Upregulation of MIF as a defense mechanism and a biomarker of Alzheimer’s disease

CSF samples were obtained from patients visiting the Guangzhou General Hospital. The control group included patients who had no history or evidence of cognitive decline. CSFs were taken by lumbar puncture under anesthesia when the subjects had surgery for diseases other than inflammatory diseases of the central nervous system. CSFs from patients with AD, mild cognitive impairment (MCI), and vascular dementia (VD) were collected for neurological diagnosis. Table 1 listed the sample information regarding the sample size, sex, and age in each group. Frozen control and AD human cortices were obtained from the Department of Pathology, Columbia University. These samples were used to examine the expression of MIF by immunoblotting experiments. Table 2 listed the sample information regarding the sex, age, and brain areas used in the study.

Animal experiment protocols were approved by The University of British Columbia Animal Care and Use Committee. APP23 transgenic mice carry human APP751 cDNA with the Swedish double mutation at positions 670/671 (KM→NL) under control of the murine Thy-1.2 expression cassette [15, 16]. The PS45 transgenic mice carry human presenilin-1 cDNA with the G384A mutation [17]. Mif^+/− mice on the c57/BL6 background were generated by breeding Mif^+/− mice on BALB/c background (The Jackson Laboratory), with in-house bred c57/BL6 mice. All mice were allowed to access water and food ad libitum. APP23/PS45 double transgenic mice were bred by cross APP23 with PS45 mice. APP23/MIF^+/− mice were bred by cross APP23 with Mif^+/− mice on c57/BL6 background. Positive pups were determined by genotyping [18].

The Morris water maze test was performed as previously described [19]. Briefly, the test was performed in a 1.5-m diameter pool with a 10-cm diameter platform placed in the southeastern quadrant of the pool. The procedure consisted of 1 day of visible platform tests and 4 days of hidden platform tests, plus a probe trial 24 h after the last hidden platform test. In the visible platform test, mice were tested for five continuous trials with an inter-trial interval of 60 min. Mouse behavior including distance traveled and escape latency was automatically video-recorded by automated video tracking (ANY-maze, Stoelting). The tests were performed on APP23/MIF^+/− mice, and APP23 mice, which were negative littler mates of APP23/MIF^+/− mice. The tests were performed at the ages between 13 and 14 months.

Half brains were fixed in 4% PFA in PBS. Fixed brains were either dehydrated in 30% sucrose solution followed by cryosectioning at 30 μm thickness or prepared for paraffin-embedded sectioning at 5 μm thickness. For immunohistochemistry, the brain slices were incubated with 3% hydrogen peroxide in PBS for 10 min, permeabilized in 0.3% Triton X-100 in PBS (PBS-Tx) for 30 min, and blocked with 5% BSA in PBS for 1 h at 22 °C. Next, the slices were incubated with primary antibodies at 4 °C overnight. After rinsing with PBS-Tx for three times, the slices were applied with secondary antibodies for 1 h at 22 °C, followed by 30-min incubation with avidin-biotin-peroxidase complex (ABC, Vector Laboratories). Color development was achieved by the DAB (Vector Laboratories) method. After rinsing with ddH₂O, brain sections were subjected to additional hematoxylin staining to visualize nuclei. The sections were observed under traditional microscopy. For Aβ plaque detection, the procedure for secondary antibody incubation was omitted, and the slices were proceeded for ABC incubation and color development. The number of plaques was quantified manually following qualification under × 40 magnification. The area of the plaques was quantified using ImageJ. Images were taken from 10 matching areas (5 slices with 540-μm intervals for each mouse) between APP23 and APP23/MIF^+/− transgenic mice. All the images used for plaque quantification were taken at the same time with the same exposure level. For immunofluorescent staining, the brain slices were permeablized in PBS-Tx for 30 min followed by sequential incubation with primary and fluorescent-labeled secondary antibodies as above. After rinsing with PBS, brain sections were coverslipped using the VECTASHILD® mounting medium with DAPI (Vector Laboratories) and observed under fluorescent microscopy. The primary antibodies are rabbit anti-MIF antibody (Torrey Pines Biolabs), mouse anti-GFAP antibody, biotinylated 4G8 antibody, rabbit anti-Iba-1 (DAKO). Secondary antibodies are biotinylated swine anti-rabbit IgG (DAKO) for immunohistochemistry, Alexa 488-labeled goat anti-rabbit IgG (Invitrogen), Alexa 594-labeled goat anti-rabbit IgG (Invitrogen), Alexa 488-labeled goat anti-mouse IgG (Invitrogen), and Alexa 594-labeled goat anti-mouse IgG (Invitrogen) for immunofluorescent staining. Primary and secondary antibodies were diluted in PBS with 1% BSA.

