Chemical Tagging of Membrane Proteins Enables Oriented Binding on Solid Surfaces.

In biological systems, membrane proteins play major roles in energy conversion, transport, sensing, and signal transduction. Of special interest are the photosynthetic reaction centers involved in the initial process of light energy conversion to electrical and chemical energies. The oriented binding of membrane proteins to solid surfaces is important for biotechnological applications. In some cases, novel properties are generated as a result of the interaction between proteins and solid surfaces. We developed a novel approach for the oriented tagging of membrane proteins. In this unique process, bifunctional molecules are used to chemically tag the exposed surfaces of membrane proteins at selected sides of membrane vesicles. The isolated tagged membrane proteins were self-assembled on solid surfaces, leading to the fabrication of dens-oriented layers on metal and glass surfaces, as seen from the atomic force microscopy (AFM) images. In this work, we used chromatophores and membrane vesicles containing protein chlorophyll complexes for the isolation of the bacterial reaction center and photosystem I, from photosynthetic bacteria and cyanobacteria, respectively. The oriented layers, which were fabricated on metal surfaces, were functional and generated light-induced photovoltage that was measured by the Kalvin probe apparatus. The polarity of the photovoltage depended on the orientation of proteins in the layers. Other membrane proteins can be tagged by the same method. However, we preferred the use of reaction centers because their orientation can be easily detected by the polarity of their photovoltages.

Dissection of membrane-binding and -remodeling regions in two classes of bacterial phospholipid N-methyltransferases.

Bacterial phospholipid N-methyltransferases (Pmts) catalyze the formation of phosphatidylcholine (PC) via successive N-methylation of phosphatidylethanolamine (PE). They are classified into Sinorhizobium-type and Rhodobacter-type enzymes. The Sinorhizobium-type PmtA protein from the plant pathogen Agrobacterium tumefaciens is recruited to anionic lipids in the cytoplasmic membrane via two amphipathic helices called αA and αF. Besides its enzymatic activity, PmtA is able to remodel membranes mediated by the αA domain. According to the Heliquest program, αA- and αF-like amphipathic helices are also present in other Sinorhizobium- and Rhodobacter-type Pmt enzymes suggesting a conserved architecture of α-helical membrane-binding regions in these methyltransferases. As representatives of the two Pmt families, we investigated the membrane binding and remodeling capacity of Bradyrhizobium japonicum PmtA (Sinorhizobium-type) and PmtX1 (Rhodobacter-type), which act cooperatively to produce PC in consecutive methylation steps. We found that the αA regions in both enzymes bind anionic lipids similar to αA of A. tumefaciens PmtA. Membrane binding of PmtX1 αA is enhanced by its substrate monomethyl-PE indicating a substrate-controlled membrane association. The αA regions of all investigated enzymes remodel spherical liposomes into tubular filaments suggesting a conserved membrane-remodeling capacity of bacterial Pmts. Based on these results we propose that the molecular details of membrane-binding and remodeling are conserved among bacterial Pmts.

Falsirhodobacter sp. alg1 Harbors Single Homologs of Endo and Exo-Type Alginate Lyases Efficient for Alginate Depolymerization.

Alginate-degrading bacteria play an important role in alginate degradation by harboring highly efficient and unique alginolytic genes. Although the general mechanism for alginate degradation by these bacteria is fairly understood, much is still required to fully exploit them. Here, we report the isolation of a novel strain, Falsirhodobacter sp. alg1, the first report for an alginate-degrading bacterium from the family Rhodobacteraceae. Genome sequencing reveals that strain alg1 harbors a primary alginate degradation pathway with only single homologs of an endo- and exo-type alginate lyase, AlyFRA and AlyFRB, which is uncommon among such bacteria. Subsequent functional analysis showed that both enzymes were extremely efficient to depolymerize alginate suggesting evolutionary interests in the acquirement of these enzymes. The exo-type alginate lyase, AlyFRB in particular could depolymerize alginate without producing intermediate products making it a highly efficient enzyme for the production of 4-deoxy-L-erythro-5-hexoseulose uronic acid (DEH). Based on our findings, we believe that the discovery of Falsirhodobacter sp. alg1 and its alginolytic genes hints at the potentiality of a more diverse and unique population of alginate-degrading bacteria.

Computationally Efficient Multiscale Reactive Molecular Dynamics to Describe Amino Acid Deprotonation in Proteins.

