Sporadic Parkinson’s disease derived neuronal cells show disease-specific mRNA and small RNA signatures with abundant deregulation of piRNAs

Induced pluripotent stem cells (iPSC) were generated from eight sporadic PD patients and six healthy controls by retroviral transduction of the four transcription factors (Oct 3/4, c-Myc, Sox2 and Klf4) as previously described [15]. The Institutional Review Board approval (Nr. 4120: Generierung von humanen neuronalen Modellen bei neurodegenerativen Erkrankungen) and informed consent forms are on file at the movement disorder clinic at the Department of Molecular Neurology, Universitätsklinikum Erlangen (Erlangen, Germany). All the hiPSC lines were cultured in mTeSR (Stemcell Technologies) or MACSbrew (Miltenyi) on Geltrex (Gibco® Thermo Fisher Scientific) and splitted every five to seven days using Collagenase IV or Gentle Cell Dissociation Reagent (Stemcell Technologies). Pluripotency and a stable karyotype were tested by flow cytometry and G-banding chromosomal analysis, respectively. One to two hiPSC clones per individual (Additional file 1: Table S1) were differentiated into small molecule neural precursor cells (smNPCs) following the protocol published by others [50] with some adaption as described in [59]. The smNPCs were further differentiated to midbrain neurons within three weeks of maturation [50, 59]. Briefly, 70% confluent iPSC were detached by collagenase IV (Gibco® Thermo Fisher Scientific) treatment for 20 min at 37 °C, 5% CO₂. Cell colonies were cultured as free-floating aggregates in human embryonic stem cells (hESC) medium (80% KO-DMEM, 20% KO serum replacement, 1% non-essential amino acids, 1% Penicillin/Streptavidin (all from Thermo Fisher Scientific), 1 mM ß-Mercaptoethanol (Sigma-Aldrich) supplemented with the small molecules 1 μM LDN (Stemgent), 10 μM SB, 3 μM Chir, and 0.5 μM Purmorphoamine (PMA, all from Tocris) on ultra-low adhesion plates. After two days of incubation at 37 °C, 5% CO2, the cell colonies were centred and the medium was changed to N2B27 medium (50% DMEM/F12, 50% Neurobasal Medium, 1:200 N2, 1:100 B27 (all from Thermo Fisher Scientific) supplemented with the same small molecules. On day four, the medium was changed to smNPC medium (N2B27 medium supplemented with 3 uM Chir, 0.5 uM PMA and 150 uM Ascorbic acid (AA; Sigma-Aldrich). After a total of six days of suspension culture, cell colonies were replaced on geltrex-coated (Gibco® Thermo Fisher Scientific) 12-well plates in smNPC medium supplemented with Rho kinase inhibitor Y27532 (RI, Axxora) for 24 h. Medium was changed every other day and cells were passaged once a week by accutase treatment. After at least five passages, smNPCs were differentiated into MN. Therefore, two days after passaging, the medium was exchanged to N2B27 medium supplemented with 100 ng/ml FGF8 (Peprotech), 1 μM PMA and 200 μM AA. On day 10 of differentiation, medium was supplemented with 100 ng/ml FGF8, 10 ng/ml GDNF (Peprotech), 10 ng/ml TGFb (eBioscience), 200 uM AA, and 500 μM Dibutyryl-cAMP (dbcAMP; Sigma-Aldrich. On the next day, cells were passaged at ratios of 1:2–1:3 as single cells after accutase treatment (Sigma-Aldrich), plated onto geltrex-coated four-well chamber slides (Ibidi) or 12-well plates and further cultured for at least two weeks in maturation medium (N2B27 medium plus 100 ng/ml FGF8, 10 ng/ml GDNF (Peprotech), 10 ng/ml TGFb (eBioscience), 200 uM AA, and 500 μM Dibutyryl-cAMP (dbcAMP; Sigma-Aldrich) with two times media change per week.

