Neuronal sphingosine kinase 2 subcellular localization is altered in Alzheimer’s disease brain

E. coli BL21 strains were transformed by one of the pJ414 constructs - PET21 vectors expressing either SphK1 (403AA) or SphK2 (674AA) with an HIS tag. Forty ml of an overnight pre-culture of transformants grown at 37 °C in DYT medium supplemented with kanamycin (100 μg/ml) was used to inoculate 200 ml of DYT-KAN media. After 2 h of growth at 37 °C, SphK proteins expression was induced by 5 mM of Isopropyl β-d-1-thiogalactopyranoside (IPTG). Aliquots of the cultures were made 5 h after induction. Protein expression was confirmed by western blot analysis with mouse anti-histidine (Cusabio, Ref. CSB-MA000011M0m). The specificity of the anti-SphK2CT rabbit polyclonal antibody (Sigma, Ref. SAB4502433) was confirmed by western blot.

Paraffin embedded human brain tissues were provided by certified French biological resource centers from Lille (Neurobank Lille DC-2008-642) and Toulouse (Brain bank AC-2009-973) for immunohistochemistry and immunofluorescence studies. For western blot studies, human brain frozen tissues were provided by the national brain bank GIE Neuro-CEB (AC-2007-5) and the biological resource center of CHU Toulouse (Brain bank AC-2009-973). The utilization of postmortem material was approved by the corresponding biobank ethic committees. All cases were scored according to current criteria of NIA-Alzheimer’s Association [39]. The assessment included Braak and Thal neuropathology stages [5, 50].

Post-mortem tissues from 25 AD patients were included in the immunohistochemical study. Characteristics of patients (age, gender, post mortem interval, Braak and Thal stages) are summarized in Table 1. Hemi-brains were fixed with formalin (4% in PBS) during approximately 1 month. Paraffin-embedded, formalin-fixed sections (4 μm) from frontal cortex and hippocampal area (entorhinal cortex and hippocampus) were deparaffinized in xylene and rehydrated in ethanol. Antigen retrieval was performed by immersing sections in boiling EDTA buffer (pH 9.0). Endogenous peroxidase and alkaline phosphatase were blocked by incubation of the sections for 5 min in Dual Endogenous Enzyme Block (Dako, Denmark). Sections were initially incubated with a primary antibody directed against β-amyloid (Dako, mouse clone 6 F/3D, Ref. M0872, 1:100) during 3 h at room temperature (RT). Sections were washed twice during 7 min in Tris-buffered NaCl solution with Tween 20 (pH 7.6, Dako). Immunostaining was revealed using BrightVision poly HRP-Anti-Mouse IgG (Immunologic, Netherlands) during 30 min at RT and treated with diaminobenzidine/hydrogen peroxide (DAB, Dako) for 5 min. After this first step, sections were washed for 10 min with Tris buffer saline (pH 7.6, Dako) before incubation with a primary rabbit polyclonal antibody directed against SphK2 (Sigma, Ref. SAB4502433, 1:50) overnight at 4 °C. The sections were washed twice during 7 min in Tris buffer saline (pH 7.6, Dako). Immunostaining was exposed using BrightVision poly AP-anti-Rabbit IgG (Immunologic) during 45 min at RT and treated with Liquid Fast Red (Abcam) for 30 min. Sections were counterstained with hematoxylin then mounted in Faramount Aqueous Mounting Medium (Dako). Once mounted, slides were scanned with a digital scanner NanoZoomer (Hamamatsu, Ref. [2].0-RS: C10730–13) to obtain high resolution virtual slides. Digitalized slides were analyzed with an imaging analysis system (ImageJ®). Morphometric investigations were carried out by a semi-automatic procedure on ImageJ. The number of neurons expressing SphK2, the percentage of Aβ stained surface and the number of amyloid deposits were quantified among the different cortical layers of the frontal and entorhinal cortex as well as in the CA1. Columns constituted of contiguous microscopic fields, from the pial surface to the white matter were drawn on each slide. As the fields were examined at a magnification of × 400, each field was 300 μM × 150 μM in size. As the thickness of the cortex appeared to be variable between the different sections, after the counting step, the columns were standardized to 10 fields [8, 18]. For cortical areas, field 1 corresponded to the cortex immediately under the pial surface and field 10 reached the white matter (Additional file 1). In each field, the number of neurons expressing a negative and a positive SphK2 stain were counted. Moreover, the number of Aβ focal deposits and the percentage of surface labeled by amyloid per field were determined and were reported on a database.

