Sampling was carried out in the central-northern section of the Puyuhuapi Fjord (Figure 1A), at a single sampling site where an oceanographic buoy equipped with a YSI 6600 V2-4 multiparameter probe has been continuously recording meteorological and hydrographic data since 2010¹ Puyuhuapi Fjord is ca. 90 km in length, situated between 44° 19′ and 44° 57′ S in Chilean Patagonia, and connected with the coastal ocean through the Jacaf and Moraleda channels to the north and south respectively (Figure 1A). The fjord is characterized by a complex bathymetry, with maximum depths of ca. 400 m and the presence of several sills that restrict the exchange of waters between inner basins and adjacent channels (Figure 1A). Puyuhuapi Fjord receives freshwater from the discharges of the Cisnes and Ventisquero rivers (annual average of 218 and 40 m⁻³ s⁻¹, respectively; Calvete and Sobarzo, 2011), combined with direct precipitation (≥2,000 mm per year), runoff, ice, and snow melting and input of minor rivers and streams. The water column of Puyuhuapi Fjord shows a seasonal stratification/mixing cycle varying from highly stratified during spring and summer to partially mixed in winter (Schneider et al., 2014).

Water samples were collected monthly from June 2013 to September 2014 using Niskin bottles deployed from a boat, both from near surface water (2 m) and from 20 m depth (Table 1). Additionally, in July 2015 the water column was sampled every second day over 7 days at 5 standard depths between 1 and 100 m (Table 1). Water samples were stored in the dark at in situ temperature in acid-cleaned containers prior to processing. Hydrographic data were recorded during the 2013-2014 period using a CTD Ocean Seven 304 (IDRONAUT, Italy) and in winter 2015 using a CTD SeaBird 25. Oxygen concentrations in surface and subsurface waters were determined using the Winkler method (Strickland and Parsons, 1972).

Within 4 h after collection, 1 L aliquots of water samples from the Niskin bottles were filtered through 0.22 μm sterile membrane filters (Millipore) and stored at −20°C prior to processing for DNA analysis. Samples for determination of chlorophyll-a and inorganic nutrients were collected by filtration of 1 L seawater through pre-combusted (4 h at 450°C) 0.7 μm glass fiber filters (Whatman GF/F). Chlorophyll-a was analyzed using the fluorometric method of Parsons et al. (1984), and inorganic nutrients by spectrophotometry (Solorzano, 1969; Strickland and Parsons, 1972).

DNA on filters was extracted using a PowerWater® DNA Isolation Kit and cleaned using a Power Clean® DNA Clean-up Kit (MOBIO Laboratories). Prokaryotic (Bacteria plus Archaea) 16S rRNA genes were amplified using primer set 515F (5′-GTGCCAGCMGCCGCGGTAA-3′) and 806R (5′-GGACTACHVGGGTWTCTAAT-3′). Amplification and sequencing on an Illumina MiSeq platform were conducted in a commercial laboratory (Research and Testing Laboratory, Lubbock, TX, USA).

The full MiqSeq data set is available at the National Center for Biotechnology Information Sequence Read Archive under the accession numbers SRR5155509 to SRR5155558. Paired Illumina reads were processed using QIIME software package version 1.9.1 (Caporaso et al., 2010). Default QIIME parameters (r = 3, p = 0.75, q = 3, n = 0) were adopted for quality filtration as recommended by Bokulich et al. (2013). Removal of potential chimeras was carried out using VSEARCH software version 1.10.2 (VSEARCH GitHub repository) and Uclust was used to cluster Operational Taxonomic Units (OTUs) at a 97% similarity threshold. Representative sequences from each OTU were classified by comparison with the Greengenes database (DeSantis et al., 2006).

