Warming mediates the resistance of aquatic bacteria to invasion during community coalescence

The immigration history of communities can profoundly affect community composition. For instance, early‐arriving species can have a lasting effect on community structure by reducing the invasion success of late‐arriving ones through priority effects. This can be particularly important when early‐arriving communities coalesce with another community during dispersal (mixing) events. However, the outcome of such community coalescence is unknown as we lack knowledge on how different factors influence the persistence of early‐arriving communities and the invasion success of late‐arriving taxa. Therefore, we implemented a full‐factorial experiment with aquatic bacteria where temperature and dispersal rate of a better adapted community were manipulated to test their joint effects on the resistance of early‐arriving communities to invasion, both at community and population level. Our 16S rRNA gene sequencing‐based results showed that invasion success of better adapted late‐arriving bacteria equaled or even exceeded what we expected based on the dispersal ratios of the recipient and invading communities suggesting limited priority effects on the community level. Patterns detected at the population level, however, showed that resistance of aquatic bacteria to invasion might be strengthened by warming as higher temperatures (a) increased the sum of relative abundances of persistent bacteria in the recipient communities, and (b) restricted the total relative abundance of successfully established late‐arriving bacteria. Warming‐enhanced resistance, however, was not always found and its strengths differed between recipient communities and dispersal rates. Nevertheless, our findings highlight the potential role of warming in mitigating the effects of invasion at the population level.


