Environ. Sci. Technol. 2005, 39, 489-497
Assessment of the Geochemical Role of Colloids and Their Impact on Contaminant Toxicity in Freshwaters: An Example from the Lambro-Po System (Italy) D A V I D E A . L . V I G N A T I , * ,† T A M A R A D W O R A K , † B E N O ˆI T F E R R A R I , † B R A H I M K O U K A L , † J E A N - L U C L O I Z E A U , †,‡ M A R I O N MINOUFLET,† MARINA I. CAMUSSO,§ STEFANO POLESELLO,§ AND JANUSZ D O M I N I K †,‡ Institut F.-A. Forel and Centre d’Etudes en Sciences Naturelles et de l’Environnement, University of Geneva, 10 Route de Suisse, CH-1290 Versoix, Switzerland, and Water Research Institute (IRSA-CNR), Via della Mornera 25, 20047 Brugherio (MI), Italy
The role of colloids in regulating element transport, behavior, and bioavailability in aquatic systems is now wellestablished. It appears that further progress in this research field is being slowed by (i) a limited integration between the geochemical and the biological aspects of the research on colloids and (ii) a persistent gap between wellcontrolled laboratory studies and real field situations. This paper presents a simultaneous evaluation of the role of colloids in controlling element environmental fate and bioavailability at the confluence between a major river and a polluted tributary. Fractionation of trace elements among suspended particulate matter, colloids, and true solution suggests that colloids may play a role in the removal of trace elements from the water column to bed sediments during the mixing of the two rivers. Toxicity testing of water samples indicates that, in this specific system, contaminants associated with colloids can contribute to water toxicity for the rotifer Brachionus calyciflorus but not for the green alga Pseudokirchneriella subcapitata. To the best of our knowledge, the results for B. calyciflorus are the first ones pointing to the possible contribution of colloid-bound contaminants to water toxicity in environmental samples. Despite the uncertainties associated with field variability, the results of chemical analysis and toxicity testing show several points of convergence. Following these observations, a few innovative research approaches are suggested to improve the understanding of trace element biogeochemistry in real field situations.
Introduction The importance of colloids, defined as solids sized between 1 nm and 1 µm (1), in the geochemical cycles of trace elements * Corresponding author telephone: +41 022 950 97 24; fax: +41 022 755 13 82; e-mail:
[email protected]. † Institut F.-A. Forel, University of Geneva. ‡ Centre d’Etudes en Sciences Naturelles et de l’Environnement, University of Geneva. § IRSA-CNR. 10.1021/es049322j CCC: $30.25 Published on Web 12/07/2004
2005 American Chemical Society
is now well-established (2-6), and constantly increasing efforts are directed toward their study in aquatic environments (7-14). Colloidal matter also constitutes a potential pathway of contaminants to filter feeders (15, 16) and can regulate, often in contrasting ways, contaminant bioavailability to unicellular organisms (17, 18). However, to date, little effort has been directed toward the combined study of the role of colloids in element cycles and bioavailability. Furthermore, integration between laboratory experiments carried out under well-controlled conditions (e.g., refs 3 and 19-23) and field studies (e.g., ref 24) is still rather limited. In particular, a stronger feedback from field studies to laboratory experiments is needed to increase our understanding of the role of colloids in regulating element behavior and bioavailability in real field situations. In this paper, we present a simultaneous evaluation of how natural colloids can affect the environmental fate and bioavailability of trace elements in a freshwater system subject to multiple anthropogenic stresses. The possible overall role of colloids in natural systems is highlighted, and ways to improve the environmental relevance of laboratory-based studies, as evinced from the experimental observations, are proposed.
