The +OSC and −OSC strains were constructed for this study (see Table 1) and were made using PCR mutagenesis in which the primers contained the changes of interest (see Table 2 for the primers). The +OSC strain was constructed to contain the +1T frameshift downstream of codon CAA-Q37 while keeping the two OSCs naturally found in the B. subtilis argF sequence intact. The −OSC strain was constructed to contain the same +1 frameshift downstream of codon Q37, and additional synonymous changes were made to codons to remove the OSCs but retain the amino acid sequence. In brief, we designed primers with the desired changes to argF that would introduce the +1 frameshift and remove the OSCs (Table 2). To construct the +OSC strain, primers (1) and (6) were first used to amplify a 942-bp segment using NEB Phusion High-Fidelity DNA Polymerase with B. subtilis genomic DNA as the template. After purifying the PCR product using a Promega PCR clean-up kit, the 942-bp fragment was used as the template for PCR 2 using Phusion High-Fidelity DNA Polymerase. PCR 2 used primers (2) and (6) to amplify a 965-bp fragment. We then used primers (3) and (4) to amplify a 593-bp fragment with Phusion High-Fidelity DNA Polymerase. All PCR reactions were cleaned up as described above, and the concentration was determined. A 1:1 ratio of fragments 593-bp and 965-bp were combined in a new PCR reaction, mixed with NEB GoTaq reaction reagents, and amplified in seven cycles of denaturing, annealing, and extension. During these steps, two products would denature into single-strand DNA, anneal over 21b (overlap), and amplify during extension. After seven cycles, primers (5) WT argF F (SalI site) and (6) WT argF R (SphI site) were added to the reaction, and PCR proceeded for 16 cycles. The 1,056-bp product was resolved on a gel, excised, and purified. The product was cloned into pdr111 amyE-hyper-SPANK (containing a spectinomycin resistance cassette) plasmid using the SalI and SphI restriction sites and transformed into CV4000 (arg::neo) and YB9801 (mfd::tc) using standard B. subtilis transformation techniques. Transformants were selected on tryptose blood agar base medium (TBAB; Difco Laboratories, Detroit, Mich.) plates containing spectinomycin (YB955) and tetracycline (YB9801) antibiotics. The strains were verified by Sanger sequencing. The same strategy was used to construct the −OSC strain, with primers 1, 2, and 4 substituted with 7, 8, and 9 respectively.

The protocol established by Sung and Yasbin (2002) was followed. Strains were streaked onto TBAB plates containing neomycin and spectinomycin for the +OSC and −OSC strains and tetracycline when working with the Mfd counterpart; they were incubated overnight at 37 °C. The next day, a colony from the TBAB plate was used to inoculate 2 mL of Penassay Broth (PAB; antibiotic medium 3, Difco Laboratories, Sparks, MD, USA), which was grown overnight at 37 °C with aeration. An aliquot of the overnight culture was then added to a 250 mL Erlenmeyer culture flask containing 10 mL of PAB medium and 10 μL of 1,000× Ho-Le trace elements (Sung and Yasbin, 2002). Cells were grown at 37 °C with aeration (250 rpm) to stationary phase. Growth was monitored with a spectrophotometer measuring O.D. at 600 nm (OD600). Cells were harvested 90 min after the end of exponential growth and centrifuged at 4,000 rpm for 5 min; after decanting the supernatant, the cells were resuspended in Spizizen minimal salts (SMS). This process was repeated twice. Initial cell titers were determined by serial dilution, and 100 μL of cells were plated on minimal media supplemented with 100 μg/mL arginine. Colonies were counted after 48 h of incubation at 37 °C. After that, 100 μL aliquots were plated on minimal media containing 0.1 mM IPTG and no arginine (n = 5) and incubated for 12–14 days at 37 °C. Plates were scored daily for the appearance of Arg⁺ colonies, and initial titers were used to normalize the cumulative number of revertants per day to the number of the CFU plated. To assay background cell survival, viability was tracked by taking agar plugs every other day, serial diluting, and plating on minima media supplemented with 100 g/mL arginine. The experiments were repeated four times.

