A 60-L surface water was collected using a clean bucket from a sampling station (SNUTY) located in Buksin Bay off Tongyeong, Korea on May 30, 2024, and then gently screened with a 100-μm sieve (Fig. 1A). The dissolved oxygen (DO), water temperature, and salinity were measured using a YSI EXO 1 instrument (YSI Inc., Yellow Springs, OH, USA). Water was directly transported to the laboratory and placed overnight in a temperature-controlled chamber at the same water temperature as the water sampled at the site. The water sampled at the site exhibited a DO of 5.8 mg L⁻¹, a pH of 8.4, a water temperature of 21.8°C, and a salinity of 31.6.

The experiment was designed to explore changes in the structure of protist communities during the incubation of seawater for seven days under the normoxic (7.0 mg L⁻¹) and hypoxic (1.5 mg L⁻¹) conditions using a metabarcoding method.

The water in the jar was equally distributed into six experimental wide-mouth 5-L Duran bottles (Duran GLS80; Schott, Mainz, Germany). Triplicate bottles for the normoxic condition were placed on shelves outside an oxygen-controlled glovebox in a temperature-controlled chamber, and the ones for the hypoxia condition inside the oxygen-controlled glovebox (Fig. 1B). All bottles were incubated at 22°C with a 14 : 10 light : dark cycle illuminated by 30 μmol photons m⁻² s⁻¹ of cool white fluorescent light. The DO sensor (Multilab FDO 4410 IDS sensor) inside each bottle was calibrated and assembled into a Multilab 4010-3W (YSI Inc.), and the real-time DO was recorded. A stirring magnet was placed inside the bottle to maintain the desired oxygen level throughout the incubation period. Simultaneously, the pH and water temperature in the bottles were recorded using Multilab IDS 4110 pH and temperature sensors connected to a Multilab 4010-3 W (YSI Inc.).

To create the hypoxic condition without altering the pH, 400 parts per million (ppm) of CO₂ and the rest balanced with N₂ were added to a compressed gas cylinder (Clark and Gobler 2016, Bausch et al. 2019, Eom et al. 2024). The mixture of CO₂ and N₂ released from the cylinder was sent to an oxygen-controlled glovebox (Fig. 1C). A Coy Oxygen Controller (Coy Laboratory Products, Grass Lake, MI, USA) placed between the cylinder and glovebox maintained the target hypoxic condition continuously in the glovebox during the entire experiment. To avoid possible shock to different O₂ concentrations, protist communities were acclimated to the normoxic and hypoxic conditions for two days before the experiment.

After thorough stirring inside each bottle, an aliquot of 50 mL was taken from each bottle every day, except on day 6. The water incubated under the normoxic condition was sampled in a temperature-controlled chamber, whereas sampling under the hypoxic condition was conducted in an oxygen-controlled glovebox. To eliminate possible sudden oxygen changes inside the glovebox during sampling, sampling tubes or bottles were placed in a passbox connected to the glovebox (Fig. 1D). After maintaining the oxygen condition inside the passbox and glovebox, the water was subsampled using sampling tubes or bottles that were moved from the passbox to the main chamber of the glovebox.

A 20-mL aliquot of the 50-mL subsampled water was fixed using 5% Lugol’s solution and stored in polyethylene vials for comparative microscopic observation and cell enumeration. To determine whether detected amplicon sequence variants (ASVs) truly represent actual cells, micrographs were taken and cell enumerations were performed. For cell micrograph analysis, samples were placed on confocal dishes with a cover glass at 1,000× magnification using a digital camera (Zeiss Axiocam 506 and Zeiss Axiocam 820 color; Carl Zeiss Ltd., Göttingen, Germany) attached to an inverted light microscope (NFEC-2024-12-301531; Zeiss Axiovert 200M and Zeiss Axio Observer 7; Carl Zeiss Ltd.). For cell enumeration, dominant dinoflagellates, diatoms, ciliates, and cercozoans under the hypoxic condition on day 7 were determined by counting >200 cells or all cells in triplicate 1-mL Sedgwick Rafter chambers under a compound microscope. Only morphologically intact cells were carefully counted.