Experimental Section Study Area. The Lambro-Po system (Figure 1) indicates the middle course of the river Po (the largest Italian river; watershed area over 70 000 km2) at the confluence between the Po itself and its most polluted tributary, the river Lambro. The latter (with a watershed of about 3000 km2) contributes about 50% of the total Cd load generated in the entire river Po basin; 40% of Pb; 20% of Cu and Zn; and 10% of As, Cr, and Ni (25). Biological effects of anthropogenic contamination in the Lambro-Po system include chronic (and occasionally acute) toxicity, as determined in laboratory tests, and reduced biodiversity in the field (26, 27). Sampling. Sampling sites (Figure 1) were selected following previous extensive monitoring studies (26). Sample collection was carried out in July 2001 under medium-low flow conditions (Figure S1 in Supporting Information). Raw water samples (about 40 L) were filtered directly in the field through a 1.2 µm polypropylene cartridge filter (Calyx, MSI) and immediately transported to the laboratory for subsampling and ultrafiltration. Ultraclean procedures were followed during sample collection, handling, processing, and analysis as detailed in ref 13. Possible contamination from the pumping system was found to be negligible for all elements except Zn (0.5-0.6 µg L-1) whose concentrations were corrected accordingly. Suspended particulate matter (SPM) was collected for organic carbon and element analysis by continuous-flow centrifugation (CFC) using two Westphalia separators (type KA2-06-175) operated in parallel under the optimal conditions set in previous studies (28, 29). Colloid Separation. Colloids were isolated by tangential flow filtration (TFF) using two regenerated cellulose membranes (Millipore, Prepscale, surface area 0.54 m2) operated in parallel and having 10 and 1 kDa nominal molecular weight cutoffs. Methodological details about cleaning, preconditioning, and operation of the TFF membranes can be found elsewhere (13, 14). TFF (concentration mode) was started within 2 h and completed within 10 h after sampling, with concentration factors (CF, see Terminology) for colloidal matter ranging from 3.8 to 6.4. Small CFs may cause overestimation of the colloidal fractions of OC and trace elements (30, 31) because of the partial retention of truly VOL. 39, NO. 2, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
489
FIGURE 1. Schematic representation of the Lambro-Po system, Northern Italy. Sampling points: site 1, Le Gabbiane (Po River); site 2, Ca’ il Masero (Po River); site 3, Livraga (Lambro River). Arrows indicate the direction of river flow. dissolved species, which have a permeation coefficient (Pc) always smaller than 1 (30). Assuming Pc values of 0.45 for truly dissolved OC and between 0.56 and 0.94 for truly dissolved trace elements (as estimated by refs 30 and 32 using the permeation model), 82% of truly dissolved OC and (for the lowest Pc value) 91% of truly dissolved elements would have passed through the TFF membrane after three ultrafiltration cycles. These calculations yield a worst-case overestimation of 20% for colloidal OC; comparable with the acceptability range of ultrafiltration mass balances (see Element and Organic Carbon Distribution). Chemical Analysis. Total filterable, HMW, LMW and truly dissolved element concentrations (see Terminology) were assayed by FAAS (Ca, K, Mg, Na, and Si) or ICP-MS (Al, Ti, V, Cr, Mn, Fe, Co, Ni, Cu, Zn, As, Cd, Sb, Pb, and U). Hg was assayed by an automated Hg analyzer (AMA254, FKV, Bergamo, Italy). Instrumental accuracy, as determined by the analysis of certified reference material (CRM) 1643d from NIST, was within 5% of certified values for all elements except Fe, whose measurements by ICP-MS are affected by interferences (33). Fe results were therefore used solely to quantify the relative importance of colloidal and truly dissolved fractions without any reference to absolute concentrations. Total filterable, HMW, LMW, and truly dissolved organic carbon were measured using a Shimadzu TOC 5000A analyzer. Particulate major elements (Al, Ti, Ca, Fe, K, Mg, Na, and Si) were determined by X-ray fluorescence (XRF) using a Philips 2400 apparatus. Relative standard deviations (2 s) for the corresponding metal oxides were 1% for Al and K; 2% for Ca, Fe, and Mg; 3% for Na; 0.7% for Si; and 0.02% for Ti. Other elements (except Hg) were determined by ICP-MS after microwave-assisted digestion. Aliquots of 0.2 g SPM (wet weight) were dissolved with 5 mL of Suprapur concentrated HNO3 (VWR). Water content of SPM slurries was determined and used to convert element concentrations on a dry weight 490
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 2, 2005
basis. Digestions were performed in triplicate, and procedural blanks were included to check for ambient or reagent contamination. Total Hg was determined directly on dried solid samples using the AMA254 analyzer (34). Recovery efficiency of the microwave digestions and Hg determinations was checked using CRM STSD-2 (stream sediment) from the Geological Survey of Canada (35). Percentage recoveries agreed with certified values within 1 SD. Particulate organic carbon was determined by a VG Isochrom elemental analyzer (University of Bourgogne, Dijon, France) after pretreatment with H3PO4 for removal of carbonates. Analytical reproducibility of particulate organic C measurement was within 2%. Estimation of Colloidal Mass. The mass of CSS was estimated from the colloidal concentrations of OC, Ca, K, Mg, Na, and Si (Table S1 in Supporting Information) under the simplifying assumption that they occur as CH2O, CaO, K2O, MgO, Na2O, and SiO2. The contribution of colloidal Al, Fe, Mn, and Ti to CSS can be neglected since colloidal concentrations of these elements are 3 orders of magnitude lower than those of OC and major elements. The colloidal Ca in the LMW fraction at sites 1 and 2 (Table S1 in Supporting Information) may be partly bound to organic matter. At these sites, molar ratios of LMW Ca versus LMW OC are approximately 2.6 and 0.8, respectively. Assuming a cation exchange capacity of 300 mequiv/g of organic matter (36), a Ca/C ratio of about 0.3 would be expected. It is therefore possible that some precipitation of CaCO3, likely because of Ca rejection by TFF membranes (37), occurred during the ultrafiltration process leading to a slight overestimation of LMW colloidal masses. On the other hand, the presence of HMW Ca at sites 1 and 3 (Table S1 in Supporting Information) is probably caused by the use of a filter with 1.2 µm pore size (see Sampling), which may retain less fine carbonates as compared with the most commonly used 0.45 or 0.22 µm filters.