An important challenge in the simulation of biomolecular systems is a quantitative description of the protonation and deprotonation process of amino acid residues. Despite the seeming simplicity of adding or removing a positively charged hydrogen nucleus, simulating the actual protonation/deprotonation process is inherently difficult. It requires both the explicit treatment of the excess proton, including its charge defect delocalization and Grotthuss shuttling through inhomogeneous moieties (water and amino residues), and extensive sampling of coupled condensed phase motions. In a recent paper (J. Chem. Theory Comput. 2014, 10, 2729-2737), a multiscale approach was developed to map high-level quantum mechanics/molecular mechanics (QM/MM) data into a multiscale reactive molecular dynamics (MS-RMD) model in order to describe amino acid deprotonation in bulk water. In this article, we extend the fitting approach (called FitRMD) to create MS-RMD models for ionizable amino acids within proteins. The resulting models are shown to faithfully reproduce the free energy profiles of the reference QM/MM Hamiltonian for PT inside an example protein, the ClC-ec1 H(+)/Cl(-) antiporter. Moreover, we show that the resulting MS-RMD models are computationally efficient enough to then characterize more complex 2-dimensional free energy surfaces due to slow degrees of freedom such as water hydration of internal protein cavities that can be inherently coupled to the excess proton charge translocation. The FitRMD method is thus shown to be an effective way to map ab initio level accuracy into a much more computationally efficient reactive MD method in order to explicitly simulate and quantitatively describe amino acid protonation/deprotonation in proteins.

How Cellulose Elongates--A QM/MM Study of the Molecular Mechanism of Cellulose Polymerization in Bacterial CESA.

The catalytic mechanism of bacterial cellulose synthase was investigated by using a hybrid quantum mechanics and molecular mechanics (QM/MM) approach. The Michaelis complex model was built based on the X-ray crystal structure of the cellulose synthase subunits BcsA and BcsB containing a uridine diphosphate molecule and a translocating glucan. Our study identified an SN2-type transition structure corresponding to the nucleophilic attack of the nonreducing end O4 on the anomeric carbon C1, the breaking of the glycosidic bond C1-O1, and the transfer of proton from the nonreducing end O4 to the general base D343. The activation barrier found for this SN2-type transition state is 68 kJ/mol. The rate constant of polymerization is estimated to be ∼8.0 s(-1) via transition state theory. A similar SN2-type transition structure was also identified for a second glucose molecule added to the growing polysaccharide chain, which aligned with the polymer 180° rotated compared to the initially added unit. This study provides detailed insights into how cellulose is extended by one glucose molecule at a time and how the individual glucose units align into cellobiose repeating units.

Crystallographic snapshot of cellulose synthesis and membrane translocation.

Cellulose, the most abundant biological macromolecule, is an extracellular, linear polymer of glucose molecules. It represents an essential component of plant cell walls but is also found in algae and bacteria. In bacteria, cellulose production frequently correlates with the formation of biofilms, a sessile, multicellular growth form. Cellulose synthesis and transport across the inner bacterial membrane is mediated by a complex of the membrane-integrated catalytic BcsA subunit and the membrane-anchored, periplasmic BcsB protein. Here we present the crystal structure of a complex of BcsA and BcsB from Rhodobacter sphaeroides containing a translocating polysaccharide. The structure of the BcsA-BcsB translocation intermediate reveals the architecture of the cellulose synthase, demonstrates how BcsA forms a cellulose-conducting channel, and suggests a model for the coupling of cellulose synthesis and translocation in which the nascent polysaccharide is extended by one glucose molecule at a time.

Crystallization and preliminary crystallographic studies of FoxE from Rhodobacter ferrooxidans SW2, an Fe(II) oxidoreductase involved in photoferrotrophy.

FoxE is a protein encoded by the foxEYZ operon of Rhodobacter ferrooxidans SW2 that is involved in Fe(II)-based anoxygenic photosynthesis (`photoferrotrophy'). It is thought to reside in the periplasm, where it stimulates light-dependent Fe(II) oxidation. It contains 259 residues, including two haem c-binding motifs. As no three-dimensional model is available and there is no structure with a similar sequence, crystals of FoxE were produced. They diffracted to 2.44 Å resolution using synchrotron radiation at the Fe edge. The phase problem was solved by SAD using SHELXC/D/E and the experimental maps confirmed the presence of two haems per molecule.

Functional characterization of the FoxE iron oxidoreductase from the photoferrotroph Rhodobacter ferrooxidans SW2.

