Microbial assemblages in Arctic coastal thermokarst lakes and lagoons

Abstract Several studies have investigated changes in microbial community composition in thawing permafrost landscapes, but microbial assemblages in the transient ecosystems of the Arctic coastline remain poorly understood. Thermokarst lakes, abrupt permafrost thaw features, are widespread along the pan-Arctic coast and transform into thermokarst lagoons upon coastal erosion and sea-level rise. This study looks at the effect of marine water inundation (imposing a sulfate-rich, saline environment on top of former thermokarst lake sediments) on microbial community composition and the processes potentially driving microbial community assembly. In the uppermost lagoon sediment influenced from marine water inflow, the microbial structures were significantly different from those deeper in the lagoon sediment and from those of the lakes. In addition, they became more similar along depth compared with lake communities. At the same time, the diversity of core microbial consortia community decreased compared with the lake sediments. This work provides initial observational evidence that Arctic thermokarst lake to lagoon transitions do not only substantially alter microbial communities but also that this transition has a larger effect than permafrost thaw and lake formation history.


Introduction
Global climate warming is acceler ating permafr ost degr adation.Gr adual degr adation is manifested b y top-do wn permafrost thawing and thickening of the active la yer.T hermokarst processes lead to r a pid and deep thawing of permafr ost and the de v elopment of thermokarst ponds and lakes, which is extremely common in ice-and or ganic-ric h permafr ost (Gr osse et al. 2013, Olefeldt et al. 2016, Strauss et al. 2017 ).In Alaska, for example, thermokarst lakes have doubled in number and incr eased a ppr oximatel y by 37.5% in area from 1949 to 2009 (Walter Anthony et al. 2021 ).Thermokarst lakes in Siberian ice-rich permafrost have generally de v eloped since the early Holocene (Jongejans et al. 2020 ).Arctic thermokarst lakes contribute to ∼80% of Arctic contemporary CH 4 hotspot emissions and gener all y r elease lar ge amounts of methane r elativ e to CO 2 , and thus have a dispr oportionatel y high climate effect (Walter Anthony et al. 2018, 2021, Knoblauch et al. 2018 ).
Coastal erosion in the pan-Arctic can establish periodical or perennial connection of thermokarst lakes to the sea, which converts these lakes to lagoons .T hermokarst lagoons were estimated to account for 54% of the estimated total of ∼470 la-goons, whic h wer e identified along the Arctic coastline by remote sensing as of 2021 (Angelopoulos et al. 2021, Jenrich et al. 2021 ).Thermokarst lagoons represent a transitional state between freshwater thermokarst lakes and a fully marine environment.In these coastal lagoons, the hydrological connection to the sea plays a crucial role in facilitating the exchange of abiotic and biotic conditions between the two ecosystems (Gianuca et al. 2017 ).Vertical diffusion of marine water generates a sulfate-rich saline gradient on the top part of pr e vious fr eshwater sediments (Schindler 2019 ).Along with the transition, microbial methane cycling comm unity c hanges , for example , can influence carbon turnover and greenhouse gas emission (Yang et al. 2023 ).In an earlier study, w e sho w ed that within the sulfate zone, spatial cooccurrence of methane and sulfate thermodynamically favours sulfate-dependent anaerobic oxidation of methane, which mitigates methane emissions from thermokarst lagoons (Yang et al. 2023 ).
Thermokarst lakes and lagoons can serve as a natural laboratory to disentangle the mechanisms of microbial species replacement and e v aluate the envir onmental contr ols on micr obial comm unity assembla ge in r a pidl y degr ading permafr ost landsca pes.
Permafr ost usuall y limits dispersal of species due to its frozen state (Bottos et al. 2018 ), while thawing will alleviate the dispersal constraints on microbes .T he lateral and vertical expansion of thermokarst lakes pr esumabl y r e works the sediments to more homogeneous conditions than the pr e viousl y fr ozen gr ound.The infiltration of saline marine water into the thawed sediment will not onl y r e work the geoc hemical pr ofile in the lake, but also introduce marine microbes to the newly formed lagoon ecosystems.Subsea permafrost was found to contain an enormous amount of organic carbon (Miesner et al. 2023 ) originating from onshore terr estrial permafr ost, wher e micr obial dynamics wer e found to be linked with changes of geochemical conditions along the sedimentation history (Mitzscherling et al. 2019 ).Ho w ever, little is known about the changes of microbial structure and interspecies connection during the tr ansition fr om thermokarst lakes to lagoons.
