Abstract
Intron retention (IR) in cancer was for a long time overlooked by the scientific community, as it was previously considered to be an artifact of a dysfunctional spliceosome. Technological advancements made in the last decade offer unique opportunities to explore the role of IR as a widespread phenomenon that contributes to the transcriptional diversity of many cancers. Numerous studies in cancer have shed light on dysregulation of cellular mechanisms that lead to aberrant and pathologic IR. IR is not merely a mechanism of gene regulation, but rather it can mediate cancer pathogenesis and therapeutic resistance in various human diseases. The burden of IR in cancer is governed by perturbations to mechanisms known to regulate this phenomenon and include epigenetic variation, mutations within the gene body, and splicing factor dysregulation. This review summarizes possible causes for aberrant IR and discusses the role of IR in therapy or as a consequence of disease treatment. As neoepitopes originating from retained introns can be presented on the cancer cell surface, the development of personalized cancer vaccines based on IR-derived neoepitopes should be considered. Ultimately, a deeper comprehension about the origins and consequences of aberrant IR may aid in the development of such personalized cancer vaccines.
Introduction
Since the Nobel prize winning discovery of mRNA splicing more than four decades ago (1, 2), the concept of alternative splicing (AS) has evolved to become a research focus in the study of gene regulation in humans. Breakthroughs in high-throughput sequencing have shown that in humans 92% to 95% of multiexonic genes undergo AS to generate multiple isoforms from a single gene (3, 4). AS is a tightly regulated process involved in a myriad of conserved cellular processes (4–6). It is now recognized as being frequently altered in disease states, such as cancer (7, 8). Indeed, cancer cells often exhibit profound splicing alterations, including intron retention (IR), and many isoforms are specifically associated with cancer progression and metastasis (9).
Different archetypes of AS have been defined over time, with exon skipping being the most frequent AS event in humans, followed by alternative donor and acceptor site selection, and IR. Although the first three forms of AS have been given prime attention and studied extensively, IR was not considered as an important contributor of informational diversity in normal biology and disease until recently. Interestingly, although Dvinge and colleagues (10) demonstrated that exon skipping was the most common form of AS in the transcriptomes of many primary cancers, IR was found to be the most imbalanced form of AS when comparing the transcriptomes of solid tumor samples with the adjacent healthy tissue. The absence of mutations directly affecting the RNA splicing machinery also shows that intron-containing mRNAs contribute to the transcriptional diversity of many cancers even in the absence of direct mutational errors.
IR is defined by a mature mRNA transcript that retains at least one intron—in a sense holding on to its junk bond(s) because introns have in some circles been referred to as junk DNA. IR has been shown to affect as many as approximately 80% of human protein-coding genes (11). The observation of widespread IR in normal biology and cancer had to wait until advanced next-generation sequencing technologies became available (10, 12, 13). Indeed, IR is a prevalent phenomenon in cancer and disease that is featured in this review. In normal human biology, landmark discoveries have shown the prevalence of the IR phenomenon (reviewed in refs. 14–16). For instance, IR is tissue and cell-type specific. It has a higher incidence rate in neural and immune cell types, but is less frequently observed in embryonic stem cells and muscle (13). Furthermore, IR appears to play a dominant role in various stages of cell differentiation and development. In hematopoiesis, IR is widespread in the myeloid lineage, where it affects the maturation of granulocytes (17), megakaryocytes (18), and erythrocytes (19, 20), but also in the lymphoid lineage during T- and B-cell activation (21, 22). In addition, polyadenylated intron-containing transcripts can stably accumulate in the cell nucleus and rapidly undergo activity-dependent splicing, cytoplasmic export, and active translation. This process mobilizes a pool of mRNAs to allow rapid protein synthesis in response to external stimuli (21, 23–25). Furthermore, temporal regulation of specific gene expression via IR was observed during spermatogenesis; nuclear detained IR transcripts, which have long half-lives and are more stable, were recruited to polyribosomes days after their synthesis (26).
Diverse fates await intron-retaining transcripts in cancer and disease (Fig. 1). They can be detained in the nucleus (a process referred to as intron detention) where they can be targeted, degraded by nuclear degradation pathways, and lead to the inactivation of tumor-suppressor genes (TSG; Fig. 1D). Alternatively, they can be stored in the nuclear compartment to be rapidly processed, exported, and translated in the cytoplasm upon environmental stimuli (15, 16). When introns are retained in the mature mRNA transcript and exported to the cytoplasm, they can be detected by the cytoplasmic surveillance machinery and degraded via nonsense-mediated decay (NMD) given that they typically contain in-frame premature termination codons (PTC; Fig. 1F; refs. 27, 28). IR transcripts can also generate peptides for endogenous processing, proteolytic cleavage, and presentation on the cell surface in complex with MHC-I (Fig. 1E; refs. 29–32). Indeed, IR transcripts go through a first round of translation in which peptides are produced that can bind to MHC-I (30). In addition, they can give rise to new protein isoforms (i.e., oncoproteins), which can differ from the canonical protein either in their function and/or localization (Fig. 1E; refs. 33–35). Hypothetically, they could generate noncoding RNA(s) involved in the regulation of oncogenes and/or TSGs, but the biological relevance of this mechanism of gene regulation requires experimental validation (Fig. 1G).
The fate of an intron to be or not to be retained within the mature mRNA transcript, is the result of a combination of complex regulatory mechanisms, which involve both cis- and trans-acting modulators that allow organisms to respond rapidly to physiological changes (16). The dysregulation of such cis- and trans-modulators can lead to aberrant IR, which is often observed in cancer (10). Features of retained introns that have been characterized in global gene regulation studies (10–13) include the presence of weaker 5′ and 3′ splice sites, higher GC content, and shorter lengths. Furthermore, Attig and colleagues (36) observed that the retention of introns flanking a retrotransposed Alu-exon is observed in a significant number of Alu-exon–containing transcripts. It has also been shown that cryptic exons can function as decoys that engage intron-terminal splice sites and compete with cross-intron interactions required for intron excision, thereby promoting IR (37). Moreover, enrichment for serine–arginine protein-binding sites was observed in retained introns in contrast to constitutively spliced introns. In addition, trans-regulators and epigenetic factors such as DNA methylation, histone marks, and the availability of splicing factors can influence IR levels. The influence of epigenetic alterations as driving forces of tumor initiation has clearly emerged in the recent years, and drugs targeting reversible epigenetic abnormalities are being increasingly explored to reverse epigenetic effects in cancer (38, 39).
