Association of GLOD4 with Alzheimer’s Disease in Humans and Mice

Background Glyoxalase domain containing protein 4 (GLOD4), a protein of an unknown function, is associated with Alzheimer’s disease (AD). Three GLOD4 isoforms are known. The mechanism underlying GLOD4’s association with AD was unknown. Objective To assess GLOD4’s role in the central nervous system by studying GLOD4 isoforms expression in human frontal cerebral cortical tissues from AD patients and in brains of Blmh−/−5xFAD mouse AD model of AD. Methods GLOD4 protein and mRNA were quantified in human and mouse brains by western blotting and RT-qPCR, respectively. Mouse brain amyloid β (Aβ) was quantified by western blotting. Behavioral assessments of mice were performed by cognitive/neuromotor testing. Glod4 gene in mouse neuroblastoma N2a-APPswe cells was silenced by RNA interference and Glod4 protein/mRNA, Aβ precursor protein (Aβpp)/mRNA, Atg5, p62, and Lc3 mRNAs were quantified. Results GLOD4 mRNA and protein isoforms were downregulated in cortical tissues from AD patients compared to non-AD controls. Glod4 mRNA was downregulated in brains of Blmh−/−5xFAD mice compared to Blmh+/+5xFAD sibling controls, but not in Blmh−/− mice without the 5xFAD transgene compared to Blmh+/+ sibling controls. The 5xFAD transgene downregulated Glod4 mRNA in Blmh−/− mice of both sexes and in Blmh+/+ males but not females. Attenuated Glod4 was associated with elevated Aβ and worsened memory/sensorimotor performance in Blmh−/−5xFAD mice. Glod4 depletion in N2a-APPswe cells upregulated AβPP and downregulated autophagy-related Atg5, p62, and Lc3 genes. Conclusions These findings suggest that GLOD4 interacts with AβPP and the autophagy pathway, and that disruption of these interactions leads to Aβ accumulation and cognitive/neurosensory deficits.


INTRODUCTION
Alzheimer's disease (AD) is the primary cause of dementia, affects 55.0 million individuals worldwide and ranks as the fifth leading cause of death globally [1].In the United States alone, an estimated 6.9 million individuals aged 65 and older (2% of the US population) live with AD dementia today, and this number is expected to grow to 13.8 million by 2060.AD is characterized by the extracellular accumulation of amyloid β (Aβ) and the intracellular accumulation of neurofibrillary tangles of hyperphosphorylated tau protein leading to neuronal death.Mutations in amyloid β protein precursor protein (AβPP), presenilin 1 (PSEN1), and presenilin 2 (PSEN2) lead to the familial early-onset AD, which is relatively rare [2].Although lifestyle and environmental factors have been recognized as modifiers of the susceptibility to AD [3], the causes of the most prevalent sporadic late-onset AD are largely unknown.Because there is no proven way to prevent AD and there is currently no cure available, it is important to identify factors influencing AD and their mode of action.
There is evidence suggesting that the glyoxalase domain containing protein 4 (GLOD4), a protein of unknown function, may play a significant role in the central nervous system (CNS) and may be linked to AD.For instance, an intronic SNP in GLOD4, rs2750012, was associated with increased risk of AD in the Arab population of northern Israel.This association has been replicated in metanalysis performed in seven independent GWAS datasets [4].Levels of one of the Glod4 protein isoforms, isoform 3, were significantly elevated in APPSwDI/NOS2 -/-mouse model of AD compared to controls [5].Proteomic studies of microglial cells BV-2 showed that treatments with the anti-inflammatory drug polyphenol pentagalloyl glucose (PGG) downregulated GLOD4 while treatments with proinflammatory LPS/IFNγ upregulated GLOD4 [6].In addition to GLOD4, PGG downregulated septin-7, ataxin-2, and adenylosuccinate synthetase isozyme 2 (ADSS), which were linked in earlier studies to neurodegenerative diseases such as AD, Parkinson's, Huntington's, Down's syndrome, and frontotemporal dementia [6].GLOD4 was also identified among 38 duplicated genes in multiple cases of autism spectrum disorder (ASD) [7].
Methylglyoxal, a byproduct of glucose metabolism [10] responsible for protein glycation can causes misfolding [11].Glycated proteins easily aggregate, which can give rise to amyloid plaques, a hallmark of AD [12].Although human GLOD4 was suggested to take part in the methylglyoxal detoxification [5], this still is to be shown.A yeast two-hybrid screen showed that GLOD4 protein interacts with mitochondrial ADP-ribose pyrophosphatase NUDT9, which may regulate cell growth [13].
The human GLOD4 protein has 3 isoforms, differing in amino acid sequence (www.uniprot.org/blast/).The isoform 1 consists of 313 amino acids and contains motifs of the glyoxalase domain [9], antibiotic resistance domain, and a domain characteristic of the family of proteins resistant to the cytotoxic anticancer drug bleomycin (www.ncbi.nlm.nih.gov/protein/NP_057164.3).
Bleomycin hydrolase (BLMH) plays a key role in the CNS and is linked to AD.For example, BLMH can process AβPP to Aβ [14] and to further process Aβ [15].Enzymatic activities of BLMH are significantly downregulated in the human AD brain [16].Blmh −/− mice show astrogliosis, behavioral changes [17], and diminished ability to detoxify Hcy-thiolactone, which elevates brain Hcy-thiolactone level and increases its neurotoxicity [4].Proteomic studies of Blmh −/− mouse brain show that Blmh interacts with diverse cellular processes, such as synaptic plasticity, cytoskeleton dynamics, cell cycle, energy metabolism, and antioxidant defenses that are essential for brain homeostasis [18].
Here we examined the role of GLOD4 in the CNS by studying GLOD4 expression in AD brains.We also examined Glod4 expression in Blmh −/− 5xFAD mouse model of AD in relation to cognitive/neuromotor performance in these mice and biochemical consequences of Glod4 gene silencing in mouse neuroblastoma N2a-APPswe cells.Our findings show that GLOD4 is important for CNS homeostasis.

