Sampling for this study was performed at 14:00 on May 17 2023 in the northeastern East China Sea at station E48 (33.0°N, 126.8°E). The survey was conducted aboard the research vessel (R/V Saedongbaek) operated by Chonnam National University.

At the sampling station, a conductivity–temperature–depth (CTD) profile was collected using a CTD system (SBE 911plus, Sea-Bird Electronics, Bellevue, WA, USA), while chlorophyll fluorescence was measured with a CTD equipped with a fluorescence sensor (ECO-AFL/FL, WET Labs, Philomath, OR, USA). The depth of the DCM layer was identified from the in situ CTD profiles. For biological and chemical analyses, seawater samples were collected using Niskin bottles (General Oceanics, Miami, FL, USA) mounted on the CTD rosette.

For nutrient analysis, 15 mL of seawater per sample was filtered through 0.7 µm GF/F filters (25 mm; Whatman, Florham Park, NJ, USA) using acid-washed syringes into polypropylene Falcon™ tubes. Concentrations of dissolved inorganic nitrogen (DIN; NO₃ + NO₂), dissolved inorganic phosphate (DIP; PO₄), and dissolved silicic acid (DSi; Si(OH₄)) were measured using a nutrient auto-analyzer (New QuAAtro39, SEAL Analytical, Orpington, UK). Quality control was maintained using certified reference standards for each nutrient.

For microscopic identification and enumeration of cells, seawater samples were preserved with Lugol's iodine solution to a final concentration of 2%. Subsequently, protist cells in the preserved samples were identified to the species level and counted. For enumeration, 500 mL subsamples preserved in Lugol’s solution were concentrated five- to tenfold using a 2-day settling method. After thorough mixing, either all or a minimum of 300 cells from three to five 1 mL Sedgwick–Rafter counting chambers were counted at 100× or 200× magnification using a light microscope (Nikon, Tokyo, Japan).

For pigment analysis, 2 L of seawater was collected from the surface, DCM, and bottom layers, and filtered through pre-combusted GF/F filters (25 mm; Whatman Inc., Florham Park, NJ, USA). Filters were sonicated for 5 min in 95% acetone, with extraction continuing for 24 h at 4 °C in darkness. After extraction, samples were centrifuged and passed through 0.2 μm PTFE syringe filters (Merck Millipore, USA) to remove particulates. High-performance liquid chromatography (HPLC) analysis followed the protocol described by Zapata et al. (2000), using an Agilent 1200 HPLC system (Agilent Technologies, USA) equipped with a Waters C8 column (150 mm × 4.6 mm, 3.5 µm particle size, 0.01 Å pore size).

Phytoplankton community structure and relative algal biomass were estimated from pigment composition determined by HPLC using CHEMTAX-based calculations (Mackey et al., 1996). CHEMTAX is a chemotaxonomic matrix factorization approach that partitions pigment concentrations into taxonomically distinct groups based on characteristic pigment ratios, providing estimates of the relative biomass of major phytoplankton classes. In this study, initial pigment-to-chlorophyll-a ratios were assigned based on published values for representative phytoplankton groups, including dinoflagellates, diatoms, pelagophytes, haptophytes, chlorophytes, cryptophytes, and cyanobacteria. Optimization was performed independently for each sample type (Surface, DCM, and Bottom) using a constrained nonlinear optimization algorithm (sequential least squares quadratic programming, [SLSQP]) implemented in Python (Python Software Foundation, Wilmington, DE, USA) with the SciPy library (SciPy Community, Austin, TX, USA). The resulting output provides the relative proportions of phytoplankton classes, enabling comparative analysis of community composition across water column depths.

For metabarcoding analysis (18S rRNA gene amplicon sequencing) of protistan communities, 1 L of seawater was collected from each depth and filtered through 0.45 µm membrane filters (47 mm; Whatman, Florham Park, NJ, USA) under gentle vacuum pressure (<100 mmHg). Filters were immediately stored at -80 °C until DNA extraction. Genomic DNA was extracted using the DNeasy PowerSoil Kit (Qiagen, Hilden, Germany) following the instructions of the manufacturer and quantified with a Qubit fluorometer using the Quant-iT PicoGreen dsDNA assay kit (Invitrogen, Carlsbad, CA, USA).