Fresh samples from frontal cortex and hippocampal area of 10 AD cases and 6 age-matched controls were used. No significant differences in sex and post-mortem interval were noted between AD patients and the control group. There was a difference in the age medians. As expected, AD patients had significantly higher Braak and Thal stages than the control group (Table 2). Samples of hippocampal area contained both entorhinal cortex and Ammon’s horn (including CA1). Moreover randomly selected samples of cerebellum from of AD (n = 5) and control (n = 3) were included as the cerebellar tissue exhibits a weak density of Aβ deposits even in the late phase of disease [50]. Frozen tissue samples were processed with Precellys® homogenizer and CK14 lysing kit (Bertin Instruments, Paris, France) containing lysis buffer (50 mM Tris-HCl pH 8.0, 5% SDS, 0.25 M saccharose, 200 mM NaCl and EDTA-free protease inhibitor cocktail). Samples were sonicated at 4 °C then centrifuged at 13000 g for 10 min.

The same cases except one control (AD = 10, Control = 5) were included in this experiment. Frozen tissue samples were lysed as aforementioned in HEPES buffer (pH 7.4) (20 mM HEPES, 10 mM KCl, 2 mM MgCl₂, 1 mM EDTA, 0.25 M saccharose, and EDTA-free protease inhibitor cocktail). Nuclei-enriched fractions and cytoplasmic-enriched fractions were prepared by sequential centrifugation. Nuclei were pelleted by centrifugation (500 g for 10 min at 4 °C) then sonicated at 4 °C and centrifuged at 9000 g for 30 min, in Tris buffer (pH 7.4) (50 mM Tris-HCl, 250 mM NaCl, 1 mM EDTA, 0.1% Triton and EDTA-free protease inhibitor cocktail). Supernatants were removed, and the membrane and cytoplasmic fractions were separated by centrifugation at 16,000 g (30 min at 4 °C). Isolation of nuclear and cytosolic fractions was confirmed by western blot with primary antibodies directed against rabbit polyclonal PARP (GeneTex, Ref. GTX100573, 1:1000) and rabbit polyclonal LDH (GeneTex, Ref. GTX114525, 1:10,000).

Total protein concentration was assessed on the supernatant with the BCA Protein Assay (Interchim). Samples were heated at 95 °C for 5 min with 5% β-mercapto-ethanol, 0.05% bromophenol blue. 150 μg of total proteins were loaded by lane in precast gels with Stain-Free technology (4–15% Mini-PROTEAN® TGX Stain-Free™, Ref. 456–8083). The proteins were separated by electrophoresis at 90 V in a MiniProtean Tetra System (Bio-Rad Laboratories, Irvine, CA). After migration and semi-dry transfer (Transblot Turbo, Bio-Rad), nitrocellulose membranes were blocked with 5% skimmed milk, and washed 3 times with Tris-buffered saline buffer containing 0.05% Tween-20 (TBST). Blots were probed with a primary antibody directed against rabbit polyclonal SphK2 (Sigma, Ref. SAB4502433, 1:500) and β-amyloid (Sigma, Ref. A8354, 1:1000). After an overnight incubation at 4 °C, the membranes were washed with TBST, labeled with a peroxidase-conjugated anti-rabbit (dilution 1:3000) secondary antibody (Bio-Rad) and revealed by chemiluminescence (Clarity kit, Bio-Rad). Bands were quantified by densitometry with ImageLab (Bio-Rad) software. The density of total protein, detected by UV light, was used to normalize the signals [20, 23].

