Introduction

Diatoms play an important role in the phytoplankton community in most marine and freshwater ecosystems, and they are particularly important for the new production, e.g. during the spring bloom in temperate areas. In contrast to most other groups of phytoplankton, diatoms need dissolved silicate (DSi) to form their siliceous cell wall (frustules). The metabolic cost of using DSi as cell wall material is far less than other building materials, and this feature has been suggested as one of the leading factors for the global success of the diatoms (Raven 1983; Smetacek 1998; Martin-Jézéquel et al. 2000). This can also be a disadvantage, however, since they require this additional nutrient that competing phytoplankton groups do not need. Indeed, loss of competitive advantage for diatoms due to DSi limitation has been demonstrated to take place both experimentally (Egge and Aksnes 1992) and in nature (Nelson and Dortch 1996).

Concentrations of DSi in the world’s oceans have on a geological time scale been much higher than current levels, and this decrease could be related to diatom populations (Canfield et al. 2005). Sedimentation of diatoms transports biogenic silicate (BSi) to the sea floor, which may function as a sink for some of this BSi. If the burial rate of BSi is larger than the input of DSi, then the DSi concentration will decrease over time. Anthropogenic eutrophication has been hypothesized to speed up this process as the increased input of nitrogen (N) and phosphorus (P), increases the diatom production, which in turn, increases the transport of BSi to the sea floor (Schelske and Stoermer 1971; Conley et al. 1993).

Another process that influences the amount of bioavailable silicate is the flux of terrestrial DSi into the sea. The majority of terrestrial DSi entering marine ecosystems originate from chemical weathering of rock. There is little anthropogenic input of DSi into aquatic systems compared with the input of N and P. Damming and other artificial reconstruction of rivers are known to reduce the amount of DSi entering the aquatic system (Humborg et al. 1997). These processes are caused by either silicate retention within river systems as a result of damming, or reduce the release of silicate from the soils due to diminished contact with water, such as when waterways are straightened. Either way, both of these processes reduce the amount of DSi entering aquatic systems.

In the Baltic Sea, the onset of the diatom spring bloom is triggered by the increasing irradiance and the stability of the water column resulting from increased freshwater discharge (Stipa 2002). The spring bloom, which is the annual productivity maximum (Heiskanen 1998), starts in the south and moves northwards (Kahru and Nõmmann 1990). In the Baltic Proper and Gulf of Finland, the spring bloom is terminated when nitrogen is depleted (Tamminen 1995; Höglander et al. 2004; Tamminen and Andersen 2007). After the spring bloom, much of the diatom production sediments quickly out of the water column; this sedimentation event is one of the main sources of organic matter to the sea floor in the Baltic Sea (Heiskanen 1993, 1998; Olli and Heiskanen 1999).

The environment in the Baltic Sea has over the last century been affected by anthropogenic loadings of inorganic N and P, and the first signs of eutrophication started to appear in the 1960s (Schernewski and Neumann 2005). This excess loading of N and P has increased the N:Si and P:Si ratios. Along with this, DSi concentrations have also decreased in absolute numbers over the past 30 years in the central Baltic Sea (Wulff and Rahm 1988; Papush and Danielsson 2006). Damming of rivers and burial of BSi in the sediment have been suggested to be the main causes for the decreased DSi concentration (Conley et al. 1993; Humborg et al. 2000).

The increased N:Si ratio associated with eutrophication has been discussed as one possible scenario that would favor non-siliceous phytoplankton production over that of diatoms (Officer and Ryther 1980; Smayda 1990; Egge and Aksnes 1992), as this change in ratio may result in a shift from N to DSi limitation of diatom growth (Gilpin et al. 2004). DIN:DSi ratios of >2, or absolute concentrations of <2 μmol l−1 DSi, have been suggested as limiting for diatoms (Egge and Aksnes 1992; Gilpin et al. 2004). However, diatoms have been shown to be highly capable of acclimating to DSi stress and can sustain high growth rates despite low external DSi concentrations (Olsen and Paasche 1986; Brzezinski et al. 1990).