The V4 region of the 18S rRNA gene was amplified using the universal eukaryotic primers TAReuk454FWD1 (5′-CCAGCASCYGCGGTAATTCC-3′) and V4 18S Next.Rev (5′-ACTTTCGTTCTTGATYRATGA-3′) (Stoeck et al., 2010; Piredda et al., 2017). Sequencing libraries were prepared following the Illumina metagenomic library preparation protocol, while paired-end sequencing (2 × 250 bp) was conducted on an Illumina MiSeq platform (San Diego, CA, USA) at CJ BioScience (Seoul, South Korea).

Raw sequencing reads were demultiplexed and processed using QIIME 1.9.1 and Cutadapt for primer removal and quality trimming. Subsequent analyses were conducted with the QIIME 1.9 pipeline (Martin, 2011; Bolyen et al., 2019). Amplicon sequence variants (ASVs) were inferred using the DADA2 pipeline, which performs denoising, read merging, and chimera removal (Callahan et al., 2016). To account for variation in sequencing depth, rarefaction was performed ten times, and the resulting ASV tables were merged. Taxonomic classification was conducted using the PR2 4.14.0 database (Guillou et al., 2012), and ASVs assigned to metazoans were excluded from downstream analyses. Seo et al. (in preparation) provided further details regarding PCR conditions, library preparation, and quality filtering procedures for the same raw sequence dataset.

For metatranscriptome analysis, 2 L of seawater was collected per sample. Total RNA was extracted from samples filtered onto 0.45 µm membrane filters (47 mm; Whatman, Florham Park, NJ, USA) using the RNAqueous-4PCR Kit (Thermo Fisher Scientific, Waltham, MA, USA) following the instructions of the manufacturer. Cell disruption was facilitated with glass beads to lyse organisms with robust cell walls (e.g., diatoms). RNA concentration and purity were assessed using a NanoDrop 2000 spectrophotometer (Thermo Fisher Scientific™, Waltham, MA, USA).

Ribosomal RNA was removed using the Illumina Ribo-Zero Plus rRNA Depletion Kit, and the remaining total RNA was processed with the TruSeq Stranded Total RNA Library Prep Gold Kit, following the instructions of the manufacturer. All biological replicates were maintained separately throughout the workflow. Sequencing was conducted on an Illumina NovaSeq 6000 platform in a 2 × 151 bp paired-end configuration to generate high-throughput data suitable for downstream metatranscriptomic analyses.

Raw sequencing reads were quality-trimmed and adapter sequences removed with Trimmomatic v0.39 (Bolger et al., 2014) using the following parameters: ILLUMINACLIP : TruSeq3-PE.fa:2:30:10:2:True, LEADING:3, TRAILING:3, SLIDINGWINDOW:4:20, and MINLEN:36. Subsequently, read quality was assessed with FastQC v0.12.1 (Andrews, 2010). De novo transcriptome assembly was performed using rnaSPAdes v4.0.0 (Bushmanova et al., 2019), while contigs were clustered at 99% similarity using CD-HIT-EST v4.8.1 (Li and Godzik, 2006) to eliminate redundancy. Assembled contigs were merged into a combined reference assembly. Read counts and transcript abundance (measured as transcripts per million [TPM]) were quantified by mapping trimmed reads to the assembled contigs using Salmon v1.10.0 in quasi-mapping mode (Patro et al., 2017) to generate gene-level expression estimates.

Taxonomic and functional annotations were performed with the DIAMOND aligner v2.1.11.165 (Buchfink et al., 2021) in BLASTX mode using an E-value threshold of <10^-5. Taxonomic classification was based on phyloDB v1.076 (https://github.com/allenlab/PhyloDB), a custom reference database containing sequences from the Marine Microbial Eukaryote Transcriptome Sequencing Project (MMETSP) (Keeling et al., 2014). Functional annotation of contigs was conducted using the UniProtKB/TrEMBL database (release 2024-06; accessed February 18, 2025; UniProt Consortium, 2025). Each contig was annotated based on its top-scoring BLASTX hit against the reference database.

