Environmental driving forces and phytoplankton diversity across the Ross Sea region during a summer–autumn transition

The Ross Sea is a highly productive system characterized by a seasonal succession of phytoplankton groups. However, most of the current understanding is based on observations on the continental shelf in spring and summer at relatively coarse taxonomic resolution. Here, we characterize community composition (class to species) using V4‐18S rRNA gene metabarcoding on transects to and across the Ross Sea (shelf‐slope and oceanic subregions) during two voyages covering the Austral summer–autumn progression in successive years. Phytoplankton composition shifted from low‐diversity diatom‐dominated (50%) communities during the summer to more diverse dinoflagellate‐dominated (48%) systems during the lower‐productivity autumn season. Prymnesiophyceae abundance was low on both voyages (10%), except on the southeast shelf‐slope, where Phaeocystis antarctica dominated a shallow mixed layer, contrary to its reported preference for deep‐mixing conditions. Amplicon sequence variant analysis identified distinct spatial patterns for two P. antarctica genotypes yet clustered certain species of Bacillariophyta and Prymnesiophyceae, indicating similar environmental preference for genotypes in these groups. Multivariate analysis of environmental drivers found a higher correlation of community composition variation with salinity and macronutrients, but less so with mixed layer depth, considered a primary determinant of taxonomic composition in the Ross Sea. Comparison between years established that community composition was temporally more stable in oceanic relative to shelf‐slope waters. This study of seasonal variation of phytoplankton community composition at finer taxonomic resolution provided insights into species‐ and strain‐specific distribution, ecological preferences, and relationships to environmental conditions in the wider Ross Sea to inform modeling and projection of future regional changes.

The wider Ross Sea region (RSR) is delimited by the southern boundary of the Antarctic Circumpolar Current (ACC) (Llopis Monferrer et al. 2021) and encompasses the continental shelf (and shelf-slope) and off-shelf oceanic environments.The continental shelf has been extensively studied due to the high productivity and massive seasonal phytoplankton blooms associated with polynyas (Smith and Gordon 1997), with the primary temporal focus on the transition from spring to summer (Peloquin and Smith 2007).Conversely, the oceanic off-shelf waters are relatively less studied and are regarded as a zone of constant lower productivity relative to the shelf (El-Sayed et al. 1983).This difference in the level of scrutiny has hampered the development of seasonal descriptions of phytoplankton dynamics across the RSR.
Seasonality of environmental drivers in the Ross Sea is more evident than in other Antarctic regions (Holland et al. 2017;Fogt et al. 2022).This seasonality influences the timing of polynya formation (Arrigo et al. 1998) and affects the physicochemical environment and biological processes such as phytoplankton bloom dynamics and composition (Smith et al. 2000).Biomass accumulation in the Ross Sea is associated with increases in temperature, solar irradiance, and iron concentration in spring and early summer (Smith et al. 2014) with a peak in biomass usually observed in December (Smith et al. 2000).These conditions support the bloom formation of Prymnesiophyceae (Phaeocystis antarctica), and Bacillariophyta (diatoms)-two phytoplankton classes with differing contribution to trophic and vertical carbon export (Smetacek 1999;Schnack-Schiel and Isla 2005;Wolf et al. 2016).These two groups also have differing temporal and spatial blooming patterns, which have been attributed to their physiological traits (Mills et al. 2010), historically ascribed to the dominant influence of mixed layer depth (MLD) (Arrigo et al. 1999;Goffart et al. 2000).P. antarctica blooms occur in the south-central shelf region when sea ice cover melts in early spring (Arrigo et al. 1999) in a deeper surface mixed layer, reflecting its tolerance for low light levels (Garcia et al. 2009).Diatom blooms tend to follow in summer in coastal areas and under more stratified conditions (Saggiomo et al. 2021) due to their preference for higher light levels (Arrigo et al. 1998(Arrigo et al. , 2010)).From late summer to early autumn, phytoplankton biomass and productivity decline associated with a decrease in irradiance due to mixed layer deepening and sea ice formation (Smith et al. 2000), while iron limitation peaks in summer and remains low during autumn (Ryan-Keogh and Smith 2021).These conditions shift community composition toward heterotrophs such as dinoflagellates (Dennett et al. 2001).However, despite these accepted seasonal patterns and the crucial role of mixing and irradiance (Arrigo et al. 1998), phytoplankton blooms can occur under compact sea ice (Horvat et al. 2022) and summer blooms of P. antarctica in stratified coastal waters have also been reported (Mangoni et al. 2019), supporting that the spatiotemporal distribution of phytoplankton groups may result from more complex interaction of environmental drivers (van Hilst and Smith 2002;Peloquin and Smith 2007).
The phytoplankton community in the Ross Sea has typically been described by microscopy (Smith et al. 2003) and diagnostic pigment analysis (Arrigo et al. 1999;Goffart et al. 2000).These methods have been effective in linking the compositional and functional aspects of phytoplankton communities.However, the relatively coarse taxonomic resolution of pigment analysis to class level and the inability of microscopy to describe taxa without distinct morphological features (generally small-sized taxa) or identify cryptic species (Piredda et al. 2018) have hindered the description of the Ross Sea phytoplankton community.This lack of insight at finer taxonomic scale limits understanding of groups with distinct biogeographic patterns and functional diversity (Bishop et al. 2022), particularly in low biomass regions where small cells tend to dominate (Cabré et al. 2016).
The application of genomic approaches to characterize phytoplankton community composition has identified high intraspecific diversity and also spatiotemporal trends previously undetected by pigment or microscopy (Santoferrara et al. 2020;Sow et al. 2020;Catlett et al. 2023).Consequently, genetic characterization is critical to understanding controls of community composition and function, and prediction of ecosystem response to climate change (Fuhrman and Campbell 1998;Farrant et al. 2016;Bishop et al. 2022).Current descriptions at such fine taxonomic level are scarce in the Southern Ocean (Lopes dos Santos et al. 2022) and primarily focus on the Antarctic Peninsula (Maria et al. 2021;Trefault et al. 2021;Liu et al. 2022).In the case of the RSR, phytoplankton taxonomic characterizations have been conducted mainly on the continental shelf (Arrigo et al. 2000;Mangoni et al. 2017), but not in the shelf-slope and oceanic subregions, limiting the determination of climate change impacts in the wider region.To address this gap, we carried out a taxonomically comprehensive characterization of the phytoplankton community from class to species and strain level to determine diversity trends among phytoplankton groups, with special attention given to diatoms and Prymnesiophyceae.This facilitated the examination of the relationship between phytoplankton community composition and environmental variability and how this varied from summer to autumn and across subregions of the Ross Sea, including stations in the less-studied oceanic and shelf-slope.

