Structure and dynamics of the contractile vacuole complex in Tetrahymena thermophila

Although essential for life, water also endangers any cell that cannot offset diffusional water uptake across its plasma membrane. Inward water diffusion occurs whenever the tonicity of the cell cytoplasm exceeds that of the extracellular fluid, and this can result in catastrophic swelling for cells that lack a rigid cell wall (Ritter et al., 2021). This challenge existed during the evolution of many unicellular protists adapting to fresh-water environments, and is today reflected in the phylogenetically widespread presence of osmoregulatory organelles called contractile vacuole (CV) complexes (CVCs) (Allen, 2000; Jimenez et al., 2022; Patterson, 1980). The CVC consists of one or more large vesicles, sometimes called cisternae or bladders (a name we will use), which are connected to membrane tubules that extend as a network (spongiome) into the cytoplasm (Allen and Naitoh, 2002). In current models, water absorbed into the spongiome from the cytoplasm fills the expanding CV, which subsequently contracts during the phase when its contents are expelled through a transient opening in the plasma membrane (Plattner, 2015). As befits an osmoregulatory organelle, the periodicity of the contractile cycle is sensitive to the extracellular tonicity (Allen, 2000; Docampo et al., 2013; Heuser et al., 1993; Gabriel et al., 1999).

CVCs were first recognized as prominent cellular features more than two centuries ago, but our current understanding of mechanisms underlying cyclic expansion and contraction is still fragmentary, including of the mechanics of contraction itself (Tani et al., 2001, 2000, 2002; Naitoh et al., 1997; Spallanzani, 1776). The most detailed molecular studies have been pursued in Dictyostelium discoideum, Paramecium (species multimicronucleatum and tetraurelia), and the trypanosomid Trypanosoma cruzi, which represent three highly unrelated lineages (Amoebozoa, Alveolata and Rhizaria, respectively). In all three, water transfer from the cytosol into the CVC lumen involves the activity of proton-pumping vacuolar (V-)ATPases in the organellar membrane (Fok et al., 1995; Gronlien et al., 2002; Temesvari et al., 1996). However, because the CVC lumen is not detectibly acidified (Stock et al., 2002a), it is believed that the proton gradients produced by these pumps are harnessed to produce coupled gradients of other ions, which are more directly involved in transport of water (Jimenez and Docampo, 2015) and other small molecules (Stock et al., 2001, 2002a,b; Heuser et al., 1993). Bladders can also undergo fusion with cytoplasmic organelles called acidocalcisomes that independently accumulate water-drawing ions (Rohloff et al., 2004; Marchesini et al., 2002; Montalvetti et al., 2004). However, the details are unclear and might differ between organisms, between which there are also notable morphological differences (Allen and Naitoh, 2002). For example, in trypanosomes, the bladder is localized by attachment via an electron-dense structure to the flagellar pocket, in which transient pores are posited to form during bladder emptying (Girard-Dias et al., 2012). A similar arrangement exists in Paramecium, except that the bladder localization is via linkage to stable plasma membrane pores via dedicated cytoskeletal connections (Allen and Naitoh, 2002). In contrast, in Dictyostelium, both the number and the position of bladders is variable, and their emptying appears to involve kiss-and-run fusion at contingent sites on the plasma membrane (Becker et al., 1999; Essid et al., 2012; Heuser, 2006).

Similarly, the anatomy of the tubular network surrounding the bladder, the spongiome, as well as the connections between the bladder and spongiome, display significant differences between these organisms (Allen and Naitoh, 2002). In Dictyostelium, both the bladder and spongiome are rich in V-ATPase, and the two structures might interconvert during contractile cycles (Gerisch et al., 2002). In T. cruzi the V-ATPase appears to be concentrated in the bladder (Ulrich et al., 2011), whereas the Paramecium V-ATPase is strictly concentrated in a subdomain of the spongiome, called ‘decorated’ (Fok et al., 1995). The Paramecium CVC, which altogether appears to be more highly structured than in Dictyostelium or trypanosomes, also includes a third distinct zone comprising undecorated tubular arms that link the bladder and decorated spongiome (Ishida et al., 1996; McKanna, 1976). The arms appear to become cyclically isolated from the bladder in a way that limits fluid backflow during contraction (Tominaga et al., 1998b; Gronlien et al., 2002; Tani et al., 2002). A similar dynamic arrangement has been described in Trypanosomatids (Linder and Staehelin, 1979).

CV-associated proteins have been identified in Dictyostelium, trypanosomes and Paramecium, but only a small subset of these appeared uniquely localized to the CVC (Plattner, 2013; Bush et al., 1994; Du et al., 2008; Manna et al., 2023; Plattner, 2015; Schonemann et al., 2013; Ulrich et al., 2011; Gerald et al., 2002; Gabriel et al., 1999). This subset has provided key tools both for live imaging of CVC dynamics in the form of fluorescently tagged transgenes, as well as for advancing informative candidates for functional studies (Becker et al., 1999; Du et al., 2008; Gabriel et al., 1999; Gerisch et al., 2002; Harris et al., 2001). In a prior study, we identified such a candidate in the ciliate Tetrahymena thermophila, as part of a survey of CORVET complexes in this organism (Sparvoli et al., 2020). Tetrahymena, like Paramecium, belongs to the Oligohymenophorean subgroup of ciliates, but the two genera diverged from one another hundreds of millions of years ago (Warren et al., 2017). VPS8D encodes one subunit of a hetero-hexameric CORVET complex whose homologs in other organisms are involved in vesicle fusion in the endolysosomal pathway (van der Beek et al., 2019). In Tetrahymena, one of six such complexes appeared to localize to the CVC (Sparvoli et al., 2020). Although the CVC in Tetrahymena species was historically the subject of both physiological and ultrastructural studies (Elliott and Bak, 1964), Vps8Dp (the protein product of the VPS8D gene), together with a small set of Rab GTPases (Bright et al., 2010), were the first potential protein markers specific to this compartment. Recently, other potential candidates were advanced in genetic studies that uncovered two genes, DOP1 and VMA4, that were linked with CVC-related phenotypes (Cheng et al., 2016). DOP1 belonged to the conserved DOPEY family, of which members in yeast and animals were associated with membrane trafficking (Moliere et al., 2022). VMA4 encoded the E-subunit of a V-ATPase complex (Collins and Forgac, 2020).

