ORAI2 modulates store-operated calcium entry and T cell-mediated immunity

Store-operated Ca2+ entry (SOCE) through Ca2+ release-activated Ca2+ (CRAC) channels is critical for lymphocyte function and immune responses. CRAC channels are hexamers of ORAI proteins that form the channel pore, but the contributions of individual ORAI homologues to CRAC channel function are not well understood. Here we show that deletion of Orai1 reduces, whereas deletion of Orai2 increases, SOCE in mouse T cells. These distinct effects are due to the ability of ORAI2 to form heteromeric channels with ORAI1 and to attenuate CRAC channel function. The combined deletion of Orai1 and Orai2 abolishes SOCE and strongly impairs T cell function. In vivo, Orai1/Orai2 double-deficient mice have impaired T cell-dependent antiviral immune responses, and are protected from T cell-mediated autoimmunity and alloimmunity in models of colitis and graft-versus-host disease. Our study demonstrates that ORAI1 and ORAI2 form heteromeric CRAC channels, in which ORAI2 fine-tunes the magnitude of SOCE to modulate immune responses.

C a 2 þ release-activated Ca 2 þ (CRAC) channels mediate Ca 2 þ influx in many cell types. They are formed by the tetraspanning plasma membrane proteins ORAI1, ORAI2 and ORAI3. These ORAI proteins mediate Ca 2 þ influx by store-operated Ca 2 þ entry (SOCE), so named because it depends on the filling state of intracellular Ca 2 þ stores 1 . Upon cell stimulation through receptors that are linked to phospholipase C and the production of IP 3 , Ca 2 þ is released from the endoplasmic reticulum (ER) via the opening of IP 3 receptors. The reduction in the ER Ca 2 þ concentration is followed by activation of two transmembrane proteins located in the ER membrane, stromal interaction molecule 1 (STIM1) and STIM2 (ref. 1). Dissociation of Ca 2 þ from the EF hand domains of STIM1 and STIM2 results in conformational changes that enable them to bind and open CRAC channels in the plasma membrane 1 .
CRAC channels are hexameric complexes composed of individual or potentially multiple ORAI homologues 2 . Structure-function analyses of Drosophila Orai channels as well as human and mouse ORAI1 have shown that they constitute the pore of the CRAC channel [2][3][4] . The transmembrane domains are highly conserved between all three ORAI homologues. The first transmembrane domain of ORAI1 lines the channel pore and contains a glutamate residue that constitutes a high-affinity Ca 2 þ binding site and confers strong Ca 2 þ selectivity to the CRAC channel 2,3,5-7 . All three ORAI homologues can function as Ca 2 þ channels when overexpressed 8,9 . The properties of ectopically expressed ORAI1, ORAI2 and ORAI3 channels are similar to those of endogenous CRAC channels 10,11 , including activation by Ca 2 þ store depletion, high Ca 2 þ selectivity, inward rectification and Ca 2 þ -dependent inactivation (CDI) 1 . However, the three ORAI homologues differ in some of their channel properties, including fast and slow CDI and their sensitivity to pharmacological inhibitors such as 2-aminoethoxydiphenyl borate 8,9 . CRAC currents (I CRAC ) and SOCE have been reported in many cell types, consistent with the ubiquitous expression of all three ORAI homologues 1,[12][13][14] . Transcript levels of murine Orai1 are particularly high in immune cells and those of Orai2 are high in the brain, lung, spleen and small intestine, whereas Orai3 mRNA is abundant in many solid organs 1,13,15,16 . Expression of Orai2 is also reported in platelets, melanocytes, B cells, dendritic cells, macrophages and mast cells 13 . ORAI1 is the best-characterized ORAI homologue in terms of its physiological functions, whereas less is known about ORAI2 and ORAI3. Patients with null or loss-of-function mutations in ORAI1 present with a complex disease syndrome, CRAC channelopathy, which is characterized by immunodeficiency, autoimmunity, muscular hypotonia and ectodermal dysplasia because ORAI1 has critical functions in T cells, skeletal muscle cells, dental enamel-forming cells and eccrine sweat glands [17][18][19] . ORAI1 is of particular importance for T cell function as emphasized by the lack of CRAC currents and SOCE in T cells of patients with null or loss-of-function mutations in ORAI1 (refs 17,20,21). The mutations cause a severe combined immunodeficiency-like disease characterized by impaired T cell proliferation, reduced cytokine production, abolished antibody responses and severe viral and bacterial infections 17,18 . In mice, the dependence of CRAC channel function on ORAI1 appears to be less pronounced as naive T cells from Orai1 À / À mice and Orai1 R93W knock-in mice (that express a non-functional ORAI1 p.R93W protein) have residual SOCE [22][23][24] , reduced but not abolished cytokine production and the ability (upon differentiation into proinflammatory T helper 1 (T H 1) and T H 17 cells) to cause experimental autoimmune encephalomyelitis 25,26 . These findings suggest that residual SOCE and T cell function in the absence of ORAI1 may be mediated by ORAI2 and/or ORAI3. Additional functions of ORAI1 in smooth muscle cells, endothelial cells, platelets, mast cells and secretory cells have been described 19,23,[27][28][29][30] . ORAI2 and ORAI3, by contrast, are not as well-defined functionally due to the lack of patients with null mutations, gene-deficient mouse models and selective inhibitors of individual ORAI homologues. Whereas ORAI3 has been shown to mediate SOCE in breast, lung and prostate cancer cells and to promote their growth and invasiveness 31 , the physiological role of ORAI2 is unclear.
Here we show that ORAI2 is highly expressed in naive T cells, but downregulated in effector T cells, resulting in an increased ORAI1:ORAI2 ratio and stronger ORAI1 dependence in effector T cells. Whereas genetic deletion of Orai1 reduces SOCE and CRAC currents, deletion of Orai2 enhances both. These distinct effects are explained by the ability of ORAI2 to form multimeric channel complexes with ORAI1, in which ORAI2 attenuates steady-state CRAC channel currents likely due to its altered inactivation properties compared to ORAI1. Using Orai1-or Orai2-deficient mice, we find that ORAI1 and ORAI2 mediate SOCE and T cell function. Only combined, but not individual, deletion of Orai1 and Orai2 abolishes SOCE completely and interferes with protective antibody responses against viral infection, and prevents autoimmune and alloimmune inflammation in models of inflammatory bowel disease (IBD) and graftversus-host disease (GvHD), respectively. Collectively, our data demonstrate that ORAI1 and ORAI2 form heteromeric CRAC channel complexes, in which ORAI2 modulates the magnitude of SOCE to control T cell-mediated immune responses.

Dynamic regulation of ORAI1 and ORAI2 expression in T cells.
To understand the contributions of the three ORAI homologues to immunity, we analysed the expression of Orai1, Orai2 and Orai3 genes in various immune cell subsets using public databases (Immgen.org; BioGPS.org: gene 109305 (Orai1), 269717 (Orai2) and 269999 (Orai3)). All three homologues were expressed broadly in immune cells with the highest levels of Orai1 and Orai2 found in granulocytes. Orai2 was expressed above average in T cells compared to most other immune cells ( Supplementary Fig. 1a). The analysis of Orai1, Orai2 and Orai3 mRNA expression in FACS-sorted thymic and splenic CD4 þ and CD8 þ T cell subsets from wild-type (WT) mice by quantitative real-time PCR (qRT-PCR) confirmed that Orai1 and Orai2 are highly expressed in all mature T cell subsets compared to Orai3 ( Supplementary Fig. 1b).
