**4.1 Poly-L-lysine (PLL)**

The amino acid sequence of PLL and all other peptides subsequently discussed can be found in Table 1. PLL is the first peptide-based vector to be studied intensively. The molecular weight of PLL spans a wide range and depends on the number of repeating units within a chain, or its degree of polymerisation (DP). For the sake of discussion, the molecular weight of PLL is arbitrarily classified as follows: oligolysine (DP<20), low (20≤DP<50), medium (50≤DP≤400) or high (DP>400). Most studies use PLL of medium molecular weight.

PLL has been successfully employed to deliver a host of different cargos. By covalently coupling human serum albumin or HRP to PLL via amide bonds in a 1-ethyl-3-(3 dimethylaminopropyl) carbodiimide (EDC) catalysed reaction, proteins were delivered into L929 mouse fibroblasts for the first time using such a PLL-mediated strategy (Shen & Ryser, 1978). More recently, streptavidin-conjugated quantum dots (QD) were also attached to biotin-tagged PLL and delivered into HeLa cells (Mok et al., 2008). However, PLL is most frequently used to deliver nucleic acids where the electrostatic attachment is preferred. Although polylysine made out of D-amino acids has been suggested to work better – presumably because they are more resistant to enzymatic degradation (Mitchell et al., 2000) – PLL continues to be favoured in transfection studies. Wu and partner provided one of the earliest examples of PLL being used to condense plasmids for receptor-mediated delivery (Wu & Wu, 1987). In that study, asialoorosomucoid (ASOR) was first covalently attached to PLL with *N*-succinimidyl 3-(2-pyridyldithio) propionate (SPDP) as a linker and then used to

Peptides as Promising Non-Viral Vectors for Gene Therapy 627

groups on PLL to form a thiourea bond and used to mediate gene expression in the livers of rats for up to 140 days (Perales et al., 1994). Mannose and lactose were also used to target cell lines that have receptors for the specific sugar (Midoux et al., 1993). ASOR was coupled onto PLL in an EDC catalysed reaction and intravenously injected into rats (Wu et al., 1989). However, reporter gene expression by liver cells disappeared by day four. Transferrin was coupled to PLL via disulfide bonds in a SPDP-mediated reaction to increase its accumulation within cells (Wagner et al., 1991). Insulin was also coupled to PLL in an SPDPmediated reaction and used with adenovirus (biotin-streptavidin conjugated) to deliver plasmids into pre-implantation mammalian embryos (Ivanova et al., 1999). PEG was grafted onto PLL via amide bonds in an *N*-hydroxysuccinimide (NHS) catalysed reaction and helped in preventing particle aggregation (Rimann et al., 2008). Stearyl-PLL mixed with a low density lipoprotein was also used to condense plasmids (~600 nm) into ~100 nm complexes, as measured by atomic force microscopy (Kim et al., 1998). Further, the lipoprotein was found to be essential for efficient uptake. Finally, peptides themselves can serve as targeting signals. For example, a short peptide sequence (GACRRETAWACG) suggested to target α5β1 integrins was linked to K16 and mixed with Lipofectin® to transfect neuroblastoma cells (Lee et al., 2003). Several groups have also used the RGD motif to target integrins. A recent example involved RGD being linked to K16 peptides, which were then used with an intracellularly cleavable PEG-lipid formulation to deliver plasmids into mice bearing subcutaneous tumors (Tagalakis et al., 2011). The integrin-targeting signal was shown to be important. Upon augmentation by the enhanced permeation and retention

Full length tat has 86 amino acids (sequence 1-86) and it is a regulatory protein encoded by HIV-1 that transactivates viral gene expression. The ability of tat to cross cell membrane was initially observed by two groups independently in 1988 (Frankel & Pabo, 1988; Green & Loewenstein, 1988). In one study, tat was simply added to a culture of HL3T1 cells modified to contain an integrated copy of chloramphenicol acetyltransferase (CAT, a reporter gene) under the control of the HIV-1 LTR (long terminal repeat) promoter and CAT expression was unexpectedly detected (Frankel & Pabo, 1988). Crucially, the amount of CAT expression was dependent on the dosage of tat. Unlike earlier experiments, tat did not require any help from the scrape loading technique to enter cells. Tat 1-86 was further dissected into five regions and region II and III, together spanning roughly residues 38-62, were identified to be essential and sufficient for transactivation activity (Green & Loewenstein, 1988). Region III (roughly residues 49-62), in particular, was interesting and was rich in arginine and lysine. Furthermore, replacing the three arginines at position 55-57 with alanines drastically reduced transactivation activity. Tat was also observed to localise in the nucleus and

The region surrounding the arginine- and lysine-rich domain of tat was keenly evaluated as a gene carrier. Today, full length tat is seldom used and among the many truncated versions being studied (Table 2), tat 47-57 and 48-60 are the two most popular sequences. The *in vivo* half life of free tat 47-57 was calculated to be ~3.5 minutes (Grunwald et al., 2009) and tat 48- 60 exists as an unstructured random coil in buffer solutions or when bound to lipid vesicles (Caesar et al., 2006). The secondary structure of a carrier has been suggested to affect its membrane translocation ability. However, it remains unclear if it is better for a structure to

effect, the complexes were mainly distributed to the tumor.

thought to be a NLS (Green & Loewenstein, 1988).

**4.2 Tat** 

bind a reporter plasmid. Complexes then accumulated within HepG2 cells (ASOR receptor positive) but not in SK-Hep1 cells (ASOR receptor negative).

Plasmids dissociated more slowly if bound to a higher molecular weight PLL and this can negatively affect transfection (Schaffer et al., 2000). To facilitate plasmid release, cysteine residues were added to oligolysine, which was then cross-linked using disulfide bonds (McKenzie et al., 2000). The intention was to make use of the intracellular environment – which is much more reductive due to its elevated levels of glutathione (Lee et al., 2007) – to break down the disulfide bonds for polymer disintegration and plasmid release. The number of lysine, cysteine and histidine (for buffering capacity) residues used and the spacing and ordering between the residues were found to influence the final performance of the vector. Upon optimisation, some constructs transfected at levels that were comparable or even higher than LipofectAce, a commercial reagent used as a positive control (McKenzie et al., 2000). Disulfide-linked oligolysine was then further functionalised with triantennary Nglycan signals to target hepatocytes and evaluated *in vivo* (Kwok et al., 2003). However, contrary to results *in vitro*, the particles were not stable enough in the reductive intracellular liver environment and premature plasmid release ultimately limited gene expression.

PLL by itself is generally not considered to be an efficient vector (Meyer et al., 2008) and is frequently coupled with other agents. The imidazole headgroup of histidine (pKa~6) can provide endosomal buffering and has been added to PLL to boost transfection (Midoux & Monsigny, 1999). Histidine was also added onto a K15-based oligopeptide which was then self assembled for drug and gene co-delivery (Wiradharma et al., 2009). The HA2 subunit of the hemagglutinin glycoprotein present on the surface of influenza virus plays an important role in the endosomal escape of viruses. To do this, HA2 exhibits a pH-dependent membrane fusion activity. HA2 is not normally lytic at neutral pH. However, protonation of its acidic residues during endo/lysosomal transition triggers HA2 to adopt a more α-helical secondary conformation. At the same time, it exposes a highly-conserved hydrophobic sequence which then interacts and destabilises the endosomal membrane (Wagner et al., 1992a). To exploit this membrane-disrupting mechanism, 20 amino acids corresponding to the N-terminus of HA2 were attached to PLL via cysteine-mediated disulfide bonds. The HA2 functionalised PLL was then shown to induce pH-dependent liposomal leakage (Wagner et al., 1992a) and to augment transfection to a greater extent than the use of chloroquine (Midoux et al., 1993). Melittin, a major component of bee venom, is another peptide with fusogenic activity but unlike HA2, melittin is unresponsive to pH. As such, it remains fusogenic at pH 7 and can indiscriminately disrupt plasma membranes (Chen et al., 2006). To confine fusogenic activity within the endosomes, dimethylmaleic anhydride protecting groups were attached to mask the activity of melittin at neutral pH. Upon cleavage of the protecting groups under acidic conditions, the activity of melittin was restored (Meyer et al., 2008). This protected form of melittin was then coupled to PLL via a cysteine-mediated disulfide bond and the entire construct mediated siRNA knockdown more efficiently than PEI. Adenovirus, known to display pH-dependent membrane disruption as part of its infectious cycle (Curiel et al., 1991), has also been coupled to PLL to improve transfection efficiency. Human adenovirus type 5 (dl312) was either biotinylated and coupled to a streptavidinylated PLL (Wagner et al., 1992b), or simply added as free viral particles into the culture medium to be taken up together with the PLL complexes (Curiel et al., 1991). In another study, chicken embryo lethal orphan virus, an adenovirus from chicken, was attached to PLL and augmented transfection as well as human adenovirus (Cotten et al., 1993).

