1. Introduction
Today there is a fast growing number of nucleic acid-based strategies to modulate a vast variety of cellular functions (for a review see: [1,2]). Several classes of oligonucleotides like aptamers, transcription factor-binding decoy oligonucleotides, ribozymes, triplex-forming oligonucleotides (TFO), immunostimulatory CpG motifs, antisense oligonucleotides including peptide nucleic acids (PNAs), small interfering RNAs (siRNAs) and antagomirs have attained much interest as a research tool owing to their highly specific mode of action. Even more important, these oligomeric nucleic acids do have a considerable potential to be used as therapeutics. Figure 1 provides an overview of such oligonucleotides and their target sites within the cell.
Although quite different in their mode of action, oligomeric nucleic acids have several features in common. Essentially, they can either be rationally designed (e.g. antisense oligonucleotides or siRNAs) or selected in vitro (e.g. aptamers or ribozymes). These are major advantages compared to traditional small molecule drug screening approaches. In general, these macromolecules show remarkably high specificity for their targets accompanied by low probability of generating side-effects. Additionally, nucleic acids are virtually non-immunogenic compared to protein- or peptide-based approaches. On the downside, considerations like stability, bio-availability and pharmacokinetics come into play. Though, these drawbacks can be resolved by appropriate chemical modifications. Nuclease resistance for instance can be achieved by alkyl modifications at the 2′-position of the ribose. In recent years, valuable progress has been accomplished through the development of novel chemically modified nucleotides with improved properties such as enhanced serum stability, higher target affinity and low toxicity. Pharmacokinetic parameters can be rationally improved by increasing the molecular size, e.g. by site-specific conjugation of polyethylene glycol (PEG). In spite of this, the major obstacle for turning oligomeric nucleic acids into drugs is efficient cellular delivery. Due to this limitation, for a long time nucleic acids were not considered to have a significant therapeutic prospective, though their efficiency has been proven by countless in vitro studies. This obvious dilemma urgently calls for safe and efficient nucleic acid delivery systems.
Essentially, the nucleic acid delivery techniques available today comprise various physical and chemical methods, viral and non-viral vector systems, and uptake of naked nucleic acids (Figure 2). They all have certain advantages and disadvantages and might only be appropriate if particular requirements are fulfilled. In general, physical and chemical methods like microinjection, electroporation or particle bombardment as well as calcium phosphate coprecipitation are highly efficient but rather harmful for the target cells and lack the potential to be applicable in vivo. There is general consent that viral vector systems are the most efficient vehicles to deliver nucleic acids into cells. However, despite substantial efforts over the last 15 years, up to now research has failed to develop suitable and especially safe viral systems (for a review see: [3,4]). On the contrary, the field has experienced several setbacks causing important clinical trials to be put on hold [5–8]. As a result of the difficulties encountered with these viral vectors (e.g. mutagenesis and immune responses) much attention was paid to the development of allegedly safer non-viral delivery systems. This conception includes an assortment of fairly unrelated approaches yielding various degrees of enhanced cellular uptake of nucleic acids. Currently, liposomes and cationic polymers are used as a standard tool to transfect cells in vitro. These approaches are yet characterised by a significant lack of efficiency accompanied by a high level of toxicity rendering them mostly inadequate for in vivo applications.
Peptides acting as shuttles for a controlled cellular delivery of nucleic acids represent a new and innovative concept to bypass the problem of poor bio-availability and clinical efficacy of such macromolecules. The idea of using peptides as carriers goes back some 20 years when two groups discovered by chance that the HIV-1 transactivating protein Tat is taken up by mammalian cells [10,11]. Just a few years later, the Antennapedia homeodomain of Drosophila melanogaster was shown to act similarly [12]. Later on, it could be shown that peptides derived from Tat and Antennapedia as well as other proteins are capable of transporting macromolecular cargo molecules into cells [13–15]. Based on such promising results, a rapidly expanding field focusing on the so-called cell-penetrating peptides (CPPs), also referred to as protein transduction domains (PTD) began to develop.
In this review we will report about recent progress in the field of peptide-mediated delivery of nucleic acids, highlighting the development of several new CPPs, and discuss mechanisms for cellular internalisation. Additionally, we will present own data on peptide-mediated siRNA delivery and briefly discuss them in the given context.
3. Mechanism of cellular translocation
Despite the widespread interest in peptide carriers the mechanisms underlying the cellular translocation of CPPs are poorly understood and subject to controversial discussions (Figure 3). Early work relied upon fluorescence imaging or flow cytometry analysis of chemically fixed cells to examine intracellular localisation of fluorescently labelled peptides in the absence or presence of cargo. According to these experiments peptides seemed to be internalised very rapidly within minutes even at 4 °C. From such observations it was concluded that CPPs penetrate cell membranes by an energy-independent mechanism [18–22]. Although it had been reported quite early on that certain fixation procedures may cause artefacts leading to an overestimation of the cellular uptake rates [23–25] the dimension of this problem was not commonly recognised until a side by side comparison of fixed and living cells was published [26]. The authors convincingly demonstrated that the distribution pattern of Tat48–60 and R9 as well as their conjugates was completely different in living versus fixed cells. They stress that an important source of misinterpretation arises from experimental difficulties to distinguish cell surface-associated CPP from CPP internalised into cytoplasmic compartments. Most notably it could be shown that endocytotic transport is a key component of the internalisation process of a Tat derived peptide [26] and penetratin [27]. Prior to endocytosis CPPs interact electrostatically with the extracellular matrix of the cell surface mostly through binding to negatively charged glycosaminoglycans, i.e. heparan sulfate proteoglycans [28–31].
