Objectives Altered signalling in B cells is a predominant feature of systemic lupus erythematosus (SLE). The genes BANK1 and BLK were recently described as associated with SLE. BANK1 codes for a B-cell-specific cytoplasmic protein involved in B-cell receptor signalling and BLK codes for an Src tyrosine kinase with important roles in B-cell development. To characterise the role of BANK1 and BLK in SLE, a genetic interaction analysis was performed hypothesising that genetic interactions could reveal functional pathways relevant to disease pathogenesis.
Methods The GPAT16 method was used to analyse the gene–gene interactions of BANK1 and BLK. Confocal microscopy was used to investigate co-localisation, and immunoprecipitation was used to verify the physical interaction of BANK1 and BLK.
Results Epistatic interactions between BANK1 and BLK polymorphisms associated with SLE were observed in a discovery set of 279 patients and 515 controls from northern Europe. A meta-analysis with 4399 European individuals confirmed the genetic interactions between BANK1 and BLK. As BANK1 was identified as a binding partner of the Src tyrosine kinase LYN, the possibility that BANK1 and BLK could also show a protein–protein interaction was tested. The co-immunoprecipitation and co-localisation of BLK and BANK1 were demonstrated. In a Daudi cell line and primary naive B cells endogenous binding was enhanced upon B-cell receptor stimulation using anti-IgM antibodies.
Conclusion This study shows a genetic interaction between BANK1 and BLK, and demonstrates that these molecules interact physically. The results have important consequences for the understanding of SLE and other autoimmune diseases and identify a potential new signalling pathway.
Statistics from Altmetric.com
Systemic lupus erythematosus (SLE) is a complex autoimmune disease in which B-cell activity plays a major role in its development and clinical expression through the production of auto-antibodies and antigen presentation. Therefore, susceptibility genes co-expressed in B cells are interesting candidates to be tested for genetic and functional interactions.
In humans, polymorphisms of the BANK1 gene have been associated with susceptibility to SLE in European and Asian populations.1,–,3 BANK1 is located on chromosome 4q24 and codes for an adaptor/scaffold protein of 785aa (full-length isoform) primarily expressed in B cells. BANK1 protein has 13 tyrosines susceptible to phosphorylation, two ankyrin repeats, a conserved Dof, BCAP and BANK (DBB) domain, and a coiled-coil motif. It was identified as a binding partner of LYN, and it is also phosphorylated by SYK.4 BANK1 protein binds the IP3 receptors type 1 (IP3R-1) and 2 (IP3R-2) and promotes their LYN-mediated phosphorylation to induce Ca2+ mobilisation from endoplasmic reticulum stores.4 However, Ca2+ mobilisation was not impaired in a Bank1 knock-out mouse.5 Furthermore, the Bank1-deficient mouse showed a slight increase in germinal centre formation and increased T-dependent responses with activation of Akt dependent on CD40 signalling. These features were subtle and no autoimmune phenotype was investigated.
BLK was also recently identified as a susceptibility gene for SLE.6,–,8 The genetic polymorphisms of BLK associated with SLE, rs1327713 and its proxy rs2736340, are located in the promoter of BLK and the risk genotypes are correlated with reduced gene transcript levels. BLK is a Src tyrosine kinase specifically expressed in the B-cell lineage.9 A knockout mouse for Blk did not show any phenotype and BLK was deemed to be redundant in B-cell development and immune responses.10
In this study we tested whether BANK1 and BLK polymorphisms associated with SLE showed a genetic epistatic interaction, but we also extended our study to analyse whether BANK1 and BLK, like LYN and BANK1,4 could show a protein–protein interaction. While we identified an interaction between polymorphisms in both genes, we also found that both proteins immunoprecipitated and their co-expression influenced the subcellular location of the kinase. As the genetic interaction involves risk variants correlated with gene expression, the genetic interaction might reflect an imbalance in gene expression. The relative amounts of the gene products could be important to maintain the homeostasis of a common pathway.
