Phenogenomic Characterization of a Newly Domesticated and Novel Species from the Genus Verrucosispora

ABSTRACT

The concept of bacterial dark matter stems from our inability to culture most microbes and represents a fundamental gap in our knowledge of microbial diversity. Here, we present the domestication of such an organism: a previously uncultured, novel species from the rare Actinomycetes genus Verrucosispora. Although initial recovery took >4 months, isolation of phenotypically distinct, domesticated generations occurred within weeks. Two isolates were subjected to phenogenomic analyses, revealing domestication correlated with enhanced growth rates in nutrient-rich media but diminished capacity to metabolize diverse amino acids. This is seemingly mediated by genomic atrophy through a mixed approach of pseudogenization and reversion of pseudogenization of amino acid metabolism genes. Conversely, later generational strains had enhanced spore germination rates, potentially through the reversion of a sporulation-associated kinase from pseudogene to true gene status. We observed that our most wild-type isolate had the greatest potential for antibacterial activity, which correlated with extensive mutational attrition of biosynthetic gene clusters in domesticated strains. Comparative analyses revealed wholesale genomic reordering in strains, with widespread single nucleotide polymorphism, indel, and pseudogene-impactful mutations observed. We hypothesize that domestication of this previously unculturable organism resulted from the shedding of genomic flexibility required for life in a dynamic marine environment, parsing out genetic redundancy to allow for a newfound cultivable amenability.

IMPORTANCE The majority of environmental bacteria cannot be cultured within the laboratory. Understanding why only certain environmental isolates can be recovered is key to unlocking the abundant microbial dark matter that is widespread on our planet. In this study, we present not only the culturing but domestication of just such an organism. Although initial recovery took >4 months, we were able to isolate distinct, subpassaged offspring from the originating colony within mere weeks. A phenotypic and genotypic analysis of our generational strains revealed that adaptation to life in the lab occurred as a result of wholesale mutational changes. These permitted an enhanced ability for growth in nutrient rich media but came at the expense of reduced genomic flexibility. We suggest that without dynamic natural environmental stressors our domesticated strains effectively underwent genomic atrophy as they adapted to static conditions experienced in the laboratory.

INTRODUCTION

The Great Plate Count Anomaly has confounded microbiologists since the 1800s and the introduction of solid growth media (1). The first recorded instance was in 1932 by Razumov, who was unable to equate viable plate counts with the microscopic enumeration of environmental isolates (2). Advances in next-generation sequencing and metagenomics in the modern era have only confirmed this culturable discrepancy, leaving the longest unresolved question in microbiology unanswered. Recent analyses of genomic data produce a bleak picture, highlighting that the current rate of characterization at 600 new species per year will require more than 1,000 years to complete (3). The knowledge from this microbial “dark matter” has untold and innumerable consequences. Unlike the hypothetical understanding of dark matter in astrophysics, concrete microbial dark matter discoveries demonstrate that the uncultured majority is a dominant factor within the universe (4). Technological advancements reveal that this uncharacterized diversity also encodes the potential to produce countless bioactive compounds that must be investigated (5), both for bacterial taxonomy and potentially to restock our chemical arsenal in a postantibiotic age (6).

While the postantibiotic age is a relatively new threat, humans have been using bacterial natural products since ancient Mesopotamia (7). More recently, however, we have been able to optimize bacterial biosynthetic potential for myriad applications. This domestication, or taming, is analogous to domestication of farm animals and crops. Human selection has caused “wild-type” organisms to evolve away from their highly variable and complex natural environments: without dynamic natural environmental stressors necessitating fitness traits, domesticated organisms effectively undergo genomic atrophy as they adapt to static, controlled conditions experienced in the laboratory (8).

During our explorations of cryptic microbial dark matter, we have previously reported recovery and whole-genome sequence of Verrucosispora sp. strain CWR15 isolated from a Gulf of Mexico sponge (9). Although markedly understudied, the Verrucosispora genus has attracted some attention due to the demonstratable bioactive potential of its secondary metabolites. Since the taxon was established in 1998, there have been discoveries of novel antibiotic (10), antitumor (11), anti-influenza A (12), and anti-HIV (13, 14) activities. Despite this, the overwhelmingly marine-derived genus is among the “rare actinomycetes” recalcitrant to laboratory culturing and domestication (15).

Here, we report the domestication of Verrucosispora sp. CWR15 by sequencing laboratory subpassaged, phenotypically distinct generations and their functional characterization. Genomic exploration reveals laboratory-mediated genetic changes in key functions such as metabolism, environmental information processing, antibacterial gene cluster activity, and antimicrobial resistances. This results in a domesticated, previously unculturable organism that appears to be a novel species, here named Verrucosispora sioxanthis. This domestication and species determination serve as an example of how to decrease the microbial dark matter knowledge gap with the aim of restocking the antibiotic medicine cupboard in a postantibiotic age.

RESULTS

Cultivation and domestication of a novel isolate of the Verrucosispora genus.