Another 20-mL aliquot of the 50-mL subsampled water was filtered through a GF/F membrane filter (Whatman Inc., Clifton, NJ, USA), and the filtrate in polyethylene vials was used for nutrient analyses. The concentrations of NO₃ + NO₂ (hereafter NO₃), NH₄ (ammonium), PO₄, and SiO₂ were measured using a 4-channel nutrient auto-analyzer (QuAAtro; Seal Analytical GmbH, Norderstedt, Germany) (Ok et al. 2021).

Subsequently, the filter was placed in a 15-mL conical tube, and 10 mL of 90% acetone was added. Then, the tubes were sonicated for 10 min, wrapped in aluminum foil, and stored in a 4°C chamber overnight. Afterwards, the tubes were then centrifuged to pipe 5 mL of the supernatant to measure chlorophyll-a (Chl-a) concentration using a 10-AU Turner fluorometer (Turner Designs, Sunnyvale, CA, USA).

A 10-mL aliquot of the 50-mL subsampled water was placed on a Petri dish, and living organisms were roughly observed to decide which samples should be analyzed for metabarcoding. Samples taken on days 0, 3, and 7 were decided to be used for metabarcoding analysis because notable changes in protist composition were observed between consecutive-day samples. Moreover, two consecutive single-cell isolations were performed on dinoflagellates and diatoms under the hypoxic condition on day 7 to establish clonal cultures.

For eDNA, an additional 500-mL aliquot was taken from each experimental 5-L Duran bottle and was filtered using 25-mm GF/C membrane filters (Whatman Inc.) (Ok et al. 2023a).

The eDNA in the membrane filters was extracted using a DNeasy PowerSoil Pro kit (Qiagen, Hilden, Germany). The extracted eDNA was then quantified using a Qubit fluorometer with Quant-IT PicoGreen (Invitrogen, Waltham, MA, USA), and stored at −20°C until polymerase chain reaction (PCR) could be performed.

Metabarcoding libraries were generated using a two-step PCR method as detailed by Ok et al. (2023a). Briefly, a sequencing library was prepared to amplify the target genes using the Illumina metagenomic sequencing library protocols (San Diego, CA, USA). For the first PCR, the 5 ng of genomic DNA was amplified with a 5× reaction buffer, 1 mM dNTP mix, and 500 nM of universal primers (forward primer TAReuk454FWD1, 5′-CCAGCASCYGC GGTAATTCC-3′ and reverse primer V4 18S Next.Rev, 5′-ACTTTC GTTCTTGATYRATGA-3′) targeting the V4 region of the 18S rRNA gene for protists (Stoeck et al. 2010, Piredda et al. 2017), with Herculase II fusion DNA polymerase (Agilent Technologies, Santa Clara, CA, USA). After the first PCR was conducted according to the methods in Ok et al. (2023a), the amplicons were purified using AMPure beads (Agencourt Bioscience, Beverly, MA, USA). Then, using 10 μL of the first PCR product with NexteraXT Indexed Primers, the second PCR was performed under the same conditions with the exception that 10 cycles were run instead of 25 cycles. The purified products were quantified using quantitative PCR following the protocols of the KAPA Library Quantification kits for Illumina sequencing platforms. It was qualified using a TapeStation D1000 ScreenTape (Agilent Technologies, Waldbronn, Germany). Paired-end sequencing was conducted using an Illumina MiSeq platform (Illumina) at Macrogen (Seoul, Korea).

Raw sequencing data from Illumina MiSeq were sorted by sample using index sequences, generating paired-end FASTQ files for each sample. After sequencing, Cutadapt (ver. 3.2) was used to remove the adapter and primer sequences and trim both forward (Read 1) and reverse (Read 2) reads to 240 and 200 bp, respectively. DADA2 software package (ver. 1.18.0) was used to correct errors by excluding sequencing with an expected error rate ≥2 in the amplicon sequencing data (Callahan et al. 2016). Erroneous reads were denoised using the established error model for each batch. Paired-end sequences were assembled into single sequences, chimeras were removed using the DADA2 Consensus method with the remove BimeraDenovo function, and ASVs were clustered. Normalization was performed by subsampling based on the lowest number of reads across all samples using QIIME (ver. 1.9.0) software (Caporaso et al. 2010). BLAST+ (ver. 2.9.0) was used for the taxonomic assignment of each ASV sequence using a reference database (Camacho et al. 2009). ASVs were not assigned if the query coverage or identity was <85%. ASVs that were not assigned to the species level in the PR2 database after using BLAST+ were further annotated using the National Center for Biotechnology Information (NCBI) database. ASVs assigned to metazoans were omitted from further analyses. Raw sequencing data have been deposited in the NCBI Short Read Archive (accession No. PRJNA1218482).

