Diatom populations in an upwelling environment decrease silica content to avoid growth limitation

Summary

A mix of adaptive strategies enable diatoms to sustain rapid growth in dynamic ocean regions, making diatoms one of the most productive primary producers in the world. We illustrate one such strategy off coastal California that facilitates continued, high, cell division rates despite silicic acid stress. Using a fluorescent dye to measure single-cell diatom silica production rates, silicification (silica per unit area) and growth rates we show diatoms decrease silicification and maintain growth rate when silicon concentration limits silica production rates. While this physiological response to silicon stress was similar across taxa, in situ silicic acid concentration limited silica production rates by varying degrees for taxa within the same community. Despite this variability among taxa, silicon stress did not alter the contribution of specific taxa to total community silica production or to community composition. Maintenance of division rate at the expense of frustule thickness decreases cell density which could affect regional biogeochemical cycles. The reduction in frustule silicification also creates an ecological tradeoff: thinner frustules increase susceptibility to predation but reducing Si quotas maximizes cell abundance for a given pulse of silicic acid, thereby favoring a larger eventual population size which facilitates diatom persistence in habitats with pulsed resource supplies.

Introduction

Diatoms, silicified unicellular phytoplankton, are among the most ecologically diverse and successful organisms in the ocean (Kooistra et al., 2007; Malviya et al., 2016), generating ~40% of global marine primary production (Nelson et al., 1995). Diatoms thrive in dynamic ocean environments (Nelson et al., 1995) where physical mixing brings pulses of nutrients to the sunlit-surface ocean. In these regions, diatoms create dense blooms using a complex mix of r-selected strategies—which prioritize rapid and sustained cell division. This is contrasted by organisms that use K-selected strategies which prioritize persistence in environments near carrying capacity, e.g. dinoflagellates (Margalef, 1978). Among the r-selected strategies employed by diatoms are thought to be physiological adjustments for maintaining growth with rapidly decreasing nutrient availability (e.g. Guillard et al., 1973; Paasche, 1975).

The capacity of phytoplankton cells to maintain division rates as resources are depleted influences not only succession among classes of phytoplankton (Egge and Aksnes, 1992) but also succession within a class due to interspecies differences in nutrient use (Tilman, 1977; Kilham and Tilman, 1979). Diatoms are unique among phytoplankton in their obligate requirement for silicic acid, Si(OH)₄, which is an essential component of the diatom siliceous cell wall or frustule. In all ocean provinces examined to date, including highly productive coastal (Nelson and Dortch, 1996) and polar regions (Nelson and Tréguer, 1992; Nelson et al., 2001), silicic acid concentration, [Si(OH)₄], episodically or chronically limits diatom silica production rates (hereafter referred to as ‘silicon stress’). The unique requirement of silicon and the pervasiveness of silicon stress has led to the conclusion that [Si(OH)₄] contributes to phytoplankton succession (Egge and Aksnes, 1992; Lochte et al., 1993; Sieracki et al., 1993; Merico et al., 2004). However, culture experiments have found that diatoms are capable of decoupling silica production rates and cell division rates when Si(OH)₄ becomes scarce—implying that cells decrease their silicon requirement to delay the onset of growth limitation (Guillard et al., 1973; Paasche, 1973a; Brzezinski et al., 1990).

The physiological progression of diatoms first reducing frustule silica content and then decreasing division rate was described in culture experiments 45 years ago (Paasche, 1973a). The decoupled decrease in rates of silicon uptake (determined by net drawdown of Si(OH)₄) and division (determined from cell abundance) implied a decrease in silicification and frustule thinning. Subsequent laboratory studies documented the same hierarchical response to silicon stress in a number of species (Martin-Jézéquel et al., 2000 and references therein) suggesting decreasing silicification could be a common strategy for avoiding silicon growth limitation in the ocean. Cultured strains were shown to decrease their frustule silicon content three- to four-fold under severe silicon stress without significant decreases in growth rate (Guillard et al., 1973; Paasche, 1975; Brzezinski et al., 1990). This implies that in natural diatom populations [Si(OH)₄] would need to limit silica production rates to 25–33% of maximum before division rates would slow significantly. Given that diatom silica production follows Michaelis-Menten kinetics (Sup. Info.) with a median half-saturation constant of ~2 μM Si(OH)₄ (Martin-Jézéquel et al., 2000), the plasticity in silica content facilitates near maximum division rates until Si(OH)₄ decreases below 1 μM. Directly assessing diatom-specific growth limitation and changes in silicification (Si μm⁻²), has largely eluded the field due to methodological constraints. Thus, whether sacrificing silicification to maintain division rate is a common adaptive strategy that influences population growth and succession within diverse natural diatom assemblages remains to be directly evaluated.

