Spatial and temporal changes of parasitic chytrids of cyanobacteria

Parasitism is certainly one of the most important driving biotic factors of cyanobacterial blooms which remains largely understudied. Among these parasites, fungi from the phylum Chytridiomycota (i.e. chytrids) are the only eukaryotic microorganisms infecting cyanobacteria. Here, we address spatiotemporal dynamics of the cyanobacterial host Dolichospermum macrosporum (syn. Anabaena macrospora) and its associated chytrid parasites, Rhizosiphon spp., in an eutrophic lake by studying spatial (vertical, horizontal) and temporal (annual and inter-annual) variations. Our results show homogenous chytrid infection patterns along the water column and across sampling stations. However, the prevalence of infection presented drastic changes with time, at both intra- and inter-annual scales. In 2007, a maximum of 98% of vegetative cells were infected by R. crassum whereas this fungal species was not reported seven years later. In opposite, R. akinetum, a chytrid infecting only akinetes, increased its prevalence by 42% during the same period. High chytrid infection rate on the akinetes might have sizeable consequences on host recruitment (and proliferation) success from year to year, as supported by the recorded inter-annual host dynamics (affecting also the success of other chytrid parasites). The spatial homogenous chytrid infection on this cyanobacterium, coupled to both seasonal and inter-annual changes indicates that time, rather than space, controls such highly dynamic host-parasite relationships.


Results and Discussion
The cyanobacterial host community was exclusively composed of D. macrosporum, which largely dominated the phytoplankton community, accounting up to 87.41 ± 0.02% (i.e. mean for triplicates ± SD) of the phytoplanktonic cells on Oct. 14 th . Based on the morphology of the sporangium and the type of infected cell (vegetative cells or akinetes), two parasitic chytrid species of D. macrosporum were identified: Rhizosiphon crassum and R. akinetum. R. crassum is reported mainly on vegetative cells and infects several cells with a tubular rhizoid system, whereas R. akinetum infected exclusively single mature akinetes without expanding to adjacent cells 23 . In addition, both species presented significant differences in terms of both prevalence of infection and abundance, with a maximal abundance of R. crassum approximately 11 fold lower than that recorded for R. akinetum in 2011. R. crassum infected a maximum of 0.68 ± 0.04% of vegetative cells reaching maximal sporangia density of 16 ± 8 × 10 3 sporangia.l −1 recorded on Oct. 7 th at 2 m in the central station (CS). Because all life stages of R. crassum were not observed, and this chytrid presented a very low abundance in 2011, vertical and horizontal distributions of sporangia were only described with more details for the R. akinetum-D. macrosporum pairing.
A stratified host community homogenously infected. Akinetes distribution in the water column did not show any significant difference within the two stations (CS and LS) (Mann-Whitney; P > 0.05). In the same way, we did not report significant differences in vegetative cell abundances. Nonetheless, we noticed an interesting difference by taking into account the vertical distribution since akinete abundances were significantly lower at the deepest layers regardless of the stations (Mann-Whitney; P = 0.03, Fig. 1). For example, on Oct. 14 th , 2.65 ± 0.21 × 10 5 ak.l −1 and 2.68 ± 0.36 × 10 5 ak.l −1 were recorded at 0.5 m at CS and LS, respectively, whereas less than the half were reported for the deepest depths (CS: 1.12 ± 0.22 × 10 5 ak.l −1 and 1.02 ± 0.03 × 10 5 ak.l −1 at 6 m and 9 m, respectively, and LS: 1.15 ± 0.03 × 10 5 ak.l −1 at 4 m). This trend was reported for all dates and stations but was less pronounced at the last sampling date, probably due to an increase of the mixing layer (from 5 m on Oct. 7 th to 8 m on Oct. 21 st ).
