Microbial succession and dynamics in meromictic Mono Lake, California

Abstract Mono Lake is a closed‐basin, hypersaline, alkaline lake located in Eastern Sierra Nevada, California, that is dominated by microbial life. This unique ecosystem offers a natural laboratory for probing microbial community responses to environmental change. In 2017, a heavy snowpack and subsequent runoff led Mono Lake to transition from annually mixed (monomictic) to indefinitely stratified (meromictic). We followed microbial succession during this limnological shift, establishing a two‐year (2017–2018) water‐column time series of geochemical and microbiological data. Following meromictic conditions, anoxia persisted below the chemocline and reduced compounds such as sulfide and ammonium increased in concentration from near 0 to ~400 and ~150 µM, respectively, throughout 2018. We observed significant microbial succession, with trends varying by water depth. In the epilimnion (above the chemocline), aerobic heterotrophs were displaced by phototrophic genera when a large bloom of cyanobacteria appeared in fall 2018. Bacteria in the hypolimnion (below the chemocline) had a delayed, but systematic, response reflecting colonization by sediment “seed bank” communities. Phototrophic sulfide‐oxidizing bacteria appeared first in summer 2017, followed by microbes associated with anaerobic fermentation in spring 2018, and eventually sulfate‐reducing taxa by fall 2018. This slow shift indicated that multi‐year meromixis was required to establish a sulfate‐reducing community in Mono Lake, although sulfide oxidizers thrive throughout mixing regimes. The abundant green alga Picocystis remained the dominant primary producer during the meromixis event, abundant throughout the water column including in the hypolimnion despite the absence of light and prevalence of sulfide. Our study adds to the growing literature describing microbial resistance and resilience during lake mixing events related to climatic events and environmental change.

to transition from annually mixed (monomictic) to indefinitely stratified (meromictic). We followed microbial succession during this limnological shift, establishing a two-year (2017-2018) water-column time series of geochemical and microbiological data. Following meromictic conditions, anoxia persisted below the chemocline and reduced compounds such as sulfide and ammonium increased in concentration from near 0 to ~400 and ~150 µM, respectively, throughout 2018. We observed significant microbial succession, with trends varying by water depth. In the epilimnion (above the chemocline), aerobic heterotrophs were displaced by phototrophic genera when a large bloom of cyanobacteria appeared in fall 2018. Bacteria in the hypolimnion (below the chemocline) had a delayed, but systematic, response reflecting colonization by sediment "seed bank" communities. Phototrophic sulfide-oxidizing bacteria appeared first in summer 2017, followed by microbes associated with anaerobic fermentation in spring 2018, and eventually sulfate-reducing taxa by fall 2018. This slow shift indicated that multi-year meromixis was required to establish a sulfate-reducing community in Mono Lake, although sulfide oxidizers thrive throughout mixing regimes. The abundant green alga Picocystis remained the dominant primary producer during the meromixis event, abundant throughout the water column including in the hypolimnion despite the absence of light and prevalence of sulfide. Our study adds to [Corrections added on 25 March 2021, after first online publication: At the request of the corresponding author two important changes have been made to the first page of the article.
In affiliation 2 "Earth Systems Process Division" and "United States Geological Survey" have been transposed and the USGS nonenforcement disclaimer has been added.]

| INTRODUC TI ON
The interplay between microbial communities, biogeochemical cycling, and changing environmental conditions is of long-standing interest. Related questions such as the pace and dynamics of microbial succession have been less explored, presumably due to the inherent difficulty of observing natural ecosystems in the midst of dramatic changes. Existing studies of aquatic microbial succession in lakes have focused mainly on loss of stratification, presumably for practical reasons of timing and access. These studies (Hollibaugh et al., 2001;Jones et al., 2008;Shade et al., 2011Shade et al., , 2012 suggested that aquatic communities were generally quite stable in the face of disturbance due to mixing, with high resistance and resilience. We are not aware of any previously published time series examining succession across the onset of stratification and anoxia. Given that climate models and field observations alike suggest future increases in lake stratification with global warming (Arvola et al., 2010;Peeters et al., 2002), this represents a significant knowledge gap. Here, we report a two-year time series (2017-2018) of microbial diversity and geochemistry from Mono Lake, California, that captures one such transition-the lake's shift from annually mixed (monomictic) to indefinitely stratified (meromictic), following an exceptionally large snowmelt in 2017. This dataset provided a unique opportunity to study both shifts in microbial composition and lake geochemistry that accompany the onset of stratification, as well as the timing of these changes.
Mono Lake is a hypersaline (~85 ppt, ~3x seawater salinity), alkaline (pH 9.8), closed-basin lake located in the eastern Sierra Nevada Mountains of California, USA (38°N, 119°W; Figure 1). Water enters the lake primarily from snowmelt via streams (Wilson Creek, Mill Creek, Rush Creek, Lee Vining Creek) and exits by evaporation (~1 m/yr). This hydrologic condition, combined with weathering of Sierran granite in the lake's watershed, produces the high salinity and pH characteristic of Mono Lake (Garrels & MacKenzie, 1967). These properties also tightly tie lake levels and water chemistry to climate, with modern and Pleistocene high stands correlated with wet Sierra Nevada conditions (Benson et al., 1998). Further, primary productivity in Mono Lake is extremely high (269-1060 g C m −2 yr −1 ) due largely to the activity of the green alga Picocystis sp. strain ML (Jellison & Melack, 1993). The brine shrimp Artemia monica is the predominant metazoan predator in Mono Lake, feeding on Picocystis and in turn supplying a food source to hundreds of thousands of migrating birds that flock to the lake annually on the Pacific Flyway (Wiens et al., 1993). Thus, despite its modest size and extreme chemistry, Mono Lake serves an essential ecological role beyond the growing literature describing microbial resistance and resilience during lake mixing events related to climatic events and environmental change.

