G A Cumme

Friedrich-Schiller-University Jena, Jena, Thuringia, Germany

Are you G A Cumme?

Claim your profile

Publications (46)72.1 Total impact

  • [Show abstract] [Hide abstract]
    ABSTRACT: Biomarker search using multidimensional native liquid fractionation of serum in microplates was evaluated. From different donors, homologous sample fractions with UV absorbance depending on state of illness were selected, and their constituents were identified and quantitated by MS. Analysis of sera of patients with Alport syndrome and severe inflammation proved the reliability of the method by confirming characteristic alterations. Moreover, 23 new marker candidates were detected for Alport syndrome, some of them being involved in matrix degradation and repair, and 33 new candidates for severe inflammation, among them alpha1B-glycoprotein cysteine-rich secretory protein and an apparently low molecular-weight albumin variant.
    Journal of Chromatography B 11/2008; 876(1):31-40. DOI:10.1016/j.jchromb.2008.10.014 · 2.69 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: The microplate-based method developed by our group for non-denaturing multidimensional proteome separation was improved on using improved column arrays and a newly developed robot. Currently size exclusion, anion exchange and lectin affinity chromatography are combined orthogonally. Different samples run simultaneously to enhance reliability of intercomparison. LC-ESI (electro-spray ionization) MS/MS analysis of selected fractions identified 32,288 peptides matching 2,669 serum proteins. The present contribution (I) shows the characteristics of the method, whereas "prove of principle" by applying it to search for biomarker candidates with model diseases is reported in an accompanying paper (II).
    Journal of Chromatography B 11/2008; 875(2):567-72. DOI:10.1016/j.jchromb.2008.09.040 · 2.69 Impact Factor
  • Source
    Gerhard A Cumme · Stefan Kreusch · Matthias Nagel · Heidrun Rhode
    [Show abstract] [Hide abstract]
    ABSTRACT: 2-D native LC yields thousands of fractions especially when applied to sera of different origin. Checking reproducibility of repeated separation of the same serum or searching for biomarker candidates and fractions containing them requires finding, selection, and comparison of interesting data subsets out of huge data volumes. An innovative software package is applied that markedly enhances simplicity, velocity, and reliability of (i) check of reproducibility of the separation method and (ii) analysis of proteomes pertaining to different disease states.
    PROTEOMICS 01/2008; 8(1):37-41. DOI:10.1002/pmic.200700453 · 3.97 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: How to effectively mix small volumes of liquids within microplate wells is a still underestimated and often neglected challenge. The method the authors introduce here relies on violent turbulent motion within a liquid caused by spotting an organic solvent drop onto its surface. The amount needed, less than 1 to 3 microL, is generally small enough not to alter bioactive molecules. Moreover, a solvent may be selected for its compatibility with assay components. The method was tested with layers of aqueous liquids that differ in pH and concentration of a pH-dependent dye, allowing mixing to be monitored optically. Rapid mixing was caused by spotting drops of alcohols, acetone, acetonitrile, and aqueous solutions of these, as long as the difference of surface tension between the drop and the uppermost layer of the bulk liquid surpassed 30 dynes/cm. Along with this difference, position and velocity of spotting, as well as viscosity and geometry of the bulk liquid volume, may influence the turbulence evoked. No significant difference was found for the activity of aspartate aminotransferase, alanine aminotransferase, and alkaline phosphatase when measured after mixing by shaking and after mixing by spotting 1 microL of methanol onto assays within 96-well microplates.
