Environ. Sci. Technol. 2010, 44, 5430–5436
Partition Coefficients of Organic Contaminants with Carbohydrates HSU-WEN HUNG,† TSAIR-FUH LIN,† AND C A R Y T . C H I O U * ,†,‡ Department of Environmental Engineering and Sustainable Environment Research Center, National Cheng Kung University, Tainan 70101, Taiwan, and U.S. Geological Survey, Denver Federal Center, Denver, Colorado 80225
Received February 8, 2010. Revised manuscript received May 3, 2010. Accepted May 10, 2010.
In view of the current lack of reliable partition coefficients for organic compounds with carbohydrates (Kch), carefully measured values with cellulose and starch, the two major forms of carbohydrates, are provided for a wide range of compounds: short-chain chlorinated hydrocarbons, halogenated benzenes, alkyl benzenes, polycyclic aromatic hydrocarbons (PAHs), polychlorinated biphenyls, and organochlorine pesticides. To ensure the accuracy of the Kch data, solute concentrations in both water and carbohydrate phases are measured by direct solvent extraction of the samples. For a given compound, the observed partition coefficient with cellulose (Kcl) is virtually the same as that with starch (Kst). This finding expedites the evaluation of organic contamination with different forms of carbohydrates. The presently determined Kch values of 13 PAHs are substantially lower (by 3-66 times) than the literature data; the latter are suspect as they were obtained with (i) presumably impure carbohydrate samples or (ii) indirectly measured equilibrium solute concentrations in carbohydrate and water phases. Although the Kch values are generally considerably lower than the respective Kow (octanol-water) or Klipid (lipid-water), accurate Kch data are duly required to accurately estimate the contamination of carbohydrates by organic compounds because of the abundance of carbohydrates over lipids in crops and plants. To overcome the current lack of reliable Kch data for organic compounds, a close correlation of log Kch with log Kow has been established for predicting the unavailable Kch data for low-polarity compounds.
Introduction Carbohydrates form a major fraction of natural organic matter. They comprise a wide variety of substances in which the H:O atomic ratios are the same as that in water. Depending on the structural complexity, carbohydrates are divided into three major groups: monosaccharides, disaccharides, and polysaccharides. Monosaccharides (e.g., glucose and xylose) and disaccharides (e.g., sucrose and lactose) are highly soluble in water, whereas polysaccharides (e.g., starch and cellulose) are largely insoluble. In many crops (e.g., wheat and rice) and plants (including grasses), carbohydrates are predominant organic constituents, with lipids (and lipid-like substances) being at trace levels to yield high * Corresponding author phone: +886 6 2353710; e-mail:
[email protected]. † National Cheng Kung University. ‡ U.S. Geological Survey. 5430
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 44, NO. 14, 2010
carbohydrate-to-lipid ratios (often >20) (1-4). Thus, although the partition capacity of a sparingly water-soluble organic compound with a unit mass of cellulose is far less than that with lipids (5-8), the total contaminant uptake by cellulose/ starch relative to that by lipids could become significant with many food crops because of the relatively high carbohydrate content. A close evaluation of the crop contamination by diverse organic compounds from different sources thus requires accurate partition coefficients with cellulose (Kcl) and starch (Kst). Besides the abundance in crops/plants, cellulosic substances also form a large fraction of waste-paper products disposed at various municipal landfill sites (9). This situation raises a concern for whether these substances strongly retain certain organic contaminants and thus influence their transport and fate. Resolution of this question requires well determined or accurately estimated partition coefficients of the organic compounds with celluloses (Kcl). Another merit of the Kcl data has to do with the fact that celluloses are commonly used as the base filter material for separating suspended solids from solution. In this case, it is important to know whether a substance in solution sorbs significantly to the filter to affect its concentration and recovery. At this time, the partition coefficients with carbohydrates (Kch) are unavailable for the majority of common organic compounds. To our knowledge, the existing Kch data consist only of benzene and carbon tetrachloride (5) and some 16 polycyclic aromatic hydrocarbons (PAHs) (e.g., phenanthrene and pyrene) with cellulose (4, 7, 8, 10); there have been virtually no reported partition coefficients with starch despite the importance of this information for assessing food contamination. Moreover, the reported Kch values for PAHs from different sources are widely inconsistent, leading to drastically different estimates of PAH partitioning with carbohydrates. For instance, the reported Kcl for phenanthrene ranges from 156 (7), 170 (8), 952 (10), to 1106 (4), with no clear indication of the reliability of any particular value. The reported Kcl values for pyrene suffer similar uncertainties. These highly discrepant Kcl values result in highly inconsistent correlations between Kcl and Kow (octanol-water) for predicting unavailable Kcl. Here, for example, the correlation reported by Jonker (8) indicates that