Identification of positive and negative regulators in the stepwise developmental progression towards infectivity in Trypanosoma brucei

Trypanosoma brucei is a protozoan parasite that causes important human and livestock diseases in sub-Saharan Africa. By overexpressing a single RNA-binding protein, RBP6, in non-infectious procyclics trypanosomes, we previously recapitulated in vitro the events occurring in the tsetse fly vector, namely the development of epimastigotes and infectious, quiescent metacyclic parasites. To identify genes involved in this developmental progression, we individually targeted 86 transcripts by RNAi in the RBP6 overexpression cell line and assessed the loss-of-function phenotypes on repositioning the kinetoplast, an organelle that contains the mitochondrial genome, the expression of BARP or brucei alanine rich protein, a marker for epimastigotes, and metacyclic variant surface glycoprotein. This screen identified 22 genes that positively or negatively regulate the stepwise progression towards infectivity at different stages. Two previously uncharacterized putative nucleic acid binding proteins emerged as potent regulators, namely the cold shock domain-containing proteins CSD1 and CSD2. RNA-Seq data from a selected group of cell lines further revealed that the components of gene expression regulatory networks identified in this study affected the abundance of a subset of transcripts in very similar fashion. Finally, our data suggest a considerable overlap between the genes that regulate the formation of stumpy bloodstream form trypanosomes and the genes that govern the development of metacyclic form parasites.

Trypanosoma brucei is a protozoan parasite that causes African sleeping sickness in humans and together with T. congolense and T. vivax remains a public health concern in sub-Saharan Africa by afflicting livestock with nagana. During its complex life cycle between the mammalian host and the blood-feeding tsetse fly vector (Glossina sp.), T. brucei transitions through many distinct developmental forms to adapt to changing environments. In the infected mammal, T. brucei is coated with a variant surface glycoprotein 1 (VSG) and exists as both a proliferative slender and a quiescent stumpy form. The transition from slender to stumpy form parasites is triggered in a cell-density dependent manner through a quorum sensing mechanism 2,3 . Stumpy form parasites are primed for survival in the tsetse fly as evident from the increased expression of mitochondrial proteins 4 , which prepares parasites to exploit oxidative phosphorylation for ATP production and use of proline as the new carbon source in the tsetse fly, and the protein associated with differentiation 1 (PAD1) 5 , which internalizes cis-aconitate and citrate that act as signaling molecules to trigger development into procyclic parasites. Upon ingestion of bloodstream form trypanosomes by the tsetse fly, the stumpy form parasites that withstand the digestive proteases 6,7 , insect immune system 8 , and alkaline conditions 9 found in the midgut will differentiate into procyclic forms and lose mammalian infectivity 1,10 . This differentiation process is accompanied with major morphological and metabolic changes that include repositioning of the kinetoplast (an organelle that contains the mitochondrial genome) between the posterior pole and the nucleus, expansion of the mitochondrial cisternae for an oxidative phosphorylation-dependent metabolism, loss of the VSG coat, and expression of a glycoprotein coat composed mainly of procyclins rich in Gly-Pro-Glu-Glu-Thr repeats (GPEET) and procyclins rich in Glu-Pro repeats (EP) 1,[11][12][13] .
