Protracted Fate Acquisition and Epigenetic De-Aging During Induced Neural Stem Cell Conversion of Human Blood Cells
Transcription factor-based direct conversion of somatic cells represents an interesting avenue for generating induced neural stem cells (iNSCs) from peripheral blood without transit through a pluripotent stage. While this paradigm has been shown to be associated with epigenetic de-aging, the dynamics of this process have remained unclear. Here, we used overexpression of the two reprogramming factors SOX2 and cMYC to generate iNSCs from erythroid progenitors of donors ranging from neonatal to 101 years of age. Using an epigenetic aging clock algorithm, we corroborated our previous finding that iNSCs generated from aged donors show pronounced epigenetic de-aging, preserving around 13% and 5% of the original donor age at low and high passages, respectively. Studying the dynamics of epigenetic de-aging during iNSC conversion across time, we found that this process is largely protracted, continuing for several weeks and even under proliferation-inhibiting conditions. Transcriptomic differences between young and old donor-derived iNSCs dissipate with extended time in conversion, too. Concordant with this observation, established iNSC lines lack age-associated cellular hallmarks, similar to induced pluripotent stem cells and their derivatives. Interestingly, time course analysis of DNA methylation and RNA sequencing data revealed that acquisition of a bona fide NSC signature extends greatly beyond the time point when proliferative PAX6-positive iNSCs emerge. The unexpected slow dynamics of these processes make iNSC conversion an attractive model for dissecting the mechanisms underlying somatic transdifferentiation and epigenetic de-aging.
1 Introduction
The rise of cell programming, encompassing all techniques that aim at manipulating cellular identity such as induced pluripotent stem cell (iPSC) reprogramming as well as direct cell fate conversion, has largely facilitated the derivation of human neural cells for applications such as disease modeling and regenerative cell replacement. Significant remodeling events of transcriptomic and epigenetic signatures play a pivotal role in the process of cell programming. Yet, the dynamics and underlying mechanisms of these processes are only partially understood. During iPSC reprogramming, genome-wide remodeling is induced, first leading to the silencing of the cell of origin's transcriptional program, followed by the activation of transient enhancers and eventually upregulation of pluripotency-associated genes such as Sall4, Nanog, and Dppa2 that initiate, mature and/or stabilize the acquired pluripotent state (Buganim et al. 2013; Deng et al. 2021). Direct conversion into post-mitotic induced neurons (iN), which was first successfully achieved by the overexpression of the transcription factor cocktail Ascl1, Brn2 plus Myt1l in mouse fibroblasts (Vierbuchen et al. 2010), offers a more direct route for generating neural cells without a pluripotent intermediate.
Interestingly, iPSC reprogramming and direct conversion into iNs have vastly diverging effects on diverse aging signatures: Biological age, if assessed using DNA methylation (DNAm)-based algorithms (Horvath 2013; Horvath et al. 2018), has been shown to be reverted to an embryonic stage upon iPSC reprogramming (Horvath 2013; Lo Sardo et al. 2017; Weidner et al. 2014), while iNs retain the DNAm age of their source cells (Huh et al. 2016). Notably, DNAm age can only be considered a proxy for biological age, since it is not a direct measurement of age-related cell (dys)functions. However, differential preservation of age-associated cellular traits between iPSCs and iNs is also reflected by the length of telomeres, mitochondrial health, nuclear organization and the extent of transcriptional senescence signatures (Kim et al. 2018; Lapasset et al. 2011; Mertens et al. 2015; Miller et al. 2013; Prigione et al. 2011; Tang et al. 2017; Yang et al. 2015).
In 2011, Kim et al. introduced the concept of directly converting somatic cells into still proliferating induced neural stem cells (iNSCs; Kim et al. 2011). While originally overexpressing the full cocktail of transcription factors used for iPSC reprogramming, namely Oct4, Sox2, Klf4, and c-Myc, in mouse fibroblasts, several groups have meanwhile published protocols using reduced and/or altered transcription factor cocktails in human cells (reviewed in Flitsch and Brüstle 2019). In a previous study (Sheng et al. 2018), we have shown that easily accessible human PBMC-derived erythroid progenitor cells (EPCs) can be converted into clonally growing iNSCs by Sendai virus (SeV)-mediated overexpression of the two transcription factors SOX2 and cMYC only. iNSCs exhibit tripotent differentiation capacity, with neuronal differentiation being amenable to regional patterning, and can be employed for disease modeling and neurotransplantation. Indeed, iNSC-derived neurons were shown to be electrophysiologically functional after maturation in vitro as well as in vivo (Berg et al. 2024; Sheng et al. 2018). While we could further demonstrate that iNSCs are overall similar to iPSC-derived NSCs, not only with regard to transcriptomic, but also epigenetic and cellular age-associated signatures (Sheng et al. 2018), the question of how signatures of cellular identity and age are remodeled in the process of conversion was not yet addressed. Here, we set out to explore the dynamics of cell fate conversion on the one hand and loss of aging signatures on the other. To this end, we generated and characterized isogenic pairs of iNSCs and iPSC-derived NSCs from blood of newborns to high-age donors, representing the entire human life span. We then dissected the blood-to-iNSC conversion process by performing DNAm and RNA sequencing analyses across the time course of transdifferentiation. To our surprise, our findings demonstrate that both processes, epigenetic de-aging—which we define as a reduction of epigenetic age according to the results of the epigenetic clock algorithm published in Horvath et al. (2018)—and cell fate consolidation, are largely protracted, with epigenetic de-aging even continuing under conditions that inhibit proliferation.