the Kcl of PAHs are lower than the respective Kow all by about 400 times, whereas the correlation of Zhang and Zhu (4) suggests that the Kcl of PAHs are lower (than the Kow) by 26-41 times. The sharp contrasts in both measured and predicted Kcl values lead to contradictory estimates that carbohydrates may be contaminated by organic compounds. Accurate Kch data are thus critically needed to resolve the confusion. Measurements of accurate Kch values for organic compounds require careful analytical steps because the Kch are usually very small that special care is needed to ensure their reliabilities. In view of the serious shortage of consistent Kch data for assessing the food-crop and waste-site contamination, this study was aimed at providing carefully measured Kch values for a large group of organic compounds to facilitate the contamination assessment. Many classes of compounds were examined, including halogenated benzenes, alkyl benzenes, aniline derivatives, short-chain chlorinated hydrocarbons, PAHs, polychlorinated biphenyls, organochlorine pesticides, and phthalate esters. Besides checking against the available Kch for consistency, this study examined whether the Kch values with starch and cellulose, the two major forms of carbohydrates, differ significantly to necessitate separate accounts. Starch and cellulose are known to have identical molecular compositions, (C6H10O5)n, but the glucose units in 10.1021/es1004413
2010 American Chemical Society
Published on Web 06/22/2010
starch are connected by an alpha linkage and those in cellulose by a beta linkage. The measured Kch values with starch (Kst) and cellulose (Kcl) serve to scrutinize any structural effect between them on contaminant partitioning. With carefully measured Kch values for the selected classes of compounds, a rigorous correlation between Kch and Kow could thus be developed for estimating unavailable Kch values.
Experimental Section Materials. The organic chemicals used as test solutes in this study are of reagent grades or analytical standards from Aldrich, Supelco, Dr. Ehrenstorfer, and Chem Service. Two forms of carbohydrates, starch and cellulose, were obtained from Sigma-Aldrich Chemical Co. and used without further purification. The starch was a pure amylopectin derived from the unmodified Waxy corn, and the cellulose was derived from pure cotton linters in the form of a 20-µm microcrystalline powder. The moisture contents in starch and cellulose were determined to be 7.8% and 1.4%, respectively, by heating the samples overnight at 105 °C. The dry weights of starch and cellulose were corrected for the moisture contents. Sorption Experiments. The partition experiments of organic solutes with either starch or cellulose were performed with a batch equilibration technique. A total of 0.5-10 g of the sample (starch or cellulose) was added to 30-mL screwcapped centrifuge tubes containing 0.005 M CaCl2 water solution to which varied amounts of test solutes in methanol were added to give initial solute concentrations at 20-60% of the water solubility. The amount of methanol added was less than 0.2% (v/v) in water solution in order to minimize the cosolvent effect on solute water solubility. Samples were prepared in triplicate. The pHs of solute-solid suspensions were between 5.2 and 6.0 in all experiments, not adjusted. The samples were equilibrated for 2-3 days on a rotary mixer at temperature of 25 ( 2 °C; the solute concentrations in water and solid phases appeared to reach equilibrium over this period, and no significant degradation of test solutes was observed. After the equilibration, the solid and solution phases were separated by centrifugation at 3500 rpm (2460 g) for 1 h. An aliquot of the supernatant (normally, 10 mL) was withdrawn and quickly transferred (to minimize the solute volatilization) to a tube containing hexane to extract the solute for determining the concentration. To determine solute concentrations in the solid phase (i.e., the precipitated cellulose or starch), the excess water in sample tubes was quickly decanted after water samples were withdrawn for analyses. The sample tube containing a wet solid phase was then weighed to determine the amount of residual water in the solid before the wet solid was extracted by 20 mL of a 1:4 acetone/hexane solvent mixture; the extraction was carried out with a rotary mixer for 2 days. Both the water-phase and solid-phase extracts were analyzed by a gas chromatograph (GC). The concentrations of most short-chain chlorinated hydrocarbons (SCCHs), halogenated benzenes (HABZs), alkyl benzenes (ALBZs), phthalate esters, and aniline derivatives (ANDVs) were analyzed by using a Thermo Finnigan Trace GC System (Thermo Electron Corp., Italy) equipped with a flame ionization detector. For the PAHs, a Varian 4000 GC/MS system, comprised of a CP-3800 GC and a 4000 ion-trap mass spectrometer (Varian Inc., USA), was used to measure the concentrations; the same Varian GC system equipped with an electron capture detector was employed to quantify the concentrations of the selected polychlorinated biphenyls (PCBs) and organochlorine pesticides (OCPDs). For all GC systems, a DB-5MS capillary column (30 m long ×0.25 mm inner diameter with a 0.25 µm film thickness) from Agilent Technologies, Inc., U.S.A., was used. The GC carrier gas was either ultrahigh-purity helium or nitrogen (Jing-Shang Corp., Kaohsiung, Taiwan) maintained at a constant flow rate of 1 mL/min.
