Insights into the role of endonuclease V in RNA metabolism in Trypanosoma brucei

Inosine may arise in DNA as a result of oxidative deamination of adenine or misincorporation of deoxyinosine triphosphate during replication. On the other hand, the occurrence of inosine in RNA is considered a normal and essential modification induced by specific adenosine deaminases acting on mRNA and tRNA. In prokaryotes, endonuclease V (EndoV) can recognize and cleave inosine-containing DNA. In contrast, mammalian EndoVs preferentially cleave inosine-containing RNA, suggesting a role in RNA metabolism for the eukaryotic members of this protein family. We have performed a biochemical characterization of EndoV from the protozoan parasite Trypanosoma brucei. In vitro, TbEndoV efficiently processes single-stranded RNA oligonucleotides with inosine, including A to I-edited tRNA-like substrates but exhibits weak activity over DNA, except when a ribonucleotide is placed 3′ to the inosine. Immunolocalization studies performed in procyclic forms indicate that TbEndoV is mainly cytosolic yet upon nutritional stress it redistributes and accumulates in stress granules colocalizing with the DEAD-box helicase TbDhh1. RNAi-mediated depletion of TbEndoV results in moderate growth defects in procyclic cells while the two EndoV alleles could be readily knocked out in bloodstream forms. Taken together, these observations suggest an important role of TbEndoV in RNA metabolism in procyclic forms of the parasite.

equal efficiency 19 . Similarly, the enzyme of the Pyrococcus furiosus archaeon shows strong activity toward RNA substrates as well as DNA substrates 20 . In contrast, mouse and human EndoVs exhibit very weak activity over deoxyinosine and DNA substrates in general 21,22 whereas they incise inosine-containing RNA very efficiently, suggesting a role in RNA metabolism instead of DNA repair for the mammalian enzymes 19,23,24 .
Trypanosoma brucei is a protozoan parasite belonging to the Kinetoplastida order and the causative agent of African trypanosomiasis and nagana in humans and animals, respectively 25 . During its life cycle, T. brucei alternates between an insect and a mammalian host and the adaptation to the environmental conditions requires significant structural and physiological changes. During infection, an essential component of the primary immune response is the production of nitric oxide (NO) by activated phagocytes that generates reactive oxygen and nitrogen species (ROS and RNS). However, the impact of DNA repair mechanisms such as endonuclease V to counteract host-generated oxidative and nitrosative stress has not been yet investigated in these human pathogens and may be relevant to understand their capacity to survive in such adverse environment. Here, we have purified and characterized the catalytic properties of the endonuclease V encoded by Trypanosoma brucei genome (TbEndoV). In vitro, TbEndoV efficiently processes inosine-containing single-stranded RNA oligonucleotides, including A to I-edited tRNA-like substrates. On the contrary, a weak activity was observed over inosine in DNA, except when a ribonucleotide was placed 3′ to the lesion. Analysis by immunofluorescence microscopy indicates that TbEndoV localizes predominantly to the cytosol. The enzyme appears to be non-essential in bloodstream forms (mammalian-stage parasites) where the two EndoV alleles could be readily knocked out. In contrast, protein depletion in procyclic forms (insect-stage parasites) by RNA interference led to impaired growth and defects in cell cycle progression, thus suggesting a specific and vital role for TbEndoV at this life stage.

Identification of a putative endonuclease V in Trypanosoma brucei. Homology searching of the
Trypanosoma brucei brucei genome database, using the human endonuclease V ortholog as the query sequence, identified a protein annotated as a putative endonuclease V (Tb927. 10.6860) composed of 316 amino acids and a calculated mass of 34.9 kDa. The T. brucei EndoV protein sequence was aligned with other six characterized orthologs from prokaryotic and eukaryotic organisms (Fig. 1). Three amino acids identified as crucial for endonuclease activity in Thermotoga maritima, Asp43, Glu89 and Asp110, are conserved in TbEndoV. These residues Figure 1. Sequence alignment of endonuclease V orthologs from several organisms including T. brucei. The alignment was made in ClustalOmega and visualized with JalView. Amino acids are colored using the clustal color scheme (for more details check the following web page: http://www.jalview.org/help/html/colourSchemes/ clustal.html) and the identity threshold was set at 50%. The amino acids involved in metal cofactor interaction are indicated by red asterisks, the amino acids responsible for substrate-enzyme interaction are indicated by green asterisks and the amino acids involved in product-enzyme interaction are marked with blue asterisks. Protein sequences were retrieved from NCBI RefSeq: T_brucei, Trypanosoma brucei (XP_822927.1); H_sapiens, Homo sapiens (NP_775898.2); M_musculus, Mus musculus (NP_001158108.1); P_furiosus, Pyrococcus furiosus (AFN03782.1); B_subtilis, Bacillus subtilis (NP_391498.2); E._coli, Escherichia coli (NP_418426.2); T_maritima, Thermotoga maritima (NP_229661.1).
coordinate the interaction with the metal cofactor which is essential for catalysis 26 . Sequence comparison revealed that, TbEndoV exhibits a high similarity to the human ortholog (identity = 37%; E = 3 × 10 −41 ; bits = 143).
