The taxonomic and functional biogeographies of phytoplankton and zooplankton communities across boreal lakes

Strong trophic interactions link primary producers (phytoplankton) and consumers (zooplankton) in lakes. However, the influence of such interactions on the biogeographical distribution of the taxa and functional traits of planktonic organisms in lakes has never been explicitly tested. To better understand the spatial distribution of these two major aquatic groups, we related the distributions of their taxa and functional traits across boreal lakes (104 for zooplankton and 48 for phytoplankton) to a common suite of environmental and spatial factors. We directly tested the degree of coupling in their taxonomic and functional distributions across the subset of common lakes. Phytoplankton functional composition responded mainly to properties related to water quality, while zooplankton composition responded more strongly to lake morphometry. Overall, the spatial distributions of phytoplankton and zooplankton were coupled at taxonomic and functional levels but after controlling for the effect of environmental drivers (water quality and morphometry) and dispersal limitation, no residual coupling could be attributed to trophic interactions. The lack of support for the role of trophic interactions as a driver coupling the distribution of plankton communities across boreal lakes indicates that taxon-specific and functional trait driven ecological interactions may not modulate large-scale spatial patterns of phytoplankton and zooplankton in a coordinated way. Our results point to community structuring forces beyond the phytoplankton-zooplankton trophic coupling itself, and which are specific to each trophic level: fish predation for zooplankton and resources for phytoplankton.


Introduction
dispersal-limited (Beisner et al. 2006, De Bie et al. 2012, Padial et al. 2014. Such differential 108 response patterns could also interfere with a strong biogeographical coupling of the groups. In 109 fact, differential relative responses of phytoplankton and zooplankton to environmental factors 110 and distances between lakes, make it necessary to control for their respective effects before 111 assessing the importance of trophic interactions on the biogeographical coupling of the two 112 plankton groups. 113 Here, we conducted the first comprehensive study that directly considers the relative 114 importance of trophic interactions between planktonic organisms and relates these to the effects whether the signature of trophic interactions could be captured at a regional scale, we assessed 120 the coupling between the taxonomic and functional distribution of lake zooplankton and 121 phytoplankton communities.By using a stepwise framework, we first accounted for the effect of 122 habitat selection and dispersal, and then assessed whether residual variation in the observed 123 coupling between species and traits could be attribution to trophic interactions. This approach 124 also allowed us to investigate the environmental and spatial factors influencing each group 125 (phytoplankton or zooplankton) separately as well. Our study covers a large biogeographical 126 range (longest distance between two lakes is 1 228 km) and examines more than 100 lakes 127 clustered in three regions that together characterize geological and environmental variation in the 128 boreal belt across one of Canada's largest provinces (Québec). Because trophic interactions are 129 mediated by functional traits, we expected to observe a stronger residual coupling between the distribution of zooplankton and phytoplankton traits compared to the taxonomical coupling. 131 Furthermore, we expected the coupling to be even stronger when only traits specific to trophic 132 interactions between zooplankton and phytoplankton were considered (e.g. pigment type, feeding 133 strategy). USA), hauled vertically from 1 m above the sediments to the surface. Zooplankton samples were 144 anaesthetized using carbonated water and were preserved in 75% (final concentration) ethanol.

145
In a subset of 48 lakes, the phytoplankton community was also simultaneously sampled over the 146 photic zone using a flexible PVC sampler tube and an integrated subsample (250 ml) was 147 preserved in Lugol's solution.

148
Crustacean zooplankton were identified at the species level (but aggregated for analyses 149 at the genus level to correspond to the phytoplankton data), using an inverted microscope (50-150 400X) and individuals were counted until a total of 200 individuals had been enumerated. For 151 each taxon present in a lake, the length of 20 mature individuals was measured and biomass by taxon was estimated using length-dry-mass regressions (McCauley 1984, Culver et al. 1985.

