Non-Cell Autonomous Mechanism of Parkinson’s Disease Pathology Caused by G2019S LRRK2 Mutation in Ashkenazi Jewish Patient: Single Cell Analysis

Abstract

Parkinson’s disease (PD) is the second most prevalent neurodegenerative disease, characterized by the loss of the midbrain dopaminergic neurons, which leads to impaired motor and cognitive functions. PD is predominantly an idiopathic disease, however about 5% of cases are linked to hereditary mutations. The most common mutation in both familial and sporadic PD is the G2019S mutation of leucine-rich repeat kinase 2 (LRRK2) with high prevalence in Ashkenazi Jewish patients and in North African Arab patients. It is still not fully understood how this mutation leads to PD pathology. In this study, we derived induced pluripotent stem cells (iPSCs) from an Ashkenazi Jewish patient with G2019S LRRK2 mutation to isolate self-renewable multipotent neural stem cells (NSCs) and to model this form of PD in vitro. To investigate the cellular diversity and disease pathology in the NSCs, we used single cell RNA-seq transcriptomic profiling. The evidence suggests the existence of three subpopulations within the NSCs: a committed neuronal population, intermediate stage population and undifferentiated stage population. Unbiased single-cell transcriptomic analysis revealed differential expression and dysregulation of genes involved in PD pathology. The significantly affected genes were involved in mitochondrial function, DNA repair, protein degradation, oxidative stress, lysosome biogenesis, ubiquitination, endosome function, autophagy and mitochondrial quality control. The results suggest that G2019S LRRK2 mutation may affect multiple cell types in a non-cell autonomous mechanism of PD pathology and that unbiased single-cell transcriptomics holds promise for personalized medicine.

1. Introduction

Parkinson’s disease (PD) is a neurodegenerative disorder characterized by the progressive degeneration of not only the midbrain dopaminergic system but also of other neural systems and organs (Braak et al., 1995; Jellinger, 1991). It is characterized by tremors, rigidity, bradykinesia, akinesia, and non-motor symptoms including cognitive impairment, depression and dementia (Emre et al., 2007; Gelb et al., 1999; Hughes et al., 1992; Jankovic, 2008; Remy et al., 2005). Even though PD is considered an idiopathic disease, at least 5% of cases are hereditary (de Lau and Breteler, 2006). A common genetic mutation found in familial PD was identified in the leucine-rich repeat kinase 2 (LRRK2) gene (Gilks et al., 2005; Healy et al., 2008; Paisan-Ruiz et al., 2004; Zimprich et al., 2004). LRRK2 has known kinase and GTPase activity, with functions linked to transcription, translation, autophagy, mitochondrial function, and vesicular transport (Cookson, 2010; Ho et al., 2009; Imai et al., 2008; Kanao et al., 2010; Migheli et al., 2013; Smith et al., 2005). The most common LRRK2 mutation is the G2019S residue located in the kinase domain (Gilks et al., 2005; Zimprich et al., 2004) in both familial and sporadic PD with high prevalence in Ashkenazi Jewish patients and in North African Arab patients (Lesage et al., 2006; Ozelius et al., 2006) while less common in Asian population (Fung et al., 2006; Gandhi et al., 2009; Kalinderi et al., 2007; Tan et al., 2005; Tan et al., 2007; Tomiyama et al., 2006). This mutation was shown to cause hyper-kinase activity of LRRK2 (Greggio et al., 2006; Luzón-Toro et al., 2007; Smith et al., 2006; West et al., 2005) with a disease phenotype comparable to idiopathic PD (Beilina et al., 2014; Marras et al., 2016). However, it is still unknown how this mutation causes PD (Klein and Westenberger, 2012) and how it impacts the transcriptional profile of neural cells.

