RNA-binding proteins in disease etiology: fragile X syndrome and spinal muscular atrophy
- Department of Biochemistry and Biophysics, Perelman School of Medicine, University of Pennsylvania, Philadelphia, Pennsylvania 19104, USA
- Corresponding author: gdreyfuss{at}hhmi.upenn.edu
Abstract
All RNAs exist in complexes (RNPs) with RNA-binding proteins (RBPs). Studies in my lab since the 1980s have identified, sequenced and characterized the major pre-mRNA- and mRNA-RBPs (hnRNPs/mRNPs), revealing RNA-binding domains and common features of numerous RBPs and their central roles in posttranscriptional gene regulation. The first links between RBPs and RNPs to diseases emerged serendipitously for fragile X syndrome, as its gene (FMR1) encoded RBP (FMRP), and spinal muscular atrophy (SMA), caused by deficits in survival motor neurons (SMN). Discoveries of the SMN complex and its unanticipated function in RNP assembly, essential for spliceosomal snRNP biogenesis, advanced understanding of RNA biology and pathogenesis. I reflect on how these and other contributions (e.g., nucleocytoplasmic shuttling, telescripting) originated from curiosity-driven exploration and highly collaborative lab culture. The vast RNA and RBP assortments are beneficial, but increase complexity and chances of disorders, making the RNP sphere a rich source for future discoveries.
INTRODUCTION
The field of RNA and disease has come to include many different pathogenic phenomena linked to RNA biology. In living organisms, cells, and viruses, all RNAs exist and function as part of RNA–protein complexes (RNPs), with RNA-binding proteins (RBPs) playing crucial roles in every aspect of RNA life. Mutations or aberrations in both coding and noncoding RNAs, as well as in RBPs, are now recognized as the underlying causes of numerous diseases and developmental abnormalities. These alterations can disrupt RNA biogenesis, processing, localization, or stability, including transcription elongation and termination, splicing, and RNA modification. Pathogenic mechanisms also include nucleotide repeat expansions, imbalances in the expression or local abundance of RBPs and their cognate RNAs, and other RNP perturbations, all of which can have deleterious cellular outcomes.
Since the field's early days, in the 1990s, the universe of RNA- and RBP-related pathogenic aberrations has increased enormously, driven by remarkable technological advances. Innovations in RNA sequencing, bioinformatics, mass spectrometry, and methods for structure determination and prediction have vastly improved our understanding of RNA biology. Additionally, advances in genomics and medical genetics facilitated the discovery of associations between basic research and disease, accelerating the identification of therapeutic targets and the development of transformative therapeutic modalities such as gene therapy, gene editing, and antisense oligonucleotides. This progress represents the fruits of basic research, which remains essential for uncovering molecular mechanisms and a source of applications for treating RNA- and RBP-related diseases.
Here, I reflect on the discoveries made in my lab regarding the survival motor neurons (SMN) complex and its link to spinal muscular atrophy (SMA), as well as the identification of the fragile X mental retardation protein (FMRP) as an RBP. I focus on the lab's thinking stream, a crucial part behind research that tends to be left out of published papers to conform with journals’ guidelines, to illustrate the serendipitous origin of the earliest links between RBPs and diseases. These discoveries emerged from the lab's long-standing interest in heterogeneous nuclear ribonucleoprotein (hnRNP) proteins and RBPs in general, and they are products of the efforts and creativity of many lab members working as a team. To keep this piece concise, I reference and name only a few contributors as needed. This reflection is intentionally limited to only a few aspects of our work and is not intended as a review of the literature.