Similar to other parasites, chytrids are known to be closely linked to their host density [24][25][26] . Thus, we attempted to elucidate if akinetes and their parasites R. akinetum were produced simultaneously in the same place. Based on akinetes/vegetative cells ratios ( Supplementary Fig. 2) it appears that akinetes differentiation and maturation first occur in the upper layers of lake. Despite numerous studies on this topic, the external factor triggering the process of akinete differentiation is still not well understood. Nutrient depletion [27][28][29] , low temperature 30, 31 and high light intensity 32 have been identified as drivers inducing akinete differentiation. In our study, the temperature was homogenous along the epilimnion, allowing its mixing. Thus, we may hypothesize that the nutrient concentrations would not have been enough contrasted to explain such akinete distribution. Light availability is higher at the lake surface which may explain why akinetes were produced in the upper layer of the water column. Light intensity is known to play a crucial role in the chytrid infection as well 24 . Also, fungi exhibit phototaxis 33 , implying that infection parameters (Prevalence (Pr) and Intensity (I) of infection) should be higher in the upper water layers. However, for all the sampling periods, except for one-time point and depth (on Oct. 7 th at 9 m), a homogenous fungal epidemic along the water column was reported ( Fig. 2) suggesting that phototaxis would be a minor driving factor for the vertical distribution of R. akinetum infection. Van den Wyngaert et al. 34 investigated the role of chemotaxis in chytrid-phytoplankton interactions. These authors showed that chemotaxis becomes more relevant under low host density condition due to decreasing chance contact rates between host and parasite. Then, the uniform fungal infection observed along the water column may be explained by the high capacity of R. akinetum to detect its host even at low host densities, as reported in the deepest water layers.

R. akinetum infection efficiency: when time prevails on space. The chytrid infection was spatially
homogenous during the sampling period in 2011, but presented significant temporal variations. Indeed, only 4.7 ± 1.2% of akinetes were infected by chytrids at the first date whereas an average value of 31.9 ± 3.2% was reported along the water column at CS, on Oct. 21 st (Fig. 2a-f).
Moreover, the intensity of infection did not reveal any difference between the first two dates but presented significantly higher values within the lake at last one (1.42 ± 0.06 spor.ak −1 ; P = 0.04). Hence, the R. akinetum population significantly increased throughout the sampling period (P = 0.018) from an average along the water column of 1.12 ± 0.23 × 10 3 spor.l −1 on Oct. 7 th to 1.33 ± 0.08 × 10 5 spor.l −1 on Oct 21 st . Additionally, all R. akinetum parasitic life cycle stages were observed for each date in all samples and the comparison of their relative abundances did not reveal significant vertical or horizontal differences (Fig. 2a-c), highlighting a synchronized parasite population within the two sampled stations.
The chytrid population increase was not linear and presented a significantly higher increase during the first seven days. The density of chytrids increased approximately 30-fold between the 7 th and the 14 th of October, but only 5-fold during last week (Figs 2a,b and 3). The first important increase of R. akinetum could be explained through the combination of both the life cycle duration and the success of infection. Based on the life cycle duration that we previously estimated at 3 days for R. crassum in the same lake 23 , we can expect the achievement of at least a whole life cycle in one week for R. akinetum. On 7 th October, R. akinetum presented a majority of sporangia involved in the maturation phase (48.6 ± 9.4% and 57.2 ± 3.3% for CS an LS, respectively ( Fig. 2b). At this date, the biovolume of mature sporangia averaged 1749.8 ± 618.1 µm 3 at CS and did not differ (P > 0.05; Kruskal-Wallis) from the value observed at the LS (1647.7 ± 594.9 µm 3 ). By using the conversion factor 35 (CF) of 0.0172 zoosp.µm 3 we established a theoretical capacity of each infecting sporangium to produce 30.01 ± 10.61 and 28.3 ± 10.2 zoospores for CS and LS, respectively. Bruning 24 reported a mean infective lifetime of Rhizophydium planktonicum zoospores roughly about 8 days under laboratory conditions. Once released, these zoospores infect akinetes, explaining the dominance of young stages among the population of R. akinetum one week later on Oct. 14 th (Fig. 2b, 59.9 ± 4% and 58.1 ± 4.1% for Cs and LS, respectively). The 30-fold increase in the sporangia abundance observed during the first week (which is in the same range than the estimated zoospore content) combined with the composition change of R. akinetum population, suggests that almost all zoospores released in their environment may have caused successful infections within a week, illustrating efficient transmission of chytrid infection for this period. Chytrid fecundity did not significantly differ between the October 7 th (30 ± 9 zoosp. spor −1 ) and the October 14 th (26 ± 6 zoosp.spor −1 ). Then we could expect an increase of the R. akinetum population roughly about 30-fold, as reported between the first dates. However, the chytrid population reported on the last date was only 5-fold higher that of Oct. 14 th . Then, it appeared that only one over the six zoospores putatively released on Oct. 14 th would have been responsible for a successful infection on Oct. 21 st .
