Cell-Penetrating Streptavidin: A General Tool for Bifunctional Delivery with Spatiotemporal Control, Mediated by Transport Systems Such as Adaptive Benzopolysulfane Networks.

In this report, cell-penetrating streptavidin (CPS) is introduced to exploit the full power of streptavidin-biotin biotechnology in cellular uptake. For this purpose, transporters, here cyclic oligochalcogenides (COCs), are covalently attached to lysines of wild-type streptavidin. This leaves all four biotin binding sites free for at least bifunctional delivery. To maximize the standards of the quantitative evaluation of cytosolic delivery, the recent chloroalkane penetration assay (CAPA) is coupled with automated high content (HC) imaging, a technique that combines the advantages of fluorescence microscopy and flow cytometry. According to the resulting HC-CAPA, cytosolic delivery of CPS equipped with four benzopolysulfanes was the best among all tested CPSs, also better than the much smaller TAT peptide, the original cell-penetrating peptide from HIV. HaloTag-GFP fusion proteins expressed on mitochondria were successfully targeted using CPS carrying two different biotinylated ligands, HaloTag substrates or anti-GFP nanobodies, interfaced with peptide nucleic acids, flipper force probes, or fluorescent substrates. The delivered substrates could be released from CPS into the cytosol through desthiobiotin-biotin exchange. These results validate CPS as a general tool which enables unrestricted use of streptavidin-biotin biotechnology in cellular uptake.

Introduction

Streptavidin[1−4] is a 52 kDa β-barrel tetramer that binds one biotin per monomer with exceptionally high affinity because the barrels close upon binding (Figure A). The advantages of such encapsulation combined with tetravalency, stability, and ease of use empower the high versatility of biotin–streptavidin technology. For cellular uptake, streptavidin has been used to noncovalently couple biotinylated substrates to biotinylated transporters such as cell-penetrating peptides (CPPs)[5] and other dynamic covalent systems,[6] including cell-penetrating poly(disulfide)s,[7,8] related disulfide-containing systems,[9,10] and cyclic oligochalcogenides (COCs, Figure B).[11−15] However, this approach is limited to the reliable delivery of only one functionality because it is difficult to control the interfacing of more than two different ligands with the streptavidin tetramer. This limitation is occasionally overcome by covalent linking of fluorophores to proteins.[14] A more powerful solution would be cell-penetrating streptavidin (CPS) with covalently attached transporters and all four binding sites free to harness the full power of streptavidin–biotin technology for bifunctional delivery (Figure C) with, for example, retention-using-selective-hooks (RUSH)-like[16] spatiotemporal control (Figure D).

Figure 1

(A) Molecular model of wild-type streptavidin tetramer with bound biotins (yellow) and lysine residues (red, front view) used to (C) covalently attach transporters. (B) Conventional use of streptavidin to interface biotinylated transporters with one biotinylated substrate. (C) Cell-penetrating streptavidin (CPS) with all biotin binding sites free to interface with two different biotinylated substrates S1 and S2 for bifunctional uptake with (D) spatiotemporal control.

Covalent protein modification has been used previously for cellular uptake. Examples include supercharging of proteins by either addition of positive or removal of negative charges[17] or simple covalent conjugation to possible new transporters such as boronic acids or halogen-bond donors.[5−10,18]

To elaborate on the idea of CPS, we selected COCs as transporters. COCs such as asparagusic acid (AA),[11,12] epidiketodithiopiperazines,[13] diselenolipoic acid (DSL),[14] or the most recent benzopolysulfanes (BPS)[15] are currently being explored to access increasingly unorthodox dynamic covalent oligochalcogenide exchange chemistry on the way into the cytosol. Such thiol-mediated uptake of COCs[11−15] and related transporters[7−10] has allowed delivery of not only small molecules but also larger substrates such as DNA,[9b] antibodies,[9c] quantum dots,[8] other nanoparticles,[9c] liposomes, and polymersomes[11d] to the cytosol without significant capture within endosomes. Mechanistic hypotheses envision COCs as molecular walkers, walking along disulfide tracks in membrane proteins[12,14,19,20] through the transient micellar pores known from CPPs but also from disulfide-rich scramblases or voltage-gated ion channels (Figures B and 2C).[14,20,21] Driving the growing impact of dynamic covalent chemistry to cellular uptake[6−14] to the extreme, BPS have been hypothesized to act by forming adaptive networks of rare sulfur species such as macrocycles 1 containing up to 19 sulfur atoms for cells to select from (Figure D).[15] BPS are known to occur in marine natural products,[22] have attracted early attention in total synthesis,[23] and appeared top in recent library screens to reverse cognitive defects in mouse models.[24]