    Journal of Biomolecular Screening 05/2007; 12(3):361-9. DOI:10.1177/1087057106297565 · 2.01 Impact Factor
  • Shock 10/2006; 26(Supplement 1). DOI:10.1097/00024382-200610001-00003 · 2.73 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: A method is introduced to evaluate protein concentrations using the height sum of all MALDI-MS peaks that unambiguously match theoretic tryptic peptide masses of the protein sought after. The method uses native chromatographic protein fractionation prior to digestion but does not require any depletion, labeling, derivatization, or preparation of a compound similar to the analyte. All peak heights of tryptic peptides are normalized with the peak height of a unique standard peptide added to the MALDI-MS samples. The sum of normalized peak heights, S(n), or the normalized mean peak height, M(n), reflects the concentration of the respective protein. For fractions containing various proteins, S(n) and M(n) can be used to compare concentrations of a protein between different fractions. For fractions with one predominating protein, they can be used to estimate concentration ratios between fractions, or to quantify the fractional protein concentration after calibration with pure protein solutions. Initial native fractionation retains the possibility to apply all conventional analytic procedures. Moreover, it renders the method relatively robust to MS mass accuracy. The method was validated with albumin, transferrin, alpha1-antitrypsin, and immunoglobulin G within highly complex chromatographic fractions of pathological and normal sera, which contained the respective intact native protein in dominating as well as minor concentrations. The correlation found between S(n) and the protein concentration as determined with ELISA showed that the method can be applied to select markers for distinguishing between normal and pathological serum samples.
    PROTEOMICS 07/2006; 6(13):3909-17. DOI:10.1002/pmic.200500747 · 3.97 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: A versatile, multidimensional, and non-denaturing proteome separation procedure using microplate technology is presented, yielding a digitized image of proteome composition. In the first dimension, the sample under study is separated into 96 fractions by size exclusion chromatography (SEC). In the second dimension, the fractions of the first dimension are transferred by the liquid-handling device CyBi-Well (CyBio AG, Jena, Germany) to 96 parallel anion exchange chromatography columns. In this way the proteins are conserved in their native states and are distributed in 2400 liquid fractions with high recovery rates and sufficient reproducibility. The resulting fractions are subjected to protein quantitation and identification. Spectrophotometrical and immunological methods and enzyme activity measurements are used for quantitation. To identify proteins, the fractions are subjected to MALDI-MS, and their tryptic digests to both MALDI- and LC-ESI-MS/MS. All preparation steps except the first are applied in parallel to sets of multiples of 96 samples. The procedure may be refined by adding more separation steps and may be adapted to various protein amounts and to various proteomes. Moreover, the method offers the opportunity to investigate functional protein complexes. The method was applied to separate the normal human serum proteome. Within 255 fractions exhibiting the highest protein concentrations, 742 proteins were identified by LC-ESI-MS/MS peptide sequence tags.
    PROTEOMICS 01/2006; 6(2):559-70. DOI:10.1002/pmic.200500142 · 3.97 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: An efficient method is presented to determine precision and accuracy of multichannel liquid-handling systems under conditions near to application. The method consists of gravimetrical determination of accuracy and optical determination of precision based on the dilution of absorbing and fluorescent dye solutions in microplates. Mean delivery volume per well can be determined with precision better than a 0.04% coefficient of variation (CV). Optical signal precision, CV(S), is improved by multiwavelength measurements. Precision of absorbance measurement yields a better resolution than precision of fluorescence measurement (0.3% and 1.5%, respectively), indicating that absorbance measurements should be preferred. From CV(S), an upper bound of the precision of the volumes delivered is derived. Method performance is demonstrated with the dispenser CyBi-Drop and the pipettor CyBi-Well using different ejection principles; with commonly used fluids; with 96-, 384-, and 1536-well microplates; and with photometric and fluorometric indicators. Precision of the volumes delivered, as obtained with optimized methods, all plate formats, and both devices, is better than 2% CV with 2 microL set volume and about 1% CV with higher set volumes.
