Ncs2* mediates in vivo virulence of pathogenic yeast through sulphur modification of cytoplasmic transfer RNA

Abstract Fungal pathogens threaten ecosystems and human health. Understanding the molecular basis of their virulence is key to develop new treatment strategies. Here, we characterize NCS2*, a point mutation identified in a clinical baker's yeast isolate. Ncs2 is essential for 2-thiolation of tRNA and the NCS2* mutation leads to increased thiolation at body temperature. NCS2* yeast exhibits enhanced fitness when grown at elevated temperatures or when exposed to oxidative stress, inhibition of nutrient signalling, and cell-wall stress. Importantly, Ncs2* alters the interaction and stability of the thiolase complex likely mediated by nucleotide binding. The absence of 2-thiolation abrogates the in vivo virulence of pathogenic baker's yeast in infected mice. Finally, hypomodification triggers changes in colony morphology and hyphae formation in the common commensal pathogen Candida albicans resulting in decreased virulence in a human cell culture model. These findings demonstrate that 2-thiolation of tRNA acts as a key mediator of fungal virulence and reveal new mechanistic insights into the function of the highly conserved tRNA-thiolase complex.


INTRODUCTION
Fungal pa thogens threa ten human nutrition and health as well as animal biodi v ersity ( 1 , 2 ). In general, mammals are resistant to most fungi and only very few species are part of their microbiomes where they usually remain outside epithelial barriers as benign colonizers ( 3 ). Some fungal pathogens can cause superficial infections, like Candidaassociated thrush or dermatophyte-associated skin disorders. Howe v er, if a fungus establishes a systemic bloodstream infection, e v en optimal medical care does not prevent high mortality rates ( 4 ). A key factor underlying the basal anti-fungal resistance is the high body temperature of mammals, to which most fungi are sensiti v e. This has led to a stark interest in identifying genes r equir ed for hightemperature growth, since such genes likely facilitate systemic infections ( 5 , 6 ). Candida albicans is frequently present in the human gastrointestinal tract as a part of the regular gut microbiome of healthy individuals ( 3 ). While it harmlessl y inhabits m ucosal surfaces, under certain predisposing conditions it can enter the bloodstream and lead to infection of internal organs and trigger septic shock. Indeed, it is one of the most common causes of nosocomial fungemias. C. albicans is a polymorphic organism that grows , among others , in a yeast form and in a filamentous form, which are both critical during different stages of infection. The ability of C. albicans to trigger the morphological switch as a rapid response to varying growth conditions has been linked to its virulence ( 7 ). In contrast, baker's yeast is uni v ersally associated with human lifestyle through different fermentation processes and is Generally Recognized As Safe (GRAS) ( 8 ). Howe v er, if ab le to grow at high tempera tures, Sacchar om y ces cer evisiae can turn into an opportunistic pathogen and infect immunocompromised patients ( 9 ). This phenomenon is becoming incr easingly r elevant in the clinical routine due to rising numbers of patients with an impaired immune system ( 10 ).
Transfer RNAs (tRNAs) are the adaptor molecules that physically link messenger RNA (mRNA) codons to their respecti v e amino acid during ribosomal decoding. Importantl y, tRN As carry a plethora of chemical modifications tha t af fect all aspects of tRN A biolo gy ( 11 ). Modifications in the anticodon loop tune codon-anticodon interactions and their absence can perturb mRNA translation by affecting the accuracy, speed, and efficiency of protein synthesis ( 12 , 13 ). Ther efor e, perturbations in tRNA modifications are linked to a growing list of degenerati v e and metabolic human diseases as well as cancer (14)(15)(16). A highly conserved group of chemical tRNA modifications are those of wobble uridine (U 34 ) (Figure 1 A and B). Nearly all U 34 in eukaryotic tRNA are modified by the Urm1-or the Elp pathway (for a recent re vie w, see ( 17 )). The Urm1 pathway consists of four essential core members. U biquitin r elated m odifier 1 (Urm1) and its activating enzyme Ub iquitin-a ctivating 4 (Uba4) receive and activate sulphur from the cysteine desulphurase N i FS -like (Nfs1) enzyme ( 18 ). The activated sulphur is subsequently transferred to the N eeds-C la4-to-S urvi v e 2 / 6 (Ncs2 / Ncs6) complex. Ncs6 is a Fe / S-cluster-containing enzyme ( 19 , 20 ) that adenylates target tRNAs and performs the final sulphurtransfer reaction to generate 5-methoxycarbonylmethyl-2-thiouridine (mcm 5 s 2 U) ( Figure 1 A and B) ( 17 ). Ncs2, the homologous binding partner of Ncs6 has lost its enzymatic function and likely acts as a structural scaffold or provides target specificity ( 21 ). Ne v ertheless, Ncs2 is essential for 2thiolation and ncs2 yeast lacks mcm 5 s 2 U-modified tRNA ( 21 , 22 ). The Elongator protein (Elp) complex catalyses the first steps to synthesize 5-methoxycarbonylmethyluridine (mcm 5 U) and 5-carbamoylmethyluridine (ncm 5 U) on U 34 . The dodecameric Elp complex contains two copies of each six individual protein subunits (Elp1-6), all of which are essential for ncm 5 U / mcm 5 U forma tion ( 23 , 24 ). W hile s 2 is not r equir ed for mcm 5 synthesis, the absence of Elongator function leads to a significant decrease in s 2 le v els in yeast. Remar kab ly, the lack of mcm 5 s 2 U 34 induces a codon-specific slo w do wn of mRNA translation ( 12 , 13 ) and thereby triggers protein-homeostasis defects ( 13 ). Furthermore, the absence of U 34 modifications induces complex phenotypes ( 25 , 26 , 21 ). Interestingly, a mutation in NCS2 was reported to underlie the ability of a clinical isolate of S. cerevisiae to grow at high temperatures ( 5 ). This observation prompted us to ask whether tRNA 2-thiolation directly contributes towards the virulence potential of pathogenic yeasts.
Here, we demonstrate that a single nucleotide polymorphism (SNP) in NCS2 , found in a patient-deri v ed baker's y east str ain, is sufficient to promote yeast virulence. Specifically, yeast harbouring this H71L mutation ( NCS2 *) grows a t high tempera tur e and is r esistant to oxidati v e stress, inhibition of nutrient signalling, and cell-wall stress in a 2thiolation-dependent manner. NCS2 * is characterized by increased le v els of 2-thiolation at body temperature compared to wild type. We find that Ncs6 binds Ncs2* more tightly than Ncs2 and that the thiolase complex remains more stable providing a molecular explanation for the increased activity of the 2-thiola tion pa thway in NCS2 * yeast. Furthermore, w e observe low er le v els of protein aggregates in NCS2 * yeast when challenged with high temperatures compared to the wild type. Finally, we show that in the absence of NCS2 function S. cerevisiae and the common pathogen C. albicans are less virulent in vitro and in vivo . These findings re v eal the physiological importance of the highly conserved tRNA modification mcm 5 s 2 U as a key mediator of fungal virulence.
(Formedium) at 30 • C, 200 rpm unless stated otherwise. Yeast overnight cultures were diluted to OD 600 0.2, grown at the indicated temperatures to OD 600 0.8-1.0 and harvested by centrifugation for 3 min at 5000 g unless stated otherwise. Yeast pellets were snap frozen in liquid nitrogen and stored at −80 • C. Pellets of C. albicans were washed once with ethanol to kill the cells before freezing. For some experiments pr e-cultur es wer e grown at 37 • C to ensur e a full reduction of 2-thiolation le v els in wild-type yeast. All yeast strains are listed in Supplementary Table S2.

