The butyrophilin 3A1 intracellular domain undergoes a conformational change involving the juxtamembrane region

Abstract

Small isoprenoid diphosphates, such as (E)-4-hydroxy-3-methyl-but-2-enyl diphosphate (HMBPP), are ligands of the internal domain of BTN3A1. Ligand binding in target cells promotes activation of Vγ9Vδ2 T cells. We demonstrate by small-angle X-ray scattering (SAXS) that HMBPP binding to the internal domain of BTN3A1 induces a conformational change in the position of the B30.2 domain relative to the juxtamembrane (JM) region. To better understand the molecular details of this conformational rearrangement, NMR spectroscopy was used to discover that the JM region interacts with HMBPP, specifically at the diphosphate. The spectral location of the affected amide peaks, partial NMR assignments, and JM mutants (ST²⁹⁶AA or T³⁰⁴A) investigated, confirm that the backbone amide of at least one Thr (Thr³⁰⁴), adjacent to conserved Ser, comes close to the HMBPP diphosphate, whereas double mutation of nonconserved residues (Ser/Thr^296/297) may perturb the local fold. Cellular mutation of either of the identified Thr residues reduces the activation of Vγ9Vδ2 T cells by HMBPP, zoledronate, and POM₂-C-HMBP, but not by a partial agonist BTN3 antibody. Taken together, our results show that ligand binding to BTN3A1 induces a conformational change within the intracellular domain that involves the JM region and is required for full activation.—Nguyen, K., Li, J., Puthenveetil, R., Lin, X., Poe, M. M., Hsiao, C.-H. C., Vinogradova, O., Wiemer, A. J. The butyrophilin 3A1 intracellular domain undergoes a conformational change involving the juxtamembrane region.

T cells that express the Vγ9Vδ2 T-cell receptor are activated by (E)-4-hydroxy-3-methyl-but-2-enyl diphosphate (HMBPP) (1–4) and related essential intermediates (5–7) of isoprenoid metabolism (also known as phosphoantigens) in a manner that is independent of the major histocompatibility complex. The ability to detect the presence of these small molecules in target cells implies a role for Vγ9Vδ2 T cells in response to certain pathogens (8) and malignant cells (9, 10). Several groups have investigated these molecules and their analogs as potential chemoimmunotherapeutic agents (11–17). Thus, it is important to understand the mechanisms of this unique activation process.

Recent studies have identified the B7 family protein, butyrophilin (BTN)-3 isoform A1 (BTN3A1; also known as CD277), as essential for the activation of Vγ9Vδ2 T cells by HMBPP (18–20). BTN3A1 is one of 3 BTN-3 isoforms (BTN3A1, -3A2, and -3A3) found in humans and other primates and is essential in cells that can trigger Vγ9Vδ2 activation upon exposure to HMBPP (18, 21, 22). After removal of its signal peptide, BTN3A1 is an ∼54 kDa, type I receptor glycoprotein of the Ig family containing an IgV/IgC extracellular domain attached to a B30.2 (PRY/SPRY) domain at its cytosolic end (Fig. 1A). The intracellular B30.2 domain is connected to the transmembrane helix by a 68-aa juxtamembrane (JM) region (Fig. 1B), the structure and function of which remains unknown. Different modeling approaches (23–26) fail to converge in the prediction of its secondary structural elements (Fig. 1C).

Figure 1.

The JM region of BTN3A1 connects the transmembrane domain to the B30.2 domain. A) Domain architecture of BTN3A1. B) Sequence of the 68-aa JM region of BTN3A1 and -3A3 highlighting the 2 Ser/Thr residues used in this study (red line). C) Simplistic BTN3A1 protein structure consists of 2 Ig-like extracellular domains (IgV, IgC), a TM and a JM region, and an intracellular B30.2 domain. The JM region is shown in model 1 as determined by a homology-based approach (25), reflecting a conformation with potential kink within the helical-coiled region suggested previously (22), or in a folded model 2 and extended model 3 as determined by I-TASSER. D) Structures of HMBPP and DMAPP. V, IgV domain; C, IgC domain; TM, transmembrane helix.

HMBPP (Fig. 1D) (22, 27) binds to BNT3A1 within the shallow basic pocket of the BTN3A1 B30.2 domain. However, it is unclear how binding of the ligand triggers Vγ9Vδ2 activation (22, 28) and whether the JM region is involved in that process. We have demonstrated by NMR spectroscopy that binding of HMBPP to the full intracellular domain (BFI) perturbs residues found within the JM region (29), suggestive of a ligand-induced conformational change to the intracellular domain of BTN3A1 beyond just the B30.2 domain, which may involve a direct interaction with the ligand. The comparison of crystal structures (27) of the apo (4N7I) and C-HMBPP bound (4N7U) B30.2 domain suggests that ligand binding can occur to some extent in the absence of significant conformational rearrangement within the B30.2 itself, although crystals for determination of the bound form were obtained from the cross-linked apo form further soaked with the phosphoantigen and the structure derived may not truly represent the native activated state. More important, these data cannot reveal whether ligand binding facilitates a major conformational change involving the JM region. This intracellular conformational change could affect the binding of BTN3A1 to potential intracellular binding partners, such as periplakin (PPL), a cytoskeletal adaptor protein (21), or RhoB, a small GTPase that has been reported to contribute to an extracellular conformational change (30).

