Pulsed export of carbon in the north-western Mediterranean Sea

The short term (hourly scale) variability of heterotrophic prokaryote (HP) vertical distribution and respiratory activity, was investigated in the north-western (NW) Mediterranean Sea. HP vertical distribution was determined on board by flow cytometry analysis of seawater samples collected by series of CTD casts. Cell counts and viability were determined for all samples. HP respiratory rates were determined later in the laboratory from filtered seawater samples (23 dm3) from 300–1 150-m depth. The average cell viability was 94.8%±2.2% (n=240). There was no accumulation of dead cells, due to quick decay of damaged cells. In the epipelagic layer, three HP groups were distinguished, two (HNA1, HNA2) whose cells exhibited a high nucleic acid content and one (LNA) with low nucleic acid content cells. HNA2 was most populated at 50 m but not detected at 90 m and below, presumably aerobic anoxygenic photoheterotrophic bacteria (AAPs). The variability in HP abundance was mainly confined in the upper 80 m. A few secondary peaks of HP abundance were observed (80–150 m) in connection with abundance troughs in the surface layer. HP cells were continuously present in a wide layer around 500 m (mean 191×103 cells/cm3). Below this layer, HP abundance randomly exhibited peaks, coupled to respiratory rate peaks. The HP abundance and variability in the water column was suppressed during a strong wind event. The observed sporadic variability was tentatively interpreted through a pulsed carbon-export mechanism induced by the microorganism production of dissolved polysaccharides, followed by flocculation and rapid sinking. This mechanism would thus contribute to (i) preventing organic matter accumulation in the epipelagic layer, (ii) seeding the water column with live HP cells, and (iii) supplying the aphotic water column with fresh and labile organic matter. This important vertical flux mechanism needs further observations and modelling.


INTRODUCTION
Biological activity in the ocean occurs primarily in the sunlit layer. The ocean occupies about 70% of the Earth surface, the oceanic primary production comprises about half that of the planet (Field et al., 1998), driven by phytoplankton photosynthetic activity. The short life cycle of phytoplankton cells implies they are highly biodynamic, enabling them to sustain the oceanic primary production from a phytoplankton, which represents only 0.2% of the total world biomass. In a fi rst phase, most of the eff orts of the scientifi c community focused on characterizing and quantifying oceanic primary production. Then the fate of this primary production was assessed in the sunlit layer where biological activity seemed concentrated and organic matter mineralisation became an increasing topic of interest, unveiling the role of heterotrophic microbes ; see also a review by del Giorgio and Duarte, 2002). To investigate the export of the primary production to the deep ocean, the biological activity in the mesopelagic (200-1 000 m) and bathypelagic (1 000-4 000 m) realms has to be quantifi ed, because it is there that heterotrophic prokaryotes (HP) are the main agents of organic matter mineralisation, the key diff erence between the mesopelagic layer and deeper being the water mass residence-time, decades for the former, centuries for deeper layers . Whereas it was long considered that biological activity was negligible in deep waters (Menzel and Ryther, 1968), in-situ determinations in the mesopelagic layer and calculation of vertically integrated respiratory activity (Packard et al., 1988;Lefèvre et al., 1996;Boyd et al., 1999) supported the idea that biological activity below the epipelagic layer could be of the same order as biological activity in oceanic upper 200 m. This dark-ocean activity was thought to be fuelled by export of particulate organic carbon (Boyd et al., 1999), but it has since been shown that dissolved organic carbon is the main source of organic matter mineralisation in the mesopelagic layer (Lefèvre et al., 1996;Karl et al., 1998). Primary production, calculated from the dynamics of particulate organic matter variation, is not the only source of dissolved organic matter; phytoplankton exudation, which has rarely been included in budgets of primary production, is now suspected of playing a signifi cant auxiliary role in the export of organic matter. Indeed, transparent exopolymeric particles (TEP), formed from dissolved polysaccharide precursors, have received much attention (Passow, 2002;Verdugo et al., 2004), and their implication in organic matter export is now clearly established (Engel et al., 2004).
The present study was a contribution to a national programme dedicated to the Dynamics of rapid processes in the water column. We report here experimental evidence of pulsed export of HP cells from the epipelagic layer that did not exhibit a clear periodicity. This pulsed export was revealed by investigating the short-term (hourly scale) variation of the vertical HP distribution from the surface down to 1 150 m. The distribution and viability of HP cells was established by on-board fl ow cytometry analysis of freshly collected samples. Their respiratory activity was assessed later in the laboratory. Results are interpreted by hypothesising a mechanism of TEPmediated particle export to the mesopelagic layer in line with that demonstrated in mesocosm experiments by Engel et al. (2004).
During each leg, seawater was collected in the mesopelagic layer at 300,350,430,510,590,670,750,830,910,990, 1 070, and 1 150 m, over a period of 36 h, with CTD casts every 3 h.

