Research ArticleDRUG DELIVERY

Cytosolic delivery of proteins by cholesterol tagging

See allHide authors and affiliations

Science Advances  19 Jun 2020:
Vol. 6, no. 25, eabb0310
DOI: 10.1126/sciadv.abb0310


Protein-based imaging agents and therapeutics are superior in structural and functional diversity compared to small molecules and are much easier to design or screen. Antibodies or antibody fragments can be easily raised against virtually any target. Despite these fundamental advantages, the power and impact of protein-based agents are substantially undermined, only acting on a limited number of extracellular targets because macrobiomolecules cannot spontaneously cross the cell membrane. Conventional protein delivery techniques fail to address this fundamental problem in that protein cargos are predominantly delivered inside cells via endocytosis, a remarkably effective cell defense mechanism developed by Mother Nature to prevent intact biomolecules from entering the cytoplasm. Here, we report a unique concept, noncovalent cholesterol tagging, enabling virtually any compact proteins to permeate through the cell membrane, completely bypassing endocytosis. This simple plug-and-play platform greatly expands the biological target space and has the potential to transform basic biology studies and drug discovery.


Living cells are the basic building blocks of the human body and are spatially and temporally complex organisms that perform a diverse spectrum of functions, including growth, mass transport, energy production, metabolism, and reproduction. These functions are generally encoded by the DNAs and RNAs and realized by proteins. Dynamic monitoring of the protein activities allows an in-depth understanding of the nature of cell physiology and pathology, whereas modulating their activities provides a direct means to precisely control many biological processes and the associated human diseases. In principle, these ambitions in biology and medicine can be readily achieved with immunological agents such as antibodies and antibody fragments because of their broad availability (and in case not available, reasonably easy to raise).

For example, in fluorescence microscopy, tagging biomarkers of interest with fluorescently labeled antibodies allows imaging biomarker localization, colocalization, translocation, and expression levels with high sensitivity and high spatial and temporal resolution. Similarly, in drug development, new therapeutics based on immunological agents are becoming increasingly attractive. With a higher degree of complexity, protein therapeutics offer tunable binding affinity, improved binding specificity, and lower side effects compared to many small-molecule drugs, promising a paradigm shift in both drug discovery and disease treatment (14). It is worth mentioning that biological techniques for raising antibodies, nanobodies, or screening functional peptides (e.g., hybridoma and phage display) have matured over the years. Since the early 1980s, over 230 therapeutic proteins (including peptides) and their 380 drug variants have received regulatory approval for human uses (5). Despite this rapidly increasing number, we are still missing out on the majority of cell signaling nodes and drug targets inside the cells.

The fundamental limitation of the current immunological diagnosis and interventions is their incapability of interrogating or modulating intracellular targets in live cells. Considering the three-dimensional structure of cells, there are far more intracellular targets with biological significance than their cell membrane counterparts. For immunofluorescence microscopy, cells have to be fixed first, thus only providing a snapshot of the dynamic and evolving cell signaling process. For therapeutics, hydrophilic macromolecular protein drugs (e.g., antibodies, Fc fusions, hormones, cytokines, and enzymes) that cannot cross the cell membrane spontaneously only act on the limited number of extracellular targets.

The significance and urgent need for intracellular protein delivery are widely recognized in biology and medicine (68). A variety of technologies have been developed in the past several decades that partially address this chronic problem. For example, microinjection can inject any biologics directly into the cytosol but is of very low throughput (one cell at a time) (9). High-throughput approaches to punch transient pores in cell membrane have also been developed based on electroporation, toxin, hypertonic solution, and mechanical forces (1018). Compromising the cell membrane integrity not only allows material exchange between the cytoplasm and the outside media but also is toxic to cells and affects cell normal physiology. Another popular category of protein delivery technologies uses chemical agents such as liposomes, polymers, and cell-penetrating peptides (1925). These compounds do not punch holes in the cell membrane, but a common issue shared by these technologies is that protein cargo molecules are predominantly delivered inside cells via endocytosis, a remarkably effective cell defense mechanism to prevent intact foreign biomolecules from entering the cytoplasm. As a result, the vast majority of ingested biomacromolecules are degraded inside endosomes and lysosomes, limiting the bioavailability of the imaging or therapeutic agent. To address this challenge that developed by Mother Nature over millions of years of evolution, here, we report a unique, simple solution based on noncovalent cholesterol tagging. It enables proteins to directly permeate through the cell membrane without generating transient pores.