Central to the mechanism of activation of Vγ9Vδ2 T cells through BTN3A1 is the binding of a ligand to the intracellular B30.2 domain. The present study sheds light on how this binding contributes to the activation process. We hypothesize that the JM region plays a vital role in conformational rearrangement of the activated receptor. We examine, in more detail, the role of the BTN3A1 JM region in HMBPP-induced activation of Vγ9Vδ2 T cells.

MATERIALS AND METHODS

Reagents

RPMI-1640, fetal bovine serum (FBS), nonessential amino acids, pyruvate, penicillin streptomycin, 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid, 2-ME, guanosine diphosphate, sodium diphosphate, and zoledronate were obtained from Thermo Fisher Scientific (Waltham, MA, USA). HMBPP and DMAPP were obtained from Echelon (Salt Lake City, UT, USA). POM₂-C-HMBP (29) was a gift from Prof. David Wiemer at the University of Iowa (Iowa City, IA, USA). IL-2 and the MACS γδ T-cell negative selection kit were obtained through Miltenyi Biotec (Somerville, MA, USA). K562 cells were from ATCC (Manassas, VA, USA), and blood was procured from Research Blood Components (Boston, MA, USA). pcDNA3.1 and DNA primers were obtained from Thermo Fisher Scientific; restriction enzymes including XhoI, EcoRV, and NheI from New England Biolabs (Ipswich, MA, USA); G418 from Cayman Chemical (Ann Arbor, MI, USA); Myc-tag rabbit antibody (2272) from Cell Signaling Technologies (Danvers, MA, USA); and anti-CD277 (BT3.1), phycoerythrin (PE)-conjugated anti-CD277 (BT3.1), and PE-conjugated mouse IgG1 isotype control (MOPC-21) from BioLegend (San Diego, CA, USA).

Cell culture

Human peripheral blood mononuclear cells were isolated from blood with Lymphoprep (Axis-Shield, Oslo, Norway). Cells were aliquotted in freezing medium (10% DMSO, 20% FBS, and 70% medium) and stored in liquid nitrogen. Cells were resuspended at 1 × 10⁶ cells/ml in fresh T-cell medium (RPMI-1640, 10% heat-inactivated FBS, 1× 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid, pyruvate, nonessential amino acids, and 2-ME). Cells were stimulated with 0.01 μM HMBPP for 3 d and cultured for another 4–18 d after compound removal. Fresh IL-2 (5 ng/ml) was provided every 3 d. After 7–14 d, Vγ9Vδ2 T cells were purified by negative selection. Experiments were performed at least 3 times with at least 2 donors. K562 cells were cultured in RPMI-1640, 10% heat-inactivated FBS, and 1% penicillin-streptomycin and maintained at between 0.2 and 1 × 10⁶ cells/ml.

Molecular cloning

Recombinant B30.2 (hBTN3A1^340–513) domain and BFI (hBTN3A1^272–513), the intracellular domain of human BTN3A1 containing both the JM and B30.2 domains, had been cloned into the pET21b vector (29). The JM region (hBTN3A1^272–339) of BTN3A1 was cloned into the NdeI and XhoI restriction sites of pET15b with the forward primer 5′-GCTATCCATATGCAACAGCAGGAGGAAAAA-3′ and the reverse primer 5′-GCTATCCTCGAGCTAAGGCTTGAAGAGGGCCTTT-3′. Two separate mutants, JM-ST²⁹⁶AA and JM-T³⁰⁴A, were created from the JM plasmid in pET15b. The 296 mutation was generated with the forward primer 5′-GTTGAGAGAAATGGCATGGGCCGCAATGAAGCAAGAACAAAG-3′ and the reverse primer 5′-CTTTGTTCTTGCTTCATTGCGGCCCATGCCATTTCTCTCAAC-3′. The 304 mutation used the forward primer 5′-GAAGCAAGAACAAAGCGCAAGAGTGAAGCTCC-3′ and the reverse primer 5′-GGAGCTTCACTCTTGCGCTTTGTTCTTGCTTC-3′.

Periplakin constructs (hPPL^212–495 and hPPL^495–865) were cloned into the NdeI and NotI cut sites of pET21b. PPL^212–495 was generated with the forward primer 5′-GCTATCCATATGCTGGCCAAGGACGGGGAC-3′ and the reverse primer 5′-GCTATCGCGGCCGCGCTGTCAGCCAGAGCCAGG-3′, and the PPL^495–865 construct was generated with the forward primer 5′-GCTATCCATATGAGCCTGGGCAGCCAGTACC-3′ and the reverse primer 5′-GCTATCGCGGCCGCTTCCTTCACTTTGGTGGCAG-3′. All clones were confirmed through forward and reverse Sanger sequencing.

We have reported the generation of BTN3A1-deficient K562 cells by CRISPR/Cas9 and a pcDNA3.1-BTN3A1 rescue construct (31). Point mutations harboring either ST²⁹⁶AA or T³⁰⁴A in pcDNA3.1-BTN3A1 were generated with the same primers as above.