If the sequence of physiological responses to Si stress observed in culture is widely applicable to natural assemblages it would point to an r-selected mechanism whereby diatoms delay silicon growth limitation to maximize ultimate population size for a given pulse of Si(OH)₄. It would also suggest a shifting ecological role for diatoms based on the degree of silicon stress. When the severity of silicon stress is sufficient to decrease diatom division rates, interspecific differences in silicon kinetics would influence both diatom-species (Paasche, 1973b; Tilman, 1977; Kilham and Tilman, 1979) and phytoplankton succession (Egge and Aksnes, 1992); these changes affect higher order consumers (Finney et al., 2002), and the cycling of carbon (Tréguer et al., 2017). Mild silicon stress, that stunts silica production rates and decreases silicification but does not alter division rate, would exert little bottom-up pressure on phytoplankton community structure. However, it could indirectly affect top-down pressure on the community because more lightly silicified cells are more susceptible to predation by a variety of zooplankton groups (Assmy et al., 2013; Liu et al., 2016; Zhang et al., 2017) and changes in predation rate could affect the fate of diatom biomass and community structure. Additionally, mild silicon stress could have biogeochemical implications as decreased silicification reduces cell density, such cells would be more likely to remineralize and dissolve at shallow depths compared to heavily silicified individuals (Passow et al., 2011).

Here, we use a recently developed method to directly quantify taxon-specific silicon stress in a coastal upwelling system. We examine the extent to which cells alter silicification and division rate with the increase of silicon stress that follows an initial nutrient injection. The results reveal high interspecific variation in the susceptibility of taxa to silicon stress but a shared physiological response to intensifying silicon stress.

Results

Environmental setting

Samples for incubation experiments were collected during the IrnBru (MV1405) cruise aboard the R/V Melville in July 2014. Water was collected during two transects off the California coast; each generally followed plumes of upwelled water that extended perpendicularly away from the coast (Fig. 1a) with the exceptions of station 2 in the northern transect and the transition to more oligotrophic water in station 5–7 along the southern transect. The recently upwelled water nearshore was cold with a high concentration of macronutrients, while the offshore water was warmer with lower nutrient concentrations. A more thorough description of the environmental conditions can be found in McNair et al. (2018). [Si(OH)₄] concentration was highly correlated with [NO₃⁻] (R = 0.84, p < 0.001) and [PO₄^3-] (R = 0.96, p < 0.001) with all three nutrients decreasing in concentration away from shore (McNair et al., 2018). Along the northern transect [Si(OH)₄] ranged from 21.2 μM near shore to 2.0 μM offshore; along the southern transect [Si(OH)₄] ranged from 15.4 μM to 2.3 μM near shore to offshore, respectively (Fig. 1b).

Figure 1:

(a) MODIS 8-day composite of sea surface temperature (SST) for the northern and southern transect with stations denoted as circles, station 1 was closest to shore and 7 was farthest offshore. The black line at 40° latitude breaks the 8-day composite centered on July 20, 2014 for the northern transect from the composite centered on July 7, 2014 for the southern transect. Map adapted from McNair et al., (2018). (b) Relative degree of silicon stress (V_+Si/V_amb) along the northern transect (top panel) and along the southern transect (bottom panel). White circles denote total community production as measured with ³²Si tracer, black circles denote average, weighted, taxa-specific silica production from the fraction of the community measured with PDMPO. Error bars represent standard error. Silicic acid concentration (x’s) in μM.

Community-level changes in silica production

The degree of silicon stress was characterized by comparing the rate of silica production (V) in a +Si treatment, where sufficient Si(OH)₄ was added to saturate silica production rates (V_+Si), to the rate of silica production under ambient nutrient condition (V_amb), which was not amended with Si(OH)₄. The influence of community composition and environmental conditions on V_amb can be found in McNair et al. (2018). The ratio V_+Si/V_amb quantifies the degree to which silica production was limited by [Si(OH)₄] under ambient conditions; larger V_+Si/V_amb indicates more severe silicon stress and V_+Si/V_amb close to one indicates that the rate of silica production was insensitive to the additional Si(OH)₄ and thus was near maximum under ambient [Si(OH)₄].

