Role of V-ATPase a3-Subunit in Mouse CTL Function

Cytotoxic T lymphocytes are an essential part of the adaptive immune system and clear the body of virally infected or tumorigenic cells through the release of cytotoxic components at the cell–cell contact zone, the immune synapse (IS). The IS compartmentalizes the pore-forming protein, perforin, and degradative enzymes, the granzymes, which are released by CTLs and deployed unidirectionally across the synapse to puncture and kill the target cell (1, 2). Prior to release, the cytotoxic components are stored in specialized organelles, cytotoxic granules (CG). CG share many features with lysosomes and are thus often termed secretory lysosomes or lysosome-related organelles (2–6). In contrast to classical lysosomes, CG are dual function organelles that combine degradative function with secretory function (7). Their acidic pH of ∼5.5 (8, 9) enables not only the degradation of proteins through hydrolases (10) but is also required to process precursors of granzymes and perforin to their mature form (11, 12). Some studies also report that acidification of the granules is necessary to keep perforin initially inactive, whereas others state that low pH may increase perforin’s cytolytic activity (2, 13).

The acidification of tissue environments and organellar lumen is achieved via the vacuolar-type (H⁺)–adenosine triphosphatase (V-ATPase or proton pump), which moves protons across biological membranes (14–16). The V-ATPase is a large, complex, membrane-spanning, ATP-driven rotary pump that comprises two multisubunit domains: the transmembrane (V_o) and cytosolic (V₁) domains. The V-ATPase complexity is further increased by the regulated expression of several isoforms and splice variants of individual subunits (17, 18), and its activity is regulated at many levels, including the separation of V_o and V₁ domains for individual regulation and trafficking (19). The specific localization of the V-ATPase to distinct cellular compartments is determined by the expression of the a-subunit of the V_o domain. For example, the a1-subunit is expressed on synaptic vesicles (20, 21), the a2-subunit on vesicles like Golgi and early endosomes (22), the a3-subunit on the plasma membrane of osteoclasts (23, 24), and the a4-subunit on the plasma membrane of renal intercalated cells (25–27).

Considering the importance of pH for CG and CTL function, it is surprising that the identity of the a-subunit localized on CG is not known. Despite intensive studies of the V-ATPase in other tissues, and even in other secretory granules (28), its role in the critical acidic compartment of cell–cell killing has remained unknown. In this study, we characterized both CG lumen acidity and V-ATPase location and function throughout the maturation and transport of CG and their docking and fusion at the IS. We show that three a-subunits, a1, a2 and a3, are expressed in CTLs, but that the acidic pH of ∼6 inside the CG lumen is generated exclusively by V-ATPase complexes containing the a3-subunit. Knockdown of the a3-subunit leads to reduced killing of target cells, reduced fusion of CG, reduced number of CG at the IS, reduced interaction with the microtubule network during IS formation, and reduced condensation of the dense core in individual CG. Thus, we conclude that acidification of CG is important for multiple maturation steps and function of CG.

C57BL/6 mice were purchased from the Jackson Laboratory, and the generation of the synaptobrevin2-mRFP knock-in (SybKI) mouse was reported before (29). Granzyme B (GzmB)–monomeric teal fluorescent protein (mTFP) knock-in (KI) mice were generated by CRISPR/Cas9 technology using guide RNA 5′-GTC CAG GAT TGC TCT AGG AC-3′ (PAM = AGG). For inserting mTFP in frame with exon 5 of GzmB, a homology directed repair fragment was designed using sense wild-type (wt) primer 5′-CAA CAG CTC AGT GCC TTG TAT CCA-3′ or sense KI primer 5′-ACC GCA TCG AGA TCC TGA ACC-3′ and antisense wt primer 5′-TTC ACA AGG ACC AGC TCT GT-3′.