In this manuscript, we provide evidence that Vps8Dp, Dop1p and Vma4p, together with a novel predicted lipid scramblase Scr7p, are bona fide CVC proteins in T. thermophila. Vps8D and Dop1p also have large dispersed cytoplasmic pools, which undergo exchange with the CVC-localized cohort. By exploiting fluorescently tagged copies of these proteins for live-cell imaging, we developed models for both the anatomy and the contractile cycle of the CVC in this organism. Cells expressing the tagged proteins also provided new insights into the response of the CVC to osmotic stress, and CVC duplication during cell division. Notably, although all four proteins localize to the CVC, they occupy different zones. Dop1p and Vps8Dp both localize to the bladder periphery, but Vps8D is shifted distally. This difference between localization of Vps8Dp and Dop1p becomes more pronounced under hypo-osmotic stress, supporting the idea that the proteins occupy two functionally distinct regions. In addition, Vps8Dp and Dop1p localize to the spongiome, likewise each displaying a distinct distribution. Instead, the two appear locally concentrated within distinct but overlapping zones. Scr7p localization coincides with that of Vps8Dp, whereas Vma4p is strictly localized to the spongiome but not the bladder.

T. thermophila contains a prominent CVC, whose structure, based on past studies, consists of a central bladder that is bridged with the plasma membrane via pores, generally two in number. The bladder is surrounded by a tubulovesicular spongiome. This structure is shown in Fig. 1A. By standard light microscopy, only the bladder is visible, as a round structure in the cell posterior. The bladder goes through cycles of collapse and expansion on a time scale of tens of seconds (Fig. 1B, top row). As previously reported, cells lacking the DOP1 gene instead develop an enormously enlarged bladder, which grows at least in part by undergoing fusion with other large vesicular structures, and only undergoes cyclic collapse on a time scale of minutes (Fig. 1B, bottom row; Cheng et al., 2016). To ask whether the protein product of DOP1 is directly associated with the CV, we targeted the macronuclear DOP1 locus to endogenously tag the Dop1p C-terminus with mNeon. (The micronuclear germline DOP1 locus, which was not modified by our approach, is silent in vegetative cells; Cassidy-Hanley et al., 1997.) Cells expressing Dop1p–mNeon showed bright fluorescence in the cell posterior. In cross sectional views, the bright fluorescent signal was tightly localized to the periphery of the CV bladder, from which it extended both anteriorly and posteriorly as a stripe tracing the cell cortex (Fig. 1C). In cells that are rotated 90° around their long axis, relative to the perspective in Fig. 1C, the Dop1p signal is again most concentrated at the CV bladder and extends in an irregular reticulum that radiates in all directions (Fig. 1D). In near-tangential views, the fluorescent signal appeared as two (or sometimes three) rings whose diameters were smaller than that of the bladder, and whose positions suggested their potential association with the pores bridging to the plasma membrane through which the bladder empties (Fig. 1E). To investigate this further, we took advantage of the fact that bundles of microtubules are known to emanate from the pore junctional zone linking the bladder and plasma membranes (Gaertig et al., 1995; Frankel, 2000), and thus the structures can be stained using anti-α-tubulin antibodies. In cells where we visualized both Dop1p and α-tubulin, the rings formed by the two proteins were concentric but not significantly overlapping (Fig. 1F,G). These images suggest that the junctional pore membrane might have a different composition from the bladder, and that Dop1p localizes preferentially to the latter. In addition to this concentrated fluorescence at the bladder and surrounding spongiome, Dop1p puncta were also lightly dispersed throughout the cell cytoplasm. These puncta showed minor overlap with endocytosed FM4-64 and therefore appear primarily to be associated with non-endosomal compartments (Fig. S1).

The availability of Dop1p as a bright and relatively compartment-specific marker for the CVC offered a new addition to the cell biology reagents available for ciliates, and it allowed us to monitor aspects of the contractile cycle in live cells. In optical sections corresponding roughly to the midpoint of an expanded bladder, one sector of the circumference of the expanded bladder is apposed to the cell periphery. That sector appears to be relatively stable when the CV contracts, while the rest of the bladder apparently collapses down upon it. This results in a brightly fluorescent crescent, which in three dimensions would appear as a lens. That is, the replacement of a CV sphere by a lens can be most easily explained by directional collapse of the bladder so that the cortex-distal zone draws close to the cortex-proximal zone (Fig. 1H). This model suggests that the plasma membrane-adjacent zone of the bladder is selectively stabilized, which might involve the aforementioned microtubules that emanate from the CV–plasma membrane junctions (Fig. 1G). The lens serves as a platform for the refilling of the bladder. As seen in cross section, refilling involves the emergence from the crescent of what appear as blebs, which subsequently expand (Fig. 1H).

The same cycle could also be viewed tangentially in living cells, in optical sections just below the surface that are roughly co-planar with the plasma membrane in the region of its junction with the CV (see cartoon in Fig. 1I). From this perspective, the ring corresponding to the bladder periphery fluctuates in fluorescence intensity over the contractile cycle but shows only minor fluctuations in size (Fig. 1I). These images are consistent with the idea that the crescent seen in the 3rd and 4th panels of Fig. 1H is a cross-section of a lens-shaped compartment that results from collapse of the originally spherical bladder (Fig. 1I). From the same tangential perspective, the blebbing seen in the orthogonal view (Fig. 1H, 4th panel) manifests as twin circles (Fig. 1I, 4th panel), suggesting that this phase of membrane expansion might be structured by the pores. A model for the contractile cycle of the bladder based on these complementary perspectives is shown in Fig. 1J.

During cell division, a subset of structures present in a cell must be duplicated and/or partitioned so that they are inherited by both daughters. In T. thermophila cell division (Fig. 2A), key features of the cell are duplicated along an anterior–posterior axis prior to cytokinesis (Cole and Gaertig, 2022). These include the CV pores, in which a new anterior pair appears at a stage when the micronucleus (MIC), but not macronucleus (MAC), has undergone mitosis. The pre-existing CV remains in the posterior while a second CV forms in the anterior and at the same relative position on the cell cortex (Ng, 1977; Ng and Frankel, 1977; Frankel, 1992, 2000).