T. brucei endonuclease V preferentially cleaves inosine-containing RNA. To characterize the enzymatic properties of TbEndoV, the corresponding open reading frame was amplified by PCR and cloned into an E. coli vector that allows expression of the protein fused to the maltose-binding protein in order to increase its expression and solubility. The MBP-TbEndoV protein was overexpressed in the E. coli TB1 strain and purified to over 99% purity by amylose affinity and gel filtration chromatography.
The purified recombinant protein was initially assayed for endonuclease activity on 5′-radiolabeled 21-nucleotide single-and double-strand DNA substrates in which one strand contained hypoxanthine or uracil in a central position. Oligonucleotide sequences can be found in Fig. S1. Double-strand substrates were designed to produce Hx:T (dsDNA dI:dT), Hx:C (dsDNA dI:dC), U:G (dsDNA dU:dG) and U:A (dsDNA dU:dA) base pairs that mimic potentially physiological DNA substrates. The dsDNA dI:dT and dsDNA dU:dG substrates may result from adenine and cytosine deamination at A:T and C:G base pairs respectively while dsDNA dI:dC and dsDNA dU:dA can be generated during replicative synthesis of hypoxanthine-or uracil-containing DNA strands, respectively. In general, TbEndoV showed poor or no endonuclease activity on these DNA substrates ( Fig. 2a-f). Indeed, ssDNA dI and dsDNA dI:dT were only cleaved at very high enzyme concentrations while practically no cleavage was observed with dsDNA dI:dC (Fig. 2a-c). The trypanosomal enzyme also exhibits residual activity on uracil in ssDNA substrates while no activity was detected over the uracil-containing dsDNA substrates (Fig. 2d-f). To rule out a possible contamination with bacterial endonuclease V during purification, we generated a D95A catalytic mutant by site-directed mutagenesis. The aspartate at this position is a highly conserved residue responsible for the coordination of the metal cofactor during substrate cleavage 27 . As shown in Fig. 2, replacement of aspartate 95 by alanine fully abolished TbEndoV activity in agreement with studies on homologous enzymes 19,28,29 . Hence, the activity detected is intrinsic to TbEndoV.

Figure 2.
TbEndoV acts efficiently upon inosine-containing ssRNA. DNA substrates with (a-c) inosine or (df) uracil, and (g) single and (h,i) double-strand RNA substrates or (j,k) RNA-DNA hybrids were incubated with increasing amounts of TbEndoV (0-2000 nM, as indicated). (l) Plot of the percentage of substrate cleaved by endonuclease V activity versus the total amount of protein used in the reaction. Each point represents the mean (±s.d.) from three independent experiments. The D95A catalytic mutant was used as negative control. Reaction products were analyzed by 20% denaturing PAGE. Image in j was cropped and edited for easier comparison. Positions of the reaction substrate (S) and product (P) are indicated.
While bacterial EndoV cleaves both DNA and RNA substrates, the human enzyme has been reported to act more efficiently upon inosine-containing RNA substrates 19,23 . Indeed, riboinosine can arise by spontaneous or chemically-induced adenine deamination, as a consequence of enzymatic RNA-editing or by erroneous incorporation of ITP during transcription 30 . We therefore tested the activity of TbEndoV using different RNA substrates (Fig. 2). As shown in Fig. 2g, trypanosomal EndoV efficiently cleaves single-strand RNA substrates with inosine (ssRNA rI) but is not able to process double-strand RNA when Hx pairs with U (dsRNA rI:rU) or cytosine (dsRNA rI:rC) (Fig. 2h,i). As expected, the incubation of the RNA substrates with the D95A catalytic mutant did not result in product formation.
We have also considered the possibility that riboinosine could be incorporated into nascent RNA during transcription giving rise to a transient RNA:DNA hybrid where Hx would establish a stable base pair with C. However, the cleavage assay did not reveal significant endonuclease activity using the dsRNA:DNA rI:dC substrate and only minor incision was detected with a similar substrate containing an Hx:T mispair (dsRNA:DNA rI:dT) (Fig. 2j,k).