The pH and chloride sensor ClopHensorN was purchased from Addgene (50758). ClopHensorN was cloned in the pMAX vector by digesting the pCDNA3-ClopHensorN plasmid with HindIII and blunted the DNA with DNA polymerase 1, large (Klenow) fragment followed by digestion with NotI. The same was performed for pMAX vector, and both the vector and insert were ligated. ClopHensorN(Q69M) construct was generated by mutating glutamine (aa 69 in E²GFP) to methionine by using forward primers 5′-ATG TAT ACT AGC TAG CTG GAG CCA CCC GCA GTT C-3′and 5′-ACG GCG TGA TGT GCT TCA-3′ and reverse primers 5′-TGA AGC ACA TCA CGC CGT-3′and 5′-ATG TAT ACG CGG ATC CGC GCT TGT ACA GCT CGT CCA T-3′. Citrine was amplified from D1ER (kind gift from A. Cavalie) using forward primer 5′-ATG TAT ACC GGA ATC CAT GGT GAG CAA GGG CGA G-3′ and reverse primer 5′-ATG TAT ACG CGG ATC CTT ACT TGT ACA GCT CGT CCA T-3′ and cloned in pMAX vector (30). GzmB-citrine was cloned by replacing mCherry at the C terminus of GzmB with forward primer 5′-ATG TAT ACG CGG ATC CAC CGG TCG CCA CCA TGG TGA GCA AGG GCG AG-3′ and reverse primer 5′-ATG TAT AAA AGC GGC CGC TTA CTT GTA CAG CTC GTC-3′. Citrine-TdTomato was cloned by replacing E²GFP from pMAX-ClopHensorN plasmid with TdTomato by using forward primer 5′-ATG TAT ACC GGA ATT CGG TGG GAG CGG CGG AAG CGG CGG TAA GCT TAT GGT GAG C-3′and reverse primer 5′-ATG TAT ACG CGG ATC CCT TGT ACA GCT CGT CCT T-3′. Citrine–TagRFP-T was cloned by amplifying TagRFP-T with forward primer 5′-ATG TAT ACG CAC CGG TAT GGT GTC TAA GGG CGA A-3′ and reverse primer 5′-ATG TAT ACC GCT CGA GTT ACT TGT ACA GCT CGT C-3′. mt-ClopHensorN(Q69M) was cloned by amplifying a mitochondrial signal sequence with forward primer 5′-ATG TAT ACG GGG TAC CGC CAC C ATG TCC GTC CTG ACG CCG-3′ and reverse primer 5′-ATG TAT ACC GGA ATT CCC CCA ACG AAT GGA TCT T-3′. The following primary Abs were used: anti-a1 (ab105937; Abcam), anti-a2 (ab96803; Abcam), anti-a3 (polyclonal; Thomas Jentsch, Leibniz-Institut für Molekulare Pharmakologie, Berlin, Germany), anti–SNAP-23 (111203; Synaptic Systems), anti-GAPDH (14C10, 2118L), (2368; Cell Signaling Technology), HRP-conjugated anti–Strep-tagII (2-1509-001; IBA Lifesciences), and anti-mitochondria (ab3298; Abcam). Secondary Abs were HRP-conjugated donkey anti-rabbit and Alexa647 goat anti-mouse IgG (H + L) (Thermo Fisher Scientific). For structured illumination microscopy (SIM), an Alexa647-coupled anti–GzmB (GB11; BioLegend) Ab was used. For total internal reflection fluorescence microscopy (TIRFM), a hamster anti-mouse CD3ε (clone145-2C11; BD Pharmingen) Ab was used for coating coverslips and stimulating cells.

For functional effector/target cell conjugations, p815 target cells were incubated with 30 μg of anti-CD3ε Ab for 15–30 min at 37°C and washed with 1× PBS, and 20,000 cells were plated on the 12-mm glass coverslips. Carefully, 0.2 × 10⁶ CTLs were added on top of settled target cells. After 15 min, cells were fixed with 4% paraformaldehyde and stained with indicated Abs.

An in vivo pH calibration curve in primary CTLs was generated using the ratio of the fluorescence intensity with illumination at 488 nm to that at 458 nm of the E²GFP moiety of ClpH. The AIMV medium was removed and replaced with Ringer solution, and the cells were placed under the microscope. The extracellular solution was then replaced with a pH-clamping solution. Solutions with pH values of 4, 4.5, 5, 5.5, 6, 6.5, 7, 7.5, 8 and 8.5 were made as follows. The base solution consisted of 30 mM NaCl, 100 mM KCl, 2 mM MgCl₂, and 10 mM glucose, as well as 10 μM nigericin, 4 μM valinomycin, and 5 μM CCCP. Depending on the target pH, lactate (20 mM [pH 4]), MES (20 mM [pH 5 and 6]), or HEPES (20 mM [pH 7–9]) were included. After allowing equilibration of the intracellular pH (>5 min), images were acquired with sequential illumination at 488, 458, and 561 nm with closed pinhole. The entire cell was considered for ratiometric calibration. Ratios of fluorescence versus pH values were fitted using a single binding site Hill model in Igor Pro (WaveMetrics). For pH measurements, the coverslips were placed in the observation chamber, filled with extracellular solution, and imaged at room temperature (RT). Images were acquired at 488, 458, and 561 nm. Ratios of fluorescence at 488 and 458 nm were calculated, and the pH was interpolated from the calibration-curve fit. Solutions with chloride concentrations of 0, 10, 20, 50, 70, 90, 100, 120, and 140 mM were made as follows. The desired [Cl⁻]_i and pH_i were controlled by equilibrating extra- and intracellular ion concentrations using the K⁺/H⁺ exchanger nigericin (5 μM), the protonophore carbonyl cyanide p-chlorophenylhydrazone (CCCP) (5 μM), the K⁺ ionophore valinomycin (5 μM) and the Cl⁻/OH⁻ exchanger tributyltinchloride (10 μM) in the presence of high-K⁺ 20 mM HEPES buffer containing 0.6 mM MgSO₄, 38 mM sodium gluconate, and 100 mM potassium gluconate. The specified amount of gluconate anion was replaced by Cl⁻, and the pH was adjusted with small amounts of NaOH.