We have previously detailed the isolation and genome sequencing of a novel isolate from the genus Verrucosispora, Verrucosispora sp. strain CWR15, isolated from a Gulf of Mexico sponge (9). Given the understudied nature of the Verrucosispora genus, and the rarity with which they are cultured, we present here a detailed phenogenomic exploration of the generational domestication of this organism. Following initial isolation of Verrucosispora sp. CWR15, we observed that its continued culturing within the lab proved difficult. Very often the organism failed to disperse in broth, and medium preparations would evaporate or desiccate during extended (7+ days) incubations. To explore this, we initially tried different standard laboratory media (tryptic soy broth [TSB] and Luria-Bertani broth [LB]) without the addition of the environmental sponge extract that Verrucosispora sp. CWR15 was initially cultured on (Fig. 1A and B). In so doing, we noted minimal growth in TSB and LB alone that was only improved upon inclusion of both sucrose and glass beads, alongside extended culture times. Toward the latter point, sucrose is known to incite faster/nonaggregative growth in similar hypha-growing organisms such as Streptomyces spp. (16), while sterile glass beads physically disrupt biomass in liquid media and mechanically prevent aggregation. We next tried more specialized media typically used to culture organisms such as Verrucosispora spp. (17–19). Here, we found that AMM and ISP-2 (Fig. 1C and D) media were much better suited for rapid biomass accumulation, as evidenced by the 3- to 5-day increase in optical density for AMM, and for ISP-2 after the 6-day time point. Unfortunately, although these media produced more robust growth, they are not amenable to downstream experimentation due to component solubility and instability, particularly during long growth windows. Indeed, such instability was evidenced by the drastic and sometimes unpredictable growth patterns observed. Thus, all downstream experimentation was performed with TSB plus sucrose and beads, yielding reproducible and steady growth.

FIG 1.

Verrucosispora growth optimized with chemical and mechanical supplementation. The isolate was cultured with sterile glass beads (dotted lines, circles) or without (solid lines, squares) in four different growth media—TSB (A), LB (B), AMM (C), and ISP-2 (D)—with 20% sucrose solution (unfilled) or without supplementation (darkened). The data are presented as averages of five technical replicates, with error bars indicating ± the standard errors of the mean (SEM).

Serial subculturing yields generationally distinct isolates with disparate growth rates.

Using supplemented TSB, we were able to subpassage Verrucosispora sp. CWR15. This domestication, via gradual introduction to laboratory culturing, yielded visually distinct generations over time that grew to different biomasses with various efficiencies. Strikingly, the first (G1), third (G3), and fifth (G5) generations were notable as being the most phenotypically distinct. Initial differences were observed in dispersal, biomass, and rate of growth between each generation (data not shown). This is remarkable since (i) no record exists of this rarely cultured taxon displaying such laboratory attenuation, (ii) such drastic phenotypic changes are typically described for bacteria spanning years not days (20), and (iii) the elucidation of laboratory-attenuation mechanisms could provide the basis for understanding why this rarely cultured taxon resists traditional culturing. To avoid complications from colony aggregation, we first measured growth rates of our generational isolates using a modified methylene blue absorption protocol. We observed that the three generations desorb methylene blue at different concentrations based on their biomass (Fig. 2). Although G1 may follow the same overall trends as the other isolates, it never accumulates the same biomass as its offspring generations. Indeed, even after 4 days, G1 has significantly less biomass than G3 and G5. In contrast, G3 has the most biomass at all time points, although G3 and G5 eventually converge. Together, these initial growth trends suggest that the most “wild-type” organism, G1, struggles to achieve the biomass of the more lab-attenuated G3 and G5. Such growth reticence of earlier generations likely contributes to the taxon’s culturing infrequency.

FIG 2.

Increased domestication increases biomass. The growth of three generational isolates of the Verrucosispora isolate (1, blue circles; 3, green squares; 5, red triangles) was monitored by a modified methylene blue absorbance protocol. The data are presented as averages of three technical replicates ± the SEM.

Whole-genome sequencing of generational isolates.

With generationally distinct strains in hand, we next performed whole-genome sequencing on our two most phenotypically distinct isolates (G1 and G3). This was done to shed light on the rapid domestication observed and to glean insight as to why this taxon resists culturing. Although we have previously published a hybridized Nanopore and Illumina genome for H-G3 (9); here, we performed IonTorrent-based sequencing for G1 and G3 for ease of comparison. These were individually remapped to the hybridized G3 (H-G3) genome to provide standardized generational comparisons (Fig. 3). Since the published H-G3 genome consisted of 35 contigs in no order, comparable universal position numbers were derived using progressiveMauve to align contigs to the single contig scaffold reference, Verrucosispora CNZ293. To account for differences between G1 and G3 genomes, all analyses examined only genes present within both. While G1 has a total of 5,018 coding sequences (CDS) annotated, G3 only has 4,762 CDS. Among these, 4,420 genes were present in both strains (88% of G1 CDS), presenting only a minor limitation to these analyses. Generational mapping coverage depths and statistics are shown in Table S1 in the supplemental material, while the bioinformatics pipeline used here is presented in Fig. S1.

FIG 3.

Circos plot summary of Verrucosispora sioxanthis domestication-induced genomic changes. Genomic map identifying genomic positions of domestication interest. From outermost ring to innermost: contig-containing karyotype (thick black line), uncategorized KEGG genes (gray), KEGG-designated human diseases/antibiotic resistance genes (blue), KEGG-designated cellular processing genes (green), KEGG-designated environmental information processing genes (yellow), KEGG-designated genetic information processing genes (orange), KEGG-designated metabolism genes (red), pseudogene (purple) or nonpseudogene (yellow-orange) status, indel frequency histogram (magenta; scale, 0 to 20%), SNP frequency histogram (blue; scale, 0 to 1%), and coverage histogram (black; scale, 0 to 100× depth).

Domestication results in SNPs, indels, and widespread pseudogene alterations throughout the genome.