Photoferrotrophy is presumed to be an ancient type of photosynthetic metabolism in which bacteria use the reducing power of ferrous iron to drive carbon fixation. In this work the putative iron oxidoreductase of the photoferrotroph Rhodobacter ferrooxidans SW2 was cloned, purified, and characterized for the first time. This protein, FoxE, was characterized using spectroscopic, thermodynamic, and kinetic techniques. It is a c-type cytochrome that forms a trimer or tetramer in solution; the two hemes of each monomer are hexacoordinated by histidine and methionine. The hemes have positive reduction potentials that allow downhill electron transfer from many geochemically relevant ferrous iron forms to the photosynthetic reaction center. The reduction potentials of the hemes are different and are cross-assigned to fast and slow kinetic phases of ferrous iron oxidation in vitro. Lower reactivity was observed at high pH and may contribute to prevent ferric iron precipitation inside or at the surface of the cell. These results help fill in the molecular details of a metabolic process that likely contributed to the deposition of precambrian banded iron formations, globally important sedimentary rocks that are found on every continent today.

Carotenogenesis gene cluster and phytoene desaturase catalyzing both three- and four-step desaturations from Rhodobacter azotoformans.

A carotenogenesis gene cluster from the purple nonsulfur photosynthetic bacterium Rhodobacter azotoformans CGMCC 6086 was cloned. A total of eight carotenogenesis genes ( crtA , crtI , crtB , tspO , crtC , crtD , crtE , and crtF ) were located in two separate regions within the genome, a 4.9 kb region containing four clustered genes of crtAIB - tspO and a 5.3 kb region containing four clustered genes of crtCDEF . The organization was unusual for a carotenogenesis gene cluster in purple photosynthetic bacteria. A gene encoding phytoene desaturase ( CrtI ) from Rba. azotoformans was expressed in Escherichia coli. The recombinant CrtI could catalyze both three- and four-step desaturations of phytoene to produce neurosporene and lycopene, and the relative contents of neurosporene and lycopene formed by CrtI were approximately 23% and 75%, respectively. Even small amounts of five-step desaturated 3,4-didehydrolycopene could be produced by CrtI . This product pattern was novel because CrtI produced only neurosporene leading to spheroidene pathway in the cells of Rba. azotoformans. In the in vitro reaction, the relative content of lycopene in desaturated products increased from 19.6% to 62.5% when phytoene reduced from 2.6 to 0.13 µM. The results revealed that the product pattern of CrtI might be affected by the kinetics.

Biosynthesis of caffeic acid in Escherichia coli using its endogenous hydroxylase complex.

BACKGROUND: Caffeic acid (3,4-dihydroxycinnamic acid) is a natural phenolic compound derived from the plant phenylpropanoid pathway. Caffeic acid and its phenethyl ester (CAPE) have attracted increasing attention for their various pharmaceutical properties and health-promoting effects. Nowadays, large-scale production of drugs or drug precursors via microbial approaches provides a promising alternative to chemical synthesis and extraction from plant sources. RESULTS: We first identified that an Escherichia coli native hydroxylase complex previously characterized as the 4-hydroxyphenylacetate 3-hydroxylase (4HPA3H) was able to convert p-coumaric acid to caffeic acid efficiently. This critical enzymatic step catalyzed in plants by a membrane-associated cytochrome P450 enzyme, p-coumarate 3-hydroxylase (C3H), is difficult to be functionally expressed in prokaryotic systems. Moreover, the performances of two tyrosine ammonia lyases (TALs) from Rhodobacter species were compared after overexpression in E. coli. The results indicated that the TAL from R. capsulatus (Rc) possesses higher activity towards both tyrosine and L-dopa. Based on these findings, we further designed a dual pathway leading from tyrosine to caffeic acid consisting of the enzymes 4HPA3H and RcTAL. This heterologous pathway extended E. coli native tyrosine biosynthesis machinery and was able to produce caffeic acid (12.1 mg/L) in minimal salt medium. Further improvement in production was accomplished by boosting tyrosine biosynthesis in E. coli, which involved the alleviation of tyrosine-induced feedback inhibition and carbon flux redirection. Finally, the titer of caffeic acid reached 50.2 mg/L in shake flasks after 48-hour cultivation. CONCLUSION: We have successfully established a novel pathway and constructed an E. coli strain for the production of caffeic acid. This work forms a basis for further improvement in production, as well as opens the possibility of microbial synthesis of more complex plant secondary metabolites derived from caffeic acid. In addition, we have identified that TAL is the rate-limiting enzyme in this pathway. Thus, exploration for more active TALs via bio-prospecting and protein engineering approaches is necessary for further improvement of caffeic acid production.