To conjugate CTLs with P815 target cells, CTLs were mixed with target cells at a 10:1 ratio and plated onto 0.01% poly–l-ornithine or 30 μg anti-CD3ε coated glass coverslips and incubated at 37°C for 5, 10, and 15 min. Cells were fixed in ice-cold 4% PFA in Dulbecco 1× PBS (Thermo Fisher Scientific) and stained with Abs. The SIM setup was an Elyra PS.1 obtained from Carl Zeiss Microscopy, Jena, Germany. Images were acquired with a 63× Plan-Apochromat (NA 1.4) with laser excitation at 488, 561, and 635 nm and then processed to obtain higher resolutions (Zen 2012; Carl Zeiss Microscopy). For analysis of colocalization, Pearson and Manders coefficients of correlation (31, 32) were determined using the JACoP plugin of ImageJ v1.46 (33).

The TIRFM setup from Visitron Systems (Puchheim, Germany) was based on an IX83 microscope body (Olympus) equipped with the Olympus autofocus module, a UAPON100XOTIRF NA 1.49 objective (Olympus), a 488-nm 100-mW laser emitting at 561 nm, the iLAS2 illumination control system (Roper Scientific, Evry, France), the evolve-EM 515 camera (Photometrics), and a filter cube containing Semrock (Rochester, NY) FF444/520/590/Di01 dichroic and FF01-465/537/623 emission filters. The setup was controlled by Visiview software (version 4.0.0.11; Visitron Systems). CTLs isolated from GzmB-mTFP KI mice were electroporated with either nonsilencing small interfering RNA (ns-siRNA) (control) or small interfering (siRNA) 1 or 2 against a3-subunit. For other experiments, CTLs were transfected with GzmB-mCherry. After 12–16 h of transfection, 0.3 × 10⁶ CTLs were resuspended in 30 μl of extracellular buffer (2 mM HEPES, 140 mM NaCl, 4.5 mM KCl, and 2 mM MgCl₂) containing no Ca²⁺ and allowed to settle for 1–2 min on anti-CD3ε Ab (30 μg/ml) coated coverslips. Cells were then perfused with extracellular buffer containing 10 mM Ca²⁺ for CG fusion. Depending on the constructs used and the experimental conditions, cells were imaged for 7 min at RT either at 561 or 488 nm wavelength alternating between both illuminations. Unless specified otherwise, acquisition frequency was 10 Hz. Images and time-lapse series were analyzed using ImageJ (34) or the FIJI package of ImageJ. CG fusion analysis was performed using ImageJ with the plugin Time Series Analyzer. A sudden drop in GzmB-mTFP or GzmB-mCherry fluorescence occurring within 300 ms (three acquisition frames) was defined as fusion (35).

Splenocytes were isolated from 8- to 18-wk-old C57BL6/N, SybKI mice or granzyme B–mTFP KI mice, as described before (30). Briefly, naive CD8⁺ T cells were isolated from splenocytes using Dynabeads FlowComp Mouse CD8⁺ Kit (Thermo Fisher Scientific). The isolated CD8⁺ T cells were cultured at a 1:0.8 ratio for up to 14 d (37°C, 5% CO₂) with anti-CD3/anti-CD28 activator beads (Thermo Fisher Scientific) at a density of 1 × 10⁶ cells per milliliter in AIMV (Thermo Fisher Scientific) containing 10% FCS, 100 U/ml recombinant mouse IL-2, and 50 mM 2-ME to generate effector CTLs. P815 target cells were cultured in RPMI 1640 medium (Thermo Fisher Scientific) containing 10% FCS, 1% penicillin/streptomycin, and 10 mM HEPES.

Total RNA was extracted with TRIzol (Thermo Fisher Scientific) and reverse transcribed (SuperScript II; Thermo Fisher Scientific) using random hexamer primers. Semiquantitative PCR was performed using specific a1 forward primer 5′-TCA GTA CCT GAG GAA GAA GC-3′ and reverse primer 5′-CTG GTG GAC CAT GGT GTC GC-3′, a2 forward primer 5′-CGA GAA GTG ACG TGT GAG GA-3′ and reverse primer 5′-ACT GAA CTT GGA GGA GAG CA-3′, a3 forward primer 5′-CGA ACC ACC TGA GCT TTC TC-3′ and reverse primer 5′-CCC ATG GAA GAG CAG ATG AT-3′, and a4 forward primer 5′-GCA GTG CAT CAT TGC CGA GAT C-3′ and reverse primer 5′-GAA CAT AGG CTG GAC ACT CCA AG-3′. GAPDH was used as loading control using forward primer 5′-ACC ACA GTC CAT GCC ATC AC-3′ and reverse primer 5′-TCC ACC ACC CTG TTG CTG TA-3′. Data were normalized to GAPDH.