A very simple small-molecule tag was designed by conjugating Coomassie blue (CB) with cholesterol, as shown in Fig. 1A. CB is a protein staining dye commonly used in biochemistry laboratories. It is also a dye for human use under the trade name Brilliant Peel for retinal surgery and a drug candidate to treat spinal injuries (26). CB binds noncovalently with protein surface via the combination of hydrophobic interactions and heteropolar bonding (27). In parallel, cholesterol is a natural component (~30% by weight) of all animal cell membranes, essential to maintaining membrane structural integrity and fluidity (28). Two copies of cholesterol and one CB molecule are joined through a short bifunctional linker (the aminated linker increases tag solubility and helps mask the strong negative charges from the sulfonate groups in CB). Upon mixing of the tag with proteins, the CB end of the tag anchors onto protein molecules independent of the protein sequence, endowing proteins with a cholesterol-decorated surface compatible with the cell membrane. We speculate that similar to transmembrane proteins, the protein-tag complex can insert between the lipid bilayers (Fig. 1B), a behavior we have also observed on cholesterol-covered small interfering RNA (siRNA) macromolecules (29). Because of the noncovalent nature of protein-tag interaction, we believe that the cholesterol-based tags should be eventually pulled away by the cell lipid bilayer. The dissociation will leave behind a hydrophilic protein embedded between the hydrophobic lipid bilayers. Following the like-dissolves-like chemistry rule, the protein molecules will be “spit” out of the membrane, slipping directly into the cytosol.

Fig. 1 Cytosolic delivery of protein by cholesterol tagging.

(A) The protein delivery tag is designed by joining two copies of cholesterol with a CB G250 molecule via an aminated cross-linker. (B) Proposed cell entry mechanism. Upon mixing with proteins, the CB end anchors onto the protein surface, whereas the exposed cholesterol makes the cargo protein compatible with the cell membrane lipid bilayers. The protein-tag complex embedded between the bilayers should eventually dissociate because of the noncovalent nature of the protein-tag binding, leaving a hydrophilic macromolecule in the hydrophobic bilayer. This incompatibility would render the protein molecule to slip out of the membrane, either into the extracellular space or directly into the cytosol avoiding endocytic sequestration. (C) Confocal fluorescence and bright-field micrographs of live HeLa cells treated with tagged proteins of different molecular weights. All proteins were labeled with a green fluorescent dye (AF488) and tagged at appropriate molar ratios (5:1, 8:1, 12:1, and 16:1). The increasing tag ratio for large proteins was used to compensate for the increase in protein surface areas. Scale bar, 5 μm. Zoomed-out images with a larger field of view are available in fig. S1.

To establish the validity of this technology, we tested proteins of multiple sizes (molecular weight ranging between 6.5 and 150 kDa) and found that this unique cytosolic delivery mechanism is selective in protein size, likely determined by the degree of lipid bilayer deformation (bulging). For easy visualization, all proteins were fluorescently labeled. Confocal fluorescence microscopy revealed distinct intracellular fluorescence patterns of the proteins tested (Fig. 1C and fig. S1). Compact proteins were strongly favored for transmembrane delivery. As the protein size increased, the percentage of protein entering cells via direct permeation reduced, while that via endocytosis increased. For example, aprotinin with a molecular weight of 6.5 kDa showed bright and homogeneous fluorescence intracellular distribution, indicating the absence of endocytosis; lysozyme of 15 kDa generally showed a diffused distribution while having a few punctate bright spots (signature of endosomal sequestration). This result was further validated by costaining endosomes and lysosomes using a LysoTracker emitting a different color. Inside cells, the delivered proteins did not show colocalization with the endosomes and lysosomes (fig. S2). In comparison, large proteins such as bovine serum albumin (BSA; 66 kDa) and immunoglobulin G (150 kDa) exhibited punctate intracellular fluorescence only, showing that endocytosis was the dominating cell entry mechanism. As a control experiment, the small protein, aprotinin, without the cholesterol tag was also used to treat cells under the same condition. As expected, only a few punctate fluorescent spots were observed in cells (fig. S3), showing poor cellular uptake and endosomal escape.