The discovery of a new class of IR-associated diseases has emerged over the last decade (14, 15). Hence, there is an imperative to achieve a better understanding of the mechanisms leading to aberrant IR in order to provide approaches to control IR-associated oncogenic and pathogenic changes. In this review, we discuss the impact and contribution of IR in disease as well as its potential to provide diagnostic, prognostic, and therapeutic tools for cancer control.
Causes and Consequences of Aberrant IR in Disease
Epigenetics, epitranscriptomics, and IR
DNA packaging into nucleosomes and chromatin affects the processes of gene transcription and cotranscriptional splicing (40, 41). In an analysis of chromatin accessibility in 42 clear cell renal cell carcinoma and 7 normal kidney samples, Simon and colleagues (42) found tumor-specific differences in chromatin accessibility. In a subgroup of tumors, nucleosome depletion was associated with target sites of the HIF transcription factor family linked to genes involved in the response to hypoxia. They also identified two tumor subpopulations with either normal levels of the SETD2 protein, a histone methyltransferase that trimethylates H3K36, or SETD2 depletion. As H3K36 can be methylated by mechanisms independent of SETD2, this prompted a further investigation of the open chromatin regions marked by H3K36me3 in SETD2-deficient tumors. RNA sequencing (RNA-seq) analysis of tumors with normal and depleted levels of H3K36me3 revealed differential IR and aberrant splicing between the two groups, and H3K36me3-deficient tumors had a dramatic increase in open chromatin regions immediately upstream of intron/exon junctions (Fig. 1A). IR was also dramatically increased in the H3K36me3-deficient tumors at 95% of the transcripts (6,551 in total) marked by H3K36me3 in normal kidney. These results show that IR and splicing defects affect a large fraction of genes with altered chromatin accessibility in SETD2-mutant tumors (42).
Accumulating evidence suggests that SETD2 regulates RNA splicing and modulates AS of the genes implicated in oncogenesis. In SETD2-deficient intestines in a colorectal cancer mouse model, Yuan and colleagues found that 279 genes were undergoing IR (39.2%), 198 with exon skipping (27.8%), and 161 with mutually exclusive exons (22.8%). Interestingly, reduced IR levels were observed in 81.8% of IR-affected genes, whereas 68% of SETD2-influenced SE genes exhibited an increase of exon inclusion in the SETD2-deficient mouse model. Furthermore, SETD2 depletion led to a more than 2-fold reduction in H3K36me3, compared with the controls. The production of intron-loss or exon-inclusion transcripts seems to be facilitated by the loss of SETD2 or H3K36me3 (43). Yuan and colleagues further investigated the interesting case of IR in Dvl2, whereby the lack of retained introns becomes disease causing. Indeed, the authors show that Dvl2 exhibits robust IR in control samples and is subjected to NMD as retention of intron 2 introduces a PTC. Setd2 depletion in mouse intestine leads to a remarkable increase in Dvl2 pre-mRNA transcripts without intron 2. They also showed that Setd2 is directly involved in regulating the processing of Dvl2 pre-mRNA as there was a marked decrease of H3K36me3 modifications within the IR areas of the Dvl2 gene locus. They concluded that upregulation of DVL2 in the absence of SETD2 causes the hyperactivation of Wnt signaling in colorectal cancer (43). Surprisingly, their analysis of Pol II elongation rate revealed that Setd2 depletion leads to Pol II pausing at intron 2 of Dvl2, suggesting that the slower rate of transcription elongation might facilitate the removal of Dvl2-intron 2. However, these results contradict observations made in other studies where RNA Pol II stalling near the splice junctions and within introns represses the recruitment of splicing factors and thereby stabilized IR (13, 44). Because there is no consensus on the role of Pol II elongation in IR regulation, additional studies are needed to decipher the mechanisms linking Pol II rate to pathologic IR changes.
Dysregulation of epigenetic readers can also alter IR in disease. For example, Guo and colleagues (45) found that the H3.3 lysine 36 trimethylation reader BS69 (a.k.a. ZMYND11) promotes IR in HeLa cells by interacting with splicing factors including EFTUD2. BS69 binds to H3K36me3 and H3.3K36me3 marks in intron-retaining genes. Therefore, BS69-mediated splicing regulation occurs in a SETD2-dependent manner (Fig. 1A). BS69 knockdown in lung cancer cells led to reduced IR, whereas the knockdown of EFTUD2 resulted in the upregulation of IR, suggesting that BS69 suppresses splicing via EFTUD2 (45).
BRD4 (bromodomain protein 4) is an epigenetic reader that binds to acetylated histones and moderates the Pol II elongation rate by interacting with a subunit of the positive transcription elongation factor b (pTEFb) complex (46). Hussong and colleagues (47) have demonstrated that BRD4 regulates stress-induced splicing (including IR). Heat shock–mediated increase in IR was enhanced through BRD4 knockdown in WI38 cells. Indeed, under heat shock, a total number of 5,879 retained introns exhibited a 2-fold increase in IR, whereas the number reaches 7,421 when heat shock is combined with BRD4 knockdown. BRD4 knockdown alone did not show significant changes in IR events. Upon heat shock, BRD4 was recruited to nuclear stress bodies, which are enriched in acetylated histones (H4K8ac and H4K16ac). Treatment with bromodomain and extraterminal motif inhibitors blocked BRD4-acetylated histone binding and reduced the formation of BRD4-containing nuclear stress bodies. Given that a significant fraction of IR events in lung adenocarcinomas (48) overlap with heat shock and BRD4 knockdown-induced IR (Table 1), Hussong and colleagues argued that bromodomain and extraterminal motif inhibitors might induce IR in oncogenes and thereby suppress tumor growth (47).
