We immunized an alpaca with 250 ug of the purified extracellular portion of MICA*009 (obtained by baculovirus expression in the lab of K.W. Wucherpfennig (39)) comprising the α1, α2, and α3 domains in alum adjuvant, followed by 3 booster injections at 2-week intervals. Immunizations were carried out by Camelid Immunogenics. The immune response of the animal was checked by immunoblot ( Supplementary Figure 1 ). Briefly, 1 μg of antigen was resolved by SDS PAGE and transferred to a PVDF membrane. The membrane was incubated with at 1:5000 dilution of alpaca serum collected 2 weeks after the last boost. HRP-linked goat-anti-llama (0.05 μg/mL; Bethyl, NC9656984) was used as the secondary antibody. Membranes were developed with ECL Western Lightning Plus. Mononuclear cells from peripheral blood of the immunized alpaca were isolated by Ficoll gradient separation. The VHH library was generated according to an established protocol (Maas et al., 2007). Briefly, RNA was extracted (RNeasy RNA purification kit, Qiagen) and cDNA was prepared (Superscript III first-strand synthesis system, Invitrogen). The DNA sequences from conventional and heavy-chain only Ig genes are not distinguishable based on the use of specific primers, but two distinct hinge regions are generated between the VHH domain and the CH2 region. We amplified the VHH repertoire from the alpaca using VHH-specific primers that target these hinge sequences ( Supplementary Table 1 ). We pooled the VHH PCR products and ligated them into a phagemid vector in-frame with the pIII gene of the M13 phagemid to construct a phagemid library display. We performed two rounds of panning against MICA*009 immobilized on an ELISA plate, following previously described protocols (40).

DNA from positive clones was sequenced and 9 clones were selected for further characterization. The relevant VHH sequences were subcloned into a pHEN6 expression vector with C-terminal modifications, so that each nanobody sequence included an LPETG motif recognized by sortase A, followed by a (His)₆-tag to facilitate recovery and purification. Briefly, VHH sequences were amplified from the phagemid vector by PCR (primers in Supplementary Table 1 ) and the pHEN6 vector was linearized using the NcoI and BstEII restriction enzymes. Gibson assembly was performed following manufacturer’s directions (Gibson Assembly^® Master Mix, NEB). Positive VHH clones were expressed in WK6 E.Coli in terrific broth and periplasmic protein expression was activated by induction with isopropyl β-D-thiogalactopyranoside (1 mM) at an OD600 of 0.6. VHHs were harvested from the periplasm by osmotic shock. The C-terminal (His)₆-tag allows purification of the recombinant proteins using Ni-NTA agarose beads (Qiagen), followed by FPLC purification on an S75 column by FPLC (ÄKTA, Cytiva Life Sciences). Sortase reactions were performed by incubating each nanobody with a 10-fold molar excess of GGG-nucleophile in the presence of 25 µM Sortase 7M (41) overnight at 4°C. Because the LPETG sequence is cleaved during transpeptidation, the (His)₆-tag immediately C-terminal of the LPETG motif is lost. This allows enrichment of the desired modified product by depletion of His-tagged sortase and unreacted nanobody on a NiNTA matrix, while the unbound fraction contains the modified nanobody.

An ELISA was performed to determine the concentration at which each biotinylated nanobody showed ~80% binding to recombinant MICA*009 (5 mg/mL) immobilized on an ELISA plate. Biotinylated nanobody at a concentration that yielded 80% of the maximum attainable binding value was then mixed with a 500-fold excess of unlabeled competitor nanobody and allowed to compete for binding to 5 μg/mL MICA*009 coated on an ELISA plate. Plates were incubated with streptavidin-HRP (0.00025 μg/mL) for 45-60 minutes at room temperature. After addition of TMB substrate, absorbance was read out at 450 nm on a Spectramax iD5 plate reader (Molecular Devices). If the unlabeled nanobody binds to an epitope distinct from that recognized by the biotinylated nanobody, no diminution of the signal at 450 nm is expected. Nanobodies that recognize the same epitope as that seen by the biotinylated nanobody will show a reduction in the signal at 450nm.

We estimated the binding affinity of VHH-A1 and VHH-H3 by performing an ELISA as previously described (42). Briefly, we incubated plates coated with 100μL PBS containing 2.5 μg/mL recombinant MICA*009 or GFP as negative control with biotinylated VHH-A1 and VHH-H3 in various concentrations (10-fold serial dilutions; 0.000001 nM – 1000 nM). Streptavidin-HRP at 0.00025 μg/mL was used as detection agent. After addition of TMB substrate, absorbance was read at 450 nm on a Spectramax iD5 plate reader (Molecular Devices). Binding affinity was estimated by calculating the IC50 obtained from three experimental replicates with each sample added in duplicates. Recombinant MICA*009 was produced by transfection of EXPI-293 cells with pcDNA3.1(+) vector encoding for extracellular MICA*009 containing a C-terminal LPETG sortase motif followed by a His (6)-tag to facilitate recovery and purification on a NiNTA matrix ( Supplementary Figure 2 ). EXPI-293 cells were transfected using the ExpiFectamine™ 293 Transfection Kit, according to manufacturer’s directions (Gibco).

