Introduction

Diatoms are a species-rich group of microeukaryotes contributing to ca. 20 % of global carbon fixation [1] and are important contributors to the ocean carbon pump. Due to the formation of their siliceous cell wall from dissolved silicate, they also affect the global Si-cycle [2]. Although fundamental aspects of diatom physiology have become increasingly well understood [3, 4], our knowledge of their diplontic life cycle is still very limited. Open questions include, for example, cell size-dependent changes in cell physiology and behavior, dormancy, and auxosporulation [5, 6]. In those diatoms for which data are available, relatively long periods of vegetative growth alternate with short episodes of sexual reproduction [5, 7]. During vegetative cell division, one of the two biomineralized halves of the cell wall (thecae) is inherited by each of the daughter cells. The production of the new thecae occurs within the confines of the parental walls, leading to a gradual decrease in the mean cell size of the population [5]. Upon reaching a sexual size threshold (SST), restoration of the original cell size can occur via sexual reproduction, a key process in the life cycle of many diatoms [5]. After mating, a zygote is formed that develops into an auxospore, which upon germination forms a large initial cell that starts again to divide mitotically. Sexual reproduction of pennate diatoms can sometimes be triggered by environmental cues but recent findings show that pheromones play a key role in the control of sexuality [5, 8, 9].

Successful mating requires synchronized behavior between mates that leads to physical encounters [10]. This necessarily implies that one or both mating partners are motile and have the ability to locate their partners. Motile male gametes of the diatoms Pseudostaurosira trainorii locate stationary female gametes via pheromone sensing while non-directed random walk is observed in Tabularia fasciculata [9, 11]. These two species are araphid pennate diatoms that exhibit anisogamy wherein the pairing gametes differ behaviorally and morphologically. In contrast, most raphid pennates, including the emerging model species Seminavis robusta produce morphologically and physiologically similar isogamous gametes after gametangiogamy [5]. In this species, the induction of the sexual phase as well as the mate-finding under the control of pheromones is confined to gametangial cells. Gamete production only takes place after successful pairing of gametangial MT+ and MT− cells. The first diatom pheromone was only recently identified as the l-proline-derived diketopiperazine (in the following abbreviated as diproline) in S. robusta using a metabolomics-enabled approach [8]. Diproline is released by the mating type MT− after perceiving chemical signals of the opposite mating type MT+ that indicate its presence. Attraction assays showed that MT+ cells exhibit a finding behavior towards their mating partner. This behavior can also be observed upon the addition of polymer beads releasing the pheromone diproline as cells accumulate around the pheromone-loaded beads [8]. Diproline is part of a complex signaling system with two other putative pheromones that induce sexualization in the opposite mating types [8, 12].

Raphid pennate diatoms like S. robusta move by gliding through adhesive mucilage excreted from elongated slits in the cells’ frustules (raphe). The extruded extracellular polysaccharides (EPS) are not only responsible for forward and backward movement but can also form pseudopods or stalks that adhere to the substratum and produce a torque that supports cell rotation. This results in a back and forth movement wherein cells can reverse direction after each stop [13]. Diatoms can utilize directed motility to respond to light [14], monosaccharides [15], and dissolved silicate [16], by biasing their movement towards or away these signals. In this contribution, we record, compare, and analyze the movement of the attracted mating type MT+ of S. robusta in the presence and absence of diproline gradients generated from diproline-loaded polymeric beads [8]. We show for the first time using a pure pheromone, delivered in defined concentrations how motile diatom cells exhibit a chemotactic and chemokinetic response.

Methods

Strains and Culture Conditions

We used the S. robusta strains 85A (MT+) and 84A (MT−) both below their sexual size threshold (∼30 μm). The strains are maintained in the BCCM/DCG diatom culture collection at Ghent University (http://bccm.belspo.be/about-us/bccm-dcg). Stock cultures were grown in buffered artificial sea water (ASW) as described by Maier and Calenberg [17]. For tracking experiments, cultures were prepared by inoculating 15 % stock culture to fresh ASW and kept in standard tissue culture flasks or well plates (Sarstedt, Nümbrecht, Germany). Both were grown at 18 °C in a 12:12 h light:dark regime with cool-white fluorescent lamps at approximately 35 μmol photons m−2 s−1.

