Arabidopsis thaliana extracts optimized for polyphenols production as potential therapeutics for the APOE-modulated neuroinflammation characteristic of Alzheimer’s disease in vitro

Although the cause of Alzheimer’s disease (AD) is unknown, glial-induced neuroinflammation is an early symptom. Familial AD is caused by increases in amyloid-beta (Aβ) peptide, particularly soluble oligomeric (oAβ), considered a proximal neurotoxin and neuroinflammatory stimuli. APOE4, a naturally occurring genotype of APOE, is the greatest genetic risk factor for AD; increasing risk up to 12-fold compared to APOE3 and APOE2. oAβ-induced neuroinflammation is greater with APOE4 compared to APOE3 and APOE2. As sinapates and flavonoids have anti-inflammatory properties, a protocol was developed for optimizing polyphenol production in seedlings of Arabidopsis thaliana (A. thaliana). Three mutants (cop1, prn1, xpf3) were identified, and the extracts treated with liver microsomes to mimic physiological metabolism, with HPLC and MS performed on the resulting metabolites for peak identification. These extracts were used to treat primary glial cells isolated from human APOE-targeted-replacement (APOE-TR) and APOE-knock-out (KO) mice, with neuroinflammation induced by lipopolysaccharide (LPS) or oAβ. The dose-response data for TNFα secretion demonstrate the followed the order: APOE-KO > APOE4 > APOE3 > APOE2, with xpf3 the most effective anti-neuroinflammatory across APOE genotypes. Thus, the plant-based approach described herein may be particularly valuable in treating the APOE4-induced neuroinflammatory component of AD risk.


Results
UV-C survival screen for A. thaliana mutant seedlings with increased polyphenol production. We screened 6 day (d)-old seedlings of specific mutants of A. thaliana and identified 3 mutants that survived UV-B irradiation comparable to wild type (wt) (Fig. 1A, top) and survived UV-C irradiation significantly better than wt and the adt3 mutant, a mutant designed for reduced phenylalanine levels, thus the negative control (Fig. 1A, bottom). These responses are indicative of effective polyphenol accumulation or activity 11,73 . The mutants are: 1. Pirin1 (transcriptional co-factor and quercetinase; prn1); 2. constitutive photomorphogenic1 (E3 ubiquitin-protein ligase, acts as a repressor of photomorphogenesis; cop1); and 3. excision repair homologue 3 (homolog of the human xeroderma pigmentosum group F DNA repair; xpf3).
The CellRox assay was used to assess oxidative stress, an additional measure of enhanced polyphenols in the mutants, as polyphenols can be potent scavengers of reactive oxygen species (ROS). Post UV-C irradiation, epidermal optical slices (1 μ m thick) were stained for the CellRox assay with flourescence by Cy5 indicating ROS levels (670 nm, pink), while DAPI (358 nm; blue) and FITC (448 nm; green) capture the natural fluorescence of phenylpropanoids 11 . Representative images were captured on a deconvoluting microscopy for wt, wt + UV-C and xpf3 + UV-C (Fig. 1B, top). Quantification of fluorescence for Cy5 (650 nm, pink, ROS) and DAPI + FITC (359-494 nm; green-blue, polyphenol levels) are represented as bar graphs (Fig. 1B, bottom). With UV treatment, the xpf3 mutant had significantly reduced signs of oxidative stress compared to wt. The xpf3 mutant seedlings also produced significantly more phenylpropanoids than wt.
Using principal components analysis (PCA), independent batches of seedling extracts clustered based on mutations (Fig. 2). For PCA of the mutants xpf3, prn1, adt3, cop1 and wt samples, using relative content of peaks from UV spectroscopy (under UV 254 nm , UV 280 nm and UV 320 nm ) as variables, demonstates that the samples formed their own cluster within a small region significantly seperated from each other. The exception was the similarity of Scientific RepoRts | 6:29364 | DOI: 10.1038/srep29364 prn1 to wt as these 2 clusters overlapped in 2-D mapping. In addition to distinguishing among the mutants, PCA also indicates the batch-to-batch consistency of a specific mutant based on the spread of the cluster.
