[version 1; peer review: 2 approved]
KSH, IMG and PMM conceived the study. IMG, PMM and MEL conducted ESC culture, produced neurons and extracted the RNA. KSH, IMG and PMM organized the data sets and prepared the manuscript. All authors approved the final manuscript for publication.
Using paired-end RNA sequencing, we have quantified the deep transcriptional changes that occur during differentiation of murine embryonic stem cells into a highly enriched population of glutamatergic cortical neurons. These data provide a detailed and nuanced account of longitudinal changes in the transcriptome during neurogenesis and neuronal maturation, starting from mouse embryonic stem cells and progressing through neuroepithelial stem cell induction, radial glial cell formation, neurogenesis, neuronal maturation and cortical patterning. Understanding the transcriptional mechanisms underlying the differentiation of stem cells into mature, glutamatergic neurons of cortical identity has myriad applications, including the elucidation of mechanisms of cortical patterning; identification of neurogenic processes; modeling of disease states; detailing of the host cell response to neurotoxic stimuli; and determination of potential therapeutic targets. In future work we anticipate correlating changes in longitudinal gene expression to other cell parameters, including neuronal function as well as characterizations of the proteome and metabolome. In this data article, we describe the methods used to produce the data and present the raw sequence read data in FASTQ files, sequencing run statistics and a summary flatfile of raw counts for 22,164 genes across 31 samples, representing 3–5 biological replicates at each timepoint. We propose that this data will be a valuable contribution to diverse research efforts in bioinformatics, stem cell research and developmental neuroscience studies.
Transcriptional profiling by RNA sequencing (RNAseq) enables the sensitive and accurate characterization of the transcriptome
We have described a method to produce large quantities of highly enriched, electrically active glutamatergic neurons (ESNs) from suspension-adapted mouse embryonic stem cells (ESCs)
Although aspects of corticogenesis have been replicated
To evaluate the transcriptional changes associated with differentiation of cortical glutamatergic neurons, we conducted a longitudinal expression profile of the deep transcriptome during neurogenesis from days
In addition to providing the basis for a study to characterize transcriptional processes involved in corticogenesis and neuronal maturation, we can foresee several other applications for this data. First, the identification of genes and isoforms that exhibit differential expression in synchronization with known markers is expected to provide novel insight into molecular mechanisms of neurogenesis and neural patterning. Second, a few algorithms are available for statistical determination of differential gene expression within longitudinal data sets. The depth and quality of these data suggests it may be well-suited for the development and validation of such methods. Third, we envision this dataset facilitating inter-specific comparisons of transcription during neurogenesis and corticogenesis, providing insight into transcriptional mechanisms common to and unique in the evolution of the mammalian cerebral cortex. For these reasons, we are making this data publically available for research efforts in bioinformatics, stem cell research and developmental neuroscience studies.
ESCs were adapted to feeder cell-free, suspension culture and maintained as previously described
ESCs were differentiated into neurons between 5–30 passages after adaptation to suspension culture. A modified 4/4 protocol was used for neuron differentiation
On DIV 0, aggregates were dissociated with TrypLE Express for 5 min at 37°C. Trypsinization was halted with 5 mL of 1% soybean trypsin inhibitor (Life Technologies), the aggregates were gently dissociated by trituration with a 10 mL pipet, and the cell suspension was filtered through a 40 µm cell strainer (Thermo Scientific, Waltham, MA). Cells were pelleted for 5 min at 300 x g, washed in N2 medium (Neurobasal-A medium with 1 x N2 vitamins, 2 mM glutamine and antibiotics [Life Technologies]) and counted manually using a hemocytometer. Neuronal progenitors were plated at 1.5×10 6 cells/cm 2 in poly-D coated dishes. Complete washes with N2 medium were conducted at 4 h and 24 h to remove residual serum and non-adherent cells. At DIV 2, N2 was replaced with B27 medium (Neurobasal-A supplemented with antibiotics, 2 mM glutamine and 1 x B27 vitamins [Life Technologies]). Subsequently, ESNs underwent full medium changes with B27 on DIV 4, 8 and 12. On DIV 8 the media was supplemented with 30 µM 5-fluoro-2’-deoxyuridine and 70 µM Uridine (Sigma-Aldrich) to select against any remaining glia. Following DIV 12, ESNs were left undisturbed until RNA harvest. Cultures remained healthy and viable until at least DIV 28 under these conditions.
