Light availability impacts structure and function of phototrophic stream biofilms across domains and trophic levels

Abstract Phototrophic biofilms are ubiquitous in freshwater and marine environments where they are critical for biogeochemical cycling, food webs and in industrial applications. In streams, phototrophic biofilms dominate benthic microbial life and harbour an immense prokaryotic and eukaryotic microbial biodiversity with biotic interactions across domains and trophic levels. Here, we examine how community structure and function of these biofilms respond to varying light availability, as the crucial energy source for phototrophic biofilms. Using metatranscriptomics, we found that under light limitation‐dominant phototrophs, including diatoms and cyanobacteria, displayed a remarkable plasticity in their photosynthetic machinery manifested as higher abundance of messenger RNAs (mRNAs) involved in photosynthesis and chloroplast ribosomal RNA. Under higher light availability, bacterial mRNAs involved in phosphorus metabolism, mainly from Betaproteobacteria and Cyanobacteria, increased, likely compensating for nutrient depletion in thick biofilms with high biomass. Consumers, including diverse ciliates, displayed community shifts indicating preferential grazing on algae instead of bacteria under higher light. For the first time, we show that the functional integrity of stream biofilms under variable light availability is maintained by structure–function adaptations on several trophic levels. Our findings shed new light on complex biofilms, or “microbial jungles”, where in analogy to forests, diverse and multitrophic level communities lend stability to ecosystem functioning. This multitrophic level perspective, coupling metatranscriptomics to process measurements, could advance understanding of microbial‐driven ecosystems beyond biofilms, including planktonic and soil environments.


