The Ribbon-Helix-Helix Domain Protein CdrS Regulates the Tubulin Homolog ftsZ2 To Control Cell Division in Archaea

Healthy cell growth and division are critical for individual organism survival and species long-term viability. However, it remains unknown how cells of the domain Archaea maintain a healthy cell cycle. Understanding the archaeal cell cycle is of paramount evolutionary importance given that an archaeal cell was the host of the endosymbiotic event that gave rise to eukaryotes. Here, we identify and characterize novel molecular players needed for regulating cell division in archaea. These molecules dictate the timing of cell septation but are dispensable for growth between divisions. Timing is accomplished through transcriptional control of the cell division ring. Our results shed light on mechanisms underlying the archaeal cell cycle, which has thus far remained elusive.

division are largely conserved with eukaryotes (2)(3)(4)(5). In contrast, the cell cycle in the lineage Euryarchaeota retains features of all three domains, including a bacterial FtsZ system of cell division (6), a eukaryotic system for DNA replication (5), and archaeaspecific firing of replication origins (7).
The tubulin homolog FtsZ has been studied in detail in many bacteria for its central role in cell division. In most bacteria, FtsZ monomers assemble into short filaments that form the cytokinetic ring at midcell, which constricts to divide the mother cell into two daughters of equal size (8)(9)(10)(11)(12). FtsZ in archaea appears to function similarly, as previous fluorescence imaging experiments in fixed (13)(14)(15)(16) and live (17)(18)(19) hypersaline adapted archaeal cells (species Haloferax volcanii [Hfx. volcanii]) demonstrated Z-like rings forming at midcell. However, all known halophilic archaeal genomes encode multiple tubulin-like proteins (13,20), so the function and mechanism of these proteins in cell division remain unclear.
Halobacteria, a clade of hypersaline-adapted Euryarchaeota, provide excellent model systems for understanding cell cycle mechanisms and how they are regulated. In particular, for the model species Halobacterium salinarum (Hbt. salinarum), Hfx. volcanii, and Hfx. mediterranei, large and facile toolkits enable genetic manipulation (knockouts, overexpression, etc.) (21)(22)(23)(24). For Hbt. salinarum strain NRC-1, large systems biology data sets, including transcriptomic profiles under a wide array of growth and stress conditions, enable rapid hypothesis generation regarding gene functions (25,26). In previous work, we developed live-cell, time-lapse microscopy methods for hypersaline-adapted archaea to overcome the challenges of rapid salt crystallization on microscopy slides (27). Salt-impregnated agarose microchambers were fabricated using soft lithography, which support up to six generations of growth for Hbt. salinarum.
Using these tools, we demonstrated that single, rod-shaped Hbt. salinarum cells grow (elongate) exponentially, adding a constant volume between divisions (the "adder" model of cell size control [28]). However, the size distribution and division site placement at midcell demonstrated greater variance than bacterial cells that maintain their size in a similar fashion (27). Here, we adapt microfluidics for Hbt. salinarum and leverage the existing genetics and systems biology toolkits to interrogate the regulation of the archaeal cell cycle.
Cell cycle progression in eukaryotes is known to be exquisitely regulated, and DNA replication and cell division are coordinated in bacteria (29). However, despite recent progress regarding cell growth and size control in archaea, the underlying molecular mechanisms regulating these processes remain unknown. Gene expression profiling experiments suggest that archaea possess the capability for oscillating gene expression patterns, a hallmark of genes with cell cycle-related functions in eukaryotes (30). For example, our prior work with transcriptomics in Hbt. salinarum provides evidence for temporally coordinated induction of hundreds of genes during the resumption of growth following stasis (31). Oscillating gene expression was observed in Hbt. salinarum cultures entrained to day-night cycles (32). Cyclic gene expression patterns have also been observed in synchronized cultures of the crenarchaeon Sulfolobus solfataricus (3).
Gene regulatory networks (GRNs), comprised of interacting transcription factors (TFs) and their target genes, are central to the process of dynamic, physiological response to a variable environment. Archaeal transcription proteins resemble those of both bacteria and eukaryotes at the level of amino acid sequence. Basal transcriptional machinery required for transcription initiation in archaea, like that of eukaryotes, consists of transcription factor II B, a TATA binding protein, and an RNA-Pol II-like polymerase (reviewed in reference 33). The proteins that modulate transcription (e.g., activator and repressor TFs) typically resemble those of bacteria, with the majority of these proteins possessing helix-turn-helix (HTH) or winged-HTH DNA binding domains (34). Our recent studies on GRNs in Hbt. salinarum systematically investigated the function of transcription factors using high-throughput phenotyping of TF knockouts (35,36). This study implicated the putative TF DNA binding protein VNG0194H (VNG_RS00795) as a candidate regulator of multiple stress responses: deletion of VNG0194H led to a growth defect under multiple stress conditions, including oxidative stress, low salinity, and heat shock (35). Intriguingly, the VNG0194H gene is located upstream of ftsZ2 (37), suggesting additional roles for VNG0194H in cell growth and/or division. An additional putative DNA binding transcriptional regulator VNG0195H is encoded upstream.
To address knowledge gaps regarding archaeal cell division mechanisms, we investigated here the cell growth and division functions of FtsZ2, VNG0194H (CdrS [cell division regulator short]) and VNG0195H (CdrL [cell division regulator long]). We combine a battery of assays, including genetic knockouts, quantitative time lapse microscopy of single cells, custom microfluidics technology, gene expression profiling, and TF-DNA binding ChIP-seq experiments. The resultant data demonstrate that CdrS and FtsZ2 are required for normal cytokinesis but not cell elongation. This regulation is accomplished via (i) CdrS activation of ftsZ2 and other cell cycle-related genes and (ii) CdrL direct regulation of the cdrS-ftsZ2 operon. The CdrSL GRN system is highly specific to regulation of ftsZ2 at the level of transcription.

