MAJIQlopedia: an encyclopedia of RNA splicing variations in human tissues and cancer.

Quantification of RNA splicing variations based on RNA-Sequencing can reveal tissue- and disease-specific splicing patterns. To study such splicing variations, we introduce MAJIQlopedia, an encyclopedia of splicing variations that encompasses 86 human tissues and 41 cancer datasets. MAJIQlopedia reports annotated and unannotated splicing events for a total of 486 175 alternative splice junctions in normal tissues and 338 317 alternative splice junctions in cancer. This database, available at https://majiq.biociphers.org/majiqlopedia/, includes a user-friendly interface that provides graphical representations of junction usage quantification for each junction across all tissue or cancer types. To demonstrate case usage of MAJIQlopedia, we review splicing variations in genes WT1, MAPT and BIN1, which all have known tissue or cancer-specific splicing variations. We also use MAJIQlopedia to highlight novel splicing variations in FDX1 and MEGF9 in normal tissues, and we uncover a novel exon inclusion event in RPS6KA6 that only occurs in two cancer types. Users can download the database, request the addition of data to the webtool, or install a MAJIQlopedia server to integrate proprietary data. MAJIQlopedia can serve as a reference database for researchers seeking to understand what splicing variations exist in genes of interest, and those looking to understand tissue- or cancer-specific splice isoform usage.

Contrasting and Combining Transcriptome Complexity Captured by Short and Long RNA Sequencing Reads.

Mapping transcriptomic variations using either short or long reads RNA sequencing is a staple of genomic research. Long reads are able to capture entire isoforms and overcome repetitive regions, while short reads still provides improved coverage and error rates. Yet how to quantitatively compare the technologies, can we combine those, and what may be the benefit of such a combined view remain open questions. We tackle these questions by first creating a pipeline to assess matched long and short reads data using a variety of transcriptome statistics. We find that across datasets, algorithms and technologies, matched short reads data detects roughly 50% more splice junctions, with 10-30% of the splice junctions included at 20% or more are missed by long reads. In contrast, long reads detect many more intron retention events, pointing to the benefit of combining the technologies. We introduce MAJIQ-L, an extension of the MAJIQ software to enable a unified view of transcriptome variations from both technologies and demonstrate its benefits. Our software can be used to assess any future long reads technology or algorithm, and combine it with short reads data for improved transcriptome analysis.

Mapping and modeling human colorectal carcinoma interactions with the tumor microenvironment.

The initiation and progression of cancer are intricately linked to the tumor microenvironment (TME). Understanding the function of specific cancer-TME interactions poses a major challenge due in part to the complexity of the in vivo microenvironment. Here we predict cancer-TME interactions from single cell transcriptomic maps of both human colorectal cancers (CRCs) and mouse CRC models, ask how these interactions are altered in human tumor organoid (tumoroid) cultures, and functionally recapitulate human myeloid-carcinoma interactions in vitro. Tumoroid cultures suppress gene expression programs involved in inflammation and immune cell migration, providing a reductive platform for re-establishing carcinoma-immune cell interactions in vitro. Introduction of human monocyte-derived macrophages into tumoroid cultures instructs macrophages to acquire immunosuppressive and pro-tumorigenic gene expression programs similar to those observed in vivo. This includes hallmark induction of SPP1, encoding Osteopontin, an extracellular CD44 ligand with established oncogenic effects. Taken together, these findings offer a framework for understanding CRC-TME interactions and provide a reductionist tool for modeling specific aspects of these interactions.

Alternative splicing of its 5'-UTR limits CD20 mRNA translation and enables resistance to CD20-directed immunotherapies.