| INTRODUCTION
Taxonomic and phylogenetic diversity in natural microbial communities is massive (Locey & Lennon, 2016;Torsvik, Goksøyr, & Daae, 1990). Despite the methodological leaps that have enabled description of this biodiversity, from deep sequencing of the prokaryotic rare biosphere (Sogin et al., 2006) to phylogenetic inference of novel prokaryotic and eukaryotic lineages (Hug et al., 2016;Jones et al., 2011), a tendency to address prokaryotes and eukaryotes separately has prevailed in most biodiversity studies. This inhibits the understanding of the ecology of complex microbial communities that include organisms across all domains of life on several trophic levels.
Metatranscriptomics has the potential to overcome these barriers by addressing structure and function of entire microbial consortia while avoiding common biases associated with PCR amplification of marker genes (Urich et al., 2008). This is made possible by parallel analysis of small subunit ribosomal RNA (SSU rRNA), as a common currency for living biomass of all three domains of life and of the mRNA from actively transcribed functional genes providing insights into the functional capacity of the community. In addition, metatranscriptomics and related molecular techniques have advantages over traditional microscopic identification of (eukaryotic) morphospecies, which may underestimate microbial diversity due to time-consuming manual counting of cells, and preferential detection of taxa with distinct morphologies over inconspicuous cells which can comprise hidden phylogenetic and functional diversity (e.g., Jones et al., 2011;Liu et al., 2009).
In complex microbial habitats such as soils, metatranscriptomics has contributed to the discovery of relatively abundant but previously largely undetected taxa (Geisen et al., 2015). Metatranscriptomics has also elucidated the functional adaptations of microbial communities in marine plankton, for example in response to iron limitation (Marchetti et al., 2012), and has helped to unravel complex interactions between diatoms and bacteria (Amin et al., 2015) and resource partitioning between diatoms (Alexander, Jenkins, Rynearson, & Dyhrman, 2015). For biofilms, metatranscriptomics has mainly been employed to study clinically relevant biofilms, such as gene expression in oral biofilms associated with periodontitis (Frias-Lopez & Duran-Pinedo, 2012;Yost, Duran-Pinedo, Teles, Krishnan, & Frias-Lopez, 2015). However, studies that link structure and function in complex environmental biofilms are rare (but see Lindemann et al., 2017), despite the importance of such biofilms for primary production, carbon and nutrient cycling in various aquatic ecosystems.
Stream biofilms are a prime example of phylogenetically diverse, multitrophic level communities. They form the trophic basis of the stream ecosystem contributing the bulk of primary production and biomass (reviewed in (Battin, Besemer, Bengtsson, Romani, & Packmann, 2016). Stream biofilms harbour diverse phototrophic algae, cyanobacteria, heterotrophic bacteria and various eukaryotic consumers, among them even early-instar insect larvae. Several studies have highlighted the phylogenetic diversity and important roles of each of these functional groups separately (Besemer et al., 2013;Bott & Kaplan, 1990;Schmid-Araya, 1994;Wellnitz & Rader, 2003), yet only rarely in combination (Piggott, Salis, Lear, Townsend, & Matthaei, 2014;Romani et al., 2014). Thus, there is limited understanding of how environmental variability and change affects structure and function of organisms of different biofilm functional groups simultaneously, including interactions between them, and how this influences their collective contribution to biogeochemical fluxes.
Light is a primary resource in streams, which fuels photosynthesis by algae and cyanobacteria in biofilms and therefore lays the basis of the carbon (C) and nutrient (N and P) cycles in streams in addition to terrestrial C inputs (allochthonous C). Although the importance of light in regulating stream food webs and biogeochemical fluxes is widely acknowledged, most studies to date have focused on one or a few functional groups when assessing the effects of altered light regimes on stream organisms (Hill, Ryon, & Schilling, 1995;Romani et al., 2014;Ylla, Borrego, Romani, & Sabater, 2009 Hill & Fanta, 2008;Hill, Roberts, Francoeur, & Fanta, 2011).
As predicted by this hypothesis, increased production and exudation of C-rich photosynthates by phototrophs under high light, P-depleted conditions, also fuel bacterial heterotrophic metabolism. However, it is currently unclear how this affects the uptake of limiting P by phototrophs and heterotrophic bacteria in the biofilm. Moreover, increased light availability may cause a shift in the relative importance of autochthonous (stream phototroph-derived) and allochthonous (terrestrial-derived) carbon sources for bacterial heterotrophs, with unknown consequences for biofilm structure and function. Understanding how light modulates stream biofilm structure and function is fundamental, as global climate change and human activities alter light regimes in many streams through for example deforestation (increased light), but also through upward shifts of the tree line in alpine areas due to global warming (decreased light; Lenoir, Gegout, Marquet, de Ruffray, & Brisse, 2008). Ultimately, light availability in stream biofilms is analogous to primary resource variability in many microbially driven ecosystems. For example, planktonic communities experience resource variability in light and nutrient levels resulting in bloom and nonbloom states with significant impacts on structure and function of both algal, bacterial and consumer communities. In soils, variations in above-ground primary productivity also affect microbial biomass and community composition, and vice versa (Wardle et al., 2004). Despite this, studies investigating the impacts of resource variability on several microbial functional groups and trophic levels at the same time remain rare in any ecosystem, hindering insights into the links between structure, function and biogeochemical processes in complex microbial communities.
We experimented with phototrophic biofilms initially grown in headwater streams and further incubated in microcosms to determine how light availability impacts structure and function of biofilms across all three domains of life, assessed using metatranscriptomics. We expected that light availability would impact all microbial functional groups, leading to shifts in both community composition and function. More specifically, we hypothesized that functional groups (i.e., phototrophs, heterotrophic bacteria and microbial consumers) would respond differently reflecting their unique roles within the biofilm matrix: • Overall, we expected that shifts in community composition in response to light availability would be more pronounced for phototrophs than for heterotrophic bacteria and consumers, as the latter functional groups should display only a secondary response to light mediated by changes in phototroph biomass, activity and community composition.
• Phototrophs (algae and cyanobacteria) should display a shift in community composition to more shade-tolerant taxa under lower light availability accompanied by functional shifts in expression of genes involved in photosynthesis. Additionally, genes involved in P uptake should be overexpressed in treatments with higher light availability, reflecting the lower availability of this limiting nutrient relative to light (Sterner et al., 1997).
• Heterotrophic bacteria should also display a shift in community composition due to increased availability of photosynthetic exudates in higher light availability relative to allochthonous C sources. Moreover, genes involved in P uptake should be overexpressed in higher light availability as a result of reduced relative P availability and increased bacterial P demand due to metabolic stimulation by increased phototroph exudation.
• Biofilm consumers such as microbial grazers will experience a shift in their food source as light availability alters the community composition within the phototroph and heterotrophic bacteria functional groups, causing a secondary shift in the community composition of consumers.

| Biofilm establishment
Biofilms were grown in a flume with diverted stream water from the Oberer Seebach (Lunz am See, Austria) during a period of 4 weeks in June 2012. Glass tiles of approximately 1 cm 2 were used as a substrate and were covered by shading foils to manipulate the light availability for the growing biofilms. Biofilm-covered tiles from three different light conditions, 92%, 51% and 7% light transmission, were transferred into triplicate microcosms. For a detailed explanation of the biofilm establishment, see Wagner, Besemer, Burns, Battin, and Bengtsson (2015).