l-Diproline Attraction Assays

The attraction assay methods were modified after Gillard et al. [8]. MT+ cells were grown for 2 days in 24-well plates (1.7 mL well−1) and thereafter dark-synchronized for 36 h [18]. MT− cells were grown in culture flasks for 3 days. MT+ cells were induced by adding 600 μl of filtered MT− medium into each well, and the plates were kept in the light for 6 h. For bead preparation, aliquots of 0.5 mg HLB-SPE (hydrophilic-lipophilic balanced solid phase extraction, Oasis®, Waters, Eschborn, Germany) material were incubated in 1 ml deionized H2O with or without synthetic diproline (2 nmol) for 1 h at room temperature. Five micrograms of beads were added into each well of induced MT+ culture and immediately afterwards, the swimming behavior of MT+ cells was recorded for 600 s. Observations, cell culture photography, and video recording were done on an inverted Leica DM IL LED light microscope (Heerbrugg, Switzerland) mounted with a Nikon DS-Fi2 CCD camera (Tokyo, Japan).

Determination of the Pheromone Gradient

The amount of diproline diffusing from these beads and the concentration gradient this process produced were determined through a combined experimental and calculation approach. Diproline and control beads were prepared as described above. Pheromone-loaded beads (167 μg) were added to 50 mL fresh ASW and after 10 min exposure, the beads were filtered out. The equivalent amount of diproline on the beads (0.668 nmol) was also added to 50 mL fresh ASW to determine the maximum amount that could be adsorbed by the bead. Prior to extraction, 1.5 nmol of caffeine was added to each 50 mL sample as internal standard. The diproline concentration in both experiments was determined after SPE extraction according to an established method by using gas chromatography coupled to mass spectrometry on an ISQ Trace GC-Ultra GC-MS system (Thermo Fisher, Dreieich, Germany) [8].

Based on this concentration, we calculated the total flux of diproline that diffused from the bead (i) after 10 min using the formula from Barbara and Mitchell [19]:

$$ C=i/4\pi rD $$

where C is the measured concentration that diffused from the bead into the medium, r is the radius of the bead (15 μm), and D is the diffusion coefficient for small molecules (10−5 cm2 s−1) [20]. From the calculated i, we then determined the steady state concentration C (r) using the same formula mentioned above. To correct for the two dimensionality of the biofilm, we replaced r with the square root of the radius (√r) of the observation area (250 μm) [21]. The time to steady state was determined as the time equal to or greater than d 2 /D where d is the diameter of the whole observation area (500 μm).

Video and Statistical Analyses

In total, we recorded five videos for diproline and four for control treatments. On each field of view, 1–7 beads could be used in individual observation fields resulting in a total of 24 beads for diproline and 7 for control. The open-source software Fiji [22] with the plug-ins Cell Counter and TrackMate (http://fiji.sc/TrackMate) was used for cell counting and tracking, respectively. All videos were used for cell counting whereas only three videos each from control and treatment were randomly chosen for cell tracking. Movies were taken for 600 s with a frame rate of 100 frames per second. The first 60 s of the movie were excluded from all statistical analyses as cells were disturbed during and shortly after the sinking of the introduced beads. Cell accumulation around the particles was determined from a microscopic picture frame extracted every 30 s from the movies. We determined the number of cells within a circular area with a radius of 50 μm around the beads (7.85 × 10−3 mm2). For cell tracking, the circular observation area covered had a radius of 250 μm around the bead (0.196 mm2). Within this area, 10 randomly selected moving cells were tracked every second using the semi-automatic settings that allow track editing in case of obviously broken trajectories.

From the cell track data recorded for 600 s, parameters such as speed, angular orientation, and sum distance of cells were calculated as follows. All data analyses were done using the open-source statistical and graphic software R version 3.2.1 [23] and the package ggplot2 [24]. Calculated parameters were fitted either with a linear or a mixed model to account for the nested and longitudinal characteristic of the data. The optimum model was chosen based on the Akaike information criterion (AIC). Linear mixed effects (LME) modeling was done through the package nlme [25] and general additive mixed (GAMM) modeling through the package mgcv [26]. If needed, correlated data between independent variables were corrected by an autoregressive order 1 (AR-1) correlation structure. A constant variance function structure (varIdent) was also added to the model to correct residual spreads. We also inspected residual plots and verified that the model residuals are not deviating from normality and homoscedasticity. For each optimum model, a Wald test was also performed to determine the significance of the fitted estimates on each term.