    Journal of Biomolecular Screening 01/2005; 9(8):726-33. DOI:10.1177/1087057104269496 · 2.01 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Digestion of calf intestine alkaline phosphatase with pronase and subsequent dephosphorylation of the released peptidyl-(Etn-P)2-glycosyl-PtdIns with HF generated 8 glycosyl-Ins species the largest of which (G1 and G2) have the following proposed structures:G3 and G5 are lower homologues of G1 and G2, respectively, being one αl-2 linked mannopyranosyl residue shorter. G4 is analogous to G2 lacking the N -acetylgalactosaminyl residue and G6 is the next lower homologue of G4. Most of G4 and G6 occur substituted with a palmitoyl (G4, G6) or a myristoyl residue (G6) probably attached to the inositol moiety. Thus, the basic ManxGlc-Ins species are either substituted with an N -acetylgalactosaminyl residue or a fatty acid ester. The structures were deduced from compositional analysis, molecular-mass determination by matrix-assisted laser desorption MS, sequential hydrolysis with appropriate exoglycosidases and treatment with CrO3. Purification of the glycosylinositol species was achieved by a novel reverse-phase HPLC technique using fluorescent fluoren-9-yl-methoxy-carbonyl (Fmoc) derivatives. These stable derivatives were susceptible to hydrolysis with exoglycosidases which allowed sequential cleavages to be carried out and kinetics to be followed at the picomole level.We observed recently that native alkaline phosphatase separates on octyl-Sepharose into four distinct fractions of increasing hydrophobicity (F1–F4). Here we show that all four fractions contain G1–G6. The acylated species G4 and G6 were restricted to F2 and F4 which had been shown earlier to contain, on average, 2.5 and 3 fatty acid residues/subunit, respectively. In all four fractions the diradylglycerol moiety was predominantly diacylglycerol, alkylacylglycerol being less than 10% which is in contrast to most glycosyl-PtdIns–anchored proteins of mammalian origin.
    08/2004; 238(1):259 - 269. DOI:10.1111/j.1432-1033.1996.0259q.x
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: An UV spectrophotometric method for protein determination using microplates is described. Using the SPECTRAmax PLUS reader, the UVStar 96- and 384-well microplates and a 96 or 384 parallel channel liquid handling technique, large-scale determinations can be performed with intraassay precision better than 3% CV (coefficient of variation) in the range from 1 to 8000 microg of protein/ml, measuring at 205, 215, and 280 nm and using different volume-dependent light-path lengths. Since the absorbance coefficient at 205 nm is found to be 30 ml/(mgxcm) for eight different proteins with a CV of 5.6% only with the Path Check option of the reader, protein concentration can be determined without any individual calibration. Samples in the volume range of 60-250 microl can be analyzed without time-consuming and expensive treatment and without sample loss. Using a special 96 or 384 parallel dialyzing device, low molecular weight substances which interfere with the analysis by their UV absorbance, such as buffers and detergents, can effectively be removed. Application examples for serum protein separation are also shown in the presence of the strongly UV absorbing detergent Triton X-100.
    Analytical Biochemistry 03/2003; 313(2):208-15. DOI:10.1016/S0003-2697(02)00460-8 · 2.22 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: Surprisingly alkaline phosphatase (AP) (EC 3.1.3.1) of calf intestine is found in large amounts, e.g. 80%, within chyme. Most of the enzyme is present as a mixture of four differently hydrophobic anchor-bearing forms and only the minor part is present as an anchorless enzyme. To investigate whether changes in the N-glycosylation pattern are signals responsible for large-scale liberation from mucosa into chyme, the glycans of the two potential glycosylation sites predicted from cDNA were investigated by matrix-assisted laser desorption/ionization and electrospray ionization mass spectrometry in combination with exoglycosidase treatment after tryptic digestion and reversed-phase chromatography. The glycans linked to Asn249 are at least eight different, mainly non-fucosylated, biantennary or triantennary structures with a bisecting N-acetylglucosamine. For the most abundant glycopeptide (40%) the following glycan structure is proposed: [carbostructure: see text]. The glycans linked to Asn410 are a mixture of at least nine, mainly tetraantennary, fucosylated structures with a bisecting N-acetylglucosamine. For the most abundant glycopeptide (35%) the following glycan structure is proposed: [carbostructure: see text]. For the structures the linkage data were deduced from the reported specificities of the exoglycosidases used and the specificities of the transglycosidases active in biosynthesis. The majority of glycans are capped by alpha-galactose residues at their non-reducing termini. In contrast to the glycans linked to other AP isoenzymes, no sialylation was observed. Glycopeptide 'mass fingerprints' of both glycosylation sites and glycan contents do not differ between AP from mucosa and chyme. These results suggest that the observed large-scale liberation of vesicle-bound glycosylphosphatidylinositol (GPI)-anchored AP from mucosa into chyme is unlikely to be mediated by alteration of glycan structures of the AP investigated. Rather, the exocytotic vesicle formation seems to be mediated by the controlled organization of the raft structures embedding GPI-AP. (c) 2001 John Wiley & Sons, Ltd.
