Mechanisms and heterogeneity of in situ mineral processing by the marine nitrogen fixer Trichodesmium revealed by single-colony metaproteomics

Oceanographic context of the sampling location

All Trichodesmium colonies used in this study were collected from the same phytoplankton net which sampled a surface-ocean Southern Caribbean Sea community (Fig. 1a). At the sampling station the phosphate concentration was low (0.13 μM at 100 m) as is typical in an oligotrophic environment, while the surface dissolved iron concentration was relatively high (2.02 nM at 100 m), consistent with coastal or atmospheric inputs being mobilized in this region (Fig. 1a). By far the most abundant Trichodesmium species at this location was an uncharacterized Trichodesmium thiebautii species, as determined by Trichodesmium-specific metagenome-assembled-genome recruiting (see Table S1).

Thirty individual colonies of mixed morphology were separated by hand-picking, immediately examined, and photographed by fluorescent microscopy (385 excitation, >420 nm emission), then frozen individually for particle characterization and metaproteomic analysis (Fig. S1). All colonies used in this study presented as healthy with reddish-orange pigmentation and well-defined shape. When the particles were present they auto-fluoresced in the visual light range, appearing as yellow, red, or blue dots. In general, the particles were concentrated in the center of puff-type colonies, though they were also present in tufts but in smaller numbers. Strikingly, colonies either had many such particles or none at all. Based on prior experimental evidence demonstrating that Trichodesmium colonies can capture mineral particles and access iron from them [14,15,16, 21], we hypothesized that these particles were terrestrially derived minerals (Fig. 1c–h). Therefore, we embarked to understand the morphological heterogeneity by characterizing the particles and the colony’s molecular response to them.

Mineralogical characterization of the colony-associated particles

To find out whether these natural colonies of Trichodesmium had captured iron-rich mineral particles, we performed synchrotron-based micro-X-ray fluorescence (μ-XRF) element mapping of representative colonies with the observed particle associations. Prior evidence of Trichodesmium–particle associations has been based mainly on experimental “feeding” of dust to cultured or captured colonies [15, 16, 20, 22,23,24,25], and it was therefore important to establish these specific Trichodesmium–particle relationships, which developed in nature. We examined one tuft- and two puff-type colonies, all of which had particles associated with them. The element maps were consistent with the hypothesis that there were mineral particles enriched in iron (Fe), copper (Cu), zinc (Zn), titanium/barium (Ti/Ba, which cannot be distinguished by this method), manganese (Mn) and cobalt (Co), though the concentrations approached the limit of detection for the latter two elements (Fig. 2, Figs. S2 and S3). Iron concentrations were particularly high in the particles. Micro-X-ray absorption near-edge structure (μ-XANES) spectra for iron were collected on six particles—three each from the two puffs (Fig. 2 and Fig. S4). The particles contained mineral-bound iron with average oxidation states of 2.6, 2.7, two of oxidation state 2.9, and two of oxidation state 3.0 (Table S2, Fig. S5). While the mineralogy of these particles could not be definitively resolved using μ-XANES, the structure of the absorption edge and post-edge region provided insight into broad mineral groups. Both Fe(III) (oxy/hydro)oxides and mixed-valence iron-bearing minerals consistent with iron silicates were present, suggesting heterogeneous mineral character. While we could not positively identify the silicate mineral phases based on XANES, the spectroscopic similarity of some samples to iron-smectite and the geologic context suggest iron-bearing clays were present (Fig. S5). In this geographic region, iron oxides and clays could be sourced from atmospheric dust deposition, which is common in this region [27, 28] and/or from riverine sources such as the Orinoco and/or Amazon rivers [29, 30].

These colony-associated mineral particles likely serve as a simultaneous source of nutritional (Fe, Ni, Co, Mn) and toxic (Cu) metals to the colonies. The elemental composition of the particles is similar to a recent characterization of Trichodesmium-particle associations in the South Atlantic [30]. Release of metals from the particles likely vary over time, with copper, nickel, zinc, and cobalt continually leaching and iron leaching initially, then re-adsorbing back onto particles unless organic chelates assist in solubilization [31].

