Efficient delivery of heterologous molecules for treatment of cells is a great challenge in modern medicine and pharmacology. Cell-penetrating peptides (CPPs) may improve efficient delivery of a wide range of macromolecular cargos, including plasmid DNA, small interfering RNA, drugs, nanoparticulate pharmaceutical carriers, and anticancer drugs. In this paper, we present the history of CPPs’ discovery with special attention drawn to sequences of viral origin. We also describe different CPP families with regard to their physicochemical properties and numerous mechanisms of CPP cell uptake by direct penetration and endocytotic pathways. A detailed description is focused on formation of carrier-cargo complexes, which are needed for practical use of CPPs in medicine and biotechnology. Examples of successful application of CPPs in treatment of human diseases are also presented, including decreased tumor growth and induction of cancer cell death. Finally, we review modern design approaches to novel CPPs and prediction of their activity. To sum up, the current review presents a thorough and up-to-date knowledge of CPPs and may be a valuable source of information for researchers in pharmacology designing new therapeutic agents.
Efficient delivery of therapeutic chemical compounds or diagnostic molecules to cells is a laborious challenge. In recent years, cell-penetrating peptides (CPPs) received great attention as efficient cellular delivery vectors due to their intrinsic ability to enter cells and mediate uptake of a wide range of macromolecular cargos, such as plasmid DNA, small interfering RNAs, drugs, nanoparticulate pharmaceutical carriers, and anticancer drugs. The “cargo” is associated with the CPP via covalent bonds or through noncovalent interactions. CPPs are alternatively called protein transduction domains. The first CPPs found were parts of naturally occurring proteins that have the ability to enter cells as transactivator of transcription (Tat) proteins of Human immunodeficiency virus 1 (HIV-1) and a pANT 16-mer peptide termed penetratin derived from the third helix of the homeodomain of Drosophila Antennapedia (Derossi et al., 1994; Lebleu, 2013; Sagan et al., 2013) (Fig. 1). Both proteins are the transcription factors that bind DNA. The following studies indicated that an arginine-rich motif in Tat (Futaki et al., 2001) and the third helix (amino acids 43–58) of the pANT homeodomain are necessary and sufficient for translocation into cells. Further studies on arginine-rich peptides showed a key role of guanidium side chains of arginine residues for cellular uptake (Futaki et al., 2013). Recently, a new class of cell-penetrating peptides represented by the short peptide Xentry (LCLRPVG) derived from an N-terminal region of the X protein of Hepatitis B virus has been described (Meloni et al., 2014) (Fig. 1). Many other animal viruses, such as Flock house virus or Vaccinia virus, also use protein transduction mediated by the CPP motif to enter animal cells (Mercer and Helenius, 2008; Nakase et al., 2009). Plant viruses' entry mechanism into cells is not well known. It was indicated that the intact virion and a recombinant capsid protein (CaP) from the plant-infecting nonenveloped icosahedral RNA virus Brome mosaic virus (BMV) could penetrate the membranes of plant protoplasts and deliver BMV RNA into cells. The N-terminal tail of CaP, containing the arginine-rich motif, is required to penetrate cellular membranes (Qi et al., 2011). It is interesting that CPPs are present in viruses that can infect many different hosts, from human to insects to plants, and the CPPs derived from viruses (HIV-1, BMV) are also involved in an interaction with viral RNA during encapsidation (Frankel and Pabo, 1988; Hema et al., 2010). Furthermore, un-natural CPPs, such as model amphipathic peptide (MAP) or transportan, which do not have any natural parent protein, were synthesized de novo.
Generally, CPPs are relatively short cationic (10–30 amino acid residues in length) and/or amphipathic water-soluble peptides, and are rich in basic amino acids, such as arginine and lysine. The common feature of these peptides is the ability to pass through cell membranes and transport attached macromolecules from extracellular space through the cell membrane into cytoplasm. Therefore, CPPs can be used as in vitro and in vivo delivery vectors. However, this approach is limited by a lack of cell specificity in CPP-mediated cargo delivery, and the mechanisms by which it occurs are poorly understood. CPPs are translocated across biologic membranes via an energy-independent cellular process. Many studies suggest endocytosis as the translocation pathway, whereas others suggest that receptor-mediated and energy-independent nonendocytotic mechanisms are used. The formation of micromicelles at the membrane or direct translocation through the lipid bilayer has also been described (Derossi et al., 1996; Thoren et al., 2005; Lindgren and Langel, 2011). CPPs also use different endocytotic routes via caveolae, micropinocytosis, a clathrin-dependent pathway, and a cholesterol-dependent clathrin-mediated pathway (Madani et al., 2011). Which of these mechanisms a CPP will use depends on such parameters as size (with cargo), temperature, cell type, and modifications of CPPs or their cargo (Lindgren and Langel, 2011).
Cell-Penetrating Peptide Families
CPPs are derived from natural or un-natural protein or chimeric sequences. Although there is no established classification of CPPs, depending on their physicochemical properties or origin, they can be divided into several classes or groups (Table 1) (Milletti, 2012; Bechara and Sagan, 2013).
