|Home | About | Journals | Submit | Contact Us | Français|
ATM is a PI3 kinase family member that plays an important role in DNA double strand break repair. ATM deficiency leads to ataxia-telangiectasia (AT), a syndrome involving cancer susceptibility, hypersensitivity to ionizing radiation, immune deficiency and sterility1, 2 – phenotypes that can straightforwardly be attributed to a defective response to DNA damage. Yet patients with AT also suffer from ataxia, speech defects and abnormal body movements3–5 – neurological phenotypes whose origins remain largely unexplained. Compounding the discordance, ATM mutations in mouse interfere with DNA repair but have only mild neurological symptoms6–9 suggesting that the link between DNA damage and the death of neurons and astrocytes can be broken10–12. We find that in neurons ATM protein has a significant cytoplasmic distribution. The functional importance of this cytoplasmic ATM is supported by our finding that in Atmtm1Awb mice long term potentiation in the Schaffer collateral pathway, induced by theta burst stimulation, is significantly reduced, as is the rate of vesicular dye release. ATM can be isolated from synaptosomes where it is found in a complex with two synaptic vesicle proteins: VAMP2 and synapsin-I, both of which must be phosphorylated to bind ATM. The neurological symptoms of ataxia-telangiectasia may thus result from defective cytoplasmic functions of ATM as well as defective DNA repair.
ATM is best known for its critical role in the DNA damage response where, after autophosphorylation on serine 1981 (S1981), the ATM dimer dissociates into two catalytically active monomers13. While ATM is a predominantly nuclear protein, in neurons studies have shown significant amounts of cytoplasmic ATM protein, the function of which has not been completely identified14–18. We performed subcellular fractionations of tissue extracts from mouse brain, spleen and thymus (Figure 1A). All three tissues had nuclear ATM (lanes 4–6), while in spleen and thymus, cytoplasmic ATM was nearly undetectable (lanes 2, 3). By contrast, in brain tissue, cultured neurons or N2a cells (Figure 1C, lanes 1, 2) significant quantities of cytoplasmic ATM were present. Antibody specificity was confirmed by the absence of a band in Atmtm1Awb tissues (Supplementary Figure 1C). Immunohistochemistry of cryostat sections (Figure 1B1, 1B2; Supplementary Figure1A and B) and of cultured cells (Figure 1D4, 1D5) was consistent with these findings. A similar nuclear/cytoplasmic distribution was found with a GFP-ATM fusion protein, transiently expressed in either cortical neurons (Figure 1D1) or N2a cells (Figure 1D2). Non-neuronal cells such as NIH3T3 and HeLa had predominately nuclear ATM (Figure 1C, lanes 3, 4; Figure 1D3). Further, in NIH3T3 cells (Figure 1D6), GFP-ATM was overwhelmingly nuclear. Finally, ATM could be found in association with synaptic vesicles and synaptic membranes in synaptosomal fractions from mouse brain (Figure 1E). Thus, four independent lines of evidence support the existence of neuronal cytoplasmic ATM.
The neuron-specific occurrence of cytoplasmic ATM suggests a function distinct from its role in the DNA damage response. If primary neuronal cultures are exposed to 5 Gy of radiation, the phosphoS1981-ATM epitope appears in the nuclei neurons or N2a cells as reported previously19, but not in the cytoplasm (Figure 1F). These results were verified using etoposide (Figure 1G and Supplementary Figure 2) and in vivo using ionizing radiation (Supplementary Figure 3A). ATM activation in the nucleus was demonstrated by the enhanced phosphorylation of two ATM targets: p53 and Chk2 (Supplementary Figure3B). Thus, neither irradiation nor topoisomerase inhibitors activate cytoplasmic ATM, suggesting its function is unrelated to the DNA damage response.
