Seasonal variation in near-surface seasonally thawed active layer and permafrost soil microbial communities

Understanding how soil microbes respond to permafrost thaw is critical to predicting the implications of climate change for soil processes. However, our knowledge of microbial responses to warming is mainly based on laboratory thaw experiments, and field sampling in warmer months when sites are more accessible. In this study, we sampled a depth profile through seasonally thawed active layer and permafrost in the Imnavait Creek Watershed, Alaska, USA over the growing season from summer to late fall. Amplicon sequencing showed that bacterial and fungal communities differed in composition across both sampling depths and sampling months. Surface communities were most variable while those from the deepest samples, which remained frozen throughout our sampling period, showed little to no variation over time. However, community variation was not explained by trace metal concentrations, soil nutrient content, pH, or soil condition (frozen/thawed), except insofar as those measurements were correlated with depth. Our results highlight the importance of collecting samples at multiple times throughout the year to capture temporal variation, and suggest that data from across the annual freeze-thaw cycle might help predict microbial responses to permafrost thaw.


Introduction
Permafrost serves as a massive store of organic carbon (C) [1], covering around 15% of the exposed Northern Hemisphere land area [2]. Near-surface permafrost-up to 3 m deep-contains around 10 12 tons of C globally, or about 33% of the total soil pool [3]. Climate change driven permafrost thaw may generate substantial climate feedbacks, as carbon dioxide and methane released by thawing permafrost contribute to further warming [4]. Permafrost thaw can also alter surface geomorphologic processes and affect surface water chemistry [5,6].
Understanding the implications of thawing permafrost for landscape biogeochemistry requires knowledge of how soil microbes will respond to warming, since they mediate soil processes [3,7]. Microbial metabolic activity increases rapidly with higher temperatures during thaw [8]. Bacterial and fungal communities also show shifts in both composition and function [8][9][10][11][12][13][14][15]. Changes to longterm soil temperature averages, particularly at the 'transition zone' (where the seasonally-thawed active layer meets permafrost) can therefore be expected to impact microbial activity in permafrost regions [16,17]. Some changes are consistently observed, such as increases in soil respiration, and in genes for organic C degradation [9,14,15]. But outcomes such as changes to nitrogen (N) cycling and production of greenhouse gases like methane and nitrous oxide are likely to depend on changes to microbial community composition following thaw [18][19][20].
Understanding seasonal variation in soil microbial communities is key to anticipating long-term responses to permafrost thaw, since climate warming will initially entail lengthening summer season thaw and downward expansion of the active layer. Information on permafrost-affected microbial communities Not subject to copyright in the USA. Contribution of US Army Engineer Research and Development Center Environ. Res. Lett. 18 (2023) 055001 C C M Baker et al derived from summer fieldwork is readily available (e.g. 10,17), as are data from in situ artificial warming studies (e.g. 16,21). But despite evidence that Arctic soil microbial communities show seasonal variation in composition and abundance [22], our understanding of this variation remains limited, due in part to the relative difficulty of sampling from cold environments outside of summer. Indeed, a recent review of seasonality in Arctic soil microbial ecology [23] identifies an ongoing and pressing need to identify the dominant drivers of seasonality in Arctic soil microbial community composition, and to identify linkages with biogeochemical processes.
Previous work by Barker et al found seasonal changes in surface water metal concentrations at Imnavait Creek that coincided temporally with seasonal changes in thaw depth [24]. Other studies also report substantial increases and decreases in surface water nutrient and metal concentrations accompanying permafrost thaw [25,26]. However, soil microbes may be influenced by many potentially co-varying parameters including soil pH, temperature, water availability, vegetation, nutrient availability, and soil chemistry [22,[27][28][29]. While empirical evidence points to distinct bacterial communities in frozen and thawed surface soils [22], it is unclear whether these changes result from changing soil temperatures per se, or other environmental parameters such as the increased availability of organic matter during warmer periods due to plant and microbial photosynthesis. Furthermore, changes in soil parameters are likely to vary through the soil profile, with potentially different effects on microbial communities at different depths.
To explore the connection between seasonal thaw processes and soil microbial communities, we undertook a study at Imnavait Creek on the North Slope of Alaska, USA. We examined changes in the depth profile of microbial communities over the growing season from summer through fall, and sought to relate temporal changes in those communities to environmental parameters such as temperature, nutrients and metal availability. Our study therefore adds to the growing literature on microbial responses to permafrost thaw by (a) conducting observational field sampling during the relatively inaccessible fall season; (b) assessing both bacterial and fungal communities; and (c) seeking links between microbial community composition, biogeochemical processes and other drivers such as temperature, using pore water metal and soil nutrient concentrations as well as soil temperature and thaw depth data.
We set out with the following questions and hypotheses: (i) How do microbial communities vary with depth? We expected community composition to change, and diversity to decrease, with depth, in line with existing studies [13,[30][31][32][33]; or alternatively, for composition to differ between the active layer and permafrost, with higher diversity in the active layer [34][35][36]. (ii) How does the depth profile of microbial communities vary across the growing season? We expected community composition in the active layer to be more variable across the growing season than in permafrost on account of more variable abiotic conditions (repeated freezethaw, temperature variation, changes in metal concentrations etc) known to influence soil microbes [28,29]. (iii) How do soil microbial changes correlate with soil nutrients and chemistry? We expected samples with high metal concentrations to have distinctive community composition, dominated by microbial groups specialized in the use of those metals (e.g. Fe oxidizers and reducers for samples dominated by Fe) [28], and with lower diversity on account of the toxicity of metals to many microbes [37,38]. We also expected variation in nutrients and pH to be associated with variation in microbial community composition [39][40][41].