In the Baltic Sea, it is not yet known how the spring diatoms respond to the increased N:DSi ratio, or whether the observed decreasing trend in DSi availability may indeed result in a reduced competitive ability in diatoms due to DSi limitation. At present, there is evidence suggesting that recruitment and build-up of biomass at the initial phases of the spring bloom determines which species dominates the bloom peak (Kremp et al. 2008). However, if DSi continues to decrease, it can be expected that diatom growth and their capacity to form viable resting spores could be negatively affected at some point before the peak of the spring bloom.

The ecological consequences of potential diatom silicate limitation in the Baltic Sea are poorly known but could potentially have cascading effects through both the pelagic and the benthic food webs. Ecological theory states that the best competitor for a given nutrient is the species which can maintain a certain growth rate at the lowest external nutrient concentration (Sterner and Elser 2002). Knowing the species-specific silicate uptake kinetics could thus provide information about the species’ growth ability in an environment with low DSi concentrations. The aim of this study was to estimate DSi uptake kinetic parameters for dominant Baltic Sea spring diatom species. We were interested in maximum growth rates (μmax), maximum DSi uptake (V m ), half-saturation constants (K s ) and the concentrations of DSi where growth of the species decreases to zero. In this paper, we present estimates of silicate uptake properties for 8 spring diatom species from the Baltic Sea, derived from a time series based model of DSi depletion and biomass accumulation.

Materials and methods

Culture conditions

Non-axenic monocultures of Chaetoceros wighamii Brightwell, Diatoma tenuis Agardh, Melosira arctica (Ehrenberg) Dickie, Nitzschia frigida Grunow, Pauliella taeniata (Grunow) Round and Basson, Skeletonema costatum (Greville) Cleve, Thalassiosira baltica (Grunow) Ostenfeld and T. levanderi van Goor were isolated from natural communities off the southwest coast of Finland. Cultures in early stationary growth phase were inoculated into 250-ml tissue culture flasks using a modified f/2 medium (Guillard and Ryther 1962) produced from natural seawater from the Baltic Sea (salinity~6 PSU). The stoichiometric N:P ratio was adjusted according to the Redfield ratio, and the DSi concentration was reduced sufficiently to ensure that the cultures were not limited by any other nutrient. The bottles were submerged in a water bath to maintain the temperature at a constant 4°C. Light was provided on two sides with 16:8 light/dark cycle at ~100 μmol photons m−2 s−1. Aeration was applied in order to keep cells in suspension and to provide sufficient CO2 to prevent excessive pH changes. Each species was grown in three separate bottles, which were used as replicates in the data analysis.

The DSi uptake measurements were performed at two levels of nutrient addition. For the first set, we used 20 μmol l−1 DSi for all species and measured DSi levels three times a week. Because several species (T. baltica, T. levanderi, C. wighamii, C. gracilis and D. tenuis) had such a rapid uptake at this level, we repeated the measurements for these species. In the second set of measurements, ~40 μmol l−1 DSi was spiked, and the concentration was determined daily.

Samples for cell enumeration were taken at the time of the DSi measurements. Cells were enumerated under an inverted microscope (Lomo, St. Petersburg, Russia), using a 1-ml counting chamber (Sedgewick Rafter). We determined the particulate nutrients twice: once during the initial exponential growth phase and once when the stationary phase was reached. The BSi cell−1 relationship during exponential growth was used as a proxy for the BSi development during growth, for example when transforming cell counts to BSi estimates. We also measured dissolved inorganic N and P at the time of sampling to preclude the possibility of N or P being depleted to growth-limiting levels.

Nutrient determination

Inorganic nutrients were measured according to Grasshoff et al. (1983). For particulate carbon (POC) and nitrogen (PON), we used pretreated (acid-washed and combusted) GF/F filters (Whatman, Middlesex, UK). POC and PON filters were allowed to dry and stored at room temperature (20°C) until C and N concentrations were determined using a mass spectrometer (Europa Scientific ANCA-MS 20-20, SerCon, Cheshire, UK). Polycarbonate filters (Poretics, GE Osmonics, Minnetonka, MN, USA) with a pore size of 0.8 μm were used for BSi analysis, using the method described by Krausse et al. (1983). Briefly, the polycarbonate filters were placed directly into digestion vessels (plastic scintillation vials) and leached with 0.2 M NaOH in a boiling water bath followed by neutralization with 1 M HCl. The resulting solution was then analyzed for dissolved silicate (DSi) using standard colorimetric procedures.

