Temperature dependences of growth rates and carrying capacities of marine bacteria depart from metabolic theoretical predictions

Using the metabolic theory of ecology (MTE) framework, we evaluated over a whole annual cycle the monthly responses to temperature of the growth rates (μ) and carrying capacities (K) of heterotrophic bacterioplankton at a temperate coastal site. We used experimental incubations spanning 6ºC with bacterial physiological groups identified by flow cytometry according to membrane integrity (live), nucleic acid content (HNA and LNA) and respiratory activity (CTC+). The temperature dependence of μ at the exponential phase of growth was summarized by the activation energy (E), which was variable (−0.52 to 0.72 eV) but followed a seasonal pattern, only reaching the hypothesized value for aerobic heterotrophs of 0.65 eV during the spring bloom for the most active bacterial groups (live, HNA, CTC+). K (i.e. maximum experimental abundance) peaked at 4 × 10⁶ cells mL⁻¹ and generally covaried with μ but, contrary to MTE predictions, it did not decrease consistently with temperature. In the case of live cells, the responses of μ and K to temperature were positively correlated and related to seasonal changes in substrate availability, indicating that the responses of bacteria to warming are far from homogeneous and poorly explained by MTE at our site.

marine bacteria, coastal ocean, flow cytometry, activation energy, carrying capacity, metabolic theory of ecology

INTRODUCTION

Marine bacteria play key roles in biogeochemical cycles, and in the context of global warming their response to temperature is crucial to our understanding of the structure and functioning of the ocean in the future (Sarmiento et al.2004; O'Connor et al.2009). The metabolic theory of ecology (MTE, Brown et al.2004) provides a useful framework to study these effects by supplying a simple theoretical approach and specific predictions against which datasets can be tested, although it is not free from controversy (Martínez del Río 2008; Price et al.2012; Glazier 2014). In essence, the MTE states that individual metabolism can be described by combining the thermodynamic laws that govern the biochemistry of enzymes and the allometric scaling of body mass. These same principles can be used to describe the energy flow at higher levels of organization, which would link the metabolism of organisms, populations and, to a narrower extent, communities (Sibly, Brown and Kodric-Brown 2012). The MTE predicts that individual metabolic rates increase exponentially with temperature (Gillooly et al.2001; Brown et al.2004) and scale as a power law of mass (Kleiber 1932; West, Brown and Enquist 1999; Anderson-Texeira and Vitousek 2012), and that the temperature dependence of mass-corrected metabolic rates is universal for all aerobic organisms living under normal temperature ranges (0–40ºC), despite different lifestyles or levels of organization (Brown et al.2004). The aforementioned premises are expressed in the MTE central equation (1):

\begin{equation} I = I_0 M^{\rm a} {\rm e}^{ - E/kT} , \end{equation}

(1)

where I is the individual metabolic rate; I₀ is a taxon-specific, mass and temperature-independent normalization constant; M is body mass; a the scaling exponent; e^−E/kT is the Boltzmann factor, where E is the average activation energy of metabolism, k is Boltzmann's constant (8.62 × 10⁻⁵ eV K⁻¹) and T is absolute temperature in Kelvin. Growth rates are the equivalent parameter of metabolism at the population level (Savage et al.2004) and in the MTE are described as the sum of the mass-corrected individual metabolic rates, shown in equation (2):

\begin{equation} {\rm \mu } = \sum {(I/M)\sim(M^{\rm a} /M)} {\rm e}^{ - E/kT} \end{equation}

(2)

The baseline rules of MTE hold for populations and μ is expected to increase exponentially with temperature according to the activation energy (E). The MTE proposes an average universal E value of ∼0.65 eV for heterotrophic organisms (Gillooly et al.2001; Brown et al.2004; Yvon-Durocher et al.2012) following the kinetic activation energies of the enzymes involved in respiration, considered the driver of metabolism in the heterotrophic aerobic lifestyle. Activation energies for unicells support the MTE in large database comparisons (Dell, Pawar and Savage 2011; Yvon-Durocher et al.2012) but a wider range of values (from 0.28 to 1.14 eV) have been reported in specific studies for aquatic bacteria (Sinsabaugh and Shah 2010; Morán, Ducklow and Erickson 2011).