153
Phytoplankton were enumerated at the genus level using the Ütermohl method on an inverted 154 microscope at 400X magnification. Phytoplankton biomass was estimated from biovolume 155 computed using cell and colony length measurements and corresponding geometric forms 156 (Hillebrand et al. 1999). We also measured key limnological variables to characterize the lake 157 and catchment environments. We used a multiparameter sonde (YSI,Yellow Springs 158 Instruments, OH, USA) to measure pH (at 0.5m depth) and temperature (at 0.5m depth intervals, 159 then averaged over the water column). Water samples were collected at 0.5m depth to measure 160 the concentration of chlorophyll a (Chl-a), total phosphorus (TP), total nitrogen (TN), dissolved 161 organic carbon (DOC) and coloured dissolved organic matter (CDOM ). Chl-a was extracted 162 with 90% hot ethanol and absorption was measured spectrophotometrically before and after 163 acidification to account for phaeophytin (Lorenzen 1967, Nush 1980; TP was measured from 164 water samples using the molybdenum-blue method following persulfate digestion (Cattaneo and 165 Prairie 1995); TN was measured using nitrates after persulfate digestion; DOC concentration was 166 measured on an O.I. Analytical (Texas, USA) TIC/TOC using 0.45 μm filtered water after 167 sodium persulfate digestion; CDOM was measured using a UV/Vis UltroSpec 2100 168 spectrophotometer (Biochrom, Cambridge, UK) at 440 nm. Missing values (see Table 1) were 169 imputed using an approach based on random forest (missForest R package, Stekhoven and 170 Bühlmann 2012, NRMSE : 0.038). Lake depth was measured at sampling point using a Portable 171 Water Depth Sounder Gauge (Cole-Parmer). Lake area was derived using ArcGIS V10 software 172 (ESRI Inc., Redland, CA, USA) and catchment slope was estimated using a Digital Elevation 173 Model (Canadian Digital Elevation Data). To visualize environmental differences between the 174 three regions, we used a principal component analysis (PCA) using the rda function (vegan R package, Oksanen et al. 2015). Finally, because our sampling was discontinuous on the 176 landscape we used the Euclidean distance between lakes to characterize the effect of dispersal 177 limitation on the distribution of taxa and functional traits.

178
Functional trait composition and diversity 179 Given that our objective was to test for a significant coupling between adjacent trophic 180 groups, we selected functional traits ( Table 2) that explicitly characterize the grazing interaction 181 between phytoplankton and zooplankton, as well as their interaction with other trophic levels in 182 the food web. For phytoplankton, we selected traits that define how they interface with resources 183 (i.e. nutrients and light) and the zooplankton grazers (motility, edibility, colony formation). For 184 zooplankton, we focused on traits that define how they consume phytoplankton (i.e. feeding type  (Table 2). 195 To visualize how phytoplankton and zooplankton taxa and traits were distributed on the 196 landscape, we used the percentage of lakes in which they occurred. We tested for differences in taxonomic or functional composition between the three regions using the constrained ordination 198 technique Canonical Analysis of Principal Coordinates (CAP BiodiversityR package ;Anderson 199 and Willis 2003). Using the CAP leave-one-out allocation success (% correct, Anderson and 200 Willis 2003) we assessed the distinctiveness of regional composition using the proportion of 201 correct allocation, which can be interpreted as the strength of the compositional differences 202 between regions. Prior to the CAP ordination, taxonomic composition was Hellinger-transformed 203 to reduce the effect of double zeros (Legendre and Gallagher 2001) and Bray-Curtis distance was 204 used for the CAP ordination.

205
To test whether the distributions of phytoplankton and zooplankton communities were 206 coupled across the landscape, we used a hierarchical framework ( Figure 2). First, we tested for 207 significant coupling between the composition of the two groups at taxonomic and functional trait 208 levels using a Procrustes analysis (Mardia et al. 1980) on the subset of lakes (48) for which both 209 phytoplankton and zooplankton were sampled. Specifically, we tested the degree of concordance 210 between the PCA ordinations of phytoplankton and zooplankton taxonomic and functional 211 community compositions. For this analysis, species composition was transformed using the 212 Hellinger transformation Legendre 1998, Legendre andGallagher 2001), and 213 functional trait composition was transformed using a logit transformation. We tested the 214 significance of the Procrustes statistic using a permutation procedure (9999 simulation, protest 215 function in vegan R package, Oksanen et al. 2015). To further investigate whether the coupling 216 could be explained by a similar response of the two groups to environmental variation or by 217 dispersal limitation (see Figure 2) we also tested for a significant coupling, using Procrustes 218 analysis, after controlling for environmental factors (water quality characteristics and morphometry) and space (using between-lake distance based on latitude and longitude 220 coordinates).