To investigate the role of ORAI2 in T cells, we generated Orai2 À / À mice by replacing the protein-coding exons of Orai2 with a LacZ reporter ( Supplementary Fig. 2a,b). X-gal (5-bromo-4-chloro-3-indolyl b-D-galactopyranoside) staining of E17.5 heterozygous Orai2 LacZ/ þ embryos showed Orai2 expression in many organs including skin, kidneys and spleen ( Supplementary  Fig. 2c). FDG (fluorescein di-[b-D-galactopyranoside]) staining of isolated immune cells demonstrated high Orai2 expression in bone marrow-derived macrophages (BMDMs) ( Supplementary  Fig. 2d), thymocytes and splenic T cells (Fig. 1a,b). Orai2 expression was induced in thymocytes of Orai2 LacZ/ þ mice from the CD4 À CD8 À to CD4 þ CD8 þ stage, consistent with Orai2 mRNA expression in FACS-sorted thymocytes ( Supplementary  Fig. 1b). In peripheral T cells, Orai2 expression was highest in CD44 lo CD62L hi naive T cells and lower in CD44 hi CD62L lo effector cells (Fig. 1a,b). Downregulation of Orai2 was confirmed at the mRNA level ( Fig. 1c and Supplementary Fig. 1b). In contrast, Orai1 mRNA expression was lower in naive than effector CD4 þ T cells ( Fig. 1c and Supplementary Fig. 1b), resulting in a higher Orai1:Orai2 expression ratio in effector versus naive CD4 þ T cells (Fig. 1d). The Orai1:Orai2 mRNA expression ratio also increased after activation of naive CD4 þ T cells in vitro to generate effector T cells (Fig. 1e). By contrast, the levels of Orai3 mRNA remained unchanged (Fig. 1c). Together, these results show that T cells coexpress ORAI1 and ORAI2, especially in naive T cells, whereas ORAI1 expression dominates in effector T cells, which may contribute to the modulation of SOCE and T cell function.
ORAI2 controls the magnitude of SOCE in naive T cells. Homozygous Orai2 LacZ/LacZ (hereafter called Orai2 À / À ) mice lacked Orai2 mRNA and protein expression ( Supplementary  Fig. 2e,f). Orai2 À / À mice were born at normal Mendelian ratios without gross morphological or histological abnormalities. To determine the role of ORAI1 and ORAI2 in T cells, we analysed T cells from Orai1 fl/fl Cd4cre mice that lack Orai1 in all T cells 25 , Orai2 À / À mice, and Orai1 fl/fl Orai2 À / À Cd4cre mice, whose T cells lack both ORAI proteins. When SOCE was induced by depletion of ER Ca 2 þ stores with thapsigargin, which inhibits sarcoplasmic/endoplasmic Ca 2 þ ATPases, ablation of Orai1 reduced SOCE in naive CD4 þ and CD8 þ T cells by B30-50% (Fig. 2a,b and Supplementary Fig. 3a,b) consistent with previous results in T cells from Orai1 À / À (refs 22,25,26,32) and Orai1 R93W knock-in mice 24 . Unexpectedly, naive CD4 þ and CD8 þ T cells from Orai2 À / À mice had markedly increased SOCE compared to WT cells, whereas combined deletion of both Orai1 and Orai2 completely abolished SOCE (Fig. 2a,b and Supplementary Fig. 3a,b). To determine whether increased Ca 2 þ levels in Orai2 À / À T cells were due to increased SOCE, we analysed Ca 2 þ influx in Mn 2 þ quench experiments. The quench rate of Fura-2 fluorescence in naive CD4 þ T cells from Orai2 À / À mice was significantly higher than in WT T cells stimulated with thapsigargin in the presence of extracellular Mn 2 þ , indicating increased SOCE in Orai2 À / À T cells (Fig. 2c). Increased SOCE was observed in naive Orai2-deficient T cells, but not in in vitro differentiated effector T cells. Orai2 À / À effector T cells had SOCE comparable to WT cells, whereas deletion of Orai1 resulted in a 460% reduction of SOCE ( Supplementary  Fig. 4a,b) consistent with results in Orai1 À / À (refs 22,25,26,32) and Orai1 R93W mice 24 . The smaller effect of Orai2 deletion in effector T cells is in line with its reduced expression in effector T cells. These findings suggest that ORAI2 is part of the CRAC channel complex in naive (but not effector) T cells and exerts an inhibitory effect on SOCE.
Increased SOCE was not restricted to Orai2 À / À CD4 þ and CD8 þ T cells (Fig. 2a,b and Supplementary Fig. 3a,b), but was also observed in Orai2-deficient BMDMs (Fig. 3a,b) and BM-derived dendritic cells (BMDCs) ( Supplementary Fig. 3c-f). These findings point to a universal role of ORAI2 in attenuating ORAI1-mediated CRAC channel function. This effect is conserved in human fibroblasts because short hairpin RNA (shRNA)-mediated knockdown of ORAI2 caused increased SOCE when compared to fibroblasts transduced with scrambled shRNA (Fig. 2d-f). By contrast, knockdown of ORAI1 in human fibroblasts reduced SOCE consistent with previous observations in fibroblasts from ORAI1-deficient patients 20 . We investigated whether increased SOCE in the absence of ORAI2 is due to   compensatory upregulation of ORAI1. ORAI1 mRNA levels were comparable to those in control fibroblasts (Fig. 2d). Furthermore, cell surface expression of ORAI1 was unchanged in shORAI2-treated fibroblasts (Fig. 2g). Likewise, Orai1, Orai3, Stim1 or Stim2 mRNA levels were not altered in primary Orai2 À / À T cells ( Supplementary Fig. 3g)   findings show that increased SOCE in the absence of ORAI2 is not due to compensatory upregulation of ORAI1 or other CRAC channel components but rather indicate a negative regulatory function of ORAI2.
Orai2 deletion in macrophages enhances CRAC channel function.
To understand the mechanisms responsible for increased SOCE in the absence of ORAI2, we investigated BMDMs that, like naive T cells, have high ORAI2 expression (Supplementary Fig. 2d-f). Deletion of Orai2 resulted in elevated SOCE in ionomycinstimulated BMDMs over a range of extracellular Ca 2 þ concentrations ( Fig. 3a,b). It has been suggested that ORAI2 may function as a Ca 2 þ release channel in the ER 33 . Orai2 deletion might therefore result in increased filling of ER Ca 2 þ stores and potentially affect activation of CRAC channels. Depletion of Ca 2 þ stores with ionomycin in the absence of extracellular Ca 2 þ , however, showed only a very moderate increase in Ca 2 þ store content in Orai2 À / À compared to WT BMDMs (Fig. 3a,b) arguing against ORAI2 as an ER release channel. Furthermore, the Mn 2 þ quench rate of Fura-2 fluorescence was significantly higher in Orai2 À / À than WT BMDMs supporting the idea that ORAI2 inhibits SOCE through ORAI1 channels (Fig. 3c). We next analysed CRAC currents (I CRAC ) in Orai2-deficient BMDMs to determine how ORAI2 affects I CRAC (Fig. 3d,e). WT BMDMs had currents consistent with the properties of I CRAC including an inwardly rectifying current-voltage (I-V) relationship, permeation of Na þ seen in a divalent-free (DVF) solution, and blockade by La 3 þ . In Orai2 À / À BMDMs, these CRAC currents were significantly larger, but otherwise retained the characteristic features of I CRAC in WT BMDMs including inwardly rectifying I-Vs in 20 mM extracellular Ca 2 þ and DVF solutions or La 3 þ blockade (Fig. 3d,e). Enhanced I CRAC in Orai2 À / À BMDMs (Fig. 3e) provides direct evidence that the deletion of ORAI2 enhances CRAC channel function in BMDMs. The fact that the properties of I CRAC in Orai2 À / À BMDMs were indistinguishable from those of ectopic ORAI1 expression 8,9 suggests that enhanced I CRAC in Orai2 À / À BMDMs is mediated by ORAI1. It is noteworthy that measurements of I CRAC in naive T cells from WT and Orai2 À / À mice were not successful due to their small whole-cell currents. Together, we show that ORAI2 attenuates CRAC channel function in BMDMs and naive T cells, presumably by replacing the more efficient ORAI1 subunit in heteromeric ORAI1:ORAI2 complexes.
