Genetic susceptibility to lupus: the biological basis of genetic risk found in B cell signaling pathways

Review on B cell signaling pathways in lupus risk genes, possibly representing a unique therapeutic opportunity.

Abstract

Over 50 genetic variants have been statistically associated with the development of SLE (or lupus). Each genetic association is a key component of a pathway to lupus pathogenesis, the majority of which requires further mechanistic studies to understand the functional changes to cellular physiology. Whereas their use in clinical practice has yet to be established, these genes guide efforts to develop more specific therapeutic approaches. The BCR signaling pathways are rich in lupus susceptibility genes and may well provide novel opportunities for the understanding and clinical treatment of this complex disease.

Introduction

SLE (or lupus) is an archetypical systemic autoantibody-mediated autoimmune disease. Autoantibody production in lupus leads to direct autoantibody-target tissue damage and to indirect autoantibody-autoantigen (immune) complex formation and deposition, both resulting in multisystem tissue injury. Lupus is highly variable in severity and in the range of clinical manifestations, as evidenced by allowing classification as lupus with any combination of four of 11 criteria [1, 2].

Lupus is tenfold more frequent in women and three- to seven fold more frequent in non-European ancestries [3]. Despite its heterogeneity, SLE has a strong genetic component, with heritability estimates as high as 66%, a high sibling risk ratio (8:29), and a concordance rate in monozygotic twins (∼30%), 10 times higher than that of dizygotic twins (∼3%) [4]. The most powerful evidence for a genetic component for lupus susceptibility, however, is the ∼50 genetic variants that are now convincingly established to be associated with lupus (Table 1). In aggregate, these independent associations account for a minor fraction of the genetic risk of lupus [4], suggesting that other risk factors and better models of genetic effect remain to be discovered.

Table 1. Genes Associated with Lupus.

Gene	Location	Odds ratio	Best P value	Population	References	SNP with best P value (risk allele) reference	Risk allele frequency of best SNP	Causal variant (minor allele)
PTPN22	1p13.2	1.4	3.4 × 10−12	EU, HA	[5–7]	rs2476601 (A) [7]	0.10	rs2476601 (A) PTPN22 R620W
FCGR2A, FCGR3B	1q23	0.74	6.8 × 10−7a	EU, AA, AS	[6, 8–10]	rs1801274 (T) [6]	0.43	rs1801274 (T) FCGR2A H166R
NCF2	1q25	1.19	4.62 × 10−20	EU, AS	[7, 11–13]	rs17849502 (G) [13]	0.109	rs17849502 (G) NCF2 H389Q
CRP	1q21	0.49	9.2 × 10−14	EU, AA	[14]	rs3093061 (G) [14]	0.1925
TNFSF4	1q25	1.46	2.5 × 10−32	EU, AS, HA	[6, 7, 12, 15, 16]	rs2205960 (A) [16]	0.347
IL10	1q31–q32	1.19	4.0 × 10−8	EU, AA	[7, 12, 17, 18]	rs3024505 (A) [7]	0.16
Complement genes	1p36		Convincing family studiesb	EU	[19–23]			Multiple, usually leading to a deficiency in protein production
RASGRP3	2p24.1	0.7	1.3 × 10−15	AS	[16]	rs13385731 (G) [16]	0.07
IFIH1	2q24	1.11	1.6 × 10−8	EU	[24]	rs1990760 (T) [24]	0.61
STAT4	2q32.2	1.55	5.17 × 10−42	EU, AA, AS, HA	[6, 12, 15, 25–28]	rs7582694 (C) [16]	0.4227
PXK	3p14.3	1.25	7.1 × 10−9	EU	[6, 12]	rs6445975 (C) [6]	0.32
TREX1	3p21.31	44.65	8.5 × 10−11	EU	[29, 30]	rs3135945 (A) [29]	0.01	TREX1 R11H (very rare; thus, the overall association is probably a result of other causal variants)
BANK1	4q24	1.31	2.62 × 10−13	EU, AA, AS, HA	[12, 15, 31, 32]	rs10516487 (A) [32]	0.13	rs10516487 (A) BANK1 R61H
IL2/IL21	4q26	1.16	2.2 × 10−8	EU, AA, AS	[11, 33]	rs907715 (G) [33]	0.69
TNIP1	5q32	1.27	1.67 × 10−9	EU, AA, AS	[7, 12, 16, 34–36]	rs10036748 (G) [6]	0.45
Mir146a	6q5	1.29	2.74 × 20−8	AA, AS	[37, 38]	rs57095329 (G) [37]	0.21	rs57095329 affects protein binding, including the binding of Ets12 to the promoter
HLA and other genes	6p21.3	2.35	1.27 × 10−51	EU, AS, AA, HA	[6, 7, 12, 16, 34, 35, 39, 40]	rs1270942 (G) [6]	0.2
ATG5	6q21	1.25	5.2 × 10−12	EU, AS	[6, 12, 16]	rs548234 (G) [16]	0.30

Gene	Location	Odds ratio	SNP with best	Population	References	SNP with best P value (risk allele) reference	Minor allele frequency of best SNP	Causal variant (minor allele)
TNFAIP3	6q23	1.72	1.3 × 10−17	EU, AA, AS	[7, 12, 16, 41–43]	rs2230926 (C) [16]	0.07	TT > A polymorphic dinucleotide (deletion T followed by A transversion) in promoter; decreased NF-κB binding, reduced transcript, and protein expression
IKZF1	7p13	0.72	2.8 × 10−23	AS, EU	[12, 16, 36]	rs4917014 (C) [16]	0.25
JAZF1	7p15.2	1.19	1.5 × 10−9	EU	[7, 12]	rs849142 (T) [7]	0.49
IRF5	7q32	1.54	3.611 × 10−19	EU, AA, AS, HA	[6, 7, 12, 44–47]	rs12537284 (A) [6]	0.19	Three independent effects, ranging from differential splicing to increased mRNA expression and differences in polyubiquitination
XKR6	8p23.1	1.23	2.5 × 10−11	EU	[6, 48]	rs6985109 (G) [6]	0.51
BLK	8p23	0.69	2.1 × 10−24	EU, AA, AS, HA	[6, 16]	rs7812879 (A) [16]	0.117
LYN	8q13	0.77	5.4 × 10−9	EU, AA, AS	[6, 49]	rs7829816 (C) [6]	0.18
LRRC18, WDFY4	10q11.23	1.24	7.2 × 10−12	AS	[12, 16, 50]	rs1913517 (A) [16]	0.33
CD44	11p13	0.71	4.0 × 10−12	EU, AS, AA	[51, 52]	rs507230 (G) [51]	0.44
PHRF1/IRF7/KIAA1542	11p15.5	0.78	3 × 10−10	EU, AA	[6, 7, 12, 16]	rs4963128 (T) [16]	0.05	IRF7 Q412R was reported by Fu et al. [3]; resulted in a twofold increase in insulin response element transcription
ETS1	11q24.3	1.37	1.8 × 10−25	AS	[16, 34, 36, 50]	rs6590330 (A) [16]	0.41
SLC15A4	12q24.32	1.26	1.77 × 10−11	AS	[16, 36]	rs1385374 (A) [16]	0.25
ELF1	13q13	1.26	1.5 × 10−8	AS	[53]	rs7329174 (G) [53]	0.28
ITGAM	16p11.2	1.62	1.61 × 10−23	EU, AS, HA	[6, 7, 12, 15, 50, 54, 55]	rs9888739 (T) [6]	0.19	ITGAM R77H compromises leukocyte adhesion
PRKCB	16p11.2	0.81	1.4 × 10−9	AS	[56]	rs16972959 (A) [56]	0.23

