Diurnally dynamic iron allocation promotes N2 fixation in marine dominant diazotroph Trichodesmium

Trichodesmium is the dominant photoautotrophic dinitrogen (N2) fixer (diazotroph) in the ocean. Iron is an important factor limiting growth of marine diazotrophs including Trichodesmium mainly because of high iron content of its N2-fixing enzyme, nitrogenase. However, it still lacks a quantitative understanding of how dynamic iron allocation among physiological processes acts to regulate growth and N2 fixation in Trichodesmium. Here, we constructed a model of Trichodesmium trichome in which intracellular iron could be dynamically re-allocated in photosystems and nitrogenase during the daytime. The results demonstrate that the dynamic iron allocation enhances modeled N2 fixation and growth rates of Trichodesmium, especially in iron-limited conditions, albeit having a marginal impact under high iron concentrations. Although the reuse of iron during a day is an apparent cause that dynamic iron allocation can benefit Trichodesmium under iron limitation, our model reveals two important mechanisms. First, the release of iron from photosystems downregulates the intracellular oxygen (O2) production and reduces the demand of respiratory protection, a process that Trichodesmium wastefully respires carbohydrates to create a lower O2 window for N2 fixation. Hence, more carbohydrates can be used in growth. Second, lower allocation of iron to nitrogenase during early daytime, a period when photosynthesis is active and intracellular O2 is high, reduces the amount of iron that is trapped in the inactivated nitrogenase induced by O2. This mechanism further increases the iron use efficiency in Trichodesmium. Overall, our study provides mechanistic and quantitative insight into the diurnal iron allocation that can alleviate iron limitation to Trichodesmium.


Introduction
Trichodesmium is a dominant cyanobacterial dinitrogen (N 2 )fixing microorganism and is widely distributed in tropical and subtropical oligotrophic oceans [1]. Iron (Fe) is a key micronutrient needed by Trichodesmium mainly for its photosystems and nitrogenase (the enzyme catalyzing N 2 fixation) [2][3][4][5]. Given that the nitrogenase contains high amount of Fe (2 Fe-protein dimers and 1 MoFe protein per complex) [2] and that the solubility of Fe is low in the ocean [6], Fe is generally a limiting factor for N 2 fixation and growth of Trichodesmium [7][8][9].
Many studies have investigated the response of Trichodesmium to Fe limitation. Trichodesmium generally grows slower and fixes less N 2 fixation with intensifying Fe limitation, mainly attributed to the downregulation of Fe quota in both photosystems and nitrogenase [8,[10][11][12][13][14][15][16]. Furthermore, some studies have shown that Trichodesmium can change the allocation of intracellular Fe in different physiological processes under varying ambient Fe concentration [8,10,16]. Under ocean acidification, the low pH decreases the efficiency of nitrogenase, causing the upregulation of nitrogenase under both Fe-depleted and Fe-replete conditions [8,10,16].
The majority of these findings of Trichodesmium were based on observations without resolving diurnal variations in Fe quota in photosystems and nitrogenase. An exception is the study by [10], showing that the level of photosynthetic proteins decreases and that of nitrogenase increases in Trichodesmium cells during the daytime. These results suggest the existence of diurnal reallocation of Fe in photosystems and nitrogenase in Trichodesmium, which may increase Fe use efficiency of the organism particularly under Fe limitation.
Trichodesmium also faces another challenge that its nitrogenase is irreversibly inactivated by oxygen (O 2 ) [17] especially because it performs oxygenic photosynthesis and N 2 fixation concurrently during the daytime [1]. It has been proposed that Trichodesmium can solve the conflict by temporally segregating photosynthesis and N 2 fixation in different phases and/or spatially segregating them among cells [9,18], while a recent model study suggests that the spatial segregation may be not necessary [19]. The pattern of diurnal variations in Trichodesmium photosystems and nitrogenase mentioned above [10] is consistent with some observed temporal segregations, where photosynthetic and N 2 fixation rates of Trichodesmium were elevated in early and late daytime, respectively [17,20]. These findings tentatively suggest that diurnal Fe reallocation may also play a role in intracellular O 2 management in Trichodesmium, indirectly regulating photosynthesis and N 2 fixation.
Nevertheless, it still lacks a systematical understanding of how and to what degree Trichodesmium can benefit from the diurnal Fe allocation. A recent model simulated diurnal fluxes of carbon, nitrogen, O 2 , NADPH (nicotinamide adenine dinucleotide phosphate hydrogen) and ATP (adenosine triphosphate) in Trichodesmium, but did not simulate its intracellular Fe [19]. In this study, we improved this model by incorporating major intracellular Fe pools including photosystems, nitrogenase, storage and maintenance, so that the diurnal cycles of photosynthesis, N 2 fixation and O 2 management could be modulated through the diurnally dynamic allocation of Fe in different pools. The improved model was simulated under both Fe-replete and Fe-depleted conditions. Model experiments were also conducted with diurnally fixed Fe in different pools. The comparison of model results provided us with a mechanistic and quantitative understanding of the role of dynamic Fe allocation in Trichodesmium. This framework is expected to improve the projections of N 2 fixation by this dominant diazotroph in the ocean with variations in Fe availability over space and time.