2 Methods
2.1 Cell Culture Techniques
2.1.1 EPC Induction
Peripheral blood mononuclear cells (PBMCs) were isolated from fresh blood draws using vacutainer CPT tubes (BD Bioscience, San Jose, USA), which were centrifuged for 30 min at 1800 rcf. Isolated mononuclear cells were washed twice with 1× DPBS (Thermo Fisher Scientific, Waltham, USA) before 3–5 × 106 PBMCs were either frozen in 90% FCS (Thermo Fisher Scientific) plus 10% DMSO (Sigma-Aldrich, St. Louis, USA) or directly resuspended in EPC medium consisting of StemSpan SFEM (Stemcell Technologies, Vancouver, Canada) supplemented with 1% CD Lipid (Thermo Fisher Scientific), 2 U/mL recombinant human (rh) erythropoietin (Bio-Techne, Minneapolis, USA), 100 ng/mL rhSCF (Bio-Techne), 40 ng/mL rhIGF1 (Bio-Techne), 1 μM dexamethasone (Sigma-Aldrich) and 10 ng/mL rhIL3 (Bio-Techne) for subsequent EPC induction, which was performed according to a published protocol with minor modifications (van den Akker et al. 2010). In short, cells were cultured in uncoated 6-well cell culture plates at 6% CO2 and 37°C. From Days 1 to 6 of EPC genesis, 1 mL EPC medium was added freshly per day and well. At Day 7, EPCs were enriched by applying a Percoll (Sigma-Aldrich) density gradient with a density of 1.075 ρ. After centrifugation, the interphase ring containing EPCs was collected. Cells were washed twice before EPCs were resuspended in EPC medium lacking rhIL3 and plated in uncoated 6-well cell culture plates at a density of 1–1.5 × 106 cells/6-well for further maturation. Enriched EPCs were expanded until Day 9 of EPC genesis, performing medium addition on Day 8. On Day 9, EPC-containing cell suspensions were collected and 1–2 × 106 cells were frozen in 90% EPC medium lacking rhIL3 but supplemented with 10% DMSO.
2.1.2 iPSC Reprogramming and Differentiation Into NSCs
EPCs were thawed in DMEM/F12 (Thermo Fisher Scientific). After 1-h incubation at 6% CO2 and 37°C, 3 × 105 EPCs were aliquoted in 15 mL-tubes and centrifuged for 5 min at 1500 rpm. After removing the supernatants, cell pellets were resuspended in 250 μL EPC medium without rhIL3 containing the CytoTune 2.0 (Thermo Fisher Scientific) SeVs—that is, a polycistronic vector for KLF4-OCT3/4-SOX2 overexpression, as well as single vectors encoding for cMYC and KLF4—diluted to a multiplicity of infection (MOI) of 5:5:1. Solutions containing cells and viruses were then transferred to uncoated 4-wells and incubated for approximately 24 h at 6% CO2 and 37°C. The next day, infected cells were collected and centrifuged for 5 min at 1500 rpm. The virus-containing supernatants were discarded, cell pellets resuspended in 500 μL EPC medium without rhIL3 and plated in uncoated 4-wells. On Day 3 of iPSC reprogramming, cells were collected and centrifuged 5 min at 1300 rpm. Supernatants were discarded and cells were replated in EPC medium without rhIL3 on 10 cm cell culture-dishes coated with Matrigel (Corning Inc., Corning, USA; 1:60 dilution in DMEM/F12). Medium was gradually changed to StemMACS iPS-Brew XF (incl. supplement; Miltenyi Biotec, Bergisch Gladbach, Germany) from Days 5 to 11 of iPSC reprogramming by performing partial medium changes with increasing concentrations of StemMACS iPS-Brew XF. Around Day 20 of iPSC reprogramming, up to 24 colonies were manually picked under microscopic control and plated in StemMACS iPS-Brew XF supplemented with 10 μM Rock-inhibitor Y-27632 (Cell Guidance Systems, Cambridge, UK) onto Matrigel-coated 24-well plates (1:60 dilution in DMEM/F12). In order to eliminate SeV transgenes, this procedure was repeated every week until five single colony picking cycles were completed. After the fifth picking cycle, cells were allowed to grow confluent. Once confluency was reached, cells were dissociated by 3-min-long incubation with 0.5 mM EDTA (Carl Roth, Karlsruhe, Germany). After removing the EDTA, cells were detached by pipetting and replated in cell culture formats appropriate for the number of cells detached. Transgene-free iPSC lines were routinely expanded in StemMACS iPS-Brew XF on vitronectin-coated (Thermo Fisher Scientific; 1:100–1:200 dilution in 1× DPBS) cell culture formats before being frozen in 90% StemMACS iPS-Brew XF plus 10% DMSO for cryo-preservation.
