CRISPR-Cas13d functional transcriptomics reveals widespread isoform-selective cancer dependencies on lncRNAs

Key Points

• •Tumor cells exhibit widespread lncRNA dependencies uncovered by CRISPR-Cas13d screening.
• •Our IsoScan platform precisely maps functional lncRNA isoforms driving oncogenic activity.

Visual Abstract

Abstract

Long noncoding RNAs (lncRNAs) are a significant yet largely uncharted component of the cancer transcriptome, with their isoform-specific functions remaining poorly understood. In this study, we used RNA-targeting CRISPR-Cas13d to uncover and characterize hundreds of tumor-essential lncRNA (te-lncRNA) isoforms with clinical relevance. Focusing on multiple myeloma (MM), we targeted the lncRNA transcriptome expressed in tumor cells from patients with MM and revealed both MM-specific and pan-cancer dependencies across diverse cancer cell lines, which we further validated in animal models. Additionally, we mapped the subcellular localization of these te-lncRNAs, identifying >30 cytosolic isoforms that proved essential when targeted by cytosol-localized Cas13d. Notably, a specific isoform of small nucleolar RNA host gene 6, enriched in the endoplasmic reticulum, interacts with heat shock proteins to maintain cellular proteostasis. We also integrated functional and clinical data into the publicly accessible LongDEP Portal, providing a valuable resource for the research community. Our study offers a comprehensive characterization of te-lncRNAs, underscoring their oncogenic roles and therapeutic potential.

The isoform-specific functions of long noncoding RNAs (lncRNAs) remain poorly understood. Morelli et al used RNA-targeting CRISPR-Cas13d to identify and characterize lncRNA isoforms that are essential for multiple myeloma cell survival. This work provides a comprehensive catalog of tumor-essential lncRNAs in multiple myeloma, laying the groundwork for future research to elucidate the distinct functional roles of individual lncRNA isoforms in tumor cell viability.

Introduction

Long noncoding RNAs (lncRNAs) are abundant, nonprotein-coding elements of the human genome that undergo extensive alternative splicing, generating multiple isoforms with diverse functions, from molecular scaffolding to enzymatic activity.^1,2 These transcripts can originate from intergenic or antisense regions and may contain additional genetic elements such as small RNAs.³ Moreover, lncRNAs operate in distinct subcellular compartments, with their function often linked to their localization.⁴

Emerging evidence underscores the pivotal roles of lncRNAs in cancer, positioning them as promising therapeutic targets.⁵ This is particularly evident in multiple myeloma (MM), a challenging hematologic malignancy⁶ in which alterations in intergenic lncRNA expression in newly diagnosed patients serve as independent predictors of clinical outcomes.⁷ Several lncRNAs, including MALAT1,^8,9 NEAT1,¹⁰ SMILO,¹¹ RP11-350G8.5,¹² and MIR17HG,^13,14 have been implicated in MM cell growth. Notably, RP11-350G8.5 and MIR17HG were identified via CRISPR-CRISPR–associated protein 9 (Cas9) and CRISPR interference (CRISPRi) screens, respectively, approaches that have only revealed a few functional lncRNAs and struggle with those from antisense loci. Further studies indicate that MIR17HG drives tumor growth through the multifaceted roles of its isoforms, functioning as precursors of cytosolic microRNAs¹³ or as molecular scaffolds for chromatin interactions; these latter isoforms are also referred to as lnc-17-92 or RROL (RNA Regulator of Lipogenesis).^14,15

Given this complexity, advanced screening platforms are needed to capture the full spectrum of lncRNA diversity. To address this, we adapted RNA-targeting CRISPR-Cas13d^16,17 to precisely target lncRNA isoforms across genomic contexts and subcellular compartments. Our novel platform, IsoScan, was applied to MM and non-MM cancer models, revealing both common and unique vulnerabilities. Importantly, the ability to selectively target individual isoforms without affecting other homologous variants addresses a key challenge in the field, because isoforms may have distinct or even opposing functions within the same gene.¹⁴ Our approach leverages isoform-specific guide RNAs to mitigate off-target effects, and we systematically validated our findings by targeting alternative isoforms that do not elicit the same phenotypes. Further refinement of these strategies will be crucial to developing therapeutics that selectively target biologically significant isoforms while minimizing unintended consequences.

We have made our findings accessible via the LongDEP Portal (LongDEP Portal), providing a comprehensive resource on lncRNA isoform functions and clinical associations.

Methods

Cells

Cell lines and patient cells were grown at 37°C at 5% CO₂ and were cultured in RPMI 1640 medium (Gibco Life Technologies, Carlsbad, CA) supplemented with 10% fetal bovine serum (Lonza Group Ltd, Basel, Switzerland) and 1% penicillin/streptomycin (Gibco, Life Technologies), as previously described.^13,14,18 More detailed information is included in the supplemental Methods, available on the Blood website.

Experiments with CRISPR-Cas13d

The generation of cell lines stably expressing Cas13d and their use for viability screens followed our established protocols^14,19 and is detailed in the supplemental Methods.

RNA expression analysis

Total RNA was extracted from cells with TRIzol reagent (Thermo Fisher Scientific), according to the manufacturer’s instructions, as previously described.^13,14,18 Then, RNA expression was evaluated by quantitative reverse transcription polymerase chain reaction (qRT-PCR) or RNA sequencing (RNA-seq), as detailed in the supplemental Methods. Subcellular RNA-seq (subRNA-seq) was performed following the protocol described elsewhere.²⁰ Cytoplasm, endoplasmic reticulum (ER), and mitochondria fractionation as well as ascorbate peroxidase (APEX) assay are explained in the supplemental Methods.

RNA-protein and protein-protein interaction assays

Comprehensive identification of RNA-binding proteins (ChIRP), RNA immunoprecipitation, and coimmunoprecipitation experiments are detailed in the supplemental Methods.

