**4. Discussion**

The prevalence of neurodegenerative disorders is growing [2,38] in parallel to the urgency to find new compounds for the treatment of such diseases, in which oxidative stress is a common hallmark and has been suggested to play a causative role [39,40]. Unfortunately, the screening of drug leads and natural compounds to counteract neurodegeneration using 2D cell cultures often results in the unsuccessful translation of data to clinics. Neurons are strongly influenced by their immediate extracellular environment, and there is a grea<sup>t</sup> need to develop new culture systems that more faithfully reproduce the complexity of this milieu in vivo. Human 3D cell culture models are a good alternative to animal models [41,42]. In contrast to 2D cell cultures, 3D cell cultures do not overlook the physical interactions existing between cell–cell and cell–matrix and have a higher resemblance to the in vivo phenotype. Ideal scaffolds for neuronal tissue or disease modelling should exhibit suitable 3D architecture for in vitro manipulation, should facilitate cell adhesion while promoting neurites outgrowth, and have high biocompatibility [43]. Collagen type I is highly used as scaffold because of its abundance and ubiquity in most of the hard and soft tissues in the human body [44]. Porous collagen sponges have been used to grow various cell types in vitro [45] and collagen derived scaffolds have been widely used in neural tissue engineering for drug delivery [46]. Furthermore, the extracellular matrix (ECM) in nerves is mainly composed of type I collagen [47] and is a commonly used material in nerve tissue engineering [47] and for peripheral nerve regeneration [48]. ECM geometry and composition are well known to influence cell morphology and gene expression. It has been shown that SH-SY5Y cells extended longer neurites in 3D collagen I hydrogel cultures than in 2D cultures [26]. On this basis, we used equine native collagen, commercially available for clinical application, as sca ffold to support 3D cultures of di fferentiated SH-SY5Y cells.

Our aim was to study the neuroprotective activity of a combination of SF, EGCG, and PB in counteracting peroxide-induced damage in 3D cultures of di fferentiated SH-SY5Y cells.

Taking into account previous studies showing the protective e ffects of these compounds against oxidative stress [49,50], we decided to treat SH-SY5Y cells with specific concentrations of SF, EGCG, and PB to better mimic concentrations that could be measured in plasma after oral intake of the three compounds [51]. We selected a porous instead of a hydrogel sca ffold because it easily permits the removal of apoptotic blebs and dead cells by washing during medium exchange. The used collagenous sca ffold was found to be highly biocompatible since it supported the adhesion and proliferation of SH-SY5Y cells in the 3D environment.

Our data demonstrated that SEP co-treatment was significantly more e ffective against oxidative stress than the single treatments of PB, EGCG, or SF, suggesting a synergistic protective mechanism of the co-treatment. In particular, SEP was more e ffective in limiting cell injury induced by H2O2 exposure. These data were also demonstrated in our 2D cell model, and we confirmed the superior efficacy in enhancing GSH levels by SEP co-treatment compared the single treatments both in 2D and 3D models (Figure S3). Although di fferent reports have discussed the neuroprotective e ffect of PB, EGCG, and SF against brain-induced toxicity [52–54], there is no documented work on the e ffect of their combination. Our results are in agreemen<sup>t</sup> with other papers demonstrating the synergistic protective e ffect of di fferent combinations of natural compounds against neurodegeneration [55–57]. In general, the superior protection of co-treatments compared to the single treatments could be probably ascribed to the concurrent modulation of di fferent molecular targets involved in the pathogenesis and progression of these multi-factorial diseases.

To better elucidate the mechanisms behind SEP protection against H2O2 in SH-SY5Y cells, we investigated the e ffect of the co-treatment on the expression of di fferent antioxidant enzymes: heme oxygenase 1 (HO1), NADPH: quinone oxidoreductase 1 (NQO1), glutathione reductase (GR) and thioredoxin reductase (TR).

The enzyme HO1 converts heme to three end products, namely biliverdin, CO, and ferrous ion [58], then biliverdin reductase activity produces the antioxidant bilirubin. NQO1 is a highly inducible detoxifying flavoenzyme. It catalyzes the reduction of quinones generating stable hydroquinones and possesses superoxide scavenging activity [59]. GR is responsible for maintaining a storage amount of reduced glutathione [60]. The thioredoxin (Trx) system, composed of Trx, TR, and NADPH as a cofactor, is a cellular defense system that is ubiquitously involved in converting ROS to nontoxic metabolites [61]. In such a system, the Trx in reduced status can be oxidized into oxidized Trx during the degradation of H2O2 and then reduced by TR [62].

