The Effect of Glyceraldehyde-Derived Advanced Glycation End Products on β-Tubulin-Inhibited Neurite Outgrowth in SH-SY5Y Human Neuroblastoma Cells

Nutritional factors can affect the risk of developing neurological disorders and their rate of progression. In particular, abnormalities of carbohydrate metabolism in diabetes mellitus patients lead to an increased risk of neurological disorders such as Alzheimer’s disease (AD). In this study, we investigated the relationship between nervous system disorder and the pathogenesis of AD by exposing SH-SY5Y neuroblastoma cells to glyceraldehyde (GA). We previously reported that GA-derived toxic advanced glycation end products (toxic AGEs, TAGE) induce AD-like alterations including intracellular tau phosphorylation. However, the role of TAGE and their target molecules in the pathogenesis of AD remains unclear. In this study, we investigated the target protein for TAGE by performing two-dimensional immunoblot analysis with anti-TAGE antibody and mass spectrometry and identified β-tubulin as one of the targets. GA treatment induced TAGE-β-tubulin formation and abnormal aggregation of β-tubulin, and inhibited neurite outgrowth in SH-SY5Y cells. On the other hand, glucose-derived AGEs were also involved in developing AD. However, glucose did not make abnormal aggregation of β-tubulin and did not inhibit neurite outgrowth. Understanding the underlying mechanism of TAGE-β-tubulin formation by GA and its role in neurodegeneration may aid in the development of novel therapeutics and neuroprotection strategies.


Introduction
The balanced intake of basic nutrients including carbohydrates is necessary to maintain human health. There have been many reports regarding whether type 2 diabetes mellitus (DM) is a clinical risk factor for Alzheimer's disease (AD), given that the incidence of AD is as much as 2-5 times higher in DM patients [1]. Reactive derivatives from non-enzymatic sugar-protein condensation reactions, as well as lipids and nucleic acids exposed to reducing sugars such as glucose, form a heterogeneous group of irreversible adducts called advanced glycation end products (AGEs) [2]. AGEs form and subsequently accumulate in various tissues during normal aging, and their formation and accumulation is highly accelerated in DM patients [3,4].

Slot Blotting Analysis for TAGE
Cells were harvested and homogenized after being treated within 12 h with GA. An equal amount of protein was applied to a Hybri-SLOT apparatus (Gibco BRL) and transferred to a nitrocellulose membrane (Whatman, Tokyo, Japan) by vacuum filtration. After blocking with 3% bovine serum albumin for 1 h at room temperature, samples were incubated with the anti-TAGE antibody at 4 • C overnight, followed by incubation with an anti-rabbit IgG antibody (Sigma Aldrich Saint Louis, MO, USA). Antibody-bound protein bands were detected using a BCIP-NBT Kit and densitometrically analyzed.

Protein Identification
Protein samples from control or GA-treated cells were subjected to 2-DE as described above and the spots were manually excised from the 2D gels stained with Coomassie Brilliant Blue R-250 (BioRad, Tokyo, Japan). The gel pieces were destained with 50% acetonitrile in 25 mM ammonium bicarbonate at room temperature, then desalted with 100% acetonitrile and then 50 mM ammonium bicarbonate. The gel pieces were digested with 10 µg/mL trypsin (Promega, Madison, WI, USA) solution at 37 • C overnight. The extracted peptides were subjected to matrix-assisted laser desorption/ionization time of flight tandem mass spectrometry (MALDI-TOF/TOF MS). The data were submitted to a MASCOT search engine for identification (https://www.matrixscience.com).

Neurite Outgrowth
SH-SY5Y cells were differentiated with 1% fetal bovine serum to avoid overgrowth of cells and retinoic acid (RA) at 10 µM for 24 h. The effects of GA were observed following addition to the culture medium for 12 h during RA-induced differentiation. Neurite outgrowth stained with anti-α-internexin was observed by fluorescence microscopy and assessed by Scion Image software. Neurite outgrowth was quantified by obtaining 20 random images per dish from five independent dishes and assessing the longest neurites in each image (n = 100) [19]. We defined the control axon length to 100% and the average neurite length was expressed as the mean ± SEM. We did experiments with masking test by two coauthors (R.N. and Y.K.).

Immunocytochemistry
Cells were cultured and fixed in 0.1% glutaraldehyde containing phosphate saline buffer (pH 7.4). The cells were blocked with Blocking One (Nacalai Tesque) and incubated with primary anti-β-tubulin (1:500, D3U1W, CST) and anti-α-internexin (1:500, Abcam, Tokyo, Japan, ab10830) antibodies. The cells were then incubated with Alexa fluoro 488 anti-IgG (Molecular Probes, Eugene, OR, USA) and the cell nuclei were stained with acridine orange (AO, Dojindo, Tokyo, Japan). To confirm there were no immunoreactivities of non-specific or auto fluorescence, we used a secondary antibody as a negative control.