Toxicity Testing. Toxicity of filtered water samples was assessed using the microplate bioassay for Pseudokirchneriella subcapitata (see ref 38) and the standard AFNOR test for the reproductive success of the rotifer Brachionus calyciflorus (39). Details of the experimental and evaluation procedures for the bioassay results are given as Supporting Information. Individuals of B. calyciflorus were also exposed to the retentates (see Terminology) of the 1 kDa TFF membrane for the three sampling points. Independent aliquots of each retentate were ultracentrifuged (Centrikon T-1080) at 26 000 rpm (105 000 g; mean value) for 56 h (T ) 5 °C), and the supernatant was assayed for toxicity. Assuming spherical particles with a density of 1.1 g cm-3, this ultracentrifugation procedure removes colloidal particles larger than 20 nm (14). All subsamples for bioassays were immediately frozen after sample fractionation. This preservation step can drastically perturb aggregate structure (40), but it was unavoidable to prevent excessive bacterial growth during the time span between sample collection and laboratory bioassays. Filtered waters and raw retentate suspension were tested for toxicity upon thawing. Ultracentrifugation (UCF) of retentate aliquots was started immediately after thawing under conditions (see above) minimizing bacterial growth (41). Toxicity testing of these aliquots was performed immediately after UCF.
Results and Discussion Element and Organic Carbon Distribution. Acceptable mass balances (100 ( 20%) were generally obtained for all elements and OC, although Al, Pb, and Ti occasionally exhibited significant losses (>30%) for the 1 kDa cartridge (Figure S2 and Tables S2 and S3 in Supporting Information). Some high OC blanks were found for both TFF membranes, but except for the 1kDa membrane at site 1, OC mass balances in natural samples usually remained within 100 ( 20%. TFF blanks for trace elements were negligible except for Cr, for which blankcorrected concentrations (with propagated errors) have been used. The concentrations of elements and OC in the two aliquots of filtered water, independently fed to the two TFF membranes, generally agreed within 10% or less (Tables S2 and S3 in Supporting Information). The homogeneity of filtered waters ensures an accurate determination of the LMW fraction, which is calculated by difference between the total colloidal and the HMW colloidal concentrations. Higher concentrations of total filterable OC (sites 1 and 2) and Cr (sites 2 and 3) were measured in the aliquot fed to the 10 kDa cartridge, and the LMW fractions may therefore be underestimated in these cases. On the basis of their distribution among the various fractions (Figure 2 and Table S1 in Supporting Information), most elements can be assigned to one of the following groups: Group 1: major elements (Ca, Mg, K, and Na) preferentially transported in the truly dissolved phase (>90%) and showing little association with colloids. Group 2: terrigenous elements (Al, Ti) essentially associated with the particulate phase (>95%) and with 40-60% of the total filterable concentrations in the HMW colloidal fraction. Group 3: trace elements (Cd, Co, Cu, Fe, Mn, Ni, Pb, and Zn) characterized by varying distribution patterns and by a larger truly dissolved fraction at site 3 as compared with sites 1 and 2. In the case of Mn, this situation probably arises from the dissolution of manganese oxides, which are also scavengers for Co and Ni (25, 42) in the oxygen-depleted conditions of site 3 (see Table S4 in Supporting Information). For the other elements, complexation with ligands passing through the 1 kDa TFF membrane at site 3 cannot be excluded, since as in conventional filtration, the cutoff of TFF membranes is operationally defined (13). The different
element distributions between sites 1 and 2 (river Po) and site 3 (river Lambro) may then originate from differences in the size and nature of the colloidal material in the two rivers (see the discussion on OC below). Note that total filterable Cd concentration at sites 1 and 2 is close to the analytical detection limit (0.007 µg L-1, calculated as 3 times the standard deviation of the analytical blank) and Cd cannot be accurately apportioned between colloids and true solution. On the other hand, at site 3, total filterable Cd is mainly associated with HMW colloids (Figure 2), suggesting a peculiar mechanism of interaction with colloidal matter. According to the classification of Pearson (see, e.g., ref 43), Cd is a “soft acid”, which prefers binding sites containing sulfur rather than oxygen; while Co, Cu, Ni, Pb, and Zn are “borderline acids”, which have higher affinity to oxygen-containing binding sites. Group 