Effects of GADL1 overexpression on cell migration and the associated morphological changes

Lithium has been used for maintenance treatment of bipolar disorder, but drug response varies among patients. Single-nucleotide polymorphisms in glutamate decarboxylase–like protein 1 (GADL1) are found to be associated with lithium response in Han Chinese bipolar patients. In this study, we assessed GADL1 function using a neuroblastoma cell line that stably overexpressed GADL1. Genes encoding factors involved in cell migration, such as FN1, ITGA2, ITGAV and CCL2, were downregulated in GADL1-overexpressing cells. GADL1 overexpression indeed suppressed cell migration. Cell migration speed and perimeter length exhibited similar trends, both of which were decreased under GADL1 overexpression or lithium treatment but increased upon stimulation with CCL2. Secreted GADL1 or its enzyme product, taurine, in the conditioned medium might exert only mild effects on the observed changes. Compared with SH-SY5Y cells, GADL1-overexpressing cells were much more sensitive to CCL2 treatment but less sensitive to lithium, indicating that the level of GADL1 expression can affect cell sensitivity to lithium or CCL2 treatment. Together, these results suggest that cell migration and related morphological changes might provide good indicators of the sensitivity toward lithium treatment, and the GADL1 stable overexpression cell line might serve as a useful platform to screen novel therapeutics for bipolar disorder.

receptors that couple the extracellular matrix to the cytoskeleton to regulate cell migration 22,23 . The chemokine (C-C motif) ligand 2 (CCL2) also regulates neuron migration 24,25 . Treatment of neurons in culture with CCL2 leads to a significant, dose-dependent increase in the number of migrating neurons and the average distance they travel 25 . Neurons that have undergone in vitro transdifferentiation from bipolar patient skin cells exhibit significantly different cell-adhesion phenotypes between lithium responders and nonresponders, indicating that cell adhesion is associated with clinical response to lithium treatment 26 . To help understand GADL1 function, GADL1 was stably overexpressed in the human neuroblastoma cell line, SH-SY5Y. We assessed the impact of GADL1 overexpression or of treatment with lithium or CCL2 on cell migration and related morphological changes.
To demonstrate a causative relationship between GADL1 overexpression and the cellular phenotypes, we further reduced GADL1 expression in the GADL1-overexpressing cell line using small interfering RNA (siRNA) knockdown. The RNA expression changes of GADL1, FN1, ITGA2, ITGAV, and CCL2 after GADL1 knockdown (siGADL1) in the GADL1-overexpressing cell line were examined using RT-qPCR analysis, showing that GADL1 was knocked down to 67.2% relative to RISC-free control siRNA (Fig. 1c). Effects of GADL1 overexpression on cell number, migration, and morphology. Next, cell migration was compared between GADL1-overexpressing and SH-SY5Y cells using label-free, real-time, holographic imaging for 48 h (Fig. 2). Cell counts and morphology were also monitored at the same time. Cell number (Fig. 2a) and thickness (Fig. 2d) did not differ significantly between the GADL1-overexpressing cells and SH-SY5Y cells. Moreover, GADL1 overexpression significantly decreased cell migration (Fig. 2b), area (Fig. 2c), volume (Fig. 2e), and perimeter length (Fig. 2f).
Two parameters were also used to assess changes in cell shape: irregularity and eccentricity. Irregularity, calculated as 1-4π(area)/(perimeter length) 2 , is how much the circumference of the cell deviates from the circumference of a perfect circle. A value of 0 means the cell is circular, whereas a higher value means a more irregular (and hence longer) perimeter. Eccentricity, referring to the elongation of a cell, is calculated as the square root of 1 − w 2 /h 2 where h and w are the height and width of the minimum rectangle, w ≤ h. A value of 0 means that cells are essentially square, whereas a higher value means that cells are more elongated, i.e., rectangular. GADL1 overexpression significantly decreased both cell irregularity (Fig. 2g) and eccentricity (Fig. 2h), which could be observed by phase contrast microscopy as shown in Figs S1 and S2.
