Embed Size (px)
RNA Second Messengers andRiboswitches: Relics from the RNAWorld?Some riboswitches and their ligands may be relics of signaling networksused when organisms relied on RNA instead of DNA and proteins
Ronald R. Breaker
Compelling evidence supports thehypothesis that modern cells de-scended from organisms whose com-ponents were made entirely of RNA.Scientists who helped formulate this
“RNA World” hypothesis for the evolution oflife have identified many characteristics of mod-ern cells that strongly suggest RNA once ruledthe planet. For example, the existence of self-cleaving and self-splicing ribozymes that cut andligate RNAs demonstrates that at least some
RNA-processing reactions can be catalyzedwithout the need for proteins. Also, the exis-tence of ribosomal RNAs that build all geneti-cally encoded proteins emphasizes the fact thattoday’s organisms depend on RNA to carry outfundamental biochemical processes. Even manynucleotide-like coenzymes that are near univer-sal in all life forms likely first appeared in RNAWorld organisms. The RNA World hypothesisis appealing because it helps explain the arcanecharacteristics of genetic and biochemical pro-
cesses in modern cells. Why do all cells useribosomal RNAs, and not protein enzymes,to stitch together amino acids to makepolypeptides? Most likely, the ribosome’speptidyl transferase center is an evolution-ary descendent of an ancient ribozyme thatfirst catalyzed this reaction before proteinenzymes evolved. Why do so manycoenzymes, such as adenosylcobalamin(AdoCbl), thiamin pyrophosphate (TPP),flavin mononucleotide (FMN), S-adenosyl-methionine (SAM), molybdenum cofactor(Moco), and tetrahydrofolate (THF), allcarry recognizable fragments of RNA nucle-otides or derive from nucleotide precursors?Perhaps these compounds served as coen-zymes for metabolic ribozymes of the RNAWorld, and they have been preserved aslegacies from a time before protein enzymesexisted.
The study of RNAs and RNA-derivedsmall molecules may shed some light on thefunctions that RNA carried out before pro-teins emerged. We can use such insights toseek functional RNAs in modern cells that
• Modern cells are widely believed to have de-scended from RNA World organisms that usedfunctional RNA molecules to catalyze reactionsand sense various metabolites.
• Riboswitches are structured RNA domainspresent in the mRNAs of many bacterial speciesthat sense small molecules and control the ex-pression of associated genes.
• Some modern riboswitches may have an RNAWorld heritage, and therefore may represent thetypes of structures and functions that were use-ful for ancient life forms.
• The bacterial second messenger c-di-GMP issensed by riboswitches, and this partnershipbetween a small RNA signaling molecule and alarge RNA aptamer may constitute one of theearliest signaling networks.
• Additional second messenger and riboswitchinteractions likely remain to be discovered, andthese may expand our understanding of bothmodern and ancient forms of signaling withnucleic acids.
Ronald R. Breakeris Henry Ford IIProfessor in theDepartment of Mo-lecular, Cellular andDevelopmental Biol-ogy; Department ofMolecular Biophys-ics and Biochemis-try; and is an Inves-tigator with theHoward HughesMedical Institute,Yale University,New Haven, Conn.
Volume 5, Number 1, 2010 / Microbe Y 13
remain hidden, or to better harness the power ofstructured RNAs through molecular engineer-ing efforts. Here I focus on the possibility thatsome natural metabolite-sensing RNAs calledriboswitches may be primordial in origin. Anewfound class of riboswitches that sense thebacterial second messenger cyclic-diguanylate-monophosphate (c-di-GMP) could be a modernmanifestation of an RNA World signaling part-nership.
Riboswitches and the RNA World
Simple bacterial riboswitches are usually foundin the 5� untranslated region (UTR) of a messen-ger RNA, where they form a single RNAaptamer that selectively binds a target metabo-lite. Metabolite binding favors a folding change
in the adjoining nucleotides, called theexpression platform, which changesthe amount of protein produced fromthe downstream open reading frame(ORF). Riboswitches usually exploitmetabolite-regulated transcription ter-mination or translation initiation toregulate ORF expression. However,bacteria (and some eukaryotes) also useother mechanisms such as mRNA self-cleavage, transcriptional interference,and alternative splicing.
