Specificity Landscape of Ribonuclease P Processing of Pre- Trna Substrates by High-Throughput Enzymology

SPECIFICITY LANDSCAPE OF RIBONUCLEASE P PROCESSING OF PRE- TRNA SUBSTRATES BY HIGH-THROUGHPUT ENZYMOLOGY

COURTNEY NICOLE NILAND

Submitted in partial fulfillment of the requirements

For the degree of Doctor of Philosophy

Dissertation Advisor: Dr. Michael E. Harris

Department of Biochemistry

CASE WESTERN RESERVE UNIVERSITY

January 2017

CASE WESTERN RESERVE UNIVERSITY

SCHOOL OF GRADUATE STUDIES

We hereby approve the thesis/dissertation of

Courtney Nicole Niland

candidate for the degree of Doctor of Philosophy.

Committee Chair

Hung-Ying Kao, Ph.D.

Committee Member

Michael E. Harris, Ph.D.

Committee Member

Eckhard Jankowsky, Ph.D.

Committee Member

Blanton Tolbert, Ph.D.

Committee Member

Michael A. Weiss, Ph.D.

Date of Defense

October 25, 2016

*We also certify that written approval has been obtained for any proprietary

material contained therein.

Dedication

Firstly, I dedicate this body of work to my wonderful parents, Timothy and

Jeannie Niland, because without their love and support it would not exist. From an early age, they instilled in me a strong moral compass and a high value for education. I saw the value of hard work manifested in their own actions as I grew- up and it was through their encouragement that I went to college and subsequently, graduate school. They have been my rock throughout this sometimes extremely challenging endeavor. They have shared in my triumphs and success and comforted me in times of worry and struggle. They have been my biggest cheerleaders, closest confidants, trusted counselors, and above all my best friends. The accomplishment of a PhD is as much theirs as mine.

Secondly, I dedicate this to my amazing friends and family who have showered me with their pride and enthusiasm throughout this time. Everyone from aunts, uncles, cousins, grandparents, family friends, and others to my friends both in Cleveland and abroad who have become a second family. To my lab family for fun outings, emotional support, and constant team enthusiasm. To my great mentor, Dr. Michael Harris for his wonderful mentorship and kind patience and understanding throughout my graduate schooling, even before I became a member of the lab. I could ask for no better lab experience in graduate school and will look back on my time here with fond memories as I move on to future careers.

Table of Contents

List of Figures ...... 3 Acknowledgements ...... 5 Specificity Landscape of Ribonuclease P Processing of Pre-tRNA Substrates by High- Throughput Enzymology ...... 6 Abstract ...... 6 Chapter I: Introduction ...... 8 Introduction ...... 9 Ribonuclease P...... 13 RNase P:pre-tRNA Interactions ...... 16 RNase P Mechanism ...... 18 Measuring the True Specificity of an Enzyme ...... 23 High-Throughput Enzymology to Measure RNase P Specificity...... 24 Discussion ...... 25 Chapter 2: Determination of the Specificity Landscape for Ribonuclease P Processing of Precursor tRNA 5' Leader Sequences ...... 27 Abstract ...... 28 Introduction ...... 29 Results and Discussion ...... 34 RNase P Processing of Pre-tRNA Substrate Pools ...... 34 Modeling of RNase P Substrate Specificity ...... 36 Specificity Landscape of Substrate Processing ...... 40 Methods ...... 48 Multiple-Turnover and Single-Turnover Reactions ...... 48 High-Throughput Sequencing Kinetics (HTS-Kin) ...... 49 Quantitative Modeling of Sequence Specificity ...... 50 Supplementary Methods and Data ...... 51 Preparation and Purification of Protein and RNA ...... 51 Multiple turnover and single turnover reaction kinetics ...... 54 High-Throughput Sequencing Kinetics (HTS-Kin) ...... 56 Quantitative modeling of sequence specificity ...... 62 Chapter 3: Optimization of high-throughput sequencing kinetics for determining enzymatic rate constants of thousands of RNA substrates ...... 71 Abstract ...... 72

Introduction ...... 73 Results and Discussion ...... 77 Determination of relative rate constants for in vitro RNA processing reactions by HTS-Kin ...... 77

The magnitude of krel is independent of the distribution of substrate mole fractions in the initial precursor RNA population...... 83 Optimization of reaction kinetics and choice of internal reference for calculation

of krel ...... 86 Reliability of first-strand cDNA synthesis and quantitative PCR for Illumina library preparation ...... 90

Robustness of Illumina sequencing for reproducible determination of krel values ..... 95 Evaluation of experimental error ...... 96 Chapter 4: The contribution of the C5 protein subunit of Escherichia coli ribonuclease P to specificity for precursor tRNA is modulated by proximal 5’ leader sequences ...... 104 Abstract ...... 105 Introduction ...... 107 Results ...... 114 Discussion ...... 130 RNA and Protein Preparation ...... 134 RNase P Reactions ...... 134 High-Throughput Sequencing Kinetics ...... 135 RNA Sequence Specificity Modeling ...... 136 Supplementary Data ...... 137 Chapter 5: Summary and Future Directions ...... 141 Summary ...... 142 Identification of Rate Limiting Step in RNase P Processing of 5’ Leader Variants in Pre-tRNA by HTS-Kin ...... 143 Investigation of the Sources of Error in HTS-Kin and Optimal Conditions for Quantitative Analysis ...... 144 Observation of Energetic Coupling in RNase P Specificity for Pre-tRNA Substrates ...... 146 Discussion ...... 148 Future Directions ...... 149 Bibliography ...... 151

List of Figures

Figure 1-1: Complexity of RNA protein interactions in a strongly simplified model system…………………………………………………………………………………...12

Figure 1-2: Diagram of the three-dimensional structure of Thermotoga maritima P RNA bound to pre-tRNA………...... ……………………………………14

Figure 1-3: The X-ray crystal structure of RNase P-product complex from T. maritima at 4.2Å………………………………...……………………………………………19

Figure 2-1: Recognition of the pre-tRNA 5′ leader sequence by the RNA and protein subunits of ribonuclease P………………………………………………………..…..30

Figure 2-2: Determination of the effects of all possible pre-tRNA 5′ leader sequence variants on the relative kcat/Km for RNase P processing………………………....35

Figure 2-3: Quantitative modeling of 5′ leader sequence specificity of RNase P……..38

Figure 2-4: Effect of sequence variation in the 5′ leader is primarily on substrate affinity and not on catalysis…………………………………………………………………….42

Figure 2-5: Reaction free-energy profile for RNase P processing of pre-tRNAMet82 and effects of sequence variation defining the specificity landscape...... …44

Figure 2-S1: Analysis of raw HTS-Kin datasets obtained with the 21A or 21B terminal 5’ sequence………………………………….………………………………………..65

Figure 2-S2: Quantitative analysis of single turnover HTS-Kin datasets performed with the 21A terminal 5’ sequence…………………………………………………………….67

Figure 2-S3: Single substrate reactions using pre-tRNAMet with varying 5’ leader sequences……………………………………………………………………………….68

Figure 2-S4: The 5’ leader sequences from endogenous pre-tRNA from E. coli show variation in specificity………….………………………………………………………...70

Figure 3-1: High-throughput sequencing kinetics (HTS-Kin) measures processing rates of thousands of RNA substrates using internal competition kinetics…...78

Figure 3-2: Analysis of the dependence of the observed krel on the distribution of substrate mole fractions in the initial precursor RNA population………………...………84

Figure 3-3: Optimization of reaction kinetics and choice of internal reference for calculation of krel…….……………………..………………………………………………..88

Figure 3-4: Analysis of the reliability of first-strand cDNA synthesis and

3 quantitative PCR for Illumina library preparation………………………………...…………..93

Figure 3-5: Illumina sequencing errors contribute to imprecision in measurement of low krel values……………………….………………….………………………………….....97

Figure 3-6: HTS-Kin replicates show reproducible determination of substrate variant krel except for slowest reacting substrates…………………………………………...99

Figure 4-1: Bacterial RNase P is an essential ribonucleoprotein RNA processing enzyme…………………………………………………………………………...109

Figure 4-2: Sequence determinants for holoenzyme and ribozyme reactions are unique and reveal changes in specificity……………………………..…………………116

Figure 4-3: Plotting the coupling coefficients from the PWM+coupling coefficient model identifies key altered energetic in holoenzyme and ribozyme reactions…...... ….119

Figure 4-5: Additional secondary structure in the acceptor stem of the pre-tRNAMet formed between 5’ leader nucleotides and the 3’ACCA slows processing rate constants…………………………………………………………………….125

Figure 4-6: Coupling between nucleotides in the RNA and protein binding sites in the 5’ leader of pre-tRNA is a key component of RNase P specificity...... 128

Figure 4-7: A mechanistic model for modulation of interdependence of proximal and distal leader sequence specificities …………….………………………………………132

Figure 4-S1: Schematic of the HTS-Kin procedure………………………………………137

Figure 4-S2: Dividing the holoenzyme HTS-Kin data into subsets reveals extensive coupling between the sequence identity in the RNA binding site on sequence specificity of the protein binding site in the 5’ leader of pre-tRNA……………..138

Figure 4-S3: Position weight matrix values of holoenzyme HTS-Kin data subsets reveals changes in sequence specificity of 5’ leader nucleotides in the protein binding site……………………………………………………………………………………..140

Acknowledgements

For helpful discussions on the manuscripts presented in this thesis, including critical reading and discussion of results, I thank members of the Harris lab including Dr. Michael Harris, Andrew Knappenberger, Jing Zhao, Hsuan-Chun

Lin, Edward Ollie, and Dr. Daniel Kellerman. I would like to especially point out the contribution of Jing Zhao to the data in the second chapter herein in which she performed all studies using the transversed substrate population (21B) in Figure

2-3. In addition, the essential work in Figure 2-2 using the pre-tRNA population randomized at N(-8 to -3) was the result of experiments by Dr. Ulf-Peter Gunther and Dr. Frank Campbell. I would also like to thank the members of my thesis committee for their support and guidance over these years: Dr. Hung-Ying Kao,

Dr. Michael Harris, Dr. Eckhard Jankowsky, Dr. Blanton Tolbert, and Dr. Michael

Weiss. Finally, I would like to thank members of the Biochemistry Department at

Case Western Reserve University for providing a collegial environment in which to perform this exciting work and for helpful feedback at annual seminars.