Neurons were harvested from 6 cm dishes at DIV -8, -4, 0, 1, 7, 16, 21 and 28 (n=3 to 5) and RNA was isolated by QIAcube (Qiagen, Valencia, CA) using the RNeasy mini kit protocol (Qiagen) and submitted to Expression Analysis, Inc. (Durham, NC) for library preparation and sequencing. Typical recovery was ~3 µg per 6 cm dish, with an RNA integrity number exceeding 8 and a 260:280 of 1.8 – 2.2. PolyA+ RNA was purified from 500 ng of total RNA using polyA selection, chemically fragmented and reverse transcribed using random hexamers. This was followed by second strand synthesis and end repair. Libraries were prepared for paired-end sequencing using the TruSeq™ RNA sample prep kits (Illumina, San Diego, CA) per manufacturer’s instructions, and library size and integrity were determined using the Agilent Bioanalyzer 2100 (Santa Clara, CA). Libraries were bound to flow cell surfaces using the Standard Cluster Generation Kit v5 (Illumina). Flow cells were transferred to the Illumina HiSeq 2000 and run using TruSeq SBS Kits (Illumina). Paired-end sequencing data were generated over 2x50 sequencing cycles. Sequence information and quality scores are available at SRA in FASTQ format, with the forward and reverse reads appended with an "F" or "R" (e.g., DIV7.1F and DIV7.1R). Raw sequence reads were aligned using the University of California, Santa Cruz’s mouse knownGENE track, and transcriptome abundance estimation was performed on completed alignments using RSEM (RNA-Seq by Expectation Maximization)
We thank Angela Adkins, Cindy Kronman and Marian Nelson (USAMRICD, MD) for administrative, logistical and technical assistance. We also thank Expression Analysis, Inc. (Durham, NC) for assistance with RNA sequence generation. The views expressed in this article are those of the authors and do not reflect the official policy of the Department of Army, Department of Defense, or the U.S. Government.
The depth and temporal nature of the dataset presented in this paper will be beneficial to any researcher interested in cortical development in general, and potentially lead to many new insights and avenues to pursue.
A point of note, in my experience differences in passage number of the cells used for differentiation can affect gene expression levels throughout. The authors state “ESCs were differentiated into neurons between 5-30 passages after adaptation to suspension culture.”, I wonder if that is why the DIV21 samples cluster in between the DIV16 and DIV28 when performing a PCA analysis on the transcript read counts (obtained from Data File 2)?
Related question, how raw are the transcript read counts in Data File 2, as I thought raw counts would have to be integers whereas the counts given have decimal points?
Finally, with regard to the previous Ref Report (Ragsdale and Albertin; 12 March 2013), have you considered comparative analysis using the Allen Brain Atlas/ Mouse Brain expression data for the thalamic and cortical areas and see which region your samples resemble most?
Reviewer Expertise:
NA
I confirm that I have read this submission and believe that I have an appropriate level of expertise to confirm that it is of an acceptable scientific standard.
There are a growing number of protocols for differentiating stem cells into particular neural cell types. This paper demonstrates the great potential of RNAseq technologies for assessing the identities of such differentiated cells in culture. The authors’ goal is an
Reviewer Expertise:
NA
We confirm that we have read this submission and believe that we have an appropriate level of expertise to confirm that it is of an acceptable scientific standard.
Despite the presence of a number of transcripts suggestive of telencephalogenesis, the singular derivation of vglut2+/vglut1- neurons has bothered us as well, and was one of the motivating factors underlying this experiment. While vglut2 expression has been reported in the neocortex, transcripts have generally been described to appear postnatally (e.g., see allen brain institute expression data for slc17a6 in the developing mouse brain) and vglut1 expression is much stronger than vglut2 in cortical glutamatergic neurons. We are still wading our way through the differential gene expression data, but preliminary, analysis suggests that the RNAseq data may be more consistent with a thalamic origin than a neocortical origin. Not only is thalamogenesis consistent with a vglut2+/vglut1- phenotype, but it is also less well described than neocortical development. We are currently working toward a clear differential diagnosis based on gene expression (should one exist) in an attempt to more clearly elucidate the structural origin of neurons derived in our model.