Magnesium Modulates Bacillus subtilis Cell Division Frequency

ABSTRACT

By chance, we discovered a window of extracellular magnesium (Mg²⁺) availability that modulates the division frequency of Bacillus subtilis without affecting its growth rate. In this window, cells grown with excess Mg²⁺ produce shorter cells than do those grown in unsupplemented medium. The Mg²⁺-responsive adjustment in cell length occurs in both rich and minimal media as well as in domesticated and undomesticated strains. Of other divalent cations tested, manganese (Mn²⁺) and zinc (Zn²⁺) also resulted in cell shortening, but this occurred only at concentrations that affected growth. Cell length decreased proportionally with increasing Mg²⁺ from 0.2 mM to 4.0 mM, with little or no detectable change being observed in labile, intracellular Mg²⁺, based on a riboswitch reporter. Cells grown in excess Mg²⁺ had fewer nucleoids and possessed more FtsZ-rings per unit cell length, consistent with the increased division frequency. Remarkably, when shifting cells from unsupplemented to supplemented medium, more than half of the cell length decrease occurred in the first 10 min, consistent with rapid division onset. Relative to unsupplemented cells, cells growing at steady-state with excess Mg²⁺ showed an enhanced expression of a large number of SigB-regulated genes and the activation of the Fur, MntR, and Zur regulons. Thus, by manipulating the availability of one nutrient, we were able to uncouple the growth rate from the division frequency and identify transcriptional changes that suggest that cell division is accompanied by the general stress response and an enhanced demand to sequester and/or increase the uptake of iron, Mn²⁺, and Zn²⁺.

IMPORTANCE The signals that cells use to trigger cell division are unknown. Although division is often considered intrinsic to the cell cycle, microorganisms can continue to grow and repeat rounds of DNA replication without dividing, indicating that cycles of division can be skipped. Here, we show that by manipulating a single nutrient, namely, Mg²⁺, cell division can be uncoupled from the growth rate. This finding can be applied to investigate the nature of the cell division signal(s).

INTRODUCTION

The abundant cation Mg²⁺ is perhaps most appreciated for its role as an enzymatic cofactor, as it supports catalysis in hundreds of biochemical reactions (1–3). However, Mg²⁺ has diverse biological functions and is also critical for the assembly of ribosomes (4, 5), chelation, the stabilization of ATP and other polyphosphates (6), the regulation of phosphate uptake (7, 8), osmotic adaptation (9), the setting of the circadian period in plants (10), and the support of envelope integrity in bacteria (11–13). Due to the central place of Mg²⁺ in physiology, cells must be able to respond rapidly when availability fluctuates, as these fluctuations sometimes occur over several orders of magnitude. In human serum, where Mg²⁺ concentration is tightly controlled, 0.7 to 1.0 mM is considered homeostatic (14). In contrast, Mg²⁺ in the digestive tract of animals or in soil and aquatic environments is much more variable, ranging from micromolar to millimolar levels.

Whereas free-living bacteria can adapt to large fluctuations in extracellular Mg²⁺, they keep intracellular levels relatively constant. Under replete conditions, cell-associated Mg²⁺ in E. coli is estimated to be 20 to 100 mM (15, 16), of which only 1 to 10 mM is considered labile (16–19). Of the remaining pool, approximately half is associated with nucleic acid, proteins, and ribosomes. The other half is found complexed with the enzymatically relevant form of ATP, the relatively stable chelate Mg²⁺-ATP (20). Not surprisingly, ATP synthesis is tightly coordinated with Mg²⁺ availability (6). In fact, cells will scavenge Mg²⁺ from ribosomes at the expense of protein synthesis before allowing intracellular Mg²⁺ to fall to levels that are insufficient to support ATP chelation (6, 9, 21).

Intracellular Mg²⁺ is acquired using importer proteins that, in Gram-negatives, are often regulated by PhoPQ two-component systems (22, 23). These systems sense and respond to changes in both external Mg²⁺ availability and cellular demand primarily by regulating the transcription of Mg²⁺ transporters. B. subtilis lacks a homologous two-component system and controls the expression of its major Mg²⁺ importer (MgtE) via a riboswitch. The M-box riboswitch attenuates the transcription of mgtE when intracellular Mg²⁺ is sufficient by forming a terminator (4). Two other importers, YfjQ and CitM, also contribute to Mg²⁺ uptake (24). YfjQ is a minor importer, whereas CitM is a symporter that allows for the cotransport of Mg²⁺ and citrate (25, 26). Under hyperosmotic conditions and concomitant with potassium influx, B. subtilis can also efflux Mg²⁺ through an exporter called MpfA (27).

Aside from its role in supporting basic physiological functions, Mg²⁺ is known to suppress phenotypes associated with the inactivation of cell envelope-related genes. The provision of 10 to 25 mM Mg²⁺ (more than an order of magnitude higher than the concentrations found in typical media) restores both viability and rod shape to strains with deletions in the morphogenes mreB, mbl, mreBH, mreC, and mreD (28–30). Mg²⁺ restores a regular rod shape to strains with deletions in ponA (encoding PBP1A) and lytE (a major d, l-endopeptidase) (11, 31) as well as to strains with mutations in teichoic acid synthesis genes (12, 32, 33). A ΔglmR mutant, which is unable to upregulate gluconeogenesis, is inviable when grown on a gluconeogenic carbon source (34). Remarkably, millimolar Mg²⁺ can suppress this lethality (34). Mg²⁺ even increases resistance to the cell wall-targeting antibiotic methicillin (35). Thus, Mg²⁺ elicits cellular changes that allow it to function as a general suppressor of a wide variety of envelope-related defects.

The mechanism by which Mg²⁺ promotes envelope integrity is unclear. B. subtilis grown with higher levels of Mg²⁺ display lower levels of amidated meso-diaminopimelic acid in their peptidoglycan (PG) (36); however, reduced amidation itself is unlikely to account for Mg²⁺ rescue, as a mutant lacking the modification (ΔasnB) also has shape defects that are suppressed by Mg²⁺ (36). Mg²⁺ also reduces the dysregulated d, l-endopeptidase activity that is associated with deletion of mreB (37); it is unknown whether the impact of Mg²⁺ on endopeptidase activity is direct or is a consequence of other effects that Mg²⁺ has on the cell.

