The mechanisms regulating the subcellular localization of AID

Abstract

Activation induced deaminase (AID) is a unique enzyme that directly introduces mutations in the immunoglobulin genes to generate antibody diversity during the humoral immune response. Since this mutator enzyme poses a measurable risk of off-target mutation, which can be deleterious or transforming for a cell, several regulatory mechanisms exist to control its activity. At least three of these mechanisms affect AID subcellular localization. It was recently found that AID is actively imported into the nucleus, most likely through importin-α/β recognizing a structural nuclear localization signal. However, AID is largely excluded from the nucleus in steady state thanks to two mechanisms. In addition to nuclear export through the exportin CRM1, a mechanism retaining AID in the cytoplasm exists. Cytoplasmic retention hinders the passive diffusion of AID into the nucleus playing an important role in the nuclear exclusion of AID. Subcellular localization of AID also determines its stability. The regulation of the nuclear fraction of AID by these many mechanisms has functional implications for antibody diversification.

To generate the almost infinite number of specific receptors required to recognize any antigen, B-lymphocytes use the mechanism of V(D)J-recombination to rearrange their immunoglobulin (Ig) genes.¹ This combinatorial assembly of the Ig variable region is catalyzed by the RAG endonuclease in each B lymphocyte precursor and generates the primary repertoire of antibodies, of random specificity. Rearranged V(D)J regions are further diversified by somatic hypermutation (SHM), a second mechanism of genetic modification that introduces point mutations during B-cell clonal expansion. In some species with limited V(D)J diversity, SHM or its related mechanism Ig gene conversion generate much of the primary repertoire.^2,3 However, in most species, the predominant role of SHM is in generating the secondary repertoire of antibodies by underpining the maturation of the affinity of the primary response.⁴ SHM is initiated by Activation Induced Deaminase (AID),⁵ which deaminates deoxycytidine to deoxyuridine in Ig genes. The deoxyuridine is recognized as a foreign base in DNA and mutagenically processed at the Ig loci by specific DNA repair enzymes, which leads to the full spectrum of SHM (reviewed in refs. ⁶ and ⁷). At the IgH locus AID also initiates isotype-switching⁵ by targeting the DNA regions that precede the exons encoding for the different types of constant region. Processing of deoxyuridine by DNA repair enzymes then produces the DNA breaks necessary for class switch recombination (CSR) (reviewed in ref. 8). This allows the B-cell to switch the production of the default IgM antibodies to another isotype (IgG, IgE, IgA), thus acquiring specialized biological properties. The importance of AID for immunity is evident from its biological functions, but as a trade-off AID facilitates neoplastic transformation by mutating proto-oncogenes and tumor suppressor genes or initiating chromosomal translocations.^9–11 Unlike the site-specific recombinase RAG, AID has no sequence specificity and therefore it is more inclined to act off-target. Not surprisingly, there are many steps at which AID is regulated: mRNA stability,^12–14 protein stability,¹⁵ activity by phosphorylation^16,17 and subcellular localization.^18,19

AID is a Nucleocytoplasmic Shuttling Protein

Compartmentalization is a common strategy to regulate enzymatic activity. It was early on observed that overexpressed AID-GFP was cytoplasmic, despite which it increased SHM in a B cell line.²⁰ It was later found that the last ∼10 residues of AID contain a Leucine-rich nuclear export signal (NES) that is recognized by the exportin CRM1.^18,19,21 Indeed, inhibition of CRM1 by the antibiotic leptomycin B led to a substantial increase in nuclear AID-GFP and demonstrated that AID is a nucleocytoplasm shuttling protein.^18,19,22 In those reports, the mechanism for AID nuclear import was not explored in depth. A bipartite nuclear localization signal (NLS)¹⁸ as well as passive diffusion^19,22 were equally proposed as ways for AID to access the nucleus. In any case, AID nuclear export seemed to be dominant over any import pathway so it was assumed that nuclear export essentially determined AID subcellular localization, which was consistent with the available data. However, we have recently found that AID compartmentalization is also determined by active nuclear import and cytoplasmic retention mechanisms,²³ which we discuss herein.