Gene	Location	Odds ratio	Best P value	Population	References	SNP with best P value (risk allele) reference	Minor allele frequency of best SNP	Causal variant (minor allele)
IRF8	16q24.1	1.16	2.3 × 10−9	EU	[24]	rs280519 rs2280381 (A) [24]	0.62
TYK2	19p13.2	1.20	3.88 × 10−8	EU	[24, 57]	(A) [24]	0.47
CD40	20q12	0.63	2.0 × 10−8	EU	[58]	rs4810485 (T) [58]	0.24
UBE2L3	22q11.21	0.78	1.48 × 10−16	EU, AS	[6, 16]	rs463426 (G) [16]	0.41
TLR7	Xp22.3	1.67	6.5 × 10−10	AS	[59]	rs3853839 (G) [59]	0.81
IRAK1/MECP2	Xq37	1.39	6.65 × 10−11	EU, AS, HA	[7, 12, 15, 60, 61]	rs1734787 (C) [61]	0.17

^aP < 5 × 10⁻⁸.

^bGenetic variants at 36 loci covering more than 50 genes have an established role in the etiology of lupus through association studies, replication in populations of diverse ancestry, and family studies. Some of the associated loci, such as 1q21 and Xq37, contain multiple genes. Until follow-up studies demonstrate or exclude the possibility of independent effects, all of the genes in the associated regions will remain candidate lupus genes. For HLA and complement, there are many different independent effects, which for simplicity, are summarized in a single row. Variants in CFHR3/CFHR1, NMNAT2, ICA1, IKBKB, and SCUBE1⁴ have P values approaching genome-wide significance (5×10⁻⁸<P<10⁻⁷) and are the focus of current replication studies. Population ancestry: EU, European; AA, African; AS, Asian; HA, Hispanic American. FCGR, Low-affinity Ig-γ FcR; NCF2, neutrophil cytosolic factor 2; CRP, C-reactive protein; TNFSF, TNF superfamily⁵

To date, consensus surrounding the etiology and pathogenesis of lupus remains elusive. There is great potential in using genetic studies to highlight signaling pathways that likely contribute to lupus pathogenesis. Genetic association data inform hypothesis-driven questions for functional biologists to elucidate underlying disease mechanisms.

In this review, we present the many genetic regions now identified as associated with lupus. We then discuss how the putatively responsible genes are often coexpressed and cooperate in shared signaling pathways. Particular attention is focused on the convergence of lupus-associated variants in the BCR signaling pathway, as well as costimulatory signaling pathways, such as CD40, which synergizes with the BCR. The enrichment of genes in these pathways provides insight into SLE pathogenesis and offers potential targets for future therapeutic strategies.

GENETIC ASSOCIATIONS WITH SLE

Over the past two decades, lupus genetic studies have transitioned from identifying linkage effects using variable number tandem repeats in less common multiplex families to the statistical association of alleles defined by SNPs. The standard study design for association studies compares allele frequency of SNPs (or other variant markers) in cases and controls. These studies can be performed as candidate hypothesis-driven studies on one SNP or a handful of SNPs in one gene. Alternatively, in hypothesis-generating GWA studies, the assessment of millions of human SNPs facilitates the discovery of statistically associated genetic variants that serve as markers for the variants that are causal. Presently, deep sequencing studies are underway that allow the exploration of the entire variant genotypic complement of an individual. The incredible technical revolution of the past 7 years, analogous to switching from horseback to a modern sports car seemingly overnight, fundamentally changes the conceptual character of the genetic approach. Contemplating a complete identification of the major genetic associations and their interactions is daunting; however, given the current trajectory of innovation, these types of studies might well be possible within the foreseeable future.

For lupus, fewer than 5% of the over 150 peer-reviewed and published associations are confirmed in the GWA era of genetic studies [62]. As the genotyping technology has become more sophisticated, the weak performance of the previous candidate gene and linkage eras became obvious for many disease-related phenotypes [63, 64]; however, those genes that have been replicated, such as IRF5 and STAT4, have modest effects and should not be discounted. In the spirit of a reemphasis on the scientific method and the development of more rigorous statistical criteria before an association would be accepted as probably present, Pe'er et al. [65] developed a genome-wide criteria for significance to diminish the consequences of overtesting the data. They reasoned that the genome contains ∼1 million independent elements, given the linkage disequilibrium present. Therefore, the threshold for having a one in 20 chance of a false-positive result (α error of 0.05) should be a probability of ∼5 × 10⁻⁸ [65]. The problem is that this level of rigor requires large collections of cases and controls to achieve adequate power with moderate expectations for effect size. Even then, the case-control design only has adequate power with a limited set of mechanisms of inheritance [66]. In fact, statistical power is very poor to detect recessive and epistatic effects, despite their presumed role in lupus etiology. The sample sizes required for most alleles operating through these mechanisms of inheritance are far beyond our current capacity. Notwithstanding the challenges, the application of these rules has allowed for the identification of more than 50 statistically convincing and replicated associations for the development of lupus (Table 1).