| INTRODUC TI ON
Variation in the composition of ecological communities can be the product of historical processes such as immigration, extinction and speciation (Fukami et al., 2007). The timing of immigrants' arrival (immigration history) can profoundly affect community structures and maintain diversity of communities in a landscape via a process known as priority effects (Fukami, 2015). Priority effects imply that early-arriving species, through niche modification and/or resource depletion, gain advantage and become resistant to invasion of latearriving ones, and therefore maintain high relative abundances over time (Lockwood et al., 1997;Rillig et al., 2015). Successful local adaptation by early-arriving species initiates priority effects that can reduce the establishment success of late-arriving species that are otherwise well-adapted to the local environment (Loeuille & Leibold, 2008). On the other hand, priority effects can be absent or weak when late-arriving species generate species replacements. In the latter case, dispersal initiates species sorting processes even at very low rates of dispersal (Declerck et al., 2013), which eventually selects species that are better adapted to the given environment.
All these processes can play an important role whenever mixing of communities (known as "community coalescence") occurs (Rocca et al., 2020). Such community interchange events vary in the extent to which the environments of the coalescing communities are involved in the exchange, the mixing ratio of the communities, as well as temporal aspects of the event, which then all influence the resulting species establishments, exchanges and extinctions (Rillig et al., 2015). In cases where coalescence does not result in substantial environmental alteration and the mixing ratio is skewed towards the early-arriving community (e.g., inflow of a stream into a lake or the movement of propagules on sea splash by wind) the outcome of the community coalescence is expected to be influenced by priority effects.
Previous studies have demonstrated that organisms with high growth rates have the capability to facilitate strong priority effects (De Meester et al., 2002;Lee et al., 2013;Peay et al., 2012;Ruiz-González et al., 2015;Tucker & Fukami, 2014). Hence, any possible environmental factor that increases the abundance and growth of early-arriving species could possibly enhance priority effects, and thus increase community persistence and make the recipient communities resistant to further immigration (Chase, 2010;Rudolf & Singh, 2013). Therefore, priority effects are expected to be more important in species with short generation time (e.g., bacteria) (De Meester et al., 2016). Aquatic bacteria are most probably continuously subject to priority effects due to mixing of waterbodies, for instance at riverlake interfaces. Two recent studies aimed to identify distinct roles of bacterial groups in priority effects during community succession in a biofilm (Brislawn et al., 2019) and experimental freshwater bacterial communities (Rummens et al., 2018). Brislawn et al. (2019) found that the majority of taxa had been replaced after a 56-day period of succession, thus were not persistent over time despite some indication of priority effects. In the study of Rummens et al. (2018), time lags in inoculation history resulted in priority effects but the responses of individual bacteria to immigration was diverse, meaning that the initial relative abundances of either the early-arriving or late-arriving bacteria could not predict the outcome of invasion, highlighting the unpredictability of a multispecies system. Focusing on macroorganisms, a recent study by Grainger et al. (2018) showed that warming increased the competitive exclusion of a late-arriving competitor and increased the importance of priority effects by aphids that arrived early.
The impact of ongoing climate change leads to increased mean water surface temperatures (IPCC, 2014) that can increase the growth rates of aquatic organisms. However, microorganisms are generally understudied in climate change-context studies (Cavicchioli et al., 2019) and there is, to our knowledge, not a single study investigating the effects of warming on the outcome of community coalescence in aquatic bacteria. Although several studies suggest that priority effects occur in a variety of aquatic bacterial communities (Andersson et al., 2014;Lee et al., 2013;Rummens et al., 2018;Svoboda et al., 2018;Tan et al., 2012), it is currently unclear how the joint effect of warming and mixing ratio influences priority effects and the outcome of coalescence of aquatic bacterial communities. Furthermore, we lack knowledge about the identity of key bacteria that can persist or successfully invade a local community following coalescence. Therefore, we aimed to investigate whether the fates and roles of distinct aquatic bacteria during community coalescence differ in response to warming and if aquatic bacteria can resist immigration, for example due to priority effects. We performed a full-factorial experiment, where bacteria from three Swedish lakes were inoculated and grown in cell-free Baltic Sea medium at three temperature levels. These lake communities represented the early arrivals that were allowed to colonize the "foreign" (Baltic Sea) medium to which they were not priori adapted. After initial growth and establishment, these communities became the "recipient communities" that were exposed to invasion by Baltic Sea bacteria ("dispersal source community"). These late-arriving communities were mixed into the recipient communities at three different mixing (dispersal) rates. Here, we wanted to focus on how potential priority effects by recipient communities can diminish the advantage of late-arriving Baltic Sea bacteria that are historically adapted to the environmental conditions (i.e., Baltic sea medium). The original recipient communities differed in their proximity to the Baltic Sea in order to assess whether the geographic distance to the origin of the invading bacteria might affect the strength of community resistance. We expected that recipient communities closer to the Baltic Sea might have been exposed to dispersal from the Baltic Sea in their recent history to a greater extent than those farther from the Baltic Sea. This could have resulted in a larger shared species pool, including larger numbers of bacteria of Baltic Sea origin in local lake seed banks (Comte et al., 2014). At the end of the experiment 16S rRNA-based bacterial community composition was analysed, and the persistence of recipient communities and the immigration success of late-arriving bacteria were investigated both at community and population level.
We hypothesized that warming should increase resistance of aquatic bacterial communities (e.g., due to potential priority effects).
In other words, late-arriving species from the Baltic Sea should be less successful in colonizing the recipient communities at higher temperature as a consequence of niche modification and/or resource depletion by the early-arriving bacteria in the recipient communities. Moreover, we assumed that recipient communities with lake inocula that are geographically closer to the Baltic Sea should contain larger numbers of species adapted to the applied Baltic Sea medium, and thus negatively impact the invasion success of their late-arriving peers.

