Deep mutational scanning of the Trypanosoma brucei developmental regulator RBP6 reveals an essential disordered region influenced by positive residues

Abstract

To regain infectivity, Trypanosoma brucei, the pathogen causing Human and Animal African trypanosomiasis, undergoes a complex developmental program within the tsetse fly known as metacyclogenesis. RNA-binding protein 6 (RBP6) is a potent orchestrator of this process, however, an understanding of its functionally important domains and their mutational constraints is lacking. Here, we perform deep mutational scanning of the entire RBP6 primary structure. Expression of libraries containing all single-point variants of RBP6 in non-infectious procyclic forms and subsequent purification of infectious metacyclics supports the existence of an RNA-recognition motif (RRM) and reveal an N-terminal intrinsically disordered region (N-IDR). In contrast to the RRM, the N-IDR is more tolerant to substitutions; however, a handful of positions contain a third of all deleterious mutations found in the N-IDR. Introduction of positively charged residues in the N-IDR dramatically alters the normal metacyclogenesis pattern. Our results reveal an essential N-IDR, possibly playing a regulatory role, and an RRM likely involved in protein-RNA interactions.

Little is known about the presence and role of intrinsically disordered regions in RNA-binding proteins in trypanosomes. Here, the authors provide evidence for an intrinsically disordered domain in the essential N-terminal segment of RBP6.

Introduction

Trypanosoma brucei is a protozoan parasite that causes devastating diseases in humans and cattle and is transmitted by the tsetse fly vector. In T. brucei gene expression is regulated almost entirely at a posttranscriptional level and thus, RNA-binding proteins (RBPs) are particularly relevant factors that orchestrate posttranscriptional control^1,2. For example, in vitro overexpression of RBP6 in replicative T. brucei procyclic cells triggers a step-wise cascade of events that largely recapitulates the parasite differentiation occurring in the tsetse fly with the highest numbers of epimastigotes and metacyclic cells appearing at day 3 and 6 after induction, respectively (Fig. 1a)^3,4. Metacyclic parasites become infective through the expression a single metacyclic variant surface glycoprotein (mVSG)^4–8, whereas other trypanosome cell types found in the tsetse lack this feature and are unable to cause infection in the mammalian host. In the RBP6-induced in vitro metacyclic development, mVSG 397 is the most frequently expressed mVSG³. Although several negative and positive regulators of this developmental pathway have been recently identified^9,10, we still do not fully understand RBP6’s mechanism of action, nor do we know which segments of the protein are functionally important. Interestingly, expression of the RBP6 mutant Q109K (RBP6^Q109K) resulted in an accelerated parasite development, with metacyclics appearing as early as 48 h after induction and reaching a peak after 3 days¹¹. Furthermore, metacyclic cells generated through overexpression of this mutant can continue the T. brucei life cycle as they are capable of transforming into the mammalian bloodstream form in vitro¹¹.

Fig. 1. Deep mutational scanning of RBP6 and composition of mutant-RBP6 libraries.

a Scheme of RBP6-induced developmental progression. Procyclics expressing RBP6 develop into epimastigotes and then into metacyclics, which express the surface markers BARP and mVSG, respectively. b Overview of the DMS approach for obtaining and analyzing deleterious, neutral, or beneficial mutations in RBP6. c Number of mutated codons per clone in mutant-RBP6 entry libraries showing an average of 1.7 mutated codons per clone and a Poisson distribution of the number of mutations per clone (47 clones from n = 3 experimental replicates). d Mean of frequency of mutations across RBP6 in plasmid and T. brucei libraries at day 3 or 6 post-induction (n = 3 experimental replicates) in comparison to the mean of single replicates corresponding to plasmid, T. brucei, input and metacyclic WT controls. e Frequency of each type of mutation in mutant-RBP6 libraries or WT controls (mean of n = 3 experimental replicates for mutant-RBP6 libraries; one replicate shown for each WT library). Source data are provided as a Source Data file.

RBP6 was first identified through an in silico search for RBPs containing an RRM in the genome of T. brucei¹², although its RRM is atypical (see below). This domain is present in RBPs from all life kingdoms; and in eukaryotes is the most abundant classical RNA-binding domain (RBD)^13,14. The canonical RRM consist of 80 to 100 residues and contains four β-strands that form an antiparallel β-sheet packed against two α-helices in a β1-α1-β2-β3-α2-β4 configuration. Typically, the β-sheet and loops connecting the secondary structure elements provide a platform for direct interactions with RNA targets. In this conformation, the central β1- and β3-strands contain the highly conserved RNP2 and RNP1 sequences, respectively. In the archetypical interface between RRM and RNA, aromatic residues in position 2 of RNP2 and position 5 of RNP1 form stacking interactions with nucleobases of the RNA molecule. Additionally, the aromatic residue in position 3 of RNP1 makes hydrophobic interactions with two RNA residues; and finally, a positively charged amino acid in position 1 of RNP1 interacts with a phosphate group of the RNA backbone through a salt bridge¹³. Interestingly, intrinsically disordered regions (IDRs) are highly enriched in mRNA-interacting RBPs, with glycine (G), arginine (R), lysine (K) and tyrosine (Y) being more prevalent^14,15. Furthermore, several patterns of repetitive peptides, like serine-arginine (SR), arginine-glycine-glycine (RGG) and K-rich segments, were identified in IDRs of RBPs that are not necessarily evolutionarily related¹⁴. Some IDR repetitive sequences from RBPs have been found to be functionally important, to confer stability and to be associated with phase separation, in vivo protein aggregation, RNA binding and RNA folding^16–20. Thus, IDRs have emerged as important modules in RBPs that contribute highly to the diverse functional roles of these proteins.

Deep mutational studies focused on unveiling the link between RBP6’s primary structure and its function would allow us to provide insights into the mechanism of this master regulator and would reveal functionally relevant regions and their mutational limitations. Thus, we set out to map the functional landscape of RBP6’s primary sequence and measure the phenotypical consequences of every possible single residue change through deep mutational scanning (DMS)²¹. Our results support the existence of a single C-terminal RRM likely involved in direct RNA-target interactions and an N-terminal IDR (N-IDR) with a possible regulatory role. Furthermore, we proved the existence of strong combinatorial effects of single point mutations located in both the RRM and the N-IDR that were associated with an accelerated and increased production of metacyclics.

Results

Experimental strategy for DMS of RBP6

Our design (outlined in Fig. 1b) relied on the expectation that procyclic cells expressing RBP6 with deleterious mutations would fail to develop into metacyclics, whereas neutral or beneficial mutations would not impede metacyclic production. DMS libraries for assessing the effects of every possible single amino acid substitution in RBP6 were generated with PCR-based codon mutagenesis (Fig. 1b)^22,23, producing libraries comprised of approximately 5.8 × 10⁷ (±0.3 × 10⁷) clones (Supplementary Data 1) that contained, on average, a single codon mutation (Fig. 1c) introduced at any position in the entire coding sequence (Fig. 1d and Supplementary Fig. 1). Libraries for transfection contained an average of 5.6 × 10⁷ (±2.8 × 10⁷) clones (Supplementary Data 1) and yielded T. brucei libraries of 3.5 × 10⁵ (±0.3 × 10⁵) individual clones (Fig. 1b and Supplementary Data 1) in which almost all codon mutations were detected >15 times (Supplementary Fig. 2).