Once an association is established, fine mapping studies are initiated with the aim of identifying genetic variants that are biologically responsible for the genetic association. Fine mapping studies are usually done by designing genotyping arrays that use the same chemistry as the GWA study arrays to genotype as many SNPs as possible around the detected lupus-associated variant. Many such array-based fine mapping studies limit the location of the causal variant to small physical regions of DNA. LLAS-1 and LLAS-2, encompassing 20,000–30,000 fine mapping markers and up to 18,000 subjects, have been effective for fine mapping-identified loci (described in refs. [7, 29, 52]). Subsequent studies currently underway with the 125,000 useful fine mapping markers on the ImmunoChip are expanding our ability to rapidly fine map associated regions, as has been accomplished recently for Celiac's disease [67]. Another more recent approach is to obtain deep DNA sequence of physical regions around the associated variants in a sample of subjects to provide a characterization of the variation differences between the risk and nonrisk haplotypes.

Fine mapping studies of the strongest associations have revealed that identifying the causal variant is a nontrivial process for many genes. In fact, amino acid-changing SNPs associated with lupus have only been identified and subsequently became candidate causal variants in eight out of the 50 associated loci (Table 1). The linkage disequilibrium within associated regions results in many variants demonstrating statistical associations at a single locus. The evolutionarily diverse and fourfold longer population history of African ancestry makes possible important insights into variant identification that would not otherwise be possible from physical mapping experiments in Europeans; however, even when the genetic effect is localized to a few hundred bases, the functionality of the variant is not always directly apparent. Whereas coding variants, such as nonsense and mis-sense SNPs, are annotated easily, variants that affect transcription factor binding sites, micro-RNA regulatory sites, RNA stability, or epigenetic regulation of a locus are more difficult to annotate given the current level of understanding.

Oftentimes, finding the functional effect of an association through experimentation with genotyped samples can inform the analytical determination of the causal variant. For example, rs9888739 results in an arginine to histidine amino acid change at position 77 in ITGAM, putatively compromising leukocyte adhesion, which at present, is accepted as the causal variant responsible for the genetic association in the genomic region around ITGAM [68]. In other cases, such as IRF5 and TNFAIP3, the association is a result of the biological impact of multiple causal variants that affect the mRNA expression and stability [69, 70].

Organizing the associations into testable pathways provides a framework from which to ask hypothesis-driven, biologically relevant questions. In lupus, several groups have suggested collections of associated genes that work together in a common pathway and activated by common stimuli. For example, of the ∼50 genes now known from published and preliminary data to be convincingly associated with SLE, it is striking that 30 are directly or indirectly involved in the NF-κB signal transduction pathway related to IRF5 and IRF7.

PATHWAYS IN LUPUS: HIGHLIGHTING GENES IN THE BCR SIGNALING PATHWAY

Reducing the complexity of disease-associated genetic variants by identifying cell types and pathways is a well-established approach to dissect lupus pathogenesis. Categorizing statistically associated genetic variants may reveal and highlight potentially important disease mechanisms. For example, Hu et al. [71] use a novel statistical method to identify cell-specific expression of lupus variants and find a convincing enrichment of gene expression in transitional B cell genes. Lupus susceptibility genes tend to cluster in the TLR/type 1 IFN pathway, immune-complex processing, immune signal transduction, and BCR signaling pathway [72].

Intuitively, one might assume that multiple variants within the same category would synergistically increase risk of disease. Geneticists use complex statistical approaches to measure epistasis between variants in the same pathway. Despite studies, including up to 15,000 subjects, the association results, to date, generally support additive relationships between lupus susceptibility genes instead of synergistic, multiplicative, or epistatic models [73]. In the context of current findings consistent with additive models of association, our current working hypothesis is that small changes to specific pathways permit the events that result in broken tolerance and overt autoimmunity.

To underscore significant lupus-associated genetic variants in B cell signaling pathways, we highlight the presence of proteins with established lupus associations in the BCR signaling pathway (Fig. 1) and interconnected costimulatory pathways (Fig. 2) with brown shading in the figures and bold font in the text.

Figure 1. Lupus risk genes regulate B cell signaling.

Engagement of cell surface receptors, inhibitory or excitatory, induces signaling events involving numerous lupus-associated genes. LYN can bind to excitatory receptors, such the BCR, and inhibitory receptors, such as CD22 and FcγRIIb. In both cases, it acts as a kinase to initiate downstream activity, with the end result of activation or inhibition depending on the context of the receptor. BLK is also a PTK but does not interact with inhibitory receptors, resulting in the propagation of excitatory signals only. LYP is a PTP and thus, dephosphorylates second messengers in downstream activation pathways, resulting in reduced activation. BANK1 is a scaffold protein that brings PTKs into proximity with targets such as IP₃R. BANK1 can also suppress cellular activation, for example, by inhibiting CD40-dependent Akt phosphorylation. Proteins shaded in brown are coded by lupus risk-associated genetic loci. NFAT, Nuclear factor of activated T cells.

Figure 2. Lupus risk genes regulate NF-κB activation.

Ubiquitination (Ub) is a key regulatory step for numerous cell-signaling events. A20 and TNIP1 form a complex that cleaves activating K63-linked ubiquitin chains (depicted in the figure by ovals) and add K48-linked ubiquitin chains (depicted in the figure by squares). Whereas the exact mechanism is not known, A20 and TNIP1 act in a MyD88-dependent manner to regulate inflammation, in part, by inhibiting TRAF6, which is an E3 ubiquitin ligase that is known to act downstream of CD40 and MyD88-linked extracellular (TLR1/2, TLR2/6, TLR4, and TLR5) and intracellular (TLR3, TLR7/8, and TLR9) TLRs. TRAF6 activation leads to the phosphorylation of IRF5 and IRF7, which among other functions, activate NF-κB signaling. Of note, each of these genes participates in many interactions beyond those depicted in this figure. Proteins shaded in brown are coded by lupus risk-associated genetic loci.

B CELLS IN LUPUS

Patients with lupus progressively accumulate a number of autoantibodies directed against self-antigens, in particular, anti-nuclear proteins [74]. Whereas it is not clearly understood what leads to this break in self-tolerance, autoantibody production precedes overt disease manifestations and clinical diagnosis [75]. Autoantibodies are thought to mediate many of the clinical sequelae of lupus, indirectly and directly. Renal disease in lupus is primarily caused by immune complex deposition in the glomerulus, leading to fibrosis and loss of glomerular filtration capacity [76]. Autoimmune hemolytic anemia is thought to be a result of clonal autoantibodies to cell surface antigens. Some lupus patients develop antiphospholipid syndrome, in which antibodies against the phospholipid-β-2 glycoprotein I complex are associated with thrombotic events, placental infarcts leading to miscarriage, and liver dysfunction [77]. Given the role of autoantibodies in SLE pathogenesis, B cells have long been thought to play a crucial role in mediating disease.