With the equilibrium solute concentrations in water (Cw) and in solid cellulose or starch (Cs) determined, the related solute partition coefficients (Kcl or Kst) were then computed as the ratios of Cs to Cw. To improve the accuracy of Kcl or Kst, all solute concentrations in the solid phase were corrected for the amounts of solutes in the residual water of the solid. The overall recoveries of solutes from respective watersolid mixtures based on total detected solute masses were generally in the range of 90-110%. The efficiencies of solutes recovered by the solvent extraction of the liquid (water) phase were in the range of 89-115%, and those by the extraction of the solid phase were in the range of 85-112%. These ranges included errors from transferring the solutes from stock (methanol) solutions to the sample tubes. No correction of Kcl or Kst for the solute recovery with each phase was made, considering that both phases were solvent-extracted and analyzed. According to these recovery data, the estimated analytical errors on Kcl or Kst were less than (15% for the majority of compounds. With the direct solvent extraction of both water and solid samples and high overall and single-phase solute recoveries, the measured Kcl or Kst values in triplicate for given solutes were usually very comparable. Blank experiments consisting of starch or cellulose (10 g) mixed with 30 mL of 0.005 M CaCl2 water solution without added solutes were performed to determine the level of cellulose or starch dissolved in water, quantified as the dissolved organic carbon (DOC). This adopted sample-towater ratio represented the upper limit of the DOC formed in all carbohydrate-water partition experiments. The DOC levels in water with the amounts of starch and cellulose added were found to be 143 and 13 mg/L, respectively. With such low DOC levels and relatively small solute Kch values (shown later), the solubility enhancement of the studied solutes by DOC (11) and the mass loss of applied carbohydrates to solution could both be safely ignored in the Kch determination. The partition coefficients with cellulose (Kcl) and starch (Kst) were both measured for all compounds using the same procedure so that any significant difference between measured Kcl and Kst values could be discerned. The log Kcl or log Kst values from triplicate samples were averaged and reported. To ensure the data quality, the average log Kcl or log Kst values were accepted only when the standard deviations of measured log Kcl or log Kst in triplicate were less than 0.20; otherwise, the experiments were repeated.
