Arch. Environ. Contam. Toxicol. 32, 232–245 (1997)

A R C H I V E S O F

Environmental Contamination a n d Toxicology r 1997 Springer-Verlag New York Inc.

Distribution of Polychlorinated Naphthalene Congeners in Environmental and Source-Related Samples U. Ja¨rnberg1, L. Asplund1, C. de Wit1, A.-L. Egeba¨ck1, U. Wideqvist1, E. Jakobsson2 1 2

Laboratory for Analytical Environmental Chemistry, Institute of Applied Environmental Research, Stockholm University, S-106 91 Stockholm, Sweden Department of Environmental Chemistry, Wallenberg Laboratory, Stockholm University, S-106 91 Stockholm, Sweden

Received: 29 June 1995/Revised: 16 May 1996

Abstract. Polychlorinated naphthalene (CN) congener profiles in environmental and source related samples were compared graphically and by principal component analysis. Samples investigated included biological, sediment, water, and air samples, technical polychlorinated biphenyl (PCB) and polychlorinated naphthalene (PCN) formulations, as well as municipal waste incineration (MWI) fly ash and graphite electrode sludge. Biological samples showed a preferential enrichment of planar, 1,3,5,7-substituted tetra-, penta-, and hexachlorinated congeners and most of these samples showed profiles that displayed some similarity to those found in the technical PCB formulations. Sediment samples representing diffuse pollution, i.e., sediment samples from remote sites, showed an elevated abundance of the planar hexa- and heptaCN congeners (1,2,3,4,6,7-/1,2,3,5,6,7- and 1,2,3,4,5,6,7-). The CN congener profile found in these sediment samples and the two air samples were more similar to the technical PCB formulations than to the investigated MWI and graphite sludge samples. Samples from three PCB contaminated lakes displayed similar congener profiles as Aroclor 1242, 1254 and Clophen A40. Two sediment samples and a pike sample collected from the vicinity of a chloroalkali plant showed profiles that were closely related to the investigated graphite electrode sludge sample. None of the environmental samples displayed profiles similar to low or medium chlorinated technical PCN (Halowax 1099, 1013, and 1014).

Polychlorinated naphthalenes (PCNs) are a group of compounds consisting of two fused aromatic six membered rings where the hydrogen atoms have been replaced with chlorine atoms in one to eight positions. In all there are 75 possible chlorinated naphthalenes (CNs). PCNs were first discovered in the middle of the 19th century (Laurent 1833) and subsequently were found to possess several valuable technical properties such as good electrical insulation properties, excellent weather

Correspondence to: U. Ja¨rnberg

resistance and low flammability. Therefore, several technical PCN formulations were manufactured from the beginning of this century by, among others, the Koppers Company in the United States. They produced PCN mixtures with differing chlorine content, known as Halowaxes, which were used in electrical and electronic equipment (Kover 1975). There exists no estimate of the cumulative world production of PCNs, but from available figures (de Voogt et al. 1989; Crookes and Howe 1993), it may be estimated that several hundred thousand tonnes have been produced, most of which dates back to the years between 1920 and 1950. It was recognized early that workers exposed to fumes of PCN developed chloracne, serious illness, and acute liver damage that was lethal in a number of cases (Greenburg 1939). The acneigenic properties and acute liver damage have been attributed to technical products containing penta- and hexachlorinated CNs (Shelley and Kligman 1957). It appears from several reports (Kover 1975; Brinkmann and Reymer 1976; Crookes and Howe 1993) that most of the toxic properties are associated with the higher clorinated CNs (penta- to heptaCNs). Because of their high acute toxicity in humans, PCNs gradually were replaced by less toxic materials. They have never been banned and though most of the production has now ceased the occurrence of PCNs in modern electronic equipment was reported as late as 1992 (Weistrand et al.). A few recent toxicological studies have focused on the dioxin-like toxicity of PCNs, since these are planar compounds with structures similar to PCDD/PCDF (Hanberg et al. 1990; Engwall et al. 1994). These investigations indicated that a few hexachlorinated CNs (1,2,3,4,6,7-, 1,2,3,5,6,7-, 1,2,3,5,7,8-, and 1,2,3,4,5,6-HxCN) and one heptachlorinated CN (1,2,3,4,5,6,7-HpCN) induce EROD (7-ethoxy-resorufin-o-deethylase) and AHH (aryl hydrocarbon hydroxylase) enzyme activity and that the apparent dioxin-like activity for these congeners may be of the same order as that previously established for octachlorodibenzo-pdioxin (toxic equivalency factor of 0.001 relative to 2,3,7,8TCDD). The distribution of PCNs to the environment is not well understood. Several potential sources of PCNs in the environment have been suggested apart from those associated with the production and continued use of PCN formulations. For ex-

