Expansion microscopy facilitates quantitative super-resolution studies of cytoskeletal structures in kinetoplastid parasites

Expansion microscopy (ExM) has become a powerful super-resolution method in cell biology. It is a simple, yet robust approach, which does not require any instrumentation or reagents beyond those present in a standard microscopy facility. In this study, we used kinetoplastid parasites Trypanosoma brucei and Leishmania major, which possess a complex, yet well-defined microtubule-based cytoskeleton, to demonstrate that this method recapitulates faithfully morphology of structures as previously revealed by a combination of sophisticated electron microscopy (EM) approaches. Importantly, we also show that due to the rapidness of image acquisition and three-dimensional reconstruction of cellular volumes ExM is capable of complementing EM approaches by providing more quantitative data. This is demonstrated on examples of less well-appreciated microtubule structures, such as the neck microtubule of T. brucei or the pocket, cytosolic and multivesicular tubule-associated microtubules of L. major. We further demonstrate that ExM enables identifying cell types rare in a population, such as cells in mitosis and cytokinesis. Three-dimensional reconstruction of an entire volume of these cells provided details on the morphology of the mitotic spindle and the cleavage furrow. Finally, we show that established antibody markers of major cytoskeletal structures function well in ExM, which together with the ability to visualize proteins tagged with small epitope tags will facilitate studies of the kinetoplastid cytoskeleton.

PG, 0000-0002-5119-7144; MP, 0000-0002-7146-991X; HV, 0000-0003-0405-5693; MO, 0000-0002-2788-7188; MSC, 0000-0001-9053-8928; VV, 0000-0002-0384-1281 Expansion microscopy (ExM) has become a powerful super-resolution method in cell biology. It is a simple, yet robust approach, which does not require any instrumentation or reagents beyond those present in a standard microscopy facility. In this study, we used kinetoplastid parasites Trypanosoma brucei and Leishmania major, which possess a complex, yet well-defined microtubulebased cytoskeleton, to demonstrate that this method recapitulates faithfully morphology of structures as previously revealed by a combination of sophisticated electron microscopy (EM) approaches. Importantly, we also show that due to the rapidness of image acquisition and three-dimensional reconstruction of cellular volumes ExM is capable of complementing EM approaches by providing more quantitative data. This is demonstrated on examples of less well-appreciated microtubule structures, such as the neck microtubule of T. brucei or the pocket, cytosolic and multivesicular tubule-associated microtubules of L. major. We further demonstrate that ExM enables identifying cell types rare in a population, such as cells in mitosis and cytokinesis. Threedimensional reconstruction of an entire volume of these cells provided details on the morphology of the mitotic spindle and the cleavage furrow. Finally, we show that established antibody markers of major cytoskeletal structures function well in ExM, which together with the ability to visualize proteins tagged with small epitope tags will facilitate studies of the kinetoplastid cytoskeleton. parallel evenly spaced microtubules, which subtend the entire cytoplasmic membrane [2]. Another prominent feature is the presence of the flagellum, the principal motility organelle, which contains the evolutionarily conserved microtubulebased axoneme, and the kinetoplastid-specific paraflagellar rod [2]. The flagellum specifies, via association with its base, the position of single-copy organelles, such as the kinetoplast (i. e. mitochondrial DNA) and the flagellar pocket [3]. Moreover, in T. brucei, the attachment of the flagellum to the cell body via the structure called the flagellum attachment zone (FAZ) defines cell shape, polarity and the starting point of cytokinesis [4]. Replication of these cytoskeletal structures is coordinated and tightly linked to the cell cycle, and occurs with high fidelity [5]. Thus, kinetoplastid cells of a proliferative population show a remarkable uniformity of their cytoskeletal architecture.
Studies of the kinetoplastid cytoskeleton have been heavily reliant on electron microscopy (EM), mainly due to the spacing of various structural components, such as the corset microtubules, axonemal microtubule doublets and domains of the FAZ, being below the resolution limit of conventional light microscopy, which is about 180 nm in the lateral and 500 nm in the axial direction [6]. Indeed, EM provided a considerable understanding of the kinetoplastid cytoskeletal architecture (e.g. [5,7,8]). However, the EM approaches require specialized reagents, instrumentation, and are time demanding, making certain types of studies challenging. These include for example an ultrastructural investigation of cells that are rare in a population; yet these may be key for understanding life cycle transition processes. Another example is studies aiming to determine the localization of specific molecules in the entire cellular volume. Similar limitations also apply to major super-resolution optical microscopy techniques, such as STED and STORM [6].
Over recent years, another super-resolution method, termed expansion microscopy (ExM), was developed [9,10]. It is based on embedding a specimen into a swellable polymer and its subsequent isotropic expansion. These specimens, typically expanded 4-5 times in each direction, are thereafter imaged with a confocal microscope; this produces a corresponding increase in resolution and is comparatively rapid.
ExM has been extensively employed to study mammalian cells and tissues (e.g. [11][12][13]). By contrast, only a limited number of studies of parasitic protists have been conducted so far. Yet investigations of Giardia intestinalis [14] and Plasmodium [15], as well as the recent study of a mitochondrial genome segregation factor in T. brucei [16], demonstrated that due to small sizes of their cells ExM offers clear opportunities for analysing structural features of these important parasites. Moreover, due to a well-defined morphology of their cellular cytoskeleton protists enable assessing the ExM approach at cellular dimensions.
In this work, we set out to optimize and validate ExM for Trypanosoma brucei and Leishmania major cells, which possess a complex, highly defined and relatively well-characterized microtubule-based cytoskeleton, as described above. We show that data obtained by ExM recapitulate the previous EM observations with regard to the presence, positioning, orientation and shape of individual structures. However, caution should be observed when using ExM to determine distances, as individual structures may expand to a different extent. This was previously noticed in other experimental systems (e.g. [17,18]). Importantly, we observed that ExM enables resolving individual cytoskeletal components, such as the microtubules.
This, together with the quantitative aspect of the approach, facilitated a better understanding of less well-appreciated structures, such as the neck microtubule in T. brucei, or the pocket, cytoplasmic and multivesicular tubule-associated microtubules in L. major. We further observed in a subset of L. major cells short microtubules or plate-like structures associated with basal and pro-basal bodies. Moreover, ExM facilitated studies of rare cell types in a population, providing detailed information on the structure of the mitotic spindle and cleavage furrow in L. major.
Finally, we demonstrate that a number of commonly used antibody markers can be used to visualize the respective cytoskeletal structures by ExM as well as localize epitopetagged proteins in the entire cellular volume. We envisage that, due to its robustness, flexibility and rapid image acquisition, ExM will become a valuable approach to study the biology of the kinetoplastid parasites.