Redesign, reconstruction, and directed extension of the Brevibacterium linens C40 carotenoid pathway in Escherichia coli.

In this study, the carotenoid biosynthetic pathways of Brevibacterium linens DSMZ 20426 were reconstructed, redesigned, and extended with additional carotenoid-modifying enzymes of other sources in a heterologous host Escherichia coli. The modular lycopene pathway synthesized an unexpected carotenoid structure, 3,4-didehydrolycopene, as well as lycopene. Extension of the novel 3,4-didehydrolycopene pathway with the mutant Pantoea lycopene cyclase CrtY(2) and the Rhodobacter spheroidene monooxygenase CrtA generated monocyclic torulene and acyclic oxocarotenoids, respectively. The reconstructed beta-carotene pathway synthesized an unexpected 7,8-dihydro-beta-carotene in addition to beta-carotene. Extension of the beta-carotene pathway with the B. linens beta-ring desaturase CrtU and Pantoea beta-carotene hydroxylase CrtZ generated asymmetric carotenoid agelaxanthin A, which had one aromatic ring at the one end of carotene backbone and one hydroxyl group at the other end, as well as aromatic carotenoid isorenieratene and dihydroxy carotenoid zeaxanthin. These results demonstrate that reconstruction of the biosynthetic pathways and extension with promiscuous enzymes in a heterologous host holds promise as a rational strategy for generating structurally diverse compounds that are hardly accessible in nature.

Revealing linear aggregates of light harvesting antenna proteins in photosynthetic membranes.

How light energy is harvested in a natural photosynthetic membrane through energy transfer is closely related to the stoichiometry and arrangement of light harvesting antenna proteins in the membrane. The specific photosynthetic architecture facilitates a rapid and efficient energy transfer among the light harvesting proteins (LH2 and LH1) and to the reaction center. Here we report the identification of linear aggregates of light harvesting proteins, LH2, in the photosynthetic membranes under ambient conditions by using atomic force microscopy (AFM) imaging and spectroscopic analysis. Our results suggest that the light harvesting protein, LH2, can exist as linear aggregates of 4 +/- 2 proteins in the photosynthetic membranes and that the protein distributions are highly heterogeneous. In the photosynthetic membranes examined in our measurements, the ratio of the aggregated to the nonaggregated LH2 proteins is about 3:1 to 5:1 depending on the intensity of the illumination used during sample incubation and on the bacterial species. AFM images further identify that the LH2 proteins in the linear aggregates are monotonically tilted at an angle 4 +/- 2 degrees from the plane of the photosynthetic membranes. The aggregates result in red-shifted absorption and emission spectra that are measured using various mutant membranes, including an LH2 knockout, LH1 knockout, and LH2 at different population densities. Measuring the fluorescence lifetimes of purified LH2 and LH2 in membranes, we have observed that the LH2 proteins in membranes exhibit biexponential lifetime decays whereas the purified LH2 proteins gave single exponential lifetime decays. We attribute that the two lifetime components originate from the existence of both aggregated and nonaggregated LH2 proteins in the photosynthetic membranes.

Crystal structure of a protein, structurally related to glycosyltransferases, encoded in the Rhodobacter blasticus atp operon.

The F1-ATP synthase atp operon in the proteobacterium Rhodobacter blasticus contains six open reading frames, encoding six hypothetical proteins. Five of these subunits, in the stoichiometry (alphabeta)3gamma delta epsilon make up the catalytic F1-ATP synthase complex similarly in bacteria, chloroplasts and mitochondria. The sixth gene of the R. blasticus atp operon, urf6, shows very little sequence homology to any protein of known structure or function. The gene has previously been cloned, the product (called majastridin) has been heterologously expressed in Escherichia coli, and purified to high homogeneity [M. Brosché, I. Kalbina, M. Arnfelt, G. Benito, B.G. Karlsson, A. Strid, Occurrence, overexpression and partial purification of the protein (majastridin) corresponding to the URF6 gene of the Rhodobacter blasticus atp operon, Eur. J. Biochem. 255 (1998) 87-92]. We have solved the X-ray crystal structure and refined a model of majastridin to atomic resolution. Here we present the crystal structures of apo-majastridin and the complex of majastridin with Mn2+ and UDP and show that it has extensive structural similarity to glycosyltransferases (EC 2.4). This is the first structure determined from a new group of distantly related bacterial proteins of at least six members. They share the identical amino acids that bind Mn2+ and a triplet of amino acids in the putative sugar-binding site.