Statistics
All results are reported as mean ± SEM. Differences between groups were analyzed using one-way ANOVA, followed by Dunnett's multi-comparison test with PASW Software (SPSS Inc., Chicago, IL, USA). p values < 0.05 were considered statistically significant.

Detection and Identification of TAGE Proteins
In our previous study, 1 mM GA increased the formation of TAGEs in SH-SY5Y cells and showed cell death within 24 h [9], but it was unclear which proteins were the targets of TAGE formation. GA treatment induced neurotoxicity within 24 h ( Figure 1A) and TAGE formation was significantly observed from 12 h ( Figure 1B). We comprehensively detected TAGE protein before cell death, by 2D immunoblot analysis with anti-TAGE antibody at 12 h treatment of GA. The pattern of TAGE proteins differed between GA-treated for 12 h and vehicle control SH-SY5Y cells ( Figure 1C,D). Individual spot volumes represent the amount of TAGE protein levels. In the GA-treated cells, the intensities of seven spots increased (data not shown). To identify these proteins, MALDI-TOF/TOF MS analysis were performed by searching a sequence database using specific peptide mass data (Supplementary data Figure S1A,B). We identified several protein spots by amino acid sequence analysis, including human β-tubulin (MW: 50 kDa, pI: 4.78) ( Figure 1C,D, arrows).

TAGE Formation and Abnormal Aggregation of β-Tubulin by GA
We examined the formation of TAGE by incubating recombinant tubulin proteins with 0, 0.3, 0.7, or 1 mM GA for 12 h. GA dose-dependent increased the formation of TAGE (Figures 2A-D). Interestingly, the levels of 55 kDa monomer tubulin protein (

TAGE Formation and Abnormal Aggregation of β-Tubulin by GA
We examined the formation of TAGE by incubating recombinant tubulin proteins with 0, 0.3, 0.7, or 1 mM GA for 12 h. GA dose-dependent increased the formation of TAGE (Figure 2A-D). Interestingly, the levels of 55 kDa monomer tubulin protein (Figure 2A  The monomer band dose-dependently decreased after GA-treatment ( Figure 2E,H) whereas all polymer β-tubulin bands become denser in a dose-dependent manner ( Figure 2E,G; lower band, Figure 2E,F; upper band). Next, we confirmed the formation of TAGE-β-tubulin by GA treatment in SH-SY5Y cells ( Figure 2I-M). We used β-actin antibody for confirmation of loading control in Figure 2I. As the band size is similar between β-actin ( Figure 2I) and β-tubulin monomer band ( Figure 2J), we used the same lot samples with another loading membrane ( Figure 2I,J). Non-specific bands were seen because we used large quantities of protein (180 µg) in Figure 2I to indicate aggregation of β-tubulin bands clearly in Figure 2J. The monomer band significantly decreased after GA-treatment ( Figure 2J,M) whereas all polymer β-tubulin bands ( Figure 2J,L; lower band, Figure 2J,K; upper band) become denser after GA-treatment.

GA Inhibits Neurite Outgrowth From SH-SY5Y Cells
As β-tubulin is present in growing neurites as well as in cell bodies with various localization profiles, we observed the localization of β-tubulin using an immunocytochemical study. In the first, we checked non-specific immunofluorescence using only secondary antibody ( Figure 3A,B). SH-SY5Y cells have short neurite in vehicle treatment ( Figure 3A). Both auto fluorescence and non-specific fluorescence staining could not be seen in Figure 3B. In the vehicle control, β-tubulin positive staining was seen in the growth cone, which is involved in stabilizing the microtubule population and axonal growth ( Figure 3C). In differentiated axons treated with RA, strong staining of β-tubulin was observed in the axons and growth cones ( Figure 3D). However, in RA-differentiated cells treated with GA, β-tubulin staining was observed in the cytosol and in the axon hillock area, which might be a starting point for neurite outgrowth ( Figure 3E). Furthermore, we evaluated neurite length by immunocytochemistry of α-internexin, which is one of the axon marker proteins [20]. We investigated whether GA suppresses neurite outgrowth in differentiated SH-SY5Y cells. At 10 µM, RA significantly induced neurite outgrowth ( Figure 3G,I) compared with a vehicle treatment ( Figure 3F,I) whereas 1 mM GA inhibited neurite outgrowth by RA in SH-SY5Y cells ( Figure 3H,I). The cell death of SH-SY5Y treated with RA and/or GA was evaluated using an MTT assay. RA and/or GA treatment did not have any effect on cell viability relative to the vehicle control ( Figure 3J).
( Figures 2E,G; lower band, Figure 2E,F; upper band). Next, we confirmed the formation of TAGE-βtubulin by GA treatment in SH-SY5Y cells (Figures 2I-M). We used β-actin antibody for confirmation of loading control in Figure 2I. As the band size is similar between β-actin ( Figure 2I) and β-tubulin monomer band ( Figure 2J), we used the same lot samples with another loading membrane ( Figures  2I,J). Non-specific bands were seen because we used large quantities of protein (180 μg) in Figure 2I to indicate aggregation of β-tubulin bands clearly in Figure 2J. The monomer band significantly decreased after GA-treatment ( Figures 2J,M)

