Mixed clonal-aggregative multicellularity entrained by extreme salinity fluctuations in a close relative of animals

Multicellularity evolved multiple times independently during eukaryotic diversification. Two distinct mechanisms underpin multicellularity: clonal development (serial cell division of a single precursor cell) and aggregation (in which independent cells assemble into a multicellular entity). Clonal and aggregative development are traditionally considered to be mutually exclusive and to result from independent acquisitions of multicellularity. Here, we show that the choanoflagellate Choanoeca flexa, a close relative of animals that forms contractile monolayers of cells (or “sheet colonies”), develops by an unconventional intermediate mechanism that we name “clonal-aggregative multicellularity”. We find that C. flexa sheets can form through purely clonal processes, purely aggregative processes, or a combination of both, depending on experimental conditions. To assess the ecological relevance of these findings, we characterize the natural context of multicellular development in the native environment of C. flexa on the island of Curaçao. We show that the C. flexa life cycle is environmentally regulated by extreme salinity fluctuations in splash pools undergoing cycles of evaporation and refilling. Upon desiccation, C. flexa colonies dissociate into drought-resistant quiescent cells, which resume activity and reform multicellular sheets after rehydration. We hypothesize that clonal-aggregative development reflects selection for fast transitions into and out of multicellularity in the ephemeral context of coastal splash pools. Our findings underscore the potential of the exploration of biodiversity for revealing new fundamental biological phenomena and expand the known option space for both multicellular development and for the origin of animal multicellularity.


Introduction
The transition to multicellularity was a key event in animal evolution.Multicellularity has evolved independently over 45 times across eukaryotes (1) and can develop by two main mechanisms canonically depicted as mutually exclusive (2)(3)(4): clonal development or aggregative development (also referred to as "staying together" versus "coming together", respectively (5)).During clonal development, a single founder cell undergoes serial cell divisions without separation of sister cells, resulting in a genetically identical multicellular structure.In contrast, aggregative development refers to the formation of multicellular entities by grouping and coalescence of independent cells.Contrary to clonal development, aggregation allows genetically distinct cell populations to combine into chimeric multicellular structures (6).The potentially low degree of genetic relatedness within aggregates can result in evolutionary conflict (e.g.vulnerability to "cheaters"; see glossary) and has been argued to limit the potential for aggregation to give rise to complex multicellularity, alongside the fact that aggregative multicellularity tends to be transient and facultative (7)(8)(9)(10)(11).Indeed, all five eukaryotic lineages with "complex multicellularity" (i.e.controlled three-dimensional morphogenesis and spatial cell differentiation), including animals, fungi, and land plants, develop clonally rather than by aggregation (3,8).
Despite the apparent evolutionary limitations of aggregative multicellularity, one remarkable feature of aggregation is the speed with which aggregative organisms can complete multicellular development from a unicellular state (often over a few hours (6,6,12)).
Aggregative multicellularity is often a stress response to environmental changes such as nutrient depletion (13,14) or the presence of predators (15)(16)(17).In such situations, aggregation allows for rapid establishment of a multicellular structure that confers some selective advantage (e.g.increased size, shielding of inner cells by a protective cell layer, or elevation from the substrate via formation of a stalk).This mechanism is faster than clonal division, and might be especially beneficial when cell proliferation (and therefore clonal development) is infeasible (6,13,17).Thus, it has been argued that aggregation can be ecologically more advantageous than clonal development in fast-fluctuating environments, including conditions of acute stress (as an "emergency response") (6,10,18).
Efforts to reconstruct the evolution of multicellularity in the stem lineage of animals have benefited from the study of their closest living relatives (the unicellular and facultatively multicellular holozoans (see glossary)) and notably of their sister group: the choanoflagellates (Figure 1A) (19)(20)(21)(22).Choanoflagellates are bacterivorous aquatic microeukaryotes bearing an apical flagellum surrounded by a collar of interconnected actin-filled microvilli (Figure 1B-C) (23).Flagellar beating causes a water flow that carries bacterial prey toward the collar, where the bacteria are trapped and ultimately phagocytosed.Choanoflagellates can differentiate into diverse cell types, and can develop into facultative multicellular forms in many species (3,(23)(24)(25)(26).As in animals, multicellular development is clonal in all choanoflagellate species investigated so far, which contrasts with the occurrence of aggregative multicellularity in more distantly related lineages, such as filastereans (27)(28)(29)(30) and dictyostelid amoebae (13,31).These observations have inspired the hypothesis that animals evolved from organisms with a simple form of clonal multicellularity, akin to that observed in certain modern choanoflagellates (3,(32)(33)(34).
Here, we describe an unusual type of multicellular development in the colonial choanoflagellate Choanoeca flexa (35), which challenges previous generalities about choanoflagellates.We report that C. flexa colonies develop by a combination of clonal and aggregative development, which we refer to as 'clonal-aggregative multicellularity'.This mixed mode of development has never been formally described in choanoflagellates or other protists to our knowledge and nuances the textbook dichotomy between clonal and aggregative