Spot-dilution assays
Overnight cultures in YPD were diluted to OD 600 0.75 in sterile water. A serial dilution of 1:5 was pr epar ed in 96-well plates and spotted onto YPD agar plates adding drugs as indica ted. The pla tes were briefly dried a t room tempera ture and subsequently incubated for 2-3 days at the designated tempera ture. To investiga te the ability of baker's yeast to invade agar, cultur es wer e spotted on YPD plates. Following incubation for 2 days at the appropriate temper ature, y easts were washed off with sterile water and the plates were dried and incubated for one additional day at 30 • C.

Isolation and purification of tRNA and total RNA
Bulk tRNA was isolated as previously described ( 30 ). To isolate total RNA, smaller culture volumes (2-4 ml) were used. Frozen cell pellets were resuspended in 500 l of 0.9 % NaCl, adding 500 l acidic phenol and 75 l 1-bromo-3-chlor opr opane (BCP). One vortexing step was carried out for 10 min and the aqueous phase was transferred to tubes containing 500 l acidic phenol and 50 l BCP. RNA was precipitated overnight in 2 ml ethanol and pellets were washed with 80 % ethanol. The dried RNA was resuspended in 10 l RNase-free water. tRNA 2-thiolation analysis 0.4 g of bulk tRNA or 1 g of total RNA per sample were heated for 5 min at 80 • C and applied onto denaturing polyacrylamide gels (8 % polyacrylamide (19:1), 7 M urea, with or without 0.05 % ([ N -acryloylamino]phenyl)mercuric chloride (APM) in 0.5 × TBE) ( 31 ). RNA was transferred onto Immobilon NY + membranes (Millipore) by semi-dry transfer with 0.5 × TBE. Northern blot was performed essentially as described using specific DNA-oligo probes ( 21 ). Signals were collected and analysed using a Fuji FLA 7000 phosphoimager (Fujifilm). Digital image analysis was performed using Fiji-ImageJ2 (The Fiji Project).

Isolation of endogenous protein aggregates from yeast
To isolate the endogenous protein aggregates, overnight cultures grown at the indicated temperatures were diluted to OD 600 0.05 in 100 ml prewarmed YPD and harvested by filtra tion, a t OD 600 0.5. The yeast pellet was snap frozen in liquid nitrogen and stored at −80 • C. Aggregate isolation was performed essentially as described ( 32 ). Lysates were pr epar ed by r esuspending frozen yeast pellets in 1 ml resuspension buffer (20 mM NaPi pH 6.8, 1 mM EDTA, 10 mM DTT, 0.1 % Tween 20) that contained a cocktail of protease inhibitors (0.5 mM AEBSF, 10 g / ml aprotinin, 0.5 mg / ml benzamidine, 20 M leupeptin, 5 M pepstatin A). 60 U of zymolyase was added and cells were incubated for 20 min at 22 • C, 1400 rpm in a thermo-shaker (Eppendorf) and afterwards chilled on ice. Lysates were transferred to a cold 50 ml reaction tube containing 1 ml of resuspension buffer and sonicated with 8 pulses at le v el 4, duty cycle 50 % with a tip sonicator (Branson) and chilled on ice. Centrifugation for 20 min at 200 g, 4 • C was carried out in 2 ml reaction tubes. Protein concentration of the supernatant was determined by Bradford analysis. Subsequently, total protein of all samples was adjusted with buffer AIB to equal amounts (1-2.5 mg per sample). 50 l of this solution was kept as total-protein control. The remaining lysates were centrifuged for 20 min at 16 000 g, 4 • C . The superna tant was discarded, and the pellets were w ashed in w ash buffer 1 (20 mM NaPi pH 6.8, 2 % NP-40 cOmplete protease inhibitor cocktail (Roche)), vortexing the samples till the pellets were dislodged and sonication with 6 pulses at le v el 4, duty cy cle 50 % followed by centrifugation as before. This step was repeated once. The supernatant was discarded and 2 ml of wash buffer 2 (20 mM NaPi pH 6.8, cOmplete protease inhibitor cocktail (Roche)) was added, the samples wer e vortex ed and sonicated with 4 pulses at le v el 2, duty cycle 65 %, transferred to 2 ml reaction tubes and centrifuged for 20 min at 16 000 g, 4 • C. The supernatant was discarded and all remaining liquid carefully aspira ted. Protein-aggrega te pellets wer e r esuspended in 100 l 4 × sample buffer containing 8 M urea, to 50 l of the total protein control samples 25 l of the same buffer were added and all samples boiled for 10 min at 95 • C and stored at −20 • C. 10 l of each sample was applied to 12 % polyacrylamide gels and stained with Colloidal Blue (Invitrogen).