Overexpression of BTN3A1 and mutants

In BTN3A1-deficient K562-knockout (KO) cells, BTN3A1, ST²⁹⁶AA, and T³⁰⁴A were reintroduced by transfection and selection. KO cells were resuspended in serum-free medium at 2.5 × 10⁷ cells/ml. Ten micrograms of plasmid was added to 400 μl of cells for 10 min at room temperature. Electroporation was performed on the GenePulser XCell system (Bio-Rad, Hercules, CA, USA) exponential protocol at 316 V and 500 μF and allowed to incubate at room temperature for 15 min. The cells were allowed to recover for 18–24 h before selection with G418 at 400 μg/ml. These stable cell lines were made once and maintained in G418 at 200 µg/ml after the initial selection period. Expression was confirmed by Western Blot for the myc tag after separation on 10% SDS-PAGE gels and by flow cytometry for surface expression after forward and side scatter gating with mean intensities of PE-conjugated BT3.1 staining vs. isotype control.

IFN-γ ELISA

Wild-type (WT) K562 cells, BTN3A1 KO K562 cells+empty vector, BTN3A1 KO+BTN3A1 (WT), BTN3A1 KO+BTN3A1 (ST²⁹⁶AA), and BTN3A1 KO+BTN3A1 (T³⁰⁴A) were resuspended in T-cell medium at 10⁵ cells/ml and treated with various concentrations of test compounds or antibody for 2 h. After incubation, the cells were washed 3 or 5 times and mixed with purified Vγ9Vδ2 T cells at an effector:target ratio of 3:1. Cell mixtures were incubated for 20–24 h at 37°C. IFN-γ was measured by ELISA, according to the manufacturer’s protocol (BioLegend). Experiments were performed at least 3 independent times with T cells from a minimum of 2 donors.

Protein purification

After sequencing, all the aforementioned recombinant clones were transformed into Escherichia coli BL-21 (DE3) cells and grown on an agar plate with antibiotics specific to the vector. A single colony was transferred to Luria-Bertani (LB) broth and grown overnight at 37°C to be transferred to LB or M9 minimal medium. For M9 minimal medium, ¹⁵NH₄Cl or [¹³C]-glucose both were added as the sole source of nitrogen and carbon. Cells were induced with 1 mM isopropyl β-d-1-thiogalactopyranoside, once they reached an optical density of 0.5, and were harvested after 4 h.

Recombinant BFI and B30.2 were purified with the protocol described in our previous study (29). Harvested cells overexpressing recombinant PPL^212–495 or PPL^495–865 were lysed by 3 passages through a French press (Thermo Fisher Scientific) in Tris-NaCl buffer (50 mM Tris and 100 mM NaCl, pH 7.5). Cell debris and insoluble precipitates were removed by centrifugation at 12,000 rpm at 4°C. The supernatant containing the His-tagged soluble protein was mixed with Ni-NTA resin (Qiagen, Germantown, MD, USA) for an hour, washed with 20 mM imidazole, and eluted in 400 mM imidazole in Tris-NaCl buffer. The eluate was finally buffer exchanged to 50 mM Tris (pH 8.0), 100 mM NaCl, and 5 mM 2-ME on a Superdex-75 column (GE Healthcare, Waukesha, WI, USA). Peak fractions containing the purified protein were pooled and concentrated to be used for further experimentation.

The JM region was purified under denaturing conditions from recombinant expression within inclusion bodies. A base buffer containing 8 M urea, 100 mM Tris, and 200 mM NaCl (pH 7.4) was used. Cells were resuspended in lysis buffer (base buffer +5 mM imidazole) and lysed by 3 passages through the French press (Thermo Fisher Scientific). The lysate was centrifuged at 12,000 rpm. Supernatant containing the denatured protein was mixed with Ni-NTA resin (Qiagen) at room temperature for 2 h. The resin was washed with 10 mM imidazole and eluted with 300 mM imidazole in base buffer. The eluate was loaded onto a Proto 300 C4 column (Higgins Analytical Inc., Mountain View, CA, USA) and purified in a 10–40% gradient formed by buffer A (90% H₂O, 10% acetonitrile, and 0.1% trifluoroacetic acid) and buffer B (90% acetonitrile, 10% H₂O, and 0.1% trifluoroacetic acid). Protein fractions were collected and lyophilized for later use. SDS-PAGE and Coomassie staining was performed to analyze the purity of the purified protein.

Small-angle X-ray scattering

Purified BFI in buffer containing 20 mM Tris, 100 mM NaCl, 5 mM 2-ME, and 5% glycerol (pH 7.5), in 3 concentrations ranging from 0.6 to 3 mg/ml, was used for small-angle X-ray scattering (SAXS) experiments. Duplicates of BFI samples were prepared with and without HMBPP. SAXS was performed at the Chess G1 station (Cornell University, Ithaca, NY, USA), equipped with a 9.8 keV 250 μm² X-ray beam. Data were collected on a Pilatus 100 K detector (Dectris, Baden-Daettwil, Switzerland) with a sample-to-detector distance of 1.5 m. Exposures of 15 s at 1-s intervals were taken for the protein and its corresponding buffer samples loaded into an in vacuo oscillating flow cell. Radiation damage was accessed by overlaying the first and the last exposure profile obtained from the sample data. Buffer measurements were performed twice, before and after the sample, to monitor changes in the background scattering. Two-dimensional scattering from the detector was converted to a one-dimensional scattering image as a function of momentum transfer s [s = 4π sin (θ)/λ, where 2π is the scattering angle and λ is the wavelength], with the software BioXTAS RAW and PRIMUS (32). Individual scattering profiles were averaged for both buffer and sample. Background scatter from the buffer was subtracted from the sample profile to obtain the final protein scattering image. The radius of gyration (R_g) was determined from Guinier analysis using AUTORG (33) in the low-q portion of the data, where R_g · q ≤ 1.3, and compared with the values calculated from pair–distance distribution functions [P(r)] using the program AUTOGNOM (34). Fifteen independent runs in DAMMIN (35) were performed, and the most probable model was computed with DAMAVER (36). The final envelope representation was obtained using Situs (37).