Prior to differential expression analysis, read counts were imported and aggregated into gene-level estimates for each taxonomic group (diatoms, dinoflagellates, chlorophytes, ciliates, cryptophytes, haptophytes, Stramenopiles, Syndiniales, and other eukaryotes) using Tximport v3.21 (Soneson et al., 2015). Differential expression analysis was conducted using DESeq2 v1.49.3 (Love et al., 2014) by aggregating contig counts within each taxonomic group. To minimize noise from low-abundance genes, a pre-filtering step was applied before differential expression analysis. Genes with <10 read counts in at least three samples were excluded from downstream analyses. Raw read counts were normalized using size factors estimated by DESeq2, accounting for sequencing depth and RNA composition bias across samples. Differential expression between two groups was assessed using the Wald test, while genes with an adjusted p-value < 0.05 were considered significantly differentially expressed.

Log₂(fold change) and log₂(average normalized count) values were calculated for genes associated with Photosystem I and II subunits, light-harvesting complexes (LHCs), the ATP synthase complex, and nitrogen transport and metabolism. For UniProt entries annotated with the same protein function (Supplementary Table S1), values were aggregated into a single representative gene. The representative log₂(fold change) was calculated as a weighted average: ∑(log₂(fold change) × baseMean) / ∑(baseMean), and the corresponding log₂(average normalized count) was calculated as log₂(∑baseMean), where ‘baseMean’ represents the mean size-factor-normalized count of DESeq2 across all samples. Because baseMean is averaged across conditions, this summary represents an overall expression proxy rather than condition-specific abundance.

To visualize gene expression patterns, a bubble heatmap was generated using the “bubbleHeatmap” R package (Boxall et al., 2023) in R version 4.5.0. The analysis included only genes expressed in at least two species within each of the seven major eukaryotic lineages—diatoms, dinoflagellates, chlorophytes, ciliates, cryptophytes, haptophytes, and Stramenopiles. Furthermore, only genes with expression values available in at least three of these lineages were retained for the final visualization.

The water column at the surveyed station exhibited marked physical and biogeochemical contrasts between the surface and DCM layers. Chl-a fluorescence peaked at approximately 45 m, reaching over ten times the surface intensity before rapidly declining with increasing depth (Figure 1A). Temperature sharply decreased, while salinity rose slightly, reflecting weak stratification between the surface and DCM layers. Beyond 60 m, these parameters stabilized at approximately 15.5 °C and 34.5, respectively (Figure 1B). Dissolved inorganic nutrient concentrations varied markedly between the surface and DCM layers (Figure 1C). Nutrient levels were relatively low at the surface but increased at the DCM, where DIN, DIP, and DSi reached 3.21, 0.24, and 11.52 µM, respectively. Below the DCM, concentrations continued rising, indicating progressive nutrient accumulation in deeper waters.

Microscopic observations revealed clear vertical variation in protist community structures across depths (Table 1). Total protist abundance increased with depth, peaking at the DCM with 65.0 cells mL^-1, over three times the surface value of 19.8 cells mL^-1. This increase was largely due to picoeukaryote proliferation (< 2 µm), which dominated total cells at all depths and peaked at the DCM.

Among identifiable taxa, cryptophytes were consistently observed throughout the water column, peaking between 20–30 m (Table 1). Diatoms were predominantly confined to subsurface layers below 50 m, including Nitzschia, Synedra, and Thalassiosira, with maximum abundance at 75 m (5.6 cells mL^-1). Dinoflagellates such as Gymnodinium and Gyrodinium appeared sporadically, with Gymnodinium spp. peaking at 30 m (8.5 cells mL^-1). Ciliates, primarily oligotrichs, were present across depths but in low densities (< 1.1 cells mL^-1). These depth profiles showed a layered community structure shaped by vertical gradients.