Data analysis

The presented growth rates were based on cell counts and calculated according to the following equation:

$$ \mu = {\frac{{\ln (N_{t} ) - \ln (N_{0} )}}{t}} $$
(1)

where μ is the growth rate (day−1), N t is concentration of cells at time t in days, N 0 is the initial concentration of cells at time 0, and t is the time interval between t 0 and t in days.

Silicate uptake rates were calculated from the time series measurements of the depletion of (DSi) and increase in BSi, calculated from the cell counts and average BSi content cell−1. In short, we used a modified Monod equation with an introduction of the S min parameter, which is the DSi concentration where nutrient uptake stops. Since the system is closed and both state variables are expressed in the same units (μmol Si l−1), the dynamical mass balance equations become very simple: any uptake leading to a decrease in DSi must lead to a corresponding increase in BSi.

$$ {{\text{dDSi}} \mathord{\left/ {\vphantom {{\text{dDSi}} {{\text{d}}t}}} \right. \kern-\nulldelimiterspace} {{\text{d}}t}} = - V\left( {\text{DSi}} \right)\;{\text{BSi}} $$
(2)
$$ {{\text{dBSi}} \mathord{\left/ {\vphantom {{\text{dBSi}} {{\text{d}}t}}} \right. \kern-\nulldelimiterspace} {{\text{d}}t}} = V\left( {\text{DSi}} \right)\;{\text{BSi}} $$
(3)

where V is the specific DSi uptake rate (day−1). Uptake is related to DSi concentration as:

$$ V = V_{m} {\frac{{{\text{DSi}} - S_{\min } }}{{K_{S} + \left( {{\text{DSi}} - S_{\min } } \right)}}} $$
(4)

where V m is the maximal specific uptake rate (day−1), K s is the half-saturation parameter (μmol l−1), and S min is the threshold DSi concentration for positive net uptake (Fig. 1). As the non-linear differential equations relating DSi depletion (Eq. 2) to BSi accumulation (Eq. 3) do not have a closed form analytical solution, they were solved numerically with the Runge–Kutta method using the ‘ode23s’ function in Matlab (The MathWorks, Natick, Ma, USA). Unknown model parameters (V m , K s , S min) were estimated by least-squares fitting model solutions to observations using a derivative-free optimization routine (Matlab ‘fminsearch’). Confidence intervals of parameter estimates were constructed by a Monte Carlo method described by Silvert (1979).

Fig. 1
figure 1

An example of the modified Michaelis–Menten uptake kinetics of dissolved silicate (DSi) derived from the model. The new S min parameter is the DSi concentration where positive net uptake of DSi stops. The potential significance of this parameter is discussed in the text. If there are physiological barriers for uptake of DSi below the S min concentration, the actual K s for growth would be the sum of the modeled S min and K s (K s  + S min)

Full size image

The two parameters of the Monod function (Eq. 4) control different phases of the depletion trajectory, with V m determining the saturated growth at high concentrations and V m /K s controlling the nutrient limited growth at low concentrations. The two parameters can only be successfully estimated if the uptake actually is close to saturated at the start of the experiment (i.e. if initial DSi is substantially larger than K s ). This will be equivalent to having a sharp inflection point in the depletion curve, representing the transition from nutrient-saturated growth to nutrient-limited growth. A smooth depletion curve with a weak inflection point probably reflects that the initial DSi concentration was too low compared to the K s of the species, such that only affinity (V m /K s ), but not K s by itself, can be precisely identified from the data. A graphical presentation of the relationship between the inflection point and K s is presented in Fig. 2.

Fig. 2
figure 2

Scenarios of the model fit showing different transition from nutrient-saturated growth to nutrient-limited growth and its effect on the modeled K s . The left panel shows the nutrient draw down (biomass increase has for simplicity been left out) and the right panel the corresponding modeled nutrient uptake kinetics. The critical part for determining K s is the inflection point in the depletion curve, and consequently, lack of data in this region gives high uncertainty of the estimated K s

Full size image

We introduced upper and lower parameter boundaries to constrain the fitting algorithm from producing arbitrary high V m and K s estimates under such conditions. The high parameter boundary (HPB) for K s was set to 10 μmol DSi l−1, which is in the high end of published K s values for DSi uptake in diatoms (Martin-Jézéquel et al. 2000), and the low parameter boundary (LPB) was set to 0.01 μmol DSi l−1, below that of published K s values.