Once ssRNA rI was established as the preferred substrate for TbEndoV (Fig. 2l), reaction conditions were empirically varied to determine how they affected its incision activity (Fig. S2). We found that TbEndoV acts more efficiently at pH 7.5 when using 1 mM MnCl 2 , the preferred metal cofactor (Fig. S2a,c). On the other hand, addition of high concentrations of NaCl or KCl (≥100 mM) greatly inhibited TbEndoV activity and therefore, salts were not included in the final reaction buffer (Fig. S2b). We also noticed that the ssRNA rI substrate was refractory to undergo complete cleavage by TbEndoV even at high enzyme concentrations and under different reaction conditions tested. In contrast, total incision was achieved with an alternative substrate sequence, indicating that substrate recognition and/or incision is sequence-dependent and activity could be influenced by the secondary structure of the substrate (Fig. S2d).
Analysis of the substrate determinants for the inosine recognition and cleavage. Both mammalian and prokaryotic EndoVs recognize inosine and cleave at the second phosphodiester bond 3′ to the lesion 14,19,20 . However, while human and mouse enzymes act only upon RNA, bacterial EndoV operates efficiently on DNA and RNA. The strong preference of trypanosomal EndoV for inosine in RNA suggests that the 2′ hydroxyl group of the ribose is required for an efficient catalysis. Here, we have investigated the effect of specific substrate determinants on TbEndoV catalytic function. A set of DNA oligonucleotides was designed to examine how the presence of a ribose either at the lesion site or at the 3′-adjacent nucleotide affected the incision of in situ deamination products. As shown in Fig. 3, DNA substrates (ssDNA rIdG and dsDNA rIdG:dTdC) containing a riboinosine but followed by a deoxyribose are not efficiently cleaved by TbEndoV (Fig. 3a,b,e). However, when a ribonucleotide is placed at the 3′ position of a deoxyinosine lesion, TbEndoV incision activity is greatly enhanced, regardless the DNA substrate is single (ssDNA dIrG) or double-stranded (dsDNA dIrG:dTdC) (Fig. 3c,d,e).
TbEndoV acts upon tRNA substrates containing inosine. RNA-editing is a potential source of inosine in trypanosomes since ADAT proteins have been characterized in these organisms 4,31 . In particular, threonine coding tRNA showed high editing levels at positions 32 (C-to-U editing) and 34 (A-to-I editing) in T. brucei 4

.
To determine whether this could be a biologically relevant RNA substrate, we assayed the TbEndoV incision activity over non-edited, single-and double-edited tRNA-mimicking substrates (Fig. 4). As expected, TbEndoV did not cleave the non-edited (Fig. 4a,e) or the C to U edited substrates with adenine at position 34 (Fig. 4c,e). In contrast, the tRNA containing Hx at position 34 (C 32 I 34 ) was efficiently processed in vitro (Fig. 4b,e) and could be a potential cellular target for endonuclease V in T. brucei. Notably, the additional presence of U at position 32 had a negative impact on the endonuclease ability to cleave this tRNA substrate (U 32 I 34 ), suggesting that cytosine deamination at position 32 may act as an inhibitory mechanism to prevent the processing of A to I-edited tRNAs by EndoV. Subcellular localization of TbEndoV. Endonuclease V has been localized to the cytosol of human cells 23 but also associated to the nucleoli 32 . In order to determine the subcellular localization of the parasite enzyme, specific rabbit polyclonal antibodies against TbEndoV were raised using soluble recombinant His-TbEndoV as antigen. To ensure higher specificity, antibodies were purified from blood serum by affinity chromatography against the same pure recombinant protein. Immunoblotting analysis failed to detect EndoV protein in procyclic and bloodstream whole cell extracts, even though the antibody does react efficiently with twenty nanograms of purified recombinant protein (Fig. 5a), suggesting that EndoV expression levels are extremely low in trypanosomes. To overcome this limitation for subcellular localization studies, we generated procyclic T. brucei cell lines stably overexpressing TbEndoV (PF/TbEndoV) or a C-terminal Myc-tagged version of TbEndoV (PF/TbEndoV-myc). Protein levels were monitored using the purified anti-TbEndoV and monoclonal anti-myc antibodies, respectively (Fig. 5b,c). Both, polyclonal and monoclonal antibodies, recognize a single and highly intense band in western blots. In spite of the high levels of protein synthesis, TbEndoV overexpression did not appear to have any impact on proliferation and therefore these transgenic cells were considered suitable for cellular localization studies.