Generally, CPPs belonging to this group contain a stretch of positively charged residues in their primary structure, which plays a crucial role in the cellular uptake (Milletti, 2012). The most well known cationic CPP is CPPTat, derived from the HIV-1 Tat protein, which is a transcription-activating factor involved in the replication of HIV. It contains 86–102 amino acids, depending on the viral strain and three different functional domains: 1) an acidic N-terminal region important for transactivation, 2) a cysteine-rich DNA-binding region (22–37 amino acids) with a zinc-finger motif, and 3) a basic region (49–58 amino acids) responsible for nuclear import and a 48–60 Tat sequence, which contains six to nine arginines and acts as a CPP (Vives et al., 1997; Lindgren and Langel, 2011).
Studies focused on arginine homopolymers derived from CPPTat [Rn(n = 6–12)] have shown that cellular uptake strongly depends on the number of arginine residues, and the minimal sequence required for the efficient uptake is octa-arginine (Mitchell et al., 2000). The effectivity of arginine homopolymers uptake increases along with the peptide length up to 15 amino acids. Furthermore, homopolymers composed of the same number of lysine, ornithine, or histidine residues are less efficient in entering the cell than arginine polymers (Mitchell et al., 2000). M918 is a cationic CPP which consists of 22 residues of amino acids, including seven arginines, and derives from the C terminus of the tumor suppressor protein p14ARF (Chugh et al., 2010).
The hydrophilic face of amphipathic CPPs contains mostly lysine residues (Chugh et al., 2010). Amphipathic helices with hydrophilic and hydrophobic faces are probably important to cross the plasma membrane at neutral pH. This class includes primary amphipathic and secondary amphipathic CPPs. Sequences of primary amphipathic CPPs, such as transportan 10 (TP10), contain hydrophobic and hydrophilic motives (Madani et al., 2011). These peptides can be divided into two subclasses. The first one encompasses CPPs derived from naturally occurring proteins such as pVEC (CPP derived from murine vascular endothelial cadherin) (Elmquist et al., 2001) or alternate reading frame protein (1–22) (Johansson et al., 2008), and the second one consists of chimeric peptides containing a hydrophilic domain: the nuclear localization sequence (NLS) of simian virus 40 large T antigen (Goldfarb et al., 1986). On the other hand, MPG and Pep-1 have hydrophobic domains derived from the HIV-1 fusion protein glycoprotein 41 and the tryptophan-rich cluster, respectively. Both CPPs contain a spacer domain (WSQ) separating hydrophilic and hydrophobic motifs (Deshayes et al., 2006). Some primary amphipathic CPPs, such as TP10, are toxic to cells even at low concentrations. They interact with both neutral and anionic lipid membranes (Ziegler, 2008).
Properties of secondary amphipathic CPPs (such as penetratin, pVEC, M918) result from the conformational state that allows for presenting of hydrophobic and hydrophilic residues on opposite sides of the molecule upon interaction with the cell membrane (Madani et al., 2011). These CPPs have been reported to create an α-helix or a β-sheet structure, e.g., MAP model amphipathic peptide (Oehlke et al., 1998) or VT5 (Oehlke et al., 1997). They bind to model membranes with a certain fraction of anionic lipids (Ziegler, 2008).
Classification Based on CPPs’ Origin.
This classification distinguishes three major classes: 1) peptides derived from proteins, 2) chimeric peptides that are composed of two natural sequences, and 3) synthetic peptides which are designed based on structural-activity studies. The first class of CPPs includes peptides such as Tat from HIV, penetratin, VP22 from Herpes simplex virus, or a 22-residue peptide derived from the N-terminal region of CaP from a plant-infecting BMV (Qi et al., 2011) (Fig. 1). The next group of CPPs is chimeric peptides, which are partly derived from naturally occurring peptides or protein and bear two or more motifs from other peptides. For example, transportan is a 27 amino acid–long peptide containing 12 functional amino acids from the amino terminus of the neuropeptide galanin and mastoparan (a peptide derived from the venom of the Vespula lewisii wasp) in the carboxyl terminus connected via a lysine (Fig. 1). TP10 is a 21-residue peptide derived from mastoparan. A 14-residue peptide comes from wasp venom and is linked through an extra lysine residue to a six-residue sequence from the neuropeptide galanin. Certain chimeric CPPs contain signal sequences, which are recognized by acceptor proteins on the membrane of the appropriate intracellular organelles and direct proteins into an intracellular compartment (for example, NLS derivatives) (Deshayes et al., 2006).
Another class of cell-penetrating peptides is the disulfide-rich, head-to-head cyclic peptides (cyclotides), which contain approximately 30 amino acids with six conserved cysteine residues forming a cyclic cysteine knot at the core of their structure (Craik et al., 1999). These peptides were isolated from plants: MCoTI-II comes from the seeds of plant Momordicaco cochinchinesis (trypsin inhibitor II) and kalata B1 from Oldenlandia affinis. Another plant-derived cyclic peptide is sunflower trypsin inhibitor 1 (SFTI-1). It contains 14 amino acids and can also penetrate cell membranes. Although SFTI-1 has a cyclin backbone, it is structurally different from other cyclotides (Luckett et al., 1999).
