The use of cell-penetrating peptides (CPPs) as biomolecular delivery vehicles holds great promise for therapeutic and other applications, but development has been stymied by poor delivery and lack of endosomal escape. We have developed a CPP-adaptor system capable of efficient intracellular delivery and endosomal escape of user-defined protein cargos. The cell-penetrating sequence of HIV transactivator of transcription was fused to calmodulin, which binds with subnanomolar affinity to proteins containing a calmodulin binding site. Our strategy has tremendous advantage over prior CPP technologies because it utilizes high-affinity non-covalent, but reversible coupling between CPP and cargo. Three different cargo proteins fused to a calmodulin binding sequence were delivered to the cytoplasm of eukaryotic cells and released, demonstrating the feasibility of numerous applications in living cells including alteration of signaling pathways and gene expression.
In recent years a number of peptides that are rapidly internalized by mammalian cells have been discovered or designed (Fonseca et al., 2009; Sebbage, 2009; Johnson et al., 2011). Cell-penetrating peptides (CPPs) are capable of mediating penetration of the plasma membrane, allowing delivery of macromolecular cargos to which they are attached to cell interiors. CPPs are typically 10–30 amino acids long and fall into one of three major categories: arginine-rich, amphipathic and lysine-rich, and hydrophobic (Gautam et al., 2012). CPP delivery of cargos to the interior compartments of cells is potentially transformative as a research tool, diagnostic aid and therapeutic mechanism.
More than 25 CPP clinical trials are underway, including a Phase III trial (Glogau et al., 2012; Lonn and Dowdy, 2015). However, CPPs have largely disappointed (Palm-Apergi et al., 2012) for a variety of reasons including non-penetration (Lundberg et al., 2003), limited endosomal escape (Erazo-Oliveras et al., 2012) and requirements for hydrophobic cargos (Hirose et al., 2012). Described CPP technologies are reliant on covalent crosslinking or non-specific hydrophobic interactions (Koren and Torchilin, 2012).
The work presented herein describes a novel technology that solves or ameliorates all of these problems. We have designed a CPP-adaptor fusion protein, TAT-calmodulin (TAT-CaM), which non-covalently binds, delivers and releases cargo into the cytoplasm. Three different cargo proteins were chosen to reflect a range of characteristics including size, oligomerization and structure. The strategy is generally applicable to any soluble protein and might also be used to deliver non-protein cargos such as RNA. Assays can be performed in real time with live cells without significant cytotoxicity, offering an alternative to transfection. Our advance greatly expands the applications and effectiveness of CPPs.
RESULTS AND DISCUSSION
Schematic diagrams of the CPP-adaptor and cargo proteins are shown in Fig. 1A. Our prototype CPP-adaptor, TAT-CaM, consists of the cell-penetrating sequence from the HIV transactivator of transcription (Green and Loewenstein, 1988) fused to calmodulin (CaM). Calmodulin was selected as the prototype adaptor not only because it binds its partners with high affinity in the presence of calcium but also because mammalian cells typically maintain low resting cytoplasmic Ca2+ levels, allowing rapid release of cargo after internalization as the affinity of CaM for its ligands is negligible in the absence of Ca2+. Cargos are fused to an N-terminal CaM-binding sequence (CBS) encoded by the vector.
The affinity of TAT-CaM for a natural ligand was first examined. TAT-CaM bound to neuronal nitric oxide synthase (nNOS, also known as NOS1) via the native CaM-binding site with affinity similar to wild-type CaM as assayed with biolayer interferometry (BLI), an optical biosensing technique similar to surface plasmon resonance (Fig. 1B) (Abdiche et al., 2008; Sultana and Lee, 2015; McMurry et al., 2011). Model cargos were then examined for CaM-binding affinity and kinetics. Cargos myoglobin (CBS-Myo), horseradish peroxidase (CBS-HRP) and β-galactosidase (CBS-β-Gal) all bound CaM with low nanomolar affinity and expected fast-on, slow-off kinetics (Fig. 1C–E). TAT-CaM and cargo proteins dissociated rapidly upon exposure to EDTA (koff≈0.1 s−1; Fig. 1F), indicating that the TAT-CaM–CBS interactions function essentially indistinguishably from those of wild-type CaM. All analytes exhibited negligible binding to sensors without TAT-CaM. Rate and affinity constants determined from single-state global fits are listed in Table 1.
