A rationally designed and highly versatile epitope tag for nanobody-based purification, detection and manipulation of proteins

Specialized epitope tags are widely used for detecting, manipulating or purifying proteins, but often their versatility is limited. Here, we introduce the ALFA-tag, a novel, rationally designed epitope tag that serves an exceptionally broad spectrum of applications in life sciences while outperforming established tags like the HA, FLAG or myc tags. The ALFA-tag forms a small and stable α-helix that is functional irrespective of its position on the target protein in prokaryotic and eukaryotic hosts. We developed a nanobody (NbALFA) binding ALFA-tagged proteins from native or fixed specimen with low picomolar affinity. It is ideally suited for super-resolution microscopy, immunoprecipitations and Western blotting, and also allows in-vivo detection of proteins. By solving the crystal structure of the complex we were able to design a nanobody mutant (NbALFAPE) that permits efficient one-step purifications of native ALFA-tagged proteins, complexes and even entire living cells using peptide elution under physiological conditions.


Introduction
Epitope tags are employed in virtually any aspect of life sciences 1,2 . They are used in biotechnology to facilitate the expression and purification of recombinant proteins 1 or in cell biology to monitor the biogenesis or spatial organization of a given protein of interest (POI) 3,4 . Other uses include, e.g., the immunoprecipitation of protein complexes studied by mass spectrometry 5,6 , or protein manipulations using tag-binding reagents in living cells 7,8 .
While a given tag might be ideal for a specific application, it may completely fail in others. This is a result of how the tags have been generated -typically as byproducts while developing antibodies against specific POIs (for example the myc-tag 9 , the HA-tag 10 or the Spot-tag ®11,12 ). Other tags, like the His-tag 13 , have been rationally designed for a specific application. None of the available tags covers the full range of current biological applications (see Table S1 for details). For example, the His-tag provides poor results in immunostaining and imaging applications, albeit it is excellent for protein purification. The FLAG, myc and HA-tags have often been used for immunostainings, but due to the large size of the antibodies used as binders, they are suboptimal for super-resolution microscopy and exceedingly difficult to express within cells. These tags therefore cannot be used to reveal or manipulate POIs in living cells. Similar considerations apply for the Twin-Strep tag, which is in addition fixation-sensitive, and therefore not suitable for immunostainings.
More recently the EPEA tag 14 (also known as C-Tag) and the Spot-tag ®11, 12 have been identified as tags recognized by camelid single-domain antibodies (sdAbs, also known as nanobodies 15 ). In contrast to conventional antibodies, sdAbs are monomeric, small, show a superior performance in super-resolution microscopy 16,17 and can , in principle, even be used for intracellular applications 7,8 . Unfortunately, both of these systems have several problems that override the potential advantages given by their nanobody binders. For instance, both nanobodies detect endogenous proteins (α-synuclein and β-catenin, respectively) and display comparably poor affinities when used as monovalent binders (Table S1) implying that they are suboptimal for advanced microscopy applications or pull-downs of low abundant proteins.
Additionally, both tags have so far not been reported to work in living cells, and the EPEA tag can only be used at the C-terminus of target proteins.
To overcome the limitations of the available epitope tags, a first step is to define a set of desired features. A truly versatile tag should not affect the structure, topology, localization, solubility, oligomerization status or polar interactions of the tagged protein 18,19 . It should therefore be small, monomeric, highly soluble 20 and electroneutral. For highest versatility, it should in addition be resistant to chemical fixation. To avoid specific background signals, the optimal tag should be unique in eukaryotic and prokaryotic hosts, while being well expressed and protease resistant. Similar, the molecule binding such tag should fulfill certain characteristics. It should not only be small for an optimal access to crowded regions and have a minimal linkage error in super-resolution microscopy 21,22 , but also specifically bind the tag with high affinity. For in-vivo imaging and manipulations, the binder should be genetically accessible and fold properly within various host organisms. For biochemical applications, the preferred binder should allow for both, stringent washing and native elution of immunoprecipitated proteins or complexes. Strikingly, currently existing epitope tag systems fail to fulfill the complete set of mentioned criteria (Table S1). To manufacture an epitope tag system with the ultimate versatility, the only way is to design it de novo.
With this clear objective we created the ALFA-tag. This 15 amino acids sequence is hydrophilic, uncharged at physiological pH and devoid of residues targeted by amine-reactive fixatives or cross-linkers. It has a high propensity to form a stable α-helix that spontaneously refolds even after exposure to harsh chemical treatment. Due to its compact structure, the tag is physically smaller than most linear epitope tags. The ALFA-tag is compatible with protein function and can be placed at the N-or C-terminus of a POI, or even in between two separately folded domains.
As a counterpart binding the ALFA-tag, we developed a high-affinity nanobody (NbALFA), which proved to be suitable for super-resolution imaging and intracellular detection of ALFA-tagged proteins and allowed very efficient and clean immunoprecipitations and Western blots. However, it was virtually impossible to separate NbALFA from the ALFA-tag under native conditions, hindering its application for the purifications of native protein complexes, organelles or cells. Based on the crystal structure of the NbALFA-ALFA peptide complex we engineered a nanobody variant (NbALFA PE ; "Peptide-Elution") suitable for highly specific purification of ALFA-tagged proteins, protein complexes and even living cells under physiological conditions. The rationally designed ALFA system presented here serves an exceptionally broad range of applications from biotechnology to cell biology. A single tag can therefore replace a great variety of traditional epitope tags.

