A model suite of green algae within the Scenedesmaceae for investigating contrasting desiccation tolerance and morphology

Microscopic green algae have long been known to inhabit microbiotic crusts in arid regions world-wide (e.g. Friedmann et al., 1967; Metting, 1981). Over the last two decades, researchers combining morphological and molecular information have demonstrated that these green algae are remarkably diverse (e.g. Lewis and Lewis, 2005), belonging to at least six different green algal classes (Cardon et al., 2008). Phylogenetic analyses further revealed that, even within families and genera (e.g. Lewis and Flechtner, 2002, 2004; Fucíková et al., 2013), multiple desert lineages have evolved from aquatic ancestors independently, resulting in suites of extant related green algal species that have desert-dwelling and aquatic representatives. These related species harbor rich, natural variation that has developed in the presence of selective pressures within aquatic and desert environments.

Lewis and Flechtner (2004) described three independently evolved lineages of desert-dwelling green algae that fell within the genus Scenedesmus: S. rotundus, S. deserticola and S. bajacalifornicus, that were isolated from the Sevilleta desert, San Nicolas Island (off the coast of California) and Baja California, respectively. These organisms are particularly intriguing because aquatic relatives within the Scenedesmaceae have been studied for decades by a number of researchers, who focused on genetic and physiological controls over cell division by multiple fission (e.g. Bišová and Zachleder, 2014), abiotic and biotic environmental drivers of phenotypic plasticity (e.g. Trainor et al., 1976; Lampert et al., 1994), the evolution of terrestriality within green plants (e.g. Lewis and Flechtner, 2004; Lewis and Lewis, 2005; Gray et al., 2007; Cardon et al., 2008) and, more recently, biofuels development (e.g. Ho et al., 2014) and bioremediation (e.g. Suresh Kumar et al., 2015). We are building on this broad foundation, studying the three desert-derived taxa described by Lewis and Flechtner (2004) together with two sister aquatic Scenedesmus Meyen sensu lato (s.l. Scenedesmoideae) (An et al., 1999; Lewis and Flechtner, 2004; Hegewald et al., 2010) as a new model suite of organisms. Although not yet genetically tractable, the common genetic background of these five species in the Scenedesmaceae, coupled with their ecological and morphological distinctiveness, make them a promising group for comparative analyses from the variety of biological perspectives noted above.

Evolutionary relationships among our five target organisms are firmly established (Hegewald et al., 2013; Sciuto et al., 2015; Wynne and Hallan, 2015). Desert-derived S. deserticola and S. bajacalifornicus (Lewis and Flechtner, 2004) – recently renamed Acutodesmus deserticola and Acutodesmus bajacalifornicus, respectively (Hegewald et al., 2013) – fall within the same genus as aquatic S. obliquus, which itself was recently renamed Tetradesmus obliquus (Wynne and Hallan, 2015; see Materials and Methods for discussion of taxonomic nomenclature in flux). Desert-derived S. rotundus – recently renamed Fletchneria rotunda (Sciuto et al., 2015) – and aquatic Scenedesmus costatus – recently renamed Enallax costatus (Hegewald et al., 2013) – are consistently more basal within the family Scenedesmaceae, no matter which gene region was used for phylogenetic analysis (Sciuto et al., 2015).

Here, we test several hypotheses regarding cellular and physiological traits in this target suite of five ecologically and morphologically contrasting desert and aquatic taxa. Lewis and Flechtner (2004) noted that A. bajacalifornicus, A. deserticola and F. rotunda are unicellular, but intracellular membrane characteristics and details of division have not been previously explored. T. obliquus and E. costatus are coenobial: each mother cell produces daughter cells that are permanently physically joined side by side in a sheet-like, multicellular coenobium. Because of their contrasting morphology, these five taxa provide intriguing potential as a system to study determinants of multicellularity within a common genetic background.

We first tested the hypothesis that all five taxa divide by multiple fission, as expected considering many other Scenedesmus sp. (Bišová and Zachleder, 2014). During multiple fission, a single mother cell commonly gives rise to more than two daughter cells during asexual reproduction. The nucleus and chloroplast in the mother cell divide into multiple copies, and cytokinesis compartmentalizes new daughter cells prior to release when the mother cell wall ruptures. Bišová and Zachleder (2014) reviewed a large body of literature focused on multiple fission in Chlamydomonas and Scenedesmus quadricauda, noting that, in Chlamydomonas, each of the multiple daughter cells is independent and unicellular. In S. quadricauda, multiple daughter cells are physically joined in a coenobium (although cytokinesis is complete). We examined whether the cells in our five target species divided by multiple fission by using overnight, time-lapse imaging of cell division and release of unicellular or coenobial daughter cells in gas-permeable polymethylsiloxane (PDMS) microfluidic chambers (Bascom et al., 2016). We then explored intracellular organellar and membrane patterns during division in more depth, using confocal microscopy for all five species and transmission electron microscopy for the three species F. rotunda, A. deserticola and E. costatus.

Although A. deserticola, A. bajacalifornicus and F. rotunda were isolated from desert microbiotic crust, DNA-based identification does not in itself demonstrate that the sequenced organisms are a permanent part of the crust community (i.e. adapted to desert conditions). Sampled organisms could be transient visitors that were, for example, deposited by chance by wind. To test for evidence of adaptation to desert conditions, Gray et al. (2007) assayed the resilience of photosynthetic activity to desiccation and rehydration in representatives from six independent desert green algal lineages and their close aquatic relatives. Among the desiccation-tolerant desert taxa analyzed by Gray et al. (2007) was F. rotunda (isolated by Lewis and Flechtner, 2004). After desiccation, photosynthesis in desert-derived algae rebounded more strongly and quickly upon rapid rehydration than it did in aquatic sister taxa. This rebound in photosynthetic capacity occurred in the vegetative stage of the desert-derived organisms; desiccation tolerance did not require production of a desiccation-tolerant spore that could germinate upon rehydration. The clear desiccation tolerance observed by Gray et al. (2007) supported the notion that the desert-derived algae tested by them, including F. rotunda, were physiologically adapted to tolerate desiccation that is common to desert habitats; they were not just coincidental, short-term visitors recently deposited on crust.