Animal studies

Six-week-old female immunodeficient NOD.CB17-Prkdcscid/NCrCrl (NOD/SCID) mice (Charles River) were housed in our animal facility at Dana-Farber Cancer Institute (DFCI). All experiments were performed after approval by the animal ethics committee of the DFCI and performed using institutional guidelines, as previously described.^13,14,18 More detailed information is included in the supplemental Methods.

Statistical analysis

Statistical analysis is described in the supplemental Methods.

Results

Identification of tumor-essential lncRNA (te-lncRNA) isoforms through an IsoScan CRISPR-Cas13d screen

We used RNA-seq to assess the expression of lncRNA isoforms in MM cells from 319 patients with newly diagnosed MM (NDMM), focusing on transcript isoforms originating from both intergenic and antisense lncRNAs. Our analysis identified a total of 5327 lncRNA isoforms (transcripts per million [TPM] >0.5) derived from 3793 lncRNA genes in at least 10% of patient samples (Figure 1A; supplemental Table 1). These transcripts were composed of both intergenic (46%) and antisense (54%) lncRNAs. We also validated the expression of these transcripts in the MM cell lines (MMCLs) used for subsequent functional screens (supplemental Figure 1A).

Figure 1.

Identification of te-lncRNA isoforms through IsoScan CRISPR-Cas13d screen. (A) RNA-seq analysis of 319 patients with NDMM identified 5327 lncRNA isoforms as expressed (TPM >0.5) in at least 10% of patients. (B) Schematic of lncRNA isoform-targeting strategy for the design of libraries of gRNAs (upper panel) and relative composition of IsoScan-L1 in isoform-specific and multi-isoform gRNAs (lower panel). (C) Schematic of the screening procedure and analysis. (D) Identification of te-lncRNA isoforms in 1 (orange), 2 (purple), 3 (light blue), 4 (green), or 5 (red) Cas13d-MMCLs, using IsoScan-L1. (E) Schematic of secondary screens in vitro with IsoScan-L2. Overlapping te-lncRNA isoforms in primary and secondary screens are shown for each cell line. (F) Robust ranking aggregation–based ranked analysis of te-lncRNA isoforms in the secondary screens with H929^Cas13d cells in vitro and in xenografts. The Venn diagram highlights the hit overlap in vitro and in xenograft. (G) Identification of strong hits in the primary screen with IsoScan-L1. The positive control lncRNA dependencies (isoforms of MALAT1, NEAT1, and MIR17HG) and mRNA dependency (IRF4) are highlighted. The dotted red line indicates a logFC depletion of –1, set to identify the strong hits. L1, library 1; L2, library 2; NGS, next-generation sequencing.

To target the lncRNA transcriptome using CRISPR-Cas13d, we constructed a pooled lentiviral library (IsoScan-L1) containing 9 guide RNAs (gRNAs) per target, designed to be isoform-selective by base pairing at unique exon-exon junctions or distinctive noncoding exons (Figure 1B). The library included 55 405 gRNAs targeting the MM lncRNA transcriptome, along with various controls, such as nontargeting gRNAs as negative controls and gRNAs targeting essential factors such as the transcription factors IRF4 and lncRNA isoforms of MALAT1, MIR17HG, and NEAT1, as positive controls (supplemental Table 2).

To enhance targeting precision and minimize off-target effects, we implemented a rigorous selection pipeline for gRNA design, prioritizing sequences with minimal homology to other isoforms or overlapping protein-coding transcripts. As a result, >95% of the gRNAs were predicted to bind exclusively to a single transcript within the assessed lncRNA transcriptome (Figure 1B).

We engineered 5 MMCLs to express Cas13d (Cas13d-MMCLs: AMO1, H929, KMS11, OPM2, and R8226) and confirmed system efficiency using gRNAs against IRF4 messenger RNA (mRNA), a known strong protein-coding dependency in MM,²¹ which reduced IRF4 expression and affected cell viability and IRF4 canonical targets in Cas13d-MMCLs but not in Cas13d^– MMCLs (supplemental Figure 1B-D). Subsequently, we performed a viability screen by infecting the Cas13d-MMCLs with IsoScan-L1 and identifying depleted gRNAs through deep DNA sequencing and MAGeCK (model-based analysis of genome-wide CRISPR/Cas9 knockout) analysis (Figure 1C). Cas13d^– AMO1 MM cells served as negative controls, yielding no significant hits (supplemental Figure 1E). Our analysis revealed significant depletion (false discovery rate [FDR] <0.05) of gRNA pools targeting 598 lncRNA isoforms in at least 1 Cas13d-MMCL, constituting 12% of the MM lncRNA transcriptome. These 598 te-lncRNA isoforms were derived from 479 lncRNA genes and encompassed a total of 1180 isoforms targeted by IsoScan-L1. These isoforms included both intergenic (n = 343 [58%]) and antisense (n = 255 [42%]) transcripts. Notably, only some transcripts had previously established tumor-essential activity in MM, such as isoforms of MIR17HG, MALAT1, and NEAT1. Importantly, te-lncRNAs showed significant overlap across Cas13d-MMCLs: 101 were in all 5 MMCLs, whereas an additional 54 were in at least 4, 53 in at least 3, and 107 in at least 2 MMCLs (Figure 1D).

For validation, we performed secondary screens in vitro and in xenografts, targeting all 315 te-lncRNAs common to at least 2 Cas13d-MMCLs (Figure 1E). Infection of this secondary library, comprising ∼3500 gRNAs (IsoScan-L2; supplemental Table 3), confirmed depletion of gRNA pools targeting most te-lncRNA isoforms in each cell line cultured in vitro (ranging from 78% in KMS11% to 98% in R8226; Figure 1E). For the xenograft plasmacytoma model, H929^Cas13d cells were infected with IsoScan-L2 and subcutaneously injected into NOD SCID mice. Analysis of tumor growth revealed 146 te-lncRNA isoforms, with depletion scores generally stronger compared with those observed in vitro (Figure 1F). Overall, there was substantial overlap between the significant hits identified in the in vitro and in vivo screens with H929^Cas13d cells (75%).