Isothermal titration calorimetry

Calorimetric measurements of BTN3A1 domains to the potential binding partners (including PPL and phosphoantigens) were performed on a low-volume Nano isothermal titration calorimetry (ITC) instrument (TA Instruments, New Castle, Delaware, USA). All experiments were performed at 25°C with a stirring speed of 200 rpm. A buffer composition of 50 mM Tris (pH 7.5), 100 mM NaCl, and 5 mM 2-ME was chosen for all ITC experiments, except those involving PPL where a buffer of 50 mM Tris (pH 8.0), 100 mM NaCl, and 5 mM 2-ME was used. For competitive binding experiments, the concentration of sodium diphosphate was 100 times the concentration of HMBPP. Each titration consisted of 20 2.5-µl injections at 300-s time intervals. Protein concentrations were determined by A₂₈₀. Titrants were prepared in the same buffer by dilution from a 100 mM stock or by concentrating proteins (PPL) with a Centricon filter (Millipore, Billerica, MA, USA), with the appropriate molecular weight cutoff. The concentrations of BTN3A1 ranged from 30 to 60 μM. The concentration of protein was examined after each ITC experiment to demonstrate good solubility (>85%) under the experiment conditions. The analysis of the data was done in NanoAnalyze Software (TA Instruments) suite using the “independent” model algorithm.

NMR spectroscopy

All NMR experiments were performed on a Varian Inova 600-MHz spectrometer (Agilent Technologies, Santa Clara, CA, USA) at 25°C. Protein–protein interactions were determined through ¹H-¹⁵N HSQC experiments collected at 2048 · 128 increments. The interaction between JM and HMBPP, DMAPP, GDP, and sodium diphosphate was accessed in buffer containing 100 mM MES, 50 mM NaCl, 5 mM 2-ME, and 8% D₂O at pH 6.5. HMBPP, DMAPP, GDP, and sodium diphosphate was added in a molar excess of 10 times the concentration of the JM region protein.

Spectra for backbone assignments [HNCACB, HNCA, HN(CO)CA, CBCA(CO)NH] were collected using 1024 · 32 · 16 increments in the direct, carbon, and nitrogen dimensions, respectively. HNCO experiments were executed with lower increments in the carbon dimension (n_i = 16). Experiments were performed in the presence of 5 mM dodecylphosphocholine (Avanti Polar, Alabaster, AL, USA). Peak picking was performed with the CcpNmr software suite (38), and chemical shifts of assigned peaks were used to predict the protein’s secondary structures through TALOS+ webserver (39).

RESULTS

HMBPP binding induces a conformational change to the globular intracellular domain of BTN3A1

We initiated our investigation by ascertaining the effects of the binding of HMBPP to BTN3A1, using SAXS. The full intracellular domain of BTN3A1 consisting of the JM region along with the B30.2 domain was used. SAXS allowed for the generation of a low-resolution structural envelope of BFI in the presence and absence of HMBPP (Fig. 2). In contrast to models that predict an extended helical conformation for the JM region (22), our SAXS data indicate relatively globular shapes for both the apo and ligand-bound BFI. This result could imply a partially unstructured JM region (Fig. 1), acting as a hinge that aids in the reorientation of B30.2.

Figure 2.

The BTN3A1 intracellular domain is a globular structure that condenses upon HMBPP binding. A) Scattering curve for BFI in the absence (green) and presence (blue) of HMBPP. B) Overlay of the normalized pair–distance distribution curves P(r) of the scattering curves. C) Surface representation of the average ab initio model for BFI and BFI+HMBPP rendered through SITUS. D) Rearrangement of the BFI region (JM+B30.2 domain) in the presence of HMBPP. Also shown is a model of the JM (red) and B30.2 domain (green/blue) fitted into the SAXS envelope.

Ab initio modeling was performed on the scattering curves shown in Fig. 2A. Overlay of the P(r) distance distributions obtained for BFI in the presence of HMBPP showed a shift toward a more compact structure (Fig. 2B). Data analysis revealed a reduction in R_g by ∼2 Å in the presence of HMBPP, further corroborated by an identical difference in the R_g obtained from the Guinier estimation. A more striking difference was noted for D_max, which was reduced by ∼8 Å. Taken together, these differences manifest into a smaller envelope for BFI in the presence of ligand. Figure 2C depicts the changes in the shapes of the ab initio envelopes for apo and ligand-bound states obtained using Situs (37). Overall, the SAXS results were consistent with the hypothesis that HMBPP binding to BTN3A1 induces a measurable conformational change to the intracellular domain (Fig. 2D).