Procyclic parasites undergo an intricate developmental process known as metacyclogenesis, whereby noninfectious trypanosomes develop into infectious metacyclic parasites, and regain mammalian infectivity through the expression of a specific subset of VSGs known as metacyclic VSGs (mVSGs). As procyclic parasites transit to a portion of the foregut in the tsetse fly known as the proventriculus, they elongate, reposition the kinetoplast anterior to the nucleus, and become epimastigotes, which undergo asymmetric division to produce long and short epimastigotes 1 . Epimastigotes travel to the salivary gland, where they express a family of alanine-rich surface proteins known as brucei alanine rich protein (BARP) 14 . A subset of these epimastigotes will attach to the epithelia of the salivary gland and gain the capacity to undergo either symmetric division to colonize the salivary gland or asymmetric division to produce infectious, quiescent metacyclic parasites 1 . Major morphological and metabolic changes occur in metacyclic parasites, which include a morphologically regressed mitochondrion that primes the parasite for a glycolysis-based metabolism 15 , increased endocytosis 16 , a repositioned kinetoplast at the posterior end of the parasite 1 , and mVSG expression 17 for survival in the mammalian host. The molecular mechanisms underlying this developmental process remain largely unknown. Previously, this dearth of knowledge was in part due to the lack of an in vitro differentiation system, which prevented the use of large-scale genetic and biochemical experiments. This limitation was overcome with the establishment of an in vitro differentiation system, where procyclic parasites overexpressing the RNA-binding protein 6 (RBP6) progress to become metacyclic parasites 18 . This system recapitulates most of the T. brucei developmental processes from procyclic to metacyclic parasite and captures many of the in vivo hallmarks of differentiation 18 . Since then two additional in vitro differentiation systems have been developed. The overexpression of RBP6 with a single-point mutation (Q109K) skipped the intermediate epimastigote stage and the resulting metacyclics were able to progress to bloodstream forms in vitro 19 . In addition, RBP10 was shown to promote the bloodstream form differentiation state, when overexpressed in procyclic parasites [20][21][22] .
Among the three in vitro differentiation systems, the wild-type RBP6 overexpression system best reflects the development of trypanosomes within the tsetse fly and has opened up a better understanding of metacyclogenesis. For instance, metacyclics obtained from this system are quiescent, i.e. they are non-dividing and display a dramatic reduction in transcription, protein synthesis and the amount of ribosomal proteins 23 . Transcriptomic and proteomic analysis revealed an increase in glycolytic proteins and their corresponding mRNAs in metacyclic parasites relative to procyclic parasites, which is consistent with the progression to glycolysis as the primary mode of ATP generation 23 . Furthermore, metacyclic parasites expressing one of the eight major T. brucei Lister 427 mVSGs 24 were able to establish an infection in mice 18 . Recently, mitochondrial metabolic changes in the proline oxidation pathway and tricarboxylic acid (TCA) cycle have been documented using transcriptomic, proteomic, and metabolomic approaches 25 . Upon RBP6 induction, there was an upregulation of proline catabolism and components of the TCA cycle, which fuels oxidative phosphorylation-based ATP production. Changes in the electron transport chain increased the levels of reactive oxygen species, which act as signaling molecules to drive cellular differentiation from epimastigotes to metacyclic parasites 25 . Utilizing this in vitro differentiation system, we performed a targeted RNAi screen to identify genes involved in regulating metacyclogenesis. We individually targeted 86 transcripts by RNAi in the inducible RBP6 overexpression cell line and assessed the loss-of-function effects on kinetoplast repositioning, and BARP and mVSG expression, resulting in the identification of 22 genes that positively or negatively regulate different steps of metacyclogenesis.

Results
RBP6 induction alters the mRNA translation landscape. The goal in this project was to identify genes downstream of RBP6 that are involved in the developmental program transforming procyclic parasites into quiescent, mammalian-infectious metacyclic parasites, also referred to as metacyclogenesis. In a first set of experiments, we investigated the impact of RBP6 induction on changes in translation efficiencies, i.e. changes which are not paralleled by corresponding changes in cytoplasmic mRNA levels. Un-induced and RBP6-induced cells for 48 h were treated with the translation elongation inhibitor cycloheximide to immobilize ribosomes on mRNA and cytosolic extracts were separated on 15-50% linear sucrose density gradients by ultracentrifugation to separate polysomes from monosomes, ribosomal subunits and messenger ribonucleoprotein particles (mRNPs) 26 . This enables discrimination between efficiently translated from poorly translated mRNAs. The polysome profile derived from each sample (Fig. 1a,c) was used to determine the fractions that corresponded to mRNPs (fractions 1-2), 40S (fractions 3-4), and ribosomal subunits 60S and 80S (fractions 5-7). Using the sedimentation of α-tubulin mRNA, assayed by Northern blot (Fig. 1b,d), as a guide, we arbitrarily defined fractions 8-12 and 13-22 as light and heavy polysomes, respectively. Fractions containing heavy polysomes were pooled and along with total RNA samples analyzed by RNA-Seq (see "Methods" section). We performed three biological replicates and identified 990 and 12 transcripts whose polysome association was increased or decreased, respectively, at least threefold relative to un-induced cells (Supplementary Table S1). The abundance of 304 transcripts was upregulated at 48 h 19 and 238 of these revealed