RNA sequencing (RNA-seq) applied to a whole cell population or tissue measures the average expression levels for genes across a population of cells in the culture system or tissue. It is useful for comparative transcriptomics and quantifying expression patterns (Mortazavi et al., 2008). As bulk RNA-seq represents a large population of cells, this may lead to bias towards a subpopulation with high expression and therefore be insufficient when dealing with heterogeneous systems (Bengtsson et al., 2005). Single cell RNA-seq (scRNA-seq) is a relatively new technology that overcomes this limitation seen with bulk RNA-seq. Unlike bulk RNA-seq, scRNA-seq measures the distribution of gene expression levels in single cells allowing for the observation of cell-specific transcriptomic changes. We can identify cell types in heterogeneous cell populations, measure the stochastic expression profile of individual cells, study cell heterogeneity, and identify highly variable genes across the sample population (Tang et al., 2009).

In this study we sought to identify unique single cell gene expression signature in induced pluripotent stem cells (iPSC) generated from a PD patient with the G2019S LRRK2 mutation. To explore the effects of this mutation on the expression profile of neural stem cells (NSCs), we performed scRNA-seq on NSCs derived from iPSCs. We utilized single-cell transcriptomics to study the heterogeneity and gene expression signature of differentiated NSCs in an unbiased manner to better understand the disease pathology.

2. Results

2.1. Single-cell sequencing demonstrates existence of 3 cell types in the NSC neurospheres

Induced pluripotent stem cells harboring the G2019S LRRK2 mutation were differentiated into NSCs and processed for single-cell RNA-seq (Fig. 1). The RNA from 67 individual cells was sequenced and aligned to the GRCh38.p10 reference genome. The result was 854 unique transcripts that annotated to 572 genes. Unsupervised hierarchical clustering of the NSCs revealed that the individual cells were grouped into three subpopulations (Fig. 1C). Principal component analysis (PCA) recapitulated the presence of the three subpopulations in the NSC cultures (Fig. 1D). These subpopulations will henceforth be referred to as subpopulation 1 (SP1), subpopulation 2 (SP2) and subpopulation 3 (SP3). SP1 clustered independently from the two other subpopulations SP2 and SP3. Statistical analysis of transcript abundances between the eight cells to all the other samples revealed 140 genes significantly up-regulated and 61 genes significantly downregulated.

Figure 1: Experimental design and single cell analysis of global gene expressions in NSC neurospheres.

(A) Workflow of single cell analysis of iPSCs harboring the LRRK2 G2019S mutation were used to generate NSC neurospheres (Adapted from (Kim and Daadi, 2019b)). The neurospheres were dispersed into single cells and then subjected to single-cell RNA-seq using the C1 single-cell auto prep system. (B) Representative immunostaining of NSCs stained for Beta-tubulin III (TUJ1, red), Nestin (NES, green), SOX2 (red), and Vimentin (VIM, green). NSCs exhibit expression of specific neural progenitor markers. White bar indicates 20 µm. (C) Heatmap reveals three subpopulations after performing hierarchical clustering. The three subpopulations are named SP1 (red circle, n=8), SP2 (blue square, n=30), SP3 (green triangle, n=29). (D) Principal component analysis (PCA) recapitulates the three subpopulations: SP1 (red), SP2 (blue), and SP3 (green). SP1 is clusters separately from SP2 and SP3, which are clustered close to each other.

2.2. Single-cell transcriptional profile of LRRK2 NSC neurospheres reveals subclasses of neural lineages with gene signature of PD.

We then correlated cluster gene signatures between SP1 and SP2, 139 genes were significantly upregulated and 74 genes were significantly downregulated (Fig. 2A). Of the 213 differentially expressed genes, 23 genes overlapped with neuronal differentiation including NAV3, NEDD4, and IGSF9 (Fig. 2A, Table 1). While 37 genes coincided with disease phenotype, including MT-ND4 and SLC25A5 (Fig. 2A). Genes implicated for disease phenotype were chosen based on genes present in the KEGG pathway for PD, as well as genes involved in Golgi apparatus (Lin et al., 2009), mitochondrial function (Cooper et al., 2012), ubiquitin-mediated proteolysis (Pan et al., 2008), and vesicular transport (Biskup et al., 2006; Shin et al., 2008). Among the 23 neuronal genes (Fig. 2B), 18 were upregulated and 5 were downregulated in SP1 when compared to SP2 (Fig. 2D). Of the 37 disease phenotype genes (Fig. 2E, Table 1), 26 were upregulated and 11 were downregulated (Fig. 2G). In both gene sets, SP1 and SP2 clustered separately during unsupervised hierarchical clustering (Fig. 2B and 2E). These data suggests that SP1 had more of a neuronal identity.