In contrast to μ, the MTE states that the carrying capacity (K), another key population parameter, should decrease with increasing temperatures (Gillooly et al.2001; Brown et al.2004; Marquet, Labra and Maurer 2004; Isaac et al.2011) proportionally to the supply rate (assumed to be constant) of the limiting resource for the study population. Nonetheless, in natural environments with fluctuating conditions, authors studying a variety of organisms have observed a mixture of increasing, decreasing and modal responses of K to temperature (Fox and Morin 2001; Jiang and Morin 2004; Isaac et al.2011). For the specific case of bacteria, K is generally studied associated with bottom-up or top-down effects (Pace and Cole 1994; Sonderggard and Middelboe 1995; Vázquez-Domínguez et al.2008; Šolić et al.2009), and in the few studies that address temperature, increases and decreases have been also reported (Li, Head and Harrison 2004; Sarmento et al.2010).

The variable responses of growth rates and carrying capacities to temperature suggest the lack of universality of the MTE predictions for aquatic bacteria. However, to the best of our knowledge, few studies have addressed specifically the temperature dependences of μ and K simultaneously (Ducklow and Hill 1985; Shiah and Ducklow 1994), despite temperature being a key environmental driver already in the present ocean and even though bacterioplankton is predicted to achieve even more dominant roles in the oceans under a warming scenario (Daufresne, Lengfellner and Sommer 2009; Sarmento et al.2010; Morán et al.2015).

In this study, we addressed the temperature effects on the carrying capacities and growth rates of four physiological populations of marine bacteria (studied with flow cytometry) under the variety of trophic regimes found annually in the temperate NE Atlantic. Surface waters of the southern Bay of Biscay continental shelf change predictably from a winter mixing period, in which nutrients are available throughout the water column, to summer stratification with nutrient depletion in the upper layers (Calvo-Díaz and Morán 2006). In 2012, we performed experimental incubations of natural bacterial communities, after removing predators, at three temperatures: the sampling temperature and 3ºC above and below. This allowed us to assess the temperature dependences of different physiological populations and test the MTE predictions and their ecological significance with global warming, namely (1) if E consistently attains a value of 0.65 eV and (2) if K is inversely related to temperature.

MATERIAL AND METHODS

Sampling area and experimental design

Samples were taken from the Southern Bay of Biscay continental shelf at a 110 m depth station (43.675°N, 5.578°W), 37 km off Gijón/Xixón (Spain), for which oceanographic conditions have been extensively studied previously within the RADIALES time series program of the Spanish Institute of Oceanography (www.seriestemporales-ieo.com). The site exhibits characteristic stratification and mixing periods through the year. Surface water was sampled monthly in 2012 (two samples in May, on the 2nd and the 23rd were referred to as ‘April’ and ‘May’, respectively) from R/V ‘José de Rioja’ with 5 L Niskin bottles in a rossette sampler attached to a SeaBird25 CTD probe.

Upon collection, seawater was filtered on board through a 0.8-μm pore cartridge (PALL Corporation) to remove larger planktonic organisms such as flagellates and phytoplankton and isolate the bacterial community from both predators and primary producers. Bacterioplankton abundance in filtered water represented on average 86.3 ± 3.1% of the unfiltered, whole bacterial community. Water was kept in darkened 20 L Nalgene polycarbonate carboys until arrival at the laboratory ca. 4 h later. Once in the laboratory, water was transferred to acid-washed, 4 L Nalgene polycarbonate bottles and placed in three incubators set with the sampling day photoperiod and at three different temperatures: 3ºC below in situ temperature, in situ temperature and 3ºC above in situ temperature. We followed microbial growth by sampling twice per day until bacterial abundance reached a plateau or started to drop (usually between 4 and 7 days) and we did not consider secondary growth phases.

Samples for size fractionated chlorophyll a (20, 2 and 0.2-μm-pore size filter) and nutrient measurements (NO₂, NO₃, PO₄) were collected at the beginning of the experiment and analyzed following the same methodology as in Calvo-Díaz and Morán (2006).