Field site
We measured soil temperatures and collected samples in the Imnavait Creek Watershed, in the northern foothills of the Brooks Range, Alaska, USA (latitude 68.61 • longitude −149.32 • ; figure 1(a)). The area is formed from Sagavanirktok glacial till [42] and overlies continuous permafrost 250-300 m deep [43]. Soils are poorly drained silty loams covered by a peat layer rich in organic matter. Parts of the soil profile have been described as having high chroma color, suggesting oxidation of iron minerals, while adjacent zones are gleyed, suggesting reducing conditions [44]. Vegetation includes tussock sedge tundra [45] and sphagnum moss/ericaceous plants [46].

Soil temperatures
Pairs of Onset HOBO U23 Pro v2 dual external temperature sensors (Bourne, MA, USA) were installed at depths of 20, 40, 60, 80, 100, and 120 cm at a single location adjacent to the sampling pits. To install these thermistors, we used a slide hammer and rod to excavate a 0.75 cm diameter hole and inserted plastic sleeves to protect the sensors. Hourly measurements were taken from each sensor from 6 June 2019 onwards. Twelve months of data are reported here, but measurements were ongoing at the time of writing (August 2022).

Soil sampling
Sampling occurred on 6 June, 5 August, and 8 October 2019. June and August represent early-and late summer, with temperatures above freezing and precipitation slightly lower than the annual peak in July (figures 1(b) and (c)). By October, conditions include below-freezing maximum temperatures and moderate precipitation. Summer 2019 was relatively wet compared to the past 30 years (figure 1(c)), especially in August, at which time the precipitation consisted of rainfall rather than snow.
In each sampling month, we excavated a single 1 m × 1 m pit to a depth of 1 m (figure S1), using cold chisels and a jackhammer cleaned with 70% isopropanol, DNA AWAY, and RNase AWAY (Thermo Fisher Scientific, Waltham, MA USA). All sampling was conducted wearing Tyvek suits, surgical masks, and nitrile gloves. Probing with a 1.7 m/1 cm diameter metal frost probe before digging revealed thaw depths of 5 cm, 46 cm and 58 cm in June, August, and October, respectively.
We collected samples at five depths: 0-20, 20-40, 40-60, 60-80 and 80-100 cm. At each sampling depth, we collected three samples of approximately 5 kg for the chemical analyses described in section 2.4. Samples were individually collected for each chemical analysis, since different assays required different collection methods (e.g. some required homogenization while others required soil heterogeneity remain intact), precluding sampling in bulk followed by homogenization. Each sample spanned the full 20 cm of the depth range it represented. Samples thus covered the entirety of the soil profile in 20 cm increments.
At each sampling depth, we collected a further three samples of approximately 500 g for the sequencing described in section 2.5. These samples were also collected individually from the pit, i.e. without prior homogenization. A jack hammer was used to chip out separate samples for each replicate, with each sample spanning the full 20 cm depth range. These samples were placed in sterile Nasco™ Whirl-pak bags (Thermo Fisher Scientific, Waltham, MA, USA) and shipped frozen (−4 • C) to our lab in Hanover, NH, USA. Subsamples of approximately 10 g were taken for amplicon sequencing, with processing conducted in a −12 • C cold room using oven-sterilized drill bits and chisels. Each subsample was placed in a sterile Whirl-pak bag and homogenized at room temperature by crushing with a mallet. Homogenized samples were transferred to 50 ml Falcon tubes and stored at −80 • C until DNA extraction.

