Effects of Phosphorus Limitation on the Bioavailability of DOM Released by Marine Heterotrophic Prokaryotes

Heterotrophic prokaryotes (HP) contribute largely to dissolved organic matter (DOM) processing in the ocean, but they also release diverse organic substances. The bioavailability of DOM released by HP under varying environmental conditions has not been fully elucidated. In this study, we investigated the bioavailability of DOM released by a single bacterial strain (Sphingopyxis alaskensis) and 2 natural HP communities grown under P-replete and P-limited conditions. The released DOM (HP-DOM) was used as a substrate for natural HP communities at a coastal site in the Northwestern Mediterranean Sea. We followed changes in HP growth, enzymatic activity, diversity, and community composition together with the consumption of HP-DOM fluorescence (FDOM). HP-DOM produced under P-replete and P-limited conditions promoted significant growth in all incubations. No clear differences in HP-DOM lability released under P-repletion and P-limitation were evidenced based on the HP growth, and P-limitation was not demonstrated to decrease HP-DOM lability. However, HP-DOM supported the growth of diverse HP communities, and P-driven differences in HP-DOM quality were selected for different indicator taxa in the degrading communities. The humic-like fluorescence, commonly considered recalcitrant, was consumed during the incubations when this peak was initially dominating the FDOM pool, and this consumption coincided with higher alkaline phosphatase activity. Taken together, our findings emphasize that HP-DOM lability is dependent on both DOM quality, which is shaped by P availability, and the composition of the consumer community.


Introduction
Dissolved organic matter (DOM) is a large pool of reduced carbon consisting of diverse molecules with various degrees of reactivity. A significant portion of this DOM is recalcitrant to microbial degradation, thus serving as a carbon sequestration mechanism [1]. Heterotrophic prokaryotes (HP) play an important role in DOM processing in the ocean [2] and are well known to uptake DOM for biomass production and respiration. However, HP metabolism also results in the generation of DOM. This process, termed the "microbial carbon pump" (MCP) [3], was suggested to make an important contribution to carbon sequestration [4]. The MCP encompasses different pathways, including the successive microbial transformation of labile DOM into recalcitrant DOM and the direct release of recalcitrant DOM by HP [5][6][7][8][9][10][11]. These processes sustain a large DOM pool, including DOM compounds that can be refractory in specific environmental conditions but labile in others and DOM compounds present at concentrations below the HP uptake limit [12]. DOM bioavailability to HP can then be viewed as a relative concept that can be challenging to address. Additional information on the DOM released by HP, the factors influencing its release, and its further bioavailability are crucial for a better understanding of carbon sequestration through the MCP.

3
DOM lability in the ocean can be controlled by the composition and metabolic capacities of the degrading community [13][14][15][16], but it can is also be shaped by environmental variables such as inorganic nutrients or temperature [17,18]. Nutrient availability, for example, can alter both the suite of DOM molecules released to the environment and their uptake. Phosphorus (P) limitation has been reported in previous studies [19], particularly in the Mediterranean Sea [20][21][22]. Previous studies showed that the condition in which DOM is released can influence its availability to bacteria [23,24]. Using exudates of diatoms grown under P-enriched and P-depleted conditions, Puddu et al. [24] showed that bacterial growth was significantly lower in the exudates from P-depleted diatoms. In spite of previous studies showing that P-limitation can influence the HP extracellular metabolome [25] and its optical properties [26,27], no studies have investigated the effects of P-limitation on the bioavailability of DOM released by HP.
In a previous work, we showed the effect of P-limitation on the quantity and fluorescence properties of DOM released by HP (a single bacterial strain and 2 natural HP communities from different seasons) grown under P-limitation (C:P ratio 45:1) and P-repletion (C:P ratio 374:1) [26]. We have shown that under P-limitation, HP-DOM was enriched in humic-like fluorescence [26], which is largely considered refractory [28][29][30][31], whereas under P-repletion, protein-like fluorescence, which is considered labile [32][33][34], prevailed. The present study builds on these findings to determine the effect of P-limitation on the bioavailability of this HP-DOM. We hypothesize that (i) DOM released under P-limitation would thus promote lower HP growth and that (ii) the labile protein-like fluorescence would decrease during the incubation contrary to the likely refractory humic-like DOM. To test these hypotheses, we carried out short-term incubations of HP communities from the NW Mediterranean Sea with HP-DOM collected during the stationary phase under different culturing conditions. HP-DOM consumption was followed by changes in DOM fluorescence spectroscopy (FDOM). The response of the HP communities to the different types of DOM was investigated using flow cytometry, enzymatic activity, and 16S rRNA sequencing.

