April 1, 2016 (Vol. 36, No. 7)
Richard A. A. Stein M.D., Ph.D.
A Rogue’s Gallery of Malignant Outliers May Hide in Transcriptome Profiles That Emphasize Averages
In recent years, scientists have adopted a gene-centric view of cancer, a tendency to see each malignant transformation as the consequence of alterations in a discrete number of genes or pathways. These alterations are, fortunately, absent from healthy cells, but they pervert malignant cells.
The gene-centric view takes in molecular landscapes illuminated by genomic and transcriptomic technologies. For example, genomes can be cost-effectively sequenced within hours. Such capabilities have made it possible to interrogate associations between genotypes and phenotypes for increasing numbers of conditions, and to collect data from progressively larger patient groups.
As genomic and transcriptomic technologies rise, they reveal much—but much remains hidden, too. Perhaps these technologies are less like the sun and more like the proverbial streetlight, the one that narrows our searches because we’re inclined to stay in the light, even though what we hope to find may lie in the shadows.
“Each individual study that looks at the cancer transcriptome is impressive and tells a convincing story, but if we put several high-quality papers together, there are very few genes that overlap,” says Henry H. Heng, Ph.D., professor of molecular medicine, genetics, and pathology at Wayne State University. “This shows that something is wrong.”
One of the major observations in Dr. Heng’s lab is that the intra- and intertumor cellular heterogeneity results in nearly every cancer cell having a unique, distinct karyotype, that is, an important but often ignored genotype. “Biological systems need a lot of heterogeneity,” notes Dr. Heng. “People like to think that this is noise, but heterogeneity is a fundamental buffer system for biological function to be achievable. Moreover, it is the key agent for cellular adaptation.”
To capture the degree of genomic heterogeneity at the genome level and its impact on cancer cell growth, Dr. Heng and colleagues performed serial dilutions to isolate single mouse ovarian surface epithelial cells that had undergone spontaneous transformation. Spectral karyotyping revealed that within a short timeframe each of these unstable cells exhibited a very distinct karyotype. In these unstable cells, cloning at the level of the karyotype was not possible.
Stable cells exhibited a normal growth distribution, i.e., no subset of stable cells contributed disproportionately to the overall growth of the cell population. In contrast, unstable cell populations showed a non-normal growth distribution, with few cells contributing most to the cell population’s growth. For example, a single unstable colony contributed more than 70% to the cell population’s growth. This finding suggests that although average profiles can be used to describe non-transformed cells, they cannot be taken to represent the biology of malignant cells.
“Most people who study the transcriptome want to get rid of the noise, but the noise is in fact the strategy that cancer uses to be successful,” explains Dr. Heng. “Each individual cancer cell is very weak but together the entity becomes very robust.”
In a recent model that Dr. Heng and colleagues proposed, system inheritance visualizes chromosomes not merely as the vehicle for transmitting genetic information, but as the genetic network organizer that shapes the physical interactions between genes in the three-dimensional space. Based on this model, individual genes represent parts of the system. The same genes can be reorganized to form different systems, and chromosomal instability becomes more important than the contribution of individual genes and pathways to cancer biology.
The vital link between genomic instability and cancer progression is transcriptome dynamics, and the shifts in those dynamics that contribute to cancer evolution may come down to statistical outliers.
“Transcriptome studies rarely focus on single-cell analyses, which means important outliers are frequently ignored,” declares Dr. Heng. “This preoccupation with uninformative averages explains why we have learned so little despite having examined so many transcriptomes.”
Chimeras and Fusion Genes
“Our focus is on chimeric RNA molecules,” says Laising Yen, Ph.D, assistant professor of pathology at Baylor College of Medicine. “This category of RNAs is very special because their sequences come from different genes.”
In a study that was designed to capture chimeric RNAs in prostate cancer, Dr. Yen and his colleagues performed high-throughput sequencing of the transcriptomes from human prostate cancer samples. “We found far more chimeric RNAs, in terms of abundance, and a number of species that are not seen in normal tissue,” reports Dr. Yen. This approach identified over 2,300 different chimeric RNA species. Some of these chimeras were present in prostate cancer cell lines, but not in primary human prostate epithelium cells, which points toward their relevance in cancer.
“Most of these chimeric RNAs do not have a genomic counterpart, which means that they could be produced by trans-splicing,” explains Dr. Yen. During trans-splicing, individual RNAs are generated and trans-spliced together as a single RNA, which provides a mechanism for generating a chimera.
“The other possibility is that in cancer cells, where gene–gene boundaries are known to become broken, chimeras can be formed by cis-splicing from a very long transcript that encodes several neighboring genes located on the same chromosome,” informs Dr. Yen. Chimeric RNAs formed by either of these two mechanisms can potentially translate into fusion proteins, and these aberrant proteins may have oncogenic consequences.
Another effort in Dr. Yen’s laboratory focuses on chromosomal aberrations in ovarian cancer. One of the hallmarks of ovarian cancer is the high degree of genomic rearrangement and the increased genomic instability.