The BTN3A1 JM region interacts with the HMBPP diphosphate moiety

Previously, we compared the effects of ligand binding on the BFI and B30.2 domains through NMR-based chemical-shift mapping experiments. In the presence of HMBPP, we observed the movement of additional residues in the ¹⁵N-HSQC spectrum of BFI that were not observed in the spectrum of the B30.2 domain, suggesting the involvement of the JM region in interacting with the ligand (29). This finding set the premise for the current work where we parsed out the molecular details from the JM region relevant for this interaction. We suspected the strong binding to the basic binding pocket in the B30.2 domain, identified through X-ray and NMR studies (27) would favorably position the ligand to allow for a weaker interaction between the ligand and the JM region. Because NMR is an excellent tool for studying weak interactions (40), we performed extensive NMR studies to investigate the interaction of the JM with antigenic and nonantigenic diphosphates.

Because of severe spectral overlaps, evaluation of the JM region within the BFI spectrum was not feasible. We worked around this challenge by investigating the 68-aa JM region in isolation. pH incompatibilities leading to uncontrollable precipitation diminished efforts to directly investigate the interaction between the JM region and the B30.2 domain. We were able to directly analyze the ability of the JM region to associate with HMBPP. Most of the JM amides in the ¹⁵N-HSQC spectrum were sharp and were within the random coil chemical shifts values (8.5–7.5 ppm), indicating a predominantly unstructured ensemble of conformers in fast exchange, with several broad peaks indicating an exchange in the intermediate regimen. HMBPP was titrated in 10-fold excess to facilitate the detection of weak binding. The JM spectra obtained in the presence and absence of HMBPP showed few perturbations, indicating a weak association of HMBPP with the JM region (Fig. 3A). The major effects were characterized by differences in line shapes, such as peak sharpening, rather than chemical shift changes (Supplemental Fig. 1), suggesting a dynamic nature of the ensemble with different conformers in exchange. One affected amide, marked by the asterisk in Fig. 3A and located within the expected positions of Ser/Thr residues, indicated that HMBPP binding selectively favored a specific conformation for this residue from the several conformations that were in intermediate exchange.

Figure 3.

The JM region interacts with HMBPP, DMAPP, and nonantigenic diphosphates. Major differences are marked by asterisks. A) HSQC of JM alone and with HMBPP at a 1:10 ratio, indicating weak association of HMBPP with JM. B) The overall effect of binding was nearly identical for JM:DMAPP, which implied the irrelevance of the hydroxyl group of HMBPP in binding to the JM region. C, D) Binding to the JM region was observed in response to the nonantigenic diphosphates GDP (C) and sodium diphosphate (D).

To better characterize the functional group that associates with the JM region, we used a series of molecules similar to HMBPP. DMAPP, a molecule identical to HMBPP without the allylic alcohol (Fig. 1D), displayed similar chemical shift perturbations in its ¹⁵N-HSQC spectrum as HMBPP (Fig. 3B), indicating the allylic alcohol in HMBPP is not essential for the interaction with the JM region. The JM region also weakly associated with GDP and sodium diphosphate (Fig. 3C, D), 2 ligands that do not activate BTN3A1. The common functional group among all the ligands was the negatively charged diphosphate, suggesting a possible interaction of the diphosphate moiety with the JM region. Comparable peak sharpening effects were also observed for the alternate molecules DMAPP, GDP, and diphosphate (Supplemental Fig. 1)

The interactions we observed by NMR are weak, with minimal spectral changes. This precluded the use of ITC to measure binding affinities directly to the JM region construct. To further confirm that the diphosphate associates with the JM region, we performed competitive binding assays by ITC (Supplemental Fig. 2 and Table 1), using the BFI and B30.2 constructs, which differ only by the presence or absence of the JM region. Binding thermodynamics for each construct were similar to our previously reported values (29). The addition to BFI of 100× molar excess of sodium diphosphate relative to HMBPP was able to reduce the ΔH from −53.7 to −32.9 kJ/mol, while becoming more entropically favorable as TΔS increased from −21.6 to −1.1 kJ/mol, suggesting that diphosphate alone competitively binds to BFI. In comparison, the addition to B30.2 of a 100× molar excess of sodium diphosphate relative to HMBPP showed a less pronounced difference in the ΔH (−53.7 vs. −38.4 kJ/mol) and TΔS (−23.0 vs. −7.7 kJ/mol). Thus, the effect of the diphosphate is lessened in the absence of the JM region, which indirectly substantiates that the JM is involved in the interaction with the diphosphate.

TABLE 1.

Summary of the thermodynamic data of HMBPP binding to BTN3A1 domains derived from ITC measurements in different conditions

BTN3A1 domain	ΔH (kJ/mol)	TΔS (kJ/mol)	Kd (μM)	n
BFI	−53.7	−21.6	2.36	0.845
BFI+diphosphate	−32.9	−1.1	2.61	0.860
B30.2	−53.7	−23.0	4.27	0.994
B30.2+diphosphate	−38.4	−7.7	4.10	1.070

The BTN3A1 JM region was incapable of direct binding to PPL spectrin repeats 1/2 or 3/4

A recent report suggested that the internal domain of BTN3A1 associates with PPL via a dileucine motif within the JM region to promote T-cell activation (21). To confirm this interaction, we examined the ability of PPL^212–495 (containing spectrin repeats 1 and 2) or PPL^495–865 (containing spectrin repeats 3 and 4) to bind with BFI using ITC (Supplemental Fig. 2). No binding was observed between the BTN intracellular domain and either PPL repeats 1–2 or 3–4. Further experiments were conducted with prior addition of HMBPP to BFI, mimicking the activated state of the receptor, followed by the titration with PPL^495–865, which also yielded negative results. A possible explanation for this outcome could be envisioned through the competition of B30.2 domain with PPL for binding to the JM region. Hence, we tested the ability of PPL to interact with the isolated JM construct, which again failed to register an association. Based on the above experiments, we were inclined to conclude the absence of a direct interaction between these PPL constructs and BTN3A1, suggesting either a possible need for additional binding partners or a different region of PPL that was not included in our construct design.