Results
C. flexa was discovered in 2018 in splash pools (see glossary) on the rocky coast of the Caribbean island of Curaçao in the form of curved multicellular "sheet" colonies (Figure 1A-D) (35).Cells within C. flexa colonies are held together through direct collar-collar interactions, forming a concave monolayer of cells with aligned apico-basal polarity (Figure 1B-D) (35,36).
C. flexa sheets display collective contractility under the control of environmental cues (including light-to-dark transitions), inverting their curvature and switching from a feeding state (relaxed form, flagella-in; Figure 1B) to a swimming state (contracted form, flagella-out; Figure 1D) (35,37).In an earlier study, we established stable cultures of C. flexa sheets from a single isolated cell, indicating that sheets can arise from individual cells (35).Nevertheless, the mechanisms of C. flexa multicellular development remained unexplored.
(Figure 1 legend on the next page) are the sister group to animals (Metazoa).The phylogenetic relationships depicted are based on several recent phylogenomic studies (3,25,27,35,(38)(39)(40)(41)  colonies, we monitored small-and medium-sized colonies (between 6 and 46 cells) by timelapse microscopy and regularly observed clonal expansion by cell division both at the core and periphery of the sheets (Figures 1G and S1; movies S2-S3).However, unexpectedly, we also captured instances of free-swimming single cells and doublets meeting colonies, attaching to their periphery, re-orienting their apico-basal polarity to align with neighboring cells, and seemingly integrating into the sheet (Figures 1G and S1; movies S2-S3).Taken together, these results show that C. flexa colonies can form clonally but can also expand by aggregation.This motivated us to test whether C. flexa could also acquire multicellularity purely by aggregation (Figure 1H).

Aggregation is sufficient for multicellular development in C. flexa
To test whether C. flexa colonies can form by aggregation, we mechanically disassembled colonies into free-swimming single cells and performed live imaging of the dissociated cells (Figure 2A; movie S4).We found that C. flexa single cells aggregated within minutes into irregular masses of cells that continued incorporating additional cells and underwent repeated fusion into larger groups (a process known as agglomeration (42); Figure 2A; movie S4).To determine the fate of these aggregates, we labeled two populations of dissociated single cells with different fluorophores.Mixing both populations was followed by aggregation of cells of both colors within minutes (Figure 2B; movie S5).After 24 hours, we observed chimeric colonies displaying the polarized cup-shaped morphology typical of canonical C. flexa sheets and comprising cells of both colors (Figures 2C and S2A; movies S6-S8), indicating that they had developed by aggregation.To better understand the dynamics of aggregative development, we fixed aggregates at different stages and imaged them by Airyscan confocal microscopy.We found that early aggregates (fixed up to 6 hours post-dissociation (6 hpd)) displayed a variable and irregular morphology with cells frequently showing unaligned apicobasal polarity, likely resulting from collision and attachment of cells in diverse random orientations (Figures 2D and S2B; movies S9-S12).Colonies observed 24 hpd, on the other hand, consistently showed bona fide C. flexa sheet morphology with regularly aligned cells (Figures 2D and S2B; movie S13), suggesting that early irregular aggregates had by then completed their maturation into canonical sheets.At that stage, colonies comprised about 50 cells on average and as many as 120 cells (Figure 2E).This size further supported the idea that these sheets could not have formed exclusively by clonal development within 24 hours, given the cell cycle duration in C. flexa (at least 8 hours, which could result in sheets comprising at most 16 cells assuming maximal and synchronized proliferation; see Figure 1F).Indeed, treatment with the cell cycle inhibitor aphidicolin abolished cell proliferation (Figure S3) but did not prevent colony formation from dissociated cells (Figure 2F), confirming that aggregation was sufficient for C. flexa multicellular development.
Our observation of aggregative development in C. flexa contrasts with earlier observations in other species of multicellular choanoflagellates, which had so far been found to develop strictly clonally (3,23,24).Given that other protists often deploy aggregative development as an "emergency response" to fluctuating environments (4,14), while clonal development is thought to more frequently result in stable structures capable of feeding and cell proliferation, we were curious about the natural conditions of C. flexa mixed clonal-aggregative development.Thus, we set out to investigate the natural ecological context of C. flexa multicellular development.