Despite the large range of [Si(OH)₄] across both transects the bulk measurement of diatom community silica production from ³²Si measurements, showed the stress (V_+Si/V_amb) of the total diatom community was relatively constant among stations (Fig. 1b). The correlation between [Si(OH)₄] and V_+Si/V_amb in the northern transect was low and not significant (R = −0.46, p = 0.36). The correlation along the southern transect was positive and significant (R = 0.83, p = 0.05). On average, in the northern transect, community silica production rates increased by 43 ± 20% (S.E.) with the addition of 19 μM Si(OH)₄ (Fig. 1b). Along the southern transect, community silica production rates were similar, on average, in both treatments and only increased by 10 ± 15% (S.E.) with added Si(OH)₄.

Taxon-specific measures of silicon stress (measured with PDMPO and microscopy) were combined and averaged for each station then compared to the bulk-measurement (³²Si) of total community silicon stress (Fig. 1b). The averages were weighted by a taxon’s contribution to silica production. Limited time and resources precludes making single-cell measurements for all taxa within a community as such, taxa were targeted for microscopy based on their likely contribution to silica production (i.e. abundant and/or large). Although only a fraction of the community was included in the weighted averages, the reconstructed estimates of community-level silicon stress at each station generally agreed with the direct total community measurements made with ³²Si (Fig. 1b). For the taxon-weighted community measurements, the addition of Si(OH)₄ increased community silica production by 27 ± 250% across stations on the northern transect and 45 ± 224% on the southern transect (Fig. 1b). The taxon-specific average silica production rates were not correlated with [Si(OH)₄] (northern transect, R = −0.42, p = 0.41, southern transect, R = −0.17, p = 0.75).

Taxon-specific changes in silica production, silicification, new frustule surface area, and division rate

The changes in silica production rates of individual taxa showed a wide range of responses to the addition of Si(OH)₄ (Fig. 2) even though the community response was relatively constant (Fig. 1b, white dots). Taxon-specific data were pooled across transects because there was no significant difference in the slopes of the regressions for any taxon-specific rate parameter with [Si(OH)₄] between transects including the Si uptake ratio (V_+Si/V_amb, F = 0.12, p = 0.73), change in silicification (Z_+Si/Z_amb, F = 0.01, p = 0.90), the surface area of new frustule created during the incubation (SA_new+Si /SA_new-amb, F = 1.01, p = 0.31), or division rate (μ_+Si/μ_amb, F = 1.16, p = 0.29). Within a station, taxa exhibited values of V_+Si/V_amb from ~1 to ~3 (Fig. 2a). The variability of V_+Si/V_amb among taxa generally increased as [Si(OH)₄] decreased.

Figure 2:

Relative change in taxon-specific physiological parameters with silicic acid concentration. (a) The relative increase in silica production rate (b) the relative increase in new frustule SA, (c) the relative increase in silicification and (d) the relative change in growth rate. Colors correspond to taxa as shown in the key; Thalassiosira-like cells make up the centrics group. The ratio is between the ambient and enhanced treatments for taxa, the x-axis is the concentration of silicic acid in μM in the ambient condition.

Taxon-specific V_+Si/V_amb varied among stations as well (Sup. Table 1). Some taxa, such as Skeletonema sp. and Chaetoceros spp. <20 μm appeared Si-replete across the gradient in [Si(OH)₄], with V_+Si/V_amb generally close to one (Fig. 2a, Sup. Table 1). Silica production rates for other taxa, such as the centrics (Thalassiosira-like cells) and Chaetoceros spp. >20 μm were more silicon stressed, showing a nearly 3-fold silica production increase in response to added Si(OH)₄ (Fig. 2a). The relative interspecific sensitivity to silicon stress was examined by evaluating the fraction of stations where each species exhibited significant Si stress. The number of instances where a specific taxon’s V_+Si/V_amb was greater than 1.3 (i.e. one plus the average standard error of V_+Si/V_amb, which was 30%) was divided by the number of stations for which we had data for that taxon, to quantify the fraction of stations where limitation was observed for each taxon (Fig. 3). By this metric silica production by Fragilariopsis sp., Skeletonema sp., and Chaetoceros spp. <20 μm was least often silicon stressed, while Proboscia alata, Chaetoceros spp. >20 μm and F. pseudonana was most often silicon stressed. The remaining taxon spanned a continuum.