We tested the hypothesis that all three of our focal taxa isolated from deserts exhibit stronger desiccation tolerance than their two aquatic relatives – only F. rotunda had been tested previously. Because during desiccation and rehydration a significant source of damage can be the build-up of reactive oxygen species (ROS) within cells (Hoekstra et al., 2001; Oliver et al., 2004; Farrant and Moore, 2011), we further tested the hypothesis that the extent of desiccation tolerance across the five target taxa is mirrored by differential accumulation of cellular ROS during desiccation and rehydration.

We used confocal microscopy to examine whether the algae divided by multiple fission. Nuclei were stained with SYBR Safe; cell membranes were stained with the lipophilic dye FM4-64; and chloroplasts exhibited chlorophyll autofluorescence. We found that, in all species, the nucleus of each mother cell divided first into multiple nuclei, followed by chloroplast division and, finally, cytokinesis within the mother cell wall. Fig. 1 shows representative images for two stages of division. T. obliquus has undergone nuclear but not chloroplast division, and cell membranes have not yet formed around daughter cells. For the other species, Fig. 1 shows daughter cells held within the mother cell wall – having completed nuclear and chloroplast division, and cytokinesis – but not yet released from the mother cell wall. In all species, nuclear division occurred in the presence of one large chloroplast with an intact pyrenoid (e.g. T. obliquus, Fig. 1, pyrenoids indicated by white arrows; chlorophyll autofluorescence channel). The pyrenoid is a subcellular structure within the chloroplast where ribulose 1,5-bisphosphate carboxylase/oxygenase (RuBisCO) is concentrated; pyrenoids are commonly part of the CO₂-concentrating mechanism in green algae and appear electron dense in transmission electron microscopy (TEM). After nuclear division, the chloroplast divided, and the daughter cells were packaged by membranes delineating daughters (e.g. FM4-64 channel, all species except T. obliquus, Fig. 1). The pyrenoid reformed after chloroplast division and was visible in daughter cells still enclosed within the mother cell wall (Fig. 1, white arrows). Confocal z-stacks (Movies 1-5) and rendered volumes from them (Movies 6-10) reveal the intracellular 3D morphology.

We used TEM to examine details of intracellular structures during cell division in two desert and one aquatic species. Aquatic E. costatus was chosen because it is substantially larger than T. obliquus and, thus, easier to work with for TEM. Desert F. rotunda is more closely related to E. costatus; A. deserticola is more distantly related. In these three taxa, the chloroplast was easily recognizable, with internal thylakoid membranes (Fig. 2A, white arrows), starch granules and an electron-dense, prominent pyrenoid (Fig. 2A; see labels ‘s’ and ‘P’, respectively). We sampled cells late in the day when we expected division to be underway. Multi-nucleate cells were clear in A. deserticola and E. costatus (Fig. 2A,B,E,F; nuclei labeled ‘N’). At the time of sampling, F. rotunda had already completed cytokinesis but daughter cells were still confined within the mother cell wall (Fig. 2C,D). E. costatus and A. deserticola had undergone nuclear division but had not yet divided the chloroplast or performed cytokinesis (Fig. 2A,B and E,F, respectively). In all cases, Golgi stacks were closely associated with the nucleus (Fig. 2, blue arrowheads). The perinuclear position of these Golgi stacks was very similar to that observed with FM4-64 staining, which showed small, brightly stained membrane accumulation near the nuclei (Fig. 1, blue arrowheads). We also observed development of membrane separating daughter cells (Fig. 2E, yellow arrows), with nuclei sometimes visible and sometimes not depending on the plane of sectioning. Although the chloroplast has not divided on the right-hand side of the cell in Fig. 2E, two nuclei are visible, and membrane strands are beginning to develop between them (Fig. 2F, purple arrows). Observations such as these indicate that membrane deposition and organization began before division of chloroplasts was complete and, possibly, even before all nuclear divisions had occurred.

In all five species, we also used time-lapse imaging of chlorophyll autofluorescence to observe dynamics of division in living cells. In these taxa, a single chloroplast occupies a significant fraction of the volume of each cell. Therefore, chloroplast morphology and autofluorescence served to indicate the shape and location of the mother and then daughter cells in unstained, living organisms undergoing division in PDMS chambers overnight. Interestingly, just prior to or during cytokinesis, the cell mass within the mother cell wall (as indicated by chloroplast autofluorescence) rotated as a unit. These rotations are most easily discerned in the supplementary movies (Movies 11-15) but Fig. 3 presents selected movie frames. Yellow outlines on Fig. 3 track the location of a single daughter chloroplast (or two daughters, in the case of T. obliquus) within a group of daughters still inside a mother cell wall. In these frames, chloroplast division has occurred, but the frame is captured just prior to or during cytokinesis. The location of the yellow-outlined daughter cell shifted as the whole cell mass reoriented within the mother cell wall (Fig. 3, Movies 11-15).

Ultimately, when the mother cell wall finally split (not shown in Fig. 3), daughter cells in F. rotunda, A. deserticola and A. bajacalifornicus were released as identical, independent individuals (see Fig. S1, Movie 13 for an example in A. deserticola). In contrast, multiple daughter cells already joined in a coenobium were released from each mother cell in E. costatus and T. obliquus coenobia. For example, Fig. S1 shows a daughter coenobium still folded up within an intact mother cell wall in T. obliquus at 230 min. Fifteen minutes later, at 245 min, the mother cell wall had ruptured, and the daughter coenobium had been released and had unfolded into a coenobial sheet of four joined cells (Fig. S1, Movie 11). A similar series is shown for E. costatus (Fig. S1, Movie 12). In contrast, four daughter cells produced within an A. deserticola mother cell are released as individuals at time 780 min, and only one remains in the field of view at time 805 min (Fig. S1, Movie 13).