The ALFA-system
The ALFA-tag sequence (Fig.1a) is inspired by an artificial peptide reported to form a stable α-helix in solution 23 . It was selected based on the following properties: i) It features a high alpha-helical content, ii) The sequence is absent in common eukaryotic model systems, iii) It is hydrophilic and neutral at physiological pH while retaining moderate hydrophobic surfaces and iv) It does not contain any primary amines that are modified by aldehydecontaining fixatives. In order to develop nanobodies binding the ALFA-tag, we immunized alpacas and selected nanobodies specifically targeting ALFA-tagged proteins by a novel nanobody selection method (see Online Methods). The best binder (NbALFA; Fig.1b) was genetically modified with ectopic cysteines allowing for a site-specific immobilization or fluorophore attachment 24 .

Detection of ALFA-tagged proteins by direct immunofluorescence
In immunofluorescence (IF) applications on PFA-fixed mammalian cells, fluorescently labeled NbALFA specifically recognized target proteins harboring ALFA tags at different locations within the proteins (Fig.S1). Importantly, all tested proteins showed their characteristic localization (Tom70-EGFP-ALFA: mitochondrial outer membrane; ALFA-Vimentin: filamentous structures; EGFP-ALFA-TM: plasma membrane), the ALFA-tag was therefore compatible with the folding and proper localization of the tagged proteins. In a quantitative assay, the nucleocytoplasmic distribution of EGFP carrying an ALFA tag at either terminus was indistinguishable from non-tagged EGFP (Fig.S2). An atypical interaction to cellular membranes or organelles was not observed. Moreover, gel filtration of a recombinant ALFA-tagged mEGFP variant confirmed its monomeric state indicating that the ALFA-tag does not induce multimerization (data not shown). We therefore concluded that the ALFA-tag seems not to impair the physiological behavior of the tagged POIs. Sequence of NbALFA. Grey boxes indicate CDRs 1-3 (AbM definition). c; COS-7 cells transfected with Tom70-EGFP-ALFA were fixed with paraformaldehyde (PFA) and stained with NbALFA coupled to AbberiorStar635P (NbALFA-Ab635P). Left to right: NbALFA-Ab635P; intrinsic EGFP signal; overlay incl. DAPI stain; sketch illustrating the detection of Tom70-EGFP-ALFA. d; ALFA-Vimentin was detected with NbALFA-Ab635P after fixation with 4% PFA, 2% glutaraldehyde (GA), or 100% Methanol (MeOH). e, STED and confocal images of COS-7 cell transiently transfected with ALFA-Vimentin and stained with NbALFA-Ab635P. f; HeLa cells transfected with ALFA-vimentin were stained with NbALFA bearing a 10-nucleotide single stranded DNA before imaging by 3D DNA-PAINT. The histogram refers to a region (small yellow rectangle) where 2 vimentin filaments are resolved although being only ~90nm apart. The localization precision was 5.2nm. g; COS-7 cells were co-transfected with an NbALFA-mScarlet-I fusion and ALFA-Vimentin. NbALFA-mScarlet-I co-localizes with ALFA-Vimentin detected by immunofluorescence using NbALFA-Ab635P. This shows that NbALFA expressed in the cytoplasm of mammalian cells can be used for targeting ALFAtagged proteins in living cells.