Analysis of beta and alpha diversity (Chao1) was carried out after removal of sequences identified as Chloroplast and after resampling with the rarefaction method using the minimum number of sequences per sample (13036). Similarity was estimated at the OTU level based on the Bray-Curtis distance matrix index and then used as input to carry out Non-Metric Multidimensional Scaling (NMDS) ordination analysis in R version 3.1.2 using the package vegan (Oksanen et al., 2013). Environmental variables were fitted onto NMDS plot using the function envifit of the package vegan. Statistical differences in the community composition between water masses were tested by PERMANOVA analysis in R. OTU heatmaps were produced using filtered OTU table to identify the contribution of the main representative individual OTUs (representative OTUs defined as those containing more than 1,000 sequences) to the overall community composition of each sample. Rarefaction curves for each prokaryotic community were generated from the means of 10 randomized data sets in Qiime. The Mann-Whitney non-parametric U-test was applied to test for statistical differences among depths (surface: 1–5 m n = 22, subsurface: 8–20 m n = 19 and deep: >50 m n = 7) for environmental variables and community diversity, and the Kruskal-Wallis non-parametric test for differences in OTU abundance among water masses, between cold (June-September, n = 19) and warm (October-May, n = 7) periods, and differences in winter temperatures between 2010 and 2016. The Spearman correlation index was calculated to analyze associations between abundance of representative OTUs and environmental variables.

Vertical sections of salinity in the water column (Figures 1B,C) showed variable contributions of Estuarine Fresh Water (EFW, 11–21 psu) and Estuarine-Salty Water (ESW, 21–31 psu) within the top 5 m. Between 5 and 30 m depth, ESW dominated during winter, whereas the influence of Modified Sub-Antarctic Waters (MSAAW, 31–33 psu) increased from austral spring 2013 and autumn 2014 (Figure 1B). In July 2015 (austral winter), MSAAW was distributed from ~30 to ~100 m depth and Sub-Antarctic Water (SAAW, >33 psu) was present in waters below 100 m (Figure 1C). Temperature distribution in the top 30 m was characterized by a strong seasonality (Figures 1E,D), with cold waters (<10°C) in surface and subsurface layers during austral winter followed by an increase in temperature (>15°C) in the top 10 m during austral summer (December 2013–March 2014; Figure 1D). In austral winter, low temperatures coincided with periods of increased river discharge and precipitation (Figures 1D,H). Oxygen concentration was higher in surface (ca. 7 to 10 mL L⁻¹) than subsurface waters (3 to 7 mL L⁻¹, Figure 1F), where oxygen concentrations lower than 5 mL L⁻¹ were observed between December 2013 and April 2014 coinciding with the presence of higher salinity MSAAW (Figure 1F). During July 2015, the presence of hypoxic waters (<2 mL L⁻¹) was observed below 120 m (Figure 1G).

Nitrate (1 to ~20 μM) and phosphate (~0 to 4.3 μM) concentrations were lower in surface (1–5 m) than subsurface and deeper waters during most of the study period (Table 1). Silicic acid concentrations were highly variable (<1 to >100 μM) with maximum values detected in austral winter in surface waters, and during winter and spring in subsurface waters (Table 1). Ammonia concentration ranged from 0.2 to 1.5 μM with maximum values in spring and summer (October–December; Table 1). Chlorophyll-a concentrations varied from ~1 to >25 μg L⁻¹ in surface waters, with maximal values observed in autumn, winter and spring (Table 1). There were significant differences in nitrate, phosphate and chlorophyll-a concentrations between surface and subsurface waters, and in nitrate, phosphate, and ammonia between surface and deep waters (Table 2).