Pulse-chase pr otein-turnov er experiments
Yeast overnight cultures were diluted to OD 600 0.2 in YPD and incubated at 30 • C or 37 • C. After 3.5 h, cy clohe ximide was added to a final concentration of 200 g / ml. Samples were harvested by centrifugation, cell pellets were resuspended in PBS and adjusted to the 9 OD 600 units. Proteins were isolated as described previously ( 33 ), applied to 10 % polyacrylamide gels and analysed by Western blot.

Yeast two-hybrid assays
Y1026 y east tr ansformed with the y east-two-hybrid plasmids was grown in SRaf-HT medium overnight. On the next morning, the cultures were diluted in SRaf-HT to OD 600 0.2 and incubated at 30 • C to an OD 600 0.55-0.65). Subsequently, galactose was added to a final concentration of 2 % and incubated for 1.5 h. The OD 600 was measured and used for calculating Miller Units (MU). All subsequent steps were carried out on ice. 1 ml of each sample was centrifuged for 3 min at full speed. The pellets were washed with 1 ml resuspension buffer (60 mM Na 2 HPO 4 , 10 mM KCl, 40 mM NaH 2 PO 4 , 1 mM MgSO 4 ) and centrifuged as above. The supernatant was discarded, and pellets were resuspended in 150 l of fresh resuspension buffer containing 0.27 % ␤-mercaptoethanol. After adding 50 l BCP and 20 l of 0.1 % SDS samples were mixed for 10-15 s using a vortex. The reaction was started by adding 700 l of prewarmed ONPG. When the reaction mix turned yellow, the reaction was stopped by adding 500 l fresh 1 M Na 2 CO 3 . Samples were centrifuged for 3 min at full speed and OD 420 of the supernatant was determined using a Synergy Mx plate reader (BioTek). Miller Units were calculated as MU = (OD 420 *1000) / (OD 600 * x min*1 ml), where x is the time (in min) that passed between adding ONPG and Na 2 CO 3 .

Quantification of cytotoxicity (LDH assays)
Confluent monolayers of A498 or C2BBe1 cells in 96-well plates were infected for 24 h with C. albicans (MOI = 1) from an overnight culture, incuba ted a t 37 • C in YPD. Cytotoxicity was determined as r eported befor e by a colorimetric assay that measures lactate dehydrogenase (LDH) release using a cytotoxicity detection kit (Roche) according to the manufacturer's instructions ( 34 ). Damage by mutant strains was calculated (after subtraction of spontaneous release in all samples) as percent of LDH release caused by the isogenic wild-type strain SN87HL.

In vivo infection experiments
For in vivo infection experiments strains carrying two copies of NCS2 , NCS2 *, or ncs2 were deri v ed from YJM128. 3 ·10 7 yeast cells of each strain were resuspended in 200 l sterile PBS and injected into the tail vein of 3-4-week-old female NOD.CB17 / scid mice ( n = 18 animals per group). Mice were allowed to eat and drink ad libitum and were monitor ed thr ee times daily. The experiment was terminated either when mice showed signs of clear distress or when the endpoint was reached after 5 days. The experiments were performed in a b linded manner. Survi val curv es were analysed using Prism (GraphPad). The use of mice followed the ethical guidelines of the European Laboratory Animal Science Associations (FELASA). All experiments recei v ed the ethical approval issued by the Landesamt f ür Natur, Umwelt und Verbraucherschutz (LANUV) of the state of North Rhine-Westphalia, Germany (Permit 84-02.04.2013.A018).

Ribosome profiling
Ribosome profiling libraries were prepared as described previousl y (35)(36)(37). Briefly, wild-type and ncs2 / C. albicans y east str ains wer e grown over night, r e-diluted and grown in YPD to OD 600 0.4. Cells were harvested by vacuum filtration using a 0.45 m nitrocellulose filter (GE Healthcare) and immediately flash frozen in liquid nitrogen. The cells were lysed using a Freezer-Mill (SPEX Sam-plePrep) with the settings: 2 cycles of 2 min precooling, 2 min at 5 CPS and 1 min intermittent cooling. Lysates were thawed in lysis buffer (20 mM Tris-HCl pH 7.4, 5 mM MgCl 2 , 100 mM NaCl, 1 % Triton, 2 mM DTT, 100 g / ml CHX and clarified by two rounds of centrifugation (3 min; 4 • C; 3000 g and 5 min, 4 • C and 10 000 g). 10 A 260 units of cleared lysates were digested with 250 U Ambion RNase I (ThermoFisher) for 1 h at 22 • C and 1400 rpm agitation. The reaction was stopped by adding 15 l SuperaseIn (ThermoFisher). To isolate monosomes, the digested RNA was loaded on a 10-50 % sucrose gradient and spun in an ultracentrifuge (SW41 TI rotor; 3 h; 4 • C; 35 000 rpm). Fractions containing monosomes were collected using a piston gr adient fr actionator (Biocomp) and RNA was extr acted with acidic phenol and BCP. Purified RNA was separated on a 15 % PAA gel and ribosome protected fragments (27-29 nt) were excised and precipitated using 0.3 M NaOAc. RPF libr ary prepar ation was carried out as described before using a 3 -adapter carrying 4 randomized positions at the 5 end ( 36 , 37 ). Sequencing was performed on an Illumina HiScanSQ instrument.