Materials and methods
Patients and controls
We extracted data from an Affymetrix 100 k single nucleotide polymorphism (SNP) genome-wide association scan conducted in 279 cases with SLE and 515 controls from northern Europe.1 Individuals used for the 100 k GWAS have been described.1 Two independent sets of cases and controls were used for replication. Set 1 (‘USA’) is a European–American multicentre cohort of 621 cases and 774 controls. The second set (‘Europe’) comprised 1697 SLE cases and 1550 sex and ethnically matched controls from a European multicentre collection (BIOLUPUS) including German, Italian, Argentinean and Spanish individuals.
Genetic outliers with less than 90% European ancestry were removed, as estimated using principal component analysis and the clustering algorithms implemented in EIGENSTRAT and STRUCTURE software, respectively, based on genotype data from 350 ancestry informative markers or genome-wide data (available for the Argentineans and north Europeans). All SLE cases met at least four of the 11 classification criteria of the American College of Rheumatology.11 All individuals provided informed consent as approved by the recruiting site institutional review boards at each of the affiliate institutions. All clinical investigation has been conducted according to the Declaration of Helsinki.
The Swedish individuals were genotyped using the 100 k Affymetrix SNP array as described.1 The previously associated SNP for BLK (rs2736340), which is not included in the 100 k Affymetrix SNP array, was genotyped by TaqMan (ABI, Foster City, California, USA) predesigned genotyping assays. SNP showing genetic interaction with BANK1 in the 100 k were selected for replication. Replication set 1 (‘USA’) SNP were genotyped on the BeadExpress Illumina system (San Diego, CA, USA). SNP rs10516483 (BANK1) was not available for this dataset. Genotyping of set 2 (Europe) was performed for SNP rs10516487 and rs10516483 (BANK1), rs1478895 and rs2736340 (BLK) also using TaqMan. Only individuals with a genotyping rate greater than 90% were used for analysis.
From the 100 k Affymetrix SNP array data, nine tag SNP in BANK1 (rs7675129, rs10516487, rs10516483, rs2850390, rs1872701, rs10516490, rs1395306, rs871153 and rs238486) were individually tested for 16 types of interaction against seven tag SNP in BLK (rs1478895, rs1478890, rs2252534, rs1382566, rs9329246, rs7014565 and rs2061830). SNP were filtered as following Hardy–Weinberg equilibrium in controls (p>0.01) and having a missing data rate per SNP of less than 5%. Only markers with minor allele frequencies greater than 30% in controls and greater than 10% in cases, and minor genotype frequencies greater than 10% in controls and greater than 5% in cases were used. The rationale was that we wanted to screen only common variants of the general population (controls) in order to have enough two-SNP combinations and we did not want to miss some SNP that would be less common in the SLE population. Linkage disequilibrium blocks were determined using the method of Gabriel et al12 and tag SNP were selected not to be in strong linkage disequilibrium (r2<0.80). BLK SNP covered 22% while BANK1 SNP covered 44% of the alleles in each genomic region at a r2 greater than 95%.
For the replication stage, SNP following Hardy–Weinberg equilibrium in controls (p>0.001) and with missing data rates per SNP of less than 10% were included in the analysis. None of the SNP had significant differences in missing data between cases and controls (p>0.05).
Genetic interaction analysis
We used the GPAT16 method of Wirapati et al.13 In brief, this method tests the genetic interaction between every pair of non-correlated SNP (r2<0.8) by recording the 16 possible contingency tables formed by the combinations or co-occurrences of alleles or genotypes of both SNP under dominant and recessive models. For each contingency table, a Pearson score S is computed with its corresponding P value. A P<1×10–5 was considered significant. A significant interaction reflects the sum of additive (or main effects) and epistatic effects for a specific genotype combination (dominant or recessive). In this particular experiment our total number of tests performed was 504 (nine BANK1 bait SNP × seven BLK SNP × 16 tests/2). GPAT16 makes 16 tests, but the total number is divided by 2 because each interaction is tested in only one direction.