Furthermore, using MALAT1 as a positive control, we observed that significantly depleted gRNAs in the screens effectively inhibited MM cell growth, with a strong correlation between the growth phenotype and MALAT1 downregulation measured by qRT-PCR (R² = 0.94; P < .0001; supplemental Figure 1F).

Remarkably, several te-lncRNA–targeting gRNA pools exhibited depletion levels equal to or greater than the gRNA pool targeting IRF4 mRNA. In the primary screen, 272 te-lncRNAs demonstrated substantial depletion (log fold change [logFC] less than –1 and FDR of <0.05; Figure 1G). This included positive controls such as isoforms of MALAT1, NEAT1, and MIR17HG; of these, 183 te-lncRNAs (67%) were validated as strong hits by IsoScan-L2.

These data highlight a broad and strong dependency on lncRNA transcripts in MM.

Validation of the isoform-selective targeting activity

We investigated the isoform-selective activity of our gRNAs by individually targeting 7 isoforms, small nucleolar RNA host gene 6–003 (SNHG6-003), SNHG8-005, GAS5-003, DANCR-001, ASH1L-AS1-001, LINC-PINT-004, and KDM4-AS1-003, which emerged as hits in our screens (Figure 2A). These isoforms, derived from 7 distinct lncRNA genes, represent 31 isoforms within the MM lncRNA transcriptome targeted in our study. For each isoform, we validated the phenotypic changes observed in the screens (ie, inhibition of cell proliferation) and confirmed effective isoform-specific downregulation (Figure 2B).

Figure 2.

Validation of the isoform-selective targeting activity. (A) Heat map showing the depletion score (logFC) from the primary screen with IsoScan-L1 of 7 te-lncRNAs and their isoforms. The isoforms validated in Figure 1B are underlined. (B) CCK-8 proliferation assay (black bars) and qRT-PCR analysis (red line) of H929^Cas13d cells infected with gRNAs targeting the reported te-lncRNA isoforms. Two gRNAs per te-lncRNAs were used. Data are reported as the percentage of NT gRNAs. (C) Expression of SNHG6 isoforms (-003, -005, -006) in a cohort of 319 patients with NDMM. We reported the median TPM expression on the y-axis and the percentage of patients expressing >0.5 TPM of a particular isoform on the x-axis. (D) Schematic (simplified) of SNHG6 locus. Red lines indicate the gRNAs used in the screen. (E) Depletion scores (logFC depletion) of the targeted isoforms of SNHG6 in the primary and secondary screens in vitro and in xenograft. (F) RNA-seq analysis of H929^Cas13d cells infected with 2 gRNAs specific for SNHG6-003 or NT gRNAs. The relative expression of SNHG6 isoforms -003, -005, -006 is shown. (G) qRT-PCR analysis of SNORD87 in H929^Cas13d cells infected with 2 gRNAs specific for SNHG6-003 or NT gRNAs. ns means P > .05 after the Student t test. (H) RNA-seq analysis of H929^Cas13d cells infected with 2 gRNAs specific for SNHG6-005 or NT gRNAs. The relative expression of SNHG6 isoforms -003, -005, and -006 is shown. (I) CCK-8 proliferation assay (live cells) and qRT-PCR analysis of SNHG6 isoforms -003, -005, and -006, of H929^Cas13d cells infected with gRNAs targeting SNHG6-006. Two gRNAs targeting SNHG6-006 were used. Data are reported as the percentage of NT gRNAs. ns, not significant; NT, nontargeting; pts, patients.

Among these 7 validated hits, SNHG6-003, SNHG8-005, GAS5-003, and DANCR-001 are SNHGs, which have recently been implicated in cancer pathogenesis because of their unique genomic sequences that also host small nucleolar RNAs.²² Furthermore, some SNHGs (eg, GAS5 and SNHG8) are antisense to other genes. Consequently, we focused on 2 SNHG lncRNAs (SNHG6 and GAS5) to further explore isoform selectivity in depth.

The lncRNA SNHG6 expresses 3 isoforms (-003, -005, -006) in the MM transcriptome, with nearly all patients with MM expressing each (TPM >0.5). SNHG6-003 shows the highest expression (median TPM = 31.52), followed by SNHG6-006 (median TPM = 11.88) and SNHG6-005 (median TPM = 5.31; Figure 2C). Isoform-selective gRNAs targeting these transcripts revealed that SNHG6-003 is the predominant functional isoform in vitro and in xenografts (Figure 2D-E), with validation of its antiproliferative and knockdown effects shown in Figure 2B. RNA-seq confirmed that SNHG6-003 gRNAs selectively downregulated this isoform. In contrast, the other isoforms were upregulated in H929^Cas13d and KMS11^Cas13d cells, suggesting negative regulation among isoforms (Figure 2F; supplemental Figure 2A). Additionally, SNHG6 serves as the host for SNORD87, which remained unaffected by SNHG6-003 knockdown (Figure 2G). Similarly, gRNAs targeting SNHG6-005 and SNHG6-006 demonstrated isoform-selective activity (Figure 2H-I). The protein-coding gene MCMDC2 is located antisense to SNHG6. However, we could not detect its expression in wild-type KMS11 and H929 cells by RNA-seq (data not shown).