Multicellular C. flexa sheets occur in a restricted salinity range in their natural environment
C. flexa was originally discovered in its multicellular form on the Caribbean island of Curaçao (35), a semi-arid tropical island that experiences dry and rainy seasons (Figure 3A) (43).Unlike other multicellular choanoflagellate species (20,44), C. flexa has been repeatedly re-isolated in the wild since its discovery in 2018 and can thus be studied in its natural environment.C. flexa sheets are found in splash pools on the wind-exposed northern part of the island (Figures 3A-B and S4A-B) that undergo natural cycles of desiccation and refilling (45,46).Water-filled splash pools experience gradual evaporation, leading to increasing salinity and, occasionally, complete desiccation (Figure 3C).Splash pools refill from crashing waves, splash, or rain, restoring lower salinity levels (Figures 3C and S4B; movie S14).
Splash pools are, consequently, ephemeral habitats in which organisms often experience extreme and recurrent hypersaline and hyperosmotic stress (46)(47)(48)(49).Thus, we investigated how this highly fluctuating environment might influence the life history and multicellular development of C. flexa.
We focused our studies on Shete Boka (meaning "seven bays" in Papiamento) National Park, where C. flexa is reliably found (Figure 3A).We surveyed the distribution of C. flexa in 150 splash pools and measured splash pool salinity in two different field expeditions: Exped-A and Exped-B (Figure 3D).During Exped-A, we randomly sampled 79 splash pools (numbered Sp1 to Sp79) along about 2 kilometers of coastline in Shete Boka (Figure 3D; see methods).While seawater collected from the neighboring inlets or bokas had a stable salinity of 40 parts per thousand (ppt), splash pool salinity ranged from below average seawater salinity (25 ppt) to saturation (≥ 280 ppt) (Figures 3E and S4C).
In 10 out of 79 splash pools, we found sheets that exhibited the stereotypical C. flexa cupshaped morphology and inversion behavior in response to light-to-dark transitions (35) (Figures 3F and S5A-C; movie S15).18S ribosomal DNA sequencing of manually isolated sheets confirmed their identification as C. flexa (Figure S5D; supplementary files S1-S4).
Interestingly, although the salinity of surveyed splash pools ranged from 0.6-fold to 7-fold of that measured in seawater from the neighboring bokas, C. flexa sheets were only found in splash pools with <2-fold seawater salinity (<73 ppt; Figures 3E and S4C).To independently test whether the presence of multicellular C. flexa was constrained by an upper bound of salinity, we employed a different sampling strategy during a second expedition (Exped-B) in which we exhaustively sampled all splash pools within a 4-meter by 10-meter quadrant (n=71, numbered Sp1 to Sp71; Figure 3D; see methods).We found sheets in 14 out of 71 splash pools, with a 94 ppt salinity upper bound for sheet occurrence (Figures 3E and S4C).Across both expeditions, C. flexa sheets were found in splash pools with an average salinity of 61.42 ± 17.54 ppt, 1.56-fold higher than natural seawater from the bokas (Figures 3E and S4C) and significantly lower than the salinity of sampled splash pools that did not contain sheets (155.1 ± 107.2 ppt; p=3.1e-04 by the Mann-Whitney U test).We never observed sheets in splash pool water with a salinity above 94 ppt (2.35-fold seawater salinity; Figures 3E and S4C).(E) Distribution of salinity of splash pools surveyed in Exped-A (circles) and Exped-B (triangles).For daily monitored splash pools, measurements from the first day of monitoring are shown.Magenta and turquoise colors indicate that sheets were respectively found and not found in splash pool samples.The observed natural limit of salinity where sheets were found (94 ppt) is indicated by an orange dotted line, and the average seawater salinity measured in the bokas with a gray dotted line.Gray area: salinity saturation (>280 ppt).(F) Brightfield image of a sheet observed in a splash pool sample.(G) Representative images of a splash pool near Boka Kalki (Sp15) followed for eight days, showing recorded salinity in the upper left.This splash pool was completely desiccated at day 5 (right).Dashed line: splash pool outline.(H) Salinity (upper panel) and depth (lower panel) measurements in three representative splash pools followed over eigth days.Shown are one splash pool that experienced evaporation but not complete desiccation (Sp50), one splash pool that experienced complete desiccation (Sp15), and one splash pool that experienced both desiccation and refilling (Sp69).Filled and empty circles indicate that sheets were respectively found and not found in splash pool samples.Gray area: salinity saturation.Dashed lines: dry periods.Orange circles: refilling events.
(I) Recovery of sheets from soil samples collected from a splash pool (Sp12) during its dry period.Soil samples were collected every day for eight days and were independently rehydrated in the laboratory with filtered seawater from the bokas, reconstituting salinity back to 42 ppt.Each rehydrated soil sample was monitored over five days for sheet re-appearance.

Transitions into and out of multicellularity correlate with the natural evaporation-refilling cycle of splash pools
We then set out to study the natural evaporation-refilling cycle of splash pools and how it may impact the presence/absence of sheets.We monitored ten splash pools in which C. flexa sheets were found and five more randomly selected splash pools daily over eight days (Figures 3G-H and S6).We screened samples for sheets once per day and measured the salinity and maximum depth of each splash pool (Figures 3G-H and S6).We observed a gradual decrease in depth and concomitant increase in salinity in all 15 splash pools (presumably due to evaporation; Figure S6A-B), six events of complete desiccation (Figure S6C-D), and four events of refilling (Figure S6E-F; see three examples in Figure 3H).In all cases, C. flexa sheets were no longer observed after salinity crossed a ~100 ppt threshold during gradual evaporation (Figures 3H and S6), consistent with results from Exped-A and Exped-B ( Figures 3E-H and S6).Moreover, artificial evaporation of a natural splash pool sample containing sheets in the laboratory led to sheet disappearance (Figure S7).These observations further reinforced the idea that the multicellular form of C. flexa might not tolerate high salinity levels.
Interestingly, in two dry splash pools that underwent refilling during our study (restoring salinity down to ~50 ppt), sheets were re-observed 48 hours after refilling (Sp69 and Sp70; Figures 3H and S6E-F).The new incoming sheets may have arrived from waves or splash, or may have been transferred from neighboring splash pools.Alternatively, C. flexa might have persisted in the soil of desiccated splash pools in a hitherto unrecognized resistant form, perhaps in the form of unicellular cysts (which have been morphologically reported in other choanoflagellates (50)(51)(52)(53)(54)(55)).Under this hypothesis, colonies would disassemble and transition into resistant single cells under high salinity, and those cells would then develop back into sheets once salinity decreased.To explore this possibility in situ, we collected soil samples from six desiccated splash pools in which C. flexa had been previously observed, rehydrated them in the lab (down to ~65 ppt salinity) and monitored the rehydrated samples for several days (Figure 3I-J; movie S16).Surprisingly, sheets could be recovered in various soil samples collected from one splash pool after two to seven days of desiccation (Figure 3I-J).
Sheets consistently appeared 48-72 hours post-rehydration, suggesting that a resistant form of C. flexa can survive complete desiccation in the wild for at least a week (Figure 3I-J).
Interestingly, the presence of choanoflagellates in soil might be a more general phenomenon: metagenomic analyses of soil samples from diverse environments have revealed the presence of choanoflagellate species yet to be studied (56,57), which might represent a significant and poorly understood ecological niche for choanoflagellates (58).
Given that salinity fluctuations in splash pools seemed to dramatically impact C. flexa multicellularity in nature, we next set out to investigate the phenotypic response of C. flexa to evaporation-refilling cycles in a laboratory context.

