Skip to main content
Premium Trial:

Request an Annual Quote

CRISPR Screening Helps Parse Noncoding Genomic Regions


NEW YORK (GenomeWeb) – That CRISPR/Cas9 is useful for interrogating protein-coding genes is more than evident. Now, researchers are proving CRISPR can be equally useful for characterizing the noncoding elements of genomes as well.

Two teams from the Broad Institute published papers today in Science demonstrating the power of unbiased CRISPR-based screening to tease out the roles of noncoding genomic elements.

While some studies have used CRISPR to disrupt known functional elements, the new studies are systematic looks at large stretches of the genome and pick out any elements that affect gene activity. Using single guide RNA (sgRNA) tiling arrays, these screens can disrupt sequences hundreds of kilobases long.

In one paper, scientists led by first authors Neville Sanjana, now at the New York Genome Center, and Jason Wright and senior author Feng Zhang used a high-resolution CRISPR/Cas9-based knockout screen to interrogate the regions surrounding three genes implicated in resistance to a cancer drug. Knockouts to those genes often lead to BRAF inhibitor resistance in melanoma; the screen revealed several regions in which indel mutations led to a similar phenotype as knockout of the gene itself.

"This is a peek at what we can do," Sanjana told GenomeWeb. There's an enormous amount of territory in the noncoding genome to cover — 98 percent of the approximately 3 billion base pairs in the human genome — but doing the screens 100 kilobases at a time is manageable, he said, though time and resource intensive. "What we need is to look at larger regions at a coarser resolution, then maybe zoom in," he said.

The other paper could offer the perfect complement, a high-throughput, less fine-grained screening method using CRISPR interference (CRISPRi) via a chromatin regulation, rather than knockout. Led by senior authors Eric Lander and Jesse Engreitz, the other Broad Team mapped more than 1,000 kilobases of non-coding genome thought to regulate the transcription factors MYC and GATA1.

"CRISPRi has a slightly larger blast radius," Engreitz said. "That means lower resolution, but a better quantitative estimate of the effect." It's powerful enough to provide a set of principles that can predict the existence of enhancers in different cell types. "As we expand these mapping experiments to additional cell types and gene loci, we may even be able to build models that forgo screens and find target genes," he said.

The papers suggest a systematic way to map the functions of non-coding regions. There may even be immediate practical applications for the results in drug discovery and patient care, considering the genes are involved in cancer.

Engreitz said that while his group came at the screen more interested in uncovering the regulatory wiring of the genome than in technology development, they adopted a similar approach as Zhang's team, born out of lengthy discussions between the researchers.

"We've been throwing ideas back and forth," he said.

Both studies can trace their roots back to CRISPR screens developed by Broad researchers. The current set of papers echoes a pair of papers from Zhang, Sanjana, and Lander published in 2014 establishing the idea of the pooled lentiviral knockout screen, where the viral vector delivers an sgRNA to not only direct Cas9 but also serve as a barcode.

In Sanjana and Zhang's 2014 paper, as in the new one, the barcode gets amplified in cells where loss of function of a particular gene confers resistance to vemurafenib, a BRAF inhibitor used to treat melanoma. Those cells proliferate, increasing the number of sgRNA barcodes in each pooled dish, indicating which genes contributed to that resistance.

The new paper from Sanjana and Zhang simply recapitulates that experiment, but with a twist. "The idea is to detect elements in the noncoding genome where modulating expression of the gene has a similar effect to knocking out the gene," Sanjana said.

Sanjana had also done screens to knock out a predetermined list of microRNAS in mouse and to zero in on an element that regulated the human gene BCL11A, which blocks expression of fetal hemoglobin.

"Here we took the approach of, 'What if we didn't use any prior information?'" Sanjana said.

The screen turned up many regions, not all of which the team could validate. Sanjana said they curated a list of 25 that could be binned into several categories, like sites on the promoter, in the UTR, and far away on the 5' end. But of the 25 elements they chose to follow up on, 24 led to resistance when knocked out on their own. "In every category you can slice and dice this region into, we were able to find elements enriched in the screen that seem to, when mutagenized individually, really confer resistance."

Sanjana suggested that the screen yielded some immediately actionable insights into cancer biology. "It can be used in a preclinical sense, before a pharma company invests a significant amount of time into a drug, they can understand how mutations in genes and outside of genes might impact that drug's performance," and how tumors gain resistance to a drug, he said. "For many treatments there's just one drug. Life finds a way. [CRISPR screens] can be used as a preclinical tool to find where that resistance is coming from."

He added that once whole-genome sequencing becomes more routine in cancer care, mutations in noncoding regions that have the same resistance-conferring effect as a protein-coding mutation could be used to stratify patients.

Engreitz's study also used a lenti-CRISPR screen; however, in addition to using CRISPRi, it employed a negative selection screen. Instead of looking at elements that helped cells proliferate, it looked for cells that caused loss of function of MYC, a key oncogene that drives cell proliferation and also a gene with an incredibly complex regulatory structure. "We took the MYC locus on in part because it was so complex. There are potentially hundreds of enhancers predicted to work in the region," Engreitz said. "If we could figure out the things that control MYC, then any other gene should be easy."

Specifically, the screen used nuclease-null Cas9 (dCas9) fused to KRAB chromatin remodeling domain. "This recruits a series of proteins that essentially shut down enhancers at the level of chromatin regulation," he said.

The screen identified a specific set of seven regulatory elements that dialed MYC down by varying amounts. "Some of these elements were extremely far away [from the gene]," Engreitz said. "We found a constellation of elements that seemed to control the gene and validated that they indeed do."

But only a subset of the predicted MYC enhancers actually affected the gene. "It's critical for the experiment to test the specific functions of these elements, as opposed to elements that are correlated," such as chromatin marks.

Sanjana also noted the importance of showing a direct relation between the noncoding element and a phenotype. "A lot of the assays we have to think about noncoding function [such as DNase1 hypersensitivity, ATAC-Seq, and ChIP-Seq] really are indirect readouts of function," he said.

Engreitz envisioned using the screen to identify the noncoding regions that affect any given gene. While not all genes are so intertwined with selective pressures like cell proliferation, he suggested that coupling green fluorescent protein to any gene could serve as an artificial selection pressure.

Sanjana concurred. "The idea is to create a sea of mutations and apply any selective pressure that is relevant to the biology you want to study or a particular disease," he said. "There's so much that can be done."