In order to generate a stable NSC population from iPSCs, a protocol adapted after Reinhardt et al. (2013) was applied. First, iPSCs were singularized by 5–15 min-long incubation with StemPro Accutase (Thermo Fisher Scientific) at 37°C. Cell suspensions were diluted in DMEM/F12 and centrifuged for 5 min at 1000 rpm. Afterwards, 2 × 106 iPSCs were resuspended in 1xN2B27 (1:1 DMEM/F12: Neurobasal with 0.5× N2 supplement, 0.5× B27 supplement without vitamin A, 2 mM L-glutamine, 0.0025% BSA and 1× Pen/Strep; all materials from Thermo Fisher Scientific) containing 3 μM CHIR99021 (Miltenyi Biotec), 0.5 μM purmorphamine (Miltenyi Biotec), 10 μM SB431542 (Axon Biotech, Hengersberg, Germany) and 1 μM dorsomorphine (Tocris Bioscience, Bristol, UK) and supplemented with 10 μM Y-27632, and transferred to AggreWell 800 microwell culture plates (Stemcell Technologies), which were prepared according to the manufacturer's instructions. AggreWell 800 plates were centrifuged for 3 min at 300 rpm. The next day, a half medium change with 1xN2B27 containing CHIR99021, purmorphamine, SB431542 and dorsomorphine was performed. Approximately 48 h after aggregation of iPSCs to embryoid bodies, these were carefully dislodged from the AggreWells and transferred to uncoated 6 cm Petri-dishes. One day later, medium was changed to 1xN2B27 medium supplemented with 3 μM CHIR99021, 0.5 μM purmorphamine and 64 μg/mL LAAP (Sigma-Aldrich; subsequently called 1xN2B27 + CPL). After another medium change on Day 4 of NSC generation, embryoid bodies were triturated in small pieces on Day 5 using a 1 mL-pipette. Triturated embryoid bodies from one AggreWell were distributed to a full Matrigel-coated (1:60 dilution in DMEM/F12) 6-well cell culture plate and seeded in 1xN2B27 + Pen/Strep + CPL supplemented with 10 μM Y-27632. From this day onwards, NSCs were routinely cultivated on Matrigel-coated cell culture formats (1:60 dilution in DMEM/F12) in medium without Pen/Strep. Medium was changed every 1 to 2 days, and cells were passaged when reaching confluency using Accutase. Splitting ratios typically ranged from 1:4–1:12. Up until passage 5, 10 μM Y-27632 was added for replating.
2.1.3 iNSC Conversion and Propagation
EPCs were converted into iNSCs according to the procedure published in Sheng et al. (2018) with minor modifications. On Day 0 of iNSC conversion, 1.5 × 105 EPCs were thawed as described and spin-infected in EPC medium without rhIL3 containing SeV-SOX2 and SeV-cMYC (each at MOI 5, both single vector-SeVs included in the CytoTune 1.0 kit (Thermo Fisher Scientific, now purchased from ID Pharma, Tokyo, Japan)) for 30 min at 1500 rcf and 32°C. Cell pellets were resuspended in their same virus-containing supernatant after centrifugation and plated into uncoated 4-wells. Cells were kept in a humidified incubator for approximately 24 h at 37°C and 21% O2. The next day, suspension cells were collected, pelleted by 5-min-long centrifugation at 1500 rpm, resuspended in 75% iNSC conversion medium, consisting of 2× N2B27 (1:1 Advanced DMEM/F12 (Thermo Fisher Scientific): Neurobasal with 1× N2 supplement, 1× B27 supplement without vitamin A, 2 mM L-glutamine and 0.0025% BSA) supplemented with 3 μM CHIR99021, 1 μM purmorphamine, 0.5 μM A83-01 (Tocris Bioscience), 10 ng/mL human LIF (Novoprotein, Fremont, USA), 64 μg/mL LAAP (Sigma-Aldrich), 1 μg/mL laminin (Sigma-Aldrich) and 5 μM tranylcypromine (Enzo Biochem Inc., Farmingdale, USA), plus 25% EPC medium without rhIL3 and plated onto Matrigel-coated cell culture formats (1:60 dilution in DMEM/F12) at a density of about 1.3 × 104 cells/cm2. From this day onwards, converting cells were cultured at 37°C and 5% O2. On Day 3 of conversion, medium consisting of 75% iNSC conversion medium plus 25% EPC medium without rhIL3 was added. From Day 5 onwards, emerging adherent iNSCs were cultured in 100% iNSC conversion medium, performing full medium changes every other day. Starting from Day 10 to Day 11, iNSCs were cultivated in iNSC expansion medium (2× N2B27 supplemented with 3 μM CHIR99021, 0.5 μM purmorphamine, 0.5 μM A83-01, 10 ng/mL human LIF and 64 μg/mL LAAP). Colonies consisting of neuroepithelial-like shaped cells were picked during the first 3 weeks of iNSC conversion under microscopic control as described. In order to