increased polysome association. In contrast, 752 transcripts with increased polysome association did not change transcript abundance levels (Supplementary Table S1), thus highlighting translation as an important regulatory step. Targeted RNAi screen identifies 22 genes that play a role in metacyclogenesis. Since we know that the induction of RBP6 expression for only one day results in the generation of metacyclics (Supplementary Fig. S1), we hypothesized that RBP6 is triggering a cascade of gene expression, biochemical, and morphological events. Initially, we attempted a genome-wide RNAi screen to identify genes downstream of RBP6, www.nature.com/scientificreports/ but encountered technical difficulties, likely due to an apparent interference between the RNAi library and the RBP6 overexpression system. We therefore designed a targeted RNAi screen based mainly upon our available transcriptomic 19,23 and proteomic 23 datasets that documented differential gene expression during the transformation from procyclic to epimastigote and, subsequently, into metacyclic parasites. We primarily chose candidates that encode protein kinases, protein phosphatases, cell-cycle regulators, and DNA-and/or RNA-binding proteins, because these categories of proteins are often associated with developmental regulation in eukaryotic organisms 3,27-30 . We included 61 genes whose transcripts were at least two-fold upregulated after RBP6 overexpression for 24 h or in metacyclic parasites 19 . We also included 3 genes that increased in protein abundance by at least two-fold in metacyclic parasites relative to procyclic parasites 23 . Additionally, 8 genes whose transcripts were approximately two-fold downregulated in metacyclic parasites compared to procyclic parasites were added to the screen, since these genes could be negative regulators of metacyclic development 19 . Finally, based on the experiments described above, we included 25 genes whose transcripts increased in polysome association by at least two-fold in RBP6-induced cells for 48 h when compared to un-induced parasites (Supplementary Table S1). From this set of 97 RNAi targets, we eliminated 11 genes, whose down-regulation has been shown to cause growth arrest or parasite death by a previous genome-wide RNAi screen for procyclic parasite fitness 31 . Of the 86 total candidate genes (Supplementary Table S2), 15 genes were previously identified to play an important role in the development of bloodstream parasites from slender to stumpy forms 3 . Since both short stumpy and metacyclic parasites are quiescent and pre-adapted for transmission, we hypothesized that these genes required for stumpy bloodstream form development may also be involved in the production of metacyclics. Because T. brucei mRNA degradation by the RNAi machinery is efficiently and specifically triggered by double-stranded RNA produced in vivo as a long hairpin 32,33 , we created plasmid constructs that utilize a tetracycline-inducible promoter to transcribe hairpins targeting the open reading frame. Our list included 4 gene families where individual members were targeted simultaneously, namely genes coding for Proteins Associated with Differentiation (PAD 1-8), two hexokinases (HK1-2), a gene family of protein phosphatases 1 (PP1 4-6), and a gene family of NEK kinases (NEK17 1-3). The remaining transcripts were targeted individually, resulting in a total of 74 RNAi cell lines. Each RNAi cassette was stably integrated in the non-transcribed 177-bp repeat satellite DNA of T. brucei Lister 427  cells, while the tetracycline-inducible RBP6 construct was stably integrated in the non-transcribed rDNA spacer. Therefore, upon doxycycline exposure, each cell line simultaneously expressed RBP6 to trigger development from procyclic to metacyclic forms and a double-stranded RNA hairpin targeting a transcript of interest for RNAi downregulation. In this screen, we did not monitor the efficacy of RNAi depletion of the targeted transcripts, and thus, if an RNAi cell line did not display a phenotype, we cannot conclude that the targeted gene is not involved in metacyclic development. Additionally, the degree of RNAi www.nature.com/scientificreports/ efficiency for different mRNAs could influence the severity of the observed phenotype. Lastly, our experimental design was aimed at analyzing genes acting downstream of RBP6, so any genes operating upstream will not be detected, particularly because protein overexpression generally occurs more rapidly than RNAi in T. brucei. The RBP6 overexpression system we use will result in approximately 30% of the cells developing into epimastigotes after 2-3 days and approximately 50% of the cells will be infectious metacyclic parasites after 4-6 days. We used two experimental methods to quantify the efficacy of this developmental program (Fig. 2). First, we Hoechst stained the differentiating population and used fluorescent microscopy to distinguish and quantify epimastigote parasites based on the relative position of the parasite's kinetoplast to its nucleus. Whereas the kinetoplast of both procyclic and metacyclic parasites is positioned posterior to the nucleus, the kinetoplast of epimastigotes is localized anterior to the nucleus. Second, we used Western blot analysis to monitor the expression of specific surface proteins. Since epimastigotes express BARP 14 and metacyclic parasites express one of the eight major mVSGs, the expression of BARP and mVSG397, the most abundantly expressed mVSG and present on approximately 50% of metacyclics 19,34 , served as a positive readout for the quantity of epimastigote and metacyclic parasites, respectively. Taken together, Fig. 2 illustrates our assays in a 6-day time course of RBP6 overexpression in procyclic parasites. The first event to happen is the repositioning of the kinetoplast and this occurs before BARP expression, which peaks on the third day. Finally, mVSG397 expression is highest on the 6 th day, which correlates with the morphological counts of metacyclics 34 .
