Exploring the Mechanism of Biocatalyst Inhibition in MicrobialDesulfurization

Andres Abin-Fuentes,a Magdy El-Said Mohamed,b Daniel I. C. Wang,a Kristala L. J. Prathera

Department of Chemical Engineering, Massachusetts Institute of Technology, Cambridge, Massachusetts, USAa; R&DC, Saudi Aramco, Dhahran, Saudi Arabiab

Microbial desulfurization, or biodesulfurization (BDS), of fuels is a promising technology because it can desulfurize compoundsthat are recalcitrant to the current standard technology in the oil industry. One of the obstacles to the commercialization of BDSis the reduction in biocatalyst activity concomitant with the accumulation of the end product, 2-hydroxybiphenyl (HBP), duringthe process. BDS experiments were performed by incubating Rhodococcus erythropolis IGTS8 resting-cell suspensions withhexadecane at 0.50 (vol/vol) containing 10 mM dibenzothiophene. The resin Dowex Optipore SD-2 was added to the BDS experi-ments at resin concentrations of 0, 10, or 50 g resin/liter total volume. The HBP concentration within the cytoplasm was esti-mated to decrease from 1,100 to 260 �M with increasing resin concentration. Despite this finding, productivity did not increasewith the resin concentration. This led us to focus on the susceptibility of the desulfurization enzymes toward HBP. Dose-re-sponse experiments were performed to identify major inhibitory interactions in the most common BDS pathway, the 4S path-way. HBP was responsible for three of the four major inhibitory interactions identified. The concentrations of HBP that led to a50% reduction in the enzymes’ activities (IC50s) for DszA, DszB, and DszC were measured to be 60 � 5 �M, 110 � 10 �M, and50 � 5 �M, respectively. The fact that the IC50s for HBP are all significantly lower than the cytoplasmic HBP concentration sug-gests that the inhibition of the desulfurization enzymes by HBP is responsible for the observed reduction in biocatalyst activityconcomitant with HBP generation.

Biodesulfurization (BDS) is a process in which microorgan-isms, typically, bacteria, are used to reduce the level of sulfur

in fuels derived from crude oil, including diesel and gasoline. Overthe last 20 years, interest in BDS as an alternative to hydrodesul-furization (HDS), which is the current desulfurization standard inthe oil industry, has increased. HDS uses a metal catalyst alongwith hydrogen gas (H2) at high temperature and pressure to re-move sulfur from organic sulfur compounds and generate H2S gas(1). Major drawbacks of HDS include steric hindrance of themetal catalysts by certain recalcitrant compounds and large en-ergy consumption due to process operation at high temperatureand pressure (1). Recalcitrant compounds include the parent mol-ecule dibenzothiophene (DBT) and some of its alkylated deriva-tives, such as 4-methyldibenzothiophene (4-DBT) and 4,6-di-methyldibenzothiophene (4,6-DBT). BDS can potentially be usedto remove the sulfur that cannot be removed by HDS, though it islikely not a replacement for the current HDS infrastructure. Theuse of more than one desulfurization technology may be necessaryto meet the increasingly stringent sulfur regulations (1).

There is a wide range of microorganisms known to have BDScapability (2). Such microorganisms typically desulfurize DBT byone of two pathways: the Kodama pathway or the 4S pathway (1).The Kodama pathway is a destructive BDS pathway in which car-bon-carbon bonds in the DBT molecule are broken and sulfur isnot selectively removed from the organic molecule. Due to itsdestructive nature, the Kodama pathway reduces the caloric valueof the fuel that is being desulfurized. As a result, the majority of thefocus in the past 20 years has been on the 4S pathway, which is anoxidative desulfurization pathway that cleaves the carbon-sulfurbond in DBT and leaves the carbon structure intact. The 4S path-way is a four-step enzymatic pathway that converts DBT to 2-hy-droxybiphenyl (HBP) and sulfate (Fig. 1). The first two steps arethe conversion of DBT to DBT sulfoxide (DBTO) and then toDBT sulfone (DBTO2). These steps are catalyzed by the enzymes

DszC monooxygenase and DszD oxidoreductase in synchrony.The third step is the conversion of DBTO2 to 2-(2=-hydroxy-phenyl) benzene sulfinate (HBPS), which is catalyzed by DszAmonooxygenase and DszD oxidoreductase in synchrony. Thefinal step is the conversion of HBPS to HBP and sulfite by DszBdesulfinase (3).

The isolation and purification of the desulfurization enzymesfrom Rhodococcus erythropolis IGTS8 have been reported (4, 5).Preliminary characterization of these enzymes showed that DszBcatalyzes the slowest reaction with a turnover number (kcat) ofabout 2 min�1 (4). Furthermore, it was reported that DszA has akcat of about 1/s (4). Following the finding that DszB catalyzes theslowest step, the kinetics of DszB were investigated in detail (6).DszB kinetics were modeled appropriately by the Michaelis-Men-ten model with a Michaelis constant (Km) of 0.90 � 0.15 �M anda kcat of 1.3 � 0.07 min�1 (6). The kinetics of DszA, DszC, andDszD from R. erythropolis IGTS8 have yet to be investigated indetail.

Two major obstacles facing BDS commercialization are that (i)biocatalyst activities are significantly lower than what is needed forBDS rates to match HDS rates (2) and (ii) biocatalysts cannotmaintain activity for a long period of time. Considerable effort hasgone into trying to understand and overcome the limitations tothe desulfurization activity of the biocatalysts. Several studies have

Received 12 August 2013 Accepted 2 October 2013

Published ahead of print 4 October 2013

Address correspondence to Kristala L. J. Prather, kljp@mit.edu.

Supplemental material for this article may be found at http://dx.doi.org/10.1128/AEM.02696-13.

Copyright © 2013, American Society for Microbiology. All Rights Reserved.

doi:10.1128/AEM.02696-13

December 2013 Volume 79 Number 24 Applied and Environmental Microbiology p. 7807–7817 aem.asm.org 7807

on July 18, 2020 by guesthttp://aem

.asm.org/

Dow

nloaded from

shown that genetic engineering can lead to higher desulfurizationactivities (2). For example, the desulfurization activity of R. eryth-ropolis KA2-5-1 was increased from 50 to 250 �mol DBT/g dry cellweight (DCW)/h by providing multiple copies of the desulfuriza-tion genes and placing them under the control of alternative pro-moters (2).

A reduction in biocatalyst activity over time has been widelyreported in BDS processes (2, 7, 8). The reduction in activity istypically correlated to the accumulation of HBP in the mediumduring the BDS process. The desulfurization activity of a cell sus-pension of R. erythropolis IGTS8 with a cell density of 66 g DCW/liter mixed with hexadecane (at a 1:1 [vol/vol] ratio) containing19,000 �M DBT was found to follow first-order decay with a decayconstant of 0.072 h�1 (7). It was suggested that the loss of biocat-alyst activity might be due to exposure to increasing HBP concen-trations, although no experiments were done to validate this hy-pothesis (7). The cells were active for 24 h, and only 7,000 out of19,000 �M DBT in oil was consumed. The final concentration ofHBP accumulated in the oil phase after 24 h was only 3,300 �M.Although aqueous-phase accumulation of DBT or HBP was notmeasured, the contribution from the aqueous phase concentra-tion to the overall concentration of either compound could nothave been more than 5%. This was estimated on the basis of thepartition coefficients of DBT and HBP between hexadecane (oil)and water (PO/W), which are 21,000 and about 30 to 50, respec-tively (9). As a result, approximately 3,600 to 3,700 �M DBT orHBP is unaccounted for at the end of the BDS process. Possiblereasons for the discrepancy between DBT disappearance and HBPaccumulation are accumulation of pathway intermediates or re-tention of DBT and/or HBP within the biocatalyst. R. erythropolis,being a Gram-positive bacterium, has a cytoplasmic membranesurrounded by a thick cell wall made mostly of peptidoglycan. Thecell wall is composed of a thick peptidoglycan structure to whichfatty acid molecules are attached. In the case of R. erythropolis (amember of the Actinomycetes family), the cell wall is composed ofmycolic acids that are perpendicular to the cell surface (10). These

mycolic acids range from 30 to 54 carbon atoms in length, andthey make the cell wall of R. erythropolis highly hydrophobic. It ispossible that a significant concentration of DBT initially in the oilphase would have been retained within the cell wall of R. erythro-polis IGTS8 as it entered the cell. It is also possible that a significantfraction of HBP produced by the biocatalyst would have beentrapped within the biocatalyst cytoplasmic membrane or cell wallas it exited the cell. The amount of DBT or HBP retained by thebiocatalyst depends on the partition coefficients between the bio-catalyst and the oil or aqueous phases, PC/O or PC/W, respectively(where C, O, and W represent cells, oil, and water, respectively).