generate polyclonal cell lines, remaining colonies were dissociated on Day 21 of conversion with Accutase as described, pooled and plated in Matrigel-coated cell culture formats (1:60 dilution in DMEM/F12) at a density of approximately 1 × 105 cells/cm2. Medium of iNSC lines was changed every 1 to 2 days. Lines were passaged at least once a week, ideally when cells had reached a confluency > 100%, using Accutase. Typical splitting ratios ranged from 1:3–1:9. From Day 35 of conversion onwards, iNSCs were cultured under normoxic conditions at 37°C and 21% O2. SeV removal was achieved by prolonged cultivation (typically between 4 and 7 weeks) of established iNSC lines at 39°C. During this period, SeV-mediated transgene expression was monitored weekly by RT-PCR. Once expression of both SeVs was below the detection level of the RT-PCR, cells were transferred back to 37°C. One week thereafter, another RT-PCR was performed in order to confirm the persistent absence of SeVs.
For manipulating proliferation speed during EPC-to-iNSC conversion, cells were treated with either 2% (v/v) glycerol (Sigma-Aldrich) or 2 mM thymidine (Sigma-Aldrich) starting from Day 14 of conversion. During this time, iNSCs were replated once a week and cell numbers were estimated at each replating step using a Neubauer counting chamber (Paul Marienfeld, Lauda-Königshofen, Germany) in order to calculate population doublings. Glycerol was added to the medium with every medium change but was omitted on days when splitting was performed. Thymidine treatment always started the day after splitting and was discontinued after 48 and 24 h for Weeks 2–5 and 6–7 of conversion, respectively.
2.1.4 iNSC Differentiation by NGN2 Overexpression
A transcription factor-mediated forward programming approach was implemented for rapid differentiation of iNSCs into post-mitotic neurons early in the conversion process. To this end, at Day 7 of conversion, iNSCs were infected with two lentiviruses (1:50–1:80 dilution), encoding for a reverse tetracycline-controlled transactivator (rtTA) and a tetracycline-responsive promoter element (TRE)-regulated NGN2 expression cassette. Afterwards, iNSCs were routinely cultured up until Day 21 of conversion, when iNSCs were replated at a density of 8.3 × 104 cells/cm2. The day after, forward programming was induced by changing medium to NGMC containing 10 ng/mL BDNF, 10 ng/mL GDNF, 64 μg/mL LAAP, 5 μM DAPT and 1 μg/mL DOX. Differentiation of iNSCs proceeded under normoxic conditions at 37°C and 21% O2. From Day 5 to Day 7 of differentiation, 5 μM Cytarabin/Ara-C (Merck Millipore) was added to the differentiation medium in order to kill mitotically active cells. Medium was changed three times a week during the 20-day-long differentiation period.
2.2 Molecular Biology Techniques
2.2.1 DNAm Analysis
DNA extraction was performed using Qiagen's (Hilden, Germany) Blood & Tissue kit according to the manufacturer's instructions. The concentration and purity of extracted DNA were determined using a Nanodrop 2000c. For DNAm profiling, DNA samples were diluted to a concentration of 55 ng/μL with AE buffer (Qiagen) and provided to the Next Generation Sequencing Core Facility of the University of Bonn. DNA samples were then bisulfite-converted before being profiled on Infinium MethylationEPIC 850 k chips (Illumina). Raw data were processed using the minfi package (Aryee et al. 2014) in the R environment (R Core Team 2019). Afterwards, retrieved beta values were used to calculate the samples' epigenetic age (also called DNAm age) using a predictor algorithm based on (Horvath 2013 and Horvath et al. 2018). In addition, DNAm data were assessed for differentially methylated probes and regions (DMPs and DMRs, respectively) in R using ChAMP (Morris et al. 2014; Tian et al. 2017). ChAMP's gene set enrichment analysis was followed up with rrvgo (Sayols 2020), and Venn diagrams were plotted using an online tool provided by the Ghent University ( http://bioinformatics.psb.ugent.be/webtools/Venn/). In addition to retrieving gene ontology (GO) terms within ChAMP, significant most variable positions (MVPs) identified by this package were mapped to genes using the missMethyl package (Phipson et al. 2016), prior to pathway analysis using Reactome PA (Yu and He 2016) in R.