    Journal of Mass Spectrometry 08/2001; 36(8):960-72. DOI:10.1002/jms.200 · 2.71 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: The interaction of a glycosylphosphatidylinositol (GPI) protein with different detergents was studied for the first time with a purified protein. Four differently hydrophobic fractions of GPI-alkaline phosphatase (GPI-AP) from calf intestine were used as model proteins. The mode of interaction was determined by investigating (i) the self-aggregation behaviour of the GPI-AP fractions, (ii) the interference of detergents with GPI-AP binding to octyl-Sepharose, and (iii) the elution of GPI-AP bound to octyl-Sepharose. It was shown that polyoxyethylene-type detergents surprisingly interact much stronger than n-octylglucoside with GPI-AP, which is in contrast to the known behaviour of GPI-proteins in natural membranes. Gel filtration chromatography of Triton X-100 at concentrations above the critical micellar concentration yields three different micelle species with apparent molecular weights of about 166, 54, and 16 kDa. GPI-AP fraction II, which is shown to bear only one anchor per dimer, does not bind to any of these micelles. We demonstrate that a complex is formed containing about 150 Triton X-100 molecules and about 4700 molecules of water per molecule of GPI-AP dimer. The experimental findings are in accordance with a simple geometrical model based on the physical data of fatty acids and the arrangement, mean size, and shape of Triton X-100 molecules.
    Biological Chemistry 03/2000; 381(2):161-72. DOI:10.1515/BC.2000.022 · 2.69 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: The enzymatic properties of glycosylphosphatidylinositol-specific phospholipase D (EC 3.1.4.50) were characterized using a 6,000-fold purified enzyme. This was obtained in 100 microg amounts from human serum with a recovery of 35%. Pure alkaline phosphatase containing one anchor moiety per molecule was used as substrate. The enzyme is stimulated by n-butanol, but in contrast to other phospholipases this activation is not produced by a transphosphatidylation reaction. The previously reported non-linearity of the specific activity with respect to phospholipase concentration in the test was no longer observed upon purification, indicating inhibitor removal. The serum inhibitor(s) co-chromatograph with serum proteins and lipoproteins. The main part of the inhibitory activity was found in the lipid fraction after protein denaturation and can be subfractionated into acid phospholipids, cholesteryl esters and triacylglycerides. Added phosphatidyl-serine, phosphatidylinositol, phosphatidylglycerol, gangliosides, cholesteryl esters, and sphingomyelins turned out to be strong inhibitors, as well as phosphatidic acid. Phosphatidylethanolamine and various monoacylglycerols were found to be activators. The low glycosylphosphatidylinositol-specific phospholipase activity found in native serum did not increase significantly upon 90% removal of phospholipids by n-butanol. High serum concentrations of strongly inhibiting compounds, complex kinetic interactions among aggregates of these substances, and compartmentalization effects are discussed as possible reasons for the observed inactivity.
    Biological Chemistry 01/2000; 381(5-6):471-85. DOI:10.1515/BC.2000.062 · 2.69 Impact Factor
  • Gerhard Arnim Cumme · Eva Blume · Renate Bublitz · Horst Hoppe · Anton Horn
    [Show abstract] [Hide abstract]
    ABSTRACT: Matrix-assisted laser desorption/ionization mass spectrometry (MALDI-MS) can be used to determine the distribution of single polymer species of non-ionic detergents of the polyoxyethylene type (Triton X-100 and 114, Tween 20 and Brij 35). Thin-layer chromatographic (TLC) and reversed-phase chromatographic (RPC) methods are presented which may separate single polymer species as verified by MALDI-MS. Comparison of chromatographic and MALDI-MS data show Poisson-like distributions for Triton X-100 without marked differences between different methods. Distribution parameters obtained with Triton X-100 charges from different suppliers are very similar. The RPC method used can be scaled up for preparation of pure detergent species.
