Introduction

Trichodesmium surface blooms (hereafter, referred to as ‘blooms’) were reported and described near the Torres Straits as ‘sea-sawdust’ by Captain James Cook and Sir Joseph Banks on the Endeavor’s first voyage (Banks, 1770), and near the Abrolhos Islands (east of Bahia, Brazil) in 1832 by Charles Darwin (Darwin, 1909). More recently, remote sensing has revealed the scope of these massive blooms in the subtropical and tropical oceans, extending over 5.4 × 106 km2 (Subramaniam et al., 1999; Westberry and Siegel, 2006) and consisting of up to 1.2 × 107 cells l−1 (Capone et al. 1998; Lugomela et al., 2002; Rodier and Le Borgne, 2008, 2010). These blooms develop swiftly and are characterized by high rates of CO2 and N2 fixation, up to 640 pg C per cell per day and 29 pg N per cell per day, respectively, as well as rapid doubling times (∼2.6 day) (Capone et al., 1998; Rodier and Le Borgne, 2008, 2010; Luo et al., 2012). The elicitors and signals causing these rapid surface-bloom formations of Trichodesmium are still vague (Berman-Frank et al., 2007; Rodier and Le Borgne, 2010; Bergman et al., 2012).

Notably, these massive Trichodesmium blooms often collapse abruptly (within 3–5 days), with mortality rates paralleling bloom development rates (Rodier and Le Borgne, 2008, 2010; Bergman et al., 2012) and with different processes (not necessarily mutually exclusive in natural populations) being implicated in their termination. Bloom crashes may result from viral lysis (Hewson et al., 2004) or autocatalytic programmed cell death (PCD), which is induced by nutrient (iron (Fe) starvation) or high light (oxidative) stress in both laboratory and natural populations (Berman-Frank et al., 2004, 2007). PCD has indeed been well documented in a variety of diverse phytoplankton lineages including bloom formers like Trichodesmium (Berman-Frank et al., 2004; Bidle and Falkowski, 2004), implicating it as an important feature of algal ecophysiology. However, the nature and mechanistic controls of PCD in phytoplankton are not well understood and, moreover, the consequences of this cellular process on ecosystem and biogeochemical dynamics are virtually unexplored.

Indeed, the fate and quantitative contribution of Trichodesmium blooms to the vertical export or recycling of newly fixed nitrogen (N) and carbon (C) in the ocean following bloom collapse is poorly studied (Mulholland, 2007; Bergman et al., 2012), despite its global-scale importance (62–137 Tg N per year and 2.1–18 Tg C per year) (Luo et al., 2012). Trichodesmium may release up to 50–80% of recently fixed N2 in natural communities and cultures as dissolved organic N and ammonium (NH4+) (Mulholland, 2007). Low δ15N of exported material collected from the HOTS and BATS time series stations indicates that recently fixed N is transferred out of the euphotic zone (Karl et al., 2002), although Trichodesmium is rarely recovered in sediment traps and is only grazed by three representatives of the Miraciidae family of pelagic harpacticoid copepods, mostly by Macrosetella gracilis (O’Neil, 1998).

Our previous findings indicate that PCD in Trichodesmium may trigger rapid sinking due to concomitant internal cellular degradation, vacuole loss and increased production of extracellular polysaccharide aggregates, operationally defined as transparent exopolymeric particles (TEPs) (Berman-Frank et al., 2004, 2007). In the present study, we used a customized, experimental water column to specifically test and quantify for the first time whether PCD-induced-bloom demise of Trichodesmium mechanistically regulates increased vertical fluxes of C and N, thereby facilitating its rapid export from surface waters to depth.