The infection efficiency is submitted to diverse factors including the susceptibility of the host, its density, the temperature, the light, the nutrient concentration, as well as the grazing pressure 25,[36][37][38] . Among these different factors, the temperature is the unique abiotic factor driving the chytrid parasitism at each life phase. It impacts the sporangia maturation time, the zoospore infective lifetime and potentially the number of zoospores per sporangium 25 . Between the first two dates of our sampling, temperatures in the epilimnion did not vary significantly (−1 °C), whereas the last week was characterized by a sudden temperature drop (−3 °C). Such temperature decrease could impact the infective life time of the zoospores of R. akinetum, reducing transmission efficiency. Bruning 25 showed that the infective life time of the zoospores of Rhizophydium planktonicum, a parasite of the   diatom Asterionella formosa, was reduced by a temperature increase of 4 °C. Also, Ibelings et al. 21 showed that mild winters have strong influences on both host (A. formosa) and parasite (Zygorhizidium planktonicum) which consequently does not reach epidemic levels. Diatoms are known to grow better at lower temperatures than cyanobacteria and green algae 39,40 . We thus hypothesize that life cycle traits (i.e. temperature optima) of parasitic chytrids vary according to their hosts. This suggestion is supported by a previous study on Rhizophydium sphaerocarpum-Spirogyra sp. pairing, where authors reported that optimal conditions for chytrid infection of Spirogyra sp. was 30 °C 41 which was largely higher than what was described for other algal parasites like R. planktonicum 25 . A loss of zoospores, resulting in a chytrid infection reduction, can also be due to zooplankton grazing. In Lake Aydat, zooplankton community associated to cyanobacterial bloom is mainly composed of cladocerans (Daphnia sp., and Ceriodaphnia sp., Thouvenot A., personnal. communication). Kagami et al. 42 , Agha et al. 43 , and Schmeller et al. 44 showed that both cladocerans 43,45 and protozoa 44 can actively graze and/or grow on zoospores. As our last sampling week was marked by the decline in the cyanobacterial bloom, corresponding usually to an increase of zooplankton community 46 , we cannot reject the hypothesis of a massive grazing loss of zoospores between the 14 th and the 21 st of October.
An overview of D. macrosporum-Rhizosiphon spp. pairings changes over 7 years. At the annual scale, chytrid infection is homogenously distributed within the lake, independently of their location in the water column. Based on this result, and with the aim to get an idea on the Rhizosiphon sp. -D. macrosporum changes in Lake Aydat, we analyzed data on these pairings for a 7-year period from 2007 to 2014. Globally, it appears that the infection of D. macrosporum akinetes by R. akinetum highly increased whereas parasitism of vegetative cells by R. crassum dropped (Fig. 4).
