Limited reversibility of bioconcentration of hydrophobic organic chemicals in phytoplankton.

Aging, reversibility, and desorption rates for the binding of hydrophobic chemicals (HOC) to phytoplankton cells have not been directly measured. Here the effect of bioconcentration time on subsequent desorption of hexachlorobenzene (HCB) and polychlorinated biphenyls (PCBs) was studied for the alga Monoraphidium minutum. Cell suspensions were exposed to HCB and PCBs spanning a range of log Kow values of 5.7 to 8.2, for 0.13 to 14 d. Subsequently, reversibility and desorption rates were assessed by extracting the chemicals from the cells using infinite sink extractions with Tenax beads or Empore disks employed in the cell suspension. Uptake was biphasic with constant relative contributions of fast surface sorption. Desorption was biphasic too and well fitted to a first order two compartment model. Increasing exposure times resulted in increasing slowly desorbing chemical fractions and decreased desorption rates from these fractions. For the most hydrophobic PCBs, slowly desorbing fractions were >80-90%, whereas desorption half-lives from these fractions ranged up to 120 days. The slow desorption rates directly prove that bioconcentration to algae can be rate limited and imply that already after a few hours of exposure, HOCs may become practically unavailable for repartitioning.

Introduction

Binding to phytoplankton is an important process in the cycling of hydrophobic organic chemicals (HOCs) in aquatic ecosystems. Biomass and growth rate of phytoplankton have been found to affect air–water exchange dynamics of HOCs such as PCBs.[1] Scavenging of HOCs by algae determines the uptake of HOCs in aquatic food webs,[2] and thus may affect toxicant concentrations present in top predators, and determines HOC settling fluxes. In natural and laboratory systems, the interactions between HOCs and phytoplankton are highly dynamic, and the condition of constant exposure probably is rare. Algae grow at variable rates, whereas aqueous phase HOC concentrations rapidly drop during algal growth.[1] In aquatic systems also, the periods of uptake in algal cells preceding desorption, may be highly variable. This implies that HOC bioconcentration experiments under dynamic nonequilibrium conditions may yield insights which remain invisible in steady state or constant exposure experiments, and which may be more relevant for natural conditions.

Bioconcentration in algae received considerable interest during the past two decades. The current understanding includes a two-step uptake mechanism,[3−7] traditionally modeled using a two compartment first-order approach. In addition, release mechanisms such as exudate mediated release and phagotrophic transfer have been reported.[8] Exudates may accelerate the release of PCBs because of their high affinity for dissolved carbon.[8,9] Pioneering work by Swackhamer and co-workers suggested that bioconcentration kinetics for PCBs is slow enough to cause growth dilution, leading to nonlinear Log BCF – Log Kow relationships.[10,11] Other authors observed faster bioconcentration kinetics and suggested that nonlinear Log BCF – Log Kow relationships may also relate to binding to dissolved organic carbon (DOC).[9,12,13] It has been suggested that a dual resistance model can be used to link uptake and desorption rates to undisturbed boundary layer (UBL) and lipid membrane transport resistances.[14,15] The diffusion as well as the dual resistance model approach predict that contact time has an effect on bioconcentration but implicitly assumes that bioconcentration is reversible. We hypothesize that if the kinetics is slow, an “aging” effect may occur similar to that observed for sediments and soils.[16] A key parameter in algal bioconcentration dynamics is the desorption rate constant. If this constant is low relative to the growth rate of the algae, then algal cells do not reach thermodynamic equilibrium.[5,11] This underlines the need for accurate measurements of desorption rates from algal cells. To date, however, desorption rates for HOCs like PCBs have been fitted mainly from BCF data obtained from uptake experiments and direct measurements are lacking.

Batch or gas purge methods are the traditional approaches to study the uptake and desorption kinetics of HOCs to phytoplankton.[3,4,9,10,17] These methods, however, have several drawbacks. In most published batch experiments, emphasis is on the uptake phase. Accordingly, the net equilibration rate is the resultant of simultaneous dynamic uptake and desorption processes and assessment of kinetic parameters requires complex modeling. Furthermore, aqueous phase measurements in growing cultures include DOC associated chemicals so that it is difficult to distinguish between physical and exudate-mediated desorption.[4,5,8,9,12,13,18] This may limit the interpretation of kinetic parameters derived from batch experiments. Gas purging of PCBs from cells may provide more direct kinetic data,[3,12] but is complicated for HOCs because of their low volatility.[12] In the past years, the application of artificial partitioning media in sorption and bioconcentration studies has received much interest. Infinite sink depletive methods like XAD, Tenax, or Empore disk solid phase extraction (SPE) have been used for desorption studies with soils and sediments,[19−23] but their potential has not been tested with respect to phytoplankton biosorption studies. SPE methods are not subject to the DOC artifact and allow direct determination of desorption rate constants. The higher extraction efficiencies of these methods may provide a higher resolution with respect to the detection of kinetic domains compared to gas purging, which typically has a low removal rate.[12] If desorption from phytoplankton is slow, as is mainly inferred from uptake data that use BCF as end point, then directly measured desorption rates should be low and match the desorption rates presently modeled from BCF data (e.g.., ref (11)). We are not aware of studies that have experimentally determined desorption rates using infinite sink extractions and that tested the reversibility of bioconcentration of PCBs to phytoplankton.

