Multifunctional streptavidin–biotin conjugates with precise stoichiometries

Streptavidin is ubiquitously used to link different biotinylated molecules thanks to its tetravalent binding to biotin. An unwanted side-effect is the resulting statistical mixtures of products. Here, a general approach to form multifunctional streptavidin conjugates with precise stoichiometries and number of open binding pockets is reported. This method relies on an iminobiotin-polyhistidine tag, which allows separating streptavidin conjugates with different numbers of tags, and later reopening binding pockets at lowered pH to introduce a second functionality. Pure fluorescently labelled mono-, diand trivalent streptavidin– biotin conjugates prepared in this way were used for imaging biotinylated cell surface molecules with controlled clustering. Furthermore, these conjugates were functionalized with a second biotinylated molecule, folic acid–biotin, to investigate the importance of multivalent binding in targeted delivery of cancer cells. These streptavidin–biotin conjugates with precise stoichiometries combined with a variety of biotinylated molecules render this method a diverse and powerful tool for molecular biology and biotechnology.


Introduction
The binding of streptavidin to biotin is well known for the strong noncovalent interaction with femtomolar affinity (K d ¼ 10 À14 ). 1 The high affinity, slow exchange rate, and good specicity of the biotin-streptavidin interaction has resulted in a wide range of biotechnological applications including extracellular and in vitro labelling, 2,3 therapeutics, biosensing and biofunctionalization. 4,5 The large range of biotinylated small molecules, peptides, proteins, antibodies and nucleic acids as well as materials adds to the diversity of the biotin-streptavidin chemistry. Each streptavidin tetramer has four independent biotin binding sites. This tetravalence poses a conundrum. On the one hand, it allows streptavidin to be used as a linker between a wide variety of biotinylated molecules with targeting, sensing, diagnostic and therapeutic functionalities and assembling them into one molecule in a modular fashion. On the other hand, tetravalence can be a disadvantage as it leads to statistical mixtures of conjugates when multiple biotinylated molecules are conjugated as well as unwanted cross-linking and aggregation. 6,7 A partial solution to the problems arising from the multivalence are genetically engineered monovalent versions of streptavidin with only one active biotin binding site per tetramer, as well as a monomeric biotin binder, rhizavidin, which can almost achieve multimeric streptavidin-like binding stability for biotin conjugates. [6][7][8][9][10] Alternatively, (strep)avidins composed of different subunits have been used to integrate different functionalities within one conjugate. 11,12 These streptavidins have been used to form structurally dened one-toone streptavidin-biotin conjugates and image biotinylated cell surface receptors without the formation of articial receptor clustering. However, the monovalent streptavidins are unt as linkers to assemble streptavidin conjugates with multiple functional groups, or with multiple copies of one functional group. Given this, a method of assembling multifunctional streptavidinbiotin conjugates with precise stoichiometries still remains a big challenge and if successful, it would signicantly expand the potential of streptavidin-biotin based technologies.
In this study, we report a new strategy to assemble multifunctional streptavidin-biotin conjugates with precise stoichiometries and different valencies. Using native streptavidin we were able to produce streptavidin (S) conjugates with one (SA 1 ), two (SA 2 ) or three (SA 3 ) copies of a biotinylated molecule (A), where the residual biotin binding pockets remain open to introduce a second biotin conjugated functionality. Using this method, we were able to produce precise uorescent streptavidin conjugates for cell surface labelling and to investigate how the number of targeting ligands per conjugate affects cellular uptake. These initial illustrations show how this method can be used to address a wide range of questions in molecular and cell biology. chemistries; the separation of proteins with different numbers of His-tags using Ni 2+ -NTA (nitrilotriacetic acid) columns, and secondly, the pH-dependent binding of iminobiotin to streptavidin (basic pH, K d $ 10 À11 M; acidic pH, K d $ 10 À3 M). 13 In this method, varying numbers of biotin or iminobiotin conjugated His-tags (Bio-His-Tag or Ibio-His-Tag, respectively) were rst introduced on to streptavidin (Fig. 1). This allowed for the easy separation of species with differing numbers of His-tags and open biotin binding pockets on a Ni 2+ -NTA column using an imidazole gradient. Subsequently, a biotin conjugated molecule of choice (A) was coupled to the open binding pockets yielding pure streptavidin conjugated with precise stoichiometries (S(His-Tag) 3 A 1 , S(His-Tag) 2 A 2 , S(His-Tag) 1 A 3 ). In the case of the iminobiotin conjugated His-tag complexes, the His-tag was released by decreasing the pH to 3.5 and a second biotin conjugated molecule (B) was added with a precise stoichiometry (SA 1 B 3 , SA 2 B 2 , SA 3 B 1 ).