A number of studies have investigated the effect of HBP onbiocatalyst activity. HBP has been shown to inhibit the growth ofR. erythropolis IGTS8 at concentrations greater than 200 �M (11).Preliminary characterization of the DszB enzyme from R. erythro-polis KA2-5-1 showed that its activity is reduced by 50% at HBPconcentrations of about 2,000 �M (12). In other studies, the effectof exogenously added HBP on the BDS process has been de-scribed. R. erythropolis IGTS8 aqueous resting-cell suspensions of2 g DCW/liter were prepared and supplied with only one of the 4Spathway compounds (either DBT, DBTO, DBTO2, or HBPS)(13). HBP was also added at concentrations of either 0 or 50 �M ineach experiment. The disappearance rate of each compound wasmonitored over a short period of 15 min. It was found that thedisappearance rates of DBTO and HBPS were significantly re-duced by the presence of 50 �M HBP. These results suggested thatHBP might be inhibitory to the enzymes responsible for DBTOand HBPS consumption, which are DszC and DszB, respectively(Fig. 1). In a different study, the DBT desulfurization rate of rest-ing cells of Microbacterium sp. strain ZD-M2 was found to de-crease significantly when HBP was added exogenously at concen-trations ranging from 0 to 2,000 �M (14). Results fromexperiments where HBP is added exogenously cannot be used toquantitatively predict the effect of HBP on a typical BDS processwhere DBT is added exogenously and HBP is generated endoge-nously within the cytoplasm of the cells. When HBP is added

FIG 1 The four-step biodesulfurization 4S pathway. The first two steps, catalyzed by both DszC and DszD, are the conversion of DBT to DBT-sulfoxide (DBTO)and then to DBT-sulfone (DBTO2). The third step, catalyzed by both DszA and DszD, is the conversion of DBTO2 to HBPS. The final step is the conversion ofHBPS to HBP by DszB. DBT-MO, DBT monooxygenase.

Abin-Fuentes et al.

7808 aem.asm.org Applied and Environmental Microbiology

on July 18, 2020 by guesthttp://aem

.asm.org/

Dow

nloaded from

exogenously, a significant fraction may be retained by the cell walland may never reach the cytoplasm, where inhibition of the de-sulfurization enzymes could occur. On the other hand, HBP gen-erated endogenously within the cytoplasm is immediately at thelocation where it can be inhibitory to the desulfurization enzymes.Therefore, the specific HBP loading of the biocatalyst (mg HBP/gDCW) that leads to a certain level of reduction in BDS activity maybe significantly larger when HBP is added exogenously.

In summary, there is sufficient evidence to suggest that HBP(either generated endogenously or supplied exogenously) reducesthe overall biocatalyst activity. In this work, a series of commer-cially available adsorbents was screened for the affinity and selec-tivity of the adsorbents to adsorption of HBP. The best adsorbentwas used to determine whether or not HBP-selective adsorbentscould help mitigate the loss of desulfurization activity correlatedwith HBP accumulation during the BDS process. The kinetics andinhibition of the four enzymes in the 4S pathway by the interme-diate compounds were next investigated in detail, in an attempt tounderstand the mechanism for the reduction in biocatalyst activ-ity that is correlated with HBP accumulation.

MATERIALS AND METHODSBacterial strains, vectors, media, and chemicals. The DBT-desulfuriz-ing strain used in this study was Rhodococcus erythropolis IGTS8 ATCC53968, purchased from the American Type Culture Collection. Thedefined minimal medium for cultivation contained the following (perliter of deionized water): glucose, 30.0 g; NH4Cl, 3.0 g; K2HPO4 · 3H2O,6.75 g; NaH2PO4 · H2O, 2.25 g; MgCl2, 0.245 g; FeCl3, 4 mg; CaCl2, 4 mg;Na2SO4, 0.14 g; ZnCl2, 32 mg; MnCl2 · 4H2O, 1 mg; CuCl2 · 2H2O, 5 mg;Co(NO3)2 · 6H2O, 0.7 mg; Na2B4O7 · 10H2O, 0.7 mg; (NH4)6Mo7O24 ·4H2O, 1 mg; EDTA, 12 mg. Cryogenic stocks were prepared by addition of15% (vol/vol) glycerol (final concentration) to mid-log-growth-phasecultures in minimal medium, which were then kept at �80°C for long-term storage. Medium components were obtained from VWR Interna-tional. Christine Nguyen, from Stephen Buchwald’s research group atMIT, kindly provided HBPS. All other chemicals were obtained fromSigma-Aldrich. The strains used for expression of the desulfurizationgenes were Escherichia coli MAX Efficiency DH10B and E. coli BL21Star(DE3) One Shot from Invitrogen. The vector used for all molecularmanipulations was pETDuet-1 from Novagen. The medium used to cul-ture these strains was Luria-Bertani (LB) broth from Difco supplementedwith ampicillin (working concentration, 100 �g/ml) for selection ofclones containing the pETDuet-1 vector.

Testing HBP adsorbents. Activated charcoal, molecular sieves (poresizes, 4, 5, and 13 Å), Diaion HP-20, Dowex Optipore L-493, DowexOptipore SD-2, Biobeads, Amberlite XAD4, Amberlite IRC86, and Am-berlite IRA958 were all obtained from Sigma-Aldrich. Resins were testedfor their ability to adsorb HBP from a hexadecane-water solution. Theadsorption experiments were carried out in 20-ml scintillation vials bymixing 5 ml of water with 5 ml of a hexadecane solution containing 10,000�M DBT and 10,000 �M HBP with either 0.1 or 1.0 g of resin. Mixtureswere equilibrated for 24 h at 30°C, with agitation at 250 rpm. After mea-suring solute concentrations before and after sample equilibration with aparticular resin concentration (Xr), specific loadings (Lr) were determinedby the following relationships:

Xr �mr

Vtotal(1)

Lr ��CHBP�t0

� CHBP�teq�MMHBP

Xr(2)

where CHBP represents the total concentration of HBP (in oil and water),Vtotal is the total volume of solution (oil plus water), mr is the mass of resinused, Xr is the concentration of resin employed, MMHBP is the molecular

mass of HBP (170 g/mol), and t0 and teq represent the times at initial andequilibrium conditions, respectively.

Resting-cell preparation. R. erythropolis IGTS8 cultures were grownin 400 ml of minimal medium in a 2-liter shake flask for a period of 40 to48 h, during which the cell density increased from approximately 0.03 gDCW/liter to 3 g DCW/liter. Cultures were centrifuged at 5,000 rpm and4°C for 15 min, and spent medium was discarded. Biocatalyst pellets wereresuspended to the experimental cell density in 5 g/liter glucose and 20mM phosphate buffer, pH 7.0.