Materials and methods

Experimental setup

Triplicate experiments were conducted with Fe-free medium to determine the relationship between PCD-induced bloom termination, and C and N export fluxes. In all experiments, measurements were performed to assess culture state (cell abundance and Chl a), PCD induction (caspase-specific activity and metacaspase gene expression) and organic export (total organic C (TOC), particulate organic C (POC), dissolved organic C (DOC), TEP and particulate organic N (PON) levels). Prior to surface-bloom induction, Trichodesmium erythraeum IMS101 was cultivated as ‘pre-bloom’ exponential-phase batch cultures (0.5 × 106 cells ml−1) in a 10-l air-bubbled YBC II medium (Chen et al., 1996) at 25 °C with a 12:12 light/dark cycle at ∼80 mMol quanta m−2s−1. Although Trichodesmium IMS101 stock cultures were not axenic, cohabitating bacteria accounted for <1% of the total biomass during exponential growth (as determined by epifluorescence microscopy counts). All culture transfers were performed ascetically and media in water columns was kept sterile throughout the experiments.

Experiments were all conducted in a custom-made water column (200 cm high, 13.8 cm (∅) cylinder, five sampling ports) that was fabricated to resemble and model a water column (Figure 1, Supplementary Figure S1). The cylinder was kept in dark (using a black-out jacket) except for surface illumination in a 12:12 light:dark cycle at ∼600 mMol quanta m−2s−1. The custom-made external halogen light source was positioned 0.5 m from the surface water column, illuminating fixture by a fiber optic cord to avoid any thermal stress. Experiments were set by pumping 30 l of sterile YBC II (Fe-free) medium from the bottom of the cylinder using a peristaltic pump (Masterflex 6600, Cole-Parmer Instruments, Vernon Hills, IL, USA). After the water column was filled with YBC II, ∼600 million Trichodesmium cells (from the ‘pre-blooming’ batch culture) were poured from the top. Each experiment was conducted over 5 days and five depths were routinely sampled: surface layer (0.5 cm), 50, 100, 150 and 200 cm deep. All experiments were initiated (T0) after a bloom was formed (3 h since Trichodesmium addition), defined as >95% of the cells in the water column above 0.5 cm. To minimize bacterial contamination, the water column system was cleaned and disinfected with 10% HCl for 3–5 h prior to each run.

Figure 1
figure 1

Schematic overview of the experimental water column setup. The numbers correspond to sampling depths: 1(surface), 2 (50 cm), 3 (100 cm), 4 (150 cm) and 5 (bottom, 200 cm).

Full size image

RNA extraction

Surface filaments (10 ml) were collected onto 25-mm diameter, 5-μm pore-size polycarbonate filters placed in a 2-ml eppendorf tube and snap-frozen in liquid N2. Total RNA was extracted from Trichodesmium cells using TRIzol Reagent (Invitrogen, Life Technologies, Santa Cruz, CA, USA) followed by RNeasy mini Kit (Qiagen, Valencia, CA, USA). DNase (Turbo DNase, Ambion, Life Technologies) was used to any gDNA contamination from the RNA sample and RNeasy MinElute Cleanup Kit (Qiagen) to concentrate and clean the samples. Total RNA concentration was determined using a NanoDrop ND-1000 Spectrophotometer (peqLab Biotechnologie, Erlangen, Germany). First-strand complementary DNA was generated from total RNA by reverse transcription using SuperScript III First-Strand Synthesis System (Invitrogen Life Technologies). Reverse transcription reactions used equal total amounts (500 ng) of RNA and were primed with random hexamers, allowing the investigation of multiple candidate genes from the same source first-strand complementary DNA.

Quantitative real-time polymerase chain reaction (qRT-PCR)

