Ceramide and Related-Sphingolipid Levels Are Not Altered in Disease-Associated Brain Regions of APPSL and APPSL/PS1M146L Mouse Models of Alzheimer's Disease: Relationship with the Lack of Neurodegeneration?

There is evidence linking sphingolipid abnormalities, APP processing, and neuronal death in Alzheimer's disease (AD). We previously reported a strong elevation of ceramide levels in the brain of the APPSL/PS1Ki mouse model of AD, preceding the neuronal death. To extend these findings, we analyzed ceramide and related-sphingolipid contents in brain from two other mouse models (i.e., APPSL and APPSL/PS1M146L) in which the time-course of pathology is closer to that seen in most currently available models. Conversely to our previous work, ceramides did not accumulate in disease-associated brain regions (cortex and hippocampus) from both models. However, the APPSL/PS1Ki model is unique for its drastic neuronal loss coinciding with strong accumulation of neurotoxic Aβ isoforms, not observed in other animal models of AD. Since there are neither neuronal loss nor toxic Aβ species accumulation in APPSL mice, we hypothesized that it might explain the lack of ceramide accumulation, at least in this model.


Introduction
Alzheimer's disease (AD) is the most common form of dementia in adults. Pathologically, the hallmarks of AD are amyloid plaques and neurofibrillary tangles, associated with widespread neuronal loss. Its fundamental causes and the pathological cascades leading to symptoms, however, remain poorly understood. Lipids and lipid peroxidation products have important roles in central nervous system homeostasis. Extensive evidence supports an important role of cholesterol in the development and possibly progression of AD [1][2][3]. The role of other lipids, such as ceramides and relatedsphingolipids, (sphingomyelins, sulfatides, and glycosphingolipids) is also emerging. Ceramides are the core constituent of most sphingolipids. They can be produced by hydrolysis of sphingomyelin (SM) via sphingomyelinases (SMases) or synthesized de novo from fatty acyl CoA and sphingosine. Conversely, in the Golgi, ceramides may be transformed into SM by the addition of phosphorylcholine. Additionally, glycosyltransferases can attach sugar to ceramide, turning it into glucosylceramide or galactosylceramide (Galcer), a key step in the generation of complex glycosphingolipids [4,5].
Ceramides are important second messenger molecules that regulate diverse cellular processes including cell growth, differentiation, and apoptosis. Ceramide levels also increase in response to aging and various age-related stress factors and are directly involved in apoptotic signaling in various cell types, including neurons [6][7][8].
There is evidence linking sphingolipid abnormalities, APP processing, and neuronal death in AD. In vitro, it has been shown that β-amyloid added to cultured neurons [9,10] or oligodendrocytes [11] increase SMase activity, leading to an increase in ceramide levels [12]. Additional reports have found that ceramide levels increase β-amyloid synthesis [12,13] and favor gamma secretase processing of APP [14][15][16] so that inhibition of ceramide synthesis 2 International Journal of Alzheimer's Disease confers protection against β-amyloid [9]. Thus, it has been suggested that ceramides and β-amyloid may synergize to induce neuronal death.
Several studies have examined the sphingolipid alterations in human AD brain. For example, the total phospholipid and sulfatide content in AD was decreased as compared to normal [17][18][19], while the ceramide and galactosylceramide levels were elevated [9,18]. Enhanced ceramide levels have been reported in the cerebrospinal fluid (CSF) of patients with AD [20] and changes in the activity of several key enzymes of ceramide metabolism, in gene expression of pathways associated to sphingolipid metabolism have been found in brains of AD patients [21,22].