Human brain samples
Human frontal cerebral cortical tissues were obtained at autopsy by the Case Western Reserve University Brain Bank under an IRB-approved protocol, from clinically and pathologically confirmed cases of AD (n = 6) using criteria set up by the National Institute of Aging (NIA) and Consortium to Establish a Registry for Alzheimer's Disease (CERAD) [21].Brain tissues were also obtained from non-AD control patients (n = 6).Samples of frontal cortex, coded prior to shipment to keep laboratory personnel blinded as to diagnosis, were shipped on dry ice to the New Jersey Medical School and stored at−80•C until used.The AD patients were 84±8-year-old, 67% female, post-mortem interval was 8±5 h.Controls were 77±6-year-old (P = 0.100), 83% female (P = 0.549), post-mortem interval was 16±8 (P = 0.103).

Mice
Blmh -/-5xFAD mice, obtained by crossing Blmh −/− [22] and 5xFAD [23] animals, were characterized in an earlier study [20].Blmh and 5xFAD genotypes were established by PCR using primers listed in Table 1.Mice were kept at the Rutgers-New Jersey Medical School Animal Facility as previously described [20].5xFAD mice carry a transgene with the K670 N/M671 L (Swedish), I716V (Florida), and V717I (London) mutations in human AβPP(695), and M146 L and L286V mutations in human PS1 associated with familial early-onset AD. 5xFAD mice accumulate elevated levels of Aβ42 beginning around 2 months of age [23].Four groups of 5-month-old mice of both sexes were used in experiments: Blmh −/− , their wild type Blmh +/+ littermates, Blmh −/− 5xFAD, and their Blmh +/+ 5xFAD littermates.The groups were derived from multiple litters and equal number of females and males were used in each group.Animal procedures were approved by the Institutional Animal Care and Use Committee at Rutgers-New Jersey Medical School.

Behavioral testing
Cognitive and neuromotor performance was evaluated using the novel object recognition (NOR) [24] and the hindlimb [25] tests as previously described [20].Briefly, NOR evaluates recognition memory in two sessions.During the familiarization session, a mouse freely explores two identical objects for 20 s.
Six hours later, during the test session with one object replaced by a novel object, a mouse is allowed to explore for 20 s.The time spent exploring each object is calculated from video recordings.Cognitively healthy mice spend more time exploring a novel object, while cognitively impaired mice do not differentiate between novel and familiar objects.
The hindlimb test evaluates neurodegeneration.A mouse is suspended by tail and videorecorded for 10 s in three trials over three days.Hindlimb clasping is scored from 0 to 3: score 0, hindlimbs splayed outward and away from the abdomen; score 1, one hindlimb retracted inwards towards the abdomen for at least 50% of the observation period; score 2, both hindlimbs partially retracted inwards towards the abdomen for at least 50% of the observation period; score 3, both hindlimbs completely retracted inwards towards the abdomen for at least 50% of the observation period.The scores were averaged for the three trials.A higher score indicates worse neurodegeneration.

