My entry into the RNA world began in 1967 when a visiting lecturer at the University of Toronto, the late Dr. Gordon Tener, inspired my interest in RNA. Various aspects of RNA biology, from ribosomes to plant viruses, occupied my attention during the early stages of my career. It was only sometime later that I became fascinated by mRNA turnover in bacteria as it represented an intellectual challenge and a research opportunity.
My subsequent path as a graduate student at Cornell with David Wilson allowed me to witness early developments in mRNA decay in Escherichia coli without participating directly. The methodology of the era, pulse-labeling and DNA-RNA hybridization, showed that mRNAs in E. coli were typically short-lived with average half-lives of about 2 min. Moreover, it was widely believed that degradation of an mRNA began even before its transcription was completed. Curiously the direction of degradation, 5′ to 3′ or 3′ to 5′, seemed to preoccupy the field and precipitated fierce debates. Nonetheless, the late David Apirion formulated a model for mRNA turnover that, although simplified and short on mechanistic detail, provided high level direction to the field. He postulated that the physical degradation of an mRNA entailed a series of endonucleolytic cleavages followed by scavenging of the fragments by known 3′-exonucleases such as polynucleotide phosphorylase (PNPase) and RNase II. The key challenge seemed to lie in identifying “the” initiating endonuclease.
Development of more sensitive analytical techniques (e.g., Northern blotting, nuclease protection, and primer extension) and phage-encoded RNA polymerases rejuvenated the investigation of prokaryotic mRNA metabolism in the early 1980s. Fortunately, this coincided with my entry into the field, propelled by a need to explain how translational repression led to mRNA destabilization. To my pleasant surprise, my first Northern blot of bacterial RNA showed that the two nested rpsT mRNAs, encoding ribosomal protein S20, could be detected cleanly, not as a partially degraded smear. Moreover, mutational inactivation of PNPase led to the accumulation of a stable degradative intermediate that could only be generated by endonucleolytic cleavage, consistent with Apirion's model. The obvious next step was to identify the activity responsible. As often happens in research, an unrelated study of the expression of T4 phage gene 32 from Henry Krisch's group provided the clue. They showed that gene 32 mRNA is shortened to a stable intermediate by RNase E, an endonuclease then believed to process only non-coding RNAs. This precipitated a flurry of work from multiple groups, mine included. Collectively, we implicated RNase E in the degradation of a number of mRNAs and provided evidence, although not definitive, that it was a sought-after initiating mRNase. Although RNase E now appears paramount in this regard, other enzymes, including RNase G (a paralog of RNase E), RNase P, RNase III, and RNases from the toxin-antitoxin family, can catalyze the initial cleavage of mRNAs in vivo. But how an mRNA transitions from active translation to vulnerability to the first cleavage remains hazy.
Further progress came with difficulty: RNase E and its gene provided remarkably recalcitrant to characterization. Two approaches ultimately prevailed: somewhat brute force renaturation of over-expressed, intact RNase E from an SDS gel (by my lab) and classical purification with vigorous suppression of proteolysis (by A.J. Carpousis). A.J.’s work excitingly revealed that RNase E co-purified with PNPase. Subsequently, this complex, termed the RNA degradosome, was found to contain a DEAD-box RNA helicase and a glycolytic (!) enzyme (RhlB and enolase, respectively, in E. coli). More surprisingly, this complex resides on the inner bacterial membrane. Alas, the degradosome was dogged by initial controversy. Such concerns have now been put to rest. My group demonstrated the functional reconstitution of the degradosome from purified components while Ben Luisi's employed crystallographic and biophysical methods to define its protein–protein interactions. Degradosome-like complexes can be found in other bacteria, most strikingly in Bacillus subtilis which lacks a homolog to RNase E but encodes RNase Y, an endonuclease without sequence or structural similarly to RNase E (work by Stulke's group). RNase Y organizes a membrane-bound complex whose composition and properties broadly resemble those of the E. coli degradosome. Thus, macromolecular RNA processing machines are a common feature of RNA metabolism in a variety of bacteria. Curiously, however, neither RNase E nor RNase Y has “graduated” into a eukaryotic RNase; in contrast, RNase P, RNase III, and PNPase have evolved significant roles in higher organisms.
