Laboratory divergence of Methylobacterium extorquens AM1 through unintended domestication and past selection for antibiotic resistance

Whole-genome sequencing reveals extensive genomic divergence in the Modern AM1 lineage

To explore the extent to which AM1 has diverged at the genomic level, we used whole-genome sequencing to compare the Archival genome to a previously sequenced Modern reference [21]. Illumina sequencing reads were analyzed both by mapping onto the Modern reference, and through de novo sequence assembly. For sites in which these strains differed, we compared the mutational state at each site (i.e., Archival or Modern) to other previously sequenced strains of M. extorquens to determine whether substitutions occurred in either the Modern or Archival lineage (Figure 1). While this analysis cannot identify changes that occurred between the divergence of “wild” AM1 from the Archival/Modern laboratory ancestor, these mutations, if any, would only add to the ways in which AM1 has evolved in the lab.

Some Modern mutations are likely to have far-reaching physiological consequences. Certain mutations target highly pleiotropic genes such as rplJ, which encodes the ribosomal subunit protein L10; rpoB, the beta subunit of RNA polymerase; and recG, the primary DNA helicase involved in recombination repair and other functions. The insertion of mobile elements might also have altered physiology in the evolution of Modern AM1, jumping into several putative protein-coding genes. Still other mutations resulted in the loss of a substantial amount of DNA from Archival to Modern AM1. Deleted regions were typically flanked by insertion sequences and other low complexity DNA, but in most instances we could infer the likely genomic location and content of these deletions (Table 1). Those genes deleted in Modern AM1 are predicted to perform a variety of functions, and appear in some instances to be homologs of genes found only in distantly related members of the Methylobacterium genus. Further insights into the extensive loss of DNA in Modern AM1, as well as the functional and evolutionary consequences of Modern mutations, will require considerable future work. Still, these results clearly show that the Modern AM1 lineage has undergone a substantial degree of evolutionary divergence and domestication over fifty years of growth and storage in the lab.

Archival is faster and fitter under most standard growth conditions

To begin to explore the phenotypic differences arising in Modern AM1, we compared the performance of Modern and Archival strains grown on standard growth substrates in “Hypho” minimal medium. Two primary metrics were used to assess growth: specific growth rate, determined from the increase in optical density (OD₆₀₀) for a particular strain/condition using a custom-built, high-throughput microbial culturing system and analysis software [27, 44]; and relative fitness, which analyzes performance over multiple phases of bacterial growth (i.e. lag, exponential, and stationary) using a head-to-head competition of strains in co-culture [30].

Our initial hypothesis was that Modern AM1 would outperform the Archival strain, owing to an increased likelihood of mutations in Modern AM1 that could facilitate adaptation to laboratory conditions. However, contrary to our expectations, we found that the Archival strain was both faster and fitter under most conditions tested. The Archival strain was considerably faster growing on the C₁ compounds methanol (42%) and methylamine (12%), as well as the multi-C substrate succinate (52%; Figure 2A and B). In head-to-head fitness competitions, the Archival strain showed roughly a 30% advantage across the substrates tested (Figure 2C), suggesting that analyses of growth rate and fitness are not entirely correlated. Nonetheless, these results show that the Archival strain outperforms Modern under most standard growth conditions, and suggest that the Modern lineage became slower and less fit during its time in the lab.

Modern outperforms the Archival AM1 on nutrient broth

We next sought to compare Modern and Archival AM1 using less traditional growth substrates. In contrast to “Hypho” minimal medium, nutrient broth (NB) is a rich medium composed of partially digested proteins (peptone), but its exact nutritional components are largely undefined. Today, NB is rarely used as a growth substrate for Methylobacterium, and is used almost exclusively to support growth in co-cultures with E. coli during conjugal matings for genetic manipulation. It is possible, however, that Modern AM1 was more frequently cultured on NB in the past and may have adapted to growth on this medium. To test this hypothesis, we assessed the growth and fitness of Modern and Archival AM1 cultured on NB.