RNA isolation, cDNA synthesis, RT-qPCR analysis
Total RNA was isolated using Trizol reagent (MilliporeSigma).RNA concentration was measured using NanoDrop (Thermo Fisher Scientific).andintegrity was assessed by agarose gel (1%) electrophoresis.cDNA was synthesized using Revert Aid First cDNA Synthesis Kit (Thermo Fisher Scientific) according to manufacturer's protocol.RT-qPCR reactions (2 ng/µL cDNA, primers 0.3 µM each, iTaq 5 µL in a total volume of 10 µL) were performed in two technical replicates using SYBR Green Mix and the CFX96 Touch Real-Time PCR Detection System (Bio-Rad).RT-qPCR primer sequences are listed in Table 1.
For mouse Actb, Gapdh, Glod4_1, Glod4_2, Glod4_3 the following reaction conditions were The 2 (-ΔΔCt) method was used to calculate the relative expression levels [27].Data were analyzed using CFX Manager™ Software and Microsoft Excel.

Data analysis
Normality of distributions was evaluated with the Shapiro-Wilk's statistic.For variables with normal distributions, the results were calculated as mean±SD for variables with normal distributions or medians for non-normally distributed variables.A parametric unpaired t test was used for comparisons between two groups of variables with normal distribution.A Mann-Whitney rank sum test was used for comparisons between two groups of non-normally distributed variables.Data were analyzed using GraphPad Prism7 software (GraphPad Holdings LLC, San Diego CA, USA, https://www.graphpad.com)and MS Excell.

GLOD4 is downregulated in AD brains
Three isoforms of GLOD4 protein exist in humans: 34.8 kDa isoform 1 (GLOD4_1), 33.2 kDa isoform 2 (GLOD4_2), and 21.3 kDa isoform 3 (GLOD4_3).We quantified mRNAs for GLOD4 isoforms in frontal cerebral cortical tissues from brains of six AD patients and six non-AD controls by using RT-qPCR.
We found that GLOD4_1 mRNA level was the highest in control human brains; GLOD4_2 and GLOD4_3 mRNAs were present at 10-fold lower levels (Figure 1A).Similar pattern of GLOD4 mRNA isoforms expression was seen in AD brains (Figure 1A).However, relative levels of GLOD4 mRNA isoforms were significantly reduced in AD brains: from 10.31 to 3.83 (by 63%) for GLOD4_1 mRNA, from 1.00 to 0.51 (by 49%) for GLOD4_2 mRNA, and from 1.10 to 0.30 (by 73%) for GLOD4_3 mRNA (Figure 1A).GLOD4 protein, quantified by western blotting, was similarly downregulated in frontal cerebral cortex tissues from human AD brains.SDS-PAGE analyses revealed two bands reacting with anti-GLOD4 antibody in the cortical brain protein extracts: an upper band of 35/33 kDa corresponding to GLOD4_1 and GLOD4_2 and a lower band of 21 kDa corresponding to GLOD4_3, based on the molecular weight (Figure 1B).Quantification showed that GLOD4_1/2 protein level was significantly reduced in AD brains compared to controls (0.62 vs. 1.00) as was the level of GLOD4_3 protein (0.63 vs. 1.00) (Figure 1C).

Glod4 level is associated with Aβ and cognitive/neuromotor performance in Blmh −/− 5xFAD mice
To determine whether downregulated GLOD4 levels seen in AD patients are associated with hallmarks of AD, we quantified Glod4 mRNA level in Blmh −/− 5xFAD mice, a model of human AD; Blmh +/+ 5xFAD littermates were used as controls [20].The choice of the Blmh −/− 5xFAD mouse model [20] for studies of Glod4 was based on our previous findings showing that Blmh activity was reduced in human AD brains [16].Because Aβ starts to accumulate in 5xFAD mice at 2 months [23], this model allowed examination of a relationship between Glod4 expression, behavioral performance, and Aβ accumulation at an early age.Earlier analyses of gene expression in mouse liver showed that Glod4_1 mRNA is a major isoform while Glod4_3 is a minor isoform while Glod4_2 isoform was not detectable by PCR and that Glod4_3 mRNA isoform was not expressed Blmh -/-mouse liver [18].
Using RT-qPCR we found that Glod4_1 was the most abundant isoform and Glod4_3 was present at 25-fold lower level than Glod4_1 in the mouse brain (0.041 vs. 1.00,Table 2).The Glod4_2 mRNA isoform was essentially absent in the mouse brain, with an estimated level 1250-fold (0.0008 vs. 1.00) and 51-fold (0.0008 vs. 0.041) lower than those of Glod4_1 and Glod4_3 mRNAs, respectively (Table 2).  1. Six to seven-week-old Blmh +/+ (n=12) and Blmh -/-(n = 10) mice of both sexes were used in the experiments.Signals for Glod4_3 were observed in brain samples from all wild type Blmh +/+ mice and in two out of ten Blmh -/-mice.# Estimated from RT-qPCR measurements in duplicates with 10 samples containing 5-80 ng cDNA prepared from wild type Blmh +/+ mouse brain RNA.Signals for Glod4_2 were observed for 4 out of 10 samples.
Behavioral testing showed that Blmh −/− 5xFAD mice with lower Glod4 level also had neurological impairments characteristic of human AD.Specifically, these mice spent the same time exploring novel and familiar objects, i.e., did not differentiate between novel and familiar objects in the NOR test, which shows impaired recognition memory.In contrast, Blmh +/+ 5xFAD mice that had higher Glod4_1 level spent more time exploring a novel object, i.e., had normal preference for novelty (Figure 2C).
Glod4_1/3 mRNA was significantly downregulated also in Blmh +/+ 5xFAD mice compared with non-transgenic wild type Blmh +/+ animals (Figure 3A), although the difference was attenuated compared to the difference seen between Blmh -/-5xFAD and Blmh -/-mice.In contrast to Blmh -/-mice, the influence of the 5xFAD transgene on Glod4 mRNA expression in Blmh +/+ mice was sex-dependent: observed in males but not in females (Figure 3C).