Post-mortem tissues from 9 AD and 6 controls were included in the immunofluorescence study (Table 3). There were no significant differences in age, sex and post mortem interval between the groups used for the immunofluorescence study. Most of the AD subjects were staged Braak V-VI and Thal 4 to 5. Control subjects included cases with respectively low Braak (I-II) or Thal (1–2) stages. Immunofluorescent staining of SphK2 and MAP2 was performed on paraffin-embedded, formalin-fixed human brain sections. The human brain sections were deparaffinized in xylene and rehydrated in ethanol. Antigen retrieval was performed by immersing sections in boiling EDTA buffer (pH 9.0). Human brain sections were washed in PBS, permeabilized with PBS-Triton 0.1% for 10 min. To avoid the autofluorescence, sections were incubated in True black 1X (Ozyme) for 1 min. Nonspecific antibody reactions were blocked by incubation in solution containing 5% BSA for 1 h. Sections were incubated with mouse monoclonal primary anti-MAP2 (Life technologies, Ref. MA512826, 1:100) and anti-SphK2 (Sigma, Ref. SAB4502433, 1:50) antibodies. Secondary antibodies included goat anti-rabbit IgG (Life technologies, Alexa Fluor® 543 conjugate, Ref. A-11010, 1:1000) and labeled goat anti-mouse IgG (Life technologies, Alexa Fluor® 488, Ref. A-11001, 1:1000). DAPI was used as a nuclear counterstain (final concentration of 1 μg/mL). It also allowed to visualize the amyloid deposits (Additional file 2). Human brain sections were mounted with Fluorescence mounting medium (DAKO S3023). Confocal laser microscopy was performed on a Zeiss, LSM 780 apparatus. Several fields of view (> 300 cells) were analyzed for each group. The confocal composite image (merge) was analyzed using ImageJ 1.51o software. Morphometric investigations were carried out to quantify the percentage of surface stained by SphK2 antibody. SphK2 staining in neurons was estimated on the total surface of each neuron, using a staining threshold taking into account the background. The percentage of SphK2 stained surface was further analyzed in two cell compartments leading to a ratio of nuclear versus cytoplasmic stained surfaces. The same nuclear/cytoplasmic ratio was applied for analysis of intensity of SphK2 staining. To overcome experimental biases, the percentage of colocalization between MAP2 and SphK2 and the percentage of colocalization between DAPI and SphK2 were determined.

We used a rabbit polyclonal SphK2 antibody raised against amino acid 580–629 sequence. The western blots carried out on SphK1 and SphK2 recombinant proteins and on human brain lysate, confirmed the specificity of the antibody for SphK2 recombinant protein with no cross-reaction for the SphK1 isoenzyme (Fig. 1a2). No other bands were observed with SphK2 antibody in whole control brain lysate.

Most of the subjects presented a high number of neurofibrillary tangles as well as of amyloid deposits (Braak V-VI and Thal 4 to 5; Table 1). In both studied cortical areas, thickness variability was noticed and could be related to atrophy, which is a common feature in AD. The Aβ deposits were more frequent in cortical layers II and III, mostly represented in fields 2 to 6 (Additional file 3). There was no correlation between the density of neurons and Aβ deposits (in three locations, p > 0.5 for all regressions), whether Aβ deposits were evaluated as the number of focal deposits or the percentage of amyloid areas (p > 0.5 for all regressions). SphK2 staining was mainly observed in neurons (Fig. 2a), and likely in the nucleus of cells exhibiting the hallmarks (size, shape, specific location) of oligodendrocytes. SphK2 staining was also observed less frequently in astrocytes. In neurons, the immunostaining of SphK2 was both nuclear and cytoplasmic. Because the nuclear immunostaining of SphK2 was difficult to quantify, our analysis focused on cytoplasmic SphK2 expression to avoid experimental bias. The cytoplasmic SphK2 expression was heterogeneous between the three analyzed structures (F_(2,647) = 350.8, p <  0.0001). The percentage of SphK2 positive neurons was higher in CA1 (vs frontal or entorhinal cortex, p <  0.0001 for both comparisons) and lower in frontal vs entorhinal cortex (p <  0.0001).

The percentage of SphK2 positive neurons (cytoplasm) was inversely correlated with the amount of amyloid deposits (Aβ surface percentage) in the frontal cortex (r = 0.037, p = 0.007; Fig. 2b1), in the entorhinal cortex (r = 0.062, p <  0.001; Fig. 2b2) and in CA1 (r = 0.29, p <  0.001; Fig. 2b3). This negative correlation was also found for the focal plaques (frontal cortex: r = 0.044, p = 0.003; entorhinal cortex: r = 0.064, p <  0.001; and CA1: r = 0.38, p <  0.001; Fig. 2c). From the coefficient values, we noted that the relationship between the amount of amyloid and the percentage of SphK2 positive neurons was stronger in the CA1 area than in the cortical regions (frontal and entorhinal).