The lncRNA GAS5 produces >90 transcripts; our RNA-seq detected 9 isoforms (TPM >0.5) in >10% of patients with MM (-001, -003, -008, -014, -015, -016, -020, -027, and -028; supplemental Figure 2B). These isoforms vary in expression from a median TPM of 49.21 (GAS5-001) to 0.13 (GAS5-027) and are present in 13% to 100% of patients. GAS5-003 emerged as the most prominent hit (supplemental Figure 2C), and its selective knockdown was validated (Figure 2B). RNA-seq confirmed that GAS5-003 gRNAs preferentially reduced GAS5-003 relative to other GAS5 isoforms and the antisense transcripts GAS5-AS1, ZBTB37, and DARS2 (supplemental Figure 2D), and qRT-PCR showed no effect on the intronic SNORD76 (supplemental Figure 2E).

These results demonstrate the effectiveness and selectivity of IsoScan gRNAs in targeting specific lncRNA isoforms, revealing their roles as effectors of lncRNA activity.

Identification of cancer-common and MM-selective te-lncRNA isoforms

To extend the investigation of te-lncRNA isoforms across different cancer types, we introduced IsoScan-L2 into 6 additional cancer cell lines expressing Cas13d and with diverse origins including SU-DHL-4 (diffuse large B-cell lymphoma), KARPAS-422 (diffuse large B-cell lymphoma), RAJI (B-cell lymphoma, Burkitt lymphoma), A549 (non–small-cell lung cancer), MCF7 (breast cancer), and HCT-116 (colorectal cancer). In this expanded screening, 257 te-lncRNA isoforms (82% of those tested) emerged as strong hits in at least 1 cell line (Figure 3A). Notably, among them, 42 lncRNA isoforms (16%) were strong hits across all 11 tested cancer cell lines, whereas 100 lncRNA isoforms (33%) exhibited strong hits in at least 6 cancer cell lines, including at least 3 MMCLs and 3 non-MMCLs, collectively designated as “cancer-common” te-lncRNA isoforms (highlighted in red in Figure 3A). Examples of cancer-common te-lncRNA isoforms are those originating from lncRNAs with (1) established roles in cancer, such as MALAT1,^8,9 NEAT1,¹⁰ and MIR17HG^13,14; (2) lncRNAs previously not associated with cell proliferation, such as DANCR²³; (3) those identified as tumor suppressors, such as GAS5²⁴ and GMDS-AS1²⁵; and (4) SNHG6-003, a previously validated lncRNA isoform (in Figure 2 and highlighted in Figure 3A, and previously reported by Cao et al²⁶). This expanded screen also identified 8 “MM-selective” te-lncRNA isoforms as strong hits in at least 2 MMCLs with no impact on other cancer cell lines (highlighted in blue in Figure 3A). The positive control IRF4 mRNA was confirmed to be MM-specific in this screen.

Figure 3.

Identification of cancer-common and MM-selective te-lncRNA isoforms. (A) Identification of strong hits in 11 cancer cell lines expressing Cas13d. Strong hits identified te-lncRNA isoforms as cancer common (red: strong hits in >8 cancer cell lines) and MM selective (blue: at least in 2 MMCLs and none of the non-MMCLs). The black bars indicate the te-lncRNA isoforms not included in these 3 categories. (B) Differential expression of cancer-common te-lncRNA isoforms in MM vs non-MM CCLs. (C) Differential expression of MM-selective te-lncRNA isoforms in MM vs non-MM CCLs. (D) Heat map showing the expression of te-lncRNA isoforms in the MM and non-MM CCLs. Red and orange dots indicate strong hits (red) and hits (orange) in the screen. (E) CCK-8 proliferation assay (gray bars) and qRT-PCR analysis (red rectangles) of the reported Cas13d-CCLs infected with a gRNA targeting SNHG6-003. Analysis was performed 3 days (KMS11) or 6 days (other CCLs) after infection. Data are reported as the percentage of an NT gRNA. ∗P < .05 by Student t test. (F) Correlation between the dependency score and the expression levels of SNHG6-003 across the panel of CCLs. ∗P < .05 by Student t test. ns, not significant by Student t test; NT, nontargeting.

We assessed the expression of te-lncRNA isoforms in the expanded panel of cancer cell lines by RNA-seq. The cancer-common te-lncRNA isoforms exhibited heterogeneous expression, ranging from undetectable levels in some cell lines to very high expression (maximum TPM = 21 115.71). There was no significant difference in expression between MM and non-MM cancer cell lines (CCLs; P = .59; Figure 3B). In contrast, the MM-selective te-lncRNA isoforms were significantly more expressed in MM than in non-MM CCLs (P = .028; Figure 3C). Although expression alone could not fully explain the observed dependencies, we found consistency between the expression levels and the essentiality of MM-selective te-lncRNA isoforms (Figure 3D). Some te-lncRNA isoforms, such as AP003900.6-001, RP11-305L7.1-001, and LINC01021-001 and -006, were not detectable in most non-MM CCLs. Notably, AP003900.6-001, LINC01021-001, and -006 were hits in all MMCLs except KMS11, which does not express these lncRNAs (Figure 3D). Interestingly, LINC01021, also known as P53 upregulated regulator of P53 levels,²⁷ is not detected in KMS11, because it is a TP53-null cell line. TPTEP1-012, which was a hit in all MMCLs but not in non-MM CCLs, had its preferential expression in MMCLs confirmed by qRT-PCR (supplemental Figure 3A).

