The Metabolic Capability and Phylogenetic Diversity of Mono Lake During a Bloom of the Eukaryotic Phototroph Picocystis strain ML

Algal blooms in lakes are often associated with anthropogenic eutrophication; however, they can occur naturally. In Spring of 2016 Mono Lake, a hyperalkaline lake in California, was near the height of a rare bloom of the algae Picocystis strain ML and at the apex of a multi-year long drought. These conditions presented a unique sampling opportunity to investigate microbiological dynamics during an intense natural bloom. We conducted a comprehensive molecular analysis along a depth transect near the center of the lake from surface to 25 m depth during June 2016. Across sampled depths, rRNA gene sequencing revealed that Picocystis associated chloroplast were found at 40-50 % relative abundance, greater than values recorded previously. Despite the presence of the photosynthetic oxygenic algal genus Picocystis, oxygen declined below detectible limits below 15 m depth, corresponding with an increase in microorganisms known to be anaerobic. In contrast to previously sampled years, metagenomic and metatranscriptomic data suggested a loss of sulfate reducing microorganisms throughout the lake’s water column. Gene transcripts associated with Photosystem I and II were expressed at both 2 m and 25 m, suggesting that limited oxygen production may occur at extremely low light levels at depth within the lake. Oxygenic photosynthesis under low light conditions, in the absence of potential grazing by the brine shrimp Artemia, may allow for a cryptic redox cycle to occur in an otherwise anoxic setting at depth in the lake with the following effects: enhanced productivity, reduced grazing pressure on Picocystis, and an exacerbation of bloom. IMPORTANCE Mono Lake, California provides habitat to a unique ecological community that is heavily stressed due to recent human water diversions and a period of extended drought. To date, no baseline information exists about Mono Lake to understand how the microbial community responds to drought, bloom, and what genetic functions are lost in the water column. While previously identified anaerobic members of the microbial community disappear from the water column during drought and bloom, sediment samples suggest these microorganisms seek refuge at lake bottom or in the subsurface. Thus, the sediments may represent a type of seed bank which could restore the microbial community as a bloom subsides. Our work also sheds light on the activity of the halotolerant algae Picocystis strain ML during a bloom at Mono Lake, its ability to potentially produce oxygen via photosynthesis even under extreme low-light conditions, and how the remainder of the microbial community responds.