The aim of the current work was to test the reversibility of HOC bioconcentration in phytoplankton by measuring the effect of uptake time on the characteristics of subsequent desorption from the cells, thus testing the hypothesis of slow desorption from algal cells. In addition, the applicability of novel SPE methods for algae was explored. The performance of Tenax and Empore disk extractions was compared. Reversibility of bioconcentration was tested and parameters provided for selected HOCs, i.e., hexachlorobenzene (HCB) and six PCBs covering a Kow range of average to very high hydrophobicity (Kow = 105.7–108.2), by consecutive adsorption–desorption experiments with varying adsorption times.

Materials and Methods

Algae

Monoraphidium minutum with largest diameter of 7–20 μm was obtained from the Norwegian Institute for Water Research (NIVA) and was preserved in a continuous culture in 25% Z8 mineral medium[24] prepared axenicly in 0.2 μm filtrated Nanopure water at 17 °C under continuous light at 100 μE/m2. Prior to the use of the algae in the experiments dry weight and cell numbers were determined. The algae were used for two experiments: a validation of extraction method and an exposure time experiment. All representative subsamples of cultures were obtained using a Retsch (Dassel, Germany) suspension divider.

Characterization of Algae

Algal density (AD), i.e., cell numbers and sizes were determined using a Coulter Multisizer II (Coulter Electronics, Mijdrecht, The Netherlands) in the size window from 2 to 14.95 μm. TOC and DOC analysis were performed using an OIC (college station, TX, U.S.A.) Model 700 TOC analyzer in total and 0.45 μm filtered (glass fiber, Whattmann GF/C, Maidstone, U.K.) culture subsamples, respectively. Dry weights (DW) were determined by filtration over the same filters followed by drying at 105 °C until constant weight. State of the cells was inspected using light microscopy.

Chemicals

All water used in the experiments was Barnstead Nanopure water (Sybron-Barnstead, Dubuque, IA, U.S.A.). Hexachlorobenzene (HCB, >98%) was obtained from BDH Chemicals (The Netherlands). 2,4,4′- Trichlorobifenyl (PCB-28), 2,2′,5,5′- tetrachlorobifenyl (PCB-52), 2,2′,3,4,4′,5′- hexachlorobifenyl (PCB-138), 2,2′,4,4′,5,5′- hexachlorobifenyl (PCB-153), 2,2′,3,4,4′,5,5′- heptachlorobifenyl (PCB-180) and 2,2′,3,3′,4,4′,5,5′,6,6′- decachlorobifenyl (PCB-209) were obtained from Dr. S. Ehrenstorfer (Augsburg, Germany). Picograde or p.a. quality organic solvents were obtained from Promochem, The Netherlands (acetone, diethyl ether) and Mallinckrodt, The Netherlands (2,2,4-trimethylpentane). Other chemicals and materials used were: aluminum oxide-Super I (ICN Biomedicals, Eschwege, Germany), silica gel 60 (63–200 mesh; Merck), Cu-powder (99.7%, Merck), Tenax beads (60–80 mesh; Chrompack, Middelburg, The Netherlands), and Empore disks (47 mm, J.T. Baker, The Netherlands). Prior to use, silica gel was activated overnight at 180 °C, aluminumoxide was deactivated with 10% (w/w) Nanopure water, Cu-powder was Soxhlet-extracted with hexane for 4 h and Tenax beads and Empore disks were Soxhlet-extracted with acetone and hexane (2 h each) and dried overnight at 75 °C.

