Investigation of lncRNA expression in newly diagnosed multiple myeloma reveals a LINC01432-CELF2 axis as an inhibitor of apoptosis

Abstract

Multiple myeloma (MM) is an incurable malignancy of plasma cells, with over 35,000 new cases diagnosed annually in the United States. Despite an expanding arsenal of approved therapies, nearly all patients relapse, and mechanisms underlying disease progression remain poorly understood. In particular, the role of long non-coding RNAs (lncRNAs) in MM progression and treatment response is largely unexplored. To address this gap, we performed transcriptome sequencing of newly diagnosed MM (NDMM) patient samples and compared individuals with short progression-free survival (PFS; <24 months) to those with prolonged PFS (>24 months) following standard first-line therapy. We identified 157 lncRNAs upregulated in patients with short PFS, and prioritized the most significantly upregulated transcript, LINC01432, for functional characterization. CRISPR-mediated knockdown of LINC01432 expression results in upregulation of genes associated with interferon-α/γ responses and increases apoptosis. Targeting LINC01432 with locked nucleic acid antisense oligonucleotides also induces apoptosis, which can be rescued by LINC01432 overexpression. Mechanistically, we discovered that LINC01432 binds the RNA-binding protein CELF2 directly. Transcriptomic analysis following depletion of either LINC01432 or CELF2 revealed 108 overlapping target genes, indicating that this lncRNA–protein complex regulates transcriptional programs governing immune activation, stress response, and cell survival. In summary, this study identified lncRNAs associated with NDMM and characterized LINC01432 as a critical regulator of MM cell survival, acting in complex with CELF2 to repress pro-apoptotic and immune response pathways. These findings highlight LINC01432 as a potential therapeutic target for overcoming resistance in MM.

Introduction

Multiple myeloma (MM) is the fifteenth leading cause of cancer-related deaths in the United States [1]. While prolonged event-free survival rates have improved due to advances in the standard treatment approach for stem cell transplant-eligible MM patients, nearly all patients become refractory and die from the disease or its sequelae [2–4]. Standard treatment consists of a three-dose induction therapy, including a proteasome inhibitor, an immunomodulatory drug, and a steroid, followed by treatment with high-dose melphalan and autologous stem cell transplantation. In addition, while some improvements in patient outcomes have been achieved using novel immunomodulatory agents, new treatment approaches are needed due to high toxicity and the development of drug resistance [5–7]. Thus, understanding the mechanisms and biomarkers of treatment resistance in MM patients is critical to the development of novel MM therapies.

Long non-coding RNA (lncRNA) is defined as RNA greater than 200 nucleotides in length that is not translated into functional proteins. Prior studies report that lncRNAs can promote the pathogenesis of all cancer types, including MM [8–15]. Many lncRNAs have been shown to promote MM drug resistance, including NEAT1, ANRIL, MEG3, LINC00461, H19, and PCAT1 [12, 16–20]. Subcellular localization is highly important to the biological function of lncRNAs, which may include transcriptional regulation, translational regulation, and interaction with RNA binding proteins [21, 22]. Further, recent advances in understanding the roles lncRNAs play in promoting cancer, including MM, increases their potential as targets for RNA-based therapeutics [10, 23–25].

In this study, we generated and analyzed RNA sequencing (RNA-seq) data from a cohort of newly diagnosed MM (NDMM) patients to identify lncRNAs that were associated with a short progression-free survival (PFS). We identified several lncRNAs that were highly upregulated in patients with short PFS, as compared to prolonged PFS, and determined that LINC01432 binds to the RNA-binding protein, CELF2, to repress immune activation and apoptosis-associated gene expression.

Methods

RNA sequencing data, patient samples, and cell lines

RNA-seq data from NDMM patient samples (n = 135) were obtained from the Multiple Myeloma Research Foundation (MMRF) Clinical Outcomes in Multiple Myeloma to Personal Assessment of Genetic Profiles (CoMMpass) study (https://registry.opendata.aws/mmrf-commpass), accessed in February of 2021 (Supplementary Tables 1 and 2). MM cell lines were generously provided by Dr. John DiPersio at Washington University School of Medicine (WashU, RPMI 8226, U266B1, MM.1S, and OPM2) and cultured in RPMI 1640 media (Invitrogen, Carlsbad, CA, USA) supplemented with 15% fetal bovine serum (Invitrogen) and 1% penicillin/streptomycin (Invitrogen). MM.1R cell lines were purchased from ATCC (catalog number CRL-2975) and authenticated. All cell lines were tested for mycoplasma contamination.

Full length LINC01432 transcript was amplified via PCR and cloned into the pCFG5-IEGZ-GFP vector (generously provided by Dr. Christopher Maher, WashU) to create the pCFG5-IEGZ-GFP-LINC01432 vector (pCFG5-LINC01432), as previously described [10]. Full vector length was confirmed by GeneScript.

LINC01432 knockdown CRISPR/Cas9 cells were generated using the RPMI 8226 cell line. The sgRNAs were generated by the Genome Engineering and Stem Cell Center (WashU) and cloned into the pLV hUbC-dCas9 KRAB-T2A-GFP plasmid (Addgene, Watertown, MA, USA, catalog number 672620). A non-targeting sgRNA was used as the control. Sequences are listed in Supplementary Table 4. HEK 293T cells were infected with this lentivirus to induce expression of dCas9-KRAB [26], followed by transduction into RPMI 8226 cells and validation of knockdown of LINC01432 expression via RT-qPCR.

Transfection of locked nucleic acid antisense oligonucleotides

Locked nucleic acid GapmeR antisense oligonucleotides (LNA ASOs) targeting LINC01432 (LINC01432 LNA ASO 1 or LINC01432 LNA ASO 2), CELF2 (CELF2 LNA ASO 1 or CELF2 LNA ASO 2), and negative control LNA ASOs, were designed using the Qiagen Antisense LNA GapmeR Custom Builder (https://www.qiagen.com), sequences and catalog numbers are listed in Supplementary Table 4. MM cells were seeded at a density of 500,000 cells/well in 6-well plates, transfected with respective ASOs at 50 nM and 100 nM concentration using Lipofectamine 2000, and incubated for 48–72 h. Cells were harvested and target knockdown was validated via RT-qPCR.