Methods
The model in this study was constructed by modifying a previous model [19] with new representations of intracellular Fe pools and Fe-related processes. In the following, we briefly describe the model schemes focusing on the new schemes, while the full model description, parameter values and variables can be found in Supplementary Methods and Tables S1 to S4.

General model framework
The model (Fig. 1) simulates diurnal cycles of carbon assimilation into carbon skeleton and carbohydrate and N 2 fixation in Trichodesmium trichome. The rates of these processes are controlled by dynamic allocations of Fe, ATP and NADPH across different metabolic processes and by the regulation of intracellular O 2 . The model simulates daily growth rate of Trichodesmium as a function of dissolved inorganic Fe concentration (Fe′) in the extracellular environment. The model runs for 12 h light period, while biosynthesis is designed to occur during the dark period and the temporal variation of the biosynthesis is not simulated [20]. Instead, the accumulated carbon skeletons, carbohydrates and fixed N at the end of the light period are used to calculate the synthesized biomass and the growth rate. The model parameters are optimized to maximize the growth rate.
Photosynthetic electron transfer (PET) includes linear PET (LPET) and alternative electron transfer (AET) in our model. Both LPET and AET produce ATP, while NADPH required by carbon and N 2 fixations is only produced by LPET [21,22]. ATP is consumed by carbon and N 2 fixations, CCM, maintenance, and biosynthesis. N 2 fixation can occur only when the intracellular O 2 concentration is low. The O 2 produced by LPET [21,22] can physically diffuse between cells and the extracellular environment. O 2 can be also consumed by respiratory protection, a process that actively respires carbohydrates to consume intracellular O 2 and protect N 2 fixation, with the produced energy lost in the form of heat to the environment [17,19,[23][24][25].
The model calculates the total intracellular Fe from Fe′ using a previous model scheme [16] based on observations [10]. A portion of intracellular Fe is stored and does not involve in metabolic processes, while a small, fixed portion of metabolic Fe is used in maintenance (Fig. 1B). The remaining Fe is dynamically allocated among photosystems, active and inactivated nitrogenase, and buffer during the daytime (Fig. 1B). Active nitrogenase is inactivated by intracellular O 2 [17]. The decomposition of nitrogenase seems to occur at night [26] and therefore is not considered during the light period in our model. That is, the Fe in the inactivated nitrogenase is not reused in the light period. This model case with diurnally dynamic Fe allocation is referred to as the "dynamic-Fe" case hereafter.
To further quantitatively explore the role of dynamically diurnal Fe allocation, we set up another model case with fixed Fe allocation (referred to as "fixed-Fe" case), in which the Fe in photosystems, To simplify the plot, arrows that represent the ATP production by PET and ordinary respiration, the production NADPH by LPET, and the NADPH consumption by carbon and N 2 fixations (dotted frames) are omitted. ATP produced by RP is wasted as heat and not counted. The pentagrams with Fe indicate processes requiring Fe. Dark orange frame, cell membrane; oval blobs, biochemical pools; rectangles, metabolic processes; solid arrows, mass or energy fluxes; dashed arrows, inhibition effects of RP on PET (gray) or O 2 on N 2 fixation (blue). LPET: linear photosynthetic electron transfer; AET: alternative electron transfer; NADPH: nicotinamide adenine dinucleotide phosphate hydrogen; ATP: adenosine triphosphate; DIC: dissolved inorganic carbon; CCM: CO 2 concentrating mechanism; CF: carbon fixation; NF: N 2 fixation; RP: respiratory protection; RESP: ordinary respiration; CH 2 O: carbohydrate; CS: carbon skeleton; N: fixed nitrogen; MT: maintenance; BIO: biosynthesis; G: growth rate. (B) The model represents fundamental intracellular Fe pools, including storage and five metabolic pools (photosystems, active and inactivated nitrogenase, maintenance, buffer and storage). Note that all inactivated nitrogenase by O 2 does not decompose and Fe in it is not recycled during model period (daytime). nitrogenase and buffer are constant during the modeled light period, except for the O 2 -induced inactivation of the active nitrogenase.
Both model cases were run under a Fe-depleted condition (Fe′ = 40 pM) and a Fe-replete condition (Fe′ = 1250 pM).
is the initial slope of PET versus light curve, and [mol C (mol C) -1 s] represents the degree of the inhibition from RP on PET.
The PET electrons are further differentiated into two pathways: LPET and AET. The fraction of electrons flowing into LPET and AET is calculated at each time step to fulfill the immediate intracellular demands for ATP and NADPH, as LPET produces both ATP and NADPH while AET only produces ATP [19].