Cell Counts

The raw cell count data was normalized according to standard Z-score calculation per treatment.

$$ Z=\left(X-\mu \right)/\sigma $$

where μ is mean, X is score, and σ is standard deviation. For the starting point to be normalized to 0, we subtracted the standardized cell count on each time point to the value at T = 0 s. A value of 0 indicates that the cell density is equal to the mean. A positive value indicates that the cell count is above the population mean while a negative value indicates a value below the mean. To compare pheromone and control treatments, normalized cell counts were fitted on a LME model with an interaction term between time and treatment as independent fixed variables and replicate ID as a random factor. Since successive measurements were done on each replicate per treatment, an AR-1 correlation structure was added to the model. A varIdent variance structure on treatment was also added.

Speed

A Wilcoxon rank sum test was performed to compare the overall mean cell speeds from diproline and control treatments. We also determined the effect of time along with treatment on averaged cell speeds in intervals of 30 s of all cell tracks via GAMM. The log +1 transformed speed was fitted to a smoothed factor interaction of time and treatment penalized with cubic regression splines and track ID (i.e., unique cell ID) as random factor. An AR-1 correlation structure between treatment and track ID and a varIdent variance structure on treatment were also added on the model. Additive models connect individual data points by smoothing and setting the population mean of each data set to zero. Computed p values indicate the significance of the smoothing terms and low p values indicate that the fitted model splines used in the function have low likelihood to be zero [26].

Angular Orientation

The angular orientation of the cells towards the gradient was computed by determining the sine angle from vectorized cell tracks. A first vector is defined by the coordinate position of the bead to a cell position at T = 0 s. A second vector points from this position to the coordinate of the cell at T = 1 s. The sine angle between these two defined vectors is determined. This procedure is repeated every second for the entire observation period [27]. A cell is moving towards the gradient if the calculated sine angle has a positive value and away from it in the case of a negative value. We performed LME on the mean sine angle over intervals of 30 s with the interaction term of treatment and time as independent variables and track ID as random factor. The same correlation and variance structure as with the speed model were used.

Sum Distance

Within the observation period, we determined every second the sum distance of cells by taking the sum of the calculated Euclidean distance of each cell coordinate relative to the center of the bead for each treatment. This gives insight into the net migration preference of the cell. The sum distance was fitted into a simple linear model with the interaction of time and treatment as explanatory variables.

Motility Parameters

Motility parameters were determined after a modified protocol described in Bondoc et al. [16]. We computed standard motility parameters using the pooled data tracks from three videos by fitting the root mean square (RMS) of the net distance as a function of time to Taylor’s equation [28]

$$ RMS={\left[2{v}^2\tau \left(t-\tau \left(1-{e}^{-t/\tau}\right)\right)\right]}^{0.5} $$

where v is the effective swimming speed, τ is the decorrelation time scale, and t is the time. From this, we were able to compute the decorrelation time (τ) and length (λ = vτ) scale, which corresponds to the time and distance, respectively, wherein there is directional persistence in swimming over the period 60 to 410 s. We also calculated the effective diffusivity of motility (D = v 2  × τ / n) to describe the spread of the cell tracks, and the encounter kernel (β = 4πRD) to determine the water volume screened by cells within the observation time. We also compared the mean net distance of cells used in the RMS fitting via Wilcoxon rank sum test.

Results

Diproline Release from the Beads

The total diproline administered to each bead was 26.7 ± 9.12 fmol and within the 600-s observation time, 7.40 ± 0.81 fmol diproline diffused out from each bead, corresponding to ∼30 % of the loaded amount. The calculated initial flux, i was 6.06−18 ± 6.66−19 mol s−1. This created a steady state gradient after 62.5 s with a concentration of ∼5 nM at the surface of the bead and decreasing to 0.30 nM on the edge of the observation area (Fig. 1).