Study site and sampling strategy
Samples were collected during two research voyages conducted in the RSR onboard the R/V Tangaroa.In 2018 (TAN1802, February-March 2018) the voyage covered the late summer/early autumn period as in previous studies (Gardner et al. 2000) and focused on the northern continental shelf-slope of the Iselin Bank (70-73 S) and off-shelf waters between 65 S and 70 S.In 2019 (TAN1901, January-February 2019), the voyage reflected summer conditions.TAN1901 spatially overlapped with TAN1802 on the Iselin Bank and off-shelf area and extended the survey from the previous year to include Cape Adare and the southeast region of the Ross Sea (73-76 S) and off-shelf waters of the Ross Gyre (70-75 S) (Fig. 1).
CTD stations were regionally classified into the sub-Antarctic zone (SAZ), polar front zone (PFZ), southern Antarctic Circumpolar Current (sACC), and RSR, based on their geographical distribution following Llopis Monferrer et al. ( 2021), with oceanographic fronts delimited according to Orsi and Harris (2019).We further partitioned stations in the RSR into Oceanic (off-shelf including the Ross Sea Gyre influenced station, CTD26 and 27), and Cape Adare, Iselin Bank, and Southeast Shelf-Slope (shelf-slope) subregions based on geographical location (Supporting Information Tables S1, S2).The sACC included CTD 02, 40, 41, and 43 from TAN1802, located south of the oceanographic boundary but clustered separately from the RSR stations based on physicochemical conditions (nutrients and salinity; Supporting Information Fig. S1), and CTD 05 from TAN1901 according to Llopis Monferrer et al. (2021).
During both voyages, a Sea-Bird Electronics (SBE) 911plus CTD combined with a 24 Â 10-L external-spring Niskin-type bottle rosette (Ocean Test Equipment Standard 10 SBE) was used to obtain salinity, temperature, dissolved oxygen depth profiles and to collect seawater from 3 to 8 depths in the upper 500 m.The CTD sensor was configured with SBE 3plus, SBE 4, and SBE 43 dual sensors for the parameters above, and a seapoint fluorescence sensor, and a photosynthetically active radiation (PAR) sensor (Biospherical Instruments QCP-2300L-HP).DNA samples were collected on all stations during TAN1901 (31 stations) and only in 8 out of the 27 stations sampled during TAN1802 (Supporting Information Table S1).

Irradiance, mixing, and dissolved oxygen
The MLD was calculated using a criteria of 0.03 kg m À3 change in potential density anomaly relative to the surface (considered as 10 m) (Gardner et al. 1995), consistent with other studies in the region (Kaufman et al. 2014).The depth of the euphotic zone (Z eu ) was calculated as À4.6/log of slope of broadband (400-700 nm) downwelling irradiance measured as quanta (PAR) (i.e., 1% surface light depth based on estimated diffuse downwelling, broadband attenuation coefficient).Dissolved oxygen anomaly (O 2 anomaly) was calculated using temperature and salinity (Garcia and Gordon 1992).

DNA sampling and extraction procedures
DNA samples were collected from 5 to 7 depths during TAN1802 and 3 to 6 depths during TAN1901 by filtering 1.5 to 2 L of seawater from each depth through 0.2 μm polyethersulfone SterivexTM filter units (Millipore), using a peristaltic pump (Masterflex L/S; Cole-Parmer).The Sterivex units were filled with RNAlater, flash-frozen in liquid nitrogen, and stored at À80 C until analysis.In total, 48 samples for DNA were collected from 8 stations during TAN1802 and 92 samples from 31 stations during TAN1901.DNA was extracted using the "Easy Blood and Tissue Kit" (Qiagen) according to methods described in Gutiérrez-Rodríguez et al. (2022).

