Article Text
PDF
Abstract
Background Aberrant Wnt pathway activation is a key driver of colorectal cancer (CRC) and is essential to sustain tumour growth and progression. Although the downstream protein-coding target genes of the Wnt cascade are well known, the long non-coding transcriptome has not yet been fully resolved.
Objective In this study, we aim to comprehensively reveal the Wnt-regulated long non-coding transcriptome and exploit essential molecules as novel therapeutic targets.
Design We used global run-on sequencing to define β-catenin-regulated long non-coding RNAs (lncRNAs) in CRC. CRISPRi dropout screens were subsequently used to establish the functional relevance of a subset of these lncRNAs for long-term expansion of CRC.
Results We uncovered that LINC02418 is essential for cancer cell clonogenic outgrowth. Mechanistically, LINC02418 regulates MYC expression levels to promote CRC stem cell functionality and prevent terminal differentiation. Furthermore, we developed effective small interfering RNA (siRNA)-based therapeutics to target LINC02418 RNA in vivo.
Conclusion We propose that cancer-specific Wnt-regulated lncRNAs provide novel therapeutic opportunities to interfere with the Wnt pathway, which has so far defied effective pharmacological inhibition.
- colorectal cancer
Data availability statement
Data are available in a public, open access repository. All data relevant to the study are included in the article or uploaded as supplementary information. The sequence libraries generated in this study are publicly available through the National Center for Biotechnology Information (NCBI) Gene Expression Omnibus (GEO) under accession code: GSE206582.
This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.
Statistics from Altmetric.com
WHAT IS ALREADY KNOWN ON THIS TOPIC
Hyperactivation of the Wnt pathway is common in patients with colorectal cancer (CRC), and plays a central role in promoting cancer stem cell renewal and cell proliferation.
Therapeutic inhibition of the Wnt pathway has proven to be challenging due to its influence on normal stem renewal and tissue homeostasis; however, long non-coding RNAs recently emerged as important tissue-specific regulators of gene expression, which could be used as novel therapeutic targets.
WHAT THIS STUDY ADDS
Using global run-on sequencing, we shed light on the Wnt-regulated non-coding transcriptome across a large panel of CRC cell lines, representing all consensus molecular subtypes.
CRISPRi dropout screens uncovered a subset of long non-coding RNAs that is essential to CRC cells.
Transcription of LINC02418 is reactivated in CRC to maintain MYC expression levels.
LINC02418-targeted therapy holds the potential to halt tumour growth by interfering with the Wnt signalling through MYC in a cancer-specific manner.
HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY
This work highlights the importance of the non-coding transcriptome in supporting oncogenic pathways, and reinforces the need to globally characterise non-coding RNA expression patterns and functions.
We bring forward a novel cancer-specific RNA interference-based therapy, which could improve the clinical outcome of patients with CRC.
Introduction
Colorectal cancer (CRC) results from the stepwise accumulation of genetic aberrations.1–3 The majority of CRCs harbour APC mutations, leading to hyperactivation of the Wnt pathway, which in turn promotes dedifferentiation, proliferation and resistance to cell death.4–6 Protein-coding Wnt target genes have been extensively studied and some of them (eg, MYC) proved to be key downstream mediators of the pathway in oncogenesis.7–10 Given the importance of the Wnt signalling in CRC as well as in other cancer types, great efforts have been made to develop inhibitors, such as monoclonal antibodies (mAbs) and small molecule inhibitors (SMIs).11 However, downstream activation of Wnt/β-catenin signalling pathway in CRC hinders the therapeutic benefits of blocking extracellular or membrane-bound targets with monoclonal antibodies. Furthermore, while SMIs or peptides targeting intracellular proteins or protein-protein interactions (eg, β-catenin/T-cell factor and lymphoid enhancer-binding factor (TCF-LEF)) represent a possible approach, clinical translation of effective molecules remains an important challenge due to the central influence of Wnt signalling on normal stem cell dynamics and tissue homeostasis.12 Therefore, the development of novel therapeutic strategies to target the Wnt pathway in a cancer-specific manner is particularly relevant.
Recent advances in high-throughput sequencing technologies have greatly enhanced our understanding of the genome. In particular, RNA sequencing has been instrumental in revealing the diverse tissue- or cell-specific transcriptional profiles of the human genome. Interestingly, long non-coding RNAs (lncRNAs) seem to display high tissue and/or cell specificity as compared with protein coding RNAs.13 In addition, lncRNAs have been reported to play key regulatory functions in a wide range of cellular processes, and to be deregulated in various diseases, including cancer.14 So far, several lncRNAs have been characterised for their significant functions in CRC.15 However, a thorough assessment of essential lncRNA molecules in CRC remains elusive. Here, we sought to comprehensively define the Wnt-regulated long non-coding transcriptome by profiling a panel of CRC cell lines following β-catenin knockdown. Furthermore, we assessed the therapeutic potential of identified molecules by performing CRISPRi screens, and initiated the development of a novel RNA interference (RNAi)-based therapy targeting the cancer-specific LINC02418.
Results
Define Wnt transcriptional programmes in CRC
To comprehensively characterise Wnt transcriptional programmes in CRC, we engineered a panel of 15 cell lines to silence β-catenin expression using a doxycycline-inducible shRNA (figure 1A and online supplemental figure S1A). Effective β-catenin knockdown and protein depletion was obtained after 4 days in all cell lines (figure 1B and online supplemental figure S1B,C). Consistent with the essential role of the Wnt pathway in CRC, loss of β-catenin led to reduced proliferation and clonogenicity in 11 out of 15 lines (figure 1C and online supplemental figure S2A–D).
