Environ. Sci. Techno!. 1995, 29, 933-940
Dynamics of Olganic MicropoIIutant Biosorption to Cyanobacteiia and Detritus A L B E R T A. K O E L M A N S , * SANDRA F. M. ANZION, AND LAMBERTUS LIJKLEMA Department of Water Quality Management and Aquatic Ecology, Wageningen Agricultural University, P. 0. Box 8080, 6700 DD Wageningen, The Netherlands
Equilibrium and rate parameters of chlorobenzene biosorption to the cyanobacterium Anabaena and various types of algal detritus were measured using gas purge desorption experiments. Organic carbon normalized sorption coefficients (KJ and slopes of log KO, - log KO, plots for algae and detritus increased significantly with increasing C/N atomic ratios of the adsorbents. A statistical analysis showed that the desorption kinetics could be described best with a twocompartment model. The compartments were interpreted as a lipid pool in the surface of the cells and a similar pool deeper in the cell interior. The kinetic parameters measured were such that in most aquatic systems biosorption equilibrium can be assumed.
Introduction In the aquatic environment, the transport and food web accumulation of hydrophobic organic compounds (HOCs) is strongly influenced by sorption to bacteria, phytoplankton, and detritus ( 1 , 2 ) . In earlier reports (reviewedin (ref 3), the uptake and release by phytoplankton are found to be fast, limiting the urge for kinetic data. Consequently, because kinetics are fast, in many transport models the lipid-normalized bioconcentration factor, BCF (or Kip for detritus), or the organic carbon-normalized sorption coefficient, KO,,is considered a constant under most environmental conditions and is thought to be linearly correlated to &,with a slope close to unity. However, this equilibrium partitioning concept is not necessarily valid for phytoplankton and its detrital products for two reasons. First, it is unlikely that BCF, &ip, or remain constant upon aging, death, and mineralization of algal biomass or are identicalfor different algal species. It has been described by severalinvestigators that GC varies for dissolved organic matter, soils, and sediments of different origin (4-9). The differences are usudy attributed to variations in the polarity of the organic matter. Second, phytoplankton grows, which results in a ‘diluting’ effect on the chemical concentration in the cells. This can be described by the following equation, assuming that both biosorption and algal growth are first-orderprocesses and that growth does not change the sorption kinetic constants (1, 2):
dCA/dt= k12C, - kzlCA- kGcA
(1)
where CAhglg) is the concentration in the algae, CW(UglL) the aqueous concentration, k12 (L g-’ day-’) is a pseudofirst-order adsorption rate constant linear in the solid-towater ratio, and k21 (day-l) is a first-order rate constant for desorption. k~ (day-’) is the rate constant for algal growth. This model has the following steady-state solution:
B C F ~= c;/Ge=
k12
- BCF~- k21 k21
+ kc
(2)
in which BCFG is the bioconcentration factor for growing phytoplankton, BCFN is the bioconcentration factor at nongrowth conditions, equal to kdk21, and the superscript e refers to equilibrium concentrations. If k~ cannot be neglected compared to kzl,then k~ affects the bioconcentration factor. In recent publications of Swackhamer and Skoglund (10,111, BCFs for growing phytoplankton were invariant with KO, at KO, > 105.5.Furthermore, higher BCF values were found at lower phytoplankton growth rates. These findings suggest that the biosorption rate for the more hydrophobic HOCs is slow compared to the algal growth rate and, therefore, confirm the model as condensed in eq 1. However, the evidence is indirect since no kinetic experiments allowing the estimation of distinct rate pa-
rameters (k12, k21) were performed. The objective of our study is to quantify chlorobenzene biosorption rate and sorption affinity for the cyanobacterium Anabaena spp. at nongrowth conditions and for * Corresponding author; FAX: +31-(0)8370-84411; e-mail address:
[email protected].