To explore domestication mutations, we subtracted Mauve-identified single nucleotide polymorphisms (SNPs) from the CLUSTAL-identified mutations, allowing indel identification. In so doing, it appears that the majority of genetic changes between our isolates are not due to SNPs (285/4,420 genes, ∼6.4%) but instead are indels (2,570/4,420 genes, ∼58.1%). When calculating mutation frequencies per gene (accounting for various gene lengths), we noted the SNP frequencies ranged from 0 to 3.5%, while the indel frequencies ranged from 0 to 54%. To link KEGG ontological groups with CDS via BLASTn, we next pooled G1 and G3 CDS and aligned them with the genomes of Verrucosispora maris (now Micromonospora maris) and Verrucosispora NA02020. While these references only confer KEGG ontology to 1,181/4,420 G1+G3 genes (∼26.7%), it is not due to a lack of homology. The average nucleotide identity (ANI) values are 84.52% for Verrucosispora/Micromonospora maris and 84.91% for Verrucosispora NA02020, indicating similarity. Indeed, only 6 total CDS were not matched in the reciprocal BLASTn searches. The most significant contribution to the lack of KEGG ontology is that ∼40 and ∼41% of Verrucosispora/Micromonospora maris and Verrucosispora NA02020 CDS, respectively, are not associated with a KEGG ontological group. Another point to note is that indels caused significant generational DNA dissimilarity, as determined via CLUSTAL Omega. Here, indel-identified genes ranged from 46 to 99.97% DNA similarity: 1,778/4,420 genes (∼40.2%) contained ≤ 10 indels, 546/4,420 genes (∼12.4%) contained 11 to 50 indels, 210/4,420 genes (∼4.8%) contained 51 to 199 indels, and 36/4,420 genes (∼0.8%) contained 201 to 1611 indels. Overall, only 271/4,420 genes (∼6.1%) contained both SNP and indel mutations.

Since pseudogenes present within the unsequenced originating sponge strain cannot be identified, the presence or absence of extraneous stop codons in G1 cannot be determined. Instead, we explored the change of pseudogene status from G1 to G3 in the context of laboratory attenuation. To do this, we defined a pseudogene as any protein sequence with a predicted early stop codon prior to the end of a given annotation. This includes the spectrum of possibilities, ranging from protein sequences with major truncations—those affecting predicted active domains—and minor truncations, which are not predicted to drastically affect function. Upon analysis, we observed that 598/4,420 genes (13.5%) changed pseudogene status from G1 to G3. Of these genes, 191/598 genes (∼31.9%) changed from pseudogene status in G1 to nonpseudogene status in G3.

Pseudogene alteration in amino acid metabolism genes results in measurable phenotypic outcome.

The largest KEGG ontological group within the references (and therefore G1 and G3) belongs to metabolism, due to the overrepresentation of metabolism within available annotations. Strikingly, numerous amino acid metabolism genes were identified as undergoing mutations in our strains that changed pseudogene status (see Table S2). These included genes that specify proteins required for the utilizations of arginine, glutamine, proline, and serine. G1 contained six nonpseudogenes and three pseudogenes, while the G3 pseudogene profile flipped to majority pseudogene status (six pseudogenes and three nonpseudogenes); all changes resulted from indels rather than SNPs. To determine whether these events had phenotypic outcomes, the amino acid utilization profile of G1 and G3 were evaluated. In addition, although G5 was not subject to whole-genome analysis, it was also included to provide context and insight to continued domestication. We determined that G1 grew under all tested conditions, including basal/nitrogen-deficient media (Fig. 4). In comparison, G3 and G5 were incapable of growth without amino acid supplementation and failed to demonstrate G1-comparable biomass even in the presence of diverse amino acids. Of note, although G1 grew under all conditions, the preferred nitrogen source appeared to be arginine, which also elicited growth for the other generations as well, although not to the same extent. G3 grew on a larger variety of sources (all but glycine) compared to G5, which was only able to grow on arginine, glutamic acid, and glycine.

FIG 4.

Increased domestication results in an impaired ability to metabolize diverse amino acids sources. The three generations were grown in nitrogen-deficient medium that was supplemented with the amino acids (0.1% [wt/vol]) noted. Cultures were plated in biological duplicate within two technical, replicate plates, and the CFU were counted. The data are shown as box-and-whisker plots with bars denoting the minimum and maximum spread of replicate values (n = 4).

An interesting consideration for these findings, and our study of domestication in general, is whether the genomic changes between G1 and G3 could potentially be considered reversible. Although we are unable to examine whole-genome data for G5, we were able to generate sequences for two amino acid metabolism genes (ENC19_09540 and PdxT-homologue) from this strain. Table S3 contains the comparative DNA similarities, indel frequencies, and number of extra stop codons for these genes between G1:G3, G3:G5, and G1:G5. Upon analysis, we note that the G3:G5 sequences for both genes are identical, indicating both genes accrued stable mutations as G3 arose from G1. When reviewing mutations in the context of active domains, we observe that for ENC19_09540, G1 has four early stop codons, three of which precede the proposed enzymatic component of the protein (see Fig. S2). Conversely, all four stop codons are absent in G3/G5, indicating the reading frame has been restored in this gene, and thus the product produced is almost certainly active and functional. In the context of PdxT, we find that the predicted active domain of this protein is somewhat shorter in G1; however, mutations in G3 and G5 result in an extended and native active site that restores a missing enzymatic residue. Conversely, G3/G5 do have an early termination codon that is not present in G1; however, this location is very close to the end of the gene, beyond that which codes for the active site, and is thus not expected to impact function. Therefore, at least in the context of these analyses, it would appear that domestication/evolution of our Verrucosispora isolate can be considered linear (from G1>G3>G5), rather than something that switches back and forth in a reversible manner.