Disease . | Cause . | Gene(s)/Intron(s) . | Disease causality . | Consequence(s) . | Reference . |
---|---|---|---|---|---|
Myeloproliferative diseases | Genetic variation | ABCC3 & HOOK1/V | Not established | Not characterized | (89) |
6 cancer types | Multiple/Multiple | Not established | Inactivation of TSGs | (53) | |
Pyruvate carboxylase deficiency | PC/XV | Established | Intron containing in-frame PTC (in-silico) | (90) | |
The congenital long QT syndrome | KCNH2/IX | Not established | Not characterized | (91) | |
Malignant melanoma | LKB1 & STK11/IV | Not established | Not characterized | (92) | |
DM2, FECD, and C9-ALS/FTD | Multiple/Multiple | Not established | CNBP Intron 1 containing in-frame PTC (in-silico) | (55) | |
Fish eye disease | LCAT/IV | Established | Intron-containing in-frame PTC; null allele | (58) | |
Von Willebrand disease | VWF/XXXXIV | Not established | Truncated VWF protein | (57) | |
Late infantile neuronal ceroid lipofuscinosis | CLN2/VII | Not established | Production of an alternative CLN2 protein isoform | (93) | |
Autosomal-recessive gray platelet syndrome | NBEAL2/IX | Putative | Intron-containing in-frame PTC (in-silico) | (56) | |
Cowden syndrome | PTEN/IV & VII | Not established | Intron-containing in-frame PTC (in-silico) | (94) | |
31 cancer types | Multiple/Multiple | Not established | Not characterized | (54) | |
Chronic lymphocytic leukemia | Splicing factor dysregulation | Multiple/Multiple | Not established | Not characterized | (71) |
Amyotrophic lateral sclerosis | Multiple/Multiple | Not established | Nuclear loss of SFPQ protein | (73) | |
Prostate cancer | CYCLIN-D1/IV | Not established | Production of an alternative Cyclin-D1 protein isoform | (62) | |
Breast cancer | Multiple/Multiple | Not established | Not characterized | (64) | |
Head and neck cancer | EIF2B5/XII | Not established | Intron-containing in-frame PTC; Truncated eIF2Bϵ protein | (63) | |
Inflammatory bowel disease | Multiple/Multiple | Not established | Intron-containing in-frame PTC (in-silico) | (72) | |
Lymphoma | Multiple/Multiple | Not established | Production of an alternative Dvl1 protein isoform | (65) | |
Myelodysplastic syndrome | Multiple/Multiple | Not established | Not characterized | (69) | |
Pancreatic cancer | CCK-B/IV | Not established | Not characterized | (61) | |
Uveal melanoma | ABCC5/V | Not established | Not characterized | (70) | |
Liquid and solid tumors | Multiple/Multiple | Not established | Not characterized | (10) | |
Breast and lung cancer | Epigenetics | Multiple/Multiple | Not established | Degradation via Nuclear RNA Surveillance Complex | (51) |
Colorectal carcinoma | HDAC6 & TP53BP1/Multiple | Not established | Production of an alternative HDAC6/TP53BP1 protein isoform | (95) | |
Breast cancer | Multiple/Multiple | Not established | Not characterized | (49) | |
Kidney cancer | Multiple/Multiple | Not established | Nucleosome destabilization | (42) | |
Intestine cancer | DVL2/II | Not established | Increased DVL2 expression/Upregulated Wnt activity | (43) | |
Lung cancer | Multiple/Multiple | Not established | Not characterized | (45) | |
Breast and pancreatic cancer | FBXL2, ULK1, CARD10/XXII; XVII | Not established | Degradation via NMD is prevented when CELF2 is silent | (77) | |
Breast, colon, lung cancer | Not determined | KAT/III | Not established | Truncated KAT protein | (96) |
Melanoma | c-MYC & Sestrin-1/II & III; IX & X | Not established | Introns-containing in-frame PTCs and source of miRNAs (in-silico) | (66) | |
Colon cancer | CD44/IX | Not established | Not characterized | (97) | |
Colon cancer | OGT/IV | Not established | Nuclear detention of OGT mRNA transcript | (24) | |
Colon cancer | CD44/XIIX | Not established | Not characterized | (98) | |
Esophageal, colonic, bladder, and breast cancer | CD44/IX | Not established | Not characterized | (99) | |
Gastrointestinal stromal tumor | CCK2R & CCKBR/IV | Not established | Not characterized | (100) | |
Hepatocellular carcinoma | Multiple/Multiple | Not established | Intron-containing in-frame PTC and degraded via NMD (in-silico) | (85) | |
Kidney, liver, and pancreatic cancer | SLC28A1/IV | Not established | Not characterized | (101) | |
Leukemia | CD19/II | Not established | Intron-containing in-frame PTC; Untranslated IR transcript (ribosome profiling) | (82) | |
Lung cancer | Multiple/Multiple | Not established | Intron-containing in-frame PTC and degraded via NMD (in-silico) | (48) | |
Multiple tumors | Multiple/Multiple | Not established | Source of neoepitopes | (29) | |
Prostate cancer | KLK 1-5 & 15/III | Not established | Intron-containing in-frame PTC; production of alternative protein isoforms | (102) | |
Prostate cancer | Multiple/Multiple | Not established | Not characterized | (103) | |
Prostate cancer | CLK1-5/IV | Not established | Intron-containing in-frame PTC; truncated CLK1–5 protein | (104) | |
Neurodevelopmental psychiatric disorders | Multiple/Multiple | Not established | Nuclear retention in a subset of transcripts | (105) | |
Alzheimer's disease | Multiple/Multiple | Not established | Intron-containing in-frame PTC; production of alternative protein isoforms | (106) |
Disease . | Cause . | Gene(s)/Intron(s) . | Disease causality . | Consequence(s) . | Reference . |
---|---|---|---|---|---|
Myeloproliferative diseases | Genetic variation | ABCC3 & HOOK1/V | Not established | Not characterized | (89) |
6 cancer types | Multiple/Multiple | Not established | Inactivation of TSGs | (53) | |
Pyruvate carboxylase deficiency | PC/XV | Established | Intron containing in-frame PTC (in-silico) | (90) | |
The congenital long QT syndrome | KCNH2/IX | Not established | Not characterized | (91) | |
Malignant melanoma | LKB1 & STK11/IV | Not established | Not characterized | (92) | |
DM2, FECD, and C9-ALS/FTD | Multiple/Multiple | Not established | CNBP Intron 1 containing in-frame PTC (in-silico) | (55) | |
Fish eye disease | LCAT/IV | Established | Intron-containing in-frame PTC; null allele | (58) | |
Von Willebrand disease | VWF/XXXXIV | Not established | Truncated VWF protein | (57) | |
Late infantile neuronal ceroid lipofuscinosis | CLN2/VII | Not established | Production of an alternative CLN2 protein isoform | (93) | |
Autosomal-recessive gray platelet syndrome | NBEAL2/IX | Putative | Intron-containing in-frame PTC (in-silico) | (56) | |
