Multi-step recognition of potential 5' splice sites by the Saccharomyces cerevisiae U1 snRNP

  1. Sarah R Hansen
  2. David S White
  3. Mark Scalf
  4. Ivan R Corrêa
  5. Lloyd M Smith
  6. Aaron A Hoskins

  • Curated by eLife

    eLife logo

    Evaluation Summary:

    This study extends previous work from the same group on the mechanism of 5' splice site recognition using co-localization single-molecule spectroscopy. There are three important conclusions: 1) the association of the U1 snRNP with the 5' splice site is largely determined by the snRNP itself and does not require other splicing factors; 2) sequence features of the 5' splice site determine whether a short-lived complex with U1 dissociates or transitions into a longer-lived, "productive" complex, potentially mediated by stabilized contacts with U1 associated proteins; and 3) the ability to form the longer-lived complex cannot be accurately predicted by base-pairing potential alone, as presumed by many predictive algorithms. Currently, a test for the role of specific protein-RNA contacts is lacking; additionally, a comparison with other nucleic acid recognition events is missing, particularly those also showing a two-step binding mechanism. This work will be of interest to colleagues in the splicing field as well as to others in fields where nucleic acid recognition by snRNPs plays a major role.

    (This preprint has been reviewed by eLife. We include the public reviews from the reviewers here; the authors also receive private feedback with suggested changes to the manuscript. Reviewer #1 and Reviewer #2 agreed to share their names with the authors.)

Reviewed by

Read the full article See related articles

Listed in

Log in to save this article

Abstract

In eukaryotes, splice sites define the introns of pre-mRNAs and must be recognized and excised with nucleotide precision by the spliceosome to make the correct mRNA product. In one of the earliest steps of spliceosome assembly, the U1 small nuclear ribonucleoprotein (snRNP) recognizes the 5' splice site (5' SS) through a combination of base pairing, protein-RNA contacts, and interactions with other splicing factors. Previous studies investigating the mechanisms of 5' SS recognition have largely been done in vivo or in cellular extracts where the U1/5' SS interaction is difficult to deconvolute from the effects of trans -acting factors or RNA structure. In this work we used colocalization single-molecule spectroscopy (CoSMoS) to elucidate the pathway of 5' SS selection by purified yeast U1 snRNP. We determined that U1 reversibly selects 5' SS in a sequence-dependent, two-step mechanism. A kinetic selection scheme enforces pairing at particular positions rather than overall duplex stability to achieve long-lived U1 binding. Our results provide a kinetic basis for how U1 may rapidly surveil nascent transcripts for 5' SS and preferentially accumulate at these sequences rather than on close cognates.

Article activity feed

  1. Version published to 10.7554/elife.70534 on eLife
  2. eLife

    Author Response

    Reviewer 1

    In this manuscript, Hansen and coworkers make use of the powerful, single-molecule assay CoSMoS to study the recognition of the 5' splice site by the U1 snRNP. Specifically, they investigate how 5' splice site oligos interacts with purified U1 snRNP to isolate 5' splice sitebinding from other factors, including the CBC, BBP, and any other factors in whole cell extract that may impact binding; previous studies have investigated binding in vivo or in cellular extracts or with limited quantitative capabilities. The authors find evidence for a reversible, two-step, binding reaction in which a short-lived interaction precedes a longlived interaction and in which binding depends on the 5' splice site sequence and the 5' end of U1. The data further suggests a compelling kinetic framework for how U1 surveys nascent transcripts for a bona fide 5'SS; specifically, both authentic and inauthentic 5' splice sites form the short-lived complexes but whereas the inauthentic complex preferentially dissociates, the authentic complex preferentially proceeds to a stable complex. Using oligos with different mutations to limit base-pairing they find that at least six potential base-pairs are required for association but that a stretch of seven base-pairs, with a maximum of one mismatch, is required for the long-lived interaction, with residues near the 5' splice site playing more important roles and with length being a stronger predictor of complex lifetime than thermodynamics, with implications for splice site predictions.

    The work focuses on the determinants and mechanism of the first and a pivotal step in splicing, in a manner that completes recent structural advances. The work extends findings presented in a previous publication from the lab (Larson and Hoskins, 2017) studying binding of U1 snRNP to the 5' splice site in extract. In that study, the authors provided early evidence of two-step U1 snRNP binding in the absence of the cap binding complex or the branch point binding protein, with a more stable state following a weaker state; although factors in the extract may have influenced binding, the results are not qualitatively different here. The authors also showed some evidence in the previous study that longer binding depended on crossing a threshold and did not increase further with greater stabilization. Still, this new work is of high quality with conclusions justified by the data and of significant interest to the splicing field and of general interest to those investigating binding of snRNPs to nucleic acid.