18S rRNA gene amplification, sequencing, and processing
The V4 hypervariable region of the 18S rRNA gene was amplified using the eukaryotic primers TAReuk454FWD1 (CCAGCASCYGCGGTAATTCC) and the 18S Next.Rev (ACTTTCGTTCTTGATYRATGA) with partial overhang adapters for Illumina indexing modified from Piredda et al. (2017).PCR amplification was conducted in 50 μL total volumes according to Gutiérrez-Rodríguez et al. (2022).Amplicons were visualized on a 1% agarose gel prior to submission for clean-up, barcode indexing, library construction and 2 Â 250 bp pair-end sequencing on an Illumina MiSeq at the Genotoul GeT core facility (Toulouse, France).
Fastq files were processed using DADA2 workflow in R (Callahan et al. 2016) using the ABiMS platform from Station Biologique de Roscoff.Sequences were truncated to 240 bp and forward and reverse primers trimmed by their length (20 and 21 nucleotides, respectively).Pair-end reads were then denoised, and a sequence variant inference algorithm was applied to generate an amplicon sequence variant (ASV; 100% similarity) before merging and removing chimeras.The taxonomic assignation was performed with PR2 database version 4.12 (https://pr2-database.org/;Guillou et al. 2012), which includes 8 levels of taxonomic structure (Kingdom/Supergroup/ Division/Class/Order/Family/Genus/Species), and the RDP naïve Bayesian classifier method (Wang et al. 2007) implemented in DADA2.Taxonomic assignation was followed by a bootstrapping approach to determine the confidence of taxonomic assignation.In this case, taxonomic assignation was retained for bootstrap confidence of 80% (Catlett et al. 2020).
ASVs of interest were BLASTed against GenBank for taxonomic confirmation.The metabarcode annotated dataset was prefiltered to select photosynthetic taxa classified within the phytoplankton classes Chlorophyta, Dinoflagellata, Cryptophyta, Haptophyta, and Ochrophyta ranked in PR2.In addition, we included Cercozoa and Radiolaria due to the presence of mixotrophic taxa (i.e., organisms that engage in both photoautotrophy, sometimes through photosymbiosis, and heterotrophy; Mitra et al. 2016).Strict heterotrophic and parasitic groups such as Syndiniales and Sarcomonadea were excluded from further analysis.

ASV and diversity analysis
The ASV table was standardized by the median sequencing depth across samples to operate with species abundance rather than percentage (Gutiérrez-Rodríguez et al. 2022).The analysis of ASV was conducted using the R packages Phyloseq (McMurdie and Holmes 2013), DESeq2 extension (Love et al. 2014), andMicroViz (Barnett et al. 2021).Taxonomy plots were made in MicroViz, selecting the 12 most abundant taxa at the class level.Identification of ASV with a significant fold change between voyages (considered as p < 0.05) was conducted using the DESeq2 extension.Settings for DESeq2 included the Wald test and parametric fit for dispersion.Results from DESeq2 are shown for the ASV with a significant log2 fold change between À10 and 10.
Alpha diversity was assessed based on the Chao1 index using estimate_richness in Phyloseq, considering only samples within the euphotic zone to allow comparisons between all stations and voyages.To determine if differences observed by Chao1 index, MLD and Chl a were statistically significant between voyages, a one-way analysis of variance (ANOVA) with Tukey's honestly significant difference (HSD) test was calculated using the function aov and TukeyHSD, respectively, in R. Permutational multivariate ANOVA (PERMANOVA) analysis was applied at species level using the MicroViz package.Distance was calculated based on Bray-Curtis similarity and redundancy analysis conducted with 999 permutations.
Due to the relevance of Bacillariophyta (here it includes Mediophyceae, Bacillariophyceae and Coscinodiscophyceae classes) and Prymnesiophyceae in the Ross Sea (Arrigo et al. 2000), correlation analysis for ASV belonging to these two classes and environmental parameters was carried out separately by voyage using sparse partial least squares (mixOmics package on R; Lê Cao et al. 2016) applied in regression mode as in similar analysis (Guidi et al. 2016;Tréguer et al. 2018).Results are presented for the 30 ASV with the highest correlation scores independently of their abundance for each voyage (Supporting Information Tables S3, S4).

Curation of taxonomic assignation
The verification of the taxonomic assignment of ASV of interest confirmed ASV0069 as Fragilariopsis cylindrus.In the case of ASV0110 (assigned as Fragilariopsis sublineata) the sequences also matched with Fragilariopsis kerguelensis and Fragilariopsis curta (Supporting Information Table S5), indicating that the taxonomic assignation ASV0110 cannot be verified (Gutiérrez-Rodríguez et al. 2022).Henceforth, ASV0110 is referred to as Fragilariopsis sp.The results from DESeq2 analysis are shown at the genus level with species, and ASV is mentioned for specific genera.In all of those cases, ASV taxonomic assignation to species level was confirmed by blastn.