Eukaryotic cells take up macromolecules and particles from the surrounding medium by a process called endocytosis. Besides phagocytosis, which is only relevant for specialised cells like macrophages, this process occurs in all cells by at least four basic mechanisms: macropinocytosis, clathrin-mediated endocytosis, caveolin-mediated endocytosis, and clathrin- and caveolin-independent endocytosis (Figure 4; for a review see: [32,33]). These mechanistically diverse and highly regulated endocytic pathways contribute to control complex physiological processes, such as hormone-mediated signal transduction, immune surveillance, antigen-presentation, and homeostasis. The best-studied endocytosis pathway is the clathrin-coated pit pathway (for a review see: [34]). Many receptors and their associated ligands cluster into clathrin-coated pits by association with clathrin adaptor proteins such as the four-subunit complex AP2. Clathrin adaptors in turn bind to the clathrin lattice which is thought to provide the force required to deform the membrane into a curved bud. The large GTPase dynamin is then involved in pinching off the coated pit to form a clathrin-coated vesicle. Such vesicles are then uncoated by the chaperone hsc70 and the DNA-J domain co-chaperone auxillin. Uncoated endocytic vesicles then fuse with one another and with early endosomes in a reaction requiring the small GTPase Rab5. Eventually, early to late endosome transport may be mediated by small vesicular intermediates. Late endosomes are then thought to fuse with pre-lysosomes to form ‘hybrid’ organelles which mature back into lysosomes through sorting and fission.
The second best studied pathway depends on caveolin (for a review see: [36]). Oligomerised caveolin associates at the plasma membrane with proteins, e.g. PTRF-cavin [37], and lipids to form stable functional units referred to as caveolae. This caveolar unit is also maintained as a stable structure upon endocytosis. Budded caveolae can fuse either with the caveosome in a Rab5-independent manner or with the early endosome in a Rab5-dependent manner. Caveolae remain at the plasma membrane for long periods, but various agents can stimulate their internalisation. These include the SV40 virus as well as sterols and glycosphingolipids which use caveolae for entry into cells and stimulate caveolar budding. Despite the different agents used to stimulate caveolar internalisation there are common mechanisms involved in these pathways with a crucial role for dynamin, Src kinases, protein kinase C and actin recruitment. Besides these two pathways, there are other less well-defined pathways (i.e. clathrin- and caveolin-independent, lipid-raft endocytosis) that differ from the routes described above [32].
As most of these pathways have several characteristics in common, it remains complicated to precisely determine which endosomal route is followed [38–40]. This in part can be attributed to the fact that endocytosis itself is far from being understood in detail. Differentiation between alternative endocytosis pathways involved in cellular uptake of cargo molecules can be accomplished by the use of small molecule inhibitors or by colocalisation studies with specific markers following fluorescence microscopy or cell fractionation. As the fate of the cargo (e.g. degradation in lysosomes, storage in caveosomes or transport to the Golgi complex or the ER etc.) largely depends on which endocytosis pathway is involved in uptake, it would be highly desirable to induce internalisation via a particular pathway in order to better control trafficking through the cell. Currently, this is not easily achieved. However, there are certain parameters which influence the mode of uptake like particle size, the interaction of the complexes with cell surface components as well as the physiological state of the target cells. In that respect there are several options to manipulate cell trafficking.
Based on the findings described above, many groups began to re-examine their data. However, despite considerable technical improvements, there are still puzzling controversial results concerning the exact mechanism of CPP uptake. Though in most cases endocytosis has been suggested to be the main route of internalisation, substantial difficulties are encountered identifying the exact pathway. Recent studies indicate that the uptake mechanism of CPPs can be influenced by the attachment of cargos. For example, Richard et al. [26,41] reported a co-localisation of Tat48–59 with markers of clathrin-mediated endocytosis, whereas Fittipaldi et al. [42] found a caveolae/lipid raft-dependent process for a Tat-GFP fusion protein and Wadia et al. [43] described a macropinocytotic uptake pathway for a fusion construct of Tat peptide with Cre recombinase.
To avoid the complexity inherent to cell culture-based experiments, biophysical investigations of CPP/membrane interactions with different model membranes were performed. According to these kind of studies, a direct penetration of several CPPs through the lipid layer was proposed [44–48]. However, it is highly problematic to transfer these data to the cellular level, as natural membranes are composed of a heterogeneous mixture of proteins and lipids including distinctive microdomains such as lipid rafts (microdomain enriched in cholesterol and sphingolipids [49]). What is even more important, cells are surrounded by an extracellular matrix consisting of miscellaneous proteins and receptors. Thus, it is obvious that simple model membranes do not reflect this situation. Nevertheless, such studies could in part mimic processes taking place at endosomal membranes and might help to understand how CPPs escape from these vesicles.
Taken together, it has to be concluded that the precise mechanism of internalisation remains elusive and strongly depends on the properties of both CPP and cargo as well as on the transfection conditions and the cell lines used [50–54]. Furthermore, it is reasonable to speculate about multiple pathways involved in cellular entry and that when blocking one of them an alternative pathway may become active [54].