Libraries for next-generation sequencing were prepared from 1 μg total RNA with the TrueSeq RNA library preparation kit v2 according to the manufacturer’s instructions (Illumina, San Diego, CA, USA). Briefly, poly-A RNA was purified from the total RNA preparation with magnetic oligo-dT beads. The RNA-bead mixture was incubated at 65 °C for five minutes to denature the RNA. Then, the mixture was incubated for five minutes at RT to allow the RNA to bind to the beads. Afterwards, the beads were washed with bead washing buffer and resuspended in elution buffer for two minutes at 80 °C. Bead binding buffer was added to allow rebinding of the eluted RNA. The beads were washed with bead washing buffer and resuspended in elute, prime, fragment mix. The RNA was eluted at 94 °C for eight minutes, and first strand cDNA synthesis was performed with SuperScript II (Thermo Fisher Scientific, Waltham, MA, USA) in first strand mix supplied by Illumina. Second strand synthesis was performed with the second strand mix for one hour at 16 °C and the resulting double stranded cDNA was purified using AMPure XP beads (Beckman Coulter, Brea, CA, USA). Then, end repair was performed for 30 min at 30 °C with the provided end repair mix. The end repaired DNA was again cleaned up with AMPure XP beads. A-tailing was performed with the A-tailing mix at 37 °C for 30 min followed by 70 °C for five minutes. Indexed adapters were ligated to the end-repaired A-tailed cDNA at 30 °C for ten minutes. Next, stop ligation buffer was added and the libraries were cleaned up with AMPure XP beads. PCR amplification was performed with the provided PCR reagents and the following cycling conditions: Denaturation at 98 °C for 30 s and then 15 cycles of1) 98 °C for 10 s, 2)60 °C for 30 s and 3)72 °C for 30 s. Afterwards, a final extension at 72 °C for five minutes was performed and the amplified libraries were purified again with AMPure XP beads. Finally, quality control was performed with a Bioanalyser® (Agilent).

For all tissue samples (RIN usually < 8), we used the TrueSeq RNA Access Kit (Illumina) to prepare libraries from 100 ng total RNA, which is suitable for degraded RNA. First, first strand cDNA was synthetized with the Elute, Prime, Fragment High Mix and super script II in First Strand Master Mix (25 °C for 10 min, 42 °C for 15 min and 70 °C for 15 min). Then, 20 μl Second Strand Marking Master Mix were added and the second strand was synthetized for 1 h at 16 °C. The cDNA was cleaned up with AMPure XP beads and A-tailing was performed with A-tailing mix (37 °C for 30 min followed by 5 min at 70 °C). Following this, adapter ligation was performed with Ligation Mix for 10 min at 30 °C after which the reaction was stopped with Stop Ligation Buffer. The libraries were cleaned up with AMPure XP beads and amplified with PCR Master Mix and PCR Primer Cocktail (98 °C for 30 s, 15 cycles of 98 °C for 10 s, 60 °C for 30 s, 72 °C for 30 s and 72 °C for 5 min). Following a further clean up with AMPure XP beads, the libraries were quantitated with a Bio-Analyser and 4 libraries were pooled at equal concentrations (200 ng each) for exome capture, which was repeated twice. For exome capture, Coding Exome Oligos were added to the pooled libraries together with Capture Target Buffer 3 and incubated for 95 °C for 10 min and 18 cycles of one minute incubations, starting at 94 °C, then decreasing 2 °C per cycle with a final incubation for 58 °C for 90 min. Immediately afterwards, the hybridized probes were captured with streptavidin magnetic beads for 25 min at RT and washed twice with Enrichment Wash Solution (incubation at 50 °C for 20 min). Finally an Elution Premix was prepared with Enrichment Elution Buffer 1 and NaOH. The streptavidin beads were resuspended in this Elution Premix, incubated for two minutes at RT, the supernatant was separated from the beads on a magnetic stand and Elute Target Buffer 2 was added to the supernatant. After the second exome capture, the libraries were cleaned up with AMPure XP beads. Finally, a second enrichment was performed with the Enhanced PCR Mix (98 °C for 30 s followed by 10 cycles of 98 °C for 10 s, 60 °C for 30 s and 72 °C for 30 s, with a final extension at 72 °C for 5 min) and the amplified libraries were purified again with AMPure XP beads. Finally, quality control was performed with a Bioanalyser® (Agilent).