Flow cytometry
For fl ow cytometry analysis, seawater samples were collected at each depth in 150-cm 3 fl asks not completely fi lled to enable homogenisation. Seawater was fi ltered through a 100-μm mesh size net to prevent clogging of the fl ow cytometer, and homogenised before taking subsamples for the subsequent fl ow cytometry analysis following vortexing. Results are therefore representative of abundances averaged over the initial volume (between 100 and 150 cm 3 ). Under these conditions, replicates would improve neither the information nor the accuracy defi ned by the reproducibility of the instrument and the Poisson law regarding count levels (Shapiro, 2003).
The HP analyses by fl ow cytometry were carried out on board, directly upon collection of the seawater from the Niskin bottles, after each CTD cast, by using a fl ow cytometer (Cytoron Absolute, Ortho Diagnostic Systems) equipped with an air-cooled 488-nm argon laser. The sample and sheath rates were 1 and 100 mm 3 /s respectively. For each cell, 5 optical variables were recorded: two scatter variables, namely forward angle scatter (related to particle size) and right angle scatter (related to cell structure), and three fl uorescence variables measured emissions in the red (>620 nm), orange (565-592 nm) and green (515-530 nm) wavelength ranges. Data were collected and stored in list-mode with the IMMUNOCOUNT ® software (ORTHO Diagnostic Systems) that also provided the cell concentrations (cells/cm 3 ) of the cell groups resolved by the cluster analyses.
Two protocols were applied to each sample. One subsample was stained with SYBRGreen™ (Molecular Probes ® solution diluted 1/5 000 in fi nal solution; see also Zhao et al., 2020), staining nucleic acids for the overall cell counting and cell group resolution. For this purpose, each 1-cm 3 seawater sample was supplemented with 10-mm 3 SYBRGreen™ and incubated 15 min in the dark before analysis. Counting reproducibility was better than 3.7 standard deviation (SD) ( n =23) over a range of concentrations extending one order of magnitude (Grégori et al., 2001).
The other subsample was subjected to the nucleic acid double staining (NADS) experiment to test cell viability. These subsamples were supplemented with SYBRGreen™ and propidium iodide (PI, Sigma), according to Grégori et al. (2001).

Respiratory activity
For respiratory activity measurements, two Niskin bottles were fi lled at each depth and these seawater samples were further fi ltered onto 47-mm GFF glass fi bre fi lters. Filters were then preserved in cryovials, freeze trapped and stored in liquid nitrogen until their use in the laboratory where they were submitted to the electron transport system (ETS) assay of Packard (1971), following the upgraded protocol (Packard and Williams, 1981;Packard et al., 1996;Packard and Christensen, 2004) based on the kinetics of INT (iodonitrotetrazolium salt) reduction by NADH dehydrogenase upon nicotinamide adenine dinucleotide (NADH) addition.

HP Integrated abundance
Integrated abundances of HP were calculated according to the trapezoidal method for the diff erent ocean layers that were assessed: epipelagic (0-200 m), mesopelagic (200-1 000 m), bathypelagic (1 000-1 150 m). HP abundances at 200 and 1 000 m were derived by interpolation from the study data.