5. CPP-mediate delivery of nucleic acids
In the following section we will focus on CPP-mediated cellular transport of oligomeric nucleic acids including siRNA, antisense, decoy and triplex forming oligonucleotides. Selected examples of the most recent literature will be briefly described. Additionally, examples of plasmid delivery are listed owing to their importance as vehicles for intracellular expression of small RNAs. Table 3 to 5 give an overview of all the approaches illustrated. Another important group of oligomeric nucleic acids are aptamers. These small oligonucleotides derived from an in vitro evolution process called SELEX are promising therapeutic and diagnostic agents. Yet, to our knowledge there are no convincing reports about peptide-mediated delivery of functional aptamers yielding observable biological effects in cellular systems.
5.1. Plasmids
To date there are many examples of CPP-mediated delivery of plasmid DNA into cultured cells and also in vivo. Due to the size of plasmids and the resulting high number of negative charges, only a non-covalent approach has proved feasible.
It has been shown that Tat peptides bind to DNA as well as other polyanions to form complexes which then interact with the membrane of different cells followed by internalisation through endocytosis [91]. According to microscopic studies, Tat peptide/DNA complexes accumulated in acidic vesicles from which they were eventually set free. Applying a similar approach, Ignatovich et al. [92] observed moderate reporter gene expression after incubation of cultured cells with Tat peptide/plasmid complexes. On the other hand, intravenous injection of such complexes into mice yielded only very low expression levels, predominantly in the liver. The distribution of plasmid DNA and the expression levels did not differ significantly from those obtained with naked DNA [92]. This was attributed mainly to non-specific hepatic uptake of macromolecular compounds as well as rapid clearance due to interactions with serum albumin [191,192]. Furthermore, Rudolph et al. [93] presented data supporting a non-covalent plasmid DNA complex formation with oligomeric Tat47–57. Here, dimers and trimers of Tat47–57 were found to be more efficient than tetramers. Another intriguing study compares the potential of high molecular weight forms of Tat and Tat peptides to form stable non-covalent cell transfecting complexes with plasmid DNA [193]. This so-called POLYTAT consists of a mixture of linear polymers of Tat peptide molecules cross linked by disulfide bridges. The diameter of such complexes was determined to be in the range of 200 nm. In contrast, complexes consisting of monomeric Tat peptides and plasmid DNA tended to aggregate into much bigger particles. POLYTAT yielded about 100-fold increased transfection rates as compared to monomeric Tat. This could be further increased by the addition of the lysosomotropic reagent chloroquine to even exceed the level of PEI-mediated transfection. The authors hypothesised that POLYTAT promotes the formation of more stable complexes that expose free basic amino acids on the surface, thereby increasing transfection efficiency [193]. This system has been further developed to form reducible layer-by-layer films with plasmid DNA in a self assembly process, which has yet to be tested for delivery [194]. Though, the idea to construct a carrier system out of several Tat peptide monomers was not new. Some years before, in 2002, it was found that branched peptides containing 8 Tat moieties had considerable transfection potential [195]. Further investigations with this branched 8Tat in the presence of chloroquine revealed that in different cell lines alternative pathways of intracellular trafficking might be relevant [196].
For R8-mediated plasmid delivery, low transfection efficiencies could be increased by two orders of magnitude by the attachment of a stearyl group to the N-terminus [197]. Probably due to the amphipathic character, these modified peptides reached the same level of plasmid transfection as LF and similar results were obtained for Tat48–60. Here, endocytosis was proposed to be the mechanism of uptake. Another example is the peptide MPG that interacts through its positively charged NLS sequence with the negatively charged plasmid DNA. Such peptide/nucleic acid complexes have been reported to be taken up efficiently by mammalian cells [130]. The non-covalent complexation of the prion protein derived peptide bPrPp with plasmid DNA led to endocytotic uptake of the particles and protein expression from the plasmid. The expression levels, however, were much lower than measured after LF-mediated transfection even with peptide concentrations up to 29 μM [161]. Similarly, none of the branched human calcitonin derivatives hCT9–32-2br and hCT18–32-k7 could reach the transfection rate of Lipofectamine™ 2000 (LF2000) in the cell lines HEK 293, HeLa, MCT-7, COS-7 and SK-N-MC or in primary cells. Although the branched peptides were more efficient than Tat48–60, the transfection rate was only about 40 % of that seen with LF2000, even when tested in the presence of 125 μM chloroquine and at the optimal peptide/plasmid charge ratio (30:1). Fluorescence microscopic localisation studies led to the conclusion that both branched CPPs delivered the plasmids into all cells via endocytosis with an unsatisfying extent of endosomal escape. Interestingly, the protein expression level in neuroblastoma cells was 1.5-fold higher than after LF2000-mediated transfection [142].
A novel arginine-grafted polymer for gene delivery with low cytotoxicity was developed by Kim et al. [198]. This copolymer consists of two PAMAM moieties separated by PEG and flanked by arginine residues. It forms nanosized polyplexes with plasmids that were delivered into various cell lines. The precursor polymer lacking the arginines showed only a minor capability to promote gene expression from the delivered plasmid, presumably because it cannot escape from endosomes. In contrast to this, transfection with the arginine modified polymer led to significantly enhanced gene expression. Based upon extensive studies with inhibitors of endocytosis, the authors suggested that this delivery system enters the cell via a combination of pathways [198].
From the literature currently available, it seems that for in vivo studies CPP-based polymers or complex systems are preferable as opposed to simple CPP approaches. To give an example, the histidine/lysine-polymer H2K4b proved to be useful for plasmid delivery into several cell lines as well as in mouse models. In comparison to other versions, this 4 branched system had the highest efficiency [183]. In vivo, after systemic administration of a luciferase plasmid in mice with a xenograft tumour, gene expression was mostly detected in lung and spleen, in addition to tumour tissue. In two models with different growth rates the tumour size was significantly reduced when the transfected plasmid encoded antitumour genes [184].
