FLEXamers: A Double Tag for Universal Generation of Versatile Peptide-MHC Multimers

A T cell’s function is determined largely through the affinity of the TCR to Ags presented on the MHC (peptide-MHC [pMHC]) of cells. Analyses of TCR:pMHC interactions have been challenging as the affinity of monomeric pMHC molecules is not strong enough for stable binding. Scaffolds allowing multimerization enable analyses of weak and transient interactions of molecules through a gain in avidity through multivalent binding. Soluble pMHC monomers (biotinylated, e.g., via an Avi-tag) can be multimerized on a dye-conjugated streptavidin backbone (“tetramer”) (1). This enables sensitive identification and isolation of Ag-specific T cells and has opened up new avenues for in-depth T cell analysis in basic research and immune monitoring in a clinical setting (2).

However, stable binding of pMHC tetramers can also deteriorate T cell functionality in vivo (3, 4). The fact that pMHC ligand binding to the TCR is stable in its multivalent form but can be reversed upon monomerization has been exploited for the development of clinical cell selection and processing technologies (5–8) and is further used for in-depth characterization of TCR:pMHC interactions. Reversible pMHC multimer reagents, such as “Streptamers,” allow traceless isolation of cell products with no functional difference compared with cells that have never bound pMHC multimers (5–9). When reversible pMHC monomers themselves are labeled with a fluorophore, their dissociation from TCRs on living T cells can be tracked over time (10, 11). Through this, absolute and reproducible measurements of TCR:pMHC dissociation (k_off) rates can be achieved in a relatively easy and high-throughput compatible manner. TCR-ligand k_off rates indicate TCR avidity and are predictive of T cell functionality (10).

Until now, the versatility of pMHC multimer reagents comes at the cost of distinct generation processes for each application (Supplemental Fig. 1). Separate recombinant protein expression, in vitro refolding, and pMHC purification processes make the synthesis of pMHC-based reagents laborious, time-consuming, and prone to batch-to-batch variability. The cumbersome and expensive effort to generate distinct pMHC constructs for each application has so far prevented many laboratories to make broad use of the versatility of pMHC multimer reagents. Ideally, the three distinct constructs should emerge from one common pMHC precursor, thus streamlining the production process while simultaneously providing full flexibility to generate all other pMHC multimer types. Flexibility can be achieved by enzymatic functionalization tags (12, 13). In addition, Strep- and His-tags can be used to generate reversible pMHC multimer reagents (14). So far, however, no approach has provided an all-in-one solution to produce versatile pMHC-based reagents within a simple generation process. In this article, to our knowledge, we present a novel approach to generate distinct pMHC multimer reagents from a single, highly functional, double-tagged pMHC precursor protein (“FLEXamer”). FLEXamers can be used without further modification for traceless isolation of T cells but can also be conjugated 1) with biotin for stable identification of Ag-specific T cells, 2) with fluorescent dyes to track dissociation of monomeric pMHC molecules for TCR avidity measurement, or 3) with any probe of interest.

Tubulin tyrosine ligase (TTL) was expressed and purified as follows, according to a published protocol (15). The TTL (Canis lupus) coding sequence was amplified from a mammalian expression vector (16), cloned into a pET28-SUMO3 (EMBL-Heidelberg, Protein Expression Facility), and expressed in Escherichia coli BL21(DE3) as a Sumo-TTL fusion protein with an N-terminal His-tag. Expression was induced with 0.5 mM isopropyl β-d-thiogalactoside and incubated at 18°C for 18 h. Lysis was performed in the presence of lysozyme (100 μg/ml), DNAse (25 μg/ml), and PMSF (2 mM), followed by sonication (Branson Sonifier; five times, 7 × 8 s, 40% amplitude) and debris centrifugation at 20,000 × g for 30 min. His-Sumo-TTL was purified using a 5 ml HisTrap (GE Healthcare). Purified protein was desalted on a PD10 column (GE Healthcare); buffer was exchanged to MES/K (pH 7.0; 20 mM MES, 100 mM KCl, 10 mM MgCl₂) supplemented with 3 mM 2-ME, 50 mM l-glutamate, and 50 mM l-arginine. Protein aliquots were shock frozen and stored at −80°C.