    Specific Points:

    1. To test and define the role of protein in the snRNP, the authors need to investigate the roles of Yhc1 and Luc7 in 5' splice site binding in this assay, particularly with respect to defining the basis of asymmetry and snRNP destasbilization.

    See Reviewer Comment #1.

    1. The similarity or difference of the two-step recognition mechanism described here to the recognition mechanisms of other nucleic acids by other RNP complexes is unclear. The authors need to put their findings into a larger context, relating their findings to studies of analogous systems described in the literature.

    See Reviewer Comments #2 and #4

    1. It is important that the authors address whether they can rule out that the exclusively long-lived complexes skip the short-lived conformation.

    See Reviewer Comment #5. Overall, a model with reversible connection between the unbound state and the long-lived bound state (U->B*) is less likely to explain our data.

    1. Given the co-transcriptional nature of many splicing events, the authors should discuss how recruitment by RNAP II might impact the two-step process. For example, fast dissociation by short duplexes might be countered by retention of U1 locally via RNP II.

    See Reviewer Comment #4.

    Reviewer 2

    In this work, the authors use co-localization single-molecule spectroscopy (CoSMoS) to dissect the sequence-directed nature of pre-mRNA 5' splice site recognition by U1 snRNP using purified, surface-tethered U1 snRNP complexes and truncated substrate RNA oligonucleotides containing the 5'splice site (SS) consensus sequence. The senior author previously has extensively published on related findings using the CoSMoS approach (PMIDs 23569281, 24075986, 27244240), and the current work is a logical extension. Here the authors find that the U1 snRNP reversibly selects a suitable 5' SS in a sequencedependent, two-step mechanism. They derive a kinetic selection scheme that suggests initial base pairing at particular positions, followed by a commitment to a longer-lived complex that enters the chosen 5' SS into the splicing cycle. This type of scheme is widespread among nucleic acid-binding enzymes and sometimes referred to "conformational proofreading". The work could be further strengthened by making more connections to existing kinetic selection schemes for other enzymes.

    In the following, major suggestions for improvements are summarized.

    1. The model described in the paragraphs starting with line 262 through 280 to interpret the observation of long and short complex lifetimes is not entirely clear. There are at least two potential models that can be considered to fit the observations: a linear and a circular model. A linear model would be one where U1 and substrate RNA are not associated (state 1), then they partially associate (state 2), and finally they isomerize to the completely associated/fully hybridized complex (state 3). The circular model is the same, except that it would additionally allow switching between states 1 and 3 directly (bypassing the partially associated state). To differentiate between these two scenarios, the authors would have to vary the concentration of the RNA probe and see if there is a uniform change in a single kon rate or if two kon rates start to appear. These rate subpopulations would be much easier to detect by fitting with hidden Markov models. It would seem unjustified to decide between these two models without obtaining such additional supporting data.

    See Reviewer Comment #5.

    1. In the section describing U1/5'SS duplexes destabilization in U1 snRNP (line 281) an underlying assumption is that the binding of two RNAs (in the absence of the spliceosomal proteins) would share the same characteristics or trends as two identical RNAs incorporated into the U1 snRNP. While this may be a rhetorical device to increase the clarity/connection between the concepts of predicted binding free energies and the residence time of hybridized oligonucleotides, it does not address the possible reasons for the discrepancy observed in RNA oligonucleotide versus U1 snRNP binding. The authors should point to a reference and derive a physical model from the available cryo-EM structures to show that the U1 snRNA is, most likely, being constrained by its associated proteins in such a way that it increases the binding affinity to complementary RNA oligonucleotides.

    See Reviewer Comment #2

    1. While the two-factor authentication metaphor of Figure 7 is charming, it seems off-topic. Instead, the authors should review the literature for examples of short, exploratory binding events involving an RNA:protein complex, followed by more stable, accommodated binding events, see e.g., the work by Sarah Woodson on 30S ribosomal subunit assemble and on Hfq function, work on kinetic proofreading of the ribosome, work on Cas9-based recognition of its target site, and many others. A potential descriptive framework to be used here is that of "conformational proofreading".

    See Reviewer Comment #4.