Immunofluorescence microscopy images obtained from the parental procyclic cell line (PF) (Fig. 5d, upper row) and the TbEndoV-overexpressing cell line (middle row) revealed a granular staining pattern with predominant localization in the cytosol and at a much lesser extent in the nucleus. The same distribution pattern was verified in an independent experiment using a monoclonal anti-myc antibody on PF/TbEndoV-myc cells (Fig. 5d, lower row). These results demonstrate that TbEndoV performs an important function in the cytoplasm although a certain role in the nucleus cannot be excluded. Our analysis provides the first experimental evidence of the subcellular localization of a eukaryotic EndoV under endogenous expression conditions, avoiding this way potential overexpression-induced artifacts.
TbEndoV depletion impairs proliferation of T. brucei procyclic forms. In order to assess whether T.
brucei requires the EndoV activity for normal proliferation and survival, we proceeded to generate EndoV-null mutants in bloodstream and procyclic forms of the parasite. For bloodstream forms, replacement cassettes contained a puromycin-resistance gene (PAC) and a phleomycin-resistance gene (BLE), flanked by 550 bp of the EndoV 5′-UTR and 3′-UTR sequences (Fig. S3). All DNA fragments were obtained by PCR amplification of T. brucei 427 genomic DNA and primers were designed from the sequence obtained from the GeneDB database. Linear targeting fragments containing each of the inactivation cassettes were initially transfected into T. brucei bloodstream strain yielding recombinant clones, which were efficiently isolated in selective media with puromycin or phleomycin. One of the puromycin-resistant clones was subjected to a second round of inactivation with the BLE cassette obtaining this way transfectants resistant to puromycin and phleomycin. Integration of the PAC or BLE markers was monitored by PCR using primers that hybridize with the flanking genes upstream and downstream from the TbEndoV gene in combination with marker-specific primers (Fig. S3). As expected, no expression of TbEndoV mRNA was observed by RT-PCR in the selected clone (Fig. 6a). Despite lacking TbEndoV, double knockout cells exhibited similar proliferation rate as parental cells (Fig. 6b), providing direct evidence that the enzymatic activity of TbEndoV is not essential for viability or proliferation of bloodstream trypanosomes in vitro.
Inactivation of TbEndoV by homologous recombination was also attempted in procyclic T. brucei cells; however, two-allele replacement could not be achieved after several attempts. We therefore used an alternative knock-down approach that allows monitoring the cellular events that follow the reduction of EndoV activity. We therefore proceeded to silence TbEndoV expression using the RNA interference (RNAi) methodology. Due to the low constitutive expression of TbEndoV, we monitored the inhibition of expression using qRT-PCR. As shown in Fig. 6c, TbEndoV could be successfully depleted in doxycycline-induced RNAi-transfected cells. Determination of the TbEndoV mRNA levels at two, four and six days after doxycycline addition revealed a decrease of 25%, 60% and 49% respectively, in comparison to the parental strain. A time course analysis of parasite growth was carried out during 16 days to determine the effect of TbEndoV depletion. Cells where TbEndoV was down-regulated exhibited a moderate but consistent defect in proliferation with respect to non-induced or parental control cells (Fig. 6d).