We used an assay of photosynthetic efficiency in the desiccated state and upon rehydration to test for desiccation tolerance in the five species. Samples from three replicate culture flasks were tested. Fig. 4 shows, for each species, the quantum yield of photosystem II (PSII) photochemistry (Φ_PSII) measured in the desiccated ‘dry’ state and at 3 min after rehydration. Φ_PSII, as defined by Genty et al. (1989), reflects the efficiency of electron transport per quantum absorbed by PSII. Φ_PSII is a widely used metric of photosynthetic activity and efficiency in the plant physiological and ecological literatures because it can be compared among samples (e.g. Genty et al., 1989; Maxwell and Johnson, 2000; Gray et al., 2007). All species had high Φ_PSII values (0.6–0.7) prior to desiccation. The Φ_PSII of aquatic T. obliquus was very low (<0.1) in both the dry and subsequent rehydrated states. E. costatus retained slightly higher Φ_PSII values immediately upon rehydration after desiccation, but values fell to Φ_PSII <0.1 within 3 min after rehydration and never recovered. Desert taxa had very low Φ_PSII when dry (tan bars, Fig. 4, A. deserticola, F. rotunda and A. bajacalifornicus) but, upon rehydration, Φ_PSII rebounded notably. The increase of Φ_PSII in rehydrated (relative to desiccated) algae was significantly larger in desert taxa (A. bajacalifornicus, F. rotunda and A. deserticola) than in aquatic taxa (E. costatus and T. obliquus) [n=3 flasks per species, tested by using generalized linear regression (GLM) with species nested within their native desert or aquatic habitat, P<0.05, Fig. 4].

Fig. S2 presents data from more-intensive testing of desiccation tolerance, from a fourth flask per species, during which algae were desiccated and rehydrated multiple times. Even after three cycles of desiccation and rehydration the Φ_PSII in desert species recovered upon rehydration (Fig. S2B). Fig. S2 also shows representative raw chlorophyll fluorescence data for each species that were used to quantify Φ_PSII during desiccation/rehydration cycles.

To test whether the differential desiccation tolerance detected from photosynthetic assays of the five taxa (Fig. 4) was mirrored by differential accumulation of intracellular reactive oxygen species (ROS), we used the ROS-sensitive dye DCFH₂-DA loaded into cells and either quantified at the population level using a Biotek Synergy plate reader, or imaged and quantified per cell on a confocal microscope. De-esterification and oxidation of DCFH₂-DA to highly fluorescent DCF indicates that ROS are present within cells.

Fig. 5A shows representative confocal images of DCF fluorescence (yellow) and chlorophyll fluorescence (red) in desiccation-intolerant T. obliquus and in desiccation-tolerant A. bajacalifornicus prior to desiccation (top two panels), and after slow desiccation (bottom two panels). Images for all five species without desiccation, after slow desiccation (over ∼12 h) and after fast desiccation (in <30 min) are shown in Fig. S3. Although dye gets into all cells (see controls in Fig. S4), strong and pervasive intracellular DCF fluorescence appeared most clearly in aquatic taxa after desiccation (see yellow fluorescence, Fig. 5A, T. obliquus, bottom left panel).

Fig. 5B presents data from a 96-well plate-based assay of ROS in samples of algae taken from three separate culture flasks per species. DCF fluorescence from algae that were never dried (pre-desiccation) was minimal (Fig. 5B, blue bars). Fluorescence was increased from algae that were desiccated and then rehydrated with the dye solution (tan bars, Fig. 5B). This increase in DCF fluorescence (indicating ROS production), associated with the desiccation step, was significantly stronger in aquatic taxa (E. costatus and T. obliquus) than in desert taxa (A. bajacalifornicus, F. rotunda and A. deserticola). Samples were taken from n=3 flasks per species. The increase in DCF fluorescence caused by desiccation was significantly larger in aquatic than in desert species, as tested by GLM with species nested within desert or aquatic source habitat (P<0.05, Fig. 5B).

A more labor-intensive, in-depth exploration of DCF fluorescence at the single-cell level was conducted using confocal microscopy of samples obtained from a fourth flask per species. The intensity of yellow DCF fluorescence per cell, in all five species, is shown in Fig. S4 in box and whisker plots. Again, all five species had low DCF fluorescence per cell under fully hydrated conditions, indicating low levels of ROS (Figs S3 and S4). The desert taxa did not develop a strong increase of DCF fluorescence per cell after fast or slow desiccation (Fig. S3 and S4) unless they were actively dividing at the time of desiccation (e.g. Fig. 5A, A. bajacalifornicus, left bottom corner of image).

Results from several controls are also presented in Fig. S4, showing that DCFH₂-DA was inside cells of all species and very little fluorescence was detected when cells were not treated with DCFH₂-DA.

As a model suite of organisms, the five species within Scenedesmaceae examined here offer clear differences in desiccation tolerance (Fig. 4), as well as substantial morphological diversity (Figs 1–3, Fig. S3). To date, they are not all genetically tractable, but we are working with the U.S. Department of Energy's Joint Genome Institute (Walnut Creek, CA) to sequence the genomes of all but A. bajacalifornicus. Guo et al. (2013) and Chen et al. (2016) have reported some success with plasmid-based transformation of T. obliquus. Species within the Scenedesmaceae share many of the characteristics that supported the development of Chlamydomonas as a model system (Jinkerson and Jonikas, 2015). Algae within the Scenedesmaceae and Chlamydomonas are micro-eukaryotes, and are easy to grow in synchronized vegetative culture (Bišová and Zachleder, 2014; Jinkerson and Jonikas, 2015). Vegetative cells are haploid, facilitating mapping of genotype and expressed phenotype. They are in the class Chlorophyceae and share a number of cellular traits with higher plants, including having chlorophyll a and b, starch as a storage compound in chloroplasts, and a cellulosic cell wall.