Optimizing expansion microscopy for kinetoplastid cells
To optimize and validate the ExM protocol for kinetoplastids, we used procyclic T. brucei brucei cells, which naturally occur in the tsetse fly midgut. They can, however, be readily cultured in a laboratory, and hence their morphology, cellular architecture and changes to these during the cell cycle have been extensively characterized [5,7]. We employed a variation of ExM termed ultrastructure ExM, which is based on fixation of cells with 4% formaldehyde and 4% acrylamide, followed by gelation with 19% sodium acrylate, 10% acrylamide and 0.1% N, N'-methylenebisacrylamide, denaturation with 200 mM sodium dodecyl sulfate at 95°C, and antibody staining [10]. Following a described protocol for handling free-swimming Chlamydomonas cells [10], we adhered procyclic T. brucei cells to a glass coverslip and further processed for ExM. To visualize expanded microtubule-based structures to validate the procedure, we tested several antitubulin antibodies: the polyclonal rabbit anti-alpha tubulin antibody (table 1 for specification), the mouse monoclonal anti-alpha tubulin antibody TAT1 [19], the mouse anti-beta tubulin antibody KMX-1 [20] and the mouse monoclonal anti-acetylated tubulin antibody C3B9 [19].
A three-dimensional volume reconstructed from a series of confocal z-planes of expanded cells revealed that the antitubulin signal of each of the antibodies outlined the entire microtubule corset (figure 1a; electronic supplementary material, figure S1). The corset retained its typical elongated shape, including being wider at the posterior cell end than at the anterior end, as demonstrated previously by light microscopy [25], as well as by EM [5]. To estimate the expansion factor, we turned the z-stacks into two-dimensional images by applying maximum intensity projection and measured the length of the corsets; these were 70-110 µm long. Corsets of formaldehyde-fixed non-expanded cells imaged by confocal microscopy (figure 1b) were 16-24 µm long, yielding the expansion factor of approximately 4.5. This is in agreement with the measured 4.7-fold physical expansion (4.58, 4.66 and 4.71-fold expansion in three experiments) of the gel in water prior to imaging.
Tilting three-dimensional reconstructed corsets revealed that they were significantly flattened on the side juxtaposed royalsocietypublishing.org/journal/rsob Open Biol. 11: 210131 to the surface of the gel (figure 1a,c). We assumed this flattening resulted from the adhesion of live cells to the glass surface at the beginning of the ExM procedure. Indeed, we observed similar flattening of non-expanded cells, which were attached to glass alive (figure 1b). Such a reduction in the cell height may be beneficial, as imaging thick specimens is subject to a significant signal loss and aberrations [26]. However, to obtain a more realistic visualization of the shape of the corset, we adapted the procedure for the preparation of specimens for scanning EM [3]. Live cells were first fixed in solution, settled onto a coverslip and thereafter processed for ExM. This indeed led to only a very limited area of a cell being flattened (figure 1d; electronic supplementary material, figure S1). To mitigate the signal loss we tried to match the refractive index of the gel to the one of the oil immersion by incubating it with a solution of 84% sucrose [27]. This indeed partially reduced the signal loss, but in accordance with previous observations [27] also led to a 10% reduction in the gel size with the final expansion factor of 4.3. Therefore, the approach was not universally employed in this study.
In contrast with non-expanded cells, in which the antitubulin antibodies stained the axonemal microtubules only weakly (figure 1b; electronic supplementary material, figure  S2, see also [19]), there was a strong signal associated with this structure in expanded specimens (figure 1a,c,d; electronic supplementary material, figure S1). Hence, the axoneme could be traced continuously from the point of its exit from the corset, running in an incomplete turn alongside the corset, and extending beyond its anterior end (figure 1a,c,d; electronic supplementary material, figure S1), consistent with previous observations [25]. A gap between the corset and the axoneme was always present. This space is occupied by the cellular and flagellar membranes and also contains the junction complex of the FAZ [28] (see below for antibody staining against the FAZ constituents).
Dividing cells were readily identifiable by the presence of two axonemes. The distal end of the new axoneme was found in the proximity of the old axoneme with a gap separating the two (figure 1c,d; electronic supplementary material, figure S1). It had been shown that this space is occupied by the membrane junction termed the flagella connector [29]. Due to the ability to rapidly screen cells in a population at a low magnification, it was straightforward to identify and image rarer cell types, such as cells in various stages of cytokinesis (figure 1c,d; electronic supplementary material, figure S1B). In these cells, the cleavage furrow was clearly discernible and the non-equivalence of nascent daughter cells apparent, with the new flagellum daughter having a longer free flagellum overhang and a non-tapered anterior end (figure 1c,d; electronic supplementary material, figure S1B) [24]. Similarly to the procyclic cells, staining expanded bloodstream T. brucei cells with the anti-tubulin antibodies facilitated visualization of the shape of the corset and the axoneme (figure 1e). The increased resolution made discernible the area of missing anti-tubulin signal surrounding the distal end of the new axoneme (figure 1f ). This corresponds to the groove, an indentation of the cell body membrane with an associated gap in the subpellicular corset originally observed by serial block-face scanning EM [30]; ExM enabled for the first time visualization of the groove by light microscopy. In addition to the groove, a line devoid of the signal extended alongside each axoneme; this area was likely occupied by the FAZ filament (figure 1f ).
As L. major cells are of significantly smaller dimensions than T. brucei cells, their expansion prior to imaging would be very beneficial. Staining with the anti-tubulin antibodies indeed facilitated a three-dimensional reconstruction of corsets and axonemes of expanded promastigote L. major cells. We observed slender cells with a single flagellum as well as shorter and rounder cells in cytokinesis (figure 1g,h), recapitulating the previously described changes to the cell shape associated with the cell cycle progression [31,32]. The expanded cells were between 30 and 60 µm long, which, compared to non-expanded cells (6.5-13 µm) (electronic supplementary material, figure S3), implicated the expansion royalsocietypublishing.org/journal/rsob Open Biol. 11: 210131 factor of approximately 4.6, similar to the one observed for procyclic T. brucei cells.
Three-dimensional reconstruction of cytoskeletons of cells in cytokinesis revealed that the cleavage furrow partitioned the cytoskeleton along the anterior-posterior axis of the cell body between the two flagella (figure 1h). The last point of attachment of two daughter cells was at their very posterior end (figure 1h), as was previously described by scanning EM [32] and at a lower magnification also by differential interference contrast microscopy [31].

