The RNA-binding protein RBP10 controls a regulatory cascade that defines bloodstream-form trypanosome identity

Gene expression control in the pathogen Trypanosoma brucei relies almost exclusively on post-transcriptional mechanisms, so RNA binding proteins must assume the burden that is usually borne by transcription factors. T. brucei multiply in the blood of mammals as bloodstream forms, and in the midgut of Tsetse flies as procyclic forms. We show here that a single RNA-binding protein, RBP10, defines the bloodstream-form trypanosome differentiation state. Depletion of RBP10 from bloodstream-form trypanosomes gives cells that can grow only as procyclic forms; conversely, expression of RBP10 in procyclic forms converts them to bloodstream forms. RBP10 binds to procyclic-specific mRNAs containing an UAUUUUUU motif, targeting them for translation repression and destruction. Products of RBP10 target mRNAs include not only the major procyclic surface protein and enzymes of energy metabolism, but also protein kinases and stage-specific RNA-binding proteins: consequently, alterations in RBP10 trigger a regulatory cascade.

Their life cycles involve a series of unidirectional transitions, which are driven by 37 environmental changes as they move between mammalian and invertebrate hosts. 38 The kinetoplastid studied in this paper, Trypanosoma brucei, causes human sleeping 39 sickness in Africa and diseases of ruminants throughout the tropics. T. brucei multiply 40 as bloodstream form trypomastigotes in the blood and tissue fluids of mammals, 41 escaping immunity by antigenic variation of Variant Surface Glycoprotein (VSG) [1]. 42 in procyclic-form-specific mRNAs, and increases in bloodstream-form specific 233 mRNAs (Fig 3 B and S4 Table). A subset of mRNAs showed opposite reactions in 234 the bloodstream RNAi and procyclic over-expression experiments, consistent with 235 being immediate RBP10 targets (Fig 3 C). 236 We had previously attempted to identify RBP10-bound mRNAs, but the methods that 237 we used had insufficient sensitivity [37]. Since a subsequent study showed that 238 RBP10 is indeed bound to mRNA [40], we made another attempt. This time, we used 239 our bloodstream-form cell line in which all RBP10 bears a cleavable tag (S2 Fig A), 240 and compared bound and unbound mRNAs by RNASeq. 400 mRNAs were identified 241 (S5 Fig A and S5 Table, sheet 3). 12% of them are more abundant in procyclic forms 242 than in bloodstream forms (Fig 3 D). Consistent with the tethering results, 15% of the 243 RBP10 targets increased after rbp10 RNAi in bloodstream forms, but only 1% 244 decreased (Fig 3 E). Only 16 target mRNAs decreased after RBP10 expression in 245 procyclic forms (Fig 3 F); this low number might be due to the very short duration of 246 RBP10-myc expression. The RBP10-bound mRNAs that are more abundant in 247 procyclic forms and increased after rbp10 RNAi (Fig 3 G) are listed in Table 1: six of 248 them are also decreased after RBP10 expression in procyclics (Fig 3 H)  Overall, the RBP10-bound mRNAs are significantly less abundant than unbound 262 mRNAs (Fig 3 I). At the proteome level [16], the products of RBP10-bound mRNAs 263 are enriched in proteins that increase within 2-24h of the initiation of stumpy form 264 differentiation, or early during conversion to procyclic forms. There is no enrichment 265 for proteins that appear later in differentiation, or only in mature procyclics, and 266 unregulated proteins or those that are decreased in procyclic forms are under-267 represented (S5 Table, sheet 6). 268

RBP10-bound mRNAs share a UAUUUUUU motif 269
A comparison of the 3'-untranslated region (3'-UTR) sequences of the most reliably 270 RBP10-bound mRNAs with those of unbound mRNAs (S5 Table, sheet 5 and S1-S3 271 texts) revealed that the motif UA(U) 6 is highly enriched in the 3'-UTRs of RBP10 272 targets (Fig 4 A). It is present 1-9 times in 225 out of the 255 most strongly bound 3'-273 UTRs (Fig 4 B, S1 text). Transcriptome-wide, the number of UA(U) 6 motifs was 274 strongly associated with RBP10 binding (Fig 4 C) Table S5 sheet 5) were compared with those of mRNAs 283 that did not bind (<0.7x enrichment in both experiments) using DREME. Only 3'-284 UTRs annotated in tritrypDB were used (S1 text, S2 text, S3 text). The best-scoring 285 motif found is shown. 