Cortical tissues were homogenized by sonication with 5X (v/w) RIPA-DOC lysis buffer supplemented with a complete mini protease inhibitor cocktail tablet (Roche Molecular Biochemicals). The samples were then centrifuged at 16,000×g at 4 °C for 30 min. The supernatants were removed and added to 2X Novex® tricine SDS sample buffer (Invitrogen) followed by boiling at 100 °C for 2 min. The samples were resolved in 12% tris-tricine gels and transferred to PVDF-FL membranes (Millipore). The membranes were blocked with 5% non-fat milk and incubated with primary antibodies for MIF (Torrey Pines Biolabs) and β-actin (Sigma, AC-15). To detect the proteins, IDye680-labeled goat anti-rabbit and IDye800-labeled goat anti-mouse antibody were used. The blots were scanned using the Odyssey Imager (Licor).

The mouse microglia cell line BV-2, mouse macrophage cell line RAW264.7, human neuroblastoma cell line SHSY-5Y, and a stable cell line overexpressing MIF (SYMS) were maintained in Dulbecco’s modified Eagle’s medium supplemented with 10% fetal bovine serum, 1 mmol/L of sodium pyruvate, and 2 mmol/L of l-glutamine (Invitrogen). Cells were seeded onto 96-well plates and cultured at 37 °C in an incubator supplemented with 5% CO2. Aβ oligomers were prepared as previously described with modification [20, 21]. Briefly, synthetic Aβ1-42 was dissolved in 1,1,1,3,3,3-hexafluoro-2-propanol (HFIP, Fluka), vacuum dried, and dissolved in DMSO as a 5 mM stock. Aβ oligomer was prepared by diluting the stock Aβ in sterile PBS to 100 μM and incubated at 4 °C for 12 h. The oligomers were further diluted to 10 μM or 50 μM by culture medium to treat the cells. LPS was used as a positive control for MIF secretion and was used at the concentration of 100 ng/mL. Sixteen hours after Aβ treatment (10 μM), the culture medium was collected and centrifuged at 3000×g for 2 min at 4 °C prior to assays. MIF concentrations in culture media were measured by a human (R&D systems) or mouse (Mybiosource) MIF ELISA kit following the manufacturer’s instruction. Culture media were diluted five and two times prior to the assays to measure human and mouse MIF concentration, respectively. To assess cell membrane integrity, LDH assay (Promega) following the manufacturer’s instruction was performed using the same batch of culture medium. SHSY-5Y and SYMS were treated with Aβ oligomers (50 μM). To detect Aβ-induced cytotoxicity, MTS assay was performed following the manufacturer’s protocol (Promega).

To prepare the membrane for the dot blot assay, 2 μL of oligomerized Aβ peptide (100 μM) or purified green fluorescent protein (GFP) protein (approximately 50 μM) were spot on a nitrocellulose membrane and were let dry. The membrane was then blocked in 0.3% BSA in PBS for 1 h at room temperature prior to incubation with mixed proteins of purified hMIF and GFP at the concentration of approximated 5 μM at 4 °C for overnight. The membrane was then washed, and immunoblotting was performed to detect MIF and GFP. The primary antibody to detect MIF was a monoclonal anti-MIF antibody (D-2, Santa-Cruz). The primary antibody to detect GFP was a polyclonal anti-GFP antibody.