This size effect suggests that the noncovalent cholesterol tagging technology is most suitable for small imaging and therapeutic proteins such as antibody fragments, single-domain antibodies, nanobodies, peptides, and some enzymes. Large proteins that attached to the cell membrane (unlikely to be fully embedded between the lipid bilayers) eventually enter cells through endocytosis. Although the actual protein size cutoff that allows cell membrane encapsulation depends on protein structures because proteins fold into different shapes, thus interacting with the cell membrane differently, the protein size tests suggest that the size cutoff for transmembrane permeation is likely below 60 kDa. Previous studies using rigid inorganic nanoparticles also offer some similar insights. It has been shown that gold nanoparticles coated with dodecanethiol can be embedded between cell lipid bilayers when the particle size is below 5 nm (30).

The above confocal microscopy studies demonstrated the effectiveness of our cytosolic protein delivery technology. To further characterize the cell entry process, we hypothesized that the cargo proteins should have a certain amount of cholesterol coverage on the surface to be embedded between the bilayer. To probe this effect, we mixed the midsize protein lysozyme with the cholesterol tag at various molecular ratios (Fig. 2A and fig. S4). At low tag/protein ratios (molar ratios 1 and 3), only punctate intracellular fluorescence distribution was observed. As the ratio increased to 6, diffused fluorescence signal started to appear, indicating some contribution from direct membrane permeation, whereas a molar ratio of 8 led to substantial direct membrane permeation. This trend is interesting in that it revealed the significance of cholesterol density on protein surface for cell membrane permeation. When the density is low, cargo proteins attach to the outer leaflet of the plasma membrane, similar to the way that antibodies recognize cell surface receptors (Fig. 2A, right schematic panel). As a result, the protein molecule enters cells through endocytosis (similar to receptor-mediated endocytosis; fig. S5). A couple of details in these experiments deserve discussions. First, the optimal ratios for new cargo proteins should be probed and optimized since different proteins come in with different sizes, shapes, surfaces, and folding properties. For example, a molar ratio of 5 is sufficient for transmembrane delivery of the smaller aprotinin in contrast to the ratio of 8 for lysozyme. Second, at the optimal tagging ratios, the cargo proteins with several copies of the tag remained dispersed in solution rather than forming large aggregates as revealed by dynamic light scattering analysis. Only a small size increase was observed after cholesterol tagging (fig. S6).

Fig. 2 Effect of tag density on delivery and confirmation of delivery route.

(A) Titration of the tag/protein molar ratio for cytosolic delivery (top panels). Cell boundaries were circled in red. Scale bar, 10 μm. High tag/protein ratio producing sufficient cholesterol coverage of protein surface should favor direct membrane permeation (bottom left schematic), whereas insufficient cholesterol coverage likely will lead to initial protein anchoring on cell surface followed by endocytosis (bottom right schematic). Images of larger field of view are shown in fig. S4. (B) Cytosolic delivery independent of the endocytic pathway. Low temperature (4°C) that inhibits endocytosis had virtually no effect on cytosolic delivery of the tagged aprotinin (high tag/protein ratio at 5). Aprotinin without tags or undertagged (low tag/protein ratio of 2) was not uptaken by cells efficiently or confined in the cell membrane. Aprotinin was fluorescently labeled with AF488 (green), and the cell membrane was highlighted with a membrane marker wheat germ agglutinin (WGA-AF594; red). Scale bar, 10 μm. Broad field-of-view images are available in fig. S7.

To further rule out the scenario that the tagged proteins initially enter cells via endocytosis followed by an efficient endosomal escape, we performed the confocal microscopy experiment again using the small protein aprotinin at 4°C, a temperature blocking endocytic membrane trafficking. Without the tag, aprotinin uptake by cells was negligible, whereas at low tag/protein ratio, aprotinin was virtually confined to the cell membrane (Fig. 2B and fig. S7). Remarkably, at the optimal tag/aprotinin ratio, bright and homogeneous intracellular fluorescence was still observed at this low temperature, confirming that the protein-tag complex was not initially taken up by cells via endocytosis. To evaluate the generality of this protein tagging technology, we tested it using nine additional cell types besides HeLa cells. Efficient cytosolic delivery of aprotinin was observed, independent of the cell lines (Fig. 3A). Protein transfection directly into the cytosol was also achieved in primary cells, including human adipose–derived stem cells, primary human dermal fibroblasts, and primary human umbilical vein endothelial cells (HUVECs) (Fig. 3B). It is worth noting that the delivered aprotinin was found not only in the cytosol but also in the nucleus. This observation is consistent with previous reports, showing protein diffusion and exchange through the nucleus membrane pores when the protein size is below 60 kDa (31).