Aberrant skipping of coding exons in CD19 and CD22 compromises the response to immunotherapy in B-cell malignancies. Here, we showed that the MS4A1 gene encoding human CD20 also produces several messenger RNA (mRNA) isoforms with distinct 5' untranslated regions. Four variants (V1-4) were detected using RNA sequencing (RNA-seq) at distinct stages of normal B-cell differentiation and B-lymphoid malignancies, with V1 and V3 being the most abundant. During B-cell activation and Epstein-Barr virus infection, redirection of splicing from V1 to V3 coincided with increased CD20 positivity. Similarly, in diffuse large B-cell lymphoma, only V3, but not V1, correlated with CD20 protein levels, suggesting that V1 might be translation-deficient. Indeed, the longer V1 isoform contained upstream open reading frames and a stem-loop structure, which cooperatively inhibited polysome recruitment. By modulating CD20 isoforms with splice-switching morpholino oligomers, we enhanced CD20 expression and anti-CD20 antibody rituximab-mediated cytotoxicity in a panel of B-cell lines. Furthermore, reconstitution of CD20-knockout cells with V3 mRNA led to the recovery of CD20 positivity, whereas V1-reconstituted cells had undetectable levels of CD20 protein. Surprisingly, in vitro CD20-directed chimeric antigen receptor T cells were able to kill both V3- and V1-expressing cells, but the bispecific T-cell engager mosunetuzumab was only effective against V3-expressing cells. To determine whether CD20 splicing is involved in immunotherapy resistance, we performed RNA-seq on 4 postmosunetuzumab follicular lymphoma relapses and discovered that in 2 of them, the downregulation of CD20 was accompanied by a V3-to-V1 shift. Thus, splicing-mediated mechanisms of epitope loss extend to CD20-directed immunotherapies.

Alternative splicing of its 5'-UTR limits CD20 mRNA translation and enables resistance to CD20-directed immunotherapies.

Aberrant skipping of coding exons in CD19 and CD22 compromises responses to immunotherapy for B-cell malignancies. Here, we show that the MS4A1 gene encoding human CD20 also produces several mRNA isoforms with distinct 5' untranslated regions (5'-UTR). Four variants (V1-4) were detectable by RNA-seq in distinct stages of normal B-cell differentiation and B-lymphoid malignancies, with V1 and V3 being the most abundant by far. During B-cell activation and Epstein-Barr virus infection, redirection of splicing from V1 to V3 coincided with increased CD20 positivity. Similarly, in diffuse large B-cell lymphoma only V3, but not V1, correlated with CD20 protein levels, suggesting that V1 might be translation-deficient. Indeed, the longer V1 isoform was found to contain upstream open reading frames (uORFs) and a stem-loop structure, which cooperatively inhibited polysome recruitment. By modulating CD20 isoforms with splice-switching Morpholino oligomers, we enhanced CD20 expression and anti-CD20 antibody rituximab-mediated cytotoxicity in a panel of B-cell lines. Furthermore, reconstitution of CD20-knockout cells with V3 mRNA led to the recovery of CD20 positivity, while V1-reconstituted cells had undetectable levels of CD20 protein. Surprisingly, in vitro CD20-directed CAR T cells were able to kill both V3- and V1-expressing cells, but the bispecific T cell engager mosunetuzumab was only effective against V3-expressing cells. To determine whether CD20 splicing is involved in immunotherapy resistance, we performed RNA-seq on four post-mosunetuzumab follicular lymphoma relapses and discovered that in two of them downregulation of CD20 was accompanied by the V3-to-V1 shift. Thus, splicing-mediated mechanisms of epitope loss extend to CD20-directed immunotherapies. Key Points: In normal & malignant human B cells, CD20 mRNA is alternatively spliced into four 5'-UTR isoforms, some of which are translation-deficient.The balance between translation-deficient and -competent isoforms modulates CD20 protein levels & responses to CD20-directed immunotherapies. Explanation of Novelty: We discovered that in normal and malignant B-cells, CD20 mRNA is alternatively spliced to generate four distinct 5'-UTRs, including the longer translation-deficient V1 variant. Cells predominantly expressing V1 were still sensitive to CD20-targeting chimeric antigen receptor T-cells. However, they were resistant to the bispecific anti-CD3/CD20 antibody mosunetuzumab, and the shift to V1 were observed in CD20-negative post-mosunetuzumab relapses of follicular lymphoma.

Discovery of antibodies targeting multipass transmembrane proteins using a suspension cell-based evolutionary approach.