Experimental Design
To evaluate the bioavailability of DOM released by HP under P-repletion and P-limitation, we conducted two-step experiments. First (Fig. 1a), a single bacterial strain (Sphingopyxis alaskensis) and 2 natural HP communities from different seasons (SOLA-fall and SOLA-spring) were grown in minimum media with glucose (200 μMol C L −1 ) as the sole carbon source under P-replete and P-limited conditions (coded balanced "B" and unbalanced "U"). Cultures were harvested at the stationary phase, and DOM in the media was separated from the cells by 0.2-μm filtration. Since little glucose (1.8 to 23.4 μmol C L −1 ) was remaining at the end of the cultures of the 1st step, measured dissolved organic carbon (DOC) was assumed to reflect that released by the HP, and the quality of this HP-DOM was characterized using DOM fluorescence. The detailed experimental design of the 1st step can be found in the Supporting Information, and its results are described elsewhere [26]. In the 2nd step of the 3− experiment, the DOM derived from the bacterial strain and the HP communities were used as a substrate for biodegradation experiments (Fig. 1b). Inocula of in situ HP communities from the Mediterranean Sea were added to the DOM collected in the release experiments, i.e., DOM released by the strain and the two HP communities under P-limited (U-DOM) or P-replete (B-DOM). In situ DOM (IS-DOM) alone or in situ DOM supplemented with 20 μmol L −1 of glucose (G-DOM) were used as additional culture media in each of the experiments. The G-DOM treatment served as a control to rule out that HP grew only on glucose leftovers from the first step of the experiment. IS-DOM was collected from the same location and date as the HP inoculum of the respective experiment ( Fig. 1; Table S1).
To prepare the inoculum, seawater collected at the Service d'Observation du Laboratoire Arago (SOLA) observatory (Northwestern Mediterranean Sea, 42° 29′ 300 N-03° 08′ 700 E) was first prefiltered through 0.8 μm to remove phytoplankton and grazers. Of the prefiltered water, 2 to 3 L was then reduced to 200-250 mL by filtration through 0.2μm filters using a vacuum pump to concentrate the inoculum (Table S1). We then stopped the filtration and resuspended the cells stuck to the filter by pipetting out and in several times. This volume was then distributed to the different treatments as inoculum, with an aliquot reserved for 16S rRNA gene sequencing.
The incubations consisted of 10% (vol/vol) of the inocula and 90% of DOM from each treatment. Inorganic nutrients (NH 4 + and PO 4 − ) were added to force carbon limitation (Table S2). The incubations were performed in triplicate at in situ temperature in the dark. Cell abundance was followed through the incubation, and samples for FDOM, exoenzymes activity, and diversity were taken at the onset (T 0 ) and end of the incubations (T f ).