K E Y W O R D S
environmental microbiology, geochemistry, limnology, microbial ecology, microbial succession

Summary statement
We investigated microbial succession in hypersaline, alkaline Mono Lake, California, following a climatic perturbation. By combining geochemical and microbiological data, we were able to link depth-dependent microbial community shifts to limnological changes. F I G U R E 1 Bathymetric map of Mono Lake after Bruce, Jellison, Imberger, & Melack (2008). Location is indicated on the California map insert with a blue star. All samples were collected from Station 6 in the south basin, marked with a white star. White areas on the map correspond to Paoha island (larger island) and Negit Island to the north F I G U R E 2 Vertical profiles of (a) salinity, (b) temperature, (c) dissolved oxygen, (d) nitrate, (e) ammonium, (f) sulfide, (g) particulate organic carbon (POC) (h), POC δ 13 C VPDB , and (i) dissolved organic carbon (DOC) concentrations at Station 6, Mono Lake, CA. Profiles were measured in spring 2017 (red circles), summer 2017 (orange inverted triangles), fall 2017 (yellow squares), spring 2018 (green triangles), summer 2018 (blue diamonds), and fall 2018 (purple pentagons) (Award #12520045). Any use of trade, firm, or product names is for descriptive purposes only and does not imply endorsement by the U.S. Government.
the Sierra Nevada. Seasonal surveys of primary production, Artemia, and birds such as grebes, gulls, and phalaropes have been conducted since at least 1980 (Cooper et al., 1984;Melack et al., 2017), though they have not provided much insight into Mono Lake's microbial community.
Mono Lake is typically monomictic, with thermally driven summer stratification disrupted by winter, wind-driven overturn. However, in years with very high spring-summer freshwater inputs, often correlated with El Niño-Southern Oscillation (ENSO) events, a buoyant layer with >10 ppt lower salinity can form at the surface of the lake that withstands winter mixing. This leads to protracted stratification that can continue for five or more years, depending on continued water balance and storm severity. During this time, the hypolimnion becomes euxinic, with sulfide concentrations reaching hundreds of micromolar (Miller et al., 1993 (Carini et al., 2005), sulfate reducers (Scholten et al., 2005), nitrifiers (Carini & Joye, 2008), and methylotrophs (Nercessian et al., 2005).
Recently, two studies of Mono Lake have used next-generation sequencing technology, including 16S rRNA gene amplicon sequencing, environmental metagenomics, and transcriptomics (Edwardson & Hollibaugh, 2017;Stamps et al., 2018), but these investigations focused on single time points. To supplement this literature, we monitored Mono Lake at the onset of a new meromictic period that began in 2017. Our two-year (thus far) time series reflects six seasonal sampling dates with physical, chemical, and biological data to characterize Mono Lake during this transition.

| Water sampling
All water sampling was conducted at "Station 6" (37.95739 N, −119.0316 W), a representative site in the deepest part of the south basin, marked by a buoy (Figure 1). Samples were collected on May Salinity profiles were initially calculated from conductivity after Jellison et al. (1991). While this produced accurate bottom-water salinity estimates (as compared to handheld refractometer measurements of 86 ± 2 ppt; Atago S-28E), it also yielded sharp fluctuations in salinity across the thermocline, which we do not believe are physically plausible and which made interpreting the structure of the chemocline difficult. Instead, we used the default SeaBird seawater calibration of our CTD, then scaled each profile linearly to a bottom-water salinity of 86. This produced a smoother chemocline structure ( Figure 2a), but potentially at the expense of less accurate surface salinity values. Discrete water samples were collected using a 5-L Niskin bottle or a 2" diameter submersible well pump at five to eight depths. When the pump was used, water flowed through the pump and tubing at each depth for at least two minutes before sampling. Artemia monica (brine shrimp) were removed from sampled water using a 10µm Nitex-screen filter. Water aliquots were taken for dissolved inorganic carbon (DIC), dissolved organic carbon (DOC), dissolved organic nitrogen (DON), sulfate and sulfide, nutrients and major ions, particulate organic carbon (POC) and fatty acids (FAs), 16S rRNA gene amplicon sequencing, and incubations. Not all parameters were sampled on all dates, as described in Table S1.