Here, we investigate a phenotype not previously associated with Mg²⁺. We identify a window in which increasing Mg²⁺ availability increases the frequency of cell division without affecting the growth rate. Our results suggest that as Mg²⁺ availability decreases, cells prioritize the maintenance of cell elongation and growth over cell division. This prioritization of cell resources, which transcriptional profiling suggests is accompanied by changes in metal homeostasis and in the general stress response, results in longer cells with more nucleoids and fewer Z-rings.

RESULTS

Mg²⁺ supplementation leads to cell shortening in rich media in both domestic and undomesticated B. subtilis strains.

In 2017, we obtained a new bottle of premade LB-Lennox powder from Sigma. While examining micrographs of membrane-stained B. subtilis 168 cells that were grown in liquid LB made from the new powder (LB*), we noted that the cells appeared qualitatively longer, compared to cells grown in previous lots of LB. LB is a rich medium consisting of only NaCl, tryptone, and yeast extract. We reasoned that the longer cells most likely resulted from a difference in trace element content. Mg²⁺ was a priority candidate both because tryptone-based media, such as LB, tend to be low in Mg²⁺ (37, 38) and because prior studies had shown that low Mg²⁺ media results in B. subtilis filamentation (39–41). We could not directly compare our results to those of published studies, as their microscopy methods did not include the visualization of septa present along filaments. To test whether the longer cells could be the result of low Mg²⁺ content in the new LB, we supplemented the medium with 10.0 mM MgCl₂ and imaged cells following membrane staining. As shown in Fig. 1A and B, cells grown with supplemental Mg²⁺ were 36% shorter on average, consistent with the possibility that the cells were longer in the new medium due to the reduced Mg²⁺ content.

FIG 1.

Cell length following growth in LB. WT B. subtilis 168 (BJH004) was grown at 37°C to the mid-exponential phase in LB* medium without or with 10.0 mM MgCl₂ supplementation. LB* indicates that the phenotype was lot-specific and is not generalizable to all LB. (A) Micrographs following the membrane staining and epifluorescence microscopy (scaled identically). (B) Scatterplots showing the distribution of cell lengths quantitated for 300 cells from each condition. The bars represent the means of 300 cells ± the SDs. (C) Representative growth curves. ****, P ≤ 0.0001.

Since the growth rate can affect the cell size (42–48), we next assessed whether the Mg²⁺ addition affected the doubling time of the cells in the new LB. We found that the growth rates of cells cultured without and with 10.0 mM MgCl₂ were identical during the exponential stage (Fig. 1C), the same phase of growth used for cell imaging. From these results, we concluded that there is a growth rate-independent effect of Mg²⁺ availability on cell length, at least in the window tested.

The phenotype that we initially observed was specific to only one batch of LB. So, we next investigated whether the Mg²⁺ responsive phenotype could be observed in another rich medium. CH (casein hydrolysate) is an amino acid-based medium (49) that is commonly used in B. subtilis studies. We found that B. subtilis grew robustly in CH without the standard addition of Mg²⁺ and Mn²⁺ salts (Fig. 2A) and designated this medium CH*. Similar to the results with LB*, the cells grown in CH* exhibited a Mg²⁺-responsive phenotype. B. subtilis 168 cells cultured in CH* with 10.0 mM Mg²⁺ were approximately 2-fold reduced in average cell length, compared to those cultured in CH* only (Fig. 2B and C). No difference in doubling time (Fig. 2A) or cell width (Fig. 2D) was detected between CH* and CH* supplemented with Mg²⁺.

FIG 2.

Cell length of three B. subtilis strains following growth in CH* medium. WT B. subtilis 168 (BJH004), PY79 (BJH001), and 3610 (BJH403) were grown at 37°C to the mid-exponential phase in CH* supplemented with 10.0 mM MgCl₂ as indicated. (A) Representative growth curves for the WT. (B) Representative micrographs following the membrane staining and epifluorescence microscopy (scaled identically). (C) Scatterplots showing the distribution of cell with quantitated for 300 cells from each condition. The bars represent the means of 300 cells ± the SDs. (D) Scatterplots showing the distribution of cell widths for 200 cells grown without or with 10.0 mM MgCl₂. The bars represent the means of 200 cells ± the SDs. ns, P > 0.05.

Next, we investigated whether the Mg²⁺-responsive phenotype observed in the B. subtilis 168 cells was present in two other commonly utilized B. subtilis strains: the SPβ-cured laboratory strain PY79 (50) and the undomesticated strain NCIB 3610 (50). Independent of strain, the addition of 10.0 mM MgCl₂ to CH* consistently resulted in cells that were approximately 2-fold reduced in length, compared to CH* alone (Fig. 2B and C). These results suggest that the effect of Mg²⁺ is likely to be a phenomenon that is generalizable to the species.

Mn²⁺ and Zn²⁺ supplementation also elicit cell shortening.

We wondered whether the growth rate-independent cell shortening was specific to Mg²⁺ or could also be induced by providing other metals in excess. For these experiments, we again utilized CH*. Cells were grown as before, but instead of supplementing cultures with MgCl₂, salts of Ca²⁺, Cu²⁺, Fe²⁺, Fe³⁺, Mn²⁺, and Zn²⁺ were added. Of the metals tested, only Mn²⁺ and Zn²⁺ elicited cell shortening similar to that observed for Mg²⁺ (Fig. 3; Fig. S1); however, Zn²⁺ strongly reduced the growth rate at the concentrations required to observe the cell shortening effect (0.2 mM) (Fig. S1). While Mn²⁺ did not reduce the doubling time of cells once the culture reached the exponential phase, the precultures containing Mn²⁺ remained in lag for several hours longer than did the unsupplemented cultures. Due to the pleiotropic effects of Zn²⁺ and Mn²⁺ on the growth, we chose to focus on Mg²⁺ for the remainder of the study.

FIG 3.

Effect of divalent metals on cell length. WT 168 (BJH004) cells were cultured in CH* without or with the following concentrations of divalent cation salts: 0.1 mM MnSO₄, 0.2 mM ZnCl₂, 10.0 mM MgCl₂. (A) Micrographs following the membrane staining and epifluorescence microscopy (scaled identically). (B) Scatterplots showing the distribution of cell lengths quantitated for 300 cells from each condition. The bars represent the means of 300 cells ± the SDs. *, 0.01 < P ≤ 0.05; ns, P ≥ 0.05. For all pairwise comparisons to the control without additional metal (−), a P value of <0.0001 was calculated.

The Mg²⁺-responsive phenotype occurs in minimal medium and does not require the addition of exogenous amino acids.