Soil and pore water chemical analyses
Total C, total N, loss-on-ignition (LOI), and ammonium (NH 4 -N) were measured for each of the triplicate soil samples. We also extracted pore water from one sample at each depth in each sampling month for analyses of 23 elements: Al, As, Ba, Cr, Cu, Fe, Mn, Ni, Sr, Ti, V, Zn, Au, Cd, Co, Hg, Mo, Pb, Rb, Sb, Se, Sn and Tl. Details of these chemical analyses are provided in File S1 Supplementary Methods.

Amplicon sequencing
We extracted genomic DNA from the 10 g subsamples. DNA was sent for amplicon sequencing of the 16S rRNA gene (V4 region) with 515FY and 806RB primers to profile bacterial communities, and the internal transcribed spacer (ITS) region with modified ITS1F and ITS2 primers to profile fungal communities [47][48][49][50][51][52]. Libraries were sequenced at 2 × 251 bp on an Illumina MiSeq. Details of extractions and library preparation are provided in File S1 Supplementary Methods.

Statistical analysis 2.7.1. Soil temperature
We averaged the pairs of thermistor readings to generate a single temperature time series for each depth. To interpret these data, we compared the means to an assumed freezing point of 0 • C. However, we note that the true freezing point may not necessarily be 0 • C due to the presence of solutes. Furthermore, there is likely small-scale thermal heterogeneity that is not captured at the scale of our measurements. Sampling depths that measured close to 0 • C were likely partly frozen and partly liquid at a macro scale.

Soil and pore water chemistry
Technical replicate measurements of C, N, LOI, pH and NH 4 -N were averaged prior to analysis (with the exception of LOI for 0-20 cm in October, for which only one measurement was obtained). For the trace metals, measurements for 12 elements (Al, As, Ba, Cr, Cu, Fe, Mn, Ni, Sr, Ti, V, Zn) were above the corresponding detection limits (i.e. levels at which measurements can be reliably distinguished from zero) and reporting limits (i.e. levels at which measurements are reliably quantitative). Some or all of the measurements for the remaining 11 elements (Au, Cd, Co, Hg, Mo, Pb, Rb, Sb, Se, Sn and Tl) were below either the reporting limit or the detection limit, and we excluded these from further analysis. Scatterplots were used to visualize variation in each metal's concentration with depth and time. To visualize variation among samples, and look for patterns among the metals, we drew principal components biplots in R [60], using points to represent the samples, and arrows to represent the measured soil and pore water characteristics.

Microbial diversity
To quantify within-sample microbial diversity (alpha diversity [61]), we first normalized samples using total-sum scaling to account for read count variation (see File S1 Supplementary Methods for discussion of this normalization choice). We then calculated Shannon and inverse Simpson diversity using vegan::diversity() [62]. These diversity measures were chosen for their lower sensitivity to rare taxa compared to observed or estimated richness, since the dada2 pipeline is not intended to quantify rare taxa accurately (e.g. sample-level singletons are discarded). To evaluate the effect of sampling month and depth on the two diversity measures, linear mixed models were estimated with sampling time/depth as a random effect to account for our sampling design, and the significance of month and depth assessed using likelihood ratio tests. For bacterial and fungal Shannon diversity, we compared a model with depth, month and their interaction as fixed effects to a model with only the random effects. Fungal Shannon diversity was clearly associated with depth during August, and we repeated our analysis excluding that month to assess the effect of depth in June and October. However, since the interaction of depth and month was not significant, we compared a model with depth and month fixed effects to one with only the random effects. For bacterial Simpson diversity, diversity was so elevated at 0-20 cm in October that a linear model was clearly inappropriate. We therefore excluded the October 0-20 cm samples; however, models including month as a predictor invariably gave a singular fit, so we compared a model with depth fixed effects to a model with only the random effects. For fungal Simpson diversity, the full model including month/depth interaction gave a singular fit; we therefore compared a model with month and depth fixed effects to one with only the random effects. We repeated our analysis excluding the August samples, comparing a model with month fixed effects to one with only the random effects, and a model with depth fixed effects to one with only the random effects.

