Microbial Eukaryotes in Natural and Artificial Salt Marsh Pools

: Microscopic eukaryotes are important components of coastal wetland ecosystems. The goal of this study was to investigate the diversity of microeukaryotes in the tidal pools of a New Jersey salt marsh and to compare the assemblages of natural and artificial pools excavated for controlling mosquito populations. We evaluated microeukaryotic assemblages using the amplicon sequencing of 18S and rbc L DNA markers and the microscopic identification of diatoms in water and sediment samples. 18S unique amplicon sequence variants (ASV) representing ciliates, dinoflagellates, diatoms, and cercozoans were the most diverse, while the reads of dinoflagellates, diatoms, ciliates, and nematodes were the most abundant. The dominant ASVs were attributed to organisms that are characteristic of coastal plankton and sediments or those known for their resistance to salinity, desiccation, hypoxia, and UV stress. The sediment assemblages were more diverse compared to those from the water column and contained a larger portion of ASVs that were not assigned to any low-rank taxa, reflecting the current gaps in understanding the diversity of microeukaryotes. Most taxonomic groups were significantly different in their abundance and composition between natural and artificial pools. Dinoflagellates, haptophytes, chrysophytes, pelagophytes, and raphidophytes—the groups that include a large proportion of mixotrophic taxa and species known for forming harmful algal blooms—were more abundant in the artificial than in the natural pools. Fungi, labyrinthulomycetes, and peronosporomycetes were also more abundant in artificial pools, which may be related to organic matter enrichment. Diatoms and foraminifera showed an opposite trend of higher abundance in natural pools.


Introduction
Tidal pools are distinct habitats within coastal salt marshes that are critically important for marsh biodiversity, biogeochemistry, and stability [1].Natural pools are integral components of marsh food webs and serve as shelters, nurseries, and feeding grounds for birds, fish, and aquatic invertebrates [2][3][4].In addition to plants and macroscopic animals, pools and other salt marsh habitats harbor a much higher, but understudied, diversity of microorganisms that are no less important for ecosystem functioning [5].Among salt marsh microbes, bacteria and fungi have been studied most extensively considering their key role in decomposition and nutrient cycling [6][7][8][9].Both the structure and function of their assemblages have been shown to change in response to nutrient enrichment [10][11][12][13], chemical pollution [14], biological invasions [15,16], sea level rise [17], and marsh restoration [18,19].
Organisms inhabiting salt marsh pools are expected to be mostly recruited from tidal waters and from the surrounding marsh soils.The environment of the pools is harsh, and their biota must be uniquely adapted to these unfavorable conditions.High levels of ultraviolet radiation are characteristic for pools as with any other intertidal habitats [53].Small water volume and a lack of shading cause high rates of evaporation in tidal pools, leading to fast and pronounced salinity and temperature changes [54,55].Microbial mats covering sediment surfaces may cause severe diurnal fluctuations in oxygen [1,56].Desiccation affects tidal pools differentially depending on the frequency of their inundation; therefore, marsh elevation gradient and hydrologic modifications are expected to influence their biota [57].
Many salt marshes of the North American east coast have been severely altered for the purposes of mosquito control, first by the construction of parallel ditches in the first half of the 20th century, and then by the excavation of artificial pools connected by ditches, known as open marsh water management (OMWM) [58].In the New Jersey coastal marshes, approximately 7000 OMWMs were excavated between 1970 and 2010, and this activity was shown to negatively influence a marsh's ability to sequester carbon [59].While these drastic hydrological modifications visibly affect marsh animals and vegetation [57,[60][61][62], less is known about their role in restructuring microbial assemblages.
The goal of our study was to assess the diversity of microbial eukaryotes in tidal pools located in a salt marsh in southern New Jersey, USA using a DNA metabarcoding approach combined with the microscopic evaluation of diatom assemblages.We also aimed to compare assemblages of natural pools and OMWMs in terms of their diversity and taxonomic composition.

Sampling
Ten pools located in the Great Bay Boulevard Wildlife Management Area, a protected salt marsh near Tuckerton, New Jersey, were sampled on 6 October 2018 (Figure 1, Table S1).This marsh is dominated by cordgrass Spartina alterniflora, with the tall form growing in low-elevation and the short form in higher-elevation areas.Six pools are natural, and four are artificial (OMWMs).The pools have an area of ~50-400 m 2 and are 20-60 cm deep.In each pool, three sediment samples and one surface water sample were collected for a total of 40 samples.Sediment samples were collected near shore using turkey basters, at three equidistant locations in each pool, and placed in sterile whirl-pak bags.Water samples were collected by directly immersing 1 L sterile carboys beneath the water surface to a depth of 0.1 m.Samples were transported to the laboratory in a cooler on ice within 4 h of collection.The water samples were fractionated by filtering through 8 µm and 0.2 µm polycarbonate filters.The samples were stored at −80 • C until processing.Water salinity, pH, dissolved oxygen concentration, and temperature were measured in the field with a YSI ProPlus multiparameter meter, YSI Inc., Yellow Springs, OH, USA (Tables 1 and S1).