RNA sequencing data analyses

RNA-seq data from the MMRF CoMMpass study were processed and analyzed to investigate transcriptomic alterations. Reads were aligned to the GRCh37 (hs37d5) [27] reference genome using Ensembl version 74, with additional transcript annotations provided in the accompanying GTF file (https://github.com/tgen/MMRF_CoMMpass.git). Quality control of RNA BAM files was performed using Picard RNA metrics and BamTools IgCounts to ensure data integrity. Differential expression (DE) analysis was performed with edgeR, and genes with a p-value < 0.05 and an absolute log₂FC ≥ 2 were significantly differentially expressed. Gene Set Enrichment Analysis (GSEA) was conducted to identify enriched biological pathways. Gene Ontology (GO) and pathway enrichment analyses were performed using DAVID [28], while POSTAR3 [29] was used to explore predicted LINC01432 RNA:protein interactions.

RNA-seq analysis was conducted on RPMI 8226 cell lines transfected with 50 nM of either LINC01432 (n = 2), CELF2 (n = 2), or Control LNA ASOs (n = 2). Total RNA integrity was determined using Agilent Bioanalyzer or 4200 Tapestation. Ribosomal RNA was removed using RiboErase kits (Kapa Biosystems, Wilmington, MA, USA). Total RNA was fragmented and reverse transcribed using SuperScript III (Life Technologies, Carlsbad, CA, USA) and random hexamers. After second-strand synthesis, ds-cDNA was end-repaired, A-tailed, and ligated to Illumina adapters. Libraries were PCR-amplified with dual-index primers and sequenced on an Illumina NovaSeq X Plus with 150 bp paired-end reads. Basecalling and demultiplexing were performed using bcl2fastq (1 mismatch allowed). Reads were aligned to the Ensembl GRCh38.113 genome using STAR (v2.7.11b), and gene-level counts were obtained with featureCounts (v2.0.8). Transcript isoform quantification was performed with Salmon (v1.10.0). Sequencing quality was assessed using RSeQC (v5.04).

Multiplexed fluorescent RNA in situ hybridization (mFISH)

mFISH was performed using RNAscope 2.5 HD Reagent Kit Red assay combined with Immunohistochemistry (Advanced Cell Diagnostics [ACD], Newark, CA USA, catalog numbers 323180 and 322372), as previously described [30]. Bone marrow aspirates or tumor tissues were fixed to slides, deparaffinized, treated with hydrogen peroxide, and underwent target retrieval. Slides were incubated overnight at 4 °C with CUGBP2 (CELF2) antibody (Proteintech, Rosemont, IL, USA catalog number 12921-1-AP). LINC01432 probes (ACD catalog number 878271) were then hybridized, followed by RNA amplification and Fast Red staining. Alexa Fluor 488 secondary antibody (Abcam, Cambridge, United Kingdom, catalog number ab150081) was applied for one hour at room temperature in the dark. Slides were counterstained with DAPI (Sigma, St. Louis, MO, USA catalog number D9542), mounted with ProLong Gold Antifade (Invitrogen, catalog number P36930), and imaged using EVOS M5000 Imaging System (Invitrogen). Spot counts per cell and cellular distribution were quantified using QuPath v0.5.1, with multiplex and compartmental analysis to assess lncRNA localization across tissues or cell lines.

In vitro apoptosis assays and rescue assays

Apoptosis was measured by flow cytometry using BD Horizon V450 AnnexinV (BD Biosciences, Franklin Lakes, NJ, USA). 500,000 MM cells were seeded in a 6-well plate for 24 h, harvested, stained per manufacturer’s instructions, and assessed on a flow cytometer machine (Novios, Becton Dickinson, Franklin Lakes, NJ, USA) at the Flow Cytometry Core of Siteman Cancer Center, Washington University in St. Louis. A minimum of 50,000 cells per sample were measured in triplicate using FlowJo Version 10 (Becton Dickinson).

Apoptosis was also measured using the ApoTox-Glo Triplex Assay (Promega, Madison, WI, USA) per manufacturer’s instructions. Control or LINC01432-modified RPMI 8226 cells were seeded at 20,000 cells/well in triplicate in 96-well plates in 100 µls complete media per well. To measure apoptosis, 100 µl Caspase-Glo 3/7 reagent was added, mixed by orbital shaking for ~20 s, then incubated at room temperature for 30 min. Luminescence (relative luminescence units) was measured using the Varioskan LUX microplate reader.

Apoptosis was further evaluated using 5 μM CellEvent™ Caspase-3/7 Red Detection Reagent (Thermo Fisher, St. Louis, MO, USA, catalog number C10430), with a 30-minute incubation at 37 °C. FAM-labeled LNA ASOs and caspase activation were imaged using an EVOS M5000 (excitation/emission: ∼502/530 nm for FAM, ∼590/610 nm for CellEvent™) and quantified using ImageJ.

Loss-of-function and rescue experiments were performed using FAM-labeled LNA ASOs and lncRNA overexpression (OE) plasmids. MM cell lines were seeded at 300,000 cells/well in 6-well plates and transfected with 50 nM LINC01432-targeting LNA ASO1 or ASO2 or a non-targeting control using Lipofectamine 2000, then incubated for 72 h. For rescue assays, cells were transfected with 2 μg of LINC01432 OE plasmid or empty vector and incubated for an additional 48–72 h.

