Positive colonies were selected and grown at 37 • C in LB medium containing kanamycin (30 µg/mL). When the OD 600 reached 0.4-0.6, protein expression was induced by isopropylβ-D-thiogalactopyranoside (IPTG, 1 mM) for 4 h at room temperature. Cells were harvested by centrifugation at 10,000× g for 10 min, and the pellet was stored at −20 • C. To isolate the soluble proteins, the cell pellet was lysed by the BugBuster mix in the presence of a protease inhibitor EDTA-free cocktail. The cell lysate was centrifuged at 20,000× g for 20 min at 4 • C to remove insoluble cell debris. The soluble extracts were loaded onto a nickel resin column and eluted with His•Bind ® buffer. The eluted protein was dialyzed against PBS containing 10% glycerol and 0.1% (v/v) β-ME for 24 h. Protein concentrations were determined with the Bio-Rad Protein Assay using BSA as the standard. Purified protein was probed using SDS-PAGE.

Measurement of Hydrodynamic Diameter and Charge
The hydrodynamic diameter and zeta potentials of purified monomer dual-RGD protein (0.1 mg/mL) in PBS buffer was detected by a Zetasizer Nano ZS analyzer (Malvern, Worcestershire, UK).

SiRNA Binding Assay
To assess siRNA binding capability of the dual-RGD carrier, a gel retardation assay was performed. EGFR siRNA was labeled with FAM fluorophore at the 5 end of the sense strand. FAM-siRNA was incubated with dual-RGD at different protein/siRNA molar ratios (1:1, 2:1, a d 4:1) for 30 min at room temperature. Bound and unbound siRNA was quantified using 1% agarose gel electrophoresis. Images first were captured under a UV transilluminator. Then, the same gel was stained with Coomassie blue to reveal the protein location, and the stained gel was imaged again under bright light.

RNase Resistance and Serum Stability
siRNA was labeled with FAM at the 5 terminus of the sense strand. Dual-RGD/siRNA complex or siRNA only was spiked with 50% (v/v) fresh pooled human serum at different times (1 h, 4 h, and 24 h); 1% agarose gel electrophoresis was performed to monitor the dissociation.

Western Blot
Cell lysates or tumor tissues were lysed in lysis buffer (M-PER Mammalian Protein Extraction Reagent, Thermo Fisher Scientific) containing 1× Halt Protease Inhibitor Cocktails. The cell lysates were kept on ice for 40 min, vortexed 3 times, and centrifuged at 12,000× g for 10 min at 4 • C. The concentration of supernatant protein was determined with the Bio-Rad Protein Assay. Samples were separated on 8% SDS-PAGE and transferred to the PVDF membrane. The membranes were blocked with 5% non-fat milk and then incubated with primary antibodies for 2 h at room temperature, followed by incubation with quantum dot-conjugated secondary antibodies for 2 h at room temperature. The blots were imaged under a UV table. Western blot was quantified using ImageJ version 10.2 (NIH).

Cell-Type Specific Binding Assay
Cell lines with different integrin α v β 3 expression levels were grown to 80% confluence and trypsinized from the dishes. The collected cells were fixed in 10% buffered formalin for 30 min. The siRNA sense strand was labeled with fluorophore FAM or Cy5 at the 5 end. FAM-siRNA (5 nmoles of siRNA) was loaded into dual-RGD at a molar ratio of 1:1. The complex of dual-RGD/FAM-siRNA was incubated with fixed cell lines at 37 • C for 1 h, and the binding specificity of dual-RGD was detected by flow cytometry (BD FACSCalibur Cell Analyzer). In the competitive assay, the RGD peptide (10 nmoles) was treated cells for 1 h before adding the complex of dual-RGD/Cy5-siRNA.

Tumor-Targeting and Biodistribution
Five-week-old female athymic nu/nu mice were injected with MDA-MB-231 cells (5 × 10 6 ) mixed with Matrigel (v/v 1:1) (Corning, NY, USA) subcutaneously in their backs. After 4 weeks, tumor-bearing mice were tail-vein injected with 100 µL of Cy5-EGFR siRNA/dual-RGD complex (5 nmoles) or equal moles of Cy5-siRNA/dsRBD-18His complex. At time points of 3 h, 6 h, 12 h, and 24 h, the Cy5 fluorescence signals of mice were detected with the Xenogen IVIS100 imaging system (Caliper Life Sciences, Waltham, MA, USA) to monitor the siRNA distribution in the whole body. At 12 h and 24 h, major organs were removed and homogenized in buffer (10 mM Tris pH 7.4 and 0.5% Triton X-100) with a mortar and pestle at a ratio of 100 mg of tissue per mL of buffer. Subsequently, 100 µL of tissue homogenate was loaded into a 96-well plate. The plate was measured by a Tecan Infinite F200 Pro Microplate Reader (Tecan Company, Mannedorf, Switzerland).

In Vivo Gene Knockdown
MDA-MB-231 tumor-bearing mice (n = 5 per group) were iv treated with saline, EGFR siRNA only, EGFR siRNA/dsRBD-18His, or EGFR siRNA/dual-RGD twice a week for four weeks. Five nanomoles of siRNA were added in each group. The molar ratio of siRNA to dual-RGD or dsRBD-18His was 1:1.