Flow cytometry physiological groups

Four bacterial populations were studied as physiological groups (del Giorgio and Gasol 2008) by their signature on a BD FACSCalibur flow cytometer equipped with an argon 488 nm laser. Live or membrane-intact cells were distinguished from dead or membrane-compromised cells with a combination of propidium iodide (Sigma Chemical) and SybrGreen (Molecular Probes) stains (Grégori et al.2001; Falcioni, Papa and Gasol 2008); CTC+ or actively respiring cells were detected by counting the deposition of oxidized crystals of the CTC-tetrazolium salt (Polysciences; Posch et al.1997; Sherr, del Giorgio and Sherr 1999); high nucleic acid (HNA) and low nucleic acid (LNA) cells were distinguished by the difference in fluorescence intensity after SybrGreen staining (Marie et al.1997). Live-dead and CTC+ cells were analyzed in vivo, while HNA-LNA cells were previously fixed with a final concentration of 1% paraformaldehyde and 0.05% glutaraldehyde, frozen in liquid nitrogen and stored at −80ºC until analysis following Gasol et al. (1999).

Each of the four physiological populations occupies a distinct position in the activity continuum of the bacterioplankton community but with a certain degree of overlap between groups. The whole bacterial community is represented by both the sum of live and dead cells and, separately, by the sum of HNA and LNA cells. CTC+ cells represent a fraction of the live cells, and therefore CTC+ and live cells are spread across the HNA and LNA groups. According to del Giorgio and Gasol (2008), CTC+ cells are the most active and LNA cells are the least active of the community. The former measure directly actively respiring cells and the later have been related to less active bacterial groups such as SAR11 (Schattenhofer et al.2011) or SAR 86 (Vila-Costa et al.2012). HNA and live cells comprise more heterogeneous compositions, with their activities generally falling between those of CTC+ and LNA cells. However, different systems exhibit their own particularities with different responses of each of the physiological groups (Scharek and Latasa 2007; Ortega-Retuerta et al.2008; Horňák, Jezbera and Šimek 2010; Morán, Ducklow and Erickson 2011).

Bacterial growth rates and carrying capacity

Bacterial growth rates (μ) for each physiological group were calculated using the R (R Core Team 2014) package grofit (Kahm et al.2010) by fitting a free cubic spline to the experimental data of ln-transformed abundance against time in days with a smoothing parameter of 0.5, which was defined empirically as the more appropriate to match the observed data. Using the fitted values, we estimated the growth rates as the slope of the linear phase of the natural logarithm of abundance against time in days, calculated by ordinary linear regression (see example in Fig. S1, Supporting Information).We selected this automated method in order to eliminate any subjectivity in the selection of the data points that were included as part of the exponential growth phase and because it is reproducible. Carrying capacity (K) was measured from the experimental data of abundance against time in days as the maximum abundance reached at the plateau stages of each experiment.

MTE temperature dependence parameters

The temperature dependence of bacterial growth rates (E) was analyzed by rearranging equation 2 and ln transforming the terms on both sides, as in equation 3:

\begin{equation} \ln ({\rm \mu }/M^b )\sim - E \times 1/kT{\rm ,} \end{equation}

(3)

where b represents the new scaling exponent (b = a–1 obtained from M^a/M). E was therefore calculated for each month and bacterial group as the slope of ln-transformed growth rates at each experimental temperature against 1/kT (known as an Arrhenius plot). Growth rates from HNA and LNA bacterial groups were initially corrected by changes in mass as proposed by MTE using the mean bacterial cell volume during the growth phase. Due to open controversy regarding the universality of the scaling exponent (e.g. Glazier 2005; DeLong et al.2010; Sinsabaugh et al.2015), we tested b = −0.25, the scaling exponent proposed by MTE for population growth rates (since a is assumed to be 0.75 from fractal-like distribution networks, West, Brown and Enquist 1999), as well as b = 0.73, a value specifically found for prokaryotes (DeLong et al.2010) and b = 0 (no mass correction). However, partially due to the small range in cell size of bacteria, the effect of mass correction of μ was negligible on E (paired t-tests showed mean differences of 0.02–0.03 eV between scaling exponents). Probably, our sample size was, in any case, too small to detect these differences. In addition, we performed a model where the parameters b and E were fit freely, and E values were again non-significantly different from the previously mentioned models. However, b did show high variation indicating that our data were not appropriate to test b differences or propose a new scaling exponent. Assuming that we are unable to detect mass-related differences, and for the sake of comparability across all physiological groups (since a reliable size could not be estimated for CTC+ and live cells), we decided to use uncorrected values of E. The slope of the Arrhenius plot was expected to be negative (growth rates increase with increasing temperatures) with higher slope values indicating a greater dependence to temperature.