0013-936x/95/0929-0933$09.00/0
a 1995 American
Chemical Society
VOL. 29. NO. 4. 1995 / ENVIRONMENTAL SCIENCE &TECHNOLOGY
933
detritus and to evaluate the relevance of these two factors. No previous studies have examined HOC sorption to cyanobacteria and the effects of mineralization in terms of multiple compartment kinetic models. The results, Le., parameters for rate and affinity, are compared to similar sorption data for Scenedesmus spp., which were reported by Koelmans and co-workers (12). The biosorption dynamics were investigated using gas purge induced desorption experiments. This method is very suitable for the delicate Anabaena cells because phase separation is not required (12, 13). Phase separation is always incomplete (14)and implies a serious risk of mechanical stress and damage to the cells. By incomplete phase separation, the sorption coefficient may become dependent on the solid concentration (14).Gas purging, also known as gas sparging or dynamic headspace partitioning, is a well-established technique used to measure aqueous fugacities,Henry's law constants (H),or characteristics of HOC binding to DOC, algae, or sediments (e.g., refs 12, 13, and 15-18). In a previous study with Scenedesmus (121, a 100%increase of KO, with aging was observed. To be able to explain KO, trends from polarity changes, in the current study with Anabaena, a more extensive adsorbent characterization including lipids and elemental C, H, and N was used.
Materials and Methods Origin and Handling ofAdsorbents. Adsorbents were fresh
Anabaena spp. cells, Anabaena detritus derived from the fresh cells, and Scenedesmus detritus. Fresh Anabaena (originally PCC 7120) was cultured in 40 L of 2-8 medium (12)under continuous light at 20 "C until a stationaryphase was obtained. The medium was prepared in 0.2-pm filtered Nanopure water (Sybron-Barnstead, Dubuque, IA) to minimize the contribution of bacteria to the total biomass. From the resultant algal suspension, a representativesample of 6.4 L was taken. After two more days in continuous light, the light was turned off (mineralization time = 0) to induce mortality and mineralization, and 2 L of filtered (Schleicher and Schuell,Dassel, Germany;Ref. No. 300412) lake water (Lake VolkeraklZoom, The Netherlands) was added as a natural source of bacteria. Because of the dilution of the lake water in the algal suspension, the algaerelated carbon represented more than 95% of total organic carbon in the system. The 6.4-Lsample was divided further in eight representative subsamples. The subsamples were used for a triplicate purge experiment and for characterization of the adsorbent. Characterization and purge experimentswere performed on the same day. Similar 6.4-L samples of mineralizingAnabaena were taken after 18,47, and 124 days and used for the determination of adsorbent and sorption characteristics. During mineralization, the suspension was aerated continuously. The resultant volume reduction of the suspension partially compensated the loss of biomass due to mineralization. For the samples used for sorption measurements (Table 1: samples ANAOa, ANA18, ANA47, and ANA124), a variation in solid concentration of a factor 2 remained. Theoretically, such a variation might innuence the sorption coefficient. However, known mechanisms for an effect of solids concentration on sorption coefficients, such as incomplete phase separation (14) or nonequilibrium ( I n , play no role in our experimentaldesign. To test this, sorption coefficientswere determined for one of the detritus samples at the two extreme adsorbent concentrations. For this purpose, a detritus sample (466-dayaged Scenedesrnus) that was left 934 * ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 29, NO. 4, 1995
TABLE 1
Characteristics of Anabaena and Detritus Samples sample codea
DWb
TOCC
lipidsd
(mg/L)
(mg/L)
(YO)
bSn) h g R )
ANAOa ANAOb ANA18 ANA47 ANA124
82.3 106.6 39.8 41.8 98.2
34.7 50.5 27.9 22.9 30.0
3.7 3.1 4.1 2.6 1.1
1302 1693 103 6 11
91.2 114.8 55.5 61.1 86.3
55.9
SC466
620
72.5
0.6
276
296.7
35.7
sample codea
CHV DWh
CHU, foci PHA' (TOC/DW)
ANAOa ANAOb ANA18 ANA47 ANA124
15.8 15.9 2.6 0.14 0.11
5.4 5.6 1.3 0.13 0.21
0.42 0.47 0.70 0.55 0.31
0.28
4.72 5.07 6.47 6.28 6.87
f0.02
1.69 1.69 1.81 1.77 1.92
0.45
0.53
0.12
0.16
9.50f0.11
1.96
SC466
CHLg
fock 0.43
COD'
Ci" (fsd, n = 2) iO.02
i 0.03 ?c
0.04
f 0.01
Lln
(YO)
HI C'
aCapital letters refer to algal species: ANA = Anabaena, SC = Scenedesmus. Numbers refertotheageofthealgae(days). The letters a and b for ANA0 refer to the culture 2 days prior to mineralization (a) and 0 days prior to mineralization (b). Concentration of suspended solids(filtrationover0.45pn anddrying at 105°C. CTotalorganiccarbon concentration. Total lipid content of particles. e Chlorophyll concentration. 'Chemical oxygen demand. Loss of weight after drying at 550 "C. Chlorophyll to dry weight ratio. Chlorophyll to phaeophytin ratio. i Organic carbon fraction of particles calculated as TOC/DW. Organic carbon fraction measured by elemental analysis of particles. 'Calculated from elemental analysis results.
