The RNA-binding protein RBP42 regulates cellular energy metabolism in mammalian-infective Trypanosoma brucei

RNA-binding proteins (RBPs) are key players in coordinated post-transcriptional regulation of functionally related genes, defined as RNA regulons. RNA regulons play particularly critical roles in parasitic trypanosomes, which exhibit unregulated co-transcription of long unrelated gene arrays. In this report, we present a systematic analysis of an essential RBP, RBP42, in the mammalian-infective bloodstream form of African trypanosome and show that RBP42 is a key regulator of parasite’s central carbon and energy metabolism. Using individual-nucleotide resolution UV cross-linking and immunoprecipitation to identify genome-wide RBP42-RNA interactions, we show that RBP42 preferentially binds within the coding region of mRNAs encoding core metabolic enzymes. Global quantitative transcriptomic and proteomic analyses reveal that loss of RBP42 reduces the abundance of target mRNA-encoded proteins, but not target mRNA, suggesting a positive translational regulatory role of RBP42. Significant changes in central carbon metabolic intermediates, following loss of RBP42, further support its critical role in cellular energy metabolism. Trypanosoma brucei infection, transmitted through the bite of blood-feeding tsetse flies, causes deadly diseases in humans and livestock. This disease, if left untreated, is almost always fatal. Existing therapies are toxic and difficult to administer. During T. brucei’s lifecycle in two different host environments, the parasite progresses through distinctive life stages with major morphological and metabolic changes, requiring precise alteration of parasite gene expression program. In the absence of regulated transcription, post-transcriptional processes mediated by RNA-binding proteins play critical roles in T. brucei gene regulation. In this study, we show that the RNA-binding protein RBP42 plays crucial roles in cellular energy metabolic regulation of this important human pathogen. Metabolic dysregulation observed in RBP42 knockdown cells offers a breadth of potential interest to researchers studying parasite biology and can also impact research in general eukaryotic biology.

R NA regulons that coordinately regulate the production of functionally related proteins are emerging as key regulatory modules in eukaryotes (1).In contrast to the prokaryotic "operons" in which functionally related gene clusters are co-transcribed and co-translated, RNA regulons co-regulate gene cohorts post-transcriptionally by dynamic RNA-protein and RNA-RNA interactions to modulate critical regulatory steps, including messenger RNA (mRNA) maturation, localization, translation, and decay (2).Thus, RNA regulons allow rapid, yet precise response to both intra-and extra-cellular signals that trigger readjustment of entire biochemical pathways (3).Accordingly, RNA regulons have been discovered in diverse cell systems performing crucial regulatory functions (4,5).