Prokaryotic diversity was examined by comparing the distribution of 3250730 16S rRNA gene sequences of Bacteria and Archaea (Table 3), which corresponded to 13807 OTUs assigned to Bacteria and 413 to Archaea after removal of chimera and chloroplast sequences. Rarefaction curves (Supplementary Figure 1) showed that all samples had reached the curvilinear phase of sampling effort, and that most had started to plateau. Of a total of 9376 different OTUs identified after resampling, 11% of these were common to surface, subsurface and deep waters, with the highest percentage of shared OTUs (28%) common to surface and subsurface waters (Figure 2). OTU richness estimated by Chao1 ranged from 623 to 2,914 in surface waters, between 962 and 1,798 in subsurface waters and from 481 to 1,270 in deep waters (Figure 3). Chao1 values estimated for surface and subsurface waters were significantly different to those estimated for deep waters (Table 2). The highest values of Chao1 (>2,000) in surface waters were observed during winter 2013, 2014, and 2015, whereas the lowest values of OTU richness were found in summer 2014 coinciding with the seasonal increase in surface water temperature (Figure 3A). Significant differences (Mann-Whitney U = 23, p = 0.03) were observed for Chao1 values for surface waters between cold (June-September) and warm (October-May) periods (duration of cold and warm periods was defined according to short-term monthly average temperature reported by Schneider et al., 2014). In subsurface waters, values of Chao1 were less variable, with lower values (<1,500) found in summer and coinciding with increases in salinity and temperature and a decrease in oxygen concentration (Figures 1, 3B). In July 2015, with the exception of July 14 when extremely low values were observed in the top 5 m, vertical distribution of Chao1 was characterized by high values in surface waters and a decrease with depth (Figures 3C–F). Chao1 was negatively correlated with temperature (Spearman R = −0.53, p < 0.001), positively correlated with oxygen (Spearman R = 0.40, p = 0.005) and showed a weak correlation with ammonia concentration (Spearman R = 0.30, p = 0.035; Table 4).

At phylum and class levels, the composition of prokaryotic assemblages in surface waters was characterized by predominance of classes Flavobacteriales (Bacteroidetes) and Alpha- and Gammaproteobacteria, which together accounted for up to 90% of total sequences during Austral summer (Figure 3A). Subsurface waters were characterized by Thaumarchaeota which accounted for 5 to 40% of prokaryotic sequences, and by Gammaproteobacteria and Bacteroidetes which accounted for 18 to ~45% and up to 30% of total sequences, respectively (Figure 3B). With a relatively low proportion of total sequences, Verrumicrobia, Cyanobacteria (mostly picocyanobacteria of order Synechococcales), Actinobacteria, and Betaproteobacteria were more abundant in surface than in subsurface waters, whereas Planctomycetes and Deltaproteobacteria were more frequently detected in subsurface waters (Figures 3A,B). In July 2015, with the exception of July 14, prokaryotic composition in the top 20 m was similar to that observed in the 2013-2014 time series, whereas in deep waters OTUs of Thaumarchaeota and Gammaproteobacteria predominated and sequences of Epsilonproteobacteria increased their relative abundance (Figure 3C). Relative abundances of Thaumarchaeota, Thermoplasmata, Planctomycetes, and Deltaproteobacteria were significantly and positively correlated with salinity, nitrate and phosphate, and Bacteroidetes and Alpha- and Betaproteobacteria positively correlated with oxygen concentration (Table 4). On the contrary, significant negative correlations were observed for relative abundance of Thaumarchaeota, Planctomycetes, and Delta- and Gammaproteobacteria with oxygen, Bacteroidetes, Alpha- and Betaproteobacteria with salinity, and Thermoplasmata and Deltaproteobacteria with chlorophyll-a (Table 4).

