The SH-SY5Y cell line in Parkinson’s disease research: a systematic review

The most-reported source for access to the SH-SY5Y cell line is the American Type Culture Collection (ATCC, CRL-2266, deposited by JL Biedler). Other sources concern retrieval from other cell banks, such as the European Collection of Authenticated Cell Cultures (ECACC, Catalog number: 94030304, deposited by PFT Vaughan) or the German collection of Microorganisms and Cell Cultures (DSMZ, ACC 209). Cells were also obtained through gifts from colleague scientists (Fig. 1). However, in 455 out of 962 publications the cell origin was not specified.

Recommendations regarding the composition of the growth medium for propagation of SH-SY5Y cells vary among the various cell line distributors: ATCC recommends MEM/F12 supplemented with 10% fetal bovine serum, ECCAC recommends MEM/F12 with 2 mM glutamine, 1% non-essential amino acids and 15% fetal bovine serum, and DSMZ recommends MEM plus 15–20% fetal bovine serum. In the actual protocols employed in the PD-related publications, DMEM was used most (434 out of 962 publications), followed by DMEM/F12 (230 out of 962 publications), MEM/F12 (68 out of 962 publications), DMEM high glucose (46 out of 962 publications), RPMI 1640 (37 out of 962 publications), Cosmedium-001 (21 out of 962 publications) and MEM (20 out of 962 publications) (Table 1). Furthermore, media used were supplemented with antibiotics/antimycotics (65.6% of the articles), glutamine (23.9% of the articles), non-essential amino acids (9.8% of the articles), sodium pyruvate (6.3% of the articles) or other components, such as HEPES, sodium carbonate, uridine or L-lysine, in different combinations (Table 1; for more detailed information of the media composition of each article, see Additional file 2). Careful choice of medium type and composition is crucial, e.g. the use of DMEM or RPMI changed the metabolome and the differentiation capacity of a number of cell lines [12, 13]. Evidence is now accumulating that nutrient availability, and also the degree of oxygenation, affects wiring through different metabolic pathways and that the intracellular levels of glutamine, alpha-ketoglutarate, pyruvate and NAD+/NADH redox ratio and concentration are determining factors in the epigenetic control of gene expression during differentiation, e.g. via effects on histone-lysine demethylases and DNA demethylases [14]. Furthermore, supplementing media with sodium pyruvate has been shown to be protective against oxidative stress [15, 16] and as such may affect the outcome of experiments that study the involvement of oxidative damage in PD. In 80% of the articles, the media described were supplemented with 10% fetal bovine serum, although some authors use other concentrations of serum, ranging from 5 to 20%, or serum from other species, such as horse (Table 1). Taking into account that serum includes growth factors, hormones, amino acids and lipids that can influence cell growth and differentiation, the use of different serum concentrations, serum from different species or even serum from different batches may influence the outcome [17]. It is of note here that variability in cell growth and sensitivity to various compounds depending on the media and substrate used for the growth of SH-SY5Y cells has already been reported [18].

To appropriately acknowledge the various effects that differences in culture media composition can have on the cellular phenotype, it is thus imperative to systematically identify in future studies how various metabolic intermediates, ions, serum constituents and substrates influence aspects of growth and differentiation of SH-SY5Y cells. Only then the protocols for experiments with PD cell models can be widely standardized.

The use of the SH-SY5Y cell line is not restricted to PD-research; this cell line has also been used in other areas of neuroscience, including research on Alzheimer’s disease, neurotoxicity, ischemia or Amyotrophic Lateral Sclerosis, among others [11, 19, 20]. To obtain derivative cells with a neuronal phenotype, multiple differentiation protocols have been described [21‐25], but details about the final population of cells, regarding the fate-choices and fate-specification of cells that undergo terminal differentiation, have not been systematically reported. Since PD is characterized by the death of SN DAergic neurons, the degree to which this cell line displays a DAergic phenotype is a key aspect regarding the validity of the model. In this respect, 392 out of the 962 papers state that SH-SY5Y cells have a DAergic phenotype without actually showing supporting evidence. Only a few papers cite previous work showing the DAergic phenotype. Another large proportion of articles (432 out of 962) does not provide any statement about the DAergic phenotype or the rationale behind the choice of the cell line for use in PD-research. Among the remaining publications, in 76 papers the SH-SY5Y lineage is represented as a PD-model with DAergic properties or a toxin-induced PD-like phenotype, another 56 publications report the analysis of the DAergic phenotype, and in 7 papers cholinergic, neuronal or noradrenergic phenotypes are mentioned (Table 2).