Experimental evaporation-refilling cycles cause reversible transitions into and out of multicellularity
To better understand the response of C. flexa to evaporation, desiccation, and refilling, we designed an experimental setup mimicking the natural evaporation-refilling cycles of Shete Boka splash pools (Figures 4A and S8).As a first step, we calculated the empirical evaporation rate of 12 splash pools in our longitudinal field survey (see methods; Figure S8A-B).We then subjected C. flexa cultures to evaporation in an incubator (Figure S8C), matching both the temperature (30°C) and evaporation rate to that observed in natural splash pools (see methods; Figure S8D).This setup resulted in complete desiccation after four days of evaporation (starting from artificial seawater (ASW) with a salinity of 35 ppt, referred to as 1X salinity in the rest of the text).Under these conditions, C. flexa sheets gradually dissociated into non-motile single cells (Figures 4A-B and S9; movie S17), with more than 50% of the cells being solitary when salinity crossed the limit of sheet occurrence observed in the field 14 (94 ppt; Figures 3E and 4A).Almost all sheets had dissociated into single cells when salinity reached saturation (Figures 4A-B and S9; movie S17).As a negative control, we monitored identically treated cultures that were not undergoing evaporation and observed that colonies remained multicellular throughout the experiment (Figure S10).This suggested that sheet dissociation in the gradual evaporation setup was caused by the increasing salinity.To independently test this hypothesis, we raised the salinity of C. flexa cultures by directly adding seawater salt (without evaporation) and similarly observed dissociation of multicellular sheets into single cells (Figure S11A-C).
To test whether single cells resulting from sheet dissociation in the lab were viable and could survive desiccation, we mimicked splash pool refilling by rehydrating gradually evaporated samples with artificial seawater three hours after full desiccation.This brought salinity back down to ~50 ppt and was followed by the reappearance of small sheets as soon as 24 hours post-rehydration.The fraction of cells in colonies increased gradually until ~50% of cells were colonial on day 7 and ~75% on day 9 (Figure 4A-B).Taken together, our results show that C. flexa sheets dissociate into non-motile single cells during gradual evaporation, that these solitary cells can survive complete desiccation under laboratory conditions, and that sheets reform after rehydration through both clonal division and aggregation.These findings are consistent with our field observations, where we recovered sheets after rehydrating soil samples from a dry splash pool (Figure 3I-J

C. flexa differentiates into solitary cyst-like cells at high salinity
We then set out to further investigate the phenotype of the desiccation-induced single cells.In diverse protists, resistance to desiccation is often achieved by differentiation into cysts (59)(60)(61), which are dormant cells with reduced metabolic activity and minimal or arrested proliferation.This process, known as encystment, often entails significant morphological changes such as rounding of the cell body, changes in cell volume, flagellar loss, and formation of a protective cell wall (53,59,(62)(63)(64)(65)(66)(67)(68).First, we determined whether C. flexa desiccation-resistant cells (hereafter 'cyst-like cells') actively proliferate by quantifying cell growth during gradual evaporation in the laboratory.We found that growth was arrested above 2X salinity, with net cell loss above 3X salinity (Figures 4D and S12).These effects may be attributable to a halt in the cell cycle and cell death due to hypersaline stress, similar to processes observed during encystment in other protists (13,59,62,68,69).
We next monitored cellular morphology by DIC microscopy during gradual evaporation and observed structural changes that occurred asynchronously, resulting in a heterogeneous cell population (Figure 4E).At 3X salinity, most cells had dissociated from their colonies and bore a round cell body lacking microvilli and, most often, lacking a flagellum (Figure 4Eii).
Additionally, some cells exhibited multiple filopodia-like protrusions (Figure 4Eiii).Membrane and F-actin staining of cyst-like cells confirmed that they lacked a collar complex (Figures 4F     and S13) but also revealed that they transiently formed an F-actin cortex, detectable at 3X salinity but lost above 6X (close to saturation; Figures 4F-G and S13).This transient actin cortex might contribute to protecting the plasma membrane against osmotic stress during the initial differentiation stages (70)(71)(72)(73).
Finally, morphometric analysis of cell body and nucleus volume showed that cyst-like cells had a larger nucleus-to-cytoplasm ratio compared to flagellate cells cultured in 1X salinity (p=3.4e-05;Figures 4H and S14).An increase in the nucleus-to-cytoplasm ratio frequently correlates with cell quiescence in other systems (74), consistent with the growth arrest in C. flexa cyst-like cells (Figures 4D and S12).To confirm that these changes were induced by hypersaline stress, we directly increased the salinity of C. flexa cultures by addition of seawater salt, and observed similar morphological changes and arrest in cell growth (Figure S11).
In sum, hypersalinity caused drastic and reversible changes in cell architecture and growth: loss of the collar and flagella (and thus of motility), formation of a transient actin cortex, increased nucleus-to-cytoplasm ratio, and arrest of cell growth (Figure 4I), resembling the changes that occur during encystment in diverse protists (59,62).Thus, during the evaporation-refilling cycle of a splash pool, C. flexa likely alternates between a multicellular flagellated form (at low salinity) and a unicellular cyst-like form (at high salinity), suggesting these two phenotypes benefit from an adaptive advantage in their respective environment.
We thus decided to directly test for these putative advantages in laboratory conditions.