Figure 2: Comparative analysis between NSC-derived SP1 and SP2.

(A) Volcano plot reveals 139 upregulated genes and 74 downregulated genes where the fold change threshold is set at 2 and p<0.05. (B-D) Differential expression analysis using neuronal gene set. (B) Heatmap of gene expression between SP1 (red circle) and SP2 (blue square) after hierarchical clustering of samples according to specific gene sets. (C) Violin plot of the relative expression levels of the neuronal genes of SP1 (red) and SP2 (blue). (D) Bar graphs of log fold change of neuronal genes displaying upregulated (blue) and downregulated (red) expression. (E-G) Differential expression analysis using disease phenotype gene set. (E) Heatmap of gene expression between SP1 (red circle) and SP2 (blue square) after hierarchical clustering of samples according to specific gene sets. (F) Violin plot of the relative expression levels of disease phenotype genes SP1 (red) and SP2 (blue). (G) Bar graphs of log fold change of disease phenotype genes displaying upregulated (blue) and downregulated (red) expression.

Table 1.

Comparison of Log₂ expressions of statistically significant differential expression (based on the p value) of neuronal and disease phenotype genes between SP1 and SP2 and fold change (Log₂) differences. Downregulated (red) and upregulated (green) genes are determined by fold change values.

Comparing SP1 to SP3, 193 genes were differentially expressed with 150 genes significantly upregulated and 43 genes significantly downregulated (Fig. 3A). From these genes, 25 exhibited neuronal function (example: NEGR1, ASTN2, NEDD4, Fig. 3A) and 36 were related to relevant PD phenotype, including mitochondrial functions (example: MT-CO2 and MT-ND1, Fig. 3A). SP1 clustered separately from SP3 with the neuronal gene set (Fig. 3B), with one gene downregulated (NRSN1) and 24 were upregulated, such as SRRM4 and SHISA9 (Fig. 3C, 3D, Table 2). Among the 36 disease phenotype genes (Fig. 3E, Table 2), 10 were downregulated, including NDUFA11 and UBE2V1, which are involved in mitochondrial function and DNA repair, respectively (Berger et al., 2008; Sancho et al., 1998) while 26 genes were upregulated, including CLN5, SOD2 and BLOC1S6 (Fig. 3G, Table 2) which are involved in protein degradation, oxidative stress and lysosome biogenesis, respectively (Isosomppi et al., 2002; Setty et al., 2007; Zelko et al., 2002).

Figure 3: Comparative analysis between NSC-derived SP1 and SP3.

(A) Volcano plot reveals 150 upregulated genes and 43 downregulated genes where the fold change threshold is set at 2 and p<0.05. (B-D) Differential expression analysis using neuronal gene set. (B) Heatmap of gene expression between SP1 (red circle) and SP3 (green triangle) after hierarchical clustering of samples according to specific gene sets. (C) Violin plot of the relative expression levels of the neuronal genes of SP1 (red) and SP3 (green). (D) Bar graphs of log fold change of neuronal genes displaying upregulated (blue) and downregulated (red) expression. (E-G) Differential expression analysis using disease phenotype gene set. (E) Heatmap of gene expression between SP1 (red circle) and SP3 (green triangle) after hierarchical clustering of samples according to specific gene sets. (F) Violin plot of the relative expression levels of disease phenotype genes of SP1 (red) and SP3 (green). (G) Bar graphs of log fold change of disease phenotype genes displaying upregulated (blue) and downregulated (red) expression.

Table 2.

Comparison of Log₂ expressions of statistically significant differential expression (based on the p value) of neuronal and disease phenotype genes between SP1 and SP3 and fold change (Log₂) differences. Downregulated (red) and upregulated (green) genes are determined by fold change values.