Supplemental material
Supplemental material
β-Catenin transcriptional programmes in CRC. (A) Workflow overview. A panel of CRC cell lines was engineered with a doxycycline (dox)-inducible short hairpin RNA (shRNA) against β-catenin. β-Catenin-depleted cell lines were profiled using global run-on sequencing (GRO-seq), and regulated lncRNAs were identified. (B) β-Catenin expression at the RNA (top, n=3) and protein (bottom) levels in control (shCtrl) and β-catenin (shCTN) knockdowns. SW1463 and MDST8 cells were treated for 4 days with 1 µg/mL dox. RNA levels are depicted as fold change (FC) expression levels and normalised to each dox-treated shCtrl sample. Protein quantifications (ratio β-catenin/GAPDH) were normalised to the corresponding minus dox (−) reaction, which was arbitrarily set to 1. (C) Colony formation assays (top) and limiting dilution assays (bottom) to evaluate the impact of control and β-catenin knockdowns on cell proliferation. (D) Scheme depicting the workflow used to process the GRO-seq data and identify Wnt-driven lncRNAs. (E) Fraction of detected coding and long non-coding transcripts, including de novo identified enhancer RNAs (eRNAs). (F) Negative enrichment score (ES) of downregulated protein-coding transcripts previously defined as Wnt target genes22 in β-catenin knockdown (shCTN) versus control (shCtrl), using Gene Set Enrichment Analysis (GSEA). (G) Heatmap representing the signature of β-catenin-regulated protein-coding transcripts (top 50 downregulated). mRNAs with the highest concordance across profiled cell lines (minimum 20% reduction following β-catenin knockdown in at least 13 cell lines) were selected. Intersection with the dnTCF-dependent Wnt signature is included (right side). P values (two-tailed Student’s t-test, n=3): *p<0.05, ***p<0.001.
Next, we profiled our panel of engineered CRC cell lines using global run-on sequencing. In contrast to conventional RNA sequencing, this technique allows the detection of nascent, lowly abundant and non-polyadenylated transcripts.16 17 The GRO-seq data were then processed to quantify the expression of annotated coding genes and lncRNAs, and to perform de novo detection of enhancer RNAs (eRNAs) (figure 1D). For most cell lines, the number of detected lncRNAs was approximately 25%–40% of the number of identified protein-coding RNAs (mRNAs) (figure 1E and online supplemental table S1). Gene Set Enrichment Analysis (GSEA) confirmed that previously established Wnt mRNA signatures were significantly downregulated in β-catenin knockdown samples (figure 1F,G). Comparison of the most strongly regulated lncRNAs (top 50 per cell line) revealed the high context dependence of the β-catenin transcriptional programme, consistent with previously described expression patterns of non-coding transcripts13 (figure 2A–C and online supplemental table S2). However, despite a clear cell line-specific impact of β-catenin knockdown, we identified a set of 40 Wnt-driven lncRNAs showing consistent downregulation following β-catenin loss across cell lines (figure 2D and online supplemental table S3). Importantly, this core set of Wnt-regulated lncRNAs was enriched in CRC patient samples as compared with normal colon tissue, suggesting their relevance in the disease (figure 2E).
Supplemental material
Supplemental material
Supplemental material
Identification of a β-catenin-driven lncRNA signature with enriched expression in CRC. (A) Heatmap showing the top 50 β-catenin downregulated lncRNAs in each cell line. (B,C) Examples of (B) shared and (C) cell-specific β-catenin regulation. Control (Ctrl) and β-catenin knockdown (Ctn) tracks are displayed for all profiled cell lines. Positive (+, red) and negative (−, blue) strands are indicated. (D) Heatmap representing the signature of β-catenin-regulated lncRNAs across the panel of CRC cell lines. lncRNAs (top 50) with the highest concordance are depicted. (E) Enrichment plot of the Wnt/lncRNA signature in CRC versus normal tissue (The Cancer Genome Atlas (TCGA) data). ES, enrichment score. (F) Volcano plot depicting expression of selected lncRNA candidates in CRC versus normal tissue (TCGA data). (G) Distance between β-catenin-regulated lncRNA loci and CRC single nucleotide polymorphisms (SNPs) or control SNPs shown in base pairs (bp). abs, absolute; padj, adjusted p value (Benjamini-Hochberg); ***p<0.001.
Global functional assessment of Wnt-regulated lncRNAs
Given the extensive cell line heterogeneity, we further analysed lncRNAs that exhibited modest to high level of downregulation (fold change (FC)>1.8) in at least two cell lines following β-catenin knockdown (online supplemental figure S3 and online supplemental tables S4 and S5). These lncRNA candidates also showed enrichment in CRC tumours compared to healthy tissue (figure 2F). Additionally, we detected that CRC-associated single nucleotide polymorphisms (SNPs) were significantly enriched in regions surrounding these candidate lncRNAs as compared with control SNPs (figure 2G). This further supports the notion that selected Wnt-regulated lncRNAs might be functionally involved in colorectal carcinogenesis.
Supplemental material
Supplemental material
Supplemental material
To assess the functional relevance of the Wnt-driven lncRNAs in vitro, we performed CRISPRi-based dropout screens in six CRC cell lines (online supplemental figure S4A,B). The custom pooled guide RNA (gRNA) library encompassed 745 non-coding loci and 55 positive control regions (essential genes) as well as >2000 negative (scrambled) control gRNAs (figure 3A and online supplemental table S6). Using the MAGeCK robust rank aggregation approach,18 we analysed the dropouts for each cell line and identified on average >50 critical regions where at least two guides reached a two fold downregulation with a false discovery rate (FDR) <0.05 (online supplemental figure S4B). To focus on functional regions common to all CRC cell lines, we analysed dropouts across the screened cell lines using the MAGeCK maximum likelihood estimation method.18 This analysis identified 92 significant dropout loci with an FDR <0.05, which represent common vulnerabilities for CRC cell lines (figure 3B,C, online supplemental figure S5 and online supplemental table S7).
Supplemental material
Supplemental material
Supplemental material
Supplemental material
CRISPRi dropout screen reveals functional Wnt-regulated lncRNAs in CRC cells. (A) Schematic illustrating the CRISPRi screening workflow. The pooled guide RNA (gRNA) library was introduced in CRC cell lines expressing a dox-inducible dCas9-KRAB construct. After selection with puromycin, untreated and dox-treated (0.5–1 µg/mL) cells were harvested at days 0, 8 and 16 for analysis. (B) Heatmap representing the CRISPRi hits (92 loci) obtained across six CRC cell lines. Depleted gRNAs were identified by MAGeCK maximum likelihood estimation (MLE), visualising the difference in log2 fold change (dLFC) at d8 or d16 along with the corresponding false discovery rate (FDR). The bottom panels show representative negative and positive control gRNAs. (C) Detailed view of the top two dropout regions (#1:PPP1R35-AS1 or ENSG00000240211.1_9 and #2:KPNB1-DT or ENSG00000263766.6_10) for the various cell lines, conditions (−/+ dox) and time points (d0, d8 and d16). Hits are represented with the average read counts of the most depleted two gRNAs across the three replicates and in all tested cell lines (colour coded). (D) Limiting dilution assays to validate two gRNAs targeting SUZ12P1 and NBR1-AS in COLO320-HSR and (E) CO100 CRC cell lines. (F) Limiting dilution assays following the suppression of SUZ12P1 and NBR1-AS with shRNAs in LS411N CRC line. Control (Ctrl) and targeting gRNAs/shRNAs were activated with dox for 14 days. P values (two-tailed Student’s t-test, n=3): *p<0.05; **p<0.01, ***p<0.001.