RESULT
A grid of 16 stations was sampled 3 times to locate the most suitable station for this study, which was chosen as the cruise Time Series Station (TSS), 43°25ʹN, 8°00ʹE, between Nice and the Corsica island, the central point ( Fig.1), where horizontal gradients were observed to be weak (Andersen et al., 2009). Such conditions favoured the investigation of short-term processes in the vertical dimension, with the least interference possible from horizontal eff ects such as lateral advection. (ii) damaged cells, distinguished by their red fl uorescence from PI which only penetrates cells when they are damaged. These cells strongly bind nucleic acids as an intercalant, indicating their strong affi nity for one another. PI emits red fl uorescence upon excitation by the 488-nm laser. If PI binds a nucleic acid close (<70 Å) to a bound SYBRGreen™ probe, it induces a fl uorescence resonance energy transfer from SYBRGreen™ to propidium that triggers the PI red fl uorescence emission and prevents the green fl uorescence emission of SYBRGreen™. This is why when the cell damage is large enough, the damaged cells are characterised not only by red fl uorescence but also by a lower green fl uorescence than the live cells. The cytograms also distinguished (iii) rapidly decaying cells for which both green and red fl uorescence signals are lower than those of the damaged cells exhibiting the maximum red fl uorescence. There was no accumulation of dead cells. The nucleic acid degradation decreases the number of binding sites for both PI and SYBRGreen™. The average percentage of live cells through all the study was 94.8%±2.2% ( n =240). The viability control does not distinguish the clusters resolved with a specifi c protocol (Fig.3). It is therefore reasonable to extend to each cluster the overall percentage of live cells (94.8%).

Community structure
Three distinct HP groups were resolved in the upper 150 m of the epipelagic layer (Fig.3). In addition to the usual (Gasol et al., 1999) high nucleic acid content (here labeled HNA1) and low nucleic acid content (LNA) cells, a second HNA group, labeled HNA2 ( Fig.3) was observed. Its nucleic acid content (green fl uorescence signal) and its side scatter signal were higher than those of the HNA1 group but its cell abundance was lower and depth-dependent, with a maximum at about 50-m depth and rapidly declining to become undetectable at 90-m depth (Fig.3).

Vertical distribution variability
The vertical distribution of HP in the epipelagic (0-150 m) layer is displayed in Fig.4. The fi gure regroups observations made during the four sampling cycles carried out in the epipelagic layer during the cruise. As a general trend, HP was mainly present in the upper 80 m where their abundance exhibited marked variability. For instance, in Cycle 1 at 50-m depth, the maximum abundance (1.82×10 6 cells/cm 3 , Julian day 264.21) was about twice (1.9) the minimum abundance (0.97×10 6 cells/cm 3 ) observed on Julian day 264.95. This variability quickly vanished in Cycle 4 after Julian day 287.45, following the sharp decrease in abundance, down to values about eight times less than the abundance value of Julian day 287.45. This large decrease in abundance appeared as a consequence of a strong wind event occurring on Julian day 287 with a wind speed >25 m/s (Andersen et al., 2009).
A few secondary peaks of HP abundance were observed below 80 m, particularly during cycles 1 and 2, for instance at Julian days 265.72 and 271.72 respectively. A close examination of the related vertical distributions revealed that all these deeper abundance peaks were associated with an abundance trough at 60-80 m as shown in Fig.5.

Fig.2 Flow cytogram visualisation of heterotrophic prokaryote viability
All samples were subject to this kind of fl ow cytometry according to the double staining procedure of Grégori et al. (2001). The Green fl uorescence versus red fl uorescence cytograms enabled the distinction of (i) live cells, only exhibiting the green fl uorescence of SYBRGreen™, (ii) damaged cells exhibiting red fl uorescence from propidium iodide and a lower green fl uorescence under the occurrence of fl uorescence resonance energy transfer (FRET) between SYBRGreen™ and propidium iodide that penetrates damaged membranes, and (iii) rapidly decaying cells for which both green and red fl uorescence are decreasing compared to damaged cells that exhibit the maximum red fl uorescence signal. Damaged cells were decaying quickly enough to prevent accumulation of dead cells which refl ects the loss of binding sites on nucleic acids for both PI and SYBRGreen™. a.u.: arbitrary unit. No.1 DENIS et al.: Pulsed export of carbon in the north-western Mediterranean Sea

Vertical distribution variability
The vertical distribution of HPs in the 300-1 150-m layer established during the cruise over two periods of 36 h, through CTD casts run every 3 h, is displayed in Fig.6. One notable feature is the sustained and large abundance of HPs at the 500-m level where their concentration reached values more than twice the background level in deeper water. Another striking observation is the presence of a number of abundance peaks found at diff erent depths below 500 m (Figs.6 & 7) and occurring without apparent periodicity.

