Soils associated to different tree communities do not elicit predictable responses in lake bacterial community structure and function

ABSTRACT

Freshwater bacterioplankton communities are influenced by the inputs of material and bacteria from the surrounding landscape, yet few studies have investigated how different terrestrial inputs affect bacterioplankton. We examined whether the addition of soils collected under various tree species combinations differentially influences lake bacterial communities. Lake water was incubated for 6 days following addition of five different soils. We assessed the taxonomic composition (16S rRNA gene sequencing) and metabolic activity (Biolog Ecoplates) of lake bacteria with and without soil addition, and compared these to initial soil communities. Soil bacterial assemblages showed a strong influence of tree composition, but such community differences were not reflected in the structure of lake communities that developed during the experiment. Bacterial taxa showing the largest abundance increases during incubation were initially present in both lake water and across most soils, and were related to Cytophagales, Burkholderiales and Rhizobiales. No clear metabolic profiles based on inoculum source were found, yet soil-amended communities used 60% more substrate than non-inoculated communities. Overall, we show that terrestrial inputs influence aquatic communities by stimulating the growth and activity of certain ubiquitous taxa distributed across the terrestrial-aquatic continuum, yet different forest soils did not cause predictable changes in lake bacterioplankton assemblages.

bacterial communities, freshwater ecosystems, terrestrial inputs, Biolog Ecoplates, illumina sequencing, microbial dispersal; IDENT

INTRODUCTION

The presumed high dispersal potential of microorganisms has led to the idea that local environmental conditions structure microbial communities through selection of species from a microbial pool that could theoretically be present (or transported) anywhere (Womack, Bohannan and Green 2010; Caporaso et al.2012; Gibbons et al.2013; Comte et al.2014; Ruiz-González et al.2017a). Studies focusing on microbial community dynamics and assembly, however, are often restricted to single types of ecosystems (only lakes, only soils), thus disregarding the potential linkages between communities from different habitats, which may play important roles in the structuring of local assemblages. Freshwater ecosystems are a good example of this as they are integrated in a terrestrial landscape that continuously exports soil-derived material and microorganisms to the water. Although several studies have shown how these tight land-water linkages may influence downstream aquatic microbial communities in terms of function and composition (Besemer et al.2013; Berggren and del Giorgio 2015; Savio et al.2015; Wilhelm et al.2015; Niño-García, Ruiz-González and del Giorgio 2016a), it has only recently been demonstrated that freshwater bacterioplankton communities are numerically dominated by a few taxa present in the soils, which, once inoculated into the water, can gradually increase their abundances during transit through aquatic networks (Crump, Amaral-Zettler and Kling 2012; Ruiz-González, Niño-García and del Giorgio 2015a). This suggests that the aquatic recruitment, and subsequent selection of bacteria able to persist in dry soils and sediments (Fazi et al.2008) is a major process structuring freshwater bacterioplankton communities. A question that remains unresolved, however, is to what extent differences in these terrestrial inputs of bacteria (and the associated soil material) may lead to selection of taxonomically and/or functionally distinct bacterioplankton communities in the receiving ecosystems.

Soil bacterial communities are known to be highly diverse and heterogeneous over small spatial scales, in part due to the large microscale environmental heterogeneity that exists in soils (Torsvik, Sorheim and Goksoyr 1996; Philippot et al.2009; Vik et al. 2013; Vos et al.2013; Shen et al.2015). Compared to other ecosystems, soils seem to harbor a greater fraction of inactive or dormant bacteria, which can potentially reactivate when soil conditions become favorable (e.g. Lennon and Jones 2011; Placella, Brodie and Firestone 2012; Aanderud et al.2015), but also when transported to different habitats such as freshwater ecosystems (Crump, Amaral-Zettler and Kling 2012; Ruiz-González, Niño-García and del Giorgio 2015a; 2017a). There is some evidence that the pool of soil bacteria that can be resuscitated upon changing conditions may be site-specific, suggesting that the taxa responding to particular changes in conditions will differ among different soils (e.g., Aanderud et al.2015). However, whether this also applies to the soil taxa that, once transported to the water, have the potential to grow and dominate aquatic ecosystems (Ruiz-González, Niño-García and del Giorgio 2015a) remains unknown. Experimental studies assessing the colonization of sterile aquatic media by airborne, sediment or pelagic bacteria have shown that different inocula can lead to taxonomically different communities even under identical conditions (Langenheder, Lindström and Tranvik 2006; Comte et al.2014), as well as that the same inoculum can generate similar communities despite being exposed to different environments (Lee et al.2013). However, the importance of the inoculum is less clear in non-sterilized ecosystems where immigrants must compete with resident communities (Jones and McMahon 2009; Lee et al.2013; Székely, Berga and Langenheder 2013). In general, most of these studies find that the observed responses are largely driven by the growth of generalist or widely distributed species with opportunistic growth strategies, which also seems to be the case for the soil taxa capable to recruit in aquatic communities (Ruiz-González, Niño-García and del Giorgio 2015a, 2017a). Thus, if the pool of terrestrial bacteria that can thrive in aquatic conditions is comprised of generalist taxa present in different soils, one might expect a weak influence of the nature of the terrestrial inputs on the ultimate structure and function of the bacterioplankton communities developing in aquatic systems. In contrast, if different types of soil harbor different pools of potentially aquatic bacteria, changes in the terrestrial sources will significantly impact the receiving freshwater communities.

The diversity of tree species can influence the taxonomic composition and activity of soil bacteria by altering the soil physico-chemical properties; also probably through inoculation of bacteria from falling leaves or the rain-washed tree canopy, since phyllosphere bacterial assemblages are host-specific (Kim et al.2012; Kembel et al.2014; Rivest et al.2015). For example, soil microbial communities differ between conifer and broad-leaved forests (Ushio et al.2008; Jiang et al.2012), and different tree species have been shown to distinctly affect the biomass, structure, and functional responses of the soil microbial communities (Jiang et al.2012; Orwin and Wardle 2005; Rivest et al.2015). We could thus expect that lakes surrounded by different tree species types, particularly between conifer vs. broad-leaved, will be subjected to immigration of specific bacterial communities, yet so far no study has explored the compositional responses of freshwater bacterioplankton communities to inputs of soils associated to different tree species.

