This web page was produced as an assignment for an undergraduate course at Davidson College.
Programmable aligned riboregulators of eukaryotic gene expression (View abstract)
Travis S Bayer & Christina D Smolke
Nature Biotechnology Volume 23 Number 3 March 2005 337-343
This paper describes the design and manipulation of an antiswitch, which is a synthetic ligand controlled RNA construct capable of regulating gene expression. Travis Bayer and Christina Smolke have created an inducible antiswitch called s1, which when activated, possesses the ability to silence GFP expression in transgenic yeast ( Saccharomyces cerevisiae). The s1 antiswitch has an apatmer domain, which binds to the ligand/ effector theophylline, and an antisense domain, which is designed to base pair with target mRNA. In the absence of theophylline, the s1 antisense domain base pairs internally as an antisense stem. However, the conformation changes when s1 becomes activated by theophylline such that the antisense domain can silence GFP mRNA. In this paper, Bayer and Smolke explain the overall mechanism of antiswitches, specifically detail the sequence of s1 in its alternate "on and "off" conformations, and examine tweaked antiswitches s2, s3, s4, s5, s6, s6, and s7 that are predictably more or less effective than s1. They also create and test a suppressible antiswitch construct, s8. Finally, Bayer and Smolke demonstrate that two different antiswitches (s1 and s9) can be employed to effectively and specifically control the expression of two separate proteins (GFP and YFP).
Figures: Description of Data and Evaluation of Data
Part A is an illustration of an inducible antiswitch in alternate "off" and "on" states. When the effector/ligand is not bound, the inducible antiswitch is in an inactive state: the antisense RNA base pairs internally with other nucleotides in the antiswitch sequence forming a stem, while the aptamer domain remains unbound. Thus, the antisense portion of the antiswitch cannot bind to the target mRNA and expression of the target mRNA proceeds normally. On the other hand, when an effector is present, the antiswitch becomes activated. The effector binds to the antiswitch and changes its conformation such that the aptamer domain internally base pairs within the antiswitch sequence, while the antisense domain becomes free. Thus, the antisense domain can bind to its target mRNA and inhibit translation like RNAi. This is a simple cartoon, but highly effective. It clearly and quickly provides a basic overview of the design and mechanism of an antiswitch.
Part B of Figure 1 provides a detailed nucleotide map of the s1 antiswitch construct. It depicts the specific nucleotide sequence of s1 in the "off" conformation, in which the antisense stem is internally bound and the aptamer domain is free. It also depicts the conformation of the sequence in the presence of the effector, theophylline, in which the aptamer domain binds internally and the antisense domain becomes able to bind to the target mRNA translation initiation sequence. This figure is also useful, but it would be helpful for the authors to include a visualization of where/how the theophylline binds.
Part C is a graph of relative GFP expression at increasing effector concentrations. It shows very steady minimal GFP expression in the presence of antisense RNA and theophylline as a negative control (i.e. antisense RNA effectively silences GFP regardless of effector concentration as in other studies of RNAi (Fire A, et al., 1998)). As a positive control, this figure depicts a very steady high expression of GFP in the presence of the aptamer and increasing theophylline concentration (i.e. the ligand binding aptamer sequence has a minimal effect on GFP expression, regardless of effector concentration). As another positive control, this figure depicts a steady, slightly lower steady GFP expression in the presence of the s1 antiswitch and the incorrect aptamer, caffeine (i.e. s1 is able to reduce the GFP expression about 20% even in its inactivated form and increasing caffeine concentration has no observable effect on GFP expression). Finally, this figure shows that GFP production is slowly and steadily reduced with the addition of s1 and 0.01-0.07 mM of Theophylline. Furthermore, GFP production sharply stops in the presence of s1 and about 0.8 mM of theophylline (i.e. s1's silencing capacity of GFP depends on the concentration of theophylline present and there exists an in vivo critical threshold of 0.8 mM Theophylline to induce GFP expression silencing). This figure is very effective because it provides adequate controls and a striking result. Unlike the controls, which either allow or prevent GFP production, scientists can uniquely silence GFP expression by turning s1 "on" with the appropriate theophylline concentration.
