We initially tagged trypanosomes with different colours in order to study genetic exchange in tsetse flies. For this purpose plasmids encoding different fluorescent proteins (GFP and DsRED) were integrated into defined loci on chromosomes 6 and 10 (see Materials and methods). When flies were co-infected with these tagged procyclic forms, we observed that the growth rates of individual clones differed and one clone/colour overgrew the other. To identify pairs with similar growth rates in culture, pairs of red and green procyclic forms were mixed and their relative numbers monitored by fluorescence microscopy. Unexpectedly, we observed that approximately 1% of the cells were positive for both fluorescent proteins and appeared yellow in merged images. After repeating this experiment several times we discovered that the transfer of cells to fresh medium with fresh FBS led to robust and reproducible production of yellow cells. While most yellow cells observed after 24 hours were single cells with no interacting partner, some were in intimate contact, forming characteristic pairs ( Figures 1A & B). This is in contrast to dividing cells, which have two clearly separate flagella and joined posterior ends ²⁰ . Yellow cells apparently connected by their anterior ends could also be observed ( Figure 1C).

Procyclic form trypanosomes are covered by several million copies of glycophosphatidylinositol (GPI)-anchored surface glycoproteins known as EP and GPEET procyclins ³⁹ . To see if these proteins were required for the formation of yellow cells we used a procyclin null mutant (Δproc; ⁴⁰ ) tagged with either GFP or DsRED. Mixed cultures of the mutant gave rise to double-positive cells at even higher frequencies than wild-type trypanosomes, reaching 5–7% of the population after 24 hours ( Figure S1). No morphological differences could be observed between the procyclin knockout and wild-type parasites or the pairs that they formed. Because of the increased frequency of double-positive cells we used this mutant for some of the experiments reported here. Addition of the chelating agents EDTA or EGTA abolished the formation of yellow cells, indicating a requirement for extracellular calcium ( Figure S1).

Since yellow cells are used as a read-out for mating ^{36, 41} , we initially thought that trypanosomes had undergone some form of genetic exchange, although this has never been documented for procyclic forms. The parasites were tagged with different selectable markers (rendering them resistant to blasticidin and G418, respectively), either at the same locus on the diploid copies of a chromosome or at different chromosomal loci. However, despite repeated attempts, we were never able to isolate genetic hybrids that were resistant to both drugs.

To better investigate how double-positive cells arose we performed time-lapse microscopy using wild-type trypanosomes expressing cytoplasmic DsRED or GFP, together with a GFP-tagged version of Histone 2B to visualize the nucleus. In order to track cells for extended periods we cultured them for 4 hours in medium without FBS, which causes them to adhere to the surface of the culture flask (see Materials and methods). These experiments revealed that the trypanosomes exchanged cytoplasmic proteins in both directions. The interacting cells became double positive within 10 minutes (the interval between images) and they could separate again ( Figure 2, Video 1 and Video 2). Some cells stayed connected for more than 7 hours ( Video 1), while others separated within 20 minutes ( Video 2). Neither cell-cell fusion nor nuclear exchange was observed, arguing against this phenomenon being mating.

If proteins, but not DNA are exchanged between two cells, it would be expected that they lose one fluorescent protein once they separate. To test this we enriched for double-positive cells and monitored how long they retained both colours. After one round of FACS 26% of trypanosomes in the culture were double-positive ( Figure 3). Cells expressing GFP only were sorted as a control. Cell numbers and the percentage of yellow cells were determined every day. While the cell number more then doubled within the first 24 hours, the percentage of yellow cells decreased only slightly, indicating that the yellow cells still proliferated and were not simply overgrown by single positive cells ( Figure 3). After 3 days, however, the percentage of yellow trypanosomes returned to background level, consistent with turnover of the transferred proteins.