Phenotypic validation of essential lncRNAs
As a proof of concept and to confirm the quality of our screens, we validated the dependency of CRC cells on three dropout hits: SUZ12P1, RP11-242D8.1 (hereafter referred to as NBR1-AS) and LINC00173. While SUZ12P1 and NBR1-AS transcripts have not been previously described in the context of CRC, LINC00173 was reported to contribute to CRC survival and progression, and was therefore included as a positive control.19 To confirm the screen results, we first tested the best two gRNAs/transcript for their ability to mediate both locus-specific transcriptional silencing and to impact cell proliferation (figure 3D and online supplemental figure S6A–C). Next, we extended the validation to two additional cell lines not included in the screen (CO100 and SW48) and found that the tested gRNAs effectively reproduced their phenotypic effects (figure 3E and online supplemental figure S6D–H). Similar results were also obtained by silencing lncRNA candidates with doxycycline-inducible shRNAs (figure 3F and online supplemental figure S6I). Together, these analyses demonstrate that the Wnt-driven long non-coding transcriptome in CRC cell lines contains numerous essential transcripts that are efficiently identified with our screening platform. Most importantly, we confirmed that these vulnerabilities are generalisable to other CRC models and robust to changes in the targeting strategy, which facilitates therapeutic applications.
Supplemental material
CRC-specific expression of LINC02418
To focus our efforts on functional transcripts with high CRC specificity, we then intersected our dropout hits with their differential expression in patient-derived healthy tissue and CRC tumour samples (figure 4A). Interestingly, while previously validated transcripts LINC00173, SUZ12P1 and NBR1-AS showed moderate accumulation in tumour samples, LINC02418 revealed the most striking expression pattern with a highly significant ~8 log2 fold increase in CRC (figure 4A). Moreover, a detailed analysis of LINC02418 expression across tissues and different malignancies revealed its high CRC specificity (figure 4B–D and online supplemental figure S7A–C). As we did not find any, or very low expression of LINC02418 in normal adult tissues, we hypothesised that this transcript could be involved in a developmental transcriptional programme. In agreement, we detected LINC02418 expression by quantitative PCR and RNA in situ hybridisation (ISH) in human embryonal intestinal tissue at 13–20 weeks of gestational age (figure 4E,F and online supplemental figure S7D,E). These data suggest that LINC02418 is involved in early embryonic development in the intestine and reactivated in CRC.
Supplemental material
Expression of LINC02418 is cancer-specific and essential in CRC. (A) Intersection of CRISPRi dropout hits with their differential expression (log2 fold change (FC)) in patients with CRC versus normal samples (TCGA data). (B) Expression levels of LINC02418 lncRNA in various cancerous and normal tissues (TCGA data). Samples are ranked based on mean expression level in cancer samples. (C) Detection of LINC02418 transcripts (red dots) in a human tissue sample containing both healthy (H) and cancerous parts (C, delineated by dotted line), using RNA in situ hybridisation (ISH). Scale bar: 50 µm. (D) LINC02418 expression levels determined by quantitative PCR (qPCR) in five matched healthy and cancerous tissues. (E, F) Detection of LINC02418 in human embryonic tissue using (E) RNA-ISH and (F) qPCR. Scale bar: 50 µm. (G) Visualisation of LINC02418 cellular localisation (red) by RNA-ISH in LS180 cells with co-staining of E-cadherin (membrane, green) and Hoechst (nucleus, blue). (H) Relative abundance of LINC02418 RNA in both the nucleus and the cytoplasm. Nuclear (U2) and cytoplasmic (S14) RNAs were also monitored to control the cell fractionation. (I, J) Effect of shRNA-mediated LINC02418 depletion on cell proliferation in (I) LS180 and (J) CO100 cells after 4, 7 and 14 days of dox treatment. Scale bar: 100 µm. (K) ATP assays measuring the influence of a control (shCtrl) or shLINC02418-1 and shLINC02418-2 (sh1 and sh2) on the proliferation of CO100, a primary (spheroid) CRC cell line. The quantification was performed at indicated time points over 14 days. The relative luminescence unit (RLU) of each dox-treated sample was normalised to their respective untreated counterpart. P values (two-tailed Student’s t-test, n=3): *p<0.05; **p<0.01. nd, not determined.
Next, we studied the subcellular location of LINC02418 to determine the ideal silencing strategy. We found that the majority of RNA molecules localised to the cytoplasm (figure 4G,H and online supplemental figure S7F,G), which facilitates their direct targeting by shRNAs. Indeed, efficient knockdown of LINC02418 by specific dox-inducible shRNAs reduced the outgrowth of both adherent and suspension CRC cultures (figure 4I–K, online supplemental figures S7H and S8A). These observations confirm LINC02418 as a critical lncRNA in CRC and emphasise its potential as a therapeutic target.