N 2 fixation
N 2 fixation requires both ATP and NADPH [27,28]. The maximal potential of N 2 fixation rate [V NF max , mol N (mol C) -1 s -1 ] is determined when the produced ATP and NADPH from PET are completely consumed by N 2 fixation [19].
The N 2 fixation rate [V NF , mol N (mol C) -1 s -1 ] is also limited by the Fe quota in nitrogenase [Fe NF , μmol Fe (mol C) -

Carbon fixation
Carbon fixation also requires both NADPH and ATP [30]. The carbon fixation rate is solved at each time step assuming that total NADPH and ATP production by PET are immediately and fully utilized by intracellular process [19]. Carbohydrates, produced by carbon fixation, stimulate the production of carbon skeletons which are downregulated by its own accumulation.

Respiratory protection and O 2 diffusion
RP is a process wastefully respiring carbohydrates in order to reduce intracellular O 2 for supporting N 2 fixation. The RP rate increases with the requirement for N 2 fixation but decreases as intracellular O 2 level rises [19].
The rate of O 2 diffusion between intracellular cytoplasm and ambient environment is parameterized by adopting the scheme of [31].

Intracellular Fe pools and translocation
Trichodesmium can take up more Fe than that required for its metabolism (called 'luxury uptake') especially in high-Fe environments, and the excess Fe is stored for surviving in low-Fe environments [32,33]. The fractions of total intracellular Fe used in metabolism and storage are calculated using a previous scheme [16]. ] is the maximal N storage. Hence, this scheme assumes that the increase of light and the production of CS stimulates the requirement of N 2 fixation while the accumulation of fixed N lowers the requirement.
Active nitrogenase is inhibited upon exposure to O 2 , flowing into the pool of inactivated nitrogenase [17] at the rate [μmol Fe (mol C) -1 s -1 ]: where T NF NA max [μmol Fe (mol C) -1 s -1 ] is the maximal inactivation rate of nitrogenase. It should be noted that in the fixed-Fe model case, photosystems and total nitrogenase are set diurnally constant, but the inactivation of nitrogenase still occurs. The Fe in the inactivated nitrogenase is not released for reuse during the model period, as discussed above.