Reduced representation bisulfite sequencing allows for highly accurate and efficient analysis of methylation patterns at single base pair resolution with a focus on CpG islands [21]. Digestion was performed with 2.5 μg gDNA in buffer 4 (NEB) and 400 units MSPI O/N at 37 °C. Afterwards, the DNA was purified from this reaction with the PCR purification Kit (Qiagen) according to the manufacturer’s instructions. Briefly, 5 volumes buffer PB were added to the MSPI digest, applied to a spin column provided in the kit and centrifugation was performed for 1 min at 9500 × g. Then, the column was washed with 750 μl buffer PE, dried by centrifugation for 1 min at 9500 × g and the DNA was eluted with 30 μl buffer EB. Library preparation was then performed with the NEXTflex™ Bisulfite Library Prep Kit (BIOO Scientific) according to the manufacturer’s instructions with some modifications. Briefly, end repair was performed with 500 ng digested, purified DNA in end repair buffer mix and end repair enzyme mix in a total volume of 50 μl. The reaction was incubated at 22 °C for 30 min and then cleaned up with the MinElute® PCR Cleanup Kit. Then, 16.5 μl of the eluate were mixed with 4.5 μl of adenylation mix and the reaction was incubated for 30 min at 37 °C. Afterwards, 31.5 μl ligation mix and 2.5 μl of individual adapters (diluted 1:2) were added, and adapter ligation was performed for 15′ at 22 °C. Afterwards, the DNA was cleaned with AMPure XP beads and size selection for fragments from 175 to 400 bp was performed with a gel purification step. The libraries were separated on a 2% low melt agarose gel (Sigma-Aldrich), the cut out gel fragments were dissolved for 10 min at RT in DNA binding buffer and 150 μl ethanol were added. Then, the solution was applied to a clean-up spin column and centrifuged at 18500 xg until the complete volume was processed. Afterwards, the column was washed twice with DNA wash buffer, dried by centrifugation and the DNA was eluted with column elution buffer. Then, bisulfite conversion of the DNA was performed with the EZ Methylation Gold Kit (Zymo Research) according to the manufacturer’s instructions. Briefly, 130 μl conversion reagent were added to 20 μl purified DNA. The reaction was incubated for 10 min at 98 °C and for 2.5 h at 64 °C. Then, the samples were loaded on spin columns containing 600 μl M-Binding buffer and mixed by inverting. The DNA was bound to the column by centrifugation for 30 s at 18620 x g. Then, the column was washed with 100 μl wash buffer, and 200 μl desulphonation buffer were added. The desulphonation buffer was incubated for 17 min at RT, and then removed by centrifugation. The column was washed twice with 200 μl wash buffer, and dried by centrifugation for 10 s. Finally, 17 μl elution buffer were added to the column, incubated for 1 min and the DNA was eluted by centrifugation. Afterwards, PCR amplification of the bisulfite converted libraries was performed with PfuCx Hot Start (Agilent). 15 μl DNA were amplified with the NEXTflex™ Primer Mix. Cycling conditions were: Initial denaturation for five minutes at 95 °C. Then 18 cycles of 95 °C for 30 s, 65 °C for 30 s and 72 °C for 45 s. Afterwards a final extension at 72 °C for 7 min. The libraries were purified again using AMPure XP beads (Beckmann Coulter). Finally, quality control with a Bioanalyser® (Agilent) was performed.

Small RNA libraries were prepared from 1 μg total RNA containing small RNAs with the TrueSeq Small RNA kit (Illumina) according to the manufacturer’s instructions. Briefly, a 3’adapter was added to the RNA, denaturation was performed for two minutes at 70 °C and afterwards the mixture was put on ice immediately. Then, the adapter was ligated with a T4 RNA Ligase2 deletion mutant (epicentre) for 1 h at 28 °C. Then the reaction was stopped with stop solution for 15 min at 28 °C, the previously denatured 5′ adapter was added together with ATP and T4 DNA Ligase and ligated to the RNA for 1 h at 28 °C. After that, cDNA synthesis was performed with super script II and Illumina supplied dNTPs for 1 h at 50 °C. Afterwards, the cDNA was amplified and indexed with the primers and PCR mix supplied in the kit (eleven cycles of denaturation at 98 °C for 10 s, annealing 60 °C for 30 s and extension at 72 °C for 15 s with a final extension at 72 for 10 min) and size selection was performed on 5% TBE acrylamide gels (Bio-Rad, Munich, Germany). Here, the region marked by Illumina’s custom ladder between 145 bp and 160 bp were cut out and pooled for sequencing. The gel was homogenized by centrifugation at 20000 x g through a gel breaker tube (Bio-Cat) and 300 μl ultrapure H₂O were added to elute the DNA O/N in a DNA LoBind tube (Eppendorf). The following day, the gel debris was separated from the water by centrifugation for 10 s at 600 xg through a 5 μm filter tube (Bio-Cat). Then, 2 μl glycogen (CALBIOCHEM), 30 μl sodium acetate (Thermo-Fisher Scientific), 2 μl 0.1× pellet paint and 975 μl prechilled 100% ethanol were added and the mixture was incubated for 20 min at − 80 °C. Then, the library was pelleted by centrifugation for 20 min at 4 °C and 20,000 xg, washed with 70% ethanol and recentrifuged for 5 min. Finally, the ethanol was removed, the pellet dried for five minutes at 37 °C and resuspended in 15 μl ultrapure water.

Illumina deep sequencing as well as quantification of small RNA content were performed at a genomics core facility: Center of Excellence for Fluorescent Bioanalytics (KFB, University of Regensburg, Germany). For deep sequencing, all libraries were quantified using the KAPA SYBR FAST ABI Prism Library Quantification Kit (Kapa Biosystems, Woburn, MA, USA). Equimolar amounts of each library were used for cluster generation on the cBot with the TruSeq SR Cluster Kit v3 (Illumina, San Diego, CA, USA). The sequencing run was performed on a HiSeq 1000 instrument (Illumina, San Diego, CA, USA) using the indexed, 50 cycles single read (SR) protocol and the TruSeq SBS v3 Kit (Illumina, San Diego, CA, USA). Image analysis and base calling resulted in .bcl files which were then converted into .fastq files by the CASAVA1.8.2 software.