In a rst step, to produce stoichiometrically dened streptavidin conjugates with one functionality, streptavidin, (S ¼ 30 mM) was incubated with a biotin conjugated His6-tag peptide (biotin-(His) 6 , Bio-His-Tag ¼ 90 mM), which yielded a statistical mixture of S, S(Bio-His-Tag) 1 , S(Bio-His-Tag) 2 , S(Bio-His-Tag) 3 and S(Bio-His-Tag) 4 . Subsequently, the mixture was passed over a Ni 2+ -NTA agarose column; unbound molecules without Histags such as S were washed off and different species were eluted using an imidazole gradient, where species bearing more tags eluted at higher imidazole concentrations. In the chromatogram there were four clearly separated peaks that eluted at different imidazole concentrations (relative integrated areas: 19.4% 1 st peak, 14.3% 2 nd peak, 20.6% 3 rd peak, 45.7% 4 th peak) (Fig. 2a). Presumably the ratio of these peaks can be changed using different S to Bio-His-tag ratios. To demonstrate how each  of these peaks corresponds to a single species with a dened stoichiometry, the open biotin binding pockets of S(Bio-His-Tag) 1-4 were titrated with biotin-5-uorescein where the uorescence is quenched upon streptavidin binding (Fig. 2b-e). 14,15 The streptavidin conjugate in the rst peak required 3 equivalents of the dye to saturate all biotin binding sites and therefore was assigned as the S(Bio-His-Tag) 1 . Likewise, the molecules in the second, third and fourth peaks required 2, 1 and 0 equivalents of dye to saturate all biotin binding sites, respectively, and they corresponded to S(Bio-His-Tag) 2 , S(Bio-His-Tag) 3 and S(Bio-His-Tag) 4 , respectively. This assignment is also consistent with streptavidin conjugates eluting at higher imidazole concentrations having more Bio-His-Tags.
The identity of the different species was further conrmed with MALDI-TOF mass spectroscopy. While S and Bio-His-tag have molecular weights of 52 905 and 1500 Da (Fig. S1 †), respectively, the different conjugates of S and Bio-His-Tag have higher molecular weights (Fig. 2f). The isolated species with one to four Bio-His-tags had maximal peaks 54 405 Da (S(Bio-His-Tag) 1 ), 55 905 Da (S(Bio-His-Tag) 2 ), 57 405 Da (S(Bio-His-Tag) 3 ) and 58 905 Da (S(Bio-His-Tag) 4 ), respectively, which are well in agreement with the theoretically expected values. Moreover, for species with different numbers of Bio-His-Tags lower molecular weight peaks were observed due to the dissociation of the Bio-His-Tag from S in the mass spectrometer. More specically, there were two peaks for S(Bio-His-Tag) 1 (S and S(Bio-His-Tag) 1 ) and three peaks for S(Bio-His-Tag) 2 (S, S(Bio-His-Tag) 1 and S(Bio-His-Tag) 2 ), while no peak of S(Bio-His-Tag) 2 was observed for S(Bio-His-Tag) 1 . These results conrm and fully resolve the molecular structure of the prepared streptavidin conjugates with dened stoichiometries.