This study investigates how microbial communities, beyond those involved in methane cycling, shift along the transition from coastal thermokarst lakes to thermokarst lagoons in the Arctic.We presume that the restratification of geochemical profiles following thermokarst lake to lagoon transitions result in restructuring and conv er gence of the core consortia and address how micr obial comm unities r espond to the div er ging geoc hemical conditions between thermokarst lakes and lagoons.We studied sediments of two thermokarst lakes and a lagoon from the Bykovsky Peninsula in northeastern Siberia where lagoons are extensively distributed and many thermokarst lagoons started to emerge about 2 ka before present (BP) (Jongejans et al. 2020 ) utilizing deep amplicon sequencing, and multiple numeric ecological appr oac hes.

Study site and sampling
Sediment cores of three thermokarst bodies were retrieved on the Bykovsk y P eninsula in the La pte v Sea, northeastern Siberian permafr ost r egion.Lake Golzov o y e (LG) and Northern Polar Fo x Lak e (LNPF) ar e fr eshwater thermokarst lakes while Polar Fox Lagoon (PFL) is a thermokarst lagoon to the south of LNPF (Fig. 1 ).Details about the thr ee r esearc h sites can be found in Yang et al. ( 2023 ).P aleoclimatic pr oxies suggested thermokarst erosion to LG and LNPF since 8 cal ka BP and lagoon formation of PFL started about 2 cal ka BP (Jongejans et al. 2020 ).The PFL has more dynamic environmental conditions because of seasonal hydrological connection to Tiksi Bay, which is broken by ice in winter (Schirrmeister et al. 2018, Jenrich et al. 2021 ), while the thermokarst lakes maintain gener all y stable fr eshwater conditions.
Sampling and subsampling were performed during a field expedition in April 2017.Thr ee cor es (PG2420, PG2426, and PG2423) wer e r etrie v ed for a total length of 5.2 m, 5.4 m, and 6.1 m, respectiv el y, fr om sediments of lake LG, LNPF, and PFL, using an UWITEC piston cor er.Subsequentl y, based on the specific r esearc h objectiv es of differ ent participants during the joint field campaign, the core segments were either stored in N 2 -flushed, vacuum sealed bags at ∼4 • C for pore-water analysis or sediment plugs were taken with sterile syringes dir ectl y in the field and subsequently frozen until further processing.The cores for microbial studies were divided into 49 samples, r epr esenting v arious depths in the sediment cores: 13 samples were retrieved from lake LG, 17 from LNPF, and 19 from PFL.In our recent study (Yang et al. 2023 ), we analyzed a subset of 23, which encompassed complete dataset of both geoc hemical and micr obial information.In the curr ent study, all the 49 microbial samples were used, independent of completeness of geochemical data, in order to obtain comprehensive information about microbial composition.

Bulk parameters and pore water chemistry
Briefly, total carbon , total organic carbon, and total nitr ogen wer e measured on bulk material using Elementar Micro Vario elemental analyzer (Elementar Analysensysteme, Hanau, Germany).The por e w ater w as drained into a vacuum syringe in an anaerobic glove box (N 2 :H 2 , 95%:5%).The corresponding analyses included alkalinity, sulfate, c hloride, nitr ate, ferric, and ferrous iron.Alkalinity was measured by colorimetric titration, cations and anions wer e measur ed with suppr essed ion c hr omatogr a phy, while the dissolv ed ir on (ferric and ferr ous) concentr ations in por e water wer e measur ed via spectr ophotometry by the ferr ozine method (Viollier et al. 2000 ).All samples were measured in triplicates, the geochemical data together with detailed method description have been deposited at GFZ Data Services ( https:// doi.org/ 10.5880/ GFZ.3.7.2022.001).