Changes in mutation frequencies after induction of mutant-RBP6 libraries and selection of metacyclic cells support the existence of a single C-terminal RRM in RBP6

Expression of WT-RBP6 in procyclic cells generates a peak of metacyclic forms at day 6 post-induction⁴ and expression of the mutant Q109K (RBP6^Q109K) results in an accelerated development with metacyclics reaching a peak at day 3 of RBP6^Q109K expression¹¹. As it is possible that certain mutations in our libraries will accelerate metacyclogenesis, we decided to purify metacyclics 3 days after induction of the T. brucei libraries in addition to the typical 6 days of RBP6 expression (Fig. 1b). We analyzed high-throughput sequencing (HTS) reads from mutant plasmids and T. brucei libraries, input and purified metacyclics and corresponding WT controls using dms_tools2²⁴. The overall frequency of mutations in RBP6 showed no significant differences between mutant plasmids and uninduced T. brucei libraries (Fig. 1e and Supplementary Fig. 3). Interestingly, while we did not observe an appreciable increase in the mutation frequency of the input cells induced for 3 days when compared to the uninduced library, a 6-day induction resulted in an average of 19% increase in the frequency of mutations (Fig. 1e and Supplementary Fig. 3). Purified metacyclics selected after 6 or 3 days of induction exhibited a decreased mutation frequency compared to the uninduced library by 23% or 12%, respectively (Fig. 1e and Supplementary Fig. 3). Additionally, input cells contained the highest numbers of stop codons which were drastically reduced in metacyclic fractions (Fig. 1e and Supplementary Fig. 3), indicating that deleterious mutations, like premature stop codons, inhibited the function of RBP6.

To identify sites subjected to high mutational constraints we analyzed changes in mutation frequencies across RBP6 relative to uninduced T. brucei libraries (Fig. 2, Supplementary Fig. 4 and Supplementary Data 2). The most pronounced reductions of mutation frequencies [log2(FC) < −1.0] in the metacyclic fractions and highest increments on mutation frequencies [log2(FC) > 0.5] in the input cells induced for 6 days, were located exclusively in the C-terminal portion of RBP6 (Fig. 2, Supplementary Fig. 4 and Supplementary Data 2). To shed light into the biological meaning of these findings, we performed in silico analyses of the primary structure of RBP6. Alignment of RBP6 CDSs of the Trypanosoma genus showed that their overall identity ranges between 47% and 87%, with T. vivax and T. grayi sequences sharing the lowest identity, and T. brucei and T. congolense being the most identical (Supplementary Fig. 5a). The most conserved region comprised residues 135 to 223 of the C-terminal portion; in contrast, positions 1 to 134 and 224 to 240 showed mild levels of conservation (Supplementary Fig. 5a). The only predicted domain in 19 RBP6 homologs of the Trypanosomatidae family was a putative RRM in the most conserved part of the protein spanning positions 140 to 218 in T. brucei (Supplementary Fig. 5a). The RRM of T. brucei and T. congolense is connected to the N-terminal portion through a variable linker containing a tract of 12 or 5 consecutive glutamine residues, respectively. However, this tract is absent in T. vivax and is located upstream in T. grayi and T. theilieri (Supplementary Fig. 5a). The N-terminal portion of T. brucei RBP6 and the last 21 residues of the C-terminus do not appear to contain any predictable structural modules. However, the N-terminal portion was enriched in glutamines, prolines, and serines, with composition percentages of 23.7%, 10.8% and 8.6%, respectively (Supplementary Fig. 5b). As these residues have been classified among the top disorder-promoting amino acids²⁵, we sought to score the disorder propensity across RBP6. The computational tool flDPnn²⁶ predicted residues from two discrete regions of RBP6 as disordered: a C-terminal intrinsically disordered region (IDR) from position 225 to 240, and a longer IDR comprising most of the residues from position 1 to 135, except V78, T79 and P80 (Supplementary Fig. 5c). As the latter IDR contains the entire N-terminal portion of RBP6, including the Q-tract and fifteen amino acids (positions 121 to 135) of the so-called C-terminal portion, we henceforth refer to this region (positions 1 to 135) as N-terminal IDR (N-IDR). Thus, in silico analyses suggested the presence of a C-terminal RRM in the most conserved portion of the protein, and two IDRs in the less conserved regions, a C-terminal IDR (C-IDR) and an N-IDR. Given that the most pronounced changes in mutation frequencies observed with our DMS approach overlapped with several residues of the predicted RRM (Fig. 2, Supplementary Fig. 4 and Supplementary Data 2), these results strongly supported the presence of this domain in RBP6 and indicate a high selective pressure against mutations in this protein module.

Fig. 2. Changes of mutation frequencies throughout RBP6.

Plot showing the log2 of the fold-change (FC) of the mean frequency of mutation across RBP6 in input and metacyclic fractions induced for 6 (top) or 3 days (bottom), relative to uninduced T. brucei libraries. Means derived from 3 biological replicates (n = 3). Diagram on top depicts RBP6 domains. Source data are provided as a Source Data file.

An essential role for the N-IDR in RBP6 function

While most positions in the N-IDR displayed minor changes in mutation frequencies when compared to uninduced libraries, 14 positions showed polarized results in the metacyclic fractions, i.e., significant increments of mutation frequencies at day 3, but a decrease at day 6 of selection (Fig. 2 and Supplementary Data 2). Intriguingly, we detected a trend of enrichment for positively charged amino acids spanning almost all residues of the N-IDR at day 3 of metacyclic purification that disappeared upon purification of metacyclics at day 6 of induction (Fig. 3, Supplementary Fig. 6 and Supplementary Data 3). Among these, 25 substitutions in 20 positions showed the most statistically significant results with values of differential selection < −0.84 at day 6 of metacyclic purification and scores > 0.89 at day 3 of selection (Supplementary Data 3) and comprised the following mutations: F2K, F2R, I16K, Y22K, M24K, M25K, F36R, A37K, A43K, L45K, L45R, Y48K, Y50K, P53K, E57K, F61K, Y73K, Y73R, V74K, M76K, M76R, V78K, V78R, V81K and M86K (Supplementary Data 3). Although at day 3 of purification the p values of these mutations did not reach a confidence cutoff, we identified one biological replicate that consistently showed lower values of differential selection (Supplementary Data 3). We obtained the means and p values of the two more similar replicates and filtered out A37K, L45K, P53K, V74K, V81K and M86K that still showed p values > 0.05 and FDR > 0.15 (Supplementary Data 3). As most single point mutations in the N-IDR portion of RBP6 did not show scores indicating deleterious effects on development (Fig. 3, Supplementary Fig. 6 and Supplementary Data 3), it seems that the amino acid content, rather than the presence of distinct domains, plays a dominant role in this portion of the protein. These results suggested that the introduction of positively charged residues in the N-IDR, in a similar fashion to RBP6^Q109K, could trigger accelerated development resulting in production of metacyclics at day 3 of induction, but absence or low numbers of this life cycle stage at day 6 of forced RBP6 expression.