| Microcosm incubations
Using laboratory microcosm incubations after biofilm establishment, we could measure the contribution of the biofilms to essential processes such as primary production and nutrient uptake. Microcosms consisted of 0.74 L, gas-tight transparent plexiglas chambers, equipped with magnetic stirring and a perforated support which accommodated the biofilm-covered glass tiles (Wagner et al., 2015).
The top of each microcosm was covered with the respective shading foil (92%, 51% and 7% light transmission), and all microcosms were illuminated with the same fluorescent lights. The resulting light intensities ranged between 10 and 150 micromole photons m À2 s À2 with slight variability within the three light treatments caused by differences in the placement of microcosms in relation to the light source ( Table 1). The light intensities are within the variability that may be encountered in a stream draining a forested catchment, where riparian vegetation shades the streambed from direct sunlight (Hill et al., 1995). Microcosm incubations were carried out in a climate chamber at 17.5°C (AE2°C). There was no difference in temperature between light treatments, yet the temperature was elevated in all treatments compared to stream conditions during biofilm establishment (10.2°C AE 1.6°C). Each microcosm was filled with 90 tiles (each~1 cm 2 ) at the start of the 7-day experiment, and three tiles were harvested on days 1, 3 and 7, flash-frozen in liquid N 2 and stored at À80°C until nucleic acid extraction. Microcosms were filled with a stream water medium consisting of raw oligotrophic groundwater from the region supplemented with an aged extract of crack willow (Salix fragilis) leaves to mimic terrigenous DOM (Wagner et al., 2014). The entire volume in each of the microcosms was exchanged at the start of each day and night incubation, which lasted for 6 hours. Concentrations of O 2 , DOC, NO 3 and PO 4 were measured at the start and end of each day and night incubation (as described in Wagner et al., 2015), and the uptake/release of solutes was calculated as the difference between start and end, normalized to the biofilm area and expressed as an hourly rate. Gross primary production was calculated from the daytime oxygen release (net primary production) T A B L E 1 Biofilm biomass parameters and biofilm-mediated fluxes in response to light availability. Values represent means of treatment replicates (n = 3), AE1 SD. All light treatments represent light levels typical of stream reaches shaded to various degrees (not exposed to full sunlight)

| Metatranscriptomics
Total nucleic acids (NA) were extracted from each tile with a phenol-chloroform-based method featuring mechanical lysis (Urich et al., 2008). The extracts of tiles from the different time points from each microcosm were pooled using equal NA amounts. Thus, one timeintegrated sample from each microcosm (n = 9) comprised NA extracted from approximately 9 cm 2 of biofilm. Nucleic acid extracts were treated with DNAse, and RNA was purified using the AllPrep kit (Qiagen) and subjected to linear amplification using MessageAmp II kit (Ambion, Tveit, Urich, & Svenning, 2014
Ribosomal RNA reads classified on different taxonomical levels were exported from MEGAN, while function-assigned reads (mRNA) were exported from MGRAST. To link functional gene tags to microbial taxa, functionally annotated and taxonomically assigned mRNA reads were exported from MGRAST using the workbench feature. To address differential abundance of microbial taxa (rRNA) and differential expression of microbial functional categories (mRNA) in response to light, read count data were analysed using a statistical framework for RNA-Seq data implemented in the EDGER package in R (R Develop-

| Light availability influenced biofilm-mediated biogeochemical fluxes
We found that experimental shading resulted in several changes in biofilm properties and biofilm-mediated fluxes, summarized in Table 1.
Notably, primary production and biofilm biomass were reduced by a factor of 3 between the highest and the lowest light availability treatments, also visible to the naked eye as a thicker biofilm cover in the   Table S2). For example, several genus-level taxa within

| Dominant phototrophs responded to light limitation through shifts in expression of functional genes involved in photosynthesis
Betaproteobacteria and Verrucomicrobia increased in abundance when light availability and hence primary productivity were low (Figure 4c, Table S2); they may be primarily involved in decomposition of allochthonous or detrital organic matter (Newton, Jones, Eiler, McMahon, & Bertilsson, 2011).
A parallel study employing the same microcosms demonstrated a temporal shift in the bacterial communities during the week-long experiment (Wagner et al., 2015). Due to our time-integrated sampling strategy (time points were pooled owing to the relatively time-and resource-intense sample processing for metatransciptomics), we cannot address temporal shifts in structure and function that may be caused for example by natural succession or temperature and light regime differences between the stream and the microcosms. Our results therefore reflect changes in response to light availability that are strong enough to be detected despite possible temporal shifts.