| Geochemical measurements
2.2.1 | Sulfide and sulfate Water (7 ml) was immediately preserved for sulfide analysis by filtering through a 0.22µm polyvinylidene difluoride (PVDF) syringe filter into a 15-ml falcon tube containing 3 ml zinc acetate (1 M), precipitating zinc sulfide and zinc carbonate. Samples were kept at room temperature until analysis. Concentrations were measured using a modified spectrophotometric assay (Cline, 1969)

| Dissolved organic carbon (DOC)
On the boat, 500 ml of water was subsampled for DOC analysis into acid-cleaned 1-L amber bottles (Nalgene, HDPE) and placed on ice.
Directly after returning to the lake shore, this water was filtered through pre-washed 0.22 µm PES syringe filters into acid-washed 60-ml amber bottles (Nalgene) and frozen at −20°C. Immediately before analysis, samples were thawed and acidified to pH 2 with 10 M hydrochloric acid (HCl  (Nydahl, 1978;Wang et al., 2015), followed by quantification of the resulting nitrate with the chemiluminescence method (Braman & Hendrix, 1989) at Princeton. In a pre-combusted 4-ml glass vial, 0.1 ml of Mono Lake water was combined with 1 ml persulfate oxidizing reagent (1 g recrystallized low-N K 2 S 2 O 8 and ACS grade 1.5 g NaOH in 100 ml ultrapure water) and autoclaved for

| Fatty acids (FAs)
Cultured Picocystis biomass and Mono Lake POM were analyzed for fatty acid compositions via gas chromatography-mass spectrometry (GC-MS) at Caltech. Picocystis biomass (~10 mg) was freeze-dried, then ground in a solvent-washed mortar and pestle. Freeze-dried POM filters were cut in half, with one half archived while the other was extracted. Samples were extracted and derivatized as methyl esters following standard protocols described in the SI.

| Picocystis cultures
We obtained Picocystis sp. strain ML pure culture (Roesler et al., 2002) from R. Oremland (USGS) and grew cells in batch culture under white light on a rotary shaker at 180 rpm and 30°C. L1 minimal medium was prepared and used as described in Guillard and Hargraves (1993), substituting artificial seawater (35 g/L Instant Ocean Sea Salt) in place of filtered natural seawater. The growth medium was supplemented with salt to reach 70 g/L NaCl. Cell growth was monitored by measuring absorbance at 600 nm, and purity was checked by microscopy. Cultures were harvested in the late exponential phase (OD 0.4-0.6) and stored at −25°C until lyophilization, extraction, and analysis.

| Sulfate reduction incubations
Incubations for sulfate-reducing potential were started in fall 2018 by addition of 30 ml Mono Lake water from 12, 17, 20, and 35 m water depth to N 2 -sparged 60-ml serum bottles with blue butyl stoppers at the time of sampling. Samples were kept on ice overnight until return to Caltech, when they were incubated at 4°C in the dark. Incubations were conducted in duplicate with four conditions: lactate addition (2 mM final concentration), pyruvate addition (2 mM final concentration), no addition, and killed control. Killed controls were autoclaved 48 hr after sample addition. Incubations were monitored by measuring dissolved sulfide concentration using the Cline assay (described in SI) at three days, two weeks, one month, and six months.