The cytoplasmic/synaptosomal presence of ATM in neurons prompted us to ask whether ATM might play a role in synaptic function. We recorded from in vitro slices of adult wild type and Atmtm1Awb hippocampus, but found no deficit in baseline synaptic transmission. Both genotypes had similar input-output curves (Figure 2A), although there was a tendency for the mutants to show a plateau in responsiveness at lower stimulus currents than wild-type mice. We next tested the consequences of ATM deficiency on long-term potentiation (LTP) at the Schaffer collateral-CA1 synapse. Examination of theta burst stimulation (TBS)-induced LTP revealed that Atmtm1Awb mice have a considerable deficit as compared to wild-type animals (Figure 2B). In control mice TBS produced a maintained elevation in synaptic strength; the same stimulation protocol produced only a modest long-term increase in Atmtm1Awb mice (Figure 2B). Previous studies indicate a defect in exocytosis in lymphocytes of A-T patients20. This led us to question whether a defect in vesicle recycling might occur in Atm mutant neurons. We monitored FM4-64 dye uptake and release from cultured neurons derived from either wild type or Atmtm1Awb mice. There was no significant difference in FM4-64 dye uptake between Atmtm1Awb and wild type neurons or wild type neurons treated with RNAi (not shown). By contrast, spontaneous dye release was significantly slower in Atmtm1Awb and ATM siRNA-treated cultures than in wild type, untreated dishes (Figures 2C and Supplementary Videos). Thus, in the absence of ATM, spontaneous vesicle release is reduced in culture and the ability to establish and maintain LTP is significantly impaired.
We sought to determine the mechanistic basis of these functional deficits. Recently, hundreds of potential ATM/ATR targets have been identified21. In light of our findings, two proteins from this list seemed especially noteworthy: VAMP2 and synapsin-I. Both are vesicle proteins that are involved in vesicle trafficking and release. VAMP2 (synaptobrevin) forms a central part of the SNARE complex that mediates synaptic vesicle fusion with the cell membrane during neurotransmitter release. By immunocytochemistry, ATM partially co-localizes with VAMP2 in cultured neurons (Figure 3A). This association was confirmed by co-immunoprecipitation with antibody to either ATM or VAMP2 from protein extracts of adult mouse brain (Figure 3B). Having demonstrated the physical association of VAMP2 and ATM we immunoprecipitated VAMP2 from cytoplasmic fractions of mouse neocortex and cerebellum and examined the precipitates on Western blots with an antibody against [S/T]Q phosphorylated sites – the canonical ATM/ATR target sequence. A strong band of staining was found (Figure 3C). The amino acid sequence of VAMP2 reveals two potential ATM/ATR phosphorylation sites: T35 and S75. We constructed non-phosphorylatable alanine substitution mutants at these sites and expressed them as Flag-tagged derivatives in N2a cells, which we then differentiated with dibutyryl-cAMP. Blots of Flag immunoprecipitates probed with either phospho-[S/T]Q or a specific phospho-T35-VAMP2 antibody (P-T35) revealed that T35 is significantly phosphorylated in normal neuroblastoma cells (Figure 3D). We observed that blocking VAMP2 T35 phosphorylation had two significant consequences. First, immunoprecipitation with ATM antibody reveals that ATM does not bind to the T35A-VAMP2 mutant, suggesting that ATM binds primarily to the phosphorylated form of VAMP2 (Figure 3D). Second, over-expression of the VAMP2 T35A mutant, but not the S75A mutant or wild type VAMP2, in N2a cells was associated with a decrease in the rate FM4-64 dye release from differentiated N2a cells (Supplementary Figure 4). Thus, phosphorylation of VAMP2 at T35 induces ATM binding and leads to a change in neuronal function.
Synapsin-I is an abundant neuronal phosphoprotein that is associated with synaptic vesicles22. Synapsin-I/ATM co-immunoprecipitations reveal that they are physically associated (Figure 3E). To identify the synapsin-I kinase, we immunoprecipitated synapsin-I from wild type or Atmtm1Awb brain and probed with the [S/T]Q antibody (Figure 3F). In the mutant, immunoreactivity for the [S/T]Q antibody was reduced to nearly undetectable levels. Similar results were found in primary cultures after ATM siRNA introduction (Figure 4A). Serine 656 of synapsin-I is a strong theoretical candidate for an ATM/ATR phosphorylation target, a prediction we confirmed by site-directed mutagenesis. Furthermore, the non-phosphorylatable S656A synapsin-I mutant can no longer pull down ATM as efficiently as wild type (Figure 3G). As with VAMP2, blocking synapsin-I phosphorylation has significant functional consequences. Normally synapsin-I and ATM partially co-localize in cultured cortical neurons where they are found in a punctate pattern along the neurites (Figure 3H). In addition to inhibition of ATM binding, the subcellular distribution of both synapsin-I and ATM is altered when S656 phosphorylation is blocked. Co-expression of an mCherry tagged wild type synapsin-I with GFP tagged ATM shows a number of puncta where the two proteins co-localize (Supplementary Figure 5). By contrast, when -synapsin-I carries the S656A mutation, the co-localization with ATM is reduced.