Validation of Extraction Methods with Tenax Beads and Empore Disks

Subsamples of the Monoraphidium minutum continuous culture were transferred to two glass stoppered 1 L flasks and diluted with 10% Z8 until a density of 5 mg/L was obtained. Each flask was spiked with 20 μL acetone spike solution, containing 65 mg/L of the test compounds HCB, PCB28, PCB52, PCB138, PCB153, PCB180, and PCB209. The volume of acetone (spiking solvent) was always less than 2 × 10–3 % (v/v) to avoid solvent effects. After 5 days of uptake (8:16 h light:dark cycle, Philips TLD 18 W lamp, 17 °C), each flask was split into four subsamples of 250 mL. To ensure that algal culture subsamples were identical, all subsample preparations used a Retsch (Haan, Germany) suspension divider. Two of the resulting eight subsamples were used for characterization of algae, three were used for a triplicate desorption experiment with Tenax, and three were used for an identical triplicate desorption experiment with Empore disks. These desorption experiments were performed in 250 mL glass-stoppered flasks. Tenax or Empore disks were removed at 1.5, 3, 5.5, 8, 11.5, 25, 50, and 75 h desorption time after which new were added. Agitation was with magnetic stirring bars (Tenax) or with a Gerhardt LS30 shaker table (Empore disks).

Exposure Time Experiment

Monoraphidium minutum was used for consecutive uptake–desorption experiments with HCB, PCB28, PCB52, PCB138, PCB153, PCB180, and PCB209. At start, 15 identical cultures were prepared with a density of 5 mg/L using a Retsch sample divider. Three of these were spiked immediately. Nine of these remaining 12 systems were spiked at later times so that total uptake times were 336, 120, 24, and 3 h (each n = 3) before the simultaneous start of desorption for all 12 spiked system. The three remaining cultures were treated like the others but received no acetone spike. Samples from these cultures were used for analytical blanks and to check whether the acetone spikes caused effects on growth. After this uptake phase, test compounds were sequentially extracted from the cells with Empore disks as described below. Disk replacement times were 1.5, 3, 5.5, 8, 11.5, 25, 50, and 100 h. Further conditions were as follows: Gerhardt LS30 shaker table for agitation, 8:16 h light/dark cycle (Philips TLD 18 W lamp) and a temperature of 17 °C. In addition, at start of desorption and after the extraction, cells were sampled for PCB analysis, by filtering (Whattmann GF/C, Maidstone, U.K.) 75 mL aliquots of the cultures.

General Extraction Procedure with Tenax Beads and Empore Disks

For the Tenax extractions, 25 mg was used per 250 mL subsample. This means that at algal densities of 5–20 mg/L and assuming 50% algal carbon, the Tenax to algal carbon weight ratio was >10, which is assumed to be a safe excess to maintain the maximum PCB concentration gradient between algae and water.[21] Tenax was removed by transferring the 250 mL culture into a separatory funnel. Tenax particles floated and adhered to the glass walls of the funnel. The algae therefore could be removed from the funnel through the tap back into the original 250 mL flask after which new clean Tenax was added. Tenax beads were flushed from the wall of the funnel with 10 mL hexane and subsequenly vortexed in a test tube. After removal of the hexane with a Pasteur pipet, the remaining Tenax was washed twice with 2 mL hexane.

Empore disk extractions used one disk per 250 mL subsample. They were weighted with a stainless steel clip to prevent floating at the water surface. Disks were transferred to and from the flasks using metal tweezers. Tenax beads and disks were cleaned prior to use, by repeated sonication in acetone. All extractions were performed in airtight bottles, originally designed for determination of dissolved oxygen. To prevent leakage these bottles have their own uniquely polished glass stopper. Mass conservation of chemicals in these batch systems was demonstrated in earlier work.[25]

Analytical Procedures and Quality Assurance

Analytical procedures and quality assurance were as reported before[8,13] and are provided in detail as Supporting Information (SI). In short, Tenax beads and Empore disks were hexane extracted and 4 h Soxhlet extracted (hexane–acetone), respectively. Algal samples were Soxhlet extracted for 16 h. Extracts were concentrated using Kuderna-Danish equipment, cleaned (algal samples) and analyzed using GC-ECD. Results were corrected for replicated blanks and recoveries. Congeners with contaminated blanks (>10% of signal) were removed from the data set. This was the case for PCB28 and PCB52 in the Empore disk samples from the exposure time experiment. Clean-up recoveries were between 81.4 ± 3.3 (HCB) and 92.3 ± 3.6 (PCB209) %. Mass balances were calculated as total amounts extracted with Empore disks plus Soxhlet, divided by spiked mass, and ranged 72–91, 69–91, 65–77, and 66–81% for the 0.13, 1, 5, and 14 day exposure experiments, respectively.