The isolated S(Bio-His-Tag) 1 , S(Bio-His-Tag) 2 and S(Bio-His-Tag) 3 conjugates are equivalent to trivalent, divalent and monovalent streptavidin and can be used to form precise conjugates with a large variety of biotinylated molecules including small molecules, peptides, proteins, DNA and antibodies. As an example, biotinylated mOrange (orange uorescent proteins, O, 31.3 kDa) was reacted with the streptavidin conjugates of varying valences. Each of the conjugates eluted at a different volume on a size exclusion column, and the molecular weights determined, based on a protein standard, were in agreement with the theoretically expected values (S(Bio-His-Tag) 1  For the success of this approach, the slow exchange rate of the Bio-His-Tag with streptavidin is required so that species isolated in different peaks do not interconvert. To test the kinetic stability, puried S(Bio-His-Tag) 2 was stored at 4 C, room temperature (RT), and at 37 C. Aer 1 and 5 days, the samples were loaded onto the Ni 2+ -NTA agarose column and eluted using an imidazole gradient (Fig. 3). The S(Bio-His-Tag) 2 stored at 4 C and RT for up to ve days eluted as a single peak like the fresh sample (0 day), demonstrating the long-term stability of isolated species. Even at 37 C aer one day, there was no signicant change for S(Bio-His-Tag) 2 , while aer ve days at 37 C, 30% of S(Bio-His-Tag) 2 converted into S(Bio-His-Tag) 1 and S(Bio-His-tag) 3 . Therefore, a single species prepared using this method is kinetically stable enough to be used in future studies.
In the second step, in order to produce precise streptavidin conjugates with two different functional groups, we used an iminobiotin conjugated His-tag (Ibio-His-Tag) instead of the Bio-His-Tag. The Ibio-His-Tag is removable from the streptavidin complex at a lower pH (pH 3.5) and can be used to reopen biotin binding pockets in isolated streptavidin conjugates with precise stoichiometry and a rst functionality. The reopened biotin binding pockets can then be used to introduce a second biotin conjugated molecule. In initial experiments, we noticed that the streptavidin-Ibio-His-Tag complexes had a lower affinity to the Ni 2+ -NTA column. A potential reason for this could be the different orientations of biotin and iminobiotin in the streptavidin binding pocket. While serine 27 of streptavidin acts as a H-bond donor for biotin, it is a H-bond acceptor for iminobiotin. 16 Consequently, the exposure of the connected His-tags could also be affected. To increase the affinity to the column and later separation of different species, the His6-tag was extended to a His12-tag, which increases the number of ligands that can bind to the column and the Ni 2+ -NTA column was replaced with a Cu 2+ -NTA column, as Cu 2+ has a higher binding affinity towards His-tags than Ni 2+ . 17 To produce the Ibio-His-tagged streptavidins, streptavidin (S ¼ 30 mM) was incubated with the Ibio-His-Tag (iminobiotin-(His) 12 ¼ 90 mM) for 15 minutes. Additionally, to simplify the procedure, the rst biotin conjugated molecule, atto-565-biotin (A ¼ 50 mM), was subsequently added to the reaction mixture and incubated for 15 minutes before the separation of different species (S(Ibio-His-Tag) 1 A 3 , S(Ibio-His-Tag) 2 A 2 , S(Ibio-His-Tag) 3 A 1 , S(Ibio-His-Tag) 4 ) over the Cu 2+ -NTA column using an imidazole gradient. In the chromatogram, four peaks were visible in the absorbance channel at 280 nm, but only the rst three absorbed at 563 nm, Fig. 3 Long time stability test at the absorbance wavelength of 280 nm. (a) S(Bio-His-Tag) 2 was stored at 4 C, room temperature (RT) and 37 C separately for one day, and then was loaded on the Ni 2+ -NTA column and washed with elution buffer. There is no change for samples stored at 4 C compared with the sample stored at room temperature, and there is a very little change at 37 C compared with the fresh sample (0 day). (b) There is no change for samples stored at 4 C and room temperature after five days, while 70% of S(Bio-His-Tag) 2 remained stable when stored at 37 C for five days, and 8.4% and 21.6% of S(Bio-His-Tag) 2 changed to S(Bio-His-Tag) 1 and S(Bio-His-Tag) 3 , respectively (black arrow).
