Degree of differentiation impacts neurobiological signature and resistance to hypoxia of SH-SY5Y cells

Objective. SH-SY5Y cells are valuable neuronal in vitro models for studying patho-mechanisms and treatment targets in brain disorders due to their easy maintenance, rapid expansion, and low costs. However, the use of various degrees of differentiation hampers appreciation of results and may limit the translation of findings to neurons or the brain. Here, we studied the neurobiological signatures of SH-SY5Y cells in terms of morphology, expression of neuronal markers, and functionality at various degrees of differentiation, as well as their resistance to hypoxia. We compared these to neurons derived from human induced pluripotent stem cells (hiPSCs), a well-characterized neuronal in vitro model. Approach. We cultured SH-SY5Y cells and neurons derived from hiPSCs on glass coverslips or micro-electrode arrays. We studied expression of mature neuronal markers, electrophysiological activity, and sensitivity to hypoxia at various degrees of differentiation (one day up to three weeks) in SH-SY5Y cells. We used hiPSC derived neurons as a reference. Main results. Undifferentiated and shortly differentiated SH-SY5Y cells lacked neuronal characteristics. Expression of neuronal markers and formation of synaptic puncta increased during differentiation. Longer differentiation was associated with lower resistance to hypoxia. At three weeks of differentiation, MAP2 expression and vulnerability to hypoxia were similar to hiPSC-derived neurons, while the number of synaptic puncta and detected events were significantly lower. Our results show that at least three weeks of differentiation are necessary to obtain neurobiological signatures that are comparable to those of hiPSC-derived neurons, as well as similar sensitivities to metabolic stress. Significance. This indicates that extended differentiation protocols should be used to study neuronal characteristics and to model brain disorders with SH-SY5Y cells. We provided insights that may offer the basis for the utilization of SH-SY5Y cells as a more relevant neuronal model in the study of brain disorders.


Introduction
Human in vitro neuronal models have emerged as a powerful tool to investigate brain disorders over the past decades [1,2].Among these models, SH-SY5Y cells have received significant attention, because of their ease of maintenance, rapid expanding nature, and relatively low costs.They have been broadly used to study molecular and cellular abnormalities and to test treatment strategies in disorders such as Alzheimer's disease, Parkinson's disease, and cerebral ischemia [3][4][5][6][7][8].
Despite the potential of SH-SY5Y cells to exhibit neuronal-like properties, the majority of the studies affirming to use a neuronal model have been conducted with undifferentiated SH-SY5Y cells (e.g. about 90% of studies on cerebral ischemia and 80% on Parkinson's Disease) [3,6,8].When neuron-like cells have been used, the degree of differentiation varied widely (i.e. from 1 to 21 d of RA stimulation) [3,6].Differences in expression of neuronal markers and properties, as well as in gene expression have been observed in undifferentiated and differentiated SH-SY5Y cells, indicating that SH-SY5Y cells move towards a neuronal state through the differentiation process [21].Thus, since the phenotype and response to drugs depend on cellular properties, the use of cells lacking neuronal properties as a neuronal model is debatable.Furthermore, due to its cancerous origin, the SH-SY5Y cell line displays a number of genetic aberrations that influence cellular properties (i.e.differentiation fate, viability, growth performance and metabolic aspects), so that the use of such an oncogenically transformed cell line remains a controversial issue [6,22].
With the recent advent of human induced pluripotent stem cells (hiPSCs) technology, the establishment of human in vitro neuronal models in which the genetic characteristics of the donor (i.e.healthy subject or patient) are preserved has become possible [1,4].Several differentiation protocols have been developed to obtain mature functional neurons which have been extensively characterized in terms of morphology, expression of neuronal markers and electrophysiological activity [23][24][25].HiPSC-derived neuronal networks have been used to model neurological disorders, to study molecular properties underpinning phenotypic abnormalities and to test treatment strategies [1,24,26,27].Thus, hiPSCderived neuronal networks have become a widely accepted benchmark for human in vitro neuronal models [1,2].