Effects of lithium on cell number, migration, and morphology. The single-nucleotide polymorphisms in GADL1 have been found to be associated with lithium response in bipolar patients of Han Chinese descent 8 . As shown in Fig. 1, GADL1 overexpression downregulated certain genes, including FN1, ITGA2, and ITGAV, involved in cell adhesion and migration 22 . Thus, we hypothesized that lithium could affect cell adhesion/ migration, which might be affected by the cellular level of GADL1. To test this idea, we monitored cell migration, cell counts, and morphology using label-free, real-time, holographic imaging of GADL1-overexpressing and SH-SY5Y cells in the presence or absence of 20 mM lithium chloride for 48 h (Fig. 2). Treatment with lithium at 1 mM did not have powerful effects on cellular phenotypes for both cells, except on cell area and perimeter length (Fig. S3).
For both SH-SY5Y and GADL1-overexpressing cells, lithium exposure significantly decreased cell number (Fig. 2a), irregularity (Fig. 2g), and eccentricity (Fig. 2h), but lithium had the effect of increasing cell thickness (Fig. 2d). Lithium had only mild effects on cell area (Fig. 2c) and cell volume (Fig. 2e). Lithium-induced morphological changes on SH-SY5Y and GADL1-overexpressing cells could be also observed by phase contrast microscopy as shown in Figs S1 and S2, respectively. Treatment with lithium significantly decreased cell migration (Fig. 2b) and perimeter length (Fig. 2f) only in SH-SY5Y cells.
www.nature.com/scientificreports www.nature.com/scientificreports/ Effects of CCL2 treatment on cell number, migration, and morphology. CCL2 can regulate neuronal migration in vitro 24,25 , and we found that GADL1 overexpression downregulated CCL2 expression ( Fig. 1). Thus, we assessed the effects of CCL2 alone or in concert with GADL1 overexpression on cell migration, cell counts, and morphology using label-free, real-time, holographic imaging of GADL1-overexpressing and SH-SY5Y cells in the presence or absence of 50 ng/ml CCL2 for 72 h (Fig. 3). Treatment with CCL2 at 25 ng/ml did not significantly affect the examined parameters (data not shown).
The effects of GADL1 overexpression on cell migration and morphology as well as the sensitivity toward lithium are cell autonomous. Because GADL1 is an enzyme 11 , we speculated whether the effects of GADL1 are cell autonomous or non-autonomous. To address this question, conditioned medium (CM) was collected from 2-to 3-day cultures of GADL1-overexpressing (GADL1-CM) or SH-SY5Y (5Y-CM) cells. Figure 4 shows that neither CM substantially affected SH-SY5Y cells in terms of cell number ( (Fig. 4f). However, the decrease in cell area and perimeter length induced by GADL1-CM was much smaller than that induced by GADL1-overexpressing cells, as shown in the comparisons in Fig. 2c,f, respectively. These results suggested that the GADL1 overexpression-induced decrease in cell area and perimeter length was mainly dependent on the 'intracellular form' of GADL1 rather than the 'secreted form' .
The effects of GADL1 overexpression on the sensitivity to CCL2 are cell autonomous. To understand if the differential sensitivity to CCL2 was mediated by the intracellular or secreted form of GADL1, CM was collected from 2-to 3-day cultures of GADL1-overexpressing (GADL1-CM) or SH-SY5Y cells (5Y-CM). Then, SH-SY5Y cells were cultured in the GADL1-CM or 5Y-CM in the presence or absence of CCL2 (50 ng/ml).
