You'd Prefer An Argonaute

Rapid Construction of Empirical RNA Fitness Landscapes

Jason N. Pitt and Adrian R. Ferré-D’Amaré

Science Vol. 330, 376 – 379, 15 October 2010.
DOI: 10.1126/science.1192001

This week’s instructive summary and analysis by Xuebing Wu:

This paper is one of the newest examples of how next-generation sequencing technology is enabling us to do experiements previously not possible. Combining sequencing with in vitro selection, Pitt and Ferré-D’Amaré demonstrate that it’s now possible to measure fitness for millions of genotypes and build an empirical fitness landscape.

A fitness landscape is a map that connects genotypes and the fitness of their corresponding phenotypes, and it has been proposed that evolution is an adaptive walk through such a landscape, toward higher fitness. It would be cool to actually see such a landscape to decipher whether there are, for example, multiple optimal states in the path , and how evolution jumps from one state to the next. Constructing such a landscape, however, presents a huge challenge. For example, for a simple 20mer RNA molecule, there are roughly 10²⁰ genotypes whose fitness you would need to measure to make a fitness landscape. More than 20 years ago we already had the technology to generate such vast genotypic space, by either chemically synthesizing DNA oligos, or large-scale mutagenesis of a template. So the real challenge is efficiently measuring the fitness for each genotype. “Fitness” itself is easily generalized–depending on the system and phenotype you are studying, fitness may be defined in different ways. It is, however, a widely accepted notion that in evolution, or population genetics, when there is selection pressure, the frequencies of genotypes with higher fitness increase. Therefore upon selection, an increase in genotype frequency can be used as a surrogate for fitness. This translates the problem of measuring the fitness of each genotype to measuring the frequency of each genotype, now feasible with next-generation sequencing technology.

The authors tested this idea using the Class II ligase, a ribozyme that catalyzes the ligation of its 5’ end to a substrate sequence. The authors had a couple reasons for choosing this molecule: it is short enough that its full length can be sequenced with high quality; and it is highly active, almost optimal in terms of catalytic activity such that its peak in a fitness landscape should be clear. Interestingly, this ligase (also known as “a4-11”) was isolated in David Bartel’s lab about 15 years ago, through multiple rounds of in vitro selection from a pool of totally random sequences. So the goal in the present study was to construct an empirical fitness landscape for this RNA ligase.

Recall that the change in genotype frequencies before and after in vitro selection serves as a measure of fitness. To select sequences capable of performing ligation, the authors incubated a pool of random RNA sequences with substrates covalently attached to beads, so that molecules with more ligase activity are more easily ligated to the substrates immobilized to beads. These selected RNAs are then reverse transcribed, PCR amplified, and sequenced.

The authors showed that selection enriches for sequences similar to the a4-11 wild-type sequence, and that sequences which come out earlier in their “serial depletion” are more biochemically active, which is not surprising. By creating ~160 single point mutants of the ligase and measuring their activity, they also showed that the frequency of genotypes in the selected pool was positively correlated with experimentally measured rate constants. This observation supports the use of genotype frequency as a surrogate for fitness. However, in my opinion, it would have been better if they had also shown that the change in genotype frequency upon selection positively correlates with measured rate constants, since it’s the change in frequency, not the frequency itself, which indicates fitness.

Overall, I think this is an interesting paper. The fitness landscape yields much more information than could be obtained from traditional in vitro selection experiments. However, outside of the ribozyme field, I don’t see clearly how such landscape can be used.

Role of a ribosome-associated E3 ubiquitin ligase in protein quality control

Mario H. Bengtson & Claudio A. P. Joazeiro

Nature 467, 470–473, 23 September 2010.
doi:10.1038/nature09371

This week’s insightful summary and analysis by David Weinberg:

In their 2010 Nature paper, Bengtson & Joazeiro demonstrate that proteins being translated from non-stop mRNAs are targeted to the proteasome by the E3 ubiquitin ligase Ltn1.  A non-stop mRNA is defined as any mRNA that lacks a stop codon that is recognized by the ribosome.  Natural causes of non-stop mRNAs may include mutations in termination codons (at the DNA level or due to transcription errors), readthrough of bona fide termination codons, premature or alternative cleavage and polyadenylation within coding regions, or the initiation of 3′-5′ mRNA decay on messages being translated.  Because a non-stop mRNA can only be recognized as such after at least one round of translation, non-stop protein products are necessarily generated from non-stop mRNAs.  The translation of non-stop mRNAs is problematic for the cell because it results in the production of aberrant proteins and, perhaps more importantly, sequesters ribosomes as a result of the failure to recruit release factors.  While the quality control pathway that recognizes non-stop mRNAs in eukaryotes has come into focus over the past decade, the details of how non-stop protein products are recognized and degraded has been relatively under-studied.  Here, the authors synthesize previously-published results with their own observations to provide the first comprehensive picture of the non-stop protein decay pathway in eukaryotes.

The story began in 2009 when Joazeiro’s lab identified the E3 ubiquitin ligase Listerin in a forward genetics screen for neurodegeneration in mice.  Seeking to gain insight into the cellular function of Listerin, the authors turned to its homolog Ltn1 in the budding yeast S. cerevisiae.  Since Ltn1 had been previously pulled out in a yeast genetic screen for non-stop decay genes, Bengtson & Joazeiro go after the precise role of Ltn1 in this relatively-uncharacterized pathway.  After verifying previously-published results that convincingly demonstrate a role for Ltn1 in the quality control of non-stop proteins, the authors go one step further and implicate its ability to specifically bind to and ubiquitinate non-stop proteins as an essential part of this pathway.

Although the E2-binding RING domain in Ltn1 is shown to be required for its function in non-stop protein decay, the identity of the E2 binding partner is not addressed here.  A traditional pulse-chase experiment is used to show that Ltn1 promotes the turnover of newly-synthesized non-stop proteins, which raises the question of how non-stop proteins are recognized as aberrant and thereby targeted for degradation.  A hint (or perhaps even the answer) came from previously-published observations that hard-coding 12 Lys residues in an otherwise-normal protein causes instability and that long tracts of Lys or Arg cause translational arrest (presumably due to electrostatic interactions between the nascent peptide and the ribosome exit tunnel).  If a ribosome were to translate through the 3′-UTR of a non-stop mRNA and reach the poly-A tail, translation through the poly-A tail would naturally generate a C-terminal poly-Lys tract in the protein product that might similarly stall translation.  Indeed, the authors show that hard-coding Lys residues recruits Ltn1 and leads to ubiquitination.  Intriguingly, products associated with apparently-stalled ribosomes are specifically targeted to Ltn1, while the protein products from ribosomes that efficiently translate through the poly-K tract are not.  This suggests that the translational stall, rather than the poly-Lys tract, is the signal for Ltn1 recruitment.  Unfortunately the authors don’t address perhaps the most interesting follow-up question here:  Does any translational stall (e.g., one caused by stretches of rare codons) trigger Ltn1-dependent ubiquitination, or is it somehow specific for non-stop proteins?  Aside from identifying the poly-K tract as sufficient for Ltn1 recruitment, no additional insight is provided into how this is accomplished at the molecular level.  Additional experiments show that the nascent non-stop protein is associated with ribosomes and, moreover, that Ltn1 itself is predominantly associated with ribosomes.

The paper concludes with an attempt to demonstrate the biological relevance of Ltn1 by identifying a phenotype in the ltn1 knock-out strain.  While the strain shows no growth defect in standard media, the addition of either an antibiotic or nonsense suppressor mutation – both which would facilitate stop codon readthrough – reveals a slow-growth phenotype for the knock-out strain.  Thus, the authors conclude that Ltn1 confers resistance to stress caused by the production of non-stop proteins, but it is unclear if the slow-growth phenotypes are due to the accumulation of the non-stop proteins themselves or the depletion of translation-competent ribosomes.