Besides these direct bacterial inputs, soils also export large amounts of dissolved organic matter (DOM) and nutrients that will likely differ depending on tree species types (e.g., Lennon and Pfaff 2005). Thus, differences in tree species composition in the surrounding terrestrial landscape may trigger changes in taxonomic composition and functioning of aquatic bacterial communities, not via bacterial inoculation, but by in-lake species sorting based on the relative delivery rates of different substrates or via physico-chemical environmental modification (Lennon and Pfaff 2005; Judd, Crump and Kling 2006; Berggren, Laudon and Jansson 2007, 2010; Ruiz-González et al.2015b; Berggren and del Giorgio 2015). Disentangling these complex linkages is crucial to improving knowledge of freshwater bacterioplankton community assembly and dynamics, as well as for a better understanding on how changes in the surrounding watershed or in run-off regimes may influence the nature of terrestrial inputs and their impacts on aquatic communities.

Here, we present an experimental study designed to assess the influence of different soil biota and physico-chemical properties as affected by differing tree species composition on the establishment and functioning of lake bacterioplankton communities. Our experiment benefitted from the International Diversity Experiment Network with Trees (IDENT) site located in Southern Québec, which established several small patches of high-density tree communities of varying tree species composition to study how tree species diversity affects ecosystem functioning. After only 5 years, many studies have reported significant changes in the soil biota and physico-chemical properties (Rivest et al.2015; Khlifa et al.2017). We collected five different soils plugs associated to different tree species or pairs of species and incubated them in non-sterilized water from a boreal lake for 6 days, mimicking soil derived inputs to lakes. Since all tree species were initially planted on the same soil type, any differences in soil bacterial communities will be due to the direct or indirect influences of the tree species that developed in the patch. By comparing the resulting bacterioplankton communities with those of the un-inoculated lake water at the end of the experiment, we were able to explore the influence of different terrestrial sources on the taxonomic structure (Illumina sequencing of the 16S rRNA gene) and functioning (use of different carbon substrates) of the newly developed bacterioplankton assemblages. Our experimental design did not allow for the differentiation between the actual recruitment of bacteria from soils and the indirect changes in lake communities caused by amendments of different nutrient or DOM inputs. However, both bacteria and DOM are naturally imported together from soils, and our focus was on establishing whether differences in the soil sources are translated into predictably different lake bacterial communities.

MATERIALS AND METHODS

Experimental design

The experiment consisted of 24 1L microcosms to which 0.65 g of soil collected from terrestrial plots was added to lake water, to assess whether the inoculation of different soils could trigger the development of taxonomically and functionally different lake bacterioplankton communities. The microcosms were sampled at the start (time 0) and after 6 days, a time period based on preliminary incubation tests (see “Establishment of the incubation period” section). The water used as medium in the microcosms was collected from a small oligotrophic lake (Lake Croche) located in the temperate Laurentian region 60 km north of Montréal; soils were collected from the Montreal IDENT long-term experimental site, also located within the same general region.

The general characteristics of Lake Croche (45.99°N, 74.00°W) are presented in Vachon and del Giorgio (2014). Briefly, it is a small (18.1 ha) and relatively deep (mean depth 6 m) headwater lake surrounded by a pristine watershed dominated by maple (Acer saccharum) and yellow birch (Betula alleghaniensis). The lake thermally stratifies from late June through September, and is covered with ice from October to May. The mean annual surface water temperature is 15º ranging from 2ºC in January to 24ºC in July. Its mean water retention time is 1.1 year, and the lake has relatively low and stable DOC concentrations (ca. 4–5 mg C L⁻¹). Lake water was collected once in July 2014 at 0.5 m depth in the middle of the lake, and transported to the lab in acid-rinsed carboys.

Soil samples were collected on the same day at the Montreal IDENT tree-diversity experimental site located on the Macdonald College Campus of McGill University (Sainte-Anne-de-Bellevue, Quebec, 45.5ºN, 73.9ºW). IDENT was established in 2009 on a former agricultural field, and currently ca. 10 000 trees consisting of 12 North American temperate forest species are represented. Saplings were planted at 50 cm intervals, and each plot contained 64 individuals in different species and functional combinations (for details on the soil characteristics and the overall design of the plantation see Tobner et al.2014 and Rivest et al.2015). For our study, we collected soil from five different plots containing monocultures or pairs of deciduous and conifer trees (Table 1) as follows: (1) Acer saccharum [Ac], (2) Larix laricina [La], (3) Pinus strobus [Pi], (4) A. saccharum + L. laricina [LaAc] and (5) L. laricina+ P. strobus [LaPi]. In each plot, soil samples from three randomly chosen locations were collected from the top layer (0–10 cm) of the soil (leaf litter was brushed aside), and were kept in the dark for a maximum of 4 hours, until the start of incubations.

Table 1.

Characteristics of the studied samples or treatments. Treatment description, average prokaryote abundances (Prok.) measured at the beginning (T₀) and at the end (T₆) of the experiment, mean number of OTUs, chao1 richness index and taxonomic evenness (Pielou’s index). Values are means of three replicates per treatment (± standard deviation). Different letters in the same column represent means that are significantly different from each other (Tukey’s post hoc text, P < 0.001).

Sample ID	Treatment	Soil type (tree sp. in the plot)	Functional catergory of trees	Analyses	Prok TO(106 ml−1)	Prok T6(106 ml−1)	Number of OTUs	Chao1	Evenness
Soil Ac	Soil Ac	Acer saccharum	Deciduous broadleaf	Sequencing	–	–	1155 (31)c	1641 (56)a	0.85 (0.01)a
Soil La	Soil La	Larix laricina	Decidous conifer	Sequencing	–	–	1108 (32)cd	1673 (72)a	0.82 (0.01)a
Soil Pi	Soil Pi	Pinus strobus	Evergreen conifer	Sequencing	–	–	1203 (18)d	1703 (80)a	0.85 (0.01)a
Soil LaAc	Soil LaAc	A. saccharum/L. laricina	Dec. broadleaf/Dec. conifer	Sequencing	–	–	1172 (27)d	1711 (58)a	0.85 (0.00)a
Soil LaPi	Soil LaPi	P. strobus/L. laricina	Dec. conifer/Ever. conifer	Sequencing	–	–	1187 (7)d	1696 (34)a	0.85 (0.00)a
In situ	Unfiltered Lake water	–	–	Bact abund	1.01	–	329a	442b	0.56a
Control	<1um Lake water	–	–	Bact abund/sequencing/biolog	0.98 (0.08)a	1.88 (0.27)a	356 (4)a	524 (40)b	0.57 (0.01)b
Ac	<1um Lake water + soil Ac	Acer saccharum	Deciduous broadleaf	Bact abund/sequencing/biolog	0.91 (0.09)a	2.14 (0.20)ab	914 (46)b	1610 (156)a	0.66 (0.02)b
La	<1um Lake water + soil La	Larix laricina	Decidous conifer	Bact abund/sequencing/biolog	1.02 (0.06)a	2.36 (0.37)ab	964 (86)bc	1618 (119)a	0.70 (0.05)b
Pi	<1um Lake water + soil Pi	Pinus strobus	Evergreen conifer	Bact abund/sequencing/biolog	0.90 (0.06)a	2.38 (0.04)ab	959 (74)b	1550 (138)a	0.69 (0.01)b
LaAc	<1um Lake water + soil LaAc	A. saccharum/L. laricina	Dec. broadleaf/dec. conifer	Bact abund/sequencing/biolog	0.86 (0.29)a	2.81 (0.54)b	983 (53)bc	1717 (29)a	0.68 (0.04)b
LaPi	<1um Lake water + soil LaPi	P. strobus/L. laricina	Dec. conifer/ever. conifer	Bact abund/sequencing/biolog	0.93 (0.04)a	2.97 (0.26)b	989 (88)bc	1728 (165)a	0.67 (0.02)b

