Presenilins (PS) are highly homologous polytopic membrane proteins that play a critical role in intramembranous processing of amyloid precursor proteins (APP), leading to the production of Aβ peptides (for reviews, see 1234). Inheritance of mutations in PSEN1 and PSEN2, that encode PS1 and PS2, respectively, cause familial forms of Alzheimer's disease (FAD) 56, and do so by elevating the ratio of Aβß₄₂/Aβ₄₀ peptides 78910. Hence, it has been argued that FAD-linked PS1 variants cause disease primarily through a gain of function mechanism. However, recent studies have shown that these PS variants also result in a significant loss of γ-secretase function that may critically contribute to the pathophysiology of FAD (for a review, see 11).

Over the last several years we assessed the gene expression profiles of PSEN1 knockout mice and PSEN1 mutant transgenic animals, and have defined a set of PSEN1-dependent genes 12131415. However, it is important to point out that due to redundant functions and distributions of PSEN1 and PSEN2 1617, cortical PSEN1 ablation in mice results in only a limited phenotype. In contrast, forebrain PSEN1/PSEN2 double KO mice show a strong and progressive neurodegenerative phenotype which is characterized by both anatomical and behavioral changes 1118. Although anatomical changes are not apparent at 2 months of age, these mice already show mild memory impairments at this age. The behavioral phenotype progresses with time and the mutant mice begin to exhibit excessive grooming behavior, increased stereotypy in the open field, and reduced latency in the rotarod test at the age of 6 months 18. The anatomical changes develop over time, and the progressive thinning of cortical layers becomes prominent by the age of 6 months.

To gain a better understanding of presenilin-dependent neurodegeneration in the cortex, we performed an expression profiling of the hippocampus (HC) and frontal cortex (FC) of mice with forebrain ablations of PSEN1 and PSEN2. We used DNA microarrays to analyze the transcriptome at five developmental time points. We chose to examine the temporal nature of gene expression changes because the PSEN1 allele is conditionally deleted by a CamKII-cre allele that promotes recombination of "floxed" loci first in the hippocampus, and then spreads throughout the entire forebrain. We focused our attention on the following questions: 1) what are the gene expression consequences of PSEN1/PSEN2 ablation; 2) do the altered genes share common functional characteristics; 3) how does gene expression change over time as a function of PSEN1/PSEN2 deletion; 4) are the molecular signatures of neurodegeneration similar in the HC and FC and 5) is presenilin-dependent molecular neurodegeneration progressing similarly in FC and HC?

The experimental design and analysis strategy are outlined in Figure 1. Briefly, frontal and hippocampal cortices of PSEN1/PSEN2 KO (PSKO) and matched control animals of different postnatal ages were analyzed by MG_430 oligonucleotide arrays by Affymetrix. An RMA-based normalization was followed by a multi-level analysis that compared the samples based on the genotype, age and brain region.

Over the past few years, we have used gene microarray strategies to examine transcript levels in brains of mice with conditionally inactivated PSEN1 (cPS1 KO) alleles. While these studies have proven informative, redundant function and distribution of PSEN1 and PSEN2 in hippocampus and cortex 1617 has limited the interpretation of our findings. Indeed, while cPS1KO mice fail to exhibit any neuroanatomical or behavioral alterations 1920, cPS1KO mice on a PSEN2 KO background show a strong and progressive neurodegenerative phenotype that is characterized by both anatomical and behavioral changes 1118. These animals do not exhibit anatomical changes at 2 months of age, yet show mild memory impairments at this age. The behavioral phenotypes progress over time and are paralleled by considerable cortical atrophy and neurodegeneration. In the present study, we performed expression profiling in order to develop an understanding of the mechanism(s) that might mediate these morphological and behavioral observations, and we now offer several novel insights. First, we show that simultaneous ablation of PSEN1 and PSEN2 leads to a wide variety of gene expression changes that progress over time. Second, the frontal cortex and hippocampus are both affected in PSKO brains, albeit by the notable difference that gene expression changes in the FC are characterized by elevated transcript levels, while the HC is mostly characterized by reduced transcript levels. Third, the transcriptome changes suggest a progressive neurodegenerative process with strong immune system activation. Fourth, based on the BioCarta pathway analysis, it appears that GABA-ergic neurons may be a preferential early target of the neurodegenerative events. Fifth, the Complement and Cathepsin systems appear to be critical contributors to neurodegeneration in both the hippocampus and frontal cortex.

One of the most intriguing findings of the manuscript is that transcript repressions dominate in the FC, while transcript inductions are the hallmark of the HC tissue. Although the most prominent presenilin deletion induced transcript changes may be different between these two brain regions, the deletion signature of the FC expression change was strong in HC and vice versa. We interpret these findings that the neurodegenerative events in FC lag behind of those in HC, suggesting that the early stages of degeneration are characterized by a strong, widespread repression of mRNA species. However, as the pathophysiological process proceeds, in addition to the transcript repression, a de novo inflammation-related transcript induction occurs, which becomes a dominant feature of a gene expression profile in the HC. Alternatively, the differential cytoarchitecture and molecular composition of FC and HC provide a differential framework for the adaptational responses of these two brain regions, and results in a different gene expression profile.