Specific oligonucleotide DNA primer sets were designed as metacaspase 2 and 9 genes (MC2 (NCBI Accession no. YP_722351) and MC9 (NCBI Accession no. YP_724070; Table 1) (Berman-Frank et al., 2004). Expression levels of target genes (TG; metacaspase and iron stress-inducible genes) were normalized to 16S rRNA housekeeping gene (HK) and calibrated to an initial time point (Table 1). Triplicate first-strand complementary DNA aliquots from each sample served as templates in real-time PCR. Reactions containing either ‘no template’ (NTC) or DNase-treated RNA (to exclude genomic DNA contamination) as template served as negative controls. All NTC and RNA-only samples were always below the detection limit. Brilliant II QPCR Reagents (Stratagene, La Jolla, CA, USA) for in situ monitoring of PCR progression. Reaction mixture (total of 10 μl) contained 4 μl diluted complementary DNA (∼20 ng RNA), 5 μl SYBRgreen (Brilliant II QPCR Reagents, Stratagene), 0.5 μl per primer (250 nM) and 0.5 μl DEPC water. Transcript levels were determined by Corbett 6000 Rotor-Gene Q real-time PCR cycler using Brilliant SYBR Green (Brilliant II QPCR Reagents, Stratagene). Q-RT-PCR-running conditions were: 7 min in 95 °C followed by 45 cycles of 10 s in 95 °C, 30 s in 60 °C and 20 s in 72 °C ending in melting, 72 °C–95 °C (1 °C steps). Samples were then analyzed by Rotor-Gene Q software subjected to the CT method Statistics (Stratagene). Standard curves were generated using serially diluted linear plasmid DNA that contained each gene amplicon (Bidle and Bender, 2008) with measured amplification efficiencies being between 88% and 93%. Melt curves were performed with one distinct peak for both metacaspase and 16S primers, whereas a small second peak was also apparent for the IsiBTery primer set (Supplementary Figure S2). The copy numbers of TG relative to housekeeping genes (HK) were estimated as 2[−1(CtHK—CtTG)] (shorthand as 2(−ΔΔCt)) according to Pfaffl (2001).

Table 1 Primer sequences used for qRT-PCR and gene expression of biomolecular markers for T. erythraeum Fe stress and the associated programmed cell death during bloom determination
Full size table

Measurements of caspase activity

Surface filaments (15 ml) were collected onto 25-mM diameter, 5-μM pore-size polycarbonate filters, placed in a 2-ml eppendorf tube and snap-frozen in liquid N. Cells were resuspended in Lauber buffer (50 mM HEPES (pH 7.3), 100 mM NaCl, 10% sucrose, 0.1% 3-[(3- cholamidopropyl)-dimethylammonio]-1 propanesulfonate and 10 mM dithiothreitol) and sonicated on ice (four times, each for a 30 S cycle) using an ultracell disruptor (Sonic Dismembrator, Fisher Scientific, Waltham, MA, USA). Cell extracts were centrifuged then (10 000 g, 2 min), and supernatant was collected for caspase activity and protein assays. Caspase-specific activity was determined using the fluorogenic, peptide substrate IETD-AFC (Calbiochem, Rockland, MA, USA) at a 50 mM final concentration. Kinetic analysis was performed on substrate cleavage over a 4-h period at 26 °C with measurements taken at 15 min intervals using a Fluoroskan Ascent Fl (Labsystems, Waltham, MA, USA) plate reader (Ex 405 nM, Em 485 nm). Cleavage rates were normalized to total protein, as determined by BCA Protein Assay (Thermo Scientific, Rockford, IL, USA).

Cell measurements

Cells were counted using a Sedgwick-Rafter Cell (Pyser-SGI, Kent, UK) and a light microscope (Nikon, Melville, NY, USA; Eclipse 80i). Filament number and size were counted using Nikon plan fluor × 4/0.75 lens and cell length was measured by Nikon plan fluor X40/0.75 lens. Image analyses were done using Image J software (http://rsbweb.nih.gov). Trichodesmium cells are usually cylindrical, therefore, cell volume of the cylinder (length × π × r2) was calculated accordingly: Trichodesmium cell length × π × (1/2 × the cell width)2. Cellular C content was determined according to the equation C=0.433·V0.863 (Verity et al., 1992) as, where C represents cell C content (Ccell, pg) and V represents cell volume (μM3).