Figure 3:

Relative stress levels for taxa across all stations. The average standard error of V_+Si/V_amb among taxa was ~30% thus, a taxon was considered notably stressed when V_+Si/V_amb >1.3. The bars depict the quotient of the number of occurrences of V_+Si/V_amb greater than 1.3 divided by the number of stations in which that taxon was present, counts are given over each bar in parentheses.

The amount of silica that a cell produces over a given time is the mathematical product of the surface area of the new frustule created (SA_new, μm²) and the silicification (Z, Si μm⁻²) of the new frustule. On average SA_new+Si was 8% more than SA_new-amb but was variable within a taxon at a given station with no trend in magnitude or variance with [Si(OH)₄] (Fig. 2b). Z increased within the +Si treatment by an average of 36% across taxa and stations (Fig. 2c). Similar to V_+Si/V_amb, the variability in Z_+Si/Z_amb generally increased among taxa with decreasing [Si(OH)₄]. The ~2-fold average increase in Z_+Si/Z_amb with declining [Si(OH)₄] was less than observed for V_+Si/V_amb as expected (see explanation below), except for the high value for centric diatoms at ~10 μM Si(OH)₄ (Fig. 2c).

Generally, division rate was altered little with additional Si(OH)₄. However, P. alata and Chaetoceros spp. >20 μm, which were the taxa most frequently silicon stressed (Fig. 3) showed a 0.5 to 2-fold increase in growth rate with added Si when ambient Si(OH)₄ was <10 μM (Fig. 2d). Division rate was estimated by the PDMPO-labelling patterns of partially, half, and fully labeled cells as in McNair et al. (2018). Division rate generally ranged from one to three divisions per day across taxa (McNair et al., 2018).

The relationship between silica production, silicification, surface area and division rate

The coordination of physiological responses (i.e. silicification (Z), new surface area produced (SA_new), growth rate (μ)) as a function of silicon stress were examined among all taxa (Fig. 4). As described in the methods, all regressions were forced through the coordinate (1, 1). The regressions for both Z and SA_new with the silicon stress index (V_+Si/V_amb) were significant, while that between μ and V_+Si/V_amb was not:

$${\text{Silicification: Z}}_{\text{+Si}}{\text{/Z}}_{\text{amb}}=0.58\pm 0.06\times \left({\text{V}}_{\text{+Si}}{\text{/V}}_{\text{amb}}\right)+0.42\left(p<0.001\right)$$

$${\text{New surface area: SA}}_{\text{new+Si}}{\text{/SA}}_{\text{new-amb}}=0.29\pm 0.04\times \left({\text{V}}_{\text{+Si}}{\text{/V}}_{\text{amb}}\right)+0.71\left(p<0.001\right)$$

$${\text{Growth: µ}}_{\text{+Si}}{\text{/µ}}_{\text{amb}}=0.08\pm 0.05\times \left({\text{V}}_{\text{+Si}}{\text{/V}}_{\text{amb}}\right)+0.92\left(p=0.12\right)$$

Figure 4:

Relationships between changes in silicification (Z, red), new frustule SA (SA_new, blue) and growth rate (μ, black) with increased silicon limitation (V_+Si/V_amb). All least squares linear regressions were forced through (x, y) = (1, 1).

Mathematically changes in silicification and the surface area of new frustule created contribute equally to the rate of silica production, such that the degree of silicon stress can theoretically be related to changes in Z and SA_new by:

$${\text{V}}_{\text{+Si}}{\text{/V}}_{\text{amb}}=\left({\text{SA}}_{\text{new+Si}}{\text{/SA}}_{\text{new-amb}}\right)\times \left({\text{Z}}_{\text{+Si}}{\text{/Z}}_{\text{amb}}\right)$$ (1)

This equation holds for the regression lines in Figure 4 such that the product of Z_+Si/Z_amb and SA_new+Si /SA_new-amb well predicts the measured V_+Si/V_amb (slope = 1.001, R² = 0.93). The slope of the relationship between silicification and the silicon stress index was nearly twice that for the new surface area, indicating that the dominant response across taxa to increasing silicon stress was a reduction in cell silicification.

