Absolute Measurements of mRNA Translation in Caulobacter crescentus Reveal Important Fitness Costs of Vitamin B12 Scavenging

Caulobacter crescentus is a model system of the bacterial cell cycle culminating in asymmetric cell division, with each daughter cell inheriting a distinct set of proteins. While a genetic network of master transcription factors coordinates the cell cycle timing of transcription for nearly 20% of Caulobacter genes, we lack knowledge of how many of each protein “part” encoded in the genome are synthesized. Therefore, to determine the absolute production rates across the genome, we performed ribosome profiling, providing, for the first time, a quantitative resource with measurements of each protein “part” needed to generate daughter cells. This resource furthers the goal of a systems-level understanding of the genetic network controlling asymmetric cell division. To highlight the utility of this data set, we probe the protein synthesis cost of a B12 utilization pathway and provide new insights into Caulobacter’s adaptation to its natural environments.

I n bacterial systems biology, global mRNA translation measurements are critical for understanding how cells utilize their resources to achieve their evolutionarily selected cell growth and division cycles. To complete the bacterial cell cycle, the protein parts encoded within the genome must be transcribed into mRNAs that are translated into the appropriate number of proteins for the daughter cells to be generated. Genomewide absolute quantitation of protein level measurements has allowed the monitoring of protein resource allocation (1,2), revealing that these cells allocate resources for optimal growth. As the ribosome content is positively correlated with the growth rate (3), cells must optimize the fraction of protein synthesis needed to make new ribosomes (enzymes that make proteins) versus the protein synthesis needed to produce the proteomes of the daughter cells to achieve short generation times (2,4). Optimality has also been observed at the protein-complex level, as translation of a stoichiometric amount of protein subunits to the overall multiprotein complexes has been observed (2), with different posttranscriptional strategies across species utilized to achieve the optimal protein concentration (5). Therefore, to understand the mechanisms controlling the growth and division cycles of diverse bacteria, we must understand how bacteria are able to optimize their protein synthesis resources for maximal fitness.
Caulobacter crescentus is an oligotrophic alphaproteobacterium with a carefully orchestrated cell cycle yielding asymmetric cell division and a model organism for the study of the bacterial cell cycle (6,7). In C. crescentus, cells undergo changes in gene expression of ϳ20% of their entire genome during the process of the cell cycle (8,9). Timing of 57% of the cell cycle-regulated mRNAs is controlled at the transcription level by a master regulatory circuit that is composed of 4 transcription factors (DnaA, GcrA, CtrA, and SciP) and a DNA methylase (CcrM) (6,10), and 49% of those cell cycleregulated mRNAs are additionally regulated at the level of mRNA translation (8). Importantly, global C. crescentus studies have focused solely on the control of the timing of gene expression in the cell cycle, and thus, little is known about the absolute levels of protein synthesis, or how the protein synthesis resources are allocated across the proteome.
Here, we utilize ribosome profiling to achieve a quantitative genome-wide absolute measure of protein synthesis in C. crescentus. This resource provides the absolute protein synthesis rate of each protein expressed from the C. crescentus genome and a global map of protein synthesis resource allocation. Absolute levels of mRNA translation of cell cycle master regulators showed higher levels of mRNA translation compared to their known DNA binding sites for all but CcrM and a relatively low level of mRNA translation of CtrA regulatory proteins relative to the CtrA master regulator itself. PopZ, a polar protein scaffold that recruits asymmetric cell fate specification proteins (11), is at a limiting concentration compared to its client proteins, suggesting that these clients compete for access to the cell pole. Surprisingly, we discovered that the btuB vitamin B 12 importer and the metE methionine-biosynthetic gene were among the most highly translated genes in the absence of B 12 , showing that the C. crescentus B 12 -scavenging pathway requires a surprisingly large amount of the cell's protein synthesis resources. The high cost of protein synthesis of the B 12 -scavenging pathway is reduced in the presence of B 12 by riboswitches in the 5= untranslated region (UTR) of these two genes. The widely utilized lab strain NA1000 is a facultative B 12 scavenger due to the metE gene, which produces methionine in the absence of B 12 , yet many natural Caulobacter isolates are obligate B 12 scavengers (12). We show that the facultative B 12 -scavenging lifestyle generates a fitness tradeoff, where in the absence of B 12 there is a positive fitness advantage from MetE's B 12 -independent methionine production, while in B 12 's presence there is a fitness disadvantage due to the wasted cost of MetE's protein synthesis, providing an explanation for why many isolates have lost the metE gene to become obligates for B 12 .