Microbial community composition
We constructed bar plots to compare relative abundances of the most abundant bacterial phyla and fungal classes (based on total-sum-scaled read counts). Between-sample variation in microbial diversity (beta diversity [61]) were visualized with nonmetric multidimensional scaling (NMDS) ordinations of Bray-Curtis dissimilarities [63] based on total-sum-scaled read counts. For bacterial communities, we additionally used NMDS to visualize weighted UniFrac distances [64] calculated using phyloseq::UniFrac() from phylogenetic placements determined with SEPP (see File S1 Supplementary Methods).
Compositional differences between sampling months and sampling depths were tested for significance using permutational multivariate analysis of variance (permanova) with sampling month and depth as predictors, and 999 permutations [65,66]. We tested for homogeneity of multivariate dispersion across combinations of month and depth using a permutational test. We used the same methods to visualize and test for compositional differences between frozen and thawed samples, by using observed thaw depth (i.e. measured with a frost probe) as the permanova predictor. The magnitude of compositional variation attributable to sampling month at different depths was examined by fitting a separate model with sampling month as the predictor to the microbial data at each sampling depth. We then took the sum of squares and the R 2 for month from each permanova, as well as the total sum of squares, and plotted these as a function of sampling depth. Distance-based redundancy analysis (dbRDA [67]) was used to quantify the variation attributable to measured soil and pore water characteristics. dbRDA performs constrained ordinations with arbitrary dissimilarity measures by using metric scaling to ordinate the dissimilarity data and then analyzing the results with conventional redundancy analysis. Mantel tests were used to evaluate the Pearson correlation between Bray-Curtis dissimilarities for bacterial and fungal communities. Finally, DESeq2 was used to determine a list of ASVs that showed significant logfold changes between sampling months, with each pair of sampling months considered separately.

Soil temperatures
Thermistors showed decreasing temperatures with depth throughout the growing season (figure 2). The mean measured temperature in 2019 was above 0 • C from mid-June until end-September at 20 cm, and from the beginning of June to end-September at 40 cm. Once above 0 • C, temperatures rose steeply, peaking as high as 8.7 • C at 20 cm and 4.2 • C at 40 cm in the latter half of July. Temperatures in the top 40 cm fell to ∼0 • C by the end of September and remained there until November, at which time they fell below those of the deeper soils. Soil at 60 cm and below remained <0 • C throughout the period of sampling. These deeper soils did not exhibit the substantial summer temperature rise seen in the top 40 cm, but did show a slow rise of 1 • C-2 • C from June until their peak in late November [68], before falling to around −10 • C by March. Since soils at 60 cm and below never rose above 0 • C over the course of the year, we interpret this as indicating that the 60-80 cm and 80-100 cm samples are permafrost, with the shallower depths constituting the active layer.
Most of the 12 trace metals retained in our analysis showed sharply elevated concentrations at 80-100 cm in October (figure 4). As and Cr showed a different pattern, with the highest concentrations for these metals occurring above 80 cm in August, and no sharp peak occurring at 80-100 cm in October.

Microbial diversity
In the 16S dataset, Shannon diversity decreased slightly with sampling depth in the June and October samples ( figure 5(a), left panel). Inverse Simpson diversity was elevated in the October 0-20 cm samples but otherwise invariant to depth ( figure 5(b), left panel). In the ITS dataset, both Shannon and inverse Simpson diversity showed an increase with depth in August, but June and October samples showed no change in diversity with depth (figures 5(a) and (b), right panels).