Figure 2

(A) Synthesis of CPSs 3–6: (a) PBS, pH 7.4, rt, 2 h; (b) CuSO4·5H2O, BTTAA, sodium ascorbate, aminoguanidine hydrochloride, PBS, pH 7.4, rt, 1 h; (c, d) ref (15). (B–D) Working hypotheses for the modes of action of COCs as molecular walkers (B) first exchanging with exofacial thiols and (C) then walking along disulfide tracks and transient micellar pores into cells and of (D) BPS as an adaptive network including macrocycles as large as 1.

The objective of this study was to create a general tool which warrants unrestricted availability of streptavidin–biotin technology in cellular uptake. With COCs as a timely example of freely variable transporters, bifunctional delivery with spatiotemporal[16] controllability is tackled as a functional challenge of biological relevance (Figure D). Specifically, HC-CAPA, a new combination of the chloroalkane penetration assay (CAPA)[25,26] with automated high content fluorescence microscopy,[27] is introduced first to secure direct quantitative functional evidence for cytosolic delivery of CPS. On the basis of these results, compatibility of CPS with targeted delivery, the interfacing of nanobodies,[8] peptide nucleic acids (PNAs),[28] and mechanosensitive fluorescent flipper probes[29] with HaloTag-GFP fusion proteins,[25,30] and their controlled release through intracellular desthiobiotin–biotin exchange are demonstrated.[3,4,16,29a]

Results and Discussion

Design and Synthesis of CPS

Wild-type streptavidin tetramer 2 contains 16 lysine residues (Figures A and 2A),[1−4] which were partially derivatized to give CPS with different numbers of COCs. We chose to chemically modify amines rather than carboxylates to preclude misleading CPP-like contributions[5−8,21] to the uptake activity of CPS due to increased net positive charge.[26,17] The synthesis of CPSs 3–6 is shown as an example (Figures A and 3, Schemes S7 and S8). The reaction of 2 and a controlled amount of activated ester 7 gave modified streptavidin 8 with up to ca. eight azide groups. The degree of functionalization was estimated using a colorimetric amine assay (Figure S1). The obtained key intermediate 8 underwent copper-catalyzed alkyne–azide cycloaddition reaction (CuAAC) quantitatively with alkynylated transporter 9 under mild conditions to give CPSs with ca. one (3, BPS1), two (4, BPS2), four (5, BPS4) or eight covalently linked BPSs (6, BPS8, Figure A). BPS 9 was prepared by multistep synthesis from catechol 10 following previously reported procedures.[15] CuAAC between pre-CPS 8 and other alkynylated transporters and controls afforded CPSs covalently linked with four or eight AAs (11, AA4; 12, AA8), four AA-arginines (13, RAA4), four or eight DSLs (14, DSL4; 15, DSL8), four DSL-arginines (16, RDSL4), and four or eight myristic acids (17, MA4; 18, MA8, Figure , Schemes S1, S7, and S8). An overview of CPS with different transporters, CPS complexes with different substrates, and controls described in the following is provided in Figure . A comprehensive summary of all structures used can be found in Figure S39, Supporting Information.

Figure 3

Structure of monomers, CPS with n transporters T, CPS complexes with n substrates S1 and S2, and controls made and used in this study with representative assembly of 39 and indication of pertinent figures (Fig) reporting on activity. n is the stoichiometry used for complex preparation (35: n = 2–4, see text); 47 and 49 are a mixture of m/p-carboxyfluorescein isomers.