Cell-specifi c respiration rate
The respiration rate of HPs in the dark layer was determined as detailed in Section 2. Figure  activity between 300-and 1 150-m depth, on Julian day 282.979. This distribution followed the same pattern as the HP abundance vertical distribution. The respiratory activity was always high at the 500-m depth level and peaks of respiratory activity were observed at greater depths in a non-periodic way and coinciding with the related peaks of HP abundance. On the same graph, with a distinct respiration rate scale, is represented the vertical distribution of the related specifi c respiration rate (fmol O 2 /(h·cell)). The comparison of these two vertical distributions highlights two remarkable features: (i) the trough in the respiration rate occurring in the 400-500-m layer coincides with a trough in HP cell-specifi c activity at 430 m, and (ii) the peaks of HP respiration rate observed at 590-, 750-, and 1 000-m coincide with peaks of HP specifi c activity at the same respective depths. The mean value of these 3 peaks of HP specifi c activity diff ers by only 2.5% from the specifi c activity at 300 m, assigned to cells sinking from the surface layer and penetrating into the mesopelagic layer. Figure 8b displays the time course of the mean cell-specifi c rate of each cast during leg 2. The highest mean cell-specifi c activity displayed in Fig.8b corresponds to the vertical distribution in Fig.8a (Julian day 282.979). Within 6 h, the mean cellspecifi c activity decreased from the highest observed value to a kind of base level. Indeed, for that lowlevel period, the HP cell-specifi c activity exhibited small variations with depth around the mean value (see Supplementary Fig.S1).

Time course of HP abundance at diff erent depths
The abundance of HPs was monitored at 510 and 1 000-m depth between Julian days 277 and 290, taking advantage of all casts run during that period. The abundance variability at both depths is displayed in Fig.9, together with that observed at 50-m depth. Fourier analysis of these three time-series could detect no potential periodicity. The average abundance (1.13×10 5 cells/cm 3 , n =56) at 1 000-m depth, represented 10.8% of the average abundance (1.05×10 6 cells/cm 3 , n =43) at 50 m. Table 1 is given the ranges of abundance variation at 50, 510, and 1 000 m and their respective standard deviations.
The ranges of abundance variation at 510 and 1 000 m represent a very similar percentage (close to 7%) of the abundance variation range at 50-m depth, pointing to the epipelagic layer as the main factor responsible for the HP abundance variability in the mesopelagic realm.
Most of the extreme values (peaks and troughs) at 1 000-m depth correspond to similar extremes (peaks or troughs) at 510-m depth though the amplitudes were diff erent.
The time course variations at the 3 selected depths in Fig.9 reveal that the wind event that occurred on Julian day 287 not only induced a sharp (3 to 5-fold decrease between 20-70 m within less than 3 h) abundance decline in the epipelagic layer, but also simultaneously aff ected the whole dark-sampledlayer where HP abundance sharply declined to a depth-dependent baseline level during the remaining days of the experiment.
To document the characteristics of the widely accepted ocean layers defi ned with respect to depth , i.e., epipelagic (0-200 m), mesopelagic (200-1 000 m), and bathypelagic (1 000-4 000 m) within the limit of the study-sampled water column, we calculated the vertically integrated HP abundances over each layer as defi ned in Material and Method. Results are reported in Table 2. The lower limit of the mesopelagic layer (1 000 m) was interpolated between values determined at 990 and 1 070 m. HP abundance values for the bathypelagic layer were available only between 1 000 and 1 150 m. From our own observations, HP integrated abundance through the mesopelagic layer was quite similar to the one calculated for the epipelagic layer (Table 2). The comparison was derived from the 43 casts at these 3 depths that included seawater samples submitted to on board fl ow cytometry analysis. The variation ranges at 510 and 1 000 m represented similar percentages of the variation range at 50 m, which supports the major role of the epipelagic layer in the HP abundance variability in the mesopelagic layer. σ represents standard deviation. In the absence of any sample between 150-and 300-m depth, HP abundance at 200 m was derived by linear interpolation from abundances at 150 and 300 m. HP abundance at 1 000 m was also determined by linear interpolation. The HP abundance in the bathypelagic layer was only partially defi ned between 1 000-and 1 150-m depth.

