1 Characterizing the Function of Alanyl-Trna Synthetase Activity In

Characterizing the Function of Alanyl-tRNA Synthetase Activity in Microbial Translation

Dissertation

Presented in Partial Fulfillment of the Requirements for the Degree Doctor of Philosophy in the Graduate School of The Ohio State University

Paul Michael Kelly, MS

Graduate Program in Molecular, Cellular and Developmental Biology

The Ohio State University

2020

Dissertation Committee:

Dr. Michael Ibba, Advisor

Dr. Kurt Fredrick

Dr. Jane Jackman

Dr. Natividad Ruiz

Copyrighted by

Paul Michael Kelly

2020

Abstract

Across all domains of life, mechanisms have evolved to ensure faithful translation of the genetic code. One integral step in the transition from a nucleic acid encoded-genome to functional proteins is the aminoacylation of tRNA molecules. To perform this activity, aminoacyl-tRNA synthetases (aaRSs) activate free amino acids in the cell forming an aminoacyl-adenylate before transferring the amino acid on to its cognate tRNA. These newly formed aminoacyl-tRNA (aa-tRNA) can then be used by the ribosome during mRNA decoding. In Escherichia coli, there are twenty aaRSs encoded in the genome, each of which correspond to one of the twenty proteinogenic amino acids used in translation.

Given the shared chemicophysical properties of many amino acids, aaRSs have evolved mechanisms to prevent erroneous aa-tRNA formation with non-cognate amino acid substrates. Of particular interest is the post-transfer proofreading activity of alanyl-tRNA synthetase (AlaRS) which prevents the accumulation of Ser-tRNAAla and Gly-tRNAAla in the cell. Upon mis-activation of the non-cognate serine or glycine, AlaRS will transfer the amino acid to the 3’ end of tRNAAla where it will translocate into a distinct editing active site in the enzyme. Recognition of the non-cognate substrate will lead to hydrolysis of the amino acid from the tRNA leaving both substrates available for subsequent rounds of aminoacylation. It has been previously shown that disruption of the cytosolic and mitochondrial AlaRS editing domains in mice leads to neurodegeneration and embryonic

ii lethality respectively. Here we show that perturbation of the E. coli AlaRS editing domain causes gross perturbation to the E. coli proteome causing a variety of pleiotropic phenotypes including growth defects, motility impairment, and antibiotic sensitivity.

Furthermore, we have identified second-site suppressors within the AlaRS editing domain that alleviate these defects. This work highlights the importance of AlaRS fidelity in E. coli and characterizes novel elements within this aaRS editing domain.

Given the essentiality for maintaining protein synthesis, aaRSs have been promising targets for therapeutic discovery. Leishmania major is a eukaryotic parasite that infects upwards of a million individuals a year. Current therapeutic options remain limited and have adverse effects on humans. While antibiotics are able to target prokaryotic cellular components specifically, developing therapies against eukaryotic pathogens is more challenging due to conservation between potential pathogen drug targets and their host counterparts. Previous studies have identified anti-fungal therapies that target pathogen aaRS with some degree of specificity, indicating that these essential enzymes could be developed as drug targets for a variety of infections. Using bioinformatic approaches, we have identified several aaRSs, including AlaRS in L. major that appear to be potential targets for anti-leishmanial therapies. Preliminary biochemical analysis have supported these in silico predictions and led to the identification of novel aaRS inhibitors in vitro.

iii

Dedication

The body of work presented herein is dedicated to my Grandmothers (Patricia Kelly and

Mary Sopira). This degree would have never been possible without their love, support,

and continued belief in me. I hope I’ve made them proud.

Acknowledgments

First and foremost, I’d like to acknowledge my Mother (Donna Bittner), Father

(Shawn Kelly), and Stepfather (Butch Bittner) for raising me in a home in which my aspirations were always supported and met with love and encouragement. They may have not understood what I was doing, but they were always the first to ask “How are your fishes doing?” when I was monitoring kidney regeneration in zebrafish, or “Are the E. coli growing?” while I was studying the role of prokaryotic translational fidelity (by the time you finish Chapter 3, you will see that the E. coli were in fact not always growing particularly well). Jokes apart, the pursuit of my academic ventures has rarely been easy and most of the time, incredibly isolating. Without the love and support of the many family and friends who have always been there for me, this journey would have likely ended long ago.

I would also like to acknowledge my committee members, Drs. Kurt Fredrick, Jane

Jackman, and Natacha Ruiz for their continued guidance over the last few years. Their advice has been invaluable towards the completion of my dissertation.

To my lab mates, words cannot describe how fortunate I feel to have gotten to work with you all. The friendships made over the last six years are some of the best I’ve ever had, and I consider you all like family (albeit a very loud and dysfunctional one). The dynamic of the lab has changed over the years and I am grateful for everyone who has been

v a part of this journey. Early on, I had the great fortune to learn from Drs. Tammy

Bullwinkle and Kyle Mohler. They taught me how to think about science in new and interesting ways and I would not be the scientist I am today without their tutelage. The lab then transitioned to what I consider the “core” years when I got to witness the creativity and tenacity of Drs. Annie Witzky, Becky Steiner, and Rodney Tollerson II on a daily basis. The best way to describe Annie, Becky, and Rodney is that they are “winners,” and that winning mindset continues to inspire me as I plan for the next chapter of my life. Most recently, my role in the lab has changed to that of a mentor for our younger graduate students. I have had the opportunity to witness Nien-Ching Han, Aru Kavoor, and Mary

Cranley grow as researchers and begin to find their scientific voice. Of the many great things that have happened in the lab, knowing that I may have played any role in their development is one of my proudest achievements. To any other lab members that I didn’t mention by name (sorry, I figured I should wrap this up because I’m sure Mike is already starting to doze off), just know that you have all contributed to this work and I am grateful for the time we have spent together.

Finally, I owe my greatest debt to Dr. Michael Ibba because without him, this body of work would not exist. Throughout my life, I have been plagued with unrelenting self- doubt. Mike has always been there to support and encourage me even when I had given up on myself. For the first few years, I was convinced that I had “tricked” Mike into letting me join his lab. I couldn’t figure out how I could possibly be good enough to get to work with these amazing people and I kept waiting for the day that he would realize his mistake and I would finally be asked to leave. Well, that day never came and here I am ready to

vi move on to the next stage of my life, knowing that I provide value and without any fear of my inadequacies. Mike is the greatest mentor I could have ever asked for, he has changed the trajectory of my life, and I am eternally grateful.

vii

Vita

2008 ...... Taylor Allderdice High School

2010 ...... Community College of Allegheny County

2012 ...... B.S. Biology, Robert Morris University

2014 ...... M.S. Biology, Indiana University of Pennsylvania

2014 to present ...... Graduate Teaching and Research Associate,

Molecular, Cellular and Developmental Biology,

The Ohio State University

Publications

Kelly, P., Hadi-Nezhad, F., Liu, D., Lawrence, T., Linington, R., Ibba, M., Ardell, D., (2020) Targeting tRNA-Synthetase Interactions towards Novel Therapeutic Discovery Against Eukaryotic Pathogens. PLoS Negl. Trop. Dis. 14(2): e0007983

Kelly, P., Backes, N., Mohler, K., Buser C., Kavoor, A., Rinehart, J., Phillips, G., Ibba. M., (2019) Alanyl-tRNA Synthetase Quality Control Prevents Global Dysregulation of the Escherichia coli Proteome. mBio. 10 (6)

Kelly, P., Ibba, M., (2018) Aminoacyl-tRNA Quality Control Provides a Speedy Solution to Discriminate Right from Wrong. J. Mol. Biol. 430(1).

viii

Fields of Study

Major Field: Molecular, Cellular, and Developmental Biology

Table of Contents

Abstract ...... ii Dedication ...... iv Acknowledgments ...... v Vita ...... viii List of Tables...... xiv List of Figures ...... xv List of Symbols and Abbreviation ...... xvii Chapter 1. Introduction ...... 1 1.1 Aminoacyl-tRNA Synthetases ...... 2 1.1.1 Aminoacyl-tRNA Synthetase Activity ...... 2 1.1.2 Aminoacyl-tRNA Synthetase Proofreading ...... 5 1.1.3 Aminoacyl-tRNA Freestanding Proofreading Factors ...... 9 1.2 Alanyl-tRNA Synthetase ...... 11 1.2.1 Alanyl-tRNA Synthetase Structure and Regulation ...... 11 1.2.2 Alanyl-tRNA Synthetase Substrate Recognition ...... 12 1.2.3 Alanyl-tRNA Synthetase Proofreading ...... 14 1.2.4 Alanyl-tRNA Synthetase-Associated Proofreading Factors...... 18 1.2.5 Exploring the Physiological Cost of Alanyl-tRNA Synthetase Errors ...... 20 Chapter 2 Alanyl-tRNA Synthetase Quality Control Prevents Global Dysregulation of the Escherichia coli Proteome ...... 21 2.1 Introduction ...... 21 2.2 Results ...... 25 2.2.1 High levels of serine miscoding are toxic to E. coli ...... 25 2.2.2 AlaRS proofreading is required for optimal growth in E. coli...... 28 2.2.3 AlaRS fidelity is required for maintaining proteome homeostasis ...... 32 x

2.2.4 Mistranslation disrupts regulation of the translational machinery ...... 35 2.2.5 Reduced AlaRS fidelity impairs swimming motility ...... 39 2.2.6 AlaRS-mediated mistranslation alters the cell membrane ...... 42 2.3 Discussion ...... 45 2.3.1 AlaRS fidelity is evolutionarily protected ...... 45 2.3.2 Toxic mistranslation may provide a new target for drug discovery ...... 46 2.3.3 A threshold exists for neutral/beneficial mistranslation events ...... 46 2.4 Methods ...... 48 2.4.1 Strain Construction and Reagents ...... 48 2.4.2 Growth Analysis ...... 52 2.4.3 Construction of Mistranslating tRNASer Plasmids ...... 52 2.4.4 In vivo Mistranslation Reporter ...... 53 2.4.5 Total Proteome Analysis ...... 53 2.4.6 Immunoblotting ...... 55 2.4.7 Transcript Analysis ...... 56 2.4.8 Recombinant AlaRS Purification ...... 57 2.4.9 Active site and Thermal Stability...... 58 2.4.10 Swimming Motility ...... 59 2.4.11 Antibiotic Sensitivity ...... 60 2.4.12 Transmission Electron Microscopy ...... 60 Chapter 3 Second-Site Suppressors in AlaRS Rescue AlaRS C666A-Associated Phenotypes ...... 61 3.1 Introduction ...... 61 3.2 Results ...... 63 3.2.1 Identification of Suppressor Mutations that Alleviate the AlaRS C666A Growth Defect ...... 63 3.2.2 Second-site AlaRS mutations alleviate the slow-growth AlaRS C666A phenotype ...... 67 3.2.3 AlaRS C666A suppressors do not prevent serine mistranslation ...... 70 3.2.4 Phenotypes associated with AlaRS-mediated proteome dysregulation are alleviated in with the second-site suppressor mutations ...... 72 3.2.5 In vitro characterization of the AlaRS C666A suppressors ...... 75 3.3 Discussion ...... 79

3.3.1 AlaRS R561C reduces Ser-tRNAAla formation ...... 79 3.3.2 AlaRS R561S has elevated proofreading activity ...... 80 3.3.3 AlaRS D551Y increases Ser-tRNAAla deacylation in trans ...... 81 3.3.4 Conclusions...... 81 3.4 Methods ...... 82 3.4.1 General Methods ...... 82 3.4.2 Suppressor Mutant Characterization ...... 82 3.4.3 Strain Construction...... 83 3.4.4 Growth Analysis ...... 84 3.4.5 Motility Assays ...... 84 3.4.6 Antibiotic Sensitivity...... 84 3.4.7 In vivo Mistranslation Reporter ...... 85 3.4.8 Preparation of Recombinant Protein and in vitro Transcribed tRNAAla ...... 85 3.4.9 Proofreading Activity ...... 86 Chapter 4 Targeting tRNA-Synthetase Interactions towards Novel Therapeutic Discovery Against Eukaryotic Pathogens ...... 88 4.1 Introduction ...... 88 4.2 Results ...... 92 4.2.1 Custom Annotation of tRNA Genes and Gene Clusters in TriTrypDB Genomes ...... 92 4.2.2 Divergent Class-Informative Features between Humans and TriTrypDB Genomes ...... 99 4.2.3 AaRS Activity Screen Identified Leishmania major AlaRS Inhibitors ...... 108 4.2.4 Natural Product Library Inhibitors of Leishmania major ThrRS...... 111 4.2.5 Predictive Network Interactions Identified Broad-Spectrum Anti- Trypanosomal Targets...... 113 4.2.6 Separation of Active Components from Natural Products Extracts ...... 115 4.3 Discussion ...... 119 4.3.1 Systems-biology driven identification of Trypanosome-specific drug targets ...... 119 4.3.2 Inhibition of aminoacyl-tRNA synthetases ...... 120 4.3.3 Chemotherapeutic inhibition of multiple aminoacyl-tRNA synthetases may be relatively resistance-proof ...... 121 4.4 Methods ...... 122 xii

4.4.1Annotation, Clustering and Filtering of tRNA Genes in TriTrypDB Genomes ...... 122 4.4.2 Prediction of Divergent tRNA Class-Informative Features (CIFs) in Humans and Parasites ...... 123 4.4.3 AaRS Cloning and Protein Purification ...... 126 4.4.4 Preparation of in vitro Transcribed tRNA ...... 127 4.4.5 Marine Natural Product Library...... 128 4.4.6 Screen for Aminoacylation Inhibitors ...... 129 4.4.7 Pyrophosphate Exchange ...... 130 4.4.8 Bacterial Fermentation and Natural Product Extraction ...... 130 4.4.9 HPLC-UV-MS and UPLC-ESI-qTOF-MS Analyses ...... 132 Chapter 5. Discussion and Future Outlooks ...... 133 5.1 Regulation of post-translational modification of aaRS ...... 134 5.2 Broader roles for aaRS quality control ...... 137 5.3 Targeting aaRS fidelity for therapeutics ...... 140 5.4 Conclusion ...... 145 Bibliography ...... 147

xiii

List of Tables

Table 1. Classes of aminoacyl-tRNA synthetases ...... 4 Table 2. KEGG pathway analysis of proteome changes ...... 34 Table 3. Strains and plasmids used in this study ...... 51 Table 4. Statistics on final tRNA annotation gene set ...... 94 Table 5. Clades and genomes analyzed with statistics on CIF estimation gene sets ...... 98

xiv

List of Figures

Figure 1. Aminoacyl-tRNA synthetase activities ...... 3 Figure 2. Serine miscoding is toxic in E. coli...... 27 Figure 3. AlaRS fidelity is required for optimal growth in E. coli...... 31 Figure 4. AlaRS-mediated mistranslation disrupts proteome homeostasis ...... 33 Figure 5. AlaRS fidelity regulates AlaRS protein levels ...... 37 Figure 6. The AlaRS C666A proteins does not have elevated stability ...... 38 Figure 7. Swimming motility is impaired in the absence of AlaRS fidelity ...... 41 Figure 8. The E. coli membrane is affected when AlaRS fidelity is impaired ...... 44 Figure 9. AlaRS C666A suppressors map to the AlaRS proofreading domain ...... 66 Figure 10. Suppressor mutants alleviate AlaRS C666A-associated growth defects ...... 69 Figure 11. AlaRS C666A second-site suppressors do not prevent mistranslation ...... 71 Figure 12. Suppressor mutants alleviate AlaRS C666A-associated phenotypes ...... 74 Figure 13. In vitro characterization of AlaRS C666A suppressors ...... 78 Figure 14. tRNA gene cluster size distribution for Leishmania, Trypanosoma, and other TriTrypDB genomes ...... 95 Figure 15. Conserved divergence of parasite tRNAAla ...... 101 Figure 16. Conserved divergence of parasite tRNAThr ...... 102 Figure 17. Adenine function logos for humans and four clades of Leishmania ...... 103 Figure 18. Function logos for tRNA Class-Informative Base-Pairs and Mis-Pairs in humans ...... 104 Figure 19. Function logos for tRNA Class-Informative Base-Pairs and Class-Informative Mis-Pairs in L. major clade ...... 105 Figure 20. Function logos for tRNA Class-Informative Base-Pairs and Class-Informative Mis-Pairs in T. cruzi clade ...... 106 Figure 21. CIFs for tRNATyr and tRNATrp are strongly conserved between parasites and humans...... 107 Figure 22. Identification of Leishmania major AlaRS inhibitors ...... 110 Figure 23. Identification of Leishmania major ThrRS inhibitors ...... 112 xv

Figure 24. Leishmania major and Trypanosoma cruzi AlaRS have conserved tRNA identity elements...... 114 Figure 25. Workflow for the extraction of RL12-182-HVF-D ...... 117 Figure 26. Marine natural product extract 2096D F10 inhibits Leishmania major AlaRS aminoacylation...... 118 Figure 27. AlaRS fidelity is required for accurate tmRNA aminoacylation...... 139 Figure 28. Proposed model for targeting AlaRS fidelity for chemosensitization ...... 144

xvi

List of Symbols and Abbreviation

AMP adenosine 5’-monophosphate

ATP adenosine 5’-triphosphate

Ala L-alanine

AlaRS alanyl-tRNA synthetase aaRS aminoacyl-tRNA synthetase aa-tRNA aminoacyl-tRNA

Amp ampicillin

Asn L-asparagine

Asp L-aspartic acid

β beta

BME β-mercaptoethanol

BSA bovine serum albumin

CaCO3 calcium carbonate

CaCl2 calcium chloride

Cam chloramphenicol

CD circular dichroism

CIMP class-informative mis-pairs

CIPB class-informative base-pairs

CIF class-informative features

xvii

Cys L-cysteine

CysRS cystinyl-tRNA synthetase

°C degrees Celsius

DAP diaminopimelic acid

CH2Cl2 dichloromethane

DMSO dimethyl sulfoxide

DDT dithiothreitol

Δ delta/deletion

DNA deoxyribonucleic acid

Dtd D-aminoacyl-tRNA deacylase

EtOH ethanol

EtOAc ethyl acetate

EDTA ethylenediaminetetraacetic acid

EF-Tu elongation factor Tu

Gln L-glutamine

Glu L-glutamic acid

Gly L-glycine

GlyRS glycyl-tRNA synthetase

GMP guanosine 5’monophosphate

GTP guanosine 5’-triphosphate

HPLC high-performance liquid chromatography

HisRS histidyl-tRNA synthetase

HRP horseradish peroxidase

HCl hydrochloric acid xviii

ID information difference

IleRS isoleucyl-tRNA synthetase

IPTG isopropyl β-D-thiogalactoside

Kan kanamycin kDa kilodalton

KLD Kullback-Leibler divergence

LB lysogeny broth

LeuRS leucyl-tRNA synthetase

MgCl2 magnesium chloride

MgSO4 magnesium sulfate

MNP marine natural product

MS mass spectrometry

MeOH methanol

µ micro

µL microliter

µM micromolar mL milliliter mM millimolar

M molar ng nanogram nm nanometer

NMR nuclear magnetic resonance

Nva norvaline

OD optical density xix

ORF open reading frame

PCR polymerase chain reaction

Phe L-phenylalanine

PheRS phenylalanyl-tRNA synthetase

KCl potassium chloride

Pro L-proline

ProRS prolyl-tRNA synthetase

PPi inorganic pyrophosphate

RNA ribonucleic acid

Ser L-serine

NaCl sodium chloride

SDS sodium dodecyl sulfate

NaF sodium fluoride

Thr L-threonine

ThrRS threonyl-tRNA synthetase tRNA transfer RNA tmRNA transfer-messenger RNA

TEM transmission electron microscopy

TCA trichloroacetic acid

TFA trifluoroacetic acid

Tyr L-tyrosine

TyrRS tyrosyl-tRNA synthetase

ValRS valyl-tRNAsynthetase

Chapter 1. Introduction

Throughout all domains of life, mechanisms have evolved to minimize errors during cellular growth. With strong selective pressures to prevent errors directly to the genome, errors in replication occur most infrequently, with mutations occurring ~10-10 per nucleotide per generation [1]. After DNA replication, errors in transcription have been reported to occur between 10-4-10-5 per nucleotide [2], finally followed by translational errors which have historically been reported to occur 10-4 times per amino acid [3]. Recent advances in quantitative proteomics have further suggested that initial reports of translational errors may have been under-estimated, with error rates closer to 1 in every

1000 translated amino acids [4]. One possible explanation for the high translational error rates is the complex multi-step process by which protein synthesis occurs. The synthesis of new proteins is an orchestrated process requiring specific, and ordered interactions between free amino acid substrates, tRNAs, enzymes, and energy sources to generate aminoacyl-tRNAs (aa-tRNAs) that can be used for peptide bond formation in the ribosome.

While the steps required for accurate peptide bond formation present their own unique and complex challenges, the focus of this work will be on the regulation of the cellular aa- tRNA pool by the aminoacyl-tRNA synthetases (aaRSs).

1.1 Aminoacyl-tRNA Synthetases

1.1.1 Aminoacyl-tRNA Synthetase Activity

The evolutionarily conserved aaRSs are responsible for pairing free amino acids in the cell to their correct tRNA. In Escherichia coli, there are twenty aaRSs, each corresponding to one of the twenty proteinogenic amino acids. AaRS perform their canonical activity in two steps (reviewed in [5]): first, free cognate amino acid is activated by ATP in the aaRS active site, forming an aminoacyl-adenylate. Second, the amino acid is transferred to the 3’ end of the cognate tRNA (Fig. 1). This newly formed aa-tRNA is now able to participate in ternary complex formation with elongation factor (EF-Tu in bacteria) and GTP for ribosomal decoding. Despite their shared enzymatic activity in ligating free amino acids to their corresponding tRNAs, the aaRSs can be divided into two separate and evolutionarily distinct classes of enzymes, the Class I aaRSs and the Class II aaRSs (Table 1) [6].

Figure 1. Aminoacyl-tRNA synthetase activities Aminoacyl-tRNA synthetases function in two steps. First, cognate amino acids are activated with ATP creating an aminoacyl- adenylate. Second, the amino acid is transferred to the cognate tRNA to be used in translation. Due to the similarities between many amino acid substrates in the cell, some aaRSs have evolved secondary activities to prevent mis-acylation on tRNAs. Pre- transfer proofreading occurs by hydrolyzing the mis-activated amino acid. Alternatively, mis-aminoacylated tRNA substrates can be hydrolyzed following the transfer onto tRNA. 3

Table 1. Classes of aminoacyl-tRNA synthetases

Class I aaRSs are characterized by their Rossmann nucleotide-binding domain and the presence of the conserved KMSKS and HIGH amino acid motifs [7]. Mutational analysis of these motifs has suggested they are important for aaRS function [8]. Another distinguishing feature of Class I aaRSs is the position of amino acid linkage on the tRNA.

All Class I aaRS transfer their amino acid substrate on the 2’ OH of the ribose of the terminal adenosine of the tRNA [9].

In contrast, Class II aaRS are structurally characterized by a series of antiparallel

β-strands in the active site and three degenerate sequence motifs (numbered 1, 2, and 3) that facilitate subunit dimerization, adenylate formation, and tRNA binding [10,11]. With the exception of PheRS which has been shown to aminoacylate phenylalanine on the 2’

OH of tRNAPhe, all other Class II aaRSs ligate their corresponding amino acids on the 3’

OH of the terminal ribose [9]. Regardless of initial position of aminoacylation, all amino acids will eventually be positioned on the 3’ OH via transesterification to participate in peptide bond formation in the ribosome.

1.1.2 Aminoacyl-tRNA Synthetase Proofreading

Given the many diverse metabolites found across the cellular milieu, aaRSs generally have high specificity for their cognate substrates [12]. While most substrates are occluded from aaRS active site, due to the shared physicochemical properties of many amino acids, some non-cognate substrates can interact within the amino acid binding pocket. To prevent errors associated with mis-activation, aaRSs have evolved mechanisms to discriminate against non-cognate substrates prior to their participation in translation. It

5 has been suggested that if a non-cognate substrate has an amino acid specificity value (i.e.

(kcat cognate/KM cognate)/(kcat non-cognate/KM non-cognate)) less than 3000:1, the enzyme will likely require proofreading mechanisms to prevent the accumulation of mis- aminoacylated tRNA products [13,14]. Two distinct proofreading activities have been observed to contribute to accurate aaRS fidelity (reviewed in [15]). Some aaRSs, including several of the Class I aaRS, utilize pre-transfer proofreading to prevent the accumulation of mis-activated adenylate products. In this process, the non-cognate substrate will be mis- activated, but prior to its transfer onto tRNA, the adenylate will be hydrolyzed, releasing free amino acid and AMP back into the cell. Alternatively, several aaRSs utilize a distinct post-transfer proofreading mechanism to maintain accurate aa-tRNA fidelity. For aaRS which employ post-transfer proofreading, the mis-activated adenylate will be transferred to the 3’ end of the tRNA. Upon transfer, the mis-aminoacylated tRNA will translocate into a separate active site which monitors accurate aa-tRNA products. Mis-aminoacylated products will be hydrolyzed releasing amino acid and tRNA back into the cell for iterative rounds of amino acid activation and aminoacylation (Fig. 1).

Some of the first characterized proofreading activities were those of the Class I aaRSs which are responsible for pairing the branch chained amino acids, ValRS [16,17],

LeuRS [18], and IleRS [19]. Early observations indicated that all three of these aaRS have the potential for pre-transfer proofreading and post-transfer proofreading through their conserved CP1 insertion domain (reviewed in [15]). While the specificity for the non- cognate branch chain amino acids would suggest that LeuRS and IleRS will both require proofreading, recent work has highlighted a new non-cognate substrate for both LeuRS and

IleRS, the non-proteinogenic amino acid norvaline. Compared to the specificity for isoleucine, norvaline is mis-activated by LeuRS with a discrimination factor of 116 compared to ~30000, respectively [20,21]. This suggests that norvaline may be the primary proofreading substrate for LeuRS in vivo. The biochemical mechanisms for these norvaline discrimination events has been interrogated using both pre-steady state and steady state kinetics [22]. Beyond the mechanistic insights gained from these works, it was observed in both bacterial [21,23] and eukaryotic models [24] that genetic perturbation to the intrinsic proofreading activities in LeuRS led to a growth defect upon norvaline exposure.

In addition to the proofreading activities that have been characterized for Class I aaRSs, the proofreading mechanisms for many of the Class II aaRSs have also been well investigated. With phenylalanine and tyrosine differing by only the addition of the terminal hydroxyl group, PheRS employs post-transfer proofreading to prevent the accumulation of

Tyr-tRNAPhe in the cell [25]. Besides GlyRS, PheRS is the only other known aaRS that can oligomerize into a heterotetramer, with both the activation site domain and proofreading domains being translated from two different genes. Interestingly, many PheRS proteins across the phylogenetic landscape encode for sufficiently-discriminatory catalytic active sites which greatly diminishes the need for PheRS proofreading for proteinogenic para- tyrosine [26]. In contrast, it was recently shown that this discriminatory active site does not prevent the mis-activation of another tyrosine isomer, meta-tyrosine, which is a non-protein amino acid produced by some grass species [27] or during phenylalanine oxidation [28].

Abolishment of E. coli PheRS proofreading led to significant growth defects [28] and dysregulation of stringent response activation when supplemented with meta-tyrosine [29].

Similar to the observations from those studying LeuRS and IleRS proofreading, the role for the evolutionary conservation of many aaRSs proofreading activities is not fully yet known as they may be required for non-protein amino acid discrimination [30].

The Class II aaRS ThrRS has also served as a model for studying post-transfer proofreading, first for the mechanistic detail in which the proofreading activity was characterized and subsequently for its unique potential in modulating proofreading activity in vivo. Given the shared terminal hydroxyl group of threonine and serine, serine is believed to be the primary proteinogenic non-cognate substrate for ThrRS discrimination.