HMBPP alters the environment of Thr residues in the nonhelical center of the JM region

Based on our chemical-shift data, the perturbed residue marked by the asterisk in Fig. 3A hinted toward either a Ser or Thr residue that was most likely involved in the interaction. However, the sparing solubility of JM with a high propensity for aggregation, led to difficulties with executing triple-resonance NMR experiments (Supplemental Fig. 1). Therefore, we performed backbone assignments of JM in the presence of dodecylphosphocholine (DPC) micelles, which improved the overall solubility and stability of the peptide (Supplemental Fig. 3). Nevertheless, the degeneracy of the primary sequence with a large number of sequentially repeating amino acids (QQQEEKK, RKKKR, and so on), combined with the low solubility and conformational heterogeneity, deterred the complete assignment of all resonances. From partial backbone assignments, we were able to predict certain secondary structural features with the JM region (Supplemental Fig. 1). Coupled with the narrow chemical shift pattern from ¹⁵N-HSQC, we found that most of the JM maintained a random coil conformation, with the C terminus displaying secondary structural characteristics.

We next tested the effect of detergent molecules on the JM-HMBPP interaction and found that zwitterionic DPC competes with negatively charged diphosphate ligands for JM binding in a concentration-dependent manner. At a low JM:DPC ratio (1:40), several peaks broadened upon addition of HMBPP, whereas a higher ratio (1:120) showed no changes, possibly because of a saturating concentration of detergent micelles that occlude ligand binding (Supplemental Fig. 3). Based on the low JM:DPC ratio, we focused on identifying the residues involved in binding to HMBPP. The JM construct contains 3 Thr residues (at positions 279, 297, and 304; Fig. 1B) of which Thr²⁷⁹ was successfully assigned. HMBPP binding perturbed residues that were ambiguously assigned as Thr²⁹⁷ or Thr³⁰⁴. Because the spectral region in Fig. 3A displayed perturbation belonging to the Ser/Thr residues, we focused our attention on the Ser residues that precede the 2 Thr^297/304 residues. The Ser preceding Thr³⁰⁴ was conserved among BTN3A3, a BTN receptor that is unresponsive to phosphoantigen. Thus, we designed 2 mutants (S/T²⁹⁶AA and T³⁰⁴A) and probed their role in the binding of JM to HMBPP.

Thr mutants abolished binding of HMBPP to the JM region of BTN3A1

To confirm the relevance of the identified Thr residues, we purified JM peptides mutated at the specific Thr/Ser residues. As shown in Fig. 4A, B, both mutations clearly altered the NMR spectrum in the expected region, which allowed for the identification of Ser²⁹⁶, Thr²⁹⁷, Ser³⁰³, and Thr³⁰⁴. The labels pinpoint the position of each mutated residue in the WT spectrum that is identified through their absence in the mutant spectrum. The point mutations led to chemical-shift changes beyond sequentially adjacent residues, suggesting their involvement in maintaining a local conformation within the JM region.

Figure 4.

Specific mutations perturb the JM Thr spectra and abolish the ability of HMBPP to interact with the BTN3A1 JM region. Based on the chemical shift changes, peaks corresponding to the mutated serine/threonine residues were identified, where solid lines represent confident assignment while dashed represent lower confidence. A) ¹⁵N-HSQC comparison of JM and its ST²⁹⁶AA mutant. B) ¹⁵N-HSQC comparison of JM and its T³⁰⁴A mutant. C, D) ¹⁵N-HSQC spectra of HMBPP titrated into ST²⁹⁶AA (C) and T³⁰⁴A (D) mutant at a 10:1 HMBPP:JM ratio showed no perturbation and hence a loss of binding capacity.

We next assessed the interaction of HMBPP with the mutated JM constructs and found that neither of the 2 mutant constructs was able to bind to HMBPP, as shown by the lack of perturbations in the ¹⁵N-HSQC spectral overlay after ligand titration (Fig. 4C, D). Taken together, both Thr residues were deemed important for the binding of HMBPP to the JM region of BTN3A1.

Thr mutations in the BTN3A1 JM hinder the ability of phosphoantigen-containing cells to activate T cells

We then assessed the role of these residues in cellular activity assays (Fig. 5). Plasmids containing the empty vector, the WT BTN3A1, or the ST²⁹⁶AA or T³⁰⁴A mutations were transfected into K562 cells that were deficient in functional BTN3A1 (31). The 3 constructs expressed at the expected molecular weight and at similar levels, as demonstrated by Western blot analysis and expressed at similar surface levels, albeit slightly lower than parental cells, as demonstrated by flow cytometry (Supplemental Fig. 4). The transfected K562 cells were tested for their ability to trigger IFN-γ production in primary Vγ9Vδ2 T cells in response to preloading with varied HMBPP concentrations according to the timeline shown in Fig. 5A. As expected, parental K562 cells showed the strongest stimulation of IFN-γ production by the Vγ9Vδ2 T cells among all conditions (Fig. 5B). Two-way ANOVA demonstrated that the treatment conditions significantly differed, with the BTN3A1-deficient cells displaying the reduced effects of HMBPP. The WT BTN3A1 construct produced a strong rescue of the effects of BTN3A1 depletion, and both Thr mutants gave intermediate phenotypes.