RESULTS
cdrS encodes a conserved, putative transcription factor co-transcribed with the tubulin-encoding ftsZ2 gene. Our previous genetics experiments indicated an important role for putative DNA binding protein VNG0194H in stress response and growth physiology of Hbt. salinarum (35,36). We first used bioinformatics to generate hypotheses regarding the physiological function of VNG0194H and its encoding gene locus (Fig. 1A). VNG0194H is predicted to encode a small 55-amino-acid, single-domain protein that exhibits Ͼ99% structural homology to other ribbon-helix-helix (RHH) domain transcriptional regulators of the RHH_1 family (Pfam accession PF01402, E value of primary sequence homology 5.3 ϫ 10 Ϫ5 ; 99.6% confidence in structural homology to transcription factor NikR), suggesting that it may function as a DNA binding transcriptional regulator or in protein-protein interactions (38). Encoded immediately upstream, the VNG0195H protein is predicted to contain an N-terminal RHH_1 domain (95.9% confidence in structural homology to NikR) and a C-terminal double zinc ribbon domain (DZR, PF12773, E value 3.1 ϫ 10 Ϫ7 ). The presence of VNG0195H in this locus appears to be unique to halophiles (see Fig. S1 [all supplemental material can be accessed via the FigShare repository associated with this study at https://doi.org/10.6084/m9.figshare .12195081]). In addition, a tubulin FtsZ homolog is encoded immediately downstream of the VNG0194H gene (ftsZ2; VNG0192G; Fig. 1A). Hbt. salinarum FtsZ2 exhibits strong primary sequence identity to known tubulin components of the cell division ring (tubulin/FtsZ GTPase domain PF00091, E value 7.4 ϫ 10 Ϫ77 ; tubulin C-terminal domain PF03593, E value 6 ϫ 10 Ϫ32 [18]). Taking these bioinformatic analyses together, we renamed VNG0194H as "CdrS" for "cell division regulator Short" and VNG0195H as "CdrL" for "cell division regulator Long." Previous tiling microarray studies indicated cotranscription of cdrS and ftsZ2 but were inconclusive regarding the inclusion of cdrL in this operon (39). To further investigate the transcriptional status of this locus, we examined the expression of the three genes from 1,154 microarray and three RNA-seq transcriptome profiles for Hbt. salinarum grown under a wide variety of environmental and genetic perturbations (25,26) (Fig. 1B). Across all conditions, cdrS and ftsZ2 were strongly and significantly correlated (Spearman's ϭ 0.923, 95% confidence interval [CI] ϭ 0.915 to 0.932), whereas cdrL was weakly but significantly correlated with cdrS ( ϭ 0.275, 95% CI ϭ 0.221 to 0.327) and ftsZ2 ( ϭ 0.299, 95% CI ϭ 0.245 to 0.350). As a control, we calculated the correlation of these genes with an unrelated gene located elsewhere in the genome (trmB VNG1451C), which exhibited weakly negative correlation with the locus (cdrS, ϭ Ϫ0.275, 95% CI ϭ Ϫ0.327 to Ϫ0.221; ftsZ2, ϭ Ϫ0.291, 95% CI ϭ Ϫ0.343 to Ϫ0.238). The weak but significant correlation of cdrL with ftsZ2 and cdrS is driven by strong coexpression of the three genes in response to stress such as metal overload (50 transcriptome profiles, ϭ 0.771, CI ϭ 0.627 to 0.864; Fig. 1C, right). In contrast, cdrL is not coexpressed with ftsZ2 or cdrS under conditions that foster rapid growth (58 profiles; ϭ Ϫ0.158, 95% CI ϭ Ϫ0.400 to 0.104; Fig. 1C, left). Previous statistical models that inferred the global gene regulatory network of Hbt. salinarum also predicted coregulation of cdrS and ftsZ2 under all growth conditions, whereas cdrL was only coregulated with the other two genes under a subset of conditions (25). Together, these results suggest that cdrS and ftsZ2 are cotranscribed from a polycistronic operon that is coregulated under all growth conditions. cdrL, in contrast, is conditionally coregulated with the other two genes.
To determine how broadly this locus is conserved outside Hbt. salinarum, we investigated the sequence conservation of each gene and the genomic synteny of the gene pair across archaeal genomes. The cooccurrence of cdrS with ftsZ2 homologs was detectable across representatives of all known archaeal clades except DPANN ( Fig. 1D; see also Table S1 in the supplemental material). FtsZ in the absence of CdrS was also widely distributed. Conservation of the cdrS-ftsZ2 locus was particularly strong across the Euryarchaeota, with wide conservation across the halophilic archaeal clade (including the classes Halobacteria, Natrialbales, and Haloferacales) and neighboring phylogenetic class Methanomicrobia (see Fig. S1). CdrL was widespread across the Halobacteria but absent from all other archaeal clades. Taken together, these results suggest that (i) cdrS exhibits a strong primary and secondary structural homology to transcriptional regulators of the RHH family; (ii) the cdrS-ftsZ2 locus encodes a highly conserved, coregulated transcriptional unit; and (iii) CdrL is a putative transcription regulator unique to halophiles and appears to be conditionally coexpressed with cdrS-ftsZ2.
CdrL is a specific and direct regulator of the cdrS-ftsZ2 operon.   (78)(79)(80). Light purple boxes indicate clades with genomes found to encode at least one FtsZ protein but no CdrS (see also Table S1). Blue and dark purple boxes indicate clades with genomes that encode genes homologous to cdrS and ftsZ2 in synteny (see also Fig. S1). Darnell et al. by chromatin immunoprecipitation coupled to sequencing (ChIP-seq; see Materials and Methods). Given the synteny and cdrL conditional coexpression with cdrS-ftsZ2, we reasoned that CdrL may play a role in regulation of the locus. The FLAG epitope was integrated into the chromosome at the 3= end of the native cdrL locus, with the resultant strain encoding a C-terminal CdrL-FLAG translational fusion (see Materials and Methods; see Table S2). In both the mid-logarithmic and the stationary phases of growth, the region upstream of the cdrS-ftsZ2 locus was the only significant CdrL binding site reproducibly detected throughout the genome (Fig. 2). Significant binding at other locations was detected in some ChIP-seq samples. However, these binding events were detected in genomic regions with redundant genes or poor coverage in the input sample. Other binding regions were not reproducible across replicate samples (see Fig. S2). We conclude that CdrL is a specific and direct regulator of cdrS-ftsZ2 expression, binding exclusively and reproducibly upstream of this locus. The cdrS-ftsZ2 locus is important for maintaining cell size and biomass in bulk culture. To test the function of the cdrL-cdrS-ftsZ2 locus, we constructed independent gene deletion mutants in each coding region. We previously reported a ΔVNG0194H strain (35,36); however, that strain included a start site for the ftsZ2 gene that was misannotated in the NCBI database, which we have corrected here ( Fig. 1A; see also Fig.  S1). This enabled a more precise, conservative deletion within cdrS to avoid polar effects by keeping the putative ftsZ2 ribosome binding site intact ( Fig. 1A; see also Tables S2 to S4). Because halophilic archaea are highly polyploid (40), stringent quality controls were implemented for both ΔcdrS and ΔftsZ2 strains, including PCR, Sanger sequencing, and whole-genome Illumina resequencing (see Materials and Methods). These controls confirmed that the cdrS and ftsZ2 coding genes were removed from all genome copies and that no second-site mutations had accumulated (Table S5). However, second-site mutations were detected in the ΔcdrL mutant (Table S5). We therefore focused our remaining analysis on CdrS and FtsZ2.  Table 1). These length and area measurements are consistent with previous observations for Hbt. salinarum growth and division (27). Also, 10.3% of the cells were longer than one standard deviation above the geometric mean; these longer cells in the rightward skew of the distribution are commonly seen during routine culturing and contribute to the noise in the Hbt. salinarum cell division model (27). Similar cell lengths were measured at time points sampled in mid-log phase and stationary phases of the growth curve (Table 1; see Fig. S3). In contrast, early exponential-phase cultures of the ΔcdrS strain cells were significantly longer (Welch's P Ͻ 2 ϫ 10 Ϫ16 ; medium effect size, 0.760) and larger in area (P Ͻ 2.67 ϫ 10 Ϫ16 ; medium effect, 0.665) than those of the parent strain, with the longest cells being Ͼ40 m. Large variation in size distribution of mutant length and area were also detected ( Table 1; Fig. 3A, middle panel; see also Table S6). Similarly, the ΔftsZ2 strain cell size was significantly larger and more variable than that of the parent strain (length, P Ͻ 2.2 ϫ 10 Ϫ16 ; large effect, 0.832; area, P Ͻ 3.52 ϫ 10 Ϫ15 , medium effect, 0.571; Fig. 3A, bottom panel). These differences in cell sizes between the parent and mutant strains were consistent across growth phases. Together, these microscopy results indicate a role for ftsZ2 and cdrS in maintaining wild-type cell size in Hbt. salinarum.