Due to their critical functions in cell sensing and signal processing, membrane proteins are highly preferred as pharmacological targets, and antibody drugs constitute the fastest growing category of therapeutic agents on the pharmaceutical market. However, major limitations exist in developing antibodies that recognize complex, multipass transmembrane proteins, such as G-protein-coupled receptors (GPCRs). These challenges, largely due to difficulties with recombinant expression of multipass transmembrane proteins, can be overcome using whole-cell screening techniques, which enable presentation of the functional antigen in its native conformation. Here, we developed suspension cell-based whole-cell panning methodologies to screen for specific binders against GPCRs within a naive yeast-displayed antibody library. We implemented our strategy to discover high-affinity antibodies against four distinct GPCR target proteins, demonstrating the potential for our cell-based screening workflow to advance the discovery of antibody therapeutics targeting membrane proteins.

A Bayesian model for unsupervised detection of RNA splicing based subtypes in cancers.

Identification of cancer sub-types is a pivotal step for developing personalized treatment. Specifically, sub-typing based on changes in RNA splicing has been motivated by several recent studies. We thus develop CHESSBOARD, an unsupervised algorithm tailored for RNA splicing data that captures "tiles" in the data, defined by a subset of unique splicing changes in a subset of patients. CHESSBOARD allows for a flexible number of tiles, accounts for uncertainty of splicing quantification, and is able to model missing values as additional signals. We first apply CHESSBOARD to synthetic data to assess its domain specific modeling advantages, followed by analysis of several leukemia datasets. We show detected tiles are reproducible in independent studies, investigate their possible regulatory drivers and probe their relation to known AML mutations. Finally, we demonstrate the potential clinical utility of CHESSBOARD by supplementing mutation based diagnostic assays with discovered splicing profiles to improve drug response correlation.

FBXW7ß isoform drives transcriptional activation of the proinflammatory TNF cluster in human pro-B cells.

Noncanonical exon usage plays many important roles in cellular phenotypes, but its contribution to human B-cell development remains sketchily understood. To fill this gap, we collected various B-cell fractions from bone marrow (BM) and tonsil donors, performed RNA sequencing, and examined transcript variants. We identified 150 genes that harbor local splicing variations in all pairwise comparisons. One of them encodes FBXW7, an E3 ubiquitin ligase implicated as a driver in several blood cancers. Surprisingly, we discovered that in normal human pro-B cells, the predominant transcript used an alternative first exon to produce the poorly characterized FBXW7ß isoform, previously thought to be restricted to neural tissues. The FBXW7ß transcript was also abundant in cell lines and primary samples of pediatric B-cell acute lymphoblastic leukemia (B-ALL), which originates in the BM. When overexpressed in a heterologous cell system, this transcript yielded the expected protein product, as judged by anti-FLAG immunoblotting and mass spectrometry. Furthermore, in REH B-ALL cells, FBXW7ß mRNA was the only FBXW7 isoform enriched in the polyribosome fraction. To shed light on possible functions of FBXW7ß, we used gain- and loss-of-function approaches and identified an FBXW7-dependent inflammatory gene signature, apparent in a subset of B-ALL with high FBXW7ß expression. This signature contained several members of the tumor necrosis factor superfamily, including those comprising the HLA Class III cluster (LTB, LST1, NCR3, LTA, and NFKBIL1). Our findings suggest that FBXW7ß expression drives proinflammatory responses, which could contribute to normal B-cell development, leukemogenesis, and responses to anticancer therapies.

A convergent malignant phenotype in B-cell acute lymphoblastic leukemia involving the splicing factor SRRM1.

