A synthetic ‘essentialome’ for axenic culturing of ‘Candidatus Liberibacter asiaticus’

‘Candidatus Liberibacter asiaticus’ (CLas) is associated with the devastating citrus ‘greening’ disease. All attempts to achieve axenic growth and complete Koch’s postulates with CLas have failed to date, at best yielding complex cocultures with very low CLas titers detectable only by PCR. Reductive genome evolution has rendered all pathogenic ‘Ca. Liberibacter’ spp. deficient in multiple key biosynthetic, metabolic and structural pathways that are highly unlikely to be rescued in vitro by media supplementation alone. By contrast, Liberibacter crescens (Lcr) is axenically cultured and its genome is both syntenic and highly similar to CLas. Our objective is to achieve replicative axenic growth of CLas via addition of missing culturability-related Lcr genes. Bioinformatic analyses identified 405 unique ORFs in Lcr but missing (or truncated) in all 24 sequenced CLas strains. Site-directed mutagenesis confirmed and extended published EZ-Tn5 mutagenesis data, allowing elimination of 310 of these 405 genes as nonessential, leaving 95 experimentally validated Lcr genes as essential for CLas growth in axenic culture. Experimental conditions for conjugation of large GFP-expressing plasmids from Escherichia coli to Lcr were successfully established for the first time, providing a practical method for transfer of large groups of ‘essential’ Lcr genes to CLas.


Introduction
'Candidatus Liberibacter' spp. are a versatile group of fastidious, Gram-negative, psyllid-transmitted and phloem-limited α-Proteobacteria (order Rhizobiales). 'Ca. Liberibacter' spp. have a wide host range and are associated with several plant diseases of variable economic consequence, some high enough to warrant regulatory action. Huanglongbing (HLB) or citrus 'greening' is associated with 'Ca. L. asiaticus' and 'Ca. L. americanus' (CLas and CLam, both vectored by Asian citrus psyllid Diaphorina citri) and 'Ca. L. africanus' (vectored by African citrus psyllid Trioza erytreae) [1]. Aberrant assimilate partitioning and nutrient transport leads to progressive decline in productivity and eventual death of the HLBinfected trees. Liberibacter crescens (Lcr) strain BT-1 (NC_019907.1) was originally isolated from leaf sap of a diseased Babaco Mountain papaya (Carica stipulata×C. pubescens) and has been axenically cultured in vitro [2]. Lcr BT-1 has no known plant or insect host and serves as a surrogate gene expression host and model for functional genomics of CLas [3]. Comparative metagenomic analyses suggested stepwise reductive evolution of all 'Ca. Liberibacter' spp. including 24 fully sequenced CLas strains (all genomes ∼ 1.2 Mb) from a common ancestor following an initial split of Lcr (1.5 Mb genome) from other Rhizobiales [4]. All attempts to fulfill Koch's postulates or to culture CLas in axenic media have failed. Only inconsistent, transient and very low "titer" co-cultures [5][6][7][8][9] have been obtained, rendering their use impractical for most functional genomics purposes designed to understand host/pathogen/vector interactions and implement effective disease mitigation strategies.

HLB pathosystem (CLas/citrus host/psyllid vector) exists as a 'holobiont'
CLas survival and slight growth within complex hostderived microbial communities was detected by PCR in cocultures of Ishi-1 [6] and psy62 [7,8]. CLas strain A4 titers increased modestly in leaf disc explants incubated in the presence of glucose and the antibiotic amikacin under microaerobic conditions [10]. Li et al. [11] reported a 419-fold increase of CLas density without any corresponding increase in other citrus phloem-associated microflora in dodder (Cuscuta campestris) tendrils trained on CLas-infected citrus.
These observations indicate that the HLB pathosystem (CLas/citrus host/psyllid vector) exists as a 'holobiont' (host/vector with its endo-and extracellular microbiome) [12]. Metabolic and ecological interactions (mutualistic, synergistic or competitive) between the microbial community members within the HLB pathosystem are paramount for the survival of CLas with its highly reduced genome [4]. Genome reduction is a dominant mode of evolution in intracellular pathogenic/endosymbiotic bacteria, providing robust niche-specificity by virtue of increased metabolic efficiency and decreased transcriptional and regulatory costs associated with a streamlined genome [13][14][15].