As it has been already mentioned in the chapter ‘polymers and complex systems’, the ‘multifunctional envelope-type nano devices’ (MEND) mimic virus particles [185]. In order to compare the transfection efficiency of different MEND types for gene delivery, DNA coding for luciferase was used as a cargo. Out of several different combinations of peptides and lipids tested in cell culture the highest luciferase activity was achieved with MEND3, which are poly-L-lysine/DNA particles coated with endosomolytic lipids and stearyl-R8 (DOPE/CHEMS/STR-R8). Most impressively, the efficiency of this transfection method was comparable with that of adenovirus without inducing any cytotoxicity. Additionally, the authors present a successful gene delivery into hair follicles of mice after topical application at a rate of transfection much higher than achieved with LF [179].
5.2. Antisense oligonucleotides
By far the most common use of oligonucleotides as inhibitors of gene expression is the so-called antisense approach (for a review see: [199,200]). Antisense oligonucleotides are complementary to the RNA of interest, therefore specificity is mediated through Watson-Crick base pairing of the oligonucleotide with the target RNA. The three principle ways that antisense oligonucleotides have been used to disrupt protein production are: (I) the oligonucleotide/RNA duplex may form a substrate for endogenous RNase H, leading to mRNA cleavage; (II) the oligonucleotide/RNA duplex may prevent the productive assembly of the ribosomal complex or arrest a ribosomal complex already engaged in translation, in both cases affecting protein biosynthesis; (III) the oligonucleotide/RNA duplex may alter pre-mRNA splicing in the nucleus. Early studies demonstrating antisense oligonucleotide-mediated effects (i.e. inhibition of neurite growth via downregulation of the amyloid precursor protein [14] or cell death after downregulation of a Cu/Zn superoxide dismutase [201]) were performed with penetratin-DNA conjugates. In addition, conjugates of Tat peptide or penetratin with phosphorothioate modified oligonucleotides were effective in antisense inhibition of P-glycoprotein expression, a membrane ATPase associated with multidrug resistance in tumour cells [86]. However, most peptide-based delivery studies of antisense oligonucleotides were conducted with peptide nucleic acids (PNAs) [17,85,105,112,202–205]. PNAs are nucleic acid mimics in which the ribose-phosphate skeleton has been exchanged with a simpler polyamide backbone [206]. PNAs bind to both single-stranded DNAs and RNAs with high affinity and sequence specificity. Furthermore, PNAs bind to double-stranded DNAs through the unique mechanism of so-called strand invasion. In addition to their remarkable hybridisation properties, PNAs are resistant to nucleases and proteases because they lack anomeric carbon atoms and standard amino acids. Owing to their inability to activate RNase H, in contrast to unmodified antisense oligonucleotides, PNAs merely act as a steric block.
The first study showing CPP-mediated antisense activity for a PNA was conducted by Pooga et al. [112]. Here, a suppression of the galanin receptor expression in cell culture as well as in a rat model was achieved by coupling a corresponding 21mer PNA to penetratin or transportan. Additionally, it was reported that a 16mer PNA coupled to transportan and targeted to the HIV-1 transactivator responsive region (TAR) RNA was efficiently internalised into cultured cells [203]. Examination of the functional efficacy of the PNA-transportan conjugate in cell culture using a luciferase reporter gene assay revealed a significant inhibition of Tat-mediated transactivation of HIV-1 long terminal repeat. Furthermore, the conjugate substantially inhibited HIV-1 production in chronically HIV-1 infected H9 cells [203].
More recently, a PNA with an almost identical sequence, disulfide-linked to either transportan or the chimeric peptide R6-penetratin was shown to exhibit dose-dependent inhibition of Tat-mediated transactivation in a HeLa cell assay when incubated for 24 h [207]. When chloroquine was co-administered, transactivation activity was already reached within 6 h. Interestingly, fluorescein-labelled stably linked conjugates of Tat, transportan or TP10 with the same PNA were inactive when delivered alone, but attained transactivation inhibition in the presence of chloroquine. The data presented indicate that a cleavable bond is not essential for activity in this assay. Moreover, confocal microscopy showed that fluorescently labelled CPP-PNA conjugates were sequestered in endosomal or membrane-bound compartments of HeLa cells, which varied in appearance depending on the CPP. Coadministration of chloroquine was seen in some cases to release fluorescence from such compartments into the nucleus, but with different patterns depending on the CPP. These findings of Turner et al. [207] are inconsistent with observations of Tripathi et al. [205]. The latter suggested a non-endocytotic pathway for the uptake of disulfide-linked conjugates of anti-TAR PNA with several CPPs as illustrated by flow cytometry analysis. Additionally, an inhibitory effect on HIV-1 replication with IC50 values in the submicromolar range as well as viricidal activity in the low nanomolar range for the conjugates tested was reported. Very recently, they evaluated pharmacokinetic properties of the anti-TAR PNA-penetratin conjugate in Balb/C mice and concluded that the construct should be nontoxic in the concentration range predicted for a future therapeutic use [107].