SIGNALING THROUGH THE BCR

The BCR-mediated signaling cascade is a highly complex series of inhibitory and activating interactions that occur in spatial and temporal patterns (reviewed in refs. [78, 79]). BCR-mediated signaling begins when the antigen binds and cross-links the BCR. Src-family PTKs, such as Lyn, phosphorylate ITAMs on Igα and -β, which then serve as docking sites for cytosolic SH2 domain-containing PTK Syk, resulting in phosphorylation of Syk (Fig. 1). Upon activation, Syk is central to multiple downstream signaling events (PLCγ2-SLP65/BLNK complex), ultimately resulting in cytoskeletal rearrangement, gene-expression regulation, and cell-fate decisions. As PTKs are central to positive BCR signaling, PTPs, such as Lyp, encoded by PTPN22, play an opposing role and serve as a balance to maintain homeostasis. PTPs are recruited to the BCR signaling complex through ITIMs on inhibitory coreceptors such as CD5 and CD22 (Fig. 1). ITIMs are also targeted by Src-family PTKs, however, ITIM phosphorylation results in recruitment of SH2 domain-containing PTPs, which counteract the kinase activity of PTKs. Another mechanism by which BCR signaling is negatively regulated is by activation of the low-affinity FcγRs, specifically FcγRIIb. This receptor is unique among FcγRs, in that it has an ITIM, and phosphorylation of this motif results in attenuation of BCR signaling (Fig. 1).

Stimulation of B cells on coreceptors, such as CD40, or through TLR activation synergizes with BCR signaling and leads to B cell proliferation (Figs. 1 and 2), cytokine production, antibody secretion, and isotype switching, each of which play a potential role in SLE pathogenesis. Different combinations of costimulatory signals allow B cells to produce cytokines and antibody isotypes for a specific inflammatory response. Genetic variants in these costimulatory pathways subtly affect how B cells generate an immune response. These changes could prejudice B cell responsiveness toward inappropriate breaks in tolerance, leading, in some cases, to lupus.

One distinct feature of the BCR signaling pathway is that in addition to second messenger-mediated signal transduction, the BCR-antigen complexes are physically internalized and trafficked through the endocytic pathway to the MIIC. Here, antigen is digested and loaded onto MHCII molecules prior to surface expression [80].

BCR SIGNALING IN B CELL SELECTION

The mechanisms that allow autoreactive B cells to escape negative selection and become antibody-secreting plasma cells in the periphery are poorly understood. It is clear that tolerance is broken in these autoreactive B cells and that immune dysregulation is central to the etiology of lupus. Currently, we do not have clear mechanistic insight into how genetic and environmental factors sometimes result in the loss of immunological tolerance of self. B cell selection occurs in two general steps: early selection prior to development into mature B cells, and later, a second selection step occurs, as mature B cells circulate in the periphery. Early B cell selection consists of two checkpoints. First, immature B cells in the BM can undergo three distinct fate decisions based on binding affinity to self-antigen: high-affinity binding results in deletion, moderate-affinity binding leads to receptor editing via heavy- and light-chain gene rearrangement, whereas low-affinity binding leads to anergy. The second checkpoint is less well understood but is believed to occur as newly emerged transitional B cells circulate in the periphery prior to maturation. This is supported by the decreased numbers of autoreactive clones in the mature B cell pool when compared with the number of autoreactive clones in the pool of new emigrant B cells, suggesting that some deletion of autoreactive B cells has occurred in transit. Many suspect that this may result from an encounter of a new emigrant B cell with autoantigen, leading to subsequent negative selection [81].

SPECIFIC GENES WITH A POTENTIAL ROLE IN B CELL SIGNALING AND SURVIVAL

LYN

LYN encodes a Src-family PTK that acts early in the B cell signaling pathway (Fig. 1). Polymorphisms within LYN are associated with lupus in European-American women, and the risk allele is the major allele [49, 82]. The specific genetic variant responsible for the lupus association with LYN is unknown, although Lyn expression is decreased in B cells isolated from lupus patients. The decrease in expression also correlates with exclusion of Lyn from the lipid raft fraction in B cell digests [83, 84]. Lyn is one of the central mediators of signaling in B cells as it phosphorylates ITAMs and ITIMs (Fig. 1).

Early results with LYN^−/− mice revealed a lupus-like phenotype that suggested a role for Lyn in lupus pathogenesis. LYN^−/− mice have decreased B cells in the periphery that display decreased proliferation and diminished phosphorylation of CD22 and FcγRIIb in response to anti-IgM stimulation [85]. Autoantibody levels are increased in LYN^−/− animals, as is total IgM in the serum, which corresponds with an increase in the IgM+ B cell fraction [86]. Whereas Lyn can phosphorylate ITAMs and ITIMs of cell surface receptor complexes, its role in activating positive signaling pathways downstream of the BCR is redundant with Blk and Fyn [87]. Importantly, Lyn has a unique role in mediating negative signaling by phosphorylation of ITIMs in CD22 and FcγRIIb [88, 89]. In an effort to understand the mechanism of the B cell depletion in LYN^−/− mice, Gross et al. generated mixed BM chimeras to define the competitive advantage or disadvantage of Lyn deficiency in B cells. The loss of LYN^−/− B cells is a B cell-intrinsic defect, as LYN^−/− follicular B cells were reduced compared with WT in mixed BM chimeras. Mechanistically, this B cell contraction correlates with increased expression of the proapoptotic molecule Bim [90].