Fig. 3. Effects of positively charged amino acids in the mutation tolerant N-IDR of RBP6.

a Heatmaps showing differential selection scores for every mutation in the N-IDR at day 6 (top) or 3 (bottom) after purification of metacyclics. Gray rectangles indicate WT residues at each position. The scheme on top delimits the position of the Q₍₁₂₎-tract. Scores derived from 3 biological replicates (n = 3). b Logo plot indicating amino acid preferences from position 2 to 30 of RBP6 in the metacyclic fraction purified at day 3 post-induction. A clear trend of preferences for positively charged residues is observed. Scores derived from 3 biological replicates (n = 3). c Densitometry analysis of expression of RBP6 (left) or mVSG 397 (right) in WT- or mutant-RBP6 cell lines carrying a triple or single lysine substitutions of residues located in the N-IDR. Western blots used for this analysis are shown in Supplementary Fig. 7. Source data are provided as a Source Data file.

To test this hypothesis, we generated cell lines that independently expressed the single-point mutants F2K, M24K, F36K, L45K, Y48K and M76K in the N-IDR. Strikingly, induction of these mutants for 6 days showed an altered pattern of differentiation (Fig. 3c and Supplementary Fig. 7). First, we observed higher levels of expression of the metacyclic marker mVSG 397 at 48 h and 72 h after induction when compared to the WT control Second, the mVSG 397 signals peaked between day 2 and 3 of induction and significantly decreased at later days. The highest levels of expression of mVSG 397 obtained with the mutant variants were from ~25 to ~80% lower than the highest levels of expressions resulted from the WT-RBP6 induction (Fig. 3c and Supplementary Fig. 7). Taken together, these results confirmed that single point substitutions with positively charged amino acids in the N-IDR promote accelerated but diminished developmental progression. We next investigated the effect of combining three single-point mutations by generating a cell line that inducibly expressed RBP6 with F2K, L45K and M76K. Surprisingly, although the levels of expression of this mutant were slightly lower than those obtained with WT-RBP6 (Fig. 3c and Supplementary Fig. 7), we observed early expression of mVSG 397, which was detected as soon as 48 h after induction and, in contrast to single lysine substitutions, we noticed a sustained and enhanced production of metacyclics at later days of the 6-day time-course experiment (Fig. 3c and Supplementary Fig. 7). Together, these results demonstrated an additive effect of introduced positively charged residues in the N-IDR in altering the typical progression of development and suggested that this region plays an important regulatory role.

The mutation-tolerant N- and C-IDRs contain hot-spot residues under high selective pressure

To distinguish mutations causing a significant effect on RBP6’s function, we set as cutoffs scores of differential selection above 0.88 or below −0.84 and p values < 0.05. Using these criteria, 25 out of 304 (~8%) amino acid substitutions in the C-IDR and 226 out of 2318 (~10%) possible amino acid changes in the N-IDR showed significant negative scores in metacyclic fractions and were classified as deleterious (Fig. 3a, Supplementary Fig. 6 and Supplementary Data 3). However, it was surprising that despite the observation that these regions tolerated a much higher proportion of mutations than the RRM (Supplementary Fig. 6), just a limited number of residues (F2, R27, R44, P47, Q64, Y73, Q75, M76 and R77) contained ~33% of all deleterious mutations in the N-IDR and only two positions (R236 and A239) made up for 56% of all deleterious mutations in the C-IDR (Figs. 3a and 4b, Supplementary Fig. 6 and Supplementary Data 3). Interestingly, some of these residues (R27, R44, P47, Q75 and R77) showed a significant decrease of mutation frequencies in both metacyclic fractions and input populations (Fig. 2 and Supplementary Data 2). In fact, 22 mutations showed significant negative values of differential selection for both days of metacyclic purification and negative scores in the input libraries, nine of which showed p values < 0.05 (Supplementary Data 3). Surprisingly, from this subset only substitution L147W resided in the RRM while the remaining mutations were found in the N-IDR: 6 mutations in Q75; 3 in R77; 2 in R44, P47, and G60; and 1 in T14, R27, Q28, A43, Q56 and Q64 (Fig. 4b and Supplementary Data 3). These mutations consisted mainly of substitution to aromatic amino acids (9 mutations to W, 5 to F, and 1 to Y) or aliphatic residues (3 substitutions to M, 2 to I and 1 to L, or V) (Fig. 4b and Supplementary Data 3). Taken together, these data indicated the presence of residues subjected to high selective pressures in the IDR regions that may play key functional roles during development.

Fig. 4. Most deleterious mutations in the mutational tolerant N-IDR are found in a handful of residues.

a The N-terminus of RBP6 has limited conservation. RBP6 sequences from five Trypanosoma species were aligned with the MUSCLE software. The alignment was used to generate a sequence logo which is displayed bellow the alignment and shows the frequency of each amino acid. b Logo plots showing negative scores of differential selection in purified metacyclics (top) and input cells (bottom) induced for 6 days. Residues containing ~33% of all deleterious mutations in the N-IDR are indicated with arrows connected by dashed lines in (a) and (b). Scores derived from 3 biological replicates (n = 3). c Expression of WT-RBP6 and variants R27V, P47V or Q75V, before and after 1 day of induction with doxycycline at 10 ng/µl. A cross-reacting band is used as loading control. d Western blot analyses that show expression of WT- and mutant-RBP6 variants at day 1 post-induction (p.i.) with varying concentrations of doxycycline. At day 5 p.i. the presence of metacyclic cells in the cultures induced with the indicated concentration of doxycycline was assessed by abundance of mVSG 397. Paraflagellar rod (PFR) protein was used as loading control. e Western blot analyses to detect mVSG 397 in cultures induced with 10 ng/µl of doxycycline to express WT-RBP6 or the indicated mutants for 6 days. A cross-reactive band was used as loading control. Blots shown in c–e were repeated at least one time obtaining consistent results. f Growth curve of four clonal cell lines (n = 4 biological replicates) induced (Dox +), or not (Dox −), to express RBP6^R27V. Cells were counted and diluted daily to 2 × 10⁶ cells/ml for 6 days. Each time point represents the mean cumulative cell density multiplied by the dilution factor and error bars represent standard deviation. Source data are provided as a Source Data file.

To test if single amino acid substitutions in the N-IDR were sufficient to disrupt RBP6’s function, we generated cell lines that independently expressed RBP6 with mutations R27V, P47V and Q75V. According to our screen, these mutations showed the most negative values of differential selection at day 6 of metacyclic selection (scores < −4.98) and strong negative values at day 3 of purification (values < −4.25, Supplementary Data 3). Furthermore, they also displayed negative values in the input libraries induced for 6 days (Fig. 4b and Supplementary Data 3). The expression of RBP6^R27V, RBP6^P47V and RBP6^Q75V showed a ~60% decrease in protein abundance relative to WT levels after 24 h of induction (Fig. 4c). To match the levels of expression of the mutants, we induced WT-RBP6 cell lines with 80-fold lower concentration of doxycycline (125 ng/ml). Under these conditions, WT-RBP6 was expressed at the mutant protein levels with only a slight decrease of mVSG 397 expression (Fig. 4d). In contrast, overexpression of the three mutant proteins resulted in undetectable levels of mVSG 397 over 6 days of expression (Fig. 4d, e), confirming that the N-IDR is necessary for in vitro development and that conserved N-IDR residues are likely involved in regulatory processes. Because these mutants also showed negative differential selection scores in input populations induced for 6 days, we investigated the effect of RBP6^R27V on the growth rates of procyclic cells. We found a reduction of proliferation relative to uninduced cells that is readily detected after 48 h post-induction (Fig. 4f), indicating that in addition to impeding the function of RBP6, the R27V mutation negatively impacts the growth rate of procyclic cells.