ORAI1 and ORAI2 form heteromeric channels. We next tested if ORAI2 inhibits I CRAC and SOCE by forming heteromeric CRAC channel complexes with ORAI1. We used primary CD4 þ T cells from Orai1 fl/fl Cd4cre mice, which have reduced SOCE that is mediated by ORAI2 and retrovirally transduced them with either empty vector (EV), WT ORAI1 (ORAI1 WT ) or the non-functional pore-mutant ORAI1 E106Q that is expressed in the plasma membrane but unable to conduct Ca 2 þ (ref. 5). We hypothesized that if ORAI1 and ORAI2 form heteromeric complexes, ectopic expression of ORAI1 E106Q together with endogenous ORAI2 results in ORAI1 E106Q :ORAI2 channels that are unable to conduct Ca 2 þ . Expression of ORAI1 WT (but not EV) strongly enhanced SOCE in Orai1-deficient T cells, whereas ORAI1 E106Q completely suppressed residual ORAI2-mediated SOCE (Fig. 4a). To confirm this finding, we retrovirally transduced CD4 þ T cells from Orai2 À / À mice with either EV, ORAI2 WT or the non-functional pore-mutant ORAI2 E80Q (ref. 14). Orai2-deficient T cells transduced with EV had strong SOCE mediated by homomeric ORAI1 channels. Expression of ORAI2 WT in Orai2 À / À T cells markedly reduced SOCE consistent with the idea that ORAI2 inhibits ORAI1 function, whereas expression of ORAI2 E80Q completely abolished SOCE (Fig. 4b). Taken together, these data show that pore-dead mutants of one ORAI isoform suppress the function of the other isoform, likely because ORAI1 and ORAI2 form heteromeric complexes. An alternative explanation is that overexpression of the ORAI1 E106Q mutant in Orai1-deficient T cells sequesters the majority of STIM1 required for activation of endogenous ORAI2 (and likewise expression of ORAI2 E80Q in Orai2 À / À T cells sequesters STIM1 necessary for ORAI1 activation). We therefore mutated leucine (L) 273 in the C terminus of ORAI1 WT and ORAI1 E106Q to aspartate (D) to abolish ORAI1 binding to STIM1 (refs. 34,35). ORAI1 L273D was expressed at the plasma membrane ( Fig. 4c) but completely abolished SOCE in T cells from Orai1 fl/fl Cd4cre mice compared to EV-transduced cells (Fig. 4e,g). SOCE suppression was comparable to that found in Orai1-deficient cells transduced with ORAI1 E106Q (Fig. 4a,d,g) or the ORAI1 E106Q/L273D double mutant that can neither conduct Ca 2 þ nor bind to STIM1 (Fig. 4f,g). These findings argue against STIM1 sequestration by pore mutant ORAI proteins as the cause of suppression of SOCE (Fig. 4a,b), but instead support the model that ORAI1 and ORAI2 form heteromeric channels (Fig. 4h). The fact that ectopic expression ORAI1 L273D in T cells of Orai1 fl/ fl Cd4cre mice almost completely inhibited SOCE is consistent with a report demonstrating that STIM1 binding to all subunits of hexameric ORAI complexes is required for channel activation 36 .
To understand if the functional properties of heteromeric ORAI1:ORAI2 channels differ from ORAI1 or ORAI2 homomeric channels, we analysed I CRAC properties in HEK293 cells that coexpressed ORAI1 and/or ORAI2 together with STIM1. Coexpression of STIM1, ORAI1 and ORAI2 resulted in large CRAC currents, which could be blocked with La 3 þ or 2-aminoethoxydiphenyl borate and showed Na þ permeation in DVF solution, thus resembling endogenous I CRAC (Fig. 4i). Furthermore, ORAI1, ORAI2 and ORAI1/ORAI2 channels had similar inwardly rectifying I-V relationships typical of I CRAC (Fig. 4j). Importantly, coexpression of ORAI2 WT with the poredead ORAI1 E106A mutant 5,37 strongly attenuated I CRAC (Fig. 4j,k), confirming our T cell data ( Fig. 4a-h). Peak current amplitudes were not different in cells coexpressing STIM1 together with either ORAI1, ORAI2 or both (Fig. 4k), indicating that homomeric and heteromeric ORAI channels have similar abilities to conduct Ca 2 þ . By contrast, steady-state I CRAC was significantly reduced in cells coexpressing ORAI1/ORAI2 compared to ORAI1 alone (Fig. 4l). Further analysis of fast CDI of CRAC currents during hyperpolarizing voltage steps showed that ORAI1 currents inactivated by B20% within 100 ms, whereas ORAI2 currents inactivated more strongly by B45% (Fig. 4m,n). These findings are consistent with reported differences in fast CDI for ORAI1 and ORAI2 (refs 9,38). Importantly, CRAC currents in cells coexpressing ORAI1 and ORAI2 also had more pronounced fast CDI compared to cells expressing ORAI1 alone (Fig. 4m,n), likely accounting for their smaller steady-state current amplitudes (Fig. 4l). Together, these data show that ORAI1 and ORAI2 form heteromeric channels and suggest that differences in fast CDI between ORAI1 and ORAI2 likely contribute to distinct CRAC current and SOCE levels.
Orai1/Orai2 deletion disrupts peripheral immune homeostasis. We next investigated the role of ORAI1 and ORAI2 in T cell development and function. Orai2 À / À mice had normal populations of immature CD4 À CD8 À DN, CD4 þ CD8 þ DP, CD4SP and CD8SP T cells in the thymus and similar T cell numbers in spleen and lymph nodes (LNs) compared to littermates (Fig. 5a,b and Supplementary Fig. 5). Furthermore, the frequencies of CD44 hi CD62L lo effector CD4 þ and CD8 þ T cells as well as CD44 hi CD62L hi memory CD8 þ T cells were unchanged in Orai2 À / À mice (Fig. 5b). Unperturbed thymic T cell populations, T cell numbers and frequencies of effector and memory T cells were also observed in Orai1 fl/fl Cd4cre mice ( Fig. 5a and Supplementary Fig. 5a,b).
By contrast, combined deletion of both Orai1 and Orai2 in T cells caused splenomegaly and lymphadenopathy in Orai1 fl/fl Orai2 À / À Cd4cre mice at 43 months of age with elevated numbers of immune cell populations in the spleen and LNs ( Fig. 5a and Supplementary Fig. 5e), which is reminiscent of the phenotype observed in animals with combined T cell-specific deletion of Stim1 and Stim2 (ref. 39). Orai1 fl/fl Orai2 À / À Cd4cre mice had no defect in conventional T cell development as thymic cellularity ( Fig. 5a) and frequencies of DN, DP, CD4SP and CD8SP T cells were unaltered (Supplementary Fig. 5a,b). In the spleen, the frequencies of T cells were also comparable in Orai1 fl/fl Orai2 À / À Cd4cre and WT mice. This is consistent with data from Stim1/Stim2-deficient mice and ORAI1-or STIM1-deficient human patients, providing further evidence that SOCE is dispensable for conventional T cell development 39,40 . Orai1 fl/fl Orai2 À / À Cd4cre mice, but not single knockout mice, had elevated numbers of CD4 þ and CD8 þ T cells with CD44 hi CD62L hi memory and/or CD44 hi CD62L lo effector phenotypes, in line with their splenomegaly (Fig. 5a,b). A similar immune phenotype was not observed in either Orai1 fl/fl Cd4cre or Orai2 À / À mice. The disruption of the peripheral immune homeostasis in Orai1 fl/fl Orai2 À / À Cd4cre mice was likely due to the decreased numbers of Foxp3 þ T regulatory (Treg) cells in the thymus, spleen and LNs compared to WT, Orai1 fl/fl Cd4cre and Orai2 À / À mice ( Fig. 5c and Supplementary Fig. 5e). The development and maintenance of Treg cells is dependent on interleukin-2 (IL-2) 41 . Orai1/Orai2deficient T cells were unable to produce IL-2 in vitro (data not shown), which likely contributed to decreased Treg cell numbers. Collectively, our data show that deletion of Orai1 and Orai2 together (but not individually) impairs thymic Treg cell development resulting in perturbed immune homeostasis.