Identification of extractable constituents of A. thaliana mutant xpf3 extract, pre-and post-treatment with liver microsomes. HPLC revealed 13 identifiable peaks from 10 samples of each mutant extract. A sample trace for xpf3 extract (Fig. 3) pre-(lower) and post-(upper) treatment by murine liver microsomes (LM) used in order to mimic aspects of metabolism. LM treatment induced a quantitative increase in most of the 13 identified constituents, particularly peak 7, barely visible in the untreated extract (lower trace). The contents of the peaks were analyzed by HPLC-DAD-MS/MS (described in Methods), and structures confirmed by NMR (Table 1).
Plant extracts dose-dependently attenuate APOE-modulated LPS-induced neuroinflammation (APOE-KO > APOE4 > APOE3 ≥ APOE2). LM-treated mutant extracts were used to treat the primary mixed glial cultures from APOE-TR or APOE-KO mice consisting of ~95% astrocytes + 5% microglia, the approximate proportion found in both human and mouse brain. In this human apoE-relevant in vitro model, we measured the amount of TNFα released into the media, based on our previous work demonstrating that the levels (a) 6d-old seedlings of specific A. thaliana mutants quantified 24 h after sub-lethal (3 min) UV-B (top) or lethal (8 min) UV-C radiation (bottom) demonstrate that mutants cop1, prn1 and xpf3 survive a dose of UV-C that is lethal to wt and the adt3 negative control. The number of surviving seedlings for the mutants and wt, for a sub-lethal dose (top) and a lethal dose (bottom); n = 3, expressed as means ± SEM, *vs. wt. p < 0.05 via paired student's t-test (two-tailed) using GraphPad Prism 5. (b) Post UV-C irradiation, epidermal optical slices (1 μm thick) were stained for the CellRox assay with flourescence by Cy5 indicating ROS levels (670 nm, pink,), while DAPI (358 nm; blue) and FITC (448 nm; green) capture the natural fluorescence of phenylpropanoids 11 . Representative images were captured on a deconvoluting microscopy for wt, wt + UV-C and xpf3 + UV-C (Fig. 1B, top). Quantification of fluorescence for Cy5 (670 nm, pink, ROS) and DAPI + FITC (359-494 nm; green-blue, polyphenol levels) are represented as bar graphs (Fig. 1B, bottom). With UV treatment, the xpf3 mutant had significantly reduced signs of oxidative stress compared to wt. The xpf3 mutant seedlings also produced significantly more phenylpropanoids than wt. Data expressed as means ± SEM, for n = 20, * vs. wt and wt + UV-C, p < 0.05 via paired student's t-test (two-tailed) using GraphPad Prism 5. of this cytokine are elevated significantly, likely due to the feedback loop with activated astrocytes stimulating microglia of release of TNFα . Thus, TNFα levels were a measure of the relative anti-neuroinflammatory properties of the mutant A. thaliana extracts. These cultures are routine in our lab 16,18,[74][75][76][77][78][79][80][81][82][83] .
The cultures were pre-treated for 1 hour (h) with LM-digested wt, prn1, cop1, xpf3 and adt3 plant extracts (log scale dilution: 1:5000, 1:1000, 1:500, 1:200 and 1:100) prior to LPS (100 ng/ml) treatment for 16 h to elicit a neuroinflammatory response. As these are immune cells, their response to stress (LPS or oAβ ) is to become activated, as evidenced by the often observed change in morphology 16,18,[74][75][76][77][78][79][80][81][82][83] , as well as the measured secretion of TNFα. An MTT assay was preformed for cell viability, specifically mitochondral function, in the mixed glial cultures isolted from APOE4-TR mice, the most vulnerable genotype (Fig. 4). MTT, expressed as %control for LPS or oAβ, did not change significantly in cells treated +/− the A. thaliana mutant extracts at the highest dose, the 1:100 dilution. TNFα levels in the media were measured in duplicate by ELISA, and expressed as %APOE3 LPS-vehicle control (VC) (Fig. 5). The inset is an enlarged graph that more clearly shows the separation between APOE3 and APOE2. The LPS-induced inflammatory response is dose dependent and APOE-modulated (APOE-KO > APOE4 > APOE3 ≥ APOE2) with the greatest change in APOE-KO over the dilution range. At all dilutions, xpf3 was effective in attenuating LPS-induced TNFα secretion and, as expected, adt3, was the least effective extract. Thus, the data in this figure is based on duplicates for a full dose-response experiment with LPS treated with 4 mutant + wt plant extracts in cultures from 4 APOE genotypes.