Figure 5.

Point mutations in the JM region show loss of function in response to HMBPP stimulation but not antibody stimulation. A) Timeline for expansion, purification, and assay of primary human Vγ9Vδ2 T cells. B) Quantification of IFN-γ secretion after 2 h exposure to HMBPP and 20 h incubation with effector T cells (n = 3). Final concentrations of HMBPP were 3-fold dilutions between 500 pM and 10 µM. C) Quantification of IFN-γ secretion after 2 h exposure to POM₂-C-HMBP and 20 h incubation with effector T cells (n = 3). Final concentrations were 1, 10, 33, and 100 nM. D) Quantification of IFN-γ secretion after 2 h exposure to zoledronate and 20 h incubation with effector T cells (n = 3). Final concentrations were 1, 10, 33, and 100 µM. E) Quantification of IFN-γ secretion after 2 h exposure to HMBPP, followed by additional wash steps, and 20 h incubation with effector T cells (n = 3). Final concentrations of HMBPP were 3-fold dilutions between 100 nM and 10 µM. POM₂-C-HMBP (100 nM) and zoledronate (100 µM) were used as positive controls. F) IFN-γ secretion following 2 h exposure to anti-human BTN3A1 and 20 h incuvation with effector T cells (n = 3). Final concentrations were 0, 0.01, 0.1, and 1 µg/ml. Letters on the graphs indicate significantly different group results, as determined by 2-way ANOVA. P < 0.05.

At high doses of HMBPP, some residual activation occurred, likely because of BTN3A1 being present in the T-cell population of this coculture leading to autoactivation by residual HMBPP remaining after the wash steps. As BTN3A1 seems to be essential for activation, we examined 2 different ligands in this assay: POM₂-C-HMBP (29) (Fig. 5C) and zoledronate (Fig. 5D). Both of these ligands induced a similar pattern of activity, though with a stronger response and with less background activity. Furthermore, additional wash steps (5 total) were able to reduce the activity of HMBPP in the knockout cells to baseline (Fig. 5E). The extra wash steps did not reduce the activity of POM₂-C-HMBP or zoledronate, indicating that the effect was related to removal of HMBPP rather than loss of cells. Taken together, both ST²⁹⁶AA and T³⁰⁴A have reduced ability to trigger activation of Vγ9Vδ2 T cells in response to treatment with intracellular ligands. In contrast, when cells were stimulated with a BTN3 antibody, which functions as a partial agonist by binding to the extracellular domain, no significant differences were observed in the mutants (Fig. 5F).

DISCUSSION

The functions and antigen-specific activation of γδ T cells have been mysterious until now (41, 42). BTN3A1, an integral membrane protein within the B7 family, has garnered therapeutic interest for its unique role in responding to phosphoantigens, enabling the cell to be recognized by Vγ9Vδ2 T cells and targeted for elimination. Although it is now clear that BTN3A1 ligands, including HMBPP, bind to the surface pocket of the BTN3A1 B30.2 domain, the exact molecular details of the signal transduction mechanism, required for activating Vγ9Vδ2 T cells, remain elusive.

Here, we present direct evidence that HMBPP binding induces a conformational change within the full intracellular domain of BTN3A1. The SAXS surface envelope model depicts that the apo intracellular domain exists in an elongated, but not fully extended state. It then compacts upon complexing with HMBPP, as demonstrated by the noticeable ∼2-Å reduction in the radius of gyration. This significant conformational change was not observed in the crystal structure of the isolated B30.2 domain upon phosphoantigen binding (27). In fact, the calculated backbone RMSDs between the apo and complexed B30.2 (4N7I vs. 4N7U) are almost nonexistent. The lack of conformational change in the crystal structure could be related to methodology, where the protein was crosslinked. Although this structure is useful in determining the binding site, the conformation observed may not be an accurate representation of the activated ligand-bound state of the receptor. Therefore, it is possible that the change we observed in our SAXS model was caused by either a conformational change within the B30.2 itself or the relative orientation/positioning of the B30.2 and the JM region, or a combination of both.

Furthermore, our NMR results demonstrate a weak association between the JM region and HMBPP. Previously reported data showed that the binding stoichiometry of HMBPP:BFI is 1:1 (29). This result is consistent with the possibility of 1 molecule of HMBPP interacting with residues within both the B30.2 and JM region through different points of contact. This notion is also consistent with the B30.2/C-HMBPP crystal structure. A closer look at the electron density of the bound ligand only shows a well-defined tetrahedral moiety suitable for fitting the β-phosphate functional group, suggesting that the other portion of HMBPP was flexible as it was not visualized by X-ray crystallography (27). This finding could be related to nonoptimal cross-linked conformation of the apo B30.2 used for crystallization as discussed above, which was not able to promote additional contacts, such as with H381 or other nearby residues (43), including the ones from the JM region. Our data suggest that the JM region may coordinate with the second exposed phosphate given that DMAPP, which lacks the allylic alcohol, GDP and even simple diphosphate still show the same weak association with the JM region.