Using an improved purification protocol, Turner et al. [208] synthesised several CPP-conjugated 2′-O-methyl RNA oligonucleotides (OMe) and OMe/locked nucleic acid (LNA) mixmers as well as OMe-phosphorothioate RNA oligomers targeted to HIV-1 TAR. Although all oligonucleotides had previously shown activity in the HIV-1 transactivation assay after cationic lipofection [209], no activity was detectable for the highly pure conjugates. In agreement with this, only vesicular uptake, but no nuclear import was observed by confocal microscopy. In order to generate a net positive charge of the conjugates, additional basic amino acids were introduced into the peptide sequence, which did not enhance bio-availability of the oligonucleotides either. Interestingly, the rate of uptake was dramatically enhanced by addition of free CPP to the conjugates, though still no biological activity was observed, indicating a possible lack of endosomal escape [208]. The authors suggested that these free CPPs form complexes with CPP-cargo conjugates, which play a significant role in the uptake process and concluded that care has to be taken during conjugate purification. Furthermore, this study shows that the uptake pattern strongly depends on the cell line analysed.
Astriab-Fisher et al. [87] described delivery of OMe RNA phosphorothioate oligonucleotides linked via disulfide bridge to Tat peptide and penetratin. As a biological readout a splice correction assay [210] was applied. This assay uses antisense-mediated rescue of an introduced aberrant splice site which otherwise leads to an inactive reporter enzyme, in this case luciferase. Such an approach is particularly interesting since the reporter gene activity is turned up rather than turned down upon application of the appropriate antisense oligonucleotide. Thus, negative effects caused by experimental constraints leading to reduced protein expression, which are not due to the applied nucleic acid, will not cause false interpretation of the experimental data. The CPP-oligonucleotide conjugates progressively entered cells in a matter of hours and were detected both in cytoplasmic vesicles and in the nucleus [87]. The conjugates targeted to the aberrant splice site, but not the mismatched controls, caused an increase in luciferase activity in a dose-responsive manner. These findings are in contrast to the results of Turner et al. [208] described above, who could not find a biological effect. This discrepancy may simply be due to differences between the two biological assays.
Phosphorodiamidate morpholino oligomers (PMOs) are similar to DNA with two major structural differences: the negatively charged phosphorodiester internucleoside linkage in DNA is replaced by the neutral phosphorodiamidate linkage and the five-membered ring of deoxyribose in DNA is replaced by the six-membered ring of morpholine. The uncharged and hydrophilic PMOs are highly resistant to enzymatic degradation. Using the same splice correction assay as described above, Moulton et al. [211] could show that missplicing of pre-mRNA was corrected upon addition of a R9F2-PMO conjugate into cell culture medium at low micromolar concentrations. Delivery of PMOs to the cell nucleus and cytosol required conjugation rather than complexation of peptides to PMOs. Furthermore, the arginine-rich peptide R9F2 showed higher transfection rates than conjugates with Tat peptide, penetratin or a Tat peptide analogue. The comparison of conjugates with various linkers revealed increased antisense activity of R9F2-PMO conjugates with longer spacers whereas variation in conjugation chemistry did not result in any differences [211]. Additional studies of the same group show inhibition of coronavirus, flavivirus, Dengue virus and West Nile virus replication by CPP-mediated antisense PMO delivery [212–216].
Besides the addition of chloroquine, different endosome disrupting strategies have been evaluated using the splice correction assay, for example cotreatment with endosome-disruptive peptides [115] or photochemical internalisation [129]. However, the most promising results, especially concerning future in vivo applications of steric block oligonucleotides, have been achieved with two newly developed derivatives of classical CPPs (reviewed in [217]). The modification of oligoarginines with non-natural, uncharged amino acids [218] led, amongst others, to the peptide (R-Ahx-R)4, in which Ahx represents a six-atom aminohexanoic acid spacer. Abes et al. demonstrated that in contrast to Tat or oligoargine, PMO-conjugates of this peptide led to dose-dependent splice correction at low micromolar concentrations in the absence of endosomolytic agents. The underlying mechanism for this superior activity is not clear yet, as the uptake of (R-Ahx-R)4 constructs was less efficient than the uptake of Tat or oligoarginine constructs and also involved endocytotic routes [219]. The second peptide is a derivative of penetratin, to which six arginine residues were added at the N-terminus (R6Pen). R6Pen-PNA conjugates were shown to inhibit HIV-1 Tat-dependent transactivation [207] as well as promote efficient splice correction, in both cases at low concentrations and in the absence of endosomolytic agents [109]. Again, uptake of R6Pen-conjugates seemed to involve endocytosis and there was nearly a difference in splice correcting activity regardless of the nature of the linker used for conjugation, like a stable thioether versus a reducible disulfide linker [217]. CPP-PMO or -PNA-conjugates are beginning to advance from research tools to therapeutic application. In vivo efficacy has already been demonstrated in mouse models for muscular dystrophy and coronavirus infection using another variant of the (R-Ahx-R)4-peptide described above ([220] and references therein).
In this regard, another very interesting approach is followed by Morris et al. [134] using negatively charged PNA-like DNA mimics called HypNA-pPNA, which consist of phosphonate analogues of PNA and PNA-like monomers on the basis of trans-4-hydroxyl-L-proline [221]. Based on a previous peptide from the Pep family [133], the group designed a new CPP, Pep-3, which forms stable non–covalent complexes with charged as well as uncharged PNAs [134]. Efficient cellular uptake of an antisense HypNA-pPNA was demonstrated via downregulation of cyclin B1 in HeLa cells as well as in suspension and primary cells. Furthermore, a mouse xenograft tumour model of human prostate carcinoma was used to analyse in vivo delivery of Pep-3/HypNA-pPNA complexes. Intravenous administration reduced tumour growth by approximately 20 %, but after intratumoural injection, a specific and concentration-dependent inhibition of tumour growth up to 90 % was detected, which could be further improved by stabilising the complex through PEGylation at the N-terminus.