Domestication results in enhanced spore germination rates.

As domesticated generations accumulate greater biomass independent of amino acid metabolism, we next focused on growth from spore solutions. While the underannotated nature of the taxon is a limitation, several sporulation-associated genes were identified. One in particular, cotH, a kinase regulator implicated in spore coat assembly and germination (21), notably underwent domestication-related pseudogene alteration. Strikingly, this gene had only 80.95% nucleotide similarity between G1 and G3 due to ∼19.0% indels that cause change from pseudogene to nonpseudogene upon further domestication (see Table S2). The G1 sequence contains an early stop codon prior to the predicted CotH PFAM domain (see Fig. S2), likely inactivating protein function. However, upon passaging to G3, no such early stop codons are now present in cotH. To explore this observation further, spore germination assays were performed with each generational strain (Fig. 5). We observed that G1 experienced delayed and minimal germination, and subsequent growth, as visualized by a single suspended orange colony. In contrast, G3 and G5 germinate and begin rapidly growing at around 14 days, with optical densities far exceeding those of the essentially dormant G1. It is noteworthy that domestication-related indel accumulation within cotH results in G3 encoding a complete, nonpseudogene protein. This alteration could be indicative of a broad selection pressure placed on sporulation and germination genes, as functional CotH is required for efficient germination of Bacillus subtilis spores (21). Furthermore, CotH is required for outer spore coat structure and integrity (22). A G1 cotH pseudogene, particularly one missing the CotH-associated PFAM domain, could thus contribute to decreased heat resistance and subsequent minimal growth of heat-treated spores.

FIG 5.

Domesticated generational strains display increased germination rates and subsequent growth efficiency. (A) Mean OD₆₀₀ values for each generation are plotted over time as the inoculated spores initially germinate and then continue to grow, marked by days. Each data point is the average of three independently repeated experiments with six technical replicates (n = 18), with 95% confidence interval error bars. (B) Endpoint photos of the three generations (1, blue circles; 3, green squares; 5, red triangles). No contaminants were observed, and all growth was determined to originate from the heat-treated spore solutions since any contaminating planktonic cells would rapidly overgrow and present as medium turbidity.

As with our study of metabolic genes, we also considered whether the changes observed in cotH sequence between G1 and G3 were potentially reversible. Interestingly, in this scenario, G1:G5 have the greatest DNA similarity, compared to G1:G3 or G3:G5, and the lowest indel frequency. This would tend to indicate some level of reversion toward G1 sequence for this gene as strain G3 evolves to G5. Importantly, however, the location of stop codons is markedly different between G1 and G5. While these are present in G1 before the predicted active domain of CotH, the four that occur in the G5 sequence are after this domain (see Fig. S2), indicating that they would most likely not cause dysfunction of the protein. As such, although there appears to be some tendency toward sequence reversion for this gene across our strains, this ultimately does not appear to impact protein function, and thus is likely not driven by a selection pressure to revert to G1 status.

Generational strains display differential aggregation phenotypes.

Next, to explore macroscopic aggregative growth phenotypes, fluorescence microscopy with membrane and cell wall staining was performed (Fig. 6). We observed that G1 displayed a compact, nondiffuse phenotype that was not present within the more filamentous, elongated G3. G5 was observed to have a phenotype similar to G1, though with more filaments branching from core aggregated cells. We thereby classified G5 to have a phenotype averaging both G1 and G3 dispersals. Overall, these images suggest that G1 growth is more aggregative than the more domesticated generations. This could explain the limited biomass accumulation of this strain, as well as its general failure to disperse. Indeed, the overlapping filaments demonstrate a propensity to grow inward rather than outward, causing eventual growth limitation and subsequently decreased biomass.

FIG 6.

Fluorescence microscopy reveals mycelial growth patterns of Verrucosispora sioxanthis. Fluorescence micrographs of cells from generations 1, 3, and 5 of V. sioxanthis. Left to right: columns show cells with FM4-64 membrane stain (red), BODIPY FL cell wall stain (green), overlay, and differential interference contrast (DIC). Scale bar, 5 μm.

Domestication limits the antibacterial capacity of Verrucosispora sp. CWR15.

Given that the genus has contributed novel chemical structures in the past, we next investigated the antibacterial potential of our isolates. We began by analyzing biosynthetic gene clusters (BGCs) using antiSMASH for H-G3, identifying 18 clusters, including those for NRPS, T1PKS/T2PKS/T3PKS, siderophores, and bacteriocins. This low number is likely a result of the limited taxonomic investigations and annotations of the genus. Indeed, ∼72% of the BGCs do not exceed >50% similarity within available sequences, and five clusters returned no significant similarity scores at all (see Table S4). These five included syntheses of a lasso peptide, NAAGN, two separate terpenes, and a class I lanthipeptide.