AID Nuclear Import

The molecular mass of AID, 24 kDa, is well below the nuclear pore cut-off.²⁴ So, it was indeed a possibility that AID would passively diffuse into the nucleus. However, the fact that treating cells with leptomycin B causes human AID-GFP to accumulate into the nucleus,²³ actually suggests that AID is actively imported into the nucleus. Would AID passively diffuse through the nuclear pores one would expect it to follow the mass action law reaching equilibrium across the nuclear membrane once nuclear export is inhibited. In contrast, nuclear accumulation suggests an import rate higher than passive diffusion and therefore an active mechanism.

Additional lines of evidence indicated that AID nuclear import is an active process.²³ First, AID was able to mediate nuclear import of proteins as large as 175 kDa, which could not proceed by passive diffusion. Second, AID interacted with importin-α proteins, known mediators of facilitated nuclear import. Third, the nuclear accumulation of AID-GFP after leptomycin B treatment could be inhibited by oxidative stress, which is known to inhibit all importin-α-mediated nuclear import^25,26 but it is unlikely to inhibit passive diffusion. Finally, we found that depleting cells of energy by inhibiting glycolysis and mitochondrial respiration led to constitutively nuclear AID variants diffusing out of the nucleus. The effect was observed in ∼40% of the cells rather than the 100% of the NLS^SV40-GFP control, but it was statistically significant and reversible. We left open the possibility of nuclear retention contributing to the observed accumulation based on this quantitative difference but the dependence of AID nuclear localization on energy still indicated an active process. We found that the kinetics of AID nuclear accumulation differs between cell lines, being notably slower in HEK293 than in Hela or Ramos cells. This might reflect differences in the nuclear import machinery but we cannot rule out that this is due to differential AID cytoplasmic retention (see below).

Despite considerable evidence for an active nuclear entry mechanism, prediction algorithms fail to consistently identify a conserved NLS in AID (not shown). However, the interaction with importin-α would suggest the presence of a positively charged NLS in AID.²⁷ The first 26 amino acids of AID have been proposed to contain a bipartite NLS¹⁸ but we rather found that the minimal region of AID that was sufficient to mediate the transport of large proteins into the nucleus was much larger, encompassing most of AID. The N-terminal part of AID indeed constitutes most of this NLS and is required for binding to importin-α but we found that changes in at least three positively charged motifs between AID residues 19 and 54 compromised AID nuclear import. Based on the analysis of a large panel of AID mutants we proposed a structural NLS for AID. The identity of the residues making contact with importin-α versus those required to display these residues in the right conformation remains to be determined. Moreover, we did not do an exhaustive mutational analysis so additional residues may also contribute to the NLS we reported. In any case, our results are best explained by AID nuclear import requiring a certain conformation rather than being mediated by a linear signal. This contrasts with the NES, which functions autonomously as a 16 or even 10mer peptide.^18,19

The exact mechanism mediating AID nuclear import was not characterized beyond the interaction with importin-α, but the formation of the importin-α/β heterodimer would be expected.²⁷ Based on their sequence similarity, importins-α are subdivided into three subgroups that usually share specificity for their cargos.²⁸ AID was pulled down in vitro with similar efficiency by importin-α1, -α3 and -α5, representatives of each of these subgroups, but there may still be some preference in vivo.

Cytoplasmic Retention of AID

We demonstrated that AID and AID-GFP could diffuse into and out of the nucleus through nuclear pores using energy depletion experiments (see above) and by comparative analysis with APOBEC2 (an AID paralog of very similar size and structure whose subcellular distribution is achieved by passive diffusion).²³ Therefore, it was an exciting finding from our study that passive diffusion of AID from the cytoplasm into the nucleus was somehow hindered, thus implying the existence of cytoplasmic retention.