Effects of GADL1 overexpression or treatment with lithium or CCL2 on cell migration speed and perimeter length as assessed in independent experiments. Independent experiments were carried out to confirm our observed effects of GADL1 overexpression or treatment with lithium or CCL2 on cell migration speed (Fig. 6a,c,e, respectively), cell perimeter length recorded at 42 h (Fig. 6b,d,f, respectively), cell irregularity recorded at 42 h (Fig. S4a,c,e, respectively), and cell eccentricity recorded at 42 h (Fig. S4b,d,f, respectively). Effects   www.nature.com/scientificreports www.nature.com/scientificreports/ of GADL1 overexpression on cell migration speed (Fig. 6a) were calculated from differences between SH-SY5Y and GADL1-overexpressing cells without any treatment (−5.2 ± 1.46 μm/h) or from differences of SH-SY5Y cells cultured in 5Y-CM vs. GADL1-CM (−0.9 ± 0.47 μm/h). Effects of GADL1 overexpression on cell perimeter length (Fig. 6b) were calculated from differences between SH-SY5Y and GADL1-overexpressing cells without any treatment (−13.1 ± 4.52 μm) or from differences of SH-SY5Y cells cultured in 5Y-CM vs. GADL1-CM (−1.3 ± 1.46 μm). Effects of GADL1 overexpression on cell irregularity (Fig. S4a) and eccentricity (Fig. S4b) were calculated from differences between SH-SY5Y and GADL1-overexpressing cells without any treatment or from differences of SH-SY5Y cells cultured in 5Y-CM vs. GADL1-CM, showing that only cell irregularity was affected by GADL1 overexpression. These results suggested that the GADL1 overexpression-induced decrease in cell migration speed (Fig. 6a), Figure 6. Effects of GADL1 overexpression or lithium or CCL2 treatments on cell migration speed and perimeter length from independent experiments. Data from different batches of experiments were statistically analyzed to confirm GADL1 overexpression and the effects of treatment with lithium (20 mM) or CCL2 (50 ng/ml) on cell migration speed (a, c, e, respectively), and perimeter length recorded at 42 h (b, d, f, respectively). The Student's t test was used to compare the differences between SH-SY5Y (5Y) cells and GADL1-overexpressing cells (GADL1) ( * p < 0.05). Data for SH-SY5Y cells cultured in the conditioned medium (CM) from SH-SY5Y cells (5Y-CM) vs. from GADL1-overexpressing cells (GADL1-CM) were also compared using the Student's t test, which revealed no significant differences with respect to changes in cell migration speed and perimeter length for cells treated with lithium (c,d) or CCL2 (e,f). Effects of GADL1 overexpression on cell migration speed (a), and perimeter length (b) were calculated from differences between SH-SY5Y and GADL1-overexpressing cells without any treatment (cells). The same calculations were also done for SH-SY5Y cells cultured in 5Y-CM vs. GADL1-CM (CM). The Student's t test was used to compare the differences calculated from cells vs. CM ( *** p < 0.001). (2019) 9:5298 | https://doi.org/10.1038/s41598-019-41689-x www.nature.com/scientificreports www.nature.com/scientificreports/ perimeter length (Fig. 6b), and irregularity (Fig. S4a) were cell autonomous, depending mainly on the intracellular form of GADL1. Secreted GADL1 might have mild effects on the decrease in cell migration speed (Fig. 6a).

Discussion
GADL1 overexpression downregulated FN1, ITGA2, and ITGAV, which involved in cell adhesion and migration processes as well as maintenance of cell shape 22 . In our study, GADL1 overexpression suppressed cell migration, area, volume, perimeter length, irregularity and eccentricity. The RNA expressions of FN1, ITGA2, ITGAV, and CCL2 were upregulated after GADL1 knockdown in the overexpression cells, suggesting that the observed cellular phenotype and migration changes upon GADL1 overexpression were indeed triggered by GADL1 overexpression itself. Cell migration speed and perimeter length exhibited similar trends, and both of them were decreased under GADL1 overexpression or lithium treatment but increased upon stimulation with CCL2. Several reports show that cancer cells with the longer perimeter length or a greater value of irregularity can move faster, and thus resulting in the enhanced ability for invasion and metastasis [27][28][29] . The relationship between cell shape and migration in cancer cells is similar to our observations. For both SH-SY5Y and GADL1-overexpressing cells, lithium treatment decreased cell migration and perimeter length. Lithium has been reported to inhibit invasion and migration of glioma cells 30 . Lithium can regulate a cytoskeletal modulator 31 and may also play a role in the neuron migration.
CCL2 has been found to regulate neuron migration 24,25 and can stimulate embryonic hypothalamic neurons to migrate greater distances 25 . For both SH-SY5Y and GADL1-overexpressing cells, stimulation with CCL2 also increased cell number and migration speed but decreased cell thickness. In comparison, CCL2 exposure increased cell area, perimeter length, and irregularity only in GADL1-overexpressing cells, whereas it decreased cell volume only in SH-SY5Y cells. These results indicate that GADL1 overexpression increased cell sensitivity to CCL2 treatment. The findings were echoed with the observation that GADL1 overexpression downregulated CCL2 expression.