Model parameter values
Some model parameters were optimized to maximize Trichodesmium growth rate (Table S1). In the fixed-Fe case, three parameters, including the maximal respiratory protection rate (v RP max ), the initial fraction of metabolic Fe in photosystems ( f Fe PS 0 ), and the initial fraction of metabolic Fe in nitrogenase ( f Fe NF 0 ), were optimized (Table S1). In the dynamic-Fe case, f Fe PS 0 and f Fe NF 0 were not optimized but were set using observations in a laboratory culture study [10]. In addition to v RP max , the maximal synthesis (T PS BF max ) and decomposition (T BF PS max ) rates of photosystems and the maximal synthesis rate (T NF BF max ) of nitrogenase were also optimized (Table S1). The optimization was conducted using the global optimizer Multi-Start in MATLAB.
Other parameters that were not optimized (Table S2) were either adopted from previous studies or derived from our constant-light model experiments that fit the observed growth and N 2 fixation rates and the observed diurnal Fe in photosystems and nitrogenase [10] (see Results). The Fe in nitrogenase and photosystems from [10] were estimated from observed protein content, based on Fe atoms in per protein. PSII, Cyt b6f, PSI and Ferredoxin together represents photosystems. Cyt b6f and Ferredoxin were not measured in [10] but estimated by assuming Cyt b6f:PSII = 1:1 in Fe quota and Ferredoxin:PSI = 1:1 in protein content. Further details are in the supplementary information in [16].

Simulated growth rate, carbon and N 2 fixation rates and O 2 concentration
We first used the observational data from a culture experiment [10] to constrain and evaluate the model results. Because the light intensity was constant in this culture experiment, we used constant light intensity (90 μmol m -2 s -1 ) and dynamic-Fe case in this model exercise, so that the model results and the observational data can be compared. By tuning values of some model parameters (Table S2), the model results well fitted the observed daily growth rates (Table 1). Additionally, the model also captured the diurnal variations in the observed photosystem and nitrogenase Fe pools (for photosystem Fe, R 2 = 0.47 and 0.93 under low and high Fe, respectively; for nitrogenase Fe, R 2 = 0.53 and 0.82 under low and high Fe, respectively) (Fig. 2). The observations and the model both showed that nitrogenase increased and photosystem proteins decreased over the day period under the low-Fe and high-Fe conditions. These results partly support the robustness of our model. We used these tuned model parameters in our following simulations, with the exception of those key model parameters optimized for maximal growth rates (see Section 2.7 and Table S1).
We then ran the model with diurnally dynamic light intensity using the sine function during a 12-hour light period [6] to simulate more natural conditions. The simulations were extended to ten Fe levels ranging from 20 pM to 1800 pM. The model results showed that N 2 fixation and growth rates of Trichodesmium increased with the increase of the Fe concentration (Fig. 3), which generally captured the trends of observed N 2 fixation and growth rates in previous culturing experiments [8]. Furthermore, with increasing Fe concentration, the impact of dynamic Fe allocation in promoting growth rates decreased from 21% to 3% (Fig. 3), highlighting that the benefit of the dynamic Fe allocation tended to be more pronounced under low Fe and could be marginal under high Fe.
In the following, we analyzed the effect of dynamic Fe allocation by comparing the dynamic-Fe and fixed-Fe model cases using diurnally dynamic light intensity. We focused on the simulation results at two Fe levels (40 pM and 1250 pM) that were used in the laboratory experiments [10]. Although the growth rates were higher in the dynamic-Fe cases than in the fixed-Fe cases, the gross carbon fixation rates in the former were even lower under both low-and high-Fe conditions ( Table 2). In other words, the dynamic Fe allocation improved the carbon use efficiency (ratio of net to gross carbon production) ( Table 2). Additionally, the simulated higher carbon use efficiency under higher Fe was also consistent with previous studies [13,34].
The modeled carbon and N 2 fixation rates exhibited diurnal variations (Fig. 4). Under low Fe, carbon fixation in the two model cases (Fig. 4A) increased and reached a maximal level within the first 1.5 h, decreased by approximately 50% in the next 1.5 h, maintained nearly unchanged for another 6 h, and then decreased again until it ceased at the end of the daytime. N 2 fixation mainly occurred during the midday and late light period (Fig. 4C) when carbon fixation was downregulated (Fig. 4A) and intracellular O 2 was low (Fig. 4E). Compared to the fixed-Fe case, the carbon fixation rate during the low-O 2 window period was slightly lower in the dynamic-Fe case (Fig. 4A, E); the N 2 fixation rate in the dynamic-Fe case peaked later but at a higher level and thus achieved a higher daily-integrated rate ( Fig. 4C and Table 2). Under high Fe, modeled carbon and N 2 fixation rates were higher than those under low Fe, while their diurnal patterns were similar under both Fe levels ( Fig. 4A-D).