| 16S rRNA Illumina TAG sequencing
Water samples for high-throughput Illumina 16S rRNA gene amplicon sequencing were collected in 50-ml falcon tubes and stored on ice or at 4°C for ~24 hr until benchtop centrifugation (5250 g, 20 min, 4°C). Supernatant was discarded until a pellet and ~1 ml liquid remained. Pellets were resuspended and transferred to 2-ml tubes and centrifuged (16 000 g, 10 min). Supernatant was decanted, and pellets were kept at −20°C until DNA extraction following Zhou et al. (1996; details available in the SI). Following the initial PCR, amplicons were barcoded using Illumina Nextera P5/P7 primers and pooled prior to sequencing and subsequent sample demultiplexing at Laragen Inc. Sequence data were analyzed using QIIME 1.8.0 (Caporaso et al., 2010), with a 99% OTU cutoff and taxonomic assignment using the SILVA138 NR99 database (Quast et al., 2013).
OTU counts were binned by taxonomic assignment at the genus rank. Using R, counts for each genus were normalized to relative abundance per dataset. After normalization, chloroplasts (representing mainly the eukaryote Picocystis) and fully unassigned sequences (<90% identity to any SILVA sequence) were removed for clarity, and the data were binned in 3 groups: genera represented by >1% of the sequences of any one sample, genera represented by 0.1%-1% of the sequences of any one sample, and genera not present in more than 0.1% in any sample. The resulting abundance tables were used to generate heat maps in MATLAB (code in SI). Statistical analyses (NMDS, ANOSIM, ADONIS, Pearson correlation, SIMPER) were carried out in R using the vegan package (Oksanen et al., 2019). Alpha diversity was analyzed by rarefying each sample to 1,370 sequence depth using QIIME 1.8.0 and calculating richness (number of distinct OTUs), evenness (Shannon equitability), and diversity (Shannon-Wiener Index) following Shannon and Weaver (1949). Dissolved sulfate (~105 mM), silica (540 ± 60 µM), and phosphate (697 ± 44 µM) remained high throughout the time series (Table S2) Table S2).

| Physical and chemical properties
The total hydrolyzable fatty acid (FA) content of cultured Picocystis sp. strain ML is shown in Figure 3 and described in detail in the SI. Extractable fatty acids from Mono Lake POM were consistent with Picocystis being the main source ( Figure 3, Table S3).
Saturated FA between C 14 and C 20 accounted for a slightly larger normalized abundance than observed in Picocystis culture, potentially reflecting significant contributions from bacterial sources and/ or degradation of short-lived double bonds (Rontani et al., 2012).
Monounsaturated FAs between C 14 and C 16 not synthesized by Picocystis varied between 2% and 4% in the hypolimnion and reached up to 7% in the epilimnion. Iso and anteiso-C 15 FAs were uniquely present in POM and reached up to 5%-6% in hypolimnion samples. Long-chain fatty acids (>C 20 ) were also identified in variable quantities (~1-10 ng/ml); these molecules were likely sourced from terrestrial or littoral vegetation and were removed before normalization (Table S3).

F I G U R E 3
Relative abundance of fatty acids (FAs) detected in Picocystis cultures and Mono Lake water-column POC, analyzed as fatty acid methyl esters. For the purpose of comparison, FAs are grouped into five categories shared between Picocystis and POC extracts: C 14 -C 20 saturated (dark blue), C 16 and C 18 monounsaturated (turquoise), C 16 , C 18 , and C 20 polyunsaturated (green), C 14 , C 16 , and C 18 β-hydroxy (light green), and C 16 and C 17 branched (yellow). Categories of FAs unique to POC samples included C 14 , C 16 , and C 18 monounsaturated (orange), C 17 monounsaturated (red), and C 15 branched (maroon)

| Microbial community composition
The similarity in 16S rRNA gene amplicon diversity for our time- hypersaline lakes, with Archaea frequently dominating under extremely high salinity (Jiang et al., 2006;Oren, 2008). 98.7% of chloroplast sequences were assigned to the dominant phototrophic alga, Picocystis. We therefore used chloroplast relative abundance as a proxy for Picocystis, which varied between 17% and 40% relative abundance in the epilimnion (Figure 6a; Figure S1a) and 32%-38% in the hypolimnion (Figure 6b; Figure S1)  Picocystis (dark blue) abundance was inferred from chloroplast sequences. Some phyla were not included, as their relative abundance was too low to be discernable in this visualization method. These groups included the following: Euryarchaeota, Actinobacteria, Deinococcus-Thermus, Planctomycetes, Spirochaetes, Verrucomicrobia, and Tenericutes (a) (b)

F I G U R E 7
Measurements of water-column sulfate reduction potential in incubation experiments. Samples were collected in spring 2018 from 12, 17, 20, and 35 m water depths. Each depth was monitored for sulfide production over the course of six months. Values are the average of replicates with vertical error bars displaying 1σ standard deviations. The 12 m sample was from the epilimnion, while other samples were collected from the hypolimnion and at time zero began with background sulfide concentrations in the lake (0.5 mM). 2 mM lactate addition (purple pentagons), 2 mM pyruvate (blue diamonds), no addition (yellow squares), and the autoclaved killed control (red circles) nevertheless emphasizes that sulfate reduction rates in the water column are likely very slow.

| D ISCUSS I ON
Our dataset for Mono Lake offers insight into aquatic microbial succession during the onset of meromixis. Within just one month of stratification, community shifts were apparent, and 16 months later, these successional dynamics were still occurring (Figures 4-6). The epilimnion and hypolimnion microbial communities followed distinct paths (Figures 4 and 6), and-with the possible exceptions of Picocystis and Cyanobacteria-there appeared to be relatively little biological exchange of these communities across the chemocline.
We therefore considered the following four aspects of our data: (a) the dynamic succession in the epilimnion, (b) the more predictable succession patterns in the hypolimnion, (c) the temporal and limnological drivers of sulfur cycling, and (d) the enigmatic ecophysiology of Mono Lake's prolific primary producer, Picocystis, and its resistance to stratification.