DISCUSSION
The sampling strategy for the 2004 DYNAPROC2 cruise (present study) was designed to document potential rapid processes involving HP through their abundance and respiration rate. If respiration rates were determined in the laboratory after the cruise from seawater samples collected every 3 h over 36 h, HP abundances were established by on board fl ow cytometry analysis in almost real time through the 1 200-m water column at high frequency (up to every 3 h, based on Fourier analysis of the HP abundances). It rapidly became clear that the variation of the HP vertical distribution was not periodic (see Figs.4, 6, & 7). In the same sea area as the present study, Méjanelle and Dachs (2009) found that variation was aperiodic in both biomass and vertical fl ux of various lipid biomarkers. This is consistent with the fi nding in the present study that variation of HP abundance was aperiodic in both the epipelagic layer (Fig.4) and in deeper water (Fig.6).
Our fl ow cytometry analysis of HP encompassed both free and attached cells as demonstrated by Mével et al. (2008) during the same cruise. This results from vortexing samples before analyzing them with the fl ow cytometer and indicates that cells were not tightly bound to the particles. Consequently, the abundances reported here encompass systematically attached and free cells. The observed percentage of viable HP through the study was very high (94.8%±2.2%; n =240). This is very similar to the percentage (96.7 %) observed in a Marseille yacht harbor in February 2000, but higher than the percentage (41.0%) of viable HP found in open sea during the same investigation by the end of March 2000 (Grégori et al., 2003). In fact, the real diff erence comes from the observation of dead cells (53.3% in March 2000) whereas none was detected in the present study, as if, once damaged, their decay was too rapid to let dead cells accumulate. Nevertheless, it is very important to know that almost all HP cells analysed down to 1 200 m were live cells.

Epipelagic layer (0-200 m)
The HP community structure in the epipelagic layer appeared to be depth-dependent and apparently light dependent for the group labeled HNA2. The fact that this group was only present in the upper 90 m with a maximum abundance at 50-m depth (Fig.3) makes it a good candidate for being aerobic anoxygenic photoheterotrophic bacteria (AAPs) (Kolber et al., 2000). This question remains open since this group was not sorted for identifi cation and analysis of fi ngerprinting profi les by capillary electrophoresis single strand conformation polymorphism (CE-SSCP) conducted by Ghiglione et al. (2008) in samples from the same cruise. Our suggested identity (AAPs) thus remains tentative.
The HP abundance was highly variable in the upper 80 m but rarely declined below a relatively steady base-level abundance of around 0.5×10 6 to 1.0×10 6 cells/cm 3 (Fig.4). Further secondary peaks of abundance were observed in the 80-150-m subsurface  Fig.5. Interestingly, the vertical profi les of HP abundance exhibiting such peaks like the profi les established on Julian days 265.72 (Fig.5a) and 271.72 (Fig.5b) also exhibited an abundance trough in the surface layer, strongly suggesting that peaks in cell abundance appearing apparently randomly in subsurface layers were sinking from their upper location. This is consistent with the upcast seawater collection enabling to sample the trough generated by the sinking cells before it becomes fi lled by the gradient concentration eff ect.

Mesopelagic layer (200-1 000 m)
The HP vertical distribution in the mesopelagic layer did not decrease continuously with depth like one would expect but revealed a layer of high abundance around 500 m, permanently occupied by HP (Fig.6) with a mean abundance value of 1.91× 10 5 cells/cm 3 ( n =56) at 510 m, more than twice the mean abundance value (7.9×10 4 cells/cm 3 ( n =56)) at 1 000 m. This layer is known to host a mesozooplankton community, a large fraction of which is undergoing diel vertical migrations (Andersen and Sardou, 1992;Andersen et al., 2001a, b;Raybaud et al., 2008;Isla et al., 2015;Kelly et al., 2019). Their abundance is large enough to maintain the presence of the prokaryotic community through their exudation, supplying prokaryotes with fresh and labile dissolved organic matter. In contrast, peaks of HP abundance were observed deeper in an apparently random way regarding time and depth, reaching values as high as in the 500-m layer (Figs.6-7). These cells were living (Fig.2) and their respiratory activity was in agreement with earlier observations in this area (Savenkoff et al., 1992(Savenkoff et al., , 1993Lefèvre et al., 1996). The reported deep abundances of HP are consistent with the results from the particle fl ux study at 200-m depth conducted at 6-h intervals during the cruise (Méjanelle and Dachs, 2009). No day-night periodicity could be inferred from the observations and those authors concluded that the short-term temporal variability of export fl uxes was depending primarily on physical constraints exerted by carrier particle dynamics (Méjanelle and Dachs, 2009). The HP abundances exceeding the base line for abundance at each particular depth would thus be drawn down by particles sinking essentially from the epipelagic layer. This is consistent with the similarity of percentages at 510-and 1 000-m depth of abundance variation ranges with respect to that in the epipelagic layer (Table 1). The vertical distribution of HP strongly suggests that these cells were sinking in variable numbers of particles, that they were overwhelmingly alive thus preserving their capacities for respiration (Fig.8a) and mineralisation activity. There is no reason to think that the particles would have been perfectly identical, carrying the same numbers of cells and sinking simultaneously from the same level. Furthermore, ongoing mineralisation during sinking may modify the particle morphology and their rheology (Jenkinson, 1986;Jenkinson et al., 1991) as may also increase hydrostatic pressure. Consequently, it is reasonable to consider that a sinking event starting in the epipelagic layer would give rise to sinking trails in the mesopelagic and bathypelagic layers, spreading vertically the pulsed nature of the initial event at the surface. The bathypelagic layer bottom depth (4 000 m) is close to the average