RESULTS
Absolute quantitation of mRNA translation rates. Ribosome profiling provides a global direct measure of the protein synthesis rate by sequencing ribosome-protected mRNA footprints (2,13,14). To determine absolute rates of translation in C. crescentus, ribosome profiling was performed in unsynchronized C. crescentus NA1000 cells grown in M2G minimal medium. In fast-growing bacteria where the rate of translation is the main driving force of protein levels, protein degradation can be negligible, and therefore, the main driving force of protein levels is mRNA translation (2). This is largely true in C. crescentus, as Ͼ95% of proteins were found to have half-lives longer than the cell cycle (15). First, we examined the ribosome footprint density along each open reading frame (ORF) on mRNAs as a relative measure for translation (Fig. 1A). For example, in the divK/pleD polycistronic operon, we find that divK has 2.0-fold-higher ribosome density than pleD (Fig. 1A). For absolute quantitation, it is assumed that the average elongation rate is constant for each mRNA, which would allow the average ribosome density to be directly proportional to the rate of protein synthesis of each ORF (2). It is also assumed that all ribosomes will finish translation and make the full-length protein (2). Next, to reduce the impact of fast-and slow-moving ribosomes in the ribosome occupancy profiles along ORFs on the quantitative level of translation, we used winsorization to correct the average ribosome footprint density of each ORF (see Table S1 in the supplemental material). Start codon and stop codon regions were omitted from the analysis to avoid biases in slow-moving ribosomes that are initiating or terminating (13,16).
To convert ribosome density to absolute mRNA translation rates, we measured the average protein mass of C. crescentus cells, which was multiplied by the fractional ribosome density measurement of each gene and divided by the molecular weight (k i ϭ i P/mW i ) (2). This measure of the average number of proteins translated per cell correlated well between the protein concentrations reported in the literature as well as protein concentration measurements reported here using yellow fluorescent protein (YFP) intensity of C-terminally tagged gene fusions ( Fig. 1B; Fig. S1 and Table S2). As expected, FtsZ, the key cell division protein which is known to be a substrate of cell cycle-dependent proteolysis (17), has a 13-fold-larger amount of translated protein than protein observed in the cell, while stable proteins ranged between 0.65-and 2.3-fold. The same phenomenon of higher translation levels than protein levels was also observed for the proteolyzed cell cycle regulators DnaA and CcrM, and upon deletion of the Lon protease, which is known to be responsible for their proteolysis, the indicate the standard deviation in YFP intensity or standard deviation for the Western blots, while horizontal error bars indicate the standard deviation from ribosome profiling replicates (n ϭ 3). FtsZ is expected to deviate from the line since its protein levels are under proteolytic control (17). Data are in Tables S1 and S2.
Absolute Quantitation of C. crescentus mRNA Translation correlation was restored (Peter Chien, personal communication) (Fig. S1) (18,19). These data suggest that the absolute measures of mRNA translation are reflecting the absolute protein synthesis rate for each ORF in C. crescentus and provide a reasonable measure of steady-state protein levels for stable proteins (data can be found in Table S1).
Global analysis of C. crescentus absolute mRNA translation levels. C. crescentus cells dedicate a significant percentage of their protein synthesis to several major cellular processes associated with cell growth ( Fig. 2A; Fig. S2). Across major KEGG categories, we analyzed the percentage of ribosome footprints to understand their allocation of protein synthesis capacity (Table S3). Nutrient transporters (11.8%), the ribosome (11.2%), and the cell envelope (9.8%) represent the largest classes of protein production for C. crescentus. A significant fraction of protein synthesis capacity (25.5%) is allocated to produce proteins of unknown function, showing that a significant fraction of the cell's protein synthesis capacity is not understood. By comparing the fraction of the translation KEGG category across minimal medium (M2G) (15.9%) and a richer complex medium (peptone-yeast extract [PYE]) (22.8%), we find that the cells dedicate a larger amount of protein synthesis capacity to making translational machinery in rich medium, similar to Escherichia coli (3). Additionally, we see that a larger amount of protein synthesis capacity in "cell growth and death" is observed in M2G (11.3%) than in PYE (6.45%), owing largely to increased protein synthesis capacity of an operon of cell-contact-dependent toxins and immunity proteins that are known to be expressed in stationary phase (Fig. S2) (20).