We identified 42 genes differentially expressed between SP2 and SP3. Thirty-six genes were significantly upregulated and six genes were significantly downregulated (Fig. 4A). Eight genes exhibited neuronal function, such as NEFM and NETO2 (Fig. 4B) and all were upregulated in SP2 (Fig. 4D, Table 3). Seven genes were related to disease phenotype (Fig. 4E, Table 3) with four of them upregulated in SP2, including CALB1, SLC25A29 and GBF1 which are involved in mitochondrial amino acid transport and vesicular trafficking, respectively (Claude et al., 1999; Porcelli et al., 2014). While downregulated genes included HUWE1 (ubiquitination), COPB1 (endosome function and autophagy) and BNIP3L (mitochondrial quality control) (Fig. 4G, Table 3) (Razi et al., 2009; Schweers et al., 2007; Zhao et al., 2008). SP2 and SP3 did not distinctively cluster in the heatmap when comparing expression of neuronal and disease phenotype genes (Fig. 4B and 4E).

Figure 4: Comparative analysis between SP2 and SP3.

(A) Volcano plot reveals 49 upregulated genes and 6 downregulated genes where the fold change threshold is set at 2 and p<0.05. (B-D) Analysis using neuronal gene set. (B) Heatmap of gene expression between SP2 (blue square) and SP3 (green triangle) after hierarchical clustering of samples according to specific gene sets. (C) Violin plot of the relative expression levels of the neuronal genes of SP2 (blue) and SP3 (green). (D) Bar graphs of log fold change of neuronal genes displaying upregulated (blue) and downregulated (red) expression. (E-G) Analysis using disease phenotype gene set. (E) Heatmap of gene expression between SP2 (blue square) and SP3 (green triangle) after hierarchical clustering of samples according to specific gene sets. (F) Violin plot of the relative expression levels of disease phenotype genes of SP2 (blue) and SP3 (green). (G) Bar graphs of log fold change of disease phenotype genes displaying upregulated (blue) and downregulated (red) expression.

Table 3.

Comparison of Log₂ expressions of statistically significant differential expression (based on the p value) of neuronal and disease phenotype genes between SP2 and SP3 and fold change (Log₂) differences. Downregulated (red) and upregulated (green) genes are determined by fold change values.

Gene ontology (GO) term enrichment analysis was performed using DAVID functional annotation. The genes upregulated in SP1 when comparing to SP2 were enriched for neuron development (Fig. 5A). The genes within the GO for neuron development were LST1, NEDD4, OPHN1, IGSF9, CLN5 and SOD2. This same enrichment was observed when comparing SP1 to SP3, albeit to a greater extent (Fig. 5B). The genes within the neuron development GO were similar except for the addition of CELSR2 and LAMB2. There was no significant gene set enrichment when comparing SP2 to SP3.

Figure 5: GO term enrichment analysis using DAVID functional annotation.

(A) Gene set of all significantly upregulated genes in SP1 compared to SP2 was used. Gene ontology selected was based on p<0.05. (B) Gene set of all significantly upregulated genes in SP1 compared to SP3 was used. Gene ontology selected was based on p<0.05.

3. Discussion

Differentiation of neural lineages from pluripotent stem cells is hindered by a lack of reproducibility and a limited understanding of the cellular composition that constitutes the neural stem cell population. We have established a defined method to reproducibly generate neural stem cells from iPSCs (Daadi, 2019). Single cell RNA-seq is a powerful tool to understand the cellular diversity and identify the composition of neural lineages differentiated from iPSCs. We have demonstrated that single cell analysis of a population of NSCs differentiated from human iPSCs harboring the G2019S LRRK2 mutation revealed three subpopulations with statistically significant differential expression patterns (based on the p value, see Method Section). The significant upregulation of neuronal genes in SP1 when compared to SP2 and SP3 suggest that SP1 is more committed in the neuronal lineage than the other two subpopulations. SP2 also had significant upregulation in a few neuronal genes, which suggests that it represents an intermediate subpopulation between SP1 and SP3. There was an increase in differential expression of disease related genes in SP1. Genes present in the KEGG pathway for Parkinson’s disease were significantly upregulated in SP1. Together, these data suggest that the SP1 is a more differentiated neuronal population with susceptibility to express the disease phenotype. The SP2 sub-population expressed genes with neuronal function and disease phenotype, however SP2 segregated closer with SP3 than SP1. SP3 did not show neuronal phenotype and instead showed more of an undifferentiated state. These findings suggest that SP2 may represent a transient neuronal population between the undifferentiated state SP3 and the differentiated state SP1. However, we cannot rule out the possibility that SP2 and SP3 could be 2 different subpopulations, undifferentiated NSCs and glial cells, as it has been observed that quiescent NSCs and astrocytes could not be distinguished with PCA (Dulken et al., 2017).