Within the ThrRS proofreading domain, a catalytic active site of conserved HXXXH and

CXXXH motifs are utilized to hydrolyze Ser-tRNAThr. Upon translocation of mis- aminoacylated Ser-tRNAThr from the active site to the proofreading domain, the 3’end of the aa-tRNA is stabilized by the binding of Cys182 to the 2’ OH of the terminal ribose, and serine is hydrolyzed from tRNAThr [31]. Alanine substitutions at Cys182 led to an increase in Ser-tRNAThr product formation in vitro, suggesting an important role for tRNA stabilization prior to hydrolysis [31]. More recent characterization of this proofreading activity in vivo has suggested that the Thr to Ser substitution is well tolerated in the proteome. Chromosomal substitution of the C182 to an alanine, led to no noticeable defects in E. coli growth [32]. To confirm that the ThrRS C182A variant was causing Thr to Ser translational errors, a mass spectrometry-based reporter was employed to confirm that mistranslation was occurring [32]. Beyond errors mediated by genetic manipulation of aaRS proofreading domains, it was further shown that oxidative stress causes oxidation of

C182 which also perturbs ThrRS fidelity [33]. While the role for this environmentally-

8 regulated change in translational fidelity is yet to be found, it does suggest that some translational errors are not only tolerated, but possibly beneficial under certain environmental conditions.

The suggestion that some translational errors may be evolutionarily tolerated is not entirely circumstantial (reviewed in [34]). Analyses of aaRS sequences across many parasitic organisms have indicated that several proofreading domains have been lost

[35,36]. Interestingly, these degenerate proofreading domains have primarily been observed in either the Class I LeuRS, or the Class II PheRS and ThrRS [37]. Furthermore, loss of aaRS proofreading has been observed to occur in up to two of the three aaRSs listed in any species, but never all three. This observation leads to an interesting interpretation that despite the loss of some aaRS proofreading, there still remains a threshold for translational fidelity.

1.1.3 Aminoacyl-tRNA Freestanding Proofreading Factors

Beyond the intrinsic proofreading found in half of the aaRSs in the cell (Table 1), several freestanding homologs of aaRS proofreading domains have been observed across many phyla, presumably to prevent particularly detrimental amino acid substitutions in the proteome. The identification of freestanding proofreading enzymes, also known as trans- editing factors led to the proposal of the “triple sieve model” for translational accuracy

[38]. In this model, all aaRSs will first discriminate most non-cognate substrates at the initial step of amino acid activation because they will not fit within the amino acid binding pocket. Secondly, if a non-cognate substrate can be mis-activated, additional proofreading

9 activities will have evolved to recognize those erroneous products and they will be removed from the substrate pool. Finally, if a non-cognate substrate can bypass initial active site discrimination and subsequent intrinsic aaRS proofreading, then additional factors will be required to correct these errors. The most characterized example for these activities are those found encoded in ProRS and its various trans-editing associated factors that aid in proline translational fidelity in the proteome. In E. coli, ProRS can mis-activate two non- cognate substrates, alanine and cysteine. Upon alanine mis-activation, E. coli ProRS has both pre- and post-transfer proofreading activities through its widely conserved INS domain [39]. While ProRS has proofreading mechanisms to prevent Pro to Ala mistranslation, ProRS itself is unable to discriminate against Cys-tRNAPro. To prevent cysteine mistranslation events, the trans-editing factor Ybak will recognize the mis- aminoacylated tRNA and hydrolyze the cysteine by thiol side chain discrimination [38,40].

Some organisms such as Caulobacter crescentus have truncated INS domains and encode a secondary trans-editing factor, ProXp-Ala which acts a freestanding INS domain to deacylate Ala-tRNAPro in the cell [41]. In addition to the ProRS-associated trans-editing factors that have been identified to date, most of the other characterized freestanding proofreading enzymes are those that prevent errors mediated by AlaRS. In the following sections, the structure, substrate recognition, and proofreading activities of AlaRS will be described in more detail.

1.2 Alanyl-tRNA Synthetase

1.2.1 Alanyl-tRNA Synthetase Structure and Regulation

Once assembled into its oligomeric structure, AlaRS is the largest aaRS in the cell.

In E. coli, one AlaRS monomer (96 kDa) is translated from the 2631 bp gene alaS. AlaRS was originally characterized as a homotetramer [42], with four AlaRS monomers oligomerizing primarily by interactions in the C-terminal domain [43] and contacts in motif

1 [44]. The structural designation as a homotetramer was somewhat controversial as several other groups had determined that AlaRS existed as a homodimer [45,46]. However, recent evidence has shown that AlaRS can either exist as a homodimer or a homotetramer in the cell and these changes in AlaRS oligomeric structure may play an important role in regulating its cellular activity [47].

It was previously observed that AlaRS has DNA binding potential and its canonical binding site was the 12 bp palindromic sequence found upstream of the transcription start site of the alaS gene [48]. Furthermore, this binding interaction was found to be dependent on the concentration of alanine present in the reaction [48]. The finding that amino acid substrate levels can regulate cognate aaRS transcription is not unique as the PheRS- encoding genes are also regulated by phenylalanine concentrations and translational readthrough of an upstream leader peptide [49]. However, the observation that AlaRS could directly act as a transcription factor for its own gene was the first example of this non-canonical function of an aaRS.

Recently the AlaRS:DNA binding interactions were comprehensively explored.

Through a series of biophysical experiments, it was shown that DNA binding was only

11 possible when AlaRS was in the higher-ordered tetrameric structure [47]. Furthermore, this binding was not alanine-dependent. The authors note that this discrepancy in alanine- dependency may be due to technical limitations in their experimental setup, however, it still raises the possibility that AlaRS-mediated alaS regulation may not be through intracellular alanine concentrations as has previously been assumed.

1.2.2 Alanyl-tRNA Synthetase Substrate Recognition

The formation of Ala-tRNAAla requires two separate interactions between AlaRS and its substrates. As a Class II aaRS, the AlaRS active site is composed of seven β-strands which include the three characteristic recognition motifs [50,51]. Structural studies have elucidated the AlaRS:amino acid active site interactions which facilitate cognate alanyl- adenylate formation, but also permit non-cognate substrate activation of both glycine and serine [52]. The active site pocket is primarily formed by R69 which interacts with the α- carboxyl group of all amino acid substrates and forms a salt bridge with the adenylate.

Other important interactions include N212 which binds between the aa-AMP and D235 which interacts with the α-amino group of all substrates. Importantly, D235 is evolutionarily conserved across all AlaRS species. This acidic group also facilitates the potential for hydrogen bonding with the terminal -OH group of non-cognate serine, and the consequence of serine mis-activation will be discussed in detail later [52].

In addition to discriminating cognate alanine against the other proteinogenic and non-proteinogenic substrates during amino acid activation, AlaRS must also maintain faithful interactions with cognate tRNAs in the cell. E. coli encodes five tRNAAla genes

(alaT, alaV, alaU, alaW, and alaX) which transcribe to two different tRNA isoacceptors that are able to base pair either complementarily or through wobble interactions with the four alanine codons in the cell (GCA GCG, GCU, and GCC).

Accurate aaRS:tRNA interactions are facilitated by “tRNA identity elements.”

These chemical and/or structural features inherent to a class of tRNAs will act as either positive or negative specificity determinants for aaRS interactions. Many tRNA identity elements are found at two distinct regions of the tRNA, in the acceptor stem and in the anticodon loop (reviewed in [5]). The utilization of two discriminatory tRNA features aids in accurate aaRS:tRNA interactions. Notably, AlaRS:tRNAAla interactions only require one identity element for AlaRS recognition and that is the unique evolutionarily conserved

G3:U70 wobble base pair in the acceptor stem found only in tRNAAla [53]. Furthermore, mutation of other tRNA acceptor stems to contain this G3:U70 base pair can facilitate alanylation on non-tRNAAla tRNAs [54]. Through a series of enzymatic analyses using wild-type and AlaRS variants from E. coli and humans, it was recently shown that despite the conservation of the G3:U70 base pair, there are mechanistic distinctions for how this recognition event occurs [55]. In E. coli, D400 in AlaRS interacts with the amino group on

G3 in the minor groove. This interaction acts as a positive discriminator for aaRS:tRNA interaction, as substitutions of this position greatly reduced aminoacylation activity. The other primary contact of this identity element is N303 with O4 of U70 in the major groove.

In contrast to D400, substitutions to N303 led to an increase in aminoacylation of the near- cognate G3:C70 Watson Crick base pair that is found in many tRNAs. Together, these results suggest that D400:G3 interaction acts as a primary determinant for AlaRS binding,

13 whereas the N303:U70 interaction is a negative determinant to prevent near-cognate aminoacylation. Despite the conservation of the Asp and Asn residues in human AlaRS, neither residue appeared to act as a positive determinant for tRNA binding as substitutions to either of these positions did not significantly affect aminoacylation. In contrast, both residues appear to act as negative determinants for near-cognate G3:C70 base pairs. This work provided insight into how evolutionary sequence conservation may not always be suggestive of mechanistic conservation [55]. More specifically, this work suggests that in higher eukaryotes, prevention of translational errors caused by mis-alanylation is potentially more important than elevated catalytic efficiency as observed in the bacterial enzyme.

1.2.3 Alanyl-tRNA Synthetase Proofreading

Similar to some of the other Class II aaRSs, AlaRS encodes for a distinct secondary proofreading active site to provide surveillance of accurate Ala-tRNAAla formation. As was described above, in addition to accommodating the cognate amino acid alanine in the primary active site, the AlaRS active site will also mis-activate two non-cognate substrates, serine and glycine [56]. While alanine is activated with a KM of 50 µM, serine and glycine are both relatively poor substrates for AlaRS, with KM values of 31 mM and 26 mM respectively [52]. Despite being poor substrates, there is a growing body of evidence that

AlaRS proofreading is required for active cellular growth as defects associated with AlaRS errors have been observed from bacteria to mice.

Similar to the characterized proofreading domain of ThrRS, AlaRS also has an evolutionarily conserved cysteine residue in the proofreading domain that is required for post-transfer non-cognate amino acid hydrolysis [31]. Early biochemical studies indicated that substitutions of that cysteine to an alanine (C666A in E. coli) led to elevated Ser- tRNAAla formation that would otherwise be removed by AlaRS proofreading [57]. AlaRS proofreading is facilitated by the coordination of the conserved HXXXH and CXXXH motifs. Furthermore, several additional residues have been characterized to participate in non-cognate substrate discrimination, including Q584 which interacts with the β-hydroxyl group of serine. Biochemical characterization of C666 suggests that this residue is predicted to make three possible interactions: coordination of a zinc atom in the active site, binding of the 2’ OH of the tRNA, and the α-amino group of serine [58].

Subsequent experimentation sought to understand the role of AlaRS proofreading in vivo. To address this question, a ΔalaS E. coli strain was used to show that when the native alaS gene is removed and E. coli is complemented with mutated alaS alleles encoding AlaRS proofreading-deficient variants, E. coli is sensitive to elevated serine stress [57]. This experiment provided the first evidence for the important role of AlaRS fidelity in vivo and suggested that the Ala to Ser substitution was not well tolerated in the cell.

In the following years, the Ackerman group published a series of papers which for the first time addressed the consequence of AlaRS proofreading defects in higher eukaryotes. The authors identified a mutant mouse during a genetic screen which appeared to have rough and matted fur. This mutant allele was termed sti and is more commonly

15 known as the sticky mouse given its irregular fur patterning. Genetic mapping of this mutant allele indicated a C2201A mutation in the Aars gene (mouse AlaRS gene). This mutation caused the substitution A734E in the AlaRS proofreading domain. Further investigation of the sti/sti homozygous mice indicated these mice presented with enhanced neurodegeneration by three weeks of age [59]. Additionally, histological analyses suggested that the observed neurological defects were due to the accumulation of protein aggregates, presumably from the accumulation of misfolded proteins caused by this mutant

AlaRS allele. Despite the striking neurodegenerative phenotype, biochemical analyses of the sticky AlaRS variant only led to a subtle defect in mis-serylation in vitro [59]. Together, these results suggested that minor defects in AlaRS proofreading likely leading to low levels of protein mistranslation are not well tolerated in terminally differentiated neuronal cells.

Subsequent experimentation looked to further exacerbate the mistranslation- associated phenotypes in the sticky mutant by further decreasing AlaRS fidelity in mice.

Preliminary attempts to generate homozygous AlaRS C723A (homologous to C666A in E. coli) variants in mice indicated that the combination of both alleles was embryonic lethal

[60]. Two additional mutants were generated that produced viable offspring in expected

Mendelian ratios but exhibited further tissue-specific defects beyond the sti/sti mouse.

Using a Cre/lox conditional knock-in allele, the authors could either repress one copy of the Aars gene, termed stop/sti, or induce expression of their mutant cassette, creating a heterozygous AlaRS C723A/sti mouse. Beyond the neurodegenerative defects associated with the homozygous sti mouse, these additional mutants presented with cardiac fibrosis

16 and protein aggregation in cardiomyocytes [60]. These bodies of work indicated that there is small margin in which defects in AlaRS proofreading can be tolerated in higher eukaryotes and at least one copy of partially-accurate AlaRS must be present for cytoplasmic translation in mice. Furthermore, phenotypes associated with AlaRS errors are going to present first in neurons then subsequently in cardiomyocytes. It is still unclear if tissue-specific expression of more error-prone AlaRS variants will exhibit cellular defects and protein aggregation.

Despite the neuron-specific phenotype exhibited by cytoplasmic homozygous sti/sti

AlaRS alleles, it was recently shown that attempts to generate the homologous mutation in the mitochondrial AlaRS gene, Aars2, led to embryonic lethality [61]. Consistent with observations from other mitochondrial aaRSs, mitochondrial translation appears to be hyper-sensitive to certain translational errors [26]. The extent of mitochondrial fidelity has yet to be explored in-depth but these experiments highlight the importance of AlaRS proofreading across differentiated cell types and organelles in higher eukaryotic systems.

Beyond differentiated cells, there is recent evidence that perturbations in AlaRS fidelity also causes defects in stem cell differentiation. Studies exploring the pluripotent potential of hematopoietic stem cells from sti/sti mice showed defects in differentiation [62]. These stem cells exhibited elevated protein aggregation which perturbed regulated protein degradation of other cell cycle factors likely through sequestration of proteolytic machinery. The observation that perturbations to translational fidelity generate problems beyond local protein-specific defects and instead can affect global protein regulation

17 provides new context for the role of translational fidelity in maintaining cellular homeostasis.

1.2.4 Alanyl-tRNA Synthetase-Associated Proofreading Factors

Beyond the intrinsic proofreading activity found in AlaRS, several notable examples of additional freestanding proofreading proteins have been characterized that further prevent mistranslation at alanine positions in the proteome. Most organisms across all three domains of life encode a freestanding AlaRS-proofreading domain, AlaXp.

Through a series of biochemical and complementation experiments, it was shown that

AlaXp proteins can proofread Ser-tRNAAla in vitro and AlaXp from Methanosarcina mazei can complement proofreading-deficient AlaRS in E. coli [63]. It was further shown that mammalian AlaXp is transcriptionally fused to p23H, a HSP90 co-chaperone homolog.

Translated full length p23H-AlaXp transcripts do not contain proofreading activity but, through alternative splicing, freestanding AlaXp is transcribed and subsequently translated into an active Ser-tRNAAla deacylase [64]. These authors were able to show that non-p23H fused AlaXp protein products were present across an array of tissue samples including the brain and heart which have already been described to be sensitive to AlaRS errors [64].

In addition to the Ser-tRNAAla deacylase activity observed in AlaXp homologs, a novel trans-editing activity was recently identified from further genetic investigation of the sticky mouse [65]. Subsequent experimentation moving the sti/sti alleles into other mouse genetic backgrounds indicated that not all mice elicited the neurological phenotypes associated with the original sticky mouse. Mapping of the genetic variations among the

18 different mice backgrounds indicated a SNP in the Ankrd16 gene which caused an alternative splicing event, subsequently encoding a premature stop codon. Through a series of detailed biochemical and cellular experiments, the authors showed that Ankrd16 binds directly to AlaRS. Unlike other previously characterized proofreading factors which deacylate non-cognate substrates once transferred to the 3’end of the tRNA, Ankrd16 binds mis-activated Ser-AMP, preventing it from being aminoacylated on to tRNAAla [65].

Early characterization of the AlaRS active site indicated that two proteinogenic substrates can be accommodated for AlaRS activation [56]. Much of the focus over the last fifteen years has been the consequence of serine mistranslation caused by AlaRS errors. It had been assumed that mistranslation caused by glycine mis-aminoacylation was well tolerated in the proteome as defects could only be observed at high levels of non-cognate supplementation [57]. Recently, it was identified that the evolutionarily conserved D- aminoacyl-tRNA deacylase (Dtd) is important beyond preventing the accumulation of D- amino acids in the proteome, as it also acts on Gly-tRNAAla substrates [66]. Subsequent characterization of Dtd in E. coli highlighted the importance of this factor as glycine stress led to severe growth defects when AlaRS proofreading was perturbed [66]. The role of Dtd was further biochemically characterized for higher eukaryotic systems indicating that the

Gly-tRNAAla deacylase activity is also evolutionarily conserved. Taken together, the identification of redundant, yet distinct mechanisms to prevent AlaRS-mediated translational errors highlights the potentially detrimental consequence of mistranslation at alanine positions in the proteome.

1.2.5 Exploring the Physiological Cost of Alanyl-tRNA Synthetase Errors

In the aforementioned sections, our current mechanistic and functional understanding of AlaRS proofreading has been described. While those studies have been informative, they have yet to explore the direct physiological cost of AlaRS-mediated translational errors. It has been assumed that several predictive unfolded protein response factors may be induced upon mistranslation, but the global landscape of these translational errors has yet to be determined. In the following sections work highlighting our new understanding of how E. coli responds to AlaRS-mediated errors will be described.

Additionally, we have characterized novel functional regions within AlaRS that alleviate

AlaRS C666A-associated phenotypes. Finally, given the importance for AlaRS activity, we have identified AlaRS as a promising new therapeutic target against the eukaryotic parasite Leishmania major.

Chapter 2 Alanyl-tRNA Synthetase Quality Control Prevents Global Dysregulation of the Escherichia coli Proteome1

2.1 Introduction

Throughout all domains of life, mechanisms have evolved to minimize errors in protein synthesis. Translational fidelity is maintained at two distinct steps of protein synthesis: surveillance of accurate aminoacyl-tRNA (aa-tRNA) pairing and cognate A site tRNA recognition during ribosome decoding [34]. While it has been previously shown that translational errors occur more frequently than errors in replication and transcription, the extent of these errors had not been well defined. Through advances in proteome-wide mass spectrometry, the prevalence of certain translational errors has been extensively characterized [4]. Observations from these efforts suggest that errors in ribosomal decoding by near cognate anticodons are far more frequent than errors likely caused by mis- aminoacylated tRNAs. This result suggests that errors in decoding are evolutionarily better accommodated than aminoacyl-tRNA errors which can occur at higher frequencies if unchecked by proofreading.

Accurate aa-tRNA pairing is maintained by aminoacyl-tRNA synthetases (aaRSs).

These enzymes are responsible for pairing free amino acids in the cell and ligating them

1 The work presented in this chapter was originally published in mBio (10:e02921-19) and done in collaboration with Nicholas Backes, Kyle Mohler, Christopher Buser, Arundhati Kavoor, Jesse Rinehart, Gregory Phillips, and Michael Ibba. Experiments performed by co-authors are credited in the corresponding figure legend. All other experiments were performed by Paul Kelly. 21 on to their cognate tRNAs [5]. AaRSs perform their function in two distinct steps, first, free amino acid is activated in an ATP-dependent manner forming an aminoacyl-adenylate.

Upon amino acid activation, the amino acid is transferred on to its cognate tRNA which can then be released to be used in translation. The E. coli genome encodes twenty aaRS genes, one for each of the proteinogenic amino acids. As a result of the shared chemicophysical properties of many amino acids, half of the aaRS enzymes can potentially mis-activate numerous non-cognate amino acids (Reviewed in [15]). To prevent erroneous translation, aaRSs have evolved proofreading mechanisms to prevent mis-activated amino acids from being transferred on to tRNAs and subsequently released to the translation machinery for protein synthesis. AaRS-catalyzed proofreading mechanisms (commonly referred to as “editing”) can occur immediately following amino acid activation in which the aminoacyl-adenylate will be hydrolyzed, releasing the amino acid back into the pool of free metabolites. For example, IleRS utilizes pre-transfer proofreading to prevent Val-

AMP from being transferred on to tRNAIle [67]. Alternatively, some aaRS genes encode a second, distinct catalytic active site to monitor aminoacyl moieties following the transfer onto the 3’ end of the tRNA. The aforementioned mechanism of post-transfer proofreading is widespread and has been well characterized for several aaRSs to discriminate non- cognate amino acids including: Tyr-tRNAPhe [25], Nva-tRNAIle/Leu [22,68], Ser-tRNAThr

[31], and Ser-tRNAAla [57,58]. In addition to proofreading activities by the aaRS, several free-standing enzymes are genomically encoded which have activity on mis-aminoacylated tRNA species following release by the aaRS. Some of the more widely characterized trans- editing factors are members of the INS-like family of enzymes that share similar

22 proofreading active site architecture to the prolyl-tRNA synthetase enzymes [41].

Additionally, free-standing AlaRS proofreading domains are present among the AlaXp family of enzymes. AlaXp is an evolutionarily conserved factor which contributes to the accuracy of tRNAAla aminoacylation. Interestingly, E. coli is an outlier among most organisms in that it does not encode an AlaXp homolog [63]. The absence of this factor makes E. coli a strong model for studying AlaRS mistranslation as there is not a redundant mechanism to correct Ser-tRNAAla product formation. Recently, a novel trans-editing factor, ANKRD16 was identified in vertebrates which binds to Ser-AMP in complex with

AlaRS preventing the transfer on to tRNAAla [65]. Finally, the D-tyrosyl deacylase (Dtd) whose function was originally characterized to prevent D-amino acid aminoacylation, was found to have proofreading activity against Gly-tRNAAla [66]. Taken together, the redundancy in proofreading factors, specifically those whose activity is to prevent erroneous tRNAAla aminoacylation, highlights the potential cost of alanine mistranslation events.

In addition to the biochemical identification and characterization of redundant tRNAAla proofreading factors, the physiological cost of AlaRS errors has been described in higher eukaryotic model organisms. A mutant AlaRS allele (sti) in mice was shown to lead to neurodegeneration [59] and cardioproteinopathy [60] due to the accumulation of misfolded proteins. Interestingly, in vitro characterization of the mutant AlaRS protein showed only partial loss of proofreading activity compared to the wild-type enzyme, suggesting that low frequency AlaRS errors are costly to the mammalian proteome.

Furthermore, recapitulation of the sti allele into the mitochondrial AlaRS lead to embryonic

23 lethality [61] suggesting that the mitochondrial proteome is even more intolerant to AlaRS errors.

Despite the importance for AlaRS proofreading and the presumed negative impact on proteome homeostasis of Ala mistranslation events, evidence for beneficial mistranslation has also recently been observed. During oxidative stress, a critical cysteine in the E. coli threonyl-tRNA synthetase (ThrRS) proofreading site becomes oxidized, leading to an overall decrease in ThrRS fidelity [33]. Additionally, oxidative stress causes elevated mis-methionlyation on non-cognate tRNAs in both bacteria and eukaryotes which serves as a protective mechanism against reactive oxygen species [69,70]. In addition to cysteine oxidation, it was recently identified during a screen for aaRS acetylation that

ThrRS can be post-translationally acetylated at K169 leading to a decrease in ThrRS accuracy [71]. Taken together, it appears that during protein synthesis, specific translational errors may be regulated and provide some benefit for the cell under certain environmental conditions. While recent advances in proteome mass spectrometry has led to greater quantification of mistranslational errors, the physiological consequences of these errors has not been extensively explored.

Here we report the global consequences of translational errors in E. coli on cellular physiology and fitness. Despite recent evidence that elevated ThrRS-mediated mistranslation errors occur during both oxidative stress and a regulated protein acetylation event, we show that high levels of serine misincorporation at threonine codons are detrimental to E. coli. Furthermore, we show that growth defects caused by AlaRS-

24 mediated errors are not only due to the accumulation of mistranslated proteins but rather a gross perturbation to proteome homeostasis.

2.2 Results

2.2.1 High levels of serine miscoding are toxic to E. coli

Previous independent reports studying the effects of translation errors suggest that uncoded Thr to Ser substitutions are better tolerated by the cell than Ala to Ser substitutions

[33,57]. It can be speculated that the cause for this difference is the shared terminal hydroxyl group among threonine and serine functional groups leading to a more conservative substitution. Given that ThrRS-mediated mistranslation may be up regulated during cellular stresses such as elevated reactive oxygen species, we wanted to determine the tolerance for serine mistranslation in E. coli. To study the substitution-specific effects of mistranslation, wild-type tRNASer or tRNASer mutants which decode at either alanine or threonine codons were expressed in wild-type E. coli. Plasmid constructs expressing tRNASer variants were generated by cloning the entire serW transcription unit which encodes for one of the five tRNASer genes (Ser5) [72] into a low copy plasmid under control of the native serW promoter and terminator allowing for reliable tRNA processing. Based on previous reports, serW abundance [73] and aminoacylation levels [74] are similar to those of other tRNASer isoacceptors providing a good model for studying global serine mistranslation events. This approach has previously been used to show that Ser to Ala mistranslation led to elevated tumorigenesis in mice [75]. When miscoding tRNASer mutants were expressed in E. coli, both led to a similar decrease in growth rate compared

25 to exogenously expressed wild-type tRNASer (Fig. 2). Results from this experiment indicate that elevated levels of serine miscoding, regardless of predicted translational error will lead to an overall growth defect compared to wild-type E. coli when grown rich media.

Figure 2. Serine miscoding is toxic in E. coli. Chimeric tRNASer variants were generated to decode at either Ser (Anticodon: GGA), Ala (UGC), or Thr (GGU) codons (top). Mutant tRNAs were expressed in wild-type E. coli and growth was monitored (bottom). Both tRNA mutants caused growth defects in MG1655. All growth experiments were performed in triplicate and error bars indicate the standard deviation of the replicates.

2.2.2 AlaRS proofreading is required for optimal growth in E. coli.

To compare the cost of Ala to Ser versus Thr to Ser mistranslation events in E. coli, previously characterized mutations in both the AlaRS and ThrRS proofreading domains were made in isogenic MG1655 genetic backgrounds. Architecturally, these enzymes are predicted to share similar proofreading domains [57], both of which have critical cysteine residues in the active site that coordinate the 3’ end of the tRNA for non-cognate hydrolysis

[31,58]. In vitro experimentation has shown that mutations of these cysteine residues to alanine will partially eliminate the proofreading activity of these enzymes [31,57]. Through the use of a mass spectrometry-based reporter, it was shown that upon mutation of C182 in ThrRS, there was an increase in Thr to Ser substitutions, but these mistranslation events had no effect on cell viability [32]. In comparison, using a temperature-sensitive alaS allele, and AlaRS variants expressed on low copy plasmids, it was shown that when C666

(homologous to C182 in ThrRS) was mutated to alanine, E. coli was sensitized to non- cognate serine stress [57]. Because these independent efforts utilized different genetic approaches and growth conditions, we sought to generate isogeneic E. coli aaRS mutants to directly compare the phenotypic cost of low-level serine mistranslation at both alanine and threonine codons. In vitro kinetic analyses of the corresponding proofreading-defective

ThrRS and AlaRS variants showed that despite serine acting as a non-cognate substrate for amino acid activation, serine is a poor substrate for activation compared to Thr and Ala, respectively. This would suggest that any phenotypes associated with these variants would be representative models for studying the effects of low frequency translation errors

[33,52].