Specificity Landscape of Ribonuclease P Processing of Pre-tRNA Substrates by High-Throughput Enzymology

Abstract

COURTNEY NICOLE NILAND

To fully understand the roles of RNA processing enzymes in cellular processes and human health, it is essential to dissect their substrate specificity. Using

Ribonuclease P (RNase P) as a model system, this body of work seeks to better understand multiple substrate recognition by RNA processing enzymes. The ubiquitous endonuclease RNase P removes the 5’ leader from all pre-tRNAs in the cell. To understand how variation in the 5’ leader is accommodated by the active site of RNase P, we comprehensively determined the processing rates of pre-tRNA substrates with all possible sequence combinations in the 5’ leader. This quantification involved Illumina® sequencing of the residual substrate population at different reaction times to monitor substrate depletion and calculate relative rate constants, a technique termed HTS-Kin. Additionally, the 5’ leader of pre-tRNA is recognized by both the catalytic RNA subunit (P RNA) and smaller protein subunit of RNase P, C5. We therefore hypothesized that variation in substrate sequence contacting one enzyme subunit may alter the recognition or energetic contribution of contacts to the other enzyme subunit. Upon comprehensive determination of

RNase P specificity for 5’ leader sequences, we have determined that this enzyme is tuned for specificity at association as the catalytic rate constant is unaffected by substrate variation. We have also performed mechanistic analysis to identify the key sources of error in the HTS-Kin technique: experimental error and Illumina sequencing error. Finally, we ascertained that the sequence identity of 5’ leader nucleotides contacting P RNA does not alter C5 protein specificity but rather modulates its energetic contribution to the processing rate.

Chapter I: Introduction

Introduction

Recent decades have opened a wealth of knowledge and excitement in the field of RNA biology and biochemistry. From the discovery of catalytic RNA, the spliceosome, and RNA editing decades ago to the recent discovery of microRNAs, long non-coding RNAs, and CRISPR RNAs, the field of RNA biochemistry has flourished with discoveries. Not only did these discoveries exterminate the central dogma that RNAs sole purpose is to encode for protein through translation, they ushered in a profound recognition of the role of RNA in gene regulation. Through careful experimentation by brilliant scientists, these intriguing discoveries are now transforming our view of human health and disease. For example, defects in function of many of these RNAs are associated with various forms of cancer and neurological disorders (1,2). Additional evidence for the importance of these RNAs in biology is provided by studies that implicate defects in RNA-binding proteins in various diseases (3-6). The association of these RNAs with many disease states begs a mechanistic explanation of their function, particularly with regard to substrate recognition and structure-function relationships.

The gateway to the RNA world was opened with the identification that RNA could act as a catalyst for biological reactions in the cell, much like its protein counterparts. The first demonstration of the catalytic power of RNA for a substrate in trans was shown for the enzyme RNase P in the lab of Dr. Sydney Altman in

1983 (16,17). Along with the discovery of catalytic RNAs in the form of the hammerhead (7), hairpin (8,9), Varkud satellite (10), and hepatitis delta virus (11) ribozymes, larger and more complex ribozymes were discovered, including group

I and II self-splicing introns (12-15). These large ribozymes contain distinct domains and utilize nucleophiles in the form of metal ions or water molecules to catalyze hydrolysis or phosphoryl transfer reactions (18).

The myriad of functions performed by RNA in the cell necessitate its formation of complex three-dimensional structures to maintain its biological roles.

The making or breaking of interactions can result in regulation of structures in the

RNA molecule that direct its function; such is the case for riboswitches, RNA thermometers, and small RNAs that regulate bacterial gene expression (19-21).

This complexity translates from these small RNAs to motifs of large RNA complexes described above. A number of crystal structures have been determined for these large and small RNAs that shed light on the intricacy of their structure and its relation to substrate recognition and catalysis. For example, structures of the ribosome, group I - and group II introns were determined some time ago along with more recent structures of the hammerhead ribozyme, HDV ribozyme, Varkund satellite ribozyme, and RNase P (22-28). In each of these complexes, an intricate hydrogen-bonding network is apparent along with the necessity of metal ions in the folded structure. A network of RNA secondary structures including loops and bulges, helices, pseudoknots, kinks, and turns are pieced together to obtain these complex structures.

In recent decades, the powerful combination of RNA and protein in recognition of various substrates has become well-established (Figure 1-1).

Examples of biologically important ribonucleoprotein complexes (containing catalytic and non-catalytic RNA) include the ribosome, spliceosome, RNA

10 interference, and others. The textbook example of the ribosome shows that whereas the RNA subunits together are catalytic, the various ribosomal proteins serve to stabilize the structure. The spliceosome of eukaryotes is comprised of several small nuclear ribonucleoproteins, snRNPs – each made of small nuclear

RNA and various proteins, that function to regulate gene expression by removing introns from pre-mRNA transcripts and performing alternative splicing to produce multiple genes from a single RNA transcript. Heterogeneous nuclear ribonucleoproteins (hnRNPs) are a diverse class of complexes that guide pre- mRNA folding, regulate splicing, and transport mRNA from the nucleus to the cytoplasm. Additional examples of RNA-binding proteins and RNA-protein complexes are vast in biology.

Another member of this family is the bacterial RNase P, which functions as a complex of a large catalytic RNA and a smaller protein subunit, both important for substrate recognition. A common theme among these RNA-binding proteins and ribonucleoprotein complexes is that their biological function necessitates their recognition of multiple substrates, many of which vary in their sequence and/or structure. This fact begs a reconciliation of the need for enzyme specificity while accommodating cellular needs. In many cases a quantitative mechanistic determination of these enzymes specificity is lacking. In the work presented herein, we employ RNase P as a model system to better understand multiple substrate recognition and the coordinated functions of enzyme subunits.

Figure 1-1. Complexity of RNA protein interactions in a strongly simplified model system. RNAs are represented as structures, proteins as colored circles. Black lines indicate RNA-protein interactions, dotted red lines RNA-RNA interactions, dotted blue lines protein-protein interactions. Created by Eckhard Jankowsky and Michael Harris.

Ribonuclease P

Ribonuclease P (RNase P) is an essential enzyme in all domains of life, consisting of a catalytic RNA subunit, P RNA (~400 nucleotides) and a small protein subunit, C5 Protein (~100 amino acids) in bacteria (29) (Figure 1-2). This complexity is increased in higher order archaeal and eukaryotic versions of RNase

P which include several protein subunits and a structurally similar single catalytic

RNA subunit (30,31). In the mitochondria of humans and plants, a unique form of

RNase P has been identified that contains no RNA subunit, called protein-only

RNase P (PRORP), but maintains a similar catalytic mechanism to its bacterial form (32,33). In all forms of life, RNase P is responsible for the removal of the 5’ leader of all precursor tRNAs (pre-tRNAs) in the cell (34,35). The P RNA subunit of bacterial RNase P can perform endonucleolytic cleavage of the pre-tRNA in vitro without the C5 protein under conditions of high ionic strength (17,36). Whereas hydrolysis of the phosphodiester bond is performed by the P RNA subunit, the C5 protein increases substrate affinity and the processing rate constant as well as the association of catalytically important divalent cations (37,38).

A comparison of the sequence from over 200 P RNAs species of bacteria revealed high conservation of nucleotides in the catalytic domain and led to a consensus structural model of the enzyme consistent with biochemical evidence

(39,40). The P RNA subunit of RNase P consists of several distinct domains with individual functions. Two distinct domains of P RNA are identifiable upon inspection of its secondary and tertiary structure: the catalytic domain and specificity domain. Region P7-P11 contains nucleotides in the specificity domain

Figure 1-2. Diagram of the three-dimensional structure of Thermotoga maritima P RNA bound to pre-tRNA based on the crystal structure of (Torres-Larios et al. 2005). Helices of P RNA are shown as individually colored cylinders. The tRNA- body of the substrate is depicted by a black ribbon and the P protein subunit (C5 in E. Coli) is shown as a gray sphere. The 5′-leader sequence of pre-tRNA is shown as a dashed line. (Reprinted with permission from the RNA Journal, citation: Sun L. and Harris ME. RNA (2007) 13(9): 1505-15. Copyright Cold Spring Harbor Laboratory Press).