Analysis of poly-A RNA, RNA enriched with the coding exome oligos and small RNA data was performed using the Genomatix software (Genomatix, Munich, Germany). For poly-A RNA and RNA enriched with the coding exome oligos, the .fastq files were mapped to the human genome assembly GRCh38 (annotation based on ElDorado 6–2015) using the Genomatix Mining Station Mapper v3.7.6.3 allowing one mismatch. We sequenced all poly-A RNA libraries until at least 15 * 10⁶ and all exome capture libraries until at least 10 × 10⁶ unique hits per sample could be mapped. All unique hits were further processed using the Genomatix Genome Analyser v3.51106 which was used to create count tables and RPKM expression values for all samples. Reads were counted locus-based, i.e. for unions of exons of genes. All further analyses based on these count tables were performed with the free software R v3.1.1, Bioconductor v3.0 [20] and the package DESeq2 v1.6.3 [36]. Gene set enrichment analysis [53] was performed with GenePattern and the GSEA module v18 [53] with RPKM (reads per kilobase of exon model per million mapped reads) values and gene sets c2.Biocarta as well as c2.KEGG using standard settings. For small RNA analysis, we mapped .fastq files against a small RNA library based on GRCh38 (annotation based on ElDorado 6–2015) allowing one mismatch. Afterwards, count tables were created for each small RNA type (piRNA, mature miRNA, miRNA hairpin structures) with the small RNA workflow available in the Genomatix Genome Analyser. We sequenced all samples until at least 1 × 10⁶ counts for mature miRNAs were reached when reads were counted, with a mean of unique hits for mature miRNAs of 3,037,413 ± 2088,481and for piRNA 9,754,231 ± 4,096,709. All further single locus based analyses were performed as described for poly-A RNA. For further analysis of piRNAs, we matched the RNAdb2.0 identifiers [45], on which Genomatix is based, to the pIRbase [66], which provides information on the genomic localization, sequence as well as the elements from which the respective piRNAs are derived. Parts of the poly-A RNA-Seq dataset were generated in a collaborative project and three poly-A RNA-Seq runs from neurons are also contained/analysed in [59].

For RRBS, .fastq files were trimmed with TRIM Galore v0.4.2 in the rrbs mode and mapped with BISMARK v0.16.3 [28] to the human genome GRCh38.84 as obtained from www.​ensembl.​org. Then, the methylation values were extracted with the BISMARK methylation extractor again removing 2 bp at the five prime end of the reads to remove the methylation bias in untrimmed reads stemming from the end repair procedure. We sequenced the libraries until at least 1 × 10⁶ alignments with a unique best hit were found by BISMARK. All further analyses were performed based on extracted coverage files with RnBeads v1.2.0 [4].

All normalized NGS data were deposited in GEO (URL: https://​www.​ncbi.​nlm.​nih.​gov/​geo) under the super series accession GSE110720. Coding exome RNA-Seq data is deposited under accession GSE110716, Poly-A RNA-Seq data is deposited under accession GSE110717, RRBS data is deposited under accession GSE110718 and small RNA-Seq data under accession GSE110719.

Due to historical reasons, the mRNA data for quality control analyses of iPSCs were mapped to the human genome GRCh37 without mismatch. We included all samples that we analysed, as long as they met the following criteria: iPSCs had to show a marker expression of the pluripotency markers DNMT3B, SOX2, NANOG, OCT4 and REX1 within the range of a previously analysed high quality ESC cohort and to fall at least in the safety margin of a previously established database driven PluriTest adaption [57]. Furthermore, viral transgene silencing had to be comparable to a published cohort for which NGS raw data was freely available [1]. We analysed viral transgene silencing both by comparison of RPKM values of established markers of viral transgene silencing [27] as well as by direct mapping of the plasmids used for reprogramming counting multiple and unique hits when mapping against the vector sequence (excluding the coding regions of the pluripotency factors) normalized to the number of unique hits of the respective sequencing run when mapped to the genome. The upper-limit cut-offs shown in the supplementary material were calculated as described in [57]. SNPs were called within the Genomatix Genome analyser with a workflow based on samtools, with at least 4 x coverage per SNP and exclusion of indels. We excluded five iPSC lines based on these criteria.

The Institutional Review Board approval (Nr. 4120: Generierung von humanen neuronalen Modellen bei neurodegenerativen Erkrankungen) and informed consent forms are on file at the movement disorder clinic at the Department of Molecular Neurology, Universitätsklinikum Erlangen (Erlangen, Germany). All procedures involving patient samples (tissues or cells) were approved by the local institutional review board (Ethikkommission Regensburg), approval 14–101-0216. The experiments involving embryonic stem cells were approved by the Central Ethics Committee for Stem Cell Research in Germany according to StZG (AZ: 3.04.02/0121). Tissue samples were obtained from the Netherlands brain bank as fresh frozen tissue. iPSCs were generated from skin biopsies of PD- and control-patients by the ForIPS core project as described elsewhere [15].