Resistance to amine-reactive fixatives
In order to facilitate optimization of fixation conditions, it is advantageous if a given epitope tag is compatible with various fixation procedures. In contrast to many established epitope tags (Table S1), the ALFA-tag does not contain lysines by design. Consequently, it could be detected after standard fixation with paraformaldehyde or methanol, and was even resistant to 2% glutaraldehyde (Fig.1d). The ALFA-tags is thus compatible with standard fixation methods and has the potential to be employed in electron microscopy, where glutaraldehyde is preferred due to its superior preservation of cellular structures.

Super-resolution microscopy
Due to the small size of both the ALFA-tag and NbALFA, the ALFA system results in a minimal linkage-error and it thus perfectly suited for super-resolution microscopy. As examples, we imaged cells transfected with ALFA-Vimentin by either STED microscopy 16 or 3D DNA-PAINT 25 using NbALFA directly coupled to a "STEDable" fluorophore or a short single-stranded oligonucleotide, respectively (Fig.1e,f). Our results show that the ALFA system is compatible with the demanding conditions of these fluorescent super-resolution microscopy techniques.

Detecting and manipulating ALFA-tagged proteins in vivo
Some nanobodies are functional within mammalian cells and can thus be used to detect or manipulate target proteins in living cells 8,26 . Indeed, when co-expressing ALFA-tagged target proteins and NbALFA fused to mScarlet-I 27 in mammalian cells, NbALFA-mScarlet-I robustly co-localized with ALFA-tagged target proteins with minimal off-target signals, resulting in crisp detection of ALFA-tagged structures (Fig.1g, Fig.S3). This demonstrates the ability of NbALFA to bind ALFA tagged proteins in living cells. This feature is very attractive to be used in combination with genome editing tools like CRISPR-Cas 28 and allows manipulation of ALFA-tagged proteins in vivo at their endogenous levels.

Western Blot
We next tested the ALFA system in Western blots applications and could specifically detect ALFA-tagged Vimentin in lysates from transfected cells using NbALFA labeled with IRDye800CW ( Fig.2a, Fig.S4a). To compare the performance of the ALFA system with common epitope tags, we employed a fusion protein harboring HA, myc, FLAG ® and ALFA tags (Fig.2b). At identical concentrations of primary antibody or nanobody, the obtained ALFA-tag signal was 3-10-fold stronger than the signal obtained for all other epitope tags ( Fig.2c, Fig.S4b). This is striking since the ALFA-tag signal exclusively relied on the NbALFA fluorophores, while the signal of all other epitope tags was amplified using polyclonal secondary antibodies. Without further optimization, NbALFA yielded a remarkably linear signal over at least three orders of magnitude ( Fig.2d) and was able to detect as little as 100pg of target protein. The detection limit was thus ~10-times better than observed for all other epitope tags. Established monoclonal antibodies (M2, 9E10 and F-7) were used together with an antimouse IgG coupled IRDye800CW to detect the FLAG ® , myc and HA-tags respectively. The ALFA-tag was detected using NbALFA coupled to IRDye800CW. Fig.S4b shows the complete experiment including internal controls. d; Double-logarithmic plot showing quantification of signals obtained in c. Lines represent linear fits to the obtained values. Even without signal amplification by a secondary antibody, signals obtained using NbALFA were 3-to >10-times stronger than by established reagents recognizing the other epitope tags. At the same time, detection with NbALFA was 10-fold more sensitive and showed an excellent linearity over ~3 orders of magnitude. . NbALFA: orange with CDRs 1-3 colored in yellow. ALFA peptide in blue. For both molecules, the N-terminal is oriented left, the C-terminal right. b; Polar interactions within the Nterminal region of the ALFA peptide. ALFA peptide residues are denoted by an apostrophe. S2' and E5' form hydrogen bonds with CDR2 (S57, E58, R59 and N61). R3 ' reaches out to CDR3 and interacts with the backbone of V107 and the side chain of D105. c; R3' and E7' sandwich F110 on the nanobody, forming a cation-Pi interaction (reviewed in 29 ); Illustration of a hydrophobic cluster (L4', L8' and L12') facing the nanobody's hydrophobic cavity. e; Polar interaction near the C-terminal of the ALFA peptide. The backbone of E14' forms hydrogen bonds with R65 while the side chain of R11' interacts with D112 and Y42 on the five-stranded b-sheet. Interestingly, Y42 has been described as a conserved residue in nanobodies of this particular architecture 8 .