To determine the epistatic effect, that is, the increase in risk and an association odds ratio higher than expected under the null hypothesis of independence, each interaction is computed as the difference between the observed Pearson score S of each contingency table and the expected Pearson score S0 under the null hypothesis of no epistasis.14 By doing so, it derives an epistasis-like score (Se=S–S0). An epistasis P value (Pe) is obtained through permutation. A Pe<1×10–3 was considered significant. This score is the difference of two dependent scores, each one following asymptotically a 1–df c2. Therefore it does not follow any known statistical law and p values pet have to be empirically determined by permutation. If two genotypes when combined have a significant association (S score significant, P<1×10–5) but there is no significant epistatic effect (Pe>1×10–3) we conclude that such an association is mainly due to the sum of the individual or marginal effects of the associated genotypes. If the epistatic effect is significant (Pe<1×10–3) we then refer to it as a genetic epistatic interaction.
Protein interaction experiments
The synthesised peptide ETKHSPLEVGSESSC was used to immunise rabbits to generate polyclonal anti-human BANK1 anti-sera (ET-BANK antibody) and affinity purified using the SulfoLink Kit (Pierce, Thermo Scientific, Rockford, IL).Additional antibodies include anti-mouse and anti-rabbit Alexa Fluor488, anti-mouse and anti-rabbit Alexa Fluor647, anti-V5 (Invitrogen, Carlsbad, California, USA), anti-Flag M2 monoclonal and rabbit anti-Flag (Sigma, Sigma-Aldrich, St Louis, MO), anti-rabbit and anti-mouse IgG HRP (Zymed, San Francisco, California, USA). Mouse anti-human BLK antibodies were from Santa Cruz Biotechnology (Santa Cruz, California, USA) and anti-β-tubulin from Sigma-Aldrich (St Louis, Missouri, USA).
BANK1 and BLK sequences were amplified by PCR using complementary DNA from human blood and the BJAB cell line, respectively, and open reading frames were cloned in pcDNA3.1D/V5-His (Invitrogen) and confirmed by sequencing. Proteins tagged by V5 and His epitopes at the C-terminal were produced by stop codon deletion. The N-terminal FLAG-tagged BANK plasmids were constructed by sequential PCR using overlapping primers. The amplified product coding FLAG fused to BANK1 variants was cloned into pCR4-TOPO (Invitrogen) excised by EcoRI and BamHI and directionally subcloned into pIRESS2-EGFP (Clontech, Mountain View, California, USA). Sequences of the constructs are available upon request.
Co-immunoprecipitation and immunoblot analysis
Embryonic kidney HEK293T cells were seeded on six-well plates and transfected with 4 µg of expression plasmids containing FLAG-tagged BANK1 and V5-tagged BLK using Lipofectamine 2000. At 40 h cells were solubilised in triton X-100 buffer (1% triton X-100, 50 mM HEPES pH 7.1, 150 mM NaCl, 1 mM EDTA, 2 mM Na3VO4, 10% glycerol, 0.1% sodium dodecylsulphate (SDS)) containing protease inhibitors (Roche, Indianapolis, Indiana, USA) and 1 mM phenyl-methylsulphonyl fluoride. Aliquots of precleared lysates were saved for input analysis and the remaining lysate was incubated with rabbit anti-FLAG or mouse anti-V5 and immobilised with A or G-sepharose beads (GE Heathcare, Uppsala, Sweden), respectively. Beads were washed with 1:1 triton X100 buffer:phosphate-buffered saline and immunoprecipitates eluted with SDS sample buffer boiling 5 min. SDS–polyacrylamide gel electrophoresis (PAGE) and immmunoblotting were carried out using standard protocols.
Primary B-cell separation and purification
Peripheral blood mononuclear cells from buffy coats were isolated by Ficoll-Paque Plus (GE Healthcare) density gradient centrifugation. For the preparation of purified, unmanipulated naive B cells, peripheral blood mononuclear cells were subjected to negative selection using the naive B cell isolation kit II (Miltenyi Biotec, Auburn, California, USA). For depletion of CD10+ transitional B cells from negatively selected CD19+CD27-naive B cells, selected cells were incubated with anti-human CD10 microbeads (Miltenyi Biotec). Cells were magnetically separated with MACS columns and MACS separator (Miltenyi Biotec). The negatively selected naive B cells consisted of more than 95% CD19+CD27– cells.