It has been reported that taurine retards radial migration of neurons in the developing mouse cerebral cortex 32 . However, we found that GADL1 overexpression affected cell migration, perimeter length, and the differential sensitivity to lithium and CCL2, all of which were dependent mainly on the intracellular form of GADL1. Effects of GADL1 overexpression on cell migration speed or perimeter length were much more obvious in the differences between SH-SY5Y and GADL1-overexpressing cells than in the differences of SH-SY5Y cells cultured in 5Y-CM vs. GADL1-CM. These results suggested that secreted GADL1 or its enzyme product, taurine, in the conditioned medium might exert only mild effects on the observed changes.
A previous study using 3-week-old mice shows that Gadl1 expression is higher in the olfactory bulb 18 , which is an active zone for neuron regeneration in the adult mammalian forebrain. Neuroblasts migrate tangentially along the rostral migratory stream until they reach the olfactory bulb, where they then migrate radially to complete their differentiation into neurons [19][20][21] . The inability of newly generated neurons in the brain to migrate to their target locations might result in improper neural circuitry maintenance and function, and thus might contribute to the emergence of neuropsychiatric disorders, including epilepsy, schizophrenia and autism [33][34][35][36] . Schizophrenia patient-derived cells are less adhesive and more mobile than cells derived from healthy control subjects 37 . Disrupted in schizophrenia1 (DISC1) regulates neuron migration, and loss of function of DISC1 may lead to schizophrenia 33 . Postmortem brains from psychiatric patients show alterations in the polysialylated neural cell adhesion molecule, a protein which has a key role in cell migration 38,39 . The level of GADL1 expression could affect cell migration, indicating that GADL1 might play an important role in the disease development of bipolar disorder.
Exposure to 20 mM lithium decreased the number of both SH-SY5Y and GADL1-overexpressing cells, while treatment with lithium at 1 mM did not have powerful effects on cellular phenotypes for both cells, except on cell area and perimeter length. The possibility that lithium at the concentration of 20 mM is somewhat toxic and may have pleiotropic effects on many biological processes including cell migration could not be ruled out. To correlate the results of this experiment to clinical studies, dose-response analyses, examining the effects of lower concentrations of lithium, would be necessary in the future studies.
We found that GADL1 overexpression decreased cell sensitivity to lithium, as reflected by its effects on cell migration speed. It has been reported that neurons transdifferentiated from bipolar patient skin cells displayed a significantly different cell-adhesion phenotype between lithium responders and nonresponders 26 . These findings together indicate that cell adhesion or migration might serve as a good indicator for the sensitivity to lithium treatment. As compared with the time-consuming and labor-intensive processes to induce neurons from patient skin cells 26 , our GADL1 stable overexpression cell line might therefore be used as a fast screening platform for novel therapeutics for bipolar disorder. It allows high throughput chemical or RNAi-based screens of potential drug candidates for bipolar disorder. The extent of its utility will be clarified by measurement of cell-migration phenotype across different treatments; at minimum, it should facilitate efforts to elucidate mechanism of action of the gold-standard treatment for psychiatric disease.

Methods
Cell culture and establishment of stable cell lines. SH-SY5Y cells, a human neuroblastoma cell line, were grown in Dulbecco's Modified Eagle Medium (DMEM)/F12 medium (1:1) (Life Technologies, USA) supplemented with 10% fetal bovine serum (FBS) (Life Technologies), 2 mM d-glutamine, 100 U/ml penicillin, and 100 μg/ml streptomycin (Life Technologies) at 37 °C in a humidified incubator containing 5% CO 2 . Full-length GADL1 was cloned into a vector bearing an enhanced GFP reporter and the neo gene from Tn5 encoding an aminoglycoside 3′-phosphotransferase, which conferred resistance to G418. The plasmid with GADL1 was transfected into SH-SY5Y cells using Xfect TM transfection reagent (Clontech, USA). At 48 h post-transfection, enhanced GFP-positive cells were sorted and pooled using a FACS Aria II cell sorter (BD Biosciences, USA). The establishment of stable cell lines was achieved by selection with G418 at 300 μg/ml for 2 weeks, after which cells were maintained with 100 μg/ml G418. RNA expression array. Total RNA was extracted SH-SY5Y or GADL1-overexpressing cells pooling from sextuplicate wells using the NucleoSpin RNA/protein isolation kit (MACHEREY-NAGEL, Germany). RNA from each cell line was subjected to single microarray chip analysis. Total RNA (10 µg) was used for cDNA synthesis, and cDNA was labeled via in vitro transcription followed by fragmentation according to the GeneChip Expression Analysis Technical Manual rev5 (Affymetrix, USA). Labeled samples (11 µg each) were hybridized to GeneChip Human Transcriptome Array 2.0 (HTA 2.0; Affymetrix) at 45 °C for 16.5 h. The wash and staining steps were performed with a Fluidic Station-450, and the GeneChip HTA 2.0 was scanned with an Affymetrix GeneChip Scanner 7 G. Gene expression changes for the GADL1-overexpressing cells vs. SH-SY5Y cells were analyzed with GeneSpring software (Agilent, USA) and Ingenuity Pathway Analysis (IPA). Fold changes in gene expression were depicted using Prism 5 software (GraphPad, USA).