We analyzed the effect of temperature on carrying capacity (K) as the slope of linear regressions of ln-transformed K against experimental temperature in the three treatments (see Fig. S2B and D, Supporting Information). We used the direct effect of temperature on ln-K instead of the equation proposed by MTE (Brown et al.2004; Savage et al.2004) because we were lacking relevant information as the supply rate of the limiting resource and a non-controversial scaling exponent (a similar problem as with growth rates). We assumed that supply rate was equal for all the treatments, since they all had the same initial conditions, and that size had a negligible effect due to the small size range considered, being most of the changes in K associated with temperature. According to Savage et al. (2004), carrying capacity should decrease with increasing temperatures (and with increasing body size, although not the scope of this study), resulting in our direct approach in negative K vs temperature slopes. When necessary we will refer to this slope of K against temperature as K-TR (from K-temperature relationship).

RESULTS

Environmental conditions and bacterial physiological structure

We found the expected marked seasonality for a temperate, coastal site with most of our data well in accordance with previous reports (Franco-Vidal and Morán 2011; Morán, Ducklow and Erickson 2011; Calvo-Díaz, Franco-Vidal and Morán 2014; Fig. 1). Temperature ranged from 12.7 to 21.2ºC with minimum values in February and maximum in August. Total chlorophyll a ranged from 0.14 to 1.8 μg L⁻¹, peaking during the spring bloom in March–April as well as in the late autumn mixing phase, and dropping to minimum values in July (Fig. 1A).

Figure 1.

Sampling conditions at the beginning of each of the 12 experiments carried out in 2012. (A) Selected environmental variables (temperature, chlorophyll a and nitrate). (B) Bacterial abundances of the studied flow cytometric physiological populations (live + dead, HNA + LNA and CTC+ cells) and (C) their relative contributions. Each data point represents the mean of three measurements, but since replicates were so similar, SE error bars are not visible.

The initial total heterotrophic bacterial abundance ranged from 0.17 to 1.10 × 10⁶ cells mL⁻¹. Live bacteria contributed on average 89.4 ± 4.5% to the total abundance (live + dead cells). Live and CTC+ abundances showed very similar (r = 0.82 P < 0.01, n = 12) bimodal distributions with peaks in April and summer (Fig. 1B), although CTC+ cells were usually a minor contribution to total numbers (mean 5.6 ± 0.6% of live initial bacterial counts, Fig. 1C). HNA and LNA bacteria showed contrasting unimodal distributions, in which HNA cells were responsible for the live peak in abundance during April–May, whereas LNA cells dominated during the summer months (Fig. 1B). Accordingly, % HNA varied noticeably year round (23–96%, Fig. 1C) matching previous observations where recurrent peaks appeared in April–May (Morán, Calvo-Díaz and Ducklow 2010; Franco-Vidal and Morán 2011). However, % HNA in May reached the highest value ever reported (above 90%) since the start of the time series in 2002.

Bacterial growth rates and carrying capacities

Overall community growth rates ranged from near 0 to almost 2 day⁻¹. Despite similarities in ranges and mean annual values, the seasonal patterns diverged between physiological groups (Fig. 2), which only shared relative maximum values in May. None of the groups had consistently higher growth rates than the others but LNA cells were generally outgrown by the other groups. Live and HNA growth rates followed a similar pattern, but live values were lower during the months with high LNA cell contributions. CTC+ cells shared with live and HNA cells the low rates in summer and, indeed, growth rates of all these three groups were negatively correlated to ambient temperature (Table 1).

Figure 2.

Bacterial growth rates by physiological group (A: live, B: CTC+, C: HNA and D: LNA) for each of the 12 experiments. Closed symbols represent the mean of triplicate growth rate measurements of the in situ treatment (SE error bars narrower than data points). Monthly ambient temperature is represented by the dashed line.

Table 1.

Pearson correlation coefficients for selected variables of the in situ treatment pooling together the 12 experiments (n = 108). See footnotes below for definition of terms.

Physiological group		μ	K	Initial abundance
CTC+		–0.73**	–0.55**	+0.50**
Live		–0.49**	ns	+0.67**
	Temperature
HNA		–0.52**	–0.56**	ns
LNA		ns	+0.75**	+0.80**
CTC+			+0.34*	–0.43**
Live			+0.60**	–0.43**
	μ
HNA			+0.75**	–0.83**
LNA			–0.41**	–0.37**
CTC+				–0.57**
Live				ns
	K
HNA				+0.48**
LNA				+0.86**

Variables and units: Temperature (ºC); growth rates (μ, day⁻¹); carrying capacity (K, cells mL^-1); initial abundance (cells mL⁻¹).

