| 25 Aug 2025
Spatiotemporal heterogeneity in diazotrophic communities reveals novel niche zonation in the East China Sea
Abstract. The East China Sea (ECS) is a hotspot for studying nitrogen fixation in the marginal seas of the western Pacific, where this microbially mediated process is profoundly influenced by both the coastal and oceanic current systems. Yet, how physical forcing controls the biogeography of diazotrophs and regional nitrogen budget in the ECS remains poorly characterized. Here, we carried out a cross-season survey and demonstrated dynamics in diazotrophic communities that is tightly linked to distinct water masses in the ECS. An overall spatial heterogeneity among some of the major diazotrophic ecotypes was unveiled, with the filamentous cyanobacteria Trichodesmium and diatom-diazotroph symbioses (Het-1 and Het-2) dominating the upper 30 m of the warm, saline, N-limited offshore water intruded by the Kuroshio and Taiwan warm current, whereas the unicellular cyanobacterial diazotrophs (UCYN-A, UCYN-B and UCYN-C) and the non-cyanobacterial diazotroph (γ-24774A11) extending their distribution further down to 60 m of the Kuroshio surface and subsurface waters. The diazotrophic abundances and nitrogen fixation rates were generally higher in autumn than in spring, suggesting a seasonal variability primarily regulated by hydrographic conditions (mainly temperature and salinity) associated with water mass movement. Modeling the distribution of diazotrophs in the water masses identified five taxon-specific niches occupied by eight distinct diazotrophic ecotypes. Taken together, our analyses provide mechanistic insights into the role of dominant forms of physical forcing in driving the spatiotemporal variability in diazotrophic distribution and activity in the ECS, which is of important reference in assessing diazotrophs adaptation in a changing marine ecosystem.
General comments:
Mai et al. investigated diazotroph abundance and activity in the East China Sea using qPCR and in situ incubations, and attempted to elucidate niche zonation with a maximum-entropy ecological model. While the data collection effort is acknowledged, the data analysis and discussion are regrettably inadequate. The modeling framework is introduced without sufficient explanation or validation, and the many results diverge from established patterns (e.g., this study suggests Trichodesmium favors low temperature and high nutrients). Inconsistent results with previous knowledge can be valuable, but the authors do not examine them critically and instead cite mostly only studies that support their observations. Consequently, the discussion remains largely descriptive and does not elucidate or discuss the physio-ecological mechanisms underlying diazotroph niche partitioning. Because reports of diazotrophy around the Kuroshio region and importance of Kuroshio intrusion are no longer novel, the modeling approach could have been the key contribution of this study. However, the absence of careful interpretation and the lack of clear validation leave the novelty and significance of the work unclear.
Major Comments:
1. Model justification and validation
The rationale for using the maximum-entropy model is not sufficiently explained, nor is its validity demonstrated. The authors should describe the model in more detail and clarify why it is appropriate for their dataset. Also I suppose that such models are not necessarily valid for all dataset, and thus model significance validation should be done carefully. For example, Brun et al. (2015) evaluated model performance/validity using the AUC of the ROC curve and a student’s t-test with randomizations to exclude invalid models. A comparable validation is needed here. Without it, the model results cannot be considered reliable.
2. Seemingly selective and sometimes inappropriate citations
References often appear cherry-picked to match the authors’ results. For instance, in discussing phylotype-specific niche partitioning, they cite only studies whose outcomes align with their data, ignoring meta-analyses and culture experiments that already provide robust estimates of optimal temperature and nutrient ranges (detail are mentioned in the specific comments). They should revise the citation thoroughly.
3. RDA analysis
The RDA explains little of the variance of dataset (autumn: ~12 %, spring: ~34 %), yet the discussion relies on such weak relationship. RDA compresses relationships among multiple variables, but authors use them to claim direct correlations between specific taxa and single environmental parameters from RDA results. Also, collinearity among variables such as SRP, NOx, and depth is not tested or reported, VIF threshold should be set and reported. The authors should (i) report overall R² and significance of their RDA, (ii) test the significance of each variable’s significance and contribution in RDA. More simple analysis such as simple correlation and multiple-variable linear regression such as GLM may be better for understanding controlling factor on each diazotroph.