Metabarcoding
Genomic DNA was extracted with a Takara NucleoSpin Soil DNA extraction kit (Takara Bio Inc., Shiga, Japan) following the manufacturer's instructions.Five sediment samples were extracted twice to estimate variability among extraction replicates, resulting in total of 35 DNA extracts from sediment samples.The 0.2 µm and 8 µm fractions of water samples were extracted separately.DNA yield was quantified with Qubit 3.0 (Invitrogen, Life Technologies, Grand Island, New York, NY, USA), and one extraction (0.2 µm fraction of water sample from the pool "AR") with very low DNA yield was discarded.The total number of samples used for metabarcoding was 54.
Two markers were used for metabarcoding, the 96-134 base pair (bp) V9 region of the 18S rRNA gene (18S_V9) and a 312 bp fragment of the rbcL plastid gene most often used for diatom metabarcoding [63][64][65][66].The 18S_V9 region was amplified using the "paneukaryotic" primers 1391F and EukB, widely used in protist metabarcoding [46,67].18S_V9 libraries were constructed, and sequencing was conducted at the University of Minnesota Genomic Center (UMGC) following protocols by Gohl et al. [68].The rbcL region was amplified with KAPA HiFi HotStart ReadyMix (Kapa Biosystems, Wilmington, MA, USA) and an equimolar mix of the forward primers Diat_rbcL_708F_1, 708F_2, and 708F_3 and the reverse primers R3_1 and R3_2 [66] modified according to the Illumina protocol by adding universal Illumina tails.
The three replicates of each sample were pooled, purified using the AMPure XP Beads (Agencourt Bioscience Corp., Beverly, MA, USA) according to manufacturer's instructions, and sent for the last steps of library construction and sequencing to UMGC.The amplicon libraries were sequenced on an Illumina MiSeq platform using the V2

Metabarcoding
Genomic DNA was extracted with a Takara NucleoSpin Soil DNA extraction kit (Takara Bio Inc., Shiga, Japan) following the manufacturer's instructions.Five sediment samples were extracted twice to estimate variability among extraction replicates, resulting in total of 35 DNA extracts from sediment samples.The 0.2 µm and 8 µm fractions of water samples were extracted separately.DNA yield was quantified with Qubit 3.0 (Invitrogen, Life Technologies, Grand Island, New York, NY, USA), and one extraction (0.2 µm fraction of water sample from the pool "AR") with very low DNA yield was discarded.The total number of samples used for metabarcoding was 54.
Two markers were used for metabarcoding, the 96-134 base pair (bp) V9 region of the 18S rRNA gene (18S_V9) and a 312 bp fragment of the rbcL plastid gene most often used for diatom metabarcoding [63][64][65][66].The 18S_V9 region was amplified using the "paneukaryotic" primers 1391F and EukB, widely used in protist metabarcoding [46,67].18S_V9 libraries were constructed, and sequencing was conducted at the University of Minnesota Genomic Center (UMGC) following protocols by Gohl et al. [68].The rbcL region was amplified with KAPA HiFi HotStart ReadyMix (Kapa Biosystems, Wilmington, MA, USA) and an equimolar mix of the forward primers Diat_rbcL_708F_1, 708F_2, and 708F_3 and the reverse primers R3_1 and R3_2 [66] modified according to the Illumina protocol by adding universal Illumina tails.
The three replicates of each sample were pooled, purified using the AMPure XP Beads (Agencourt Bioscience Corp., Beverly, MA, USA) according to manufacturer's instructions, and sent for the last steps of library construction and sequencing to UMGC.The amplicon libraries were sequenced on an Illumina MiSeq platform using the V2 paired-end sequencing kit (2 × 300 bp).Demultiplexed raw Illumina MiSeq reads were archived in the National Center for Biotechnology Information (NCBI) Sequence Read Archive (SRA) under bioproject number PRJNA1035775.

Bioinformatics Processing and Taxonomic Assignment
Illumina paired-end reads were processed using the dada2 package version 1.18.0 in R version 4.3.1.[69].Primer sequences were removed with cutadapt version 2.8 [70] using the default parameters (maximum error rate = 10%) and the -g flag, which removes any base upstream of the primers.Read quality was visualized with the plotQualityProfile function.Reads were filtered using the filterAndTrim function, adapting parameters (truncLen, minLen, truncQ, maxEE) according to overall sequence quality.Merging of the forward and reverse reads was carried out with the mergePairs function using the default parameters (minOverlap = 12, maxMismatch = 0).Chimeras were removed using remove-BimeraDenovo with default parameters.Parameters in the dada2 script used for processing rbcL reads were those recommended for diatom metabarcoding and available on Github (https://github.com/fkeck/DADA2_diatoms_pipeline,accessed on 1 October 2023).

Diatom Enumeration
Sediment subsamples were treated with nitric acid, followed by six rinses with distilled water to remove organic matter.Permanent diatom slides were prepared with Naphrax ® mounting medium and examined under an AxioImager A1 (Zeiss, Oberkochen, Germany) light microscope equipped with differential contrast optics and oil immersion at 100× objective.At least 400 diatom valves were identified and counted in each sample using several identification resources [74][75][76][77][78].

Data Analysis
All data manipulations and numerical analyses were conducted in the R environment using packages vegan version 2.6-4 [79], ALDEx2 version 1.32.0 [80,81], indicspecies version 1.7.14 [82], and packages commonly employed for data handling and visualization.
To account for variability in sequencing depth, the raw ASV numbers were rarefied to the smallest number of ASVs per sample in each data subset, namely sediment or water (rrarefy function in vegan).A Welch t-test was used to test for significant difference in mean read abundances of major taxonomic groups in natural and artificial pools.Permutational multivariate analysis of variance (PERMANOVA) with Bray-Curtis dissimilarity (adonis2 function in vegan) was employed to test for differences in assemblage composition [83] using several datasets: all microeukaryotic 18S_V9 metabarcoding data, subsets of 18S data corresponding to major taxonomic groups, rbcL diatom metabarcoding data, and diatom count data.Non-metric multidimensional Scaling (NMDS) with Hellinger-transformed data and a Bray-Curtis distance matrix was used for visualizing differences in assemblage composition (metaMDS function in vegan).The Envfit procedure in vegan was employed to explore correlations between NMDS axes and measured environmental characteristics.
To identify individual ASVs with significantly different abundances in natural and artificial pools, we used the ANOVA-like differential expression tool for high throughput sequencing data implemented in the ALDEx2 Bioconductor R package [80,81], considered as one of the most conservative and reliable tools currently available for differential abundance analysis with high throughput sequencing data [84].ASVs significantly different in their abundance were identified based on Benjamini-Hochberg corrected p-values of the Welch t-test statistic.To determine which morphologically defined diatom species were characteristic of natural or artificial pools, an indicator species analysis [82] was carried out with the indicspecies package.