Most mutation-intolerant RBP6 residues are located in the β-strands of the RRM

The tertiary structure of RBP6’s atypical RRM predicted by AlphaFold2 displayed a non-canonical β1-α1-β2-β3-α2-β3’-β4 configuration (Fig. 5a). In this model, instead of the four canonical β-strands that form the antiparallel β-sheet packed against the two α-helices, an additional short β-strand is formed in loop 5 that, following previously used nomenclature¹³, we refer to as β3’. In this divergent conformation, β4 is located between the central β1-strand and the external β3’-strand (Fig. 5a). Canonically, the central β-strands 3 and 1 contain the conserved motifs RNP1 and RNP2, respectively. Commonly, the aromatic residues in position 2 of RNP2 and positions 3 and 5 of RNP1 are exposed on the surface of the β-sheet to directly interact with RNA targets^13,27. The RNP motifs of RBP6 are highly conserved across the Trypanosomatidae family (Fig. 5b). However, while the RNP1 motif of RBP6 completely matched the consensus sequence, the highly conserved sequence in the β1-strand only partially resembled the RNP2 motif and position 2 contained an isoleucine residue instead of an aromatic amino acid (Fig. 5b). The regions with the lowest number of functionally neutral mutations were the β-strands of the RRM, with β1 and β3 displaying the lowest proportions: β3’, ~54%; β2, ~29%; β4, ~21%; β3, ~14% and β1, ~6% (Fig. 5c and Supplementary Data 3). In contrast, 62% and 69% of all possible mutations in the α1- and α2-helices of the RRM did not reach the functional significance cutoffs (Fig. 5c and Supplementary Data 3). From all possible substitutions within the RRM that do not comprise α-helices or β-strands ~56% were neutral (Fig. 5c and Supplementary Data 3). The highest percentages of deleterious mutations per motif were observed within the β-strands of the RRM, which consistently showed higher percentages than α-helices or other residues within this domain (Fig. 5c, d and Supplementary Data 3): Compare ~94% (89/95), ~85% (130/152), ~79% (105/133), ~71% (94/133) and ~43% (33/76) in β1, β3, β4, β2 and β3’ to ~34% (72/209) and 27% (57/209) in α1 and α2, respectively. Despite a modest 41% (281/684) of all possible mutations of miscellaneous residues in the rest of the RMM classified as deleterious, there were some residues flanking the predicted β-strands with high accumulation of deleterious mutations, e.g., positions 140 and 146-148 flanking β1, and 216-220 flanking β4 (Fig. 5c and Supplementary Data 3). The overlay of cumulative differential selection per residue on the tertiary structure of RBP6’s RRM showed that the highest levels of deleterious effects are caused by mutations in the β-sheet and buried residues that are an integral part of the hydrophobic structural core of the domain (Fig. 5d). In contrast, cumulative differential selection scores pointing towards beneficial effects were located in loops and solvent exposed residues of α -helices (Fig. 5e). Combined, these results supported the existence of a C-terminal RRM and indicated that β-strands and surrounding residues likely play a direct role in protein-target RNA interactions. As in vivo RNA binding substrates of RBP6 have only been predicted bioinformatically²⁸, we decided to use an unbiased approach to confirm that RBP6 can directly contact RNA. We used a randomly chosen RNA from a pool of 43 sequenced clones after 7 rounds of Systematic Evolution of Ligands by EXponential enrichment (SELEX)²⁹ for in vitro transcribed RNA molecules containing a variable 50-nt segment with affinity for recombinant maltose-binding protein (MBP)-RBP6. After incubation of internally ³²P-labeled RNA with MBP-RBP6 and UV irradiation, a covalent bond between RBP6 and an RNA fragment was detectable as radiolabeled protein (Fig. 5f), demonstrating the close proximity of bound RNA substrate and RBP6. Along with core residues that contribute to the structural integrity of the domain, positions in the predicted β-sheet were subjected to the highest mutational pressure in the molecule. Furthermore, we found only 16 sites that consistently showed preference for the WT residue at both days of metacyclic purification (Supplementary Fig. 8 and Supplementary Data 4), suggesting that our cell culture-based experimental system is likely less restrictive compared to the demanding environmental conditions in the tsetse, under which the protein evolved. All residues from this last cohort were found in the RRM and the majority (10 positions) were located in β-strands; 1 in β2 and 3 in each: β1, β3 and β4 (bottom logos in Fig. 5b, Supplementary Fig. 8 and Supplementary Data 4). In contrast, α1- and α2-helices contained 2 and 1 residues, respectively (Supplementary Fig. 8 and Supplementary Data 4). The remaining 3 were G164 in loop 2; and N218 and P219, located 3 residues downstream of β4 (Supplementary Fig. 8 and Supplementary Data 4). The WT residues showing the strongest preferences at both days of purification were Q141 in β1 (position −1 in RNP2), F185 and Y187 in β3 (positions 5 and 7 in RNP1) (bottom logos in Fig. 5b, Supplementary Fig. 8 and Supplementary Data 4). F185 is the well-conserved residue in position 5 of RNP1 that forms stacking interactions with a nucleobase of the RNA targets in other RBPs^13,27,30. While not all positions in the RNP motifs showed a strong preference for the WT residue, a clear preference for amino acid sharing similar physicochemical properties to the WT residue was observed across these motifs (Fig. 5b, Supplementary Fig. 8 and Supplementary Data 4). Therefore, our results suggested that as in other RRMs, the β-strands contain the key residues responsible for sequence-specific interactions with target transcripts and structural integrity of the domain.

Fig. 5. The strongest mutational constraints are found in the RRM of RBP6.

a, d, e Alphafold prediction of the tertiary structure of RBP6’s RRM shown as ribbon diagram (left) or space-filled model (right). A view of a 180° rotation is shown below. a Indicates distinct α-helices and β-strands. The overall negative or positive scores of differential selection per residue in metacyclics purified at day 6 or 3 post-induction is shown in (d) or (e), respectively. b Analysis of RNP motifs. Consensus sequences of RNP2 and RNP1, found in β1- and β3-strands, are shown on top. Positions that directly interact with RNA molecules are highlighted in red. X: any amino acid. Sequence logos show the frequency of residues per each RNP position in a curated master alignment of 347 RRM sequences (top), or in an alignment of 19 RBP6 trypanosome homologs (middle), and amino acid preferences in RNPs at day 3 of purification of metacyclics (bottom). c Number of substitutions with significant scores (p < 0.05) of positive (above 0.88) or negative (below −0.84) differential selection per residue obtained with input cells, or pure metacyclic fractions, at days 3 or 6 post-induction. p values were derived from a two-sided t-test in which the means of differential selection scores from three biological replicates (n = 3) of input or metacyclic mutant libraries were contrasted against the respective mean from plasmid libraries. Only residues 121 to 240 are shown to highlight mutational constraints in the RRM. A scheme on top delineates the different elements of the secondary structure. L1 to L5: loop 1 to loop 5. f UV crosslinking assay to demonstrate RBP6’s ability to directly contact RNA. Internally ³²P-labeled RNA was mixed with the indicated proteins, with or without UV treatment, followed by degradation of the RNA by RNase A and SDS-PAGE. g Western blot analyses to detect mVSG 397 expression upon a 6-days induction of WT- or mutant-RBP6 variants L147R and F185L. Blots shown in f and g were repeated at least one time obtaining consistent results. h Quantification of expression of WT- or mutant-RBP6 variants L147R and F185L after 1 day of induction. Bars show the mean of n = 8 experimental replicates and error bars represent standard deviation. A representative western blot to detect RBP6 in each cell line is shown below. Asterisks in g and h indicate a cross-reactive band that served as loading control. Source data are provided as a Source Data file.