The G2019S mutation is associated with hyper-kinase activity of LRRK2 (Greggio et al., 2006; Luzón-Toro et al., 2007; Smith et al., 2006; West et al., 2005), which results in cellular toxicity in vitro (Iaccarino et al., 2007; Smith et al., 2005). A recent study demonstrated that this mutation does not impact dopaminergic neurons identity in iPSC-derived neural cells (Sandor et al., 2017). Neurite shortening, however is a common phenotype seen with G2019S neurons (Borgs et al., 2016; Macleod et al., 2006; Marrone et al., 2018; Plowey et al., 2008; Sánchez-Danés et al., 2012; Winner et al., 2011). Neurensin 1(NRSN1) is believed to regulate neurite extension (Araki et al., 2002; Ida et al., 2004). It was significantly downregulated in SP1 when compared to both SP2 and SP3. A downregulation of NRSN1 in the SP1 subpopulation may be indicative of the neurite shortening exhibited in G2019S neurons. Serine/Arginine Repetitive Matrix 4 (SRRM4) is another gene that regulates neurite extension (Ohnishi et al., 2017). However, SRRM4 was significantly upregulated in SP1. This may be due to a possible compensatory mechanism between SRRM4 and NRSN1, which could explain the discrepancies in the results obtained on the neurite outgrowth assaying the LRRK2 role in neurite extension (Garcia-Miralles et al., 2015). Further studies addressing interactions between NRSN1 and SRRM4 would shed light on the effects of G2019S mutation on neurite outgrowth. Interestingly, one of the genes upregulated in SP1 is Moesin (MSN), which is a known target of LRRK2 phosphorylation (Jaleel et al., 2007). MSN is part of the ERM protein family. It has been previously shown that there is a significant increase in pERM in G2019S hippocampal neurons, which correlates with neurite shortening (Parisiadou et al., 2009).

Mitochondrial dysfunction and perturbed oxidative phosphorylation is another hallmark phenotype of G2019S mutation. Specifically, complex I deficiency has been noted (Banerjee et al., 2009; Hanagasi et al., 2005; Schapira et al., 1990). SP1 exhibited a significant up-regulation in mitochondrial genes: MRPS5, MORC4, MT-CO2, MT-CO3, MT-CYB, MT-ND1, MT-ND4, MT-ATP8, and MT-ATP6. Similar expression profile of mitochondrial genes had been previously reported with substantia nigra tissue samples from PD patients (Noureddine et al., 2005). The increased expression of mitochondrial genes may be suggestive of oxidative stress. These findings are consistent with a recent report describing the correlation between transcriptional profiles of dopaminergic neurons derived from iPSCs carrying LRRK2-G2019S mutation and those induced by rotenone exposure (Sandor et al., 2017), a pesticide known to cause Parkinson’s disease through inhibition of mitochondrial respiratory chain complex I (Sherer et al., 2003). Besides complex I deficiency, there are also reports of complex IV, cytochrome c oxidase (COX), deficiency in Parkinson’s disease patients (Bindoff et al., 1991; Blin et al., 1994). The COX subunits MT-CO2 and MT-CO3 both exhibited a significant upregulation in SP1. On the other hand, COX6A1 exhibited a significant downregulation in SP1. COX6A1 deficiency has been shown to lead to a reduction in COX activity (Tamiya et al., 2014). Furthermore, COX biogenesis is dependent on copper. SLC25A3 is a phosphate transporter that also acts as a copper transporter. Knockdown of SLC25A3 resulted in deficiency in COX activity (Boulet et al., 2017). Our data also exhibits a downregulation of SLC25A3 in SP1. Together with COX6A1 downregulation, our data suggest COX deficiency within SP1. Other critical PD-associated genes that were down-regulated in SP2 vs SP3 included HUWE1 involved in ubiquitination and subsequent degradation of proteins, COPB1, involved in autophagy by playing a role in early endosome function and BNIP3L involved in mitochondrial quality control. Interestingly, BNIP3L has also been shown to be a substrate for PARK2, promoting mitophagy through the PINK1/PARK2 pathway (Gao et al., 2015). Together, these data suggest that the iPSC-LRRK2-derived NSCs express the critical pathways affected in PD.