Minimal media (MM) allows for the manipulation of individual medium components but generally results in a slower doubling time because it requires that cells undertake extensive de novo synthesis. To test whether Mg²⁺ could modulate cell length in a defined medium, we utilized a phosphate-buffered glucose MM. We began with a base medium that contained 13 amino acids and 50.0 μM Mg²⁺ (MM-13aa). In MM-13aa, the doubling time for the B. subtilis 168 prototroph (functional trpC+) was 53 min, both with and without the addition of 10.0 mM MgCl₂ (Fig. 4A). The cell length in MM-13aa was assessed using microscopy and quantitation. As shown in Fig. 4B and C, the cells grown with supplemental Mg²⁺ were 45% shorter than those grown in MM-13aa only.

FIG 4.

Cell length of B. subtilis following growth in minimal medium (MM). WT B. subtilis 168 (trpC prototroph) (BTG169) was cultured in a base MM containing 50.0 μM Mg²⁺ supplemented with 10.0 mM MgCl₂ and amino acids as indicated. (A) Representative growth curves. (B) Micrographs following the membrane staining and epifluorescence microscopy (scaled identically). (C) Scatterplots showing the distribution of cell lengths quantitated for 250 cells from each condition. The bars represent the means of 250 cells ± the SDs.

To test whether the Mg²⁺-responsive phenotype was dependent on the presence of amino acids in the medium, the experiments were repeated in the base MM without amino acid supplementation. This modification increased the doubling time to 59 min, but the growth rate again remained unchanged, both with and without Mg²⁺ supplementation (Fig. 4A). Similar to the pattern observed in the media that contained amino acids (LB, CH*, and MM-13aa), the cells were shorter in the Mg²⁺ supplemented medium, compared to those in the unsupplemented control (Fig. 4B and C). These results suggest that the Mg²⁺-responsive phenotype is generalizable to both complex and defined media and does not depend on the addition of amino acids.

Mg²⁺-responsive cell length changes follow a dose-response curve.

Next, we wanted to know whether cell length decreased proportionally with increasing extracellular Mg²⁺ or, alternatively, there was a threshold at which cells underwent a switch in cell length. To investigate this, we cultured cells in media across a range of Mg²⁺ concentrations, from 6.25 μM MgCl₂ (at which differences in the growth rate begin to emerge at the densities at which we collected the cells) (Fig. 5A) up to 25.0 mM. Cells from each condition were imaged, and the cell lengths were quantified (Fig. 5). We found that cell shortening was neither directly proportional to Mg²⁺ availability nor a switch. Instead, cells became progressively shorter from 12.5 μM to 4.0 mM, with the largest decrease occurring between 0.2 μM and 2.0 mM (Fig. 5C and D).

FIG 5.

Correlation analysis of extracellular Mg²⁺ availability with cell length and with a reporter of intracellular Mg²⁺. WT cells harboring a reporter for P_mgtE-lacZ (BTG182) were grown in MM-13aa with the indicated concentrations of supplemental MgCl₂. (A) Representative growth curves and (B) representative micrographs following the membrane staining (scaled identically) from across the range of Mg²⁺ concentrations examined. (C) Mean cell lengths for 500 cells ± the SD for each condition (top) and the β-galactosidase activity assay for the mgtE riboswitch transcriptional reporter (bottom). (D) Scatterplots showing the distribution of cell lengths at each Mg²⁺ concentration. The bars represent the mean cell lengths for 500 cells ± the SDs. (E–G) WT (BTG182) or the ΔmpfA mutant (BTG333) harboring a reporter for P_mgtE-lacZ was grown in MM-13aa supplemented with 10.0 mM MgCl₂. Samples were taken during the exponential growth phase. (E) β-galactosidase activity assay. The bars represent the means from 2 biological replicates ± the SDs. (F) Scatterplots showing the distribution of cell lengths. The bars represent the mean cell lengths for 200 cells ± the SDs. (G) Representative micrographs of cells stained with TMA and scaled identically. *, 0.01 < P ≤ 0.05; **, 0.001 < P ≤ 0.01; ***, 0.0001 < P ≤ 0.001; ****, P < 0.0001; (ns) P ≥ 0.05.

Cells in the 0.2 μM to 2.0 mM window grew at equivalent doubling times (Fig. 5A), suggesting that the cells were not yet experiencing significant intracellular Mg²⁺ limitation. To independently assess this, we introduced a promoter fusion (P_mgtE-lacZ) that reports on changes in internal Mg²⁺ availability. The promoter region includes a riboswitch that terminates transcription when the intracellular Mg²⁺ levels are sufficient (51, 52). Using this reporter, Dann et al. observed an approximately 10-fold induction of LacZ activity following the extended growth of cells with 5.0 μM Mg²⁺ (growth rate-limiting), compared to a Mg²⁺ excess (2.5 mM in their study) (Fig. 1B in reference [51]). Consistent with these findings, we observed that LacZ activity was also induced in medium containing 6.25 μM Mg²⁺ (Fig. 5B). Notably, this induction coincided with the time when the growth rate effects began to be observable (Fig. 5A).

The steepest decline in cell-shortening occurred between 0.2 and 2.0 mM MgCl₂. To our surprise, no substantial differences in P_mgtE-lacZ activity were detectable in this window, suggesting that the cell shortening is unlikely to be attributable to increased levels of labile intracellular Mg²⁺. We considered the possibility that the riboswitch of the P_mgtE-lacZ reporter may already be saturated (terminator form) in this range and thus may be insensitive to further increases. To test this, we deleted the gene for the Mg²⁺ exporter protein MpfA, which was previously shown to increase intracellular Mg²⁺ (53). Even when cells were grown in 10.0 mM MgCl₂, the ΔmpfA mutant displayed an additional 4-fold reduction in LacZ activity, compared to the wild-type (Fig. 5E), indicating that the reporter retained sensitivity. This result suggests that if there are changes in intracellular Mg²⁺ in the range in which we observe the most dramatic decreases in cell length, then the changes are not detectable with our reporter. Notably, even though intracellular Mg²⁺ can be inferred to have increased in the ΔmpfA mutant, compared to the wild-type, no differences in average cell length were observed (Fig. 5F and G). Collectively, these results suggest that under our growth conditions, changes in intracellular labile Mg²⁺ are neither necessary nor sufficient to elicit cell shortening. At the same time, we do not exclude the possibility that the cells may be responding to changes in intracellular Mg²⁺ that are not detectable by P_mgtE-lacZ, such as an increase in the nonlabile pool. An alternative possibility is that the response is driven by differences in extracellular, or at least extracytoplasmic, Mg²⁺.