Multicellular sheets are advantaged at low salinity and unicellular cyst-like cells are advantaged at high salinity
If C. flexa cyst-like cells represent a desiccation-resistant form, we could expect these cells to be more resistant to hypersaline stress and desiccation compared to sheets.To test this, we compared the survival of both sheets and cyst-like cells after desiccation.We subjected sheets to fast evaporation, which resulted in complete desiccation from 1X salinity over 20 hours (see methods).Under these conditions, cells did not acquire a cyst-like morphology but instead retained an observable flagellum, collar, and multicellular morphology even after complete desiccation (Figure S15).This suggests that the formation of cyst-like cells requires gradual evaporation at a comparable rate to that of natural splash pools.In parallel, we induced the differentiation of C. flexa into cyst-like cells by slow evaporation following the protocol detailed in the previous section (from 1X salinity to desiccation in 72 hours).After desiccation, we rehydrated both types of cells by adding 1X ASW and monitored their recovery.We found that desiccated sheets having undergone rapid evaporation never developed into viable cells after rehydration (Figure 5A).By contrast, as in the experiments detailed earlier, rehydrated cyst-like cells consistently gave rise to viable sheets, confirming that they had survived desiccation (Figure 5A).These observations show that cyst-like cells, unlike flagellates, are equipped to survive hypersalinity and desiccation, and suggest that differentiation into cyst-like cells confers a selective advantage during evaporation.The loss of multicellularity during differentiation into cyst-like cells is likely linked to the retraction of the microvillous collar that connects cells within sheet colonies.Because unicellular cyst-like cells seemed to have a survival advantage over multicellular flagellates in the hypersaline phase of the evaporation-refilling cycle, we wondered whether multicellularity, by contrast, was advantageous in the other phase of the cycle, marked by low salinity.It has been speculated that multicellularity in choanoflagellates might enhance feeding via cooperative hydrodynamic interactions between flagella, increasing the flux of bacterial prey towards the collar (75-77); however, this concept remains uncertain in the model choanoflagellate Salpingoeca rosetta, with different studies having come to contrasting conclusions (75,78,79).To test for a feeding advantage in multicellular sheets of C. flexa, we quantified the capture of fluorescent bacteria in sheets and dissociated single flagellates (Figure 5B-C).We found that sheets captured more than twice as many fluorescent bacteria per cell as single cells (Figure 5B-C).This suggests that multicellularity confers a prey capture advantage at the salinity levels that are compatible with cells maintaining a functional collar complex and, therefore, with feeding.
Taken together, our findings support a model of C. flexa life history that correlates with the evaporation-refilling cycle of splash pools characteristic of their natural habitat (Figure 5D).
C. flexa sheets are found in water-filled splash pools with salinities below ~2.3-fold that of natural seawater.As gradual evaporation proceeds, salinity increases, and sheets dissociate and differentiate into solitary, non-motile, and non-proliferative cyst-like cells (some of which undergo cell death).These cyst-like cells can survive complete desiccation and persist in the soil of the splash pools.After splash pool refilling, cyst-like cells regrow a collar complex and transition back into free-swimming flagellates that develop into multicellular sheets by aggregation and/or clonal division, a phenotype that more efficiently captures bacteria and presumably supports faster cell proliferation.The selective advantage of multicellularity has been the subject of intense debates (76,(80)(81)(82) and multiple studies in choanoflagellates (75,76,(83)(84)(85)(86)(87).In the case of C. flexa, the prey capture efficiency of cells within sheets surpasses that of solitary flagellated cells, suggesting a feeding (and likely reproductive) benefit of multicellularity.Cooperative feedinga trait observed among various protists-has been postulated as a selective factor for the evolution of multicellularity (88).However, other advantages of multicellularity might exist in C. flexa.Prominently, C. flexa colonies exhibit collective behavior through coordinated contractility that implements whole-colony inversion.Inversion allows reversible transitions between feeding and swimming states and is modulated by several signals, including mechanical stimuli, heat shocks, and nitric oxide (37).This switch is also regulated by lightto-dark transitions, thus enabling photokinesis (preferential accumulation of sheets under light) (35).Our previous research indicates that individual cells from dissociated sheets are no longer capable of photokinesis, suggesting this behavior requires multicellularity (35).

Discussion
Moreover, multicellular groups show decreased vulnerability to size-selective predators in different protists and predator avoidance has been proposed as a potential selective force for evolving multicellularity (6,15,76,81) (but see (86)).
More generally, the environmentally entrained life history transitions of C. flexa are in line with other recently described examples.Unicellular-to-multicellular switches are controlled by fluctuating environmental parameters in other species, such as salinity for cyanobacteria in brackish environments (89) and periodic flooding for cave bacteria attached to surfaces (90).A selective advantage for regulated life cycles in fluctuating environments has also been supported by laboratory experiments in yeast (91) and by theoretical models (5,92,93).The phenotypic plasticity of C. flexa might thus enable it to reap both the benefits of multicellularity (for feeding and collective behavior) under permissive conditions, and of differentiation into solitary cyst-like cells for individual survival under harsh conditions.
Beyond choanoflagellates, both clonal and aggregative multicellularity are widely present, with a discontinuous occurrence, across unicellular holozoans (99-104, 28, 42, 27, 29, 30) and eukaryotes as a whole (3,(105)(106)(107).The scattered distribution of clonal and aggregative multicellularity is currently interpreted as reflecting independent evolutionary origins of both types of development (4)..However, the discovery of clonal-aggregative multicellularity in a eukaryote raises the intriguing alternative possibility of evolutionary interconversions between aggregative and clonal multicellular development (via a mixed intermediate).Future studies of clades that contain both clonal and aggregative forms (such as Holozoa and Holomycota) might help test this possibility.Although mixed clonalaggregative multicellularity has not been formally reported before in other protists to our knowledge, some parallels can be drawn with processes described in other taxa or contexts: for example, clonality and aggregation might cooperate in the formation of bacterial boofilms (108) and in the development of (unicellular but multinucleated) syncytia in reticulopodial amobae (1).This raises the possibility that mixed clonal-aggregative development might be more widespread than currently appreciated.
Interestingly, the phenotype of C. flexa colonies differs in some important respects from that of most previously described aggregative structures.The paradigmatic model of aggregative multicellularity is fruiting body formation (see glossary), which has been wellcharacterized in dictyostelid amoebae (4,14,109) and evolved convergently multiple times across eukaryotes (110,111) (and even at least once in bacteria (112)).Fruiting bodies are masses of dormant cells (sometimes together with cells having undergone cell death) that disperse into single cells before resuming metabolic activity and proliferation upon the restoration of favorable environmental conditions.By contrast, C. flexa sheets are proliferatively and behaviorally active, and appear to survive and reproduce in a multicellular form as long as permissive conditions persist.Thus, the ecological niche accessible to mixed clonal-aggregative multicellularity might differ from that typical of purely aggregative forms.
Aggregative multicellularity comes with a well-known evolutionary challenge: one aggregate can combine cells of different ancestries and potentially different genotypes, thus raising the possibility of conflict (113)(114)(115).Aggregates are notably vulnerable to cheater genotypes, which reap the benefits of multicellularity without contributing their fair share of the cost of multicellular development.Such cheaters have, for example, been described in Dictyostelium (116), in the fruiting body-forming bacterium Myxococcus (117), and in flocculating yeast (10).The evolutionary challenge posed by cheater mutants can be addressed by strategies that limit aggregation to fellow co-operators, either directly by cooperator recognition (so-called "green beard" mechanisms (118,119)) or indirectly by preferential aggregation with close relatives.Restriction of aggregation to relatives can either be enforced actively by kin recognition (mediated by polymorphic loci (120)(121)(122)) or be facilitated passively by spatial structuration of the environment that can limit dispersal (123,124).The latter might be relevant to C. flexa, as splash pools are collections of disconnected environments that might lead to geographic divergence of genotypes (as shown for other organisms in similar environments, such as Daphnia metapopulations in tide pools (125)).Splash pools might thus favor the evolution of aggregative multicellularity by mitigating the need for strong kin recognition mechanisms.This is in line with recent models supporting an "ecological scaffolding" function for patchy environments in the emergence of multicellularity (126).In the future, studies of potential kin recognition mechanisms, natural genetic diversity and dispersal in C. flexa might shed further light on these questions.

