Structure of the Dispase Autolysis-inducing Protein from Streptomyces mobaraensis and Glutamine Cross-linking Sites for Transglutaminase*

Transglutaminase from Streptomyces mobaraensis (MTG) is an important enzyme for cross-linking and modifying proteins. An intrinsic substrate of MTG is the dispase autolysis-inducing protein (DAIP). The amino acid sequence of DAIP contains 5 potential glutamines and 10 lysines for MTG-mediated cross-linking. The aim of the study was to determine the structure and glutamine cross-linking sites of the first physiological MTG substrate. A production procedure was established in Escherichia coli BL21 (DE3) to obtain high yields of recombinant DAIP. DAIP variants were prepared by replacing four of five glutamines for asparagines in various combinations via site-directed mutagenesis. Incorporation of biotin cadaverine revealed a preference of MTG for the DAIP glutamines in the order of Gln-39 ≫ Gln-298 > Gln-345 ∼ Gln-65 ≫ Gln-144. In the structure of DAIP the preferred glutamines do cluster at the top of the seven-bladed β-propeller. This suggests a targeted cross-linking of DAIP by MTG that may occur after self-assembly in the bacterial cell wall. Based on our biochemical and structural data of the first physiological MTG substrate, we further provide novel insight into determinants of MTG-mediated modification, specificity, and efficiency.

absence of primary amines resulting in the formation of deamidated proteins. In the case of S. mobaraensis, substrate hydrolysis or intramolecular cross-linking are the preferred reactions of MTG in submerged cultures as shearing forces most likely prevent the attachment of a lysine donor substrate to the acylenzyme intermediate (20).
The active site of MTG has been engineered to accelerate protein modification by improving specificity (21). Heat-resistant and heat-sensitive MTG variants were both obtained by random mutagenesis (22). A variety of approaches has been pursued to unravel binding characteristics of glutamine and lysine donors using inhibitory proteins from streptomycetes (23), fluorescent derivatives of the artificial substrate N ␣benzyloxycarbonyl-L-glutaminylglycine (24), and diverse acyl acceptor molecules (25). The application of phage display and mRNA display techniques revealed glutamine consensus sequences for MTG that are quite different from those of the mammalian enzymes (26,27). However, the MTG cross-linking sites in physiological substrates have not been determined despite the fact that patterns of the tertiary structure can be more important for protein-protein interaction than the mere amino acid sequence (28).
We have purified and characterized several proteins from S. mobaraensis that are glutamine-and lysine-donor substrates of MTG (20, 29 -31). The dispase autolysis-inducing protein (DAIP, P84908) has a molecular mass of 37 kDa containing 5 glutamine and 10 lysine residues (32). Substrates are metalloproteases of the thermolysin family M4 (P00800) that undergo rapid autolysis in the presence of DAIP. A mechanism to trigger self-degradation has not yet been published, but proteolytic activity was excluded, and the influence of Ca 2ϩ and Zn 2ϩ on DAIP activity was negligibly small (29).
The aim of the present study was to determine the structure and glutamine cross-linking sites for the first physiological substrate of MTG and to contribute to more site-specific protein modification procedures. For this purpose, a production procedure for DAIP was established in Escherichia coli. DAIP glutamines were replaced by asparagines through consecutive sitedirected mutagenesis, keeping one of the five available residues in place. Additionally, we succeeded in determining the sevenbladed ␤-propeller structure of DAIP to 1.7 Å resolution and correlated the structural data with the mutagenesis data and efficiency of MTG-mediated biotin cadaverine incorporation. Enzymatic linkage of biotin cadaverine to the DAIP variants preferentially occurs on glutamine cross-linking sites of the microbial transglutaminase that cluster on one side of the DAIP ␤-propeller structure, suggesting a site directed and orientation-specific cross-linking.