Akinetes are the only cells which overcome unfavorable conditions and lead to the colonization of the water column by Nostocalean cyanobacteria when favorable conditions for growth return 47 . Legrand et al. 48 investigated the proportion on live (intact) and dead (lysed) akinetes of D. macrosporum in surface sediment after the sedimentation of the entire cyanobacterial bloom in Lake Aydat (December 2014). Then, we investigated the chytrid parasitism on akinetes from their sediment samples from the Central Station (for more details about the sampling strategy see Legrand et al.) 48 . On the 92.3% of lysed akinetes reported we showed that 45.6% were due to R. akinetum parasitism. By impacting akinetes, the fungal parasitism could be responsible for an important loss of the inoculum size, and thus could delay or be responsible for a decrease in the competitive ability of their next year's cyanobacterium host populations. Actually, Kravchuk et al. 49 underlined that the size of the inoculum (akinete density in sediment) is a critical factor determining the dominance of Dolichospermum flos aquae in the phytoplankton community. Additionnally, Tsujimura et al. 50 suggested that the start of bloom formation was related to the quantity of cyanobacteria colonies in the sediment. In Lake Aydat, increasing infection by R. akinetum results in increasingly severe akinete losses, which likely reduces the pool of overwintering inoculum, and might delay bloom formation. The long term survey of the phytoplankton community in Lake Aydat is consistent with this idea, showing a shift of the blooming period (threshold fixed at 40 µg eq. chlo a.l −1 ) from the end of August in 2007 9 (Fig. 4), underlining that both abiotic and biotic conditions were optimal for bloom development. However, the bloom duration decreased from one month in 2007 to one week only in 2014, narrowing the window of opportunity for infection of vegetative cells and could be one explanation for the R. crassum decrease. In 2015, one year after the maximal chytrid infection reported on akinetes, the cyanobacterial population did not exceed 10 µg eq. chl a l −1 . Although this change can be partly explained by the inter-annual variations of abiotic factors, the parasitism on more than a third of resting cells reported in 2014 has to be considered as an important driver of the "non-bloom" situation observed one year later.
To conclude, it clearly appears that time prevails on space in such highly dynamic relationships. This finding suggests that sampling strategies aiming to capture the dynamics of cyanobacteria-chytrid pairings should focus on temporal, rather than spatial resolution. Nonetheless, we recommend that studies on sediment compartment should be maintained to clarify the importance of benthic phase in the entire chytrid life cycle. Inter-annual variations in chytrid infection highlight the importance of long term monitoring of chytrid-phytoplankton pairings to obtain a global view of the system. Behind the influence of the host density, different interwoven factors and processes underpin the inter-annual variations. Here we show that such variations partly result from the impact of parasitism on resting cells pool, which modulate not only host densities, bloom intensity, but could also impact the success of other chytrid parasite infecting the same host species.

Materials and Methods
Study site and sample collection. Samples were collected in Lake Aydat (45°39′48″N, 002°59′04″E) (Fig. 1). The central point corresponds to the area of maximum depth, whereas the maximal depth of the littoral sampling station was 5 m. For each date, the CS was sampled at five different depths (0.5, 2, 4, 6 and 9 m), and LS at three different depths (0.5, 2 and 4 m). For each sampling depth, 20 liters of lake water were sampled using an 8-L Van Dorn bottle. To eliminate the metazoan zooplankton, immediately after being collected the samples were prefiltered through a 150 µm-pore-size nylon filter, poured into clean transparent recipients, and then transferred to the laboratory for processing. The ≥150-µm fraction was checked to make sure that it did not contain any cyanobacterium. Back in the laboratory, samples were treated: (i) to study the host community (triplicate180-ml aliquots of the raw samples were fixed with Lugol's iodine (Sigma catalog no. 62650)), (ii) to investigate both the prevalence and the intensity of infection as well as the chytrid fecundity using a double staining method 35 . Physical parameters. For each sampled depth and station, water transparency was measured in situ using a Secchi-disk (Z s ) and the depth of the euphotic zone (Z eu ) was calculated according to Reynolds 53 : Z eu = 1.7 × Z s . Temperature and dissolved oxygen profiles were obtained using a multiparametric probe ProOdO TM (Ysi, Germany). A vertical pigment profile was obtained by using a BBE Fluoroprobe ® (Moldaenke, Germany) ( Supplementary Fig. 3).