Four-component biodesulfurization experiments. BDS experimentsin the presence of Dowex Optipore SD-2 resin were carried out in 250-mlshake flasks containing 20 ml of a 15.5-g DCW/liter resting-cell suspen-sion mixed with 20 ml of hexadecane containing 10,000 �M DBT. Theresin was added to the mixture at Xrs of 0, 10, or 50 g/liter in the shakeflask, and the flasks were incubated at 30°C for 26 h. After the 26-h incu-bation period, the mixtures were centrifuged for 15 min at 5,000 rpm, andthe four components (oil, water, cells, and resin) were separated. The DBTand HBP concentrations in the oil and aqueous components were mea-sured directly after centrifugation. The DBT and HBP concentrations inthe cells and resin were measured by extraction into ethanol. Ethanol wasselected as the best extractant among a variety of solvents tested, includingacetone, methanol, hexane, hexadecane, squalane, and toluene. Cell wallsolubilization by treatment with an enzymatic mixture of lysozyme (100mg/ml) and mutanolysin (5,000 U/ml) did not increase the concentrationof DBT or HBP extracted from the cells. Sonication of the cells also did notincrease extraction concentrations. Loadings of HBP on the biocatalyst(Lc) and resin (Lr) were calculated from the expressions

Lc ���i

Cextract, i�Vextract

mcells(3)

Lr ���i

Cextract, i�Vextract

mr(4)

where Cextract, i is the concentration of HBP in the ith extract, Vextract is thevolume of the extract (constant for each fraction), mcells is the mass of cellsextracted, and mr is the mass of resin extracted.

DNA manipulations and preparation of E. coli recombinant strains.Molecular biology manipulations were performed according to standardprotocols. Genomic DNA was extracted from R. erythropolis IGTS8 usingPromega’s Wizard genomic DNA purification kit. Each of the desulfuriza-tion genes was PCR amplified from the extracted genomic DNA withappropriate primers (see the supplemental material). Restriction sites forEcoRI and HindIII were included in the primers. PCR amplicons werethen cut with the appropriate restriction enzymes and ligated with thepETDuet-1 vector that had previously been cut with the correspondingrestriction enzymes. The sequences of the resulting vectors containing thefour genes (dszA, dszB, dszC, and dszD) were verified. Each of the fourvectors was then used to transform E. coli DH10B by electroporation using1-mm-path-length cuvettes. E. coli BL21 Star(DE3) was used as the ex-pression host for each of the four desulfurization proteins and trans-formed according to the manufacturer’s protocol (Invitrogen). Selectionwith ampicillin (100 �g/ml) was used to identify successful transfor-mants.

Protein expression and purification protocol. The following proto-col was followed to express and purify each of the four desulfurizationenzymes:

1. A transformed E. coli BL21 Star(DE3) colony from an LB broth-ampicillin plate was picked and grown in 10 ml LB broth withampicillin at 30°C overnight.

2. On the next day, 1 liter of LB medium with ampicillin was inoc-ulated with 10 ml of an overnight culture, and cells were grown at30°C with shaking at 250 rpm.

3. The 1-liter culture was induced with 0.3 mM IPTG (isopropyl-�-

Biocatalyst Inhibition in Microbial Desulfurization

December 2013 Volume 79 Number 24 aem.asm.org 7809

on July 18, 2020 by guesthttp://aem

.asm.org/

Dow

nloaded from

D-thiogalactopyranoside) when the optical density at 600 nm ofthe culture was 0.5 to 0.6. Growth was continued for 18 h at 20°C.

4. Cells were harvested by centrifugation at 5,000 rpm for 15 min at4°C.

5. The supernatant was discarded, and the cell pellet was frozen at�20°C overnight.

6. From this point on, all cells and protein samples were maintainedat 4°C or on ice to avoid protein degradation. The cell pellet wasresuspended using a volume of the 1� His buffer (with 10% glyc-erol) equal to at least 2.5 times the mass of the cell pellet. Lysozyme(1 mg/ml) and DNase I (500 �g/ml) were added, and the cellsuspension was incubated for 30 min on ice.

7. The cell suspension was sonicated for 15 to 30 min at 90 to 100%amplitude with a 1-s on/1-s off cycle.

8. Cell debris was pelleted by centrifugation at 20,000 rpm for 30min at 4°C.

9. The supernatant (20 to 40 ml) was mixed with 5 ml Ni-nitrilotri-acetic acid agarose (Qiagen) in order to bind the histidine-taggeddesulfurization protein to the resin. The time for this binding stepwas 1 h at 4°C.

10. After the 1-h binding step, the protein-resin mixture was pouredonto a Pierce centrifuge column (total volume, 10 ml), and theresin was allowed to settle. The flowthrough was processed bygravity flow and collected.

11. The resin was washed with 10 ml of 1� His buffer containing 7.5mM imidazole, and the flowthrough was collected. Then, the resinwas treated with five different solutions of 1� His buffer contain-ing increasing imidazole concentrations of 40, 60, 100, 250, and500 mM imidazole. Each solution was 5 ml in volume. The eluentof each fraction was collected.

12. A diagnostic SDS-polyacrylamide gel was run to determine thefractions with the highest purity. These fractions were combinedin one piece of snakeskin dialysis tubing, and the protein samplewas dialyzed overnight in 1 liter of 50 mM Tris-HCl–50 mMNaCl–10% glycerol buffer in a cold room with mild stirring.

13. Finally, the dialyzed protein was flash-frozen using liquid nitro-gen in small aliquots for long-term storage at �80°C.

Enzyme assays. All enzyme assays were performed in 20 mM phos-phate buffer at pH 7.0 and 30°C. The assay volume was 1,000 �l for allassays. DszD assays were performed in semimicrocuvettes and main-tained at room temperature. DszA, DszB, and DszC assays were per-formed in 1.7-ml Eppendorf tubes in a rotary shaker incubator at 30°Cand 250 rpm and were quenched either by addition of 1% of a 12 M HClsolution or by use of an Amicon Ultra 0.5-ml 10-kDa spin filter device toremove the enzyme. For DszD enzyme assays, either the concentration ofNADH was varied from 0 to 400 �M (at a fixed flavin mononucleotide[FMN] concentration of 100 �M) or the concentration of FMN was var-ied from 0 to 100 �M (at fixed a NADH concentration of 150 �M), theconcentration of DszD enzyme was 1.2 �g/ml, and the assays were run for10 min. For DszC enzyme assays, the concentration of NADH was 500�M, the FMN concentration was 10 �M, the DBT concentration wasvaried from 0 to 5 �M, the DszD concentration was 0.47 �g/ml, the DszCconcentration was 3.1 �g/ml, and the assays were run for 5 min. Assayconditions were identical for determining the mechanism of DszC inhi-bition by HBPS and HBP, except for the addition of those compounds atconcentrations of 0 to 200 and 0 to 1,000 �M, respectively. For DszAenzyme assays, the NADH concentration was 500 �M, the FMN concen-tration was 10 �M, the DBTO2 concentration was varied from 0 to 80 �M,the DszD concentration was 1.4 �g/ml, the DszA concentration was 6.7�g/ml, and the assays were run for 3 min. For DszB enzyme assays, theHBPS concentration was varied from 0 to 200 �M, the DszB concentra-tion was 3.5 �g/ml, and the assays were run for 5 min. For all four en-

zymes, dose-response-type assays were performed to measure the inhibi-tory concentration leading to a 50% reduction in enzyme activity (IC50).For DszD dose-response assays, the concentration of NADH was 500 �Mand that of FMN was 200 �M, the concentration of the various pathwayintermediates was varied (DBT, 0 to 5 �M; DBTO2, 0 to 80 �M; HBPS, 0to 2,000 �M; HBP, 0 to 2,000 �M), the concentration of DszD was 4.7�g/ml, and the assays were run for up to 5 min. For DszC dose-re-sponse assays, the NADH concentration was 500 �M, the FMN con-centration was 10 �M, the DBT concentration was 5 �M, the DszDconcentration was 0.47 �g/ml, the DszC concentration was 0.3 mg/ml,the concentration of the pathway intermediates was varied (DBTO2, 0to 80 �M; HBPS, 0 to 1,000 �M; HBP, 0 to 2,000 �M), and the assayswere run for 10 min. For DszA dose-response assays, the NADH con-centration was 500 �M, the FMN concentration was 10 �M, theDBTO2 concentration was 80 �M, the DszD concentration was 1.4�g/ml, the DszA concentration was 67 �g/ml, the concentration of thepathway intermediates was varied (DBT, 0 to 5 �M; HBPS, 0 to 500�M; HBP, 0 to 1,000 �M), and the run time was 3 min. For DszBdose-response assays, the HBPS concentration was 200 �M, the DszBconcentration was 58.5 �g/ml, the concentration of the pathway in-termediates was varied (DBT, 0 to 5 �M; DBTO2, 0 to 80 �M; HBP, 0to 1,000 �M), and the assays were run for 5 min.