Various signals have also been added onto PLL to improve its uptake or target specificity. For instance, galactose containing an isothiocyanate group was reacted with the amine groups on PLL to form a thiourea bond and used to mediate gene expression in the livers of rats for up to 140 days (Perales et al., 1994). Mannose and lactose were also used to target cell lines that have receptors for the specific sugar (Midoux et al., 1993). ASOR was coupled onto PLL in an EDC catalysed reaction and intravenously injected into rats (Wu et al., 1989). However, reporter gene expression by liver cells disappeared by day four. Transferrin was coupled to PLL via disulfide bonds in a SPDP-mediated reaction to increase its accumulation within cells (Wagner et al., 1991). Insulin was also coupled to PLL in an SPDPmediated reaction and used with adenovirus (biotin-streptavidin conjugated) to deliver plasmids into pre-implantation mammalian embryos (Ivanova et al., 1999). PEG was grafted onto PLL via amide bonds in an *N*-hydroxysuccinimide (NHS) catalysed reaction and helped in preventing particle aggregation (Rimann et al., 2008). Stearyl-PLL mixed with a low density lipoprotein was also used to condense plasmids (~600 nm) into ~100 nm complexes, as measured by atomic force microscopy (Kim et al., 1998). Further, the lipoprotein was found to be essential for efficient uptake. Finally, peptides themselves can serve as targeting signals. For example, a short peptide sequence (GACRRETAWACG) suggested to target α5β1 integrins was linked to K16 and mixed with Lipofectin® to transfect neuroblastoma cells (Lee et al., 2003). Several groups have also used the RGD motif to target integrins. A recent example involved RGD being linked to K16 peptides, which were then used with an intracellularly cleavable PEG-lipid formulation to deliver plasmids into mice bearing subcutaneous tumors (Tagalakis et al., 2011). The integrin-targeting signal was shown to be important. Upon augmentation by the enhanced permeation and retention effect, the complexes were mainly distributed to the tumor.

#### **4.2 Tat**

626 Non-Viral Gene Therapy

bind a reporter plasmid. Complexes then accumulated within HepG2 cells (ASOR receptor

Plasmids dissociated more slowly if bound to a higher molecular weight PLL and this can negatively affect transfection (Schaffer et al., 2000). To facilitate plasmid release, cysteine residues were added to oligolysine, which was then cross-linked using disulfide bonds (McKenzie et al., 2000). The intention was to make use of the intracellular environment – which is much more reductive due to its elevated levels of glutathione (Lee et al., 2007) – to break down the disulfide bonds for polymer disintegration and plasmid release. The number of lysine, cysteine and histidine (for buffering capacity) residues used and the spacing and ordering between the residues were found to influence the final performance of the vector. Upon optimisation, some constructs transfected at levels that were comparable or even higher than LipofectAce, a commercial reagent used as a positive control (McKenzie et al., 2000). Disulfide-linked oligolysine was then further functionalised with triantennary Nglycan signals to target hepatocytes and evaluated *in vivo* (Kwok et al., 2003). However, contrary to results *in vitro*, the particles were not stable enough in the reductive intracellular liver environment and premature plasmid release ultimately limited gene expression. PLL by itself is generally not considered to be an efficient vector (Meyer et al., 2008) and is frequently coupled with other agents. The imidazole headgroup of histidine (pKa~6) can provide endosomal buffering and has been added to PLL to boost transfection (Midoux & Monsigny, 1999). Histidine was also added onto a K15-based oligopeptide which was then self assembled for drug and gene co-delivery (Wiradharma et al., 2009). The HA2 subunit of the hemagglutinin glycoprotein present on the surface of influenza virus plays an important role in the endosomal escape of viruses. To do this, HA2 exhibits a pH-dependent membrane fusion activity. HA2 is not normally lytic at neutral pH. However, protonation of its acidic residues during endo/lysosomal transition triggers HA2 to adopt a more α-helical secondary conformation. At the same time, it exposes a highly-conserved hydrophobic sequence which then interacts and destabilises the endosomal membrane (Wagner et al., 1992a). To exploit this membrane-disrupting mechanism, 20 amino acids corresponding to the N-terminus of HA2 were attached to PLL via cysteine-mediated disulfide bonds. The HA2 functionalised PLL was then shown to induce pH-dependent liposomal leakage (Wagner et al., 1992a) and to augment transfection to a greater extent than the use of chloroquine (Midoux et al., 1993). Melittin, a major component of bee venom, is another peptide with fusogenic activity but unlike HA2, melittin is unresponsive to pH. As such, it remains fusogenic at pH 7 and can indiscriminately disrupt plasma membranes (Chen et al., 2006). To confine fusogenic activity within the endosomes, dimethylmaleic anhydride protecting groups were attached to mask the activity of melittin at neutral pH. Upon cleavage of the protecting groups under acidic conditions, the activity of melittin was restored (Meyer et al., 2008). This protected form of melittin was then coupled to PLL via a cysteine-mediated disulfide bond and the entire construct mediated siRNA knockdown more efficiently than PEI. Adenovirus, known to display pH-dependent membrane disruption as part of its infectious cycle (Curiel et al., 1991), has also been coupled to PLL to improve transfection efficiency. Human adenovirus type 5 (dl312) was either biotinylated and coupled to a streptavidinylated PLL (Wagner et al., 1992b), or simply added as free viral particles into the culture medium to be taken up together with the PLL complexes (Curiel et al., 1991). In another study, chicken embryo lethal orphan virus, an adenovirus from chicken, was attached to PLL and augmented transfection as well as

Various signals have also been added onto PLL to improve its uptake or target specificity. For instance, galactose containing an isothiocyanate group was reacted with the amine

positive) but not in SK-Hep1 cells (ASOR receptor negative).

human adenovirus (Cotten et al., 1993).

Full length tat has 86 amino acids (sequence 1-86) and it is a regulatory protein encoded by HIV-1 that transactivates viral gene expression. The ability of tat to cross cell membrane was initially observed by two groups independently in 1988 (Frankel & Pabo, 1988; Green & Loewenstein, 1988). In one study, tat was simply added to a culture of HL3T1 cells modified to contain an integrated copy of chloramphenicol acetyltransferase (CAT, a reporter gene) under the control of the HIV-1 LTR (long terminal repeat) promoter and CAT expression was unexpectedly detected (Frankel & Pabo, 1988). Crucially, the amount of CAT expression was dependent on the dosage of tat. Unlike earlier experiments, tat did not require any help from the scrape loading technique to enter cells. Tat 1-86 was further dissected into five regions and region II and III, together spanning roughly residues 38-62, were identified to be essential and sufficient for transactivation activity (Green & Loewenstein, 1988). Region III (roughly residues 49-62), in particular, was interesting and was rich in arginine and lysine. Furthermore, replacing the three arginines at position 55-57 with alanines drastically reduced transactivation activity. Tat was also observed to localise in the nucleus and thought to be a NLS (Green & Loewenstein, 1988).

The region surrounding the arginine- and lysine-rich domain of tat was keenly evaluated as a gene carrier. Today, full length tat is seldom used and among the many truncated versions being studied (Table 2), tat 47-57 and 48-60 are the two most popular sequences. The *in vivo* half life of free tat 47-57 was calculated to be ~3.5 minutes (Grunwald et al., 2009) and tat 48- 60 exists as an unstructured random coil in buffer solutions or when bound to lipid vesicles (Caesar et al., 2006). The secondary structure of a carrier has been suggested to affect its membrane translocation ability. However, it remains unclear if it is better for a structure to

Peptides as Promising Non-Viral Vectors for Gene Therapy 629

was also conjugated onto tat 49-57 and used for plasmid transfection (Saleh et al., 2010). Antisense oligonucleotides were linked via disulfide bonds to cysteine-modified tat 49-60 and delivered into cells to inhibit the expression of P-glycoprotein, a transmembrane pump that is responsible for the multidrug resistance phenotype of tumor cells (Astriab-Fisher et al., 2000). siRNA has also been delivered by tat 47-57 and interestingly, photostimulation promoted the escape of tat complexes into the cytosol (Endoh et al., 2008). Reactive oxygen species produced during laser illumination was postulated to damage the endosomal vesicles for the enhanced cytosolic access. Finally, tat has also played the role of a helper. As an internalisation enhancer, tat 43-60 was displayed on the surfaces of phage particles to augment the delivery of plasmids encapsulated within the phages (Eguchi et al., 2001). In another study, tat 47-57 was used more as a NLS during the transfection of CHO cells (Moore et al., 2009). Unlike many other systems, the presence of serum augmented tat-