In my opinion, the most interesting aspect of the pathway characterized in this paper is how it compares to the analogous pathway used by prokaryotes.  The ssrA/tmRNA pathway in prokaryotes similarly depends on the tagging of stalled nascent polypeptides and their subsequent degradation by energy-dependent proteases.  However, in the case of prokaryotes – whose mRNA lack poly-A tails – the tagging sequence is provided in trans by a tmRNA molecule that recognizes ribosomes that have reached the end of an mRNA.  In contrast, eukaryotes appear to take advantage of the existing poly-A tail to accomplish a similar feat without the need for a trans-acting factor.  Interestingly, the tmRNA includes a stop codon that triggers translation termination of non-stop messages, while the eukaryotic pathway appears to never ‘officially’ terminate translation.  This key difference perhaps warrants further investigation, as it seems unlikely that eukaryotes would altogether bypass a requirement for translation termination to recycle ribosomes from non-stop mRNAs.

While many questions remain – including how this function for Ltn1 is related to the neurodegeneration phenotype observed in Listerin mutant mice – this paper provides a satisfying initial characterization of the eukaryotic non-stop protein decay pathway, albeit with the help of many previously-published results and limited novel insight.

Quality control by the ribosome following peptide bond formation

Hani S. Zaher & Rachel Green

Nature Vol 457, 8 January 2009.
doi:10.1038/nature07582

This week’s cogent summary and analysis by Josh Arribere:

The authors initiate the paper with a discussion of known quality control mechanisms in protein synthesis. They present the overall rate in vivo as being in the range of 6e-4 and 5e-3, and state that their own in vitro measurements of fidelity are in the range of 1e-4 and 2e-3. From the overlap of these two ranges, it is not readily apparent that a new quality control mechanism need exist, but the true motivation for the study becomes apparent shortly. In the process of making an oligopeptide in vitro, the authors failed, and instead observed a miscoding event that led to premature termination.

It is from this observation the authors begin the paper. They demonstrate that although the rate of RF2-stimulated hydrolysis (release of the nascent peptide) is comparable for correct vs. miscoded events (fig 1c), the K_m is ~10 fold less (fig 1d). Such a difference is surprising given the one base pair change between the two constructs. Furthermore, they demonstrate that the miscoded construct is subject to RF2-mediated release (increase in the rate of hydrolysis, fig 1f), albeit inefficiently, even when the A site lacks a stop codon. The K_m is also decreased in the miscoded event, leading to an overall ~300 fold increase in the second order rate constant (fig S4). Different mismatch events and A-site codons argue that the observed phenomenon is not a peculiarity of their original construct (fig 1f). Moreover, of all the mismatches in the P-site, one is tolerated, namely, the G:U wobble base pair in the third position, as to be expected given the degenerate nature of the genetic code (fig S8). Thus the proposed mechanisms are compatible with known biology.

Primer extension assays demonstrate the ribosome has not shifted frame (fig S5), and although P-site tRNA dropoff is increased in the miscoded case, the rate of dropoff is ~2.5 fold slower than RF2-mediated release. 2.5 fold is a rather small gap, and the rate of RF2-mediated release is still two orders of magnitude slower than the rate of elongation. However, upon addition of RF3 (a class II release factor) and RF2, the rate of release increases another 10 fold (fig 2). This puts release following a miscoding event on par with the rate of chain termination, but still slower than elongation (~2/sec). So what happens if the tRNA beats the RFs to the ribosome? The rate of peptidyl transfer is not inhibited (fig 3a black bars), though the ribosome has ~10x diminished capacity to correctly incorporate the next amino acid following a miscoding event (fig 3a white bars). Examining the nature of the peptides formed reveals predominantly the correct tripeptide product in the correctly coded case. However, multiple seemingly random tripeptide products are formed following a P-site mismatch (fig 3b). Thus a single mismatch in the P site leads to a general loss of ribosome fidelity.