Supplemental material
LINC02418 regulates MYC expression levels
To gain insight into the mechanism by which LINC02418 regulates CRC cell expansion, we performed both RNA-seq and Ribo-seq in LS180 cells (figure 5A). GSEA revealed several key gene sets downregulated following LINC02418 knockdown, including E2F targets, G2M checkpoint and two distinct MYC target gene signatures (figure 5B,C and online supplemental figure S8B–D). Furthermore, LINC02418 knockdown led to decreased translational efficiency (TE) of ribosomal subunit transcripts (figure 5D,E). Translation is a process known to be heavily impacted by MYC activity.20 In line with these observations, we found that LINC02418 knockdown could destabilise MYC at the protein level (figure 5F and online supplemental figure S8E–G), while only a modest effect was observed at the RNA level (online supplemental figure S8H,I), a feature reminiscent of microRNA (miRNA)-dependent regulation. Interestingly, others have previously reported the ability of LINC02418 to sponge miRNAs.21–24 To address this possibility, we tested whether LINC02418 mediates its function via the 3′-untranslated region (3′UTR) of MYC. Importantly, deletions in the 3′UTR of MYC alleviated its sensitivity to LINC02418 knockdown (figure 5G and online supplemental figure S9A,B). We then investigated whether LINC02418 could bind specific miRNAs known to regulate MYC. Previous reports have shown that MYC can be targeted by miRNA-24, miRNA-34a and let-7.25–27 We found that miRNA-24 specifically binds LINC02418 (figure 5H–J and online supplemental figure S9C). However, droplet digital PCR analyses revealed that LINC02418 transcripts are ~20 times less abundant than MYC RNA molecules, indicating that a direct transfer of miRNAs is unlikely to be a plausible mechanism (online supplemental figure S9D).28 Interestingly, previous efforts to model competing endogenous RNA (ceRNA) networks (ceRNETs) highlighted transcription and gene regulation as key biological processes modulated by these networks.29 In line with its influence on MYC, we hypothesised that LINC02418 could mediate its function through the regulation of a ceRNET. Thus, we listed all significantly downregulated genes following LINC02418 knockdown, and used TargetScan to predict the miRNA-dependent regulation of each transcript. Network analysis revealed that miR-24 was the most connected miRNA with a regulatory potential over 125 transcripts out of 213 (figure 5K and online supplemental table S9). Supporting this finding, concomitant interference with miR-24 and LINC02418 function restored MYC protein levels (figure 5L). Moreover, inhibition of miR-24 in the presence of LINC02418 was ineffective in modulating MYC protein levels (online supplemental figure S9E). Similarly, exogenous expression of LINC02418 could partially rescue MYC repression (online supplemental figure S9F–H). Together, these experiments indicate that LINC02418 is part of a ceRNET, which is strongly connected to miR-24. We suggest that silencing LINC02418 influences the balance of the ceRNET, which ultimately increases the availability of miR-24 and facilitates the repression of MYC. The miR-24-dependent repression of MYC reduces its transcriptional activity, further maintaining the ceRNET in a repressed state.
Supplemental material
Supplemental material
LINC02418 modulates MYC expression levels. (A) Workflow of RNA sequencing and ribosomal profiling following LINC02418 depletion in LS180 cells. CHX, cycloheximide. (B) Downregulated gene signatures identified by Gene Set Enrichment Analysis following LINC02418 depletion (shLINC02418). ES, enrichment score; FDR, false discovery rate; q.val, q value. (C) Enrichment plot for MYC target gene signature v1 in LINC02418 knockdown (shLinc) versus control (shCtrl). (D) Translational efficiency (TE) of small and large ribosomal protein genes (RPGs) following LINC02418 depletion (sh1 and sh2) in LS180 cells. For each transcript, the TE was normalised to the control knockdown and results are presented as log2 fold change (FC). Dots represent individual transcripts of the small and large ribosomal subunits (SUs). (E) Example of a ribosomal protein gene (RPL5) regulated at the translation level. Read coverage of RPL5 in both Ribosome sequencing (Ribo-seq) and RNA sequencing (RNA-seq) is displayed for each condition (control and LINC02418 knockdowns). (F) Western blot analysis showing MYC protein levels following the induction of a control (shCtrl) or two shRNAs targeting LINC02418 in LS180. Cells were treated with dox (1 µg/mL) for 4 days prior to sample preparation. Protein quantifications (ratio MYC/GAPDH) were normalised to the minus reactions. (G) Western blot analysis as performed in (f), but with a cell clone harbouring deletions in the MYC 3′UTR. (H) Regulation of LINC02418 luciferase reporter by a control (Ctrl) or three specific miRNAs (miR-24, miR-34a and Let-7) known to bind the MYC 3′UTR. For each reaction, the relative luciferase activity (the ratio of renilla to firefly) was normalised to the control reaction, which was arbitrarily set to 1. P values (two-tailed Student’s t-test, n=3): *p<0.05. (I) Illustration of LINC02418 luciferase reporter and the location of potential miR-24 (red) binding sites (MBS1-3). Coordinates of each binding site are indicated. (J) Luciferase assays showing the targeting of LINC02418 by miR-24. Luciferase assays were performed using both wild-type (wt) and mutant (mut) binding sites. The relative luciferase activities were processed as in (H). P values (two-tailed Student’s t-test, n=3): *p<0.05. (K) Regulatory network representing the interaction between shLINC02418-downregulated transcripts and the five most connected/predicted to target miRNAs (miR-24, miR-128-3 p, miR-129-5 p, miR-150-5 p and miR-7). Red edges and red boxes represent predicted targets of miR-24, while grey edges are connection mediated by the other four miRNAs. Yellow boxes are transcripts not predicted to be targeted by miR-24. (L) Rescue of MYC expression levels following concomitant silencing of LINC02418 and inhibition of three specific miRNAs (miR-34a, miR-135b and miR-24). A control anti-miR (anti-miR-Ctrl) was also used. Protein quantifications (ratio MYC/GAPDH) were normalised to the minus reaction.
LINC02418 expression promotes CRC stem cell functionality and prevents terminal differentiation
To further investigate the essential role of LINC02418 in CRC cell proliferation, we then assessed its potential influence on known stem cell markers and genes expressed in more differentiated cell types.30 31 We found that by interfering with LINC02418 expression several stem cell markers and Wnt target genes (eg, LGR5, ASCL2, AXIN2, EPHB2) were downregulated, whereas a subset of differentiation markers showed increased expression at the RNA (eg, KRT20, MUC2 and LYZ) and protein level (eg, KRT20) (figure 6A,B and online supplemental figure S10A,B). Furthermore, we analysed RNA polymerase I subunit A (PolR1A) levels in LINC02418 depleted cells. While high levels of PolR1A, which is involved in ribosomal DNA transcription and protein synthesis, are observed in CRC stem-like cells with high biosynthetic capacity, loss of PolR1A promotes terminal differentiation and growth arrest.32 Consistent with the elevation of specific differentiation markers, we found decreased PolR1A expression (figure 6B) as well as reduced cancer stem cell frequencies in limiting dilution assays in vitro (figure 6C and online supplemental figure S10C). Together these findings indicate that LINC02418 is essential for the maintenance of a population of CRC cells with high clonogenic and biosynthetic potential.