RBP42 binds within the coding region of transcripts encoding cellular primary metabolic enzymes
To accurately identify RBP42's targets in mammalian-infective slender bloodstream form T. brucei, we performed UV cross-linking and immunoprecipitation (CLIP) by incorporat ing two strategies.First, to eliminate non-specific protein-antibody interactions from our analysis, we used a second independent antibody reagent, a mouse monoclo nal antibody to Ty1-epitope (anti-Ty1), in addition to a rabbit polyclonal antibody to bacteria-derived recombinant RBP42 protein (anti-RBP42) (29).Transgenic cell line (3Ty1-RBP42), producing triple-Ty1 epitope-tagged RBP42 protein, was generated by homologous recombination-mediated in situ tagging of both alleles of the diploid parasite (Fig. 1A).Genomic polymerase chain reaction (PCR) and immunoblot analy ses of two representative clonal cell lines show proper integration and expression Complexes were separated by denaturing sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) and transferred to nitrocellulose membranes.
A PhosphorImage scan of 32 P-labeled protein-RNA complexes purified from both wild-type (WT) cells and 3Ty1-RBP42 cells is shown.In the absence of ribonuclease (RNase I), the protein-RNA complexes are too large, which fail to resolve by gel chromatography, and appear as broad smear.While anti-Ty1 (BB2) antibody immunoprecipitated only the tagged protein, anti-RBP42 antibody immunoprecipitated both WT and tagged proteins.(F) Optimization of RNase I concentration to obtain optimal size RNA fragments for iCLIP library preparation.A representative PhosphorImager scan, similar to panel E, is shown.
Protein-RNA complexes were treated with decreasing concentrations of RNase I before immunoprecipitation.Complexes from the marked gel area (red dotted box) were excised from the nitrocellulose membrane, and extracted RNAs were reverse transcribed to prepare cDNA library.Normal mouse antibody serves as control.
We used WT cells in combination with anti-RBP42 antibody and 3Ty1-RBP42 cells in combination with anti-Ty1 antibody, to identify RBP42's cellular RNA targets (Fig. 1E and  F).Second, since UV-irradiated cross-linking of protein-RNA interactions is known to be very inefficient (35,36), we captured RBP42-RNA interactions at two increasing UV doses, at 150 mJ/cm 2 and at 300 mJ/cm 2 constant energy.We reasoned that authentic RBP42-RNA interactions, as opposed to non-specific, cryptic interactions, would be captured at both conditions.We identified RBP42-RNA interactions by generating and deep-sequencing cDNA libraries from CLIP-purified RNA (Fig. 1F).cDNA libraries were made using individualnucleotide resolution CLIP (iCLIP) method (37).The iCLIP, by incorporating a cDNA circularization step, takes advantage of the frequent termination of reverse transcriptase (RT) at the crosslink site, which therefore corresponds to the nucleotide preceding the cDNA start.Thus, iCLIP enables the position of RBP-RNA interaction to be precisely mapped, and each unique cDNA molecule in the library denotes an individual cross link event.We produced four independent iCLIP libraries: two for each of anti-RBP42 antibody (Samples 1 and 3) and anti-Ty1 antibody (Samples 2 and 4) at both 150 mJ/cm 2 (Samples 1 and 2) and 300 mJ/cm 2 UV doses (Samples 3 and 4) (Fig. 2A).We obtained 29,464 and 26,055 unique cDNA molecules in Samples 1 and 2, respectively, and 867,214 and 493,361 unique cDNA molecules in Samples 3 and 4, respectively (see Fig. S1a at https://doi.org/10.6084/m9.figshare.21737321).The large, ~20-fold more unique cDNA molecules in Samples 3 and 4, compared to Samples 1 and 2, may occur as a batch effect, i.e., minor variations in enzymatic activities and/or reaction conditions.It is also possible that higher, 300 mJ/cm 2 , UV doses in Samples 3 and 4 produced more cross link events (see Fig. S1b at https://doi.org/10.6084/m9.figshare.21737321).Genome-wide comparison of mapped crosslink sites reveals moderate to high degree of correlations (Spearman correlations from 0.40 to 0.94 between different samples) among all four iCLIP libraries (see Fig. S1c at https://doi.org/10.6084/m9.figshare.21737321).Distribution of crosslink sites over annotated genomic features show that the vast majority of RBP42 interactions occur within the mRNA coding sequences (Fig. 2B).This result, in accordance with previously published data in fly-infective procyclic forms (29), confirms that RBP42 predominantly binds within the target mRNA-coding sequences.
We identified RBP42-target transcripts by recording crosslink sites to annotated T. brucei 927 genome (TriTrypDB.org;V45).The number of crosslink events at each recorded crosslink site was calculated by counting the number of stacked unique cDNAs with coinciding start sites (see Materials and Methods) (Fig. 2C and see Fig. S2 at https:// doi.org/10.6084/m9.figshare.21737321).Evaluation by PureCLIP algorithm (38), designed to analyze iCLIP data, identified similar crosslink sites.To identify potential RBP42 crosslink sequence motif, we sampled ~1,100 high-confidence crosslink site sequences, each 11 nucleotides (nt) long, consisting of crosslink site plus 5 nt flanking sequences, and searched for motif discovery using MEME suite (https://meme-suite.org/tools/ meme).A consensus hexanucleotide sequence, centered around a CC dinucleotide, with highly significant enrichment (E = 6.7e−010) is revealed as the preferential RBP42 crosslink sequence motif (Fig. 2D).Since RBP42 harbors single RNA recognition motif (RRM) domain, the shortness of the sequence motif is consistent with RNA recognition by other RRM domain-containing proteins (39).In vitro binding assays show specific binding of RBP42 to a candidate target sequence (see Fig. S3 at https://doi.org/10.6084/m9.figshare.21737321).Substitution of the CC dinucleotides within the target sequence decreased but did not abolish the binding affinity (Fig. 2E).
Tabulation of RBP42 targets, compiled from all four samples, resulted a combined and overlapping set of 2,145 transcripts: 796 in Sample 1, 903 in Sample 2, 1,449 in Sample 3, and 1,020 in Sample 4 (see Fig. S1d at https://doi.org/10.6084/m9.figshare.21737321).To   undertake the functional significance of RBP42 binding on target transcripts, we classified the "most reliable" RBP42-target set using a stringent criterion-189 congruent transcripts identified in all four iCLIP libraries.The "most reliable" RBP42-target mRNAs represent a set of relatively stable (median half-life of 40 min compared to 12 min for the non-target set) (40), long mRNAs (median length of 2.5 kb compared to 1.6 kb for the non-target set) with long 3′ untranslated region (UTR) (median 3′UTR length of 800 nt compared to 360 nt for the non-target set) (Fig. 3).We validated RBP42's in vivo associa tion by performing independent CLIP experiments in which we measured the enrich ment of a representative set of target transcripts using anti-RBP42 antibodies relative to non-specific antibody control.Indeed, all 10 target transcripts were enriched, from 2.5-to 7.5-fold, but 3 non-target transcripts were not (see Fig. S4a  include 16 of unknown function and therefore annotated as hypothetical protein.Gene ontology (GO) analysis of 165 known protein groups shows several enriched GO terms associated with primary metabolic processes (see Fig. S4b at https://doi.org/10.6084/m9.figshare.21737321).Similar result was reported with RBP42-targets in fly-infective procyclic forms (29), indicating a conserved RBP42 function in both procyclic and slender bloodstream stages of the parasite.We also observed that RBP42-targets include transcripts encoding several RNA binding proteins, 18 in the "most reliable" 189-target set including its own, and several translation factors, indicating a critical RBP42 function in modulating parasite's RBP-mediated post-transcriptional regulation.Interestingly, many tRNAs, five tRNAs in the "most reliable" set, were crosslinked to RBP42 (see Table S1 at https://doi.org/10.6084/m9.figshare.21737321).Since RBP42 is known to be associated with translating polysomes (29), crosslinking to tRNAs is not surprising.However, the exact mechanism of how RBP42 connects both mRNA and tRNA within the translating polysomes is not known.

RBP42 is essential for slender bloodstream form T. brucei survival
To investigate the functional significance of RBP42 binding on target gene expression, we generated an RBP42 conditional knockdown slender bloodstream form cell line (RBP42 Ty1 ) in which cell growth depends upon a tetracycline-regulated exogenously expressed Ty1-tagged version of RBP42 (Fig. 4A and D).Two antibiotic resistance genes replaced two native RBP42 alleles, which were confirmed by genomic PCR (Fig. 4B).Immunoblot analysis confirmed that only the tagged RBP42 protein is expressed log cell number, ml-1 5 in RBP42 Ty1 cells (Fig. 4C).In the presence of tetracycline, added to the growth media, RBP42 Ty1 cells grew normally.However, loss of RBP42, in the absence of tetracycline, caused RBP42 Ty1 cells to stop dividing after 2 days, as reflected by the growth curve, and eventual death (Fig. 4D).Immunoblot and immunofluorescence microscopic analyses confirmed the expected reduction of tagged-RBP42 protein in RBP42 Ty1 cells (Fig. 4E and see Fig. S5a  RBP42 knockdown triggered similar phenotypic alterations of procyclic forms (29), indicating its essential role in both mammalian and insect stages of the parasite.
To seek answers to how RBP42 regulates T. brucei gene expression, we measured global changes in cellular transcriptome, proteome, and metabolome in RBP42 Ty1 cells following loss of RBP42 (Fig. 4F).