The experiments described above show that procyclic form trypanosomes are capable of exchanging soluble proteins that are mainly present in the cytoplasm. To test whether surface proteins could also be exchanged we first used cells tagged with a GFP-procyclin fusion protein. In trypanosomes, newly synthesised GPI-anchored proteins gain access to the cell surface via the flagellar pocket. This is an invagination of the plasma membrane where the flagellum emerges from the cell body, and is the only known site of endo- and exocytosis. On exiting the pocket GPI-anchored proteins are distributed along the flagellum to the cell surface ⁴² . In cells expressing GFP-procyclin the flagellar pocket is seen as an intensely fluorescent signal ( Figure 4A). When trypanosomes expressing cytoplasmic DsRED and GFP-procyclin were mixed together, we observed that DsRED was equally distributed between two interacting cells, but GFP-procyclin was only transferred to the flagellum of the recipient and not the rest of the cell surface ( Figure 4A). These results indicate that proteins on the outer leaflet of the plasma membrane can be transferred between cells, but transfer appears to be restricted to the flagellum.

Calflagin-44 is an acylated protein that is anchored to the flagellar membrane ⁴³ . To obtain more information on the involvement of the flagellum we generated stable transformants expressing calflagin-GFP and calflagin-mCherry fusion proteins. These were correctly localised to the flagellum, but the protein was sometimes also detected in the cell body, possibly due to its overexpression. When trypanosomes expressing calflagin-GFP and calflagin–mCherry were co-cultured, interacting pairs had both proteins in their flagella, but calflagins in the cell bodies were not exchanged ( Figure 4B). Based on these results we hypothesised that surface proteins that were excluded from the flagellum would not be exchanged. To test this we mixed cells expressing DsRED with trypanosomes expressing a GFP-tagged version of the nucleoside transporter NT10, which is restricted to the cell body ⁴⁴ . Interacting pairs from this mixture had DsRED distributed equally between the two cells while NT10-GFP remained only on one cell ( Figure 4C and Figure S2). Taken together, these data suggest that the flagellum is the primary site of interaction and protein exchange. In this context it is important to note that soluble GFP and DsRED are able to cross the diffusion barrier of the flagellar transition zone and are therefore present in the flagellar matrix as well as in the cytoplasm.

RNA-binding proteins play a major role in regulating gene expression in trypanosomes. They can alter the stability or translational efficiency of individual mRNAs or mRNA cohorts ⁴⁵ , or they can silence them by RNA interference ⁴⁶ . To test if RNA-binding proteins could be exchanged between cells we used a tagged form of Argonaute (Ago) with GFP fused to its N-terminus (GFP-Ago). Cytoplasmic proteins >75kDa are normally excluded from flagella and cilia unless they contain targeting signals ^{47– 49} . One rationale for choosing the GFP-Ago fusion was that the protein is 130 kDa and should therefore require active transport; the second rationale was that it is a catalytic component of the RNA-induced silencing complex. When cells expressed GFP-Ago, the protein was detected in the flagellum as well as in the cytoplasm ( Figure 5A). Moreover, when these cells were mixed with cells expressing DsRED, bidirectional exchange of both proteins was observed ( Figure 5B). In summary, trypanosomes have the capacity to transport Ago into the flagellum and to transfer it to another cell. This transfer could potentially reprogram gene expression in the recipient cell by transferring small interfering RNAs.

To investigate protein exchange with higher temporal resolution, we performed time-lapse imaging with a lattice light sheet microscope ⁵⁰ using an image acquisition interval of one minute. Again we mixed trypanosomes expressing cytoplasmic GFP or calflagin-mCherry. The exchange of calflagin-mCherry occurred in less than a minute ( Figure 6), while it took approximately 2–3 minutes until GFP was equally distributed between two cells ( Figure 6 and Video 3). These experiments further demonstrated that interacting pairs were still moving, indicating that the motility function of the flagellum was not impaired ( Video 3). Furthermore, complete divisions of interacting cells could be observed ( Video 4). While some interacting trypanosomes stayed together for an extended period of time, other interactions were very transient. The shortest interaction we observed was for 1–2 minutes ( Video 5).