In order to avoid some of the concerns with setting parameter boundaries in the model, we also present the terms: maximum observed uptake (V mo ) and observed half-saturation constant (K so ). These parameters are the observed maximum uptake and half-saturation constant for the concentration of DSi used, i.e. no extrapolation to higher DSi concentrations. In other words, V mo  = V at the starting DSi concentration (Table 1), and K so is the DSi concentration where the depletion rate is equal to V mo /2. An example using different HPB for K s and the effect on the respective nutrient uptake kinetics is presented in Fig. 3.

Table 1 Parameter estimates of dissolved silicate (DSi) uptake kinetics for Baltic Sea spring diatoms
Full size table
Fig. 3
figure 3

Scenarios of the model fit using high parameter boundaries (HPB) for K s of 10 (solid line) and 100 (stippled line). The model fit to concentration of dissolved silicate (DSi) (filled points) and biogenic silicate (BSi) (open points) in left panel, and the respective uptake kinetics of dissolved silicate (DSi), using the HPB for K s of 10 and 100 (right panel). V m is modeled maximum uptake, V mo is observed maximum uptake within the DSi concentration range, K s is at HPB, and K so is the DSi concentration where the depletion rate is equal to V mo /2

Full size image

Results

DSi uptake patterns and growth kinetics

The increase in cell density generally followed a sigmoidal growth curve, which differs from a logistic model by not being antisymmetrical around the inflection point. The modeled uptake parameters are presented in Table 1 and Figs. 4 and 5. The highest observed uptake rates, V mo , were found in D. tenuis and T. levanderi. The maximum uptake rates were somewhat reflected in the maximum growth rate as the species with low V m also had low μmax (Table 2). Most of the species with high V m also had high μmax with the exception of D. tenuis, which had a relatively high V m but a low μmax. The highest K s values were found in D. tenuis and M. arctica, while the rest of the species had K s values < 2 μmol DSi l−1.

Fig. 4
figure 4

Modeled dissolved silicate (DSi) uptake kinetics of Melsoira arctica, Nitzchia frigida, Pauliella taeniata, Skeletonema costatum and Thalassiosira levanderi grown in 100 μmol photons m−2 s−1, 16:8 light/dark cycle. Left panels show the DSi draw down (filled points) and biogenic silicate (BSi) (open points), and the model fit (solid lines). Right panel shows the respective DSi uptake kinetics. Dotted lines represent the 90% confidence intervals

Full size image
Fig. 5
figure 5

Modeled dissolved silicate (DSi) uptake kinetics of Diatoma tenuis, Chaetoceros wighamii and Thalassiosira baltica grown in 100 μmol photons m−2 s−1, 16:8 light/dark cycle. All other notations are as in Fig. 4

Full size image
Table 2 Growth rate and silicon stoichiometry in Baltic Sea diatoms listed according to maximum growth rate, which is based on cell enumeration in culture
Full size table

Growth of all cultures with excess N and P ceased before DSi was depleted, with the residual DSi (S min) ranging from 1.7 to 5.6 μmol DSi l−1 (Table 1). This should not be considered to be the result of any analytical/methodological interference between P and Si measurements because in previous experiments with natural plankton assemblages, using the same method, results have been as low as 0.5 μmol DSi l−1 (Kremp et al. 2008).

Si stoichiometry

The average initial C: Si ratio during exponential growth was 20 (Table 2) for the examined species. There was no apparent change in size of the cells during the measurement period, suggesting that the change in C:Si ratio was caused by changes in degree of silicification. M. arctica, D. tenuis and N. frigida had the highest degree of silicification (lowest C:Si), while C. wighamii, P. taeniata and T. baltica were the least silicified species as reflected by the C:Si ratio. There was a general trend of increasing C:Si ratio when the cultures reached the stationary phase in the growth curve, but there were large differences between the species. For example, P. taeniata and D. tenuis had only a slight change in the C:Si ratio between exponential and stationary growth phase, while the C:Si ratio in M. arctica increased by 390%.