Cell shortening following a shift from lower to higher Mg²⁺ is rapid.

In the experiments above, the cell lengths were determined during steady-state growth, and care was taken to collect the cells at equivalent densities, as Mg²⁺ continues to be depleted from the medium with time. To assess the transition from longer to shorter, we grew cells in CH* and in MM-13aa, and we monitored the lengths of the cells at 10 min intervals after the addition of 10.0 mM MgCl₂ (Fig. 6). For both media, the mean cell length continued to decrease for 40 to 50 min, at which point the cells achieved a mean length similar to those of cells always grown with excess Mg²⁺ (Fig. 6). Remarkably, 54% and 40% of the total decrease in the mean cell length occurred within the first 10 min of adding Mg²⁺ in the CH* and MM-13aa media, respectively (Fig. 6). Thus, the initial response to Mg²⁺ is relatively rapid and is well below the doubling time of the cells. After the initial rapid decrease, a more gradual decline in average cell length was observed. These results suggest that the adjustment to higher Mg²⁺ may occur through two distinct mechanisms: a dramatic initial decrease in cell length resulting from rapid division onset and a second, more protracted period in which the average cell length decreases more incrementally. These observations suggest that the initial response may have a more biophysical or enzymatic basis, whereas the second requires biosynthesis and outgrowth to achieve a new steady-state length.

FIG 6.

The B. subtilis cell length response to a change in extracellular Mg²⁺ is fast and is independent of the growth rate. WT 168 (BJH004) was cultured in the indicated media without Mg²⁺ supplementation (–Mg²⁺). Following addition of 10.0 mM MgCl₂, the cells were imaged at 10 min intervals. (A and B) The bars represent the mean cell lengths for 300 cells ± the SDs. (C) Representative micrographs. The membranes are stained with TMA, and the images are scaled identically. *, 0.01 < P ≤ 0.05; **, 0.001 < P ≤ 0.01; ****, P < 0.0001; (ns) P ≥ 0.05.

Mg²⁺ modulates the frequency of Z-ring assembly.

A rod-shaped bacterium can increase cell length without altering other dimensions by elongating faster, dividing less frequently, or both. Based on the optical density, the growth rate of the cells in our experiments was equivalent, both with and without Mg²⁺ supplementation. Although absorbance readings are the most widely accepted method for monitoring cell growth, we considered the possibility that the populations that we were comparing might absorb light differently enough to mask differences in mass accumulation. To independently test this, we examined the abundance of the constitutively expressed protein SigA via a western blot analysis. In both CH* and MM-13aa, the SigA levels were equivalent, with and without Mg²⁺ supplementation, when cells were normalized to each other using OD₆₀₀ values (Fig. S2A). As an additional control, we compared the dry weights of samples grown in MM-13aa with and without excess Mg²⁺ (Fig. S2B). The dry weights were indistinguishable in the two conditions, further supporting the conclusion that the optical density provides an accurate approximation of the cell mass. These results strongly support the conclusion that Mg²⁺ supplemented cells become shorter as a result of more frequent cell divisions, and, conversely, that cells divide less often under lower Mg²⁺ conditions.

A delay in septation could occur before or after the assembly of the divisome. To assess this, we grew wild-type harboring a GFP fusion to ZapA, an early-arriving cell division protein that colocalizes with FtsZ as part of the “Z-ring” (54, 55). Cells expressing P_xylA-GFP-zapA grew equivalently, with and without 10.0 mM MgCl₂ (Fig. 7A), and retained the Mg²⁺-responsive reduction in cell length (Fig. 7B and C). Epifluorescence microscopy was performed, and images were captured of membranes, DNA (nucleoids), and Z-rings from both conditions (Fig. 7B). Overlays of the micrographs were used to quantitate both the number of distinct nucleoids per cell (Fig. 7D) and the number of Z-rings per unit cell length (Fig. 7E).

FIG 7.

Mg²⁺ modulates the frequency of Z-ring assembly. WT cells harboring P_xyl-GFP-zapA (BTG186) were cultured in CH* with 2.0 mM xylose. 10.0 mM MgCl₂ was added as indicated. (A) Representative growth curves. (B) Scatterplots showing the distribution of cell lengths. The bars represent the mean cell lengths of 800 cells ± the SDs. (C) Representative micrographs following the staining of membranes and nucleoids with FM4-64 and DAPI, respectively. The images are scaled identically. The arrowheads indicate examples of cells with four nucleoid masses and Z-rings that lack septa (yellow), 1 nucleoid mass (white), 2 nucleoid masses (blue), or partial septa (green). (D) Fraction of cells with the indicated number of nucleoids from three independent biological replicates. (E) Average number of Z-rings per unit cell length ± SD. Each circle represents the mean of 300 cells from three independent biological replicates per condition. (F) Average fraction of cells with coalesced ZapA-GFP that presented the indicated septum type ± the SD. Each circle represents the mean of 300 cells from three independent biological replicates per condition. *, 0.01 < P ≤ 0.05, **, 0.001 < P ≤ 0.01; (ns) P ≥ 0.05.

Regardless of the condition, both populations possessed a large proportion of cells with two nucleoids (35% and 59% for the samples with and without supplemental Mg²⁺, respectively). The most striking differences became apparent when comparing the proportions of cells with one or four nucleoids. While only 2% of the cells grown in CH* had one distinct nucleoid mass, this proportion increased 30-fold (to 65%) in the Mg²⁺-supplemented medium. Conversely, cells with four nucleoids were relatively frequent (38%) in the base medium but were rare (<1%) in the Mg²⁺-supplemented cultures (Fig. 7D). These results indicate that division is more frequent when Mg²⁺ is in excess. Consistent with this idea, Z-rings were more frequently observed along the lengths of Mg²⁺-supplemented cells, and fewer Z-rings were observed along the lengths of cells grown in unsupplemented CH* (Fig. 7E). Moreover, the Z-rings that did form were less likely to be associated with a partial (incomplete) septum (Fig. 7F). Similar results were observed when FtsZ was tracked directly with an FtsZ-GFP fusion (Fig. S3). We conclude that there is a window of Mg²⁺ availability in which the cell division frequency can be modulated before the growth rate is impacted.

Overexpressing undecaprenyl pyrophosphate synthetase (UppS) results in the loss of the Mg²⁺-responsive phenotype and in constitutively short cells.