Mouse CTLs were homogenized with a syringe in lysis buffer (50 mM Tris [pH 7.4], 1 mM EDTA, 1% Triton X-100, 150 mM NaCl, 1 mM DTT, 1 mM deoxycholate, protease inhibitors, and PhosSTOP; Roche) on ice. Lysates were rotated for 10 min at 4°C, and insoluble material was removed by centrifugation at 15,000 × g. The protein concentration was determined using Quick Start Bradford 1× Dye Reagent (5000205; Bio-Rad Laboratories). Proteins were separated by SDS-PAGE (NuPAGE; Thermo Fisher Scientific), transferred to nitrocellulose membranes (Amersham), and blocked by incubation with 5% skim milk powder in 20 mM Tris, 0.15 M NaCl (pH 7.4), and TBS for 1–2 h and blotted with specific Abs. Blots were developed using ECL reagents (SuperSignal West Dura Chemiluminescent Substrate; Thermo Fisher Scientific) and scanned. For expression analysis, the area and mean fluorescence intensity were determined with ImageJ v1.46.

CTLs from day 3 to 12 were used for transfections. An amount of 5 × 10⁶ CTLs were transfected with 1–2 μg of plasmid DNA (Mouse T Cell Nucleofector Kit, VPA-1006; Lonza). Cells were imaged 10–12 h after transfection. For knockdown of protein expression, 3–5 × 10⁶ CTLs were transfected with 2 μM of siRNAs (control siRNA: 5′-UUC UCC GAA CGU GUC ACG UTT-3′, anti-a1 siRNA1: 5′-AAG GTC ATT TAC AAT TTG CTA-3′, anti-a2 siRNA1: 5′-CTG GGT AGA ATT TCA GAA CAA-3′, anti-a3 siRNA1: 5′-CAT GCT CAC CCT GAA CCC TAA-3′, anti-a3 siRNA2: 5′-CTG GCC ATG GTC CTC ACT TGA A-3′) were purchased from QIAGEN (Mm_Atp6v0a1_3, Mm_Atp6v0a2_2, Mm_Tcirg1_6, Mm_Tcirg1_5, SI00907249, SI00907270, SI05141577, and SI05141570; QIAGEN). Knockdown efficiency was assessed by preparing whole-cell lysates of 1 × 10⁶ CTLs boiled in SDS-loading buffer containing 4% 2-ME and 1 mM DTT. Lysates were sonicated, proteins were separated by SDS-PAGE, and expression was analyzed by Western blotting.

The killing assay was performed as previously described (30). Briefly, p815 target cells were loaded with calcein-AM (500 nM; Life Technology) in serum-free AIMV for 15 min at RT, washed once with PBS, and plated into 96-well black plates with clear bottoms (BD Falcon). Target cells were lysed with 0.1% Triton X-100 to calculate maximum target cell lysis. Sixteen to eighteen hours after transfection with ns-siRNA, anti-a1 siRNA, anti-a2 siRNA, anti-a3 siRNA1, or anti-a3 siRNA2, CTLs were added to target cells (10:1 or 20:1 ratio; 0.2 × 10⁶ cells per well) to measure killing at 37°C for 4 h. Readings were measured at 485-nm excitation wavelength and 535-nm emission wavelength by GENios Pro plate reader. The fluorescence for the experimental conditions was adjusted by the parameter γ according to the live target cell control fluorescence. The γ value was measured at time zero: γ = F_{live (0)}/F_{exp (0)}. The cytotoxicity was calculated from the loss of calcein fluorescence in target cells using the following equation: percentage of target cell lysis = (F_live − γ × F_exp)/(F_live − F_lyse) × 100%. Abbreviations in the equation are as follows: fluorescence of only target cell controls (F_live), CTLs with target cells (F_exp), and maximum target lysis (F_lyse). All experiments were performed in duplicates from three independent preparations.

Isolated mouse CD8⁺ cells from GzmB-mTFP KI mouse 5 d after activation by anti-CD3/anti-CD28 beads were treated with either anti-a3 siRNA1 or siRNA2. Sixteen hours after transfection, 4000 CTLs in AIMV with 30% FCS were seeded onto poly–l-ornithine (0.1 mg/ml)–coated 1.4-mm sapphire discs in flat specimen carriers (Leica). The cells were allowed to settle for 20 min at 37°C with 5% CO₂. Samples were vitrified in a high-pressure freezing system (Leica EM PACT2/RTS) in AIMV with 30% FCS. All samples were further processed in an automatic freeze-substitution apparatus (AFS2; Leica) as described in (29). In brief, all samples were transferred into the precooled (−130°C) freeze-substitution chamber of the AFS2. The temperature was increased from −130 to −90°C over 2 h. Cryosubstitution was performed from −90 to −70°C over 20 h in anhydrous acetone and from −70 to −60°C over 20 h with 0.3% (w/v) uranyl acetate in anhydrous acetone. The samples were infiltrated with increasing concentrations (30, 60, and 100% for 1 h each) of Lowicryl (3:1 K11M/HM20 mixture) with 0.3% uranyl acetate. After 5 h of infiltration with 100% Lowicryl, samples were UV polymerized at −60°C for 24 h and for additional 15 h while the temperature was raised linearly to 5°C. Ultrathin sections were cut using a Leica EM UC7 and collected on pioloform-coated copper grids. After contrasting with uranyl acetate and lead citrate sections were analyzed with a Tecnai12 Biotwin electron microscope (Thermo Fisher Scientifics).