One of the consequences of these multiple miscoding events is a further stimulation of release by RF2 and RF3 (fig 4b,c,d). Peculiarly, an E-site mismatch alone does not stimulate release (except in the “buffer-dependent” instance of fig4b), but does stimulate release together with a P-site mismatch. This begs the question: how is an E-site mismatch only sensed together with a P-site mismatch and not by itself? Frame maintenance is somewhat compromised only in the doubly miscoded case (fig S13). Of interest, the only case where an E-site mismatch alone led to stimulation of hydrolysis (MNKF, fig 4b), also exhibits an abnormal primer-extension banding pattern similar to the doubly miscoded event (fig S13b, compare last two lanes). At any rate with this further RF2/3-stimulated increase in release for the doubly miscoded event, the rate of peptide hydrolysis (~1/sec) is now on par with elongation (~2/sec), making it a kinetically viable pathway in protein synthesis.

All of the above observations, together with the rate constants and concentrations of protein synthesis factors, are incorporated into a model (fig 5a). Testing the model with a S100 extract (supernatant of a 100,000g cell lysate) confirmed some of the predictions of the model. Following an initial miscoding event, the next correct amino acid is added ~30% of the time, and subsequently a relatively low loss of yield for this product is observed (fig 5c 3^rd columns). Since the doubly miscoded event contains multiple species (many AAs possible), it is not possible to measure the “Incorrect PT” arrow in fig 5a with the TLC assay (and the authors note this). One confusing point is the apparent increase in MN-matched formation between the di- and tripeptide (fig 5c 2^nd column of di-, tripeptide), though the authors do not comment on this.

The authors come an incredibly long way from a failed experiment (oligopeptide production) to discover proofreading by the ribosome. They keep an eye on rate constants to demonstrate the phenomena they are studying are kinetically relevant. Different in vitro translation labs each favor particular buffer systems (buffers A, B, C, D in this paper), and the authors quell arguments by repeating some of their observations in multiple buffer systems (for instance, fig S15). This is important since each buffer seems to have its own peculiarities (see fig 4b, S1, S3, S11), and it is not readily apparent what this means, nor which buffer, if any, is the “correct” one.

There are many future directions for further study, and some are mentioned in the paper. One that was not mentioned is the fate of the released peptide. A miscoded, truncated peptide is a potential dominant negative nightmare for the cell. Clearly there must be a tight coupling in the cell between peptide release and degradation. Subsequent unpublished experiments have shown that the mRNA is destabilized following miscoding, though I do not know about the fate of the nascent peptide. It would be very interesting to know what discerns miscoding and RF2/3 stimulated release from normal stop-codon mediated release.

Genome-wide measurement of RNA secondary structure in yeast

Michael Kertesz, Yue Wan, Elad Mazor, John L. Rinn, Robert C. Nutter, Howard Y. Chang & Eran Segal

Nature Vol 467, 2 September 2010.
doi:10.1038/nature09322

This week’s summary and analysis by David Garcia:

In contrast to experimental methods for probing RNA secondary structure such as footprinting or SHAPE, the novel method described in this paper, called PARS (Parallel Analysis of RNA Structure), offers a significant advancement: the ability to work on a grand scale. The authors applied PARS to thousands of mRNAs simultaneously from S. cerevisiae, but the technique could in theory be applied to any population of RNAs for which sequence is known, and which can be selected and folded in vitro.

That the technique analyzes RNAs folded in vitro is a valid concern, as we might not want to get too excited about the fidelity of mRNA structure formed in a test tube versus how it actually happens, probably co-transcriptionally, in the cell, especially on this scale. But to my knowledge, all other currently available methods for analyzing RNA secondary structure are in vitro too. And it doesn’t exclude the authors from noting some interesting similarities to a published in vivo ribosome profiling dataset. When a full-blown genome-wide in vivo structure approach arrives, PARS data will be a useful comparison as well.

At the core of the method is detection of which nucleotides in RNAs are either paired or unpaired, to reveal a picture (relatively low resolution in this iteration) of secondary structure on a genome-wide level. It relies on the different specificities of two nucleases, RNase V1 which preferentially cleaves phosphodiester bonds 3’ of double-stranded RNA, and S1 nuclease which preferentially cleaves phosphodiester bonds 3’ of single-stranded RNA. The authors subjected a pool of poly-A selected yeast mRNA to either enzyme, followed by base hydrolysis mediated random fragmentation to generate smaller molecules amenable to cloning and sequencing by SOLiD. After aligning the reads, they produced profiles for each RNase that, based on where and how frequently reads clustered along an mRNA in either the V1 or S1 libraries, represent which portions of the RNA were double or single stranded. A ratio of signals from each library is expressed in the PARS score (log₂ ratio of V1 over S1), such that a larger/positive score represents a more double-stranded region, a smaller/negative score more single-stranded.