Discussion
The studies described herein were designed to test the potential of plant-based materials for the prevention of chronic inflammation as the A. thaliana plant extracts were added to the cultures prior to the treatment with LPS or oAβ . Flavonoids can play critical roles in reducing negative effects of oxidative stress in plant cells, but also act as potential anti-inflammatories for human use [84][85][86][87] . A. thaliana is a genetic model with an established phenylpropanoid synthesis pathway that can be manipulated, and developed for specific chemical applications. Herein, we identified A. thaliana mutants, and extractable compounds that may be useful for anti-inflammatory    amounts following oral administration 90 . In the current in vitro study, plant extracts were treated with LM to obtain metabolites potentially active in vivo. LM contains enzymes like cytochrome P450s (CYP) that metabolize the plant extracts via hydroxylation and oxidative methylation [91][92][93] . Hence, the metabolic activation by LM mimic  aspects of physiological ingestion and metabolism, and produce metabolites that may survive in vivo. This is important as some compounds lose their activity post-liver processing and metabolism, while others may assume an active form. These data provide evidence that enhancing the anti-inflammatory properties of A. thaliana via polyphenols and associated molecules have potential as dietary supplements as part of a therapeutic approach to neurodegenerative disorders 1,2 . Furthermore, incorporating analogous mutations into consumable plants could benefit patients at a higher risk for AD. Acylation of several phenylpropanoids is typically seen in plant metabolism, dependent upon environmental cues 94,95 , and produces many compounds of potential importance to medicine, which may have been promoted by our growth methods in young seedlings to promote a stress-resistant (survive UV-C) chemotype. Chemicals like 1,2 di-O-sinapoyl-β -glucose and sinapoyl malate, known for UV-inducible responses 96 , were enhanced and accumulated in the seedling tissue under our methods, particularly in xpf3 mutants. As well, the N,N'-di-sinapoyl spermidine, while a minority chemical in an already potent flavonoid-based extract, was elevated post LM treatment in specifically xpf3. Thus, the chemical analysis correlates with the function of the mutants, as mutant xpf3 had the greatest anti-inflammatory effect, while adt3 was the negative control with impaired polyphenol production. N,N'-di-sinapoyl spermidine is a hydroxycinnamic acid conjugated to a polyamine, referred to as hydroxycinnamic acid amide (HCAA), accumulates in A. thaliana seed 97 . HCAA have been implicated in oxidative stress defence in plants and utilized as a potential antioxidant, chemotherapeutic agent, and in a cardiovascular disease model 98 . This type of chemical is anti-inflammatory in human monocytes (clovamide) and mouse microglia (clovamide derivate) 99,100 , and neuroprotective in vitro 101 , suggesting a potential to reduce chronic inflammatory responses. PRN1 specifically cleaves quercetin and it is possible that while prn1 mutants accumulate more phenylpropanoids, this increase may only be in very specific quercetin(s) according to the PCA and biological data 11 . It is currently unknown how changes in polyphenol ratios can influence inflammation, but individivual peaks are being investigated.
An important and fundamental question regarding apoE is whether the presence of apoE4 causes a loss of positive function or a gain of toxic function. For development of future therapeutics, it is critical to understand whether apoE4 represents loss of positive function or gain of toxic function. If apoE4 is a gain of toxic function, then apoE4 will induce a detrimental response compared to a lack of apoE, experimentally defined as a functional improvement in APOE-KO mice compared to APOE4-TR mice. If apoE4 is a loss of positive function, the prediction is a neutral or improved response in APOE4-TR mice compared to APOE-KO mice. Thus, a comparison between apoE4 and apoE3 for a given function is not sufficient to determine if apoE4 is worse vs. not as effective compared to apoE3. The necessary three-way comparison is APOE-KO vs. APOE4 vs. APOE3. In this culture model, both LPS and oAβ induce neuroinflammation that is APOE-modulated, with TNFα secretion: APOE-KO > APOE4 > APOE3 ≥ APOE2, suggesting that APOE4 confers a loss of positive function. This is consistent with previous studies of both LPS and oAβ -induced nueroinflammation in vivo 55 and in vitro data 18 .