pET3a expression vectors containing the coding sequence of Strep-tagged HLA allotypes, including murine H2-K^b, served as parental plasmid to insert the Tub-tag or sortase A (SrtA)-tag sequence seamlessly downstream of Strep-tag. All insertions were performed using the Q5 Site-Directed Mutagenesis Kit (New England BioLabs) following manufacturer’s protocol. Insertion primers (Sigma-Aldrich) contained 18 bp of plasmid binding sequence flanking the integration site and encoded one-half of the Tub- or SrtA-tag sequence.

All pMHC monomers described in this report, including the double-tagged pMHC molecules, were generated as previously described (2, 5). In brief, recombinantly expressed and purified human as well as murine MHC H chain (HC) and β2 microglobulin were denatured in urea and subsequently refolded into the heterotrimeric pMHC complex in the presence of an excess of peptides (synthetic peptides purchased by Peptides and Elephants). Correctly folded pMHC monomers were purified using size exclusion chromatography, concentrated, and stored at −80°C or in liquid nitrogen. All conventional dye-conjugated Streptamers used for k_off rate measurements were generated by Maleimide chemistry using a solvent-exposed artificial cysteine residue as described (10, 17).

TTL-catalyzed ligation of 3-azido-l-tyrosine (Watanabe Chemical Industries) to Tub-tagged FLEXamers was performed in 25–100 μl consisting of 20 μM FLEXamer, 5 μM TTL, and 1 mM 3-azido-l-tyrosine in TTL-reaction buffer (20 mM MES, 100 mM KCl, 10 mM MgCl₂, 2.5 mM ATP, and 5 mM reduced glutathione) at 25°C for 3 h followed by buffer exchange to 20 mM Tris HCl and 50 mM NaCl (pH 8) by size-exclusion chromatography (Zeba Spin desalting columns, 7K MWCO; Thermo Scientific). Azido-FLEXamers were stored at 4°C or directly used for click functionalization.

Azido-FLEXamers were functionalized by incubation of 20 μM azido-FLEXamer with either 400 μM DBCO-PEG₄-Biotin, 400 μM DBCO-sulfoCy5, or 200 μM DBCO-PEG₄-Atto488 (Jena Bioscience) for 18 h at 16°C followed by buffer exchange to 20 mM Tris, 50 mM NaCl (pH 8), and storage at −80°C. Conjugation was analyzed by reducing SDS-PAGE and Coomassie staining. Conjugation efficacies were assessed from scanned Coomassie-stained SDS-PAGE gels. For this, the Gel Analyzer plugin of the Fiji software was used to quantify band intensities of unconjugated and conjugated HC. The efficiency was calculated using the following equation: intensity-labeled HC/(intensity-labeled HC + intensity-unlabeled HC). To confirm the identity of the attached functional groups, biotinylated FLEXamers were plotted on a nitrocellulose membrane, stained with a streptavidin-Alexa Fluor 594 (Dianova) conjugate, and detected on an Amersham Imager 600 system (GE Healthcare). In-gel fluorescence of fluorophore-labeled FLEXamers was directly detected using the same instrumentation.

A total of 10 μM SrtA-tagged FLEXamer was incubated with 1 mM Gly₅-FITC or 1 mM Gly₅-biotin peptide (JPT Peptide Technologies GmbH) and 30 μM SrtA (kindly provided as purified enzyme derived from Staphylococcus aureus by Dr. H. Meyer; Technical University of Munich) in 20 mM HEPES and 5 mM CaCl₂ (pH 7.5) at 25°C for 18 h. Ni-NTA Agarose–based pulldown (Quiagen) in PBS and 20 mM imidazole (pH 8) at 4°C for 30 min was used to remove His-tagged SrtA and SrtA-tagged FLEXamer educts still carrying the His-tag. Purified functionalized SrtA-tagged FLEXamers were buffer exchanged after functionalization to 20 mM Tris and 50 mM NaCl (pH 8). Conjugation and purification were analyzed by SDS-PAGE, followed by detection of in-gel fluorescence and Coomassie staining.