During the last years, numerous mouse transgenic lines have been created and have been screened for various aspects of AD pathology [23,24]. Unfortunately, very little work has been done on determining if sphingolipid content is likewise perturbed in these rodent models of AD. In one study, longchain ceramides were shown to be elevated in presenilin 1 (PS1M146V) mouse brain and to induce apoptosis in PS1 astrocytes [25]. In a second study, sulfatides, a class of sulfated galactocerebrosides, were found to be decreased in brain tissues from two APP transgenic mice (i.e., APPV717F and APPsw), whereas no significant changes in the content of other sphingolipid classes including SM, Galcer, and ceramides were noted [26]. By contrast, using the new mouse mutant APP SL /PS1Knock-in line, we have recently found an early and significant increase of ceramides in the cortex of these mice, without significant changes in other sphingolipid levels [27]. However, the APP SL /PS1Ki model is unique for its drastic neuronal loss, not observed in other animal models of AD [28]. The discrepancy in the few data available and the lack of knowledge of sphingolipid levels in the brain of other rodent models of AD prompted us to investigate whether the sphingolipid composition is altered in the brain of two other mouse models of AD: single APP SL and double APP SL /PS1 M146L transgenics. Concentrations of ceramides, SM, Galcer, and sulfatides were determined in three brain regions: the cortex and the hippocampus, the two brain regions typically associated with the disease, and the cerebellum, a nonvulnerable region with no Aβ plaques. For all analyses, age-matched PS1 M146L (amyloid free) mice as well as nontransgenic wild-type mice (WT) were used.

Transgenic Mice.
Generation and detailed neuropathological analyses of the APP SL and the APP SL /PS1 M146L transgenic mice have been described earlier [29,30]. In brief, these mice express human APP751 with Swedish and London mutations (Thy1 promoter) either alone or in combination with human presenilin-1 (M146L, HMG-CoA promoter). In this study, 12-month-old (n = 5) APP SL /PS1 M146L double transgenic, 12-month-old (n = 5) PS1 M146L single-transgenic littermates, and 12-month-old (n = 6) nontransgenic mice as well as 24 month-old (n = 5) APP SL single transgenic and 24-months (n = 5) nontransgenic littermates were used (generous gift of Sanofi-Aventis, Vitry sur Seine, France). APP SL mice were analyzed at 24 months of age and APP SL /PS1 M146L mice were assessed at 12 months of age because they revealed comparable levels of amyloid plaques in the brain at these respective ages. The genetic background of the mice was mixed CBA/C57BL/6. All the mice used in this study were female, because a gender effect with female mice displaying more severe pathology than male has been mentioned in several studies [27,31,32].
All experiments were performed in compliance and following approval of the Sanofi-Aventis Animal Care and Use Committee and in accordance with standards of the guide for the care and use of laboratory animals (CNRS ILAR) and with respect to French and European Community rules.

Brain Tissue Preparation.
The mice were anesthetized with pentobarbital (40 mg/kg, IP) and sacrificed. Brains were removed immediately, and cortex, hippocampus, and cerebellum were dissected. These cerebral regions were homogenized by 10 up-and-down strokes of a prechilled Teflonglass homogenizer in 20 volumes of buffer (25 mM Tris-HCl, 150 mM NaCl, 1 mM EDTA, pH 7.4) and supplemented with 50 mM sodium fluoride, 1 mM phenylmethylsulfonyl fluoride, protease, and phosphatase inhibitor cocktails (50 μL/g tissue and 10 μL/mL lysis buffer, resp.). The homogenates were centrifuged at 15,000 g for 15 min at 4 • C. The resulting pellet was resuspended in 3 volumes of ice-cold water, altogether 1.5 mL, homogenized at 4 • C (10-15 strokes) and then sonicated for 20 s using a sonifier (Branson Ultrasonics, sonifier 450, Danbury, CT). Samples from each mouse brain were analyzed separately and in duplicate.