The effect of the apoE isoforms on activation of glial cells can vary based on the cellular composition of the culture model. In vitro data from astrocytes demonstrate that LPS-induced secretion of proinflammatory cytokines including TNFα , IL1β and IL6 is APOE-modulated, with the levels of the cytokines higher in APOE2, compared to APOE3 and APOE4 102 . In contrast, levels of proinflammatory cytokines from LPS-induced microglial cultures are higher in APOE4 cultures, compared to APOE3 and APOE2 103 . Mixed glial cultures serve as a physiologically relevant model as it includes the individual activation of the microglia and astrocytes, as well as the paracrine-like signaling between these two major immune cells in the brain. Indeed, studies in mixed glial cultures demonstrate that LPS-induced neuroinflammation is APOE-modulated (reviewed in ref. 56), with levels of proinflammatory markers (TNFα , IL1β , NO, IL6, etc.) higher in APOE4 than APOE3 18,[78][79][80][81] . In addition, mixed glial cultures from APOE-KO mice showed higher LPS-induced NO production compared to glial cells from APOE3-TR mice 104 . Importantly, in microglia/neuron co-cultures, LPS-induced loss of synaptic protein follows the order: APOE4 > APOE3 > APOE2, recapitulated in hippocampal slice cultures 56 .
One pathway through which APOE4 may modulate neuroinflammatory pathways is via TLR4-dependent signaling. Indeed, our published data in APOE/FAD-Tg mice demonstrate that microgliosis and astrogliosis are greater with APOE4 compared to APOE3 105 . In addition, inflammatory mediators, as a result of TLR4 activation, are increased with APOE4 compared to APOE3 18 . Complementing these in vivo findings, our in vitro data reveal that stimulation of TLR4 in mixed glial cultures with LPS or oAβ induces a similar neuroinflammatory response, as characterized by secretion of TNF-α , an effect more pronounced with APOE4 than APOE3 18 . Importantly, both LPS-and oAβ -induced responses are inhibited by LPS-RS. Recent in vitro data also demonstrate potent TLR4 antagonism of oAβ -induced responses using the synthetic small molecule IAXO (Innaxon), developed at the University of Milano-Bicocca, and shown to inhibit innate or auto-inflammatory processes [106][107][108] .
These data can be used to further understand how to potentiate polyphenols in seedlings, and aid in the quest to develop plant-based materials suitable to prevent chronic inflammation. Furthermore, by studying the further actions of the phytochemicals, it may be possible to identify their targets and understand how they may exert their actions, as has been done for individual flavonoid chemicals in vitro 109 . Importantly, nutraceuticals are an inexpensive and effective option that could be potentially used to prevent neuroinflammation, hence delaying the onset of AD, specifically for APOE4-carriers, who are at a higher risk of both AD and non-AD APOE4-associated neuroinflammatory processes, including recovery from head injury, cerebral haemorrhage, stroke and cerebral artery amyloidosis.

