Postdoc, laboratory of David Sabatini, Whitehead Institute for Biomedical Research
• PhD, experimental pathology, Harvard University — 2008
• BS, biology/chemistry, Massachusetts Institute of Technology — 2001
Researchers from the Whitehead Institute, in collaboration with the Massachusetts Institute of Technology and the Broad Institute, reported this month on the discovery of an essential pathway in breast cancer using a negative-selection RNAi screen in vivo.
Using the approach, the team screened a set of metabolic genes associated with aggressive breast cancer and stemness to find ones required for in vivo tumorigenesis, according to their paper, which appeared in Nature.
One gene in particular, phosphoglycerate dehydrogenase, was found to be up-regulated in estrogen receptor-negative breast cancers, and experimentation showed that certain of these cancers are dependent on increased serine pathway flux triggered by PHGDH over-expression.
This week, Gene Silencing News spoke with Richard Possemato, the lead author of the study, about the findings.
Let's start with some background on the research that led to this work.
We at the Broad and in the Sabatini lab had undertaken a variety of in vitro screens, mostly array-based … but we were wondering whether we could screen directly in vivo and, if we did a large set of genes, whether there would be any striking difference with what we found in vitro.
The thought process was that we would choose a set of genes that were very likely to be essential in the ER-negative breast cancer model we were using … as a pilot screen. If it worked in that cell line, we'd then expand it to a much larger set of genes to ask whether screening directly in vivo versus in vitro would yield different results.
How is the screening approach you took different than ones done by the lab before?
Instead of doing the screen in culture, we're taking cells that have been recently infected with a pooled shRNA library.
[The shRNAs were injected] into animals forming tumors, then [we] captured the DNA from pre-injected cells and the tumors [themselves]. Using a PCR technique we developed in collaboration with the Broad, [we aimed to] figure out what the fold change is from before and after the culture.
In order to do that, we had to get better at doing the PCR detection technique, which we hadn't done before, and then we would have to show that we could actually apply that to in vivo screening.
For the PCR technique, [conducting] Illumina-based deep sequencing is a little bit expensive — it's about $1,000 per run — so there is no way we'd ever be able to do this project. We had to engineer [into the deep sequencing approach] a way to amplify the hairpin region that we were detecting, along with a barcode for the sample, so we could then run multiple samples in one lane.
We've done up to about 30 samples per lane, [while] in the paper we were doing 16. Now, the scientists at the Broad are in the range of 50 or so.
Was the Broad's RNAi Consortium involved in the work?
They gave us advice on how to do the PCR detection technique.
They have published on pooled RNAi screens … using the chip-based detection method, and that's great if you're doing a whole-genome screen. But if you only care about 133 genes, as we did here, it's hard to justify having an Affymetrix chip as an output because you're wasting all that space. So we put together the Illumina detection method with a pooled screening method.
Can you give a snapshot of the study's findings?
First, we showed that we were able to do pooled screening in vivo using the Illumina detection method … getting good replicates across the tumor samples we [tested]. Using the data we gathered from that screen, we identified a set of 12 to 15 high-priority genes. We showed that, in fact, the data we got from the screen was robust and the hits — we ended up looking at about six of them — behaved as we expected.
[Using] our hit list of genes that we identified in the screen … [we identified a] set of 133 genes that were associated with ER-negative breast cancer, that were up-regulated globally in cancer, and were associated with a stemness phenotype.
We wanted to prioritize the hit list … and found that one gene in particular, PHGDH, was in a peak of amplification on chromosome 1 in breast cancer and some other types of cancer. It was very strongly associated with ER-negative disease, more than any of the other genes in the study.
We took a set of breast cancer cell lines, either amplified for PHGDH or not, and the amplified cell lines had quite a bit more expression than the non-amplified cell lines. There were, it seems, a set of ER-negative breast cancer cell lines that, despite not being amplified, also up-regulated expression. Both cell lines that had amplified PHGDH at the genomic level and those that had up-regulated PHGDH through some other mechanism were sensitive to suppression of PHGDH, much more so than cell lines with little or no PHGDH expression.
Overall, using gene-expression data and some immunohistochemistry staining for patient samples, we determined that approximately 70 percent of ER-negative cancer has elevated levels of this gene.
This, itself, is associated with elevation of a lot of the other genes in the serine biosynthesis pathway, and it really drives flux through that pathway as it comes directly off of glycolysis. There is a substantial amount of flux through glycolysis, and we calculated that this pathway is about 8 to 9 percent of glycolytic flux, which is a substantial amount [due to] the fact that most of the glucose that is taken up by the cancer cell is immediately converted to lactate.
Are there plans to follow up on these findings with regards to the screening method?
For the screening method, there are plans to do the screen in a larger set of genes to really ask this in vivo-versus-in vitro question more thoroughly.
[Ultimately,] we want to have a comprehensive study of cancer metabolism. You could imagine that there might be a pathway where all the genes are essential, and there may be other pathways where key enzymes are essential. And so by having a look at all 2,800 enzymes and transporters that we've [previously] identified, we're hoping to have a very broad idea of which pathways are the most essential to get us a little bit away from looking at central carbon metabolism, which is obviously very important, [so that we can] cast a wider net into other areas of metabolism that haven't already been mined.
It's interesting with studying metabolism, in a sense there are a lot of genes with unknown function, but many of them fall very nicely into this network. Hopefully we'll be able to use the screening data with the network map to pull out these interesting associations.