In rich media, the AlaRS C666A variant had a marked decrease in growth compared to wild-type E. coli (Fig. 3A). This result was unexpected as this strain is neither starved for cognate alanine nor supplemented with excess serine. As serine is a poor substrate for

AlaRS activation, we would predict a highly accurate Ala-tRNAAla pool in the AlaRS

C666A strain, suggesting that very low levels of serine misincorporation at alanine codons is detrimental to E. coli. The observed cellular growth defect could be restored by complementing the wild-type alaS allele on a low copy plasmid (Fig. 3A). As expected, the ThrRS C182A variant had no change in growth compared to wild-type MG1655 E. coli.

Growth analyses were repeated in minimal media to determine which non-cognate stress is responsible for this defect. As it has been previously shown that one of the roles of E. coli Dtd is to prevent Gly-tRNAAla accumulation in the cell [66], glycine supplementation was included in our analyses. While exogenous glycine caused a subtle perturbation to growth, only serine supplementation led to a large growth defect (Fig. 3B).

These results suggest that rich media contains sufficient serine to promote mistranslation at a level resulting in a significant growth defect.

Having confirmed that in the absence of AlaRS editing, serine stress impairs cellular growth, it was important to determine if the serine-specific growth defect correlated with serine mistranslation in the proteome. A similar mass-spectrometry based reporter used to show serine mistranslation in the ThrRS C182A strain [32] was used in the AlaRS mutant. Unfortunately, the overexpression of the reporter caused severe cellular growth impairment (data not shown) and was therefore not suitable for in vivo analyses.

To monitor possible Ala to Ser mistranslation in vivo, a β-lactamase-based reporter was repurposed from previous work studying Thr to Ser mistranslation [76]. β-lactamase is an enzyme responsible for cleaving beta-lactam rings, a common class of antibiotic drugs.

Within β-lactamase is an essential serine residue that when mutated, leaves the enzyme inactive and cells are unable to grow in the presence of beta-lactams (e.g. ampicillin) [77].

This essential serine codon was mutated to encode for alanine and thus, cells should only be able to grow in the presence of beta-lactams if Ala to Ser mistranslation occurred at this position. The AlaRS C666A variant but not the wild type, was able to grow on ampicillin while expressing the β-lactamase S68A variant (Fig. 3C). This result provided direct evidence of AlaRS-mediated serine mistranslation in E. coli. These experiments also strongly suggest that in the absence of AlaRS proofreading, E. coli is sensitive to serine stress likely due to elevated mistranslation levels.

As the role of heat shock response proteins was expected to be influential in the maintenance of optimal E. coli growth when translational fidelity was perturbed [78], wild- type, ThrRS C182A, and AlaRS C666A variants were grown in rich media at 42°C and growth was monitored over time. While there was no difference for the ThrRS C182A strain when compared to wild type, the AlaRS C666A variant was further sensitized and impaired in growth at the elevated temperature (Fig. 3D). This observation suggested that the burden caused by mistranslation was likely leading to a global defect preventing the cells from mounting an adequate response to heat stress.

Figure 3. AlaRS fidelity is required for optimal growth in E. coli. A) AaRS mutants and the AlaRS C666A complement strain (pLK-alaS) were grown in LB and their growth was monitored. In rich media, AlaRS fidelity is required for optimal growth. B) AlaRS C666A was grown in M9 minimal media supplemented with exogenous amino acids to determine the toxicity of non-cognate stress. C) Serine mistranslation was observed in the AlaRS C666A mutant using a β-lactamase S68A mistranslation reporter. D) The severity of the AlaRS C666A growth defect was enhanced when grown at 42°C. All growth experiments were performed in triplicate and error bars indicate the standard deviation of the replicates.

2.2.3 AlaRS fidelity is required for maintaining proteome homeostasis

Having observed that AlaRS fidelity was required for optimal growth in rich media and that heat stress further necessitated the requirement for AlaRS aminoacylation accuracy, total proteome analysis was performed to determine which proteins were enriched or under-represented in the absence of AlaRS proofreading. This analysis gave insight into the array of stress responses that can be influenced by protein mistranslation.

In total, 833 proteins were significantly different in the AlaRS C666A variant compared to the wild-type control, 502 were enriched and 331 were under-represented (Fig. 4). KEGG pathway analysis of the total proteome dataset highlights the diverse determinants of cellular homeostasis that translational fidelity can impact. Notably, many of the most enriched pathways are those which are involved in metabolism (Table 2).

Figure 4. AlaRS-mediated mistranslation disrupts proteome homeostasis Total proteome analysis was performed on wild-type and AlaRS C666A E. coli. Depicted is a volcano plot of the 833 significantly enriched or under-represented proteins when AlaRS fidelity is impaired. Proteomics was performed by Kyle Mohler.

Table 2. KEGG pathway analysis of proteome changes

2.2.4 Mistranslation disrupts regulation of the translational machinery

From the KEGG pathway analysis, it was noted that aminoacyl-tRNA biogenesis proteins were under-represented in the AlaRS C666A proteome. Upon further investigation, it was clear that many aaRS and tRNA modification proteins were depleted in the strain. However, it was particularly interesting to note that AlaRS was ~2.3x enriched in the AlaRS C666A strain. To validate this result, steady-state immunoblot analysis was performed to measure AlaRS protein levels in these strains. Recapitulating the proteomic data set, AlaRS protein levels were ~2x higher in the AlaRS C666A background compared to the wild-type strain (Fig. 5A). Interestingly, by complementing the AlaRS C666A strain with a plasmid expressing the wild-type AlaRS gene, the total steady-state AlaRS levels were reduced to a level more similar to the wild-type strain.

AlaRS is known to auto-regulate alaS transcription through alanine-dependent binding upstream of the alaS transcription start site [48]. In the presence of high intracellular alanine, AlaRS will repress active alaS transcription, presumably leading to a decrease in AlaRS protein levels. As many changes to metabolism were observed, it was of interest to know if perturbation to amino acid biosynthesis was responsible for the increased AlaRS protein levels. To determine if the increase in AlaRS protein levels were due to transcriptional changes to alaS expression, qRT-PCR was performed. Intriguingly, there was no observed significant difference in transcript levels (Fig. 5B). These results suggest that AlaRS levels are post-transcriptionally elevated in the absence of AlaRS proofreading. As the steady-state analysis is influenced by both the rate of protein decay and active translation, it was of interest to know if the rate of AlaRS decay was modulated

35 by changes in AlaRS fidelity. Translation was stopped in actively growing E. coli cultures by the addition of chloramphenicol and AlaRS protein levels were monitored over time.

One hour post-antibiotic treatment, ~70% of AlaRS protein had decayed in the wild-type strain compared to ~30% in the error-prone AlaRS C666A variant (Fig. 5C). To determine if changes in protein stability are caused by the C666A substitution, in vitro active site stability and in vitro thermal melting assays were performed on recombinant protein and no change in protein stability was observed (Fig. 6). While a direct mechanism for the elevated and stabilized AlaRS levels remains unclear, contribution by other protein factors which were also dysregulated are predicted to play a role in this effect. Two such factors that may contribute to the observed AlaRS stability are the DnaK-associated factors, GrpE and the molecular chaperone, ClpB, both of which were enriched in our data set, 1.8 and

1.9 fold respectively [79]. These proteins are functionally associated with those which were observed during heat stress in a ribosomal decoding mutant [78].

Figure 5. AlaRS fidelity regulates AlaRS protein levels A) Representative image (left) and quantification (right) of steady-state AlaRS protein levels in wild-type, AlaRS C666A, and AlaRS C666A complemented strains. In the absence of AlaRS fidelity, AlaRS protein levels are elevated. B) qRT-PCR analysis of alaS indicates that transcript levels are unaffected in the AlaRS C666A mutant. C) Representative image (left) and quantification (right) of native AlaRS decay upon treatment of a translation inhibitor. AlaRS protein levels are stabilized when AlaRS fidelity is perturbed. All experiments were performed in triplicate and error bars indicate the standard deviation of the replicates. 37

Figure 6. The AlaRS C666A proteins does not have elevated stability In vitro stability assays of both wild-type and AlaRS C666A recombinant proteins indicated no inherent difference in A) active site or B) thermal stability. Thermal stability assays were performed by Arundhati Kavoor.

2.2.5 Reduced AlaRS fidelity impairs swimming motility

Aside from changes in metabolism, another notable pathway enriched in the under- represented proteins was that involved in flagellar assembly, including the master regulator

FlhD. Perturbation to swimming motility in response to protein mistranslation has previously been observed in a ribosomal decoding mutant [80]. To determine if the under- representation of flagellar assembly proteins led to a change in swimming motility, mistranslating strains were grown on swimming agar plates. As anticipated from the total proteomic data set, loss of AlaRS fidelity led to a decrease in swimming motility (Fig. 7A).

The decrease in motility was restored when complemented with the wild-type alaS allele.

Interestingly, there was no decrease in swimming motility in the ThrRS C182A strain indicating that not all error-prone strains will lead to changes in motility.

In the previous work that implicated translational fidelity and motility impairment, the authors noted the role of the small RNA DsrA to be responsible for these effects. DsrA activity is normally dependent on the small RNA chaperone Hfq [80]. To determine if

DsrA or other small RNAs are influencing the swimming phenotype in the AlaRS C666A strain, the assay was repeated in an AlaRS C666A background in which hfq was deleted.

Deleting hfq resulted in an overall decrease in motility in both the wild-type and AlaRS mutant strains (Fig. 7B). This observation was consistent with previous reports that Hfq- dependent small RNAs act as both positive and negative regulators of E. coli motility [81].

However, these results do suggest that binding of the small RNA DsrA is likely not sufficient to inhibit flagellar synthesis. Aside from post-transcriptional regulation of flagellar genes by small RNAs, motility is regulated by many other environmental and

39 regulatory factors (Reviewed in [82]). As AlaRS-mediated mistranslation led to gross homeostatic perturbation, identification of a solitary mechanistic event leading to motility impairment may not be feasible.

Figure 7. Swimming motility is impaired in the absence of AlaRS fidelity A) Representative image (top) and quantification (bottom) of swimming motility data. In the absence of AlaRS proofreading, E. coli has a swimming defect, and this can be rescued my complementation. B) The observed swimming defect is not exclusively due to small RNA inhibition as an hfq mutant was unable to rescue the defect. All data was plotted relative to a wild-type control for each experiment. Bar graphs are average data collected from three experiments and error bars represent the standard deviation of those experiments.

2.2.6 AlaRS-mediated mistranslation alters the cell membrane

Pathway enrichment during mistranslation also highlighted that fatty acid biosynthesis was perturbed. Furthermore, investigation of stress response activators in the

AlaRS C666A proteome indicated that all five envelope stress response regulators were significantly enriched in the error-prone strain, including (fold enrichment in parentheses):

Sigma E (2.6x), CpxR (1.3x), RcsB (2.6x), BaeR (3.3x), and PspF (1.7x) [83]. It has been suggested that mistranslation of membrane proteins may cause particularly detrimental effects to the cell given the requirement for proper folding and stability across the membrane [84]. Together, these observations led to the investigation of membrane integrity in the AlaRS C666A strain. Membrane defects were surveyed using an array of antibiotic sensitivity assays. For all antibiotics screened, the AlaRS C666A strain was significantly more sensitive (Fig. 8A). The observed broad-spectrum sensitivity suggests that higher concentrations of antibiotics are likely able to cross through the cellular envelope, rather than a targeted sensitivity for the antibiotic mechanism of action. Further support for this hypothesis was the enrichment of the outer membrane porin OmpF which is known to transport antibiotics across the cell membrane [85,86].

It has also been shown that genetic loci of membrane proteins will spatially coordinate towards cellular membranes, consistent with the transertion model of co- transcriptional and co-translational processing directly into the inner membrane [87]. In addition to coordination of translation machinery, the transertion model also causes predictive effects on nucleoid dynamics which are disrupted following treatment with transcription or translation inhibitors [88]. To determine if there were any changes to

42 ribosome organization in the AlaRS C666A strain, all error-prone strains of interest were subject to transmission electron microscopy (TEM). Under normally growing conditions, there is a heterogeneous distribution of the nucleoid and proteins. However during times of stress, nucleoid and ribosomal organization will change leading to noticeable compartmentalization of DNA and ribosomes [88]. Following TEM analysis, both the wild-type and ThrRS C182A strains had no discernable patterning of organization with heterogeneous distribution throughout the cellular milieu. In contrast, in the absence of

AlaRS proofreading, a clear re-arrangement of the cellular nucleoid can be observed, with the ribosomes being sequestered towards the periphery of the cell (Fig. 8B). It remains unclear if this nucleoid rearrangement is solely due to enrichment of the nucleoid associated proteins or if this is in part due to disruption of the translational machinery.

Figure 8. The E. coli membrane is affected when AlaRS fidelity is impaired A) The AlaRS C666A mutant (blue) is sensitive to an array of antibiotics as observed using a disk diffusion assay compared to wild-type (black) and ThrRS C182A (red) E. coli. All experiments were performed in triplicate with error bars indicating the standard deviation. The asterisks indicate statistical significance as determined using one-way ANOVA with Tukey post-hoc comparison (p < 0.05). B) TEM analysis indicated altered nucleoid morphology in the absence of AlaRS fidelity. TEM was performed by Christopher Buser.

2.3 Discussion

2.3.1 AlaRS fidelity is evolutionarily protected

Across all domains of life, mechanisms have evolved to prevent AlaRS-mediated mistranslation. In addition to endogenous proofreading activity, AlaRS can re-sample aa- tRNAs leading to mis-aminoacyl-tRNA hydrolysis. Furthermore, at least three freestanding enzymes have been identified in vivo to prevent mis-charged tRNAAla species from accumulating including, AlaXp [63], Dtd [66], and ANKRD16 in vertebrates [65].

Additional factors such as ProXPST1 have also been identified to have proofreading activity in vitro against these aa-tRNAs but their role in vivo has yet to be well characterized

[89].

Despite the poor specific activity for non-cognate aa-tRNA synthesis by AlaRS, results from this work highlight the incredibly high cost of Ala to Ser mistranslation in E. coli, which has also been suggested from several studies in higher eukaryotes. The identification of the sti allele in mice and its neurodegenerative phenotype provided the first insight into the possible cost of AlaRS-mediated mistranslation in eukaryotes [59].

Interestingly, this AlaRS allele only exhibited a minor defect in Ser-tRNAAla proofreading in vitro. Together, these results suggested that low frequency AlaRS errors are incredibly costly to terminally differentiated neuronal cells. Consistent with that observation, mice homozygous with the AlaRS C723A (corresponding to E. coli C666A) mutation were embryonic lethal [60]. While the aforementioned studies observed indicators of elevated protein misfolding, the work described in this report show global dysregulation of the

45 proteome resulting from loss of AlaRS proofreading which may also contribute to the phenotypic defects observed in higher eukaryotes.

2.3.2 Toxic mistranslation may provide a new target for drug discovery

As AlaRS-mediated errors led to global changes in proteome regulation, this suggests that tRNAAla fidelity may act as a key checkpoint for cellular homeostasis.

Furthermore, this leads to the possibility of targeting AlaRS fidelity for novel antimicrobial discovery. Given their essential role in protein synthesis, aaRS have been a promising drug discovery target, with several compounds on the market including the topical antifungal agent Tavaborole. Tavaborole functions by binding to the LeuRS proofreading site which locks the enzyme in a non-productive conformation [90]. While this drug essentially inactivates the enzyme, the observations from this work suggest that a novel class of aaRS inhibitors could be screened for whose mechanism would block proofreading activity releasing elevated mis-aminoacyl tRNA into the cell for translation. As mistranslation has been shown to lead to increased antimicrobial resistance [91], targeting translational fidelity as a monotherapeutic would likely be ineffective. However, chemically inducing mistranslation may act as an agent for chemosensitization that could be exploited for combination therapies with pre-existing compounds [92].

2.3.3 A threshold exists for neutral/beneficial mistranslation events

Considerable efforts have been devoted to the characterization of beneficial mistranslation events (Reviewed in [34]). The identification of these events leads to several

46 interesting interpretations, some of which have been characterized, including changes in antibiotic resistance [91] and antigen diversity [93]. As our understanding of beneficial mistranslation is still in its infancy, two questions have remained to be explored: can we begin to predict novel beneficial mistranslation events and do these beneficial mistranslation events provide some consequence to the cell. Evolutionarily, incidence of genomically encoded error-prone aaRSs have been observed across several phyla of intracellular pathogens, including Microsporidian [35,36] and Mycoplasma [37] species.

An interesting observation from these studies is the propensity for the same error-prone aaRS to arise in these organisms. Of the systems that have been explored, LeuRS, ThrRS, and PheRS have the potential to lose proofreading activity [37]. This suggests that errors mediated by these aaRS are better accommodated by the proteome globally. It is also likely that the proofreading activity of these aaRS are most easily subject for regulated fidelity as they bridge the gap between completely degenerate proofreading domains and fully active proofreading function. This is further supported by the high tolerance for PheRS-mediated errors under nutrient-stable conditions in E. coli [28] and the recently observed modulation of Salmonella PheRS fidelity upon oxidative stress [94].

Results from this work and others suggests that due to the high proteotoxic cost of

AlaRS-mediated mistranslation, environmentally regulated Ala to Ser substitutions will likely not be observed in nature as the potential benefit (e.g. antigen diversity), does not support the penalty of an inaccurate proteome. This is further supported by the numerous genomically-encoded secondary mechanisms to minimize Ala mistranslation, which to date is unique to AlaRS.

In contrast to the requirement for AlaRS fidelity, E. coli ThrRS has recently been shown to lose proofreading activity through at least two different mechanisms, during oxidative stress [33] and post-translational acetylation [71]. Loss of ThrRS proofreading activity is not exclusive to E. coli as several Mycoplasma [37,95] and yeast [96] species encode ThrRS genes that are naturally proofreading-deficient in cytoplasmic and mitochondrial translation, respectively. Another goal of this work was to determine if sufficiently high levels of Thr to Ser mistranslation (i.e. the resulting mistranslation event following ThrRS modification), would be tolerated by E. coli. By generating a chimeric tRNASer variant which would translate at Thr codons, it was shown that this mistranslation event does cause growth defects. This suggests that despite the shared terminal functional group, these amino acids are not neutral to the proteome. While it is possible that other factors temporally aligned with native ThrRS modification may influence this result, overall our findings suggest that not all beneficial mistranslation events are ubiquitously advantageous.

2.4 Methods

2.4.1 Strain Construction and Reagents

To construct new alaS and thrS E. coli mutants, elements of the “gene gorging” method [97] were combined with other approaches [98], that use I-SceI nuclease to introduce double-strand breaks into the bacterial chromosome to select for cells that have undergone homologous recombination events. The improved method, which will be described in further detail separately, offers the advantage that specific DNA sequence

48 alterations can be made to the chromosome without the need to make any other base pair changes to the DNA. Briefly, PCR products representing the wild-type alaS (~2.7kbp) and thrS (~1.9kbp) were amplified from MG1655 [99] genomic DNA and cloned into an R6K- based suicide vector constructed specifically for allelic exchange [100]. This mobilizable plasmid encodes resistance to chloramphenicol (Cam) and includes the 18bp recognition site for the I-SceI nuclease [101]. The desired mutations were then introduced to alaS

(1996TGTGCG) and thrS (1544TGCGCG) by inverse PCR [102]. DNA sequencing was used to confirm that only the desired changes had been made to alaS and thrS (DNA

Facility, Iowa State University, Ames, IA). The resulting plasmids were then transformed into the diaminopimelic acid (DAP) auxotroph donor strain MFDpir [103]. These transformants were then used as a donor strains to introduce the R6K plasmids into

MG1655 by conjugation. CamR colonies that grew in the absence of DAP were selected at

37°C, which represented recombinants where the suicide vector had integrated into the E. coli chromosome by a single-crossover event. CamR recombinants were then transformed with the helper plasmid pSceH, a derivative of pSLTS [98], which expresses the I-SceI nuclease under control of TetR from a temperature-sensitive pSC101-derivative plasmid and imparts ampicillin (Amp) resistance. AmpR transformants were selected at 30°C in the presence of anhydrotetracycline (aTc) to induce I-SceI expression. Surviving colonies were then screened to identify recombinants that had lost the CamR marker, indicative that the integrated R6K vector had been deleted by a second recombination event. Multiple CamS recombinants were subsequently tested by PCR and Sanger sequencing to identify mutants that had inherited the new alaS and thrS alleles.

To determine the relationship between small RNA binding and AlaRS fidelity for swimming motility, hfq was deleted in the alaS mutant background by P1 transduction of an hfq::kan allele from NR633 [104]. A complete list of strains and plasmids used in this work can be found in Table 3.

Lysogeny broth (LB) was used for all experiments in rich media. M9 minimal media was prepared for all minimal media experiments. M9 contained 1x M9 salts, 2 g/L glucose, 1 mg/mL thiamine, 1 mM MgSO4 and 0.1 mM CaCl2 [29,105]. Amino acid supplementation was performed when indicated. Antibiotic supplementation was performed as follows: kanamycin (RPI - 25 µg/mL final), ampicillin (RPI - 100 µg/mL final for selection, 20 µg/mL final for mistranslation reporter), and chloramphenicol (200

µg/mL final to halt translation). Unless otherwise noted, all reagents and oligos were purchased from Sigma Aldrich.

Table 3. Strains and plasmids used in this study

2.4.2 Growth Analysis

For all growth experiments, overnight cultures of the respective strains were grown to saturation in either LB or M9 minimal media with antibiotics supplemented when applicable. Prior to commencing growth analysis, OD600 measurements were taken for each saturated culture and starting inoculums were normalized to an OD600 value of 0.05.

Cultures were grown with aeration at either 37°C or 42°C for heat stress analysis. OD600 values were measured at the indicated time points using a CO8000 Cell Density Meter

(WPA). Data plotted is the average of at least three biological replicates with error bars indicating the standard deviation of the replicates.

2.4.3 Construction of Mistranslating tRNASer Plasmids

A 195bp DNA fragment encoding the serW transcription unit was designed across two partially overlapping synthetic DNA oligos with 5’ phosphate modifications added.

The serW gene encodes for one of the five serine tRNAs in E. coli [72]. The aforementioned oligos were used in PCR to generate the wild-type full-length serW amplicon and was subsequently cloned into the pSMART-LCKan blunt cloning vector following manufacturer’s recommendations (Lucigen). To generate mistranslating tRNASer variants, the pLK-serW (Ser) vector was used as a template for site-directed mutagenesis

(Stratagene) and the anticodons were mutated to translate at either alanine or threonine codons. All three serW variant plasmids were transformed into MG1655 for growth analysis.

2.4.4 In vivo Mistranslation Reporter

It has previously been shown that mutation of an essential serine in β-lactamase

(Bla) will inactivate the enzyme [77] which can then be used as a useful tool for studying missense serine mistranslation [76]. The promoter and β-lactamase-encoding gene (bla) were amplified from pUC18 and cloned into pSMART-LCKan by blunt ligation (Lucigen).

The resulting pLK-Amp vector was then subjected to site-directed mutagenesis to generate the inactive Bla S68A variant. Both, the pLK-Amp and pLK-Amp S68A vectors were transformed into MG1655 and MG1655 AlaRS C666A while maintaining selection using the kanamycin-resistance cassette on the plasmid.

In vivo mistranslation was monitored by streaking MG1655 pLK-Amp/Amp S68A and MG1655 AlaRS C666A pLK-Amp/Amp S68A on LB plates containing either kanamycin or kanamycin and ampicillin. Plates were grown for two days at 37°C to allow sufficient time to observe growth by the MG1655 AlaRS C666A strains.

2.4.5 Total Proteome Analysis

To monitor changes in protein abundance, 20 mL E. coli cultures were grown to an

OD600 of 1.0 and harvested by centrifugation. The resulting pellet was frozen at -80 °C for downstream processing. For cell lysis and protein digest, cell pellets were thawed on ice and 2 µl of cell pellet was transferred to a microcentrifuge tube containing 40 μl of lysis buffer (10 mM Tris-HCl pH 8.6, 10 mM DTT, 1 mM EDTA, and 0.5 % ALS). Cells were lysed by vortex for 30 s and disulfide bonds were reduced by incubating the reaction for

30 min. at 55 °C. The reaction was briefly quenched on ice and 16 μl of a 60 mM IAA

53 solution was added. Alkylation of cysteines proceeded for 30 min in the dark. Excess IAA was quenched with 14 μl of a 25 mM DTT solution and the sample was then diluted with

330 μl of 183 mM Tris-HCl buffer pH 8.0 supplemented with 2 mM CaCl2. Proteins were digested overnight using 12 μg sequencing grade trypsin. Following digestion, the reaction was then quenched with 12.5 μl of a 20% TFA solution, resulting in a sample pH<3.

Remaining ALS reagent was cleaved for 15 min at room temperature. The sample (~30 μg protein) was desalted by reverse phase clean-up using C18 UltraMicroSpin columns. The desalted peptides were dried at room temperature in a rotary vacuum centrifuge and reconstituted in 30 μl 70% formic acid 0.1% TFA (3:8 v/v) for peptide quantitation by

UV280. The sample was diluted to a final concentration of 0.2 μg/μl and 5 μl (1 μg) was injected for LC-MS/MS analysis.

LC-MS/MS was performed using an ACQUITY UPLC M-Class (Waters) and Q

Exactive Plus mass spectrometer. The analytical column employed was a 65-cm-long, 75-

μm-internal-diameter PicoFrit column (New Objective) packed in-house to a length of 50 cm with 1.9 μm ReproSil-Pur 120 Å C18-AQ (Dr. Maisch) using methanol as the packing solvent. Peptide separation was achieved using mixtures of 0.1% formic acid in water

(solvent A) and 0.1% formic acid in acetonitrile (solvent B) with either a 90-min gradient

0/1, 2/7, 60/24, 65/48, 70/80, 75/80, 80/1, 90/1; (min/%B, linear ramping between steps).

Gradient was performed with a flowrate of 250 nl/min. At least one blank injection (5 μl

2% B) was performed between samples to eliminate peptide carryover on the analytical column. 100 fmol of trypsin-digested BSA or 100 ng trypsin-digested wild type K-12

MG1655 E. coli proteins were run periodically between samples as quality control standards.

The mass spectrometer was operated with the following parameters: (MS1) 70,000 resolution, 3e6 AGC target, 300–1,700 m/z scan range; (data dependent-MS2) 17,500 resolution, 1e6 AGC target, top 10 mode, 1.6 m/z isolation window, 27 normalized collision energy, 90 s dynamic exclusion, unassigned and +1 charge exclusion. Data was searched using Maxquant version 1.6.1.0 with Acetyl (N-Term), Deamidation (NQ),

Oxidation (M), and Phospho (STY) as variable modifications and Carbamidomethyl (C) as a fixed modification with up to 3 missed cleavages, 5 AA minimum length, and 1% FDR against the Uniprot E. coli database. Searches were analyzed with Perseus version 1.6.2.2.

2.4.6 Immunoblotting

Steady-state AlaRS levels were determined by harvesting growing E. coli cultures when they reached an OD600 of 0.7. Cell pellets were re-suspended in SDS loading dye and boiled for 10 minutes before equal volumes of total cellular material was loaded on a

10% SDS-polyacrylamide gel and separated by electrophoresis. Proteins were transferred onto a 0.45 µm nitrocellulose membrane (Amersham) before blocking for an hour in 5% milk in TBST. E. coli AlaRS (96 kDa) was probed with an anti-E. coli AlaRS antibody

(1:1,000 – ProSci custom antibody) and anti-rabbit horseradish peroxidase (1:5,000 – GE

Healthcare). Membranes were stripped using the Abcam mild stripping protocol and re- probed with an HRP-conjugated anti-Sigma 70 (70 kDa) antibody (1:4,000 – Biolegend) as a loading control. HRP signals were developed using Clarity ECL substrate (Bio-Rad)

55 and monitored using a ChemiDoc and accompanying software (Bio-Rad). Western blot densitometric quantification was performed using ImageJ software [106].

