Gene expression profiling of neuronal subtypes could advance our understanding of normal brain development and function, reveal distinctions between healthy and diseased neurons, and aid in the identification of diagnostic and prognostic biomarkers. But creating such profile catalogs has proven challenging since the neuronal processes connecting cells in the adult brain are fragile and easily damaged by enzymatic dissociation and mechanical trituration. While fluorescence activated cell sorting (FACS) has long been considered inapplicable for isolating neuronal populations from adult brain since shearing forces during sorting are likely to damage the fragile cells, recent reports have suggested the feasibility of coupling FACS with neuronal isolation through gradient separation, followed by cell culture with factors needed for regenerating the neuronal processes. Although promising, isolation rates using these approaches have been too low to permit sequence-based expression profiling. The availability of multiple mouse strains with fluorescently labeled neuronal subpopulations, along with the attractive information that could be gained through expression profiling of neuronal sub-populations, prompted Saxena et al. (RIKEN Yokoham Institute, Japan) to explore methods for improved FACS of neurons. The authors reasoned that trehalose, an agent known for its ability to preserve cell viability during heat stress and cryopreservation through chaperonin-like activity, might stabilize neurons during tissue digestion and dissociation. To test their idea, they supplemented all solutions used in a neuron disaggregation procedure with 0.132M trehalose, and found that this increased cell viability from 58% to 81%, when compared to the same protocol without trehalose. FACS gating removed the dead cells, which no longer fluoresced, resulting in a 10-fold overall improvement in cell yield. Most importantly for expression analysis, the RNA extracted from these isolated cells was of sufficient quality and quantity for profiling by cDNA array, nanoCAGE, and RNA seq, retaining RNA transcripts localized to neurites. This novel approach will expand researchers’ capability for transcriptome studies, and should enable single neuron sequencing studies as well as assays of other complex heterogeneous tissues containing fragile cell types.
Large-scale human transcriptome studies using microarrays often require different sample types obtained from numerous patients. Unfortunately, even with carefully implemented standard operating procedures, a significant percentage of mislabeled samples can occur and impede the accurate analysis of gene expression patterns. It is therefore imperative to accurately track patient samples in microarray transcriptomic studies — ideally using some intrinsic feature of the trancriptome — in order to identify any mislabeled samples. In this month's issue, W. Zhao and colleagues at Stanford University (Palo Alto, CA) demonstrate that single nucleotide polymorphisms (SNPs) present in coding sequences, i.e. coding SNPs (cSNPs), can be used for this exact purpose. Individuals can be identified by tens to hundreds of the SNPs in his or her genome, so authors hypothesized that cSNPs transcribed into mRNAs should allow microarray samples from different subjects to be distinguished from one another. In addition, the phenomenon of allele-specific expression, which has been shown to occur throughout the genome and in cell type-specific patterns, could be helpful in using cSNPs. To test cSNPs as intrinsic markers of mRNA samples, the authors incorporated probes for 89,000 cSNPs onto a microarray used in an ongoing transcriptome study, and set up a test experiment consisting of 91 samples taken from five randomly chosen patients who each had been each sampled for three cell types and up to seven time points. After microarray hybridization and analysis, they quantified the allelic imbalance score (AIS) of each cSNP, which indicated the genotype of a homozygous cSNP or the allele-specific imbalance of a heterozygous cSNP and demonstrated that mislabeled samples could be detected as outliers. Using the 500 cSNPs with the most variable AIS, the authors carried out hierarchical clustering and observed that most samples clustered well by subject, with the exception of three samples, suggesting that these were not from the five patients. Reprocessing of the correct three samples showed that they were now correctly clustered with the other samples. Finally, they determined that the number of cSNPs could range from the 50 to 10,000 most variable cSNPs and still allow the same outlier detection as seen using the 500 cSNPs. This approach for sample tracking with microarray experiments is easy to implement without requiring additional information about patient genotypes.