Substrate Interaction during Aerobic Biodegradation of Creosote

Sci. Technol. 1995, 29, 1944-1952. Substrate Interaction during. AerobicBiodegradation of. Creosote-Related. Compounds: A Factorial BatchExperiment. D...
0 downloads 0 Views 3MB Size
Environ. Sci. Techno/. 1995, 29, 1944-1952

Substrate Interaction duringAerobic Biodegradation of Creosote-Related Compounds: A Factorial Batch Experiment DENIS MILLETTE,*,+ JAMES F. BARKER,+ YVES C O M E A U , + BARBARA J . B U T L E R , + EMIL 0. F R I N D , + BERNARD CLEMENT," AND REJEAN SAMSON+,Waterloo Centre for Groundwater Research, University of Waterloo, Waterloo, Ontario, Canada N2L 3G1, Biotechnology Research Institute, National Research Council Canada, 6100 Avenue Royalmount, Montreal, Quebec, Canada H4P 2R2, and Department of Civil Engineering and Department of Applied Math, Ecole Polytechnique, P.O. Box 6079, Station "Centre V i W , Montreal, Quebec, Canada H3C 3A7

The interactions among seven creosote-related compounds (phenanthrene, fluorene, p-cresol, pentachlorophenol, carbazole, dibenzothiophene, and dibenzofuran) during their aerobic biodegradation in groundwater were studied in factorial experiments. Three separate experiments were conducted in simple batch systems with phenanthrene, p-cresol, and carbazole as the response variable for the first, second, and third experiments, respectively, and other compounds as factors. In general, the more hydrophobic and recalcitrant compounds were more affected by substrate interaction. Mineralization of p-cresol was not affected by substrate interaction. In contrast, p-cresol inhibited mineralization of phenanthrene, but other compounds did not. Mineralization of carbazole was severely affected by the presence of other compounds, p-Cresol was the main inhibitor. Phenanthrene inhibited biodegradation, but to a lesser extent, whereas fluorene enhanced mineralization of carbazole. Pentachlorophenol and dibenzofuran caused an increase in the lag time but did not affect mineralization of carbazole afterward.

lntrodnction Coal tar creosote, in its pure form or mixed with petroleum hydrocarbons (e.g., fuel oil no. 2),has been used as a wood * Corresponding author address: 302-2655 Rufus Rockhead, Montreal, Quebec, Canada H3J 2W8;e-mail address: 75714,655@compuserwcom; telephone: (514) 393-1000, ext 70644; fax: (514) 3924758. University of Waterloo. Biotechnology Research Institute. Department of Civil Engineering, Ecole Polytechnique. ' I Department of Applied Math, Ecole Polytechnique. - Present address: NSERC Industrial Chair in Site Bioremediation, Department of Chemical Engineering, Ecole Polytechnique de Montreal, Montreal, Quebec, Canada H3C 3A7. +

*

1944 ENVIRONMENTAL SCIENCE &TECHNOLOGY / VOL. 29, NO. 8,1995

preservative in North America for approximately 100 years (1). Creosote, a bulk distillate of coal tar (21, is a complex mixture of over 200 major individual compounds with varying molecular weights, polarities, and functionalities along with dispersed solids and products of polymerization. The exact composition depends on the parent coal tar (3), but in general, it is composed of approximately 85% polycyclic aromatic hydrocarbons (PAHs), 10% phenolic compounds, and 5% N-, S-, and 0-heterocycles ( 4 ) . In the subsurface, small but significant quantities of creosote can dissolve into groundwater leaching from creosote-contaminated sites. Since some of the components of creosote have drinking water standards of less than 1 ppb (e.g., chrysene, benzo[alpyrene; ref 51, there is a serious potential for groundwater contamination originating from wood preserving facilities;such contamination has been reported (3, 6- 10). Creosote-contaminated groundwater contains numerous aromatic compounds in addition to ammonia and other inorganics. At the Pensacola, FL, wood preserving facility, the contaminated groundwater was enriched with organic acids, phenolic compounds, single- and double-ring nitrogen, sulfur- and oxygen-containing compounds, and single- and doublering aromatic hydrocarbons. Pentachlorophenol (PCP) is often found concomitantly with creosote in contaminated groundwater (6, 7, 10) because it is used as an alternate wood preservative at wood treating facilities. The potential of subsurface bacteria to remediate creosote-contaminated groundwater has been demonstrated (10-13). However, optimization of the bioremediation of environments contaminated with mixtures of organic pollutants requires a sophisticated understanding of several types of different substrates and of their interaction (14).Substrate interaction during aerobic or anaerobic biodegradation of organic compounds has been observed by various researchers (10, 15-21). Kompala ef al. (15) showed that preferential substrate metabolism is a function of the bacterial growth rate supported by an individual compound and that compounds that support the fastest growth rate will be preferentially degraded first. Results of microcosm assays revealed that methanogenic biodegradation of creosote-related contaminants in aquifer material and contaminated groundwater occurred in the following three-step sequence: (1) quinoline, isoquinoline, benzoic acid, and C3-C6 volatile fattyacids; (2) phenol; and (3) 2-, 3-, 4-methylphenol, 2(1H)quinolinone, and 1(2H)-isoquinolinone (10). Biodegradation of PAHs was not monitored. These results corroborated well with actual field data collected in the methanogenic portion of the aquifer. Most of the organic acids and phenolic compounds and some of the nitrogen-containing heterocyclic compounds were found to have been completely depleted while being transported in the first 150-m portion of the aquifer. The concentration of PAHs, PCP, benzothiophene, dibenzofuran, and 2-methylquinoline decreased in proportion to the conservative tracer in this area, suggesting that they were not degraded during downgradient transport in this portion of the aquifer. Arvin etal. (16)demonstrated interaction among mixed substrates during aerobic biodegradation of benzene. The presence of toluene and o-xylene alone stimulated the rate of