| Experimental design
In total, our experimental design resulted in 132 communities, three sets of recipient communities, exposed to three levels of dispersal and three temperatures, each with four replicates. For each temperature there was a dispersal source with four replicates and a control with four replicates consisting of cell-free medium ( Figure 1).
For the preparation of the Baltic Sea incubation medium used in this experiment, 120 L seawater was collected at the Swedish Baltic Sea coast on 19 June, 2018, at Barnens Ö (N 59°55′11.9″, E 18°54′52.2″). The water was filtered through a 20 μm net in situ to remove zooplankton and kept in the dark at 4°C overnight. Then, the medium was autoclaved (121°C for 40 minutes) and its pH was adjusted to its original level (pH = 8.18) by HCl addition. Afterwards, the medium was filtered through sterile 0.2 μm 47 mm membrane filters , and distributed into sterile 1,000 ml glass bottles, and autoclaved once more at 121°C for 20 minutes in order to achieve a sterile cell-free incubation medium. Until inoculation, the bottles containing the sterile medium were kept in the dark at 4°C.
For the preparation of the inoculum communities, water samples were collected from three Swedish lakes (Lötsjön -N 59°51′44.0″, E 17°56′37.6″; Erken -N 59°50′09.2″, E 18°37′57.9″; and Grytsjön -N 59°52′21.1″, E 18°52′53.6″) and from the Baltic Sea (same location as above) on 4 July, 2018 ( Figure S1). The distances of the three lakes from the Baltic Sea sampling location were 54.5 km (Lötsjön), 18.3 km (Erken) and 5.6 km (Grytsjön). The chemical characteristics of these lakes were slightly different; the concentrations of total carbon (TOC) and PO 4 3− was higher, while NO 3 − was lower in lakes located closer to the sea coast (Table S1). All samples were F I G U R E 1 Experimental design of the study. The recipient communities were comprised of three different lake inocula (Lötsjön, Erken, or Grytsjön; indicated by cross, square and triangle symbols, resp.) inoculated separately into "foreign" Baltic Sea incubation medium. The three lake inocula differed in their geographical distance from the Baltic Sea, with Grytsjön (triangles) being closest and Lötsjön (crosses) farthest away. The dispersal source constituted of the Baltic Sea community (blue cells) inoculated into cell-free incubation medium. Both the recipient (early-arriving lake bacteria) and the dispersal source (late-arriving Baltic Sea bacteria) communities were incubated at three different temperatures (15, 20 and 25°C). Coalescence (dispersal) events were applied at three rates by replacing 0%, 5% and 20% cells in the recipient communities with cells from the dispersal source (indicated by the different colours of the recipient community symbols). Black and grey arrows represent the direction of the coalescence. The recipient communities (n = 4, for each lake inocula) were always mixed with the corresponding dispersal source replicate (n = 4) at the respective temperature level sequentially filtered to remove bacterial grazers, first through a 20 μm net in situ to remove zooplankton and then through GF/F filters (0.7 μm, Whatman) prior inoculation to remove protozoans.
The dispersal source communities were established by inoculating 100 ml of GF/F filtered Baltic seawater into bottles containing 900 ml of cell-free Baltic Sea medium on day 0. The batch cultures were incubated at three different temperature levels (15, 20 and 25°C) in the dark with four replicates at each incubation temperature. The established dispersal source communities were used in the dispersal treatments and represented the late-arriving species arriving at different rates ( Figure 1).
To create the recipient communities of early-arriving species 50 ml of GF/F filtered lake water inocula was added to bottles containing 450 ml cell-free Baltic Sea medium, and incubated at three different temperature levels (15, 20 and 25°C) in dark with four replicates at each incubation temperature. The incubation of recipient cultures was started with one day delay (day 1) so that cell abundance would most likely be lower compared to the dispersal sources, in order to avoid strong dilution of the medium during the coalescence process.

| Coalescence event
On day 7, after the successful establishment of recipient communities measured as bacterial abundances ( Figure S2), community coalescence was performed by adding the dispersal source communities to the recipient communities. The coalescence event consisted of one dispersal event at three different rates: no, low and high dispersal, wherein 0%, 5% and 20% of the cells were exchanged with cells from the respective dispersal source (Figure 1).
For this, each replicate "A" of the three recipient communities at the different incubation temperatures received cells from replicate "A" of the dispersal source at the respective temperature level.
Likewise, each replicate "B" of the recipient communities received cells from replicate "B" of the dispersal source and so on. The volume that needed to be exchanged in order to apply the 5 or the 20% dispersal rates was calculated based on the measured bacterial abundances in all cultures on day 7. To reach an equal final volume (564 ml) in all cultures the differences were compensated by adding additional cell-free incubation medium that was kept at the same conditions throughout the entire experiment. One additional medium bottle (kept at 20°C) broke during the experiment, hence, a mixture of the two other medium bottles (kept at 15 and 25°C) were used after the dispersal treatments to reach equal volume in each incubation bottle. Both the cell exchange and the supply of additional medium were carried out under sterile conditions using sterile disposable pipettes.
To follow changes in environmental conditions in the cultures, samples for chemical analyses were collected three times: on day 1 after lake inocula were distributed into the medium, after the dispersal treatment (day 7), and on the last day of the experiment (day 22). Total phosphorus (TP), total nitrogen (TN) and total carbon (TOC) were measured spectrophotometrically (Perkin Elmer, Lambda 40, UV/VIS Spectrometer) using the molybdenum-blue method (Menzel, 1965)