The last class of CPPs is synthetic peptides such as MAP or polyarginine peptides (Bechara and Sagan, 2013). Pep-1 is the first synthesized 21–amino acid peptide containing three domains: a hydrophobic domain of many tryptophan residues, a hydrophilic lysine-rich domain derived from NLS of Simian virus 40 large T antigen, and a spacer domain. The last one enhances the flexibility and integrity of the other two domains (Chugh et al., 2010).
Native CPP-Like Peptides.
Peptides similar to CPPs can be found in prion protein (PrPsc), dynorphin neuropeptides, and some antimicrobial peptides (AMPs). Peptides derived from the N terminus of the prion protein are composed of a hydrophobic sequence followed by a basic domain (KKRPKP, residues 25–30) and have cell-penetrating ability similar to cell-penetrating peptides (Lofgren et al., 2008). The prion protein–derived CPPs have antiprion properties and significantly reduce PrPsc levels in prion-infected cells, but have no effect on cellular prion protein levels in noninfected cells. Dynorphins are endogenous opioid peptides, which are ligands for κ-opioid receptors. They are short and highly basic peptides which exert a number of important functions in the brain. Dynorphin A (Dyn A) and big dynorphin (which consists of a Dyn A and Dyn B sequence) can translocate across plasma membranes of live neurons and non-neural cells. Big dynorphin has a higher potency to penetrate than Dyn A, which is similar to a prototypical CPP transportan 10 (Hugonin et al., 2006). AMPs are short cationic and hydrophobic peptides, which display antimicrobial activity at micromolar concentrations and below (bacterial membrane damage). The distinction between AMPs and CPPs is not clear—CPPs are evaluated against mammalian cells, whereas the target of AMPs is a bacterial cell. The antimicrobial activity of some CPPs consists in membrane permeabilization, whereas toxicity of some AMPs does not and is caused by their different biologic activity (Wadhwani et al., 2012). There are also CPPs (for example, pVEC, TP10) capable of killing bacteria by permeabilizing their membranes. Thus, AMPs and CPPs could be grouped in a more general class of “membrane active peptides.” Several AMP-derived CPPs, such as Bac7, pyrrhocoricin, crotamine, melittin, human lactoferrin, or bovine lactoferricin, are known (Kerkis et al., 2004; Smith et al., 2008; Duchardt et al., 2009; Fang et al., 2013). It is noteworthy that AMP peptides are components of the natural immune system (Last et al., 2013).
Translocation of CPPs into Cells
CPPs can traverse membranes to enter cells by different mechanisms. Despite many studies on CPPs, the mechanisms by which CPPs enter the cells have not been completely understood. Different biologic and biophysical methods need to be used, as there is no specific method that could give a complete answer for each of the cell uptake mechanisms of CPPs. Moreover, depending on the experimental conditions, CPPs can enter cells via two or more pathways (Madani et al., 2011; Cleal et al., 2013; Martin et al., 2013). For a long time, it was believed that CPPs would most likely enter cells by a passive process, which is temperature and receptor independent. However, these initial studies were performed with a fixation step prior to confocal laser scanning microscopy of cells incubated with fluorescence-labeled CPP, leading to artifactual results (Richard et al., 2003).
Endocytosis, being the most common process used by cells to absorb materials from their environment, can also be used as the translocation pathway of CPPs into cells (Choi and David, 2014). Endocytosis is a generic term for several different processes, such as phagocytosis for large particles and pinocytosis for smaller ones, as well as receptor-mediated endocytosis in which clathrin or caveolin pits are involved (Madani et al., 2011). Several receptors were uncovered to be involved in internalization of CPPs, such as chemokine receptors (Tanaka et al., 2012), syndecans (Letoha et al., 2010), neuropilins (Prud'homme and Glinka, 2012), the family of integrins (Mokhtarieh et al., 2013), homing sequences, and positively-charged scavenger receptors (Reissmann, 2014). Micropinocytosis appears to be a pathway for the MCoTI-II peptide to enter into the cell. It is mediated by positively-charged residues interacting with phosphoinositides. On the contrary, kalata B1, by targeting phosphatidylehanolaminephospholipids, interacts directly with the cell membrane. SFTI-1 does not interact with any of the phospholipids, and its mechanism of penetration appears to be distinct from McoTI-II and kalata B1 (Cascales et al., 2011; Nawae et al., 2014).