The fact that trypanosomes exchange proteins anchored to the outer surface of the flagellar membrane (EP procyclin) or the inner surface (calflagin), suggested either fusion of these membranes or a short-range exchange of vesicles had taken place. To distinguish between these possibilities we performed electron microscopy (EM). Since only a few percent of a mixed population is double-positive, and only some of the cells interact at a given time-point, we again enriched for yellow cells by FACS. Using scanning EM we observed the same characteristic pairs of trypanosomes seen by fluorescence microscopy ( Figure 7A). The flagella of these pairs appeared thicker than the flagella of single cells, which could be compatible with fusion along their entire length. From these images alone, however, we could not conclude unambiguously that the membranes were fused; it remained possible that they were merely in very close contact. To better characterise the interaction between flagellar membranes we performed transmission EM ( Figure 7B and Figure S3). These images demonstrated unequivocally that pairs of cells shared a flagellum with two axonemes and two paraflagellar rods bounded by a single membrane ( Figure 7B and Figure S3A). This differs from what happens during cell division, in which the trypanosomes produce a separate new flagellum posterior to the old one. Thus, the most plausible explanation is that the two flagella have fused. We also documented examples of triple and quadruple fusions ( Figures S3B & S3C), albeit at a much lower frequency.

Structured Illumination Microscopy (SIM) provides a resolution doubling when compared to diffraction limited imaging and complements scanning EM data by delivering information on protein exchange. We therefore applied this method to co-cultures of trypanosomes expressing cytoplasmic DsRED or calfalgin-GFP. This confirmed that large parts of the flagella fuse into a single structure in double-positive cells ( Figure 7C). Once again, fusion of multiple cells could be observed ( Figure S4A). Pairs with a second flagellum forming on one cell, double-positive cells undergoing cytokinesis and fused cells giving rise to daughter cells were also seen ( Figures S4B & C). Taken together, these data suggest that flagellar fusion is a guided process that is part of normal cell functions.

Unless otherwise specified, chemicals were obtained from Sigma-Aldrich, Switzerland. Oligonucleotide primers were synthesised by Microsynth AG (Balgach, Switzerland). Enzymes were purchased from New England Biolabs.

Procyclic forms of T. b. brucei AnTat 1.1 ⁶⁴ and genetically manipulated derivatives were used in this study. The deletion mutant Δproc ⁴⁰ was described previously. Trypanosomes were cultured at 27°C in SDM-79 (Amimed, Cat. No. 9-04V01M) supplemented with 10% heat inactivated foetal bovine serum ⁶⁵ .

Conditions for double-positive cells: 5 × 10 ⁶ cells expressing DsRED were mixed with 5 × 10 ⁶ cells expressing GFP and centrifuged at 1300g for 6 minutes in a 15ml Falcon tube. The cells were resuspended in 10ml phosphate-buffered saline and centrifuged again. The washed trypanosomes were resuspended in 2.5ml SDM-79 with 10% fresh FBS and cultured overnight at 27°C in one well of a six-well plate. The cells tend to adhere weakly to the bottom of the well, so before analysis they were gently flushed loose with a 1ml pipette.

Stable transformation of procyclic form trypanosomes was performed as described ⁶⁶ . To clone the plasmids pG-EGFP-Blast-ΔLII and pG-DsRED-Blast-ΔLII, pG-EGFP-ΔLII ⁶⁷ and pG-DsRED-ΔLII ⁶⁸ were digested with the restriction enzymes NheI and ClaI to cut out the neomycin resistance gene. The blasticidin resistance gene was amplified by PCR using the primers NheIBlast (GC TAGCTAGCATGGCCAAGCCT) and ClaBlast (CC ATCGATACTCACAGCGACTA) and pC-EP2-ΔLII-Blast as template ⁴⁰ the product was digested with NheI and ClaI and ligated into pG-EGFP-ΔLII and pG-DsRED-ΔLII.