The presence of Lewy bodies in the substantia nigra and more importantly in the cingulum was verified with stained sections from the Netherlands brain bank (NBB). For those cases were no staining was available, we obtained paraffin sections from the NBB and performed a staining with an antibody directed against aggregated α-synuclein (anti-human α-synuclein 5G4, mouse monoclonal, analytikjena, Jena, Germany). After deparaffinization, antigen retrieval was performed by cooking in citrate buffer for 20 min and DAB staining was performed with the Envision Dual Link System-HRP DAB+ Kit (Agilent, Santa Clara, CA, USA) according to the manufacturer’s instructions. Briefly, the sections were blocked with Dual Endogenous Enzyme Block for 10 min and rinsed with PBS. Then, primary antibody was applied (diluted 1:500 in PBS + 1% BSA) for 40 min and again the slides were washed with PBS. Afterwards, labelled polymer was added for 30 min. After a further washing step, substrate was added for five minutes. Afterwards, the slides were washed in running tap water, counterstained in hemalaun (Dako/Agilent) and again rinsed in running tap water. Finally, the slides were dehydrated in increasing ethanol concentrations and xylol and mounted in entellan mounting medium (Merck).

For analysis of methyl-cytosine content in tissue, we cut frozen tissue sections from the cingulate gyrus material. The DAB staining was performed with the Envision Dual Link System-HRP DAB+ Kit (Agilent, Santa Clara, CA, USA) according to the manufacturer’s instructions with some modifications. After cutting and thawing, the sections were first fixed with 4% PFA for 15 min, which was necessary as we retrieved unfixed material. Then the sections were washed three times for five minutes in TBS. Afterwards, the slides were incubated for 30 min in 2 N HCl for antigen retrieval. The slides were washed twice with PBS and blocking was performed with the Dual Endogenous Enzyme Block reagent for 10 min at RT. After further washing in TBS, anti-methyl-cytosine (Epigentek, mouse monoclonal, clone 33D3) antibody was added in 1:400 dilution and incubated O/N at 4 °C in 1% normal goat serum (PAN Biotech, Aidenbach, Germany) in TBS + 0.3% TritonX. A mouse IgG (Thermo-Fisher) was used as negative control. The next day, the slides were washed three times, covered with labelled polymer and incubated for 30 min at RT. Then, after one washing with TBS, the sections were covered with chromogen for 10 min. Afterwards, the slides were washed in running tap water, counterstained in Mayer’s hemalaun and again rinsed in running tap water. Finally, the slides were dehydrated in increasing ethanol concentrations and xylol (Carl-Roth) and mounted in entellan mounting medium (Merck).

cDNA synthesis was performed from 200 ng total RNA using random hexamer primers (Gene Link, Hawthorne, NY, USA) and the SuperScript™ II Reverse Transcriptase (Life Technologies) according to the manufacturer’s instructions. Real-time RT-PCR was performed with the SensiFAST™ SYBR Hi-Rox Kit (Bioline, London, UK) on the StepOnePlus™ cycler (Life Technologies). Relative expression values were calculated with the ΔΔC_T (analysis of relative gene expression) method [35] using ARF-1 as the reference transcript. Primers used for ARF-1 were 5´-GACCACGATCCTCTACAAGC (forward) and 5´-TCCCACACAGTGAAGCTGATG (reverse), for TH 5´-CCAAGACCAGACGTACCAGT (forward) and 5′- CGTGAGGCATAGCTCCTGAG (reverse). Primers for analysis of SYP, MAP2, TUBB3, EN1, NURR1, KCNJ6 and FOXA2 were described by others [34]. Primers for WNT3 5´-GGAGAGGGACCTGGTCTACTA (forward) and 5′- CTTGTGCCAAAGGAACCCGT (reverse) were and downloaded from the PrimerBank database [63]. Analysis of mtDNA content was performed with the real-time PCR conditions as described above and primers for mtDNA and nuclear DNA as described by others [51], as was analysis of mtDNA deletions [11]. Semiquantitative PCR was performed with the HotStarTaq polymerase (Qiagen) according to the manufacturer’s instructions and the equivalent of 125 ng cDNA per reaction. An annealing temperature of 55 °C and 35 cycles were used for amplification. Primers for semiquantitative PCR were designed with Primer3Plus [61]. Primer sequences for PIWIL2 were 5´-TTGGATTCGAAAATGGCTTC (forward) and 5´-AGCCAGGAAGCGGTTATTTT (reverse), PIWIL4 5´-CAAGGACGTGATGGTTGTTG (forward) and 5´-ACCGACAGTCTTGAGCTGGT (reverse). GAPDH was used as reference with the primers 5´-CCATCTTCCAGGAGCGAGAT (forward) and 5´-ATGATGTTCTGGAGAGCCCC (reverse).