Results and Discussion
Primary Structure of the DAIP-DAIP was originally purified from culture broth of S. mobaraensis (29). Determination of the N-terminal amino acid sequence suggested a novel protein, as the sequence was not related to any protein characterized so far. Because attempts to identify the DAIP-encoding gene using an appropriate DNA probe failed, the entire genome sequence of S. mobaraensis 40847 was determined to obtain the fulllength sequence ( Fig. 2A). In parallel, the genome sequence of the same strain was published by Yang et al. (32). Both genome sequences perfectly coincided and depicted the DAIP encoding gene without any deviation (not shown).
DAIP consists of 348 amino acids preceded by a 26-mer signal peptide ( Fig. 2A). Ala-Gln-Ala is most likely the binding site of the putative signal peptidase, which is in accordance with the experimentally determined N-terminal peptide sequence. The mature protein contains 5 glutamines and 10 lysines. Consensus sequences for MTG were not detectable in the potential glutamine and lysine cross-linking sites (Table 1). Moreover, amino acids, predicted by phage display and mRNA display libraries (26,27), were not found at the proposed positions referring to the central glutamines Gln-39, Gln-65, Gln-144, Gln-298, and Gln-345. The DAIP motifs contain neither aromatic amino acids (26) nor arginines in position Ϫ3 (27). Proline (27) and arginine (26) are not preferred amino acids for position ϩ1, and even double-Q pairs that have been found in random peptides as preferred substrate sites are missing (27). It thus seems plausible that the interaction of MTG and its physiological substrate DAIP is governed by yet unidentified properties directing efficient modification by MTG at the glutamine cross-linking sites.
Production of ⌬ Q -deficient DAIP Variants-C-terminally His 6 -tagged rDAIP turned out to be unsuitable for the characterization of glutamine binding sites, as modification of rDAIP with a C-terminal His 6 tag resulted in partial reduction in dispase-autolysis activity. We thus used untagged DAIP derivatives to avoid any impact of the tag on activity and MTG-mediated modification efficiency. Tag-free rDAIP was produced in E. coli using the optimized gene as described under "Experimental Procedures." The described procedure resulted in high yields of purified wt-rDAIP ( Fig. 2B) with similar activity compared with DAIP from S. mobaraensis (Sm-Q 5 ), suggesting that the overall conformation was not altered (Table 2).
Next, four of the five DAIP glutamines were substituted by asparagines via site-directed mutagenesis to yield Q 1 variants containing only a single glutamine residue. The ⌬ 4Q -deficient proteins are denoted as Gln-39, Gln-65, Gln-144, Gln-298, and Gln-345. Besides wt-rDAIP, still possessing five glutamines (Q 5 ), a null-Q variant (Q 0 ) was generated as control. All productions resulted in considerably lower, but sufficient amounts of modified rDAIP to perform the experiments described below (Table 2). Functional conformation of the ⌬ Q -deficient proteins was verified by DAIP activity. All tag-free variants destroyed dispase as fast as wt-DAIP ( Table 2).
Alkylation of Microbial Transglutaminase-Access of microbial transglutaminase to the still available glutamines of modified rDAIP was examined by labeling and blotting procedures (20,29). However, MTG and DAIP are comparable in size, and the enzyme MTG is capable of autocatalytic incorporation of lysine equivalents such as MBC. To distinguish unambiguously between biotinylated glutamine-deficient DAIP variants and the modifying enzyme, masking of MTG glutamines was essential. Carboxamide transamidation was the most appropriate method, as transglutaminases are thought not to react with alkylated glutamines (8). The MTG-mediated self-modification was studied by incorporation of several alkyl amines into the accessible glutamines. Amphiphilic N-lauroylsarcosine was added to enhance efficiency of MTG alkylation (20). After incubation for several hours and dialysis against water to remove excess of alkyl amines and N-lauroylsarcosine, MTG activity was comparable to non-alkylated MTG (supplemental Fig. S2  A). In contrast, self-biotinylation of alkylated MTG was considerably impaired, as indicated by significantly weakened bands (supplemental Fig. S2B). Because impact on self-biotinylation activity was comparable between the different alkylated MTG, methylated MTG (hereafter referred to as MTG met ) was used for subsequent labeling reactions.