Chemical and Biochemical Analysis
HP cell abundances were followed by flow cytometry using a BD Biosciences FACSCanto. Samples were fixed using glutaraldehyde (0.5% final concentration) and preserved at −80°C until analysis. Samples were thawed and stained with SYBR Green (0.025% final concentration) [35,36]. Fluorescent beads were added as internal standards, and prokaryotic cells were identified in plots of side-scattered light vs. green fluorescence. Since we did not discriminate between autotrophic and heterotrophic prokaryotes in these cytograms, and given our experimental design (dark incubations), autotrophic prokaryotes are considered negligible.
Samples for FDOM were filtered through 0.2-μm polycarbonate filters and analyzed within a few hours upon sampling. Excitation emission matrices (EEMs) were recorded using a JASCO FP-8500 Spectrofluorimeter in a 10-mm quartz cell at a scan speed of 5000 nm min −1 . Excitation scans ranged between 240 and 450 nm at 5-nm intervals, and emission scans were recorded between 300 and 560 nm at 2-nm intervals. MilliQ EEM blanks were recorded in every batch of analysis. Fluorescence intensities were reported in Raman units (RU), obtained by dividing the fluorescence units by the MilliQ blank peak area (Raman scatter) excited at 350 nm. EEMs were separated into 6 distinct components using parallel factor analysis (PARAFAC) under the drEEM toolbox in MATLAB (Murphy et al., 2013). These results are described elsewhere [26]. We used the protein-like component C 340 (ex 275/ em 340) as an indicator of labile DOM and the microbial humic-like C 398 (ex 314/em 398) as an indicator of recalcitrant DOM.
Extracellular enzyme activities were quantified using fluorogenic substrates in a plate reader. Each sample was pipetted in triplicate into 96-well black plates and added to the following substrates: 4-methylumbelliferyl phosphate (for alkaline phosphatase (APA)) and l-leucine-7-amido-4-methyl coumarin (for leu-aminopeptidase (AMA)) (final concentration of 125 μmol L −1 ). Fluorescence was measured immediately after the addition of the substrates, after 1.5 h, and after 3-to 4-h incubations in the dark at the same temperature as for the experimental incubations. Fluorescence readings were done with a JASCO FP-8500 Spectrofluorimeter at 365/450 nm ex/em wavelengths. The increase in fluorescence units during the period of incubation was converted into enzymatic activity (μM h −1 ) with standard curves prepared with 7-amino-4-methylcoumarin for AMA and 4-methylumbelliferone for APA.
Samples for HP diversity (70 to 140 mL) were filtered onto 0.2 μm polycarbonate filters that were stored at −80°C until analysis. The frozen filters were then cut with sterilized scissors. Cell lysis was conducted in 2 steps: adding 50 μL of lysozyme (20 mg mL −1 ) and incubating for 45 min at 37°C followed by a 1-h incubation at 55°C after adding 10-μL proteinase K (20 mg mL −1 ). DNA was then purified using Quick DNA™ fungal/bacteria miniprep kit (Cat. No.: D6005 Zymo Research). Amplification and sequencing were done at LGC Genomics GmbH (Germany). 16srRNA gene amplicon sequencing (300 bp paired-end read Illumina Miseq V3) was done using the primer pair 515FY-926Rjed [37]. Sample demultiplexing using Illumina bcl2fastq, as well as adapter and primer clipping, were performed by the sequencing company. Further processing, including filtering and trimming, dereplication, sample inference, merging pair reads, and removing chimeras, was done under R using the DADA2 package (v1.18.0, [38]). Obtained amplicon sequence variants (ASVs) were taxonomically assigned using the Genome Taxonomy Database (GTDB) [39].

Data Analyses
Differences in cell abundance between treatments were tested using repeated measures ANOVA and post hoc pairwise t-tests (package rstatix in R, v0.7.0; Kassambara, [40]). Differences in exoenzymatic activities were tested using two-way ANOVA and post hoc pairwise t-tests (package rstatix in R, v0.7.0; Kassambara, [40]). To see dissimilarities in HP communities' structures, we used a non-metric multidimensional scaling (NMDS) based on Bray-Curtis dissimilarity matrices, and then differences in communities' structures were tested by ANOSIM tests (package vegan, v2.5.7; Oksanen et al., [41]). This was performed on the ASV relative abundance obtained after rarefaction to the minimum number of reads per sample using the rrarefy function from package vegan (v2.5.7; Oksanen et al., [41]). An indicator ASV analysis was run using indval function from package labdsv (v2.0.1 Roberts, [42]) based on the absolute abundance after rarefaction. Only ASVs with more than 100 reads were considered for this analysis.