Proteome composition is altered by particle presence

To understand the impact of the particles on colony diversity and function, we performed comparative metaproteomic analysis of the individual Trichodesmium colonies and their microbiota. Seven puffs without particles, 14 puffs with particles, and 4 tufts with particles were analyzed by a new single-colony metaproteomic method. This approach allowed for the first time the molecular profiles of heterogeneous Trichodesmium colonies to be examined individually. Compared to bulk population-level metaproteomes from this location, which achieved deeper resolution of low-abundance proteins by integrating biomass from 50 to 100 colonies (4478 proteins identified) [32], proteome coverage for the low-biomass single colonies was lower yet sufficient for characterizing colony function (2078 proteins identified, Fig. S6) [32]. In total, 1591 Trichodesmium and 487 epibiont proteins were identified across the 25 single-colony metaproteomes versus 2944 Trichodesmium and 1534 epibiont proteins across triplicate population-level metaproteomes (Tables S3 and S4). Phylogenetic exclusivity was checked such that peptides used to identify epibiont proteins were not present in the Trichodesmium genome (Table S5 and Fig. S7) [33, 34].

Trichodesmium’s epibiont community plays crucial roles in colony health and physiology, and together the single-colony proteomes demonstrated a diverse and functionally active microbiome associated with the colonies (Fig. 1b and Fig. S8). The proteomic analysis generally reflected the more abundant, “core” members of the epibiont community as was expected given their low-biomass proportion relative to Trichodesmium cells. Many commonly identified epibiont groups were present including Alphaproteobacteria, Microscilla, and non-Trichodesmium cyanobacteria [12, 35, 36]. In general, epibiont abundance was unaffected by particle presence, with one exception: Firmicute proteins were more abundant in tufts and puffs with particles, suggesting enhanced, possibly anaerobic, metabolism. Greater differences were identified between the puff and tuft morphologies, independent of particle presence and consistent with prior characterizations finding that puffs and tufts harbor distinct epibiont communities [12]. Specifically, eukaryotic proteins were more abundant in puffs compared to tufts. These proteins likely represent copepods due to sequence similarity to the model organism Calanus finmarchicus, and this result is consistent with observed associations between copepods and puffs at this location (Fig. S8B). Notably, proteins from the PVC superphylum, particularly an uncharacterized eukaryote pathogen species related to Chlamydia, were also more abundant in puffs. Eukaryotes are often observed in association with Trichodesmium colonies, but are not always identified due to differences in sampling protocols that could wash them away [12], as well as due to biases in analytical methods, for instance in studies with a focus on bacterial 16S or metagenomic analyses. Overall, the differences in the epibiont community were small, suggesting that these do not explain the observed morphological heterogeneity. We therefore turn our attention to describing how the particles impacted the proteome of Trichodesmium specifically.

Mineral presence was associated with significant differences in the Trichodesmium proteome. In total, 131 proteins were differently abundant in puffs with particles versus without particles (p < 0.05, FDR-controlled Welch’s unequal variances t-test). Proteome differences were distributed across a variety of biogeochemically relevant proteins, particularly metalloproteins containing iron, nickel, copper, and zinc. Photosynthesis and carbon fixation proteins including Rubisco (p = 0.04), citrate synthase (p = 0.08), and the accessory pigment allophycocyanin (p = 0.1) were more abundant when minerals were present. In fact, most of the differentially abundant proteins were more abundant in the particle-containing colonies suggesting that the colonies with particles were more metabolically active. Particle presence did not affect nitrogenase abundance, consistent with prior laboratory experiments in which Trichodesmium erythraeum filaments were fed concentrated dust [24]. Because nitrogen fixation is a critical process, Trichodesmium may distribute iron to nitrogenase at a steady rate, while altering the activity of other systems in response to the particles. Alternatively, it is possible that nitrogenase activity was being regulated post-translationally [37] and there were indications of enhanced use and recycling of fixed nitrogen compounds; the nitrogen assimilation proteins glutamine synthetase (p = 0.008), spermidine synthase (p = 0.005), and a urea transporter (p = 0.02) were enriched in colonies with particles (see Fig. S9) [32].

