Enrichment cultures consisting of motile, filamentous cone-forming cyanobacteria and other microbes (Electronic Supplement, [34]) were enriched by serial harvesting and culturing of cones in a modified Castenholz medium D [17], as well as by the selection of the fastest-gliding phototactic filamentous cyanobacteria [34]. These organisms were identified as cyanobacteria using 16S rRNA sequences and chlorophyll a (Chl a) autofluorescence (described in more detail in [34] and Electronic Supplement). Cyanobacterial clumping in our laboratory enrichment cultures occurred in morphologically and chemically distinct stages and depended on culturing conditions. When growing on flat surfaces and in initially anaerobic solutions equilibrated with 5% CO₂, the mats first spread laterally (Figure 2A), then formed regularly spaced clumps (Figure 2B), and finally produced topographically distinct pinnacles and cones (Figure 1D–F). In contrast, when aerobic enrichment cultures were initially equilibrated with air, mats were visible only as clumps and cones, and only later spread laterally (Figure 2C). During the lateral spreading and thickening of flat mats in the initially anaerobic enrichment cultures, the concentration of oxygen increased, but the appearance of regularly spaced clumps coincided with a decrease in net oxygen production (Figure 2D). At a given light intensity, clumps always appeared at similar oxygen concentrations, emerging one week earlier and exhibiting a larger spacing (2.21 ± 0.81 mm, average ± standard deviation for all the remaining results, N = 71) at 160 μE/m²/s than at 10 μE/m²/s (0.82 ± 0.16 mm, N = 114, p < 0.01). Clumping in anaerobic laboratory cultures occurred in the presence of 5 mM nitrate and 28 mM DIC. Because these concentrations were much greater than the K_m for nitrate (several tens of μM, [35]) and the K_m for CO₂ and HCO₃⁻ (in the μM range, [36]), clumping under our experimental conditions could not be interpreted as a direct response to low concentrations of nitrate and inorganic carbon.

To the naked eye, clumps appeared as circular features darker than the adjacent flat mats (Figure 2B). Confocal microscopy showed that the newly discernible clumps were not elevated by more than 19 ± 8 μm (N = 6) above the mat, but contained 1.83 ± 0.44 (N = 6) times denser cyanobacterial cells than the surrounding mats (p < 0.01). Cyanobacterial filaments in clumps were randomly oriented, bowed vertically and entangled (Figure 3). In contrast, those around clumps were gently curved, loosely intertwined, predominantly horizontal (Figure 1C, Figure 2 and Figure 3), and often aligned into dense bundles that bridged adjacent clumps (Figure 1B). The randomly oriented, three-dimensionally bent filaments in clumps (Figure 3) were surrounded by abundant, presumably heterotrophic, small cells that did not exhibit Chl a autofluorescence (Figure 3). Analyses of molecular diversity of laboratory enrichment cultures confirmed the presence of various non-photosynthetic bacteria (Electronic Supplement). Clumps also contained 17 ± 9 (N = 4) times denser cyanobacterial sheaths stainable by Calcofluor White than the surrounding areas (Figure 4, p < 0.01). Most sheaths within the clumps were abandoned, i.e., they had a filamentous form, but did not contain detectable Chl a or cells. At the same time, stainable sheaths were rare where cyanobacterial filaments were well-aligned and horizontal (Figure 4). These observations confirmed that clumps did not form as small blisters around gas bubbles, although bubbles lift up some cyanobacterial mats into cm-scale towers [37,38]. The presence of abandoned sheaths in the centers of clumps also demonstrated the continuing motility of three-dimensionally bent filaments within clumps (Figure 4 and Figure 5, Electronic Supplement).