Fig. 3 Validation of the cholesterol tag–based protein delivery platform on various cells.

(A) Protein delivery in cell lines. (B) Protein delivery in stem cells and primary cells. The tagged aprotinin (molar ratio, 5:1) was incubated with cells for 2 hours at 37°C followed by counterstaining of the cell boundaries with WGA. Full names of the primary cells in (B) are human adipose–derived adult stem cells, primary human dermal fibroblasts, and primary HUVECs, respectively. Protein, aprotinin-AF488; WGA, WGA-AF594. Scale bar, 30 mm.

A remaining issue before we proceed to live-cell intracellular imaging and therapy is whether cholesterol tagging affects protein functions. To probe the effect, we tested two protein molecules, horseradish peroxidase (HRP) and green fluorescence protein (GFP), because their functions can be readily measured using optical methods (colorimetric and fluorescent assays, respectively). As shown in fig. S8, the fluorescence of GFP before and after cholesterol tagging exhibited no detectable difference. Similarly, the enzymatic activity of HRP was not affected either. These results showed well-preserved biological functions after cholesterol adsorbed onto protein surfaces.

Last, we proceeded to demonstrate two applications that should have a major impact on biology and medicine: immunofluorescence in live cells and delivery of therapeutic proteins inside cells. For immunofluorescence, one of the most common techniques in biological and medical research, currently, it is either used to profile cell surface markers on live cells (e.g., many flow cytometry studies), missing all the key cell signaling nodes inside cells, or used on fixed cells, missing out the temporal dynamics of cell signaling that determines cell fate, behavior, communications, and growth. To demonstrate immunofluorescence imaging of intracellular biomarkers in live cells, LifeAct, a fusion protein (molecular weight, ~30 kDa) of actin-bind peptide and GFP, was used because of the distinctive actin structure in cells and because of the significance of actin dynamics (32). As shown in Fig. 4A, when LifeAct-GFP was mixed with the cholesterol tag, it not only entered cells but also showed the unique filament structure, indicating preserved LifeAct functionality and labeling specificity. In contrast, LifeAct alone did not show any intracellular labeling (Fig. 4B). To confirm the labeling specificity, cells were pretreated with cytochalasin D before incubating with LifeAct-Tag. The filament structure was completely disrupted (Fig. 4C), demonstrating the feasibility of using our tagging technology in drug response tests.

Fig. 4 F-actin labeling in live cells using LifeAct-GFP (green).

(A and B) HeLa cells were incubated with cholesterol-tagged LifeAct-GFP or LifeAct-GFP alone. Actins’ signature filament structure was only observed in (A), where LifeAct-GFP was cholesterol-tagged (molar ratio, 9:1). (C) HeLa cells pretreated with cytochalasin D (an actin polymerization inhibitor) before the addition of cholesterol-tagged LifeAct-GFP. The disrupted actin structure shows the live-cell intracellular labeling specificity. (D) LifeAct-GFP delivered by Fuse-It-P; some filament structure was observed over a diffused hazy background. Some endosomal sequestration was also observed inside the cells (the bright dots highlighted by the white arrows). The cell nuclei were counterstained by Hoechst 33258 (blue).

We further compared our tagging technology with Fuse-It-P, a fusogenic liposome that is recommended and provided by the LifeAct-GFP company. As shown in Fig. 4 (A and D), under the same LifeAct-GFP concentration, our cholesterol tagging technology produced significantly better contrast. It is also worth mentioning that although, in theory, fusogenic liposomes should directly offload the encapsulated proteins in the cytosol upon fusion with the cell membrane, some liposomes still entered cells via endocytosis, where cargo molecules (LifeAct-GFP in this case) were trapped in endosomes and lysosomes. This sequestration can lead to two unfavorable outcomes: degradation of fluorescent probes, which creates diffused hazy background signals (Fig. 4D), and ultra-bright endosome and lysosome compartments, which create dotted/concentrated background inside cells (Fig. 4D, arrows).