Environmental conditions
During autumn (TAN1802) surface temperature varied from 7.4 C in the northernmost station of the PFZ to À1.1 C in the southern RSR, while the wider latitudinal coverage of the early summer voyage (TAN1901) increased the temperature range surveyed between the SAZ and the RSR (1.0-10.8C) (Fig. 2).MLD was on average shallower during TAN1901 (mean AE SD = 38 AE 16 m) compared to TAN1802 (54 AE 19 m) and was significantly different between voyages within the RSR (ANOVA, F 1,40 = 11.8, p < 0.05).During TAN1901 in the RSR, the MLD was deeper in Oceanic and shallower in the Southeast Shelf-Slope and the Oceanic Ross Gyre influenced stations (CTD27 and CTD28) (Supporting Information Table S2).TAN1802, on the other hand, had a deeper MLD in Iselin Bank relative to Oceanic.The depth of the euphotic zone was, on average, similar between the two voyages but more variable in TAN1901 (Z eu_TAN1802 = 92 AE 6 m, Z eu_TAN1901 = 87 AE 19 m).
On both voyages, DRSi surface concentrations (10-25 m) were lowest (< 3 μmol L À1 ) in SAZ and PFZ, increasing across the sACC and into the RSR where they remained stable (60-70 μmol L À1 ), whereas NO 3 concentrations were lowest in the SAZ (< 20 μmol L À1 ) and increased from the PFZ southwards (Fig. 1).During TAN1901, two stations had distinct physical and chemical conditions: (1) CTD27, a station near the Ross Gyre that had low DRSi (42 μmol L À1 ) and NO 3 concentrations (26 μmol L À1 ), similar to sACC waters further north; and (2) CTD26, located on the Southeast Shelf-Slope, had the lowest surface NO 3 (16 μmol L À1 ) and the highest NH 4 concentrations (2.6 μmol L À1 ) of all stations in the RSR (10-fold higher than the average RSR-NH 4 concentration during the TAN1901).
Size-fractionated Chl a identified different community size structure between voyages and regions (Fig. 3).Picophytoplankton (0.2-2 μm) (mean AE SD = 36% AE 20%) and microphytoplankton (> 20 μm) (33% AE 21%) contributed similarly to TChl a during TAN1901, while the contribution of larger cells (> 20 μm, 43% AE 15%) was higher during TAN1802, despite the lower Chl a levels in the latter (Fig. 3).During TAN1802 the size structure was similar in the sACC and RSR, while it differed during TAN1901.Similarly, spatial variability within the RSR was greater during TAN1901 than TAN1802 with differing regional trends between seasons.For example, during TAN1901, microphytoplankton dominated at Cape Adare (67% AE 4%) and accounted for < 30% of the TChl a on the Iselin Bank, whereas this pattern was reversed in TAN1802 (Fig. 3).Stations along the Southeast Shelf-Slope also showed a co-dominance of pico-and micro-phytoplankton size fractions, except at CTD26 where picophytoplankton (98%) dominated the community.
Reads belonging to Bacillariophyta, Dinophyceae, and Prymnesiophyceae classes dominated the phytoplankton metabarcode dataset, but with variations in their relative contribution between voyages and regions (Fig. 4).Within the euphotic zone, autumn (TAN1802) metabarcode was dominated by Dinophyceae (mean 43%, range 26%-55% of total standardized read abundance) followed by Bacillariophyta (27%, range 7%-47%).The relative dominance of Dinophyceae was more pronounced in the PFZ region, which also had the highest relative contribution of Pelagophyceae and Mamiellophyceae (Chlorophyta) (Fig. 4A).The relative contribution of Prymnesiophyceae during autumn was lower (10%, range 3%-27%) than Bacillariophyta and Dinophyceae, with higher relative abundance in the sACC and Oceanic subregion of the RSR relative to shelf-slope on and near the Iselin Bank.The relative contribution of Bacillariophyta was lowest at the northernmost stations in the PFZ, increased south of the PF and remained relatively constant (mean AE SD = 35% AE 11%) through the Oceanic subregion and shelf-slope waters of the RSR.Cryptophyceae were consistently present at relatively low abundance (< 3%) across stations south of the PF.Samples collected in the lower euphotic zone and below (100-200 m) showed increased relative abundance of Dinophyceae and radiolarian groups (Polycystinea, Acantharea, and RAD-B).
In summer (TAN1901) there was a greater relative contribution of Bacillariophyta reads with a wider range (mean 50%, range 2%-92% of total standardized read abundance across all stations), a corresponding reduction in Dinophyceae (23%, 3%-45%) and a similar relative abundance of Prymnesiophyceae (14%, 1%-70%) in the euphotic zone.Bacillariophyta relative contribution to the metabarcode increased southwards from SAZ through to the sACC (Fig. 4B), consistent with increasing availability of silicate south of the PF (Fig. 2).In the RSR, the relative reads abundances of Bacillariophyta and Dinophyceae were inversely related: Bacillariophyta were higher on Iselin Bank and at Cape Adare, coinciding with the lowest abundance of Dinophyceae, whereas Dinophyceae were higher in Oceanic subregion to the north (CTDs 06, 29, 30) and along the Southeast Shelf-Slope (CTDs 24, 25, 26) where it had the lowest relative abundance (Fig. 4B).During this voyage, Prymnesiophyceae relative contribution peaked at the southernmost station (CTD26) located at the Southeast Shelf-Slope, where they accounted for 70% of total reads, with a lower relative contribution (2%-23%) in the northern part of Iselin Bank and other subregions within the RSR.Pelagophyceae and Mamiellophyceae had a greater presence in the RSR, especially in Oceanic and Iselin Bank.As with TAN1802, Cryptophyceae represented a small fraction of the community, and was highest ($ 5%) in the Oceanic subregion.
The genera with significant differences in standardized read abundance between voyages belonged mainly to Dinophyceae (more abundant in TAN1802-autumn), and Bacillariophyta (more abundant in TAN1901-summer) (Fig. 5A).Within the RSR, the seasonal trends showed a significantly higher abundance of ASV belonging to the diatom genera Cymbella, Pseudo-Nitzschia, and Chaetoceros in summer, and of Asteroplanus, Navicula, Gyrosigma, Fragilariopsis, and Minidiscus during autumn (Fig. 5B).Among Corethron, two ASV of Corethron inerme were identified, one for each voyage (ASV0304 for TAN1802 and ASV0291 for TAN1901).Three ASV of Chaetoceros were identified at higher abundance in TAN1901 (C.atlanticus ASV38547, Chaetoceros neogracilis ASV2304, and C. dichaeta ASV38538) and one in TAN1802 (C.dichaeta, ASV0163).In the RSR, all the Dinophyceae genera identified with a significant   higher reads abundance were found during TAN1802 (Fig. 5B).Conversely, in summer there was a higher abundance of two ASV of Chlorophyta; one belonging to Mamiellophyceae (Micromonas, ASV1599) and the other to Pyramimonadophyceae (Pyramimonas, ASV8140).In addition, one ASV of the Prymnesiophyceae (Phaeocystis, ASV1343) was found in higher abundance in the Ross Sea during summer TAN1901 voyage.Symbols represent the subregions within the RSR: Cape Adare (cross), Iselin Bank (triangle), Oceanic (circle), and Southeast Shelf-Slope (square).Arrows indicate the variables that were significant (p < 0.05) as identified by PERMANOVA (Supporting Information Table S8), with the length representing the strength of the correlation between community composition at each location and each environmental factor.Z eu = depth of the euphotic zone, MLD = mixed layer depth.S3, S4).
Prymnesiophyceae had a total of four ASVs identified as P. antarctica during TAN1901 and five ASVs in TAN1802.In both cases, two ASVs (ASV0010, ASV0028) were most abundant (a remaining ASV had relative contribution < 0.005% per sample) and showed distinct distributional patterns between genotypes (Fig. 6).
While both ASVs coexisted in the PFZ, ASV0028 was more abundant in the SAZ and ASV0010 increased south of the PF, reaching a maximum in Southeast Shelf-Slope.Among Bacillariophyta, C. inerme ASV0291 and ASV0033 were found in the same locations but with differences in read abundance; ASV0291 peaked in the sACC while ASV0033 showed highest read abundance in the Southeast Shelf-Slope and Cape Adare where it dominated the metabarcode.In the case of Chaetoceros dichaeta, ASV0113 and AS0080 had similar read abundance between south of the sACC and north of the RSR.Pseudo-nitzschia sp.(ASV0057) was more abundant along the central part of the Iselin Bank (CTD15-18) but decreased in the southernmost part of the bank during TA1901.The Dinophyceae Gyrodinium helveticum (ASV0018) was identified in all regions, but with higher reads abundance in Oceanic and Southeast Shelf-Slope subregions of the RSR.