Data Analysis

Uptake Stage

From the HCB and PCB data relating to uptake, apparent bioconcentration factors (BCF, L/kg) were calculated as sorbed concentration divided by aqueous concentration, the latter from difference between added and sorbed PCB amounts, for each of the time points. From these BCF values, apparent kinetic parameters were obtained assuming that bioconcentration kinetics has two components: an instantaneously fast cell surface component (BCFs) and a slow component (BCFi) relating to uptake by the interior of the cells. Surface sorption has been shown to occur in minutes,[6,7,15] which therefore can be assumed to be at equilibrium before our first BCF determination after 3 h. The overall kinetic BCF is defined as follows:Bioconcentration in the cell interior (Ci/Cw) is modeled as a first order process:in which t is time, Ci is the concentration in the interior of the algae (μg/kg), CW (μg/L) is aqueous phase concentration, ku (d–1) is the uptake rate constant, kl (d–1) is the apparent loss rate constant which can be split into an elimination rate constant kd and a growth rate constant kG (kl = kd + kG) and DW is algal density (kg/L). Estimates of kG were obtained from the algal dry weight measurements. The ratio ku/DW is the pseudo first order rate constant ku* (L kg–1 d1). For a spiked, closed system with zero concentration in the algae at t = 0, eq 2 has an analytical solution[26] (math provided as SI), which can be substituted for BCFi in eq 1:Parameters BCFs, and ku in eq 3 were fitted to the experimental BCF data using the solver function in Microsoft Excel. The parameter kl was constrained as ku/BCFi( and the fraction BCFs/BCFkin is referred to as the fraction of surface sorption (x1).[5−7] Confidence intervals (CI90) in the parameters were obtained by refitting the parameters to meet the criterion:[27]where SS90 is the residual sum of squares at the 90% confidence contour, SSmin is the sum of squares at the optimum parameter set, n is the number of data points used in the optimization, equal to the number of triplicate data points in time (i.e., 12), p is the number of fitted parameters and F (p, n – p,90%) is the F distribution.

Desorption Stage

Like the uptake stage, desorption was biphasic so that the Empore disk concentration data were analyzed using a two-compartment model.[3−5] Estimates of desorption rate constants and slow desorbing HOC fractions (xs) bound to the algae were obtained by fitting desorption curves per HOC to the equation:in which Q (μg) is the quantity of chemical extracted from the cells by Tenax or Empore disks as a function of time (t, d). It is assumed that the extraction is rate limited by the algae, at least for the slow desorbing chemicals, and can be described by two apparent first order desorption rate constants; one for the “fast extracted fraction” (k1, d–1) and one for the “slow desorbing fraction” (k2, d–1) with size xs and that backward sorption is negligible. Q0 is the quantity of the test compound at time zero. Q and t are measured (n = 24 per chemical) so that k values and x could be estimated accurately.

Results and Discussion

Methods Evaluation

Uptake Capacity of Tenax Beads versus Empore Disks

To test the uptake efficiency of Tenax beads as compared to Empore disks, desorption rates were compared after 5 days of incubation. At start, all systems were identical but subsequent growth was less in the system with Tenax beads (SI Table S1). In the Tenax system a 50% reduction in DW is observed, while the Empore disk systems show a 4-fold increase of DW (SI Table S1). The decrease in DW may be explained from decay of algae due to unfavorable conditions in the presence of Tenax, like for instance photoinhibition. The desorption curves fitted well to eq 5 (to be discussed in next sections) so that accurate xs, fast- (k1) and slow (k2) release rate coefficients could be calculated. The errors in the parameters (error bars in SI Figure S1) are about two times higher for the triplicate Tenax systems than for the triplicate Empore disk systems. The rate constants themselves were not very different for Tenax versus Empore disk systems; a factor 0.3 to 2 (k1) or at most a factor 0.3 to 3 (k2) higher for Tenax systems. The slightly faster release from the cells by Tenax for higher Kow PCBs is consistently counteracted by lower fast desorbing fractions. The differences between the Tenax and the Empore disk data may relate to extraction efficiencies or to differences in cell properties. The Tenax beads provide a much larger surface area than the single Empore disk and thus will bind the chemicals more efficiently from the water phase. Prior blank tests[28] yielded a first order uptake rate constant of 18 d–1 for the depletion of phenantrene from the same 250 mL water volume (no algae) due to uptake in a free floating 47 mm Empore disk. In identical but stirred systems an uptake rate constant of 46 d–1 was measured. The increased uptake upon stirring can be explained from the reduced thickness of the UBL. For Tenax beads at a concentration of 2 g/L, first order extraction rate constants for aqueous PCB and PAH of 360–504 d–1 have been measured.[21] These rate constants for disks and Tenax are significantly higher than the desorption rate constants measured for the algae (SI Figure S1) or those reported in the literature.[1,6,13] This implies that (a) desorption from the algae is the rate limiting step in the uptake to both solid phase media, and (b) that the fitted constants indeed are direct estimates of desorption rate constants. Furthermore, if desorption from the cells would not have been rate limiting, the differences in k values between beads and disk would have been much larger than the factors 0.3–3 measured here. Finally, Tenax extraction would have been faster for all HOCs studied, which is not the case as Tenax extraction is slower for HCB. This implies that the measured differences relate to the conditions of the cells. This is consistent with the DW measurements: in the systems with Tenax, reduced growth may have preserved high specific surface areas which facilitated release fluxes from the cells.[14] In summary, we conclude that for desorption studies with algae, Tenax or Empore disk extraction methods are useful. Empore disks are easier to use, yield more accurate results, are as effective as Tenax beads and seem to interfere less with algal growth