Structure of the NbALFA bound to the ALFA peptide
To better understand the interaction between NbALFA and the ALFA tag, we solved the crystal structure of NbALFA in complex with the ALFA peptide ( Fig.3, Table S2). NbALFA adopts a canonical IgG fold 30 comprising two b-sheets connected by a disulfide bond. The paratope accepting the ALFA peptide extends from the nanobody's N-terminal cap to the side formed by the five-stranded b-sheet. It is mainly built from complementarity determining regions (CDRs 15 ) 2 and 3 and, interestingly, also involves residues within the five-stranded bsheet forming a hydrophobic cavity. As a result, the ALFA peptide is oriented parallel to the central axis of NbALFA. The ALFA peptide forms an α-helical cylinder with a length of ~2nm and a diameter of ~1.3nm, that is stabilized by a complex network of intramolecular interactions (Table S3). In addition, the peptide forms multiple polar and hydrophobic contacts with NbALFA ( Fig.3b-e, Table S4).

Capture of ALFA-tagged target proteins using ALFA Selector resins
For biochemical purifications, we site-specifically immobilized NbALFA on an agarosebased resin (Fig.4a). The resulting resin efficiently and strongly bound an ALFA-tagged GFP variant (shGFP2 31 ): Even after 1h competition with an excess of free ALFA peptide at 25°C, most of target protein remained on the resin (Fig.4b, red solid line). In line with these observations, SPR assays indicated that the affinity of NbALFA to an ALFA-tagged target was ~26 pM (Fig.S5). We therefore called the NbALFA-charged resin "ALFA Selector ST " (for Super-Tight). To allow for an efficient competitive elution of ALFA-tagged target proteins, we intended to weaken NbALFA. For that, we followed a rational approach based on the NbALFA-ALFA peptide structure: NbALFA variants featuring single or combined amino acid exchanges in spatial proximity to the ALFA peptide were individually tested for their binding and elution properties. The successfully weakened binder NbALFA PE (for Peptide Elution) carries several mutations with respect to NbALFA, thereby removing specific interactions of NbALFA with the ALFA peptide. As a consequence, the affinity of NbALFA PE to a fusion protein harboring a C-terminal ALFA-tag was reduced to ~11 nM ( Fig.S5). As intended, a NbALFA PE -charged resin (ALFA Selector PE ) efficiently and stably captured shGFP2-ALFA while allowing an efficient release within ~15-20 minutes by competition with free ALFA peptide (Fig.4b,c). Similar elution kinetics were found when the ALFA-tag was placed between two folded domains, while elution of ALFA-shGFP2 from ALFA Selector PE was slightly quicker (Fig.S6). Remarkably, in the absence of competing peptide, spontaneous elution of all target proteins from both ALFA Selector variants was insignificant (Fig.4b, c and Fig.S6). To estimate off-rates, the resins were incubated with an excess of free ALFA peptide at 25°C. Control reactions were carried out without peptide. shGFP2 released from the resin was quantified using its fluorescence. The graph (b) shows mean fluorescence readings as well as standard deviations (n=3). Lines represent fits to a single exponential. A photo was taken upon UV illumination after 3h of elution (c). d; Resistance towards stringent washing steps. Both ALFA Selector variants were charged with either ALFA-shGFP2 or shGFP2-ALFA and incubated with a 10-fold volume of the indicated substances for 1h at 25°C with shaking. Without further washing steps, photos were taken upon UV illumination after sedimentation of the beads. e; Resistance towards various pH: Similar to d, here, however, the resins were washed to remove non-bound material after incubating at indicated pH for 30min. Photos were taken after re-equilibration in PBS to allow for recovery of the GFP fluorescence.

Stringent washing and pH resistance
We next performed stringent washing steps on both ALFA Selectors variants bound to either ALFA-shGFP2 or shGFP2-ALFA (Fig.4d). At 25°C, all substrate-resin interactions were completely resistant to significant concentration of salt, Guanidinium-HCl, nondenaturing detergents or reducing agents. A small fraction of substrate was released from ALFA Selector PE , but not from ALFA Selector ST , upon treatment with up to 6M urea.
In a similar assay, the loaded ALFA Selector resins were exposed to buffers adjusted to various pH (Fig.4e). The interaction was stable at pH7.5 or 9.5, and only slightly affected at pH4.5. However, even after neutralization, both ALFA Selector resins remained completely non-fluorescent when washed with 100mM Glycin at pH2.2. The eluted material, in contrast, successfully recovered its fluorescence at neutral pH (not shown), indicating that acidic elution with Glycin at pH2.2 is possible even from ALFA Selector ST .