Changes in Heterotrophic Prokaryote Abundance
Heterotrophic prokaryotes from the Mediterranean Sea grown on B-DOM and U-DOM (DOM released under P-repletion and P-limitation, respectively) from S. alaskensis reached significantly higher abundances than if grown on IS-DOM and on G-DOM ( Fig. 2a; repeated measurement ANOVA (Table S3) and post hoc pairwise t-test ( Fig  S1)). In contrast, DOM released by natural HP communities SOLA-fall and SOLA-spring in both B-and U-DOM treatments supported lower carrying capacities as compared to the IS-DOM and the G-DOM (Fig. 2b, c; Table S3; Fig  S1). No clear differences in cell abundances between B-and U-DOM treatments were observed in the S. alaskensis DOM biodegradation experiment. Cells grew slightly faster in the B-DOM treatment during the exponential phase but tended to be similar to the U-DOM treatment in the stationary phase. However, significantly higher cell numbers (Table S3; Fig S1) on U-DOM as compared to B-DOM were detected in both SOLA-fall and SOLA-spring experiments (Fig. 2b, c).   (Fig. 3a, b). In SOLA-fall, a pronounced decrease in C 340 was observed in the B-DOM treatments, while it was less depleted in the U-DOM treatments (Fig. 3a). The humic-like C 398 was only consumed in the U-DOM treatment (Fig. 3b). In SOLA-spring, a C 340 consumption was detected in the B-DOM treatments but not in the U-DOM, and there was no decrease of C 398 (Fig. 3a, b). In parallel, APA was slightly higher in the S. alaskensis B-DOM than in the U-DOM; however, this difference was not statistically significant. APA was higher in the SOLA-fall U-DOM, where humic-like FDOM C 398 was consumed than in the B-DOM (Fig. 3c). AMA did not differ among treatments or experiments (Fig. 3b).

Changes in Heterotrophic Prokaryote Community Composition
Samples for community composition were only taken in the S. alaskensis and SOLA-fall DOM experiments. Changes in diversity and composition were observed in all treatments ( Table 2; Fig. 4). At T 0 , HP communities growing on S. alaskensis and SOLA-fall were both highly dominated by Vibrionaceae (92% and 56-95%, respectively), and minor contributions from Flavobacteriaceae, Pelagibacteriaceae, and Rhodobacteraceae were observed (Fig. 4a, b). The dominance of Vibrionaceae at the onset of the experiments could be due to the concentration procedure of the inocula, which would favor the presence of larger cells such as Vibrio species (see "Experimental Design"). However, the HP taxa that were initially present at low relative abundances made up the majority of the community composition at T f . This led to an increase in diversity (either expressed as Shannon or evenness) between T 0 and T f in both experiments (Table 2). At T f , there were no significant differences in diversity (evenness, Shannon) between the B-DOM and U-DOM treatments (p value >0.05). The S. alaskensis-derived DOM led to a lower diversity compared to its in situ DOM treatment while HP community-derived DOM (SOLA-fall) led to similar diversity as in situ DOM (Table 2).
In the S. alaskensis DOM degradation experiment, the B-DOM treatment was dominated by Alteromonadaceae (> 20%) and Methylophagaceae (> 30%), except for replicate 2, where Methylophagaceae were not present. In the U-DOM treatment, Vibrionaceae and Rhodobacteraceae were the dominant families, contributing more than 29% and 23% to the community composition, respectively (Fig. 4a). In the SOLA-fall experiment, both B-DOM and U-DOM showed relatively high proportions of Alteromonadaceae, Flavobacteraceae, and Rhodobacteraceae, but the U-DOM treatment was characterized by the presence of Methylophilaceae (Fig. 4b). In both experiments, the G-DOM treatment was dominated by Vibrionaceae at the end of the incubation (24 to 83%). The NMDS ordination showed that T 0 samples of both experiments grouped together but showed separation among experiments at T f (ANOSIM, p value <0.05) and separation among treatments as replicates grouped together (ANOSIM, p value <0.05) (Fig. 4c). Therefore, we further studied the two experiments independently to identify which taxa are specifically associated to each treatment.
To identify which taxa were specifically associated with a DOM source, an indicator taxa analysis was run at the ASV level for each experiment separately. In the S. alaskensis experiment, the number of indicator ASVs in the B-DOM and U-DOM was 3 and 2 ASVs, respectively, while the Gand IS-DOM treatments had a higher indicator ASV number (4 and 27, respectively). B-DOM indicator ASVs accounted for 32%, on average, of the community in this treatment and consisted of 3 families: Alteromonadaceae, Marinomonadaceae, and a few Rhodobacteraceae (Fig. 5a). Indicator ASVs of U-DOM represented up to 23% of the U-DOM community, mainly belonging to the family Rhodobacteraceae and few Flavobacteriaceae (Fig. 5b). G-DOM indicator ASVs belonged to Vibrionaceae, Flavobacteriaceae, and Sphingomonadaceae, while the IS-DOM indicators were representative of several families, including Pelagibacteraceae and Flavobacteraceae (Fig 5c, d).
Indicator taxa analysis for the SOLA-fall experiment showed significant indicator ASVs for all treatments. The indicator ASVs of B-DOM (8 ASVs) represented up to 44%   mainly of Vibrionaceae and showed a high contribution to the other treatment communities at T f (14 to 45%). Even though many families were repeatedly shown as indicators in different treatments of both experiments (e.g., Rhodobacteraceae, Methylophilaceae, etc.), the ASVs were different as illustrated by the genus level taxonomy (Fig. S2). For example, Rhodobacteraceae were respectively represented by Litoreibacter and Pelagimonas in S. alaskensis U-DOM and B-DOM, while it was rather Thalassobacter and Pseudophaeobacter in SOLA-fall U-DOM and B-DOM, respectively.

