Main

During formation of the pre-immune repertoire, highly abundant self-antigens induce the negative selection of immature B cells through deletion, receptor editing or anergy1, while some less-abundant or intracellular self-antigens trigger the positive selection of the polyspecific innate-like B1 B cell subset, at least during early ontogeny2. Later, once the repertoire has been formed, some high-avidity and regularly arrayed antigens can generate a sufficiently strong B cell receptor (BCR) signal to induce proliferation and plasma cell differentiation in the absence of T cell help3. However, the majority of B-cell activation involves interactions with lower avidity and/or low-affinity antigens such as soluble proteins, which may be tolerogenic4, or immunogenic antigens coupled via complement or Fc receptors and displayed on the membranes of macrophages, non-cognate B cells or follicular dendritic cells5,6. Little is known about how B cells discriminate between these different antigenic forms, especially when compared with T cell development, partly because of the limited number of tools available to study B cell cognate responses in vivo and in vitro.

Themis2 (Icb-1), a protein that is primarily expressed in B cells and monocytes7, is a member of the newly described Themis family of proteins, that is found in animal species from mammals to cnidarians and is characterized by the presence of one or two copies of a cysteine-containing globular CABIT domain8. Themis2 shares 29% identity and 65% homology at the amino acid level with its T cell homolog, Themis1 (ref. 8). Although not much is known about the function of Themis2, Themis1 has been shown to be critical for positive selection of CD8+CD4+ T cells in the thymus and their transition to the CD8intCD4+ transitional stage8,9,10,11,12. Themis1 is constitutively associated with the signal adaptor protein Grb2 via a PxRPxK 'proline-rich' domain8,9,11,13 and, following T cell activation, it is rapidly tyrosine phosphorylated by the protein tyrosine kinases Lck and ZAP70 and recruited by Grb2 to the transmembrane adaptor LAT13,14,15,16. It has been proposed that Themis1 acts as an inhibitory analog-to-digital converter in developing T cells, ensuring that low-affinity peptide-MHC and T cell antigen receptor (TCR) interactions do not transduce high-intensity signals that result in negative selection, but instead generate signals appropriate for positive selection17. Furthermore, Themis1 and 2 are functionally interchangeable in T-cell development, as transgenic expression of Themis2 can rescue the development and signaling defects in Themis1−/− mice15. Given this evidence of shared function, we hypothesized that Themis2 might be a regulator of the B-cell response to different antigens during development and activation.

We investigated the function of Themis2 in B lymphocytes. We found that Themis2 regulates the positive selection of B1 B cells by self-antigen and the positive selection of germinal center (GC) B cells in response to foreign antigen. Themis2 lowered the threshold for B-cell activation by low-avidity antigens, but was not required for activation by high-avidity antigens. Themis2 bound and facilitated the activation of PLC-γ2, downstream Ca2+ mobilization and activation of Erk1/2 signaling pathways. These data provide insights into how B cells discriminate between qualitatively different antigenic signals via the same BCR and highlight how antigens of different quantity and quality may be able to fine-tune the immune response.

Results

The expression and gene targeting of Themis2

To investigate the function of Themis2, we first used flow cytometry to isolate distinct B and T cell subsets from C57BL/6 (B6) mice and quantified Themis1 and Themis2 mRNA expression by quantitative PCR (qPCR). We defined the subsets as pro/pre (B220+CD43+IgM−IgD−), immature (B220+CD43−IgM+IgD−), B1 (FSChiB220loCD19+IgMhiIgD−), follicular (B220hiCD23+CD21+) and marginal zone (MZ, B220hiCD23−CD21hi) B cells; and double-negative (DN, CD4−CD8−), double-positive (DP, CD4+CD8+) and CD4+ and CD8+ single-positive (SP) thymocytes. GC (B220hiGL7+CD95+) and class-switched (B220hiIgG1+) B cells were isolated from B6 mice that had been immunized with sheep red blood cells (SRBCs). We detected Themis2 mRNA in all of the B cell populations that we analyzed, with expression being 5–10-fold higher in immature, follicular and B1 B cells than MZ, GC or class-switched B cells (Fig. 1a). There was minimal expression of Themis2 mRNA in DP thymocytes (Fig. 1a). In contrast, Themis1 was expressed in DN, DP, and CD4+ and CD8+ SP thymocytes, and its absolute expression in the T-cell lineage was up to tenfold higher than that of Themis2 in B cells (Fig. 1b). Themis1 mRNA was not detected in B cells (data not shown).

Figure 1: The expression and gene targeting of Themis2.
figure 1

(a,b) mRNA expression of Themis2 (a) and Themis1 (b) in flow-sorted cells from wild-type (WT) mice. (a) Themis2 expression in BM pro/pre (B220+CD43+IgM−IgD−) and immature (B220+CD43−IgM+IgD−) B cells, peritoneal B1 (B220loCD19+IgM+IgD−) B cells, splenic follicular (B220hiCD23+CD21+) and marginal zone (B220hiCD23−CD21hi) B cells, and germinal center (B220+GL7+CD95+) and IgG1 (B220+IgG1+) B cells following immunization with SRBC. (b) Themis1 expression in thymic cell populations. Columns show means from triplicate sorts and qPCR with maximum and minimum measurements. (c) Relative expression of Themis2 splenic mRNA across indicated exon junctions. Bars show means with 95% confidence limits and comparison by unpaired t test. *P < 0.01, **P < 0.001. (d) Representative immunoblot analysis of Themis2 and β-actin protein expression in splenocytes from mice carrying wild-type, heterozygote (Themis2+/−) and homozygote (Themis2−/−) targeted alleles of Themis2 stained with a rabbit anti-Themis2 antibody14.

Full size image

To investigate the function of Themis2 in B cells, we generated Themis2-deficient mice (called Themis2−/− hereafter) by gene targeting and deleting Themis2 exon 4 in B6 × S129/Sv F1 embryonic stem cells (Supplementary Fig. 1a). Exon 4 encodes the C terminus of the first CABIT domain and the entire second CABIT domain and PxRPxK motif, and its deletion introduced a frame-shift mutation8. Themis2 mRNA and protein were absent in Themis2−/− splenocytes, as assessed by qPCR and immunoblotting (Fig. 1c,d). Liquid chromatography–tandem mass spectrometry (LC-MS/MS) peptide mapping identified 12 high-scoring peptides across Themis2 in wild-type splenocytes, including seven between amino acids 1 and 163 in exons 1–3, but none of these peptides were sequenced by MS/MS in Themis2−/− splenocytes (Supplementary Fig. 1b). In summary, we found differential expression of Themis2 and Themis1 in B cells and T cells, respectively, and confirmed the absence of Themis2 protein expression in Themis2−/− exon 4 knockout mice.