Interestingly, while the cell cycle is a major area of study in C. crescentus, the cell cycle genes represent only a small fraction of protein synthesis capacity (1.78%), and yet these genes play a critical role in shaping the growth and division cycles. The cell cycle-regulatory circuit itself is composed of four transcription factors (DnaA, GcrA, CtrA, and SciP) and a DNA methylase (CcrM) whose spatiotemporal activation facilitates cell cycle progression (6). For master regulator proteins, the cell produces between ϳ3,000 and 30,000 copies of each protein, which corresponds to between 71-and 648-fold more proteins than the number of known DNA binding sites that they control (10), with the exception of CcrM (Table S4). Three thousand two hundred eighty CcrM proteins are translated to methylate the 4,542 GANTC sites per chromosome (Fig. 2B). While the number of CcrM proteins is approximately one-third the number of GANTC sites present after DNA replication, CcrM is a processive enzyme (21), suggesting that each CcrM may on average methylate ϳ3 GANTC sites. While the number of CtrA proteins translated (25,400 proteins) corresponds closely with the amount measured in predivisional cells (18,000 to 22,000 [22,23]), CtrA is produced at a significantly higher level than its collection of regulatory kinases, phosphotransferases, and proteolytic adapters that control its cell cycle-dependent activity (Fig. 2B). GcrA interacts with the RNA polymerase/ 70 complex to activate transcription of target promoters (24), where an ϳ4-fold excess of GcrA over 70 is produced, suggesting that excess GcrA may accelerate binding to the RNA polymerase holoenzyme (25) to facilitate subsequent recruitment of 70 .
As many proteins were found to have distinct subcellular patterns of protein accumulation in C. crescentus as determined in reference 26, we compared protein synthesis capacity to the localization patterns of proteins observed in this data set (Fig. 2C). Seven percent of protein synthesis occurs for "localized proteins" in C. crescentus. Of those localized proteins, most are "patchy/spotty," while a significant fraction has a subcellular address where the protein accumulates (pole, stalk, or center) (Fig. 2C). Many proteins are specifically required to form asymmetric polar protein complexes that function to determine cell fate upon division (6). Many of these polarly localized proteins are recruited to the cell pole through the multimeric hub protein PopZ (27)(28)(29). Interestingly, by examining PopZ and its known client proteins, we find that PopZ is made in limiting amounts (Fig. 2D), suggesting that the clients compete for PopZ binding in vivo.

Analysis of vitamin B 12 and methionine metabolism.
Analysis of the most highly translated proteins found that RsaA, the surface layer protein, was the most highly translated protein in the cell (Fig. 3A) (30). Elongation factor Tu was the third most abundant cytoplasmic protein owing to its requirement to deliver aminoacyl-tRNAs to the ribosome during translation (31). Surprisingly, we also observed that the homolog of the B 12 importer (btuB, second highest) and the methionine-biosynthetic gene (metE, fourth highest) were among the most highly translated proteins. Vitamin B 12 is an important enzymatic cofactor that in C. crescentus is used for the biosynthesis of methionine, deoxynucleoside triphosphate (dNTP) production, tRNA modification, and isomerization of methylmalonyl coenzyme A (CoA) to succinyl-CoA (Fig. 3) (32). C. crescentus cannot synthesize B 12 de novo but can import it through the BtuB protein (33). In the cytoplasm, both MetE and MetH perform the rate-limiting step of methionine biosynthesis, where MetH requires B 12 but has a higher specific activity than MetE (Fig. 3B) (34,35). Of note, both BtuB and MetE are translated at much higher levels than the other components related to methionine biosynthesis (Fig. 3B). Both the btuB and metE genes are the only two genes in C. crescentus with B 12 riboswitches encoded in their 5= UTRs (36). We tested the function of these riboswitches by creating 5= UTR fusions to the mCherry gene driven by the vanillate promoter and subjecting the cells to various concentrations of B 12 in the form of cyanocobalamin (Fig. 3C). Both the metE and btuB 5= UTR reporters showed high translation in the absence of B 12 and exhibited a B 12 concentration-dependent translational shutoff (Fig. 3C). The metE riboswitch appears to be more sensitive to B 12 concentration, with a K 1/2 of 0.062 nM, while the btuB riboswitch K 1/2 was 0.19 nM, both in line with the concentrations found in aquatic ecosystems (Fig. S3) (37,38). Taken together, these data show that C. crescentus cells are investing a large amount of their protein synthesis capacity toward B 12 uptake and the B 12 -independent methionine pathway in the absence of B 12 . We therefore hypothesized that the cells are wasting energy in the absence of B 12 by producing these very costly proteins.