Silicon stress effects on contribution of taxa to community silica production and abundance

To better understand the role of [Si(OH)₄] in regulating community silica production and community composition, the fractional contribution of each taxon to total silica production and to total cell abundance was compared in the ambient and +Si treatments (Fig. 5). Across taxa the slope of the reduced major axis regression between a taxon’s contribution to silica production (Fig. 5a) in the ambient and +Si treatment was close to unity, 0.97 ± 0.10 (95% C.I.) (R² = 0.91) despite individual taxa exhibiting a two- to three-fold shift in silica production rate between treatments. This signifies that a taxon’s contribution to community silica production was essentially unaltered, even though total community silica production rates generally increased with the addition of Si(OH)₄. Similarly, the fractional abundance of taxa at the end of the incubation, which was calculated using division rate and initial cell abundance (McNair et al., 2018), was not significantly different between the ambient and +Si treatment (Fig. 5b). The slope of the reduced major axis regression for the abundance measures was 1.01 ± 0.08 (S.E.) (R² = 0.96).

Figure 5:

Changes in fractional production and abundance of taxa in the ambient and +Si treatments. (a) The fractional contribution of taxa to community silica production. (b) The fractional abundance of taxa. In both plots the 1:1 line is depicted as a solid black line.

Discussion

Prolonged rapid growth in dynamic ocean regions make diatoms one of the most productive primary producers globally. Our field data show for the first time that the decrease in silicification in lieu of decreased growth rate first inferred in laboratory experiments 45 years ago, is a coping mechanism employed by multiple taxa within natural diatom communities. Both the data presented here and those from culture studies (Guillard et al., 1973; Paasche, 1975) find that diatoms first respond to reductions in silicon availability by thinning their frustules, which minimizes the impact of diminished Si uptake rates on division rates (Fig. 4). The more lightly silicified cells observed in the ambient treatment indicate that, as silica production rates became limited in situ by [Si(OH)₄], cells decreased their silicon quota such that the experimental addition of Si(OH)₄, i.e. release from silicon stress, restored maximum silica production rates and maximum silicification during the 24 h incubation. This is direct and quantitative field evidence that diatoms maximize growth rate at the expense of other aspects of cell physiology, specifically their frustule properties. Such a response is consistent with the r-strategy of diatom adaptation whereby they are able to respond quickly to inputs of nutrients (Sommer, 1983) and maintain high growth rates as nutrients decrease (this study), and thus thrive in turbulent environments (Margalef, 1978). Along with their intrinsic high division rates, the prioritization of cell division rate produces the maximum number of individuals for a given pulse of Si(OH)₄ favoring persistence in environments with episodic nutrient input.

Diatom division rates across the transects significantly correlated with macronutrient concentrations (McNair et al., 2018). However, the results presented here imply that decreasing silicic acid, and thus increasing silicon stress, was not the cause of observed decreases in division rates (Fig. 4). Growth limitation by factors besides silicon (light, iron, nitrate, etc.) often inversely correlate with silicification because decreased growth rate gives a cell more time to acquire silicon, which could result in a fully silicified cell even when [Si(OH)₄] is low (Claquin et al., 2002). All other things being equal, decreasing growth rate would decrease silicon stress (V_+Si/V_amb). Within the dataset, silicon stress generally increased and silicification decreased with decreasing Si(OH)₄ (Fig. 2). This suggests that any decreases in growth did not slow the cell cycle sufficiently to allow cells with less than maximum Si uptake rates to acquire enough silicon to produce fully silicified frustules in ambient [Si(OH)₄]. This ability to produce less silicified frustules in favor of progressing through the cell cycle enabled most taxa in this study to avoid silicon growth limitation for [Si(OH)₄] down to 2 μM.

The lack of decrease in growth rate with increasing silicon stress is likely reflective of the ambient [Si(OH)₄] not being sufficiently low to induce the severe silicon stress that would overcome the compensatory mechanism of frustule thinning and lower division rates. In culture, diatoms are capable of thinning their frustules by a factor of 3–4 in response to low [Si(OH)₄] (Guillard et al., 1973; Paasche, 1975; Brzezinski et al., 1990) suggesting that V_+Si/V_amb would have to exceed a value of 3–4 before reductions in division rate would be expected. In our dataset V_+Si/V_amb was < 3 (Fig. 4) and direct quantifications of cell thinning were generally less than a factor of two (Fig. 4). By both criteria, the observed minimal impact to division rates was expected.