ORAI1 and ORAI2 control immunity to infection. Patients with loss-of-function mutations in ORAI1 that abolish SOCE exhibit a severe combined immunodeficiency-like disease phenotype with recurrent and chronic viral, bacterial and fungal infections 17,[42][43][44][45] . ORAI1-deficient human T cells proliferate poorly in response to antigen or mitogenic stimulation and have impaired cytokine production 17,20,21 . The proliferation of CD4 þ T cells from either Orai1 fl/fl Cd4cre or Orai2 À / À mice after stimulation with anti-CD3/CD28 was comparable to WT cells. By contrast, deletion of both Orai1 and Orai2 severely compromised proliferation ( Fig. 6a and Supplementary Fig. 6). To test if this defect was secondary to decreased IL-2 production      (n) Summary of fast inactivation during the 100 ms pulse measured as: 1 À I ss /I peak , where I peak and I ss are the peak and steady-state currents during the pulse to À 100 mV. k,l,n are means±s.e.m. *Po0.05; **Po0.005; ***Po0.001 in (a,b,g,k,l,n) using unpaired Student's t-tests.
that functions as an autocrine growth factor 41 , we stimulated Orai1/Orai2-deficient T cells in the presence of recombinant IL-2. However, IL-2 supplementation only partially rescued proliferation in Orai1/Orai2-deficient T cells (Fig. 6a), indicating an intrinsic, IL-2-independent role of SOCE in T cell proliferation. Production of cytokines by CD4 þ helper T cells (T H ) is dependent on SOCE 45 and is impaired in ORAI1-deficient patients 46 . We therefore differentiated naive T cells from WT, Orai1 fl/fl Cd4cre, Orai2 À / À and Orai1 fl/fl Orai2 À / À Cd4cre mice into T H 1, T H 17 and T H 2 cells in vitro and analysed the expression of interferon-g (IFNg), IL-17 and IL-4, respectively (Fig. 6b,c). Cytokine production was significantly reduced in Orai1and Orai1/Orai2-deficient T H 1, T H 17 and T H 2 cells, whereas Orai2 À / À T cells had normal cytokine expression compared to WT cells. ORAI1 and STIM1-deficient patients have defective humoral immune responses and lack antigen-specific antibodies 17,21 . We recently reported that STIM1 and STIM2 are required for germinal centre (GC) reactions and the production of virusspecific antibodies by controlling the differentiation of T follicular helper (T FH ) cells 18 . To test how ORAI1 and ORAI2 contribute to T cell-dependent humoral immunity, we infected WT, Orai1 fl/fl Cd4cre, Orai2 À / À and Orai1 fl/fl Orai2 À / À Cd4cre mice with lymphocytic choriomeningitis virus (LCMV) and analysed T FH cells, the GC reaction and anti-LCMV antibody production. At 10 days postinfection, CD4 þ CXCR5 hi PD-1 hi T FH cells were significantly reduced in Orai1 fl/fl Orai2 À / À Cd4cre mice compared to WT controls (Fig. 7a). In contrast, Orai1-or Orai2deficient mice had normal numbers of T FH cells. Analysing CD19 þ CD38 À GL.7 þ GC B cells by flow cytometry (Fig. 7b) and GCs within B-cell follicles using immunohistochemistry (Fig. 7c), we observed a strongly impaired GC reaction in Orai1 fl/fl Orai2 À / À Cd4cre but not single knockout or WT mice (Fig. 7b,c). Consequently, LCMV-specific, class-switched immunoglobulin-G (IgG) antibodies were markedly decreased in Orai1 fl/fl Orai2 À / À Cd4cre (but not single knockout) mice, whereas LCMV-specific IgM antibodies were normal (Fig. 7d). Taken together, our data demonstrate that both ORAI1 and ORAI2 are required for T FH cell-dependent humoral immunity.
Deletion of Orai1 and Orai2 prevents T cell-induced colitis. T cells provide immunity to infection but also mediate inflammation in a variety of autoimmune diseases 47,48 . To elucidate the role of Orai1 and Orai2 in autoimmunity, we used an adoptive transfer model of IBD that tests the function of ORAI1 and ORAI2 in pathogenic donor T cells 49 . We injected naive CD4 þ T cells from WT, Orai1 fl/fl Cd4cre, Orai2 À / À and Orai1 fl/fl Orai2 À / À Cd4cre mice into lymphopenic Rag1 À / À mice to Cd4cre mice lost B10% of their original body weight (Fig. 8a).
Transfer of Orai2-deficient T cells caused more pronounced (but statistically not significant) weight loss compared to WT cells. Importantly, T cells from Orai1 fl/fl Orai2 À / À Cd4cre mice did not induce any weight loss (Fig. 8a). Macroscopic inspection of colons after 12 weeks showed colon shortening, thickening and bleeding after transfer of WT or Orai2-deficient T cells, and to a lesser extent of Orai1-deficient T cells (Fig. 8b). In contrast, colons appeared healthy in mice that were injected with Orai1 fl/fl Orai2 À / À Cd4cre T cells (Fig. 8b). The histopathological examination of colonic sections after transfer of Orai1/Orai2deficient T cells revealed no IBD symptoms (such as immune cell infiltration, epithelial hyperplasia, goblet cell depletion and/or ulceration) resulting in low IBD scores of o1 (Fig. 8c,d).
Consistent with reduced colitis, the frequency and absolute numbers of CD4 þ T cells in the spleen, mesenteric LNs and lamina propria (LP) were significantly reduced in mice that had received T cells from Orai1 fl/fl Orai2 À / À Cd4cre mice compared to WT or single knockout mice ( Fig. 8e and Supplementary  Fig. 7). Proinflammatory cytokines including IFNg, tumournecrosis factor-a (TNFa) and IL-17, which are produced by colitogenic T H 1 and T H 17 cells, are critically involved in IBD pathogenesis. Orai1-and Orai1/Orai2-deficient donor T cells isolated from the spleen, mesenteric LNs and LP had strongly impaired production of IFNg, TNFa and IL-17 after restimulation ex vivo compared to WT or Orai2-deficient T cells (Fig. 8f). It is noteworthy that although cytokine production was similarly impaired in Orai1-and Orai1/Orai2-deficient T cells (Fig. 8f), only mice that had received Orai1/Orai2-deficient T cells were fully protected from colitis, which is likely explained by the combination of abrogated cytokine production, proliferation and decreased tissue homing, all of which were more pronounced in Orai1/Orai2-deficient compared to Orai1-deficient T cells.
Deletion of Orai1 and Orai2 in T cells prevents GvHD. GvHD is a complication of BM transplantation during which allogenic T cells in the transplant cause inflammation of the recipient's liver, skin and gastrointestinal tract 50 . To investigate the role of ORAI1 and ORAI2 in donor T cells after allogenic hematopoietic cell transplantation (allo-HCT), we transplanted total T cells from either WT, Orai1 fl/fl Cd4cre, Orai2 À / À or Orai1 fl/fl Orai2 À / À Cd4cre C57BL/6 mice, together with BM cells from WT C57BL/6 mice into lethally irradiated allogenic BALB/c host mice. Transplantation of WT or Orai2 À / À T cells caused fulminant GvHD (using a composite clinical GvHD score as described in ref. 51) after allo-HCT, whereas Orai1 fl/fl Cd4cre T cells resulted in markedly reduced immunopathology (Fig. 9a,b). Strikingly, Orai1/Orai2-deficient T cells did not cause any clinical GvHD symptoms including weight loss (Fig. 9a,b) and were indistinguishable from control mice that had been transplanted with BM from C57BL/6 mice only.