At the OTU level, non-metric multidimensional scaling (NMDS, based on Bray–Curtis distance matrix) evidenced variations in composition of prokaryotic communities among water masses, with distinguishable patterns of distribution for samples from EFW and MSAAW, whilst communities from ESW overlapped those from EFW and MSAAW (Figure 4). Segregation of prokaryotic communities amongst these waters masses was mostly explained by variations in salinity and oxygen concentration, whilst temperature and ammonia concentration were likely responsible for variations of composition between cold and warmer periods, particularly in EFW (Figure 4). PERMANOVA analysis showed significant differences between compositions of prokaryotic communities among water masses (p-value < 0.001, Pseudo-F = 5.73) and cold and warm periods (p-value < 0.001, Pseudo-F = 5.71). Among OTUs with more than 1000 counts (Figure 5, Table 5), 62% showed significant differences in their abundance (Kruskal-Wallis, p < 0.05) between EFW and MSAAW, 47% between ESW and MSAAW, 42% between EFW and ESW, and ~33% between EFW, ESW and MSAAW. Based upon the distribution of OTU abundances among samples, and upon correlations with principal explanatory environmental variables, we identified taxa that were representative of the two water masses (Figure 5, Table 5).

At higher taxonomic levels, temporal variability of community composition in surface waters was mostly characterized by opposing changes in the relative abundance of Bacteroidetes and Gammaproteobacteria (Figure 3A). These changes coincided with an increase in temperature between late austral winter 2013 (September) and summer 2014 (March) and a decrease in OTU richness (Figure 3A). Additionally, sequences of Cyanobacteria were more abundant during winter and spring 2014 and Verrumicrobia increased sharply during spring 2013 and autumn 2014 (Figure 3A). In subsurface waters, the proportion of Archaea increased sharply in spring 2013, and between autumn and winter 2014 (Figure 3B), coinciding with increases in salinity (Figure 1B).

At the OTU level, temporal variations were evidenced by loss of richness and decreased abundance in more than 65% of the representative OTUs of surface waters during austral summer (December 2013-March 2014; Figure 6, Table 5). Compared to surface waters, a low proportion of representative OTUs varied throughout the year in subsurface waters (Figure 7). The majority of OTUs that exhibited temporal changes were negatively correlated with temperature and ammonia concentration and positively with silicic acid concentration (Figure 6). Significant correlations were also detected between OTUs showing seasonal variability and abundance of diatoms and dinoflagellates (Supplementary Table 1) in surface and subsurface waters (Figures 6, 7). Significant differences (p < 0.05) were observed for OTU abundances between cold (June-September) and warm periods (October-May) in ~50% of representative OTUs of surface waters (Figure 6).

The hydrographic structure of the water column observed during the study period was consistent with the estuarine circulation and water mass distribution previously described for Patagonian fjords generally (Silva et al., 1998; Sievers and Silva, 2008), and more specifically for Puyuhuapi Fjord (Schneider et al., 2014). Saline stratification of the water column of Puyuhuapi Fjord resulted in a vertical segregation of dissolved inorganic nutrients and chlorophyll-a concentrations, consistent with previous observations for a variety of biogeochemical parameters in other Patagonian fjords (González et al., 2013; Silva and Vargas, 2014; Gutiérrez et al., 2015; Ríos et al., 2016).