Quantitative data showing the lower sensitivity of growth rate than silica production rate to changes in Si(OH)₄ availability (Fig. 2) provides the opportunity to quantitatively evaluate another aspect of diatom silicon physiology using the response of multiple co-occurring species in nature. An explicit outcome of the thinning of frustules in response to diminishing [Si(OH)₄] is that the half-saturation for growth (K_μ) is less than that for silica production (K_s) (Martin-Jézéquel et al., 2000). Using the kinetic equations for growth (Monod (Monod, 1942)) and silica production (Michaelis-Menten (Michaelis and Menten, 1913)) we can estimate the relative magnitude of the average K_s and average K_μ for taxa in our dataset. By solving both equations for [Si(OH)₄], the setting them equal to each other, and solving for μ_m/μ (Sup. Info.) we form a relationship with K_μ/K_s:

$$\frac{{\text{μ}}_{\text{m}}}{\text{μ}}=\frac{{\text{K}}_{\text{μ}}}{{\text{K}}_{\text{s}}}\left(\frac{{\text{V}}_{\text{m}}}{\text{V}}\right)-\left(\frac{{\text{K}}_{\text{μ}}}{{\text{K}}_{\text{s}}}-1\right)$$ (2)

where μ_m and μ are approximated by the measured μ_+Si and μ_amb, respectively and V_m and V are approximated by V_+Si and V_amb, respectively. This equation is in the same form as the regressions from Figure 4 (μ_+Si/μ_amb = 0.08 ± 0.05 × (V_+Si/V_amb) + 0.92; p = 0.12) such that the slope of V_+Si/V_amb versus μ_+Si/μ_amb is the ratio of K_μ to K_s. The slope of this relationship (0.08) was significantly less than one (t-test, t ratio 15.36, p <0.001). Assuming the Monod and Michaelis-Menten models are accurate for field diatoms, this directly confirms that K_μ is considerably less than K_s and is the underlying reason for the quantitatively measured decoupling of silica production and division rates.

Ecological and biogeochemical implications of silicon stress

Shifting silicon availability was not a dominant force in shaping diatom community composition. Increased availability of Si(OH)₄ minimally affected division rates such that at the end of the 24-hr incubation the fractional distribution of taxa was not different in the ambient and +Si samples (Fig. 5). However, small changes in growth compound over time so any difference in growth rate between the ambient and +Si treatments may be more pronounced on longer timescales. In environments where the ambient [Si(OH)₄] is very low, e.g. <0.5 μM observations in the California Current (Krause et al., 2015) or Mississippi River Plume (Nelson and Dortch, 1996), our results predict that [Si(OH)₄] could severely limit diatom growth and affect community structure on shorter time scales.

Given that silicon stress did not affect division rate, the absence of changes to community composition over our one-day incubations was expected (Fig. 5b). However, community and taxon-specific silica production rates were consistently higher in the +Si treatment (Fig. 1b). This combined with the variability of silicon stress among taxa at a station suggested that the distribution of silica production among diatom taxa would be altered with decreasing [Si(OH)₄]. Yet, there was not a significant difference between treatments in the fractional amount of silica produced by taxa (Fig 5a). The changes in silica production rate that occurred with increased Si(OH)₄ were too small to overcome the large initial differences in fractional contributions to silica production.

While community dynamics were not altered with silicon stress, the coordinated decrease in silicification likely affects regional nutrients cycles and food-web dynamics. More lightly silicified frustules result in decreased cell density which decreases sinking speed. The potential decrease in sinking speed can be calculated using the modified Stokes equations from Miklasz and Denny (2010) assuming decreased silicification was the only physiological change induced with silicon stress. Across the size range of cells in this study sinking speeds would decrease 15–35% for Z_+Si/Z_amb = 1.5 which would increase the likelihood for diatom-bound nutrients being retained in the surface ocean (Tréguer et al., 2017). Changes in silicification could additionally affect flows of energy through the food-web. Studies that measured changes in grazing pressure with changes in silicification found that decreasing silicification led to increased predation from microzooplankton (Zhang et al., 2017) and copepods (Liu et al., 2016). The ratios of silicification from those studies (Z_+Si/Z_amb~1.4 and ~2, respectively) falls within the range observed here, suggesting that top-down forces may become an increasingly important loss for diatoms that are silicon stressed.