However, it remained unclear whether the drastic increase in cell size together with morphology defects in ΔftsZ2 and ΔcdrS mutants were due to unregulated growth (cell elongation or biomass accumulation) or a decrease in cell division (fewer septation events). To compare the rate of biomass accumulation between wild-type and mutant strains, we measured growth rates by determining the optical density at 600 nm (OD 600 ) for each of the parent, ΔftsZ2, and ΔcdrS strains in batch culture under aerobic conditions in rich media (see Materials Methods). The maximum instantaneous growth rate of the Δura3 strain was 0.152 Ϯ 0.004 h Ϫ1 (see Table S6), consistent with previous observations (35,36). Neither the ΔcdrS nor the ΔftsZ2 mutant strain exhibited a growth rate defect measured by OD 600 and reached similar carrying capacities (OD 600 ϭ 1.73 to 2.62 for all three strains after 94 h; see Table S6; Fig. 3B). This suggests that deletion of neither cdrS nor ftsZ2 reduces biomass as measured from the OD.
However, in the spectrophotometer, elongated cells scatter light differently than short cells (41), which can obfuscate true defects in cell size and/or division. Therefore, in addition to OD readings, we also plated for CFU. As the largest range of cell lengths occurred during early exponential phase, we plated multiple time points in the linear OD range between lag phase and an OD of 0.2. We detected a strong and significant positive correlation between log 2 -transformed OD and log 10 -transformed CFU/ml for the Δura3 parent strain (Pearson's ϭ 0.9561, P ϭ 1.188 ϫ 10 Ϫ6 ; Fig. 3C). Similar correlations were detected for each of the ΔcdrS ( ϭ 0.8166, P ϭ 1.19 ϫ 10 Ϫ3 ) and ΔftsZ2 ( ϭ 0.8509, P ϭ 4.494 ϫ 10 Ϫ4 ) strains. These strong correlations enabled direct comparison between strains of the CFU normalized by the OD (see Materials and Methods). The ΔftsZ2 strain yielded 2.2-fold fewer CFU/ml per log 2 (OD) compared to the Δura3 parent, and the ΔcdrS strain had 2.6-fold fewer. These results suggest that fewer viable individual cells are present in cultures of the mutant strains compared to the Δura3 strain, which is likely the result of larger mutant cell size (more biomass per CFU). Together, these colony counts, cell density, and quantitative microscopy results support the hypothesis that cdrS and ftsZ2 gene products are important for cell division but not cell area increase in batch culture. ⌬cdrS cells are phenotypically insensitive to aphidicolin. To further test whether growth and cell division are decoupled in the ΔcdrS mutant, we synchronized populations of cells by treating with the cell cycle inhibitor aphidicolin, which specifically The CdrSL System Regulates Archaeal Cell Division ® targets DNA polymerase ␣ in eukaryotes (42). Aphidicolin has been shown to impair DNA replication and cell division but not elongation in wild-type Hbt. salinarum (15,43,44). Here, we quantified cell area prior to aphidicolin addition, after 6 h of cell cycle block in the presence of drug, and 11 h after drug removal (Fig. 4). After aphidicolin treatment, the Δura3 average cell area increased significantly from 4.42 to 7.44 m 2 (geometric mean; Fig. 4A; P Ͻ 2.0 ϫ 10 Ϫ16 ; large effect size [ES] of 1.146), suggesting continued elongation in the absence of division and consistent with previous observations (15). After aphidicolin removal, Δura3 cells returned to an average area indistinguishable from that of pretreatment values (4.31 m 2 ; P ϭ 0.59; negligible ES 0.026). Cells undergoing septation were also observed, indicating a recovery of cell division (see Fig. S4). In contrast, the distribution of the ΔcdrS cell area remained largely unchanged by aphidicolin treatment (geometric mean of 8.73 m 2 before addition and 9.43 m 2 after 6 h of treatment, respectively; Fig. 4B  To ensure that the insensitivity to aphidicolin was specific to the deletion of cdrS, we generated a complementation strain by integration of P cdrS -cdrS into the chromosome at a neutral locus (NC_002607.1, 1245981 to 1247318; see Materials and Methods) (23). Prior to aphidicolin treatment, the complemented strain cell area was intermediate between that of the Δura3 parent and ΔcdrS mutant strains (6.45 m 2 ), suggesting partial complementation (Fig. 4C, Table 2). However, after exposure to aphidicolin, the average cell area increased to 9.52 m 2 (P Ͻ 3.9 ϫ 10 Ϫ14 , small ES 0.448, Table 2). After aphidicolin washout, the cell area mean returned to a smaller area of 4.78 m 2 (2 ϫ 10 Ϫ16 , large ES 0.887). These shifts in distribution across the time course indicate that the complemented strain was responsive to aphidicolin, suggesting that complementation was achieved under native transcriptional control at the ectopic site. Therefore, while the area of the Δura3 and the cdrS complementation strain were affected by aphidicolin treatment, the ΔcdrS mutant remained insensitive. Consistent with results from batch culture growth, these data suggest that CdrS is important for cell division but not elongation. In addition, given that aphidicolin treatment causes elongation of Δura3, phenocopying ΔcdrS, these data support the idea that cdrS may act as a checkpoint regulator, acting in the pathway that arrests cell division when DNA replication is perturbed.
