Two hundred microliter of 10³ CFU/ml BM CMCC55210 suspension was inoculated on TSA plates, spread well, and incubated at 37°C for 3 days. Single-colony colonies were picked and inoculated onto a coverslip, 50 μl of acridine yellow hydrochloride solution (1:500, Sigma-Aldrich, USA) was added, the coverslip was covered and then gently ground. The colonies observed visually with noticeable agglomerated particles or flocculent particles were judged as rough mutants. The remaining colonies of the mutant were picked and inoculated on TSA plates and incubated at 37°C for 2 days. Then, the bacteria were harvested by scraping the moss, added to 1 ml of preservation solution (TSB solution containing 20% glycerol), and then frozen at −80°C in a refrigerator. Crystal violet staining was performed to further verify the rough phenotype of the selected mutant strains (6).

One hundred microliter of 10³ CFU/ml BM CMCC55210a rough bacterial suspension and BM CMCC55210 smooth bacterial suspension were inoculated on TSA plates separately, spread well, and incubated for 3 days at 37°C. Subsequently, 2 ml of staining solution [0.05% crystal violet (Sigma-Aldrich, USA), 25% methanol] was poured over the plates. After 15 s, the staining solution was discarded and poured into a disinfectant (75% ethanol). In contrast to the light purple background, smooth colonies were stained a translucent light blue-green, whereas rough colonies showed red to blue-red. Following the staining process, the peripheral colors of smooth colonies progressively became deeper blue-violet, but the uniform coloration of rough colonies was maintained for 4 days.

The bacterial suspension of smooth strain and selected natural mutant strain was inoculated on TSA medium containing basic fuchsin and thionin (1:50,000 w/v, Sigma-Aldrich, USA), respectively, and then incubated at 37°C for 3 days. Bacteria growth was determined by colony counting. Both BM strains should grow well on the TSA plates, adding basic fuchsin and thionin as normal TSAs.

The strains BM CMCC55210 and CMCC55210a were inoculated on TSA plates. Lead acetate test papers (Sigma-Aldrich, USA) were placed on the surface of the agar and co-incubated at 37°C. The color of the test papers was checked after 2 days. The paper would turn black when hydrogen sulfide is produced. The cultivation of Brucella melitensis should be free of hydrogen sulfide production.

Phage propagation and titration were those described in previous references (7). One hundred microliters of 10⁹ CFU/ml BM CMCC55210 and CMCC55210a were inoculated on the TSA medium and distributed evenly. Each plate was dripped with 5 μl of Brucella bacteriophage Tb, Wb, Fi, and BK2 at a routine test dilution (RTD) concentration. After incubating at 37°C for 2 days, the plaques by phage lysis were visually determined. Phages Tb, Wb, and Fi were unable to lyse Brucella melitensis, and no phage plaque was present on the plates. Phage BK2 was capable of lysing Brucella melitensis, and neatly edged, translucent plaques were presented.

Female BALB/c mice (~42 days old), purchased from Slack Laboratory Animal Co., Ltd (Shanghai, China), were kept in an individually ventilated cage (IVC) rack system of the ABSL-3 laboratory of Fudan University. Animal experiments were reviewed and approved by the Animal Welfare and Ethics Committee of the School of Basic Medical Science of Fudan University (Approval No. 20201105-001).

The strains BM CMCC55210 and CMCC55210a were cultured on TSA for 3 days, and the bacterial suspension was harvested in NS and adjusted to 10⁹ CFU/ml as previously. Mice were challenged by intraperitoneal injection with 100 μl bacterial suspension of CMCC55210 strain and CMCC55210a strain. The number of mice in the negative control group was injected intraperitoneally with 100 μl of NS. Actual infection doses as above were further confirmed by serial dilution, culture, and CFU counting. At 2-, 14-, 28-, 42-, and 56-day post-infection, 14 mice (including six of the CMCC55210 group, six of the CMCC55210a group, and two of the NS group) were weighed and euthanized with an overdose of sodium pentobarbital (200 mg/kg body mass) and then immersed in 75% alcohol for 3 min. Afterward, the spleens were aseptically obtained, weighed, examined for gross pathology, photographed, and then homogenized in 1 ml PBS buffer. The spleen homogenates were diluted 10-fold with 1 × PBS in sequence and plated on TSA to determine the spleen bacteria load by counting CFU. The splenomegaly index was performed as spleen/mouse weight × 100%.