We followed this process in cells expressing Dop1p–mNeon. In particular, because previous studies had focused exclusively on the pores themselves as prominent cortical landmarks in fixed cells, the availability of a CV marker allowed us to visualize other compartments and in live cells. At an early stage in the process when only the MIC had undergone mitosis, the Dop1p signal extended anteriorly from the maternal CV beyond the elongating cell midline, to two circular structures located at the cortical position that will be occupied by the newly forming CV (Fig. 2B). These circular structures might be related to the pores but their diameter (2–3 µm) is too large to correspond to mature pores. We favor the idea that the Dop1p-labeled structures instead reflect the beginning of bladder formation, an idea more consistent with the bladder localization data in Fig. 1G. To get better time resolution for when this bladder formation occurs, we simultaneously imaged bladder formation with either newly forming pores, using antibodies against acetyl-α-tubulin (Fig. 2C), or with the newly forming oral apparatus, using antibodies against centrin (Fig. 2D) (Jiang et al., 2020). Surveying a large number of dividing cells, we found that all cells showing newly forming bladders also had new pores (Fig. 2C; Fig. S2) and a fully developed new oral apparatus (Fig. 2D; Fig. S2).

Many of the Dop1p-labeled puncta and Vma4p-labeled reticulum seen in dividing cells are aligned along the cytoskeletal ‘ribs’, called meridians, that underlie the Tetrahymena cortex (Fig. 2B,C and Fig. S3) (Frankel, 2000). Interestingly, DOPEY family members in other organisms can recruit kinesin, a cytoskeletal-based motor protein (Mahajan et al., 2019). We therefore wondered whether T. thermophila Dop1p might associate with the cytoskeleton by a similar mechanism. A pilot immuno-isolation of epitope-tagged DOP1, expressed at the endogenous locus, was analyzed by liquid chromatography coupled to tandem mass spectrometry (LC-MS/MS) to identify potential interacting proteins. We detected two histidine triad motif (HIT)-domain-containing proteins, with corresponding peptide intensities and sequence coverage suggestive of stochiometric complex formation (Fig. S4A–D, Tables S4, S5). HIT domain proteins are phosphatases with a broad substrate range and the two candidate interactors exhibit a domain composition consistent with a predicted role in scavenger mRNA decapping. A further potential interactor, a predicted polytopic transmembrane protein, was detected with much lower confidence. (Fig. S4A–D, Tables S4, S5). Based on a conserved domain search (NCBI), this membrane protein might contain a PHM7_cyt superfamily domain, which is a characteristic of ion transporters including osmosensitive and mechanosensitive ion channels (Wang et al., 2023). To ask whether this protein associates with Dop1p at the CV bladder, we endogenously tagged it with mNeon. At steady state, the tagged protein was not visible at the CVC but instead appeared localized to peripheral structures that are likely to be the Golgi, as well as heterogeneous mobile compartments deeper in the cytoplasm (Fig. S4E; Movie 17). Given this lack of localization to the CV, together with the lack of an obvious connection to RNA metabolism, we did not further pursue these pulldown-defined candidates.

CORVET is a widely conserved hetero-hexameric complex, characterized most extensively in yeast and animals, that facilitates vesicle-vesicle tethering and fusion in endosomal trafficking (Balderhaar and Ungermann, 2013). We previously established that T. thermophila expresses six distinct CORVET complexes, called A through F, and we found that the D complex was likely to associate with CVs by localizing its Vps8 subunit (Vps8Dp) (Sparvoli et al., 2020). To ask how Vps8Dp localization compared with that of Dop1p, we analyzed cells expressing VPS8D endogenously tagged with mNeon. Consistent with previous results, Vps8Dp–mNeon appeared to localize both around the CV bladder and to the spongiome (Fig. 3A,B), and like Dop1p, also revealed the twin connections between the bladder and plasma membrane pores (Fig. 3B, third panel). However, whereas in live-cell imaging the Dop1p fluorescent signal over the bladder and reticulum surfaces appeared relatively continuous, Vps8Dp labeling was distinctly punctate on all structures (Fig. 3B–D). In addition, heterogeneous puncta of Vps8Dp–mNeon were conspicuous throughout the cytoplasm.

As with cells expressing tagged Dop1p, the contractile cycle could be visualized in Vps8Dp–mNeon cells. Analysis of these videos confirmed that contraction is asymmetrical, with the bladder collapsing toward the plasma membrane (Fig. 3C). The images are consistent with the crescent/lens intermediate discussed above, but this structure appears less well-defined because although Dop1p tightly traces the contour of the bladder membrane, Vps8Dp puncta are irregularly distributed near the bladder membrane (Fig. 3B, middle panel). When the contractile cycle is viewed along the axis of the pores, and just beneath the plasma membrane, the bladder is deformed but without undergoing dramatic contraction, consistent with imaging in Dop1p–mNeon cells (Figs 3C,D and 1).

Although the CV bladder and spongiome both contain Dop1p and Vps8Dp, these two compartments are likely to serve non-identical functions and therefore differ in their molecular composition. V-ATPases have been shown to localize to a spongiome in both the amoebozoan Dictyostelium discoideum and the ciliate Paramecium tetraurelia (Fok et al., 1995; Heuser et al., 1993; Fok et al., 1993; Nolta et al., 1993). In T. thermophila, the V-ATPase E-subunit-encoding gene, VMA4, is a potential CVC determinant (Cheng et al., 2016). To test this, we disrupted all macronuclear copies of VMA4. The resulting Δvma4 cells showed greatly slowed cycles of CV contraction (Fig. 4A). To ask whether Vma4p localizes to the CVC, we integrated the mNeon tag at the endogenous VMA4 locus. The fluorescent signal was visible only at the CVC and exclusively within the spongiome. That is, unlike Vps8Dp or Dop1p, Vma4p was not concentrated at the bladder periphery (Fig. 4B,C). Imaging of the edges of the spongiome in live cells revealed that Vma4p-labeled tubules show dynamic extension, retraction and branching (Fig. 4D). The localization of Vma4p to the spongiome but not the bladder underscores the functional difference between those compartments.

The Vma4p–mNeon signal appeared to be evenly distributed along spongiome tubules, which is different to what is seen with either Vps8Dp or Dop1p. To ask whether the same tubules contain Vma4p, Dop1p and Vps8Dp, we expressed Vps8Dp–mCherry in pairwise combination with either Vma4p–mNeon or Dop1p–mNeon (Fig. 4E,F). In both cases, the images are consistent with the idea that all proteins are present and show overlapping distributions throughout the spongiome, but that there are local differences in their intensities (Fig. 4E,F), potentially reflecting membrane domains. These conclusions were consistent with expansion microscopy images of cells co-expressing Vps8Dp–mCherry and Dop1p–mNeon (Fig. 4G).