To assess if the cells are wasting energy from the B 12 -related pathways, we examined the fitness of C. crescentus cells with disruptions in the nonessential components of these pathways in the absence of B 12 (Fig. 4A). For this, we used published transposon sequencing (Tn-seq) data sets (39) and analyzed B 12 -related genes whose disruption would not have polarity effects (single genes or last genes in operons) in M2G minimal medium or PYE rich medium, neither of which contains B 12 (12,40). In M2G minimal medium, cells require the metE gene to make methionine (33), while the other nonessential B 12 -related components showed increases in fitness when disrupted that were proportional to their protein synthesis costs (Fig. 4A). In PYE rich medium, which contains methionine in the peptone, the metE gene is no longer essential, but instead, its disruption leads to higher fitness (39). Indeed, all the components of the methionine pathway led to increases in fitness proportional to their protein synthesis cost when disrupted in PYE (Fig. 4A) (39). These data show that in the absence of B 12 , the excessive translation of these proteins leads to unnecessary costs of protein synthesis that limit the fitness of C. crescentus cells.
To test the effects of B 12 on C. crescentus cell growth, we examined the growth rate of cells cultured in M2G medium containing B 12 in the form of cyanocobalamin. Here, we observe faster growth in a B 12 concentration-dependent manner ( Fig. 4B; Table S5), with up to a 21% faster doubling time observed at 1 nM B 12 in M2G medium. Eleven percent acceleration of cell growth rates also occurs in PYE medium, which contains methionine in the peptone, suggesting that the growth enhancement of B 12 is likely caused in part by reduced protein synthesis costs from btuB and metE riboswitches. We attempted to separate B 12 's effects on methionine synthesis from its effects on other pathways by addition of exogenous methionine in the presence or absence of B 12 , but exogenous methionine leads to a dramatic decrease in growth by an unknown mechanism (41). Overall, we find that B 12 significantly enhances the growth rate of NA1000 cells.   (Table S1) on the x axis versus the Tn-seq-derived fitness values for the btuB B 12 importer and methionine-biosynthetic genes under growth in minimal or rich medium as measured previously (39) (biological replicates, n ϭ 2 for M2G, n ϭ 10 for PYE). Black points represent nonessential genes for methionine biosynthesis, and red points represent genes required for methionine biosynthesis under the specified growth condition. Error bars represent standard deviation. Curve fits were performed only on nonessential genes. (B) Doubling times of C. crescentus cells in M2G and PYE media with indicated concentrations of B 12 . Error bars represent the standard deviation for doubling time measurements (biological replicates, n ϭ 3 for each condition). (C) Tn-seq-derived fitness values for metE, btuB, and metH under growth in Lake Michigan lake water as measured previously (57). (D) Fitness tradeoff of facultative versus obligate B 12 scavenging. Relative fitness shown for species with metE (facultative) or without metE (obligate) in environments lacking or containing sufficient B 12 . (E) Phylogenetic tree of all Caulobacter species with completed genomes based on btuB and metH protein sequences. Each species is labeled by its NCBI accession identifier, and the scale represents the Kimura distance. Marks next to species represent the presence of a metE, btuB, or metH gene. All species have a predicted B 12 riboswitch upstream of btuB and metE genes (not shown), except the species noted with the yellow asterisk. A list of species names can be found in Table S7.

DISCUSSION
Absolute quantitation of protein synthesis in C. crescentus. As the cost of protein synthesis is a significant output of the cell's energy, absolute quantitation of translation is a powerful method to measure gene expression and resource allocation. Here, we present an absolute protein synthesis resource for C. crescentus generated by ribosome profiling which will be vital for systems modeling efforts for the C. crescentus cell cycle (42)(43)(44)(45) and in the subsequent optimization of synthetic Caulobacter genomes (46,47). Across the proteome, the mRNA translation resource allocation showed that 1.4% of the translation machinery is dedicated to translation of cell cycle-regulatory genes. We observed that for all the cell cycle master regulators, except CcrM, the number of proteins translated dramatically exceeds the total number of DNA binding sites in the genome (Fig. 2B). We hypothesize that the high concentrations of these factors may facilitate rapid activation of target gene transcription during each phase of the cell cycle.