In the lab, lake water was filtered through 1 µm pore-size filters to remove most phytoplankton and bacterivore protists but allow most free-living bacteria to pass, and 24 1L-Nalgene autoclaved bottles were filled with 1L of this filtered lake water. We added 0.65 g of wet soil from the different plots to the bottles in triplicate, such that 3 bottles were inoculated with each type of soil. Three additional bottles containing only filtered lake water were kept as control (see Table 1 for details on experimental treatments). Bottles were mixed gently, and once the samples for the initial prokaryotic abundances were taken from each bottle (see “Prokaryotic abundances” section), they were kept in the dark in a temperature-controlled chamber at 24ºC. After 6 days samples were taken for final prokaryotic abundances, DNA analyses, and substrate utilization patterns.

Establishment of the incubation period

Prior to the experiment, two bottles (one containing only 1 µm-filtered lake water and one inoculated with soil) were incubated under the same conditions as in our main experiment and sampled daily to test for changes in prokaryotic abundance through time (Fig. S1A, Supporting Information). A 6-day incubation was chosen in order to give enough time for the full development of the bacterial community (and to ensure that initially rare soil taxa could attain high enough abundances) while avoiding the eventual collapse of the populations due to nutrient limitation. In the main experiment, we sought to avoid collection of daily samples to prevent contamination and to maintain water volumes. We therefore also ran the test incubation to provide samples for daily prokaryotic abundances during the incubation of the main experiment (Fig. S1B, Supporting Information).

Prokaryotic abundances

Prokaryotic abundances were estimated by flow cytometry (Gasol and del Giorgio 2000) both at the beginning and at the end of the experiment, and daily in the two bottles that were kept to track the changes in prokaryotic abundance during the experiment (Fig. S1B, Supporting Information). Samples of 2 mL were preserved with 1% gluteraldehyde (final concentration) and kept frozen at −80ºC until analysis with a Becton–Dickinson FACSCalibur flow cytometer after staining with SYTO-13 (Molecular Probes, Eugene, OR).

Bacterial substrate utilization patterns

The substrate utilization patterns were obtained with Biolog EcoPlates™ (CA, USA) both from the lake water at the start of the experiment and from the different treatments after incubation. Unfiltered water (125 µl) from each bottle was inoculated to each well of the Ecoplates, which contain 31 different carbon sources in triplicate. The bacterial respiratory activity associated to substrate use reduces a tetrazolium dye, producing color measurable as the absorbance at 595 nm. One plate per bottle was incubated in the dark and at room temperature and the overall color development of the plates (average well color development, AWCD) was determined daily using a microplate reader. The mean color development of each compound was calculated as the blank-corrected mean absorbance of each substrate measured at the time when the AWCD was closest to 0.5 (Garland, Mills and Young 2001), usually between days 2 and 5. We also registered the maximum AWCD attained by each sample as a proxy of the maximum average rate of substrate utilization.

Bacterial community composition

Either 0.25 g of the original soil samples or 300 mL of water filtered onto 0.22 µm filters were used for genomic analyses. Bacterial DNA was extracted using MoBio PowerWater DNA extraction kits following the manufacturer’s protocol. The V5-V6 region of the 16S rRNA gene was amplified using the bacterial primers (799F and 1115R; Reysenbach and Pace 1995; Chelius and Triplett 2001) in a one-step PCR designed to attach a 12-base pair barcode and Illumina adaptor sequence to the fragments during amplification. The primer selection was based on our intention to compare these samples with phyllosphere samples collected at the same sites, so they are widely-used chloroplast-excluding primers (e.g. Laforest-Lapointe et al.2017). As a result they do not capture Cyanobacteria but an in silico test against the SILVA database using the online tool TestPrime (Klindworth et al.2012) shows that they recover 80% of the total bacterial sequences, showing good coverage of most dominant phyla in soils and water. In support of this, the bacterial community from the studied lake (lake Croche) addressed with this primer pair and the more commonly used 515F-806R pair (Caporaso et al.2011) shows largely similar taxonomic composition (García-Chaves unpublished data), thus suggesting that the use of these primers does not provide a significantly different depiction of communities.

We prepared multiplexed 16S libraries by mixing equimolar concentrations of DNA, and sequenced the DNA library on an Illumina MiSeq using a paired-end approach (Caporaso et al.2012). Paired-end reads were assembled with PEAR (Zhang et al.2013) and sequences between 250–290 base pairs in length were used for downstream analyses in QIIME in order to remove primers, low-quality and archaeal reads (Caporaso et al. 2010). We eliminated chimeric sequences using the Uclust and Usearch algorithms (Edgar 2010). The remaining sequences were binned into operational taxonomic units (OTUs, ≥ 97% similarity) using UCLUST v1.22q (Edgar 2010) and identified taxonomically using the BLAST algorithm by comparison with the SILVA v.123 reference alignment. In order to remove potentially spurious OTUs due to sequencing or PCR errors, OTUs represented by 20 or fewer sequences were removed prior to subsequent analyses. In order to enable comparisons between samples for beta-diversity analyses, the OTU table was randomly subsampled to ensure an equal number of sequences per sample, based on the sample with the least number of reads, and the resulting rarefied OTU table had 7106 sequences per sample. Raw sequence data have been deposited in the Figshare data repository (https://doi.org/10.6084/m9.figshare.6654287.v1).