We performed a fluctuation test on each strain to determine the mutation rates for arginine prototrophy during growth. In brief, an aerated culture grown overnight in 2 mL of PAB medium at 37 °C, was diluted 1:10 in 1× SMS, vortexed, and diluted 1:10 again in this solution. Aliquots of the second 1× SMS solution (0.5 mL) were added to 49.5 mL of PAB supplemented with 0.1 mM IPTG. Cells were vortexed, and 1 mL of the 10^–4 diluted cell mixture was dispensed into 40 different 18 mm test tubes. Cells were grown for 12 h at 37 °C with aeration (250 rpm). Each test tube culture was pelleted, washed, resuspended in 100 µL of 1× SMS, and individually plated onto minimal media lacking arginine with 0.1 mM IPTG. Separately, 1 mL aliquots of the 10^–4 culture, individually dispensed into individual tubes, were serially diluted and plated on minimal media supplemented with 100 μg/mL arginine to determine cell count. After incubating at 37 °C for 48 h, the colonies were counted. Mutation rates were determined using the Lea–Coulson method of the median (Foster, 2006; Rosche and Foster, 2000).

Arg⁺ colonies that appeared between days 5 and 12 in the SPM were picked and patched on minimal media lacking arginine and containing 0.1 mM IPTG to ensure stable mutations. Arg⁺ colonies were taken from four separate SPM experiments. After patches were grown for 48 h s, a small sample was then grown in minimal broth; genomic DNA was extracted and used for Gotaq PCR using Pspac F primer and WT argF rev (SphI site). PCR products were sent to Functional Biosciences (Madison, WI) for Sanger sequencing. Routinely, argF sequencing 1RV was chosen as the sequencing primer as it covered 286 bp of the gene and the promoter. To ensure capture of the whole gene argF sequencing, 2RV was also sequenced for some Arg⁺ mutants. See Table 3 for list of sequencing primers.

We took an indel event that occurred in the +OSC and −OSC strain and reconstructed it in the opposite strain by inserting the sequence at the same site at which it occurred in the parent strain. We then translated the sequence to determine its effect on the protein loop composition.

Significance was determined using an unpaired Student’s t-test to compare two strains and an ordinary one-way analysis of variance to compare more than two strains. Each comparison had a minimum of three trials.

The B. subtilis 319 amino acid ArgF protein sequence was input into pBLAST as the query sequence and blasted into the UniProtKB/Swiss-Prot (SwissProt) database. The output identified 100 ArgF amino acids (complete list in supplemental). We used the pBLAST distance tree results to identify the domain and phylum that each sequence fell into. We then exported the pBLAST multiple sequence alignment into SnapGene for further analysis. Additionally, positions 57–89 from the alignment corresponding to the protein loop region were then imported into WebLogo (Crooks et al., 2004) to measure amino acid conservation at those positions (bits).

We took four entries out of the 100 that were representative of different loop lengths and phylogenetic groups and looked at their Swiss-Model structures under the 2D structure mode. The 2D structure identified and colored alpha-helices (blue/purple) and beta-sheets (green). We then located the loop sequence we had identified for each entry using the protein alignment and confirmed that it was indeed found in a loop in the model. We zoomed in on the loop and labeled the first and last position. The Swiss-Model Identifiers for the four entries were P18186 (OTC_BACSU), Q5N1N8 (OTC_SYNP6), Q8TKT5 (OTC_METAC), and Q2G3J8 (OTC_NOVAD).

The percentage likelihood of forming OSCs was calculated in two ways. First, we considered all possible codon pairs at positions 35, 36, 64, and 65 (referencing the Weblogo figure, which shows the OSCs of interest) and the likelihood of forming an OSC, then we considered the frequency of each amino at each given position. This analysis considered all possible amino acid pairs for positions 35, 36, 64, and 65. In the second method, we calculated the likelihood of an OSC for each of the 100 sequences based on their specific amino acid pairs at these positions and then found the likelihood for these 100. This analysis excluded non-occurring amino acid pairs.

In Bacillus subtilis and other bacteria, the argF gene encodes ornithine carbamoyltransferase (Otc), a protein that catalyzes the sixth step in arginine biosynthesis. We previously showed that engineered mutations in argF can render bacterial cells auxotrophic for arginine and also score mutations that restore growth (Ermi et al., 2021; Martin et al., 2019). After identifying 46 OSCs in argF (Supplementary Figures S1 and S2), we determined whether OSCs occur in conserved protein regions by analyzing B. subtilis OSCs and 100 Otc protein sequences from diverse prokaryotes (list found S10). We found 14/46 (∼30%) of B. subtilis OSCs mapped to conserved amino acid codons (Figure 1), with 13 of the 14 conserved OSCs being opal stops (TGA) and one an ochre stop (TAA) (Supplementary Figure S3). An amino acid was defined as conserved if >50% of protein sequences had the same amino acid.