For a proof-of-principle application in therapeutic protein delivery, we selected cytochrome c (Cyt c), an intracellular signaling protein for apoptosis (33), as a model cargo molecule. To track Cyt c delivery and intracellular distribution, it was labeled with a fluorescent dye (AF488, green). Compared to the control (no treatment), cells treated with Cyt c alone showed minor dotted intracellular green fluorescence. This is expected because Cyt c is a highly water-soluble macromolecule (molecular weight, ~12 kDa) that cannot diffuse through the cell membrane by itself. Instead, a minute amount of Cyt c was uptaken by cells but trapped in endosomes/lysosomes (Fig. 5A). In contrast, the addition of our cholesterol tag to Cyt c resulted in strong perinuclear fluorescence, likely due to Cyt c binding with intracellular apparatus such as the endoplasmic reticulum, initiating the apoptosis process (34). The cell viability study showed dose-dependent cell death (Fig. 5B and figs. S9 and S10), similar to previous reports using microinjected Cyt c (35). As an example, at a Cyt c concentration of 2 μM for cell incubation, no sign of toxicity was detected for cells that were untreated or treated with Cyt c alone (Fig. 5, C to E), while near-complete cell death was found for cells treated with the cholesterol-tagged Cyt c. This study proves the feasibility and potential of specific targeting of certain cell signaling nodes inside cells with protein drugs.

Fig. 5 Delivering Cyt c, a model intracellular protein drug, into living tumor cells using our tag.

(A) Fluorescence micrographs revealed that the Cyt c protein (AF488-labeled, green) cannot enter cells by itself. Aided by the cholesterol-CB tag (molar ratio, 8:1), Cyt c entered the cytoplasm and showed a perinuclear localization. The cell nuclei were counterstained by Hoechst 33258 (blue). (B) Cyt c–triggered cell apoptosis. Dose-dependent cell apoptosis was observed only in tagged Cyt c, but not Cyt c alone (average/SD from six measurements). (C to E) Representative cell viability micrographs obtained at Cyt c concentration of 2 μM (live-dead cell staining; live cells, green; dead cells, red). The fluorescence images for other concentrations in (B) are shown in figs. S9 and S10.


Although DNA and RNA provide the genetic codes, biological structures and functions are mostly realized by proteins and protein interactions. Direct observation, perturbation, modulation, and control of the molecular interactions provide a means for understanding important biological processes and treatment of diseases. For functional protein production, the biology of raising monoclonal antibodies, antibody fragments, nanobodies, and screening for functional peptides has become relatively mature and straightforward that an antibody or peptide can be found for virtually any target molecule. Toward cellular applications, however, the power and potential of functional proteins and peptides are substantially limited. One of the biggest problems is that most signaling nodes and drug targets are inside cells, inaccessible to macromolecular agents.

To overcome the barriers of cell membrane and endocytosis that are extremely effective in preventing intact biomolecules to enter the cytosol, we have developed an intracellular protein delivery technology based on noncovalent cholesterol tagging. The tag is simple to make by linking CB with cholesterol, both of which are biocompatible (fig. S11), and the technology is broadly applicable to any compact proteins and cell types. In contrast to conventional protein delivery technologies that are mostly based on endocytosis, our small-molecule tag enables proteins to permeate through the cell membrane without generating pores that cause cytotoxicity. This platform technology should open a whole new realm for basic biology studies and translational medicine.

For example, a fluorescently labeled nanobody or single-chain fragment variable (scFv) will enable direct visualization and dynamic tracking of intracellular targets in live cells, unlike conventional immunohistochemistry (IHC; based on fixed cells) that only provide a snapshot of a continuous event. For live-cell imaging, the concentration of protein imaging probe delivered inside cells should be controlled (e.g., by controlling the probe concentration in media and incubation time). Unlike conventional IHC where excess imaging probes can be washed away, unbound probes in live cells cannot be removed, potentially reducing the imaging signal-to-noise ratio. Fortunately, because of the excellent binding affinity of immunological agents (KD often in the nanomolar range), as long as the imaging probe is not in large access to the corresponding intracellular targets, the vast majority of the delivery probe molecules should remain in the bound state rather than in the unbound state at equilibrium. This is what we observed in actin labeling (Fig. 4).