Lipid Extraction from Tissues and Thin-Layer
Chromatography. Total lipids from brain homogenates were prepared according to previously described procedures [27,33]. Briefly, the homogenates were added to 8 vol. of rapidly stirring methanol, then 4 vol. of chloroform were added. The mixture was stirred overnight, and then centrifuged at 1,000 g for 10 min. The pellet was re-extracted with chloroform-methanol-water (4 : 8 : 3, v/v/v) and the two supernatants combined, evaporated to dryness. Partitioning was carried out using diisopropyl ether/1-butanol/50 mM aqueous NaCl (6 : 4 : 5, v/v/v) according to the method of Ladish and Gillard [34]. The total upper phase was evaporated to dryness and taken up in chloroform. The lipids were fractionated using solid-phase extraction on International Journal of Alzheimer's Disease 3 100 mg LC-NH 2 cartridges (Supelco, L'Isle d'Abeau, France) as previously described [27]. The eluted fractions containing, respectively, free ceramides, galactosylceramides (Galcer), alkali-stable phospholipids (SM), and sulfatides were applied to HPTLC plates with a Linomat 5 (Camag, Switzerland). Prior to SM analysis, the phospholipid-containing fraction was subjected to mild alkaline methanolysis (treatment with chloroform: 0.6 N NaOH in methanol 1 : 1 (v/v) at room temperature for 1 h) to remove glycerophospholipids. To quantify each lipid species (ceramides, SM, Galcer, and sulfatides), calibration curves were obtained by running in parallel known amounts of purified lipid standards. For free ceramides, 15-20 mg of brain tissue (wet weight) was spotted per lane. The plates were developed with chloroformmethanol 50 : 3 (v/v) and visualized by charring for 10 min at 180 • C with 3% cupric acetate in 8% phosphoric acid solution. For SM analysis, 6-9 mg of tissue was spotted per lane. The plates were developed with chloroform-methanolwater 60 : 35 : 8 (v/v/v). SM was visualized with sulfuric acid-CuSO 4 -ammonium molybdate spray reagent followed by heating at 110 • C for 15 min. Galcer and sulfatides (about 0.4 mg and 2 mg of tissue per lane, resp.) were developed in chloroform-methanol-water 65 : 25 : 4 (v/v/v), sprayed with orcinol-H 2 SO 4 reagent, and then heated at 150 • C for 2 min.
Each sphingolipid species was quantified by scanning densitometry of the plates at 396 nm for ceramides, 540 nm for Galcer and sulfatides, and 750 nm for SM with the Camag TLC Scanner 3/WinCats software system.

Statistical Analysis.
The results were presented as mean ± S.D. values. Owing to the small number of animals per group, statistical analysis of the data was performed using a nonparametric Kruskall-Wallis test followed by a posthoc test of Dunn for multiple comparisons. For the comparison of two means, the Mann-Whitney test was used. All calculations were performed using GraphPad Prism software 3.02 (GraphPad Software, Inc.). P values less than .05 were considered as significant.

Results
In single APP SL mice (Figure 1(a)), a moderate but not significant elevation of NFA-ceramides was seen in the cortex (+22%), with no changes of the 2-HFA-ceramide level. Conversely, in the hippocampus (Figure 1(b)), the NFAceramide level did not differ between WT and APP SL mice, whereas a tendency towards an increase of 2-HFA-ceramides was noted (+19%), although it was not significant.
Surprisingly, as shown in Figure 1(c), the level of 2-HFAceramides in the cerebellum of APP SL mice was significantly increased in comparison to the counterpart of their WT littermates (+50%, P < .05). Conversely, the NFA-ceramides showed a slight but not significant decrease compared to WT control values. However, no difference in the total ceramide content was noticed in the cerebellum of both WT (51.35 ± 1.45 nmol/mg tissue) and APP SL mice (50.28 ± 3.31 nmol/mg tissue).
Because the double-transgenic mouse model APP SL / PS1 M146L develops neuropathological features of AD earlier than the single APP SL mice [29], the sphingolipid analysis was performed in the brain regions of this model in younger animals (12 months of age). As shown in Figures 2(a) and 2(b), concentrations of NFA-ceramides as well as those of 2-HFA-ceramides did not differ between wild-type, PS1 M146L , and APP SL /PS1 M146L mice in the two disease-associated brain regions (cortex and hippocampus). In contrast to the changes of ceramide content in cerebellum of APP SL mice, the NFA-ceramide as well as the 2-HFA-ceramide levels were unchanged in this brain region in APP SL /PS1 M146L mice relative to their WT controls and PS1 littermates ( Figure 2(c)). However, NFA-ceramide content showed a tendency towards a decrease, while 2-HFA ceramides tended to slightly increase in the cerebellum of the APP SL /PS1 M146L mice.