0013-938x/35/0929-1944$09.00/0

%. 1995 American Chemical

Societv

biodegradation of benzene, but pyrrole strongly inhibited benzene degradation. Likewise, Alvarez and Vogel (I7) observed that toluene enhanced biodegradation of benzene during aerobic biodegradation of BTX. However, recent studies indicated that when toluene and benzene were present concomitantly, toluene was catabolized and benzene oxidation was delayed (19, 21). Thus, interaction among substrates can result in inhibition or enhancement of the mineralization of a given compound. Coal tar creosote being a complex mixture of numerous chemicals, it is conceivable that substrate interaction plays a key role when creosote is degrading in an oligotrophic groundwater environment. Various numerical models have been developed and used to predict biodegradation in aquifer systems (18,2228). Bacterial growth and substrate utilization is usually modeled using Monod or dual-Monod kinetics. Generally, a single organic substrate is considered (23, 25) although some models can simulate the biodegradation of more than one substrate (IS,24, 27, 28). The model of Semprini and McCarty (18) accounts for interaction among the growth and secondary substrates but in general, the above numerical models do not specifically address interaction among a number of substrates present simultaneously. Numerical models are important tools to assist in the understanding of complex flow systems and to assess the efficiency of groundwater remediation scenarios. However, the application of many existing numerical models is limited because they do not address the complexities of substrate interaction. The accuracy of these models could be improved by additional research to identify interactions among substrates of a contaminant matrix and to assess their impact on the overall efficiency of remediation of creosote-contaminated groundwater. The objective of the present project was to investigate some of the interactions occurring during the aerobic biodegradation of selected creosote-related compounds using a factorial experimental design. It was decided to focus on the more soluble compounds because they are more likely to leach into groundwater systems. The array of compounds selected for this study covers the three classes of compounds found in contaminated groundwater at wood treating facilities. Phenanthrene (PHEN) and fluorene (FLUOR) were selected as PAHs, p-cresol (PC) and pentachlorophenol (PCP) were selected as phenolic compounds, and finally, carbazole (CARB), dibenzothiophene (DBT), and dibenzofuran (DBF) were selected as N-, S-, and 0-heterocycles, respectively. PCP is not a constituent of creosote, but it was included in the study because it is used as an alternative wood preservative and is often detected in creosote-contaminated groundwater.

Materials and Methods Chemicals. The radioactive substrates used included [9-I4Clphenanthrene(8.3mcilmmol), p-[UL-'4C]cresol(5.6 mCilmmol), and [ U L - ' 4 C l c a r b ~ ~(7.9 l e mcilmmol), all purchased from Sigma Chemical Co., Radiochemical Division, St. Louis, MO. The purity of the labeled chemicals, as specified by the manufacturer, was >98%. High purity (99+%)phenanthrene, carbazole, dibenzothiophene, and dibenzofuran were obtained from Aldrich Chemical Co., Milwaukee,WI. Fluorene (98.3%pure) was obtained from Sigma Chemical Co. High-purity pentachlorophenol (99+%) was obtained from Supelco Inc., Bellefonte,PA. High-purity p-cresol (99+%) was purchased from Fluka Chemika-

BioChernica, Switzerland. Dichloromethane, Accusolv grade (purity >99.7%), was obtained from Anachemia Science,Montreal, Quebec, Canada. Toluene, HPLC grade (purity > 99.7%),was purchased from Caledon Lab. Ltd., Georgetown, Ontario, Canada. Preparation of Stock Solutions. A modified version of the method developed by Mihelcic and Luthy (29)was used to prepare the nonradioactive stock solutions of the compounds of interest. A known amount of crystalline compound was added to a sterile mineral salt medium solution (MSM),stirred at 40 "C for 24 h, and then stirred at room temperature for an additional 24 h, in a closed volumetric flask covered with aluminum foil. The solution was then filtered through an extra-thick borosilicate glass fiber filterwithout binder (No. 66209; Gelman Sciences Inc., Ann Arbor, MI) with a nominal retention rating of 1ym to remove any undissolved, suspended crystals. This resulted in stock solutions of concentrations near the maximum solubility, as confirmed by GUMS. To further increase the efficiencyof dissolution of PHEN, the crystals were first ground with a mortar and pestle. Because of the relatively hydrophilic nature of PCP and PC, stock solutions of those compounds only required stirring for 20 min, at room temperature, and filtering was not necessary. Radioactive compounds were either supplied in a crystalline form or dissolved into toluene. The crystalline compounds were first dissolved into toluene. Then, preparation of aqueous radioactive stock solutions was achieved by: (1) adding a known volume of the toluene solution into a 60-mL glass tube with Teflon-lined screw cap; (2) evaporating toluene by slowly hand rotating the tube; (3) adding a known volume of sterile MSM; and (4) agitating the mixture for 20 min at full speed on a Wrist Action Shaker Model 75 (Burrell Co., Pittsburgh, PA). The final activity of each aqueous radiolabeled stock was determined by measuring the radioactivity of a 100-pL sample aliquot with a scintillation counter. The residual concentration of toluene in the aqueous radioactive stock solutions was verified and found to be below the limit of detection of the instrument. Aqueous radioactive stocks were prepared within 24 h of microcosm preparation, and required volumes were added to the microcosms by syringe. AnalyticalMethods. SolutionsofPAHs and heterocyclic compounds were extracted by liquid-liquid extraction following U.S. EPA Method 3510. Dichloromethane was used to extract the samples,and4-fluorobiphenylwas added as the recovery standard. The concentration of PAHs and heterocyclic compounds in the dichloromethane extract was determined by GUMS (Hewlett Packard 5970 MSD connected to a HP 5890 GC) following U.S. EPA Method 8270 and using 1,2,3-trichlorobenzene as an internal standard. One microliter of the extract was injected onto a 30 m x 0.25 mm DB-5 capillary column (J&WScientific Inc., Rancho Cordova, CAI maintained at 55 "C for 3 min and elevated to 280 "C at a rate of 7.5 "C min-I. The carrier gas was helium, and injector and detector temperatures were maintained at 275 and 280 "C, respectively. Radioactivity of the trapped I4CO2was determined by liquid scintillation counting. The KOH samples were transferred to ACS scintillation cocktail (Amersham,U.K.) and analyzed using a Packard Tri-Carb Model 4530 liquid scintillationcounter (Packard Instrument Co. Inc., Downers Grove, IL) with the appropriate quench curve. Preparation of Microcosms. The aquifer material used in this studywas collected from an unconfined sandy aquifer VOL. 29, NO. 8,1995 /ENVIRONMENTAL SCIENCE &TECHNOLOGY