The phenotype of SH-SY5Y cells can be manipulated by inducing different programs of terminal neural differentiation. However, in 81, 5% of the published studies no differentiation regime was used (Fig. 2), for which in only seven publications a reason was given. Among the studies that do report on forced differentiation, the most common method employed is the addition of retinoic acid (RA) in concentrations ranging from 5 μM to 100 μM, for a period of time from 24 hours to 21 days, and, sometimes, a reduction of the concentration of serum in the media (Fig. 2). It has been reported that RA treatment upregulates expression of neuronal and DAergic markers and increases susceptibility to DAergic neurotoxins [26]. However, other studies have observed increased neuronal markers upon RA differentiation, but no change in DAergic markers and decreased susceptibility to DAergic neurotoxins [27]. The phenotypic effect of RA on SH-SY5Y cells has been systematically studied, including the induction of a terminal neural phenotype with, specifically, a DAergic-like character [28]. Conversely, RA-mediated differentiation of SH-SY5Y cells has been associated with the induction of a cholinergic rather than DAergic phenotype [29]. Here it is important to note that RA has been found to partially protect SH-SY5Y cells against proteasome inhibitors [30]. In view of this finding, the results of studies examining proteasomal dysfunction and involving RA-differentiated SH-SY5Y cells as PD-model should be interpreted with care. The second method of choice to differentiate SH-SY5Y cells is a sequential treatment with RA, usually 10 μM, and 12-O-Tetradecanoylphorbol-13-acetate (TPA), mostly added in a concentration of 80nM (Fig. 2). This protocol has been demonstrated to differentiate SH-SY5Y cells more efficiently to DAergic-like neurons [31‐33]. Early studies on the use of RA and TPA (alone or in combination) to differentiate SH-SY5Y cells have shown that these compounds induce various neuronal-like populations, with a strong increase of NA content when using only TPA [7]. In view of these differences, it is important to realize that a set of neurons each synthesizing a separate neurotransmitter (s) has a distinct transcriptional profile [34]. Even neurons synthesizing a specific neurotransmitter can be classified into several subpopulations, each with a clearly defined signaling function in a particular (brain) region and an explicit vulnerability for stress factors [35]. The third approach that is commonly used for differentiation induction involves the sequential treatment with RA, usually 10 μM, and 10-100 ng/mL of brain-derived neurotrophic factor (BDNF) (Fig. 2). This procedure leads to a homogeneous neuronal population with expression of neuronal markers and decreased proliferation [21]. The phenotypic outcome of this RA/BDNF differentiation protocol is, however, still somewhat controversial as it has been described as sympathetic cholinergic, based on evidence from target-directed qPCR and microarray studies which pointed into the direction of increased levels of acetylcholine transporter, choline acetyl transferase and neuropeptide Y [36, 37], but also as dopaminergic by others [38]. Moreover, inhibition of cell growth has not always been replicated when employing this procedure [24]. Additional protocols used for differentiation may involve combinations of the above-mentioned methods, or a combination of 10 μM RA and 0.3-5 mM dibutyryl cyclic adenosine monophosphate (dbcAMP) [39, 40], or of 10 μM RA for 3 days and 80nM tissue plasminogen activator [41] or the protocol was not specified. Differentiation may also be caused by 200 ng/mL growth/differentiation factor 5 (GDF5) [42], recombinant bone morphogenetic protein 2 (BMP2) [42], staurosporine [43, 44] or 50 ng/mL glial cell line-derived neurotrophic factor (GDNF) [45]. The pros and cons of the differentiation of the SH-SY5Y cell line to obtain a relevant model for PD have been reviewed more extensively elsewhere [46]. Again, proper characterization of the differentiated cells is crucial. Next to the more conventional “ensemble” approaches, such as western blotting, transcriptomics or proteomics, the use of novel single-cell microscopy and single-cell RNAseq approaches will become instrumental to provide more detailed phenotypic profiling of the differentiated cell population [47, 48].