The amount of Aβ was assessed in human brain lysates from the frontal cortex, hippocampal area (entorhinal cortex and CA1) in AD and control groups. To better appreciate the relation between SphK2 and Aβ deposits, we analyzed samples of cerebellum, a brain area where Aβ deposits are scarce and appear late in the evolution of the disease [50] (Fig. 3a, top row). Western blot studies revealed the presence of the β-amyloid precursor protein (APP, approximately 130 kDa band) along with a set of apparent supramolecular Aβ assemblies. High-molecular weight Aβ oligomers or multiples of trimeric Aβ oligomers have been previously characterized in human AD brain tissue [32, 51]. Quantitative analyses did not demonstrate significant differences between control and AD groups for APP (p > 0.5).

In addition to 150–250 kDa bands corresponding to higher MW Aβ oligomers, bands of 27 kDa were detected. This molecular masses apparently correspond to multiples of trimeric Aβ and could thus represent hexameric (Aβ*27) Aβ assemblies. High levels of total Aβ (high MW oligomers and Aβ*27) were observed in AD group in comparison with the control group (F_(1,38) = 10.7; p = 0.002). When each structure was considered separately (Fig. 3b), statistical analysis revealed a significant increase of total Aβ in frontal cortex (p = 0.009) and an increase at the margin of statistical significance in the hippocampal area (p = 0.06). No difference was found in the cerebellum (p > 0.5).

With regard to the Aβ*27 oligomers, there was a significant difference between groups (F_(1,38) = 7.9; p = 0.008), with a significant increase in hippocampal area (p = 0.028; Fig. 3c) and an increase in the frontal to the margin of the significance (p = 0.08). No difference was found in cerebellum (p = 0.99).

Higher MW Aβ oligomers (F_(1,38) = 8.6; p = 0.006) content were markedly increased in frontal cortex (p = 0.02) but not significantly in hippocampal area (p = 0.095) and in the cerebellum (p = 0.4), in AD group (data not shown).

These data are in accordance with the evolution of Aβ deposits and the distinct sequences in which the regions of the brain are hierarchically involved according to the Thal stages (Phase 1: neocortical Aβ deposits, Phase 2: hippocampal area deposits, Phase 5: cerebellum is the latest structure affected by Aβ deposits) [50].

The SphK2 immunoblots showed two major bands in brain lysates, a 75 kDa band corresponding to the large isoform of SphK2 (SphK2b) and a protein of lower molecular weight which was considered as a cleaved fragment of SphK2 (Fig. 3a, low row). A cleaved fragment of SphK2 has been previously described in extraneural cell models (Jurkat T cells, murine fibroblasts, spleen cells) in relation to apoptotic process [53].

The quantification of total SphK2 (full length and cleaved forms) revealed an increase of its expression in AD samples, in frontal cortex (F_(1,14) = 5.4; p = 0.04) but not in the hippocampal area and cerebellum (p > 0.5 for both comparisons). However, full length SphK2 was not different between AD and control groups (data not shown) (p > 0.05). In contrast, the cleaved SphK2 fragments (Fig. 3d) was higher in AD extracts as compared to control (F_(1,38) = 7.6; p = 0.009). More specifically, an increase was observed in the frontal cortex (F_(1,14) = 5.8; p = 0.03) and in the hippocampal area, to the margin of the significance (F_(1,14) = 4.1; p = 0.06). No difference was found in cerebellum (p > 0.08).

Correlation analyses were made to confirm relationships between individual expression of a cleaved SphK2 form and Aβ levels (Fig. 3e). The individual expression of cleaved SphK2 form correlated positively with Aβ*27 levels in frontal cortex (r = 0.35, p = 0.016; Fig. 3e1) and in hippocampal area (r = 0.48, p = 0.003; Fig. 3e2), but not in cerebellum (p > 0.5; Fig. 3e3).

A positive correlation was also noted between higher MW Aβ oligomers and cleaved SphK2 in frontal cortex (r = 0.47, p = 0.003) and in hippocampal area (r = 0.73, p = 0.001) (data not shown). In the cerebellum, higher MW Aβ oligomers were expressed at a low level in one AD case (Fig. 3a3). Cleaved SphK2 was also found in this case.

To analyze the subcellular localization of both full length and cleaved SphK2, cytoplasmic- and nuclei-enriched fractions were prepared from frontal cortex and hippocampal area. In line with data presented in Fig. 3, the cleaved SphK2 expression was markedly increased in AD samples when compared to controls (F_(1,58) = 5.7; p = 0.02; Fig. 4a) (data not shown).