To confirm the results from our DMS approach, we tested mutations L147R and F185L located at position 6 and 5 of RNP2 and RNP1, respectively. Our criteria for selection was based on: (1) both mutations were significantly depleted at both days of metacyclic purification (Supplementary Fig. 6 and Supplementary Data 3); (2) these positions were highly intolerant to mutations (Figs. 2 and 5b); (3) F185L was enriched in the input population induced for 6 days (Supplementary Fig. 6 and Supplementary Data 3); and (4) F185 was one of the top 3 WT residues with strongest amino acid preferences at both days of metacyclic selection (Fig. 5b, Supplementary Fig. 8 and Supplementary Data 4). We induced the expression of these mutant versions of RBP6 over the course of 6 days and surveyed the production of metacyclics using as read-out the expression of mVSG 397. Contrary to the WT control, we did not detect the expression of mVSG 397 with either mutant (Fig. 5g). However, the levels of RBP6^L147R and RBP6^F185L after 24 h of induction were drastically lower than those observed with WT-RBP6 (Fig. 5h). As the levels of expression of WT-RBP6 are directly correlated with the production of metacyclic cells³, we couldn’t conclude that the lack of metacyclics is a direct effect of inhibiting RBP6-RNA interactions. However, it was clear that these mutations dramatically affected the expression levels of the protein, which alone can explain the lack of developmental progression.

Substitution of the RRM residue E165 in RBP6 with aromatic amino acids enhances cell differentiation

In purified metacyclic samples we observed a tendency of enrichment for substitutions of negatively charged residues in the RRM to aromatic residues that explains the consistent increments of mutation frequencies in D150, E165 and E199 at both days of purification (Fig. 2, Supplementary Fig. 4 and Supplementary Data 2). While D150 and E165 are located in the loops flanking α1-helix, E199 is found inside α2-helix (Supplementary Fig. 5a). Importantly, substitutions of E165 with aromatic residues showed the highest values of differential selection in all of our data sets: we obtained a differential selection score of ~3.5 for E165Y; ~3.4 for E165W and ~3.0 for E165F, all with p values < 0.05 (Supplementary Fig. 6 and Supplementary Data 3). To confirm these results, we investigated the effects of expressing the single-point mutants E165F, E165W and E165Y. Surprisingly, inducible expression of constructs with aromatic residues in procyclic cells over a 6-day time-course (Fig. 6 and Supplementary Fig. 7) showed an accelerated and dramatically increased mVSG 397 expression when compared to the WT-RBP6, RBP6^E165A and RBP6^E165K constructs, which was readily detected as soon as 48 h after induction (Fig. 6 and Supplementary Fig. 7). Intriguingly, expression of all constructs except for RBP6^E165A, sped up and increased the signal of the short epimastigote marker BARP (Fig. 1a)³¹, when compared to the WT-RBP6 construct (Fig. 6 and Supplementary Fig. 7). These results confirmed that substitutions of E165 with aromatic residues enhance the magnitude and pace of RBP6’s function.

Fig. 6. Beneficial effects of substitutions of E165 with aromatic residues.

Densitometry analysis of the expression of RBP6, BARP and mVSG 397 upon a 6-day induction of WT- or mutant-RBP6 with the indicated substitutions of E165 and F2. Western blots used for this analysis are shown in Supplementary Fig. 7. Source data are provided as a Source Data file.

Because we identified independent mutations in both the N-IDR and RRM of RBP6 which stimulate its function, we sought to characterize the effects of combining these substitutions. Therefore, we generated a cell line that inducibly expresses an RBP6 variant bearing substitutions F2K (in the N-IDR) and E165Y (in the RRM). Surprisingly, induction of this variant resulted in the highest levels of mVSG 397 observed in all of our in vitro overexpression experiments (Fig. 6 and Supplementary Fig. 7). However, as we observed a peak of mVSG 397 expression at days 3 and 4 after induction of this double mutant, the pattern of mVSG 397 signal better approximated that obtained by overexpression of the RBP6^E165Y mutant, rather than the one observed with RBP6^F2K (Figs. 3c and 6 and Supplementary Fig. 7). These results pinpointed mutations in the N-IDR and RRM that together can confer strong effects and improve RBP6’s function during in vitro metacyclic development.

Discussion

Notwithstanding the proven importance of RBP6 as a key factor driving metacyclogenesis in T. brucei, the regions important for this function and their mutational constraints were previously undetermined. Here, we mapped the functional landscape of full-length RBP6 through DMS. Supported by a recently performed DMS of the T. brucei RNA editing protein KREPB4³², our results demonstrated that DMS approaches are feasible in trypanosomes for the assessment of the effects of all possible single amino acid mutations in a protein in an in vivo context. Most mutation-intolerant residues in T. brucei RBP6 resided in the β-sheet of the RRM, specifically in the β1 and β3-strands that carry the RNP motifs. Furthermore, residues in α-helices whose side chains are oriented toward the hydrophobic core of the domain showed lower tolerance to mutations than solvent-exposed moieties. A comparison between our DMS results and the enrichment scores obtained for the RRM2 of PABP1 from S. cerevisiae³³ show very similar mutational constraints in both RRMs (Supplementary Fig. 9). Together, these observations support the existence of a single RRM in RBP6 and suggest that direct RNA interactions mediated by the RNP motifs are critical for the function of this RBP. Furthermore, the high mutational constraints found in both RRM2 of PABP1 and the RRM of RBP6 highlight the importance for the evolutionary conservation of the archetypical structure of this domain in distantly related organisms.

The most conserved structural elements among RBP6 homologs in trypanosomes are the β1- and β3-strands of the RRM, which contain the highly conserved RNP2 and RNP1 motifs, respectively (Fig. 5b and Supplementary Fig. 5a). As supported by our results and the predicted tertiary structure of RBP6, it is possible that, similarly to RNP1s of other RBPs, hydrophobic residues in positions 2, 4 and 6 of RBP6’s RNP1 preserve the hydrophobic core of the RRM, while positions 1, 3 and 5 could play a critical role through direct contact with RNA targets. Despite the highly conserved hallmarks of sequence and tertiary structure of RRMs, small variations of this domain allow for a high degree of functional plasticity^13,34,35. The RNP2 motif of RBP6 only partially matched the consensus sequence, with positions 2 and 5 diverging from the canonical F/Y and N residues, respectively. A logo from a curated alignment of 347 RRM sequences³⁴ showed that, after F and Y, I is the third most frequent amino acid in position 2 of RNP2 (Fig. 5b), reflecting that although rare, this residue can be found in this position of RRMs as in the case of RBP6. Furthermore, N and G are the two most frequent residues in position 5 of RNP2 (Fig. 5b), however, the homologous residue in RBP6 is Y and we observed strong preferences for aromatic residues in this position (Fig. 5b). Interestingly, N is found right upstream of position 5 in RNP2, suggesting that this residue could play a similar role to that of the more conserved N residue in RNP2 of other RRMs. Strikingly, Q141 which is predicted to be the first amino acid in the β1-strand and is located at position −1 of RNP2, was one of the most highly preferred WT residues, but as this position is much less conserved, it does not appear in previously reported consensus sequences (Fig. 5b). Intriguingly, the side chains of the residues in RNP2 that diverge from the consensus sequence (positions −1, 2, 4 and 5) were predicted to be exposed to the RNA-interacting surface of the β-sheet (Supplementary Fig. 9b). In contrast, all conserved positions in RNP2 represented hydrophobic residues predicted to be buried in the structure. Thus, while the divergence of the RNP2 sequence in RBP6 might have a role in the specificity and affinity of distinct RNA-protein interactions, these functional implications are without a precedent. The highly conserved residues are likely structurally involved in the formation of the hydrophobic core of the RRM domain.