Now the first issue to be raised is that they did not perform a minus nuclease negative control, as is standard in footprinting experiments. This would help reveal how much of their library results from endogenous degradation products (or during cell lysis) which have 5’-phosphoryl ends and make it through selection. While this “contamination” is probably small, the control seems basic to me. On the plus side they did check for several other biases in their method, but I won’t go into detail here.

Next they compared PARS and traditional footprinting profiles for several endogenous mRNAs, as well as other RNAs they spiked into their library (domains from HOTAIR and the Tetrahymena group I ribozyme). They see strong overlap between the profiles. They also show strong agreement between PARS scores and known secondary structures for a few well-characterized domains of endogenous mRNAs. This data represents a convincing proof of principle, and now the task is, of course, to see if there’s a tangible way to assess PARS’s accuracy throughout a large dataset.

While they saw an overall strong correlation between PARS and Vienna scores (predicted double-stranded probability), even when they looked at only nucleotides with very strong PARS scores (high or low), a little less than half in each set could still fill out the entire distribution of Vienna scores, meaning a decent fraction were contradictory. It’s hard to conclude too much from these apples and oranges comparisons, but hopefully the two methods will be complementary in many cases, as the authors stress.

Using their PARS dataset, they highlight five global properties of yeast mRNAs. Number one: based on PARS scores, the CDS was more structured than UTRs. I found this to be quite intriguing, and perhaps I haven’t appreciated how intrinsic structure is to the sequence of raw nucleotides, absent of proteins. Unfortunately, what they did not address with this result is how much structural differences relate to sequence composition differences between the UTRs and the CDS. Since UTRs are more AU rich, could this explain the result? Or what fraction does it not explain? I realize this gets into a kind of a chicken and egg debate, because it has been shown by many that the CDS and UTRs differ in numerous ways, which are likely highly intercorrelated, and so one cannot really say what is controlling what. Still, I think this should have been checked.

Finding number two: when they looked at average PARS scores along the CDS (not in the UTRs), they saw the strongest periodic signal in 3-nt cycles, with the first position of each codon scoring the lowest average PARS score. They also saw a strong correlation between the amplitude of this 3-nt cycle and translational efficiency, as measured by average ribosome occupancy from Ingolia et al. Thus this cycle could in some way facilitate ribosome translocation, and messages that utilize it most effectively are rewarded with increased translation. It’s an interesting observation made by linking an in vivo and in vitro dataset. The system seems all so intelligently designed.

Finding number three: a small anti-correlation between mRNA structure around the translation start site and translation efficiency (again, via ribosome density from Ingolia et al.). It was clearest when the authors clustered subsets of messages into groups where the average PARS scores where distinct. Finding number four, which the authors describe as a “rich picture of biological coordination,” didn’t make much sense to me, it involved GO analysis. Maybe it was too rich for me.

Their last finding was that transcripts that encode signal peptides had less structure in portions of the 5’ UTR and the first ~30 nucleotides in the CDS compared to non-signal peptide encoding transcripts. They might have checked to see whether this effect was due in part to the sequence/codon constraints in these regions required to code for the signal peptide itself.

PARS should be a highly useful method for probing RNA structure on a genome-wide scale. While this study has nucleotide resolution, it’s low, and so better suited for systematic analysis rather than molecule-by-molecule structure determination. More controls, testing conditions, and deeper sequencing will reveal more. In the absence of any directly comparable dataset, the authors present some intriguing similarities to the Ingolia et al. dataset, implying that a measureable fraction of RNA function in vivo is inherent to sequence itself, perhaps no big surprise, but cool to ponder nonetheless. The findings could have benefited from more computational rigor, with respect to sequence constraints that may partly explain structural differences.

The in vivo main course could take a while–snack judiciously on PARS in the meantime.

