Sample Processing Obscures Cancer-Specific Alterations in Leukemic Transcriptomes

Substantial effort is currently devoted to identifying cancer-associated alterations using genomics technologies. While the genome is inherently stable over short time frames, the transcriptome is dynamic and potentially susceptible to alteration as samples move from the patient to the lab bench. Here, we show that standard blood collection procedures rapidly change the transcriptional and post-transcriptional landscapes of hematopoietic cells, resulting in biased activation of specific biological pathways, up-regulation of pseudogenes, antisense RNAs, and unannotated coding isoforms, and inhibition of RNA surveillance. These artifacts affect almost all published leukemia genomics studies, explaining up to 40% of putative cancer-associated differential expression and alternative splicing.

To determine how standard blood collection procedures affect hematopoietic transcriptomes, we collected whole blood from four healthy donors in anticoagulant blood collection tubes. We then left this blood at room temperature for defined lengths of time (0-48h) in order to mimic the variable incubation that patient samples are typically subjected to during transfer from the primary treating physician to a research center, isolated peripheral blood mononuclear cells (PBMCs), and performed deep RNA-seq (Figure A). Contrary to the common assumption that RNA degrades during incubation, we observed no evidence of decreased RNA quality. Nonetheless, rapid and dramatic changes affected virtually every level of the gene expression process. The changes were highly biased; for example, pseudogenes and antisense RNAs were preferentially up-regulated, while cassette exons were preferentially repressed (Figure B).

Incubation-induced changes in the transcriptome confound the identification of true cancer-specific alterations. Many genes affected by incubation participate in biological pathways of current interest in leukemia, including cytokine production, NF-κB signaling, chromatin modification, and RNA splicing. Furthermore, incubation for as little as 4 hours, our shortest time point, introduced dramatic changes in the post-transcriptional landscape. We observed widespread isoform switches, wherein isoforms that were rare or even undetectable at 0h became the major isoform after 8 to 24h of incubation, in genes such as NOTCH2, LEF1, and PHF20 that have been previously used as leukemic biomarkers (Figure C). Perhaps most surprisingly, incubation rapidly inhibited RNA surveillance, leading to genome-wide expression of normally degraded RNAs. This highly abnormal RNA surveillance inhibition was readily detectable in all published leukemia genomics datasets that we analyzed, with the exception of TCGA, and undetectable in any lymphoma or solid tumor dataset (Figure D).

Together, our data show that incubation-induced changes in the transcriptional and post-transcriptional landscapes generate substantial artifacts that complicate the interpretation of leukemia genomics studies. To facilitate the incorporation of sample processing artifacts into downstream analysis, we provide biomarkers that detect prolonged incubation of individual samples. We furthermore show that the simple expedient of keeping blood on ice drastically reduces changes to the transcriptome. Our findings have important implications for the interpretation of published and ongoing leukemia genomics studies. Figure. (A) Leukemic samples are frequently subject to an ex vivo incubation period of variable length. (B) Ex vivo incubation causes biased up-regulation of pseudogenes and antisense RNAs and repression of cassette exons. (C) Incubation causes complete isoform switches in the LEF1 and PHF20 genes. (D) RNA surveillance is inhibited after only 4h of incubation. This abnormal inhibition of RNA surveillance is visible in all analyzed leukemia genomics datasets, with the exception of TCGA, and not in any lymphoma or solid tumor. Numbers above each x axis label indicate the number of samples in each dataset.

Figure 1

View largeDownload slide

Figure 1

View largeDownload slide

Close modal