| Higher light availability led to increases in expression of functional genes involved in phosphorus metabolism in bacteria
Thick biofilms developing in our high light treatment depleted P-PO 4 concentrations in the experimental microcosms (Figure 1b), which likely also led to even more severe local phosphorous (P) depletion within the biofilm where mass transfer becomes the limiting step compared to thinner biofilms (Battin, Kaplan, Newbold, & Hendricks, 2003). As in many oligotrophic freshwater ecosystems, P is the limiting inorganic nutrient in our study stream, where it puts constraints on primary production and biomass build-up. P limitation as a result of increased light availability is a key prediction of the nutrient-light hypothesis (Sterner et al., 1997), and we hypothesized that P limitation would lead to changes in expression of functional genes involved in P uptake. Biofilms exposed to the highest light availability indeed displayed a 1.5-fold increase (p < .05) in relative abundance of mRNAs related to P metabolism (Table S3). The protein functions contained within P metabolism were diverse but included periplasmic phosphate-binding proteins (TC 3.A.1.7.1), PO 4 transport system regulatory proteins and alkaline phosphatase (EC 3.1.2.1).
Most of the mRNAs involved in P metabolism were classified as bac- F I G U R E 5 The relative abundances of biofilm functional gene transcripts. The Krona chart (centre) displays overall relative abundance of mRNA reads assigned to functional categories (SEED subsystems) of all samples (n = 9) while a-c displays mRNA reads involved in the key functions photosynthesis (a,c) and phosphorous metabolism (b) in response to the different light conditions. Read counts were normalized to assigned mRNA reads per sample and expressed as a percentage. Bars are shaded to indicate the light condition (darker shade = lower light intensity). Error bars refer to standard deviation of the mean (n = 3), and significant differences between the light treatments are indicated with asterisks (*p < .05) according to subsystem-wise negative binomial GLMs with Benjamini-Hochberg correction for multiple testing (as implemented in EDGER) more relevant in biofilms due to the close proximity of algal and bacterial cells allowing exchange of solutes within the biofilm matrix.
However, we cannot exclude the possibility that mRNAs involved in P metabolism could not reliably be assigned to diatoms due to the low representation of annotated diatom genomes in the public databases compared to bacteria.

| Community shifts in eukaryotic biofilm consumers may reflect shifts in abundance of their prey along the light gradient
We detected diverse eukaryotic micrograzers, including ciliates (Ciliophora), Cercozoa and Amoebozoa, and even larger predators such as chironomid larvae and rotifers (Figure 3 centre). In addition, we detected many potentially parasitic taxa, such as chytrid fungi (Chytridiomycota) and vampyrellid rhizarians, which likely reduce the productivity of the dominant algae (Kagami, Miki, & Takimoto, 2014). On average, these consumers comprised 15%-25% of the biofilm SSU rRNA (Figure 3 centre, Figure S3, red shades), and the relative abundance of this functional group as a whole did not vary along the light gradient, but appeared to be relatively stable compared to phototrophic eukaryotes and bacteria ( Figure S3). However, as hypothesized, light availability induced striking community shifts within specific taxa. For example, among the ciliates, several genera increased or decreased significantly with light ( Figure 4c, Figure S4

| Biofilm functional groups responded to light availability in unique ways
As detailed above, all functional groups responded to light availability through community shifts, consistent with our hypotheses.
However, surprisingly, the most abundant algal and cyanobacterial taxa did not change in relative abundance ( (Kolber et al., 2001;Stoecker, 1998) and may contribute to the observed patterns to some extent.
Studies that link structure and function of phototrophic biofilms on a molecular level are rare, yet the few existing studies report a composition of organisms and gene expression under stable conditions that is a broadly similar to our results (Krohn-Molt et al., 2013;Nakamura et al., 2016;Schneider, Reimer, Hahlbrock, Arp, & Daniel, 2015). In a recent study, Lindemann and colleagues investigated protein expression patterns during succession in cyanobacterial biofilms (Lindemann et al., 2017). They found that P starvation during 28 days of succession caused expression of similar P uptake genes in all bacteria, while nitrogen (N) uptake strategies varied more among microbial taxa. They suggest that, under severe P limitation, niche complementarity allowing for coexistence may be achieved by differing N uptake strategies.
In contrast to the sparsity of biofilm "omics" studies, a considerable body of literature deals with plankton, including conditions characterized by contrasting resource availability such as phytoplankton blooms and diel dynamics. In a diatom-dominated marine phytoplankton bloom, phosphorous depletion was followed by increased expression of several different bacterial phosphorous acquiring proteins, showing that bacteria are essential for P cycling under high resource (C) availability also in this environment (Teeling et al., 2012). In an oligotrophic mountain lake, bacteria appeared to employ different P acquiring strategies during day (high resource availability) than during night (low resource availability), relying on inorganic P during the day and organic P during the night (Vila-Costa, Sharma, Moran, & Casamayor, 2012). Although we found no indications of a similar P uptake strategy in our high-light vs. low-light-grown biofilms in this study, we did detect an intensified utilization of allochthonous carbon during the night and under darker conditions in a parallel study from the same experiment (Wagner et al., 2017).
This indicates that scavenging of organic molecules during dark conditions may release additional P to biofilms. biofilms are functionally adapted to persist and carry out essential biogeochemical processes even under very limiting light. This resilience of biofilm communities likely contributes to the stability of stream ecosystems, and their ability to recover from events that alter light regimes on the short term (such as shading by suspended solids after storms), and adapt to long-term changes, including effects of riparian de-and reforestation (Hill et al., 1995;Stephenson & Morin, 2009).