We found over multiple different growth experiments that Modern AM1 consistently outperformed Archival on NB, although both strains grew more poorly than on Hypho medium. Modern displayed a clear growth advantage on NB in 48-well plates (Figure 3A), which could be traced in part to its relative insensitivity to perturbations arising during the later stages of NB growth. While monitoring the increase in OD₆₀₀ over time, we observed that NB cultures tended to slow down during the later stages of exponential growth, and that the Archival strain was hindered to a greater extent than Modern. This suggested that Modern may have adapted to yet unknown components of growth in NB; however, due to this decrease during late exponential phase, we were unable to accurately assess differences in specific growth rate, and sought instead to quantify performance using a head-to-head competition of Modern and Archival co-cultures.

The performance of Archival AM1 co-cultured with a fluorescently labeled Modern reference was monitored over the course of several days during growth in NB flasks. Co-cultures were sampled periodically to monitor changes in the ratio of nonfluorescent (Archival) to fluorescent (Modern) cells using flow cytometry. These results suggest that the Archival AM1 holds an early growth advantage in NB that quickly decreases from 1 to 4 days until the strains reach stationary phase (Figure 3B). As a control, a co-culture of nonfluorescent Modern mixed with the same fluorescent Modern reference showed little change over the course of the experiment. Using the ratio of nonfluorescent to fluorescent cells at the start and the end of one growth cycle, we can calculate the fitness of the Archival AM1 relative to Modern assuming a 64-fold (2⁶) increase in the population (see Methods), and find that Archival is only 61% as fit as Modern during growth on NB. Overall, the improved performance of Modern on NB is suggestive of adaption, either specifically to this medium, or to yet unknown but similar growth conditions.

Archival and Modern are similar in terms of long-term growth and storage

Another dimension in which AM1 may have adapted to life in the lab is through improved performance during long-term growth and storage. We compared the Modern and Archival strains grown for extended periods both in flasks and on plates by creating co-cultures of each strain with a fluorescently labeled Modern reference, and monitoring the change in fluorescent to non-fluorescent cells over time using flow cytometry. Growth in flasks was performed over 14 days with continual shaking using succinate as a growth substrate, while growth on methylamine plates was carried out at 30°C for 4 days, followed by up to 60 days at 4°C to simulate long-term storage.

Here, Modern and Archival AM1 were roughly equivalent in terms of growth and survival during stationary phase. In flasks, the ratio of unlabeled Archival cells to labeled Modern cells was steady for 12 days after the initial 2 days of exponential growth (Figure 4A), while the unlabeled Modern control remained unchanged for the duration of the experiment. Archival held a similar advantage during storage at 4°C on agar plates for up to 60 days (Figure 4B). Collectively these results suggest that, at least under these conditions tested, the extent to which laboratory domestication improved the long-term growth and survival of AM1 is limited, and that the major difference between these strains lies in decreased exponential phase growth in Modern under standard conditions, and slightly increased growth on NB.

Growth of AM1 is significantly hampered by selection for rifamycin resistance

To explain the decreased performance of Modern AM1 under most growth conditions, we returned to the genomic changes identified in this lineage. All but one of the 29 mutations arising in Modern AM1 can be attributed to unintended laboratory domestication; this single exception, however, was central to the development of Modern AM1. In 1984, Fulton and colleagues selected for rifamycin resistance (Rif^R) to facilitate genetic manipulations in AM1 using conjugative, tri-parental matings [45]. Across numerous systems, mutations conferring Rif^R most often occur in the beta subunit of RNA polymerase – encoded by rpoB – and typically give rise to fitness tradeoffs between survival in the presence of antibiotics versus decreased growth in their absence [46]. Compared to the Archival strain, Modern AM1 has two mutations to the rpoB locus (Table 1), one of which (Q521R) falls in a region that confers Rif^R in a variety of other organisms [47], while the effect of the other (Q1081R) is yet unknown. At both positions the amino acid state of the Archival strain is universally shared with other non-AM1 strains of M. extorquens, suggesting that these mutations arose exclusively in the Modern lineage. Thus, these mutations to rpoB, particularly Q521R, are strong candidates for decreased overall performance in Modern AM1: offering resistance in the presence of Rif but slower growth in its absence.

To explore the potential costs associated with antibiotic resistance, we recapitulated the evolution of Rif^R in independent replicate cultures of the Archival strain. Out of 36 independent populations grown to saturation, only 7 produced a small number of spontaneous, resistant colonies when plated on Rif agar plates. Each independent population was streaked to purity, analyzed in terms of growth rate on succinate (with no antibiotic), and sequenced along with the Modern and Archival strains at the rpoB locus.