There are numerous distinct classesof riboswitches (Fig. 1); the list of com-pounds that are sensed by these isnearly as long. What is it about a givenriboswitch class that may implyan ancient origin? Certainly, not all ex-pression platform mechanisms existedin a purely RNA World organism. In-trinsic transcription terminators, com-prised of a strong stem followed byseveral U residues, interact with RNApolymerase to end transcription. How-ever, protein-based RNA polymeraseenzymes would not have existed untilorganisms began to produce polypep-tides.
In contrast, the aptamers of manyriboswitches are strikingly widespreadand well conserved among numerousbacteria. Horizontal transfer could ex-plain some of this distribution and ap-parent conservation, but it seems morelikely that at least a few riboswitcheswere present very early in the evolution-
ary radiation of bacterial species. Indeed, TPPriboswitches are exceptionally common in bac-teria, and representatives are found in at leastsome organisms from all three domains of life.This class is a candidate for an RNA Worldriboswitch, since its distribution suggests that itmay have been present in the last common an-cestor of all life.
TPP riboswitches also sense a coenzyme thatmay have been used by primordial ribozymes.All of the putative RNA World coenzymes havemodern riboswitch partners that sense and re-spond to their changing cellular concentrations.None of these riboswitches is assuredly a directdescendant from the RNA World, but their ex-istence in modern cells shows that RNA se-quence-space could have offered sensing and
F I G U R E 1
Validated riboswitch classes are plotted in rank order relative to the number of knownrepresentatives. Riboswitch classes are named for the ligand they sense. The search forriboswitch representatives was conducted with RefSeq version 32, which has more than5.1 � 10�9 nucleotides and includes 764 bacteria whose genomes have been completelysequenced.
14 Y Microbe / Volume 5, Number 1, 2010
Breaker: from Chicken Breeding to an Interest in Very Ancient BiologyAs a youth, Ronald Breakerearned a dubious reputation fromhis informal efforts to breedchickens. “I knew I’d pushed mygenetics tinkering too far whenone year my father—seeing theseason’s young and very strangelooking roosters stroll by—de-cided I needed to stop my experi-ments,” he says. “Each new gen-eration of roosters sprouted evermore interesting tufts of featherson their bodies, and more elabo-rate combs on their heads. At theend of this evolutionary line was arooster that was very large in size,visually terrifying, and [with] anasty disposition that causedother small farm animals to fleefrom its path.”
The line became extinct at hisfather’s insistence, and by hispending departure for college.“Certainly I was doing whatfarmers have done for 10,000years—although my selectivebreeding work was . . . not at allfor increased crop yields or forimproved domestication traits,”Breaker says. Today, his studiesare considerably less frightening,with potentially more valuableapplications.
Breaker and his collaboratorsat Yale University, where he is aprofessor of Molecular, Cellular,and Developmental Biology andan investigator for the HowardHughes Medical Institute, arestudying natural and engineerednucleic acids. “In some ways,members of my laboratory arelike molecular archaeologists,” hesays. “We are digging around theunexplored regions of genomes,looking for hidden treasures, orfor clues about the nature ofearth’s earliest organisms. I . . .am intrigued by the likelihoodthat traces of our earliest ances-
tors are retained as useful compo-nents in modern cells.”
Breaker, 45, was drawn to thisresearch because of his longstand-ing interest in the origins of life.“One of the most compelling pro-posals, called the RNA World hy-pothesis, argues that the earliestlife forms were made entirely ofRNA,” he says. “If . . . true, thenRNA must be able to carry out farmore biological functions thanthose we know exist from studieson modern organisms.”
One goal is to determinewhether RNA molecules act asmolecular sensors and switches.“RNA World creatures must havehad the ability to sense and re-spond to their surroundings, andso I felt this was not a high-riskproject,” he says. “When we vali-dated the existence of the first ri-boswitches, we recognized thatthey control genes whose func-tions are sometimes essential forthe survival of bacteria. There-fore, new forms of antibiotics canbe created that target ribo-switches and kill disease-causingbacteria.” Recently, he and hisresearch colleagues foundedBioRelix, a New Haven biotechcompany that aims to develop justsuch antimicrobial compounds.