DN A extr action and libr aries prepar a tion for Illumina sequencing
Total nucleic acids were extracted in duplicates using the Po w erSoil-Kit (MO-Bio) accor ding to the manufacturer's protocol.Amplicon libr aries wer e pr epar ed b y using bar coded primer pair sets (Uni515-F[5 -GTGTGYCAGCMGCCGCGGTAA-3 ]/Uni806-R[5 -CCGGA CTA CNV GGGTWTCTAAT-3 ]), with duplicates for each sample.PCR reactions (50 μl) contained 10 × Pol Buffer C (Roboklon GmbH, Berlin, Germany), 25 mM MgCl 2 , 0.2 mM dNTP mix (Ther-moFisher Scientific), 0.5 mM each primer (TIB Molbiol, Berlin, Germany), and 1.25 U of Optitaq Polymerase (Roboklon, Germany).The PCR pr ogr am included an initial denatur ation step at 95 • C for 7 min, follo w ed b y 33 c ycles at 95 • C for 15 s, annealing at 60 • C for 30 s, extension at 72 • C for 30 s and a final extension step at 72 • C for 5 min.After purification with the Agencourt AMPure XP kit (Beckman Coulter, Switzerland), the recovered PCR products wer e equilibr ated into compar able equal amounts befor e pooling together with positive and negative controls.For the positive controls, we utilized a commerciall y av ailable moc k comm unity (Zy-moBIOMICS Micr obial Comm unity DN A Standar d II).As for the negativ e contr ols, they consisted of the DNA extraction buffer and the PCR buffer.Sequencing was run in paired-end mode (2 × 300 bp) on Illumina MiSeq platform by Eurofins Scientific (Konstanz, Germany).

Da ta processing, numeric, and sta tistical anal ysis
Raw sequences were demultiplexed by a custom Python script which used the 'make.contigs'function (pdiff = 2, bdiff = 1, other settings by default) in Mothur (v.1.39.5) (Schloss et al. 2009 ) to gener ate r eport files, upon whic h the r aw sequences wer e demultiplexed into individual samples.After orientation correction with 'extract_bar codes.py' in QIIME1 (Ca por aso et al. 2010 ), the sequences were processed by D AD A2 (maxN = 0, maxEE = 2, truncQ = 2, and minLen = 175) and the output was reported in the format of an amplicon sequence variant (ASV) table (Callahan et al. 2016 ).The taxonomy was assigned against the SILVA138 database (Quast et al. 2013 ).Negative controls were emplo y ed to assess the contamination during DNA extraction and PCR processes, positiv e contr ols ensur ed that the sequencing itself did not introduce noticeable errors .Moreo ver, the sequencing duplicates demonstrated high consistency ( Figure S1 , Supporting Information ).The contribution of different community members to the total abundance and beta diversity (Bray-Curtis dissimilarity, BC) was summarized by using R pac ka ge otuSummary (version 0.1.1)(Yang 2020 ).The data obtained fr om eac h of the 49 samples, including their r espectiv e duplicates, wer e combined.The v ery r ar e ASVs with a cum ulativ e count less than 10 across all samples were removed, resulting in the retention of a total of 25 880 ASVs .T he microbial community dissimilarity was explored by nonmetric multidimensional scaling (NMDS) by using R pac ka ge v egan (v ersion 2.5.7)(Oksanen et al. 2019 ) based on the BC dissimilarity from Hellinger transformed data to mitigate the excessive effect of r ar e taxa.Following the clustering in NMDs, a hier arc hical clustering ( Figure S2 , Supporting Information ) was performed to identify the grouping feature of samples by R base pac ka ge (R Core Team 2014 ).With that, permutational MANOVA was completed by 'adonis2' function of vegan package with BC matrix.To detect taxa, which were significantly enriched in the freshwaterand marine water-influenced sediments, linear discriminant analysis (LDA) effect size (LEfSe) was performed by using R pac ka ge microbiomeMarker (v1.1.2),based on normalized data by using a negative binomial model (Cao 2021 ).