Results and Discussion The measured log Kcl and log Kst values for ALBZs, HABZs, and ANDVs are listed in Table 1. Similar sets of data for SCCHs, phthalate esters, PCBs, and OCPDs are listed in Table 2. In Table 3, the measured log Kch of 13 PAHs from this work and the literature values of 16 PAHs are listed for comparison. Also included in Tables 1-3 are the log Kow and log Sw (water solubility) of the compounds. The standard deviations (SD) of the log Kch for all compounds are e0.05, except for 2-toluidine (0.09); N,N-dimethylaniline (0.10); fluorobenzene (0.07), dimethyl phthalate (0.13); and hexachlorobenzene (0.12). The numbers of compounds tested are 10 for ALBZs; 10 for HABZs; 4 for ANDVs; 6 for SCCHs; 3 for phthalate esters; 13 for PAHs; 7 for PCBs; and 6 for OCPDs, which add up to a total of 59 compounds. The presently measured log Kch range from -0.22 (benzene) to 2.96 (2,2′,4,5,5′-PCB), with the corresponding log Kow extending more widely from 2.13 to 6.50. It is observed that the log Kcl are virtually identical to the respective log Kst. This finding indicates that the minor structural difference between cellulose and starch poses no significant impact on solute partitioning. The Kch of a compound is taken as the average of its Kcl and Kst. To our knowledge, no previous studies have systematically studied the solute partition efficiencies with cellulose and starch. VOL. 44, NO. 14, 2010 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
5431
TABLE 1. Water Solubilities (mol/L) and Partition Coefficients (L/kg) of ALBZs, HABZs, and ANDVs with Carbohydrates compound
log Swa
log Kowa
log Kcl
log Kst
log Kchb
ALBZs benzene toluene styrene ethylbenzene 1,2-xylene 1,4-xylene 1,3,5-trimethylbenzene 1-ethyl-2-methylbenzene 1,2,4,5-tetramethylbenzene hexamethylbenzene
-1.64 -2.25 -2.54 -2.82 -2.78 -2.83 -3.24 -3.21 (-4.02) (-4.68)
2.13 2.69 2.95 3.15 3.13 3.18 3.42 3.53 4.10 4.61
-0.23 0.11 0.37 0.46 0.45 0.50 0.70 0.74 1.11 1.54
-0.22 0.12 0.40 0.46 0.46 0.53 0.71 0.78 1.13 1.61
-0.22 0.12 0.39 0.46 0.46 0.52 0.71 0.76 1.12 1.58
HABZs fluorobenzene chlorobenzene bromobenzene iodobenzene 1,2-dichlorobenzene 1,4-dichlorobenzene 1,2,3-trichlorobenzene 1,2,3,4-tetrachlorobenzene pentachlorobenzene hexachlorobenzene
-1.79 -2.38 -2.58 -2.95 -2.98 (-3.03) (-3.79) (-4.59) (-5.20) (-5.71)
2.27 2.84 2.99 3.28 3.38 3.37 4.04 4.60 5.03 5.50
-0.18 0.35 0.42 0.57 0.63 0.69 1.07 1.44 1.90 2.23
-0.13 0.35 0.48 0.58 0.68 0.71 1.11 1.43 1.93 2.19
-0.15 0.35 0.45 0.58 0.66 0.70 1.09 1.44 1.92 2.21
-0.41 -0.82 -1.53 -2.04
1.09 1.42 1.90 2.31
-0.06 -0.12 0.14 0.29
-0.06 -0.12 0.20 0.24
-0.06 -0.12 0.17 0.27
ANDVs aniline 2-toluidine 2-chloroaniline N,N-dimethylaniline
a Values from Chiou et al. (12). The log Sw enclosed with parentheses are the supercooled liquid solubilities. 0.5(Kcl + Kst).
b
Kch )
TABLE 2. Water Solubilities (mol/L) and Partition Coefficients (L/kg) of SCCHs, Phthalate Esters, PCBs, and OCPDs with Carbohydrates compound
log Swa
log Kowa
log Kcl
log Kst
log Kchb
SCCHs carbon tetrachloride trichloroethylene tetrachloroethene 1,1,1-trichloroethane 1,1,2,2-tetrachloroethane pentachloroethane
-2.28 -1.99 -2.92 -2.00 -1.75 -2.61
2.73 2.53 3.38 2.47 2.39 2.89
0.22 0.03 0.58 -0.08 -0.09 0.31
0.23 0.08 0.63 -0.09 0.04 0.31
0.23 0.06 0.61 -0.08 -0.02 0.31
Esters dimethyl phthalate diethyl phthalate dibutyl phthalate
-1.59 -2.31 -4.33
1.86 2.47 4.08
-0.34 -0.16 1.15
-0.31 -0.07 1.03
-0.32 -0.11 1.10
PCBs biphenyl 4-PCB 2,4′-PCB 4,4′-PCB 2,4,4′-PCB 2,2′,5,5′-PCB 2,2′,4,5,5′-PCB
(-3.95) (-4.67) (-5.34) (-5.56) (-6.02) (-6.19) (-7.01)
4.04 4.61 5.10 5.33 5.62 5.81 6.50
1.03 1.51 1.98 2.10 2.25 2.36 2.91
1.06 1.44 2.08 2.06 2.23 2.41 3.01
1.05 1.48 2.03 2.08 2.24 2.39 2.96
OCPDs lindane dieldrin chlordane heptachlor p,p′-DDE p,p′-DDT
(-3.62) (-5.22) (-6.03) (-6.05) (-6.15) (-6.79)
3.72 5.10 5.54 5.73 5.77 6.36
1.01 2.06 2.49 2.37 2.66 2.84
1.04 2.03 2.45 2.36 2.62 2.92
1.03 2.05 2.47 2.37 2.64 2.88
a Values from Chiou et al. (12) and Mackay et al. (13). The log Sw enclosed with parentheses are the supercooled liquid solubilities. b Kch ) 0.5(Kcl + Kst).