Given that distinct CVC proteins appear to be differentially distributed within the spongiome, we wondered whether the same might be true of the bladder membrane. To investigate this possibility, we asked whether Dop1p and Vps8Dp differed in their distributions at the bladder periphery. We co-expressed Dop1p–mNeon and Vps8Dp–mCherry from their endogenous loci, and then traced the intensity of each signal along a vector drawn outward from the center of the bladder (Fig. 5A). This analysis revealed that the peak of Dop1p was proximal to that of Vps8Dp (Fig. 5B,C). This could be explained if Dop1p is bound at the bladder membrane, whereas Vps8Dp is associated with a peripheral compartment. The Vps8Dp compartment appears to be less spatially restricted than the Dop1p compartment, given that the intensity profile of Dop1p showed a narrow peak relative to that of Vps8Dp (Fig. 5B). We did the same pairwise comparison between Vps8Dp and Vma4p, the latter shown above to be associated with the spongiome, and as expected we found that Vma4p in such radial plots is yet further spaced from the center (Fig. 5D–F).

Our results suggest that the bladder periphery can be conceived of as several concentric zones, for which Dop1p, Vps8Dp and Vma4p are markers that can be used to map the distributions of additional proteins. For example, we recently identified TTHERM_00999060, which we named SCR7 (Scramblase 7; not published, D.P.R.), as a novel predicted lipid scramblase that localizes in part to the CVC (Fig. S5, Movies 19 and 20). When we co-expressed Scr7p–mNeon with Vps8D–mCherry, the two proteins showed near-complete colocalization (Fig. 5G,H; Fig. S5). Thus, Scr7p appears to localize to the same CVC zone as Vps8Dp. This finding also provides evidence that the distinct localizations seen for other pairs of proteins does not reflect a technical artifact (e.g. homophilic clustering induced by fluorescent protein tags).

Many observations of ciliates by simple light microscopy have revealed that osmotic stressors result in morphological and dynamic changes to the CV bladder (Allen and Naitoh, 2002). The proteins we have identified in Tetrahymena allowed us to extend these observations in live cells by visualizing the spongiome, as well as distinguish the Vps8Dp- and Dop1p-defined zones of the bladder. We exposed the dual-labeled cells described above to mild osmotic stress for a brief period, as described in Materials and Methods, and then measured the same radial plots for the pairs of CV proteins. Interestingly, the distance between the peak intensities of Dop1p and Vps8Dp increased upon osmotic stress (Fig. 5I,J) whereas that between Vps8Dp and Vma4p decreased (Fig. 5K,L). The change in the distribution of Vma4p upon osmotic shock is driven primarily by a striking response to the osmotic stress, namely what appears to be the contraction or collapse of the spongiome (Fig. 5K, right panel). As a result of this contraction, Vma4p signal becomes much more heavily concentrated near the bladder periphery. A composite model including the distributions of Dop1p, Vps8Dp and Vma4p within the CVC is shown in Fig. 5M.

That the CVC is a highly dynamic structure is apparent both from its periodic contraction as well as the tubule extension and branching visualized above. Numerous Rabs and SNAREs localize to the CVC in ciliates, suggesting that the mechanisms underlying CVC dynamics involve active vesicle budding and fusion (Plattner, 2013, 2010; Bright et al., 2010; Turkewitz and Bright, 2011; Schonemann et al., 2013). Similarly, Vps8Dp, as part of CORVET, is likely to be directly involved in vesicle trafficking.

We asked whether Vps8Dp, together with other CVC proteins identified in this study, are dynamically associated with CVC membranes. We performed fluorescence recovery after photobleaching (FRAP) by briefly bleaching a defined sector of the CVC in immobilized cells and measuring the time of fluorescence recovery, which depends on exchange with fluorescent proteins from outside of the bleached area. A limitation of these experiments is that we although we could be certain of bleaching the entire CV this was not possible for the spongiome, given the expanse of the latter. These experiments were done using a confocal microscope, so the appearance of the fluorescent signals differs for some tagged proteins compared to the epifluorescence imaging shown in earlier figures.

 Fig. 6A–F shows FRAP results for Dop1p and Vps8Dp, both of which are peripheral membrane proteins. The data reveal that both proteins dynamically exchange with pools in the non-bleached area, but that exchange of Vps8Dp appears more rapid than that of Dop1p (Fig. 6B,E; Fig. S6A–C). The extent of the recovery of each protein after bleaching might be limited by the relative sizes of the CV-associated versus non-CV pools. To gain some insight into this, we used these images to estimate the relative amounts of each protein pool at steady state that is localized to the CV versus non-CV area, subject to the limitation regarding defining the precise spongiome borders. This analysis suggests that both Dop1p and Vps8D include significant pools of cytosolic protein, but that this fraction is higher for Vps8Dp (Fig. 6C,F; Fig. S6D–F).

To ask more directly whether the cytosolic pool of Vps8Dp exchanges with the CVC pool, we used fluorescence loss in photobleaching (FLIP). We repeatedly bleached the CVC of immobilized cells expressing Vps8Dp–mNeon over a period of minutes, while monitoring the Vps8Dp–mNeon fluorescence at a distant site in the cell anterior (Fig. 6G). The results are consistent with exchange between cytoplasmic and CV-localized pools, given that Vps8D fluorescence in the cell anterior decreases upon bleaching of the CV pool (Fig. 6H,I).

Vma4p, unlike Dop1p and Vps8Dp, is a subunit of an integral membrane protein complex. In addition, unlike Dop1p and Vps8Dp, the entire pool of detectible Vma4p is localized to the CVC. We used FRAP to ask whether Vma4p is mobile within the spongiome membrane (Fig. 7A,B; Fig. S6G,H). After bleaching a subsection of the spongiome, we observed that Vma4p–mNeon fluorescence initially begins to recover at one edge and then progressively spreads (Fig. 7C). This is consistent with the idea that Vma4p is mobile within the plane of the membrane, and that exchange is primarily due to diffusion along the tubules. In contrast, when the same analysis was performed on cells expressing Dop1p–mNeon, the fluorescence recovery was not directional but instead occurred in patches throughout the bleached area (Fig. 7D). This result is consistent with the idea that Dop1p exchange occurs via a cytosolic pool, which could be either soluble or vesicle associated. This difference between the spatial patterns of fluorescence recovery of Vma4p and Dop1p could also be seen in kymographs plotting the recovery of fluorescence (Fig. 7E).