Glod4 gene silencing affects APP and autophagy related Atg5 and p62 mRNAs in mouse neural cells
To elucidate the mechanism by which Glod4 can affect behavioral hallmarks of AD, we first examined whether our findings of downregulated GLOD4 expression in brains of AD patients (Figure 1A) can be recapitulated in mouse neuroblastoma cells N2a-APPswe carrying a mutated human AβPP transgene.We silenced the Glod4 gene by RNA interference using siRNA Glod4#1 and siRNA Glod4#2 for transfections; controls were transfected with scrambled siRNA (siRNAscr).Gene expression was examined by RT-qPCR using Actb mRNA as a reference.
We found that the Glod4 mRNA level was significantly reduced in Glod4-silenced cells (by 76% for siRNA Glod4#1 and by 98% for siRNA Glod4#2; Figure 4A).Glod4 protein was also significantly downregulated in the silenced cells (Figure 4B).The downregulation of Glod4 expression was accompanied by upregulation of AβPP mRNA (Figure 4C) and AβPP protein (Figure 4D).Representative images of western blots are shown in Figure 4E.
To find whether the effects of Glod4 silencing on autophagy could be mediated by mTOR signaling, we quantified mRNAs levels for mTOR and PHF8 (a positive regulator of mTOR expression) in Glod4-silenced and control cells.We found that mTOR and Phf8 mRNA levels were not affected by the Glod4 gene silencing (not shown).