Primary naive B cells (3×106 per condition) were treated without (–) or with (+) aIgM (10 µm/ml) for 10 min in serum-free RPMI medium. Cell extracts were made from the treated cells and subjected to immunoprecipitation and western analysis. Antibodies against human BANK1 and BLK were purchased from Santa Cruz Biotech, Inc. and Abnova Corporation (Heidelberg, Germany), respectively. Recombinant protein-G sepharose 4B beads were obtained from Invitrogen. Cell extracts were prepared using the lysis buffer containing 1% triton X100, 50 mM Tris pH 7.4, 50 mM NaCl, 1 mM EDTA, 2 mM Na3VO4 and protease inhibitor cocktail from Roche. Immunoprecipitation was carried out using anti-human BANK1 antibody overnight. The immunocomplexes were precipitated using protein-G beads and washed three times with lysis buffer.
The precipitated complexes were mixed with SDS–PAGE sample buffer from Invitrogen and the proteins were resolved in 4–12% gradient NuPAGE gel (Invitrogen). Western blot was carried out using standard protocols.
Transfected cells were fixed for 20 min at room temperature with 3.7% paraformaldehyde in phosphate-buffered saline/0.18% triton X and permeabilised in ice-cold 50:50 methanol-acetone at –20°C for 10 min. After blocking in 3% bovine serum albumin, 3% goat serum in PBS with 0.1% Tween, antibodies were diluted in blocking buffer and incubated overnight at 4°C. Fluorochrome-conjugated secondary antibodies were incubated for 2 h at room temperature and counterstained with SlowFade antifade with DAPI (Invitrogen). Fluorescence fusion proteins were visualised directly after fixation, FX enhancer treated (Invitrogen) and mounted in Vectashield (Vector Laboratories, Burlingame, California, USA).
Confocal microscopy was performed using a Zeiss 510 meta confocal scanning microscope with Zeiss plan-Apochromat 63× oil-immersion objective. Dual or triple-colour images were acquired by consecutive scanning with only one laser line active per scan to avoid cross-excitation. Image analysis was prepared using ImageJ and Adobe Photoshop.
Genetic interactions with BANK1
In the initial gene interaction analysis performed in the north European set, we observed a genetic interaction between BLK and BANK1, with the strongest epistatic effect between the BANK1 SNP rs10516483 and the BLK SNP rs1478895 (Pe=0.0001) (table 1). Two SNP in BANK1 (rs10516483 and rs10516487, D‘=0.86, r2=0.36) and two in BLK (rs1478895 and rs2736340, D‘=0.93, r2=0.06) were involved in significant interactions although they did not reach the Pe<10–3 threshold. Given the moderate sample size of the north European dataset, we chose these four SNP for replication in two larger and independent sets of cases and controls of European ancestry. We observed significant interactions between BANK1 and BLK across all datasets (table 1). The strongest association was displayed by the combination of recessive genotypes of BANK1 rs10516487 (GG) and dominant genotypes of BLK rs2736340 (TT+TC) (Pmeta-analysis=1.75×10–15) by using a total of 4399 samples. A significant Pe was demonstrated for this association in replication set 2 from Europe (Pe=0.0013) (table 1). In this set, a significant epistatic interaction was also observed between BANK1 rs10516483 (CC) and BLK rs2736340 (TT+TC) genotypes (Pe=0.0024).
Biochemical interaction between BANK1 and BLK proteins
The fact that BANK1 was identified as a partner of LYN,4 a Src tyrosine kinase, led us to test whether BANK1 would show a similar interaction with Blk, also a Src tyrosine kinase. We found that BANK1 and BLK co-immunoprecipitated each other in co-transfected HEK293T cells (figure 1A,B). As the products of co-transfection could result in an enhanced artifactual binding, we then tested whether the endogenous proteins co-immunoprecipitate in the B-cell line Daudi and in isolated naive B cells. We demonstrated co-immunoprecipitation between the endogenous BANK1 and BLK in the B-cell line (figure 1C) and in primary, naive B cells (figure 1D). We further showed that the binding was enhanced by stimulation through the B-cell receptor using anti-IgM antibodies (figure 1C,D) suggesting that activation of BANK1 or BLK may be required to enhance protein–protein interaction.