Identification of ‘reactive’ OTUs

We identified as ‘reactive’ OTUs those that showed significant increases in relative abundance over the course of the experiment relative to their abundances at the beginning of the experiment (T₀, immediately upon inoculation with soil for the five soil-amended treatments), and also with respect to the control treatment. Reactive OTUs were detected using the Bioconductor package edgeR (Robinson, McCarthy and Smyth 2010), a software that allows finding significant changes between two or more groups when at least one of the groups has replicated measurements. We applied the glmLRT function to compare the abundances of each OTU between the different samples, and OTUs were regarded as ‘reactive’ when showing significant increases (adjusted P values < 0.05) in the amended treatments relative to the initial conditions and the control. In order to explore whether the same OTUs responded similarly or differently across treatments, this analysis was performed separately for each of the different five soil additions. To identify the ‘reactive’ OTUs in the control treatment, we simply extracted the taxa that increased their abundances relative to the in situ lake community. Because the portion of the spiked soil bacteria of all bacteria in the microcosm (lake + soil bacteria) was not quantified at the beginning of the experiments, we approximated the initial relative abundance of the inoculated soil taxa in the mesocosm water by assuming the potential dilution of the bacterial biomass in the 0.65g soil plug added. We used the bacterial biomass values reported for the same soils in these same experimental plots by Rivest et al. (2015) in conjunction with the bacterial biomass present in the lake medium at the beginning of the experiment; the latter calculated from the measured cell abundances assuming 20 fg C cell⁻¹ (del Giorgio et al. 1997). This calculation suggested an average 25-fold dilution of soil bacteria by lake water medium in the microcosms at the start of the incubations, and we used this dilution factor to estimate the initial abundances of all soil-derived taxa.

Statistical analyses

Bray–Curtis distances were used as an estimate of taxonomic and functional dissimilarity between samples. Community dissimilarities were visualized using non-metric multidimensional scaling (NMDS) analyses. Differences in bacterial taxonomic composition or functional profiles between treatments were tested with permutational multivariate analyses of variance (PERMANOVA). Links between differences in taxonomic composition and metabolic dissimilarity were estimated using Mantel tests. Pielou’s evenness index was calculated as an estimate of bacterial taxonomic evenness, and the Chao1 estimator was computed as a diversity index. All the analyses were performed using R 3.0.0 software (R Core Team, 2013) and the R vegan package (Okasanen et al.2015)

RESULTS

Bacterial community composition

We recovered 430 739 high quality sequences that clustered into 2,765 OTUs (≥97% sequence similarity). After subsampling our OTU table to 7106 reads per sample, 244 596 sequences were retained that clustered into 2732OTUs. The investigated soil communities clearly differed between the five soils considered (Fig. 1A, PERMANOVA R² = 0.51 (i.e., 51% of the variation in distances was explained by the soil type), P < 0.001). Although at the phyla, class and order level (Fig. 2), they were roughly comparable, we observed significant changes in the abundance of different genera across soils (Figure 2, Supporting Information). 45% of the soil OTUs (representing 90% of total soil sequences) were present across the five soils, but there were also taxa restricted to certain soils (details not shown). Taken together, 61 bacterial classes were found across soil samples, which were mainly dominated by the class Actinobacteria (38% of the total soil sequences), followed by Thermoleophilia (11%), Alphaproteobacteria (9%), Betaproteobacteria (7%), Acidimicrobiia (5%) and Chloroflexi (4%). In contrast, the lake community was less diverse, and comprised only 18 bacterial classes, the most dominant being Actinobacteria (53% of lake sequences) and Betaproteobacteria (32% of all sequences).

Figure 1.

Compositional differences between the studied bacterial communities. NMDS plots based on Bray–Curtis distances of taxonomic composition of bacterial communities from the original soils (A) and the bacterioplankton communities after 6 days of lake water incubations (B). Symbols indicate the different soil communities (A) and the different experimental treatments (B). Stress values are shown for each plot.

Figure 2.

Taxonomic composition of bacterial communities. Data are presented as percent contribution of each bacterial taxonomic rank to the total community sequences pooling the three replicates together. The classification was performed at the Phylum (P.) level in some cases, but the Phyla Actinobacteria, Bacteroidetes and Proteobacteria (separated by dashed lines) were split into classes (C.) or orders. Act: C. Actinobacteria; Alph: C. Alphaproteobacteria; Bet: C. Betaproteobacteria. Treatment acronyms are explained in Table 1.

Incubations of lake water amended with different soils

The prokaryotic abundance in the epilimnion of Lake Croche at the time of the experiment was 1.01 × 10⁶ cells ml⁻¹. The 1 µm-prefiltration of lake water used to generate the incubation medium did not result in a significant decrease in prokaryotic abundance (Table 1). Similarly, the soil additions did not result in a measurable increase in prokaryotic abundance in the bottles at the beginning of the experiment, relative to the ambient lake water (Table 1) even though it is possible that there were large aggregates that were not quantified by the flow cytometer. This lack of increase in prokaryotic numbers due to soil additions is not unexpected; based on the amount of soil added and the soil bacterial biomass of these same soils (Rivest et al.2015), we were probably adding between 30 000–60 000 cells per ml of lake water, which is smaller than the variability between replicates (Table 1).

The 6-day incubation of the samples resulted in an average 2-fold increase in prokaryotic abundance in the control treatment (lake water only) relative to unfiltered lake water, whereas the increase in prokaryotic abundance in soil-inoculated treatments was generally higher, ranging from 2.4 to 3.8-fold (Table 1). Interestingly, the largest increases in abundance occurred in the treatments amended with soil associated to combinations of tree species (LaPi and LaAc, respectively), whereas the lowest occurred in the treatment amended with soil associated to the single deciduous species A. saccharum (Ac, Table 1).

The incubation of filtered lake water in the microcosms for 6 days (control treatment) did not result in major changes in taxonomic composition, richness, or evenness of the bacterial communities relative to the in situ lake communities (Fig. 1B, Table 1). In contrast, the inoculation of lake water with soil (all types) led to clearly different bacterial communities after 6 days relative to unfiltered lake and control communities (Fig. 1B). The communities that developed in the soil-amended bottles had on average 2.6-fold more OTUs compared to ambient lake water (Table 1), probably because all soil inocula had many more bacterial taxa than the ambient lake community.

Despite the clear differences in bacterial community structure between the different soils that were used as inocula (Fig. 1A), the structure of the communities that developed in the soil-amended incubations did not have a clear origin-dependent pattern (Fig. 1B), although there were still significant differences between the resulting communities based on PERMANOVA analysis (Fig. 1B, R² = 0.37, P < 0.01). In terms of taxonomic composition, the classes Actinobacteria and Betaproteobacteria still dominated the communities that developed in the soil-amended communities, but other classes like Cytophagia, Thermoleophilia, Acidimicrobiia and Alphaproteobacteria were more abundant in the soil-amended samples relative to the unfiltered and control water incubations (Fig. 2).