221
To directly evaluate the relative importance of environmental and spatial variables as drivers 222 of the taxonomic and functional composition of phytoplankton and zooplankton, we used 223 distance-based redundancy analysis for taxonomic composition (dbRDA, Legendre and 224 Legendre 1998) and multiple regression for each functional trait; both followed by variation 225 partitioning (Borcard et al. 1992). We separated the environmental variables into two groups;

255
For both phytoplankton and zooplankton, all functional traits were present in all three 256 regions. All phytoplankton functional traits occurred in more than 90% of lakes, with the 257 exception of two traits associated with cyanobacteria: presence of a vacuole for motility (63% of 258 lakes) and the potential to fix nitrogen (58% of lakes). The occurrence of zooplankton functional 259 traits ranged between 18% for carnivores and omnivore-herbivores, and 100% for herbivores, 260 with the average occurrence of zooplankton traits being 64% (median 75%). Functional 261 composition between regions did not differ for either phytoplankton, or zooplankton (% correct = 262 35%, p=0.69 and 49%, p=0.15 respectively, Figures 3c and 3d).
In the taxonomic Procrustes analyses we found a significant correlation between 264 phytoplankton and zooplankton taxa (Table 3), indicating a taxonomic coupling between the 265 distribution of the two groups. However, the correlation was not significant after controlling for 266 water quality, morphometry and space together, suggesting that the observed taxonomic coupling 267 cannot be attributed purely to trophic interactions but rather to both groups responding to the 268 same environmental and spatial drivers. On the other hand, using functional traits, there was a 269 significant coupling between the plankton groups only when traits related to trophic interactions 270 between phytoplankton and zooplankton (i.e. without phytoplankton resource acquisition traits; 271 including only motility, colonial and biovolume; Table 2) were used, but not when all traits were 272 considered ( Table 3). The significant trait correlation did not remain after controlling for water 273 quality, morphometry or space (independently or together), indicating that the observed composition and most functional traits, a large portion of the variation was attributable to lake 304 morphometry (yellow bars; Figure 5). However, for most traits, and taxonomic composition, 305 some variation was either shared or explained independently by water quality.

306
In the subsequent RDA of zooplankton taxonomic composition, constrained by 307 morphometric and water quality variables (Figure 6b), variables related to lake productivity 308 (Chla, TP and TN), carbon content (DOC) and temperature loaded on the first axis, while variables relate to lake morphometry (lake depth and area) loaded more strongly on the second 310 axis. For functional traits (Figure 6d), the first axis was mainly related to lake depth, with the 311 second being related to lake area. The raptorial feeding trait was highly correlated with deep 312 lakes and C-filtration with shallow lakes. B-filtration was positively correlated with lake area and 313 size was negatively correlated with lake area while D-filtration was related to large and deep 314 lakes and Stationary suspension to small and shallow lakes. In the taxonomic and functional 315 RDA, the first two axes were significant.

317
We observed a significant coupling between the landscape distributions of phytoplankton 318 and zooplankton at both taxonomic and functional levels. However, after taking into account the 319 coupling explained by environmental filtering and limitations to dispersal, we showed that in 320 boreal lakes, no residual coupling between phytoplankton and zooplankton could be attributed to 321 trophic interactions directly. Overall, the distribution of phytoplankton and zooplankton taxa and 322 traits was strongly influenced by water quality and lake morphometry and to a lesser extent by 323 dispersal limitation and these emerge as the main drivers coupling the distribution of 324 phytoplankton and zooplankton across boreal lakes. Moreover, phytoplankton and zooplankton 325 taxonomic composition displayed strong regional differences across the entire studied landscape, 326 but not when considered through a functional trait lens. Thus, although we observed important 327 taxonomic differences between regions, these differences did not translate into functional 328 differences between regions, despite the large environmental gradients covered.

329
Regional differences in composition and environment three regions indicate a role for environmental factors or dispersal in influencing distributions at 332 the regional scale. On the other hand, functional composition did not differ between regions, 333 indicating that environmental variation or dispersal distances could not prevent any traits from 334 being distributed across the entire landscape. However, the biogeographical overlap in traits 335 does not mean that functional composition was similar across all lakes. Environmental factors 336 related to water quality and lake morphometry explained a significant amount of variation in 337 relative biomass of most functional traits across both plankton groups ( Figure 5), indicating that 338 the control of functional composition acts at sub-regional scales. We also observed that control 339 of plankton taxonomic and functional composition by these lake characteristics was far more 340 important than was the effect of dispersal limitation. Finally, comparing across plankton groups, 341 dispersal limitation was more important for zooplankton than phytoplankton, supporting 342 previous work (Beisner et al. 2006, De Bie et al. 2012, but here now also including functional . This is to be expected based on previously established strong relationships 350 between composition and lake nutrient status for broad taxonomic groups, groups which were 351 reflected in our functional (pigment) categorization (Watson et al. 1997). On the other hand, morphometry. However,some water quality effects were also evident for zooplankton, indicating 354 overall an integrated response of zooplankton functional and taxonomic composition to their 355 proximal environment (water quality) and habitat characteristics (lake morphometry).