In the current study, we aimed to compare the neurobiological signatures of SH-SY5Y cells at varying levels of differentiation in terms of morphology, expression of neuronal markers and functionality.Furthermore, we aspired to investigate the effect of various degrees of differentiation on responses to hypoxia-induced metabolic stress.We studied responses to metabolic stress, since cerebral ischemia is a much-modeled condition in SH-SY5Y cells and metabolic stress is an important final common path in various other brain diseases [3,[28][29][30][31].Since the susceptibility of SH-SY5Y cells to ischemia in relation to differentiation timing remains poorly understood, this presents a significant obstacle in interpreting and comparing research outcomes and potential treatment strategies [6].We employed hiPSC-derived neuronal networks from healthy subjects as a reference neuronal in vitro model, because of their wellestablished and characterized neuronal nature [1,23].
We found that prolonged differentiation of SH-SY5Y cells (i.e. at least three weeks) is necessary to obtain neurobiological signatures that are comparable with hiPSC-derived neurons.Furthermore, we showed that the degree of differentiation impacts the SH-SY5Y cells' susceptibility to hypoxia, with prolonged differentiation periods leading to a lower resistance to hypoxia, comparable to the responses to hypoxia of hiPSC-derived neurons.Our findings indicate that the degree of differentiation is a critical factor when working with SH-SY5Y cells and advocate for the use of extended differentiation protocols to study neuronal characteristics.

Experimental protocol
Investigation of neuronal properties (i.e.expression of neuronal markers, formation of synaptic puncta and exhibition of electrophysiological activity) was conducted at various degrees of differentiation (DIV0, DIV1, DIV3, DIV7, DIV14 and DIV21) for SH-SY5Y cells and at DIV49 for hiPSC-derived neuronal networks.To investigate electrophysiological activity, we monitored activity of SH-SY5Y throughout the various degrees of differentiation.
Computer-controlled climate chambers were used to establish normoxia (20% O 2/ 75% N 2/ 5% CO 2 ) or hypoxia (2% O 2 /93% N 2 /5% CO 2 ).Temperature was maintained stable at 37 • C, as previously described [27].SH-SY5Y cells and hiPSCderived neuronal networks were first exposed to normoxia.Expression of neuronal markers and cell viability were assessed through microscopy after immunocytochemical staining.Electrophysiological activity was recorded with MEAs for 10 min, after 30 min of acclimatization in a climate-controlled chamber under normoxia.Afterwards, cultures were exposed to hypoxia, where immunocytochemistry assays were performed after 24 and 48 h of exposure.
Images were taken at a 60x magnification with the use of a Nikon Eclipse 50i Epi-fluorescence microscope (Nikon, Japan).Quantification of the mean intensity of MAP2 was computed with custommade MATLAB scripts (The MathWorks, Natick, MA, USA) and corrected for number of cells identified in each image.Each group was normalized against the mean intensity of SH-SY5Y cells at DIV0.Synaptic puncta were counted manually, by assessing the number of synapsin1/2 positive puncta per 10 µm of MAP2 positive neurite.

Cell viability
Cell viability was assessed through a live/dead assay, including apoptosis.Cell Event (1:500, Thermoscientific) was added to the cultures and cultures were incubated for 30 min at 37 • C to stain the apoptotic cells.Next, the hypoxic period started for 24 h or 48 h.At the end of the hypoxic period, propidium iodide (PI, 1:1000, Invitrogen) was added and incubated for 15 min at 37 • C to stain dead cells.Cells were carefully washed with dPBS (Gibco) and fixated with 3.7% PFA for 15 min at RT. Lastly, nuclei were stained with DAPI (1:1000, Sigma Aldrich) for 10 min at RT.Samples were carefully washed again, mounted with Mowiol (Sigma Aldrich), and stored.
