Groundwater bacterial communities evolve over time, exhibiting oscillating similarity patterns in response to recharge

Time series analyses are a crucial tool for uncovering the patterns and processes shaping microbial communities and their functions, especially in aquatic ecosystems. Subsurface aquatic environments are perceived to be more stable than oceans and lakes, due to the lack of sunlight, the absence of photosysnthetically-driven primary production, low temperature variations, and oligotrophic conditions. However, periodic groundwater recharge should affect the structure and succession of groundwater microbiomes. To disentangle the long-term temporal changes in groundwater bacterial communities of shallow fractured bedrock community, and identify the drivers of the observed patterns, we analysed bacterial 16S rRNA gene sequencing data for samples collected monthly from three groundwater wells over a six-year period (n=230) along a hillslope recharge area. We show that the bacterial communities in the groundwater of limestone-mudstone alternations were not stable over time and showed oscillating dissimilarity patterns which corresponded to periods of groundwater recharge. Further, we observed an increase in dissimilarity over time (generalized additive model P < 0.001) indicating that the successive recharge events result in communities that are increasingly more dissimilar to the initial reference time point. The sampling period was able to explain up to 29.5% of the variability in bacterial community composition and the impact of recharge events on the groundwater microbiome was linked to the strength of the recharge and local environmental selection. Many groundwater bacteria originated from the recharge-related sources (mean = 66.5%, SD = 15.1%) and specific bacterial taxa were identified as being either enriched or repressed during recharge events. Overall, similar to surface aquatic environments, groundwater microbiomes vary through time, though we revealed groundwater recharges as unique driving factors for these patterns. The high temporal resolution employed here highlights the complex dynamics of bacterial communities in groundwater and demonstrated that successive shocks disturb the bacterial communities, leading to decreased similarity to the initial state over time.

indicating that temporal variation might be system specific, and that not all aquatic bacterial 68 communities show seasonal response patterns. 69 Unlike limnic and marine systems, groundwater microbiomes have been largely 70 ignored in temporal studies. Subsurface waters are not driven by seasonally varying primary 71 5 production due to the lack of sunlight, there is little temperature variation, presumably little to 72 no input of surface-derived fresh, easily available organic carbon (Akob and Küsel, 2011;Benk 73 et al., 2019), and groundwater residence time can exceed those of rivers and lakes. It was 74 therefore assumed that groundwater microbial communities would be relatively stable over 75 time (Griebler and Lueders, 2009). However, these systems appear to be less constant than first 76 assumed and there is evidence from short-term studies (< 1 year) that groundwater contrasting depths, surface connectivity, and hydrochemistry. We hypothesise that (1) the long 98 observation time (> 6 years) and high temporal resolution of our sampling will allow us to 99 detect temporal variations within the groundwater microbiome, (2) temporal patterns will vary 100 between wells due to varying local conditions and environmental selection strength, and (3) 101 successive disturbances, in the form of groundwater recharge events, will drive the temporal 102 variation through their impact on microbial diversity and abundance, including the introduction 103 of surface-derived taxa.

105
Study site 106 In the Hainich CZE, a multi-storey fractured aquifer system within Triassic limestone-    Supplementary Table S1) were sequenced in-house using library preparation       Figure S4C).

Specific bacterial taxa can be associated with recharge phases 302
The contribution of recharge-related taxa to the baseline communities (i.e. those present 303 during recession phases) in the groundwater were well-and period-dependent (Figure 4).

304
Among the three wells, bacterial communities in H43 received the highest taxa contribution 305 from the recharge-related source (mean = 41.0%, SD = 16.5%) over the investigated periods, 306 followed by H41 (mean = 33.5%, SD = 23.3%), and significant differences in the contribution 307 of the recharge-related taxa between periods were evident in these wells. In P18, the 308 overwhelming majority of bacteria originated from the recharge-related source in both H41 309 (mean = 66.5%, SD = 15.1%) and H43 (mean =57.8%, SD = 15.4%). The contribution of 310 15 recharge-related taxa was smallest in the periods between P15 and P17 when the water level 311 change during recharge was only minor. The contribution of recharge-related taxa was 312 consistently low (mean = 9.1%, SD = 4.6%) over the periods in well H52. 313 We observed considerably more taxa, across all taxonomic levels, that were 314 consistently associated with recharge events in well H41 (231) compared to wells H43 and H52    The discrete but recurring disturbances, in the form or groundwater recharge, are an

626
(2017), B) daily (peaks) and accumulated (grey shaded area) precipitation during the investigation period, C) dynamic changes of groundwater 627 level in different wells on a daily basis. The mean water level of the investigation period for each well was indicated with a value box and a dashed 628 line. "Recharge phases" were defined for hypothesis testing based on water level changes (increase: recharge; decrease: recession). A period, 629 labelled as "P + two-digit year", included one recharge phase and one recession phase, except for P13 which only had a recharge phase.

630
Abbreviations: bgl depth in meters below ground level (m), MAMSL meters above mean sea level.