To explore antimicrobial potential, G1 was grown in a variety of media prior to crude extraction of secondary metabolites. These extracts were assayed against a panel of ESKAPE pathogens (see Table S5) at a concentration of 250 μg/ml (Fig. 7A). Differential antibacterial activities were observed between culture conditions, with ISP-2, AMM, and 2YT demonstrating minimal inhibitory effects. TSB conditions, however, elicited profound antibacterial activity against Enterococcus faecium. This inhibition (∼91%) is notable for a crude extract, since no fractionation or purification was undertaken to strengthen activity. Next, all three generations were examined for antibacterial production in TSB. While all showed activity, G1 exhibited the greatest anti-E. faecium effects (Fig. 7B), readily exceeding that of G3 (∼80%) and G5 (∼67%). This finding is fascinating since we demonstrate that the most wild-type-like strain (G1) has the greatest capacity for antibacterial effects, a phenotype that is lost with increasing laboratory domestication. Although these observations are outside the boundaries of statistical significance, repeated assays maintained the trend of greater anti-E. faecium activity within G1 extracts. Such a finding has significant potential ramifications for environmental-microbe natural products drug discovery. Specifically, although domestication allows for easier manipulation and increased biomass/growth efficiency, excessive domestication can also disrupt biosynthetic pathways and prevent identification of already-cryptic genes through genome-wide alterations. This presents a challenge to discovery efforts and suggests that one must take a balanced approach to domestication, taking care to not extensively subpassage isolates, lest biosynthetic potential be negatively impacted.

FIG 7.

Biosynthetic potential and anti-ESKAPE activity. Generation 1 was grown for 3 weeks in 4 fermentation broths (International Streptomyces Project 2 [ISP-2], tryptic soy broth [TSB], AMM, and 2× yeast extract tryptone [2YT]) prior to crude extract assay against ESKAPE panel. (A) The percent inhibition was determined by spectrophotometric comparison to a DMSO-treated control of two repeat experiments with pathogen biological triplicates at 250 μg/ml. (B) Upon TSB-specific activity, the three generations were examined for the most active generation. (C) The generation 1 TSB fermentation crude extract was chosen for dose dependence against Enterococcus faecium in a pathogen biological triplicate of two repeat experiments, and the percent inhibition was compared to a no-drug control. Results are presented as means ± the SEM.

The PGM G1 and G3 genomes were uploaded to antiSMASH for investigation of the differential generational bioactivity. The passage of G1 to G3 resulted in both SNPs and indels in BGCs; however, as with the rest of the genome, the majority of the mutations were indels. These mutations contributed to antiSMASH predicting three clusters (producing xanthoferrin, formicamycins A to M, and isorenieratene) in G1 that had no equivalent in G3 (see Table S6 and Table S7). Conversely, 7/18 BGCs noted above for H-G3 were absent from G1, again due to domestication mutations (see Table S6).

We next undertook a dose-response study with G1 (Fig. 7C), noting that the observed anti-E. faecium activity persists even as the tested concentration decreases. The inhibitory effects of G1 crude extracts continued at >50% until only after they were diluted to a concentration of 100 μg/ml. Such activity is indeed noteworthy, since, to our knowledge, no anti-E. faecium activity has ever been reported within Verrucosispora spp.

Verrucosispora sp. CWR15 is a novel species.

As a final step, using data from our genomic analyses, we created a 16S rRNA phylogenetic tree with the H-G3 genome (Fig. 8) comparing our novel isolate to other members of the genus and the closely related Micromonospora. This revealed that Verrucosispora sp. CWR15 is most similar to Micromonospora sediminimaris, previously recognized as Verrucosispora sediminis. Due to the taxon undergoing rearrangement with Micromonospora, we also predicted the digital DNA-DNA hybridization (dDDH) to determine novelty of the organism (see Table S8). Although the taxonomy of the genus may be in flux as Verrucosispora spp. are increasingly being acknowledged as heterotypic synonyms to Micromonospora, the highest dDDH values of 63.9, 46.3, and 61.1% (d₀, d₄, and d₆, respectively) do not exceed the accepted threshold of 70%, indicating that the isolate is taxonomically distinct. Furthermore, the 92.04% ANI and 94.1% average amino acid identity (AAI) to Micromonospora sediminimaris fall within the threshold for species determination. Collectively, these genomic and in silico assessments support the notion that our isolate is a novel species, which we here term Verrucosispora sioxanthis based on the sioxanthin pigment that bestows the organism with its vibrant orange appearance.

FIG 8.

Taxonomy of Verrucosispora sioxanthis. A tree inferred using FastME 2.1.6.1 from GBDP distances was calculated from 16S rRNA gene sequences. The branch lengths are scaled in terms of GBDP distance formula at day 5. Numbers above branches indicate GBDP pseudo-bootstrap support values of >60% from 100 replications, with an average branch support of 66.9%. The tree is rooted at the midpoint. The tree was resolved and edited using Interactive Tree Of Life (iTOL).

Given the myriad genome-wide alterations between G1 and G3, we next sought to determine whether the accrued changes were so extensive that these two isolates could themselves be considered separate species from each other. Standard comparisons of the 16S rRNA gene through BLASTn similarity (percent identity) show that the generations are within the same species thresholds (see Table S9). This holds true even when considering the G5 16S rRNA sequence. To ensure our conclusions were robust, we submitted the G1 and G3 whole-genome data to dDDH, ANI, AAI, and percentage of conserved protein (POCP) analyses. As seen in Table S9, dDDH (>70%) (23), ANI (>95%) (24), AAI (>95%) (25), and POCP (>63%) (26) values all fall within the thresholds to still be considered the same species.