| Ecological implications of the functional integrity of biofilms along changing light regimes
Variations in primary resource availability are ubiquitous in nature. As already mentioned, phytoplankton blooms result from seasonal changes in light and nutrient availability. In soils, variations in inputs from above-ground primary production cause differing resource availability for soil microbial communities (Wardle et al., 2004). There are some notable and fundamental differences between these situations and the stream biofilms that we have experimented on under variable light conditions. For example, while we observed a stable relative contribution of dominant phototroph taxa along the light gradient, phytoplankton blooms are characterized by a distinct succession of dominant taxa as the bloom proceeds, likely due to the different temporal and physiochemical mechanisms underlying blooms in addition to just light availability. On the other hand, soils exposed to reduced organic resource input due to tree girdling experience a rapid drop in microbial biomass (Dannenmann et al., 2009), but not necessarily in microbial community composition (Dannenmann et al., 2009;Wu et al., 2011). It is possible that biofilm and soil microbial communities are more buffered against changes in primary resource input than planktonic communities, responding mainly with changes in biomass rather than community composition. If so, it may be due to their structured nature, where microbial cells coexist in close proximity, offering opportunities for interactions (Shade et al., 2012). In the case of the stream biofilms studied in this work, a few dominant phototrophic taxa with high phenotypic plasticity appear to confer this stability by providing a structured habitat and fixed carbon to numerous biofilm heterotrophs, possibly in exchange for P recycling by biofilm bacteria.

| Biofilms as "microbial jungles"
Our view into the structure and function of complex biofilms across all domains of life, which was enabled by metatranscriptomics, highlights that structured microbial consortia in nature are best understood as multitrophic entities whose function is determined by microbial interactions across trophic levels and domains. Such "microbial jungles" are the rule rather than the exception in microbial habitats, which rarely exclusively contain organisms from one domain or functional group (such as algae) although it may seem convenient to limit the scope when studying these communities.
Analogous to macroscopic jungles where trees provide the threedimensional structure that houses tremendous biodiversity (Ellison et al., 2005;Lowman & Rinker, 2004), dominant diatoms and cyanobacteria in stream biofilms act as foundation taxa that build a structured habitat facilitating other organisms such as heterotrophic bacteria and microscopic consumers. We have shown how these dominant phototrophs offer stability to the biofilm via their phenotypic plasticity, similarly to how long-lived trees in a forest withstand short-term climatic variability and thereby protect its associated flora and fauna (Bormann & Likens, 1994). Long-term diversity and stability in forests can be governed by species interactions such as insect herbivory and fungal pathogenicity (Bagchi et al., 2014;Janzen, 1970), whereas in biofilms, microscopic consumers such as ciliates and fungal parasites may regulate biomass and diversity. In conclusion, phototrophic biofilms contain more diversity than just the dominant phototrophs that first meet the eye, and a multitrophic level perspective such as is ubiquitous for macroscopic habitats could significantly enhance our understanding of these microscopic worlds. As studies which address structure and function across domains and trophic levels become more common, it will become apparent how interactions within complex microbial jungles such as stream biofilms regulate their response towards environmental change, and how this impacts the ecosystem fluxes that they mediate.

ACKNOWLEDG EMENTS
The authors wish to acknowledge Nancy Burns, Erik Herberg, Celina Pieber and Roseanne Kreimel for participation in the microcosm experiment, Kristin Rath for assistance in the laboratory and Stephanie Markert for constructive comments on an earlier version of the manuscript. This study was funded through a Marie-Curie Individual Fellowship awarded to MMB (EU, FP7, PIEF-GA-2010-274895) and through two Austrian Science Fund projects awarded to TJB (P23420-B17 and Y420-B17).

DATA ACCESSIBILI TY
The raw metatranscriptomics sequences are available at the NCBI

AUTHOR CONTRI BUTION
The