The communities that developed in the soil-amended incubations were dominated by OTUs that were originally detected in the soil but not in the lake water (i.e., ‘soil-unique’ OTUs, Fig. 3A), although in terms of sequences, they were dominated by OTUs that were detected in both the lake water and in the soils at the beginning of the experiment (i.e., ‘shared’ OTUs, Fig. 3B). ‘Lake-unique’ OTUs, i.e., those OTUs that were found in the lake water either at the beginning or the end of the experiment, but never in the original soil inocula, accounted for ca. 30% of the total sequences in the control treatment and decreased to 20% in the five soil-amended treatments (Fig. 3B). Interestingly, >70% of the lake bacterial sequences (both in in situ lake water and in the control treatment) belonged to OTUs that were also detected in the ambient soils but at low abundances (Fig. 3B).

Figure 3.

Distribution of taxa based on their presence across the studied communities. (A) Proportion of OTUs categorized depending on whether they were initially found only in soils (Soil unique), only in the lake water (either at the beginning or the end of the experiment in the Control treatment, Lake unique OTUs) or that were detected in both lake water and in one or more of the initial soils. (B) Proportion of sequences associated to those taxa. For the control and each of the five soil-inoculated treatments, the values are expressed as a fraction of the total community OTUs (A) or sequences (B), pooling the 3 replicates together.

Identification of taxa positively responding to the soil additions (i.e., ‘reactive’ OTUs)

To further identify the main taxa behind the differences in taxonomic resulting from the soil amendments, we identified OTUs that were the most responsive in each of the soil treatments and in the control; responsiveness defined by significant increases in sequence abundance relative to that at time 0 of the incubation (see Methods). Pooling the five experimental treatments plus the control, we recovered 252 OTUs that showed significant increases, ranging from118 in LaAc to 132 in Pi treatment. Only 4 OTUs were identified in the control treatment (details not shown). Many of the OTUs detected, however, were relatively rare and showed small increases in abundance, so for subsequent analyses we focused on the 50 OTUs that showed the largest average increases in abundance (hereafter ‘reactive’ OTUs, Fig. 4). By doing this, only one OTU was retained as ‘reactive’ in the Control treatment, whereas the number of ‘reactive’ OTUs across the five soil-amendments varied from 46 to 49.

Figure 4.

Identification and responses of ‘reactive’ taxa. Average change in relative abundance of the 50 bacterial OTUs that showed the largest increases in abundance relative to the communities at the beginning of the experiments (i.e., “reactive” OTUs). The change was calculated as the difference of relative abundance of each OTU at the T6 to its initial abundances in the water after inoculation with the soil (see Results for further details), and is presented as the average of the three replicates for each treatment. Colors indicate whether an OTU was initially detected only in soils (white), in lakes (black), or was present both in the lake and any soil (grey). The same individual OTUs are represented in all graphs and are ranked along the X axis according to their overall mean change in relative abundance, so that gaps indicate that a given OTU was not identified as ‘reactive’ in that particular treatment. The asterisks (*) in the first panel indicate which of those ‘reactive’ OTUs were identified as ‘reactive’ across the five treatments. The numbers on top of each graph indicate the number of OTUs that were identified as ‘reactive’ in each treatment.

In order to investigate whether the responses of the OTUs categorized as ‘reactive’ differed across the five soil treatments, we ranked all the ‘reactive’ OTUs by their average increase in abundance during the experiment, and compared these among treatments (Fig. 4). Interestingly, roughly the same OTUs behaved as ‘reactive’ across the five soil additions showing responses of similar magnitude (Fig. 4). Most of the ‘reactive’ OTUs were initially present both in the soils and the lakes (i.e., ‘shared’ OTUs, Fig. 4), and indeed most were detected across the five soils (details not shown). Soil-unique and lake-unique ‘reactive’ OTUs were generally less responsive to soil additions and showed smaller changes in abundance (Fig. 4). In terms of composition, most sequences associated to these ‘reactive’ OTUs belonged to the orders Corynebacteriales, Burkholderiales, Cytophagales and Rhizobiales across all soil treatments, whereas the only ‘reactive’ OTU from the control treatment was associated to the order Sphingonadales (Fig. 5). A detailed exploration of the dynamics of the 10 ‘reactive’ OTUs showing the largest increases during incubations (first 10 OTUs from Fig. 4) showed that although they behaved as ‘reactive’ regardless of the soil type in most cases, the abundances reached varied between treatments (Fig. S3, Supporting Information). Interestingly, moreover, their abundances in the original soils varied from very low to relatively abundant.

Figure 5.

Taxonomic composition of ‘reactive’ taxa. Taxonomic composition of the bacterial taxa that showed significant increases in relative abundance during the course of the experiments experiments (i.e., ‘reactive’ OTUs, see Methods for further details). Data are presented as percent contribution of each bacterial taxonomic rank to the sequences of the pool of reactive taxa considering the three replicates together. The classification was performed at the Phylum (P.) level in some cases, but the Phyla Actinobacteria, Bacteroidetes and Proteobacteria (separated by dashed lines) were split into classes (C.) or orders. Act: C. Actinobacteria; Alph: C. Alphaproteobacteria; Bet: C. Betaproteobacteria. Treatment acronyms are explained in Table 1.

The ‘shared’ and ‘lake-unique’ subsets of ‘reactive’ OTUs did not show a consistent pattern related to soil type in the soil-inoculated aquatic communities (PERMANOVA P > 0.05, Table 2), whereas ‘soil-unique’ reactive OTUs showed a certain segregation depending on the type of soil. However, both the ‘shared’ and ‘soil-unique’ reactive taxa did show some significant soil-specific taxonomic segregation in the original soils before inoculation in the water (Table 2). This loss of the soil signature during incubation suggests that different soil inputs triggered the response of a core pool of ‘reactive’ OTUs that behaved similarly regardless of the soil addition (i.e., the most responsive ‘shared’ OTUs shown in Fig. 4), but also a few ‘soil-unique’ reactive OTUs that did reflect their source of origin. However, the latter were less abundant on average and showed smaller increases during incubation (Fig. 4), explaining why the pool of ‘reactive’ taxa did not retain such a soil-specific signature after incubations (Table 2).

Table 2.

Taxonomic differences between soil treatments in the ‘reactive’ taxa pools. Permutational multivariate analysis of variance (PERMANOVA) to examine differences in taxonomic composition of different subsets of bacteria between the five different types of soils, both in the original soils (Soil bacterial communities) and after 6 day incubation of soil in lake water (Soil-derived bacterioplankton communities). All R² for Bray–Curtis distance matrices values were calculated using 9999 permutations. Asterisks indicate significant differences (***P < 0.0001;P < 0.001**; P < 0.05*) in taxonomic composition between the five different soils or five different soil-inoculated treatments. n.s. = not significant; n.a. = not applicable.