As Lyn can act as a positive and negative regulator, LYN^−/− mice are used to investigate the balance between positive and negative signaling in peripheral B cells. Newly arrived splenic transitional (T1) B cells are progressively less sensitive to BCR-induced signaling, as they mature to follicular B cells. This is demonstrated by increased Ca²⁺ flux and second-messenger phosphorylation in T1 B cells compared with follicular B cells following IgM stimulation. The increased signaling is most likely a result of the up-regulation of the CD22-SHP-1-Lyn regulatory axis, which is an important source of negative signaling (CD22 is increased in murine follicular B cells compared with T1 B cells). T1 B cells from LYN^−/− mice show no change in BCR-induced signaling compared with WT, whereas LYN^−/− follicular B cells are much more active than WT cells [91]. These results further suggest a role of Lyn in maintaining negative signaling in B cells, particularly follicular B cells. Whereas these results highlight a role for Lyn in the suppression of follicular B cell activation, loss of LYN does not affect tolerance during B cell maturation as LYN^−/−Ig^HEL mice have no defect in negative selection [92]. These data identify Lyn as a key mediator of the FcγRIIb and CD22-negative signaling pathways, which appear to be more important in mature B cells compared with early transitional cells, where stimulatory signals predominate, promoting negative selection. The increase in positive signaling observed in LYN^−/− B cells may be enough to increase negative selection in early transitional B cells, leading to the decreased pool of transitional B cells in LYN^−/− mice. The explanation for the emergence of autoreactive B cells and autoantibodies is not as clear, given the decrease in B cells. Increased BCR signaling as a result of the decreased negative signaling in mature B cells because of Lyn deficiency is one explanation, but it may also be a product of the effect of Lyn deficiency in myeloid cells, which increases the inflammatory tendency of the cellular milieu [93]. LYN^−/−MyD88^−/− mice do not produce autoantibodies [94] as a result of the suppression of plasma cell differentiation. The development of autoimmunity in LYN^−/− animals most likely arises as a result of the combination of the increased sensitivity of mature B cells to positive signaling pathways and the concurrent increase in inflammation mediated by LYN^−/− myeloid cells and TLR-dependent stimulation.

• Lyn is a Src-family PTK that can phosphorylate ITAMs and ITIMs (Fig. 1).

• Lyn-mediated phosphorylation of ITIMs is crucial to dampen activation of follicular B cells in mice.

• Decreased levels of Lyn are associated with lupus.

• The major allele in the population is the risk variant for LYN.

BLK

Blk is a Src-family PTK expressed predominately in B cells [95]. The genetic association between BLK and lupus in individuals of European descent was first identified at a SNP ∼7 kb upstream of BLK and confirmed in subsequent studies [82]. BLK expression is genotypic, with people who carry two copies of the risk allele expressing more Blk than people carrying one copy or two copies of the protective allele [54]. The association of BLK with lupus has been replicated in Japanese [96] and Chinese [54] populations.

BLK interacts with BANK1 genetically (with epistasis between the risk variants, despite no increased lupus risk) and physically (by colocalizing; Fig. 1) [97]. Coimmunoprecipitation studies demonstrated a basal interaction between Blk and BANK1 that was enhanced by IgM stimulation. Additionally, the expression of BANK1 appears to alter Blk localization, and Blk localizes to cytoplasmic compartments instead of the plasma membrane in the presence of BANK1 [97].

BLK^−/− mice have no observable immune phenotype [98]; however, as Blk is functionally redundant with Lyn and Fyn, this is not surprising [87]. Whereas Blk does not apparently have a unique role in mice, decreases in Blk expression may still have a significant impact on human disease, as is seen with another PTK family gene product, Btk. Mutations in Btk in humans result in X-linked agammaglobulinemia, which can result in a complete lack of B cells. The phenotype in BTK^−/− mice is not as severe, suggesting significant differences between PTK signaling pathways in humans and mice [99].

• Blk is a PTK that can bind ITAMs and BANK1 (Fig. 1).

• A polymorphism in BLK associated with lupus is also associated with decreased expression of Blk.

BANK1

BANK1 is a B cell scaffold protein that binds the Src family PTK Lyn and the IP₃R in distinct regions. Upon BCR stimulation, BANK1 is phosphorylated, favoring Lyn binding (Fig. 1). Dual binding of Lyn and IP₃R permits Lyn-mediated phosphorylation of IP₃R, leading to Ca²⁺ mobilization from intracellular ER stores [100]. BANK^−/− mice have been developed and found to have slightly increased numbers of mature B cells, increased serum IgG2a, and increased formation of germinal centers. Following CD40 stimulation, B cells from BANK1^−/− mice demonstrate enhanced proliferation and survival. BANK1^−/−, CD40^−/− mice rescue the phenotype. The increased survival and proliferation of B cells depend on CD40 activation of TRAF6, which is essential for Akt ubiquitination, membrane recruitment, and phosphorylation. Further mechanistic characterization of the BANK1^−/− CD40^−/− mice suggests that BANK1 acts as a negative regulator of CD40-induced survival signals via the suppression of Akt [101]. The connection between the role of BANK1 in Lyn-mediated signaling and the role of BANK1 in CD40-mediated signaling within B cells remains to be explored, as does the possibility that BANK1 has actions that influence the BCR and CD40 pathways distinctly at different times over the life of a B cell. Future studies will also need to address the role of BANK1 in B cell signaling pathways apart from CD40 and Lyn. Two major isoforms are identified for BANK1: one full-length isoform and a truncated isoform lacking all of exon 2, named Δ2 [31].

Kozyrev et al. [31] identified three potential, functional variants associated with lupus in Europeans, a finding that we replicated subsequently [102]. rs17266594 is in a putative branch point site in intron 1 but was later shown not to be essential for splicing [103]. rs10516487 codes for an R61H variant in exon 2, and rs3733197 codes a A383T variant in exon 8 [31]. The nonsynonymous SNP rs10516487 is located in the second exon and is in high-linkage disequilibrium with rs17266594 (r²=0.90). The protective minor allele A results in a coding change from Arg to His at position 61, resulting in the creation of a recognition site for the splicing enhancing factor SRp40. The presence of this binding site appears to inhibit splicing of exon 2. In addition, the A allele was associated with lower levels of BANK1 expression at the message and protein level. The exact relationship between the R61H variant and the Δ2 isoform is not firmly defined, but the Δ2 isoform lacks the IP₃R binding domain and when present, alone or with full-length BANK1, is diffusely distributed in the cytoplasm instead of clustered into punctae, as is the full-length isoform when present alone. Thus, heterologous dimer formation may be an important mechanism of BANK1 regulation, determining cellular location and access to interacting proteins [103].

Whereas the identification of these potential causal variants is promising, the exact relationship between SLE-associated variants in BANK1 and disease pathogenesis remains unclear. Further studies focused on understanding the effect of these functional variants on the CD40 signaling pathway will help clarify the role of BANK1 in SLE pathogenesis.

• BANK1 is a scaffold protein that can bind PTKs and IP₃R (Fig. 1).

• Multiple functional polymorphisms in BANK1 are associated with lupus, although the casual variants are not yet identified.

