Cultivating the uncultured: growing the recalcitrant cluster-2 Frankia strains

The repeated failures reported in cultivating some microbial lineages are a major challenge in microbial ecology and probably linked, in the case of Frankia microsymbionts to atypical patterns of auxotrophy. Comparative genomics of the so far uncultured cluster-2 Candidatus Frankia datiscae Dg1, with cultivated Frankiae has revealed genome reduction, but no obvious physiological impairments. A direct physiological assay on nodule tissues from Coriaria myrtifolia infected with a closely-related strain permitted the identification of a requirement for alkaline conditions. A high pH growth medium permitted the recovery of a slow-growing actinobacterium. The strain obtained, called BMG5.1, has short hyphae, produced diazovesicles in nitrogen-free media, and fulfilled Koch’s postulates by inducing effective nodules on axenically grown Coriaria spp. and Datisca glomerata. Analysis of the draft genome confirmed its close proximity to the Candidatus Frankia datiscae Dg1 genome with the absence of 38 genes (trehalose synthase, fumarylacetoacetase, etc) in BMG5.1 and the presence of 77 other genes (CRISPR, lanthionine synthase, glutathione synthetase, catalase, Na+/H+ antiporter, etc) not found in Dg1. A multi-gene phylogeny placed the two cluster-2 strains together at the root of the Frankia radiation.

Ever since Robert Koch first proposed a series of postulates to establish a causal link between a microbe and a disease, a basic tenet of microbiology has been the isolation of the causative microbes in pure culture 1 . Because many microbes such as fungi (Gigaspora, Rhizophagus) or bacteria (Mycobacterium leprae, Wolbachia, etc), cannot be cultivated in pure culture, it is now widely accepted these postulates should not be strictly considered 2 . Nevertheless, growth in pure culture remains an irreplaceable tool to study the physiology of an organism. Reasons behind these failures are varied including fragility of cells buffered within host tissues, cumulative damages by reactive compounds to cellular constituents such as DNA and membranes, or loss of genes resulting in auxotrophy 3 .
Frankia is a slow-growing actinobacterium. First described in 1866 as the causative agent of nodules found on the roots of Alnus and Elaeagnus 4 , it resisted cultivation attempts for a century. Because of its importance in ecological successions, studies were made using crushed nodules. However, the lack of isolates caused erroneous conclusions on plant host range and phylogenetic affiliation of the microbes present 5 . In 1978, the first Frankia isolation was reported 6 using an enzymatic maceration of host tissues and a complex growth medium, and has led to many subsequent successful reports on the isolation of strains from the different host plants. Molecular phylogenetic approaches using both cultured and uncultured bacteria have grouped Frankia into 4 clusters 7 and 3 of these clusters now routinely yield isolates from most provenances. However in spite of repeated attempts in numerous laboratories, cluster-2 strains that nodulate Rosaceae, Coriariaceae, Datiscaceae and Ceanothus have consistently resisted isolation. Nevertheless, the genome of a cluster-2 strain was determined by directly sequencing DNA isolated from nodule tissues 8 and a comprehensive study of its metabolic pathways permitted to conclude there was no missing function that would impede cluster-2 strains to grow in pure culture 9 .
One of the aims of the cultivation of a microbe is to provide knowledge on its physiological properties, but this information is also necessary to tailor the growth media for isolation of this given microbe. This circular logic is usually broken through trials and errors. The Biolog (Hayward CA) assay, which allows the assessment of physiological properties of cells, does not depend on growth, but rather on the ability of cells to reduce an indicator dye, tetrazolium to the purple insoluble formazan 10 . We decided to use this system to study the physiology of cluster-2 Frankia strains and thus adapt our isolation conditions to fit the endophytic cells.