18S_V9 Metabarcoding
The 18S_V9 sequencing of 54 water and sediment samples yielded a total of 6,123,792 raw reads.A total of 3,604,113 reads remained after quality filtering, merging, denoising, and chimera removal.The total number of ASVs in all 54 samples was 18,455.The removal of ASVs that were not taxonomically assigned to Eukaryota with both reference databases, not assigned to any eukaryotic phylum, or representing groups of predominantly macroscopic organisms, resulted in a dataset of 1,642,367 reads and 4215 ASVs.The rarefaction curves demonstrate a greater number of reads and ASV diversity in sediment compared to water samples, with richness approaching saturation in all samples (Figure 2).
Coasts 2024, 3, FOR PEER REVIEW 5 sequencing data implemented in the ALDEx2 Bioconductor R package [80,81], considered as one of the most conservative and reliable tools currently available for differential abundance analysis with high throughput sequencing data [84].ASVs significantly different in their abundance were identified based on Benjamini-Hochberg corrected p-values of the Welch t-test statistic.To determine which morphologically defined diatom species were characteristic of natural or artificial pools, an indicator species analysis [82] was carried out with the indicspecies package.

18S_V9 Metabarcoding
The 18S_V9 sequencing of 54 water and sediment samples yielded a total of 6,123,792 raw reads.A total of 3,604,113 reads remained after quality filtering, merging, denoising, and chimera removal.The total number of ASVs in all 54 samples was 18,455.The removal of ASVs that were not taxonomically assigned to Eukaryota with both reference databases, not assigned to any eukaryotic phylum, or representing groups of predominantly macroscopic organisms, resulted in a dataset of 1,642,367 reads and 4215 ASVs.The rarefaction curves demonstrate a greater number of reads and ASV diversity in sediment compared to water samples, with richness approaching saturation in all samples (Figure 2).The majority of the ASVs represented ciliates (578), dinoflagellates (427), diatoms (321), and cercozoans (321).A large portion of ASVs could not be assigned to any class within the TSAR (Telonemia + Stramenopila + Alveolata + Rhizaria) phylum (497), the phylum Excavata (85), and below the class Alveolata (145).The proportions of ASVs representing major taxonomic groups were not very different between water and sediment samples, with diatom and dinoflagellate ASVs being slightly more diverse in water, and ciliates in sediment (Figure 3, left panel).Sediment samples had proportionally more reads that belonged to diatoms, nematodes, platyhelmintes, and raphidophytes and fewer green algae (Chlorophyta) reads compared to water samples (Figure 3, right panel).The majority of the ASVs represented ciliates (578), dinoflagellates (427), diatoms (321), and cercozoans (321).A large portion of ASVs could not be assigned to any class within the TSAR (Telonemia + Stramenopila + Alveolata + Rhizaria) phylum (497), the phylum Excavata (85), and below the class Alveolata (145).The proportions of ASVs representing major taxonomic groups were not very different between water and sediment samples, with diatom and dinoflagellate ASVs being slightly more diverse in water, and ciliates in sediment (Figure 3, left panel).Sediment samples had proportionally more reads that belonged to diatoms, nematodes, platyhelmintes, and raphidophytes and fewer green algae (Chlorophyta) reads compared to water samples (Figure 3, right panel).
Dinoflagellates were represented by three classes: Dinophyceae (74% of ASVs, 94% of reads), Syndiniales (20% of ASVs, 2.6% of reads), and Oxyrrhea (0.7% of ASVs, 2% of reads).The most abundant genera of dinoflagellates were Durinskia (13.7% of all dinoflagellate reads), Levanderina (6.8%), Kryptoperidinium (4.8%), and Alexandrium (4.0%).The first three genera were each represented by a single species in each, namely Durinskia dybowskii, Kryptoperidinium foliaceum, and Levanderina fissa, and these were the most abundant protistan ASVs identified to the species level in the entire 18S dataset.Alexandrium spp.were proportionally more abundant in sediments, while D. dybowskii and K. foliaceum had the highest contribution in the 8 µm water fraction (Figure 4, left panel).Both D. dybowskii and L. fissa were present in all pools except one located closest to the shore (natural pool "SB"), while K. foliaceum was found only in natural pools.The 0.2 µm water fraction was dominated by reads identified as Oxyrrhis marina.Dinoflagellates were represented by three classes: Dinophyceae (74% of ASVs, 94% of reads), Syndiniales (20% of ASVs, 2.6% of reads), and Oxyrrhea (0.7% of ASVs, 2% of reads).The most abundant genera of dinoflagellates were Durinskia (13.7% of all dinoflagellate reads), Levanderina (6.8%), Kryptoperidinium (4.8%), and Alexandrium (4.0%).The first three genera were each represented by a single species in each, namely Durinskia dybowskii, Kryptoperidinium foliaceum, and Levanderina fissa, and these were the most abundant protistan ASVs identified to the species level in the entire 18S dataset.Alexandrium spp.were proportionally more abundant in sediments, while D. dybowskii and K. foliaceum had the highest contribution