14 that contact the t-stem loop of the tRNA; the catalytic domain in regions P1-P5 is where the chemistry is performed and catalytic metal ions bind (41,42). The P1-

P4 multihelix junction is highly conserved among species and contains residues involved in binding of magnesium cations through their non-bridging oxygen (43).

Specifically, the presence of a bulged nucleotide in the P4 helix is correlated with binding of magnesium ions and substrate cleavage (44). These divalent metal ions potentiate local changes in structure that are necessary for tertiary folding of the P

RNA as well as substrate binding and catalysis (44,45).

As described above, the C5 protein subunit of RNase P is not involved in cleavage of the 5’ leader from pre-tRNA. However, it is responsible for recruitment of divalent metal ions and substrate recognition and therefore is necessary for in vivo function (46). The C5 protein of RNase P recognizes the pre-tRNA substrate only in the 5’ leader at nucleotides distal to the site of cleavage (47). A pronounced single-stranded RNA-binding cleft and significant conservation of amino-acid residues is evident from the X-ray crystal structure of the protein subunit of

Thermatoga maritima RNase P and phylogenetic sequence conservation analysis

(26). Due to the high degree of sequence variation in the 5’ leader of pre-tRNAs, it has been assumed until recently that the C5 protein was a non-specific RNA- binding protein.

RNase P is distinguished from the other aforementioned ribozymes in its ability to perform multiple rounds of catalysis, i.e., multiple turnover reactions, due to its recognition of its substrate in trans (48). This is contrary to the single turnover self-cleavage reactions common to most ribozymes that act in cis. Another key

15 feature of RNase P as mentioned above is its biological necessity to recognize and process all pre-tRNAs in the cell. In the E. coli cell there are 87 pre-tRNAs which vary in sequence and structure; all must be processes by RNase P (49). Due to the significant variation in pre-tRNA substrates, the specificity of RNase P is poorly understood.

RNase P:pre-tRNA Interactions

Despite the lack of sequence conservation in pre-tRNA substrates, previous biochemical data have identified key determinants for RNase P recognition. The

P RNA subunit is thought to base-pair with the T-stem loop of the tRNA body through its specificity domain (50,51). For those pre-tRNAs encoded with a 3’-

CCA, the L15 region of the P RNA also makes direct base-pairs with these nucleotides (52-55). In addition, the G(+1)/C(+72) base-pair at the beginning of the acceptor stem of the tRNA is a positive determinant for substrate cleavage

(37). Previous studies from our lab also identified binding interactions between the

P RNA and proximal nucleotides in the 5’ leader of pre-tRNA using crosslinking and mutational analysis (56,57).

The C5 protein subunit of RNase P is also known to contact the 5’ leader of pre-tRNA at nucleotides distal to the cleavage site (47,58). In experiments measuring the equilibrium dissociation constant for pre-tRNA substrates with all identified determinants (cognate substrates) or lacking one or more (non-cognate substrates), substantial variation was observed in their affinity for P RNA and their

16 observed processing rate constant (59). Inclusion of the C5 protein provided uniformity in substrate binding and catalysis by increasing the affinity of all pre- tRNA substrates tested to similar values (59).

Biochemical data from our lab and others have shown that despite their natural variation, nucleotides in the 5’ leader of pre-tRNA substrates are important for substrate specificity and recognition. For example, hydroxyl radical protection assays showed increased protection of nucleotides N(-6) to N(-3) in the 5’ leader in the presence of the C5 protein compared to the P RNA alone, indicating direct contacts between the distal region of the 5’ leader and the C5 protein. In another study from the Fierke lab, structural information was used to guide mutagenesis experiments, which revealed that an α-helix of the C5 protein beginning with arginine-asparagine-arginine contained specific hydrogen bonds to the N(-4) of the

5’ leader and that these interactions are necessary for binding and conformational change (60).

Our lab previously identified a potential direct base-pairing interaction between the N(-1) position in the 5’ leader of pre-tRNA and an adenosine residue,

A248, in the J5/15 region of P RNA (56). This study investigated the effect of this mutation on catalysis by the ribozyme and the importance of this contact in the holoenzyme and its role in substrate recognition are unknown. Another set of studies also identified the N(-2) position of the 5’ leader as a possible interaction with the J18/2 region of P RNA; however, the exact residues involved were unidentified (61-63). The importance of contacts between the 5’ leader of pre- tRNAs and RNase P is underscored by studies that mutated the 5’ leader

17 sequences of cognate pre-tRNA substrates to those found in non-cognate pre- tRNAs and vice versa (59). In general, the 5’ leaders from non-cognate pre-tRNAs increased substrate affinity while those from cognate substrates decreased affinity, possibly reflecting the necessity of contacts to the 5’ leader in the enzyme- substrate complex when other optimal contacts are unavailable.

In addition to these direct contacts, the X-ray crystal structure of T. maritima enzyme-product complex suggests intimate contacts between the P RNA and the proximal region of the 5’ leader and between the C5 protein and the distal nucleotides of the 5’ leader (Figure 1-3; (26)). Unfortunately, atomic resolution of these interactions is unavailable from the data as the resolution of structure at the nucleobases of the 5’ leader is low, and thus many of the molecular interactions remain elusive. The importance of these individual contacts in the RNase P reaction and how they might work together to aid in substrate recognition are unexplored.

RNase P Mechanism

Biochemical studies to date of the RNase P reaction mechanism have supported a two-step pre-tRNA substrate binding mechanism (Scheme 1-1). In this model the RNase P-pre-tRNA complex is formed upon an initial binding event followed by a conformational change that increases affinity of the complex as additional contacts are made (64). Upon formation of the high-affinity complex, irreversible cleavage of the substrate occurs, and the 5’ leader and tRNA products

Figure 1-3. The X-ray crystal structure of RNase P-product complex from T. maritima at 4.2Å. The protein subunit of RNase P (pink) and P RNA subunit (blue) both make intimate contacts with the 5’ leader (green) of the pre-tRNA substrate. The tRNA body is colored orange. Individual nucleobases in the 5’ leader are not present because their precise position could not be determined with the obtained spectral data. Adapted from Reiter et al. Nature 2010 using PyMol (PDB: 3Q1R).

Scheme 1: In the holoenzyme, the pre-tRNA is bound with higher affinity than the ribozyme and the conformational change of the enzyme with substrate bound, k2, is promoted relative to ribozyme. Product release is much slower in the ribozyme than the holoenzyme and is possibly the rate limiting step in the ribozyme reaction.

20 are released from the RNase P active site (37). For substrate recognition, this conformational change in the enzyme-substrate complex upon binding has been observed for the Bacillus subtilis RNase P using fluorescence anisotropy; the results support an induced fit model of substrate recognition (64). Thus, two parameters in this reaction become important for its characterization: (i) the first- order rate constant of substrate cleavage (kcat) measuring the reaction of the enzyme-substrate complex to free enzyme and products (ii) and the second-order rate constant (kcat/Km) measuring the reaction of free enzyme and substrate through the reaction to free enzyme and products.

A simple chemical mechanism has been proposed for endonucleolytic cleavage of pre-tRNA by RNase P (48). RNase P hydrolyzes the phosphodiester bond between the first nucleotide in the tRNA body N(1) and the beginning of the

5’ leader, N(-1). Upon RNase P recognition, a transition state is formed with Mg2+ ion coordination of bridging and non-bridging oxygens of the phosphate backbone.

Binding of divalent cations promotes donation of a proton by the water nucleophile to the 2’ oxygen, which in turn facilitates protonation of the phosphate leaving group at the 5’ end of the tRNA body. Due to the slow rate constant observed in multiple-turnover reactions with RNase P, Fierke and colleagues have suggested that the enzyme is tuned for recognition of multiple substrates rather than rapid catalysis (65).

The high degree of sequence and structural variation in the pre-tRNA substrates, particularly in the 5’ leader, raises significant questions regarding the mechanism of RNase P cleavage. As stated above, early experiments using in

21 vitro transcribed RNA established that the P RNA subunit of RNase P is responsible for catalysis. Steady-state and pre-steady-state kinetics performed on the ribozyme (P RNA) alone identified product release as the rate-limiting step of the reaction (65-67). In fact, the equilibrium dissociation constant of the tRNA product with P RNA was lower than that of the pre-tRNA substrate (59). Pre-steady state reactions in the presence of the C5 protein subunit show an increase in affinity of RNase P for its pre-tRNA substrate compared to P RNA alone (49,65-

67). In steady-state reactions presence of the C5 protein also increases the catalytic rate constant likely through recruitment of magnesium ions and through the promotion of conformational change of P RNA (37,38,59). For the E. coli

RNase P holoenzyme, it was recently shown that the rate-limiting step is substrate association (49).

Dissecting the molecular recognition of RNase P for its pre-tRNA substrates using combinations of kinetics, thermodynamics, and structure-function assays provides a system in which to globally understand specificity of ribonucleoproteins.