Presenilins associate with Aph1, Pen2 and Nicastrin in a large complex that mediates intramembranous, "γ-secretase" 1 proteolysis of a variety of type-I membrane proteins, including Notch, APP, E-cadherins, N-cadherins, and p75NTR, amongst others (reviewed by 21). For most of these proteins, γ-secretase-mediated processing results in the generation of intracellular derivatives, termed ICDs, that have been shown to initiate complex signaling cascades, including those involved in cell adhesion, lateral inhibition of cell fate decisions, neurotrophin signaling and cell differentiation 2223. It has also become increasingly evident that PS have functions that go beyond their role as the catalytic entity of γ-secretase activity. For example, PS located in the endoplasmic reticulum (ER) and Golgi apparatus regulate calcium homeostasis 24, findings that may be equally relevant for neurodegeneration observed in the PSKO mice. Not surprisingly, with increasing age, the PSKO mice develop striking neurodegeneration of the cerebral cortex and worsening impairments of memory and synaptic function 18. However, it is important to point out that gene expression changes arise before noticeable anatomical neurodegeneration. The progressive gene expression changes that we observed in both hippocampus and frontal cortex also appear to suggest that the cortical GABA-ergic interneurons may be one of the early targets of neurodegeneration mediated by the ablation of PSEN genes. As the GABA-ergic system has a strong influence on cognitive behaviors 2526, the expression changes in interneurons are likely to be related to the altered cognitive processes reported in these mice. On the other hand, interneurons have been shown to exhibit enhanced vulnerability to various insults 27. These events likely result in necrosis-induced microglial activation, accompanied by overexpression of the complement system transcripts, leading to neuroinfammatory processes 28, thus further accelerating the neurodegenerative processes.

The mechanism(s) by which ablation of PSEN genes leads to neurodegeneration are not fully understood, but likely involve several converging pathways. In this regard, our dataset is highly informative: first, we observe a strong and progressive increase in expression of transcripts encoding molecules in the calpain-cathepsin system. These results confirm and expand the previously reported upregulation of Cathepsin S in these mice 29. Both calpains and cathepsins belong to the papain superfamily of cysteine proteases that are primarily localized in lysosomes and are critical for intracellular protein catabolism. Calpain-mediated extralysosomal release of cathepsins is a critical event in necrosis, and cathepsin inhibitors show strong neuroprotection in models of acute neuronal insults 3031. Thus, we propose that neurodegeneration in these mice is a result of a progressive neuronal calpain-cathepsin system activation over time. How does loss of PS function lead to induction of the calpain-cathepsin system? We offer two proposals that are not mutually exclusive. In the first, we suggest that in the absence of PS/γ-secretase activity, extremely high levels of membrane-tethered stubs derived from type 1 membrane proteins hyperaccumulate in endosomal-lysosomal compartments, hence "saturating" those proteases that would otherwise be necessary for maintaining intracellular homeostasis. In this setting, the induction of the calpain-cathepsin system would thus reflect a compensatory mechanism. This hypothesis is consistent with the model that a function of PS/γ-secretase is to serve as a membrane-bound proteosome that promotes intramembranous proteolysis in order to enhance clearance of membrane-tethered stubs derived from ectodomain shedding of type 1 membrane proteins 32. Indeed, we have shown that cultured mammalian cells expressing dominant-negative PS1 variants harboring mutations at the critical aspartate 257 and 385 residues necessary for γ-secretase activity, exhibit unusual accumulations of distended intramembranous organelles resembling multivesicular bodies that are immunoreactive with antibodies specific for membrane-tethered stubs 33. The second model to explain our findings is in the absence of PS expression, the cell surface levels of receptors and ion channels in either pre- or postsynaptic membranes is altered in manner that gives rise to inappropriate signaling events and alterations in calcium homeostasis. In support of this notion, we 33 and others 34 have shown that cells that express dominant-negative forms of PS1 or that lack PSEN alleles, respectively, exhibit elevated levels of a variety of type I membrane proteins and their membrane-tethered stubs.

Inhibition of PS/γ-secretase activity has been debated as a promising therapeutic strategy for treatment of Alzheimer's Disease 3536. In this context, our results offer a cautionary note and suggest that the preclinical evaluation of such agents should include close monitoring for signs of calpain-cathepsin system activation, and development of a peripheral tissue assays on the complement-cathepsin system expression as informative tools for assessment of potential adverse effects that may also be present is the CNS.

SSS discloses that he is a paid Consultant of Neuropharma, Inc., Torrey Pines Therapeutics and Eisai Research Labs Inc, but does not a shareholder in any company that is a maker or owner of a FDA-regulated drug or device.

All authors read and approved the final manuscript. KM and SSS designed the study and supervised all steps of the experiments. KM wrote the initial draft of the manuscript. KG carried out the molecular biology part of the studies, SHC and XZ generated the animals and harvested the brain tissue, PE and KM performed the statistical analyses, while EMN performed the Western blotting for C1q and Ctsd.