Microbial community composition
In the bacterial dataset, Acidobacteria and Verrucomicrobia were the most prominent phyla in the top 40 cm of the soil profile, while Caldisericota and Firmicutes were prominent below 60 cm ( figure  S2(a)). Acidobacteria and Caldisericota had especially high relative abundance in June compared to August and October. Verrucomicrobia showed increasing relative abundance from June to October, while Firmicutes had similar relative abundances across the three sampling months. In the fungal dataset, Agaricomycetes and Leotiomycetes were present throughout the soil profile in all three months ( figure S2(b)).
Bacterial and fungal community composition varied with both depth and sampling month (figures 6(a) and (b); UniFrac distances for bacterial communities showed similar variation, figure  S3). Since bacterial community composition exhibited a broadly depth-related gradient, samples that were thawed (surface) or frozen (deeper) at the time of collection tended to have different community compositions ( figure S4(a)). However, composition also separated by depth within the thawed or frozen samples, suggesting that depth rather than frozen/thawed condition was the chief driver of community composition. Further supporting this position, the two depths that thawed between June and October, i.e. 20-40 cm and 40-60 cm, did not show substantial changes in composition as they went from frozen to thawed. Fungal communities, on the other hand, showed a weaker relationship with depth and frozen/thawed condition ( figure S4(b)). Mantel testing failed to reveal any significant correlation between bacterial and fungal pairwise Bray-Curtis dissimilarities (r = 0.001, p = 0.481).
Month-to-month variation in both bacterial and fungal community composition was greatest in the top 20 cm of the soil profile, and accounted for a smaller fraction of total variation as depth increased ( figure 6(c)). There was no indication that this month-to-month variation differed discretely between the 0-40 cm depths and the 40-100 cm depths (which exhibited distinct temperature behaviors). The variation at 0-20 cm was especially notable in the bacterial dataset, with the October 0-20 cm samples clustering quite separately from other months at that depth. DESeq2 analysis identified a number of bacterial taxa that showed changes in abundance between June/August and October at 0-20 cm (File S3 DESeq2 Results). These included decreases in Acidobacteriota (including Occallatibacter and Candidatus Koribacter), Actinobacteriota (including Oryzihumus) and Desulfobacteriota (including Smithella), and changes in several Verruomicrobiota (especially Pedospharaceae) and Proteobacteria (including decreases in Rhodoferax ferrireducans and Pseudolabrys).
dbRDA indicated that measured soil and pore water characteristics were significantly associated with variation in overall community composition (figures 7 and S5). In the bacterial dataset, variation in community composition with depth was associated with nutrients (C, N, NH 4 -N) that were elevated towards the surface, as well as metals such as Fe and Al that spiked higher at 80-100 cm in October. On the other hand, nutrients showed limited temporal variation at the surface (figure 3) and consequently did not help explain variation in surface bacterial community composition.