Genomic details of the strain BM CMCC55210 were unclear, and the genome information of reference strain BM 16M (21) was generally used for analysis. Thus, we performed genome sequencing of both strains through Illumina and Pacbio platforms. The genomes of CMCC55210 and CMCC55210a were composed of two chromosomes; the lengths of Chr1 and Chr2 were 2.12 and 1.18 M, and the GC contents were 57.16 and 57.34%, respectively (Table 2). The average nucleotide identity between the two strains was 99.99%. Then, we performed COG function and genomic islands analysis within the genomes of CMCC55210 and CMCC55210a, and the results showed that there were no significant differences at the genomic level (Figure 3).

To define genomic variations in mutant strain CMCC55210a, we performed comparative genomics between CMCC55210 and CMCC55210. The genome of CMCC55210 was used as analysis control, and 19 insertions (INs), 11 deletions (DELs), and 3 single nucleotide mutations occurred in the strain CMCC55210a. Considering the genome variations occurring in the intergenic region did not affect protein functions; therefore, we focused on mutations occurring in the protein-coding regions, including 4 INs, 8 DELs, and 2 single nucleotide mutations involving seven genes, namely trmFO, trmD, xerC, pntA, hsdM, livM, and manB (Table 3). From the gene length statistics, three genes, including the xerC (BRUCa_1896), trmD (BRUCa_1892), and hsdM (BRUCa_2973) of CMCC55210a strain, became shorter, whereas the pntA (BRUCa_3109) and livM (BRUCa_2789) were lengthened. The manB (BRUCa_2508) were the same length in all three samples, with only localized base and amino acid differences.

The protein XerC sequences of strains 16M (BME_RS00695) and CMCC55210 (BRUC_1898) were identical. They contained two large functional domains, Xer and Phage integrase, whereas the corresponding gene, BRUCa_1896, in strain CMCC55210a, had a deletion of cytosine (C) at site 160, resulting in a shortening of the 5′ end of the gene. Other variations occurred in xerC of CMCC55210a, including guanine (G) to C at site 256, resulting in the change of amino acid from alanine (A) to proline (P) and the insertion of C at site 328 (Figure 5). The most critical variant may be the incomplete Xer functional domain affecting the gene's normal function. The XerCD site-specific recombination system is important for the stable inheritance of circular replicons, such as plasmids and bacterial chromosomes (29). Similar to other bacteria, Brucella species harbor circular chromosomes. The formation of chromosome dimers is caused by odd numbers of crossovers during homologous recombination (30). Translocase FtsK activates XerCD recombination (31, 32) at a specific site, dif, so that recombination occurs at the right time and place. Recent studies revealed that XerD of Staphylococcus aureus and Bacillus subtilis unloads structural maintenance of chromosome (SMC) complexes through the binding of XerD to additional chromosomal loci, not dif, a different action that does not depend on XerC (33). We hypothesize that the redundancy mechanism of XerC-XerD also exists in Brucella to ensure bacterial chromosome stabilization.

The protein TrmFO sequences of strains 16M and CMCC55210a were identical, and both contained complete GIDA functional domains. In contrast, the corresponding gene BRUC_0891 in CMCC55210 strain was unable to form a normal functional protein due to a code-shift mutation at the 5′ end of the gene because of a deletion of a base C at the 27th base and an insertion of a base C at the 371st base, which resulted in the formation of normal functional proteins, and a 180 amino acid deletion at the 5′ end of the gene relative to the proteins in the other two samples, which resulted in an incomplete GIDA functional domain. It thus affected the regular expression and functional roles of the protein (Figure 6).