8 density estimated by bacterial and archaeal 16S rRNA gene copy number varied by less than 161 10% from 2 to 25 m. In contrast, a eukaryotic 18S rRNA gene copy number maximum was 162 present at 20 and 25 m (Figure 2b). Major anions including sodium (Na + ) were consistent, and 163 near previously reported values (Table 1). Only minimal differences in anion or cation 164 concentrations were detected within Mono Lake. Nitrate, nitrite, and sulfate were elevated at 10 165 m relative to 2, 20, and 25 m. No phosphate was detectable by ion chromatography (IC) from 2 166 to 25 m within Mono Lake, though surface water taken near shore had an average value of 0.02 167 mM (Table 1). Total dissolved phosphorus (potentially including phosphate and 168 organophosphorus) measured by ICP-AES ranged from 0.59 to 0.63 mM (±0.08 mM) from 169 surface to 25 m depth, respectively (Table 1). Most major anions and cations, and dissolved 170 inorganic carbon, were below detectable limits in the sampled stream water and well water, with 171 the exception of calcium which was elevated relative to Mono Lake water samples (Table 1). 172 Individual replicate results for ICP-AES and IC are shown in supplemental table S1. After quality control a total of 694,948 DNA sequence reads were obtained, clustering into 831 177 operational taxonomic units (OTUs). Additional summary statistics are found in Supplementary 178 Table S2. Chloroplast sequences were abundant across all lake water samples and were removed 179 from further analysis. The bacterial and archaeal community differed in structure above and 180 below the oxycline (Figure 3a). Samples taken from sediment at 10 m depth near the water 181 sampling site also were distinct in bacterial, archaeal, and eukaryotic community structure from 182 those in the sampled water column. Two OTUs most closely related to genera within the order 9 Bacteroidetes decreased in relative abundance steadily with depth: Psychroflexus and ML602M-184 17, whereas unclassified Bacteroidetes remained relatively constant in abundance throughout the 185 water column (Figure 3a). An OTU most closely related to the genus Thioalkalivibrio increased 186 in abundance as depth increased. Unique to the sediment were the Euryarchaeota and the 187 bacterial genus Desulfonatroibacter. An increase in the relative abundance of chloroplast 188 sequence was noted at 20 m, increasing from 39.7 at the surface to 48.4 at 10 m, and then to 61.9 189 percent relative abundance at 20 m (Supplemental Figure S1). Well water taken to compare to Compared to the observed bacterial and archaeal community, the eukaryotic community 204 contained far fewer OTUs. Within the water column at Mono Lake, an almost homogenous 205 distribution of OTUs most closely related to the genus Picocystis was observed at all depths, 206 with a maximum of to 97.9% relative abundance at 10 m depth (Fig 3b)  suggest it is most likely Artemia monica, endemic to Mono Lake, although because of the short 210 sequence read length of 250 bp the identification is ambiguous. Influent stream water samples 211 were distinct from the water and sediment of Mono Lake, with few overlapping OTUs among the 212 samples (Fig 3b). Specifically, multiple OTUs most closely related to the Ochrophyta 213 (Heterokont algae), Ciliophora, and Chytridiomycota were unevenly distributed across the 214 stream and well water sampled. Community membership and distribution within the water 215 column at Mono Lake was significantly influenced (0.017, R 2 = 0.61) by depth and the transition 216 to anoxia visualized by weighted UniFrac PCoA ordination and a corresponding ADONIS test, 217 although less significantly than the bacteria and archaeal community (Figure 4b). 218 219

