The endoplasmic reticulum-associated mRNA-binding proteins ERBP1 and ERBP2 interact in bloodstream-form Trypanosoma brucei

DNA manipulation and trypanosomes

Lister 427 strain trypanosomes expressing the tet repressor were used for all experiments, and were cultivated and transfected as described previously (Alibu et al., 2004). All plasmids and oligonucleotides are listed in Table S3. Expression from tetracycline-inducible promoters was induced with 100 ng/ml tetracycline, and all growth studies were performed in the absence of selecting antibiotics. Cultures that were used for RNA and protein analysis had a maximum density of 1.5 × 10⁶/ml.

Western blotting

3–5 × 10⁶ cells were collected per sample, resuspended in 6x Laemmli Buffer and heated at 95 °C for 10 min. The samples were subjected to SDS-PAGE gel electrophoresis using 10% polyacrylamide gels. The gels were then stained with SERVA blue G or blotted on a 0.45 µm nitrocellulose blotting membrane (Neolabs). To verify the protein transfer, the membrane was stained with Ponceau S (SERVA). The membrane was blocked with 5% milk in TBS-Tween and incubated with appropriate concentrations of first and secondary antibodies. Western Lightning Ultra (Perkin Elmer) was used as a chemiluminescence system and signals were detected with the LAS-4000 imager (GE Healthcare) and CCD camera (Fujifilm). Antibodies used were: rabbit anti-Aldolase (1:50000) (Clayton, 1987); mouse anti-myc 9E10 (Santa Cruz, 1:200); rabbit Peroxidase anti-Peroxidase (Sigma, 1:20000); rat anti-ribosomal protein S9 (1:1000); anti-Trypanothione Reductase (rabbit, gift from L. Krauth-Siegel, BZH Heidelberg); mouse anti-V5 (Biorad, 1:2000); anti-SCD6 and anti-DHH1 (from S. Kramer, University of Wurzburg, 1:10000 and 1:15000 respectively) and rabbit anti-BiP (from J. Bangs, University of Buffalo, 1:1000).

Digitonin and stress granule fractionation

For each sample, 3 × 10⁷ cells were collected by centrifugation at 2,000 g for 10 min at 4 °C. The pellet was resuspended in 100 µl 1x PBS and centrifuged at 10,000 g for 5 min at 4 °C. The pellet was then resuspended in 50 µl STE buffer (10 mM Tris–HCl pH 8.0, 150 mM NaCl, 1 mM EDTA) and centrifuged at 10,000 g for 5 min at 4 °C. A 10 µg/µl digitonin stock solution was heated at 98 °C for 5 min and cooled down before use. Seven different digitonin containing solutions, ranging from 0–1.65 µg/µl digitonin, were prepared and each pellet was resuspended properly in 60 µl of one solution. The samples were incubated at 25 °C for 5 min and then centrifuged immediately at 10,000 g and 4 °C for 5 min. The supernatant was transferred to another tube containing 20 µl 4x SDS-PAGE sample buffer. The pellet was washed twice with 1 × PBS by centrifugation (4 °C, 10,000 g, 5 min) and finally resuspended in 80 µl 1 × Laemmli buffer. Samples were analyzed by Western Blotting. Stress granules were purified exactly as described in Fritz et al. (2015), and immunoprecipitations were done as described in Singh et al. (2014).

Immunofluorescence microscopy

Tissue culture glass slides with 8 chambers were treated with 0.1% Poly-Lysine (Sigma, P-8920). 2.5 × 10⁶ formaldehyde-fixed T. brucei were allowed to adhere to poly-lysine-treated chambered slides (Falcon, 354108), permeabilised with 0.2% (w/v) Triton X-100 then incubated with protein-specific antibodies followed by fluorescently-labelled second antibodies in PBS containing 0.5% gelatin. DNA was stained with 100 ng/ml DAPI (D9542, Sigma-Aldrich). Mitochondria were detected by addition of Mitotracker Red CMXRos (50 nM, Thermo Fisher Scientific) to the cells 5 min prior to fixation. Images were examined using the Olympus IX81 microscope, 100x Oil objective with a numerical aperture of 1.45. Digital images were taken with ORCA-R2 digital CCD camera C10600 (Hamamatsu) and using the Xcellence RT software. The bright-field images were taken using differential interference contrast (DIC). Fluorescent images were taken as Z-Stacks with a height of roughly 4 µm and a step width of 0.2 µm. The images were deconvoluted (Wiener Filter, Sub-Volume overlap: 20) and then processed using ImageJ. The background was subtracted and brightness and contrast were adjusted automatically. The most in-focus image of the deconvoluted stack was used.

RNA preparation and Northern blotting

5 × 10⁷ cells were used for the extraction of total mRNA using peqGold Trifast (PeqLab). RNA was separated on agarose-formaldehyde gels and blotted on a nylon membrane (Amersham Hybond-N+, GE Healthcare, RPN203B). RNA was cross-linked on the membrane by UV light (0.2 J/cm²) and stained afterwards with methylene blue (SERVA) before hybridization with ³²P-labelled probes, made using either the Prime-IT RmT Random Primer Labelling Kit (Stratagene) or, for oligonucleotides, [γ³²P]ATP and T4 polynucleotide kinase (New England Biolabs).