in the 8 µm water fraction (Figure 4, left panel).Both D. dybowskii and L. fissa were present in all pools except one located closest to the shore (natural pool "SB"), while K. foliaceum was found only in natural pools.The 0.2 µm water fraction was dominated by reads identified as Oxyrrhis marina.Among ciliates, most reads were attributed to the genera Fabrea (4.1%), Strombidium (3.9%), Frontonia (3.8%), and Mesodinium (2.5%), with Fabrea, Frontonia, and Strombidium being the most abundant in sediments and Mesodinium and Zoothamnion in water (Figure 4, right panel).Besides these genera, in the 8 µm water fraction, reads assigned to the genus Strombidium were relatively abundant, while in the 0.2 µm fraction, Chlamydodon and Ephelota had higher proportions.The most abundant ASVs identified to the species level belonged to Fabrea salina, Strombidium guangdongense, and Mesodinium rubrum.The proportion of dinoflagellate and ciliate reads not assigned to any genus was higher  Dinoflagellates were represented by three classes: Dinophyceae (74% of ASVs, 94% of reads), Syndiniales (20% of ASVs, 2.6% of reads), and Oxyrrhea (0.7% of ASVs, 2% of reads).The most abundant genera of dinoflagellates were Durinskia (13.7% of all dinoflagellate reads), Levanderina (6.8%), Kryptoperidinium (4.8%), and Alexandrium (4.0%).The first three genera were each represented by a single species in each, namely Durinskia dybowskii, Kryptoperidinium foliaceum, and Levanderina fissa, and these were the most abundant protistan ASVs identified to the species level in the entire 18S dataset.Alexandrium spp.were proportionally more abundant in sediments, while D. dybowskii and K. foliaceum had the highest contribution in the 8 µm water fraction (Figure 4, left panel).Both D. dybowskii and L. fissa were present in all pools except one located closest to the shore (natural pool "SB"), while K. foliaceum was found only in natural pools.The 0.2 µm water fraction was dominated by reads identified as Oxyrrhis marina.Among ciliates, most reads were attributed to the genera Fabrea (4.1%), Strombidium (3.9%), Frontonia (3.8%), and Mesodinium (2.5%), with Fabrea, Frontonia, and Strombidium being the most abundant in sediments and Mesodinium and Zoothamnion in water (Figure 4, right panel).Besides these genera, in the 8 µm water fraction, reads assigned to the genus Strombidium were relatively abundant, while in the 0.2 µm fraction, Chlamydodon and Ephelota had higher proportions.The most abundant ASVs identified to the species level belonged to Fabrea salina, Strombidium guangdongense, and Mesodinium rubrum.The proportion of dinoflagellate and ciliate reads not assigned to any genus was higher Among ciliates, most reads were attributed to the genera Fabrea (4.1%), Strombidium (3.9%), Frontonia (3.8%), and Mesodinium (2.5%), with Fabrea, Frontonia, and Strombidium being the most abundant in sediments and Mesodinium and Zoothamnion in water (Figure 4, right panel).Besides these genera, in the 8 µm water fraction, reads assigned to the genus Strombidium were relatively abundant, while in the 0.2 µm fraction, Chlamydodon and Ephelota had higher proportions.The most abundant ASVs identified to the species level belonged to Fabrea salina, Strombidium guangdongense, and Mesodinium rubrum.The proportion of dinoflagellate and ciliate reads not assigned to any genus was higher (approximately 2/3 of all reads) in sediment samples but also substantial (approximately 1/3 of all reads) in water (Figure 4).
Within Cercozoa, the relatively highest proportions of reads belonged to the "Protaspa lineage", Ebria tripartita, and Placopus pusillus, a parasite of Tetraselmis which was found to be most abundant among 18S green algal reads.The most abundant green algal ASVs identified to the species level belonged to Tetraselmis marina, Ostreococcus mediterraneus, O. tauri, and Pyramimonas disomata.The most common euglenid ASVs assigned lowrank taxonomy were Notosolenus urceolatus, N. ostium, Eutreptiella sp., and Rapaza viridis.Among other Euglenozoa classified at least to the genus level, the most abundant were kinetoplastids Neobodo sp., Rhynchomonas nasuta, and Klosteria bodomorphis and diplonemids Rhynchopus serpens, Hemistasia phaeocysticola, and Diplonema aggregatum.
Only 25% of all foraminiferan ASVs were taxonomically assigned at the genus level, and at the class level, 22% of ASVs were assigned to Monothalamids, 16% to Globothalamea, and 6% to Tubothalamea.Within fungi, most reads identified to the class level belonged to Ascomycota (35%), followed by Basidiomycota (8%), while other fungal classes did not reach 1% of relative abundance and 55% of all reads were not assigned to any class.

Diatom rbcL Metabarcoding
RbcL sequencing yielded a total of 2,520,050 raw reads.A total of 1,569,601 reads remained after quality filtering, merging, denoising and chimera removal.Out of 2900 ASVs, 2434 were taxonomically assigned at least to the phylum level with 2404 ASVs classified as diatoms (Bacillariophyta), 29 as other ochrophytes (including 11 Chrysophyceae, five Raphidophyceae and one Xanthophyceae), and one as a red alga (Rhodophyta).A total of 1240 diatom ASVs were unique to benthic samples, 784 were unique to water, and 380 were shared among both habitats.