ORIGINS OF THE LINKS BETWEEN RBPs AND DISEASES
Since the early 1980s, my lab has maintained a strong interest in hnRNP proteins, the RBPs that associate with nascent transcripts, including pre-mRNAs and long noncoding RNAs, and mRNAs. It seemed reasonable to believe that they have crucial roles in mRNA biogenesis and functions, yet up to that time they remained enigmatic and controversial, and their sequences were unknown. Our work described and molecularly defined the major hnRNP proteins (hnRNPA-U) (Choi and Dreyfuss 1984; Dreyfuss et al. 1984; Choi et al. 1986; Piñol-Roma et al. 1988; Swanson and Dreyfuss 1988a) and mRNP proteins, particularly the poly(A)-binding protein (Adam and Dreyfuss 1987; Adam et al. 1986a,b), noting key features and characteristics, including the main RNA-binding domain (RBD/RRM) (Adam et al. 1986b; Nakagawa et al. 1986; Swanson et al. 1987), modular architecture, multi-RBDs, and auxiliary domains with simple composition (low complexity), such as RG/G-rich region (RGG box) (Ishikawa et al. 1993; Takimoto et al. 1993; Burd and Dreyfuss 1994). In addition to studying their roles in mRNA biogenesis and functions, for example, Choi et al. (1986), we have also been interested in hnRNPs’ cellular localization and dynamics, including the nucleo-cytoplasmic shuttling of RBPs, as it appeared to be important for mRNA transport and could change the distribution of splicing-regulating hnRNP proteins between the nucleus and cytoplasm and thus their stoichiometry in each cellular compartment. Biochemical experiments revealed that hnRNP proteins have RNA-binding specificity (Adam et al. 1986b; Swanson and Dreyfuss 1988a,b; Burd et al. 1989, 1991), RNA annealing activities (Portman and Dreyfuss 1994; Portman et al. 1997), and capacity to bring together disparate segments of RNA or bridge RNAs owing to their multi-RBD architecture and propensity for self-association (Burd et al. 1991; Matunis et al. 1992). They could thereby sculpt pre-mRNAs for splicing and other processing reactions and promote or hinder recruitment of spliceosomes and other trans-acting machineries to specific sites in cognate RNAs. The investigations on hnRNPs continued for decades and spawned many projects.
A new interest in the lab in the early 1990s was to determine the nuclear import and export signals (NLS and NES, respectively) in shuttling RBPs (Piñol-Roma and Dreyfuss 1992, 1993). For that, we initially focused on the highly abundant hnRNPA1, a prototype of the large A/B family members, including RBPs that have historically different names, such as TDP-43/TARDBP, and a common 2xRRM-RGG architecture (Nakagawa et al. 1986). They contained no typical positively charged NLS, the only type known at the time. The shuttling signals turned out to be in the middle of the C-terminal RGG domain, and we went on to identify the receptors (transportins) that mediate their translocations through nuclear pores (Michael et al. 1995, 1997; Siomi and Dreyfuss 1995; Nakielny et al. 1996; Pollard et al. 1996).
At the same time, many abundant hnRNP proteins remained to be characterized, notably hnRNPK and hnRNPU, neither of which had RBD/RRM consensus sequences. We investigated both proteins in parallel and as always in a highly interactive and collaborative spirit, which multiplies excitement and productivity. In every case, we cloned, sequenced, expressed recombinant proteins, studied RNA and protein binding, and produced monoclonal antibodies for immunoblotting, microscopy and to purify complexes. Both hnRNPU and hnRNPK gave us unexpected gifts. In hnRNPU, we found that it bound RNA via its C-terminal RGG region, providing evidence for its role as an RBD (Kiledjian and Dreyfuss 1992). HnRNPK and its related protein hnRNPJ, contained three ∼45 amino acid repeats, which were highly similar to sequences found in several nucleic acid-binding proteins across diverse organisms. This discovery defined a novel RBD that we named KH (K homology), after hnRNPK (Siomi et al. 1993a). In addition to KH, hnRNPK also had an RGG domain. The discoveries of signature motifs revealed hundreds of RBPs in every eukaryotic organism and impacted many fields as it pointed to their involvement in diverse biological processes and every aspect of the life of mRNAs. Nearly a thousand genes in humans have RBP signature motifs, highlighting the central role of posttranslational processes in gene expression regulation.
Follow-up studies unexpectedly connected the above discoveries to two neurological diseases, fragile X syndrome and SMA. Curiosity and a tradition of scholarship in the lab helped draw our attention to scientific news, even from unconventional sources like The New York Times, that we might not have noticed otherwise.