Chlorophyll a concentrations

Filaments were collected on GF/F filters (Whatman, Kent, UK) and Chl a was extracted by boiling for 6 min in 90% methanol. Pigment concentration was analyzed spectrophotometrically at 664 and 750 nM (CARY100, Varian, Santa Clara, CA, USA) according to Tandeau de Marsac and Houmard (1988).

Particulate organic C

Samples (50 ml) were filtered through pre-combusted (4 h, 450 °C) GF/F filters, dried overnight at 60 °C and stored in a desiccator until further analysis. POC was determined using a CHN analyzer Perkin Elmer (Waltham, MA, USA) CHNS/O elemental analyzer, PE 2400, after carbonate removal from the filters using overnight fuming with concentrated HCl vapor.

Dissolved organic C

Duplicate water samples (40 ml) were first pre-filtered through combusted GF/F filters to remove particulates and collected in acid-washed glass tubes. HCl (37%) was added (40 μl) to each sample and stored at 4 °C until analysis by TOC analyzer (Shimadzu ASI-L Autosampler, Columbia, MD, USA). TOC was determined as POC+DOC.

TEP determination

Water samples (20 ml) were filtered gently (∼100–150 mbar) onto 0.4-μm polycarbonate filters and TEP concentrations (μg gum xanthan equivalents l−1) were measured according to Passow and Alldredge (1995). A conversion factor of 0.51 was employed to convert from milligrams of gum xanthan equivalents to milligrams of C (Engel and Passow, 2001).

Results and discussion

Formation of Trichodesmium surface bloom

We experimentally induced T. erythraeum IMS101 blooms (Figure 1, Supplementary Figure S1) by introducing a dense culture (3 × 106cells l−1) into an Fe-free water column. Immediately after addition, the trichomes were uniformly distributed (1 × 106–2 × 107 l−1) throughout the upper 100 cm (Figure 2a). The customized water column was operated without turbulence, at an uniform temperature (26 °C), darkened from all sides and illuminated from the surface by ∼600 μMol quanta m−2 s−1 (12:12 h light/dark). Bloom initiation always followed cell aggregation from single trichome morphology into puff and tuft colony formation (Figure 2b). An extremely dense (1.3 × 109±3.3 × 107 l−1) ‘slick’-like bloom developed at the surface (upper 0.5 cm) from these ascending and aggregated colonies at a rate of 4 × 108cells h−1 (Figure 2c). Integrated cell abundance remained constant (3 × 108±9 × 107 m−3) in the water column throughout bloom development simulations, whereas surface cell numbers increased rapidly, indicating that active ascent was due to positively buoyant cells. After ∼3 h, 96–98% of the cells had concentrated at the surface, incorporating significant concentrations of TOC and total organic N (0.72±0.09 mg C m−2 and 0.13±0.03 mg N m−2, respectively), whereas <2% of the cells were evenly distributed below the upper 0.5 cm throughout the rest of the water column (Figure 2a).

Figure 2
figure 2

Dynamics of Trichodesmium bloom formation in our fabricated water column. (a) Depth (surface, 10, 25, 50 and 100 cm) measurements of cell abundance immediately after batch culture addition (T0) and after the establishment of an aggregated bloom (T3h). (b) Surface cell abundance during 3 h of bloom development, along with a representative image (c) taken after the bloom was fully developed (T3h).

Full size image

Although artificially expedited using dense batch cultures (Figure 1), bloom formation in our water column possessed notable similarities to natural blooms observed in tropical oceans, where surface slicks of dense Trichodesmium aggregations develop under calm conditions with sea surface temperatures of ±26 °C, high solar irradiance and high concentrations of Fe and P (Karl et al., 2002; Rodier and Le Borgne, 2008, 2010; Bergman et al., 2012). In these oceanic blooms, the vertical ascent of Trichodesmium cells due to positive buoyancy (Villareal and Carpenter, 2003; Rodier and Le Borgne, 2008, 2010) occurs simultaneously with rapid growth rates, both processes accounting for the hastened, ultra-dense accumulation of cells at the surface (Rodier and Le Borgne, 2008, 2010; Bergman et al., 2012).