The consistent decrease in silicification (Si μm⁻²) in response to silicon stress with the maintenance of division rate confirms the prevalence of an adaptive strategy in nature that facilitates the formation of large diatom blooms in episodically nutrient-rich habitats. Because division rates were generally unaltered by [Si(OH)₄] across taxa, silicon stress had little influence on community composition. The impact on taxon-specific contribution to community silica production was also small in our incubations, despite interspecific differences in the degree of silicon stress among co-occurring diatoms. In these environments with intermittent nutrient infusions diatom community composition is more likely a reflection of interspecific differences in maximum, unrestricted division rates as well as differential grazing pressure rather than a reflection of cells best adapted to low nutrient conditions. Any effects of resource limitation on community succession is likely consequence of taxonomic-specific responses to resources other than dissolved silicon. The decrease of frustule strength caused by decreased silicification could make cells more susceptible to predation and more likely to fuel regenerative production, which together alter food web dynamics and regional nutrient cycles. Yet, the increased risk of predation is opposed by the r-selected strategy of maximizing population size for a given pulse of silicon; thereby favoring diatom persistence in pulsed resource environments such as coastal upwelling systems.

Experimental Procedures

Sample collection

Data for this study was collected during the IrnBru (MV1405) cruise aboard the R/V Melville in July 2014 along two transects off the California coast. The northern transect extended ~100 km southwest from Cape Blanco, Oregon (USA) and the southern transect extended ~200 km west from Point Arena, California (USA) (Fig. 1a). Macronutrient and biogenic silica (bSiO₂) samples were collected along both transects as described in (McNair et al., 2018).

At each station, silica production rates, growth rates, and silicification were measured in incubations of paired ‘ambient’ and ‘+Si’ treatments to characterize the degree of silicon stress. Water for the incubations was collected at roughly two-hour intervals while the ship was underway (Fig. 1a) using a surface-towed, trace metal clean, water pump, (i.e. the Geo Fish, (Bruland et al., 2005)). The fluorescent tracer, PDMPO [2-(4-pyridyl)-5-((4-(2-dimethylaminoethylamino-carbamoyl)methoxy)phenyl)oxazole] was added to each pair of bottles to a final concentration of 0.158 μM. The ambient treatment was not altered further. The +Si treatment was spiked with a Chelex-cleaned sodium metasilicate stock solution such that the final concentration was 19 μM greater than the ambient of [Si(OH)₄]. To measure the silica production rates of the total diatom community at each station, an additional pair of ambient and +Si bottles were spiked with the radioisotope ³²Si (15,567 Bq μg⁻¹). 261 Bq of ³²Si was added to samples from the northern transect and 293 Bq of ³²Si was added to samples from the southern transect.

All the bottles were incubated for 24 h in deck-board, flow-through incubators with circulation surface seawater to maintain temperature and processed as in McNair et al., (2018). PDMPO fluorescence was used to estimate cell-specific silica production rates, growth rates and silicification as described in McNair et al., (2018). Samples for ³²Si production rate measurements were filtered through 25 mm diameter, 1.2 μm pore size, filters, air dried and mounted on planchettes (Krause et al., 2011). ³²Si activity was measured using low-level beta detection after aging samples into secular equilibrium between the ³²Si and its short-lived daughter isotope, ³²P, and production rates calculated as in (Krause et al., 2011).

Metrics of Si stress

A normalization procedure was used to compare the responses to silicon stress among taxa as prior results indicated silica production rates spanned three orders of magnitude (McNair et al., 2018). The relative degree of silicon stress for each taxon was quantified by comparing the rate of silica production in the +Si treatment to the rate of silica production in the ambient treatment (ρ_+Si/ρ_amb) where both measures are in units of mol Si cell⁻¹ h⁻¹. Assuming the silica content for all cells of a given taxa was the same at the start of the incubation in both the ambient and +Si treatments, the ratio of ρ_+Si/ρ_amb = (V+Si^…bSiO₂) ÷ (Vamb bSiO₂) is equivalent to V_+Si/V_amb

The experimental addition of Si(OH)₄ in the +Si treatment is assumed to relieve cells from any silicon stress that may be occurring in the ambient conditions and thus V_+Si/V_amb approximates V_m/V (eq. 1). The normalization places the response of all taxa on the same relative scale that can be directly related to changes in silicon production kinetics (eq. 1). V_+Si/V_amb scales linearly with the degree of Si stress, with larger values indicating greater stress. Similarly, changes in measures of silicification (Z), new frustule SA production rate (SA_new), and growth rate (μ) were normalized by dividing the value of each parameter in the +Si treatment by the paired measurement in the ambient treatment.