The quantification of total SphK2 (full length and cleaved forms) revealed a differential expression between the two subcellular fractions in control group (F_(1,18) = 12.9; p = 0.002). SphK2 was less expressed in the nuclei-enriched fraction (1.22^E8 ± 0.23^E8 densitometry units) than in the cytoplasmic-enriched fraction (2.47^E8 ± 0.25^E8 densitometry units). This difference was found in the frontal cortex (p = 0.04) and in the hippocampal area (p = 0.01). In AD group (F_(1,38) = 0.35; p = 0.5), no difference was found between the nuclear (2.75^E8 ± 0.63^E8 densitometry units) and the cytoplasmic (3.22^E8 ± 0.48^E8 densitometry units) (data not shown).

For each subcellular fraction, we analyzed separately full length and cleaved SphK2 fragments in AD group as compared to control. The results are expressed as percentage variation in AD versus control group (Fig. 4b and c).

Full length SphK2 expression was increased in the nuclei enriched fraction (F_(1,28) = 3.9; p = 0.050) but was decreased in cytoplasmic enriched fraction (F_(1,28) = 5.4; p = 0.027) in AD group relative to the control (Fig. 4b). Post-hoc analyses showed a reduction of cytoplasmic full length SphK2 mainly in the hippocampal area (p < 0.001).

A rise of cleaved SphK2 (Fig. 4c) was observed in nuclear fractions (F_(1,28) = 10.4; p = 0.003) and in cytoplasmic fractions (F_(1,28) = 3.9; p = 0.050) in AD group relative to control group especially in frontal cortex (nuclear p = 0.0015 et cytoplasmic p = 0.08; Fig. 4c1). To a lesser extent, yet not significantly, this change of expression was seen in hippocampal area (nuclear p = 0.17 and cytoplasmic p = 0.13; Fig. 4c2).

These data suggest a shift of full length SphK2 from cytosol to the nucleus examined in all brain structures, accompanied by an accumulation of the cleaved SphK2 in the nucleus that would be more significant in the frontal cortex, in AD.

The density of neurons and of amyloid deposits were inversely correlated in the frontal (r = 0.14, p = 0.0004) and entorhinal cortex (r = 0.15, p = 0.002), but not in the CA1 (r = 0.02, p = 0.25). This negative correlation was only related to the presence of focal deposits in all structure (frontal cortex: r = 0.13, p = 0.0006, entorhinal cortex: r = 0.13, p = 0.0045, CA1: r = 0.11, p = 0.005) (data not shown).

We first examined the distribution of SphK2 in the frontal cortex, entorhinal cortex and CA1 then its subcellular localization. With SphK2 immunostaining both full length and cleaved fragments were stained. The immunostaining of SphK2 was mainly seen in neurons and oligodendrocytes (Fig. 5a). In neurons, SphK2 was essentially seen in the nucleus in AD contrary to the control group in which SphK2 was mostly expressed in the cytosol (intensity: p < 0.001; neuron area (%): p < 0.001 for both comparisons). At regional level, SphK2 was not homogeneous (intensity: F_(2,704) = 8.9; p = 0.05; neuron area (%): F_(2,704) = 5.5; p = 0.005). Indeed, SphK2 was more expressed in the CA1 (intensity: p = 0.026 vs frontal cortex, p = NS vs entorhinal; neuron area (%): p = 0.05 vs frontal cortex, p = 0.001 vs entorhinal) (data not shown) as compared to both cortical regions.

In agreement with data shown in Fig. 2, immunofluoresence microscopy revealed that the percentage of SphK2 stained surface in neuron (nuclear and cytoplasmic compartments) was decreased (F_(1,837) = 49.0; p < 0.0001; Fig. 5b) in the AD group as compared to control for all structures (frontal cortex: p = 0.033; entorhinal cortex and CA1: p < 0.0001). The nuclear/cytoplasmic ratio of SphK2 stained surface was increased in AD group (F_(1,837) = 113.5; p < 0.0001). This change of expression was seen in the frontal cortex, the entorhinal cortex and the CA1 (p < 0.001 for all comparisons) as compared to the control subjects, demonstrating a disturbance of the SphK2 subcellular localization (Fig. 5c). This observation was reinforced by the increase of SphK2 nuclear/cytoplasmic ratio of staining intensity in AD group, for the three structures (p ≤ 0.001 for the three structures) (Fig. 5d).