Time-lapse microscopy reveals cell division defects in ⌬cdrS and ⌬ftsZ2 mutants. In previous work, we designed agarose chambers that allowed real-time microscopic observation of unperturbed growth of H. salinarum (27). However, these chambers were square (10 by 10 m), precluding growth of the filamentous ΔftsZ2 and ΔcdrS strains and necessitating a new device. We adapted the mother machine microfluidic device for real-time growth observation of these mutant Hbt. salinarum strains. The mother machine design used for hypersaline adapted archaea here is the same as those used previously for bacteria (45), consisting of linear channels 1.5 m wide, 1 m deep, and a selection of lengths (80 m used here; see Fig. S5). The mold for the chip was fabricated via photolithography. The chip is arrayed with ϳ200 troughs per feeding channel (four feeding channels per slide). Each chip was connected to microfluidics, supplying the growing cells with fresh medium throughout the course of each experiment (see Materials and Methods). Under steady-state growth conditions in the mother machine, the Δura3 cell area doubling time (6.85 Ϯ 1.98 h; Fig. 5A) was similar to that in batch culture (6.68 h; Fig. 1B; see Table S6). Mother machine growth rates also reflected previous chamber growth measurements (6 Ϯ 1 h) (27). Like the square chambers, the mother machine supports up to 6 generations of growth (see Movie S1 in the supplemental material). We conclude that the Hbt. salinarum parent strain grows optimally in the mother machine.
To compare division events from single cells across strains in real time using the mother machine, phase-contrast time-lapse images were used to quantify the cell growth parameters of elongation rate, division ratio, and interdivision time. The cell area doubling times (proportional to elongation rate) of the ΔcdrS (geometric mean, 7.88 h Ϯ 2.21 h) and ΔftsZ2 (8.04 Ϯ 2.03 h) mutants were statistically indistinguishable from the Δura3 parent strain (6.86 Ϯ 1.98 h; Fig. 5A; see Fig. S6; P values comparing each mutant to the parent of Ͼ0.76; effect size, 0 to 0.1). This corroborates batch culture results that CdrS and FtsZ2 are not required for cell elongation. In contrast, division was strongly impaired for each mutant relative to the parent strain. Division of the ΔcdrS strain (30 of 108 cells divided) was observed at 35% the frequency of Δura3  (27). In contrast, ΔcdrS cells divided asymmetrically, with few cells observed to divide in the center (Fig. 5B). This asymmetric division pattern was not  (Fig. 5B). The mean area of the ΔcdrS mother cells at the time of division (geometric mean, 19.64 m 2 ; ϭ 17.57) was significantly larger and more variable than that of Δura3 ( Fig. 5C; 8.79 m 2 ; ϭ 2.52; Welch's P ϭ 1.7 ϫ 10 Ϫ7 ; effect size, 1.50). ΔcdrS daughter cell mean area was twice as large as that of Δura3 ( Fig. 5D; 8.81 m 2 ; ϭ 13.67 versus 4.32 m 2 ; ϭ 1.60, respectively; Welch's P ϭ 2.3 ϫ 10 Ϫ10 and effect size 1.16), albeit the smallest ΔcdrS cells were within the same size range as Δura3 cells (ϳ3 to 10 m). This suggests that asymmetric division leads to variable cell sizes of mothers and daughters, with a tendency toward larger cell size in ΔcdrS relative to that of the Δura3 parent ( Fig. 5C  and D).
In the same cells visualized for quantitation with phase-contrast imaging, we tracked FtsZ1 division rings to differentiate active division events from cell fragmentation. The gene encoding monomeric superfolder GFP (msfGFP) was integrated at the native chromosomal locus by translational fusion to FtsZ1 in each of the three strain backgrounds (see Materials and Methods; see Table S2 in the supplemental material). In the Δura3 strain, we observed that cell division was preceded by helical assembly of the FtsZ1 ring ( Fig. 6A; see Movie S1 in the supplemental material). In contrast, deletion of ftsZ2 abrogated ring formation in some cells, with a diffuse msfGFP-FtsZ1 signal observed throughout the cell (see Movie S3; Fig. 6B). In other ΔftsZ2 cells, division rings formed but constriction was not observed (see Movie S4). These rings appeared decondensed relative to the parent strain. Despite these division defects, ΔftsZ2 cells continued to elongate for the duration of the imaging experiments, filling the chamber (see Movie S3; Fig. 6B). For ΔcdrS cells that were able to divide (though at a lower frequency than Δura3 cells), each division event was preceded by formation of a msfGFP-FtsZ1 ring ( Fig. 6C; see Movie S5a in the supplemental material). However, not all ring formation resulted in division: in ΔcdrS cells that were not observed to divide, rings often formed but later disassembled ( Fig. 6D; see Movie S5b in the supplemental material). Taking these fluorescence images together with the quantitative analyses of the phase-contrast images (Fig. 5), these data strongly suggest that CdrS is an important regulator of cell division, and FtsZ2 is required for triggering cytokinesis at midcell. However, elongation appears to proceed independently of these factors.
CdrS specifically regulates cell division and other cell cycle genes. To determine how CdrS regulates cell division, we compared gene expression of the ΔcdrS strain to the isogenic parent Δura3 over the growth curve and in response to cell cycle arrest by aphidicolin. Given the cell division defects of the ΔcdrS strain, we focused on 20 genes known or predicted to be involved in growth and division in other microbial systems, including all known ftsZ and cetZ paralogs encoded in the Hbt. salinarum genome. We used NanoString probe-based mRNA counting technology. Previous work demonstrated that this method is successful for accurate quantification of gene expression over time in Hbt. salinarum, enabling direct counting of mRNA molecules (46). Of the genes tested, 10 were significantly differentially expressed in the Δura3 parent strain in batch cultures over the course of the growth curve, including ftsZ2 and three CetZ homologs-cetZ1 (VNG1933G), cetZ2 (VNG0265G), and cetZ5 (VNG6260G) (see Fig. S7A and Table S7 in the supplemental material). Relative to the Δura3 control strain, ftsZ2, cetZ1, and sojA were significantly differentially expressed in response to cdrS deletion during growth (Fig. 7A). The protein product of plasmid-encoded sojA is a predicted member of the SIMIBI superfamily (NCBI accession cl28913), encompassing NTP-ases involved in a wide array of cellular functions, including the plasmid partitioning ParA AAA-type ATPase widely conserved in bacteria (47). ftsZ1 was not differentially expressed either over the growth curve or in the ΔcdrS versus the parent strain (Fig. 7A, upper left). In contrast, ftsZ2, cetZ1, and sojA were dynamically expressed throughout growth, and expression levels were lower in ΔcdrS during early log phase (Fig. 7A). For example, ftsZ2 expression steadily increased ϳ1.8-fold throughout growth, reaching its peak during the transition to stationary phase (Fig. 7A, upper right; see also Table S7). In the ΔcdrS strain, ftsZ2 expression followed a similar growth-dependent expression pattern. However, expression magnitude ranged from 1.3-to 2.2-fold lower across all growth time points in the ΔcdrS strain, with the largest defect in gene activation observed at the early log time point, with cetZ1 and sojA following similar patterns (Fig. 7A). In response to cell division block by aphidicolin and subsequent release into growth, 11 genes were significantly differentially expressed in the Δura3 parent strain (see Fig.  