The very close phylogenetic relationships and shared genetic background among the five taxa targeted for study here make these algae promising organisms for identifying controls over production of unicellular versus coenobial progeny. Daughter cells are physically joined in a coenobium in aquatic taxa E. costatus and T. obliquus, but daughter cells are released as independent, unicellular individuals in desert taxa F. rotunda, A. deserticola and A. bajacalifornicus. While no coenobial taxa within the Scenedesmaceae have been isolated from desert microbiotic crust, the tremendous diversity of terrestrial green algae is clear (e.g. Lewis and Lewis, 2005) and a coenobial representative of the family might ultimately be isolated. Numerous publications from the 1960s to 1990s also indicate that, within some aquatic species in the family Scenedesmaceae, there can be substantial plasticity in spiny ornamentation of cells located at the ends of groups of cells joined single file into coenobia (e.g. Trainor et al., 1976; Lampert et al., 1994). Among the five taxa studied here, we observed notable differences in the numbers of daughter cells produced from each mother cell (4–8 in T. obliquus, E. costatus and A. bajacalifornicus; 2–4 in A. deserticola; and 8–32 in F. rotunda; see Figs 1–3, Fig. S1), but the unicellular versus coenobial fate of daughters appeared to be fixed.

Important advances in understanding of the genomic signature of multicellularity have been achieved previously by the study of the volvocine algae (Kirk, 2005), including Chlamydomonas. Within that clade, extant volvocine species are arrayed across a number of intermediates ranging from unicellular (independent, identical daughters as in Chlamydonomas) to multicellular (coenobial, functionally differentiated daughters as in Volvox) species (Herron and Michod, 2008). Genomic changes correlated with the degree of multicellularity (and cellular differentiation) across the clade are localized dominantly in a few main protein groups (Prochnik et al., 2010). Developmental innovations leading to multicellularity did not involve major changes in the ancestral protein repertoire but, instead, involved expansion of lineage-specific gene families and co-opting of ancestral genes into new developmental processes (Prochnik et al., 2010). A strength in this evolutionary, clade-based approach lies in the common genetic background shared by the organisms, enabling identification of genetic differences correlated with morphologically intermediate stages. Even closer phylogenetic relatedness is found among the five morphologically distinct taxa within the Scenedesmaceae introduced here, providing the opportunity for a distinct future window into determinants of coenobial multicellularity.

We took advantage of the development of gas-permeable PDMS microfluidic chambers (Bascom et al., 2016) to conduct live-cell, time-lapse imaging of cell division in these five species over many hours, from late afternoon, overnight, to morning (Fig. 3, Fig. S1, Movies 11-15). By bathing the PDMS chambers in fresh nutrient solution, we were able to maintain algae in fresh, oxygenated medium. We found they all divided as predicted – by multiple fission (Bišová and Zachleder, 2014). Results from live cells stained with SYBR and FM4-64, and from TEM, revealed that the single chloroplast in a mother cell of these species divides after nuclear division, producing multiple chloroplasts that are then each partitioned into each daughter cell during cytokinesis. Time-lapse imaging (without staining) showed that just before or during cytokinesis, the cell mass rotated within the mother cell wall (Movies 11–15), although the five taxa have cell shapes that vary from round (F. rotunda and A. bajacalifornicus) to oblate (A. deserticola and E. costatus) to very elongated (T. obliquus). It is possible that, in oblate and elongate species, physical forces stemming from the 3D shape of the mother cell wall drive rotation of cell mass during division and packaging into daughter cells. However, it is unclear what forces would drive the rotation of the intracellular contents in the spherical taxa F. rotunda and A. bajacalifornicus. The similar rotation in all five taxa suggests that similar mechanisms act on the cell mass during packaging of the divided chloroplasts and cytosolic components within new membranes and walls; observation of the cytoskeleton during division is clearly a next step for research.

The combination of TEM (Fig. 2) and live-cell FM4-64 staining (Fig. 1) also suggest that Golgi stacks are consistently located close to nuclei as multiple fission progresses. A small, bright FM4-64-stained structure was consistently seen in all five species near daughter cell nuclei (Fig. 1), and further investigation using TEM revealed the consistent presence of Golgi stacks near each daughter nucleus (Fig. 2). Even prior to division of the mother cell chloroplast, TEM revealed that future cell membranes were beginning to be laid down between nuclei (e.g. Fig. 2E), and vesicles are clearly seen between the Golgi stacks and the newly forming membrane. We know of no other example of integrated live-stained cell, long-term time-lapse and TEM imaging of such a suite of closely related eukaryotic microalgae of various morphologies, to examine dynamics and structural details of multiple fission at cellular and whole organism levels.

Confocal imaging of intracellular fluorescence from the ROS-sensitive dye DCFH₂-DA added a physiological perspective to cellular observations. The five taxa were isolated from extraordinarily different aquatic and desert environments, and our chlorophyll fluorescence-based assay of photosynthetic efficiency demonstrated that the desert-derived taxa exhibited strong desiccation tolerance (Fig. 4) – even in the face of repeated cycles of desiccation and rehydration (Fig. S2). Accompanying work in which cells were exposed to 50 µM DCFH₂-DA in H₂O, then imaged using a laser scanning confocal microscope, showed that cell walls were permeable to solution (Fig. S4, positive controls). Overall, the patterns in DCF fluorescence across species, observed in plate-based assays (Fig. 5B) and confocal measurements (Fig. S4), paralleled results from the chlorophyll fluorescence measurements. Desiccation tolerance assessed from Φ_PSII data indicated T. obliquus<E. costatus<<A. deserticola, F. rotunda and A. bajacalifornicus (Fig. 4). Confocal imaging showed accumulation of DCF fluorescence, particularly in the pyrenoid area, within the algal cells (e.g. in T. obliquus and E. costatus; Fig. 5A, Fig. S3), as in lichenized Trebouxia (Candotto Carniel et al., 2015), perhaps related to compartmentalizing oxidized mRNAs in the pyrenoid (Zhan et al., 2015). Fluorescent DCF also accumulated in the periplasmic space in the aquatic, but not the desert, taxa (Fig. S3).