Discussion
Our study identified spatiotemporal variation in soil microbial communities during summer through fall in the top 100 cm of active layer and near-surface permafrost. Community alpha diversity was mostly invariant to depth or sampling month, with the exception of elevated bacterial diversity in the October 0-20 cm communities and increasing fungal diversity with depth in August. But community composition varied with both depth and sampling month, especially in surface soils. Our results highlight the importance of sampling across multiple seasons to understand microbial communities and their response to their environment-especially as climate warming causes deepening of the active layer and shifts the timing of annual active-layer freeze-thaw cycles.
Our alpha diversity findings diverged from our expectation that diversity would decrease with depth, based on previous studies of bacteria in permafrostaffected soils [32,35,69]. One explanation may be differences in the range of depths considered. For example, in some studies that report decreasing diversity with depth, in temperate or cold-regions soils, the overall trend is strongly influenced by the top ∼10 cm (e.g. 31,70), which in our study may be mixed with slightly deeper soils in our 0-20 cm samples. The relationship between fungal diversity and depth in permafrost-affected soils is less wellstudied, but Chen et al report higher diversity in permafrost compared to active layer [36], consistent with our results.
Although community composition varied with depth, variation did not closely mirror measured soil temperatures, nor the inferred active layer/permafrost boundary. Thermistors recorded higher and more variable temperatures at 0-40 cm compared to 40-120 cm. We hypothesized that this would be reflected in compositionally distinct 0-40 and 40-100 cm communities, and greater temporal variation for 0-40 cm communities compared to those at 40-100 cm. But our results indicated that 0-20 cm and 20-40 cm samples were in fact distinct from one another and Fungal Simpson diversity also varied by month and showed an overall increase with depth (χ 2 = 12.8, df = 3, p = 0.005); this was driven by the August samples, without which month was borderline significant (χ 2 = 3.7, df = 1, p = 0.06) but depth did not significantly affect diversity (χ 2 = 1.7, df = 1, p = 0.19). from the deeper samples; and that temporal variation was much greater for the 0-20 cm samples than for the 20-40 cm samples. Similarly, recorded thaw depths (i.e. 5 cm, 46 cm and 58 cm in June, August, and October, respectively) were also not reflected in community composition or temporal variability.
Together, these observations suggest that seasonal thaw and refreezing are not the primary drivers of permafrost microbial community composition. Other depth-correlated parameters may have more explanatory power. For example, inputs of organic matter from plants and photosynthetic microbes likely influence soil microbes more at the surface than at depth [71]. Our soil nutrient measurements were suggestive of this, as C and N concentrations decreased with depth ( figure 3). We hypothesized that nutrients might help explain the high temporal variation at 0-20 cm, since elevated photosynthesis over the summer may, by late fall, contribute C to the upper parts of the soil profile [22]. However, C and N measurements were not higher in October at 0-20 cm, failing to provide evidence that these nutrients were driving temporal variation in surface microbial communities. This may reflect that soil C and N are unlikely to change substantially on the time scale of our study, and that total C and N are not good proxies for microbe-available nutrients. Future work examining changes over multiple years, and looking at measures such as dissolved organic/inorganic nutrients, would be helpful in clarifying our results.
Temporal variation in microbial community composition was also not obviously associated with variation in other measured environmental covariates. While community composition at 0-20 cm showed large shifts across the growing season, metal concentrations at that depth showed relatively little change, and changes in nutrient concentration did not correlate with variation in community composition. Conversely, several metals showed large changes at 80-100 cm, especially between August and October, while corresponding microbial communities showed little change. The metal and nutrient concentrations that we measured appear to have limited explanatory power over temporal variation in community composition. We suggest that future work test other soil characteristics that might be more relevant to microbes than total elemental or nutrient concentrations. Such characteristics might include metal redox states; dissolved organic and inorganic C; and measurements of phosphorus. On the other hand, recent work indicates that stochastic microbial community assembly processes are more important than deterministic ones during these the initial stage responses to permafrost thaw [13], in which case environmental characteristics may not show any strong relationships with community composition.
One caveat to our results is the effect of persistent DNA on our sequencing results, which might be expected to mask temporal variation in our sequencing results. Although this does not appear to be a  Each data point represents a single sample, with shapes denoting sampling month and colors denoting sampling depth. Shaded convex hulls are used to visually group replicates from the same month and depth. Note that only 11 constraining axes are used due to collinearity, and only 12 of the 15 depth/month combinations are included due to missing trace metal concentration data. See figure S4 for a constrained ordination of all 15 depth/month combinations using soil characteristics, for which data from all depths and months were available. general problem in our study, since we did detect community variation over time, we cannot rule out the possibility that it accounts for the differences in temporal variation at different depths. In particular, we might expect greater DNA stability at the lower soil temperatures found at greater depths. As a result, there may be more persistent DNA at those depths, and intra-annual variation, as a fraction of the community would appear to be lower.
Finally, sampling from even later in the year may help link our results with previous findings that soil temperatures in late fall were higher at intermediate depths than at the surface [24]. Thermistor data from this study showed decreasing temperatures with depth from June through October when the microbial samples were collected (figure 2). Measurements later in the year show the expected temperature inversion at the surface, but only after we had finished microbial sampling. Since temperature directly affects microbial metabolism and growth, sampling later in the year has the potential to reveal further changes to those identified in this study.

Conclusions
Our results show that the composition of soil microbial communities in the top 100 cm of these Arctic soils varies with both depth and time over the growing season from summer through fall. Variation was greatest close to the surface in the seasonally thawed active layer. Neither temperature nor measured environmental characteristics alone could explain community variation. Our results highlight the importance of sampling at different times of the year in order to assess temporal variation in active layer microbial community composition and particularly late in the season (late fall to early winter), while the active layer is at its deepest yearly extent.

Data availability statement
Raw sequence data are available from the NCBI's Sequence Read Archive under BioProject PRJNA934356. ASV tables, thermistor data, and metals/nutrients data are available with this article as supplementary datasets (Files S4 through S12). Climate data (figures 1(b) and (c)) are available from NOAA NCEI and the USDA NRCS SNOwpack TELemetry Network. Code for bioinformatic processing and statistical analysis is available from GitHub [54].
The data that support the findings of this study are openly available at the following URL: www.ncbi. nlm.nih.gov/sra/PRJNA934356. Data will be available from 01 January 2024.