The genome of Candidatus Frankia datiscae Dg1, an uncultured cluster-2 strain, was analyzed to provide information on these symbionts for their isolation. Although no metabolic pathways appeared to be incomplete, the genome contained a reduced number of genes involved in stress responses 8 . With this information, a high-throughput microplate assay was used to first assess the viability of the microsymbiont. These live/dead assays were performed under different pH conditions, with various carbon sources and in the presence or absence of oxygen scavengers and osmolytes. Strikingly, the viability of these endophytic cell clusters was highly sensitive to pH. Viable endophytic cells were recoverable only from PM10 microplate wells with a pH higher than 8.5 for up to two months as shown by the live/dead assay. Therefore, buffers used to release vesicle clusters from root nodules and to inoculate Biolog microplates were adjusted to pH 9 for all subsequent experiments, which permitted a substantial extension of cell endophyte viability in numerous wells in PM1 and PM2A microplates. Pyruvate, propionate, succinate, and acetate are all generally used to cultivate other Frankia strains [11][12][13] , and were identified as potential carbon and energy sources for the isolation medium.
Based on this information, a modified BAP medium 13,14 buffered to pH 9 and supplemented with several organic acids to replace the usual propionate as carbon source and several oxygen scavengers was used, and yielded very slow-growing brownish colonies. These colonies required a 14 month incubation period to yield sufficient material for subculturing. One colony, termed BMG 5.1, was selected, and gently fragmented for propagation. The isolate was maintained in pure culture and found to be free of contaminants after plating a sample of it on solid LB medium. The isolate grew slowly, forming hard-to-fragment colonies in liquid medium (Fig. 1A). Under phase contrast microscopy, the isolate was observed to consist of a mycelial mat containing short irregular filaments ( Fig. 1B) with thick elbow-like regions that could be due to discontinuous growth (Fig. 1C). Globular diazovesicles, the site of nitrogen fixation in Frankia under symbiotic and pure-culture conditions 15 , were produced in nitrogen-deficient (BAP medium without NH 4 Cl) medium (Fig. 1D). No sporangia were observed under the conditions tested. The culture grown on nitrogen-deficient medium was also able to reduce acetylene (135 + /− 11 nmoles h-1 mg-1 protein), a proxy for nitrogen-reduction, as soon as vesicles were formed. The results from the Biolog assays on isolate BMG5.1 grown in pure culture (Table S1) showed a similar pattern of physiological properties as predicted from the viability screening of microsymbiont clusters. Furthermore, isolate BMG5.1 showed an optimum growth yield at pH 9.5 ( Fig. 2). In contrast, the other Frankia strains showed optimum growth yield around pH 7-7.5. Isolate BMG5.1 failed to grow in medium at pH 6.5 that was usually used for other Frankia isolates 14 .
Phylogenetic analysis shows that BMG5.1 strain belongs to Frankia cluster-2 and shares 99.6%, 98.16%, 98.76% and 99.4% identity with the Candidatus Frankia datiscae Dg1, respectively on the basis of 16S rRNA, nifH, gyrB and glnA gene sequences. A multigene analysis using concatenated atpI, gyrA, ftsZ, secA and dnaK genes confirms the close proximity of the new isolate with Candidatus Frankia datiscae Dg1 at the root of the genus radiation (Fig. 3).
The draft genome of Frankia sp. strain BMG5.1 was determined and found to consist of 5,795,263 bp with a 71% G+ C and 5,245 predicted protein-coding genes, characteristics which are very close to those of the Candidatus Frankia datiscae Dg1 genome ( Table 1). All of the generated 500000 reads were found to be Frankia sequences indicating the absence of potential contaminants. Comparative analysis showed that the two genomes had an average nucleotide identity (ANI) of 96.93%, which would correspond to the same species 16 and ANI scores of 82.32, 82.97, 83.05 and 83.68% with Frankia sp. strains EuI1c, ACN14a, EAN1pec and CcI3 genomes, respectively 17 . Several genes were found to be present in BMG5.1, but were absent in Dg1 (Table S2) including catalase, choline dehydrogenase and glutathione synthase that would correspond to a higher level of resistance to in vitro culture conditions for BMG5.1. Interestingly, the Frankia sp. BMG5.1 genome contains multiple Na+ /H+ antiporter genes (Fig. S1).