Fig. 1
figure 1

Concentration gradient produced by the diproline-loaded bead around the observation area, calculated for a distribution within a radius from the bead of 250 μm in the two-dimensional biofilm. This steady state with a steep change of diproline concentration from the surface of the bead to the edge of the observation area is reached after 62.5 s. Shaded area represents the SD of the mean of three replicates

Full size image

Characterizing Motility Patterns in S. robusta MT+

Cells accumulate in the immediate vicinity of diproline-beads confirming chemoattraction towards the pheromone source while no attraction towards control-beads was observed (Fig. 2a; Movies S1 and S2) [8]. Increasing cell density within the observation area over time was found in the diproline treatments while the controls exhibited constant cell counts (LME; p < 0.0001, Fig. 2b; Table S1). The overall mean speed of cells under the influence of the diproline gradient was 69.2 ± 22.2 % higher compared to the control (Wilcoxon rank sum test, p < 0.0001; Fig. 3a). Increased speed in diproline treatments compared to control was observed from 90 s after addition of pheromone loaded beads with a maximum around ca. 300 s and lasting until 510 s. After this, a mean speed comparable to the control treatment was observed. Cells in the control had a constant mean speed over the entire observation time (GAMM: pcontrol = 0.04, pdiproline = <2−16; Fig. 3b; Table S2). In the statistical test GAMM, each group (control and diproline) has a p - value. It shows the connectedness of each mean speed and compare it over the population mean for that group. In this case, because of random movement of cells that slow down, e.g., when they bump another cell, a statistically relevant p - value is also observed for the control though it is lower compared to the diproline treatment. When the model fit was plotted (Fig. S1), the speed values in the control did not deviate from 0, which is the population mean. On the other hand, the mean speed in the diproline treatment deviated from 0 and even showed an increasing pattern during the first 300 s. This analysis documents the effect of the pheromone on cell speeds. Cells did not have a directed orientation towards the bead as the calculated sine angle did not differ between control and diproline treatments (LME, p = 0.38; Fig. 4; Table S3). However, the calculated sum distance traveled by all the tracked cells showed that there is a preferential overall movement over time towards diproline gradients (linear model, p = 1.147−11; Fig. 5; Table S4). This indicates that cells moved closer towards the diproline-bead over time compared to the control and that they bias their back and forth movement towards the pheromone gradient Fig. 6.

Fig. 2
figure 2

Chemoattraction of Seminavis robusta mating type MT+ towards a diproline-loaded bead. a MT+ cells approach and accumulate around the bead. Scale bar = 50 μm. b Significant increasing normalized cell density recorded every 30 s in the observation area (depicted as the circle in a) was observed over time when cells were exposed to diproline compared to control beads (Linear mixed effects (LME) modeling, p < 0.0001). A positive normalized cell count indicates that cell density is above the population mean while a negative one indicated values below the mean. Error bars indicate SE. The gray shaded area illustrates the LME model fit from 60 to 600 s with 95 % confidence intervals (Table S1)

Full size image
Fig. 3
figure 3

Speed of MT+ cells exposed in the diproline gradient for 600 s. a The overall mean speed in the diproline treatment is significantly higher compared to the control (Wilcoxon rank sum test: p < 0.0001). b The mean cell speed over intervals of 30 s from 60 to 600 s was taken and fitted using General additive mixed modeling (GAMM, Table S2, Fig. S1). Cells moved faster over time when exposed to the pheromone gradient while there was approximately constant speed for cells in the control treatment (GAMM: pcontrol = 0.04, pdiproline = <2−16). Error bars indicate SE

Full size image
Fig. 4
figure 4

The mean sine angle values of each cell’s coordinate position relative to the position of the bead averaged over intervals of 30 s. A cell moving towards the bead has a positive value while a cell moving away has a negative value. Cells in the diproline treatment did not differ from the cells exposed to a control bead (LME, p = 0.38). Shaded area represents the LME model fit from 60 to 600 s with 95 % confidence intervals (Table S3) and error bars indicate SE

Full size image
Fig. 5
figure 5

The sum distance of all cell’s coordinate position relative to the position of the bead was averaged over intervals of 30 s. The sum distance of cells in the diproline treatment significantly decreased over time, signifying that cells go nearer the bead compared to the control (linear model, p = 1.147−11). Shaded area shows the linear model fit from 60 to 600 s with 95 % confidence interval (Table S4) and error bars represent SE