MicroRNAs (miRNAs), a conserved class of small non-coding RNAs, participate in the post-transcriptional regulation of many cellular processes and are also involved in the emergence of tumours or metabolic diseases [222]. Specific miRNA-silencing can be achieved through the administration of antisense oligonucleotides, so called antagomirs [223]. Along these lines, Fabani et al. investigated the blocking activity of an anti-miR-122 PNA conjugated to the R6-modified penetratin described above [108]. Knockdown of miR-122 to a very low level after incubation with the inhibitory construct was verified by Northern blot analyses as well as by up-regulation of mRNAs normally negatively regulated by miR-122. Strikingly, incubation with an unconjugated PNA which was only modified with 4 lysine residues, led to a complete knockdown of miR-122. Cellular uptake of similar PNA-Lys oligonucleotides had been observed earlier [207,219], but in these cases was apparently not sufficient to obtain the desired biological activity in the nucleus.
5.3. Transcription factor decoy oligonucleotides and triplex forming oligonucleotides
Double-stranded transcription factor decoy oligonucleotides are a powerful tool to modulate gene expression. Decoys compete with response elements within the promoter regions of genes that bind transcription factors.
Coupling a NFκB specific decoy to either transportan or TP10, Fisher et al. [113] could show efficient cellular translocation of this construct going along with an inhibition of interleukin(IL)-1β-induced NFκB activation and NFκB downstream effects. The coupling was achieved by hybridising a PNA to the decoy, which contained a corresponding single strand overhang, while the PNA itself was linked to the CPP via a disulfide bridge. Recently, the authors applied this principle to a model of Alzheimer'sdisease in primary rat glial cells which are much harder to transfect. This resulted in 80 % inhibition of the NFκB binding activity leading to a decrease of the IL-6 mRNA expression by 50 % [224].
Using a similar approach, El-Andaloussi et al. [225] presented uptake experiments comparing the effect of a Myc decoy either covalently or non-covalently attached to TP10. The Myc protein had been shown to be overexpressed in 50 – 60 % of human tumours known today and, being well characterised, represents a favourable target for anticancer therapies. Besides the biological effect, the intracellular amount of decoy was determined by fluorescence measurement. Compared to non-covalent complexation of TP10 and decoy, the oligonucleotide was taken up about 50-fold less when hybridised to the TP10-PNA conjugate. However, this did not correlate with biological activity, as the conjugate was just 2-fold less effective. Intriguingly, various endocytosis inhibitors had no effect on the rate of internalisation suggesting that peptide/cargo complexes bypass the endosomal pathway to a large extent. Therefore, the low bio-availability must be caused by other unknown intracellular processes.
Triple helix-forming oligonucleotides (TFOs) bind as a third strand in the major groove of duplex DNA to form triplex DNA in a sequence-specific manner (for a review see: [226]). Attractive applications of this TFO approach include not only inhibition of the target gene, but also induction of point mutations at predetermined sites. A TFO coupled to penetratin by disulfide linkage was shown to internalise into cells [227]. In this approach the supFG1 reporter gene, encoding an amber suppressor tRNA, was used as target and the number of mutations induced by the TFO was determined with the help of an E. coli strain expressing β-galactosidase with an amber mutation. After treatment of cultured mouse cells, the rate of mutation was documented to be enhanced by a factor of 20 compared to basal levels. Confocal microscopy studies confirmed uptake into the nucleus and also revealed that the sequence of the oligonucleotide cargo affected the efficiency of delivery and the pattern of intracellular distribution (e. g. a higher GC-content seemed to reduce nuclear accumulation).
5.4. siRNA
In addition to the antisense applications described above, siRNAs represent a further valuable antisense tool to inhibit the expression of a target gene in a sequence-specific manner. These small RNA molecules induce a process termed RNA interference (RNAi) resulting in mRNA degradation (for a review see: [228,229]).
The first study about CPP-mediated delivery of siRNA was published in 2003 by Simeoni et al. [131]. Here, siRNAs were non-covalently complexed with the peptide MPG leading to a strong down regulation of the target protein. Interestingly, a mutation in the NLS sequence of the carrier peptide (MPGΔNLS) resulted in a slight increase of the RNAi effect. When siRNA was associated with MPG at a 1:10 ratio of negative to positive charges and applied to Cos-7 or HeLa cells, a decrease of about 80 % in luciferase activity was detected. This effect was further enhanced to about 90 % by MPGΔNLS [131].
Other studies describe covalent attachment of cargo and carrier. In one approach, anti-GFP or anti-CDK9 siRNA were cross-linked to Tat47–57, but significant down-regulation of the target protein could only be observed for high concentrations of nucleic acids (about 300 nM, [90]). Simple mixing of siRNA and Tat peptide did not lead to any measurable RNAi effect. LF- or Tat47–57-mediated transfection resulted in a perinuclear localisation of siRNA. In contrast, fluorescently labelled Tat47–57 without cargo was mainly found in the nucleolus. A significant increase in the rate of uptake of siRNAs targeted against luciferase or GFP could be observed after disulfide coupling the 5′-end of the sense strand to penetratin or transportan [104]. Compared to LF2000, slightly higher levels of transfection were achieved. Interestingly, after LF2000-mediated transfection, basal luciferase activity returned to normal levels one day earlier than after CPP-mediated transfection although the same concentration of siRNA was applied.