1945

at Canadian Forces Base Borden, located approximately 100km northwest of Toronto, Canada. The aquifer material consists of medium- to fine-grained sand of glaciofluvial origin (30) with low organic carbon and clay contents (31). Pristine material was collected approximately 60 cm below the water table in June 1992 and stored at 4 "C until required in January 1993. Studies were conducted using simple batch systems in 100-mLserum bottles sealed with Mininert valves (Supelco Inc., Bellefonte, PA). The glassware and equipment used in microcosm preparation were sterilized by autoclaving, and the microcosnis were prepared in a laminar-flow hood. All microcosms of an experimental block were prepared concurrently. A sample of 50 g of moist aquifer material, 17 mL of MSM, and the required volume of stock solutions of radiolabeled and unlabeled compounds were added to each bottle, leaving a headspace of approximately 85 mL. The headspace air was calculated to contain more than sufficient oxygen to allow for complete mineralization of the organic matter inherent to the aquifer material plus all added carbon. The actual oxygen concentration in the headspace was measured periodically,and it was confirmed that the actual decrease in the oxygen concentration of the headspace was neghgible throughout the experiment. Each microcosm also included a 14C02trap consisting of a 5-mL glass test tube filled with 1 mL of 0.5 N KOH and placed within the serum bottle. The MSM was modified from Greer et al. (32) and contained (mg L-I of deionized water) the following: Na H2PO4, 870; K2HP04, 2260; (NH4)2S04, 1100; and MgS04.7 H20, 97. To this solution was added 1 mL L-' of a trace metals solution composed of (mg L-') the following: Co(N0I2.6HnO,291; AlK(S04)2.12H20,474; CuS04,160;ZnSOc7 H20, 288; FeS04.7 HnO, 2780; MnS04.H20, 1690; NanMoO4-2H20,482; and Ca(N0&*4H20,2362. The final pH of this medium was 7.5. After preparation, the microcosms were shaken at 160 rpm for 10 min to ensure complete mixing of the liquid phases and then incubated in the dark at 10 "C. All microcosms were hand shaken for 2 min daily until most of the mineralization had occurred and every second day afterward. Factorial EsqKriment Design. Three separate fractionalfactorial experiments were conducted. In each, there were six independent variables (factors) and one dependent variable (response variable; Figure 1). The percent mineralization of PHEN, PC, and CARB were the response variables for the first, second, and third experiments, respectively. The remaining compounds were the independent variables. The experimental procedure for each of the three response variables was designed as a 26-2 fractional-factorial design (Le.,one-fourth of a full factorial design with two levels) (33). The levels were defined as the concentration of the factors. Each experiment was divided into two independent blocks of eight runs each. The experimental design for the first block was a nonsaturated two-level eight-run Plackett-Burman design (Table 1).In Table 1, the three-factor interactions (e.g.,ABD, CDF, etc.) are not shown because generally, of all multi-factor interactions, onlythe two-factor interactions are considered the most likely to occur (33). The design for the second block was a reflection of the first block. Reflection is achieved by inverting the level of the factors in all runs; this results in a switch in the sign of the two-factor interactions (Le., contrast equation A - BC - DE becomes A + BC + 1046 1 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 29. NO. 8 , 1 9 9 5

Independent variables (Factors)

9.r:,,

..,.'...

'.'a*

Phenanthrene

.

'. :' . :' I

14C Labeled

'h

I.

Response variables

Fluorene

CH 3D

O

H

pCresol

clcc CI

CI

PCP

carbazole

/;r;;;.'

Dibenzothiophene

/'

Note: Factors at two levels (+) or (-), model compounds at one level (+I only

DibenGfuran

FIGURE 1. Combination of compounds used for the study. TABLE 1

Array for Basice Eight=RunPlackett-Burmann Design for Six Factors contrast labels and aliasesb A -BC -DE response variable

B C -AC -AB -DF -EF

__

3 4

5 6

7 8

-AD

-BF

-CF

F G C -BD -AF -CE -BE -C D

compound assigned to each contrast label

phenanthrene FLUOR PC PHEN FLUOR pcresol carbazole PHEN FLUOR run 1

2

E

D

-AE

-

+ + + +

-

PCP CARB DBT DBF PCP CARB DBT DBF PCP DBT DBF PC

-

+

+

-

+ f + + -

+

-

+

-

f

+

+ -

-

-

+

4

-

T

+

-

-

-

-

+ -

f

-

+ -

a (+) high level; (-1 Low level. In the reflected design, all of the plus and minus signs in the array are inverted. This results in a switch in the sign of the two-factor interactions (e.g., equation A - BC - DE becomes A + BC DE). Three-factor interactions (e.g., ABD, DEF, BCF, etc.) not shown in this table because their effect is normally considered negligible (33). Although contrast label G remained unassigned, the two-factor interactions associated with this contrast label must still be included in the statistical analysis.