It should be noted that, although cortex and hippocampus of WT mice displayed almost identical ceramide content (Figures 1(a), 1(b), 2(a), and 2(b)), a relatively lower NFAceramide content was manifest in the cerebellum (Figures 1(c) and 2(c)). In normal human brain, the ceramide levels were also reported much higher in the cortex versus the cerebellum [18].
Typical ceramide profiles from either cortex, hippocampus, or cerebellum of the different transgenic mice and wild-type controls were shown in Figure 3. Since 2-HFAceramides are present in extremely low levels leading to a weak staining on the HPTLC plate, densitometric scanning of the plate was shown to visualize the peak corresponding to the 2-HFA-ceramide species (Figure 3).
To test whether other related-sphingolipids could be altered in the brain of transgenic mice, the content of SM, Galcer, and sulfatides in lipid extracts of the three examined brain regions was assessed by HPTLC analysis. Figures 1(d)-1(f) show that sphingolipids (SM, Galcer, and sulfatides) did not display significant changes in both cortex, hippocampus, and cerebellum of APP SL mice compared to the WT littermates. Similarly, the levels of SM, Galcer, and sulfatides did not differ among nontransgenic, PS1 M146L and APP SL /PS1 M146L mice in all brain region examined (Figures 2(d)-2(f)). It should be noted that there are differences of SM, Galcer, and sulfatide concentrations between cerebral tissues (cortex and hippocampus) and cerebellar tissues. In both models, cerebellum displayed higher levels of Galcer and sulfatides than cerebral tissues. This is in accordance with Cheng et al. [26] who also reported a ∼2-fold higher level of sulfatides in the cerebellum compared to the cortex, in two other transgenic mouse models of AD. In contrast, SM levels were almost identical in hippocampus and cerebellum, but higher than those of cerebral cortex (Figures 1(d)-1(f) and 2(d)-2(f)). Examples of HPTLC profiles of sulfatides, Galcer, and SM and from either cortex, hippocampus, or cerebellum were represented in Figures 4(a), 4(b), and 4(c), respectively. It should be mentioned that the HPTLC profiles of each sphingolipid class were similar in all examined brain regions from both mouse models.

Discussion
Using the same methodology as in our previous work [27], we herein analysed the sphingolipid composition of two additional models (i.e., single APP SL and double APP SL /PS1 M146L mice) in which the time-course of pathology is much closer to that seen in the majority of currently available models. The main results of this study are (1) a moderate but not significant increase of NFA-ceramides and 2-HFA-ceramides in the cortex and the hippocampus respectively, of the APP SL mice, (2) unaltered ceramide levels in the two disease-associated brain regions from APP SL /PS1 M146L mice, (3) unexpected alterations of the ceramide profile in the cerebellum of APP SL mice, a region with no Aβ pathology, and (4) no significant changes in the other related-sphingolipids in all brain structures examined of both transgenic models. Based on our results and those from the literature, we will first discuss the possibility of a relationship between neurodegeneration, toxic Aβ accumulation, and ceramide content. For the second time, why an amyloid-free brain region such as cerebellum showed ceramide alterations is discussed.