Discussion
In line with previous studies [10,32], our experiments allowed us to confirm that a fraction of DOM released by heterotrophic prokaryotes is labile, promoting significant cell growth in all incubations. In contrast to our hypothesis, clear differences in DOM lability released under P-repletion and P-limitation (B-DOM and U-DOM, respectively) could not be evidenced based on cell growth, and P-limitation was not shown to decrease the lability of the released DOM.

FDOM Consumption
Although no clear differences in cell growth were observed between U-DOM and B-DOM, different patterns of FDOM consumption were detected. The protein-like component C 340 , usually considered a proxy of labile DOM [34,43], was only partially used on the time scale of our incubations. Lønborg et al. [32], using long-term incubations, suggested that a fraction of bacterial-derived protein-like FDOM could be refractory. However, in contrast to our initial hypothesis, the microbial humic-like component C 398 was consumed by the communities from the Mediterranean Sea in some of the incubations. This component has been considered refractory that resists bacterial degradation [29,30]; however, we have evidence that the microbial humic-like FDOM fluorophore released by HP is not always refractory and can be partially used when this DOM component dominates. Previous studies in marine and freshwater environments reported bacterial consumption of similar humic-like fluorescence (summarized in Table 3). Based on these studies, the humic-like fluorescence was used to a different extent (Table 3) depending on its molecular weight [46], source [44], or degrading community [10]. In our incubations, only 32.5 to 78.3% of the humic-like DOM was consumed. The percentages of HP degradation of microbial humic-like FDOM derived either from bacteria or phytoplankton (Table 3) are higher than the degradation of DOM directly isolated from natural waters [45,46]. This is most likely because humic-like DOM isolated from natural systems is more processed than DOM produced over short-time incubations. Together with our results, this also suggests that similar humic-like FDOM peaks could be of different bioavailability and likely different chemical composition since fluorescence spectroscopy captures only a bulk signal from DOM. Further DOM analyses (e.g., highresolution mass spectrometry) would be needed to elucidate this point.
Our results indicate that humic-like FDOM use is also related to the lack of other DOM compounds. In our experiment, the humic-like FDOM component was consumed only when the presumably labile protein-like component was initially not prevailing, implying that heterotrophic prokaryotes are able to use less labile carbon sources when the presumably labile ones are scarce. This is in line with the results by Xie et al. [48], who followed the bioavailability of Synechococcus-derived DOM to estuarine and coastal bacteria for 180 days and showed that a similar humic-like fluorophore can be produced by some bacterial groups when labile Synechococcus-derived DOM is available, then re-used by other bacterial groups when the labile DOM is depleted. Microbial humic-like consumption was also accompanied by a consistently high alkaline phosphatase activity. Although high alkaline phosphatase activity could be interpreted as a response to P-limitation, this was not the case in our incubations since inorganic P was added (Table S2). Studies demonstrated that HP could use alkaline phosphatases to unbind bioavailable organic carbon [49]. We hypothesize that alkaline phosphatase was used in our experiments to break down P-containing DOM polymers to use them as a substrate.