Glucose Did Not Induce Abnormal β-Tubulin Aggregation and Did Not Inhibit Neurite Outgrowth
To know whether GA specifically induced abnormal β-tubulin aggregation and inhibited neurite outgrowth, we further checked these effects by glucose. We examined the aggregation of β-tubulin by incubating recombinant tubulin proteins with 1 mM glucose for 12 h. The levels of β-tubulin were detected by anti-β-tubulin antibody (Figures 4 A-C). The levels of the monomer bands ( Figure 4B, 55 kDa) and lower bands ( Figure 4C, 154 kDa) of β-tubulin were detected by western blot using an anti-

Glucose Did Not Induce Abnormal β-Tubulin Aggregation and Did Not Inhibit Neurite Outgrowth
To know whether GA specifically induced abnormal β-tubulin aggregation and inhibited neurite outgrowth, we further checked these effects by glucose. We examined the aggregation of β-tubulin by incubating recombinant tubulin proteins with 1 mM glucose for 12 h. The levels of β-tubulin were Nutrients 2020, 12, 2958 9 of 13 detected by anti-β-tubulin antibody (Figure 4 A-C). The levels of the monomer bands ( Figure 4B, 55 kDa) and lower bands ( Figure 4C, 154 kDa) of β-tubulin were detected by western blot using an anti-β-tubulin antibody. Glucose treatment did not affect the levels of both monomer and lower bands of β-tubulin compared to vehicle treatment. Next, we evaluated neurite length by immunohistochemistry of α-internexin. We investigated whether glucose changed neurite outgrowth in differentiated SH-SY5Y cells. At 10 µM, RA significantly induced neurite outgrowth ( Figure 4E,G) compared with a vehicle treatment ( Figure 4D,G). Using 1 mM glucose did not change neurite outgrowth by RA in SH-SY5Y cells ( Figure 4F,G).
Nutrients 2020, 12, x FOR PEER REVIEW 9 of 13 β-tubulin antibody. Glucose treatment did not affect the levels of both monomer and lower bands of β-tubulin compared to vehicle treatment. Next, we evaluated neurite length by immunohistochemistry of α-internexin. We investigated whether glucose changed neurite outgrowth in differentiated SH-SY5Y cells. At 10 μM, RA significantly induced neurite outgrowth (Figures 4E,G) compared with a vehicle treatment (Figures 4D,G). Using 1 mM glucose did not change neurite outgrowth by RA in SH-SY5Y cells (Figures 4F,G).