Protein stability assays were performed essentially as described above with minor modifications. Cells were grown to an OD600 of 0.5 and 1 mL of cells were removed, pelleted, and frozen as T0 samples. Simultaneously, chloramphenicol was added to the remaining cultures (200 µg/mL final) and continued to grow. At the indicated time points, samples were removed, pelleted, and frozen until all samples were collected. To quantify the relative protein stability, densitometric analysis was performed and normalized to the abundance of protein at T0 for a given biological replicate. The data plotted is the average stability of three independent biological replicates with error bars indicating the standard deviation of the replicates.

2.4.7 Transcript Analysis

From saturated overnight cultures, strains were normalized, back-diluted, and cultures were grown to an OD600 of 0.7 prior to pelleting. Bacterial pellets were re- suspended in RNAlater (Ambion) and stored at 4°C overnight. Pelleted cellular material was re-suspended in buffer containing 20 mM sodium acetate pH 5.2, 1% SDS, and 0.3M sucrose and extracted in equal volume acid phenol chloroform at 65°C. The aqueous phase was subsequently re-extracted with acid phenol chloroform at room temperature before one final chloroform extraction. RNA was precipitated in 1 volume isopropanol and 1/10th volume sodium acetate. DNA was removed from samples using Turbo DNase (Invitrogen) and RNA was extracted using acid phenol chloroform prior to ethanol precipitation.

Reverse transcription was performed using 100 ng RNA and Superscript IV (Invitrogen) following manufacturer’s recommendations. Transcript abundance was determined using primers specific for target mRNAs, SsoAdvanced Universal SYBR Green Supermix (Bio-

Rad), and analyzed using a CFX96 Real-Time PCR Detection System. (Bio-Rad). Data was analyzed using the Pfaffl method [107] and is the result of three technical replicates from three independent biological replicates.

2.4.8 Recombinant AlaRS Purification

The effect of the AlaRS C666A substitution on protein structure was determined by monitoring the stability of recombinant protein. Wild-type or mutant alaS genes were amplified from their respective E. coli strains and cloned into pET21b at NdeI and XhoI restriction cloning sites. The resulting expression construct generated an in-frame C- terminal His tag for metal affinity purification. AlaRS proteins were expressed in BL21

(DE3) after growing cells to at 37°C to mid-log and subsequently inducing expression with

500 µM IPTG for 4 hours. Harvested cells were re-suspended in Buffer A (50 mM Tris

HCl pH 8.0, 300 mM NaCl, and 10 mM imidazole) and EDTA-free cOmplete mini protease inhibitor (Sigma) prior to lysis by sonication. Clarified lysate was passed over a TALON metal affinity column (Takara), washed with Buffer A, and finally eluted with Buffer B

(50 mM Tris HCl pH 8.0, 300 mM NaCl, and 250 mM imidazole). Proteins were dialyzed in two stages to remove imidazole and to store the enzyme in 50% glycerol.

Initial enzyme concentrations were determined by active site titration [108].

Briefly, enzyme was incubated with 8 mM ATP, 150 µM [14C]-alanine (Perkin Elmer),

57 pyrophosphatase, and 1x buffer (100 µM HEPES pH 7.2, 30 mM KCl, and 10 mM MgCl2).

Reactions were incubated at 37°C for 20 minutes before filtering through a Protran BA 45 nitrocellulose membrane (Whatman). Filter discs were prewashed with 0.5x buffer, and subsequently washed three times with 0.5x buffer after sample filtration. Discs were dried, and radiolabeled signal was quantified using scintillation counting.

2.4.9 Active site and Thermal Stability

To monitor changes in protein stability, active site titration was performed as above using 5 µM enzyme before (T0) and after (T60) incubating the enzyme at 37°C for 60 minutes prior to analysis. The stability of the enzyme was determined by plotting the change in activity after incubation at 37°C (Activity at T60/Activity at T0) [109].

Changes in protein stability were also monitored using circular dichroism (CD).

Wild-type and mutant proteins were re-suspended in 100 mM Potassium Dihydrogen

Phosphate and transferred to a 0.5 mL Amicon Ultra Centrifugal Filter tubes. Both proteins were then centrifuged at 10,000 x g for 15 minutes at 4°C to remove Tris and glycerol, which are incompatible with CD analysis. The centrifugation was repeated three times with the addition of 200 µl of 100 mM potassium dihydrogen phosphate after each spin. The concentration was determined using a NanoDrop and a final concentration of 0.5mg/ml was used for the CD experiments.

A variable temperature measurement was performed on the Jasco J-815 Circular

Dichroism Spectrometer where the change in molar ellipticity of the protein was measured as a function of temperature to determine the melting temperature (Tm). The molar

58 ellipticity was measured at 222 nm which captures the extent of alpha helicity present in the protein. A range of temperatures between 25°C-95°C was selected. The values of molar ellipticity obtained were converted to fraction folded using the following equation:

α = (θT – θU) / (θF – θU)

α is the fraction of folded protein, θT is the molar ellipticity at 222 nm at any temperature,

θF is the molar ellipticity at 222 nm of the completely folded form (at 25°C) and θU is the molar ellipticity at 222 nm of the completely unfolded form (at 95°C). The fraction folded vs temperature was plotted where the Tm, the temperature at which 50% of the protein is folded, was estimated.

2.4.10 Swimming Motility

Swimming plates were prepared in LB broth as described above and 0.2% agar.

From overnight saturated cultures, cells were normalized to an OD600 of 0.5 and 5 µL of cells were spotted on freshly solidified swimming plates. Plates were incubated at 37°C for

4 to 8 hours prior to imaging. Swimming distance was calculated using ImageJ and the relative swimming distance was determined by comparing the mutant genotype of interest to a wild-type control. Bar graphs indicate the average relative motility for three independent biological replicates and error bars indicate the standard deviation of the replicates.

2.4.11 Antibiotic Sensitivity

Saturated overnight cultures were struck on LB agar plates with cotton swabs creating a bacterial lawn. Prior to culturing at 37°C, an array of antimicrobial sensitivity disks (Oxoid) were placed on the bacterial lawn. Following the overnight incubation, plates were imaged and the distance of disk diffusion was measured using ImageJ.

2.4.12 Transmission Electron Microscopy

Strains were back-diluted from a saturated overnight culture and grown to an

OD600 of 0.8. Cells were gently pelleted at 3000 rpm for 5 minutes. Pellets were prefixed in 4% EM-grade buffered formaldehyde for shipping and storage. Bacterial pellets were subsequently fixed in 2.5% glutaraldehyde/0.1 M phosphate buffer at 4°C. Pellets were then washed three times for 5 minutes in 0.1 M phosphate buffer, placed in 1% aqueous osmium tetroxide for 90 minutes and then reduced with ferrocyanide for 60 minutes at room temperature. Pellets were washed three times with water, dehydrated in a graded acetone series and embedded in Spurr’s resin. Sections were cut 60 nm thin, collected on formvar-filmed copper grids, stained with 2% uranyl acetate and Reynold’s lead citrate.

Bacterial sections were imaged at 80 kV in a Zeiss EM10 using a Gatan Erlangshen CCD camera.

Chapter 3 Second-Site Suppressors in AlaRS Rescue AlaRS C666A-Associated Phenotypes2

3.1 Introduction

Across all domains of life, protein coding sequences have evolved to maintain activity of essential biological processes. Specific residues or modular domains with high evolutionary conservation within a polypeptide sequence can provide insight into regions of functional essentiality of a given protein [110,111]. Because of this, mechanisms have also evolved to minimize errors in protein synthesis (reviewed in [34]).

Translational fidelity is maintained at several distinct stages of protein synthesis, with one of the key steps being the accuracy of aminoacyl-tRNA synthesis by aaRSs.

AaRSs are responsible for pairing free amino acids in the cell to their cognate tRNAs, with the product then released for participation in ternary complex formation with elongation factor and GTP (reviewed in [5]). This complex is then recruited to the A-site of the ribosome to facilitate peptide bond formation. Most organisms encode at least one distinct aaRS gene for each of the twenty proteinogenic amino acids in the cell. In complex with

ATP, aaRSs bind to free cognate amino acids leading to synthesis of an aminoacyl- adenylate. This activated amino acid will then be transferred to the 3’ end of the tRNA to be released in the cell for translation. AaRSs have evolved mechanisms to prevent the

2 The work presented in this chapter was done in collaboration with Arundhati Kavoor, Nicholas Backes, Gregory Phillips, and Michael Ibba. Experiments performed by co-authors are credited in the corresponding figure legend. All other experiments were performed by Paul Kelly. 61 accumulation of mis-aminoacylated tRNAs. Half of the aaRSs in E. coli are able to occlude non-cognate amino acid activation in the primary active site, however the other ten aaRS will mis-activate non-cognate substrates (reviewed in [15]). To prevent non-cognate aa- tRNAs from being released into the cell, many aaRSs have evolved secondary mechanisms to hydrolyze non-cognate products. Several synthetases, including IleRS have evolved hydrolytic activity against non-cognate amino acids at the amino acid activation site. Upon non-cognate adenylate recognition, the amino acid and AMP will be hydrolyzed and released back into the cell [19]. In contrast, several aaRS including, PheRS [25], ThrRS

[31], and AlaRS [57] encode a distinct secondary active site which contains proofreading activity against non-cognate amino acid-containing substrates after the amino acid is transferred to the 3’ end of the tRNA.

While considerable work has been devoted to characterizing the enzymatic activity of aaRSs, the exploration of the role of these enzymes in vivo has only been recently investigated [28,29,112,113]. It was previously observed that an alanine substitution of an essential cysteine in the AlaRS proofreading domain leads to mis-aminoacylation of serine on tRNAAla in vitro [57]. E. coli with the same substitution (AlaRS C666A) show defects in maintaining proteome homeostasis in E. coli [113]. To expand on these efforts, we sought to identify suppressors of the AlaRS C666A phenotype in hopes of further characterizing the mechanisms by which AlaRS fidelity promotes cellular homeostasis. In this work we identify and characterize two additional residues in E. coli AlaRS that play roles in enzymatic activity. Furthermore, the observations presented herein suggest that a slight increase in AlaRS proofreading activity may be sufficient to alleviate the AlaRS

C666A-associated physiological defects. More broadly, this work demonstrates the utility of applying genetic approaches to further elucidate novel mechanistic features of aaRSs.

3.2 Results

3.2.1 Identification of Suppressor Mutations that Alleviate the AlaRS C666A Growth Defect

It has been previously observed that perturbation of AlaRS proofreading in E. coli leads to growth defects when compared to wild-type E. coli [113]. To determine if secondary mutations in the E. coli genome could suppress the growth defects associated with a deficiency in AlaRS proofreading, suppressor mutants that restore near-normal growth compared to their proofreading-deficient strain were screened for. Due to the severe growth defects associated with AlaRS-mediated mistranslation, suppressor mutations were anticipated to act in reducing the mistranslational load on the proteome. It was speculated that this suppression could possibly occur by increasing the thermodynamic barrier for EF-

Tu discrimination and thus preventing mis-acylated tRNA accumulation into active ternary complex [114-116]. Additionally, the E. coli protein ProXP-ST1 was previously shown to have activity against Ser-tRNAAla in vitro [89]. Therefore it was possible that mutations in the yeaK gene (gene encoding ProXP-ST1) could increase ProXP-ST1 proofreading activity on mis-serylated tRNAAla substrates.

Suppressor colonies of the MG1655 AlaRS C666A strain accumulated on LB agar plates without the need for additional mutagenic manipulation. One initial wild-type-like colony was selected from the mixed population, colony purified, and genomic DNA

(gDNA) was extracted. The isolated gDNA was further subjected to whole genome 63 sequencing (WGS) for variant analysis. When compared to another control sample that was also analyzed by WGS, only one additional mutation was identified in the suppressor mutant. The identified single nucleotide polymorphism (SNP) was a G1651T mutation in the alaS gene that when translated, caused a D551Y substitution in AlaRS. Having identified second-site alaS suppressors, two additional suppressor mutant colonies were isolated from LB agar plates. Rather than subjecting the two additional mutants to WGS analysis, gDNA was extracted from these mutants and the DNA was used as the template for PCR to amplify the alaS gene. This alaS amplicon was used as the template for a Sanger sequencing reaction using a primer with homology to alaS. Sanger sequencing of the two additional suppressors reported that both of the mutants also contained secondary mutations in the alaS gene, with both SNPs occurring at nucleotide position C1681, ultimately resulting in two different substitutions at R561, R561C and R561S.

While the alaS gene encodes one complete monomeric unit of AlaRS, this gene contains at least three characterized domains responsible for distinct enzymatic activity including, amino acid activation and tRNA aminoacylation [51,52], proofreading [57], and a C-terminal C-Ala domain required for oligomerization that also participates in aminoacylation [43,117]. The three suppressor mutations all occur in the proofreading domain of AlaRS (Fig. 9A). To date, these two amino acid positions have yet to be characterized or predicted to participate in any proofreading activity. The coordination of the proofreading domain active site has previously been explored and has indicated that residues involved in mis-acylated tRNA hydrolysis are evolutionarily conserved (e.g. C666 in E. coli) [58]. To determine if the two amino acid residues identified in the suppressor

64 screen are also conserved, AlaRS proofreading domains were aligned across several prokaryotic and eukaryotic model systems. With the exception of the closely related

Salmonella enterica, these amino acid positions do not appear to be evolutionarily conserved (Fig. 9B). Currently, no structural data for the E. coli AlaRS proofreading domain exists. Using Phyre2 modeling predictions, the position of the suppressor residues were mapped on the AlaRS structure. Based on these predictions, both residues, D551 and

R561 are positioned in potentially productive regions in the AlaRS proofreading domain for tRNA body interactions, but likely unable to directly participate in mis-aminoacyl- tRNA hydrolysis (Fig. 9C). These preliminary observations provided the first evidence of second-site AlaRS suppressors that alleviated the growth defect associated with AlaRS mistranslation in E. coli.

Figure 9. AlaRS C666A suppressors map to the AlaRS proofreading domain A) Suppressor mutations map to the AlaRS proofreading domain. B) The sites for suppressor mutants do not appear to be conserved across all domains of life. C) Suppressor variants were modeled on the AlaRS proofreading domain using Phyre2 (96% of residues modeled at >90% accuracy).

3.2.2 Second-site AlaRS mutations alleviate the slow-growth AlaRS C666A phenotype

While the sequencing data indicated second-site mutations in the alaS gene were likely responsible for the suppression of the AlaRS C666A slow-growth phenotype, it was possible that other mutations in the suppressor strains may have been missed by our sequencing analysis and the second-site mutations were a coincidence but not responsible for the suppression. To confirm that the three identified mutants were acting as the

MG1655 AlaRS C666A suppressors, the three mutations were made in isogenic MG1655 and MG1655 AlaRS C666A strains, creating single and double mutants for the suppressors.

One possibility for these mutants was the potential for the second-site mutants acting as compensatory mutations for the AlaRS C666A substitution, while single mutants alone may elicit a separate growth defect.

Upon generation of the six isogenic strains, growth of all of the strains was monitored in LB liquid media at 37°C. This growth condition has previously been used to highlight the AlaRS C666A growth defect [113]. Interestingly, all single and double mutant suppressor strains grew as well as wild-type E. coli (Fig. 10A). This observation indicated that suppressor mutations identified from the various sequencing analyses were acting as suppressors. Furthermore, substitutions at D551 or R561 alone do not cause any change in growth at 37°C in rich media. As previously eluded to, perturbation to AlaRS proofreading leads to defects in growth in rich media. It was further shown that this defect is serine dependent, presumably through elevated levels of Ala to Ser mistranslation [57,113]. To determine if the suppressor mutants also exhibit a serine-dependent growth defect that may have been masked in rich media, the suppressor mutants were grown in minimal media 67 supplemented with exogenous serine and their growth was monitored. Despite the excess non-cognate amino acid stress, the single and double suppressor mutants were able to grow just as well as wild-type E. coli (Fig. 10B), which may suggest that the level of serine mistranslation was reduced in the presence of the second-site substitutions.

Figure 10. Suppressor mutants alleviate AlaRS C666A-associated growth defects Suppressor mutants alleviated the AlaRS C666A growth defects in A) LB and B) M9 minimal media supplemented with non-cognate serine.

3.2.3 AlaRS C666A suppressors do not prevent serine mistranslation

To directly observe the potential for Ala to Ser mistranslation in the suppressor strains, a recently described mistranslation reporter [113] was applied to all of the single and double mutants. The mistranslation reporter utilizes a mutated β-lactamase gene in which an essential serine residue has been mutated to encode an alanine [76]. Under conditions of high AlaRS fidelity, the mutant β-lactamase product will be translated and the cells will be unable to grow on β-lactam antibiotics (e.g. ampicillin). If Ala to Ser mistranslation is occurring, the wildtype β-lactamase protein will be made and the cells will grow in the presence of the antibiotic. This reporter has recently been used to show that in the AlaRS C666A variant strain, Ala to Ser mistranslation is occurring in vivo [113].

Upon transformation of the reporter and plating of the suppressor strains, all variants that contained the C666A allele were able to grow on ampicillin (Fig. 11). This observation suggests that the C666A substitution is sufficient to cause protein mistranslation even in the presence of additional second-site suppressors. None of the single AlaRS suppressor variants were able to grow in the presence of antibiotic suggesting that these substitutions alone do not negatively affect the hydrolytic proofreading activity of these enzymes in vivo.

One limitation of this qualitative assay is the inability to discern quantifiable changes in levels of mistranslated proteins. It is possible that despite the accumulation of mistranslated proteins in the second-site suppressors, the levels of mistranslation are still reduced compared to the single AlaRS C666A variant.

Figure 11. AlaRS C666A second-site suppressors do not prevent mistranslation Serine mistranslation was observed in the AlaRS C666A suppressors using a β-lactamase S68A mistranslation reporter.

3.2.4 Phenotypes associated with AlaRS-mediated proteome dysregulation are alleviated in with the second-site suppressor mutations

Previous efforts studying the effect of AlaRS mistranslation in E. coli identified that not only was protein mistranslation occurring, but the accumulation of mistranslated proteins led to dysregulation of the E. coli proteome [113]. This observation suggested that regardless of the individual protein that may contain a misincorporated amino acid, the Ala to Ser substitution itself was not tolerated globally and caused activation of several stress response cascades that could be reproducibly observed with population-level analyses. The first indication that protein regulation may be altered in the E. coli AlaRS strain was the observation that in the absence of AlaRS proofreading, E. coli exhibited an accentuated slow growth phenotype when grown in rich media at 42°C. The delayed growth response to the heat stress potentially suggested that these AlaRS proofreading-deficient cells were unable to mount an adequate response to thermal stress. Total proteome analysis of wild- type and AlaRS C666A strains further revealed that in the absence of AlaRS proofreading, more than 800 proteins were differentially enriched or under represented [113].

Despite the observation that serine mistranslation was still occurring in the suppressor mutants, it was of interest to determine if the phenotypes associated with the proteome dysregulation in the AlaRS C666A strain were still present in the suppressor strains. To determine if the suppressor strains were able to respond to a shift in thermal stress, the suppressor strains were grown in LB media at 42°C. As had been previously observed, the AlaRS C666A strain has an elevated growth sensitivity at 42°C compared to

37°C [113]. However, none of the single or double suppressor mutants elicited any deviation in growth when compared to wild-type E. coli (Fig. 12A). This observation 72 suggested that the dysregulation associated with the single AlaRS C666A allele was not present in the suppressor strains.

From the original total proteome analysis, several distinct physiological pathways were identified to be enriched or under-represented. In addition to many other biological processes, there appeared to be a defect in motility and membrane integrity that was further assayed for, and confirmed using direct phenotypic analyses [113]. To further support that the global dysregulation was alleviated in the AlaRS C666A suppressors, all strains were plated on LB agar swimming media and motility was monitored. The only strain to exhibit a statistically significant change in motility was the single AlaRS C666A mutant, as had previously been observed. All of the suppressors were able to swim as well as wild type

(Fig. 12B). Perturbations to membrane integrity were monitored by antibiotic disk diffusion assays. Once again, all suppressor strains were equally sensitive to both ertapenem and polymixin B as wild-type E. coli, whereas the AlaRS C666A variant was significantly more sensitive to antibiotic exposure (Fig. 12C). The results presented herein highlight that Ala to Ser mistranslation is not sufficient to cause global proteome dysregulation. Having demonstrated that the second-site AlaRS suppressors were able to not only alleviate the slow growth defect associated with the AlaRS C666A variants (Fig.

10), but also likely restore the global proteome homeostasis (Fig. 12) but still not prevent

Ala to Ser mistranslation (Fig. 11), the mechanism for this suppression still remained unclear.

Figure 12. Suppressor mutants alleviate AlaRS C666A-associated phenotypes Second-site suppressors alleviated the A) heat stress, B) swimming, and C) antibiotic sensitivities previously observed in the AlaRS C666A mutant. Statistical significance was determined by one-way ANOVA with Tukey post-hoc comparison (*: p < 0.0005).

3.2.5 In vitro characterization of the AlaRS C666A suppressors

Having previously demonstrated that the AlaRS C666A suppressors do not prevent mistranslation in vivo, biochemical analyses were performed to potentially identify differences in the formation of the Ser-tRNAAla substrates in vitro. Recombinant AlaRS variants were expressed and purified from E. coli and in vitro transcribed E. coli tRNAAla was used as a tRNA substrate for all cognate and non-cognate experimentation. To determine if there were any differences in non-cognate Ser-tRNAAla formation, in vitro mis-aminoacylation experiments were performed using all of the AlaRS variants. The

AlaRS C666A editing-defective variant was able to mis-serylate tRNAAla as expected, and all of the AlaRS suppressor enzymes were also able to mis-aminoacylate tRNAAla (Fig.

13A). While the AlaRS suppressor variants were able to mis-serylate tRNAAla, AlaRS

C666A R561C did exhibit a ~50% reduction in Ser-tRNAAla product formation (Fig. 13B).

Consistent with observations using the in vivo mistranslation reporter, this in vitro analysis suggests that the AlaRS suppressor variants are unable to prevent mis-aminoacylation but a substitution of R561C can potentially lead to an overall decrease in mis-serylated tRNA formation in the cell.

In addition to the direct contribution of Ser-tRNAAla product formation, it was possible that the mechanism for the suppressors could be occurring due to changes in Ser- tRNAAla proofreading. Pre-formed Ser-tRNAAla was incubated with all of the AlaRS variants and deacylation was monitored over time. None of the second-site suppressors were able to proofread the mis-serylated tRNA product similar to wild-type AlaRS levels

(Fig. 13C). The AlaRS R561S variant enzyme exhibited elevated levels of cognate Ala-

75 tRNAAla deacylation, with the percent of deacylated substrates reaching ~40% after 30 minutes (Fig. 13D). Upon further inspection of the initial rates of Ser-tRNAAla deacylation, two of the single mutant suppressors, AlaRS D551Y and AlaRS R561S appear to be proofreading faster than the wild-type enzyme (Fig. 13E). Furthermore, evaluation of the initial deacylation rates for the double mutants suggest that AlaRS D551Y C666A and the

AlaRS R561S C666A enzymes have slightly elevated proofreading activities compared to the editing-defective AlaRS C666A enzyme (Fig. 13F). An additional observation made from the in vitro deacylation experiment is that the single AlaRS R561C variant was unable to deacylate Ser-tRNAAla in trans. As previously described, the AlaRS R561C variant was unable to form mis-serylated tRNAAla product in vitro, suggesting that an additional conformational change may have occurred in this enzyme leading to greater fidelity during initial serine activation that no longer requires proofreading activity in the editing domain.

This hypothesis is further supported by results from the in vivo reporter experiment, which did not reveal Ala to Ser mistranslation in the MG1655 AlaRS R561C mutant background

(Fig. 11).

A secondary measure for determining aaRS proofreading activity is through the monitoring of ATP futile cycling. As has been previously described, the first step in aa- tRNA formation is the activation of free amino acid with ATP to form an aminoacyl- adenylate. During standard cognate aa-tRNA reactions, one molecule of ATP is activated per one aa-tRNA formed, ultimately reaching equilibrium when all tRNA substrates are aminoacylated. Alternatively, when aaRS proofreading is occurring, the mis-activated

76 substrate will be hydrolyzed and can subsequently be re-activated, leading to iterative rounds of ATP consumption.

Observations from monitoring ATP cycling suggest that AlaRS R561S was the most proofreading-proficient AlaRS variant as ATP consumption levels surpassed those of the wild-type enzyme (Fig. 13G). Furthermore, AlaRS R561S displayed increased rates of

ATP consumption when incubated with cognate alanine, consistent with observations that

R561S leads to cognate Ala-tRNAAla deacylation (Fig. 13H). While the AlaRS D551Y enzyme displayed the highest rate of Ser-tRNAAla deacylation in trans, similar levels of

ATP consumption were observed when compared to wild-type, suggesting an imbalance in the rate of deacylation and re-activation. Finally, enzymes containing the R561C variant had the lowest rate of ATP consumption, which is consistent with deacylation studies suggesting that these enzyme are unable to proofread Ser-tRNAAla.

Figure 13. In vitro characterization of AlaRS C666A suppressors The three identified AlaRS C666A suppressors were assayed for A) mis-serylation activity and results from the B) AlaRS R561C mutants were further analyzed. The proofreading activity of the enzymes were examined over a time course for C) Ser-tRNAAla and D) Ala- tRNAAla substrates. Furthermore the initial rates of Ser-tRNAAla deacylation were monitored for the E) single mutants and F) double mutants. Finally, ATP consumption was observed for reactions including tRNAAla and G) serine or H) alanine. Experiments performed by Arundhati Kavoor. 78

3.3 Discussion

In this report we have identified three new suppressor mutations of the MG1655

AlaRS C666A strain. While extensive effort has been put forth to characterize the required residues within AlaRS that maintain accurate translation [57,58], to date, the two new residues described herein had yet to be identified. Previous work has shown that defects in

AlaRS proofreading lead to a variety of pleiotropic phenotypes in E. coli [113]. All of the identified suppressors were able to rescue the phenotypes associated with the AlaRS

C666A strain but surprisingly all still had sufficient levels of Ala to Ser protein mistranslation to cause β-lactam resistance in our in vivo mistranslation reporter. To discern the mechanism by which the three suppressor mutations are rescuing the AlaRS C666A defects, the suppressor mutant enzymes were investigated biochemically. In the following sections, the predicted mechanisms of the three suppressors are discussed.

3.3.1 AlaRS R561C reduces Ser-tRNAAla formation

Biochemical characterization of the AlaRS R561C variants suggests this enzyme has an overall reduction in the potential to form Ser-tRNAAla, as is evident in figure 13B.

Interestingly, the R561C substitution led to a reduction in overall proofreading activity even with C666 present (Fig. 13C). This interpretation is consistent with observations monitoring ATP consumption (Fig. 13G). Taken together, the biochemical analyses suggest that AlaRS R561C is leading to increased discrimination of serine, independent of

AlaRS proofreading.

3.3.2 AlaRS R561S has elevated proofreading activity

In contrast to observations studying the effect of the AlaRS R561C on aaRS discrimination, the AlaRS R561S variant appears to be hyper-active in AlaRS proofreading. The AlaRS R561S C666A variant is able to mis-serylate tRNAAla which is consistent with the observation of Ala to Ser mistranslation in vivo (Fig.11). However, the levels of Ser-tRNAAla do appear to be slightly reduced compared to the AlaRS C666A single mutant (Fig. 13A). The elevated proofreading activity is more obvious when monitoring the deacylation activity directly in editing assays. The R561S variant has elevated rates of Ser-tRNAAla deacylation compared to wild-type AlaRS (Fig. 13E), and the R561S C666A enzyme has elevated proofreading activity compared to editing- defective AlaRS C666A variant (Fig. 13F). Consistent with these observations is the elevated levels of ATP consumption in enzymes containing the R561S substitution. The single AlaRS R561S variant had the highest rate or ATP consumption and the R561S

C666A variant had elevated ATP consumption compared to the C666A enzyme (Fig. 13

G). Interestingly, the R561S enzyme had elevated levels of deacylation (Fig. 13D) and

ATP consumption for cognate Ala-tRNAAla substrates (Fig. 13H). This observation provides the first evidence of a hyper-active proofreading enzyme. Furthermore, the contribution of AlaRS R561S together with the editing-defective C666A substitution may lead to an overall reduction of Ser-tRNAAla in the cell.