where atto-565 absorbs (Fig. S4a, † relative integrated areas at 280 nm: 5.1% 1 st peak, 42.6% 2 nd peak, 26.7% 3 rd peak and 25.6% 4 th peak). To conrm the identity of the species in each peak, rst the Ibio-His-Tag was removed by lowering the pH to 3.5 and by dialysis, and subsequently the streptavidin species were titrated with biotin-5-uorescein as described above (Fig. S3b-e †). The streptavidin species in the rst peak reacted with 1 equivalent of biotin-5-uorescein and was therefore identied as SA 3 , which carried 1 Ibio-His-Tag before the acid-ication. Similarly, the species in the following peaks reacted with 2, 3 and 4 equivalents of biotin-5-uorescein, respectively, and were identied as SA 2 (two open biotin sites), SA 1 (three open biotin sites) and S (four open biotin sites). Moreover, this assignment was also supported by the relative absorbance of the conjugates in the UV-vis at 280 nm where streptavidin and atto-565 absorb and 532 nm where only atto-565 absorbs (Table S1 †). Using the Ibio-His-Tag, we were able to prepare uorescently labelled mono-(SA 3 ), di-(SA 2 ) and tri-(SA 1 ) valent streptavidins without genetic manipulation. Moreover, the atto-565-biotin used in this protocol can easily be replaced by another biotin conjugated molecule of choice. The straightforward preparation of these functionalized streptavidin derivatives with dened valences for the introduction of a second biotin coupled molecule offers us a new domain to expand and develop the many applications of biotin-streptavidin chemistry.
One area where uorescently labelled monovalent streptavidins are especially useful is in the imaging of biotinylated cell surface molecules. Unlike monovalent streptavidins, multivalent streptavidins lead to articial clustering of the biotinylated surface molecules, resulting in altered biological responses and cell uptake. Given this, the atto-565-biotin labelled monovalent streptavidin SA 3 described above is ideal for the visualization of biotinylated cell surface molecules. For this purpose, the cell membranes of MDA-MB-231 cells were rst randomly biotinylated using sulfo-NHS-LC-biotin 10 and subsequently the cell membranes were labeled with SA 3 , SA 2 or SA 1 , which was visible under the confocal microscope (Fig. 4, Movies S1-S3 †). However, the uorescence staining appeared different for different atto-565-biotin-streptavidin conjugates over time, as membrane proteins quickly internalize when they are crosslinked. 7,10 The cells labeled with monovalent SA 3 were stained at the membrane's periphery and even aer 30 minutes hardly any labeled protein was internalized. On the other hand, cells incubated with divalent SA 2 and trivalent SA 1 showed signicant protein internalization aer just 10 minutes, where this effect was more pronounced for SA 1 . The conclusions were also supported quantitatively based on the intracellular uorescence intensity (Fig. S4 †). The uorescently labelled monovalent streptavidin, SA 3 , can actually be used at concentrations as low as 0.05 mM for clearly cell surface labelling with high sensitivity (Fig. S5 and S6 †). Thus, it is a good alternative to current monovalent streptavidins, which require an additional labelling step with the uorophore and are only labelled statistically.  The open biotin binding pockets in the uorescently labelled streptavidins can also be used to introduce a second functionality with precise molecular stoichiometry. For instance, we added folic acid, an active and selective targeting molecule for aggressive cancer cells that overexpress the folic acid receptor on their surfaces. [19][20][21] The folic acid and atto-565 labelled streptavidins allowed us to determine whether the number of folic acid groups per molecule impacts cellular uptake by using the signal from the uorescent label. For this purpose, pure SA 3 , SA 2 and SA 1 were each incubated rst with excess folic acid-PEG-biotin (F) for 20 minutes and then excess F was removed by dialysis. The nal conjugates, SA 3 F 1 , SA 2 F 2 and SA 1 F 3 were tested on two different cell lines: folate receptor-positive MDA-MB-231 and folate receptor-negative MCF-7. Aer 4 hours of incubation the uorescence signal from the atto-565 was much brighter in the MDA-MB-231 cells than in the MCF-7 cells (Fig. 5a). The uptake was clearly due to the folic acid as cells incubated with SA 1 , SA 2 and SA 3 were not signicantly uorescent ( Fig. S7 and S8 †). For quantication, the uorescence intensities of SA 1 F 3 , SA 2 F 2 and SA 3 F 1 in both MDA-MB-231 and MCF-7 cells were measured and normalized to the relative brightness of the streptavidin-atto-565 species (Fig. 5b and Table S2 †). This analysis shows that all three streptavidin-folic acid conjugates were taken up equally well by the folic acid receptor positive cell line MDA-MB-231 independent of the number of folic acids in its structure.