S7B and Table S7). ftsZ2 expression was also significantly reduced in the ΔcdrS strain relative to the Δura3 strain in response to aphidicolin (Fig. 7B). Three par family paralogs (sojA, sojB, and sojE) were also significantly misregulated in the ΔcdrS strain relative to the parent control under these conditions (Fig. 7B). Across both the growth and the aphidicolin treatments, we observed that absolute expression levels of ftsZ2 were extremely high (mean count, Ͼ2 ϫ 10 6 , Table S7), ranking as the third highest expression of the genes tested in the parent strain. In contrast, ftsZ1 counts were 2 orders of magnitude lower (ϳ9,000, twelfth highest expression). Since ftsZ2 but not ftsZ1 expression was reduced in the ΔcdrS mutant, CdrS may act to maintain the normal balance between ftsZ1 and ftsZ2 levels. Together, these expression data indicate that CdrS is The CdrSL System Regulates Archaeal Cell Division ® important for wild-type expression magnitude but not growth-dependent expression change of the ftsZ2, cetZ1, and SIMIBI family protein-coding genes. These results are consistent with the hypothesis that CdrS is a specific regulator of the cell division ring and other putative cell division-related functions. cdrS homologs in other Haloarchaea are required for maintaining cell shape and size. The cdrS-ftsZ2 locus was detected in all known haloarchaeal genomes (Fig. 1), and protein alignments showed strong conservation in model species across the clade (Fig. 8A). The beta-sheet region of the RHH protein was perfectly conserved, and only nine residues of the alpha-helical regions varied across these species (Fig. 8A). Given this strong conservation, we hypothesized that CdrS plays a conserved functional role as a regulator of cell division across hypersaline-adapted archaeal species. However, multiple attempts to delete cdrS in the genetically amenable model species Hfx. volcanii (HVO_0582; HVO_RS07500) and Hfx. mediterranei (HFX_0561, HFX_RS02725) were unsuccessful despite using a selection-counterselection scheme routinely used in the field (24,48). In Hfx. volcanii, 48 clones were screened by PCR across three transformations, and 292 clones were screened by PCR across four transformations in Hfx. mediterranei. No Hfx. mediterranei knockout candidate clones were detected. Eight Hfx. volcanii candidates were identified; however, Sanger sequencing detected many point mutations throughout the locus. These results, corroborated by a parallel study on cdrS in Hfx. volcanii (V. Vogel et al., unpublished data), suggest that cdrS is required for viability under laboratory conditions in these species.
Instead, we overexpressed cdrS in these species to investigate how its role in cell division is conserved. We compared phenotypes to a cdrS overexpression strain of Hbt. In contrast, strains carrying the cdrS overexpression plasmid (cdrS ϩ ) exhibited significant growth inhibition in all three species, indicating the importance of tight control of cdrS expression levels (Fig. 8B). Like the Hbt. salinarum deletion strain, the overexpression strain is elongated and filamentous relative to the rod-shaped control (Fig. 8C).
Similarly, in both Haloferax species, severe morphological defects were observed in cdrS overexpression strains compared to the disc-shaped control strain ( Fig. 8C; Table 3). cdrS ϩ cells were also 3-fold longer than the disc-shaped control, suggesting that the area increase was primarily due to elongation of the cell body. However, thickness varied along the length of the cdrS ϩ cells, often resulting in club-like shapes (Fig. 8C). Similar to Hbt. salinarum ΔcdrS ( Fig. 3 and 5), the cdrS ϩ cell area was more variable than that of the empty vector cells, suggesting impaired regulation (Table 3; see Fig. S8). Similar results were obtained with the Hfx. mediterranei system, with a 3-fold larger cell area and increased variance observed relative to the control strain (Welch's P Ͻ 2.2 ϫ 10 Ϫ16 ; effect size, 1.536 [large]; Table 3; Fig. 8C; see Fig. S8). However, unlike Hfx. volcanii cdrS ϩ overexpression cells, the Hfx. mediterranei cdrS ϩ cells were, on average, only 1.5-fold longer than control cells. We observed that cdrS ϩ in Hfx. mediterranei exhibited two major forms: one with an increase in cell area across two planes, generating plate-like cells, and the other elongated and/or club-shaped (Fig. 8C). Taken together, these data suggest that CdrS is required for maintaining wild-type disc-like cell shape and size in other halophilic archaea. Given the similarity of these phenotypes with those of Hbt. salinarum, these data are consistent with the hypothesis that CdrS in both Haloferax species is also required for cell division regulation but may play an additional role in maintaining cell shape and viability.

DISCUSSION
Growth and division are precisely controlled to ensure the coordination of cellular events. However, such regulation thus far has been unexplored in archaea. In the present study, we identify and characterize the highly conserved CdrSL gene regulatory network (GRN). Quantitative microscopy on cells from bulk culture and single-cell time-lapse images demonstrates the requirement of CdrS for cell division in the model archaeal species Hbt. salinarum ( Fig. 3 and 6). Specifically, deletion of cdrS or ftsZ2 impairs cell division but does not affect cell elongation rate, providing strong evidence that the CdrS regulatory system and FtsZ2 itself are required for coupling cell growth and division. Intriguingly, while FtsZ2 is absolutely required for cell division, a small The CdrSL System Regulates Archaeal Cell Division ® subset of ΔcdrS cells are still able to divide (Fig. 5 and 6), hinting that other mechanisms regulating archaeal cell division await discovery. Our genetic evidence, transcript profiling, and protein-DNA binding suggest that regulation is achieved via CdrS transcriptional activation of genes that encode proteins predicted to function in critical aspects of cell division (Fig. 2 and 7). These include the cell division ring (FtsZ2), cell shape maintenance (CetZ1) (18), and DNA partitioning (SojABE). One caveat is that the soj genes are encoded on the Hbt. salinarum pNRC100 and pNRC200 megaplasmids. These genomic elements are subject to frequent copy number variation (26), so further evidence is needed to determine the definitive mechanism by which CdrS affects their expression. Nevertheless, CdrS exerts its effect on these genes in early log phase and after release from a chemical cell division block, suggesting that CdrS acts during the transition from stasis to growth. CdrL provides a second level of regulation by binding the region upstream of the cdrS-ftsZ2 operon (Fig. 2). Together, these data suggest a mechanism by which the CdrSL system controls cell division.