Cowden syndrome | PTEN/IV & VII | Not established | Intron-containing in-frame PTC (in-silico) | (94) | |
31 cancer types | Multiple/Multiple | Not established | Not characterized | (54) | |
Chronic lymphocytic leukemia | Splicing factor dysregulation | Multiple/Multiple | Not established | Not characterized | (71) |
Amyotrophic lateral sclerosis | Multiple/Multiple | Not established | Nuclear loss of SFPQ protein | (73) | |
Prostate cancer | CYCLIN-D1/IV | Not established | Production of an alternative Cyclin-D1 protein isoform | (62) | |
Breast cancer | Multiple/Multiple | Not established | Not characterized | (64) | |
Head and neck cancer | EIF2B5/XII | Not established | Intron-containing in-frame PTC; Truncated eIF2Bϵ protein | (63) | |
Inflammatory bowel disease | Multiple/Multiple | Not established | Intron-containing in-frame PTC (in-silico) | (72) | |
Lymphoma | Multiple/Multiple | Not established | Production of an alternative Dvl1 protein isoform | (65) | |
Myelodysplastic syndrome | Multiple/Multiple | Not established | Not characterized | (69) | |
Pancreatic cancer | CCK-B/IV | Not established | Not characterized | (61) | |
Uveal melanoma | ABCC5/V | Not established | Not characterized | (70) | |
Liquid and solid tumors | Multiple/Multiple | Not established | Not characterized | (10) | |
Breast and lung cancer | Epigenetics | Multiple/Multiple | Not established | Degradation via Nuclear RNA Surveillance Complex | (51) |
Colorectal carcinoma | HDAC6 & TP53BP1/Multiple | Not established | Production of an alternative HDAC6/TP53BP1 protein isoform | (95) | |
Breast cancer | Multiple/Multiple | Not established | Not characterized | (49) | |
Kidney cancer | Multiple/Multiple | Not established | Nucleosome destabilization | (42) | |
Intestine cancer | DVL2/II | Not established | Increased DVL2 expression/Upregulated Wnt activity | (43) | |
Lung cancer | Multiple/Multiple | Not established | Not characterized | (45) | |
Breast and pancreatic cancer | FBXL2, ULK1, CARD10/XXII; XVII | Not established | Degradation via NMD is prevented when CELF2 is silent | (77) | |
Breast, colon, lung cancer | Not determined | KAT/III | Not established | Truncated KAT protein | (96) |
Melanoma | c-MYC & Sestrin-1/II & III; IX & X | Not established | Introns-containing in-frame PTCs and source of miRNAs (in-silico) | (66) | |
Colon cancer | CD44/IX | Not established | Not characterized | (97) | |
Colon cancer | OGT/IV | Not established | Nuclear detention of OGT mRNA transcript | (24) | |
Colon cancer | CD44/XIIX | Not established | Not characterized | (98) | |
Esophageal, colonic, bladder, and breast cancer | CD44/IX | Not established | Not characterized | (99) | |
Gastrointestinal stromal tumor | CCK2R & CCKBR/IV | Not established | Not characterized | (100) | |
Hepatocellular carcinoma | Multiple/Multiple | Not established | Intron-containing in-frame PTC and degraded via NMD (in-silico) | (85) | |
Kidney, liver, and pancreatic cancer | SLC28A1/IV | Not established | Not characterized | (101) | |
Leukemia | CD19/II | Not established | Intron-containing in-frame PTC; Untranslated IR transcript (ribosome profiling) | (82) | |
Lung cancer | Multiple/Multiple | Not established | Intron-containing in-frame PTC and degraded via NMD (in-silico) | (48) | |
Multiple tumors | Multiple/Multiple | Not established | Source of neoepitopes | (29) | |
Prostate cancer | KLK 1-5 & 15/III | Not established | Intron-containing in-frame PTC; production of alternative protein isoforms | (102) | |
Prostate cancer | Multiple/Multiple | Not established | Not characterized | (103) | |
Prostate cancer | CLK1-5/IV | Not established | Intron-containing in-frame PTC; truncated CLK1–5 protein | (104) | |
Neurodevelopmental psychiatric disorders | Multiple/Multiple | Not established | Nuclear retention in a subset of transcripts | (105) | |
Alzheimer's disease | Multiple/Multiple | Not established | Intron-containing in-frame PTC; production of alternative protein isoforms | (106) |
It is known that DNA methylation can modulate IR (44). In an analysis of four breast cancer types, Kim and colleagues (49) observed a negative correlation between CpG-site methylation and IR in a small population of retained introns. Given that some identified methylated CpGs were located within splicing enhancer regions, the authors concluded that IR is not regulated solely by DNA methylation but rather through complex interactions between DNA methylation and splicing regulatory elements (49).
Kamdar and colleagues (50) assessed the distribution of methylation and hydroxymethylation in prostate cancer and observed a shift of 5hmC enrichment from exonic regions in healthy cells to the intronic regions in cancer cells. Their analysis suggests intronic 5hmC is associated with reduced expression of genes involved in signaling, regulation of cellular proliferation, and cAMP biosynthesis regulation in prostate cancer (Fig. 1A; ref. 50).
Posttranscriptional RNA modifications together with RNA-binding protein activity can also contribute to the regulation of pathologic IR events. Fish and colleagues demonstrated that TARBP2 can regulate m6A deposition co-transcriptionally and promote methylation of nascent RNA (51). The authors demonstrated that TARBP2 directly controls the stability of its bound targets via co-transcriptional recruitment of the METTL3 methyltransferase complex, resulting in intron methylation and subsequent retention of the intron. Indeed, TARBP2 binds to introns in pre-mRNAs in breast cancer cells and recruits the RNA methylation machinery to deposit m6A marks on these transcripts. These m6A marks repel splicing regulator binding, resulting in increased IR and degradation of pre-mRNA transcripts by the nuclear exosome (Fig. 1B; ref. 51). Taken together, these results establish for the first time a link between an RNA-binding protein (TARBP2), m6A methylation, and controlled IR. However, research linking epitranscriptomic alterations to pathologic IR needs further exploration. Future studies should examine the mechanisms that underlie posttranscriptional RNA modifications. In addition, further improvements in functional validation methods of RNA editing events will be required to measure the impact and contribution of these epitranscriptomic alterations to pathologic IR events.