Ribosome profiling analysis
Ribosome profiling reads were processed as described ( 37 ): Adapter sequences were clipped and the 4 randomized nucleotides trimmed using the FASTX-Toolkit ( http://hannonlab.cshl.edu/fastx toolkit , June 2017), version 0.0.13. The processed r eads wer e mapped to ORFs (cgdGene) using bowtie. Reference ORFs (Allele A, C albicans SC5314 version A22) were extended by 18 nt into the UTRs. A-site codons -excluding footprints that mapped to the first and last 15 codons of a transcript -were mapped according to the frame of the 5 end of footprints and an appropriate offset was defined ( 13 ). For differential gene-expression analysis, gene count tables were generated and the analysis was performed using DESeq2 ( 38 ). For altered transcripts, the padj a-threshold was set to 0.05. To identify unaltered transcripts, the althypothesis function was set to 'lessAbs'. Gene ontology (GO) enrichment analysis used TopGO. Redundant GO-terms wer e r emoved using Revigo ( 39 ). were analysed using T-Cof fee (tcof fee.crg.ca t) using the standard MSA settings ( 40 , 41 ). The colour of the amino acids indicates the likelihood that the residues are correctly aligned (b lue: 0, dar k red: 9). Values > 5 (y ellow, or ange, and r ed) ar e likely corr ectly aligned. The alignment was manually edited to match the format size and minor alignment changes were introduced in low-quality regions of the alignment.

Structural models
We have built structural models of the Ncs2 / 6 heterodimer using the AlphaFold 2 Protein Structure Database ( 42 ) and the Protein Homolo gy / analo gy Reco gnition Engine (Phyre2) ( 43 ). Primary structural features of the complex were obtained from the crystal structures of TtuA (PDB: 3VRH, 5ZTB), an archaeal structural homolog of Ncs6 in T. thermophilus ( 44 , 45 ). The tRNA positioning was inferred from the crystal structure of TilS-tRNA (PDB: 3A2K), a lysidine synthetase from Geobacillus kaustophilus , superimposed on the Ncs6 subunit.

GST-pulldown for ncs2 / 6 interaction
CtNcs2 or ChtNcs6 proteins were incubated with GST-CtNcs6 and GST-CtNcs2 respecti v ely, in reaction buffer (20 mM Tris pH 8, 150 mM NaCl and 1 mM DTT). The reaction mix was subsequently added to equilibrated Glutathione Sepharose 4B beads (GE Healthcare) and incubated on a rotating wheel for 120 min at 4 • C. After binding, glutathione beads were collected by gentle spinning at 500 g and subsequently washed three times with a 0.05 % (v / v) Tween 20-containing reaction buffer. Bound proteins wer e denatur ed at 95 • C in the pr esence of Laemmli sample buffer and analysed on BoltTM 4-12 % Bis-Tris Plus Gels (Thermo Fisher Scientific). For protein visualization, the gels were stained with Coomassie Brilliant Blue. Inputs were collected after the reaction and before the pull-down and r epr esent 7 % of the pull down.

Thermal-shift assays
To stabilize CtNcs6, 10 g of either CtNcs2, CtNcs6 or CtNcs2 / 6 were suspended in 50 mM HEPES pH 8, 150 mM NaCl, 2 mM DTT. The influence of nucleotides on the thermostability of CtNcs2 and CtNcs2 / 6 was tested in the stora ge b uffer supplemented with 2 mM MgCl 2 . Each protein was measured in a buffer without nucleotide, with 1 mM ADP, 1 mM ATP or 1 mM AMP-PNP. Samples were gradually heated 4-98 • C for 2 h. The denaturation process was tracked using the hydrophobic fluorescent dye SYPRO Orange (Sigma Aldrich). Fluorescence was measured using a Bio-Rad CFX96 thermocycler. Melting temperatur es wer e calculated from the average of the peaks of the first derivati v e from at least three technical replicas.