Metagenomic and Transcriptomic Profiling of Mono Lake and Sediments 220
A summary of assembly statistics for sediment and water samples are available in Table S3. The 221 abundance of sulfate (> 100 mM) and the lack of oxygen beg the question of whether active 222 sulfate reduction is occurring in the dissolved organic carbon (DOC) rich waters of Mono Lake. 223 No sulfite oxidase genes (sox) were identified, however genes for the complete reduction of 224 sulfate to sulfide were identified in the sediment metagenome, and genes for reverse-225 dissimilatory sulfite reductases (dsrA) were identified in water metagenomes. No true reductive 226 dsrA genes were identified in the water metagenomes. Dissimilatory sulfite reductase genes 227 within the sediment metagenome had high (> 80 %) homology to known Deltaproteobacterial 228 sulfate reducing microorganisms. Reductive dsrA/B genes were identified within the water 229 column, identified putatively via BLAST that most closely related to known Thioalkalivibrio 230 dsrA/B genes. Sulfite reductase genes did not appear to be expressed within the 231 metatranscriptome (Table S4). Nitrate and nitrite reductases were found at 20, 25 m and within 232 the sediment, while nitric oxide reductase (nor) was only identified within the sediment (Table  233 S3). Genes associated with nitrogen fixation, including nifH, D, and K were found at 20 m within 234 the water, and within the sediment metagenome. No genes associated with ammonium oxidation 235 by bacteria or archaea (AOB/AOA) were identified. Formate-dependent nitrite reductases were 236 identified as both genes and transcripts (Supplemental Table S4 identified MAGs, 38 were greater than 50 percent complete, and less than 10 percent 243 contaminated with other DNA sequence. A subset of XX of these MAGs contained rRNA gene 244 sequence, and a putative identification was produced from these data (Supplemental figure S2). MAGs were unique to the sediment, including a Euryarchaeon ( Figure 5). No archaea were 248 found in abundance throughout the sampled water column. However, no genes associated with 249 the production of methane were identified. Multiple MAGs were recovered from uncultivated 250 orders within the Actinobacteria, Gammaproteobacteria, and Bacteroidetes (Table S3) including 251 MAGs with 16S rRNA gene sequence previously identified by rRNA gene clone library 252 sequencing at Mono Lake such as ML602J-51 (14). A summary of each genome is available in 253 Supplementary Table S3, and figure 5. No MAGs were identified with the genes required for 254 sulfate reduction, with only reverse-dsr genes found in MAGs. Nitrogen fixation genes (nifH, D, 255 and K) were identified within 3 MAGs, two within the Gammaproteobacteria (Bin 10 and 23), as 256 well as a single unclassified bin (Bin_11_2). One bin (Bin 45) contained photosystem II 257 associated genes, identified within the Epsilonproteobacteria (Table S3, Figure 5). Three MAGs 258 were identified in the EukRep filtered metagenomic sequence. A single MAG was identified 259 with 18S rRNA gene sequence closely related to that of Picocystis strain ML (Supplemental 260 Figure S2). However, this MAG appears to be contaminated with bacterial sequence, although 261 putative searching of the identified sequence returns homology to other known algae. While the 262 MAG should be interpreted with caution, it represents a partial genome sequence of Picocystis sp. 263 from Mono Lake. The annotated genome contained no genes related to sulfur cycling, and other 264 incomplete metabolic cycles (Supplemental Figure S3).  Table S4). More 271 transcripts identified within the co-assembled metatranscriptome were significantly upregulated 272 at 25 m relative to 2 m ( Figure 6). Genes associated with Photosystem I and II pathways were 273 expressed at both sampled depths (Supplemental table S4, Table 2). Expression values for 274 photosystem I and II transcripts including psaA/B, psbA/B, and psbC were significantly 275 13 upregulated at 25 m relative to 2 m depth (Table 2). In addition, several light-independent 276 protochlorophyllide reductase transcripts were significantly upregulated at 25 m, while no 277 transcripts related to chlorophyll production were significantly upregulated at 2 m 278 (Supplementary table S4). concentrations were ten times higher, 33.9 µM (23). The elevated chlorophyll a concentration 286 and Secchi disk values (indicative of lake clarity) above 1 m suggest that Mono Lake was well 287 within a bloom of Picocystis. The relative abundances of microorganisms presented here and the 288 well-mixed major ions of Mono Lake relative to previous work (11, 14), indicated that our 289 sampling represents the first high-throughput molecular study of Mono Lake during a Picocystis 290 bloom and concurrent monomixis. Genes required for sulfate reduction to sulfide were detected 291 only in the sequenced lake sediment, while both metagenomic and 16S rRNA gene sequencing 292 indicated a near complete loss of the anaerobic sulfate reducing potential within the water 293 column of Mono Lake. Instead, a mixed algal and facultatively anaerobic microbial community 294 was present below the detectable oxycline, more similar to the near-surface microbial 295 community than previously reported (11). It is yet unknown how the microbial community of 296 Mono Lake will rebound after such a significant algal bloom and a decline in the population of 297 Artemia within the lake. 298 299 Our survey allowed for a comprehensive evaluation of the genomic potential, and expressed 300 genes associated with metabolic processes throughout the water column. Dissimilatory nitrate 301 reduction to ammonium (DNRA) appeared active, with formate-dependent cytochrome c nitrite 302 reductases detected within the transcriptome (Table S4) and formate-dependent nitrite reductase 303 subunits within the assembled metagenomes (Table S3). No genes associated with ammonium 304 oxidation (AOB) were identified in contrast to previous years (24) in either the transcriptome or 305 metagenome, suggesting that the ammonia produced within the lake was assimilated, likely by 306 the dense population of growing Picocystis. In addition to nitrate reduction another key 307 anaerobic respiratory process, sulfate reduction, was largely absent from the water column. 308 309 Previous work during meromixis/non-bloom intervals has shown that sulfate reduction is a key 310 respiratory process in Mono Lake, supporting the growth of multiple species of sulfide oxidizing 311 aerobic microorganisms above the oxycline (11). We found that microorganisms capable of 312 sulfate reduction were only identified in sediment metagenomic samples during the bloom. 313 Dissimilatory-type reverse sulfite reductases associated with sulfur oxidizing 314 gammaproteobacterial (25) taxa were identified at 20 and 25 m, but no true reductive sulfite 315 reductases were found in sequenced water samples. Taxa known to reduce sulfate were also only 316 identified by 16S rRNA gene sequencing in stark contrast to previously sampled years (11, 26). 317 Instead, the most abundant microorganisms with identifiable dsrA/B gene clusters were reverse-318 dsr type reductases identified previously in the Gammaproteobacterium genus Thioalkalivibrio 319 (25). While lake sulfate reduction rates are typically very low (27) our data suggest a complete 320 loss of sulfate reducing activity in the water column during a bloom. It is likely that during a 321 bloom, sulfate reduction is repressed as more oxidizing conditions are present throughout the 322 water column due to an increased abundance of oxygenic photosynthetic algae. Members of the 323 Bacteroidetes were in high abundance throughout the water column, including OTUs most 324 closely related to ML310M-34, which remained abundant through the water column and 325 Psychroflexus, which decreased in abundance from 2 to 25 m as oxygen levels declined. The 326 eukaryotic microbial community was more evenly distributed throughout the water, with 327 Picocystis detected in near equivalent relative abundance throughout the water column ( Figure  328  previous studies may also slow the metabolism of Artemia, resulting in reduced fecundity and 340 increased mortality (22, 28). A decline in Artemia population could also impact bird mortality, 341 though this was outside the scope of this study, and should be investigated at a later date. 342