In the CH* shifting experiment, more than half of the decrease in the average cell length occurred within 10 min of adding the Mg²⁺ (Tdoub = 33 min) (Fig. 6), suggesting that a majority of the cells were generally poised to divide but were somehow inhibited. In a 1969 study, Garrett showed that Mg²⁺, in the same ranges of interest as were examined in our study see Table 2 in reference [56]), caused B. subtilis (W23) to accumulate the cytoplasmic PG precursor UDP-MurNAc-pentapeptide (56). This result suggests that the next step in the PG synthesis pathway, namely, the generation of Lipid I via the conjugation of the MurNAc-pentapeptide to Und-P (Fig. S4A), is sensitive to Mg²⁺ availability.

Und-P is generated through the dephosphorylation of Und-PP, the latter of which is either synthesized de novo by UppS or is regenerated following the transfer of Und-P-linked precursors to acceptor molecules (Fig. S4A). To test whether increasing the Und-PP pools from the de novo synthesis pathway would promote more frequent division, we overexpressed uppS from an IPTG-inducible promoter. uppS overexpression resulted in shorter cells, compared to the uninduced control in CH*, and the overexpressing cells were equivalently short, irrespective of Mg²⁺ supplementation (Fig. 8A and B). In contrast, overexpressing mraY, which encodes the enzyme that reversibly transfers phospho-MurNAc-pentapeptide to Und-P (Fig. S4A) (57), did not shorten the cells, and the cells remained responsive to Mg²⁺ (Fig. S4B and 4C). In contrast, cells overexpressing bcrC, which encodes the major Und-PP phosphatase (58–60) (Fig. S4A), were 37% shorter in CH*, compared to the uninduced control (Fig. S4C and D), consistent with enhanced division. However, unlike uppS, bcrC-overexpressing cells were still able to decrease in length following the addition of Mg²⁺ (Fig. S4D). The growth rate of the cells overexpressing uppS was slower, but it was equivalent with and without Mg²⁺ supplementation (Fig. 8C), suggesting that the cell length decrease is not simply attributable to the lower growth rate. At the same time, we cannot exclude the possibility that slow growth reduces the cells to a length that makes differences too subtle to be measured with our assay. Together, these results suggest that the availability of Und-P is rate-limiting for cell division in the unsupplemented medium.

FIG 8.

UppS overexpression results in constitutively short cells. WT cells harboring P_hy-uppS (BTG708) were cultured in CH* with MgCl₂ (10.0 mM) and IPTG (0.5 mM) added as indicated. (A) Representative micrographs of cells stained with TMA and scaled identically. (B) Scatterplots showing the distribution of cell lengths quantitated for 300 cells from each condition. The bars represent the mean cell length ± the SD. (C) Representative growth curves. ****, P < 0.0001; (ns) P > 0.05.

RNA-seq analysis of cells grown with and without excess Mg²⁺.

To further explore the Mg²⁺-responsive phenotype, we used an RNA-seq analysis to determine whether there were relative changes in transcription between the short and long cells that could provide insight into the mechanism. RNA-seq was performed using samples collected from exponential-phase cells that were growing in CH* or in CH* supplemented with 10.0 mM MgCl₂. Genes expressed either more or less in the Mg²⁺-supplemented CH*, relative to CH* alone, were categorized based on known regulatory information that was retrieved from SubtiWiki (Tables S1 and S2) (61, 62).

As expected, based on the riboswitch data (Fig. 5), the transcriptional changes related to Mg²⁺ homeostasis were not significant. To our surprise, a number of other regulons that are related to metal homeostasis showed significant and internally consistent relative shifts in transcription. In particular, the profiles indicated the upregulation of the genes repressed by Fur (Fe²⁺ acquisition, sequestration, and sparing), MntR (Mn²⁺ uptake), and Zur (Zn²⁺ acquisition). In addition, there was enhanced expression of a large number of genes, consistent with a response to enhanced general stress (SigB). A subset of genes from two different prophages (PBSX and SPβ) were also expressed at higher levels but not in a manner consistent with prophage induction (Table S1). Conversely, and consistent with the observed upregulated genes, we observed the downregulation of the CzrA repressed genes cadA, czcD, and trkA (Zn²⁺ efflux/resistance to toxic metals), mneP (MntR-activated Mn²⁺ efflux), a large number of genes under the control of various cell envelope stress sigma factors (SigM/X/W/V), and those activated by the two-component response regulator YvrHb (63). Notably, several regulatory genes showed decreased transcription, including rsiX (encoding a SigX anti-sigma factor), sigX, and abh (a regulator of the transition state). Although we screened strains deleted for genes that showed differential expression via RNA-seq, those that were part of an annotated two-component system, and a strain that was cured of all prophages, each of the mutants retained the Mg²⁺-responsive phenotype (Tables S1–S3). Thus, although metal homeostasis and general stress response genes are clearly impacted by the changes in Mg²⁺ availability, we currently lack data that link any of these changes to enhanced division.

DISCUSSION

The findings in this study surprised us in several ways. First, the discovery of the cell division phenotype was itself unanticipated. Although LB is known to be low in Mg²⁺ and is generally discouraged by cell physiologists for various reasons (nicely outlined in a Small Things Considered blogpost by Hiroshi Nikaido [64]), LB has also been the default medium for the routine propagation of B. subtilis and E. coli for decades - not necessarily a context in which one expects to observe a novel phenotype. We cannot definitively say whether or not the phenotype observed in LB (Fig. 1) was due to the unusually low level of Mg²⁺ in the batch, as the powder was exhausted, and its bottle and lot number were discarded long before we appreciated the significance of the phenotype. Nonetheless, the data strongly implicated Mg²⁺ and motivated further experiments, which ultimately demonstrated that reduced Mg²⁺ availability impacts division before the growth rate and cell elongation.

The finding itself makes sense intuitively, as cell division is not a prerequisite for the replication of genetic material, growth, or survival. In fact, many bacteria are known to subsist as filamentous forms that divide infrequently. An extreme version of the filamentous lifestyle was recently discovered in the marine organism Thiomargarita magnifica, which displays average lengths of nearly a centimeter (65). This remarkable bacterium was found to harbor approximately 37,000 copies of its genome per millimeter of cell length - evidence of a dramatic uncoupling between growth and cell division. It is notable that many of the longest bacteria ever recorded, including the giant bacterium Beggiatoa that was discovered by Winogradsky, are sulfur-oxidizers (66). The evidence for an intimate relationship between sulfur metabolism and cell division is not restricted to giant bacteria. E. coli mutants with a reduced ability to convert ATP and the sulfur-containing amino acid methionine into S-adenosyl methionine (SAM) are inhibited for cell division but continue to grow and replicate DNA (67). The capacity of cells to uncouple division from other biosynthetic processes makes sense from an evolutionary perspective; it could provide cells with a mechanism by which to continue generating new “units” as resources deplete, while at the same time reserving the option to rapidly separate into individual cells when conditions are favorable.