Environmental drivers of community composition variability in the RSR
Redundancy analysis of species composition following PER-MANOVA (Supporting Information Table S8) clustered TAN1802 stations together while TAN1901 had distinct subregional patterns (Fig. 7).Comparison of the overlapping regions between the two voyages showed that stations from the Iselin Bank formed separate clusters in each voyage, whereas Oceanic stations clustered together.Repeat analysis conducted for each voyage separately confirmed this trend with the off-shelf and shelf-slope stations clustering together during TAN1802, whereas analysis of TAN1901 alone showed regional clustering supporting this differentiation between the shelf-slope and offshelf waters (Supporting Information Fig. S3).
Analysis of the physicochemical parameters yielded statistically significant relationships, yet the proportion of community composition variability explained by individual parameters (R 2 ) was low (Supporting Information Table S9).Among the variables analyzed, salinity (PERMANOVA, F = 10.18,p < 0.001) and O 2 anomaly (PERMANOVA, F = 7.92, p < 0.001) emerged as the most relevant environmental parameters during TAN1901 but only explained 9% and 7% of community variability, respectively.TAN1802 showed the highest R 2 values with DRSi being the primary factor explaining 12% of compositional variability (PERMANOVA, F = 7.63, p < 0.001).MLD was statistically significant (p < 0.05) on the two voyage-specific analysis, but with a R 2 value ranked below macronutrients and salinity (Supporting Information Table S9).
The relationship between ASV of Bacillariophyta and Prymnesiophyceae and environmental drivers formed three main clusters within each voyage, with correlations between ASV reads abundance and environmental factors being systematically stronger during TAN1802 relative to TAN1901 (Fig. 8).In both cases the correlation of ASV with MLD was remarkably low.During TAN1802, cluster 1 comprised ASV primarily consisting of Bacillariophyta that showed a positive correlation with macronutrients, salinity, TChl a and temperature, and a negative correlation with Z eu , the absolute contribution of the nano-size fraction of Chl a, NH 4 and O 2 anomaly (Fig. 8A).Conversely, cluster 3, which contained smaller size diatoms (including three ASVs of Fragilariopsis) and two ASVs of Prymnesiophyceae, showed an inverse relationship with these physicochemical variables.
For TAN1901, cluster 1 comprised mostly Bacillariophyta ASVs and two Phaeocystis ASVs (Fig. 8B).This cluster had a positive correlation with salinity, Z eu , NO 3 and DRP and negative correlation with O 2 anomaly, NH 4 , temperature, total and sizefractionated Chl a.As with TAN1802, cluster 3 showed an opposing trend to cluster 1, and contained mostly Prymnesiophyceae including the most abundant ASV of P. antarctica (ASV0010).This ASV dominated CTD26 on the Southeast Shelf-Slope where NH 4 concentration was considerably higher than other regions.Cluster 3 also contained three ASV of Bacillariophyta including Fragilariopsis sp.The smaller cluster 2, containing the ASV0033 of C. inerme and ASV13195 of C. neogracilis, dominated the assemblage at Cape Adare and showed a strong positive correlation with both total and > 20 μm Chl a concentrations.