Diatom Counts
A total of 229 diatom morphotaxa were found in 30 sediment samples with 37-69 species per sample (Table S2).The most common and abundant were several species of the genera Navicula (N.salinicola, N. consentanea, N. johnssonii, N. hamiltonii, and N. halinae) and Halamphora (H.adumbratoides, H. tenerrima, and H. subtropica), as well as Serratifera spp., Cocconeis neothumensis var.marina, Nitzschia frustulum, and Opephora spp.A total of 66 morphotaxa did not fully fit descriptions of known species or represented complexes of similar species that were impossible to separate with light microscopy either because of their small size or insufficient information about species delimitation.

Assemblage Composition
Several taxonomic groups identified by 18S metabarcoding showed significant differences in read abundances between natural and artificial pools (t-test, Table 2).These differences were most pronounced in sediment samples where diatoms, euglenoids, foraminifera, and nematodes were more abundant in natural pools, while several groups of mixotrophic flagellates, such as dinoflagellates, pelagophytes, raphidophytes, and some other groups such as green algae, rotifers, fungi, and fungi-like stramenopiles (peronosporomycetes and labyrinthulomycetes) were proportionally more abundant in OMWMs (Figure 6).Ciliates were more abundant in natural pools, although this trend was marginally significant (Table 2).In water samples, the same main trends were observed, although the differences were rarely statistically significant because of the lower number of samples (Table 2).The NMDS plot of sediment samples (Figure 7, left panel) shows a strong separation of samples from natural and artificial pools along the first NMDS axis, which also Differences in the composition of the whole microeukaryote assemblage between natural and artificial pools were likewise most pronounced in sediment and similar, but often non-significant, in water samples (PERMANOVA, Table 2).The difference in microeukaryotic assemblage composition between natural and artificial pools was mostly driven by dinoflagellates, cercozoans, diatoms, and apicomplexans in sediment and by green algae, diatoms, and cercozoans in water (PERMANOVA, Table 2).The NMDS plot of sediment samples (Figure 7, left panel) shows a strong separation of samples from natural and artificial pools along the first NMDS axis, which also corresponds to a gradient of water salinity, pH, and turbidity.Samples from the same pools, shown as circles connected by "spider" lines in Figure 7, tend to cluster together, thus showing a greater similarity of assemblages within pools as compared to among-pool similarity.The four samples positioned on the left of the NMDS plot are from a natural pool located closest to the open water of the Barnegat Bay with a sandy bottom and experiencing the most flushing.This indicates that the main gradient of microeukaryotic assemblage variation may also be related to substrate texture and other unmeasured environmental characteristics related to pool position on the marsh.The second NMDS axis appears to be underlain by the water depth gradient and associated characteristics such as temperature (higher in shallow pools) and dissolved oxygen (higher in deeper pools).
Diatom assemblages characterized by rbcL metabarcoding were also found to be significantly different in their composition (Table 2, Figure 7, right panel), while diatom count data revealed only a marginally significant difference between natural and artificial pools (Table 2, Figure S1).
ural pool located closest to the open water of the Barnegat Bay with a sandy bottom and experiencing the most flushing.This indicates that the main gradient of microeukaryotic assemblage variation may also be related to substrate texture and other unmeasured environmental characteristics related to pool position on the marsh.The second NMDS axis appears to be underlain by the water depth gradient and associated characteristics such as temperature (higher in shallow pools) and dissolved oxygen (higher in deeper pools).Diatom assemblages characterized by rbcL metabarcoding were also found to be significantly different in their composition (Table 2, Figure 7, right panel), while diatom count data revealed only a marginally significant difference between natural and artificial pools (Table 2, Figure S1).
In water samples, only the dinoflagellate Kryptoperidinium foliaceum was higher in abundance in natural pools as compared to OMWMs, while only one green algal ASV classified as "Chlorellales_X" had a higher abundance in OMWMs.
Thirteen diatom rbcL ASVs had significantly higher abundance in sediments of the natural pools.Among these, four AVSs were assigned taxonomy at the species level: Minidiscus trioculatus, Lyrella hennedyi, Nitzschia draveillensis, and Thalassiosira minima.Twenty-four ASVs were more abundant in OMWMs, and six of them were classified at species level: Chaetoceros muelleri, Haslea howeana, Cylindrotheca closterium, Halamphora
In water samples, only the dinoflagellate Kryptoperidinium foliaceum was higher in abundance in natural pools as compared to OMWMs, while only one green algal ASV classified as "Chlorellales_X" had a higher abundance in OMWMs.
Thirteen diatom rbcL ASVs had significantly higher abundance in sediments of the natural pools.Among these, four AVSs were assigned taxonomy at the species level: Minidiscus trioculatus, Lyrella hennedyi, Nitzschia draveillensis, and Thalassiosira minima.Twenty-four ASVs were more abundant in OMWMs, and six of them were classified at species level: Chaetoceros muelleri, Haslea howeana, Cylindrotheca closterium, Halamphora foramina, Serratifera andersonii, and Nitzschia laevis.In water samples, no diatom rbcL ASVs were significantly more abundant in one type of pool compared to another if judging by Benjamini-Hochberg corrected p-values, but some of the same ASVs that differentiated pool types in the sediment dataset showed the same trends in water samples based on uncorrected p-values.For example, Minidiscus trioculatus was found to be more abundant in natural samples in comparison to OMWMs, and Chaetoceros muelleri was more abundant in OMWMs.