Postmortem cortical tissues collected from AD patients and controls were assayed by immunoblotting to evaluate the expression of MIF (Fig. 1a). The expression level of MIF was significantly elevated for 1.58 ± 0.14 folds of the controls (p < 0.05) (Fig. 1b). To assess the MIF level in AD patients, CSF was collected from control subjects and patients diagnosed with MCI and AD. CSF from patients with VD was also collected. The age and sex distribution of the patients are listed in Table 2. Our results showed that the MIF level in the CSF was significantly upregulated in the patients diagnosed with AD (14.62 ± 1.15 ng/ml) compared to the control subjects (10.07 ± 0.60 ng/ml) (p < 0.05) (Fig. 1c). There were no significant differences in the concentration of MIF between the MCI, VD patients (9.89 ± 1.48 and 9.86 ± 0.83 ng/ml, respectively) and the control subjects (p > 0.05). However, the MIF level in the MCI and VD patients was significantly lower than that in AD patients (p < 0.05) (Fig. 1c). Notably, the CSF concentration of MIF elevated approximately 1.4 folds in AD patients compared to the that in control subjects, which was similar as the fold increase in the brain tissue, indicating the level of MIF in CSF could be an indicator of the level of MIF in the brain tissue. Our results demonstrated that MIF was upregulated in AD, and the elevation of MIF in CSF could be a biomarker of AD.

To determine the change of the MIF expression during AD pathology, we measured the expression of MIF in the brain tissue from APP23/PS45 double transgenic mice at different ages. APP23/PS45 mice have an accelerated AD-like pathological progression [17, 22, 23], and by the age of 3 months, the mice have developed a significant amount of plaques and cognitive impairments in behavioral tests. Brain slices obtained from APP23/PS45 mice were stained with thioflavin S and 4G8 for plaque detection. In 2-month old mice, a few plaques appeared in the cortical and hippocampal regions and were scattered in the cortical layers (Fig. 1d). As expected, the number of plaques significantly increased at 3 months of age in all brain regions as indicated by both thioflavin S and 4G8 staining (Fig. 1d). MIF expression levels in cerebral parenchyma were also measured using the same group of mice. Immunoblotting results showed that MIF protein level in the brain tissue was similar between wildtype and APP23/PS45 mice at the age of 2 months (1.00 ± 0.03 vs 1.02 ± 0.02 folds, p > 0.05) (Fig. 1e). However, at 3 months of age, MIF expression in the APP23/PS45 mice increased to 1.37 ± 0.05 fold (p < 0.05) of the WT mice at the same age (Fig. 1e). Our results demonstrated that upregulation of MIF in the APP23/PS45 mice occurred at the late stage of the disease when a large amount of amyloid plaques had been formed.

MIF is a secretory protein and exerts functions via autocrine/paracrine mechanisms. To examine whether Aβ could trigger MIF secretion, we first tested whether MIF could be secreted after Aβ treatment on mouse microglia and mouse macrophage cell lines, BV-2 and RAW 264.7, respectively. LPS was used as a positive control as it triggers MIF secretion in RAW 264.7 [24, 25]. Baseline expression of MIF was detected in the medium from RAW 264.7 at the concentration of 11.73 ± 1.54 pg/mL (Fig. 2a), while not detectable in that from BV-2 (Fig. 2b). After the LPS treatment, the concentration of MIF significantly increased in the culture medium of both cell lines to 18.83 ± 0.88 ng/mL (p < 0.05) (Fig. 2a) for RAW 264.7 and 14.35 ± 0.86 ng/ml for BV-2 (p < 0.05) (Fig. 2b). However, Aβ treatment did not significantly alter the concentration of MIF in the medium from RAW 264.7 (9.43 ± 1.24 pg/mL) compared to controls, but it was significantly lower than that from LPS treatment (p < 0.01) (Fig. 2a). To our surprise, Aβ treatment did not elevate the concentration of MIF above the detection limit in the culture medium from BV-2 (Fig. 2b). LDH assays were performed to examine the cell membrane integrity that could affect the release of MIF. Our result demonstrated that Aβ treatment did not result in cell membrane leakage in both cell lines.