An Allosteric Self-Splicing Ribozyme Triggered by a Bacterial Second Messenger

Elaine R. Lee, Jenny L. Baker, Zasha Weinberg, Narasimhan Sudarsan, Ronald R. Breaker

Science Vol. 329. no. 5993, pp. 845 – 848, 13 August 2010.
DOI: 10.1126/science.1190713

This week’s methodical summary and analysis by Alex Subtelny:

From the lab that discovered riboswitches comes this paper, which describes a bacterial riboswitch that allosterically controls the self-splicing of a ribozyme located immediately downstream. This unusual tandem arrangement was discovered upstream of a putative C. difficile virulence gene (CD3246) during a computational search for new riboswitches, including those for cyclic di-guanosyl 5’-monophosphate (c-di-GMP), an important bacterial second messenger that regulates the transition between motile and biofilm states. Interestingly, the riboswitch in question was located far (~600 nucleotides) upstream of its associated ORF and appeared to lack the typical expression structures associated with riboswitches. Instead, the intervening sequence between the riboswitch and the ORF contained what looked like a group I ribozyme. This raised two intriguing possibilities: i) that the c-di-GMP aptamer allosterically regulates self-splicing of the ribozyme, and ii) that unlike most group I ribozymes, which are part of selfish genetic elements, this one might perform a beneficial function for its host.

The authors first demonstrate that the putative riboswitch aptamer indeed binds c-di-GMP with high affinity and specificity. Then, they dissect the mechanism of the tandem riboswitch-ribozyme through a beautiful series of in vitro experiments with mutants that disrupt or restore key secondary structure elements. Binding of c-di-GMP to the aptamer stabilizes a base-pairing architecture that favors splicing of the region upstream of the ribozyme (the 5’ exon) to the region downstream (the 3’ exon), which contains the ORF for the virulence gene. In the absence of the ligand, a different base-pairing structure is favored, leading to the formation of an alternative excision product consisting of a fragment of the 3’ exon. The authors support their splicing assays with kinetic experiments showing that c-di-GMP causes a ~12-fold increase in the rate of 5’-3’ spliced product formation and a modest decrease in the rate of formation of the alternative 3’ excision product. Finally, the authors present an elegantly convincing model to explain how alternative processing of the mRNA might affect the expression of the virulence gene. 5’-3’ splicing, which is favored in the presence of c-di-GMP, generates a ribosome binding site situated an optimal distance from the start codon, which in the precursor mRNA is concealed by being part of a stem-loop. In contrast, the alternative 3’ excision product lacks a ribosome binding site (since only five nucleotides are left upstream of the start codon), preventing translation of the downstream ORF. Thus, according to this model, the mRNA for the virulence gene is competent for translation only in the presence of c-di-GMP.

While the authors do an excellent job of showing that c-di-GMP regulates alternative ribozyme self-splicing in vitro and present a highly plausible model for how this might regulate virulence gene expression, they stop there. They provide little evidence to support the in vivo relevance of the riboswitch-regulated ribozyme, and, in particular, to show that it performs a beneficial function for the host. In one of their supplemental figures, the authors show that the major RT-PCR product for CD3246 (using primers corresponding to the aptamer and the interior of the ORF) is 5’-3’ spliced, and that the extent of splicing increases with culture age, which is associated with an increased concentration of c-di-GMP. However, they do not show that 5’-3’ splicing results in increased protein output. This could conceivably be accomplished by placing the riboswitch-ribozyme (or mutants thereof) upstream of a reporter gene, introducing this fusion into their C. difficile strain or another bacterial species, and measuring levels of the reporter normalized to another, control reporter. Moreover, the authors do not address in the paper the (rather unlikely but) possible existence of alternative transcriptional start sites within the body of the riboswitch-ribozyme that, if highly used, might call into question the relevance of their model for the translational regulation of CD3246 expression. In addition, we are left with several other key questions: what is the function of CD3246? And why is it important for its expression to be regulated by c-di-GMP? Insight into these questions, as well as those discussed earlier, would strengthen the authors’ hypothesis that group I ribozymes can be co-opted into performing beneficial functions for their hosts.