Polychlorinated Naphthalene Congeners in the Environment

ample, elevated environmental levels of PCNs have been reported in sediment and fish samples from locations near a chloroalkali plant (Ja¨rnberg et al. 1993) and a magnesium refinery (Schlabach et al. 1994). One route to the environment is also through the use of polychlorinated biphenyls (PCB), since PCNs exist as microcontaminants in PCB formulations (Haglund et al. 1993). High levels of PCNs have, for instance, been found in fish and sediment samples from a lake contaminated with PCB from a recycled paper plan (Ja¨rnberg et al. 1993). Other authors have reported on the occurrence of PCNs in fly ash from municipal waste incinerators and in waste from a copper ore smelter indicating formation during thermal processes (Oehme et al. 1987; Benfenati et al. 1991; Wiedmann and Ballschmiter 1992, 1993; Theisen et al. 1993; Nakano et al. 1993; Takasuga et al. 1994; Imagawa et al. 1993). Formation of chlorinated naphthalenes from pyrolysis of chlorinated solvents such as tetrachloroethylene and polyvinylidene chloride has also been suggested (Tirey et al. 1990; Yasuhara and Morita 1988). The environmental levels found in piscivorous birds from the Baltic Sea indicate that the TCDD-like toxicity of the hexa- and heptaCNs may contribute considerably to the total TCDD-like toxicity (Ja¨rnberg et al. 1993). This may also be the case for fish collected near the point sources at the chloroalkali and magnesium plants (Ja¨rnberg et al. 1993; Schlabach et al. 1994). In order to understand how PCNs are distributed into the environment, more information is needed on the relation between the congener profiles found in potential source samples and samples from remote sites with no known point sources (background). Unfortunately, data on CN congener profiles in environmental samples are very limited, especially from background sites. The purpose of the present investigation was first of all to complement the available data on CN congener profiles and levels with samples from background sites and some potential point sources as well as some additional predatory species. The second purpose was to provide an overview of the different CN congener profiles found in environmental samples and to try to assess relationships between the different samples with special reference to potential sources. Similar to PCDD/Fs and PCBs, the large number of PCN congeners typically present in environmental samples complicates direct comparison of congener profiles in a large set of samples. Multivariate statistical tools such as principal component analysis (PCA) can be used to calculate new variables, principal components (PC) that carry the essential information of the original variables. A basic introduction to PCA is given by Wold et al. (1987). The principal components can then be used to provide useful two-dimensional plots for similarity and classification studies. This has been successfully demonstrated for investigating polychlorinated biphenyls in biota (Schwartz and Stalling 1991) and polychlorinated dibenzo-p-dioxin (PCDD) and dibenzofuran (PCDF) congener profiles in sediment (Cash and Breen 1992; Wenning et al. 1993) and air samples (Tysklind et al. 1993).

Material and Methods Samples The PCN data covered by this investigation were extracted from two previous publications ( Ja¨rnberg et al. 1993; Haglund et al. 1993) and supplemented with data from more recently analyzed samples.