4: trace elements (As, Mo, Sb, U, and V) showing minor association with colloids and preferentially transported in the truly dissolved fraction (>75%). These elements exist as anionic complexes in most surface waters and hence show little affinity for natural colloids, most probably bearing negative charges at the circumneutral pH (Table S4 in Supporting Information) observed in this study (44). The likely presence of the poorly adsorbed uranium carbonate and uranium phosphate complexes (45, 46) may explain why the significant pools of colloidal U reported for other systems (ref 47 and references therein) are not observed in the Lambro-Po (Figure 2). As for the remaining elements, Si shows little association with colloids, and the truly dissolved phase accounts for 30% of the total concentration at sites 1 and 2 and 55% at site 3; the rest being in the particulate phase. Total filterable Hg was always below the detection limit (0.01 ng of Hg, absolute detection limit); while an adequate discussion for Cr would require the knowledge of its redox speciation between the (+3) and (+6) valence states. At site 3, Cr probably occurs as particle-reactive Cr(III), since 90% of total filterable concentration is associated with colloids (Figure 2). On the contrary, Cr(VI) should predominate at site 1 where 90% of total filterable Cr is in the truly dissolved fraction, although the occurrence of Cr(III) in the truly dissolved fraction has also been reported (48). Site 2 (30% of total filterable Cr associated with colloids) represents an intermediate situation resulting from the mixing of waters from sites 1 and 3 and, possibly, in situ redox processes. More research on Cr speciation is clearly needed in this system. Note that mafic and ultramafic rocks are present in the Po basin (49) and increase the particulate Cr fractions at sites 1 and 2. Finally, the majority of total filterable organic carbon (OC) is transported in the truly dissolved phase, although significant quantities of colloidal OC are present at all sites (Figure 2). The interpretation of OC distribution is difficult because of the heterogeneous nature of organic matter in aquatic systems (50). The decreasing δ15N values for SPM in the order site 1 (δ15N ) 6.52‰), 2 (4.67‰), and 3 (-1.40‰; our unpublished data) indicate the presence of sewagederived material (see, e.g., ref 51) at sites 2 and 3, in agreement with known inputs of sewage from site 3 (25, 26). A similar trend possibly occurs for CSS. Role of Colloids in Element Geochemical Behavior. At site 2, the relatively unpolluted waters of the river Po (site 1) mix with the polluted waters of the Lambro (site 3). The fraction of Lambro waters (XLambro) in the mixing zone can be estimated using Na as a conservative tracer (25):
XLambro )
([Na+]site 2 - [Na+]site 1)
(a)
([Na+]Lambro - [Na+]site 1)
In the present case, eq a yields an XLambro of 0.17. The theoretical concentrations of elements and OC in case of VOL. 39, NO. 2, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
491
FIGURE 2. Percentage distribution of selected trace elements and organic carbon (OC) among the particulate (black), HMW colloidal (hatched), LMW colloidal (dotted), and truly dissolved fractions (white) in the Lambro-Po system. Particulate Al is not shown since it constitutes over 99% of the total. Truly dissolved OC at site 1 is given as the fraction less than 10 kDa, and only total filterable concentrations (bars with square pattern) are given for Cd at sites 1 and 2 (see text for explanations). conservative mixing between the waters of the rivers Po and Lambro can be calculated by rearranging eq a as follows:
concnsite 2 ) concnLambro × XLambro + concnsite 1 × (1 - XLambro) (b) Significant differences between measured and calculated 492
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 2, 2005
concentrations would indicate non-conservative behavior in the mixing zone (site 2). Elements belonging to groups 1 and 4 (see previous section) tend to behave conservatively (e.g., U in Figure 3). On the other hand, for elements of groups 2 and 3 (and Cr), experimentally measured concentrations tend to be lower than those predicted by eq b (Figure 3). Fractiona-
FIGURE 3. Experimental (white) vs calculated (hatched) concentrations of organic carbon (OC) and selected trace elements at site 2. Calculated concentrations were obtained assuming conservative mixing of the river Po and river Lambro waters at site 2 (formula “b”). Calculation was not possible for truly dissolved (Tr. Diss.) OC due to evident contamination of the permeate of the 1 kDa TFF membrane for site 1 (see text). Vertical bars indicate 1 SD (propagated errors are used for calculated concentrations).