Pull-down of ALFA-tagged target proteins from complex lysates
To address the specificity of the ALFA Selector resins in native pull-downs, we spiked E.
coli or HeLa lysates prepared in PBS with recombinant ALFA-shGFP2 (Fig.5a, left lane). The target protein specifically bound to both ALFA Selector variants but not to a control resin without coupled nanobody. As expected, ALFA-shGFP2 efficiently eluted from ALFA Selector PE using 200µM of ALFA peptide in PBS. In contrast, significant elution from ALFA Selector ST was only observed upon treatment with SDS sample buffer. Importantly, pulldowns from both, E. coli (Fig.5a) and HeLa lysates (Fig.5b) were highly specific: After peptide elution from ALFA Selector PE essentially all visible bands could be attributed to the input protein, and even in the SDS eluate, the number and strength of detectable impurities originating from lysate proteins were very low.
In a more delicate co-immunoprecipitation experiment, we tried to pull-down the E. coli YfgM-PpiD inner membrane protein complex 32 under native conditions (Fig.5c). For this, either wild-type YfgM or YfgM-ALFA was expressed in a ΔyfgM strain under the control of its endogenous promoter 32 . ALFA Selector PE was able to specific pull-down the YfgM-PpiD complex in a detergent-resistant manner from the lysate containing YfgM-ALFA. The native membrane protein complex could be recovered from ALFA Selector PE by peptide elution under physiological conditions showing that the ALFA-tag together with ALFA Selector PE resin is perfectly suited for native pull-downs of challenging (membrane) protein complexes. were incubated with ALFA Selector ST , ALFA Selector PE or an analogous resin without immobilized sdAb (Selector Control). After washing with PBS, the resins were incubated with 200µM ALFA peptide for 20min before eluting remaining proteins with SDS sample buffer. Indicated fractions were analyzed by SDS-PAGE and Coomassie staining. Eluate fractions correspond to the material eluted from 1µL of resin. c; Native pull-down of an E. coli inner membrane protein complex. Left: Sketch of the target protein complex. Right: Detergent-treated lysates from a ΔyfgM strain complemented with either C-terminally ALFA-tagged (left panel) or untagged YfgM (right) were incubated with ALFA Selector PE . After washing with PBS, bound proteins were eluted using 200µM ALFA peptide. Samples corresponding to 1/800 of the input and non-bound material or 1/80 of eluate fractions were resolved by SDS page and analyzed by Western blot. ALFA Selector PE specifically immunoprecipitated the native protein complex comprising ALFA-tagged YfgM and its interaction partner PpiD. In the control reaction (no ALFA tag on YfgM), both proteins were absent in the eluate. Complete blots in Fig.S7.

Isolation of live lymphocytes
An envisioned application for the ALFA Selector PE is the specific enrichment of cells under physiological conditions. This may be particularly interesting e.g. for the generation of chimeric antigen receptor-modified T (CAR-T) cells, the precursors of which are usually obtained from blood 33 . To investigate if the ALFA system can be applied to enrich live blood cells, human peripheral blood mononuclear cells (PBMCs) were passed through an ALFA Selector PE column pre-charged with an ALFA-tagged nanobody recognizing CD62L, a surface marker typically present on naïve T cells 34 (Fig.6a). After washing, bound cells were eluted using ALFA peptide, stained with antibodies recognizing CD62L, the pan T cell marker CD3 and the pan B cell marker CD19, and analyzed by FACS (Fig.6). Total PBMCs served as a control. Using this strategy, CD62L + lymphocytes were enriched from 71.8 to 97.7% (Fig.6b). In addition, we confirmed that the vast majority of ALFA peptide-eluted cells were CD3-positive T cells, while B cells represented a minor population of the isolated cells ( Fig.6c).