Expansion microscopy resolves individual structural components
Inspecting confocal planes revealed that in many individual corset microtubules could be discerned (figure 2a). This was particularly the case when the microtubules were oriented parallel to the imaging plane and close to the surface of the gel, with the concomitantly high-signal intensity. To estimate the lateral spacing of the microtubules, we performed the forward Fourier transform of images of 10 such regions (for an example see figure 2b). The resulting frequency-domain images showed strongly represented frequencies with periodicities within an interval with boundaries of 270 ± 30 nm (mean ± s.d.) and 330 ± 50 nm, and with the mean value of 300 nm. These frequencies indeed accounted for the majority of observed microtubule-associated signal (figure 2b). The spacing of corset microtubules was previously determined by transmission EM as being 24 nm [7]. Taking into account the microtubule diameter of about 25 nm, the centre to centre distance of two adjacent microtubules in an unexpanded cell is about 50 nm. This implicates an expansion factor of 6.0, significantly higher than the expansion factor of the gel (4.7). The antibodies KMX-1, TAT1 and C3B9 produced staining of a higher density and lower background than the rabbit anti-tubulin antibody. However, using none of these anti-tubulin antibodies, it was possible to reliably determine where individual microtubules terminated in corsets of T. brucei (e.g. figure 2a,b). This was, however, possible in L. major, mainly due to larger distances between corset microtubules in expanded cells ( periodicities within an interval of 310 ± 20 nm to 470 ± 30 nm, n = 11 cytoskeletal regions; see also electronic supplementary material, figure S4). This, for example, enabled visualization of the high concentration of microtubule ends at the posterior end of the cell (figure 2c). royalsocietypublishing.org/journal/rsob Open Biol. 11: 210131 In individual confocal planes axonemal microtubule doublets were readily noticeable (figure 2a). In regions where the axoneme was oriented perpendicular to the imaging plane, it was possible to discern all nine doublets and the central pair, which was typically seen as a unit (figure 2d). The three-dimensional reconstruction revealed that in other regions not all the doublets were resolved (figure 2d ), owning to the significantly worse axial than lateral resolution in light microscopy [6].
Based on transmission EM data, the axoneme is considered to be circular in its cross-section [33], enabling us to assess whether the expansion in ExM was isotropic. We analysed 27 axonemal regions of procyclic T. brucei specimens expanded either in water or in 84% sucrose, in which the axonemes were oriented perpendicular to the imaging plane. Fitting an ellipse to the axonemal cross-sections revealed that the aspect ratio between the major and minor axis was in all cases close to one (figure 2e; electronic supplementary material, figure S5; i.e. the expanded axonemes were nearly circular). There was no significant difference between specimens expanded in water (1.099 ± 0.037, n = 9) and in sucrose (1.096 ± 0.044, n = 18).
Furthermore, the diameter calculated from the circle area was 1.060 ± 0.022 µm for nine specimens expanded in water, and 0.946 ± 0.010 µm for 18 specimens expanded in sucrose. When divided by the diameter of the axoneme (220 nm according to [34]), it yielded the expansion factor of 4.8 in water and 4.3 in sucrose, well in accord with the measured expansion factors of gels (4.7 and 4.3, respectively).