Here, we show that FtsZ proteins have distinct but interrelated roles in cell division, which are reflected in their differential regulation by CdrS ( Fig. 5 and 7). Previous work in halophilic archaea suggests that a large class of tubulin-like proteins, FtsZ and CetZ, function in cell division and cell morphology, respectively (13,18). Here, we build on this knowledge, demonstrating that ftsZ1 expression levels are independent of CdrS regulation and remain relatively constant at different growth rates (Fig. 7). FtsZ1 rings form just prior to cell division events (Fig. 6). In contrast, ftsZ2 transcript levels fluctuate depending on the presence of CdrS, growth phase, and chemical perturbations, indicating a growth-sensitive mechanism of transcriptional regulation (Fig. 7). Although ftsZ2 can be deleted in Hbt. salinarum (Fig. 3 and 6; see also Table S5), it is essential for triggering the constriction of the cytokinetic ring during exponential growth, further suggesting condition-specific functions for FtsZ2 (Fig. 6). Therefore, the multiple copies of FtsZ within the Halobacteria, and likely other Euryarchaeota, do not appear to be redundant but instead may represent a case of subfunctionalization. CdrS plays a key role in regulating the interrelated but separate functions of the two FtsZ proteins, as ΔcdrS cells fail to activate ftsZ2 during rapid growth (early log phase), delaying cell division ( Fig. 5 and 7). Given the independence of cell elongation from regulation by CdrS and FtsZ2 ( Fig. 3 and 5) and that haloarchaeal cells grow by inserting new surface layer (S-layer) material at midcell (17), cell elongation and cytokinesis may occur at the same cell region (Z ring) but may be temporally sequential and separately regulated. CdrS appears to play a key role in coordinating these events.
Our results evoke a function for archaeal FtsZ proteins analogous with those required for chloroplast division in land plants and some bacterial species. In chloroplasts, FtsZ1 and FtsZ2 have interrelated but nonredundant functions in cell division. The two FtsZ homologs form copolymers, with one thought to be involved in divisome structure, the other involved in dynamic GTP turnover and constriction (51). Similarly, in the alphaproteobacterium Agrobacterium tumefaciens, although two FtsZ proteins copolymerize at midcell, only one is required for constriction (52). Recent results in Hfx. volcanii suggest colocalization of FtsZ1 and FtsZ2 at midcell (53). Consistent with these multi-FtsZ models of cell division, here we observed defects in Hbt. salinarum FtsZ1 ring assembly in the absence of FtsZ2 or CdrS, suggesting that FtsZ1 and FtsZ2 could also form copolymers whose stoichiometry is balanced by CdrS regulation.
The CdrS-FtsZ2 system is widely conserved across the Archaea at both the protein structural and functional levels, with CdrL restricted to the Halobacteria ( Fig. 1 and 8). At the level of primary structure, the ribbon-helix-helix (RHH) CdrS protein is detected in every major known taxonomic group of archaea except DPANN (Fig. 1). Further, RHH protein CdrL is encoded in all sequenced members of Halobacteria, though often only annotated by the C-terminal double-zinc-ribbon (DZR) domain. The CdrS and CdrL clades are phylogenetically distinct within the RHH superfamily (PF01402, Fig. 1; see also Fig. S1), suggesting independent evolutionary history. Restriction of cdrL to the Halobacteria further supports this hypothesis. Therefore, we surmise that the locus acquired cdrL after the divergence of methanomicrobial and halobacterial ancestors.
CdrS is also conserved at the functional level, since we demonstrate that CdrS is required for proper cell division across multiple species of haloarchaea, including Hbt. salinarum, Hfx. volcanii, and Hfx. mediterranei (Fig. 8). This was corroborated independently in a recent study that also demonstrated that cdrS is an essential gene whose product is required for cell division in Hfx. volcanii (Vogel, unpublished data). Given the essentiality of cdrS in both Haloferax species and the polyploidy of Halobacteria, we have included whole-genome sequencing (WGS) as an essential step of strain construction (see Table S5 in the supplemental material). Particularly with seemingly essential genes, we have found WGS more sensitive than standard PCR and Sanger sequencing in detecting all copies of a target gene, as well as ruling out second-site mutations (see also reference 54). Taking our phylogenetic and genomics evidence together, we conclude that the CdrS-FtsZ2 system is widely conserved and important for cell division across hypersaline-adapted archaea.

Bioinformatic prediction and phylogenetic analysis. Protein structural predictions of CdrS and
CdrL were conducted using the Phyre2 server (55) using default parameters, access date 2/12/2020. The top hit reported in the text was the structure RHH DNA binding domain of the NikR transcription factor (protein data bank [PDB] identifier 2BJ7; rscb.org, original structure published in reference 56). Protein primary sequence predictions are reported for protein families (PFAM; pfam.xfam.org) (57), E values of significance of the matches for each protein were found in the Hbt. salinarum genome database (37; https://baliga.systemsbiology.net/projects/halobacterium-species-nrc-1-genome/). Gene expression correlations in Fig. 1C were calculated and visualized using the corrplot and psychometric packages in the RStudio coding environment, R version 3.6.1. Synteny of the cdrL-cdrS-ftsZ2 locus (37; identifiers VNG0195H-VNG0194H-VNG0192G; NCBI identifiers VNG_RS00800-VNG_RS00795-VNG_RS00790) was determined using the SyntTax database using default parameters (https://archaea.i2bc.paris-saclay.fr/ SyntTax/) (58). All 384 archaeal genomes housed in the SyntTax database were searched with the FtsZ2 (VNG0192G) protein sequence of Hbt. salinarum. Detection of locus homologs and synteny for those genomes not included in SyntTax (Bathyarchaeota, Korarchaeota, Asgard, 20 genomes) were found using NCBI genomes database using BLAST to detect FtsZ2 homologs (sequence similarity cutoff Ͼ200 bits). Subsequent manual inspection in the NCBI genome browser (https://www.ncbi.nlm.nih.gov/genome/) detected synteny of the locus. Locus identifiers for cdrS-ftsZ2 across 93 archaeal genomes and UniProt protein identifiers for FtsZ-family homologs (including FtsZ and CetZ-like proteins) in the absence of cdrS across 1,497 genomes are given in Table S1 in the supplemental material. The CdrS protein sequence alignments shown in Fig. 8 were determined using Clustal Omega with default parameters in the DNAstar MegAlign software package.
Strains, plasmids, and primers. Halobacterium salinarum NRC-1 (ATCC 700922) was the wild-type strain used in this study. Gene deletions and chromosomal integrations were performed using two-stage selection and counterselection homologous recombination in the ΔpyrF (Δura3) strain isogenic parent background, as described previously (23), updated in (59), and subject to whole-genome resequencing here (see Table S5, Sequence Read Archive PRJNA614648). Plasmids were constructed using isothermal assembly (60) and propagated in Escherichia coli NEB5␣ (see the primer list in Table S3 for details). Primers were ordered from Integrated DNA Technologies (Coralville, IA). Final strain genotypes were verified using site-specific PCR and Sanger sequencing by Eton Biosciences, Inc. (San Diego, CA), and genomic DNA extraction, followed by Illumina sequencing (see below). The plasmids used in cloning are presented in Table S4, and the resultant strains are presented in Table S2.
For overexpression studies, Haloferax volcanii DS2 and Haloferax mediterranei ATCC 33500 were the wild-type strains. Plasmids were constructed from pJAM809 using restriction enzymes NdeI and KpnI to remove the resident open reading frame (ORF) and replace it with the cdrS gene from each species. ΔpyrE2 derivatives of Haloferax species (22,24) were transformed with NEB5␣-propagated plasmid. Due to concerns about the higher mutation rate in methylase-deficient E. coli, we opted not to passage plasmids through a Δdam Δdcm strain, in contrast to what is commonly done (61).