• B cells from BANK1^−/− mice proliferate in a CD40-dependent manner.

PTPN22

PTPN22 encodes Lyp, a PTP found in lymphocytes [104]. A variant of PTPN22 (1858C/T) is associated with multiple autoimmune diseases, including RA [105], SLE [5], T1D [106], Graves disease [107], and myasthenia gravis [108]. The association between PTPN22 and SLE is well established in multiple ancestries [109] and seems to be explained entirely by the 1858C/T polymorphism. This genetic variant results in the amino acid substitution of R620W, producing Lyp 620W.

Lyp interacts with Csk, a PTK family member, to regulate phosphorylation of Src family kinases downstream of cell-surface receptors, including the TCR, BCR, and CD40 (Fig. 1). Early studies demonstrated that the R620W substitution disrupts Lyp binding to Csk [106]. The Lyp 620W variant has been studied in lymphocytes isolated from patients with lupus, T1D, and myasthenia gravis. The functional consequences of the Lyp 620W variant in autoimmune disease were first reported in T cells isolated from T1D patients [106, 110]. In these cells, TCR engagement resulted in decreased phosphorylation of downstream targets, decreased Ca⁺² flux, and decreased IL-2 production [110]. Given these findings, the Lyp R620W mutation was proposed to be a gain-of-function mutation, resulting in decreased signaling through the TCR in cells carrying the disease-associated phenotype [110]. Subsequent work in isolated T cells appears to have confirmed these findings [111, 112]. However, other groups have reported the opposite phenotype in PBMCs and T cell lines simultaneously transfected with Lyp and Csk [113, 114]. Whereas T cells isolated from individuals carrying the Lyp R620W mutation may represent a more physiological assessment than overexpression of Lyp in vitro, these differences remain to be reconciled.

Current efforts to resolve these differences have focused on the mouse homologue of Lyp: PEP. Zikherman et al. crossed PEP^−/− mice with CD45^w/w mice, which contain a point mutation in the CD45 juxtamembrane wedge domain (E613R), in which hyperactive CD45 signaling leads to B cell-driven autoimmunity [114]. The PEP^−/−CD45^w/w animals develop a lupus-like disease, similar to CD45^w/w mice, but with increased anti-dsDNA IgG. B cells from PEP^−/−CD45^w/w mice are similar to those from CD45^w/w mice. Whereas attempts to understand the effect of PEP deficiency on the development of autoimmunity have not been impressive, to date, other mouse models have been developed to specifically test the hypothesis that the Lyp 620W mutation will lead to increased autoimmunity. To test this hypothesis, mice expressing a PEP variant similar to Lyp 620W, PEP 619W, were developed. Expression of the Lyp 620W/ PEP 619W variant results in lower Lyp/PEP protein expression as a result of increased calpain- and proteasome-mediated cleavage and degradation, causing an increased activation of lymphocytes in these mice [115].

The Bruckner lab has produced numerous studies over the last few years examining the phenotype of B cells isolated from human subjects carrying the 1858C/T PTPN22 polymorphism. Individuals carrying at least one copy of the risk allele have reduced memory B cells in the periphery [112], along with increased transitional and anergic, naïve B cells [116]. Strikingly, autoreactive B cell clones are more frequent within these pools of transitional and mature, naïve B cells. These B cells are also more sensitive to CD40 ligand and IgM cross-linking by measure of surface activation markers [117]. In contrast to what occurs in the transitional and anergic, naïve B cell pools, the memory B cell population of individuals carrying the lupus-risk allele demonstrates a decrease in Ca²⁺ mobilization and proliferation following treatment with anti-IgM in vitro. In addition, stimulation of these memory B cells results in a reduction in phosphorylation of downstream targets, including Syk and PLCγ. The genotypic attenuation in these signaling pathways can be restored following treatment with the Lyp inhibitor I-C11 [118]. These studies suggest that PTPN22 plays a crucial part in early B cell tolerance checkpoints, with Lyp 620W, permitting progression of autoreactive clones to the periphery. These results also highlight the B cell-intrinsic effects of Lyp; however, PTPN22 mutations in T cells may still be important in the modulation of B cell activation. Decreased signaling through the TCR could also affect Treg function, thereby allowing autoreactive B cells that escaped to the periphery as the result of PTPN22 mutations during selection, to be activated and drive SLE pathogenesis.

If Lyp 620W does represent a gain-of-function mutation, then one explanation for the phenotype observed in individuals with the 1858C/T polymorphism would be that a hyperfunctional Lyp restricts activation of key second messengers, as evidenced by decreased phosphorylation in cells with Lyp 620W, raising the threshold for cellular activation. A higher activation threshold in transitional and naïve B cells could allow persistence of B cell clones that would otherwise be deleted through high-affinity binding. This is reflected in the increase of autoreactive cells in the pool of transitional B cells and importantly, may be a key event in the breakdown of tolerance and progression toward autoimmunity, as the increase in autoreactive clones mirrors what occurs in T1D [117].

In an additional layer of complexity, gene expression analysis identifies IRF5 as one of the genes up-regulated in B cells from 1858C/T individuals [117]. IRF5 is well established as being an important SLE risk gene. Increased IRF5 expression is associated with type I IFN production [119], suggesting that the PTPN22 risk allele may also promote cellular activation in the periphery. This represents a further mechanism by which alterations in PTPN22 polymorphisms promote autoimmunity.

• PTPN22 encodes Lyp, a PTP important for negative regulation of B cell signaling (Fig. 1).

• A PTPN22 risk allele, the Lyp 620W-putative, casual variant, is associated with numerous humoral autoimmune diseases.

• Lyp 620W is associated with increased activation of transitional and follicular B cells and decreased activation of memory B cells.

TNFAIP3

TNFAIP3, encoding the ubiquitin-modifying protein A20, was first identified as a risk gene in RA [120, 121] and has since been associated with multiple autoimmune diseases, including lupus [43, 54], Celiac disease [122], Crohn's disease [123], T1D [124], and psoriasis [125]. Recently, a TT > A deletion polymorphism was proposed to be the causal variant responsible for disease association in lupus. This deletion is downstream of the TNFAIP3 gene in an apparent regulatory region. Oligonucleotides with the deletion display decreased binding of the NF-κB subunits p50, p65, and cREL and reduced overall TNFAIP3 mRNA expression according to the number of risk chromosomes (0, 1, or 2) present in the sample [41]. These results suggest that decreased A20 levels are a risk factor for lupus.