Significance levels: ns = non significant; **P < 0.01; *P < 0.05.

The carrying capacities varied across physiological groups and seasons (Fig. 3). As with growth rates, seasonality was very similar in live and HNA cells (r = 0.87 P < 0.01, n = 12), with relative increases in abundance with respect to initial values ranging from 1- to 14-fold (in September–October and May, respectively). K of HNA, live and CTC+ cells peaked in April (3.36 ± 0.03 × 10⁶, 3.93 ± 0.03 × 10⁶ and 3.22 ± 0.03 × 10⁶ cells mL⁻¹, respectively) while that of LNA cells was found in August (1.09 ± 0.03 × 10⁶ cells mL⁻¹), with high values also found from June through September, reflecting to a large extent the initial abundances (Table 1, Fig. 3D). As a result of this seasonality, carrying capacities decreased with higher ambient temperatures for CTC+ and HNA cells, while LNA bacteria reached higher K at higher temperatures and no pattern was observed for live cells (Table 1). Also, except for live cells, K was correlated to the initial abundance (Table 1). Correlations between in situ carrying capacities and growth rates were significant for all groups, positive for live, HNA and CTC+ cells and negative for LNA cells (Table 1).

Figure 3.

Bacterial abundances by physiological group at ambient conditions (A:live, B: CTC+, C: HNA and D: LNA). Open symbols represent initial abundances and closed symbols the reached carrying capacity. SE error bars of the triplicate measurements are narrower than data points. Monthly ambient temperature is represented by the dashed line.

Temperature dependence

For consistency with prior studies using the MTE framework, we report positive activation energy (E) values as increases in growth rates with temperature and negative values as decreases, which correspond to negative and positive slopes, respectively, of the relationship between ln-transformed μ and 1/kT (equation 3) as in the example (see Fig. S1A and C, Supporting Information). Experimental E values ranged from −0.52 to 0.72 eV for the four physiological groups, although a vast majority (81%) of the values were positive as expected (i.e. growth rates increased with increasing temperature). The seasonal variability of E for live, HNA and CTC+ cells was similar (Fig. 4, with significant correlations of r = 0.68 between HNA and CTC+ and r = 0.63 between HNA and live, P < 0.05, n = 12). Only the highest E values for these groups matched or were close to the MTE prediction of 0.65 eV and were usually reached during the early spring bloom, while values not significantly different from 0 or even negative typically concentrated around summer. In contrast, LNA cells μ did not show a strong temperature dependence and E values were non-significantly different from 0 in 75% of the experiments. LNA E showed no correlations to any of the other physiological groups variation.

Figure 4.

Monthly activation energies (E in eV) by physiological group (A: live, B: CTC+, C: HNA and D: LNA), calculated as the slope of Arrhenius plots of the ln-transformed growth rates of each temperature treatment against 1/kT. Higher slopes (E values) indicate higher temperature dependence. Closed and open symbols represent significant and non-significant slopes, respectively. Error bars represent 95% confidence intervals of significant values. The gray line represents the MTE average predicted E value of 0.65 eV for heterotrophic organisms. The shaded area represents values where growth rates were lower at higher incubation temperatures, opposite to MTE predictions.

The responses of K to temperature (K-TR) were assessed as the slopes of linear regressions, overall ranging from −0.04 to 0.1 cells mL⁻¹ ºC⁻¹. Positive K-TR values indicate increases of K with increasing temperature while negative slopes indicate decreases (see Fig. S2 B and D, Supporting Information). Despite the wide variation found, most of the relationships between carrying capacity and temperature were not significant (Fig. 5), and among significant values, the general trend was positive (i.e. higher K values at higher incubation temperatures), inconsistent with the MTE prediction. For live cells, K-TR values were positively correlated to E values (r = 0.59 P < 0.05, Fig. 6A) and negatively to ambient temperature (r = −0.69 P < 0.05, Fig. 6B), indicating that when the temperature dependence of growth rates was high, K increased accordingly, usually found during the colder months.

Figure 5.

Monthly carrying capacity temperature response (K-TR in cells mL^-1 ºC^-1) by physiological group (A: Live, B: CTC, C: HNA and D: LNA), calculated as the slope of linear regressions of ln-transformed K vs temperature (ºC). Closed and open symbols represent significant and non-significant slopes, respectively. Error bars represent 95% confidence intervals of significant values. The shaded area represents values where K was higher at higher incubation temperatures, opposite to MTE predictions.