	PERMANOVA R2 values
	Soil bacterial communities	Soil-derived bacterioplankton communities
All ‘reactive’ OTUs	0.50***	n.s.
Shared ‘reactive’ OTUs	0.49**	n.s.
Soil-unique ‘reactive’ OTUs	0.52***	0.39*
Lake-unique ‘reactive’ OTUs	n.a.	n.s
All non-reactive OTUs	0.51***	0.38**

Effects of soil inoculants on the functional capacities of lake communities

The shifts in community composition resulting from soil amendments also triggered major changes in the functional capacities of communities in terms of organic substrate use (Biolog substrates), relative to the control treatment. By the end of the experiment, soil-amended communities clearly diverged from the control assemblages (Fig. 6A), although there was no significant segregation of communities on the basis of soil treatment (PERMANOVA P > 0.05). There were no large differences between the control and any of the soil treatments in the number of substrates used and functional diversity (details not shown), but soil-amended assemblages showed on average 60% greater maximum average substrate utilization (maximum AWCD) than both the initial lake and the control communities (Fig. 6B), suggesting that the addition of soil significantly enhanced overall bacterial substrate uptake in the incubations. Compared to the control treatment, the use of some substrates such as D-cellobiose, glycogen and phenylalanine was significantly enhanced in the soil-amended communities (Fig. 6C), whereas other substrates were less utilized (e.g., asparagine, arginine). Interestingly, the direction of these changes was the same across the five soil treatments, and even the magnitude of the change was relatively similar in most cases(Fig. 6C).

Figure 6.

Effect of soil treatments on functional profiles. (A) NMDS of samples based on the metabolic profiles obtained with the Biolog Ecoplates. Symbols indicate the different treatments. (B) Maximum average colour development (maximum AWCD, a proxy of the maximum substrate use) reached by the plates in the different treatments. (C) Mean absolute change in substrate used for the 19 substrates that showed significant increases (positive values) or decreases (negative values) with respect to the control treatment. Values are means (±standard error) of triplicate samples.

Links between taxonomic composition and functional profiles

To explore whether the observed changes in taxonomic composition could explain the shifts in substrate uptake patterns in the soil amended incubations, we estimated the Mantel correlation between the overall functional dissimilarity matrix and the taxonomic dissimilarity matrices calculated for different subsets of bacteria (i.e, all OTUs, all ‘lake-unique’ OTUs, all ‘reactive’ OTUs –including ‘soil-unique’ or ‘shared’ reactive OTUs-, and all ‘non-reactive’ OTUs –i.e., all OTUs not categorized as reactive-). We found that changes in metabolic profiles observed at the community level in the five soil treatments were not significantly associated with the overall changes in community composition (Table 3). Instead, metabolic differences were most strongly correlated to shifts in the composition of ‘reactive’ OTUs and among those, the ‘shared reactive’ and ‘soil-unique’ OTUs, although the significance of these correlations was slightly above the 0.05 level. (Table 3).

Table 3.

Links between taxonomy and function. Variation in the R coefficients of the Mantel correlations between the taxonomic dissimilarity matrices calculated for different subsets of bacteria, and the overall functional dissimilarity matrix.

	OTU number	Mantel R (p value)
All OTUs	2507	0.31 (0.10)
Non-reactive OTUs	2457	0.17 (0.21)
All reactive OTUs	50	0.45 (0.07)
Reactive ‘shared’ OTUs	33	0.42 (0.06)
Reactive ‘soil unique’ OTUs	13	0.55 (0.06)
Reactive ‘lake unique’ OTUs	4	0.14 (0.29)

The correlations were performed only for the five treatments inoculated with soil (n = 15). Bold values indicate correlations below the P = 0.1 significance level. OTU number indicates the number of OTUs that were considered for each individual analysis.

DISCUSSION

We show that soil inputs can significantly influence freshwater bacterioplankton communities taxonomically and functionally, by providing terrestrially derived material and delivering novel immigrant taxa that may develop in the new aquatic environment, in agreement with recent reports (Judd, Crump and Kling 2006; Kritzberg, Langenheder and Lindström 2006; Perez and Sommaruga 2006; Berggren, Laudon and Jansson 2007; Crump, Amaral-Zettler and Kling 2012; Besemer et al.2013; Fasching et al.2014; Berggren and del Giorgio 2015; Ruiz-González, Niño-García and del Giorgio 2015a; Wilhelm et al.2015). Our results, however, do not support the hypothesis that inputs from soils associated with different tree species or communities will lead to predictably distinct bacterioplankton assemblages under the tested circumstances. Although there were large differences in the physico-chemical properties between the five soil treatments (see Rivest et al.2015) and in the community structure of the resident soil microbial assemblages (Fig. 1A; Fig. S2, Supporting Information), we did not find a comparable source-specific signature in the functional or taxonomic responses of the communities that developed in the soil-amended lake water. Thus, while soil additions significantly influenced lake communities, the taxonomic and functional shifts could not be predicted based on the type of forest soil involved.

When lake water is amended with soil as in our experiment, in addition to inoculating terrestrial bacteria, nutrients and a variety of carbon substrates are added, potentially influencing the growth of local lake bacterial communities (e.g., Eiler et al.2003; Lennon and Pfaff 2005; Kritzberg, Langenheder and Lindström 2006; Perez and Sommaruga 2006). Even though there is experimental evidence showing that bacteria able to persist in dry soils and sediments can readily colonize water (e.g. Fazi et al.2008), the relative importance of direct recruitment of soil bacteria versus the effect of soil-derived resources on local communities is difficult to disentangle. However, in natural ecosystems bacteria are washed into aquatic ecosystems together with soil-derived material, and thus the purpose of our study was thus not to separate these processes, but rather to assess (1) the overall responses of lake bacteria to soil additions, and (2) whether differences in the original soil inoculants, driven entirely by differing tree composition would trigger distinct taxonomic and or functional responses in the receiving lake bacterial communities. Hereafter, we will discuss these two main issues separately.