Mats first appeared as clumps in cultures that were initially equilibrated with air (Figure 2C). In contrast, those in initially anaerobic cultures developed clumps only after the concentration of oxygen stopped increasing (Figure 2D). The concentration of O₂ in the initially anaerobic medium increased from 0 to more than 200 μM, the concentration of DIC in the same medium decreased from 29 to 28 mM, and the concentration of nitrate remained in the mM range. These observations suggested a link between the formation and persistence of clumps and the changing amounts of O₂, a major photosynthetic product. To determine the influences of O₂ and DIC, respectively, we measured cyanobacterial horizontal gliding rates away from clumps in media containing different amounts of O₂ or DIC, respectively (Figure 5, Experimental Section 3.3). At low oxygen concentrations, filaments migrated away from clumps fast, but hardly at all at 95% O₂ (Figure 5). The rates of clump dispersal also decreased with the increasing concentrations of DIC in the medium (Figure 5), but clumps dispersed over the entire range of experimentally examined DIC concentrations when oxygen was absent (Figure 5, Experimental Section 3.3).

The notable influence of oxygen on the persistence of clumps (Figure 5) suggested that the cycling of carbon, oxygen and nutrients within clumped mats also may depend on the concentration of oxygen. To test this, we examined the cycling of carbon and oxygen in clumps and adjacent areas, and between clumped and mechanically dispersed, formerly clumped mats (Figure 6, Experimental Section 3.4). The gross photosynthetic rates were always higher in clumps, scaling with the higher filaments density in clumps (Figure 6A). These measurements indicated that filaments in clumps and flat areas had comparable gross photosynthetic rates. More surprisingly, oxygen concentrations at the surfaces of clumps were lower than those at the surfaces of mats (Figure 6B, Experimental Section 3.4), although clumps grew upward faster than the surrounding areas (Figure 1).

To learn more about the cycling of carbon and oxygen in clumped mats, we measured the gross and the net photosynthetic rates in clumped and dispersed mats at different concentrations of O₂ in the medium (Experimental Section 3.4). Net oxygen production reflects the amount of fixed inorganic carbon that is not remineralized in the mat, while the difference between the gross and the net photosynthetic rate depends on the intensity of oxygen-removing processes, photorespiration and aerobic respiration (Equation 1). Photorespiration occurs in organisms which incorporate inorganic carbon using ribulose-1,5-bisphosphate carboxylase/oxygenase (RuBisCo), an enzyme that can act both as a carboxylase and an oxygenase. During photorespiration, the oxygenation of RuBisCo is accompanied by the excretion of soluble organic carbon (e.g., [39,40]), reducing the incorporation of inorganic carbon into cellular biomass. Increasing O₂ concentrations can thus lead to higher rates of oxygen removal and lower net photosynthetic rates (Equation 1).

The gross photosynthetic rates of dispersed mats, measured in the presence of 154, 446 and 577 μM O₂ (Experimental Section 3.4), did not depend on the concentration of O₂ (p > 0.2, N = 10 measurements per condition). Therefore, high O₂ concentrations did not measurably impact the production of oxygen by cyanobacterial cells. In turn, net photosynthetic rates decreased with the increasing O₂ concentration in both intact and dispersed samples (Figure 6C). This indicated an increase in the combined respiratory activity at increasing O₂ concentrations (Equation 1). A similar trend between net photosynthetic rates and O₂ concentrations was reported previously in microbial mats dominated by cyanobacteria which do not form clumps and cones [41]. In anaerobic solutions, mechanically dispersed mats had higher net photosynthetic rates than clumped mats (Figure 6C, p < 0.05, N = 9). These rates did not differ between dispersed and clumped mats at higher oxygen concentrations (p > 0.1, N = 13, Figure 6C). Because cyanobacteria in clumped and dispersed mats exhibited similar gross photosynthetic rates (Figure 6A), and these mats had similar net photosynthetic rates in oxic media (standardized by total Chl a or cyanobacterial density, Experimental Section 3.4), Equation 1 shows that clumping did not impact the sum of photorespiration and aerobic respiration in oxic media. Furthermore, the observed accumulation of particulate organic carbon in clumps (Figure 3, Figure 4 and Figure 6B) was not consistent with the observed lower oxygen concentrations above the clumps (Figure 6B). Therefore, we hypothesized that more soluble organic carbon may be respired aerobically within clumps.