Themis2 is not essential for B-cell development or survival

Analysis of B-cell development in wild-type and Themis2−/− mice showed similar numbers of pro/pre, immature and mature B cells in the bone marrow (BM), B220+CD23−CD21− transitional, follicular and MZ B cells in the spleen and lymph nodes, and B1 B cells in the peritoneal cavity (Fig. 2a and Supplementary Fig. 2a). Serum IgM antibody titers in unimmunized wild-type and Themis2−/− mice were also equivalent (Fig. 2b).

Figure 2: Themis2 is not required for the generation of a pre-immune B cell repertoire, but does reduce BCR expression on mature B cells.
figure 2

(a) Number of B cells in six wild-type and five Themis2−/− mice, gated on transitional (B220hiCD23−CD21lo), follicular (B220hiCD23+CD21+) and MZ (B220hi CD23−CD21hi) B cells in the spleen. Columns show means and s.e.m. (b) Serum IgM antibody titers in 9 wild-type and 21 Themis2−/− adult mice, aged 8–14 weeks. Bars show means and 95% confidence limits. (c) Flow cytometry of wild-type and Themis2−/− B220+ BM B cells showing the frequency and gating of pro/pre (IgM−IgD−), immature (IgM+IgD−) and mature (IgM+IgD+). Histograms are representative of more than ten mice. (d) Mean cell surface IgM expression on B220+IgD+ wild-type and Themis2−/− B cells in BM and spleen, expressed as mean fluorescent intensity (MFI) by flow cytometry and gated as in c. Bars show means and 95% confidence limits and comparison by unpaired t test. *P < 0.05, ***P < 0.001. Data are representative of results from three independent experiments.

Full size image

Examination of the phenotype of B cells revealed higher expression of IgM on Themis2−/− B220+IgD+ follicular B cells in the BM and spleen compared with wild-type follicular B cells (Fig. 2c,d). High IgM expression has been previously associated with reduced spontaneous BCR signaling18, suggesting that Themis2 deficiency might affect BCR signaling and the selection of B cells.

To exclude the possibility that alterations in the B-cell repertoire in the Themis2−/− mice might be masked by compensatory B cell proliferation during development, we generated mixed BM chimeric mice. Equal mixes of CD45.1 wild-type and CD45.2 Themis2−/− whole BM cells were injected intravenously into lethally irradiated CD45.1 wild-type recipient mice, and these animals were compared with mice reconstituted with CD45.1 wild-type and CD45.2 wild-type donors. We found that, 8 weeks after BM transfer, the reconstitution of pro/pre, immature, transitional, follicular and MZ B cells derived from CD45.2 Themis2−/− did not differ between subsets and was similar overall to that of CD45.2 wild-type BM in the control mice (Supplementary Fig. 2b). These findings confirmed that B-cell development and B-cell expansion were normal in the absence of Themis2. In the mixed chimeras, Themis2−/− B220+IgD+ follicular B cells again showed higher expression of IgM than their wild-type counterparts, indicating that this effect was B-cell intrinsic (Supplementary Fig. 2c).

Although our qPCR analysis failed to detect Themis1 mRNA in selected B-cell subsets, it is possible that low or stage-specific expression of Themis1 might compensate for the loss of Themis2. To test this, we intercrossed Themis1−/− and Themis2−/− strains to generate Themis1−/−Themis2−/− double-deficient mice. The Themis1−/−Themis2−/− mice had a block in T-cell development at the transition from the DP to the SP stage, which was the same as that seen in Themis1−/− mice, and normal B-cell development, as seen in the Themis2−/− strain (Supplementary Fig. 3a–d). These findings identified separate and non-redundant functions for Themis1 and Themis2 in T and B cells, respectively. Although Themis2 was not essential for B-cell development in naive mice, its deficiency enhanced IgM expression on follicular B cells, a sign that has been correlated elsewhere with decreased levels of intrinsic BCR signaling18.

Themis2 increases positive selection to self and foreign antigens

Immunization of Themis2−/− mice with the T-independent antigen 4-hydroxy-3-nitrophenylacetyl (NP)-Ficoll or T-dependent antigen NP coupled to chicken gamma globulin (NP19-CGG) plus adjuvant revealed no defect in the antibody response compared with wild-type mice (Supplementary Fig. 4a,b). However, immunization with intraperitoneal SRBCs alone induced fewer GC and switched IgG1 B cells after 8 d in Themis2−/− mice than in wild-type mice (Fig. 3a,b). Immunization with a lower avidity form of the NP antigen, NP3-CGG, plus adjuvant also induced lower antibody titers in Themis2−/− mice than in wild-type mice after 7 and 14 d (Fig. 3c).

Figure 3: Reduced germinal center and class-switched B cells in the absence of Themis2.
figure 3

(a) Flow cytometry of wild-type and Themis2−/− splenic B220+ B cells in naive mice and in mice 8 d after intraperitoneal (IP) immunization with SRBCs showing the gating and typical frequency of GC (B220+GL7+CD95+) B cell numbers. Data are representative of more than 12 mice in 3 independent experiments. (b) Number of GC (B220+GL7+CD95+) B cells (top) and class-switched (B220+IgG1+) B cells (bottom) in wild-type and Themis2−/− mice, 8 d after immunization with SRBCs, with bars showing means and s.e.m. Data are representative of results from three independent experiments. (c) IgM and IgG1 anti-NP25 antibody titers in WT and Themis2−/− mice 7 and 14 d after primary immunization with NP3-CGG. Symbols represent individual mice, bars means and 95% confidence limits and comparison by unpaired t test. *P < 0.05, **P < 0.01 and ***P < 0.001