A20 is a ubiquitin-editing enzyme induced by CD40 activation and signaling by proinflammatory cytokines, such as TNF-α and IL-1β. A20 modulates cell signaling by removing K63-linked ubiquitin chains from signaling intermediates and by adding K48-linked ubiquitin chains [126]. In general, the addition of K63-linked ubiquitin chains is activating, whereas K43-linked polyubiquitination targets proteins for proteosomal degradation. The main byproduct of ubiquitin-editing events is to limit NF-κB activation, aiding in the maintenance of immune homeostasis (Fig. 2) [127]. Recent evidence suggests that inhibition of IKKβ phosphorylation by A20 may be independent of A20 enzymatic activity, although still dependent on ubiquitin binding [128]. A20^−/− mice develop massive systemic inflammation, resulting in cachexia and multiorgan failure, leading to early death, mediated by failure to restrict NF-κB signaling [129]. Crossing A20^−/− mice with MyD88^−/− mice rescues the phenotype by restricting ubiquitination of TRAF6 [130], suggesting that A20 plays a prominent role in dampening homeostatic, MyD88-mediated inflammatory signaling (for a recent review, see ref. [131]).

Mice with TNFAIP3 conditionally deleted in B cells (CD19+) were developed recently by two separate groups [132, 133]. Both report significant changes in B cell populations, including increased plasma cells in older mice, immune complex deposition in the kidney, and increased IL-6 production. Tavares et al. [133] reports significant autoantibody production, but it is not clear whether these are IgM or IgG autoantibodies. As some IgM autoantibodies may be protective [134], the implications of this finding are not immediately obvious. The TNFAIP3^−/− mice studied by Chu et al. [132] had high levels of IgM in the sera and class-switched anticardiolipin IgG in older mice.

Furthermore, both groups also report changes relevant to splenic B cells in the B cell conditional TNFAIP3^−/− mice. Chu et al. [132] finds that marginal zone B (CD1d+) cells are located in the follicle of the spleen, with little to no B cells in the marginal zone. In addition, they demonstrate that splenic B cells fail to differentiate into Blimp1⁺ plasmablasts following LPS stimulation in vitro. Tavares et al. [133] also identify alterations in splenic B cells, noting an increase in germinal center B cells.

To test the hypothesis that A20 plays a role in Fas-mediated PCD in germinal center-negative selection [135], PCD was induced by agonist anti-CD95 stimulation in vitro. Indeed, A20^−/− cells are resistant to Fas-mediated PCD. The resistance to Fas-induced apoptosis correlates with an increase in the antiapoptotic molecule Bcl-x, a molecule that is normally suppressed by the NF-κB pathway [132]. Notably, both groups observe a dose effect in response to A20 deficiency in B cells, consistent with a pathogenic role for alleles, producing less-active A20, leading to B cell hyper-reactivity in human autoimmunity.

Taken together, A20 is potentially critical to the homeostatic dampening of NF-κB signals, and decreased protein expression is sufficient to tip B cells toward an autoimmune phenotype and loss of tolerance. Whereas the exact mechanism remains to be identified, the results from conditional knockout mice suggest that decreased expression of A20 may result in the production of B cells resistant to negative selection in the spleen. In addition, increased IL-6, produced by these animals, may also promote the survival of autoreactive plasma cells and T cell-mediated pathogenesis. Obviously, much work remains to connect these murine phenotypes with the genetic associations reported in human SLE. However, given the findings that SLE-associated genotypes correlate with lower A20 expression and that decreased A20 in murine B cells results in resistance to negative selection, in vivo studies aimed at tying together these observations will be an exciting next step in using B cell signaling pathways to understand statistical genetic associations.

• TNFAIP3 encodes A20, a ubiquitin-editing enzyme that can add and remove ubiquitin chains from target molecules (Fig. 2).

• A20 can regulate inflammation via attenuation of MyD88-dependent NF-κB signaling in mice.

• A lupus-associated proposed causal variant in TNFAIP3 is statistically associated with lower expression of A20.

TNIP1

TNIP1 encodes a ubiquitin-binding adapter protein ABIN-1, which binds and cooperates with TNFAIP3 in suppressing NF-κB (Fig. 2). Also known as Nef-associated factor-1, polymorphisms in TNIP1 are associated with lupus in European-derived and Asian populations [7, 16]. TNIP1 can be differentially spliced into at least 11 different isoforms [136], each of which suppress NF-κB activation with different levels of inhibition. Whether the lupus-associated risk variant of TNIP1 alters splicing or NF-κB activation is not yet known.

Deubiquitination and degradation of ubiquitinated signaling proteins are key mechanisms by which ABIN-1 cooperates with A20 to limit T lymphocyte receptor (TCR) signaling as part of a negative-feedback loop. Coexpression of ABIN-1 together with A20 in human embryonic kidney 293 cells contributes to the negative regulatory function of A20 in TNF-α-mediated NF-κB activation [137]. Mechanistically, ABIN-1 interacts with A20, independent of the ubiquitin-binding domain, to mediate the NF-κB inhibitory function of A20 [138].

ABIN-1 also binds the MAPK ERK2 and causes retention of ERK2 in the cytoplasm, leading to reduced signaling [139]. Small interfering RNA-mediated knockdown of ABIN-1 results in increased IFN-β production following viral infection or polyinosinic:polycytidylic acid administration in vivo, which is dependent on the ubiquitin-binding domain of ABIN-1, suggesting that ABIN-1 suppresses viral-induced IFN-β production [140]. Interaction between ABIN-1 and the TLR pathways is further supported by the finding that ABIN-1 recognizes K63-linked polyubiquitin chains on NEMO, which is the IKKγ subunit of the NF-κB signaling complex. NEMO activation is necessary for the nuclear translocation of NF-κB. Whereas K48-linked polyubiquitination of NEMO targets it for subsequent proteasomal degradation, K63-linked polyubiquitination of NEMO regulates its activity but not the half-life [141].