Approximately 7% of protein synthesis capacity is dedicated to genes whose protein products were found to be localized in one of several modes of subcellular organization (Fig. 2C) (26). Recent reports suggest that some of these foci are formed by liquid-liquid phase separation of the proteins into membraneless organelles (48,49). For C. crescentus BR bodies, the concentration of the condensate-forming protein RNase E (6.3 M) appears to correspond closely to the transition boundary for liquid-liquid phase separation (48), potentially allowing control of the assembly of these bodies. Interestingly, we observed that the polar protein scaffold PopZ, which facilitates recruitment of asymmetrically localized signaling proteins to the cell poles (11,27,29), is present at approximately one-half the concentration of its client proteins, suggesting that clients compete for PopZ access (Fig. 2D). Dynamic competition of clients for PopZ may be important for the ordered assembly of unique proteins at each cell pole and may impact the spatial activation of downstream signaling outputs (27,50,51).
Implications of B 12 -scavenging pathway for environmental fitness. B 12 is an important enzymatic cofactor that is required for the activity of enzymes involved in biosynthesis of methionine, dNTP production, tRNA modification, and isomerization of methylmalonyl-CoA to succinyl-CoA ( Fig. 3B and D) (32). Like many bacteria, C. crescentus cannot produce B 12 but can scavenge it from the environment (33), where it can increase the growth rate of C. crescentus by up to 21% (Fig. 4B). btuB, the B 12 importer, and metE, the B 12 -independent methionine synthase, are among the most highly expressed genes, accounting for ϳ4.5% of all protein synthesis capacity (Fig. 3A). To counteract the protein synthesis demand, btuB and metE genes also contain B 12 riboswitches that reduce translation when sufficient B 12 enters the cytoplasm (Fig. 3C). Freshwater bodies typically have B 12 concentrations in the range from 0.11 nM to below the level of detection (Ͻ0.1 pM) (37,38,(52)(53)(54)(55), suggesting that high levels of BtuB may help facilitate import. Importantly, the conserved metE and btuB riboswitches are sensitive to B 12 in physiologically relevant ranges (K 1/2 ϭ 0.062 nM and 0.19 nM, respectively), and their different sensitivities suggest that metE translation would be downregulated before shutting off the btuB importer. At high concentrations of B 12 that saturate the riboswitches (Fig. 4B), a significant portion of the increase in growth rate is likely due to the liberation of protein synthesis resources on these two highly expressed genes ( Fig. 3A and C). When grown in M2G or PYE medium lacking B 12 , disrupting the btuB gene increases fitness by freeing up wasted protein synthesis resources (Fig. 4A) (39). Similarly, metE disruption leads to increased fitness in PYE, which contains methionine, but metE disruption becomes essential in M2G minimal medium as it is required to make methionine (Fig. 4A) (39). Why then does C. crescentus have a facultative B 12 lifestyle containing both B 12 -dependent and -independent methionine biosynthesis pathways?
Perhaps B 12 -independent and B 12 -dependent pathways exist to buffer fluctuations in environmental B 12 concentrations. The concentration of available B 12 in a freshwater body shows variation of up to 40-fold between different sampling locations and at the same sampling location at different times (38,(52)(53)(54)(55)(56). Having both methionine biosynthesis pathways adds flexibility to generate methionine under either high-or low-B 12 conditions; however, these pathways have different protein synthesis costs. The B 12 -independent pathway requires 1.57% of the total protein synthesis capacity to make sufficient MetE, while the B 12 -dependent MetH pathway uses only 0.156% (Fig. 3B). Interestingly, disrupting the metE gene in lake water leads to a fitness advantage; however, disrupting btuB or metH leads to a fitness decrease as determined in reference 57 (Fig. 4C). Although not measured directly, the gene fitness signature from the experiment leads us to infer that physiologically relevant B 12 concentrations were present in the sampled lake water (57). The increased fitness of metE disruptions in lake water suggests that increased biosynthetic flexibility comes with a negative fitness cost from protein synthesis resources wasted on MetE (Fig. 4D). Forty-seven percent of fully sequenced Caulobacter species have lost the metE gene but not the btuB and metH genes, suggesting that the observed environmental fluctuations in B 12 concentration (38,52) alter the selective pressure on metE (Fig. 4E).