EndoV-deficient cells were further analyzed by quantifying the DNA content by flow cytometry together with DAPI staining of nuclei and kinetoplasts. FACS analysis revealed a significant increase of EndoV-deficient trypanosomes in the S-phase (Fig. 6e), which is consistent with the accumulation of cells with an enlarged kinetoplast (1N1K*) observed in cells that have already initiated nuclear DNA duplication (Fig. 6f,g). At the same time, cells in which EndoV was knocked-down also displayed a decrease in 1N1K (corresponding to cells in G1) and a concomitant increase in the population of 1N2K cells (G2 phase) (Fig. 6e,f). The major anomaly was the presence of cells with one kinetoplast but without nucleus (0N1K), also known as zoids (Fig. 6f,g), that may account for the sub-G1 population detected by flow cytometry. Similar cell phenotypes have been described in procyclic T. brucei cells after treatment with aphidicolin or bleomycin where the inhibition of nuclear S phase was followed by defective mitosis 33,34 . TbEndoV localizes to stress granules after starvation. Human EndoV has been shown to relocate to cytoplasmic RNA-containing granules in response to cellular stress caused by sodium arsenite where it could be contributing to the elimination of damaged inosine-containing transcripts 24 . Stress granules (SGs) are transient aggregates of stalled translation initiation complexes whose function is to preserve the mRNAs and prevent their translation until the normal conditions are restored. Oxidative stress, glucose deprivation, heat-shock, viral infections etc. are demonstrated inducers of stress granule formation 35 . In trypanosomes, SGs have been observed in the procyclic forms that have to cope with periods of starvation and temperature changes inside the insect 36 . To address whether TbEndoV might be relocalizing in SGs after starvation-induced stress, cells were cultured in carbon-free buffer (PBS) for different time periods and subsequently monitored by fluorescence microscopy (Fig. 7a). After 120 min in PBS, endogenous TbEndoV protein accumulated into discrete foci which were bigger in size than the granule particles observed in cells grown in normal rich medium (untreated) (Fig. 7c). Under the same conditions, the DEAD-box RNA helicase TbDhh1, a well-characterized component of starvation SGs 37 , also relocates into foci although with different kinetics, thus validating the protocol used to induce starvation (Fig. 7a). To validate the specificity of the signal detected, cells where TbEndoV expression was silenced by RNAi were analyzed by immunofluorescence microscopy (Fig. 7b). Quantitative analysis revealed that trypanosomes expressing low levels of TbEndoV exhibited a significant reduction in the number and size of TbEndoV-positive granules after 120 min in the absence of a carbon source thus confirming the specific accumulation of this endonuclease in starvation-induced granules (Fig. 7c).
We next proceeded to establish whether TbEndoV does actually colocalize with the RNA-binding protein TbDhh1 in SGs upon starvation. To perform the colocalization experiments, the anti-TbEndoV antibody was non-covalently labeled with Alexa Fluor ® 594 while the polyclonal anti-TbDhh1 was detected with an Alexa Fluor ® 488-conjugated secondary anti-rabbit antibody. Immunostaining showed extensive colocalization of TbEndoV and TbDhh1granules at 120 minutes after induction of starvation (Fig. 7d). Pearson's correlation and Manders' overlap coefficients were determined to establish the degree of colocalization. Manders' coefficient indicate that 74% of the signal corresponding to TbDhh1 overlaps with TbEndoV (M1 = 0.74 ± 0.13) and vice versa, 70% of TbEndoV fluorescence ovelaps with TbDhh1 (M2 = 0.70 ± 0.11). The colocalization is also supported by the Pearson's correlation coefficient (0.87 ± 0.04). These data demonstrate the presence of TbEndoV in SGs and a role for this protein in the stress response to starvation.

Discussion
Trypanosoma brucei is an extracellular parasite under constant exposure to oxidative and nitrosative stress either generated as a byproduct of its own metabolism or produced by immune effector cells in response to parasite infection 38,39 . It is well established that ROS and RNS can damage vital molecules, such as proteins, lipids, or nucleic acids 40 . Therefore, protozoan survival and genomic stability greatly depends on DNA repair systems, which for that reason could be excellent targets for chemotherapeutic interventions 41 .
Deaminated DNA bases such as uracil and hypoxanthine arise as result of spontaneous hydrolysis or induced by oxidative or nitrosative agents. These lesions are the substrate of specialized DNA glycosylases as part of the BER system, which carries out the correction of DNA alterations that do not distort the overall structure of DNA helix. T. brucei expresses a single uracil-DNA glycosylase activity (UNG) 42 responsible for excision of U but no glycosylases with capacity to excise Hx, generally an alkylpurine-DNA glycosylase 43 , have been identified in kinetoplastid parasites 44 . While in bacteria, EndoV provides a back-up mechanism for the repair of Hx and U 45 , trypanosomal EndoV is not endowed with significant activity over DNA substrates and therefore it is unlikely that this enzyme performs an important role in DNA repair. Nevertheless, we cannot exclude the possibility that neglected factors in our in vitro cleavage assay may enhance the enzyme activity over DNA in vivo. Similarly to mammalian EndoVs which have been associated to a possible role in RNA processing 19,23 , TbEndoV exhibits strict specificity for inosine in RNA suggesting that this might be the potential bona fide substrate in physiological conditions. Several potential biological targets can be anticipated for TbEndoV. Deamination of adenine can occur in RNA and their precursors via hydrolytic and nitrosative reactions upon exposure to endogenous or exogenous agents 46 . T. brucei is exposed to RNS in the early steps of an infection, after nitric oxide (NO) and ROS are released by phagocytes as part of the primary immune response 47 . RNS can react with nucleic acids generating strand breaks and base modifications including deaminations 48 , which constitute the primary cause of NO-induced mutagenicity in a variety of organisms 49,50 . The nucleotide pool may also undergo base deaminations under the same circumstances leading to the generation of non-canonical ITP that could be erroneously incorporated by the RNA polymerase during transcription. Supporting this notion, E. coli and yeast cells defective in inosine triphosphate pyrophosphatase (ITPase) exhibit increased levels of inosine in their RNA 51 . The presence of Hx in mRNAs could have mutagenic consequences during gene expression since it is recognized as guanine during protein translation 5 . Likewise, the presence of inosine in tRNAs and rRNAs might modulate or interfere with protein translation as described for other modifications 52 . Recognition and cleavage by EndoV endonuclease activity of adenine-deaminated transcripts would prevent the formation of mutant proteins thus acting as a mechanism to eliminate potentially harmful mRNAs.