Breaker grew up in central Wis-consin, the only scientist amongtwo brothers and a sister. His twobrothers work in constructionand manufacturing, and his sisteris a traveling nurse who providesin-home care for the elderly. Hecomes from several generations ofdairy farmers and lumberjacks.His father still owns the familyfarm, but gave up dairy herds togrow only crops. His mother re-cently retired from her job as ateller and loan officer for the localbank.
Breaker was a rare teenager
who preferred farm work to par-tying. “I was surrounded by biol-ogy—the seasonal cycle of cropplanting, growth and harvest, andthe frequent births of calves orhatching chickens,” he says. “Asa high school student, I wouldsometimes volunteer to assist acow giving birth, rather thanspend a Saturday night in town.My friends put a stop to this bydriving out to the farm and insist-ing I get in the car.”
He received his B.S. degree in1987 from the University of Wis-consin, Stevens Point, where hismajor was biology and his minor,chemistry. He earned a doctoratein biochemistry in 1992 from Pur-due University, and did researchas a postdoctoral fellow at theScripps Research Institute in LaJolla, Calif. from 1992–1995. Hejoined the Yale faculty in 1995.
He and his wife Michelle, amath instructor at Gateway Com-munity College, have two chil-dren. Their daughter is a fresh-man in the honors program at theUniversity of Connecticut, withan interest in archaeology and bi-ology. Their son is a high schoolsophomore, and a pitcher on thebaseball team. Breaker also is in-terested in that sport, and helpscoach Little League baseball. Healso enjoys visits to Wisconsin tosee friends, fish for trout, and pur-sue his interests as an amateurgeologist. “I’ve begun to learn alittle about the geology of glaciersthat have left so much evidence oftheir existence in Wisconsin, in-cluding dropping billions ofstones on farm fields—stones thatwe needed to remove by hand be-fore each year’s planting,” he says.
Marlene CimonsMarlene Cimons lives and writes in
Volume 5, Number 1, 2010 / Microbe Y 15
switching functions to the most primitive organ-isms. Moreover, not all presumably ancientcompounds sensed by riboswitches are coen-zymes, as exemplified by the curious case ofc-di-GMP.
Riboswitches That Sense aBacterial Second Messenger
Bacteria usually do not come to mind whenthinking of organisms with diverse signalingcompounds and pathways. However, quorumsensing discoveries in recent years revealed nu-merous compounds that are involved in bacte-rial cell-to-cell communication. Moreover, nu-cleotide-based compounds such as cAMP andppGpp serve as bacterial signaling compoundsin a manner similar to eukaryotic second mes-sengers. One of the most widespread bacterialsecond messengers, first encountered in the mid-1980s, is c-di-GMP (Fig. 2). Diguanylate cyclaseenzymes use two GTP molecules and two phos-phoester transfer reactions to form the circularRNA dinucleotide c-di-GMP linked via two3�,5�-phosphodiester bonds.
Homologs of the cyclase and nuclease en-zymes can be found in a staggering diversity ofboth gram-positive and gram-negative bacteria,suggesting that c-di-GMP is an exceptionallywidespread second messenger. This distribu-tion implicates c-di-GMP as a key regulator ofcellular physiology since a very early stage inbacterial evolution. The mechanisms by whichthis second messenger effects wide-ranging
physiological changes remained puz-zling for more than two decades. Al-though protein binding and regulationwere identified, it was unclear howchanging c-di-GMP concentrationscontrols genes. At least part of this mys-tery can be explained by the discoveryof a new class of riboswitches that se-lectively binds c-di-GMP (Fig. 3).
As with many other riboswitches, thec-di-GMP aptamer domain has se-quence and structural features that arehighly conserved across a great diver-sity of bacteria (Fig. 3A). Each aptamerfolds to form a three-stem junction,with conserved nucleotides sprinkledthroughout the structure. An atomic-resolution structure (Fig. 3B) revealsthat some of these conserved nucleo-
tides serve as key components that hold thethree-dimensional structure together. For exam-ple, the loop of stem P2 almost always forms a“GNRA” tetraloop (where R is either G or A),which commonly docks with a tetraloop recep-tor located near the loop of the adjacent P3stem. These same two stems are also held closelyby a surprising long-distance base pair formedby strictly conserved C and G nucleotides frombulges in the P2 and P3 stems, respectively.