NCS2* confers cellular fitness under stress
A screen for mutations that enable clinical isolates of baker's yeast to grow at high temperatures identified a SNP in NCS2 (A212T), which leads to a histidine-to-leucine substitution (H71L) ( 5 ). Since Ncs2 is a component of the tRNA-thiolase complex ( 17 ), we sought to investigate the role of this highly conserved tRNA modification in yeast virulence. We used the common laboratory yeast background S288c to generate isogenic strains that only differ by this SNP (further called NCS2 for the wild-type allele and NCS2* for the H71L mutant; Supplementary Figure  S1A). First, we grew the strains on YPD at 30 • C, 37 • C and 40 • C and confirmed that NCS2* yeast grew better than NCS2 or ncs2 yeast at 40 • C, while we observed no difference at lower temperatures (Figure 1 C). Next, we exposed the strains to conditions that r equir e 2-thiolation. NCS2 * grew better than the wild type when challenged with the TOR inhibitors rapamycin and caffeine, the sulfhydryloxidizing agent diamide, the cell-wall damaging reagent calcofluor white (CFW) or paromomycin at 37 • C (  Table S1). Furthermor e, we tr eated yeast with zymocin, a toxin that specifically cleaves mcm 5 s 2 U-modified tRNA, while hypomodified tRNA is resistant to nucleolytic cleavage (46)(47)(48). Strikingly, NCS2 * yeast did not grow at 37 • C ( Figure  1 E) while NCS2 and ncs2 yeast grew under these conditions. When performing similar experiments at 30 • C, the difference between NCS2 and NCS2 * was negligible (Figure 1 E). Notably, in wild-type yeast, s 2 U levels in tRNA ar e r educed a t eleva ted tempera tures, which is consistent with the observed effect in response to zymocin ( 30 , 49-51 ).
Ther efor e, we compar ed the le v els of tRNA 2-thiolation in NCS2 and NCS2* yeast a t dif ferent tempera tures by ([ Nacryloylamino]phenyl)mercuric chloride (APM)-affinity gel electrophoresis and northern blotting ( 31 ). 2-thiolation levels were similar in NCS2 and NCS2* at 25 • C and at 30 • C (Figure 1 F, G and Supplementary Figure S1C-S1F). Howe v er, at 37 • C 2-thiolation was significantly decreased in NCS2 , while remaining stable in NCS2* yeast (Figure 1 F, G and Supplementary Figure S1C-S1F). Our results show that the NCS2 * mutation confers resistance to temperature and chemical stress and that this effect is likely mediated by 2-thiola tion. Furthermore, our da ta confirm tha t eleva ted temperatur es r egulate the activity of the modifica tion pa thway consistent with previous findings ( 30 , 49-51 ).
The cellular effects of NCS2* are mediated through 2thiolation of tRNA Since Ncs2 is r equir ed for 2-thiolation of U 34 , we asked whether NCS2 * confers cellular fitness by affecting tRNA functionality. Ther efor e, we deleted genes encoding for members of the Elongator complex ( ELP4 and ELP6 ) and the Urm1 pathway ( URM1 and NCS6 ) in NCS2 or NCS2 * yeast. In the absence of ELP4 or ELP6 , xm 5 U 34 modifications are not f ormed. Theref ore, s 2 le v els are reduced ( 17 , 52 ). Inter estingly, the str ess sensitivity of elp4 and elp6 yeast was rescued in NCS2* but not in NCS2 yeast ( Figure 2 A and Supplementary Figure S2A), strengthening a direct link of the phenotype to s 2 U le v els. Ne xt, we deleted URM1 or NCS6 , which are essential for 2-thiolation and found that the NCS2 * urm1 and NCS2 * ncs6 yeasts were indistinguishable from NCS2urm1 and NCS2ncs6 yeast under stress at 37 • C, thus, demonstrating that NCS2 * exerts its function through 2-thiolation (Figure 2 B and Supplementary Figure S2B). Finall y, tRN A is not 2-thiolated in NCS2 * urm1 or NCS2 * ncs6 yeast (Figure 2 C), showing that Ncs2* does not bypass the r equir ement of the Urm1 pathway. Overall, these findings establish that NCS2* enhances cellular fitness through 2-thiolation of U 34 .
The NCS2 * mutation is specific S. cer evisiae na turall y and predominantl y exists in a diploid f orm. Theref ore, we asked whether a single allele of NCS2 * is sufficient to confer a fitness advantage in S288c. To address this question, we generated a series of isogenic diploid strains with all possible combinations of the NCS2* , NCS2 and ncs2 alleles. While ncs2 / ncs2 yeast was sensiti v e to stress when grown at 37 • C, a single copy of the NCS2 allele partially rescued growth at high temperatures (Figure 3 A). Howe v er, a single copy of the NCS2 * allele resulted in a stronger rescue when combined with either ncs2 Δ or NCS2 , recapitulating the high-temperature growth phenotype (Figure 3 A). Remar kab ly, this shows that NCS2 * is dominant over the wild-type allele and suggests that a single copy of NCS2 * is sufficient to confer a growth advantage.
To understand the molecular mechanism by which Ncs2 * exerts its fitness advantage we analysed its sequence. Ncs2 does not contain known functional domains. Howe v er, the H71L substitution occurs within a short region that is relati v ely well conserv ed in Ncs2 (Supplementary Figure  S3A; black box). We generated three-dimensional models of Ncs2* in complex with Ncs6 and a tRNA. The first model was based on the structures of TtuA from Thermus thermophilus --a sulphurtr ansfer ase that is homologous to Ncs2 / Ncs6 --and tRNA-bound TilS from Geobacillus kaustophilus (Figure 3 B), and the second very consistent model is based on an AlphaFold 2 prediction (Supplementary Figure S3B) ( 44 , 45 ). The anticodon loop of the tRNA fits into the catalytic pocket of Ncs6 in the vicinity of its ATPase domain and its 4Fe / 4S-iron-sulfur cluster. The Cterminus of Ncs2 contacts the core of the tRNA, thereby likely stabilizing the tRNA-thiolase complex in line with its potential role as a specificity factor ( 21 ). The H71L substitution is located ∼25 Å away from the Ncs2-tRNA interface in a region that is homologous to the ATPase domain of Ncs6. We next asked whether the high-temperature growth phenotype of NCS2 * yeast was specific to the H71L substitution or whether any other amino acid substitution was suf ficient to recapitula te the phenotype. Ther efor e, we r eplaced H71 with alanine (H71A) or isoleucine (H71I). Howe v er, yeast carrying these substitutions did not show a fitness advantage (Figure 3 C). Furthermore, 2-thiolation levels wer e r educed in both mutants at 37 • C (Figure 3 D). Our results show that the enhanced fitness is specific to the H71L substitution and that it occurs by increasing 2-thiolation le v els at high temperatures.