A nonendocytotic, receptor-free, energy-independent cellular process is another mechanism of the CPP translocation across biologic membranes, including formation of inverted micelles, direct translocation through the lipid bilayer, and pore formation on the membrane (Derossi et al., 1996; Thoren et al., 2005; Lindgren and Langel, 2011; Choi and David, 2014). Furthermore, the carpet-like and the membrane thinning models were described as pathways of direct CPP penetration through cell membranes (Pouny et al., 1992; Lee et al., 2005). CPPs with high content in cationic residues are first absorbed at the cell surface to the numerous anionic moieties, such as sialic or phospholipidic acid or heparan sulfate proteoglycans (Pujals et al., 2006). CPPTat was shown by NMR spectroscopy and electron microscopy to form inverted micelles in DMPC (1,2-dimyristoyl-sn-glycero-3-phosphocholine) membranes (Afonin et al., 2006). The formation of the inverted micelle decreases the potential energy of the peptide. The inverted micelle structure is consistent with the experimental results where Tat peptides and Antennapedia induce the nonlamellar phase of the lipid bilayer membrane (Kawamoto et al., 2011). Inverted micelles could be formed easier when the cationic group of CPPs rich in basic amino acids (arginine or lysine) is involved, including peptides such as penetratin, members of the polyarginine family, and Tat. The pathway that this peptide follows for entry into the cell has been the subject of strong controversy in the last decade. The Tat peptide is highly basic and hydrophilic. Therefore, a central question needs to be answered: how this highly hydrophilic peptide is able to cross the hydrophobic barrier imposed by the cell membrane. A mechanism for the spontaneous translocation of the Tat peptides across a lipid membrane was proposed. The Tat peptides interact with the phosphate groups on both sides of the lipid bilayer. The insertion of charged side chains initiates the formation of a transient pore. This step is followed by the translocation of the Tat peptides that diffuse on the pore surface (Herce and Garcia, 2007). The pore formation model occurs through disruption of the lipid bilayer initiated by interactions between the side chains of basic amino acids such as Arg9 and phospholipid phosphate groups of cell membranes (Sgolastra et al., 2013). In this model, hydrophobic residues are near the lipid chains, and hydrophilic residues help in formation of the central pore. There is also an alternative proposal for this model in which lipids bend in a way that the CPP is always close to the head group. It seems that this mechanism could also be used by the amphipathic group of CPPs including the following peptides: MAP, MPG, CADY, and Pep-1. The third hydrophobic group comprising peptides such as TP10, C105Y, and Bax-inhibiting peptide (Choi and David, 2014) was reported to use different uptake pathways, depending on the content of hydrophobic residues (Bechara and Sagan, 2013).
Which of these mechanisms a CPP will use is dependent on such parameters as concentration, size (with cargo), temperature, cell type, and modifications of CPPs or their cargo (Lindgren and Langel, 2011). “Magic arginine” and “tryptophan power” as well as secondary structures of CPPs were described as especially important in translocation through a cellular membrane (Bechara and Sagan, 2013). Different CPPs can localize to different cellular compartments. Some of them generate pore-forming complexes resulting in the destruction of cells, whereas others localize to various organelles (for example, herpes simplex virus type 1 VP22 peptide reaches the nucleus to exert its biologic activity). Endosomal escape of CPPs is an important issue, thus they can avoid degradation in lysosomes and carry the cargo to its extraendosomal target and retain its biologic activity (Bechara and Sagan, 2013). One of the biggest drawbacks of CPP usage as therapeutic agents results from the uptake into intracellular endosomes. The release from these cell structures requires their destabilization by the addition of auxiliary compounds or charged polymers such as polyethylenimines. All of these auxiliary compounds may be cytotoxic. To avoid this endosomal uptake, new CPPs with nonendosomal uptake mechanisms should be developed (Reissmann, 2014). Some preclinical and clinical developments of CPP conjugates demonstrate their potential as therapeutic agents for drug discovery. There is increasing evidence to suggest that CPPs have the potential to cross several biobarriers (e.g., blood-brain barriers, intestinal mucosa, nasal mucosa, and skin barriers) (Shi et al., 2014). Electroporation has recently been demonstrated as an effective approach for delivering cell-penetrating peptide conjugates of peptide nucleic acids (CPP-PNAs) into a complex cellular system for antisense application. The electroporation improves the bioavailability of CPP-PNA and its antisense effects (Ma et al., 2014). Figure 2 summarizes all of the previously described mechanisms of CPP uptake into cells.
Formation of Carrier-Cargo Complexes
CPPs have the ability to transport attached macromolecules from extracellular space through the cell membrane into the cytoplasm; therefore, they can be used as efficient carriers for the delivery of drugs and biologically active substances, and the improvement of intracellular trafficking is highly expected. There is a large range of chemical compounds that can be considered as cargos, including plasmids, DNA, oligonucleotides, small interfering RNA (siRNA), PNA, proteins, peptides, liposomes, low-molecular-weight drugs, and nanoparticles.
On the basis of association of CPPs with their cargo, they could be classified as covalently and noncovalently bound. Covalent interaction is a conjugation by chemical cross-linking, such as sulfosuccinimidyl suberate linkage, carbodiimide conjugation, and thiol-amine coupling, but also cloning that is followed by the expression of the CPP fusion protein (Wagstaff and Jans, 2006; Liu et al., 2010). On the other hand, the noncovalent linkage includes electrostatic and biotin-streptavidin interactions or can be metal affinity–based (Heitz et al., 2009; Liu et al., 2010).
The covalent conjugation has been the main method used in the past for cargo coupling. It results in stable and chemically well defined conjugates, involving mainly disulfide or thio-ester linkages. For this type of conjugation, CPPs such as Tat, pANT, transportan, pVEC, and arginine-rich peptides were commonly used (Muratovska and Eccles, 2004; Hu et al., 2009; Wei et al., 2009; Sawant and Torchilin, 2011; Regberg et al., 2012). The advantages of the covalent bonding for in vivo applications are reproducibility of the procedure as well as control of the stoichiometry of as-prepared conjugates. This strategy was successfully used to internalize into cells uncharged oligonucleotides (ONs) such as PNA and phosphorodiamidate morpholino oligomer (Laufer and Restle, 2008; Moulton and Moulton, 2008; Kenien et al., 2012; Margus et al., 2012) as well as small-molecule drugs (Bolhassani, 2011) and nanoparticles (Delehanty et al., 2010; Lee and Tung, 2012; Chakrabarti et al., 2014).