TAT-CaM and fluorescently labeled cargo protein (1 µM of each) in buffer containing 1 mM CaCl2 was added to subconfluent BHK21 cells and incubated for 1 h, after which cells were washed and imaged by fluorescence confocal microscopy. Uptake of cargo into cell interiors was assayed using an inverted Zeiss (Jena, Germany) LSM700 confocal microscope equipped with a 40× EC Plan-Neofluar objective (NA=1.3). Z-stacks of both TAT-CaM-treated and untreated cells were acquired and analyzed for incorporation of fluorescently labeled cargo into the cytoplasm. Orthogonal projections of Z-stacks were generated using Zeiss ZEN software, which allowed for viewing both treated and untreated cells alike at the same depth within the cell relative to the diameter of the nucleus. As shown in Fig. 2, all cargo proteins were delivered to the interiors of the cells and showed significant cytoplasmic distribution, indicating efficient penetration and escape from endosomes. The fluorescently labeled cargo proteins without TAT-CaM showed a very small degree of adherence to the surfaces of cells, but no penetration into the cell, as observed by the absence of fluorescence at the same cytoplasmic depth as fluorescence was observed in cells treated with TAT-CaM (Fig. 2). Cargos reflect an array of characteristics, e.g. myoglobin is small, all-alpha helical and monomeric while β-galactosidase is large, structurally complex and forms a 464 kDa tetramer (Jacobson et al., 1994). TAT-CaM delivers all cargos so far examined, including nNOS (data not shown).
To demonstrate the broad utility of cargo uptake into different types of cells, subconfluent HEK and HT-3 cells were treated with CBS-myoglobin cargo under identical conditions, and assayed using the same conditions as above (Fig. 3). HT-3 is a human retinoblastoma line of cervical-type cells and HEK cells are human, embryonic and have an epithelial morphology whereas BHK cells have fibroblast morphology. As observed with BHK cells, CBS-myoglobin was also delivered to the interior of these different mammalian cell lines, indicating that cargo delivery can occur regardless of the cell type used.
Other CPP-cargo strategies rely on covalent linkage or non-specific hydrophobic linkers that are susceptible to getting trapped by membrane association as the CPP itself might not be released into the cytoplasm. Even non-hydrophobic CPPs are likely to be tightly bound to membrane proteins involved in endocytosis or transmembrane translocation, which would be expected to greatly hinder endosomal escape by covalently attached cargo proteins. One group estimates that less than 1% of TAT-fused cargos escape endosomes (Lonn and Dowdy, 2015). Our strategy of high-affinity but reversible non-covalent attachment of cargos overcomes trapping effects via Ca2+-dependent dissociation, allowing rapid and efficient cargo distribution to the cytoplasm even though TAT might remain trapped in the endosome.
Cytotoxicity is also a major drawback with current CPPs, particularly given the concentrations necessary to attain observable endosomal escape. Cytotoxic effects of TAT become significant in the tens of µM (Cardozo et al., 2007). That TAT-CaM effects significant cytoplasmic distribution of cargos at 1 µM without an increase in cell death as measured by Trypan Blue exclusion (data not shown) is particularly exciting.
The time frame of cargo delivery using CPP-adaptors is under an hour, and likely much faster. This is in stark contrast to the time needed to transfect cells; CPP-adaptor delivery is roughly two orders of magnitude faster than transfection, making possible time course experiments that could not previously be attempted.
The array of applications made possible by our method is vast. For example, CPP-adaptors can be developed that allow for subcellular addressing, e.g. delivery of a transcription factor to the nucleus. Delivery of antibodies, enzymes, nucleic acids and small molecules are all potentially transformative methods. Follow-up studies to address kinetics, dosing, toxicity and other parameters as well as delivery of a wide array of cargos are underway.
MATERIALS AND METHODS
Expression and purification
Our prototype CPP-adaptor, TAT-CaM (New Echota Biotechnology, Kennesaw, GA, USA), is encoded by a pET19b-based vector containing a cleavable His-tag and the cell-penetrating sequence from the HIV transactivator of transcription (Green and Loewenstein, 1988) fused to calmodulin via a GGR linker. TAT-CaM was expressed and purified from Escherichiacoli BL21(DE3)pLysS using metal affinity chromatography essentially as described (McMurry et al., 2015). Cargo proteins were expressed in BL21(DE3)pLysS from synthetic, E.-coli-optimized genes cloned into pCal-N-FLAG-based plasmids (Agilent Technologies, CA, USA), which contain vector-encoded N-terminal calmodulin binding sequences. Cargos were purified to near-homogeneity using a calmodulin Sepharose column (GE Life Sciences, Pittsburgh, PA, USA) and dialyzed into binding buffer (10 mM HEPES, pH 7.4, 150 mM NaCl, 10% glycerol, 1 mM CaCl2, 1 mM DTT). Cargos were myoglobin (CBS-Myo, GenBank accession no. AH002877), horseradish peroxidase (CBS-HRP, GenBank accession no. E01651) and β-galactosidase (CBS-β-Gal, GenBank accession no. M22590). nNOS was expressed and purified as described (Gerber et al., 1995; Roman et al., 1995) and exchanged into binding buffer as above. For cell-penetration assays, cargo proteins were labeled with DyLight 550 (ThermoFisher, USA) according to the manufacturer's protocol. Unreacted label was removed via a dye binding column.