A remarkably strong RNAi effect in hard to transfect primary neuronal cells was reported by Davidson et al. [230]. Here, siRNAs directed against several endogenous proteins were coupled to penetratin via a disulfide bond. The observed down regulation of the target proteins after peptide-mediated siRNA delivery was found to be far more effective compared to LF2000. This in part was attributed to the toxicity of the lipids.
Dowdy and his group [231] presented a rather critical point of view referring to previous studies with CPP-siRNA-conjugates. They claim that the successful delivery described therein is solely the result of excess free peptide which leads to additional complexation and thereby cellular import of the siRNA. This is in accordance with Turner et al. [208], who were the first to observe that careful purification of CPP-antisense-conjugates abrogates their biological effect.
Lundberg et al. [76] rationally modified penetratin to form a CPP (termed EB1) with improved endosomolytic properties. They achieved a pH-dependent conformational change of the peptide to a higher degree of helicity by the replacement of two basic amino acids with histidines and the N-terminal addition of six amino acids. In this study, several CPPs were compared in a non-covalent approach by measuring the overall cellular uptake via fluorescence and biological effect of siRNA targeted to the luciferase mRNA. Penetratin- as well as TP10-mediated transfection did not lead to any silencing of luciferase gene expression, despite high amounts of intracellular siRNA [76] and in contrast to previous achievements with siRNA-penetratin-conjugates [230] or TP10/DNA-complexes [225]. EB1 showed improved delivery with a reduction of luciferase activity to approximately 50 % at 100 nM siRNA. The peptide also induced RNAi in HepG2 cells but to achieve this, the transfection protocol had to be changed, i.e. the preincubation volume had to be increased, emphasising differences in uptake properties of different cell lines. As it had been described earlier, that addition of a pH-sensitive peptide derived from haemagglutinin (HA2) can promote endosomal escape [43], the authors linked HA2 to penetratin [76]. It turned out that although HA2-penetratin improved the silencing effect when coincubated with penetratin, EB1 was more potent than this combination of peptides. Together with confocal microscopy studies the authors concluded that the lack of biological effect after penetratin-mediated siRNA delivery is due to a lack of endosomal escape and that EB1 has a superior endosomolytic activity in comparison to HA2-penetratin. In addition to EB1, MPGΔNLS and bPrPp were analysed in this study. For all three peptides, a much lower silencing effect was seen compared with LF2000 after transfection of 100 nM siRNA [76].
In a recent study Nakamura et al. reported condensing of siRNA with 3 types of positively charged agents: poly-L-lysine, stearyl-R8 (STR-R8) and protamine [187]. These siRNA cores were packaged into a R8-grafted lipidic envelope to form R8-MEND. It was shown that R8-MEND particles with STR-R8-condensed siRNA yielded the smallest complexes and the most efficient silencing effect in HeLa cells. At a concentration of 60 nM siRNA 80 % of gene silencing was achieved. Transfection with siRNA/STR-R8 cores lacking the lipidic envelope did not lead to any biological effect, probably due to inefficient endosomal escape [187]. In spite of this, another study provides evidence that siRNA/STR-R8 particles alone can elicit RNAi in primary rat neuron cells [232].
A promising result with a prospect for cell specific siRNA delivery was presented by Leng et al. [172]. In an attempt to optimise a branched histidine/lysine-polymer for siRNA delivery, the authors found that different 8 branched versions (H3K8b) yielded up to 80 % knockdown of the target gene in several cell types. A 4 branched H2K4b, on the other hand, turned out to be a suitable carrier for plasmids [184] but not for siRNA. Structure-function studies revealed an important role of the composition of the histidine-rich domain as well as its position within the peptide and the branches for siRNA delivery, whereas size and surface charge did not have any effect. Furthermore, the toxicity was much lower than for the commercial cationic lipids Oligofectamine and LF2000 [172]. The attachment of the tripeptide RGD, an integrin-ligand, only slightly enhanced siRNA delivery, but turned this carrier into a cell-specific system [172].
As one of the first groups to report on Tat48–60- or penetratin-mediated siRNA delivery in vivo, Moschos et al. showed, that intratracheal administration of the conjugates did not lead to any intensification of the knockdown of the target gene p38 mitogen-activated protein kinase in mouse lungs in comparison to unmodified non-formulated siRNA [106]. They give an overview of in vivo studies with siRNA alone or with several non-peptidic carrier systems showing, that even unmodified and non-formulated siRNA can exert a significant silencing effect [233]. Strikingly, it was found that the peptides alone triggered a detectable decrease in target gene expression and that the penetratin-conjugate induced elevated levels of the immune markers IFN-α, TNF-α, and IL-12p40 in lung tissue [106]. This emphasises that a conclusion cannot easily be extrapolated from in vitro experiments and applied to the in vivo situation and that the experimental conditions have to be carefully controlled.
Meanwhile, CPP-mediated siRNA delivery has been shown to be successful at least in some in vivo studies. In one of them, subcutaneous injections of siRNA non-covalently complexed with cholesteryl oligo-D-arginine (Chol-R9) in a mouse model successfully targeted the angiogenic growth factor VEGF (vascular endothelial growth factor). Seventeen days post-administration, target protein expression in the tumour decreased to approximately 40 %. More importantly, a tumour regression by a factor of 7 was measured [121].