Analytical methods. For DszD assays, the absorbance at 340 nm wasmeasured using a spectrophotometer to calculate the NADH concentra-tion, and this was used to calculate the activity. The concentrations ofDBT, DBTO2, HBPS, and HBP were measured using reversed-phase high-pressure liquid chromatography (HPLC) with a Zorbax SB-C18 column ata flow rate of 1 ml/min. For DszC and DszB assays, the mobile phase was50% acetonitrile–50% water and the temperature of the column wasmaintained at 65°C. For DszA assays, the mobile phase was 30% acetoni-trile–70% 10 mM tetrabutylammonium bisulfate, the mobile phase was15 mM acetic acid (pH 5.0), and the temperature of the column wasmaintained at 45°C.

RESULTSScreening HBP-selective adsorbent resins. A variety of commer-cially available resins were screened for their ability to take up HBPand DBT from a 0.50 (vol/vol) hexadecane-water solution with aninitial concentration of 10,000 �M DBT and 10,000 �M HBP inhexadecane. The resins screened in this study were chosen becausethey span a wide range of resin hydrophobicity and ionic func-tionalities, and most have been previously investigated in thePrather lab (15). We found that poly(styrene-co-divinylbenzene)-derived resins generally had the greatest affinity for HBP (Fig. 2A;Xr � 10 g resin/liter total volume for each resin). For instance,Dowex Optipore L-493 and SD-2 were able to achieve specificloadings of HBP of 40 and 50 mg/g resin, respectively. In additionto having the highest affinity for HBP of all the resins tested,Dowex Optipore L-493 and SD-2 also possessed the highest selec-tivity for HBP relative to DBT, with HBP loadings that were 2.1and 2.5 times greater than their respective DBT loadings. Theseresins were also found to have the highest affinity for butanol, withspecific loadings of butanol of 175 and 150 mg/g resin, respectively(15). The high affinity of these resins for butanol has been attrib-uted to strong hydrophobic interactions between the resins’ �-�bonds of the aromatic side groups and the alkyl chain of butanoland to the resins’ high specific surface area (15). The affinity ofthese resins for HBP is likely due to strong hydrophobic interac-tions between the resins’ aromatic side groups and the biphenylpart of HBP, though the specific loadings are reduced relative tothat of n-butanol.

Four-component BDS experiments. We previously reportedthat the DBT desulfurization activity of R. erythropolis IGTS8 is

Abin-Fuentes et al.

7810 aem.asm.org Applied and Environmental Microbiology

on July 18, 2020 by guesthttp://aem

.asm.org/

Dow

nloaded from

significantly reduced as HBP accumulates in the medium duringthe BDS process (A. Abin-Fuentes, J. C. Leung, M. E. S. Mohamed,D. I. C. Wang, and K. L. J. Prather, submitted for publication). Inparticular, we found that the activity of a resting-cell suspension of15.5 g DCW/liter was decreased by 90% when the HBP concen-tration in the aqueous medium reached 40 �M (Abin-Fuentes etal., submitted). In previous work, the use of Dowex OptiporeSD-2 in butanol fermentations led to a decrease in the aqueous-phase concentration and an increase in productivity (15). In thecurrent work, this resin, which had the best affinity and selectivityfor HBP of all resins studied, was used to try to mitigate the reduc-tion in biocatalyst activity correlated with HBP accumulation dur-ing the BDS process, thereby increasing the conversion of DBT. R.erythropolis IGTS8 resting-cell suspensions of 15.5 g DCW/literwere mixed with hexadecane (0.50, vol/vol) containing 10,000�M DBT, and Dowex Optipore SD-2 was added at an Xr of 0, 10,

or 50 g/liter. The four-component (oil, water, cells, and resin)mixtures were incubated at 30°C in a rotary shaker at 250 rpm for26 h. The concentration of DBT and HBP in each component wasmeasured at the end of the incubation period. The partition coef-ficient of HBP between one component (component 1) and an-other component (component 2) in the four-component BDSexperiments (Pcomp1/comp2) is expressed as

Pcomp1⁄comp2 �CHBP, comp1

CHBP, comp2(5)

where CHBP, comp1 is the HBP concentration measured in compo-nent 1 and CHBP, comp2 is the HBP concentration measured incomponent 2. For example, PR/C is the partition coefficient of HBPbetween the resin and the biocatalyst (cells). Similarly, PO/W is thepartition coefficient of HBP between hexadecane (oil) and water.The partition coefficients calculated at the different resin concen-

FIG 2 (A) Specific loading of HBP (white) and DBT (black) by the various resins tested at a resin concentration of 10 g/liter from a 0.50 (vol/vol) hexadecane-water solution initially containing 10 mM DBT and 10 mM HBP. (B) Total amount of HBP produced (white) and cytoplasmic HBP concentration (black) in thefour-component BDS experiment at 15.5 g DCW/liter and with an oil fraction of 0.50 (vol/vol) and 10 mM DBT. Specific HBP loadings are shown above the blackcolumns. Data shown are the averages and standard deviations of 3 replicates.

Biocatalyst Inhibition in Microbial Desulfurization

December 2013 Volume 79 Number 24 aem.asm.org 7811

on July 18, 2020 by guesthttp://aem

.asm.org/

Dow

nloaded from

trations are shown in Table 1. The resin Dowex Optipore SD-2had the highest affinity for HBP relative to the other componentsof the system. This resin had log PR/C and log PR/O values of about2, which means that the resin’s affinity for HBP was about 100times greater than the affinity of either the hexadecane oil phase orthe biocatalyst. Furthermore, this resin had a log PR/W of about 4,which means that its affinity for HBP was about 10,000 timesgreater than that of the aqueous buffer. The partition coefficientsof HBP between oil and the aqueous buffer (PO/W) and betweenthe biocatalyst and the aqueous buffer (PC/W) were very similar(Table 1). The log PO/W and log PC/W in the absence of resin were1.7 and 1.8, respectively. Also, the partition coefficient of HBPbetween the biocatalyst and oil (PC/O) was calculated to be 1 in theabsence of resin. The finding that PO/W was approximately equalto PC/W and that PC/O was approximately equal to 1 indicates thatthe oil and biocatalyst components have very similar affinities forHBP. Both the oil and the biocatalyst have an affinity for HBP thatis 50 to 60 times greater than that of the aqueous buffer in theabsence of resin (Table 1).