    1. There is significant concern that the single molecule sampling rate used to acquire the CoSMoS data is too slow to accurately measure the shortest lifetimes observed, which are only ~10 seconds long. According to the Nyquist sampling criterion, the sampling rate needs to be (at least) twice the frequency of the event being measured, implying that the authors cannot meaningfully observe any lifetime shorter than ~10 seconds given their limited sampling rate. Further considering that at minimum two consecutive data points are needed for observing a 10 second lifetime, artifacts (e.g., camera noise) could make up a disproportionate amount of the signal observed in their data for these short lifetimes. For an accurate measurement, the authors need to repeat the experiments at a higher sampling rate to make sure that there are no faster, transient interactions than those currently reported, and that the values reported are accurate.

    See Reviewer Comment #6.

    1. The authors have chosen to extrapolate rates via exponential fitting to dwell time distributions. This is a reductive approach that ignores the relationship between consecutive events. It is strongly recommended that the authors consider using a hidden Markov modeling (HMM) approach instead. HMMs have long become the gold standard in single molecule biophysics. Even better, a Bayesian approach could help analyze entire datasets at the same time. In this reviewer's opinion, the ebFRET software package from the Gonzalez lab at Columbia University could, for example, work well here.

    See Reviewer Comment #7

    1. The manuscript would be majorly strengthened if the authors were testing their hypothesis that Yhc and Luc7 contribute to U1 snRNA:5'SS stabilization, by generating (e.g., temperature sensitive) mutant strains that allow them to interfere with this function of the two proteins, either in purified U1snRNPs or whole cell extracts. Alternatively, the authors could choose to test the role of trans-activing factors such as BBP/Mud 2. Without such data, and given the extensive work the authors have previously performed to already demonstrate that U1 snRNP binds to a 5'SS reversibly, with fast and slow dissociation events, one can argue that the current work falls somewhat short in providing major new biological insights. More generally, the plethora of recent cryo-EM structures gives a wonderful opportunity to ask incisive mechanistic questions, which the authors do not fully leverage.

    See Reviewer Comment #1.

    Reviewer 3

    This study of U1 snRNP interaction with the 5'ss is an interesting and exciting piece of work. In particular, the data support two important conclusions of general importance to the field: 1) the association of the U1 snRNP with the 5'ss is largely determined by the snRNP itself and does not require other splicing factors and 2) the ability to form "productive" (i.e. longlived) interactions between the U1 snRNP and the 5'ss cannot be accurately predicted by base-pairing potential alone. This second point is particularly important as many algorithms for predicting splicing efficiency are based on base-pairing strength between the U1 snRNA and the 5'ss sequence. The data immediately suggest two additional questions.

    1. The authors repeatedly speculate that the benefit of basepairing toward the 3' end is due to the activity of Yhc1. If this model is true, these 3' end basepairs should not influence binding for a U1 snRNP with a mutant Yhc1. Since the authors have used mutant Yhc1 in other studies it seems possible to test this prediction.

    See Reviewer Comment #1.

    1. Since splice sites are often "found" in the context of alternative or pseudo/near-cognate splice sites, it would be interesting to know how the "rules" identified in the experiments presented in this study influence splice site competition and whether both the short- and longlived states are subject to competition or, rather, only the short-lived complexes. Is it possible to repeat the CoSMoS experiment with two oligomer sequences of different colors?

    See Reviewer Comment #3.

    1. Finally, the authors should say more about the particular requirement for basepairing at position 6, especially in the context of the experiments in Figure 5. This is particularly striking as this position is not well conserved in natural 5'ss, at least compared to position 5.

    See Reviewer Comment #8

  3. eLife

    Evaluation Summary:

    This study extends previous work from the same group on the mechanism of 5' splice site recognition using co-localization single-molecule spectroscopy. There are three important conclusions: 1) the association of the U1 snRNP with the 5' splice site is largely determined by the snRNP itself and does not require other splicing factors; 2) sequence features of the 5' splice site determine whether a short-lived complex with U1 dissociates or transitions into a longer-lived, "productive" complex, potentially mediated by stabilized contacts with U1 associated proteins; and 3) the ability to form the longer-lived complex cannot be accurately predicted by base-pairing potential alone, as presumed by many predictive algorithms. Currently, a test for the role of specific protein-RNA contacts is lacking; additionally, a comparison with other nucleic acid recognition events is missing, particularly those also showing a two-step binding mechanism. This work will be of interest to colleagues in the splicing field as well as to others in fields where nucleic acid recognition by snRNPs plays a major role.

    (This preprint has been reviewed by eLife. We include the public reviews from the reviewers here; the authors also receive private feedback with suggested changes to the manuscript. Reviewer #1 and Reviewer #2 agreed to share their names with the authors.)