CMV-reactive T cell clones were generated and cultured as described previously (10). Primary T cells reactive for CMV were derived from healthy CMV-seropositive donors. Written informed consent was obtained from the donors, and usage of the blood samples was approved according to national law by the local Institutional Review Board (Ethikkommission der Medizinischen Fakultät der Technischen Universität München). Blood was diluted 1:1 with sterile PBS and PBMCs isolated by density gradient centrifugation using Leucosep tubes (Greiner Bio-One) following manufacturer’s protocol.

All reversible pMHC monomers (with and without dye) were multimerized on Strep-Tactin APC or Strep-Tactin PE (IBA) by incubating 1 μg of reversible pMHC monomer and 1 μl of Strep-Tactin APC or PE in a total volume of 50 μl of FACS buffer for 30 min on ice in the dark. Conventionally biotinylated pMHC monomers for generation of nonreversible multimers were generated as described (2). Subsequently, all biotin functionalized pMHC monomers described in this report were multimerized by incubation of 0.4 μg of biotinylated pMHC monomers with 0.1 μg of streptavidin-BV421 (BioLegend), 0.25 μg of streptavidin-PE (eBioscience), or 0.1 μg of streptavidin-APC (BioLegend) in a total volume of 50 μl of FACS buffer for 30 min on ice in the dark. For k_off rate measurements, up to 5 × 10⁶ cells were incubated with dye-conjugated reversible pMHC multimers for 45 min on ice in the dark. Ab staining (CD8 eF450 eBioscience, Thermo Scientific) was added after 25 min, and cells were incubated for an additional 20 min. If combinatorial staining with nonreversible pMHC multimers was performed, cells were washed and incubated for 10 min with nonreversible pMHC multimers on ice in the dark. For live/dead discrimination, cells were washed in propidium iodide solution. When solely performing pMHC multimer staining with a combination of nonreversible pMHC multimers, staining was incubated for 30 min on ice in the dark. We routinely stain the pMHC multimer conjugated to the smaller dye first. After incubation, cells were washed and stained with the second pMHC multimer for 30 min. Ab staining was added after 10 min, and cells were incubated for an additional 20 min. When cells were stained with reversible pMHC multimers for traceless cell isolation, samples were incubated for 45 min with the multimer regent. After 25 min, Ab staining was added, and cells were incubated for an additional 20 min. All FACS data were analyzed with FlowJo software (FlowJo).

Acquisition of FACS samples was done on a CyAn ADP Px9 color flow cytometer (Beckman Coulter). Flow sorting was conducted on a MoFlo legacy (Beckman Coulter). k_off rate measurements were performed as described (18). In brief, samples were transferred into precooled FACS tubes containing a total volume of 1 ml of FACS buffer and placed into a Peltier cooler (qutools GmbH) set to 5.5°C. After 30 s acquisition, 1 ml of cold 2 mM d-biotin was added into the ongoing measurement. Dissociation kinetics were measured for 15 min. For analysis of k_off rate data, fluorescence data of Ag-specific cells were exported from FlowJo to PRISM (GraphPad Software). The t_1/2 was determined by fitting a one-phase exponential decay curve.