The partition coefficient PC/W is defined as

PC ⁄ W �CHBP, intracellular

CHBP, water(6)

where CHBP, water is the HBP concentration in the water phase andCHBP, intracellular is the HBP concentration within the biocatalyst.The cell can be viewed as being composed of two components: theinner cytoplasmic space and the outer envelope/shell that encom-passes the cytoplasm, which is the cell wall. The partition coeffi-cient of HBP between the cytoplasm and water (Pcytoplasm/W) wasestimated from solving for its value in the expression

PC ⁄ W � Pcell wall ⁄ Wfcell wall � Pcytoplasm⁄Wfcytoplasm (7)

where Pcell wall/W is the partition coefficient of HBP between thecell wall and the water. The value of Pcell wall/W was estimated to be550 (see the supplemental material). fcell wall and fcytoplasm are thefractions of the total volume of a single cell that are occupied bythe cell wall and cytoplasm, respectively. The values of fcytoplasm

and fcell wall were calculated from the expressions

fcytoplasm ��Rcell � W�3

�Rcell�3 (8)

fcell wall � 1 ��Rcell � W�3

�Rcell�3 (9)

where W is the thickness of the cell wall, which has been estimatedto be approximately 10 nm for Rhodococcous species (16). Rcell isthe radius of the cell, which was estimated to be 0.5 �m (17). Thecytoplasmic HBP concentration (CHBP, cytoplasm) was calculatedfrom the following expression:

CHBP, cytoplasm � Pcytoplasm⁄W CHBP, water (10)

The value of CHBP, cytoplasm was calculated to be 1,100, 330, and260 �M at resin concentrations (Xr) of 0, 10, and 50 g/liter, re-spectively (Fig. 2B). The corresponding HBP loadings on the bio-catalyst (Lc) were calculated to be 1.6, 0.5, and 0.2 mg HBP/g DCWusing equation 3 (Fig. 2B). These values show that the resin waseffective in reducing HBP retention within the cytoplasm of thebiocatalyst, which is where the desulfurization enzymes are pres-ent (4). Despite the significant decrease in HBP retained withinthe cytoplasm, the total amount of HBP produced in the systemdid not increase with increasing resin concentration (Fig. 2B).Therefore, it was postulated that the biocatalyst and, in particular,the desulfurization enzymes might be susceptible to cytoplasmicHBP concentrations of less than 260 �M.

Enzyme kinetics. To study the kinetics of each enzyme, recom-binant protein was expressed in and purified from E. coli (see thesupplemental material). For DszA, DszB, and DszD, the kineticdata obtained were appropriately modeled by the Michaelis-Men-ten model (Fig. 3A, B, and D). The equation for this model is

� �kcat�E0��S�Km � �S� �

Vmax�S�Km � �S� (11)

where is the enzyme activity, kcat is the turnover number, [E0] isthe enzyme concentration in each assay, [S] is the substrate con-centration, Km is the Michaelis constant, and Vmax is the maxi-mum enzyme activity. The kinetics of DszC could not be modeledaccurately with a simple Michaelis-Menten model because it wasfound that DszC was inhibited by its substrate, DBT (Fig. 3C). Asa result, a substrate inhibition model was used to fit the data ob-tained with DszC (18):

� �Vmax�S�

Km � �S� ��S�2

KSI

(12)

where KSI is the substrate (DBT) inhibition constant.The fitting of the models to the data was done using the enzkin

package in the MATLAB program. The kinetic data obtained forall four desulfurization enzymes are shown in Fig. 3. The kineticparameters of each enzyme are summarized in Table 2 and com-pared to parameters from other studies, where available. To ourknowledge, the only prior detailed characterization of a desulfu-rization enzyme from R. erythropolis IGTS8 found kcat to be equalto 1.3 � 0.07 min�1 and Km to be equal to 0.90 � 0.15 �M forDszB (6). In the current work, kinetic characterization of DszByielded a kcat value of 1.7 � 0.2 min�1 and a Km value of 1.3 � 0.3�M, which are in good agreement with values from the previousstudy (6). Rough preliminary characterization of DszA and DszDfrom R. erythropolis IGTS8 yielded kcat values of about 60 min�1

and approximately 300 min�1, respectively (4). The correspond-ing values of kcat for DszA and DszD measured in this work were

TABLE 1 Comparison of partition coefficients between the four different components in the four-component BDS experiments in the presence ofDowex Optipore SD-2 resina

Resin concn(g/liter) PC/O PC/W PO/W PR/C PR/O PR/W Log PC/O Log PC/W Log PO/W Log PR/C Log PR/O Log PR/W

0 1 60 52 0.1 1.8 1.710 1 47 42 88 99 4,116 0 1.7 1.6 1.9 2 3.650 3 160 54 99 296 15,906 0.5 2.2 1.7 2 2.5 4.2a C, cells; O, oil; W, water; R, resin.

Abin-Fuentes et al.

7812 aem.asm.org Applied and Environmental Microbiology

on July 18, 2020 by guesthttp://aem

.asm.org/

Dow

nloaded from

11 � 2 min�1 and 760 � 10 min�1, which are of the same order ofmagnitude as those rough estimates from preliminary character-ization studies (4). DszC was inhibited by its substrate (DBT), andthe kinetic parameters obtained were as follows: kcat was equal to1.6 � 0.3 min�1, Km was equal to 1.4 � 0.3 �M, and KSI was equalto 1.8 � 0.2 �M. The activity of DszC from R. erythropolis D-1 wasreported to be 30.3 nmol DBTO2/mg DszC/min (19), which is inclose agreement with the maximum activity measured in this workof 31.3 nmol DBTO2/mg DszC/min (Fig. 3C).

The catalytic efficiency of an enzyme is best defined by the ratioof the kinetic constants, kcat/Km (20). The catalytic efficiency ofeach desulfurization enzyme was calculated (Table 2). The cata-lytic efficiencies of DszA, DszB, and DszC were calculated to be

3.1, 1.3, and 1.1 �M�1 min�1, respectively. The catalytic effi-ciency of DszD on NADH and FMN was calculated to be 6.7and 100 �M�1 min�1, respectively. Therefore, the enzymes canbe listed in order of decreasing efficiency as DszD DszA DszB � DszC.

Identification of four major inhibitory interactions in the 4Spathway. All of the possible interactions among the 4 differentdesulfurization enzymes and the four compounds in the pathway(DBT, DBTO2, HBPS, and HBP) were studied to determine themajor inhibitory interactions. DBTO was not included because itis not typically observed during the BDS process because its rate ofconsumption is much faster than its rate of generation (4). Thestrength of inhibition was studied by means of a dose-response

FIG 3 Enzyme kinetics for DszA (A), DszB (B), DszC (C), and DszD (D). Diamonds, the data; solid lines, Michaelis-Menten model fits for DszA, DszB, and DszDand substrate inhibition model for DszC. Data shown are the averages and standard deviations of 3 replicates.

TABLE 2 Main properties of the Dsz enzymes purified in this work, including molecular weights, stock concentrations, and measured kineticconstants

EnzymeMol wt(103)

Stockconcn(mg/ml)

This work Other authors

kcat (min�1) Km (�M) kcat/Km (�M�1 min�1) kcat (min�1) Km (�M)

DszA 49.6 1.3 11 � 2 3.6 � 0.5 3.1 �60a �1a

DszB 39.0 1.2 1.7 � 0.2 1.3 � 0.3 1.3 1.3 � 0.07b 0.90 � 0.15b

DszC 45.0 6.1 1.6 � 0.3 1.4 � 0.3e 1.1 NAf 5a

DszD 25.0 4.7 760 � 10 114 � 5 (NADH),7.3 � 0.5 (FMN)

6.7 (NADH), 100 (FMN) �300c 208 (NADH), 10.8 (FMN)d

a The strain used was R. erythropolis IGTS8 (4).b The strain used was R. erythropolis IGTS8 (6).c Calculated estimate based on data in reference 4. The strain used was R. erythropolis IGTS8.d The strain used was R. erythropolis D-1 (25).e KSI was equal to 1.8 � 0.2 �M.f NA, not available.