Examination of the measured log Kch against the respective log Kow in Tables 1-3 reveals a close correspondence between log Kch and log Kow for all low-polarity compounds. As seen, 1,2,4,5-tetramethylbenzene, 1,2,3-trichlorobenzene, fluorene, 5432
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 44, NO. 14, 2010
biphenyl, and dibutyl phthalate, which show similar log Kow at 4.04-4.18, exhibit comparable log Kch at 1.05-1.12. This consistency facilitates the prediction of Kch from available Kow for nonpolar compounds. Since the log Kow of compounds
TABLE 3. Water Solubilities (mol/L) and Partition Coefficients (L/kg) of Selected PAHs with Carbohydrates from Different Sources compound
log Swa
log Kowa
log Kclb
log Kchc
log Kchd
indane naphthalene acenaphthene fluorene 2,6-dimethylnaphthalene 1,5-dimethylnaphthalene phenanthrene anthracene fluoranthene pyrene benz[a]anthracene chrysene benzo[e]pyrene benzo[b]fluoranthene benzo[k]fluoranthene benzo[a]pyrene benzo[ghi]perylene dibenz[ah]anthracene indeno[123,cd]pyrene
-3.04 (-3.09) (-3.89) (-4.14) (-4.25) (-4.19) (-4.48) (-4.63) (-5.15) (-4.92) (-5.89) (-6.11) (-6.82) -
3.33 3.36 3.92 4.18 4.31 4.38 4.46 4.54 5.16 5.18 5.61 5.73 6.44 6.20 6.20 6.34 7.04 6.84 7.04
2.23 2.34 2.32 2.39 3.11 3.15 3.74 3.76 3.85 3.93 4.42 4.42 4.47
1.75 2.35 2.57 3.04 3.71 -
0.52 0.67 0.91 1.07 1.27 1.30 1.40 1.54 1.89 1.89 2.36 2.51 2.77 -
a Values from Chiou et al. (12) and Mackay et al. (14). The log Sw enclosed with parentheses are the supercooled liquid solubilities. b Values with cellulose from Jonker (8). c Values with carbohydrates from Zhang and Zhu (4). d Kch as the average of Kcl and Kst from this study.