In acute GvHD, levels of IFNg and TNFa correlate with disease activity 50,51 . IFNg and TNFa levels in the sera of recipient mice (Fig. 9c) as well as IFNg and TNFa production by donor CD4 þ and CD8 þ T cells restimulated ex vivo (Fig. 9d) were strongly impaired in mice that had received Orai1-deficient T cells or Orai1 fl/fl Orai2 À / À Cd4cre T cells (Fig. 9c,d). In contrast, transfer of Orai2 À / À or WT T cells resulted in comparable and robust cytokine production. The histopathological analysis of liver, small intestine and large intestine showed pronounced immune cell infiltration and inflammation in mice transplanted with WT or Orai2 À / À T cells, whereas less or no immunopathology was apparent in recipients of Orai1-or Orai1/Orai2-deficient T cells, respectively (Fig. 9e). Although mice that had received Orai1-deficient T cells survived significantly longer than recipients of WT or Orai2 À / À T cells, only animals transplanted with Orai1/Orai2-deficient T cells were completely protected from lethal GvHD (Fig. 9f). Taken together, loss of Orai2 moderately exacerbated and loss of Orai1 ameliorated lethal GvHD, whereas combined deletion of both Orai1 and Orai2 in T cells fully protected recipient mice from GvHD, demonstrating that ORAI1 and ORAI2 together regulate pathogenic T cell function.

Discussion
The importance of ORAI1 for SOCE and immune function is well established 17,44,45 , but that of ORAI2 and ORAI3 remains essentially unknown. Here we find that ORAI1 and ORAI2 are highly expressed in mouse T cells. Deletion of Orai1 moderately decreased SOCE in naive and more strongly in effector T cells. By contrast, deletion of Orai2 enhanced SOCE in naive but not effector T cells. Combined deletion of Orai1 and Orai2 completely abolished SOCE, providing strong evidence that both ORAI1 and ORAI2 mediate SOCE in T cells, whereas ORAI3 is dispensable. We here show that ORAI1 and ORAI2 form a heteromeric channel complex, in which ORAI2 attenuates the function of ORAI1 and limits SOCE. This effect is at least in part due to smaller steady-state CRAC currents in cells coexpressing ORAI1 and ORAI2, likely caused by enhanced fast CDI in these cells. Abolishing SOCE by combined deletion of Orai1 and Orai2 resulted in severe defects in T cell function in vitro and T cell-mediated immune responses in vivo including antibody production after viral infection, colitis and GvHD. Individual deletion of ORAI1 or ORAI2 had less pronounced or no effects, respectively.
The inhibitory role of ORAI2 in CRAC channel function is not limited to naive T cells. We also observed increased SOCE in Orai2-deficient neutrophils, astrocytes, dendritic cells and macrophages as well as shORAI2-treated human fibroblasts. A similar enhancement of SOCE has recently been reported in Jurkat T cells and a chondrocyte cell line after knockdown of ORAI2 by RNA interference 52,53 . These data indicate that ORAI2 attenuates CRAC channel function and SOCE in many cell types. We show that increased SOCE in Orai2-deficient cells is due to increased I CRAC mediated by ORAI1. CRAC currents were increased in Orai2-deficient BMDMs and had biophysical properties comparable to native CRAC currents. Furthermore, Fura-2 quench rates by extracellular Mn 2 þ were larger in Orai2-deficient BMDMs and T cells. Enhanced SOCE in ORAI2-deficient cells was not due to greater expression of ORAI1 or altered Ca 2 þ content of ER stores. ORAI2 has been proposed to be an ER release channel that controls the Ca 2 þ filling state of the ER and may thereby control SOCE 33 . If this was the main function of ORAI2, we would have expected that ORAI2-deficient cells have more Ca 2 þ stored in their ER, which was, however, not the case.
Our finding that combined deletion of ORAI1 and ORAI2 abolishes I CRAC and SOCE, whereas single deletion of ORAI2 enhances both is best explained by a model in which ORAI1 and ORAI2 form heteromeric channels. CRAC channels are formed by multimers of ORAI proteins, which most likely form hexameric complexes 1,2 . While most studies have investigated homomeric ORAI channels by ectopic expression of either ORAI1, ORAI2 or ORAI3, it is likely that CRAC channels are heteromers of different ORAI homologues. This idea is supported by coexpression of different ORAI homologues in many tissues 14 as well as co-immunoprecipitation 14 and fluorescence resonance energy transfer 53 between ectopically expressed ORAI1 and ORAI2 proteins. We here demonstrate that ORAI1 and ORAI2 together form functional heteromeric CRAC channels in naive T cells (and other cell types) in which both ORAI1 and ORAI2 subunits form the channel pore. Previous studies have revealed that ectopically expressed ORAI2 generates CRAC currents with similar properties but smaller amplitudes than ORAI1 (refs 8,9). Thus, deletion of ORAI2 protein from heteromeric CRAC channels in naive T cells might be expected to produce homomeric ORAI1 channels with larger currents resulting in increased SOCE, which is consistent with our findings. Although the presence of ORAI2 in heteromeric ORAI1:ORAI2 channels has an inhibitory effect on CRAC channel function and SOCE, ORAI2 is a functional CRAC channel. This is apparent from the fact that residual SOCE in T cells from Orai1 fl/fl Cd4cre mice is abolished when Orai2 is also deleted and that overexpression of ORAI2 results in currents resembling those of endogenous CRAC channels consistent with published reports 8,9 . Evidence supporting direct interaction of ORAI1 and ORAI2 in heteromeric complexes comes from our experiments in which expression of the pore-dead ORAI1 E106Q mutant in Orai1deficient T cells (whose residual SOCE is mediated by ORAI2) abolished SOCE. Conversely, expression of the pore-dead ORAI2 E80Q mutant in Orai2-deficient T cells (whose enhanced SOCE is mediated by ORAI1) also abolished SOCE. These effects of ectopically expressed pore-dead ORAI mutants on SOCE are unlikely due to their sequestration of STIM1. Rather, we conclude that pore-dead ORAI proteins directly assemble with their endogenous homologues into heteromeric channels and abolish Ca 2 þ conductance. This conclusion is supported by experiments in which we overexpressed ORAI1, ORAI2 or both ORAI1/ORAI2 together with STIM1. Whereas coexpression of WT ORAI1 and ORAI2 results in very large CRAC currents, coexpression of the pore-mutant ORAI1 E106A together with ORAI2 abolished I CRAC . Since STIM1 is abundant under these conditions, the effects of ORAI1 E106A are not due to STIM1 sequestration away from ORAI2 but instead support the model of direct ORAI1/ORAI2 interaction in a heteromeric channel.
Enhanced I CRAC and SOCE in ORAI2-deficient cells and evidence for heteromeric ORAI1:ORAI2 channels suggest that ORAI2 is less efficient than ORAI1 or ORAI1:ORAI2 channels in conducting Ca 2 þ . It is therefore noteworthy that peak current amplitudes were comparable in cells expressing ORAI1, ORAI2 or ORAI1/ORAI2, indicating that homomeric and heteromeric ORAI channels have similar abilities to conduct Ca 2 þ . By contrast, steady-state currents were significantly larger in cells expressing ORAI1 compared to ORAI2, which can be explained by the more pronounced fast CDI observed in cells expressing ORAI2 compared to ORAI1 consistent with previous reports 9, 38 . More studies are needed to elucidate the mechanistic differences in fast CDI and ion conduction between ORAI1 and ORAI2. An alternative explanation for enhanced SOCE in Orai2 À / À T cells that merits further investigation is that gating of ORAI1 channels by STIM1 is more efficient than that of ORAI2 channels 38,54 .