Since we have previously observed that low levels of RBP6 expression inefficiently stimulate metacyclic development 18 , we first assessed RBP6 expression levels in the generated RNAi cell lines after 24 h of doxycycline induction. In comparison to the parental RBP6 expressing cell line, each of the RNAi cell lines expressed RBP6 at a similar level (Supplementary Figs. S2 and S3). Next, we assessed whether RNAi affected the morphological appearance of epimastigote parasites. The most severe phenotype we observed was caused by the down-regulation of cold shock domain-containing protein 2 (CSD2; Tb927.4.4520) with a 66% reduction in the number of epimastigote parasites, relative to the RBP6 overexpression cell line (Fig. 3a). Additionally, in 5 RNAi cell lines we observed a slight, but statistically significant effect on the morphological appearance of Figure 2. Targeted RNAi screen overview. BARP and mVSG397 are surface markers that serve as a positive readout for epimastigote and metacyclic production, respectively. Epimastigotes were quantified on the third day and metacyclics on the sixth day, since these are the times where BARP and mVSG397 expression was highest in our 6-day time course. Each RNAi cell line was induced with doxycycline to overexpress RBP6 to stimulate development and to express an RNAi hairpin that down-regulates a specific transcript of interest. After three days of induction, cells were scored as epimastigotes, if their kinetoplast localized anterior to the nucleus and these results were compared to the positive control, the wild-type RBP6 overexpression cell line. Induced cells from the targeted RNAi screen and the positive control were also harvested on the third and sixth day of doxycycline induction for Western blot analysis to quantify BARP and mVSG397 expression, respectively. Fulllength blots are shown in Supplementary Fig. S10 S4). Taken together, we identified 6 positive and 3 negative effectors of kinetoplast repositioning in the initial phase of the developmental program. Next, we evaluated by Western blotting which of the 74 RNAi cell lines affected BARP expression after three days of RBP6 overexpression. The RNAi-induced down-regulation of 18 individual transcripts resulted in a statistically significant decrease in BARP expression relative to the 3-day induced RBP6 overexpression cell line (Fig. 3b, Supplementary Fig. S5 and Supplementary Table S3). Among these 18 genes were the 9 RNAi cell lines we have shown above to positively or negatively affect the morphological appearance of epimastigotes. In addition, we identified 9 cell lines that displayed normal kinetoplast repositioning, but BARP expression was decreased by approximately 40% (Fig. 3b, Supplementary Fig. S5 and Supplementary Table S3). These 8 genes code for a GPI-anchor transamidase subunit 8 (GPI8, Tb927.10.13860), a Pumillio/PUF RNA-binding protein 11 (Puf11, Tb927.11.10810), a hypothetical protein 2 (Hyp2, Tb927.9.4080), an AMP-activated protein kinase β subunit (AMPKβ, Tb927.8.2450), an AMP-activated protein kinase γ subunit (AMPKγ, Tb927.10.3700), a cold shock domain-containing protein 1 (CSD1, Tb927.8.7820), a cyclin 8 (CYC8, Tb927.7.1390), a Wee1-like protein kinase (Wee1, Tb927.4.3420), and a DYRK protein kinase (DYRK, Tb927.10.15020).