Discussion

DSi uptake kinetics and stoichiometry

The maximum uptake and half-saturation constants for DSi were highly variable but were generally within the range commonly observed in diatoms (Martin-Jézéquel et al. 2000). The modeled K s values were determined by the smoothness of the curve at the transition from saturated uptake at maximal growth rate to Si-limited growth. Consequently, the K s parameter estimates are very sensitive to lack of data at this transition point, making K s the parameter with the greatest uncertainty (Figs. 1 and 2). Changing the parameter boundaries of K s had relatively little effect on V m . For example, tuning the maximum ‘allowed’ K s values from 10 down to 5 in the model for M. arctica (which had parameter estimates at the upper parameter boundary for K s ) reduced the V m estimate by approximately 10%. This would not lead to changes in the general conclusions about maximum uptake for any of the other species; however, a better estimate of maximum uptake for M. arctica is probably the observed V mo , which does not extrapolate the model results outside the range of DSi concentrations used.

Residual DSi that has not been taken up by diatoms in situations where growth is not limited by other nutrients has been reported in earlier culture work (Paasche 1973; Kudo 2003), and Paasche (1973) suggested this represents non-reactive silicate. An alternative explanation would be dissolution of BSi back to the DSi pool. However, since dissolution of BSi is a relatively slow process at the low temperature we were keeping the cultures (Kamatani 1982), dissolution of BSi is probably not the reason for the observed S min. It therefore appears that there is some unexplained factor that shuts down the uptake of DSi at a relatively high concentration (>1 μM), and this concentration seems to be variable even for the same algal cultures (unpublished data). However, if the remaining silicate, S min, is unavailable for the diatom cells because of physiological limitations, the actual K s value for growth is the sum of our modeled K s and S min values.

For the species where the C:Si and N:Si ratios increased from exponential to stationary growth phase, the ratios reflected accumulation of C and N compared with Si. However, the C:Si and N:Si ratios changed relatively little (<30%) between exponential growth and stationary phase for D. tenuis, N. frigida, T. baltica and P. taeniata; in particular D. tenuis and N. frigida had a relatively low C:Si ratio, which might be an indication of these species having a high Si requirement. The stoichiometric relationships were determined in early stationary phase, and more measurements are needed in order to strengthen this hypothesis. M. arctica also had a low C:Si ratio during exponential growth, for example before DSi depletion, but the C:Si and N:Si ratios increased considerably after silicate uptake had ceased. The result suggests that M. arctica has a more flexible C, N:Si ratio compared with D. tenuis and N. frigida.

Interestingly, the much used molar N:Si ratio of 1 (Brzezinski 1985; Levasseur and Therriault 1987), as a standard for diatom silicate uptake, was shown to be too low for the species we investigated, with the exception of N. frigida. All the other species had N:Si ratio above 1 such that an uptake ratio of 2–3 seems to be more appropriate (Table 2), which is also similar to the reported N: Si ratio for Chaetoceros socialis (Kudo 2003). This pattern of higher than usual ratios was also seen in the C:Si ratio of 12–25 during exponential growth phase, which was higher than the average (~8) found in 27 different diatoms by Brzezinski (1985). Also worth noting is that we only evaluated the nutrient stoichiometry (C:Si and N:Si ratio), which may not reflect the degree of silicification using other criteria, such as relating it to cell volume. For example, T. baltica has relatively low biomass yield to DSi uptake compared with C. wighamii and S. costatum, indicating a higher degree of silicification (Olli et al. 2008), but this was not reflected in our stoichiometric data.

The BSi estimates (Figs. 4 and 5) are based on cell counts, and the increase in the BSi pool is directly coupled with the depletion of the DSi pool. Thus, increases in BSi concentration, higher than the decrease in DSi concentration, can only be the result of decreasing BSi concentration cell−1. Most of the species appeared to become less silicified after depletion of DSi with the exception of N. frigida and P. taeniata.