Abrupt and synchronized bloom demise

Once surface blooms were established, we utilized the expression levels of isiB, an archetypical prokaryotic iron stress-response protein, as a diagnostic indicator of Fe stress (Chappell and Webb, 2010), along with the levels of metacaspase gene expression (Berman-Frank et al., 2004; Bidle and Falkowski, 2004) and caspase-like catalytic activity (Berman-Frank et al., 2004, 2007) as diagnostic subcellular markers of PCD (see Materials and methods). Whereas cell numbers remained stable in our experimental water column during the first 24 h (T1d) of Fe limitation, slight increases in transcript expression levels of isiB, MC2 and MC9 genes (by 0.5–0.9 logΔΔCt) were already detectable (Table 1, Figures 3a and b). After 2 days (T2d) of Fe limitation, the expression levels of isiB, MC2 and MC9 genes increased significantly with respect to T1 (three-, six- and five-fold, respectively) (Figure 3b). The changes in MC gene expression were paralleled by 26-fold higher caspase activity rates (718±268 RFU mg per protein h−1, Figure 3b).

Figure 3
figure 3

(a) Daily changes in the abundance of Trichodesmium cells from the induction of Fe stress after bloom initiation to the biomass crash after 5 days. (b) Associated changes in various biomolecular PCD markers, including caspase catalytic activity (C.A.) and the expression of selected metacaspase genes (MC2 and MC9). Expression of a known Trichodesmium Fe-stress responsive gene (isiB) was used to confirm subcellular Fe stress.

Full size image

In the aquatic environment, natural populations of Trichodesmium encounter similar Fe limitation where they can mobilize and reincorporate Fe from intracellular storages, such as ‘DNA-binding proteins from starved cells’, to meet cellular demands for growth (Castruita et al., 2007). The observed lag time in our system (with no changes in cell abundance; Figure 3a) between bloom initiation (T0) and the induction of PCD (T2d) may have resulted from the consistent slow release of intracellular Fe (Castruita et al., 2007) prior to the PCD-induced collapse. Following bloom collapse (T3d), MC gene expression and caspase activity rates were significantly reduced from their maximal levels (T2d) by 76–100% and 40–80%, respectively (Figure 3b). The low metacaspase expression and caspase-specific activity levels measured in the remaining surface trichomes were consistent with the low constitutive levels of these genes and protein activities, respectively.

These observations demonstrate that an Fe-stress induced, PCD cellular response mechanistically triggered an abrupt, synchronized Trichodesmium bloom collapse. The drastic reduction (by 71%) in cell abundance (Figure 3a) was characterized by a notable loss of recognizable trichome cellular structure and the transition to green amorphous aggregates composed of dying and decomposing biomass (Figure 3a, Supplementary movie S1). We suggest that the remaining 29% (5.1 × 108cells l−1) of the biomass helped to ensure the survival of the Trichodesmium population.

In different species, cysts, spores and hormogonia cells are known to be resistant to environmental stress and may also resist PCD, thus forming an inoculum for subsequent blooms (Vardi et al., 1999). Careful examination of microscopic cell images (∼100 of × 40 magnification) showed sporadic short filaments during and after the crash, although these could not be definitively characterized as hormogonia. Thus, the surviving trichomes probably avoided PCD through an as yet unknown mechanism(s).

The subcellular triggers, signal transduction and regulatory networks of PCD in Trichodesmium are currently unknown. One intriguing aspect of this PCD response is the potential role that quorum sensing (QS) may have in triggering its onset. QS is known to regulate the cell–cell communication and population response through the activation of different metabolic pathways including cell death (Dandekar et al., 2012), and QS has been recently documented to control phosphorus acquisition among Trichodesmium-epibiont consortia (Hewson et al., 2009b; Hmelo et al., 2012). Further investigation is needed to identify whether QS has any role in the PCD of Trichodesmium.