Extraction of mRNA from cells and real-time quantitative PCR (RT-qPCR).
Total RNA was extracted SH-SY5Y or GADL1-overexpressing cells pooling from sextuplicate wells using the NucleoSpin RNA/ protein isolation kit (MACHEREY-NAGEL, Germany). The extracted RNA was reverse transcribed into cDNA using a reverse transcription kit (Roche, Switzerland). Expressions of GADL1 (PPH22451A), FN1 (PPH00143B), ITGA2 (PPH00625F), ITGAV (PPH00628C), and CCL2 (PPH00192F) were examined with SYBR Green (Qiagen, Germany) using gene-specific primers (all designed by Qiagen) in triplicates. The primers and probe (Roche) for ACTB were used for the relative quantification of transcription. RT-qPCR was performed with an ABI 7500 system (Applied Biosystems, USA). siRNA knockdown in the GADL1-overexpressing cell line. 24 hr after cell seeding, GADL1-overexpressing cells were transfected with RISC-free negative control siRNA or siRNA targeting GADL1 at 0.1 μM using DharmaFECT1 transfection reagent. Medium was changed 24 hr after transfection. Two days post-transfection, cells from sextuplicate wells were harvested and pooled for subsequent RNA extraction and reverse transcription, followed by RT-qPCR analysis for GADL1, FN1, ITGA2, ITGAV and CCL2. The fold-change value for each gene was normalized to ACTB expression. Digital holographic imaging. Cell migration distance and morphological changes of cells including cell area, thickness, volume, perimeter length, irregularity, and eccentricity were measured using real-time, three-dimensional holographic imaging (HoloMonitor M4; Phase Holographic Imaging, Sweden) 40 . SH-SY5Y or GADL1-overexpressing cells were seeded at 3 × 10 5 cells/well on a laminin-coated 6-well plate (Corning, USA) and maintained in DMEM/F12 (1:1) with 3% FBS inside an incubator at 37 °C in 5% CO 2 . At 4-5 h after plating cells, lithium chloride (1 or 20 mM, Sigma Aldrich, USA) or CCL2 (25 or 50 ng/ml, R&D Systems, USA) was added to the cells for 48 or 72 h, respectively. Images were acquired at 20-min intervals for 48-72 h. The images were analyzed using Hstudio software (Phase Holographic Imaging). The dose of 20 mM lithium used in the experiment was based on a previous study showing that, at this dose, lithium activated MAPK and inhibited GSK-3β in SH-SY5Y cells, with no evidence of cytotoxicity 41 . Besides, 20 mM lithium treatment resulted in an intracellular lithium concentration of 3.2 ± 0.2 mM as measured by a previous study 42 . Collection of conditioned medium (CM). SH-SY5Y and GADL1-overexpressing cells were cultured in DMEM/F12 (1:1) with 3% FBS. After 2-3 days of culture, the resultant CMs (5Y-CM and GADL1-CM) were harvested and sieved using a 0.22 μm filter. SH-SY5Y cells seeded at 3 × 10 5 cells/well on a laminin-coated 6-well plate (Corning) were cultured in the presence of CM vs. fresh complete medium (2:3) and subjected to 2-3 days of continuous real-time holographic imaging. Statistical analysis. Repeated measure ANOVA with Tukey's multiple comparison test in GraphPad Prism was used to analyze the results from holographic imaging between different cell lines or different treatments. The Student's t test was used to compare the results from different batches of experiments.

Data Availability
The RNA expression array datasets generated and analyzed in this study are available from the corresponding authors on reasonable request.