Cells were prepared for high-pressure freezing as described before, but samples were vitrified in a high-pressure freezing system (Leica EM PACT2/RTS) 5 min after addition of CellMask deep red (1:1000; Invitrogen) in AIMV for plasma membrane staining. After freeze substitution, the samples were kept in the dark at 4°C until further processing. One hundred– to one hundred and twenty–nanometers ultrathin sections were cut by using an EM UC7 (Leica). The sections were collected on carbon-coated 200 mesh copper grids (Plano). Fluorescence analysis was performed within 1 d after sectioning to avoid loss of fluorescence signals. For correlative fluorescence and electron microscopy, the grids were place in a drop of water between two coverslips, sealed with silicone (picodent twinsil). High resolution SIM (ELYRA PS.1; Carl Zeiss Microscopy) images were acquired by using the 63× Plan-Apochromat (NA 1.4) objective with excitation light of 488- and 642-nm wavelengths and processed. After imaging in the bright-field mode for grid orientation, CellMask deep-red image (642 nm) was recorded to identify both the image plane and the outline of CTLs by plasma membrane staining. In a z-stack analysis three to eight anti-RFP488 images (488 nm) and CellMask deep-red images (642 nm) were recorded with a step size of 100 nm to scan the cells of interest. For data acquisition and image processing for higher resolution, ZEN 2013 software (Carl Zeiss Microscopy) was used. After fluorescence imaging, the same grids were stained with 2% uranyl acetate and lead citrate and recorded with a Philips Tecnai12 Biotwin electron microscope (Thermo Fisher Scientifics). Only CTLs with well-conserved membranes, cell organelles, and nuclei were analyzed and used for correlation. For correlation, the CellMask deep-red image that shows the labeled plasma membrane of the cells was used to find the best overlap with the electron microscope image. The final alignment with the anti-RFP488 image defines the position of the fluorescent signal within the cells of interest. Images were overlaid in Corel Draw.

Confocal and stimulated emission depletion (STED) images of microtubules and CG were acquired on a commercial STED microscope (Abberior Instruments, Göttingen, Germany). Activated GzmB KI T cells were labeled with 5 μM silicon rhodamine (SiR)–tublin for 1 h at 37°C in AIMV medium with 10% FCS and placed on anti-CD3 coated glass coverslips for live imaging at RT in AIMV medium. All images were acquired within 10–15 min after placing on the coated coverslips. Images were recorded with a 100× oil immersion objective (NA 1.4; Olympus UPLSAPO 100XO), excitation wavelengths 485 nm for mTFP granules and 640 nm for SiR-tubulin, detection windows 498–520 and 650–720 nm (respectively), and pixel size 40 nm. STED beams of 595 and 775 nm were used. Images were linearly deconvolved (Wiener filtered) with Matlab.

To measure the distance between granules and microtubules, a line was drawn perpendicular to the microtubule and through the CG. The absolute fluorescence intensity values along the line were measured in both channels using the FIJI plot profile. Using Excel, we determined for each channel the position on the line where the fluorescence signal was maximum. Then we calculated the distance between the two specific positions in each channel.

Statistical significance of differences was calculated with Student t test for normally distributed data. For nonnormally distributed data the statistical test used is indicated in the figure legends. Data were analyzed with ImageJ v1.46 (34), Excel (Microsoft), and SigmaPlot 13 and graphed using Affinity Designer Software (Serif).

There is no specific sensor to accurately measure pH in the CG lumen. While the GFP-derivative citrine is reported to be pH-sensitive, ClopHensorN, a genetically encoded ratiometric pH and Cl⁻ sensor, has been developed for measurements in the cytoplasm of cells (36). However, these were not suitable for our purposes, as calibration curves generated upon transfection of these constructs into primary CTLs revealed suboptimal results. Citrine had a small dynamic range for pH changes, and ClopHensorN had a better dynamic range but a pKa value of 7.2 unsuitable for pH measurements in acidic compartments (Supplemental Fig. 1A, 1B). Therefore, we mutated glutamine at position 69 of E²GFP in ClopHensorN to methionine to lower the pKa value (37) (Fig. 1A). After verification by sequencing, we fused this construct to the C terminus of GzmB to localize it to the CG lumen (GzmB-ClopHensorN [Q69M]; Fig. 1A). Primary murine CTLs were then transfected with GzmB-ClopHensorN(Q69M), fixed, and stained with an anti–GzmB Ab coupled to Alexa647 to test for localization to CG. SIM revealed a punctate-staining pattern for GzmB-ClopHensorN(Q69M) (Fig. 1B; green and red channel) and a complete overlap with the signal of the anti-GzmB Ab, which detects both the transfected and endogenous GzmB (Fig. 1B; magenta). Quantitative analysis by Pearson coefficient of correlation (Fig. 1C; 0.70 ± 0.04; n = 16) and Manders overlap coefficient (Fig. 1D; 0.71 ± 0.03 [GzmB to GzmB-ClopHensorN(Q69M)] and 0.84 ± 0.03 [GzmB-ClopHensorN(Q69M)] to GzmB; n = 16) confirmed these findings. We further tested the specific localization and function of ClopHensor by targeting it to mitochondria through addition of an N-terminal mitochondrial signal sequence and measuring the pH in mitochondria as 7.57 ± 0.87 (Supplemental Fig. 1C–F). Having confirmed the specific localization of GzmB-ClopHensorN(Q69M) to the CG lumen, we next determined the pH and Cl⁻ concentration inside the CG. For that purpose, we measured the ratio of fluorescence signal at 488 and 458 nm for pH and 458 and 561 nm for [Cl⁻], respectively, and compared it to the respective in vivo calibration curves at these ratios (Fig. 1E, 1F; see 2Materials and Methods for details). We determined a pH_CG of 6.11 ± 0.28 (n = 137; mean ± SD) and an intralumenal [Cl⁻]_CG of 73.78 ± 0.12 mM (n = 171; mean ± SEM). Thus, our newly generated GzmB-ClopHensorN(Q69M) sensor allowed us to accurately measure the pH and [Cl⁻] inside CG of CTLs in vivo.