Supplemental material
Loss of LINC02418 drives terminal differentiation in vivo and halts tumour growth. (A) Volcano plot showing differentially regulated genes in LINC02418 depleted LS180 cells. Regulated Wnt-associated genes and stem cell markers are highlighted in red and differentiation markers are highlighted in green. (B) Regulation of cytokeratin 20 (KRT20) and RNA polymerase I subunit A (POLR1A) at the protein level following LINC02418 knockdown induction (4 days of dox, 1 µg/mL) in LS180 cells. Protein quantifications (POLR1A/GAPDH or KRT20/GAPDH ratios) were normalised to the minus reactions. (C) Influence of control (shCtrl) or shRNAs targeting LINC02418 (shLinc-1/shLinc-2) on the clonogenic potential of LS180 (left) and CO100 (right) cell lines as measured by limiting dilution assays. P values (two-tailed Student’s t-test): ***p<0.001. (D) Expression of KRT20 (left) and LGR5 (right) in xenograft tumours derived from LS180 CRC cells. RNA levels were normalised to GAPDH. P values (two-tailed Student’s t-test):*p<0.05, ***p<0.001. (E) Detection of KRT20 (green) and POLR1A (purple) by immunofluorescence and of LGR5 (turquoise) by RNA in situ hybridisation (ISH) on xenograft tumours used for gene expression analysis in (d). Nuclei (blue) were stained with Hoechst. Representative images of the same tumour area were derived from consecutive slides. Scale bars: 25 µm. (F) Cartoon displaying the time frame used for the pre-induction setup. (G) Outgrowth of xenografted CO100 (left) and LS180 (right) cells. Cells were treated with dox for 96 hours prior to subcutaneous injection. Dox was continuously supplied in the drinking water. Cross-sectional analyses (shCtrl vs shLinc) were performed using the Wilcoxon rank sum test, *p<0.05. (H) Cartoon displaying the time frame used for the intervention (established tumours) setup. (I) Influence of LINC02418 knockdown on pre-established xenograft tumours from CO100 (left) and LS180 (right) cell lines. Dox was continuously supplied in the drinking water when tumours were established. Tumour volumes are depicted over time as mean±SEM. Tumour growth data are fitted to exponential growth model: y=a×e(Kx) with a=starting point and K=growth rate. Cross-sectional analyses (shCtrl vs shLinc) were performed using the Wilcoxon rank sum test, *p<0.05.
LINC02418 is essential for CRC tumour growth
Next, we evaluated the impact of blocking LINC02418 on tumour growth in vivo using two xenograft models. Here, knockdown of LINC02418 resulted in marked differentiation as evidenced by decreased LGR5 and increased KRT20 expression (figure 6D,E and online supplemental figure S10D,E). In line with these observations, areas with high KRT20 levels showed reduced PolR1A expression (figure 6E). In addition, we found that the outgrowth of xenografts with LINC02418 knockdown was significantly delayed (figure 6F,G), while depletion of LINC02418 in established tumours impaired further expansion (figure 6H,I). Next, we sought to translate our findings by developing an exogenous RNAi-based therapeutic approach to target LINC02418 transcripts in vivo. To this end, we designed and in vitro tested a panel of 45 small interfering RNAs (siRNAs) (figure 7A and online supplemental figure S11A) and further validated 15 of them in two different CRC cell lines (figure 7B and online supplemental figure S11B). The best six siRNAs were subsequently evaluated for their phenotypic impact (figure 7C and online supplemental figure S11C,D). Next, we selected two different on-target, functional siRNAs to be formulated with lipid-based nanoparticle emulsion. We confirmed that lipid nanoparticle (LNP)-formulated siRNAs were able to decrease LINC02418 levels in vitro (figure 7D,E). Importantly, both LNP-formulated siRNAs significantly inhibited xenograft tumour outgrowth when administered intravenously (figure 7F,G). Moreover, these results recapitulate the phenotypic impact observed with the suppression of LINC02418 by shRNAs (figure 7H and online supplemental figure S11E,F), providing direct evidence for the therapeutic potential of Wnt-driven lncRNAs in CRC.
Supplemental material
LINC02418-targeted therapy. (A) Distribution of all 45 siRNAs targeting LINC02418. Coverage of LINC02418 locus (global run-on sequencing) and LINC02418 annotated exons are also displayed. (B) LINC02418 expression levels following the transfection of sixteen LINC02418-specific siRNAs and one control siRNA (Ctrl1) in the LS180 cell line. Selected siRNAs (F3 and F5) are highlighted in blue while the least efficient LINC02418-directed siRNA (H1) is highlighted in black. (C) Proliferation assays (incucyte) following LINC02418 knockdown in LS180. The best six LINC02418-directed siRNAs were compared with the least efficient LINC02418-directed siRNA (H1). Proliferation curves of selected siRNAs (F3 and F5) are indicated with blue arrows. (D) Cartoon showing the in vitro assessment of formulated siRNAs. (E) Silencing of LINC02418 following incubation of LS180 cells with a control (siCtrl) or two LINC02418-specific (F3 and F5) LNP-formulated siRNAs. P values (two-tailed Student’s t-test, n=3): *p<0.05. (F) Illustration of treatment setup using intravenous (IV) injection of lipid nanoparticles (LNPs). (G) Tumour growth curves following treatment with two siRNAs against LINC02418 or a control siRNA. Tumour volumes are depicted over time as mean±SEM. Cross-sectional analyses (siCtrl vs siLinc) were performed using the Wilcoxon rank sum test, *p<0.05. (H) Detection of LINC02418 (RNA), KRT20 and POLR1A in control siRNA (siCtrl, upper panels) and siLINC02418-treated tumours (lower panels). Zoom in locations are indicated with dotted rectangles. Scale bars (50 µM) are displayed. (I, J) Cartoon showing (I) the regulatory connections between β-catenin/TCF4, MYC, LINC02418, LINC02418/MYC-associated ceRNET and miR-24 in CRC and (J) following the delivery of LNP-formulated siRNAs targeting LINC02418.