Mean DCF fluorescence (indicating the presence of ROS) in aquatic T. obliquus and desert A. deserticola was stronger in slowly desiccated cells than in rapidly desiccated cells (Fig. S4, ‘experimental data’ boxes). This pattern was unexpected. In mosses, slow desiccation is less damaging because slow rates provide time for the synthesis of protective compounds (Cruz De Carvalho et al., 2012). We speculate that the surprising result here is a function of the daily division cycle in these rapidly growing algae. We took cells from rapidly growing cultures in late morning, and the slow-drying treatment (over 12 h) allowed more cells to enter the division process while desiccation progressed, making them more vulnerable to damage. Fast drying in 30 min, in contrast, was likely complete prior to division being initiated in many cells, so the vulnerable dividing phase was never faced with desiccation. We base our speculation on the observation that large dividing cells e.g. in A. deserticola, F. rotunda and A. bajacalifornicus appear more vulnerable to build-up of ROS (Fig. 5A, Fig. S3). The photoprotective mechanism that controls the production of intracellular ROS in desert species during desiccation and rehydration remains unknown; patterns in chlorophyll fluorescence data obtained at room temperature (RT) for the three desert taxa (Fig. S2) were very similar to those observed by Veerman et al. (2007) in the desiccation-tolerant lichen Parmelia sulcata, which houses the green algal symbiont Trebouxia. Φ_PSII was greatly diminished because a photoprotective mechanism was engaged during desiccation (Fig. S2). Veerman et al. (2007) used time-resolved, chlorophyll fluorescence spectroscopy to show that in Parmelia sulcata, photoprotection resulted from the induction of an unknown mechanism dissipating energy far faster than it could be used by PSII. Slavov et al. (2013) hypothesized that a majority of this protective mechanism is spillover (sharing) of energy from the PSII antenna to PSI, although the mechanism remains unresolved.

In conclusion, these five taxa within the Scenedesmaceae are a promising group for comparative analyses targeting basic cellular and physiological questions regarding form and function of green microalgae. They share genetic background, are ecologically and morphologically distinct, and there is a rich literature describing the biology of Scenedesmus Meyen sensu lato from multiple perspectives. Well-supported ITS2, rbcL, tufA and 18S phylogenetic trees (Sciuto et al., 2015), as well as cell wall characteristics (Hegewald et al., 2013), unite three of the species (A. bajacalifornicus, A. deserticola, T. obliquus) in a single genus (nomenclatural amendment pending; Wynne and Hallan, 2015). The remaining two species, F. rotunda and E. costatus, are not in the same genus; they were chosen for their contrasting ecology, morphology and basal location in the phylogeny of Scenedesmaceae. Further comparative investigation will aim to understand the genetic, ultrastructural and physiological characteristics supporting unicellular versus multicellular coenobial morphology, as well as the remarkable ability of representatives in the Scenedesmaceae to colonize ecologically diverse, even extreme, habitats.

Each of the desert lineages used in this work is monophyletic (Lewis and Flechtner, 2004) and all have been granted species status. The nomenclature for the taxa, however, is in flux. Ongoing attempts to reorganize the family Scenedesmaceae have resulted in substantial variation in the number of genera recognized by different authors. In all papers, however – irrespective of author or genus name applied – there is no debate about the placement of A. deserticola within the same genus as aquatic T. obliquus and desert A. bajacalifornicus, whether that genus is labeled Scenedesmus (e.g. Lewis and Flechtner, 2004), Acutodesmus (Hegewald et al., 2013) or Tetradesmus (Wynne and Hallan, 2015). Currently, Algaebase and NDBI databases assign a different genus name to T. obliquus, A. bajacalifornicus and A. deserticola. This is not because there is argument regarding the evolutionary relatedness of the species. They are consistently placed within one genus based on multiple published biological metrics. Rather, the mismatch in genus names has resulted from a violation of nomenclatural code pointed out by Wynne and Hallan (2015) that stranded A. bajacalifornicus and A. deserticola in the Acutodesmus but allowed T. obliquus to be moved to the new genus Tetradesmus. This difference in genus name does not reflect the evolutionary history or relatedness of these three taxa. Close relatedness of A. bajacalifornicus, A. deserticola and T. obliquus is important for our work, since unicellular (A. deserticola and A. bajacalifornicus) and coenobial (T. obliquus) taxa are members of the same genus, no matter what that genus is currently called. (T. obliquus differs in 18S sequence by 8 nucleotides – 99.6% identity – from A. bajacalifornicus, and by 5 nucleotides – 99.7% identity – from A. deserticola). The other two algal species we include in our analyses (F. rotunda and E. costatus) occupy a more basal position within the subfamily Scenedemoideae. Sciuto et al. (2015) recently renamed Scenedesmus rotundus as Flechtneria rotunda, and Hegewald et al. (2013) suggest the related, basal aquatic species Scenedesmus costatus should revert to the genus name Enallax. Desert F. rotunda and aquatic E. costatus differ in 18S sequence by 23 nucleotides (98.6% identity).

Below, the most-recent authority for each taxon is included within brackets, together with names considered synonymous in the NCBI database and Algaebase.