To determine which a-subunit of the V-ATPase is responsible for localization to the CG and for establishing the acidic pH in the CG lumen, we first performed RT-PCR analysis of primary, activated CTLs. We could detect the RNA of all except the a4-subunit (Fig. 2A). We then designed siRNAs against the a1-, a2-, and a3-subunits and transfected them into CTLs one at a time together with the GzmB-ClopHensorN(Q69M) sensor. Because the a-subunits differ considerably in their protein sequence (only between 47 and 62% identity) and we selected unique sequences for the selection of siRNAs (Supplemental Fig. 2), we can assume that the knockdown of individual a-subunits was highly selective. After confirming the efficient knockdown of the respective a-subunit by Western blot with isoform-specific Abs (>75%; Fig. 2B–D), we measured the pH inside the CG lumen. We found that knockdown of the a1- or a2-subunit did not change the acidic pH of CG (Fig. 2E). In contrast, knockdown of the a3-subunit led to a 10-fold reduction of the [H⁺] inside CG (from pH ∼6 to pH ∼7; n = 137; ***p < 0.001; Fig. 2E) compared with CG from control CTLs transfected with an ns-siRNA. We also stained CTLs isolated from SybKI cells, which endogenously express a specific marker for CG coupled to mRFP (29), with a polyclonal Ab against the a3-subunit and found significant colocalization of CG with a3-subunit with a Pearson coefficient of correlation of 0.64 ± 0.08 (Fig. 2F). We next tested the effect of a3-subunit knockdown on CTL function by performing a calcein-based killing assay (38). CTLs transfected with ns-siRNA or siRNA against a1-, a2-, or a3-subunits were incubated with p815 target cells at two different CTL/target cell ratios (10:1 and 20:1). We found that knockdown of a1- and a2-subunit had no effect on target cell lysis and CTLs killed target cells with a similar efficiency as wt CTLs and CTLs transfected with ns-siRNA (Fig. 2G). In contrast, knockdown of the a3-subunit by either of the two anti-a3 siRNAs led to a highly significant reduction in target cell killing at both ratios (n = 3; ***p < 0.001 [t test]; Fig. 2G).

Thus, we conclude that the a3-subunit of the V-ATPase is localized on the membrane of CG, that complexes containing this subunit are required for establishing a mildly acidic pH of ∼6 in the lumen of CG, and that its knockdown leads to reduction in killing efficiency.