Assemblage Composition
Our metabarcoding results show that the dominant microscopic eukaryotes in studied pools were dinoflagellates, diatoms, ciliates, green algae, cercozoans, and nematodes, the groups commonly reported as dominant in metabarcoding surveys of marine plankton [40,43,44,85,86], deep sediments [41,42], and intertidal mudflats [48].A large proportion of ASVs recovered in this survey were not attributed to low-rank taxa, which is characteristic of barcoding and metabarcoding surveys regardless of used DNA markers or target taxonomic groups [40,45].Forster et al. [42] found that about 70% of planktonic and 33% of benthic RNA sequences recovered in the metabarcoding of marine coastal environments could be assigned taxonomy using contemporary reference databases.This ratio reflects the existing knowledge gaps of the taxonomic diversity of benthic organisms, especially protists, and is practically the same as what we found in the water and sediment samples of the tidal pools.The ASV diversity in studied pools was higher in sediment than water, which is also concurrent with findings by Forster et al. [42], who explained this phenomenon by the greater heterogeneity of sediment habitats.
Ciliate ASV richness was the highest among all major taxonomic groups in the studied pools.The most abundant reads were attributed to ciliates known from both coastal plankton and intertidal or inland habitats with high and fluctuating salinity.A characteristic representative of the latter group is Fabrea salina, a euryhaline heterotrich notorious for its ability to withstand high salinity and ultraviolet radiation [87][88][89].Several other genera abundant in studied pools, such as Strombidium, Chlamydodon, and Frontonia, are common in salt marshes [33,90] and coastal plankton [91], and include species known for their tolerance of low oxygen levels [92].In particular, Strombidium purpureum, an anaerobic species harboring symbiotic purple bacterium [93], was detected in two artificial pools with relatively low oxygen levels (1.9 and 3.5 mg L −1 ).The largest proportion of ciliate reads in the water samples of studied pools were attributed to the photosynthetic Mesodinium rubrum, a cosmopolitan and occasionally bloom-forming species common in coastal marine plankton including estuaries of the mid-Atlantic region of USA [94,95].The presence of amplicons assigned to sessile ciliate genera Zoothamnion and Ephelota in water samples requires further study.Snyder et al. [96] found an unexpected number of non-planktonic ciliates in their survey of microzooplankton in the Gulf of Mexico and explained this by the association of ciliates with floating algae.The high proportion of reads of sessile ciliates in 0.2 µm water fraction in our study is probably due to the capture of extracellular DNA.Zoothamnion DNA, for example, was found in copepod guts [97] and, therefore, may appear in water after excretion.
Dinoflagellate reads were the most abundant in the entire 18S dataset and several dinoflagellate ASVs, all from the class Dinophyceae, were dominant in water, sediments, or both.Their ASV richness was also high, which is often the case in metabarcoding surveys of marine waters, including low-oxygen areas [46].Dinoflagellates are consistently reported as having a relatively high abundance of reads in metabarcoding surveys, even when their cell numbers are moderate, as their cell DNA content is often extremely high [85].In studied tidal pools, we also observed dense populations of live and encysted dinoflagellates with light microscopy.The three dinoflagellate species that were most abundant in studied pools, Durinskia dybowskii (syn.D. baltica, Peridinium balticum), Levanderina fissa (syn.Gyrodinium uncatenum, G. instriatum, G. pavillardii), and Kryptoperidinium foliaceum (syn.K. triquetra), have been reported from various coastal systems across the world, including the tidal ponds and coastal lagoons of eastern North America [98][99][100].Several species of Dinophyceae found in this survey, such as Akashiwo sanguinea, Karlodinium veneficum, and Pfiesteria piscicida, are toxicogenic [101], while others such as Levaderina fissa may cause HABs without producing toxins [102].
The presence of heterotrophic dinoflagellate Oxyrrhis marina in the 0.2 µm fraction of water samples is difficult to explain as this is a benthic species with a body size larger than 8 µm [103].Dinoflagellate reads including those of Oxyrrhis in this fraction may represent extracellular DNA.While the read numbers of the dinoflagellate class Syndiniales, which includes only parasitoid taxa [104], were not especially high, their high diversity (20% of dinoflagellate ASVs) indicates the complexity of ecological networks in tidal pools.
Among diatoms, the genera Chaetoceros, Thalassiosira, Skeletonema, Minidiscus, and Cyclotella that both rbcL and 18S metabarcoding showed to be abundant in the water samples of the studied pools have been reported as dominant phytoplankters in New Jersey coastal lagoons adjacent to our sampling area [95,105] and in other coastal regions of eastern North America [106].RbcL ASVs of diatoms Minutocellus polymorphus and Arcocellus mammifer were abundant in the 0.2 µm water fraction in our samples.These diatoms are likely to thrive too in the open water of adjacent coastal lagoons, as M. polymorphus has been reported from Barnegat Bay, while Arcocellus could have been previously reported as Phaeodactylum tricornutum [95,105].These diatoms are extremely small and, therefore, easy to misidentify or underestimate in routine microscopy analysis.They were also not recoded in our diatom morphology-based counts of sediment samples, indicating that their lightly silicified frustules easily dissolve after cell death.The 8 µm water fraction had a high abundance of ASVs classified as the diatom Nitzschia draveillensis, commonly reported in the plankton of eutrophic inland waters [75].We also found N. cf.draveillensis, morphologically slightly different from the freshwater N. draveillensis, in our pool sediment samples.Diatoms identified as N. cf.draveillensis have been previously reported from oceanic plankton [107] and identified by the denaturing gradient gel electrophoresis band sequencing from an intertidal mudflat in South Carolina [50].In sediment samples, our metabarcoding results largely paralleled the morphology-based identification of diatoms at the genus level with Navicula, Halamphora, and Nitzschia as the dominant genera both in terms of morphotaxa and ASV diversity and abundance.These same genera appeared to be dominant in a metabarcoding study of salt marsh diatoms in South Carolina [52].Neritic planktonic diatoms also appeared abundant in sediment, which can be expected, as many of them (Chaetoceros, Thalassiosira, and Skeletonema) can either produce resting cells and spores [108] or live on the surface of intertidal sediments (Minidiscus) [26].At the species level, not much concordance between morphological and molecular identifications was found, which is expected given the low coverage of diatom diversity in reference databases [64,65].Morphology-based identifications revealed that a high proportion of taxa found in studied pools are common for the marsh soils and mudflats of the mid-Atlantic region [26,109].
Small flagellate (Mamiellophyceae, Chlorodendrophyceae, Pyramimonadophyceae, and Chlorophyceae) and coccoid (Trebouxiophyceae) green algae appeared very abundant in the water column of studied pools.They are frequently identified microscopically in coastal plankton and tidal pools [110] including New Jersey coastal lagoons [95,105], have been noted among dominant groups in metabarcoding surveys of marine coastal plankton [111], and were shown to be less abundant in benthic samples than in plankton [42].
Cerocozoans have been noted as one of the most diverse but less studied groups of protists in soils and marine and freshwater sediments [42,112], and we found a high diversity of cerocozoan ASVs in studied pools.Some of the most abundant cercozoan amplicons were attributed to the Protaspa lineage known to parasitize diatoms [113] and identified in metabarcoding surveys of marine plankton [104,113] and intertidal sediments [48].Another cercozoan abundant in the pools was Placopus pusillus, known for parasitizing flagellate green algae from the order Chlorodendrales [114].The most abundant free-living cercozoan was Ebria tripartita commonly detected in marine and inland saline waters [110].Eugenozoa likewise included organisms with a wide range of lifestyles such as free-living (e.g., Eutreptiella) and parasitic (e.g., Rapaza viridis) euglenids [115], mostly free-living and predatory diplonemids, and a variety of kinetoplastids.
There was no correspondence between our metabarcoding identifications of foraminifera apparently dominated by monothalamids and the lists of visually identified species from New Jersey salt marshes [25,27], where multichambered Globothalamea are most common.A similar discord in the taxonomic composition of foraminifera identified morphologically and with metabarcoding was found by Frontalini et al. [116], who also noted a prevalence of soft-walled monothalamids in their 18S metabarcoding dataset from the Adriatic Sea.
The relatively high abundance of nematode ASVs was expected, as they are often the dominant group in the meiofauna of coastal marshes [22,33].Within fungi, the prevalence of Ascomycota followed by Basidiomycota that we found in studied pools is characteristic of marine waters [86,117] and salt marshes [7,11], in contrast to freshwater habitats where other fungal groups are more common [118].