We hypothesized that combining a site-specific functionalization tag with a reversible multimerization tag to a double-tagged FLEXamer will unite reversibility with the opportunity to equip the pMHC with any desired additional functionality (Fig. 1, Supplemental Fig. 1). For site-specific conjugation, we first chose a new chemoenzymatic system, termed Tub-tag (15), which is based on a short hydrophilic, unstructured sequence recognized by TTL (19). TTL-catalyzed attachment of tyrosine derivatives, such as 3-azido-L-tyrosine, allows subsequent addition of a variety of functional groups, such as biotin or dyes, by highly efficient and mild click chemistry (15). Similar to other chemoenzymatic approaches, the TTL reaction is not reversible; however, the product does not suffer from hydrolysis, and the substrate tyrosine derivatives represent compounds that are easy to synthesize (15) or commercially available (see 2Materials and Methods).

We performed proof-of-concept experiments to test if we could use this strategy to conjugate biotin or dyes to Strep- and Tub-tagged FLEXamers. We generated two different FLEXamers for the HLA class I HC B*07:02 and B*08:01, which present CMV pp65 and IE1, respectively (Fig. 2A). Enzymatic activation of the common precursor FLEXamer and subsequent conjugation with biotin, Atto488, or sulfo-cyanine5 was highly efficient, with conversion rates >95% based on Coomassie-stained SDS-PAGE gel band intensities (Fig. 2B; see 2Materials and Methods).

We then tested whether the nonreversible biotinylated FLEXamer, the reversible dye-conjugated FLEXamer, and their reversible FLEXamer precursor could fulfill their distinct functions. Biotinylated FLEXamers stained B*07:02/pp65_417–426–specific T cells from peripheral blood of a CMV-seropositive donor with high sensitivity and no difference to conventionally biotinylated tetramers (Fig. 3). An irrelevant epitope/MHC combination (A*02:01/Her2neu_369–377) served as control for unspecific staining (Fig. 3).

To test for possible interference of the functionalization tag with reversibility, we stained and flow sorted B*07:02/pp65_417–426–specific CD8⁺ T cells from peripheral blood of a CMV-seropositive donor either with conventional Streptamers or FLEXamers (Fig. 4). FLEXamers could stain B*07:02/pp65_417–426-specific T cells and allowed high purity flow cytometric sorting like conventional Streptamers (Fig. 4B). Upon addition of d-biotin, the pMHC label could be detached. Complete removal of pMHC monomers from the cells is demonstrated by the inability to restain the cells by solely adding the Strep-Tactin backbone, whereas addition of the multimerized FLEXamer resulted in efficient restaining (Fig. 4B).

Conjugation of dyes to Streptamer pMHCs allows direct tracing of pMHC monomer dissociation kinetics after addition of d-biotin, to measure TCR:pMHC k_off rates for TCR structural avidity estimation (10, 18) (Fig. 5). When a B*07:02/pp65_417–426 T cell clone was stained with dye-conjugated Streptamers or FLEXamers, the dye-conjugated pMHC molecules showed monomeric pMHC dissociation after initial dye dequenching, as previously described (10) (Fig. 5B). The k_off rates determined by fitting of exponential decay curves were identical for dye-conjugated Streptamers and FLEXamers (Fig. 5D). We next tested the functionality of murine FLEXamers and therefore generated FLEXamers for H2-K^b/OVA_257–264. We measured k_off rates of OT-I transgenic T cells, which were as fast as expected (20) and did not differ between conventionally generated dye-conjugated Streptamers and dye-conjugated FLEXamers (Supplemental Fig. 2A).