Media and growth conditions. Hbt. salinarum strains were routinely grown using CM medium ( 6.8). Media were supplemented with 50 g/ml uracil (Sigma) to complement the uracil auxotropy of the Δura3 background. During knockout and integrant strain construction, the first stage of selection was performed on mevinolin (10 g/ml; AG Scientific) plates (CM with 20 g/liter agar; Difco), and the second stage of counterselection on 5-fluoroorotic acid (300 g/ml; ChemImpex) was performed in agar plates. All growth was performed at 42°C, and liquid cultures were shaken at 225 rpm under ambient light. Self-replicating Hbt. sailnarum plasmids were maintained using 1 g/ml mevinolin in liquid culture. Self-replicating Hfx. volcanii and Hfx. mediterranei plasmids were maintained using 0.1 g/ml novobiocin. E. coli was grown in LB medium with carbenicillin (50 g/ml; Sigma) to maintain plasmids. Maximum The CdrSL System Regulates Archaeal Cell Division ® July/August 2020 Volume 11 Issue 4 e01007-20 mbio.asm.org 17 instantaneous growth rates were calculated as described previously (50). Raw data are provided in Table  S6 in the supplemental material. For statistical analysis of the OD versus CFU data shown in Fig. 3, we noted that the slopes of the regression lines were similar between Δura3 and each mutant, so we fit a linear model using the log 2 (OD) and genotype to predict the log 10 (CFU). In a two-way analysis of variance (ANOVA) test, we found no evidence of interaction between strain and OD versus CFU slope (Δura3 versus ΔcdrS, P ϭ 0.225; Δura3 versus ΔftsZ2, P ϭ 0.95). Therefore, a new model was fit that constrained equal slopes for the regression line for each strain, allowing us to determine the difference in CFU/ml per unit log 2 (OD) between strains. These differences are reported in the text.
gDNA extraction and Illumina sequencing. Hbt. salinarum strains were grown to mid-logarithmic phase (OD 600 ϳ0.7), and 1 ml was pelleted by centrifugation. Pellets were stored at -20°C until processed. DNA was extracted using a phenol-chloroform method. Briefly, pellets were lysed in dH 2 O and treated with RNase A and proteinase K. DNA was extracted using phenol-chloroform-isoamyl alcohol (25:24:1; Fisher Scientific) in Phase Lock Gel tubes (QuantaBio) and ethanol precipitated. DNA pellet was resuspended in modified TE buffer (10 mM Tris-HCl [pH 8.0], 0.1 mM EDTA). Purified DNA was quantified using a NanoDrop system (Thermo Scientific) and sonicated in a Diagenode sonicating water bath for 20 cycles on high. DNA quality was assessed by using a bioanalyzer using a high-sensitivity DNA chip (Agilent). Samples were submitted to Duke Center for Genomic and Computational Biology core sequencing facility for adapter ligation with TruSeq (Illumina) adapters and library amplification. Samples were pooled and run in a single lane on an Illumina HiSeq 4000. 50-bp reads were assessed for quality using FastQC, and adapter sequences were trimmed using Trim Galore and Cutadapt. Reads were aligned to the Hbt. salinarum NRC-1 genome using Bowtie2 within the breseq package, as described previously (54,62). The breseq results were analyzed for incomplete gene conversion, SNPs, and possible genomic rearrangements. The results are given in Table S5; the code is freely available on github (https://github .com/amyschmid/aglB-WGS-growth), and raw data are accessible through the sequence read archive (SRA) under accession number PRJNA614648.
Microscopy and quantitative analysis of cell morphology. Cell culture preparation for microscopic imaging. (i) Deletion mutants. Strains ΔcdrS, ΔftsZ2, and Δura3 cells were collected at various stages throughout the growth curve (see Table S6 for the OD 600 values at harvest).
(ii) Aphidicolin treatment. Cells were cultured to stationary phase and then subcultured in 20 ml of CM medium supplemented with uracil to an OD 600 of ϳ0.01. Subcultures were incubated at 42°C with 225 rpm shaking until reaching an OD 600 of ϳ0.3, at which point 30 M aphidicolin (30 mM stock in DMSO; Sigma) was added. After 6 h of incubation with drug, the remaining culture was washed twice by centrifugation (4,500 ϫ g, 6 min), resuspension in fresh CM, and incubation for 20 min at 42°C with 225 rpm shaking. Samples were harvested for microscopy (5 l) prior to aphidicolin addition, prior to washing, and 11 h after washing (release) from aphidicolin treatment. The rationale for the time point selection was based on a previously published study (15).
(iii) Overexpression strains. Five biological replicate overnight cultures of Hfx. volcanii strain H26 harboring the pAKS78 plasmid and Hfx. mediterranei WR510 harboring the pAKS77 plasmid were grown in rich Hv-YPC medium supplemented with novobiocin to maintain the plasmid and 50 g/ml uracil to complement the auxotrophy of the ΔpyrE2 background. Image quantitation and significance of the difference between cdrS ϩ overexpression strain cells and empty vector control strain in Haloferax species was calculated as described above for knockout mutants and aphidicolin experiments. To obtain micrographs of the Hbt. salinarum cdrS ϩ overexpression Δura3 strain harboring the pAKS0194-Chis plasmid, cultures were grown in rich CM medium supplemented with mevinolin to maintain the plasmid and 50 g/ml uracil to complement the uracil auxotrophy.
Microscopy. For microscopy, 5-l portions of three to five biological replicate culture samples (grown as described above) were mounted on 1% agarose pads equilibrated with basal salt buffer (medium without peptone or yeast extract) overnight at room temperature, except that the Hbt. salinarum cdrS ϩ overexpression strain was imaged on wet-mount slides. Phase-contrast images were taken with a Zeiss Axio Scope.A1 microscope (Carl Zeiss, Oberkochen, Germany) and a PixeLINK CCD camera (PixeLINK, Ottawa, Canada) at ϫ40 magnification.
Image quantitation and statistical analysis. Quantification of cell length and area was calculated using the Fiji distribution (63) of ImageJ (64) and plug-in MicrobeJ (65). For batch culture experiments (overexpression and deletion mutants), the significance of the difference in cell size (length and/or area) between mutant versus parent strain cells was calculated using Welch's modified t tests on orderedquantile-normalized data, and the effect sizes of these differences were calculated using Cohen's D test. For aphidicolin experiments, mixed-effects ANOVA with Welch's post hoc t tests on ordered-quantilenormalized data were used to determine statistical significance of the differences between strains and time points. In all cases, P values were adjusted for multiple hypothesis testing using the Benjamini-Hochberg correction.
Single-cell microfluidics time-lapse microscopy. (i) Strains and growth conditions. All strains analyzed in the mother machine experiments were streaked fresh from -80°C stocks and inoculated in 50 ml of CM supplemented with uracil in beveled flasks and grown at 42°C to an OD 600 of 0.6 to 0.8 before loading into the microfluidic chip.