Although tat can bring nucleic acids and other macromolecules into a cell, the pathway(s) which tat exploits to do so is ambiguous. There is little consensus in the literature, except for the observation that heparan sulphate, an anionic cell membrane glycosaminoglycan, is crucial for uptake (Sandgren et al., 2002; Tyagi et al., 2001). Besides being rapidly internalised via endocytosis (Tyagi et al., 2001), heparan sulphate is probably also involved in the initial binding step before internalisation (Ferrari et al., 2003). It can also act as a coreceptor for endocytosis (Leonetti et al., 2010). Interestingly, other glycosaminoglycans such as chondroitin sulphate and dermatan sulphate have been ruled out in the binding/internalisation of full length tat 1-86 (Tyagi et al., 2001) and tat 37-57 (Leonetti et

Classical endocytosis is an energy- and ATP-dependent mechanism. For the sake of discussion, endocytosis can be further classified into clathrin-mediated endocytosis, caveolae-mediated endocytosis, macropinocytosis and other forms of clathrin- and caveolaeindependent endocytosis. Prompted by the observation that tat 48-60 was internalised at 4°C, a temperature normally considered inhibitory for endocytosis, an energy-independent pathway was initially suggested to be responsible (Futaki et al., 2001b). Another group similarly detected low-temperature internalisation of tat 43-60 and reported that uptake was independent of endocytosis but required the presence of caveolae (Eguchi et al., 2001). The ability of free tat 47-57 to enter cells in the presence of sodium azide, which blocks intracellular ATP synthesis, also led the authors to preclude endocytosis as the uptake

Several models of energy-independent uptake have been proposed. A common theme seems to involve the peptide sticking to the cell membrane and creating a local mass imbalance. At the same time, it transforms itself into a more non-polar compound either due to charge neutralisation (Su et al., 2009) or hydrogen bonding (Rothbard et al., 2004) with the anionic membrane proteins. The peptide then partitions into the hydrophobic lipid bilayer (Rothbard et al., 2002) and translocates through the membrane in a process that is driven by the voltage potential across membrane (Rothbard et al., 2004) and/or the need to relieve membrane curvature stress caused by the mass imbalance in the initial step (Su et al., 2009). In stark contrast, Richard and colleagues reported that the uptake of tat 48-59, in both its free and PNA-conjugated forms were sensitive to low temperature and sodium azide, indicative of classical endocytosis (Richard et al., 2003). Endocytosis is also believed to be responsible for the uptake of protein-conjugated tat 37-72 in another study (Fawell et al., 1994). Having agreed that the uptake of tat requires energy, opinions are once again divided

mediated transfection (Astriab-Fisher et al., 2000; Eguchi et al., 2001).

al., 2010), but not for the shorter tat 48-60 (Sandgren et al., 2002).

mechanism (Ignatovich et al., 2003).

be rich in α-helices or unordered. On the one hand, structures rich in α-helices have been suggested to be more efficient in inserting themselves and crossing lipid bilayers (Almeida & Pokorny, 2009). This is consistent with the view that α-helical structures are responsible for the membrane disrupting property of fusogenic peptides such as HA2 (Wagner et al., 1992a), melittin (Chen et al., 2006) and GALA (Subbarao et al., 1987). On the other, α-helices has also been negatively correlated with uptake (Ye et al., 2010). Instead, an unordered structure was preferred as it afforded the flexibility needed to adopt the most energetically favourable conformation during membrane translocation (Caesar et al., 2006).


Table 2. Heterogeneity in tat sequences being reported in the literature. All sequences are with respect to the original sequence 1-86 in the first entry. Only one reference is provided for each entry, although there may be more discussed in text.

Proteins have been delivered by tat and are usually covalently attached. Fawell and colleagues were the first to chemically conjugate several different proteins onto tat 37-72 (Fawell et al., 1994). One of the model proteins used was β-galactosidase (β-gal), which was grafted onto tat using succinimidyl 4-(N-maleimidomethyl)-cyclohexane-1-carboxylate (SMCC) as a heterobifunctional cross-linker. The β-gal/tat conjugate was then intravenously injected into mice and found to accumulate mainly in the heart, liver and spleen. In a separate study, β-gal/tat 47-57 conjugates were generated by plasmid fusion and injected intraperitoneally into mice (Schwarze et al., 1999). The conjugates were then found to be distributed to all tissues, even across the tight blood-brain-barrier. GFP/tat 47-57 fusion protein was also delivered into cells and uptake was discovered to increase if the cells had been pretreated with 10% dimethylsulfoxide (DMSO) for one hour prior to transfection (Wang et al., 2010). Other macromolecules delivered by tat include peptide nucleic acid (PNA), which are artificial nucleic acid-mimicking polymers. PNA was either conjugated to tat 48-59 using 2-aminoethoxy-2-ethoxyacetic acid as a cross-linker (Richard et al., 2003) or linked via a disulfide bond to a cysteine-modified tat 48-60 (Lundin et al., 2008).

Nucleic acids, on the other hand, are mostly electrostatically carried by tat. For instance, tat 48-60 was used to deliver reporter plasmids into HeLa cells and it was shown that its efficiency can be further improved by adding the SV40 NLS and a dendrimer of seven lysine residues (Yang et al., 2009). A short membrane active peptide, LK15 (KLLKLLLKLLLKLLK)

be rich in α-helices or unordered. On the one hand, structures rich in α-helices have been suggested to be more efficient in inserting themselves and crossing lipid bilayers (Almeida & Pokorny, 2009). This is consistent with the view that α-helical structures are responsible for the membrane disrupting property of fusogenic peptides such as HA2 (Wagner et al., 1992a), melittin (Chen et al., 2006) and GALA (Subbarao et al., 1987). On the other, α-helices has also been negatively correlated with uptake (Ye et al., 2010). Instead, an unordered structure was preferred as it afforded the flexibility needed to adopt the most energetically

> 1 1-86 (Green & Loewenstein, 1988) 2 1-72 (Frankel & Pabo, 1988) 3 37-72 (Fawell et al., 1994) 4 37-57 (Leonetti et al., 2010) 5 43-60 (Eguchi et al., 2001) 6 47-57 (Wang et al., 2010) 7 48-57 (Wadia et al., 2004) 8 48-58 (Fittipaldi et al., 2003) 9 48-59 (Richard et al., 2005) 10 48-60 (Eiriksdottir et al., 2010) 11 49-57 (Saleh et al., 2010)

12 49-60 (Astriab-Fisher et al., 2000)

Table 2. Heterogeneity in tat sequences being reported in the literature. All sequences are with respect to the original sequence 1-86 in the first entry. Only one reference is provided

Proteins have been delivered by tat and are usually covalently attached. Fawell and colleagues were the first to chemically conjugate several different proteins onto tat 37-72 (Fawell et al., 1994). One of the model proteins used was β-galactosidase (β-gal), which was grafted onto tat using succinimidyl 4-(N-maleimidomethyl)-cyclohexane-1-carboxylate (SMCC) as a heterobifunctional cross-linker. The β-gal/tat conjugate was then intravenously injected into mice and found to accumulate mainly in the heart, liver and spleen. In a separate study, β-gal/tat 47-57 conjugates were generated by plasmid fusion and injected intraperitoneally into mice (Schwarze et al., 1999). The conjugates were then found to be distributed to all tissues, even across the tight blood-brain-barrier. GFP/tat 47-57 fusion protein was also delivered into cells and uptake was discovered to increase if the cells had been pretreated with 10% dimethylsulfoxide (DMSO) for one hour prior to transfection (Wang et al., 2010). Other macromolecules delivered by tat include peptide nucleic acid (PNA), which are artificial nucleic acid-mimicking polymers. PNA was either conjugated to tat 48-59 using 2-aminoethoxy-2-ethoxyacetic acid as a cross-linker (Richard et al., 2003) or

linked via a disulfide bond to a cysteine-modified tat 48-60 (Lundin et al., 2008).

Nucleic acids, on the other hand, are mostly electrostatically carried by tat. For instance, tat 48-60 was used to deliver reporter plasmids into HeLa cells and it was shown that its efficiency can be further improved by adding the SV40 NLS and a dendrimer of seven lysine residues (Yang et al., 2009). A short membrane active peptide, LK15 (KLLKLLLKLLLKLLK)

for each entry, although there may be more discussed in text.

favourable conformation during membrane translocation (Caesar et al., 2006).

Entry Sequence Reference

was also conjugated onto tat 49-57 and used for plasmid transfection (Saleh et al., 2010). Antisense oligonucleotides were linked via disulfide bonds to cysteine-modified tat 49-60 and delivered into cells to inhibit the expression of P-glycoprotein, a transmembrane pump that is responsible for the multidrug resistance phenotype of tumor cells (Astriab-Fisher et al., 2000). siRNA has also been delivered by tat 47-57 and interestingly, photostimulation promoted the escape of tat complexes into the cytosol (Endoh et al., 2008). Reactive oxygen species produced during laser illumination was postulated to damage the endosomal vesicles for the enhanced cytosolic access. Finally, tat has also played the role of a helper. As an internalisation enhancer, tat 43-60 was displayed on the surfaces of phage particles to augment the delivery of plasmids encapsulated within the phages (Eguchi et al., 2001). In another study, tat 47-57 was used more as a NLS during the transfection of CHO cells (Moore et al., 2009). Unlike many other systems, the presence of serum augmented tatmediated transfection (Astriab-Fisher et al., 2000; Eguchi et al., 2001).