Ceramides were shown to accumulate in AD human brain regions and their levels vary by disease severity suggesting that they could be indicators of AD progression [9,18,35]. Similarly to what happens in human AD, we previously found that ceramides increase very early in the cortex of the APP SL /PS1Ki mouse model, preceding the neuronal loss [27]. By contrast, our present results reveal that in single APP SL mice, ceramide levels were not significantly altered in disease-associated brain regions (e.g., hippocampus or cortex). This is consistent with the findings of Cheng et al. [26] who reported no changes in ceramide content in any brain region from APPsw and APP V717F transgenic mice. An important difference between these single APP mouse lines and the APP SL /PS1Ki model is that the latter develops an early and massive neuronal loss which correlates with strong accumulation of intracellular Aβ42, Aβ aggregates, and Aβ42 oligomers [28,36]. Although intraneuronal Aβ accumulation has also been documented in various APP models [29,37,38], a striking difference between the models used in the present study and the APP SL /PS1Ki mice is the nature of the Aβ peptides which accumulate. Indeed, in APP SL /PS1Ki mouse brain, extremely high levels of Ntruncated Aβx-42 variants and abundant oligomers were detected, which closely resembles that found in AD brain. By contrast, in APP SL mouse brain, with the same total Aβ levels as in APP SL /PS1Ki mice, only very limited levels of Aβ42 N-terminal truncated isovariants were detected [28]. In the APP SL /PS1Ki mice, Aβx-42 is the major form accumulated with a ratio Aβx-42/total Aβ close to 1. In comparison, this ratio is approximately of only 0.2-0.3 in the APP SL mice, and ∼0. 3-0.4 in the double transgenic APP SL /PS1 M146L mice, similar to the range of values reported for a large number of other APP-based transgenics [28,29].
N-truncated Aβ peptides are known to aggregate more readily and are considered to be very toxic species. Thus they might play a key role in the neurotoxicity observed in the APP SL /PS1Ki model. In particular, the pyroglutamate modified N-terminal truncated form of Aβ at position 3 (Aβ 3(pE)-42 ), which represents a major species in the brain of AD patients [39], was recently shown to induce a severe neuron loss in the TBA2 mouse model, a new model expressing only N-truncated Aβ 3(pE)-42 in neurons [40]. Interestingly, there is also a coincidence of considerable amounts of Aβ 3(pE)-42 and massive neuron loss in the APP SL /PS1Ki mouse model [41]. On the basis of these findings, it is attractive to speculate that in APP SL /PS1Ki mice, the strong accumulation of intracellular toxic forms of Aβ induces early elevation of ceramides. Extensive increase of ceramides could therefore activate proapoptotic pathways, leading to neuronal death. Conversely, other APP transgenic mouse models including the APP SL mice have been reported to show no [42] or very low Aβ 3(pE)-42 levels [43]. This is due to the lack of posttranslational modifications such as Nterminal degradation and pyroglutamyl formation in these mice. Because APP SL mice display neither neuronal loss nor accumulation of highly toxic Aβ 42 isoforms, this may at least in part explain why no significant accumulation of ceramides occurred in the disease-associated brain regions of these mice.
Thus, despite numerous neuropathological, biochemical, and even behavioral changes representative of AD developed by these APP mice [23, 28-30, 33, 44], they may not constitute a relevant model to further explore the role of ceramides in AD pathology. However, answering the question of the relationship between neurodegeneration, toxic Aβ accumulation, and ceramide elevation could be facilitated using restricted models either lacking (i.e., single APP mice) or mimicking only some specific AD-related neuropathological alterations (i.e., TBA2 mice mentioned above). Indeed, owing to the simultaneous occurrence of numerous pathological features of AD, the connection between them is often difficult to unravel.