Simulated diurnal Fe allocation
In the dynamic-Fe case, initial levels of Fe in photosystems and nitrogenase at the beginning of the light period were set based on observations in [10] under both low and high Fe. Fe in photosystems decreased in the whole daytime, with the decrease rate being slowed after 3 h (Fig. 5A, B). This decomposition of photosystem Fe can then supplement the buffer pool in the dynamic-Fe case (Fig. 5I, J). Active nitrogenase was continuously inactivated during the whole light period in both model cases, while the inactivation was slower in middle and late daytime when intracellular O 2 was low (Fig. 5G,  H). Consequently, active nitrogenase decreased over time in the fixed-Fe case (Fig. 5E, F). In the dynamic-Fe case, however, although initial nitrogenase was less than that in the fixed-Fe case, Fe from buffer pools could support continuous synthesis of new nitrogenase (Fig. 5C, D) and the active nitrogenase became more than that in the fixed-Fe case during the period of active N 2 fixation (3 −7 h) (Fig. 5E, F). It was worth noting that in the dynamic-Fe case, because of the lower level of nitrogenase in the early light period (Fig. 5C, D), less Fe was trapped in the inactivated nitrogenase in most time during the day (Fig. 5G, H), promoting overall Fe use efficiency in the model. This diurnal pattern of Fe pools largely determined the modeled photosynthesis and N 2 fixation. During the early daytime (0 −3 h), large and close amount of Fe in photosystems in the fixed-Fe and dynamic-Fe cases (Fig. 5A, B) ensured high and similar photosynthesis rates (Fig. 4A, B). After 3 h, the reduced Fe in photosystems in the dynamic-Fe case (Fig. 5A, B) resulted in a lower photosynthesis rate than that in the fixed-Fe case (Fig. 4A, B). In the dynamic-Fe case, nitrogenase can start at a very low level because of unfavorable condition for N 2 fixation in the morning (0 −3 h) (Fig. 4E, F; 5 C, D); in approximately 3 −4 h, higher levels of active nitrogenase supported higher N 2 fixation rate in the fixed-Fe case (Fig. 4C, D; 5E, F). However, with the synthesis of nitrogenase in the dynamic-Fe case (Fig. 5C, D), its active nitrogenase exceeded that in the fixed-Fe case after 4 h (Fig. 5E, F), achieving a higher N 2 fixation rate (Fig. 4C, D). The accumulated fixed N in the dynamic-Fe then became more than that in the fixed-Fe case after 7 h (Fig. S1E).
Comparing the low-Fe and the high-Fe simulations, the Fe allocations presented similar patterns, except for higher magnitudes of variations in the high-Fe simulation (Fig. 5). In the dynamic-Fe case, a higher fraction of metabolic Fe was allocated to initial nitrogenase under high Fe than under low Fe (41% versus 15%) (Fig. 5D). Consequently, the relative difference in inactivated nitrogenase between the fixed-Fe and dynamic-Fe cases was smaller under high Fe than under low Fe (Fig. 5G, H), which weakens the potential role of dynamic allocation in promoting N 2 fixation and growth rates ( Table 2).

Discussion
In this study, we constructed an eco-physiological model of Trichodesmium trichome to quantitatively study how its intracellular Fe was diurnally allocated to photosynthetic and N 2 -fixing apparatuses under Fe limitation and repletion during the light period (Fig. 1). The model also integrated intracellular management including the formation of the temporal segregation between photosynthesis and N 2 fixation and the creation of the low-O 2 window (Fig. 4). The model results reproduced the observed diurnal variations of Fe in photosystems and nitrogenase of Trichodesmium [10], showing that photosystems decrease while nitrogenase increases continuously during the daytime (Figs. 2 and 5). It also captured the general diurnal patterns of carbon and N 2 fixations found in some observations of Trichodesmium [17,20]: high photosynthesis and low N 2 fixation in the morning, and moderate photosynthesis and high N 2 fixation in the noon and afternoon. The model results showed that the dynamic Fe allocation moderately increased the Trichodesmium growth rates (Table 2).
We also designed a new model experiment using the dynamic-Fe case but optimizing the initial Fe in photosystems and nitrogenase for maximal growth rate, instead of setting the initial Fe to observed values. Compared to the standard dynamic-Fe case, this setup further increased the growth rate by 18% and 10% under the low and high Fe, respectively. In other words, if the initial Fe pools could be more flexible, the dynamic Fe allocation would generate even higher potential of benefit for Trichodesmium (Figs. S2−4).