Loss of RBP42 leads to decreased levels of target mRNA-encoded protein
To assess the effect of RBP42 on mRNA translation, we measured overall protein synthesis rate before (day 0) and after 1 (day 1) and 2 (day 2) days of RBP42 depletion, using a recently developed non-radioactive method known as the surface sensing of translation (SUnSET) (42), which relies on the principle that puromycin, owing to its structural analogy to tyrosyl t-RNA, is incorporated into elongating peptide chains.By detecting newly synthesized puromycin incorporated peptides, with a monoclonal anti-puromycin antibody in immunoblot assays, we evaluated cellular translational activity.Total lysates from puromycin-treated cells detected a trail of puromycin-incor porated newly synthesized peptides (see Fig. S8a at https://doi.org/10.6084/m9.figshare.21737321).No signal was detected in control lysates, prepared from untreated cells, and therefore ruling out the possibility of non-specific antibody binding to existing peptides.We observed ~25% diminished rate of protein synthesis following 2 days of RBP42 knockdown (see Fig. S8b at https://doi.org/10.6084/m9.figshare.21737321).Reduced protein synthesis was also observed in procyclic forms following loss of RBP42 (29), indicating its possible role on mRNA translation in both mammalian and insect stages of the parasite.
To determine the specific effect of RBP42 on its target mRNA translation, we measured proteome levels in RBP42 Ty1 cells before (day 0) and after 2 days of RBP42 depletion (day 2) using Isobaric Tags for Relative and Absolute Quantitation (iTRAQ)based LC/MS/MS method.Total cellular proteins from eight samples, four replicates of each condition, were labeled with eight unique iTRAQ reagents and analyzed (see Fig. S9a at https://doi.org/10.6084/m9.figshare.21737321).All eight samples exhibit similar iTRAQ intensity distribution, with a median value ~100, indicating robust, reprodu cible measurement (see Fig. S9b at https://doi.org/10.6084/m9.figshare.21737321).As expected, there is >5-fold drop in RBP42 protein following 2 days of deple tion (see Fig. S9c and d  We observed major alterations in cellular proteome following loss of RBP42.Significant changes (P value < 0.01) were observed for 1,650 proteins (~30% of quantified proteome), of which 340 proteins showed ≥1.2-fold upregulation, and 226 proteins showed ≤0.8-fold downregulation (Fig. 5A and see Table S1 at https://doi.org/10.6084/m9.figshare.21737321).Of the 183 "most reliable" RBP42-target mRNA, identified by the iCLIP, our proteomic analysis quantified 178 mRNA-encoded proteins.Majority of the target-encoded proteins were downregulated; 109 proteins, of which 23 proteins <0.8-fold, in contrast to 69 upregulated proteins, of which 7 proteins >1.2-fold (see Fig. S10 at https://doi.org/10.6084/m9.figshare.21737321).The noticeable changes in RBP42-target mRNA-encoded proteins following loss of RBP42, but not target mRNA abundance, indicate a possible translational regulatory role of RBP42 in T. brucei gene expression.Although it is possible that RBP42 exerts distinct translational regulations, both positive and negative on discrete sets of mRNA targets, akin to human antigen R (HuR) protein (43,44), for the majority of its targets, RBP42 acts as a positive regulator of translation.Since RBP42-targets include several RBPs with known post-transcriptional regulatory roles, the observed changes in global transcriptome (see Table S1 at https:// doi.org/10.6084/m9.figshare.21737321)following loss of RBP42 is most likely a secondary effect.
To evaluate the effects of loss of RBP42 on specific cellular processes, we performed GO term enrichment analysis of both upregulated and downregulated proteins (Fig. 5A).Many of the 340 most significant upregulated proteins (>1.2-fold, P < 0.01) are involved in membrane lipid and GPI anchor synthetic process, as well as membrane transporters, and show enrichment of a few general (higher level) GO terms.These include carbohy drate derivative biosynthetic process (GO:1901137, P 4.41e−7), and lipid metabolic process (GO:0006629, P 3.27e−6).In contrast, the 226 most significant downregulated proteins (<0.8-fold, P < 0.01) show many enriched GO terms, both general (higher level) and specific (lower level), with high significance (P < 1e−05) that are associated with primary metabolism (Fig. 5B).Some very specific categories include tricarboxylic acid cycle (GO:0006099, P 2.34e−09), alpha amino acid metabolic process (GO:1901605, P 2.76e−09), and aerobic respiration (GO:0009060, P 1.23e−08).To obtain a meaningful biological interpretation, we analyzed these enriched GO terms using TopGO algorithm (45), which uses underlying GO graph topology to improve GO group scoring and reduce redundancy.The resulting GO graph of the Biological Process (BP) category, illustrated in Fig. 5C, shows clear downregulation of the central carbon and energy metabolic pathway following loss of RBP42.
To examine which specific processes are mostly affected, we compiled, using published data set (23,25,46), a set of gene cohorts associated with central carbon and energy metabolic pathway.Analysis shows downregulation of many metabolic cohorts, but not control cohorts (Fig. 6A).The two most affected cohorts are tricarboxylic acid cycle (TCA) and alpha amino acid oxidation.We observed significant downregulation of enzymatic proteins that are part of the glycolytic, TCA cycle, and the alpha amino acid oxidation processes (Fig. 6B).As anticipated, the levels of mRNAs encoding these enzymes did not show any significant changes (see Fig. S11 at https://doi.org/10.6084/m9.figshare.21737321).Although the existing metabolic model for slender bloodstream form does not support mitochondrial substrate-level or oxidative phosphorylation-medi ated ATP production, recent proteomic studies revealed that most, if not all, enzymes involved in the production of succinate and acetate are expressed (47)(48)(49).Importantly, several of these enzymes are also essential (47,50,51).For example, we observed noticeable downregulation of these essential enzymes following loss of RBP42: PDH-E2  Increasing coloring toward red represents increasing significance levels.The GO descriptions of each node along with significance levels and ratio of hits over total are also shown.