FIGURE 1. Plot of log Kch versus log Sw for the low-polarity compounds. are inversely related to their liquid-solute water solubilities (log Sw) (12, 15), a similar relation between log Kch and log Sw is expected for the nonpolar solutes, despite that the Kch are considerably lower than the respective Kow due to sharply contrasting polarities of octanol and carbohydrates. Correlation of the log Kch and log Sw data for the low-polarity compounds, excluding ANDVs, in Tables 1-3 gives log Kch ) -0.595log Sw- 1.21 (n ) 55; R2 ) 0.988) (1) with a standard error of 0.103 for the observed vs estimated log Kch. The correlation spans about 5.4 orders of magnitude in Sw (mol/L) and 3.2 orders of magnitude in Kch (L/kg), as shown in Figure 1. Whereas the solute molar volume is known to have a minor effect on the partition coefficient (12, 16), it is not addressed here as it is not the primary focus of this work. As mentioned, there is currently a lack of literature Kch data for common organic compounds, other than benzene and carbon tetrachloride (CT) (5) and some 16 PAHs (see Table 3), that are available for comparison with the present data. The earlier reported Kch values for benzene (0.56) and
FIGURE 2. Plots of log Kch versus log Kow for selected PAHs based on present and literature data: 9 Zhang and Zhu (4); 0 Jonker (8); and b this study. CT (1.75) with cellulose (5) are well consistent with the present values of 0.60 and 1.70, respectively. Moreover, the small Kch values of benzene and CT are closely correlated with their partition coefficients with other natural organic matters (Kom) (e.g., muck and peat) with respect to the organic-matter composition/polarity (5). Considering that the Kch of all nonpolar compounds including benzene and CT from the present study are well correlated with their respective Sw and Kow, this consistency along with the rigorous analytical method used may be taken as an unambiguous support of the credibility of the present Kch data. It is noted that the literature Kch for PAHs (4, 8) are higher than the present Kch by 3-66 times (see Table 3 and Figure 2). The high literature Kch are suspect from the standpoint of solute partitioning, considering the profound disparity in polarity between nonpolar solutes (e.g., PAHs) and extremely polar carbohydrates (5). Resolution of such discrepancies is essential to a reliable estimation of the contaminant uptake by carbohydrates. For example, based on relatively high Kch values of the five PAHs studied, Zhang and Zhu (4) contended that the uptake of PAHs by carbohydrates is significant relative to that by lipids in plants (ryegrass roots) because of the VOL. 44, NO. 14, 2010 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
5433
generally high carbohydrate-to-lipid weight ratios in crops and plants. The validity of this view for most PAHs with common crops/plants may be questioned according to the present Kch that are 12-66 times lower than claimed by Zhang and Zhu (4). A more detailed account of this subject is presented later using the Klipid (lipid-water) and present Kch data of the organic compounds involved. Judging from the experimental steps employed by Zhang and Zhu (4) to measure the Kch, we speculate that their carbohydrate (cellulose) samples prepared with a series of solvent extractions of ryegrass roots were not sufficiently pure or that the sample-to-water ratios they used (1:400 to 1:2000) were far too small to significantly reduce the solute concentration in water to be accurately quantified by the concentration change. The effects of the sample purity and sample-to-water ratio on the solute concentration reduction may be illustrated using phenanthrene (PHN) as a model solute along with a carbohydrate-to-water ratio of 1:2000 (as used for PHN) and Kch ) 25 for PHN from this work. A massbalance calculation with the said carbohydrate-to-water ratio indicates that the PHN concentration in water at equilibrium will decrease only 1.25% relative to that of the carbohydratefree control sample. Thus, if the carbohydrate samples of Zhang and Zhu (4) were pure, the decrease in PHN concentration in water would be far too small to be quantified accurately, considering a generally much greater analytical error. On the other hand, if these authors observed a considerably higher (than 1.25%) PHN concentration decrease at equilibrium that could be accurately quantified, their samples could not be a pure carbohydrate. Similar arguments would discredit the relatively high Kch values for other PAHs, such as pyrene. Although the Kcl values measured by Jonker (8) with a cellulose sample are considerably lower than the Kch by Zhang and Zhu (4) for the same PAHs (e.g., by 6.5 times for PHN and 21 times for pyrene), they are still appreciably higher (by 3-15 times) than the present Kch. The cellulose used by Jonker was an oak-derived, chlorine-free sample supplied by a paper factory that was further extracted with hexane to remove contaminants. It was probable that this sample contained residual lignin (a much less polar substance than cellulose) that affected the solute partitioning. Meanwhile, in Jonker’s experiments, the PAH concentrations in cellulose were based again on changes of solute