TNIP1^−/− mice develop a progressive, lupus-like inflammatory disease. The development of the lupus-like disease is dependent on leukocytes, as BM transplantation of WT mice with TNIP1^−/− BM is sufficient to initiate autoimmunity. ABIN-1 is recruited to the MyD88 signaling complex and controls TLR-mediated activation of the major isoform of C/EBPβ, known as liver-enriched activator protein, which has been shown to mediate TLR-dependent inflammatory cytokine production [142]. In fact, mice carrying a mutant TNIP1, which cannot bind ubiquitin, also display an autoimmune phenotype. Ubiquitin binding is rescued by crossing the mice onto a MyD88^−/− background [143]. Stimulation with TLR ligands, BCR cross-linking, or a CD40 agonist caused enhanced phosphorylation and activation of the NF-κB pathway in ABIN-1 mutant mice compared with WT mice [143]. These murine studies lead us to conclude that ABIN-1 cooperates with A20 in BCR signaling through interaction with the signaling pathway components downstream of CD40 and the BCR.

We recently reported a potential genetic association between SLE and another protein involved in ubiquitination, TRAF6 [144]. TRAF6 is an E3 ubiquitin ligase that has been identified as being important in numerous immune functions by activating NF-κB, following stimulation of BAFFR, TNFRs, TLRs, and CD40 [145, 146]. Whereas the results of these first association studies will need to be replicated in many more people to confirm this association, they are consistent with the reported association of TRAF6 with RA, also a humoral autoimmune disease. These results emphasize the importance of ubiquitin-mediated modulation of cellular-activation pathways. The associations between these genes and lupus suggest a fruitful area of research in working to elucidate the mechanisms by which disruptions in ubiquitination can facilitate the onset of autoimmunity (see Fig. 2).

• TNIP1 encodes a ubiquitin-binding adapter protein—ABIN-1, which binds ubiquitin and A20 (Fig. 2).

• Several genes involved in the ubiquitin regulatory pathway (e.g., TNIP1 and TNFAIP3) are associated with lupus, highlighting the importance of this pathway in the development of autoimmunity.

• ABIN-1 has anti-inflammatory activities that are dependent and independent of ABIN-1 ubiquitin-binding in mice.

• ABIN-1 attenuates MyD88/NF-κB inflammation in mice, in part, by targeting NEMO for proteasomal degradation; furthermore, ABIN-1 acts independently of MyD88 downstream of BCR and CD40 signaling.

CONCLUDING REMARKS

B cells are gaining newfound respect, now being appreciated as more than Ig factories. Indeed, the genetic evidence supplements other data supporting their critical role. Early case reports for small, clinical studies using B cell depletion therapies showed promise that these strategies would prove to be highly effective in treating lupus [147–149]. Unfortunately, larger, random control studies did not replicate these findings [150, 151]. Similarly, the first drug approved for lupus in 50 years, Belimumab, targets BAFF, a cytokine essential for B cell survival and proliferation, and has shown modest benefit in clinical trials [152]. These results underscore the fact that there are other factors that contribute to inflammation and SLE pathogenesis. These targeted approaches will most likely be more effective in a subset of SLE patients. B cell-related genetic risk variants may, one day, be used in more sophisticated and successful clinical approaches, including identifying which patients will be most likely to respond to specific therapies.

We are early in the effort to understand how sequence changes to specific genes can affect lupus risk. As each of the lupus associations is evaluated progressively, many complex, interconnected, and sometimes contradictory stories are likely to emerge. The controversies and apparent contradictions in studies of PTPN22 and LYN may foreshadow potential contradictions as we learn more about less-studied associations, such as BANK1. In fact, one decade ago, PTPN22 was thought to have a mechanism that operated through T cells. Now, the evidence is just as strong that lupus risk related to PTPN22 is operating through B cells. The history surrounding studies that ascribe molecular mechanisms to explain genetic effects gives us a glimpse at the challenges of forming a reasonable, conceptual framework to understand how risk of disease is generated.

EBV may play a significant role in the etiology of lupus. Briefly, the very earliest autoantibodies to appear in preclinical lupus patients are cross-reactive with EBV proteins, including EBV nuclear antigen-1, an EBV-encoded molecular mimic of SLE-associated autoantigens, such as Sm and Ro [75, 153]. In addition, EBV seroconversion [154] and viral load are increased in lupus cases compared with controls [155]. Several of the BCR-associated lupus genes discussed in this review play a role in EBV replication and immune evasion. For example, LYN has been shown to be the preferred signaling protein for the major EBV latency protein LMP2A, which acts as a BCR functional mimic, permitting B cell survival and preventing apoptosis of EBV-infected B cells, at least in part, by co-opting the B cell signaling apparatus, including lipid rafts [156]. A20 has also been shown to be important in EBV infection of B cells, as A20-mediated ubiquitination of IRF7 is stimulated by the EBV protein LMP1 [157], which functions as a CD40 functional mimic to stimulate B cells. A20 competes with the LMP1 complex in activation of the NF-κB pathway, altering the ability of LMP1 to drive NF-κB activation.

The biological impact of lupus-associated genetic variants in genes known to be important for EBV infection, such as LYN and A20, will be of special interest moving forward, as an individual's immune response to EBV infection (and the immune dysregulation caused by EBV-encoded mimics of key B cell activators) may provide important insight into EBV-associated lupus pathogenesis.

This review focused on the signaling in B cells and the corresponding lupus risk genes (Figs. 1 and 2). Other aspects of B cell signaling, including BCR internalization and delivery of antigen to the MIIC, are also known to have major roles in regulating B cell responses [158]. Genes in these pathways remain likely candidates for lupus. Also, other lupus genes, such as STAT4 and IRF5, are important transcription factors that mediate the response of many different cells, including B cells, to inflammatory stimuli. Yet, other genetic associations are in genes, for which little is known. Further investigation might well reveal newly recognized proteins with critical B cell functions. As more sophisticated techniques become available, we may find that risk polymorphisms in separate pathways act in concert to facilitate cellular changes conducive to tolerance breakdown and autoimmunity.

Indeed, all of the known genes together only account for a minority of the genetic heritability, as has been discussed for complex genetic diseases in general [159]. Together, small changes in proteins caused by common genetic variants alter B cell behavior toward a breakdown in tolerance and autoimmunity. Even after each of the individual associations is understood mechanistically, the practice of identifying the etiological mediators of lupus in patients presenting in the clinic will require a sophisticated understanding of the interactions of the proteins in molecular pathways, such as the BCR signaling cascade and the regulation of ubiquitin modification. By integrating genetic and biological data through multidisciplinary studies, we will prime ourselves for success in the coming age of personalized genomic medicine.

AUTHORSHIP

S.E.V. and J.B.H. developed the theme. S.E.V. and L.C.K. wrote and prepared the manuscript. J.B.H. and M.E.M. provided critical feedback and revised the manuscript.