Affinity purification and mass spectrometry

To purify TAP-ERBP1 for mass spectrometry, the protein was subjected to two steps of affinity purification (Estévez et al., 2003). Briefly, the cleared lysate was incubated with IgG sepharose beads, washed, and then bound proteins were released using TEV protease. The resulting preparation was then allowed to adhere to a calmodulin affinity column, and proteins were eluted with EGTA. Co-purifying proteins from three independent experiments were analyzed by LC/MS by the ZMBH Mass Spectrometry facility. Cell lines expressing TAP-GFP served as control. Raw data were analyzed using MaxQuant 1.5.8.3, with label-free quantification (LFQ), match between runs (between triplicates), and the iBAQ algorithm enabled. The identified proteins were filtered for known contaminants and reverse hits, as well as hits without unique peptides.

Affinity purification and RNASeq

For identification of associated RNAs (Droll et al., 2013), 1 × 10⁹ cells were resuspended in 50 ml ice-cold PBS, and UV-crosslinked (2 × 0.24 J/cm², Stratagene UV crosslinker) in two P15 Petri dishes on ice. They were then pelleted, snap-frozen and stored in liquid nitrogen before use. The TAP-ERBP1 was allowed to bind to IgG beads, and the unbound lysate was retained as one of the controls. The bound protein was released using TEV protease. Cross-linked proteins were removed by proteinase K digestion (Droll et al., 2013) and RNA was extracted using peqGold Trifast (peqLab) according to manufacturer’s protocol. rRNA was depleted as required, by incubation with complementary oligonucleotides and RNaseH (Minia et al., 2016). The NEBNext Ultra RNA Library Prep Kit for Illumina (New England BioLab) was used for library preparation, prior to sequencing with sequenced at EMBL (HiSeq 2000) to generate 50-base reads. Data were analyzed using an in-house tool (10.5281/zenodo.165132) (Mulindwa et al., 2018).

ERBP1 is conserved in Trypanosoma

To investigate the function of ERBP1, we first analyzed its sequence. ERBP1 is a 55-kDa protein with an isoelectric point of 5.73. It is predicted (by Phyre2) to consist predominantly of alpha-helices. Homologues are found at the equivalent chromosomal position in other Trypanosoma species, and in Endotrypanum, Paratrypanosoma, Blechomonas and Bodo, but the gene appears to have been lost in Leishmania. The alignment (Fig. S1) shows that the proteins share extremely acidic C-termini: for example, in T. brucei, the last 33 amino acid residues include 5 aspartates and 14 glutamates, as opposed to 3 lysines. Outside the Kinetoplastida, the nearest matches were not with yeast Bfr1p, but with proteins of unknown function in Galdieria sulphuraria, an acidiphilic red alga, and organisms from the SAR group that probably have incorporated red algal endosymbionts: the oomycete Phytophthora parasitica, the brown alga Ectocarpus siliculosus and the diatom Phaeodactylum tricornutum. The alignment with all these proteins, as well as Bfr1p, showed little sequence identity between the kinetoplastid sequences and all the others (Fig. S1).

ERBP1 is required for normal growth but is not essential

We first confirmed that ERBP1 is indeed required for normal growth of bloodstream-form trypanosomes. To do this, we created bloodstream forms in which one ERBP1 gene was modified to encode a protein with an N-terminal V5 tag (V5-ERBP1). These were transfected with a tetracycline-inducible construct for production of ERBP1 dsRNA. Depletion of ERBP1 clearly inhibited cell proliferation but the cells nevertheless survived (Fig. 1).

Since some protein always remains after RNAi, we also tested the ability of the cells to survive in the absence of the protein. Bloodstream form cells without ERBP1 (Figs. 2A and 2B) also grew slower than wild-type (Fig. 2C), and this could be compensated by re-expression of the protein (Fig. 2D). Cells containing only a single tetracycline-inducible copy of ERBP1-myc (inducible on a knock-out background, cKO) grew at similar rates to wild-type with or without tetracycline, but the protein was clearly detectable in the absence of induction. The double band from ERBP1 was not always seen but might be caused by acetylation (Moretti et al., 2018) or phosphorylation (Urbaniak, Martin & Ferguson, 2013). Despite these initial results, upon prolonged cultivation, the cells lacking ERBP1 gradually increased their growth rate to wild-type. Survival and recovery of the cells after starvation was also indistinguishable from wild-type (Fig. S2A). In procyclic forms, the endogenous ERBP1 gene could be deleted only if an inducible ERBP1-myc gene was present. Depletion of the ERBP1-myc had no effect on growth or recovery from starvation (Fig. S2B), but some residual ERBP1-myc was doubtless present. It is, therefore, possible that ERBP1 is essential in procyclic forms, but the failure of the deletion could also have been due to a technical problem.