To fulfill Koch's postulates, plant infectivity experiments were performed on seedlings of Coriaria japonica, Coriaria myrtifolia and Datisca glomerata, all of which are hosts of cluster-2 Frankia. The seedlings were grown axenically and transferred to growth pouches prior to inoculation with the isolate. After inoculation with isolate BMG5.1, the plants produced root nodules within 30-35 days. These nodules were actively fixing nitrogen as determined by the better growth of inoculated seedlings in N-free medium relative to non-inoculated controls (Fig. 4) and by the result of the acetylene reduction assay 18  reisolated from these nodules using the same isolation methods, grown in pure culture, and shown by PCR-sequencing to have identical markers (16S rRNA, nifH, gyrB and glnA).
It is a stressful situation for a bacterium buffered by host tissues and under oxygen levels kept low by neighboring mitochondria 19 to grow on synthetic medium. Down-regulation of many stress proteins in symbiotic Frankia alni 20 also indicates that symbiosis may be less stressful than saprotrophic growth. In the case of alders, defense peptides have been shown to increase Frankia porosity 21 and would render symbiotic cells particularly fragile. Thus, if in vitro growth conditions greatly differed from those found in the host tissues, isolation would be extremely difficult and would probably not occur. This is the reason why it is necessary to modify growth media to make them as comparable to the physiological conditions existing inside host tissues as possible. In the case of cluster-2 Frankia, these changes to the growth medium included the addition of oxygen scavengers and the use of an alkaline pH. Physiological assessment of uncultivated cells can thus provide reliable physiological data, leading to testable hypotheses for pure culture conditions.
This new Frankia isolate may form, together with Candidatus Frankia datiscae Dg1 strain, a different species that is mildly alkaliphilic. A number of alkaliphiles have been found among various lineages, many of which are actinobacteria in arid soils 22 . Adaptation to an alkaline biotope occurs through either local acidification of the medium or modification of the wall. Among the genes present in the BMG5.1 genome but absent in Dg1 were the Na+ /H+ antiporters, which have been described as permitting cells to adjust to high pH 22 . A pure culture of Frankia BMG5.1 should be helpful in experiments directed toward our understanding this requirement for these unusual physiological conditions. Clearly, the use of bioinformatics analysis of the genome and physiological bioassays has resulted in the isolation of a previously recalcitrant microbe. This dual approach could be extended to other microbes considered as yet uncultivable such as Mycobacterium leprae or Rhizophagus intraradices, which should markedly facilitate their study and biotechnological uses.  Methods Vesicle clusters viability and growth requirement screening. Root nodules obtained from Coriaria japonica 23 growing in Japan (Tosa district) were used to inoculate Coriaria myrtifolia seedlings grown on sterile sand and maintained in a greenhouse. Root nodules were harvested, surface disinfected (Na hypochlorite 1.05% for 30 min) and washed several times with sterile distilled water. Preparation of vesicle clusters was performed as described previously 24 . Lobes were crushed in 0.05 M sodium phosphate buffer, pH 7.2 using a Broeck tissue homogenizer and incubated with the following scavengers for hydrogen peroxide (10 mM N,N'-dimethylthiourea) hydroxyl radical (0.28 M dimethylsulfoxide), for singlet oxygen (0.02 M L-histidine), for superoxide anion (10 mM 4,5-dihydroxybenzene-1,3-disulfonate), for peroxyl radical (100 μ M α -tocopherol) and for peroxynitrite anion (100 μ M uric acid). Released microsymbiont vesicle clusters were retained by gravity filtration on the 52-μ m nylon screen and used to assess cells viability using PM1 and PM2A microplates (Biolog) for carbon sources, as well as PM9 and PM10 microplates (Biolog) for pH and NaCl tolerance, respectively. Viability of endophytic cell clusters was weekly assessed in a subset of PM1, PM2A, PM9 and PM10 microplates using only the live/dead assay BacLight Bacterial Viability Kit for microscopy (Life Technologies, Paisley, UK).