Full size image

Because subtle effects on the B cell response are difficult to detect in a polyclonal repertoire, we crossed Themis2−/− mice with the MD4 transgenic strain (hereafter referred to as IgHEL), which expresses a high-affinity IgMaIgDa BCR specific for hen egg lysozyme (HEL; Ka = 2 × 1010 M−1) derived from the BALB/c hybridoma HyHEL-10 (ref. 19). The development of pro/pre, immature, follicular and MZ B-cell subsets was similar in wild-type IgHEL and Themis2−/− IgHEL mice (Fig. 4a). To assess negative selection by self-antigens, we crossed wild-type IgHEL and Themis2−/− IgHEL mice to transgenic mice expressing either soluble HEL (sHEL) as neo-self antigen under the metallothionein promoter or membrane-bound antigen HEL (mHEL) under the MHC class I promoter20. Mature B220+IgD+ B cells from wild-type and Themis2−/− IgHEL-sHEL double transgenic mice showed down-modulated IgM expression, characteristic of B cell anergy19 (Supplementary Fig. 5a,b). Likewise, self-reactive immature IgHEL cells were deleted from the repertoire by self-antigen with equal efficiency in wild-type and Themis2−/− IgHEL-mHEL double transgenic mice20 (Supplementary Fig. 5c).

Figure 4: Themis2 is required for B cell selection by self and foreign antigen.
figure 4

(a) Numbers of B cell subsets in the BM and spleens of wild-type, Themis2+/− and Themis2−/− IgHEL transgenic mice (gated as in Fig. 2). Symbols represent individual mice and bars represent means and s.e.m. (b,c) Phenotype and enumeration of peritoneal B cells from wild-type and Themis2−/− IgHEL and IgHEL-mHELKK mice showing the flow cytometry gating (b) and enumeration (c) of HEL-binding B220hiIgMa+ follicular B cells in IgHEL mice (left histograms and upper graphs) and B220loIgMa hi B1 B cells in IgHEL-mHELKK mice (right histograms and lower graphs). Columns show means and s.e.m. and data are representative of three experiments (n > 10). (d) Serum anti-HEL IgMa in adult wild-type and Themis2−/− IgHEL and IgHEL-mHELKK mice. Symbols represent individual animals and bars show means and s.e.m. (e) Wild-type B6 mice were injected intravenously with 50:50 mixtures of CD45 allotype-marked WT:WT or WT:Themis2−/− IgHEL B cells and OT-II T cells and immunized with OVA-HEL/RIBI 1 d later. Flow cytometry shows typical gating of GC B cells at 8 d and their resolution into HEL+CD45.1 (wild-type) and HEL+CD45.2 (wild-type or Themis2−/−) cells. (f) Percentage of wild-type or Themis2−/− HEL+CD45.2 GC B cells at input and in the GC 8 d after immunization. Columns show means and s.e.m. and symbols represent individual mice at 8 d or separate input samples. Data were combined from three experiments and compared by unpaired t test. *P < 0.05, **P < 0.01 and ***P < 0.001.

Full size image

To explore the role of Themis2 in B-cell-positive selection, we crossed Themis2−/− IgHEL mice with transgenic mice expressing the mHELKK variant of mHEL, which is expressed under the same MHC class 1 promoter as mHEL, but is restricted to the endoplasmic reticulum by a C-terminal cytoplasmic di-lysine retention motif21. The mHELKK self-antigen induces the positive selection of IgHEL B1 cells in IgHEL-mHELKK double transgenic mice and promotes the generation of large numbers of anti-HEL-IgMa-secreting plasma cells21. We detected fewer peritoneal B1 cells by flow cytometry and lower serum anti-HEL IgMa antibody by ELISA in Themis2−/− IgHEL-mHELKK mice compared with wild-type IgHEL-mHELKK controls (Fig. 4b–d). Lower serum titers of IgMa were also seen in naive Themis2−/− IgHEL mice compared with wild-type IgHEL controls (Fig. 4d), suggesting that Themis2 is also involved in the differentiation of plasma cells in the absence of the cognate antigen.

To further assess the B-cell response to a T-dependent antigen, we immunized wild-type mice with monovalent chimeric HEL-ovalbumin (HEL-OVA) in RIBI adjuvant 1 d after adoptive transfer of equal mixes of wild-type splenic IgHEL B cells (CD45.1) and wild-type or Themis2−/− splenic IgHEL B cells (CD45.2) together with splenic CD4+ T cells from OT-II mice, which express an MHC class II I-Ab-restricted TCR transgene specific for OVA peptide 323–339. In this experimental system, the naive adoptively transferred IgHEL B cells were avidly recruited to the GC, where they bound HEL and received T cell help from the OT-II T cells. The ratio of CD45.1 wild-type:CD45.2 Themis2−/− B220hiGL7+CD95+HEL+ IgHEL GC B cells was then compared to the ratios of input cells and CD45.1 wild-type:CD45.2 wild-type controls. 8 d after HEL-OVA immunization, the number of Themis2−/− IgHEL GC B cells was significantly reduced compared with wild-type IgHEL GC B cells (Fig. 4e,f). The same experimental protocol using HEL coupled to SRBCs rather than OVA-HEL also resulted in fewer Themis2−/− IgHEL GC B cells compared with wild-type IgHEL (Supplementary Fig. 6a). Collectively, these observations indicate that, although Themis2 is not required for B-cell negative selection by systemic antigens, it increases the positive selection of B1 cells and plasma cells by intracellular self-antigens and the generation of IgM-secreting plasma cells in unimmunized mice. Themis2 also increases the magnitude of the antibody and GC B cell response to foreign antigens, notably when the avidity of these antigens is low.

Themis2 regulates the BCR response to low-avidity antigen

To investigate the response of wild-type and Themis2−/− IgHEL B cells to different forms of antigen in vitro, we stimulated 50:50 mixtures of CD45.1 wild-type and CD45.2 Themis2−/− IgHEL splenic B cells overnight with either sHEL (1–1,000 ng/ml), anti-IgM F(ab')2 fragments (0.1–10 μg/ml) or the Toll-like receptor 4 agonist lipopolysaccharide (LPS; 0.1–1,000 ng/ml) and measured the induction of the activation markers CD69 and CD86. We found that the threshold for B-cell activation by the low-avidity antigen sHEL was significantly higher in Themis2−/− IgHEL B cells than in wild-type IgHEL B cells (Fig. 5a). Similar results were found using soluble duck egg lysozyme, which has a ∼1,500-fold lower affinity for IgHEL (∼Ka = 1.3 × 107 M−1)22 (Fig. 5a). The maximum sHEL-induced CD69 and CD86 expression was also lower in Themis2−/− IgHEL B cells than in wild-type IgHEL B cells (Fig. 5b). In contrast, there was no difference in the response of wild-type or Themis2−/− IgHEL B cells to high-avidity anti-IgM F(ab')2 or to LPS at any concentration (Fig. 5c).

