Gramicidin is a polypeptide which consists fo 15 hydrophobic amino acids with alternating L-D-configuration. It forms homodimers in vivo that function as ion channels in lipid bilayers. The primary strucutre of Gramicidin A has been determined as formyl-Val1-Gly2-Ala3-DLeu4-Ala5-DVal6-Val7-DVal8-Trp9-DLeu10-Trp11-DLeu12-Trp13-DLeu14-Trp15-ethanolamine. The peptide is protected from degradation by N-formylation of the N-terminus and ethanolamine at the C-terminus. In the work presented here a gene cluster of 61 kb was identified with the help of a fosmid gene bank from Bacillus brevis ATCC 8185. Sequencing revealed the Gramicidin A nonribosomal peptide synthetase. A multienzyme complex consisting of the four NRPSs LgrABCD which encode 2, 4, 6 and 4 modules, respectively, was determined to be responsible for assembly and modification of the pentadecapeptide. In silico investigations of the substrate specificity of all 16 modules were confirmed by biochemical analyses of the first three adenylation domains. By this, the principle of colinearity of this NRPS was proven. The putative C-terminal Reductase (R) domain which is fused to module 16 was thought to set the N-formylated peptide chain free as alcohol. Biochemical investigations using the recombinant R-Domain with a peptidyl-substrate variant showed that the enzyme conducts only a single-step reduction using NAD(P)H to the aldehyde intermediate. By using different substrats, a broad substrate tolerance of the R domain was demonstrated. The search for a second reductase which supposedly converts the peptidyl-aldehyde to the final product peptidyl-ethanolamine led to the discovery of the aldoreductase LgrE. In coupled assays using both reductases peptidyl-ethanolamine was produced. In in vitro investigations of LgrE with purified, enzymatically produced aldehyde-intermediate, a strict NADPH-dependence of the enzyme was shown. The aldoreductase showed a more stringent substrate specificity than the R domain. The second part of this work showed the development of a method for module exchanges in NRPSs at the genetic level. At first a gene fragment coupled to a selectable marker is introduced into the genome via homologous recombination. In a second step the marker is lost again via loop out using direct repeats. The gene fragment remains in the chromosome, and the new NRPS product can be investigated.