Cell Viability
The proliferation and cytotoxicity of dual-RGD protein were quantified by measuring WST-8 formazan using Cell Counting Kit-8 (CCK-8). C4-2 and HEK293 cells in 10% FBS containing RPMI 1640 culture media were seeded in 96-well plates at a density of 2 × 10 3 in a 5% CO 2 incubator for 24 h at 37 • C. Cells were treated with varying concentrations of dual-RGD for 72 h. CCK-8 reagent (10 µL/well) was added to each well for 4 h. Absorbance was measured at 450 nm on a Tecan infinite M200 microplate reader (Tecan Company, Mannedorf, Switzerland).

In Vivo Toxicity Assay
Athymic mice (n = 6 per group, 3 male and 3 female) were iv treated with saline or dual-RGD (10 nmoles) or an equal volume of saline twice a week for 4 weeks. Major organs were collected for histology assay.

Histology Assay
Tumors and organs (spleen, lung, kidney, intestine, heart, liver, and brain) were collected and fixed with 4% paraformaldehyde. Sections (6 mm) were cut and mounted Pharmaceutics 2021, 13, 2182 5 of 18 on slides and deparaffinized in xylene and ethyl alcohol. Major organs were stained with hematoxylin for 10 s and eosin for 30 s followed by dehydration. For tumor IHC assay, sections were blocked with 3% goat serum for 2 h and incubated with anti-EGFR antibody (1:500). After washing, the sections were incubated with biotinylated secondary antibody (1:200) (Vector Labs, Burlingame, CA, USA) for 1 h. Following washing, the sections were incubated with VECTASTAIN ABC reagents (Vector laboratories, Burlingame, CA, USA) for 30 min. The images were captured with a Nuance fluorescence microscope with a bright field imaging system (Caliper LifeSciences, Waltham, MA, USA). IHC quantification was performed with the ImageJ plugin IHC profiler (NIH) version 10.2. Images were changed to an 8-bit grayscale type and inverted under the "Edit" menu of ImageJ. After inverting, the DAB stain areas were bright, and the unstained areas were dark. The mean intensity was measured. 8-10 fields of each treatment group were assessed.

Statistical Analysis
The results were expressed as means ± SEM. All data were analyzed using a twotailed Student's t-test (Graph Pad Prism version 9) by comparing with the control group, and p < 0.05 was considered statistically significant.

Construction and Expression of Dual RGD-dsRBD-18his (Dual-RGD) Recombinant Protein
In our previous studies, we cloned a dsRBD-18 His protein that contained a doublestranded RNA binding domain and 18His [20]. In this study, we genetically engineered two RGD peptides into N-and C-termini of dsRBD-18His protein, respectively. The N-terminal of dsRBD-18His was appended a cyclic RGD peptide (CDCRGDCFC (RGD-4C)). RGD-4C was cloned with TNFα and tumor necrosis factor-related apoptosis-inducing ligand (TRAIL). RGD-4C-containing TNFα and TRAIL have shown strong tumor selectivity and binding affinity, and also retain natural activities of TNFα [31] or TRAIL [28]. Meanwhile, a GRGDS (Gly-Arg-Gly-Asp-Ser) peptide was added to the C-terminal of dsRBD-18His since GRGDS is relatively conserved in natural RGD-containing proteins, such as fibronectin and α-lytic protease [32,33]. The gene structure of dual-RGD is illustrated in Figure 1A. Briefly, the dsRBD-18His gene in PET28a plasmid [20] was selected as the cloning PCR template. RGD genes and BamH1/Xhol1 restriction enzyme recognition sites were introduced at the 5 and 3 ends of the primers, respectively. The recombinant protein was expressed in E. coli BL21 (DE3), induced by 1 mM of IPTG. Expressed protein was purified by an Ni-NTA affinity column. The purified protein was analyzed on SDS-PAGE with Coomassie blue staining. As shown in Figure 1B, there are two high-density bands in the non-reducing condition (Lane 2). After adding β-mercaptoethanol (β-ME) and heating, the larger-sized protein disappeared, and the small-sized protein remained (Lane 3), which indicates that dual-RGD protein can form a homodimer under nonreducing conditions. SDS-PAGE with a molecular weight (MW) marker showed the expected molecular weights of~26 kDa (monomer) and~52 kDa (dimer). On the basis of the above protein conformation, in the following protein purification, we added β-ME to keep dual-RGD protein as a monomer for function characterization. Furthermore, the hydrodiameter and charge were measured by a ZetaSizer analyzer. As shown in Figure S1, the hydrodynamic diameter of the monomer dual-RGD protein is about 23.5 nm, and the zeta potential is −7.85 mv. This indicates that the new dual-RGD protein is nano-sized with a slightly negative charge that is close neutral. This feature is in contrast with cationic polymer carriers.
weights of ~26 kDa (monomer) and ~52 kDa (dimer). On the basis of the above protein conformation, in the following protein purification, we added β-ME to keep dual-RGD protein as a monomer for function characterization. Furthermore, the hydrodiameter and charge were measured by a ZetaSizer analyzer. As shown in Figure S1, the hydrodynamic diameter of the monomer dual-RGD protein is about 23.5 nm, and the zeta potential is −7.85 mv. This indicates that the new dual-RGD protein is nano-sized with a slightly negative charge that is close neutral. This feature is in contrast with cationic polymer carriers.