Using an isoform-selective gRNA, we validated the role of the cancer-common lncRNA isoform SNHG6-003 as essential in KMS11, SU-DHL-4, A549, and HCT-116 cell lines, in addition to H929 (Figure 2B), and as dispensable in KARPAS-422 and MCF7 cell lines (Figure 3E). In all cell lines, the expression of the gRNA significantly knocked down SNHG6-003 (Figure 3E). Interestingly, SNHG6-003 is expressed by all CCLs at relatively high levels, ranging from 42.5 TPM in R8226 (not dependent) to 505.5 TPM in H929 (dependent), and a significant correlation is observed between its expression and dependency score across the CCLs (r = 0.63; P = .002; Figure 3F). This suggests that tumor cells may become dependent on certain lncRNA isoforms such as SNHG6-003 only when expressed at high levels. Additionally, we validated the MM-selective essentiality of TPTEP1-012 using 2 gapmeR antisense oligonucleotides (ASOs) in 3 MM (AMO1, R8226, KMS11) and 2 non-MM CCLs (A549, HCT116) (supplemental Figure 3B).

These data highlight that lncRNA isoforms, like protein-coding genes,²⁸ maintain tumor-essential roles across different cancer contexts and/or exhibit selective essentiality in MM.

Mapping the subcellular site of essentiality of te-lncRNA isoforms

The function of lncRNAs is linked to their subcellular localization.^4,29 Thus, we used subRNA-seq to measure the relative expression of te-lncRNA isoforms in the chromatin-bound, nuclear-soluble, and cytoplasmic fractions of AMO1, H929, and KMS11 MMCLs (Figure 4A). As expected, most te-lncRNA isoforms had the highest expression in chromatin-bound fractions (157 in AMO1, 167 in H929, and 100 in KMS11) or nuclear-soluble fractions (76 in AMO1, 46 in H929, and 119 in KMS11; Figure 4A-B). These included isoforms of MALAT1, MIR17HG, and NEAT1, which have nuclear functions in MM and other cancers. Fewer te-lncRNA isoforms (35 in total: 18 in AMO1, 28 in H929, and 19 in KMS11) displayed prevalent expression in the cytosol (Figure 4A-B), with 12 of these isoforms common across the MMCLs tested.

Figure 4.

Mapping the subcellular site of essentiality of te-lncRNA isoforms. (A) Subcellular localization for te-lncRNA isoforms in AMO1, H929, and KMS11 cells based on relative abundances from RNA-seq fractionation data. (B) Representation of CB-, NS-, and CY-enriched te-lncRNA isoforms in AMO1, H929, and KMS11 cells. (C) Results of viability screens in AMO1, H929, and KMS11 expressing either NES or NLS and infected with IsoScan-L2. Significant cytosolic hits are colored in green. (D) Overlap of CY-essential te-lncRNA isoforms in the 3 MMCLS, with 6 “core” CY-essential te-lncRNA isoforms highlighted. (E) Depletion of the gRNA pools targeting SNHG6-003 in viability screens using AMO1, H929, and KMS11 expressing either NES or NLS and infected with IsoScan-L2. An asterisk (∗) refers to a FDR <0.3 in the screen after MAGeCK analysis. (F) Relative expression of SNHG6-003 in the CB, NS, and CY fractions in AMO1, H929, and KMS11. (G) Expression of te-lncRNA isoforms in the cytosolic fractions of AMO1, H929, and KMS11 cells. SNHG6-003 is highlighted in green as the te-lncRNA isoform with the highest expression in the cytosolic fractions. (H) CCK-8 proliferation assay (white bars) and qRT-PCR analysis (green rectangles) of AMO1, H929, and KMS11 transfected with a siRNA targeting SNHG6-003. Analysis was performed 2 days after infection. Data are reported as the percentage of an si-NC. ∗P < .05 by Student t test. CB, chromatin-bound; CY, cytoplasmic; NES, Cas13d^NES; NLS, Cas13d^NLS; NS, nuclear-soluble; si-NC, nontargeting small interfering RNA; siRNA, small interfering RNA.

To explore the essentiality of te-lncRNA isoforms in the cytosol, we engineered H929, KMS11, and AMO1 MMCLs to express CRISPR-Cas13d with a nuclear export signal for cytosolic accumulation (Cas13d^NES), in contrast to the standard Cas13d with a nuclear localization signal (Cas13d^NLS) used in previous screens (supplemental Figure 4A). After infection with IsoScan-L2, we conducted viability screens and found 33 te-lncRNA isoforms, with 6 of them common to the 3 MMCLs (Figure 4C-D). The depletion scores (logFC depletion) of the gRNA pools targeting these te-lncRNA isoforms were generally weaker than those produced in the viability screens with Cas13d^NLS (Figure 4C), suggesting that these isoforms may play only a partial role in the cytosol or that Cas13d^NES is less effective than Cas13d^NLS, as previously reported.³⁰ Nevertheless, the positive control IRF4 mRNA was a significant hit in KMS11^Cas13d-NES and H929^Cas13d-NES cells. In contrast, the nuclear-retained isoforms of MIR17HG and NEAT1 were not a hit in any cell lines expressing Cas13d^NES.

Interestingly, SNHG6-003 was identified as a hit in all 3 MMCLs expressing Cas13d^NES (Figure 4E). It was expressed predominantly in the cytosolic fraction (Figure 4F), in which it exhibited the highest expression among all te-lncRNA isoforms (Figure 4G), as assessed by subRNA-seq. Using small interfering RNAs, which use the cytosolic RNAi machinery, we confirmed a cytosolic dependency on SNHG6-003 in the same MMCLs (Figure 4H).

These data indicate cell nuclei as the primary site of te-lncRNA isoform expression and activity but highlight a subset of te-lncRNA isoforms, particularly SNHG6-003, that potentially play important roles in the cytosol as well.

SNHG6-003 is an ER-enriched te-lncRNA isoform affecting the unfolded protein response pathway

Among the identified te-lncRNA isoforms, many have unknown functions, offering opportunities for further study. Here, we focus on SNHG6-003, the predominant SNHG6 isoform with tumor-essential functions in MM and other cancers, which is notably localized in the cytosol, in which it contributes to growth dependency.