Conclusion
Our study documents an unusual mode of multicellular development that combines features of clonal and aggregative multicellularity in a choanoflagellate, belonging to the sister-lineage of animals.This complex life cycle further underscores the emerging concept of considerable phenotypic and developmental plasticity among close unicellular relatives of animals, supporting the existence of a complex life cycle in unicellular ancestors of animals (19,127).
Beyond this, our study establishes C. flexa as a model to study the facultative multicellular development of a close relative of animals in its natural context.This contrasts with other well-characterized facultatively multicellular holozoans, such as S. rosetta (20) and C. owczarzaki (128), which could both be isolated only once from their natural environment and in which studies of unicellular-to-multicellular transitions are thus inevitably restricted to laboratory setups.In the future, we expect the dialogue between field and lab studies of C. flexa to continue, clarifying questions such as the selective advantage(s) and ecological implications of multicellularity and unicellularity, as well as the existence of possible natural kin recognition mechanisms and cheater mutants.

Box 1. Glossary
Cheaters: in social evolution theory, organisms that receive a benefit at the cost of other organisms of the same species within the context of a collective behavior.In the case of aggregative multicellularity, cheaters are cells within a chimeric multicellular aggregate that impair collective outcomes to their own benefit.For example, in sorocarp-forming eukaryotes, cheaters differentiate disproportionately into spores (reproductive cells) rather than other cell types (such as stalk cells), giving them a higher likelihood of reproducing than non-cheaters.
Cheaters can exploit shared resources in a way that enhances their own fitness at the expense of the collective.The existence of cheaters may impede group-level selection of traits and result in evolutionary conflict (116,117).
Fruiting body: a spherical mass of stress-resistant cells (spores or cysts), often in a quiescent state, that eventually undergoes disassembly and dispersal (129).Fruiting bodies often result from aggregative development and are sometimes equipped with a stalk providing elevation from the substrate, and facilitating dispersal by the wind after dissociation.
Holozoa: eukaryotic clade encompassing all species more closely related to animals than to fungi.Holozoa comprises various clades, including Metazoa (animals), Choanoflagellata, Filasterea, Corallochytrea/Pluriformea, and Ichthyosporea.Unicellular eukaryotes (a.k.a.protists) within the Holozoa clade are pivotal for elucidating the evolutionary transitions that facilitated the emergence of multicellularity in animals from unicellular ancestors (27,130,38).Splash pools: shallow seawater pools in rocky shores inhabited by prokaryotic and eukaryotic organisms that are occasionally filled with seawater by oceanic waves and splash (and sometimes with rainwater).In Curaçao, the substrate of splash pools is composed mainly of limestone.Their sizes range from a few centimeters to a few meters.In splash pools, environmental conditions such as seawater volume, temperature, salinity, and oxygen fluctuate as they undergo natural evaporation-refilling cycles over a few days or weeks.Thus, splash pools are considered extreme environments.
modes of multicellular development.We show that C. flexa clonal-aggregative development occurs in the context of a life cycle entrained by salinity fluctuations encountered in their natural habitat: splash pools that undergo cycles of evaporation and refilling on the timescale of a few days.We propose that this extreme, fast-fluctuating environment might have contributed to selecting for this unusual mode of multicellularity.Our observations establish C. flexa as a new model for the ecological context and developmental mechanisms of facultative multicellularity across close relatives of animals and expand the range of possible scenarios for the emergence of animal multicellularity.