Two-tailed independent-sample t-tests were conducted to determine the differences between the initial chemical properties under the normoxic and hypoxic conditions. All statistical analyses were performed using SPSS ver. 29.0 (IBM-SPSS Inc., Armonk, NY, USA), with p-value <0.05 as a statistical significance criterion.

Throughout the incubation period, DO, pH and water temperature were maintained at the target levels under the normoxic and hypoxic conditions (Table 1, Fig. 2A). Furthermore, the initial concentrations of Chl-a, NO₃, NH₄, PO₄, and SiO₂ under the normoxic condition were not statistically different from those under the hypoxic condition (two-tailed t-tests; p = 0.53, p = 0.17, p = 0.06, p = 0.72, and p = 0.30, respectively) (Table 1, Fig. 2B). Thus, only the DO levels under the normoxic and hypoxic conditions differed.

The Chl-a concentration (mean ± standard error) under the normoxic condition increased from 34.7 ± 0.8 on day 0 to 58.8 ± 5.4 μg L⁻¹ on day 7, while that under the hypoxic condition increased from 37.9 ± 3.4 to 81.9 ± 21.6 μg L⁻¹ (Table 1, Fig. 2B). Over the incubation period, the concentrations of NO₃, NH₄, PO₄, and SiO₂ in both normoxic and hypoxic conditions almost continuously decreased.

Across all the samples, 54,588 ± 1,481 reads per sample were obtained from sequencing the V4 region of the 18S rRNA gene. A total of 1,280 ASVs were detected and 1,058 ASVs were assigned, corresponding to 26 annotated phyla and 398 assigned species (Supplementary Table S1).

Among the phyla, the relative ASV abundances in Dinoflagellata under both oxygen conditions were the highest on all days (Fig. 3A, Supplementary Table S2). The relative ASV abundances of Perkinsozoa, Cercozoa, Bigyra, and Chlorophyta increased under normoxia from day 0 to day 7, whereas that of other phyla declined or remained similar (Fig. 3A, Supplementary Table S2). Under hypoxia, the relative ASV abundances of Cercozoa and Ochrophyta increased under hypoxia from day 0 to day 7, whereas that of other phyla declined or remained similar (Fig. 3A, Supplementary Table S2).

On day 0, the relative read abundance of Dinoflagellata under both conditions was the highest (Fig. 3B, Supplementary Table S3). On day 3, the relative read abundances of Perkinsozoa increased under both conditions and were the highest. However, on day 7, the relative read abundance of Perkinsozoa was the highest under the normoxic condition, whereas that of Cercozoa was the highest under the hypoxic condition.

Among the dinoflagellate orders, the relative ASV abundances of the order Syndiniales under both oxygen conditions were the highest on days 0 and 3, whereas that of the order Peridiniales were the highest on day 7 (Fig. 4A, Supplementary Table S4). The relative ASV abundance of Prorocentrales increased under hypoxia but decreased under normoxia from day 0 to day 7 (Fig. 4A, Supplementary Table S4).

On day 0, without unassigned taxa, the relative read abundances of the orders Gymnodiniales and Prorocentrales under the normoxic condition were the highest and second highest, respectively, but they were very similar (Fig. 4B, Supplementary Table S5). However, on day 7, the relative read abundance of the order Suessiales was the highest under the normoxic condition, whereas that of the order Prorocentrales was the highest under the hypoxic condition. The survival of species belonging to the order Prorocentrales under the hypoxic condition was confirmed by microscopic observations (Supplementary Fig. S1).

In total, 272 dinoflagellate ASVs were assigned to the phylum Dinoflagellata in the present study. In all the samples, 46 species were taxonomically identified at the species level (Tables 2 & 3). Among the 46 identified species, 31 were detected under both the normoxic and hypoxic conditions, 11 under the normoxic condition only, and four under the hypoxic condition only on days 0, 3, and 7 (Tables 2 & 3). Prior to the experiment, species detected on day 0 had been exposed to their respective oxygen conditions through a two-day preliminary incubation. Thus, under the normoxic condition, 42 species were taxonomically identified, whereas 35 species were identified under the hypoxic condition on days 0, 3, and 7.