To determine whether more dissolved organic compounds leaked from the dispersed mats, we measured the production of glycolate, a soluble product of photorespiration. Net glycolate production in the medium is described by:

where F describes the fraction of total aerobic respiration that uses glycolate as a substrate. After a 1-h incubation in the oxic media, dispersed mats produced 69 ± 7 nmoles of glycolate per μg Chl a, whereas clumped mat released less glycolate, 43 ± 3 nmoles per μg Chl a (N = 4, p < 0.01). The lower net leakage of soluble organic carbon from clumped mat could lead to the accumulation of more particulate organic carbon there, even though both clumped and dispersed mats produced similar amount of oxygen in oxic media. Three mechanisms can account for this observation. First, clumping reduces photorespiration and enhances the aerobic respiration. These two processes are not independent, because aerobic respiration lowers the O₂/CO₂ ratio, which, in turn, can reduce photorespiration. Second, clumping might increase F in Equation 2, if clumped mats contain more heterotrophic bacteria that prefer glycolate to other organic compounds. This scenario is unlikely under our experimental conditions, because clumped and dispersed mats preserved the ratios of non-photosynthetic cells, cyanobacteria and particulate material larger than ~1 μm (Experimental Section 3.4). We thus expect F (Equation 2) to be the same in clumped and dispersed mats. Third, the diffusion coefficients for the diffusion of small molecules and ions in clumps may be lower than those of the surrounding areas. This mechanism augments the first one by increasing the availability of glycolate and oxygen for aerobic respiration within the clumped mats.

The characteristic pattern in clumped cyanobacterial mats (Figure 1B and Figure 2B) arises from the ability of filamentous non-heterocystous cyanobacteria to bow, bend and reverse gliding directions. These processes initially distribute filaments in clumps, ridges and less dense surrounding areas and continue to influence the orientation of filaments during later growth (Figure 5, Electronic Supplement). The persistence of newly formed clumped mats strongly depends on the concentration of O₂ (Figure 5), because high O₂ concentrations in the medium around clumps reduce the dispersal of clumps (Figure 7A). At this point, it remains unclear whether high oxygen concentrations promote clumping directly, or indirectly, by influencing the production of other diffusible metabolites and signaling molecules, which are then sensed by cyanobacteria. In any case, horizontally gliding filaments are more likely to reverse their gliding direction when oxygen is present in the medium outside of clumps (Electronic Supplement), whereas the already bent and bowed filaments are less likely to leave the clump (Figure 5 and Figure 7A). The bent and bowed filaments in clumps also are more likely to span diffusive boundary layers (e.g., [30,42,43]), compared to the predominantly horizontal filaments in the surrounding areas.

Cyanobacterial filaments also align into ridges that bridge clumps (Figure 1B), and into thin laminae on the sides of larger cones (Figure 8). Currently, the formation of ridges is attributed to undirected gliding motility [44], but this model explains neither the denser concentration and alignment of cells in ridges, nor the thickness and the regular spacing of the ridges. The organization and accumulation of cyanobacterial filaments in these areas may depend on cell-to-cell contact and pilus-mediated gliding motility in a similar fashion to the alignment and persistence of aligned regions of much smaller and non-filamentous myxobacteria [45].

In quiet, stratified solutions, the flow of nutrients into the mat and the flow of waste products out of the mat depend on the microbial organization within the mat [30,42,43,46]. Assuming that the initial transition from the “flat” mats (Figure 2A) to the clumped ones is a behavioral adaptation that allows a better use of resources, one can develop a simplified model of clump spacing. In this geometric model, a flat mat becomes clumped when the surface area to volume ratio of the clumps is higher than that of a flat mat. Consider a volume of bacteria V distributed over an area A. The bacteria can either be distributed as a flat sheet of thickness δ or organized into hemispherical clumps of radius r. In flat mats, the surface area to volume ratio is 1/δ. In clumped ones, the surface area to volume ratio is 3/r. Consequently, if r/δ> 3, nutrients and waste products flow more easily into and out of a flat mat, but the biofilm becomes more efficient by forming clumps if r/δ< 3. Thus, the biofilm should form clumps when

At the very transition from flat to clumped mats, the thickness of biomass from clumps, if spread over the entire area occupied by a clumped mat, is equal to the thickness of the flat mat spread over the same area. If the clumps are arranged on a hexagonal lattice (spaced by λ; Figure 1B and Figure 2B), then the area per clump is:

If each clump were redistributed uniformly over this area a, the thickness Δ of the resulting mat would be:

Clumping occurs when Δ > 3/r. This requirement and the rearranged Equations 4 and 5 give the distance between newly formed clumps:

Note that the ratio described by Equation 6 will be greater if the clump height is smaller than the clump diameter, or if clumps contain denser filaments than mats. The value predicted by Equation 6 is generally consistent with the observed ratios of clump spacings and diameters in newly clumped mats (Figure 1B and Figure 2B). A more precise prediction would require a more detailed model of how nutrients and waste products flow through flat mats and within clumps. In general, regularly spaced clumps are more likely to develop on flat mats than in the presence of three-dimensional photosynthetically active structures (e.g., preexisting cones), because these structures create more complex chemical gradients in the solution (e.g., [47]).

An additional constraint is required to understand the magnitude of clump diameters. A simple and plausible possibility is that the size of a bent filament constrains the minimum size of structurally stable clumps. In this case, the size of nascent clumps is determined by the material properties of the bacteria. This, in combination with the surface area argument, determines λ. Alternatively, the spacing may be constrained by the flow of nutrients around aggregates [41].

Once the clumped pattern is established (Figure 1), clumps continue to grow upward. This growth pattern relies on the growth of new cyanobacterial and non-photosynthetic cells, the accumulation of cyanobacterial sheaths (Figure 3 and Figure 4), and the presence of upright or bent filaments in clumps (Figure 1 and Figure 3). Similar processes may build laminae composed of clumped or randomly oriented filaments in mature cones [34] and form small clumps on the sides and tips of mature cones ([24], Figure 8). The photosynthetic activity and cyanobacterial density in larger clumps and cones does not remain constant, but declines below a ~0.2 mm thick surface layer [34].

In oxic solutions, clumping reduces net photorespiration, measured as the excretion of soluble organic carbon above the mat, and the loss of organic carbon from the mat (Figure 2 and Figure 6). The secretion of glycolate in our cultures shows that clump-forming cyanobacteria cannot completely abolish the photorespiration pathway, even though cone-forming cyanobacteria are able to concentrate carbon [34], and clumped mats grow in the presence of 25 mM DIC. The requirement for photorespiration appears to be a general property of cyanobacterial CCMs (e.g., [40]). The close spatial proximity of cyanobacteria and other microbes (Figure 3 and Figure 7B), the availability of sheaths as attachment surfaces for aerobic heterotrophic microbes and the retention of nutrients and photosynthates within the dense exopolymeric matrix [48] may lower the O₂/CO₂ ratios in clumps more than those in the surrounding areas, promoting the accumulation of organic carbon in clumps (Figure 7B). It remains to be seen whether cyanobacteria continue to form clumps in the absence of heterotrophic microbes. To date, our attempts to obtain pure cultures in which to test this have remained unsuccessful.

At first glance, the continued growth of clumps is at odds with the lower O₂ concentrations above clumps (Figure 6B). Namely, less O₂ leaving the clumps may suggest a more extensive organic degradation rather than a faster accumulation of particulate organic carbon. Due to the small size of clumps and surrounding areas, we were unable to compare the net photosynthetic rates of clumps and the surrounding areas, or compare the release of glycolate by clumps and the surrounding areas. However, if clumps release less glycolate out of the microbial mat than the surrounding areas, and this intense recycling leads to the production of more particulate carbon, this may account for the lower oxygen concentrations above clumps. This hypothesis can be further explored by using enzyme-based microsensors specific to glycolate (e.g., [49]).

Clumping occurs only at certain stages of biofilm growth, and only in limited areas. At low oxygen concentrations, or in the presence of sulfide [33], heterotrophic respiratory activities can be limited both by the availability of O₂ and by the supply of organic substrates. Under these conditions, each cyanobacterial filament would maximize the availability of nutrients in the growth medium, forming a “flat” mat. Indeed, when incubated in the initially anoxic solutions, dispersed mats release more oxygen into the medium per unit time than clumped mats (Figure 6C). This suggests that the benefit of clumping is smaller in the absence of oxygen, but promotes the growth of clumps in oxic media (Figure 7B).