Using double staining with a nonreversible biotinylated pMHC multimer and a reversible dye-conjugated pMHC Streptamer, dissociation kinetics can be tracked without previous purification on a flow cytometer through continuous gating on the nonreversible pMHC multimer⁺ T cell population (18). This emphasizes that not only the different pMHC constructs themselves but also their combinatorial use enable in-depth T cell characterization. We costained a heterogeneous B*07:02/pp65_417–426–specific T cell population directly ex vivo with both nonreversible pMHC conjugated to biotin and reversible pMHC conjugated to Atto488 (Fig. 5C). Nonreversible pMHC multimers allowed continuous gating on the Ag-specific T cell population after the addition of d-biotin (Supplemental Fig. 2B), whereas the reversible fluorophore-conjugated pMHC monomers dissociated over time (Fig. 5C, Supplemental Fig. 2E). The heterogeneous T cell populations specific for B*07:02/pp65_417–426 entailed two distinct kinetics (Fig. 5C, Supplemental Fig. 2D). We retrieved T cell clones from both kinetics and stained them with dye-conjugated Streptamers and FLEXamers. Again, we obtained highly comparable dissociation rates resembling those of the parental T cell population (Supplemental Fig. 2D). After 60 min of d-biotin addition, both kinetics reached baseline, validating that both populations represented true dissociation kinetics (Supplemental Fig. 2C). The combinatorial use of pMHC reagents therefore allows visualization of subpopulations with different dissociation kinetics from a common native heterogeneous T cell population directly ex vivo. FLEXamers enable universal generation of these different pMHC constructs from a common precursor protein.

Next, we set out to test the general applicability of our double-tag approach. Therefore, we cloned and refolded an HLA-A*02:01 FLEXamer harboring a SrtA recognition tag for versatile protein conjugation via transpeptidation (13). This construct is additionally equipped with a His-tag for fast and efficient protein purification after transpeptidation (Supplemental Fig. 3A). We stained PBMCs with a transgenic TCR specific for A*02:01/pp65_495–503 with Tub-tag– or SrtA-biotinylated tetramers (Supplemental Fig. 3B) and also tested reversibility of the SrtA-tag carrying FLEXamer precursor (Supplemental Fig. 3C). Furthermore, Tub- or SrtA-tag dye-conjugated reversible FLEXamers were tested for characterization of TCR:pMHC k_off rates (Supplemental Fig. 3D, 3E). In each case, SrtA FLEXamers, Tub-tag FLEXamers, and their biotin- or dye-conjugated downstream pMHC products performed equally well (Supplemental Fig. 3B–E) independent of the respective functionalization strategy. However, compared with Tub-tag, SrtA-mediated pMHC functionalization is less efficient overall, which had to be compensated by significantly increased educt consumption. We therefore focused on Tub-tag technology to generate FLEXamers.

Encouraged by the much simpler generation process of different pMHC multimer reagents from a single double-tagged FLEXamer precursor, we also generated FLEXamers for other epitope–HLA combinations. For B*08:01 presenting IE1_199–207K, we validated the equal functionality of nonreversible, reversible, and fluorophore-conjugated FLEXamers (Supplemental Fig. 4). To even further extend the set of available FLEXamers, we folded 26 FLEXamers in total, covering nine HLA class I HC as well as the murine HC H2-K^b (Fig. 6A). The conjugation efficacy with fluorophore or biotin was consistently high for all FLEXamers (Fig. 6B). Because of the skewed frequency distribution of HLA class I alleles, the nine human HLA HC together cover 76.5% of the European Caucasian population (Fig. 6C, 6D) and also entail two allotypes (A*24:02 and A*11:01), which are highly prevalent in Asian populations. This set of FLEXamers can serve as precursors for any kind of pMHC reagent.

The heterogeneity of infectious agents and cancers is met by the adaptive immune system’s ability to present and recognize many different targets. The total epitope repertoire has been estimated to be between 10⁶ and 10¹¹ in mice (21) and is likely similarly, if not even more diverse in humans. More than 13,000 HLA class I alleles have now been described for humans (22), and the total human TCR repertoire encompasses more than 10⁸ unique clonotypes (23). Customized monitoring of Ag-specific immune responses and individualized immunotherapy therefore require streamlined methods that allow flexible adaptation for each patient and disease in terms of target-specific epitopes as well as patient-specific HLA (24). The versatile applicability of pMHC multimer reagents (for T cell identification, traceless isolation, or TCR avidity measurement) makes them particularly valuable tools for the investigation and therapeutic use of T cells (25) but consequently adds even a third level of complexity as so far as the specific reagents needed to be produced separately.