Consistent with this stratification of the water column, the composition of the prokaryotic community showed a discernible pattern of vertical zonation at higher taxonomic levels, with members of Bacteroidetes and Alpha- and Gammaproteobacteria dominating surface waters, and the presence of abundant sequences of Archaea in subsurface and deep marine waters (Figure 3). Our observations are consistent with reports showing predominance of the same taxa of bacteria in large brackish environments (Riemann et al., 2008), in estuaries and coastal saline gradients (Campbell et al., 2011; Herlemann et al., 2011; Fortunato et al., 2012; Fortunato and Crump, 2015), and in polar surface waters (Zeng et al., 2009; Teske et al., 2011; Prasad et al., 2014; Signori et al., 2014; Piquet et al., 2015) and Patagonian glacial fjords (Gutiérrez et al., 2015). For archaeal OTUs, predominance of members of the phylum Thaumarchaeota (including Marine group I Archaea) is consistent with this group being a major contributor to microbial diversity in estuarine waters and sediments (Webster et al., 2015; Xia et al., 2015), in polar waters (Bano et al., 2004; Galand et al., 2008, 2009a,b), in sea ice (Collins et al., 2010; Cowie et al., 2011) and in fjord environments (Zaikova et al., 2010; Gutiérrez et al., 2015). Regarding Thermoplasmata, most sequences were affiliated to the uncultured Marine group II, which represents a significant fraction of the archaeal community in estuarine waters (Xia et al., 2015) and coastal waters of the Arctic Ocean (Galand et al., 2008, 2009a), but in fjords appear to be poorly represented (Zaikova et al., 2010). The current understanding on Marine group II is rather limited compared with information presently available on Thaumarchaeota (Zhang et al., 2015), and our study presents valuable new information on distribution of this group in fjord environments where few data are available. Among other taxa, presence of Actinobacteria and Verrumicrobia is consistent with their significant contribution to microbial communities in estuarine environments (Riemann et al., 2008; Campbell and Kirchman, 2013; Fortunato et al., 2013; Fortunato and Crump, 2015). The observation of elevated abundance of picocyanobacteria is also consistent with their contribution to microbial diversity, phytoplankton biomass and primary productivity in estuarine and fjord environments (Stal et al., 1999; Hajdu et al., 2007; Piquet et al., 2014; Xia et al., 2015), including Patagonian fjords (Gutiérrez et al., 2015). Members of the phylum Planctomycetes and the classes Delta- and Epsilonproteobacteria were poorly represented and mainly restricted to more saline waters. Although these taxa make a low contribution to overall bacterial abundance (Zinger et al., 2011; Yilmaz et al., 2016), they do appear to be ubiquitous in marine environments and can play important roles in biogeochemical cycles of carbon, nitrogen and sulfur (Campbell et al., 2006; Fuerst and Sagulenko, 2011; Yilmaz et al., 2016).

Salinity and oxygen concentration exerted major controls on prokaryotic community composition among the contrasting water masses identified during the present study in Puyuhuapi Fjord (Figure 4). Salinity is known to be a major factor influencing global biogeography of microorganisms (Lozupone and Knight, 2007; Auguet et al., 2012) and the spatial variability of microbial communities along coastal ecotones (Herlemann et al., 2011; Campbell and Kirchman, 2013; Fortunato and Crump, 2015), including polar and fjord waters (Signori et al., 2014; Gutiérrez et al., 2015). Dissolved oxygen is also known to have a major influence on microbial diversity and activity in marine environments exhibiting strong vertical oxygen gradients (Stevens and Ulloa, 2008; Wright et al., 2012). In some brackish environments, where bottom topography can restrict water circulation and lead to hypoxia in deep waters, as suggested for Puyuhuapi Fjord (Schneider et al., 2014; Silva and Vargas, 2014), oxygen can represent a major factor in the structuring of microbial communities (Zaikova et al., 2010). Segregation of microbial communities among water masses in Puyuhuapi Fjord is consistent with studies showing identifiable microbial signatures in water masses of the North Atlantic (Teira et al., 2006; Varela et al., 2008), Arctic Ocean (Galand et al., 2009a,b, 2010; Hamdan et al., 2013), and in proglacial fjords of Patagonia (Gutiérrez et al., 2015). An exception to this pattern of distribution was the microbial community associated with Estuarine Saline Waters (ESW), which consisted of a combination of OTUs representative of Estuarine Fresh Waters (EFW) and Modified Sub Antarctic Waters (MSAAW). Since the microbial community within ESW did not show a distinctive microbial signature, we hypothesize that this community is composed of cosmopolitan prokaryotes adapted to a wide range of salinity and oxygen concentrations found in EFW and MSAAW.