The H71L substitution enhances thiolase-complex stability
Since Ncs2 forms a heterodimer with Ncs6 in different organisms ( 53 , 21 ), we asked whether the formation of the thiolase complex underlies the fitness advantage of Ncs2*. Ther efor e, we probed the interaction between Ncs6 and the Ncs2 variants by yeast-two-hybrid (Figure 3 E). Strikingly, the interaction of Ncs6 with Ncs2* was a pproximatel y 2fold higher than between Ncs6 and Ncs2, while the interactions with Ncs2 H71A or Ncs2 H71I were similar or slightly weaker than with the wild-type protein (Figure 3 E). This led us to hypothesize that the stronger interaction between Ncs6 and Ncs2* stabilizes the activity of the thiolase complex. We tested this by monitoring the turnover of Ncs2, Ncs2*, and Ncs6 in vivo after blocking protein synthesis with cy clohe ximide. Ncs2 and Ncs2* showed roughly similar decay rates at 30 • C or 37 • C (Figure 3 F). Howe v er, while Ncs6 decayed quickly at 37 • C in NCS2 yeast, it was more stable in NCS2 * yeast (Figure 3 F). This shows that stabilization of Ncs6 at high temperatures results in an increase of functional Ncs6 / Ncs2*, thus leading to higher tRNA 2thiolation le v els in NCS2 * y east during high-temper ature growth. To further analyse the interaction of Ncs2 and Ncs6 we performed in vitro experiments. Ncs6 contains an oxygen-sensiti v e Fe / S cluster ( 19 , 20 , 54 ). This pre v ented us from reconstituting the baker's yeast thiolase complex in vitro . Howe v er, we succeeded in expr essing r ecombinant Ncs2 and Ncs6 from Chaetomium thermophilum . As expected, both proteins form a complex ( Supplementary Figure S3C) ( 21 , 53 ). To determine the stability of the complex and its individual subunits, we used thermal shift assays. We found that the Ncs2 / Ncs6 complex is much more stable than Ncs2 or Ncs6 alone (Supplementary Figure S3D). Even though residues critical for an enzymatic function have been mutated ( 21 ) it is possible that Ncs2 binds ATP and that this stabilises the protein. Importantly, we found that the stability of Ncs2 increases significantl y, w hen it is bound to ATP, ADP or the non-hydrolysable AMP-PNP ( ttest < 0.001, n = 3-4; Supplementary Figure S3E). Binding to these three nucleotides likewise enhances the stability of the Ncs2 / 6 complex ( t -test < 0.001, n = 3-4; Supplementary Figure S3F). These data support the hypothesis that the H71L substitution stabilizes the thiolase complex likely by modulating nucleotide binding of Ncs2.

NCS2* affects protein homeostasis at high temper atur es
Codon-specific tr anslation r ates ar e decr eased in the absence of U 34 modifications ( 12 , 13 ), thereby triggering protein-homeostasis defects ( 13 ). Ther efor e, we hypothe-sized that higher le v els of 2-thiolated tRNA leads to decr eased protein aggr egation in NCS2 * cells. To test this, we grew ncs2 , NCS2 , \ and NCS2 * yeast at 30 • C and 37 • C and monitored the presence of protein aggregates ( 32 , 13 ). Indeed, ncs2 cells, but not wild-type cells contained protein aggrega tes a t 30 • C (Figure 3 G). W hen grown a t 37 • C , both NCS2 and NCS2* cells contained a ggregates, b ut the amount of aggregated protein was significantly lower in NCS2 * yeast (Figure 3 G). The accumulation of cytoplasmic protein aggregates constitutes a significant fitness cost and likely explains why NCS2* yeast cope better with stressful environments.

NCS2 * leads to increased virulence in vivo
NCS2 * was isolated from a clinical isolate of baker's yeast. To directly link tRNA 2-thiolation to yeast pathogenicity, we analysed its role in clinical S. cerevisiae isolates ( 9 , 27 , 55 ). These pathogenic strains differ from laboratory strains in their r epertoir e of SNPs and ar e generally diploid. Ther efore, we deleted both alleles of NCS2 in se v eral clinical isolates. As expected, tRNA was not thiolated in these knockouts (Figure 4 A). Interestingly, 2-thiolation le v els differ between wild-type isolates while most of them were temperature sensiti v e (Supplementary Figure S4A). We selected YJM128 as an example of a virulent strain and YJM223 as an example of a strain with intermediate virulence ( 9 , 27 , 55 ). Next, we reintroduced NCS2 or NCS2 * into the corresponding deletion strains. Importantly, both strain backgrounds reacted to stress like the laboratory yeast, showing that the results obtained in S288c are a good proxy for pathogenic baker's yeast (Figure 4 B). We subsequently anal ysed w hether 2-thiolation le v els differ between NCS2 and NCS2 * at high temperatures in these strain backgrounds and found this to be the case (Figure 4 C and Supplementary Figure S4B). Howe v er, likely due to the composition of SNP in these pathogenic isolates, a single allele of NCS2 * was not sufficient to r estor e wild-type s 2 U levels, suggesting that the dominant effect of NCS2 * that we observed for in S288c r equir es genetic interactors beyond the H71L substitution. Finally, we tested whether the addback strains can invade agar. In the YJM128 background, we found improved attachment to the substrate in NCS2 * yeast, while none of the deri vati v es of YJM223 was able to invade agar (Supplementary Figure S4C). We next determined whether increased 2-thiolation levels in NCS2* yeast confer a fitness ad vantage w hen exposed to the immune system of a human host. Thus, we incubated the yeast cells with blood from healthy human donors and monitored the efficiency of the immune cells to kill the yeast. We observed efficient killing in all cases, confirming that the intact human immune system can defend the host efficiently against baker's yeast (data not shown). Ther efor e, we applied an in vivo infection model that mimics an immunocompromised host. We injected yeast into the tail vein of 3-4-week-old NOD.CB17 / scid mice and monitored survival for 5 days ( 56 ). Strikingly, all mice were resistant to infections by ncs2 cells (Figure 4 D). Furthermore, the mice were more se v erely affected by NCS2 * yeast compared to NCS2 yeast consistent with an increased fitness of the mutant (Figure 4 D). Collecti v ely, these e xperiments demonstra te tha t 2-thiola tion facilita tes virulence when baker's yeast infects a susceptible mammalian host.