BANK1 is classified as an adaptor/scaffold protein and as such, could function to direct other molecules towards specific subcellular compartments. Confocal microscopy showed that both BANK1 and BLK co-localised in the cytoplasm when co-expressed (figure 2A–D). Interestingly, BLK localised preferentially to the plasma membrane in the absence of BANK1 (figure 2E–G), while it was mostly retained in the cytoplasm when BANK1 was co-expressed in the same cell (figure 2G). This was not the case for an irrelevant protein of similar size as BANK1 (see supplementary figure S1, available online only). In fact, BLK was located at the plasma membrane in 95% of cells when the protein was expressed alone contrary to 27% of cells co-expressing both BLK and BANK1 (figure 2H and supplementary table S1, available online only). Our results suggest that BANK1 could modulate the subcellular localisation of BLK, which would be in agreement with the function of BANK1 as an adaptor/scaffold protein.
Here, we demonstrate that two SLE susceptibility genes showing a genetic interaction, namely BANK1 and BLK, also interact physically.
We used the GPAT16 method to test for associated genotypic interactions, a method in principle similar to the multifactor dimensionality reduction15 and slightly more powerful than standard algorithms.16 According to simulations, GPAT16 is at least as powerful as the method of Marchini et al.17 The GPAT16 method enumerates exhaustively genetically relevant genotype combinations under dominant and recessive inheritance models, resembling the Batesonian definition of epistasis. This method is different from methods that consider the Fisherian definition of epistasis such as that implemented in PLINK,18 which tests the interaction term in a logistic regression model.
The genetic interactions between BANK1 and BLK observed in the north European and European datasets follow a recessive model for the BANK1 genotypes (rs10516483 CC or rs10516487 GG) and a dominant model for BLK (rs2736340 TT+TC) genotypes. The interactions described here were not observable using logistic regression as implemented in PLINK19 (see supplementary table S2, available online only), except for a weak significant interaction using the discovery set.
True epistatic interactions have been very difficult to detect and replicate.19 20 We observed in the north European set a strong epistatic effect. As there is no established P value for genetic interaction analysis, we used replication with independent sets of cases and controls. We replicated some of the epistatic effects (represented by the Pe value) that did not, however, reach our stringent Pe limit of <10–3.21 Due to the computational characteristics of the method, a meta-analysis cannot be done.
We chose to study the interaction between BANK1 and BLK because of their functional interest in relation to SLE and their role in B-cell signalling. We believe that this way of analysing genetic interactions fits our purpose of prioritising candidate interacting genes for biological validation.22,–,24 In fact, a recent paper by Sun et al25 analysed human genome protein–protein interactions and found that physical connections were preferentially involved in gene–gene interactions. Therefore, we believe that statistical genetics may guide the identification of true functional pathways in complex diseases.
Our findings point to a B-cell-specific pathway that might be relevant in lupus pathogenesis. We showed that B-cell receptor stimulation enhances BANK1 and BLK binding. Because the engagement of the B-cell receptor with anti-IgM leads to tyrosine phosphorylation of numerous proteins including BANK1, it is likely that the interaction between BANK1 and BLK is regulated by cellular kinases. In chicken cell lines, SYK is a major player in the phosphorylation of BANK1 upon BcR stimulation.4 BANK1 is a proline and tyrosine-rich protein containing several predicted motifs for binding the SH2 and SH3 domains of Src-kinases. The binding of BANK1 to the Src-kinase LYN has been demonstrated but the precise protein domains involved in the interaction have not been defined.4 Detailed mutational analyses of BANK1 and BLK would be required to understand how BANK1 interacts with this family of kinases.