Biocatalyst Inhibition in Microbial Desulfurization

December 2013 Volume 79 Number 24 aem.asm.org 7813

on July 18, 2020 by guesthttp://aem

.asm.org/

Dow

nloaded from

plot. The dose-response equation describing the effect of the in-hibitor concentration on enzyme activity is expressed as (20)

�i

�0�

1

1 ��I�

IC50

(13)

where i is the enzyme’s activity at an inhibitor concentration of[I], 0 is the enzyme’s activity in the absence of inhibitor, and IC50

is the concentration of inhibitor required to reduce the enzyme’sactivity by 50%. The IC50 parameter is phenomenological and hasno mechanistic implications. The value of IC50 was obtained byfitting the data for i/0 versus [I] to the dose-response equationusing the nonlinear fitting package nlinfit in the MATLAB pro-gram. The four major inhibitory interactions identified were (inorder of decreasing strength) DszC inhibition by HBPS (IC50 �15 � 2 �M), DszC inhibition by HBP (IC50 � 50 � 5 �M), DszAinhibition by HBP (IC50 � 60 � 5 �M), and DszB inhibition by

HBP (IC50 � 110 � 10 �M) (Fig. 4 and Table 3). Note that theIC50s of all four inhibitory interactions are significantly less thanthe estimated cytoplasmic HBP concentration. Even in the best-case scenario, when 50 g/liter of a highly HBP-selective resin wasadded to the BDS mixture, the estimated cytoplasmic HBP con-centration was still 260 �M. This finding suggests that these fourinhibitory interactions are likely responsible for the reduction inbiocatalyst activity that is observed during a typical BDS processwhen HBP is generated endogenously from DBT within the bio-catalyst.

The only prior preliminary characterization of inhibitionwithin the 4S pathway showed that DszB from R. erythropolisKA2-5-1 had an IC50 of about 2,000 �M with respect to HBP (12).We report here three new inhibitory interactions in the pathwaythat are stronger than any previous preliminary findings. Notethat the strongest inhibitory interactions are on the first enzyme inthe pathway, DszC, and the strength of inhibition decreases far-ther down the pathway, with the last enzyme in the pathway,DszB, having the weakest HBP inhibition at an IC50 of 110 �M(Fig. 4B). Note also that the major inhibitory compounds in thepathway are the last two intermediates, HBPS and HBP. Thispattern of inhibition is typical of feedback inhibition of linearpathways; for example, in the tricarboxylic acid (TCA) cycle,the first enzyme in the pathway is strongly inhibited by the endproduct, ATP.

Mechanism of major inhibitory interactions in the 4S path-way. The telltale sign of noncompetitive inhibition is a reducedVmax without a change in Km as the inhibitor concentration is

FIG 4 Normalized desulfurization enzyme activity at different inhibitor concentrations for the four major inhibitory interactions identified in the 4S pathway.(A) DszA inhibition by HBP has an IC50 of 60 � 5 �M; (B) DszB inhibition by HBP has an IC50 of 110 � 10 �M; (C) DszC inhibition by HBPS has an IC50 of15 � 2 �M; (D) DszC inhibition by HBP has an IC50 of 50 � 5 �M. Data shown are the averages and standard deviations of 3 replicates.

TABLE 3 Summary of inhibition parameters for the four majorinhibitory interactions, including the mechanism of inhibition, Ki, �,and IC50s

Enzyme Inhibitor Mechanism Ki (�M) � IC50 (�M)

DszA HBP Not determined NAa NA 60 � 5DszB HBP Not determined NA NA 110 � 10DszC HBPS Noncompetitive 13.5 0.13 15 � 2DszC HBP Noncompetitive 40 0.4 50 � 5a NA, not applicable.

Abin-Fuentes et al.

7814 aem.asm.org Applied and Environmental Microbiology

on July 18, 2020 by guesthttp://aem

.asm.org/

Dow

nloaded from

increased (20). This phenomenon was observed in the inhibitionof DszC by both HBPS and HBP (Fig. 5A and B). The model fornoncompetitive inhibition of an enzyme that obeys Michaelis-Menten kinetics is given by (20)

� �Vmax�S�

Km�1 ��I�Ki� � �S��1 �

�I��Ki

� (14)

where [I] is the inhibitor concentration, Ki is the inhibition con-stant, and � is a parameter that reflects the effect of the inhibitoron the affinity of the enzyme for its substrate and, likewise, theeffect of the substrate on the affinity of the enzyme for the inhib-itor. Noncompetitive inhibition refers to the case in which aninhibitor displays binding affinity for both the free enzyme and theenzyme-substrate binary complex (see the supplemental mate-rial). This form of inhibition is the most general case; in fact,competitive and uncompetitive inhibition can be viewed as spe-cial, restricted cases of noncompetitive inhibition in which thevalue of � is infinity and zero, respectively (20). Since the kineticsof DszC showed substrate inhibition, a modification to the non-competitive model was derived (see the supplemental material).The noncompetitive inhibition of DszC by HBPS and HBP can beexpressed as

� �Vmax�S�

Km�1 ��I�Ki� � �S��1 �

�I��Ki

� ��S�2

KSI

(15)

The solid lines in Fig. 5A and B represent the fits for the non-competitive inhibition model (equation 15), where the values of

Ki and � were 13.5 �M and 0.13, respectively, for HBPS inhibitionand 40 �M and 0.4, respectively, for HBP. Fitting of the model tothe data was attempted using nlinfit from MATLAB but was un-successful due to convergence issues. Instead, the parameters weredetermined by a graphical method outlined in section 8.3 of ref-erence 20. The first step in this method was the construction of thedouble-reciprocal Lineweaver-Burk plot. To obtain the values ofKi and �Ki, two secondary plots were constructed. The first ofthese was a Dixon plot of 1/Vmax as a function of [I], from whichthe value of ��Ki can be determined as the x intercept. In thesecond plot, the slopes of the double-reciprocal lines (from theLineweaver-Burk plot) are plotted as a function of [I]. For thisplot, the x intercept is equal to �Ki. Combining the informationfrom these two secondary plots allows determination of both in-hibition parameters. The mechanism of DszB inhibition by HBPand DszA inhibition by HBPS could not be determined due to thelimited resolution of the HPLC detector.

The enzyme inhibition model predicts the reduction in bio-catalyst activity. A model that incorporated enzyme inhibitionwas developed to predict the volumetric desulfurization rate inbiphasic (oil-water) BDS experiments (see the supplemental ma-terial for the model equations). The model accounted for DBT andHBP in three components: oil, water, and cytoplasm. The twopathway intermediates, DBTO2 and HBPS, were assumed to bepresent only in the cytoplasm. The three major inhibitory inter-actions by HBP on DszA, DszB, and DszC were taken into ac-count. The fourth major inhibitory interaction, the inhibition ofDszC by HBPS, was ignored for simplicity. DszD activity was as-sumed to not be limiting and was therefore excluded from themodel. A loss of biocatalyst activity unrelated to HBP accumu-lation was also taken into account in the model through anexponential decay constant, kd (see the supplemental material).The concentrations of DszA, DszB, and DszC were assumed tobe all the same and equal to E0. E0 was the only floating param-eter in the model. The value of E0 was obtained by allowing itsvalue to vary until the initial desulfurization rate predicted bythe model matched the measured desulfurization rate. Thevalue of E0 obtained was 15 mg/ml, so that the total concentra-tion of desulfurization enzymes in the cytoplasm was 45 mg/ml. This value is of the same order as the total cytoplasmicconcentration of protein, which has been reported to be ap-proximately 200 mg/ml (21).