Responses in iron uptake and utilization proteins

The broadest and strongest proteome response occurred in iron-related proteins, consistent with Trichodesmium’s strong dependence on iron availability. Several iron-containing proteins were significantly more abundant when particles were present, suggesting that the particles acted as a micronutrient source. These included an iron-containing peroxidase (p = 0.03), the electron transport protein ferredoxin fdxH (p = 0.0006), and the iron storage/DNA binding protein Dps-ferritin (p = 0.002) (Fig. 3). Together, the coordinated response of each of these proteins suggests that the minerals were a nutritional source of iron. For instance, increased abundance of ferredoxin is consistent with increased iron availability, since the non-iron-containing flavodoxin is substituted for ferredoxin during iron stress [6, 38, 39]. Similarly, increased abundance of Dps-ferritin would serve to buffer and store iron acquired from the concentrated particulate metal source. In addition to binding iron, the Dps-ferritin may also serve to protect DNA from photooxidative damage, and this in addition to the peroxidase signal may indicate increased oxidative stress in the colonies with particles [40]. The clear increase in Dps-ferritin protein abundance in the presence of mineral particles differs from a prior report, which found no change in bacterioferritin transcript abundance when cultured T. erythraeum was fed Saharan desert dust [24]. It is possible that Dps-ferritin and bacterioferritin respond to distinct cellular conditions, for instance Dps-ferritin may have been preferred by these colonies because of its dual function for iron storage and protection against oxidative stress. This difference highlights the potential challenges in comparing laboratory and field studies, as well as the often-suggested importance of post-transcriptional and post-translational controls in Trichodesmium [32, 37, 41]. Given the high iron demand of the nitrogenase metalloenzyme, the ability to store iron from rich but episodically available mineral particles could provide an important ecological niche in oceanic environments where iron can be scarce and its solubility is low.

Multiple uptake mechanisms were involved in obtaining iron from the mineral particles. While iron acquisition in Trichodesmium is not well characterized, at least three systems are known to exist in the genome: the FeoB system (Fe(II)), the IdiA system (Fe(III)), and uptake via Fe-siderophores [9, 11]. The single-colony metaproteomes provided evidence for the deployment of the latter two mechanisms in the natural environment; FeoB is rarely identified in field metaproteomes of diazotrophs, possibly due to its low copy number and high efficiency (Fig. 4) [38, 39, 42]. Despite evidence that the particles provided iron to the colonies, the periplasmic iron transport protein IdiA was more abundant during particle associations (p = 0.16) (Fig. 3d). IdiA is often used as a biomarker of iron stress because it is responsive to/more abundant in low iron conditions [32, 38, 39]. Increased IdiA abundance was confusing because it suggested that the particle-associated colonies were more iron limited, a conclusion that is not supported by the increase in iron storage and utilization proteins described above. One interpretation is that IdiA serves a function in acquiring iron from the mineral particles. Due to its binding preference for Fe(III) [11] this could include uptake from a ligand-bound Fe(III) state that may aid in particle dissolution [11, 43, 44]. In this way, IdiA may have a different functional response to particulate versus dissolved iron. There was also evidence that iron-binding siderophore systems synthesized by the bacterial epibionts were involved: a Firmicute acyl carrier protein putatively involved in siderophore production was enriched in puffs with particles (Fig. 3f, p = 0.04). Trichodesmium does not seem to produce siderophores, but it can acquire siderophore-bound iron produced via mutualistic interactions with epibionts, especially when provided with concentrated dust [9, 22, 23, 43]. Corroborating this, a Trichodesmium TonB-dependent transporter (TBDT) for ferrienterochelin/colicins was identified only in puffs with particles [45].