| Bacterial community composition
At the end of the experiment (day 22), the cultures (564 ml Sequences were processed using the DADA2 pipeline (Callahan et al., 2016) in R on the server of the Uppsala Multidisciplinary Center for Advanced Computational Science (UPPMAX). First, forward and reverse sequences were trimmed to 280 and 220 bp long, respectively, after quality filtering (truncQ = 2) with maximum expected errors set to 2 and 5 for forward and reverse sequences, respectively. Secondly, sequences were dereplicated and sequence variants were inferred. Finally, chimeric sequences were removed and the final amplicon sequence variants (ASVs) were assigned against SILVA 132 core reference alignment (Quast et al., 2013).

| Data analyses
All statistical analyses and visualizations were conducted in R we assumed that the early-arriving bacteria had enough community resistance (e.g., through priority effects by niche modification and/ or resource depletion) to avoid complete replacement by the better adapted late-arriving bacteria.
To quantify community resistance, we performed a conservative mixing model following Székely and Langenheder (2017) in order to investigate whether the observed invasion success of the late-arriving bacteria is lower or higher than expected based on the applied cell exchange (dispersal) rates. For this, we calculated the expected proportion of ASVs in the coalesced communities based on the ASV proportions in the recipient communities and in the corresponding dispersal source, and the applied cell exchange rates (i.e., 5 or 20%). Thereafter, we calculated the Bray-Curtis dissimilarities between the coalesced communities and the corresponding dispersal sources for the measured and expected data matrices, respectively. The deviation of the measured Bray-Curtis dissimilarity from the calculated expected dissimilarity was multiplied by -1 to express 'invasion success' and tested using paired t-tests. A significant positive deviation (p < .05) indicates that the invasion success of the late-arriving bacteria from the dispersal source was higher than expected. In contrast, a significant negative deviation indicates that the late-arriving bacteria established less successfully than expected, which could be a consequence of priority effects.
Resistance at population level was investigated by determining the relative abundance of early-arriving lake ASVs of the recipient communities that persisted after the coalescence event. We further investigated the invasion success of late-arriving Baltic Sea ASVs of the dispersal source that were mixed into the recipient communities. For this, we first identified ASVs that were unique in either the recipient or the dispersal source communities and fell in the abovementioned categories by performing differential abundance analyses at each temperature level using the "DESeq2" package (Love et al., 2014). First, we selected the unique abundant ASVs (> 0.5% relative abundance) in each recipient community and the dispersal source. Then, we determined separately for each recipient community whether the relative abundances (as a proxy for population size) of the abundant early-arriving ASVs changed after the effective dispersal treatments (i.e., 5% and 20% dispersal rate treatments) compared to their relative abundances in the no dispersal (0%) communities. Here, we interpreted the lack of significant (adjusted p < .05; Benjamini and Hochberg method) negative differences in relative abundances as a sign of persistence, and grouped the cor-