Although tat can bring nucleic acids and other macromolecules into a cell, the pathway(s) which tat exploits to do so is ambiguous. There is little consensus in the literature, except for the observation that heparan sulphate, an anionic cell membrane glycosaminoglycan, is crucial for uptake (Sandgren et al., 2002; Tyagi et al., 2001). Besides being rapidly internalised via endocytosis (Tyagi et al., 2001), heparan sulphate is probably also involved in the initial binding step before internalisation (Ferrari et al., 2003). It can also act as a coreceptor for endocytosis (Leonetti et al., 2010). Interestingly, other glycosaminoglycans such as chondroitin sulphate and dermatan sulphate have been ruled out in the binding/internalisation of full length tat 1-86 (Tyagi et al., 2001) and tat 37-57 (Leonetti et al., 2010), but not for the shorter tat 48-60 (Sandgren et al., 2002).

Classical endocytosis is an energy- and ATP-dependent mechanism. For the sake of discussion, endocytosis can be further classified into clathrin-mediated endocytosis, caveolae-mediated endocytosis, macropinocytosis and other forms of clathrin- and caveolaeindependent endocytosis. Prompted by the observation that tat 48-60 was internalised at 4°C, a temperature normally considered inhibitory for endocytosis, an energy-independent pathway was initially suggested to be responsible (Futaki et al., 2001b). Another group similarly detected low-temperature internalisation of tat 43-60 and reported that uptake was independent of endocytosis but required the presence of caveolae (Eguchi et al., 2001). The ability of free tat 47-57 to enter cells in the presence of sodium azide, which blocks intracellular ATP synthesis, also led the authors to preclude endocytosis as the uptake mechanism (Ignatovich et al., 2003).

Several models of energy-independent uptake have been proposed. A common theme seems to involve the peptide sticking to the cell membrane and creating a local mass imbalance. At the same time, it transforms itself into a more non-polar compound either due to charge neutralisation (Su et al., 2009) or hydrogen bonding (Rothbard et al., 2004) with the anionic membrane proteins. The peptide then partitions into the hydrophobic lipid bilayer (Rothbard et al., 2002) and translocates through the membrane in a process that is driven by the voltage potential across membrane (Rothbard et al., 2004) and/or the need to relieve membrane curvature stress caused by the mass imbalance in the initial step (Su et al., 2009).

In stark contrast, Richard and colleagues reported that the uptake of tat 48-59, in both its free and PNA-conjugated forms were sensitive to low temperature and sodium azide, indicative of classical endocytosis (Richard et al., 2003). Endocytosis is also believed to be responsible for the uptake of protein-conjugated tat 37-72 in another study (Fawell et al., 1994). Having agreed that the uptake of tat requires energy, opinions are once again divided

Peptides as Promising Non-Viral Vectors for Gene Therapy 631

(Macia et al., 2006). The same is true for methyl-β-cyclodextrin, which depletes cholesterol from the membrane (Richard et al., 2005) and cytochalasin, which disrupts actin formation (Belting et al., 2005). Both drugs can affect all three forms of endocytosis. Moreover, any drug that interferes with uptake is usually toxic and a safe effective dosage must be established in order not to affect the conclusion. There is also possible redundancy in the uptake mechanisms whereby shutting off one pathway may activate another (Rodal et al.,

Both flow cytometry and confocal microscopy are fluorescent-based techniques and rely on either the carrier or/and its cargo to be labelled with a dye. However, the process of tagging the peptide with a dye can already affect its property. For example, results from isothermal titration calorimetry showed that unlabelled tat 47-57 binds to heparan sulphate much more strongly that tat 47-57 labelled with fluorescein isothiocyanate (FITC), a florescent dye whose hydrophobicity has been implicated in the weakening of binding affinity (Ziegler et al., 2005). In another study, FITC labelling was also observed to alter the trafficking property of an octaarginine conjugate (Puckett & Barton, 2009). Significantly, fixing of cells has been shown to be undesirable and even mild fixing can cause artifactual intracellular accumulation (Richard et al., 2003). As a result of this revelation, the validity of earlier publications has been challenged, especially in studies where cells were fixed and intracellular accumulation could still be observed despite low temperature incubation. Surface-bound peptides are another source of confusion which must be separated from those which have been truly internalised (Richard et al., 2003). A brief trypsin wash is

Above all, it is fair to say that our understanding of endocytosis is still not perfect. As mentioned earlier, it is still not clear which other pathways, besides the clathrin-mediated one, can produce vesicles that undergo acidification. Whether vesicles from different origin eventually merge is another unresolved issue. Furthermore, both the size (Choi et al., 2006) and shape (Sharma et al., 2010) can affect the uptake of a particle and its *in vivo* distribution (Decuzzi et al., 2010). However, these physical parameters of tat or other CPP complexes are seldom reported. Great care must also be taken when reporting particle sizes based on light scattering as a recent survey has estimated that up to 90% of the published figures can be erroneous due to inappropriate assumptions being made during measurements (Keck & Muller, 2008). With such a plethora of factors that can affect outcome, it seems more realistic to accept that the uptake of tat or its conjugates cannot be ascribed any particular route.

The potential of using arginine for gene delivery was firmly established when it was shown that replacing most of the arginine residues in tat 1-86 with alanine drastically reduced internalisation (Tyagi et al., 2001). Replacing the arginine residues within tat 48-60 with lysine had a similar effect and abolished internalisation (Thoren et al., 2004). Clearly, arginine residues are crucial. However, another important (but less obvious) message from these experiments is that arginine is more than just a cationic residue as replacing it with

Oligoarginine of various lengths have been evaluated. In one study, oligoarginine with 4, 6, 8, 10, 12 or 16 residues were compared and it was found that both arg4 and arg16 were poorly taken up by cells (Futaki et al., 2001b). Instead, arg8 exhibited the highest rate of internalisation both in its free form or when linked via a disulfide bond to a model carbonic anhydrase protein. This demonstrated that cationic charge was not sufficient to afford

typically enough to digest and remove most of the surface-bound peptides.

1999), further complicating interpretation.

**4.3 Oligoarginine (Argn, usually n = 7 to 9)** 

lysine, another cationic residue, is not good enough.

regarding which of the specific endocytosis pathways involved. For instance, a study singled out clathrin-mediated endocytosis as the pathway responsible for the uptake of unconjugated tat 48-59 (Richard et al., 2005). In others, caveolae-mediated endocytosis was suggested to be the dominant form of uptake for GFP/tat 48-60 fusion protein (Ferrari et al., 2003), and also for GFP/tat 48-58 and GFP/tat 1-86 fusion proteins (Fittipaldi et al., 2003). To confound matters, macropinocytosis was reported to be mainly responsible for the uptake of PNA-conjugated tat 48-60 (Lundin et al., 2008) and tat 48-57 fusion protein (Wadia et al., 2004). In yet another twist, multiple pathways involving both clathrin-mediated endocytosis and macropinocytosis were proposed for the internalisation of unconjugated tat 48-60 (Räägel et al., 2009).

With this amount of confusion in the literature, is there hope of reconciliation? It seems that the first step is to recognise that there may not be a single mechanism responsible for the uptake of tat or indeed, other CPP that will be discussed. CPP thus cannot be taken in the same light as transferrin (Rejman et al., 2004) or cholera toxin (Torgersen et al., 2001), which are accepted to be internalised exclusively via the clathrin- and caveolae-mediated pathway respectively in a wide variety of cells. The outcome being observed therefore depends on the combination of cell/peptide/detection method chosen for that experiment and heterogeneity in conclusion can be expected due to the following reasons.

Cell: Different cell lines can use different pathways to internalise tat (Ignatovich et al., 2003). The nature of cells (primary versus immortalised or adherent versus suspension) (Eiriksdottir et al., 2010) and culture conditions such as the type of growth medium (Moulton et al., 2004), passage number of cells or even the Young's modulus (Kong et al., 2005) of the substrate were suggested to influence plasmid transfection and uptake.

Peptide: As can be seen from Table 2, there are at least 12 distinct sequences under the same umbrella name of tat – more if one takes into account chain-end modifications. Considering that simply modifying a few residues can drastically affect the property of tat, as discussed earlier, it is understandable why the literature cannot seem to agree. The concentration of peptide used was also shown to be important. For instance, it was reported that below 10 mM, unconjugated tat 47-57 exploited both macropinocytosis and caveolae-mediated endocytosis to gain entry but once above 10 mM, a non-endocytosis mechanism seemed to operate (Duchardt et al., 2007). Perhaps there is a concentration threshold above which there is sufficient amount of surface-bound peptide to cause the mass imbalance and trigger direct transport through the membrane. Finally, the absence or presence and the nature of the cargo can also influence the mechanism of uptake. For example, unconjugated tat 47-57 was reported to enter cells using an energy-independent mechanism but endocytosis was responsible after a plasmid was bound (Ignatovich et al., 2003).