2.2.2 RNA Sequencing Analysis
RNA extraction was performed with Qiagen's RNeasy kit according to the manufacturer's instructions. The concentration and purity of RNA samples were analyzed with a Nanodrop 2000c. RNA samples were stored at −80°C. For bulk RNA sequencing, RNAs were diluted to a concentration of around 50 ng/μL with RNA-free H2O (Qiagen) and provided to the Next Generation Sequencing Core Facility of the University of Bonn, where samples were subjected to QuantSeq 3′mRNA-Seq library preparation (Lexogen GmbH, Vienna, Austria) before being sequenced on a HiSeq 2500 V4 (Illumina) in high output mode.
Raw data were processed by the Bioinformatics Core Facility of the University of Bonn. In short, after trimming of the Illumina Universal Adapter with cutadapt (Martin 2011), reads have been aligned to the human genome (GRCh38) with STAR (Dobin et al. 2013). featureCounts (Liao et al. 2014) was used to assign reads to genes as defined by Ensembl. A read is counted, first, if it is uniquely mapped, second, if it matches the strand of the gene, and third, if it overlaps with only one gene, that is, the read can be nonambiguously assigned to a single gene. Only genes with at least 100 read counts across samples and at least 2 samples each with a minimum count of 40 were considered in the downstream statistical analysis which was performed in the R environment with the Bioconductor package limma (Huber et al. 2015; Ritchie et al. 2015). As samples from the same donor were measured repeatedly over time and cellular state, donor was considered as a random effect in the statistical model and passed to the function duplicateCorrelation as blocking factor. In addition, sequencing batch and sex were modeled as fixed effects to correct for unwanted variation. Statistical contrasts were calculated for the main fixed effect defined by combining age class, cell type and time in a single variable. The Benjamini-Hochberg method was used to calculate p-values adjusted to multiple testing (false discovery rate, FDR) for each contrast. Data visualization, such as volcano plots and heatmaps were generated using R-packages ggplot2 (Wickham 2016) and ComplexHeatmap (Gu et al. 2016), respectively. Pathway enrichment analysis (Fisher test) for differently expressed genes (FDR < 0.05) was performed using the Bioconductor package clusterProfiler (Wu et al. 2021). Only pathways with a set size of at least 10 or at most 500 genes were considered. For some analyses, enriched GO terms were depicted as treemap plots with rrvgo as described.
2.3 Immunocytochemistry and Fluorescence Imaging
For staining cells within the first 2 weeks of conversion, both adherent cells and cells in suspension were collected. To this end, the supernatant was harvested and adherent cells were dissociated with Accutase as described. The adherent and nonadherent fractions were pooled and subsequently spun down on slides at 500 g for 7 min using appropriate spinning cuvettes. Afterwards, primary mouse monoclonal IgG antibody to CD71 (1:100; BD Pharmingen, Franklin Lakes, USA) in HBSS was incubated for 30 min at room temperature. Antibody solution was washed off using HBSS before cells were fixed with 4% PFA (Thermo Fisher Scientific) for 5 min at room temperature. Following another washing step with 1× DPBS containing 1% BSA, slides were incubated with 1× DPBS containing 1% BSA and 0.5% Triton-X100 (Sigma-Aldrich) for 30 min at room temperature. Afterwards, cells were stained with rabbit polyclonal IgG antibody to DACH1 (1:100; Proteintech, Planegg-Martinsried, Germany) in 1× DPBS containing 1% BSA and 0.3% Triton-X100 for 2 h at room temperature. After washing in HBSS, secondary antibodies goat anti-mouse IgG (H + L) Alexa Fluor 555 and goat anti-rabbit IgG (H + L) Alexa Fluor 488 (both from Thermo Fisher Scientific) in 1× DPBS containing 1% BSA and 0.3% Triton-X100 were incubated for 30 min at room temperature. Cells were washed again in 1× DPBS containing 1% BSA before nuclei were stained with 2 μg/mL 4′,6-Diamidino-2-phenyl-indol-dihydrochloride (DAPI; Sigma-Aldrich) for 5 min at room temperature. Finally, slides were washed in 1× DPBS and mounted with cover slips using 133.33 mg/mL Mowiol-488 plus 25 mg/mL DABCO (both from Carl Roth).