Having identified ATM as the source of S656 synapsin-I phosphorylation we next sought to determine the identity of the VAMP2 kinase. Proteins were immunoprecipitated from primary cortical neurons that had been treated with ATM siRNA. While synapsin-I [S/T]Q reactivity was largely blocked by ATM knockdown, [S/T]Q staining of VAMP2 was only slightly reduced (Figure 4B). A similar result was found when VAMP2 was immunoprecipitated from Atmtm1Awb brain (Figure 3G). Thus, it is unlikely that ATM is a major VAMP2 kinase. We next asked whether ATR, a closely related member of the PI3 kinase family, might be involved. Application of ATR RNAi all but eliminated the phospho-[S/T]Q and phospho-T35 reactivity of VAMP2 immunoprecipitates and immunostaining of primary neurons (Figure 4B & C and Supplementary Figure 6), pointing to ATR as the predominant VAMP2 PI3 kinase. There would thus appear to be a reciprocal relationship among the four proteins with VAMP2 relying on ATR and synapsin-I relying on ATM as a PI3 kinase.
Its participation in the phosphorylation of VAMP2 suggested that ATR might also be a VAMP2 binding partner. Immunostaining of brain tissue or cultured neurons demonstrated that ATR has a significant cytoplasmic distribution (Figure 4D1–3); the distribution of exogenous GFP-ATR suggested a similar conclusion (Figure 4D4). Further, synaptosomal preparations contained ATR in the same fractions as ATM (Figure 1E). ATR physically associates with VAMP2 and synapsin-I, as antibodies to ATR pulled down both synaptic proteins. Reverse immunoprecipitations were also successful (Figure 4E). These multiple interactions led us to ask whether all the proteins co-existed in a single complex. VAMP2 can immunoprecipitate synapsin-I from synaptosomal preparations of cultured cortical neurons, and the reverse immunoprecipitation confirmed the synapsin-I/VAMP2 association (Figure 4F). Yet, when ATM siRNA is used to block the synthesis of ATM in the cultured neurons, VAMP2 and synapsin-I fail to immunoprecipitate each other (Figure 4F). This is strong evidence that ATM acts as a physical link between VAMP2 and synapsin-I. We next asked whether ATM and ATR were associated with each other. We co-expressed ATM and ATR, each with a different epitope tag, in N2a cells. We performed immunoprecipitations with one of the epitope tags and did Western blots using an antibody to the other epitope as probe. A strong band of staining was found in cytoplasmic but not nuclear fractions (Supplementary Figure 7). Reversing the epitope tags produced a similar result. This suggests that the ATM and ATR proteins do indeed interact, but that their physical association in neurons occurs primarily in the cytoplasm.
Neurons and neuronal cell lines appear to be unique in that a substantial fraction of their ATM protein is cytoplasmic, a conclusion supported by three different lines of evidence: subcellular fractionation, immunocytochemistry and the distribution of an exogenous GFP-ATM fusion protein. The prominence of this cytoplasmic component is cell-type specific as neither spleen, nor thymus nor cell lines of non-neuronal origin have significant amounts of ATM outside of their nucleus.
We postulate that this non-nuclear ATM is different in its function as well as in its location. The biochemical and physical relationships among ATM, ATR, VAMP2 and synapsin-I, coupled with the defects in LTP and spontaneous dye release observed in ATM-deficient neuron are strong evidence that cytoplasmic ATM plays cellular roles in neurons that are unrelated to the DNA damage response. Cytoplasmic ATM shows no evidence of activation as assessed by phosphorylation at S1981. Yet, others have speculated that ATM kinase activity unrelated to the phosphorylation of S1981 exists23, 24, and the ATM-dependent phosphorylation of synapsin-I at S656 would tend to support this suggestion.