Comparison of Assemblages from Natural and Artificial Pools
The natural and artificial pools in this study were found to sustain significantly different microeukaryotic assemblages in terms of the differential abundance of individual ASVs, several major taxonomic groups, and overall assemblage composition.The limited number of sampled pools in our study and their spatial arrangement in two clusters requires this finding to be tested with a larger and spatially and environmentally randomized set of pools.Moreover, our survey was not designed to assess the temporal variation in species composition, while salt marsh microbial assemblages are known to experience considerable seasonal changes [9,48,119].While further confirmation is needed, these first metabarcoding data provide interesting insights into the impact of hydrological modification on the microeukaryotic diversity of salt marshes.The most striking difference between the two types of pools was the increased abundance of several groups of mixotrophic flagellates in OMWMs.While mixotrophic dinoflagellates, haptophytes, chrysopytes, and raphidophytes are mostly characterized as phagotrophs and less frequently as osmotrophs [120,121], osmotrophy is characteristic of pelagophytes [122] as their growth can be stimulated by dissolved organic compounds [123,124].The harmful blooms ("brown tides") of pelagophytes Aureococcus anophagefferens and Aureoumbra lagunenesis in the coastal waters of the eastern USA are associated with elevated concentrations of dissolved organic matter (DOM) [125,126].
DOM concentration was not measured in our study, but it is likely to be higher in OMWMs compared to natural pools because of their relatively greater ratio of shoreline length to pool area (Figure 1D).Narrow ditches with steep banks excavated in marsh peat should inevitably have water that is high in organic matter and, thus, stimulate mixotrophs and decomposers.Besides the direct consumption of DOM, bacterivorous mixotrophs such as dinoflagellates and raphidophytes may benefit from elevated organic matter via an increase in available bacterial biomass [120].Among bacterivorous mixotrophs, toxigenic raphidophytes Chattonella subsalsa and Chloromorum toxicum showed a remarkable increase in OMWMs in our study.These species are known for their toxicity to fish and mollusks and have been causing HABs with significant economic impacts in the coastal waters of the eastern USA [102,127].
We also found an increase in fungi and fungi-like protists such as peronosporomycetes (water molds) and labyrinthulomycetes in OMWMs, which may also be related to the higher organic matter availability to decomposers.Despite the considerable functional diversity within these groups, together with bacteria, they are ultimately responsible for the degradation of particulate and dissolved organic matter in aquatic ecosystems [128][129][130].
Besides the increased abundance of protistan groups known as predominantly mixotrophs and decomposers, OMWMs were different from natural pools in the composition of several taxonomic groups, for example, diatoms (Figure 7, right panel).It is not obvious from our data what factors may be responsible for this distinction, but the effect of added organic matter and increased turbidity in OMWMs cannot be excluded.For example, some diatom species are known for being partially or even entirely osmotrophic [131], but the extent of osmotrophy in different diatom taxa is largely unknown.As our results show, almost all taxonomic groups of microeukaryotes showed shifts in their assemblage composition if the sample size in terms of ASV and read numbers was sufficiently large.This is not surprising considering that the effects of organic matter increase may propagate through the entire food web, indirectly affecting predatory, parasitic, and symbiotic microeukaryotes through changes in their food sources and hosts [132].
Lewitus et al. [133] found abundant populations of HAB-causing flagellates in artificial ponds excavated in residential areas and golf courses in the coastal areas of South Carolina.Some of the species that formed blooms in these ponds, such as Chattonella subsalsa, were the same as in the tidal pools that we studied in New Jersey.Lewitus et al. [133] further hypothesized that these ponds could be "natural incubators" of harmful bloom-forming protists that proliferate in sheltered and nutrient-rich pond water and subsequently are spread over coastal areas by tides.The ability of most HAB-causing protists to form cysts also makes tidal pools a suitable refuge for surviving unfavorable conditions.Our results support the idea that tidal pools can serve as nurseries for bloom-forming protists, and artificial pools may further promote the growth of mixotrophic flagellates implicated in HABs.As blooms of harmful pelagophytes and raphidophytes began in coastal areas of the eastern USA in the 1980s [102,134], the question arises about the potential link between these blooms and the elevated organic matter concentrations in estuaries and coastal lagoons due to salt marsh modifications and destruction caused by sea level rise.
In conclusion, our survey revealed a diverse assemblage of microbial eukaryotes in salt marsh pools and confirmed the power of the metabarcoding approach for revealing the effects of environmental variation on the biota.It also raised questions about the impacts of salt marsh modifications on coastal ecosystems that could be addressed by further studies.