Pigment concentrations from surface, DCM, and bottom samples revealed distinct vertical variation in phytoplankton community structure (Figure 2A). Chl-a, a proxy for total phytoplankton biomass, peaked at the DCM (~1.9 µg L^-¹), was substantially lower at the surface (0.2 µg L^-¹), and minimal at the bottom (0.09 µg L^-¹). Chlorophyllide-a (Chlide-a), a degradation product linked to cellular breakdown or stress, occurred in low concentrations only at the DCM and bottom. Accessory pigments, indicative of specific phytoplankton taxa, exhibited distinct vertical distribution patterns. For instance, 19′-hexanoyloxy-fucoxanthin (19′-Hex-fuco), typically marking haptophytes but also present in certain dinoflagellates (e.g., Kareniaceae due to tertiary endosymbiosis), was strongly enriched at the DCM. Similarly, fucoxanthin (Fuco), mainly associated with diatoms, and 19′-butanoyloxy-fucoxanthin (19′-But-fuco), linked to pelagophytes, were significantly enriched at the DCM.

CHEMTAX-based estimates highlighted vertical differentiation in protist assemblages (Figure 2B). At the surface, haptophytes (38.4%) and cyanobacteria (19.7%) dominated, with chlorophytes (14%), diatoms (13%), and cryptophytes (7%) present in smaller proportions. At the DCM, haptophytes remained dominant, rising to 51.7%, alongside notable proportions of chlorophytes (22.1%) and pelagophytes (10.9%). In contrast, at the bottom, the community shifted to pelagophytes (42.2%) and diatoms (23.9%), whereas haptophytes declined to 18.2%. Dinoflagellates and cyanobacteria notably declined consistently with depth.

To examine protist community composition and transcriptional activity across depths, we conducted DNA metabarcoding and metatranscriptomic analyses (Figure 3). Metabarcoding revealed that surface waters were dominated by dinoflagellates with Stramenopiles and chlorophytes also abundant, whereas the DCM layer was enriched in chlorophytes and Syndiniales (Figure 3A). Overall, dinoflagellates and Stramenopiles were more abundant in the upper layers (0–30 m), while Syndiniales predominated below the DCM. Species richness, inferred from the number of ASVs, was consistently higher at the DCM and deeper layers than at the surface (Supplementary Table S2). Dinoflagellates, diatoms, ciliates, and Syndiniales displayed marked increases in richness at the DCM.

Metatranscriptomic analyses revealed that dinoflagellates contributed the largest share of total transcript abundance in surface and DCM communities, followed by diatoms, Stramenopiles, haptophytes, and chlorophytes (Figure 3B). Diatoms and haptophytes, despite their relatively low abundance in the 18S rRNA amplicon data, exhibited disproportionately high transcript levels, indicating substantial metabolic activity. In contrast, Syndiniales, although abundant in the DNA-based profiles at the DCM, displayed negligible transcript expression. The magnitude of depth-related transcriptional shifts, measured based on the proportion of significantly differentially expressed genes (DEGs), varied markedly among taxa. Diatoms showed the highest proportion, with 47.9% of annotated transcripts significantly upregulated or downregulated at the DCM compared to those at the surface (Table 2). Conversely, dinoflagellates, despite their dominance in transcript abundance, exhibited the lowest proportion of DEGs (9.0%), followed by Syndiniales (12.4%). Among the DEGs, diatoms, chlorophytes, and ciliates were distinguished according to disproportionately high proportions of upregulated genes at the DCM.

At the genus level, multi-omics analyses revealed pronounced depth-dependent restructuring of protist communities (Figure 4). Metabarcoding showed that surface waters were enriched in dinoflagellates, particularly Gyrodinium, Prorocentrum, and Lepidodinium, whereas the DCM was dominated by chlorophytes (Ostreococcus, Bathycoccus, Micromonas) along with Syndiniales groups I and II. Diatoms (Chaetoceros, Pseudo-nitzschia) and Stramenopiles (Pelagomonas, MAST lineages) contributed at moderate levels. In contrast, metatranscriptomic profiles revealed disproportionately high transcriptional activity in diatoms (Chaetoceros, Pseudo-nitzschia), haptophytes (Chrysochromulina, Phaeocystis), chlorophytes (Micromonas), and specific dinoflagellates (Karlodinium, Alexandrium, Karenia), whereas Syndiniales (Amoebophrya) exhibited minimal transcript expression despite their substantial DNA-based abundance.