We observed a tendency for beneficial effects for substitutions of negatively charged residues in the RRM with aromatic amino acids. Although the extent varied among mutations, they explain high levels of mutation frequencies in D150, E165 and E199 in purified metacyclics and we confirmed that substitution of E165 with aromatic amino acids enhanced in vitro metacyclogenesis. Interestingly, these residues are found in loops 1 and 2 of the RRM and the solvent-exposed face of α2-helix. Residues outside of the β-sheet have been shown in other RRMs to be involved in RNA interactions. For example, in the splicing regulator Rbfox, while the last three nucleotides of the target sequence UGCAUGU are recognized by the β-sheet, the first four nucleotides (UGCA) contact loops connecting β1 with α1, β2 and β3, and α2 with β4³⁶. Therefore, although it is possible that beneficial substitutions for aromatic residues in RBP6 might promote stability of the protein through inter- or intramolecular interactions with other residues or protein partners, a direct role in protein-RNA contacts cannot be ruled out.

In contrast to domains with stable structures, IDRs present a lower level of conservation, are more tolerant to mutations, and their diverse amino acid composition has been associated with different biological functions^37,38. It has been recently proposed that as discrete regions inside an IDR can have independent functional, chemical, and structural properties subjected to different evolutionary pressures, they should be referred to as distinct domains coexisting in a disordered context³⁹. IDRs can operate as intramolecular linkers, regulatable binding scaffolds for protein or RNA interactors, intracellular sensors of physicochemical changes, and nucleation factors driving granule formation^37,38,40. Mammalian mRNA interacting-RBPs are enriched in IDRs¹⁴ and in silico analyses have shown that 30 out of 230 RRM-containing RBPs in the human genome contain highly disordered regions (the remaining 200 do not), with IDRs being more prevalent in RBPs containing a single RRM⁴¹. Additionally, approximately 50% and 30% of RNA-binding sites identified by proteome-wide screens were located in disordered regions of RBPs from H. sapiens and D. melanogaster, respectively^42,43. Our findings suggest that the presence of functionally important IDRs in RBPs can be extended to trypanosomes. We showed that although the slightly-conserved N-terminal portion of RBP6 is highly tolerant to mutations, 9 positions contained around a third of all deleterious substitutions in this region. Additionally, introduction of positively charged amino acids across the N-terminal portion highly altered the pattern of metacyclic development, and substitutions of all R residues with aromatic or aliphatic amino acids resulted in scores pointing to deleterious effects (Fig. 4). These results indicated an essential role for the N-terminal segment of RBP6, influenced by positively charged residues and provide strong evidence for an intrinsically disordered state. A number of motifs in IDRs have been directly linked to biological functions. In IDRs of RBPs, sequence repeats containing positively charged amino acids like RGG, SR, R- and K-rich motifs have been found to be engaged in non-canonical interactions with RNA^42,44. It has been hypothesized that similar to lysines in histone tails which interact with DNA⁴⁵, poly(K) patches in IDRs of RBPs could interact with the phosphate backbone of RNA targets¹⁴. However, the low mutational tolerance of arginines in the N-IDR and the effects of introducing positive charges in this region might hint towards a regulatory mechanism involving different conformational states perhaps in a manner dependent of specific intracellular signals. This is in agreement with the rapid and drastic environmental changes that T. brucei encounters during metacyclogenesis. Therefore, studies focused on elucidating the specific roles of the IDRs of RBP6 as intracellular sensors and effectors mediating posttranscriptional regulation are needed.

Methods

PCR-based codon-mutagenesis and obtention of mutant RBP6 entry libraries

RBP6 codon-mutant libraries were generated by adapting a published PCR-based codon-mutagenesis protocol²². First, we TA-cloned the WT-RBP6 sequence flanked by attB1 and attB2 sites into the pGEM®-T vector (Promega). This insert was obtained using primers R6B1Fw-1 (5’-GGGGACAAGTTTGTACAAAAAAGCAGGCTTCATGTTCTACCCCAACAGCCCGCAGACC-3’) and R6B2Rv-1 (5’-GGGGACCACTTTGTACAAGAAAGCTGGGTTCAACCAGCGGCACCGCGGGAACCCGCATG-3’), and genomic DNA (gDNA) from T. brucei Lister 427. Sanger sequencing of three independent clones showed that the RBP6 sequence from our strain lacks one of the five CAG triplets that code for glutamines in the Q-tract of the RBP6 sequence annotated in the TritrypDB database (https://tritrypdb.org/tritrypdb/app). Therefore, the numbering of residues in RBP6 was adjusted to the sequence in our strain. Three independent preparations of the resultant construct were used to mutagenize the RBP6 ORF as described previously²². Primers pGEMT7fw1 (5’-CTATAGGGCGAATTGGGCCCGACGTC-3’) and pGEMTSP6rv1 (5’-CAACGCGTTGGGAGCTCTCCCATATG-3’) were used to obtain linear templates for the first mutagenic PCR and as end primers in mutagenic and joining reactions. Mutagenic primers tiled the RBP6 coding-sequence, except for start and stop codons, and contained NNN random triplets. These were designed with the create_primers.py script (https://github.com/jbloomlab/CodonTilingPrimers) using default settings to achieve similar melting temperatures²³ and were ordered in a 96-well plate format (Integrated DNA Technologies). We performed two rounds of mutagenesis, each with 7 and 20 cycles for mutagenic and joining PCRs, respectively. The PCR conditions were identical to the ones used in the published procedure²² except for the annealing temperature, which was increased to 55 °C, and the elongation time, which was reduced to 15 s. To generate entry mutant libraries, agarose-purified amplicons from the second round of joining PCRs were cloned into pDONR-Express⁴⁶: BP reactions containing 150 ng of mutagenized amplicons, 150 ng of pDONR-Express, 2 µl of Gateway™ BP Clonase™ II Enzyme mix (Invitrogen), and TE to 10 µl, and incubated at room temperature (25 °C) for 20 h⁴⁷. Reactions were cleaned and concentrated with one volume of AMPure XP reagent (Beckman Coulter). Then, 20 µl of MegaX DH10B™ T1R Electrocomp™ Cells (Invitrogen) were transformed with 1 µl of cleaned DNA containing half of the products from the BP reaction. Transformed cells were plated on LB-agarose containing 100 μg/ml spectinomycin and grown overnight at 37 °C. To estimate the total number of transformants (Supplementary Data 1), dilutions of an aliquot of recovered cells were plated. Colonies were scraped and plasmids extracted through alkaline lysis following standard methods. To investigate the composition of the entry libraries, we randomly picked 47 colonies for Sanger sequencing and analyzed them with the analyze_library.py script (https://github.com/jbloomlab/SangerMutantLibraryAnalysis)⁴⁸. For the WT control, one WT-RBP6 entry library was prepared following the same procedure described for mutant libraries except that a WT amplicon obtained with primers pGEMT7fw1 and pGEMTSP6rv1 was used instead.