    Journal of Chromatography A 12/1997; 791(s 1–2):245–253. DOI:10.1016/S0021-9673(97)00800-5 · 4.26 Impact Factor
  • A Klemm · T Steiner · U Flötgen · G A Cumme · A Horn
    Methods in Enzymology 02/1997; 280:171-86. · 2.19 Impact Factor
  • E Müller · H Hegewald · R Klinger · R Wetzker · G A Cumme · H Frunder
    [Show abstract] [Hide abstract]
    ABSTRACT: Effects of intracellular pH and Mg2+ on turnover and extent of metabolic compartmentation of phosphomonoester groups of phosphoinositides and phosphatidate were investigated in human erythrocytes by short-term and equilibrium labeling with [32P]Pi under steady-state conditions. At pH 6.7, the specific radio-activities of phosphoinositides reached apparent equilibrium values, in the range of 70% of that ATP-gamma-P after long-term labelling. At pH 7.2, these values were in the range of 40-50% of that ATP-gamma-P. This demonstrates a decreased accessibility of phosphoinositides to enzymatic phosphorylation and dephosphorylation at the transition from acidic to normal incubation conditions. These changes were more pronounced at pH 7.8. High intracellular [Mg2+] initially activated the turnover of phosphoinositides at pH 7.2. The activation changed into an inhibition and an increase of metabolic compartmentation after three hours preincubation of erythrocytes at high [Mg2+]. The long-term Mg2+ effects are reversible to a great extent by a subsequent re-reduction of [Mg2+]. In conclusion, deprotonation of phosphoinositides by low [H+] or high intracellular [Mg2+] at normal pH induces a decreased accessibility for their specific lipid kinases and phosphatases. This effect may be the result of lateral phase separation of acidic phospholipids as a consequence of divalent cation complexation under both experimental conditions, high pH and high [Mg2+] at normal pH.
    Biological Chemistry 01/1997; 377(12):851-6. · 2.69 Impact Factor
  • Andree Klemm · Thomas Steiner · Uwe Flötgen · Gerhard A. Cumme · Anton Horn
  • [Show abstract] [Hide abstract]
    ABSTRACT: Digestion of calf intestine alkaline phosphatase with pronase and subsequent dephosphorylation of the released peptidyl-(Etn-P)2-glycosyl-PtdIns with HF generated 8 glycosyl-Ins species the largest of which (G1 and G2) have the following proposed structures: [sequence: see text] G3 and G5 are lower homologues of G1 and G2, respectively, being one alpha 1-2 linked mannopyranosyl residue shorter. G4 is analogous to G2 lacking the N-acetylgalactosaminyl residue and G6 is the next lower homologue of G4. Most of G4 and G6 occur substituted with a palmitoyl (G4, G6) or a myristoyl residue (G6) probably attached to the inositol moiety. Thus, the basic ManxGlc-Ins species are either substituted with an N-acetylgalactosaminyl residue or a fatty acid ester. The structures were deduced from compositional analysis, molecular-mass determination by matrix-assisted laser desorption MS, sequential hydrolysis with appropriate exoglycosidases and treatment with CrO3. Purification of the glycosylinositol species was achieved by a novel reverse-phase HPLC technique using fluorescent fluoren-9-yl-methoxy-carbonyl (Fmoc) derivatives. These stable derivatives were susceptible to hydrolysis with exoglycosidases which allowed sequential cleavages to be carried out and kinetics to be followed at the picomole level. We observed recently that native alkaline phosphatase separates on octyl-Sepharose into four distinct fractions of increasing hydrophobicity (F1-F4). Here we show that all four fractions contain G1-G6. The acylated species G4 and G6 were restricted to F2 and F4 which had been shown earlier to contain, on average, 2.5 and 3 fatty acid residues/subunit, respectively. In all four fractions the diradylglycerol moiety was predominantly diacylglycerol, alkylacylglycerol being less than 10% which is in contrast to most glycosyl-PtdIns--anchored proteins of mammalian origin.