Genetic variants causing IR
The efficiency of splicing is known to be influenced by conserved elements located at the 5′ and 3′ splice sites, branchpoint consensus region, and the polypyrimidine tract of pre-mRNA sequences (52). Therefore, one of the most conspicuous and intuitive manifestations of aberrant IR in disease is the presence of mutations affecting these conserved elements. Several studies have reported mutations within the gene body that would trigger IR and disease (Table 1; Fig. 1B).
Recent studies have shown that a reduced splice site strength can lead to higher relative frequencies of IR and potentially inactivate TSGs in various cancers (53, 54). In their comprehensive analysis of somatic single nucleotide variants (SNV) resulting in aberrant splicing, Jung and colleagues (53) found that SNVs causing IR were enriched in TSGs. Jung and colleagues have analyzed The Cancer Genome Atlas (TCGA) patient samples from six distinct cancer types (breast, colon, liver, lung, ovarian, and uterine cancer). Their RNA-seq analysis of 1,812 cancer patient samples showed that 97% of detected IR-associated SNVs generate PTCs, which lead to an insufficient level of tumor-suppressor protein. In contrast, only 50% of exon-skipping events generate PTCs. However, IR-transcripts can also escape NMD and result in a partial inactivation of tumor suppressors such as TP53, ARID1A30, and VHL31 (53). Their work therefore exposed, for the first time, IR as a common mechanism of TSG inactivation. The association of IR and TSGs was also demonstrated in a separate study carried out by Shiraishi and colleagues (54). In their whole-exome and transcriptome sequencing analysis of 8,976 cancer samples, they generated a catalog of 14,438 splicing-associated variants. Shiraishi and colleagues have analyzed 31 of the 33 cancers represented in TCGA including those 6 analyzed by Jung and colleagues. Their results suggest that IR, as a consequence of splicing-associated variants, is occurring in 30 of the 31 cancer types affecting 1,428 unique genes. The only exception is thymoma. In 4 of 119 thymoma samples, the authors detected 9 splicing-associated variants, none of which produces IR. TP53 was found to be one of the most frequently targeted TSGs with multiple recurrent variants at splice donor and acceptor sites leading to abnormal IR events. They concluded, in agreement with Jung and colleagues, that IR was a major mechanism of splicing-associated variant-induced TSG inactivation and provided novel insights into IR-related transcriptional deregulation in cancer.
Genetic diseases frequently show aberrant splicing that leads to IR (Table 1). For example, Sznajder and colleagues demonstrated that disease-associated GC-rich intronic microsatellite expansions are selectively associated with host IR in a variety of affected patient cells and tissues with hereditary diseases such as myotonic dystrophy type 2, Fuchs endothelial corneal dystrophy, amyotrophic lateral sclerosis, and frontotemporal dementia (55). The mutation causing the repetition of GC across the intronic sequence could be partially responsible for the retention of certain introns as GC microsatellite expansions are predicted to form highly stable RNA secondary structures (hairpins and G-quadruplexes). Thus, the presence of these RNA structures induced by the GC-rich microsatellite expansion would have an inhibitory effect on splicing of a given host intron by preventing the binding of trans-acting factors and by slowing down RNA polymerase II elongation.
Furthermore, several studies have reported mutations located either at the exon–intron boundary, or beyond the core splice sites and at the branchpoint region that can inhibit splicing and lead to IR (Table 1; Fig. 1B).
RNA-seq analysis performed on an individual with Gray platelet syndrome showed an abnormal distribution of reads mapping to NBEAL2, a gene encoding the neurobeachin-like 2 protein, which is critical for the development of platelet alpha-granules. Investigation of the NBEAL2 locus by genomic DNA sequencing from the same individual identified two mutations, one at the exon–intron 9 boundary and the other outside of the core splice sites in exon 28, both inducing IR in the NBEAL2 mRNA (Table 1; Fig. 1B). The retention of introns 9 and 28 within NBEAL2 mRNA transcripts introduces PTCs that trigger NMD. Consequently, NBEAL2 protein expression is reduced, which is considered the main cause of autosomal-recessive gray platelet syndrome (56).
Yadegari and colleagues have identified a heterozygous silent mutation (C7464>T) in exon 44 of the von Willebrand factor gene, which triggers the retention of intron 44 in a family with type 1 von Willebrand disease (57). The aberrant intron-containing transcript encodes for a truncated protein that lacks the C-terminal end of the Willebrand factor and accumulates abnormally in the endoplasmic reticulum. By using a combination of in silico secondary and tertiary structure analysis of the pre-mRNA, they demonstrated that the mutation has a long-distance/allosteric influence on the RNA structure by inhibiting the accessibility of the core splice site. Their study described a new molecular pathologic mechanism by which a silent mutation outside the core splice sites triggers IR and disease. Hence, the impact of silent/exonic variations on pre-mRNA structure and splicing is likely to be uncovered as more targeted approaches are applied in cancers.
Mutations in the intronic branchpoint region, a region that plays a central role in the splicing mechanism, have been observed in several diseases and are associated with IR (Table 1; Fig. 1B). A heterozygous T to C nucleotide substitution, 22 bases upstream of the acceptor splice site of intron 4, was found in the LCAT gene, which encodes a plasma glycoprotein that is involved in the metabolism of fish eye syndrome (FED). This point mutation in the branchpoint consensus sequence caused a null allele due to complete IR in 3 patients suffering from FED (58).
IR induced via splicing factor dysregulation
Dysregulation of genes involved in RNA processing can cause aberrant splicing in diseases such as cancer (Table 1; Fig. 1C). An important resource for tumor-specific splicing alterations is the TCGA SpliceSeq database (59). This interactive website serves as a repository for TCGA splicing analysis results generated using the SpliceSeq analysis software (60).
In a pan-cancer analysis of AS, Dvinge and Bradley found that the expression of some splicing factors (among other genes) is strongly associated with aberrant IR in cancers such as acute myeloid leukemia, colon cancer, and breast cancer (10). These splicing factors, which include SF3A1, SF3B1, and SF3B2, are involved in 3′ splice site selection (Table 1).
The RNA-induced silencing complex member TARBP2 binds to introns of pre-mRNAs, which leads to intron detention and decreased transcript stability of hundreds of TARBP2 targets (51). In this context, it was shown that nuclear TARBP2 interacts with mRNA processing and export factors. Overexpression of TARBP2 is associated with several cancers including lung adenocarcinoma, lung squamous cell carcinoma, breast invasive carcinoma, prostate adenocarcinoma, uterine corpus endometrial carcinoma, and thymoma. Moreover, TARBP2 has been shown to affect lung tumor growth and breast cancer metastasis (51).
