PHARMACEUTICAL COMPOSITIONS CONTAINING SERMs FOR THE TREATMENT OF ALZHEIMER'S DISEASE Field of the invention The present invention refers to a new therapeutic application of drugs named Selective Estrogen Receptor Modulators (SERMs), in particular for the treatment of Alzheimer's disease. State of the art Alzheimer's disease (AD) is the most prevalent form of late-life mental failure in humans and is characterized by a progressive impairment of cognitive functions, such as memory and language. The histopathoiogical hallmark of AD is represented by the accumulation of extracellular β-amyloid (βA) plaques and intracellular neurofibrillary tangles, which are responsible for a complex inflammatory response leading to neuronal degeneration and cell death. The fact that, to date, there is no accurate way to predict the likelihood that an individual will develop AD, to prevent or to effectively arrest the progression of the disease, results in a profound health, social and economic burden. The toxicity of βA is mediated by its interaction with the cell membrane, which results in calcium influx and loss of intracellular calcium homeostasis. Changes in the lipid composition of the neuronal membrane may therefore facilitate βA toxicity. Indeed, it has been recently demonstrated in a feocromocitoma cell line (PC12) that the cells became resistant to βA citotoxicity when the surface membrane was enriched in cholesterol levels. Conversely, the cells were more vulnerable to the action of βA after cholesterol extraction or by inhibiting the de novo synthesis. The hypothesis is that cholesterol deprivation modifies the fluidity of the plasma membrane, thus promoting the incorporation and pore formation of βA into cell membranes. Recently, a novel gene, named seladin-1 (for Selective Alzheimer's Disease lndicator-1 ), has been isolated and found to be down-regulated in brain regions affected by AD. Remarkably, overexpression of seladin-1 in brain tumoral cells conferred protection against βA-mediated toxicity and from oxidative stress. In addition, seladin-1 effectively inhibited caspase 3 activity, a key mediator of apoptosis, and protected from apoptotic death. This gene has marked sequence homology to the Diminuto/Dwarfi gene, described in plants (i.e. Arabidopsis thaliana) and in Caenorhabditis elegans. In plants the product of Diminuto/Dwarfi gene is an enzyme required for the synthesis of brassinosteraids, which are plant sterols essential for normal growth and development. A subsequent study demonstrated that seladin-1 has enzymatic activity, too. In fact, this gene encodes 3-beta-hydroxysterol delta-24-reductase (DHCR24), which catalyzes the reduction of the delta 24 double bond in desmosterol to produce cholesterol. Mutations of this gene have been found in desmosterolosis, a rare severe multiple-congeπital- anomaly syndrome, including developmental and growth retardation. Desmosterolosis belongs to a growing group of genetic diseases due to defective cholesterol synthesis, in which mental and psychomotor retardation are constantly observed, thus supporting the hypothesis that cholesterol plays an important role in preserving neuronal cell integrity. As mentioned above, an effective pharmacological approach for the prevention or treatment of AD does not exist so far. However, there is in vitro evidence, mostly using animal models, that estrogen exerts neurotrophic and neuroprotective effects and a favorable estrogen effect in protecting from AD has been claimed. AD is more common in women and decreased estrogen levels after menopause is a risk factor for the disease. Several studies indicate that estrogen treatment may decrease the risk or delay the onset of AD in postmenopausal women. Other pharmaceutical compounds, which interact with estrogen receptors, have been designed over the years. These compounds are named Selective Estrogen Receptor Modulators (SERMs) and have been widely used in the treatment of breast cancer and osteoporosis. Examples of SERMs according to the invention are: tamoxifen, raloxifene, toremifene, ardoxifene, ospenifene, idoxifene. The effects of SERMs in the brain are less well understood. Nonetheless, there are experimental observations in animals that the SERMs tamoxifen or raloxifene stimulate neurite outgrowth and prevent brain damage in models of ischemic stroke. No trial primarily focused on the role of SERMs in the prevention/treatment of AD has been performed, so far. In a recent large trial of raloxifene in osteoporotic women (7478 patients treated for three years) a trend toward a beneficial effect of this drug limited to some specific tests, such as attention and verbal memory test, was observed. However, these few data do not represent a clear cut indication for a pharmacological approach with SERMs in AD patients, so far, and underscore the need of additional specifically designed experimental studies and clinical trials. SERMs might turn out to be very interesting drugs for the treatment of neurodegenerative diseases, considering also the fact that estrogen/progesterone replacement therapy in post-menopausal women has been shown to increase the risk of breast cancer. Noticeably, in 2002 the estrogen plus progestin therapy in the Women's Health Initiative trial was discontinued because of breast cancer risk. Detailed description of the invention In order to assess the neuroprotective effect of the SERMs (in particular tamoxifen and raloxifene) we used two different neuronal cell models. The first model was represented by neuroblast long-term cell cultures from human fetal olfactory epithelium, previously established, cloned and propagated (1 , 2). These cells (FNC) show unique features, because they synthesize both neuronal proteins and olfactory markers and respond to odorant stimuli, suggesting their origin from the stem cell compartment that generates mature olfactory receptor neurons, the only type of neuron in the olfactory epithelium. In addition, FNC cells express both the subtypes of estrogen nuclear receptors (ER), i.e. a and β (2). Thus, they represent an ideal, and most