Another source of inosine in RNA relies on the enzymatic activity of specific adenosine deaminases acting on mRNA (ADAR) and tRNA (ADAT) 1 . These enzymes play an important role introducing inosine in specific positions of RNA, a key process for correct splicing and codon diversity 53 . While no ADAR protein has been identified in T. brucei, adenine deamination in tRNAs by ADAT enzymes has been extensively described 4,31 . In particular, a heterodimer made up of two subunits (ADAT2 and ADAT3) is responsible for specific A to Hx deamination 3 in T. brucei. Eight different tRNA species (for Ala, Arg, Ile, Leu, Pro, Ser, Thr and Val amino acids) have adenine at the wobble position of the anticodon and therefore, potential A-to-I editing sites. Indeed, threonine coding tRNA is highly edited at position 34 (A-to-I) but also at position 32 (C-to-U) in T. brucei 4 . In vitro, TbEndoV cleaves the trypanosomal A-to-I edited tRNA substrates with high efficiency, suggesting that the turnover of edited tRNAs may be initiated by this inosine-specific endonuclease activity. The presence of U at position 32 was somewhat detrimental for activity, which suggests that EndoV-mediated cleavage may be modulated by editing events.
Beyond a potential role in tRNA degradation, TbEndoV ability to cleave tRNAs is reminiscent of the endonucleolytic activity of yeast Rny1 and mammalian ribonuclease angiogenin. These nucleases carry out the incision of tRNAs within the anticodon loop generally as part of a stress response but in this case, tRNA cleavage is not intended to promote tRNA degradation since full-length tRNA levels are not substantially affected. Instead, tRNA-derived small RNAs themselves may function to inhibit translation or promote the degradation or repression of specific mRNAs under particular physiological conditions 54,55 . The presence of half-tRNA molecules have been detected in Tetrahymena, Aspergillus and Trypanosoma cruzi following nutritional stress [56][57][58] while Saccharomyces, Arabidopsis and human cells accumulate cleaved tRNAs in response to oxidative stress 59 . Interestingly, T. cruzi secretes extracellular vesicles containing tRNA halves that may reach host mammalian cells and alter the expression of genes involved in the immune response in order to facilitate the infection 60 . Although the form of tRNA processing is essentially different from the more specific endoV-mediated cleavage of A-to-I edited tRNAs, it raises the question as to whether cleaved tRNAs generated by the action of TbEndoV might perform analogous functions and participate in non-canonical regulatory pathways.
Subcellular localization analysis revealed that TbEndoV is primarily a cytosolic enzyme. Likewise, human EndoV has been localized to the cytosol but also associated to the nucleoli, suggesting an involvement in ribosomal RNA metabolism 23,32 . Furthermore, arsenite-induced oxidative stress induces the accumulation of hEndoV in cytoplasmic granules where it colocalizes with polyadenylate-binding protein C1 (PABPC1) 24 . In procyclic trypanosomes, TbEndoV colocalizes with TbDhh1 in cytoplasmic stress granules upon starvation. Similarly to sodium arsenite, glucose deprivation also induces ROS production which in turn may cause harmful base deaminations in mRNAs 61 . While SGs are mainly involved in storage and regulation of translation of mRNAs, Pbodies contain mRNAs targeted for degradation as well as factors involved in mRNA turnover. These granules are not hermetic entities neither in mammalian cells, where SGs may transiently establish physical and functional interactions with P bodies 35 , nor in T. brucei where SGs and P bodies share several of their components including TbDhh1 62 . Likewise, it has been suggested that in T. brucei, some of the stress granules might arise from preexisting P bodies as described for yeast 63 . Although whether TbEndoV is a constitutive component of P bodies remains unclear, its presence in starvation-induced SGs suggest that it could be performing quality control of deaminated mRNAs and regulating the transfer of damaged transcripts to adjacent P bodies for their elimination.