The first step in the biphasic BDS experiments was growth ofbiocatalyst in a 4-liter bioreactor to a cell density of 29 to 35 gDCW/liter. Next, 10% (vol/vol) hexadecane containing 100,000�M DBT was added to the bioreactor. The desulfurization rate ofDBT in the bioreactor was monitored thereafter from the mea-sured concentrations of DBT and HBP in the hexadecane andaqueous phases. For the first 2 h after hexadecane addition, themixing speed was maintained at 200 rpm. The mixing speed wasincreased to 500 rpm thereafter to minimize mass transport limi-tations in the system. The enzyme inhibition model was used topredict the desulfurization rate only after the shift in mixing speedto 500 rpm. The model predicts the concentration of HBP in theoil and water phases and the volumetric desulfurization rate accu-rately (Fig. 6A to C). The coefficient of determination (R2) be-tween the model and the data was calculated to be 0.96 using theMATLAB function corr(X,Y).

FIG 5 Noncompetitive inhibition of DszC by HBPS and HBP. (A) DszCactivity over a range of DBT substrate concentrations from 0 to 5 �M andHBPS concentrations of 0 �M (closed diamonds), 5 �M (closed squares), 25�M (closed triangles), and 50 �M (crosses). (B) DszC activity over a range ofDBT substrate concentrations from 0 to 5 �M and HBP concentrations of 0�M (closed diamonds), 100 �M (closed squares), and 500 �M (closed trian-gles). Solid lines are model fits from equation 15.

Biocatalyst Inhibition in Microbial Desulfurization

December 2013 Volume 79 Number 24 aem.asm.org 7815

on July 18, 2020 by guesthttp://aem

.asm.org/

Dow

nloaded from

DISCUSSION

Dose-response experiments were performed to investigate the po-tential inhibitory effect of the various intermediates in the 4Spathway on the desulfurization enzymes. Four major inhibitoryinteractions were identified, all of which had IC50s under 110 �M.Three of the four major inhibitory interactions were carried out byHBP. HBP was found to inhibit DszA, DszB, and DszC with IC50sof 60, 110, and 50 �M, respectively. The fourth major inhibitoryinteraction identified is the inhibition of DszC monooxygenase byHBPS. This interaction is the strongest of the four since it has thelowest IC50 of 15 �M. The HBPS concentration during the four-component BDS experiments was not measured in this work. TheIC50s of HBP on the desulfurization enzymes were all significantlylower than the minimum cytoplasmic HBP concentration esti-mated during the four-component BDS experiments, which was

260 �M. This suggests that enzyme inhibition by HBP was respon-sible for the reduction in biocatalyst activity during the four-com-ponent BDS experiments, even in the presence of 50 g/liter ofHBP-selective resin (Dowex Optipore SD-2). In the n-butanolfermentation by Clostridium acetobutylicum ATCC 824, n-butanolproductivity was increased 2-fold upon addition of Dowex Opti-pore SD-2 to the medium (15). However, the n-butanol toxicitythreshold of Clostridium is generally considered to be about 1.3%(wt/vol), which is approximately 180 mM (15). The fact that theinhibitory levels of HBP on the R. erythropolis IGTS8 desulfuriza-tion enzymes are at least 3 orders of magnitude smaller than thetoxicity levels of n-butanol on Clostridium might explain why then-butanol productivity by Clostridium was enhanced upon addi-tion of Dowex Optipore SD-2 and the HBP productivity by R.erythropolis IGTS8 was not enhanced.

In another study, R. erythropolis IGTS8 lysates were suppliedwith 200 �M DBT initially and the concentrations of DBT,DBTO2, HBPS, and HBP were monitored over time (4). After 10min, all DBT was depleted, the HBPS concentration was approx-imately 130 �M, and the HBP concentration was approximately50 �M (4). From 10 to 60 min, the HBPS concentration decreasedsteadily from 130 to 0 �M and the HBP concentration increasedfrom 50 to 200 �M. No DBTO2 was detected at any time (4). All ofthese results agree with the kinetic data obtained in this work. Firstof all, since the kcat of DszA is approximately 7 times that of DszC(Table 2), the DBTO2 consumption rate is expected to be signifi-cantly greater than its generation rate, which agrees with the factthat no DBTO2 was detected. Second, the buildup of HBPS withinthe first 10 min is consistent with the fact that its consumption rate(DszB kcat � 1.7 min�1) is significantly slower than its generationrate (DszA kcat � 11.2 min�1) (Table 2). The fact that the HBPconcentration accumulates to over 130 �M within the first fewminutes indicates that DszC would have been severely inhibitedby HBPS at that point in time. As HBPS is consumed, DszC inhi-bition by HBPS is relieved, but then HBP inhibition of DszC (andDszA and DszB) becomes more significant. Therefore, we expectthat during a BDS experiment, all four major inhibitory interac-tions reported here would be important. HBPS inhibition of DszCis responsible for maintaining the BDS rate low at the beginning ofthe BDS process, when HBP levels are still low. Once HBP levelsrise, HBP inhibition of DszA, DszB, and DszC is mostly responsi-ble for inhibition of the BDS rate.

Due to the biocatalyst’s high affinity for HBP relative to that ofthe aqueous buffer (PC/W � 60 at Xr � 0 g/liter), the intracellularHBP concentration (CHBP, intracellular) is much higher than theHBP concentration in the aqueous buffer (CHBP, water). In previousstudies, HBP inhibition in biphasic (oil-water) BDS experimentshas been downplayed due to the partition coefficient of HBP be-tween oil and water (PO/W) being high (about 40 to 50) (22). It hasgenerally been assumed that since the biocatalyst resides in theaqueous phase, the intracellular concentration of HBP is similar tothe aqueous-phase concentration. This assumption is made inpart because the biocatalyst’s affinity for HBP has not been quan-tified in previous studies. We have shown that this affinity for HBPis in par with that of the oil phase. As a result, HBP retentionwithin the biocatalyst and HBP inhibition of the desulfurizationenzymes are obstacles even in biphasic BDS experiments.

The affinity of the biocatalyst for HBP is an intrinsic propertyof the biocatalyst and likely varies depending on the biocatalystemployed. Interactions between cyclic hydrocarbons and biolog-

0

50

100

150

200

250

300

0 10 20 30 40 50

VB

DS (

µmol

e H

BP/

L/h

)

time (hours)

0 50

100 150 200 250 300 350

0 10 20 30 40 50

HB

P in

wat

er (u

M)

0

2000

4000

6000

8000

10000

0 10 20 30 40 50

HB

P in

oil

(uM

)

time (hours)

B

C

FIG 6 Enzyme inhibition model predictions for BDS bioreactor experiment.(A) Volumetric desulfurization rate data (circles) and model prediction (line).VBDS, volumetric desulfurization rate of BDS. (B) Concentration of HBP in theoil phase data (circles) and model (line). (C) Concentration of HBP in thewater phase data (circles) and model (line). Data shown are the averages andstandard deviations of 3 replicates.

Abin-Fuentes et al.

7816 aem.asm.org Applied and Environmental Microbiology

on July 18, 2020 by guesthttp://aem

.asm.org/

Dow

nloaded from

ical membranes have been previously investigated (23). The par-tition coefficients of a range of cyclic hydrocarbons (includingaromatics) between liposomes prepared from E. coli phospholip-ids and an aqueous phosphate buffer were measured. From thesemeasured partition coefficients, a correlation for predicting thepartition coefficient of any cyclic hydrocarbon between the lipo-some (membrane) and the aqueous buffer (PM/B) based on theoctanol-water partition coefficient (PO/W) of that cyclic hydrocar-bon was developed. The liposome-buffer partition coefficient ofHBP [log(PM/B)] is predicted to be 2.2, given the known value ofthe octanol-water partition coefficient [log(PO/W)] of 3.09 (24).This is in good agreement with the value of the partition coeffi-cient of HBP between the biocatalyst and the buffer measured inthis work [log(PC/W) � 1.8 to 2.2] (Table 1). The liposomes pre-pared by Sikkema et al. (23) are representative of the cytoplasmicmembrane of a Gram-negative bacterium. R. erythropolis, being aGram-positive bacterium, has a thin cytoplasmic membrane sur-rounded by a thick cell wall made mostly of a peptidoglycan struc-ture and mycolic acids (10). The fact that the value of log(PC/W)measured in this study agrees well with the predicted value oflog(PM/B) suggests that the HBP retained by R. erythropolis IGTS8might reside mostly within the cytoplasmic membrane as opposedto the cell wall, which is structurally very different. This idea agreeswith the mechanism of DBT-to-HBP conversion during the BDSprocess. When DBT is added exogenously, it is transported intothe cell (likely by passive diffusion). Inside the cytoplasm, the DBTis transformed to HBP via the 4S pathway and then the HBP in-teracts first with the cytoplasmic membrane as it attempts to exitthe cell.