In view of the relevance of the DAergic system in PD, it is striking that among the papers using SH-SY5Y cells in PD research only 55 out of 962 examine the DAergic phenotype of the cell line (Table 2; for a list and details of the articles that checked the DAergic phenotype see Additional file 3). Remarkably, in some cases the DAergic phenotype has been studied but the results were not shown. The methods used to determine the SH-SY5Y DAergic phenotype include immunocytochemistry (ICC), western blot (WB), quantitative polymerase chain reaction (qPCR) and DA uptake/content. Of note, “ensemble” methods like WB and qPCR do not allow a distinction between expression changes in the whole population or in just a subset of cells. In this respect, ICC represents a more reliable technique to check the phenotype of the cell population. ICC with DAergic markers has been used in 26 out of the 56 articles: 10 in undifferentiated and 16 in differentiated cells (in general compared with undifferentiated controls). Comparison of all the provided ICC images, taking into account the cell source, media composition and differentiation method (when applicable), did not allow us to draw definitive conclusions about the phenotype of the undifferentiated cells, or about differences in outcome between various differentiation methods. Of note, no or only a few positive control images are shown in most articles, hindering the comparative literature survey. Also, SH-SY5Y cells are known to respond inconsistently to the same differentiation treatment, depending on their cell source [49] or possibly passage number. Therefore, the variation reported may be at least in part due to the origin of the cells and different aspects regarding their handling, highlighting the importance of the proper reporting of all protocols involved.

In order to create SH-SY5Y-derived cell models that mimic PD, strategies are used based on drug treatment and/or genetic approaches with the manipulation of expression of candidate genes that have emerged from genetic studies in PD-families. Most papers (800 out of 962) choose one pharmacological or genetic strategy to force manifestation of a PD-like phenotype, but also multiple variants or combinations of these strategies have been used. The most-used compounds in drug-based approaches are 1-methyl-4-phenylpyridinium (MPP+), 6-hydroxydopamine (6-OHDA) and rotenone, which dysregulate multiple cellular pathways, focusing on mitochondrial dysfunction and oxidative stress (Table 3). MPP+ is the toxic metabolite of 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP), a by-product in the synthesis of 1-methyl-4-phenyl-4-propionoxy-piperidine (MPPP), a synthetic analog of heroin, which causes severe parkinsonism in humans when injected intravenously [50]. Since SH-SY5Y cells do not have the machinery to transform MPTP to MPP+, the metabolite itself is administered. Once inside the cell, MPP+ inhibits complex I of the electron transport chain, impairing mitochondrial respiration and increasing reactive oxygen species (ROS) production, and redistributes DA to the cytosol, where it is oxidized and generates more ROS (reviewed in [51]). 6-OHDA is a catecholaminergic neurotoxin and its molecular mechanisms of action have been reviewed elsewhere [52]. Briefly, 6-OHDA enters catecholaminergic neurons via DA or NA transporters; it accumulates inside of the cell and triggers the formation of ROS and catecholamine quinones, leading to oxidative stress and cell death. Furthermore, 6-OHDA may inhibit complexes I and IV of the electron transport chain [53], although this has not been confirmed by others [54]. Exposure to rotenone, a highly lipophilic insecticide, has been linked to sporadic PD [55] and can directly enter cells, independently from transporters, to inhibit complex I, impairing mitochondrial respiration and enhancing ROS production, and inhibit proteasomal activity (reviewed in [56]).