2-thiolation mediates virulence in C.albicans
Although S. cerevisiae can infect immunosuppressed patients, baker's yeast is Generally Recognized As Safe (GRAS). Ther efor e, w e used C . albicans to understand how the absence of 2-thiolated tRNA affects the virulence of a genuine fungal pathogen. First, we grew wild-type C. albicans at 25-45 • C, purified tRNA from the cells and assessed its thiolation status. This established that 2-thiolated tRNA is present in C. albicans and tha t 2-thiola tion levels remain constant at different temperatures (Supplementary Figure S5A). Next, we identified the C. albicans orthologue of NCS2 (orf19.4399) and knocked out both alleles ( 28 , 29 ). tRNA isolated from ncs2 / cells are not thiola ted, demonstra ting the conserva tion of Ncs2 function ( Figure 5 A). To test whether the ncs2 / mutant is sensiti v e to thiolation-linked stress, we challenged it by rapamycin. The ncs2 / strain was sensitive to the drug at 25-37 • C, but the effect was less pronounced at higher temperatur es (Figur e 5 B). To distinguish 2-thiolation-specific phenotypes from a general stress response, we exposed the strains to the antifungal drugs fluconazole, nystatin, and natamycin (Supplementary Figure S5B). Howe v er, we did not observe a difference in sensiti vity towar ds these drugs, suggesting that the lack of 2-thiolation does not lead to general stress sensitivity. To investigate whether a lack of thiola ted tRNA af fects high-tempera ture growth we grew the wild-type , heterozygous , and full deletion strains at 25-44 • C (Figure 5 C and Supplementary Figure S5C). At temperatures exceeding 37 • C ncs2 / colonies appeared smoother than the wild type or heterozygous strains. Colony morphology is a good proxy for the ability of the strains to form hyphae. Hence, we grew the strains under hyphae-inducing conditions. Indeed, the ncs2 / colonies neither formed radial hyphae nor central wrinkling, while wild-type colonies showed robust hyphae formation on YPD or Choco medium ( Figure 5 D). Finally, we tested whether 2-thiolation-deficient C. albicans is able to damage epithelial cells in a cell-culture model. Importantly, we found that ncs2 / cells exhibit a striking reduction in their ability to damage kidney (A498) or colon (C2BBe1) cells, indica ting tha t they are less virulent than the wild type ( Figure 5 E). These findings underscore the importance of tRNA modifications for virulence also in the human pathogen C. albicans .

Thiolation-deficient C. albicans shows translation defects
Notably, the presence of 2-thiolation of tRNA has been shown to tune codon-specific translation in different organisms ( 12 , 13 , 57 ). Ther efor e, we asked whether a lack of mcm 5 s 2 U 34 similar ly affects tr anslation dynamics in C. albicans . To address this question, we performed ribosome profiling of wild-type and ncs2 / cells. This method generates high-resolution snapshots of translation through the sequencing of ribosome-protected mRNA fragments ( 58 ) and has been used to characterize translation defects in tRN A modification m utants ( 12 , 13 , 59 ). Consistent with baker's yeast, we observed increased codon occupancy for AAA and CAA in the ribosomal A site of ncs2 / C. albicans at 37 • C (Figure 6 A) indicating that ribosomes in the mutants r equir e mor e time to decode codons that depend on hypomodified tRNA ( 12 , 13 ). Next, we performed differ ential-expr ession analyses of the ribosome profiling data to identify pathways that differ in activity between the mutants and the wild type. Interestingly, the factor that best explained differences in the gene-expression between the samples was temperature (Figure 6 B). Next, we conducted gene ontology (GO) analyses of the genes that are differ entially expr essed between ncs2 / cells and wild type. Mutant cells are characterized by changes in their metabolic programs indicated by a reorganization of glucose and amino acid metabolism and the downregulation of genes that encode for proteins that localize to mitochondria (Figure 6 C, D and Supplementary Figure S6A-S6D). Furthermore, genes belonging to GO terms that are linked to membrane biology, the hyphal cell wall and biofilm formation were consistently downregulated in the m utant, w hich is compatible with altered colony morphology and a decrease in hyphal growth. These findings suggest that the observed codon-specific translation defect in C. albicans mutants partially blocks the morphological switch to hyphal growth, ther eby r endering C. albicans less pathogenic. Like pathogenic baker's yeast, C. albicans virulence is se v er ely r educed in the absence of mcm 5 s 2 U.