Loss of RBP42 impairs cellular central carbon and energy metabolism
To determine the effect of RBP42 on cellular metabolism, we measured levels of intracellular metabolites, including intermediates from glycolysis and TCA cycle, organic acids, and nucleotides.We reasoned that downregulation of the many metabolic enzymes following loss of RBP42 will alter the levels of these metabolic intermediates.Using mass spectrometry method, we generated metabolite profile of RBP42 Ty1 cells before and after 2 days of RBP42 depletion.Compounds were extracted from four replicate samples of each condition and analyzed to measure quantitative changes in intermediary metabolites of glycolysis, TCA cycle, amino acids, and nucleotides.Analysis revealed noticeable reduction in many glucose-derived metabolites; out of a total of 138 quantified metabolites, 40 show significant decrease (P < 0.05) (Fig. 7A and see Table S1 at https://doi.org/10.6084/m9.figshare.21737321).
We observed noticeable (~20%) decline in pyruvate (Fig. 7B), which is the major end product of catabolized glucose in slender bloodstream form.Pyruvate also serves as a major source of alanine, which declined 35% following loss of RBP42.We also observed large decreases in oxaloacetate, malate, and succinate that are produced from the "succinate branch" of the TCA cycle.Oxaloacetate-derived aspartate is a known precursor of pyrimidine synthesis via dihydroorotate (52).Significant decreases are also observed in purine and pyrimidine nucleotides.In the absence of de novo synthesis, trypanosomes rely on purine salvage (53), which requires ribose 5-phosphate that is synthesized via the oxidative branch of pentose phosphate pathway (54).We observed ~60% decrease in ribose 5-phosphate, indicating that dysregulation of glucose metabolism also caused aberrations in pentose phosphate pathway.In addition to a steep decrease in ATP (~40%), we observed significant decreases in critical cofactors, NAD and NADP, and the methyl donor S-adenosyl methionine (Fig. 7B), all of which are essential for normal cellular metabolic activities.Taken together, these results show that RBP42 ensures proper regulation of core metabolic enzymes, which is critical for the parasite's survival in the diverse nutritional environments encountered throughout its life cycle.

DISCUSSION
RBP-mediated post-transcriptional regulation play key roles in trypanosome gene expression.Although several important RBPs have been studied to date, details of their regulatory roles remain elusive.Here, employing a detailed analysis that combines in vivo RNA target identification with global transcriptomic, proteomic, and metabolic profiling, we provide evidence that RBP42 acts as a critical regulator of T. brucei central carbon and energy metabolism in mammalian-infective slender bloodstream forms (Fig. 8 and see Fig. S12 at https://doi.org/10.6084/m9.figshare.21737321).Our analysis reveals that RBP42 targets mRNAs encoding enzymes involved in core metabolic processes.This finding is consistent with previously identified targets in the fly-infective procyclic forms (29), indicating that RBP42 plays a conserved role in regulating metabolic genes in both stages of the parasite.Using a very stringent iCLIP criteria, we identified 189 transcripts as "most reliable" RBP42 targets.Although iCLIP is extremely powerful in identifying crosslink sites at single nucleotide resolutions, various technical limitations hinder an accurate cataloging and precise enrichment estimation of all mRNA targets relative to RNA abundance (36).This is particularly challenging for genes that are present in multiple identical or almost identical copies in the genome (55).For example, alpha tubulin mRNAs, a very abundant mRNA, is not identified as RBP42 target in our "most reliable" set, contrary to a recent in vivo mRNP capture study that indicated that RBP42 is associated with alpha tubulin mRNAs (56).Only future studies with improved techniques and computational strategies can reveal the true comprehensive set of RBP42 interacting transcripts.
Our study revealed that RBP42 binds within the coding region of target mRNAs.Analysis of RBP42 crosslink sites on mRNA shows preference to a hexanucleotide sequence motif centered around a di-cytidine (Fig. 2D).However, a majority of the crosslink sites on RNA are devoid of this motif, raising the question of how RBP42 associates with specific sets of mRNAs.Since in vitro assays show RBP42 binding to mutated target RNA (Fig. 2E), albeit with reduced affinity, it is possible that RBP42 interacts with cryptic RNA motifs, distributed on target mRNA after initial loading, or recruited to some target mRNA via protein-protein interactions.
Having defined RBP42's RNA target set, we were motivated by curiosity to understand its impact on the regulation of these genes.We focused on the transcript and protein levels of the significant subset of iCLIP-identified targets following loss of RBP42, using a conditional knockdown strategy.Depletion of RBP42 had little effect on steady-state levels of target mRNAs, but major effect on the target proteome, indicating a possible translational regulatory role of RBP42.This is consistent with polysomal association of RBP42 observed in a previous study (29).Since no major changes are observed in the steady-state levels of RBP42-target transcripts following loss of RBP42, we speculate that additional factors are involved that possibly sequester these target transcripts in mRNA storage granules.
Loss of RBP42 caused a moderate reduction in overall cellular translation rate, ~15% on day 1, and ~25% on day 2 (see Fig. S8 at https://doi.org/10.6084/m9.figshare.21737321).This cell-wide reduction in translation rate can be a result of either a uniform partially reduced translation rate of each mRNA, ~25% on day 2 in this instance, or severely reduced translation rate of a subset of mRNAs while the other mRNAs are translated normally.Our data suggest that there are more pronounced translational defects for the RBP42-target mRNAs.However, RBP42-targets include several translation factors of which three are known to be essential from a genome-scale "loss of fitness" study (57) (EIF3L, Tb973.10.4640;EIF3B, Tb927.5.2570;EIF4A1; Tb927.9.4680).Even a slight reductions in these essential translation factors are expected to impair translations of the majority of cellular mRNAs.Therefore, ~25% reduction in overall translation rate after 2 days of RBP42 knockdown is probably a combined effect, a more pronounced direct effect of loss of RBP42 on its target mRNAs, plus a secondary effect on other non-target mRNAs because of diminishing essential translation factors.Dissecting this is an important future goal.Our analysis indicates that loss of RBP42 caused clear downregulation of many enzymes involved in core metabolic processes.Along with glycolysis, the two most affected pathways are mitochondria-resident interconnected energy metabolic pathways, i.e., TCA cycle and amino acid oxidation.It is increasingly recognized that many cellular mRNAs are translated within subcellular domains that allow pre cise localization and regulation of the newly made proteins (58).RBP42 may permit translocation of mitochondrial-resident protein synthesis, while protecting mRNA from degradation.Similar coordinated control of mitochondrial respiratome components is proposed for T. brucei ZC3H39/40 RNA-binding complex (25).RBP42 is known to be phosphorylated (59), which is a well-recognized mechanism that controls the activity of many RBPs (60)(61)(62).It is possible that phosphorylation of RBP42 is regulated by nutrient availability, which in turn may regulate its activity as a translational regulator to fine-tune cellular metabolic activity.