These structural features reside at a distancefrom the c-di-GMP binding pocket, which isformed by conserved nucleotides in the regionsthat join the stems together. Since c-di-GMP isa circular version of an RNA dinucleotide, wespeculated that riboswitch aptamers simply useWatson-Crick base pairing to recognize both Gresidues of the ligand. However, there werenot two strictly conserved C residues that couldbase pair with c-di-GMP. The atomic-resolutionstructure again provided the answer—twoseemingly imperfectly conserved nucleotidesand a perfectly conserved A nucleotide form thehighly selective binding pocket for c-di-GMP.
Watson-Crick base pairing does occur be-tween one G of the ligand and a C residue of theaptamer (position 92 in Fig. 3B). Although thenucleotide at this position appears to be mutatedin rare instances, in such cases a C residue usu-ally resides at a nearby location. This C likelymoves into the binding site to retain ligand-aptamer base pairing. The other G residue of theligand is usually recognized by a nonstandardbase pair with a G from the aptamer (position
F I G U R E 2
Chemical structure of c-di-GMP. The guanyl base, ribosyl and phosphatyl moieties for thetwo 5�-guanosine monophosphate substructures are similarly colored.
16 Y Microbe / Volume 5, Number 1, 2010
20 in Fig. 3B), although sometimes the apta-mer nucleotide can be an A residue. Thestrictly conserved A residue (position 47 inFig. 3B) intercalates between the two bases ofc-di-GMP to help form a continuous base-stacking architecture involving aptamer andligand nucleotides. Several additional ligandcontacts are made by aptamer ribose residuesor by divalent metal ions, which complete thelocal structure of the binding pocket. Suchmixtures of familiar and unpredictable waysin which RNA can fold to grip a ligand arefound with other riboswitch classes as well,indicating that RNA has a robust collection offolding tricks to form precision binding pock-ets for many different metabolites.
There are extensive differences in the struc-tures of c-di-GMP aptamers when probed inthe absence or presence of c-di-GMP. The atom-
ic-resolution structural model reveals that c-di-GMP binding stabilizes an otherwise weak P1stem, which is characteristic of many other ri-boswitch classes. Usually, some nucleotides in-volved in P1 formation could bind nucleotides inthe expression platform to form alternativestructures, such as terminator or antiterminatorstems that influence gene expression.
The mutually exclusive formation of ligand-bound aptamer versus antiterminator structureis a mechanism proposed for a c-di-GMP ribos-witch from the bacterial pathogen Clostridiumdifficile that terminates transcription when c-di-GMP levels are elevated (Fig. 4A). This ribos-witch controls expression of a large operon thatcodes for many of the proteins needed to buildflagellar organelles. The ligand-triggered tran-scription termination mechanism would helpexplain how increasing c-di-GMP concentra-
F I G U R E 3
Sequence and structural features of c-di-GMP riboswitches. (A) Consensus sequence and structure of c-di-GMP riboswitches based oncomparative sequence analysis and x-ray structure data. (B) Sequence and atomic-resolution structure model for Vc2 RNA, which representsone of two c-di-GMP riboswitches present in Vibrio cholerae. Note that the original sequence is engineered at the tip of P3 to carry a U1Aprotein binding site, which has been used to facilitate crystallization. G-U wobble base pairs are indicated with filled circles, and othernon-standard base pairs are indicated with hexagons. The c-di-GMP ligand is depicted in red, and other nucleotides in the secondarystructure model are colored to correspond to the similarly colored regions present in the tertiary structure model. Nucleotides are numberedaccording to the original nomenclature for Vc2. Tertiary structure model courtesy of Scott Strobel.
Volume 5, Number 1, 2010 / Microbe Y 17
tions can be harnessed by C. difficile to changefrom a motile to a sessile life style.
C. difficile has at least 12 versions of thisc-di-GMP riboswitch class encoded at variousplaces in its genome (Fig. 4B). Since most ofthese riboswitches are associated with genesof unknown function, c-di-GMP can have mul-tiple effects on bacterial physiology. For exam-ple, riboswitches 2 and 12 appear to be locatedclosest to bacteriophage genes, which suggeststhat some bacteriophages can use riboswitches
to read out c-di-GMP levels and spy onthe physiological changes in its bacte-rial host. Also, riboswitches 4 and 11appear to be unaffiliated with a codingregion. Perhaps these unaffiliated ribos-witches control the expression of non-coding RNAs that act in trans to regu-late or participate in other biochemicalprocesses.