| The dynamic surface community
Mono Lake's epilimnion bacterial community experienced drastic changes throughout our two-year time series, demonstrating low resistance and resilience (Figures 4a and 6a). Other surveys of stratified lakes have also found that epilimnion bacteria were more variable over time than hypolimnion populations , reasoning that surface communities face more variability from climatic interactions. Indeed, seasonal Mono Lake surface water temperatures fluctuated between ~0°C and 20°C and winter storms increased terrestrial organic matter delivery, as inferred from increased longchain fatty acids (>C 20 ) (Table S3). Taken together, it was perhaps unsurprising that Mono Lake's surface microbial community shifted as lake stratification stabilized. What was unexpected is the direction of this shift: a reduction in relative numbers of heterotrophs, with concomitant increase in primary producers.
At the start of our time series, during monomixis, the epilimnion microbial community was characterized by the phototrophic alga Picocystis (~20% relative abundance) and aerobic or aerotolerant heterotrophic bacteria (~40% relative abundance) within Alphaproteobacteria, Gammaproteobacteria, Cytophagia, Actinobacteria, Bacteroidia, and Flavobacteriia ( Figures 5 and 6a, Figure S2; see SI for taxa specific details). This early epilimnion diversity comprised a suite of saline-and alkaline-adapted heterotrophic bacteria likely capable of degrading and remineralizing complex organic matter such as that supplied by the dominant Picocystis.
The drastic shoaling of the oxycline in spring-summer 2017 ( Figure 2c) reduced the niche for aerobic metabolisms, but the composition of the epilimnion microbial community was only minimally altered (Figures 4-6). By fall 2017, however, putative heterotrophs had fallen drastically to ~10% relative abundance and would remain at that level throughout the remainder of our time series.  (Figure 4d). C 14 -C 18 monounsaturated fatty acids also increased in the epilimnion during this phototrophic bloom (orange bars, Figure 3) and were tentatively attributed to the Cyanobacteria. Group 1 cyanobacterial species, like those in the genus Synechococcus, are up to 50% C16:1 fatty acid by weight (Murata et al., 1992). As this cyanobacterial bloom is repre-  Table S2). Further, both Picocystis and Cyanobacteria are capable of utilizing ammonium, which appears to be continuously diffusing into surface waters (Figure 2e). Finally, cyanobacterial success post-bloom is often attributed to their high surface area to volume ratios that outcompete larger cells for nutrients. This advantage becomes less relevant when compared to Mono Lake's small (<3 µm) Picocystis. Although nitrogen limitation has been favored in other systems for explaining cyanobacterial blooms, it seems unlikely to be the main driver in Mono Lake.
We hypothesize that top-down controls from predation by Artemia, rather than bottom-up factors such as nutrients, were responsible for increased phototrophic niche space. Surveys of Mono Lake's algae have documented that net primary production, as measured by chlorophyll a abundance, drops by a factor of two or more during meromixis (Melack et al., 2017). The response of Artemia is a single population peak in the spring, rather than two smaller generations in the spring and summer ( Figure S3b). This pattern is consistent with our PAR data ( Figure S1c It is rare for species competing for the same resources, such as light and nitrogen, to coexist during steady state (Hardin, 1960), although dynamic transitions may create additional niche spaces during succession for organisms to co-occur (Flöder & Sommer, 1999). In our fall 2018 data, Mono Lake's meromictic epilimnion was characterized by two abundant phototrophs. The codominance of algae and Cyanobacteria has been previously reported in another hypersaline system: At Lake Dziani Dzaha, authors found that Picocystis salinarum and Arthrospira fusiformis spanned the water column of a monomictic lake (Bernard et al., 2019). They hypothesized that this coexistence related to distinct light adaptations. Indeed, Picocystis is specifically adapted for low photon flux and blue-green wavelengths (~400-500 nm), allowing survival deeper in the epilimnion (Roesler et al., 2002). The most abundant Cyanobacteria in our study belonged to an uncharacterized genus within the non-diazotrophic

| Ecological shifts in the hypolimnion
The water column in spring 2017 was largely oxic to hypoxic (Figure 2c), and the hypolimnion bacterial community resembled that of the epilimnion (Figure 4, Figure S2). With the development of a sharp oxycline, aerobic and aerotolerant populations shoaled.