All adherently growing cell preparations were fixed with 4% PFA for 10 min at room temperature before being washed with 1× DPBS once. Cells were blocked with 10% FCS in 1× DPBS for 2 h at room temperature. Blocking solutions were supplemented with 0.5% Triton-X100 for cell permeabilization. After blocking, cells were incubated overnight at 4°C with primary antibodies diluted in 1× DPBS containing 5% FCS + 0.3% Triton-X100: chicken polyclonal IgY to MAP2 (1:1000; Bio-Techne), mouse monoclonal IgG to γH2AX (1:250; Merck Millipore), mouse monoclonal IgG to LMNA/C (1:200; Abcam, Cambridge, UK), mouse monoclonal IgM to O4 (1:500; Bio-Techne), mouse monoclonal IgG to S100β (1:1000; Sigma-Aldrich), mouse monoclonal IgG to SOX2 (1:100; Bio-Techne), mouse monoclonal IgG to TUBB3 (1:1000; BioLegend, San Diego, USA), rabbit polyclonal to GFAP (1:1000; Merck Millipore), rabbit polyclonal IgG to LAP2a (1:250; Abcam), rabbit polyclonal IgG to NES (1:200; Bio-Techne) and rabbit monoclonal IgG to DCX (1:200; Abcam). The next day, cells were washed three times with 1× DPBS + 0.3% Triton-X100 before secondary antibodies, diluted in 1× DPBS containing + 0.3% Triton-X100, were incubated for 2 h at room temperature: goat anti-chicken IgY (H + L) Alexa Fluor 647 conjugate, goat anti-mouse IgG (H + L) Alexa Fluor 488 as well as 555 conjugates, goat anti-mouse IgM (H + L) Alexa Fluor 488 conjugate and goat anti-rabbit IgG (H + L) Alexa Fluor 488 as well as 555 conjugates (all from Thermo Fisher Scientific). After three additional washing steps with 1× DPBS + 0.3% Triton-X100, 2 μg/mL DAPI was incubated for 5 min at room temperature to counterstain nuclei. Cells were finally washed once with 1× DPBS before mounting with a medium containing 133.33 mg/mL Mowiol-488 and 25 mg/mL DABCO. Mounted specimens were dried at least overnight at 4°C before imaging.
Image acquisition of immunofluorescent stainings was either performed manually using Zeiss' (Oberkochen, Germany) AxioImager or automatedly utilizing the IN Cell Analyzer 2200 (GE Healthcare Bio-Sciences Corp, Picataway, USA). For qualitative image presentation, color brightness and contrast were either adjusted right after image acquisition with the software associated with the utilized microscope or retrospectively using ImageJ (National Institutes of Health, Bethesda, USA; version 1.52a Java 1.8.0_181, 64-bit). Quantification of immunofluorescent stainings was performed on unmodified raw data using CellProfiler (Kamentsky et al. 2011; Broad Institute, Cambridge, USA; version 2.2.0) pipelines, which automatically and equally applied algorithms for primary object identification and subsequent pixel intensity quantification within identified primary objects to all evaluated pictures according to pre-defined parameters. After image quantification, CellProfiler pipelines were further used to automatically and equally apply background subtraction and color brightness adjustment to all pictures for respective qualitative display.
2.4 Data Processing and Statistical Analyses
Numerical raw data were processed as described in the respective sections with Microsoft Excel (Microsoft, Redmond, USA; Office version 16.16.27). Dot plots were produced in Microsoft Excel, while box plots were created using the web-tool of BoxPlotR ( http://shiny.chemgrid.org/boxplotr/). For box plots, center lines show the medians, box limits indicate the 25th and 75th percentiles as determined by R software, whiskers extend 1.5 times the interquartile range from the 25th and 75th percentiles, outliers are represented by dots and data points are plotted as open circles. If indicated, statistical analyses were conducted in R. First, to determine whether variables met the assumptions of linear models, Kolmogorov–Smirnov's or Shapiro Wilk's and Levene's test or F-test for normal distribution and homogeneity of variance were performed, respectively. If the variables analyzed were normally distributed, Welch two sample t-test and independent one-way ANOVA were conducted in order to compare two and more than two groups, respectively. ANOVA was followed-up by pairwise comparisons using the Tukey-honest significant difference and Games-Howell tests for homogenous and nonhomogeneous variances, respectively. In case analyzed variables were not normally distributed, the nonparametric equivalents, namely the Wilcoxon rank sum and Kruskal Wallis tests, were calculated. The Kruskal Wallis test was followed-up by multiple comparisons using the Wilcoxon rank sum test. Significance level was set at p < 0.05 and adjusted by default when multiple comparisons were performed. Significance levels are indicated in figures as follows: *: p < 0.05; **: p < 0.01, ***: p < 0.001, ****: p < 0.0001. Figures were assembled using Inkscape (Free Software Foundation Inc., Boston, USA; version 0.92.2).
