Shao and Luo compile published abundance data of Gamma A (qPCR nifH gene counts), a putatively heterotrophic diazotroph widespread in the oceans. Using ancillary data, atlas, satellite products and models, they perform a thorough statistical analyses to infer relationships between Gamma A and environmental variables. Their results suggest that Gamma A benefits from primary production by-products and is mostly dominant in warm and iron-poor waters of the ocean. The data analyses are extensive and the results worth publishing. However, the authors should improve the comparison between their study and Langlois et al. 2015, who also compiled Gamma A data and performed statistical analyses to define their niche. How does the present study build up on previous ones? I also found several mis-citations, where the wrong citations are given to justify a statement or where the message of a given paper was not well understood. In all, the exercise seems statistically correct but of relatively poor ecological interpretation significance unless several points are improved. Below I provide a list of comments.

ABSTRACT

L7: Delete “the” in “to the global marine”.

L10: What is the carrying capacity? This term is used throughout the manuscript whereas it is never really explained.

L15: “in addition” not “in additional”.

L17: Eddies are not short-term features, they may last several months. Please choose another term.

L18: “organic matter” not “organic matters”

L19: Weird wording, please rephrase.

L20: “sampling” not “samplings”, and delete “better” from the end of the sentence.

INTRODUCTION

L32: “oxygen deficient zones”

L33: “heavy” sounds weird, please rephrase.

L33: I would not say heterotrophic N₂ fixation is “not well quantified”, it’s just not quantified at all. There is -currently- no assay able to isolate heterotrophic N₂ fixation from autotrophic N₂ fixation.

L41: This is not what these papers really say. In Benavides 2018b Gamma A was not detected, so it does not necessarily mean it was not stimulated by DOM, it just was not present in the samples at all. In Benavides 2015 N₂ fixation in dark waters was stimulated by amino acids.

L42-43: NCDs are thought to be attached to particles, but they haven’t been found to be attached to particles.

L45: Bonnet 2016 find that N released by Trichodesmium is taken up by diatoms. Not N released by NCDs.

L45: “to equip” sounds weird.

L52-53: How did those studies look at DIN inhibition of individual NCDs strains? It seems this is not what these studies really did.

L58: This is quite unfair to say, please cite:

Bombar, Deniz, Ryan W. Paerl, and Lasse Riemann. 2016. “Marine Non-Cyanobacterial Diazotrophs: Moving beyond Molecular Detection.” Trends in Microbiology 24 (11): 916–27.

Cornejo-Castillo, Francisco M., and Jonathan P. Zehr. 2020. “Intriguing Size Distribution of the Uncultured and Globally Widespread Marine Non-Cyanobacterial Diazotroph Gamma-A.” The ISME Journal. https://doi.org/10.1038/s41396-020-00765-1.

Langlois, Rebecca, Tobias Großkopf, Matthew Mills, Shigenobu Takeda, and Julie LaRoche. 2015. “Widespread Distribution and Expression of Gamma A (UMB), an Uncultured, Diazotrophic, γ-Proteobacterial NifH Phylotype.” PloS One 10 (6): 1–17.

Moisander, Pia H., Mar Benavides, Sophie Bonnet, Ilana Berman-Frank, Angelicque E. White, and Lasse Riemann. 2017. “Chasing after Non-Cyanobacterial Nitrogen Fixation in Marine Pelagic Environments.” Frontiers in Microbiology. https://doi.org/10.3389/fmicb.2017.01736.

Riemann, Lasse, Hanna Farnelid, and Grieg F. Steward. 2010. “Nitrogenase Genes in Non-Cyanobacterial Plankton: Prevalence, Diversity and Regulation in Marine Waters.” Aquatic Microbial Ecology: International Journal 61 (3): 235–47.

L59: change “can” to “may”.

L63: You may cite Benavides, M., and J. Robidart. 2020. “Bridging the Spatiotemporal Gap in Diazotroph Activity and Diversity With High-Resolution Measurements.” Frontiers in Marine Science 7. https://doi.org/10.3389/fmars.2020.568876.

L69: “suggesting” would be fairer than “revealing” here. Note that nif genes can be used for other purposes.