Figure 5: Themis2 sets the threshold for B cell activation by soluble and low-abundance membrane-bound antigen.
figure 5

(a) Percentage of CD69+CD86+ activated wild-type and Themis2−/− IgHEL B cells following 16-h stimulation with sHEL (left) or sDEL (right). Data are representative of six (sHEL) and three (sDEL) independent experiments and comparison by unpaired t test. *P < 0.05, **P < 0.01. (b) The expression of CD69 (left) and CD86 (right) on wild-type and Themis2−/− IgHEL B cells following 16-h activation with sHEL. Comparison by unpaired t test. **P < 0.01. (c) Percentage of wild-type and Themis2−/− IgHEL B cells activated (CD69+CD86+) following 16-h stimulation with anti-IgM F(ab')2 (left) or LPS (right). Data are representative of six experiments. (d) pBMN retrovirus expressing mHEL variants linked to three tandem intracellular GFP molecules was stably transfected into NIH-3T3 cells, which were incubated overnight with CD45.2 and CD45.1 allotype-marked WT:WT or WT:Themis2−/− IgHEL B cells. The percentage of activated IgHEL B cells (CD69+CD86+) following overnight incubation with NIH-3T3 cell clones expressing different low levels of mHEL (two clones), and variants mHEL2X and mHEL3X (two clones each). Data are representative of three separate experiments with different clones. Symbols represent separate cultures and bars show means from triplicate assays. The mean levels of mHEL expression on each clone relative to untransfected cells are displayed below the x axis and were calculated from flow cytometric staining with the anti-HEL antibody HyHEL9.

Full size image

We next tested whether a similar effect might occur following an encounter of rare membrane-bound antigens of the sort expressed on follicular dendritic or other lymphoid cells in vivo. To assess the response to a membrane-bound antigen with both low abundance and variable affinity for the IgHEL receptor, we generated a series of NIH-3T3 cell lines expressing comparably low amounts of mHEL variants with different affinities for the IgHEL receptor: mHEL-3T3 (Ka = 2 × 1010 M−1), mHEL2X-3T3 cells (Ka = 8 × 107 M−1) and mHEL3X-3T3 (Ka = 1.5 × 106 M−1)23. The mHEL expression was quantified by measuring the fold increase in fluorescence, compared with untransfected NIH-3T3 cells, generated by three tandem copies of GFP linked to the intracellular C terminus of each mHEL variant, and by extracellular staining with the HEL-specific antibody HyHEL9, whose binding is unaffected by the HEL2X or HEL3X mutations (Supplementary Fig. 6b). Overnight incubation of 50:50 mixtures of CD45.1 wild-type and CD45.2 Themis2−/− IgHEL splenic B cells with the mHEL-3T3 cells induced the expression of CD69 and CD86. We found that fewer CD45.2 Themis2−/− IgHEL splenic B cells upregulated CD69 and CD86 in response to the lower affinity antigens (mHEL2X or mHEL3X) at low density (fluorescence < 3× background) than wild-type IgHEL B cells (Fig. 5d). In contrast, wild-type and Themis2−/− IgHEL B cells were equivalently activated by mHEL at low or high density or mHEL3X at high density (Fig. 5d). Similar results were found when the mHEL-3T3, mHEL2X-3T3 and mHEL3X-3T3 cells coexpressed a constant amount of the integrin receptor ICAM1, indicating that failure of the B cells to engage ICAM1 on NIH-3T3 cells was not a limiting factor (Supplementary Fig. 6c). These findings demonstrate that Themis2 increases the B-cell response to low-avidity soluble antigens, regardless of affinity, and to membrane-bound antigens, when they are present at low affinity and low abundance.

Themis2 activates PLC-γ2 and downstream pathways

We then used IgHEL B cells to investigate the molecular basis of Themis2 function during BCR signaling. To assess calcium mobilization, we loaded equal mixes of CD45.1 IgHEL splenocytes and CD45.2 wild-type IgHEL or Themis2−/− IgHEL splenocytes with Pluronic F-127 and the calcium indicator Fura Red-AM and assessed calcium flux induced by sHEL or anti-IgM F(ab')2. Themis2−/− IgHEL B cells showed lower peak calcium flux in response to sHEL than wild-type IgHEL B cells, without a reduction in the time to the peak response (Fig. 6a,b). In contrast, the calcium response to anti-IgM F(ab')2 was identical in wild-type and Themis2−/− IgHEL B cells (Fig. 6a).

Figure 6: Themis2 sets the calcium, PLC-γ2 and Erk response to low-avidity, but not high-avidity, antigen.
figure 6

(a) Median calcium flux in mixed naive CD45.2 Themis2−/− and CD45.1 wild-type IgHEL splenic B cells in response to stimulation with sHEL or anti-IgM F(ab')2. Data are representative of four experiments. (b) Peak median height and time to peak height of calcium flux in response to varying concentrations of sHEL. Symbols represent individual samples. (c) Mean phospho-specific antibody binding to intracellular signaling molecules downstream of the BCR, 5 min after stimulation of mixed MACS sorted IgHEL CD45.2 Themis2−/− (open circles) and CD45.1 wild-type (closed circles) B cells with sHEL. (d) Time course of activation of p-Erk and pCD19 in CD45.2 Themis2−/− and CD45.1 wild-type IgHEL B cells following stimulation with 100 ng/ml sHEL. Results derived from triplicate samples and bars indicate 95% confidence limits with unpaired t tests. *P < 0.05, **P < 0.01 and ***P < 0.001. Graphs are representative of at least three separate experiments in each case. (e) Mean SOS fluorescence intensity in confocal images from mixed wild-type and Themis2−/− IgHEL B cells stimulated for 10 min with 5 ng/ml or 1 μg/ml HEL, and identified as wild-type or Themis2−/− by counterstaining with B220 alone (KO, Themis2−/−) or B220 and CD45.1 (WT, wild type). (f) Mean anti-SHP1 and SHP2 phospho-specific antibody binding following stimulation of 50:50 mixtures of CD45.1 wild-type and CD45.2 Themis2−/− IgHEL B cells with sHEL. Data show means from triplicate samples and graphs are representative of three or more experiments.