siRNA Binding Capability
Next, we evaluated the functionalities of three domains in dual-RGD protein, including dsRBD, 18His, and RGD. First, we detected the dsRBD domain for siRNA binding. Selected as a siRNA model, EGFR siRNA was labeled with FAM fluorophore at the 5′ end of the sense strand. FAM-labeled EGFR siRNA was incubated with dual-RGD at different protein/siRNA molar ratios (1:1, 2:1, and 4:1) for 30 min at room temperature. Bound and unbound siRNA was quantified using 1% agarose gel electrophoresis. As shown in Figure  2A (left), under UV transilluminator, the mobility of the complex of siRNA and dual-RGD was much slower than the control of siRNA only. At a 1:1 molar ratio, all siRNA was bound to dual-RGD, and no free siRNA was detectable. To further detect if slowly moving siRNA had indeed bound to dual-RGD protein during the migration, the same gel was stained with Coomassie blue to identify the location of dual-RGD protein. As shown in Figure 2A (right), the bright field image shows that the protein location (Coomassie blue) was colocalized with the siRNA location (UV-revealed). The result clearly indicates that the dsRBD domain in dual-RGD maintains siRNA binding capability. In the previous

siRNA Binding Capability
Next, we evaluated the functionalities of three domains in dual-RGD protein, including dsRBD, 18His, and RGD. First, we detected the dsRBD domain for siRNA binding. Selected as a siRNA model, EGFR siRNA was labeled with FAM fluorophore at the 5 end of the sense strand. FAM-labeled EGFR siRNA was incubated with dual-RGD at different protein/siRNA molar ratios (1:1, 2:1, and 4:1) for 30 min at room temperature. Bound and unbound siRNA was quantified using 1% agarose gel electrophoresis. As shown in Figure 2A (left), under UV transilluminator, the mobility of the complex of siRNA and dual-RGD was much slower than the control of siRNA only. At a 1:1 molar ratio, all siRNA was bound to dual-RGD, and no free siRNA was detectable. To further detect if slowly moving siRNA had indeed bound to dual-RGD protein during the migration, the same gel was stained with Coomassie blue to identify the location of dual-RGD protein. As shown in Figure 2A (right), the bright field image shows that the protein location (Coomassie blue) was colocalized with the siRNA location (UV-revealed). The result clearly indicates that the dsRBD domain in dual-RGD maintains siRNA binding capability. In the previous studies [34] with cognate HIV TAR RNA (59NT), which contains bulges and an internal loop, using isothermal titration calorimetry, a model was fit for the two binding sites in TAR for the interaction of TAR RNA with dsRBD. One is a single high-affinity site and the second is a low-affinity site. At a 1:1 molar ratio of dsRNA to dsRBD, a type 1 complex with fast gel mobility was formed. At a higher ratio of 1:2, a type 2 complex was formed with slow gel mobility. The dissociation constant for the type 1 complex was 70 nM with a 1:1 stoichiometry. By contrast, the dissociation constant for the type 2 complex was 23 µM with 1:2 stoichiometry, which is 300-fold higher than that of the type 1 complex. Importantly, the previous studies [22] also demonstrated that: 16-20 bp dsRNA binding to dsRBD resulted in only the formation of complex 1 at a 1:1 molar ratio, while 22-24 bp dsRNA binding to dsRBD could form complex 2 at a molar ratio 1:2 of dsRNA to dsRBD. In our report, we only observed complex 1 (1:1 molar ratio) ( Figure 2A) because our siRNAs were 19 bp [35]. Our results are in agreement with those of previous studies. Meanwhile, complex 1 had a smaller size than complex 2, making it suitable for tissue penetration in vivo.
with slow gel mobility. The dissociation constant for the type 1 complex was 70 nM with a 1:1 stoichiometry. By contrast, the dissociation constant for the type 2 complex was 23 µM with 1:2 stoichiometry, which is 300-fold higher than that of the type 1 complex. Importantly, the previous studies [22] also demonstrated that: 16-20 bp dsRNA binding to dsRBD resulted in only the formation of complex 1 at a 1:1 molar ratio, while 22-24 bp dsRNA binding to dsRBD could form complex 2 at a molar ratio 1:2 of dsRNA to dsRBD. In our report, we only observed complex 1 (1:1 molar ratio) (Figure 2A) because our siR-NAs were 19 bp [35]. Our results are in agreement with those of previous studies. Meanwhile, complex 1 had a smaller size than complex 2, making it suitable for tissue penetration in vivo.  To localize dual-RGD protein position, the same gels were stained with Coomassie blue. Two images show that dual-RGD and siRNA were at the same location, indicating dual-RGD binding to siRNA. (B) Protection of siRNA against nucleases. Naked siRNA or siRNA/dual-RGD complexes were incubated with 50% (v/v) human serum for different time periods at 37 • C. The siRNA degradation was probed using agarose gel electrophoresis.

