CRISPR-Based Functional Transcriptomics Defines the Tumor-Dependency and Molecular Determinants of Long Noncoding RNAs in Multiple Myeloma

Long noncoding RNAs (lncRNAs) constitute most elements in the human transcriptome, yet we lack a comprehensive understanding of their role in tumor cells. Here, in multiple myeloma (MM), we exploited the RNA-targeting activity of the CRISPR-Cas13d endonuclease to develop a functional transcriptomics screening platform that enabled not only the transcriptome-wide identification of tumor-promoting-lncRNAs (tp-lncRNAs) in MM cells – outperforming any prior attempt – but also provided mechanistic information about their sub-cellular site-of-function, the isoform-conferring activity, and their downstream molecular pathways.

Using RNA-seq data from CD138+ plasma cells from 319 clinically annotated, homogeneously treated, newly diagnosed MM patients, we first identified 5,304 lncRNAs that were expressed in ≥10% of patients, and then designed and generated a pooled lentiviral library with 9 sgRNAs per identified lncRNA, ~500 non-targeting sgRNAs as negative controls, and 9 sgRNAs targeting known protein-coding oncogenes in MM, such as IRF4, as positive controls. We infected this library at low MOI (<0.3) in five MM cell lines (AMO1, H929, KMS11, OPM2, R8226), each expressing a Cas13d-GFP fusion protein. After 3 weeks, we tested for sgRNAs that were relatively depleted or enriched in the MM cell population using deep sequencing and the MAGeCK robust rank aggregation algorithm.

From this, we identified ~600 tp-lncRNAs upon which MM cell lines exhibit significant growth dependency, making up ~12% of the identified lncRNA transcriptome, significantly larger than the previous estimate of 3–5%. The tp-lncRNAs overlapped across MM cell lines: 17% were observed in all cell lines tested, 26% in four cell lines, 34% in at least three cell lines, and 53% in at least two cell lines. The dependency score of many tp-lncRNAs was on a par or even superior to the dependency score of the positive control IRF4 and, as expected, the top-scoring tp-lncRNAs included RROL, MALAT1, and NEAT1. We used a small library of 3,500 gRNAs to validate the dependency on 314 tp-lncRNAs and further validated the dependency on 12 tp-lncRNAs by individually targeting them using antisense oligonucleotides or CRISPR-Cas13d. These studies were followed by downstream transcriptomic analyses using either qRT-PCR or RNA-seq, which validated the growth phenotypes and informed on function.

Interestingly, when comparing CRISPR-Cas13d to the current state-of-the-art CRISPR interference screens, we found that the CRISPR-Cas13d platform was superior, with a 4- to 30-fold increase in identifying tp-lncRNAs among 913 lncRNAs of intergenic origin in the same cell lines, and with the additional ability to identify tp-lncRNAs generated at gene loci that are antisense to protein-coding genes.

We further investigated the sub-cellular localization and site-of-function of the tp-lncRNAs by using a Cas13d that accumulates either in the cytosol or in the nucleus as well as sub-cellular RNA-seq and single molecule RNA FISH. This way, while confirming that most tp-lncRNAs were functional and enriched in the cell nuclei, we also identified 20 hits providing dependency in the cytosol, including the small nucleolar RNA host genes 6 and 10. We observed that the lncRNA transcriptome investigated in this study contained several transcript variants produced from the same lncRNA genes, which differ in their sequences and likely in their structures and functions. Importantly, our screening platform was able to precisely identify tp-lncRNAs isoform(s) conferring dependencies even among many isoforms, a result that was corroborated by RNA-seq and/or qRT-PCR analysis after individually targeting the tp-lncRNA isoforms with Cas13d.

Finally, from our screens, we identified and validated the MM dependency on the novel lncRNA Plasma Cell Specific RNA 1 (PLASP-1), which is one of the most abundant lncRNAs in MM and normal plasma cells (median TPM>25), whereas it has significantly lower expression or is absent in 54 normal tissues and cell types. PLASP-1 is an un-spliced transcript that directly interacts with spliceosome proteins, as established in living cells. We are currently investigating its functional impact on the splicing machinery and the results will be presented.

Overall, this novel platform comprehensively defines tumor dependency on lncRNAs in MM and provides information to guide further mechanistic investigation.