Discussion
In this study, we have comprehensively characterised the Wnt-regulated long non-coding transcriptome in CRC. We provided direct evidence that these Wnt-driven lncRNAs are functionally relevant in CRC, as dropout screens identified a substantial panel of targets that were robustly associated with decreased expansion of CRC cell lines. We regard this as key evidence that lncRNAs downstream of the β-catenin/TCF complex are bona fide components of the Wnt signalling axis, and essential for mediating its oncogenic effect in CRC. Furthermore, we have established that the Wnt-regulated lncRNA LINC02418 is specifically expressed in CRC and seems to be involved in intestinal development during the first trimester. This phase is characterised by rapid expansion of the embryonal gut, driven by sustained and intense Wnt activation, as well as stem cell expansion similar to that observed in CRC growth.33–35 LINC02418 has previously been described in the context of CRC as well as lung cancer to regulate cell expansion by miRNA sponging.21–24 Here, we identified LINC02418 as a Wnt target gene and further extended its importance to the maintenance of high MYC expression levels in CRC (figure 7I). Known as a global transcription amplifier, MYC has a broad impact on many cellular processes including proliferation, biosynthesis and differentiation.36 37 In this respect, silencing LINC02418 decreases MYC expression levels, which subsequently impacts cancer cell proliferation in vitro and tumour growth in vivo. While many other lncRNAs have been associated with the regulation of MYC in CRC,38–41 LINC02418 stands out as an attractive therapeutic target, due to its exquisite cancer-specificity.
Mechanistically, LINC02418 appears to regulate MYC expression levels in a miRNA-dependent manner. Importantly, the stoichiometry between LINC02418 and MYC disfavours the possibility of a simple sponging mechanism. Previous efforts to quantitatively model miRNA sponging further support our observation, as most endogenous RNA molecules are unlikely to mediate significant competitive effects.28 42 Alternatively, complex cross-regulation within challenged ceRNETs may help explain connections or co-dependencies between specific transcripts.29 We suggest that silencing LINC02418 could potentially disturb the homeostatic state of its associated regulatory network. Interestingly, we found that miR-24 was the most connected miRNA in the LINC02418-associated ceRNET, and further highlighted its role in the regulation of MYC. The connection between MYC/E2F and the miR-24 was previously reported.26 However, it remains unclear how the LINC02418 connects with the ceRNET and MYC, and how its suppression initiates the downregulation of the network. Consitent with its impact on MYC expression levels, silencing of LINC02418 causes rapid differentiation and loss of stemness in CRC, which highlights its crucial role in CRC biology.
Finally, we provide evidence that lncRNAs are essential regulatory molecules for established tumors that can successfully be targeted by RNAi-based approaches (figure 7J). This opens new avenues for the development of a novel class of anticancer agents that target the long non-coding transcriptome downstream of various oncogenic pathways (eg, EGFR and NOTCH).43 44 We propose that the method presented here, based on the use of GRO-seq analyses of genetically perturbed in vitro models, paired with the analysis of patient samples and functional custom screens, can identify key vulnerabilities in other malignancies. Therapeutic exploitation might be facilitated by the extensive tissue- and context-specificity of lncRNAs, as compared with protein coding mRNAs, thus limiting toxicity in other tissues.
Novel developments in RNA interference as therapeutic modality in an increasing number of diseases further enforces the promises of this approach, in line with the here provided proof of concept for siRNA-based therapeutic delivery to tumours.45 46
Materials and methods
Cell culture
Colon cancer cell lines (Sanger) were cultured in Dulbecco’s Modified Eagle Medium/F12 (LoVo, HT55, SW1463, LS180, CL-40, HuTu-80, HCT116, T84, SW48, Caco-2) or Roswell Park Memorial Institute (LS411N, LS1034, MDST8, COLO320-HSR) containing 10% fetal calf serum (Life Technologies), penicillin and streptomycin. Previously established human primary colon cancer lines CO100 and RC511 were cultured as previously reported.47 48
Cell fractionation
Cells were harvested, washed in phosphate-buffered saline and incubated on ice with 20 mL of swelling buffer (10 mM Tris-HCl pH 7.5, 2 mM MgCl2, 3 mM CaCl2) for 10–15 min. To isolate nuclei, pellets were resuspended in 200 μL lysis buffer (10 mM Tris-HCl pH 7.5, 2 mM MgCl2, 3 mM CaCl2, 0.5% octylphenoxypolyethoxyethanol (IGEPAL), 10% glycerol). After centrifugation, the cytoplasmic fraction was transferred to a new tube and mixed with 3 volumes of TrizolLS (Qiagen). To obtain a clean nuclei fraction, pellets were resuspended in 1 mL of lysis buffer and spun down. The nuclei pellet was lysed in TRIsure. Samples were stored at −80°C until RNA isolation.
RNA interference
siRNAs were designed, synthesised, characterised and purified according to previously published protocols49 (online supplemental table S8). Transfection of siRNAs (50 nM) in LS180, T84 and HT55 CRC cell lines was performed using Dharmafect (1 or 2) following instructions provided by the manufacturer. LNP-formulated siRNAs were prepared as described elsewhere.50 Delivery of LNP-formulated siRNAs in vitro was achieved by incubating siRNAs (30 nM) with LS180 cells for 48 hours.
Supplemental material
Global run-on sequencing and library preparation
GRO-seq and preparation of sequencing libraries were performed as described previously.16 17 RNA-seq libraries were prepared using the TruSeq Stranded mRNA Library Prep Kit (Illumina) and quantified with the NEBNext Library Quant Kit (New England Biolabs). Barcoded samples were pooled equimolarly and multiplexed sequenced on the Illumina HiSeq 4000 platform.
GRO-seq analysis
The quality of single-end reads was assessed using FastQC V.0.11.3 (http://www.bioinformatics.bbsrc.ac.uk/projects/fastqc). Low-quality sequences were removed by Trimmomatic51 and high-quality reads were aligned to the human reference genome, V.GRCh37 (hg19), by using bowtie2 V.2.3.0.52 The generated SAM files were processed and converted to BED files using the function bamtobed from the BEDTools toolset.53 The orientation of the mapped reads was adjusted in the BED files using a bash script. The quantification of the nascent transcription was performed with the nascent RNA-seq analysis (NRSA) tool.54 The parameters corresponding to the minimum gene length (-l option) and minimum distance from enhancer to the transcription termination sites (TTS) (-dtts option) were adjusted to 200 bp and 5000 bp, respectively. We used the Human GENCODE lncRNA gene set (release 19) as reference for the quantification of lncRNAs and the hg19 annotation provided by NRSA for quantification of eRNAs and coding genes. The outputs generated by NRSA for the studied cell lines were integrated using an in-house python script available on https://github.com/vermeulenlab/lncRNA/. The data are available on NCBI GEO (GSE206582).