Cytosolic Delivery of CPS Evidenced by HC-CAPA

CAPA has been introduced recently by the Kritzer group to quantify cytosolic delivery reliably based on function.[25] This is important because the commonly used fluorescence intensity-based assays are often significantly affected by the environment[2,11−15] and, in general, demonstrate neither cytosolic location nor intact functionality.[25,26] CAPA operates with HGM cells, e.g., HeLa cells that have a fusion protein of HaloTag and GFP (Figure F, green) expressed on the outer surface of mitochondria (Figure A).[25,30] Delivered chloroalkanes, such as 19, react in cytosol with HaloTags (Figure B) and thus inhibit reaction with the subsequently added fluorescent chloroalkane 20 (Figure C and 4H vs 4I, red). Thus, the delivery of chloroalkanes can be quantified from the fluorescence intensity of 20 on mitochondria. To secure quantitative data of the highest possible quality, we adapted CAPA to an image-based assay using automated high-content fluorescence microscopy (referred to also as high-content screening (HCS) or high-content analysis (HCA)),[27] preserving the power of flow cytometry to record and analyze thousands of cells in a short period of time without losing the subcellular spatial resolution of fluorescence imaging. For HC-CAPA, the mitochondria with HaloTags were first segmented applying a top-hat transform to the unmodifiable fluorescence of the GFP (Figure F). Then this mask was used to collect the intensity of 20 only in the region of mitochondria, thus excluding any possible false positives from off-target signals (Figure G, Figures S2–S5).

Figure 4

(A–C) Schematic representation of CAPA: (A) HGM cells were incubated with (B) 21–31 and then (C) 20. (D) HC-CAPA dose–response curves for 23 (●), 25 (▼), and 28 (■). (E) CP50 values of 21–26 and 28. Statistical significance (**** p < 0.0001) was estimated by the t test using GraphPad Prism. (F–I) Representative images for HC-CAPA: (F) top-hat transform (white) of the image generated from the emission of GFP (green); (G) cell body (light blue) and mitochondria (yellow) masks applied to selectively extract from the mitochondria the emission intensity in the rhodamine channel; (H) rhodamine channel (red) from labeling with 20 without incubation with transporters and (I) after incubation with 20 μM 23 (nuclei in blue; Hoechst 33342, poststaining, scale bars 50 μm).

The newly devised HC-CAPA was then used to assess cytosolic delivery of CPS-chloroalkane complexes 21–31, prepared by mixing the corresponding CPS with biotinylated chloroalkane 19 (Figure , Scheme S9). The absence of unbound 19 was assured using a substoichiometric 3 equiv of ligand per CPS and by meticulous purification using centrifugal filters. Results from HC-CAPA were quantified as CP50, which is the CPS concentration needed for 50% inhibition of HaloTagging by rhodamine 20. The best CP50 = 2.1 ± 0.3 μM was found for BPS423, with the onset of activity detectable already in the nanomolar range (Figures D (●), 4E, and 4I). This activity exceeded the CP50 = 3.1 ± 0.5 μM reported for chloroalkane-tagged Tat peptide, the original CPP from the HIV virus.[25] Although direct quantitative comparison of these two values has to be made with due caution considering the different methods used for data collection, it perfectly illustrates the high efficiency of cytosolic delivery with BPS423. BPS together with a large 52 kDa protein outperforming this small undecapeptide in cytosolic delivery is particularly interesting considering that uptake efficiencies in general decrease rapidly with increasing substrate size.[5,8,9,11]