Gene 'essentiality' is non-binary and context-specific
Peterson and Fraser [16] have argued against a universal, theoretically rigid 'minimal genome' or 'essentialome' design for achieving an autonomous self-replicating cellular unit based on gene conservation criteria across large phylogenetic distances. Mounting evidence suggests that the gene 'essentiality' concept is neither binary nor static and evolves [17,18] under specific environmental and contextual genomic constraints [19,20]. For instance, approximately one-third of the essential genes in E. coli are non-essential in Bacillus subtilis and vice versa [21]. Likewise, ~ 17% of genes considered essential in the budding yeast Saccharomyces cerevisiae are non-essential in the fission yeast Schizosaccharomyces pombe and ~ 27% of essential fission yeast genes are non-essential in the budding yeast [22]. Some 'essential' genes are dispensable in the context of another missing gene (or pathway) because the former might encode protective functions towards the (likely) toxic effects of the latter. For example, glyoxalase I (GloA) is a biologically fundamental and ubiquitously conserved enzyme for detoxification of methylglyoxal, a cytotoxic byproduct of glycolysis. However, absence of gloA is well tolerated in CLas because of transcriptional downregulation of glycolysis and subsequent reliance on scavenging ATP from host cells by virtue of an nttA-encoded ATP/ADP translocase present in the uncultured pathogenic 'Ca. Liberibacter' spp. [23].

Successful axenic culturing of CLas requires a synthetic 'essentialome'
Genomic and metabolic pathway comparisons between Lcr and all pathogenic 'Ca. Liberibacter' spp. revealed a trend for the reduction or complete absence of multiple biosynthetic pathways, metabolic enzymes and secretion systems consistent with their intracellular lifestyle [4]. Even though the genomes of Lcr and pathogenic 'Ca. Liberibacter' spp. are highly similar and microsyntenic [2,4], the core Liberibacter genomes share only 658 genes, and most of the species-specific genes encode hypothetical proteins. Bioinformatic analysis revealed that 37% of the functionally annotated genes in the Lcr genome are species-specific in comparison to only 17% in CLas and 9% in CLam [24].
Based on genome scale metabolic modeling and largescale gene 'essentiality' data sets across 79 bacterial and archaeal domains, the 'essentiality' patterns cluster together phylogenetically in silico as well as experimentally at the metabolic pathway level [25]. It is therefore axiomatic that the genes validated as 'essential' for Lcr growth in vitro [23,[26][27][28][29] but absent in CLas, are likely indispensable for maintaining CLas in replicative cultures. Under this premise, unique gene loci in Lcr were identified using a custom Perl script and reciprocal blasts implemented in the OrthoMCL software [30] at e-value cut offs of < 3e-30 and < 40% identity. Functional annotation by InterProScan [31] and Prokka [32] revealed 405 unique ORFs (including 104 hypothetical proteins) in Lcr that were missing in all sequenced strains of pathogenic Liberibacters (Additional File 1). Out of these, 120 ORFs (81 annotated and 39 hypothetical proteins) were completely absent (Additional File 2) whereas 286 ORFs (221 annotated and 65 hypothetical proteins) were truncated in the genomes of all sequenced CLas strains. Classification of the 302 annotated 'essential' genes of Lcr into Cluster of Orthologous Groups (COG) is presented in Fig. 1. Notably, a relatively high percentage of these were involved in membrane or envelope biogenesis and partitioning. Tan et al. [33] very recently reported a similar number (323) of COGs unique to Lcr using very different methodology.
The 'essentiality' of 405 genes for Lcr growth was validated by targeted site-directed marker interruption [3] and compared with previously published EZ-Tn5 transposon mutagenesis dataset [27]. EZ-Tn5 mutagenesis data were also manually examined to determine if the location of the Tn5 insertions within each ORF likely affected the expression of the predicted conserved domains in the ORF, either by polar effects or by direct disruption. Some Tn5 insertions occurring in the terminal 20% of the 3ʹ region of the target ORF and outside of known functional protein domains were considered not likely to produce a mutated phenotype [34] and were experimentally validated by multiple failed knockout attempts, as were observed for Lcr glyoxalase (gloA) [23] and Kdo2-lipid IVA lauroyltransferase (lpxXL) [28,29] genes. Out of the 405 unique Lcr genes, 310 were eliminated in this study as being likely non-essential, leaving 95 genes (65 annotated and 30 hypothetical proteins) that were either 'essential' or quasi-essential for Lcr growth (Additional File 3).