Exceptions included the phototrophic Cyanobacteria and alga
Picocystis, which maintained substantial abundance in the dark, anoxic hypolimnion throughout the time series (Figure 6b, Figure S1a), an enigma discussed further in Section 4.4. The overall pattern as meromixis developed was a shift from aerobic heterotrophs to anaerobic fermenters and lithotrophic sulfur oxidizers to, eventually, sulfate reducers. In particular, there was a slow rise in obligately anaerobic bacteria, especially the Firmicutes and Bacteroidia. These organisms were likely utilizing ample deep-water DOC and DON ( Figure 2i, Table S2) and/or the degradation products of sinking algal debris and fecal pellets. Branched C 15 fatty acids also increased over time ( Figure 3); Firmicutes genera have been documented to contain 45% C 15 branched FAs by weight (Chan et al., 1971) and were previously implicated in hypersaline environments with the presence of branched compounds (Jiang et al., 2006). These fermenters in turn would have potentially supplied labile organic substrates for sulfate reducers, as implied by our incubation experiments (Figure 7), which only appeared in significant abundance after prolonged stratification. This transition was surprisingly slow, with the first sulfate reducers appearing three months after hypolimnion anoxia was established, and others appearing after a year of anoxia. We discuss the temporal evolution of sulfur cycling in more detail in Section 4.3.
Interestingly, most of the OTUs observed in the meromictic hypolimnion have been previously identified in Mono Lake sediments (Rojas et al., 2018). To confirm this connection, we characterized the microbial community in sediment core tops in spring 2018 and found persistent symmetry in sediment and hypolimnion bacteria, including genera with fermenting and sulfate-reducing potential ( Figure S5).
We hypothesize that the bacterial community developing in Mono Lake's meromictic deep water directly reflects sedimentary species colonizing this expanding, anoxic niche. Previous studies support this claim, finding high resistance in sediment bacterial communities while water-column communities are disturbed (Diao et al., 2017).
These authors postulated that sediments in stratified lakes act as seed banks for anaerobic bacteria and recent reviews have implicated dormancy in seed banks as a mechanism for species resuscitation following environmental changes (Lennon & Jones, 2011).
Indeed, spore-forming Clostridia, which were detected in sediment core tops, were abundant OTUs in the anoxic hypolimnion community of Mono Lake from fall 2017 onward. Taken together, our results for the hypolimnion appear consistent with the idea that sediment seed banks act as a buffer when disturbance alters the chemical composition of a stratified lake.

| Temporal evolution of sulfur cycling
Mono Lake supports an active microbial sulfur cycle, with ~110 mM sulfate concentration (Table S2)  In contrast, multiple studies of the lake under sustained meromictic conditions (>1 year after the onset of stratification) have observed microbial sulfate reduction by diverse deltaproteobacteria at depth (Edwardson & Hollibaugh, 2018;Humayoun et al., 2003;Oremland et al., 2000;Scholten et al., 2005), with sulfide oxidation restricted to the chemocline edge (Edwardson & Hollibaugh, 2017). Our time series provides a detailed view of the ecological transition between these two lake states.
In spring 2017, Mono Lake was monomictic and resembled that described by Stamps et al. (2018), with ~13% of total 16S rRNA reads belonging to Thioalkalivibrio ( Figure 5). No OTUs representing putative sulfate reducers were observed. By summer 2017, a shallow oxycline had developed, providing a niche with concurrent sulfide and light but minimal oxygen. Ectothiorhodospira, an anoxygenic phototroph capable of using reduced sulfur or arsenite as an electron donor, appeared in shallow waters. By fall 2017, Mono Lake had a ~5 m oxycline and hydrogen sulfide was first detected by smell. Thioalkalivibrio remained abundant below the chemocline, Ectothiorhodospira disappeared, and 16S rRNA sequences for putative sulfate-reducing organisms appeared for the first time, albeit at low (<1%) relative abundance ( Figure S2) Thioalkalivibrio remained abundant at depth accounting for 2%-4% of reads.
Not until one year after thermal stratification-and at least ten months after anoxia pervaded the hypolimnion-did sulfate reducers reach >1% of the microbial assemblage. This observation of a slow transition from sulfur oxidation to sulfate reduction is in contrast to numerous studies of freshwater lakes, which document rapid expansion of water-column sulfate reduction following seasonal stratification and anoxia, typically within a few weeks to a month (Cappenberg, 1972;Cook & Schindler, 1983;Hartland et al., 2015;Ingvorsen et al., 1981). Mono Lake's more gradual shift has several implications. First, there is a clear need to consider diversity in the context of time since last turnover in these alkaline, seasonally stratified lakes. Single snapshots of diversity are not necessarily comparable even given a similar lake mixing regime because of hysteresis effects. In Mono Lake, for example, we do not know how long meromixis is required to reach a stable microbial ecosystem.
Second, under monomixis, it is unlikely that sulfate reduction would ever become strongly established in Mono Lake's water column. One should therefore be cautious in assuming that seasonally stratified soda lakes will harbor significant populations of sulfate reducers.
Third, and more generally, it is unknown whether marine populations of sulfate reducers will adjust quickly like those in freshwater lakes, or very slowly like those in Mono Lake. The implications are significant for generation of toxic concentrations of hydrogen sulfide in seasonal ocean anoxic and hypoxic zones (Thrash et al., 2017;Ulloa et al., 2012).
Limited sulfate reduction in hypersaline lakes has previously been attributed to less favorable energetics of sulfate reduction relative to demands of osmoregulation at high salt stress (Oremland et al., 2000;Oren, 1999). We note that sulfate reducers in Mono Lake multiplied only after the appearance of abundant fermenters, particularly Clostridia and Bacteroidia, in fall 2017 (Figures 5 and   6b). Fermenters typically fill an important ecological function of converting polymeric organic molecules into small organic acids that serve as substrates for anaerobic respiration. We hypothesize that the establishment of sulfate reducers in Mono Lake may be limited more by suitable organic substrates than by the metabolic demands of high salinity. This idea is supported by our microcosm incubations (Figure 7), where sulfide production was significantly enhanced in anoxic Mono Lake water by the addition of lactate and pyruvateyielding sulfide concentrations up to 3x more than unamended controls. While DOC concentrations in Mono Lake are extremely high (>10 mM), its composition could be highly polymeric and thus largely inaccessible to most sulfate reducers. Future studies could further probe the molecular composition and reactivity of Mono Lake's DOC to study relationships between the available organics and the community of fermenters and sulfur cyclers.