Part D is a graphical comparison of the GFP expression vs. time in samples in which s1 is absent or present and to which theophylline is added after about 3 hours. The GFP expression over time is similarly increasing for both the s1 containing and s1 absent samples until the hour theophylline is added. GFP expression continues to increase, then peak, and finally decline around 6.5 hours for the s1-free sample after theophylline is added. In contrast, once the theophylline is added to the s1 containing sample, GFP expression steadily declines to about 0 RFU/OD. The authors explain that the GFP expression declines in the s1 containing sample at a level corresponding to the half-life of their particular GFP variant. Thus, in this sample, the yeast do not making any more GFP and GFP present degrades at its predicted rate. Panel D is a generally useful figure to show that theophylline activates s1 and s1 silences RNA immediately when it becomes activated. This figure suggests that GFP protein itself is not destroyed by s1, but new GFP expression ceases. However, neither the figure nor the authors explain why there is a slight decline in the GFP expression after 6.5 hours in the s1-free strain. Perhaps the yeast cells being monitored begin to die or unable to produce GFP after 6.5 hours. Furthermore, because GFP is the protein being monitored and the technology is available, it may have been useful to show images or create a movie of the yeast cells as GFP degrades. Cole N, et al. (1996) very effectively monitored protein production by creating movies of still images.Part E is an in vitro affinity assay. It attempts to demonstrate the conformational change of the antiswitch molecule in the presence of an appropriate concentration of theophylline concentration along with the target RNA. The authors explain that 5 nM amounts of both target mRNA and radiolabeled s1 exist in each lane, while the concentration of theophylline varied by lane from 0.2 to 200 micromolar. The bands are more like smears in each lane, but there appears to be a be a molecular weight shift between the 2 micromolar and 10 micromolar theophylline lanes. The bands in the 10, 20, and 200 micromolar lanes are broader indicating that some antiswitches have traveled less than most antiswitches in the 0.2 and 2 micromolar lanes (i.e. the radiolabeled antiswitches have higher molecular weight when when concentration of theophylline is at least 10 micromolar). It makes sense that there would be a shift in molecular weight of the antiswitch because figure 1c showed there was a critical concentration/binary action of theophylline required to activate s1 and silence GFP production. The authors argue that the shift in molecular weight exists because at the appropriate theophylline concentration, both the theophylline and the target mRNA can bind to the radiolabeled antiswitch, increasing its molecular weight. However, this figure is poorly presented and does not contain adequate controls. It would be interesting to calculate the individual molecular weights of s1, theophylline, and the target RNA, and see if the combined weight or a multiple of the weight equals the molecular weight of the band in question. However, there is no molecular weight marker lane and there is no lane with 0 micromolar concentration of theophylline. There is no definite indication which way the molecules traveled through the gel. Worst of all, there is no way to know conclusively whether the antiswitch conformation changed at high theophylline concentration because both the target molecular and theophylline bound. Perhaps, at 10, 20, and 200 micromolar concentrations simply one tier more theophylline bound to the antiswitch shifting the molecular weight. Indeed, this figure cannot definitively show that the theophylline at an appropriate concentration changes the conformation of antiswitch molecule by allowing target RNA to bind. More rigorous testing would be necessary to illustrate their claims. Finally, the authors only speculate that a lesser concentration is necessary to change the conformation of s1 in vitro than in the in vivo experiments because theophylline does not readily cross semispherical membranes. An experiment should be conducted to determine the exact intracellular concentration of theophylline necessary to activate s1.
Part A of Figure 2 compares the nucleotide sequence and free energy of s1, s2, s3, and s4 antiswitch constructs. S2 has one less internal base pairs in the antisense stem as compared to s1, making that stem portion less stable. The scientists predicted that s2 would be easier to induce/activate than s1. S3 contains five more internal base pairs in the antisense stem and three less nucleotides in the aptamer domain as compared to s1, making the s3 antiswitch more stable. Bayer and Smolke predicted that s3 would be more difficult to induce/active than s2 and s1. Finally, the s4 construct contains a larger loop in the antisense stem than s1, making it less stable than s1 and s2. The researchers anticipated that s3 would be easier to induce than s1 and s2. This figure is well presented and useful. However, the figure legend refers to s1, s2, s3, and s4 as "predicted structures," so sequencing testing should be conducted to confirm that the scientists test the expected antiswitches.