+

DE). The experimental design for each block was of resolution 111. A design of resolution I11 does not have any

of its main effects confounded with other main effects but only with two-factor or higher order interactions (33).For

TABLE 2

TABLE 3

Concentration of Compounds for Study compound PAHs phenanthrene fluorene phenols pCresol PCP Heterocycles carbazole (N) dibenzothiophene (S) dibenzofuran (0)

concn at saturation (%.I’ (ppb)

Results of Mass Balance Experiment

concnb low I-) hiah - .(+I ppb PPb

660 95 1

33 48

165 238

405,655 3,277

20,283 164

101,414 819

662 43 159

33 10 105

166 50 527

a Calculated using Raoult’s law; mole fractions estimated based on the typical creosote composition presented in Mueller et al. ( 4 ) . Concentrations correspond to s1$4 (high) and ~ ~ $ 2(low) 0 except for dibenzothiophene, where values were increased in order to be well above the detection limit, and for dibenzofuran where s1$4 and ~ ~ $ 2 0 values are based on an experimentally determined sleof 2000 ppb (35).

example,the main effect Ain the basic design is confounded with the -BC and -DE interactions (Table 1). In the reflected design, main effect A is confounded with BC and DE. Thus, by combining the effects of the two blocks, the design becomes one of resolution IV,and main effects can be separated from two-factor interactions. When some of the main effects are negligible, combining the effect of two blocks can allow the identification of significant two-factor interactions, which may be indicative of synergistic or antagonistic effects. To set the concentrations of the compounds to be used in the experiments, a concentration at saturation (s3, defined as the dissolved concentration of the compound, in pure water when in equilibrium with the creosote matrix was first calculated for each individual compound using Raoult’s law. Mole fractions of the compounds of interest were calculated according to the typical creosote composition of Mueller et al. ( 4 ) . Liquid-phase solubility rather than crystal solubility was used to calculate sle because compounds in the creosote matrix are already in a liquid state (34). A higher concentration (+) and a lower (-) concentration were defined for each compound based on SI?(Table2). Generally, the higher level was 114 slL.to reflect contamination near the source, and the lower level was 1/20 sle to represent conditions farther downgradient of the source of contamination. Exceptions to this were dibenzothiophene, where added amounts were increased to avoid working near the analytical detection limit, and dibenzofuran, which was observed to have a slemuch higher than the calculated value in laboratory experiments (35). “Control” (response variable compound + aquifer material MSM) and “poisoned control” (the same, plus 0.2% sodium azide) microcosms were prepared for each block. Four replicate microcosms were prepared for every experimental condition. Mass Balance Jlxperiment. M e r each fractional-factorial experiment was completed, a mass balance was conducted on two control microcosms and on one poisoned control. A total of nine microcosms were therefore evaluated. Recovery of radioactive material proceeded as follows: (1) microcosms were acidified to pH 2 and were connected immediately to a series of three traps (one containing a solvent, two containing Carbo-Sorb (Packard Instrument Co. Inc., Downers Grove, IL), a COz absorber)

+

typeof compd microcosm

4 (dpm)

mass A, A, balance (dpm) (dpm) (dpm) (dprn) (YO)

40, AI

76.9 PHEN control 90 015 61 290 1 241 90 6 574 90015 63073 2 148 119 7 187 80.6 control PcontroP 90015 2 6 2 5 3 8 1 0 8 67987 82.7 mean: 80.0 PCb control 100894 82832 428 132 6 6 7 2 89.3 control 100894 77814 608 163 10793 88.6 mean: 89.0 CARE control 61 096 31 848 507 73 28866 100.3 control 61 096 32 276 518 8 13614 76.0 PcontroP 61 096 154 5 0 2 1 36 44273 81.0 mean: 85.8 a Poisoned control. *Vial of the poisoned control broken while conducting the mass balanceexperiment. A,, initial activityintroduced intothe microcosm;~~2,finalcummulativeactivitymineralizedto14CO1; A,, residual activity in the liquid phase; A., activity released during acidification; A,, residual activity sorbed onto the soil matrix.

to collect any I4CO2existing as dissolved or precipitated carbonate; (2) the Mininert valve was removed, and the liquid phase was filtered; total radioactivity in the liquid phase was then determined: (3) the filter was mixed with the soil in the microcosm, and the mixture was dried by adding 50 g of anhydrous sodium sulfate (Anachemia, Montreal, Quebec, Canada); (4) solvent was then added to the soil and filter mixture, microcosms were capped with a Teflon-lined septum and were agitated overnight on a Wrist Action Shaker; (5) the solvent was removed by syringe, and radioactivity recovered in the solvent was determined; (61 the extraction procedure was repeated twice more, but agitation lasted only 2 h and (7)1 g of dehydrated soil was placed directly into a scintillation vial, and residual activity in the soil was measured.