concentrations in water after equilibration; the concentrations in water were not directly measured but were estimated rather by conversion from the determined solute concentrations in solid-phase microextraction (SPME) fibers added to the water-solid mixtures. Estimated PAH concentrations in cellulose, and thus the measured Kcl, could suffer significant uncertainties because the errors introduced by these steps might be propagated or even magnified if the potential solute losses were not closely accounted for. It is difficult, however, to pinpoint the exact causes for the high Kcl. In the present study, the Kch were achieved with the uses of highly pure starch and cellulose and solvent-extracted solute concentrations with both water and solid phases. The sample purities ensured that no gross measurement errors were introduced. The solvent extraction of both phases enhanced the analytical accuracy and minimized the effect of potential solute losses on the measured Kch. These precautions are essential to quantifying the small solute partition coefficients with cellulose and starch. We now consider the correlation between log Kch and log Kow for various organic compounds. On the partitioning of 13 PAHs with cellulose, Jonker (8) developed the following log Kcl ) 1.0 log Kow - 2.6 (n ) 13; R2 ) 0.95)
5434
9
(2)
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 44, NO. 14, 2010
which asserts that the Kcl are uniformly lower than the respective Kow by about 400 times for any of the 13 PAHs studied. On the partitioning of five PAHs with carbohydrates, Zhang and Zhu (4) observed instead the correlation log Kch ) 1.23 log Kow - 2.42 (n ) 5; R2 ) 0.975)
(3)
which indicates a significantly different slope (>1). Over the same range of log Kow for the five PAHs (3.30-5.18) studied by Zhang and Zhu (4), the Kch (or Kcl) predicted by eq 3 is 8-23 times higher than those by eq 2. For the 13 PAHs presently examined, many being identical to the ones selected by Jonker (8) and Zhang and Zhu (4), a distinct correlation is found log Kch ) 0.770 log Kow - 2.04 (n ) 13; R2 ) 0.994) (4) in which the slope is clearly less than 1 and the intercept is significantly different from those in eqs 2 and 3. A comparison of the correlations represented by eqs 2-4 is illustrated graphically in Figure 2. Not only that eq 4 predicts considerably lower log Kch than eqs 2 and 3 for PAHs over the same log Kow range, the less-than-1 slope in eq 4 implies that as log Kow increases (or log Sw decreases) within an extensive set of low-polarity compounds, there is a progressive decrease in the affinity of the compound with carbohydrates. This inferred characteristic is in keeping with the fact that when log Kow increases in a chemical series, there is a concomitant increase in solute molecular size (or molar volume) that acts to amplify the incompatibility between a nonpolar solute and a highly polar carbohydrate phase. Therefore, the range of partition coefficients embraced by a group of nonpolar solutes with carbohydrates (Kch) would be compressed relative to that of the corresponding partition coefficients with considerably less polar octanol (i.e., Kow). In view that eq 2 (with a slope of 1) implies that the Kch and Kow span over the same order of magnitude while eq 3 (with a slope >1) suggests that the Kch spread over a greater range than the Kow, the credibility of reported log Kch data for PAHs is thus highly suspect. To substantiate the preceding argument, one observes with the data from this work that the Kow span about 4.4 orders of magnitude from benzene to 2,2′,4,5,5′-PCB, while the corresponding Kch span about only 3.2 orders of magnitude. In other words, the Kow/Kch ratio increases with increasing Kow within a series of the low-polarity compounds. For instance, based on the present Kch with PAHs, this ratio is 490 for naphthalene (log Kow ) 3.36), 1150 for PHN (log Kow ) 4.46), 1950 for pyrene (log Kow ) 5.18), and 3720 for benzo[a]pyrene (log Kow ) 6.34). For ALBZs, the ratio is 223 for benzene (log Kow ) 2.13), 457 for 1,4-xylene (log Kow ) 3.18), 513 for 1,3,5-trimethylbenzene (log Kow ) 3.42), and 1070 for hexamethylbenzene (log Kow ) 4.61). Similar effects are evident for compounds from other series. Thus, the correlations with eqs 2 and 3, which imply either a fixed or a decreasing Kow/Kch ratio as the Kow increases, do not properly depict the actual relative change between Kch and Kow. In particular, the error on estimated log Kch by use of eq 3 is magnified as the log Kow increases. Since the log Kow are lower than the log Klipid by less than 0.20 units for most low-polarity compounds (17), the observed increase in Kow/Kch with increasing Kow thus affects the fractional uptakes of contaminants by carbohydrates and lipids in a crop/plant, assuming that lipids and carbohydrates are the principal partition components in dry crop/plant matter (6). This means that as the Kow of a compound increases, the contribution of lipids relative to that of carbohydrates to the compound uptake by a crop/plant increases, making the latter effect increasingly less important.
FIGURE 3. Plot of log Kch versus log Kow for the low-polarity compounds, with the dotted lines showing the 95% confidence region.