D-D-glucose
Importantly, loss of RBP42 resulted in significant reduction of many intermediary metabolites.Energy metabolism, i.e., production and utilization of ATP, is a highly complex, dynamic process involving numerous factors that respond to intra-and extra-cellular signals, nutrient availability, and cellular physiological and developmental status.Slender bloodstream form T. brucei relies mainly on glucose as the major carbon source and current metabolic model suggests that ATP is exclusively produced via glycolysis, pyruvate being the major (~85%) end product (48).The high ATP demand of proliferating cells is achieved by ~10-fold greater glycolytic rate in slender bloodstream form, compared to procyclic form.However, recent proteomic studies have confirmed that although mitochondrial oxidative-and substrate-level phosphorylation is not utilized for ATP production, various mitochondrial activities are essential for slender form survival.Metabolomic studies confirmed that slender form excretes significant levels of alanine, acetate, and succinate that are produced in the mitochondria.Mitochondrial production of acetate is essential for the slender bloodstream form T. brucei (47).Glucose-derived pyruvate and threonine are two main sources of acetate, which is produced via the "acetate branch" of TCA cycle.Therefore, marked reduction of both pyruvate dehydrogenase (PDH) and threonine dehydrogenase (TDH), following RBP42 knockdown, is expected to severely weaken acetate production.
RBP42 knockdown reduced the levels of many glycolytic enzymes.This modulation of glycolysis may impede the proper function of the oxidative branch of pentose phosphate pathway and therefore reduce ribose 5-phosphate and nucleotide salvage pathways.A marked, ~40%, drop in ATP level was observed.Cellular metabolic activities primarily rely on ATP as energy carrier that drives anabolic reactions critical for cell survival and proliferation.Importantly, ATP also works as structural precursor of important cellular cofactors, including NAD, NADP, and the methyl donor S-adenosyl methionine (SAM).Therefore, a decrease in ATP level is expected to cause widespread disruption of metabolic activity.For example, a decline in SAM is expected to cause major alterations in activities of many methylation-dependent RBP.Absence of protein arginine methyla tion is known to cause striking changes in cellular energy metabolism (27).
In conclusion, it is clear that RBP42 allows proper expression of metabolic enzymes involved in T. brucei central carbon and energy metabolism.By undertaking a broad approach, we show that RBP42-mediated regulation of metabolic networks is critical for the parasite.Importantly, RBP42 homologs are also present in other trypanosomatid organism, including parasitic T. cruzi and Leishmania spp (63).We anticipate that RBP42 homologs play similar metabolic regulatory roles in these related organisms.Although RBP42's precise mode of action remains to be discovered, our analysis hints an apparent translational regulatory role of RBP42.

T. brucei strain and growth analysis
T. brucei Lister 427 bloodstream form wild-type cells, single marker (SM) cells (64), and all stable transgenic cell lines were grown in HMI-9 medium supplemented with 10% fetal bovine serum and 10% serum plus at 37°C in a humidified incubator con taining 5% CO 2 .SM cells that co-express T7 RNA polymerase and the Tet repressor with NEO resistance gene were maintained with 2.5 µg/mL G418.Recombinant DNA constructs were introduced by nucleofection using Amaxa Human T-solution following the manufacturer's instruction.Transgenic cell lines were selected and grown by the addition of puromycin, phleomycin, and blasticidin as required at 0.1, 1.25, and 5 µg/mL, respectively.Homozygous triple-Ty1-tagged RBP42 cell line was generated by N-termi nal epitope tagging of both RBP42 alleles at the native loci.RBP42 conditional knock down cell line (RBP42 Ty1 ) was generated by replacing two native RBP42 alleles with BLE and PAC selectable marker cassettes and introducing an ectopic inducible Ty1-tagged RBP42 allele.The inducible expression construct was introduced into a single RBP42 allele null strain prior to deletion of the second RBP42 allele, while maintaining RBP42 expression by addition of tetracycline to the culture media.RBP42 Ty1 cells were grown in the presence of 1 µg/mL tetracycline to maintain inducible expression of exogenous RBP42 transgene.To shut down the expression of the exogenous RBP42 transgene, cells were washed twice to remove tetracycline and resuspended in culture medium lacking tetracycline.Cultures were seeded at 1 × 10 5 cells/mL and growth analysis was carried out by counting cell density on a hemocytometer.