Organelles in the endomembrane network are maintained by a balance of inward and outward membrane traffic. The relevant trafficking pathways for the CVC are poorly understood in any organism. Recent analysis in Dictyostelium points to the involvement of endosomal fusion machinery, although the specific roles are as yet unexplored (Manna et al., 2023). Similarly, an important role for endosomal trafficking to the ciliate CVC is consistent with the markers we have established above, given that CORVET in other organisms functions as an endosomal or endolysosomal tether–fusion complex, and the DOPEY family proteins are similarly endosome associated. Both V-ATPases and lipid scramblases function in many compartments, including endosomes (Marshansky and Futai, 2008; Hankins et al., 2015).

In Tetrahymena, as in many other cells, fluorescent styryl dyes can intercalate into the outer leaflet of the plasma membrane and be taken up into endocytic vesicles. Such dyes can therefore be used to trace endocytic trafficking (Betz and Bewick, 1992; Betz et al., 1996; Elde et al., 2005). One would also expect that FM dye in the cell medium could diffuse into the CV through the open pore during bladder emptying, and this has been documented in Dictyostelium (Nishihara et al., 2007). We confirmed those results, detecting clear CVC labeling after a 5 min incubation of Dictyostelium with red fluorescent FM4-64 (Fig. S7A). We then similarly incubated Tetrahymena, expressing tagged CVC markers, with FM4-64, and then looked for overlap between the FM dye and either Dop1p (Fig. 8A) or Vma4p (Fig. 8B). Strikingly, although a large range of cytoplasmic vesicles were labeled, there was no detectible overlap with the CVC markers after either 15 or 30 min of FM update, with the same result obtained after an overnight incubation (Fig. S7B, Movies 18A,B). Thus, the CVC in Tetrahymena appears to be isolated from bulk endocytic traffic, and in addition, some barrier appears to exist to free diffusion through the pore. Our data could also be explained if an unusual lipid composition in the luminal leaflet of the CVC bilayer resists intercalation of FM dyes.

Although the CV has been recognized by biologists for more than three centuries, the structure, and particularly the mechanisms underlying periodic contraction, are still incompletely understood. This lack of knowledge is partly attributable to the fact that although CVs are widely distributed among eukaryotes, they are not present in fungi, animals or plants, where most cell biology has been pursued. A corollary hurdle is that many of the organisms that do bear CVs have been inaccessible to many genetic and molecular approaches that are crucial for the analysis of such complex cell biological structures. CVs are prominent in both marine and fresh-water ciliates, and experiments and observations in Paramecium spp. have been important in generating our current understanding of the CV (McKanna, 1976; Kitching, 1939). Among the ciliates, the species with the best developed toolbox is T. thermophila (Ruehle et al., 2016). In this article, we establish a set of live markers for the CV in this organism. Whereas some CV-associated proteins have previously been identified in both Tetrahymena and Paramecium, the uniqueness of the localization has rarely been established, and/or other limitations precluded their use as live markers. We previously reported a set of Rab GTPases which, as over-expressed GFP-tagged proteins, appeared to be highly concentrated at the CV (Bright et al., 2010). However, these tagged proteins might not be readily visible when expressed at endogenous levels, a problem we have encountered with many Rabs in this organism (A.P.T., unpublished observations).

As shown in this manuscript, four proteins, two of which are peripherally associated with the membrane and two of which are membrane embedded or subunits of integral membrane complexes, could be tagged and used to visualize CVC dynamics when expressed at endogenous levels. These results significantly expand our ability to analyze the CVC in ciliates. Although the bladder could already be visualized even in the 18th century by light microscopy, our ability to capture videos of fluorescent bladders has revealed new details of the contractile cycle. The observation that Dop1p localized to a subtly different position at the bladder periphery compared to Vps8Dp and Scr7p, and that these zones show differential responses to osmotic challenge, should allow the mapping of subdomains in this complex organelle. In Paramecium, where the CVC is morphologically more complex than in Tetrahymena, there is a notable substructure within the spongiome where distal versus peripheral tubules show differential occupancy by V-ATPase as well as an aquaporin (Ishida et al., 2021).

Fluorescently tagged proteins that label the CVC structures also permitted us to recognize novel features of CVC duplication in mitotic cells, including the discovery that newly forming Dop1p-labeled bladders appear as two rings intermediate in size between pores and mature bladders. Previous studies of CVC duplication have relied on the pores, visible using silver staining, which appeared in relatively late stages of cell division (Frankel, 1992, 2000; Cole and Gaertig, 2022). We found that all dividing cells showing newly forming bladders also have new CV pores and a fully developed new oral apparatus. The identification of CVC-specific genes and proteins now presents opportunities for exploring developmental regulation, extending recent studies on cortical determinants for position-specific organelle biogenesis in ciliates. These studies have highlighted the crucial role of conserved signaling proteins in defining cortical pattern formation during cell division (Jiang et al., 2017, 2019a,b, 2020). One open question is whether CVC-specific proteins interact with the newly identified morphogens to influence the positioning and assembly of the new CVC. The precise positioning of new CV pores has been proposed to be regulated by their positioning relative to other organelles, rather than being fixed to a specific cortical pattern (Ng, 1977; Ng and Frankel, 1977). Intriguingly, we found that tubules labeled with Dop1p puncta as well as Vma4p, extend along meridians during cell division and potentially establish a connection between the pre-existing and new CVCs, so that the latter might derive in part from the former either informationally and/or compositionally.

The mechanisms underlying bladder contraction are still under investigation (Velle et al., 2023). CV bladders that were micro-surgically excised from Paramecium briefly continue to undergo periodic contraction, suggesting that remodeling might not depend on an organized cytoskeleton (Tominaga et al., 1998a; Tani et al., 2000, 2002). Ultrastructural studies have revealed what could be an unusual packing of CV membranes during contraction, with possible similarity to lipid cubic phases (McKanna, 1976; Tominaga et al., 1999; Allen, 2000). A distinct lipid composition of the Paramecium bladder has been proposed to directly drive contractile cycles in a model that could involve flipping across the membrane bilayer (Tani et al., 2002), but as yet this model is untested. Our finding that the predicted lipid scramblase Scr7p is strongly concentrated at the CVC corroborates that lipid transfer plays a role in this organelle, which we hope to elucidate in the future. We found that Scr7p closely colocalizes with Vps8Dp, distal to the localization of Dop1p at the bladder periphery, but higher resolution imaging will be needed to understand precisely what structures are involved.