Surprisingly, a recent survey of available metagenomic 16S rRNA sequencing data showed that Caulobacter is more abundant in soil/compost than in aquatic ecosystems (58). Soil has been shown to have B 12 levels that can range from 20 nM to 0.3 nM, correlated with levels of organic matter (59), while bodies of freshwater typically have B 12 concentrations in the range of 0.11 nM to below the level of detection (0.1 pM) (37,38,(53)(54)(55)(56). The higher B 12 concentration in soil will enhance the growth rate and may explain the increase in relative abundance in this environment.

Bacterial strains and cell growth.
A list of all bacterial strains used here can be found in Table S6 in the supplemental material (76)(77)(78)(79)(80)(81). C. crescentus cells were grown in M2G or PYE growth medium (40) and supplemented with the appropriate antibiotic concentrations (60). E. coli cells used for cloning were grown in LB medium and supplemented with the appropriate antibiotics.
Ribosome profiling. Ribosome profiling was performed similarly to procedures in references 8 and 36, except that contaminating rRNA fragments generated during micrococcal nuclease (MNase) digestion were depleted to allow deeper quantitation of resulting mRNA translation similarly to reference 2. For a detailed protocol for the procedure, see reference 16. Five hundred milliliters of NA1000 cells were grown in M2G medium to an OD 600 of 0.5, treated with 100 g/ml chloramphenicol for 2 min, and then harvested by centrifugation and flash-frozen in liquid nitrogen. Cells were then lysed on a mixer mill (Retsch mm400) for 6 cycles of 3 min at 15 Hz and thawed, membranes were pelleted, and the supernatant was footprinted by addition of MNase (Roche). After footprinting, MNase was quenched with EGTA, and samples were separated by sucrose gradient fractionation. 70S peaks were purified, phenol chloroform extracted, and ethanol precipitated (16). Resulting mRNA fragments were size selected by 10% acrylamide-1ϫ TBE-7 M urea PAGE, end repaired, 3= adapter ligated, reverse transcribed, circularized, and depleted of rRNA fragments (2,14,16). rRNA cDNA fragments were removed using biotin-linked DNA oligonucleotides (oCaulo1, 5=/5Biosg/CGCTTACGGGGCTATCACCCA; oCaulo2, 5=/5Biosg/TGGCAACT AATCACGAGGGTT; oCaulo3, 5=/5Biosg/CTCATCTGGTTGCCCAAAAGA; oCaulo4, 5=/5Biosg/TGGTTCAGGAA TATTCACCTG) and MyOne streptavidin C1 Dynabeads (Invitrogen) as in reference 2. Resulting circular cDNAs were amplified by PCR using Phusion DNA polymerase (Fermentas) with indexing primers (61), pooled, and sequenced on an Illumina HiSeq 2000. Data for three ribosome profiling replicates were deposited in the Gene Expression Omnibus under accession number GSE126485. The three M2G replicates were further analyzed together with a PYE data set collected previously (36).
Ribosome footprint reads were mapped to the genome as center-weighted reads (13), and extremely fast and slow codons were corrected for by winsorization of the bottom 5% and top 95% of nucleotides, respectively. The resulting fraction of ribosome footprints ( i ) of each gene (i) compared to the total ribosome footprint total was calculated and converted into the number of molecules of protein translated per cell (k i ) by the equation k i ϭ i P/mW i , where P is the average protein mass per cell and mW i is the protein product's molecular weight as originally described in reference 2. Average protein mass per cell (P) was measured as follows. Five-milliliter mid-log cultures of NA1000 cells were grown in M2G or PYE medium overnight in a roller wheel at 28°C. Once the cells reached an OD 600 of 0.3, a 100-l aliquot of the cells was diluted and counted on PYE plates to measure the number of viable cells and another aliquot was saved for protein concentration measurements. To measure protein content, 100 l cells was spun down in a microcentrifuge at 14,000 rpm for 30 s, the supernatant was removed, and the remaining cell pellets were resuspended in 100 l of 1ϫ Laemmli sample buffer lacking any dyes. After resuspension, samples were boiled for 5 min at 95°C and then placed on ice. Lysate protein concentrations were measured using the Pierce 660-nm protein assay (Thermo Fisher) and with comparison of the lysate A 600 with a linear curve fit of bovine serum albumin (BSA) standards. The mass of protein from the sample was then divided by the number of viable cells to yield the average protein mass/cell. The resulting average protein masses per cell were (492 Ϯ 170) ϫ 10 Ϫ15 g/cell in M2G and (519 Ϯ 228) ϫ 10 Ϫ15 g/cell in PYE.