Images were taken at a 20x magnification with the use of a Nikon Eclipse 50i Epi-Fluorescence microscope (Nikon, Japan).Pre-processing of images was conducted through custom-made MATLAB scripts (The MathWorks, Natick, MA, USA).Pipsqueak Pro was used to identify and quantify DAPI labeled cells.Quantification of apoptotic and dead cells was assessed manually.Cells were considered live when only labeled with DAPI (blue), apoptotic when labeled with DAPI (blue) in combination with Cell Event (green) and dead when labeled with DAPI (blue) and PI (red) with or without Cell Event (Green).Each group was compared to its normalized normoxia.Wells that displayed insufficient quality, (i.e.low cell density, cell clumping) were discarded.Inclusion criteria ensured that neuronal networks belonging to distinct groups are comparable and viable (i.e.95% of live cells during normoxia).

MEA recordings and data analysis
Electrophysiological activity was measured by the Multiwell-MEA system using the Multiwell-Screen software (Multi Channel Systems, Reutlingen, Germany).Raw signal was sampled at 10 kHz and filtered with a high-pass second order Butterworth filter with 100 Hz cut-off and a low-pass fourth order Butterworth filter with a 3.5 kHz cut-off.Data analysis and detections of spikes were performed with the use of Multiwell Analyzer software (Multichannel systems, MCS GmbH, Reutlingen, Germany) in combination with custom-made MATLAB scripts (The MathWorks, Natick, MA, USA), as previously described [27].The parameters extracted were mean firing rate (MFR)-number of detected events in time per electrode averaged among all electrodes and percentage of active electrodes.To visualize the shape of events detected in hiPSCs-derived neurons and SH-SY5Y cells grown on MEAs, we computed the average spike shape.We detected the spike peak for each event exhibited in one electrode and we then averaged the 100 ms long spike waveforms (i.e.50 ms before and after the peak).Stringent inclusion criteria have been applied [1].First, neuronal networks were included only if showing good quality (i.e.cell density allowing a proper neuron-electrode coupling and even distribution of cells).Only 8.5% of the SH-SY5Y cultures and 100% of hiPSCs-derived neuronal networks plated on MEA had good quality.In addition, in normoxia neuronal networks should exhibit MFR > 0.1 spike/s and number of active channels >80% [27].

Statistical analysis
The statistical analyses were conducted using GraphPad Prism 9 (GraphPad Software, Inc. California, USA).Significance was determined by p-values < 0.05.We tested for normality of residuals and opted for non-parametric analysis due to the distributions of our data.Mann-Whitney-U test was used to compare detected events and active channels between differentiated SH-SY5Y (three weeks) and hiPSC-derived neuronal networks (DIV49).Kruskal-Wallis and Dunn's multiple comparisons tests were used to compare groups of SH-SY5Y cells of various degrees of differentiation and hiPSC-derived neurons.To facilitate comparisons, all parameter values during different phases of the experiment were normalized to their normoxia (control) values, and the data are presented as mean ± standard error of the mean.Samples sizes are reported in the corresponding figure legends.Effect size (eta-squared from the Kruskal-Wallis test) of our dataset has been estimated using R (Team 2023) [32].All exact p-values and effect sizes of the statistically different comparisons are reported in tables S1 and S2, respectively.

Neurobiological signatures of SH-SY5Y cells evolve during differentiation
To determine the stage at which SH-SY5Y cells acquire neuronal properties, we investigated their neurobiological signature at various degrees of differentiation (DIV0, DIV1, DIV3, DIV7, DIV14, DIV21).We compared the observed properties with those of hiPSC-derived neuronal networks (exact pvalues are reported in table S1).