286 B) Numbers of UA(U) 6 motifs in 255 manually annotated bound 3'-UTRs (S1 text, S5 287 To test whether the 26mer was required for binding of the EP1 3'-UTR to RBP10, we 328 used cell lines that expressed chloramphenicol acetyltransferase (CAT) reporter 329 mRNAs with either the intact wild-type EP1 3'-UTR, or a version without the 26mer 330 (ep1∆26) (Fig 4 D) [34]. Co-immunoprecipitation of the CAT mRNA with RBP10 331 depended on the presence of the 26mer (Fig 4 E). Depletion of RBP10 by RNAi also 332 increased expression from the CAT-EP reporter but not from CAT-ep1∆26 (Fig 4 F). 333 These results demonstrated both that RBP10 binds to the EP 3'-UTR via the 26mer 334 sequence, and that the 26mer is necessary for RBP10-mediated regulation of an 335 mRNA containing the EP1 3'-UTR. 336

Expression of RBP10 determines bloodstream form identity 337
To test the role of RBP10 in differentiation, we first examined conversion from the 338 bloodstream form to the procyclic form. Our monomorphic bloodstream forms with 339 rbp10 RNAi, or treated with the differentiation-inducer cis aconitate, are unable to 340 multiply under procyclic culture conditions. We therefore used differentiation-341 competent trypanosomes (EATRO 1125, Antat1.1 strain). As expected [3], growth of 342 these cells at high density in methyl cellulose for three days resulted in expression of 343 the a -form marker PAD1 [54] and stumpy morphology; RBP10 was also reduced 344 (Fig 5 A, lanes 2 &3), consistent with its absence in the stumpy-form proteome [54]. 345 The cells were now treated with 6mM cis-aconitate in bloodstream-form medium at 346 27°C for 17h. Upon transfer to procyclic medium without cis-aconitate, the cells 347 started to grow as procyclics within 24h (Fig 5B).  proteins. As expected, cells without RNAi induction died (Fig 5 E). The RNAi 390 experiment had to be done without high density pre-incubation in methyl cellulose 391 since RNAi can only be induced in growing trypanosomes. As a control, we therefore 392 grew differentiation-competent cells to maximum density in liquid medium, and 393 treated them with cis-aconitate for 17h at 27°C [56] before transfer to procyclic 394 medium without cis-aconitate, also at 27°C. In these cells, weak RBP10 expression 395 persisted for 2 days (Fig 5 F, lanes 8-13) but the growth and procyclin expression 396 kinetics were similar to those seen after rbp10 RNAi (Fig 5 E, F, G). 397 The rather slow resumption of growth after both rbp10 RNAi and 17h cis-aconitate 398 treatment, and the persistence of RBP10 in the cis-aconitate experiment, suggested 399 that the population might be a mixture of differentiation-competent and incompetent 400 cells. Indeed, both differentiating cultures were mixtures of normal-looking, 401 proliferating, and dying or abnormal forms, and even after 3 days in procyclic culture, 402 about 9% of the cells in both cultures retained bloodstream-form morphology and 403 PGK localization (S6 Fig D). In bloodstream forms, the kinetoplast is at the posterior 404 end of the cell (Fig 5 G), which means that the distance between the posterior and 405 the kinetoplast (P-K) is much less than the distance from the kinetoplast to the We next inducibly expressed RBP10-myc in differentiation-competent procyclic 413 forms. For the results shown, we made bloodstream forms with inducible RBP10-414 myc, differentiated them into procyclic forms, grew these for more than 3 months, 415 and then induced RBP10 expression. However similar results were obtained by 416 differentiating bloodstream forms into procyclic forms, and then making three new 417 cell lines with the inducible RBP10-myc plasmid. After RBP10-myc induction, cell 418 growth was inhibited, as expected [37] (Fig 6 A). After 48h, expression of procyclin 419 was strongly reduced, alternative oxidase increased, and untagged RBP10 was 420 detected (Fig 6 B, C). At this point the cell population was very heterogeneous. Some 421 cells had normal procyclic morphology and expressed EP and GPEET procyclin; 422 some were clearly abnormal; while another subset had clear bloodstream-form 423 morphology ( Fig 6D). This was reflected in the P-K/K-N ratios, which ranged from 424 procyclic to bloodstream-form patterns (Fig 5 G, S7 Fig F). Flow cytometry analysis 425 revealed that about 50% of cells had decreased, or absent, procyclin expression ( Fig  426   6 E, F). Cells with bloodstream-form morphology were usually (but not always) 427 procyclin negative (S7 Fig G). Notably, VSG mRNA was present (Fig 6 C). We do not 428 know which VSGs are expressed in the bloodstream-form-morphology cells, but a 429 subset of them stained with a polyclonal antibody against VSG117 (Fig 6 D, G, H). 