Gene expression profiles across all protist taxa revealed distinct, taxon-specific transcriptional shifts between the surface and DCM layers (Figure 5). In diatoms, chlorophytes, and dinoflagellates, genes involved in photosynthetic electron transport and ATP synthesis—particularly those encoding Photosystem I components (PsaA, PsaB, PsaF) and chloroplast ATP synthase subunits (AtpB, AtpD, AtpH, AtpI)—were strongly upregulated at the DCM. Conversely, mitochondrial ATP synthase genes (e.g., ATP5PB and ATPGMC1) were downregulated or largely unchanged, indicating a metabolic shift toward chloroplast-dominant energy production under subsurface conditions.

Photosystem II genes (PsbC, PsbV) exhibited group-specific responses: diatoms and dinoflagellates were clearly upregulated at the DCM, whereas chlorophytes displayed minimal change. In contrast, cryptophytes uniquely upregulated Photosystem II genes without corresponding increases in Photosystem I, while Stramenopiles generally downregulated genes from both photosystems and ATP-related pathways.

LHC genes were broadly upregulated at the DCM across all groups except ciliates, indicating a shared transcriptional response to low light. Nitrogen acquisition and assimilation genes—including nitrate/nitrite transporters and reductases—were markedly upregulated in diatoms, chlorophytes, and dinoflagellates, but downregulated or largely unchanged in haptophytes, Stramenopiles, and cryptophytes. Urea transporter (UT) genes increased specifically in Stramenopiles, while UT and silicate transporter (SIT) genes were upregulated in diatoms at the DCM.

At the study site, the DCM occurred at 45–50 m depth, where the differences in temperature (2 °C) and salinity (0.2) between surface and DCM waters indicate weak vertical stratification. Consequently, protist communities are unlikely to be sharply divided by physical barriers, allowing diel vertical migration and resource exchange across layers. This interpretation aligns with water-mass classification from the same survey Seo et al. (in preparation), which identified surface and DCM waters as Kuroshio Surface Water, underlain by denser Kuroshio SubSurface Water. The proximity of this denser layer may further promote biomass accumulation at the DCM through internal wave dynamics or passive retention at the density interface (Muacho et al., 2013).

In addition to these physical features, the sharp increase in nutrient concentrations at the DCM—particularly DIN (3.21 µM), phosphate (0.24 µM), and silicate (11.52 µM)—provides a nutrient-rich environment that supports high primary productivity. These values exceed the half-saturation constants (Ks) for nutrient uptake in several protists, suggesting that cells at this depth can sustain higher photosynthetic efficiency and growth (Paasche, 1973; Jeong et al., 2015). Thus, while light at the DCM is lower than that at the surface, it remains sufficient to support photosynthesis when coupled with nutrient enrichment, producing a trade-off that promotes niche specialization.

Collectively, these vertical oceanographic features establish the DCM as an optimal ecological niche for protists: suboptimal but sufficient light for basic photosynthesis, nutrient levels above physiological requirements, and hydrodynamic conditions that enhance biomass retention. This combination of drivers likely shapes the distinct community composition and functional gene expression at the DCM.

Microscopic observations highlight the DCM as a key ecological niche characterized by peak protist biomass, primarily driven based on pico-sized phototrophs. While these cells could not be taxonomically resolved via light microscopy, their vertical abundance is consistent with molecular evidence for picochlorophyte dominance—particularly Ostreococcus and Micromonas—at this depth. Their prevalence at the DCM highlights the competitive advantage of small phototrophs in exploiting subsurface resource gradients.