Finally, we monitored each of the 74 RNAi cell lines by Western blotting for their ability to express mVSG397 after 6 days of RBP6 overexpression (Fig. 3c, Supplementary Fig. S5 and Supplementary Table S3). We identified at total of 18 cell lines that displayed at least a 50% reduction in mVSG397 expression, when compared to the 6-day induced RBP6 overexpression cell line. Downregulation of AGC/RSK, CYC8, and Wee1 only resulted in a slight, but statistically significant decrease in mVSG397 expression (Fig. 3c, Supplementary Fig. S5 and Supplementary Table S3). In contrast, RNAi reduction of the remaining identified 14 transcripts that affected epimastigote kinetoplast repositioning and/or BARP expression impaired mVSG397 expression up to 90%. This final phenotypic analysis also revealed four cell lines with normal kinetoplast repositioning and BARP expression, but decreased mVSG397 expression. The four targeted transcripts encode RNA binding protein 3 (RBP3, Tb927.11.530), protein tyrosine phosphatase 1 (PTP1, Tb927.10.6690), a rac serine-threonine kinase (RAC, Tb927.6.2550), and a metacaspase (MCA1, Tb927.11.3220), and these proteins are likely to play a role in the later step(s) of the developmental progression to metacyclics. Unexpectedly, we noticed that the down-regulation of DYRK, which displayed normal kinetoplast repositioning ( Fig. 3a and Supplementary Table S3), but reduced BARP expression (Fig. 3b, Supplementary Fig. S5 and Supplementary Table S3), did not result in a significant reduction of mVSG397 expression.
RNA-seq analysis of selected cell lines. The targeted RNAi screen described above identified 22 genes, which are required at different steps in the development of infectious metacyclics (Fig. 5 and Supplementary  Table S3). To begin to gain an understanding of the specific role(s) these genes play, we concentrated on RNA/ DNA binding proteins and a signal transducer and performed RNA-Seq on the RNAi cell lines of AMPKα2, ZC3H45, CSD1, and CSD2, and the RBP10 KO cell line. It is important to note that the initial phenotypic analysis for RBP10 was done with the RNAi cell line. The subsequent establishment of an RBP10 KO cell line in the RBP6 overexpression background confirmed the RNAi results (Supplementary Tables S4 and S5) and thus we used the KO cell line in subsequent analyses. Three biological replicates of un-induced, day 1-, 2-, 3-, 4-and 6-induced cells were processed (Supplementary Tables S6, S7, S8, S9 and S10), and their transcriptomes were compared to the parental RBP6 over-expressing cell line. To minimize potential secondary effects of RNAiinduced down-regulation, we restricted the analysis reported here to days 1 and 2. During this time period overexpression of RBP6 revealed 304 transcripts with a more than two-fold higher abundance, as compared to uninduced cells, and 50 transcript had a lower abundance 19 . In our initial screen, CSD2 down-regulation displayed the most severe phenotype affecting the early repositioning of the kinetoplast (Fig. 3a) and this was reflected in the RNA-Seq data with only 44 and 11 transcripts having a higher and lower abundance, respectively ( Fig. 4 and Supplementary Table S11). Most interestingly, during the first two days the abundance of 5 mVSG transcripts (mVSG397, mVSG653, mVSG1954, mVSG531, mVSG639) increased to levels comparable to the parental RBP6 cell line, indicating that CSD2 does not play a major direct role in mVSG activation. In contrast, in the RBP10 KO cell line the 5 mVSG transcripts were downregulated (Supplementary Table S12), indicating that RBP10 may have a role in stabilizing mVSG transcripts. The mVSG abundance did not change in the CSD1 (Supplementary  Table S13), AMPKα2 (Supplementary Table S14) and ZC3H45 RNAi cells (Supplementary Table S15) when compared to un-induced cells, suggesting that these three proteins play a role in mVSG activation directly or indirectly.