3.3.3 AlaRS D551Y increases Ser-tRNAAla deacylation in trans

Compared to observations from the R561 variants, the AlaRS D551Y suppressor is unique in that the double mutant suppressor enzyme had no change in Ser-tRNAAla formation (Fig. 13A). The mis-serylation experiments including the AlaRS D551Y C666A enzyme were the most variable of all other enzymes tested. To circumvent this problem, this experimental condition was tested nine total times compared to the three replicates of the other enzymes. After much consideration, we believe that this variability is not an artifact of technical error but rather, suggestive of perturbation to the overall rate of AlaRS activity. When studying proofreading activity directly, AlaRS D551Y had the fastest rate of Ser-tRNAAla deacylation of all enzymes tested (Fig. 13E). Interestingly, the elevated rate of deacylation was not observed when monitoring the rate of ATP consumption (Fig. 13G).

Together these results suggest that AlaRS D551Y is able to efficiently deacylate Ser- tRNAAla in trans, but is unable to equivalently re-activate serine for iterative rounds of activation and deacylation. It still remains unclear if this discrepancy is due to changes in activation or tRNA transfer, but does suggest that the AlaRS D551Y substitution may reduce the overall Ser-tRNAAla burden caused by the C666A substitution.

3.3.4 Conclusions

Defects associated with errors in AlaRS proofreading have been continually observed from prokaryotes [113] to eukaryotes [59]. In the previous sections we have described for the first time suppressor mutations that are able to alleviate phenotypes associated with AlaRS errors in E. coli. While the identified mutants were unable to

81 completely prevent protein mistranslation, through a series of biochemical experiments, we believe we have identified three unique mechanisms by which AlaRS can reduce the amount of Ser-tRNAAla in the cell. Furthermore, this work highlights the utility of coupling classical genetic approaches and biochemical analyses to further characterize the enzymatic activity of aminoacyl-tRNA synthetases that has previously been unexplored.

3.4 Methods

3.4.1 General Methods

Lysogeny broth (LB) was used for all growth experiments in rich media. M9 media was prepared for the minimal media experiment supplemented with exogenous serine

[113]. SOB and SOC media was used for CRISPR/Cas9-mediated strain engineering.

When applicable, antibiotics were supplemented at the following concentrations: kanamycin, 25 μg/ml; ampicillin, 100 μg/ml for selection and 20 μg/ml for mistranslation reporter; chloramphenicol 30 μg/ml; spectinomycin 50 μg/ml. All DNA oligonucleotides were synthesized by Sigma Aldrich.

3.4.2 Suppressor Mutant Characterization

Suppressor mutant colonies were isolated and genomic DNA was collected using the Wizard Genomic DNA Purification Kit (Promega) following manufacturer recommendations. For whole genome sequencing analysis, 50 ng of genomic DNA was sheared using a Covaris E220 ultrasonicator and sequencing libraries were subsequently prepared using the KAPA Hyper Prep Kit (KAPA Biosystems). Sequencing was performed

82 on an Illumina HiSeq4000 sequencer following the PE150 protocol. Sequencing data were uploaded to the public server at usegalaxy.org and sequencing fragments ranging between

35-151 bp were analyzed using the Galaxy web platform [118]. AlaRS C666A suppressor mutations were also identified using Sanger sequencing of PCR-amplified fragments of the

AlaRS editing domain.

AlaRS protein sequences were aligned using the EMBL-EBI Clustal Omega multiple sequence alignment tool [119] and visualized using Jalview [120]. Positions of the suppressor-substituted amino acids were modeled using Phyre2 [121] and visualized using PyMOL software.

3.4.3 Strain Construction

The AlaRS C666A suppressor in alaS at nucleotide position G1651T (encoding

AlaRS D551Y) was made in an isogenic MG1655 strain using the “gene gorging” method as has previously been described [113]. Suppressor mutations encoding substitutions at

R561 were also generated in isogenic MG1655 derivative strains. These four strains (single

R561C and R561S variants and double R561C/S C666A variants) were constructed using

CRISPR/Cas9 genome engineering [122,123]. Following the cloning strategies described by Reisch and colleagues, the suppressor strains were constructed and validated by Sanger sequencing of the AlaRS proofreading domain amplicon.

3.4.4 Growth Analysis

For all growth experiments, saturated overnight cultures were normalized and back- diluted to an OD600 of 0.05 in the respective experimental media. Cultures were then incubated at either 37°C or 42°C and OD600 values were monitored using a CO8000 cell density meter (WPA) at 30’ time intervals. The data plotted are the averages of three biological replicates with error bars indicating the standard deviation of the replicates.

3.4.5 Motility Assays

To monitor changes in E. coli motility, swimming plates were prepared using LB media supplemented with 0.2% agar. Saturated overnight cultures were normalized to an

OD600 of 0.5 and 0.5 µL of cells were spotted on the agar plates. The swimming plates were incubated at 37°C for 8 hours prior to imaging. E. coli motility was quantified by measuring the swimming diameter using ImageJ.

3.4.6 Antibiotic Sensitivity

Sensitivity to antibiotic exposure was monitored by streaking bacterial lawn cultures on LB agar plates and co-incubation with antibiotic disks (Oxoid). Plates were incubated overnight at 37°C and the sensitivities to ertapenem and polymixin B were quantified by measuring the diameter of antibiotic clearing using ImageJ.

3.4.7 In vivo Mistranslation Reporter

The construction and utility of the pLK-Amp S68A mistranslation reporter has been previously described [113]. In brief, β-lactamase, a gene products used for conferring ampicillin resistance, contains an essential serine residue at position S68. By mutating the gene to encode for an alanine residue at the 68 position, active β-lactamase translation is dependent upon mistranslation of the alanine to serine at that position. The aforementioned mistranslation event can be monitored by the ability for microbial growth in the presence of ampicillin. Reporter plasmids containing either wild-type or mutant β-lactamase genes were transformed into all of the MG1655 AlaRS C666A suppressor strains. The resulting strains were struck onto LB agar plates containing ampicillin and growth was monitored after 48 hours of growth at 37°C.

3.4.8 Preparation of Recombinant Protein and in vitro Transcribed tRNAAla

Genes encoding wild-type AlaRS or the AlaRS C666A variant were cloned into pET21b at NdeI and XhoI restriction sites. The aforementioned cloning strategy generated an in-frame C-terminal His-tag for metal affinity purification as previously described

[113]. To characterize the biochemical activity of the suppressor mutations, site directed mutagenesis was used to create single and double mutant AlaRS expression constructs.

Active enzyme concentrations were determined by active site titration [108,113].

In vitro transcribed E. coli tRNAAla was prepared by slow cooling partially overlapping synthetic DNA oligonucleotides which encode for a T7 promoter and the most abundant tRNAAla isoacceptor in E. coli (genes: alaT, alaU, and alaV). The annealed

85 fragments were then ligated and cloned into EcoRI and XbaI restriction sites in pUC18.

Plasmid containing the E. coli tRNAAla gene was used as a template for PCR amplification of the T7 promoter and tRNA. The PCR-amplified product was subsequently used for T7 runoff transcription. In vitro transcription was performed as previously described [94] and the tRNA was purified using anion exchange chromatography.

3.4.9 Proofreading Activity

To determine if the AlaRS suppressor variants led to changes in mis- aminoacylation in vitro, all single and double mutant AlaRS variants were used for in vitro mis-serylation experiments. To monitor mis-aminoacylation, 5 μM AlaRS was incubated with 5 μM tRNAAla, 750 μM [3H]-Ser, aminoacylation buffer (100 µM HEPES pH 7.2, 30 mM KCl, and 10 mM MgCl2), and initiated with 8 mM ATP. Mis-serylation reactions were carried out for at 37°C 15 minutes and radiolabeled signal was determined by TCA precipitation and subsequent quantification by scintillation counting.

Beyond mis-aminoacylation, trans proofreading activity was monitored to determine if the suppressor mutants altered aminoacyl-tRNA re-binding and subsequent proofreading. To generate preformed aa-tRNAAla, recombinant EF-Tu was first activated

(EF-Tu expression construct was generously provided by Dr. Kurt Fredrick). EF-Tu activation was performed by incubating 50 μM EF-Tu with 50 mM Tris pH 7.8, 100 μM

DTT, 68 mM KCl, 6.7 mM MgCl2, 2.5 mM phosphoenol pyruvate, 30 μg pyruvate kinase, and 500 μM GTP. EF-Tu activation reactions were incubated at 37°C for 30 minutes and used immediately for mis-aminoacylation. Preformed Ser-tRNAAla was subsequently

86 generated by incubating 10 μM tRNAAla with 950 μM [3H]-Ser, 5 μM activated EF-Tu, 5

μM E. coli AlaRS C666A, 8 mM ATP, and aminoacylation buffer. Cognate Ala-tRNAAla was preformed in buffering conditions as described above and with 150 μM [14C]-Ala, 200 nM E. coli AlaRS, and 5 μM tRNAAla. Aminoacylation reactions were incubated for 1 hour at 37°C and subsequently quenched by adding equal volume of acid phenol chloroform.

Following acid phenol chloroform extraction, aa-tRNAAla was EtOH precipitated, before finally being re-suspended in 100 mM sodium acetate pH 4.5. To monitor aa-tRNAAla deacylation activity, 150 nM AlaRS variants were incubated with aminoacylation buffer and preformed aa-tRNAAla. Aliquots of the reaction were quenched on 5% TCA presoaked filter paper at various time points between 0 and 20 minutes. Quenched reactions were then washed 3x in 5% TCA, EtOH, dried, and quantified using scintillation counting. Percent deacylation was determined by the decrease in radiolabeled signal relative to T0.

To monitor ATP consumption, reactions including: 2 μM tRNAAla, 1x aminoacylation buffer, 1 μM AlaRS variants, 0.2 U PPiase, 3 mM [32P – γ] ATP, and either

10 mM alanine or 100 mM serine were incubated at 37°C. At time points ranging between

3’-30’, aliquots of the reaction mix were removed and quenched in equal volume of glacial acetic acid. To monitor the consumption of ATP, migration of the quenched reactions were monitored by thin layer chromatography on PEI cellulose plates. Plates were developed in dipotassium phosphate and imaged using phosphor imaging.

Chapter 4 Targeting tRNA-Synthetase Interactions towards Novel Therapeutic Discovery Against Eukaryotic Pathogens3

4.1 Introduction

Developing therapies against eukaryotic pathogens has proven challenging due to high conservation between the infectious agent drug target and their host counterpart [124].

Of particular concern is the Trypanosome parasite Leishmania that infects upwards of 2 million individuals every year and accounts for more than 50,000 deaths annually [125].

While current treatments of amphotericin B and miltefosine are commonly prescribed to patients with Leishmania infections, they have undesired off-target cytotoxicity, leading to poor patient compliance and low-dose administration [126], and ultimately contributing to the rise of strain-dependent drug resistance [127,128]. There is a strong need for new non- toxic drugs with broad-spectrum activity against different species of Leishmania and other

Trypanosomes [129,130].

Given their essential role in protein synthesis, aminoacyl-tRNA synthetases

(aaRSs) have been an attractive target for antimicrobial therapeutics [131]. AaRSs are essential enzymes found in all domains of life that are responsible for the correct pairing of free amino acids in the cell to their cognate tRNA [5]. AaRSs perform their activity in

3 The work presented in this chapter was originally published in PLoS Neglected Tropical Diseases (14(2): e0007983) and was done in collaboration with Fatemeh Hadi-Nezhad, Dennis Y. Liu, Travis J. Lawrence, Roger G. Linington, Michael Ibba, and David H. Ardell. Experiments performed by co-authors are credited in the corresponding figure legend. All other experiments were performed by Paul Kelly. 88 two steps: first, a free amino acid is activated by the enzyme through the hydrolysis of

ATP, forming an aminoacyl-adenylate. Second, the amino acid is transferred to its corresponding tRNA before being released into the aminoacyl-tRNA pool [5]. Given the complex pool of free amino acids and uncharged tRNAs in the cell, aaRSs have coevolved discrete mechanisms to ensure mutually exclusive amino acid activation and cognate tRNA recognition [132]. The sequence/structural determinants (or anti-determinants) that lead to accurate aaRS:tRNA recognition are also known as the tRNA identity elements. The primary tRNA identity elements that aid in cognate aminoacylation have been extensively studied for several decades [133,134]. For example, across all three domains of life, all tRNAAla isoacceptors contain a conserved G:U base pair in the acceptor stem that is recognized by alanyl-tRNA synthetase (AlaRS), leading to accurate Ala-tRNAAla synthesis in the cell [54,55,135].

While some aaRS inhibitors have successfully made it to the clinic, including the

IleRS-targeting mupirocin [136], ProRS inhibitor halofuginone [137], and the LeuRS inhibitor tavaborole [90,138], there are likely many potential aaRS drugs still to be identified. Target-based approaches relying on structural data and sequence identity have previously been used to try and predict novel Trypanosome aaRS drug targets [139-141], with some recent success [142]. While structure-based approaches have their utility, exploiting aaRS:tRNA interactions has been under-explored for their therapeutic potential.

In particular, while interactions with small molecules are expected to be quite conserved across species, the evolutionary diversification of tRNA identity element interactions through coevolution with aaRSs opens the possibility of greater species-specific inhibition.

While major identity elements have been experimentally characterized for many aaRS:tRNA pairs in various model systems, much less is known about how tRNA identity elements evolve and diverge over the Tree of Life. Recent theoretical advances explain how tRNA identity elements can evolve and diverge in a phylogenetically informative way, even while under strong selective constraints [143]. Earlier work, developed a bioinformatic method to predict tRNA identity elements [144]. These bioinformatic predictions are called Class-Informative Features (CIFs), based on the statistic of structure- conditioned Shannon Information [145], and visualized through graphs called Function

Logos [144]. Later work, applied two other statistics, Information Difference and

Kullback-Leibler Divergence, to facilitate pairwise comparisons of CIFs between two taxa, in two new visualizations called Information Difference (ID) logos and Kullback-Leibler

Divergence (KLD) logos, respectively [146]. ID logos visualize gains and losses of CIFs, which in this work we call gains and losses of information, while KLD logos visualize the functional conversion of CIFs from one functional type of tRNA to another, which in this work we call change of functional information. In the present work, we integrate together all three statistics (structure-conditioned information about function, ID, and KLD) and apply it to the problem of identifying parasite-specific tRNA identity elements. Our approach visualizes functionally informative features in parasite tRNAs that have either gained or retained functional information relative to humans, altered functional associations, or both, since divergence from their common ancestor with humans.

Our modeling approach integrates genomic tRNA sequence variation across multiple tRNA gene families of different functions, revealing potentially useful

90 information about the specification of substrate identity for all aaRSs simultaneously. The multiplicity of aaRSs in cells provides multiple potential targets for inhibition of essential parasite enzymes, opening the door to improved combination chemotherapies. Advances in systems biology and chemogenomics have fueled interest in combination chemotherapies to benefit from synergistic drug interactions [147-150] and combat the evolution of resistance [151]. Combination chemotherapies are naturally effective, for example, in the pathogenic defenses of arthropods [152] and have yielded exciting antifungal [153] and antihelminthic [154] therapies. Additionally, artemisinin-based combination therapies (ACTs) are the primary treatment plan for Plasmodium falciparum malaria infections [155,156].

Here, we report our improved annotation of TriTrypDB genomes and new methodologies for predicting conserved identity elements across biological domains. As proof of principle, we screened for identity element divergence between Trypanosomes and humans to search for new therapeutic targets for these eukaryotic pathogens.

Validating our computational approaches, we found several natural product fractions that inhibit Leishmania major AlaRS activity but have no effect on the homologous human enzyme. The fractions we identified also have inhibitory activity against Trypanosoma cruzi AlaRS, showing that our approach holds promise towards identifying new broad- spectrum anti-Trypanosomal therapies.

4.2 Results

4.2.1 Custom Annotation of tRNA Genes and Gene Clusters in TriTrypDB Genomes

We obtained 4381 unified gene records from the raw output of two tRNA gene- finders, Aragorn and tRNAscan-SE v.2.0, to TriTrypDB v.41. Of these, 3597 were found by both gene-finders, 750 were found by Aragorn only, and 34 were found by tRNAscan-

SE 2.0.0 only (Table 4). We identified the same 76 genes as initiator tRNA genes, using either tRNAscan-SE 2.0.0’s profile-based predictions or our own edit-distance-based clustering approach, by finding the unique set of genes carrying conserved initiator tRNA features as described in [157].

To further refine the final annotated gene set, we identified tRNA gene clusters in

TriTrypDB genomes using a maximum intergenic distance criterion of 1000 bp on either strand. Doubling this distance criterion did not substantially increase cluster number or size. After filtering 4381 gene records by their gene-finder scores as described below, 3616 high-confidence gene records remained, of which 77% occur in clusters of size two or greater (Fig. 14). The largest tRNA gene clusters were of size ten, accounting for 9% of total genes. We used Jaccard distance as a gene functional content distance to hierarchically cluster tRNA gene clusters with similar gene functional contents, and found that distance cutoffs between 0.680 and 0.692 defined intuitively reasonable similar, distinct, and putatively homologous tRNA gene cluster variant groups that we found to be conserved either within each of the Leishmania and Trypanosoma genera, or across both genera. The conservation of tRNA gene cluster variants and groups spanning TriTrypDB genome assemblies of different genera is indisputable, but further statistical and phylogenetic 92 characterization of them may best be undertaken via long-read genome resequencing, as tRNA gene clusters can be difficult to assemble reliably from short-read sequencing data.

Table 4. Statistics on final tRNA annotation gene set

Figure 14. tRNA gene cluster size distribution for Leishmania, Trypanosoma, and other TriTrypDB genomes Green labels at tops of stacks show percentages of total tRNA genes in clusters of a given length. Numbers within each bar show frequencies of gene clusters of that length for the given clade. Gene clusters from the Trypanosoma clade are represented in teal, clusters from the Leishmania clade in yellow, and other kinetoplastid clades (including Leptomonas and Crithidia) are in gray.

Based on this evidence, we retained 3616 genes from 46 TriTrypDB genome assemblies that had an Aragorn bit-score of at least 107 or a tRNAscan-SE 2.0 bit-score of at least 50, including 36 genes found by Aragorn only and 1 gene found by tRNAscan-SE

2.0 only.

The median number of genes per genome in our raw union annotation gene set was

82. Among these were 45 functionally ambiguous but high scoring genes, including 2 with identity unassigned by both gene-finders, 6 marked as pseudogenes or truncated by tRNAscan-SE 2.0.0, 4 containing sequence ambiguities, and 33 with conflicting structural and anticodon annotations. Furthermore, ten intron-containing genes were predicted as tRNATyr genes, all from genomes in the American Trypanosoma clade [158]. Our score- filtered union annotation gene-set was further filtered and pooled into defined clade gene- sets and mean and standard deviations of tRNA gene number and pooled composition statistics by clade were determined (Table 5). The gene compositions of the Leishmania clades display some similarities, yet also show unique differences compared to those of both humans and Trypanosoma. Human gene composition is most divergent with African

Trypanosoma gene composition, and second most divergent from those of Leishmania.

American and African Trypanosoma compositions are less divergent from each other, however, Trypanosoma and human tRNA gene compositions are different from those of

Leishmania in different ways. Human tRNA genes are richer in purines while

Trypanosoma tRNA genes are richer in G and C. American Trypanosoma are also GC- rich, but less so than African Trypanosoma. There is little heterogeneity of gene set

96 compositions by genome assembly within clades, with the greatest variation appearing in

Trypanosoma, particularly African Trypanosoma.

Table 5. Clades and genomes analyzed with statistics on Class-Informative Features estimation gene sets .

4.2.2 Divergent Class-Informative Features between Humans and TriTrypDB Genomes

We developed a bioinformatic workflow that combines information from tRNA function logos estimated from a parasite clade and Information Difference (ID) logos [144] and Kullback-Leibler Divergence (KLD) logos between the parasite clade and humans

[146]. The workflow quantitates tRNA features that are functionally informative in the parasite clade and have either gained or retained functional information or altered functional association since divergence of the parasite clade and humans from their common ancestor. We found many examples of highly informative Trypanosome CIFs that have been gained, retained or changed functional information since divergence from their common ancestor with humans, and most of these divergent CIFs have been strongly conserved in Trypanosomes over 231–283 million years of evolutionary divergence between Leishmania and Trypanosoma [159], for example among alanine tRNAs (Fig. 15) and threonine tRNAs (Fig. 16). Even though they are calculated at single-site resolution,

CIF divergences are correlated across structurally paired sites. Inspection of single-site function logos across taxa confirms the conservation of parasite-specific CIFs and reveals

A- and U-containing features underlying the signals shown in Figures 15 and 16, including some sharing of divergent features between tRNAAla and tRNAThr functional classes, for example at Sprinzl coordinate 39 (Fig. 17). This can be seen more easily by inspecting function logos for paired sites directly. Figures 18-20 show base-pair function logos for humans, the L. major clade and the American Trypanosoma respectively, showing that both Class-Informative Base-Pairs (CIBPs) and Class-Informative Mis-Pairs (CIMPs) can be relatively conserved, and that recurring hot-spots of CIF evolution appear in the data, 99 yielding insight to mechanisms of CIF evolution. Inspection of Class-Informative Base-

Pairs and Mis-Pairs shows that a U:A informative base-pair diverged in tRNAThr to an adjacent site-pair, from 31:39 to 30:40, and that a U:U informative Mis-Pair was gained in tRNAAla at site-pair 6:66 in Trypanosomes relative to humans (Figs. 18-20).

Our computational screen for tRNA CIF divergence indicated that tRNAAla and tRNAThr are among those tRNA functional types that have the greatest number of sites and site-pairs with the largest CIF divergence relative to humans, and would be good potential candidates for therapeutic targeting. Contrast, for example, our results for Trypanosome tRNATyr or tRNATrp, which show Trypanosomal tRNA CIFs that are strongly conserved with humans, as shown in figure 21. These observations led us to follow-up and investigate tRNAAla and tRNAThr because both of these tRNA types, and their accompanying cognate synthetases, are readily reconstituted in vitro [31,33,57,59].

100

Figure 15. Conserved divergence of parasite tRNAAla CIFs across eight phylogenetic clades of Leishmania and Trypanosoma. Analysis performed by Ardell lab.

101

Figure 16. Conserved divergence of parasite tRNAThr CIFs across eight phylogenetic clades of Leishmania and Trypanosoma. Analysis performed by Ardell lab.

102

Figure 17. Adenine function logos for humans and four clades of Leishmania The total height of a stack of letters at any site quantifies the information potentially gained about the functional type of a tRNA by a tRNA-binding protein if it recognizes the specific feature corresponding to that site and logo. The letters within each stack symbolize functional types of tRNAs, wherein IUPAC one-letter amino acid codes represent elongator tRNA aminoacylation identities and “X” symbolizes initiator tRNAs. The relative heights of letters within each stack quantifies the over-representation of tRNA functional types carrying that feature relative to the background frequency determined by gene frequencies of functional types (as calculated through the normalized log-odds). Analysis performed by Ardell lab.

103

Figure 18. Function logos for tRNA Class-Informative Base-Pairs and Mis-Pairs in humans Details regarding the interpretation of this data are the same as described in figure 17. Analysis performed by Ardell lab.

104

Figure 19. Function logos for tRNA Class-Informative Base-Pairs and Class- Informative Mis-Pairs in L. major clade Details regarding the interpretation of this data are the same as described in figure 17. Analysis performed by Ardell lab.

105

Figure 20. Function logos for tRNA Class-Informative Base-Pairs and Class- Informative Mis-Pairs in T. cruzi clade Details regarding the interpretation of this data are the same as described in figure 17. Analysis performed by Ardell lab.

106

Figure 21. CIFs for tRNATyr and tRNATrp are strongly conserved between parasites and humans. Conserved divergence of parasite tRNA CIFS for A) tRNATyr and B) tRNATrp. Analysis performed by the Ardell lab.

107

4.2.3 AaRS Activity Screen Identified Leishmania major AlaRS Inhibitors

Using a pre-validated MNP library (described in more detail in section 4.4.5), we developed a medium-throughput phosphor imaging-based aminoacylation screen to identify aaRS inhibitors in vitro (Fig. 22A). From the one hundred and twenty complex inhibitory mixes tested in the MNP library, we qualitatively identified four potential L. major AlaRS inhibitors as determined by a decrease in the overall tRNA aminoacylation signal (Fig. 22B). These four candidates were then re-screened using time-dependent quantitative approaches and we concluded that three of the four mixes, 1881C, 2059D, and

2096B were altering aminoacylation, with inhibitory activities ranging between 80% and

99% (Fig. 22C).

Since the aminoacylation screen discerns total net changes to the aaRS activity, we attempted to identify which part of the two-step aaRS catalyzed reaction is being affected by the MNPs. To observe any tRNA-independent effects on aaRS function, we used pyrophosphate exchange to monitor ATP-dependent amino acid activation. From this experiment, we were able to conclude that our inhibitors were perturbing amino acid activation, with lead compounds ranging in inhibitory activity between 45% and 95%. The differences in MNP activity between amino acid activation and tRNA-dependent aminoacylation highlight the multiple aaRS activities that can be targeted in our network predictions. To validate the predictive tool for identifying anti-Trypanosomal drugs, we counter-screened the newly identified L. major AlaRS inhibitors against the human AlaRS enzyme. Treatment of the human AlaRS enzyme with the MNP inhibitors had no effect on amino acid activation (Fig. 22D). Combined with our original screening data, these results

108 show the utility of our computational and biochemical workflow to identify new novel therapeutics that have minimal cross-reactivity with the human homolog of the parasite drug target.

109

Figure 22. Identification of Leishmania major AlaRS inhibitors A) Workflow to identify aminoacylation inhibitors (details described in Methods). B) Representative image of the MNP chemical screen. The spot boxed in red is an example of a predicted inhibitor depicted by the decrease in signal intensity. DMSO positive control (+). C) Three of the four identified inhibitors prevented the accumulation of Ala-tRNAAla formation, 1428B was a false-positive result from our preliminary screen. D) The three identified inhibitors perturbed L. major AlaRS activation (black) but had no effect on human AlaRS (gray). The relative amino acid activation is plotted relative to the DMSO control. Error bars indicate the standard deviation of three replicates.

110

4.2.4 Natural Product Library Inhibitors of Leishmania major ThrRS

As our network predictions identified CIF divergence among many functional classes of tRNAs, we also wanted to determine if our MNP library screen would find inhibitors against non-AlaRS aaRS. The most concentrated mixes from our MNP library were re-screened against L. major ThrRS and tRNAThr aminoacylation (Fig. 23A). The preliminary screen led to the identification of eight extracts with inhibitory activity. Those compounds were re-analyzed using quantitative aminoacylation reactions and the results show that that two of the candidates did not inhibit aminoacylation, two inhibited the reaction at ~50%, and four had greater than 75% inhibition (Fig. 23B). In addition, two of the four most active inhibitors (2059D and 2096B) also had activity against L. major AlaRS

(Fig. 23C). The cross-reactivity of these inhibitors may be a consequence of the extensively conserved aaRS architecture found between AlaRS and ThrRS [57,160].

111

Figure 23. Identification of Leishmania major ThrRS inhibitors A) The MNP library was re-screened at the highest concentrations to qualitatively identify Leishmania major ThrRS aminoacylation inhibitors. Plate IDs reference the position within the original library and not library IDs. B) Eight inhibitors were qualitatively identified from the preliminary screen. Two of the candidates did not inhibit aminoacylation (black), two inhibited at ~50% activity (gray), and four inhibited at greater than 25% (white).C) All four active inhibitors continued to perturb aminoacylation over a time course experiment. Error bars indicate the standard deviation of three replicates.

112

4.2.5 Predictive Network Interactions Identified Broad-Spectrum Anti- Trypanosomal Targets

The aaRS:tRNA network analyses suggested that parasite-specific tRNAAla identity elements were highly conserved between the Leishmania and Trypanosoma genera (Fig.