All statistical analyses of NGS data were performed with the software packages described above and standard settings, unless otherwise indicated. A (small-) RNA locus was defined as differentially regulated when it was deregulated at a log2FC of ≥0.6 and an adjusted p-value of < 0.1. For analysis of methylation data a cytosine was accepted as differentially methylated when it passed the threshold Δ-meth ≥0.2 and an adjusted p-value of < 0.1. A gene set was accepted as differentially enriched at a p-value < 0.05 and a false discovery rate (FDR) < 0.25. Otherwise, all statistical tests used are provided in the figure legends and the results part together with the sample size. All bar graphs show the mean + standard deviation (SD) unless otherwise indicated.

We gained skin fibroblasts from six healthy control- and nine sporadic Parkinson’s disease (PD)-patients within the Bavarian Research Network Induced Pluripotent Stem Cells (ForIPS). All samples were derived from inner upper arm skin biopsy derived fibroblasts. Early passage fibroblast samples were reprogrammed to iPSCs using retroviral reprogramming Yamanaka factors. From all except one patient (PD7) iPSC lines were established, and finally a subset of these lines was further differentiated to midbrain neurons (MN) (Additional file 1: Table S1 for details on the patient cohort).

Analysis of the transcriptome showed that fibroblasts, iPSCs and differentiated neuronal cells -as expected- had globally distinct gene expression patterns (Fig. 1a). Our neurons clustered well in a principal component analysis with dopaminergic neurons published by others [34] (Fig. 1b). We then validated that all iPSCs corresponded to high quality pluripotent cells according to our recent PluriTest adaption for NGS [57] (Fig. 1c). This means that all cell preparations mapped at least within a calculated statistical cut-off, with most preparations also meeting the requirements of an empirical cut-off of high quality pluripotent cells (for further information see [57]). We checked for the expression of pluripotency marker genes (in the range of high quality pluripotent cells, Additional file 2: Figure S1), ensured that cells showed viral transgene silencing (examined by direct mapping to the vector sequences used for reprogramming) and excluded that markers of insufficient viral transgene silencing were overexpressed [27] (Fig. 1d and e). In addition, we performed a SNP based paternity testing for every reprogrammed and differentiated line. All cell lines used for further analyses met these stringent quality criteria.

The neuronal cells showed robust induction of the dopaminergic neuron marker TH, and there was no significant difference between the control- and PD-group in the expression of TH, EN1, MAP2, FOXA2, KCNJ6, SYP2, TUBB3 and NURR1 (two-sided Mann-Whitney test, p > 0.05 for all comparisons, n = 8 CTRL and 7 PD derived MN neurons, Additional file 3: Figure S2). In addition the cells were not showing morphological differences or increased cell death without stressor [59] as has been described for genetic models of the disease [42, 53].

A hierarchical clustering based on mRNA data (TOP2000 variable genes, rlog-normalized) clustered the samples by cell type, but no distinct groups were detected for PD- and control-patient derived cells (Fig. 2a). We then went on to examine differential gene expression between control- and PD-derived cell populations within every cell type and adjusted our model for differential expression for the covariate gender and for the iPSCs additionally for passage number. As ageing marks have been reported to be removed in four-factor reprogrammed cells [39], we did not include age as a covariate in the analysis of iPSCs and differentiated neurons, but only for fibroblasts.

It became evident that relevant differential gene expression between the PD- and control-group could only be observed in neuronal cells with 97 deregulated loci, but not in fibroblasts with 3 or iPSCs with 11 deregulated loci (log2FC ≥ 0.6, p-adj. < 0.1, Fig. 2b and Additional file 4: Table S2). Analysing these alterations in neuronal cells on the single gene level, genes that belong to the WNT-pathway were deregulated, i.e. WNT3 was upregulated in PD-patient derived neurons (two-sided Mann-Whitney test, p < 0.05, n = 8 CTRL neurons and 7 PD neurons, Fig. 2c). On the pathway level, the NOS1- and CREB-pathways as well as the pathway regulating PGC1α (among others) were significantly inactivated in PD-patient derived neurons (p < 0.05, FDR < 0.25, n = 8 CTRL neurons and 7 PD neurons, Figs. 2d and e and Additional file 5: Table S3). Both, the PGC1α- and CREB-pathway are well-known and important regulators of neuronal cell survival [23, 67]. As such, these findings provided first implications for the usefulness of neuronal models derived from sporadic PD-patients via the iPSC stage for disease modelling and prompted us to further examine the epigenome of our cell lines across differentiation stages.