Results and Discussion Mass Balance Experiment. Results of the mass balances conducted on nine microcosms are shown in Table 3. Mass balances of 80%, 89%,and 85.8% were obtained for PHEN, PC, and CARB, respectively. In the nonpoisoned controls, the final cumulative activity mineralized to l4CO2during the course of the experiment was relatively high compared to the initial radioactivity introduced into the microcosm (A”). The activity released during acidification (A,) was negligible compared to A,, indicating that the amount of l4CO2in a precipitated or dissolved carbonate form was negligible and that the concentration of COZ in the head space was low. This was expected considering that COz of the head space was trapped in a KOH solution. In all microcosms, residual radioactivity in the liquid phase (A3 was considerably lower than that sorbed onto the aquifer material (As),indicating that a large proportion of the compound was sorbed onto the aquifer material. The large proportion of PHEN and CARB sorbed onto the aquifer material of the poisoned controls was consistent with the sorption distribution coefficients measured (data not shown). In a separate experiment (data not shown), mass balances were conducted on microcosms amended with radiolabeled FLUOR and PCP, and results of 108%and 91% were obtained, respectively. Mass balances for DBT and DBF were not determined because radiolabeled compounds VOL. 29. NO. 8. 1995 / ENVIRONMENTAL SCIENCE 81 TECHNOLOGY

1

1947

~,>

.' ,,,...

40

m MAX%

&? 20 10

........ MAX% 0.4

os

"8

I

FIGURE 2. Graphical representation of the response parameters.

were not commercially available. To assess the extent of partitioning of DBT and DBF into the headspace, sterile 200-mLsolutionsofeachwerepreparedandstoredat room temperature in sterile 500-mL amber glass bottles. After 3 months, the concentration of the solution of DBT and DBF remainedunchanged,indicatingthat compoundlosses to the headspace were minimal. Henry's law constant for DBT was not available, but that of DBF was estimated as lo-' aun m3 mol-' using vapor pressure data of Yaws (36) and assuming Henry's law constant as the ratio of vapor pressure to solubility. Again, thissuggests that partitioning of DBF into the headspace was not important. Thus, it is unlikely that either DBT or DBF was lost from the closed serum bottle system during the experiment. Fractional-Pactorial Jkperiment. The results of this experiment were interpreted using the following response parameters (Figure 2): (1) MAX% is defined as the final % of the added radioactivity accumulated as W02;(2)LAG is defined as the time required to reach 115 MAX%; 13) RMAX is defined as the rate of mineralization between LAG and 2/3 MAX%); (4) RAVG is defined as the average rate of mineralization between 0% and 2/3 MAX% (4) BDI is definedasthe biodegradationindex;and (5) NBDI isdefined as the normalized biodegradation index The relationship defining BDI is BDI = RAVG x MAX% 100% ~

The equation defining NBDI is NBDI =

BDI, BDI,

where I stands for run and c stands for control. Thelagperiod before detectablemineralizationisagood indicator of the ability of the indigenous microbial consortium to adapt to biotransform the compound. LACwas defined as 115 MAX% as an analytical convenience to estimate the point where the lag had ended and detectable COz evolution ensued. The interval from 0 to 2/3MAX% was selected to calculate RAVG because this interval encompasses the lag period and the period during which most of the degradation occurs. The value of 2/3 MAX% g e n e d y coincided with the upper point of inflection of the curve of mineralization. RAVG is therefore an indicator of the ability of the microbial consortium to mineralize the majority of the mass of a compound introduced into a system. MAX% is also an important parameter. A small MAX% is agood indicatorofthe assimilationofthesubstrate into biomass andlor of incomplete mineralization &e,, possibility of the presence of intermediary compounds). 1948 m ENVIRONMENTAL SCIENCE &TECHNOLOGY I VOL. 29. NO. 8,1995

BDI is a lump parameter combining response parameters RAVG, MAX%. and implicitly LAC and R M A X This index may prove to be a useful indicator of the overall ability of the microbial consortium to mineralize a specific compound. NBDl wasdeveloped to study theeffect ofsubstrate interaction on the minemlizationofthe compound selected as response variable. A value of Iis indicative of minor or no substrate interaction (Le. the curve of mineralization obtained is similar to that of the control). A value >I is indicative of positive substrate interaction. and a value of c I is indicative of inhibition. Curves of mineralization of PHEN. PC, and CARB (basic design) are shown in Figure 3. Curves of mineralization of the reflected design are not shown. but results were consisrent with those of the basic design. Data used to generate thecurveswere obtained fromfourreplicatesand were corrected for a poisoned control. The total activity recovered from the '*CO2 traps of the poisoned controls was minimal and never exceeded 3.1%. 1.5%. and 0.2% for PHEN. PC, and CARB, respectively. Variability among the four replicatesoftheAlA,. ratios corrected for the poisoned control. expressed as the average standard deviation ofA/&, was 4.5%. 1.5%. and 1.6% for PHEN. PC. and CARB, respectively. Phenanthrene. A qualitative examination of the curves of mineralization of PHEN (Figure 3a) suggest that its biodegradation was affected by the presence of other compounds. A similar behavior was observed with the reflected design (data not shown). In general. substrate interaction caused an increase in LAG andlor MAX%. The curve of mineralization for run 1 (basicdesign), which had all factors at the lower level (Table I), is very similar to that ofthecontrol,indicatingthatfactorsatalowconcentration did not alter degradation of the response variable. As the concentration of other factors increased, curves of mineralization of PHEN diverged from that of the compound alone 1e.g.. runs 2 and 6). Bar graphs of response parameters MAX%. LAG,RMAX, RAVG. and NBDl for the basic and the reflected design of PHEN were developed. An illustrativeexample for variable R M A X is shown in Figure 4a (left). Except for MAX%. the responseobtained with the reflected design was practically aperfecl reflectionofthat ofthe basicdesign. For example, with the reflected design, R M A X for runs 3-6 was higher than for other runs 1i.e.. ;;.5.5% d compared to ~ 3 . 5 % d-.'),but with the basic design. it was the contrary. In all four runs, only the PC concentration remained constant (Table l), suggesting that PC inhibited biodegradation of phenanthrene. A n analysis of variance (ANOVA) was conducted by combining results of the basic and the reflected design. This allowed determination of the level of significance of eachmain- and two-factor interactionand thedevelopment of scree plots 1e.g.. Figure 4a; right) to determine which of the significant interactions represented a real signal. Sometimes,apvalue indicatesstatisticalsignificance where the difference is not practically significant. This situation may occur with large data sets or when there is a small amount of variation in the data (37). In scree plots. contribution to the total sum of the squares ICTSS) of the main. and two-factor interactions are plotted in descending order of magnitude. The objective is to find an elbow in the curve. Points above the elbow will be likely to indicate a signal of practical significance. while those below will most likely be noise. A close analysis of the scree plots of

'

A

t

80

Y

Run 1

Run 6

+ J(t

~

Run 7

2+ + Run 3

Run 8 -8Run 4 Model Alone Run 5

+ I"

*"

LY

I""

8-

.