By using a reasonably well-studied enzyme, we hope to elucidate the mechanism by which ribonucleoproteins can retain specificity for their cognate substrates among non-cognate binding sites in the cell. These studies not only have direct implications for understanding enzyme specificity, but also are likely to provide a more comprehensive view of how perturbations of this mechanism might promote disease states.

Measuring the True Specificity of an Enzyme

One of the key questions regarding enzymes that process multiple substrates is how they retain their specificity for various cognate substrates while avoiding recognition of other non-cognate molecules in the cell. Therefore, a true measure of enzyme specificity must include competition between potential substrates as this is what occurs in vivo (68). As previously described by Fersht, competition of two substrates for the same enzyme is described by a ratio of the reaction rates as given by a ratio of second order rate constants (kcat/Km) of each substrate multiplied by their respective concentrations (Equation 1-1). The derivation of this equation has been published by Fersht and emphasizes the fact that substrates are in competition at the level of their “specificity constant” or ratio of kcat/Km (68,69).

(푘푐푎푡⁄ ) 푣표푏푠2 퐾푚 2 푆2 = 푘 ( ) Equation 1-1 푣표푏푠1 ( 푐푎푡⁄ ) 푆1 퐾푚 1

As discussed above, the second-order rate constant of a reaction includes all reaction steps from enzyme and substrate collision to regeneration of free enzyme upon release of products. Therefore, measuring the second-order rate constant of a reaction will quantify the rate-limiting step for that substrate up to and including the first irreversible step. By its very nature then, it is possible that two competing substrates in an enzymatic reaction could have different rate-limiting steps (i.e., substrate 1 may be limited by association and substrate 2 by catalysis).

By placing substrates into competition with one another for the enzyme, each substrate acts as a competitive inhibitor to all others.

Previously, others have derived these equations in relation to kinetic isotope effect studies and using competitive alternative substrates (70-73). The results of this derivation have provided insight into the conditions under which these measurements can be made. These equations are valid for any concentration of substrate, regardless of its apparent Km, since substrates are in competition for association with the enzyme. While differences in substrate concentration affect the reaction rate, the specificity constant (krel) remains unchanged (49). In addition, this approach is also independent of enzyme concentration because although present in the equation, it cancels out in the final derivation. Thus, this equation is also applicable to single-turnover reactions in which enzyme is in large excess over substrate. Finally, this theory is valid for any number of substrates and can be expanded far above comparison of two. As shown previously by our lab, additional substrates in the enzyme reaction continue to act as competitive inhibitors for all other substrates in the reaction. This results in a decrease in the processing rate of all substrates equally, and thus the quantified krel is unaltered

(49).

High-Throughput Enzymology to Measure RNase P Specificity

Obtaining a true measure of enzyme specificity has been a significant challenge in biochemistry. Previous attempts have focused on either the effect of single point mutations on enzyme processing or in qualitative approaches that select only the most favorable substrates. In recognition of the above knowledge of biological specificity and advantage of a competitive approach to analyzing

24 substrate recognition, we have developed a high-throughput enzymology approach to quantitatively define enzyme specificity for RNase P.

To accomplish this, we used a combination of enzymology and high- throughput sequencing to monitor the processing rates of thousands of pre-tRNA substrate variants in a single reaction. The details of this technique are described throughout the remaining chapters of this thesis and have been published (49,74-

77). An initial study from our lab provided a proof of concept for the use of competitive alternative substrate reactions with RNase P. These studies showed that pre-tRNAMet and pre-tRNAFMet (canonical and non-canonical RNase P substrates respectively) each have a rate-limiting step at association (49). In addition, this study provided direct evidence that relative rate constants could be obtained by placing these substrates into competition with one another at a variety of concentrations to obtain a similar krel (49). As expected, addition of a third pre- tRNA substrate did not affect this parameter (49). These studies laid the foundation for scaling up this analysis to testing thousands of substrates as described herein, which we term High-Throughput Sequencing Kinetics (HTS-Kin).

Discussion

Here, we present the results of studies aimed at a comprehensive determination of specificity in RNase P. We set out to answer the following questions: 1) How is variation in the 5’ leader nucleotides of pre-tRNA accommodated by RNase P? 2) What is the effect of variation in reaction

25 parameters and experimental set-up on HTS-Kin results? 3) Does the sequence identity of contacts between the 5’ leader and P RNA alter specificity of contacts between the C5 protein and 5’ leader? Each of these questions is addressed in its own chapter.

Chapter 2: Determination of the Specificity Landscape for Ribonuclease P Processing of Precursor tRNA 5' Leader Sequences

Reprinted with permission from Niland, C.N.; Zhao, J.; Lin, H-C.; Anderson, D.R.; Jankowsky, E.; Harris, M.E. “Determination of the Specificity Landscape for Ribonuclease P Processing of Precursor tRNA 5' Leader Sequences” ACS Chem Biol (2016) 11(8):2285-92 Copyright 2016 American Chemical Society

Abstract

Maturation of tRNA depends on a single endonuclease, ribonuclease P

(RNase P), to remove highly variable 5′ leader sequences from precursor tRNA transcripts. Here, we use high-throughput enzymology to report multiple-turnover and single-turnover kinetics for Escherichia coli RNase P processing of all possible

5′ leader sequences, including nucleotides contacting both the RNA and protein subunits of RNase P. The results reveal that the identity of N(−2) and N(−3) relative to the cleavage site at N(1) primarily control alternative substrate selection and act at the level of association not the cleavage step. As a consequence, the specificity for N(−1), which contacts the active site and contributes to catalysis, is suppressed. This study demonstrates high-throughput RNA enzymology as a means to globally determine RNA specificity landscapes and reveals the mechanism of substrate discrimination by a widespread and essential RNA- processing enzyme.

Introduction

To function in the cell, most RNA-processing enzymes and ribonucleases must recognize many alternative RNA substrates. Multiple substrate recognition is an inherent and essential function characteristic of complex ribonucleoprotein machines such as the spliceosome, ribosome, and enzymes involved in processing tRNAs, snRNAs, and siRNAs. A major challenge they all face is to recognize their cognate substrates among the excess of non-cognate binding sites in the transcriptome. Therefore, a complete understanding of RNA metabolism and gene expression requires characterizing in detail the substrate RNA sequences and structures that drive enzyme specificity. Achieving a comprehensive understanding of the mechanistic basis for substrate discrimination by RNA-processing enzymes is challenging but important for realizing their potential as drug targets (78-80) and platforms for engineering synthetic biology (81-83).

Ribonuclease P is a ubiquitous and essential tRNA-processing endonuclease that generates the mature 5′ end of tRNA by removal of the 5′ leader sequence (84). In bacteria, RNase P is composed of a large catalytic RNA subunit

(P RNA) and a smaller protein subunit, both of which are responsible for substrate recognition. In Escherichia coli, a single RNase P enzyme must process all 87 pre-tRNAs, which vary greatly in sequence and length of their 5′ leaders with some base conservation observed only at nucleotides proximal to the cleavage site

(49,58) (Figure 2-1A). Despite this lack of 5′ leader sequence conservation, experimental approaches such as cross-linking, chemical protection, mutagenesis,

Figure 2-1. Recognition of the pre-tRNA 5′ leader sequence by the RNA and protein subunits of ribonuclease P. (a) A sequence alignment of all 5′ leaders in Escherichia coli pre-tRNAs reveals no identifiable conserved sequence motif. (b) X-ray crystal structure of the Thermotoga maritima RNase P-product complex from Reiter et al. P RNA subunit is shown in cyan, protein subunit in pink, tRNA body in orange, and green 5′ leader.

30 as well as X-ray crystallography of the Thermotoga maritima RNase P-product complex, demonstrate both the protein subunit and RNA subunit make direct contact with 5′ leader nucleotides N(−8) to N(−1) (16,34,35,37,56,59,85-90).

Comparison of the kinetics of genomically encoded E. coli pre-tRNAs reveals that they react in vitro with essentially equivalent kcat/Km values (37,59). Similar multiple-turnover reaction kinetics are due, in part, to variation in the strength of 5′ leader sequence interactions that may compensate for weaker affinities of different tRNAs (49,59). Bioinformatic analyses suggest that sequence-specific contacts between the protein subunit and the 5′ leader sequences of pre-tRNAs may be common in bacterial RNase P and may lead to species-specific substrate recognition (58).

Current structure models of pre-tRNA bound to bacterial RNase P indicate that the two nucleotides, N(−2) and N(−1), immediately 5′ to the cleavage site interact directly with the catalytic P RNA subunit consistent with biochemical data.