Detection method: The majority of experiments aimed at studying the endocytosis pathways of peptides uses one or both of the following techniques: 1) assaying the effects of drugs that are known to inhibit specific endocytosis pathways at the uptake (by flow cytometry or confocal microscopy) and/or the gene expression level, or 2) using a confocal microscope to trace for any co-localisation with known markers for endocytosis pathways, e.g., transferrin (clathrin), cholera toxin (caveolae) or dextran (macropinocytosis). While both are sensible experiments, one must also be aware of their limitations. Inhibitory drugs are usually pleiotropic and affect more than one pathway concurrently. For instance, dynasore has been used to shut down the clathrin-mediated pathway (Gratton et al., 2008a). However, dynasore is a drug that interferes with the activity of dynamin, an enzyme that is needed in both clathrin- and caveolae-mediated pathway to pinch off vesicles from the cell membrane

regarding which of the specific endocytosis pathways involved. For instance, a study singled out clathrin-mediated endocytosis as the pathway responsible for the uptake of unconjugated tat 48-59 (Richard et al., 2005). In others, caveolae-mediated endocytosis was suggested to be the dominant form of uptake for GFP/tat 48-60 fusion protein (Ferrari et al., 2003), and also for GFP/tat 48-58 and GFP/tat 1-86 fusion proteins (Fittipaldi et al., 2003). To confound matters, macropinocytosis was reported to be mainly responsible for the uptake of PNA-conjugated tat 48-60 (Lundin et al., 2008) and tat 48-57 fusion protein (Wadia et al., 2004). In yet another twist, multiple pathways involving both clathrin-mediated endocytosis and macropinocytosis were proposed for the internalisation of unconjugated tat

With this amount of confusion in the literature, is there hope of reconciliation? It seems that the first step is to recognise that there may not be a single mechanism responsible for the uptake of tat or indeed, other CPP that will be discussed. CPP thus cannot be taken in the same light as transferrin (Rejman et al., 2004) or cholera toxin (Torgersen et al., 2001), which are accepted to be internalised exclusively via the clathrin- and caveolae-mediated pathway respectively in a wide variety of cells. The outcome being observed therefore depends on the combination of cell/peptide/detection method chosen for that experiment and

Cell: Different cell lines can use different pathways to internalise tat (Ignatovich et al., 2003). The nature of cells (primary versus immortalised or adherent versus suspension) (Eiriksdottir et al., 2010) and culture conditions such as the type of growth medium (Moulton et al., 2004), passage number of cells or even the Young's modulus (Kong et al.,

Peptide: As can be seen from Table 2, there are at least 12 distinct sequences under the same umbrella name of tat – more if one takes into account chain-end modifications. Considering that simply modifying a few residues can drastically affect the property of tat, as discussed earlier, it is understandable why the literature cannot seem to agree. The concentration of peptide used was also shown to be important. For instance, it was reported that below 10 mM, unconjugated tat 47-57 exploited both macropinocytosis and caveolae-mediated endocytosis to gain entry but once above 10 mM, a non-endocytosis mechanism seemed to operate (Duchardt et al., 2007). Perhaps there is a concentration threshold above which there is sufficient amount of surface-bound peptide to cause the mass imbalance and trigger direct transport through the membrane. Finally, the absence or presence and the nature of the cargo can also influence the mechanism of uptake. For example, unconjugated tat 47-57 was reported to enter cells using an energy-independent mechanism but endocytosis was

Detection method: The majority of experiments aimed at studying the endocytosis pathways of peptides uses one or both of the following techniques: 1) assaying the effects of drugs that are known to inhibit specific endocytosis pathways at the uptake (by flow cytometry or confocal microscopy) and/or the gene expression level, or 2) using a confocal microscope to trace for any co-localisation with known markers for endocytosis pathways, e.g., transferrin (clathrin), cholera toxin (caveolae) or dextran (macropinocytosis). While both are sensible experiments, one must also be aware of their limitations. Inhibitory drugs are usually pleiotropic and affect more than one pathway concurrently. For instance, dynasore has been used to shut down the clathrin-mediated pathway (Gratton et al., 2008a). However, dynasore is a drug that interferes with the activity of dynamin, an enzyme that is needed in both clathrin- and caveolae-mediated pathway to pinch off vesicles from the cell membrane

2005) of the substrate were suggested to influence plasmid transfection and uptake.

heterogeneity in conclusion can be expected due to the following reasons.

responsible after a plasmid was bound (Ignatovich et al., 2003).

48-60 (Räägel et al., 2009).

(Macia et al., 2006). The same is true for methyl-β-cyclodextrin, which depletes cholesterol from the membrane (Richard et al., 2005) and cytochalasin, which disrupts actin formation (Belting et al., 2005). Both drugs can affect all three forms of endocytosis. Moreover, any drug that interferes with uptake is usually toxic and a safe effective dosage must be established in order not to affect the conclusion. There is also possible redundancy in the uptake mechanisms whereby shutting off one pathway may activate another (Rodal et al., 1999), further complicating interpretation.

Both flow cytometry and confocal microscopy are fluorescent-based techniques and rely on either the carrier or/and its cargo to be labelled with a dye. However, the process of tagging the peptide with a dye can already affect its property. For example, results from isothermal titration calorimetry showed that unlabelled tat 47-57 binds to heparan sulphate much more strongly that tat 47-57 labelled with fluorescein isothiocyanate (FITC), a florescent dye whose hydrophobicity has been implicated in the weakening of binding affinity (Ziegler et al., 2005). In another study, FITC labelling was also observed to alter the trafficking property of an octaarginine conjugate (Puckett & Barton, 2009). Significantly, fixing of cells has been shown to be undesirable and even mild fixing can cause artifactual intracellular accumulation (Richard et al., 2003). As a result of this revelation, the validity of earlier publications has been challenged, especially in studies where cells were fixed and intracellular accumulation could still be observed despite low temperature incubation. Surface-bound peptides are another source of confusion which must be separated from those which have been truly internalised (Richard et al., 2003). A brief trypsin wash is typically enough to digest and remove most of the surface-bound peptides.

Above all, it is fair to say that our understanding of endocytosis is still not perfect. As mentioned earlier, it is still not clear which other pathways, besides the clathrin-mediated one, can produce vesicles that undergo acidification. Whether vesicles from different origin eventually merge is another unresolved issue. Furthermore, both the size (Choi et al., 2006) and shape (Sharma et al., 2010) can affect the uptake of a particle and its *in vivo* distribution (Decuzzi et al., 2010). However, these physical parameters of tat or other CPP complexes are seldom reported. Great care must also be taken when reporting particle sizes based on light scattering as a recent survey has estimated that up to 90% of the published figures can be erroneous due to inappropriate assumptions being made during measurements (Keck & Muller, 2008). With such a plethora of factors that can affect outcome, it seems more realistic to accept that the uptake of tat or its conjugates cannot be ascribed any particular route.

#### **4.3 Oligoarginine (Argn, usually n = 7 to 9)**

The potential of using arginine for gene delivery was firmly established when it was shown that replacing most of the arginine residues in tat 1-86 with alanine drastically reduced internalisation (Tyagi et al., 2001). Replacing the arginine residues within tat 48-60 with lysine had a similar effect and abolished internalisation (Thoren et al., 2004). Clearly, arginine residues are crucial. However, another important (but less obvious) message from these experiments is that arginine is more than just a cationic residue as replacing it with lysine, another cationic residue, is not good enough.

