GSK-3[agr]/[bgr] kinases and amyloid production in vivo

GSK-3β is believed to be central to the pathogenesis of Alzheimer’s disease, linking amyloid and tau pathology2, 3, 4, 5. Unlike GSK-3β, neither physiological functions nor pathological roles of GSK-3α are well explored6. To analyse the function of both GSK-3 kinases in brain, we generated mice that were completely deficient in GSK-3α (denoted as Gsk3aKO) as well as mice with neuron-specific GSK-3α or GSK-3β deficiency using Cre/loxP as for presenilin 1 (ref. 7). The contribution of GSK-3α to the processing of APP was analysed by Phiel et al.1 by silencing with short interfering RNA in cells and by pharmaceutical inhibition with Li+ ions (an inhibitor of both GSK-3 isozymes). We approached this problem genetically by generating two strains of GSK-3α-deficient mice, denoted as Gsk3aKO and Gsk3an−/− (complete and neuron-specific GSK-3α deficiency, respectively). In Gsk3aKO mice, the gene was inactivated completely in all tissues, demonstrated by genotyping, immunohistochemistry and western blotting (data not shown). In the brains of Gsk3an−/− mice, some residual activity was expected because GSK-3 enzymes are expressed in non-neuronal cell types in the CNS5, 8 (Fig. 1). GSK-3α deficiency had no evident impact on GSK-3β levels, demonstrating their independent regulation (Fig. 1). First, we analysed biochemically the proteolytic processing of endogenous mouse APP in brains of adult Gsk3an−/− mice by measuring the APP metabolites in total brain extracts by validated methods5, 7, 9, 10, 11. Neuronal deficiency of GSK-3α did not affect levels of immature and mature APP, APP phosphorylated at T668 (pT668), secreted ectodomain APP, nor of carboxy-terminal fragments (Fig. 1a). Levels of mouse amyloid peptides in the brain, measured by specific enzyme-linked immunosorbent assay (ELISA), were unaffected in Gsk3an−/− mice relative to age- and gender-matched control mice with floxed Gsk3a genes but not expressing Cre recombinase (Fig. 1a). We went on to administer AAV-APP.SLA viral vectors (adeno-associated viral vector expressing triple mutant APP with Swedish, London and Austrian mutations) by intra-hippocampal injection into Gsk3an−/− mice, to analyse the processing of human mutant APP.SLA (ref. 9). At 3 weeks after injection, biochemical analysis of hippocampal extracts of AAV-APP.SLA-injected Gsk3an−/− mice revealed no significant differences for any of the APP metabolites (Fig. 1b). We continued to analyse mice with a complete deficiency in GSK-3α generated serendipitously during the expansion of the Gsk3an−/− colony (data not shown). In addition, we generated bigenic mice by crossing the Gsk3aKO mice with our App(V717I) mice, a validated model for amyloid pathology5, 7, 9, 10, 11. We carried out a biochemical analysis of metabolites of mouse APP and human mutant APP in brains of Gsk3aKO mice and Gsk3aKO×App(V717I) bigenic mice, respectively. None of the APP metabolites derived from either mouse or human APP was affected by the complete deficiency of GSK-3α (Fig. 1c). We subsequently extended the study to the GSK-3β isozyme by generating Gsk3bn−/− mice with a neuron-specific deficiency of GSK-3β. GSK-3β deficiency had no evident impact on the expression of the GSK-3α isozyme, further corroborating their independent regulation (Fig. 2). Notably, no significant deviation was noted in the brain levels of mouse APP metabolites in Gsk3bn−/− mice (Fig. 2). In addition, we attempted to generate mice deficient in both GSK-3 isozymes in their central neurons. The combination of mice containing both floxed Gsk3 genes and the Thy1-Cre recombinase transgene yielded 354 offspring over a time span of 16 months that were all genotyped for the five genes of interest: wild-type and floxed GSK-3 isozymes and Thy1-Cre. None of the pups contained the wanted combination with both Gsk3 genes recombined homozygously. The outcome proved, not unexpectedly, that complete deficiency of GSK-3 activity in central neurons is lethal for mice. We conclude that the GSK-3 isozymes do not contribute significantly to the processing of APP in mouse brain in vivo. Of interest, the combination of Gsk3bn−/− with App(V717I) did not yield viable offspring (data not shown), in contrast to viable offspring yielded by the Gsk3aKO×App(V717I) combination (Fig. 1c). This underlines the substantial functional difference of the GSK-3 isozymes in brain; a difference that we believe is not, however, related to APP processing. The combined outcome substantiates the notion that GSK-3β is physiologically and pathologically the dominant isozyme, notwithstanding the fact that both are activated by amyloid5. The latter point testifies that both kinases contribute, downstream of APP, to tauopathy and cognitive defects in bigenic and viral mouse models3, 5, 12, 13 (data not shown) and by extrapolation in human disease2, 4, 13. Download references