MTG-specific Glutamine Cross-linking Sites-Sorting of DAIP glutamines was carried out by MTG met -mediated incorporation of MBC into tag-free ⌬ Q -deficient variants. Labeling mixtures were separated by electrophoresis, blotted if necessary, and stained by a fluorescent streptavidin conjugate, Coomassie, and polyclonal antibodies as described under "Experimental Procedures." The Gln-39, Gln-298, and Gln-345 variants showed intensively stained biotinylated proteins, identifying these glutamine residues as preferred cross-linking sites of MTG (Fig.  2C). Moreover, it became obvious that MTG has access to Gln-65 and Gln-144 as well, at least under the conditions used. In contrast, the null-Q variant could not be detected by the fluorescent probe, excluding unspecific incorporation of MBC.
Labeling Kinetics-The tag-free DAIP variants were also studied to determine semiquantitatively the rate of MBC incorporation by MTG at pH 8.0. Highly purified DAIP from Sm-Q 5 , the null-Q variant (Q 0 ), and a ␤-glucosidase from almonds were used as controls. Reaction mixtures contained 8 M DAIP and 0.2 mM MBC. The reaction was started by the addition of 0.5 M MTG met , and various samples were taken in the first hour and after incubation overnight (Fig. 2D). Constantly increasing flu-orescent bands, already indicated after 3 min, showed that Gln-39 was biotinylated almost as fast as the wild type protein Sm-Q 5 . In contrast, significantly stained protein bands of Gln-298 and Gln-345 only emerged when the incubation mixtures were allowed to react for at least 30 min. MTG-mediated linkage of the biotinylated amine to Gln-65 and Gln-144 obviously occurred with even lower velocity, and the controls Q 0 and ␤-glucosidase clearly revealed the absence of any cross-linking site for bacterial transglutaminase. The latter enzyme was not labeled by MTG despite 14 glutamines, of which at least 10 are solvent-accessible in a predicted structure model based on the structure of the related ␤-glucosidase from Oryza sativa (Fig. 3, A and B). Weak bands of the control proteins became only visible after incubation overnight.
There was a linear increase in the degree of labeling over time as the emission intensity of the fluorescent streptavidin conjugate suggested (Fig. 2E). The highest biotinylation rate was monitored for DAIP from S. mobaraensis (Sm-Q 5 ), corresponding with an increase in fluorescence of 278 rfu/min after protein blotting and streptavidin conjugate treatment. Enzymatic linkage of MBC to the DAIP variants was slower, but Gln-39 was the best cross-linking site by far. The labeling rate of Gln-39 was equivalent with a fluorescence increase of 159 rfu/min and reached about 57% that of Sm-Q 5 . Assuming similar protein blotting and biotin orientation, preference of MTG for the DAIP glutamines was on the order of Gln-39 (159 rfu/min, 57.2%) Ͼ Ͼ Gln-298 (48 rfu/min, 17.5%) Ͼ Gln-345 (33 rfu/min, 12.0%) ϳ Gln-65 (33 rfu/min, 11.8%) Ͼ Ͼ Gln-144 (5 rfu/min, 1.8%).
Molecular Structure of DAIP-To gain more insights into the overall structure, the location of glutamines and lysines in DAIP as well as to identify potential structural properties governing efficient modification by MTG, we determined the crystal structure of DAIP to 1.7 Å resolution by x-ray crystallography ( Fig. 4 and supplemental Table S2). The structure was solved by ytterbium single-wavelength anomalous diffraction phasing and comprises two molecules in the asymmetric unit with residues 2-346 built from a total of 348 residues, with a cis-peptide bond between Ser-40 and Val-41. The structure of DAIP shows the classical seven-bladed ␤-propeller (Fig. 4A). In contrast to the blades I-VI, which contain four sequential ␤-strands A-D, the last blade VII is composed of two ␤-strands derived from the C terminus (A-B, amino acids 326 -340) and one from the N terminus (C, amino acids 20 -25). These two elements from both the N and C terminus interlock and form blade VII, thus closing and stabilizing the ring-shaped propeller structure. The DAIP structure is further stabilized by two cysteines, Cys-12 and Cys-72, which form a single disulfide bridge covalently linking the N-terminal tail at position 12 to the ␤-turn between C I and D I (amino acids 68 -74) of the first propeller blade. Four elongated loops (D IV -A V , BC V , BC VI , and C VII -A I ) are found on the top of the propeller (Fig. 4B), two of which, namely C VII -A I and BC VI , contain Gln-39 and Gln-298, respectively.