The N(−1) nucleobase interacts with a universally conserved adenosine in the

J5/15 region of P RNA (56,86,91). N(−2) is proposed to interact with J18/2 although the chemical details of this interaction are not defined (86,89,92). More distal 5′ leader sequences N(−8) to N(−3) are contacted by the small, but essential protein subunit (termed C5 in E. coli) (47,66,93). Biochemical and biophysical data show that the 5′ leader binds to a cleft on the surface of P protein in an extended conformation (35,47,94). An X-ray crystal structure of T. maritima RNase P with a

5′ leader sequence oligonucleotide soaked into the crystal is consistent with these intimate contacts (26) (Figure 2-1B). In vitro structure–function studies

31 demonstrate that sequence variation at 5′ leader nucleotides involved in both

RNA–RNA and RNA–protein interactions can have significant functional effects on

RNase P processing rates as well as cleavage site specificity (58,59,95). For example, analysis of 5′ leader point mutations and amino acid substitutions in P protein support favorable hydrogen bonding interactions between an adenosine at

N(−4) and the protein subunit of Bacillus subtilis RNase P; however, the sequence preference of E. coli RNase P at this position is diminished (58). Additionally, mutation of the 5′ leader at N(−1) showed altered processing rates and that are partly rescued by compensatory mutations P RNA in J5/15 (56,57). At present, the 5′ leader sequence specificity of RNase P enzymes and the manner in which sequence variation affects the reaction mechanism are not well-defined. As a consequence, we lack the information necessary to predict the effects of variation in the sequences of genomically encoded tRNA precursors in vivo and the understanding to identify the range of cognate substrates in the transcriptome.

As an initial step toward addressing this problem, we determined the sequence specificity of E. coli RNase P for 5′ leader sequence positions N(−8) to

N(−3) in a non-initiator pre-tRNAMet82, which includes binding sites for both the

RNA and protein subunits (75). Previously, we measured the relative kcat/Km values for RNase P processing of all possible substrate variations in the C5 binding site using high-throughput sequencing kinetics (HTS-Kin). HTS-Kin combines

Illumina sequencing and internal competition kinetic analysis to measure the relative second-order rate constant (kcat/Km) for thousands of alternative RNA substrates simultaneously. Quantitative modeling of the resulting rate-constant

32 distribution revealed the functional sequence specificity of C5 protein. However, the specificity for nucleotides proximal to the cleavage site that interact with P

RNA, as well as the mechanism by which discrimination between alternative substrates is achieved remain unknown.

Here, we determine the effects of all possible sequence variants at N(−6) to

N(−1) in the 5′ leader that contact the P RNA and C5 protein subunits on both multiple-turnover and single-turnover reaction kinetics. The observed specificity of RNase P under multiple-turnover conditions reveals N(−2) and N(−3) as the key determinants of alternative substrate discrimination. Moreover, we find that competition for alternative substrates occurs at the level of substrate association, not catalysis, and that this conceals specificity for N(−1) at the cleavage step.

Thus, a combined approach of high-throughput RNA enzymology and quantitative modeling of sequence specificity provides a general means to globally determine

RNA specificity landscapes. Taken together, the results reveal the key specificity determinants for an essential RNA-processing enzyme and support a general specificity landscape and mechanism by which discrimination is achieved.

Results and Discussion

RNase P Processing of Pre-tRNA Substrate Pools

To characterize the specificity of E. coli RNase P for pre-tRNA 5′ leader sequences, we randomized nucleotides N(−6) to N(−1) that include positions interacting with the RNA subunit (N(−2) and N(−1)), as well as N(−6) to N(−3) that interact with the C5 protein subunit, providing 4096 substrate variants (Figure 2-

2A) (56,59,62,87-90). Analysis of multiple-turnover kinetics of the pre-tRNAMet

N(−6 to −1) randomized population demonstrated that this pool of substrates reacts with kinetics that are overall slower than the native pre-tRNAMet82 with its genomically encoded 5′ leader sequence (Figure 2-2B). Similarly slow reactivity was observed for the pre-tRNAMet N(−8 to −3) substrate population randomized at positions N(−8) to N(−3) as reported in Guenther et al (75). Thus, randomization of the 5′ leader from N(−6) to N(−1) or from N(−8) to N(−3) affects RNase P multiple-turnover processing kinetics similarly, likely due to their commonly randomized region (N(−6) to N(−3)).

Met Next, the relative kcat/Km for all 4096 sequences in the pre-tRNA N(−6 to

−1) randomized population were measured using HTS-Kin. HTS-Kin measures changes in the relative abundance of alternative RNA substrates over reaction time using next-generation sequencing, and these data are used to calculate relative rate constants using internal competition kinetics (68,69,96,97). A relative second- order rate constant, krel, was calculated for each substrate that is normalized to the rate constant for the genomically encoded sequence (i.e., krel = (kcat/Km (variant))/( kcat/Km (AAAGAU)). A histogram showing the distribution of average relative rate

Figure 2-2. Determination of the effects of all possible pre-tRNA 5′ leader sequence variants on the relative kcat/Km for RNase P processing. (a) The six nucleotides randomized in pre-tRNAMet85N(−6 to −1) and pre-tRNAMet85N(−8 to −3) substrate pools are indicated by a green box, RNase P cleavage site indicated with arrow. (b) Multiple-turnover kinetics of genomically encoded pre-tRNAMet(WT)21A substrate (black) to the randomized pre-tRNAMet N(−8 to −3)21A (red) and pre- tRNAMet N(−6 to −1)21A (blue) populations. The inset shows typical results from RNase P reactions run on a denaturing polyacrylamide gel, each lane shows a time point that separates substrate and product. (c) Histogram of the number of substrate variants with a particular krel value from HTS-Kin showing the average of three experiments with the pre-tRNAMet N(−6 to −1) and two with pre-tRNAMet N(−8 to −3). (d) Sequence probability logos identify sequence preference in the fastest 1% of substrates in the HTS-Kin reactions. Black letters show the wildtype sequence and indicate nonrandomized positions. (e) The location of the 256 sequences that are in common between the randomized pre-tRNAMet N(−8 to −3)21A and pre-tRNAMet N(−6 to −1)21A populations at positions N(−6) to N(−3) are indicated by a green box. (f) The observed krel values from the two separate independent experiments and distinct randomized pools are plotted and fit to a linear function, red line.

35 constants for all substrate variants from three replicate experiments is shown in

Figure 2-2C. This histogram defines a rate-constant distribution that describes the entire range of effects of sequence variation on kcat/Km. The rate-constant distribution for the pre-tRNAMet N(−6 to −1) population is similar in shape to that previously observed for pre-tRNAMet N(−8 to −3) randomized population (75). A significant number of sequences (ca. 200) in the pre-tRNAMet N(−6 to −1) population reacted with krel greater than 2-fold faster than the reference, indicating that the genomically encoded 5′ leader sequence at N(−6) to N(−1) is not optimal for kcat/Km.

Modeling of RNase P Substrate Specificity

To identify the sequence determinants optimal for kcat/Km, we calculated sequence probability logos from 5′ leader substrate variants with the top 1% of krel values (98) (Figure 2-2D). Surprisingly, the results indicated little sequence preference at N(−1) and N(−4) which were previously indicated to contribute to specificity (56-58). In contrast, N(−2) and N(−3) showed strong preferences for adenosine and uridine, respectively. Optimal sequence determinants in the protein binding site (−6 to −3) of the 5′ leader are similar to those previously observed for these positions in HTS-Kin results obtained using the pre-tRNAMet N(−8 to −3) population (75). A subset of 256 sequence variants is common to both the pre- tRNAMet N(−6 to −1) and pre-tRNAMet N(−8 to −3) randomized populations (Figure

2-2E). As an internal check on accuracy, we compared the krel values measured for these sequences in the two independent randomized populations (Figure 2-

2F). The results show that the measured krel in both HTS-Kin experiments are highly correlative, demonstrating the reproducibility of the technique.

To amplify pre-tRNAs for high-throughput sequencing without ligation bias, an additional 21 nucleotides were added to the terminal 5′ end of pre-tRNAMet N(−6 to −1) substrates. Cross-linking, X-ray crystallography, and FRET studies demonstrate that the 5′ leader sequence is bound in an extended single-stranded conformation (35,47,94), therefore these additional 21nt (termed 21A) could result in formation of unfavorable secondary structure in the ground state that would complicate identification of intrinsic RNase P sequence specificity. We tested for

Met these effects by comparing the krel values determined for a second pre-tRNA

N(−6 to −1) population in which each nucleotide in the 21nt upstream sequence was changed to its Watson–Crick complement (termed 21B). Differences in krel value for the same 5′ leader substrate variant in the two different contexts identify potential instances of inhibitory secondary structure. Comparison of the rate- constant distributions for the pre-tRNAMet N(−6 to −1)21A and pre-tRNAMet N(−6 to

−1)21B populations (Figure 2-3A) shows that the majority of substrate processing rates correlate between the 21A and 21B substrate backgrounds. However, a subset of the population of N(−6) to N(−1) substrate variants reacts with greater than 2-fold slower kinetics in the 21A leader sequence context compared to its complement 21B.

To test whether inhibitory secondary structure explains the deviation of rate constants of some substrates between the 21A and 21B populations, M-Fold was used to calculate the predicted folding free energies between the 21A sequence

Figure 2-3. Quantitative modeling of 5′ leader sequence specificity of RNase P. (a) Density plot of average krel for each substrate variant in three HTS-Kin reactions in the 21A context compared to two reactions with the 21B substrate context. The plot shows the number of substrates in each hexagonal region indicated by the scale at right (less than 1% of data omitted for clarity). (b) Plot of average krel for 5′ leader variants in the context of 21A or 21B sequence. Individual points are colored by their M-Fold predicted highest folding free energy between the 5′ leader sequence and 21A sequence (less than 1% of data omitted for clarity). (c) PWM model used to describe the HTS-Kin data evaluated by plotting the observed krel from HTS-Kin against the predicted krel from the model. (d) A PWM model with an included coupling term used to describe the data evaluated by plotting the observed krel from HTS-Kin against the predicted krel of the model. (e) Linear coefficients for each nucleotide, indicated by color, from the PWM portion of model 2 shown for each position in the 5′ leader indicated at the bottom. (f) Heatmap of coupling coefficients from the model. Position and nucleotide identity indicated on each axis and the coupling coefficient indicated at the vertex.