The unbiased single-cell transcriptomics analysis of the iPSC-LRRK2 G2019S mutant NSC progeny revealed that the significantly affected genes were involved in mitochondrial function, DNA repair, protein degradation, oxidative stress, lysosome biogenesis, ubiquitination, endosome function, autophagy and mitochondrial quality control. This study also suggests that G2019S LRRK2 mutation may affect multiple cell types in a non-cell autonomous mechanism of the disease pathology. However, further functional analysis of the identified genes would present stronger evidence of their potential role in the LRRK2-mediated PD pathology; and the comparison between mutant and gene-corrected isogenic LRRK2 NSCs may reveal additional relevant genes. The differentially expressed genes reported here are promising targets to be further explored in dissecting the mechanisms mediating disease pathology and for validation as potential therapeutic targets.

4. Experimental Procedure

4.1. Induced pluripotent stem cell (iPSC) cultures:

IPSCs-lrrk2 were derived from the fibroblasts of a 52 year old PD male patient of Ashkenazi Jewish descent (ND29802, RRID:CVCL_DD50) with heterozygous G2019S LRRK2 mutation pro cured from the NINDS Repository at the Coriell Institute for Medical Research (Camden, NJ, USA). A frozen vial of the patient’s fibroblasts was thawed in 9 ml of fibroblast media (10% Fetal Bovine Serum/DMEM high glucose) and centrifuged at 1000 rpm 5 min at room tem perature. The fibroblasts were re-suspended in fresh media and cultured then expanded for 4 passages. A stock was frozen and 1 was transfected with the reprogramming factors. We used the non-integrative episomal vector system, which employs 4 episomal vectors: pCXLE-hOCT3/4- shp53, pCXLE-hSK, pCXLE-hUL and pCXWB-EBNA1 (Addgene, Cambridge, MA, USA). The vectors were amplified in bacterial culture and purified using the Miniprep kit (Qiagen, Hilden, Germany) ac cording to the manufacturer’s instructions. The advantage of EBNA1-based episomal reprogramming system is the use of non-integrating vectors with stable extrachromosomal replication. The vector and re programming gene sequences are cleared from the cells as the iPSC colonies are expanded. Eighty-two microliters of NHDF Nucleofector solution were mixed with 18 μl of Supplement and added with 0.83 μg of each pCXLE-hOCT3/4-shp53, pCXLE-hSK, pCXLE-hUL and 0.5 μg of pCXWB-EBNA1. The fibroblasts were dissociated into single cells and collected by centrifuge 1000 rpm for 5 min at room temperature. The pellet was resuspended with the DNA-nucleofection mixture and ap plied to the U-023 program on the Nucleofector 2b device. The reprogrammed fibroblasts were cultured on mouse embryonic fibroblast feeders with expansion media consisting of DMEM-F12 (Gibco, Thermo Fisher Scientific, Waltham, MA, USA), 20% knockout serum replacement (Gibco, Thermo Fisher Scientific, Waltham, MA, USA), 10 ng/ml basic fibroblast growth factor (bFGF, Stemgent, Cambridge, MA, USA), 1% non-essential amino acids 100x (Gibco, Thermo Fisher Scientific, Waltham, MA, USA), 0.5% L-glutamine (Gibco, Thermo Fisher Scientific, Waltham, MA, USA) and 0.14% 2-mercaptoethanol (Thermo Fisher Scientific, Waltham, MA, USA) solution. The media was changed daily and cells were passaged once every 5 days. At about three weeks after transfection, iPSC colonies with human ES cell-like morphology were picked. A total of 24 clones were chosen from the transfected fibroblasts. Each clone was independently plated in a well of a 24-well plate, expanded for 3 passages, and banked. We expanded 2 clones of iPSCs with the G2019S LRRK2 mutation for 6 passages and generated a seed bank of about 25 cryopreserved vials each. The iPSCs were further expanded for multiple passages and characterized for the expression of the pluripotency markers, Oct4, Nanog, TRA160 and SSEA4. The IPSC-lrrk2 demonstrated stable growth and normal male Karyotype (Supplement figure).