The mCherry coding region was amplified by PCR using the plasmid CWP1:mCherry (a gift from Adrian Hehl, Zürich University) as a template and the primers mCherryfor (TT ACCGGTCATGGTGAGCAAGGGCG) and mCherryrev (TT GGATCCCGGGCTTGTACAGCTCGTCCATG). The PCR product was digested with BamHI and AgeI and ligated into BamHI/AgeI digested pG-EGFP-ΔLII, replacing EGFP by mCherry. To change the selectable marker of pG-mCherry-ΔLII from neomycin-resistance to phleomycin-resistance the plasmid G-BIL4-phleo was digested with XbaI and NotI to excise the phleomycin resistance gene. The phleomycin resistance gene was then ligated into the XbaI/NotI-digested pG-mCherry-ΔLII.

To tag Calflagin44 at its C-terminus with EGFP or mCherry, the coding region of Cal44 was amplified by PCR using the primers Cal44for (AT AAGCTTATGGGTTGCTCTGCATCG) and Cal44rev (AT GTCGACGATTACCTTCATTTGCTCC) and genomic DNA as the template. The PCR product was cloned into the pGEM-T Easy Vector System (Promega), subsequently excised with HindIII and SalI, then ligated into HindIII/SalI digested pG-EGFP-Blast-ΔLII and pG-mCherry-Phleo-ΔLII. To tag Ago1 at its N-terminus the Ago1 coding region was amplified by PCR using the primers Ago1SmaFor ( CCCGGGATGTCTGACTGGGAAC) and Ago1SmaRev ( CCCGGGTTATAGATAATGCATTGTTG) and the product was cloned into the pGEM-T Easy Vector System (Promega). The plasmids pGEM-Ago1 and pG-EGFP-ΔLIIγ ⁶⁹ were digested with the restriction enzyme SmaI and the coding region of Ago1 was ligated into pG-EGFP-ΔLIIγ, correct integration was confirmed by sequencing.

The constructs pG-H2B-GFP-ΔLII and pG-NT10-GFP-ΔLII were described previously ^{44, 69} .

Cells were washed in phosphate buffered saline (PBS), spread on a microscope slide and mounted with Moviol. Images were taken with a Leica DFC360FX monochrome CCD (charge-coupled-device) camera mounted on a Leica DM5500 B microscope with a 100x oil immersion objective and analysed using LAS AF 1.0 software (Leica).

Structured illumination microscopy (SIM) was performed as described in 70. Cells were fixed for 20 minutes at room temperature with 4% paraformaldehyde and 0.5% glutaraldehyde. For each axial plane of a 3D stack, raw SIM images were acquired at five phase steps spaced by 2π/5 of the illumination pattern period, and this process was repeated with the excitation pattern rotated by ±120° with respect to the first orientation. The axial stepping size was set to 160 nm. The raw data was reconstructed into the 3D super-resolution image based on the algorithm described in 71. In our study, when the result was Fourier transformed back to real space, we applied a gamma apodization function A( k)=1–( k/ k _max) ^γ , with γ = 0.4, rather than the traditional triangle apodization A( k)=1– k/ k _max ⁷¹ , where k _max is the maximum support of the expanded optical transfer function (OTF). Therefore, the higher spatial frequencies were not suppressed more than necessary. Furthermore, we strictly follow the azimuthally dependent maximum support k _max(θ) to define the endpoint of the apodization function. All of these provided a better suppression of the ringing artifacts associated with Fourier transformation. After reconstruction, SIM images achieve 110 nm lateral, 350 nm axial resolution.