233

The environmental samples previously analyzed included fresh water pike (Esox lucius) from lakes with a suspected PCB point source pollution (L. Ja¨rnsjo¨n and L. Kyrksjo¨n), one lake with a chloroalkali point source pollution (L. Va¨nern) as well as a pike sample from a remote lake with no known point source (L. Storvindeln). PCN levels and distribution in liver and muscle were studied in burbot (Lota lota) and cod (Gadus morrhua), and age dependence and sampling season in Baltic herring (Clupea harengus, spring and fall caught and four and six years old). Grey seal (Halichoerus grypus), common (harbour) porpoise (Phocaena phoceana), and guillemot (Uria aalge) were analyzed to investigate possible biomagnification. Guillemot also were analyzed to study PCN time trends. Geographical distribution was studied using samples of fresh water surficial sediment from lakes with different locations and pollution situations. These analyses provided no background data on PCN in sediment due to a too small sample amount. One sample of percolating water from a municipal waste dump site and one sample of carbon-less copy paper (which was previously in use in Sweden) also were analyzed previously. Samples analyzed within the present investigation were two homogenates of egg from Baltic white-tailed sea eagle (Haliaeetus albicilla), two homogenates of otter (Lutra lutra), surficial sediment samples from two sites with a suspected point source pollution (L. Bengtsbroho¨ljen, recycled paper plant) and two background sites (L. Storvindeln and Baltic Sea, west of Gotland), one surface fresh water sample (L. Ja¨rnsjo¨n), one MWI fly ash sample, one sample of graphite sludge from a chloroalkali process and several technical PCN mixtures with different chlorination degrees: Halowaxes 1099 (52% chlorine content), 1013 (56%), 1014 (62%), 1051 (70%) (Koppers Co. Inc. Pittsburg, PA, USA).

The relative CN congener composition of two air samples have been included in this study. These samples were collected at two background sites (Gotland, southernmost part and Ammarna¨s, alpine region of Sweden) and were representative of most of the air samples analyzed (Egeba¨ck et al. 1995). The sampling period for all samples, with the exception of the guillemot and otter samples, ranged from 1988 to 1991. Table 1 provides sample type, location and the sum of PCN with four to seven chlorines for all samples as well as the abbreviations used to identify each sample in the multivariate part of this study. The locations of the sampling sites are shown on the map of Scandinavia in Figure 1.

Analytical Procedure The cleanup and analysis of the samples in this study is essentially the same as that for the previously analyzed samples (Ja¨rnberg et al. 1993). Biological samples were homogenized and extracted with glassdistilled n-hexane/acetone and further purified by liquid-liquid partitioning between diethyl ether (HPLC-grade, Labscan, Dublin, Ireland)/nhexane and a water phase containing 0.9% sodium chloride in 0.1 M phosphoric acid. The n-hexane phase was gently evaporated to dryness and the residual lipid content determined gravimetrically. This residue was redissolved in n-hexane (Jansson et al. 1991). Wet centrifuged (1800 rpm 10 min) sediment was batch extracted in a screw-cap vial with glass distilled n-hexane/acetone and analytical grade 2-propanol (Merck, Darmstadt, Germany) according to Jensen et al. (1977). Graphite sludge and fly ash were Sohxlet-extracted for 24 hs, using analytical grade toluene (Labscan, Dublin, Ireland). Water and air were sampled on polyurethane foam plugs and Sohxlet-extracted as above with toluene and acetone/n-hexane respectively. All samples except the Halowax mixtures were treated with concentrated sulphuric acid prior to cleanup. Sulphuric acid treated and untreated Halowax 1014 were analyzed in two subsequent GC-MS runs to check that this pretreatment does not influence the congener levels or relative composition.

Table 1. Sample description Sample Id. Description

Latin Name

Location

PiJaa PiKma PiKla PiSaa PiVma PiVla PiSta PiKya PiKya BuEla BuEma BuPla BuPma BuSla BuSma He6sa He4f a He6f a Coma Cola Gu74a Gu76a Gu78a Gu82a Gu87a EaNob EaOeb OtLob OtHib SGra SJa¨1a SJa¨2a SSja SHea SRia SGo1–3a SKaa SAna SBeb SMub SStb SBab