cells /cm 3 ) at diff erent depths
The temporal variability of HP abundance followed a similar trend at 1 000-and 510-m depth and appeared similarly linked to the variability of HP abundance at 50-m depth, pointing at the epipelagic layer as the main source of the extra HP cells appearing seemingly randomly deeper in the vertical profi les. The arrows highlight the impact of the wind event through the water column evidenced by a simultaneous sharp decline of the HP abundance variability at all depths.
Vol. 41 198 J. OCEANOL. LIMNOL., 41(1), 2023 world ocean depth (3 800 m). Consequently, the HP integrated abundance for the bathypelagic layer can be considered as representative of the bathypelagic ocean (1 000-4 000 m) where HP would be more abundant than in the epipelagic layer (0-200 m), and the mesopelagic layer (200-1 000 m) as well (Table 2). Such a diff erence might suggest that the CO 2 sequestration capacity of the bathypelagic world ocean could be larger than the one of the mesopelagic layer, which remains to be substantiated.

Respiration rate
HP cell-specifi c respiration rates have been found to closely refl ect their metabolic activity (Baltar et al., 2009). The observed trough at 430-m depth in the vertical profi le of HP respiratory rate (Fig.8a) is reminiscent of the abundance troughs at 60-m depth displayed in Fig.5. This trough strongly suggests that a fraction of the HP community sustained by the presence of meso-zooplankton was drawn down by the pulse of HP loaded particles. In this Fig.8a, the vertical distribution of HP specifi c activity is displayed alongside the HP respiratory rate, exhibiting a minimum at 430-m depth where occurred the trough of HP respiratory rate. In contrast, both variables exhibited peak values at the same depths (490, 750, and 1 000 m) suggesting that HP specifi c activity remained unchanged during sinking as long as the colonised TEPs were not disintegrated. The alternating peaks and minimums in HP specifi c activity clearly illustrate the succession of pulses whose intensity may decline rapidly as illustrate by the sharp decrease of the mean specifi c activity in Fig.8b.

Wind event
The wind event that occurred on Julian day 287 with a wind speed >12.8 m/s, lasted 3 days and was stronger than the one experienced during the DYNAPROC cruise conducted at the same site in May 1995 . The maximum wind speed was 16 m/s and this wind event aff ected the suspended particles throughout the water column, decreasing their abundance and their variability as observed by using an under-water video profi ler (UVP, Stemman et al., 2000;Denis et al., 2003). In the present study, there was no deployed UVP throughout the whole water column, but the reduced variability of the vertical HP distribution upon the strong wind event suggests the same wind eff ect as the one observed with a UVP (Stemman et al., 2000;Denis et al., 2003). The wind-induced fl occulation generated in the epipelagic layer has previously been found to produce large TEPdiatom aggregates massively and rapidly sinking at the end of the bloom (Passow et al., 1994). Unfortunately, our sampling was interrupted before the return to steady conditions. Nevertheless, this wind impact on the particle vertical distribution deserves specifi c investigations to quantify its role in carbon export.