343
A key finding of this study is the confirmation that Picocystis strain ML appears capable of communication), will allow for its genome to be removed from subsequent sequencing efforts 356 which will simplify assembly, and enhance the resolution of bacterial and archaeal binning 357 efforts in the future, yielding a better understanding of the microbial community responsible for 358 the diverse metabolic potential in both the sediments and water of Mono Lake. Despite the lack 359 of a reference genome, our transcriptomic sequencing was able to recover Picocystis chloroplast 360 associated transcripts. At 25 m depth, a significant upregulation of Photosystem II was observed 361 (Table 1, Supplemental Table S4). This, combined with the 40 percent increase in the number of 362 18S rRNA gene copies at 25 m relative to 2 m suggest that there is, at a minimum, a near 363 equivalent amount of transcription of photosynthesis-associated genes throughout the water 364 column. Recently, photosynthesis in a microbial mat was shown to be capable under extremely 365 low light concentrations, although in a bacterial system (29). Still, the presented data suggest that 366 under extreme low light conditions, photosynthesis may still occur. This is the first 367 transcriptomic evidence from Mono Lake to support previous laboratory observations of 368 Picocystis growing under low light conditions (6). 369

370
Our study represents the first study of Mono Lake during the height of an algal bloom and 371 suggests significant shifts in both the bacterial and archaeal microbial community and its 372 metabolic potential from non-bloom years (11, 16). Picocystis was present throughout the water 373 column, and apparently carrying out oxygenic photosynthesis even at extremely low levels of 374 light at depth within the lake. While Picocystis bloomed throughout Mono Lake, there was also a 375 loss of sulfate reducing microorganisms. The lack of sulfate reduction at and below 20 m within 376 Mono Lake is in contrast to previous work and is possibly linked to the intense drought 377 experienced by Mono Lake from 2012 to 2016. During such a drought anaerobic microorganisms 378 may seek refuge within the underlying sediment. By sequencing nearby sediment, we have 379 shown that even if sulfate reduction is temporarily lost in the planktonic community of Mono 380 Lake, the sediment may act as a "seed bank" or refugia for organisms capable of this, and likely 381 other necessary metabolisms dependent upon overlying water / lake conditions (30). 382 Alternatively, the sulfate reducing microorganisms may find a better reduced substrate or fewer 383 inhibitors in the sedimentary environment. Furthermore, the recovery of microbial populations 384 within Mono Lake must come from its' sediment or underlying groundwater, not from the 385 streams that feed it as no overlapping taxa exist. Establishing if, and how, the chemistry and 386 microbiota of Mono Lake recover after monomixis, drought, and algal bloom should be the focus 387 of future work. Such research can be compared against our metagenomic and transcriptomic 388 during bloom as well as previous metatranscriptomic sequencing (11) to better understand how, 389 or if, the microbial community of Mono Lake returns to its previous state after extended periods 390 of both monomixis and algal bloom. After measurements were obtained water was pumped from depth to the surface at station 6 399 (37.95739,-119.0316, Figure 1), sampled at 2 m, 10 m, 20 m, and 25 m the following day (due to 400 lake conditions) using a submersible well-pump. Water was allowed to flow from the measured 401 depth for 1 to two minutes to clear any residual water from the lines prior to sampling. Artemia 402 were removed from water samples using clean cheese cloth prior to filling 1 L sterile high-403 density polyethylene containers. Samples were stored in a dark cooler until filtration occurred.