Part B shows the Relative GFP expression vs. increasing concentration of theophylline for the various antiswitches s1, s2, s3, s4. As predicted, s2 begins to silence GFP expression at a lower concentration theophylline than s1 because its antisense stem is less stable. Also, s4 silences GFP expression at an even slightly lower concentration of theophylline than s1 or s2. As expected, s4 inhibits GFP expression at a higher concentration of theophylline, because its antisense stem is more stable. This graph clearly demonstrates that Bayer and Smolke can predictably tune and rationally antiswitches. S1 activity vs. controls has been measured before in Figure 1, so no more controls are necessary. It is interesting to note that the GFP expression drops off most sharply in s1 and s3. Perhaps the drop offs appear less steep in GFP expression in s4 and s2 containing samples due to the logarithmic scale of the x axis.
Part C also clearly demonstrates that Smolke and Bayer can predictably tune antiswitches. This graph shows that s5 requires a higher concentration of theophylline to silence GFP expression than s1 because it possesses a lower affinity for theophylline. Furthermore, the authors take s5 and destabilize the the antisense stem by removing one base pair to create s6. As predicted, the graph illustrates that s6 antiswitch requires a lower theophylline concentration to silence GFP expression than s5, but a higher concentration theophylline than s1. Although not necessary, it would have been helpful to see results of the reverse of this experiment, i.e. what happens when the aptamer domain of an antiswitch has a higher affinity for the effector. Smolke and Bayer could make a domain with a higher affinity for theophylline and test the relative GFP expression while increasing concentrations of the theophylline. They would expect that GFP expression would be silenced at a lower concentration of theophylline than in the case of s1.
Part D shows the relative GFP expression vs. effector concentration for antiswitch constructs s1 and s7. The aptamer domain of s7 has been engineered to activate in the presence of tetracycline as opposed to theophylline that activates s1. Since the expression trends vs. effector concentration are similar, this graph indicates that GFP expression is reduced in a similar manner for both constructs as respective effector concentration is increased. Thus, Smolke and Bayer demonstrate that their findings about antiswitches are not limited to those controlled by theophylline. This figure is also useful and effective. In order to show that their findings are even more widely applicable, Smolke and Bayer could design and test antiswitches to affect proteins other than GFP or GFP variants.
Part A of this figure depicts the sequence and free energy of the suppressible antiswitch, s8. The s8 construct is "on"/ active with the antisense stem free to bind to target mRNA unless theophylline is present. When the theophylline effector is added, the conformation changes such that the aptamer domain binds to the antisense domain and prevents mRNA silencing. Thus, protein expression can be turned "on". This is an effective figure, very similar to ones presented before.
Part B is a graph of the relative GFP expression over increasing concentration of theophylline for samples containing either the s1 or s8 antiswitches. As before, in the s1 containing sample, the GFP expression is silenced at around 0.8 micromolar concentration of theophylline. In the s8 containing sample, however, GFP expression basically does not occur until 0.8 micromolar theophylline is added. At concentrations above 0.8 micromolar theophylline in the sample containing s8, GFP expression reaches high levels. This is one of the most striking figures of the paper because it demonstrates that the authors can not only tune an inducible antiswitch, but they can create a suppressible antiswitch. Smolke and Bayer have seemingly mastered tuning and designing antiswitches.
Part A is an illustration of the mechanism of two simultaneous inducible antiswitches that can control the expression of two separate proteins. One antiswitch (s1) is activated by theophylline and controls GFP expression. The other antiswitch (s9) is activated by tetracycline and controls YFP (yellow fluorescent protein) expression. This figure is logical, but not altogether necessary. The information it provides is very similar and only slightly expanded than figure 1a.