Additional surprising results were the degree and type of the transcriptional changes that occurred between the supplemented and unsupplemented growth conditions. We expected that if changes occurred at all, then they would be associated with envelope synthesis or Mg²⁺ uptake or efflux. Instead, we observed changes in genes related to the general stress response and to iron, Mn²⁺, and Zn²⁺ homeostasis. More specifically, we found that when Mg²⁺ is more available, the expression of the SigB regulon is enhanced, and the Fur, MntR, and Zur regulons undergo shifts consistent with acquisition, sequestration, and/or sparing responses.

At first, the metal responses were perplexing, as we are accustomed to thinking about most metal regulons in terms of responses to starvation or toxicity, conditions that were not present in our experiments. However, the responses are relative, not absolute, and we can think of several other ways to interpret the data. The first is to think of the changes as homeostatic adjustments. For example, the data indicate that in the higher Mg²⁺ condition, the cells receive signals consistent with the need to acquire, sequester, and spare iron, as well as to increase internal Mn²⁺ and Zn²⁺. The expression of genes for surfactin biosynthesis (srfABCD) and the SigB general stress response are also increased. Iron sequestration and iron-sparing responses would reduce the possibility of Fenton reactions (68). Mn²⁺ and Zn²⁺ both reduce internal reactive oxygen species (69–75), and surfactin and SigB both combat oxidative/energy stress, the former by reducing proton motive force (76, 77). The overall profile suggests that, compared to the unsupplemented cells, the Mg²⁺-supplemented cells exhibit changes consistent with the prevention of oxidative stress.

The second possibility, which is not mutually exclusive, is that at least some of the transcriptional changes are associated with events before, after, and/or during cell division itself. Cell cycle variations are typically obscured when samples are collected from pooled populations of asynchronously growing cells. The samples collected for RNA-seq were also growing asynchronously; however, because the Mg²⁺-supplemented populations were essentially “enriched” for dividing cells (Fig. 7; Fig. S3), the transcriptional changes associated with division should also be enriched. The idea that cells may adjust metal pools not only to respond to different environmental stressors but also as a regulated part of the life cycle or development is intriguing, and we think it merits further inquiry.

We noted one anomaly in the RNA-seq data that we do not understand. In the Zur regulon, we see more RNA corresponding to yczL but relatively little or no increase in other genes found in the predicted (78) folEB-yciB-yczL-zagA transcript (Table S1). The unexpected differential expression of the yczL region could be attributable to differences in RNA stability, rather than transcription itself. The gene upstream of yczL (yciB) is proceeded by four instances of putative cotranslational coupling, the last of which could lead to the translation of yczL. It may be worth exploring whether there is a functional consequence to the predicted coupling, such as the alteration of transcriptional readthrough, transcript stability, or the effects on the expression of the downstream gene zagA.

We do not know whether any of the transcriptional changes relate directly to the Mg²⁺-responsive phenotype or are incidental. We are uncertain whether differential expression would still occur if cells were grown at maximal “shortness” but different Mg²⁺ concentrations, such as the 4.0 mM to 25.0 mM range shown in Fig. 5. The cell shortening phenotype was still detected in a prophage-cured strain (Table S3). So, at the least, the prophage-related changes appear to be incidental. Consistent with this conclusion, others have documented differential expression from various B. subtilis prophage loci as having resulted from variation the in growth regime (79).

Visually screening a number of deletion strains also failed to implicate any single gene knockouts in the Mg²⁺-responsive phenotype (Tables S1–S3). The only condition identified that showed an apparent loss of Mg²⁺-responsiveness in CH* was the overexpression of uppS (Fig. 8). However, as these cells grew slowly, we could not confidently exclude the possibilities that (i) cell shortening occurred but fell outside the detection limit of our assay and/or (ii) that the low growth rate imposes an upper limit on the division frequency that is dominant over the effects of Mg²⁺.

Still, it is plausible that exogenous Mg²⁺ somehow enhances the availability of Und-P to the divisome. UppS utilizes Mg²⁺ as a cofactor (80–82), so we considered the possibility that UppS activity could be affected by altering the availability of Mg²⁺. Although we do not exclude this possibility, several observations argue against it. First, E. coli UppS achieves maximal turnover (k_cat) around 0.5 mM MgCl₂ and half-maximal turnover at 49.0 μM MgCl₂ (80). The levels of labile, intracellular Mg²⁺ in B. subtilis (bulk measurements) are estimated to be between 0.8 and 3.7 mM in media containing 1.6 mM MgCl₂ (83). Notably, we still observe cell shortening at this concentration (Fig. 5). Thus, assuming that the B. subtilis and E. coli enzymes possess similar properties, there is likely sufficient Mg²⁺ to support full UppS activity when Mg²⁺ is not limiting the growth rate. Second, based on in vitro transcription termination measurements, the P_mgtE riboswitch transitions from an anti-terminator conformation to a terminator conformation between 2.0 and 3.0 mM Mg²⁺ (51). We did not see significant induction of the P_mgtE-lacZ reporter at the concentrations of Mg²⁺ at which the most dramatic changes in cell length were observed (Fig. 5C). This suggests that the pool of intracellular Mg²⁺ does not vary dramatically at the exogenous ranges that we focused on. Moreover, this is consistent with the observations made by others that intracellular Mg²⁺ does not vary significantly during exponential growth (83). Third, cell elongation also requires Und-P, and elongation was not affected in our conditions. Of course, it is possible that division is more sensitive to Und-P availability. At least in E. coli, there is some evidence that cells have a propensity to filament before bulging when Und-P is limiting (84). Finally, the rapid onset of division that we observed in the shifting experiment suggests that the cells have the capacity to divide without significant biosynthesis. Capturing these initial divisions in real time with a flow cell would provide more control and resolution regarding their timing.