Bioconcentration Kinetics

Reversibility of sorption, i.e., the effect of uptake time on the degree of subsequent desorption within a fixed time frame, was studied by extracting previously spiked PCBs from Monoraphidium minutum cells after incremental time intervals (0.13, 1, 5, and 14 days). This subsection discusses the results from the uptake stage, which in essence are similar to those in previous reports.[6,7,10,11,19] Because of the different uptake times, the treated systems might show differences in dry weight or cell number after 14 d of uptake (which is desorption time t = 0), because of PCB or acetone toxicity. However, after 14 d the 15 systems still had identical cell densities that did not differ significantly (p > 0.05; t tests) and averaged 11.23 ± 1.0 mg/L dry weight, including the 336 h blank without PCBs (Table 1). Previous reports showed similar absence of toxicity due to PCB acetone spikes.[11] The growth data correspond to a growth rate constant (kG ± s.d.) of 0.058 ± 0.0045 d–1 (n = 5) and confirm the good state of the cells.

Table 1

Algal Cell Densities (AD), Dissolved (DOC) and Total Organic Carbon (TOC) and Dry Weight (DW) at Start of Uptake (“0”), Start of Desorption Phase (“des”) and End of Desorption Phase (“end”)a

spiked systems					blank systems
exposure times:	0.13 d	1 d	5 d	14 d	14 d
ADdes	25.7 ± 0.43	30.5 ± 2.60	44.0 ± 2.37	26.9 ± 0.67	39.6 ± 0.95
ADend	25.9 ± 1.50	25.9 ± 1.60	30.0 ± 1.95	27.9 ± 1.59	54.3 ± 21.7
TOCdes	10.56 ± 0.20	9.31 ± 0.11	10.64 ± 0.33	9.45 ± 0.25	3.39 ± 0.09
TOCend	36.17 ± 0.87	27.59 ± 7.12	24.02 ± 1.78	18.06 ± 0.61	7.46 ± 0.49
DOCdes	7.09 ± 0.05	6.68 ± 0.06	7.50 ± 0.10	6.52 ± 0.26	0.78 ± 0.19
DOCend	28.63 ± 3.06	24.06 ± 10.91	15.33 ± 2.48	4.27 ± 0.09	5.13 ± 0.57
DW0	5.00	5.00	5.00	5.00	5.00
DWdes	12.0 ± 1.03	10.7 ± 1.35	12.0 ± 0.06	11.0 ± 0.85	10.5 ± 0.61
DWend	25.50 ± 2.27	24.83 ± 8.78	32.22 ± 2.14	42.50 ± 3.17	14.53 ± 0.36

Units: AD ( × 104) cells/L; TOC, DOC and DW mg/L. Errors relate to s.d. of triplicate systems.

Overall, the apparent BCF values increased in time and with increasing log Kow (Figure 1). It appears that rapid uptake takes place within 3 h (0.13 d), followed by a slower uptake until 1 d, then for most congeners no clear increase is observed between 1 and 5 d and finally gradual uptake proceeds until 14 d. For HCB, the order between 1 and 5 d was reversed, which remains unexplained.

Figure 1

Log BCF (L/kg) versus Log Kow for polychlorinated biphenyls in Monoraphidium minutum, after 0.13, 1, 5, and 14 d of exposure. Error bars relate to s.d. of triplicate experiments.