FRAGILE X SYNDROME, THE FMR1 GENE, AND FMRP
As the RGG story was emerging in the lab, an astute postdoctoral researcher spotted that the predicted protein (FMRP) of the newly identified fragile X mental retardation syndrome gene (FMR1), which had no recognizable motifs, contained an RGG region (Siomi et al. 1993b). A low-tech discovery by visual inspection, using colored pencils, when databases were thin, and search methods were rudimentary (in 1991). We produced monoclonal antibodies to FMRP and studied its expression (everywhere, not just in neurons), localization, and interactions. This led to the identification of FMRP homologs, FXR1 and FXR2. RNA-binding assays showed that FMRP could bind RNA. In fact, as the work on hnRNPK progressed in parallel, we noticed that FMRP had KH domains. By sheer luck, my good friend and Penn/HHMI colleague, Robert Nussbaum, was studying FMR1 pathogenic mutations. Establishing the first connection between RBP and disease, we showed that a patient with severe fragile X syndrome had a mutation in a KH domain of FMRP, which impaired its RNA-binding ability (Siomi et al. 1994). The closing sentence in the discussion of one of the papers reporting these findings summarized the significance and our fascination with this discovery: “These findings reduce at least one aspect of intelligence to a molecular issue of protein–RNA interaction” (Siomi et al. 1994). Digging deeper after FMRP functions would have required setting up animal models and neurobiology experiments in isolated neurons; however, other irresistible projects bloomed in the lab at the same time, and we chose to pursue them instead.
THE SMN COMPLEX AND SMA
The RGG box kept coming up in several contexts. It was widely present in RBPs, yet the picture of its roles seemed incomplete. By coincidence, we recalled an older report describing dimethylarginines in an abundant, nuclear, RBP in Physarum polycephalum and hnRNPA/B-type protein with similar migration in SDS-PAGE and overall amino acid composition in human cells (Christensen et al. 1977). While the role of methylated arginines in proteins was unknown, we reasoned that if this posttranslational modification occurred in RGG domains and we could recreate it, it could potentially affect interactions with RNA or other proteins. This possibility seemed worth exploring, which required purifying the protein arginine methyltransferase. There was little enthusiasm in the lab for doing that because there were almost no publications on protein arginine methylation, unlike protein phosphorylation, indicating no general interest (I thought that's great, actually).
Fortunately, a gifted and courageous graduate student took on the task, and we were able to label and detect dimethylarginine modifications in the RG-rich regions of hnRNPA/B proteins, hnRNPU and other proteins HeLa cells (Liu and Dreyfuss 1995). We could also carry out the reaction in cell extracts, which allowed us to purify the enzyme to ∼1000-fold; however, achieving higher levels of purification was challenging. As an additional approach, we tried interaction cloning using the extended RGG domain of hnRNPU by two-hybrid screening in yeast against a human cDNA library. Most of the clones that came up were from the same protein, an unknown protein with no sequence similarity to anything in the databases and no recognizable motifs. The recombinant protein, which we used to make monoclonal antibodies, had no methylation activity in vitro. Just in case that might lead to something, we decided to use the unknown protein for another two-hybrid screening. Alas, that attempt netted yet another unknown protein. We made monoclonal antibodies to that too but still no methylation activity. Going through phases in new projects where almost nothing was known was a familiar, even exciting experience when explorations uncovered great breakthroughs. Until that happens, it can be challenging to explain to a student's thesis committee the logic behind hunting for an enzyme that carries out a modification of unknown function in a novel protein domain, and persisting through an unknown-interacting protein of an unknown RGG-interacting protein. It was encouraging when experiments with the reagents we generated started yielding intriguing results. And then unexpected news came.
In January 1995, two back-to-back papers were published in Cell, each reporting the identification of the SMA gene by positional cloning. The studies pointed to two adjacent genes, on an inverted duplication unique to humans on Chromosome 5 at 5q13. One claimed that the culprit was a gene called SMN (Lefebvre et al. 1995) while the other proposed neuronal apoptosis inhibitory protein (NAIP) (Roy et al. 1995). An accompanying editorial (Lewin 1995) hailed the achievement and explained the decision to publish both studies noting they used adequate methodologies and emphasizing the significance of SMA—a devastating neurodegenerative disease and the most common hereditary cause of infant mortality. It was not immediately clear which gene was truly responsible. It did not take long for us to realize that the hnRNPU RGG-interacting protein that we had identified was the same as the one predicted by the SMN gene, the gene reported by Judith Melki's group of pediatric geneticists in Paris. I had not heard about SMA previously. The closest resemblance was the widely known ALS (amyotrophic lateral sclerosis), a late-onset, mostly sporadic adult motor neuron degeneration. In collaboration with Melki's team, we used the antibodies we had produced to investigate SMN. This revealed a deficit of SMN in motor neurons of deceased infants that correlated with SMA severity, adding compelling evidence that SMN is the disease-causing gene in SMA (Lefebvre et al. 1997).