Evaluation of model

The approach we have taken in the present model is to follow the nutrient uptake from exponential growth phase into the stationary growth phase, when the nutrient in question becomes limiting, and not after a nutrient spike such as in a perturbation experiment (Caperon and Meyer 1972). By following the cultures from a high nutrient concentration through to depletion of nutrients, the natural situation of a spring bloom is mimicked. Aksnes and Egge (1991) presented a theoretical model of nutrient uptake in phytoplankton showing that the much used Michaelis–Menten function of nutrient uptake assumes several properties to remain constant: (1) the number of nutrient uptake sites, (2) handling time, (3) site area and (4) mass transfer coefficient. Any acclimation in nutrient uptake would thus interfere with the direct Michaelis–Menten type relationship between uptake and external nutrient concentration. Assuming that nutrient acclimation takes place, the nutrient uptake kinetics will change as a function of the external nutrient concentration. The model output presented here describes nutrient uptake kinetics from nutrient replete to nutrient deplete conditions, which is not identical to nutrient uptake kinetics at any particular nutrient concentration, but rather a cumulative (or integrative) presentation of kinetics of nutrient uptake throughout the period examined. This closely resembles some techniques used for measuring nutrient uptake in terrestrial plant roots (Claassen and Barber 1974).

A traditional nutrient kinetic assay is thus a snapshot of the nutrient acclimation at the time of the measurement. This is useful, for example, in describing the nutrient kinetics of algal cells exposed to a nutrient patch of a limiting nutrient but will not reflect nutrient uptake kinetics under conditions where nutrient acclimation takes place. For instance, nutrient acclimation may take place under continuously changing nutrient concentration, such as during the spring bloom in temperate areas. An alternative to the model described here would be to perform successive traditional nutrient kinetic assays throughout the depletion of the nutrient in question.

Ecological niches

Grime’s C-S-R paradigm (Grime 1974, 1977) was originally developed for terrestrial plants, but it has since been adapted to phytoplankton ecology (Reynolds 1988). The basic concept is a three-way trade-off between adaptation to high productive habitats (C), high stress habitats such as low nutrients (S), and high disturbance habitats, such as deep mixing (R). During the spring bloom in the Baltic Sea, C. wighamii, C. holsaticus, S. costatum, P. taeniata and T. baltica are often the dominating diatom species (Höglander et al. 2004; Tamelander and Heiskanen 2004). All of these species are pelagic, chain-forming, and they have good abilities for growth during the spring bloom.

The spring bloom in the Baltic Sea is a period when phytoplankton growth initially is limited by light until increasing irradiance, and initial stratification of the water column triggers a highly productive period. Thus, all of the above-mentioned species exhibit ecological strategies toward the R–C axis of the C-S-R triangle. C. wighamii had the highest maximum growth rate and had a high V m (Figs. 4 and 5), thus appearing to be the best competitor in high nutrient conditions, and consequently furthest toward the C corner in the C-S-R triangle. Among the common spring diatoms, T. baltica had the lowest maximum growth rate and a relatively low V m , indicating an opposing ecological strategy. P. taeniata and S. costatum were intermediate in terms of V m and maximum growth rate. S. costatum has often been reported to be relatively more abundant after the main peak of the spring bloom (Tamelander and Heiskanen 2004). In our measurements, S. costatum appeared to have very low K s values; this result and the observations from nature indicate an ecological niche further toward the C corner when compared with the rest of the species. Our results suggest that P. taeniata is situated somewhere between C. wighamii and T. baltica within the C–S-R triangle. The results are presented in a conceptual C-S-R triangle in Fig. 6.

Fig. 6
figure 6

Some of the main vernal diatoms in the Baltic Sea placed in a conceptual C-S-R triangle: Chaetoceros wighamii, Pauliella taeniata, Thalassiosira baltica, T. levanderi and Skeletonema costatum. The positions are only relative, see text for discussion. The corners represent high productive (C), high stress (S) and high disturbance (R) habitats. The arrow indicates the general succession pattern within the C-S-R triangle (Margalef 1978; Reynolds 1988; Wilson and Lee 2000)

Full size image

Both N. frigida and D. tenuis have a low C:Si ratio (Table 2), but our results show very different uptake characteristics of DSi between the two; D. tenuis had a relatively high V m and K s , while N frigida had low V m and K s . This indicates that D. tenuis is the most vulnerable, of the examined species, to low concentrations of DSi. Of the most commonly occurring species, P. taeniata appears to be the most rigid in terms of silicate requirements, with little variation in C:Si ratio and no notable change in BSi cell−1 during DSi limitation.