Data quality control

The taxon-specific data for silica production rate, growth rate, and silicification are inherently variable. Time and resources limit the number of cells that can be imaged on the confocal microscope for each sample. This constrains the number of cells included in the averages for each taxon and results in a large standard error (S.E.) for the mean response of a taxon when only a few individuals are imaged. Silica production rates can vary widely within a genus or taxonomic group adding to the variance of averages for taxonomic groups that are resolved only to the genus level. Variation also arises from intraspecific variability among cells of the same species. For instance, Fragilariopsis pseudonana had silica production rates that spanned an order of magnitude within a station. This variation is consistent with the findings of Durkin et al. (2012), who reported an order of magnitude range in PDMPO fluorescence in Fragilariopsis sp. (5 μm) from the subarctic Pacific when quantifying fluorescence with a flow cytometer, which analyzes significantly more individuals. The necessary grouping of multiple species into the taxonomic groups reported here such as “centrics” and “Pseudo-nitzschia spp.” compounds apparent intraspecific differences. On average, the S.E. across measurements for different taxon and taxonomic groups was 36% of the mean for V_+Si/V_amb, 19% for Z_+Si/Z_amb, 26% SA_new+Si/SA_new-amb and 12% for μ_+Si/μ_amb.

To reduce the influence of highly variable data points in our analyses, the data were filtered using two objective criteria: the number of cells included in the average needed to be greater than four in both treatments and the magnitude of the S.E. of V_+Si/V_amb had to be less than 70% of the average. This removed 23 out of 61 data points across the thirteen stations. After applying these criteria, the average S.E. relative to the mean decreased to 30% for V_+Si/V_amb, 16% of for Z_+Si/Z_amb, 23% for SA_new+Si/SA_new-amb, and increased to 13% for μ_+Si/μ_amb.

Statistics

Simple linear regressions using JMP12 were used to relate relative changes in growth rate, silicification, and the rate of new frustule surface area production to changes in relative silica production rate. Reduced major axis regressions were not used in this instance because of the asymmetric relationship between variables (Smith, 2009), i.e. X and Y variables were not interchangeable because the intent was to investigate physiological response (Z_+Si/Z_amb, SA_new+Si/SA_new-amb, μ_+Si/μ_amb) as a function of Si stress (V_+Si/V_amb). An analysis of covariance (ANCOVA) was used to test for differences in the regressions between transects (run in JMP 12). A reduced major axis regression was used to compare the relative contribution of taxa to community silica production and fractional abundance in the ambient versus the +Si treatment. This analysis was preformed using the R package lmodel2.

Under the assumption that the Si addition in the +Si treatment raised each metric to their physiological maximum, the relationship between the silicon stress metric (V_+Si/V_amb) and silicification (Z_+Si/Z_amb), new frustule SA (SA_new+Si/SA_new-amb), and growth rate (μ_+Si/μ_amb) should all mathematically converge at x, y = 1, 1 at infinite [Si(OH)₄]. The linear regressions were forced through (1, 1) by subtracting each datum from a value of one, e.g. 1-Z_+Si/Z_amb, which transposed the coordinate (1, 1) to (0, 0). Then least squares linear regression was applied to the relationship between the transformed physiological metric and the transformed silicon stress metric (V_+Si/V_amb) while forcing the intercept of the regression through zero. The resulting equation was translated back to the original parameter space giving a final equation for the line in the form:

$${\text{V}}_{\text{+Si}}{\text{/V}}_{\text{ambi}}={\text{ m* A}}_{\text{+Si}}{\text{/A}}_{\text{amb}}+\left(\text{1 – m}\right)$$ (3)

Where A is either Z, SA_new or μ and m is the slope of the line from the linear regression that was forced through zero.