So far, more than 150 familial AD-causing mutations in PSEN1 have been identified, approximately 10 additional mutations have been found in the homologous gene PSEN2 and 25 mutations have been identified in the APP gene (http://www.molgen.ua.ac.be/ADMutations). The study of PSEN1 is therefore crucial for understanding the pathogenesis of familial AD.
Most mutations in PSEN1 are simple missense mutations that result in single amino-acid substitutions in presenilin 1. Some are more complex, for example, small deletions, insertions or splice mutations. The most severe mutation in PSEN1 is a donor–acceptor splice mutation that causes two amino-acid substitutions and an in-frame deletion of exon 9 (Fig 1). Significantly, however, the biochemical consequences of these mutations for γ-secretase assembly are limited (Bentahir et al, 2006; Steiner et al, 1999). Although as many as one-third of the 467 amino acids in the open-reading frame of presenilin 1 are affected by disease-causing mutations, a truncation or absence of the protein has never been observed, indicating that haploinsufficiency does not cause AD. Rather, at first glance and from a strictly genetic perspective, these different clinical mutations all seem to lead to a specific gain of toxic function for PSEN1. Investigations over several years, however, have not succeeded in translating this genetic concept into molecular terms—that is, explaining how the mutations scattered over presenilin 1 all cause a similar gain of toxic function in the protein.
Mutations in either presenilin or APP consistently increase the relative ratio between the long (Aβ42) and short (Aβ40) amyloid peptides (Aβ42/Aβ40; Borchelt et al, 1996; Scheuner et al, 1996). Given that inactivation of Psen1 and Psen2 completely prevents Aβ generation (Herreman et al, 2000; Zhang et al, 2000), this increase can indeed be explained as a gain of toxic function. However, the change in ratio can also be the consequence of a partial loss of Aβ40 generation, as is the case with several PSEN mutations discussed in detail below. Indeed, several authors have challenged the dominant gain-of-toxic-function hypothesis over the years. First, wild-type human PSEN1 can effectively rescue the loss of its suppressor of Notch-family member lin-12 (sel-12) homologue in Caenorhabditis elegans, whereas a mutated PSEN1 is less effective or not effective at all (Baumeister et al, 1997; Levitan et al, 1996). Second, Shen and co-workers have shown that a total loss of Psen function in the forebrain of mice causes neurodegenerative disease in the absence of Aβ (Saura et al, 2004). Third, several groups have reported that specific loss of Psen1 in the mouse forebrain affects particular aspects of memory (Feng et al, 2001; Yu et al, 2001). Both neurodegeneration and memory deficits are important features of AD; however, it might be dangerous to extrapolate these observations to human pathology. Accordingly, it is difficult to correlate the total loss of four Psen alleles in a mouse model (Saura et al, 2004) with the relatively mild single mutation of one PSEN allele in familial AD patients. Indeed, neurodegenerative phenotypes have not been observed in animal models with only one allele inactivated (Psen1+/− or Psen2+/−; for a more detailed overview of the different Psen-knockout mouse models, see Marjaux et al, 2004). Furthermore, it is unclear whether the memory deficits in mice with a forebrain-specific Psen1-knockout can really be compared with the memory deficits in patients with AD. In this regard, some of the memory deficits that result from APP overexpression in AD mouse models can be alleviated by Psen1 inactivation (Dewachter et al, 2002; Saura et al, 2005). Fourth, in the past few years, some studies have indicated that genuine loss-of-function PSEN1 mutations could be involved in forms of frontotemporal dementia without the involvement of Aβ (Amtul et al, 2002; Dermaut et al, 2004; Raux et al, 2000). However, formal genetic or molecular proof that these mutations are responsible for the neurodegenerative process in these patients has not been provided. In fact, an additional mutation in the progranulin gene (Baker et al, 2006; Cruts et al, 2006) in a patient with the presenilin 1 Arg352 insertion (Boeve et al, 2006) is probably the cause of the dementia, which implies that, at least in this case, the mutation in presenilin 1 is a polymorphism. Finally, promotor polymorphisms in the PSEN1 gene that decrease its expression contribute to the risk of early-onset AD (Theuns et al, 2000); however, whether these affect amyloid generation is not yet known.
In conclusion, although the current research clearly indicates that presenilin 1 is important for maintaining the integrity of the brain, it is less clear whether the severe deficits in homozygous loss-of-function mouse models are relevant to the pathology in human patients. Furthermore, PSEN1 deficiency is unlikely to contribute to the disease process in patients with APP mutations, which implies that the effects of loss of PSEN1 function on APP processing are crucial for our understanding of the pathogenesis of AD. I do not, however, exclude the possibility that partial dysfunction of PSEN1—for example, in the Notch signalling pathway that modulates neurite outgrowth and brain repair—makes the brain more prone to Aβ toxicity. This would fit with the previously proposed ‘two-hit’ model for AD (Marjaux et al, 2004), and would also explain why familial AD generally strikes earlier and is more aggressive than sporadic AD.