To detect associations between micr oor ganisms fr om thermokarst lakes and lagoon, network analysis was implemented to explore the taxon co-occurrence patterns and the niche spaces.An initial filtering removed poorly represented ASVs with mean r elativ e abundance < 0.5% from the whole community dataset, follo w ed b y a secondary filtering to get those ASV lineages with the Spearman correlation coefficient (absolute value > 0.75) and P -value ( < .01).Afterw ar ds, a netw ork object was generated and analyzed by R package igraph (version 1.2.10) (Csardi and Nepusz 2006 ).Community modules of the network were detected with the 'cluster_edge_betweenness' algorithm of igr a ph pac ka ge .T he final network contained 194 ASVs.Based on the membership affiliation of each node (which represents individual ASVs), an NMDs plot was generated to explore the pr efer ential occurr ence of module members (ASVs) ov er differ ent samples.A nonpar ametric Welc h t -statistic was used to test the separation of each module over different groups with base pac ka ge in R. In addition, the one-dimensional dia gr am was used to display the r epr esentativ e of individual modules over samples by using the function ' linestack ' from vegan package.

En vironmental fea tures
Exploratory ordination analysis on environmental variables, whic h wer e based on the por e water geoc hemistry and C, N content of bulk sediments, suggested that the marine water influenced samples, which were entirely composed of the uppermost 3 m sediments of PFL clustered a wa y from the fresh water sediments ( Figure S3 , Supporting Information ).The marine cluster wer e c har acterized by high le v els of sulfate, salinity, and alkalinity, with highl y enric hed δ 13 C of methane ( −54 ‰∼−37 ‰) in contrast to the depletion ( −90 ‰∼−75 ‰) of freshwater sediment samples.T he marine influence , thus , had a larger effect than that of the location.
Collectiv el y, the 14 pr edominant phyla account for an av er a ge of 74% (first quantile: 66.2%, median: 75.1%, third quantile: 85.2%) to the total BC dissimilarity.The NMDs suggested two separate clusters of microbial communities, with one cluster consisting of Cummulative abundance(%) Figur e 2. T he abundance of dominant phyla with mean r elativ e abundance greater than 1% over all samples .T he 14 abundant phyla account for 90% of the total abundance .T he y -axes in the left and right denote the scales for the barplot and cumulative abundance (line in gre y), respecti vely.
Figure 3. Bubble plot showing abundance variation of the 45 abundant lineages (with mean relative abundance > 0.35%) over depths for the three thermokarst lakes (LG: Lake Golzov o y e, LNPF: Northern Polar Fox Lake, and PFL: Polar Fox Lagoon) in this study.Along the vertical axis, the taxonomy was presented at the family rank, and if assigning to the family level was not feasible, the next available higher taxonomic level was utilized.The r elativ e abundance was calculated by combining the archaeal and bacterial ASVs and then collapsed at family level for this plot.The bubble colours correspond to different phyla, while the size of the bubbles reflects the average relative abundance.
samples from the brackish layer of PFL influenced by marine water (until the sample PFL_220 r etrie v ed at depth of 220 cm), while the second cluster encompassed samples from freshwater sediments ( Figure S3 , Supporting Information ).Interestingly, this pattern aligns closely with the two clusters observed in the environmental ordination, which correspond to sediments influenced by fr eshwater and br ac kish water, r espectiv el y ( Figur e S3 , Supporting Information ).The freshwater-and saltwater-influenced microbial clusters were statistically different ( P < .001)according to adonisbased nonparametric MANOVA.
In the freshwater-influenced samples, a total of 8 c har acteristic taxa were observed with mean relative abundance > 2%, F igure 4. Netw ork sho wing the pattern of module members (A).The ASVs in this plots w er e filter ed out fr om the whole bacterial and arc haeal dataset over 49 samples.The edges within a module and between modules were coloured in black and red, respectively.The NMDs (B) shows the association between module members (ASVs, r epr esented by points) and samples (illustrated by cross symbols in the plot).The labels of the ASV lineage and samples were not shown in the plots to avoid crowdedness.