In an instant, a connection was drawn between RBPs and disease. Our seemingly blind exploration acquired a framework and significance both on the basic science in the RNA biology sphere, and potential to advance knowledge on the pathogenesis and prospects of therapy for a severely debilitating and often lethal disease that had none. Thus, from the outset we knew a lot more about SMN than what the landmark paper described and had the tools to move forward. Clearly, SMA was not due to a protein specific to motor neurons—we discovered SMN in HeLa (cervical carcinoma) cells and could detect it in all cells we tested (Liu and Dreyfuss 1995). That meant that we could use any cell type as starting material for biochemistry. We already knew that SMN had some connection with RGG domains and was not a solitary protein, rather a component of a multiprotein complex. And we had its closest interactor already, which we named SIP1, for SMN interacting protein1, and later changed to Gemin2 (Liu and Dreyfuss 1996; Liu et al. 1997).
Aided by mass spectrometry, in collaboration with Matthias Mann's group, we isolated and sequenced cDNAs of additional SMN-associated proteins (Charroux et al. 1999, 2000; Baccon et al. 2002; Gubitz et al. 2002; Pellizzoni et al. 2002a). We expressed recombinant proteins, produced antibodies, and studied their protein–protein interactions, RNA-binding properties, and cellular localization. All showed the same pattern, cytoplasmic staining, and striking nuclear dots, that look like twins of snRNP-rich nuclear-coiled bodies (later renamed Cajal bodies after Ramon y Cajal, who first described them in neurons 90 years earlier). We named them Gems, for Gemini of Cajal bodies (and their brilliant appearance in immunofluorescence) (Liu and Dreyfuss 1996). SMN and its associated components (the SMN complex) showed the same pattern, and we named them Gemins 2–7. Two additional proteins, Gemin8 and Unrip, were subsequently discovered in the labs of former postdoctoral researchers, Utz Fischer and Livio Pellizzoni (Carissimi et al. 2005, 2006; Grimmler et al. 2005).
In two-dimensional (2D) gel electrophoresis, SMN complexes had an additional cluster of low-molecular mass, positively charged proteins (Liu et al. 1997). These proteins reminded us of the Sm core proteins of small nuclear ribonucleoproteins (snRNPs), which we had seen in many talks and papers in the splicing field. Western blots and mass spectrometry proved that to be the case; however, they were substoichiometric compared to Sm proteins in mature snRNPs. Each snRNP is comprised of a small noncoding nuclear RNA (∼100–300 nt in vertebrates, named U1, U2, U4, U5, U11, U12, U4atac), a seven-membered ring of Sm proteins, and additional snRNA-specific proteins. The snRNPs are the main subunits of spliceosomes, and some have additional roles critical to transcription regulation and cleavage and polyadenylation (U1 telescripting) (Kaida et al. 2010; Berg et al. 2012; So et al. 2019; Venters et al. 2019) and specialized mRNA processing (U7 in histone mRNAs 3′-end processing). The presence of Sm proteins prompted us to probe if SMN complexes also contained snRNAs, which they did. But why, and how did they bind? Below is a brief summary of what this project discovered.
We found that the SMN complex is essential for snRNP biogenesis (Fischer et al. 1997), mediating and conferring specificity to Sm core assembly exclusively on snRNAs at specific sequences known as Sm sites (Pellizzoni et al. 2002b). This chaperone activity was unanticipated because Sm cores can spontaneously form on uridine-rich sequences. However, this spontaneous assembly poses a risk: the formation of highly stable and long-lived Sm cores on non-snRNA substrates. The SMN complex mitigates this risk by orchestrating the assembly of Sm cores through defined components and interactions, mobilizing Sm proteins, other RBPs, and their cognate RNAs (So et al. 2016). It acts as a chaperone not only for snRNPs but for other RNPs as well. Using adaptors such as Gemin5, the complex binds snRNAs via snRNA-defining RNA elements (snRNP code) (Yong et al. 2004, 2010; Golembe et al. 2005; Battle et al. 2006). Additionally, the complex interacts with diverse other RNAs, and recruits many RBPs, which it recognizes by dimethylarginine-modified RGG domains (Friesen and Dreyfuss 2000). The core SMN complex does not bind RGGs well and has no methylase activity. That function is carried out by an associated complex, the methylosome that contains the protein arginine methyltransferase JBP1/PRMT5 (Friesen et al. 2001b). We were able to capture SMN by interaction cloning with hnRNPU RGG likely because it got methylated in yeast. RG-rich synthetic peptides bound SMN poorly, which made us use them to purify the methylosome. In contrast, SMN bound the same peptides when the arginines were first methylated to symmetric dimethylarginines (Friesen et al. 2001a,b). Thus, this work also revealed the function of symmetric dimethylarginine protein modification.