Independent experiments recapitulating Rif^R in the Archival strain all selected for mutations to rpoB, and all resulted in decreased growth in the absence of antibiotic. Upon sequencing rpoB from the 7 newly evolved Archival isolates, we found that Rif^R was always associated with mutations to rpoB. One of these strains (CM4022) acquired the exact same nonsynonymous change that occurred in Modern evolution, corroborating our hypothesis that the Q521R mutation (from Archival to Modern) is causal in conferring Rif^R to Modern AM1. Indeed, all mutations observed fall within a region of rpoB that is commonly mutated to confer Rif^R across a variety bacteria [47].

The spectrum of rpoB mutations across Rif^R Archival isolates displayed highly variable effects on growth rate in the absence of antibiotic. Compared to their Archival ancestor, several Rif^R isolates show very little decrease in performance when grown on succinate, while other isolates slow to near Modern levels, and still others grow substantially worse than Modern (Figure 5B). Interestingly, the CM4022 isolate that perfectly recapitulated the change from Archival to Modern AM1 (Q521R) was slightly faster than Modern, suggesting perhaps that other mutations further hamper growth in the Modern lineage. We note, however, that a direct comparison of this strain is difficult given that many other mutations and mutational interactions were likely present in Modern AM1 during the original selection for Rif^R. Nevertheless, these results demonstrate that selection for Rif^R can substantially hinder growth of the Archival strain in the absence of antibiotic, and that this single researcher-imposed event – not laboratory domestication - is the major mechanism by which Modern AM1 became slower growing in the lab.

Bacterial strains & growth conditions

Strains relevant to this study were as follows. Our Modern AM1 strain was derived from a pink, “wildtype” M. extorquens AM1 (CM501) described elsewhere [23]. A sample of lyophilized Archival AM1 (renamed CM3944) was acquired from the National Center of Industrial Food and Marine Bacteria (NCIMB #9133, Aberdeen, Scotland), grown to saturation, and frozen at −80°C in 8% DMSO. To limit cell clumping and reduce noise in OD₆₀₀ measurements during analyses of growth rate, we utilized a strain of Modern AM1, CM2720, that lacks genes for cellulose biosynthesis [27] without a loss of growth or fitness. For competitions in co-culture, a fluorescently labeled Modern reference (CM1175) was constructed by placing the red fluorescent protein mCherry under control of a constitutive P _tac promoter at the katA locus [30]. Isolates from each of seven Archival populations selected for rifamycin resistance (described below) were numbered CM4020 through CM4026.

Standard growth conditions utilized a modified version of Hypho minimal medium consisting of: 100 mL phosphate salts solution (25.3 g of K₂HPO₄ plus 22.5 g Na₂HPO₄ in 1 L deionized water), 100 mL sulfate salts solution (5 g of (NH₄)₂SO₄ and 2 g of MgSO₄ • 7 H₂O in 1 L deionized water), 799 mL of deionized water, and 1 mL of trace metal solution [55]. All components were autoclaved separately before mixing under sterile conditions. Carbon sources added just prior to inoculation in liquid minimal media consisted of 20 mM methanol, 3.5 mM sodium succinate, or 20 mM methylamine hydrochloride. Growths in 48-well microtiter plates consisted of Hypho medium plus the appropriate carbon source to a volume of 640 μL. Agar plates consisted of growth medium plus either 125 mM succinate or 100 mM methylamine and were autoclaved with 1.6% w/v agar. Difco nutrient broth (Becton, Dickson, and Company, Franklin Lakes, NJ) was prepared according to the manufacturer’s guidelines.

All growth regimes consisted of three phases consisting of inoculation, acclimation, and experimentation growths. All strains were stored in vials at −80°C in 8% DMSO; growths were initiated by transferring 10 μL freezer stock into 10 mL of Hypho medium with methanol. Upon reaching stationary phase (~2 days at 30°C with shaking), cultures were transferred 1:64 into the appropriate medium and vessel to be tested, allowed to reach saturation in this acclimation phase, and diluted 1:64 again into fresh medium for the measured (experimental) growth.