Next to drug-based strategies also genetic approaches (e.g. knock down or forced overexpression of genes with mutations found in familial cases of PD) have been widely used to induce a PD-like phenotype (Table 3). Altogether 19 loci segregating with familial forms of PD are now known [57]. Of these only a subset have been used for reverse genetic manipulation, and overexpression of genetically encoded mutated variants of α-synuclein (A30P, A53T, E46K, G51D, H50Q, S129A, S129D, S129E) or extracellularly added α-synuclein variants is by far the most commonly used method to mimic PD in SH-SY5Y. Functional and rescue studies involving knockdown, overexpression or mutated forms of other genes, such as LRRK2 (G2019S, I2020T, R1441C, Y1699C), PINK1 (G309D, P209A, P399L, T313M), DJ-1 (A39S, C53A, C106A, L166P), ATP13A2, PLA2G6, GBA and Parkin (C289G, C431F, G328E, G430D, K161N, R42P, T240N, T240R, R265C, W453stop), have also been performed. Gene knockdown was mostly performed by transfection with siRNA or shRNA, while stable expression of mutated genes was achieved by their insertion into vectors such as pcDNA3.1 and transfection with reagents like Lipofectamine® 2000 (ThermoFisher Scientific) or FuGENE ® (Promega), adenoviral infection or lentiviral transduction. Surprisingly, novel revolutionary technologies such as the use of CRISPR/Cas9 or TALEN have not yet been applied in studies on PD, but we anticipate that reports on these tools for precision gene-editing will appear soon. Collectively, the studies mimicking the genetic mutations present in PD have allowed the identification of a diverse set of pathways and molecular processes that are involved in the manifestation of the disease [58], including mitochondrial and mitophagy dysfunction [59, 60] and proteasomal and autophagy dysregulation, leading to protein aggregation [61‐63]. One complementary strategy to study PD in cells is to interfere directly with one of these processes by administering specific compounds, with agonistic or antagonistic activity, such as hydrogen peroxide (oxidative stress), lactacystin/MG-123 (proteasome inhibitors), tunicamycin (N-glycosylation inhibitor, triggers ER stress), bafilomycin (inhibitor of vacuolar H⁺ ATPase, leading to autophagy dysfunction), thapsigargin (inhibitor of the sarco/endoplasmic reticulum Ca^{2 +} ATPase, resulting in ER stress and autophagy inhibition), carbonyl cyanide m-chlorophenyl hydrazone (CCCP) (inhibitor of oxidative phosphorylation, leading to mitochondrial dysfunction), Conduritol B epoxide (CBE) (GBA inhibitor), or salsolinol/staurosporine (cell death). Intriguingly, staurosporine, a broad-spectrum kinase inhibitor, has been used in some PD-related publications to induce cell death [64‐72], while other publications have used it to induce DAergic differentiation and study PD-related features [43, 44]. Early studies on SH-SY5Y cells showed differentiation towards a neuronal phenotype upon treatment with staurosporine [73, 74], which later has been characterized as catecholaminergic-like [75]. Therefore, the effects of different concentrations of staurosporine on SH-SY5Y cells should be characterized carefully to properly interpret studies that used this drug either as a differentiation agent or as an inducer of apoptosis. Figure 3 summarizes the now known cellular processes that are dysregulated in PD, based on the analysis of the functions of the proteins encoded by the (mutated) familial PD genes, and the use of PD-mimicking drugs in SH-SY5Y cells.

Apart from the DAergic phenotype, the ability of the cell line to reproduce the cellular abnormalities of PD is crucial for the validity of the model. One of the main hallmarks of PD is α-synuclein aggregation [5]. To mimic this pathological feature, overexpression of WT α-synuclein or stable expression of one of its familial mutations, such as A53T or A30P, has been successfully used [76‐78]. Nevertheless, these manipulations do not always inevitably lead to increased formation of inclusions. Therefore, triggers such as cell differentiation together with FeCl₂ (and H₂O₂) treatment, or Hsp70 blockage, are sometimes needed to observe α-synuclein aggregation [79‐82]. These different outcomes are possibly due to the specific α-synuclein mutation used [77] or the level of expression achieved by the various constructs [78]. Interestingly, spontaneous α-synuclein aggregation has been reported in non-transfected SH-SY5Y cells [83]. Moreover, both differentiated and undifferentiated SH-SY5Y cells are sensitive to extracellular α-synuclein-induced toxicity [84]. The cell line has also been used to study the kinetics and mechanisms of α-synuclein degradation [78, 85], the link between α-synuclein aggregation and intracellular calcium [86], and other pathological changes of α-synuclein, such as its carboxyl-terminal cleavage [32]. Other PD-related problems, such as abnormal mitochondrial function, oxidative stress and autophagy or proteasomal dysfunction, have been reproduced in SH-SY5Y cells as well. These hallmarks are usually triggered by the administration of specific drugs (Fig. 3) or, alternatively, by the knockdown of a gene corresponding to a familial PD-gene or expression of a familial PD-gene. For example, silencing of PINK1 leads to mitochondrial dysfunction [87, 88] and mutated forms of α-synuclein impair proteasomal activity [89]. The additive effects that are caused by co-expression of familial PD genes can also be observed in this cell line (e.g. cotransfection of α-synuclein and LRRK2 enhances the formation of aggregates, phosphorylation, cell-to-cell transmission and extracellular release of α-synuclein [90]). Therefore, SH-SY5Y cells represent an attractive tool to study most cellular alterations linked to PD and strategies to ameliorate their effects, but a careful experimental setting is required when analyzing α-synuclein aggregation, since the findings vary between studies. Yet, not all aspects of PD pathobiology can be faithfully studied in SH-SY5Y cells, and therefore investigations into for example electrophysiological abnormalities or neurochemical dysfunction require other and more complex models, such as primary dopaminergic cultures or brain slices, ex vivo.