To be compatible with the extreme diversity of epitopes, UV exchange (26) or dipeptide (27) technologies have been developed that can be used to load HLA class I with any epitope of interest. In addition, combinatorial pMHC staining (28, 29) and DNA barcoding (30) have massively enhanced the throughput of screening Ag-specific T cell populations and their respective TCR repertoires. Despite this progress, difficulties to generate distinct pMHC multimer reagents appropriate for each individual setting in a fast and reliable manner remain a significant challenge to personalized cell therapy and T cell–based diagnostics. Importantly, for broad applicability of pMHC multimer reagents, the production process of different kinds of pMHC multimer reagents needs to also be feasible for many different HLA class I HC.

FLEXamers combine the provision of versatility through distinct pMHC constructs with a simple generation process from a single precursor protein (Fig. 1). FLEXamers are highly functional while being produced in a faster and more standardized manner compared with conventionally generated pMHC reagents. A core feature of FLEXamers is a novel double tag that allows reversible multimerization as well as functionalization with any probe of interest. In this study, we provide proof of concept to generate biotinylated tetramers, reversible Streptamers, or reversible dye-conjugated pMHC multimer reagents from a common FLEXamer precursor protein. Notably, the use of FLEXamers is not limited to these specific reagents as the functionalization tag also allows conjugation, for example, of DNA oligonucleotide sequences, toxins (Fig. 1), and many more entities. Furthermore, FLEXamers can be readily combined with epitope exchange technologies (27, 31). Because of the combination of a simple generation process with versatile application, FLEXamers facilitate implementation of different types of pMHC reagents in routine clinical immune monitoring and immunotherapy.

The effort and costs needed to generate distinct pMHC multimer reagents for each application has so far been a key obstacle for laboratories to fully exploit the versatility of pMHC-based reagents. Double-tagged pMHC FLEXamers can be easily generated and applied. Furthermore, although the double tag is a general concept readily compatible with the use of alternative functionalization tags (e.g., SrtA-tag). Notable advantages of Tub-tag technology are mild reaction conditions combined with high conjugation efficiencies using click chemistry, whereas functionalization via artificial solvent-exposed cysteine residues generates the risk of dimer formation via disulfide bridges and changes in tertiary structure because of the introduction of a polar amino acid. Finally, we used Tub-tag technology to generate a comprehensive set of versatile pMHC FLEXamers for nine different human HLA as well as murine H2-K^b.

Multivalent binding can serve as an “on switch” to stabilize otherwise transient binding of weak interaction partners. In turn, receptor–ligand binding can be switched off via disruption of the multimeric complex, which requires that the multimerization is reversible. Versatile functionalization thereby allows further stabilization of the interaction or tracking via fluorescent dyes. The field of T cell immunology has made extensive use of this trick through multimerization of pMHC monomers. Our double-tag approach enables universal generation of different pMHC constructs, but also constitutes a flexible tool for investigation of transient protein–protein interactions in general.

We thank I. Andrä, L. Henkel, and all members of the TU Munich Flow Core Facility for cell sorting; I. Andrä for technical support in flow cytometry and J. Groffmann and J. Schwach for excellent practical support in FACS experiments; P. Lückemeier for development of k_off rate analysis software; F. Mohr for critical discussions and helpful advice in experimental design; F. Graml and A. Hochholzer for excellent technical support in the generation of pMHC monomers; and V. R. Buchholz, F. Mohr, J. Baldwin, J. Leube, and J. Schütz for critical reading of the manuscript.

The Tub-tag technology is part of a patent application filed by D.S., J.H., and H.L. and is subject to the spin-off project Tubulis. M.E., A.S, D.H.B., and L.H. applied for a patent covering the double tag. D.H.B. invented the Streptamer technology. The other authors have no financial conflicts of interest.