Aquatic bacterial responses to soil addition

The addition of soils to the lake water resulted in pronounced changes in prokaryotic abundance, taxonomic diversity and functional profiles of the communities that developed in the soil-amended incubations relative to the control treatment (Table 1, Figs 1 and 6). First, the increase in prokaryotic abundance relative to unfiltered lake water was much larger in soil-inoculated treatments than in the unamended lake water (control) treatment (Table 1). Moreover, soil additions produced a large increase in OTU number (Table 1), likely due to the higher taxonomic richness of soil communities, which was not apparent in control communities. Interestingly, however, only a small fraction of those OTUs seemed to be able to increase their relative abundances following soil addition (i.e., the “reactive” OTUs, Figs 4 and 5), suggesting that most soil-derived OTUs could probably not thrive in the new aquatic conditions but were still detected through sequencing of their 16S rRNA gene. This pattern is in accordance to previous observations showing that local freshwater bacterioplankton communities comprise a large number of presumably inactive taxa transported from the surrounding catchment and adjacent terrestrial ecosystems (Crump, Amaral-Zettler and Kling 2012; Ruiz-González, Niño-García and del Giorgio 2015a; Savio et al.2015; Niño-García, Ruiz-González and del Giorgio 2016a,b; 2017b), whereas only a small subset of these transported bacteria may grow and thrive in aquatic habitats if the water residence time is long enough (Ruiz-González, Niño-García and del Giorgio 2015a; 2017b). Finally, the fact that very few of the ‘reactive’ OTUs were ‘lake unique’ taxa suggests that many of the OTUs present in the original lake water were not particularly favored by the experimental soil addition, probably because it selected for opportunistic bacteria able to rapidly respond to shifts in conditions.

The short duration of our experiments only allows us to speculate about the capability of these taxa to persist in the lake following inoculation: it is possible that, regardless of their origin, these ‘reactive’ OTUs are opportunistic bacteria showing transient abundance increases due to a better ability to utilize the soil-derived substrates, but once they are depleted, the lake dominant taxa would outcompete them again, as has been observed previously after changes in allochthonous inputs (Crump et al.2003; Haukka et al. 2005, Neuenschwander et al.2015). In this regard, the orders Cytophagales and Burkholderiales, which accounted for an important fraction of the ‘reactive’ OTUs, have been previously associated to higher concentrations of humic DOM (Burkert et al.2003; Eiler et al.2003) or to modifications in DOM composition after flooding events in subsurface karst pools (Shabarova et al.2014).

We acknowledge that our experimental setup cannot be considered as a realistic scenario; by pre-filtering the lake water and incubating samples in the dark, we were disregarding other processes that naturally occur in these systems and that together shape the resulting bacterioplankton communities. However, our aim was to keep the experimental setup as simple as possible in order to minimize factors that would have made it extremely difficult to derive any conclusions related to the capacity of different bacterial taxa to respond to soil additions. For example, grazers might mask the observed effects of soil inputs by quickly consuming the opportunistic bacteria as they grow, as shown in other experimental studies (e.g. Simek et al.2003; García-Chaves et al.2015).Thus, the reasons for pre-filtering and keeping bottles in the dark were just to minimize these other confounding factors (such as indirect stimulation of bacterial growth due to primary production and DOC exudates, among others) in order to target the responses of communities to soil additions.

From a functional perspective, soil-amended communities also differed from both the ambient lake and the control assemblages. For example, the use of certain substrates associated with terrestrial sources (e.g., cellulose and xylose) on the Biolog plates was favored with soil amendment compared to the control, whereas the utilization of other substrates was suppressed (Fig. 6C). Moreover, the maximum average substrate utilization (maximum AWCD) was consistently higher across all soil-amended samples relative to the control (Fig. 6B). This agrees with our previous study of large scale spatial patterns in the substrate-uptake profiles of hundreds of boreal freshwater bacterioplankton communities (Ruiz-González et al.2015b), wherein communities inhabiting environments with high colored DOM (cDOM, indicative of terrestrial inputs) had consistently higher maximum AWCD values, with maximum AWCD being strongly correlated with cDOM. Our results thus support the idea that terrestrial inputs may lead to potentially enhanced bacterial utilization of DOC, in accordance with previous research showing that increases in terrestrial DOC inputs have major impacts on the overall carbon cycling of the receiving aquatic ecosystems (Lapierre et al.2013; Lapierre and del Giorgio 2014; Jones and Lennon 2015).

Lack of source-specific response of lake microcosm bacterial communities

Despite the large differences between control and soil-amended communities, however, the responses were not source-dependent and communities did not show clear patterns depending on the soil treatment considered. This is surprising because our different soil additions certainly provided the lake communities with both a different pool of terrestrial bacteria (Fig. 1A), and presumably also DOM of different quality and lability.

The differences in community structure of the five soil bacterial assemblages support previous reports that tree composition may directly or indirectly impact the associated soil microbiome (e.g., Jiang et al.2012; Rivest et al.2015). When the Montreal IDENT experimental facility was established in 2009, the soil was very homogenous due to several decades of tilling for agricultural production, and thus most of the differences in soil bacterial communities and physico-chemical properties among tree patches resulted from the influence of tree composition, which have been previously documented (see Rivest et al.2015; Khlifa et al.2017). In addition, both leaf microbiome (Laforest-Lapointe et al.2017) and precipitation runoff from the canopy of these different tree species and combination of species (Nock et al. pers. comm.) at the Montreal IDENT site have been shown to consist of taxonomically unique bacterial communities. Thus, these soils are continuously inoculated with specific bacterial pools through rain or litterfall, explaining part of the observed differences in composition, although the extent to which these phyllosphere taxa persist or thrive in the soil environment remains unexplored (but see Ruiz-González et al.2017a).

Interestingly, however, this source-dependent signature was lost after incubation in lake water (Fig. 1B). Such a pattern could be expected if the soil taxa able to thrive in lake water all belonged to a single, ubiquitous pool present across all soils, or if the most important DOM sources provided with the soils do not vary largely between them. We observed that roughly the same OTUs behaved as ‘reactive’ regardless of the type of soil added (Fig. 4), but the abundances attained by them varied between treatments (Figure 2, Supporting Information). The fact that these reactive taxa exhibited different growth patterns between different soil treatments would suggest the presence of a random component potentially linked to priority effects (sensu Fukami 2015), or that the associated changes in DOM or nutrients favored some taxa over the others. It is interesting to note, however, that most of these reactive taxa were also present in the original lake community, albeit some at low densities. Thus, it is difficult to conclude whether the soil amendments yielded viable soil-derived bacteria, or whether the responses resulted from the stimulation by soil additions of resident generalist bacteria that are ubiquitous across soils and waters. The latter would agree with the observation that ubiquitous, generalist species are often those that perform best when transposed to new conditions (Langenheder and Székely 2011; Székely, Berga and Langenheder 2013; Comte et al.2014; Ruiz-González, Niño-García and del Giorgio 2015a; 2017). Moreover, the fact that the initial lake bacterioplankton communities were numerically dominated by these ‘shared’ taxa present in soils (Fig. 3), agrees with the observation that freshwater bacterioplankton are largely composed of bacteria that have the potential to persist in soils or sediments but grow upon inoculation in the water (Crump, Amaral-Zettler and Kling 2012; Fazi et al. 2008; Ruiz-González, Niño-García and del Giorgio 2015a; 2017b). Indeed, a significant fraction of the ‘reactive’ taxa sequences belonged to Burkhorderiales and Rhizobiales, which were found among the soil-derived taxa showing increases in relative abundances from headwaters to downstream ecosystems across boreal aquatic networks (Ruiz-González, Niño-García and del Giorgio 2015a). It is important to point out that the studied soils were located ca. 60 km far from the lake, meaning that they were not directly connected to lake Croche, highlighting the persistence capacities of these ubiquitous reactive taxa.