Using single-molecule RNA fluorescence in situ hybridization, we confirmed that SNHG6-003 is enriched in the cytosol, with indications of localization to the ER (supplemental Figure 5A). To refine this observation, we combined single-molecule RNA fluorescence in situ hybridization with immunofluorescence and codetected SNHG6-003 with lysine-aspartic acid-glutamic acid-leucine (KDEL) sequence, an ER marker, revealing that ∼30% of SNHG6-003 localizes to the ER (Figure 5A; supplemental Figure 5B).

Figure 5.

SNHG6-003 is an ER-enriched te-lncRNA isoform affecting the unfolded protein response pathway. (A) Representative images of SNHG6 and KDEL in KMS-11 cells that were stained with SNHG6 fluorescence in situ hybridization probes (red), KDEL (green), and nucleus with DAPI (blue). In total, 40 296 SNHG6 signals and 13 190 KDEL signals were analyzed. The percentage of colocalization of SNHG6 with KDEL (∼33%) was calculated using ARIVIS Pro software (Zeiss). Number of cells analyzed: at least 5 per slide; ×63 magnification was used, with a digital zoom of ×2. On the right, we show a compartment analysis focusing on the ER (see “Methods” for more details). Full image is shown in supplemental Figure 5B. (B) H929 fractionation in cytosol, ER, and MITO fractions followed by qRT-PCR analysis of SNHG6-003. (C) APEX-qPCR assay in H929 cells. APEX2 localized in the cytoplasm, ERM, or MITO-M. qRT-PCR analysis assessed relative expression of SNHG6-003. Results were normalized to cells expressing the same APEX construct cultured in biotin-phenol but not treated with H₂O₂. (D) ChIRP-MS analysis of SNHG6-003 and MIR17HG-002 (for comparison) in KMS11. HSP interactors of SNHG6-003 are highlighted in red. Known interactors of MIR17HG are highlighted in blue. (E) STRING analysis of the interactors of SNHG6-003 from ChIRP-MS, focusing on GOBP. Top 5 GOBP are shown, ranked based on the FDR. (F) qRT-PCR analysis of SNHG6-003 in RNA immunoprecipitation material precipitated using anti-HSP90 α- and β-antibodies, or IgG as a control. (G) Gene set enrichment analysis of transcriptional changes produced by SNHG6-003 knockdown in KMS11^Cas13d and H929^Cas13d cells 3 days after gRNA infection. (H) Modulation of ER stress–related genes produced by SNHG6-003 knockdown in H929^Cas13d cells 3 days after gRNA infection. (I) WB analysis of CHOP/DDIT3, ATF4, and ATF3 in H929^Cas13d cells infected with gRNAs targeting SNHG6-003 or SNHG6-005, or NT-gRNAs as control. Cells were collected 3 days after infection. ACTB was used as a protein loading control. (J) Coimmunoprecipitation/MS analysis of the HSP90AB1 protein interacting in the presence (KD) or absence (WT) of SNHG6-003 depletion in H929^Cas13d cells. Interactors are divided into unchanged, lost, or gained after KD. CHOP, C/EBP-homologous protein 10; DAPI, 4′,6-diamidino-2-phenylindole; ERM, endoplasmic reticulum membrane; GOBP, gene ontology biological processes; IgG, immunoglobulin G; IP, immunoprecipitation; KD, knockdown; KDEL, lysine (K)-aspartic acid (D)-glutamic acid (E)-leucine (L); MITO, mitochondria; MITO-M, mitochondrial membrane; MS, mass spectrometry; NES, normalized enrichment score; NT, nontargeting; R, response; STRING, searches the relations in networks of genes; UPR, unfolded protein response; WT, wild type.

Subcellular fractionation via sucrose gradients and ultracentrifugation³¹ further validated its enrichment in the ER of H929 cells, with glyceraldehyde-3-phosphate dehydrogenase and mitochondrially encoded cytochrome c oxidase subunit 2 serving as controls for cytoplasmic and mitochondrial fractions, respectively (Figure 5B; supplemental Figure 5C). Additionally, APEX2-based proximity labeling³² confirmed preferential localization of SNHG6-003 to the ER membrane (Figure 5C; supplemental Figure 5D).

Comprehensive identification of RNA-binding proteins by mass spectrometry (ChIRP-MS)^33,34 analysis in live MM cells identified 94 potential interactors for SNHG6-003 (logFC >2; FDR <0.1) vs 103 for the predominantly nuclear MIR17HG-002 (Figure 5D; supplemental Figure 5E). STRING (searches the relations in networks of genes) analysis revealed that SNHG6-003 interactors were enriched in ER-related processes (eg, protein folding, stabilization, and chaperone-mediated complex assembly) and prominently featured different heat shock proteins (HSP; Figure 5E). The interactions of SNHG6-003 with HSP complexes purified with anti-HSP90α or anti-HSP90β antibodies was validated by RNA immunoprecipitation qPCR (Figure 5F). In contrast, the interactors of MIR17HG-002 included relevant positive controls such as WDR82, proteins involved in the bioprocessing of primary microRNA transcripts such as DGCR8, and established partners such as GEMIN5, FUBP1, and U2AF2 (Figure 5D), confirming the reliability of our ChIRP-MS.

Knockdown of SNHG6-003 with isoform-specific gRNAs in KMS11^Cas13d and H929^Cas13d cells led to significant activation of the unfolded protein response and ER stress pathways, coupled with suppression of protein synthesis (Figure 5G). In H929^Cas13d cells, proapoptotic ER stress genes (eg, ATF3, ATF4, and DDIT3/CHOP) were upregulated whereas ribosomal protein genes were downregulated (Figure 5H). These transcriptomic changes were supported by increased ATF4, DDIT3/CHOP, and ATF3 protein levels after SNHG6-003 knockdown but not SNHG6-005 knockdown (Figure 5I). The same effect was observed when targeting SNHG6-003 directly in the cytosol, with increased protein levels of CHOP, ATF3, and ATF4 (supplemental Figure 5F).