DISCUSSION

Bacterial domestication, although underexplored, is not unheard of. It is known that endosymbionts accrue significant pseudogenes upon liberation from their eukaryotic hosts (27–30). Indeed, pseudogene numbers within transitioning endosymbiotic bacteria can reach into the thousands, often accounting for over half of the genome (29, 30). It is also known that gene inactivation through pseudogene accumulation is correlated with reduced selective pressures on redundant genes (31). This is synonymous with laboratory attenuation in our study. As the previously sponge-associated Verrucosispora sioxanthis was passaged through high-nutrient, stable laboratory conditions, the organism no longer required the genetic flexibility that allowed it to respond to the host, environment, and surrounding sponge microbiome. Thus, we propose that pseudogenization effectively allowed the shedding of its flexible genome to adapt to a low-stress laboratory environment, parsing out genetic redundancy in an otherwise stable environment.

It is important to acknowledge that the only Verrucosispora colony recovered from our sponges was on low-nutrient media supplemented with originating sponge extract. While extended incubations are common for recovering organisms from marine or otherwise oligotrophic environs (32), our incubation time (>4 months) is on the more extreme end of the spectrum. We suggest that this organism needed both extended incubation and trace host sponge factors to accumulate biomass as it approached the laboratory-attenuated genetic profile of G1. We observed that passaging, or domesticating, the organism allowed for quicker growth following initial adaptation to the laboratory. Although each generation was given exactly 7 days between passages while maintaining all incubation parameters, the generational endpoint biomasses are incomparable. The only variable that could explain these discrepancies is thus the domestication mutations accumulated during laboratory adaptation.

The domestication documented here is marked by genome-wide alterations. The alterations occur both within annotated genes and those lacking functional annotation; this annotation limitation, and the large-scale genomic change, prevents a cohesive and specific mechanism of domestication. However, the extent of change in pseudogene status observed in our genomic comparisons was quite striking. This clearly indicates a selection pressure for activating/inactivating mutations in domesticated strains of Verrucosispora sioxanthis. This ability to either produce a fully functional protein or introduce early stop codons is in line with the genomic alterations present in former endosymbionts that shed their eukaryotic hosts. As facultative endosymbionts transition away from their hosts and host-specific selective pressures, genomic change through gene inactivation is observed, mainly through the usage of pseudogenes and truncated (and thus inactive) gene products (33).

To investigate the phenogenomic effects of domestication, biomass and metabolism assays were performed to supplement in silico study. Regarding amino acid metabolism, pseudogene-mediated flexibility seemingly contributed to the ability of Verrucosispora sioxanthis to adapt and flourish within nonnative conditions. The amino acid metabolism genes identified all underwent indel-induced pseudogene alterations in G3, resulting in less amino acid metabolic ability and an overall pseudogene-dominant profile. This is in direct contrast to G1 that had the greatest amino acid metabolic capacity and an overall nonpseudogene profile for those implicated genes. G1’s nonpseudogene profile for amino acid metabolism benefits the more wild-type organism; the native oligotrophic conditions of a marine endosymbiont require genetic flexibility to maintain nutrient requirements in a dynamic system. Without the selective pressure of nutrient limitations, lab-attenuated G3 underwent genomic atrophy that led to a decrease capacity for amino acid consumption. This is in line with the theory that genes not conferring fitness are more susceptible to inactivation or otherwise deletional mutations (34).

In addition to amino acid metabolism, we also investigated domestication-impacted secondary metabolism. Since metabolism was the most-affected ontological group, it would follow that both primary and secondary metabolic pathways would be impacted. Indeed, we demonstrate that domestication impacts the antibacterial activity of crude fermentation extracts. The anti-E. faecium activity observed decreased from G1 to G3 and G5. While we cannot sequence the originating sponge isolate, we posit that G1 is more genetically similar to the original sponge isolate than later laboratory-passaged generations. Such environmental isolates, particularly marine sponge endosymbionts, must rely upon their chemical defenses to compete and fend off predation (35), frequently producing some form of antibacterial activity. The domestication of such isolates can lead to genome atrophy, in which strains lose these defense mechanisms as they become laboratory attenuated (36). This atrophy within antibacterial potential is corroborated by antiSMASH analysis, where indel-driven mutations result in widely variable predicted BGC content between strains. Our findings should caution those within environmental-microbe antibacterial drug discovery: while domestication can provide a more laboratory-suited organism, care should be taken to minimize the genomic atrophy and evolution away from environmentally fit “wild-type” genetic profiles. As environmental isolates seemingly lose their chemical defense mechanisms required and maintained by the originating niche, investigators should focus their discovery efforts on minimally passaged strains to capitalize on chemical defenses that could confer antibacterial activity.

Notably, while less-domesticated generations retain a greater ability to metabolize amino acids, this does not directly contribute to growth rate as G1 was consistently slower at accumulating biomass. This could be explained, in part, by spore germination abilities. In G1, the cotH kinase regulator exists as a pseudogene which may result in impaired germination of it spores, a phenotype observed in our assays. In support of this, it is known that CotH influences spore germination in B. subtilis (37). In contrast, G3 accumulated indels reverting this gene to nonpseudogene status. A fully functional CotH could allow G3 the ability to germinate and propagate more rapidly as vegetative cells, ultimately explaining its less-spore-dense fluorescence images under microscopy.