38 and 5′ leaders (99). An overlay of these predicted folding free energies on a plot comparing the rate-constant distributions for the pre-tRNAMet N(−6 to −1)21A and pre-tRNAMet N(−6 to −1)21B populations is shown in Figure 2-3B. The 5′ leader sequences predicted to form the most stable secondary structure in the presence of the 21A context react with slower kinetics and show the greatest degree of displacement from the rate constant measured in the 21B context. Thus, a portion of substrates are affected in the ground state in a manner independent of the intrinsic sequence specificity of RNase P for N(−6) to N(−1). To isolate effects of

5′ leader sequence variation on RNase P processing independent of upstream sequence context, we selected and averaged the data for those substrate variants with krel values that varied by less than 2-fold between the 21A and 21B contexts for further analysis of intrinsic sequence specificity.

To comprehensively and quantitatively determine the specificity of RNase

P for 5′ leader sequences N(−6) to N(−1), we first applied a simple position weight matrix (PWM) model. PWM models are commonly used to describe DNA-binding proteins (100) and consider each position in the 5′ leader as independent and therefore uncoupled with other positions in the binding site. Fitting the HTS-Kin data to a PWM model and comparing the observed krel to that predicted by the model indicates this model only explains about half of the correlation between observed rate constant and 5′ leader sequence (Figure 2-3C). Additionally, the data were fit to a PWM model including an additional term that quantifies coupling between positions in the binding site of the 5′ leader. Including quantitative coupling coefficients (β values) in the model, which allows the sequence identity

39 at each position to modulate the effects of sequence variation at other positions in the binding site, provides a significantly better fit to the experimental data (Figure

2-3D). Importantly, similar results were obtained from fitting the complete pre- tRNAMet N(−6 to −1)21A or pre-tRNAMet N(−6 to −1)21B data sets (Figure 2-S1). This result indicates that the inaccuracies introduced by inhibitory secondary structure are relatively minor compared to the dominant contribution from intrinsic RNase P sequence specificity and reflects the overdetermination of these parameters by

HTS-Kin. Contributions to specificity are greatest at positions N(−2) and N(−3), consistent with the high probabilities of uridine and adenosine at these positions in the sequence probability logo (Figure 2-3E). The β values expressing the degree of coupling between positions are shown in a heatmap in Figure 2-3F. The strongest coupling occurs between adjacent nucleotides. Interestingly, relatively strong coupling coefficients are observed between the RNA binding site and the proximal protein binding site, in particular between the N(−2) and N(−3) positions.

This result indicates coupling at the single nucleotide level between the energetic contributions of nucleobases in the pre-tRNA 5′ leader that contact P RNA and those that contact C5 protein.

Specificity Landscape of Substrate Processing

The magnitude of kcat/Km reflects the first irreversible step, which is substrate association for the kinetic mechanism of pre-tRNAMet82. Processing of pre-tRNAMet with its genomically encoded 5′ leader sequence involves slow dissociation relative to the RNA cleavage step (49,59). However, the pre-tRNAMet

N(−6 to −1) randomized substrate pool contains all sequence combinations at positions N(−2) and N(−1) that are sites of interaction with the catalytic P RNA subunit of RNase P and could affect the substrate cleavage step. Although the protein subunit of RNase P contributes little to catalysis for optimal pre-tRNA substrates, the extent to which RNA–protein interactions broadly contribute to catalysis is not known. Additionally, it is not known whether catalytically important contacts between the 5′ leader and the P RNA subunit are stabilized or destabilized by the more distal protein contacts.

To determine the extent to which randomization of N(−6) to N(−1) affects catalysis, we performed single-turnover reactions under saturating enzyme conditions in which the observed rate constant reflects the substrate cleavage step. Similar to the multiple-turnover kinetics, the pre-tRNAMet N(−6 to −1)21A randomized pool under single-turnover conditions reacted with slower overall kinetics compared to the genomically encoded reference substrate; however, the substrate population reacted to completion (Figure 2-4A). This result is consistent with smaller overall effects of sequence variation on catalysis, or a smaller number of substrates for which this parameter is affected.

To distinguish these possibilities, we measured the relative single-turnover rate constants for all substrate variants in the pre-tRNAMet N(−6 to −1)21A population by performing HTS-Kin under saturating enzyme conditions. An overlay of histograms showing the observed rate-constant distributions from multiple- turnover versus single-turnover reactions with pre-tRNAMet N(−6 to −1)21A is shown in Figure 2-4B. A much narrower distribution of krel values is observed for single

Figure 2-4. Effect of sequence variation in the 5′ leader is primarily on substrate affinity and not on catalysis. (a) Single-turnover reactions performed at saturating enzyme conditions with pre-tRNAMet(WT)21A or pre-tRNAMet N(−6 to −1)21A from two and three experiments, respectively. (b) Histogram of the distribution of krel of substrates in a single-turnover HTS-Kin reaction compared to averaged multiple- turnover HTS-Kin reactions. (c) Comparison of krel for each substrate variant (individual points on the graph) in averaged multiple-turnover and a single-turnover HTS-Kin. (d) Single substrate reactions were used to determine the effect of 5′ leader sequence variation on the later steps in the reaction, (i.e., E·S→ E + P) or –1 the second-order rate constant (i.e., E + S→ E + P). The kcat (s ) from multiple- turnover reactions for each 5′ leader sequence listed on the x-axis from at least –1 –1 three experiments (red). The absolute kcat/Km (μ s ) measured for these sequences is shown from at least three experiments (black). Standard deviation is indicated by the error bars.

-turnover conditions demonstrating that few substrate variants in the 5′ leader alter the cleavage step, while the same substrates result in a much greater range of observed kcat/Km values. For those substrates in which the single-turnover rate constant was significantly affected, there are two possible interpretations: the catalytic rate constant may be reduced due to variation in the 5′ leader or these substrates may bind more weekly to RNase P and thus have a Km above the concentration used. A plot of observed krel values for substrates in multiple- turnover reactions versus single-turnover reactions reveals little correlation between the effects of 5′ leader variation and further illustrates the much smaller range of effects on the cleavage step (Figure 2-4C). The sequence specificity for cleavage under single-turnover conditions can also be modeled. Fitting the single- turnover HTS-Kin data to the PWM model with coupling coefficients described above, identifies N(−1) as the primary contributor to specificity (Figure 2-S2). The magnitude of the calculated PWM scores shows that uridine is optimal at N(−1) consistent with previous studies of RNase P catalysis and the current model of the

ES complex (56).

Thus, comparison of the multiple-turnover and single-turnover rate constants suggests a specificity landscape for 5′ leader sequence discrimination by E. coli RNase P (Figure 2-5) that globally describes the effects of sequence variation on the reaction mechanism. In this model, variation in proximal 5′ leader sequences have relatively small effects on the cleavage step and therefore are likely to also have small effects on kcat. In contrast, sequence variation at N(−2) and N(−3) alters kcat/Km, which controls the competition between alternative

Figure 2-5. Reaction free-energy profile for RNase P processing of pre-tRNAMet82 and effects of sequence variation defining the specificity landscape. In this mechanism, the pre-tRNA substrate combined with RNase P (P RNA is shown as a blue oval and P protein a smaller red sphere) to form the RNase P-pre-tRNA complex which undergoes turnover to form the tRNA and 5′ leader sequence products and free RNase P. For the genomically encoded pre-tRNAMet82 leader sequence, kcat is fast relative to dissociation, and the magnitude of kcat/Km reflects association. Differences in the magnitude of kcat/Km for alternative pre-tRNAs arise from possible effects of unfavorable structure in the ground state (ΔG structure), differences in interactions with the 5′ leader sequence that affect the transition state for association (ΔG RNA-protein), or large changes in the cleavage rate constant that also affect the magnitude of kcat (ΔG cleavage).

44 substrates. We tested the predicted effects on the kcat/Km and kcat values for a series of single pre-tRNA substrates selected to span the nearly 100-fold range in krel as measured by HTS-Kin. As expected, the observed kcat/Km values measured using individual substrate reactions showed significant variation based on 5′ leader

2 sequence identity and correlate well with the krel from HTS-Kin (Adj. R = 0.81)

(Figure 2-4D and Figure 2-S3). The results further demonstrate HTS-Kin as an accurate and reliable method for rapid and comprehensive determination of relative rate constants. In contrast, the kcat values measured individually for the same substrates were very similar and varied less than 2-fold. An exception is the slowest reacting 5′ leader variant that is predicted to have inhibitory secondary structure between the 5′ leader and 21A and resulted in decreases in both kcat and kcat/Km. For the majority of substrates, the primary effect of 5′ leader sequence variation is on kcat/Km with minimal effects on the substrate cleavage step and consequently minimal effects on kcat (49,59).