The structure provides first insights into the orientation of the MTG binding sites. Interestingly, four of the five glutamines are located on the same side in close proximity (Fig. 4, A and E). The 10 lysines of DAIP are instead scattered over the whole  surface of the propeller without showing a preferred surface location. Glutamines are in general well resolved in the 2F o Ϫ F c density, except for Gln-345, which shows weaker electron density for the side chain, likely due to the intrinsic flexibility of the C-terminal tail of DAIP (Fig. 4C). In contrast to the solvent-exposed and accessible residues Gln-39, Gln-298, and Gln-345, the two glutamines Gln-65 and Gln-144 are found in ␤-strands C I and C III , respectively, and are less solventaccessible (Fig. 4E).
DAIP Structure Versus Kinetic Data-The kinetic results largely correlate with the accessibility of glutamine positions found in the DAIP structure. The two fastest modified glutamines, namely Gln-39 and Gln-298, are presented on loops C VII -A I and BC VI , respectively. Those are located on top of the seven-bladed ␤-propeller and are the most solvent-accessible glutamines (Fig. 4, C-E). However, Gln-345 is also solvent-accessible and presented on the C-terminal loop, but modification efficiency by MTG is at comparably low levels as for the buried Gln-65. The least efficiently modified substrate, Gln-144, is on the opposite side of the propeller in reference to the four clustered glutamines. In search for structural determinants for specificity and modification efficiency, we identified Gln-65 and Gln-144 to be located on ␤-strands and, hence, to be less solvent-accessible, whereas Gln-39, Gln-298, and Gln-345 are located in loop regions and solvent-exposed. The location in ␤-strands and consequently the reduced flexibility and accessibility are likely to be partially responsible for the low labeling efficiency. As previously shown, the active site residues of MTG are located in a cleft, which can be accessed from two sides, of which the front vestibule, which is in close proximity to the catalytic cysteine, is mainly hydrophobic with patches of weak positive charge, whereas the back vestibule is strongly negatively charged (Fig. 5). The position of the catalytic cysteine residue and charge distribution in the front vestibule suggests an accommodation of the acyl donor (glutamine substrate), whereas the negative charge and presence of the catalytic His/ Asp suggests accommodation of the positively charged acyl acceptor (lysine substrate), resulting in the formation of a ter-nary complex (9). Thus, in line with the previous suggestion by Kashiwagi et al. (9), we observe that, compared with substrate glutamines in ␤-strands, glutamines in an accessible and flexible loop are more efficiently modified, which are likely to be more easily accommodated in the cleft to reach the active site cysteine.
However, because DAIP is a purely ␤-strand protein, we need structural and biochemical characterization of substrates with glutamines in ␣-helical secondary structure elements to analyze the impact of helical structures on MTG-mediated modification. To identify additional structural features that may guide efficient modification of Gln-39 and Gln-298 in particular, but also Gln-345, we generated a structural overlay of the three glutamines present in loop regions using the main chain and C ␤ -atoms (Fig. 4D). By comparing the conformation of the glutamines and the two residues flanking both sides, we identified similarities in the overall conformation of the Gln-39 and Gln-345. However, the conformation of Gln-298 was significantly different, whereas modification efficiency was shown to be between that for Gln-39 and Gln-345. This suggests that flexibility and accessibility are a prerequisite for efficient modification but are modulated by additional properties of the substrate sites. We thus analyzed and compared the sequence and amino acid composition (Ϯ5 amino acids) of the glutamine sites surrounding Gln-39, Gln-298, and Gln-345 (Table 1 and Fig. 4F). Gln-39 is mainly flanked by small polar or hydrophobic residues (with the exception of one aromatic residue at position ϩ4) lacking charged residues. Gln-298 is flanked by a total of three aromatic/bulky residues, whereas Gln-345 is preceded by two negatively charged glutamic acids. In agreement with compatibility to the mostly hydrophobic front vestibule, the Gln-39 site with mainly small polar or hydrophobic amino acids (lacking charged residues) is most efficiently modified, whereas acidic, or in general charged residues, are apparently the least favorable. The presence of bulky/aromatic residues in close proximity does reduce the efficiency compared with small polar or hydrophobic residues but seems to be more favorable than the presence of charged residues. This is also in line with the . ␤-Glucosidase is colored in light teal in ribbon mode. 13 of 15 glutamines (Gln-4 is located in the signal peptide and is, therefore, removed from the mature protein; Gln-543 is not included in the Phyre2 model) are highlighted in orange, of which at least 10 are predicted to be solvent accessible. N and C termini are highlighted with a blue and red sphere, respectively. C, flanking residues of the 14 glutamines of ␤-glucosidase. Aromatic/bulky residues are colored yellow, and positively and negatively charged residues are in blue and red, respectively. The only glutamine that partially meets our postulated requirements for MTG modification is Gln-272, which however still contains a bulky/aromatic amino acid at position Ϫ1 adjacent to Gln. inefficient modification of ␤-glucosidase, where glutamines located in predicted loop regions are flanked by aromatic/bulky and/or charged amino acids (Fig. 3C). Two DAIP variants further supported our conclusions. To mimic the N-sided flank of Gln-345 and Gln-298, Gly-36/Thr-37 and Thr-37/Leu-38 of Gln-39 (Q 1 variant) were replaced for two glutamates or tyrosine/phenylalanine, respectively. Although insertion of the negative charges reduced the biotinylation rate by 97%, the effect of the bulky/aromatic amino acids was slightly smaller, with reduction in velocity by 85% compared with the unmodified Gln-39 variant (Fig. 2, D and E). Interestingly, we were able to use the native MTG recognition site Gln-298 positioned in the ␤-hairpin for the design of an engineered MTG-tag to efficiently conjugate a reporter cargo onto a therapeutic monoclonal antibody (2). In this study the ␤-hairpin was mimicked by the use of a disulfide bridge, which significantly improved modification efficiency by MTG compared with the variant lacking the disulfide bridge. In our biochemical data we observed a 3-fold higher cross-linking activity for Gln-39 compared with Gln-298, suggesting that the use of the Gln-39 sequence in combination with the ␤-hairpin structure might further improve efficiency of that system. Conclusions-In this work recombinant production of the DAIP and its glutamine-deficient variants allowed the determination of the first natural glutamine-binding sites for MTG from S. mobaraensis. The sequence motifs containing Gln-39, Gln-298, and Gln-345 clearly differed from those predicted by phage display and mRNA display peptide libraries (26,27). Specifically, aromatic amino acids (position Ϫ3), arginines (Ϫ3, ϩ1), and proline (ϩ1) are not available in the DAIP glutamine motifs. Glu pairs are also missing. Thus, small peptides appear to reflect the properties of natural MTG modification sites insufficiently, suggesting that for physiological substrates such as DAIP, approach of MTG to glutamine sites must be completely different. Therefore, in addition to the experimental determination of natural glutamine-binding sites, we determined the seven-bladed ␤-propeller structure of the first physiological MTG substrate DAIP to 1.7 Å resolution by x-ray crystallography. Interestingly, four glutamines, which include the three efficiently modified glutamines, cluster on one side of the propeller, giving rise to the notion that cross-linking of DAIP occurs in a site-directed and orientation-specific manner. We further provide evidence that solvent accessibility and location outside of ␤-strands is a prerequisite and main determinant for efficient modification of glutamines in DAIP. Combining the structural data with the determined modification efficiencies, sequence of DAIP substrate glutamine sites, and data from mutagenesis experiments, we propose that the properties of amino acids flanking the Q-sites favor efficient modification in the following order: small/polar/hydrophobic/uncharged Ͼ Ͼ bulky/aromatic Ͼ charged. However, additional structural and biochemical experiments are needed in the future to identify more natural MTG sites and their properties to further validate and improve our current model regarding the influence of amino acid composition of physiological glutamine sites on MTG-mediated modification efficiency.