430 PGK staining in the bloodstream-form-like cells seemed reduced and somewhat 431 more punctate than in procyclic forms, but had not yet assumed a purely glycosomal 432 pattern (S6 Fig E). Remarkably, the RBP10-induced cells could now survive only if 433 they were transferred to bloodstream-form growth conditions ( Figure 6A The conversion of procyclic forms to bloodstream forms was highly reproducible. It 470 was not due to persistence of a few bloodstream forms in the procyclic cultures, 471 since survival always depended on induced RBP10 expression (Fig 6 A) procyclic forms to bloodstream forms, "jumping" past the non-dividing metacyclic 486 form. We therefore compared the mRNA changes after 6h RBP10-myc induction with 487 those found in salivary gland parasites (a mixture of epimastigotes and metacyclics) 488 (S4 Table,

496
Our results indicate that the presence of RBP10 defines the identity of a 497 trypanosome as a replication-competent bloodstream form. 498 RBP10 binds to procyclic-specific mRNAs with a UA(U) 6 motif in the 3'-UTR, 499 suppressing translation and causing mRNA destruction. However, many RNAs that 500 are bound by RBP10 did not change in either abundance, or in the percentage in 501 polysomes, after RBP10 RNAi or forced expression. As noted above, our assay did 502 not measure ribosome density, so it is likely that many changes in translation went 503 undetected. However, our results are completely consistent with work on other RNA-504 binding proteins: usually, only some of the bound mRNAs change after the protein is 505 removed (e.g. [57]). For most mRNAs, many different proteins can bind along the 506 same 3'-UTR -a sequence of 300nt is likely to bind at least twenty proteins -and it is 507 their combinatorial effects that determine mRNA behaviour. 508 RBP10 target mRNAs include several that have been studied in considerable detail, 509 including those encoding cytochrome complexes and the procyclins. However, the 510 direct effects of RBP10 are by no means sufficient to explain its far-reaching 511 influence, since many mRNAs that change in translation or abundance after RBP10 512 depletion are not bound by RBP10. This suggests that RBP10 depletion may initiate 513 a cascade of events. Our results show how this could occur. Notably, the direct 514 RBP10 targets include the mRNAs encoding three potential RNA-binding proteins, 515 ZC3H20, ZC3H21, and ZC3H22, These proteins are normally expressed only in 516 procyclic forms [15,16,58]. ZC3H20 binds to, and stabilises, at least two procyclic-517 specific mRNAs [58], of which one, MCP12, is not bound by RBP10 but was 518 increased after rbp10 RNAi. ZC3H22 suppresses expression in the tethering assay 519 [39], and is required for proliferation of procyclic forms [15,42]. Suppression of 520 ZC3H21 and ZC3H22 expression is clearly very important to bloodstream-form 521 trypanosome survival: their mRNA 3-UTRs contain, respectively, no fewer than 5 and 522 7 RBP10 binding motifs. 523 In addition to controlling ZC3H20-22, RBP10 binds to and represses mRNAs 524 encoding two procyclic-specific protein kinases (Tb927.8.6490, Tb927.11.15010) and 525 one procyclic specific protein phosphatase (Tb927.10.8050). Also bound by RBP10, 526 but with no detected effect of RNAi, are the transcripts encoding protein kinases 527 RDK1 and NRKA. NRKA [15] is up-regulated after the onset of stumpy form 528 differentiation; it is possible that RBP10 indeed regulates its translation but we did not 529 detect this because of the insensitivity of our assay. RDK1 is a differentiation 530 repressor [59] and is more abundant in bloodstream forms, making regulation by 531 RBP10 unlikely. Similar to RBP10, depletion of RDK1 primes bloodstream forms to 532 differentiate to procyclic forms. More than 20 mRNAs show similar changes after 533 RNAi targeting either RDK1 or RBP10, and 16 of these are both bound and regulated 534 by RBP10. Surprisingly, the RBP10 transcript was not significantly affected after 535 RDK1 RNAi; however, the protein level and phosphorylation status of RBP10 were 536 not determined. 537 The changes in RNA regulators, protein kinases and phosphatases, and perhaps 538 also the changes in metabolism that are directly induced by RBP10 loss (e.g.