24A). To test this hypothesis, we purified T. cruzi AlaRS and screened our three active L. major AlaRS inhibitors in an aminoacylation inhibition assay using T. cruzi AlaRS and tRNAAla. Supporting our network prediction, all three L. major inhibitors also had activity against the T. cruzi enzyme, with activities ranging between 40% and 95% total inhibition

(Fig. 24B). While these activities were slightly reduced compared to their effect on the L. major AlaRS enzyme (Fig. 22C), these results highlight the additional potential utility of our computational methodologies as a means of identifying broad-spectrum antimicrobials for closely-related clades.

113

Figure 24. Leishmania major and Trypanosoma cruzi AlaRS have conserved tRNA identity elements A) CIF Divergence Models for tRNAAla in Leishmania major and Trypanosoma cruzi B) The three identified Leishmania major AlaRS inhibitors also have activity against the Trypanosoma cruzi AlaRS enzyme. Error bars indicate the standard deviation of three replicates

114

4.2.6 Separation of Active Components from Natural Products Extracts

From the initial set of 120 extracts with activity against L. donovani parasites, four extracts showed corresponding activity in the initial aaRS assay. Of these, three (1881C,

2059D, and 2096B) were prioritized for chemical follow up, based on potent, dose dependent biological activity. Initially, each sample was separated into 10 sub-fractions using HPLC (Phenomenex Synergy C18, 5µ, 4.6 x 250 mm). Screening of these fractions identified one fraction (2096B F10) with potent activity. To generate additional material, the producing organism was cultured on large scale (1 L, GNZ medium with 20 g XAD-7 resin), filtered, and the resin/cell slurry extracted with organic solvents (2:1 CH2Cl2/

MeOH, 400 mL). The crude extract was fractionated using an automated Combiflash chromatography system (C18 cartridge; 20, 40, 60, 80, 100% MeOH/ H2O, 100% EtOAc) and the resulting fractions subjected to biological screening (Fig. 25). Two fractions (C and

D) showed strong activity and were subjected to subsequent separation to give 10 sub- fractions each. Of these, fraction 2096D F10 showed the strongest reproducible activity

(Fig. 26). However, subsequent fractionation steps yielded sub-fractions with very low quantities of material. Review of these sub-fractions by UPLC-ESI-qTOF mass spectrometry did not identify any individual mass signatures consistent with a candidate bioactive molecule. Similarly, lack of material precluded the identification of diagnostic signals in the NMR spectra for these subfractions. Provisional information from these analyses, including NMR and MS signatures from earlier fractions and the non-polar nature of the active fractions, suggest that the active component is likely a bioactive lipid, although the precise nature of the structure of this metabolite remains unknown. The isolate

115 producing the bioactive substance was collected on April 20th, 2012 from marine sediment off the coast of Kellet Bluff, Henry Island, WA US under the permit issuing authority of the Washington Department of Fish and Wildlife (permit # 12-034).

116

Figure 25. Workflow for the extraction of RL12-182-HVF-D Schematic diagram of the chromatographic separation of RL12-182-HVF-D. Chromatography performed by Dennis J. Liu.

117

Figure 26. Marine natural product extract 2096D F10 inhibits Leishmania major AlaRS aminoacylation. Comparison of aminoacylation reactions with or without 2096D sub-fractions indicate inhibitory activity found in fraction 10.

118

4.3 Discussion

4.3.1 Systems-biology driven identification of Trypanosome-specific drug targets

tRNA CIFs apply an information criterion using function logos, rather than a conservation criterion using conventional sequence logos, to bioinformatically predict tRNA identity elements. Even though we did not apply a conservation criterion in our predictions, when we applied our information criterion independently over different

Trypanosome clades, we found that tRNA CIFs were highly conserved over 250 million years of Trypanosome evolution. A biological interpretation of this result of tRNA CIF conservation within Trypanosomes (and also between Trypanosomes and humans) is that the information contained in tRNA CIFs is functional in specifying substrate identity to tRNA-binding proteins such as aaRSs. That is to say, tRNA-binding proteins themselves exploit the information contained in tRNA CIFs to identify their tRNA substrates against the background of all possible tRNAs, with which they must interact to varying degrees. A systems biological theory for the function and divergence of tRNA CIFs is presented in

[143].

Maintaining efficient and accurate translation is predicated on catalytically productive interactions between aaRSs and free tRNAs in the cell. While the major identity elements for a given aaRS:tRNA pair are generally conserved, here we have identified divergent features within tRNAs that apparently contribute to divergent RNA-protein interactions in Trypanosomes. Much of the focus in this work was on the phylogenetic divergence of identity elements among alanine tRNAs. This class of tRNAs strongly support the utility of our computational analyses as the tRNAAla identity elements have

119 been one of the most well characterized to date [54,135]. Interestingly, it was recently shown that the conserved G3:U70 base pair is recognized by AlaRS using three distinct mechanisms across all domains of life [55]. This observation highlights that even highly conserved identity elements may be recognized and discriminated against by distinct biophysical aaRS interactions, which may therefore be stronger potentially specific drug targets than previously anticipated. The dominant association of G3:U70 with tRNAAla is conserved among all Trypanosome clades and humans in our data (Figs.18-20).

The primary objective of this research was to develop a computational workflow to quantify divergence of functionally informative features of tRNAs across different evolutionary clades. The practical application of this work is to use the information gained from our computational analyses to identify novel therapeutic targets that may be of use in the clinic. As described above, tRNAAla and tRNAThr were specifically chosen because of their amenability for in vitro reconstitution, while the computational results further suggest that other Trypanosome aaRS:tRNA pairs could serve as additional therapeutic targets either using our MNP library or other available libraries. While interesting, those discoveries are outside the scope of the present work and left to future investigations.

4.3.2 Inhibition of aminoacyl-tRNA synthetases

A goal of this work was to identify divergent tRNA identity elements in

Trypanosome parasites. We predicted that parasite-specific aaRS:tRNA interactions would be identified, sufficiently divergent from homologous human machinery to be strong candidates for drug discovery. Interestingly, our network divergence analysis led to the

120 discovery of tRNA-independent, amino acid activation inhibitors that were specific to

Trypanosomes. We interpret this as consistent with our goal, because tRNAs and aaRSs must coevolve to accommodate changes to structure and mechanism that evolve on either side of their interactions. Presumably, divergence in the structural mechanism of amino acid activation in Trypanosome AlaRSs has also changed how they interact with their tRNA substrates. By integrating information from many tRNA functional classes, we gain leverage to interpret divergence in structure and function of the much more structurally complex ensemble of aaRSs as a system. Our tRNA-based network approach identifies potential aaRS targets that may not have been initially predicted when analyzing aaRS functional sequences in isolation.

4.3.3 Chemotherapeutic inhibition of multiple aminoacyl-tRNA synthetases may be relatively resistance-proof

Two of the compounds we described were effective inhibitors of both AlaRS and

ThrRS in parasites. Although monotherapeutic inhibitors of aaRSs are highly effective

[161], combination therapies involving multiple aaRSs have not been studied. Because aminoacylation pathways are integrated in parallel at the ribosome, the slowest aminoacylation pathway can be rate-limiting for protein synthesis and growth [162]. Thus, we expect the inhibition of multiple aaRSs to be antagonistic relative to Loewe Additivity expectations, in keeping with the Highest Single Agent (HSA) model [163]: single- or multiple-drug inhibition of multiple aaRSs should mask the potentially growth-restorative effects of resistance mutations arising in any one parasite aaRS gene. It is known that antagonistic combination chemotherapies are less prone to the evolution of resistance 121

[164,165]. Therefore, chemotherapeutic inhibition of multiple aaRSs should be relatively less prone to the evolution of resistance than monotherapeutic or synergistic combination chemotherapeutic inhibition of single aaRSs. Further work is needed to test this hypothetical benefit.

4.4 Methods

4.4.1Annotation, Clustering and Filtering of tRNA Genes in TriTrypDB Genomes

We downloaded data for 46 genome assemblies from TriTrypDB version 41 released December 5th, 2018. We ran tRNAscan-SE v.2.0.0 installed via BioConda in

February, 2019 [166] and Aragorn v.1.2.38 [167] using option “-i116” (implying a maximum intron length in search targets of 116 base-pairs) on this data. We unified gene records from the two finders if they overlapped by at least one base-pair, had consistent strand-orientations and end-displacements less than or equal to 4 bp. To independently identify initiator tRNA genes, we computed edit distances [168] of CAT-anticodon- containing genes implemented in the function stringdist from its R package v. 0.9.5.5 and clustered them agglomeratively using Ward’s minimum variance method [169] implemented in the function hclust with method ward.D2 from the base R stats package, examining clusters for the initiator-distinguishing features described in [157]. All statistical analyses and sequence processing for annotation and clustering were carried out in R v.3.6

[170].

To further investigate these gene records, we examined their genetic clustering in

TriTrypDB genomes as defined by co-occurrence within a distance of 1000 bp on either

122 strand. We computed gene function content distances of tRNA gene-clusters as pairwise

Jaccard distances considering gene clusters as sets of functions using stringdist and clustered them with Ward’s method using function hclust with method ward.D2. We finalized our annotation union gene-set by retaining 3616 genes that had an Aragorn score above 106 bits or a tRNAscan-SE2 score above 49 bits, and re-annotating sequences as described in the Results.

4.4.2 Prediction of Divergent tRNA Class-Informative Features (CIFs) in Humans and Parasites

To compare CIFs between TriTrypDB genomes and humans and to have sufficient data to estimate Trypanosome CIFs, we defined eight phylogenetic clades for 39 of the 46

Trypanosome genomes as shown in Table 5. These clades were based on a composite of phylogenetic results in the literature [171-174]. CIFs were subsequently estimated for each clade independently, by pooling tRNA genes within clades. We removed two incomplete genomes from analysis, T. rangeli SC58 and T. cruzi CL-Brenner that had fewer than half the number of tRNA genes identified in any other genome and missed more than two functional classes. We filtered the gene annotation union gene set of 3616 genes, removing selenocysteine genes, pseudogenes, truncated genes, and genes of ambiguous function, leaving 3488 high-confidence functionally annotated gene records from 44 genomes in

TriTrypDB v.41 for alignment. To this set we added 431 high-confidence human tRNA gene records downloaded from GtRNADB [175] on May 15, 2019 (in the file “hg38- tRNAs.fas”), excluding two human selenocysteine tRNA genes, to yield a grand total of

3919 tRNA genes from 45 genomes for structural alignment. We aligned this alignment 123 gene-set of 3919 genes using COVEA v.2.4.4 [176] to the eukaryotic tRNA covariance model supplied with tRNAscan-SE v.1 [177]. The output alignment was manually edited in SEAVIEW [178] to correct the misalignment of 595 human and Trypanosome tRNA genes (almost exclusively of type tRNALeu and tRNASer) at Sprinzl coordinates 45 and 47 and exclude majority-gap/insertion and variable-arm-containing sites (Sprinzl coordinates are a standardized coordinate system that encodes both the consensus universal secondary structure of tRNAs, and conserved, more functionally-specific structures like the long variable arms of tRNALeu and tRNASer [179]). Sequences were further processed with the

FAST toolkit to partition genes into clades [180]. After excluding an additional 464 genes from five genomes not included in our defined clades, 3455 aligned Trypanosome and human genes remained. More statistics on the CIF estimation gene sets by clade are shown in Table 5.

For each clade we independently computed function logos [144], Information

Difference logos and Kullback-Leibler Divergence logos [146] with a newly written

Python 3 program tSFM (tRNA Structure-Function Mapper) v0.9.14 available on github

(https://github.com/tlawrence3/tsfm), which we describe briefly here, and more fully in a forthcoming publication. tSFM provides a command-line user interface for estimating function, ID, and KLD logos using our published methods. tSFM additionally calculates tRNA CIFs for secondary-structure feature pairs, in addition to single-site features. Class-

Informative Feature Pairs are elements of the Cartesian product set C = f × f×BP, where f={A,C,G,U,-} is the set of single-site features we consider and BP is the set of structurally- paired Sprinzl Coordinates involved in potential base-pairing interactions along the four

124 arms of the planar clover-leaf consensus secondary structure of tRNAs [179]. We ran tSFM with option “-x 1” corresponding to computing exact expected entropies for samples of size one by the method of [181] or by the Bayesian Nemenman–Shafee–Bialek (NSB) entropy estimator [182] otherwise.

Briefly, we computed the gain-of-information of a CIF in a particular functional class and Trypanosome clade as its information difference in bits, with that clade as foreground and humans as background, multiplied by the normalized ratio of posterior-to- prior odds of the CIF in that functional class in Trypanosomes and humans, corresponding to letter heights in ID logos, and measured in bits. We computed change-of-function of a

CIF in a particular functional class and Trypanosome clade as its Kullback-Leibler

Divergence in bits, with that clade as foreground and humans as background, multiplied by the normalized ratio of posterior-to-prior odds of the CIF in that functional class, corresponding to letter heights in KLD logos and measured in bits. To avoid division by zero when calculating KLD, we added pseudocounts to either the background or the foreground posterior distributions when one or more of the 21 functional classes was not observed. When calculating the normalized ratio of posterior-to-prior odds for a specific functional class, we only added pseudocounts to the background posterior distribution.

Furthermore, to avoid inaccuracies, we defined the KLD of a feature to be zero when its frequency in the background is less than or equal to five.

We wrote a custom script in R 3.6 to visualize CIFs within each cluster for each functional class of tRNA in a structural context, and color the parasite CIFs according to

125 whether those CIFs have gained information or changed functional information relative to human since divergence from their common ancestor.

4.4.3 AaRS Cloning and Protein Purification

L. major AlaRS and L. major ThrRS-encoding genes were codon optimized, synthesized, and sub-cloned into pUC57 (GenScript). Engineered flanking NdeI and SmaI restriction sites were used to clone the aaRS genes into pTYB2, creating in-frame C- terminal intein fusions. The resulting expression vectors were transformed into the E. coli expression strain BL21 (DE3). The gene encoding T. cruzi AlaRS was codon optimized, synthesized, and directly cloned into NdeI and XhoI cut sites in the pET21b expression vector (GenScript). The resulting plasmid expressed T. cruzi AlaRS under T7 control and was in-frame with a C-terminal 6x-His tag. The pET21b-Tc AlaRS vector was transformed into the E. coli expression strain XJb (DE3) (Zymo Research).

Both L. major AlaRS and ThrRS were purified by growing cells to an OD600 of

~0.5 and cooling on ice for 30 minutes. Protein induction was initiated by the addition of

IPTG to a final concentration of 500 µM and cells continued to grow at 16°C for 16 hours.

Cells were harvested by centrifugation and lysed by sonication in Buffer A (25 mM HEPES pH 7.2, 500 mM NaCl, 3 mM DTT) with cOmplete mini protease inhibitor (Sigma) added.

Clarified lysate was added to a chitin resin column (NEB) and washed with Buffer A. The intein tag was cleaved by the incubation of Buffer B (25 mM HEPES pH 7.2, 100 mM

NaCl, and 100 mM DTT) on the resin bed overnight at 4°C. Protein was dialyzed in two

126 stages in Buffer C (25 mM HEPES pH 7.2, 30 mM NaCl, 6 mM BME, and 10% - 50% glycerol).

T. cruzi AlaRS-expressing cells were grown to an OD600 ~0.3 and then cooled to

18°C and induced with 500 µM IPTG. Cells were grown for an additional 16 hours at 18°C before harvesting by centrifugation. Cell pellets were re-suspended in lysis buffer [Buffer

I (500 mM Tris-HCl pH 8.0, 300 mM NaCl, and 10 mM imidazole) with cOmplete mini protease inhibitor (Sigma)], sonicated, clarified, and cell lysate passed over a TALON metal affinity column (Takara). After washing the column with Buffer I, protein was eluted with Buffer II (Buffer I with 250 mM imidazole). Protein was dialyzed in two stages to remove imidazole and to store the enzyme in 50% glycerol.

Human AlaRS was expressed in E. coli Rosetta (DE3) (Novagen) from pET21a which encodes the human AlaRS gene in-frame with a C-terminal 6x-His tag (expression plasmid provided by Karin Musier-Forsyth, Ohio State University). Cells were grown to an OD600 of ~0.5 and cooled on ice for 30’ before inducing expression with 500 µM IPTG.

Upon induction, cells grew for an additional 16 hours at 20°C before harvesting. Human

AlaRS was purified as described above with the addition of 5 mM β-mercaptoethanol to both Buffer I and Buffer II. All enzyme concentrations were determined by active site titration using [108,183] [14C]-alanine (Perkin Elmer) and [14C]-threonine (American

Radiochemicals).

4.4.4 Preparation of in vitro Transcribed tRNA

L. major tRNAAla (chr11. trna1-AlaCGC), L. major tRNAThr (chr23. trna6-

ThrTGT), and T. cruzi tRNAAla (TctRNA-Ala.03) DNA sequences were cloned into EcoRI 127 and XbaI restriction sites in pUC18 by slow cooling complementary synthetic DNA oligos and ligation as previously described [184]. PCR was used to amplify 50 µg DNA template from the pUC18-tRNA plasmids to be used for T7 runoff transcription. In vitro transcription was performed with 40 mM Tris-HCl pH 8, 2 mM spermidine, 22 mM MgCl2,

5 mM DTT, 50 µg/mL BSA, 4 mM NTPs, 20 mM 5’GMP, 20 U Protector RNase Inhibitor,

2 U pyrophosphatase, DNA template, and T7 RNAP at 42°C for 16 hours. Transcription products were purified on a Diethylaminoethyl (DEAE) Sephacel (GE Healthcare) column in 20 mM Tris-HCl pH 8.0, 5 mM MgCl2, and 250 mM NaCl. tRNA was eluted from the resin with 1 M NaCl. The RNA was precipitated overnight at -20°C in 1/10th volume sodium acetate and 3x volume ethanol and re-suspended in RNase-free H2O.

4.4.5 Marine Natural Product Library

The marine natural products screening library comprises 5,304 fractions from organic extracts of marine-derived Actinobacterial fermentations (1 liter culture, following our standard protocol [185]). All fractions are stored as concentrated stock solutions in

DMSO in standard 96-well format. The library is comprised of extracts of marine sediment-derived bacterial strains, containing a cross section of gram-positive genera and enriched in Actinobacterial strains, hand-collected from over 70 discrete dive sites on the

West coast of the United States from the Channel Islands of Southern California to the San

Juan Islands in Northern Washington.

Crude extracts were fractionated in to six sub-fractions on Seppak C18 cartridges using a stepwise elution profile (20, 40, 60, 80, 100% MeOH/ H2O, 100% EtOAc). The

128 resulting fractions were solubilized in DMSO (1 mL per fraction), 4 µL aliquots diluted

1:5 in DMSO, and arrayed in 384 well format (17 x 384 well plates). The MNP library screened in this assay consisted of a focused group of bacterial extract pre-fractions that had already demonstrated activity against Leishmania in a prior whole cell assay against

L. donovani amastigotes. The MNP library was also counter-screened in a mammalian system against HeLa cells [185]. Fractions with acute cell cytotoxicity were removed from the screening library, resulting in a set of test extracts with demonstrated activity against

L. donovani and low/ no cytotoxicity against HeLa cells. Following primary screening against L. donovani amastigotes, 120 active fractions were arrayed as serial dilutions (8 x

2-fold dilutions; 50 - 0.4 µM) in 96 well format for aaRS screening.

4.4.6 Screen for Aminoacylation Inhibitors

Serial dilutions from the marine natural product (MNP) library were screened using the following protocol. Aminoacylation reactions were performed at room temperature using 10 mM DTT, 8 mM ATP, 5 µM tRNA, 60-80 µM [14C]-Ala or [14C]-Thr, 100-500 nM aaRS, and DMSO or MNP samples. After incubating the reaction for either 15 or 20 minutes, 1 µL of the reaction was spotted on 5% pre-soaked TCA 3 MM Whatman filter paper. The precipitated tRNA-bound filter paper was washed 3x with 5% TCA, washed once with ethanol, and dried. The dried filter paper was exposed overnight on a phosphor imager screen and imaged the following day. Qualitatively, the phosphor image screen was examined for a change in signal intensity relative to the DMSO control; a decrease in phosphor image intensity indicates partial or full inhibition of the reaction in the presence

129 of the inhibitor. While active concentrations were unknown for each of the MNP mixes, the serial dilution helped prevent false-positive identification. All lead candidates from the preliminary screen were confirmed using similar reaction conditions; the reactions were monitored over a time course and placed at 37°C. Samples were quantified using a scintillation counter.

4.4.7 Pyrophosphate Exchange

Amino acid activation was monitored using ATP/PPi exchange as previously described [28]. Reactions were performed at 37°C in 100 mM HEPES pH 7.2, 30 mM KCl,

32 10 mM MgCl2, 2 mM NaF, 2 mM ATP, 2 mM [ P]-PPi (Perkin Elmer), 90 µM alanine,

160 nM AlaRS, and DMSO or aaRS inhibitor. At increasing time points, aliquots of the reaction mixture were quenched in a charcoal solution containing 1% activated charcoal,

5.6% HClO4, and 75 mM PPi. Quenched reactions were vacuum filtered on to 3MM

Whatman filter discs, washed three times with 5 mL of water and once with 5 mL of ethanol. After drying the filter discs, charcoal-bound radiolabeled ATP was quantified on a scintillation counter. Relative endpoint amino acid activation was determined by comparing the inhibitor-treated enzymes to their respective DMSO control samples.

4.4.8 Bacterial Fermentation and Natural Product Extraction

Frozen stocks of the associated producing organism RL12-182-HVF-D was streaked onto fresh Marine Broth agar plates (Difco, USA) and incubated at room temperature (~ 25°C) until discrete colonies became visible. Individual colonies were

130 inoculated into 7 mL (small-scale) of modified saline SYP (mSYP) media (10 g starch, 4 g peptone, 2 g yeast extract and 31.2 g instant ocean in 1 L of distilled water) or GNZ media (20 g starch, 10 g glucose, 5 g NZ-amine, 1 g CaCO3, 5 g yeast extract in 1 L of distilled water). Bacterial fermentation was stepped up in stages by inoculating 3 mL of the 7 mL liquid mSYP or GNZ culture into 60 mL (medium-scale) of mSYP or GNZ, respectively. This was followed by inoculating 30 mL of the medium scale culture into 1

L (large-scale) in respective media with 20 g of pre-washed XAD-7 resin (CH2Cl2, MeOH and water). Small-scale cultures were incubated for four days, medium-scale cultures for four days and large-scale cultures for seven days, at ~ 25 °C and shaken at 200 RPM.

Large-scale cultures were extracted by first filtering the cellular/resin slurry under vacuum through two layers of Whatman filter paper. The cells, resin and filter paper were extracted twice with 500 mL of 1:1 CH2Cl2:MeOH and the suspension stirred for 1 hour.

Combined organic extracts were filtered and concentrated to dryness in vacuo. Dried crude extracts for each media (mSYP and GNZ) were fractionated individually by manual solid phase extraction chromatography using Sep-Pak (SP) columns (5 g C18 cartridge, Supelco,

USA). Chromatography proceeded using a stepwise MeOH/H2¬O gradient: 40 mL of 10%

MeOH (wash), 20% (fraction A), 40% (fraction B), 60% (fraction C), 80% (fraction D),

100% (fraction E) then 100% EtOAc wash (fraction F). SP fractions A – F were concentrated to dryness in vacuo and prepared for HPLC-UV separation.

131

4.4.9 HPLC-UV-MS and UPLC-ESI-qTOF-MS Analyses

All high-performance liquid chromatography analyses were performed on an

Agilent 1200 series HPLC system equipped with both an Agilent photodiode array (PDA) detector and an Agilent 6130 single quadrupole mass spectrometer to acquire UV and MS data respectively. Samples were injected onto a C18 reverse-phase column (Synergi 10 µm

Fusion RP Column, Phenomenex, USA) using a H2O:MeOH (0.02% formic acid) elution profile: 0 – 3 mins, 5% MeOH; 3 – 25 min, linear gradient 5% to 100% MeOH; 25 – 30 min, isocratic at 100% MeOH; 30 – 35 min, isocratic at 5% MeOH, using a flow rate of 2 mL/min and positive ESI mode.

Accurate mass MS/MS data were acquired on an Acquity i-Class UPLC system

(Waters Corporation) with SYNAPT G2-Si qTOF mass spectrometer (Waters Corporation) run in HRMS positive ESI mode. The instrument was operated using a 20 µg/mL leucine enkephalin lockspray infusion injected every 10 seconds to control mass accuracy. Samples are injected onto a C18 reverse-phase column (HSS C18, 100 mm x 2.1 mm, 1.7µm, Waters

Corporation) using a H2O (0.1% formic acid):ACN (0.1% formic acid) elution profile: 0 –

0.3 min, 5% ACN; 0.3 – 4.7 min, linear gradient 5% to 90% ACN; 4.7 – 5.5 min, linear gradient 90 to 98% ACN; 5.5 – 5.8 min, isocratic at 98% ACN; 5.81 – 7.5 min, isocratic at

5% ACN, using a flow rate of 0.5 mL/min.

132

Chapter 5. Discussion and Future Outlooks

While the characterization of aaRS proofreading activities has been extensively explored (reviewed in [15]), these efforts were mostly restricted to biochemical and biophysical analyses. Through the continued expansion and accessibility of modern genetic approaches, the biological role of these enzymes has only been recently investigated

[21,28,29,112]. Insights gained from studying perturbations to translational fidelity in vivo have highlighted the complexity and severity of different translational errors. Of the aaRS quality control systems that have been explored to date, we have learned that some translational errors are at least partly tolerated by the proteome as evidenced by studies exploring the role of ThrRS proofreading [32,33]. In contrast, studies from bacteria to mice have illustrated the severity of Ala to Ser substitutions that can arise upon perturbation to

AlaRS quality control. In E. coli, abolishing proofreading by creating an AlaRS C666A variant leads to a slow growth phenotype as well as a global dysregulation of the proteome likely through activation of numerous cellular stress responses [113]. Furthermore, attempts to generate a homologous homozygous variant in mice (AlaRS C723A) led to embryonic lethality [60].

As our understanding of the physiological role of translational quality control has continued to expand, several under-investigated topics have emerged and we believe they are requiring of future investigation. Discussed below are general themes which, to date, have been under-served in relation to translational quality control.

133

5.1 Regulation of post-translational modification of aaRS

Many of the studies described herein have relied upon the utilization of genetic manipulation of model systems to investigate the role of translational quality control. While these studies have been fruitful in their goals of understanding what can happen when aaRS proofreading is perturbed, they provide little insight into physiologically relevant changes in translational quality control. In contrast to the genetic approaches that have been described, several other recent pieces of evidence have emerged in the field highlighting how different environmental stressors can modulate aaRS activity.

It has been observed that upon oxidative stress aaRS activity can be altered, and interestingly, the outcome from these changes can be speculated to provide aaRS-specific fitness advantages in maintaining protein synthesis [33]. Previous studies have described how oxidative stress will modify the endogenous amino acid pool, with one example of this modification being the shift from phenylalanine to various tyrosine isomers [30].

PheRS utilizes a post-transfer proofreading mechanism to prevent the accumulation of Tyr- tRNAPhe [25]. Together, these observations led to the inquiry of whether PheRS activity is modulated during oxidative stress to account for the change in cognate to non-cognate amino acid substrate pools. Through extensive biochemical investigation, it was observed that PheRS oxidation leads to an increase in PheRS fidelity and this increase could even be observed by the wild-type enzyme which is already proofreading-proficient [94]. This increase in proofreading activity provides a clear fitness advantage to account for the simultaneous change in the amino acid pool which would otherwise lead to translational errors.