The prokaryotic community identified in EFW (Figure 5, Table 5) was characterized by predominance of OTUs matching Candidatus Aquilina sp. (Actinomycetales) and members of the families Rhodobacteraceae, Cryomorphaceae (Alphaproteobacteria), and Flavobacteriaceae (Bacteroidetes). We suggest that the predominant physicochemical conditions of surface waters of Puyuhuapi fjord, driven mainly by riverine input and meltwater discharges, confer a distinctive microbial signature to EFW. This suggestion is supported by negative correlations between the abundance of representative taxa of EFW, and salinity, temperature, nitrate and phosphate, and positive correlations with concentrations of oxygen and silicic acid (Figure 5). Moreover, some of the representative OTUs of EFW, such as Fluviicola sp. and Candidatus Aquiluna sp., corresponded to taxa considered to be characteristic of freshwater, and of coastal environments strongly influenced by freshwater, including polar fjords (O'Sullivan et al., 2005; Kang et al., 2012). Our results for the microbial community within EFW also showed that OTUs of Rhodobacteraceae and Flavobacteriales were positively correlated with chlorophyll-a concentration. This is consistent with these taxa being considered to be specialized degraders of photosynthetic organic matter (e g., Roseobacter sp.; Buchan et al., 2014).

In saline and less oxygenated MSAAW the most abundant taxa (Figure 5, Table 5) corresponded to OTUs identified as Nitrosopumilus sp. (Cenarchaeaceae; ~21% of representative OTUs), the order Oceanospirillales (~16%) and the families Flavobariaceae and Piscirickettsiaceaea (~9% each). The order Nitrosopumilales corresponds to a group of ammonia-oxidizing Archaea widely distributed in the ocean, and playing an important role in nitrogen cycling (Offre et al., 2013), including the potential for production of the greenhouse gas nitrous oxide (Löscher et al., 2012). The high abundance of Nitrosopumilus sp. (up to 30% of total sequences in subsurface waters), suggests that microorganisms present in MSAAW contribute significantly to nitrification and nitrogen cycling in Puyuhuapi Fjord. Among other representative taxa, members of Oceanospirillales appear to play a role in degradation of hydrocarbons (Mason et al., 2012; Lamendella et al., 2014), Flavobacteriales are considered important degraders of phytoplankton detritus (Buchan et al., 2014) and Piscirickettsiaceaea include members recognized as fish pathogens and responsible for major losses in salmon farming industry (Rozas and Enríquez, 2014). The presence of this family is an important observation in the context of Puyuhuapi Fjord having a significant fraction of salmon farms in the Chilean Patagonia. The majority of representative taxa of MSAAW were significantly and positively correlated with salinity, dissolved inorganic nitrate and phosphate, and negatively correlated with dissolved oxygen. These observations are consistent with a marine source for microbial communities present in deeper layers of Puyuhuapi Fjord.

We recognize limitations of our study in terms of whether a single point sampling site is representative of the fjord as a whole. However, our data clearly demonstrate the association between distinct prokaryotic OTUs and predominant water masses, which support that our sampling strategy (two-layer sampling during 2013-2014 together with a high frequency vertical profiling in winter 2015) was able to capture most of variability of microbial communities present in waters of the fjord. These water masses have been described as representative of the majority of the Patagonian fjords (e g., Sievers and Silva, 2008; Pérez-Santos et al., 2014) and Puyuhuapi fjord particularly (Schneider et al., 2014). Considering that a parcel of water is supposed to have a relatively stable microbial composition under consistent conditions in the surrounding environment (Fuhrman et al., 2015), we argue that our observations on microbial diversity at the single station are at least highly representative of the Puyuhuapi fjord.