Part B shows the relative expression of GFP or YFP when s1 and s9 are present along with varying concentrations of effectors, theophylline and tetracycline. GFP and YFP expression is strong when no effectors are present. GFP expression is severely inhibited, while YFP expression remains high in the presence of 5 mM theophylline and 0 mM tetracycline. Alternatively, YFP expression is severely inhibited and GFP expression is unaffected in the presence of 0 mM theophylline and 5 mM tetracycline. Finally, both GFP and YFP expressions are silenced in the presence of 5 mM of theophylline and 5 mM of tetracycline. This is also a very striking figure because it clearly demonstrates that protein expression can be selectively suppressed by the activation of the appropriate antiswitch. This provides even more proof that Smolke and Bayer have excellent antiswitch technique and design.
Indeed, this paper is well written and the ideas are well conceived. It is mildly confusing that the authors simultaneously and interchangeably refer to theophylline, effectors, and ligands instead of simply using one or two terms. Smolke and Bayer generally presented figures and data that provide strong evidence to support their claims. My main critiques include: Figure 1e should be repeated to include more controls and include more expected cues for the reader, the authors should consider making movies demonstrating the changes in GFP expression, they might make an antiswitch with a higher affinity to be explored as an extension of Figure 2c, and Figure 4a is unnecessary.
It would be wise for scientists to explore how antiswitch molecules affect expression of proteins besides their target protein and even DNA sequences. Such questions could be tested by performing a BLAST search for any similar RNA sequences to the target. In order to detect non-target DNA sequences that would also bind to antiswitches experimentally, scientists could wash labeled activated antiswitches over a DNA microarray containing fragments of the entire genome of an organism. Any hybridization would indicate complementary DNA sequences to the antisense domain of the antiswitch and possible side effects to employing antiswitches in normal organisms (i.e. nontransgenic). Hopefully, antiswitch constructs only silence target mRNA and do not affect DNA, so no unexpected BLAST matches would occur and no antiswitch/DNA microarray hybridization would occur. However, if it was determined that non-target sequences did bind to antiswitches, longer antisense domains and increased effector affinity may need to be created to ensure specificity.
Future research should also involve testing antiswitches to control proteins other than GFP or its variants. GFP is not a normal yeast protein and therefore, the antiswitches cannot be generalized necessarily to control protein expression in natural protein of non-transgenic organisms. As a first step, scientists might attempt to silence a natural protein in yeast or to silence GFP expression in jelly fish that naturally produce GFP with antiswitches. Furthermore, scientist could explore antiswitch protein control in higher eukaryotic organisms (i.e. can antiswitch technology be applied in worms, mice, humans or other species?) Scientists could employ similar methods assays to those described by this paper. Organisms could be engineered to produce GFP and antiswitch constructs could be inserted into the cells to induce or suppress GFP expression using viral vectors, artificial chromosomes, or microinjection. Certainly, because antiswitches work well in yeast, it is possible that antiswitch technology could be applied to many species.
Bayer and Smolke propose several possible future research directions at the conclusion of this paper. They suggest that antiswitches could be applied as gene therapy. Scientists could attempt to induce or suppress the expression of proteins to alter the progression of diseases with antiswitches. Researchers would need to make antisense domains that correspond to mRNAs of proteins of interest in higher eukaryotes and find methods to get the antiswitches into cells (microinjection, viral vectors). Additionally, Bayer and Smolke assert that antiswitch technology could be used in the emerging field of synthetic biology. Antiswitches could be added to pathways to regulate synthesized processes in bacteria. Bacteria could be engineered to perform processes in unison, such as blinking or moving. Beginning, middle, and end functions in such pathways could theoretically be controlled by humans by means of suppressible or inducible antiswitches and their appropriate effectors.
To learn more about synthetic biology and RNAi, please visit:
The MIT synthetic biology main page
Cole N, et al. 1996. Diffusional mobility of Golgi proteins of membranes of living cells. Science 273: 797-801.
Fire A, et al. 1998 Feb 19. Potent and specific genetic interference by double-stranded RNA in Caenorhabditis elegans. Nature 391: 806-811.
Jackie's Main Page
Molecular Biology Main Page
Davidson College Biology Main Page