Full size image

To dissect the differential effects of high- and low-avidity antigen stimulation on the BCR signaling pathways, we sorted splenic follicular IgHEL B cells from CD45.1 wild-type IgHEL and CD45.2 wild-type IgHEL or Themis2−/−IgHEL mice by negative selection and stimulated equal mixes of B cells with sHEL or anti-IgM F(ab')2. Intracellular staining of the splenic follicular IgHEL B cells showed similar activation of Syk, BLNK and downstream p38 and Akt in Themis2−/− IgHEL and wild-type IgHEL B cells after stimulation with sHEL (Fig. 6c and Supplementary Fig. 7a). However, phosphorylation of both PLC-γ2 and Erk1/2 was reduced in Themis2−/− IgHEL B cells compared to the co-cultured wild-type IgHEL B cells following stimulation with sHEL (Fig. 6c,d and Supplementary Fig. 7c). These results are consistent with data indicating that upregulation of CD69 and CD86 is primarily mediated by the Erk signaling pathway24. In contrast to the findings with sHEL, there was no difference in the phosphorylation of PLC-γ2 and Erk1/2 in wild-type and Themis2−/− IgHEL B cells stimulated with anti-IgM F(ab')2 (Supplementary Fig. 6b,c).

Intracellular staining showed equivalent recruitment of the Ras guanine nucleotide exchange factor SOS1 to the cell membrane in sHEL-stimulated wild-type IgHEL and Themis2−/− IgHEL B cells (Fig. 6e), suggesting that Themis2 does not directly affect Ras-Erk activation via the alternative pathway of SOS1-dependent activation. Consistent with the lack of a discernable defect in Akt activation, CD19, which can augment calcium- and Erk-dependent signaling, was normally phosphorylated in Themis2−/− IgHEL B cells in response to sHEL (Fig. 6d). In contrast with observations reported in Themis1−/− T cells25,26, we found no observable defect in SHP1 phosphorylation in Themis2−/− IgHEL B cells compared with wild-type IgHEL B cells or in the phosphorylation of SHP2, which has been widely reported to be a positive regulator of Erk17,27 (Fig. 6f).

To gain further insight into the effect of Themis2 on Erk and Ca2+ signaling, we tested whether Themis2 associates with PLC-γ2. Immunoprecipitation of Themis2 from wild-type IgHEL B cells before and after stimulation with sHEL and anti-IgM F(ab′)2 revealed that Themis2 was constitutively associated with PLC-γ2 and the Src-kinase Lyn (Fig. 7a). Consistent with previous reports7,15, Themis2 also co-immunoprecipitated Grb2. To explore the importance of these associations in more detail, we expressed Flag-tagged Themis2, Lyn, Syk and PLC-γ2 in HEK293 cells. Confirming the results in primary B cells, Themis2 associated with both PLC-γ2 (Fig. 7b,c) and Lyn (Fig. 7d,e) in transfected HEK293 cells. A mutant Themis2 protein lacking the PxRPxK Grb2 binding motif (Themis2-dPRR) also immunoprecipitated Lyn and PLC-γ2 (Fig. 7f and Supplementary Fig. 7d,e), indicating that the association between Themis2, PLC-γ2 and Lyn was not mediated by Grb2. The association between Themis2 and PLC-γ2 in HEK293 cells was only slightly enhanced by PLC-γ2 phosphorylation (Fig. 7c); however, Lyn-mediated phosphorylation of Tyr759 on PLC-γ2, which has been shown to augment PLC-γ2 enzymatic activity28,29, and Lyn-mediated phosphorylation of Tyr1217 on PLC-γ2 were enhanced in the presence of Themis2 (Fig. 7e,g). These results demonstrate that Themis2 interacts constitutively with Grb2, Lyn and PLC-γ2 and increases the activation of PLC-γ2 and downstream pathways in response to low-avidity antigens.

Figure 7: Themis2 binds and facilitates the activation of PLC-γ2.
figure 7

(a) Immunoblot of crude lysate and Themis2 immunoprecipitates (IPs) from wild-type and Themis2−/− IgHEL B cells, stimulated for 5 min with medium, 100 ng/ml sHEL or 10 μg/ml anti-IgM F(ab')2. Data are representative of three independent experiments using four wild-type and four Themis2−/− mice each. (b) Immunoblot of anti-Themis2 and anti-PLC-γ2 IP from HEK293 cells expressing combinations of Flag-tagged Themis2 and PLC-γ2. (c) Immunoblot of anti-Themis2 and anti-PLC-γ2 IP from HEK 293 cells expressing combinations of Flag-tagged Themis2, Lyn and PLC-γ2. Y1217 phosphorylation denotes PLC-γ2 activation. (d) Immunoblot of anti-Themis2 IP from HEK 293 cells expressing combinations of Flag-tagged Themis2, Lyn, Syk and PLC-γ2. (e) Immunoblot of anti-Lyn IP from HEK 293 cells expressing combinations of Flag-tagged Themis2, Lyn and PLC-γ2. Y759 phosphorylation denotes PLC-γ2 activation. (f) Immunoblot of anti-FLAG IP from HEK 293 cells expressing combinations of Flag-tagged Themis2, Lyn and PLC-γ2. (g) Immunoblot of anti-Themis2 IP from HEK 293 cells expressing combinations of Flag-tagged PLC-γ2, Lyn and Themis2, and tagged Themis2-dPRR. Data are representative of three or more independent experiments.

Full size image

Discussion

We found that Themis2 sets the threshold for B-cell activation by low-avidity antigens, increasing the activation of PLC-γ2 and downstream pathways. Themis2 increased the positive selection of B cells by both self and foreign antigens, but was not required for negative selection. Our findings also suggest that Themis2 may increase constitutive activation by the BCR in the absence of self or foreign antigen, leading to IgM downregulation and an increased rate of spontaneous plasma cell differentiation.