Comprehensive identification of RNA-binding proteins by mass spectrometry (ChIRP-ms) by western blot

ChIRP-ms was performed using a pool of 20-mer biotinylated DNA probes targeting LINC01432 with western blotting validation using a previously published protocol [31]. RPMI 8226 cells or OPM2 cells were grown to confluency (~20 million cells), fixed with formaldehyde for 30 min, then quenched with 0.125 M glycine for 5 min. Cell pellets (~80–100 mg) were lysed and sonicated in ChIRP lysis buffer. Lysates were hybridized overnight with biotinylated probes (Supplementary Table 4), then incubated with streptavidin C1 magnetic beads (Invitrogen, catalog number 65001). For RNA recovery, complexes were eluted with Proteinase K buffer, extracted with Trizol LS, purified using RNeasy Mini columns (Qiagen, Venlo, Netherlands, catalog number 74104), and analyzed by RT-qPCR.

For protein analysis, bound proteins were eluted with biotin elution buffer, then precipitated overnight with 25% trichloroacetic acid (TCA, catalog number 8693-16). Pellets were washed with cold acetone, air-dried, and resuspended in 1× Laemmli buffer (Thermo Fisher Scientific, catalog number J60015.AC). Alternatively, 20% of bead-bound proteins were directly boiled with buffer. Cross-links were reversed at 95 °C for 30 min, and samples were analyzed by SDS-PAGE and western blot. Interactions with LINC01432 were assessed using CELF2 antibody and anti-GAPDH antibody (Cell Signaling, Danvers, MA, USA, catalog number 97166S). All experiments were performed in triplicate and independently repeated at least twice.

RNA immunoprecipitation (RIP) and RT-qPCR

RIP was performed as previously described [10, 32]. Nuclear lysates from 10 million cells were prepared using the NER-PER Extraction Kit (Thermo Fisher Scientific) and incubated overnight with 5 µg of CELF2 antibody or IgG control in RIPA wash buffer supplemented with SUPERase In RNase inhibitor (Thermo Fisher Scientific). Dynabeads™ Protein G (50 µL, Invitrogen) were added and rotated for 1–2 h at 4 °C. Beads were washed six times with RIPA buffer, and complexes were digested in Proteinase K buffer at 55 °C for 30 min. RNA was extracted using acid phenol:chloroform:isoamyl alcohol and treated with the Heat&Run gDNA Removal Kit (ArcticZymes, Tromso, Norway). cDNA synthesis was performed with SuperScript™ III, and qPCR was run using Fast SYBR™ Green Master Mix and gene-specific primers (Supplementary Table 4). Fold enrichment was calculated relative to IgG and normalized to 1% input using the Sigma-Aldrich Data Analysis Shell. A >2-fold enrichment over IgG was considered significant.

In vivo individual-nucleotide resolution cross-linking immunoprecipitation (iCLIP)

iCLIP was performed as previously described [10]. Cells were washed with cold PBS and UV crosslinked using a Stratalinker. Pellets were lysed in NP-40 buffer, incubated on ice for 10 min, and centrifuged. Supernatants were treated with 1 U/µl RNase T1 at 22 °C for 30 min. Protein G beads were pre-incubated with 5 µg antibody (Supplementary Table 5) in NT2 buffer for 1 h at room temperature. Cell lysates were added to beads and rotated for 3 h. Beads were washed with NT2 buffer and treated with 20 U DNase I at 37 °C for 15 min. Proteinase K buffer was added, and RNA was extracted using phenol:chloroform:isoamyl alcohol. RNA was reverse transcribed using SuperScript III (Thermo Fisher Scientific) and primers tiling LINC01432 (Supplementary Table 4) to detect RNA–protein interactions.

In vivo myeloma models

For subcutaneous injections, 200,000–1,000,000 respective cells were subcutaneously injected into male NOD/SCID/γc^−/− (NSG) mice (N = 5–10 per group). Resulting tumor size was quantified weekly via caliper measurements, comparing length × width × height × 0.5. For post-analyses, subcutaneous tumor tissues were removed after sacrifice, formalin fixed, and paraffin embedded. This experiment was repeated twice.

Statistics

Statistical analyses were performed using methods appropriate to the data structure and experimental design. Tests were selected based on assumptions of normality, variance homogeneity, and sample size, and are justified as suitable for the comparisons presented. All experiments were performed in triplicate and independently repeated at least twice. Data are presented as mean ± SEM, and statistical significance was assessed using unpaired two-tailed Student’s t test.

For mice studies, parental cells are estimated to metastasize at a probability greater than 95%. This allows 80% power to detect changes in tumor growth probability. Investigators were not blinded to group allocation during the experiments or outcome assessment. In addition, animals were not randomized to experimental groups. Group allocation was based on predefined experimental design and logistical considerations, including injection time and availability of age-matched cohorts.

Results

Identification of dysregulated lncRNAs in patients with short progression-free survival in response to standard MM therapy

We analyzed RNA-seq data from 115 NDMM patients in the MMRF CoMMpass study to identify differentially expressed lncRNAs associated with short PFS in response to standard MM treatment, Fig. 1a. Samples were assigned to one of two groups based on (1) short PFS, <24 months PFS from the first dose of MM treatment (N = 38), and (2) prolonged PFS, >24 months PFS (N = 77), Supplementary Tables 2 and 3. We identified 157 upregulated and 91 downregulated lncRNAs in short PFS, as compared to prolonged PFS (log₂FC ≥ ±2, p < 0.05), Fig. 1b. The most differentially expressed lncRNAs in short PFS included LINC01432, lnc-LGALS9B-7, LINC01916, Lnc-SPIDR-1, and MAGEA4-AS1, Fig. 1c. We also identified two lncRNAs previously reported to be associated with MM, MEG3 [20, 33], and H19 [34, 35], Supplementary Table 1. Next, to contextualize the potential regulatory roles of dysregulated lncRNAs in short PFS, we performed pathway analysis on all differentially expressed RNAs and identified global transcriptional programs and biological pathways associated with PFS. This revealed enrichment of gene sets associated with staphylococcus aureus infection (p = 6.75e⁻¹⁰), transcriptional dysregulation in cancer (p = 1.05e⁻⁰⁶), cytokine-cytokine receptor interactions (p = 1.42e⁻⁰⁵), IL-17 signaling pathway (p = 2.31e⁻⁰⁵), and ECM-receptor interactions (p = 5.5e⁻⁰⁵), Supplementary Fig. 1a. GO analysis showed enrichment of immune-related pathways, including B cell–mediated immunity and hemoglobin complexes, Supplementary Fig. 1b, c.