Undifferentiated SH-SY5Y cells (DIV0) display a large, flat, epithelial-like cell body with multiple short processes extending outward (figure 1(a)).During differentiation, the SH-SY5Y cells morphologically change and start to resemble neurons.In particular, during the first week (DIV1, DIV3 and DIV7) SH-SY5Y cells display more elongated processes, whereas with two or three weeks of differentiation (DIV14 and DIV21) several neuritic projections connecting to neighboring cells are formed.This is similar as the differentiation trajectory observed in hiPSCs, in which neuronal networks are formed after three weeks in vitro (figure 1(b)) [1,2,23,26].We observed that the differentiation of SH-SY5Y cells was accompanied by increased MAP2 expression (figures 1(c) and (d)).In particular, the relative MAP2 intensities observed in undifferentiated (DIV0) and shortly differentiated SH-SY5Y cells (DIV0, DIV1, DIV3 and DIV7) were significantly lower than observed at DIV21.When compared to hiPSC-derived neuronal networks, significant differences were observed in shortly differentiated SH-SY5Y cells, whereas longer degrees of differentiation (DIV14, DIV21) led to similar MAP2 intensities as in hiPSC-derived neurons.In addition, we found that synaptic puncta were not formed in SH-SY5Y cells during the first weeks of differentiation (figure 1(e)).Only at DIV21, synaptic puncta were visible, but the amount was significantly lower than in hiPSC-derived neurons.
Finally, we investigated the electrophysiological properties of SH-SY5Y cells using MEA (figures 1(f)-(i)).We observed that SH-SY5Y cells did not exhibit any activity during the first three weeks of differentiation (data not shown).After three weeks in vitro, events were detected on at most one electrode per culture (i.e.8.3% of all electrodes), while neuronal networks derived from hiPSCs showed events on almost all recording channels (i.e.97.9% of all electrodes) (figures 1(f) and (i)).In addition, when compared to hiPSCs-derived neurons, SH-SY5Y exhibited a significantly lower MFR (figure 1(i)).Furthermore, we observed that the detected events of hiPSCderived neuronal networks showed a stereotyped action potential shape while in SH-SY5Y cells the typical shape was not present (figure 1(g) and supplementary figures 1(c) and (d)).The amplitude of detected events of SH-SY5Y cells was lower and the duration of the events was longer as compared to hiPSC-derived neuronal networks (figure 1(g)-(h) and supplementary figures 1(c)-(f)).

Resistance to hypoxia of SH-SY5Y cells depends on degree of differentiation
We investigated the effect of hypoxia exposure on SH-SY5Y cells at various degrees of differentiation.Since SH-SY5Y cells lacked adequate functional activity on MEA (i.e.low number of active channels and atypical action potential shape), the resistance to hypoxia was only assessed by evaluating the number of living cells microscopically.
Under normoxia, numbers of living cells were similar for SH-SY5Y cells at various degrees of differentiation and hiPSC-derived neurons (figures 2(a)-(c)).When exposed to hypoxia, longer differentiation time was associated with lower cell survival (figures 2(a)-(e)).When compared to hiPSC-derived neuronal networks, SH-SY5Y cells at the early stages of differentiation (i.e.DIV0, DIV1, DIV3) showed a higher number of living cells at 24 h of hypoxia as compared to hiPSC-derived neurons (figures 2(a), (b) and (d)).With higher degrees of differentiation (DIV7, DIV14, DIV21) no significant difference in number of living cells was observed between SH-SY5Y cells and hiPSCs-derived neurons at 24 h of hypoxia (figures 2(a), (b) and (d)).After 48 h of hypoxia, the number of living SH-SY5Y cells at early differentiation stages (DIV0 and DIV3) was higher compared to hiPSC-derived neuronal networks (figures 2(a), (b) and (e)).Conversely, there were no differences in number of living cells between SH-SY5Y cells at DIV14 and DIV21 and hiPSC-derived neuronal networks (figures 2(a), (b) and (e)).Our results indicate that the degree of differentiation affects the response to hypoxia of SH-SY5Y cells.

Discussion
In this study, we established that the neurobiological properties of SH-SY5Y cells arise over time as they differentiate towards a neuronal-like state.Additionally, we observed that the degree of differentiation plays a role in the resistance to hypoxia, with prolonged differentiation resulting in decreased resistance.