Protection of siRNA against Nucleolytic Degradation
Following the confirmation of dual-RGD/siRNA binding, we evaluated whether the formation of complexes could protect siRNAs from RNase attack. Dual-RGD/siRNA complex or siRNA only was spiked with 50% fresh pooled human serum at different times. As shown in Figure 2B, siRNA in a complex with dual-RGD did not show degradation for the 24 h test period, while siRNA alone in 50% human serum showed degradation at the start (1 h). Therefore, the dual-RGD carrier could protect siRNA against nuclease digestion. Notably, we did not perform chemical modification for the test EGFR siRNA. The serum volume always accounts for 50% (v/v) in all test groups. Thus, it is not surprising that the serum volume in siRNA-only groups was lower than that in siRNA/dual-RGD complex groups since siRNA alone has a lower volume than the siRNA/dual-RGD complex ( Figure 2B).

RGD Binding Specificity
The binding specificity of dual-RGD protein was assessed in a series of cell lines with different levels of integrin α v β 3. To verify the expression levels of integrin α v β 3 , different cell lines were probed for integrin β 3 expression with Western blot. As shown in Figure 3A, the cell lines MDA-MB-231, Hs587T, and BT20 expressed high levels of integrin β 3 , while the cell lines C4-2, BXPC3, and HEK293T showed very low expression levels of integrin β 3 . Next, FAM-EGFR siRNA was loaded into dual-RGD at a molar ratio of 1:1. A complex of dual-RGD/siRNA was incubated with formalin-fixed cell lines at 37 • C for 1 h, and the binding specificity of dual-RGD was detected by flow cytometry. Notably, fixed cells allow RGD binding to target cells but do not induce endocytosis, thus providing a precise assessment of cell surface integrin receptor-RGD binding. As shown in Figure 3B, dual-RGD had low or no binding to C4-2 and HEK-293T cells, while it had a high binding intensity for MDA-MD-231, Hs578T, and BT-20. The correlation of single-cell fluorescence intensity and integrin β 3 protein levels demonstrated the binding specificity of the RGD domain in dual-RGD protein. Furthermore, in another independent experiment, a competitive binding assay was performed to evaluate RGD receptor-specific binding. MDA-MB-231 cells were incubated with dual-RGD/siRNA in the presence of free RGD peptide. As shown in Figure 3C, after blocking with RGD peptide, the binding intensity of dual-RGD/siRNA to MDA-MB-231 cells was significantly reduced, which is displayed by the fluorescence peak's significant shift to the left. The above experiments suggest that the binding of the dual-RGD/siRNA complex to the cells occurs through the fused RGD domain in the dual-RGD carrier.

Cellular Uptake and Subcellular Distribution of Dual-RGD Carrier
Furthermore, we assessed whether RGD-integrin binding induced cell-type-specific internalization and endosomal escape, which are the prerequisites for siRNA to trigger gene silencing. In this recombinant protein, dual-RGD was incorporated with 18His, which was expected to provide endosomal rupture through the proton sponge effect. To prove 18His endosomolytic activity, confocal microscopy was performed to visualize the subcellular location of siRNA. First, Cy5 (red) fluorophore-labeled siRNA was complexed with dual-RGD or dsRBD-18His (without the RGD-targeting moiety as a control) at a molar ratio of 1:1. Then integrin α v β 3 -positive MDA-MB-231 live cells were treated with siRNA/dual-RGD or siRNA/dsRBD-18His for 6 h in 37 • C in a CO 2 incubator followed by a DPBS wash. In the meantime, nuclei were labeled with DAPI (blue), and endosomes/lysosomes were stained with LysoTracker (green). As shown in Figure 4 (top two layers), siRNA/dual-RGD-treated MDA-MB-231 cells showed a much higher density of the red fluorescence signal in the cytoplasm compared with the siRNA/dsRBD-18His-treated control. It is worth mentioning that the yellow fluorescence signal is derived from the colocalization of Cy5-siRNA (red) with LysoTracker (green) and indicates siRNA endosomal entrapment. In siRNA/dual-RGD-treated MDA-MB-231 cells, a low amount yellow fluorescence signal indicates that less Cy5-siRNA was entrapped in endosomes, and a high amount of red signal in the cytoplasm around the nuclei indicates that most siRNA has escaped from endosomes. The above results suggested that cloned RGD peptide can lead to receptormediated endocytosis, and internalized siRNA can escape from endosomes and diffuse to the cytoplasm. Furthermore, as a cell type control, integrin α v β 3 -negative C4-2 cells were treated with Cy5-siRNA/dual-RGD for 6 h and washed with DPBS. The confocal image (Figure 4, bottom layer) showed that there was no detectable Cy5-siRNA signal in C4-2 cells, which is consistent with the binding specificity assay shown in Figure 3. These results strongly imply that dual-RGD-directed siRNA endocytosis is cell type-specific, which may avoid off-target effects during in vivo circulation.