Oligoarginine of various lengths have been evaluated. In one study, oligoarginine with 4, 6, 8, 10, 12 or 16 residues were compared and it was found that both arg4 and arg16 were poorly taken up by cells (Futaki et al., 2001b). Instead, arg8 exhibited the highest rate of internalisation both in its free form or when linked via a disulfide bond to a model carbonic anhydrase protein. This demonstrated that cationic charge was not sufficient to afford

Peptides as Promising Non-Viral Vectors for Gene Therapy 633

common feature of many efficient vectors and this was clearly shown in a recent study where a novel triblock peptide, phobic5-his4-arg8, was developed (Seow & Yang, 2009b; Seow et al., 2009). This design featured a block of five hydrophobic residues (tryptophan, phenylalanine or isoleucine, in order of increasing hydrophobicity) for enhanced uptake, a middle block of four histidines for buffering capacity and a third block of arg8 for DNA binding and membrane penetration. Removing the hydrophobic block drastically reduced the efficiency of the carrier at both the uptake and gene expression level. Transfection efficiency can also be modulated just by changing the hydrophobicity of the hydrophobic block. The buffering capacity of the carriers was also assessed in a series of acid-base titration experiments and shown to vary with the length of the histidine block used. Each block was then systematically studied and an optimised sequence was used to mediate reporter plasmid expression by direct local injection into mice bearing 4T1 tumors. In another study, cholesterol was added as the hydrophobic end

The internalisation of oligoarginine is sharply inhibited by the presence of heparin, which competes with heparan sulphate for binding (Seow et al., 2009). This shows that heparan sulphate is crucial for uptake. Low-temperature internalisation was observed and energyindependence was originally proposed to characterise the uptake of oligoarginine (Futaki et al., 2001b). However, those observation was made with fixed cells. In response to studies that had exposed the flaws of using fixed cells (Richard et al., 2003), authors of the original study re-evaluated the uptake mechanism using live cells. Consequently, they reported macropinocytosis to be a major, but not exclusive pathway for the uptake of oligoarginine (Nakase et al., 2004). Macropinocytosis was also proposed to be the main form of uptake for protein-conjugated (Takayama et al., 2009) and siRNA-bound arg8 (Akita et al., 2010). Nonetheless, an energy-independent form of uptake was still suggested if the concentration of unconjugated arg9 exceeded 10 mM – otherwise, macropinocytosis and caveolaemediated, but not clathrin-mediated endocytosis seemed to be important (Duchardt et al., 2007). Another study reported that both clathrin-mediated endocytosis and macropinocytosis play a role for the uptake of unconjugated arg9 (Räägel et al., 2009). On the other hand, all three forms of endocytosis (clathrin-mediated, caveolae-mediated and macropinocytosis) were found to be responsible for the uptake of plasmid-bound oligoarginine complexes (Seow et al., 2009). Finally, an energy-dependent but nonendocytosis mechanism was also proposed to be responsible for uptake (Mitchell et al., 2000). Like tat, there is little consensus over the internalisation pathway(s) of oligoarginine, although endocytosis is generally accepted to play a key role in most cases (Richard et al.,

Homeoproteins are regulatory proteins essential for proper physical development. The DNA binding domain of these proteins is made up of a highly conserved sequence known as the homeobox. The homeobox of the *Drosophila* antennapedia gene (pAntp) is 60-amino acid long (sequence 1-60) and contains three α-helices (Derossi et al., 1994). pAntp was first discovered to effectively translocate into nerve cells and accumulate within their nuclei (Joliot et al., 1991). To demonstrate the usefulness of pAntp as a carrier, a 33-residue peptide cargo corresponding to the C-terminus of rab3, a GTP-binding protein in human, was linked to pAntp via plasmid fusion and shown to translocate into the nuclei of myoblasts and neurons (Perez et al., 1992). A study later revealed that the third helix was actually the domain driving internalisation (Roux et al., 1993). Based on this, penetratin, a 16-residue

of a his5/10-arg10 vector for gene delivery (Guo et al., 2008).

2003; Rothbard et al., 2000).

**4.4 Penetratin** 

efficient uptake. Consistent with this, another study showed that arg7 was internalised much more efficiently into Jurkat cells than lys7 and his7 (Mitchell et al., 2000). There was therefore something unique about the chemical structure of arginine that cannot be explained simply by it carrying a cationic charge. Indeed, the guanidine (pKa~12) sidegroup of arginine was determined to be key as replacing it with a urea (pKa~0.1) sidegroup removed internalisation. Urea differs from guanidine by only a single nitrogen atom, which in the former is replaced by an oxygen atom. However, this simple modification removed not only the ability of urea to protonate at neutral pH, but also its ability to form hydrogen bonds. This ability to form stable hydrogen bonds with the anionic phosphates and sulphates on cell membrane was then suggested to be the feature that distinguished arginine from lysine and histidine in terms of translocation efficiency (Mitchell et al., 2000).

The spacing between arginine residues can also influence internalisation (Rothbard et al., 2002). To study this, a library of oligoarginines was synthesised, all with seven residues but separated by 1-6 spacers at all the possible permutations. It was argued that the addition of spacers imparted flexibility to the arginine backbone which was important for better membrane translocation.

A length of 7-9 arginines is usually preferred and the chains exist as random coils in buffer solutions and when bound to lipid vesicles (Caesar et al., 2006). Both D- and L-arginines have been explored with some authors preferring D-arginines (Hyun et al., 2010; Puckett & Barton, 2009) and others finding no difference between the stereoisomers (Mitchell et al., 2000; Nakase et al., 2004). Cargoes of different nature have been delivered. Negatively charged QCs were electrostatically bound to arg8 and delivered into adipose tissue-derived stem cells for imaging purposes (Yukawa et al., 2010). The anti-cancer drug, taxol was covalently bound to arg8 via disulfide bonds using a novel linker to increase its water solubility and uptake. The drug was also designed to be released intracellularly so as to combat multidrug resistant cell lines which otherwise have limited accumulation of chemotherapeutic drugs (Dubikovskaya et al., 2008). The immunosuppressive drug, cyclosporine A, was coupled with arg7 using a novel pH-sensitive linker and used for topical delivery in a skin inflammation model (Rothbard et al., 2000). Phosphorodiamidate morpholino oligomers (PMO) are antisense molecules that interfere with mRNA translation but structurally differ from nucleic acids in several aspects. In one study, PMO was electrostatically carried by an arg9-based carrier, but the efficiency was not as good as covalently linked ones (Moulton et al., 2004). In another study, a short peptide sequence corresponding to the C-terminus of the cystic fibrosis transmembrane regulator was joined to arg7 during synthesis and laser illumination was used to trigger the release of the conjugate into the cytosol of U2OS cells (Maiolo-III et al., 2004). siRNA was electrostatically carried by an arg8-based vector and used to transfect mouse bone marrow-derived dendritic cells (Akita et al., 2010). Finally, D-arg9 with cysteines added on both terminals (i.e., cys-arg9-cys) was crosslinked via disulfide bonds to form a mesh of reducible poly(oligoarginine) (Hyun et al., 2010). The carrier was then used to deliver plasmids encoding for heme oxygenase-1 (useful for the treatment of ischemia/reperfusion-induced brain stroke) by direct injection into the brain of mice.

Several modifications have been made to oligoarginine to improve its transfection efficiency. For example, arg8 has been combined with GALA, a pH sensitive fusogenic peptide to enhance its endosomal escaping property (Akita et al., 2010). The addition of a hydrophobic stearyl chain to arg8 was also shown to greatly increase its transfection ability (Futaki et al., 2001a). This was suggested to be due to a better association between the hydrophobic moieties on the vector and the lipid bilayer (Putnam, 2006). Indeed, a certain degree of hydrophobicity is a

efficient uptake. Consistent with this, another study showed that arg7 was internalised much more efficiently into Jurkat cells than lys7 and his7 (Mitchell et al., 2000). There was therefore something unique about the chemical structure of arginine that cannot be explained simply by it carrying a cationic charge. Indeed, the guanidine (pKa~12) sidegroup of arginine was determined to be key as replacing it with a urea (pKa~0.1) sidegroup removed internalisation. Urea differs from guanidine by only a single nitrogen atom, which in the former is replaced by an oxygen atom. However, this simple modification removed not only the ability of urea to protonate at neutral pH, but also its ability to form hydrogen bonds. This ability to form stable hydrogen bonds with the anionic phosphates and sulphates on cell membrane was then suggested to be the feature that distinguished arginine from lysine

The spacing between arginine residues can also influence internalisation (Rothbard et al., 2002). To study this, a library of oligoarginines was synthesised, all with seven residues but separated by 1-6 spacers at all the possible permutations. It was argued that the addition of spacers imparted flexibility to the arginine backbone which was important for better

A length of 7-9 arginines is usually preferred and the chains exist as random coils in buffer solutions and when bound to lipid vesicles (Caesar et al., 2006). Both D- and L-arginines have been explored with some authors preferring D-arginines (Hyun et al., 2010; Puckett & Barton, 2009) and others finding no difference between the stereoisomers (Mitchell et al., 2000; Nakase et al., 2004). Cargoes of different nature have been delivered. Negatively charged QCs were electrostatically bound to arg8 and delivered into adipose tissue-derived stem cells for imaging purposes (Yukawa et al., 2010). The anti-cancer drug, taxol was covalently bound to arg8 via disulfide bonds using a novel linker to increase its water solubility and uptake. The drug was also designed to be released intracellularly so as to combat multidrug resistant cell lines which otherwise have limited accumulation of chemotherapeutic drugs (Dubikovskaya et al., 2008). The immunosuppressive drug, cyclosporine A, was coupled with arg7 using a novel pH-sensitive linker and used for topical delivery in a skin inflammation model (Rothbard et al., 2000). Phosphorodiamidate morpholino oligomers (PMO) are antisense molecules that interfere with mRNA translation but structurally differ from nucleic acids in several aspects. In one study, PMO was electrostatically carried by an arg9-based carrier, but the efficiency was not as good as covalently linked ones (Moulton et al., 2004). In another study, a short peptide sequence corresponding to the C-terminus of the cystic fibrosis transmembrane regulator was joined to arg7 during synthesis and laser illumination was used to trigger the release of the conjugate into the cytosol of U2OS cells (Maiolo-III et al., 2004). siRNA was electrostatically carried by an arg8-based vector and used to transfect mouse bone marrow-derived dendritic cells (Akita et al., 2010). Finally, D-arg9 with cysteines added on both terminals (i.e., cys-arg9-cys) was crosslinked via disulfide bonds to form a mesh of reducible poly(oligoarginine) (Hyun et al., 2010). The carrier was then used to deliver plasmids encoding for heme oxygenase-1 (useful for the treatment of ischemia/reperfusion-induced