A crucial role for redox regulation when particles were present

The enzyme nickel superoxide dismutase (Ni SOD) was significantly more abundant in the mineral-associated colonies (Fig. 3g, p = 0.004), reflecting the need to regulate the reactive oxygen species (ROS) superoxide in the presence of particles. There are multiple reasons for elevated superoxide production in particle-associated colonies. First, metabolic electron transport is associated with incidental superoxide production, and electron transport seems to have been enhanced in the presence of particles as indicated by the enrichment of photosynthesis proteins. Furthermore, Trichodesmium has among the highest Mehler reaction activity of any photosynthetic organism, and this may contribute to internal superoxide production [46]. Indeed, this evidence of an abundant ROS detoxification enzyme is consistent with intracellular superoxide production and indicates that the traditional Mehler reaction is important in field populations. Additional redox-regulating proteins including peroxiredoxin, thioredoxin, and Dps-ferritin were more abundant when particles were present (see Fig. 3e, p = 0.03 and 0.04, respectively, for puffs with and without particles), consistent with elevated ROS in particle-containing colonies. Increased superoxide dismutase protein may also reflect enhanced nickel availability when particles were present. Nickel is an essential nutrient for Trichodesmium, and can limit nitrogen fixation due to Trichodesmium’s significant need for superoxide dismutase to protect the nitrogenase enzyme [47].

An additional explanation for increased superoxide dismutase protein abundance is that it indirectly reflected iron uptake processes and/or extracellular iron reduction. Reduced iron transported into the cell could react with intracellular oxygen, leading to increased superoxide formation; [48] this could trigger superoxide dismutase production to maintain healthy intracellular superoxide levels [49]. Puffs in particular are known to closely regulate superoxide levels with a possible link to cell signaling and growth [50]. Extracellular superoxide production may also play a role in iron uptake, as has been suggested previously [9]. This extracellular superoxide production may indirectly contribute to intracellular superoxide dismutase signals via Fe(II) uptake as described above, but because superoxide does not cross cell membranes, this does not directly explain the observed signal in intracellular superoxide dismutase [51]. Besides extracellular superoxide, other mechanisms of reductive iron uptake have been suggested including reactive metabolites and extracellular proteins on the cell surface [9]. Recent observations have tied hydrogen physiology to mineral iron uptake, however, the underlying biochemical system remains unknown [21]. Systems such as the alternative respiratory terminal oxidase (ARTO) have been invoked [52], but this protein was not detected, implying it was  not a major constituent of the proteome, though as in the above case of FeoB it is possible that ARTO was present in low concentrations and  was highly active. The putative role of ARTO is furthermore unclear because a recent study observed that ARTO transcript abundance was lower in the presence of dust, and not enhanced as might be expected if ARTO is involved in iron uptake [24].

Involvement of chemotaxis machinery in particle entrainment

A prior study captured impressive visual observations of Trichodesmium responses to dust, where experimentally provided particles were captured and transported to the colony center [15]. This behavior was seen primarily in puff-type colonies isolated from the field, indicating that these natural populations may react differently to dust than laboratory Trichodesmium strains. Consistent with prior studies [15, 20], the mineral particles were concentrated in the center of puff-type colonies and this may enhance the efficiency of extracellular iron reduction pathways by reducing diffusive loss [53]. Moreover, the proteomic characterization from this study provides molecular insight into this behavior. Here we observed that movement proteins including a SwmA-like RTX protein (p = 0.1) [54] and chemotaxis regulator CheY (p = 0.3) were enriched when particles were present (Fig. S9E–H), indicating that they were involved in the particle capturing behavior observed by Rubin et al. [15]. In previously published bulk metaproteomes of field Trichodesmium populations, there was an observed relationship between the CheY protein and the iron transport protein IdiA, suggesting a link between chemotaxis and particulate iron acquisition [32]. While there is more to be learned about how this system operates, it appears that the chemotaxis and iron regulatory systems are connected and responding directly to particulate iron.