Increase in temperature of surface waters during austral summer resulted in a dramatic reduction of Chao1 and in the abundance of more than 65% of the representative OTUs (Figures 3, 6). This relationship was further supported by a strong and significant negative correlation between OTU richness and temperature in surface waters (Spearman R = −0.71, p < 0.001). Temperature is one of the major drivers of microbial diversity along latitudinal gradients in the oceans (Fuhrman et al., 2008) and can be responsible for spatial and temporal variations of community structure in estuarine (Campbell et al., 2009; Fortunato and Crump, 2011) and in large brackish environments (Andersson et al., 2010; Herlemann et al., 2016). Phytoplankton community composition also appears to influence composition and succession of microorganisms in Puyuhuapi Fjord, which is evidenced by positive correlations of abundance of specific members of Flavobacteriaceae, Alteromonadales, and Verrucomicrobiales with diatoms in subsurface waters and of Flavobacteriales (Cryomorphaceae and Flavobacteriaceae), Rhodobacteraceae, and Pelagibacteraceae with dinoflagellates in surface waters (Figures 6, 7). We suggest that phytoplankton composition could define specific niches for microorganisms in Puyuhuapi fjord waters. The role of phytoplankton bloom dynamics on the composition of prokaryotic community has been previously described (Pinhassi and Hagström, 2000; Dang and Lovell, 2016) with particular reference to specialized taxa that can efficiently degrade phytoplankton detritus (Teeling et al., 2012; Buchan et al., 2014; Dang and Lovell, 2016). Our observations support previous findings in coastal environments where certain bacterial assemblages have been associated with spring diatom blooms and others with autumn blooms dominated by dinoflagellates (El-Swais et al., 2015). Diatoms and dinoflagellates are the major primary producers in Puyuhuapi Fjord (Montero et al., 2017a,b), and specific interactions between prokaryotes and phytoplankton and their role in carbon fluxes certainly merits further analysis.

Several environmental parameters can influence marine microbial diversity on seasonal scales (Fuhrman et al., 2006, 2015; Andersson et al., 2010; Gilbert et al., 2012; Fortunato et al., 2013; El-Swais et al., 2015; Bryant et al., 2016; Bunse and Pinhassi, 2017) and in Puyuhuapi Fjord, temperature and salinity do show a strong seasonal variability (Schneider et al., 2014). Whereas, no significant association was observed between salinity and OTU richness (Spearman R = 0.25, p = 0.29) in the present study, temperature was show to be the main factor explaining variations of microbial richness in surface waters of Puyuhuapi Fjord. The negative relationship between OTUs richness and temperature was described by a power model that explained 70% of variability of Chao1 as function of temperature and showed a major reduction associated with changes in temperature during the cold period in austral winter (Figure 8). The continuous record of surface water temperature within Puyuhuapi Fjord has shown a significant successive increase in average winter temperatures for the last 6 years (Supplementary Figure 2), suggesting a potential effect for future temperature increases on cold-adapted microbes of Patagonian fjords. Among the taxa shown to be more sensitive to increased temperature were members of the bacterial families Flavobacteriaceae and Rhodobacteraceae, and Archaea of the family Cenarchaeaceae (Figure 6, Table 5). Although richness (and composition) recovered after the warm period in our time series (Figures 3A, 6), suggesting a certain degree of resilience, we hypothesize that the disturbance associated with increasing winter temperatures could potentially result in net loss of microbial diversity in the fjord.

In aquatic environments, microbial activity and diversity are considered to be closely linked to overall ecosystem function (Finlay et al., 1997), and changes in diversity have been positively associated with organic matter availability and heterotrophic activity (Landa et al., 2016). Our data indicate a minor proportion of OTUs might be replaced during warm periods, and so we need to look into potential loss of functions in surface water microbial communities with temperature increase. The influence of long-term changes in temperature on the diversity of microbial communities is one of the possible responses of fjord ecosystems to climate variability, although the mechanism behind this is not presently understood. At the regional scale, similar trends in winter temperatures in recent years have been observed in satellite-derived sea surface temperature (off Chiloé Island, ca. 150 km north of our study area; Narváez, unpublished data). These trends coincide with a positive Pacific Decadal Oscillation (PDO³) phase, and a strong El Niño event in 2015⁴. This suggests that the warming winter trend observed in Puyuhuapi Fjord could be driven by large-scale forcing, thus supporting the notion of synchrony between open ocean and coastal-fjords environments, and the potential remote connection between climatic and oceanographic processes and fjord microbes.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.