Plasmid constructs
Epitope tagging of RBP42 alleles at the native loci was carried out using PCR-based strategy (65).Two PCR-generated DNA modules, one with NEO resistance gene and the other with PAC resistance gene, were used in two successive rounds of transfec tion and selection to tag two native RBP42 alleles.The ectopic inducible construct was made by inserting Ty1-tagged RBP42 open reading frame into plasmid pAD74 (66).The resulting construct was linearized using NotI restriction enzyme to facilitate homologous recombination into rRNA loci.A BLA resistance gene within the construct allowed selection of stable cell lines.To replace endogenous RBP42 alleles, knockout gene cassettes were generated by cloning 500 bp upstream and 1,000 bp downstream sequences to flank BLE and PAC resistance genes.Knockout cassettes were released by restriction enzyme digestion prior to transfection.

iCLIP-Seq
iCLIP-Seq libraries were prepared following the published method (67).Prior to immunoprecipitation, cellular extracts were treated with titrated amount of RNase I to generate median of 40-80 nt long RNAs.After high-salt stringent washing, RNAs on beads were ligated to an adapter at the 3′ end, and radioactively labeled on the 5′ end.Protein-RNA complexes, run on 4-12% NuPAGE Bis-Tris gel (Invitrogen, Waltham, MA) and transferred on to Nitrocellulose membrane, were detected using X-ray film.Complexes in the range of 20-40 kDa above RBP42 were excised, and RNA was recovered by proteinase K digestion.Reverse transcription of RNA was performed using primers with two cleavable adapter regions separated by a BamHI site, as well as a barcode to mark unique cDNA molecules.cDNAs were size-selected into two sizes, 80-100 nt as low (L) and 100-150 nt as high (H), using denaturing gel electrophoresis, circularized by CircLigase II ssDNA ligase (Lucigen, Middleton, WI).Subsequently, cDNAs were linearized by BamHI digestion, PCR-amplified, and sequenced on the Illumina NextSeq platform.
iCLIP-Seq data were analyzed using published "analysis pipeline" (68).Raw sequence reads, quality checked using FastQC (www.bioinformatics.babraham.ac.uk/projects/ fastqc), were trimmed to remove adapter and barcode sequences.The barcode sequence was assigned to each read, using Flexbar, as unique molecular identifier (UMI) that was later used to eliminate PCR duplicates (69,70).Sequence reads of at least 15 nt in length were mapped to a hybrid genome consisting of T. brucei 927 genome assembly (version 45) plus T. brucei Lister427 telomeric contigs using STAR aligner with parame ters set to search only unique alignment, with less than 4% mismatched bases (71).Following removal of PCR duplicates using UMI-tools, each unique cDNA molecule was counted as an independent crosslink event.Crosslinked sites, the nucleotide preceding the cDNA start, were extracted using BEDTools suit (72).Significant crosslink sites were determined using PureCLIP cluster finding algorithm (38).Crosslink sequences were extracted by adding 5 nt flanking sequences to crosslink sites to obtain 11 nt crosslink ing regions.Crosslink sequence motif discovery was performed using the MEME suit (https://meme-suite.org/tools/meme), with the parameters-classic mode, zero or one occurrence per sequence, search given strand only, minimum width 4, maximum width 10.

qRT-PCR analysis
CLIP-purified RNA, as described above, was subjected to reverse transcription (RT) with random hexamer primers and iScript reverse transcriptase (Bio-Rad, Hercules, CA).Quantitative reverse transcription PCR (qRT-PCR) was performed in a CFX96 Touch Real-Time PCR detection system using primer pairs targeted at specific transcripts (see Table S1 at https://doi.org/10.6084/m9.figshare.21737321).Data were analyzed using the CFX Manager 3.0 software.Bar plots were generated using mean ± standard errors.

mRNA-Seq
mRNA-Seq libraries were prepared from poly(A) + -containing RNA, captured by two rounds of oligo-d(T) n -bead selection of 10 µg total RNA, following Illumina small RNA library preparation method.Libraries were sequenced on the Illumina NextSeq platform.Sequence reads were mapped to the T. brucei 927 genome assembly version 45 using Bowtie2 (v2.3.5).The mapped reads were then converted to gene expression values and analyzed using DESeq2 (41) and Cuffdiff2 (73) (see Table S1 at https://doi.org/10.6084/m9.figshare.21737321).Volcano and violin plots were generated using expression value (log 2 fold change) and P value in R.
Puromycin incorporation was determined by immunoblot analysis using monoclonal antibody against puromycin (Kerafast, Boston, MA) and quantified using ImageJ.