Quantitative PCR 454
Total bacterial/archaeal and eukaryotic small subunit (SSU) rRNA gene count within the water 455 column was obtained using two TaqMan based probe assays as previously described (32, 33). 456 Briefly, both assays were carried out using 25 µL reactions containing 1x final concentration of 457 of each primer, and 225 nM of either the bacterial/archaeal, or eukaryotic probe. 459

SSU rRNA Gene Analysis 460
Sequence reads were demultiplexed in QIIME version 1.9.1 (34), and filtered at a minimum Q 461 score of 20 prior to clustering. Sequence reads were first denoised and then clustered into 462 operational taxonomic units (OTUs) using UPARSE (35). After clustering, OTUs were assigned 463 taxonomy using mothur (36) against the SILVA database (r128, (37)). Each OTU was then 464 aligned against the SILVA r128 database using pyNAST (38), filtered to remove uninformative 465 bases, and then a tree was created using the maximum likelihood method and the Jukes Cantor 466 evolutionary model within FastTree 2 (39). A BIOM formatted file (40) was then produce for use 467 in analyses downstream. To limit OTUs originating from contaminating microorganisms found 468 in extraction and PCR reagents (41) all extraction blanks and PCR controls were processed 469 separately and a core microbiome was computed. Any OTU found in 95% of controls was 470 filtered from the overall dataset. Differences in community composition were estimated using the 471 weighted UniFrac index (42). The effect of depth was tested using an adonis using the R package 472 Vegan (43) within QIIME. Taxa heatmaps and ordination plots were generated using phyloseq 473 Prior to assembly, metagenomic libraries were quality filtered and adapters removed using PEAT 496 (46). A co-assembly was produced using MEGAHIT (47) with a minimum contig length of 5000 497 basepair. After assembly, quality filtered reads from individual samples were mapped to the co-498 assembly using Bowtie2 (44). Assembled contigs greater than 5 kb in length were first filtered to 499 remove eukaryotic sequence using EukRep (48) and then binned into MAGs using CONCOCT 500 (49) and refined using Anvi'o (50), in an attempt to manually reduce potential contamination or 501 redundancy within each bin. Finally, bin quality was assessed using CheckM (51). 502

23
CheckM was also used to identify possible SSU rRNA gene fragments within each bin. 504 Putatively identified SSU rRNA gene fragments were aligned against the SILVA 132 database 505 (37) using SINA (52). After alignment, sequences were added to the SILVA tree by SINA, and 506 near relatives were included to give a putative identification of MAGs containing SSU sequence. 507 The identities of each MAG with SSU sequence are available in Supplementary Table 2. 508 509

Metatranscriptomic Analysis 510
Metatranscriptome libraries were first filtered for quality and adapter removal using PEAT (46). 511 After quality control, sequence files were concatenated into a single set of paired-end reads in 512 FASTQ format, and then assembled de novo using Trinity (53). Post-assembly the Trinotate 513 package (https://trinotate.github.io/) was used to annotate assembled transcripts. After assembly, 514 reads were mapped against transcripts using Bowtie2 (54), and differential significance was 515 assessed using DEseq2 (55). Assembly, annotation, mapping, and statistical analyses were 516 carried out using XSEDE compute resources (56).