Cytoplasmic retention is most clearly evidenced in the absence of the other two mechanisms that determine subcellular localization of AID. AID nuclear export is inhibited by leptomycin B and by mutations in the NES.^18,19 We found that AID nuclear import could be prevented either by mutations in the NLS, by N-terminal fusions (like in GFP-AID) or by oxidative stress. If active nuclear import and export were the only mechanisms determining AID subcellular localization, any combination of conditions simultaneously inhibiting them both should lead to homogeneous redistribution of AID by passive diffusion. However, we found that, in those conditions, AID-GFP and even untagged and unmutated AID would remain cytoplasmic²³ (as it is illustrated in Fig. 1).

Figure 1.

AID mutants reveal different mechanisms determining its subcellular localization. Illustrative examples of confocal pictures of HeLa cells transfected with AID-GFP and mutants thereof in steady state or after treatment with leptomycin B are shown. Cytoplasmic (C), nuclear (N) or homogeneous (N + C) distribution GFP signal is indicated. AID^ID (AID import-deficient) is an AID-APOBEC2 chimera in which residues 19–57 of AID were replaced with the homologous region of APOBEC2 (residues 60–96). AID^ID is incompetent for nuclear import and is kept in the cytoplasm by nuclear export and cytoplasmic retention and therefore it is used here to reveal mutations that affect the latter mechanism. The mutations/condition combinations more clearly indicating the existence of cytoplasmic retention are boxed in dashed lines. The distribution of AID^ID D188A + leptomycin B is labeled as passive diffusion to highlight the difference with AID D188A in steady state, which also shows C + N distribution but in that case achieved through an equilibrium between active import and export.

The regions mediating nuclear export and cytoplasmic retention partially overlap within the C-terminal domain of AID, complicating the characterization of the latter. Nevertheless, truncated AID variants showed that the C-terminal region of AID is sufficient to significantly delay GFP diffusion from the cytoplasm in the presence of leptomycin B. Furthermore, we identified separation of function point mutations that could distinguish nuclear export from cytoplasmic retention (see Fig. 1). On one hand, we used mutation L198S, affecting a critical position of the NES,¹⁹ which we confirmed by showing that it disrupts CRM1 binding. The severe impairment of import-deficient AID L198S to diffuse into the nucleus (even after leptomycin B treatment) shows that the NES can be functionally separated from cytoplasmic retention. On the other hand, we identified mutations D187A and D188A, which did not affect nuclear export as shown by the nuclear exclusion of import-deficient AID D187A/D188A and by the ability of these mutants to still bind to CRM1 as efficiently as AID. However, import-deficient AID D187A/D188A redistributed homogenously throughout the cell only after nuclear export inhibition suggesting loss of cytoplasmic retention but conserved nuclear export.

Thus, our results indicate the existence of some mechanism retaining AID in the cytoplasm and preventing its diffusion into the nucleus. The molecular basis of this retention remains to be elucidated but since it works in a variety of lymphoid and non-lymphoid cells it is probably mediated by ubiquitous factors. Interaction with the cytoskeleton²⁹ and/or the oligomeric state of AID²³ could both contribute to this cytoplasmic retention. The AID cytoplasmic retention mechanism might be similar to the one acting on the AID paralog APOBEC3G, but the molecular bases of this mechanism are also unknown.^30–32