We have shown that suppression of TbEndoV activity impairs the proliferation in vitro of procyclic parasites whereas bloodstream trypanosomes can still grow normally. While it has been reported that around 6% of the genes are differentially expressed between procyclic and bloodstream forms of T. brucei 64 , TbEndoV appears to be equally expressed in both stages of the parasite 64, 65 and thus different enzyme requirements would not explain the obtained phenotype. It is possible that TbEndoV is involved in the homeostasis of specifically edited RNAs or establishes interactions with procyclic-specific proteins. Another potential factor accounting for a distinct role of TbEndoV in procyclics may relate to the major differences in mitochondrial metabolism between mammalian and insect-stage parasites. The identification of specific cellular substrates of the enzyme and its role in the occurrence of inosine-containing RNA will contribute to the understanding of the relevance of TbEndoV in parasite survival.
In summary, the data presented in this work show that TbEndoV exhibits incision activity preferentially upon inosine-containing RNA and appears to be needed for normal proliferation of procyclic forms where it could play a critical role in the metabolism of RNAs that are specifically required or expressed in this life stage of the parasite.

Methods
Expression and purification of wild-type and mutant endonuclease V. The endonuclease V coding sequence (Tb927.10.6860) was amplified from T. brucei brucei (Lister 427) genomic DNA using specific primers 5′-GCG AAT TCA TGG GTG TGG GAG AAA TGA GT-3′ and 5′-GCG GAT CCT CAG CAT TCA TTT AGT TGA ATC-3′. The PCR product was cloned into pMAL-c2X vector that allows protein expression in fusion to Maltose-Binding Protein (MBP) at the N-terminus (New England Biolabs). A TbEndoV-D95A mutant sequence was generated using QuikChange Lightning Multi Site-Directed Mutagenesis Kit (Agilent Technologies) with the following primers: 5′-C CAC GTC GGC GGC GTT GCC ATA TCC TTT ATT CAG G-3′ and 5′-C CTG AAT AAA GGA TAT GGC AAC GCC GCC GAC GTG G-3′ (underlined characters indicate the codon change). MBP-EndoV proteins were purified by amylose affinity (New England BioLabs) followed by gel filtration chromatography. Fractions containing MBP-EndoV were pooled, dialyzed against storage buffer [10 mM Tris-HCl, pH 7.5, 200 mM NaCl, 1 mM DTT and 20% glycerol] and stored at −80 °C. box represents the interquartile range (IQR) and whiskers are maximum and minimum values. Differences in granule areas are statistically significant between untreated and starvation-induced PF cells (p = 0.002) and between PF and TbEndoV-RNAi cells in normal growth conditions or under starvation (p = 0.03). Unpaired two-tailed Student's t-test with Welch's correction; *p < 0.05 **p < 0.01. (d) Colocalization of TbEndoV and TbDhh1 in starvation stress granules. Starvation stress was induced by maintaining the trypanosomes in PBS for 120 min prior to fixation. Cells were then processed for double immunofluorescence staining using the following pairs of primary and secondary antibodies: Alexa Fluor ® 594-conjugated rabbit anti-TbEndoV and rabbit anti-TbDhh1/Alexa Fluor ® 488-conjugated goat anti-rabbit IgG. Untreated (no exposed to PBS) cells were used as control. Nuclear DNA and the kinetoplast were visualized by DAPI staining. The images were representative of each staining from at least two independent experiments.
In a standard reaction (

Antibody generation.
To obtain a polyclonal anti-TbEndoV antibody, a rabbit was immunized using soluble purified recombinant TbEndoV. Before injecting the antigen into the rabbit, 450 µg of pure protein diluted in PBS was mixed with Freund's adjuvant (1:1 ratio). Antibody-containing serum was collected and affinity-purified using pure recombinant protein coupled to Affi-Gel ® 15 resin (BioRad) following the manufacturer's instructions.
For the RNA interference construct, a region of TbEndoV of 400 bp (position 29-428 of the coding region) was amplified using the following set of primers: (forward) 5′-GCG GAT CCA AGC TTC CGC GAT AAA GTT GGA CAT G-3′ and (reverse) 5′-AAC GGG CCC GCC TCC CGG AAC GCG AGG-3′. The amplification product was cloned in pGR19 69 as previously described 70 .