In terms of the relative ASV abundance of dinoflagellates, without unassigned order, the proportion of parasitic dinoflagellates was the highest under both oxygen conditions on days 0, 3, and 7 (Fig. 5A). Under the normoxic condition, the proportion of phototrophic dinoflagellates (autotrophic or mixotrophic, not yet identified) was the second highest on days 0, 3, and 7. Under the hypoxic condition, the proportion of mixotrophic dinoflagellates was the second highest on days 0 and 7.

In the relative read abundance of dinoflagellates under the normoxic condition, the proportion of mixotrophic dinoflagellates was the highest on day 0, whereas that of phototrophic dinoflagellates was the highest on days 3 and 7 (Fig. 5B). In the relative read abundance of dinoflagellates under the hypoxic condition, the proportion of mixotrophic dinoflagellates was the highest on days 0, 3, and 7 (Fig. 5B).

The number of taxonomically identified dinoflagellate species that were detected on day 7 was 16 under the normoxic condition and 14 under the hypoxic condition (Tables 2 & 3). Under the normoxic condition, the trophic modes of the identified dinoflagellate species on day 7 were one exclusively autotrophic, nine phototrophic, four mixotrophic, one kleptoplastidic taxon, and one heterotrophic. However, under the hypoxic condition, the trophic modes of the identified dinoflagellate species on day 7 were one exclusively autotrophic, six phototrophic, three mixotrophic, one kleptoplastidic taxon, and three heterotrophic.

As the relative ASV abundances cannot directly confirm the presence or abundance of actual cells, the dominant protistan taxa under the hypoxic condition on day 7 were further quantified under a compound microscope. Among dinoflagellates, Prorocentrum triestinum exhibited the highest cell density, followed by Biecheleria-like spp. (Table 4). Other dinoflagellates such as Gyrodinium spp. and Scrippsiella sp. were also observed under the compound microscope. Among diatoms, Skeletonema spp. showed the highest density, followed by Psammodictyon-like species. Other diatoms, including Cylindrotheca closterium, Thalassiosira spp., Pseudo-nitzschia spp., and Chaetoceros spp., were also present. Ciliates, including the loricate ciliate Eutintinnus sp. and naked ciliates of various sizes, and cercozoans, including Vampyrella-like sp., were also observed. In addition, cells of two dinoflagellate species (Prorocentrum triestinum and Scrippsiella acuminata) and three diatom species (Chaetoceros curvisetus, Cylindrotheca closterium, and Skeletonema japonicum) were isolated from the water incubated under the hypoxic condition on day 7 and successfully established as clonal cultures (Supplementary Fig. S2). Each species was identified based on the internal transcribed spacer and large subunit ribosomal DNA sequences.

On day 7, under the hypoxic condition, Dinoflagellata, Cercozoa, and Ochrophyta accounted for the majority of the relative ASV abundance. Both relative ASV abundance and read abundance of Cercozoa and Ochrophyta under the hypoxic condition increased during seven-day incubation. Cercozoa has been found in the hypoxic waters and sediments of the world’s oceans (Table 5) (Stock et al. 2009, Kalu et al. 2023, Ok et al. 2023a). In particular, Cercozoa has been reported to be abundant in marine benthic and interstitial communities (Pawlowski et al. 2011, Harder et al. 2016), suggesting that they may be resilient to low-oxygen environments. The present study showed that the read abundance of the cercozoan Pseudopirsonia mucosa was the highest among cercozoans under the hypoxic condition on day 7 (Table 5). P. mucosa is known to be a parasitic protist that infects various diatoms (Kühn et al. 2004). Thus, its highest proportion in hypoxic waters on day 7 may be partially related to the increase in the relative read abundance of Ochrophyta, the majority of which were diatoms (Supplementary Table S1). Moreover, the presence of this species under the hypoxic condition has not been previously reported. Thus, the present study added P. mucosa to the list of species that are dominant under the hypoxic condition.