Conclusions
In summary, we demonstrated that the Bio-His-Tag and Ibio-His-Tag can both be used to prepare multifunctional streptavidin-biotin conjugates with precise stoichiometry and structure. This method, as demonstrated by two examples, can be widely applied and is highly adaptable. It is an ideal approach to answer questions in molecular biology and for biotechnological applications. We used uorescently labelled monovalent (SA 3 ), divalent (SA 2 ) and trivalent (SA 1 ) streptavidin for imaging biotinylated cell surface molecules and investigated the importance of multivalent cell receptor interactions with folic acid in the cellular uptake and targeting (SA 3 F 1 , SA 2 F 2 and SA 1 F 3 ). The wide variety of commercially available biotinylated molecules ranging from small molecules and peptides to proteins, nucleic acids and antibodies, as well as the Bio-His-Tag and Ibio-His-Tag, all make this method extremely versatile and accessible. The potential diversity in precise streptavidin-biotin conjugates opens the door to building new bio-and nanostructures and will play a signicant role in expanding the well-established status of streptavidin-biotin chemistry in biotechnology. 22,23 Experimental Materials Streptavidin (MW 53361 g mol À1 ) was purchased from Cedarlane Laboratories. The Bio-His-Tag (biotin-(His) 6 , MW 1500 g mol À1 , sequence: biotin-GSGSGSHHHHHH) was synthesized by Peptide Specialty Laboratories GmbH and the Ibio-His-Tag (MW 2322 g mol À1 , iminobiotin-(His) 12 , sequence: iminobiotin-GSGSGSHHHHHHHHHHHH) was synthesized by Pepscan. Folic acid-PEG-biotin (MW 2000 g mol À1 ) was purchased from Nanocs. Sulfo-NHS-LC-biotin (MW 558 g mol À1 ) was purchased from AdooQ Bbioscience. BL21 (DE3) E. coli was purchased from New England Biolabs. pET Biotin His6 mOrange LIC cloning vector (H6-mOrange) was a gi from Scott Gradia (Addgene plasmid # 29723) and BirA in pET28a (w400-2) was a gi from Eric Campeau (Addgene plasmid # 26624). DMEM and RPMI-1640 without folic acid medium were purchased from Thermo Fisher Scientic. The Ni 2+ -NTA column (HisTrap™ HP, column volume 5 mL) was purchased from GE Healthcare Life Sciences. All other chemicals were purchased from Sigma-Aldrich. Buffers and aqueous solutions were prepared with Milli-Q grade water.