The utilization of SH-SY5Y cells in neuroscientific research is hindered by limited standardization of the model, in particular related to variation in differentiation timing (i.e.ranging from undifferentiated to three weeks of differentiation) [6] and lack of uniformity and standardization of differentiation protocols [6,7,[10][11][12][13][14][15][16][33][34][35][36][37][38].Undifferentiated and differentiated SH-SY5Y cells exhibit distinct physical and functional traits.Undifferentiated cells are characterized by large cell bodies, short processes, and the absence of mature neuronal markers.This profile remains consistent in cells differentiated for less than a week.However, with longer differentiation periods (i.e. three weeks), SH-SY5Y cells display extended protrusions and increased expression of mature markers, closely resembling hiPSC-derived neurons.Our data indicate that MAP2 expression in SH-SY5Y cells increases significantly after 14 days of differentiation, approaching levels observed in hiPSC-derived neurons by day 21.Furthermore, a 3week differentiation protocol with RA induces the formation of synaptic puncta, aligning with previous studies showing that SH-SY5Y differentiation promotes mature neuritic processes, increased excitability, synapse formation, and neuron-specific gene expression [13,15,[17][18][19][20][39][40][41][42][43][44][45].In line with our  S1 and S2, respectively.findings, Avola and colleagues observed a reduction in proliferation activity in SH-SY5Y cells after 14 days of RA-induced differentiation, indicating effective induction towards a neuronal phenotype [40].In contrast, undifferentiated SH-SY5Y cells persistently exhibit proliferation and lack the expression of mature neuronal markers [12].Patch-clamp research indicates an increase in cellular excitability during differentiation, with the ability to generate action potentials in a fraction of differentiated cells after stimulation.Experimental data indicate that the characteristics of undifferentiated SH-SY5Y cells are more similar to those of cancer cells than of neurons (i.e.high potassium channel activity and absence of action potentials), while characteristics similar to those of neurons are acquired after differentiation (i.e.prominent sodium current and ability to generate action potentials) [20,[42][43][44][45][46].Furthermore, during the first week of differentiation with RA and BDNF in Neurobasal medium, SH-SY5Y cells demonstrate upregulation of various genes, including neurotrophins and those related to axonal guidance signaling pathways [21].Although SH-SY5Y cells can be differentiated into neuronal-like cells, the predominant focus in the majority of conducted studies has been on undifferentiated cells [3,6].These studies have primarily explored phenotypic signatures and responses to treatment strategies in various models of neurological disease [3,6,8].When differentiation is performed, it is often for a brief period (i.e. 3 days).Only a limited number of studies have used SH-SY5Y cells differentiated for up to 21 days [6,47,48].Differences in cellular properties and gene expression observed in undifferentiated cells and cells with various degrees of differentiation lead to distinct neurobiological signatures [13, 15, 19-21, 39, 41].Evidences reveal diverse cell subsets in differentiated SH-SY5Y cultures, such as neuronal-like cells with extended axons and neuronal marker expression, partially differentiated cells with 'stem-like' traits and possible proliferation, and undifferentiated, potentially proliferative tumor-like cells [21,49].The spread in hypoxia response we found might be due to different vulnerability of cells belonging to diverse differentiation stages.Optimization is warranted to reduce variability of the differentiation protocol.In addition, the agents used to induce the differentiation can interfere with responses to various pathological conditions [13,41,50,51].For example, RA alters mitochondrial function decreasing susceptibility to oxidative stress [50] and increasing DJ-1 protein immunocontent [41].In addition, TPA-induced differentiation of SH-SY5Y cells regulates muscarinic receptors and the acetylcholinesterase activity [13].In line with this, our results showed that the degree of differentiation affects the phenotype of SH-SY5Y cells in response to metabolic stress.In particular, we showed that number of living cells decreased during hypoxia with prolonged differentiation, indicating lower resistance of differentiated SH-SY5Y to hypoxia.This raises concerns regarding the relevance of studies conducted on undifferentiated SH-SY5Y cells.Thus, further investigations focusing on differentiated SH-SY5Y cells are needed to gain a better understanding of their relevance in various research areas.