Expansion microscopy facilitates studies of structures inside the corset
Intriguingly, ExM also facilitated visualization of tubulinbased structures located inside the corset. In both procylic (n = 38 analysed cells) and bloodstream (n = 14) T. brucei cells, these were found in the basal body (BB) area around the proximal part of the flagellum. Their identity and organization became apparent when a three-dimensional volume of the area was reconstructed (figure 3a,b), as there was a very good agreement with the previously published model based on electron microscope tomography [7]. The strongly labelled BB could be distinguished at the base of the flagellum. The BB transited via a hollow transition zone into the axoneme, which was identifiable by the presence of the central pair. The axoneme exited the corset via a space visibly devoid of the microtubules. The probasal body ( pBB) was located adjacent to each BB, and its orientation could be determined. Furthermore, a ribbon-like structure was spiralling alongside the proximal part of each axoneme. The structure started on the side of the BB facing the pBB and extended towards the corset, with which it merged. The position, shape and the fact, that it was stained by various anti-tubulin antibodies (electronic supplementary material, figure S6) were consistent with the structure being the microtubule quartet (mtQ), which subtends the flagellar pocket [7]. The mtQ is typically formed by four microtubules [5,7]. Although in ExM images, the mtQ signal was substantially wider than the signal of a corset microtubule ( figure 3a,b), it was not possible to resolve individual mtQ microtubules. This is consistent with the mtQ microtubules in the BB area being more tightly spaced than the corset microtubules (e.g. fig. 3b in [35]).
Interestingly, in expanded cells, a single microtubule extending alongside the mtQ but terminating before the mtQ integrated into the corset was noticeable (figure 3a,b; electronic supplementary material, figure S6), which corresponded to the neck microtubule previously described by electron microscope tomography [7]. In expanded procyclic cells, this neck microtubule was observed in 54 of 56 BB areas (associated with the single flagellum of non-dividing cells, the old flagellum in dividing cells with two flagella, or the new flagellum in dividing cells, which extended significantly past the exit point). Similarly, the neck microtubule was present in 21 of 22 BB areas in bloodstream cells (for an example, see figure 3b).
EM previously described an even higher complexity of the microtubule-based structures inside the corset of Leishmania [8,36]. Our ExM data were well in accord with this. Moreover, the quantitative aspect of ExM revealed a significant variability in the presence of some of these structures in a population of promastigote L. major cells. Importantly, we also observed that it was straightforward to segment individual structures inside the corset, which has a potential towards their detailed analysis in the future (figure 3c).
We reconstructed volumes of seven cells with a single flagellum and 17 cells with two flagella. All 41 axonemes present in these cells were subtended by the BB with the associated pBB ( figure 3c,d). Moreover, all the axonemes had the mtQ extending alongside their proximal part ( figure 3c,d). Unlike in T. brucei, the mtQs did not intercalate the corset ( figure 3c,d), confirming the EM observations [8]. Although the signal associated with the structure in ExM was wider than a single corset microtubule (figure 3c,d), individual mtQ microtubules could not be resolved. This is consistent with smaller distances between microtubules in the mtQ than in the corset ( fig. 4e in [36]; and fig. 3d in [8]). In four of the seven cells with a single flagellum a microtubule-based structure extending in an arch from the pBB towards the axoneme was present (figure 3c,e); it probably corresponded to the nascent mtQ previously observed by EM in this cell type [8].
Additional structures included pocket microtubules, observed alongside 20 axonemes on the opposite side to the mtQ (figure 3d), and cytoplasmic microtubules, which were also associated with 20 axonemes (only partially overlapping with the set of axonemes with the pocket microtubules), but ran orthogonal to the mtQ (figure 3c,d). Finally, we observed short microtubules or small plate-like structures attached to 14 of the 41 BBs or associated pBBs (figure 3d). It is worth noting that in a minority of cases morphology or positioning of these additional structures deviated from those described by EM, highlighting the significant degree of variability of the microtubule-based structures in the BB area of Leishmania. This, together with the lack of other positional information, such as the flagellar pocket membrane, in the ExM specimens, made in these cases classification of the structures challenging.
Intriguingly, inside the corset of 20 of the 24 reconstructed cells, one (n = 17) or two (n = 3) long microtubules were present. These microtubules were less straight than microtubules of the corset and extended from its anterior end to the central part ( figure 3c,e). Occasionally, an entire microtubule was located in the central part (figure 3e; electronic supplementary material, figure S7). These microtubules corresponded to the microtubules associated with the multivesicular tubule observed by EM in cell sections [36]. Interestingly, all four royalsocietypublishing.org/journal/rsob Open Biol.  royalsocietypublishing.org/journal/rsob Open Biol. 11: 210131 cells in which no multivesicular tubule-associated microtubules were present were in mitosis. However, in other mitotic cells, these microtubules were observed.
In mitotic Leishmania cells, ExM visualized well the mitotic spindle, which was formed by a bundle of microtubules (figure 3f ), as had been previously described by transmission  royalsocietypublishing.org/journal/rsob Open Biol. 11: 210131 EM of serial sections [37]. In expanded cells, individual spindle microtubules were discernible, particularly at spindle poles. DNA staining demonstrated that the spindle was surrounded by the nuclear DNA, segregated by the spindle into two masses (figure 3f ). Cells at later stages of mitosis possessed thin elongated spindles with flared microtubule ends localized at the poles (figure 3g,h). In observed mitotic cells, an ingression of the corset microtubules along the anterior-posterior cell axis was present (figure 3f-h). The cleavage plane was orthogonal to the axis connecting the BBs of the two flagella and to the long axis of the mitotic spindle (figure 3g,h).