Separation of streptavidin Bio-His-Tag conjugates
30 mM streptavidin was mixed with 90 mM Bio-His-Tag (biotin-(His) 6 ) for 15 min at room temperature. A Ni 2+ -NTA column (HisTrap™ HP, column volume 5 mL) was pre-equilibrated with 50 mL buffer A (50 mM Tris-HCl, 300 mM NaCl, pH 7.4) using a FPLC system (GE healthcare, AKTA explorer). Then, 0.5 mL streptavidin Bio-His-Tag reaction mixture was loaded on the Ni 2+ -NTA column and the column was washed with 25 mL buffer A. Finally, the different streptavidin Bio-His-Tag conjugates were eluted with a linear imidazole gradient from 0 to 260 mM imidazole in buffer A over 130 mL. The elution of different species was monitored by the absorbance at 280 nm, the curves were corrected for the absorbance of imidazole and each peak was collected in a separate fraction. A ow rate of 0.5 mL min À1 was used throughout the experiment. The collected samples were dialyzed (10 kDa molecular weight cut-off) against 2 L of buffer A at 4 C for at least 6 h to remove imidazole. Subsequently, the samples were concentrated using a centrifugal ltration device (10 kDa molecular weight cut-off). The protein concentration was determined by UV-Vis spectroscopy.
To analyze the kinetic stability of S(Bio-His-Tag) 2 , different samples were incubated at 4 C, room temperature and at 37 C for one and ve days and analyzed on a Ni 2+ -NTA column using the same imidazole gradient as described above.

Separation of streptavidin Ibio-His-Tag conjugates
The Cu 2+ -NTA column was prepared by removing Ni 2+ ions from a Ni 2+ -NTA column (HisTrap™ HP, column volume 5 mL) with ethylenediaminetetraacetic acid (EDTA) and reloading it with Cu 2+ ions. Firstly, 30 mM streptavidin was mixed with 90 mM Ibio-His-Tag for 15 min at room temperature to ensure the binding of iminobiotin to streptavidin. Then, 50 mM atto-565biotin was added to the reaction mixture and incubated for another 15 min. 0.5 mL of the reaction mixture was loaded on a Cu 2+ -NTA column, which was pre-equilibrated with 50 mL buffer A and then washed with 25 mL buffer A. At last, the different streptavidin conjugates were eluted with a linear imidazole gradient from 0 to 20 mM imidazole in buffer A over 100 mL. The elution of different species was monitored by the absorbance at 280 nm (streptavidin and atto-565-biotin) and 563 nm (atto-565-biotin), the curves were corrected for the absorbance of imidazole and different peaks were collected separately. A ow rate of 0.5 mL min À1 was used throughout the experiment. Each peak was rst dialyzed against (10 kDa molecular weight cut-off) 2 L buffer pH ¼ 3.5, 50 mM Tris-HCl, 300 mM NaCl solution for 1 h at 4 C to remove the Ibio-His-Tag and then dialyzed twice against 2 L buffer A for at least 6 h to remove the imidazole. The samples were concentrated using a centrifugal ltration device (10 kDa molecular weight cut-off) for further studies. The protein concentration was determined by UV-Vis spectroscopy.

Determination of open biotin binding pockets
The number of open biotin binding pockets was determined using biotin-5-uorescein, whose uorescence is quenched upon binding to streptavidin. Typically, 200 mL of 10 nM of a streptavidin conjugate in buffer A was added in a transparent 96-well plate (Greiner bio-one, F-bottom), different concentrations (0 to 50 nM) of biotin-5-uorescein were added to each well and the samples were incubated for 10 min at room temperature. The uorescence intensity of each well was measured (excitation wavelength 490 nm, emission wavelength 524 nm) using a plate reader (TECAN, innite M1000).

Mass spectrometry
MALDI-TOF was used to test the molecular weight of the samples and performed on a Bruker Daltonics Reex III spectrometer. A saturated solution of sinapinic acid dissolved in 1 : 1 water : acetonitrile with 0.1% triuoroacetic acid was used as the matrix solution. Typically, a sample solution (10 mM) was mixed (1 : 1) with the matrix solution and spotted on the steel plate. Then an additional aliquot of the matrix solution was added to dilute the sample and spotted again (repeat 3Â). Spectra can be obtained in positive mode and the data were processed in mMass and Origin.