Discussion
The present study shows four salient findings: 1) β-tubulin is one target of GA-induced AGEs; 2) GA-induces abnormal β-tubulin aggregation in a dose-dependent manner; 3) TAGE-β-tubulin formation inhibits neurite outgrowth in SH-SY5Y cells; 4) glucose did not induce abnormal β-tubulin aggregation and did not inhibit neurite outgrowth. Recent epidemiological studies have reported that the risk of developing AD is higher in DM patients (2-5 fold higher compared with the non-diabetic population). The Rotterdam study surveyed over 6000 patients and indicated a strong relationship between DM and AD, with a relative risk (RR) of 1.9 [21]. Given recent interest in the relationship between insulin and AD, it is noteworthy that patients in that study receiving exogenous insulin therapy were at the highest risk (RR 4.3) of developing dementia [22,23]. AGE levels were previously shown to be increased in the brains of diabetic patients with AD [24], and thus these reports may partly explain the clinical link between DM and AD. A relationship between AGEs and AD was suggested in several reports in 1994-1995 [5,24,25]. Furthermore, glucose-derived AGEs have effects similar to amyloid β (Aβ), namely increased neurotoxicity and glucose consumption by SH-SY5Y cells [26,27]. We previously reported [20,28] that α-hydroxyaldehyde (GA and glycolaldehyde) and dicarbonyl compounds such as glyoxal, methylglyoxal and 3-deoxyglucosone contribute to the glycation of proteins. We also reported that TAGE are stronger neurotoxins rather than Glu-AGE in a neuronal culture system [8,29]. In addition, the neurotoxic effects of serum AGEs from diabetic patients on hemodialysis were eliminated by addition of an anti-TAGE-specific antibody, but not by antibodies towards glycolaldehyde-, methylglyoxal-, glyoxal-, 3-deoxyglucosone-or Glu-AGEs [8,29]. We recently reported that the epitope structure recognized by anti-TAGE antibody is different from the previously reported GA-derived AGE structures, i.e., 3-hydroxy-5-hydroxymethyl-pyridinium compound (GLAP) and triosidines structures. We found that the anti-TAGE antibody differed from antibodies for well-defined AGEs as well as those for AGEs derived from reducing sugar/carbonyl molecules with unknown structures [16,17]. TAGE are more neurotoxic than Glu-AGEs and CML, two extensively examined AGEs species. We also reported that TAGE may be general causative agents for the development of neurodegenerative diseases such as AD [9]. On the other hand, a Glu-AGE antibody was shown to react with SPs mainly with the amyloid core, whereas the GA-AGE antibody showed no immunoreactivity with SPs [14]. These results suggest that Aβ may be glycated by glucose rather than GA. TAGE are mainly present in the neurons of the hippocampus and parahippocampal gyrus and are mainly localized in the cell body of neurons [14]. Glu-AGEs were detected in both intracellular and extracellular sites, whereas TAGE were only found intracellularly, indicating that the mechanism underlying the neurotoxicity induced by Glu-AGEs and TAGE is different. Glycation is a post-translational modification produced by a reaction between reducing carbohydrates and the amino groups, such as lysine. Recently, more interest has been paid to lysine glycation from many researchers working on metabolism. However, the systematic identification of a glycation site is still challenging because the glycated residues do not show significant patterns. In this study, GA made several β-tubulin molecules aggregate. We do not know which sites of β-tubulin are modified by GA treatment, however there are at least more than 20 lysine residues in its amino acid sequence. Further studies are needed to clarify the mechanism behind the formation of abnormal aggregation of β-tubulin. Interestingly, it was reported that β-tubulin is also glycated by glucose in a DM experimental model [30] although glucose did not trigger abnormal aggregation of β-tubulin or inhibition of neurite outgrowth under our experimental conditions (Figure 4).
GA is derived from two distinct pathways: the glycolytic pathway (glycolysis) and the fructose metabolism (fructolysis) pathway [31]. Under hyperglycemic conditions, an increase in intracellular glucose stimulates the polyol pathway to generate fructose in insulin-independent tissues, including brain and nerve tissue [32]. Fructose is phosphorylated to fructose-1-phosphate and then catabolized to dihydroxyacetone phosphate and GA by aldolase B [33], and then GA promotes the formation of TAGE. Moreover, it has been reported that aldolase B is not expressed in the rat brain [34], and fructose-1-phosphate cleavage (aldolase) activity has been detected in the human brain [35].
However, it is thought that the progression of AD will take over 10 years because the enzyme expression level is considered to be very low. Although quantification of GA is currently impossible, our previous data showing that TAGE could be detected in the brain of AD patient supports this theory [14].
AD is characterized pathologically by the presence of NFTs at intracellular sites. NFTs are composed of paired helical filaments (PHFs) and straight filaments. The major component of PHFs is the microtubule-associated protein, tau [36,37]. Tau in PHFs shows distinctive properties such as high aggregation, hyper-phosphorylation and other post-translational modifications [38]. Tau proteins are enriched in normal neuronal axons where they regulate microtubule stability. However, tau is detached from microtubules and aggregates in the cytosol in the presence of NFTs in a diseased brain. From these results, free-tau proteins may result in self aggregation and tend to be phosphorylated at disease-associated sites [39]. In our previous report, GA increased intracellular tau phosphorylation levels in SH-SY5Y cells. Microtubules are constructed by the polymerization of dimers of αand β-tubulin by self-assembly. They serve as architectural elements and support the elongated shape. Thus, we focused on the microtubule-related and tau-associate β-tubulin protein. In this study, GA significantly inhibited neurite outgrowth by RA ( Figure 3H,I). These results might implicate that abnormal aggregation of β-tubulin by GA cannot form normal heterodimers with α-tubulin and might inhibit polymerization of microtubules. This may be the reason that β-tubulin staining was observed in the axon hillock area by GA treatment in differentiated SH-SY5Y cells ( Figure 3E). The inhibition of RA-induced neurite outgrowth by GA was not dependent on cell toxicity ( Figure 3J). Further studies are needed in order to elucidate the exact mechanisms underlying β-tubulin aggregation by GA.

Conclusions
Although the exact structure and/or mechanisms of TAGE-β-tubulin and its downstream signaling pathway currently remain unclear, we found β-tubulin is one of the targets of TAGE. We also demonstrated GA but not glucose-induced abnormal aggregation of TAGE-β-tubulin or inhibited neurite outgrowth. TAGE-β-tubulin may be a useful target for understanding the mechanism of DM-related AD.