Toward drug discovery, since the first protein drug insulin (5.8 kDa) became available in the 1980s, biologics have captured increasing attention from pharmaceutical scientists and industrial professionals, because of their distinctive advantages. In particular, protein drugs offer unmatched selectivity over small molecules, the capability to reach traditionally undruggable targets such as protein-protein interactions, and simplicity in screening (e.g., phage display and immunization technologies). Over 230 therapeutic proteins (including peptides) and their 380 drug variants have received regulatory approval for human uses in recent years (5). Despite this rapidly increasing number, chemically synthesized small molecules still make up over 90% of the drugs on the market today (36). This disproportion is because the majority of the targets for current drugs are located inside cells and accessible only to small molecules (8). The simple cholesterol tag platform has the potential to fundamentally change this situation and greatly expands the therapeutic target space. For example, technologies are well established to raise peptides, scFvs, or nanobodies against many intracellular targets to gain control of specific biological events. At this time, full-sized antibody (greatest availability) cannot be delivered directly into the cytoplasm using our cholesterol tag, a limitation that requires further development and optimization to address.

In summary, we have developed a unique concept to bring imaging and therapeutic proteins directly into the cytoplasm, bypassing endocytosis sequestration. This simple yet powerful technology not only enables immunolabeling of intracellular targets in live cells for dynamic imaging but also opens exciting opportunities for the development of protein therapeutics and cell engineering.



Unless otherwise noted, all organic solvents and chemicals were purchased from Sigma-Aldrich. Cholesteryl chloroformate was purchased from Alfa Aesar. Methyl triflate (MeOTf) and phosphorus oxychloride were purchased from TCI America. CB G250 was obtained from Chem-Impex International. All chemicals were of reagent grade and used as received. LifeAct-GFP and Fuse-It-P were purchased from ibidi Technologies. Cyt c was bought from MP Biomedicals. Alexa Fluor 488 N-hydroxysuccinimide (NHS), cytochalasin D, Hoechst 33258, LIVE/DEAD Viability/Cytotoxicity Kit, and Microplate BCA Protein Assay Kit were obtained from Fisher Scientific. CellTiter-Blue Cell Viability Assay was purchased from Promega Corporation. Air- and moisture-sensitive manipulations were performed with standard techniques under argon atmosphere. Column chromatography was performed using silica gel 60 (230 to 400 mesh) from Merck, and thin-layer chromatography was carried out on 0.25-mm Merck silica gel plates (60F-254). Nuclear magnetic resonance (NMR) spectra were obtained on a Bruker AV301 spectrometer. Chemical shifts are expressed in d (parts per million) values, and coupling constants are expressed in hertz. 1H spectra were referenced to tetramethylsilane as an internal standard. The following abbreviations are used: s, singlet; d, doublet; t, triplet; quint, quintet; m, multiplet; brs, broad singlet; and brd, broad doublet. Electron spray ionization (ESI)–mass spectra (MS) were measured on a Thermo LTQ-OT/Xcalibur 2.0 DS spectrometer.

Synthesis of the protein delivery tag

The cholesterol-based tag (compound 5) was synthesized by conjugating two copies of cholesterol to one CB molecule through an aminated linker. The synthetic route (compounds 1 to 5) is available in scheme S1.

Cholestyl 3-(3-aminopropyl methyl amino)propyl carbamate (3)

Compound 2, namely, 3,3′-diamino-N-methyldipropylamine (15 mmol), was dissolved in 30 ml of dry dichloromethane (DCM). To the clear solution, 1.35 g of cholesteryl chloroformate (3 mmol) was slowly added in portions. The reaction was stirred at room temperature for 12 hours. After addition of water (30 ml), the organic layer was separated. To completely remove excess compound 2, the organic phase was washed three more times with water. The organic phase was dried with anhydrous Na2SO4 and concentrated in vacuo. The product 3 was used in the next step without further purification (1.3 g, 82% yield). 1H NMR (300 MHz, CDCl3): δ0.68 (s, 3H), δ0.88 (d, 3H), δ0.93 (d, 3H), 1.01 (s, 3H), δ1.04 to 1.67 (m, 23H), δ1.75 to 2.05 (m, 5H), δ2.17 (s, 3H), δ2.20 (d, 2H), δ2.25 to 2.45 (m, 8H), δ3.23 (m, 2H), δ4.49 (m, 1H), δ5.37 (m, 1H), δ5.51 (brs, 1H). ESI-MS calcd for C33H59N3O2 529.84, found [M + H]+ 530.3.