Time (d)

1w 80 U

Fi '=

60

$c

40

Run4 Model + Alone

s 20 0

Time (d)

+-

NH

Run3

Run8

Run4 -Y7 Run 5

Mode

*-e-

11 140

+ -

Time (d) FIGURE 3. Mineralization curves for the basic design experiments with (a) PHEN. lb) PC. and Is) CARB as model compounds. The vettical dolled line indicates a change in the time male.

the PHEN experiment revealed that PC was the only substrate causingarealsignal. The effect of this interaction on the response parameters is shown in Table 4. In this table.vaJuesforthecontrolare the average offourreplicates, means are the average of the respective response parameter in all 16 runs conducted in quadruplets, and interactions listed are all significant at the 0.01 level. Results of the ANOVA confirmed that the effect due to PC resulted in an increase in LAG and a decrease in RMAX and MVG. but MAX% for PHEN was not significantly affected thereby, increasingthe concentration of PC from the lower level (-) to the higher level (+) resulted in a decrease in NBDI of 0.44.

A close examination of the curves of mineralization of PC and PHEN alone and PHEN in a mixture (run 1R) revealedthat,whenPHENw a s d e g r a d i n g i n a m i , t h e r e was an increase in LAG. In fact,the onset ofPHEN mineralization coincided with the upper point of inflection of the

curve of mineralization of PC degrading alone. This indicates that mineralition of PHEN was delayed until most of PC had been mineralized, suggesting that PC may have been preferentially utilized by the microbial consortium. p-Cresol. Data in Figure 3b suggest that mineralization of PC was not affected by the presence of the other compounds in the mixture. This was indicated by the lack of deviation among the curves of mineraliition of each individual run. Similar results were obtained with the reflected design (datanot shown). Variationintheresponse parameters among the runs was minimal, and response due to the reflection of the design did not occur (e.g., Figure 4b; left). A detailed ANOVA was performed on the data, and results confirmed that substrate interaction did not playa major role duringaerobic biodegradationof PC (e.g., Figure 4b right). Carbazole. A qualitative examination of the curves of mineralization of CARB (Figure 3c) suggests that biodegVOL. 29. NO. 8.1995 I ENVIRONMENTAL SCIENCE &TECHNOLOGY m 1949

'J

inhibition increased. The strong variation in the response parameters among the runs of a single block (e.g., Figure 4c; left) is indicative of substrate interaction. The response obtained with the reflecteddesign was often a reflection or partial reflection of that of the basic design. However, it is not as prominent as with phenanthrene. The ANOVA, combining results of both designs, continned that there was substantial substrate interaction during aerobic mineralization of CARB (Figure 4c (right); Table 4). Data in Table 4 can be used to develop regression equations that relate the response variables to the concentration ofthe factorscausinginteractionsofsignificance importance. For example, the value of LAG for carbazole can be calculated as follows:

60 50 40

E 30

2

*O I O

nn

L4GuRB = 26.92

1 + $6.26DBF + 6.07PC t

3.91PCP

Facton

PHENxFLUOW

FIGURE 4. Illustrative examples of the variation of RMAX among runsllettl and respectivescree platslrightlfor lalPHEN.lb1 PC.and IC) CARB. Check marks indicatethatthecompound most likelycaused a signal 01 practical importance; CTSS represents the contribution to the total sum of the squares; RMAX is the maximum rate of mineralization. Table 4. Summawof me Deteded Idmdbns" Re~ponseV~rlaMes LAG

MAX%

RMAX

RAVG

NED1

radation of CARB was strongly inhibited by the presence ofsome ofthe factors. LAGgenerallyincreasedwithrespect to the control and/or MAX% decreased. AS for the PHEN experiment, the response with run 1 of the basic design was similar to that of the control, but as the concentration of some of the factors increased (e.g., runs 7 and E), 1960 m ENVIRONMENTAL SCIENCE &TECHNOLOGY I VOL. 29,

NO. 8,1995

(3)