As far as peptide/protein covalent conjugation is concerned, it could be produced as a chimeric protein in bacteria. The delivery of CPP-fusion proteins into various tissues in mice was demonstrated for the first time by Schwarze et al. (1999). Recently, such conjugate was reported as an efficient tool for the manipulation of human mesenchymal stromal cells (Jo et al., 2012, 2014) and nervous tissue (Gitton et al., 2009).
Although these covalent linkages have been successfully used for the delivery of a wide range of cargos, they may be labor-intensive, time-consuming, and expensive. Moreover, the risk of altering the biologic activity of the cargo must be considered (Juliano et al., 2008). However, these undesirable effects can be avoided through the use of noncovalent interaction, such as complexes of CPPs with biomolecules or nanoparticles. Noncovalent interaction strongly depends on tertiary structures, hydrophobic, and electrostatic interactions of these complexes. For the formation of noncovalent complexes, it is also important to consider suitable surface charges that are necessary for the proper interaction between the CPP and the cargo (Keller et al., 2013). The ratio of peptide per cargo may be optimized empirically to obtain the best cellular delivery (Jarver et al., 2012).
As first noncovalent CPPs, short amphipathic peptides, Pep-1 and MPG, were proposed. Peptide/protein and nucleic acids were delivered into the cells, both in vitro and in vivo, respectively (Morris et al., 2008). Since that time, the new CPP-based carrier-cargo complexes have been under intensive development (Jafari et al., 2013; Montrose et al., 2013; Puig-Saus et al., 2014).
The noncovalent strategy is a suitable method for delivering negatively charged ONs such as siRNA (Crowet et al., 2013; Jafari et al., 2013), plasmid DNA (Liu et al., 2013b), antisense ONs, and splicing correction ONs (Margus et al., 2012), as well as peptides and proteins (Myrberg et al., 2007; Montrose et al., 2013). However, for the latter, the noncovalent strategy for cargo and carrier has been used less often than the covalent one. CPPs could also be used to enhance other drug delivery systems such as polymer-based systems (e.g., micelles, dendrimers), liposomes (Yang et al., 2014), and inorganic carriers such as gold-, silver-, and iron-based nanoparticles (superparamagnetic iron oxide nanoparticles) (Wagner et al., 2009; Shirazi et al., 2014; Zuo et al., 2014), and quantum dots (Delehanty et al., 2006; Liu et al., 2011, 2013a; Martin et al., 2013). Most nanoparticles provide binding sites for different cargos and targeting peptides which can be used for diagnostics and therapy. For example, one of the most promising tools for siRNA delivery is a new strategy based on self-assembling CPP-based nanoparticles. The CADY peptide is a 20-residue amphipathic peptide which spontaneously self-associates with siRNA with a strong affinity and forms stable nanoparticles. Such CADY/siRNA complexes were able to enter a range of cell lines using a mechanism independent of any endocytotic pathway (Konate et al., 2013).
As mentioned before, the main advantages of the noncovalent strategy over covalent conjugation are its simplicity, versatility with respect to cargo composition, lowered concentration needed to induce biologic response, easier functionalization for specific targeting, as well as preservation of the cargo functionality. However, the difficulty to control the final ratio and orientation of the peptide and cargo remains as an important limitation of the use of such a method in some cases.
Application of CPPs in Medicine and Biotechnology
The main interest for CPPs in medicine and biotechnology lies in their low cytotoxicity and in the fact that there is no limitation for the type of cargo (Heitz et al., 2009). To facilitate gene transfer into cultured cells and living animals, a number of peptide carriers that combine DNA binding, such as electrostatic domain (polylysine and polyarginine), and membrane-destabilizing properties have been developed (Morris et al., 2008). Moreover, the association of cationic lipids, polyethyleneimine, polyamidoamine cascade polymers, poly-l-lysine/DNA, or condensing peptide/DNA with amphipathic peptides possessing pH-dependent fusogenic and endosomolytic activities (such as fusion peptide of the HA2 subunit of influenza hemaglutinin, or histidine-rich peptides) has been shown to increase transfection efficiency (Morris et al., 2008; Trabulo et al., 2013; de Figueiredo et al., 2014).
CPPs have also been used for siRNA delivery into cultured cells by direct complexation with the siRNA anionic phosphate backbone (Mo et al., 2012). siRNAs covalently linked to transportan and penetratin have been associated with a silencing response. For example, MPG peptide as well as a Tat-DRB (5,6-dichloro-1-β-d-riboforanosyl benzimidazole) construct have been reported to improve siRNA delivery into many types of cell lines (Heitz et al., 2009; Presente and Dowdy, 2013; Farkhani et al., 2014). Silencing the vascular endothelial growth factor (VEGF) gene is an effective way to treat cancers. A complex of cholesterol-R9 conjugate as a carrier with siRNA against VEGF was able to silence the VEGF in cellular studies and when tested in a s.c. CT26 murine colon cancer xenograft model (Shin et al., 2014). CPP-PNA was also used for inhibition of bacterial cells encapsulated in macrophages (Ma et al., 2014). The CPP-PNA conjugate had efficient antisense effect for essential mRNA or ribosomal RNA sequences inhibiting gene expression and thereby cell and bacterial growth.