Fig. 1. Identification of dysregulated lncRNAs associated with short progression-free survival in MM therapy and genes and pathways targeted by LINC01432.

a Pipeline of assigning eligible newly diagnosed multiple myeloma patient samples to either short progression-free survival (PFS) or prolonged PFS groups. b Schematic of the pipeline used to identify lncRNAs associated with short PFS in response to standard MM therapy. c Identification of significantly differentially expressed lncRNAs in short PFS, as compared to prolonged PFS, highlighting LINC01432. d LINC01432 expression levels in NDMM patient samples, showing higher expression in short PFS as compared to prolonged PFS. e Bulk RNA sequencing data showing differentially expressed genes (DEGs) following LNA ASO-mediated LINC01432 knockdown, as compared to Control LNA ASOs. f Expression of multiple myeloma associated genes following LNA ASO-mediated LINC01432 knockdown, as compared to Control LNA ASOs. g Gene set enrichment analysis (GSEA) showing Hallmarks of Cancer pathways. h GSEA plots for Interferon alpha and gamma response, apoptosis, and p53 pathways. i Expression of interferon alpha associated genes following LNA ASO-mediated LINC01432 knockdown, as compared to Control LNA ASOs. Gene Ontology (GO) bar plots showing activation of pathways related to j molecular function, k biological process, and l cellular components following LNA ASO-mediated LINC01432 knockdown, as compared to Control LNA ASOs. *p value < 0.05, **p value < 0.005, ***p value < 0.0005, #p value < 0.00005.

LINC01432 is upregulated in short PFS and regulates immune response and myeloma-associated pathways

We focused our subsequent analyses on LINC01432, as it showed the highest fold change and lowest p-value (fold change = 6.42, p = 6.11e⁻⁴³; Fig. 1c, d) of all dysregulated lncRNAs in patients with short PFS. Despite its clinical relevance, little is known about LINC01432, with only one genome-wide association study linking it to a SNP associated with male baldness [36]. Given the genomic heterogeneity and chromosomal abnormalities common in MM, we assessed LINC01432 expression across MM genetic subtypes. High levels of LINC01432 expression correlated with the presence of t(14;16) (r = 0.57) and Amp(1q) (r = 0.14) abnormalities, Supplementary Fig. 2a.

To further characterize LINC01432, we analyzed its expression in MM cell lines and found that it is highly expressed in RPMI 8226 and OPM2 cells, with lower expression in MM.1R, MM.1S, and U266B1 cells, Supplementary Fig. 2b. As our initial patient study focused on identifying lncRNAs associated with standard MM treatment response, we tested the effects of each component of VRd therapy on LINC01432 expression. Melphalan treatment significantly increased LINC01432 expression in RPMI 8226 cells, while Bortezomib and Dexamethasone treatment both induced modest upregulation, Supplementary Fig. 2c.

To investigate LINC01432-mediated transcriptional regulation, we conducted RNA-seq analysis following knockdown of LINC01432 expression using two different LINC01432-targeting LNA ASOs, or a non-targeting control LNA ASO in MM cells and identified 314 significantly upregulated and 90 significantly downregulated genes (adjusted p < 0.05, |log₂FC|≥ 2), Fig. 1e. Interestingly, 42 of these differentially expressed genes (DEGs) have been implicated in MM progression or regulation, including genes such as BDNF (log₂FC = −4.96, p = 0.0001), CCR7 (log₂FC = 2.83, p = 0.0015), COMP (log₂FC = 3.04, p = 0.006), PDF (log₂FC = 2.7, p = 1.34e⁻²⁶), GDF15 (log₂FC = 3.06, p = 0.007), and CXCL8 (log₂FC = 2.71, p = 0.0006), Fig. 1f and Supplementary Table 6.

GSEA following LINC01432 knockdown revealed upregulated genes were associated with pathways related to interferon-α/γ responses, TNF-α signaling via NF-κB, apoptosis, and p53 signaling, Fig. 1g, h. Downregulated genes were associated with pathways related to protein secretion, androgen response, and mitotic spindle pathways, Fig. 1g. For interferon-α response genes specifically, LINC01432 knockdown led to robust upregulation of chemokines CXCL10 (log₂FC = 3.54, p = 6.71e⁻¹³⁴) and CXCL9 (log₂FC = 2.41, p = 1.71e⁻⁵⁹), guanylate-binding proteins GBP4 (log₂FC = 2.64, p = 1.40e⁻⁶³), antiviral effectors RSAD2 (log₂FC = 2.21, p = 1.18e⁻²²) and HELZ2 (log₂FC = 1.79, p = 1.86e⁻⁵⁸), as well as interferon-induced proteins IFIT2 (log₂FC = 1.95, p = 2.23e⁻⁴⁹) and IFIT3 (log₂FC = 1.75, p = 1.21e⁻⁴⁰), Fig. 1i.

We also performed GO enrichment analysis of DEGs following LINC01432 knockdown. Enriched molecular functions included cytokine and chemokine receptor binding and signaling receptor activator activity (FDR < 0.05), Fig. 1j. Key biological processes included humoral immune response, antimicrobial activity, and cell migration, suggesting LINC01432 may impair plasma cell function, Fig. 1k. Enriched cellular components, such as cell projection membrane and plasma membrane, support its nuclear and cytoplasmic roles, Fig. 1l.