Linking PCD-induced bloom demise to particulate C and N sequestration

Our experimental water column enabled us to establish a novel mechanistic link between the PCD-induced bloom collapse and quantitative C:N export fluxes. Given this laboratory model system lacks some of the complex abiotic (physical and chemical) and biotic (microbial remineralization) interactions within a natural water column, it provides a ‘best case scenario’ for flux measurements upon Trichodesmium bloom collapse and, therefore, provides an useful constraint for future modeling efforts. During bloom initiation, TOC was predominantly incorporated within the Trichodesmium surface bloom (TOC/Ccell=1, 0.74±0.09 mg Nm−2) and was comprised mostly of POC (Figures 4a and b, Table 2). PON concentrations were 0.13±0.04 mg N m−2 and POC/PON was 6.4±1. TEP comprised a very small component (only 10%) of the TOC pool (Table 2) and DOC was below the analytical detection limits. TEPs are present in large numbers in all aquatic environments and regarded as part of the POC pool (Passow, 2002). They are intensely sticky organic microgels that promote aggregation and may act as a bridge between dissolved and particulate matter (Passow, 2002).

Figure 4
figure 4

Temporal changes in organic carbon and nitrogen in the surface bloom and exported material at the bottom of the water column during 5 days of Fe stress. (a) Calculated Trichodesmium cell carbon content, Ccell (see Materials and methods) along with (b) TOC, (c) carbon content associated with TEP (note scales), and (d and e) PON and POC, respectively. (f) A corresponding image highlighting the fate of the surface bloom following 2 days of induced Fe stress. Note the difference in scales for TEP and PON associated with surface and bottom locations.

Full size image
Table 2 Temporal changes in the relative composition (w/w) and distribution of organic carbon and nitrogen fractions within the water column during the 5 days from bloom induction to biomass crash
Full size table

Abrupt bloom collapse (T2d) from Fe-stress-induced PCD was characterized by a drastic reduction in Ccell (87%) and removal rates of 11–17 μg C m−2 h−1, as well as a dramatic increase (2.2-fold) in TEP production, which resulted in two—three-fold increases in TEP/TOC and TOC/Ccell, respectively (Table 3, Figures 4a–c). Similar to the oceanic milieu (Passow, 2002; Bar-Zeev et al., 2011) and our earlier study (Berman-Frank et al., 2007), these increased ratios reflect the cessation of active incorporation of organic C and N into Trichodesmium cells and its preferential excretion into the media as polysaccharides and detrital matter. Moreover, it led to greatly enhanced sinking rates of captured aggregates, likely resulting from TEP scaffolding (De La Rocha and Passow, 2007) of Trichodesmium detrital matter, documented as large (∼1–5 mm) greenish aggregates (Figure 4f). Twenty-four hours after the bloom collapse (T3d), we quantified the sinking velocities of these aggregates at ∼200 m per day (Supplementary movie S1). Sinking rates of the corresponding TOC (composed mainly from POC) also increased significantly (5.4±1.1 μg C m h−1, Table 3, Supplementary movie S1) at this time (T3d) and rapidly accumulated at the bottom of the column (0.32±0.21 mg C m−2). While organic C and N rapidly sank and accumulated at the chamber bottom at T3d, maximal TEP sinking rates (5.2±3.9 μg C m h−1) were measured 24 h later (T4d). We suggest that the temporal uncoupling between the maximal TOC and TEP sinking rates (T3d vs T4d, respectively) may originate from their differential densities and viscosities (De La Rocha and Passow, 2007). Nevertheless, 5 days (T5) after bloom initiation, cell abundance, organic C and N concentrations at the surface layer of the column were drastically reduced, having been efficiently exported to the bottom of the 2-m water column (Figure 4, Table 3). Notably, TOC measured at the bottom (0.9±0.1 mg C m−2) was predominantly composed of POC (68%), with DOC accounting for only 12% (Figure 4, Table 2). Likewise, TEP accounted for as much as 90% of the TOC (0.84±0.1 mg C m−2). PON-measured concentrations were 0.1±0.04 mg N m−2 and C:N ratio increased to 7.6 (Figure 4, Table 2).