TABLE 1. Estimated Masses (in mg L-1) of the Total, HMW, and LMW Colloidal Suspended Solids (CSS) at the Sampling Sitesa CSStotal (1.2 µm-1 kDa) CSSHMW (1.2 µm-10 kDa) CSSLMW (10-1 kDa)
site 1
site 2
site 3
site 2 (theoretical)b
4.13 ( 0.48 1.25 ( 0.46 2.88 ( 0.15
3.63 ( 0.34 0.4 ( 0.11 3.23 ( 0.32
7.84 ( 0.83 3.97 ( 0.78 3.86 ( 0.28
4.75 ( 0.96 1.71 ( 0.90 3.04 ( 0.31
a Errors are given as ( 1 SD. Masses of total and LMW colloidal suspended solids for site 1 are reported since the poor DOC mass balance for the 1 kDa membrane was solely due to evident contamination of the permeate line (Table S2 in Supporting Information). Expected conservative concentration of CSS at site 2 (site 2 theoretical) have been calculated using eq b (see text). b (0.83 × CSSsite 1) + (0.17 × CSSsite 3).
tion of water samples between colloids and true solution allow the evaluation of the likelihood of the various mechanisms responsible for non-conservative behavior (see Figure 3): (a) adsorption of truly dissolved elements onto SPM (or CSS) or precipitation of mineral phases; that is, only the truly dissolved concentration behave non-conservatively: Mn, Co, and Ni. (b) aggregation/coagulation of colloids; that is, only colloidal concentrations behave non-conservatively: Cu and Fe. (c) a combination of “a” and “b”: Cr, Pb, Al, Ti, and Zn. A slight addition of Cr to the HMW phase is also observed (Figure 3). This peculiarity may be linked to the complex redox chemistry of Cr in aquatic systems. However, interpretation of the results is not straightforward since mass balances indicate minor Cr losses for the 10 kDa membrane at site 1 (Table S3 in Supporting Information) and experimental concentration of LMW Cr at site 2 may be underestimated (see Element and Organic Carbon Distribution).
In the case of OC, substantial removal from the HMW fraction and possible addition to the LMW fraction (Figure 3) are observed. On the other hand, no meaningful calculations can be performed for truly dissolved OC since contamination of the permeate line of the 1 kDa membrane clearly occurred at site 1 (Table S2 in Supporting Information). Note that the concentration of total colloidal OC at site 1 (about the double of HMW OC) is geochemically coherent with the data obtained for the other sites, which should ensure the validity of results for colloidal OC. An overall appreciation of the geochemical role of colloids in the Lambro-Po system can be gained by examining the variations of the mass of CSS among the sites. Total CSS concentration at site 3 is about twice of those at sites 1 and 2, with the largest differences being observed for the HMW fraction (Table 1). The variations of the CSS masses correspond fairly well to the differences in element distribution among sites (Figure 2), even if general fractionation patterns can be influenced by element-specific properties as observed for Cd and Cr. Furthermore, removal of HMW colloids at site VOL. 39, NO. 2, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
493
TABLE 2. Concentration of Organic Carbon and Selected Trace Elements in Suspended Particulate Matter Collected in the Lambro-Po System in July 2001a site1
organic C Cd Co Cu Hg Zn
(%) µg g-1 µg g-1 µg g-1 µg g-1 µg g-1
site2
site3
mean
SD
mean
SD
mean
SD
site
code
avg offsprings (sample)
avg offsprings (control)
p-value
% inhibition
1.83 0.263 16.5 39.4 0.080 100
0.02 0.036 1.37 4.34 0.003 9.8
3.23 0.350 15.4 46.0 0.149 124
0.04 0.025 0.69 1.95 0.007 6.6
10.2 0.813 4.1 79.7 1.564 293
0.26 0.085 0.45 9.59 0.096 35.2
1 1 2 2 3 3
1R1 1UCF 2R1 2UCF 3R1 3UCF
4.5 ( 0.76 6.2 ( 0.89 5.1 ( 0.64 5.5 ( 0.93 4.8 ( 0.89 6.1 ( 0.64
7.1 ( 0.64 6.8 ( 0.71 6.5 ( 0.56 7.1 ( 0.84 7.0 ( 0.76 7.7 ( 0.71