Several modifications have been made to oligoarginine to improve its transfection efficiency. For example, arg8 has been combined with GALA, a pH sensitive fusogenic peptide to enhance its endosomal escaping property (Akita et al., 2010). The addition of a hydrophobic stearyl chain to arg8 was also shown to greatly increase its transfection ability (Futaki et al., 2001a). This was suggested to be due to a better association between the hydrophobic moieties on the vector and the lipid bilayer (Putnam, 2006). Indeed, a certain degree of hydrophobicity is a

and histidine in terms of translocation efficiency (Mitchell et al., 2000).

brain stroke) by direct injection into the brain of mice.

membrane translocation.

common feature of many efficient vectors and this was clearly shown in a recent study where a novel triblock peptide, phobic5-his4-arg8, was developed (Seow & Yang, 2009b; Seow et al., 2009). This design featured a block of five hydrophobic residues (tryptophan, phenylalanine or isoleucine, in order of increasing hydrophobicity) for enhanced uptake, a middle block of four histidines for buffering capacity and a third block of arg8 for DNA binding and membrane penetration. Removing the hydrophobic block drastically reduced the efficiency of the carrier at both the uptake and gene expression level. Transfection efficiency can also be modulated just by changing the hydrophobicity of the hydrophobic block. The buffering capacity of the carriers was also assessed in a series of acid-base titration experiments and shown to vary with the length of the histidine block used. Each block was then systematically studied and an optimised sequence was used to mediate reporter plasmid expression by direct local injection into mice bearing 4T1 tumors. In another study, cholesterol was added as the hydrophobic end of a his5/10-arg10 vector for gene delivery (Guo et al., 2008).

The internalisation of oligoarginine is sharply inhibited by the presence of heparin, which competes with heparan sulphate for binding (Seow et al., 2009). This shows that heparan sulphate is crucial for uptake. Low-temperature internalisation was observed and energyindependence was originally proposed to characterise the uptake of oligoarginine (Futaki et al., 2001b). However, those observation was made with fixed cells. In response to studies that had exposed the flaws of using fixed cells (Richard et al., 2003), authors of the original study re-evaluated the uptake mechanism using live cells. Consequently, they reported macropinocytosis to be a major, but not exclusive pathway for the uptake of oligoarginine (Nakase et al., 2004). Macropinocytosis was also proposed to be the main form of uptake for protein-conjugated (Takayama et al., 2009) and siRNA-bound arg8 (Akita et al., 2010). Nonetheless, an energy-independent form of uptake was still suggested if the concentration of unconjugated arg9 exceeded 10 mM – otherwise, macropinocytosis and caveolaemediated, but not clathrin-mediated endocytosis seemed to be important (Duchardt et al., 2007). Another study reported that both clathrin-mediated endocytosis and macropinocytosis play a role for the uptake of unconjugated arg9 (Räägel et al., 2009). On the other hand, all three forms of endocytosis (clathrin-mediated, caveolae-mediated and macropinocytosis) were found to be responsible for the uptake of plasmid-bound oligoarginine complexes (Seow et al., 2009). Finally, an energy-dependent but nonendocytosis mechanism was also proposed to be responsible for uptake (Mitchell et al., 2000). Like tat, there is little consensus over the internalisation pathway(s) of oligoarginine, although endocytosis is generally accepted to play a key role in most cases (Richard et al., 2003; Rothbard et al., 2000).

#### **4.4 Penetratin**

Homeoproteins are regulatory proteins essential for proper physical development. The DNA binding domain of these proteins is made up of a highly conserved sequence known as the homeobox. The homeobox of the *Drosophila* antennapedia gene (pAntp) is 60-amino acid long (sequence 1-60) and contains three α-helices (Derossi et al., 1994). pAntp was first discovered to effectively translocate into nerve cells and accumulate within their nuclei (Joliot et al., 1991). To demonstrate the usefulness of pAntp as a carrier, a 33-residue peptide cargo corresponding to the C-terminus of rab3, a GTP-binding protein in human, was linked to pAntp via plasmid fusion and shown to translocate into the nuclei of myoblasts and neurons (Perez et al., 1992). A study later revealed that the third helix was actually the domain driving internalisation (Roux et al., 1993). Based on this, penetratin, a 16-residue

Peptides as Promising Non-Viral Vectors for Gene Therapy 635

(Pooga et al., 1998) but more recent observations have linked endocytosis to the uptake of

VP22 is a 300-residue long peptide derived from the structural protein of HSV-1 and has been used successfully to deliver GFP as a fusion protein (Elliott & O'Hare, 1997). When COS-1 cells were microinjected with plasmids encoding for VP22, an interesting pattern was observed in which a central cell would first express VP22 and exhibit cytosolic staining when probed with anti-VP22 antibodies. The VP22 peptide was then excreted and could infect neighbouring cells before eventually localising to their nuclei. CADY is a 20-residue long peptide evaluated for siRNA delivery (Crombez et al., 2009). It changes from unordered to being α-helical in the presence of lipid vesicles which then drives its internalisation independently of the endosomal pathway (Konate et al., 2010). MPG is a 27 residue peptide designed to have a hydrophobic domain (sequence 1-17) derived from HIV gp41 and a NLS domain (sequence 21-27) derived from SV40 separated by a "trp-ser-gln" unit (Morris et al., 1997). The uptake of MPG/oligonucleotide complexes was shown to be rapid and independent of the endosomal pathway. Other carriers explored for delivery

Throughout the discussion, various peptides with fusogenic and nuclear localising properties were introduced. SV40 is the classical NLS and different sequences exist. The most commonly used sequence for SV40 is provided in Table 1. Other sequences include "glu-asp-pro-tyr" (Trentin et al., 2005) and "glu-pro-tyr-cys" (Moore et al., 2009) being added onto the C-terminus and an even longer form of SV40 has been described (Eguchi et al., 2001). INF 5 and INF 7 are examples of two commonly used fusogenic peptide and both are derived from the parent HA2 peptide (Plank et al., 1994). GALA is a 30-amino acid long pH-sensitive fusogenic peptide (Subbarao et al., 1987). Upon protonation of its glu residues, its secondary structure changes from unordered at neutral pH to being α-helical in acidic environments. Both the amphipathicity and degree of α-helicity have been correlated with the ability of GALA to interact and destabilise membranes (Parente

This chapter started by discussing the challenges and intracellular barriers associated with the delivery of nucleic acids. Strategies used to overcome these hurdles were next examined, mainly in the context of peptide-derived vectors. It is clear that peptide carriers are not characterised by any typical sequences, although the majority of current designs rely on cationic residues to bind nucleic acids. This strategy, nonetheless, may be problematic during *in vivo* applications. An area that deserves more attention is the development of vectors that can bind nucleic acids using non-electrostatic forces, e.g., by including more hydrophobic residues. The stability and *in vivo* behaviour of such vectors then need to be thoroughly evaluated. Another challenge is to further improve strategies that are already in place to shield the cationic charges, e.g., by attaching PEG chains that are intracellularly cleavable so as to increase the circulation time of the complexes without compromising excessively on transfection efficiency. Advances in either will enable peptides to further

include KALA (Wyman et al., 1997) and EB1 (Lundberg et al., 2007).

both Tp and Tp10 (Lundin et al., 2008).

**4.6 Other notable peptide sequences** 

et al., 1990).

**5. Conclusion** 

realise their potential as a class of non-viral vector.

peptide within the third helix (sequence 43-58 of the original pAntp) was described (Derossi et al., 1994). Penetratin demonstrated membrane penetrating ability but was prone to aggregation. Aggregated particles can enjoy better internalisation due to a sedimentation effect which promotes a more intimate particle-cell contact (Luo & Saltzman, 2000b). While this may have helped the internalisation of penetratin, aggregation alone was excluded as the dominant reason for its efficient internalisation (Derossi et al., 1994).