METHODS

L84: “the upper 100 m of the water column”.

L87: there is no common qPCR detection limit, it depends on the essay, the machine, the lab, the volume of water filtered.

Table 1: I am a bit puzzled at 0 m depths, this is unlikely. Please check.

L103-104: I wonder why an artificial neural network was considered for DOC concentrations when there is a now a global database available https://odv.awi.de/data/ocean/dom-compilation-hansell-et-al-2021/ Please reconsider using it instead.

L116: It is unclear how SLA data was used, data was extracted from the same days as Gamma A samples were taken? Please explain.

RESULTS AND DISCUSSION

L149: nifH abundance also decreases with depth in the North Pacific (see work from Church at station ALOHA).

L170: please explain what the carrying capacity is.

L172: Please cite Bombar 2016.

L175: How are biogeographic patterns biased by the sparse and uneven sampling in different ocean regions? Can this be assessed statistically?

L190-191: This negative correlation is because low temperature anticorrelates with NPP, right?

Section 3.4. The first sentence belongs in the methods. Why even show linear regressions at all if the model is deemed better? I would suggest just mentioning the correlations, maybe move them to the supplementary, and dive into the GAM directly in the main text. Why are, in any case, the effects found using GAM so different to the ones obtained with linear correlations? (e.g. L219).

L223: fuel with what?

L224: I suggest replacing Farnelid 2019 for Riemann 2010.

L235: This seems quite a speculative conclusion to make. DOC concentrations alone do not inform about lability, and, to date, we don’t know anything about the metabolism of Gamma A or which kind of DOM molecules they may use.

L243-244: note that NCDs also need P.

L301: this seems quite different from observations.

Figure 6: why is the abundance “annual”? it’s not a rate.

Section 3.5. the connection or justification of why the effect of SLA is tested here is hard to follow.

L316: eddies are not short term phenomena.

L320: but also the number of data points in the NH is much higher than in the SH, potential bias, how can it be assessed?

SUMMARY AND OUTLOOK

L367: “confirming heterotrophy” seems quite risky. We need genomic data and tracer experiments to confirm that.

L379: Unclear here, nifH primers are universal. There is no primer for cyanobacterial diazotrophs only. These primers target all diazotrophs with Mo nitrogenases.

Shao and Luo attempt to better constrain the environmental drivers behind the observed biogeography of gamma A, a cosmopolitan marine non-cyanobacterial diazotroph group, using a metadata analysis of previously published gamma A abundances (estimated using qPCR targeting the gamma A nifH gene along with a suite of environmental parameters derived from the world ocean atlas, MODIS and several model outputs. On this whole, this represents a valid and interesting approach to gain insight into gamma A, but there are many items I feel need to be addressed prior to being considered for publication.

In general, I am concerned with ignoring all the apparent 0s in the compiled qPCR dataset. Better justification is needed for doing this – including why we should assume that abundance data would be normally distributed (in my experience with this type of data, it certainly isn’t always), and better justification for the stated assumption that these undetects are not true 0s due to primer specificity. Is there any precedent for ignoring 0s in other published work that uses GAMs or other similar analyses?

I also find much of the discussion to be speculative, esp. when trying to relate these findings to the broader group of gamma proteobacterial diazotrophs, or NCDs in their entirety.

I suggest sticking with non-cyanobacterial diazotrophs throughout in place of “heterotrophic diazotrophs”.  

Abstract ----

Line 7 – First sentence is awkward. Perhaps “non-cyanobacterial diazotrophs (NCDs) may be contributors to global marine….”  

Line 10 – This needs definition since this is not a commonly used term for this sort of data. Is this even the right term to be used here and throughout? Aren’t you really talking simply about abundance?

Line 15 – because the GLMs only explain some of the variance in gamma A abundances, I suggest using less definite terms here and throughout, e.g. in line 18 “most likely determined by” to “influenced by”, etc.