B16F10 and EL-4 cells and their MICA⁺ transfectants were a gift from the lab of Kai Wucherpfennig. B16F10 cells were cultured in complete DMEM (DMEM with 4.5 g/L glucose, substituted with 10% Fetal Bovine Serum (FBS) and 100 U/mL penicillin/streptomycin). EL-4 cells were cultured in complete RPMI 1640 (RPMI 1640, substituted with 10% FBS and 100 U/mL penicillin/streptomycin). Cells were maintained at optimal densities in a humidified 5% CO2 incubator at 37°C.

EL-4 WT and MICA⁺ cells, or B16F10 WT and MICA⁺ cells, were stained with biotinylated VHH-A1 and VHH-H3 for 30 minutes on ice, washed, and incubated with a cocktail of Streptavidin-conjugated PE at 0.0025 μg/mL (Invitrogen) and 2 μg/mL propidium iodide (Life technologies) for EL-4 or LIVE/DEAD™ Fixable Violet Dead Cell Stain Kit (Invitrogen) for B16F10, both according to manufacturer’s directions for 30 minutes on ice. Cells were analyzed on an LSR Fortessa flow cytometer (BD Biosciences). Gating strategies were based on cell lines stained with the appropriate controls, where single cells and live cells were appropriately selected.

VHH-DM1 was produced in a sortase-mediated transpeptidation reaction. Briefly, 500-1000 μg of VHH containing a C-terminal LPETG-motif was mixed with a 10-fold molar excess of GGG-DM1 and incubated with 25 μM Sortase for 16 hours at 4°C. GGG-DM1 was produced in-house by modifying a GGG-peptide linker to contain a maleimide group and allowing it to react with the thiol group on DM1 (Broadpharm) ( Supplementary Figure 3A ). Unreacted VHH and Sortase, both containing a (His)₆-tag, were depleted by incubation with NiNTA agarose (Qiagen or Prometheus). Excess free GGG-DM1 was removed by desalting on a PD-10 desalting column (Cytiva). We plated 4000 cells/well in a 96-well plate and incubated cells with serial 3-fold dilutions of VHH-drug adduct or free DM4 (Broadpharm), a structural analog of DM1 ( Supplementary Figure 3B ) at 37°C in a humidified 5% CO2 atmosphere. After 72 hours, we measured cell viability by CellTiter GloTM assay according to the manufacturer’s directions (Promega). For co-culture experiments, MICA expression was determined after a 72-hour incubation. Each treatment was performed in duplicate. For flow cytometry, the duplicate wells of each condition were combined, and the cell mixture was stained with 0.0006 μg/mL biotinylated anti-human MICA/B antibody (Clone 6D4, Biolegend) for 30 minutes on ice. Cells were washed and incubated with Streptavidin-conjugated PE at 0.0025 μg/mL (Invitrogen) and LIVE/DEAD™ Fixable Violet Dead Cell Stain Kit according to manufacturer’s directions (Invitrogen) for 30 minutes on ice. Cells were washed and viability and MICA positivity were determined by flow cytometry on an LSR Fortessa flow cytometer (BD Biosciences).

All statistical analysis was performed with GraphPad Prism 8. Flow cytometry data was analyzed with FlowJo (v10.8.1 and v10.9.0).

We immunized an alpaca with purified recombinant MICA*009 in alum adjuvant, followed by 3 booster injections at 2-week intervals. We checked the immune response of the animal by immunoblot using serum samples collected prior to each boost. Having recorded a positive response after the 3rd boost, construction of a phage display library, followed by screening for MICA-reactive hits, yielded positive clones. DNA from positive clones was sequenced and 9 clones were selected for further characterization. Because nanobodies interact with their antigen mainly via their CDR3 region, and to a lesser extent via the germline-encoded CDR1 and CDR2 (43), we chose clones that were unique in their CDR3. A detailed comparison of the nanobody clones based on sequence similarity in the framework and CDR regions is described in the caption of Figure 1 .

Relevant VHH sequences were subcloned into a pHEN6 expression vector to encode a VHH product with C-terminal modifications, so that each VHH sequence included an LPETG motif at its C-terminus, recognized by sortase A, and a (His)₆-tag to facilitate recovery and purification ( Figure 1 ). This arrangement enables the installation of fluorophores, biotin, and other substituents by a site-specific and efficient sortase-catalyzed transpeptidation reaction (41). Because the LPETG sequence is cleaved during transpeptidation, the (His)₆-tag immediately C-terminal of the LPETG motif is lost. This allows enrichment of the desired modified product by depletion of His-tagged sortase and unreacted nanobody on a NiNTA matrix, while the unbound fraction contains the modified nanobody.