A significant proportion of infant B-cell acute lymphoblastic leukemia (B-ALL) patients remains with a dismal prognosis due to yet undetermined mechanisms. We performed a comprehensive multicohort analysis of gene expression, gene fusions, and RNA splicing alterations to uncover molecular signatures potentially linked to the observed poor outcome. We identified 87 fusions with significant allele frequency across patients and shared functional impacts, suggesting common mechanisms across fusions. We further identified a gene expression signature that predicts high risk independently of the gene fusion background and includes the upregulation of the splicing factor SRRM1. Experiments in B-ALL cell lines provided further evidence for the role of SRRM1 on cell survival, proliferation, and invasion. Supplementary analysis revealed that SRRM1 potentially modulates splicing events associated with poor outcomes through protein-protein interactions with other splicing factors. Our findings reveal a potential convergent mechanism of aberrant RNA processing that sustains a malignant phenotype independently of the underlying gene fusion and that could potentially complement current clinical strategies in infant B-ALL.

High-throughput mutagenesis identifies mutations and RNA-binding proteins controlling CD19 splicing and CART-19 therapy resistance.

Following CART-19 immunotherapy for B-cell acute lymphoblastic leukaemia (B-ALL), many patients relapse due to loss of the cognate CD19 epitope. Since epitope loss can be caused by aberrant CD19 exon 2 processing, we herein investigate the regulatory code that controls CD19 splicing. We combine high-throughput mutagenesis with mathematical modelling to quantitatively disentangle the effects of all mutations in the region comprising CD19 exons 1-3. Thereupon, we identify ~200 single point mutations that alter CD19 splicing and thus could predispose B-ALL patients to developing CART-19 resistance. Furthermore, we report almost 100 previously unknown splice isoforms that emerge from cryptic splice sites and likely encode non-functional CD19 proteins. We further identify cis-regulatory elements and trans-acting RNA-binding proteins that control CD19 splicing (e.g., PTBP1 and SF3B4) and validate that loss of these factors leads to pervasive CD19 mis-splicing. Our dataset represents a comprehensive resource for identifying predictive biomarkers for CART-19 therapy.

Loss of AID exacerbates the malignant progression of CLL.

Activation-induced cytidine deaminase (AID) has been implicated as both a positive and a negative factor in the progression of B cell chronic lymphocytic leukemia (CLL), but the role that it plays in the development and progression of this disease is still unclear. We generated an AID knockout CLL mouse model, AID-/-/Eµ-TCL1, and found that these mice die significantly earlier than their AID-proficient counterparts. AID-deficient CLL cells exhibit a higher ER stress response compared to Eµ-TCL1 controls, particularly through activation of the IRE1/XBP1s pathway. The increased production of secretory IgM in AID-deficient CLL cells contributes to their elevated expression levels of XBP1s, while secretory IgM-deficient CLL cells express less XBP1s. This increase in XBP1s in turn leads AID-deficient CLL cells to exhibit higher levels of B cell receptor signaling, supporting leukemic growth and survival. Further, AID-/-/Eµ-TCL1 CLL cells downregulate the tumor suppressive SMAD1/S1PR2 pathway and have altered homing to non-lymphoid organs. Notably, CLL cells from patients with IgHV-unmutated disease express higher levels of XBP1s mRNA compared to those from patients with IgHV-mutated CLL. Our studies thus reveal novel mechanisms by which the loss of AID leads to worsened CLL and may explain why unmutated CLL is more aggressive than mutated CLL.

Identifying common transcriptome signatures of cancer by interpreting deep learning models.

BACKGROUND: Cancer is a set of diseases characterized by unchecked cell proliferation and invasion of surrounding tissues. The many genes that have been genetically associated with cancer or shown to directly contribute to oncogenesis vary widely between tumor types, but common gene signatures that relate to core cancer pathways have also been identified. It is not clear, however, whether there exist additional sets of genes or transcriptomic features that are less well known in cancer biology but that are also commonly deregulated across several cancer types. RESULTS: Here, we agnostically identify transcriptomic features that are commonly shared between cancer types using 13,461 RNA-seq samples from 19 normal tissue types and 18 solid tumor types to train three feed-forward neural networks, based either on protein-coding gene expression, lncRNA expression, or splice junction use, to distinguish between normal and tumor samples. All three models recognize transcriptome signatures that are consistent across tumors. Analysis of attribution values extracted from our models reveals that genes that are commonly altered in cancer by expression or splicing variations are under strong evolutionary and selective constraints. Importantly, we find that genes composing our cancer transcriptome signatures are not frequently affected by mutations or genomic alterations and that their functions differ widely from the genes genetically associated with cancer. CONCLUSIONS: Our results highlighted that deregulation of RNA-processing genes and aberrant splicing are pervasive features on which core cancer pathways might converge across a large array of solid tumor types.