.
Uncertain positions are represented with polytomies.(B) Brightfield image of a C. flexa multicellular colony ("sheet") in its relaxed conformation (B').B", white square: zoom-in showing flagella (magenta pseudocolor) and direct cell-cell contacts between collars (green pseudocolor).Scale bar in B'': 10 µm.(C) Diagnostic morphological features of a choanoflagellate cell.C. flexa cells within a sheet are linked by their collars (orange arrow).(D) 3D reconstruction of an Airyscan Z-stack of a fixed sheet exhibiting an inverted conformation (D'), with cell bodies stained with a membrane/cytoplasmic dye (FM TM 1-43FX, magenta, which distributes to the membrane and cytoplasm of cells following fixation), and microvilli stained with a filamentous actin (F-actin) dye (phalloidin Alexa 488, green).(E) Stills from a brightfield timelapse movie of clonal C. flexa sheet development by serial cell division from a single C. flexa swimmer cell (white arrowhead).After each division, the sister cells remain adhered to each other by direct cell-cell contacts between collars (black arrow).Note that cells retract their flagellum during division.Time scale hh:mm.(F) Cell lineage tracing as a function of time in E shows that cells divide asynchronously during colony formation, taking ~8-10 hours between each round of division.(G) Stills from a brightfield timelapse movie depicting a medium-sized C. flexa sheet (flagella-in conformation) that expands in cell number both by cell division (pseudocoloring in orange, pink, blue and yellow) and by cellular aggregation (white arrow, pseudocoloring in green).Time scale hh:mm.(H) Schematics of C. flexa mixed clonal-aggregative multicellularity development observed under laboratory culture conditions: (1)unicellular flagellate (swimmer) cells can divide clonally to initiate a colony; in turn, colonies can increase in cell number by(2) clonal division and (3) cellular aggregation (green arrowhead).In all cases, sister cells resulting from cell division or independent cells attaching to colonies by cellular aggregation adhere to each other through direct cell-cell contacts in their collar (orange arrowhead).The hypothesis that C. flexa sheets might be able to develop purely by aggregation is tested in Figure2.Figurerelated to Figure S1 and movies S1-S3.
To understand C. flexa colony development, we first isolated and monitored single cells from mechanically disassembled sheets by time-lapse microscopy.In this context, we observed clonal development of sheets by serial cell division (Figure 1E-F; movie S1).After cell division, sister cells remained attached to each other by intermicrovillar contact.Cells continued to divide asynchronously (every ~8-10 hours), resulting in a monolayer of polarized cells with the signature curved morphology of C. flexa sheets (Figure 1E-F; movie S1).These observations show that small C. flexa sheets (up to 7 cells) can develop clonally from a single cell.To test whether clonal development can contribute to the further growth of C. flexa

Figure 2 .
Figure 2. Multicellularity in C. flexa can be established purely by aggregation.(A) Stills from a brightfield timelapse movie showing that dissociated single cells (white arrowheads) quickly re-formed colonies by cellular aggregation.Note that two colonies can fuse together by agglomeration (black arrowheads).(A') Binary image mask of stills in A. Time scale hh:mm.(B) Stills from a timelapse movie showing that two dissociated single cell populations labelled with either Cell Trace CFSE (green) or Cell Trace Far Red (magenta) aggregated into dual-labelled chimeric groups of cells (white arrowheads).Time scale hh:mm.B', white square: zoom-in showing a dual-labelled colony.Scale bar: 10 µm.(C) (Upper panels) Dissociated single flagellate cells labelled with Cell Trace CFSE (left, green) or Cell Trace Far Red (middle, magenta) form single-labelled colonies 24 hours postdissociation (hpd).When both single celled populations are mixed in a 1:1 ratio, they form dual-labelled colonies (right).(Lower panels) 3D reconstructions of Airyscan Z-stacks of sheets formed by aggregation from dissociated single cells labelled as depicted above and fixed 24 hpd, with additional filamentous actin (F-actin) staining (phalloidin Alexa 405, blue).(D) 3D reconstructions of Airyscan microscopy images of single cells fixed after 10 minutes (n=12), 30 minutes (n=12), 2 hours (n=19), 6 hours (n=20) and 24 hours (n=16) post-dissociation stained for membrane/cytoplasm (FM TM 4-64FX, magenta) and F-actin (phalloidin Alexa 488, green).Note that cells frequently show unaligned apico-basal polarity and diverse cell orientations at early timepoints (white arrowheads).Time scale hh:mm.(E) Quantification of cell number per colony during aggregation (from the experiment in D) in two independent biological replicates.Black circles: mean.Error bars: standard deviation.Diamonds: mean values of independent biological replicates.(F) Quantification of particle area during an aggregation time course of dissociated single cells pre-treated with 17 µg/mL aphidicolin overnight in three independent biological replicates.DMSO: negative control.

Figure 3 .
Figure 3. Cyclical salinity fluctuations constrain the occurrence of multicellular C. flexa sheets in their natural environment.(A) Map of Curaçao and location of Shete Boka National Park (turquoise star) where fieldwork data was collected in Exped-A and Exped-B expeditions (12° 22' 5.718" N, 69° 06' 56.916" W). (B) Representative photograph of the landscape in Shete Boka, including splash pools where C. flexa sheets can be found (white arrowheads).(C) Schematics of a splash pool natural cycle of seawater evaporation, desiccation, and refilling.(D) Maps showing the locations of sampled splash pools in Exped-A and in Exped-B expeditions.In Exped-A (D), samples were collected from splash pools along ~2 km of Shete Boka coastline (n=79).15 of these splash pools were randomly selected for daily monitoring during eight days (stars) (shown are results from the first day of monitoring).Colors indicate whether sheets were found (magenta) or not found (turquoise) upon microscopic inspection.In Exped-B (D'), a random number generator was used to select the randomized sampling location (purple pin, 204 m upstream of Boka Wandomi).Samples were collected from splash pools within an area of 10 m by 4 m (n=71).
Filled circle: sheets observed.Empty circle: sheets not observed.Dashed empty circle: sample not inspected.rh: rehydration.(J) Brightfield images of soil-recovered sheets (same experiment as in I) collected after 3 (left) and 6 (right) days of desiccation and rehydrated in the laboratory (lower left: time post-rehydration).Figure related to Figures S4-S7, movies S14-S16, Tables S1-S2 and Supplementary Files S1-S4.