In terms of relative read abundance, the top five species in Ochrophyta under the hypoxic condition belonged to diatoms according to the present and previous study (Table 5) (Ok et al. 2023a). They can tolerate hypoxia by producing oxygen. Few studies have been conducted on the effects of hypoxia on diatom growth (Wu et al. 2012, Chen et al. 2025). The growth rate of the diatom Thalassiosira weissflogii increased under hypoxia with an increased net photosynthetic rate (Sun et al. 2022). The growth rates of diatom Thalassiosira pseudonana and Skeletonema costatum were affected by hypoxia; however, the rates were still positive at the DO concentration of 1.3 mg L⁻¹ for T. pseudonana, and 0.5 and 2 mg L⁻¹ for S. costatum, respectively (Wu et al. 2012). The results of the present study provide information on diatoms which can tolerate and even show higher growth rates under hypoxia, and further studies on the growth rate of the other dominant diatoms Thalassiosira tenera, Psammodictyon panduriforme, and Minidiscus variabilis are needed.

In Perkinsozoa, the relative read abundance of the dominant species Parvilucifera infectans increased under both oxygen conditions on day 3, but decreased on day 7. Parvilucifera infectans is an endoparasitoid that infects a wide range of dinoflagellates, including those from Dinophysiales, Gonyaulacales, Gymnodiniales, Peridiniales, and Suessiales (Garcés et al. 2013, Alacid et al. 2015). The decline in the relative read abundance of Dinoflagellata under both the normoxic and hypoxic conditions on day 3 co-occurred with the increase in the relative read abundance of Perkinsozoa. This result suggests that P. infectans might have been actively infecting and potentially reducing dinoflagellate populations. However, on day 7, the population of the parasitic P. infectans declined under both the oxygen conditions, possibly due to a lack of available dinoflagellate hosts. Furthermore, the relative read abundance of Perkinsozoa on day 7 declined more under the hypoxic condition than under the normoxic condition. P. infectans has not been reported to infect Prorocentrales, including the dominant species in the present study, Prorocentrum triestinum (Garcés et al. 2013). Thus, the high relative read abundance of Prorocentrales on day 7 under hypoxia may have partially contributed to a lower relative read abundance of Perkinsozoa compared to normoxia. To verify this hypothesis, further investigation is needed to determine whether the increase in the Prorocentrum triestinum population affected the decline of Parvilucifera infectans under the hypoxic condition.

Based on relative read abundance on day 0, the dinoflagellate composition differed between the hypoxic and normoxic conditions in the present study. For example, Prorocentrales was the dominant dinoflagellate order under hypoxia, whereas Gymnodiniales dominated under normoxia. This discrepancy may be attributed to the preliminary incubation, which allowed protists to acclimate to their respective oxygen conditions before the experiment. Prorocentrales possess thick thecal plates, which may provide protection against environmental stressors such as the sudden oxygen reduction (Janouškovec et al. 2017). In contrast, Gymnodiniales have relatively thin cell walls, making them less adapted to the sudden transition to hypoxic environments (Janouškovec et al. 2017). To verify this hypothesis, further investigation into the mechanisms that enable Prorocentrales to dominate under hypoxia is needed.

In Dinoflagellata, the relative ASV abundance of the order Peridiniales was the highest under the hypoxic condition on day 7, whereas the relative read abundance of the order Prorocentrales was the highest. Prorocentrum triestinum belonging to Prorocentrales, Scrippsiella precaria and Scrippsiella acuminata belonging to Peridiniales, and Biecheleria brevisulcata and Biecheleria tirezensis belonging to Suessiales ranked in the top five dinoflagellate species in read abundance under the hypoxic condition (Table 5, Supplementary Fig. S1). Prorocentrum triestinum, B. brevisulcata, S. acuminata, and S. precaria have been found in hypoxic seawater (Table 5) (Ok et al. 2023a). Moreover, clonal cultures of P. triestinum and S. acuminata from single-cell isolation were established from day 7 under the hypoxic condition (Supplementary Fig. S2). It is highly possible that B. tirezensis is present in the hypoxic waters of natural environments.