In different diseases, including cell proliferation/cancer, asthma, apoptosis, ischemia, stimulating cytotoxic immunity, and diabetes, CPPs successfully delivered therapeutic peptides and proteins to target cells (Heitz et al., 2009; Koren and Torchilin, 2012). Most of these applications use CPPs (Tat, penetratin, polyarginine, VP22) covalently linked to peptides or as fusion proteins (Mussbach et al., 2011; Shin et al., 2014).
A large number of preclinical studies have reported on the successful applications of complexes of CPPs attached to therapeutic cargos in cancer, muscular dystrophy, cardiology, prion diseases, and both viral and bacterial infections (Copolovici et al., 2014). Properly developed CPPs and their conjugates with therapeutics offer a very promising pathway to deliver toxic drugs at lower concentrations to critical tissues, such as tumors, heart, and others. For example, the most promising treatment of Duchenne muscular dystrophy is the use of antisense oligonucleotides, which cause exon skipping by binding to a specific mRNA sequence. A range of cell-penetrating peptides directly conjugated to these molecules have been shown to enhance their delivery, providing the possibility for future efficient treatment of this disease (Betts and Wood, 2013). Currently, morpholino oligos are in a clinical trial of Duchenne muscular dystrophy, and their uptake can be greatly increased by conjugation with CPPs (Moulton, 2013). Antiapoptotic peptides represent cargos which can also be CPP conjugated and serve as therapeutics. An example is the BH4 peptidic inhibitor of the mitochondrial apoptotic pathway, which showed cardioprotective properties in a murine model of heart infarction (Boisguerin et al., 2013).
Targeted delivery of CPPs with the ability to recognize cancer cells is particularly attractive for cancer therapy. Cell-targeting peptides are usually obtained by combining a conventional CPP sequence with a tumor-targeting peptide. Different p53-derived peptides covalently linked to CPPs (such as Tat or penetratin) have been demonstrated to restore tumor suppressor p53 functions in cancer cells. Tumor growth was reduced with the use of a peptide derived from the C terminus of p53 (delivery mediated by Tat) and PNC-28, a peptide derived from the MDM-2–binding domain of p53, which was linked to penetratin (Heitz et al., 2009; Rizzuti et al., 2015). Another example comprises two constructs derived from the D isomer of a cell-penetrating peptide (Flock house virus) with sections from a penetration accelerating sequence (FFLIPKG) and the C terminus of p53 (p53CO), which induced the cell death of glioma-initiating cells (Ueda et al., 2012).
CPPs with specific cargo were used to increase apoptosis within drug-resistant cells or increase the effect of a cytostatic agent. CPPs also found applications as transporters of contrast agents and are able to label tumor cells, making these compounds important tools in cancer diagnosis (Copolovici et al., 2014).
Small chemotherapeutic drugs could also be delivered by CPPs [doxorubicin (Dox), methotrexate, cyclosporine A, paclitaxel] (Stewart et al., 2008). The use of CPPs can also help to improve chemotherapeutic methods to prevent the evolution of drug resistance in the cells. For example, Puria and colleagues (2012) used human epidermal growth factor receptor 2–specific CPP conjugated to mammalian target of rapamycin–specific zinc finger nuclease for inhibition of the phosphoinositide 3-kinase/Akt/mammalian target of rapamycin pathway essential for the growth and proliferation of cancerous cells. CPP–zinc finger nuclease anticancer therapy has a high prospect of clinical success (Puria et al., 2012). Novel CPP drug constructs, such as CPP Dox conjugates, showed an enormous potential to increase the properties of many chemotherapeutic drugs. Conjugates of Tat and penetratin with the chemotherapeutic drug doxorubicin demonstrated higher apoptotic efficiency compared with free Dox in different cell lines, which seems very promising for cancer treatment (Aroui et al., 2010). For example, the chitosan/doxorubicin/Tat hybrid displayed a more efficient cell internalization compared with free doxorubicin or chitosan/doxorubicin complex without Tat and yielded significantly higher antitumor effects (Shin et al., 2014). A different anticancer agent such as methotrexate demonstrated a 5-fold increase in cytotoxicity with a breast cancer cell line after conjugation with the CPP YTA2 (Lindgren et al., 2006). Tat peptide has also been used for the delivery of modular antigen molecules useful for the treatment of allergy and vaccine production (Kinyanjui and Fixman, 2008; Jiang et al., 2014). Superoxide dismutase fused to Tat or to Pep-1 has been shown to protect pancreatic beta cells against oxidative stress (Heitz et al., 2009; He et al., 2014). The coadministration of insulin with CPP penetratin increased intestinal and nasal insulin bioavailability to 35 and 50%, respectively (Kamei et al., 2013).