We next tested the possibility that the lack of acidification of CG through knockdown of the a3-subunit alters the biogenesis or the morphology of CG. For this, we first transfected either ns-siRNA or a3-subunit knockdown-specific siRNA into SybKI cells, fixed the cells 16–18 h after transfection, and performed immunocytochemistry with an anti-GzmB Ab coupled to Alexa488. We then verified that the colocalization of GzmB and synaptobrevin2 is not altered by the knockdown of the a3-subunit by quantifying the number of double-positive CG via SIM imaging of entire cells. There was no significant difference between the colocalization of GzmB and synaptobrevin2 in cells treated with ns-siRNA and a3-subunit–specific siRNA1 with Pearson correlation coefficient of 0.71 ± 0.03 and 0.67 ± 0.02, respectively (Fig. 3A). In addition, we found that knockdown of the a3-subunit by two different siRNAs did not lead to a change in the total number of CG per CTL (12.59 ± 1.08 [n = 87] for ns-siRNA, 12.61 ± 1.14 [n = 83] and 12.76 ± 1.26 [n = 76] for KD1 and KD2, respectively; Fig. 3B), ruling out that less granules are being made or that a faster degradation of CG occurs. To test for morphological changes of individual CG we recorded electron microscopy images of CTLs transfected with either ns-siRNA or two different anti-a3 siRNAs (KD1 and KD2). Because activated CTLs have numerous intracellular organelles (5, 39), we first identified CG by using endogenous GzmB-mTFP as a marker in correlative fluorescence light and electron microscopy (Supplemental Fig. 3). We then analyzed the electron microscopy images for CG electron density and diameter (Fig. 3C). Quantification revealed a significant reduction in CG electron density (184.9 ± 6.5 arbitrary units [a.u.], n = 18 in ns-siRNA versus 123.9 ± 4.0 a.u., n = 39 in KD1 and 121.7 ± 5.1 a.u., n = 19 in KD2; ***p < 0.001, Fig. 3D) coupled with a significant increase in CG diameter (220.7 ± 12.8 nm, n = 18 in ns-siRNA versus 353.5 ± 14.6 nm, n = 39 in KD1 and 350.5 ± 19.8 nm, n = 19 in KD2, ***p < 0.001; Fig. 3E). The observed phenotype in electron microscopy argued for a lack of condensation of the dense core, which, in CG, is almost exclusively composed of serglycin (40, 41). Because serglycin-deficient CG show, besides a loss of electron density and an increased diameter, an altered expression of perforin and GzmB (40), we tested the levels of both proteins in Western blot experiments. Consistent with previous reports (8, 40, 42) we found a downregulation of perforin and an upregulation of GzmB in CTLs transfected with a3-subunit specific siRNA or treated for 2 h with 100 nM of the general V-ATPase blocker concanamycin A (42) (Fig. 3F–H).

To travel to the IS, CG are being transported along the microtubule network (43). We asked whether this transport is disturbed upon knockdown of the a3-subunit. We therefore transfected CTLs with ns-siRNA or two different anti–a3-subunit siRNAs (KD1 and KD2), labeled CG with GzmB-mTFP, plated the transfected CTLs on anti-CD3 coated coverslips to induce an artificial IS and stained the microtubule network with SiR-tubulin. We then performed STED microscopy and followed the transport of CG in real time (Fig. 4A–C). Analysis of fluorescence-intensity profiles revealed that knockdown of the a3-subunit resulted in an increased distance between microtubules and CG (representative examples are shown in Fig. 4D–F), whereas the centrosome polarization was unchanged. Considering a CG diameter of 312 nm and a microtubule diameter of 15 nm (5, 35), we used a 40-nm bin size to obtain a distance distribution of CG to microtubules (Fig. 4G–I). Assuming that CG-microtubule interaction must occur within close proximity (i.e., sum of the radii of CG and microtubules; red line in Fig. 4G–I), we found that 66% of CG were within close proximity of microtubules in control CTLs transfected with ns-siRNA (Fig. 4J). In contrast, only 6.9 and 14.5% of CG remained in close proximity upon knockdown of the a3-subunit with two different siRNA (Fig. 4J), further supporting the conclusion that the a3-subunit of V-ATPase is directly involved in CG trafficking. Because pulldowns with Rab7 and Rab27 did not precipitate the a3-subunit (Supplemental Fig. 4), a potential interaction with the tubular network is most likely not mediated by a direct interaction between these proteins.

The observed reduction in killing efficiency upon knockdown of the a3-subunit could have any one of several causes or some combination of factors. The same number of CG might fuse at the IS; however, the loss of acidic pH might lead to reduced cytotoxic potency of individual granules. Alternatively, the number of CG fusing at the IS might be reduced, probably because of a reported direct interaction of the a3-subunit with synaptobrevin2, the v-SNARE on CG-mediating fusion with the plasma membrane at the IS of mouse CTLs (29, 44). To distinguish between these possibilities, we transfected primary mouse CTLs from a GzmB–mTFP KI mouse, in which CG are endogenously labeled, with either of the two siRNAs against the a3-subunit or an ns-siRNA. After plating CTLs on anti-CD3–coated coverslips, we followed fusion of individual CG by TIRFM (Fig. 5A). In comparison with CTLs transfected with ns-siRNA (47.1 ± 2.9%, n = 98) we observed an ∼5-fold reduction of cells showing CG fusion upon knockdown of the a3-subunit with either siRNAs (10.9 ± 1.5%, n = 161; ***p < 0.001 for a3-subunit siRNA1 and 11.0 ± 0.7%, n = 143; ***p < 0.001 for a3-subunit siRNA2; Fig. 5B). Accordingly, the mean number of fusion events per cell was reduced as well (4.2 ± 0.3, n = 98 for ns-siRNA versus 2.2 ± 0.2, n = 161, **p < 0.01 for a3-subunit siRNA1 and 2.3 ± 0.2%, n = 143, **p < 0.01; Fig. 5C). Thus, the observed reduction in killing efficiency upon a3-subunit knockdown (Fig. 2F) was due in part to a reduced number of fusing CG at the IS. We tested whether the reported interaction with synaptobrevin2 (VAMP2) is responsible for this effect. However, pulldown assays from CTLs transfected with a fusion protein for synaptobrevin2 (Syb2-Twin-Strep-tag) followed by Western blotting with anti–a3-subunit Ab did not reveal any interaction of these two proteins (Fig. 5D). Because we could detect SNAP-23, the t-SNARE functioning as a binding partner for synaptobrevin2 in fusion of CG, as a specific band in our positive control, we conclude that the proposed interaction between synaptobrevin2 and the a3-subunit is not responsible for the observed reduction in CG fusion (Fig. 5D). Rather, we found that, in comparison with wt CTLs or CTLs transfected with ns-siRNA, the number of CG accumulating at the IS over time was significantly reduced by a factor of ∼4 upon a3-subunit knockdown (after 7 min: 16.1 ± 2.4 for untreated and 16.0 ± 2.9 for ns-siRNA versus 4.7 ± 0.7 for a3-subunit siRNA1 and 4.6 ± 0.9 for a3-subunit siRNA2; ***p < 0.001; Fig. 5E). Taken together, these data demonstrate that the observed reduction in CG fusion (Fig. 5B, 5C) and target cell killing (Fig. 2F) upon a3-subunit knockdown is due to a reduced number of CG arriving at the IS caused by impaired transport along the microtubule network.