In parallel with enzymological investigations, the kinetics of decay had been determined for a modest number of monocistronic RNAs including the antisense regulator, RNA I, and the lacZ, ompA, rpsO, rpsT, and trxA mRNAs. Collectively, this small group of RNAs along with the 9S precursor to 5S rRNA provided the catalyst for many productive investigations that elucidated diverse properties of RNase E, PNPase, RNase II, Hfq, and poly(A) polymerase. One noteworthy property of RNase E is its 5′-end dependence (my work). The impetus lay in explaining why stable 5′-stem loops protect mRNAs from decay as shown by Belasco's group. Reading a paper on virusoid replication from Perreault's group prompted the idea that a circular RNA should prove resistant to RNase E if the enzyme requires a free 5′-end. Considerable effort was expended in preparing intact circular RNA, but it was indeed resistant to degradosomes. Ironically a control substrate, a 5′-mono-phosphorylated RNA (an intermediate in preparing circles), provided the proof that RNase E strongly prefers mono- over triphosphorylated substrates. The crystal structure of the catalytic domain of RNase E, a major accomplishment by Luisi's group and its collaborators, offered the structural explanation for this preference. Belasco's group subsequently identified a pyrophosphatase, RppH, that effectively activates mRNA substrates for cleavage, a process that mimics decapping. This process, along with oligo/poly-adenylation, is one of the tantalizing but incomplete parallels between prokaryotic and eukaryotic mRNA decay.
Almost all the work described above was performed on a “cottage industry model” where relatively small groups, mine included, pursued their favorite enzymes or RNAs. Now it is possible to measure the half-lives of all but the least abundant mRNAs simultaneously, to identify degradative intermediates, and to correlate any transcripts’ abundance with experimental measures of translational efficiency and secondary structure through high-throughput, genome-wide methods. Needless to say, these approaches demand careful execution and thorough bioinformatic analyses. A critical test of the “big data” approach to understanding mRNA stability will be the successful prediction of an mRNA's half-life strictly from inspection of its primary sequence, in prokaryotes or eukaryotes. Time will tell….
The adoption of genome-wide approaches begs a question: Can the investigation of mechanisms of mRNA decay survive a deluge of RNA-seq data? Let's hope so as major questions remain to be solved, several of which do not depend on global methods. A first challenge is extending the structural characterization of RNA processing complexes. The RNA degradosome is a prime candidate for high resolution cryo-EM, particularly in complex with native RNAs and their chaperones and regulators. A host of questions can be addressed if this technique proves successful. Refinement and application of more traditional methods should also reveal the detailed mechanisms of substrate recognition and phosphodiester bond cleavage. A second challenge centers on the cell biology of mRNA decay, notably on the spatial organization of the participants and rationalizing the long recognized but still unexplained links among transcription, translation, and mRNA stability. Determining the extent to which bacterial RNAs localize to specific cellular sites is now practicable, with initial exciting outcomes. Yet if degradosomes are tethered on the inner membrane, RNAs destined for processing or decay must be trafficked to them. Is the RNA binding protein, Hfq, for example, a “death chaperone” that facilitates the movement of damaged or untranslated mRNAs to degradosomes? Or does this occur randomly by diffusion? Do stable RNAs employ the same pathways or chaperones for processing? Are degradosomes agnostic or specialized as implied by some studies? How do physiological processes (e.g., cell division) and external signals influence the organization and activities of the decay machinery? A third set of questions addresses the significance of RNA processing and turnover to cellular physiology, particularly in natural settings, including mixed communities, microbiomes, and infected tissues. How does any mRNA's abundance and stability in one species respond to nutritional or environmental influences or host responses affecting the whole population? Is the relative importance of the enzymes and pathways of decay as we currently perceive them maintained in natural settings? Description of such changes begs the more difficult question: What are the underlying mechanisms?
Looking back, I thank all my past trainees for their hard work, dedication, and enthusiasm; my mentors and colleagues near and far for advice, materials, and constructive criticism; four former Institutions and my current home at UBC for fostering my training and subsequent career; and the Medical Research Council of Canada/Canadian Institutes of Health Research for providing financial support.