Insights into nitrogen fixation below the euphotic zone: trials in an oligotrophic marginal sea and global compilation

Abstract. Nitrogen (N2) fixation, the energetically expensive conversion of N2 to ammonia, plays an important role in balancing the global nitrogen budget. Defying historic paradigms, recent studies have detected non-cyanobacterial N2 fixation in deep, dark oceanic waters. Even low volumetric rates can be significant considering the large volume of these waters. However, measuring aphotic N2 fixation is an analytical challenge due to the low particulate nitrogen (PN) concentrations. Here, we investigated N2 fixation rates in aphotic waters in the South China Sea (SCS). To increase the sensitivity of N2 fixation rate measurements, we applied a novel approach requiring only 0.28 μg N for measuring the isotopic composition of particulate nitrogen. We conducted parallel 15N2-enriched incubations in ambient seawater, seawater amended with amino acids and poisoned (HgCl2) controls, along with incubations that received no tracer additions to distinguish biological N2 fixation. Experimental treatments differed significantly from our two types of controls, those receiving no additions and killed controls. Amino acid additions masked N2 fixation signals due to the uptake of added 14N-amino acid. Results show that the maximum dark N2 fixation rates (1.28 ± 0.85 nmol N L−1 d−1) occurred within upper 200 m, while rates below 200 m were mostly lower than 0.1 nmol N L−1 d−1. Nevertheless, N2 fixation rates between 200 and 1000 m accounted for 39 ± 32 % of depth-integrated dark N2 fixation rates in the upper 1000 m, which is comparable to the areal nitrogen inputs via atmospheric deposition. Globally, we found that aphotic N2 fixation studies conducted in oxygenated environments yielded rates similar to those from the SCS (< 1 nmol N L−1 d−1), regardless of methods, while higher rates were occasionally observed in low-oxygen (< 62 µM) regions. Regression analysis suggests that particulate nitrogen concentrations could be a predictive proxy for detectable aphotic N2 fixation in the SCS and eastern tropical south Pacific. Our results provide the first insight into aphotic N2 fixation in SCS and support the importance of the aphotic zone as a globally-important source of new nitrogen to the ocean.


Zehr, J. P., and Capone, D. G.: Problems and promises of assaying the genetic potential for nitrogen fixation in the marine environment, Microbial Ecology, 32(3), 263-281, https://doi.org /10.1007/bf00183062, 1996. 3) Line 60: start a new paragraph with "However, owing to the..   Table 1. Compilation of N2 fixation rate studies conducted below the euphotic zone using 15 N2 methods. For studies without explicit depth of euphotic zone, only data ≥ 200 m were included. BDL stands for below detection limit. DOM represents dissolved organic matter, and DFAA, ATP, and TEP represent dissolved free amino acid, adenosine triphosphate, and transparent exopolymeric particle, respectively. DOM addition effect "+", "0", and "-" denote positive effect, no significant effect, and negative effect, respectively. Correction made.

Methods:
1) Fig. 1. Use SCS abbreviation for consistency since it has already been introduced.
Changes made.
2) Fig. 2. Not clear what OMD is referring to.
We updated the full explanation of OMD as oxygen minimum depth in the figure caption. We added one more line in the Table 1 to clarify the cruise dates for K1 (Aug 2018) and WXS (Jul 2019) and updated the treatment column and depth column to specify the depths of additional treatments as follows: Yes, and we made it clear by "Depth integrated N2 fixation rates were calculated by multiplying the average of two adjacent rates by their depth interval (trapezoidal integration)." 8) Table 2 would be more useful in the introduction rather than the long list of citations.
Changes made as 6) in Introduction part above.

Results:
1) Figs. 3 and 4: Consider reducing the y-axis range so it's easier to discern what's happening with most of the samples. For instances where d15N is greater than 100, consider using a broken y-axis so that you can highlight both high values and the majority of the data that falls within < 100.
Following your suggestions, we changed y-axes in Figure 3 and 4 as below for better visualization. Correction made as above in Methods 2).