The inoculation with soils likely also provided the lake bacteria with different DOM and nutrients, because at least some chemical properties such as total C and N content or pH differed significantly between the five soils, as the three tree species considered are known to produce different amounts and quality of leaf litter (Rivest et al.2015). DOM quality is known to impact freshwater bacterial communities both structurally and functionally (Lennon and Pfaff 2005; Kritzberg, Langenheder and Lindström 2006; Hutalle-Schmelzer et al.2010; Fujii et al.2012; Berggren and del Giorgio 2015; Ruiz-González et al.2015b; Jones and Lennon 2015), but there is also evidence that changes in DOM quality may influence functional aspects of bacterial communities to a larger extent than their taxonomic structure (e.g. Landa et al.2014; Ruiz-González et al.2015b), likely reflecting widespread functional redundancy across these communities, at least related to carbon substrate utilization. The functional profiles of the soil-amended communities did not cluster based on the type of soil inoculated, and many aspects of the functional structure were relatively similar across the different soil treatments (Fig. 6): Besides the consistent increase in maximum AWCD mentioned (Fig. 6B), the direction and the magnitude of the functional shifts were remarkably similar between the soil treatments (Fig. 6C). This suggests that irrespective of the nature of the material or bacteria loaded with each type of soil, convergence in the functional structure of the communities developed during incubations. Previously, we found that aquatic communities inhabiting environments with high concentrations of coloured DOM (indicative of terrestrial inputs) showed rather homogeneous functional structures relative to communities from systems with less terrestrial influence (Ruiz-González et al.2015b; 2017b), suggesting that high terrestrial DOM input may result in a relatively narrow resource space for bacterial communities.

Interestingly, the increase in overall substrate utilization was not associated with the observed changes in prokaryotic abundance in the incubations (Table 1, Fig. 4B). When the maximum AWCD was normalized by the number of bacteria at the end of the experiment, we observed that the communities amended with soil associated to sugar maple (Ac) had the highest relative AWCD per bacteria, whereas amendments with soil from mixed conifer stands (LaPi) showed the lowest ratio (details not shown). Previous studies have shown that the degradability of DOC by aquatic bacteria differs depending on its terrestrial origin, especially between peatbogs and different forest soils (Lennon and Pfaff 2005; Bergren, Laudon and Jansson 2007; 2010; Guillemette and del Giorgio 2011; Berggren and del Giorgio 2015). Our results suggest that whereas the functional response to the various soil amendments was convergent, the actual metabolic performance of bacteria differed between these amendments, likely driven by differences in the overall reactivity and quality of these various DOM sources.

Finally, it is also interesting to note that a relatively small number of taxa, and in particular those ‘reactive’ OTUs initially found in the soils, were likely driving most of the observed variation in the functional structure of communities resulting from soil amendments (Table 3). This explains lack of soil-specific functional responses, since the pool of ‘reactive’ OTUs also failed to show such a soil signature (Table 2). This highlights the fact that major shifts in community-level function may be driven by responses to soil amendments of a small number of key populations, in agreement with previous reports showing that fast growing, opportunistic taxa can disproportionately contribute to shaping processes at the community level (Perez and Sommaruga 2006; Aanderud et al.2015; Neuenschander et al.2015).

In summary, we show that soil amendments to lake waters significantly impact the structure of resident bacterioplankton communities, at least in the short-term, both by inoculating a large number of soil-derived taxa to the lake assemblages, and also by stimulating the growth and activity of a core group of ubiquitous taxa that were distributed across the terrestrial-aquatic continuum. Soil amendments also resulted in a rather narrow and converging functional response, suggesting that in spite of the tree species -driven differences in soil bio-physico-chemical properties, bacteria must have reacted to a core set of organic substrates that must be common across soil types. Interestingly, there was no correlational link between the taxonomic and functional responses of lake bacterioplankton to soil amendments at the whole community level, but this link was stronger when only the “reactive” taxa were considered. In any case, these results are relevant in light of projected increases in the delivery of terrestrially derived DOC to inland waters (“browning”), which are expected to be pronounced in boreal regions (Larsen, Andersen and Hessen 2011) and may have major impacts on the overall carbon cycling of the receiving aquatic ecosystems (Lapierre et al.2013; Lapierre and del Giorgio 2014; Jones and Lennon 2015). Based on our results, however, a directional or deterministic taxonomic response of lake bacterial communities to these increases in terrestrial inputs from different soil types may not always be expected, although this may change if more different soils were targeted (i.e., soils varying largely in pH, Fierer et al.2006). Finally, our results suggest that “browning” might cause some degree of functional convergence and homogenization among the communities most influenced by terrestrial inputs, regardless of the surrounding soil types.

SUPPLEMENTARY DATA

Supplementary data are available at FEMSEC online.

ACKNOWLEDGEMENTS

We thank T. Dawson for their contribution to laboratory components of this research, and F.M. Cornejo-Castillo for his help checking primer performance. This study is part of the programs of the Carbon Biogeochemistry in Boreal Aquatic Systems (CarBBAS) Industrial Research Chair to PdG and of the controlled of urban tree growth industrial Research Chair to CM, both co-funded by the Natural Science and Engineering Research Council of Canada (NSERC) and Hydro-Quebec. The study site is part of McGill University and their support is appreciated. We acknowledge funding from FRQNT, NSERC Discovery Grants to PAG, SWK, CM and BEB, and the Canada Research Chairs program. CRG was supported by a Juan de la Cierva fellowship (IJCI-2015-23505, MINECO, Spain). CAN was funded by a FQRNT postdoctoral fellowship.

FUNDING

This work was supported by the Natural Science and Engineering Research Council of Canada (NSERC) and Hydro-Quebec, as well as by the Fons de Reserche Nature et Technologies (FRQNT) and the Canada Research Chairs program.

Conflict of interest. None declared.