Plasmid constructs
To generate shRNA constructs, we first annealed complementary oligos and then ligated them into the linearised (AgeI and EcoRI) Tet-pLKO-puro (Addgene #21915) lentiviral expression vector. Single gRNAs were cloned into the pDECKO-mCherry backbone (Addgene #78534) using BsmBI restriction sites. Luciferase reporter vectors were generated by inserting the sequence of LINC02418 or specific miRNA binding sites into the linearised (Xho1 and Not1) psiCHECK-2 plasmid. Overexpression of LINC02418 was generated by the successive insertion of three fragments into the leGO-lnc (Addgene #80624) plasmid, using Xba1 and Xho1 sites. All oligo sequences are provided in online supplemental table S8.
Generation of CRISPRi cell lines
CRISPRi cell lines were generated by transducing cells with lentiviruses expressing TRE-KRAB-dCas9-IRES-BFP (Addgene #85449). Transduced cell populations were enriched by FACS sorting BFP-positive cells following induction with 1 µg/mL doxycycline (Sigma), using the SH800 Cell Sorter (Sony). Cell sorting was repeated until most cells (>90%) were BFP-positive.
CRISPRi library generation and sequencing
Transcription start site sequences (−50 bp to 350 bp) were used to generate gRNAs with CCTop-CRISPR\Cas9 target online tool. Negative (non-targeting) and positive (directed against essential genes) control gRNAs were included in the library.55–58 Screening libraries were PCR-amplified from genomic DNA and sequenced on the Illumina HiSeq 4000 platform. We used Bowtie2 (V.2.3.5.1, arguments --norc −5 22 –3 10) to align the sequencing FASTQ files to our custom library, and SAMtools (V.1.9) to create BAM files.52 59 MAGeCK (V.0.5.9.4) was used to count reads and performed statistical analyses.18
Colony formation assay
Cells were plated at a low density (2000-5000 cells) in six-well plates (Greiner) and incubated with or without dox (0.5–0.1 µg/mL) for 2–3 weeks. Colonies were fixed using a staining solution consisting of (50% Methanol, 10% Glacial acetic acid, and 0.1% Coomassie Blue).
Limiting dilution assay
The clonogenic potential was evaluated by plating CRC cells at various density (ie, 1, 2, 4, 6, 8, 12, 16, 24, 36, 48, 96, 192 and 384 cells per well), with eight or sixteen replicates per condition. Medium with dox to a final concentration of 0.5 µg/mL was added every other day and clonal outgrowth was assessed after 2 weeks. The extreme limiting dilution analysis ‘limdil’ function was applied to determine clonal frequency and significance.60
Luciferase assay
To perform luciferase assays, 293T cells were co-transfected with a reporter (LINC02418 or a miRNA binding site), a control (Ctrl) or specific miRNAs (miR-24, miR-34a and Let-7). Analysis was performed after 72 hours using the Dual-Glo Luciferase Assay System (Promega) and the Synergy HTX Multi-Mode Microplate Reader (BioTek).
ATP assay
The ATPlite 1 step Luminescence assay (Perkin Elmer) was performed according to manufacturer’s recommendation. Luminescent signal was read out with the Synergy HT plate reader (BioTek).
Tumour outgrowth and treatment
Prior to each in vivo experiment, mice were acclimatised 1 week at Amsterdam UMC animal facility (location AMC). Randomisation and group allocation were decided and monitored by animal technicians. Cells were induced with dox (1 µg/mL) for 4 days (in vitro) prior to injection. For LS180 and CO100 cell lines, 50 000 cells in medium were mixed 1:1 with Matrigel (Corning) and injected subcutaneously into both flanks of female nude (Hsd:Athymic Nude-Fox1nu) mice (aged 6–12 weeks, Envigo). Mice (six control (shCtrl, n=12 tumours) and six with an shRNA targeting LINC02418 (shLINC02418, n=12 tumours)) received drinking water supplemented with 1% glucose and 2 mg/mL dox, which was changed twice a week. Same groups and conditions (without pre-induction) were used to measure the effects of silencing LINC02418 expression in established tumours. For the exogenous RNAi-based intervention, 50 000 LS180 cells (1:1 with Matrigel) were injected subcutaneously in both flanks. Mice groups were defined as follows: six control (siCtrl, n=12 tumours), six siLINC02418-1 (siRNA F3) and six siLINC02418-2 (siRNA F5) (siLINC02418, n=12 tumours/siRNA). Intravenous administration of LNP-formulated siRNAs (3 mg/kg) was performed twice a week. Tumour growth was measured two times per week using the formula 0.5×length×width×height with a calliper. For each group, 12 tumours were measured. Statistical analysis between groups was performed using the online tool TumGrowth (https://kroemerlab.shinyapps.io/TumGrowth/)61 and the Wilcoxon rank sum test was applied. A total of 66 mice were used in this study. Sample sizes were determined based on previously established experiments.62
Human embryonic tissue
Human fetal intestinal tissue (gestational age 13–30 weeks) was obtained from the Bloemenhove clinic (Heemstede, the Netherlands) through the HIS Mouse Facility at the Amsterdam University Medical Center (AUMC). Written informed consent was obtained from all donors for the use of the material in research. Tissues were obtained with approval of the ethical committee of the AUMC, together with approval of experimental procedures by HIS Mouse Facility (AUMC) (protocol number 2016_285, #B2017369).
RNAScope and immunofluorescent staining
Detection of lncRNA LINC02418 (custom probe) and LGR5 transcript with RNAScope 2.5 HD RED or BROWN was performed on adherent cells, paraffine-embedded tissue and PFA-fixed xenograft sections according to manufacturer’s instructions. To enhance the signal of LINC02418 custom probe on embryonic tissue section, incubation step AMP1 was extended to 1 hour. Tissue section were counterstained with haematoxylin. For immunofluorescence staining, primary antibody (E-Cadherin Ab40772, Abcam 1:200) and goat anti-rabbit Alexa fluor 488 (A11035, Invitrogen 1:500) were used. Confocal microscopes SP8-X DLS and SMD (Leica) with Leica Application Suite-Advanced Fluorescence were used for imaging and image analysis.