Discussion
In the current study, we report the characterization of the ALFA system comprising the ALFA-tag, a novel and highly versatile epitope tag, and two dedicated nanobodies. The mainly rational approach allowed us to equip the ALFA system with favorable features for a broad spectrum of applications. When selecting the ALFA-tag sequence, we not only made sure that the tag is small, monovalent, devoid of lysines, hydrophilic without carrying any net charge and absent within the proteome of relevant model organisms, but also that it would adopt a stable α-helical structure in solution. As a result, the ALFA-tag is by design highly specific (Fig.5), insensitive to amine-reactive fixatives (Fig.1c) and well tolerated by the tagged target proteins (Figs.1, S1-S3). The ALFA-tag could thus even be used for purification of a labile membrane protein complex (Fig.5c). However, as for any other tag, specific effects on given target proteins have to be analyzed on a protein-to-protein basis.
The ALFA tag with its comprehensive feature set contrasts to existing tags, which mostly fail to fulfill one or more parameters (Table S1). For instance, large structured tags like GFP can affect the tagged protein localization and function 4,18 . On the other hand, intrinsically disordered small tags can adopt multiple conformations with unpredictable effects. The HAtag, e.g., can be cleaved by caspases in mammalian cells 35 . In the worst case, tags may even target the POI for protein degradation 36 . Aldehyde-containing fixatives used for light and electron microscopy modify lysine-containing epitope tags like the myc-tag, which may impair antibody-based detection. Other tags (e.g. the FLAG ® -tag) carry significant net charges and may thus affect the POI's physiological electrostatic interactions. Tags like myctag 9 , the HA-tag 10 , the Spot-tag ®11,12 , C-tag 14 and the Inntags 37 are derived from proteins present in model organisms. Therefore, the respective binders can also recognize the endogenous host proteins by default. In contrast to the C-tag 14 that only works at a POI's Cterminal, the ALFA-tag works at all accessible locations within the target protein without interfering with the protein's localization or topology (Figs.S1-S3).
For binding the ALFA-tag we developed nanobodies 15 , because they are small, monovalent and robust, can easily be modified by genetic means and recombinantly produced in various organisms. They can therefore be site-specifically immobilized or quantitatively modified with fluorophores 24 or oligonucleotides. In comparison to conventional antibodies, nanobodies are thus superior for conventional and advanced microscopy 11,17 (Fig.1e,f) since they can find more target epitopes in crowded areas 22 , localize the fluorophore closer to the target protein and avoid artificial clustering of the POI 16,21 . Interestingly, NbALFA folds even within eukaryotic cells (Fig.1g) and can thus be used as an intrabody 7 for detecting or manipulating ALFA-tagged proteins in vivo 7,8,26 .
Nanobodies often fail to detect proteins in Western blot applications as they typically recognize three-dimensional epitopes. NbALFA, however, enables highly sensitive Western blots, suggesting that the ALFA-tag's α-helical structure refolds efficiently after SDS removal. Despite its monovalent binding mode and without further signal amplification, NbALFA significantly outperformed established anti-epitope tag tools regarding absolute signal intensity and detection limit (Fig.2). We expect a superior performance also in other high sensitivity applications like ELISA or microarray assays. Due to its resistance to aminereactive fixatives (Fig.1d), we even envision applying the ALFA system for immuno-electron microscopy and nanoSIMS 38 .
The binding of NbALFA to the ALFA-tag is exceptionally strong, which could be explained after solving the crystal structure of the NbALFA-ALFA peptide complex (Fig.3).
First, the peptide binds to NbALFA in a α-helical conformation and has an unusually high propensity to form a stable α-helix also in solution 23 . This may be explained by multiple intramolecular side-chain interactions within the peptide (Table S3). As a result, no energy needs to be spent in forming the required structure during complex formation, which would otherwise disfavor complex formation. Second, the peptide forms multiple specific contacts with the nanobody (Table S4) along the whole length of the 13 amino acid ALFA core sequence (SRLEEELRRRLTE). In sum, nearly all residues of this sequence are involved in binding to NbALFA and/or stabilizing the α-helical peptide conformation. Due to the compact structure of the complex, the maximal displacement between an ALFA-tagged target and a fluorophore attached to NbALFA is well below 5nm and can be reduced to <3nm by choosing adequate positions for fluorescent labeling. In order to minimize the potential influences of neighboring sequences on the ALFA-tag conformation, we placed the ALFA core sequence between prolines acting as "insulators". Using this approach, the interaction of NbALFA with the ALFA-tag is largely independent of its localization within the tagged protein and is efficiently recognized when placed at either terminus of the POI or even in between two separately-folded domains.
The strong interaction of NbALFA and the ALFA-tag is ideal for imaging applications, highly sensitive detection of target proteins, and purification of extremely low-abundant proteins from dilute lysates or under conditions where harsh washings with chaotropic agents are required. The slow dissociation, however, precludes a competitive elution of ALFA-tagged proteins under physiological conditions within a reasonable time frame. Based on the crystal structure of the NbALFA-ALFA peptide complex (Fig.3), we site-specifically mutated the nanobody to increase its off-rate. A resin displaying the mutant nanobody (ALFA Selector PE ) allows for native purifications of proteins and protein complexes from various lysates under physiological conditions by peptide elution (Figs.4, 5). ALFA Selector PE could even be applied for the selective enrichment of CD62L-positive lymphocytes from PBMC preparations (Fig.6). We believe that this technique can easily be transferred to the highly validated recombinant Fab and scFv fragments that are currently used for cell isolation approaches and similar purposes 39 , or to novel nanobodies recognizing surface markers that can easily be equipped with an ALFA tag. Our new technology can therefore contribute to current advances in biomedical research and therapy including the CAR-T technology 33 .
The novel ALFA tag system stands out by its exceptionally broad applicability. Using the ALFA system, a single transgenic cell line or organism harboring an ALFA-tagged target protein is sufficient for a wealth of different applications including (super-resolution) imaging, in-vivo manipulation of proteins, in-vitro detection by Western blot or even native pull-down applications aiming at detecting specific interaction partners or at isolating specific cell populations. The wide range of applications of the ALFA system provides the scientific community with a novel and highly versatile tool, which will facilitate future scientific breakthroughs.