Figure 6.

Live cells relationships between the temperature dependence parameters (K-TR and E) and ambient temperature. (A) Relationship of K-TR and E. (B) Relationship of K-TR and ambient temperature. (C) Relationship of E and ambient temperature.

DISCUSSION

The seasonal variability observed for each of the four physiological groups agreed well with previous studies carried out in the sampling area (Morán and Calvo-Díaz 2009; Franco-Vidal and Morán 2011) for both their absolute and relative abundances (Fig. 1B and C). Live cells, with intact membranes and thus, viable and fit for survival, represented on average ca. 90% of the total surface bacteria on an annual basis, which is at the high end of more variable ranges reported elsewhere (Ortega-Retuerta et al.2008; Gasol et al.2009; Lasternas and Agustí 2014). The two universal subgroups of HNA and LNA cells had similar mean relative abundances with LNA prevailing year round (55%), although HNA cells dominated always in the incubations. The responses of live and HNA cells were very similar for the studied variables suggesting that the majority of live cells corresponded to HNA bacteria. Discrepancies only arose when LNA cells had an important initial contribution to the community, like during summer months. The clear shift in dominance of HNA cells during the winter–spring mixing period and of LNA cells during the summer stratification is highly predictable at our study site (Huete-Stauffer and Morán 2012; Morán et al.2015) and responds to a shift in the dominance of phylogenetic species according to the change in environmental conditions, where LNA cells represent species better adapted to oligotrophy (Schattenhofer et al.2011; Vila-Costa et al.2012).

Actively respiring cells (CTC+) represented on average 6% of live cells (usual in coastal environments) and matched almost exactly the bimodal pattern of live cells abundance, as seen in 2007 in another seasonal cycle (Franco-Vidal and Morán 2011) but not in the previous year (Morán and Calvo-Díaz 2009). During the incubations, though, we found a completely different pattern that did not match any of the other groups. We expected to find CTC+ bacterial growth rates to be the highest (Morán, Ducklow and Erickson 2011) but in summer a noticeable drop in μ (Fig. 2B) and K (Fig. 3B) was observed, resulting in CTC+ cells being the most negatively affected by temperature (Table 1). This can be explained if the CTC reactive incorporation is taxon-selective (Gasol and Arístegui 2007; del Giorgio and Gasol 2008) with those particular groups of bacteria being less present in our samples in summer or, if during summer nutrient depletion, the metabolism of the active bacteria was too low to incorporate the tetrazolium salt during the short incubation times.

For each of the physiological groups, the effect of temperature on their growth rates was summarized by using the activation energy (E). According to enzymatic thermodynamics, bacterial growth rates were expected to increase with temperature with a negative slope in Arrhenius plots. However, even if the 6ºC experimental gradient was chosen to minimize thermal stress, we did not always find this pattern. Rather, we found that E was on half of the occasions (56% pooling all months and physiological groups) not significantly different from zero and on very rare occasions (8%) even negative (μ decreased with warming). Under steady nutrient flow, 0.65 eV has been proposed as the mean E expected for heterotrophic organisms in processes governed by respiration, including growth of marine bacterioplankton. The values we obtained usually differed substantially from this prediction and were significantly below the theoretical value (Fig. 4) except on three occasions (live cells in March and April, and HNA cells in March). Differences were noticeable between the performance of the physiological groups according to their position along the activity spectrum described by del Giorgio and Gasol (2008). LNA cells, which consistently showed the lowest growth rates year round, hardly showed any experimental response to temperature although their carrying capacity was much higher in summer (Fig. 3D). Observations in a eutrophic estuary (Morán, Ducklow and Erickson 2011) had shown a marked, higher apparent response of LNA growth rates to temperature than HNA, CTC+ and live cells (E of 1.0 eV vs 0.3–0.4 eV). Either the composition of LNA cells differed among the two sites or LNA cells were too sensitive to short-term experimental perturbations of temperature at our study site, since slower environmental warming clearly associated elevated temperatures with higher LNA cell abundances (e.g. Fig. 1C and Morán et al.2015), and even apparent growth rates (Huete-Stauffer and Morán 2012). On the high activity endpoint, CTC+ cells were expected to reflect more accurately the MTE predictions (since the tetrazolium salt interacts with the electron transport system involved directly in respiration) but E never reached past 0.46 eV, probably because physiological groups with higher energetic demands such as CTC+ cells require higher substrate concentrations than the average live or HNA cells.