Codon Optimization, Plasmids, Cloning, Transformation, and Culture-The Streptomyces gene encoding mature DAIP was adapted to E. coli codon usage by GenScript (Hongkong, China) to improve gene translation (supplemental Fig. S1). The sequence encoding daip opt was obtained by NcoI and XhoI hydrolysis of the shuttle vector. The purified fragment was then ligated into plasmid pET22b(ϩ), and chemically competent E. coli XL 1 blue cells were transformed with the resulting vector construct. Positive clones containing daip opt were identified by colony PCR and sequence analysis. For the production of FIGURE 5. Glutamine binding in the front vestibule of MTG is likely to be controlled by flexibility and amino acid composition of the substrate sequence. A-C, electrostatic surface representation of transglutaminase from S. mobaraensis (PDB entry 1IU4). The front vestibule (black triangle) carries a weak positively charged patch, which is likely to bind the acyl donor (glutamine substrate). The back vestibule (white triangle) is strongly negatively charged and is likely to accommodate the positively charged acyl acceptor (lysine substrate). D, catalytic triad residues (Cys-64, Asp-255, His-274) positions are superposed in the close-up. The front vestibule is narrowed toward the active site cysteine, suggesting that substrate glutamines in flexible loops are more easily accommodated and reach the active site cysteine for efficient modification. In addition to flexibility, small/polar/hydrophobic/uncharged residues are preferred for efficient modification, whereas bulky/aromatic and/or charged residues are likely to interfere with accommodation of the substrate glutamine in the cleft, resulting in inefficient MTG-mediated modification.
recombinant rDAIP as fusion protein, the coding sequence was cloned via the restriction sites NdeI and XhoI into pET21a-CPD-His 10 bearing the His-tagged cysteine protease domain of Vibrio cholerae (35). E. coli BL21 (DE3) pLysS was transformed with the construct using the heat-shock procedure. The cells were cultured at 37°C overnight. Production was carried out overnight using 1 mM isopropyl 1-thio-␤-D-galactopyranoside in 1 liter of LB/SB amp at 28°C. Utilization of a C-terminal His 10 tag allowed purification by immobilized metal ion affinity chromatography.
Site-directed Mutagenesis-Mutagenesis was performed by PCR using the respective primer pairs (supplemental Table S1), daip opt in pET-22b(ϩ), and Pfu DNA polymerase (Thermo Scientific) or Q5-HF DNA polymerase (New England BioLabs). Upon digestion of parental DNA by DpnI for 1 h at 37°C, E. coli XL 1 blue cells were transformed with the resulting constructs daip opt in pET21a-CPD-His 10 . Successive mutagenesis was carried out until four out of five glutamine codons were replaced by asparagine codons. A control plasmid contained daip opt without any residual glutamine codon (Q 0 ).
Purification of rDAIP and Variants-Induced E. coli BL21 (DE3) pLysS cells containing rDAIP were precipitated by centrifugation (17,000 ϫ g, 15 min, 4°C) and lysed by sonication or by high pressure homogenization (2.72 kilobars). Cellular debris was removed by centrifugation (20,000 ϫ g, 30 min, 4°C). Purification was performed by immobilized metal ion affinity chromatography using chelating-Sepharose Fast Flow (column volume: 7 ml) and 15 mM imidazole in 50 mM Tris/ HCl, pH 8.0, with 300 mM NaCl. For the purification of the tag-free variants, elution of the cleaved target protein was performed after incubation for 60 -90 min with 50 M phytic acid and subsequent release of CPD by a linear imidazole gradient of 15-500 mM. Fractions were separated by 12.5% SDS-polyacrylamide gel electrophoresis and Coomassie-stained. The combined immobilized metal ion affinity chromatography fractions were separated by Fractogel EMD SO 3 Ϫ , phenyl-Sepharose, and size exclusion chromatography.
DAIP Activity Measurements-The activity of DAIP was determined indirectly using several protease assays (29,34). Upon preincubation with dispase, residual protease activity was monitored by hydrolysis of casein, azocasein, or dabSFans.