Full size image
Fig. 6
figure 6

The difference in the root mean square (RMS) distance traveled between cells in the control and diproline treatment from 60 to 410 s. a Cells exposed to the pheromone have 2.4× higher mean net distance traveled (Wilcoxon rank sum test, p = 4.11−5). b This is also apparent in the increasing RMS over time

Full size image

Track analysis via RMS fitting showed that cells significantly changed their swimming behavior around diproline-beads. There is a 4800-fold increase in decorrelation time scale (τ) as well as a 150-fold increase in length scale (λ) (Table 1) in the diproline treatment. This higher directional persistence over time also increased the diffusivity (D) and encounter kernel (β) by 6-fold indicating that cells increased their encounter rates for diproline-loaded beads by searching larger areas along the gradient for the pheromone source. The combined increase in cell speed and higher persistence resulted in a higher net distance traveled by cells towards diproline compared to the control (Wilcoxon rank sum test: p = 4.11−5; Fig. 5a), which was apparent also in the increasing root mean square distance traveled from 60 s until 410 s (Fig. 5b).

Table 1 Motility parameters of MT+ cells exposed to control and diproline-loaded beads
Full size table

As a whole, we observed that cells moved faster with no angular orientation but with a preferred forward movement towards the pheromone source to accumulate around diproline-beads. Cells also changed their swimming behavior by increasing directional persistence, leading to more frequent encounters with the pheromone-releasing beads compared to control beads.

Discussion

Microphytobenthic communities in intertidal environments and estuaries are primarily composed of raphid pennate diatoms [29] that have the ability to colonize mobile sediments and other niches [30]. Raphe systems emerged after the evolution of gametangiogamy as a life history strategy [30]. Gametangiogamy ensures a high success of pairing and prevents inbreeding in crowded benthic communities [30, 31]. Crucial to this is pairing efficiency that can be supported by species-specific pheromone perception and motility [31]. Thus, the pheromone-mediated mating system of raphid pennates might be one of the factors that helped these microalgae to colonize benthic habitats and to diversify as the youngest lineage of diatoms (90 Myr old) [32].

This study identifies key features of the pheromone-mediated searching behavior of a benthic, raphid pennate diatom. Chemical analytics and organic synthesis allowed us to conduct the studies using a pure, synthetic pheromone. This allows excluding potential overlying effects of other signals as in previous studies where crude preparations were used [8, 33]. We determined that the concentration of diproline in the vicinity of pheromone loaded synthetic beads that elicit an attraction of the migrating S. robusta mating type (MT+) are in the nanomolar range. To characterize how migrating cells respond to diproline, we categorized their movement patterns as taxis or kinesis. These two behavioral responses, chemotaxis and chemokinesis, are two processes that have to be discussed separately, even if they are sometimes interchanged in the literature. Chemotaxis is defined as the directed orientation towards chemical gradients whereas chemokinesis is the change in motility parameters such as turning frequency and speed in response to the presence of signal molecules in sufficient concentrations [34, 35]. However, both behaviors can lead to accumulation or dispersal of cells in response to signal molecules [35]. The observed preferential migration towards the pheromone source is indicative for chemotaxis whereas changes in cell speed suggest chemokinesis. Successful mating involves pair formation of the opposite mating types. This process is supported by the increased encounter rates with the pheromone source through higher diffusivity (D) and encounter kernel (β) by MT+. Although the general concept of pheromone-mediated search behavior has been hypothesized or proven for different diatoms [8, 9, 36], attraction mechanisms seem to be species specific. When stationary female gametes of T. fasciculata are in close proximity of searching males, no guided attraction is observed, presumably due to a lack of chemical cue. As a consequence, males search randomly the area as the best strategy to maximize mate encounter [11]. On the other hand, mate-finding in male gametes of P. trainorii is pheromone-mediated and characterized by an initial random walk until the gamete is in close-range with the female. Gametes also change their shape from a globular form to an amoeboid one and form pseudopod thread-like extensions that can be used to catch mates [9]. In the case of S. robusta, diproline-mediated movement can be characterized as a biased random walk since there is no angular orientation but a clear orientation towards the pheromone source as exemplified by a preferential forward movement. Furthermore, a simultaneous tactic and kinetic mechanism is operative as a searching mechanism in MT+. Macroalgal gametes are known to employ either taxis or kinesis but, so far, a combination of both has not been observed [34, 35]. Marine bacteria, on the other hand, are known to employ simultaneous chemotaxis and chemokinesis in response to, e.g., amino acids [19] and coral hosts [37]. Since both marine bacteria and S. robusta move via a run-reverse or back and forth mechanism, it might be possible that this type of movement universally enables these cells to utilize taxis and kinesis simultaneously in response to stimulus [16, 37].