| Enigmatic ecophysiology of Picocystis
Picocystis sp. strain ML, a close relative of Picocystis salinarum, is a small (~3 µm) coccoid green alga that has been an important phytoplankton species in Mono Lake for at least 40 years (Winkler, 1977). While reports through the 1990s list it as subequal in abundance with the diatom Nitzschia (Jellison & Melack, 1993;Roesler et al., 2002), recent genetic data suggest a more dominant role for Picocystis in Mono Lake, at least during spring through fall (Edwardson & Hollibaugh, 2018;Stamps et al., 2018;this study). For example, in our 16S rRNA data, chloroplast sequences that were not attributable to Picocystis comprised <5% relative abundance, over all times and depths. Fatty acid distributions were also extremely similar between lake POC and cultured Picocystis (Figure 3). Spring chlorophyll a concentrations and turbidity have risen markedly since ~2003 (Collins, 2017) suggesting the shift is relatively recent. The ecological pressures leading to this alga's rise and the subsequent loss of diatoms are unclear but may include the ability of Picocystis to thrive in low light at the base of the chemocline throughout extended meromictic events.
Picocystis strain ML was originally isolated from a chlorophyll maximum layer near the base of the chemocline during a meromictic phase of the lake in 1997 by Roesler et al. (2002). Subsequent characterization revealed that Picocystis strain ML exhibits growth across a remarkable range of salinity (0-260 mg/g), pH (4-12), and light (0.6-12 µmol/m 2 s); this taxon is also resistant to sulfide (100 µM or more), does not require O 2 , and exhibits maximal growth rates of 1.5/d. Its low-light adaptations are likely due to a diverse suite of carotenoids and ten times higher cellular chlorophyll a concentrations compared with other green algae (Roesler et al., 2002). Its ecological niche was therefore interpreted as being throughout the photic zone but especially suited for the base of the chemocline where some sunlight penetrates, ammonia diffuses from below, and low oxygen limits grazing pressure by Artemia.
In the two decades since initial characterization, additional moni-  (Collins, 2017(Collins, , 2018(Collins, , 2019Harasick, 2016). We have compiled monthly LADWP data from 2015 to 2019, creating a time series from discrete measurements of chlorophyll a fluorescence in the water column ( Figure 8). Given that multiple genetic studies revealed Picocystis to be the only significant phototroph in present-day Mono Lake, chlorophyll a concentration provides a proxy for Picocystis abundance.
Our study also found elevated chloroplast reads throughout the hypolimnion ( Figure S1; Figure 6b), indicating that Picocystis consistently accounted for ~35% of the microbial community at depth.
Surprisingly, metatranscriptomics from deep-water samples recovered mRNA for photosynthetic genes in Picocystis, suggesting they were being actively transcribed (Stamps et al., 2018). Taken together, these results beg relatively simple questions regarding this enigmatic alga: Is Picocystis alive at the bottom of Mono Lake? And if it is alive, how is it metabolizing and possibly growing?
Despite deep chlorophyll a maxima, we do not think it plausible that Picocystis is photosynthesizing throughout the hypolimnion.
Repeated CTD measurements of PAR find <0.01 µmol/m 2 s photons (the limit of detection by our sensor) at 12-15 m ( Figure S1c). Fitting PAR data to an exponential decay, which assumes constant turbidity below the chemocline in accordance with observations during sampling, photon flux would be <10 -5 µmol/m 2 s below 17 m. Although adaptations to exceedingly low-light levels are known, for example by an anoxygenic phototrophic Chlorobium strain in the Black Sea that survives down to ~10 -4 µmol/m 2 s (Manske et al., 2005), Picocystis has only been tested as low as 0.6 µmol/m 2 s. The fact that photosynthesis genes associated with Photosystem I and II (psaA, psaB, psbA, psbB, psbC) were being actively transcribed at 25 m depth (Stamps et al., 2018) therefore suggests that those cells were recently at a shallower depth, where photosynthesis was possible, or perhaps that Picocystis photosynthetic genes are constitutively expressed. In either case, they appear to be not only alive, but thriving, in the hypolimnion.
We consider two scenarios to explain the depth distribution of Picocystis. The first is that Picocystis is actively growing at the base of the photic zone and subsequently sinking. Comparison of chlorophyll a maxima to our PAR measurements ( Figure S1) suggested that these cells are active close to a photon flux of 0.1 µmol/m 2 s, well below the previously tested minimum of 0.6 µmol/m 2 s (Roesler et al., 2002). The depth distribution then requires that those cells must sink rapidly, such that CTD profiles and discrete water samplings record high cell densities throughout the hypolimnion. The near-simultaneous appearance of Picocystis maxima throughout the hypolimnion implied settling velocities on the order of ~0.5-1 m/day, faster than the ~5-10 cm/day that have been measured for other, notably larger, green algae such as Scenedesmus (Choi et al., 2006).
However, flocculation into larger aggregate particles could perhaps achieve such rates. In this scenario, cells isolated (or sequenced) from below ~20 m depth should not be interpreted as growing at those depths, but rather as surviving in sinking particles.
A second scenario is that Picocystis growth in the hypolimnion is supported partially or entirely by non-photosynthetic metabolism.