Spatiotemporal variability across RSRs
The transition from summer to autumn was characterized by a decrease in TChl a and shifts in the dominant taxa.January and February (TAN1901) had higher biomass and encountered diatom-dominated blooms (> 1 mg m À3 surface TChl a) although the survey was conducted after the December peak primary productivity and biomass ascribed to the Ross Sea (Arrigo and McClain 1994).February to March (TAN1802) had lower TChl a levels (< 0.4 mg m À3 ) and dinoflagellates dominating at all stations.These conditions in association to the dominance of microphytoplankton (Fig. 3) are assumed to reflect the seasonal increase of mixotrophs in this region as light availability decreases in autumn (Dennett et al. 2001).The POC : Chl a ratios reflected this trophic transition from a summer autotrophic system (mean AE SD POC : Chl a = 216 AE 78 w : w, TAN1901) to a heterotrophic autumn system (mean AE SD POC : Chl a = 305 AE 100 w : w, TAN1802) in the RSR (Supporting Information Table S10), supporting the progression from a diatom bloom in TAN1901 to postbloom autumn conditions in TAN1802.
Phytoplankton taxonomic composition using metabarcoding of 18S rRNA genes is widely used today across distinct environments (Sow et al. 2020;Trefault et al. 2021;Gutiérrez-Rodríguez et al. 2022).Several well-identified limitations and biases introduced during the metabarcoding technique (e.g., primer of choice, acid nucleic extraction and preservation methods, differences in gene copy numbers among taxa) are acknowledged in the literature with respect to the quantification of marine protists based on ASV (or OTU) relative abundances (Gloor et al. 2017;Santoferrara et al. 2020).New tools and methods are being developed to ameliorate this limitation (Gong and Marchetti 2019;Martin et al. 2022), but so far, this cannot yet be applied to large datasets with diverse communities, particularly in the absence of genomic data for different species.In this regard, the prevalence and high relative contributions of dinoflagellate in this study and other metabarcoding surveys (Liu et al. 2022) would be partially due to the high number of gene copies in this group (Gong et al. 2013).However, the increase in dinoflagellates abundance from summer to autumn in the Ross Sea observed with metabarcoding data is in good agreement with previous reports based on pigments (Park et al. 2021).
During summer, the genus Pseudo-nitzschia dominated the read abundance at the Iselin bank whereas C. inerme did at Cape Adare, in agreement with reported distributions of pennate and centric diatoms in the Ross Sea (Saggiomo et al. 2021).C. inerme dominance at Cape Adare, especially at CTD08 and CTD09, is consistent with the dominance of > 20 μm Chl a fraction (Fig. 3), the positive correlation with TChl a (Fig. 8) and the higher silica uptake in this subregion (Llopis Monferrer et al. 2021).These results further highlight the important regional role of larger diatoms, such as C. inerme (Cornet-Barthaux et al. 2007), in silica cycling (Sicko-Goad et al. 1984).Autumn, on the other hand, was characterized by a significantly higher abundance of Fragilariopsis and Minidiscus genera (Fig. 5).While this may reflect regional or interannual variability in community composition due to, for example, variation in sea ice (Rozema et al. 2017;Biggs et al. 2019), it is also representative of a transition toward diatom-dominated communities with a higher tolerance to low light and longer periods of darkness (Kennedy et al. 2019), and smaller cells with lower iron requirements (de Baar et al. 2005;Petrou et al. 2016).
During TAN1802, the spatial variability between subregions in the RSR was lower relative to TAN1901, likely reflecting the reduced influence of sea ice melting and increased influence of wind mixing and deepening of the surface mixed layer in autumn (Porter et al. 2019), as expected for this seasonal transition (Smith et al. 1996).When comparing subregions sampled on both voyages, the community composition in the Oceanic subregion was more stable through the summer to autumn progression than the shelf-slope (Fig. 7), as shown by the higher number of shared ASV in the Oceanic subregion relative to the Iselin Bank (Supporting Information Fig. S4).The greater temporal stability of the Oceanic subregion relative to the shelf-slope is consistent with the persistent oligotrophic state of the former (El-Sayed et al. 1983).This differentiation is in part due to currents from the Ross Gyre that constrain elevated phytoplankton biomass to the continental shelf (Reddy and Arrigo 2006).Low and medium biomass areas tend to have more diverse phytoplankton community composition relative to blooms (Irigoien et al. 2004).A higher taxonomic diversity infers that a range of different physiological controls are operating, potentially providing resilience to changes in abiotic variables such as temperature, light, and nutrient concentrations.Consequently, higher diversity may provide a buffer to changing environmental conditions (Hong et al. 2022;Martiny et al. 2022), with inference that the two regions may have different levels of resilience to future change.