Evidence for a specific regulatory response to particulate iron in Trichodesmium

Taken together, the in situ biochemical observation points to a model in which Trichodesmium colonies differentiate between dissolved versus particulate iron. First, consistent with the mineralogical profile of the particles, the colonies engaged multiple iron uptake mechanisms and also prioritized balancing cellular redox status during enhanced productivity and iron uptake (Fig. 4). Once inside the cell, mineral-derived iron was preferentially stored via Dps-ferritin. This finding adds complexity to the canonical regulatory model that IdiA and ferritin exist on a continuum with IdiA abundant during iron limitation and ferritin present only when iron is replete. It suggests that Trichodesmium alters uptake and utilization mechanisms in response to the iron’s coordination environment, specifically to the particulate vs. dissolved phase. Whether Trichodesmium makes the distinction directly via a specific mineral sensing mechanism, or indirectly due to increased intracellular metal concentration, is not yet known. Trichodesmium has an unusually large number of uncharacterized two-component sensory systems relative to other phytoplankton, and it is possible that one or more of these systems is involved, which would further imply the involvement of metal-binding ligands [55]. Either way, it is clear that multiple metabolic and homeostasis systems were involved in this coordinated response to the presence of particulate iron, and it does seem likely that iron-specific regulatory processes are involved since in laboratory studies Trichodesmium is able to distinguish between iron-containing and non-iron-containing particulate matter [20]. A final intriguing aspect of particle-associated Trichodesmium colonies was the deployment of Cu-related proteins. Trichodesmium is known to be extraordinarily sensitive to Cu toxicity [56], and close proximity to mineral particles was associated with enrichment of the copper chaperone/homeostasis protein CopZ (p = 0.06 and Fig. S9D) [57], suggesting the need for metal detoxification when minerals were present.

Conclusion

This study provides direct evidence that Trichodesmium colonies capture and process oxide and silicate minerals in situ, and that colonies alter key aspects of their biochemistry in response. These findings highlight heterogeneity in the morphological and molecular profiles of Trichodesmium colonies in nature, revealed by a new approach in which colony proteomes are analyzed individually. The underlying mechanisms for this heterogeneity are multiple and likely intersect. One important aspect is stochastics of colony-particle encounters due to patchy dust availability. In addition, the sampled population might represent a mixture of Trichodesmium colonies advected from geographical locations or water column environments/depths with different particle loadings. Other explanations could include sub-species differences among the colonies, unresolved changes in the epibiont community, and past physiological history of the colony such as age or past iron status. Clearly much is still to be discovered about the reasons for and implications of physiological heterogeneity in natural marine microbial populations.

Trichodesmium thrives in high-dust environments and this study provides clear evidence for the biochemical basis behind this specialized niche. The results have ecological and geochemical implications beyond Trichodesmium biology. Specifically, active capture and degradation of mineral particles may increase iron availability in the oligotrophic surface ocean. In this way, abundant Trichodesmium colonies may have an important role in the leaching of particulate trace metals and the supply of bioavailable iron to the euphotic zone, not least because Trichodesmium colonies tend to be neutrally or positively buoyant [58, 59] and can therefore retain mineral matter at the surface. Mineral capture and degradation may therefore have significant implications for global iron and carbon cycling [60]. Furthermore, the coupled mineralogical and biochemical characterization of single colonies present evidence for specific responses to mineral particles that could be leveraged to improve future biogeochemical and marine ecosystem models. Particulate iron utilization by Trichodesmium appears to be a critical niche, and is likely a significant factor determining this organism’s ecological success and fixed nitrogen contributions to the global ocean.

Source: Ecology - nature.com