pike muscle pike muscle pike liver pike muscle pike muscle pike liver pike muscle pike muscle pike muscle burbot liver burbot muscle burbot liver burbot muscle burbot liver burbot muscle Baltic herring, spring 6 year Baltic herring, fall 4 year Baltic herring, fall 6 year cod muscle cod liver guillemot, pooled 10 ind. 1974 9 10 individuals 1976 9 pooled 10 ind. 1978 9 pooled 10 ind. 1982 9 pooled 10 ind. 1987 white tailed eagle, 1989 9 9 9 , 1985 otter, homogenate, 6 ind., low PCB level otter, homogenate, 9 ind., high PCB level sediment, 0–2 cm depth sediment, 0–1 cm depth sediment, 9–10 cm depth sediment, 0–1 cm depth sediment, 9 sediment, 9 sediment, 9 sediment, 9 sediment, 9 sediment, 9 sediment, 9 sediment, 9 sediment, surficial

Esox lucius 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 Lota lota 9 9 9 9 9 9 9 9 9 9 Clupea harengus 9 9 9 9 Gadus morrhua 9 9 Uria aalge 9 9 9 9 9 9 9 9 Haliaeetus Albicilla 9 9 Lutra Lutra 9 9

R. Emån, L. Ja¨rnsjo¨n L. Va¨nern, Kattfjorden 9 9 L. Va¨nern, Sandviken L. Va¨nern, Vassbotten 9 9 L. Storvindeln L. Kyrksjo¨n L. Kyrksjo¨n Bothnian Bay, Etukrunni 9 9 Torne R, Pajala 9 9 Bothnian Bay, Seskaro¨ 9 9 Karlskrona archipelago 9 9 9 9 9 9 , Utklippan 9 9, 9 Gotland, Stora Karlso¨ 9 9 9 9 9 9 9 9 9 9 9 9 Norrko¨ping, Baltic Proper ¨ sthammar, 9 O 9 Several different sites in Sweden 9 9 9 9 River Emån, L. Gro¨nskogssjo¨n River Emån, L. Ja¨rnsjo¨n, PCB polluted River Emån, L. Ja¨rnsjo¨n, PCB polluted River Emån, L. Sjunnen River Dala¨lven, Hedesundafja¨rden River Go¨ta a¨lv, Rivo¨fjorden 9 9 9 , Gothenburg harbour L. Va¨nern, Kattfjorden, Chloroalkali L. Va¨nern, Anholmsviken, Chloroalkali L. Bengtsbroho¨ljen, Chloralkali L. Munksjo¨n, PCB polluted L. Storvindeln, background site Baltic Sea, Gotland, no known point source

WaVa WaJb

percolating water, city dump site fresh water

Stockholm, Vaxholm River Emån, L. Ja¨rnsjo¨n, PCB polluted

AiAb AiHb

air sample, gas phase (PUF) 1992 air sample, gas phase (PUF) 1992

Ammarna¨s Hoburgen, Gotland

Papa Grab Mwib H99b H13b H14b H51b A30c A40c A50c 1016c 1232c 1242c 1248c 1254c 1260c

carbon-less copy paper, PCB containing, 1979 graphite sludge, chloroalkali industri fly ash, municipal waste incinerator (MWI) technical formulation HalowaxR 1099 9 9 HalowaxR 1013 9 9 HalowaxR 1014 9 9 HalowaxR 1051 technical formulation ClophenR A30 9 9 ClophenR A40 9 9 ClophenR A50 technical formulation AroclorR 1016 9 9 AroclorR 1232 9 9 AroclorR 1242 9 9 AroclorR 1248 9 9 AroclorR 1254 9 9 AroclorR 1260

a

Stockholm

Results reported in Ja¨rnberg et al. 1993 Samples analyzed in this investigation c Results reported in Haglund et al. 1993 d Sum of CNs with four to seven chlorines ng/g based on lipid weight/dry weight, respectively, or ng/L for water samples % Only relative composition available b

Total PCNd ng/g 360 210 280 13 15 17 2.6 170 130 2.0 4.9 2.0 2.9 0.98 4.4 23–26 8.4–8.7 9.6–11 10 9.8 200 220 180 130 84 130 120 7.0 2.6 6.4 41 270 2.0 3.1 0.62 1.0–1.3 8.0 260 23 18 0.23 7.6 2.6 0.89 % % 2400 150 28 % % % % 870 3 103 810 3 103 1.8 3 103 6.5 3 103 170 3 103 100 3 103 67 3 103 3.5 3 103 2.7 3 103