over from an earlier study (12)was diluted to the desired concentrations with Nanopure water. Characterization of Adsorbents. Determinations of adsorbent characteristics were performed at least in triplicate on representative subsamples of the adsorbent suspensions. Shape and size of algal cells were measured using a Galai CIS-1 particle sizer (Galai Production Ltd, Israel). In addition, an external high-quality microscope (Nikon, Melville, NYI was used. Dry weight and ash free dry weight were measured after filtration over 0.45-pm membrane filters by drying at 105 and 550 "C, respectively, till constant weight. TOC was measured on an OIC (College Station, TX) Model 700 TOC analyzer (100%conversion of algal carbon). COD was determined as oxidizability byK2Crz07 according to standard methods (19). Absorption spectra from 270 to 900 nm of total and filtered samples were measured on a Beckmann (Fullerton, CA) DU-64 spectrophotometer. The absorbance for the particles was obtained by difference. Chlorophyll and phaeophytins were extracted with acetone and measured on a Beckmann DU64 spectrophotometer according to standard methods (19). Elemental analysis (C, H, N) and total lipids were measured on adsorbent particles isolated by filtration over glass fiber filters (Whattmann, Kebo, Sweden; GFIC). Elemental content was measured with a Carlo Erba (Breda, The Netherlands) 1106 elemental analyzer. Some samples contained small glass fibers from the filter so weight percentages could not be calculated. However, since the blank filters showed no detectable C, H, or N, the atomic ratios are not influenced by this artifact. Lipids were extracted from the adsorbents with a 10:5:4 (v/v/v) chloroform/methanol/water (1% NaC1) mixture in a blender. This procedure was repeated twice. After centrifugation of the combined extracts and careful drying of the chloroform
TABLE 2
Measured Values and Literature Values of Henry's Law Constant (H in 1O-I atm m3 mol-') measureda H compound
at 21.5 "C
literatureb H at 20 "C at 22.1 "C
1,2,3,4-tetrachlorobenzene
7.2k 0.8 7.7 f 1.0 5.3 f 0.6
7.0 ( 78) 7.2 ( 78) 4.9 ( 78)
pentachlorobenzene hexachlorobenzene a
6.8( 75) 6.8 ( 75) 4.7 ( 75)
With hsd, n = 6). Literature citation in parentheses.
layer, the lipids were redissolved in hexane. Part of this hexane fraction was used for a spectrophotometric lipid determination using a commercial Total Lipids Test combination (Boehringer Mannheim, 1984 catalogue, No. 124303). This method is based on the sulfophosphovanillin complexation reaction. Gas Purge Procedure. Nanograde organic solvents (acetone, diethyl ether, and 2,2,4-trimethylpentane) were obtained from Promochem (C. N. Schmidt, The Netherlands). 1,2,3,4-Tetrachlorobenzene, 1,2,4,5-tetrachlorobenzene, and pentachlorobenzene (98%, Aldrich Europe, Belgium) and hexachlorobenzene (>98%,BDH Chemicals, England) were used without further purification. Tenax 40-60 mesh was obtained from Chrompack (The Netherlands). For the purge experiments,high-purity nitrogen was used. The gas purge method was used for the measurement of sorption characteristics and Henry's law values (H; atm m3 mole-'). For a triplicate determination of sorption characteristics, three adsorbent suspensions were placed in three parallel and thermostated (21 "C) glass stoppered 900-mL bottles. Each bottle was spiked with 50 pL of a solution of 1,2,3,4-tetrachlorobenzene (TeCB),pentachlorobenzene (QCB), and hexachlorobenzene (HCB) in acetone. The total amounts present in the system were 0.95 pg of TeCB, 1.25 pg of QCB, and 1.25 pg of HCB. Equilibration times were 24 h under continuous stirring (metal stir bar). Earlier reports (12,20) showed that 24 h is more than enough time for chlorobenzenes to reach equilibrium. Out of convenience,Scenedesmus detritus was equilibrated for 48 h. M e r equilibration, chlorobenzenes were purged onto 11 Tenax traps (trap efficiency 100%)in 5 days at a flow of 500 mL/min. In the water column, the gas bubbles followed a circular path of 60-80 cm, after which dissolved and gas-phase chlorobenzenes were at equilibrium. This was confirmed by prior experiments using various flow rates and glass fritts (Koelmans, unpublished results) by the column height required for chlorobenzenes of 40 cm, as reported by ten Hulscher et al. (15) and by the agreement of our Henry's law constants (Zflwith