High-resolution structure of a key intermediate (Zhang et al. 2011), including Gemin2 bound to SMN domain with five Sm proteins, revealed how the complex achieves the architectural feat by prearranging Sm pentamers and preventing them from nonspecifically binding non-snRNAs. snRNA binding and further assembly require ATP hydrolysis that depends on Gemin3, a DEAD box motif protein (Charroux et al. 1999). Studies on the SMN complex revealed yet another protein fold that can bind RNA directly, WD repeats domain found in Gemin5 and other RBPs (Gubitz et al. 2002; Battle et al. 2006). Mutations in several Gemins have subsequently been linked to other diseases.
A hallmark of SMA pathology is unraveling of neuromuscular junction (NMJ), a synaptopathology detected in newborns of SMA mouse model. To investigate the potential mechanism behind this, we isolated and sequenced RNA from motor neurons at postnatal day 1, which revealed many splicing changes, including skipping of exons in agrin that are necessary for strengthening NMJs (Zhang et al. 2013) and could explain an important aspect of the disease phenotype. At later times (more advanced SMA) numerous splicing abnormalities as well as changes in snRNP stoichiometries are detected in all tissues (Zhang et al. 2008). This made us ask if the changes in splicing were a direct effect of SMN deficit or a result of the changes in the snRNP repertoire. To address that we systematically changed the abundance of the major spliceosomal snRNPs. These experiments led to the surprising discovery of telescripting, showing that full-length transcription in thousands of genes in vertebrate organisms requires U1 snRNP functions (in addition to its role in splicing) suppression of numerous polyadenylation signals that otherwise cause premature transcription termination (Kaida et al. 2010; Berg et al. 2012; So et al. 2019; Venters et al. 2019; Oh et al. 2020). This too has relevance to many disorders.
PERSPECTIVE
Cells use a vast assortment of RBPs and various RNAs organized as RNPs as the main tools for mediating and regulating the biogenesis of a great diversity of mRNAs and translate them to make the specific repertoire of proteins needed for cellular structures and functions. The numerous factors and complexity of RNP-mediated posttranscriptional processes provide enormous flexibility and opportunities to adapt to environmental changes and make multiple cell types. But they also increase risks of disorders, explaining the increasing number of diseases now known to be caused by mutations in RNAs and RBPs. Research on RBPs and RNPs has already provided important insights into the flow of genetic information, posttranscriptional gene expression, and disease pathogenesis and potential therapies.
In the examples I have described, as well as in the broader research conducted in my lab, new directions evolved from asking basic questions motivated by curiosity and fascination with unexpected observation, rather than adherence to predefined aims and hypothesis-driven frameworks. Explorations were made as a highly collaborative and interactive team, with a flexible mindset and willingness to shift priorities in pursuit of what seemed most interesting, not what was most familiar, feasible, or fashionable. This led to discoveries of novel molecules, processes that there was no reason to think were needed, and connections to diseases that we had not heard of before. In the same way, there is little doubt that the vital RNP space remains a rich source of many more exciting discoveries in the years ahead.
ACKNOWLEDGMENTS
I am extremely grateful to the many talented and inspiring graduate students and postdoctoral researchers in my lab that I have been fortunate to work with. This work was supported by a grant from the National Institutes of General Medical Sciences (R35 GM139646) and the Howard Hughes Medical Institute Investigator funding to G.D.
Footnotes
-
Article is online at http://www.rnajournal.org/cgi/doi/10.1261/rna.080353.124.
-
Freely available online through the RNA Open Access option.
This article, published in RNA, is available under a Creative Commons License (Attribution-NonCommercial 4.0 International), as described at http://creativecommons.org/licenses/by-nc/4.0/.