Diatoms were mostly confined to depths < 50 m, likely owing to their higher sinking rates associated with silicified frustules and the capacity of some taxa to persist as resting stages (Agustí et al., 2015; Petrucciani et al., 2023). Conversely, the mid-depth peak of Gymnodinium spp. probably represents a morphologically indistinguishable but taxonomically diverse Gymnodiniales assemblage (Hansen et al., 2000; Thessen et al., 2012; Kang et al., 2014). Many members of this order, including Gymnodinium, Karenia, and Karlodinium, are mixotrophs capable of photosynthesis and phagotrophy. This dual strategy confers ecological flexibility and probably underlies their patchy vertical distribution across variable light and nutrient environments (Berge et al., 2008; Lee et al., 2014; Ok et al., 2023).

Ciliates and cryptophytes occurred at low but consistent densities across depths, suggesting generalist strategies and broad ecological tolerance to diverse light and prey conditions. Overall, the scarcity of large protists in surface waters, combined with the predominance of small cells at depth, indicates a size-structured vertical segregation potentially modulated via grazing pressure, nutrient dynamics, and light limitation.

Pigment analysis coupled with CHEMTAX estimations offers key insights into the vertical organization and functional specialization of marine protist communities. The pronounced Chl-a peak at the DCM indicates optimal growth conditions, sustained based on enhanced nutrient availability and sufficient irradiance compared to the nutrient-depleted surface. This increase cannot be attributed solely to higher cell abundance but also indicates photoadaptive responses, whereby phytoplankton enhance chlorophyll content and adjust pigment composition to optimize light harvesting under low-light conditions (Cullen, 1982; Falkowski and Raven, 2013; Cullen, 2015; Latasa et al., 2017).

The presence of Chlide-a—a degradation pigment—at the DCM and bottom layers suggests stress-associated processes, potentially associated with reduced irradiance, grazing pressure, or residual metabolic activity, including mixotrophy, in organisms settling into deeper waters (Llewellyn et al., 2007; Pillai et al., 2018). Pronounced increases in accessory pigments highlight depth-specific community adaptations. Elevated concentrations of 19′-Hex-fuco underscore the dominance of haptophytes while also suggesting contributions from dinoflagellates such as the Kareniaceae, reflecting notable metabolic plasticity (Penot et al., 2022; Ok and Jeong, 2025). Fucoxanthin and 19′-But-fuco, diagnostic of diatoms and pelagophytes, respectively, further illustrate niche partitioning at the DCM: diatoms capitalize on elevated nutrient availability (Domingues et al., 2023), whereas pelagophytes sustain productivity under low-light, nutrient-rich conditions and may contribute to vertical carbon flux (Cabello et al., 2016; Guérin et al., 2022). Multi-omics analyses corroborate these pigment-based patterns, revealing metabolically active mixotrophic dinoflagellates (Karlodinium, Karenia) and highlighting their ecological advantage in subsurface environments (Berge et al., 2008; Ok et al., 2023).

Our study showed pronounced mismatches between protist community composition and transcriptional activity across depths. While groups such as dinoflagellates, diatoms, chlorophytes, and Syndiniales were taxonomically present across depths, significant discrepancies were observed when comparing taxonomic abundances (18S rRNA) to transcript abundances (mRNA). These disparities demonstrate that taxonomic abundance does not necessarily indicate ecological or metabolic dominance, highlighting the importance of transcription-based approaches for understanding microbial activity within ecological contexts (Marchetti et al., 2012; Hu et al., 2016; Cohen et al., 2021).

Species richness peaked at the DCM, especially among dinoflagellates, diatoms, and Syndiniales, reflecting ecological niche differentiation driven by gradients in light availability and nutrient concentrations. Dinoflagellates and diatoms—recognized for their metabolic versatility—exhibited elevated transcriptional activity relative to their DNA abundances at the DCM, suggesting active exploitation of nutrient-rich subsurface conditions. Robust transcriptional activity from mixotrophic dinoflagellates, including Karlodinium, Alexandrium, and Karenia, indicates that mixotrophy confers a metabolic advantage under DCM conditions (Lim et al., 2019; Ok et al., 2023; Jang et al., 2025). Diatoms also exhibited disproportionately high transcript abundances, consistent with their established capacity for rapid transcriptional responses to nutrient enrichment (Alexander et al., 2015). Conversely, Syndiniales were taxonomically abundant but contributed minimally to transcript pools, consistent with previous observations of their parasitic or dormant states (Chambouvet et al., 2011; Decelle et al., 2022). This discrepancy highlights that numerical abundance can overestimate their ecological role, as transcriptional activity likely peaks only during host infection events, such as dinoflagellate blooms (Chambouvet et al., 2011; Farhat et al., 2018). Moreover, the limited representation of lineage-specific genes in current reference databases may contribute to underestimating their transcriptional activity. Although Syndiniales exhibited low transcriptional activity, their parasitic interactions with dinoflagellates may exert episodic top-down control during host blooms, contributing to population turnover and altering carbon flow within DCM communities.