The change in subcellular distribution when BLK and BANK1 are expressed simultaneously suggests two possible functional scenarios. First, BANK1 as an adaptor protein could curb the positioning of BLK at the BcR by arresting it in intracellular compartments or, alternatively, BANK1 could remove BLK from the BcR to restrict a sustained signalling. In both cases BANK1 could play an inhibitory role in B-cell activation. Supporting this idea, the bank1-deficient mouse shows an increase in B-cell activation illustrated by an increased IgM production in response to T-dependent antigens.5
It is important though to remember that the interacting SNP in BLK are located in non-coding regions. The risk genotypes of rs2736340 in BLK correlate with gene expression.6 The interacting SNP of BANK1 rs10516487 is located in exon 2 and leads to a R61H substitution but it is also a proxy of an intron 1 variant (rs17266594, r2=0.90) associated with a higher level of expression of BANK1.1 In summary, the risk allele of BLK is associated with a lower level of gene expression while the risk alleles of BANK1 are coupled with higher levels of their own gene expression (see supplementary figure S2 and supplementary methods, available online only). At this point we are unable to draw the precise mechanistic pathway to explain how the risk allele interactions lead to B-cell abnormalities. A hypothesis is that alleles affecting gene expression could impair the homeostasis of the B cell by a combinatorial inhibition model as proposed by Ferrell.25 This model claims that the signalling is impaired due to alteration of the relative concentration of the interacting proteins.
The interacting variants of BANK1 and BLK presented in this study might not be the functional variants responsible for the biological interaction effect as more extensive fine mapping and re-sequencing are required. Also, the SNP coverage would need to be increased although in detriment of multiple testing issues, particularly for whole-genome interaction analyses, which will be possible with new high-density arrays, so replication of the interactions will become even more important.
In summary, we describe here the use of a genetic interaction approach to reveal biologically relevant interactions and demonstrate that such an approach can serve to define new pathways of disease, in this particular case a B-cell-specific signalling pathway, which might be impaired in lupus patients.
The authors are indebted to the patients who have consented to have their samples used for this and other lupus genetics studies.
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Files in this Data Supplement:
- Web Only Data - This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
BAPE is the coordinator of the Argentine collaborative group listed in the acknowledgements.
Funding This work was supported partly by grants from the European CVDIMMUNE project from the European Commission LSHM-CT-2006-037227, the Swedish Research Council for Medicine, the Swedish Association against Rheumatism, the Magnus Bergwalls Foundation, the Gustaf V:e 80th-year Jubilee, the Torsten and Ragnar Söderbergs Foundation and the Marcus Borsgtröms Foundation, the NIH-NCRR/COBRE grant P20 RR020143 to MEAR (PI JBH), the OCAST grant HR09-106 and the Instituto de Salud Carlos III partly financed through FEDER funds of the European Union to MEAR. This work was also partly supported by FISM, Regione Piemonte (CIPE and grant 2008) to SDA, the BMBF Kompetenznetz Rheuma C2.12, Germany to TW, grants SAF2006-00398, CTS-1180 and RETICS Program, RD08/0075 (RIER) from Instituto de Salud Carlos III (ISCIII) to JM. BAPE is the coordinator of the Argentine Collaborative group and his work was partly supported by the Federico Wihelm Agricola Foundation Research grant. National Institutes of Health RR020143 (JMG and JBH), RR015577 (JMG, JBH, JAJ), N01 AI050026-001 (JMG and JAJ), AR053483 (JMG and JAJ), AI063274 (PMG), AI031584 (JBH, JMG, JAJ), AR052125 (PMG), AR043247 (Kathy L Moser), Kirkland Scholar awards (JBH and JAJ), AR049084 (JBH), AR42460 (JBH), AR62277 (JBH), AI24717 (JBH), AR048940 (JBH, JAJ), AI083194 (JBH), R01 DE018209 (JBH), AI082714 (JBH), Alliance for Lupus Research (JBH), the US Department of Veterans Affairs (JBH), and the OHRS award for project number HR08-037 from the Oklahoma Center for the Advancement of Science and Technology (JMG).
Competing interests JW is an employee of MerckSerono Inc, and as such, the data or software he has developed belong to MerckSerono Inc.
Ethics approval All individuals provided informed consent as approved by the recruiting site institutional review boards at each of the affiliate institutions. All clinical investigation has been conducted according to the Declaration of Helsinki.
Patient consent Obtained.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.