In the above example, the Concentration of the factors is expressed as +l (high level) or -1 (low level). For example, eq 3 predicts that when all factors of importance (DBF,PC, PCP, andPHEN)are at their lowest concentration (-11, the resulting LAG would be 20.3 d. Data in Table 4 revealed that PHEN and PC were the main inhibitors,whereas FLUOR enhanced mineralization of CARB. Increasing the concentration of PHEN from (-) to (+) resulted in a decrease in all response parameters. Although the interaction due to PHEN caused a decrease in LAG (-2.95 d), it generally inhibited mineralization of CARB, resultinginareductioninNBD1 byO.11. Inhibition due to PC was more pronounced. Increasing its concentrationfrom (-) to (+) resultedinanincreaseinLAG(+6.07 d) and a decrease in MAX% (-7.5%). RMAX (-0.4% d-9, and RAVG (-0.23% d-9, thereby reducing NBDI (-0.26). DBF and PCP both caused limited inhibition that resulted in an increase of LAG of 6.26 and 3.91 d, respectively, but other response parameters were not affected by their presence. Conversely,FLUOR enhanced mineralization of CARB. This was shown by an increase in MAX%, RMAX and RAVG, which resulted in an increase in NBDI of 0.11. LAG,on the other hand, was not influenced by FLUOR. The statistical analysis revealed that there might have been interaction among some of the independent factors. The combined effects PHEN x FLUOR and DBT x DBF was 0.19% d-' for RMAX 0.09% d-' for RAVG, and 0.10 for NBDI. The experimental design does not allow for separation of the two-factorinteractions among each other. Thus, theeffectassociatedwithcontrastequationPHEN x FLUOR DBT x DBF might in fact be the effect of the PHEN x FLUOR interaction and/or that of the DBT x DBF interaction. A close examination of the data in Table 4 reveals that, in general, main effects associated with PHEN and FLUOR prevailed over those of DBT and DBF. Hence, we can hypothesize that PHEN x FLUOR was the predominant two-factor interaction. The reader should note that CTSS of this two-factor interaction was of the same order as that of the experimental error (Figure 4c; right). This indicates that although the effect of FLUOR varied as a function of the concentration of PHEN and vice versa, this interaction generated limited variabilityin the response of the design. Factorial analysis-type experiments do not lead to "mechanistic" conclusions, but one may speculate and produce hypotheseswithrespect to the mechanism behind the observed substrate interactions based on results

+

DE1

- 2.95PHENl

obtained and compound structures, information in the literature on degradation pathways and the enzymes involved, etc. These hypotheses may then be tested by further experimentation. For example, the following speculations were addressed regarding the interactions of the CARB experiment. The molecular structure of FLUOR is very similar to that of CARB (Figure 11,and consequently, the presence of FLUOR might have induced the production of an enzyme having catalytic activity toward CARB. A similar situation, where dibenzofuran-grown cells cometabolized fluorene, has been reported (38).Inhibition due to PC and PHEN might be the result of preferential use of both substrates over CARB. When provided as sole substrates, PC and PHEN are less recalcitrant than CARB, and their degradation is faster. Kopala etal. (15)evidenced that compounds supporting the fastest growth rate will be preferentially degraded first. A simple explanation for the observed synergism between PHEN and FLUOR with respect to CARB degradation is not readily apparent. The above findings suggest that substrate interaction plays a major role during the aerobic biodegradation of complex mixtures of contaminants in groundwater and that biodegradation of the more recalcitrant compounds is unlikely to occur near a source of contamination because of the presence of easily biodegradable compounds at a high concentration. This is in general accordance with the findings of Godsy et al. (IO),who observed the disappearance of most organic acids, phenolic compounds, and some nitrogen-containing heterocycles in the first 150 m of the aquifer but the persistence of the more recalcitrant compounds such as PAHs, PCP, benzothiophene, and dibenzofuran within the same section of the aquifer. The reader should note, however, that their results were obtained under methanogenic conditions and that biodegradation of carbazole was not monitored. Provided oxygen and nutrients are not limited, our results suggest that biodegradation of the more recalcitrant compounds should happen while migrating farther downgradient in the aquifer after most of the easily biodegradable compounds have been transformed,and biodegradationof these compounds will be strongly affected by substrate interaction. Hence, this study confirmed that numerical models developed to predict the fate of multi-component contaminant plumes must address such interactions. However, before these interactions can be incorporated into models, additional experiments will be required to verify if generalizations can be developed. Additionally, some experiments should be repeated in columns to verify if the interactions observed under batch conditions are significant under hydrodynamic conditions.

Conclusion A qualitative examination of the results of this study suggests that the fractional factorial design was helpful in identifymg interactions occurring during the aerobic biodegradation of creosote-related compounds. Mineralization of p-cresol was not affected by the presence of other compounds. In contrast, there was some substrate interaction during mineralization of phenanthrene. p-Cresol was the only factor causing interaction, and its presence in the mixture inhibited mineralization by causing an increase in the lag period (LAG) and a decrease in the maximum (RMAX) and average (RAVG) rates of mineralization, hence a decrease in the normalized biodegradation index (NBDI). Substrate

interaction was more apparent during biodegradation of carbazole. This was shown by a general increase in LAG and a decrease in the maximum percentage of the added radioactivity accumulated as I4CO2(MAX%)when carbazole was degrading in a mixture. The complexity of these interactions was greater than those found during the phenanthrene experiment. p-Cresolwas the main inhibitor, but phenanthrene also inhibited biodegradation of carbazole to a lesser extent, whereas fluorene enhanced mineralization of carbazole. Interaction among independent factors (phenanthrene and fluorene)was also observed, but the overall effect of this two-factor interaction on the mineralization of carbazole was of a lesser extent. In general, the presence of other substrates affected biodegradation of the more hydrophobic, recalcitrant compounds (carbazole and to a lesser degree phenanthrene) but not p-cresol.

Acknowledgments Funding for this project was provided by the National Research Council Canada and the Natural Sciences and Engineering Research Council of Canada. The authors thank researchers and technicians at the Biotechnology Research Institute and at the University of Waterloo for their guidance and valuable assistance.