Targeting CD123 in blastic plasmacytoid dendritic cell neoplasm using allogeneic anti-CD123 CAR T cells.

Blastic plasmacytoid dendritic cell neoplasm (BPDCN) is a rare hematologic malignancy with poor outcomes with conventional therapy. Nearly 100% of BPDCNs overexpress interleukin 3 receptor subunit alpha (CD123). Given that CD123 is differentially expressed on the surface of BPDCN cells, it has emerged as an attractive therapeutic target. UCART123 is an investigational product consisting of allogeneic T cells expressing an anti-CD123 chimeric antigen receptor (CAR), edited with TALEN® nucleases. In this study, we examine the antitumor activity of UCART123 in preclinical models of BPDCN. We report that UCART123 have selective antitumor activity against CD123-positive primary BPDCN samples (while sparing normal hematopoietic progenitor cells) in the in vitro cytotoxicity and T cell degranulation assays; supported by the increased secretion of IFNγ by UCART123 cells when cultured in the presence of BPDCN cells. UCART123 eradicate BPDCN and result in long-term disease-free survival in a subset of primary patient-derived BPDCN xenograft mouse models. One potential challenge of CD123 targeting therapies is the loss of CD123 antigen through diverse genetic mechanisms, an event observed in one of three BPDCN PDX studied. In summary, these results provide a preclinical proof-of-principle that allogeneic UCART123 cells have potent anti-BPDCN activity.

Modulation of CD22 Protein Expression in Childhood Leukemia by Pervasive Splicing Aberrations: Implications for CD22-Directed Immunotherapies.

Downregulation of surface epitopes causes postimmunotherapy relapses in B-lymphoblastic leukemia (B-ALL). Here we demonstrate that mRNA encoding CD22 undergoes aberrant splicing in B-ALL. We describe the plasma membrane-bound CD22 Δex5-6 splice isoform, which is resistant to chimeric antigen receptor (CAR) T cells targeting the third immunoglobulin-like domain of CD22. We also describe splice variants skipping the AUG-containing exon 2 and failing to produce any identifiable protein, thereby defining an event that is rate limiting for epitope presentation. Indeed, forcing exon 2 skipping with morpholino oligonucleotides reduced CD22 protein expression and conferred resistance to the CD22-directed antibody-drug conjugate inotuzumab ozogamicin in vitro. Furthermore, among inotuzumab-treated pediatric patients with B-ALL, we identified one nonresponder in whose leukemic blasts Δex2 isoforms comprised the majority of CD22 transcripts. In a second patient, a sharp reduction in CD22 protein levels during relapse was driven entirely by increased CD22 exon 2 skipping. Thus, dysregulated CD22 splicing is a major mechanism of epitope downregulation and ensuing resistance to immunotherapy. SIGNIFICANCE: The mechanism(s) underlying downregulation of surface CD22 following CD22-directed immunotherapy remains underexplored. Our biochemical and correlative studies demonstrate that in B-ALL, CD22 expression levels are controlled by inclusion/skipping of CD22 exon 2. Thus, aberrant splicing of CD22 is an important driver/biomarker of de novo and acquired resistance to CD22-directed immunotherapies. See related commentary by Bourcier and Abdel-Wahab, p. 87. This article is highlighted in the In This Issue feature, p. 85.

MYC Hyperactivates Wnt Signaling in APC/CTNNB1-Mutated Colorectal Cancer Cells through miR-92a-Dependent Repression of DKK3.