“When we looked at ovarian cancers, we did not find as many chimeric RNAs,” notes Dr. Yen. “But we found many fusion genes.” Gene fusions, similarly to chimeric RNAs, increase the diversity of the cellular proteome, which could be used selectively by cancer cells to increase their rates of proliferation, survival, and migration.
A recent study in Dr. Yen’s lab identified BCAM-AKT2, a recurrent fusion gene that is specific and unique to high-grade serous ovarian cancer. BCAM-AKT2 is the only fusion gene in this malignancy that was proved to be translated into a fusion kinase in patients, which points toward its functional significance and potential therapeutic value.
“Recurrent fusion genes, which are repeatedly found in many patients in precisely made forms, indicate that there is a reason that they are present,” concludes Dr. Yen. “This might have important therapeutic implications.”
“We contributed to a study of tumor gene expresssion that we are currently revisiting because so much more data has become available,” says Barbara Stranger, Ph.D., assistant professor, Institute for Genomics and Systems Biology, University of Chicago. “The data is being processed in homogenized analytic pipelines, and we can look at many more tumor types across the Cancer Genome Atlas than a few years ago.”
Previously, Dr. Stranger and colleagues performed expression quantitative trait loci (eQTL) analyses to examine mRNA and miRNA expression in breast, colon, kidney, lung, and prostate cancer samples. This approach identified 149 known cancer risk loci, 42 of which were significantly associated with expression of at least one transcript.
Causal alleles are being prioritized using a fine-mapping strategy that integrated the eQTL analysis with genome-wide DNAseI hypersensitivity profiles obtained from ENCODE data. These analyses are focusing on capturing differences across tumors and on performing comparisons with normal tissue, and one of the challenges is the lack of normal tissue from the same patients.
“But still there is a lot of power in these analyses because they are based on large-scale genomic datasets. Also, these tumor datasets can be compared with large-scale normal tissue genomics datasets, such as the NIH’s Genotype-Tissue Expression (GTEx) project,” clarifies Dr. Stranger. “This helps us characterize differences between those tumors and normal tissue in terms of the genetics of gene regulation.”
An ongoing effort in Dr. Stranger’s laboratory involves elucidating how the effect of genetic polymorphisms is shaped by context. Stimulated cellular states, cell-type differences, cellular senescence, and disease are some of the contexts that are known to impact genetic polymorphisms.
“We have seen a lot of context specificity,” states Dr. Stranger. “Our observations suggest that a genetic polymorphism can have a specific effect in regulating a particular gene or transcript in one context, and another effect in another context.”
Another example of cellular context is sex, and an active area of investigation in Dr. Stranger’s lab proposes to dissect the manner in which sex differences shape the regulatory effects of genetic polymorphisms.
“Thinking about sex-specific differences is not very different from thinking about a different cellular environment,” notes Dr. Stranger.
The expression of specific transcription factors can be determined by sex; consequently, a polymorphism that interacts with a transcription factor may have functional outcomes that can be seen in only one of the sexes.
“There are gene-level and gene-splicing differences that we see in normal tissues between males and females, and we want to take the same approach and look at the cancer context to see whether the genetic regulation of gene expression and transcript splicing is different between individuals and whether it has a sex bias,” concludes Dr. Stranger. “Finally, we want to see how that differs in cancer relative to normal tissues.”
Early Clinical Impact
“Over the last two years,” says Andrew Kung, M.D., Ph.D., chief of the Division of Pediatric Hematology, Oncology, and Stem Cell Transplantation at Columbia University Medical Center, “we have included the cancer transcriptome as part of our precision medicine program.” Dr. Kung and colleagues developed a clinical genomics test that includes whole-exome sequencing of tumors and normal tissue and RNA-seq of the tumor.
“Our results show that the transcriptome is very important in identifying clinically impactful results,” asserts Dr. Kung. “The technology has really moved from a research tool to real clinical application.” In fact, the test has been approved by New York State for use in cancer patients.
The data from transcriptome profiling has enabled identification of translocations, verification of somatic alterations, and assessment of expression levels of cancer genes. Dr. Kung and his colleagues are using genomic information for initial diagnosis and prognostic decisions, as well as the investigation of potentially actionable alterations and the monitoring of disease response.
To gain insight into gene-expression changes, transcriptome analysis usually compares two different types of tissues or cells. For example, analyses may attempt to identify differentially expressed genes in cancer cells and normal cells.
“In patients with cancer, we usually do not have access to the normal cell of origin, making it harder to identify the genes that are over- or under-expressed,” explains Dr. Kung. “Fortunately, the vast amounts of existing gene-expression data allow us to identify genes whose expression are most changed relative to models built on the expression data aggregated across large existing datasets.”
These genomic technologies were first used to augment the care of pediatric patients at Columbia. The technologies were so successful that they attracted philanthropic funding, which is being used to expand access to genomic testing to all children with high-risk cancer across New York City.