Generation of mutant plasmid libraries and single-point mutant constructs

To generate mutant plasmid libraries for inducible expression of RBP6 variants in T. brucei, we transferred mutant-RBP6 entry libraries into a destination vector). First, to diminish variability of expression due to recombination of constructs into different rRNA spacers, we modified the destination vector pLEW100v5(BSD)GW⁴⁹ to target a unique locus for integration in chromosome 10. According to previous transcriptomic reports, this locus is transcriptionally silent in procyclics, metacyclics and bloodstream forms^11,50. Thus, we devised two PCRs with either, primers 10sn2F1 (5’-TCACTTGAGCTCTCTCTTTAACGGTAACTCCCGAAGGAG-3’) and 10FusR1 (5’-GGGAGGGACGCTAACCCTCACTTGCAGCGGCCGCGTACGCTGGAGGAAAATGCTGCGGTGAAAG-3’); or 10FusF1 (5’-CTTTCACCGCAGCATTTTCCTCCAGCGTACGCGGCCGCTGCAAGTGAGGGTTAGCGTCCCTCCC-3’) and 10sn2R1 (5’-AGAGAGCTTAAGTGAGTAACACATACGCGGGCTTGTGGC-3’). The reactions were carried out in 1x Platinum SuperFi II PCR Master Mix (Invitrogen) using T. brucei Lister 427 gDNA. Amplicons were gel-purified and eluted with the same PCR volume of water. To join both amplicons, a mixture containing 1 µl of each PCR product in 50 µl of 1x Master Mix was heated to 98 °C for 30 s, then cycled five times at 98 °C for 10 s, 55 °C for 10 s and 72 °C for 1 min. Next, primers 10sn2F1 and 10sn2R1 were added for a final concentration of 0.5 µM each and the same cycling conditions were resumed for 30 additional cycles followed by 5 min at 72 °C. The final product was cloned between the SacI and AflII sites of pLEW100v5(BSD)GW⁴⁹ and amplified in One Shot™ ccdB Survival™ 2 T1^R Competent Cells (Invitrogen). We named this construct pLEW100v5(BSD)GW-ULch10. LR reactions were performed with 150 ng of each mutant-RBP6 or WT-RBP6 entry library replicate, 150 ng of pLEW100v5(BSD)GW-ULch10, 2 µl of Gateway™ LR Clonase™ II Enzyme mix (Invitrogen), and TE to 10 µl. Incubations and inactivation by proteinase K was performed as described⁴⁷. Transformation of electrocompetent cells, estimation of total UFC, and plasmid extraction were performed as described above for generation of entry libraries except that ampicillin at a final concentration of 100 µg/ml was used instead of spectinomycin. To obtain constructs for inducible expression of RBP6 with single or multiple pre-designed mutations, an individual clone isolated from the WT-RBP6 library was used as template for site-directed mutagenesis. Additionally, primers listed in Supplementary Data 5 and the Quik-Change Lightning Site-Directed Mutagenesis Kit (Agilent) were used according to the manufacturer’s instructions.

T. brucei cell culture and transfection of individual constructs or plasmid libraries

A total of 1 × 10⁸ mid-log phase T. brucei Lister 427 29-13 procyclic cells⁵¹, engineered for inducible expression of gene products⁵², were transfected with 20 µg of Not I-linearized constructs or mutant plasmid libraries according to conditions described⁹. For generating T. brucei mutant libraries or a WT control, 10 transfections per replicate (for a total of 1 × 10⁹ cells) were carried out. Cultures were selected for 8–10 days in the presence of Blasticidin at 10 µg/ml, which was added 8 h post-transfection. Cell densities were maintained below late log-phase during selection. Stable transformants were cryopreserved in liquid nitrogen as 0.5 ml aliquots of medium containing 15% fetal bovine serum, 10% glycerol and ~1 × 10⁸ cells. To estimate number of transformants in mutant- and WT-T.brucei libraries (Supplementary Data 1), right after transfection 9.6 ml of medium at 5 × 10⁴, 2.5 × 10⁴ or 1.25 × 10⁴ transfected cells/ml were transferred to 96-well plates (0.1 ml/well) and grown for 10–13 days at 28 °C, 5% CO₂ in the presence of Blasticidin at 10 µg/ml. Wells containing blasticidin-resistant clones were counted and used to estimate library size (Supplementary Data 1) and transfection efficiency; which in average was 3.7 × 10⁻⁴ (±4.3 × 10⁻⁵, SD), aligning with reported efficiencies⁵³.

Processing of cultures expressing WT- or mutant-RBP6 variants

Expression of WT- or mutant-RBP6 variants was induced with 10 μg/ml of doxycycline (unless stated otherwise) in cultures containing ≥2 × 10⁷ total procyclic cells in mid-log phase. Cultures were diluted daily with medium containing doxycycline to maintain cell densities between 2 × 10⁶ and 1 × 10⁷ cells/ml. For DMS, mutant- or WT-T. brucei libraires were independently thawed and grown for analysis of metacyclics purified at either day 3 or 6 post-induction. Cultures were scaled up without discarding any fraction of induced cells until processing of the libraries. Metacyclics cells were purified following a previously published procedure⁴. Cells were centrifuged at 1150 × g for 10 min and resuspended in BBSG (50 mM bicine, 50 mM NaCl, 5 mM KCl, 70 mM glucose) at ~1 × 10⁹ cells/ml. Around 5 × 10⁸ cells (0.5 ml of suspension) were passed through columns prepared in Disposable Borosilicate Glass Pasteur Pipettes (Fisherbrand, Cat. No. 13-678-20A) containing ~2 ml of 0.1 mm diameter zirconia/silica beads (BioSpec Products, Cat. No. 11079101z). To prevent leaking of beads, a plug of glass wool was placed at the narrowing of the pipette tip. Before purification, columns were washed with at least 2 ml of BBSG. Immediately after the entire cell suspension entered the beads by gravity, BBSG was gently added to the column until ~1 ml of metacyclic fraction was collected. The total amount of passed cells varied among biological replicates (Supplementary Data 1) with an average of 7 × 10⁹ cells (±8 × 10⁸, SD) at day 6 and 1.2 × 10¹⁰ cells (±3 × 10⁹, SD) at day 3 of purification. After purification, metacyclic fractions were pooled together and used for gDNA extraction. On average, the yields were 3.2 × 10⁷ or 7.1 × 10⁶ metacyclic cells at day 6 or 3 post-induction, respectively. To analyze RBP6 variants in uninduced libraries or input cells, right before induction or purification of metacyclics, respectively, 1 × 10⁸ cells were used for gDNA extraction.