To examine neuronal secretion of MIF by Aβ, we treated SH-SY5Y cells. The baseline release of MIF to the culture medium from the SH-SY5Y was also observed at the concentration of 26.98 ± 0.91 ng/mL (Fig. 2c). In contrast to the increased MIF secretion in RAW264.7 and BV-2, LPS treatment slightly reduced MIF release in SH-SY5Y to the concentration of 24.36 ± 1.12 ng/mL, compared to controls (p < 0.05) (Fig. 2c). The Aβ treatment on SH-SY5Y cells induced a significant increase of MIF concentrations in the culture medium to 32.97 ± 0.79 ng/mL (Fig. 2c), and the increase was not due to the cell membrane leakage as shown by LDH assay (Fig. 2d). Taken together, our result suggested that Aβ-triggered MIF secretion by neurons.

Next, we examined whether elevated MIF secretion could protect neuronal cells from Aβ-induced toxicity. We generated an SH-SY5Y cell line stably overexpressing MIF (SYMS). We treated the SH-SY5Y and SYMS cells with Aβ oligomers (50 μM). After 24 h, the MTS assay was performed to detect cell viability. We found that Aβ treatment reduced the cell survival rate to 73.64% ± 2.49% in SH-SY5Y cells. However, there was no significant difference in cell viability between Aβ-treated SYMS cells and untreated ones, suggesting that elevated MIF secretion protects neurons from Aβ-induced cytotoxicity (Fig. 2e).

The above results suggested that neurons were responsible for the MIF secretion triggered by the Aβ deposits. To test whether MIF participates in cognitive performance during AD pathogenesis, APP23/MIF^+/− and APP23 mice were subjected to memory function assessment by the Morris water maze at the age between 12 and 13 months. In the visible platform tests, APP23/MIF^+/− and APP23 mice had similar escape latencies (40.5 ± 3.0 s and 42.8 ± 4.6 s, p > 0.05) (Fig. 3a) and path length (7.7 ± 0.7 m and 8.7 ± 0.7 m, p > 0.05) (Fig. 3b), indicating that hemizygous knockout of MIF did not affect mouse mobility or vision. In the hidden platform test, APP23/MIF^+/− mice showed significant memory impairment on the fifth day of the test. The escape latency on the last day of the hidden platform test was significantly longer than in APP23 mice (28.1 ± 2.9 vs 15.7 ± 1.4 s, p < 0.05) (Fig. 3c). The APP23/MIF^+/− mice also swam significantly longer distances to reach the platform as compared to APP23 mice (4.3 ± 0.5 vs 2.6 ± 0.1 m, p < 0.05) on the fifth day of hidden platform test (Fig. 3d). In the probe trial on the last day of testing, the platform was removed. APP23/MIF^+/− mice showed a significantly lower number of times passing through the position of the hidden platform than APP23 mice (P < 0.05) (2.2 ± 0.6 vs 0.8 ± 0.3, p < 0.05) (Fig. 3e). These results indicated that MIF deficiency affected spatial learning during the hidden platform training.

Since our results suggested that sufficient MIF is necessary for normal cognitive performance, it seems controversial that AD patients with increased MIF levels still demonstrated memory loss. In order to address this discrepancy, we thoroughly analyzed the expression patterns of MIF in the brain. Immunohistology was performed on brain sections from APP23/PS45 mice using a specific anti-MIF antibody. The pattern of MIF staining demonstrated as spotted clusters in both the hippocampal region and cerebral cortex (Fig. 4A), which remarkably resembled that of 4G8 stained Aβ plaques. Next, we performed the double staining of MIF and 4G8 on an adjacent brain section (5 μm) to the one used for the DAB staining. The fluorescent signals for MIF as pointed by white arrows (Fig. 4B) were confirmed by the DAB staining of MIF as pointed by black arrows (Fig. 4A). The double staining of MIF and 4G8 revealed the colocalization of MIF and Aβ plaques (Fig. 4B). Interestingly, two patterns of association between MIF and Aβ plaques were observed. Besides the colocalization as shown in Fig. 4B, b, some MIF staining exhibited a pattern that embraced the dense core of the Aβ plaques in hippocampal (Fig. 4B, a, insert) as well as in cortical (Fig. 4C) regions. There were a few cells stained positive for MIF (arrowheads, Fig. 4B inserts), and we did further analysis to identify the cell type.