The other high-throughput result that required verification was the effect of ERBP1 when tethered to a reporter mRNA. Full-length ERBP1 had previously been identified as conferring a selective advantage in the screen, suggesting that it could activate expression of the blasticidin resistance marker. It gave no growth advantage in control cells in which the marker mRNA had no boxB sequences, ruling out the possibility that ERBP1 expression by itself results in blasticidin resistance. However, lambdaN-ERBP1 was unable to increase expression of a chloramphenicol acetyltransferase reporter (Fig. S2A). We have no explanation of this discrepancy: for both the screen and the individual test, the lambdaN-myc sequence was placed at the N-terminus of the open reading frame.

ERBP1 is associated with the endoplasmic reticulum and with a second ER protein

To examine the location of ERBP1 in bloodstream-form cells, we examined either N-terminally V5-tagged ERBP1 (V5-ERBP1), or C-terminally myc-tagged ERBP1 (ERBP1-myc) expressed from the endogenous locus (Figs. 3 and 4). In both cases, ERBP1 clearly co-localized with the endoplasmic reticulum and to some extent could be found in close proximity to mitochondria and glycosome. We then conducted controlled-digitonin permeabilization studies to evaluate the release of the protein with the same markers (Fig. 5). ERBP1 was found to coelute partially with the cytosolic marker but fully with the endoplasmic reticulum marker, suggesting that it is loosely associated with the cytosolic face of the ER.

To investigate the interactions of ERBP1, we integrated a sequence encoding the tandem affinity purification (TAP) tag into the genome such that the tag would be at the ERBP1 N-terminus (TAP-ERBP1) while the other copy of ERBP1 was deleted. TAP-tagged GFP served as the control. Three proteins reproducibly co-purified with ERBP1: calmodulin, the ATP-dependent RNA helicase HEL67/Vasa-like (Kramer et al., 2012), and a protein of unknown function encoded by Tb927.9.9550 (Table S1). Calmodulin quite often co-purifies with TAP-tagged proteins (Schimanski, Nguyen & Günzl, 2005). The association with the RNA helicase may be significant; although we have detected it on many other mRNA-related purifications. Tb927.9.9550, however, was a novel protein partner. It is conserved throughout Kinetoplastida, activating in the tethering screen (Erben et al., 2014; Lueong et al., 2016), is in the mRNP proteome (Lueong et al., 2016), but again lacks any recognizable domains; we, therefore, call it ERBP2. It appeared to be essential in the high-throughput RNAi screen (Alsford, Glover & Horn, 2005), has a predicted signal peptide and trans-membrane domain, and a GFP-tagged version, like ERBP1, localized to the ER in procyclic form (Dean, Sunter & Wheeler, 2016). To verify the interaction we used bloodstream-form cells containing YFP-in-situ tagged ERBP1, with or without expression of ERBP2-myc. YFP-ERBP1 was pulled down with anti-myc antiserum only if ERBP2-myc was present, confirming the interaction (Fig. 6A), but the experiment also showed that—at least under the conditions used—only a very small proportion of the ERBP1 was associated with ERBP2. This may, therefore, be a transient interaction. ERBP2 is predicted to be predominantly alpha-helical, like ERBP1. The isoelectric point of ERBP2 is 11, so it might interact with the highly acidic ERBP1 C-terminus.

ERBP1 associates with mRNAs encoding ribosomal proteins, and not with starvation stress granules

Finally, we wished to know whether ERBP1 had any mRNA-binding specificity. We, therefore, sequenced mRNAs that were co-purified with TAP-ERBP1 from bloodstream-form cells, comparing them either with one unbound fraction or with the total mRNA. Unusually, the ERBP1 mRNA itself was not enriched, as would be expected from pull-down of the nascent polypeptide. However, the two bound fractions clearly separated from all the controls (Fig. 6B). Sixty four mRNAs were enriched at least 2-fold in both pull-downs relative to all controls. Strikingly, nearly half of them (29) encode ribosomal proteins. Two glycosomal membrane protein mRNAs, PMP4 and PEX11, were also enriched, but those encoding other PEX proteins—including other glycosomal membrane proteins—were not, so the significance of this is uncertain. The length of the associated mRNAs seemed to be important since both ribosomal and non-ribosomal targets are shorter than the median (not shown).

It should be noted that the enrichment of mRNAs is calculated relative to the enrichment of other mRNAs, and does not actually tell us which proportion of mRNAs with that sequence is associated with the protein in vivo.

In budding yeast, Bfr1 is associated with stress granules. To find out whether ERBP1 is stress-granule associated, we incubated procyclic forms in PBS for 2 h. Although clear granules containing the marker SCD6 were formed (Fig. 7A), V5-ERBP1 remained distributed throughout the cytosol (Fig. 7A) and was not associated with the granule fraction (Fig. 7B). There is, therefore, no evidence that ERBP1 is involved in the formation of stress granules or survival after starvation.