Pulsed carbon-export hypothesis
To account for the reported observations, we formulate the hypothesis that, similarly to the critical micelle concentration (CMC), there exists a critical TEP concentration (cTEPc) above which autoaggregation occurs to form particles dense enough to sink with the load of heterotrophic prokaryotes colonising them, following the conceptual scheme displayed in Fig.10.
The existence and production of TEP have been the object of many investigations and are well documented (see a review by Passow, 2002). The rate at which cTEPc is reached depends naturally on the production rate of TEP precursors by phytoplankton whose spatial distribution is not isotropic. The size of TEP was found to be <100 μm (Beauvais, 2003) which strongly suggests that acquiring a larger size upon aggregation triggers sinking of the particles. When the cTEPc threshold is exceeded in a given When this production exceeds 20 μmol/L C (Engel et al., 2004), TEPs are formed and colonised by HP. Beyond the hypothesised critical TEP concentration (cTEPc, defi ned by analogy with the critical micelle concentration (CMC) for lipids), auto-aggregation occurs leading to sinking of HP loaded particles. No.1 DENIS et al.: Pulsed export of carbon in the north-western Mediterranean Sea volume, TEP auto-aggregation generates one or several sinking particles that leave this volume and its close surroundings with a TEP concentration well below cTEPc, but not equal to zero. Our space and time resolution were insuffi cient to provide estimates of such elementary volumes and frequency of sinking TEP particles. However, the data shown in Fig.5a provides useful information: since the secondary peak in abundance at 100 m on Julian day 265.72 was generated by cells sinking from 60-m depth, and knowing cell concentrations with and without the sinking event in the sampling fl asks, at 3-h intervals, it is possible to calculate that the equivalent of 30 to 50 cm 3 of seawater with its HP load from 60 m before sinking had sunk to 100 m. Considering that the sinking HP are necessarily attached HP that may represent about 50% of HP in the epipelagic layer (Mével et al., 2008), the secondary abundance peak observed at 100-m depth would comprise cells initially contained at 60 m in a volume of 60 to 100 cm 3 .
This pulsed export is therefore far from negligible. It is highly likely that the TEP auto-aggregation step also entraps other organic material (including phytoplankton cells) and suspended particles, further reducing the buoyancy of the polysaccharide assemblage. The potentiality of this mechanism was soundly established by Engel et al. (2004) in a mesocosm experiment based on a triggered Emiliania huxleyi bloom. These authors highlighted a cascading aggregation mechanism, increasing collision, and coagulation rates with suspended particles, enhancing sedimentation rates. Additional TEP loading could take place when crossing the thick layer at about 500-m depth with sustained high HP abundance, through collision and coagulation, and perhaps also through the chemotactic behaviour of some HP (Smriga et al., 2016). Importantly, the concentration determined for a peak of HP abundance at depth with respect to the background level would essentially express the dissemination of the HP loading in the TEPs of the initial seawater sample. In the Fig.7 profi le (Julian day 283.979), at 750-m depth, the extra HP drawn down by TEPs would represent about 80 000 cells/cm 3 . If all these cells were loaded into a single TEP equivalent to a sphere of 500-μm radius, the HP concentration inside that TEP would be 1.53×10 8 cells/cm 3 . This is similar to the bacterial concentration (1.9×10 8 cells/cm 3 ) in marine snow aggregates reported by Smith et al. (1992). In a spherical TEP of 50-μm radius in line with the size limit of 100 μm reported by Beauvais (2003), the HP concentration inside that TEP would be 1.523×10 11 cells/cm 3 . Assuming a value for HP concentration inside the collected TEPs similar to the bacteria concentration in marine snow (Smith et al., 1992), the TEP abundance spread throughout sample collected at 750-m depth would be 800 TEPs/cm 3 .
To our knowledge, the reported results are the fi rst fi eld data providing evidence for this pulsed carbon export from the epipelagic to the mesopelagic layer, that may add to other mechanisms contributing to the export function of the biological carbon pump. Our hypothesis is consistent with the very few in-situ TEP measurements (Annane et al., 2015).
The hypothesised pulsed carbon-export mechanism is of critical importance for sustaining the oceanic biological activity: (i) it permanently contributes to fl ush down the excess of organic matter produced at the surface, preventing its accumulation, (ii) it continuously seeds the water column with live HP cells, and (iii) it also supplies the water column with fresh and labile organic matter and it could release hydrolytic enzymes like marine snow aggregates do (Azam and Smith, 1991), while being a source of dissolved organic matter and HP in the dark ocean. TEP fl ushing together with diel vertical migrations of meso-zooplankton grazers may explain why the Secchi disk depth remained unchanged over decades and was considered insensitive to climate change (Falkowski and Wilson, 1992).
The reported observations challenge the current carbon export paradigm by providing supporting evidence for two export mechanisms: "pulsed export" and "water column fl ushing" not taken into account in the conventional export paradigm and still absent in the latest review on that topic (Le Moigne, 2019). Another overlooked carbon export mechanism involving dinofl agellates was recently reported by Larsson et al. (2022), further challenging the carbon export mechanism. Our study is in line with the consensus statement expressed by Cavicchioli et al. (2019), stressing the point that microorganisms that represent 90% of the ocean living biomass should not be left aside from investigations addressing interactions between ocean and global change.
New tools are needed to better characterise and quantify this pulsed export as well as the impact of strong wind events triggering a huge particle fl ushing throughout the whole water column. An automated fl ow cytometer designed for high frequency observation of HPs and developed in collaboration with a dedicated Vol. 41 200 J. OCEANOL. LIMNOL., 41(1), 2023 company, was validated by our group (Silovic et al., 2017). It should soon be operational, opening access to the dynamics of heterotrophic microorganisms. The next objective is to make this instrument submersible and able to run through the mesopelagic layer and further down where HP are active. This objective would comply with the statement that "assessing climate change involves new ways of studying the ocean's microbes" (Marx, 2020). Meanwhile, we will undertake modelling of this mechanism that could help to quantify the reported pulsed export that is not covered by the novel Microstructure, Multiscale, Mechanistic, Marine Aggregates in the Global Ocean (M4AGO) sinking scheme, that ties ballasting mineral and particulate organic carbon (Maerz et al., 2020).