The concept of heteromeric CRAC channel complexes is intriguing because the dynamic variation of the stoichiometry of two homologues with higher (ORAI1) and lower (ORAI2) Ca 2 þ permeation provides a mechanism to regulate SOCE while maintaining essential CRAC channel properties. We found that in naive T cells both ORAI1 and ORAI2 contribute to CRAC channel function with ORAI2 assuming an inhibitory role apparent from increased SOCE in naive Orai2 À / À T cells. In effector T cells, ORAI1 was up-and ORAI2 downregulated compared to naive T cells, resulting in an increased ORAI1:ORAI2 expression ratio ( Supplementary Fig. 8). As a consequence, deletion of Orai2 had little effect on SOCE in effector T cells, whereas deletion of Orai1 suppressed SOCE more strongly than in naive T cells. This reciprocal expression of ORAI1 and ORAI2 in naive versus effector T cells provides a mechanism to fine-tune the magnitude of SOCE and thus the strength of T cell-mediated immune responses. From an immunological perspective, increasing the number of ORAI2 subunits in heteromeric CRAC channels represents a means to reduce the magnitude of SOCE and restrain the strength of TCR signalling, thus reducing the likelihood of random, detrimental activation of naive T cells. By contrast, effector T cells have to respond quickly and effectively to cognate antigens. Increasing the ORAI1:ORAI2 ratio and thus the magnitude of SOCE ensures robust T cell activation and efficient recall responses.
We here show that ORAI1 and ORAI2 form the CRAC channel complex in T cells and are required for T cell-mediated immune responses, whereas ORAI3 is dispensable. T cells from Orai1 fl/fl Orai2 À / À Cd4cre mice not only had abolished SOCE but also lacked proliferation, cytokine production, T FH cell differentiation and virus-specific antibody production. Furthermore, Orai1/Orai2-deficient T cells did not cause immunopathology in adoptive transfer models of colitis and GvHD. These findings are reminiscent of Stim1 fl/fl Stim2 fl/fl Cd4cre mice, whose T cells lack SOCE and fail to differentiate into T FH cells in response to viral infection or immunization 18 . Likewise, Orai1 fl/fl Orai2 À / À Cd4cre mice had impaired T FH cell differentiation, GC formation and antibody production. Individual deletion of Orai1 or Orai2 resulting in partially impaired or increased SOCE, respectively, had no effect on T cell-proliferation in vitro or T FH cell differentiation, GC formation and antibody production in vivo. Thus, only complete elimination of SOCE in Orai1/Orai2-or Stim1/Stim2-deficient mice compromises proliferation and T cell-dependent antibody production. Furthermore, Orai1 fl/fl Orai2 À / À Cd4cre mice had reduced Treg numbers 18,39,40 and accumulation of effector/memory T cells in secondary lymphoid organs, which caused splenomegaly and lymphadenopathy, that is also observed in Stim1 fl/fl Stim2 fl/fl Cd4cre mice 39 .
Enhanced SOCE in naive Orai2 À / À T cells might have been expected to increase T cell function in vitro and T cell-dependent immunopathologies in vivo, which was however not the case. The severity of IBD was only moderately increased after adoptive transfer of naive Orai2 À / À T cells compared to WT controls. Likewise, GvHD after transfer of Orai2-deficient T cells was not significantly more severe compared to that induced by WT cells. A likely explanation for this lack of exacerbation of immune responses by Orai2-deficient T cells is the lower expression of ORAI2 in effector T cells compared to naive T cells and the resulting predominance of ORAI1 in effector T cells. Indeed, deletion of Orai2 increased SOCE in naive but not effector T cells. Since immunopathology in the IBD and GvHD models is mediated by effector T cells that have normal SOCE in the absence of ORAI2, exacerbated immunopathology would not be expected in Orai2 À / À mice. In human T cells, ORAI2 is expressed albeit at lower levels than in murine T cells. Although somatic mutations of ORAI2 were reported in cancer 55 , no functional ORAI2 germline mutations have been reported. By contrast, null or loss-of-function mutations in the human ORAI1 gene that abolish SOCE severely impaired T cell function 17,44 . Patients with inherited mutations in ORAI1 present clinically with CRAC channelopathy characterized by a severe combined immunodeficiency-like disease, autoimmunity, anhidrotic ectodermal dysplasia and congenital myopathy 17,44,45 . Although abolished SOCE, T cell dysfunction and immunodeficiency in these patients indicate that ORAI1 is the dominant CRAC channel subunit in human effector T cells, these findings do not exclude a potential role of ORAI2 in other immune or non-immune cells in which it is expressed.

Methods
Mice. Orai1 fl/fl Cd4cre mice were described before 25 . Rag1 À / À mice (JAX strain number 002216) and BALB/c WT mice (JAX strain number 000651) were purchased from The Jackson Laboratory. Orai2 LacZ/ þ mice carrying a LacZ reporter cassette were generated using VGB6 embryonic stem cells (C57BL/6NTac) obtained from the Knock-Out Mouse Project (KOMP) repository at UC Davis (project ID VG14962, Orai2 tm1(KOMP)Vlcg ). The targeting strategy for generating Orai2 LacZ/ þ mice is shown in Supplementary Fig. 2a. Exon 4 and the 5 0 region of exon 5 representing the coding sequence of Orai2 were replaced by homologous recombination with a loxP-flanked neomycin gene (neo R ) and a LacZ reporter gene. The neo R gene was later removed by mating Orai2 LacZ/ þ offsprings to CMV-Cre mice, followed by removal of the Cre transgene by backcrossing Orai2 LacZ/ þ mice to C57BL/6 mice. The resulting offspring was intercrossed to generate Orai2 À / À (Orai2 LacZ/LacZ ) mice. Genotyping of mice for the presence of the WT allele, inserted neo R gene and deleted Orai2 À / À allele was performed using mouse tail DNA and PCR (as shown in Supplementary Fig. 2b) using the following primer pairs: WT allele 5 0 -CTCGGCAGTTGCCTGTTTG-3 0 and 5 0 -ACAGCCACCACGCTCATC-3 0 (product B89 bp); Orai2 À / À allele 5 0 -GGTAAACTGGCTCGGATTAGGG-3 0 and 5 0 -TTGACTGTAGCGGCTGATG TTG-3 0 (product B210 bp); neo R gene 5 0 -TCATTCTCAGTATTGTTTT GCC-3 0 and 5 0 -CTTTGGGAATTCACCACCC-3 0 (product B400 bp). PCR conditions for all reactions were the following: denaturation at 95°C for 10 min, followed by 40 cycles of 95°C for 60 s, annealing at 55°C for 30 s and elongation at 72°C for 30 s with an additional 10 min final elongation step at 72°C. All animals were on a C57BL/6 genetic background. Both male and female mice were used between 6 and 16 weeks of age (unless otherwise noted). Sample size for various animal experiments was chosen based on prior data generated in the laboratory, no mice were excluded from experiments. Mice were maintained under specific pathogen-free conditions. All animal experiments were approved by the Institutional Animal Care and Use Committee at New York University School of Medicine.
Plasmids. Murine stem cell virus (MSCV)-based retroviral vectors expressing ORAI1 WT , ORAI2 WT and ORAI1 E106Q were described before 5,14,20 . ORAI2 E80Q , ORAI1 L273D and ORAI1 L273D/E106Q retroviral vectors were generated using the QuikChange II XL Site-directed Mutagenesis Kit (Agilent; cat. no. 200521) according to the manufacturer's instructions. Mutations were confirmed by DNA sequencing. Expression plasmids used in patch-clamp experiments were as follows: N-terminally YFP-tagged ORAI1 was purchased from GeneCopoeia. N-terminally CFP-tagged ORAI2 was described before 5,37 . The ORAI1 E106A mutation was generated by site-directed mutagenesis using the QuickChange II XL Site-directed Mutagenesis Kit (Agilent). The indicated ORAI1 or ORAI2 constructs were transfected into HEK293 cells together with a construct expressing unlabelled STIM1 (pCMV6-XL5; OriGene Technologies) using Lipofectamine (Thermo Fisher; cat. no. L3000015). The primer sequences for site-directed mutagenesis are listed in Supplementary Table 3.