Immunohistochemistry
Sections were incubated with primary antibody (β-catenin #9561, Cell Signaling Technology) and secondary antibody (anti-rabbit, Brightvision). Slides were incubated with 3,3′-Diaminobenzidine (DAB) substrate solution (K3468, Dako) for 1–3 min in the dark, rinsed with water and counterstained with haematoxylin. After rinsing in water, sections were dehydrated using a standard protocol, cleared in xylene and mounted with Pertex mounting medium. Consecutive slides were used for staining with RNA-ISH and IHC.
RNA isolation and complementary DNA
Samples were harvested in TRIsure (Bioline), and RNA was extracted according to the manufacturer’s recommendation. For tissue samples, 30 mg of fresh frozen material was homogenised in 1 mL of TRIsure using metallic beads in the TissueLyser LT (Qiagen) for 3 min at 50 Hz. After complete homogenisation, samples were further processed as described above. Superscript reverse transcriptase was used to generate complementary DNA (cDNA), according to the manufacturer’s recommendation. Primer sequences are shown in online supplemental table S8.
Droplet digital PCR
Absolute quantification of LINC02418 and MYC RNA molecules was performed using LS180 CRC cell cDNAs (40 ng RNA equivalent per reaction) and the QX200 Droplet Digital PCR System.
Western blot analysis
Protein detection was performed using primary antibodies: β-catenin (#9561 CST, 1:1000 in 5% bovine serum albumin (BSA)); c-Myc (#9402 CST, 1:1000 in 5% BSA); GAPDH (MAB347, Merck, 1:1000 in 5% milk); cytokeratin 20 (SPM140, GTX1520 GeneTex, 1:1000 in 5% BSA); RPA194 (C1, SC-48385, Santa Cruz Biotechnology, 1:1000 in 5% BSA), followed by incubation with HRP-coupled secondary antibodies (Southern Biotech, 1:5000).
RNA-sequencing
Libraries were prepared using the KAPA mRNA Hyperprep with RiboErase (Roche) according to manufacturer’s recommendation from 1 µg total RNA. Barcoded samples were multiplexed sequenced using single-end 50 bp reads on the Illumina HiSeq 4000 platform.
RNA-seq data analysis
We used HISAT2 (V.2.0.4) for alignment and Subread (V.1.6.4) for identifying gene counts.63 64 Differential expression analyses were performed using DESeq2 (V.1.26.0).65 66 Results were annotated with org.Hs.eg.db (V.3.10.0) and plotted using ComplexHeatmap (V.2.2.0) and ggplot2 (V.3.2.1).67 68 After ranking genes according to the decreasing Wald statistics value, GSEA Preranked (V.3.0) was used for GSEA with MSigDB Hallmark gene sets69–71 or with the WNT-signature of the same cell line obtained from the shCTN analysis. We used the results of the differential expression analysis from normal and tumour samples from The Cancer Genome Atlas (TCGA) COAD/READ data to identify which lncRNAs are potentially cancer-specific. We matched the ensemble ID of targeted candidate lncRNAs to the TCGA data and plotted the Benjamini-Hochberg adjusted p values against the log2 FCs.
Ribosomal profiling sample preparation and data analysis
Analysis were performed as previously described.72 After ribosomal RNA and transfer RNA filtering, Ribosome-protected fragment (RPF) reads were aligned to hg19 using STAR. Translational efficiencies were calculated using the RiboDiff tool73 and Gene Ontology terms were determined with the ClusterProfiler tool.74
Supplemental material
Supplemental material
Data availability statement
Data are available in a public, open access repository. All data relevant to the study are included in the article or uploaded as supplementary information. The sequence libraries generated in this study are publicly available through the National Center for Biotechnology Information (NCBI) Gene Expression Omnibus (GEO) under accession code: GSE206582.
Ethics statements
Patient consent for publication
Ethics approval
All in vivo experiments were approved by the Animal Experimentation Committee at Amsterdam UMC (location: Academic Medical Center (AMC)) in Amsterdam under the nationally registered license number AVD11800202114947, and were performed in accordance with national guidelines. This study involves human participants and was approved by the Medical Ethical Committee of Amsterdam UMC. Human fetal intestinal and CRC tissues were obtained with the approval of the Committee, and participants provided informed consent prior to their involvement in the study.
Acknowledgments
We thank Daniël Miedema and Tom van den Bosch for their support with statistical analysis, and Robin Blok for arranging the collection of patient samples in the Amsterdam UMC (location Academic Medical Center). We also thank the AMC core facilities for genomics, cellular imaging and pathology and the mouse research facilities for their technical support. This manuscript is based on Ronja S. Adams’ thesis.
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
- Data supplement 1
- Data supplement 2
- Data supplement 3
- Data supplement 4
- Data supplement 5
- Data supplement 6
- Data supplement 7
- Data supplement 8
- Data supplement 9
- Data supplement 10
- Data supplement 11
- Data supplement 12
- Data supplement 13
- Data supplement 14
- Data supplement 15
- Data supplement 16
- Data supplement 17
- Data supplement 18
- Data supplement 19
- Data supplement 20
- Data supplement 21
- Data supplement 22
Footnotes
LJS, RSA and LFM contributed equally.
LV and NL contributed equally.
Contributors LJS, RSA, LFM, NL and LV conceptualised the project. LJS, NL, LEN, AL, SE and OB performed in vitro experiments. LJS visualised the data. RSA and LFM performed data analysis and visualisation of GRO-seq and RNA-seq with support from DAZ and JK. RSA analysed and visualised the CRISPRi screen. LJS, KJL, SMvN, NL and LV designed in vivo experiments. LEN, AL, NEdG and SMvN performed in vivo experiments. ES and FL-P performed ribosomal profiling, data analysis and visualisation. FG and VM provided embryonic samples. CCE, MAJvE, DMP, SMvN, JK, NL and LV provided scientific input. TD, WJB and MAM provided siRNA reagents and scientific input. LJS, NL and LV wrote the manuscript with the help of RSA, LFM, SE, CCE, VM, KJL, FL-P and JK. All authors approved the content of the manuscript. NL is the guarantor.
Funding This work is supported by the Dutch Cancer Society (KWF#12228, #15052) to NL.
Competing interests LV received consultancy fees from Bayer, MSD, Genentech, Servier and Pierre Fabre, but these had no relation to the content of this publication. Currently, LV is an employee of Genentech and shareholder of Roche. TD, WJB and MAM are currently employed by Alnylam Pharmaceuticals.
Patient and public involvement Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.