Neural stem cell (NSC) cultures:

Self-renewable NSCs were isolated from iPSCs-LRRK2 using NN1 media (NeoNeuron, San Antonio, TX, USA) and grown as neurospheres as we previously described (Daadi and Weiss, 1999). Confluent iPSC colonies were mechanically detached from culture dish with a cell lifter (Fisher Scientific, Hampton, NH, USA) and re-suspended in the chemically defined media, NN1 (NeoNeuron, San Antonio, TX, USA), supplemented with 20 ng/ml bFGF (Stemgent, Cambridge, MA, USA) and 20 ng/ml epidermal growth factor (EGF, EMD Millipore, Burlington, MA, USA). After 7 days in culture all floating neurospheres were collected by centrifugation and dissociated into single cells using accutase (Gibco, Life technologies, NY, USA), re-suspended in fresh NN1 culture media and plated in T-75 cell culture flasks for expansion (Corning, Oneonta NY, USA).

Karyotype analysis (Fig. 6):

Figure 6: Karyotype of the iPSC-LRRK2.

Cytogenetic evaluation of the iPSC-LRRK2 line at passage 35 by standard G-banding was performed. Twenty metaphase cells were analyzed and showed a normal male chromosome complement (46,XY).

Karyotyping of the iPSC line was performed as we previously reported (Daadi et al., 2008). IPSC cultures confluent at 75% were incubated at 37°C and harvested for metaphase chromosomes. Metaphase chromosomes were obtained by standard chromosome harvest methods by exposure to Colcemid at 0.1 mg/ml for 1 hour at 37°C, a 2-minute exposure to trypsin/ EDTA, hypotonized with 0.057 M KCl and fixed with 3:1 methanol:acetic acid. Slide preparations were made by dropping the fixed cell pellet onto cold, wet slides and air-dried. After incubating the slides at 90°C for 30 minutes, chromosomes were trypsin banded and then Wright/Giemsa stained for G-banding analysis. Twenty metaphase cells were completely analyzed and a normal male chromosome complement was found (46,XY) (Fig. 6).

Immunocytochemistry:

Neurospheres were plated on poly-L-ornithine coated coverslips for three hours and were fixed in 10% formalin for 15 minutes (Daadi and Weiss, 1999). The coverslips were washed with phosphate-buffered saline (PBS) three times for 5 minute each. Primary antibody solutions were made with 0.03% Triton X diluted in PBS and supplemented with 10% normal donkey serum. Coverslips were incubated with primary antibodies: β-tubulin Class-III (1:1000, mouse, Sigma-Aldrich, St. Louis, MO, USA), Nestin (1:100, rabbit, EMD Millipore, Burlington, MA, USA), SOX2 (1:100, rabbit, Abcam, Cambridge, United Kingdom), or Vimentin (1:100, mouse, EMD Millipore, Burlington, MA, USA) at 37 °C for 2 hours. Coverslips were then washed with PBS three times for 5 minute each and incubated in secondary antibodies made in PBS and 10% normal donkey serum: Donkey anti-rabbit IgG FITC (1:100, Jackson ImmunoResearch, West Grove, PA, USA), donkey anti-mouse IgG CY3 (1:100, ImmunoResearch, West Grove, PA, USA) and DAPI (1:100, Molecular Probes, Thermo Fisher Scientific, Waltham, MA, USA) for 1 hour at room temperature in the dark. The coverslips were then washed a final three times with PBS, mounted onto glass microscope slides and coverslipped using Fluorsave (Calbiochem, Burlington, MA, USA). Fluorescent microscopic images were taken using Zeiss LM-800 confocal microscope.