This overall behavior, including the downward curvature at higher log Kow, agrees to and has been discussed extensively in earlier reports.[10,13,16] At higher log Kow values, true BCF values, i.e., based on freely dissolved aqueous phase concentrations, may be higher because the present BCF values include DOC- or exudates-bound PCBs in the aqueous phase data (so-called DOC effect).[8,13] The DOC-effect, however, cannot explain the downward curvature at the highest log Kow values as was pointed out before.[13,16] Consequently, we explain the downward curvature in Figure 1 from incomplete equilibrium. The DOC concentration due to exudates was 0.78 mg/L in the blank systems (Table 1) of which 0.3 mg/L can be explained from EDTA in the Z8 medium. The DOC levels in the spiked systems were higher and were almost identical for the different treatments; 6.52–7.50 mg/L (Table 1). These higher values are caused by the acetone spike, which can explain up to 9.81 mg C/L if no volatilization of acetone would occur. The almost identical DOC concentrations further imply that differences in BCF observed among the different exposure times have to relate to kinetic processes.

In order to obtain kinetic parameters for the uptake phase the BCF data (Figure 1) were fitted according to eq 3. The model fitted the data well (0.777 < r2 < 0.997); time plots are provided as SI (Figure S2). The relative contribution of surface sorption (x1) showed no trend with chemical hydrophobicity and was 0.155 ± 0.03 (n = 7, ± s.d.) on average (Table 2). This corresponds to the previous observation that initial surface sorption is primarily determined by the area to volume (A/V) ratio of the algae,[5,6] which thus was similar for all chemicals studied. Uptake rate constants (ku*) for exchange with the cell interior ranged from 3.8 × 103 to 70.9 × 103 L kg–1 d–1 for the different chemicals, whereas the overall loss rate constants (kl) did not show a trend and averaged 0.064 ± 0.036 d–1 (Table 2). Because overall loss is composed of true desorption and growth dilution (kl = kd + kG) and the average growth constant was kG = 0.058 d–1, we argue that apparent chemical loss from the cell interior probably was dominated by algal growth. Because this was the same for all chemicals this explains the similarity of the kl values. Loss rates were an order of magnitude lower than those reported for I. galbana,[1,7,15] which can be explained from the higher specific surface area of the latter species.[15]

Table 2

Surface Sorption as a Fraction of Total Bioconcentration (x1), Uptake Rate Constants (ku*), Overall Loss Rate Constants (kl), Surface Bioconcentration Factors (BCFs), Matrix Bioconcentration Factors (BCFi) and Total Kinetic Bioconcentration Factors (BCFkin)a

	x1	ku* ( × 1000 L kg–1 d–1)	kl (d–1)	Log BCFs (L/kg)	Log BCFi (L/kg)	Log BCFkin (L/kg)
HCB	0.13 (0.03–0.36)	3.85 (0.78–36.5)	0.122 (0.025–1.16)	3.66 (3.07–4.11)	4.50 (4.37–4.55)	4.56 (4.44–4.66)
PCB28	0.14 (0.05–0.34)	7.73 (1.6–41.5)	0.046 (0.01–0.24)	4.43 (3.94–4.83)	5.23 (5.11–5.27)	5.29 (5.17–5.38)
PCB52	0.15 (0.07–0.30)	13.6 (5.0–60.0)	0.071 (0.026–0.31)	4.53 (4.19–4.83)	5.29 (5.21–5.33)	5.36 (5.28–5.43)
PCB153	0.16 (0.10–0.27)	17.1 (9.0–33.0)	0.027 (0.014–0.052)	5.10 (4.87–5.31)	5.81 (5.74–5.84)	5.88 (5.82–5.93)
PCB138	0.17 (0.10–0.28)	17.0 (9.0–33.1)	0.028 (0.015–0.055)	5.10 (4.88–5.31)	5.78 (5.72–5.81))	5.86 (5.80–5.91)
PCB180	0.13 (0.12–0.13)	70.9 (65.5–76.9)	0.100 (0.092–0.108)	5.01 (4.98–5. 04)	5.85 (5.85–5.86)	5.91 (5.91–5.92)
PCB209	0.21 (0.11–0.38)	6.63 (2.0–23.2)	0.057 (0.017–0.20)	4.50 (4.21–4.75)	5.07 (4.96–5.12)	5.17 (5.07–5.25)

90% confidence intervals (CI90) for the parameters ku* and BCFs provided between parentheses. Ranges for the composite parameters x1, kl, BCFi deduced from those determined independently for ku* and BCFs. Range for BCFkin based on Gauss propagation of error in BCFi and BCFs.

Desorption after Different Uptake Times

The systems discussed in the previous section were used for subsequent desorption experiments. Interestingly, apparent growth rates were higher during the desorption stage compared to the uptake stage, an increase that was higher for longer incubated systems (Table 1). The only difference is the longer contact with the small PCB in acetone spike. Although Rhee et al.[29] observed a positive effect of PCBs on photosynthesis, we have no conclusive explanation for this observation. AD did not increase as much as DW (Table 1), which is explained from AD being measured using a fixed window of 2–14.95 μm.