Perturbations in the expression of some splicing factors seem to trigger very specific IR events. Ding and colleagues, for example, have demonstrated that reduced levels of the U2 small nuclear ribonucleoprotein particle auxiliary splicing factor U2AF35 lead to the retention of intron 4 in the cholecystokinin-B/gastrin receptor in pancreatic carcinoma (61).
Although partial retention of intron 4 in Cyclin D1b is facilitated by the CCND1 G/A870 polymorphism, it can also be induced through the RNA-binding protein SRSF1 (62). SRSF1 expression increases during tumor progression in prostate cancer; SRSF1 binding to the CCND1 splice donor site causes expression of the oncogenic CCND1 isoform Cyclin D1b (Table 1).
Various cellular and environmental stimuli can cause splicing factor dysregulation. For instance, solid tumors react to hypoxia by impeding energy-consuming processes such as splicing and translation. In head and neck cancer, hypoxia-mediated reduction of core splicing factor expression (i.e., SF1, SRSF1, SRSF3, and SRSF7) results in nearly 90% of IR-affected genes displaying increased retention of introns in hypoxic compared with normoxic cells (63). One target of hypoxia-induced IR is the master regulator of translation initiation EIF2B5, the isoform expression of which is influenced by differential binding of SRSF3 to EIF2B5 intron 12. Truncated eIF2Bϵ protein expression leads to globally reduced protein synthesis and enhanced survival in head and neck cancer cells (63).
Increase in total pre-mRNA abundance in cancer cells can increase the burden on the spliceosome to process excessive RNA molecules. Hsu and colleagues found that spliceosomal integrity, i.e., constitutive expression of splicing factors (BUD31, SF3B1, U2AF1, EFTUD2, and SNRPF), is required for cancer cell survival (64). The proto-oncogene MYC regulates the expression of core components of the pre-mRNA splicing machinery, such as PRMT5. Perturbation of the MYC–PRMT5 axis leads to IR and exon skipping (153 AS events detected) in lymphomagenesis (65), and inhibition of the spliceosome in MYC-hyperactivated cancer cells leads to significantly increased IR in 42% of genes analyzed (2,848 of 6,861) and reduced cell viability (64). Interestingly, MYC itself can be affected by IR (66). Indeed, MYC undergoes IR specifically in melanoma, indicating potential novel functions of its noncanonical intron-containing transcripts in human melanomagenesis (see section on IR as a diagnostic tool).
Apart from dysregulated expression, splicing factors are often affected by genomic variances such as somatic mutations or copy-number variations in primary tumors and metastasis (67). Loss-of-function mutations in splicing factors can affect functions such as 3′ splice site or exon recognition and activate tumorigenic potential. Recurrent mutations in splicing factors have been found in myelodysplastic syndromes, acute and chronic myeloid leukemia, chronic lymphocytic leukemia, melanoma, as well as cancers of the pancreas, lung, breast, bladder, and others (67). Some of these frequently mutated splicing factors regulate IR, which can explain some of the aberrant IR patterns observed in cancer (10). For example, the spliceosome gene ZRSR2, which is involved in 3′ splice site recognition, is frequently mutated in patients with myelodysplastic syndrome among other affected splicing factors such as SF3B1, SRSF2, and U2AF1 (68). Madan and colleagues have shown that ZRSR2 is important for U12-type intron splicing (69). In the ZRSR2-mutant MDS, 43 of 45 introns retained were U12-dependent. ZRSR2 deficiency or loss-of-function mutations led to increased retention of U12-type introns and affect processes such as cell growth and hematopoietic cell differentiation.
Nonrecurrent SF3B1 mutations in 15% of patients with uveal melanoma were associated with improved survival and AS in multiple genes, which includes IR in the ABC transporter family member ABCC5 (70). Using long-read sequencing, Tang and colleagues have shown that SF3B1 mutations are associated with differential 3′ splice site selection and downregulation of IR events in chronic lymphocytic leukemia (71).
Although not examined to the same extent, splicing factor dysregulation can be observed in other diseases too. One example is the inflammatory bowel disease, in which Häsler and colleagues have found 47 differentially regulated splicing factors affecting 33 IR events (72). Intron-retaining genes associated with inflammatory bowel disease are involved in signal transduction, the secretory pathway, the immune system, and drug metabolism. In turn, chronic inflammatory bowel disease can predispose to colorectal cancer.
It is known that IR is associated with response mechanisms to neuronal activity (23). IR is also the predominant mode of splicing in motor neurogenesis. According to a study published by Luisier and colleagues, IR affects splicing factor expression during differentiation of induced-pluripotent stem cells derived from patients with amyotrophic lateral sclerosis (73). Increased IR in SFPQ transcripts, independent of the genetic background, results in decreased nuclear abundance of the SFPQ protein, suggesting a feedback mechanism of aberrant splicing regulation during motor neuron differentiation in amyotrophic lateral sclerosis (73).
IR in Treatment and Therapy
IR as a diagnostic tool
The previous sections have established that aberrant IR is a common molecular characteristic of cancers. RNA-seq and other technologies have demonstrated that IR is widespread across cancers and genetic diseases, while contributing to their transcriptomic diversity (10). Therefore, it is reasonable to assume that IR could be used as a disease biomarker and diagnostic tool. As previously mentioned (see section on genetic variants causing IR; ref. 55), analysis of the CCTGexp mutation, which is the largest genome-wide microsatellite expansion reported in the CNBP gene, revealed abnormally elevated levels of mis-spliced CNBP transcripts in myotonic dystrophy type 2. The expression is sufficiently high to allow RT-PCR–based detection of IR in both tissues and peripheral blood. Detection of intronic repeat expansion could therefore be used as a sensitive disease-specific diagnostic biomarker. In addition, examination of IR events through RT semiquantitative PCR protocols, using total RNA preparations derived from basal and squamous cell skin cancers, and melanoma biopsy specimens, showed that c-MYC and SESTRIN-1 genes proved to undergo IR specifically in melanoma. Noncanonical splicing of these genes appears to provide a powerful and reliable IR-based molecular signature that separates melanoma from non-melanoma tumors (66).