Phytoplankton response to environmental drivers
The temporal variations in community composition observed with metabarcoding across subregions in the RSR may be expected to correlate with changes in MLD, as this is closely related to irradiance exposure and hence to photophysiological differences between diatoms and P. antarctica (Arrigo et al. 2010).Despite significant differences in MLD between the two voyages, correlation and PERMANOVA analysis indicated that the direct influence of MLD on community composition was secondary to that of salinity during summer, and macronutrients during autumn.This disparity with the conventional view (Arrigo et al. 1998) may reflect the different seasonal and geographical coverage of the present study relative to the previous focus on the spring-summer season and the south-central polynya, which is characterized by deeper MLD and P. antarctica dominance (Arrigo et al. 1999).The less prominent role of the MLD observed here may also reflect greater taxonomic insight from ASV-based analysis illuminating trends undetected by approaches with coarser taxonomic resolution.For instance, the analysis of ASV belonging to Bacillariophyta and Prymnesiophyceae (Fig. 8) identified that, despite most species of each class clustering together, some Prymnesiophyceae clustered with diatoms.This agrees with recent reports of a similar iron retention strategy in Southern Ocean diatoms and Phaeocystis (Alderkamp et al. 2019), suggesting a preference for similar environmental conditions in certain genotypes despite their affiliation to different phytoplankton classes and "functional" groups.
Here, Prymnesiophyceae were present on both voyages but they were more abundant during summer on TAN1901 (Supporting Information Fig. S5) and ranked behind diatoms and dinoflagellates, in agreement with the seasonal decline in abundance from January onwards (Smith et al. 2003).However, there was one notable exception during summer where Prymnesiophyceae dominated on the Southeast Shelf-Slope (CTD26), despite environmental conditions that differed from their documented preference for deep mixed layers (Arrigo et al. 1998).This station had a shallow MLD of 12 m, and dominance of picophytoplankton (Fig. 2).Although P. antarctica is generally in the nano-size fraction (3.5-6.35μm) (Zingone et al. 2011), it has been reported that single cells can pass through a 2 μm filter and therefore be found in the pico-size fraction (Vaulot et al. 2008;Wright et al. 2009).
This suggests a single-celled mode of P. antarctica at this station.In addition, surface TChl a at this station was below the threshold (< 1 mg Chl a m À3 ) commonly associated with P. antarctica spring blooms, suggesting a decaying stage of a late colonial bloom at CTD26.Summer blooms of P. antarctica in the Ross Sea have been observed under stratified conditions (Fragoso and Smith 2012;Mangoni et al. 2019) and our results suggest that the trait diversity that exists among genotypes might explain these different blooming conditions.
Trait diversity in P. antarctica was apparent in the distribution patterns of two ASV across oceanographic fronts (Fig. 6), which indicated differing tolerance to environmental conditions between genotypes.In this case, the southern boundary for ASV0028 was the PF, whereas ASV0010 was present from the subtropical front to south of the RSR, consistent with previous observations identifying the PF as a transition zone where environmental conditions are optimal for both genotypes (Sow et al. 2020).This illustrates that there are spectrums of traits present within a single phytoplankton species, enabling certain genotypes to dominate a broad latitude range while others are confined to specific regions (Luxem et al. 2017).However, controlled experiments with isolates representing these ASVs would be necessary to enlighten their environmental requirements.A wide trait spectrum can also lead to differing patterns of ecotype tolerance to environmental conditions, allowing certain strains to bloom under conditions that are not previously ascribed for them.This emphasizes the value of metabarcoding techniques and curated databases to assign taxonomy that reflect ASV trends (Guillou et al. 2012;del Campo et al. 2018;Santoferrara et al. 2020) and also for further experiments that include a more comprehensive repertoire of species ecotypes.