In contrast to the observed elevated fidelity in PheRS during oxidative stress, it has been noted that ThrRS fidelity is diminished during oxidative stress. Similar to the critical 134 cysteine in the AlaRS proofreading domain, ThrRS also utilizes a cysteine residue in its proofreading active site to help coordinate the 3’ end of mis-serylated tRNAThr for hydrolysis [31]. Upon oxidation of the E. coli ThrRS enzyme, it was determined by mass spectrometry that C182 (homologous to C666 in E. coli AlaRS) becomes oxidized.

Furthermore, oxidation of this residue abolished the proofreading activity of ThrRS [33].

As noted from several previous studies, ThrRS fidelity appears to be at least partly dispensable as E. coli mutants encoding a ThrRS C182A allele have no obvious defects in growth, even upon non-cognate amino acid stress [32,113]. These observations suggest that under oxidative stress, proteome heterogeneity of low-cost amino acid substitutions may be beneficial under a given environmental condition.

While transcriptional changes during oxidative stress have been well documented

(reviewed in [186]), much less is known about how oxidation of the translational machinery itself will lead to changes in the physiological outcome during this environmental stress [187]. As oxidative stress has been shown to alter the enzyme activity of aaRS [33,94], a comprehensive analysis of all twenty aaRSs during oxidation will prove informative. These analyses will allow for monitoring not only of transcriptional and translational changes during oxidation, but also allow for the determination of the propensity for translational accuracy for a given amino acid during cellular oxidation.

In addition to modulation of aaRS activity in response to environmental stressors, it has been shown that post-translational modification of aaRSs can also alter aaRS activity.

Post-translational lysine acetylation in E. coli has been shown to decrease the canonical activity of Class I aaRSs, TyrRS [188], CysRS [71], and HisRS [71] and on the Class II aaRSs, AlaRS [189] and ThrRS [71]. Acetylation of K73 in AlaRS led to a decrease in

135 cognate activity, similar to those observed from the Class I aaRSs [189]. Interestingly, acetylation of K169 on ThrRS led to a decrease in ThrRS fidelity, resulting in the accumulation of Ser-tRNAThr in vitro [71]. While the authors of this work were unable to identify the specific mechanism for this modification, this work highlights the possibility of fine tuning aaRS activity and accuracy through regulated post-translational modifications. In the works described above, the authors searched for changes in aaRS activity at acetylation sites that were determined from global acetylome databases

[190,191]. One limitation of these studies was the characterizations of the modified residues were primarily prioritized to regions of cognate activity (e.g. KMSKS motifs in

Class I aaRSs). Acetylome searches indicate several additional modifications in aaRS regions that potentially implicate changes in quality control regulation and require further investigation.

Beyond post-translational acetylation there is evidence, particularly in higher eukaryotes for aaRS phosphorylation. It was recently shown that in eukaryotic cell culture experiments, phosphorylation of LeuRS at S391 and S720 (located within the conserved

KSMKS motif) during glucose starvation leads to a decrease in leucine binding [192]. A decrease in cognate activation ultimately leads to a decrease in protein synthesis which rapidly allows for adaption to nutrient availability [192]. This is just one example of the intricacies and regulatory potential for aaRS post-translational modification. Deeper bioinformatics analysis of pre-existing modification databases coupled with site-specific biochemical investigation is sure to provide a wealth of novel enzymatic and metabolic potential.

136

5.2 Broader roles for aaRS quality control

While a primary goal of this work was to highlight the impact of substitution- specific quality control defects, very little effort has been expended searching for non- protein consequences for translational accuracy. The importance for AlaRS quality control has been described from bacteria to mice. Interestingly, in bacteria, AlaRS serves a secondary function in regulating non-stop or non-go mRNA transcripts during translation

(reviewed in [193]). To alleviate a non-stop transcript, a unique 363 bp tmRNA (transfer- messenger RNA) will enter the A-site of the ribosome. The 3’end of this RNA mimics that of an aminoacyl-tRNA [194]. Upon entrance into the A-site, the aminoacylated tmRNA is added to the growing peptide in the ribosomal P-site [195]. Upon peptide bond formation, the tmRNA will translocate in the ribosome changing the previous erroneous mRNA to a new open reading frame (ORF) encoded in the tmRNA [196]. The new ORF adds 10 additional amino acids encoding the ClpXP protease tag, finally ending in a stop codon

[195]. Upon termination of the new ORF, the protein product will be targeted by the ClpXP protease for degradation [197].

Due to the relatively large size of tmRNA compared to a standard proteinogenic tRNA and the absence of a canonical anticodon loop, aaRS:tmRNA interactions are restricted. Because AlaRS only requires one identity element in the acceptor stem for tRNA discrimination, AlaRS has evolved as the only aaRS pairing partner for tmRNA aminoacylation [198,199]. This aminoacylation event is facilitated by AlaRS recognition of the tmRNA by the conserved G3:U70 base pair found in all tRNAAla isoacceptors [53].

Upon Ala-tmRNA aminoacylation by AlaRS, the aminoacyl-tmRNA is bound by activated

EF-Tu and the tmRNA binding partner SmpB [200,201]. This tripartite complex is now free to be recruited for ribosome recycling. Beyond the initial role of AlaRS for 137 aminoacylating tmRNA, once the tmRNA-encoding ORF is translocated to the A-site of the ribosome, the remaining ClpXP degradation tag also requires accurate AlaRS-mediated aminoacyl-tRNA formation. The canonical 11 amino acid ClpXP degradation tag is

AANDENYALAA, which implies that AlaRS activity and accuracy is required for 5/11 amino acids in the tag [197]. It has been shown that perturbation to the C-terminal A-A doublet abolishes proteolytic degradation by ClpP [202]. It is unclear if mis- aminoacylation of tmRNA may also disrupt ribosome rescue.

As a preliminary experiment, we purified both wild-type and proofreading- deficient AlaRS variants and wanted to determine if tmRNA could be mis-serylated by

AlaRS. Using in vitro transcribed tmRNA and mis-serylation experiments described in

Chapter 3, we determined that in the absence of AlaRS proofreading, tmRNAs were susceptible to mis-aminoacylation (Fig. 27A). Subsequently, tmRNAs were mis- aminoacylated with serine and Ser-tmRNA substrates were used in proofreading assays with wild-type and AlaRS C666A variants. These experiments showed that not only is tmRNA a substrate for mis-aminoacylation but if Ser-tmRNA is formed, AlaRS proofreading is required for trans proofreading (Fig. 27B). Furthermore, the addition of recombinant EF-Tu and SmpB did not prevent mis-serylation (Fig. 27C) or Ser-tmRNA deacylation (Fig. 27D) in the absence of AlaRS proofreading. Together, these experiments indicate that AlaRS proofreading is required for accurate tmRNA aminoacylation. The consequence of mis-serylated tmRNA on ribosome rescue has still yet to be determined but warrants future investigation.

138

139

Figure 27. AlaRS fidelity is required for accurate tmRNA aminoacylation In the absence of AlaRS proofreading, A) Ser-tmRNA product forms and is B) not subject for deacylation. C) SmpB and EF-Tu do not prevent mis-aminoacylation or D) deacylation.

139

Beyond mistranslation affecting the three-dimensional folding of a given protein, another under-investigated consequence of mistranslation is the potential for disruption of regulatory binding sites. While regulatory phosphorylation cascades are minimal in bacteria [203], in higher eukaryotes these post-translational modifications are essential for activation or suppression of various gene networks (reviewed in [204]). The most common classes of kinases are the serine, threonine, or tyrosine kinases. These enzymes discriminate discrete differences in kinase binding sites and add phosphoryl groups at the terminal –OH group present on these amino acid functional groups [205]. Many of the characterized Class

II aaRS proofreading mechanisms play a role in preventing the erroneous incorporation of serine or tyrosine into the proteome mediated by AlaRS/ThrRS or PheRS, respectively.

There is suggestive evidence that the role of tyrosine and TyrRS has changed over evolution, with the primary role of tyrosine in prokaryotes being used for functional changes in protein structure and regulatory roles being increasingly important in eukaryotes

[206,207]. It has yet to be explored if the generation or removal of kinase binding sites mediated by errors in aaRS-fidelity will play a consequential role in eukaryotic gene regulation.

5.3 Targeting aaRS fidelity for therapeutics

As efficient aminoacyl-tRNA formation is required for translation and inhibiting any one of the aaRS in a cell required for protein synthesis will lead to cellular inactivation, aaRS have been a promising target for antimicrobial discovery (reviewed in [131]). Several

140 aaRS inhibitors are currently on the market for either medicinal or veterinary applications, including the IleRS inhibitor mupirocin [136] and the ProRS inhibitor halofuginone [137].

One of key discoveries towards aaRS-targeting therapeutics was the antifungal

LeuRS inhibitor AN2690 (later renamed Tavaborole). Through a series of biochemical and biophysical approaches it was shown that Tavaborole, a benzoxaborole, binds to the catalytic active site of the Thermus thermophilis LeuRS editing domain [90]. The addition of Tavaborole prevented LeuRS proofreading when assayed using in vitro deacylation experiments with Ile-tRNALeu and Saccharomyces cerevisiae LeuRS [90]. Together, these results led to Tavaborole’s original description as an “editing inhibitor,” however this label was a misnomer. Because treatment of this drug led to complete enzyme inactivation by preventing release of tRNA bound in the editing active site, Tavaborole was ultimately preventing iterative rounds of enzyme catalysis, not the accumulation of mis-acylated tRNAs to be used in translation. Beyond Tavaborole’s mis-identification as a proofreading inhibitor, subsequent characterization in vivo highlighted the rapid rate of antifungal resistance to Tavaborole [208]. The propensity for LeuRS resistance mechanisms to arise upon Tavaborole treatment has limited this drug’s clinical application although it is still used as a topical treatment for onychomycosis. Despite Tavaborole’s mechanism of binding to the LeuRS proofreading active site, this discovery has led many to search for new possible inhibitors of aaRS fidelity.

As has been discussed throughout this document, one limitation of the field’s progress in characterizing translational errors is the lack in understanding of the cost of specific translational errors. There are several bodies of evidence which note the correlation

141 with antimicrobial resistance and translational errors (reviewed in [209]). The most notable example of this phenomena is the connection between rifampicin phenotypic resistance and Mycobaterium tuberculosis translation fidelity. Mycobacteria species utilize the amidotransferase complex GatCAB to convert Glu-tRNAGln to Gln-tRNAGln and Asp- tRNAAsn to Asn-tRNAAsn (reviewed in [210]). Mutations in gatA (gene encoding GatA from the amidotransferase complex) were identified from clinical isolates and indicated that many of these phenotypically resistant mutants had elevated levels of mistranslation

[211]. This observation has led many to conclude that targeting translational fidelity for therapeutic discovery is unlikely to produce novel, resistance-proof therapies [212]. We disagree with this sentiment and propose that a mistranslational error-target based approach is likely to provide new biological targets. As has been previously discussed, certain translational errors appear to be well tolerated (e.g. Ser to Thr mistranslation mediated by

ThrRS [32,113]). Furthermore, the incidence of degenerate aaRS proofreading domains of

LeuRS, ThrRS, and PheRS enzymes in various Mycoplasma species indicate the fidelity of these enzymes is not critical for parasite survival [37]. With these additional considerations, we believe that targeting high-cost translational errors are likely to identify practical therapeutic targets. We have shown that defects in AlaRS fidelity lead to gross perturbation of the E. coli proteome and growth defects in several environmental conditions including exposure to a wide array of antimicrobials [113]. Furthermore, defects in AlaRS fidelity led to an increase in AlaRS protein levels [113]. We propose a model in which targeting AlaRS fidelity as a combination therapy could decrease the dosage of existing bactericidal treatments due to the elevated sensitivity towards antimicrobials. Additionally,

142 as decreases in AlaRS fidelity led to elevated AlaRS protein levels, more antimicrobial targets will be generated creating an auto-regulatory mechanism for antimicrobial chemosensitization (Fig. 28).

143

144

Figure 28. Proposed model for targeting AlaRS fidelity for chemosensitization Targeting AlaRS fidelity has combination therapeutic potential. Defects in AlaRS proofreading led to mistranslation, antimicrobial sensitivity, and elevated AlaRS protein level making it a good target for combination therapies. Increasing antimicrobial sensitivity can lead to reduced prescribed antimicrobial dosages. Furthermore, the increase in AlaRS levels will allow for subsequent proofreading inhibitor binding.

144

5.4 Conclusion

The work described herein has provided novel insight into the exploration of substitution-specific costs of mistranslation. For the first time, the cost of AlaRS-mediated translational errors has been explored in bacteria and provides insight into the potential cost of these errors in higher eukaryotic systems. Additionally, through the combined use of genetic and biochemical approaches, a previously unexplored functional region within

E. coli AlaRS has been identified. This workflow has highlighted the utility of genetic- based methodologies to understanding aaRS biology which has historically been confined to biophysical structure-function associations.

Given the essentially of aaRSs, they have long-been promising targets for antimicrobial discovery. While some aaRS-targeting inhibitors have made it to the clinic, an exhaustive search for novel biological targets has yet to be performed. Using several recently defined computational approaches to search for tRNA identity elements across phylogenetic clades, we have identified new biological targets for anti-trypanosomal drug discovery. The development of therapies against eukaryotic pathogens has been historically challenging given the high evolutionary conservation of drug targets between host and pathogen. Through iterative rounds of biochemical screening, we have identified several parasite aaRSs that were susceptible to natural product inhibition, including AlaRS, and encouragingly, the compounds had no effects on the counterpart human enzyme.

Altogether, the body of work presented adds to the field’s understanding of one of the most well characterized aaRSs, AlaRS. While its role in maintaining proteome homeostasis through AlaRS fidelity and its potential for therapeutic discovery has been

145 elucidated, much remains to still be investigated regarding this complex enzyme. However, the data detailed in this work will provide the groundwork for future discovery.

146

Bibliography

1. Lee H, Popodi E, Tang H, Foster PL (2012) Rate and molecular spectrum of spontaneous mutations in the bacterium Escherichia coli as determined by whole- genome sequencing. Proc Natl Acad Sci U S A 109: E2774-2783. 2. Traverse CC, Ochman H (2016) Conserved rates and patterns of transcription errors across bacterial growth states and lifestyles. Proc Natl Acad Sci U S A 113: 3311- 3316. 3. Kramer EB, Farabaugh PJ (2007) The frequency of translational misreading errors in E. coli is largely determined by tRNA competition. RNA 13: 87-96. 4. Mordret E, Dahan O, Asraf O, Rak R, Yehonadav A, et al. (2019) Systematic Detection of Amino Acid Substitutions in Proteomes Reveals Mechanistic Basis of Ribosome Errors and Selection for Translation Fidelity. Mol Cell 75: 427-441 e425. 5. Ibba M, Soll D (2000) Aminoacyl-tRNA synthesis. Annu Rev Biochem 69: 617-650. 6. Burbaum JJ, Schimmel P (1991) Structural relationships and the classification of aminoacyl-tRNA synthetases. J Biol Chem 266: 16965-16968. 7. Burbaum JJ, Starzyk RM, Schimmel P (1990) Understanding structural relationships in proteins of unsolved three-dimensional structure. Proteins 7: 99-111. 8. Schmitt E, Meinnel T, Blanquet S, Mechulam Y (1994) Methionyl-tRNA synthetase needs an intact and mobile 332KMSKS336 motif in catalysis of methionyl adenylate formation. J Mol Biol 242: 566-576. 9. Sprinzl M, Cramer F (1975) Site of aminoacylation of tRNAs from Escherichia coli with respect to the 2'- or 3'-hydroxyl group of the terminal adenosine. Proc Natl Acad Sci U S A 72: 3049-3053. 10. Eriani G, Delarue M, Poch O, Gangloff J, Moras D (1990) Partition of tRNA synthetases into two classes based on mutually exclusive sets of sequence motifs. Nature 347: 203-206. 11. Cusack S, Hartlein M, Leberman R (1991) Sequence, structural and evolutionary relationships between class 2 aminoacyl-tRNA synthetases. Nucleic Acids Res 19: 3489-3498. 12. Fersht A (1985) Enzyme structure and function. WH Freeman and. 13. Loftfield RB, Eigner EA (1966) The specificity of enzymic reactions. Aminoacyl- soluble RNA ligases. Biochim Biophys Acta 130: 426-448. 14. Fersht AR (1981) Enzymic editing mechanisms and the genetic code. Proc R Soc Lond B Biol Sci 212: 351-379.

147

15. Yadavalli SS, Ibba M (2012) Quality control in aminoacyl-tRNA synthesis its role in translational fidelity. Adv Protein Chem Struct Biol 86: 1-43. 16. Fersht AR, Kaethner MM (1976) Enzyme hyperspecificity. Rejection of threonine by the valyl-tRNA synthetase by misacylation and hydrolytic editing. Biochemistry 15: 3342-3346. 17. Fersht AR, Dingwall C (1979) Evidence for the double-sieve editing mechanism in protein synthesis. Steric exclusion of isoleucine by valyl-tRNA synthetases. Biochemistry 18: 2627-2631. 18. Englisch S, Englisch U, von der Haar F, Cramer F (1986) The proofreading of hydroxy analogues of leucine and isoleucine by leucyl-tRNA synthetases from E. coli and yeast. Nucleic Acids Res 14: 7529-7539. 19. Baldwin AN, Berg P (1966) Transfer ribonucleic acid-induced hydrolysis of valyladenylate bound to isoleucyl ribonucleic acid synthetase. J Biol Chem 241: 839-845. 20. Cvetesic N, Perona JJ, Gruic-Sovulj I (2012) Kinetic partitioning between synthetic and editing pathways in class I aminoacyl-tRNA synthetases occurs at both pre- transfer and post-transfer hydrolytic steps. J Biol Chem 287: 25381-25394. 21. Cvetesic N, Palencia A, Halasz I, Cusack S, Gruic-Sovulj I (2014) The physiological target for LeuRS translational quality control is norvaline. EMBO J 33: 1639- 1653. 22. Bilus M, Semanjski M, Mocibob M, Zivkovic I, Cvetesic N, et al. (2019) On the Mechanism and Origin of Isoleucyl-tRNA Synthetase Editing against Norvaline. J Mol Biol 431: 1284-1297. 23. Cvetesic N, Semanjski M, Soufi B, Krug K, Gruic-Sovulj I, et al. (2016) Proteome- wide measurement of non-canonical bacterial mistranslation by quantitative mass spectrometry of protein modifications. Sci Rep 6: 28631. 24. Ji QQ, Fang ZP, Ye Q, Chi CW, Wang ED (2017) Self-protective responses to norvaline-induced stress in a leucyl-tRNA synthetase editing-deficient yeast strain. Nucleic Acids Res 45: 7367-7381. 25. Roy H, Ling J, Irnov M, Ibba M (2004) Post-transfer editing in vitro and in vivo by the beta subunit of phenylalanyl-tRNA synthetase. EMBO J 23: 4639-4648. 26. Reynolds NM, Ling J, Roy H, Banerjee R, Repasky SE, et al. (2010) Cell-specific differences in the requirements for translation quality control. Proc Natl Acad Sci U S A 107: 4063-4068. 27. Bertin C, Weston LA, Huang T, Jander G, Owens T, et al. (2007) Grass roots chemistry: meta-tyrosine, an herbicidal nonprotein amino acid. Proc Natl Acad Sci U S A 104: 16964-16969. 28. Bullwinkle TJ, Reynolds NM, Raina M, Moghal A, Matsa E, et al. (2014) Oxidation of cellular amino acid pools leads to cytotoxic mistranslation of the genetic code. Elife 3. 29. Bullwinkle TJ, Ibba M (2016) Translation quality control is critical for bacterial responses to amino acid stress. Proc Natl Acad Sci U S A 113: 2252-2257.

148

30. Bullwinkle T, Lazazzera B, Ibba M (2014) Quality control and infiltration of translation by amino acids outside of the genetic code. Annu Rev Genet 48: 149- 166. 31. Dock-Bregeon AC, Rees B, Torres-Larios A, Bey G, Caillet J, et al. (2004) Achieving error-free translation; the mechanism of proofreading of threonyl- tRNA synthetase at atomic resolution. Mol Cell 16: 375-386. 32. Mohler K, Aerni HR, Gassaway B, Ling J, Ibba M, et al. (2017) MS-READ: Quantitative measurement of amino acid incorporation. Biochim Biophys Acta Gen Subj 1861: 3081-3088. 33. Ling J, Soll D (2010) Severe oxidative stress induces protein mistranslation through impairment of an aminoacyl-tRNA synthetase editing site. Proc Natl Acad Sci U S A 107: 4028-4033. 34. Mohler K, Ibba M (2017) Translational fidelity and mistranslation in the cellular response to stress. Nat Microbiol 2: 17117. 35. Melnikov SV, Rivera KD, Ostapenko D, Makarenko A, Sanscrainte ND, et al. (2018) Error-prone protein synthesis in parasites with the smallest eukaryotic genome. Proc Natl Acad Sci U S A 115: E6245-E6253. 36. Melnikov SV, van den Elzen A, Stevens DL, Thoreen CC, Soll D (2018) Loss of protein synthesis quality control in host-restricted organisms. Proc Natl Acad Sci U S A 115: E11505-E11512. 37. Li L, Boniecki MT, Jaffe JD, Imai BS, Yau PM, et al. (2011) Naturally occurring aminoacyl-tRNA synthetases editing-domain mutations that cause mistranslation in Mycoplasma parasites. Proc Natl Acad Sci U S A 108: 9378-9383. 38. An S, Musier-Forsyth K (2004) Trans-editing of Cys-tRNAPro by Haemophilus influenzae YbaK protein. J Biol Chem 279: 42359-42362. 39. Beuning PJ, Musier-Forsyth K (2000) Hydrolytic editing by a class II aminoacyl- tRNA synthetase. Proc Natl Acad Sci U S A 97: 8916-8920. 40. So BR, An S, Kumar S, Das M, Turner DA, et al. (2011) Substrate-mediated fidelity mechanism ensures accurate decoding of proline codons. J Biol Chem 286: 31810-31820. 41. Vargas-Rodriguez O, Musier-Forsyth K (2013) Exclusive use of trans-editing domains prevents proline mistranslation. J Biol Chem 288: 14391-14399. 42. Putney SD, Sauer RT, Schimmel PR (1981) Purification and properties of alanine tRNA synthetase from Escherichia coli A tetramer of identical subunits. J Biol Chem 256: 198-204. 43. Jasin M, Regan L, Schimmel P (1984) Dispensable pieces of an aminoacyl tRNA synthetase which activate the catalytic site. Cell 36: 1089-1095. 44. Ribas de Pouplana L, Schimmel P (1997) Reconstruction of quaternary structures of class II tRNA synthetases by rational mutagenensis of a conserved domain. Biochemistry 36: 15041-15048. 45. Sood SM, Slattery CW, Filley SJ, Wu MX, Hill KA (1996) Further characterization of Escherichia coli alanyl-tRNA synthetase. Arch Biochem Biophys 328: 295- 301.

149

46. Dignam JD, Guo J, Griffith WP, Garbett NC, Holloway A, et al. (2011) Allosteric interaction of nucleotides and tRNA(ala) with E. coli alanyl-tRNA synthetase. Biochemistry 50: 9886-9900. 47. Banerjee B, Ganguli S, Banerjee R (2019) Biophysical characterization of interaction between E. coli alanyl-tRNA synethase with its promoter DNA. Protein Pept Lett. 48. Putney SD, Schimmel P (1981) An aminoacyl tRNA synthetase binds to a specific DNA sequence and regulates its gene transcription. Nature 291: 632-635. 49. Springer M, Trudel M, Graffe M, Plumbridge J, Fayat G, et al. (1983) Escherichia coli phenylalanyl-tRNA synthetase operon is controlled by attenuation in vivo. J Mol Biol 171: 263-279. 50. Davis MW, Buechter DD, Schimmel P (1994) Functional dissection of a predicted class-defining motif in a class II tRNA synthetase of unknown structure. Biochemistry 33: 9904-9911. 51. Shi JP, Musier-Forsyth K, Schimmel P (1994) Region of a conserved sequence motif in a class II tRNA synthetase needed for transfer of an activated amino acid to an RNA substrate. Biochemistry 33: 5312-5318. 52. Guo M, Chong YE, Shapiro R, Beebe K, Yang XL, et al. (2009) Paradox of mistranslation of serine for alanine caused by AlaRS recognition dilemma. Nature 462: 808-812. 53. Hou YM, Schimmel P (1988) A simple structural feature is a major determinant of the identity of a transfer RNA. Nature 333: 140-145. 54. McClain WH, Foss K (1988) Changing the identity of a tRNA by introducing a G-U wobble pair near the 3' acceptor end. Science 240: 793-796. 55. Chong YE, Guo M, Yang XL, Kuhle B, Naganuma M, et al. (2018) Distinct ways of G:U recognition by conserved tRNA binding motifs. Proc Natl Acad Sci U S A 115: 7527-7532. 56. Tsui WC, Fersht AR (1981) Probing the principles of amino acid selection using the alanyl-tRNA synthetase from Escherichia coli. Nucleic Acids Res 9: 4627-4637. 57. Beebe K, Ribas De Pouplana L, Schimmel P (2003) Elucidation of tRNA-dependent editing by a class II tRNA synthetase and significance for cell viability. The EMBO journal 22: 668-675. 58. Pasman Z, Robey-Bond S, Mirando AC, Smith GJ, Lague A, et al. (2011) Substrate specificity and catalysis by the editing active site of Alanyl-tRNA synthetase from Escherichia coli. Biochemistry 50: 1474-1482. 59. Lee JW, Beebe K, Nangle LA, Jang J, Longo-Guess CM, et al. (2006) Editing- defective tRNA synthetase causes protein misfolding and neurodegeneration. Nature 443: 50-55. 60. Liu Y, Satz JS, Vo MN, Nangle LA, Schimmel P, et al. (2014) Deficiencies in tRNA synthetase editing activity cause cardioproteinopathy. Proc Natl Acad Sci U S A 111: 17570-17575. 61. Hilander T, Zhou XL, Konovalova S, Zhang FP, Euro L, et al. (2018) Editing activity for eliminating mischarged tRNAs is essential in mammalian mitochondria. Nucleic Acids Res 46: 849-860.

150

62. Hidalgo San Jose L, Sunshine MJ, Dillingham CH, Chua BA, Kruta M, et al. (2020) Modest Declines in Proteome Quality Impair Hematopoietic Stem Cell Self- Renewal. Cell Rep 30: 69-80 e66. 63. Chong YE, Yang XL, Schimmel P (2008) Natural homolog of tRNA synthetase editing domain rescues conditional lethality caused by mistranslation. J Biol Chem 283: 30073-30078. 64. Nawaz MH, Merriman E, Yang XL, Schimmel P (2011) p23H implicated as cis/trans regulator of AlaXp-directed editing for mammalian cell homeostasis. Proc Natl Acad Sci U S A 108: 2723-2728. 65. Vo MN, Terrey M, Lee JW, Roy B, Moresco JJ, et al. (2018) ANKRD16 prevents neuron loss caused by an editing-defective tRNA synthetase. Nature 557: 510- 515. 66. Pawar KI, Suma K, Seenivasan A, Kuncha SK, Routh SB, et al. (2017) Role of D- aminoacyl-tRNA deacylase beyond chiral proofreading as a cellular defense against glycine mischarging by AlaRS. Elife 6. 67. Fersht AR (1977) Editing mechanisms in protein synthesis. Rejection of valine by the isoleucyl-tRNA synthetase. Biochemistry 16: 1025-1030. 68. Dulic M, Cvetesic N, Zivkovic I, Palencia A, Cusack S, et al. (2018) Kinetic Origin of Substrate Specificity in Post-Transfer Editing by Leucyl-tRNA Synthetase. J Mol Biol 430: 1-16. 69. Jones TE, Alexander RW, Pan T (2011) Misacylation of specific nonmethionyl tRNAs by a bacterial methionyl-tRNA synthetase. Proc Natl Acad Sci U S A 108: 6933-6938. 70. Netzer N, Goodenbour JM, David A, Dittmar KA, Jones RB, et al. (2009) Innate immune and chemically triggered oxidative stress modifies translational fidelity. Nature 462: 522-526. 71. Chen H, Venkat S, Hudson D, Wang T, Gan Q, et al. (2019) Site-Specifically Studying Lysine Acetylation of Aminoacyl-tRNA Synthetases. ACS Chem Biol 14: 288-295. 72. Cummings HS, Sands JF, Fraser J, Hershey JW (1994) Characterization and expression of a gene encoding serine tRNA5 from Escherichia coli. Biochimie 76: 83-87. 73. Dong H, Nilsson L, Kurland CG (1996) Co-variation of tRNA abundance and codon usage in Escherichia coli at different growth rates. J Mol Biol 260: 649-663. 74. Avcilar-Kucukgoze I, Bartholomaus A, Cordero Varela JA, Kaml RF, Neubauer P, et al. (2016) Discharging tRNAs: a tug of war between translation and detoxification in Escherichia coli. Nucleic Acids Res 44: 8324-8334. 75. Santos M, Pereira PM, Varanda AS, Carvalho J, Azevedo M, et al. (2018) Codon misreading tRNAs promote tumor growth in mice. RNA Biol 15: 773-786. 76. Ruan B, Palioura S, Sabina J, Marvin-Guy L, Kochhar S, et al. (2008) Quality control despite mistranslation caused by an ambiguous genetic code. Proc Natl Acad Sci U S A 105: 16502-16507.