Our results show that the activation energies were higher and closer to predictions during the substrate-rich spring bloom (for live, HNA and CTC+ cells) and near zero during the substrate-limited summer, which suggests that resource supply plays a fundamental role for the MTE predictions on the activation energies of growth rates to hold for natural bacterioplankton populations. In other words, lower activation energies took place because substrate supply was too low to meet the higher energetic demands that were produced by higher temperatures. Therefore, there seems to be a clear change in the regulation of bacterial growth rates between spring-winter and summer. The very weak temperature dependence we have detected here for the summer period would be associated with a strong bottom-up control in the study area, as previously suggested by Morán, Calvo-Díaz and Ducklow (2010) using the approach of Billen et al. (1990) and Ducklow (1992). Our results support the idea of a seasonal switch in bacterial growth control in temperate mesotrophic waters (Pinhassi and Hagström 2000; Pomeroy and Wiebe 2001). Based on in situ observations made in 2006 and 2007, Calvo-Díaz, Franco-Vidal and Morán (2014) suggested that temperature plays a fundamental role during the nutrient-rich conditions (winter-spring) and has little or no effect when resource supply is low (summer-autumn). In the same context, variable E (0–0.86 eV) were also found by Sinsabaugh and Shah (2010) when analyzing the temperature dependence of heterotrophic plankton from Otawa rivers together with exoenzymatic activity. The authors concluded that seasonal shifts in resource supply and community composition rather than temperature per se was the principal source of variation. In an extense and recent compilation of the state of the art of metabolic ecology (Sibly, Brown and Kodric-Brown 2012), the authors have now distinguished between the inherent E (fixed at 0.65 eV) and apparent ε (proposed by Sinsabaugh and Shah 2010), which is the real value observed in a community or population (Anderson-Texeira and Vitousek 2012) and that is allowed to vary with environmental conditions. This would imply that the extrapolation of fixed enzymatic parameters to natural population and community levels is not possible (Cyr and Walker 2004), but as seen in this study, still allows us to use MTE as a framework to identify underlying ecological processes (Harte 2004; Marquet, Labra and Maurer 2004), which is one of the strengths of a general and simple theory (Tilman et al.2004).

As with variability in growth rates, the corresponding carrying capacities varied also consistently with season among physiological groups (Fig. 3). The most noticeable difference was the large increase in abundance during spring of all groups except LNA cells (with mean five times the initial values, exceeding 3 × 10⁶ cells mL⁻¹ for HNA and live cells, and 3 × 10⁵ cells mL⁻¹ for CTC+ cells). LNA cells, despite of their low growth rates, were able to increase notably during summer (up to 1 × 10⁶ cells mL⁻¹) when all other groups were probably battling against the resource scarcity, with K values on average two times the initial LNA cell abundances. The maximum bacterial abundance that this temperate ecosystem may hold under natural conditions is ca. 4 × 10⁶ cells mL⁻¹, as found in April. Presumably, strong top-down control during spring precluded ambient bacterial abundances to exceed ∼1 × 10⁶ cells mL^−1. During experimentally induced phytoplankton blooms in the NW Mediterranean, Lekunberri et al. (2012) observed bacterial abundances up to ca. 5 × 10 10⁶ cells mL⁻¹ in a treatment with urea, i.e. three times higher than the maximum abundance found in their control treatment (200 μm pre-filtration) and five times the initial abundances. This indicates that ambient abundances could potentially be much higher, but are not found in nature due to nutritional constraints (Lekunberri et al.2012), like the situation found in summer with low inorganic nutrients and phytoplnkton biomass (Fig. 1A), or top-down control by protistan predators and viruses (Bonilla-Findji et al.2009) which was likely stronger in spring at our study site (Calvo-Díaz, Franco-Vidal and Morán 2014).