Cellular Uptake and Subcellular Distribution of dual-RGD Carrier
Furthermore, we assessed whether RGD-integrin binding induced cell-type-specific internalization and endosomal escape, which are the prerequisites for siRNA to trigger gene silencing. In this recombinant protein, dual-RGD was incorporated with 18His, which was expected to provide endosomal rupture through the proton sponge effect. To prove 18His endosomolytic activity, confocal microscopy was performed to visualize the subcellular location of siRNA. First, Cy5 (red) fluorophore-labeled siRNA was complexed with dual-RGD or dsRBD-18His (without the RGD-targeting moiety as a control) at a molar ratio of 1:1. Then integrin αvβ3-positive MDA-MB-231 live cells were treated with siRNA/dual-RGD or siRNA/dsRBD-18His for 6 h in 37 °C in a CO2 incubator followed by

Target Gene Knockdown In Vitro
After confirmation of dual-RGD led siRNA endocytosis and endosomal escape, we evaluated whether siRNA triggered target gene knockdown. As a model, EGFR siRNA was selected and formed complexes with dual-RGD at a molar ratio of 1:1. EGFR siRNA/dual-RGD complexes were incubated with MDA-MB-231 cells or Hs587 T cells for 72 h. It is worth mentioning that both MDA-MB-231 and Hs587T cells have high EGFR expression, as reported in our previous studies [36]. Cationic lipofectamine RNAi MAX was used as a control to verify siRNA functionality. The gene knockdown was probed in cell lysates with Western blot. As shown in Figure 5, the dual-RGD/siRNA complex displayed a dose-dependent gene knockdown in both cell lines. At 200 nM, siRNA/dual-RGD could silence 75% of EGFR in MDA-MB-231 cells and 65% of EGFR in Hs578T cells. It is not surprising that lipofectamine RNAi MAX showed better performance. However, siRNA uptake by cationic lipofectamine occurred via charge-charge interaction between lipofectamine (cationic) and charged cell membranes (anionic). Since cellular uptake through chargecharge interaction has no cell-type specificity, lipofectamine is incapable of in vivo siRNA targeting delivery.
sosomes were stained with LysoTracker (green). As shown in Figure 4 (top two layers), siRNA/dual-RGD-treated MDA-MB-231 cells showed a much higher density of the red fluorescence signal in the cytoplasm compared with the siRNA/dsRBD-18His-treated control. It is worth mentioning that the yellow fluorescence signal is derived from the colocalization of Cy5-siRNA (red) with LysoTracker (green) and indicates siRNA endosomal entrapment. In siRNA/dual-RGD-treated MDA-MB-231 cells, a low amount yellow fluorescence signal indicates that less Cy5-siRNA was entrapped in endosomes, and a high amount of red signal in the cytoplasm around the nuclei indicates that most siRNA has escaped from endosomes. The above results suggested that cloned RGD peptide can lead to receptor-mediated endocytosis, and internalized siRNA can escape from endosomes and diffuse to the cytoplasm. Furthermore, as a cell type control, integrin αvβ3-negative C4-2 cells were treated with Cy5-siRNA/dual-RGD for 6 h and washed with DPBS. The confocal image (Figure 4, bottom layer) showed that there was no detectable Cy5-siRNA signal in C4-2 cells, which is consistent with the binding specificity assay shown in Figure  3. These results strongly imply that dual-RGD-directed siRNA endocytosis is cell typespecific, which may avoid off-target effects during in vivo circulation.

Tumor-Targeting Capability In Vivo
After confirming the functionalities of the three domains of dual-RGD protein in vitro, we further evaluated them in vivo. MDA-MB-231 tumor-bearing mice were intravenously (iv) injected with 100 µL of Cy5-EGFR siRNA/dual-RGD complex (5 nmoles) or equal moles of Cy5-siRNA/dsRBD-18His complex as a control. At time points of 3 h, 6 h, 12 h, and 24 h, Cy5 fluorescence signals in mice were detected with the Xenogen IVIS 100 imaging system to monitor siRNA distribution in the whole body ( Figure 6). After the 3 h injection, the Cy5 siRNA signal could be detected in the tumors of dual-RGD-treated mice but not in the tumors of dsRBD-18His-treated mice. At the 6h time point, Cy5-siRNA in dual-RGD-treated mice could be clearly visualized in the tumors but with reduced signal around other organs, whereas Cy5-siRNA in dsRBD-18His-treated mice remained undetectable in tumors. The Cy5-siRNA signal in tumors could last up to 12 h in dual-RGD-treated mice. After 24 h, both dual-RGD-and dsRBD-18His-delivered Cy5-siRNA had been removed from the mice. These results indicate that dual-RGD enables siRNA to be delivered to tumor sites, where siRNA can remain for 12 h. Since we labeled the siRNA sense strand with Cy5, it is possible that the sense strand (passenger strand) of siRNA was degraded after it was loaded into the RISC (RNA-Induced Silencing Complex) followed by gene silencing, which is a natural process for RNA interference but not serum nuclease-induced degradation. The time-course imaging demonstrated that dual-RGD can guide siRNA, targeting tumors in vivo. At 12 h and 24 h, we removed the major organs from treated mice and homogenized the tissues. The homogenized tissues were detected by a fluorescence microplate reader, as shown in Supplement Figure S2; at 12 h, there were weak fluorescence signals in the sites of the intestine and the brain (around the eyes), and most organs, including the heart, lungs, kidneys, spleen, and liver, were free of Cy5 signals. We reason that the capillary blood vessels around the eyes have a slower metabolic rate than other organs may experience the process of removing all metabolites; thus, some signals were detected in these organs. At 24 h, almost all fluorescence signals had been cleared from control and dual-RGD mice.