Conclusions. In this work, the mechanism of biocatalyst inhi-bition in the BDS of DBT was investigated. Retention of the finalproduct (HBP) during the BDS process to concentrations in therange of hundreds of �M or higher leads to inhibition of the threeenzymes in the linear part of the 4S pathway. These three enzymes,DszA, DszB, and DszC, have IC50s with respect to HBP of 60, 110,and 50 �M, respectively. Host engineering to reduce retention ofHBP might mitigate inhibition of the 4S pathway by HBP. Thisposes a formidable challenge, as it has been suggested that thecytoplasmic membrane, which is ubiquitous in nature, may beresponsible for HBP retention within the biocatalyst. Protein en-gineering might also be explored as a means to overcome inhibi-tion of the pathway enzymes.

ACKNOWLEDGMENTS

A.A.-F. was supported by the Biotechnology Training Program from theNational Institutes of Health (NIH). Research support from SaudiAramco is also gratefully acknowledged.

Christine Nguyen is gratefully acknowledged for synthesizing HBPS.

REFERENCES1. Soleimani M, Bassi A, Margaritis A. 2007. Biodesulfurization of refrac-

tory organic sulfur compounds in fossil fuels. Biotech. Adv. 25:570 –596.2. Kilbane JJ. 2006. Microbial biocatalyst developments to upgrade fossil

fuels. Curr. Opin. Biotechnol. 17:305–314.3. Gray KA, Mrachko GT, Squires CH. 2003. Biodesulfurization of fossil

fuels. Curr. Opin. Microbiol. 6:229 –235.

4. Gray KA, Pogrebinsky OS, Mrachko GT, Xi L, Monticello DJ, SquiresCH. 1996. Molecular mechanisms of biocatalytic desulfurization of fossilfuels. Nat. Biotech. 14:1705–1709.

5. Xi L, Squires CH, Monticello DJ, Childs JD. 1997. A flavin reductasestimulates DszA and DszC proteins of Rhodococcus erythropolis IGTS8 invitro. Biochem. Biophys. Res. Commun. 230:73–75.

6. Watkins LM, Rodriguez R, Schneider D, Broderick R, Cruz M, Cham-bers R, Ruckman E, Cody M, Mrachko GT. 2003. Purification andcharacterization of the aromatic desulfinase 2-(2=-hydroxyphenyl)benze-nesulfinate desulfinase. Biochem. Biophys. 415:14 –23.

7. Schilling BM, Alvarez LM, Wang DIC, Cooney CL. 2002. Continuousdesulfurization of dibenzothiophene with R. rhodochrous IGTS8 (ATCC53968). Biotechnol. Prog. 18:1207–1213.

8. Naito M, Kawamoto T, Fujino K, Kobayashi M, Maruhashi K, TanakaA. 2001. Long-term repeated desulfurization by immobilized Rhodococcuserythropolis KA2-5-1 cells. Appl. Microbiol. Biotechnol. 55:374 –378.

9. Jia X, Wen J, Sun Z, Caiyin Q, Xie S. 2006. Modeling of DBT biodeg-radation behaviors by resting cells of Gordonia sp.WQ-01 and its mutantin oil-water dispersions. Chem. Eng. Sci. 61:1987–2000.

10. Lichtinger T, Reiss G, Benz R. 2000. Biochemical identification andbiophysical characterization of a channel-forming protein from Rhodo-coccus erythropolis. J. Bacteriol. 182:764 –770.

11. Honda H, Sugiyama H, Saito I, Kobayashi T. 1998. High cell densityculture of Rhodococcus rhodochrous by pH-stat feeding and dibenzo-thiophene degradation. J. Ferment. Bioeng. 85:334 –338.

12. Nakayama N, Matsubara T, Ohshiro T, Moroto Y, Kawata Y, KoizumiK, Hirakawa Y, Suzuki M, Maruhashi K, Izumi Y, Kurane R. 2002. Anovel enzyme, 2=-hydroxybiphenyl-2-sulfinate desulfinase (DszB), from adibenzothiophene-desulfurizing bacterium Rhodococcus erythropolisKA2-5-1: gene overexpression and enzyme characterization. Biochim.Biophys. Acta 1598:122–130.

13. Caro A, Boltes K, Leton P, Garcia-Calvo E. 2008. Description of by-product inhibition effects on biodesulfurization of dibenzothiophene inbiphasic media. Biodegradation 19:599 – 611.

14. Chen H, Zhang WJ, Cai YB, Zhang Y, Li W. 2008. Elucidation of2-hydroxybiphenyl effect on dibenzothiophene desulfurization by Micro-bacterium sp. strain ZD-M2. Bioresour. Technol. 99:6928 – 6933.

15. Nielsen DR, Prather KJ. 2009. In situ product recovery of n-butanolusing polymeric resins. Biotechnol. Bioeng. 102:811– 821.

16. Sutcliffe IC, Brown AK, Dover LG. 2010. The rhodococcal cell envelope:composition, organization and biosynthesis. Biology of Rhodococcus.Microbiology monographs 16. Springer-Verlag, Berlin, Germany.

17. Kilbane JJ. April 1992. Mutant microorganisms useful for cleavage oforganic C-S bonds. US patent 5,104,801.

18. Shuler ML, Kargi F. 2002. Bioprocess engineering: basic concepts, 2nd ed.Prentice Hall, Inc, Upper Saddle River, NJ.

19. Ohshiro T, Suzuki K, Izumi Y. 1997. Dibenzothiophene (DBT) degrad-ing enzyme responsible for the first step of DBT desulfurization by Rho-dococcus erythropolis D-1: purification and characterization. J. Biosci. Bio-eng. 83:233–237.

20. Copeland RA. 2000. Enzymes: a practical introduction to structure,mechanism, and data analysis, 2nd ed, p 397. Wiley-VCH, Weinheim,Germany.

21. Ellis RJ. 2001. Macromolecular crowding: obvious but underappreciated.Trends Biochem. Sci. 26:597– 604.

22. Caro A, Boltes K, Leton P, Garcia-Calvo E. 2007. Dibenzothiophenebiodesulfurization in resting cell conditions by aerobic bacteria. J.Biochem. Eng. 35:191–197.

23. Sikkema J, de Bont JAM, Poolman B. 1994. Interactions of cyclic hydro-carbons with biological membranes. J. Biol. Chem. 269:8022– 8028.

24. Hansch C, Leo A, Hoekman D. 1995. Exploring QSAR, p 97. AmericanChemical Society, Washington, DC.

25. Matsubara T, Ohshiro T, Nishina Y, Izumi Y. 2001. Purification, char-acterization, and overexpression of flavin reductase involved in dibenzo-thiophene desulfurization by Rhodococcus erythropolis D-1. Appl. Envi-ron. Microbiol. 67:1179 –1184.

Biocatalyst Inhibition in Microbial Desulfurization

December 2013 Volume 79 Number 24 aem.asm.org 7817

on July 18, 2020 by guesthttp://aem

.asm.org/

Dow

nloaded from