CB, dicholest carbamyl 3-(3-propyl methyl amino)propyl sulfonamide (4)

CB sulfonyl chloride was synthesized from CB G250 according to a previous report (37). Briefly, CB G250 (100 mg) was dissolved in 5 ml of dry DC (dimethylformamide), followed by 15 ml of dry chloroform. To the solution, 100 μl of phosphorus oxychloride was added drop by drop. The mixture was refluxed for 2 hours at 50°C and then cooled to room temperature. Cold dry ethyl ether (100 ml) was added to the reaction to precipitate the product. The precipitated sulfonyl chloride was collected, washed with ether, dried in vacuo, and resuspended in 10 ml of dry DCM. To this suspension, 10 ml of the dry DCM solution of compound 3 (500 mg) was added, followed by 200 μl of triethylamine. The reaction was stirred at room temperature overnight. The crude product 4 was precipitated out from reaction with ethyl ether and used in the next step without further purification (110 mg, 55% yield).

CB, dicholest carbamyl 3-(3-propyl dimethyl amino)propyl sulfonamide (5)

N-methylation of tertiary amine linker in compound 3 was achieved by powerful methylating agent MeOTf. Briefly, 50 mg (0.027 mmol) of compound 4 was dissolved in 2 ml of dry DCM, followed by adding MeOTf (50 mg, 0.3 mmol). The reaction was stirred at room temperature for 24 hours. After washing with water, the organic layer was separated and concentrated in vacuo. The blue solid was washed with ethyl ether and purified on C-18 chromatography to produce a blue powder (40 mg, 78% yield). 1H NMR (300 MHz, CDCl3): δ0.67 (m, 6H), δ0.75 to 1.67 (m, 40H), δ1.69 to 2.5 (m, 12H), δ2.89 to 3.75 (m, 22H), δ4.01 (d, 4H), δ4.41 (m, 6H), δ5.33 (m, 2H), δ6.0 to 7.9 (m, 13H), δ6.0 to 7.9 (m, 8H), δ8.44 (m, 1H), δ11.02 (s, 1H). ESI-MS calcd for C115H170N9O9S2 1886.77, found [M + 2H]2+ 943.9.

Cell culture

The cancer cell lines were obtained from the American Type Culture Collection (ATCC). The primary cells, including human adipose–derived stem cells, primary human dermal fibroblasts, and primary HUVECs, were purchased from Zen-Bio Inc. MCF-7 (ATCC #HTB-22), MDA-MB-231 (ATCC #HTB-26), PC-3 (ATCC #CRL-1687), SK-BR-3 (ATCC #HTB-30), and HeLa (ATCC #CCL-2) cells were cultivated in RPMI 1640 medium supplemented with fetal bovine serum (FBS; 10%), penicillin (100 U/ml), and streptomycin (100 μg/ml). DU-145 (ATCC #HTB-81) and human embryonic kidney (HEK)–293T cells (CRL-11268) were cultivated in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with FBS (10%), penicillin (100 U/ml), and streptomycin (100 μg/ml). KB cells were cultured in Eagle’s minimal essential medium supplemented with FBS (10%), penicillin (100 U/ml), and streptomycin (100 μg/ml). All the primary cells were maintained and plated in dishes with the specific plating media provided by the cell vendor (Zen-Bio Inc). All the cell lines were propagated in a humidified 5% CO2 incubator at 37°C.


A Zeiss LSM 710 confocal laser-scanning microscope fitted with a plan apochromat objective (40×) was used. Alexa Fluor 488 was excited with a 488-nm argon ion laser (5% laser power), and fluorescence was recorded through frame scan. 4′,6-Diamidino-2-phenylindole (DAPI) and Alexa Fluor 594 were excited with 405- and 594-nm laser lines, respectively. Bright-field images were recorded with the differential interference contrast (DIC) setting.

Protein labeling

Proteins were chemically labeled with Alexa Fluor 488 NHS for imaging and tracking purposes. NHS-activated Alexa Fluor 488 reacts with the NH2 group on the protein surface, resulting in the covalent labeling of dyes on the protein surface. Briefly, proteins (pure solid) were dissolved in Dulbecco’s phosphate-buffered saline (DPBS; or pure water) at 6 mg/ml. Alexa Fluor 488 NHS dye (Thermo Fisher Scientific) was added in one portion with a reaction ratio of 3:1 (dye/protein molar ratio). The reaction was gently mixed at room temperature overnight, followed by purification on PD-10 columns (Bio-Rad). Protein aprotinin was labeled in pure water, instead of DPBS, because of its better solubility in buffers of low ionic strength.