Figure 4 .
Figure 4. Experimental evaporation-refilling cycles causes reversible transitions between multicellular sheets and unicellular cyst-like cells.(A) Quantification of salinity (upper panel) and fraction of cells in colonies (magenta) or in single cells (turquoise) (lower panel) during a 9-day gradual evaporation time course.Dark gray rectangle: complete desiccation (dry period).Orange dotted line: observed natural salinity limit of sheet occurrence.Gray area: salinity saturation.Lines correspond to mean values and shadowed area to standard deviation.Experiment performed in two independent biological replicates.(B) Stills of C. flexa sheets during the same 9-day gradual evaporation timecourse in A, showing salinity (in ppt and fold-change compared to that of 1X ASW in parenthesis, upper left) and time (upper right).Sheets had completely dissociated into single cells (white arrowheads) after 80 hours of gradual evaporation.Multicellular sheets were observed again after complete desiccation and rehydration with 1X ASW.(C) Stills from a brightfield timelapse movie of cells rehydrated after complete desiccation.Recovered flagellate cells re-form colonies both by clonal division and aggregation 30 hours after rehydration.(D) Growth rate of cells at different salinities during gradual evaporation in three independent biological replicates.Black circles: mean.Error bars: standard deviation.n=28, p by the Mann-Whitney U test.Diamonds: mean values of independent biological replicates.(E) Brightfield images of C. flexa showing morphological changes during gradual evaporation.At ~1X seawater salinity (left), C. flexa occurs in the form of multicellular sheets of flagellated cells (i).Green pseudocolor: collar.Magenta pseudocolor: flagellum.During gradual evaporation (ii-iv, right panel), sheets dissociate into unicellular cyst-like cells.Cyst-like cells lack a collar, often lack a flagellum, and can exhibit filopodia-like protrusions (magenta arrowheads in iii-iv).(F) AiryScan micrographs of C. flexa cells fixed during gradual evaporation and stained with a membrane (FM TM 4-64FX, magenta) and F-actin (phalloidin Alexa 488, green) dyes, showing lack of collar and flagellum in cyst-like cells.A flagellate cell (F, low-evaporation control) exhibits distribution of F-actin distribution in the collar of microvilli (mv, black arrowhead) and filopodia (fp, black arrowhead).Gradual evaporation triggers a morphological change from a flagellate to a cyst-like cell, showing a transient actin cortex (cx) at early stages of evaporation (F', black arrow) which disappears as salinity approaches saturation (F'').(G) (resp.G', G") Line scan of F-actin fluorescence intensity along the dashed lines of interest in F (resp F', F"), showing cortical actin as two peaks where the lines intersect the cell cortex.(H) Quantification of the nucleus-to-cytoplasm ratio in cyst-like cells (n=27) and flagellate cells (n=38) in three independent biological replicates.Black circles: means.Error bars: standard deviations.Diamonds: mean values of independent biological replicates.p by the Mann-Whitney U test.(I) Schematic summarizing phenotypic changes experienced by C. flexa cells during gradual evaporation.At early stages of evaporation (turquoise), flagellate cells differentiate into cyst-like cells by losing their collar and flagellum, and show a temporal actin cortex.At later stages of evaporation (desiccation, magenta), cyst-like cells lack an actin cortex and show a higher nucleus-to-cytoplasm ratio.Rehydration (refilling, purple) induces a cyst-like cell-to-flagellate cell transition, where cells regenerate their collar and flagellum and re-form multicellular sheets.Figure related to Figures S8-S14 and movies S17-S18.
), and further support the existence of a desiccation-resistant form of C. flexa.

Figure 5 .
Figure 5. Unicellular cyst-like cells are advantaged at high salinity while multicellular sheets are advantaged at low salinity.(A) Quantification of cell survival after 12 hours of desiccation in multicellular sheets (n=6 technical replicates) to unicellular cyst-like cells (n=6 technical replicates).Diamonds: means.Error bars: standard deviations.p by the Mann-Whitney U test.(B) C. flexa multicellular sheets are more efficient at capturing bacteria than unicellular flagellates.Number of labelled bacteria captured per flagellate cells in multicellular colonies (n=9) or in cultures of unicellular flagellates (n=9) in three independent biological replicates.Black circles: means.Error bars: standard deviations.Diamonds: means of independent biological replicates.p by the Mann-Whitney U test.(C) DIC images of multicellular sheets (upper panel, white arrowheads) The ecological context of C. flexa clonal-aggregative multicellularityOur results show that C. flexa can form multicellular sheets through a combination of clonal division and aggregation following the rehydration of desiccated cyst-like cells.This mixed developmental strategy might represent an adaptation to the ephemeral nature of water-filled splash pools, allowing faster multicellular development by simultaneous action of both mechanisms.It might also serve as a versatile strategy for robust re-establishment of multicellularity across a broad range of environmental conditions: indeed, aggregative and clonal multicellularity are subjected to different constraints, as they respectively depend on cell density and on the possibility of cell division (which is in turn constrained by salinity and availability of bacterial prey).In our laboratory experiments, C. flexa appeared capable of both purely clonal development (at low cell density and under permissive conditions for proliferation) and purely aggregative development (at sufficient cell density, even if proliferation is compromised), as well as intermediate modes.