DISCUSSION
Using a model of pathogenic baker's yeast and C. albicans , we demonstrate that 2-thiolation is a key mediator of virulence. Strains that lack mcm 5 s 2 U exhibit decreased fitness a t high tempera ture and fail to infect mice and to damage human cells (Figures 4 D and 5 E). To our knowledge, this is the first time that a similar phenotype has been shown in vivo for fungal pathogens.
NCS2 * is a SNP that leads to the substitution of histidine by leucine (H71L) in a conserved region of Ncs2 (Supplementary Figure S3A) ( 5 ). The gain of function that is introduced by H71L cannot be explained by a loss of the positi v e charge, since mutations to alanine (H71A) and isoleucine (H71I) do not mimic the phenotype of H71L (Figure 3 C and D). Our structural model of Ncs2 suggests that H71L is located in a putati v e ATP binding pocket (Figure 3 B and Supplementary Figure S3B). Since Ncs2 has mutated several key residues during evolution it is unlikely to hydrolyse ATP ( 21 ). Howe v er, ATP binding may stabilize Ncs2* compared to Ncs2. This attracti v e model is in line with our observa tion tha t nucleotide binding increases the stability of Ncs2 and the thiolase complex from C. thermophilum (Supplementary Figure S3E and F). Notably, the difference between Ncs2 and Ncs2* is a new example that heritable traits giving rise to complex phenotypes like hightemperature growth and virulence can be generated by a single point mutation ( 60 , 61 ). Traits that generate a fitness advantage are likely to become fixed in a population. We found that NCS2 * confers a fitness advantage under a range of stress conditions including starva tion, oxida ti v e stress, and mistranslation ( wild yeast isolates. If carrying NCS2 * were always advantageous, we would expect to observe NCS2 * in all yeast strains. This suggests that NCS2 * does not always provide a fitness advantage or e v en confers a negati v e effect under specific gro wth conditions. Ho we v er, a t high tempera tures NCS2 * becomes advantageous as indicated by its independent isolation in e volution e xperiments performed at high temperatures ( 6 ). Defects in the Urm1 pathway likewise reduce the ability for high-temperature growth in archaea, yeasts , worms , rice, and bacteria ( 25 , 53 , 62 , 63 ), a phenotype that is further enhanced by specific stress conditions. This underscores the physiological importance of mcm 5 s 2 U, consistent with findings that wobble uridine modifications are part of the core cellular translation machinery ( 64 , 65 ). Interestingly, a decrease of 2-thiolation le v els at high temperatures is not observed in all organisms. While Saccharomyces bayanus and japonica rice show low le v els of mcm 5 s 2 U similar to baker's yeast, this is not the case in C . albicans , C . glabrata or indica rice (Supplementary Figure S5A) ( 30 , 63 ). This difference might stem from variations in the metabolic wiring of the different species or the enzymatic properties of Urm1-pathway components. In baker's yeast, Urm1 pathway members are less abundant at high temperatures ( 50 , 51 ), likely because of their decreased stability ( 30 , 50 , 51 ). We show that Ncs6 le v els depend on the stability of Ncs2 (Figure 3 F). Howe v er, it is unclear whether this effect similarly regulates the abundance of proteins like Uba4.
The 2-thiolation machinery is integrated into cellular sulphur flux es. Inter estingly, these findings support a model where mcm 5 s 2 U le v els are used by the cells to regulate global translation ( 66 , 67 ). Linking 2-thiolation le v els to metabolic states is intriguing, since this mechanism can explain how tRNA modifications regulate global translation in response to the metabolic r equir ements of the cell (68)(69)(70). Ne v ertheless, the regulation of mRNA translation by tRN A modifications acts slowl y due to the high stability of tRNA and the absence of demodifying enzymes for most tRNA modifica tions. Likewise, 2-thiola tion levels respond to high temperatures within se v eral hours in baker's yeast ( 30 ) most likely due to decreased enzyme activity at high temperatures. The fact that Candida strains and baker's yeast respond differently to high temperatures suggests that different organisms have found different solutions to their specific metabolic and environmental challenges.
Pathogenicity is a complex phenotype combining cellautonomous mechanisms with host-pathogen interactions. In addition to high-temperature growth, the ability to adhere to a substrate is a key mechanism of pathogenicity in C. albicans ( 71 ). Interestingly, NCS2 and NCS6 are upregulated during biofilm formation in Candida ( 72 ). Furthermor e, genes r equir ed for wobble-uridine modifications have been identified in screens for modifiers of colon y f ormation and have been shown to exhibit morphology phenotypes at the single cell le v el in baker's yeast ( 73 , 74 ). Consistent with this, w e show ed tha t a lack of U 34 modifica tion leads to morphology defects in C. albicans and a decreased ability to kill epithelial cells in cultur e (Figur e 5 C-E and Supplementary Figure S5C).
Wobble-uridine modifications tune codon-translation rates and are critical to ensure protein homeostasis ( 12 , 13 , 59 ). Consistent with this, we found that insoluble proteins were abundant in wild-type yeast grown at 37 • C while the le v els of such protein aggr egates wer e significantly lower in NCS2 * yeast, consistent with findings in other tRN A modification m utants ( 13 , 59 , 74 ). The accum ulation of protein aggr egates r educes cellular fitness due to the loss of functional pr oteins, pr oteotoxic stress, the inability to clear misfolded proteins, and perturbations of amino acid homeostasis (75)(76)(77). This reduces the ability of cells to adapt to challenging environments. Upon infection, oxidati v e stress imposed by the host immune system and a high bod y tempera ture will perturb protein folding. High le v els of tRNA modifications optimize translation and likely benefit the pathogen through more efficient translation. Protein homeostasis defects will not only affect the cytoplasmic proteome but also secreted cell-wall proteins, consistent with the observ ed sensiti vity to CFW (Supplementary Figure S1B) as well as morphology phenotypes and a downregulation of cell-wall specific GO terms ( Compared to baker's yeast, C. albicans appeared less affected by the absence of mcm 5 s 2 U. While the deletion of NCS2 impair ed high-temperatur e growth in S. cerevisiae , this was not the case for C. albicans . Interestingly, C . albicans has undergone a reassignment of the CTG codon during evolution leading to variable levels of amino acid misincorporation and a stochastical proteome ( 78 , 79 ). The cells ar e, ther efor e, mor e adapted to remedy protein-homeostasis defects also for the mistranslation of other codons ( 80 ).
Ne v ertheless, we observ ed a perturbation of codon-specific translation, colony-morphology defects, and decreased virulence in hypomodified C. albicans and S. cerevisiae . Our work suggests that targeting tRNA modification enzymes to reduce modification le v els is a promising strategy to treat pathogenic yeast infections. This approach is specific and acts orthogonal to current antifungal drugs (Supplementary Figure S5B). It is therefore an exciting prospect that drugging of tRNA modifications may gi v e rise to nov el antifungals.

DA T A A V AILABILITY
All oligonucleotides and plasmids used in this study are listed in Supplementary Table S3 and Supplementary table S4. The description of the algorithms applied are found in the methods section. Source codes for codon-occupancy calculations are available at ( https: //github.com/LeidelLab/Codon occupancy cal , permanent DOI: 10.5281 / zenodo.8054266). The deep sequencing data from C. albicans ribosome profiling experiments are available in the Gene Expression Omnibus database under accession code GSE199422 except for wild-type C. albicans at 30 • C, which is available under accession code GSE136940.