High-throughput sequencing of barcoded-subamplicons

To retrieve the information from the screening of mutant- and WT-RBP6 libraries with high accuracy, we employed a previously described barcoded-subamplicon sequencing strategy (https://jbloomlab.github.io/dms_tools2/bcsubamp.html^54,55 and Supplementary Fig. 10). First, we generated linear templates by PCR with oligos SRS-276F (5’-AGTAAAATTCACAAGCTATCAAACAAGTTTG-3’) and SRS-277R (5’-GCAGGATCATCGAACCACTTTGTA-3’). We used the PCR conditions described²² with a few modifications: The annealing temperature was 56°C; elongation time was 15 s. Amplicons for mutant plasmids and WT plasmid libraries were obtained using 15 ng of plasmid as template. For libraries analyzed at day 6 post-induction, we used 75 ng of gDNA from uninduced libraries, input cells and metacyclic fractions and their corresponding WT-controls. For analysis of libraries at day 3 post-induction two PCR reactions, each containing 6 ng of gDNA, were carried out with 30 cycles of amplification, rather than 25. PCR products were purified from agarose gels and used as templates in “PCR1” reactions to generate three subamplicons of ~240 bp that tile the entire coding-sequence of RBP6. PCR1 and PCR2 reactions were performed as described⁵⁴, except that the amount of template for PCR2 corresponded to 1 × 10⁶ double-stranded DNA molecules per subamplicon per sample. Primer sequences for PCR1 and PCR2 are provided in Supplementary Data 5. PCR2 subamplicons were quantified, pooled in equimolar amounts, size selected by agarose-gel extraction and cleaned with one volume of AMPure XP reagent (Beckman Coulter). Subamplicons were sequenced with an Illumina HiSeq 2500 using 2 × 250-bp paired-end reads in rapid run mode with 10% PhiX Control v3.

Analysis of high-throughput sequencing data

To analyze reads from high-throughput sequencing data we used dms_tools2 (https://jbloomlab.github.io/dms_tools2/), version 2.6.12²⁴. Briefly, occurrences of each mutation per sample were counted with dms2_batch_bcsubamp (https://jbloomlab.github.io/dms_tools2/bcsubamp.html). Mutation frequencies (Supplementary Data 2) and amino acid preferences (Supplementary Data 4) were calculated from these counts using a previously described approach²⁴. Differential selection scores (Supplementary Data 3) were calculated as described⁵⁶ (see also https://jbloomlab.github.io/dms_tools2/diffsel.html), contrasting induced input cells or purified metacyclic fractions to uninduced T. brucei libraries. We considered as significant, scores of differential selection with values above 0.88 or below −0.84 and p values < 0.05. These thresholds represent the mean of scores obtained with mutant plasmid libraries plus or minus six times its standard deviation and a ~0.56- or ~1.84-fold decrease or enrichment of variants relative to uninduced libraries, respectively.

Miscellaneous software

Heatmaps were generated in R studio using differential selection scores. Logo plots were generated with the python package dmslogo, version 0.7.0. Sequence alignments were performed using the MUSCLE tool (https://www.ebi.ac.uk/jdispatcher/msa/muscle)⁵⁷ with sequences obtained from the TriTrypDB database (https://tritrypdb.org/tritrypdb/app). Master alignment of RRMs was obtained from the supplementary information (S5 Dataset) of reference³⁴. Data for generating the plot showing log2(FC) of enrichment scores for PABP1 were extracted from the supplementary information (Supplementary_Table_2) of reference³³. Frequencies in logo plots correspond to the number of times an amino acid was present in a determined position in an alignment divided by the number of total aligned sequences. Plots depicting log2(FC) mutation frequency were generated in R studio with the ggplot2 package. Scores of predicted disorder were obtained with the computational tool flDPnn (http://biomine.cs.vcu.edu/servers/flDPnn/)²⁶. Percentages of residue content in RBP6 by disorder propensity were obtained with the idpr package⁵⁸. Images of tertiary structure of RBP6 were generated in the PyMOL Molecular Graphics System, version 2.6.0a0, using a pdb file previously generated using predictions from AlphaFold2 with optimizations for trypanosomes (downloaded from http://wheelerlab.net/alphafold/TbruceiTREU927/viewprot.php?id=Tb927.3.2930)^59,60. To generate Supplementary Fig. 9b, we downloaded a pdb file with the cocrystal structure of human poly(A)-binding protein (PABP) from https://www.rcsb.org/ with PDB ID: 1CVJ⁶¹. Densitometry analyses of protein expression were performed using ImageJ software version 1.54d, using cross-reactive bands for normalization.

Miscellaneous methods

Counting of cells was performed manually using a hemocytometer. Purification of gDNA was achieved with the DNeasy® Blood & Tissue kit (QIAGEN). DNA digestion was performed with restriction enzymes from New England Biolabs. For purification of DNA from agarose gels we used the QIAquick® Gel Extraction Kit (QIAGEN). DNA was quantified using Quant-iT™ PicoGreen® dsDNA Reagent (Invitrogen). Recombinant MBP-RBP6 was expressed in BL21-Gold(DE3) E. coli from pMAL-c5X and purified on amylose resin (New England Biolabs). All the above was carried out following manufacturer’s recommendations. In vitro T7 RNA polymerase transcription of RNA substrate (with or without ³²P-labeling), SELEX, UV crosslinking, RNase A treatment of protein-RNA mixtures and SDS-PAGE were performed using standard protocols. The RNA used for the UV crosslinking assay was 5’- gcacgacgcucuuccgaucCCGUUUUAUAUAUUAGUCCUUUCUCCGGAUAACCUGUCGCGCCUUAUGAGcuuaugagcgucuacauggauccuca-3’ (the 50-nt region randomized in the starting SELEX library is in upper case). For Western blotting, cells were centrifuged at 1150 × g for 6 min and resuspended in Laemmli Lysis-buffer at 10⁸ cells/ml. Lysates were heated at 60 °C for 10 min and 10⁶ cell-equivalents per lane were separated in 12% SDS-PAGE gels. After transfer of proteins, PVDF membranes were blocked with 5% milk for 1 h at room temperature. Membranes were rinsed three times and incubated for 1 h at room temperature with rabbit serum against RBP6³, BARP, or mVSG 397⁴ at a 1:1000 dilution. Membranes were washed five times for 3 min and incubated with either horseradish peroxidase-conjugated secondary goat anti-rabbit antibody (Invitrogen, Cat# 31460) or horseradish peroxidase-conjugated secondary goat anti-mouse antibody (Invitrogen, Cat# 31430) at a 1:5000 dilution for 1 h at room temperature. After four 10 min washes, membranes were developed by electrochemiluminescence. Milk and antibody dilutions, as well as washes were performed using TBST (137 mM NaCl, 2.7 mM KCl, 19 mM Tris, 0.1% Tween-20, pH 7.4).

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

S.R.-S., N.G.K., and C.T. wrote the manuscript. S.R.-S. and C.T. conceptualized and designed the study. S.R.-S. performed the experiments and analyzed the data. N.G.K. performed the experiment shown in Fig. 5f. Funding acquisition: C.T.

Peer review

Peer review information

Nature Communications thanks Kenneth Stuart who co-reviewed with Suzanne McDermott; and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

Raw reads from this study have been submitted to the NCBI Sequence Read Archive-SRA at http://www.ncbi.nlm.nih.gov/Traces/sra/sra.cgi with accession number PRJNA1117038. Mutant-level data were uploaded to MaveDB⁶² under experiment set number urn:mavedb:00001212. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-025-56553-y.