Discussion:
1) Line 335: citations for low [PN] and low ANF rates in aphotic zone? Combine this paragraph with the next one since it's so short and interconnected. We improved the paragraph as following: "On the other hand, detectable BNF rates require high enough ∆δ 15 N-PN to exceed the natural and methodological variance, i.e. the minimum acceptable change (Montoya et al., 1996).
There are two ways to methodologically enhance ∆δ 15 N-PN. One is to increase incubation time. We updated the following explanations in the manuscript: "… heterogeneity of suspended particles, which could result from both sampling variability and intrinsic particle characteristics. On the one hand, both low concentration and limited sampling volume (1 -4 L) resulted in heterogeneity between different sampling bottles. This is also supported by Farnelid et al. (2018) that even though the bacterial community composition in bulk particles is consistent throughout time and space, large variations exist in individual particles. On the other hand, sinking particles could also be collected in our samples occasionally, which were reported to be heterogeneous in chemical composition (Martiny et al., 2013) and bacterial community composition (Boeuf et al., 2019). This could be due to the sporadic export events (Boeuf et al., 2019). The larger variances of δ 15 N-PN in the sample after the incubation compared to T0 time points further demonstrated the heterogeneity of ANF in deep waters (Fig. 3, 4 and S2). This conclusion is further supported by a model study showing a short ephemeral window for BNF in sinking particles, which is also variable due to above heterogeneity of particles (Chakraborty et al., 2021). How to …" brings labile organic matter from the surface to the mesopelagic, which together with the high mesopelagic temperature of this basin (~13ºC), would support ANF activity, by bringing fresh organic matter and stimulating metabolic activities of diazotrophs, respectively. 7) Line 380: "Horizontally" refers to "spatially" right?
Correction made.

8) Why were parallel light incubations not carried out in the euphotic zone? This would
have allowed you to determine how important diazotrophic activity by NCDs (ANF rates) are compared to cyanobacterial diazotrophic activity (light BNF rates).
The light incubations were conducted by other colleagues on the cruises and belong to another project. The results remain unpublished yet. More parameters from these two cruises will also be published in the future, but we cannot use them for now. We can further compare our results with those data after their publication.  (Fig. 7).
12) Line 415: Could you discuss how results might have been different with just a labile C source amendment over a DFAA addition? The DFAA also provides additional labile N to the microbial community but with just a labile C addition, N would be more limiting and could potentially result in a greater stimulation of the diazotrophic community.
We added this discussion in Line 408: "… DFAA additions. Thus, we suspect that in the SCS, labile organic carbon (glucose, carbohydrate etc.) addition could be a better option than labile organic carbon and nitrogen We incorporated this in Line 431: "… energy sources. Taking this into account, a mathematical model study quantified BNF by heterotrophic bacteria in sinking particles, which is determined by polysaccharide and polypeptide concentrations, particle sinking velocity, and surrounding O2 and NO3concentrations (Chakraborty et al., 2021). However, more field studies are needed for model verification and application to global ANF quantification." 14) Line 462: Could be incorporated as the last sentence for the paragraph rather than a separate paragraph.
Changes made.
15) Line 467-468: "Taken together with previous reports from different oceanographic regions highlighted in Table 2, our study shows 11 of 18 sampling depths..." Changes made as "However, taken together with previous reports from different oceanographic regions highlighted in Table 2 Another citation (with SEATS as overlapping station) that could be useful to compare how ANF rates contribute to net community production at SEATS: Estimated net community production during the summertime at the SEATS time-series study site, northern South China Sea: Implications for nitrogen fixation. Chou et al., 2006 We incorporated this citation and updated from Line 469: "The depth integrated (200-1000 m) ANF in SCS ranged from 7-86 μmol N m -2 d -1 (36 ± 26 μmol N m -2 d -1 ), which are comparable to non-hypoxia studies listed in Table 1 The suggestion for the comparison between ANF and net community production is difficult to conduct since Chou et al. (2006) only calculated this in the mixed-layer. ANF, i.e.
BNF below 200 m, only contributes to this net community production when transported upward, and belongs to the vertical diffusion term in Chou et al. (2006). Thus, we think that such comparison may distract the readers and deviate from our aim of highlighting the importance of ANF in section 4.5.
17) Line 477 -479: How is particle export in the summer compared to other seasons? Could you elaborate more why higher particle export in winter would indicate your annual estimation is conservative? Also, how do your annual ANF contributions to SCS compare to other estimates of BNF contributions to the N budget in SCS (if they exist)?
We answered seasonal particle export in Discussion-9). We further elaborated this in Line 477: "However, seasonal variability of organic sinking particle flux in the South China Sea is generally insignificant with occasionally higher fluxes observed in some depths in winter Kao et al., 2012;Liang, 2008;Yang et al., 2017). Thus, we would expect slightly higher rates in winter, with relatively stable ANF rates throughout the year, which indicates that our annual estimation may be the conservative estimate." Here our aim is to give an overview of global reported integrated ANF rates, and we would like to point out that the Mediterranean Sea study is the one with highest ANF contribution to water column integrated rates. To make this clear, we added the reference to Table 1 in the sentence: "Globally, reported integrated ANF rates ranged from below detection limit to around 600 μmol N m -2 d -1 (Table 1), contributing up to reported 100 % of water column BNF in Mediterranean Sea (Benavides et al., 2016)." 19) General discussion needs improvement and more pointed hypothesis or predictions.
With above comments, we improved this part with more in-depth discussions, more elaborate words and more references integrated in each section.