Localizing cytoskeletal proteins in respect to microtubule-based structures
ExM has been frequently used to determine the position of various proteins in respect to microtubule-based structures (e.g. [14][15][16]38]). Procyclic T. brucei cells, due to their well-defined morphology, a considerable body of knowledge on protein localization acquired by EM, and availability of high-quality antibodies (see electronic supplementary material, figure S2) appear to be an excellent system to further validate the approach. First, we focused on FAZ and assessed whether ExM can differentiate proteins localizing to distinct domains of the structure. Staining expanded cells with the L3B2 antibody recognizing the FAZ1 protein [21,39] produced a linear signal starting below the point of the axoneme exiting the corset and intimately linked to the mtQ (figure 4a). It further extended along the corset following the axoneme and terminated at the anterior end of the corset (figure 4a). Importantly, the signal was positioned between microtubules of the corset and also extended below its surface (figure 4a). This is consistent with the localization of the FAZ1 protein to the FAZ filament domain in the cell body, as determined by immunogold EM [21,28]. In contrast with immunofluorescence microscopy, the antibody also stained the nucleus and the kinetoplast in ExM (figure 4a; electronic supplementary material, figure S2).
Another FAZ protein, ClpGM6, was previously shown by immunogold EM to constitute FAZ fibres in the FAZ flagellum domain linking the flagellar cytoskeleton to the membrane junction [40]. Staining expanded cells with the anti-ClpGM6 antibody [3] indeed produced a robust signal spanning the gap between the axoneme and the corset (figure 4b). The signal started around the point of the axoneme exiting the corset and terminated past its anterior end (figure 4b).
Next, we focused on individual structures of the flagellar cytoskeleton. First, we applied to expanded cells the L8C4 antibody, which recognizes a constituent of the paraflagellar rod [21], a kinetoplastid-specific structure extending alongside the axoneme. In ExM, the L8C4 signal was indeed juxtaposed to the axoneme (figure 4c). Despite its significant granularity, it was possible to determine that the signal started distal to the point of the axoneme exiting the corset and diminished towards the distal end of the axoneme, which is consistent with immunofluorescence observations (electronic supplementary material, figure S2; see also [21]).
As an example of a protein found inside the axoneme, we studied RSP4/6, an evolutionary conserved member of the radial spokes connecting the microtubule doublets to the central pair [34]. As no antibodies against the trypanosome RSP4/6 are available, we tagged the protein at its endogenous locus with a 3 × HA tag and stained the resulting cells after expansion with an anti-HA antibody. In both, optical cross-sections and longitudinal sections, it was apparent that the anti-HA signal was confined within the space defined by the outer doublets (figure 4d,e). Moreover, in the orthogonal sections, it was noticeable that the central part was devoid of the signal; this corresponded to the space occupied by the central pair microtubules (figure 4e). ExM also revealed that the signal started just distal to the transition zone concomitant with the central pair microtubules (figure 4d).
Next, we studied localization of the protein TbSAXO. This microtubule-associated protein had previously been demonstrated to localize to the axoneme by immunogold EM [41] and has been used as an axonemal marker in T. brucei. Expanded cells stained with the mAb25 antibody [22] recognizing TbSAXO revealed a clear co-localization of the protein with tubulin in the axoneme (figure 4f ). A further interpretation of the mAb25 staining, such as the point at which the signal started along the axoneme, was precluded by the relatively strong speckled background staining in and around the cell. Finally, we assessed the level of positional information achievable by ExM for proteins localizing to discrete regions along the flagellar cytoskeleton. First, we stained expanded cells with an antibody against FTZC, the first described marker of the transition zone in trypanosomes [23]. ExM revealed that individual puncta of the signal surrounded microtubules of the transition zone (figure 4g). Next, we generated a procyclic cell line expressing 3 × HA-tagged KIN-E, the protein localizing along with the flagellum and significantly enriched at the flagellum distal end [42]. In expanded cells, an anti-HA antibody indeed produced a robust staining at the very end of the axoneme (figure 4h). Intriguingly, the resolution provided by ExM allowed to unambiguously localize KIN-E to the distal end of each axonemal doublet as well as to the central pair (figure 4h).