Materials & Correspondence
For plasmid requests please write to materials@nano-tag.com

Data Availability
The atomic coordinates and structure factors (code 6I2G) have been deposited in the Protein Data Bank (http://www.pdb.org). Primary data of graphs shown in Figures 4b, S2, S5 and S6 are available in the Source Data file.  imidazole/HCl pH7.5 and 10mM DTT, and purified by binding to Ni(II)-chelate beads. After extensive washing, proteins were eluted by on-column-cleavage with bdSENP1 as described before 40,41 . ALFA-shGFP2-His6 and His14-bdSUMO-ALFA-shsfGFP were expressed and purified in a similar fashion; Elution was, however, performed using 250mM Imidazole in buffer LS. For affinity determinations and binding studies from complex lysates, substrate proteins were additionally purified via size exclusion chromatography on a Superdex200 10/30 column (GE Healthcare). NbALFA harboring N-and C-terminal ectopic cysteines was expressed and purified as described for non-tagged NbALFA above. Labeling with maleimide-activated fluorophores was performed as previously described elsewhere 24 .

Selection of specific sdAb clones by affinity purification of B lymphocytes
1mL of T-Catch resin (IBA Lifesciences) was washed with B cell isolation buffer (PBS pH7.4, 1% BSA, 1mM EDTA) and incubated with saturating amounts of a TwinStrepTag-bdNEDDD8-ALFA fusion protein for 30min rolling at RT. The resins were cleared from excess bait protein by extensively washing with B cell isolation buffer. 100mL of blood sample was taken from alpaca immunized with ALFA peptide and immediately incubated with 5000 IU/mL heparin (Sigma) to prevent clotting. From the fresh blood (less than 4h past sampling) PBMCs were isolated using Ficoll-Paque PLUS (GE Healthcare). To remove residual serum, PBMCs were washed three times consecutively with B cell isolation buffer.
PBMCs were passed over the loaded T-Catch resin three times before washing the resins with Glycine pH 2.2, 150mM NaCl for 2.5min. To accurately determine the dissociation constant of NbALFA, the decrease in response was followed for >7h.

Transfection of 3T3, COS-7 or HeLa cells
For immunofluorescence experiments, 3T3, COS-7 or Hela (Leibniz Institute DSMZ) cells were transiently transfected with 0.2 to 1µg of plasmids listed in DNA was premixed in a 1:1 ratio and further processed as described above.

DNA-labeling of NbALFA, DNA-PAINT imaging and data analysis
NbALFA was first coupled to a single stranded DNA as described before 48   For DNA-PAINT imaging, transfected cells were screened for a certain phenotype with 488nm laser excitation at 0.01kW/cm 2 . After acquisition of the 488nm channel, the excitation was switched to 561nm, focal plane and TIRF angle were readjusted and imaging was subsequently performed using ~2.5kW/cm 2 561nm laser excitation. P3-imager strand (5'-GTAATGAAGA-Cy3b-3') concentration was chosen to minimize double-binding events.
ALFA-tag imaging was performed using an imager concentration of 1nM P3-Cy3b in Buffer C (PBS + 500mM NaCl). All imaging was performed in 1´PCA (Sigma-Aldrich, #37580)/1´PCD (Sigma-Aldrich, #P8279)/1´Trolox (Sigma-Aldrich, #238813) in Buffer C and cells were imaged for 40,000 frames at 100ms camera exposure time. 3D imaging was performed using a cylindrical lens in the detection path as previously reported 49 . Raw data movies were reconstructed with the Picasso software suite 25 . Drift correction was performed with a redundant cross-correlation and gold particles as fiducials. Final images obtained had a localization precision of <10nm calculated via a nearest neighbor analysis 50 . Rendering was performed via the recently published SMAP software.