Absolute quantitation of C-terminal YFP fusions. Five-milliliter mid-log cultures of NA1000 cells harboring C-terminal eYFP fusions (gift of the Shapiro Lab, Stanford University) were grown in M2G or PYE medium overnight in a roller wheel at 28°C. Once the cells reached an OD 600 of 0.3, cells were spotted on M2G agarose pads for imaging. Images were collected on a Leica DM6000B microscope with a Hamamatsu C9100 electron multiplying charge-coupled device (EMCCD) camera and a 100ϫ PH3 Plan Apo 1.40-numerical-aperture (NA) objective with in a Semrock model 2427A YFP filter cube with 100-ms exposure time. Fluorescence intensity was quantified using ImageJ by segmenting the cells and measuring the average pixel intensity of the cell area. Background intensity was subtracted using the NA1000 average YFP pixel intensity. For each fusion strain, a minimum of 50 cells were used for the analysis with a minimum of two technical replicates. As MipZ molecules per cell had been previously measured (62) by quantitative Western blotting, we converted the MipZ-YFP pixel intensity to the number of molecules/cell and multiplied this conversion factor by the YFP intensities of all other C-terminal YFP fusions. The average across replicates and the standard deviation () are reported in Table S2.
Doubling time measurements. Treatments were started from log-phase cultures grown overnight in the absence of cyanocobalamin (Sigma-Aldrich) and diluted in fresh medium to an OD 600 of 0.05. Each treatment was split into a separate flask, and the correct amount of cyanocobalamin was added to the concentrations of 1 nM, 0.1 nM, 0.01 nM, and 0 nM. Fifty milliliters from each treatment was added to three different 250-ml Kimex flasks, for three replicates of each of the four treatments. An initial OD 600 measurement was taken of each replicate using a cuvette and a NanoDrop spectrophotometer. The 12 250-ml flasks were then placed in a 28°C shaker incubator at 250 rpm. OD 600 time points of each flask were taken throughout the logarithmic growth phase. An exponential regression of the log-phase time points was used to calculate the doubling time of each replicate.
Translation reporter assay. JS417, JS423, and JS440 strains were started from log-phase cultures grown overnight in the absence of cyanocobalamin and diluted in medium with vanillate and antibiotic to an OD 600 of 0.05. A dilution series of each strain was used to fill tubes with 2 ml of culture at each cyanocobalamin concentration: 27 nM, 2.7 nM, 0.9 nM, 0.3 nM, 0.1 nM, 0.033 nM, and 0 nM. The 21 2-ml-cultures were then grown and induced over an 8-h period by placing the tubes in a 28°C shaker incubator at 250 rpm for 8 h. After 8 h, 2 l from each culture was pipetted onto an M2G-agarose pad on a microscope slide. Each treatment was imaged on a microscope using both phase contrast and an mChy filter cube. Average fluorescent intensities were calculated using MicrobeJ (63) across a minimum of 100 cells. B 12 homolog identification and phylogenetic tree mapping. Protein sequences for btuB, metH, and metE were determined for each Caulobacter species with a complete genome by using the NCBI Basic Local Alignment Search Tool (BLAST) with default settings by searching the protein sequence of each NA1000 gene for homologs with an E score of Ͻ10 Ϫ19 (64). The btuB and metH genes were then used to generate the phylogenetic tree using the NCBI Genome Workbench software and the MUSCLE multiple sequence alignment package (65). The tree is a maximum likelihood generated with the default settings from MUSCLE. Riboswitches were identified using rfam (66) and by searching the upstream regions of btuB and metE genes (up to 1,000 bp upstream of their predicted operons).
Proteomap generation. Categories were taken from predefined Kyoto Encyclopedia of Genes and Genomes (KEGG) categories (67). Categories were then ranked based on priority, and any gene that might have been present in more than one category was deleted from those with lower priority. The 200 most numerous proteins were then hand checked. Any that had not been automatically assigned to a KEGG category were compared against other organisms to place them in their most appropriate category. The categorized genes along with the ribosome profiling data were used to create the Proteomap (68).