| Community-level patterns
After the initial inoculation of early-arriving bacteria in the Baltic Sea medium, all recipient communities showed typical growth patterns of dilution cultures and increased in abundance at least until day 7 ( Figure   S2). The temperature increase (i.e., 20°C and 25°C) resulted in significantly higher abundances on day 7 compared to the 15°C treatment The recipient communities exposed to dispersal (i.e., 5% and 20% dispersal rate, brown and black symbols on the NMDS plot) became more similar to the dispersal source ( Figure 2; blue circles). However, PERMANOVA results showed that the recipient communities exposed to dispersal were significantly dissimilar from the dispersal source in all cases (Table S3), thus, complete convergence to the dispersal source (i.e., complete absence of resistance or priority effects) did not occur in any of the communities.
The community dissimilarities measured between the coalesced communities and the corresponding dispersal sources were compared to the community dissimilarities calculated for the theoretical coalesced and dispersal communities (estimated based on conservative mixing models where the resulting coalesced community is expected to follow the mixing ratio of the early-and late-arriving communities). The deviation between the measured and theoretically expected dissimilarities was used as an indicator for the invasion success of the late-arriving Baltic Sea bacteria. There was no case where the deviation was significantly negative, meaning that the invasion success by the late-arriving bacteria was either as high as expected in case of conservative mixing, or even greater ( Figure 3). At 5% dispersal, we found that the invasion success was in most cases significantly higher than expected (positive deviation values), while at 20% dispersal, there was no significant deviation (p > .05), indicating that the measured Bray-Curtis dissimilarity between coalesced and dispersal source communities did not differ from those expected.

| Population-level patterns
We further examined changes in the dynamics of early-and latearriving ASVs in response to warming. On a broad taxonomical level, we found that the abundant (> 0.5% relative abundance) bacterial ASVs in the early-arriving communities belonged to the class Alphaproteobacteria, Gammaproteobacteria and Bacteroidia ( Figure   S10). The most abundant genera (top three) were Brevundimonas,

(hereafter A-N-P-R) in the Lötsjön and Erken recipient communities
and Limnobacter, Algoriphagus and A-N-P-R in the Grytsjön recipient communities ( Figure S10). The abundant (> 0.5% relative abundance) members of the dispersal source communities (i.e., late-arriving bacteria) were ASVs belonging to Alphaproteobacteria (mainly
Differential abundance analyses revealed numerous ASVs of the abundant genera (> 0.5%) that could be classified as either "persistent" or "forfeited" early-arriving ASVs, and "successful" or "unsuccessful" late-arriving ASVs (see Section 2). We identified several persistent early-arriving ASVs (belonging to the genera of A-N-P-R, Brevundimonas, Flavobacterium, Limnobacter, Novosphingobium, F I G U R E 2 Nonmetric multidimensional scaling (NMDS) plot derived from abundance-based Bray-Curtis dissimilarities of bacterial community composition by the end of the experiment and facetted at the three temperature levels. Note that cultures with Baltic Sea inoculum were used as the dispersal source (n = 12), while cultures with lake inocula (Grytsjön, Erken and Lötsjön) were used as recipient communities (n = 108). All cultures were grown in Baltic Sea medium. Each symbol represents one replicate, and is shaped and coloured by inoculum origin and dispersal rate, respectively. Goodness of fit ( (15, 20 and 25°C). ASVs are grouped by bacterial genus and were identified by differential abundance analysis (see Section 2 for the assessment procedure and Figure  S6  Linking the population-level patterns ( Figure 5) to the communitylevel patterns (Figure 3) clearly shows that, when 5% dispersal was applied, the late-arriving bacteria were much more successful and established greater populations (higher sum of relative abundances) at the lower temperatures. This phenomenon was particularly pronounced in case of Lötsjön, i.e., the lake situated the furthest away from the Baltic Sea.