Since MIF expression was tightly associated with amyloid plaques, we examined whether glial cells were responsible for the plaque-associated MIF expression. Double staining of MIF and GFAP was performed on brain sections from APP23/PS45 mice at the age of 7 months (Fig. 5a). A similar MIF expression pattern was observed, confirming the characteristics of MIF expression in the AD brains—enriched in the vicinity of amyloid plaques. Active astrocytes stained with GFAP were prevalent in the entire brain section without particular enrichment. At the site of MIF enrichment, MIF expression overlapped with GFAP staining, although the highest expression of GFAP was neither in close vicinity of nor colocalized with MIF (arrows, Fig. 5a). Due to a lack of available antibodies, double staining for colocalization study was not performed for MIF and microglia. Instead, amyloid plaques were stained with 4G8 to locate possible expression of MIF, and Iba-1 was used to detect microglia (Fig. 5b). Our result demonstrated that Iba-1 positive microglia were in closer vicinity to the amyloid plaques than astrocytes, suggesting highly possible overlapping of microglia and MIF expression. In addition, the morphology of MIF expression cells (arrowheads, Fig. 5a) was very similar to Iba-1 stained microglia (arrowheads, Fig. 5b).

Our results strongly suggested a possible association of MIF with amyloid plaques in the extracellular space. However, one could not rule out the possibility that the presence of MIF was merely due to constant secretion by the vicinity cells. In order to examine whether MIF is associated with plaques in the extracellular matrix, we first examined whether MIF could interact with Aβ using a dot blot assay. Oligomerized Aβ and GFP were spot on a membrane and incubated with the protein mixture of MIF and GFP. GFP spotted on the membrane served as the control for non-specific binding by MIF, and GFP in the protein mixture served as the control for non-specific binding by Aβ. The result demonstrated that MIF interacted with oligomerized Aβ, but neither did MIF nor Aβ interact with GFP, indicating a specific binding between oligomerized Aβ and MIF (Fig. 6a). We then studied whether MIF and Aβ were associated in the AD brain by exploring whether MIF and Aβ present at the same density fraction. Brain tissues were obtained from APP23/PS45 mice and homogenized in PBS with 0.5% Triton X-100. The homogenates were fractionated under 16,000×g for 15 min at 4 °C. The homogenates were clearly separated into four layers, namely supernatant (S), pellet layer 1 (P1), pellet layer 2 (P2), and pellet layer 3 (P3). The supernatants were collected and subjected to ultracentrifugation at 100,000×g for 1 h at 4 °C, and the second supernatants were collected and labeled as S′. The layer P3 mainly consisted of non-uniform tissue debris was discarded. Subsequently, samples collected from S′, P1, and P2 fractions were subjected to immunoblots to locate the presence of Aβ enriched fractions. To study the potential association between plaques and MIF, immunoblots were performed to detect the presence of Aβ and the level of MIF in each collected fraction. Our results demonstrated that Aβ was not detectable in the S′ fraction, but was detected in P1, P2, and P3 fractions (Fig. 6b). Concentrations of MIF were similar between controls and the APP23/PS45 mice in the supernatant, indicating AD pathology did not increase the level of free MIF in the soluble fraction (0.96 ± 0.14 fold of the control, P > 0.05) (Fig. 6b, c). Interestingly, soluble MIF tended to be enriched in the P1 fraction (1.60 ± 0.31 fold of the control, P > 0.05) and was significantly enriched in the P2 fraction (1.92 ± 0.21 fold of the control, P < 0.05), where Aβ aggregates presented (Fig. 6b, c). Taken together, these results are supportive that MIF could bind with Aβ plaques in AD brains.