The first order two-compartment desorption model (eq 5) adequately fitted the data for all test compounds and incubation times. Biphasic desorption has been reported previously for release of chlorobenzenes from algal suspensions after fixed short-term incubations, using extraction by gas purging.[3−5,13,18] The desorption rate constants for the fast desorbing fraction (k1) are rather similar among test chemicals, show no consistent trend with exposure time or chemical hydrophobicity (Figure 2A) and averaged 3.6 ± 1.4 day–1 (n = 20) across all chemicals and time points. The observed range of 2 to 7 d–1 is similar to the range of 1 to 3 d–1 measured for HCB desorption after a fixed incubation time of 3–4 h from different species.[3] In contrast, the desorption rate constants for the slowly desorbing fraction decreased from 0.017–0.15 d–1 after 0.13 d to 0.006–0.19 d–1 after 14 d and also decrease over a factor of 30 with increasing Log Kow (Figure 2B). The desorption rate constants observed in the method evaluation experiment even were a bit lower (SI Figure S1). The size of this slow desorbing fraction also appears to depend on prior accumulation time and hydrophobicity (Figure 2C). After 0.13 d, the slow fractions for HCB and PCB209 already are 0.35 and 0.85 respectively, with intermediate values for the other PCBs. In two week’s time, these slow desorbing fractions further increases from 0.35 to 0.7 (HCB) and from 0.85 to 0.94 (PCB209). It should be noted that the fast desorbing fraction (1 – xs) may include some dissolved and fast desorbing chemical from the algal cell surface,[30] especially for HCB. This implies that especially for this compound, the fast desorbing fraction may have been overestimated and thus that the actual slowly desorbing fraction in the cells even may have been larger.

Figure 2

Sizes of the slowly desorbing fraction (xs), the desorption rate constants for the fast (k1) and slow (k2) compartment as a function of log Kow and time. Error bars relate to s.d. of triplicate experiments.

The observed increase of slow desorbing fractions together with a decrease in the rate constants cause overall extra slow release after prolonged incubation. Especially for superhydrophobic PCBs, such as PCB209, the slow fraction and directly measured slow desorption rate constant were 0.85 and 0.017 d–1 already after 0.13 d. This agrees to a desorption half-life of 40 days for the major fraction of bound PCB209, which implies that this chemical is hardly available for subsequent partitioning, already after a very short uptake time of 3 h. After 14 days, this also is the case for the other PCBs, with observed desorption half-lives ranging up to 120 days for PCB209. The slow kinetics is consistent with apparent BCF values increasing over time as was shown in the literature[7,10,11,13] and in Figure 1 with apparent cell interior loss rate constants (kl) values (eq 2) having the same order of magnitude as the k2 values from the desorption experiments (eq 5).

As for the mechanism of desorption, relevant inferences can be obtained from the Log Kow dependences. The slow desorption rate constant k2 showed a significant decrease with increasing Log Kow of more than an order of magnitude (Figure 2B). This is consistent with the UBL limiting diffusion because the solubility of the very hydrophobic chemicals in water is extremely low relative to that in the cell membranes.[14,15] As a result, the UBL has limited capacity to transport the chemicals by diffusion. Hence, with increasing Log Kow, the rate of desorption decreases as the solubility of the chemical in the water drops relative to that in the membranes.

This interpretation of desorption parameters is based on the model parameters obtained from consecutively extracted amounts. Alternatively, reversibility can be judged from the total extracted fractions, i.e., the sum of the quantities of test compound extracted with Empore disks in 100 h divided by the spiked quantity (Figure 3), or by the sum of extracted quantities and quantities remaining in the cells (SI Figure S3). Here we refer to this fraction as the “labile” fraction. This labile fraction and the modeled (1 – xs) fast desorbing fraction can roughly be compared because conceptually, model estimates for the amount extracted after 100 h (using the parameters in Figure 2) might approach the extracted quantity in the Empore disks in 100 h of desorption time if exactly the fast desorbing compartment is emptied within that period of time. With fast desorption rate constants of 1 to 7 d–1 (Table 2, Figure 2A) corresponding to desorption half-lives of 0.1 to 0.7 days, this condition probably is met, although some release from the slow desorbing compartment may have contributed to the apparent labile fraction, especially for HCB. All these measurements show that the shorter the contact time and the less hydrophobic the compound is, the faster and higher is the recovery of total chemical mass from the cells (Figure 3, SI Figure S3). For HCB, labile percentages decrease from 75% after 0.13 d uptake to 62% after 14 d uptake; for PCB209 these percentages are 25 and 7.9, respectively (Figure 3). That the percentages decrease with time and with increasing hydrophobicity (Figure 3, SI Figure S3) is explained from ongoing diffusion into the algal matrix over time, whereas the more hydrophobic chemicals experience a higher resistance against subsequent desorption. Remarkably, none of the compounds approaches 100% recovery in 100 h desorption time. In other words, also from these data, it appears that already after a few hours of exposure, a part of the PCBs become practically unavailable for partitioning. The operationally defined labile fractions after 100 h indeed are a bit higher than the fast desorbing fractions obtained from kinetic modeling. The HCB labile percentages decrease from 80 to 69% (0.13–14 d, Figure 3), whereas the percentages (1 – xs) × 100% decrease from 65 to 32% over the same time period (Figure 2C). For PB209, labile percentages decrease from 27 to 8% (0.13–14 d, Figure 3), whereas the percentages (1 – xs) × 100% decrease from 16 to 6% (Figure 2C).