In the present study, 35 dinoflagellate species were detected under the hypoxic condition throughout the entire study period including days 0, 3, and 7. Of these detected species during the entire study period, 16 species were previously known to be present under hypoxia, but 19 species were newly determined (Table 6). These results extend the number of the dinoflagellate species are present under hypoxia. Among the newly discovered dinoflagellates that were present under hypoxia, autotrophic species were two, phototrophic species 12, mixotrophic species one, and heterotrophic species four (Table 6). Furthermore, of these 35 detected dinoflagellate species, 14 were present on day 7, and 6 were newly identified to be detected under the hypoxic condition: one autotrophic, four phototrophic, and one heterotrophic dinoflagellates (Table 6). Thus, some dinoflagellate species remained detectable longer than the others, suggesting differential tolerances to hypoxia. Therefore, careful consideration of the presence of dinoflagellate species in relation to the duration of hypoxia is needed when interpreting the findings. Autotrophic, mixotrophic, phototrophic, and kleptoplastidic dinoflagellates can produce oxygen and may release it into ambient water, which may alleviate hypoxia. Moreover, mixotrophic dinoflagellates are known to survive the hypoxic condition by feeding on their prey. Under the hypoxic condition, the growth rate of the mixotrophic dinoflagellate Alexandrium pohangense satiated with prey is positive, whereas that of A. pohangense starved is negative (Eom et al. 2024). Furthermore, Rocke et al. (2016) observed that mixotrophic dinoflagellates became more dominant as the hypoxia intensified, suggesting a shift from autotrophy to phagotrophy. Heterotrophic dinoflagellates can also survive the hypoxic condition by growing on their prey. That is, if the growth rate of a heterotrophic dinoflagellate species feeding on prey exceeds its mortality rate due to hypoxia, the heterotrophic dinoflagellate species can survive. Under hypoxic conditions, the growth rate of the heterotrophic dinoflagellate Gyrodinium dominans satiated with prey was positive, whereas that of G. dominans starved was negative (Eom et al. 2024). Therefore, feeding is a critical survival strategy for mixotrophic and heterotrophic dinoflagellates under the hypoxic condition.

In previous metabarcoding studies, relative ASV abundances have been used to assess marine protist communities (Vasselon et al. 2018, Santoferrara et al. 2022, Ok et al. 2023a). However, the choice of barcoding region, the efficiency and specificity of PCR primers, and the varying completeness of reference databases across protist taxa can result in underestimation or overestimation of ASVs in different protist groups (Pawlowski et al. 2016, Martin et al. 2022). Although the V4 region of the 18S rRNA gene is commonly employed in protist metabarcoding (Kezlya et al. 2023), it may lead to an overestimation of dinoflagellate ASVs, as the high variability in the rDNA regions within cells can result in the detection of multiple ASVs (Stuart et al. 2024). Furthermore, the incompleteness of a diatom barcoding database in the V4 region can lead to an underestimation of diatom ASVs (Pawlowski et al. 2016). In the present study, a higher number of diatom species under the microscopy were enumerated compared to that of dinoflagellate species, although a higher number of dinoflagellate ASVs were detected compared to the diatoms on day 7. This may have been caused by the underestimation and overestimation of ASVs in diatoms and dinoflagellates, respectively. Thus, potential taxonomic biases in metabarcoding results need to be carefully considered. Furthermore, in the present study, the quantitative cell enumeration, microscopic observations, and clonal culture establishment from single-cell isolation and identification of species confirmed that the dominant taxa were composed of morphologically intact cells or viable cells and not cellular debris (Table 4, Supplementary Figs S1 & S2). Therefore, integrating microscopic observation, cell counting, and live cell cultivation with metabarcoding analyses is essential for accurately assessing the protist community structure and distinguishing actual cells from degraded material.

Under the closed condition of the present study, the absence of protist introduction and the progressive accumulation of cellular debris may have affected the detected ASV profile, making it different from natural environments. Therefore, it is important to consider that the present study was conducted in a closed system when interpreting its results in the context of natural environments. Furthermore, in ocean, low pH conditions are frequently associated with hypoxia (Howarth et al. 2011, Gobler and Baumann 2016, Guo et al. 2022). However, in the present study, the low DO condition was established by using N₂ + 400 ppm CO₂ gas to maintain stable pH levels. This provides an ideal condition for investigating the single effects of hypoxia on protistan community structures. However, the combined effects of hypoxia and acidification resulted in different physiological responses (e.g., survival, growth, photosynthesis, and dark respiration) of a marine protist, the dinoflagellate Amphidinium carterae, compared to single hypoxic effect (Bausch et al. 2019). Therefore, when applying the findings of the present study to natural environments, the potential impact of CO₂ accumulation should be considered.

The present study provides valuable insights into the dynamics of protist communities under the hypoxic condition. The results of the present study suggest that hypoxia can alter protist community composition. Future research should explore the physiological mechanisms of protist survival under the low-oxygen condition.