Host community analysis. Triplicate 180-ml aliquots of raw samples were fixed with Lugol's iodine. For each replicate, 5 to 20 ml (depending on the phytoplankton density) were allowed to settle overnight in a counting chamber. The cells were then counted under an epifluorescence microscope (Zeiss Axiovert 200 M) following the classical Utermöhl method 54 . The entire counting chamber was inspected and D. macrosporum filaments, vegetative cells and mature akinetes were quantitatively analyzed. The distinction between mature and immature akinetes was based on their morphology (the presence of an outer envelope layer is characteristic of mature akinetes), shape (mature akinetes are ovoid whereas immature akinetes are spherical) 29 , and size (16-23 µm width and 21-28 µm length for the ovoid mature akinetes vs 13-17 µm diameter for spherical immature akinetes).
Chytrid parasitism. For chytrid infection parameters, samples were treated following the size-fractionated community method developed by Rasconi et al. 55 . Briefly, 18 L of sampled water was concentrated on 25 µm pore size nylon filter. Large phytoplankton cells (≥25 µm), including the filamentous cyanobacteria D. macrosporum, were collected by washing the filter with 0.2 µm-pore-size-filtered lake water, fixed with formaldehyde (2% final concentration), and an aliquot of 195 µl was stained for the chitin wall. The chitin walls stained with CFW were examined using UV excitation (405 nm). We carried out the observations under an inverted epifluorescence microscope Zeiss Axiovert 200 M at ×400 magnification.
We systematically inspected 200 filaments, comprising 2480 to 4996 individual cells of D. macrosporum to determine the number of infected and non-infected vegetative cells and filaments. In addition, we inspected 300 mature akinetes for the number of infected and non-infected akinetes. Each sample was analyzed in the original triplicates collected. Infection parameters were calculated according to the formula proposed by Bush et al. 56 . These parameters include the prevalence of infection (Pr), i.e., the proportion of individuals in a given population with one or more fixed sporangia or rhizoids, expressed as Pr (%) = [(N i /N) × 100], where N i is the number of infected host cells (or filaments or akinetes), and N is the total number of host cells (or filaments or akinetes). The second parameter is the mean intensity of infection (I) calculated as I = N p /N i , where N p is the number of parasites, and N i the number of the infected individuals within a host population.
Moreover, for each chytrid encountered, its life stage (stage 1 to 6) was noted and assigned to Young, Mature or Empty phase, as described in Gerphagnon et al. 23 . For each mature and empty sporangium, the biovolume of the sporangia was calculated by assimilating sporangia to spheres 57 . From the biovolume of mature and empty sporangia, we calculated the theoretical zoosporic content by using the Conversion Factor (CF) of 0.0172 zoospores per µm 3 of sporangium of Rhizosiphon akinetum established in a previous study 35 .
To get an overview on the inter-annual changes of the chytrid parasitism associated to cyanoacterial blooms in Lake Aydat, we compared the results obtained in 2011 (i.e. this study) with the reports made in 2007 9 and 2010 23 . Also, we used the samples collected for the long term survey of the phytoplankton community in Lake Aydat in 2014 and 2015. Basically, this survey consists to a bi-weekly sampling of the euphotic water column at the central station of the lake with a plankton net (25-mm mesh size) and a vertical pigment profile is obtained by using a BBE Fluoroprobe ® (Moldaenke, Germany). Samples were kept in lugol and used to investigate chytrid parasitism and its D. macrosporum host population as described above.
To investigate the prevalence of infection of akinetes in the surface sediment we followed the akinete extraction method developed by Legrand et al. 48 . Three hundred mature akinetes were inspected and Pr was investigated by staining the chitin walls of R. akinetum with CFW (4% vol/vol). Samples were examined using UV excitation (405 nm). We carried out the observations under an inverted epifluorescence microscope Zeiss Axiovert 200 M at ×400 magnification.

Statistical analyses.
Because of non-normal data, the non-parametric Kruskal-Wallis test was used to test the spatial (vertical and horizontal) variations of each variable followed by a Mann-Whitney pairwise comparison with the Bonferroni correction. All statistical analyses were conducted using PAST.