Loss of Themis2 raised the threshold for the antigen-induced upregulation of CD69, which is required for localization of activated B cells in the lymph node30, and reduced the expression of CD86, which is required for B cells to receive T cell help. Themis2 may therefore be required to increase the sensitivity of naive B cells to antigens of low abundance. Because Themis2 is a putative adaptor protein and lacks any known enzymatic activity, its binding to both Lyn and PLC-γ2 may facilitate activation of PLC-γ2 in response to BCR stimulation by stabilizing the interaction of Lyn and PLC-γ2. The precise nature of this mechanism of activation and whether the unique CABIT domains in Themis2 have a role in this process remain to be determined.

B-cell responses to antigens of relatively low avidity are likely to be important in physiological settings, especially during the initial phase of T-dependent antibody responses; however, in vitro and biochemical studies of B-cell activation have almost exclusively used high-avidity analogs of T-independent antigens, typically anti-IgM F(ab')2. Our data suggest that these forms of antigenic stimulation are likely to miss important aspects of differential BCR signaling. Antigens of different affinity are routinely used to assess T cell responses, and our findings indicate that similar approaches need to be implemented in the study of B-cell selection.

It was recently reported that Themis2 deficiency had no effect on B cell development or antibody titers following immunization with NP-Ficoll and NP21-CGG or B-cell activation by anti-IgM in vitro31. However, that study did not include an analysis of lower valency forms of antigen such as NP3-CGG or sHEL. Another recent study implicated Themis1 in transmitting TCR signals from low-affinity, but not high-affinity, antigens and in attenuating rather than increasing signal strength17. The authors compared developmentally matched OVA-specific T cells and found an increase in TCR-dependent calcium flux and Erk signaling in the absence of Themis1. From these results, they proposed that, in the absence of Themis1, signals generated from weak TCR-ligand interactions are enhanced, converting positive to negative selection, and attributed this in part to a failure to activate the inhibitory phosphatase SHP1 and/or recruit SHP1 to LAT17,25. These results seem inconsistent with our finding that Themis2 is a positive regulator of signaling in B cells. Our data indicate that loss of Themis2 does not diminish sHEL- or anti-IgM-induced phosphorylation of SHP1 in B cells. Moreover, the effects of Themis2 and SHP1 deficiency on IgHEL B cells are very different, with the latter being characterized by increased spontaneous and antigen-induced B-cell signaling to sHEL and anti-IgM F(ab')232.

The fact that Themis2 can substitute for Themis1 in T-cell development implies that, at least in thymocytes, the two proteins may have a similar mechanism of action. However, cell-specific dissimilarities, including redundancy, and a notable difference in the level of cellular expression of Themis2 in T cells and Themis1 in B cells may mean that this interpretation is too simplistic. For example, the phosphatase SHP2 binds to the same Themis1-Grb2 complex as SHP1 (ref. 17), but is a positive regulator of Erk downstream of a variety of receptors33. SHP2 activation appears to be unimpaired in Themis2-deficient B cells; in principle, however, differential signaling in B and T cells might be a result of competition between positive and negative regulators linked to Themis-family proteins. The amount of Themis-family proteins and other binding partners, including inhibitory co-receptors, could result in different outcomes in different cells.

A further layer of complexity and control is likely to be added by the cellular compartmentalization of signaling34. Most cellular Erk activation occurs on internal membranes, in a RasGRP-dependent manner, and calcium-dependent RasGAPs such as CAPRIL and RASAL are also regulated through intracellular compartmentalization35. One study linked the pattern of intracellular Ras-MAPK signaling to qualitative differences in positive and negative signaling by different antigens in the thymus36. Speculatively, Themis2 could also be affecting positive selection by recruiting PLC-γ2 locally and having specific effects on these pathways. Untangling these processes and how they are dependent on different environmental cues and developmental stages, as well as on the nature and dose of antigens, will give us a better understanding of how B cells integrate upstream signals and switch between different outcomes.

Methods

Mice.

Themis2-null mice were generated by homologous recombination and deletion of Themis2 exon 4 in C57BL/6 (B6) X S129/Sv F1 embryonic stem cells. The targeted allele was backcrossed onto the B6 background for at least 10 generations (genotyping assay available on request) before intercrossing to create homozygous Themis2 deficient animals. IgHEL (C57BL/6-Tg(IghelMD4)4Ccg/J), sHEL (C57BL/6-Tg(ML5sHEL)5Ccg/J), mHEL (C57BL/6-Tg(KLK4mHEL)6Ccg/J), mHELKK and OT-II (C57BL/6-Tg(TcraTcrb)425Cbn/Crl) mice were as described previously21 and maintained on a B6 background. For BM chimeras, B6 mice were irradiated with two doses of 4.5 Gy spaced by 3 h and were injected with at least 5 × 106 bone marrow cells (single samples or 50:50 mixture of WT B6.SJL CD45.1+ BM and either Themis2−/− or wild-type B6 (CD45.2+) BM. They were allowed to reconstitute for 8–10 weeks before immunization or analysis. All experiments included age and sex-matched littermate control animals. All experiments were approved by the NIHR or the Oxford University Ethical Review Committee and done under UK Home Office License.

Flow cytometry.