Flechtneria rotunda Sciuto & L.A. Lewis, 2015 [=Scenedesmus rotundus L.A. Lewis & Flechtner, 2004], isolate BCP-SEV3-VF49 from L. Lewis. GenBank accession (18S) AF513373.

Acutodesmus deserticola (L.A. Lewis & Flechtner ex E. Hegewald, C. Bock & Krienitz) E. Hegewald, C. Bock & Krienitz, 2013 [Basionym: Scenedesmus deserticola L.A. Lewis & V.R. Flechtner ex E. Hegewald, C. Bock & Krienitz], isolate BCP-SNI-2 from L. Lewis, GenBank accession AY510462.

Acutodesmus bajacalifornicus (L.A. Lewis & Flechtner ex E. Hegewald, C. Bock & Krienitz) E. Hegewald, C. Bock & Krienitz, 2013 [Basionym: Scenedesmus bajacalifornicus L.A. Lewis & Flechtner ex E. Hegewald, C. Bock & Krienitz], isolate BEA 0747B from the Banco Español de Algas, Genbank accession KY587455.

Enallax costatus (Schmidle) Pascher, 1943 [≡Scenedesmus costatus Schmidle, 1985; Basionym: Scenedesmus costatus Schmidle], isolate CCAP276-31 from the Culture Collection of Algae and Protozoa, GenBank accession HG514428.1.

Tetradesmus obliquus (Turpin) M.J. Wynne, 2016 [≡Acutodesmus obliquus (Turpin) Hegewald & Hanagata, 2000; ≡Scenedesmus obliquus (Turpin) Kuetzing; Basionym: Achnanthes obliqua Turpin 1828], isolate Utex 72 from the University of Texas Culture Collection, GenBank accession number X56103.

The identity of each of the unialgal cultures was confirmed by sequencing of 18S barcoding region using primers SSU1 (5′-TGGTTGATCCTGCCAGTAG-3′) and C18G (5′-TGGCACCAGACTTGCCCT-3′); N18J (5′-CAATAACAGGTCTGTGATGCCCTTA-3′) and C18J (5′-TCTAAGGGCATCACAGACCTGTTATTG-3′) (Lewis and Flechtner, 2002; Shoup and Lewis, 2003). Sequencing was carried out at the sequencing unit of the University of Chicago Comprehensive Cancer Center, UCCCC. The newly generated sequence for A. bajacalifornicus isolate BEA 0747B, strain obtained from the Banco Español de Algas, was deposited at NCBI Genbank (accession number KY587455).

All species were grown in unialgal cultures. Each of the four replicate cultures of each species was started at a different time; three of four replicate flasks were 250 ml in size, one replicate flask was 50 ml in size. This difference in timing of culture start and size of culture container was intentional. As Lazic et al. (2017) (preprint) note, for clonal organisms (or cell culture lines), the idea of biological replicates is not straightforward. By starting our cultures at different times and by varying the flask size, we had cultures with slightly different histories. The statistical significance of our physiological tests of ROS production and desiccation tolerance was, therefore, more robust. All 250-ml flasks contained 100 ml of a 1:1 mix of Bold's basal medium with micronutrients (Trelease and Trelease, 1935; Bold, 1949) and Woods Hole medium (Stein, 1973). A foam stopper in the top of each 250-ml flask allowed gas exchange. The flask and stopper were fitted and autoclaved with a Pasteur pipette (with cotton plug at top) inserted through the middle of the foam stopper. Sterile solution was added to the flask inside a sterile hood, algae were added to solution from stock cultures; the liquid solution was bubbled with room air through the Pasteur pipette; sterility of the air was maintained by the autoclaved cotton plug. 50-ml flasks were filled in the same way but with 20 ml of solution, then loosely capped with aluminium foil and shaken at ∼100 rpm on a Gyrotory (Model G2) orbital shaker (Scientific Co. Inc., Edison, NJ), without bubbling. Cultures were maintained in a Conviron CMP4030 growth chamber (Conviron, Winnipeg, Canada) with a 12:12 L/D cycle (40 µE mixed metal halide and sodium lamps), at 25°C.

Cells taken from 250-ml flasks, in liquid growth medium, were incubated with a 1:1000 dilution of SYBR Safe (Life Technologies, Carlsbad, CA) for 30 min at room temperature (RT). 15 µM FM4-64 (EMD Millipore, Darmstadt, Germany) was added just prior to imaging. Cells were sandwiched between a glass slide and coverslip sealed with VALAP (1:1:1 vaseline, lanolin and paraffin) and imaged with a Nikon A1-R laser scanning confocal microscope (Nikon Instruments, Inc., Melville, NY). Samples were imaged with a resonant scanner, 60×1.49NA objective, and illuminated sequentially with 488 and 561 nm light. Emission channels were SYBR safe 450/50 band pass filter; FM4-64 525/50 band pass filter; chlorophyll fluorescence 595/50 band pass filter. 16× line scan averaging reduced acquisition noise. For time-lapse imaging, cells were loaded into gas-permeable PDMS microfluidic chambers (Bascom et al., 2016). Chlorophyll fluorescence was imaged every 5–10 min with very low laser light for up to 12 h.