Previous mutagenesis and cross-linking experiments clearly show a U(−1) makes an optimal direct contact to a conserved adenosine residue in J5/15 of P

RNA. However, the insensitivity of kcat/Km to the identity of the N(−1) nucleobase is apparent from both the HTS-Kin data and confirmed by single-substrate assays

(55-57,91). Insensitivity to sequence variation at N(−1) observed in HTS-Kin indicates formation of contacts with this position occur at a step subsequent to the first irreversible step, which for the genomically encoded 5′ leader sequence is association. As a consequence, variation at N(−1), which contributes primarily to the cleavage step, does not significantly affect the observed kcat/Km. Conversely,

N(−2) and N(−3) contribute significantly to RNase P specificity as demonstrated by optimal sequence probability logos and quantitative modeling of the HTS-Kin data.

The resulting sequence specificity model is likely to have important implications for substrate processing in the cell. The distribution of krel values predicted for the 87 different 5′ leader sequences encoded in the E. coli genome show significant variation in magnitude occurring throughout the rate-constant distribution (Figure 2-S4). The fastest reacting 5′ leader sequences matched those in proline, glycine, and alanine pre-tRNAs, which have adenosine or uridine at

N(−2) and uridine at N(−3) and thus are predicted to be optimized for kcat/Km based on the HTS-Kin results. The slowest reacting 5′ leaders predicted by these results are found in tyrosine, methionine, and arginine which lack at least one of these specificity determinants. Importantly, there are contacts between the P RNA subunit of RNase P and the tRNA body that may also modulate the overall enzyme specificity for these substrates in vivo (92,101-103).

The general model for tRNA biosynthesis in E. coli suggests that the rate- limiting step occurs at the level of transcription. For some substrates the presence of specificity determinants in the 5′ leader sequence may act to offset favorable or unfavorable interactions with the tRNA portion of the substrate to maintain uniform rates of 5′ end maturation. However, more recent molecular genetic analyses reveal that RNase P processes multiple polycistronic transcripts in which the order of processing occurs in a strict 3′ to 5′ direction (95). Thus, for some substrates, the presence or absence of specificity determinants may modulate cleavage relative to other competing cognate binding sites in the cell. The availability of a

46 comprehensive model for 5′ leader sequence specificity allows a more accurate evaluation of the potential differences in molecular recognition in vivo.

This study demonstrates the use of high-throughput RNA enzymology methods using common molecular biology techniques and instrumentation as a general means to globally determine the specificity landscape for molecular recognition by an essential RNA-processing enzyme. The application of HTS-Kin to measure both multiple-turnover and single-turnover kinetic parameters for thousands of RNA substrate variants reveals the full range of effects and the intrinsic sequence specificity for these two kinetic parameters. The ability to measure different kinetic parameters for the same sequence variants allows the construction of simple kinetic schemes for each substrate and insight into how variation alters the free-energy landscape of the reaction. Analysis of the high- throughput biochemical data using quantitative models of sequence specificity provides a way to identify key specificity determinants and quantitatively compare results obtained between experiments. Therefore, the general approach described here for characterizing the specificity of RNase P presents a rigorous and comprehensive way to investigate RNA specificity that is likely to be applicable to a range of experimental systems.

Methods

Multiple-Turnover and Single-Turnover Reactions

E. coli C5 protein was overexpressed and purified as described (104). P

RNA and pre-tRNAMet82 were synthesized by in vitro transcription using T7 RNA polymerase and PCR or cloned DNA templates. Pre-tRNA substrates were 5′ end labeled with γ-32P using polynucleotide kinase. Multiple-turnover reactions were performed in 50 mM Tris-HCl pH 8, 100 mM NaCl, 0.005% Triton X-100, and 17.5 mM MgCl2. RNase P was assembled by heating P RNA to 95°C for 3 min then

37°C for 10 min. MgCl2 was added and incubated at 37°C for 10 min before adding equivalent concentrations of C5 protein. Substrate was prepared separately by combining pre-tRNA with trace amounts of 32P labeled pre-tRNA, heating to 95°C for 3 min, incubating at 37°C for 10 min before adding MgCl2. Equal volumes of the enzyme and substrate were mixed to achieve final concentrations of 5 or 10 nM RNase P and 1 μM pre-tRNA for randomized pools. Aliquots were taken at selected time points and quenched with an equal volume of formamide loading dye containing 100 mM EDTA. Products were resolved on 15% denaturing polyacrylamide gels and quantified using phosphorimager analysis. Individual substrate assays used 3 μM pre-tRNAMet and 5 nM RNase P, and the kinetic data were fit to the integrated Michaelis–Menten equation. Single-turnover reactions were performed in 50 mM MES pH 6.0, 100 mM NaCl, 0.005% Triton X-100, and

17.5 mM MgCl2. Reactions were performed as described above except the final concentration of holoenzyme was 100 nM and substrate concentrations were <10 nM.

High-Throughput Sequencing Kinetics (HTS-Kin)

Multiple-turnover HTS-Kin measurements were made as described in

Guenther et al. Briefly, reactions were performed as described above, except scaled up 10-fold to provide sufficient material for subsequent analysis. Aliquots were taken during the time course, RNAs were resolved by polyacrylamide gel electrophoresis, and the residual substrate population was eluted and purified.

Equal amounts of RNA from individual time points were used as templates for first- strand cDNA synthesis. Products were diluted 1:300, and 1 μL of this dilution was used for PCR amplification followed by multiplexed Illumina sequencing in 75 bp single end reads on Hi-Seq 2500 instrument. Single-turnover HTS-Kin reactions were set up as described above for single-turnover reactions and were not scaled up. Primer sequences are included in Supporting Information.

Relative rate-constant (krel) values were calculated using

(1 − 푓) 푙푛 푅 푅 푖,0 (∑푖 ) 푅 1 푅 푋 푘 = 푖 0 푟푒푙 (1 − 푓) 푙푛 푖 푅 ∑1 푋 푅0 where R is ratio of each substrate variant to the reference, R0 this same ratio at the start of the reaction, and X is the mole fraction for each substrate variant. The ratio of substrates is quantified by the number of raw sequence reads from high- throughput sequencing and the total fraction of reaction was determined by quantification of polyacrylamide gels using ImageQuant software as described above.

Quantitative Modeling of Sequence Specificity

The HTS-Kin data were fit to a simple position weight matrix model treating each nucleotide in the randomized substrate region as independent and non- interacting using

6 푙푛(푘푟푒푙) = ∑ (푎푖퐴푖 + 푐푖퐶푖 + 푔푖퐺푖 + 푢푖푈푖) 푖=1

where ai, ci, gi, and ui, are integer values (0 or 1) signifying nucleotide identity and Ai, Ci, Gi, and Ui represent the linear coefficients for that nucleotide at position i. The PWM+IC model considered not only nucleotide identity and position in the randomized region but also the position and identity of other nucleotides in the binding site using the following equation:

6 푙푛(푘푟푒푙) = ∑ (푎푖퐴푖 + 푐푖퐶푖 + 푔푖퐺푖 + 푢푖푈푖) + 훽푗퐼푗 푖=1 where the second summed terms are pairwise interaction terms. For each of these couplings, βj is the linear coefficient for interaction j, and Ij is an indicator variable. Ij is 1 for all substrates with that specific pair of nucleotides, and 0 otherwise. Each interaction term which had an absolute t-value greater than 3.5

(p < 0.005) was recorded as significant. A final model was built using stepwise regression, starting with all of the significant pairwise interactions identified in the first step.

Supplementary Methods and Data

Preparation and Purification of Protein and RNA

Expression and purification of E. coli C5 protein was done as previously described (104). The E. coli P RNA gene with an added T7 promoter sequence and Bbs1 restriction site at the 3’ end was cloned into pUC18 vector. This plasmid was then linearized with Bbs1 (NEB R0539L) using standard protocol. The linearized plasmid was then used for run-off in vitro transcription with T7 RNA polymerase (NEB M0251S) using 5-10 µg of template cDNA. The resulting RNA was purified by phenol-chloroform extraction, resolved by PAGE and gel purified by UV shadowing. Excised RNAs were eluted from the gel in 10 mM Tris-HCl pH

8, 1 mM EDTA pH 8, and 0.1% SDS and recovered by ethanol precipitation. The final purified RNA was resuspended in TE Buffer (10 mM Tris-HCl pH 8, 1 mM

EDTA pH 8) and quantified using the UV absorbance. The E. coli pre-tRNAMet82 gene was also cloned into the pUC18 vector with added T7 promoter sequence on the 5’ end and Bbs1 restriction site on the 3’ end. This plasmid was also linearized as above using Bbs1 digestion. The resulting DNA was used as a template for

PCR to create sequence variants in the 5’ leader and add the 21nt sequence. PCR conditions were as follows: 1 U Taq DNA polymerase (Roche 04638964001), 1x supplied PCR buffer, 0.2 mM dNTP mix, 0.5 µM forward and reverser primers, and

18 nM template DNA were combined and heated on a thermocycler to 95˚C for 2 min followed by 40 cycles of 95˚C for 30 sec, 55˚C for 45 sec, and 72˚C for 1 min, followed by a final incubation at 72˚C for 5 min. A portion of the PCR product was then run on a 1% agarose gel at 100 volts and visualized on a transilluminator to

51 check product size before purifying by phenol-chloroform extraction and ethanol precipitation and resuspension in TE Buffer as described above. From the resulting cDNA template, 20-25 µg was used to synthesize substrate RNA by in vitro transcription using T7 RNA polymerase as described above.