Target Gene Knockdown In Vitro
After confirmation of dual-RGD led siRNA endocytosis and endosomal escape, we evaluated whether siRNA triggered target gene knockdown. As a model, EGFR siRNA was selected and formed complexes with dual-RGD at a molar ratio of 1:1. EGFR siRNA/dual-RGD complexes were incubated with MDA-MB-231 cells or Hs587 T cells for 72 h. It is worth mentioning that both MDA-MB-231 and Hs587T cells have high EGFR expression, as reported in our previous studies [36]. Cationic lipofectamine RNAi MAX was used as a control to verify siRNA functionality. The gene knockdown was probed in cell lysates with Western blot. As shown in Figure 5, the dual-RGD/siRNA complex displayed a dose-dependent gene knockdown in both cell lines. At 200 nM, siRNA/dual-RGD could silence 75% of EGFR in MDA-MB-231 cells and 65% of EGFR in Hs578T cells. It is not surprising that lipofectamine RNAi MAX showed better performance. However, siRNA uptake by cationic lipofectamine occurred via charge-charge interaction between lipofectamine (cationic) and charged cell membranes (anionic). Since cellular uptake through charge-charge interaction has no cell-type specificity, lipofectamine is incapable of in vivo siRNA targeting delivery.

Tumor-Targeting Capability In Vivo
After confirming the functionalities of the three domains of dual-RGD protein in vitro, we further evaluated them in vivo. MDA-MB-231 tumor-bearing mice were intravenously (iv) injected with 100 µL of Cy5-EGFR siRNA/dual-RGD complex (5 nmoles) or equal moles of Cy5-siRNA/dsRBD-18His complex as a control. At time points of 3 h, 6 h, 12 h, and 24 h, Cy5 fluorescence signals in mice were detected with the Xenogen IVIS 100 imaging system to monitor siRNA distribution in the whole body ( Figure 6). After the 3 h injection, the Cy5 siRNA signal could be detected in the tumors of dual-RGD-treated mice

Target Gene Knockdown In Vivo
Furthermore, target gene knockdown was assessed in MDA-MB-231 xenografts. When tumors reached 100 mm 3 , mice were iv treated with EGFR siRNA/dual-RGD (5 nmoles) or controls of saline, EGFR siRNA, or EGFR siRNA/dsRBD twice a week for 4 weeks. At the endpoint, the tumors and major organs were removed. Tumors were analyzed by IHC and Western blot. As shown in Figure 7. IHC demonstrated a significantly reduced expression in the siRNA/dual-RGD group, but not in control groups of saline, siRNA only, or siRNA/dsRGD-18His. The Western blot assay with tumor cell lysate proved the down-regulation of EGFR protein. ImageJ quantitation indicated that after the 4-week treatment, only 30% of EGFR protein remained. These results indicate that the target gene was significantly knocked down in vivo. As a comparison, EGFR expression in the brain was also detected by Western blot. As shown in Figure S3, there was no significant difference in EGFR expression between saline-treated and siRNA/dual RGD-treated mouse brains. This result indicates that dual-RGD did not induce gene knockdown in non-target organs.
degraded after it was loaded into the RISC (RNA-Induced Silencing Complex) followed by gene silencing, which is a natural process for RNA interference but not serum nucleaseinduced degradation. The time-course imaging demonstrated that dual-RGD can guide siRNA, targeting tumors in vivo. At 12 h and 24 h, we removed the major organs from treated mice and homogenized the tissues. The homogenized tissues were detected by a fluorescence microplate reader, as shown in Supplement Figure S2; at 12 h, there were weak fluorescence signals in the sites of the intestine and the brain (around the eyes), and most organs, including the heart, lungs, kidneys, spleen, and liver, were free of Cy5 signals. We reason that the capillary blood vessels around the eyes have a slower metabolic rate than other organs may experience the process of removing all metabolites; thus, some signals were detected in these organs. At 24 h, almost all fluorescence signals had been cleared from control and dual-RGD mice. Figure 6. Time-course imaging of Cy5-siRNA/dual-RGD in vivo biodistribution. MDA-MB-231 tumor-bearing mice were iv-treated with complexes of Cy5-siRNA/dual-RGD or Cy5-siRNA/dsRBD-18His. Biodistribution was captured with Xenogen IVIS 100 system.

Target Gene Knockdown In Vivo
Furthermore, target gene knockdown was assessed in MDA-MB-231 xenografts. When tumors reached 100 mm 3 , mice were iv treated with EGFR siRNA/dual-RGD (5 nmoles) or controls of saline, EGFR siRNA, or EGFR siRNA/dsRBD twice a week for 4 weeks. At the endpoint, the tumors and major organs were removed. Tumors were analyzed by IHC and Western blot. As shown in Figure 7. IHC demonstrated a significantly reduced expression in the siRNA/dual-RGD group, but not in control groups of saline, siRNA only, or siRNA/dsRGD-18His. The Western blot assay with tumor cell lysate proved the down-regulation of EGFR protein. ImageJ quantitation indicated that after the Figure 6. Time-course imaging of Cy5-siRNA/dual-RGD in vivo biodistribution. MDA-MB-231 tumor-bearing mice were iv-treated with complexes of Cy5-siRNA/dual-RGD or Cy5-siRNA/dsRBD-18His. Biodistribution was captured with Xenogen IVIS 100 system.