High-pressure freezing (HPF) and automatic freeze substitution (AFS), followed by plastic embedding, were used to produce thin sections of respective samples of cell suspensions. Cells were left to settle in culture solution for 1 h, and 5 ml of the concentrated cell suspension was transferred into an HPF flat specimen carrier and frozen with a Leica EM PACT high-pressure freezer (Leica Microsystems, Inc., Bannockburn, IL) at a typical rate of 1700°C/s. The pods with compacted frozen cells were transferred under liquid nitrogen to the pre-cooled AFS (EM AFS; Leica), and a protocol for cell fixation, H₂O substitution by acetone and a gradual warm-up to RT was followed (Table S1). After 72 h, the samples were released from the pods, washed three times in acetone, gradually infiltrated with Spurr's low-viscosity embedding medium (Electron Microscopy Sciences, Hatfield, PA) in a series of ascending concentration (25%, 50% and 75%, followed by three washes with 100% Spurr's medium for 120 min each) and cured at 60°C for 48 h. Polymerized blocks were sectioned to 70-nm thin sections with a Leica Ultracut UCT ultramicrotome, mounted on Formvar-coated 100 mesh Cu TEM grids sputter-coated with carbon, and post-stained for 7 min with aqueous 2% uranyl acetate followed by 3 min of Reynolds' lead citrate (Reynolds, 1963) prior to TEM imaging. Samples were examined with a Tecnai T-12 TEM (FEI, Hillsboro, OR) operating at 120 kV with a LaB6 filament. Images were collected digitally with a 2×2 K Ultrascan 1000 CCD (Gatan, Pleasanton, CA).

Assays of desiccation tolerance and intracellular ROS were conducted on separate samples of algae taken from three replicate flasks per species (two 250-ml flasks and one 100-ml flask), and deposited in 96-well plates. For ROS assays, 1 ml of culture was removed from each of the three replicate culture flasks and placed into Eppendorf tubes. Cells were allowed to settle for 10 min then 100 µl of the supernatant for each flask of each species was placed in wells of a Corning 96-well special optics low-fluorescence assay plate (Thermo Fisher Scientific), for desiccation. Another 96-well set was prepared for immediate measurements of ROS in control, never-desiccated cells. (The settling step was required to help separate very large, actively dividing cells from smaller non-dividing cells.)

For photosynthetic assays, a ∼15-ml aliquot of culture was removed mid-morning from each Ehrlenmeyer culture flask and placed into a sterile 15- or 50-ml centrifuge tube (Thermo Fisher Scientific, Waltham, MA). Algae were concentrated by allowing them to settle by gravity for 1 h in low room light (∼5 µE). Approximately 1 ml was removed from the bottom of the tube, and 100-µl aliquots of each concentrated culture sample were placed into wells of another 96-well special optics low-fluorescence assay plate.

The 96-well plates for ROS assessment in desiccated cells and for photosynthetic assays were placed in a fume hood, and the disappearance of volume of liquid in test wells around the samples was tracked over time using a micropipette. Liquid loss was homogeneous across the field of wells at ∼10 µl/h, resulting in dryness of the samples by ∼10 h.

Chlorophyll fluorescence was assayed using a Walz Junior PAM chlorophyll fluorometer (Heinz Walz GmbH, Effeltrich, Germany) through the clear bottom of each sample well of the 96-well plate prepared as described above. Φ_PSII was calculated following equations in Genty et al. (1989). Briefly, the Walz fiber optic was positioned 1 mm below the well bottom. The fiber optic gathered and the control unit quantified chlorophyll fluorescence excited continuously by low-intensity-modulated (10 µE) white light, and also excited by saturating flashes of light (2000 µE, 600 ms duration). Φ_PSII was calculated as (F_m′−F_t)/F_m′, where F_m′ is fluorescence emitted during a saturating flash, and F_t is fluorescence emitted in the presence of the modulated measuring light only. (See Fig. S2 for chlorophyll fluorescence traces with annotated F_m′ and F_t, from in-depth coverslip assays described below.) The Φ_PSII for desiccated algae was assessed first, then 50 µl distilled H₂O was added to the well, and Φ_PSII assessed again 3 min after rehydration. The change in Φ_PSII upon rehydration was significantly larger in desert versus aquatic species, as assessed using GLM, with species nested within native habitat (desert or aquatic) (Systat 12, Richmond, CA; n=3 samples per species).

We used the reactive oxygen species (ROS)-sensitive dye 2′,7′-dichlorodihydrofluorescin diacetate (DCFH₂-DA, Sigma-Aldrich, Saint Louis, MO) reconstituted with DMSO, at 50 µM concentration to assay for the presence of ROS. By using the 96-well plates prepared as described above (one containing cells that were never dried, and the other containing cells that were slow-dried, with wells tracked by flask), the DCF fluorescence (excitation 500 nm, emission 529 nm) normalized to absorbance (600 nm) from each well was assessed in a Biotek Synergy HT plate reader (Gain 100, Biotek Instruments, Winooski, VT). The difference in DCF fluorescence from never-desiccated versus desiccated-then-dye-rehydrated cells, paired by flask, was assessed using GLM, with species nested within source habitat (desert or aquatic) (Systat 12, Richmond, CA, n=3 samples per species).

For examination of ROS by confocal microscopy, for all associated controls described in Fig. S4, and for chlorophyll fluorescence assays testing Φ_PSII during multiple cycles of desiccation and rehydration (Fig. S2), samples of algal were taken from only one Ehrlenmeyer flask per species, and concentrated as described above. Cell density in the concentrated sample was then adjusted to optical density 1.8 at 600 nm absorbance measured in a Biotek Synergy HT plate reader (Biotek Instruments, Winooski, VT). 40 µl drops of algal solution were then deposited onto 14 new microscope slide coverslips per species (#1.5, 18 mm×18 mm, Thermo Fisher Scientific). Seven coverslips were immediately placed in the dark to dry in a fume hood. They dried within 30 min in the flowing air, resulting in a round dot of dried algae in the center of each coverslip. (This drying was substantially faster than in the 96-well plates because the drops were unprotected on coverslips.) The remaining seven coverslips were transferred to a custom drying chamber. Light levels were less than 1 µE in the drying chamber room. Air flowed through the drying chamber, desiccating the algae slowly over 12 h. Water content of air was controlled by bubbling through warmed distilled H₂O then passing the air through a glass condenser held at 0.7°C below RT by an Isotemp 3016D water bath (Thermo Fisher Scientific). An astronomy camera (Acton PI 1 kb Versarray Cooled Camera, Princeton Instruments, Trenton, NJ, with WinView/32 software) imaged drops of algae every 30 min as they diminished in size to be sure of the slow speed of drying (12 h to full dryness). Fast-dried and slow-dried algae were stored at RT in the dark.