Characterization of a plant-like protochlorophyllide a divinyl reductase in green sulfur bacteria.

The green sulfur bacterium Chlorobium tepidum synthesizes three types of (bacterio)chlorophyll ((B)Chl): BChl a(P), Chl a(PD), and BChl c(F). During the synthesis of all three molecules, a C-8 vinyl substituent is reduced to an ethyl group, and in the case of BChl c(F), the C-8(2) carbon of this ethyl group is subsequently methylated once or twice by the radical S-adenosylmethionine enzyme BchQ. The C. tepidum genome contains homologs of two genes, bchJ (CT2014) and CT1063, that are highly homologous to genes, bchJ and AT5G18660, and that have been reported to encode C-8 vinyl reductases in Rhodobacter capsulatus and Arabidopsis thaliana, respectively. To determine which gene product actually encodes a C-8 vinyl reductase activity, the bchJ and CT1063 genes were insertionally inactivated in C. tepidum. All three Chls synthesized by the CT1063 mutant of C. tepidum have a C-8 vinyl group. Using NADPH but not NADH as reductant, recombinant BciA reduces the C-8 vinyl group of 3,8-divinyl-protochlorophyllide in vitro. These data demonstrate that CT1063, renamed bciA, encodes a C-8 divinyl reductase in C. tepidum. The bchJ mutant produces detectable amounts of Chl a(PD), BChl a(P), and BChl c(F), all of which have reduced C-8 substituents, but the mutant cells secrete large amounts of 3,8-divinyl-protochlorophyllide a into the growth medium and have a greatly reduced BChl c(F) content. The results suggest that BchJ may play an important role in substrate channeling and/or regulation of Chl biosynthesis but show that it is not a vinyl reductase. Because only some Chl-synthesizing organisms possess homologs of bciA, at least two types of C-8 vinyl reductases must occur.

Effects of expression of hemA and hemB genes on production of porphyrin in Propionibacterium freudenreichii.

The genus Propionibacterium has a wide range of probiotic activities that are exploited in dairy and fermentation systems such as cheeses, propionic acid, and tetrapyrrole compounds. In order to improve production of tetrapyrrole compounds, we expressed the hemA gene, which encodes delta-aminolevulinic acid (ALA) synthase from Rhodobacter sphaeroides, and the hemB gene, which encodes porphobilinogen (PBG) synthase from Propionibacterium freudenreichii subsp. shermanii IFO12424, either monocistronically or polycistronically in strain IFO12426. The recombinant strains accumulated larger amounts of ALA and PBG, with resultant 28- to 33-fold-higher production of porphyrinogens, such as uroporphyrinogen and coproporphyrinogen, than those observed in strain IFO12426, which harbored the shuttle vector pPK705.

Detoxification of hydrogen peroxide and expression of catalase genes in Rhodobacter.

The two related facultatively photosynthetic bacteria Rhodobacter sphaeroides and Rhodobacter capsulatus show different sensitivities against peroxide stress. R. sphaeroides is able to tolerate higher concentrations of H2O2 and exhibits higher catalase activity than R. capsulatus. The katE gene of R. sphaeroides and the katG gene of R. capsulatus are strongly induced by H2O2. This induction depends on the presence of the OxyR protein, which is able to bind to the promoter regions of these genes. In addition to katE R. sphaeroides harbours the katC gene, which shows no significant response to H2O2 but is induced in stationary phase.

Bacterial acetone carboxylase is a manganese-dependent metalloenzyme.

