Biodiversity and disease risk in an algal biofuel system: An experimental test in outdoor ponds using a before-after-control-impact (BACI) design

each sampling event to plot a time-series of infection proportions. Algal cells were categorized as infected if they had any of the following signs of infection: fungal zoospore attached to the cell wall, cellular contents displaced by the growing parasite, and/or any remaining unconsum- able cellular contents of the host algal cell called a residual body (S1 Fig in S1 Appendix) [22, 39, 40]. Community composition for the Impact treatment was quantified as the proportion of total cell density represented by each algal species. When calculating the four response vari- ables, data collected prior to the initial 70% harvest on day 8 were not used so that comparisons between each subsequent harvest represent a response in the algal cultures to the same magni- tude of a routine disturbance (i.e., 90% harvest). In addition, data collected between infection (day 23) and the first post infection harvest (day 30) were not used in the analysis. These data were excluded because the fungal pathogen was initially introduced 2 days after a harvest which allowed the algal cultures to experience 2 days of growth without the potential stress of infection. Including this particular set of dates (days 23–30) in the After-infection period would include a portion of the culture’s response to harvest without stress from a fungal patho- gen, which is more characteristic of the Before-infection period. Thus, to ensure that the data included in the After-infection period is accurately indicative of the response of infected algal

cultures to routine harvesting, measurements from the first sampling event after the day 30 harvest were only included. BACI analysis. A BACI analyses was performed to determine if the four aforementioned response variables (Section Raceway sampling and processing) showed significant differences between the Control (monoculture) and Impact (polyculture) treatments, as well as the response of those two treatments to introduction of the parasitic fungus. The BACI approach described in Smith et al. [41]was used, which involved running linear mixed models for each of the response variables with treatment (Control—monoculture vs. Impact—polyculture), infection period (Before introduction of fungus vs After infection), and Treatment × Period interaction as fixed effects, and individual raceways as a random effect to account for random variation among experiment units (“lmer function” within R package lmerTest). To examine whether fungal infection has a differential effect between Treatments (control-monoculture vs. impact-polyculture), results of the linear models were subjected to an analysis of variance (ANOVA). A significant Treatment × Period interaction for each response variable was taken as evidence that each Treatment responded differently to infection from a fungal pathogen. These tests, in conjunction with the time-series estimates of infection proportions and com- munity composition, provided insight into the biological cause for differential response in the various algal production metrics between treatments.

In addition to using BACI to compare the biological response variables, similar analyses were performed that compared the physical and chemical conditions of the raceway ponds to confirm homogeneity of conditions between treatments (See Supplemental Information). No data transformations were needed as assumptions of homogeneity of variance and approxi- mate normality were met. Results Physical and chemical conditions among replicates were statistically similar (Section Physical and chemical conditions of raceways) thus these results can reasonably be attributed to differ- ences in biodiversity among treatments (control-monoculture vs. impact-polyculture). As expected, yield, productivity, and temporal stability across both treatments were reduced fol- lowing fungal infection (Fig 2). Although we did not have replicate ponds that were uninfected with the fungus to definitively show that fungal infection was the cause of the decline in pro- duction metrics, this decline can reasonably be attributed to the fungal infection. This is PLOS ONE Biodiversity, biofuel and disease: Ecological engineering in algal feedstocks PLOS ONE | https://doi.org/10.1371/journal.pone.0267674 April 28, 2022 7 / 15