Strain construction. (i) JS417. The insert btuB_5=UTR was generated by IDT as a gBlock construct for the ϩ1 transcription start site (TSS) through the start codon of the btuB gene. The btuB_5=UTR gBlock had the sequence AAGCGTTCAATTGGATCCAATCTTGACGTCCGTTTGATTACGATCAAGATTGGATCCAGCGT CAGGTTCCTCGAAAGAGGATGAAAAGGGAACGAGGTTGAAGACCTCGGCTGCCCCCGCAACTGTAAGCGGCGA GCTTCGCGTCACATGCCACTGGGCCCAAAAGGCCTGGGAAGGCGACGCCCAGAAGCATTGACCCGTGAGCCAG GAGACCTGCCCGGCGCAGTCGTTCATCGCTCGGCCGGGGTGCGCCGAACGAACGGGATCTCCCGAGAAACGAC AGTCAACAGGCCGCGCGACGGCCTGAGCGTCCGCGTCTTCGCGGGCGGTCGGGAGGTCGCGTGGGTCGTTCAT AACGGGAAGACTGTATTATGTTAATTAATATGCATGGTAC.
The plasmid pRVChyC-2 (60) was cut with MfeI and PacI, and the gBlock segment btuB_5=UTR was inserted into the plasmid by Gibson assembly. Next, the resulting plasmid was transformed into E. coli DH5␣ cells and selected on LB-kanamycin (Kan) plates. The resulting Kan r colonies were then screened by PCR for the insert and verified by Sanger sequencing (Genewiz). The purified plasmid was then transformed into NA1000 cells by electroporation and plated on PYE-Kan plates. The resulting colonies were screened for mChy fluorescence after induction with vanillate.
(ii) JS423. The insert metE_5=UTR was generated by IDT as a gBlock construct for the ϩ1 TSS through the start codon of metE. The metE_5=UTR gBlock had the sequence AAGCGTTCAATTGGATCCAATCTTGA CGTCCGTTTGATTACGATCAAGATTGGATCCAGTCGTGGTCTGCGGACGTTCGCGTCCGGAGCTAAGAGGGAAG TCGGTGAGGGCGTGAAACCCTGAATCCGGCGCTGCCCCCGCAACTGTGAGCGGCGAGCCGCTGTCCGTTTCGTG TCACTGACGCGCCGAAGCTGGTTCGGGGATGCGTCGGGAAGGCCAGGGCAGGGGTGACGACCCGTGAGCCAG GAGACCTGCCTCGACAGATAACGTCCTCCGGCGGGGTGTCCGGTCTGGCCGCTTGCTCAGCGCGACCGGACAA The plasmid pRVChyC-6 (60) was cut with MfeI and PacI, and the gBlock segment metE_5=UTR was inserted into the plasmid by Gibson assembly. Next, the resulting plasmid was transformed into E. coli DH5␣ cells and selected on LB-chloramphenicol (Chlor) plates. The resulting Chlor r colonies were then screened by PCR for the insert and verified by Sanger sequencing (Genewiz). The purified plasmid was then transformed into NA1000 cells by electroporation and plated on PYE-Chlor plates. The resulting colonies were screened for mChy fluorescence after induction with vanillate.
(iv) JS441. The JS441 strain was generated by PCR amplifying the last 500 bp of the ␤= RNA polymerase gene into the pYFPC-1 plasmid (60). The insert was PCR amplified from the NA1000 chromosome using Betaprime_forward and Betaprime_reverse primers, and the plasmid pYFPC-1 (60) was PCR amplified by pYFPC_forward and pYFPC_reverse primers. The plasmid was then treated with restriction enzyme DpnI, and the insert was placed into the plasmid by Gibson assembly. Next, the resulting plasmid was transformed into E. coli DH5␣ cells and selected on LB-spectinomycin (Spec) plates. The resulting Spec r colonies were then screened by colony PCR for the insert and verified by Sanger sequencing (Genewiz). The purified plasmid was then transformed into NA1000 cells by electroporation and plated on PYE-Spec-streptomycin (Strep) plates. The resulting colonies were screened for YFP fluorescence.
Data availability. All ribosome profiling raw sequencing reads and normalized read count values of cells grown in M2G are deposited in the NCBI GEO database with accession number GSE126485. PYE ribosome profiling data were collected in reference 36 and were pulled from NCBI GEO database accession number GSE54883. Absolute quantitation values of mRNA translation for both M2G and PYE can be found in Table S1. Ribosome profiling data were compared to the following data sets collected in previous reports. Protein localization data are from reference 26. Tn-seq fitness data are from references 39 and 57. Gene essentiality data are from reference 69. KEGG categories are from reference 70. DNA binding site counts for cell cycle master regulators are from reference 10.