There are several additional conclusions we can draw from the RNA-Seq data. The effect on the landscape of upregulated transcripts by the various RNAi cell lines was largely restricted to transcripts changing in the parental cell line, i.e. a very limited number of additional transcripts were upregulated (Fig. 4a) and the number of mRNAs uniquely upregulated in individual RNAi cell lines was surprisingly low (Table 1). In contrast, the downregulated transcripts revealed potential new targets of regulation, including the mVSGs discussed above, since they were only downregulated in the RNAi cell lines (Fig. 4b and Table 2). Of particular interest is the down-regulation of two families of proteins containing META domains in the ZC3H45 RNAi cell line (Supplementary Table S15). These gene families are clustered at the beginning of a polycistronic transcription unit on chromosome 5 and encode either a 446 aa-long protein with two predicted META domains (Tb927.5.2190, 2220, 2250) or a 116 aa-long protein with a single predicted META domain (Tb927.5.2160, 2170, 2200, 2230, 2260). The META domain of about 100 aa has been described in secreted bacterial proteins that are implicated in motility [35][36][37] . In addition, the syntenic protein meta-1 in Leishmania major was localized in the flagellar pocket 38 , was mainly expressed in infective metacyclics 38,39 and overexpression in L. amazonensis increased virulence 40 . The abundance of the T. brucei transcripts of both families did not change during metacyclogenesis 19 , but the transcripts were significantly downregulated in purified metacyclics as compared to procyclics (between 8.6-and www.nature.com/scientificreports/ 11-fold 23 ), which was also evident at the protein level 23 , suggesting that ZC3H45 has a role in stabilizing these transcripts during metacyclogenesis. In addition, in the CSD1 RNAi cell line 58 transcripts were downregulated and this included 29 novel transcripts that are potentially regulated by CSD1 through mRNA stability (Supplementary Table S13). The analysis of the coding region and the 3′ UTR of these 29 transcripts by MEME (Multiple Em for Motif Elicitation) identified a purine-rich element (RAR GAR RAKV), but further analysis will be required to investigate a potential interaction of CSD1 with these transcripts. The motif did not match any previously identified sequences involved in the developmental regulation of T. brucei genes and Gene Ontology analysis did not reveal any significant enrichment. Finally, the differentially regulated transcripts that appeared multiple times across the RNAi and KO cell lines (Fig. 4) did not expose specific gene expression regulatory networks.

Discussion
Developmental progression to infectious metacyclic trypanosomes in the tsetse fly is a critical factor for the transmission of African sleeping sickness and nagana. We gained insight into the regulation of this process with a targeted RNAi screen which revealed genes governing different steps of the metacyclogenesis pathway. Both positive and negative regulators emerged from our assays, and additionally, genes involved in controlling the transition of bloodstream forms to stumpy forms, as well as the differentiation from stumpy to procyclic forms had an effect on metacyclic cell production. Thus, the generation of metacyclics relies on a combination of factors that are required for two distinct processes in bloodstream form trypanosome biology: (i) transition to quiescence and (ii) differentiation from bloodstream to procyclic forms. The likely explanation for this observation is that in addition to establishment of quiescence 23 , metacyclic cells also thoroughly remodel their surface with the synthesis of a VSG coat and their transcriptome largely resembles the bloodstream form transcriptome 23,41 .
The dominant group of genes that exhibited effects on metacyclogenesis in our RNAi screen were signal transducers (Fig. 5). DYRK, PKA-R, AMPKα2, AMPKβ, AMPKγ protein kinases and DS-PHOS protein phosphatase have been implicated in quorum sensing during the development of stumpy bloodstream forms 3 and all had a pronounced effect on the generation of metacyclics. Genes that have been shown to regulate bloodstream to procyclic form differentiation were also represented by signal transducers and effectors. The protein phosphatase PTP1 42 , a key component of the signaling cascade during this transition, had an effect on mVSG coat synthesis, whereas RBP10, a protein largely responsible for the establishment and maintenance of bloodstream-form-type transcriptome [20][21][22] and a strong inhibitor of the bloodstream to procyclic differentiation [20][21][22] , did exhibit a positive effect on metacyclogenesis. Additionally, our RNA-Seq data revealed that repressor of differentiation kinase 2 (RDK2) was affected in all of the analyzed RNAi cell lines (Supplementary Table S16). This kinase has been identified as a critical regulator of the differentiation from bloodstream to procyclic forms 43 and it is the most upregulated kinase in metacyclics compared to procyclics 23 .