While not every previously unculturable bacterium will behave as Verrucosispora sioxanthis, we propose this study as a model for examining genome-wide effects of domestication. The vast majority of organisms remain microbial dark matter and thus the mechanisms of domestication remain largely unknown. Taken together, the identification of SNPs, indels, and pseudogenes reveals the genetic impacts of domestication. Any of these variables alone would not allow substantial elucidation of the genomic changes required for domestication. Indeed, given that we observed SNPs accounting for 7% of genetic changes, and indels for 58%, mere genomic comparisons alone are insufficient. The coupling of genome-derived hypotheses with the phenotypic investigation described here demonstrates the holistic value of a combined comparison. Collectively, this generates new insight into the process of domesticating hard-to-tame environmental isolates that are recalcitrant to laboratory culture. We consider our finding of rapid and extensive mutation during domestication of V. sioxanthis not only fascinating but also as having potentially profound implications for genomic evolution. Whether such occurrences are unique to our organism, the Verrucosispora genus, or are a function of the wider Actinobacteria in general remains to be seen. Regardless, our work presents an important opportunity to utilize similar phenogenomic studies during domestication analysis in order to generate valuable insight into bacterial adaptation and genomic evolution.

MATERIALS AND METHODS

Serial passaging.

Three distinctive generations of Verrucosispora sioxanthis CWR15 were obtained via liquid subculturing of the single, initial recovered colony. From that initial starting colony, each generation was grown for 7 days in identical baffled flasks with 60 ml of tryptic soy broth (TSB; 30 g/liter) containing sucrose (20% [vol/vol] solution of filter sterilized 50% [wt/vol] sucrose) at 30°C at 210 rpm. Serial passaging was accomplished by withdrawing 1 ml of cultures and then inoculating them into fresh TSB/sucrose flasks every 7 days. Each generation was preserved at −80°C in 20% glycerol in Instant Ocean (IO; 36 g/liter) solution.

Spore solution generation.

Individual colonies were grown in 5 ml of TSB with sterile glass beads and sucrose for 7 to 14 days at 210 rpm and 30°C. Cultures were plated on tryptic soy agar (TSA) supplemented with sucrose and placed in a humidified incubator at 30°C for 7 to 14 days until lawns established. Spores were scraped from agar plates into filter-sterilized water (38). Cellular debris was filtered using autoclaved coffee filters, and solutions were incubated at 55°C for 10 min to kill vegetative cells prior to experimentation.

Conditional growth assay.

TSB, Luria-Bertani broth (LB; 10 g/liter tryptone, 5 g/liter yeast extract, 10 g/liter NaCl), AMM (10 g/liter starch, 4 g/liter yeast extract, 2 g/liter peptone, 36 g/liter IO), and International Streptomyces Project 2 medium (ISP-2; 10 g/liter malt extract, 4 g/liter glucose, 4 g/liter yeast extract, 36 g/liter IO) were used for these assays in 96-well plates sealed with parafilm and surrounded with a sterile moat of water to minimize evaporation over the 14-day+ experiment. Five replicates of each media (75 μl) were used with/without sucrose. Freshly processed spore solution was inoculated to each well to a final volume of 100 μl. An identical setup containing a UV-sterilized glass bead (0.5 mm) in each well was also used. Plates were incubated in a humidified shaker at 30°C and 210 rpm, with growth monitored by determining the OD₆₀₀ every 24 h using a Synergy 2 Microplate reader.

Methylene blue assay.

The absorption and desorption of germinating spore solutions was monitored as described previously (39). Five 24-well plates were inoculated with TSB and spore solution, along with three sterile glass beads (3 mm). Sealed plates were incubated in a humidified shaker at 30°C and 210 rpm. Methylene blue absorption was performed within the growth plate to limit transferal loss at 245 rpm. Complete methylene blue desorption employed two washes with 250 mM HCl. Desorption was determined by reading the OD₆₆₀ of supernatants. The exact concentrations of methylene blue absorbed and desorbed were determined using a previously established absorption coefficient (39).

Genomic DNA extraction and sequencing.

Genomic DNA was extracted from 7-day TSB/sucrose/bead cultures incubated as described above. Cultures were mechanically disrupted with sterile glass beads before undergoing phenol chloroform DNA extraction. Quality was monitored with a Thermo Scientific NanoDrop ND-1000 UV-Vis spectrophotometer. An Ion Torrent PGM Hi-Q View OT2 kit was used to generate 200 to 300-bp library fragments with ∼100 ng of input DNA. Ion Plus Fragment Library kit adapters and manufacturer’s protocols were used to ligate adapters to fragments, and Ion Xpress barcodes were attached to allow multiple samples on a single chip. Purification was performed using Agencourt AMPure XP reagent according to the manufacturer’s protocol, and library size selection was performed with an E-Gel SizeSelect agarose gel. Amplification was performed using an Ion Plus Fragment Library kit, purified with Agencourt AMPure XP reagent and then analyzed using an Agilent high-sensitivity DNA kit and an Agilent 2100 bioanalyzer instrument. Barcoded libraries for each generation were pooled, and an Ion OneTouch 2 instrument was used to prepare template-positive Ion PGM Hi-Q View ion sphere particles (ISPs). ISPs were enriched using an Ion OneTouch ES instrument prior to sequencing using an Ion 318 v2 chip (Ion Torrent).

In silico prediction of species novelty.

The hybrid assembly for generation 3 (H-G3, used for scaffolding purposes) has been described previously (9). Determination of dDDH and taxonomy was achieved by submitting the H-G3 genome through the Type (Strain) Genome Server (TYGS; https://tygs.dsmz.de/) (23). The determination of closely related type strains (40–43), pairwise comparison of genome sequences (43), phylogenetic inferences (44–46), and type-based species and subspecies clustering (47) are described in an earlier study (23). The DSMZ-generated phylogenetic tree was exported to the Interactive Tree Of Life (iTOL) for visualization (48). ANI (24) and AAI (25) of H-G3 was repeated here according to the taxonomic reorganization of Verrucosispora and Micromonospora.