iTRAQ quantitative proteomics
Quantitative proteomics using iTRAQ method (74) were performed at the Center for Advanced Proteomics Research (NJMS, Rutgers Biomedical and Health Sciences, Newark, NJ).RBP42 Ty1 cells (5 × 10 7 cells/assay) were harvested before (day 0) and after 2 days (day 2) of RBP42 knockdown and washed with ice-cold TBS.Total proteins were extracted using lysis buffer containing 100 mM TEAB, 8M urea, and protease inhibitor cocktail.About 100 µg proteins, from four replicate samples of each condition, was reduced, alkylated, and trypsin digested before subjected to labeling with 8-plex iTRAQ reagents (AB Sciex, Framingham, MA).Peptides from day 0 replicates were labeled with iTRAQ tag-113, 114, 115, and 116-whereas peptides from day 2 replicates were labeled with iTRAQ tag-117, 118, 119, and 121.Subsequently, all labeled peptides from eight samples were pooled and fractionated using high pH RPLC liquid chromatography on ACQUITY UPLC system (Waters Corporation, Millford, MA).A total of 48 fractions were collected in 60 min gradient of Solvent A (20 mM HCOONH 4 , pH 10.0) and Solvent B (20 mM HCOONH 4 in 85% ACN, pH 10.0) and pooled into 12 fractions that were subjected to LC-MS/MS analysis on an UltiMate 3000 RSLCnano coupled with Orbitrap Fusion Lumos Mass Spectrometer (Thermo Scientific).Peptides, ~1 µg from each fraction, were separated on a nano C18 column (Acclaim PepMap, 75 µm × 50 cm, 2 µm, 100 Å) using a 2-h non-linear binary gradient of mobile phase A (2% ACN and 0.1% formic acid) and mobile phase B (85% ACN and 0.1% formic acid) at a flow rate of 300 nL/min.Eluted peptides were introduced into Orbitrap Fusion Lumos system through a nanospray Flex ion source (Thermo Scientific, Waltham, MA) with the spray voltage of 2 kV and a capillary temperature of 275°C.The MS spectra was acquired in a positive mode.For MS1, peptide scan range was set to 375-1,500 with the resolution of 120,000.Peptides with charge-state of 2-7, and intensity greater than 5 × 10 3 were selected for MS/MS scan in ion-trap using collision-induced dissociation (CID) with the collision energy of 35%.The dynamic exclusion is 60 s and the isolation window is 0.7 m/z.For SPS-MS3 scan, the precursor selection range was 400-1,200 with iTRAQ ion excluded.Ten SPS precursors were selected for MS3 scan in orbitrap with resolution of 50,000.High energy collision dissociation (HCD), with the collision energy of 65%, was used for iTRAQ tag quantitation.
The iTRAQ MS data were searched against UniProt T. brucei brucei (strain 927/4 GUTat10.1)database (8,579 proteins) using Sequest search engine on Proteome Discoverer (V2.4) platform.MS1 mass tolerance was set to 10 ppm and MS2 mass tolerance was 0.6 Da. iTRAQ 8-plex (K), iTRAQ 8-plex (N-terminal), and methylthio (C) were set as fixed modification, whereas oxidation (M) and iTRAQ 8-plex (Y) as variable modifications.Two missed cleavages are allowed in trypsin digestion.The reporter ion-based quantification workflow was chosen for data analysis; the CID spectra in MS2 were used for peptide identification and the HCD spectra in MS3 were used for iTRAQ quantitation.The false discovery rate for protein and peptide was set to 1% filtered with Percolator.The protein relative quantitation was calculated based on the ratio of (average in day 2 abundance)/(average in day 0 abundance) (see Table S1 at https://doi.org/10.6084/m9.figshare.21737321).Significance (P values) was computed using paired sample t-test.

Metabolomics
Metabolomic profiling was performed at the Metabolomics Shared Resources Facility (Rutgers Cancer Institute of New Jersey, New Brunswick, NJ), following the published method (75).Briefly, cells (2 × 10 7 cells/assay) were harvested as above.Metabolites were extracted with 1 mL of 40:40:20 mixture of methanol:acetonitrile:water plus 0.5% (vol/vol) formic acid on ice for 5 min.Following neutralization of formic acid by addition of 50 µL of 15% (m/vol) NH 4 HCO 3 , cleared extracts were collected by centrifugation at 15,000× g for 10 min and stored at −80°C.Metabolomic data, designed to capture intermediary metabolites in central carbon metabolism including glycolytic intermedi ates, TCA compounds, amino acids, nucleotides, and derivatives, were obtained using hydrophilic interaction liquid chromatography separation method coupled with mass spectrometry run in negative ionization mode.Each metabolite was identified by matching of accurate mass and retention time to synthetic standards.Metabolite bar plots were generated using mean ion counts ±standard errors.Significance (P value) was computed using paired sample t-test (see Table S1

Fluorescence microscopy
Cells were fixed in 1% formaldehyde and adhered to slides coated with poly-L-lysine.RBP42 immunolocalization was performed using anti-RBP42 antibodies at 1:10,000 on cells permeabilized with 0.2% NP40 for 5 min at room temperature.FITC conjuga ted secondary antibodies are used at 1:1,000.Cells were mounted with Vectashield (Vector Laboratories, Newark, CA) containing DAPI and imaged using an Olympus BX61 microscope equipped with DAPI and FITC-sensitive filters and a Hamamatsu ORCA-ER camera.

Gene ontology (GO) and pathway enrichment analysis
GO term enrichment analysis was performed using resources available from Tri TrypDB.orgdatabase.REVIGO web tool was employed to summarize GO enrichment by eliminating redundant GO terms (76).Directed acyclic graph of enriched GO terms was produced using R Bioconductor topGO package (45).Pathway analysis was performed by R Bioconductor packages GAGE (77) and Pathview (78) using Kyoto Encyclopedia of Genes and Genomes database (79).

FIG 1
FIG 1 Crosslink immunoprecipitation of RBP42-RNA complex from slender bloodstream form T. brucei.(A) Schematic showing homologous recombinationmediated in situ tagging to generate homozygous 3Ty1-epitope tagged RBP42 cell line.PCR generated DNA cassettes, containing neomycin (NEO) and puromycin (PUR) selectable marker genes, are shown.(B) Analysis of genomic DNA by PCR using FP and RP primers shown in panel (A) verifies proper integration of the DNA cassettes in two representative clonal cell lines.(C) Immunoblot analysis of total cellular proteins confirms the presence of only 3Ty1-tagged RBP42 protein in these clonal transgenic cell lines.Whereas anti-RBP42 antibody detects both native RBP42 and 3Ty1-tagged RBP42 proteins, anti-Ty1 (BB2) antibody detects only the tagged protein.Asterisk marks non-specific binding of anti-RBP42 antibody to an unknown protein.(D) Growth analysis shows that both clonal cell lines grow comparably to wild-type (WT) cells.(E) Immunoprecipitated protein-RNA complexes that were 5′-end radiolabeled on their RNA.