    European Journal of Biochemistry 06/1996; 238(1):259-69. · 3.58 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: An electrophoretically homogeneous glycosylphosphatidylinositol- alkaline phosphatase fraction from calf intestine, obtained by hydrophobic chromatography, was used as "enzyme-labeled" substrate for testing phospholipase activity. The reaction products were separated by (i) hydrophobic chromatography in pipet tips and (ii) Triton X-114 phase partitioning. The chromatographic method presented permits high test frequencies, does not need temperature-controlled sample handling, and is only slightly disturbed by detergents, organic solvents, and proteins. The method was used to characterize phosphatidylinositol- specific phospholipase C from Bacillus cereus and phospholipase D from calf serum. Measurement of substrate hydrolysis by phospholipases is apparently linear to enzyme concentration and time. Relative activity of both enzymes is maximum at pH 6.5, corresponding to the optimal pH range found with other glycosylphosphatidylinositol substrates and phosphatidylinositol-specific phospholipases of other sources. Maximum activity of phospholipase C was found at 0.03% Triton X-100, 0.01% Brij 35, and 0.2% n-octylglucoside. The activity is not affected by Ca(2+), NaHCO(3), o-phenanthroline, or EDTA, increasingly inhibited by MgCl(2), MnCl(2), and ZnCl(2), and slightly activated by Na+ and K+. Calf serum phospholipase D shows maximum activity at 0.05% Triton X-100, 0.02% Brij 35, and 0.4% n-octylglucoside. The apparent Km values for phospholipase C (12.25 micron) and phospholipase D (4.94 micron) found with glycosylphosphatidylinositol-alkaline phosphatase are compared with values published for other glycosylphosphatidylinositol substrates.
    Analytical Biochemistry 11/1995; 231(1):99-108. DOI:10.1006/abio.1995.1508 · 2.22 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: Using the enzyme activity inhibiting monoclonal antibody IB 10B8 against alkaline phosphatase of calf intestine (AP), the interaction of a macromolecular antigen with the antibody was studied with different reaction conditions and with different conformations of the antigen, i.e. using (i) different pH values, (ii) different temperatures, (iii) different substrate saturation of the enzyme, (iv) different glycosylphosphatidyl-AP (GPI-AP) aggregates, and (v) membrane-bound species. In the case of antibody excess and negligible substrate consumption enzymic product formation proceeds according to [P] = a + b x t - c x exp(-d x t). By direct progress curve fitting and secondary data evaluation using nonlinear regression, omitting numerical derivation and graphic techniques, kinetic constants of the immune reaction have been estimated. The method does not require any artificial labelling nor any separation of bound and free entities. (i) Upon increasing pH from 9.8 to 11.0, the dissociation constant of the enzyme-antibody complex is increased strongly, mainly due to the decreasing association rate constant. (ii) A temperature increase from 25 degrees C to 37 degrees C produces a marked increase of both the association and dissociation rate constant. (iii) To differentiate between the interaction of the antibody with the free (E) and substrate-bound (ES) enzyme, experiments were done at different substrate concentrations. The results were fitted to a model allowing determination of association and dissociation rate constants of the free and substrate-bound enzyme. The inverse variation of association and dissociation rate constants caused by substrate binding produces a marked increase of the dissociation constant of the antibody-enzyme complex. The antibody-bound enzyme shows a nearly three-fold higher Km value and a six-fold lower catalytic constant as compared to the free enzyme. (iv) Investigations of the interaction of the antibody with anchorless AP, different hydrophobic aggregates of purified GPI-AP (fractions II-V). (v) Membrane-bound GPI-AP show that the epitopes of all species are fully accessible to the antibody and not cryptic. Surprisingly the insertion of the GPI-moiety into the membrane and the aggregation of the different GPI-AP fractions II-V seem to improve antibody binding. Such improvement of binding was not found in control experiments with Fab, indicating only for the bivalent antibody a stronger interaction with the multivalent antigen than with the monovalent antigen.
    Journal of Immunological Methods 06/1995; 182(1):29-39. DOI:10.1016/0022-1759(95)00015-3 · 2.01 Impact Factor

Publication Stats

240 Citations
72.10 Total Impact Points

Institutions

  • 1980–2008
    • Friedrich-Schiller-University Jena
      • • Institute of Biochemistry I and II
      • • Institute of Physical Chemistry
      Jena, Thuringia, Germany
  • 1996
    • Friedrich-Alexander Universität Erlangen-Nürnberg
      Erlangen, Bavaria, Germany