The presence of fewer and more transporters in BPS121, BPS222, and BPS824 reduced activity (Figure E). The second best COC was AA with CP50 = 8 ± 1 μM for AA826 and CP50 = 14.9 ± 0.5 μM for AA425, followed by DSL428 with CP50 = 34 ± 6 μM (Figures D–I). The presence of more transporters in DSL829 rather deteriorated the poor performance of DSL under these conditions. Insertion of guanidinium cations between the protein and COCs in RAA427 and RDSL430 was also ineffective in improving the transport activity, thus excluding contributions from CPP-like mechanisms[7−11] (CP50 = 20 ± 1 μM for 27, >35 μM for 30). Decreasing activities from BPS423 to BPS824 but not from AA425 to AA826 suggested that the higher hydrophobicity of BPS starts to hinder thiol-mediated uptake in the presence of more than four transporters per streptavidin, possibly due to decreasing solubility in water, aggregation, and even precipitation (Figure E). The best activity found with an intermediate number of BPS, or “Goldilocks” behavior,[31] implied that the right balance between sufficient hydrophilicity and number of transporters is rather subtle. Goldilocks behavior further supported that simple passive diffusion does not account for the top activity of BPS423. Corroborative experimental support for this important conclusion was secured with myristyl control MA431, which is as hydrophobic as BPS but completely inactive (Figure S6, Table S3, and Scheme S9).

We previously reported the mechanistic insights in support of thiol-mediated uptake of BPS, including partial inhibition with Ellman’s reagent which inactivates cell surface thiols, insensitivity toward endocytosis inhibitors (cytochalasin B, wortmannin, chlorpromazine, or methyl-β-cyclodextrin), the formation of complex adaptive networks of extreme sulfur species in the presence of thiols and disulfides (up to heptamers 1, Figure D), and stability in PBS buffer for >2 weeks.[15] In the current context focusing on cell-penetrating streptavidin, which is a general and unrestricted availability of biotin–streptavidin biotechnology for cellular uptake independent of the transporter used, repetition of these mechanistic studies on BPS-mediated uptake was less relevant and off topic. Nevertheless, retention on and thiol-meditated release from thiol-affinity columns, arguably the most compelling test for thiol-mediated uptake,[14a,15] was confirmed for intact BPS423 (Figure S7). As previously noted with AA,[12] the presence of 10% serum in Leibovitz’s media did not affect the activity much, resulting in a less than doubled CP50 of BPS423.

Targetless Delivery of CPS into Various Cells

The delivery of four biotinylated fluoresceins 32(14) in BPS433 into unmodified HeLa cells afforded confocal laser scanning microscopy (CLSM) images with diffusely emitting cells, including their nuclei (Figures B and S8).[15] Similar images were obtained with biotinylated rhodamine 34(2) in BPS435 inside C2C12, HeLa, MCF7, or MDCK cells (Figures A and S9; n = 2). Weaker fluorescence was observed in CLSM images of AA436 in HeLa cells (Figure C). The added cations in RAA437 and RDSL438 did not help to improve the activity significantly, confirming that CPP-like mechanisms[7−11] are negligible for COC-mediated uptake (Figures D and S8). Contrary to BPS433, images of AA436 and RAA437 contain bright punctate patterns on a diffuse background, which usually indicate contributions from endolysosomes and perhaps also precipitating aggregates besides emission from the cytosol. With flow cytometry, such contributions from endosomal capture are recorded as false positives. The here introduced HC-CAPA is powerful because all eventual false positives[25,26] are eliminated without losses in fast acquisition of statistically relevant high-content data.

Figure 5

CLSM images of (A) C2C12 cells after incubation with fluorescently labeled BPS435 (n = 2, 10 μM, 2 h), of HeLa cells after incubation with (B) BPS433, (C) AA436, and (D) RAA437 (all 10 μM, 8 h), and of HGM cells after incubation with (E) BPS441 (emission top, 40; bottom, GFP; 5 μM, 2 h), (F) 39 (top, 34; bottom, 34 (red) + GFP (green), 5 μM, 2 h), (G) 45 (top, 34; bottom, 34 (red) + GFP (green), 2.5 μM, 2 h), and (H) 46 (top, 40; bottom, 40 (red) + GFP (green), 5 μM, 2 h) in Leibovitz’s media; complex stoichiometry indicates the ratios of preparation. (F′–H′) Design of experiments F–H. Scale bars: 10 μm.