Next, we examined the small RNA expression patterns via NGS in all 15 fibroblast cell lines, 24 iPSC lines, two hESC lines and ten lines differentiated to midbrain neurons. When PD-patient derived cells were compared with controls, there were 26 deregulated miRNAs in fibroblasts, 34 in iPSCs and 40 in neurons (p-adj < 0.1, logFC ≥0.6, Additional file 6: Figure S3A and Additional file 7: Table S4). For mature miRNAs, the first important finding was that let-7 family members are upregulated in PD-patient derived neurons (Additional file 6: Figure S3B). The let-7 family has been reported to be deregulated in a C. elegans model of PD [3] and this might be another regenerative mechanism, like WNT-pathway upregulation, in PD-derived neurons.

Even more strikingly, we found a high number of PIWI interacting RNAs (piRNAs) differentially regulated in all cell populations (Fig. 3a and Additional file 8: Table S5). We next checked if genes that control piRNA biogenesis are actually expressed in cultured neurons. Indeed, PIWIL2 and PIWIL4 expression were detectable in cultured midbrain neurons (Fig. 3b) which is similar to results that others have published from tissue in mouse [41] and human [60]. Importantly, when we examined all 15 fibroblast lines and a subset of 13 PD iPSC and 10 control iPSC lines as well as two hESC lines for total small RNA content, there were no significant differences (Kruskall-Wallis test, p > 0.05, Additional file 9: Figure S4A). We also asked if our library preparation included the correct length fraction of piRNAs. With our library preparation protocol, we expect the 24–32 bp piRNAs to run approximately between 150 and 160 bp, slightly higher than 22 bp mature miRNAs. The library sizes of small RNA libraries showed peaks in this size range (Additional file 9: Figure S4B). Of course, it should be noted that other RNA species, e.g. iso-mirs can also be present in these higher size ranges. As contamination of piRNAs with other RNAs, i.e. snoRNA degradation products has been a concern for IP based approaches [49], we checked the overlap between the piRNA sequences we used for piRNA mapping [66] and a snoRNA database [32]. However, as the overlap with snoRNAs was negligible and as the piRNAs identified by us showed the typical 5-prime Uridine bias, we conclude that we identified bona fide piRNA sequences. Nonetheless, as a length of 24–32 bp is another important criterion for canonical piRNAs [68] we checked the length fractions of our sequenced libraries after adapter trimming (data not shown). Due to the fact that there were less reads in the range of 24–32 bp than unique piRNA hits, we conclude that our dataset contains canonical piRNAs and piRNA-like molecules that are abundantly expressed outside of the testes as has been described by others [64]. For simplicity, piRNAs and piRNA-like molecules are referred to as piRNAs in the rest of the manuscript.

Based on the TOP100 significantly deregulated piRNAs, PD- and control midbrain neurons formed separate clusters in a hierarchical clustering analysis (Fig. 3c). However, the fraction of memory piRNAs (i.e. piRNAs that were differentially regulated between PD- and control-patient derived iPSCs/neurons and already found deregulated in fibroblasts) was rather low and always below 10% of all deregulated piRNAs (Fig. 3d). Only two piRNAs, piR-48,442 and piR-43,518, were deregulated between PD- and control-patients across all cell types (Additional file 8: Table S5).

Among the upregulated piRNAs, there was a strong enrichment of cytosine content within the bases two to nine of the piRNAs as compared to the reference of all piRNAs (Fig. 3e).

We then analysed with the annotations provided in the piRBase [66] from which elements the deregulated piRNAs were derived. Interestingly, SINE- and LINE-derived piRNAs were significantly enriched among the downregulated piRNAs as compared to the genome-wide abundance of all piRNAs analysed in our study (two-sided chi-square test, p < 0.0001 and p < 0.01, respectively, Fig. 3f). Despite massive piRNA deregulation, however, we could not find any overlap between known PD-risk loci and predicted piRNA loci in the human genome (data not shown).

We conclude that -dependent on the cell type analysed- the aberrant small RNAome is activated at different differentiation stages but only few deregulated piRNAs are shared between these stages.

As there were striking differences in piRNA expression between PD- and control-patient derived cells, we hypothesized that piRNAs should be altered by neural induction in control cells, too. Indeed, piRNAs underwent dramatic changes after induction of pluripotency and neural differentiation and were even more dynamically regulated as mature miRNAs (Fig. 4a and Additional file 10: Table S6). As many piRNAs showed a low individual abundance, we checked the 20 most abundant and differentially expressed (logFC ≥0.6 and p-adj. < 0.1) and therefore potentially most important piRNAs in the comparison iPSC vs. neuronal cells that constitute on average 19.33 ± 7.49% (mean ± SD) of all piRNA counts across neurons and iPSCs (Fig. 4b). Importantly, among all deregulated piRNAs, again SINE-derived piRNAs were overrepresented in the comparison fibroblasts (n = 6) vs. iPSCs/ESCs (n = 16) and iPSC/ESC vs. neurons (n = 5) (two-sided chi-square test, p < 0.001 and p < 0.0001, respectively, Fig. 4c). As expected, a hierarchical clustering based on the TOP100 differentially expressed piRNAs separated control-patient derived neurons and iPSCs (Fig. 4d).