151

77. Dalbadie-McFarland G, Neitzel JJ, Richards JH (1986) Active-site mutants of beta- lactamase: use of an inactive double mutant to study requirements for catalysis. Biochemistry 25: 332-338. 78. Evans CR, Fan Y, Ling J (2019) Increased mistranslation protects E. coli from protein misfolding stress due to activation of a RpoS-dependent heat shock response. FEBS Lett. 79. Doyle SM, Shastry S, Kravats AN, Shih YH, Miot M, et al. (2015) Interplay between E. coli DnaK, ClpB and GrpE during protein disaggregation. J Mol Biol 427: 312- 327. 80. Fan Y, Evans CR, Ling J (2016) Reduced Protein Synthesis Fidelity Inhibits Flagellar Biosynthesis and Motility. Sci Rep 6: 30960. 81. De Lay N, Gottesman S (2012) A complex network of small non-coding RNAs regulate motility in Escherichia coli. Mol Microbiol 86: 524-538. 82. Soutourina OA, Bertin PN (2003) Regulation cascade of flagellar expression in Gram-negative bacteria. FEMS Microbiol Rev 27: 505-523. 83. Bury-Mone S, Nomane Y, Reymond N, Barbet R, Jacquet E, et al. (2009) Global analysis of extracytoplasmic stress signaling in Escherichia coli. PLoS Genet 5: e1000651. 84. Davis BD, Chen LL, Tai PC (1986) Misread protein creates membrane channels: an essential step in the bactericidal action of aminoglycosides. Proc Natl Acad Sci U S A 83: 6164-6168. 85. Nikaido H, Rosenberg EY, Foulds J (1983) Porin channels in Escherichia coli: studies with beta-lactams in intact cells. J Bacteriol 153: 232-240. 86. Mortimer PG, Piddock LJ (1993) The accumulation of five antibacterial agents in porin-deficient mutants of Escherichia coli. J Antimicrob Chemother 32: 195-213. 87. Libby EA, Roggiani M, Goulian M (2012) Membrane protein expression triggers chromosomal locus repositioning in bacteria. Proc Natl Acad Sci U S A 109: 7445-7450. 88. Bakshi S, Choi H, Mondal J, Weisshaar JC (2014) Time-dependent effects of transcription- and translation-halting drugs on the spatial distributions of the Escherichia coli chromosome and ribosomes. Mol Microbiol 94: 871-887. 89. Liu Z, Vargas-Rodriguez O, Goto Y, Novoa EM, Ribas de Pouplana L, et al. (2015) Homologous trans-editing factors with broad tRNA specificity prevent mistranslation caused by serine/threonine misactivation. Proc Natl Acad Sci U S A 112: 6027-6032. 90. Rock FL, Mao W, Yaremchuk A, Tukalo M, Crepin T, et al. (2007) An antifungal agent inhibits an aminoacyl-tRNA synthetase by trapping tRNA in the editing site. Science 316: 1759-1761. 91. Javid B, Sorrentino F, Toosky M, Zheng W, Pinkham JT, et al. (2014) Mycobacterial mistranslation is necessary and sufficient for rifampicin phenotypic resistance. Proc Natl Acad Sci U S A 111: 1132-1137. 92. Douafer H, Andrieu V, Phanstiel Ot, Brunel JM (2019) Antibiotic Adjuvants: Make Antibiotics Great Again! J Med Chem.

152

93. Miranda I, Silva-Dias A, Rocha R, Teixeira-Santos R, Coelho C, et al. (2013) Candida albicans CUG mistranslation is a mechanism to create cell surface variation. MBio 4. 94. Steiner RE, Kyle AM, Ibba M (2019) Oxidation of phenylalanyl-tRNA synthetase positively regulates translational quality control. Proc Natl Acad Sci U S A 116: 10058-10063. 95. Zhou XL, Chen Y, Fang ZP, Ruan ZR, Wang Y, et al. (2016) Translational Quality Control by Bacterial Threonyl-tRNA Synthetases. J Biol Chem 291: 21208- 21221. 96. Zhou XL, Ruan ZR, Wang M, Fang ZP, Wang Y, et al. (2014) A minimalist mitochondrial threonyl-tRNA synthetase exhibits tRNA-isoacceptor specificity during proofreading. Nucleic Acids Res 42: 13873-13886. 97. Herring CD, Glasner JD, Blattner FR (2003) Gene replacement without selection: regulated suppression of amber mutations in Escherichia coli. Gene 311: 153-163. 98. Kim J, Webb AM, Kershner JP, Blaskowski S, Copley SD (2014) A versatile and highly efficient method for scarless genome editing in Escherichia coli and Salmonella enterica. BMC Biotechnol 14: 84. 99. Blattner FR, Plunkett G, 3rd, Bloch CA, Perna NT, Burland V, et al. (1997) The complete genome sequence of Escherichia coli K-12. Science 277: 1453-1462. 100. Kalogeraki VS, Winans SC (1997) Suicide plasmids containing promoterless reporter genes can simultaneously disrupt and create fusions to target genes of diverse bacteria. Gene 188: 69-75. 101. Monteilhet C, Perrin A, Thierry A, Colleaux L, Dujon B (1990) Purification and characterization of the in vitro activity of I-Sce I, a novel and highly specific endonuclease encoded by a group I intron. Nucleic Acids Res 18: 1407-1413. 102. Green MR, Sambrook J (2019) Inverse Polymerase Chain Reaction (PCR). Cold Spring Harb Protoc 2019: pdb prot095166. 103. Ferrieres L, Hemery G, Nham T, Guerout AM, Mazel D, et al. (2010) Silent mischief: bacteriophage Mu insertions contaminate products of Escherichia coli random mutagenesis performed using suicidal transposon delivery plasmids mobilized by broad-host-range RP4 conjugative machinery. J Bacteriol 192: 6418-6427. 104. Ruiz N, Silhavy TJ (2003) Constitutive activation of the Escherichia coli Pho regulon upregulates rpoS translation in an Hfq-dependent fashion. J Bacteriol 185: 5984-5992. 105. Sambrook J, Fritsch EF, Maniatis T (1989) Molecular cloning: a laboratory manual: Cold spring harbor laboratory press. 106. Schneider CA, Rasband WS, Eliceiri KW (2012) NIH Image to ImageJ: 25 years of image analysis. Nat Methods 9: 671-675. 107. Pfaffl MW (2001) A new mathematical model for relative quantification in real-time RT-PCR. Nucleic Acids Res 29: e45.

153

108. Wilkinson AJ, Fersht AR, Blow DM, Winter G (1983) Site-directed mutagenesis as a probe of enzyme structure and catalysis: tyrosyl-tRNA synthetase cysteine-35 to glycine-35 mutation. Biochemistry 22: 3581-3586. 109. Walker MA, Mohler KP, Hopkins KW, Oakley DH, Sweetser DA, et al. (2016) Novel Compound Heterozygous Mutations Expand the Recognized Phenotypes of FARS2-Linked Disease. J Child Neurol 31: 1127-1137. 110. Lichtarge O, Bourne HR, Cohen FE (1996) An evolutionary trace method defines binding surfaces common to protein families. J Mol Biol 257: 342-358. 111. Tondnevis F, Dudenhausen EE, Miller AM, McKenna R, Altschul SF, et al. (2020) Deep Analysis of Residue Constraints (DARC): identifying determinants of protein functional specificity. Sci Rep 10: 1691. 112. Mohler K, Mann R, Bullwinkle TJ, Hopkins K, Hwang L, et al. (2017) Editing of misaminoacylated tRNA controls the sensitivity of amino acid stress responses in Saccharomyces cerevisiae. Nucleic Acids Res 45: 3985-3996. 113. Kelly P, Backes N, Mohler K, Buser C, Kavoor A, et al. (2019) Alanyl-tRNA Synthetase Quality Control Prevents Global Dysregulation of the Escherichia coli Proteome. mBio 10. 114. LaRiviere FJ, Wolfson AD, Uhlenbeck OC (2001) Uniform binding of aminoacyl- tRNAs to elongation factor Tu by thermodynamic compensation. Science 294: 165-168. 115. Sanderson LE, Uhlenbeck OC (2007) Directed mutagenesis identifies amino acid residues involved in elongation factor Tu binding to yeast Phe-tRNAPhe. J Mol Biol 368: 119-130. 116. Sanderson LE, Uhlenbeck OC (2007) Exploring the specificity of bacterial elongation factor Tu for different tRNAs. Biochemistry 46: 6194-6200. 117. Guo M, Chong YE, Beebe K, Shapiro R, Yang XL, et al. (2009) The C-Ala domain brings together editing and aminoacylation functions on one tRNA. Science 325: 744-747. 118. Afgan E, Baker D, Batut B, van den Beek M, Bouvier D, et al. (2018) The Galaxy platform for accessible, reproducible and collaborative biomedical analyses: 2018 update. Nucleic Acids Res 46: W537-W544. 119. Madeira F, Park YM, Lee J, Buso N, Gur T, et al. (2019) The EMBL-EBI search and sequence analysis tools APIs in 2019. Nucleic Acids Res 47: W636-W641. 120. Waterhouse AM, Procter JB, Martin DM, Clamp M, Barton GJ (2009) Jalview Version 2--a multiple sequence alignment editor and analysis workbench. Bioinformatics 25: 1189-1191. 121. Kelley LA, Mezulis S, Yates CM, Wass MN, Sternberg MJ (2015) The Phyre2 web portal for protein modeling, prediction and analysis. Nat Protoc 10: 845-858. 122. Reisch CR, Prather KL (2015) The no-SCAR (Scarless Cas9 Assisted Recombineering) system for genome editing in Escherichia coli. Sci Rep 5: 15096.

154

123. Reisch CR, Prather KLJ (2017) Scarless Cas9 Assisted Recombineering (no-SCAR) in Escherichia coli, an Easy-to-Use System for Genome Editing. Curr Protoc Mol Biol 117: 31 38 31-31 38 20. 124. Fairlamb AH, Gow NA, Matthews KR, Waters AP (2016) Drug resistance in eukaryotic microorganisms. Nat Microbiol 1: 16092. 125. World Health O (2010) Control of the leishmaniases. World Health Organ Tech Rep Ser: xii-xiii, 1-186, back cover. 126. Crowther GJ, Shanmugam D, Carmona SJ, Doyle MA, Hertz-Fowler C, et al. (2010) Identification of attractive drug targets in neglected-disease pathogens using an in silico approach. PLoS Negl Trop Dis 4: e804. 127. Polonio T, Efferth T (2008) Leishmaniasis: drug resistance and natural products (review). Int J Mol Med 22: 277-286. 128. Handman E, Kedzierski L, Uboldi AD, Goding JW (2008) Fishing for anti- leishmania drugs: principles and problems. Adv Exp Med Biol 625: 48-60. 129. Singh S, Sivakumar R (2004) Challenges and new discoveries in the treatment of leishmaniasis. J Infect Chemother 10: 307-315. 130. Davis AJ, Kedzierski L (2005) Recent advances in antileishmanial drug development. Curr Opin Investig Drugs 6: 163-169. 131. Ho JM, Bakkalbasi E, Soll D, Miller CA (2018) Drugging tRNA aminoacylation. RNA Biol 15: 667-677. 132. Salazar JC, Ahel I, Orellana O, Tumbula-Hansen D, Krieger R, et al. (2003) Coevolution of an aminoacyl-tRNA synthetase with its tRNA substrates. Proc Natl Acad Sci U S A 100: 13863-13868. 133. Giege R, Sissler M, Florentz C (1998) Universal rules and idiosyncratic features in tRNA identity. Nucleic Acids Res 26: 5017-5035. 134. Ardell DH (2010) Computational analysis of tRNA identity. FEBS Lett 584: 325- 333. 135. Hou YM, Schimmel P (1989) Evidence that a major determinant for the identity of a transfer RNA is conserved in evolution. Biochemistry 28: 6800-6804. 136. Silvian LF, Wang J, Steitz TA (1999) Insights into editing from an ile-tRNA synthetase structure with tRNAile and mupirocin. Science 285: 1074-1077. 137. Zhou H, Sun L, Yang XL, Schimmel P (2013) ATP-directed capture of bioactive herbal-based medicine on human tRNA synthetase. Nature 494: 121-124. 138. Seiradake E, Mao W, Hernandez V, Baker SJ, Plattner JJ, et al. (2009) Crystal structures of the human and fungal cytosolic Leucyl-tRNA synthetase editing domains: A structural basis for the rational design of antifungal benzoxaboroles. J Mol Biol 390: 196-207. 139. Merritt EA, Arakaki TL, Gillespie JR, Larson ET, Kelley A, et al. (2010) Crystal structures of trypanosomal histidyl-tRNA synthetase illuminate differences between eukaryotic and prokaryotic homologs. J Mol Biol 397: 481-494. 140. Larson ET, Kim JE, Zucker FH, Kelley A, Mueller N, et al. (2011) Structure of Leishmania major methionyl-tRNA synthetase in complex with intermediate products methionyladenylate and pyrophosphate. Biochimie 93: 570-582.

155

141. Larson ET, Kim JE, Castaneda LJ, Napuli AJ, Zhang Z, et al. (2011) The double- length tyrosyl-tRNA synthetase from the eukaryote Leishmania major forms an intrinsically asymmetric pseudo-dimer. J Mol Biol 409: 159-176. 142. Huang W, Zhang Z, Ranade RM, Gillespie JR, Barros-Alvarez X, et al. (2017) Optimization of a binding fragment targeting the "enlarged methionine pocket" leads to potent Trypanosoma brucei methionyl-tRNA synthetase inhibitors. Bioorg Med Chem Lett 27: 2702-2707. 143. Collins-Hed AI, Ardell DH (2019) Match fitness landscapes for macromolecular interaction networks: Selection for translational accuracy and rate can displace tRNA-binding interfaces of non-cognate aminoacyl-tRNA synthetases. Theor Popul Biol. 144. Freyhult E, Moulton V, Ardell DH (2006) Visualizing bacterial tRNA identity determinants and antideterminants using function logos and inverse function logos. Nucleic Acids Res 34: 905-916. 145. Schneider TD (2010) A brief review of molecular information theory. Nano Commun Netw 1: 173-180. 146. Freyhult E, Cui Y, Nilsson O, Ardell DH (2007) New computational methods reveal tRNA identity element divergence between Proteobacteria and Cyanobacteria. Biochimie 89: 1276-1288. 147. Fitzgerald JB, Schoeberl B, Nielsen UB, Sorger PK (2006) Systems biology and combination therapy in the quest for clinical efficacy. Nat Chem Biol 2: 458-466. 148. Stumpf MP, Robertson BD, Duncan K, Young DB (2007) Systems biology and its impact on anti-infective drug development. Prog Drug Res 64: 1, 3-20. 149. Russell RB, Aloy P (2008) Targeting and tinkering with interaction networks. Nat Chem Biol 4: 666-673. 150. Berger SI, Iyengar R (2009) Network analyses in systems pharmacology. Bioinformatics 25: 2466-2472. 151. Anderson JB (2005) Evolution of antifungal-drug resistance: mechanisms and pathogen fitness. Nat Rev Microbiol 3: 547-556. 152. Kroiss J, Kaltenpoth M, Schneider B, Schwinger MG, Hertweck C, et al. (2010) Symbiotic Streptomycetes provide antibiotic combination prophylaxis for wasp offspring. Nat Chem Biol 6: 261-263. 153. Jansen G, Lee AY, Epp E, Fredette A, Surprenant J, et al. (2009) Chemogenomic profiling predicts antifungal synergies. Mol Syst Biol 5: 338. 154. Hu Y, Platzer EG, Bellier A, Aroian RV (2010) Discovery of a highly synergistic anthelmintic combination that shows mutual hypersusceptibility. Proc Natl Acad Sci U S A 107: 5955-5960. 155. Organization WH (2015) Guidelines for the treatment of malaria: World Health Organization. 156. Organization WH (2018) Artemisinin resistance and artemisinin-based combination therapy efficacy: status report. World Health Organization.

156

157. Marck C, Grosjean H (2002) tRNomics: analysis of tRNA genes from 50 genomes of Eukarya, Archaea, and Bacteria reveals anticodon-sparing strategies and domain-specific features. RNA 8: 1189-1232. 158. Padilla-Mejia NE, Florencio-Martinez LE, Figueroa-Angulo EE, Manning-Cela RG, Hernandez-Rivas R, et al. (2009) Gene organization and sequence analyses of transfer RNA genes in Trypanosomatid parasites. BMC Genomics 10: 232. 159. Lukes J, Skalicky T, Tyc J, Votypka J, Yurchenko V (2014) Evolution of parasitism in kinetoplastid flagellates. Mol Biochem Parasitol 195: 115-122. 160. Gowri VS, Ghosh I, Sharma A, Madhubala R (2012) Unusual domain architecture of aminoacyl tRNA synthetases and their paralogs from Leishmania major. BMC Genomics 13: 621. 161. Agarwal V, Nair SK (2012) Aminoacyl tRNA synthetases as targets for antibiotic development. MedChemComm 3: 887-898. 162. Elf J, Ehrenberg M (2005) Near-critical behavior of aminoacyl-tRNA pools in E. coli at rate-limiting supply of amino acids. Biophys J 88: 132-146. 163. Lehar J, Zimmermann GR, Krueger AS, Molnar RA, Ledell JT, et al. (2007) Chemical combination effects predict connectivity in biological systems. Mol Syst Biol 3: 80. 164. Cokol M (2013) Drugs and their interactions. Curr Drug Discov Technol 10: 106- 113. 165. Hegreness M, Shoresh N, Damian D, Hartl D, Kishony R (2008) Accelerated evolution of resistance in multidrug environments. Proc Natl Acad Sci U S A 105: 13977-13981. 166. Lowe TM, Chan PP (2016) tRNAscan-SE On-line: integrating search and context for analysis of transfer RNA genes. Nucleic Acids Res 44: W54-57. 167. Laslett D, Canback B (2004) ARAGORN, a program to detect tRNA genes and tmRNA genes in nucleotide sequences. Nucleic Acids Res 32: 11-16. 168. Levenshtein VI. Binary codes capable of correcting deletions, insertions, and reversals. 169. Ward Jr JH (1963) Hierarchical grouping to optimize an objective function. Journal of the American statistical association 58: 236-244. 170. Team RC (2018) R: A language and environment for statistical computing. 171. de Souza DAS, Pavoni DP, Krieger MA, Ludwig A (2018) Evolutionary analyses of myosin genes in trypanosomatids show a history of expansion, secondary losses and neofunctionalization. Sci Rep 8: 1376. 172. Hughes AL, Piontkivska H (2003) Molecular phylogenetics of Trypanosomatidae: contrasting results from 18S rRNA and protein phylogenies. Kinetoplastid Biol Dis 2: 15. 173. Pothirat T, Tantiworawit A, Chaiwarith R, Jariyapan N, Wannasan A, et al. (2014) First isolation of Leishmania from Northern Thailand: case report, identification as Leishmania martiniquensis and phylogenetic position within the Leishmania enriettii complex. PLoS Negl Trop Dis 8: e3339.

157

174. Kelly S, Ivens A, Mott GA, O'Neill E, Emms D, et al. (2017) An Alternative Strategy for Trypanosome Survival in the Mammalian Bloodstream Revealed through Genome and Transcriptome Analysis of the Ubiquitous Bovine Parasite Trypanosoma (Megatrypanum) theileri. Genome Biol Evol 9: 2093-2109. 175. Chan PP, Lowe TM (2016) GtRNAdb 2.0: an expanded database of transfer RNA genes identified in complete and draft genomes. Nucleic Acids Res 44: D184-189. 176. Eddy SR, Durbin R (1994) RNA sequence analysis using covariance models. Nucleic Acids Res 22: 2079-2088. 177. Lowe TM, Eddy SR (1997) tRNAscan-SE: a program for improved detection of transfer RNA genes in genomic sequence. Nucleic Acids Res 25: 955-964. 178. Gouy M, Guindon S, Gascuel O (2010) SeaView version 4: A multiplatform graphical user interface for sequence alignment and phylogenetic tree building. Mol Biol Evol 27: 221-224. 179. Sprinzl M, Horn C, Brown M, Ioudovitch A, Steinberg S (1998) Compilation of tRNA sequences and sequences of tRNA genes. Nucleic Acids Res 26: 148-153. 180. Lawrence TJ, Kauffman KT, Amrine KC, Carper DL, Lee RS, et al. (2015) FAST: FAST Analysis of Sequences Toolbox. Front Genet 6: 172. 181. Schneider TD, Stormo GD, Gold L, Ehrenfeucht A (1986) Information content of binding sites on nucleotide sequences. J Mol Biol 188: 415-431. 182. Nemenman I, Shafee F, Bialek W. Entropy and inference, revisited; 2002. pp. 471- 478. 183. Ibba M, Kast P, Hennecke H (1994) Substrate specificity is determined by amino acid binding pocket size in Escherichia coli phenylalanyl-tRNA synthetase. Biochemistry 33: 7107-7112. 184. Sampson JR, Uhlenbeck OC (1988) Biochemical and physical characterization of an unmodified yeast phenylalanine transfer RNA transcribed in vitro. Proc Natl Acad Sci U S A 85: 1033-1037. 185. Schulze CJ, Bray WM, Woerhmann MH, Stuart J, Lokey RS, et al. (2013) "Function-first" lead discovery: mode of action profiling of natural product libraries using image-based screening. Chem Biol 20: 285-295. 186. Farr SB, Kogoma T (1991) Oxidative stress responses in Escherichia coli and Salmonella typhimurium. Microbiol Rev 55: 561-585. 187. Shcherbik N, Pestov DG (2019) The Impact of Oxidative Stress on Ribosomes: From Injury to Regulation. Cells 8. 188. Venkat S, Gregory C, Gan Q, Fan C (2017) Biochemical Characterization of the Lysine Acetylation of Tyrosyl-tRNA Synthetase in Escherichia coli. Chembiochem 18: 1928-1934. 189. Umehara T, Kosono S, Soll D, Tamura K (2018) Lysine Acetylation Regulates Alanyl-tRNA Synthetase Activity in Escherichia coli. Genes (Basel) 9. 190. Kuhn ML, Zemaitaitis B, Hu LI, Sahu A, Sorensen D, et al. (2014) Structural, kinetic and proteomic characterization of acetyl phosphate-dependent bacterial protein acetylation. PLoS One 9: e94816.

158

191. Schilling B, Christensen D, Davis R, Sahu AK, Hu LI, et al. (2015) Protein acetylation dynamics in response to carbon overflow in Escherichia coli. Mol Microbiol 98: 847-863. 192. Yoon I, Nam M, Kim HK, Moon HS, Kim S, et al. (2020) Glucose-dependent control of leucine metabolism by leucyl-tRNA synthetase 1. Science 367: 205- 210. 193. Moore SD, Sauer RT (2007) The tmRNA system for translational surveillance and ribosome rescue. Annu Rev Biochem 76: 101-124. 194. Komine Y, Kitabatake M, Yokogawa T, Nishikawa K, Inokuchi H (1994) A tRNA- like structure is present in 10Sa RNA, a small stable RNA from Escherichia coli. Proc Natl Acad Sci U S A 91: 9223-9227. 195. Keiler KC, Waller PR, Sauer RT (1996) Role of a peptide tagging system in degradation of proteins synthesized from damaged messenger RNA. Science 271: 990-993. 196. Williams KP, Martindale KA, Bartel DP (1999) Resuming translation on tmRNA: a unique mode of determining a reading frame. EMBO J 18: 5423-5433. 197. Gottesman S, Roche E, Zhou Y, Sauer RT (1998) The ClpXP and ClpAP proteases degrade proteins with carboxy-terminal peptide tails added by the SsrA-tagging system. Genes Dev 12: 1338-1347. 198. Barends S, Wower J, Kraal B (2000) Kinetic parameters for tmRNA binding to alanyl-tRNA synthetase and elongation factor Tu from Escherichia coli. Biochemistry 39: 2652-2658. 199. Gillet R, Felden B (2001) Transfer RNA(Ala) recognizes transfer-messenger RNA with specificity; a functional complex prior to entering the ribosome? EMBO J 20: 2966-2976. 200. Karzai AW, Susskind MM, Sauer RT (1999) SmpB, a unique RNA-binding protein essential for the peptide-tagging activity of SsrA (tmRNA). EMBO J 18: 3793- 3799. 201. Barends S, Karzai AW, Sauer RT, Wower J, Kraal B (2001) Simultaneous and functional binding of SmpB and EF-Tu-TP to the alanyl acceptor arm of tmRNA. J Mol Biol 314: 9-21. 202. Roche ED, Sauer RT (1999) SsrA-mediated peptide tagging caused by rare codons and tRNA scarcity. EMBO J 18: 4579-4589. 203. Macek B, Gnad F, Soufi B, Kumar C, Olsen JV, et al. (2008) Phosphoproteome analysis of E. coli reveals evolutionary conservation of bacterial Ser/Thr/Tyr phosphorylation. Mol Cell Proteomics 7: 299-307. 204. Cohen P (1992) Signal integration at the level of protein kinases, protein phosphatases and their substrates. Trends Biochem Sci 17: 408-413. 205. Yaffe MB, Elia AE (2001) Phosphoserine/threonine-binding domains. Curr Opin Cell Biol 13: 131-138. 206. Wakasugi K, Schimmel P (1999) Two distinct cytokines released from a human aminoacyl-tRNA synthetase. Science 284: 147-151.

159

207. Raina M, Moghal A, Kano A, Jerums M, Schnier PD, et al. (2014) Reduced amino acid specificity of mammalian tyrosyl-tRNA synthetase is associated with elevated mistranslation of Tyr codons. J Biol Chem 289: 17780-17790. 208. Sarkar J, Mao W, Lincecum TL, Jr., Alley MR, Martinis SA (2011) Characterization of benzoxaborole-based antifungal resistance mutations demonstrates that editing depends on electrostatic stabilization of the leucyl-tRNA synthetase editing cap. FEBS Lett 585: 2986-2991. 209. Witzky A, Tollerson R, 2nd, Ibba M (2019) Translational control of antibiotic resistance. Open Biol 9: 190051. 210. Rathnayake UM, Wood WN, Hendrickson TL (2017) Indirect tRNA aminoacylation during accurate translation and phenotypic mistranslation. Curr Opin Chem Biol 41: 114-122. 211. Su HW, Zhu JH, Li H, Cai RJ, Ealand C, et al. (2016) The essential mycobacterial amidotransferase GatCAB is a modulator of specific translational fidelity. Nat Microbiol 1: 16147. 212. Chaudhuri S, Li L, Zimmerman M, Chen Y, Chen YX, et al. (2018) Kasugamycin potentiates rifampicin and limits emergence of resistance in Mycobacterium tuberculosis by specifically decreasing mycobacterial mistranslation. Elife 7.

160