Specific comments:
Introduction
L69–70: Note that Shiozaki et al. (2018) also examined diazotroph abundance using qPCR.
L68–73: Because authors study phylotype-specific niches, the introduction should briefly describe the major unicellular diazotrophs rather than grouping them simply as “unicellular.”
L77: The term "ecotype" generally refers to distinct strains within a species.
L81-84. Because most readers should not be familiar with maximum entropy method, authors should explain the method more detail.
L89-90: Brun et al. studied Trichodesmium and Richelia realized niche. It should be mentioned here and in the discussion.
Materials and Methods
L103: “summer” should be “autumn.”
L114: Anderson and Sarmiento (1994) did not introduce P*. To my knowledge, Deutsch et al. (2007) is the first.
L120: 200 µm mesh was used for RNA sample as well?
L122–123: Fig. 6 here is inappropriate, considering the order of figure.
L147: Please provide the detection limit of the qPCR.
L148–150: Recent studies show that UCYN-C (Schvarcz et al. 2022, 2024) and γ-24774A11 (Tschitschko et al. 2024) are also likely diatom symbionts, almost organelle-like. This simplification should be describe more carefully.
L167: For transparency as suggested in (White et al. 2020), report the minimum quantifiable N2-fixation rate for each data and the value for each replicate as supplementary material.
L179–183: Many readers will not be familiar with the MaxEnt model; provide a detailed description. Past studies sometimes exclude absence data, did this study do the same?
L179: Define “breadth (σ).”, what does this mean?
L180: After integration, how many observations (n) were included?
L181: All parameter settings of MaxEnt should be clearly stated. Is it default setting?
L182: The GAM parameters is not explained and GAM result does not appear here after.
L191: Z-score-scaling applied to PCA, as well?
L193–194: Please clarify the VIF threshold used to avoid multicollinearity.
Results
L214–216: Provide more specific spatiotemporal deviations from the Redfield ratio (16).
L222: Because nifH was not comprehensively amplified, “nifH gene pool of targeted diazotrophs” is more appropriate.
L223–225: Present quantitative numerical values.
L238: Transects A and B should be mentioned in Methods and figures 1.
L242–243: The term “broader” is vague; specify how many of how many stations, or give quantitative metrics.
L243: UCYN-A, UCYN-C, and γ-24774A11 are now recognized as likely symbionts/organelles; grouping them as “unicellular diazotrophs” is questionable.
L243–245: This description conflicts with Fig. 5. Het-2 peaks in the subsurface, and UCYN-A2/A3/A4 show profiles similar to Trichodesmium.
L246: Sampling was not performed at 60 m. If values are interpolated, please state that and use the actual sampled depth (e.g., 50 m).
L251: The phrase “were the most abundant among” is unclear; perhaps “were detected as the most abundant.”?
L269, L272: “CDW/CW-affected regions” should be defined quantitatively, e.g., by proportion of water masses.
L270: Figure 6 does not have station 1.
L271: “About 60%” is more appropriate.
L276–277: To support this, I recommend to perform multivariable regression for each phylotype and report the coefficients of determination, so that importance of each variable can be quantified.
L281: Figure S6 should be moved into the main text; otherwise readers cannot follow the discussion of μ and σ. Also, please evaluate model validity as mentioned in major comments.
L281: From Fig. S7, n = 3 seems most appropriate; consider adding a hierarchical cluster tree to illustrate the threshold objectively.
L290: “Relatively high” should be described with quantitative data.
Discussion
L295–297: Current ecological model (Fig. S6) shows Trichodesmium preferring low temperature, low salinity, and high nitrate waters. This contradicts both claim here and previous studies.
L298: growth was not measured in this study.
L300–301: Relationships inferred from RDA are weak and not checked stastically.
L303: “Below detection limit” does not mean complete absence.