AID Phosphorylation and Subcellular Localization

The AID residues Thr27 and Ser38 are PKA targets in vivo^16,17 and could conceivably affect nuclear import. Phospho-Ser38 is particularly important for AID function^16,17,33 and is enriched in chromatin-associated AID.^33,34 Two of the AID-APOBEC2 chimeric proteins we used could affect these PKA consensus target sites [RRX(S/T)] in AID. One chimera changes AID residues 34–36, disrupting the Ser38 site (from RRDS to AQDS) and showed decreased nuclear import. Another chimera affecting the Thr27 site produces a less drastic change (from RRET to RNET) and had intact nuclear import. To directly investigate a putative role of PKA phosphorylation in nuclear import we analyzed the subcellular localization of phospho-null mutants. AID T27A, S38A and T27A/S38A behaved similarly to AID-GFP in HEK293T cells, being excluded from the cytoplasm in steady state and accumulating in the nucleus after nuclear export inhibition (Fig. 2). AID S38A showed a very modest increase in the number of cells with exclusively nuclear signal. Constructs containing T27A seemed to produce a larger proportion of cells with homogeneous signal distribution after leptomycin B but the averaged differences from three independent experiments were not statistically significant from AID-GFP (Fig. 2A). After performing these experiments we found that the nuclear accumulation of AID-GFP in HEK293 cells is quite slow compared to other cell lines.²³ In addition, it has been reported that AID is not fully phosphorylated in HEK293 cells compared to B cells.³⁵ So, we treated transfected HEK293 cells with PKA inhibitors or activators but this still had no effect on the kinetics of AID-GFP nuclear accumulation (Fig. 2B). Finally, repeating the experiments in Hela cells still showed no differences in steady state nuclear exclusion or ability to accumulate in the nucleus after leptomycin B treatment between AID and any of the phospho-null mutants (Fig. 2C). Within the C-terminal region of AID, Tyr184 is phosphorylated in vivo.^16,33 Although the biological significance of this modification is unknown, given its location it could affect nuclear export or cytoplasmic retention. However, neither Y184A (Fig. 2D) nor Y184D (not shown) affected the distribution of AID-GFP or of an import-deficient AID variant. These observations have the obvious limitations of using heterologous systems and overexpressed protein and therefore a role for phosphorylation in regulating AID localization may still exist in B cells. The availability of knock-in AID-GFP³⁶ and AID phospho-null^37,38 mice will allow revisiting this issue in a more physiological set up. Given the clear relationship between AID localization and protein stability, one would expect that if phospho-null mutations had any effect on localization this would be reflected on the overall AID levels.¹⁵ However, AID S38A shows the same protein expression levels than unmutated AID in knock-in mice.^37,38 In conclusion, although they may still play subtle roles, for instance by affecting a small proportion of AID that would not be detected in these assays, phosphorylation at Ser38, Thr27 or Tyr184 does not seem to be essential for AID shuttling.

Figure 2.

Nuclear import and cytoplasmic retention of phospho-null AID mutants. (A) C-terminally GFP-tagged AID or the indicated phospho-null variants were imaged by confocal microscopy in transiently transfected HEK293 cells in steady state or after 4 h treatment with 50 ng/ml Leptomycin B. Multiple pictures were taken and the distribution of GFP signal scored blindly as nuclear (N), nuclear and cytoplasmic (N + C) or cytoplasmic (C). Representative pictures are shown as well as the averages + SD for three independent experiments. Only differences indicated with an asterisk were significative in Student t-test (p < 0.05) compared to the same category for AID. (B) HEK293 cells transiently expressing AID-GFP were treated with 50 µM Forskolin + 100 µM 3-Isobutyl-1-methylxanthine (IMBX), to boost cAMP cellular levels or with 10 µM of the protein kinase A inhibitor H-89 and 1 h later treated with 50 ng/ml leptomycin B and aliquots were fixed and processed for confocal microscopy at the indicated times. Representative pictures of the overlay of GFP (green) and DNA stained with propidium iodide (red) signals are shown. (C) Representative confocal pictures of HeLa cells transiently expressing AID-GFP or the indicated phospho-null variants in steady state or after treatment with Leptomycin B. (D) Similar experiments to (C) using AID and AID^ID carrying or not the Y184A mutation. All pictures are at 630X magnification, scale bars, 10 µm.

Integrating AID Compartmentalization Mechanisms

The description of AID active nuclear import and cytoplasmic retention in addition to nuclear export revealed a more complex regulation of AID compartmentalization than previously suspected. Active nuclear import makes sense considering the inability of AID to passively diffuse. Indeed, we found that enzymatically active AID variants carrying mutations affecting its nuclear import fail to induce CSR. However, we cannot rule out that these AID mutants also lose the ability to interact with some unknown partner, which could also explain their functional deficiency. Analogously, CSR has some undefined requirement for the AID C-terminal region,^39,40 which seems to be different from just nuclear export.⁴¹ AID with C-terminal truncations or point mutations have increased ability to drive SHM, Ig gene conversion and ectopic mutations compared to AID.^15,18,19 However, although these experiments show that nuclear exclusion restricts AID function, some of those truncations/mutations probably affect cytoplasmic retention as well. Our data suggests that cytoplasmic retention contributes significantly to AID nuclear exclusion but those mutations we could find that reduce it also abrogate CSR, probably by affecting the CSR-specific C-terminal AID motif. Understanding the functional role and relative contribution of cytoplasmic retention to AID biology compared to nuclear export will need further research.