Literature Cited (1) Feenstra, S.; Cherry, J. A. Presented at The Eleventh Annual Meeting of the Canadian Wood Presening Association; Toronto, Ontario, Canada, Nov 6-7, 1990. (2) Roche, I. N. Presented at Forest Products Research Society-Great Lakes Section Meeting; Apr 15, 1952; p 10. (3) Arvin, E.; Flyvbjerg, 1.1. Insf. WaferEnviron. Manage. 1992, 6, 646-651. (4) Mueller, J. G.; Chapman, P. I., Pritchard, P. H. Environ. Sci. Technol. 1989, 23 ( l o ) , 1197-1201. (51 US.EPA. Drinking wafer regulations and health advisories; EPA 822-R-94-001; U. S. Environmental Protection Agency: Washington, DC, 1994. (6) Bedient, P. B.; Rodgers, A. C.; Bouvette, T. C.; Tomson, M. B.; Wang, T. H. Ground Water 1984, 22 (3),318-329. (7) Goerlitz, D. F.;Troutman, D. E.; Godsy, E. M.; Franks, B. J.Environ. Sci. Technol. 1985, 19 (lo), 955-961. (8) Thomas, J. M.; Lee, M. D.; Scott, M. 1.;Ward, C. H.J. Ind. Microbiol. 1989, 4 , 109-120. (9) Ellis, B.; Harold, P.; Kronberg, H. Environ. Technol. 1991, 12, 447-459. (10) Godsy, E. M.; Goerlitz, D. F.; Grbic-Galic, D. Ground Wafer1992, 30 (21, 232-242. (111 Ehrlich, G. G.; Goerlitz, D. F.; Godsy, E. M.; Hult, M. F. Ground Wafer 1982, 20 (61, 703-710. (12) Fredrickson, J. K.; Brockman, F. J.; Workman, D. J.; Li, S. W.; Stevens, T. 0.Appl. Environ. Microbiol. 1991, 57 (3),796-803. (13) Madsen, E. L.; Sinclair, J. L.; Ghiorse, W. C. Science 1991, 252, 830-833. (141 Rittmann, B. E. Wafer Sci. Technol. 1992, 25 ( l l ) , 403-411. (15) Kompala, D. S.; Ramkrishna, D.; Jansen, N. B.; Tsao, G. T. Biotechnol. Bioeng. 1986, 28, 1044- 1055. (16) Arvin, E.; Jensen, B. K.; Gundersen, A. T.Appl. Environ. Microbiol. 1989, 55 (121, 3221-3225. (17) Alvarez, P. J. J.; Vogel, T. M.Appl. Environ. Microbiol. 1991, 57 ( l o ) ,2981-2985. (18) Semprini, L.; McCarty, P. I.. Ground Wafer 1992,30 (1),37-44. (19) Burback, B. L., Perry, J. J.Appl. Environ.Microbiol. 1993,59 (41, 1025-1029. (20) SaBz, P. B.; Rittmann, B. E. Biodegradation 1993, 4 , 3-21. (21) Oh, Y.-S.; Shareefdeen, 2.; Baltzis, B. C.; Bartha, R. Biofechnol. Bioeng. 1994, 44, 533-538. (22) Sykes, J. F.; Soyupak, S.; Farquhar, G. J. WaferResour. Res. 1982, 18 (11, 135-145. (231 Borden, R. C.; Bedient, P. B.; Lee, M. D.; Ward, C. H.; Wilson, J. T. Wafer Resour. Res. 1986, 22 (131, 1983-1990. (24) Frind, E. 0.;Sudicky, E. A,; Molson, J. W. In Proceedings of the Symposiiini held during the Third L4HS Scientific Assembly; IAHS: Baltimore, MD, 1989; pp 89-96.

VOL. 29, NO. 8, 1995 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

1951

( 2 5 ) MacQuarrie, K. T. B.; Sudicky, E. A. Water Resour. Res. 1990,26 (21, 223-239. MacQuarrie, K. T. B.; Sudicky, E. A,; Frind, E. 0. WaterResorir. Res. 1990, 26 (2), 207-222. Chen, Y.-M.; Abriola, L. M.; Alvarez, P. J.; Anid, P. 1.; Vogel, T. M. Water Resour. Res. 1992, 28 (7), 1833-1847. McNab. W. W. , Narasimhan. T. N. Water Resoitr. Res. 1993,29 (81, 2737-2746. Mihelcic, J. R. , Luthy, R. G. Appl. Enuiron. Microbiol. 1988, 54 ( 5 ) ,1182-1187. MacFarlane, D. S.; Cherrv, J. A.; Gillham, R. W.; Sudickv. E. A. I Hydrol. 1983, 63, 1-29. Patrick, G. C.; Ptacek, C. J.; Gillham, R. W.; Barker, 1. F.; Cherry, J. A,; Major, D.; Mayfield, C. 1.; Dickhout, R. D. The behaviour

of soluble petroleum product derived hydrocarbons in groundioiater. Phase r; Pace Report No.85-3; Petroleum Association for Conservation of the Canadian Environment: Ottawa, Ontario, Canada, 1985. Greer, C. W.; Hawari, I.; Samson, R. Arch. Microbiol. 1990, 154, 317-322.

1952

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 29. NO. 8.1995

I

(33) Wheeler, D. J. Understanding industrial experimentation, 2nd. ed.; SPC Press Inc.: Knoxville, TN, 1990; 388 pp. Shiu, W.Y.; Maijanen, A,; Ng, A. L. Y.; Mackay, D. Environ. Toxicol. Chem. 1988, 7, 125-137. Miller, B. J. University of Waterloo, personal communication, 1992. Yaws, C. Lamar University, Beaumont, TX, personal communication, 1993. Schlotzhauer, S. D.; Littell, R. C. SAS system for elementary statistical analysis; SAS Institute Inc.: Cary, NC, 1987; 416 pp. Engesser, K. H.; Strudel, V.; Christoglou, K.; Fischer, P.; Rast, H. G. FEMS Microbiol. Lett. 1989, 65, 205-210.

Received for review October 17, 1994. Revised manuscript received February 3, 199.5. Accepted April 24, 199S.@

ES940639 1 @Abstractpublished in Advance ACS Abstracts, June 15, 1995.