LINC01432 inhibits apoptosis and increases tumor growth

Next, we confirmed the expression of LINC01432 in NDMM bone marrow aspirates using mFISH, Fig. 2a and Supplementary Fig. 2d. We found high expression of LINC01432 in RPMI 8226 tumors and low expression in U266B1 tumors using mFISH, Fig. 2b, c. We determined that in RPMI 8226-derived tumors, LINC01432 was detected in the nucleus of 9.50% of cells, the cytoplasm of 0.59% of cells, and in both compartments of 86.91% of cells, with 2.99% of cells showing no apparent LINC01432 expression, Supplementary Fig. 2e. In the U266B1 cell line, with low endogenous LINC01432 expression, LINC01432 was detected in the nucleus of 32.79% of cells, the cytoplasm of 0.44% of cells, and in both compartments of 15.83% of cells, while no expression was detected in 50.94% of cells, Supplementary Fig. 2f. These data indicate that LINC01432 is a novel lncRNA that is highly expressed in NDMM patients with short PFS to standard treatment and in MM cell lines.

Fig. 2. LINC01432 inhibits apoptosis in myeloma cell lines and increases tumor growth.

a–c mFISH showing localization of LINC01432 expression in a newly diagnosed multiple myeloma (NDMM) patient bone marrow aspirates, b RPMI 8226 tumors, and c U266B1 tumors. Scale bar = 20 µM and 40 µM (Zoom). d CRISPR-mediated knockdown of LINC01432 expression in RPMI 8226 cells. LINC01432 knockdown cells have increased apoptosis as measured via e ApoTox-Glo assay (% luminescence estimation caspase 3/7 signal), and f Annexin V staining flow cytometry, as compared to controls. g Confirmation of LINC01432 overexpression (OE) in U266B1 cells, as compared to empty vector controls. LINC01432 OE cells have decreased apoptosis as measured via h ApoTox-Glo assay, and i Annexin V staining flow cytometry, as compared to empty vector controls. j Representative images of in vivo tumor growth at Day 42 following subcutaneous injection of LINC01432 knockdown cells into NGS mice, as compared to control CRISPR cell-induced tumors. k Quantification of LINC01432 knockdown cell-induced tumor volumes. Control CRISPR n = 5 and LINC01432 CRISPR n = 5. l Representative images of in vivo tumor growth at Day 35 following subcutaneous injection of LINC01432 overexpression cells into NGS mice, as compared to control empty vector cell-induced tumors. m Quantification of LINC01432 overexpression cell-induced tumor volumes. Empty Vector n = 8 and LINC01432 OE n = 8, *p value < 0.05, **p value < 0.005, ***p value < 0.0005, #p value < 0.00005.

To investigate how LINC01432 may induce a short PFS in response to standard MM therapy and to evaluate it as a therapeutic target, we used two CRISPR/Cas9 approaches (CRISPR1 and CRISPR2) to knockdown LINC01432 expression in RPMI 8226 cells, Fig. 2d. We found that LINC01432 knockdown significantly increased apoptosis in both CRISPR cell lines (CRISPR1, p = 2.69e⁻⁰⁵; CRISPR2, p = 2.78e⁻⁰⁷), as compared to control CRISPR cells, Fig. 2e, as measured via ApoTox-Glo assay. We validated this increase in apoptosis further via Annexin V flow cytometry (CRISPR1, p = 0.0009; CRISPR2, p = 1.98e⁻⁰⁵), Fig. 2f. We then transfected U266B1 cells with a LINC01432 OE vector, Fig. 2g, and found significantly decreased apoptosis (ApoTox-Glo, p = 0.04; AnnexinV, p = 0.04), as compared to empty vector control, Fig. 2h, i.

Next, we investigated the role of LINC01432 in MM tumor growth by subcutaneously injecting mice with either LINC01432 knockdown RPMI 8226 cells and LINC01432 overexpression U266B1 cells. We found that LINC01432 knockdown cells induced significantly lower tumor volumes, as compared to CRISPR control cells (Day 28, p = 0.04; Day 42, p = 0.02), Fig. 2j, k. Similarly, we showed that LINC01432 OE cells induced significantly higher tumor volumes, as compared to controls (Day 14, p = 0.009; Day 21, p = 0.04; Day 28, p = 0.003; Day 35, p = 0.0003), Fig. 2l, m.

LINC01432 promotes MM cell survival by suppressing apoptosis

To evaluate whether LINC01432 functionally contributes to cell survival, we conducted LINC01432 knockdown in RPMI 8226 and OPM2 MM cells using two different LINC01432-targeting LNA ASOs, or a non-targeting control LNA ASO, followed by LINC01432 OE constructs to assess phenotypic rescue, Fig. 3a–c. The LNA ASOs used in this study are designed to induce RNase H-mediated degradation of LINC01432, and thus, are expected to target both endogenous and exogenous transcripts [37–39]. In our overexpression-rescue experiments, overexpressing LINC01432 using a CMV plasmid overwhelms the capacity of the LNA ASOs to induce target degradation, in turn increasing LINC01432 expression, as shown in Fig. 3b, c.

Fig. 3. LNA ASO-mediated LINC01432 knockdown induces apoptosis.

a Schematic of locked nucleic acid (LNA) antisense oligonucleotide (ASO)-mediated knockdown of LINC01432 and rescue with LINC01432 OE in MM cells. Created with Biorender.com. RT-qPCR quantification of LINC01432 expression in b RPMI 8226 cells, and c OPM2 cells, following treatment with Control LNA ASO, LINC01432-targeting LNA ASO1 or ASO2, and rescue with LINC01432 OE. Representative images of images of d RPMI 8226 cells, and e OPM2 cells showing increased Caspase 3/7 activity following LINC01432 knockdown using CellEvent assay. Quantification of Caspase 3/7–positive cells following LINC01432 and overexpression rescue in f RPMI 8226 cells and g OPM2 cells using ImageJ. Quantification of Annexin V positive cells following LINC01432 and overexpression rescue in h RPMI 8226 cells and i OPM2 cells. *p value < 0.05, **p value < 0.005, ***p value < 0.0005, #p value < 0.00005.