L307–309: Conclusions based solely on weak RDA correlations are too qualitative.
L309–311: To demonstrate major N2 fixers, statistically test relationships between gene (or transcript) abundance and N2-fixation rate. High nifH abundance alone is insufficient.
L314–315: Many studies detected Trichodesmium colonies in the ECS (e. g. Marumo and Asaoka 1974, Jiang et al. 2023). Richelia colonies are also expected to removed by this mesh filtering. Was microscopy performed in this study? They should provide more solid evidence of not-underestimating, otherwise it is not convincing.
L319: Jiang et al. (2018, 2019) reported highest densities at the surface and 10–50 m, consistent with your results. Carpenter et al. (2004) also found maxima at ~20 m. Discussion here seems incorrect.
L323: If eddy is important, provide and check the SSH and its anomaly information during study period.
L324: Only UCYN-C and γ-24774A11 peak at 50 m; others peak above 30 m. Thus filamentous and unicellular diazotrophs are not “in contrast.”
L329: Lu et al. 2018 does not seem to provide the information of light adaptation of unicellular cyanobacterial diazotrophs?
L332: ~22℃ for optimal temp. of UCYN-B is not common understanding. Also, Webb et al 2009 report 26-30℃ for its optimal temperature. Other studies also reported ~25–30 °C (Tang and Cassar 2019; Mauda 2024).
L334–338: Tschitschko et al. 2024 in Nature strongly (almost conclusively) indicate that γ-24774A11 is a diatom symbiont. Include this information.
L346: Koening et al. (2009) only speculated about an r-strategy for Trichodesmium; they did not conclude it. Citing this paper as proof in current manner is inappropriate.
L351–353: This statement seems contradict the Maxent based discussion.
L360: Please quantitatively compare your data values with those of Jiang et al. 2023 with actual value.
L364–367: Only one phylotype of non-cyanobacterial diazotroph (NCD) was quantified. Many other NCDs likely exist; please acknowledge this limitation.
L367–368: A simple correlation analysis between N2-fixation rate and nifH transcript abundance is needed to support this argument.
L370: Recent studies report cell-specific rates for UCYN-C (Schvarcz et al. 2022) and γ-24774A11 (Tschitschko et al. 2024). Also, please provide standard deviations.
L375: The spring value of 58.5 seems incorrect; re-check Table S3.
L376: Because this study directly measured surface N2-fixation rates as well, authour should verify consistency with values inferred from nifH copies and per-cell rates and measured rates by incubations; otherwise the subsequent quantitative discussion can be not-reliable.
L399: Is this broad niche of Trichodesmium consistent with the earlier claim of an r-strategy for Trichodesmium?
L399–404: Of course there is intra- and inter-specific metabolic variability; explain how such variability could produce the observed differences.
L411–412: Zehr et al. (2007) did not study genotype-specific optimal temperatures. Webb et al. (2009) reported an optimum of 26–30 °C, which is not as broad and is inconsistently higher than current result (<24 °C).
L414: Brum et al. 2015 analyzed diatom-associated diazotrophs, please discuss it ?
K420-425: Please refer Tschitschko et al. 2024 which report impotant ecology and physiology of γ-24774A11.
L432: To my knowledge, Cheung et al. 2019 did not mention ecological niche similarity between UCYN-C and Hets.
L438–439: Cabello et al. (2016) did not report a distinct “reproductive strategy” for UCYN-A.
L440: Cabello et al. (2016) speculated on such a strategy, but many opposing studies also exist: UCYN-A2 is considered coastal ecotype, whereas UCYN-A1 is open-ocean ecotype (e.g., Turk-Kubo et al. 2017, 2021; Henke et al. 2018). Cite a more objective literature rather than only studies consistent with your results.
Figures
Figure 4: Clarify whether data represent an average across multiple stations and include error bars if so.
Figures 5 and 6: The numbers near the pie/bar charts appear to be station numbers—please clarify in the captions.
Figure 3: If station numbers are important here, add them for consistency.