Treatment directed to IR
IR-derived neoantigens represent potential targets for immunotherapies. Indeed, Smart and colleagues were among the first to demonstrate that tumor-specific IR-derived neoepitopes could be detected in both patient samples and cell lines (29). They could computationally identify aberrant IR events generating immunogenic peptides and confirm their association with MHC-I using mass spectrometry. Thus, their data provide the first evidence that IR-derived neoepitopes can be processed and presented to the immune system through the MHC-I pathway. The immunogenicity of specific tumor IR-derived neoepitopes and evaluation of their clinical relevance in patients should be fruitful areas for further investigation.
Therapeutic targeting of IR-carrying cancer cell populations has been the subject of only a few studies to date. In a recent study, Sailer and colleagues have shown that an orally available small molecule, H3B-8800, can modulate splicing selectively in cells bearing a mutant spliceosome (74). Instead of reducing aberrant-IR harboring cells, H3B-8800 seems to do the opposite and triggers IR to induce preferential lethality in cancer cells. In a nutshell, recurrent mutations in genes encoding splicing factors, such as SRSF2, U2AF1, and SF3B1, have been identified in genomic analyses of cancers (see section on splicing factor dysregulation). Dysfunction of these splicing factors forces cancer cells to rely on the remaining functional spliceosome components. H3B-8800 interacts with the SF3B complex and modulates both wild type and mutant spliceosome activity and induces the retention of short, GC-rich introns in genes encoding splicing factors. These intron-retained mRNA sequences are then thought to be degraded through the NMD pathway. H3B-8800 appears to modulate the expression of numerous RNA splicing factors. The enrichment of IR events in such important splicing components would be more detrimental for spliceosome-mutant tumor cells, which are already deficient in splicing, in comparison with wild-type cells harboring an intact splicing machinery. The preferential killing of epithelial and hematologic spliceosome-mutant tumor cells shows great potential for the clinical testing of H3B-8800 as one of the first RNA splicing factor–targeting drugs. The development and utilization of H3B-8800-like drugs, in order to eradicate IR-bearing cancer cell populations, should be the subject of further studies.
Also, more work regarding the potential therapeutic benefits of splicing regulatory perturbations is required. If for example specific detrimental splicing events are identified as disease-causing or as therapeutic vulnerability (56), a targeted context-specific treatment might be sought. If in contrast global levels of IR matter for a disease outcome (48), perhaps a systemic intervention could be considered to adjust transcriptome-wide IR levels.
In this context, compounds such as H3B-8800 targeting the spliceosome or splicing regulators are a promising new class of anticancer agents (74). Another example is the pladienolide derivative E7107, which suppresses tumor growth in patient-derived triple-negative breast cancer xenografts by targeting SF3B1, a component of the U2 snRNP (75). The therapeutic potential of the Cdc2-like kinase (CLK) inhibitor T-025 was recently demonstrated in an allograft model of MYC-driven breast cancer treatment (76). T-025 causes AS and mediates cell death and growth suppression in vitro and in vivo. More preclinical and clinical testing of therapeutic splicing perturbations is required to answer questions regarding efficacy, efficiency, and adverse effects in a tumor-specific and personalized context.
Furthermore, Piqué and colleagues (77) observed that promoter DNA methylation silences the expression of the exon inclusion-regulator CELF2 in pancreatic, gastric, and breast cancer cells. CELF2 restoration via 5′-aza-2′-deoxycytidine treatment led to differential IR in several breast cancer–specific genes and tumor growth reduction. Moreover, epigenetic silencing of CELF2 demonstrated prognostic value and was associated with poor clinical outcomes (77).
Alternatively, antisense oligonucleotides (AON) comprise a novel class of therapeutics that have been used to induce the expression of preferred mRNA splice variants in a given gene (37, 78–80). AONs are highly specific as their mechanism of action via hybridization to target mRNA (or pre-mRNA) sequences.
AONs, also referred to as splice-switching oligonucleotides, were used by Flynn and colleagues in their model of spinal muscular atrophy (78). The homozygous loss of the survival motor neuron 1 gene (SMN1) is considered the main cause of this severe childhood disease. However, a homologous gene encoding an identical protein called SMN2 can partially rescue the loss of SMN1, but only when exon 7 of SMN2 is included in the mature mRNA. By targeting AONs to exon 8 acceptor splice site and exon splice enhancers of SMN2, the authors were hoping to block critical transacting factors and slow down the RNA polymerase elongation rate to increase exon 7 inclusion. To their surprise, AON treatment resulted in the retention of both exon 7 and intron 7. Unfortunately, the increased length of the 3′ untranslated region, due to retention of intron 7, introduced negative regulatory elements that could explain the observed decrease in SMN protein upon AON treatment. Even though the outcome of this study was not the one expected, it provided a new avenue where AONs could, for example, mediate terminal IR to repress burdensome gene expression of oncogenes in cancer.
Another example of AONs used to modulate splicing was shown in the STAT5B gene, which is linked to prostate cancer progression. A naturally occurring alternative isoform (STAT5Δ) lacking the C-terminal transactivation domain due to IR has been shown to act as a tumor suppressor. Shchelkunova and colleagues (79) used steric-blocking AONs with a complimentary sequence to an STAT5B exon–intron boundary to switch STAT5B function from tumor activating to tumor suppressing. This shows that AONs have the ability to block the splicing machinery from accessing splice sites and thereby enhance alternative intron/exon retention. Although Shchelkunova and colleagues had a limited delivery efficiency and could only induce IR in up to 10% of STAT5B mRNA transcripts in vitro, a significant decrease in cell proliferation of human PC-3 prostate cancer cells was observed.
In a more recent study, AONs were used to block decoy splice sites in endogenous pre-mRNA (80). Exons located within introns can function as decoys, which engage the intron-terminal splice sites to block intron excision (37). These cryptic noncoding cassettes are much more common in large (>1 kb) retained introns than they are in small-retained introns or in nonretained introns. By targeting the 5′ splice site of a specific decoy exon with AONs, Parra and colleagues showed that significantly reduced IR levels and increased gene expression can be achieved in human erythroblasts (80). For example, IR in the OGT gene was reduced from approximately 80% to 20% in primary human erythroblasts, which was concomitantly accompanied by an increase of spliced OGT mRNA and OGT protein expression. Although this study was not carried out in a context of disease, it demonstrates that AONs can modulate the level of a subset of IR events.
In conclusion, the functional consequences of transcript isoforms tuned by AONs show great potential and should be further implemented to reach higher modulation efficiency and specificity of IR events. However, the remaining challenge, after safe administration to patients, is to induce sufficient levels of splice modulation in target tissues (81).