In contrast, the lagoon subgroup was represented by Anaerolineaceae (Chlor oflexota), Spor osarcina, and Clostridium sensu stricto 13 (Bacillota, also known as Firmicutes).Additionall y, linea ges from Caldatribacteriota JS1 were abundant in both habitat groups.ANME-2a-2b was not highlighted as a c har acteristic linea ge of the marine-water group as they largely prevailed only at the upper tw o lay ers among the total eight marine-water-influenced group, despite of their very high abundance in two sulfate-rich depths of lagoon sediments.

Microbial co-occurrence and the environmental dri v ers
The network constituted 194 ASVs (diameter: 11.01673, mean distance: 4.688331, and av er a ge clustering coefficient tr ansitivity is 0.765) with 912 edges, which show almost entir el y positiv e association except for one negativ e inter action between ASV4 (Chloroflexota; GIF9) and ASV_193 (Actinomycetota; Cryobacterium).The network suggests nine nonrandom modules (modularity 0.5635) (Fig. 4 ).In this study, two modules (M1 and M2) exhibiting high species richness were predominantly observed in freshwater sediments, while a distinct and closely interconnected subgroup (M3) dominated the lagoon sediments influenced by marine water inundation (Fig. 4 ).The one-dimensional plot r e v ealed that subgroups M3 and M6 were predominantly present in the brackish la yers , whereas M7 was mor e commonl y found in the upper layers.On the other hand, members of M1 and M5 wer e primaril y abundant at the deeper part of freshwater sediments ( Figure S4 , Supporting Information ).
The module M3 comprised two archaeal and 40 bacterial ASVs, spanning across 10 different phyla.More than half of the ASV phylotypes were affiliated with Chloroflexota (11 ASVs, mainly from Anaerolineaceae), Caldatribacteriota (comprising eight ASVs of JS1), Pseudomonadota (also known as Proteobacteria, consisting of se v en ASVs fr om Gamma pr oteobacteria in this study) and Bacteroidota (with six ASVs from Flavobacteriaceae and Ignavibacteriaceae).Additionally, this module included two archaeal linea ges, namel y fr om Halobacter ota (one ASV fr om ANME-2a-2b) and Asgar dar c haeota (one ASV fr om Lokiarc haeia).Suc h pr eference to marine-water inundation was also reflected by LEfSe analysis (Fig. 5 ).Nonparametric Wilcoxon test implied statistical significance of the abundance between freshwater sediments and marine-water influenced lagoons for each module (Fig. 6 ).For the freshwater sediments, pairwise adonis analysis did not reveal statistical significance across different modules.

Discussion
This study demonstrates a substantial change in microbial communities following the infiltration of marine water into freshwater thermokarst lake sediments .T hese differences were greater than differences of microbial communities between the different lakes and the deeper (freshwater influenced) lagoon sediments.For the thermokarst lake sediments, multiple paleo-proxies have r e v ealed r elativ el y stable geoc hemical conditions with minor v ariations over about 8 ka BP when the studied upper 8 m of the sediments accumulated (Jongejans et al. 2020 ).In the thermokarst lagoon, marine-water inundation has generated a sulfate zone on top of the sediments since at least 2 ka BP.Both, the freshand marine-water-influenced sediments wer e pr obabl y subjected to r elativ el y stable pr ocesses during the history of lake de v elopment, meaning that those geogr a phicall y adjacent lakes have likel y r eceiv ed por e waters fr om compar able sources and hav e experienced stable hydrologic conditions according to the low and stable electrical conductivity (Jongejans et al. 2020 ).Considering the longstanding anoxic and r elativ el y stable conditions in the thermokarst lakes, the low le v el of envir onmental v ariability likel y resulted in the o verall con vergence of microbial community composition.The thermokarst lagoon has seasonal connection with Figure 5. LDA effect size (LEfSe) identified c har acteristic taxa with statistically different abundance in the freshwater sediments and saltwater-influenced saline lagoon layers (the negative and positive parts along the x -axis, respectively).Taxonomy was given along the y -axis.If assignment to the genus le v el was not possible, the lowest possible taxonomic assignment was used.The dots in the plot were coloured by the modularity membership in the network plot Fig. 4 , each point represents an ASV lineage.marine water, which not only caused more dynamic geochemical variation than the freshwater sediments, but also introduced ne w micr oor ganisms.Owing to the periodic input of marine micr oor ganisms , the sea water-affected part of the thermokarst lagoon sediment microbiome potentially experienced a greater influence of species gain or loss, in addition to the preceding effect of a homogeneous environment.