Cell suspensions from BM (one femur and tibia), spleen, thymus, mesenteric lymph nodes and peritoneal cavity were counted on a hemocytometer and stained, as described previously21 with mAbs against the following antigens (from eBioscience, unless otherwise stated): B220, RA3-6B2-allophycocyanin (APC), phycoerythrin (PE) or fluorescein isothiocyanate (FITC); CD19, eBio1D3-PE–indotricarbocyanine (Cy7); CD21/35, 7G8-FITC (BD Pharmingen); CD24, M1/69-FITC; CD4, GK1.5-FITC or PE; CD43, eBioR2/60-PE; CD45.1, A20-APC or PE-Cy7 (BD Pharmingen); CD45.2, PE-104, Pacific blue-104 (Biolegend); CD5, 53-7.3-peridinin chlorophyll protein (PerCP; BD Pharmingen); CD69, H1.2F3-FITC; CD8, 53-6.7-PE; CD86, GL1-PE; CD95, 15A7-FITC, GL-7, G7-PE; IgMa, DS-1–PE (BD Pharmingen); IgDa, AMS9.1-FITC (BD Pharmingen); IgD, 11-26-FITC (BD Pharmingen); IgM, 11/41-PE; IgG1, P3-PE; Live/Dead Indicator for 488 nm laser and 633 nm laser (Invitrogen). HEL-binding cells were detected by incubating cells with 200 ng/ml of unlabeled HEL and counterstaining with HyHEL9 Tricolor (Tc). Intracellular staining was performed using the Cytofix/Cytoperm buffer (BD Bioscience) and the following antibodies against phosphorylated epitopes, which were from Cell Signaling Technology, unless otherwise stated: p-p38, Rabbit IgG anti-pThr180/pTyr182, clone 3D7; p-Akt, Rabbit IgG anti-pSer473, clone 193H12; p-Akt, Rabbit IgG anti-pThr308, clone C31E5E; p-BLNK, Mouse IgG anti-pTyr84, clone J117-1278 (BD Pharmingen); p-Erk1/2, Rabbit IgG anti-pThr202/pTyr204, (Erk1) and anti-pThr185/pTyr187 (Erk2), clone D13.14.4E; p-PLC-γ2, Mouse IgG anti-pTyr759, clone K86-689.37.73; p-Syk, Mouse IgG anti-pTyr319/pTyr352, clone 17A/P-ZAP70 (BD Pharmingen); p-CD19, Rabbit anti-pTyr531; p-SHP1, Rabbit anti-pTyr564; p-SHP2, Rabbit anti-pTyr580. Data were acquired on a FACSCanto, FACSCalibur, FACSort or LSR II (BD) and were analyzed with FlowJo Software (Tree Star).

Quantitative RT-PCR.

Total RNA was extracted from splenocytes or flow cytometry sorted populations with RNEasy kits (Qiagen) followed by cDNA synthesis using the Superscript III First Strand Synthesis SuperMix for quantitative RT-PCR (qRT-PCR) kit (Life Technologies). Samples were analyzed on a StepOnePlus Real-Time PCR System (Applied Biosystems), using multiple wavelength detection for control and target (Themis1, Themis2 and Actb mRNA), which were probed using the TaqMan Gene Expression Assays and Power SYBR Green PCR Master Mix (Life Technologies). Data from Themis2 expression across cells were normalized to Actb and analyzed using the comparative threshold cycle method. Themis2 mRNA expression in wild-type and Themis2−/− (five mice of each genotype) was assayed using primers spanning adjacent exons (available on request) and normalized against Cd19, for example ΔCT WT1 (Exon 2-3) = CT WT1 (Exon 2-3) – CT WT1 (Cd19).

LC-MS/MS.

Splenocytes from wild-type and Themis2−/− mice (two of each genotypes) were lysed and sonicated in 50 mM Tris pH 8.0, 2%SDS. Protein extracts were submitted to cysteine reduction and alkylation with iodoacetamide, and loaded onto a 1D SDS-PAGE gel, to isolate the whole protein sample in a single band after brief electrophoretic migration. Proteins were in-gel digested overnight with sequencing grade trypsin (Promega). Resulting peptides were extracted and analyzed by nanoLC-MS/MS with an UltiMate 3000 RSLCnano system (Dionex) coupled to a Q-ExactivePlus mass spectrometer (ThermoScientific), using a 300-min acetonitrile gradient separation on an in-house packed reverse phase C-18 analytical column (75-μm inner diameter × 50 cm), to optimize proteomic coverage. Raw MS files were analyzed with the MaxQuant software (version 1.5.2.8). MS/MS sequencing data was searched with the Andromeda search engine against “mouse” entries of the UniProtKB/Swiss-Prot protein database (release 2015_05). Validation of the identifications was performed with a false discovery rate set to 1% at protein and peptide sequence match level. For label-free relative quantification of the wild-type and Themis2−/− samples, the “match between runs” option of MaxQuant was enabled to allow cross-assignment of MS features detected in the different runs, in order to confirm the absence of Themis2 peptide MS signal in the null allele samples.

Immunizations and enzyme-linked immunosorbent assays.

Mice were injected IP with two 150-μl injections of 0.9% (vol/vol) sterile saline containing a total of 50-μg NP25-aminoethyl carboxymethyl–Ficoll (Biosearch Technologies), or 50 μg alum-precipitated NP19-CGG or NP3-CGG (Biosearch Technologies). ELISA plates were coated overnight at 4 °C with NP25-BSA or NP5-BSA (5 μg/μl; Biosearch Technologies), blocked with 5% (wt/vol) milk, and serum samples were serially diluted in 1% (wt/vol) milk. Plates were incubated with horseradish-peroxidase-conjugated goat antibody to mouse IgM, IgG3 or IgG1 (Southern Biotechnology) and developed with SureBlue TMB Microwell Peroxidase Substrate and TMB Stop Solution (KPL). Absorbance was measured at 450 nm on a BioTek ELx808 ELISA plate reader. Background for assays was determined by incubation of serum from naive animals. Total IgM, and IgG ELISAs and the anti-HEL capture ELISA were performed as described previously21. For SRBC immunization, mice were given IP injection into both flanks of a total of 200 μl SRBCs (Patricell) diluted 1.5 times in PBS. Conjugation of SRBC to HEL was as described previously21.

HEL-OVA antigen was generated as a His-tagged monomeric chimeric molecule transiently expressed in HEK293T cells and purified by affinity chromatography, followed by FPLC. 500 μg purified HEL-OVA was mixed with RIBI Sigma Adjuvant System (S6322) in 3 ml PBS and mice were injected IP with two 150-μl injections containing a total of 50 μg HEL-OVA.

Ex vivo culture and stimulation.