Figure 3

Percentages of mass removed from the cells (labile fractions) for HCB, PCB 153, PCB 138, PCB 180, and PCB 209 after uptake times of 0.13, 1, 5, and 14 d. Error bars relate to s.d. of triplicate experiments.

Combining all information, the uptake data (Figure 1) clearly show prolonged uptake by the cells until 14 d exposure. Hence, the total amount of chemicals in the cells is increasing, while the fast desorbing fraction (1 – xs) is decreasing with exposure time. From the bioconcentration, the Empore disk desorption curves as well as from the modeling, we conclude that a slow “aging” effect occurs by which the chemicals are sequestered slowly in the sorbing matrix.

General Discussion

Aging is a well-researched feature of HOC sorption to sediments[17] and suspended solids but on the basis of the present results, seems to apply to planktonic particles as well. After short contact times, the chemicals probably reside at the cell surface or in the outer spheres of the cells so that the major portion of the chemical also is extracted quickly (SI Figure S4). Many researchers report rapid uptake of HOCs to reach a steady state after several hours[9,31] and previous desorption studies reported reversibility of biosorption using short incubation times.[3] With the present long uptake times, hydrophobic PCBs are spread throughout the cells and apparently increasingly reach internal domains from which subsequent desorption is very slow. In their uptake studies, Swackhamer and co-workers[10,11] observed tri- and penta chlorinated PCB homologues to reach steady state after a week, while decachlorobiphenyl did not even reach steady state in 40 days. They did not directly measure the slow desorption rate constants for the four species studied but they calibrated uptake rate and surface adsorption constants from which their equivalent desorption rate parameter (k in ref (11)) can be calculated. The resulting range of ∼0.001–0.2 d–1 agrees very well to the present range for k2 of 0.006 to 0.19 d–1 measured using infinite sink Empore disk extractions. Our previous study on long-term desorption of PCBs from phytoplankton[13] used a gas purging approach with a much lower and nondepletive extraction rate than the present one and therefore probably could not detect the biphasic desorption behavior.

Implications

Previous studies showed that desorption from outer, surface sites involves times scales of minutes, whereas exchange from internal matrix pools shows half-lives of days.[1,6,15,16] The latter half-lives match the desorption half-lives observed in the present uptake experiments and the fast desorption stage of the desorption experiments, but cannot explain the slow progressive bioconcentration with time scales of weeks and months observed and hinted at in previous studies.[7,10,11] The aging effects and directly determined slow desorption rates observed in the present study, however, can explain these previous observations.

The observation of practically irreversible bioconcentration has important implications for the modeling of HOC bioconcentration in phytoplankton and for the aquatic fate of algal-bound HOCs. First, with such desorption half-lives, equilibrium in natural waters will not be attained. Second, the present data suggest that besides surface and matrix sorption with half-lives of minutes and days respectively, a third kinetic domain may be needed. Third, previous models considered sizes of the kinetic compartments to be independent of exposure time. This feature conflicts with our current desorption data showing that the previously assigned “slow” desorbing fraction of the test compounds is released with a more or less constant rate of 3.6 ± 1.4 day–1 (n = 20) across all chemicals and time points, while the size of this compartment decreases in favor of a compartment from which the chemicals are remobilized at rates that are 2 to 3 orders of magnitude lower. Furthermore, the latter rates are so low, and the size of this very slow compartment is so high that remobilization may not occur during the life span of algae. Infinite sink SPE methods may find further application in assessing slow desorption rates for other chemicals and algal species.