Conclusions:
1) Line 495: You didn't have any data suggesting high particle flux was occurring -in fact, you suggest that particle flux is much higher in the winter months. You could more appropriately discuss the potential for higher integrated ANF rates than measured in this study to occur in the winter if seasonal ANF surveys were conducted in the SCS.
We did not intend to suggest seasonal variability here. We bring up "high particle flux" here to explain spatial variation of ANF. We would like to elaborate this sentence to be: "Horizontally, high integrated ANF rates corresponded to high sinking particle flux in the SCS, which showed its potential control on ANF." 2) Paragraphs could be better organized and compiled into fewer paragraphs.
We re-organized the conclusion as follows: "Understanding the magnitudes, mechanisms and limiting factors of ANF not only extends potential niches for BNF, but also better constrains global marine nitrogen cycle. By using persulfate-denitrifier method, we provide solid evidences for ANF in the SCS, which is the firsthand data showing substantial contribution of ANF to new nitrogen inputs in the northwestern Pacific. Horizontally, high depth-integrated ANF rates corresponded to high sinking particle fluxes in the SCS. Compiled global ANF data also showed high depth-integrated ANF rates in a highly productive region -ETSP, further supporting the potential control of sinking particle on ANF rates. In some regions, detectable ANF had significant correlations with PN concentration, suggesting that easily measured PN could be a regional predictive parameter for ANF. However, this correlation is not ubiquitous globally, suggesting the heterogeneous and complex control of ANF.
We here list several recommendations for future ANF studies: 1) In order to detect low ANF rates, we recommend the use of the persulfate-denitrifier method to measure δ 15 N-PN when sufficient PN mass for the EA-IRMS approach cannot be achieved.
2) Both higher spatial and temporal resolution and corresponding complete datasets of PN, SNFR and ANF are needed to better constrain the controlling factors of ANF and the contribution of ANF to the global N budget.
3) Further rate studies coupled to molecular approach such as nano-SIMS coupled to catalyzed reporter deposition fluorescence in situ hybridization (CARD-FISH) are needed to bridge the knowledge gap between ANF rates and diazotrophs in the deep ocean."