(ii) Microfluidic chip fabrication. Microfabrication of the master mold used to create the microfluidic chip was performed in our mother machine experiments as previously described (45,66), except that the second, wider layer in the cell chambers meant to enhance growth was not necessary to maintain the cell's growth over the timescale of the current experiment. Specifically, to fabricate the microfluidic device, the features were molded into a piece of polydimethylsiloxane (PDMS) by pouring dimethyl siloxane monomer (Dow Sylgard 184 silicone elastomer base) mixed with a curing agent (Dow Sylgard 184 silicone curing agent) in a 10:1 ratio on top of the master mold, followed by degassing under a vacuum, and curing the setup overnight at 65°C. The solidified PDMS piece was then peeled from the master and cut into approximately 1.5 ϫ 1.5-cm chips. Access holes for each of the feeding channels were punched using a 0.75-mm biopsy punch (WPI). The PDMS chip was then bonded to a KOH-cleaned 22 ϫ 60-mm glass coverslip (VWR, no. 1.5) by oxygen plasma treatment at 200 mTorr of pressure and 30 W for 30 s in a PE-50 compact benchtop plasma cleaning system (Plasma Etch). Chips were baked at 65°C for at least 1 h before use.
(iii) Loading Hbt. salinarum cells into microfluidic devices. CM medium supplemented with 1% bovine serum albumin was manually injected into the microfluidic device using a 1-ml syringe and incubated for 1 h. To avoid the crystallization of salt from the media, microcapillary pipet tips with CM media were left attached to the access holes of the chip. Prior to loading, cell cultures were filtered through a 40-m cell strainer (EASYstrainer; Greiner Bio-One) to remove large salt crystals formed during culture growth. Cultures were then centrifuged at 3,000 ϫ g for 5 min and concentrated to a final volume of 100 l. Cells were manually loaded into the main flow channel of the microfluidic mother machine device using a 1-ml syringe. To increase the number mother machine wells containing cells, the chip was briefly centrifuged in a VWR Galaxy minicentrifuge. The microfluidic chip was then connected to automatic Harvard Apparatus syringe pumps by Tygon tubing (Saint-Gobain, ID 0.020 in) and blunt end dispense tips (Fisnar, 21 gauge, 1 in.). Fresh medium was continuously pumped at 1 to 2 l/min at 37°C for 30 min to allow cells to further propagate within the mother machine wells. The system was subsequently moved to the microscope for data collection.
(iv) Microscopy. Cells in the mother machine were imaged in a Nikon Eclipse Ti microscope with a 6.5-m pixel CMOS Hamamatsu camera and a Nikon 100ϫ NA 1.45 phase-contrast objective. Images were captured every 20 min for 1 to 2 days. Exposure times for both phase-contrast and fluorescence were 200 ms. Epi-illumination was provided by a fiber coupled Agilent launch for 488 nm to image msfGFP-FtsZ1 in strains AKS137, AKS170, and AKS196.
(v) Single-cell growth analysis from the mother machine. Using phase-contrast images, 80 Δura3, 108 ΔcdrS, and 110 ΔftsZ2 cells were manually traced in Fiji image analysis software (63) to determine the cell area through pixel counting. These measurements were used to determine the cell area doubling (elongation) rates by fitting an exponential curve to the single-cell growth rate graphs of change of the cell area over time. Hbt. salinarum has previously been shown to grow exponentially through single-cell analysis (27). The area at birth was calculated from images immediately after complete the separation of daughter cells. The area during division was determined by adding the area of the two daughter cells immediately after division. The division site placement was determined by treating each pole separately and then calculating the ratio of each daughter cell to its corresponding parent. Single-cell measurements shown in the figures were plotted using ggplot2 (67) in the RStudio coding environment. Statistical analyses were conducted using the sjstats (68) and rstatix packages in RStudio. Significance of differences in doubling time between strains were modeled using 3-way ANOVA, with effect sizes calculated using an 2 test. The significances of the differences between mother or daughter cells in parent and mutant strains were calculated using Welch's modified t tests on ordered-quantilenormalized data and the effect size of these differences calculated using Cohen's D test. P values were adjusted for multiple hypothesis testing using the Benjamini-Hochberg correction.
Gene expression analysis with NanoString. To collect RNA across the growth curve, culture aliquots were collected from flask cultures at 4 phases of growth (early logarithmic, mid-logarithmic, early stationary, and late stationary phases). A 50-ml portion of culture was sampled at a low OD 600 (ϳ0.05), followed by 2 ml at a higher OD 600 (ϳ2.0). RNA was collected after 6 h of aphidicolin treatment and 11 h after release. Culture samples were centrifuged at 21,000 ϫ g for 2 min; the supernatant was removed and immediately frozen in liquid nitrogen. Pellets were stored at Ϫ80°C overnight, and RNA was purified using an Absolutely RNA Miniprep kit per manufacturer's instructions (Agilent, Santa Clara, CA). To verify the lack of DNA contamination, endpoint PCR was conducted for 30 cycles on 200 ng of RNA sample using the primers indicated in Table S3 in the supplemental material. RNA quality was determined using a bionanalyzer and RNA Nano 6000 chip according to the manufacturer's instructions (Agilent).
Gene expression was quantified using NanoString detection and a custom probe Codeset (see Table  S7 in the supplemental material) (69). Probes were designed to target 20 genes predicted to encode proteins involved in cytoskeletal and growth functions, and 3 control genes. One hundred nanograms of RNA was hybridized and quantified using the nCounter instrument by the Duke Microbiome Shared Resource core facility. Counts were normalized using three housekeeping genes (eif1a2 [VNG_RS06805], coxA2 [VNG_RS02595], and VNG1065C [VNG_RS04150]) and NanoString nSolver software and then further normalized to expression in the NRC-1 wild-type control. The significance of this relative normalized differential expression between the parent and ΔcdrS strain was assessed using the maSigPro package (70) in the R Bioconductor coding environment with the default parameters (except a Q-value of 0.01, alfa ϭ 0.01, and an R 2 cutoff of 0.7). We conducted Benjamini-Hochberg correction for multiple hypothesis testing in the context of the maSigPro package. All raw and normalized data and probe sequences are available in Table S7 in the supplemental material. The R code is presented in the github repository associated with this study (https://github.com/amyschmid/cdr).
ChIP-seq experiment and analysis. Triplicate cultures of the AKS113 (CdrL tagged at the C terminus with the FLAG epitope; see Table S2) and Δura3 control strains were grown until stationary phase and subcultured in rich media supplemented with uracil. At mid-log phase (OD 600 ϳ0.15) and early stationary phase (OD 600 ϳ1.8), cultures were cross-linked and immunoprecipitated as described previously (59), with the following exceptions: (i) cultures were cross-linked with 1% formaldehyde for 30 min at room