To overexpress native TbEndoV, the coding sequence was amplified with primers 5′-GCC ATA TGG GTG TGG GAG AAA TGA GTG-3′ and 5′-GCG GAT CCT CAG CAT TCA TTT AGT TGA ATC-3′ including NdeI and BamHI restriction sites (underlined). The PCR product was cloned into pGRV23b expression vector 42 . To generate a construct for the expression of TbEndoV fused to a C-terminal myc tag, the coding sequence lacking the stop codon was amplified using primers 5′-GCC ATA TGG GTG TGG GAG AAA TGA GTG-3′ and 5′-GCG TTA ACG CAT TCA TTT AGT TGA ATC CGT CGG A-3′ including NdeI and HpaI restriction sites (underlined). The PCR product was cloned into pGRV33 expression vector 42 .
Immunofluorescence studies. Immunofluorescence (IF) studies were performed as previously described 70 . Briefly, parasites were harvested, washed twice in PBS (137 mM NaCl, 4 mM Na 2 HPO 4 , 1.7 mM KH 2 PO 4 and 2.7 mM KCl), and fixed in 4% PFA in wash solution (PBS 1x, 0.2% Tween ® 20) on a poly-l-lysine coated slide for 30 min. Fixed parasites were washed once in wash solution and permeabilized and blocked with 1% of NP40 in blocking solution (wash solution containing 1% blocking reagent (Roche)) during 75 min. The IF was performed by incubation with either rabbit polyclonal anti-TbEndoV (1:100) or mouse monoclonal anti-c-myc (Sigma, 1:100) primary antibodies in blocking solution. Alexa Fluor ® 488 goat anti-rabbit IgG or anti-mouse IgG were used as secondary antibody (Sigma, 1:500), respectively. Preparations were finally mounted with ProLong ® Gold Antifade Reagent with DAPI (Life Technologies). Vertical stacks of up to 40 slices (0.2 µm steps) were captured using an Olympus wide-field microscope and Cell-R IX81 software. Images were deconvolved and pseudo-colored with Huygens Essential software (version 3.3; Scientific Volume Imaging).
For nutritional stress studies, 25 million log-phase parasites were washed twice in PBS, and starved by incubation in PBS during the indicated time. After starvation, IF was performed by incubation with rabbit polyclonal anti-TbEndoV (1:100) or anti-TbDhh1 (1:100) primary antibodies in blocking solution. Alexa Fluor ® 488 goat anti-rabbit IgG and Alexa Fluor ® 594 goat anti-rabbit IgG were used as secondary antibody (Sigma, 1:500), respectively. To study TbEndoV and TbDhh1 colocalization, anti-TbEndoV antibody was labeled with Alexa Fluor ® 594 using Zenon ™ Rabbit IgG Labeling Kit (ThermoFisher) following manufacturer's instructions. After incubation with both the anti-TbDhh1 primary antibody and Alexa Fluor ® 488 goat anti-rabbit IgG, samples were incubated with an excess of a nonspecific primary antibody to block the secondary antibody. Preparations were then incubated with the Alexa Fluor ® 594-conjugated anti-TbEndoV (1:10) during 1 hour and fixed in 4% PFA in PBS for 15 minutes.
All images were analyzed using Fiji/ImageJ software. Degree of colocalization was determined using Pearson's correlation coefficient and Manders' overlap coefficient. Both coefficients were calculated with JACoP plugin, using a manual threshold for Manders' coefficient, and given values are the mean of at least 18 cells. Overlapping of TbDhh1 over TbEndoV is represented by M1, while overlapping of TbEndoV over TbDhh1 is represented by M2. For the determination of the number and size of granule particles, a threshold of 0.05 microns 2 and 35.000 of signal intensity was manually set over Z-projection cell images (using the sum slice method).
Fluorescence-activated cell sorting (FACS) analysis and quantification of nuclei and kinetoplasts. PF cells (10 7 cells per sample) were harvested by centrifugation and fixed with ethanol. FACS analysis with propidium iodide staining was carried out as previously described 70 . To determine the number of nuclei and kinetoplasts per cell, trypanosomes (10 6 cells per sample) were harvested by centrifugation, fixed with paraformaldehyde and mounted with ProLong ® Gold Antifade Reagent with DAPI (Life Technologies). A Zeiss Axiophot microscope (Carl Zeiss, Inc.) was used to carry out microscopy and digital image acquisition. The quantification analysis was performed in individual cells from three independent experiments (n > 300).
Data availability. No datasets were generated or analyzed during the current study.