There is a 5-methyluridine at position 54 (T54) in the T-loop of bacteria tRNAs, and T54 methylation can stabilize the tRNA structure. The methyl group is transferred through two pathways: (1) from S-adenosylmethionine by TrmA methyltransferase (22) and (2) from 5,10-methylenetetrahydrofolate by TrmFO, a folate/FAD-dependent methyltransferase (23). In most Gram-positive and several Gram-negative bacteria, TrmFO orthologs have been identified through sequence analyses (24). TrmA orthologs, a class I SAM-dependent RNA methyltransferase, were found in Brucella species. In our study, gene trmFO of BM CMCC55210a was consistent with BM 16M. The mutations that occurred in trmFO may not be associated with the attenuation mechanisms.

The TrmD protein, encoded by the gene trmD, is a tRNA methyltransferase that suppresses translational frameshift errors at proline codons through methylation modification of tRNA m1G37, which is essential for bacterial growth (25, 26). The process of methylation modification exerted in Haemophilus influenzae by TrmD requires stabilization of S-adenosyl-L-methionine (AdoMet) at position 171 by phenylalanine (Phe) and capture of the tRNA 38-position phosphate group by glycine (Gly) at position 59 (27). Protein alignment analysis suggested that the TrmD corresponding active site of BM CMCC55210 was consistent with Haemophilus influenzae. The protein TrmD sequences of BM 16M and CMCC55210 were identical, and both contained the complete tRNA_m1G_MT functional domain. In contrast, the corresponding gene BRUCa_1892 in BM CMCC55210a had a shifted mutation due to the deletion of two bases, T and G, at positions 497 and 498, which resulted in a shifted mutation of the gene and a shortening of the 5′ end, leading to an incomplete tRNA_m1G_MT functional domain, which in turn affected the regular expression and functional role of the protein (Figure 7). This mutation is considered lethal to bacteria, but our results showed that the growth of Brucella melitensis CMCC55210a was not affected.

Another evolutionary study found that cumulative mutations were detected in the gene encoding proline-tRNA ligase (proline-tRNA ligase, proS) of restored growth of Escherichia coli trmD mutant strains (28). ProS amplifies m1G37 by distinguishing between m1G37 modified and unmodified tRNA^Pro to safeguard the translational accuracy, thereby strengthening the effects. The aminoacylation efficiency of m1G37-unmodified tRNA^Pro was lower than that of m1G37-modified tRNA^Pro, and the restored growth was detected in trmD mutant strains by the aminoacylation of m1G37-unmodified tRNA^Pro. Furthermore, sequence alignments revealed that the proS nucleotide sequence of BM CMCC55210a was unchanged, so unknown potential regulatory mechanisms that helped CMCC55210a adapt to the effects of gene trmD mutation need to be investigated.

The protein HsdM sequences of strain 16M and CMCC55210 samples were identical, containing complete functional domains of HsdM_N and N6_Mtase. In contrast, there are three deletions of 716A_del, 788A_del, and 1072C_del in the corresponding gene BRUCa_2973 of BM CMCC55210a (Figure 8). The deletion of the bases led to the coding gene shifting. The 5′ end is shortened, the HsdM_N functional domain is lost, and the N6_Mtase functional domain is incomplete, thus affecting the regular expression and functional role of the protein. HsdM, a methyltransferase (MTase) encoded by gene hsdM, is a subunit of type I restriction–modification (R-M) systems that provides a major defense against phages and invasive DNA in bacteria (36, 37). Typical R-M systems are composed of three separate subunits of DNA-specificity (HsdS), DNA-modification (HsdM), and DNA restriction (HsdR) (38, 39). Another new type I R-M system can utilize two copies of HsdM and one HsdS to form the M2-S MTase complex, one of which has NPPF in the highly conserved catalytic motif IV and modifies adenine to m6A, and one having an NPPY catalytic motif IV and modifying cytosine to m4C (40). Mutations in gene hsdM were detected in Streptococcus suis serotype 2 attenuated strains, but it was unclear whether HsdM mutations were associated with reduced resistance to phagocytosis (40). So, the relationship between hsdM and the attenuation of BM CMCC55210a in this study also needs to be further verified.