importantly human, in vitro model to study the neuroprotective role of estrogen and SERMs. The second cell model was represented by mesenchimal stem cells (MSC), obtained from aspirates taken from the iliac crest of normal young volunteers. These cells can differentiate into neurons (3). The fact that a stem cell compartment is normally maintained (4, 5) in specific brain areas affected by AD makes MSC an extremely interesting model to study the effects of neuroprotective drugs. The response of neuronal cells to estrogen [17β-estradiol (17βE2)] or SERMs (tamoxifen and raloxifene) exposure was evaluated by different experimental procedures. A wide range of concentrations was used (100 pM-100 nM). Neuronal cell growth was assessed by both 3H-thymidine incorporation assay and cell counting. We found a stimulatory effect of estrogen on ^H-thymidine uptake, which was statistically significant at all the tested concentrations (100 pM-100 nM). With regard to cell counting, a biphasic effect was observed. In fact, significantly increased cell counts were determined by exposure to 100 pM and 100 nM estrogen (158 + 3% and 142 + 2.5%, respectively, mean + SE, vs control, considered as 100%). Conversely, 1 and 10 nM 17βE2 did not significantly affect cell number, compared to untreated cells. Raloxifene and tamoxifen (100 pM-100 nM) did not determine any effect on cell counts (not shown). In order to test the protective effect of estrogen/SERMs against βA-induced toxicity, cell viability assessment was determined by MTS assay (Promega Corp., Madison, Wl), a colorimetric method for the determination of the number of viable cells in cytotoxicity experiments. In the absence of estrogen/SERMs pre-incubation, βA (10 and 100 nM) significantly and dose-depeπdently reduced cell viability (Fig. 1A- C). Pre-iπcubation with 17βE2, tamoxifen or raloxifene (100 pM-100 nM) did not directly affect FNC-B4 cell viability. Following estrogen exposure, neither 10 nM nor 100 nM βA significantly altered cell viability (Fig. 1A). The same results were obtained using tamoxifen (100 pM-100 nM) (Fig. 1B) or raloxifene (100 pM and 1 nM) (Fig. 1 C). Conversely, 10 and 100 nM raloxifene did not protect against βA- induced toxicity (Fig. 1C). It is worth noting that the peak plasma concentration of tamoxifen or raloxifene, following chronic administration at the therapeutic daily dosage of 20 mg tamoxifen (breast cancer) and 60 mg raloxifene (osteoporosis) orally, is 213 nM and 2.66 nM, respectively (according to the reports of U.S. Food and Drug Administration, Center for Drug Evaluation and Research). Therefore, very low doses of tamoxifen/raloxifene should warrant a neuroprotective effect in the clinical setting for the prevention/treatment of neurodegenerative diseases. The levels of expression of seladin-1 gene in neuronal cells were determined by a quantitative real-time RT-PCR technique for the measurement of mRNA that was optimized for the first time in our laboratory and is published in Luciani P. et al. - J. Of Endocrinology and Metabolism - Vol. 89 p.1332-39 (2004). The procedure is briefly summarized hereafter: total RNA was extracted from control cells as well as from cells treated with different concentrations of 17βE2, raloxifene or tamoxifen for 72 hours. Primers and probe were selected by the proprietary software "Primer express" (Applied Biosystems Inc., Foster City, CA). The sequences of seladin-1 primers, spanning 88 bp were: 5'- ATCGCAGCTTTGTGCGATG-3' (sense, exon 4, 5'-end at position 686); 5'- CACCAGGAAACCCAGCGT-3' (antisense, exon 5, 5"-end at position 774). The sequence of the probe, labeled with FAM, which hybridized to the exon 4-5 junction region, was: δ'-TCCGTCCGAAAACTCAGACCTGTTCTATGC-S' (5'-end at position 708). A calibration curve was generated using serial dilutions of a single-stranded sense oligodeoxynucleotide spanning the sequence included between the primers. The results were expressed as fg seladin-1 mRNA/μg total RNA. The results are shown in Fig. 2 and indicate that the amount of seladin-1 mRNA in untreated cells was 112 fg/μg total RNA. 17βE2 was able to significantly increase seladin-1 expression at all the concentrations that were used. In addition, both raloxifene and tamoxifen significantly increased the level of seladiπ-1 expression (152 + 5% and 176 + 6%, mean + SE, respectively, vs control, considered as 100%, at 1 nM, as shown in Fig. 2), with the exception of raloxifene at high concentrations. In fact, remarkably the exposure of neuronal cells to 10 and 100 nM raloxifene, which did not exert neuroprotective effects (see Fig. 1C), markedly decreased the amount of seladin-1 mRNA (36.7 + 0.5% and 44.1 + 0.7%, mean + SE, respectively) compared to untreated cells. Our results strongly indicate that estrogen and SERMs exert a neuroprotective effect in neuronal cells, yet the effect of raloxifene is limited to low concentrations (100 pM and 1 nM) of the drug. We demonstrated for the first time that estrogen and SERMs stimulate the expression of the neuroprotective factor seladin-1, suggesting that this protein is a mediator of the effect of these drugs on neuronal cells. Remarkably, this hypothesis was supported by the observation that high concentrations (10 and 100 nM) of raloxifene reduced the amount of seladin-1 mRNA, paralleling the lack of a neuroprotective effect at the same concentrations. Finally, the observation that SERMs do not increase cell proliferation further indicates that the neuroprotective effects of theses drugs are mediated by events other than cell replication, i.e. by modulating the expression of genes such as seladin-1. The formulations to be used according to the present invention are similar to those employed and commercialised usually as active compounds. However, in view of the above mentioned data, in order to obtain the expected therapeutical effect in terms of neuroprotection the active principle could be used in lower quantities. For instance, formulation containing as active compound raloxifene can be used as pills for oral administration containing 5-10 mg of active principle.