For continuously running measurements, DAIP (0 -200 nM), dispase (100 -200 nM), and 2 mM CaCl 2 in 80 l of 50 mM Tris/ HCl, pH 7.5, were preincubated for 30 min at 37°C before 80 l of substrate solution containing 86.6 M dabSFans was added. The increase in fluorescence was continuously monitored at 535 nm ( exc of 340 nm) using a Genios SpectraFluor plus fluorimeter (Tecan) or Polarstar multi-mode microplate reader (BMG Labtech). The data, obtained in triplicate, were evaluated by weighted least square minimization using GraphPad Prism.
Alkylation of Microbial Transglutaminase-Purified transglutaminase from S. mobaraensis (10.3 M), 1 mM N-lauroylsarcosine, 500 M methylamine in ethanol (or ethylamine and ethanolamine) in 50 mM HEPES, pH 8, were allowed to react up to 17 h at 37°C under shaking (300 rpm). Reagents were removed by dialysis against water.
MTG-mediated Labeling Procedure and Kinetics-Labeling of DAIP was performed according to Schmidt et al. (20) with minor modifications. DAIP or its variants (8 M) and 0.2 mM MBC in 100 mM HEPES, pH 8, were incubated with 0.5 M MTG at 37°C for up to 16 h (Fig. 2C) or the indicated time (Fig.  2D). DAIP without MBC or MTG were the controls. The reaction mixtures were separated by SDS-PAGE (12.5%) and, where applicable, blotted onto PVDF membranes for DAIP detection and biotin signaling. Coomassie Blue (Coomassie) or primary polyclonal rabbit antibodies against DAIP and secondary goat IRDye 680LT anti-rabbit IgG (DAIP-IgG) were used for the detection of protein in general or DAIP. The degree of biotinylation (Biotin) was determined using fluorescent IRDye 800CW streptavidin conjugates (33 ng/ml) in combination with the Odyssey Sa Infrared Imaging System (LI-COR Biotechnology, Bad Homburg, Germany). Fluorescence was monitored at 693 nm ( exc of 676 nm) and 789 nm ( exc of 774 nm) for IRDye 680LT (DAIP-IgG) and IRDye 800CW (Biotin), respectively. Equal areas of the fluorescent protein bands were integrated to evaluate biotinylation rates, expressed in terms of rfu/min, and corrected by the amount of unspecific streptavidin binding using the Q 0 variant as a reference. Rates for mutants Gln-39-G36E/T37E and Gln-39-T37Y/L38F were normalized using the data for Gln-39.
Crystallization and Data Processing-Purified protein from S. mobaraensis culture broth (as described in Sarafeddinov et al. (29)) was used for crystallization. DAIP crystals grew in 15% (v/v) glycerol, 8.5% (v/v) 2-propanol, 85 mM HEPES, pH 7.5, and 17% (w/v) PEG 4k in drops with an equal volume of precipitant and protein solution at a concentration of 10 mg/ml. The crystals reached full size (ϳ120 ϫ 130 ϫ 100 m) after 1 week at 20°C. Crystals were soaked with 100 mM Yb-DOTA according to the manufacturer's recommendations (Jena Biosciences) and cryo-protected using mother liquor. A ytterbium single-wavelength anomalous diffraction data set was collected at a wavelength of 1.386 Å on a PILATUS 6M detector at BL14.1 (HZB-BESSYII). Data reduction and scaling was carried out with the XDS package (36). Phasing and initial model building was done in HKL2MAP (37). Coot (38) with iterative refinement cycles in Phenix.refine (39) has been used for manual model building. The figures were created using PyMOL.
Additional Procedures-The production of MTG and DAIP from S. mobaraensis, SDS-PAGE, Coomassie, silver staining of polyacrylamide gels, MTG-mediated biotinylation, and blotting procedures using streptavidin-alkaline phosphatase conjugates or DAIP-specific antibodies as well as determination of protein content were performed as described (20, 29 -31, 33). The three-dimensional structure prediction server Phyre2 (40) was used for generating a model of the ␤-glucosidase.