B cells were sorted from RBC-lysed single-cell suspensions by magnetic negative depletion using biotinylated Abs (from eBioscience) against CD43 (eBioR2/60-Bi), CD11c (N418-Bi), CD11b (M1/70-Bi) and Ly76 (TER119-Bi) and streptavidin-Dynabeads. B cells were cultured in DMEM, 100 mM nonessential amino acids, 20 mM HEPES buffer, 10% FCS, 100 U/ml penicillin, 100 mg/ml streptomycin, 2 mM L-glutamine, 100 mM 2-mercaptoethanol. IgHEL B cells were stimulated with anti-IgM F(ab')2 (Jackson Immunoresearch), LPS (Alexis), or sHEL (Sigma-Aldrich). CD45.1/2 allotype marked mixtures of wild-type and Themis2−/− IgHEL B cells (total 2 × 106 cells) were incubated at 37 °C in 5% CO2 in 0.25 ml of complete medium. After 16 h, cells were analyzed by flow cytometry for surface expression of CD69 and CD86.

Constructs expressing mHEL were generated in the retroviral vector pBMN using cDNA constructs for wild-type HEL and HEL2X and HEL3X variants, which were generated by mutagenesis. Retroviruses were stably transfected into NIH-3T3 cells, with and without co-expression of ICAM-1. Chimeric proteins contained the MHC Class I transmembrane domain, as used previously in mHEL22 and three C-terminal tandem copies of eGFP on the intracellular surface. mHEL expression was quantified by measuring eGFP fluorescence and binding to the HEL-specific monoclonal antibody HyHEL9-Tc, whose epitope is not affected by the 2X and 3X mutations. Cell lines expressing low levels of mHEL were selected by flow cytometry and then cultured overnight with 50:50 mixtures of CD45.1 wild-type IgHEL and CD45.2 Themis2−/− IgHEL B cells.

Constructs expressing Themis2 were subcloned into pFLAG-CMV2 vector by PCR with mouse cDNA for Icb1 from Imagenes. Constructs expressing Lyn, Syk and PLC-γ2 were purchased from Origene. The Themis2-dPRR construct was generated by sequentially inserting two PCR fragments into the BglII, KpnI, and SalI sites of pFLAG-CMV2 vector using Quick-Fusion (Bimake). The first PCR fragment was amplified using primers 5′-ATTCATCGATAGATCTAGAGCCGGTGCCGCTGCAGGAC-3′ and 5′-CTCTAGAGTCGACTGGTACCGGCTTGGCTCTCAGAGGCTG-3′. The second PCR fragment was amplified using primers 5′-CTGAGAGCCAAGCCGGTATGAATAAGAAACAGCAGAACATAC-3′ and 5′-ATCCTCTAGAGTCGACTCAAATTTCTTCATAGTCATG-3′.

Calcium flux analysis.

Wild-type and Themis2−/− IgHEL B cells were first stained separately for B220 using different fluorochromes and a live/dead indicator. Cells were mixed in equal parts and resuspended in HBSS Buffer containing 0.02% Pluronic F-127 and 1 μg/ml Fura-Red, AM (Invitrogen), already warmed to 37 °C. The cells were incubated at 37 °C for 30 min before washing and resuspending at 107 cells/ml in HEPES buffered HBSS, pH 7. Analysis was performed using an LSRII equipped with the following lasers: blue (488 nm, 80 mW), green (532 nm, 150 mW), red (642 nm, 40 mW) and violet (406 nm, 25 mW). Calibration was performed using CS&T Beads (BD Biosciences). Before analysis by flow cytometry, cells and 4X stimulant were warmed for 5 min in a 37 °C water bath. Increases in FuraRed emission were monitored off the Violet laser (630LP and 660/20 BP), while a decrease in emission was detected off the Green laser (685LP and 710/50 BP). The ratiometric, 'FuraRed Ratio' was calculated as the increasing/decreasing signal using the Kinetics tool in FlowJo software version 9.3.3 (Tree Star).

Immunoprecipitations and immunoblotting.

Cells were lysed with lysis buffer (10 mM Tris pH 7.5, 150 mM NaCl, 2 mM EGTA 50 mM β-glycerophosphate, 1% Nonidet P-40, 2 mM Na3VO4, 10 mM NaF, protease inhibitor cocktail (Roche)). Cell lysates were cleared by centrifugation at 14,000 g for 10 min. Cell lysates were pre-cleared with Protein G-Sepharose beads (GE Healthcare) or Protein A-Dynabeads (Life Technologies) for 1 h and immunoprecipitated with antibodies conjugated to Protein G-sepharose beads or Protein A-Dynabeads for 4 h to overnight, followed by three washes in ice-cold lysis buffer. Antibodies used for immunoprecipitation were anti-PLC-γ2 (SC407, SantaCruz), anti-Lyn (SC15, SantaCruz), anti-FLAG (F1804, Sigma-Aldrich), and anti-Themis2 (affinity purified rabbit polyclonal raised against the C-terminal peptide residues CKISVHKKDRKPNPQTQN of mouse Themis2)14.

Immunoblots were performed as previously described21 and probed with anti-PLC-γ2 (SC407, SantaCruz), anti-Grb2 (3972, Cell Signaling), anti-Lyn (628102, BioLegend), anti-Phosphotyrosine, 4G10 Platinum (16-452, Millipore), anti-FLAG (F1804, Sigma-Aldrich), anti-Themis2 (ref. 14), anti-Myc (M047-3, MBL International), anti-pTyr759-PLC-γ2 (3874, Cell Signaling), and anti-pTyr1217-PLC-γ2 (3871, Cell Signaling). Blots were developed with HRP conjugated goat anti-rabbit or anti-mouse antibodies followed by enhanced chemiluminiscence (SuperSignal West Pico Chemiluminescent Substrate, ThermoScientific).

Confocal microscopy.

Equal mixtures of purified CD45.1 wild-type IgHEL and CD45.2 Themis2−/− IgHEL B cells in complete medium, were warmed at 37 °C and stimulated with sHEL at varying concentrations for 10 min before adding an equal volume of 4% (wt/vol) of paraformaldehyde in PBS and cyto-spinning onto poly-L-Lysine slides. Slides were stained with CD45.1-FITC, B220-APC and anti-SOS (rabbit anti-SOS1/2 Santa Cruz), followed by Goat anti-Rabbit IgG-AlexaFluor-594 before imaging on a Zeiss 510 Metahead Confocal microscope. Membrane associated SOS was quantified by measuring its mean fluorescence density.

Statistics.

GraphPad Prism Software was used for statistical analyses, and unpaired, two tailed Student's t tests were used for statistical comparison between groups, unless otherwise specifically mentioned.