RNA sequencing (RNA-Seq) measures the activity of thousands of genes simultaneously, providing a snapshot of the transcriptome. Scientists use this technique to understand biological processes and disease states by quantifying messenger RNA (mRNA) abundance. A significant challenge is the technical variability and bias introduced during sample preparation, including RNA extraction, reverse transcription, and library preparation, which can obscure true biological differences. External RNA Controls Consortium (ERCC) spike-ins are synthetic RNA molecules designed as standardized controls to measure and mathematically correct for this technical variability.
Understanding the ERCC Spike-In Mix
The ERCC spike-ins are a standardized set of synthetic RNA transcripts, not naturally found in the organism being studied. These molecules are introduced into each biological sample at known, precise concentrations to serve as an internal reference standard throughout the sequencing process. The mixture typically consists of 92 distinct transcripts designed to exhibit a wide range of physical properties.
These synthetic RNAs mimic natural eukaryotic mRNA by being polyadenylated, allowing them to be processed alongside the sample’s endogenous transcripts during library construction. The transcripts vary significantly in length (250 to 2,000 nucleotides) and guanine-cytosine (GC) content. This diversity ensures the spike-ins react to potential biases in the sequencing workflow similar to the diverse set of natural mRNAs. The 92 transcripts span a broad dynamic range of concentration, often six orders of magnitude, which helps test the linearity and sensitivity of the assay across all abundance levels.
Integrating Spike-Ins into the Sequencing Workflow
For ERCC spike-ins to function effectively, their timing of introduction is strictly governed. The synthetic mix must be added to the total RNA sample before the first technical steps of RNA processing begin, ideally before or immediately after RNA extraction and purification. This early inclusion ensures the spike-ins are subjected to the maximum number of technical challenges and biases that the endogenous RNA will encounter.
As the sample moves through the workflow, the spike-ins mirror the behavior of the native RNA. They are exposed to variations in extraction efficiency, degradation during handling, and sequence-specific biases during reverse transcription or subsequent PCR amplification. By participating in every technical step, the final sequenced count of each spike-in transcript reflects the cumulative technical efficiency of that sample’s preparation. Adding them later, such as after cDNA synthesis, would fail to capture the full scope of technical variation, limiting their utility for accurate bias correction.
Normalization and Quality Control Metrics
The utility of ERCC data lies in its application for normalization, a computational process that uses known input quantities to mathematically adjust for technical differences between samples. Since the exact starting concentration of each ERCC transcript is known, researchers compare the expected concentration with the number of sequencing reads obtained. This comparison provides a direct measurement of the technical bias and efficiency of each individual RNA-Seq library.
ERCC data is also used for robust quality control (QC). Plotting the observed read counts against the known input amounts for all 92 transcripts reveals the dynamic range, linearity, and limit of detection. A strong, linear correlation confirms that the library preparation and sequencing behaved as expected. Deviations from linearity or low read counts for highly concentrated spike-ins can immediately flag a technical failure, such as a poor reverse transcription step or an issue with sequencing depth.
For normalization, ERCC data allows for scaling read counts to account for differing sequencing depths or inter-sample technical variation. If one sample was sequenced deeper than another, the ERCC counts will be proportionally higher, and a scaling factor can be derived to bring all samples to a comparable technical level. More sophisticated normalization methods, such as regression-based models, use ERCC data to model the relationship between technical efficiency and transcript features like length or GC content. By mapping the technical efficiency for the spike-ins, these models adjust the read counts of the endogenous genes, providing a more accurate estimate of true gene expression levels.
Absolute Quantification
Beyond correcting for technical variation, ERCC spike-ins enable absolute quantification of endogenous transcripts. By establishing a standard curve—a linear relationship between the known input amount of the spike-ins and their observed read counts—researchers can estimate the absolute number of molecules for any endogenous gene. This moves the measurement beyond relative comparisons (e.g., Gene A is expressed twice as much as Gene B) to providing an estimate of the actual concentration of a transcript in the original biological sample. This approach is valued in studies where the precise concentration of transcripts, rather than just relative changes, is of interest.
Limitations and Context in Gene Expression Analysis
While ERCC spike-ins offer advantages for quality control and technical correction, they have specific limitations. The most significant limitation is that these synthetic controls only measure and account for technical variability introduced during sample processing and sequencing. They do not account for biological variability, such as differences in cell size, cell cycle stage, or overall RNA content between samples.
The synthetic nature of the ERCC transcripts means they may not perfectly mimic all aspects of endogenous RNA. They lack the complex secondary structures and associated cellular proteins found on natural mRNAs, which can affect their efficiency during reverse transcription differently than the sample RNA. This imperfect mimicry means the correction derived from the spike-ins is an approximation of the true technical bias.
For many experiments, alternative normalization methods that rely on internal scaling factors, such as Trimmed Mean of M-values (TMM) or Transcripts Per Million (TPM), are often used. These methods assume that most genes are not differentially expressed between samples and scale the data accordingly. However, ERCC spike-ins remain an independent tool, particularly when absolute quantification is desired or when rigorous, external quality control of the entire RNA-Seq workflow is required.