Targeted Delivery of Fluorophores, Nanobodies, and PNA

Using the best CPS, that is, BPS45, we explored the simultaneous delivery of two functions, here a reporter and a targeting unit. BPS45 was loaded with biotinylated chloroalkane 19 and rhodamine 34 (Figures and 5F′, Scheme S12). Incubation of HeLa cells expressing HaloTags on mitochondria with the resulting complex 39 afforded CLSM images with fluorescently labeled mitochondria (Figure F, top). The diffuse emission from cytosol and nuclei observed without targeting (Figures A and 5B) disappeared, thus confirming that delivery of complex 39 to the HaloTag on the surface of mitochondria is as specific as expected. A good colocalization with GFP emission demonstrated the targeted delivery of the complex to HaloTags (Figures F, bottom, S10A, and S10B). HaloTag targeting was substrate independent; the substitution of rhodamine 34 with flipper 40(29) gave similar images for delivery of the respective complex 41 to HaloTags on mitochondria (Figures , 5E, and S11). This result was of interest for the fluorescence imaging of precisely localized membrane tension within living cells.[29] Targeted delivery of biotinylated and Cy5-labeled 18-mer PNA 42(28) in complex 43 was also accomplished in HGM cells (vide infra, Figure S13).

Biotinylated anti-GFP nanobodies[8]44 were probed as alternative targeting units in complex 45, recognizing GFP instead of HaloTag of the fusion protein on mitochondria. The resulting images showed good colocalization with GFP together with less important endolysosomal capture and/or precipitates (red dots, Figures G and S12). These off-targeted, endolysosomal capture and/or precipitates probably originate from the larger oligomers of complex 45 cross-linked through the multivalent anti-GFP nanobodies[8]44 with more than one biotin (Figures and 5G′). Shifts in uptake mechanism from cytosolic delivery to endosomal capture have been observed frequently with increasing substrate size, although much more for CPPs[5] than for thiol-mediated uptake.[8,9,11] To minimize these side effects, complex 45 was prepared by incubation of BPS45 with 3 equiv of fluorophores 34 before incubation with the multivalent nanobodies 44, a procedure that should in turn increase formation of the targetless CPS side product 35 loaded with four fluorophores 34 only (Figure , n = 4). However, eventual contributions from off-targeted delivery with CPS 35 were very minor (diffuse red as in Figures A, 5G, and S12). Considering this situation, the targeted delivery mediated by anti-GFP nanobodies[8]44 to GFP expressed on the surface of mitochondria was remarkably efficient and supported, with red fluorescence originating from ligand 34, that the majority of complex 45 reaches the cytosol in intact functional form (yellow, Figures G, 5G′, and S12). The control complex 46 without COCs was unable to penetrate cells and showed no colocalization with GFP (Figures H, S10C, and S10D).

Controlled Intracellular Release of Substrates from CPS

Once delivered in the cells, the substrate would ideally be detached from transporter and target at a well-defined time[16] to ensure and follow its proper function on demand (Figure D). To elaborate on the timed release of substrates from CPS, fluorescein as a responsive model substrate was equipped with desthiobiotin 47, N-ethylbiotin 48, and biotin 49 (Figures and 6, Schemes S4 and S13). Desthiobiotin (Kd = 10–11 M)[3] and N-ethylbiotin (Kd = 10–9 M)[4] bind to streptavidin but with reduced affinity compared to biotin (Kd = 10–15 M).[3,4] The corresponding complexes 50–52 were readily accessible by incubation with BPS45 (Figures and 6).

Figure 6

(A) Delivery of 47 with BPS45 followed by addition of biotin 53 to release 47 from complex 50 in the cytosol. (B–D) Original image of the multiwell plate to monitor the time course of substrate release in HeLa cells by automated high-content fluorescence microscopy. Thousands of cells preincubated with (B) 51, (C) 50, or (D) 52 (10 μM, 6 h in Leibovitz’s medium) were treated with 53 (middle), 54 (bottom; 40 μM each), or none (top) and imaged every 12 min (left to right; shown are every other images). (E–G) Spatiotemporal control of (F) the delivery of 58 with 5 and 59 to the mitochondria of HGM cells (5 μM, 2 h incubation in Leibovitz’s medium) and (G) the release of 58 after addition of 53 (58 (red) + GFP (green), 40 μM). Scale bars: 10 μm.