BLI experiments were performed using a FortéBio (Menlo Park, CA, USA) Octet QK using streptavidin sensors. Assays were done in 96-well plates at 25°C; 200-µl volumes were used in each well. Ligands were loaded onto sensors for 300–900 s followed by baseline measurements in binding buffer for 300 s. Association was measured by dipping sensors into solutions of analyte protein and was followed by moving sensors to buffer only to monitor dissociation. Binding was fitted to a global 1:1 association-then-dissociation model using GraphPad Prism 5.02.
TAT-CaM and DyLight 550-labeled cargo protein (1 µM each) in binding buffer were added to subconfluent BHK21 [authenticated line purchased from ATCC (no. CCL-10) in 2010; frozen aliquots were expanded for assays], human embryonic kidney (HEK, gift from David Fulton, Medical College of Georgia) or HT-3 (purchased in 2015 from ATCC, no. HTB-32) cells and incubated for 1 h, after which cells were washed three times in phosphate-buffered saline with 1 mM CaCl2. Following treatment with 2 µM CellTracker Green CMFDA Dye (ThermoFisher) to label the cytoplasmic compartment, cells were labeled with NucBlue (ThermoFisher) and transferred to media containing 25 mM HEPES, pH 7.4, for imaging. Cells were immediately imaged on an inverted Zeiss LSM700 confocal microscope equipped with a 40× EC Plan-Neofluar objective (NA=1.3). Pinholes for each fluorophore were set at 1.0 Airy Units (29 μm). SP 490 and LP 615 filters were used to acquire the NucBlue (blue channel) and DyLight 550 (red channel) signals, respectively. CellTracker Green signal (green channel) was acquired using a BP 490-555 filter. Filters were carefully chosen to minimize spectral overlap between DyLight and CellTracker Green signals at the expense of some decline in signal quality.
Z-stacks for both TaT-CaM-treated and untreated cells were set by using the NucBlue staining of the nucleus as a reference (typical Z-stacks ranged from 6.0 to 10.0 μm). Treated and untreated cells were imaged at identical gain settings, set at sub-saturation levels on cells treated with TaT-CaM in the red emission (DyLight 550 fluorescence). Treated and untreated cells were imaged at an identical laser output level (2.0%, 555 nm laser), identical pixel dwell time of 3.15 μs, and 2× line averaging.
For analysis, images were rendered using the Orthogonal View in Zen Blue (Zeiss) software. Using the diameter of the nucleus as a landmark, the Z-plane chosen for analysis corresponded to approximately the mid-point depth of the nucleus. The DyLight 550 signal was analyzed separately (Figs 2 and 3, left panels, white signal), and merged with NucBlue [Fig. 2, right panels, red (DyLight 550) and white (NucBlue)]. Finally, the CellTracker Green channel was included to aid in visualization of the cytoplasmic compartment in treated and untreated cells.
J.C.S.’s coauthors wish to acknowledge the contributions he made to their lives and careers. J.C.S. was not only a colleague but a friend and mentor to us all. His passing is an irreplaceable loss but his presence in our lives is an invaluable blessing. We dedicate this paper to his memory.
J.C.S. conceived of TAT-CaM and contributed to the design and analysis of experiments. V.M.N. conducted many of the experiments, designed the cargos and helped write the manuscript. S.J.N. performed the confocal microscopy. C.A.C. performed tissue culture work. S.J.N. and C.A.C. consulted on experimental design and assisted in writing the manuscript. A.N.H. designed and conducted the penetration experiments in Fig. 3. J.L.M. supervised V.M.N., designed and helped conduct biosensing experiments and was the principal author of the manuscript.
This work was supported by National Institutes of Health grants GM102826 (to S.J.N.), GM111565 (to J.L.M.) and GM113522; National Science Foundation grant DBI-1229237 (to S.J.N. and J.C.S.); and by the Kennesaw State University Research & Services Foundation. Deposited in PMC for release after 12 months.
J.L.M. and J.C.S. have equity interest in New Echota Biotechnology, which has exclusive license to a pending patent on the described CPP-adaptor technology and a provisional patent on downstream applications. The other authors have no competing interests.