To determine whether the isolated MICA-specific nanobodies recognized similar or distinct epitopes on MICA, we performed cross-competition experiments by ELISA. Competition of unlabeled nanobodies with a biotinylated nanobody for binding to MICA showed that this set of nanobodies recognizes two distinct epitopes, one defined by the H3 nanobody and the second by all the other nanobodies. None of the nanobodies compete for binding with the 7C6 monoclonal antibody, an agent that inhibits shedding of MICA (45) ( Figure 2A ). Typically, not all nanobodies are suitable for use in immunoblotting experiments, but the biotinylated versions of A1 and H3 yielded a strong and specific signal in immunoblots on recombinant MICA ( Figure 2B ). The binding affinities of VHH-A1 and VHH-H3 are both in the nanomolar range, at ~0.2 and ~0.4 nM for A1 and H3 respectively ( Figure 2C ), as estimated by ELISA assay. By examining the binding of the A1 and H3 nanobodies to a subset of MICA/B allelic products, available in purified form, we conclude that the A1 and H3 nanobodies recognize the MICA*008 and MICA*009 alleles ( Figure 2D ) which, combined, cover 51.1% of the Caucasian population (46). To verify that A1 and H3 also recognize surface-disposed MICA, we used B16F10 transfectants that express MICA*009, and EL-4 transfectants that express MICA*008, with B16F10 and EL-4 wild type cells serving as negative controls. Both A1 and H3 showed excellent staining of the MICA transfectants by flow cytometry and yielded no signal for the untransfected parental cell lines ( Figure 2E ) with a significant difference determined by mean fluorescence intensity (MFI) ( Figure 2F ). Gating strategies are shown in Supplementary Figure 4 .

The reactivity of VHH-A1 and VHH-H3 make them appealing candidates for the construction of nanobody-drug conjugates. To test this, we ligated the Maytansine derivative DM1, a microtubule disrupting agent, to VHH-A1 or to a VHH that targets mouse MHC-II (VHH_MHC-II) (47) as a negative control via a sortase-mediated transpeptidation reaction ( Figure 3A ) and confirmed successful ligation with SDS-PAGE ( Figure 3B ). We performed an in vitro cytotoxicity assay by titration of VHH_MHC-II-DM1, VHH_A1-DM1, or free DM4 (a functional analog of DM1) on EL-4 WT and MICA⁺ cells. EL-4 MICA⁺ cells were sensitive to VHH_A1-DM1, with a stronger cytotoxic effect at lower doses of the VHH-drug conjugate compared to VHH_MHC-II-DM1, as estimated by IC50. The IC50 of VHH_A1-DM1 treated EL-4 MICA⁺ cells was comparable to that of cells treated with free DM4. Similarly treated WT cells showed no obvious reduction in viability with either nanobody-drug conjugate ( Figure 3C ).

To further validate selectivity of VHH_A1-DM1 for MICA⁺ cells, we co-cultured EL-4 WT and EL-4 MICA⁺ cells at a 1:1 ratio, and added VHH_MHCII-DM1, VHH_A1-DM1, or free DM4 at different concentrations. We determined the ratio of viable EL-4 WT and EL-4 MICA⁺ cells after 72 hours by flow cytometry using a live/dead cell stain. We stained the MICA⁺ cells in the co-culture with a biotinylated αMICA mAb, using streptavidin-conjugated PE as secondary reagent. Gating on live cells and MICA⁺ cells showed specific elimination of MICA⁺ cells at adduct concentrations between 1.71 nM and 416 nM for VHH_A1-DM1. A difference in ratio between WT and MICA⁺ cells was not observed in cells treated with VHH_MHCII-DM1 or free DM4. Because WT cells proliferate slightly faster than MICA⁺ cells in culture, the distribution shifted to ~65% WT and 35% MICA⁺ cells after 72 hours in culture. Thus, numbers were normalized according to the percentage of cells of either line in the untreated (“0 nM”) group ( Figure 3D ). Gating strategies are shown in Supplementary Figure 5 .

Tumor cells can downregulate surface expression of MICA through shedding, mediated by proteolytic cleavage at the α3 domain. Increased levels of soluble MICA (sMICA) in the serum of patients are associated with poor prognosis and worse disease progression (10, 48–50). To address the possible competition of sMICA for binding with the anti-MICA nanobody, we performed an in vitro cytotoxicity assay. EL-4 WT and MICA⁺ cells were incubated with VHH_MHCII-DM1, VHH_A1-DM1, or free DM4 at a fixed concentration of 2.5 nM, in the presence of sMICA at various concentrations (serial 2-fold dilutions; 0-5 nM/0-170 ng/mL). We observed no reduction in cytotoxicity of VHH_A1-DM1 on MICA⁺ cells upon addition of sMICA to the medium ( Figure 3E ). Publications report concentrations of sMICA in the serum of MICA⁺ patients in the range of 0.1-15 ng/mL (51–53) which is at least 10-fold lower than the sMICA concentration in our competition assay. We thus expect little to no impact of sMICA in patients’ serum on the ability of these nanobodies to target membrane-bound MICA in vivo.