Toxicity Assessment
Next, we evaluated the possible cytotoxicity of dual-RGD against integrin α v β 3negative cells. Since RGD as a drug has shown cytotoxicity for integrin α v β 3 -expressing tumor cells [37,38], we aimed to find whether dual-RGD protein has cytotoxicity against integrin α v β 3 -negative cells. Integrin α v β 3 -negative C4-2 and HEK293T cells were selected and treated with the varying concentrations of dual-RGD carrier for 72 h; then, CCK-8 reagent was added to detect cell viability. As shown in Figure 8, both C4-2 and HEK293T cells kept over 92% cell viability after being treated with 0.2-1.0 µM of dual-RGD carrier protein. The results showed that there was no significant toxicity in the detection range up to 1.0 µM, which is three times as high as the one used in the delivery work. Furthermore, we evaluated the possible toxicity in vivo. After mice were iv treated with dual-RGD protein (10 nmoles) or saline twice a week for 4 weeks, major organs, including the brain, heart, kidneys, liver, lungs, intestine, muscle, and spleen, were collected. And the histopathology changes were examined. As shown in Figure 8B H&E staining revealed that there was no discernible abnormality observed in dual-RGD protein-treated groups compared with the saline-treated control. These results suggest that the human-origin dual-RGD protein carrier is biocompatible and shows no or low systemic toxicity in vivo. 4-week treatment, only 30% of EGFR protein remained. These results indicate that the target gene was significantly knocked down in vivo. As a comparison, EGFR expression in the brain was also detected by Western blot. As shown in Figure S3, there was no significant difference in EGFR expression between saline-treated and siRNA/dual RGDtreated mouse brains. This result indicates that dual-RGD did not induce gene knockdown in non-target organs.

Toxicity Assessment
Next, we evaluated the possible cytotoxicity of dual-RGD against integrin αvβ3-negative cells. Since RGD as a drug has shown cytotoxicity for integrin αvβ3-expressing tumor cells [37,38], we aimed to find whether dual-RGD protein has cytotoxicity against integrin αvβ3-negative cells. Integrin αvβ3-negative C4-2 and HEK293T cells were selected and treated with the varying concentrations of dual-RGD carrier for 72 h; then, CCK-8 reagent was added to detect cell viability. As shown in Figure 8, both C4-2 and HEK293T cells kept over 92% cell viability after being treated with 0.2-1.0 µM of dual-RGD carrier protein. The results showed that there was no significant toxicity in the detection range up to 1.0 µM, which is three times as high as the one used in the delivery work. Furthermore, we evaluated the possible toxicity in vivo. After mice were iv treated with dual-RGD protein (10 nmoles) or saline twice a week for 4 weeks, major organs, including the brain, heart, kidneys, liver, lungs, intestine, muscle, and spleen, were collected. And the histopathology changes were examined. As shown in Figure 8B H&E staining revealed that there was no discernible abnormality observed in dual-RGD protein-treated groups compared with the saline-treated control. These results suggest that the human-origin dual-RGD protein carrier is biocompatible and shows no or low systemic toxicity in vivo.

Discussion
Tumor-targeted siRNA delivery is of great interest in medical research. It is highly desirable but technically challenging to engineer efficient vectors that are non-cationic, non-toxic, endosomal escapable, cell-type specific, and easy to mass-produce. To find a carrier for in vivo systemic siRNA delivery, natural human proteins are attractive because they are biocompatible, biodegradable, easy to clear from the body, and easy to massproduce by fermentation.
Human PKR has been extensively studied [24,39] as a natural carrier for in vivo siRNA delivery. Natural PKR in the body is activated by dsRNA and plays a major role