Discussion
In this work, we set out to assess the ExM approach for studying cytoskeletal structures using unicellular kinetoplastid parasites. Morphology of their cytoskeleton, particularly in the case of procyclic and bloodstream T. brucei cells, had been extensively characterized by various EM approaches, with a remarkably little variation between cells in a population. We observed that ExM is capable of faithfully reproducing information on the presence, positioning, and morphology of structures and structural components. A prime example being the mtQ which in three-dimensional reconstructed cellular volumes obtained by ExM had an identical position, helical pattern and handedness as those previously determined by EM tomography [7]. Other notable examples include the groove of the bloodstream cells, so far observed only by serial block-face scanning EM [30], and differential localization of proteins to individual domains of the FAZ, previously determined by immunogold EM [21,40]. Hence, ExM can serve as a substitute of these EM approaches, which require highly specialized instrumentation, when the lateral resolution of about 50 nm suffices.
royalsocietypublishing.org/journal/rsob Open Biol. 11: 210131 Our analysis of the axoneme shape showed that the expansion was isotropic, which is consistent with published observations of tissues, cells, organelles and structures [10,12,43]. Our data, however, also indicated that the degree of expansion was not identical for the studied cytoskeletal structures; while the axoneme expanded 4.8-fold, similar to the 4.7 expansion factor of the gel, the distance between the corset microtubules increased 6.0-fold. Leaving aside the potential caveats associated with estimating dimensions from EM data [44], which were used to calculate these expansion factors, similar and higher discrepancies in expansion factors of individual organelles were recently observed in various experimental systems and using a variety of ExM protocols [17,18,45,46]. Hence, it seems likely that based on their composition individual organelles and structures undergo differential expansion, precluding a straightforward interpretation of measured distances between entities of expanded specimens.
Despite these limitations, there are characteristics which make ExM well suited to complement EM approaches for answering specific questions.
(i) Foremost, ExM is based on immunofluorescence staining and standard confocal microscopy, with image acquisition and subsequent three-dimensional reconstruction being comparatively rapid. Hence, it is straightforward to obtain three-dimensional volumes of tens of cells, as presented here, with even hundreds of cells manageable. This would be challenging with EM approaches and facilitates quantitative super-resolution studies of cytoskeletal structures. Examples presented in this study include the observation of the neck microtubule being an integral part of the BB area in procyclic and bloodstream T. brucei cells, and that the mtQ is found alongside every axoneme in promastigote L. major cells, while pocket microtubules, cytosolic microtubules and multivesicular tubule-associated microtubules are common, but not universally present. Finally, we envisage that ExM will be in a similar way highly useful to study morphology and cellular architecture of various transient life cycle forms and the processes of cell differentiation. (iii) Protein localization by immunogold EM is challenging as antibodies frequently do not reach or recognize their epitopes in resin-embedded sections. By contrast, it has been observed that antibody markers established for the standard immunofluorescence staining generally also stain expanded specimens [11]. Our data are in line with this (table 1). Intriguingly, we observed that in expanded cells the anti-tubulin antibodies stained the axonemal microtubules well, while they have been known to produce only a weak signal in non-expanded cells (figure 1a-c; electronic supplementary material, figure  S2; see also [19]). This probably reflects an increased accessibility of their epitopes in the axoneme due its expansion. Such a positive effect of de-crowding due to ExM was previously described in other systems [12,47]. On the other hand, we observed that two tested antibodies, BBA4 staining the BBs and pBBs in T. brucei [19] and mAb62 staining the flagella connector [24] (see also electronic supplementary material, figure S2), did not produce any specific signal in ExM (electronic supplementary material, figure  S8). This may stem from a degradation of respective epitopes during the SDS-based denaturation step of ExM, not present in a standard immunofluorescence protocol. An ExM approach based on a different homogenization step could be considered [48].
The applicability of ExM to localize proteins in these kinetoplastid species is further augmented by the available repertoire of reverse genetics approaches. We used the rapid PCR-only approach to generate cell lines expressing proteins of interest tagged at their endogenous loci with a small epitope tag [49]. The resolution of ExM images was sufficient to unambiguously localize the radial spoke protein RSP4/6 to the space between the axonemal microtubule doublets and the central pair, and the kinesin KIN-E to the tips of individual doublets as well as the central pair. Hence, this approach opens up the possibility of localizing in the cellular volume and at a relatively high resolution, many proteins encoded in genomes of these parasites.
In conclusion, ExM is a powerful approach for studying cytoskeletal structures in T. brucei, L. major and probably other kinetoplastid species. It is a simple, yet robust super-resolution method, which does not require highly specialized reagents or instrumentation. ExM data recapitulate well observations by various EM approaches. Moreover, ExM is capable of complementing them due to its ability to provide a highly quantitative information.