For the evaluation of small RNA expression pattern in PD tissue, we used material from the cingulate gyrus of eight healthy and eight Parkinson’s disease patients. We verified that the PD-patient derived tissues from the cingulate gyrus were positive for Lewy bodies (Fig. 5a). When analysing differential gene expression, we realized that oligodendrocyte marker genes were overrepresented in control patients, despite of the fact that classical cell type markers (GFAP, OLIG1/2/3, TUBB3 and RBFOX3) remained unchanged. We therefore included neuronal content (calculated as described in material and methods) as a covariate. A decreased transcriptional contribution of the oligodendrocyte lineage in Parkinson’s patients had been reported previously [22]. Importantly, the mean RPKM of the marker gene sets reported by others in mouse [6] was very specific for human cells and neuronal and glial markers were highly correlated (Additional file 11: Figure S5).

With correction for cell type abundance, there were many deregulated small RNAs at logFC ≥0.6, p-adj. p < 0.1 between PD- and control-patients (Fig. 5b and Additional file 12: Table S7). In our cohort, four miRNAs overlapped between tissue and neurons when control- and PD-patient derived cells were compared (Fig. 5c). As we found many deregulated piRNAs in neurons, we tested if there is expression of PIWIL2 and PIWIL4. Indeed, expression of both genes was detectable in all tissue samples (Fig. 5d).

In addition, there was abundant deregulation of piRNAs, effectively separating PD and control cases based on the TOP100 differentially expressed piRNAs (sorted by adjusted p-value, Fig. 6a). Although -as expected- differentially expressed piRNAs were largely different between cultured cells and tissues, there was an overlap of 70 piRNAs that deserve future evaluation as diagnostic marker (Fig. 6b). We again observed an overrepresentation of cytosines among the upregulated piRNAs in the second to ninth base (Fig. 6c). Importantly, both SINE- and LINE-derived piRNAs were enriched among the downregulated piRNAs in tissue, as was described for neurons above, although this reached significance only for LINE-derived elements (two-sided chi-square test, p < 0.05, Fig. 6d).

Of note, enrichment of cytosines within the first 10 bp and overrepresentation of SINE- and LINE-derived elements among the downregulated piRNAs was not consistently retained in fibroblasts and iPSCs. Therefore, aberrant piRNA expression is unmasked in differentiated neurons, but there is no significant epigenetic memory present within the small RNA fraction.

We then examined differential methylation between control and PD cell lines. We successfully sequenced more than 400,000 CpGs at a coverage ≥5× measured in every sample. We analysed 15 fibroblast lines (n = 6 in the CTRL group and n = 9 in the PD group), 28 pluripotent stem cell preparations (n = 16 in the CTRL group and n = 12 in the PD group), ten midbrain dopaminergic neurons (n = 5 in the CTRL group and n = 5 in the PD group). Sequences were highly enriched in CpG islands, CpG island shores and functional elements, as expected. There were differences in global methylation patterns between cell types, but not between PD- and control-patient derived cells (Additional file 13: Figure S6 A-C). Even after restricting the analysis to autosomes and stringent removal of low variability regions there were only very few differentially methylated CpGs (cut-off Δmeth ≥0.2, p-adj. < 0.1, data not shown) in fibroblasts (36 CpGs), iPSCs (6 CpGs) or neurons (45 CpGs) and no difference in the mean methylation pattern of any well-covered (at least 5 CpGs per promoter with 5× coverage) gene promoter of known monogenic PD genes (Additional file 13: Figure S6D).

Thus, global methylation patterns that had been found to be reduced in late stage disease [33], were unaltered in vitro as judged by RRBS. This finding was confirmed in the tissues where no differences in methyl-cytosine staining could be observed between PD- and control-patients (Additional file 14: Figure S7). In comparison with a previous study [12] that reported global methylation differences between PD- and control-patient derived neurons we used another protocol for iPSC derivation as well as neuronal differentiation. We therefore conclude that in our experimental setup there is no methylation-based epigenetic memory or any disease-specific alteration in the CpG context in sporadic PD-patient derived cells.

Finally, we analysed mtDNA methylation patterns on the basis of our RRBS data as well as mtDNA mass and mtDNA deletions by real-time PCR. There were no significant differences between PD- and control-patient derived cells but again only between cell types themselves (Kruskall-Wallis test followed by Dunn’s multiple comparison test, Additional file 15: Figure S8).