We next determined the ceramide content in the double transgenic mouse model APP SL /PS1 M146L , but in younger animals because the APP SL /PS1 M146L mice develop neuropathological features of AD earlier than single APP SL mice [29]. At 12 months of age, these double transgenics revealed comparable levels of amyloid deposits than 24-month-old APP SL mice. Additionally, they displayed similar ganglioside alterations [33]. Our results showed that at 12 months of age, ceramide levels were unaltered in both cortex and hippocampus of APP SL /PS1 M146L mice in comparison to agematched PS1 M146L mice and wild-type controls. It should be noted that there is no neuronal loss in these mice as old as 17 months, but unfortunately, older APP SL /PS1 M146L mice were not available for this study. However, in our previous work, we demonstrated that elevation of ceramides occurred very early (3 months of age) in the cortex of the APP SL /PS1Ki model, preceding by far the neuronal loss detectable at 6 months of age. Why the APP SL /PS1Ki mice showed an increase of ceramides several months before the appearance of neuronal death while the APP SL /PS1 M146L mice did not is currently unknown. One possible explanation is that the neuronal loss reported in the APP SL /PS1 M146L mice is moderate and more restricted than in the APP SL /PS1Ki model. Indeed, the loss of neurons in the former involves only the hippocampal pyramidal cell layer (loss of ∼30% in 17-month-old animals). This may in some way account for the lack of ceramide accumulation in the cortex of these mice. By comparison, in the APP SL /PS1Ki model, the cell loss is greater (∼50% at 10 months of age) and has been shown to extend to other brain areas such as frontal cortex [36] and cholinergic system [45]. Moreover, as discussed above, accumulation of N-truncated Aβ peptides should be lesser in APP SL /PS1 M146L mice, since the ratio Aβx-42/total Aβ is of 0.3 only, versus 1 in the APP SL /PS1Ki mice. In this context, it would seem likely that the level of highly toxic Aβ 3(pE)-42 form, which is thought to be involved in neuronal death, is reduced in the APP SL /PS1 M146L model. It is also possible that, at 12 months of age, it is too early to visualize an elevation of ceramides, owing to the slowest progression of AD lesions in these mice than in the APP SL /PS1Ki model. Since we did not have older APP SL /PS1 M146L mice, we should therefore be cautious to interpret the results of ceramide composition in these mice, because we cannot exclude the possibility that ceramides increase at a later age. Further investigations on this topic are warranted.
The most unexpected finding of the present study was alteration of the ceramide composition in the cerebellum of APP SL mice, a brain region lacking Aβ deposits and regarded as nonvulnerable to the disease. Intriguingly, we noted a significant increase of 2-HFA-ceramides (+50%) which was concomitant with a slight decrease in NFAceramides. However, considering the total ceramide content, it was almost similar between wild-type and APP SL mice. Previously, we found a more dramatic 161% increase of 2-HFA-ceramides in the cortex of APP SL /PS1Ki mice but contrary to the present results, it was accompanied by an elevation of NFA-ceramides. Hydroxy FA-ceramides are relatively minor species of membrane lipids. As evident from the literature, the current knowledge about the metabolism and physiological function of 2-HFA-ceramides is very limited. In particular in AD studies, no attention was paid to 2-HFA-ceramides, rendering it very difficult to draw conclusions about the role of these ceramide species in AD physiopathology. Nevertheless, a couple of interesting facts suggest that these HFA species may participate to AD pathology: (i) it was recently found that Aβ selectively bound to sphingolipids that contained a 2-OH group on the ceramide backbone and did not effectively interact with sphingolipids that contained a nonhydroxylated fatty acid, favoring a conformational shift that disrupts membrane stability and promotes peptidepeptide interactions and oligomer formation [46]; (ii) the enzyme UDP-galactose: ceramide galactosyltransferase (CGT), which transfers galactose to both NFA-and 2-HFAceramides, was found to be upregulated in human AD brain [21]. Interestingly, overexpression of CGT in transgenic mice led to a reversal NFA : HFA-Galcer ratio which resulted in both decrease in HFA-Galcer and increase in NFA-Galcer levels [47]. Reducing the HFA-Galcer level would lead to an accumulation of their immediate precursors, 2-HFA-ceramides; (iii) there is some evidence for enhanced apoptosis-inducing activity of 2-HFA-ceramides compared to NFA-ceramides [48], and this effect seems to be cell type specific. In this sight, mouse