Conclusions
Phytoplankton community composition transitioned from an autotrophic diatom-dominated community in summer to a heterotrophic or mixotrophic dinoflagellate-dominated community in autumn.Phytoplankton community composition in the less-studied oceanic subregion of the RSR was more stable through the transition from summer to autumn than in the shelf-slope area, with persistent presence of Dinophyceae but also lower biomass indicative of lower productivity.Establishing the seasonal variability and stability of community composition across taxonomic ranks (e.g., from class to genotype level) and within subregions of the RSR is then required for more robust modeling of food web processes and biogeochemical cycles in this critical region of the Southern Ocean.
The distinct ASV distributional patterns and relationship with environmental variability suggest the incorporation of finer taxonomic description may reduce the uncertainty associated with future community response to climate change by recognizing the flexibility of the boundaries associated with single species traits.While it is not clear yet how this taxonomic resolution can be applied to earth system models (EMS), expanding the characterization of phytoplankton diversity and concomitant environmental variability in less explored areas of the Ross Sea would refine ecological observations that currently rely on simplified frameworks based on functional groups.For instance, including Prymnesiophyceae and Dinophyceae in EMS would improve regional resolution.This approach would then enable more accurate prediction of ecosystem response and subsequent implications for trophic flows and biogeochemical cycles.Zingone, A., G. Forlani, I. Percopo, and M. Montresor. 2011

Fig. 1 .
Fig. 1.Location of CTD stations.Stations with labels in red correspond to TAN1802, with red dots representing stations with DNA samples and black triangles stations without DNA samples.Stations with labels in black with blue dots correspond to TAN1901, all including DNA samples.Dashed lines represent the approximate locations of the subtropical front (STF), sub-Antarctic front (SAF), polar front (PF), and southern boundary of the Antarctic Circumpolar Current (sbACC).Zones delimited by the dashed lines are in bold and represent the SAZ, PFZ, sACC, and RSR.

Fig. 2 .
Fig. 2. Latitudinal variation of temperature, MLD, NO 3 , and DRSi.Values correspond to the surface (25 m for CTD01 and 02 from TAN1802, and 10 m for all other stations).Dashed lines mark the separation between oceanographic zones corresponding to: SAZ, PFZ, sACC, and RSR.Data includes all CTD casts on each voyage regardless of DNA sample collection.Stations with anomalous nutrient concentrations-CTD26 and CTD27-also identified.

Fig. 3 .
Fig. 3. Chl a biomass and size structure.(A) TAN1802, (B) TAN1901.Bars represent proportional (%) size-fractionated Chl a of pico-(0.2-2μm, dark gray), nano-(2-20 μm, gray with dashed lines) and micro-phytoplankton (> 20 μm white dashed lines).Integrated Chl a (Chl a int , from 0 to 100 m, pink circles) and average TChl a in the surface mixed layer (black triangles) are shown for each station.Stations (bottom axis) are organized by latitude and region (top axis).The number on top of dashed lines indicates the approximate latitude.

Fig. 4 .
Fig. 4. Phytoplankton community taxonomic composition for the 12 main classes at all stations and depths (x-axis) for (A) TAN1802 (10-200 m) and (B) TAN1901 (10-100 m).Samples are organized by latitude and region and samples collected below the euphotic zone are shown in red.Taxa classify as "other" correspond to classes with an abundance lower than 4%.

Fig. 5 .
Fig. 5. Phytoplankton genera in the euphotic zone with statistically significant differences in standardized read abundance between voyages ( p < 0.05) determined by DESeq2 analysis.(A) For all stations, (B) within the RSR.Negative log2 fold change indicates genera with a statistically significantly higher abundance during TAN1802 (autumn) and a positive log2 fold change denotes genera with statistically significantly higher abundance during TAN1901 (summer).

Fig. 6 .
Fig. 6.Spatial distribution of main ASV by CTD (organized by latitude).Abundance is represented in number of reads per sample (normalized by the mean sequencing depth) considering only samples within the euphotic zone with a relative abundance of at least 30% on at least one sample.Oceanographic zones, regions, and latitude (degrees S) are indicated above the heatmap.Blue dots represent TAN1901 and red dots TAN1802.CTDs colors specify the RSR: Oceanic (light blue), Iselin Bank (dark blue), Cape Adare (orange) and Southeast Shelf (green).

Fig. 7 .
Fig. 7. Redundancy analysis at species level in the euphotic zone for the RSR.The color specifies the voyage: TAN1802 (red) and TAN1901 (blue).

Fig. 8 .
Fig. 8. Correlation analysis for selected environmental and biogeochemical drivers in relation to ASV abundance from the class Bacillariophyta and Prymnesiophyceae for (A) TAN1802 and (B) TAN1901.Species name in black corresponds to Bacillariophyta while species in orange to Prymnesiophyceae.ASV marked in bold correspond to ASV present in both analyses.The figure shows the 30 ASV with higher correlation scores regardless of their abundance (Supporting Information TablesS3, S4).