Using publicly available Otc structures, we then asked if OSC sites in B. subtilis occur in particular protein features such as near substrate binding sites or protein loop regions (see supplemental Otc structure analysis). We found that several OSCs flanked an Otc protein region that varied in sequence and length corresponding to a loop region spanning amino acids K36–K47 in the B. subtilis protein and regions 60–84 in the alignment (Figure 1, see boxed region). Analysis of 100 sequences that included bacteria and Archaea indicated that loop length varied from 11 to 21 amino acids and variation tracked with taxonomic groups (Figures 2A and B). Bacillus species belonging to phylum Bacillota (pink bar) trended toward decreased loop size (<16 amino acids), similar to cyanobacteria, with a loop length of 11 amino acids (black bar). In comparison, Pseudomonadota (green) had loop sizes of 21 amino acids, although these loops were less frequent (Figures 2A and B). Of these 100 sequences, most had loop lengths that were 16 amino acids long (Figure 2A). While we assumed that we identified a loop region for the 100 sequences based on available Otc structures and because this region fitted the characteristics of a loop region (flexible in sequence and length), we followed this analysis up by looking at the Swiss-Model structures of a few representatives. We were able to verify that the alignment was a good predictor of protein loops (Figure 2C). Interestingly, a small beta-sheet consisting of three amino acids was predicted to be in the “loop” region of the Pseudomonadota representative (Figure 2C, last panel). This observation supports the idea that loops can evolve and potentially form new 2D structures and domains that are important for protein evolution (Blouin et al., 2004).

Since protein loops tolerate indel events and are postulated to be hotspots for protein evolution (Blouin et al., 2004), we asked whether Bacillus OSCs flanking the loop region are conserved among prokaryotes, suggesting that they have a biological function. We analyzed the frequency of each amino acid in 100 prokaryotic Otcs to determine the codon options in positions where OSCs were present. The results suggest that OSCs immediately flanking the loop region are conserved and negatively correlated to loop size (Supplementary Figures S5,S6). We became interested in the OSC preceding the B. subtilis Otc loop corresponding to L₃₅–K₃₆, because this position pair had a high chance of forming an OSC independent of codon bias in 85/100 sequences due to the presence of a NTN codon at position 35, followed by lysine AAN (Supplementary Figure S11).

Beyond in silico results, we established an experimental system to determine whether OSCs preceded the loop influence accumulation of spontaneous indel events and the production of functional Otc variants. Our studies focused on OSCs between the codons of amino acids 33–36 in the B. subtilis Otc (Figure 3A, top panel). In one case (−OSC), we introduced synonymous base substitutions in this DNA region to remove both OSCs while maintaining the amino acid sequence of Otc. In the second case (+OSC), we selected for indel events by inserting a +1T at codon 38 of the argF gene to generate an ochre stop (+1TAA), rendering the cells auxotrophic for arginine (Arg⁻) (Figure 3A). The +1 insertion was also introduced to the ORF in the −OSC strain; thus, both strains could be used in a Arg⁻→Arg⁺ selection system. Next, engineered argF variants were placed under an IPTG inducible promoter with low levels of basal transcription (Castillo-Hair et al., 2019). After IPTG induction, we measured the mutation frequency of indel events that restore arginine prototrophy during stationary growth (i.e., the generation of Arg⁺ colonies on minimal media lacking arginine) in the presence and absence of OSCs.

We used a stationary-phase assay to determine how the presence or absence of OSCs directly upstream of the ArgF K₃₆–K₄₇ loop influences the accumulation of spontaneous indel events in the Otc loop during arginine starvation. In brief, we grew cells to stationary phase, plated them on minimal media lacking arginine, and scored Arg⁺ colonies over 14 days. We found that OSC removal increased the mutation frequency ∼3.5-fold (Figure 3B).

Because Arg⁺ prototrophy could result from deletion or insertion events, we sequenced the argF gene in stationary-phase Arg⁺ revertants and found that 57% of mutants in the −OSC strain had insertion events compared to 9% in the +OSC strain (Figure 3C). To better understand these differences in indel events, we determined the site and type of indel events in +/−OSC strains. Both strains had common indel sites (Figure 5AB, black x); two indel sites were unique to the +OSC strain (Figure 5A, blue x), and six were unique to the −OSC strain (Figure 5B, red x). Three unique sites in the −OSC strain occurred before codon 35 (Figure 5A). We also queried what indel events occurred in each strain and analyzed their frequency in the mutant population. Even when the indel site and type were the same (−1A) in both strains (Figure 4BC, black x₂), the frequency varied (30% in +OSC and 10% in −OSC). Importantly, a single site contained several different indel events and was more common in the −OSC strain (Figure 5C red x₄). The −OSC strain acquired more indel events downstream of codon 35, and the insertions were more complex (larger than two nucleotides) than those in the +OSC strain (Figure 5B).