[60]) 539 will trigger downstream effects on signaling pathways. Such secondary effects of 540 rbp10 RNAi must include the increase in PIP39 (Tb927.9.6090), a key regulator of 541 differentiation [55], and decreases in mRNAs encoding the RNA-binding proteins 542 PUF11, RBP9, ZC3H31, ZC3H46, ZC3H48, DRBD5 and HNRNPH (S1 Table,  It has long been known that stumpy forms cannot revert to long-slender bloodstream 565 forms, and recent results indicate that the differentiation of stumpy forms to 566 procyclics is an irreversible bi-stable switch [15]. We suggest that expression of 567 RBP10 governs a bi-stable switch during the transition from bloodstream forms to 568 stumpy forms, and from procyclic forms to bloodstream forms. The bi-stable 569 character can be explained by the fact that RBP10 directly suppresses procyclic- HMIi-9 medium supplemented with 10% fetal bovine serum at 37 o C with 5% CO2. 581 The procyclic forms were grown at 27 o C in MEM-Pros medium supplemented with 582 10% heat inactivated fetal bovine serum. Stable cell lines were generated and 583 maintained as described in [56,67] except that selection of the pleomorphic cells was 584 done with 8µg/ml hygromycin and 2µg/ml blasticidin. 585 The tetracycline inducible constructs for RBP10 RNAi and over expression were 586 described in [37]. All new constructs and oligonucleotides are in S6 Table. A 587 bloodstream-form cell line expressing all RBP10 with a tandem affinity purification tag 588 at the N-terminus was generated by inserting the tag at the 5' end of one 589 endogenous RBP10 open reading frame (pHD2506) and deletion of the second copy 590 (pHD2061). Plasmids for the tethering assays are in S6 Table. The tethering 591 constructs were separately transfected in a cell line constitutively expressing the CAT 592 reporter with 5 copies of boxB preceding the ACT 3' UTR. Expression of the fusion 593 protein was induced for 24h using tetracycline (100ng/ml) and the CAT assays 594 carried out as described in [25]. 595 Cross-linking and RNA immunoprecipitation 596 3x10 9 bloodstream form cells expressing TAP-RBP10 were irradiated using UV (400 597 mj/cm2), washed in cold PBS and the cell pellet snap frozen in liquid nitrogen. The 598 RNA immunoprecipitation was done as described in [24]. Briefly, the extracts were 599 incubated with the beads, and the unbound fraction was collected. After washing, 600 bound RBP10 was eluted using TEV protease. After digestion with 50µg proteinase K 601 at 42°C for 15 minutes to reduce cross-linked protein, RNA was isolated from both 602 the bound and unbound fractions using Trifast reagent (Peqlab, GMBH). To assess 603 the quality of the purified RNA, an aliquot of the sample was analysed by Northern 604 blotting and the blot hybridized with a splice leader probe. Total RNA from the 605 unbound fraction was depleted of ribosomal RNA (rRNA) using RNAse H as 606 To convert procyclic to bloodstream forms, AniTat 1.1 bloodstream form cells with an 702 inducible RBP10-myc construct (pHD2098) were differentiated to procyclic forms 703 using cis-aconitate as described above. The cells were cultured in presence of 704 hygromycin (8µg/ml) and phleomycin (0.2µg/ml) for more than 3 months to generate 705 well-established procyclic forms. RBP10-myc was induced using 100ng/ml 706 tetracycline. Marker proteins were detected by Western blot. VSG transcripts were 707 detected by semi-quantitative RT-PCR as previously described [61] using primers 708 CZ6308/CZ6309. To obtain growing bloodstream forms, RBP10-myc was induced for 709 48 hours, the cells were pelleted, resuspended (2x10 5 cells/ml) in HMI-9 medium and 710 then incubated at 37 o C with 5% CO 2 . The cell density was monitored for 6 days or 711 more; wild type or uninduced cells served as control. 712