  4. eLife

    Reviewer #1 (Public Review):

    In this manuscript, Hansen and coworkers make use of the powerful, single-molecule assay CoSMoS to study the recognition of the 5' splice site by the U1 snRNP. Specifically, they investigate how 5' splice site oligos interacts with purified U1 snRNP to isolate 5' splice site-binding from other factors, including the CBC, BBP, and any other factors in whole cell extract that may impact binding; previous studies have investigated binding in vivo or in cellular extracts or with limited quantitative capabilities. The authors find evidence for a reversible, two-step, binding reaction in which a short-lived interaction precedes a long-lived interaction and in which binding depends on the 5' splice site sequence and the 5' end of U1. The data further suggests a compelling kinetic framework for how U1 surveys nascent transcripts for a bona fide 5'SS; specifically, both authentic and inauthentic 5' splice sites form the short-lived complexes but whereas the inauthentic complex preferentially dissociates, the authentic complex preferentially proceeds to a stable complex. Using oligos with different mutations to limit base-pairing they find that at least six potential base-pairs are required for association but that a stretch of seven base-pairs, with a maximum of one mismatch, is required for the long-lived interaction, with residues near the 5' splice site playing more important roles and with length being a stronger predictor of complex lifetime than thermodynamics, with implications for splice site predictions.

    The work focuses on the determinants and mechanism of the first and a pivotal step in splicing, in a manner that completes recent structural advances. The work extends findings presented in a previous publication from the lab (Larson and Hoskins, 2017) studying binding of U1 snRNP to the 5' splice site in extract. In that study, the authors provided early evidence of two-step U1 snRNP binding in the absence of the cap binding complex or the branch point binding protein, with a more stable state following a weaker state; although factors in the extract may have influenced binding, the results are not qualitatively different here. The authors also showed some evidence in the previous study that longer binding depended on crossing a threshold and did not increase further with greater stabilization. Still, this new work is of high quality with conclusions justified by the data and of significant interest to the splicing field and of general interest to those investigating binding of snRNPs to nucleic acid.

    Specific Points:

    1. To test and define the role of protein in the snRNP, the authors need to investigate the roles of Yhc1 and Luc7 in 5' splice site binding in this assay, particularly with respect to defining the basis of asymmetry and snRNP destasbilization.

    2. The similarity or difference of the two-step recognition mechanism described here to the recognition mechanisms of other nucleic acids by other RNP complexes is unclear. The authors need to put their findings into a larger context, relating their findings to studies of analogous systems described in the literature.

    3. It is important that the authors address whether they can rule out that the exclusively long-lived complexes skip the short-lived conformation.

    4. Given the co-transcriptional nature of many splicing events, the authors should discuss how recruitment by RNAP II might impact the two-step process. For example, fast dissociation by short duplexes might be countered by retention of U1 locally via RNP II.

  5. eLife

    Reviewer #2 (Public Review):

    In this work, the authors use co-localization single-molecule spectroscopy (CoSMoS) to dissect the sequence-directed nature of pre-mRNA 5' splice site recognition by U1 snRNP using purified, surface-tethered U1 snRNP complexes and truncated substrate RNA oligonucleotides containing the 5'splice site (SS) consensus sequence. The senior author previously has extensively published on related findings using the CoSMoS approach (PMIDs 23569281, 24075986, 27244240), and the current work is a logical extension. Here the authors find that the U1 snRNP reversibly selects a suitable 5' SS in a sequence-dependent, two-step mechanism. They derive a kinetic selection scheme that suggests initial base pairing at particular positions, followed by a commitment to a longer-lived complex that enters the chosen 5' SS into the splicing cycle. This type of scheme is widespread among nucleic acid-binding enzymes and sometimes referred to "conformational proofreading". The work could be further strengthened by making more connections to existing kinetic selection schemes for other enzymes.

    In the following, major suggestions for improvements are summarized.

    1. The model described in the paragraphs starting with line 262 through 280 to interpret the observation of long and short complex lifetimes is not entirely clear. There are at least two potential models that can be considered to fit the observations: a linear and a circular model. A linear model would be one where U1 and substrate RNA are not associated (state 1), then they partially associate (state 2), and finally they isomerize to the completely associated/fully hybridized complex (state 3). The circular model is the same, except that it would additionally allow switching between states 1 and 3 directly (bypassing the partially associated state). To differentiate between these two scenarios, the authors would have to vary the concentration of the RNA probe and see if there is a uniform change in a single kon rate or if two kon rates start to appear. These rate subpopulations would be much easier to detect by fitting with hidden Markov models. It would seem unjustified to decide between these two models without obtaining such additional supporting data.