Both VAMP2 and synapsin-I are best known for their roles in the pre-synaptic nerve terminal, and we have shown that ATM protein is found in synaptosomal preparations. It is unlikely, however, that the function of cytoplasmic ATM is restricted to pre-synaptic terminal as immunocytochemistry also reveals robust staining in the cell soma and dendrites. The actions of cytoplasmic ATM and ATR appear to be symmetrical. We show that all four proteins can physically associate with one another in the cytoplasm, but while the ATM kinase targets synapsin-I, ATR targets VAMP2. The involvement of the ATR kinase in the cytoplasmic function of ATM is intriguing as a partial deficiency of ATR leads to devastating neurological consequences25.
The subtle nature of the synaptic changes in ATM-deficient neurons suggests a regulatory role for cytoplasmic ATM in neuronal activity rather than a central function in synaptic connectivity. It seems likely that either phosphorylation of VAMP2 or its binding to ATM would sterically interfere with its normal function in vesicle docking and fusion. Perhaps ATM holds a pool of VAMP2 and/or synapsin-I in reserve to be released as needed. Alternatively, ATM may assist in the transport of VAMP2 and synapsin-I by keeping them in a protected stat, consistent the distribution of the non-phosphorylatable proteins in cultured neurons.
In the aggregate, these novel associations of ATM with synaptic function provide new and unexpected explanations for the severe neurological symptoms in children with A-T, and possibly open new avenues of intervention specifically related to cytoplasmic ATM function26. It would appear that in conjunction with the cell cycle related cell death27 A-T neurons are subject to a synaptic dystrophy that can only exacerbate the damage caused by cell loss.
siRNA duplexes against mouse ATM and ATR were purchased from Integrated DNA Technologies. We used mainly siRNA in our study; the ATM sense strand was: 5’-GGAGCAUGCUCUAAGGACATT-3’, and the ATR sense strand was:5’-CUCCAAAGCACCACUGAAUTT-3’. Primary cultures of mouse embryonic neurons were transfected with siRNA plus Lipofectamine 2000 first at 5 days in vitro (DIV) and again at 9 DIV. Samples were mainly collected between 10 and 14 DIV.
Dissociated cortical neurons were cultured in vitro for 14 days, then incubated in 10 µM FM4-64 (Molecular Probes). Dye uptake was induced by exposing the neurons to Tyrode’s solution containing 47 mM KCl and 2mM CaCl2 for 90 sec at room temperature. Fluorescent images of FM 4-64 labelled vesicles were captured after 15 min of perfusion with dye-free solution. For live imaging, one frame was captured every 10 seconds over a total analysis period of one hour.
Extracellular recordings of field EPSPs (fEPSPs) are made with ACSF-filled glass electrodes (5–10 µm tip diameter). Test stimuli (0.1 ms) are delivered with a bipolar platinum/iridium stimulating electrode at 1 min intervals except for specialized protocols that elicit changes in synaptic strength (see below). For recordings of CA1 activation by Schaffer collateral stimulation, recording and stimulating electrodes are both placed in stratum radiatum. Each experiment is begun by obtaining input-output relationships to establish the strength of baseline synaptic transmission. A Grass S8800 stimulator connected to a Grass PSIU6 photoelectric stimulus isolation unit is used to deliver a series of increasing intensity constant current pulses. Current magnitude is adjusted to elicit responses ranging from just-suprathreshold to near maximal. Following this, stimulus intensity is adjusted to evoke fEPSPs 30–40% of maximum, typically 30–40 µA. To elicit LTP, theta burst stimulation (TBS) is used. A single TBS consists of 12 bursts of 4 100 Hz pulses spaced 200 ms apart. Response magnitude is quantified using the slope of the field potential.
This work was supported by a grant from the A-T Children’s Project and the NIH (NS20591) to K.H.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.