Dynamic Fe allocation decreases the requirement of respiratory protection
In the dynamic-Fe case, reduced Fe in photosystems after 3 h (Fig. 5A) downregulated the rate of O 2 production by 6% and 3% compared to the fixed-Fe case under low Fe and high Fe, respectively (Fig. 6A, B). Hence, the respiratory protection needed to create a low O 2 window for N 2 fixation decreased by 9% and 5% (Fig. 6E, F). Respiratory protection has been proposed as an important intracellular O 2 management strategy in Trichodesmium and other marine nonheterocystous diazotrophs [17,19,[35][36][37]. However, respiratory  [10]. The model is simulated with dynamic Fe allocation under both low-Fe and high-Fe conditions. Light intensity in the model is diurnally constant as that in [10]. Error bars represent one standard deviation. protection is generally a high indirect cost for N 2 fixation, as evidenced by observed high daily-integrated gross fixed C:N ratios (e.g., 30 -50 under high Fe) [17,20,38,39]. The process consumes a large fraction of gross produced carbohydrate (Table 2) without supplying energy for metabolic processes [17,[23][24][25]. This indicates that dynamic Fe allocation might play a potential role in improving the carbon use efficiency of Trichodesmium by downregulating photosystems and photosynthetic O 2 production during the period of active N 2 fixation (Table 2).

Dynamic Fe allocation promotes Fe use efficiency
Our model results revealed a mechanism that dynamic Fe allocation promoted Fe use efficiency for N 2 fixation and growth in Trichodesmium. The dynamic Fe allocation allowed for relatively low Fe allocation to nitrogenase during the early daytime (Fig. 5C, D) when photosynthesis was active (Fig. 4), so that less Fe was trapped in the nitrogenase inactivated by high intracellular O 2 during this period (Fig. 5G, H).
In our further model experiment in which the inactivated nitrogenase was instantaneously decomposed and its Fe returned to the buffer pool immediately, the dynamic-Fe case then did not simulate higher growth rates than the fixed-Fe case. However, the results of this model experiment were related to the initial Fe of photosystems and nitrogenase using the results from a previous culture experiment [10] (Figs. S2−4). When the initial Fe in photosystems was maximized to 90% of total metabolic Fe (the rest 10% Fe was used in maintenance) and the instant decomposition of inactivated nitrogenase was implemented, the dynamic-Fe case could still generate substantially higher growth rates (9% and 4% under low Fe and high Fe, respectively) than the fixed-Fe case. That was mainly because the higher initial content of photosystems promoted carbon fixation and more carbohydrates were accumulated in early daytime (0 −2 h), and the large amount of Fe in photosystems was then released to support the synthesis of nitrogenase after this period (Fig.  S5). Nevertheless, a previous study showed that the decomposition of nitrogenase did not occur during the light period [26]. The benefits from the dynamic Fe allocation shown in the standard model cases likely occurs in Trichodesmium. Meanwhile, the timescales of the recovery of nitrogenase by de novo synthesis need to be further explored to better evaluate the role of dynamic Fe allocation in Trichodesmium.
In summary of these analyses, our model results suggest that diurnally dynamic Fe allocation can improve carbon and Fe use  efficiency, as well as the growth rate of Trichodesmium particularly under low Fe, thereby partly alleviating Fe limitation (Fig. 7). Our model study highlights two main potential mechanisms explaining the benefits of dynamic Fe allocation to N 2 fixation and growth in Trichodesmium.
Note that the Fe in respiratory electron transfer chain was not simulated due to lacking diurnal observational data, although it might occupy a small fraction of intracellular Fe pool (less than 10% of the total intracellular Fe) in Trichodesmium [40]. In fact, respiratory and photosynthetic electron transfer chains share several apparatuses (e.g., Cyt b6f complex and ferredoxin) in Trichodesmium [11], suggesting that Fe used only by respiratory processes could be even less.