In our findings, we observ e onl y slight differ ences in micr obial community composition across the freshwater thermokarst sediments in general.This could potentially be attributed to the relativ el y shallow depth of the sediment profiles examined, the geogr a phic pr oximity of the thr ee r esearc h sites, and the r elativ el y stable environmental conditions as mentioned above.The frozen conditions inherent to permafrost typically impose strong limitation on the spatial distribution and exchange of micr obes, r esulting in island biogeogr a phy patterns and div er gent comm unities (Bottos et al. 2018 ), while the physical constraints within thermokarst sediments were greatly alleviated, which facilitates a higher turnover of species.Although spatial distance may still influence the rate of species replacement, the local and microspatial scales in thermokarst sediments are not expected to significantly impede the vertical and lateral exchange of microorganisms .T his is especially true when there is a robust hydrological connection that facilitates species turnover within the sediments.The co-occurrence of closely related taxa, observed as module 3 in the thermokarst lagoon (Figs 4 and 5 ), further emphasizes the homogeneous nature of microbial communities in the sediments of all three research sites.
The shift from thermokarst lake (LNPF) to lagoon (PFL) resulted in a decr eased div ersity of the core microbial network (number of modules).This is manifested by the co-occurring bacterial subgroups that decreased from eight in freshwater sediments to one in the br ac kish la goon sediments.In this study, almost all members within the different modules are positively connected to each other.Positive associations can enhance biological fitness of a module thr ough m utualism or syntropy (Fisher et al. 2017 ), which often occurs in phylogenetically related microbes or is driven by similar environmental conditions and habitat niche (Weiss et al. 2016 ).Mor eov er, netw ork modules w er e often r egarded as a functional unit (Wang et al. 2017 ) and the multifunctional equivalent of trophic complementarity (Monto y a et al. 2015 ).In this case, the overwhelming module diversity of freshwater sediments suggests higher functional diversity than the marine-water-inundated sediments.Since community modules ar e gener all y gov erned by habitat featur es and nic he differ ence Fresh sediments Marine-influenced sediments Figure 6.Boxplots comparing the means of freshwater sediments and marine-influenced lagoon sediment by using nonparametric Wilcoxon test for the nine modules identified from the network.For each module the number of ASV members were given in parenthesis .T he abundance was log-transformed for better visualization, where log 10 (abundance + 1) means transforming a value plus 1, which allows for handling values of zero in micr obial comm unity data.The number of observ ations was giv en within eac h boxplot.This plot was gener ated by R pac ka ge ggpubr (v ersion 0.4.0)(Kassambara 2020 ).
sity in the br ac kish la goon sediments may be a special ada ptation to the sulfate-rich saline c har acteristics, whic h led to the observed distinct and densely clustered group separate from those of the freshwater sediments ( Figure S4 , Supporting Information ).The distinct single module among the br ac kish la goon gr oup (M3) may r epr esent a specialized functional gr oup, whic h ada pted to the sulfate-rich sediments.In line with the loss of module diversity of the network, a substantial decline in the r epr esentativ e taxa was also observed after the lagoon transition (Fig. 5 ; Figure S3 , Supporting Information ).The consistent change in microbial comm unity assembla ge pr ovides e vidence of significant habitat filtering following the thermokarst lake to lagoons transition.Members of the r epr esentativ e module in the saline layers of the lagoon (M3), including ANME-2a-2b, Sva1033, Maribacter, Psyc hr obacter, and Lokiarc haeia, hav e potential roles as carbohydr ate fermenters, r educers of sulfate, nitr ate or ir on, psyc hr ophiles, or halophiles tolerant to cold environments ( Table S1 , Supporting Information ).It is worth noting that ANME-2a-2b was particularly abundant in only two sulfate-rich sediment layers in the upper lagoon (not in the other six samples of the marine influenced module group), as highlighted previously (Yang et al. 2023 ).Ho w e v er, this linea ge was not r ecognized as c haracteristic taxon because it was not abundant in most samples within a group.The anaerobic methanotrophs ANME-2a-2b engage in methane oxidation through syntrophic cooperation with sulfate-reducing bacteria (SRB), an essential process for reducing methane emissions from the ocean into the atmosphere (Boetius et al. 2000 ).The w ell-kno wn (and potential) sulfate reducers such as Desulfobacterota SEEP-SRB1 and Sva1033 co-occurred with syntrophic partners, including members of Lokiarchaeia, Fla vobacteriaceae , Caldatribacteriota JS1, Anaerolineaceae , and SBR1031, as such both parts can benefit from their establishment in the upper lagoon sediment la yers .Additionall y, prior r esearc h on the lagoon sediments, the thermokarst lagoon water column has been associated with strong methane oxidation during winter (Spangenberg et al. 2021 ).