Our metatranscriptomic analysis reveals that protist taxa exhibit taxon-specific photo-physiological and metabolic responses to the contrasting conditions of surface and DCM layers. At the DCM, the coordinated upregulation of Photosystem I (PSI)-associated genes and chloroplast-encoded ATP synthase subunits in diatoms, chlorophytes, and dinoflagellates suggests a shared strategy for optimizing energy production under suboptimal light conditions. PSI efficiently captures long-wavelength light and sustains cyclic electron flow, enhancing ATP synthesis without concomitant NADPH production, thereby preventing over-reduction of the electron transport chain (Falkowski and Raven, 2013). Such PSI-biased adjustments have been observed in eukaryotic phytoplankton acclimating to low-light conditions (Ishii et al., 2023) and likely confer energetic advantages for maintaining biomass at depth. Consistently, the simultaneous downregulation of mitochondrial ATP synthase genes indicates a strategic shift toward chloroplast-based ATP production (Cardol et al., 2009), reducing reliance on respiration and potentially limiting ROS generation under sub-saturated light.

Taxon-specific differences in photosystem regulation highlight the ecological plasticity of distinct lineages. In chlorophytes, PSI upregulation was accompanied by a pronounced decrease in PSII expression, indicating a functional decoupling between the two photosystems. This pattern mirrors green algae in low-light or DCM environments, where PSII expression and activity are downregulated to minimize light limitation stress and reduce unnecessary energy expenditure, while maintaining PSI function to support baseline photochemistry under sub-saturating irradiance (Bonente et al., 2012). Conversely, cryptophytes exhibited an inverse strategy, upregulating PSII without a corresponding increase in PSI. Combined with their distinctive phycobiliprotein-based light-harvesting pigments (Spangler et al., 2022), this PSII-dominant response suggests cryptophytes are specialized for capturing blue–green wavelengths prevalent in subsurface environments, emphasizing niche differentiation among phototrophs within the DCM.

Beyond photosystems, the widespread upregulation of LHC genes across most taxa (except ciliates) reflects a convergent strategy to maximize photon capture under suboptimal irradiance. Such LHC expansion is a well-established hallmark of photoacclimation in low-light marine phytoplankton (Moejes et al., 2017). Collectively, these transcriptional adjustments demonstrate how distinct taxa fine-tune their photophysiology to optimize energy acquisition in the DCM.

Nutrient-related gene expression highlights depth-specific nitrogen acquisition strategies at the DCM. Dinoflagellates and chlorophytes upregulated nitrate and nitrite transporters as well as associated reductases, consistent with active utilization of regenerated inorganic nitrogen pools (Dagenais-Bellefeuille and Morse, 2013; Priyanka et al., 2024). Conversely, diatoms exhibited minimal changes in nitrate and nitrite uptake but strongly upregulated urea transporters, indicating reliance on organic nitrogen. Concurrently, silicate transporter expression increased, supporting frustule formation under nutrient-replete but light-limited conditions (Thamatrakoln and Hildebrand, 2007; Allen et al., 2011; Olesen et al., 2020). Despite divergent strategies at the uptake stage, all three groups converged in upregulating downstream assimilation genes, highlighting a shared requirement for amino acid and protein synthesis. In contrast, haptophytes, cryptophytes, and other Stramenopiles exhibited reduced or stable expression of these pathways, suggesting reliance on alternative nitrogen sources, dormancy, or phagotrophic nutrition (Terrado et al., 2015; Wördenweber et al., 2018; Wilken et al., 2020).