Broader context: Is dynamic Fe allocation a common strategy in marine cyanobacterial diazotrophs?
Although the pattern of diurnally dynamic Fe allocation in Trichodesmium is not exactly the same as those in other marine cyanobacterial diazotrophs, several potential benefits of dynamic Fe allocation revealed by our model can be common in certain marine diazotrophs ( Table 3).
The unicellular N 2 -fixing cyanobacterial group A (UCYN-A) symbioses with haptophyte algae and lacks O 2 -envolving photosystem II [9]. Similar to Trichodesmium, UCYN-A symbiosis conducts photosynthesis and N 2 fixation during daytime, while its diurnal transcriptomic patterns of genes relevant to photosystem I and nitrogenase also segregate [41]. High transcription of photosystem I in UCYN-A during the early light period may contribute to reducing intracellular O 2 by cyclic photosynthetic electron transfer; and its decreasing transcription level indicates Fe release from photosystems, which can be a stage preparing for high transcription of nitrogenase and high N 2 fixation rate later [41]. This suggests a possibility that UCYN-A can also dynamically allocate intracellular Fe between photosystem I and nitrogenase to improve the Fe use efficiency, and possibly to also lower respiratory protection as revealed by rare cytochrome c oxidase coxA gene transcription in UCYN-A during the daytime [23,41]. For diazotroph-diatom associations (DDAs) that form heterocyst to protect nitrogenase, there is no obvious dynamic Fe allocation between photosystems and nitrogenase  [42]. This suggests that the dynamic allocation of Fe may not be necessary when photosynthesis and N 2 fixation are physically segregated.
Unlike Trichodesmium and UCYN-A, Crocosphaera (UCYN-B) performs photosynthesis in the day but conducts N 2 fixation at night [41,43,44]. During the light period, nearly no Fe in Crocosphaera is allocated to nitrogenase, while most of its intracellular Fe is in photosystems [37,44]. This strategy helps prevent the inactivation of nitrogenase during the light period when photosynthetic O 2 production is active and intracellular O 2 concentration is high [37].   7. Schematic diagram illustrating diurnally fixed and dynamic Fe allocations under low-Fe and high-Fe conditions. Sizes of bubble charts represents modeled growth rates (GR). Dynamic Fe allocation can reduce the requirement of respiratory protection (RP), calculated as the ratio of daily-integrated carbon consumption rate by respiratory protection to daily-integrated gross carbon fixation rate. It can also improve the iron use efficiency based on the ratio of growth rate to intracellular Fe.

Table 3
The list of dynamic Fe allocation with potential benefits in some marine autotrophic diazotrophs.

Diazotroph
Dynamic Fe allocation Ref. No [42] During the dark period, the majority of intracellular Fe is released from photosystems and used to synthesize nitrogenase [37,44]. Additionally, respiratory protection at low levels is sufficient to protect nitrogenase and support N 2 fixation [35,37]. This dynamic allocation of Fe in the light and dark periods may also partly explain why Crocosphaera is more abundant than Trichodesmium in Fe-limited oceans (e.g., in western North Pacific Subtropical Gyre) [44,45].

Conclusions
In summary, our study provides a deep insight into how Trichodesmium trichomes diurnally allocate their intracellular Fe to promote their growth. In addition to improving the Fe use efficiencies, dynamic Fe allocation is also involved in the intracellular O 2 management, reducing the need for respiratory protection in producing low-O 2 windows for N 2 fixation. Our model framework could be further integrated with the microenvironment of Trichodesmium colonies to explore their in-depth physiological mechanisms regulating N 2 fixation. It can also serve as an eco-physiological module in the biogeochemical model to promote the predictive ability in the context of oligotrophication of surface open ocean, especially under iron limitation.

Author statement
YWL originated concept for the study. YWL and WL designed numerical model. WL coded the initial version of the model and performed numerical modeling. YWL and WL analyzed results and improved the numerical model. WL wrote the first draft of the manuscript; YWL and WL revised the manuscript.

Declaration of Competing Interest
The authors declare that they have no competing interests.