Host plant infectivity and molecular identification. Surface sterilized seeds of C. japonica, C. myrtifolia and D. glomerata were germinated on moistened filter paper in Petri dishes, at 20 °C under daylight illumination and transferred to sterile moist sand. Three week old seedlings were inoculated by applying the equivalent to 10 μ g of protein of washed cultures of the isolate previously syringed with a 21G sterile needle. Plantlets were checked for nodules every week. After 9 weeks, nodules were harvested and used for endophyte reisolation and direct DNA extraction. DNA was extracted from root nodules and isolates using Plant DNeasy kits (Qiagen, Hilden, Germany) and characterized by PCR/sequencing of 16S rRNA , nifH, gyrB and glnA genes using respectively S-D-Bact0008-a-S-20 and S-D-Bact-1495-a-A-20 26 , nifHF and nifHR 27 , FragyrBF and FragyrBR 28 and DB41 and DB44 29 . The generated sequences were then analyzed by BLASTN 30 and compared to those obtained for BMG5.1.
Acetylene reduction activity. C. japonica and D. glomerata seedlings with and without nodules were assayed for nitrogenase activity by monitoring acetylene reduction 18 . About 1g of nodules was harvested and placed in sterile Vacutainer (Beckton-Dickinson, Franklin Lakes, NJ) tubes which had 10% of the gas volume replaced with acetylene. After a 60 min incubation at 30 °C, 1 mL of the gas volume was injected into a Girdel gas chromatograph (Suresnes, France) equipped with a flame ionization detector.

BMG5.1 Dg1
Status A similar assay was made for a BMG5.1 culture in BAP medium without ammonium (BAP-). Vacutainer tubes (10 mL) were inoculated with 5 ml of the liquid cultures and after 1 week, 10% (vol/vol) of the remaining gas volume was replaced with acetylene and incubated for 60 min at 30 °C.
Biolog and growth at different pH. Frankia BMG5.1 cells were harvested by centrifugation (9000 x g, 15 min), washed three time and incubated four days in BAP-mineral solution (no carbon source) to deplete nutritional reserves. The cells were homogenized by repeated passages through a 25G needle. Biolog microplates were inoculated with cells at 10% transmittance suspended in IF0 buffer (supplied by the manufacturer) adjusted to pH 9 for PM1, PM2A and PM9 microplates and supplemented with pyruvate (10 mM final) for PM9 and PM10 microplates and incubated at 28 °C for 20 days. The new isolate BMG5.1 and strains representing the three other Frankia clusters; BMG5.12 (cluster-3), ACN14a (cluster-1a), CcI3 (cluster-1b) and EuI1c (cluster-4) were grown in BAP+ media adjusted to different pH. Organic buffers used to reach the desired pH were MES (2-N-morpholinoethanesulfonicacid) for pH 6.5, and MOPS (3-(N-morpholino)-propanesulfonic acid) for the pH range 6.5-7 and TRIS (tris-hydroxy-methyl-aminomethane) for pH above 7. After 3 weeks of growth, growth yield was determined by measuring total cellular protein. After harvesting, bacterial cell samples were suspended in 1.0 N NaOH and solubilized by incubating for 15 min at 90 °C. Total cellular protein levels were measured using the BCA method 31 .
Genome sequencing of the BMG5.1 strain. Approximately 500 mg of cells obtained from 2 L of 2 months old cultures were washed and their DNA extracted using a Plant DNeasy kit (Qiagen). Highly pure genomic DNA from strain BMG5.1 was used to generate a 300-bp library for paired-end 2 × 100 Illumina sequencing (Imagif, Gif-sur-Yvette, France) from 500000 reads. Sequences assembly was performed using the Velvet de novo assembler 32 . The assembled Frankia sp. strain BMG5.1 genome was analyzed via the Integrated Microbial Genomes (IMGer) platform of the Joint Genome Institute, Walnut Creek, CA. This whole genome shotgun sequence has been deposited at DDBJ/EMBL/GenBank under accession number JWIO00000000. The version described in this paper is JWIO01000000. The draft genome was compared to that of Candidatus Frankia datiscae to identify those genes different between the two genomes using the default values of 30% amino acid identity over 80% of the length of the shortest sequence on the IMGer platform 33 .