We recently reported that S. robusta does not only orient towards pheromone sources but that it also has the capability to sense and orient towards the inorganic nutrient dissolved silicic acid (dSi) [16]. Interestingly, the basic behavioral patterns are similar in response to both stimuli. However, when attraction patterns to dSi and diproline are compared, the response to a pheromone-gradient appears more efficient compared to the search for the nutrient. Since the methods employed on creating stimuli gradients and data analyses for the determination of the response towards dSi were similar to those in this study, we can directly compare the data sets (Table S2). In both cases, the S. robusta cell size used was below their species-specific sexual size threshold (SST, 51.6 ± 0.5 μm). When we compare the degree of change from control to dSi or diproline, a 1600-fold and 75-fold higher decorrelation time (τ) and length (λ) scales, respectively, are observed when cells are exposed to diproline compared to dSi. Moreover, cells have higher encounter rates to diproline-beads compared to dSi-beads, as indicated by a 3-fold increase on diffusivity (D) and encounter kernel (β). However, such a quantitative comparison might be somewhat oversimplified, since it has to be considered that both responses are dependent on external factors influencing the behavioral response. These factors can include the level of starvation in the case of dSi limitation as well as the cell size or cell cycle stage in the case of pheromone-mediated finding. Cells under the influence of a pheromone gradient have the need to directly encounter the source to facilitate mating. Cell contact with the pheromone-producer is needed to start mate pairing. In contrast, cells exhibiting a foraging behavior for dSi would benefit from a continuous movement within the gradient to avoid diffusion limitation during dSi-uptake.

Dusenbery [38] predicted a cellular diameter of more than 0.6 μm to be a pre-requisite for the successful employment of chemical gradients. Mating S. robusta with cellular dimensions of 12–20 μm are clearly above this threshold and thereby fulfill the theoretical requirements for efficient chemotaxis. Our observations show that directed motility strongly increases mate encounters in these diatoms. S. robusta was able to detect diproline concentrations of 0.30–5 nM within a radius of 250 μm. The active radius wherein pheromones can still be perceived has been estimated to 520 μm for aquatic gametes with a size in the micrometer range [39]. At this distance from the pheromone-loaded beads, the calculated steady state concentration is 0.24 nM, a concentration in the range eliciting an attraction response in our experiments (Fig. 1). The threshold concentration for planktonic flagellated brown algal pheromone reception is even lower, 1–1000 pM. This increased sensitivity could be seen as an adaptation to the wider area covered by their flagellated gametes [34, 35, 40, 41]. Initial evidence suggests that diproline is readily degraded after its release [8], which can be seen as a mechanism to avoid misleading information towards aged pheromone sources and to further increase the steepness of the gradient towards the attracting mating type. This additional aspect is not covered in our study using artificial pheromone sources. We also do not take into account that sex-inducing pheromones might synchronize the production of diproline within several cells of the attracting mating type [12]. This might result in additional patchiness of the pheromone landscape encountered by the searching partner. Interestingly, diproline is active in a concentration range over more than three orders of magnitude [8]. Thus, S. robusta can respond to diproline over a several hundred micrometer distance to the producer and still perceives the gradient in its immediate vicinity. Thus, only one signal is required for the entire finding process. Attracted cells can orient over this distance along the pheromone gradient by increasing speed and biasing a forward movement. No other signals are required during this mate finding process.

Conclusion

Using semi-automated cell tracking as well as statistical modeling, we were able to show that the migrating mating type (MT+) of S. robusta is orienting towards a source of the pure synthetic pheromone diproline via a biased random walk mediated through simultaneous chemotaxis and chemokinesis. This could be a characteristic pattern for cells exhibiting a back and forth motility. Cells showed a stronger response to the pheromone diproline compared to the searching behavior in a gradient of the limiting nutrient dSi. However, cells might prioritize one response above another when they are in different need of resources or reproduction. Though sexual reproduction itself is costly, the efficient search behavior explored in this study ensures that this strategy is effective.