| Effect of proximity to dispersal source
At 5% dispersal, warming seemed to restrict the invasion success of late-arriving bacteria from lake inocula (e.g., Lötsjön and Erken) that are geographically further away from the dispersal source (i.e., Baltic Sea) (Figure 3). However, we found no effect of proximity on invasion success at 20% dispersal.
Inoculum origin had no effect on the total relative abundance of the persistent early-arriving ASVs in the recipient communities with different dispersal treatments (two-way ANOVA at 5% dispersal: F Inoculum origin = 1.95, p = .159; 20% dispersal: F Inoculum origin = 0.59, p = .562, no significant interactions in either case) (Figure 4). Nevertheless, it is worth mentioning that most of the forfeited ASVs occurred in Lötsjön (n = 27) and Grytsjön (n = 51) recipient communities, while only two forfeited ASVs (Sediminibacterium grouped as "other_Bacteroidia") occurred in Erken recipient communities.
The total relative abundances of successfully established latearriving bacteria differed between recipient communities (two-way ANOVA at 5% dispersal: F Inoculum origin = 3.83, p = .033; at 20% dispersal: F Inoculum origin = 4.79, p = .015; with no significant interactions in either case) ( Figure 5). Significant differences in total relative abundances (based on post-hoc Tukey tests at p < .05) were found between Erken and Grytsjön at 5% dispersal and Lötsjön and Grytsjön at 20% dispersal. Note that Grytsjön is the closest lake to the dispersal source.

| DISCUSS ION
This study investigated how warming, dispersal rates and geographic proximity to the dispersal source influence the invasiveness of aquatic bacteria during community coalescence. At the community level, we found signs of community persistence because immigration by presumably better adapted bacteria from the dispersal source did not cause complete convergence (community turnover) towards the dispersal source community. However, invasion success always approached or even exceeded the theoretical invasion success of late-arriving bacteria estimated based on our conservative mixing model, suggesting a minor role of priority effects. By investigating population-level patterns, however, it appeared that warming had the potential to restrict the establishment of late-arriving bacteria to some extent. This was in particular the case at low dispersal rate (i.e., 5%) and in recipient communities that received inocula from lakes geographically further apart from the dispersal source (i.e., Baltic Sea). The detected patterns at both community and population level can, thus, vary and depend (a) on the dispersal rate of late-arriving better adapted communities into recipient communities, and (b) on the composition of the recipient community.
F I G U R E 5 Relative abundances of successfully established late-arriving species (amplicon sequence variants [ASVs] > 0.5% relative abundance) in the different dispersal (5% or 20%) and temperature treatments (15, 20 and 25°C). ASVs are grouped by bacterial genus and were identified by differential abundance analysis (see Section 2 for the assessment procedure and Figure S7

| Temperature-dependency of invasion success
In our study we tried to mimic natural dispersal events of bacterial communities which are often complex and involve mixing or coalescence of entire communities (Rillig et al., 2015;Rocca et al., 2020).
This study provides experimental evidence that temperaturedependency of invasion (immigration) success can occur in complex pelagic bacterial communities wherein different bacterial groups are involved in different ways. The warming effect could be seen at both community and population levels. The analyses of patterns at the community level showed that the late-arriving bacteria successfully established populations and increased well above their expected relative abundances at lower temperatures and when dispersal rate was low. However, when more cells (20%) were dispersed, further establishment (exceeding the theoretical relative abundances) was not found, indicating that invasion was restricted. One possible explanation for the lower establishment success of late-arriving bacteria and stronger persistence of earlyarriving species at higher temperatures is that the resistance of recipient communities to invasion (dispersal) by late-arriving bacteria increased because higher temperature stimulated growth of the early-arriving bacteria (see Figure S2 and Table S2). This, in turn, may have resulted in that recipient communities were closer to their carrying capacity when the dispersal (community coalescence) was performed, hence offering less opportunities for establishment.
This resonated with our findings at population level which showed that the total relative abundance of successfully established (latearriving) ASVs generally decreased with increasing temperature, whereas the sum of relative abundance of persistent early-arriving ASVs tended to show the opposite trends, even though this was not significant in any case.
Our experiment did not aim to directly test the presence of priority effects, but still allowed us to make assumptions about the consequences of the applied experimental treatments on the extent of priority effects. Priority effects can be due to two distinct mechanisms: niche-modification and niche preemption (Fukami, 2015), but providing insights into the mechanisms underlying priority effects is difficult. For example, niche modification-driven priority effects of the different lake inocula could have influenced the identities of the successfully established late-arriving bacteria (Fukami, 2015), which was, however, not the case because their identities were similar in each coalesced community ( Figure 5).  Figure   S4- Figure S9), such effects appeared to be generally stronger at 5% dispersal rate when warming occurred. This highlights that both dispersal rates and temperature are important mediators of community resistance, and that this resistance in general is likely to be greater if dispersal rates of late arrivals are relatively low (Loeuille & Leibold, 2008) and at higher temperatures. However, this was not always the case because in cases where the sources of two mixing communities are geographically located close to each other (i.e., Grytsjön -Baltic Sea) we found no clear evidence that warming influenced the invasion success of the late-arriving bacteria.