Overall, our data underscore the role of oxygen in the organization, growth and persistence of clumped mats and point to benefits of clumping that may be unique to oxic systems. Clumping, as an oxygen-dependent behavior which reduces the loss of soluble organic carbon from the mat and promotes the accumulation of particulate carbon in clumps, should be specific to mats constructed by RuBisCo-containing primary producers such as photosynthetic cyanobacteria, some anoxygenic photosynthetic bacteria, or chemolithotrophs such as Beggiatoa sp. [50]. Because sub-mm clumps spaced by ~1–2 mm are present in oxic but aphotic mats (e.g., [30]) and in oxygenic photosynthetic mats in YNP (Figure 1A in this paper and [24]), additional observations are needed to recognize photosynthetic structures. Specifically, small clumps spaced by ~1 mm can be viewed as a biosignature of oxygenic photosynthesis only if accompanied by small conical stromatolites and/or reticulate ridges. To the best of our ability, we are not aware of alternative models that could account for the entire morphogenetic sequence from clumps to cones or reticulate ridges, but would not require oxygen.

Unlike the formation of clumps, the ability of filamentous organisms to glide, clump, align, form ropy bundles or bend may have little to do with the presence of oxygen. Anoxygenic photosynthetic filaments (e.g., Chloroflexus and Roseiflexus sp.) glide [51] and form alternating thin-dense and thick, porous laminae in the presence of non-filamentous cyanobacterium Synechococcus [52]. Neither cyanobacteria nor oxygen are required for the formation of clumps by the filamentous green sulfur bacterium Chloroherpeton [53], or for cAMP-dependent aggregation of the thin, gliding, filamentous photosynthetic bacterium Chloroflexus aggregans [54,55]. Dense filament bundles of unknown organisms embedded in calcite also occur in aphotic microbialites described from cold seeps [56], whereas aligned and randomly oriented filaments of gliding, sulfate-reducing Desulfonema sp. are reported from some sulfidic marine sediments [57]. Tufted and string-like macroscopic aggregates that contain aligned filaments are also present in the laboratory enrichment cultures fueled by anoxygenic photosynthesis in the presence of low sulfate and sulfide concentrations [10]. Therefore, we advise caution when interpreting tufted textures in stromatolites because it is unclear how to distinguish cyanobacterial tufts from those formed by various gliding anaerobes.

Tufted and conical microbial mats are rather common in modern intertidal and other shallow water environments [13,58,59], but their preservation potential is very low [60] and requires early lithification. The need for early lithification may account for the relative paucity of these features in the geological record, particularly that of the Phanerozoic. A rare example of the entire morphogenetic sequence starting with small clumps and ending in cones is reported from the Early Mesoproterozoic Satka Formation [61,62]. Clumps also appear to have initiated the formation of small, laterally-linked conical stromatolites from the 1.8 Ga Mistassini Formation [63], but there, clumps are not well preserved due to recrystallization and dolomitization.

Evidence for cyanobacteria in the Mesoproterozoic and Paleoproterozoic is reassuring, but not particularly surprising. A much older morphogenetic sequence from clumps to conical and ridged stromatolites was recently reported in the 2.72 Ga Tumbiana Formation [13], i.e., 0.2–0.4 billion years before the GOE. In the Tumbiana Formation, mm-scale clumps grade vertically into centimeter-scale conical stromatolites, which are linked by regularly occurring inter-connecting laminae (Figure 9). The rarity of unambiguous detrital grains in these stromatolites indicates the laminae were likely precipitated in situ, or as fine grains in the water column directly above the stromatolites. Early lithification in this environment preserved fine-scale features such as an axial zone, mm-scale clumps, tufts and reticulates (Figure 9A). This morphological sequence argues for the presence of microbes that moved, metabolized and aggregated in the presence of oxygen in shallow water environments at 2.7 Ga. The record of Archean stromatolites at 2.7 Ga, thus, independently corroborates other proposed indicators of oxygenic photosynthesis and locally oxygenated systems 0.2–0.4 billion years before the rise of atmospheric oxygen.