Mechanistically, we hypothesized that SNHG6-003 modulates HSP activity through protein-protein interactions. Consistent with this, coimmunoprecipitation/MS of HSP90β revealed that SNHG6-003 depletion altered its interaction network, losing some interactions and gaining others, notably with negative regulators of HSP90 function such as UNC45A³ and oncogenic factors such as TRIM21³⁵ (Figure 5J; supplemental Table 4).

Collectively, these data identify SNHG6-003 as an ER-resident lncRNA isoform that plays a crucial role in maintaining protein homeostasis and mediating the response to proteotoxic stress.

SNHG6-003 is a potential therapeutic target in MM

Clinically, SNHG6-003 was the only isoform of SNHG6 differentially expressed in patients with NDMM compared with healthy donor (HD) normal plasma cells (Figure 6A). Although the expression of SNHG6 was unchanged in premalignant precursor conditions (supplemental Figure 6A), it is significantly higher in relapsed/refractory compared with NDMM using both unpaired (Figure 6B) and paired sample (supplemental Figure 6B) analysis. Remarkably, SNHG6-003 exhibited significantly higher expression in patients with high-risk MM carrying del17p³⁶ (Figure 6C).

Figure 6.

SNHG6-003 is a potential therapeutic target in MM. (A) RNA-seq analysis of SNHG6-003, SNHG6-005, and SNHG6-006 in CD138⁺ cells from patients with NDMM (n = 319) and HDs (n = 16). (B) RNA-seq analysis of SNHG6 in CD138⁺ cells from patients with MM at the diagnosis (NDMM, n = 49) or relapse (RRMM, n = 69). The unpaired analysis is shown in the main figure, whereas the paired analysis is shown in supplemental Figure 6B. (C) RNA-seq analysis of SNHG6-003 in CD138⁺ cells from patients with NDMM enrolled in the IFM-DFCI studies. (C, left) SNHG6-003 is analyzed in the IFM2009 patient dataset with (n = 28) or without (n = 291) deletion del17p. (C, right) SNHG6-003 is analyzed in another cohort of high-risk patients (n = 44) or without (n = 405) deletion del17p. (D) Flow cytometry analysis of CD138⁺ cells in bone marrow mononuclear cells from 3 patients with NDMM and 1 pt with RRMM cultured ex vivo for 6 days in the presence (+) or absence (–) of SNHG6-ASO (10 μM). NPC, normal plasma cell; pt, patient; RRMM, relapsed/refractory MM.

To explore the tumor-essential role of SNHG6-003 in primary tumor cells, we used gapmeR ASOs with validated knockdown activity in MMCLs (supplemental Figure 6C) to treat bone marrow mononuclear cells from 4 patients with MM at different disease stages ex vivo. This treatment resulted in a significant reduction in the percentage of CD138⁺ tumor plasma cells (Figure 6D). Importantly, these effects were achieved in the presence of the protective bone marrow milieu. Moreover, analysis of live/dead cells revealed no significant difference in the CD138^– populations from the same patient samples (supplemental Figure 6D).

Overall, these findings underscore the usefulness of our platform in discovering relevant te-lncRNAs and highlight SNHG6-003 as a promising target for cancer therapy.

Expression and clinical relevance of te-lncRNA isoforms in patients with MM

Finally, to provide a comprehensive picture of the te-lncRNA isoform landscape in MM, we analyzed their expression in CD138⁺ plasma cells from 319 patients with NDMM and 16 HDs, previously profiled with RNA-seq. Patients with NDMM were enrolled in the Intergroupe Francophone du Myélome (IFM)-DFCI2009 clinical trial (ClinicalTrials.gov identifier: NCT01191060) and uniformly treated and clinically annotated. Among the 315 hits identified in at least 2 MMCLs in the primary screens, 85 te-lncRNA isoforms were significantly upregulated in NDMM compared with HD plasma cells, 17 te-lncRNA isoforms were downregulated, and 213 showed no differential expression (Figure 7; supplemental Figure 7).

Figure 7.

Expression and clinical relevance of te-lncRNA isoforms in patients with MM. Circos plot showing the results of clinical and expression analyses for te-lncRNA isoforms in patients with MM, with a focus on the 154 te-lncRNAs common to at least 4 cell lines (the others are reported in supplemental Figure 6). Circle 1 defines the expression of isoforms between patients with MM vs normal plasma cells (purple, upregulated in MM; green, downregulated in MM; gray, no difference). Circles 2 through 5 show the expression of isoforms between each subgroup vs others (pink, upregulated in subgroups; blue, downregulated in subgroups; gray, no difference) and circle 6 summarizes te-lncRNA isoforms whose higher expression is associated with shorter OS. chr, chromosome; HMM, hyperdiploid MM; OS, overall survival.

Interestingly, te-lncRNA isoforms also showed preferential upregulation or downregulation in specific genomic subgroups (Figure 7; supplemental Figure 7). For example, we observed lower expression of many of te-lncRNA isoforms in patients carrying the t(4;14) translocation compared with others; 40 te-lncRNA isoforms, including 7 originating from chromosome 1q (ASH1L-AS1-001, RP11-315I20.1-003, RP11-94I2.4-001, RUSC1-AS1-001, GAS5-014, GAS5-015, and HNRNPU-AS1-001), showed selective upregulation in patients carrying 1q gain; and 4 te-lncRNA isoforms (including SNHG6-003, LINC-PINT-007, NUTM2A-AS1-009, and SNHG6-005) showed selective upregulation in patients carrying del17p.

Finally, we identified 92 te-lncRNA isoforms, almost 1 of 3 examined, with higher expression correlating with shorter overall survival (Figure 7; supplemental Figure 7).