The secondary structure of penetratin in buffer is unstructured but becomes α-helical when bound to lipid vesicles (Caesar et al., 2006). Raman microscopy on live cells further showed that penetratin within the cytosol was either unstructured or in the β-sheet conformation (Ye et al., 2010). The arginine residues within penetratin are important as replacing them with lysine greatly reduced its translocation ability (Caesar et al., 2006). Various cargoes have been carried by penetratin. Antisense oligonucleotides (Astriab-Fisher et al., 2000), PNA (Lundin et al., 2008) and luciferin protein (Eiriksdottir et al., 2010) have all been coupled onto a cysteine-modified penetratin via disulfide bonds and delivered into cells. siRNA was electrostatically bound to penetratin and although the complexes accumulated favourably within cells, there was limited silencing activity (Lundberg et al., 2007). HA2 was then added to enhance endosomal escape but produced only a modest improvement. Penetratin has been evaluated *in vivo*. In one study, penetratin was directly injected into a rat's brain and the peptide was observed to spread away from the site of injection (Bolton et al., 2000). However, dosage-dependent cell death and inflammatory responses were also provoked.

Penetratin was initially suggested to enter cells using energy-independent mechanisms (Derossi et al., 1994; Perez et al., 1992; Roux et al., 1993), but a caveat is that fixed cells were used in those studies. Nonetheless, an energy-independent mechanism was still suggested to be possible past a concentration threshold of 40 mM, below which all three forms of endocytosis should dominate (Duchardt et al., 2007). Endocytosis was also reported to be responsible for the uptake of siRNA-bound penetratin (Lundberg et al., 2007) while macropinocytosis was suggested to be the main form of uptake for PNA-conjugated penetratin (Lundin et al., 2008).

#### **4.5 Transportan (Tp) and Tp10**

Tp uses a lysine residue to join the first 12 N-terminus residues of the neuropeptide, galanin, to the 14 C-terminus residues of the wasp venom, mastoparan. Tp is thus 27 residues long (sequence 1-27) and was first shown to penetrate Bowes' melanoma cells rapidly and efficiently (Pooga et al., 1998). However, Tp at a high concentration was found to inhibit the GTPase activity of cells. To overcome this side effect, a series of Tp analogues was prepared (Soomets et al., 2000). This led to the identification of Tp10 which, upon deleting the first 6 residues of Tp (i.e., sequence 7-27 remaining), was internalised as efficiently as Tp but did not have any effect on the GTPases.

PNA (Lundin et al., 2008) and luciferin (Eiriksdottir et al., 2010) were conjugated to both cysteine-modified Tp and Tp10 via disulfide bonds. siRNA was also electrostatically bound to Tp10 but had little silencing effects when transfected into cells (Lundberg et al., 2007). This was surprising insofar as Tp10 was shown earlier to mediate good levels of plasmid expression. This however resonates with comments made by other researchers (Mahon et al., 2010; unpublished observation) who had observed that a carrier's efficiency in plasmid delivery may not automatically apply to the delivery of the shorter and stiffer siRNA molecules. Earlier studies with fixed cells reported low temperature internalisation of Tp (Pooga et al., 1998) but more recent observations have linked endocytosis to the uptake of both Tp and Tp10 (Lundin et al., 2008).

#### **4.6 Other notable peptide sequences**

634 Non-Viral Gene Therapy

peptide within the third helix (sequence 43-58 of the original pAntp) was described (Derossi et al., 1994). Penetratin demonstrated membrane penetrating ability but was prone to aggregation. Aggregated particles can enjoy better internalisation due to a sedimentation effect which promotes a more intimate particle-cell contact (Luo & Saltzman, 2000b). While this may have helped the internalisation of penetratin, aggregation alone was excluded as

The secondary structure of penetratin in buffer is unstructured but becomes α-helical when bound to lipid vesicles (Caesar et al., 2006). Raman microscopy on live cells further showed that penetratin within the cytosol was either unstructured or in the β-sheet conformation (Ye et al., 2010). The arginine residues within penetratin are important as replacing them with lysine greatly reduced its translocation ability (Caesar et al., 2006). Various cargoes have been carried by penetratin. Antisense oligonucleotides (Astriab-Fisher et al., 2000), PNA (Lundin et al., 2008) and luciferin protein (Eiriksdottir et al., 2010) have all been coupled onto a cysteine-modified penetratin via disulfide bonds and delivered into cells. siRNA was electrostatically bound to penetratin and although the complexes accumulated favourably within cells, there was limited silencing activity (Lundberg et al., 2007). HA2 was then added to enhance endosomal escape but produced only a modest improvement. Penetratin has been evaluated *in vivo*. In one study, penetratin was directly injected into a rat's brain and the peptide was observed to spread away from the site of injection (Bolton et al., 2000). However, dosage-dependent cell death

Penetratin was initially suggested to enter cells using energy-independent mechanisms (Derossi et al., 1994; Perez et al., 1992; Roux et al., 1993), but a caveat is that fixed cells were used in those studies. Nonetheless, an energy-independent mechanism was still suggested to be possible past a concentration threshold of 40 mM, below which all three forms of endocytosis should dominate (Duchardt et al., 2007). Endocytosis was also reported to be responsible for the uptake of siRNA-bound penetratin (Lundberg et al., 2007) while macropinocytosis was suggested to be the main form of uptake for PNA-conjugated

Tp uses a lysine residue to join the first 12 N-terminus residues of the neuropeptide, galanin, to the 14 C-terminus residues of the wasp venom, mastoparan. Tp is thus 27 residues long (sequence 1-27) and was first shown to penetrate Bowes' melanoma cells rapidly and efficiently (Pooga et al., 1998). However, Tp at a high concentration was found to inhibit the GTPase activity of cells. To overcome this side effect, a series of Tp analogues was prepared (Soomets et al., 2000). This led to the identification of Tp10 which, upon deleting the first 6 residues of Tp (i.e., sequence 7-27 remaining), was internalised as efficiently as Tp but did

PNA (Lundin et al., 2008) and luciferin (Eiriksdottir et al., 2010) were conjugated to both cysteine-modified Tp and Tp10 via disulfide bonds. siRNA was also electrostatically bound to Tp10 but had little silencing effects when transfected into cells (Lundberg et al., 2007). This was surprising insofar as Tp10 was shown earlier to mediate good levels of plasmid expression. This however resonates with comments made by other researchers (Mahon et al., 2010; unpublished observation) who had observed that a carrier's efficiency in plasmid delivery may not automatically apply to the delivery of the shorter and stiffer siRNA molecules. Earlier studies with fixed cells reported low temperature internalisation of Tp

the dominant reason for its efficient internalisation (Derossi et al., 1994).

and inflammatory responses were also provoked.

penetratin (Lundin et al., 2008).

**4.5 Transportan (Tp) and Tp10** 

not have any effect on the GTPases.

VP22 is a 300-residue long peptide derived from the structural protein of HSV-1 and has been used successfully to deliver GFP as a fusion protein (Elliott & O'Hare, 1997). When COS-1 cells were microinjected with plasmids encoding for VP22, an interesting pattern was observed in which a central cell would first express VP22 and exhibit cytosolic staining when probed with anti-VP22 antibodies. The VP22 peptide was then excreted and could infect neighbouring cells before eventually localising to their nuclei. CADY is a 20-residue long peptide evaluated for siRNA delivery (Crombez et al., 2009). It changes from unordered to being α-helical in the presence of lipid vesicles which then drives its internalisation independently of the endosomal pathway (Konate et al., 2010). MPG is a 27 residue peptide designed to have a hydrophobic domain (sequence 1-17) derived from HIV gp41 and a NLS domain (sequence 21-27) derived from SV40 separated by a "trp-ser-gln" unit (Morris et al., 1997). The uptake of MPG/oligonucleotide complexes was shown to be rapid and independent of the endosomal pathway. Other carriers explored for delivery include KALA (Wyman et al., 1997) and EB1 (Lundberg et al., 2007).

Throughout the discussion, various peptides with fusogenic and nuclear localising properties were introduced. SV40 is the classical NLS and different sequences exist. The most commonly used sequence for SV40 is provided in Table 1. Other sequences include "glu-asp-pro-tyr" (Trentin et al., 2005) and "glu-pro-tyr-cys" (Moore et al., 2009) being added onto the C-terminus and an even longer form of SV40 has been described (Eguchi et al., 2001). INF 5 and INF 7 are examples of two commonly used fusogenic peptide and both are derived from the parent HA2 peptide (Plank et al., 1994). GALA is a 30-amino acid long pH-sensitive fusogenic peptide (Subbarao et al., 1987). Upon protonation of its glu residues, its secondary structure changes from unordered at neutral pH to being α-helical in acidic environments. Both the amphipathicity and degree of α-helicity have been correlated with the ability of GALA to interact and destabilise membranes (Parente et al., 1990).