mutants with defective saposin D dramatically accumulate HFA-ceramides in cerebellum, resulting in a selective loss of cerebellar Purkinje cells [49]; (iv) we reported a gender-specific expression of HFA-versus NFA-ceramides in the APP SL /PS1Ki mouse model of AD, and this biochemical feature could be related to the increase propensity of females to develop earlier neuronal loss [27]. Although the degree of 2-HFA-ceramide accumulation in the cerebellum of APP SL mice is much lesser than that seen in the cortex of the APP SL /PS1Ki mice, the reasons for these biochemical changes in this amyloid-free brain area are currently unknown. Possibly, it might reflect alterations of ceramide metabolism, since there is evidence that other metabolic pathways are perturbed in the cerebellum of these mice [50]. Hydroxy FA sphingolipids are synthesized by the same set of enzymes as nonhydroxy sphingolipids, except for fatty acid 2-hydroxylase (FA2H). The expression of FA2H is highly variable among cell types and is inducible by certain stimuli [48]. One may speculate that, upon unknown signal, 2-HFA-ceramides may be preferentially synthesized instead of NFA-species. Abnormal degradation of HFA sphingolipids may be also possible. It has been shown that galactosylceramidase, which forms 2-HFA-ceramides from Galcer is up-regulated, whereas acid ceramidase which hydrolyzed 2-HFA-ceramides into HFA and sphingosine is down-regulated in human AD brain [21]. Thus, with combinatorial up-and down-regulation of enzymes involved in sphingolipid metabolism, the cell could modify the levels of 2-HFA-ceramides in response to the changing cellular environment. However, this is highly speculative and further investigation is warranted to determine whether these factors or other unknown factors contribute to such changes of the ceramide profile in the APP SL cerebellum. It should be noted that also intriguing is the substantial elevation of ceramide reported by Han et al. [18] in the cerebellum of AD patients; this point also remains to be clarified. In this study, we also examined the content of other ceramide-related sphingolipid classes including SM, Galcer, and sulfatides. Similarly to what we observed in the APP SL /PS1Ki model, we did not found any changes of SM and galactolipid levels in all brain regions from the two transgenic lines investigated. Similar findings were seen in APPsw and APPV717F transgenic mice, respectively, except for sulfatides [26]. Indeed, by contrast to our results, a loss of sulfatide content was observed in multiple brain regions of these animals. The reasons for such discrepancies are still unclear, but it may be ascribed to the different genetic background of mouse lines and/or the genetic constructs based on different APP mutation and different promoters. Supporting this, it has been shown for example, that APP SL transgenic and wildtype mice on C57BL/6 background have lower basal cholesterol levels than the Ki mouse lines which are on C57BL/6 50%-CBA 25%-129SV 25% background [44]. Another example is the difference of lipid composition reported by Sawamura and coworkers [15] between mouse lines with C57BL/6J and FVB/N backgrounds, respectively. Moreover, these authors found that PS2 transgenic mice with C57BL/6J background displayed significant phospholipid alterations, particularly of SM, as compared to their wildtype controls, while PS2 transgenic mice with FVB/N background did not. These few examples point out the difficulties to compare the results from various mouse lines and reinforce the idea that additional studies in this field are required.

Conclusion
In summary, this study extends previous observations on sphingolipid alterations in animal models of AD. Despite several limitations, in particular the lack of old double transgenics, the present results demonstrated that, in the absence of neurodegeneration, no elevation of ceramides occurred in disease-affected brain regions from single APP SL mice, thus corroborating recent findings in two other single APP mice [26]. Moreover, since both neuronal loss and accumulation of toxic N-truncated Aβ peptides are lacking in the APP SL model, this might suggest that Aβ-induced ceramide production is an important event leading to neuronal death. In the future, to support this hypothesis, it will be interesting to analyse the sphingolipid composition of the TBA2 mice, the new model expressing only N-truncated Aβ 3(pE)-42 and which develops a severe neuronal loss [40], to evaluate whether or not ceramides, especially the 2-OH species, accumulate in the brain of these mice.
Accumulating and crossing the information obtained from various animal models will help to better understand the exact mechanism by which these sphingolipids contribute to AD pathogenesis.