Desthiobiotin 47 was first delivered in complex 50; then biotin 53, a nontoxic vitamin, or its methyl ester 54 was added at a 40 μM concentration as recommended by the RUSH technology[16] to diffuse into the cell, exchange, and release substrate 47 in the cytosol (Figure A). As streptavidin strongly quenches the fluorescence of bound carboxyfluorescein,[2] intracellular desthiobiotin–biotin exchange was detectable by fluorescence recovery of the released substrate 47, whereas the biotin–streptavidin side product 55 or 56 passes undetected.

Quantitative release kinetics were obtained by automated high-content fluorescence microscopy of wells containing thousands of cells with an increasing incubation time with biotin 53 and its methyl ester 54 (Figures B–D and S14). Identical t50 = 17 min obtained for controlled release of desthiobiotin 47 from CPS 50 with 53 and 54 demonstrated that passive diffusion of the protonatable carboxylate 53 into the cell is not rate limiting (Figures C and S14). In clear contrast, the release of N-ethylbiotin 48 from complex 51 was spontaneous and not affected by the presence of either 53 or 54 (Figure B). The lower fluorescence intensity observed with 48 (Figure B) compared with 47 (Figure C) suggested that 48 dissociated from CPS already during incubation time, diffused out of cells, and was washed away before the start of the acquisition (Figure S14C). Similar spontaneous release of desthiobiotin 47 was slow on the time scale of biotin-controlled release (Figures C and S14B). This difference in spontaneous release was in agreement with the poorer binding affinity of N-ethylbiotin 48 (Kd = 10–9 M)[4] compared to desthiobiotin 47 (Kd = 10–11 M).[3] On the other hand, the controlled release of 49 from 52 by biotin–biotin exchange was negligible even after an extended incubation period (Figure D), consistent with much stronger binding of biotin 49 compared to desthiobiotin 47. Insufficient release of biotin 49 and excessive release of N-ethylbiotin 48 added up again in Goldilocks behavior,[31] identifying desthiobiotin 47 with neither too strong nor too weak binding as the ligand of choice for controlled release within cells.

To finally combine targeted delivery and controlled release, PNA was selected as a prototypical example of a useful reagent that has been limited in scope by its cellular permeability.[28] In complex 57, CPS 5 was loaded with desthiobiotinylated 18-mer PNA 58 and chloroalkane 59 (Figures and 6E, Schemes S5, S6, and S13). The good colocalization (yellow) of Cy5 attached to PNA 58 (red) and GFP (green) demonstrated targeted delivery of complex 57 to HaloTags on the mitochondria of HGM cells in intact form and with excellent selectivity (Figures F and S15). Subsequent addition of biotin 53 induced PNA release, detectable as weakened emission of Cy5 on the mitochondria due to dilution of the liberated PNA 58 into the cytosol (Figures G, 6E, and S15). These results thus supported the compatibility of CPS delivery with spatiotemporal control using biotin analogs.[16]

Conclusion

This study introduces CPS as a privileged scaffold for cellular uptake, simple and robust, with biotin–streptavidin technology made fully available for general bifunctional delivery with spatiotemporal control. HC-CAPA, the recent CAPA[25] combined with automated high-content fluorescence microscopy[27] is devised as the method of choice to quantify cytosolic delivery; the currently emerging benzopolysulfanes[15] are validated as COCs of choice for cytosolic delivery. Their performance supports the integration of increasingly unorthodox[32] oligochalcogenide exchange chemistry to find new ways to enter into cells and encourages studies on mechanistic hypotheses, i.e., molecular walkers and dynamic-covalent adaptive networks.[14,15,19−22]