Preparation of biotinylated mOrange protein
Each protein expression plasmid (mOrange and BirA) was transformed into BL21(DE3) E. coli and starting from a single colony an overnight culture in 10 mL LB medium with 50 mg mL À1 kanamycin was prepared. The overnight culture was transferred into 1 L LB medium with 50 mg mL À1 kanamycin and incubated at 37 C, 250 rpm until the OD 600 ¼ 0.6-0.8 and then the protein expression was induced with 1 mM IPTG. Then, the cultures were incubated at 18 C, 250 rpm overnight and harvested the next day by centrifugation at 6000 rpm, 4 C for 8 min (Beckman Coulter Avanti J-26S XP, JA-10 rotor). The bacteria pellet was resuspended in 20 mL buffer A supplemented with 1 mM protease inhibitor phenylmethane sulfonyl uoride (PMSF) and 1 mM DL-dithiothreitol (DTT). The bacteria were lysed by sonication and the lysate was cleared by centrifugation at 12 000 rpm (Beckman Coulter Avanti J-26S XP, JA-25.50 rotor) for 30 min, followed by ltration through a 0.45 mm lter twice. The lysate was loaded onto a 5 mL Ni 2+ -NTA agarose column (HisTrap™ HP, column volume 5 mL). The column was washed with 50 mL buffer C (Buffer A with 25 mM imidazole and 1 mM DTT) and the protein was eluted with 10 mL buffer B (Buffer A with 250 mM imidazole and 1 mM DTT). The puried proteins were dialyzed against 2 L buffer A with 1 mM DTT twice for at least 6 h at 4 C.
mOrange contains a biotinylation sequence (GLNDI-FEAQKIEWHE) at its N-terminal, which is recognized by the enzyme BirA. 18 To prepare biotinylated mOrange, 50 mM mOrange, 5 mM MgCl 2 , 1 mM ATP, 1 mM BirA and 70 mM biotin were mixed in 1 mL buffer A and incubated at room temperature with gentle mixing (100 rpm) for 1 h. Then, additional 1 mM BirA and 70 mM biotin were added to the reaction mixture and incubated for 1 h. The reaction mixture was dialyzed (10 kDa molecular weight cut-off) against 2 L buffer A twice for at least 6 h at 4 C.

Preparation and analysis of mOrange streptavidin conjugates
Typically, 1 mM S(Bio-His-Tag) 1 , S(Bio-His-Tag) 2 or S(Bio-His-Tag) 3 was mixed with 5 mM, 4 mM or 3 mM biotinylated mOrange, respectively, for at least 1 h at 4 C. Then, 400 mL of each reaction mixture was injected onto a HiLoad™ 16/600, Superdex™ 200 pg size elution column and eluted with 150 mL buffer A at a ow rate of 1 mL min À1 . The elution of different species was monitored through the absorbance at 280 nm. The molecular weight of the different species was determined based on a calibration curve established with ve standard proteins (GE healthcare life science, 5 mg mL À1 thyroglobulin: MW 669 kDa, 0.3 mg mL À1 ferritin: MW 440 kDa, 4 mg mL À1 aldolase: MW 158 kDa, 3 mg mL À1 conalbumin: MW 75 kDa and 4 mg mL À1 ovalbumin: MW 44 kDa). For this, the partition coefficient (K av ) for each protein was calculated as follows: where V o is the column void volume, V e is the elution volume and V c is the geometric column volume, which is equal to 120 mL for the HiLoad™

Labeling of biotinylated cell surface molecules
MDA-MB-231 cells were seeded at 3 Â 10 4 cells per cm 2 in an 8well cell culture plate (glass bottom, ibidi) in 300 mL Dulbecco's modied Eagle's medium (DMEM) supplemented with 10% heat inactivated fetal bovine serum (FBS, 10%) and 1% penicillin/streptomycin (P/S) and incubated at 37 C, 5% CO 2 overnight. The next day, the cells were washed three times using cold phosphate buffer saline supplemented with 1 mM CaCl 2 and 0.1 mM MgCl 2 (PBS-CM) and then the cell membrane was biotinylated using 0.25 mM sulfo-NHS-LC-biotin in PBS-CM for 30 min on