SF is known to upregulate antioxidant defense through the induction of HO1, NQO1, and GR in SH-SY5Y cells [15,63], while GR, TR, and NQO1 have been observed to be upregulated in cortical neurons [64]. EGCG induced HO1 expression in rat-cultured neurons [65] and increased protein levels in treated rats following focal cerebral ischemia [66]. PB treatment led to increased levels of HO1, NQO1, and TR in SH-SY5Y cells [17]. Interestingly, our data showed that SEP co-treatment, in the absence of oxidative stress, strongly upregulated these enzymes compared to control cells (the same results were also obtained in the 2D model (Figure S3)), and, in the presence of oxidative stress, it was able to significantly increase the expression of these enzymes compared to H2O2 exposed cells. In agreemen<sup>t</sup> with our data that showed the enhanced e ffect of a combination of di fferent compounds compared to treatment with the single compounds, in a previous work, we observed that a combination of EGCG and SF counteracts in vitro oxidative stress and delays stemness loss of amniotic fluid stem cells [67]. Moreover, a combination of berberine with resveratrol had enhanced hypolipidemic e ffects in high fat diet-induced mice and was able to decrease the lipid accumulation in adipocytes to a level significantly lower than that of the treatment with the single compounds [68].

Recently, it has been suggested that inhibition of the generation, rather than the scavenging, of ROS may be a more successful strategy to counteract oxidative stress-induced neurodegeneration [69]. ROS can be produced by many enzymes in the cells such as mitochondria respiratory complexes, NADPH oxidase, nitric oxide synthase, cytochrome 450, cyclooxygenase, lipoxygenase, and xanthine oxidase [14,70]. Interestingly, all these enzymes except NADPH oxidase produce ROS as a byproduct, while NOX enzymes generate ROS as a principal aim [71–73]. Moreover, di fferent studies have shown the involvement of NADPH oxidase family members in brain injury and neurodegenerative disorders (reviewed in [69]). Our data demonstrated that SEP co-treatment was e ffective in reducing NOX1 and NOX2 expression compared to control cells and was also able to counteract the increase of NOX1 and NOX2 expression induced by H2O2. This means that SEP not only potentiates the antioxidant defense system upregulating fundamental enzymes, but also reduces the intracellular production of ROS.

The last aim of this paper was to investigate if SEP could modulate other hallmarks of neurodegeneration in addition to its ability to counteract oxidative stress. We decided to focus our attention on IDE, the main protease responsible for amyloid β clearance [74–76]. A reduction of IDE activity in the brain with age and during the early stages of Alzheimer's disease (AD) has been observed [74], suggesting that IDE downregulation may be among the triggers of AD. Of note, SEP counteracted the strong downregulation of IDE induced by oxidative stress, maintaining IDE expression at a level comparable to control cells, suggesting a potential role of SEP in counteracting AD.

In conclusion, we highlighted that an appropriate synergistic combination of natural antioxidants such as SF, EGCG, and PB can help to rescue neuronal cells from oxidative stress cell death. The protective effect of the co-treatment was observed in a novel 3D model of SH-SY5Y cells that we developed. In agreemen<sup>t</sup> with other authors [77–79], we sugges<sup>t</sup> that a 3D culture system better mimics cell–cell interactions and cell–ECM interactions compared to the traditional 2D monolayer. In particular, our 3D model would be useful for future investigations of the neuroprotective activity of natural compounds [80]. In the present study, we observed the protective e ffect of an "acute" co-treatment with SF, EGCG, and PB but, taking into account the nature of neurodegeneration, a subchronic/chronic administration should be even more e ffective. For this reason, future studies will have to be carried out to investigate the e ffect of chronic SEP treatment against oxidative stress in neurodegeneration. The present findings underscore the importance of a combinatorial approach for e ffective treatments against oxidative damage in neurodegeneration. Moreover, 3D SH-SY5Y cell culture systems appear to be the ideal environment for in vitro assays regarding the e ffects of phytochemicals on cell viability.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2076-3921/8/10/420/s1. Figure S1: Visual explanation of MTT assay as used in this study, Figure S2: Cell viability of SH-SY5Y incubated with H2O2, Figure S3: Results obtained with SEP co-treatement in SH-SY5Y 2D model. Figure S4: Images of SH-SY5Y 3D model.

**Author Contributions:** Conceptualization, P.M. and C.A.; methodology, P.M. and C.A.; data curation, P.M. and C.A.; writing—original draft preparation, P.M.; writing—review and editing, C.A. and S.H.; funding acquisition, S.H.

**Funding:** This work was supported by MIUR-PRIN 2015 (N. 20152HKF3Z) to S.H.

**Conflicts of Interest:** The authors declare that there is no conflict of interest regarding the publication of this paper.