Expansion microscopy 4.2.1. Fixation of cells
For fixation on coverslips, 2 × 10 6 cells per specimen were used, and for fixation in solution 1 × 10 6 cells per specimen were used. The cells were pelleted by centrifugation for 5 min at royalsocietypublishing.org/journal/rsob Open Biol. 11: 210131 800 g, followed by two washes with PBS, or with PBS supplemented with 3.3 mM glucose (Penta, G01102) in the case of bloodstream T. brucei cells.
For fixation on coverslips, the cells were re-suspended in 50 µl of PBS and transferred onto a coverslip (High Precision Ø 12 mm No. 1.5H, Marienfeld) placed in a 24-well plate (Jet Biofil, TCP011024), to facilitate subsequent steps. Following a 5 min incubation, the cells were washed with PBS and fixed with 500 µl of a freshly prepared fixation solution containing 4% formaldehyde (Sigma, F8775) and 4% acrylamide (Sigma, A8887) in PBS. For fixation in solution, the cells were directly re-suspended in the fixation solution and transferred to a poly-L-lysine (Sigma, P4707) coated coverslip in a 24-well plate. In both cases, the cells were fixed overnight at room temperature, followed by a PBS wash.

Gelation
Gelation was performed on a strip of Parafilm in a wet chamber, which was initially placed on ice. The monomer solution containing 19% sodium acrylate (Sigma, 408220), 10% acrylamide and 0.1% N, N'-methylenebisacrylamide (Sigma, M7256) in PBS was aliquoted and stored at −20°C for up to a month. Immediately prior to gelation N, N, N 0 , N 0 -tetramethylethylenediamine (Sigma, T9281) and ammonium persulfate (Thermo Scientific, 17874) were added to the monomer solution to the final concentration of 0.5% each, and 50 µl of the mixture were quickly transferred onto Parafilm in the wet chamber. Coverslips with the cells facing down were placed on the drop of the monomer solution and incubated for 5 min. The wet chamber was subsequently transferred to 37°C and incubated for 30 min to allow for a proper gel polymerization.

Denaturation and expansion of the gel
The gelation chamber was brought to room temperature and several drops of ultrapure water were added to the coverslips to help the gel detach from Parafilm. Thereafter, the specimens were transferred to a well of a 12-well plate (Jet Biofil, TCP011012) containing 1 ml of the denaturation buffer consisting of 50 mM tris(hydroxymethyl)aminomethanehydrochloride (Serva, 3719202), 200 mM sodium chloride (Lach-Ner, 30093-AP0), 200 mM sodium dodecyl sulfate (Roth, 1057) and pH 9.0. At this point, gels started to expand, which facilitated their separation from the coverslips. Gels together with the remaining denaturation buffer were transferred to a 1.5 ml microcentrifuge tube, which contained 0.5 ml of the denaturation buffer, and denaturation was performed for 1 to 2 h at 95°C. The denatured gels were transferred to Ø 10 cm Petri dishes and expanded by three 20 min incubations with 15 ml of ultrapure water.

Antibody staining of expanded specimens
Water was removed and an approximately 10 × 10 mm piece of the gel was cut per each antibody staining. Staining was performed in a 6-well plate (Jet Biofil, TCP011006), in the dark while gently rocking. Both primary and secondary antibodies were diluted in 500 µl of the ExM blocking buffer consisting of 2% bovine serum albumin (Roth, 8076) in PBS. In general, we observed that both primary and secondary antibodies produced good results when used more concentrated compared to standard immunofluorescence experiments. For antibody concentrations see table 1. The gels were first incubated overnight at room temperature with 500 µl of diluted primary antibodies. Thereafter, they were washed by 3 × 20 min incubations with 3.5 ml of ultrapure water, followed by an overnight incubation with 500 µl of secondary antibodies. Subsequently, the gels were washed 3 × 20 min with 3.5 ml of ultrapure water and selected specimens were also incubated for 30 min at room temperature with 10 µg ml −1 4 0 ,6-diamidino-2-phenylindole (DAPI) (Sigma, D9542) or 10 µg ml −1 Hoechst 33342 (Thermo Scientific, 62249) in PBS while rocking. Finally, the gels were washed 5 × 15 min with 4 ml of ultrapure water. For co-staining experiments, specimens were simultaneously incubated with two primary or two secondary antibodies.

Imaging
Water was carefully aspirated, a stained piece of the gel was transferred to a centre of a glass-bottom dish (Cellvis, D35-20-1.5-N) coated with poly-L-lysine, and a small drop of ultrapure water was put on it to prevent drying. To facilitate imaging, it was important to keep track of the orientation of the gel during the procedure, such that the side of the gel with cells close to its surface was ultimately facing the glassbottom dish. Specimens were imaged using a Leica TCS SP8 confocal microscope with an HC PL apochromatic 63x/NA 1.40 oil immersion objective. Excitation was performed with a 405 nm diode laser (50 mW) in the case of DAPI and Hoechst 33342, a 488 nm solid-state laser (20 mW) in the case of Alexa Fluor 488, and a 552 nm solid-state laser (20 mW) in the case of Alexa Fluor 555. Fluorescence was detected using hybrid detectors (HyD). Z-stacks were acquired with the step size of 100 nm. The pixel size was between 52 and 97 nm, and in the case of high magnification images between 18 and 36 nm. The pixel dwell time was 400 ns. The size of the pinhole was adjusted based on the signal strength, but was typically around 0.4 Airy unit to improve resolution. No averaging was used for z-stacks. In the case of a significant lateral drift, a piece of a gel was embedded in the two-component silicone-glue Twinsil (Picodent, 13001000). Expanded gels could be stored in Twinsil for several days before imaging.