Domestication elucidation bioinformatics.

For the genetic investigation of domestication, Ion Torrent PGM reads were trimmed and processed using “map reads to contigs” of the previously published H-G3 (9) within CLC Genomics Workbench (CLCGW, v20.0.2) using default parameters. The resulting contigs for each generation were aligned to Verrucosispora CNZ293 using progressiveMauve (49) (default settings, except enabling “use seed families” option) to generate a universal positional identifying number for the generations to be plotted to. The generational genomic accession numbers are available below under “Data availability.” The new genomes were aligned (G1:G3) in progressiveMauve using previously described parameters. The mapping coverage data exported from CLCGW generated a coverage histogram for G1. Biosynthetic gene clusters (BGCs) were identified using bacterial antiSMASH v6.0.0 (50) and annotated CDS were obtained from CLCGW. For G1-G3 comparisons, homologous genes were determined by reciprocal best BLASTn hit. Gene tracks were generated from reference genomes and gene annotations were extracted from each gene track using CLCGW. Extracted annotations were used in a reciprocal BLASTn (51, 52) with then-named Verrucosispora maris (currently Micromonospora maris) and Verrucosispora sp. NA02020 genome annotations as local databases, and vice versa, using default BLASTn settings with the flag “-max_target_seqs 1” to restrict results to the lowest E value for homologous determination of KEGG ontological groupings.

To explore the domestication impact, extracted CDS nucleotide (nt) and translated amino acid (aa) sequences from G1 and G3 genomes were aligned using Clustal Omega (v1.2.3) (53), default parameters except wrapping options were enabled to prevent wrapping (–wrap = 10,000 for nt, –wrap = 5,000 for aa). This was done in a pairwise manner to compare equivalent proteins, unified across generations by annotation locus tags.

G1 and G3 genome assemblies were submitted to antiSMASH v6.0.0 (50) with “relaxed” strictness and all extra features enabled to identify BGCs. The previously described SNP and indel analyses were applied to BGCs, with a modification to prevent Clustal Omega wrapping (–wrap = 1,000,000). The GC content and contig average mapping coverage depth were determined with CLCGW v21.0.3.

To explore the domestication impact on potential taxonomic effects, PFAM and active-site residue predictions were generated using PfamScan (54). The percentage of conserved proteins (POCP) (55) and comparative generation-specific AAI values (56) were determined as reported previously.

Fluorescence microscopy.

Fluorescence microscopy was performed as previously described (57). Cultures were grown in TSB with sucrose and glass beads (30°C and 210 rpm) for 7 days. Aliquots (10 μl) of each culture were diluted in 200 μl of PBS, and then 100 μl was mixed with BODIPY FL dye (1 μg/ml), followed by incubation for 10 min at room temperature. FM4-64 (1 μg/ml) was added, and 5 μl was spotted onto a glass-bottom dish (MatTek) and covered with a 1% agarose pad. Microscopy was performed using a GE Applied Precision DeltaVision Elite deconvolution fluorescence microscope equipped with a Photometrics CoolSnap HQ2 camera. Seventeen planes were acquired for each image, each 200 nm apart. Files were deconvolved using softWoRx software v6.5.2.

Amino acid utilization.

Methodology involving nitrogen-deficient basal medium was used as described previously (58). Briefly, 7-day liquid cultures were pelleted and washed to remove growth media. Duplicate inocula were prepared as described previously (58) for duplicate unsupplemented basal medium control plates and supplemented plates. Plates were sealed with parafilm and placed in a humidified incubator at 30°C for 21 days prior to scoring the results according to the literature. Colonies were counted and reported by comparison to basal media for baseline analyses. The sources of amino acids (0.1% [wt/vol] final concentration) were as follows: β-alanine (Calbiochem); l-arginine, proline, and serine (Acros Organics); l-glutamic acid (Sigma-Aldrich); and glycine (Fisher).

Secondary metabolite extraction.

Three biological replicates of Verrucosispora sioxanthis were grown in 60-ml baffled flasks in TSB, ISP-2, AMM, or 2YT (10 g/liter yeast extract, 16 g/liter tryptone, and 5 g/liter NaCl) (59) at 30°C and 210 rpm. After 21 days of incubation, 30 ml of ethyl acetate (EtOAc) was added to cultures before the samples were returned to the incubator for 1 h. The resulting extract was filtered through commercial-grade coffee filters to remove bacterial biomass before being left to settle for 24 h. The organic layers were transferred to preweighed scintillation vials and dried overnight via an airline. An additional 30 ml of EtOAc was added to original culture media for a second extraction 24 h later. After two extractions, the final dried extract was solvated in 100% dimethyl sulfoxide (DMSO) at 5 mg/ml.

Antimicrobial testing.

Overnight cultures of ESKAPE pathogens were utilized in antibacterial screens as previously described (60). Plates were incubated at 37°C and 245 rpm overnight. The inhibitory activity was determined based on OD₆₀₀ readings within a Synergy 2 microplate reader.

Data availability.

Verrucosispora sioxanthis generational genomic contigs are deposited in NCBI under the BioProject PRJNA734818. Generation 1 (G1) is available within the BioSample SAMN19546627, and generation 3 (G3) is available within SAMN19546628. The G1 Whole Genome Shotgun project has been deposited at DDBJ/ENA/GenBank under accession JAHMAD000000000, G3 has been deposited at DDBJ/ENA/GenBank under accession JAHMAC000000000.