LINC01432 knockdown significantly increased apoptosis as measured by Caspase-3/7 activity (RPMI 8226: LINC01432 LNA ASO1, p = 1.12e⁻⁰⁷; LINC01432 LNA ASO2, p = 1.96e⁻⁰⁵. OPM2: LINC01432 LNA ASO1, p = 6.06e⁻⁰⁹; LINC01432 LNA ASO2, p = 1.61e⁻⁰⁶) and ApoTox-Glo (RPMI 8226: LINC01432 LNA ASO1, p = 0.02; LINC01432 LNA ASO2, p = 0.004. OPM2: LINC01432 LNA ASO1, p = 0.033; LINC01432 LNA ASO2, p = 0.001), as compared to controls, Fig. 3d–i.

Importantly, LINC01432 OE rescued the apoptotic phenotype as measured by both Caspase-3/7 activity (RPMI 8226: LINC01432 LNA ASO1 + OE, p = 6.39e⁻⁰⁸; LINC01432 LNA ASO2 + OE, p = 2.58e⁻⁰⁵. OPM2: LINC01432 LNA ASO1 + OE, p = 5.37e⁻⁰⁹; LINC01432 LNA ASO2 + OE, p = 1.63e⁻⁰⁶) and ApoTox-Glo (RPMI 8226: LINC01432 LNA ASO1 + OE, p = 0.04; LINC01432 LNA ASO2 + OE, p = 0.001. OPM2: LINC01432 LNA ASO1 + OE, p = 0.05; LINC01432 LNA ASO2 + OE, p = 0.006), as compared to knockdown alone, Fig. 3d–i.

LINC01432 binds to the CELF2 RNA-binding protein

lncRNAs frequently function through interactions with RNA-binding proteins (RBPs), thereby regulating downstream gene expression and contributing to cancer progression and therapeutic resistance [10, 32]. We used the POSTAR3 database [29] to identify CUGBP Elav-like family member 2 (CELF2) as a putative LINC01432 RBP based on publicly available CLIP-seq data (binding score = 0.019; Supplementary Fig. 3 and Fig. 4a). CELF2 regulates RNA splicing, stability, and translation, and influences cell proliferation, migration, and tumor growth by binding to lncRNAs in other cancer types [40–45]. However, these functions have not been studied in the context of MM. Thus, we investigated whether LINC01432 binds to CELF2 and if this mechanism plays a role in LINC01432-mediated regulation of apoptosis in MM.

Fig. 4. LINC01432 binds to the RNA-binding protein CELF2.

a POSTAR3-predicted CELF2 binding sites on LINC01432 lncRNA. b Comprehensive identification of RNA-binding proteins by mass spectrometry (ChIRP-ms) followed by western blot analysis demonstrating enrichment of CELF2 protein in LINC01432 pulldown fraction, as compared to unprocessed input lysate, with GAPDH negative control in RPMI 8226 and OPM2 cells. c Schematic of CELF2-binding sites on LINC01432 with locations of site-specific primers used for iCLIP RT-qPCR. Created with Biorender.com. CELF2 iCLIP RT-qPCR showing enrichment of LINC01432 binding in d RPMI 8826 and e OPM2 cells, as compared to IgG and GAPDH negative controls. RNA immunoprecipitation (RIP) followed by RT-qPCR in nuclear and cytoplasmic compartments indicating LINC01432 enrichment in CELF2 immunoprecipitants in f RPMI 8826 and g OPM2 cells. h Multiplex fluorescent in situ hybridization (mFISH) for LINC01432 combined with CELF2 immunohistochemistry in bone marrow aspirates from newly diagnosed multiple myeloma (NDMM) patients. i Quantification of subcellular localization of LINC01432 and CELF2 in NDMM samples using QuPath analysis. Scale bar = 40 µM *fold enrichment >1.5, **fold enrichment >5, ***fold enrichment >10.

While analysis of NDMM patient RNA-seq data revealed high levels of CELF2 expression (logCPM >50), no significant differences in expression were identified between short PFS and prolonged PFS, Supplementary Fig. 4a. CELF2 is also highly expressed across hematologic malignancies, including leukemia, lymphoma, and MM (logCPM >10; Supplementary Fig. 4b), and the Human Protein Atlas shows that CELF2 is enriched in bone marrow (nTPM = 77), Supplementary Fig. 4c.

To validate the interaction between LINC01432 and CELF2 in MM cells, we performed ChIRP-ms by western blot analysis of CELF2 protein levels. We validated pulldown of LINC01432 (average fold enrichment = 100.7 for RPMI 8226, 86.79 for OPM2; Supplementary Fig. 5a, b), then demonstrated binding of CELF2 to LINC01432 using western blot, Fig. 4b. To map the regions on LINC01432 bound by CELF2, we designed tiling primers spanning the entire LINC01432 transcript, Fig. 4c, then conducted iCLIP RT-qPCR analysis in MM cells. This showed high-fold enrichment of CELF2 binding in regions 12–31 (fold enrichment = 1.79 for RPMI 8226, 17.08 for OPM2), 19,332–19,394 (fold enrichment = 8.72 for RPMI 8226, 1.67 for OPM2), and 20,489–20,509 (fold enrichment = 7.81 for RPMI 8226, 1.58 for OPM2), as compared to IgG negative control and GAPDH, Fig. 4d, e.

Given the dual nuclear and cytoplasmic localization of LINC01432 in MM, Fig. 2a–c, we next performed RIP RT-qPCR with CELF2 antibodies in RPMI 8226 and OPM2 cells and found significant enrichment of LINC01432 in both the nuclear (fold enrichment = 17.2 for RPMI 8226, 9.22 for OPM2) and cytoplasmic fractions (fold enrichment = 5.96 for RPMI 8226, 5.36 for OPM2), as compared to IgG and GAPDH controls, Fig. 4f, g.