FIG 2
FIG 2 iCLIP-seq identifies RBP42's RNA targets in slender bloodstream form T. brucei.(A) Four iCLIP-seq cDNA libraries (Samples 1-4) were made from both wild-type (WT) cells, using anti-RBP42 antibody (Samples 1 and 3), and 3Ty1-RBP42 cells using anti-Ty1 (BB2) antibody (Samples 2 and 4), applying two increasing UV doses, 150 mJ/cm 2 (Samples 1 and 2) and 300 mJ/cm 2 (Samples 3 and 4).(B) RBP42 predominantly binds within the coding region of mRNA targets.Bar plots of cDNA read distributions, mapped to T. brucei genomic features (T.brucei brucei strain 927 genome from TriTrypDB.org-version46), are shown.CDS and intergenic denote the protein-coding open reading frames and the 5′ and 3′ untranslated regions of mRNA, respectively.tRNA/rRNAs are as defined in the database mentioned above, and no features are without any annotation.(C) IGV browser view of RBP42 crosslinking to two candidate mRNA targets, one on chromosome 10 (Tb927.10.6880-glyceraldehyde3-phosphate dehydrogenase, cytosolic) and the other on chromosome 11 (Tb927.11.5520-triosephosphate isomerase).A crosslink event is counted for each unique cDNA and assigned to the upstream crosslink nucleotide (see Materials and Methods).The red bars indicate the number of crosslink events on crosslink sites.Few neighboring non-target genes are also shown as internal controls to indicate specificity of RBP42 interactions.(D) The composite sequence motif associated with RBP42 is discovered by comparing significant crosslink clusters using DREME algorithm (meme-suite.org).(E) PhosphorImage analysis of EMSA shows complexes of RBP42 protein with target and mutated RNA substrates.RNA sequences and dissociation constants (Kd) are indicated.

FIG 3
FIG 3 RBP42-targets represent a set of relatively long, stable mRNAs.Panels A and B show the density and the cumulative distribution frequency plots of mRNA lengths, and 3′UTR lengths of 183 RBP42-target mRNAs (red) compared to non-target background sets (gray).Violin plots show the distributions of the 183 RBP42-target mRNAs and 3′UTR lengths (target, red), compared to two control mRNA sets of equal size, randomly sampled from the database (random1 and random2, gray).Panels C and D show similar plots for T. brucei bloodstream form mRNA half-lives and GC content.

FIG 4
FIG 4 RBP42 is essential for slender bloodstream form T. brucei.(A) Schematic shows the strategy used to generate RBP42 conditional knockdown cell line (RBP42 Ty1 ).Homologous recombination-mediated replacement by two antibiotic resistance genes inactivated two endogenous alleles.Cellular RBP42 expression is maintained from an exogenous tetracycline-inducible Ty1-tagged RBP42 allele inserted in an rRNA spacer region.(B) Analysis of genomic DNA by PCR, as in Fig. 1, verifies proper integration of the antibiotic resistance DNA cassettes.(C) Immunoblot confirms expression of only the Ty1-tagged RBP42 protein from the regulated allele in RBP42 Ty1 cells.Antibodies are described in Fig. 1. (D) Growth analysis shows that RBP42 is essential for cell viability.RBP42 Ty1 cells were maintained by adding tetracycline to the growth media.RBP42 expression was turned off by removing tetracycline from the growth media.Parasites were counted using a hemocytometer.(E) Immunoblot confirms regulated expression of the Ty1-tagged RBP42 protein in RBP42 Ty1 cells.α-tubulin is loading control.(F) Schematic of experimental design.Expression of RBP42 was turned off on day 0 by removing tetracycline from the growth medium.Knock down of RBP42 was indicated by gray ramp.

FIG 5
FIG5 Loss of RBP42 reduces the levels of enzymatic proteins of central carbon and energy metabolic pathways.(A) iTRAQ quantitative proteomics of total cellular proteins from RBP42 Ty1 cells before (day 0) and 2 days after (day 2) RBP42 knockdown.The number of protein groups with significantly (P ≤ 0.01, by paired sample t-test) changed levels (increased, ≥1.2-fold; decreased, ≤0.8-fold) are shown.The number of upregulated and downregulated proteins that are in iCLIP "most reliable" data set are shown in parenthesis.(B) Gene ontology (GO) enrichment analysis of 226 significantly downregulated proteins following loss of RBP42.Dot plot showing top 20 enriched GO term.The x-axis represents percent hits, which is the ratio of the number of proteins in the 226 decreased set to the number of all annotated proteins with same GO term.The sizes of the dots represent the number of downregulated proteins associated with the GO term.The colors of the dots represent adjusted P values (BH).(C) GO graph showing significantly enriched GO terms in the Biological Processes category of the 226 downregulated proteins (using TopGO).Each node marks a GO term, and each arrow indicates an "is-a" relationship.Boxes indicate 10 most significant nodes.

2 FIG 7
FIG 7 Loss of RBP42 alters cellular metabolic profile.(A) Heatmap view of metabolite levels following RBP42 knockdown.Metabolomes were quantified using LC-MS and metabolites were identified with known standards.Metabolites with significantly changed levels (P ≤ 0.05) are shown.Four replicate samples prepared from RBP42 Ty1 cells before (day 0) and after 2 days of RBP42 knockdown (day 2) are indicated at the bottom.Metabolites are labeled on the right.The abundance of each metabolite is log 2 transformed and mean-centered with blue being less abundant and yellow more abundant.The dendrograms are based on hierarchical clustering.(B) In the absence of RBP42 protein, several important molecules that determine cellular energy and redox states are significantly reduced.Bar plots showing mean ± standard error of signal intensities of six compounds before (red, day 0) and after (green, day 2) RBP42 knockdown.Significance P values, estimated by paired sample t-test, are shown.