Two previously uncharacterized putative nucleic acid binding proteins emerged as potent regulators of metacyclogenesis. The cold shock domain-containing proteins CSD1 and CSD2 exhibited very strong effects on metacyclogenesis upon RNAi knockdown, albeit at different steps of the process. Additionally, CSD1 RNAi resulted in the largest number of downregulated mRNAs, exposing a number of transcripts potentially regulated by CSD1 through stabilization (Supplementary Table S13).
Mitochondrial metabolism has a pronounced effect on trypanosome development and it was previously shown to affect the type of procyclin on the surface of procyclic cells 44 , as well as the generation of metacyclics in vitro by inducible RBP6 expression 25 . Alternative oxidase channels electrons directly to oxygen and severely limits ATP production by oxidative phosphorylation. It is highly upregulated in both metacyclics and bloodstream forms and it was recently shown that it contributes to the generation of reactive oxygen species (ROS) which play a major role in the control of metacyclic production 25 . Surprisingly, the mRNA for alternative oxidase was not affected in any of the RNAi or KO cell lines for which we performed RNA-Seq and showed normal www.nature.com/scientificreports/ up-regulation (Table 3). However, metacyclogenesis in these cell lines was diminished greatly. This indicates that ROS may be required, but they are not sufficient to trigger efficient developmental progression. One clear example of a mRNA which was affected in all of the tested RNAi cell lines (Supplementary Table S16) is coding for the mitochondrial malic enzyme (Tb927.11.5450). In T. brucei this enzyme is believed to convert l-malate to pyruvate irreversibly and its down-regulation has a strong growth phenotype 45 . As pyruvate is the main end product of metabolism in bloodstream form trypanosomes, this suggests that the balance between L-malate and pyruvate (and/or the balance of NADP+ , the cofactor in the malic enzyme reaction, and NADPH in the mitochondrion) could possibly play an important role in regulating generation of metacyclics in addition to ROS. This is supported by the observation that pyruvate dehydrogenase E1α subunit (Tb927. 10.12700) showed fourfold higher phosphorylation in metacyclics relative to procyclics 23 , which in other organisms has been demonstrated to inhibit enzyme function, i.e. pyruvate is blocked from entering the TCA cycle and it is likely to be the end product of metabolism in developing metacyclics. Our targeted RNAi screen was focused primarily on nucleic-acid-binding proteins and protein kinases and phosphatases. Interestingly, the RNA-Seq data from selected group of these cell lines revealed that these components of gene expression regulatory networks are able to affect the abundance of a subset of transcripts very www.nature.com/scientificreports/ similarly. This suggests that different post-transcriptional regulons intersect in a defined pool of mRNAs, which exhibit developmentally regulated changes in expression and are under the direct or indirect control of these different RBPs. It is very likely that most or all of these transcripts are the direct targets for several of the RBPs studied here. Interestingly, among this subset is the mRNA for hnRNP F/H, yet another protein with a role in regulating the trypanosome life cycle. HnRNP F/H is highly upregulated in metacyclic 23 and bloodstream form parasites 46 in comparison to procyclics and has been shown to influence the efficiency of pre-mRNA transsplicing and mRNA stability in a life-cycle-dependent manner 46 .
In conclusion, our data suggest a considerable overlap between the genes that regulate the formation of stumpy bloodstream form trypanosomes and the genes that govern the development of metacyclic form parasites. This indicates that the vast majority of these gene expression regulators affect processes associated with cell-cycle arrest and decrease in biosynthetic capacity in cells transitioning to quiescence, a state of T. brucei development strongly correlated with transmission in either direction between the tsetse fly vector and the mammalian host.