In this study, we have identified the a3-subunit of the V-ATPase as being responsible for localizing the V-ATPase to CG of primary, murine CD8⁺ lymphocytes and for establishing and maintaining an acidic pH of 6.1 (Figs. 1, 2). Specific knockdown of the a3-subunit leads to a neutralization of the lumen of CG. Our data demonstrate that this leads to multiple effects: 1) an increase in granule diameter coupled with a reduction in dense-core condensation (Fig. 3), 2) a loss of interaction of CG with the microtubule network (Fig. 4), 3) a reduced number of fusion events at the IS (Fig. 5), and 4) a strongly reduced ability of CTLs to kill target cells (Fig. 2).

Although the V-ATPase is a large, multisubunit complex with a molecular mass of >830 kDa (17, 45), it is well established that its a-subunit is responsible for localizing the V-ATPase to specific compartments of the cell (16, 46, 47). Of the four isoforms expressed in mammalian cells, we identified the transcripts of three of them, a1–a3, being present in activated CTLs (Fig. 2A). Although in some organelles like insulin-containing granules of pancreatic β cells some a-subunits can compensate for the loss of other a-subunits in terms of acidification (48), we found that knockdown of the a3-subunit selectively neutralized the CG lumen, whereas knockdown of a1- or a2-subunit did not (Fig. 2E). Along with the colocalization of the a3-subunit with CG markers (Fig. 2F) and given the specificity of the siRNAs (Supplemental Fig. 2), we can conclude that the a3-subunit is the only a-subunit localizing to CG, facilitating its potential use in immunotherapy (see below).

Our electron microscopy studies revealed an increased diameter and a decreased density of individual CG upon a3-subunit knockdown (Fig. 3). In CG, the dense core mostly consists of chondroitin sulfate A type proteoglycans, specifically serglycin (41, 49–51). Lack of serglycin leads to reduced cytotoxicity of NK cells and CTLs, and a complete loss of mature CG, characterized by a lack of the dense core (40). Serglycin interacts with both perforin and GzmB and is proposed to promote the processing and safe storage of these cytotoxic components in the acidic environment of the CG lumen. In accordance with published data, we find a reduction in perforin levels and an increase in GzmB levels upon neutralization through a3-subunit knockdown (Fig. 3F–H), indicating that the observed change in morphology contributes to the observed reduction in cytotoxicity (Fig. 2G).

Given the enormous size of the V-ATPase it is not surprising that numerous interactions with other proteins have been reported (17). It has been reported that the V-ATPase is directly involved in the vesicular fusion with the plasma membrane (52), probably through physical interaction with the v-SNARE synaptobrevin2 (44). However, our direct measurements of individual CG fusion events by TIRFM (Fig. 5A–C) and the lack of a3-subunit precipitates in synaptobrevin2-pulldown experiments (Fig. 5D) argue against this role. Rather, the reduction in cytotoxicity and fusion frequency upon a3-subunit knockdown appears to be due to a reduced transport of CG along the microtubules (Figs. 4, 5E). Monomeric G proteins of the Rab family have been implicated in regulating trafficking of lysosomes through interactions with the V-ATPase in various organisms and cell types (53, 54). We tested for a potential interaction of the a3-subunit with Rab7 and Rab27 in pulldown assays and were unable to detect an interaction, even with mutants arresting the Rabs in the GDP- or GTP-bound form (Supplemental Fig. 4). However, it remains possible that the interaction is indirect (e.g., through Rab-interacting proteins) (55, 56) or mediated through other G proteins (e.g., of the ARF family) (57, 58).

Given the potential use of a3-subunit modulators in immunotherapy against cancer and osteopetrosis (16, 17), our structural and functional data obtained in this study might support the development of new targets for treatment of these diseases.

We thank Margarete Klose, Anja Bergsträßer, Tamara Brück, Nicole Rothgerber, and Katrin Sandmeier for excellent technical assistance and Michael L. Dustin for giving valuable comments on the manuscript.

The authors have no financial conflicts of interest.