| Does distance matter?
The lake inocula included for the preparation of the recipient communities in this study differed in their geographic distance to the dispersal source (the late-arriving bacteria from the Baltic Sea).
We therefore presumed that the potentially higher numbers of Baltic Sea bacteria in recipient communities closer to the Baltic Sea (i.e., Grytsjön) would lead to more resistance against invasion.
Our results do, however, not clearly support this idea because the dissimilarities between the coalesced and dispersal source communities were similarly high in all cases (see Figure 2), suggesting that the potential shared species pools with the Baltic Sea, including local seed banks of Baltic Sea taxa, were equal irrespective of the distance of the lake to the Baltic Sea. There were nevertheless differences among lakes regarding the pattern of how temperature affected dispersal-induced shifts in community composition, and the total relative abundance of persistent early-arriving ASVs or successfully established late-arriving ASVs. For example, it seems that warming did not influence the invasibility of recipient communities with inoculum from Grytsjön, the lake located closest to the Baltic Sea. This difference among coalesced communities might also be the consequence of the differences of chemical characteristic of the three lakes from which their inocula originated (Table   S1), or the compositional differences of the initial recipient communities ( Figure S10). Moreover, intrinsic differences in traits (e.g., temperature optima) of ASVs that contribute to the invasibility of the different recipient communities may also be the explanation of the observed differences that, however, we cannot disentangle in our study.

| Inconsistencies in population-level dynamics
We identified several persistent early-arriving ASVs that taxonomically differed between the three sets of recipient commu- would happen in the presence of predation (e.g., bacterial grazers or phages) or multilevel trophic interactions. A few previous studies on zooplankton communities suggested that predation can be an important factor and can either reduce community resistance (Berga et al., 2015;Louette & De Meester, 2007), or, in contrast, indirectly promote it . Nonetheless, we lack a comprehensive knowledge on how predation could affect the invasibility of microbial communities and potential priority effects in those communities. Another aspect that needs to be considered is temperature fluctuation that can promote the invasion success of dispersed species and maintain multiple species coexistence, thus, reducing historical contingency (Litchman, 2010;Toju et al., 2018;Tucker & Fukami, 2014).
In conclusion, temperature has been shown to stimulate microbial invasions (e.g., spread of vibrios; Vezzulli et al., 2012) and influence the biogeographical patterns of microbes (Amalfitano et al.,

2014
). Our experimental study shows the potential of warming to mitigate the effects of invasion during events when communities are mixing. More precisely, higher temperatures changed the identity and increased the sum of relative abundances of persistent bacteria in the recipient communities, and restricted the total relative abundance of successfully established late-arriving bacteria. It is, however, important to note that warming-enhanced resistance does not necessarily challenge invasion in any case as its strength depends on several other factors such as the dispersal (mixing) ratios of communities and their initial compositions. Nevertheless, our findings highlight the potential role of warming in mitigating the effects of invasion at the population level which can impact the biogeographical patterns of microbial communities as a consequence of ongoing climate change.

ACK N OWLED G EM ENTS
We thank Vasiliki Papazachariou for her help and assistance during field sampling and the implementation of the experiment. We The spatial structure of bacterial communities is influenced by historical environmental conditions. Ecology, 95, 1134-1140.