One of our most interesting findings is that carrying capacity did not decrease consistently with temperature (Fig. 5) as hypothesized by the MTE (Savage et al.2004; Anderson-Texeira and Vitousek 2012). In almost 70% of our observations, no significant effect was found over the 6ºC temperature range, and when it was, the slope was usually positive rather than negative. Although no clear pattern could be detected for other physiological groups, live cells (which is clearly the major group in terms of its contribution to total bacterial numbers: Fig 1B) challenged the MTE prediction in a more clear way (Fig. 5B). The temperature dependences of both μ and K (represented respectively as E and K-TR) were positively correlated, that is, higher growth rates with temperature also resulted in higher carrying capacities (Fig. 6A). One intriguing observation was that the predicted decrease of K with temperature was only suggested during warm (i.e. >18ºC), nutrient-limited conditions (Fig. 6B) with an overall significant negative relationship with ambient temperature. This suggests that the MTE predictions concerning the carrying capacity of heterotrophic bacterioplankton are more likely to be met in the future, warmer ocean. However, this would be accompanied by activation energies much lower than the predicted value of 0.65 eV due to severe nutrient limitation (Fig. 6C), if warming induces stronger stratification as has been recurrently suggested (Sarmiento et al.2004; Giovannoni and Vergin 2012)

The MTE predicts that population density should decrease with increasing temperatures because the same supply of resources can support fewer organisms with higher metabolic rates (Savage et al.2004). As a rough prediction, for a 6ºC difference and an E of 0.65 eV, a reduction of 58% in abundances would have been expected (using the approximation in Sarmento et al.2010). Taking this prediction at face value, warming would result in lower abundances of planktonic bacteria. However, the few extended time-series analyses published to date do not show an unanimous response of bacteria. For instance, Sarmento et al. (2010) observed a 10% decrease in bacterial abundance per year in the NW Mediterranean over 7 yr (with an increase of 0.032ºC yr⁻¹), while the decadal study at this site showed the opposite, with a 3% increase (with a temperature increase of 0.086ºC yr⁻¹, Morán et al.2015). Among other factors, the nutritional status of these two systems may partially explain the discrepancies, and following this line, our conclusions should be taken with caution since one of the fundamental assumptions of MTE is based on a steady flow of nutrients. In our experimental approach, we confined inside a bottle a bacterial community that grew until, presumably, some resource, like inorganic nutrients or DOC, became limiting. Given that seawater was taken at the same spot and time, the initial nutritional conditions were equal for all experimental bottles. Except for the few months in which a temperature effect was observed, all treatments reached virtually the same K, independently of the increase in growth rates. Rather, the major differences we observed dealt with the time at which K was reached (earlier for the warmer incubations, data not shown). As a general pattern, raised temperature tended to accelerate growth rates in all groups except the LNA cells (Fig. 4), but the corresponding K values generally remained constant or increased (Fig. 5). Sinsabaugh and Shah (2010) hypothesized that if resource supply increases, then K may remain constant. In our experiments, that would imply that there was recirculation of nutrients at the latest growth phases, which could be possible by viral lysis at high population densities. However, analysis of viral densities inside our experiments (data not shown) did not show a consistent pattern of increase that could have supported this hypothesis.

Concluding remarks

In summary, our results show that the study of a natural mixed assemblage did not behave directly as predicted by the MTE. The temperature dependence of growth rates varied seasonally tied to substrate availability for the three most active physiological groups (CTC+, live and HNA cells) and differed from a fixed 0.65 eV value, which was only reached or approached during the two months of the spring phytoplankton bloom. As for changes in carrying capacity linked to a temperature increase, no differences between temperature treatments was the prevailing result, with even and a few significant cases showing increases of K at higher temperatures. Overall, our results stressed the importance of the nutritional status of the system for future predictions: although both heterotrophic bacterial growth rates and carrying capacities were to a variable extent affected by temperature, their response to warming will be strongly tied to resource availability.

SUPPLEMENTARY DATA

Supplementary Data.

We would like to thank the Spanish Institute of Oceanography for accepting our participation in the monthly time-series cruises off Gijón/Xixón of the RADIALES project (MINECO) and we thank all the staff of the R/V ‘José de Rioja’, for their assistance collecting samples. We are grateful to L. Alonso-Sáez, A. Calvo-Díaz and A. López-Urrutia for their help during experiments, theoretical guidance and useful comments during manuscript preparation. Finally, we thank H.W. Ducklow and two anonymous reviewers for their feedback and suggestions to improve the manuscript.

FUNDING

This work was supported by the Spanish Ministry of Economy and Competitiveness (MINECO) through funding of the COMITE (Coastal Ocean MIcrobial plankton and Temperature) project [CTM-2010–15840] and the associated PhD scholarship [BES-2011–048573] to TM Huete-Stauffer.

Conflict of interest. None declared.