IR as a consequence of therapy
IR can also be the consequence of therapy. Indeed, it can be triggered when resistance, after an ostensibly curative intervention, is acquired. The use of chimeric antigen receptor–armed T-cell therapy targeting CD-19 (CART-19) has revolutionized treatment for B-cell acute lymphoblastic leukemia. Unfortunately, patients relapse in 30% of cases, even after complete response, and often lose the CD19 epitope. Asnani and colleagues investigated the nature of AS of the CD19 transcript (82). When expressed in the CD19 knockout B-cell line Raji, analysis of the processing of CD19 exon 2 in their minigene system showed that CD19-negative cells undergo robust intron 2 retention, placing a PTC 40 amino acids downstream of the exon 2/intron 2 junction. CD19 intron–containing transcripts are found predominantly in the non-translated monosomal RNA fraction, rather than in the translationally active polysomal RNA fraction, likely due to the presence of the PTC in intron 2. This would explain the loss of the CD19 epitope recognized by the CAR-T cells. They conclude that retention of CD19 intron 2 is functionally equivalent to a nonsense mutation that would cause premature termination and either NMD of the transcript or a truncated CD19 protein, thereby contributing to CART-19 resistance in leukemias (82). The development of innovative and adaptive treatments to prevent antigen loss due to mRNA splicing–derived events such as IR will be crucial for the improvement of cell-based immunotherapies. For example, a possible approach to therapy is the development of AON-based therapies to correct pathologic IR events, which arise as a consequence of CAR-T cell treatment (Supplementary Fig. S1).
Conclusion and Perspective
Since the discovery of IR, tremendous advances have been made in understanding the roles that retained introns play in cellular biology across a wide range of taxonomic groups. In human biology, IR has been shown to be an undeniable contributor of transcriptomic and proteomic diversity, which is often altered in cancer.
ENCODE (encodeproject.org) and the Cancer Cell Line Encyclopedia (CCLE; portals.broadinstitute.org/ccle) are valuable resources for studying IR and its regulation in cancer cell lines. CCLE incorporates genomics and transcriptomics (short read RNA-seq) data of more than 1,000 cancer cell lines and has recently added DNA methylation, histone H3 modification, microRNA expression, and reverse-phase protein array data (83). These data allow for a comprehensive analysis of cancer cell–specific IR landscapes and regulatory mechanisms. ENCODE on the other hand provides more epigenetics data and has recently added valuable long-read sequencing datasets from multiple cancer cell lines.
Although the use of short-read RNA-seq has been instrumental over the last decade to successfully detect IR events, sequencing technologies are rapidly evolving and adapting to answer the needs of the scientific community. Innovative technologies have been recently developed to overcome the limitations and technical problems that short-read sequencing present (e.g., internal priming, RT template switching artifact, etc.). Long-read sequencing will offer a more effective and appropriate way to capture dynamic IR changes observed in normal biology and cancer. Many recent studies have already taken advantage of these emerging technologies to demonstrate their utility for cancer and splicing research (71, 82, 84, 85). For example, Nanopore sequencing of full-length cDNA was recently performed to identify novel alternative transcript isoforms (including IR) associated with the SF3B1-K700E mutations in chronic lymphocytic leukemia samples (71). However, this technology relies on the generation of full-length cDNAs, which can introduce biases during the amplification step. In the foreseeable future, direct RNA-seq, which enables the sequencing of native RNA without any PCR amplification steps, will become the new standard for the detection of bona fide AS isoforms. Currently only Oxford Nanopore platform offers a commercial kit for direct RNA-seq method (86). One of the many advantages of this technique is the detection of co-/post-transcriptional base modifications in RNA, or the epitranscriptome, adding yet another level of complexity to the identification of key players contributing to IR in disease (87, 88).
Specific and widespread pathogenic IR events have been widely observed in cancer, but detailed characterization of the functional consequences of such splicing alterations is often lacking. A systematic evaluation of the contributions of aberrant IR events to disease biology is generally missing in most studies. To date, very few instances have been described in which IR plays a causal role in disease emergence or progression. Most studies highlight differences in global or specific IR events but rarely disclose IR-induced phenotypic changes. This is in part caused by the challenging task of artificially inducing specific IR events. Therefore, more work into exploring the fates of IR transcripts in disease and cell physiological consequences is required. Suitable approaches for synthetic IR induction could be CRISPR-based splice site editing or AONs.
It is often presumed that intron-containing mRNA transcripts are degraded via NMD, thus causing downregulation of the host gene. Thus, to truly appreciate the level of IR, inhibition of the NMD pathway, using either broad acting agents such as caffeine or the more specific knockdown of the core NMD components, such as UPF1, is needed in order to stabilize IR transcripts and facilitate their quantification. Yet, IR is also capable of generating diverse functional protein isoforms, of which, the impact on disease has been largely neglected. The prediction of IR fate(s) with bioinformatics tools combined with high-resolution proteomic profiling analyses should be further pursued to evaluate the incidence of overlooked IR fates in disease biology. Ultimately, a better understanding of pathologic IR alterations will allow the identification of novel disease biomarkers for the development of therapeutic IR modulators.
Authors' Disclosures
U. Schmitz reports grants from Cancer Institute of New South Wales and grants from Cancer Council NSW during the conduct of the study. J.E.J. Rasko reports grants from NHMRC, grants from NSW Genomics Collaborative Grant, grants from an anonymous foundation, and grants from Cancer Council NSW during the conduct of the study and reports advisory roles in Gene Technology Technical Advisory, OGTR, and Australian Government. No disclosures were reported by the other authors.
Acknowledgments
We thank the funding agencies for research support. Figures were created with Biorender.com.
This work was supported by the National Health and Medical Research Council (Investigator Grant 1177305 to J.E.J. Rasko; Project Grants #1080530, #1061906, #1128175, and #1129901 to J.E.J. Rasko); the NSW Genomics Collaborative Grant (J.E.J. Rasko); Cure the Future (J.E.J. Rasko); Tour de Cure research grants to J.E.J. Rasko; and an anonymous foundation (J.E.J. Rasko). U. Schmitz holds a Fellowship from the Cancer Institute of New South Wales. Financial support was also provided by Cancer Council NSW project grants (RG11-12, RG14-09, and RG20-12) to J.E.J. Rasko and U. Schmitz.