CONCLUSION
The short-term (hourly scale) variation of the vertical abundance, viability and respiratory-rate distributions of heterotrophic prokaryotes (HP) in the epipelagic and mesopelagic layers revealed that: (i) the average HP cell viability for all samples was 94.8%±2.2% ( n =240); (ii) the epipelagic layer exhibited a large HP abundance variability restricted to the upper 80 m; (iii) secondary peaks of HP abundance occurred in the subsurface layer (80-150 m) temporally associated with abundance troughs in the surface layer; (iv) a large and sustained HP community was present at about 500-m depth, a layer occupied by meso-zooplankton, partially involved in diel vertical migration; (v) below 500 m, coinciding peaks of abundance and respiratory activity occurred in a non-periodic way; and (vi) the vertical distribution of HP specifi c activity provides evidence that the pulsed carbon export is triggered in the epipelagic layer. This export may aggregate additional HP and suspended material when crossing the (400-500 m) layer occupied by meso-zooplankton.
The reported observations are fully consistent with the export mechanism characterized in a mesocosm experiment to investigate TEP formation, aggregation and sinking. The fi rst in-situ evidence of this mechanism in the absence of TEP analysis was made possible by singling out pulses of HP that colonise the TEP aggregates. This pulsed carbon export mechanism is distinct from the "Enzyme Pathway" of carbon fl ow, which may proceed in parallel. It is also diff erent from the particle fl ushing mechanism induced by strong wind events through the whole water column. Our approach is comforted by the recent article reporting another overlooked carbon export mechanism involving dinofl agellates that cannot be confused with the heterotrophic prokaryotes of the present study.
All these mechanisms deserve to be further investigated to be quantifi ed and to refi ne carbon export estimates. At stake is the quantifi cation of the CO 2 sequestration capacity of the mesopelagic and bathypelagic layers and their respective roles in global change.

DATA AVAILABILITY STATEMENT
The datasets generated and/or analyzed during the current study are available from the corresponding author on reasonable request.

ACKNOWLEDGMENT
We thank the crew of the R/V Thalassa for ship operations. We thank Yunyan CHEN (MIO, IOCAS) for producing the Ocean Data View (ODV) graphs and David Nerini (MIO) for crucial Fourier analyses.
We are grateful to the chief scientist, our friend and colleague, Valérie Andersen (Laboratoire d'Océanographie, Villefranche sur mer, France), for organizing the DYNAPROC 2 cruise. She prematurely passed away in March 2007, and we aff ectionately dedicate this paper to her. We are particularly thankful to Roger Z. YU for the kind introduction in Chinese of our article in the Journal of Oceanology and Limnology.

OPEN ACCESS
This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/ licenses/by/4.0/.