Impact of ALFA-tags on the localization of EGFP
Transiently transfected 3T3 cells were imaged using an epifluorescence microscope (Axio, Zeiss) equipped with a 40x 1.3 oil lens. For cells transfected with either pCMV ALFA-EGFP, pCMV EGFP-ALFA, or pEGFP-N1, 107-133 cells were imaged in a total of six to seven individual images. For each individual image, cells were grouped and counted according to the localization of EGFP ("slightly nuclear", "equally distributed", "other"). The fraction of cells in each group was statistically analyzed using Student's t-test.

Dotblot assay
A serial dilution of MBP fused to FLAG ® -, HA-, myc-and ALFA-tags was prepared in PBS  Nikon D700 camera equipped with a 105mm macro lens (Nikon).

Resistance towards stringent washing and pH
15µL of ALFA Selector ST or ALFA Selector PE saturated with either ALFA-shGFP2 or shGFP2-ALFA were washed with PBS and incubated with 100µL of the indicated substances for 60min at RT. Photos were taken after sedimentation of the beads upon UV illumination.

One-step affinity purifications from E. coli and HeLa lysates.
To obtain defined input materials for pull-down experiments from E. coli or HeLa lysates, respective mock lysates were blended with 3µM of purified ALFA-tagged shGFP2. 1mL of each lysate/substrate mixture was incubated with 25µL of ALFA Selector ST or ALFA Selector PE for 1h at 4°C. An analogous resin without immobilized sdAb (Selector Control) served as a specificity control. After washing 3 times with 600µL of PBS, the resins were transferred into MiniSpin columns (NanoTag Biotechnologies). Excess buffer was removed by centrifugation (3000x g, 30s) before incubating twice for 10min at RT with 50µL each of 200µM ALFA peptide in PBS. Proteins remaining on the beads were afterwards eluted with SDS sample buffer. 0.5µL (E. coli) or 1.5µL (HeLa) of input and non-bound fractions were resolved by SDS-PAGE (12%) and Coomassie staining. Shown eluate fractions correspond to the material eluted from 1µL of the respective resins.
Native pull-down of the E. coli YfgM-PpiD inner membrane protein complex.
A yfgM deletion strain was complemented with either C-terminally ALFA-tagged or untagged YfgM expressed from a pSC-based low-copy vector under control of the endogenous promoter. Membrane protein complexes were solubilized from total lysates prepared in buffer LS (50mM Tris pH7.5, 300mM NaCl, 5mM MgCl2) using 1% DDM for 1h on ice 51 . Both lysates were incubated with 20µL of ALFA Selector PE resin for 1h at 4°C on a roller drum.
After washing in PBS containing 0.3% DDM, bound proteins were eluted under native conditions by sequentially incubating twice with 50µL PBS containing 200µM ALFA peptide. Samples corresponding to 1/800 of the input and non-bound material or 1/80 of eluate fractions were resolved by SDS-PAGE. Analysis was performed by Western blotting using a polyclonal rabbit antiserum raised against the YfgM-PpiD complex 32 followed by an HRP-conjugated goat anti-rabbit IgG (Dianova). Blots were developed using the Western Lightning Plus-ECL Kit (Perkin Elmer) and imaged using a LAS 4000 mini luminescence imager (Fuji Film).

Preparation of human PBMCs
Human peripheral blood mononuclear cells (PBMCs) were obtained from fresh blood using standard density gradient centrifugation. Briefly, 60mL of fresh blood was diluted with 40mL of phosphate-buffered saline (PBS) supplemented with 1mM EDTA and placed on top of a layer of CELLPURE Roti-Sep 1077 (Carl Roth) in 50mL LEUCOSEP tubes (Greiner Bio-One) and centrifuged at 800 x g for 20min at room temperature. Subsequently, the PBMCcontaining layer was collected and washed five times in cold PBS + EDTA to remove platelets.

Isolation of CD62L-positive lymphocytes
Approximately 2 x 10 7 PBMCs were passed by gravity flow through an ALFA Selector PE resin loaded with an ALFA-tagged anti-human CD62L nanobody, followed by extensive washing with PBS supplemented with 1mM EDTA and 1% (w/v) bovine serum albumin.
Subsequently, bound cells were eluted in the same buffer containing 200µM ALFA peptide.