Conclusions
We developed a protocol for growth and extraction of high-yielding polyphenols from A. thaliana mutant seedlings. An in vitro screen using a human apoE-relevant primary glial cell model demonstrated that mutant xpf3 (150 μ g/ml) has the highest anti-inflammatory effect, as measured by TNFα lelevs in the media, compared to wt (150 μ g/ml). Analysis of the ethanol extractable mutant-seedling chemicals indicated that xpf3 possessed the highest polyphenol content per tissue mass, and had a specific polyphenol profile. In addition, the findings demonstrate that APOE4 imparts a loss of positive function, compared to APOE-KO. Furthermore, oAβ induced an inflammatory response, similar to LPS, that was attenuated by LPS-RS, suggesting this response was via the TLR4 pathway. Together, this study demonstrated the use of extracts from mutant A. thaliana grown to maximize the polyphenol content as a model to identify plant targets that optimize the preventative strategy for AD patients, particularly for those at high risk for AD from the APOE4 allele.  110 . The mutant lines are homozygous null for the reported insertions. Plants intended for seed stocks were grown as described 11 . For this study, three mutants of A. thaliana (xpf3, prn1 and cop1) were selected based on UV-C survival (indicating polyphenol production). adt3 was used as a negative control (produces low levels of phenylpropanoids 12 .

Plant growth conditions for experiments. Seeds of A. thaliana wt or T-DNA insertion mutants were
prepared, sown, and grown (seedlings are 6-7d) on 0.5 X Murashige and Skoog media 0.8% agarose phytatrays, then sealed in black Plexiglas boxes as described 11 . Cold-vernalized phytatrays were then moved to continuous darkness (Dc) at 20 °C for 6d 11 . Dc seed sets were exposed to two sublethal UV treatments of 254 nm for 3 minutes (min), 315 nm (4 min) (radiation sources described in ref. 73), with 1 min between treatments, then phytatrays returned to darkness for 12 h. Sets of live seedlings were harvested immedately after irradiations for oxidative stress studies (see oxidative stress assays), or quick frozen in liquid nitrogen for extraction. For seedling survival assays (sublethal UV-B dose and lethal UV-C dose), seedlings were treated with 254 nm as described in 11 , and survival was determined by those seedlings remaining standing and growing 24 h post-treatment, as described 73 . Those seedlings that "lodged" or fell over on the plate are characterized by major damage to the cells of the shoot and cotyledons and by 24 h and later, do not continue growing 73 . Oxidative stress assays. 6d-old seedlings received UV treatments (described above) or control (no UV treatment). Following the treatment, live seedlings were immediately harvested into a solution of the cell-permeable fluorescent probe CellRox Deep Red TM , with the ROS (Life Technologies) reagent dissolved as directed into anhydrous DMSO, for 30 min incubation under rotation (50 rpm/min rotation) in darkness. Plant material was washed with PBS for 3 mins (four times) at room temperature then mounted without fixation for microscopy. The fluorescent signal from CellRox Deep Red was detected by CY5 LED (see Microscopy), and both DAPI and FITC Ex/Em data were collected for phenylpropanoid flourescence.
Scientific RepoRts | 6:29364 | DOI: 10.1038/srep29364 Microscopy. Seedlings were viewed on a Zeiss Observer.Z1 (ApoTome 2) deconvoluting microscope. Images were collected with XCite 120 LED/Lumen Dynamics filter sets (CY5, DAPI, FITC) managed by Zen pro software (2012). Images were collected at the same exposure (control and experimental samples). At least 20 seedlings were viewed per experimental replicate. In post-processing software, fluorescence intensity was measured from at least 10 cells per cotyledon of a representative 5 imaged.

Plant materials extraction.
Aerial portions ground in liquid nitrogen were weighed, and extracted with 95% aqueous ethanol by sonication (20 ml of solvent per gram of plant material, performed twice, 2 h each treatment). After filtration, the filtrate was evaporated to remove solvent using a rotavapor under reduced vacuum, then freeze-dried. The extract was then resuspended in sterile DMSO at a constant concentration (150 mg/ml) across the plant mutant samples.
Liver microsomal treatment. The plant extracts from A. thaliana mutants were digested using mouse LM (Gibco, Life Technologies, Grand Island, New York) to mimic mouse ingestion, as described 111 . Briefly, an 'NADPH generating system' was established to provide the NADPH required by the P450 enzymes in the LM. The 200 μ l reaction mixture (10 μ l LM, 20 μ l plant extract (1:1), 1.2 μ l MgCl 2 (1 M), 6.6 μ l NADP (10 mM), 6.4 μ l glucose-6-phosphate (250 mM), 4.0 μ l glucose-6-phosphate dehydrogenase enzyme (500 U/ml) in 152 μ l HEPES buffer) was incubated at 37 °C for 30 min, followed by incubation on ice for 30 min to stop the reaction, and then centrifuged (2800 rpm, 10 min, 4 °C). The supernatant consisting of the plant extract metabolites (1:10, 15 mg/ml) potentially active in vivo were used for the in vitro treatments/screen. DMSO diluted in HEPES buffer was treated with LM, and the supernatant used as LM-vehicle control (LM-VC). The LM-VC was also used as the dilutent for the plant extract dilutions.