Collectively, these findings demonstrate that protist assemblages at the DCM exhibit taxon-specific metabolic strategies, partitioning niches through distinct photosystem regulation and nutrient utilization. This metabolic complementarity likely minimizes interspecific competition, enhances resource partitioning, and supports the functional diversity that underlies the high productivity and ecological resilience of DCM communities (Finkel et al., 2010; Marañón, 2015).

Integrating microscopic cell counts, pigment-based chemotaxonomy, and multi-omics offered a comprehensive view of protist community structure across surface and DCM layers. Each method captured distinct aspects of community organization—morphological (microscopy), biochemical (pigment), taxonomic (18S rRNA), and functional (mRNA)—and their integration provided unprecedented resolution of vertical niche partitioning in microbial eukaryotes. Pigment-based CHEMTAX analysis confirmed the dominance of haptophytes and cyanobacteria under high-light, oligotrophic surface conditions, whereas metatranscriptomic data revealed that surface communities, despite lower taxonomic diversity, exhibited strong functional activity—particularly among dinoflagellates—highlighting the role of metabolic plasticity and rapid photoadaptive responses in the upper euphotic zone.

In contrast, the DCM contained markedly higher cell abundance, species richness, and functional heterogeneity. Microscopy revealed abundant small-sized picoeukaryotes and diatoms, while metabarcoding revealed increased hidden diversity, including Syndiniales and rare taxa at greater depths. Metatranscriptomic analysis revealed strong taxon-specific differential expression of genes involved in photosynthesis, ATP synthesis, and nutrient metabolism, indicating the DCM as a biogeochemical hotspot with intense primary production and high ecological versatility. Pigment and multi-omics data indicated persistent pelagophytes, such as Pelagomonas, across depths, suggesting adaptations that sustain productivity in low-light, nutrient-rich conditions and support vertical nutrient cycling and carbon export below the euphotic zone (Cabello et al., 2016; Guérin et al., 2022). Likewise, pico-chlorophytes, consistently dominant at the DCM, appear to optimize their niche by using small cell size to enhance nutrient uptake in a zone of low light but high nutrient availability.

A key insight from this integrative framework is that the presence of a taxon does not necessarily indicate its functional activity. Community dominance in terms of cell counts or DNA abundance does not always indicate metabolic dominance, as evidenced by the divergence between abundant but transcriptionally inactive groups can diverge from metabolically active yet taxonomically minor ones. This functional restructuring shows that DCM assemblages are shaped by physical and chemical gradients and the differential metabolic activity of key taxa. These insights highlight the limitations of taxonomy for predicting ecological function and emphasize the need for integrative approaches to connect protist diversity with biogeochemical processes in stratified marine ecosystems.

To facilitate cross-method comparison, we synthesized microscopy, pigment-based proxies, 18S rRNA gene metabarcoding, and metatranscriptomic results into an integrated framework (Supplementary Table S3). This side-by-side comparison highlights both concordant and contrasting depth-related patterns among major protist groups and clarifies how methodological resolution and detection biases influence interpretations of community structure and function. Rather than providing quantitative equivalence, this synthesis offers a qualitative framework for contextualizing taxon-specific trends across analytical approaches and for identifying cases where taxonomic presence and functional activity are decoupled.

Overall, the convergence of multiple lines of evidence shows that the DCM is a dynamic, functionally distinct ecological niche shaped by vertical specialization and metabolic complementarity among coexisting taxa, rather than a passive aggregation layer. This implies that scaling DCM processes to regional carbon and nutrient budgets requires consideration of taxon-specific metabolic activity, because functional contributions to biogeochemical fluxes are not necessarily proportional to organismal abundance. At the same time, because this study is based on observations from a single station, the spatial representativeness of the results is inherently limited; however, the mechanistic patterns identified here offer a useful framework for interpreting protist-mediated biogeochemical processes in other stratified systems with comparable DCM structures. This integrative perspective underscores the critical role of metabolic versatility and ecological differentiation in protist communities in contributing to marine productivity, biogeochemical cycling, and carbon sequestration in stratified marine ecosystems.