It is long known, that a peptide derived from rabies virus glycoprotein (RVG) interacts specifically with the nicotinic acetylcholine receptor (AchR) on neuronal cells to enable viral entry. Only recently, Kumar et al. used this specificity for a delivery approach into the brain [234]. Remarkably, the biotinylated form of the 29-amino-acid peptide (YTIWMPENPRPGTPCDIFTNSRGKRASNG) was taken up by neuronal cells in the brain after injection into mice. In order to transport nucleic acids with this vehicle, R9 was conjugated to RVG peptide. Systemic treatment of mice with siRNA in a non-covalent complex with this modified peptide promoted a highly specific cellular import of siRNA only into cells expressing AchR. Even more important, an antiviral siRNA treatment resulted in successful protection of mice against encephalitis caused by Japanese encephalitis virus (JEV) [234]. This is the first study to report on a non-toxic method to deliver siRNA across the blood brain barrier which could help to circumvent dangerous and ineffective injections into the brain. To date it presents one of the most promising delivery approaches which might be expandable to other in vivo applications.
So far, no CPP-mediated transfection of miRNAs has been reported on. It was only shown that a primary microRNA (pri-miRNA) comprising 183 nucleotides has been transferred into the nucleus of HeLa cells after complexation with the so called reducible copolypeptide (rCPP) [181]. This linear delivery system is composed of the lysine-containing histidine-rich peptide (HRP) and the SV40 large T-antigen NLS randomly connected via disulfide bonds and was developed by Manickam et al. for plasmid delivery [180]. The intranuclear processing of pri-miRNA into mature miRNA was possible only if the polymers contained a sufficient portion of NLS moieties. The authors also studied the efficiency of rCPPs for cellular delivery of siRNA. It turned out that a low NLS-content in the polyplex was favourable for posttranscriptional cytoplasmic RNAi, whereas a higher NLS-content promoted nuclear delivery. Thus, siRNA-mediated promoter silencing was enabled which even excelled the efficiency of the commercial transfection reagent TransIT-TKO [181].
From these examples of CPP-mediated delivery of nucleic acids it becomes obvious, that in most cases a delivery system must be adjusted to the sort of nucleic acid to be transfected. The enormous difference in size and number of charges between oligonucleotides and plasmid DNA leads to different mechanisms of non-covalent particle formation. This partly can explain cargo-dependent discrepancies that have been found also for lipidic carrier systems [235].
6. Cellular uptake versus functional effects of the nucleic acid cargo
To efficiently exert their activity, nucleic acid-based tools and drugs have to reach their cellular targets after gaining entry to the cell (for a recent review see [237]). To claim ‘proof of principle’ for a particular CPP approach, it might be sufficient to just measure uptake of a given functional nucleic acid cargo (e.g. siRNA, aptamers or antisense RNA). However, it does not provide any information about the biological activity of the cargo. Thus, a side-by-side analysis of cargo internalised and the corresponding biological effect is required to determine the overall efficacy of such a system. We have adapted a highly sensitive method described by Overhoff et al. [238] for the quantification of siRNA, enabling us to detect intracellular siRNA amounts down to ≥ 10 copies per cell. The method is based on the liquid hybridisation of a radioactively labelled probe with the corresponding antisense strand of the siRNA in cellular lysates. Additionally, a stringent washing procedure is absolutely required. As already outlined above, complexes strongly attached to the cell surface are often an underestimated source of misinterpretation of the true amount of cargo taken up by cells. By implementing a heparin wash, we could reduce the overall amount of allegedly ‘intracellular’ siRNA by about 90 % [236]. This result impressively shows the importance of such a procedure.
As opposed to the majority of CPP applications reported, which rely on covalent linkage of carrier and cargo, we used in our studies a peptide termed MPGα, a derivative of the original MPG peptide described by Morris and coworkers [19]. This peptide forms highly stable non-covalent complexes with nucleic acids. MPGα differs from MPG by six amino acids in the hydrophobic part. These changes result in an alteration of the overall structure of the peptide towards a higher tendency of adopting a helical conformation [68]. Accordingly, the two peptides behave differently with respect to their interaction with artificial lipids as well as Xenopus oocytes [239] and should not be confused as, most probably, their exact mechanism of uptake is not the same.
The delivery of anti luciferase siRNA/MPGα complexes into HeLa or ECV304 cells yielded an IC50 of about 0.8 nM (Figure 5). Although the number alone is already quite impressive and convincingly shows that this approach works very well in cell culture experiments, it does not provide information about the percentage of bio-available siRNA taken up. Comparing the amount of extracellular RNA with the amount of RNA internalised, it becomes obvious, that less than 1 % entered the cells. However, even more remarkable are the data derived from a comparison with microinjection experiments [236]. These indicated that less than 1 % of the molecules taken up enter the RNAi machinery. Thus the vast majority of molecules probably are in a physical state, which is not suitable for triggering RNAi.
Confocal microscopy studies with fluorescently labelled oligonucleotides reveal a punctual cytoplasmatic distribution of the molecules after MPGα mediated uptake (Figure 6). Together with the data described above and data concerning the effect of specific inhibitors/effectors of different endocytotic pathways [236], this strongly indicates endocytosis to be involved in uptake. Accordingly, if it would be feasible to liberate the siRNA molecules trapped in vesicular compartments, one would be able to dramatically boost the overall efficacy of this approach. Though others have reported similar findings, our study for the first time puts in numbers how much room for improvement there actually is. Altogether, the data presented clearly show the enormous potential of this particular CPP-mediated cellular delivery of siRNA, which likely is true to various degrees for other similar approaches as well, but at the same time reveal the problems to be resolved in the future.