Nutrient reprieve alone is likely insufficient for axenic growth of CLas
Large-scale computational metabolic modeling identified 372 genes driving 892 metabolic reactions involving 887 metabolites in Lcr BT-1. By comparison, only 253-285 genes, driving 814-840 reactions and producing 802-837 metabolites were identified in six different CLas strains [35]. In addition to 109 unique metabolic reactions present in Lcr, ~ 30% of the Lcr-specific reactions were associated with the cell envelope and missing in all sequenced CLas strains [35]. All CLas strains were predicted to be more heavily dependent on additional metabolites, carbohydrates, nucleotides, amino acids and vitamins, and also exhibited marked deficiencies in cell envelope biogenesis, consistent with several lines of published empirical evidence [23,28,29,36].
Of the 95 culturability-related 'essential' genes of Lcr, 30 encoded hypothetical proteins with unknown function (Additional File 3) and without sequence similarity to any prokaryotic or eukaryotic proteins discoverable by BLASTP in GenBank (Additional File 4). Ten of the 30 hypothetical proteins were predicted to be secreted either via classical [37,38] or noncanonical [39] secretion pathways and eight proteins were predicted to be membrane localized integral proteins [40] (Additional File 3). The smallest synthetic bacterial genome of Mycoplasma mycoides JCVI-syn3.0 (531 kb, 473 protein-coding and 35 RNA genes) contained 84 genes that were involved in the maintenance of cell envelope and 149 genes encoding proteins with unknown biological function [34,41].
We also analyzed the genomes of 17 uncultured bacterial species, including plant and animal pathogens and insect endosymbionts, for the presence of orthologs of all  4). Notably, 49 of the 95 predicted to be required for axenic growth had no orthologs in any of these bacteria, while the remaining 46 had orthologs scattered in one or more of these uncultured bacterial genomes (Table 1). These results support our hypothesis that host-free and autonomous axenic growth of CLas can only be achieved via simultaneous addition of multiple Lcr genes identified as 'essential' , and not by media additives or manipulation of axenic growth conditions alone.

Organism
Occurrence/disease Genbank/EMBL/ DDBJ Acc. # exoconjugants. Simulated growth modeling, accounting for the connectivity of carbon and nitrogen sources, amino acids and vitamins across physiological networks predicted that the Lcr culture medium BM7 was also optimal for in vitro growth of CLas [35]. Alternatively, modified BM7 medium (BM7A) may also be used for recovery of CLas transconjugants. BM7A medium, with increased buffering capacity and reduced medium alkalization, resulted in 1000-fold improved recovery of 'viable and culturable' Lcr cells from 10-day-old cultures [3,50].

Conclusions
Currently, 97% of 14,000 cultured bacterial species (across 3500 genera and 38 phyla) belong to just four bacterial phyla (Bacteroidetes, Proteobacteria, Firmicutes and Actinobacteria), but the vast majority remain poorly characterized in vitro [51]. Several bacteria with reduced genomes have remained recalcitrant to axenic growth in vitro, likely because of (a) metabolic deficiencies that cannot be relieved by media supplementation alone, (b) novel regulatory networks that are needed for optimum gene expression and (c) additional genes that required for essential structural, membrane barrier and unknown functions for in vitro growth. Global mutagenesis datasets and modeling of regulatory and metabolic networks in phylogenetically related culturable species can provide valuable insights into gene 'essentiality' functions and bottom-up implementation of specific synthetic 'essentialomes' for  ) recA13 supE44 ara14 galK2 lacY1 proA2 rpsL20 (Sm R ) xy15 λ − leu mtl1; and TOP10 (donor): F-mcrA Δ(mrr-hsdRMS-mcrBC) Φ80lacZΔM15 ΔlacX74 recA1 araD139 Δ(ara leu)7697 galU galK rpsL (Str R ) endA1 nupG. C-D 4 μl aliquots of overnight cultures (Abs 600 = 1.0) of helper (HB101) and donor (TOP10) E. coli strains were sequentially spotted on top of a 5-day-old 3 ml Lcr culture pellet and cocultured for 8-12 h at 28 °C. E-F The coculture mix was streaked on selective BM7 plates (6 μg/ml lomefloxacin and 2 μg/ml gentamycin) for 12 weeks, and the Lcr exoconjugants were verified for the presence of mobilized pCLL031 by restriction digestion analysis and GFP expression axenic culturing of economically important pathogens with reduced genomes such as CLas.

Limitations
Successful axenic culturing of CLas will likely require transfer and expression of a complete set of at least 95 Lcr genes, all of which are simultaneously required. Efforts are underway to obtain a comparative high density transcriptomic roadmap of Lcr and CLas to better understand (a) previously uncharacterized gene regulatory networks and (b) the significance of large numbers of species-specific hypothetical proteins of unknown function present in both the Lcr and CLas genomes.