Discussion
Tumor-targeted siRNA delivery is of great interest in medical research. It is highly desirable but technically challenging to engineer efficient vectors that are non-cationic, nontoxic, endosomal escapable, cell-type specific, and easy to mass-produce. To find a carrier for in vivo systemic siRNA delivery, natural human proteins are attractive because they are biocompatible, biodegradable, easy to clear from the body, and easy to mass-produce by fermentation.
Human PKR has been extensively studied [24,39] as a natural carrier for in vivo siRNA delivery. Natural PKR in the body is activated by dsRNA and plays a major role in the cellular antiviral response through the inhibition of eukaryotic initiation factor 2 (eIF-2) [40,41]. PKR consists of an N-terminal dsRBD, including a pair of double-stranded RNA binding motifs (dsRBM1 and dsRBM2) [42,43] and a C-terminal kinase domain [44]. dsRBM1 has a dominant role in molecular recognition of short dsRNA sequences (15-30 bp), whereas both motifs contribute to binding longer dsRNA sequences (>40 bp). dsRBD binds RNA through protein 3D conformation but not through charge-charge interaction [21]. In addition to its ability to sense dsRNA, primarily of viral origin, PKR is also activated in response to endogenous RNA, such as microRNAs [45,46]. dsRBD from PKR can bind to a broad range of dsRNA in a sequence-independent manner [47], which is unique among known RNP (ribonucleoprotein) complexes. It also has been proven that dsRBD only binds to dsRNA but not RNA-DNA hybrids or dsDNA [22], which provides the basis to generate dsRBD as an siRNA carrier. However, dsRBD lacks endosomal escape capability. To address this problem, we used polyhistidine as an endosomal escape moiety instead of peptide transduction domains (PTD) because PTD is also an arginine-and lysine-rich positively charged peptide. Histidine has a pKa value of about 6.0; at neutral or tumor environment (pH 6.5-6.9) conditions [48], they are mainly deprotonated (uncharged), while in acidic conditions, such as in late endosomes (pH 4.5-5.5) [49], histidine becomes protonated (charged) and facilitates osmotic swelling, leading to cargo release, a mechanism proposed as the proton sponge effect [50]. However, it is well known that the 6xHis tag used for protein purification does not demonstrate the capability of endosome escape. We reason that 6xHis is too short to offer enough buffering effect, and certain elongated His tags should confer cargo endosomal escape. We inserted 18 His peptide into dsRBD protein [20]. We compared the endosomal escape capability of dsRBD-18His and dsRBD-6His. Our results demonstrated that adding 12 more His could significantly improve endosomal escape capability. 18His can confer adequate buffering capacity to drive cargo endosomal escape. In contrast to cationic carriers, dsRBD-18His is uncharged in the physiological condition, has low toxicity, and is biodegradable. Based on the dsRBD-18His vector, we further equipped it with two RGD peptides for tumor targeting.
The mechanism of siRNA released from dsRBD is believed to occur via displacement by the highly abundant longer RNA molecules in the cytosol [51]. Since dsRBD recognizes RNA in a sequence-independent fashion, in the referred studies, with total RNA isolated from cells, dsRBD-siRNA complex was incubated with 80 ng/µL cellular RNAs (which is still much lower than the cytosolic RNA concentration), and the siRNA-dsRBD complex is totally disrupted. These studies support that the competitive binding and exchange of cellular RNAs to dsRBD can trigger siRNA release from dsRBD inside of cells.
In this study, we aimed to establish a general tumor-targeted siRNA delivery platform. We focused on the characterization of dual-RGD performance in vivo and in vitro, and we will evaluate tumor treatment efficacy in the next study. RGD peptides have been widely used in drug delivery and imaging and have minimized safety concerns [26,52]. RGD has high affinity and specificity for integrin α v β 3 , which is highly expressed in tumor new blood vessels and many tumor cells but is not present in resting endothelial cells or normal organ systems. New RGD carriers will simplify the process of siRNA delivery and enable targeted delivery of siRNAs to tumor blood vessels and many tumors, such as glioblastomas and melanomas and pancreatic, ovarian, breast, and prostate cancers [53][54][55]. RGD-based materials have undergone rapid development for sensitive tumor detection and tumor-targeted drug delivery in vivo [23,56,57]. RGD-PET (positron emission tomography) has been used in the clinic for tumor diagnosis [58,59]. RGD has also been widely used in developing anti-angiogenesis therapeutics [60,61] and conjugation with radionuclides [62,63] or chemotherapeutic drugs [64,65] for tumor treatment. TNFα has been fused with RGD by recombinant DNA technology and has achieved tumor targeting and reduced system damage [66,67]. Cyclic RGD peptide has shown better targeting capability in vivo. In this study, to avoid unwanted internal disulfide bonds between the N-terminal and C-terminal, we cloned one end (N-) with cyclic RGD and another end (C-) with linear GRGDS peptide. The results proved that the cloned protein possesses ideal functionality.
In terms of delivery efficacy, lipofectamine outperformed the dual-RGD carrier. This was expected since cationic charge-charge interaction between lipofectamine and the cell membrane is much more effective in gene transfection than ligand-receptor-mediated endocytosis, which is the assumed mechanism of the dual-RGD carrier. However, cationic lipofectamine cannot be used for in vivo cell-type-specific siRNA delivery because of nonspecific binding.
Cyclic peptide c(-RGDfV-), as the drug Cilengitide, has shown the inhibition of tumor growth and angiogenesis [37,38]. We envision that the fusion protein dual-RGD carrier complexed with siRNA will have increased anti-tumor efficacy in addition to siRNAinduced gene silencing.

Conclusions
Our developed protein-based dual-RGD vector has three functions: a dsRBD domain for siRNA docking, 18 His for endosomal escape, and RGD for tumor targeting. This three-in-one multidomain vector will address the problems of current siRNA carriers of positive charge, low serum stability, poor clearance, cytotoxicity, and lack of cell-type specificity. RGD is much better than transferrin and folate, which can only target cancer cells but cannot target the tumor neo-vasculature. Our cloned dual-RGD protein carrier with load-to-go capability will expedite siRNA translation to in vivo tumor therapy.
Supplementary Materials: The following are available online at https://www.mdpi.com/article/10 .3390/pharmaceutics13122182/s1, Figure S1: Detection of hydrodynamic diameter and zeta potential by Zetasizer. Figure S2: Ex vivo evaluation of biodistribution. Figure S3: Evaluation of off-target gene silencing in brain.