This scenario may be illustrated by the Klipid and Kch data of a series of compounds covering a range of log Kow, such as the data of aniline, benzene, 1,2-dichlorobenzene, and 1,2,3,4tetrachlorobenzene. With the log Klipid being 0.91, 2.25, 3.51, and 4.78, respectively, for these compounds (17) and their log Kch given in Table 1, the respective Klipid/Kch ratios are 9.33, 295, 708, and 2190. Thus, the conditions for carbohydrates in a crop/plant to exhibit an equal mass uptake as the lipid component will be when the weight-fraction ratio of carbohydrates to lipids (fch/flipid) is 9.33 for aniline, 295 for benzene, 708 for 1,2-dichlorobenzene, and 2190 for 1,2,3,4tetrachlorobenzeene. That is, for different compounds partitioned to a given dry crop/plant matter with a fixed fch/flipid ratio, the fractional uptake by carbohydrates decreases as the compound’s Kow increases. In the above example, one sees that if the fch/flipid ratio is 20, i.e., if the lipid content in a dry crop/plant sample is about 5% by weight, aniline will exhibit a greater fractional uptake by carbohydrates (68%) than by lipids (32%); benzene shows only a small fractional uptake by carbohydrates (6.3%) relative to that by lipids (93.7%); 1,2-dichlorobenzene shows no more than a trace uptake (2.7%) and 1,2,3,4-tetrachlorobenzene a negligible uptake by carbohydrates. For the five PAHs analyzed by Zhang and Zhu (4), as listed in Table 3, the relatively high Klipid/Kch ratios (417-2340) derived from the measured Klipid (4) and present Kch data will call for extremely large fch/flipid ratios (>417-2340), or exceptionally small lipid contents (, 0.25%), in dry crop/plant matter to make the uptake by carbohydrates equal to that by lipids. Such low lipid contents have hardly been observed with food crops and small plants (1-3). In view of the close relation between log Kch and log Kow for all low-polarity compounds, it is profitable to combine their log Kch and log Kow data to establish a single log Kch-log Kow correlation. For all low-polarity compounds, excluding moderately polar ANDVs, this regression gives log Kch ) 0.741 log Kow - 1.86 (n ) 55; R2 ) 0.991) (5) with a standard error of 0.086 for the observed vs estimated log Kch. The related log Kch-log Kow plot, along with the 95% confidence region, is shown in Figure 3. As noted, the slope and intercept in eq 5 are comparable with those in eq 4 despite that different chemical classes and log Kow ranges are involved. The predicted log Kch by eqs 4 and 5 for the 13 PAHs in this study are essentially the same (differing only by
0.01-0.10 log units). In particular, the consistent less-thanone slopes with eqs 4 and 5 are illustrative of the different orders of magnitude embraced by Kch and Kow. Since the correlation with eq 5 covers many classes of low-polarity compounds, it should provide a reasonable accuracy on the predicted log Kch for all relatively nonpolar compounds, which is much demanded for overcoming the current serious deficiency in Kch data. Finally, it is anticipated that the Kch for moderately polar compounds should be enhanced with respect to their Kow as compared with the Kch and Kow of the low-polarity compounds. This effect is manifested by the data of ANDVs in Table 2, in which 2-toluidine and 2-chloroaniline, for example, show a higher Kch than benzene, whereas benzene exhibits conversely a higher Kow. This characteristic reflects the improved intermolecular associations between these solutes and polar carbohydrates through polar interactions and/or H-bonding. Thus, for those compounds having a moderate polarity, a different log Kch-log Kow correlation is expected to arise. In light that only a limited amount of data is available for such compounds, no rigorous correlation could be achieved in this study. However, as noted with the data of four ANDVs, the Kch values for moderately polar compounds are not tremendously higher than those of the low-polarity compounds with similar log Kow values (e.g., N,N-dimethylaniline vs benzene). An important reason for this effect seems to be the strong association of water with highly polar carbohydrates that reduces the solute-carbohydrate interaction. Lastly, no attempt was made to measure the Kch of strongly hydrophilic solutes (e.g., small alcohols, amines, and organic acids) in this work because both their Kch and Klipid values are expected to be relatively small (