The homologs of DOP1 and VPS8D in other organisms are endosomal proteins (Balderhaar and Ungermann, 2013; Moliere et al., 2022; van der Beek et al., 2019). An inference, consistent with the implications from work in other organisms, is that the CVC is a specialized endosomal compartment. However, the CVC in Tetrahymena does not appear to directly receive traffic from endocytic vesicles, judging from our results with the endocytic tracer FM4-64. Our results also suggest that there is little or no inward diffusion of FM dye through the CV pore. However, the absence of FM staining of the CVC could alternatively be explained if an unusual lipid composition in the luminal leaflet of that organellar bilayer resists intercalation of FM dyes. A better understanding of ways in which the CVC is integrated into the endomembrane network will be facilitated by identifying and analyzing additional CVC-specific proteins in experimentally tractable organisms. In parallel, developing tools to manipulate a wider phylogenetic range of eukaryotes should make it possible to determine whether CVs in diverse organisms depend on evolutionarily shared mechanisms, and therefore to shed insight into the origins of this organelle (Manna et al., 2023).

Tetrahymena thermophila strains used in this work are listed in Table S1 with culture conditions as previously described (Gorovsky et al., 1975). Cells were grown in SPP medium [2% proteose peptone (Gibco, 211684), 0.1% yeast extract (BD, 212750), 0.2% dextrose (ACROS, 41095-5000) and 0.003% EDTA, ferric-sodium salt (Sigma)] supplemented with 100 μg ml⁻¹ Normocin^TM (InvivoGen), an antimicrobial reagent. Cells were grown at 30°C with agitation at 100 rpm. Cell densities were measured using a spectrophotometer (Thermo Spectronic Unicam), where the optical density at 500 nm (OD₅₅₀)=1 corresponds to ∼10⁶ cells ml⁻¹ (Spector et al., 1998) or a Z1 Coulter Counter (Beckman Coulter Inc., Indianapolis, Indiana). Cultures were generally used during the log phase of cell growth (cell density 2×10⁵–4×10⁵ cells ml⁻¹).

50 ml Tetrahymena cultures were grown to 500,000 cells ml⁻¹ and starved for 10–16 h in 10 mM Tris-HCl buffer, pH 7.4, at 30°C with agitation. Biolistic transformations were performed as described previously (Sparvoli et al., 2018; Cassidy-Hanley et al., 1997). Transformants were identified after 3 days selection in paromomycin sulfate (120 μg ml⁻¹ with 1 μg ml⁻¹ of CdCl₂), and then serially passaged five times per week for ∼4 weeks in decreasing concentrations of CdCl₂ and increasing concentrations of paromomycin sulfate. Reagents were from Sigma-Aldrich unless otherwise noted.

The macronuclear open reading frame (ORF) of the VMA4 gene (TTHERM_00193660) was replaced with the paromomycin (Neo4) drug resistance cassette (Mochizuki, 2008) via homologous recombination with the linearized vectors pVMA4KO-Neo4. PCR was used to amplify the fragments of the genomic regions upstream (5′UTR), 1248 bp and downstream (3′UTR), 1481 bp of the ORF. The amplified fragments were subsequently cloned into restriction sites generated by NotI/PstI for the upstream fragment and BamHI/XhoI for the downstream fragment, flanking the Neo4 cassette of the pNeo4 vector, by Quick Ligation (New England, Biolabs Inc.). The primers used to create these constructs are listed in Table S2. The construct was linearized by digestion with SacI and KpnI and transformed into CU428.1 cells by biolistic transformation. Restriction enzyme reagents were from New England BioLabs.

The p2mNeon-6myc-Neo4 vector (Sparvoli et al., 2018) was designed to integrate two mNeonGreen fluorescent tags at the C-terminus of a targeted macronuclear target gene. For each target gene, we used PCR with a proofreading Taq polymerase (Roche, Expand Long Template PCR System) to amplify the C-terminal region (∼0.5–1 kb, but not including the stop codon) as a 5′-flanking fragment, and the 3′UTR as a 3′-flanking fragment. The PCR products were verified by agarose gel electrophoresis and the purified DNA fragments were cloned into p2mNeon-6myc-Neo4 vector at the Xbal site for 5′-flanking fragments and the Xhol site for 3′-flanking fragments, respectively, by using NEBuilder HiFi DNA assembly. The final construct was verified by restriction mapping, and constructs were isolated from 50 ml bacterial cultures using a plasmid DNA extraction kit (Takara Bio USA, Inc. NucleoSpin Plasmid). DNA concentrations were measured by a NanoDrop spectrophotometer. 20 µg of each construct was linearized with PvuII and then purified for each biolistic transformation. The primers used are listed in Table S2. DNA constructs are available in the Tetrahymena stock center (https://tetrahymena.vet.cornell.edu).

pVPS8D-3xmCherry-2HA-CHX was previously constructed to endogenously express mCherry-tagged Vps8Dp (Sparvoli et al., 2020). We modified the construct to replace the cycloheximide (CHX) selection cassette (bounded by PstI and XhoI sites) with a cassette conferring resistance to puromycin, which we obtained by amplifying the selectable Pur4 marker by PCR from the pPur4-opt vector (Iwamoto et al., 2014). Assembly was using the NEBuilder HiFi DNA Assembly method. The pVPS8D-3xmCherry-2HA-PUR4 was biolistically transformed into cells already expressing Dop1p–mNeon or Vma4p–mNeon.

We used two methods for cell immobilization for live-cell imaging. For both methods, we prepared a single slide at a time and viewed it immediately. The first method relied on the pressure exerted on small volume samples under a cover slip. Cells were first concentrated by centrifugation (1000 g for 1 min) to 2×10⁶–5×10⁶ cells ml⁻¹. 6 μl of sample was applied to the slide, and immediately overlayed with a 22×22-1 cover slip. The resulting pressure frequently results in cell immobilization, which could be verified by monitoring ciliary beating on the cell surface at the same time that cell viability could be monitored by observing the periodic contraction of the CV, both in the DIC channel. For the second method, cells were pelleted (1000 g for 1 min) and resuspended in a thermoreversible gel on ice, whose viscosity increases when it is warmed to room temperature. Cells at the same high density were mixed well with CyGEL^TM (ab109204, Abcam) (the mix ratio was optimized for each experiment) and immediately mounted with a coverslip. This approach results in many cells being effectively immobilized in the thin gel. For both methods, we carefully watched for proliferation of large cytosolic vesicles. Proliferating cells have large food vacuoles derived from the oral apparatus. No new food vacuoles form in immobilized cells, so any substantial increase in the number of such vesicles is likely to represent autophagosome formation as part of a stress response. We rejected any cells showing such an increase.