Activation of Wnt signaling is among the earliest events in colon cancer development. It is achieved either via activating mutations in the CTNNB1 gene encoding ß-catenin, the key transcription factor in the Wnt pathway, or most commonly by inactivating mutations affecting APC, a major ß-catenin binding partner and negative regulator. However, our analysis of recent Pan Cancer Atlas data revealed that CTNNB1 mutations significantly co-occur with those affecting Wnt receptor complex components (e.g., Frizzled and LRP6), underscoring the importance of additional regulatory events even in the presence of common APC/CTNNB1 mutations. In our effort to identify non-mutational hyperactivating events, we determined that KRAS-transformed murine colonocytes overexpressing direct ß-catenin target MYC show significant upregulation of the Wnt signaling pathway and reduced expression of Dickkopf 3 (DKK3), a reported ligand for Wnt co-receptors. We demonstrate that MYC suppresses DKK3 transcription through one of miR-17-92 cluster miRNAs, miR-92a. We further examined the role of DKK3 by overexpression and knockdown and discovered that DKK3 suppresses Wnt signaling in Apc-null murine colonic organoids and human colon cancer cells despite the presence of downstream activating mutations in the Wnt pathway. Conversely, MYC overexpression in the same cell lines resulted in hyperactive Wnt signaling, acquisition of epithelial-to-mesenchymal transition markers, and enhanced migration/invasion in vitro and metastasis in a syngeneic orthotopic mouse colon cancer model. IMPLICATIONS: Our results suggest that the MYC→miR-92a-|DKK3 axis hyperactivates Wnt signaling, forming a feed-forward oncogenic loop.

Tilting MYC toward cancer cell death.

MYC oncoprotein promotes cell proliferation and serves as the key driver in many human cancers; therefore, considerable effort has been expended to develop reliable pharmacological methods to suppress its expression or function. Despite impressive progress, MYC-targeting drugs have not reached the clinic. Recent advances suggest that within a limited expression range unique to each tumor, MYC oncoprotein can have a paradoxical, proapoptotic function. Here we introduce a counterintuitive idea that modestly and transiently elevating MYC levels could aid chemotherapy-induced apoptosis and thus benefit the patients as much, if not more than MYC inhibition.

RNA-binding proteins of COSMIC importance in cancer.

Herculean efforts by the Wellcome Sanger Institute, the National Cancer Institute, and the National Human Genome Research Institute to sequence thousands of tumors representing all major cancer types have yielded more than 700 genes that contribute to neoplastic growth when mutated, amplified, or deleted. While some of these genes (now included in the COSMIC Cancer Gene Census) encode proteins previously identified in hypothesis-driven experiments (oncogenic transcription factors, protein kinases, etc.), additional classes of cancer drivers have emerged, perhaps none more surprisingly than RNA-binding proteins (RBPs). Over 40 RBPs responsible for virtually all aspects of RNA metabolism, from synthesis to degradation, are recurrently mutated in cancer, and just over a dozen are considered major cancer drivers. This Review investigates whether and how their RNA-binding activities pertain to their oncogenic functions. Focusing on several well-characterized steps in RNA metabolism, we demonstrate that for virtually all cancer-driving RBPs, RNA processing activities are either abolished (the loss-of-function phenotype) or carried out with low fidelity (the LoFi phenotype). Conceptually, this suggests that in normal cells, RBPs act as gatekeepers maintaining proper RNA metabolism and the "balanced" proteome. From the practical standpoint, at least some LoFi phenotypes create therapeutic vulnerabilities, which are beginning to be exploited in the clinic.

MOCCASIN: a method for correcting for known and unknown confounders in RNA splicing analysis.

The effects of confounding factors on gene expression analysis have been extensively studied following the introduction of high-throughput microarrays and subsequently RNA sequencing. In contrast, there is a lack of equivalent analysis and tools for RNA splicing. Here we first assess the effect of confounders on both expression and splicing quantifications in two large public RNA-Seq datasets (TARGET, ENCODE). We show quantification of splicing variations are affected at least as much as those of gene expression, revealing unwanted sources of variations in both datasets. Next, we develop MOCCASIN, a method to correct the effect of both known and unknown confounders on RNA splicing quantification and demonstrate MOCCASIN's effectiveness on both synthetic and real data. Code, synthetic and corrected datasets are all made available as resources.