Figure 3 .
Figure 3. Percentage of 18S ASVs (left panel) and reads (right panel) assigned to major taxonomic groups in sediment and water samples.

Figure 4 .
Figure 4. Percentages of dinoflagellate (left panel) and ciliate (right panel) 18S reads assigned to the most abundant genera in sediment and water samples.

Figure 3 .
Figure 3. Percentage of 18S ASVs (left panel) and reads (right panel) assigned to major taxonomic groups in sediment and water samples.

Figure 3 .
Figure 3. Percentage of 18S ASVs (left panel) and reads (right panel) assigned to major taxonomic groups in sediment and water samples.

Figure 4 .
Figure 4. Percentages of dinoflagellate (left panel) and ciliate (right panel) 18S reads assigned to the most abundant genera in sediment and water samples.

Figure 4 .
Figure 4. Percentages of dinoflagellate (left panel) and ciliate (right panel) 18S reads assigned to the most abundant genera in sediment and water samples.

Coasts 2024, 3 ,Figure 5 .
Figure 5. Percentages of diatom rbcL reads assigned to the most abundant genera in sediment and water samples.

Figure 5 .
Figure 5. Percentages of diatom rbcL reads assigned to the most abundant genera in sediment and water samples.

Figure 6 .
Figure 6.Differences in read abundance among natural and artificial pools (OMWM), sediment samples.The horizontal lines represent median values, boxes-interquartile ranges, vertical lines-1.5×interquartile ranges, dots-outliers.Read numbers are rarefied to the smallest number of reads in the sediment dataset.

Figure 6 .
Figure 6.Differences in read abundance among natural and artificial pools (OMWM), sediment samples.The horizontal lines represent median values, boxes-interquartile ranges, vertical lines-1.5×interquartile ranges, dots-outliers.Read numbers are rarefied to the smallest number of reads in the sediment dataset.

Figure 7 .
Figure 7. NMDS plots of 18S (left panel) and rbcL-diatom (right panel) sediment data showing position of samples (filled circles) from natural and artificial pools (shown by ellipses) separated along the first NMDS axis, which also corresponds to a gradient of water salinity, pH, and turbidity.Samples from individual pools are connected by lines that meet at pool centroids.DO: dissolved oxygen; TEMP: water temperature; TURB: turbidity.

Figure 7 .
Figure 7. NMDS plots of 18S (left panel) and rbcL-diatom (right panel) sediment data showing position of samples (filled circles) from natural and artificial pools (shown by ellipses) separated along the first NMDS axis, which also corresponds to a gradient of water salinity, pH, and turbidity.Samples from individual pools are connected by lines that meet at pool centroids.DO: dissolved oxygen; TEMP: water temperature; TURB: turbidity.

Table 1 .
Summary of the environmental characteristics of studied tidal pools (minimum-meanmaximum).

Table 1 .
Summary of the environmental characteristics of studied tidal pools (minimum-mean-maximum).

Table 2 .
Comparison of assemblages between natural and artificial (OMWM) pools.Differences in assemblage composition assessed by PERMANOVA; differences in read abundance estimated by Welch t-test."F": pseudo-F statistic; "T": T-statistic; "ns": nonsignificant; "na": analysis not applicable or not conducted because of low occurrence of ASVs in rarefied datasets.Positive values of Tstatistic indicate prevalence of a group in natural pools, and negative numbers indicate prevalence of a group in OMWMs.

Table 2 .
Comparison of assemblages between natural and artificial (OMWM) pools.Differences in assemblage composition assessed by PERMANOVA; differences in read abundance estimated by Welch t-test."F": pseudo-F statistic; "T": T-statistic; "ns": nonsignificant; "na": analysis not applicable or not conducted because of low occurrence of ASVs in rarefied datasets.Positive values of T-statistic indicate prevalence of a group in natural pools, and negative numbers indicate prevalence of a group in OMWMs.