One idea that is purely speculative is that the rapid division onset is driven by the more frequent flipping of Und-P to the cytoplasmic face of the membrane and/or the enhanced conversion of Und-PP to Und-P. There is also a pool of undecaprenyl-OH in some Gram-positives (85–88), so increasing Und-P availability through undecaprenyl-OH phosphorylation may be another route to enhancing pools. If the mechanism is not dependent on changes in cytoplasmic Mg²⁺ or on the activity of cytoplasmic enzymes such as UppS, there are also extracytoplasmic enzymes to consider. For example, the enzymes for wall teichoic acid and PG synthesis compete for Und-P, and there is at least one report that wall teichoic acid synthesis is favored over PG synthesis when Mg²⁺ is limited (89).

In summary, we show that the frequency of cell division can be uncoupled from the growth rate by manipulating a single nutrient, namely, Mg²⁺. Though we did not arrive at a mechanism, we hope that the observations documented here will inspire further studies in additional bacteria as well as more consideration of medium selection during experimentation. Rich media are convenient but have the considerable weakness of variable formulation with regard to cofactor, trace metal, and amino acid abundance. Not only can this lead to confusing and irreproducible results but also (as we unintentionally discovered), rich media may be masking some important biology.

MATERIALS AND METHODS

General methods.

The strains and the details of the strain construction can be found in the supplemental material (Table S4; Text S1). Cells were stored at −80°C in 15% glycerol (vol/vol). Strains were streaked for isolation on lysogeny broth, Lennox (LB) containing 1.5% (wt/vol) bactoagar (Sigma) and were incubated overnight (approximately 16 h) at 37°C. Cultures for experiments were begun with colonies from same-day plates by inoculating single colonies into a 20 mm glass tube that contained 5 mL of the indicated medium. The tube was incubated at 37°C in a roller drum until the exponential stage (OD₆₀₀ < 0.6). For the microscopy and RNA-seq experiments, the exponential-stage cultures (see above) were back-diluted in 25 mL of fresh medium in a 250 mL baffled flask to an optical density that would allow the cells to reach the desired growth stage (exponential) after no fewer than four doublings. The flasks were placed in a shaking water bath that was set to 280 rpm and 37°C.

The LB was made by dissolving 20 g of Difco LB Broth, Lennox (product number 240230) in 1 L ddH₂O, and this was followed by sterilization in an autoclave. The CH* utilized is a modified form of CH (49). The CH* (1L) contained 10.0 g casein acid hydrolysate (Acumedia, Lot No. 104,442B), 3.7 g l-glutamic acid (25.0 mM), 1.6 g l-asparagine monohydrate (10.0 mM), 1.25 g l-alanine (14.0 mM), 1.36 g (10.0 mM) anhydrous KH₂PO₄, 1.34 g (25.0 mM) NH₄Cl, 0.11 g Na₂SO₄ (0.77 mM), 0.1 g NH₄NO₃ (1.25 mM), 0.001 g FeCl₃•6H₂O (3.7 μM), and ddH₂O. The pH was adjusted to 7.0 with 10.0 N NaOH before the medium was sterilized in an autoclave. After autoclaving, sterile CaCl₂ was added to 0.2 mM, and l-sterile tryptophan was added to 0.1 mM. The base minimal medium (MM) contained 15.0 mM (NH₄)₂SO₄, 5.0 mM KH₂PO₄, 50.0 mM Tris-HCl [pH 7.5], 27.0 mM KCl, 0.05 mM FeCl₃, 0.01 mM MnSO₂, 0.01 mM ZnSO₄, 0.2% glucose (wt/vol), 27.0 mM sodium citrate, and 10.2 mM CaCl₂. 50.0 μM MgCl₂ was added to create the base medium used in the MM experiments, except for those in Fig. 5, which required final concentrations below 50.0 μM. In these experiments, MgCl₂ was added to the final concentrations indicated in the figure. For the MM-13aa, 10× filter-sterilized amino acids [pH 7.5] were added to a 1× final concentration. The 10× mixture consisted of the following concentrations of l-amino acids: 250.0 mM glutamate; 100.0 mM (each) alanine, asparagine, and proline; 50.0 mM (each) aspartic acid, phenylalanine, glycine, isoleucine, leucine, threonine, and valine; 10.0 mM (each) histidine and tryptophan. The amino acids were added to increase the growth rate but were selected based on their availability in our laboratory, not on logic.

β-galactosidase assays.

The B. subtilis strains were grown in 250 mL baffled flasks with 25 mL of the indicated medium at 37°C and 280 rpm to the target OD₆₀₀. The optical density was recorded, and 1 mL of culture was harvested via centrifugation at 21,130 × g for 1 min at room temperature, removing the supernatant via aspiration. The pellets were immediately stored at −80°C. To perform the assay, the pellets were thawed, resuspended in 0.5 mL Z-buffer (40.0 mM NaH₂PO₄, 60.0 mM Na₂HPO₄, 1.0 mM MgSO₄, 10.0 mM KCl, 40.0 mM β-mercaptoethanol, and 0.2 mg/mL lysozyme), and incubated at 30°C for 15 min. 100 μL of 4.0 mg/mL 2-nitrophenyl-β-d-galactopyranoside (ONPG) in Z-buffer was added, and the samples were incubated at 30°C until pale yellow. The reaction was stopped with 250 μL 1.0 M Na₂CO₃, and the reaction time was recorded. The sample was vortexed for 5 sec and centrifuged at 21,130 × g for 3 min at room temperature in a tabletop centrifuge. The supernatant (minimum volume of 0.8 mL) was transferred to a 1 mL cuvette, and the OD₄₂₀ and OD₅₅₀ absorbance readings were recorded. The β-galactosidase specific activity in Miller Units was calculated using the following formula: (OD₄₂₀ − [1.75× OD₅₅₀]) / (time [min] × volume × OD₆₀₀) × 1,000.

Western blot analysis.