Retroviral transduction of T cells. CD4 þ T cells were isolated using a mouse CD4 þ T Cell Enrichment Kit (STEMCELL Technologies) and stimulated with 1 mg ml À 1 plate-bound anti-CD3 (clone 2C11) plus 1 mg ml À 1 anti-CD28 antibodies (clone 37.51; both Bio X Cell) in the presence of 50 U ml À 1 IL-2 (Peprotech). T cells were transduced 24 h after stimulation by spin infection (2,500 r.p.m., 30°C, 90 min) in the presence of concentrated retroviral supernatant and 10 mg ml À 1 polybrene (Santa Cruz Biotechnology). Retroviral supernatant was produced in the Platinum-E retroviral packaging cell line. Platinum-E cells (ATCC) were transfected with GenJet transfection reagent (SignaGen; cat. no. SL100499) with murine stem cell virus-based retroviral vectors expressing ORAI1 WT , ORAI2 WT and ORAI1 E106Q (as described in refs 5,14) as well as ORAI2 E80Q , ORAI1 L273D and ORAI1 L273D/E106Q together with the amphotrophic packaging vector pCL-10A1. Two days after transfection, the supernatant was collected and concentrated using Amicon Ultra-15 centrifugal filters (Merck Millipore; cat. no. UFC910024). At 4 h after spin infection, viral supernatant was removed, and then replaced by fresh media. At 48 h after transduction, T cells were transferred into new plates adding fresh media containing 50 U ml À 1 IL-2. Transduced cells were FACS sorted using a sterile Sony SY3200 (HAPS1) cell sorter and Ca 2 þ influx was tested using a FlexStation 3 plate reader.
Flow cytometry and cell sorting. Cells isolated from spleen, LNs and LP were washed in cold PBS containing 0.1% BSA and unspecific binding was blocked using anti-FcgRII/FcgRIII (2.4G2; eBioscience). Staining of surface molecules with fluorescently labelled antibodies was performed at 4°C for 30 min in the dark. Intracellular cytokine (IC) staining was performed using the IC Staining Buffer Kit (eBioscience; cat. no. 00-8222-49) according to the manufacturer's instructions. Surface expression of human ORAI1 in fibroblasts was detected using two different custom-made anti-human ORAI1 monoclonal antibodies, which both recognize the second extracellular loop of hORAI1. Clone 29A2 is a mouse anti-human ORAI1 mAb (IgG1) that was generated with an ovalbumin conjugated peptide corresponding to amino acids 196 to 234 of human ORA1 using mice that overexpress the neonatal Fc receptor (FcRn) and have augmented immune response due to the role of FcRn in antigen presentation 62 . Immune sera and hybridoma supernatants were tested using human ORAI1 transfected and parenteral NIH-3T3 cells with flow cytometry analyses. A mouse IgG1 Ab was used as isotype control and a goat anti-mouse IgG conjugated to AlexaFluor 647 (Life technologies) as secondary antibody for detection. Clone 2C1.1 is a fully human anti-ORAI1 monoclonal antibody conjugated to allophycocyanin (APC) as described in ref. 57. A complete list of antibodies with the respective conjugated fluorochromes can be found in Supplementary Table 1 T cell proliferation assay. A total of 5 Â 10 6 T cells were loaded with carboxyfluorescein diacetate succinimidyl ester (CFSE) (Invitrogen; cat. no. C34554) according to the manufacturer's instructions. Cells were blocked with 50% FBS, washed twice with RPMI and stimulated with 1 mg ml À 1 plate-bound anti-CD3 (clone 2C11) plus 1 mg ml À 1 anti-CD28 antibodies (clone 37.51; both Bio X Cell) with or without 50 U ml À 1 IL-2 (NIH) for 4 days. CFSE dilution was monitored daily using an LSRII flow cytometer and FACSDiva Software (BD Biosciences) and further analysed with FlowJo Software (Tree Star).
Quantitative real-time PCR. RNA was extracted from FACS-sorted T cells and BMDMs using the RNeasy Micro Kit (Qiagen; cat. no. 74004), followed by cDNA synthesis with the iScript II Kit (Bio-Rad; cat. no. 170-8890). qRT-PCR was performed using Platinum SYBR Green qPCR SuperMix (Invitrogen; cat. no. 11744100) and an Opticon 2 thermocycler (Bio-Rad). PCR conditions were as follows: 95°C for 10 min, followed by 40 cycles (94°C for 45 s, 58°C for 30 s and 72°C for 30 s) of amplification. For quantitation, C T values were normalized to Hprt and expression was analysed using the 2 À DDCT method. Additional housekeeping genes (Actin, Nono, 18S) were used to validate Orai1 and Orai2 expression in T cells, yielding similar results compared with Hprt (not shown). A complete list of all primers used in this study can be found in Supplementary Table 2. shRNA-mediated gene knockdown. HEK293T cells (ATCC) were transfected with SGEP (ref. 19) and pLKO.1 (Addgene) lentiviral plasmids expressing shRNAs specific for human ORAI1 and ORAI2, respectively, or scrambled shRNA together with packaging plasmid psPAX2 (Addgene; 12260) and vesicular stomatitis virus-G envelope plasmid pMD2.G (Addgene; cat. no. 12259) using the GenJet transfection reagent (SignaGen; cat. no. SL100499). ORAI1 and ORAI2 shRNA targeting sequences were 5 0 -TGTCCTCTAAGAGAATAAGCAT-3 0 and 5 0 -CATCTTCGTG GTCTTCACCATC-3 0 (Sigma), respectively. Human Hs27 fibroblasts (ATCC) were spin infected with fresh lentiviral supernatants for 90 min at 2,500 r.p.m. At 72 h post-transfection, transduced cells were selected with 1 mg ml À 1 puromycin for 5 days. The knockdown efficiency of ORAI1 and ORAI2 genes was examined by qRT-PCR. All cell lines were tested for mycoplasma. Intracellular Ca 2 þ levels and Mn 2 þ quenching. Cells were labelled with 2 mM Fura-2-AM (Life Technologies; cat. no. F1201) for 30 min in RPMI medium as described earlier 56 . Cells were attached for 10 min to 96-well imaging plates (Fisher) coated with 0.01% poly-L-lysine (w v À 1 ) (Sigma-Aldrich) and washed two times with Ca 2 þ -free Ringer solution (155 mM NaCl, 4.5 mM KCl, 2 mM CaCl 2 , 1 mM MgCl 2 , 10 mM D-glucose and 5 mM Na-HEPES) solution. Changes in intracellular Ca 2 þ concentration were analysed using a FlexStation 3 plate reader (Molecular Devices) at 340 and 380 nm excitation wavelengths. Cells were stimulated with 1 mM thapsigargin or 0.3 mM ionomycin (both EMD Millipore) in Ca 2 þ -free Ringer solution and SOCE was analysed after readdition of 1 mM Ca 2 þ (final) Ringer solution (for primary T cells and BMDMs) or 20 mM (final) Ca 2 þ Ringer solution (for human fibroblast cells). SOCE was quantified by the peak or the slope of the F340/380 ratio. ER store depletion was analysed by the integrated Ca 2 þ signal (area under the curve, AUC) after thapsigargin or ionomycin stimulation and before readdition of 1 mM Ca 2 þ Ringer solution. For measurements of store-operated bivalent cation influx via CRAC channels by Mn 2 þ quenching, BMDMs were loaded with Fura-2-AM, transferred to Ca 2 þ -free Ringer solution and stimulated with 0.3 mM ionomycin or 1 mM thapsigargin. Fura-2 fluorescence was measured at an emission wavelength of 360 nm before and after addition of 1 mM Mn 2 þ .