To better understand how OSCs affect genetic variation and its consequences in amino acid composition of the loop region, we examined the impact of indel events on the codon sequence in Arg⁺ mutants. We focused on how each indel event affected the codon composition and length of the Otc loop region relative to the WT protein. We identified a 12-amino-acid Otc loop region in B. subtilis spanning residues K₃₆–K₄₇ (Figure 5D) and calculated the percentage occurrence of specific indel event types in the sequenced mutant population (Figure 5B). Specifically, we identified: 1) the indel site (first column); 2) the indel event that occurred at the site (second column); 3) the consequences of the indel event on the amino acid Otc loop (third and fourth columns); 4) the frequency of that indel event among those sequenced (fifth column labeled % occurrence). Using a sequence reconstruction analysis, we also recreated the indel event in the opposite strain to examine its consequence on the Otc loop (sixth column). We found that the Otc loop region could tolerate deviations from WT loop length and composition, which is consistent with our amino acid sequence alignment showing that this region varies in loop size and composition across prokaryotes (Figure 1).

We found three indel events common to both strains that combined to represent the most frequent indel event in the strains (Figure 5). Of note, deletion events −1A (x₂) and −1A (x₅) occurred in the context of a di-adenine run (Figure 5A). Although the loop length of these indel events (−1T x₁, −1A x₂, −1A x₅) did not deviate from the WT, they did change the amino acid composition. A −1 deletion event common to both strains (X₁ −T in black) caused different amino acid composition in the loop (arginine 35R in +OSC strain, and proline 35P in the −OSC strain), both of which deviated from the WT sequence by three amino acids. Interestingly, this deletion comprised 17.4% of the total mutants in the +OSC strain and 14.3% in the −OSC strain.

By examining the codon sequence generated by indels unique to each strain, we found that the B. subtilis Otc loop region acquired codon sequences that deviated from the WT loop. Indel events that occurred uniquely in both strains resulted in shortening of the loop by one residue and lengthening of the loop by up to two amino acid residues. However, the −OSC strain accumulated larger indels than the +OSC strain at or downstream of codon 35, except for one four-base pair deletion; these genetic changes produced longer Otc loops (Figure 5). The unique indel events in both strains were detected at similar frequencies (4.4 or 4.8%) in the sampled mutant population, suggesting no differences in fitness costs between unique mutants.

To gain insight into why an indel event was found in one strain and not the other, we performed a sequence reconstruction analysis to recreate the indel event in the opposite strain (column six) and observed consequences on the loop. In this analysis, ∼36% (4/11) of unique indel events in the −OSC strain would not be possible in the +OSC strain because they would place an out-of-frame stop codon in frame, resulting in translation termination and a non-functional truncated Otc protein (Figure 5E).

We previously showed that Mfd and transcription levels are important drivers of base substitution mutations in the B. subtilis stationary phase (Martin et al., 2011). We also asked whether these factors affect indel production in our +/−OSC argF systems. In the absence of Mfd and IPTG, which induce transcription beyond basal promoter levels, Arg⁺ mutation frequency decreased in the −OSC strain (Figure 6). Since our previous studies showed that distinct mutagenic mechanisms occur in exponentially growing cells but not in stationary-phase cells (Leyva-Sánchez et al., 2020), we asked whether Mfd affected the ArgF⁺ mutation rate in growing B. subtilis strains with or without OSC. Consistent with stationary-phase cells, the −OSC strain had a higher mutation rate than the +OSC strain. Sequence analysis was also consistent with profiles seen in stationary-phase cells, with the −OSC strain having more complex insertion events in the loop K₃₆–K₄₇ (Supplementary Figures S8-S9). Importantly, 40% of the unique indel events observed in growing −OSC strain cells would not be possible in the +OSC strain because they would place an out-of-frame stop codon in frame, resulting in translation termination and a non-functional truncated Otc protein (Supplementary Figure S9). However, the effect of Mfd was opposite, with the results showing that the absence of Mfd increased the production of ArgF⁺ revertants in the +OSC strain. In contrast, no impact was observed in the −OSC strain. Thus, in growing B. subtilis cells, mfd prevents indel events at OSC, leading to argF ⁻ gene reversions.