Finally, to visualize the spatial co-localization of LINC01432 and CELF2, we performed mFISH using LINC01432-specific probes combined with CELF2 immunohistochemistry in NDMM patient bone marrow aspirates. CELF2 protein was detected in both the nuclear and cytoplasmic compartments and found to co-localize with LINC01432, with 40.01% of cells showing co-localization in both nucleus and cytoplasm, 25.30% in nucleus only, 16.27% in cytoplasm only, and 18.41% lacking detectable expression Fig. 4h, i. mFISH confirmed high expression of CELF2 and co-localization with LINC01432 In RPMI 8226 cell-induced tumors, Supplementary Fig. 6a, b.

LINC01432 and CELF2 co-regulate immune-related and apoptotic gene networks

To determine the extent of shared regulatory targets between LINC01432 and CELF2, we performed parallel RNA-seq analysis following CELF2 knockdown in RPMI 8226 cells using LNA ASOs, with control LNA ASOs. We identified 2165 DEGs following CELF2 knockdown (1506 upregulated and 659 downregulated; adjusted p < 0.05, fold change >1.5), Supplementary Fig. 7a and Supplementary Table 7. GSEA of upregulated genes showed they were enriched for gene sets associated with MYC targets, G2M checkpoint, and DNA repair, Supplementary Fig. 7b. In contrast to our LINC01432 knockdown findings, downregulated genes following CELF2 knockdown were enriched for gene sets associated with interferon-α/γ response and TNFα signaling via NFκB, Supplementary Fig. 7b and Fig. 1g. GO analysis revealed significant enrichment of pathways related to biological processes associated with cytokine and chemokine receptor binding and signaling receptor activator activity, similar to LINC01432 regulation (FDR < 0.05), Supplementary Fig. 7c and Fig. 1k. Lastly, cellular processes were enriched for actin-based cell projection and external side of plasma membrane and cell projection membrane, Supplementary Fig. 7c.

To investigate shared regulatory targets further, we compared DEGs following LNA ASO-mediated knockdown of LINC01432 and CELF2, Fig. 5a. Overlap analysis revealed a set of 108 genes with significantly altered expression following knockdown of LINC01432 and CELF2 (adjusted p < 0.05, fold change >1.5), suggesting a coordinated regulatory relationship, Fig. 5b and Supplementary Table 8. Interestingly, this shared gene set included 41.7% lncRNAs, 41.7% protein-coding genes, 6.5% processed pseudogenes, and smaller proportions of miscRNAs, rRNAs, TEC, and snRNAs, Fig. 5c, d. As CELF2 is an RBP, this enrichment of lncRNAs suggests that CELF2 directly interacts with and modulates a distinct set of lncRNAs that includes LINC01432, Fig. 5e. Among the shared targets were several genes associated with interferon signaling or immune modulation that have been previously implicated in MM progression, including CCR7 (LINC01432 log fold change = 2.83, p = 0.002; CELF2 log fold change = 2.44, p = 0.022), BDNF (LINC01432 log fold change = −4.96, p = 0.0001; CELF2 log fold change = −2.89, p = 0.025), CXCL9 (LINC01432 log fold change = 2.41, p = 1.71e⁻⁵⁹; CELF2 log fold change = −3.10, p = 3.51e⁻²²), and GDF15 (LINC01432 log fold change = 3.06, p = 0.007; CELF2 log fold change = 4.19, p = 1.10e⁻⁰⁵), Fig. 5f. We also assessed the effects of CELF2 knockdown using two independent LNA ASOs and found that it increased apoptosis, Supplementary Fig. 8a–c. Combined knockdown of LINC01432 and CELF2 further enhanced apoptosis, supporting a pro-survival role for this pathway, Supplementary Fig. 8c.

Fig. 5. LINC01432 and CELF2 co-regulate shared gene networks in multiple myeloma.

a Heatmap of RNA sequencing data illustrating expression profiles of overlapping genes across Control LNA ASOs, LINC01432-targeted LNA ASOs, and CELF2-targeted LNA ASOs. b Venn diagram depicting significantly overlapping differentially expressed genes (adjusted p < 0.05, |log₂FC|> 1.5). c Distribution of gene biotypes among the shared targets, including long non-coding RNAs (lncRNAs), protein-coding genes, and other RNA classes. d Top overlapping significantly dysregulated genes following CELF2 and LINC01432 knockdown. e Overlapping lncRNAs with most highly altered expression following CELF2 and LINC01432 knockdown. f Shared genes with established roles in multiple myeloma pathogenesis. *log fold change >2, **log fold change >3, ***log fold change >4, #log fold change >5.

In summary, these results reveal that LINC01432 and CELF2 co-regulate a shared network of immunomodulatory and apoptotic genes, including many lncRNAs. Their cooperative suppression of interferon-stimulated genes and apoptosis-related pathways suggests that LINC01432 and CELF2 functionally interact to promote MM cell survival, Fig. 6.

Fig. 6. Schematic of LINC01432-CELF2 axis in myeloma regulation.

We identified 248 deregulated long non-coding RNAs (lncRNAs) in newly diagnosed multiple myeloma patient samples with short progression-free survival (PFS) in response to standard MM therapy, as compared to those with prolonged PFS. Among these, LINC01432 was the most significantly upregulated and was found to interact with the RNA-binding protein CELF2. Elevated expression of LINC01432, or activation of the LINC01432–CELF2 regulatory axis, modulates gene networks involved in chemokine and cytokine signaling as well as interferon-stimulated pathways, collectively promoting cell survival and inhibiting apoptosis. Created with Biorender.com.

Discussion

Recent advances in cellular immunotherapies for MM are hindered by interpatient heterogeneity [46–48], complicating the understanding of molecular mechanisms controlling disease progression. Identifying mechanisms and biomarkers of treatment resistance in MM patients is crucial for developing new treatments.

In this study, we identified differentially expressed lncRNAs in NDMM patients with short PFS on standard therapy. Despite limitations, such as only including transplant-eligible patients and limited non-White participants, we identified LINC01432 as the most significantly upregulated lncRNA in patients with short PFS. Previously uncharacterized, LINC01432 was notably overexpressed in high-risk MM patients, making it a candidate biomarker for poor prognosis and therapeutic resistance.