literature data (Table 2). The gas flow (constant within 1%) was regulated with flow controllers and was passed through vessels containing water to prevent volume reduction in the algal suspensions. Tenax traps were replaced at incremental intervals. All experiments were optimized for HCB, Le., to yield approximately the same mass of chemical in each trap. Because TeCB and QCB have different desorption kinetics than HCB, trapped amounts for these congenerswere more variable. The traps consisted of glass tubes containing glass wool and approximately0.2 g of Tenax. Test compounds were desorbed from the traps by elution with 15 mL of diethyl ether (recovery 100%). Tenax traps and extracts contained no
algal or detritus particles and were not colored so any removal of particles by the scavenging of gas bubbles as reported by Friesen et al. (21),was not significant. The extracts were concentrated slowly with nitrogen after the addition of an internal standard (45 pg/L 1,2,4,5-tetrachlorobenzene in 2,2,4-trimethylpentane) to reduce the error of subsequent GC analysis. Chlorobenzene fractions remaining in the adsorbents after purging (typically less than 1%) were recovered by thorough extraction and subsequent cleanup. A Hewlett-Packard (HP; Avondale, PA) 5890 double column capillary gas chromatography system equipped with two 63Nielectron capture detectors (ECD)was used in chlorobenzene quantization. For details on the experimental procedures, we refer to our earlier paper (12). The mass balances of 14 purge experiments were (means standard deviation) as follows: 98.9%ik 1.6 for TeCB, 96.5% f 1.8 for QCB, and 97.7% ik 4.2 for HCB. These percentages refer to the summed amounts of chemical recovered from the traps and from the adsorbent after purge as the fraction of the amounts initially added to the systems. For the determination of Henry's law constants, the chlorobenzeneswere purged from Nanopure water. Equilibration and purge time were 1 and 7 h, respectively. Extraction and GC analysis were as described above. The mass balances of six H determinations were (means f standard deviation) as follows: 99.1% f 2.4 for TeCB, 95.9% & 2.6 for QCB, and 94.2% i 2.5 for HCB. Because the balances are very close to loo%, we conclude that no significant loss of chlorobenzenes occurred. Henry's law constants were calculated as described in the next section and are compared with the literature data in Table 2. The data in Table 2 and the mass balances confirm the validity of the purge experiments. Data Analysis. Without an adsorbent, the gas purge induced removal of the test compounds from the aqueous phase is described as an irreversible first-ordervolatilization process according to
*
d&/dt = -k,&
A
k, =- FH
Rn/,
(3)
in which t is time (day), CW(ug/L) is the solution-phase chlorobenzeneconcentrationreferenced to aqueous-phase volume V, (L), k, is a fiist-order volatilization rate constant (day-'), F is the gas flow (L day-'), R is the gas constant (atm m3 mol-' K-l), Tis the absolute temperature (K), and His Henry's law constant (atm m3mol-'). When headspace can be neglected and the volatilizing compound is trapped on Tenax columns, CWcan be calculated from the cumulative trapped amounts, QTENAX, using
Q,o
&=
- QTENAX
vw
(4)
in which Qr-o (ug) is the amount of test compound initially present in the system. Hvalues were calculated from the trapped amounts using eqs 3 and 4. In case of an adsorbent suspension at sorption equilibrium, the volatilization process induces desorption. If the rate of gas purging is sufficiently slow so that equilibrium between adsorbent and water is maintained, then the 'equilibrium binding' situation as describedby Hassett and Millicic (16) can be applied VOL. 29, NO. 4, 1995 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
935
dCT/dt= - 1 + (TOC)Koc k c ,
(5)
in which CT @glL) is the sum of aqueous and sorbed chlorobenzene concentrations and TOC (g/L) is the total organic carbon concentration. The equilibriumbiosorption constant is symbolized by KO,(Llg)instead of BCF, because the latter term is reserved for living species. Furthermore, normalizationto carbon eliminatesthe influence ofpossible inorganic precipitation or dissolutionprocesses. The ability of this one-compartment model to describe our data was compared to that of a model in which biosorption takes place in two kinetically defined compartments (biphasic desorption). Sorption to the first compartment (All is considered instantaneous:
in which x1 is the fraction of sorbent for which sorption is instantaneous, CAI@g/gof organic carbon) is the sorbedphase concentration in the first compartment, referenced to the total organic carbon mass (first and second compartment). For the second compartment, two configurations can be considered: a series version (CW CAIt CM) and a parallel version (CAI C WCUI. ~ In the series version, the first compartment can be pictured as the exterior of the adsorbent. In the parallel version, the compartments resemble kinetically different sorption sites. In the series version, mass transfer to the second compartment with size (1- x l ) , is characterized by forward and reverse first-order rate constants kl and k2 (day-'):
-
dC,fdt = klCAl - kzC,
(7)
in which CM (uglg of organic carbon) is the sorbed-phase concentration in the second compartment, referenced to total organic carbon mass. At equilibrium, a uniform sorption affinity as quantified by KO,can be assumed (17). This implies a uniform distribution of the chemical in the organic matter compartments, so that
BCF~BCF~ 1.0
0.8
0.6 0.4 0.2
o !
-4
k,*& - k2C,
A
kl*= k2(I - xl)K0, (9)
in which the constant kl* has the unit of a pseudo-firstorder rate constant (Lg-' day-'). Equation 7 for the series version can be converted to eq 9 for the parallel configuration, by elimination of CAIin eq 7 using eq 6, and subsequent elimination of kl using the right hand side of eq 8. This conversion shows that although the rate expressions (eqs 7 and 9) seem different at first sight, they are mathematically equivalent. Therefore,our data do not allow discrimination between the two configurations. Several authors successfully used this model to describe sorption to sediments, soils, or algae (e.g., refs 12 and 17). The equations for the compartment models allow the prediction of the effect of algal growth on the steady-state BCF. When the equilibrium binding assumption is valid for the entire adsorbent (XI I), desorption has to be fast compared to the rate of volatilization from the purge flasks. Since the volatilization rate constants k, (eq 3) were
-
936 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 29, NO. 4,1995
,= 0.0 I
1
I
-3
-2
-1
0 Log a
FIGURE 1. Underestimation of BCF at steady state quantified as the ratio BCFG/BCFNas a function of the ratio of desorption rate and growth constants (log k d k ~ at ) various sizes of the fast sorption reservoir XI.
approximately 30 day-' for TeCB and QCB and 20 day-' for HCB, the respective desorption rate constants (k21 for the one-compartment model) have to be larger than these values. For sorption rate constants as large as these, the effectsof algal growth on the steady-state BCFGas predicted by eq 2, will be small. In the case of desorption from two compartments (eqs 6-91, the overall desorption kinetics are slower because the equilibrium binding assumption holds only for afraction (xl) of the initially sorbed chemical. In this case, the sizes of the sorption reservoirs should be considered. Combining eq 7 and the right-hand side of eq 8, and assuming first-order algal growth, access to and from the second compartment is given by
Combination of the steady-state solution of eq 10 with eq 6, yields
a
In the parallel version, the rate expression for the second compartment reads
dC,ldt=
x
+ x1
BCF~-a + l (11) in which a = k z l k and KO, is the general equilibrium sorption constant for live or dead biomass. For living phytoplankton, the use of BCFN instead of KO, is more common, which leads to the replacement of KO, on the right-hand side of eq 11. It follows from eq 11 that the influence of growth depends on the magnitudes of the k2/ k~ ratio (a)and XI. This is illustrated in Figure 1,where the underestimation of BCF at steady state (the ratio BCFG/ BCFN)is plotted as a function ofxl and the log of the k z / h ratio (a).When equilibrium binding is predominant (xl close to l),growth has a negligible effect on BCF. When slow sorption is predominant (XI close to zero), eq 11 reduces to an equation essentiallyidentical to eq 2, because both k21 and k2 quantify desorption rate. Figure 1 shows that in the latter case the relative magnitude of k2 as compared to kG is important, but only when log k ~ l