3 Results
3.1 Human iNSCs Can Be Derived From Blood Cells of Donors Aged 0–101 Years
Our previous study had provided first evidence that iNSCs generated from EPCs of 31 to 62-year-old donors exhibit a remarkable reduction of epigenetic age (Sheng et al. 2018). To study the reset of aging signatures in more detail, we first extended the donor cohort to 15 individuals covering a spectrum from 0 to 101 years of age (Table S1). For comparison we generated from the same EPC samples isogenic iPSC-derived NSCs (iPS-NSCs; Reinhardt et al. 2013), which—as iPSC derivative—undergo a complete age reset (Lo Sardo et al. 2017; Sheng et al. 2018). In both protocols, proliferation is sustained by promotion of sonic hedgehog and Wnt signaling, yielding NSCs with highly similar transcriptome and differentiation potential (Figure 1A; Sheng et al. 2018). For iNSC generation, we employed a SeV system to achieve robust overexpression of SOX2 and cMYC. Following SeV transduction, EPCs downregulated the blood cell marker CD71 within a few days, and expression of the NSC marker DACH1 could be detected in converting cells as early as 12 days post infection (Figure 1B). Already at this time point, cells also expressed PAX6 (Figure 1C). We verified that iNSCs and their isogenic iPS-NSC lines homogenously expressed the NSC markers SOX2 and NES even after transgene removal and extended in vitro cultivation (Figure 1D; Figure S1A). Quality control by SNP analysis verified that iNSCs and iPS-NSCs remained genomically intact after cell programming and subsequent expansion for at least 20 passages, corresponding to up to 6 months of continuous in vitro cultivation (Figure S1B). Furthermore, iNSCs and iPS-NSCs were successfully differentiated into TUBB3-positive neurons, S100β- and GFAP-positive astrocytes and O4-positive oligodendrocytes (Figure S1C).
3.2 iNSC Conversion Is Associated With a Protracted Loss of Aging Signatures
We first assessed the epigenetic age of our enlarged cohort including source EPCs, low passage iNSCs and iPS-NSCs using the DNAm age prediction algorithm published by Horvath et al. (2018). As expected, the DNAm age of EPCs highly correlated with the chronological age of the respective donors, demonstrating the accuracy of the age predictor algorithm. Concordant with published data, all iPS-NSC lines exhibited an embryonic-like epigenetic age, irrespective of their original donor age. Interestingly, iNSCs, too, exhibited a substantial degree of epigenetic de-aging across all donor ages, with the DNAm age of iNSCs at low passage being on average 13% of the chronological age of their respective donors (Figure 2A). Intrigued by the reproducibility of this finding using donors of different ages, we next studied how age-associated signatures are affected by blood-to-iNSC transdifferentiation across time. Assessing the DNAm age dynamic in three iNSC lines derived from donors aged 50 to 101 years of age, we found that this epigenetic de-aging represents a largely protracted process continuing for several weeks, and that it takes about 50 days of iNSC conversion until the majority of age-related epigenetic signatures are reset (Figure 2B).
Acknowledging that differences in DNAm age might not inevitably impact biological age on the functional level, we next investigated cellular hallmarks typically associated with age. Specifically, we confirmed the absence of cellular aging hallmarks in established, transgene-free (TF) iNSCs from donors of different ages (0, 87, and 101 years of age) by comparing them to their isogenic iPS-NSCs in different assays. First, we found that the nuclear lamina-associated LMNB and LAP2α, which are both negatively correlated with aging (Scaffidi and Misteli 2006; Shah et al. 2013), are similarly expressed in both cell types and age groups. Expression levels of the age-associated genes RANBP17, LAMA3, and PCDH10 (Mertens et al. 2015), as well as the senescence-related genes CDKN1a and CDKN2a, were similar in young and old donor-derived iNSCs and iPS-NSCs, too (Figure S2A). Although qualitative changes in LMNA/C expression upon aging are much better characterized than absolute differences in expression levels, expression of LMNA/C tended to be higher in iNSCs as compared to isogenic iPS-NSCs and in young as compared to old donor-derived cells—a trend we observed on both RNA and protein level (Figure S2A,B). Second, Western blot analysis of p62 abundance and LC3 conversion revealed no impairments of autophagy in iNSCs (Figure S2C). Finally, DNA damage and mitochondrial ROS production in iNSCs and iPS-NSCs were overall comparable (Figure S2D,E). Altogether, these data indicate that even iNSCs derived from very old donors are epigenetically and cellularly de-aged after direct conversion, highly resembling embryonic stage-like iPS-NSCs.