DISCUSSION
Our findings suggest that Glod4, a protein with an unknown function, may have a protective role in the CNS.Specifically, we showed that (i) GLOD4 mRNA and protein were downregulated in human AD patients compared to non-AD controls; (ii) Glod4 mRNA was similarly downregulated in Blmh -/-5xFAD mouse model of AD, mostly in females; (iii) The 5xFAD transgene downregulated Glod4 in Blmh -/-mice of both sexes and in Blmh +/+ males but not females; (iv) reduced Glod4 expression was associated with worsened memory/sensorimotor performance in Blmh -/-5xFAD mice (v); Glod4 silencing in N2a-APPswe cells upregulated AβPP and downregulated autophagy-related Atg5, p62, and LC3 genes.These findings suggest that Aβ accumulation in GLOD4-depleted brains can occur via two pathways shown in Figure 4I.
In the first pathway, GLOD4 depletion upregulates AβPP, which increases Aβ accumulation leading to cognitive/neuromotor impairments.In the second pathway, GLOD4 depletion downregulates Atg5, p62, and LC3, which impairs autophagy flux and thus increases Aβ level and cognitive/neuromotor deficits.That impaired autophagy can increase Aβ accumulation mediated by AβPP is known.For example, Becn1, a protein initiating autophagy, is decreased in human AD brains while genetic reduction of Becn1 in transgenic mice that overexpress AβPP (APP + Becn +/-mice) increased Aβ in neuronal cells [28].However, as we did not find any effect of Glod4 depletion on Becn1 expression (not shown), Glod4 depletion upregulates AβPP and Aβ by a Becn1-independent mechanism.
AβPP level is an important determinant of AD, most strikingly manifested in individuals with Down syndrome who show an early AD neuropathology caused by an extra copy of the APP gene due to trisomy of chromosome 21 [29][30][31] .Transcriptional factors (SP-1, AP-1, CTCF, HSF1,NF-kB, USF, ApoE, androgen) and translational regulators (iron, IL-1, TGFβ, FMRP, hnRNP C, Rck/54, PSF/SFPQ) of AβPP expression have been identified [32].In previous work, we have shown that deletion of genes encoding enzymes participating in homocysteine metabolism in mice such as Cbs, Pon1, or Blmh elevated levels AβPP and Aβ in the mouse brain.In the present study, we found that depletion of a protein of unknown function, Glod4, in neural cells (Figure 4) and Blmh -/-5xFAD mice (Figure 1) was associated with upregulation of AβPP.These findings identify reduced level of Glod4 as a new factor that can influence the development of AD by upregulating AβPP and Aβ.The mechanism by which Glod4 can regulate AβPP and Aβ levels remains to be elucidated in future studies.
Our findings that Glod4 was depleted in Blmh −/− 5xFAD mice vs. Blmh +/+ 5xFAD sibling controls but not in Blmh -/-vs.sibling control wild type Blmh +/+ mice (Figure 3A) suggest that Glod4 depletion may be secondary to Aβ accumulation in Blmh −/− 5xFAD animals.However, to find out whether downregulation of Glod4 precedes or is secondary to Aβ accumulation, it would be necessary to examine Aβ and Glod4 levels in Blmh −/− 5xFAD animals at different time points.Our findings also suggest that Hcy-thiolactone and/or N-Hcy-proteins, which accumulate in Blmh-deficient mice [33] are unlikely to influence Glod4 expression.The mechanism of Glod4 depletion in mouse Blmh −/− 5xFAD brain and in human AD brain remains to be studied.
Although Blmh depletion caused Glod4 depletion in Blmh −/− 5xFAD mouse brain, downstream effects of Glod4 depletion leading to Aβ accumulation are different from those of Blmh depletion.
We found that major GLOD4_1 isoform, and minor GLOD4_2 and GLOD4_3 isoforms, were expressed in the human brain.We also found that all three isoforms were similarly attenuated in AD brains (Figure 1).In mice, Glod4_2 isoform was expressed at an extremely low level in wild type animals (0.0008 of the Glod4_1 expression level (Table 2) and that Glod4_1/3 isoforms were also downregulated in the Blmh -/-5xFAD mouse model (Figure 2 and 3).These findings suggest that GLOD4 isoforms are unlikely to have an isoform-specific function in AD and thus leave open a question of what the role of GLOD4 isoforms in brain physiology is.
In conclusion, our findings suggest that Glod4 may interact with AβPP and the autophagy pathway, and that disruption of these interactions upregulates Aβ, which causes cognitive/neurosensory deficits, thereby highlighting Glod4's role in brain homeostasis.Further studies are needed to assess causality of the association of GLOD4 depletion with AD.

Figure 1 .
Figure 1.Reduced expression of GLOD4 isoforms in brains of Alzheimer's disease patients.Isoforms of GLOD4 mRNA and protein were quantified in the frontal cortex samples from AD patients (n = 6) and non-AD controls (n = 6).(A) Bar graph shows GLOD4 mRNA quantification by RT-qPCR using ACTB mRNA as a reference.(B, C) GLOD4 protein isoforms were quantified by western blotting using GAPDH protein as a reference.(B) Representative image of a western blot.(C) Bar graph shows GLOD4 protein quantification.The experiment was performed twice.p-values were calculated by unpaired two-sided Student t test.

Figure 2 .
Figure 2. Reduced expression of brain Glod4_1/3 mRNA is associated with elevated Aβ and impaired cognitive/neuromotor performance in a mouse model of Alzheimer's disease.Five-month-old mice of both sexes were used in the experiments: Blmh -/-5xFAD (n = 22) and their Blmh +/+ 5XFAD siblings (n = 15).Bar graphs illustrate quantification of (A) Glod4_1/3 mRNA using Actb mRNA as a reference, (B) Aβ, and cognitive/neuromotor performance in (C) NOR and (D) hindlimb tests.Panels B, C, and D were reproduced with permission from [20].Data points for each mouse group represent mean ± SD of two to four independent measurements for each mouse.p-values for Glod4_1/3 mRNA and the hindlimb test were calculated by an unpaired two-sided Student t test.p-values for the NOR test were calculated by the paired two-sided Student t test.p-values for Aβ were calculated by one-way ANOVA with Tukey's multiple comparisons test.NS, not significant.

Table 1
Primers used for PCR or RT-qPCR

Table 2 .
Relative expression of Glod4 mRNA isoforms in the mouse brain * Mouse brain Glod4 mRNA isoforms were quantified by RT-qPCR using Actb mRNA as a reference and primers listed in Table