Cells were fixed in a final concentration of 4% paraformaldehyde (stock: 16% paraformaldehyde solution in distilled water, EM grade, 15710, Electron Microscopy Science) for 10–30 min at room temperature. They were then washed repeatedly in PBS to reduce the background autofluorescence, before staining with DAPI at final concentration of 50 ng ml⁻¹ for 10 min for nuclear staining, and/or staining with antibodies for immunofluorescence. To improve preservation of the tubular structure of the spongiome, we fixed cells with 3% paraformaldehyde for 10 min on ice. Images were captured by using a Carl Zeiss Microscope stand Axio Observer 7 system or Marianas Yokogawa-type spinning disc inverted confocal microscope.

Cells were immobilized as described above. FRAP experiments were performed using a Marianas Yokogawa-type spinning disc inverted confocal microscope with a 100×/NA1.45 oil (Alpha Plan-Fluar) objective. The system featured fast shutter speeds and channel switching for high-speed imaging and a vector high-speed point scanner for bleaching, controlled using Slidebook software. For each experiment, we first identified a well-immobilized cell and verified the fluorescence signal with a ∼100–500 ms exposure. We then used the time-lapse mode to capture images with the fastest camera speed for a pre-photobleaching record and then used the screen selection tool and real-time imaging to draw a region of interest (ROI) for photobleaching.

The image processing package FIJI was used for image processing and analysis (Schindelin et al., 2012). Processing of raw images included: image cropping and rotation, bleach correction, background subtraction, adjustment of brightness/contrast, color switching, selection of images from z-stacks, intensity threshold adjustments, and measurements of parameters including areas, mean gray values and integrated densities. The Radial Profile plugin (https://imagej.net/ij/plugins/radial-profile.html) was used to measure and plot the intensity of fluorescence signal along specified radii. An ROI corresponding to the CV bladder was created as a circle whose center corresponded with the apparent bladder center and whose circumference included the entire visible CV periphery. The normalized integrated intensity of the tagged fluorescent protein was calculated along many radii, and the average intensity at each radial position was plotted as a function of distance from the circle center.

To quantify fluorescence intensities in the CV versus cytosol, all images were captured using the same microscopy settings. The raw images were used to adjust the threshold signal intensity for segmenting the whole cell area from extracellular background, and for segmenting the CV area from the cytosol, both using Binary selection with Fill Holes. The sum fluorescence intensities of the whole cell and the CV area were then measured. For FRAP analysis, the time-lapse images were separately adjusted by bleach correction for the pre-photobleaching series and after-photobleaching series. The intensity thresholds were adjusted for background subtraction, and intensity changes over time were measured in ROIs. Intensities were plotted using GraphPad Software Prism.

Cells expressing tagged proteins were grown to log growth phase in SPP medium. For hypo-osmotic treatment, cells were pelleted by centrifugation at 1000 g for 1 min and resuspended gently in 10 mM Tris-HCl buffer, pH 7.4 for 1 min. This process was repeated, and the cells were then incubated in the same buffer, at room temperature, for 10 min before being fixed for imaging.

Cells (2×10⁵) were fixed with 4% paraformaldehyde for 30 min at room temperature, and immunolabeled as previously described (Briguglio et al., 2013; Bowman and Turkewitz, 2001). Briefly, cells were first incubated for blocking with 5% bovine serum albumin (BSA, BP1600-1 Thermo Fisher Scientific) in PBT [PBS with 0.3% Triton X-100 (ACROS) in PBS] for 1 h at room temperature and then incubated with the primary antibodies in the same blocking buffer. Anti-α-Tubulin antibody (clone DM1A ZooMAb® Mouse Monoclonal, Sigma ZMS1039) was used at 1:2000, anti-acetyl-lysine 40-α-tubulin antibody (monoclonal 6-11 B-1, Sigma) and anti-centrin antibody (monoclonal 20H5, EMD Millipore) at 1:500 (Jiang et al., 2020) overnight at 4°C, followed by three washes in PBT. The secondary antibodies were Texas Red-conjugated anti-mouse-IgG, incubated for 1 h at room temperature (1:1000, Texas Red Goat anti-Mouse IgG, Invitrogen T2767), and similarly washed. Cell nuclei were stained with DAPI (50 ng ml⁻¹).

The method was adapted from Damstra et al. (2022). Growing cells were fixed with 4% PFA for 10 min at room temperature. Fixed cells were incubated with 0.1 mg ml-1 acryloyl X-SE (AcX, 2541676, Invitrogen) for a minimum of 3 h in a dark humidified chamber. For gelation, 194 µl (for two samples) of the monomer solution (sodium acrylate 1.1 M; acrylamide 2 M; N,N′-methylenebisacrylamide 0.009% in PBS) was prepared and kept on ice. Tetramethylethylendiamine (TEMED, 11042F, Gibco BRL, 3 µl of 10% stock in water) and ammonium persulfate (A3678, Sigma, 3 µl of 10% stock in water) were added and quickly mixed on ice, and 97 µl of gelation solution was mixed on the slide with 3 µl of concentrated fixed cells. The slides were incubated in the dark at 37°C for 1 h, after which proteinase K (42-700, Apex Bioresearch) was added (final concentration 7.5 units ml⁻¹) for at least 2 h in the dark at 37°C. Afterward, the gels were gently removed and transferred to a 15 cm Petri dish with three successive water washes water for promotion of expansion. The extent of anisotropic expansion was estimated by measuring the diameter of the gels in two orthogonal directions. For imaging, small pieces were cut from the gel and mounted (cells facing down) on the coverslip (glass bottom microwell dishes, part no. P35G-1.5-10-C), which had been pre-coated by adding ∼70 µl of 20% poly-L solution and dried for 30–60 min on a 37°C heat block. Finally, 20 µl of water was added and the sample covered with a second cover slip.

p-FLAG-ZZ-Neo4 was used to construct and generate a strain expressing endogenous Dop1p with a FLAG epitope at the C-terminus of the DOP1 genomic locus. The initial Dop1p–FLAG-expressing transformants were serially passaged at least 20 times, while increasing the concentration of paromomycin from 120 μg/ml to 1000 μg/ml and decreasing the inducer concentration of CdCl₂ from 1 μg/ml to 0.05 μg/ml. Cryomilling was based on previous procedures (Obado et al., 2016; Sparvoli et al., 2020). 10 l cultures of Dop1p–FLAG-expressing CU428 cells were grown to 2×10⁵ –5×10⁵ cells/ml, washed once with 10 mM Tris-HCl pH 7.4 and re-pelleted (1000 g for 10 min). Supernatants were rapidly aspirated to leave a dense cell slurry. The slurries were dripped from a volumetric pipette into liquid nitrogen, and the frozen beads collected and milled to powders using a Cryogenic Grinder 6875 freezer mill, and stored at −75°C.