All RNA, All the Time
If this RNA ligand and RNA receptorpartnership indeed is of ancient origin,predating even the emergence of DNAand protein in living systems, someRNA molecules would have had tofunction as c-di-GMP synthase ri-bozymes. Ribozyme-mediated biosyn-thesis of c-di-GMP from GTP, the samestarting materials used by modern pro-tein-based synthase enzymes, seemsbiochemically reasonable. Several re-search groups used test tube evolutionmethods to engineer ribozymes thatform phosphodiester bonds by cata-lyzing the nucleophilic attack of a 3�oxygen of an RNA on the triphosphatemoiety from another RNA. Simply re-peating this reaction twice with twoGTP molecules as substrates wouldyield c-di-GMP.
Hypothetical RNA World organismswould also have found it useful to de-grade c-di-GMP and reverse its bio-chemical effects on the cell, rather thansimply disposing of it outside the cell.Model building suggests that c-di-GMPcan be held in a geometry that wouldpermit efficient nucleophilic attack byone of its 2� oxygen atoms on the adja-cent phosphorus center. This reactionwill yield the linearized GpG�p dinu-
cleotide (where the 3� terminus is a 2�,3�-cyclicphosphate). To promote this reaction, a c-di-GMP phosphodiesterase ribozyme would needto hold the substrate in its reactive configura-tion, and perhaps hold a well-placed Mg2� ionor some other positively charged group, to ac-celerate c-di-GMP destruction to a speed that isbiologically relevant.
The c-di-GMP aptamers examined to datedo not alter the stability of c-di-GMP. How-
F I G U R E 4
c-di-GMP riboswitches in Clostridium difficile. (A) Regulation of the 13-gene operon forflagellum biosynthesis by a c-di-GMP riboswitch. In the absence of second messenger,a portion of the aptamer forms an anti-terminator that permits mRNA transcription andexpression of flagellar proteins. When c-di-GMP concentrations increase sufficiently,ligand binding by the aptamer prevents anti-terminator formation. Thus terminator stemformation causes transcription to halt before the coding regions of the mRNA areproduced. (B) Map of 12 c-di-GMP riboswitches and associated genes in C. difficile. Openboxes designate genes of unknown function, and phage symbols indicate the riboswitchresides next to phage genes.
18 Y Microbe / Volume 5, Number 1, 2010
ever, RNAs that simply bind c-di-GMP havewon half the battle. The X-ray model of theVibrio cholerae aptamer reveals that a smallshift in aptamer structure could produce theappropriate attack geometry; a nudge by a metalion or some RNA functional group would yielda ribozyme that could selectively cleave c-di-GMP faster than spontaneous cleavage by amillionfold or more.
Thus, there appear to be no biochemical im-pediments to the existence of ribozymes thatcatalyze c-di-GMP formation and destruction.Even if such ribozymes never existed, it seemslikely that RNA could be forced by skilled mo-lecular engineers to carry out these reactions.Similar arguments could be made for the pro-duction and breakdown of other nucleotide-based signaling compounds in bacteria such ascAMP and ppGpp. Cells from the RNA Worldmay have exploited ribozymes and receptorsthat participated in signaling networks based onthese and other nucleotide second messengers,and their evolutionary descendants may awaitdiscovery in modern cells.
Finding More Riboswitches forSecond Messengers
Although only one riboswitch class for a secondmessenger has been identified, it seems plausiblethat additional riboswitch classes that sense bac-terial second messengers will be found. Eventhough bacterial genomes are compact and thevast majority of their DNA codes for proteins,all sequenced genomes carry suspicious gaps.Some of these noncoding regions will only carrypromoters and regulatory DNA domains thatwe have come to expect will play roles in generegulation. Some other DNA gaps betweengenes might serve as templates for noncodingRNAs, including undiscovered riboswitches.
Comparative sequence analysis using bioin-formatics has been a very effective strategy forriboswitch discovery. These algorithms seek outgaps between ORFs that carry conserved se-quences and structural features. If the algorithmoutput presents an extensive consensus se-quence with features characteristic of RNA sec-ondary structure formation, then representa-tives may merit testing as riboswitch candidates.The best candidates tend to reside in the 5�UTRs of genes from the same metabolic path-way, which makes guessing the ligand that trig-
gers riboswitch action easier. In contrast, guess-ing the ligand becomes almost impossible if thecandidates are distributed among a wide rangeof apparently unrelated genes, or genes of un-known function. Unfortunately, riboswitch can-didates that bind second messengers are morelikely to fit the latter category.