The exact amino acids of the JM region that are involved in interacting with HMBPP were not defined. However, we have determined that Thr³⁰⁴, which is located roughly in the center of the JM sequence, was close to the binding interface. Based on the overall length and flexibility of the JM region, we think it is possible that the B30.2 domain is moved toward the JM region and closer to the membrane upon ligand binding, as suggested by our SAXS data (Fig. 2).

Although this finding is contrary to the various predicted models (Fig. 1C) that show that the JM region is highly helical and therefore rigid, we have found experimentally that the JM region is predominantly random and flexible, at least in the form of a separate domain (or as a constituent of the BFI in our previous study). This conclusion is based on our partial assignments and the overall lack of chemical shift dispersion in ¹⁵N-HSQC spectra, with only the C terminus likely to possess helical features. Because it is unlikely that the JM region is highly helical, the probability that it adopts a coiled-coil arrangement is low. This conclusion is substantiated by our SAXS data, which show that the probability of a dimeric coiled-coil structure is minimal, at least in the absence of the transmembrane linker. Also, the repeated HxxHCxC motif or a tryptophan/leucine zipper, which is typical of dimeric coiled coils, is not found in the JM sequence. Therefore, we believe that the JM region exists in a flexible state naturally. However, once the B30.2 clamps to it upon binding HMBPP, it may very well cause the JM region to adopt more restricted helical features, although in a manner consistent with compaction of the overall domain. We envisioned that the JM region forms additional contacts with the B30.2 domain, but this theory was not validated experimentally because of buffer incompatibilities of the isolated B30.2 and JM region.

Based on the evidence and reasoning discussed, it is clear that a conformational change occurs and that it is most likely caused by the clamping of the B30.2/HMBPP/JM complex. The main question then arises as to how this finding is relevant to the role of BTN3A1 and how it may affect the activation of Vγ9Vδ2 T cells. This important concern is addressed in part by our cellular assays of BTN3A1 mutants. The mutants ST²⁹⁶AA and T³⁰⁴A showed no perceptible association with HMBPP in vitro, which we believe would then abolish or weaken the likelihood that the intracellular domain compacts upon HMBPP binding. Indeed, the cellular assays showed a reduction in BTN3A1 activity in response to 3 different phosphoantigens, validating that the JM region is essential for Vγ9Vδ2 T-cell activation and that both Ser/Thr^296/297 and Thr³⁰⁴ are important. Our NMR data show the involvement of Thr³⁰⁴ in binding HMBPP. The role of Ser/Thr^296/297, on the other hand, could lie in its ability to stabilize some local fold or its involvement in forming a binding interface with B30.2.

Nonetheless, the 2 Thr mutations resulted in the reduced activity of BTN3A1, emphasizing that the induced conformational change was a key step toward the activation of Vγ9Vδ2 T cells. The results point toward 2 possibilities that could ensue. First, the conformational change within the intracellular domain could traverse to the extracellular domain, leading to its recognition by T cells, consistent with the conclusions from a recent report where the fluorescent resonance energy transfer approach was used to show that bisphosphonate treatment alters the extracellular conformation of BTN3A1 (30). Second, the conformational change could lead to the recruitment or dissociation or both of intracellular binding partners—for example, RhoB (30) or PPL (21)—eventually facilitating T-cell activation. Our in vitro experiment did not confirm the binding of PPL to BTN3A1; further cellular analysis is needed to assess the possibility of their interactions. Furthermore, both Ser/Thr^296/297 and Thr³⁰⁴ fall within the box 2 sequence (aa 288–308) of the BTN3A1 juxtamembrane region, which was recently reported to be necessary for full activation by nitrogenous bisphosphonates (44).

In summary, our studies indicate the importance of the BTN3A1 JM region for the receptor interaction with isoprenoid diphosphates and changes in its activation state induced by ligand binding. It will be interesting to find out how the structural changes can influence the functions of BTN3A1 and other BTNs, as our understanding of their multifaceted roles in T-cell activation and other immune signaling pathways (e.g., ref. 45) continues to evolve. A logical next step to be explored is the evaluation of the full-length BTN3A1. We present the analysis of conformational changes within the intracellular domain, but it remains to be seen how the effects propagate across membranes and to the extracellular surface. Application of membrane mimetic systems such as nanodiscs (46, 47) may provide additional insights into the unique features of the mechanism of BTN3A1 activation and the potential for cooperativity at the receptor level (43), upon binding to different phosphoantigens.

Supplementary Material

This article includes supplemental data. Please visit http://www.fasebj.org to obtain this information.

Glossary

BFI

BTN3A1 full intracellular domain

BTN

butyrophilin

DMAPP

dimethylallyl diphosphate

DPC

dodecylphosphocholine

FBS

fetal bovine serum

HMBPP

(E)-4-hydroxy-3-methyl-but-2-enyl diphosphate

ITC

isothermal titration calorimetry

JM

juxtamembrane

KO

knockout

POM₂-C-HMBP

bis (pivaloyloxymethyl) (E)-4-hydroxy-3-methyl-but-2-enyl phosphonate

PPL

periplakin

SAXS

small-angle X-ray scattering

WT

wild type

AUTHOR CONTRIBUTIONS

K. Nguyen, J. Li, and R. Puthenveetil performed the research, analyzed the data, and wrote the paper; X. Lin and C. -H. C. Hsiao performed the research and analyzed the data; M. M. Poe performed research; O. Vinogradova designed research, analyzed the data, and wrote the paper; and A. J. Wiemer designed and performed the research, analyzed the data, and wrote the paper.