ice. 10 Aer washing the cells twice using cold PBS-CM, the cells were incubated with 100 mM glycine in PBS-CM for 2 min on ice to stop further biotinylation. The cells were washed again twice with cold PBS-CM and then incubated with 1.5 mM SA 1 , SA 2 or SA 3 in DMEM + 10% FBS + 1% P/S for 20 min on ice to label biotinylated surface molecules. The cells were washed twice with PBS, the medium was exchanged with prewarmed DMEM + 10% FBS + 1% P/S and the cells were incubated for 0, 10 and 30 min at 37 C in a 5% CO 2 atmosphere. Then the cells were washed twice with PBS and xed using 4% paraformaldehyde (PFA) in PBS and stained with 1 mg mL À1 TO-PRO-3. Finally, the samples were washed twice with PBS and imaged in the atto-565 and far-red channels using a confocal laser scanning microscope (Leica TCS SP8) equipped with 561 nm and 633 nm laser lines and a 63Â H 2 O objective. For live cell imaging, the cells were treated as described above and imaged in the atto-565 channel at room temperature aer adding 1.5 mM SA 1 (Movie S1 †), SA 2 (Movie S2 †) or SA 3 (Movie S3 †). Images were analysed by Fiji ImageJ. For the analysis, the average uorescence intensities of single cells in the atto-565 channel were measured by encircling single cells and measuring their average intensities. 25 cells were analysed per sample and uorescence was corrected for the background.
Preparation of uorescently labelled folic acid streptavidin conjugates and their cellular uptake 10 mM of SA 1 , SA 2 or SA 3 (A: atto-565-biotin) was mixed with 40 mM, 30 mM or 20 mM folic acid-PEG-biotin (F), respectively and incubated for 20 min at room temperature. Each reaction mixture was dialyzed (10 kDa molecular weight cut-off) against 2 L buffer A twice for at least 6 h yielding SA 1 F 3 , SA 2 F 2 and SA 3 F 1 .
The protein concentration of each conjugate and their relative uorescence were determined by UV-Vis and uorescence spectroscopy, respectively (Table S1 †). For cellular uptake studies, MDA-MB-231 or MCF-7 cells were seeded at 5 Â 10 4 cells per well on glass coverslips (VWR, diameter 18 mm) in 12-well cell culture plates (Greiner bio-one, F-bottom) and were cultured in RPMI-1640 medium without folic acid + 10% FBS + 1% P/S at 37 C, 5% CO 2 overnight. The next day, the cells were washed twice with PBS, 500 mL of RPMI-1640 medium without folic acid + 10% FBS + 1% P/S containing 1 mM of different streptavidin conjugates (SA 1 , SA 2 , SA 3 , SA 1 F 3 , SA 2 F 2 or SA 3 F 1 ) was added to each cell type and the cells were incubated at 37 C, 5% CO 2 for 4 h. Aerward, the cells were washed twice with PBS and xed with 4% PFA in PBS for 15 min at room temperature. The cells were washed twice with PBS and were mounted on a glass slide (ROTH, 24 Â 60 mm) with 40 mL Mowiol-488 containing 1 mg mL À1 DAPI. The cells were imaged in the atto-565 and DAPI channels using a confocal laser scanning microscope (Leica TCS SP8) equipped with 405 nm and 552 nm laser lines and a 63Â H 2 O objective. Images were analysed by Fiji ImageJ. For the analysis, the average uorescence intensities of single cells in the atto-565 channel were measured by encircling single cells and measuring their average intensities. 25 cells were analyzed per sample and uorescence was corrected for the background. Then, uorescence intensities measured for SA 1 F 3 , SA 2 F 2 and SA 3 F 1 in the cells were normalized taking into account the relative uorescence brightness of SA 1 F 3 , SA 2 F 2 and SA 3 F 1 as measured in solution (Table S2 †).

Conflicts of interest
There are no conicts to declare.