The expression analysis of te-lncRNAs was also performed using an independent data set from the CoMMpass study, which included 557 patients with NDMM. In this data set, we confirmed that higher expression of most te-lncRNAs (207 of 315) was associated with poor clinical outcomes. Furthermore, we observed that 233 isoforms were less expressed in patients with the t(4;14) translocation than in other patients, with only 19 isoforms more expressed in these patients. We also found that 6 isoforms were more expressed in patients with del17p, and 2 were less expressed, whereas 111 isoforms showed increased expression in patients with 1q gain, with 20 showing decreased expression (supplemental Figure 8).

These findings offer a comprehensive understanding of the expression patterns and clinical impact of te-lncRNA isoforms in patients with MM, potentially guiding the prioritization of te-lncRNA isoforms for functional exploration and therapeutic targeting.

Discussion

Using the unique RNA-targeting activity of CRISPR-Cas13d,^16,17 this study investigated the growth-promoting potential of lncRNAs at the isoform level in cancer, with a particular focus on MM. This innovative approach addresses previously overlooked aspects of MM and cancer biology, providing valuable insights into the intricate landscape of lncRNA-mediated dependencies in cancer.

Traditionally, lncRNAs have been challenging to study because of their complex genomic organization, which frequently overlaps with other genes.1, 2, 3 lncRNA genes often produce multiple transcript isoforms with distinct sequences and structures, each potentially contributing to tumorigenesis.1, 2, 3 Additionally, lncRNA genes can function as DNA regulatory elements, acting independently of their transcripts.1, 2, 3 By leveraging Cas13d, which directly targets RNA transcripts, we overcame the limitations of DNA-targeting approaches such as CRISPR-Cas9 and CRISPRi, which often fail to distinguish lncRNAs functioning as regulatory elements or transcripts. These methods also struggle to assess the activity of lncRNA genes that overlap with other genetic elements or contain embedded regulatory sequences such as microRNAs or small nucleolar RNAs.^19,37 In contrast, Cas13d enables precise targeting of lncRNA transcripts regardless of their genomic context, allowing us to uncover hundreds of te-lncRNAs. This aligns with findings from a recent study by Liang et al,¹⁷ published during the preparation of this article. Notably, earlier CRISPR-Cas9 and CRISPRi screens identified only a handful of te-lncRNAs in MM.^12,14

Our focus on isoform-specific targeting offers significant advantages over previous studies. Unlike Montero et al,³⁸ who targeted multiple isoforms of the same lncRNA to streamline library size, we deliberately prioritized the investigation of individual isoforms. This refined approach revealed distinct cell type– and cancer-specific expression patterns and uncovered functional differences between isoforms of the same lncRNA, as exemplified by SNHG6-003 vs SNHG6-005 and -006. These findings underscore the critical role of isoform specificity in understanding lncRNA function and pave the way for identifying unique therapeutic vulnerabilities. Moreover, our results suggest potential therapeutic opportunities by indicating that specific RNA structural domains unique to individual isoforms may be targeted. Strategies such as using structure-targeting ligands,³⁹ ribonuclease-targeting chimeras,⁴⁰ or blockmeR ASOs¹⁴ could be used to inhibit lncRNA function or induce their degradation. In summary, our approach deepens the mechanistic understanding of lncRNA biology while opening new avenues for precise RNA-targeted therapies.

Our findings underscore the subcellular localization of lncRNAs as a key determinant of their functional roles and therapeutic potential. Although many te-lncRNAs likely exert their primary functions in the nucleus, we also identified isoforms with significant cytoplasmic activity, highlighting the need for tailored therapeutic strategies. Nuclear-enriched lncRNAs can be targeted with gapmeR ASOs, whereas cytosolic lncRNAs, such as SNHG6-003, may also be addressed with small interfering RNAs or other cytoplasm-specific approaches.

Notably, SNHG6-003 localizes to the ER in MM cells, in which it interacts with HSPs to modulate proteostasis and activate the unfolded protein response–critical pathways in MM due to the high immunoglobulin production characteristic of plasma cells.⁴¹ However, it remains unclear whether the ER stress observed upon targeting SNHG6-003 arises from the depletion of nuclear-, cytosolic-, or ER-localized transcripts.

The clinical significance of te-lncRNAs is further underscored by their association with aggressive disease and poor survival in patients with MM, as shown in 2 independent patient data sets.

Finally, we developed the LongDEP Portal (LongDEP Portal), an open resource for functional and clinical annotation of te-lncRNAs. This platform provides comprehensive data on expression, clinical outcomes, subcellular localization, and screening results, facilitating further research into lncRNA-driven oncogenesis.

In conclusion, this study unveils a previously overlooked aspect of MM biology, revealing a robust dependency of MM cells on lncRNA isoforms. These findings challenge existing paradigms and underscore the potential of lncRNAs as therapeutic targets in MM and other cancers.

Conflict-of-interest disclosure: N.C.M. serves on advisory boards for/as consultant to Takeda, Bristol Myers Squibb, Celgene, Janssen, Amgen, AbbVie, OncoPep, Karyopharm, Adaptive Biotechnology, and Novartis and holds equity ownership in OncoPep. K.C.A. serves on advisory boards for Janssen, Pfizer, AstraZeneca, Amgen, Precision Biosciences, Mana, Starton, and Raqia and is a scientific founder of OncoPep and C4 Therapeutics. R.A.Y. is a founder and shareholder of Syros Pharmaceuticals, Camp4 Therapeutics, Omega Therapeutics, Dewpoint Therapeutics, and Paratus Sciences and serves in consulting or advisory roles at Precede Biosciences and Novo Nordisk. E.M. and N.C.M filed a patent on MIR17HG as a target for cancer therapy and a provisional patent on SNHG6 as a target for cancer therapy. The remaining authors declare no competing financial interests.