One could envisage two possible scenarios on how the subcellular localization of AID might regulate antibody diversification. In one scenario, a significant redistribution of AID from the cytoplasm into the nucleus (e.g., at some cell cycle stage and/or following some signaling event) is necessary for efficient antibody diversification (Fig. 3, model 1). A higher amount of AID-YFP in the nucleus of DT40 cells during G₁ has recently been reported, which may result from changes on the stability of nuclear AID rather than from AID shuttling regulation.⁴² It must be noted that this was observed for overexpressed protein in letpomycin B-treated cells but it has not been found (or reported) in the steady state in any of the several works that analyzed the localization of tagged-AID in a variety of cell lines.^{18–20,22,23,29,36,42} Since these cultures are asynchronous with a good proportion of cells in each cell cycle stage, one would expect the accumulation of nuclear AID at some stage to be detectable in a proportion of the cells. Furthermore, using immunofluorescence we could not see any relocation of endogenous AID into the nucleus of Ramos B cells following inhibition of nuclear export.²³ All these studies have technical caveats so further work is required to see whether they reflect what happens for primary B cells during the germinal center reaction. So far, immunohistochemistry on human tonsils and B-cell lymphomas either failed to,⁴³ or detected only a very small proportion of B cells displaying nuclear AID accumulation.⁴⁴ Thus, there is little evidence for large variations in the subcellular localization of endogenous AID in steady state.

Figure 3.

Two possible models by which subcellular localization could regulate AID activity. B cell cytoplasm (C) and nucleus (N) are drawn schematically. The amount of AID in each compartment is represented by proportional shades of gray. The stages at which antibody diversification is expected to occur is indicated. In model 1, diversification would need not only AID expression but also some further signal to induce nuclear translocation and redistribution of AID. This state would be reversible adding to the capacity of regulation. Note that not all of AID needs to be imported into the nucleus in this model but only requires that the nuclear proportion of AID changes compared to the basal stage. In the alternative model 2, the induction of AID would be sufficient for antibody diversification. In this model the nuclear proportion of AID is constant, in equilibrium with the cytoplasmic fraction. This model predicts that variations in the overall levels of cellular AID would lead to a proportional variation in nuclear AID protein, which would affect antibody diversification in the same direction. These variations have been observed so far in pathological or experimental conditions but it is unknown whether they might happen in B cells as a physiological means of regulating AID.

An alternative scenario would envisage most of AID in the cytoplasm in Le Chatellier's equilibrium⁴⁵ with a small nuclear fraction. In this case the proportion of nuclear AID would be constant and enough for antibody diversification (Fig. 3, model 2). Any decrease or increase of the overall AID levels would lead to a proportional decrease or increase of AID in the nuclear fraction, followed by the corresponding effect on AID biological activity. This hypothesis is consistent with the quasi-proportional decrease on SHM and CSR observed in AID haploinsufficient mice^46,47 as well as with the increase found in mice overexpressing transgenic AID⁴⁸ or lacking miRNAs that negatively regulate it.^12–14 An indication that nuclear accumulation of AID is not necessary for antibody diversification is given by AID variants with reduced nuclear import, which following leptomycin B treatment are undetectable in the nucleus by microscopy but can still catalyze substantial CSR, at least when retrovirally expressed.^23,49

In conclusion, the integration of three mechanisms impinging on AID subcellular localization, surely with the contribution of the mechanisms regulating AID stability,¹⁵ seems to determine the amount of nuclear and therefore biologically active enzyme. Some of these mechanisms, as well as how the balance is achieved and regulated are still ill defined and offer an exciting field of study.

Abbreviations

AID

activation induced deaminase

Ig

immunoglobulin

SHM

somatic hypermutation

CSR

class switch recombination

NLS

nuclear localization signal

NES

nuclear export signal