The LivM protein sequences of strains 16M, CMCC55210, and CMCC55210a are not identical. At site 448, that is, before the last triplet codon of the gene of the 16M sample, a single C is inserted, which results in a shifted mutation, the disappearance of the original termination codon, and a lengthening of the gene sequence. At site 798, BM CMCC55210a inserts a single G relative to CMCC55210, thus allowing the gene sequence to continue to be extended and the total protein sequence is the longest of the three. In terms of structural domains, the transmembrane region structural domains at the 5′ end were all complete. The sequences were 100% identical, while the BPD_transp_2 structural domain, strain 16M gene, was incomplete, with a portion of the 3′ end missing. The BPD_transp_2 structural domains of the strain CMCC55210 and CMCC55210a (position from amino acid 38–278) are different at the 3′ end, but both are structurally complete (Figure 9). The mutation of the CMCC55210a sample gene at the 3′ end may affect the protein's tertiary structure or active ability, which in turn affects the protein's function. Leucine-isoleucine-valine (LIV) is the major transport system in bacteria for the branched-chain amino acids. The livM-encoded protein localized to the inner membrane of the bacteria cell is one of the membrane components LivHMGF, that is required for high-affinity leucine transport (43, 44). At least six copies of LivHMGF operons were detected in the chromosomes of BM CMCC55210a and CMCC55210. We supposed the mutation that occurred in livM of CMCC55210a had no relationship with the rough phenotype.

The gene pntA of all three strains contains three functional domains. The protein sequences of strains 16M and CMCC55210 were identical. The gene BRUCa_3109 corresponding to strain CMCC55210a is missing a single C at site 1125, resulting in a shifted mutation that affects the third functional domain differently. The third functional domain, low complexity, is located from amino acid 348 to amino acid 399 in the protein sequences corresponding to all three samples (Figure 10). The gene pntA encodes PntA, which is the α subunit of the pyridine nucleotide transhydrogenase (PntAB). As a catalyzer of energy-coupled electron transfers between NAD(H) and NADP(H), PntAB regulates electron carrier redox balance (35). Previous studies have demonstrated that pntAB adaptive truncation mutation to decrease PntAB-catalyzed NADPH biosynthesis occurred in E. coli K-12 MG1655 phosphoglucose isomerase-deficient strain with enhancing environmental adaptability through adaptive evolution (34).

The gene manB was identical in sequence length in all three strains. The corresponding gene BRUCa_2508 in strain CMCC55210a has an SNP mutation at 1067 from thymine (T) to guanine (C), which results in a change in the coding amino acid from leucine (L) to proline (P) and the functional domain corresponding to this mutation position is PGM_PMM_III (Figure 11). Phosphomannomutase (ManB), encoded by manB, converts mannose-6-phosphate to mannose-1phosphate, which is a key metabolic factor during the synthesis of lipopolysaccharide (LPS) (45). LPS is composed of three domains: lipid A, core oligosaccharide, and O-antigen (46). The lack of the O-antigen or core oligosaccharide contributes to the rough LPS phenotype of Brucella melitensis. It has been found that manB, interrupted by the transposon mutant strain of Brucella melitensis H38, confronted attenuation and lack of the core oligosaccharide and the O-antigen (47). Proline is the most common first residue of α-helix and the marginal chain of β-sheet, a disruptor in the secondary structure of the protein owing to its unique ring-loading structure of the side chain that maintains exceptional conformational rigidity (48). We hypothesized that the L356P mutation may affect the structure of α-helices 12, thereby affecting the spatial conformation of the glycan-binding region of domain PGM_PMM_III. The crystal violet staining experiment showed that CMCC55210a was a rough strain, and its rough change might be caused by the truncated LPS structure.

We summarize a brief map of key genomic differences and their mechanistic implications for bacteria–host interactions in Figure 12.