In this study we describe a transposon system applicable for efficient random mutagenesis in C. glutamicum ATCC 13032. The use of an IS6100-based transposon vector gave rise to a transposon library of statistically representative size comprising independent mutant clones.
It was shown earlier that IS6100 is capable of transposing in vivo in C. glutamicum with unique transposition events by forming a cointegrate with the chromosome . The excision frequency and therefore the stability of a replicative transposon integration is usually controlled by a resolvase protein, which is not encoded by the pAT6100 transposon vector. Nevertheless, cointegrate resolution involving either the two identical IS6100 copies or the 8-bp direct repeats might be caused by homologous recombination or by replication slippage resulting in the loss of the vector part and one or both copies of the IS6100 element. In this study an antibiotic sensitivity could not be observed after prolonged growth in the absence of antibiotic pressure during cultivation, indicating that IS6100 generates cointegrations that are stably maintained in C. glutamicum. This finding conforms with the study of Weaden and Dyson  in which the stable cointegration of IS6100 was observed in Streptomyces avermitilis.
The already developed transposon systems are not well applicable for random mutagenesis in the sequenced C. glutamicum type strain ATCC 13032 because of similar endogenous insertion sequences, pronounced target site preferences, or low transposition frequencies. For instance, seven copies of ISL3 family-like sequences in the ATCC 13032 genome prevent the usage of IS31831  and related elements (e.g. IS1207)  as well as their derived transposons (e.g. Tn5531) . Very recently a mutagenesis system was described that used IS31831- and Tn5-based minitransposons to generate a comprehensive library of the C. glutamicum strain R which covers nearly 80% of the presently unpublished genome . The C. glutamicum strain R has the advantage of not possessing insertion elements of the IS31831 family. In contrast to this, IS6100 is absent from the strain ATCC 13032 chromosome, ruling out integration by homologous recombination and thus it is suitable to perform transposon mutagenesis in this strain. Beside IS6100 quite a few active mobile elements have been published for C. glutamicum which are of restricted usability of generating a comprehensive random transposon library since they prefer specific sequences, e.g. a triple A or T (IS1249)  or palindromic sequences (e.g. IS14999) , as a target for integration. On the other hand, prior studies with IS6100-based transposons suggested independent transposition and absence of a target site preference. This was first shown in Streptomyces lividans and S. coelicolor by nucleotide sequence comparison for a small number of mutants  and, on the basis of Southern hybridization studies in S. lividans  and C. glutamicum .
In this study, the usability of IS6100 for mutagenesis in the type strain ATCC 13032 was shown by analysis of a larger number of clones. The sequence data determined from 172 insertions delivered definite position informations and, consequently, allowed comprehensive analyses on the transposon target sites. Each clone investigated carried the transposon in a different chromosomal location. The absence of regional preferences together with the absence of sequence preferences (sequence pattern, nucleotide-usage for the TSD positions or G+C content) as a target demonstrated the randomness of the library constructed and applied in auxanography analyses. Furthermore, this system might be a practicable genetic tool in other organisms as well because of the broad host-spectrum of the IS6100 element.
This study integrated genomic sequence data with genome-scale auxotrophy analyses. By means of auxanographic assays, for 63% of the 295 isolated auxotrophic mutants with distinct phenotypes various nutritional requirements could be identified. Additionally, out of this contingent the different integration loci of 101 clones were determined. The value of 2.9% auxotrophic clones only slightly differs from those in other transposon mutagenesis studies (1.3% , 2.5%  and 2% ) but is remarkably higher compared to studies in which auxotrophic mutants were obtained with frequencies of 0.2% in "Brevibacterium flavum" with IS31831 , 0.2% in C. glutamicum with IS1249 (Tn5432)  and 0.5% in Streptomyces avermitilis with IS6100 . This might be explained by the fact that latter systems use mobile elements for which an insertion sequence specificity was identified (IS31831 and IS1249).
Transposon integrations were found in a variety of known amino acid, nucleotide and vitamin pathway genes as well in genes encoding hypothetical proteins or such of presently unknown function. The vast majority of the observed auxotrophic phenotypes could be correlated and explained with the knowledge of the mutated genomic region since most genes of amino acid, nucleotide and vitamin biosyntheses are annotated in C. glutamicum. In contrast, for some observed phenotypes the connection between gene and auxotrophy is not yet clear, delivering interesting targets for further studies. The loss of gene products associated with de novo synthesis or recycling pathways may be compensated by influx of necessary metabolite entries from other pathways that share common intermediates or precursors to a given intermediate. For instance, the PurF protein, an amidophosphoribosyltransferase, is known to be involved in more than one pathway. In Salmonella typhimurium PurF is the first of five enzymes shared by the de novo purine and HMP (hydroxymethylpyrimidine) synthesis, an essential compound of thiamine biosynthesis. Thus, PurF is expected to be required for both purine and thiamine biosynthesis . It is not obvious why a purF transposon mutant is supplementable solely by hypoxanthine. Alternative pathways were discovered which could bypass the requirement for all pur genes in thiamine synthesis [61, 62]. Therefore, hypoxanthine, a purine derivative, might be sufficient to compensate the growth-deficiency of this mutant.