Selection of Community Composition According to HP-DOM Quality
HP-DOM quality driven by P availability appeared to change the taxonomic composition of the Mediterranean Sea communities growing on it. Both HP-DOM released under P-repletion and P-limitation (B-and U-DOM) lead to the growth of diverse communities, suggesting that HP-DOM is composed of a multitude of substrates. Previous studies using ultra-high resolution mass spectroscopy revealed a large number of molecules in bacterially derived DOM [25,50]. Landa et al. [51] discussed the different effects of DOM generated by diatoms and cyanobacteria on the composition and diversity of bacterial communities. According to their results, the more chemically diverse the DOM is, the higher the diversity of the community growing on it.
Even though HP diversity did not differ between B-and U-DOM, community composition analyses revealed different taxonomic groups growing on the different HP-DOM sources, indicating the different composition of the HP-DOM released under P-repletion vs. P-limitation. A few families, belonging mainly to Alphaproteobacteria and Gammaproteobacteria and a few Bacteroidia, dominated the HP-DOM incubations. However, the ASVs differed among the experiments and the treatments, suggesting that community composition plays an important role in determining DOM lability. Members of Rhodobacteraceae and Alteromonadaceae (Pseudophaeobacter, Alteromonas) were the main indicators of the protein-like rich treatment in SOLAfall B-DOM. Some members of the Alteromonadales class have already been shown to be related to the protein-like peak T, corresponding to our component C 340 , in bacterial incubations of high and low molecular weight DOM, suggesting their preference for labile DOM [46]. DOM rich in humic-like components, namely S. alaskensis DOM and SOLA-fall U-DOM, selected different indicator taxa, including mainly Methylophilaceae, Marinomonadaceae, Rhodobacteraceae, and Alteromonadaceae. A coupling between FDOM Peak M (corresponding to our C 398 ) and several members of Rhodobacterales was observed in communities growing on low molecular weight DOM [46] or on viral lysates of cyanobacteria [55]. Members of Marinomonadaceae, particularly Marinomonas, have been shown to degrade complex organic matter, such as terrestrial DOM inputs, in high latitudes [52]. However, they are also known to be associated with Cyanobacteria lysates [53]. This suggests that they can respond positively to various carbon sources, including complex DOM or humic-like DOM, as shown here. In addition, Alteromonadales, Burkholderiales (Methylophilaceae), and Rhodobacterales are among the most abundant heterotrophic marine bacteria that might hydrolyze dissolved organic phosphorus outside the cytoplasmic membrane using alkaline phosphatase [54], in line with our results showing high alkaline phosphatase activities in these treatments.

Conclusion
Overall, our results illustrate how P availability shapes HP-DOM degradation patterns, suggesting that P would play a key role in shaping HP-mediated DOM fluxes in the ocean. Although the growth of Mediterranean Sea communities did not differ between P-replete and P-limited derived DOM, different metabolisms and taxonomic groups were selected for the different DOM conditions. We could show that a considerable portion of humic-like DOM might be bioavailable when this DOM pool prevails as a substrate for HP. This bioavailability is also controlled by the release conditions and the consumer's community composition. Our results emphasize that DOM lability is context-dependent and imply that the impact of P-limitation on DOM fluxes via the MCP should be placed in its local environmental context rather than generalized to a global context.