HPLC and Mass Spectrometry (MS).
Ten replicates of each mutant and wt were assessed by absorbance spectrum and by HPLC for prominent peaks (before LM treatment), to evaluate reproducibility of independently grown wt and mutant samples. Data were based on the HPLC chromatogram under UV at 320 nm. Average values from 10 batches of each plant genotype were reviewed to identify peaks. Further analysis was done by MS/MS2, particularly for xpf3, the most effective mutant at higher dilution. xpf3 samples before and after LM treatment were analyzed by HPLC-DAD-MS. Post-LM treatment, plant extracts were evlauated by MS. A new peak identified in xpf3 samples was further evaluated by LC-MS/MS 2 . A Shimadzu UFLC system coupled with a LCMS-2020 detector, and an HPLC column (Agilent C 18 column, 3 × 150 mm, 2.7 μ m) were used for all HPLC-DAD-MS analyses. Acetonitrile (A) and 0.1% formic acid in water (B) were used for mobile phase and run under the gradient program as follows. The concentration of A was kept at 5% for the first min, increased to 15% for the second min, 20% for the following 13 min, 40% for 7 min and finally to increased to 100% over 5 min (then maintained a further 8 min). Flow rate, detection wavelength, and column oven temperature were set at 0.4 mL/min, 320 nm, and 27 °C, respectively. The LC-MS/MS 2 analyses were performed on a Shimadzu LCMS-IT-TOF Mass Spectrometry.
LPS and oAβ treatment. Secondary glial cells were seeded into 24-well tissue culture plates at a density of 1.44 × 10 6 cells per plate. After 24 h, cells were washed with PBS to remove serum, and incubated in serum-free α -MEM media for an additional 24 h prior to treatment. The LM-digested plant extracts (xpf3, prn1, cop1, adt3 and wt) or LM-VC were added to the glial cultures at the indicated dilutions, ranging from 1:100 (150 μ g/ml) to 1:5000 (3 μ g/ml). After 1 h, LPS (100 ng/ml, Sigma Aldrich) or oligomeric preps of Aβ 42 (10 μ M), prepared as described 112 , were added to the glial cultures to induce inflammation. LPS-RS (LPS from Rhodobacter sphaeroides, 10 μ g/ml, Invivogen), a commercially available TLR4 antagonist 113 , was used as a positive control. After 16 h of incubation, conditioned media from the 24-well plates was removed, flash frozen and stored at − 80 °C. TNFα levels in the media were measured using a mouse TNFα ELISA (Invitrogen), following the manufacturer's protocol. The TNFα levels are expressed as %LPS VC, or oAβ VC of each APOE genotype relative to the APOE3 VC.
MTT Assay. The MTT assay, a measure of cell viability that specifically measures mtichondrial function, was performed on mixed glial cultures isolated from APOE4-TR mice, the most vulnerable APOE genotype. Cultures were treated with LM-digested plant extracts (1:100, 150 μ g/ml) and an inflammatory stimuli (LPS or oAβ ). Manufacturers protocol was followed for the assay (Catalog No CT02, EMD Millipore, Massachusetts). Briefly, the conditioned media from mixed glial cultures was removed and 0.2 ml of the MTT solution (10% MTT in serum free media) was added to the cells. After 4 h incubation at 37 C, the media was gently removed and 0.5 ml isopropanol was added to each well. The absorbance was measured at 570 nm. The absorbance values are expressed as % of non-treated cells (% control) for LPS or oAβ . Statistical analysis. Statistical significance was determined using GraphPad Prism version 5 for Macintosh.