To assess the capacity of the various species to survive multiple cycles of desiccation and rehydration, 40 µl of algae from one flask per species were dried on three coverslips. Each coverslip, with dried algae in the center, became the bottom face of an 18 mm×18 mm×6.3 mm acrylic chamber with inlet and outlet ports for air flow (Fig. S2). The coverslip was sealed to the chamber using a thin ring of EZShape modeling clay (Polyform Products, Elk Grove Village, IL), resulting in the algae being enclosed within the chamber. Air pumped through desiccant (Drierite, Hammond Drierite Co., Xenia, OH) flowed at 1 ml/min through the chamber (over the algae), out an exit tube, and to a Licor 6262 infrared gas analyzer (Licor Inc., Lincoln, NE) connected to a Campbell Scientific CR10× datalogger (Campbell Scientific, Logan, UT) logging H₂O content of air exiting the chamber (Fig. S2).

Φ_PSII was assessed using the Walz Junior PAM chlorophyll fluorometer largely as in the 96-well plate experiment. The Walz fiber optic (outside the chamber) was positioned 1 mm below the coverslip, directly below the dot of algae that had been previously dried on top of the coverslip (Fig. S2A). The fiber optic gathered chlorophyll fluorescence excited continuously by low-intensity modulated (10 µE) light, and also excited by saturating flashes of light (2000 µE, 600 ms duration) applied every 3 min throughout the experiment. Φ_PSII was calculated as (F_m’-F_t)/F_m’.

At the beginning of the experiment, after one flash in the desiccated state, algae were hydrated by disconnecting the air exit tube, inserting a syringe needle through the tube port into the chamber, depositing 10 µl distilled H₂O directly onto the algae to rehydrate them, and rapidly reconnecting the air exit tube. This rehydration was designated ‘Cycle 1’ (Fig. S2). Saturating flashes continued to be applied every 3 min. Dry air flowed through the chamber continuously picking up H₂O and, therefore, re-drying the algae over tens of minutes. Once the Licor6262 showed the air stream was again dry, H₂O (5 µl) was injected to hydrate algae again in ‘Cycle 2’. Aquatic algae showed no capacity for charge separation (no spikes of fluorescence during the saturating pulses, and therefore a quantum yield of zero, e.g. Fig. S2) after two cycles of desiccation and rehydration, so were not subjected to a third cycle. Desert algae continued responding to rehydration with increased Φ_PSII so were subjected to a third cycle of desiccation and rehydration and rebounded well (Fig. S2).

We also used DCFH₂-DA to assay for the presence of ROS on a per cell basis for the five species (Kalyanaraman et al. 2012), using algae deposited on coverslips as described above. Four coverslips of algae per species, were used in ROS experiments. (See Fig. S4 for a pictorial guide to the experiment.) Two coverslips had 40 µl distilled H₂O containing 50 µM DCFH₂-DA added to the algal cells on the coverslip: one coverslip with fast-dried cells, and one coverslip with slow-dried cells. A third coverslip held hydrated cells that had never been desiccated and to which DCFH₂-DA was added to bring the solution to 50 µM. The fourth coverslip held hydrated cells that had never been desiccated, and no added DCFH₂-DA; these algae were used to test for any autofluorescence that could interfere with the DCF signal in the assay. The four coverslips per species were then incubated 15 min, in the dark, at RT.

Cells were imaged in a confocal scanning laser microscope Axio Imager, Zeiss LSM 780 (Carl Zeiss SAS, Jena, Germany) using ×100 magnification (Plan-Apochromat 100×/1.40 oil objective) and Argon excitation laser (λ_exc=488 nm, 0.8% laser power, frame time 18.18 s, final image size of 85 µm²). Cells in at least six fields of view were imaged per treatment, in multi-channel mode (DCF, λ_em=502-589 nm emission; chlorophyll fluorescence λ_em>652 nm emission). DCF fluorescence in confocal images was quantified using Image J 1.51j8 (Schneider et al., 2012; http://rsbweb.nih.gov/ij/). Cell outlines were identified from combined DCF and chlorophyll channel signals and separated when necessary by using the watershed option in ImageJ. DCF fluorescence intensity per cell was quantified as the sum of the DCF channel gray values of all the pixels in each cell divided by the number of pixels within each cell outline. The mean (±s.e.m.) number of cells contributing to each box labeled ‘experimental data’ in Fig. S4 was 214±32 cells (median 208 cells). The mean (±s.e.m.) number of cells contributing to each of the other boxes, from the various controls in Fig. S4, was 43±9 (median 32 cells).

All imaged cells were scored for DCF fluorescence per cell. Data from the pre-desiccation, fast desiccation and slow desiccation treatments are shown in box plots in Fig. S4. Results echo those obtained with from experiments using the Biotek plate reader but images provide the extra information that large, actively dividing cells produce ROS more readily than mature cells that are not dividing. Also, DCFH₂-DA is definitely entering cells of all species and ROS can be induced in all by minutes of exposure to high light. Controls illustrated there was no contamination of the DCF fluorescence signal by autofluorescence, and the extent of DCF fluorescence was not influenced by incubation time (Fig. S4).

Dr Louise Lewis (Department of Ecology and Evolutionary Biology, University of Connecticut) generously provided F. rotunda and A. deserticola. Dr John Oakey (Department of Chemical and Petroleum Engineering, University of Wyoming) generously provided PDMS microfluidic chambers. Suzanne Thomas provided expert technical assistance.