1 mL of culture was collected and spun at 21,130 × g for 1 min at room temperature at the exponential stage, and the OD₆₀₀ value at the time of sampling was recorded. The pellet was resuspended in the lysis buffer (20.0 mM Tris [pH 7.5], 10.0 mM EDTA, 1 mg/mL lysozyme, 10 μg/mL DNase I, 100 μg/mL RNase A, 1.0 mM PMSF, 1 μL protease inhibitor cocktail [Sigma P8465-5ML] resuspended in 1 mL lysis buffer) to give a final OD₆₀₀ equivalent of 15. The samples were incubated at 37°C for 10 min, and this was followed by the addition of an equal volume of sodium dodecyl sulfate (SDS) sample buffer (0.25 M Tris [pH 6.8], 4% [wt/vol] SDS, 20% [wt/vol], 20% glycerol [vol/vol], 10.0 mM EDTA, and 10% β-mercaptoethanol [vol/vol]). The samples were heated for 5 min at 100°C prior to loading. The proteins were separated on 12% SDS-PAGE polyacrylamide gels, transferred to a nitrocellulose membrane at 100 V for 60 min, and then blocked in 1× PBS containing 0.05% (vol/vol) Tween 20 and 5% (wt/vol) dry milk powder. The blocked membranes were probed overnight at 4°C with anti-SigA (1:20,000, rabbit, gift from Fujita Masaya, University of Houston, Houston, TX) diluted in 1× PBS with 0.05% (vol/vol) Tween 20 and 5% (wt/vol) milk powder. The membranes were washed three times with 1× PBS containing 0.05% (vol/vol) Tween 20, transferred to 1× PBS with 0.05% (vol/vol) Tween 20 and 5% (wt/vol) milk powder containing 1:5,000 horseradish peroxidase-conjugated goat anti-mouse IgG secondary antibodies (AbCam, ab205719), and incubated on a shaking platform for 1 h at room temperature. The membranes were washed 3× with 1× PBS containing 0.05% (vol/vol) Tween 20, and signals were detected using the SuperSignal West Femto Maximum Sensitivity Substrate (Thermo Fisher) and a Bio-Rad Gel Doc Imaging System.

Fluorescence microscopy.

1 mL of cultured cells at the exponential stage were harvested, concentrated via centrifugation at 6,010 × g for 1 min, and resuspended in 5 μL 1× PBS with either 1-(4-(trimethylamino)phenyl)-6-phenylhexa-1,3,5-triene (TMA-DPH) (50.0 μM) or N-(3-triethylammoniumpropyl)-4-(6-(4-(diethylamino) phenyl) hexatrienyl) pyridinium dibromide (FM4-64) (6 μg/mL) and 4′,6-diamidino-2-phenylindole (DAPI) (2 μg/mL). All dyes were purchased from Thermo Fisher. For the Zap-GFP and FtsZ-GFP experiments, the cells were mounted on 1% (wt/vol) agarose pads made with PBS [pH 7.4] and overlaid with an untreated glass coverslip. Otherwise, the cells were mounted on glass sides with poly-l-lysine-treated coverslips prior to imaging.

The cells were imaged on a Nikon Ti-E microscope using a CFI Plan Apo Lambda DM 100× objective and an X-Cite XYLIS 365 nm illumination system (Excelitas Technologies). The filter cubes utilized were C-FL UV-2E/C (DAPI), C-FL Texas Red HC HISN Zero Shift, and GFP HC HISN Zero Shift. Micrographs were acquired using a CoolSNAP HQ2 monochrome camera with NIS Elements Advanced Research, and they were analyzed in ImageJ (90).

Cell length measurements.

The TMA-DPH micrographs were analyzed in ImageJ (90). For each cell, a line that extended from pole to pole was placed. One compartment between two bright, solid septa with homogenous signals were counted as one intact cell. Cells with bright partial septa or septa interrupted with discrete dark areas were counted as cells with partial septa.

Quantitation of Z-rings per unit cell length.

The cell length lines (see above) were overlapped with the GFP micrographs of the Z-rings, and the number of Z-rings along the cell length line was counted. When two cells shared a pole and a Z-ring, each Z-ring was counted as 0.5/cell. The Z-ring number per μm was calculated by dividing the Z-ring total number by the total cell length.

Nucleoid number per cell.

The cell length line was overlapped with the DAPI micrograph, and the cells were classified based on the number of nucleoids overlapping the length label. The number of nucleoids per cell was calculated by dividing the number of cells containing each number of nucleoids (1, 2, or 4) by the total number of cells counted.

Classification of septa in cells containing coalesced Z-rings.

Micrographs of the GFP channel (ZapA-GFP or FtsZ-GFP) were overlaid with the corresponding TMA-DPH micrograph. The cells were classified as having no, partial, or full septa. The fraction of each septum type was determined by dividing the number of cells in each class by the total number of cells counted.

Statistical analysis and data plotting.

The graphs were generated and the statistical analysis was performed using GraphPad Prism version 9.4.0 for Mac (GraphPad Software, San Diego, California, USA, www.graphpad.com). The statistical analyses in the following figures were performed via one-way analyses of variance (ANOVA) followed by Tukey’s multiple-comparison tests, assuming Gaussian distributions with equal standard deviations (SDs): Fig. 3B, 5E, 6A, 6B, and 8B as well as Fig. S2B and S3C. The statistical analyses in the following figures were performed via two-way ANOVA followed by Tukey’s multiple-comparison tests, fitting a full model: Fig. 7F and Fig. S3D. The statistical analyses in the following figures were performed via two-tailed unpaired t tests, assuming a Gaussian distribution and that both populations have the same SD: Fig. 1B, 2D, 6B, and 7E.

RNA-seq.

For each condition, samples were collected from three independent biological replicates grown in either CH* medium or CH* medium with 10.0 mM MgCl₂. The cells were precultured in the same medium used for the experiment. 0.5 mL cells were collected at the exponential stage (OD₆₀₀ ~of approximately 0.2) via centrifugation at 21,130 × g for 2 min at room temperature. RNA was collected using the RNAprotect Bacteria Reagent and an RNeasy Mini Kit (Qiagen) according to the manufacturer’s instructions. RNA sequencing was performed by the Texas A&M AgriLife Research Genomics and Bioinformatics Service (College Station). The total RNA were prepared for sequencing using TruSeq Stranded Total RNA and Ribo-Zero Gold (Illumina). The RNA-seq was performed using a HISeq 4000 platform (Illumina). The RNA-seq data were processed through HISAT2 (91) and StringTie (92). A differential expression analysis was done using the Deseq2 R package (93). Regulons were assigned using the GinTool (62) starter package obtained from L.W. Hamoen (University of Amsterdam, Swammerdam Institute for Life Sciences). The cutoff for inclusion in the final analysis was set as genes with an absolute log₂-fold change (log₂FC) value of ≥0.585 and an adjusted P-value of ≤0.05. The adjusted P-values that are reported as 1.0 in the raw data indicate that the gene could not be included or excluded as statistically significant (generally due to insufficient reads in either the control or experimental samples).

Data availability.

The raw RNA-seq data may be accessed through the NCBI Gene Expression Omnibus at the following link: https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE219221.