PD is characterized by the loss of DAergic neurons from the SN [5]. Thus, primary cultures of neurons from this brain area of patients and controls may be considered the most reliable models to unravel the molecular mechanisms underlying this disease. However, the inaccessibility and lack of proliferation of such neurons largely precludes their use. Conversely, as discussed above, the use of a proliferative and more uniform model like the SH-SY5Y cell model has also limitations. Therefore in many studies other cell types have been employed in parallel. Of the articles analyzed, 67.6% reported experiments performed exclusively in the SH-SY5Y cell line and 19.7% used in parallel other cell lines with a neuronal phenotype, including rodent mesencephalic primary cultures and (mainly cortical) primary neurons, stem cells, PC12 cell line, Neuro-2a cell line or MN9D cell line (Table 4). The remaining 12.7% used cell lines that are not neuronal (−like), such as HEK293, HeLa or glial cells. Primary cell cultures are physiologically more relevant than immortalized cell lines, but they are difficult to maintain and, depending on the age of the source animals or the dissection accuracy, can introduce experimental variability [91]. Since the SN is located in the mesencephalon, or midbrain, rodent mesencephalic primary cultures are enriched for the cell population of interest, the SN DAergic neurons. Primary cultures from other brain regions (mainly cortex) have also been extensively used as complementary in vitro models, but these cultures consist of mixed populations of various neuronal subtypes. Furthermore, primary cultures of DAergic neurons from other species, such as the worm C. elegans, have been used [92, 93]. In general, the overtly less physiologically relevant permanently established cell lines are easier to maintain than cells in primary cultures (because the cell lines are generally rendered immortal) and can sometimes be differentiated into terminal neuronal populations. The PC12 cell line (ATCC® CRL-1721™) has been derived from a rat pheochromocytoma and widely used to study PD. This lineage has a chromaffin-like character, and shares its embryonic origin with DAergic neurons. Nerve growth factor treatment of PC12 cells induces differentiation into a catecholaminergic-like phenotype [94, 95]. The Neuro-2a cell line (ATCC® CCL-131™) has been derived from a mouse brain neuroblastoma and can be differentiated into neuronal-like cells [96] and, more specifically, to DAergic neurons by a dbcAMP treatment [97]. Other PD cellular models include cell hybrids, such as the MN9D cell line [98], and derivatives from the neuroblastoma SK cell line. A full list of neuronal (−like) cell models used in parallel with SH-SY5Y cells can be found in Additional file 4. It is important to note that most of the alternative lineages are derived from species other than human. Finally, patient-derived induced pluripotent stem cells (iPSCs) that can be differentiated into DAergic neurons (and possibly other relevant cell types such as glial cells) represent an emerging new category of cell models for PD [99‐101].

Apart from cellular models, about 21% of the articles that use SH-SY5Y cells also employ an animal model for PD to validate and further understand the cellular results. The animal models include mouse, rat, C. elegans and fruit fly and the PD features in these in vivo models are mimicked with MPTP, 6-OHDA, rotenone or genetic mutations (reviewed in [102, 103]).