    2. In the section describing U1/5'SS duplexes destabilization in U1 snRNP (line 281) an underlying assumption is that the binding of two RNAs (in the absence of the spliceosomal proteins) would share the same characteristics or trends as two identical RNAs incorporated into the U1 snRNP. While this may be a rhetorical device to increase the clarity/connection between the concepts of predicted binding free energies and the residence time of hybridized oligonucleotides, it does not address the possible reasons for the discrepancy observed in RNA oligonucleotide versus U1 snRNP binding. The authors should point to a reference and derive a physical model from the available cryo-EM structures to show that the U1 snRNA is, most likely, being constrained by its associated proteins in such a way that it increases the binding affinity to complementary RNA oligonucleotides.

    3. While the two-factor authentication metaphor of Figure 7 is charming, it seems off-topic. Instead, the authors should review the literature for examples of short, exploratory binding events involving an RNA:protein complex, followed by more stable, accommodated binding events, see e.g., the work by Sarah Woodson on 30S ribosomal subunit assemble and on Hfq function, work on kinetic proofreading of the ribosome, work on Cas9-based recognition of its target site, and many others. A potential descriptive framework to be used here is that of "conformational proofreading".

    4. There is significant concern that the single molecule sampling rate used to acquire the CoSMoS data is too slow to accurately measure the shortest lifetimes observed, which are only ~10 seconds long. According to the Nyquist sampling criterion, the sampling rate needs to be (at least) twice the frequency of the event being measured, implying that the authors cannot meaningfully observe any lifetime shorter than ~10 seconds given their limited sampling rate. Further considering that at minimum two consecutive data points are needed for observing a 10 second lifetime, artifacts (e.g., camera noise) could make up a disproportionate amount of the signal observed in their data for these short lifetimes. For an accurate measurement, the authors need to repeat the experiments at a higher sampling rate to make sure that there are no faster, transient interactions than those currently reported, and that the values reported are accurate.

    5. The authors have chosen to extrapolate rates via exponential fitting to dwell time distributions. This is a reductive approach that ignores the relationship between consecutive events. It is strongly recommended that the authors consider using a hidden Markov modeling (HMM) approach instead. HMMs have long become the gold standard in single molecule biophysics. Even better, a Bayesian approach could help analyze entire datasets at the same time. In this reviewer's opinion, the ebFRET software package from the Gonzalez lab at Columbia University could, for example, work well here.

    6. The manuscript would be majorly strengthened if the authors were testing their hypothesis that Yhc and Luc7 contribute to U1 snRNA:5'SS stabilization, by generating (e.g., temperature sensitive) mutant strains that allow them to interfere with this function of the two proteins, either in purified U1snRNPs or whole cell extracts. Alternatively, the authors could choose to test the role of trans-activing factors such as BBP/Mud 2. Without such data, and given the extensive work the authors have previously performed to already demonstrate that U1 snRNP binds to a 5'SS reversibly, with fast and slow dissociation events, one can argue that the current work falls somewhat short in providing major new biological insights. More generally, the plethora of recent cryo-EM structures gives a wonderful opportunity to ask incisive mechanistic questions, which the authors do not fully leverage.

  6. eLife

    Reviewer #3 (Public Review):

    This study of U1 snRNP interaction with the 5'ss is an interesting and exciting piece of work. In particular, the data support two important conclusions of general importance to the field: 1) the association of the U1 snRNP with the 5'ss is largely determined by the snRNP itself and does not require other splicing factors and 2) the ability to form "productive" (i.e. long-lived) interactions between the U1 snRNP and the 5'ss cannot be accurately predicted by base-pairing potential alone. This second point is particularly important as many algorithms for predicting splicing efficiency are based on base-pairing strength between the U1 snRNA and the 5'ss sequence. The data immediately suggest two additional questions.

    1. The authors repeatedly speculate that the benefit of basepairing toward the 3' end is due to the activity of Yhc1. If this model is true, these 3' end basepairs should not influence binding for a U1 snRNP with a mutant Yhc1. Since the authors have used mutant Yhc1 in other studies it seems possible to test this prediction.

    2. Since splice sites are often "found" in the context of alternative or pseudo/near-cognate splice sites, it would be interesting to know how the "rules" identified in the experiments presented in this study influence splice site competition and whether both the short- and long-lived states are subject to competition or, rather, only the short-lived complexes. Is it possible to repeat the CoSMoS experiment with two oligomer sequences of different colors?

    3. Finally, the authors should say more about the particular requirement for basepairing at position 6, especially in the context of the experiments in Figure 5. This is particularly striking as this position is not well conserved in natural 5'ss, at least compared to position 5.

  7. Version published to 10.1101/2021.05.18.443434v1 on bioRxiv