It is known that transposon insertions near the beginning or within an operon can attenuate or interrupt the expression of downstream genes preventing RNA polymerase readthrough [63, 64]. A polar effect on downstream genes was experimentally shown for an IS6100 integration in the gene cysI (cg3118) with transcriptional analysis using Real Time RT-PCR . This example points out that the transposon system is not only applicable for high density mutagenesis of chromosomal regions (e.g. tryptophan operon, Fig. 2) but might be utilised for the determination of operons.
The screening approach for characterization of auxotrophic mutants used in this study shows a limitation in resolving complex growth phenotypes, exhibited by one thirds of these mutants. However, determination of transposon positions in randomly selected members of this auxotrophy category revealed for example mutations within purC, purE, and purK, genes known to be involved in purine biosynthesis. Such genes, as mentioned above, might be of interest to be investigated in further experimental studies. These findings indicate that also the complex phenotypes are not a result of additional spontaneously occurring mutations, but caused by disruption of the corresponding gene or by the accompanying polar effect.
For detailed studies on gene functions, transposon mutants have to be confirmed experimentally in order to rule out secondary mutations, polar effects or leaky phenotypes. In this study, this was carried out by a concrete example in which, by the means of supplementation, genetic deletion and complementation assays, the last gap in the histidine biosynthesis pathway of C. glutamicum could be closed.
Initial similarity searches revealed that the Cg0910 protein (HisN), together with another four paralogs in the C. glutamicum genome, apparently belongs to the monophosphatase-family proteins that usually hydrolyse the ester bond of myo-inositol-1(or4) phosphate. Inositol monophosphatases (IMP) play a crucial role in the biosynthesis of inositol and inositol phospholipids . The role of IMP in bacteria is not completely clear yet. Bacteria of the genus Mycobacterium contain a number of inositol-derived cell wall constituents, like phosphatidylinositol (PI), phosphatidylinositol mannosides (PIM), lipoarabinomannan (LAM) and lipomannan (LM) [67–69]. LAM-like molecules and other inositol-containing phospholipids are not only present in mycobacteria but also in other Actinobacteria, including the genus Corynebacterium [70–72]. Actually, little is known about inositol synthesis in bacteria. The known de novo pathway in mycobacteria basically comprises the cyclization of glucose-6-phosphate to inositol-1-phosphate (I-1-P) and, subsequently, the dephosphorylation of I-1-P by IMP producing inositol .
Although the IMP-like proteins analysed in this study appear to be sequence homologs phylogenetic analysis revealed that they are significantly different as they branch into distinct classes. Despite their amino acid conservation, they seem to form a protein family of diverse functions and/or with diverse substrates. The assumption of different substrate specificities corresponds to previously published studies performed either in closely related mycobacteria or in E. coli.
The E. coli CysQ homolog is required for cysteine synthesis during aerobic growth. The sulfate assimilation branch of the cysteine pathway comprises sulfate uptake, its activation by formation of adenosine 5'-phosphosulfate (APS) and conversion to 3'-phosphoadenosine 5'-phosphosulfate (PAPS) by APS kinase, and its reduction to sulfite. It has been suggested that CysQ acts on PAPS as a target. It is proposed to help controlling the levels of PAPS, which may be toxic to the cell in high concentrations, or the generation of sulfite . The actinobacterial CysQ IMP homologs in the phylogenetic tree branch later as compared to CysQ of E. coli, thus functional differences between the E. coli and the actinobacterial CysQ proteins might be possible. In C. glutamicum an APS kinase homolog is missing and thus PAPS as an intermediate is not formed. Sulfite is released from APS by direct reduction through APS reductase . Therefore, the target substrate for a PAPS CysQ protein would be missing as well. It might be considered that CysQ in C. glutamicum and the other actinobacterial members of this phylogenetic class indeed does not possess a PAPS phosphatase activity, but most likely an inositol-phosphate phosphatase activity.
Initially, impA has been proposed to encode the missing HolPase, as it is clustered with the histidine biosynthesis genes in the Actinobacteridae . This theory has been falsified for a M. smegmatis impA mutant. This mutant is not auxotrophic for histidine, but exhibits altered cell envelope permeability properties, with a notable reduction in the synthesis of phosphatidylinositol dimannoside, the precursor of LAM .
The SuhB inositol monophosphatase activity has been characterized biochemically in M. tuberculosis. Inositol-1-phosphate was shown to be the preferred target of SuhB for dephosphorylation in order to provide the PI synthase with inositol .
The specificity of Cg0910 (HisN) for histidinol phosphate was experimentally proven for C. glutamicum. For all other members of the phylogenetic IMP class labelled Cg0910 the intrinsic HolPase function could be proposed because of the close relationship and the high sequence conservation. Furthermore, in the respective Actinobacteria a protein having the enzymatic function of histidinol phosphate dephosphorylation was not identified.
The Cg0911 class comprises only corynebacterial members and one Nocardioides subspecies. The fact that in all these organisms cg0911 and cg0910 orthologs are located next to each other leads to the assumption of a recent gene duplication that exclusively occurred in corynebacteria or in a common ancestor. However, the finding that the respective proteins of cg0910 and cg0911 are members of different IMP classes indicate that they accomplish different functions and thus solely the cg0910 gene encodes the HolPase activity in C. glutamicum. Additional evidence that the unknown Cg0911 function, in contrast to Cg0910, does not play an important role in the cell was obtained by the inactivation of the respective gene (data not shown). The mutant showed no obvious phenotype under the applied growth conditions.