F I G U R E 8
Chlorophyll a abundance over time and depth in Mono Lake, CA. Data were redrawn from publicly available (http://www. monob asinr esear ch.org/onlin erepo rts/) LADWP reports from 2016, 2017, 2018, and 2019. Chlorophyll a concentrations were measured by fluorometry after collection at 2, 8, 16, 20, 24, and 28 m water depth. Note that LADWP did not collect data in January of each year. The approximate transition from monomixis to meromixis is also indicated, but it should be noted that these regime shifts are gradual, not sudden In the absence of appreciable exogenous electron acceptors such as oxygen, nitrate, or nitrite, Picocystis could perhaps use fermentation to maintain redox homeostasis and conserve energy. The ability to survive transient anoxia via fermentation is widely known in many members of the green algae (Catalanotti et al., 2013).
Chlamydomonas reinhardtii, the best studied green alga in this regard, has a diverse suite of fermentative pathways and can thrive under anaerobic conditions. When we cross-referenced Chlamydamonas fermentation genes against the recently published draft genome of Picocystis sp. strain ML (Junkins et al., 2019), numerous homologous proteins were identified with high confidence (Table S5; Figure S3).
However, many of these enzymes also catalyze other cellular functions such as respiration and are alone not evidence of fermentation.
Confirmatory laboratory experiments that test the exogenous uptake and subsequent release of carbon compounds by Picocystis are required.
Regardless of whether or not Picocystis cells are active only at the base of the chemocline or throughout the hypolimnion, it is clear that this ubiquitous alga holds great significance for Mono Lake's biogeochemical cycles and ecology. Numerous questions warrant further investigation, especially in regard to its potential metabolic flexibility, remarkable sulfide tolerance, and transcription of photosynthetic genes in the dark. In addition, our measurements of POC suggested that Picocystis strongly fractionates carbon isotopes by up to ~35‰ relative to source DIC, far greater than is expected for algae at such a high pH where [CO 2 (aq)] is extremely low. Eukaryotic strategies for survival in extreme environments are less understood than bacteria and archaea (Weber et al., 2007); investigating Picocystis may offer novel insights that span environmental applications.

| CON CLUS IONS
This study captured the microbial community response of hypersaline, alkaline Mono Lake as it transitioned from monomixis to meromixis following climatic perturbation. Patterns of succession in the epilimnion were less predictable than the hypolimnion, presumably

ACK N OWLED G EM ENTS
We would like to thank Blake Stamps for the use of his code that was modified to generate our 16S OTU heatmaps. We also acknowledge Mono Lake parameters.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request. 16S rRNA amplicon sequencing data is available in GenBank under BioProject accession number PRJNA702881.