Bacterial acetone carboxylase catalyzes the ATP-dependent carboxylation of acetone to acetoacetate with the concomitant production of AMP and two inorganic phosphates. The importance of manganese in Rhodobacter capsulatus acetone carboxylase has been established through a combination of physiological, biochemical, and spectroscopic studies. Depletion of manganese from the R. capsulatus growth medium resulted in inhibition of acetone-dependent but not malate-dependent cell growth. Under normal growth conditions (0.5 microm Mn2+ in medium), growth with acetone as the carbon source resulted in a 4-fold increase in intracellular protein-bound manganese over malate-grown cells and the appearance of a Mn2+ EPR signal centered at g = 2 that was absent in malate-grown cells. Acetone carboxylase purified from cells grown with 50 microm Mn2+ had a 1.6-fold higher specific activity and 1.9-fold higher manganese content than cells grown with 0.5 microm Mn2+, consistently yielding a stoichiometry of 1.9 manganese/alpha2beta2gamma2 multimer, or 0.95 manganese/alphabetagamma protomer. Manganese in acetone carboxylase was tightly bound and not removed upon dialysis against various metal ion chelators. The addition of acetone to malate-grown cells grown in medium depleted of manganese resulted in the high level synthesis of acetone carboxylase (15-20% soluble protein), which, upon purification, exhibited 7% of the activity and 6% of the manganese content of the enzyme purified from acetone-grown cells. EPR analysis of purified acetone carboxylase indicates the presence of a mononuclear Mn2+ center, with possible spin coupling of two mononuclear sites. The addition of Mg.ATP or Mg.AMP resulted in EPR spectral changes, whereas the addition of acetone, CO2, inorganic phosphate, and acetoacetate did not perturb the EPR. These studies demonstrate that manganese is essential for acetone carboxylation and suggest a role for manganese in nucleotide binding and activation.

Thioredoxin can influence gene expression by affecting gyrase activity.

The expression of many genes of facultatively photosynthetic bacteria of the genus Rhodobacter is controlled by the oxygen tension. Among these are the genes of the puf and puc operons, which encode proteins of the photosynthetic apparatus. Previous results revealed that thioredoxins are involved in the regulated expression of these operons, but it remained unsolved as to the mechanisms by which thioredoxins affect puf and puc expression. Here we show that reduced TrxA of Rhodobacter capsulatus and Rhodobacter sphaeroides and oxidized TrxC of R.capsulatus interact with DNA gyrase and alter its DNA supercoiling activity. While TrxA enhances supercoiling, TrxC exerts a negative effect on this activity. Furthermore, inhibition of gyrase activity strongly reduces puf and puc expression. Our results reveal a new signaling pathway by which oxygen can affect the expression of bacterial genes.

Calcium binding and thermostability of carbohydrate binding module CBM4-2 of Xyn10A from Rhodothermus marinus.

Calcium binding to carbohydrate binding module CBM4-2 of xylanase 10A (Xyn10A) from Rhodothermus marinus was explored using calorimetry, NMR, fluorescence, and absorbance spectroscopy. CBM4-2 binds two calcium ions, one with moderate affinity and one with extremely high affinity. The moderate-affinity site has an association constant of (1.3 +/- 0.3) x 10(5) M(-1) and a binding enthalpy DeltaH(a) of -9.3 +/- 0.4 kJ x mol(-1), while the high-affinity site has an association constant of approximately 10(10) M(-1) and a binding enthalpy DeltaH(a) of -40.5 +/- 0.5 kJ x mol(-1). The locations of the binding sites have been identified by NMR and structural homology, and were verified by site-directed mutagenesis. The high-affinity site consists of the side chains of E11 and D160 and backbone carbonyls of E52 and K55, while the moderate-affinity site comprises the side chain of D29 and backbone carbonyls of L21, A22, V25, and W28. The high-affinity site is in a position analogous to the calcium site in CBM4 structures and in a recent CBM22 structure. Binding of calcium increases the unfolding temperature of the protein (T(m)) by approximately 23 degrees C at pH 7.5. No correlation between binding affinity and T(m) change was noted, as each of the two calcium ions contributes almost equally to the increase in unfolding temperature.

The solution structure of the CBM4-2 carbohydrate binding module from a thermostable Rhodothermus marinus xylanase.

The solution structure is presented for the second family 4 carbohydrate binding module (CBM4-2) of xylanase 10A from the thermophilic bacterium Rhodothermus marinus. CBM4-2, which binds xylan tightly, has a beta-sandwich structure formed by 11 strands, and contains a prominent cleft. From NMR titrations, it is shown that the cleft is the binding site for xylan, and that the main amino acids interacting with xylan are Asn31, Tyr69, Glu72, Phe110, Arg115, and His146. Key liganding residues are Tyr69 and Phe110, which form stacking interactions with the sugar. It is suggested that the loops on which the rings are displayed can alter their conformation on substrate binding, which may have functional importance. Comparison both with other family 4 cellulose binding modules and with the structurally similar family 22 xylan binding module shows that the key aromatic residues are in similar positions, and that the bottom of the cleft is much more hydrophobic in the cellulose binding modules than the xylan binding proteins. It is concluded that substrate specificity is determined by a combination of ring orientation and the nature of the residues lining the bottom of the binding cleft.