Members of the bacterial JS1 group appeared to be very important ov er all.JS1 is affiliated to Caldatribacteriota (pr e vious Atribacteriota, OP9) (Katayama et al. 2020 ), which was frequently observed abundant (31%-40%) in anoxic, organic-rich, and methanecontaining bottom sediments (Webster et al. 2007, Carr et al. 2015, Lee et al. 2018 ), as well as in Arctic marine sediment with high methane concentrations (Carrier et al. 2020 ).A recent study on Baltic Sea methane hotspots suggested that JS1 together with Dehalococcoidia in Chloroflexi was strongly correlated with anaerobic methane oxidation rates (Iasakov et al. 2022 ).As suc h, the pr e v alence of bacterial phylotypes of JS1 in both marine and freshwater sediments of the studied sediments likely highlight the ecological importance of this generalist taxon.Aside from JS1, lineages of Bathy ar cheota occurred as abundant archaeal members in the ecosystem.Bathy ar chaeotal members are able to perform acetogenesis, potentially methane metabolism, and dissimilatory nitrogen and sulfur reduction, and can interact well with anaerobic methane-oxidizing archaea, acetoclastic methanogens, and heter otr ophic bacteria (Zhou et al. 2019 ).The versatile metabolic potential of this lineage should facilitate their pr e v alence in anoxic sediments .Moreo ver, metagenomic data on the same lagoon studied here has recently explored nineteen Bathy ar chaeotal genomes, which serve as peptide degraders and acetogenic microbes (Berben et al. 2022 ).

Conclusion
This stud y re presents an exploration of the microbial composition in Arctic coastal thermokarst lakes and a lagoon and suggests substantial shifts in micr obial comm unity due to br ac kish marine water inundation in the long term.It also demonstrated distinct micr obial comm unity compositions between marine-and freshw ater-influenced lay ers of the same thermokarst lagoon sediment r epr esenting former permafr ost layers and ne wl y formed lake sediment.This suggests that lagoon formation alters microbial assembla ges mor e than thermokarst lake formation.In the uppermost lagoon sediment la yers , microbial communities adapt to the sulfate-rich conditions with a reduction in spatial variation and diversity of the core microbial population.Ho w ever, the sulfaterich conditions in the top br ac kish layer of the thermokarst lagoon result in a distinct core species assemblage prevailing at the freshwater-marine interface.

Figure 1 .
Figure 1.Maps of the study site showing (A) location with respect to the Northern Hemisphere and permafrost extent regions (B) location with respect to the Bykovsky Peninsula, and (C) r elativ e location of the lakes and the lagoon (modified from Yang et al. 2023 ).
Mean relative abundance (%) C h lo r o f le x o t a C a ld a t r ib a c t e r io t a P la n c t o m y c e t o t a A c id o b a c t e r io t a A c t in o m y c e t o t a P s e u d o m o n a d o u lf o b a c t e r o t a T h e r m o p la s m a t o t a V e r r u c o m ic r o b io t a P a t e s c ib a c t e (Lima-Mendez et al. 2015 ), a substantial decline of module diver-Yang et al.