Protein tagging

Purified CB-cholesterol tag powder was reconstituted with ethanol/water (4:1, v/v) to a concentration of 1.6 mM and stored at 4°C as a stock. Before tagging, the concentration of all the protein solutions was diluted to 25 μM. To 80 μl of protein solution, 7.5 μl of the tag solution was added. The mixture was quickly mixed by pipetting. The mixture was incubated in a water bath (37°C) for 8 min and sonicated in an ultrasonic water bath for 30 s. After another round of 8-min water bath incubation at 37°C, the protein was ready for delivery.

Cellular delivery

Cells were seeded on glass-bottom dishes (poly-d-lysine–treated; MatTek Corp.) 1 day before the experiments. The cell monolayer was washed once with serum-free RPMI 1640 media and treated with 2 ml of RPMI 1640 media containing the tagged protein at a final concentration of 1 μM for 2 hours. In the end, the cells were counterstained with Hoechst 33342 and Alexa Fluor 594–labeled wheat germ agglutinin (WGA-AF594), respectively. The cells were washed three times with RPMI 1640 and imaged on a Zeiss LSM 710 microscope.

To study the delivery effect at low temperature (4°C), cell monolayers were pre-equilibrated with cold media (RPMI 1640) for 10 min at 4°C and then treated with the cold solution of tagged protein for 3 hours. The cells were washed with cold media and imaged immediately.

Toxicity study

Cytotoxicity of the tag was evaluated on HeLa cells. Briefly, cells were seeded in 96-well plates at 3000 cells per well. CB, the tag, or tagged protein (BSA) was added at a series of concentrations from 10 μM to 3.3 nM. Untreated cells were used as control. After 48-hour incubation, 10 μl of CellTiter-Blue reagent (Promega) was added to the wells and the culture was incubated for another 1 hour. Cell viability was examined on a microplate reader (Tecan) at excitation/emission of 560/590 nm, and the toxicity was analyzed according to the manufacturer’s protocol.

Function test of the tagged protein

GFP (EMD Millipore) was dissolved in water and tagged with the protein delivery tag. The tagged protein was kept at 37°C for 3 hours and then diluted to 100 ng/ml with water. This diluted solution was imaged with a fluorescence imaging system (Light Tools Research, Pasadena, CA). The freshly tagged GFP and nontagged GFP were used as control. The function of tagged HRP (Sigma-Aldrich) was evaluated similarly. The enzyme activity was detected by incubating with a 3,3′,5,5′-Tetramethylbenzidine substrate solution (Thermo Fisher Scientific) and reading with the Tecan Infinite 200 Microplate Reader.

F-actin labeling in live cells using tagged LifeAct-GFP

HeLa cells were seeded onto glass-bottom dishes. After 1 day of culturing, cells were incubated with fresh media containing tagged LifeAct-GFP for 4 hours at 37°C, followed by washing and counterstaining Hoechst 33258. The cells were imaged on an Olympus fluorescence microscope equipped with a true-color charge-coupled device, QColor 5.

Cancer cell viability upon treatment with Cyt c

HeLa cells were seeded onto glass-bottom dishes and cultured overnight. The cells were treated with Cyt c (tagged or untagged) of various concentrations in culture media for 4 hours, washed, and cultured in fresh media for another 48 hours. Cell viability was measured using the dual-color LIVE/DEAD Viability/Cytotoxicity Kit following the manufacturer’s instruction.


Supplementary material for this article is available at

This is an open-access article distributed under the terms of the Creative Commons Attribution-NonCommercial license, which permits use, distribution, and reproduction in any medium, so long as the resultant use is not for commercial advantage and provided the original work is properly cited.


Acknowledgments: Funding: This work was supported, in part, by the Office of Research at the University of Washington and the Washington Research Foundation. Author contributions: W.T., P.Z., and X.G. conceived the idea and designed the experiments. W.T. and P.Z. performed the experiments. W.T., P.Z., and X.G. analyzed the data and wrote the paper. Competing interests: The authors declare that they have no competing interests. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials. Additional data related to this paper may be requested from the authors.

Stay Connected to Science Advances

Navigate This Article