Pulldowns were as performed described previously (Sparvoli et al., 2020); 10 g of cryomilled cell powder was weighted and immediately suspended in 50 ml dissolving buffer containing 20 mM HEPES pH 7.4 and 250 mM NaCitrate on ice, supplemented with protease inhibitor cocktail tablets (Roche), gently mixed by inversion for 1 h at 4°C, and then on ice until no solid matter was visible. The solutes were then centrifuged at 140,000 g (Beckman Instruments type 45 Ti rotor) for 1.5 h at 4°C, and the supernatants transferred to new tubes, for incubation at 4°C for 2 h with anti-FLAG beads (EZ view Red Anti-FLAG M2 affinity Gel, Sigma) which had been pre-washed with lysis buffer for 2 h at 4°C. The beads were then washed five times with 20 mM Tris-HCl pH 7.4, 1 mM EDTA, 500 mM NaCl, 0.1% NP-40, 1 mM DTT and 10% glycerol supplemented with protease inhibitor cocktail (Roche) and eluted with 60 μl of 4× LDS sample buffer (988 mM Tris, 2.04 mM EDTA, 8% lithium dodecyl sulfate, 40% Glycerol, 0.88% Commassie Brilliant Blue G250, 0.7 mM Phenol red) plus 40 mM DTT for further SDS-PAGE electrophoresis. Protein samples were electrophoresed on a 4–20% gel SDS-PAGE gel to verify pulldown efficiency by Coomassie Blue staining and western blotting using anti-FLAG antibody as below. The eluted sample (total in 60 μl sample buffer) was used to run an SDS-PAGE, with the dye front allowed to electrophorese ∼1 cm into the gel. The gel was then stained with Coomassie Blue R-250 solution (0.1% w/v Coomassie, 10% acetic acid and 50% methanol) to visualize the lane. A single 1 cm gel slice was excised and prepared for protein identification via mass spectrometry.

The protein containing gel slice was destained, and then subjected to tryptic digest and reductive alkylation using standard procedures. LC-MS/MS was performed by the OMICS Proteomics Facility at Biocev on a Ultimate3000 nano rapid separation LC system (Dionex) coupled to a Orbitrap Fusion mass spectrometer (Thermo Fisher Scientific). Mass spectra were processed using the intensity-based label-free quantification (LFQ) method of MaxQuant version 1.6.6.0 (Cox and Mann, 2008; Cox et al., 2014) searching the T. thermophila annotated protein database from https://ciliates.org/ (Eisen et al., 2006; Stover et al., 2006). The minimum peptide length was set at seven amino acids and false discovery rates (FDR) of 0.01 were calculated at the levels of peptides, proteins and modification sites based on the number of hits against the reversed sequence database. If the identified peptide sequence set of one protein contained the peptide set of another protein, these two proteins were assigned to the same protein group. Potential interactors were ranked by enrichment ratios comparing to a control sample and respective LFQ intensities. Proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository (Perez-Riverol et al., 2019) with the dataset identifier PXD043302.

Total cell protein extract and immunoprecipitated samples were analyzed by western blotting as previously described (Sparvoli et al., 2018). Proteins were resolved with the Novex NuPAGE Gel system (8% or 4–20% Tris-Glycine gels, Invitrogen), and transferred to 0.2 μm PVDF membranes (Thermo Scientific). Blots were incubated with anti-FLAG (1:2000 dilution, F1804, Sigma) antibodies for overnight at 4°C after pre-blocking with 3% BSA in TBST (1× Tris-buffered saline with 0.1% Tween 20). Secondary antibody conjugated to ECL Horseradish Peroxidase (NA931) (GE Healthcare Life Sciences, Little Chalfont, UK) was diluted 1:20,000 for incubation with the blots, and signal was visualized by SuperSignal West Femto Maximum Sensitivity Substrate (Thermo Fisher Scientific).

Proliferating Vps8Dp–mNeon-expressing cells were incubated with 5 μM FM4-64 (Invitrogen) and then washed twice with S-medium. After 30 min, cells were mounted and immobilized on slides for live imaging in a Zeiss Axio Observer 7 microscope.

To estimate the extent of colocalization, the Fiji-JACoP plugin was used to calculate Manders’ coefficients M1 and M2 (Bolte and Cordelières, 2006) between two images captured in different channels. M1 is defined as the fraction of the channel A image overlapping the channel B image, and M2 is defined conversely.

Dictyostelium discoideum wild-type strain AX2 cells (a kind gift of Andrew Maselli, Chicago State University, USA) were cultivated in HL5 medium (Ashworth and Watts, 1970) on tissue culture-treated plastic 150 mm plates (Fisherbrand Tissue Culture Dish) at 21°C. For FM4-64 staining, proliferating cells were scraped from the plate and suspended in a tube of fresh HL5 medium with 5 μM FM4-64 for 5 min. The cells were mounted on slides for live imaging in a Zeiss Axio Observer 7 microscope.

We thank Daniela Sparvoli for the Vps8D–mCherry construct and advice on expansion microscopy, and Masaaki Iwamoto (Nihon University) for the puromycin drug selection plasmid. At the University of Chicago, we thank Vytas Bindokas and Christine Labno for help at the Integrated Light Microscopy Core Facility. In the Turkewitz laboratory, we thank Patrick Jiang, Maya Waarts and Josefina Hernandez for helpful discussions.

The peer review history is available online at https://journals.biologists.com/jcs/lookup/doi/10.1242/jcs.261511.reviewer-comments.pdf