CPPs can also possess specific bioactivity analogous to antimicrobial peptides. One of the first bioactive CPPs, which showed proapoptotic activity, was a 22 amino acid–long peptide derived from the N-terminal part of the tumor suppressor protein p14ARF. It induces apoptosis and translocates into cells by endocytosis (Johansson et al., 2008).
Activatable CPPs (ACPPs) comprise in vivo targeting agents which are composed of a polycationic CPP linked by a specific proteolytically cleavable linker (succinoyl or 6-aminohexanoyl) to a neutralizing polyanionic part in a hairpin structure–based conformation. ACPPs, compared with other CPPs, have an enhanced ability to reach targeted tissues in cancer and cancer metastases due to their cleavage by disease-associated proteases (matrix metalloproteinase-2/9) (Olson et al., 2009). Furthermore, ACPPs are predicted to become highly selective vectors to deliver macromolecular cargos (Copolovici et al., 2014; Reissmann, 2014).
CPPs and Host Immune Response
Compared with recombinant viruses, as carriers of therapeutic molecules, cell-penetrating peptides can be considered much safer in terms of their immunogenicity. However, CPPs as transport peptides may still be capable of inducing humoral immune response of the host organism against the therapy, which can bring undesired secondary effects in the end (Bolhassani, 2011). The two major factors involved in the process are multiple positive charges and aromatic side chains in several CPPs such as penatratin and transportan, as suggested by analysis of protein immune properties (Hopp and Woods, 1981). On the other hand, short peptides do not readily elicit an immune response when injected on their own (Rubsamen et al., 2014). TP10 and its chemically modified derivatives, PepFects, were tested with regard to their immunogenicity. It was found that both in vivo and in vitro, all peptides were shown to be nontoxic and nonimmunogenic at concentrations of 5 and 10 µM and potentially without the risk of inflammatory reactions (Suhorutsenko et al., 2011). Moreover, in another study, it was determined that three of the most commonly used CPPs—HIVTat, anntennapedia, and transportan—linked to a model bovine serum albumin protein cargo failed to induce a significant increase in the release of the inflammatory cytokines or activate nuclear factor κB in epithelial cells in vitro (Carter et al., 2013).
The innate immune response becomes stronger while a much larger cargo is attached to a CPP (de Figueiredo et al., 2014). Even then, it is possible to modulate immune response to the cargo depending on the CPP used to deliver the therapeutic peptide (Bitler and Schroeder, 2010). This property has already been used to improve immunogenicity of DNA vaccines by VP22 against influenza virus, human papillomaviruses, and Bovine herpes virus 1 (Brooks et al., 2010). A newly discovered CPP derived from human cationic proteins (CPPecp) showed an immunomodulatory effect through upregulation of interferon-α production in THP-1 and CD14+ cells, when cocultured with mite allergen Der-p2 (Yu et al., 2015). In an older study, it was shown that conjugation of arginine oligomers (R7) to cyclosporine A leads to the inhibition of inflammation in cells in mouse and human skin. In this case, R7 acts as a carrier of cyclosporine A and does not elicit local immune response itself, thus making cyclosporine A a more effective drug in inflammatory skin disorders. Otherwise, R7 would be useless for this application as immunogenicity of highly charged arginines would compete with anti-inflammatory properties of cyclosporine A (Rothbard et al., 2000). A similar study described the successful use of STAT-6-IP, a cell-penetrating STAT-6 inhibitory peptide, for treatment of allergic airways in a model of chronic asthma (Wang et al., 2011). Yet another very interesting example of using a VP22 is induction of tumor-specific immune response by gene transfer of Hsp70-CPP fusion protein to tumors in mice without identifying or isolating tumor-associated antigens (Nishikawa et al., 2010).
The aforementioned studies show that different CPPs can become novel anti-inflammatory agents for allergic inflammation treatment in the future. However, it is beyond any doubt that the immune response to and long-term side effects of CPP usage need further study and clarification. If a given CPP shows stronger immunogenic properties, it could be easily replaced from a wide choice of CPPs available today (Snyder and Dowdy, 2004).
Design and Prediction of Novel CPPs
Discovery of the first “natural” CPPs soon entailed attempts to design synthetic or chimeric molecules with altered, desirable features, but prediction of the new peptide activity turned out to be quite a challenge. As the trial-and-error (“educated guess”) approach proved laborious, expensive, and unreliable, more formalized approaches emerged (Copolovici et al., 2014). Researchers proposed algorithms based on z-descriptors—parameters derived from multiple physicochemical characteristics of amino acids using principal component analysis (Hällbrink et al., 2005; Hansen et al., 2008). The major drawback of this strategy is that it only considers a sum of descriptors calculated for the whole peptide, ignoring its sequence. Hence, it displayed a limited efficiency (68%) in differentiation between penetrating and not-penetrating peptides. Hansen and colleagues (2008) suggested that z-score–based substitution matrices (similar to point accepted mutation or BLOcks SUbsitution Matrix) could be used to adapt alignment and clustering algorithms (commonly used in phylogenetics) for the purpose of CPP prediction. Unfortunately, to the best of our knowledge, this hypothesis has never been tested.