Patch-clamp measurements. Currents were recorded in the standard whole-cell configuration at room temperature using an Axopatch 200B amplifier (Molecular Devices) interfaced to an ITC-18 input/output board (Instrutech). Currents were filtered at 1 kHz with a 4-pole Bessel filter and sampled at 5 kHz. Stimulation, data acquisition and analysis were performed using routines developed on the Igor Pro platform by R.S. Lewis (Stanford University, Palo Alto, CA). Recording electrodes were 'pulled' from 100 ml pipettes, were coated with Sylgard and were fire polished to a final resistance of 2-5 MO. All data were corrected for the liquid-junction potential of the pipette solution relative to that of Ringer's solution in the bath ( À 10 mV) and for leak currents collected in 20 mM extracellular Ca 2 þ plus La 3 þ . The standard extracellular Ringer's solution contained 130 mM NaCl, 4.5 mM KCl, 20 mM CaCl 2 , 10 mM D-glucose and 5 mM HEPES (pH 7.4) with NaOH. The DVF Ringer's solution contained 150 mM NaCl, 10 mM HEDTA, 1 mM EDTA and 10 mM HEPES (pH 7.4) with NaOH. 10 mM TEA-Cl was added to all extracellular solutions to block voltage-gated K þ channels. The standard internal solution contained 135 mM caesium aspartate, 8 mM MgCl 2 , 8 mM BAPTA and 10 mM HEPES (pH 7.2) with CsOH. The holding potential was þ 30 mV. The standard voltage stimulus consisted of a 100-ms step to À 100 mV, followed by a 100-ms ramp from À 100 to þ 100 mV applied at 1 s intervals. I CRAC was typically activated by depleting ER Ca 2 þ stores with 1 mM thapsigargin, or by passive dialysis of BAPTA into the cell as indicated in the figure legends. Analysis of current amplitudes was performed by measuring either the peak, or in some cases, the steady-state current (as indicated in the figure legends) during the À 100 mV step pulse.
LCMV infection and detection of LCMV-specific antibodies. The LCMV Armstrong (LCMV ARM ) strain was kindly provided by R. Ahmed (Emory University, Atlanta, GA). Virus was grown in BHK-21 cells and titres (PFU) in the supernatant were determined as described before 58 . For acute viral infections, mice were injected intraperitoneally with 2 Â 10 5 PFU of LCMV ARM and analysed 10 days after infection 18,58 . LCMV-specific antibodies in the sera of mice were measured as described before 18,58 . Briefly, lysates of LCMV-infected BHK-21 cells were used as substrate and LCMV-specific IgG and IgM binding was detected in serial dilutions of serum using AP-conjugated goat-anti-mouse IgG or IgM antibodies (both Southern Biotech). Enzyme-linked immunosorbent assays (ELISAs) were developed with pNPP substrate (Thermo Scientific; cat. no. 37621) and absorbance was analysed at 405 nm using a SpectraMax M5 microplate reader (Molecular Devices).
Adoptive transfer colitis. Colitis was induced in Rag1 À / À host mice by intraperitoneal injection of 5 Â 10 5 CD4 þ CD62L þ CD25 À naive T cells isolated from LNs and spleen of WT, Orai1, Orai2 and Orai1/Orai2-deficient mice and sorted by flow cytometry using a SONY SY3200 cell sorter. Greater than 95% enrichment of CD4 þ CD62L þ CD25 À T cells was confirmed by flow cytometry. Injected Rag1 À / À mice were assessed two times a week for weight loss and other symptoms of distress over a course of 12 weeks 24,59,60 . Proximal and distal colon sections were fixed in 4% paraformaldehyde, paraffin embedded and stained with haematoxylin and eosin (H&E) using standard protocols. Colon histology was scored in a blinded fashion by two individuals using a grading system from 0 to 5: 0, no changes; 1, minimal scattered mucosal inflammatory cell infiltrates, with or without epithelial hyperplasia; 2, mild scattered to diffuse inflammatory cell infiltrates, sometimes extending into the submucosa and associated with erosions, with minimal to mild epithelial hyperplasia and minimal to mild mucin depletion from goblet cells; 3, mild-to-moderate inflammatory cell infiltrates that were sometimes transmural, often associated with ulceration, with moderate epithelial hyperplasia and mucin depletion; 4, marked inflammatory cell infiltrates that were often transmural and associated with ulceration, with marked epithelial hyperplasia and mucin depletion; 5, marked transmural inflammation with severe ulceration and loss of intestinal glands.
Graft-versus-host disease. BALB/c host mice (H-2 d ) were conditioned by myeloablative total body irradiation with a split-dose dose of 8.0 Gy total using a Gammacell 40 Exactor System (Best Theratonics). At 2 h after total body irradiation, mice were injected retroorbitally with 5 Â 10 6 C57BL/6 WT BM cells (H-2 b ) with or without 1.2 Â 10 6 C57BL/6 total T cells (H-2 b ). Animals were treated with antibiotic drinking water (sulfamethoxazole and trimethoprim; Hi-Tech Pharmacal) for 7 days before and after allo-HCT to prevent unspecific infections. Transplanted mice were assessed daily for weight loss and clinical symptoms of GvHD such as posture (hunching), fur texture, skin integrity and overall activity that were scored and used to calculate a composite GvHD score (adapted from Cooke et al. 51,61 ). Mice were killed when GvHD symptoms reached our end-point scoring criteria, analysed for organ histopathology and used for cell isolation.
Enzyme-linked immunosorbent assays. TNFa and IFNg serum levels in BALB/c host mice 12 days after allo-HCT were measured using commercial ELISA Kits (eBioscience; cat. no. 88-7324-22 and 88-8314-22) and analysed at 450 nm using a SpectraMax M5 microplate reader (Molecular Devices) according to the protocols provided by the manufacturer.
Histochemistry and immunofluorescence (IF). Livers, small intestine and colon of mice were collected, processed and H&E stained 51,59 . Whole E17.5 embryos were washed with PBS before fixation with 4% paraformaldehyde for 15 min. Tissue was washed with PBS containing 0.02% NP-40 three times for 30 min. Specimens were stained with 1 mg ml À 1 X-gal (Sigma) in X-gal buffer (5 mM K 3 Fe(CN) 6 , 5 mM K 4 Fe(CN) 6 , 2 mM MgCl 2 , 0.01% Na deoxycholate, 0.02% NP-40; all Sigma) overnight at 30°C. After washing three times with PBS, samples were stored in PBS containing 0.1% PFA or embedded with Tissue-Tek (Thermo Scientific) for histological analyses. For immunohistochemistry, tissue specimens were fixed in 4% PFA for at least 24 h and embedded in paraffin. Heat-induced antigen retrieval of tissue sections was performed in 10 mM citric acid buffer (pH 6.0) for 20 min using a high-pressure cooker (Deni). Blocking for 1 h with Antibody Diluent (Dako) was followed by overnight incubation at 4°C with the following primary antibodies: rat anti-B220 (1:400, clone RA3-8B2; eBioscience) and biotinylated PNA (1:100; Vector Laboratories). Detection of IF was carried out using the following antibodies: goat anti-rat IgG (H þ L) AlexaFluor555 (1:800) and streptavidin AlexaFluor488 (1:800; both Molecular Probes). Slides were mounted using Prolong Gold Antifade Mounting Medium (Molecular Probes). Images were captured using a Nikon microscope Eclipse TE2000. Data were analysed with NIS Elements Software (Nikon) and further processed using Photoshop.
Statistical analyses. All data are shown as mean ± s.e.m. of at least three independent experiments. No specific randomization or blinding protocols were used. Figures were prepared using GraphPad Prism 6, Adobe Photoshop and Illustrator CC2014 Software. Different groups were compared using a two-tailed paired or unpaired Student's t-test or two-way analysis of variance using Prism6 (GraphPad Software). Differences with a P value of o0.05 were considered significant: *Po0.05; **Po0.005; ***Po0.001.
Data availability. The authors declare that all data supporting the findings of this study can be found within the paper and its Supplementary Information files. Additional data supporting the findings of this study are available from the corresponding author (S.F.) upon reasonable request.