356
Zooplankton response to water quality could either occur as a direct effect, or as an 357 indirect bottom-up response to changes in phytoplankton community structure. For the more 358 prevalent response of zooplankton to lake morphometry, this reflects variation in habitat 359 characteristics, such as differences in lake physics (e.g. thermal stratification) or lake depth. We 360 consider it most likely that this morphometry effect arose indirectly, operating via the trophic 361 influence of fish predation, as previously observed (O'Brien et al. 2004). Specifically, lake depth 362 influences fish community composition (Jackson and Harvey 1989),with larger volume lakes 363 tending to have longer food chains (Post et al. 2000), thereby modulating the trophic cascade 364 effect on zooplankton through planktivore fish feeding (Carpenter et al. 1985). The variation in  (Christoffersen et al. 1993). Repercussions throughout the community were observable in the 371 ordination biplots (Figure 6d), with reductions in D-filtration being related to an increased 372 proportion of stationary suspension herbivory, dominated by calanoid copepods. Also, the 373 relative biomass of the C-filtration group was negatively related to lake depth, consistent with 374 the fact that most species within this functional feeding type are littoral species and shallow lakes 375 contain greater proportion of habitats that are littoral.
Using the common currency of functional traits, the biogeography of phytoplankton and 378 zooplankton were not coupled after controlling for environment and space. Even by only 379 including phytoplankton traits reflecting trophic interactions (grazer avoidance traits) no 380 coupling could be attributed to interactions between functional groups. Because no coupling 381 between phytoplankton and zooplankton was observed after controlling for either water quality, 382 morphometry or space, it suggests that one of these is the main driver of the observed coupling.

383
Because water quality was an important driver for most phytoplankton, as well as zooplankton 384 traits (see Figure 5), we can hypothesize that water quality was the main driver of the coupling 385 between phytoplankton and zooplankton traits. The lack of support for a coupling induced by  Taxonomically, we also observed a significant coupling between phytoplankton and 391 zooplankton distribution and one that was stronger than observed functionally. The coupling 392 between phytoplankton and zooplankton taxonomy only became non-significant after controlling 393 for the combined effects of water quality, morphometry and space (Table 3), suggesting that the 394 observed coupling resulted from a combined effect of these three groups of drivers across boreal 395 lakes. While the greatest decrease in coupling was observed after controlling for space (Table 3), 396 spatial factors could not explain any independent variation in the distribution of phytoplankton 397 taxa ( Figure 5). However, as most of the variation explained by water quality in the distribution 398 of phytoplankton and zooplankton (taxonomic) was spatially structured (light blue bars in Figure 5) the observed spatial impact is likely not a result of dispersal limitation, but rather the 400 result of water quality variables that are spatially structured. Nevertheless, the coupling we 401 observed at the taxonomic level is consistent with Bowman and colleagues (2008) who observed 402 a spatial concordance between the community composition of phytoplankton and zooplankton in 403 eight Canadian lakes.

404
It is also important to acknowledge that the coupling between plankton groups that we to vary differently across variable types, with for example, water quality variables generally 411 having stronger spatial structure than those related to lake morphometry (Lapierre et al. 2015).

412
This difference in spatial structuring could disrupt factors like trophic interactions or nutrients 413 that are likely to promote joint distributions. Proximate (water quality) variables with their strong 414 spatial structure across landscapes should therefore favour greater plankton biogeographical 415 concordance than would less structured habitat variables characterizing lake morphometry.

416
Unfortunately, mechanistic links such as these are difficult to verify with observational studies 417 like ours at these spatial scales, and would require smaller-scale experimental study to fully 418 verify.

419
The goal of including functional traits was to assess whether biogeographical coupling 420 would be more easily observed using a more mechanistic aggregation of organismal 421 differentiation than is done by pure taxonomy. To this end, we selected functional traits that are directly representative of interactions between phytoplankton and zooplankton and in resource 423 acquisition. A lack of coupled biogeographies, after controlling for environmental variables was 424 observed. Hence, for the suite of feeding related functional traits used in this study, we found no 425 support for direct reciprocal influences of zooplankton and phytoplankton community structures 426 across the landscape. Overall, our study points to other food web components such as fish 427 predation and resource availability as independent drivers of each group's biogeography .

428
Overall, although phytoplankton and zooplankton are known to share strong trophic connections 429 in individual lakes (Porter 1977, Sterner 1989, which constitutes the basis of the main pathway         was constrained by environmental variables related to water quality (in blue) and by