Later approaches used to distinguish between translocated and nontranslocated peptides often involved machine learning techniques. Both Sanders et al. (2011) and Gautam et al. (2013) constructed their prediction models using support vector machines. The first model considered basic biochemical characteristics of the peptides, and the second also included binary profiles of amino acid patterns. Combination of the latter approach with Motif Alignment & Search Tool–based CPP-like motif recognition allowed construction of the hybrid algorithm that achieved 97.40% accuracy during 5-fold cross-validation and 81.31% accuracy when tested on an independent data set (Gautam et al., 2013).
Also, artificial neural networks were used to predict CPPs. Dobchev and colleagues (2010) combined them with principal component analysis to predict CPPs based on quantitative structure-activity relationship. N-to-1 neural network was also used in a different approach in which a machine learned to compress the peptide sequence into a fixed number of informative features (Holton et al., 2013).
Despite differences currently implemented, methods tend to emphasize net positive charge, number/percentage of arginine residues (especially near ends of the sequence), hydrophobicity of the peptide, and its amphipathic character as valid descriptors (Copolovici et al., 2014). What seems to be their common problem is the construction of training sets. More often than not, these methods work better if trained on balanced data sets. Although CPP-related data can be readily found in both literature and databases, information on their nonfunctional analogs is still scarce. What is worse, penetration assays are far from being standardized, and what appears as incapable of penetration in one cell line or experiment may turn out to be a CPP in the other (Sanders et al., 2011).
While only a few of the mentioned researchers made their tools publicly available, CPP-related resources available online currently include the following:
CellPPD: a prediction server based on Gautam’s support vector machine approach. Implemented tools can assess penetration capability of mutants (with a single mutation) of one given peptide, analyze a batch of fasta-formatted sequences, and scan a protein for putative CPPs or CPP-related motifs. Of note, the authors also granted full access to the data sets used in the training procedure. Both prediction services and raw data are available at http://www.imtech.res.in/raghava/cellppd (Gautam et al., 2013).
CPPpred: a server for automatic CPP prediction based on Holton's N-to-1 artificial neural network method (available at http://bioware.ucd.ie/∼testing/biowareweb/Server_pages/cpppred.php) (Holton et al., 2013).
CPPsite: a comprehensively annotated, curated database of CPPs with a built-in analysis suite. The base stores information on sequence, structure, basic physicochemical properties, cell lines used for testing, and literature or patent cross-references. Implemented tools allow searches for CPP sequences in protein sequences and calculation of basic physical properties from the sequence. The whole suite is available at http://crdd.osdd.net/raghava/cppsite (Gautam et al., 2012).
This review focuses on different aspects of cell-penetrating peptides, including characterization of their physicochemical properties, mechanisms of cellular membrane translocation, groups/categories of CPPs, and their potential application in medicine and biotechnology. CPPs conjugated with cargos can enter cells by different mechanisms, which are still not fully recognized. CPPs are able to interact with active cargos through covalent or noncovalent binding. These unique properties of cell-penetrating peptides can be promising, valuable, and cost-effective tools for biomedicine and pharmacology. In novel approaches, CPPs are used as vectors for controlled drug delivery in targeted cellular therapy or as transporters of imaging agents in early cancer diagnosis. Moreover, these peptides may be used in gene therapy as nonviral vectors. Discovery of new peptides, further clarification of mechanism of CPP action (alone and with cargos), and optimization of CPP delivery methods can enlarge the field of their applications.
Recently, scientists have come up with a concept to link CPPs to different kinds of nanoparticles. Nanoparticles modified with CPPs can significantly improve their uptake by cells. There has been a great focus on CPPs’ functionalization with quantum dots, liposomes, and dendrimers, which could improve drug release within cells and reduce toxicity. Moreover, nanomaterials can be modified on their surface with different ligands, and these complexes conjugated with CPPs could be easily transferred through the cell membrane. The combination of benefits of nanomaterials and unique properties of cell-penetrating peptides will be very helpful to control drug delivery systems. Therefore, application of CPPs for nanoparticle transfer loaded with drug molecules can lead to the development of new strategies for more effective tumor-targeted therapy and theranostics.
The authors thank Andrzej Pietrzak, a Polish-English translator at Adam Mickiewicz University Press in Poznań, Poland, for valuable linguistic help in correction of the manuscript.
Wrote or contributed to the writing of the manuscript: Durzyńska, Przysiecka, Nawrot, Barylski, Nowicki, Warowicka, Musidlak, Goździcka-Józefiak.
- Received February 3, 2015.
- Accepted April 27, 2015.
This work was supported by the National Centre for Research and Development in Poland under research grant “Nanomaterials and their application to biomedicine,” contract number PBS1/A9/13/2012 (to Ł.P. and A.W.).
- activatable CPP
- antimicrobial peptide
- Brome mosaic virus
- capsid protein
- cell-penetrating peptide
- cell-penetrating peptide conjugates of peptide nucleic acids
- CPP derived from HIV-1 Tat protein
- Dyn A
- dynorphin A
- human immunodeficiency virus 1
- model amphipathic peptide
- nuclear localization sequence
- CPP derived from murine vascular endothelial cadherin
- sunflower trypsin inhibitor 1
- small interfering RNA
- transactivator of transcription
- transportan 10
- vascular endothelial growth factor
- Copyright © 2015 by The American Society for Pharmacology and Experimental Therapeutics