Table 3 Daily changes in the removal rates of surface (0.5 cm) Trichodesmium bloom cell biomass and sinking rates measured at the bottom (200 cm) during the 5 experimental days
Full size table

The mechanistic link between PCD cellular processes and associated pulses of C:N export underscores its fundamental role in regulating cell fate, particle flux, and upper ocean biogeochemistry of bloom-forming diazotrophs. Although export fluxes from natural Trichodesmium or other diazotrophic blooms are poorly defined (Mulholland, 2007; Bergman et al., 2012), newly fixed particulate organic matter (POM) in the pelagic surface oceans is the primary source for exported C and N to the deep ocean (Buesseler et al., 2007; Arístegui et al., 2009). Globally, the export efficiency of POM from surface waters via the biological pump varies considerably (from 2 to 50%), with up to a quarter of this material sinking below 1000 m (Boyd and Trull, 2007), depending on the magnitude of net primary production, sinking velocities and remineralization rates (Boyd and Trull, 2007; De La Rocha and Passow, 2007).

Heterotrophic bacteria have a significant role in decomposing and remineralizing POC, PON and TEP as part of the microbial loop (Passow, 2002; Azam and Malfatti, 2007; Arístegui et al., 2009; Bar-Zeev et al., 2011; Yamada et al., 2012), resulting in particle solubilization, stoichiometric transformations and changes in export fluxes (Cho and Azam, 1988; Smith et al., 1992; Schneider et al., 2003; Boyd and Trull, 2007). Heterotrophic bacteria were recently shown to have an active role in gene transcription as epibionts within the Trichodesmium consortia and are suggested to alter the Trichodesmium metabolism (Hewson et al., 2009b). Our experimental observations represent a contrasting ‘best case scenario’, whereby algal bloom, crash and subsequent vertical flux were largely isolated from these aforementioned microbial transformation processes and, consequently, provide an upper limit of potential vertical C and N fluxes following Trichodesmium bloom crashes.

Sinking velocities are governed by particle concentration (McDonnell and Buesseler, 2010), size (following Stokes law) and density (Arístegui et al., 2009). Our experimental water column simulations highlight that rapid development and abrupt, PCD-induced collapse of Trichodesmium blooms will serve to mechanistically enhance these first-order controls on particle flux. Bloom collapse via PCD commences with the internal degradation of cellular components (such as gas vesicles, Berman-Frank et al., 2004) and is coincident with the massive TEP production. The resulting high concentration of dense, detrital particulate matter coupled with the sticky nature of TEP serves to promote the formation of large aggregates (De La Rocha and Passow, 2007; Burd and Jackson, 2009; McDonnell and Buesseler, 2010) and preserves C and N in particulate forms through a prompt downward pulse of POM (Supplementary movie S1). Moreover, we suggest that the toxicity of Trichodesmium and low palatability to grazers (Rodier and Le Borgne, 2008; Kerbrat et al., 2011) combined with the sheer magnitude of POM export flux will critically limit the amount of recycled matter within the upper mixed layer.

Although the sinking rates and degree of export preservation in our model system cannot a priori be extrapolated to the ocean, our findings provide novel mechanistic context for recent observations of massive, localized Trichodesmium bloom collapses (6 μg C l−1 per day and 42 ng N l−1 per day, Rodier and Le Borgne, 2008, 2010) and reported vertical fluxes of newly fixed POC and PON from diazotrophs (17 μg C m−2 per day and 1.1 μg N m−2 per day, Karl et al., 2012). Indeed, the tight link between PCD and C:N export pulses (demonstrated here) highlights the fundamental influence that PCD may exert on the oceanic biogeochemical cycling associated with bloom-forming natural populations. Parameterizations for the effects of PCD-mediated bloom collapse should be included into future diagnostic models to constrain the relative POC and PON fluxes from diazotrophs in the modern ocean.