To date, studies involving SH-SY5Y cells have often overlooked neuronal functionality, a crucial aspect for evaluating neuronal network phenotypes and drug responses [2].Our findings indicate that in differentiated SH-SY5Y cells events can be detected on MEAs, aligning with a previous study that observed minimal MEA activity at DIV21 [19].While in hiPSC-differentiated neuronal networks spontaneous events were detected in all channels and showed a typical action potential shape, in SH-SY5Y cell cultures detections were observed in at most one channels and did not exhibit an action potential shape, even if a good cellular coverage was present (supplementary figures 1(a) and (b)).As previously reported, literature has shown that only a fraction of differentiated SH-SY5Y cells can generate action potentials and only after prolonged stimulation [20,43,46].Thus, it is possible that in absence of stimulation the amount of firing neurons-and thus synaptic inputs reaching surrounding neurons-was not enough to generate action potentials that can be detected by MEA.The fluctuations observed in SH-SY5Y cells most likely reflect depolarizations resulting from the summation of subthreshold phenomena in many neurons (i.e.post synaptic potentials, PSPs) that have previously been observed on MEA in the presence of low-asynchronous activity [52].Both the shape (i.e.fast rising, exponential decay) as the relative amplitude are typical of PSPs (supplementary figure 1(d)) [52].The emergence of fluctuations on MEA together with the increasing excitability observed in single cells through differentiation [20,43,46] suggest the potential for further maturation of SH-SY5Y cell neuronal functionality through prolonged differentiation.Nonetheless, extending SH-SY5Y differentiation beyond three weeks remains challenging due to culturing limitations, such as cell clumping and detachment, underscoring the need for refinement in differentiation protocols.
We employed hiPSC-derived neurons as an established in vitro model, known for their comprehensive characterization in terms of morphology, neuronal markers, synapse formation, and electrophysiological activity [1,23,24,26,27].Our findings demonstrate that SH-SY5Y cells, after 21 days of differentiation, share neurobiological signatures and hypoxia resistance with hiPSC-derived neurons, implying the potential for SH-SY5Y cells to acquire neuronal-like properties through extended differentiation.Thus, the use of undifferentiated or shortly differentiated SH-SY5Y cells should be avoided to model neuronallike cells.When longer differentiation protocols are used, the choice between hiPSCs and SH-SY5Y models depends on the specific study objectives.Notably, hiPSC-derived neuronal networks can be generated from both healthy individuals and patients, preserving genetic backgrounds, while SH-SY5Y cells have a cancerous origin with inherent genetic abnormalities.On the contrary, the cost of SH-SY5Y cells studies is lower as compared to the use of hiPSCs.
There are various limitations in this study that should be considered.First, we used cultures composed by different population of neurons [7,23,24].The choice of the hiPSC-derived cultures was driven by the need for a robust and well characterized neuronal model to use as reference.It is important to point out that the basic neuronal properties investigated in this study (e.g.MAP2 expression) are not altered in different neuronal populations [53].Another drawback is that the hiPSC-derived neurons were co-cultured with astrocytes, while attempts to include astrocytes in SH-SY5Y cell cultures were ineffective.Since astrocytes play a fundamental role in supporting neuronal viability and functionality [23,54], we cannot exclude that co-culturing SH-SY5Y cells with astrocytes or addition of conditioned astrocytic medium would influence neuronal maturation of SH-SY5Y.
In conclusion, this study highlights that the degree of differentiation is a critical factor when working with SH-SY5Y cells: their neurobiological signatures increase and sensitivity to metabolic stress decrease during three weeks of differentiation.Our findings highlight the potential of obtaining neuronlike cells that exhibit properties and resistance to hypoxia similar to hiPSC-derived neurons, but also stress that these only arise through prolonged differentiation.Therefore, optimizing and standardizing the differentiation protocol on longer timescales (i.e. more than three weeks) is crucial to achieve a SH-SY5Y derived human neuronal in vitro model that may translate to neurons or the brain.This provides a new starting point for utilization of the SH-SY5Y cell model in neuroscientific research.