Trypanosomes (Δproc) expressing either DsRED or GFP were washed in PBS and co-cultured in SDM-79 with 10% fresh FBS for 8 hours at a density of 10 ⁷ cells/ml. Then the cells were transferred into PBS containing 2% FBS at a density of 1.5×10 ⁷ cells/ml for sorting. Sorting was performed with a BD FACS ARIA III (BD Biosciences) equipped with a 130μm nozzle running with 10 psi pressure, flow liquid was PBS. The software used was BD FACS DIVA 6.0 (BD Biosciences). To detect GFP a 488nm blue laser with a 530/30nm bandpass filter was used, to detect DsRED a 561 yellow-green laser with a 610/20 bandpass filter was used. Double-positive cells were sorted into SDM-79 10% FBS.

Flow cytometry was performed with living cells in PBS, 10 ⁴ cells were analysed using a FACSCalibur (BD Biosciences) and analysed with CellQuest Pro (version 5.1.1).

Transmission electron microscopy. Trypanosomes were washed briefly in PBS, then fixed in 2.5% glutaraldehyde in 100 mM sodium cacodylate buffer (pH7.3) for 2 hours at room temperature, followed by post-fixation in 2% OsO4 in cacodylate buffer, pH 7.3. After three washes in distilled water, samples were pre-stained in saturated uranyl acetate solution in distilled water for 30 min at room temperature, extensively washed in distilled water, and dehydrated by stepwise incubation in ethanol (30%-50%-70%-90%-100%). Parasites were then embedded in EPON812 resin as described previously ⁷² . Following polymerisation of the resin at 60°C for 24 h, sections of 80 mm thickness were cut on a ultramicrotome (Reichert & Jung, Vienna, Austria), placed onto 300 mesh formvar-carbon-coated nickel grids (Plano, Wetzlar, Germany), and sections were stained with uranyl acetate and lead citrate ⁴⁰ . Specimens were viewed on a Phillips 400 TEM operating at 60 kV.

Scanning electron microscopy. Trypanosomes were washed briefly in PBS, then fixed in 2.5% glutaraldehyde in 100 mM sodium cacodylate buffer, pH 7.3, for 2 hours at room temperature, followed by post-fixation in 2% OsO4 in cacodylate buffer, pH 7.3. They were then extensively washed in water, dehydrated by stepwise incubation in ethanol (30%-50%-70%-90%-100%), and two short washes in 50µl hexamethyl-disilazane. Trypanosomes were then taken up in a small volume of hexamethyl-disilazane, and were allowed to settle down on glass coverslips and were air-dried under a fume hood. They were then sputter-coated with gold ⁷³ and inspected on a JEOL 840 scanning electron microscope operating at 25 kV.

To immobilise the trypanosomes approximately 3×10 ⁶ cells of each clone were mixed and centrifuged for 6 minutes at 1300g. The cell pellet was washed with 10 ml PBS and centrifuged again using the same parameters as above. The cells were resuspended in 1 ml SDM-79 without FBS and incubated for 4 hours at 27°C in a glass bottom culture dish (Willco Wells, the Netherlands). During this time the trypanosomes adhered to the bottom of the dish and were immobilised. After 4h 100μl FBS was added to the cells and the time-lapse imaging was started immediately using a Nikon TE2000E-PFS microscope with a 60x objective. DIC and fluorescent images were taken every 10 minutes for 24 hours.

The 4D lattice light sheet measurements were performed using a microscope described previously ⁵⁰ . This system illuminates the sample with a massively parallel array of coherently interfering beams comprising a non-diffracting 2D optical lattice. This creates a coherent, spatially structured light sheet that is then dithered to create uniform excitation in a ~600 nm thick plane across the entire field of view. In order to capture the statistically rare fusion events in trypanosome samples, 3D images were acquired from 15 fields of view every 60 seconds for a total duration of 9h12min. Raw data were deconvolved via a 3D iterative Lucy-Richardson algorithm in Matlab version 2013b (The Mathworks, Natick, MA) utilizing an experimentally measured point spread function using 100 nm fluorescent beads (Fluospheres, ThermoFisher Cat #: F8803). Movies were generated using ImageJ (Version 1.49m).