Line 17 – “mesoscale” in place of “short-term”

Line 18 – “matter” in place of “matters

Line 19 – “provide insight into” in place of “insight a”

Introduction ----

Line 25 – remove heterotrophic here

Line 26 – these aren’t the best papers to cite here

Line 28 – “had higher relative abundances than” in place of “were far superior in number to”

Line 30 – remove “dominant” or find a way to rephrase

Line 31 – “have also been” in place of “was also”

Line 33 – rephrase “heavy”

Line 33-36 – Marine N2 fixation by NCDs is not quantified at all – please rephrase and make it clear that there is only indirect evidence, including nifH transcription which does not “support” active N2 fixation by NCDs at all, it only provides another line of indirect evidence.

Line 36-37 – in addition to being uncultivated there are likely diverse niches and metabolic strategies used by this broad group. I might move the paragraph beginning at line 64 up, so that you can make this point and introduce gamma A earlier.

Line 38 – “Apparently” is awkward here – remove. Add “presumably” before “depending”

Line 41 – this is misleading – there was no gamma A in Benavides et al. 2018, and they did not assess other NCDs in this study

Line 46 – Bonnet et al., citation makes no sense here.

Line 46 – “equip” is awkward

Line 55 – is “supposably” needed here?

Line 69 – as above, gamma A nifH transcription doesn’t “reveal” it’s important role in marine N2 fix

Line 71 – state here that this data is compiled from nifH-based qPCR studies.

Line 85 – zero nifH copies can also be true zeros

Line 87 – all studies have different detection limits based on filter volumes, extraction volumes, the amount of template used in the qPCR, etc. This is misleading.

Line 95 – replace “were” with “have been” and you should note that these studies are specific to cyanobacterial diazos, and we do not know gene copy #s in gamma A.

Table 1 – There are additional studies represented in the Figure S5, it seems? These should be listed in Supp.

Line 118 – More description needed about how cyclonic and anticyclonic eddies were called. What does a “clear shape” mean? Why is SLA missing from Table 2?  

Results and Discussion -----

Line 148 – maybe specify it was undetected in this SPOT sample?

Line 150 – there are other studies that describe the depth distribution patterns of gamma A, eg.  Chen et al., 2019, which seems to be missing from your list of studies???

           Chen, Tien-Yi, et al. "Community and abundance of heterotrophic diazotrophs in the northern South China Sea: revealing the potential importance of a new alphaproteobacterium in N2 fixation." Deep Sea Research Part I: Oceanographic Research Papers 143 (2019): 104-114.

Figure 2 caption – it’s not clear why some data was singled out as “highest” and shown with red dots, while other high datapoints were left out – much better description needed.

Section 3.2 and elsewhere – as above, I wonder whether carrying capacity is a needed term – at minimum it needs to be better defined, especially since the term has ecological ramifications that I am not sure are relevant here.

Line 171 – “gamma A is expected to require a sufficient…”

Line 208 – not clear why linear correlations are needed if the GAM is more reliable.

Line 219 – “is presumed to be” in place of “was supposably”

Line 235 – too speculative

Section 3.4.5 – although this relationship is interesting, this discussion is speculative, thus needs to be better phrased – e.g. interpreting this as “indirect” evidence supporting the hypothesis that some NCDs are motile is misleading.  

Line 275 – Abundance does not equal active N2 fixation. No evidence that gamma A fixes anywhere, including the mesopelagic. Needs rewording.

Line 292 – was this described in the methods? More details would be helpful.

Line 300 – I would begin this discussion with an emphasis that the model predicts high abundances where gamma A is not observed, like the Southern Ocean and coastal areas.

Line 305 – remove “where”

Line 354 - I'm confused why the PCR bias is mentioned here - there is no end-point PCR data included in this study. I think a more relevant discussion could include unknown copy #s in gamma A’s genome, or even accuracy of qPCR in general, due to the reliance on standard curves.

Line 359 - N2 fixers have been shown to be very patchy in space and time, see Robidart et al., 2014.

Robidart, Julie C., et al. "Ecogenomic sensor reveals controls on N 2-fixing microorganisms in the North Pacific Ocean." The ISME journal 8.6 (2014): 1175-1185.

Line 367 – remove “confirming its heterotrophy” – over interpretation.

Line 368 – replace “include” with “suggest” or the like

Line 379 – there are many “universal” nifH primers with varying performance – do you mean a universal qPCR assay (which is unrealistic and would be difficult to interpret data from)?