Measurements of specific growth rate and relative fitness

The increase in OD₆₀₀ for strains grown in 48-well microtiter plates was measured using an automated, robotic culturing and monitoring system [27, 44]. The specific growth rate of cultures was calculated from the log-linear growth phase using custom-designed growth analysis software [27]. Growth rates reported for each strain and condition are the mean plus SEM calculated from triplicate biological replicates, unless otherwise noted. Exogenous cellulase enzyme from Aspergillus niger (Sigma-Aldrich, St. Louis, MO) was added to the medium to a final concentration of 0.1 mg/mL to further minimize cell clumping and facilitate more accurate measurements of OD₆₀₀ over time (SMC, unpublished).

Fitness measurements – which encapsulate the lag, exponential, and stationary phases of growth – were performed using a head-to-head competition of strains grown in co-culture [30]. Modern and Archival AM1 were competed against a fluorescently labeled Modern strain (CM1175) expressing mCherry [30]. Co-cultures were prepared by mixing test strains with the fluorescent Modern reference in roughly equal optical densities. A sample of this co-culture prior to competition was diluted in 8% DMSO and stored at −80°C; the rest was diluted 1:64 into 640 μL Hypho medium plus carbon in 48-well microtiter plates and incubated with shaking at 30°C for 1 growth cycle. A sample of co-culture after competition was frozen for later analysis using flow cytometry.

Analysis of long-term growth and storage

To assess the ability of strains to withstand extended periods of stationary phase growth in flasks, Modern and Archival AM1 were mixed in co-culture with the fluorescent Modern reference, and the ratio of labeled to unlabeled cells was measured periodically using flow cytometry. Flasks possessing succinate were sealed to limit evaporation, and at the conclusion of the experiment co-cultures were streaked onto nutrient agar plates to check for contamination. Growth and survival on plates was measured using a similar experimental design: co-cultures were spread onto Hypho plus methylamine plates, grown for 4 days at 30°C, and then stored at 4°C for up to 60 days with periodic sampling.

Recapitulation of Rif resistance in AM1

The re-evolution of Rif resistance in Archival AM1 was performed using 36 replicate lineages grown from single colonies in 48-well plates. After 2 days growth in liquid Hypho medium plus succinate, cultures were plated without dilution onto Hypho agar plus succinate plates containing 50 μg/mL Rif. After 5 days of growth, Rif^R colonies were obtained from 7/36 cultures and streaked twice more to ensure the purity of clonal isolates (CM4020-4026). PCR amplification plus sequencing was used to assess mutations to the rpoB locus and the growth rate of strains on succinate was determined as described above.

Whole-genome sequencing of Archival AM1

Preparation of Archival genomic DNA was performed as described in [57]. Briefly, an Archival cell pellet was lysed using bead beating (MP Biomedicals, Solon, OH) and digested using heat, SDS and Triton-X100 detergents, proteinase K, and RNase A. Cell debris was precipitated with CTAB/NaCl, removed using a phenol/chloroform/isoamyl alcohol extraction, and genomic DNA recovered using an isopropanol precipitation. Illumina HiSeq2000 sequencing was performed out-of-house using GENEWIZ, Inc (South Plainfield, NJ). A total of 23,660,656 reads were generated, and 98.2% of these were mapped against the Modern AM1 reference [21] using breseq v0.21 [58] with Bowtie 2 version 2.0.0-beta7 [59]. To identify DNA lost in Modern AM1 evolution, reads that were unmapped using breseq were analyzed using ABySS v1.3.4-kmer-96 [60] for de novo assembly of contigs,. A full de novo assembly of the Archival genome was also performed in ABySS for comparison against breseq. Contigs from ABySS were further assembled and curated using contig assembly and BLAST in Geneious [61]. To determine whether mutational differences occurred in either the Modern or Archival lineage, we compared the mutational state (Archival or Modern) at each site to other previously sequenced strains of M. extorquens using both BLAST and analyses of conserved gene synteny using the Methyloscope project in MicroScope [62]. Mutations were identified as “derived” in the Modern lineage if the Archival state was consistently shared across related strains, to the exclusion of Modern AM1. Related strains included M. extorquens PA1 (GenBank Assembly ID: GCA_000018845.1), CM4 (GCA_000021845.1), DM4 (GCA_000083545.1), and DSM 13060 (GCA_000243435.2). For all loci but two, the Archival state was universally conserved; the only exceptions were for genes with little to no homology outside of Modern and Archival AM1.