Another challenge is that not all these com-pounds have been identified. Second messengersthat are narrowly distributed among bacterialspecies, or that control more obscure biologicalprocesses, may not be on the list of possibleligands to be tested for binding by newfoundriboswitch candidates. The bacterial secondmessengers such as cAMP or ppGpp and relatedcompounds are certainly widespread in bacteriaand would be excellent candidates for detectionby riboswitches. Since these compounds are alsomade from ribonucleotides and likely have avery old evolutionary origin, any riboswitchesthat sense these compounds also could be mod-ern representatives of ancient signaling systems.
Because the second messenger c-di-GMP iswidespread in bacteria, it was likely present veryearly in the evolutionary diversification of bac-teria. Furthermore, many bacteria sense thisRNA signaling compound using a genetic switchcomposed entirely of RNA. Both these charac-teristics are precisely what one would expect ifthis signal and sensor system had its origin in anRNA World. If so, the structures and character-istics of c-di-GMP riboswitches provide infor-mation on the set of molecular signals used bysome of the earliest forms of life.
However, there is much risk of error in suchspeculation. A circular RNA dinucleotide andits corresponding riboswitch RNA could haveevolved long after proteins emerged on thescene. Some of the riboswitch RNAs we seetoday may be better suited to the roles they playthan their protein-based counterparts. Ribo-switches are small compared to the coding re-gions of protein genetic factors, are produced inthe immediate vicinity of the genes they control,and directly interface with gene expression ma-chinery. These and other features may be advan-tageous for some gene control tasks. Ratherthan a legacy system from the RNA World,c-di-GMP and its riboswitch sensor may be aproduct of modern organisms, one that is more
Volume 5, Number 1, 2010 / Microbe Y 19
optimized for its roles than protein-based sys-tems.
Regardless of whether c-di-GMP riboswitchesare directly descended from RNA World com-ponents, or whether they are refined geneticswitches with a more recent pedigree, they allow
researchers to explore the true power of RNA asa functional polymer. Perhaps the discovery ofnew riboswitch classes that sense nucleotide-likesecond messengers will provide a richer view ofhow signaling circuitry looked in some of theearliest forms of life.
RNA science in the Breaker laboratory is supported by grants from the NIH and by the Howard Hughes Medical Institute.
Benner, S. A., Ellington, A. D., and Tauer, A. 1989. Modern metabolism as a palimpsest of the RNA world. Proc. Natl. Acad.Sci. USA 86:7054–7058.Breaker, R. R. 2006. Riboswitches and the RNA World, p. 89–107. In R. F. Gesteland, T. R. Cech, and J. F. Atkins (ed.): TheRNA world. Cold Spring Harbor Laboratory Press. Cold Spring Harbor, N.Y.Joyce, G. F. 2002. The antiquity of RNA-based evolution. Nature 418:214–221.Romling, U., Gomelsky, M., and Galperin, M.Y. 2005. C-di-GMP: the dawning of a novel bacterial signaling system. Mol.Microbiol. 57:629–639.Smith, K.D., Lipchock, S.V., Ames, T.D., Wang, J., Breaker, R.R., and Strobel, S.A. 2009. Structural basis of ligand bindingby a c-di-GMP riboswitch. Nat. Struct. Mol. Biol. (in press).Sudarsan, N., E. R. Lee, Z. Weinberg, R. H. Moy, J. N. Kim, K. H. Link, and R. R. Breaker. 2008. Riboswitches in eubacteriasense the second messenger cyclic di-GMP. Science 321: 411–413.Tamayo, R., J. T. Pratt, and A. Camilli. 2007. Roles of cyclic diguanylate in the regulation of bacterial pathogenesis. Annu.Rev. Microbiol. 61:131–148.Wilson, D. S., and J. W. Szostak. 1999. In vitro selection of functional nucleic acids. Annu. Rev. Biochem. 68:611–647.White, H. B., III. 1976. Coenzymes as fossils of an earlier metabolic state. J. Mol. Evol. 7:101–104.
20 Y Microbe / Volume 5, Number 1, 2010