We showed that LINC01432 represses a broad array of genes involved in immune activation, interferon signaling, and apoptotic pathways, while promoting expression of known myeloma-promoting factors [49–51]. Some of these interferon-α response-related genes are involved in transcriptional regulation of lncRNAs while modulating lncRNA-protein interactions, promoting cell death and cytokine-induced apoptosis [31, 52–54]. Notably, LINC01432 knockdown increased apoptosis. These findings suggest that LINC01432 may contribute to immune evasion and resistance by repressing interferon-driven immune signaling in MM cells.

We showed that LINC01432 binds to CELF2, which is known to bind lncRNAs and regulate downstream mRNAs in multiple forms of cancer [42, 45, 55, 56], but has not yet been studied in the context of NDMM. In cancer, CELF2 is localized to the nucleus, where it is associated with alternative splicing and transcript editing, and the cytoplasm, where it regulates pre-miRNA maturation, translation, and alternative polyadenylation [57–59]. We found that CELF2 shows different patterns of localization in MM cell lines with differential levels of endogenous LINC01432 expression. Further studies are needed to determine if this is due to the specific lncRNA:protein interaction or is solely dependent on the increased expression of LINC01432.

LINC01432 and CELF2 exert coordinated regulation over immune response genes, with specific overlap in key MM-related genes. Notably, we observed divergent gene expression changes in key pathways associated with inflammation, interferon signaling, and cytoskeletal dynamics following LINC01432 and CELF2 knockdown. This suggests that while LINC01432 and CELF2 regulate similar targets, they may do so through distinct, and sometimes antagonistic mechanisms. In addition to immune signaling, shared regulatory targets such as BDNF, LCN2, and NUPR1 are implicated in cell survival and apoptotic resistance [60–65], suggesting that both LINC01432 and CELF2 promote MM cell fitness. Strikingly, LINC01432 and CELF2 knockdown shared 108 DEGs, nearly half of which were lncRNAs, suggesting CELF2 may function as a lncRNA-binding protein in MM, consistent with its reported roles in solid tumors [41, 58]. In future studies, we plan to investigate the effects of LINC01432 and CELF expression on cell viability and apoptosis using clonogenic and co-culture assays with primary bone marrow cells to better capture the complex cellular interactions within the bone marrow microenvironment.

The tissue and cell-specific expression patterns of lncRNAs makes them ideal targets for the development of RNA therapeutics [66, 67]. ASOs are powerful tools for therapeutically targeting lncRNAs [68, 69]. This study represents a preliminary investigation into the use of LNA ASOs to downregulate LINC01432 lncRNA. Efforts to mechanistically characterize and evaluate the regulatory roles of lncRNAs, as performed for LINC01432 in this study, go beyond expression-based associations by functionally validating their biological relevance. As demonstrated in recent CRISPR-based studies [70–73], such function-guided approaches are critical for uncovering lncRNAs with causal roles in cancer biology, thereby strengthening their translational potential as novel therapeutic targets.

In conclusion, our study provides preliminary insights into the role of lncRNA expression in NDMM patients who exhibit a short PFS in response to standard MM therapy and identifies LINC01432 and CELF2 as novel potential targets for the development of future MM therapies. Together, these data provide a framework for future studies to dissect the molecular mechanisms underpinning lncRNA-protein interactions and highlight LINC01432–CELF2 as a potential axis of therapeutic vulnerability in high-risk multiple myeloma.

Author contributions

RM wrote and edited the manuscript, conducted in vitro assays, mouse experiments, mFISH, and analyzed data. PT edited the manuscript, developed cell lines, and conducted in vitro assays. DD edited manuscript, analyzed and processed sequencing data, and mFISH images. AP conducted in vitro assays, and western blots. KB analyzed and processed RNA sequencing data. SG conducted and analyzed in vitro assays and mFISH. CS, VB, JK, SD, YR, and MA conducted and analyzed in vitro assays. YC and KW analyzed mFISH. CE and SR conducted PubMed searches of sequencing data. RJ, MF, JF provided guidance on analysis of RNA sequencing data cohorts. MS, LD, JD and RV provided project guidance. RV and JD provided samples. JS-F conducted in vitro assays, and mouse experiments, designed and directed experimental studies, wrote and edited the manuscript, which all authors reviewed and approved.

Data availability

All RNA-seq data is available at GEO under accession numbers GSE267013, GSE300566, and GSE300088. All other data are available within this article and its Supplementary Information files, and from the corresponding author, upon reasonable request.

Competing interests

MS (Consulting for DSMB and endpoint adjudication work: Sorrento, Marker Therapeutics, GSK, Kura Oncology, and NovoNordisk; Research support: Incyte, Janssen, Takeda, Ichnos Biosciences, Fate Therapeutics, and Karyopharm), RV (Consulting: BMS, Sanofi, Janssen, Karyopharm, Pfizer, Regenron; Research: BMS, Sanofi, Takeda), and JD (Consulting: Rivervest, Bluebird Bio, Vertex, HcBiosciences, SPARC; Equity-ownership WUGEN and Magenta; Research support: Macrogenics, Bioline, Incyte).

Ethics approval and consent to participate

All methods were performed in accordance with relevant guidelines and regulations. This study was conducted under the supervision of the Washington University Human Rights Protections Office and in compliance with all relevant guidelines and regulations for clinical research. All participants in CoMMpass (NCT01454297) provided informed consent for that study. Only de-identified patient-level data was accessed for these analyses and, therefore, were determined to be exempt. NDMM patient bone marrow aspirates were obtained from the participants following informed consent and registration on our internal Multiple Myeloma Tissue Banking Protocol (protocol number 201102270). All animal experiment protocols in this study were reviewed and approved by the Institutional Animal Care and Use Committee of WashU (Protocol 24-0155).

Supplementary information

The online version contains supplementary material available at 10.1038/s41389-025-00579-w.