Single cell isolation and generation of cDNA library:

NSCs from passage 4 were dissociated into single cells with Accutase (STEMCELL Technologies, Vancouver, British Columbia, Canada), suspended in NN1 media supplemented with C1 Suspension Reagent (3:2 ratio) and loaded into the prepared C1 IFC chip (Fluidigm, South San Francisco, CA, USA) as we previously described in detail (Kim and Daadi, 2019b). The IFC chip was prepared following Fluidigm’s protocol. Seventy viable cells were successfully captured among the 96 possible capture sites. RNA was extracted from these cells and cDNA libraries were generated using SMART-Seq v4 (Clontech, Takara Bio, Kusatsu, Shiga Prefecture, Japan). The concentration of cDNA libraries was quantified using the Qubit 3 fluorometer (Thermo Fisher Scientific, Waltham, MA, USA). One sample was removed due to low cDNA concentration. Remaining libraries were diluted to a range of 100–300 pg/μl and indexed with Nextera XT DNA Library Preparation kit (Illumina, San Diego, CA, USA). Indexed libraries were then pooled and quality assessed with the TapeStation 4200 (Agilent, Santa Clara, CA, USA). Dual band clean-up of the cDNA was performed using AMPure XP beads (Beckman Coulter, Brea, CA, USA), providing us with length of 300–500 bp and final cDNA concentration of 7 nM. The samples were then sequenced using the Illumina HiSeq 3000 at 100 bp paired-end reads. An average of 4 million reads per sample was obtained.

RNA-seq data analysis:

We recently reported a detailed protocol describing bioinformatics RNA-seq raw data analysis (Kim and Daadi, 2019a). Raw sequencing data was quality controlled using FASTQC and quality trimmed using BBDuk of BBtools, where PHRED score > 28. Two samples with poor quality were removed, resulting in a final number of 67 samples. Trimmed sequences were then pseudoaligned using kallisto (Bray et al., 2016) (ver. 0.43.0) with 100 bootstraps mapping to the human GRCh38.p10 reference genome from Ensembl (release 88) (Lander et al., 2001; Zerbino et al., 2018). The resultant bootstrapped transcript abundances, in transcripts per million, were quantified and filtered using Sleuth (Pimentel et al., 2017). Sleuth is a downstream program of Kallisto used to quantify the bootstraps. As Sleuth requires at least two conditions in the samples to run, samples were arbitrarily divided into two groups. This does not affect any subsequent downstream analysis as Sleuth was used only to obtain transcript abundances. During this step, the zero read transcripts were filtered out of our data set. R (ver. 3.4.1) was used for the analysis of the transcript abundances. The R package SINGuLAR (ver. 3.6.2, Fluidigm, South San Francisco, CA, USA) was used to explore the heterogeneity of the cell population through differential expression analysis. Principal component analysis and unsupervised hierarchical clustering was performed using this package. Fold change threshold was set to 2 (default parameter) and p-value < 0.05 when determining significantly differentially expressed genes between subpopulations. GO term enrichment analysis was performed using the DAVID Bioinformatics Resources with default human genome as background (Huang da et al., 2009; Huang et al., 2008) (ver. 6.7, Leidos Biomedical Research, Frederick, MD, USA).

Highlights.

• Single-cell transcriptomic enables unbiased biological discovery

• Differential gene expression revealed dysregulation in genes involved PD pathology

• Affected genes involves mitochondrial function, DNA repair, protein degradation

• The G2019S mutation affects multiple cell types in a non-cell autonomous mechanism

• Unbiased single-cell transcriptomics holds promise for precision medicine