PCR Primers used in this work: ptRNAMet(WT)21A Forward Primer

5’-

TAATACGACTCACTATAGGGAGACCGGAATTCAGATTGATGAAAAAGATGGC

TACGTAGCTCAGTTGG-3’ ptRNAMet(-6 to -1)21A Forward Primer

5’-

TAATACGACTCACTATAGGGAGACCGGAATTCAGATTGATGAANNNNNNGG

CTACGTAGCTCAGTTGG-3’ ptRNAMet(-8 to -3)21A Forward Primer

5’-

TAATACGACTCACTATAGGGAGACCGGAATTCAGATTGATGNNNNNNAUGG

CTACGTAGCTCAGTTGG-3’ ptRNAMet(-6 to -1)21B Forward Primer

5’-

TAATACGACTCACTATACCCTCTGGCCTTAAGTCTAACATGAANNNNNNGGC

TACGTAGCTCAGTTGG-3’ ptRNAMet(CGGGUA)21A Forward Primer

5’-

TAATACGACTCACTATAGGGAGACCGGAATTCAGATTGATGAACGGGTAGG

CTACGTAGCTCAGTTGG-3’ ptRNAMet(AAUGGC)21A Forward Primer

5’-

TAATACGACTCACTATAGGGAGACCGGAATTCAGATTGATGAAAAUGGCGG

CTACGTAGCTCAGTTGG-3’ ptRNAMet(AUGGUU)21A Forward Primer

5’-

TAATACGACTCACTATAGGGAGACCGGAATTCAGATTGATGAAATGGTTGGC

TACGTAGCTCAGTTGG-3’ ptRNAMet(AUAUCU)21A Forward Primer

5’-

TAATACGACTCACTATAGGGAGACCGGAATTCAGATTGATGAAATATCTGGC

TACGTAGCTCAGTTGG-3’ ptRNAMet(AAUUGC)21A Forward Primer

5’-

TAATACGACTCACTATAGGGAGACCGGAATTCAGATTGATGAAAATTGCGGC

TACGTAGCTCAGTTGG-3’ ptRNAMet(CACGAU)21A Forward Primer

5’-

TAATACGACTCACTATAGGGAGACCGGAATTCAGATTGATGAACACGATGG

CTACGTAGCTCAGTTGG-3’

53 ptRNAMet(AUAUUU)21A Forward Primer

5’-

TAATACGACTCACTATAGGGAGACCGGAATTCAGATTGATGAAATATTTGGC

TACGTAGCTCAGTTGG-3’ ptRNAMet(AUAUAU)21A Forward Primer

5’-

TAATACGACTCACTATAGGGAGACCGGAATTCAGATTGATGAAATATATGGC

TACGTAGCTCAGTTGG-3’ ptRNAMet(CACGCU)21A Forward Primer

5’-

TAATACGACTCACTATAGGGAGACCGGAATTCAGATTGATGAACACGCUGG

CTACGTAGCTCAGTTGG-3’ ptRNAMet Reverse Primer

5’-TGGTGGCTACGACGGGAT-3’

Multiple turnover and single turnover reaction kinetics

To visualize the substrate and products of the reactions the pre-tRNA substrates were 5’ end labeled with γ-32P. Multiple turnover reactions were performed in 50 mM Tris-HCl pH 8, 100 mM NaCl, 0.005% Triton X-100, and 17.5 mM MgCl2. To obtain the observed second order rate constants for the wild-type and randomized pre-tRNAMet populations, the enzyme complex was formed by adding 10 nM of P RNA, heating to 95˚C for 3 min, then incubating at 37˚C for 10 min before adding MgCl2 and another 10 min before adding 10 nM of C5 protein.

The substrate containing solution was formed separately by adding 2 µM of pre- tRNA spiked with a negligible amount of 32P labelled pre-tRNA, heating to 95˚C for

3 min, then incubating at 37˚C for 10 min before adding the MgCl2. Reactions were started by using equal volumes of the enzyme and substrate to achieve the final concentrations of 5 or 10 nM RNase P holoenzyme and 1 µM pre-tRNA substrate.

Aliquots of 5µL were taken at given timepoints and the reaction was stopped in equal volume of formamide loading dye + 100 mM EDTA on dry ice. Samples were run on a 15% denaturing polyacrylamide gel at 80 watts then dried and exposed to a phosphorimager screen. Radioactivity was quantified using

ImageQuant software and fraction of reaction calculated by taking the amount of product at each timepoint divided by the addition of substrate and product bands.

Controls in which enzyme was omitted from the reaction were also run with each experiment and showed no cleavage of substrate. Rate constants for the randomized populations were calculated by fitting the data to single exponential lines in Origin software.

For individual substrate assays to measure absolute kcat/Km, reactions were performed as above except the final concentration of substrate was 3µM and holoenzyme 5 nM. After quantification, these reactions were fit to the integrated

Michaelis-Menten equation to obtain the kcat/Km and datapoints from the end of the reaction were removed until the fit to the line reached an Adj. R2 of at least 0.98.

The single turnover rate constants from these data were taken from the initial rate

55 portion of these reactions where substrate was less than 15% reacted and were fit to a linear function.

Single turnover reactions were performed in 50 mM MES pH 6.0, 100 mM

NaCl, 0.005% Triton X-100, and 17.5 mM MgCl2. To obtain the first order rate constants for wildtype and randomized pre-tRNAMet populations, reactions were set up exactly as above with multiple turnover conditions except the final concentration of holoenzyme was 100 nM, which is expected to be saturating conditions for holoenzyme above the reported Km, and substrate concentrations were below 10 nM in these reactions and contained only 32P-labelled RNA. To confirm saturation, single turnover reactions were performed by increasing the holoenzyme concentration to 1 µM in reactions with less than 10 nM pre-tRNA, the rate constant was unchanged. Reactions were quantified as described above for multiple turnover conditions.

High-Throughput Sequencing Kinetics (HTS-Kin)

Multiple turnover HTS-Kin reactions were performed exactly as described above, except they were scaled up 10-fold in volume to allow for sufficient material for subsequent analysis by next generation sequencing. Aliquots from reaction timepoints were 160 µL and the reaction stopped by placing in 33 mM EDTA on dry ice. To concentrate samples, phenol-chloroform extraction and ethanol precipitation was performed as described. The purified RNA was then run on a

10% denaturing polyacrylamide gel, bands corresponding to the residual substrate population were identified by exposing the gel to x-ray film and were isolated (gels

56 were also exposed to phosphorimager screen for later quantification). The excised gel bands containing the residual substrate population were extracted from the gel and placed in 10 mM Tris-HCl pH 8, 1 mM EDTA, and 0.1% SDS at 4˚C overnight to elute the pre-tRNA. The substrate RNA was then equalized between timepoints based on the fraction reacted. First-strand synthesis was performed using 5 µL of the equalized RNA and 1µM reverse primer. Reactions were placed at 72˚C for

10 min and placed on ice for 1 min before adding 100U of SuperScript III reverse transcriptase, and 0.75 mM dNTP mix, and 2.5 mM DTT, and 1x supplied RT Buffer before beginning incubation at 42˚C for 10 min, 50˚C for 40 min, and 55˚C for 20 min before heat inactivating the enzyme at 95˚C for 5 min. Samples were then diluted 1:300 and 1µL of this dilution was used to amplify for high-throughput sequencing. Polymerase chain reaction was performed with 1.5 U Taq DNA polymerase, 0.5 µM forward primers that bound to the 21nt sequence at the 5’ end and contained a barcode for each timepoint and randomized dinucleotide sequence, 0.5 µM reverse primers complementary to the 3’ end, 1x supplied PCR buffer, and 0.2 mM dNTP mix. To reduce bias from amplification, the number of

PCR cycles was selected based on the minimal number needed to visualize the product on a 1% agarose gel stained with ethidium bromide. The PCR protocol consisted of incubation at 95˚C for 2 min followed by 14-18 cycles of 95˚C for 30 sec, 55˚C for 45 sec, and 72˚C for 1 min, followed by a final incubation at 72˚C for

5 min. The PCR products were then purified by centrifugation through Amicon

Ultra-0.5 Centrifugal Filter Devices. Samples were then sent for multiplexed

Illumina sequencing in 75 bp single end reads on Hi-Seq 2500 instrument using

57 equimolar amounts of each timepoint. Single turnover HTS-Kin reactions were set up exactly as described above for single turnover reactions and were not scaled up.

Primer sequences are as follows:

Reverse Transcription Primer

5’-CAAGCAGAAGACGGCATACGATGGTGGCTACGACGGGAT-3’

Forward PCR Primer for 21A Substrates (underlined letters indicate the indexed barcode)

5’-