Methods
Polysome analysis and RNA-seq. The RBP6 overexpression cell line 18 was induced for 48 h with doxycycline at a final concentration of 10 μg/μl. Un-induced and 48 h induced cells were incubated with cycloheximide at a final concentration 0.1 μg/μl for 1 h. Cytoplasmic extracts were prepared using detergent lysis and manual homogenization with polysome buffer (120 mM KCl, 20 mM Tris, pH 7.4, 2 mM MgCl 2 , 1 mM dithiothreitol, 10 μg/ml leupeptin and 0.1 μg/μl cycloheximide), containing 40 Units of RNase inhibitor and 1.2% Nonidet P-40. The cellular lysates were cleared by centrifugation at 14,000 rpm for 4 min. The concentration of nucleic acid in the lysate was measured by absorbance at 260 nm. The supernatants corresponding to 80 OD 260 were layered onto 15-50% linear sucrose gradients with 0.1 μg/μl cycloheximide in polysome buffer and centrifuged for 2 h at 36,000 rpm in a Beckman SW-41 rotor at 4 °C. The A 254 profile was recorded using the ISCO UA-6 detector. Fractions of 500 μl were manually collected from the gradient and prior to RNA extraction, 5 ng of in vitro transcribed firefly luciferase (Promega) were added to each fraction to allow for normalization. Each fraction was supplemented with 200 μg/ml of proteinase K, 1% SDS, 10 mM EDTA and incubated at 65 °C for 30 min, followed by RNA isolation using the phenol-chloroform method. Based on the sucrose gradient profile,   www.nature.com/scientificreports/ acid (Hepes), 2 mM ethylenediaminetetraacetic acid (EDTA) and 5 mM MgCl 2 , pH7.6), and resuspended in 500 μl Cytomix. Subsequently, 25 μg of linearized plasmid DNA was mixed with the solution and cells were pulsed twice at 1600 V with a time constant of 0.6 ms using a GenePulser Xcell (BioRad, Hercules, CA, USA). The selective drug, phleomycin, was added to the culture medium 24 h after electroporation at a final concentration of 2.5 μg/ml.

Scoring epimastigotes and metacyclic cells.
After 72 h of doxycycline induction of an RNAi cell line, epimastigote parasites were scored from a mixed population after DNA staining by Hoechst. These quantifications were performed with 4% paraformaldehyde fixed cells and/or live cells. Different trypanosome cell types were determined by their size, shape, and position of the kinetoplast relative to the nucleus and the posterior end of the cell.
Western blot analysis. Western blots were performed as described previously 47 . To stabilize the BARP protein, 0.5 mM of the metalloprotease inhibitor bathophenanthroline disulphonic acid (Acros Organics, catalogue no. AC164050050) was added to each 48-h RBP6 induced cell line and incubated for 24 h. These cells were then harvested for Western blot analysis. Rabbit antiserum against RBP6, BARP, and mVSG397 as well as mouse antiserum against paraflagellar rod (PFR) or elongation factor 1-alpha (EF-1α, clone CBP-KK1 from Sigma-Aldrich) was used at a 1:1000 dilution. Either PFR or EF-1α was used as a loading control for each Western blot. Pierce Protease and Phosphatase Inhibitor Mini Tablets (Thermofisher, catalog no. A32959) was dissolved into the lysis buffer following manufacturer's instruction and Western blots were probed with the rabbit polyclonal anti-p-AMPK (1:1000) antibody (Phospho-AMPKα (Thr172) (40H9) Rabbit mAb #2535, Cell Signaling technologies). The horseradish peroxidase-conjugated secondary antibody (Roche) was used at a 1:5000 dilution.
RNA preparation, RNA-Seq, read processing, and data analysis. Total RNA was prepared from approximately 2 × 10 7 to 5 × 10 7 un-induced cells, 1-, 2-, 3-, 4-, and 6-day induced cells that express both RBP6, as well as a RNAi hairpin against either AMPKα2, ZC3H45, CSD1, and CSD2, or are KO for RBP10. Three biological replicates were performed at least one week apart, and the RNA was prepared using the TRIzol reagent from Invitrogen according to the manufacturer's instructions. Isolation of poly(A) + mRNA, library preparation and sequencing on an Illumina HiSeq2500 platform were performed at the Yale Center for Genome Analysis as described previously 19 . Read processing and data analysis was performed as described 19 .

Data availability
RNA-Seq data from this study have been submitted to the NCBI Sequence Read Archive-SRA at http://www. ncbi.nlm.nih.gov/Trace s/sra/sra.cgi with accession numbers: PRJNA657432, PRJNA656316, PRJNA656556, PRJNA657048 and PRJNA657089.