Standard immunofluorescence staining
Procyclic T. brucei and promastigote L. major cells were washed and re-suspended in PBS to 1 × 10 7 cells ml −1 . 50 µm of T. brucei cells were transferred onto a cleaned Ø 12 mm coverslip, followed by a 4 min incubation. After removal of an excessive liquid, the cells were fixed for 5 min with 4% formaldehyde in PBS, washed 3 × 5 min with PBS and permeabilized for 10 min with 50 µl of 0.1% Triton X-100 (Roth, R30512) in PBS. 50 µm of L. major cells were transferred onto cleaned microscopic slides (Menzel Gläser, 630-1985), incubated for 4 min and an excessive liquid removed. The slides were transferred to an ice-cold methanol, incubated for 20 min at −20°C and rehydrated for 30 min in PBS. Both T. brucei and L. major specimens were subsequently washed for 3 × 5 min with PBS and incubated for 1 h with the IF blocking buffer consisting of 1% bovine serum albumin in PBS. Following another 3 × 5 min wash with PBS they were incubated for 1 h with primary antibodies diluted in the IF blocking buffer. Then, the specimens were washed for 3 × 5 min with PBS, incubated for 1 h with secondary antibodies diluted in the IF blocking buffer, washed, mounted into a mounting medium containing 90% glycerol with 1,4-diazabicyclo[2.2.2]octane and 100 ng ml −1 of DAPI, and stored at 4°C overnight. The specimens were subsequently imaged using a Zeiss Axioplan 2 microscope with a Zeiss 100x (NA 1.4) Plan Apochromat oil immersion objective and an Andor Zyla 4.2 camera controlled by Micromanager software [53].

Image analysis
Confocal z-stacks and individual images of non-expanded cells were processed and analysed in Fiji [54]. Three-dimensional reconstructions were performed in the 3D viewer plugin. To estimate the lateral spacing of microtubules areas of corsets with microtubules oriented parallel to the image plane were selected. Consecutive confocal z-planes encompassing a complete signal of these microtubules were projected to twodimensional using the Average intensity algorithm (Image/ Stacks/Z Project). Subsequently, the FFT command (Process/ FFT/FFT) was used to compute the forward Fourier transform and display the resulting frequency-domain image. Using the oval selection tool areas of the frequency-domain image corresponding to the most strongly represented frequencies were selected and marked for the inverse Fourier transform by filling them with white colour. Subsequently, the inverse Fourier transform command (Process/FFT/Inverse FFT) was used to compute the inverse Fourier transform of these frequencies.
To compute the inverse Fourier transform of the remaining frequencies, the areas corresponding to the most strongly represented frequencies were suppressed by filling them with black colour.
Selected z-stacks were deconvolved with Huygens Professional v. 19.10 (Scientific Volume Imaging, The Netherlands, http://svi.nl), using the CMLE algorithm. To segment selected cells their confocal z-stacks were reconstructed into a three-dimensional image using Imaris software 9.7.2 (Bitplane) and segmented by applying background subtraction (local contrast) threshold.

Assessing isotropy of expansion and extent of axonemal expansion
Isotropy of expansion was assessed on raw images (without deconvolution) based on the circular shape of the axonemal cross-section. First, axonemal regions perpendicular to the image plane were selected to avoid regions of elongated cross-sections due to the axoneme being tilted. The perpendicularity check was done in Fiji [54]. The default threshold method was applied to manually selected series of z-planes with a single axoneme present to create an axoneme binary mask. The close operation was performed on the binary mask and the analyse particle tool was used to find the position of the mask centroid. For every five consecutive planes, the positions of the first and the fifth centroid were compared; if their xy coordinates were close to each other, they were further analysed. The threshold for the xy centroid distance was chosen as 4 x z-step size/tan(85 degrees), which corresponded to a 5 degree tilt. Twenty-seven selected axonemal stacks of five z-planes were corrected for the drift (StackReg plugin based on [55], Rigid Body transformation), summed (sum slices), and the default threshold method was used to create a binary mask. The circularity of the mask was assessed using the analyse particle tool by fitting an ellipse to the mask and extracting the aspect ratio of its major to minor axis. The aspect ratio of one indicates a circular object. The analyse particle tool provided also the area of the ellipse. A diameter of a circle with an equivalent area was calculated. The diameter was corrected for blur due to light diffraction, estimated to be 200 nm (the value was obtained by distilling a point spread function in Huygens software from images of 100 nm beads acquired under conditions identical to z-stacks of cells). The corrected diameter divided by the typical diameter of the axoneme (220 nm according to [34]) yielded the expansion factor.