3.3 Epigenetic De-Aging Upon iNSC Conversion Is Not Dependent on Stem Cell Proliferation
Considering that loss of age-associated epigenetic signatures is observed upon reprogramming of somatic cells into self-renewing iPSCs (Lo Sardo et al. 2017) and direct conversion into proliferating iNSCs but not upon transdifferentiation into post-mitotic neurons (Huh et al. 2016), we hypothesized that epigenetic de-aging might be driven by the persistent mitotic activity of stem cells. We thus performed two sets of experiments manipulating iNSC proliferation using cells derived from three aged donors (81–86 years of age). First, we treated iNSCs with glycerol—which is known to reduce proliferation of various cell types—or thymidine—which is commonly used as a cell cycle-synchronizing agent, since it arrests cells in the G1/S phase of the cell cycle—starting from Day 14 of conversion. Despite their different modes of action, both treatments resulted in significantly reduced accumulated population doublings at Day 49 of conversion (Figure 2C). Yet, neither the relative DNAm age of glycerol- nor thymidine-treated cultures differed significantly from untreated cultures, although there was a slight but statistically significant increase in the relative DNAm age of glycerol-treated cultures as compared to thymidine-treated cultures (Figure 2D).
Second, we infected EPC-derived iNSCs at Day 7 of conversion with lentiviruses encoding a doxycycline-inducible NGN2 expression system. Forced induction of NGN2 has been shown to rapidly induce neuronal differentiation of human NSCs (Ho et al. 2016). After 14 additional days of expansion, a fraction of the infected cells was kept in proliferation, while the rest was subjected to DOX induction under differentiation-promoting conditions. At Day 42 of conversion (i.e., after 20 days of differentiation), the relative DNAm age of NGN2-overexpressing, differentiated cultures did not significantly differ from proliferative iNSCs (Figure 2E), despite the fact that cultures forced to differentiate by NGN2 overexpression indeed mostly consisted of TUBB3-positive neurons and few GFAP-positive astrocytes (Figure 2F). In sum, our data thus suggest that epigenetic de-aging upon iNSC conversion is protracted, and does not depend on sustained cell proliferation.
3.4 Protracted Consolidation of iNSC Identity
We next assessed DNAm changes in the process of iNSC conversion at a larger scale, including data from the cell lines programmed in the course of this project as well as data produced and published in the course of our previous publication (Sheng et al. 2018). In total, we included DNAm data from EPCs of 12 donors as well as from the low passage (LP; P3-P5) iNSC lines established thereof. For a subset of these donors, we additionally included samples collected in the early phase of iNSC conversion, that is, from emerging neuroepithelial-like cultures at Days 14 ( N = 6) and 26 ( N = 4) of conversion, as well as transgene-free iNSCs after extended cultivation (TF; P21; N = 3). After raw data filtering, in total 728,674 probes were compared across these 37 samples. We found that while EPCs and day 14-iNSCs form very distinct clusters in the multidimensional scaling plot (Figure 3A), samples collected at later time points appear quite similar to each other. We then mapped significant most variable positions (MVPs; Benjamini-Hochberg-adjusted p < 0.05) to genes, assessing how many were commonly affected by the analyzed contrasts across time in conversion (Figure 3B). Subsequent Reactome pathway analysis revealed a broad spectrum of pathways, of which most could be assigned to one of five categories: (i) neuronal system, especially neuronal migration, neurite outgrowth and pathfinding; (ii) cell migration, adhesion and division in multiple tissues, including pathways relevant to extracellular matrix (ECM) and mesenchymal-to-epithelial transition (MET); (iii) signal transduction, (iv) angiogenesis and (v) the hematopoietic system (Figure S3). Except for signal transduction and the hematopoietic system, terms of the other three categories were found to be regulated across all stages of iNSC conversion. We then conducted gene set enrichment analysis directly on the identified differentially methylated regions (DMRs; FDR < 0.05). Interestingly, focusing on the 45 gene ontology (GO) terms that were commonly enriched across all stages of iNSC conversion and in vitro cultivation, 14 were associated with nervous system development such as “neurogenesis,” “axon development” and “neuron projection morphogenesis” (Table S2). In this analysis, morphogenesis-, cytoskeleton- and signal transduction-associated terms accounted for the majority of the remaining commonly regulated GO terms as well.
In analogy to the DNAm analyses, we also assessed global transcriptomic changes upon iNSC conversion. Specifically, we included 12 EPC samples as well as 10 Day 14, 14 LP (P4–P6), and seven TF (P18–P21) iNSC samples. For five donors, we additionally included isogenic iPS-NSC samples at LP (P6) as reference. Principal component analysis (PCA) of this RNA sequencing dataset revealed that while all EPC samples formed one distinct cluster, iNSC samples were distributed across two main clusters. One cluster predominantly consisted of Day 14 iNSCs, whereas the other cluster was composed of LP and TF iNSCs, with the latter intermingling with i