Polychlorinated Naphthalene Congeners in the Environment

Fig. 1. Map of Scandinavia, showing sampling locations in Sweden and in the Baltic sea

235

236

Recovery standard, 13C12 labelled 3,38,4,48-tetrachlorobiphenyl (CB# 77, Cambridge Isotope Laboratories Inc., Cambridge, MA, USA) was added to the samples prior to the extraction with the exception of the biological samples where the standard was added to the n-hexane extract. Further purification was performed with two HPLC gel permeation columns (PL-GEL, 5 µm, 50 Å, 300 mm 3 7.5 mm Polymer Laboratories, UK) coupled in series. The eluent was dichloromethane:cyclohexane, 1:1 at a flow rate of 0.7 ml/min. This step was followed by fractionation on two 2-(1-pyrenyl)ethyldimethylsilylated silica HPLC columns (Cosmosil PYE 5 µm, 150 mm 3 4.6 mm, Nacalai Tesque, Japan) coupled in series. The eluent was n-hexane, saturated with water, at a flow rate of 0.5 ml/min. A fraction containing the CN congeners with four to eight chlorines was obtained by backflush elution at a flow rate of 1.2 ml/min. Final determination was performed on a Hewlett Packard GC-MSD (5880/5970b HP; Avondale, PA) equipped with a 60 m 3 0.25 mm 3 0.25 µm fused silica capillary column (DB5-MS, J&W Scientific; Folsom, CA). Analytes were quantified against one authentic reference substance for each homologue group (i.e., 1,3,5,7-tetra, 1,2,3,5,7penta, 1,2,3,5,6,7- and 1,2,3,4,6,7-hexa, and 1,2,3,4,5,6,7-hepta CN), using 2,28,3.38,4,5,68-heptachlorobiphenyl CB#174 Bureau Central de Reference, BCR, EEC) as volumetric standard. Individual CN standards were either obtained as gifts from Prof. Udo A Th Brinkmann, Free University, Amsterdam or synthesized according to a previously published method (Jakobsson et al. 1992). Identification of CN peaks was based on retention time compared to Halowax 1014 as well as a proper isotopic ratio. A method blank was analyzed within each sample series and solvent blanks were run immediately after the highest concentration of standard to check for possible carry over.

Multivariate Data Analysis For PCA, a total of twenty-three target peaks were selected corresponding to the CNs with four to seven chlorine substituents quantified with this method. The labelling of these peaks (Figure 2) is identical with that used for previously published results (Ja¨rnberg et al. 1993). Since the data given in Haglund et al. (1993) were presented with higher chromatographic resolution, some of the peaks reported in that study were added together and converted to make them comparable to the data reported with the older labelling system. The identities of several of the peaks found in Halowax 1014 and environmental samples have been verified using authentic reference substances (Jakobsson et al. 1993; Ja¨rnberg et al. 1994; Asplund et al. 1994; Jakobsson et al. 1994; Nikiforov et al. 1992; Shigeishi et al. 1993; Imagawa et al. 1993). A complete congener specific quantitation is not considered to be achievable at present due to lack of several reference substances and the fact that some congeners remain unresolved on a number of tested GC stationary phases (Imagawa et al. 1993; Williams et al. 1993; Ja¨rnberg et al. 1994; Asplund et al. 1994). In this investigation congener distributions were compared using peak patterns quantified for the samples without reference to relative contribution of individual congeners constituting a peak. PC models were calculated using values of zero, half the detection limit or empty cells (missing value). Using values between zero and half the detection limit resulted in exactly the same PC model, whereas leaving an empty cell in the data set had a dramatic effect on the results. Therefore, a value corresponding to half the detection limit was used for congeners that were reported as not detected. To avoid the problem of differences in absolute concentrations between samples obscuring pattern differences, raw data were normalized as follows; for each sample, the sum of all the quantified peak areas was normalized to the sum of one and each peak concentration expressed as a fraction of the total (Sharaf et al. 1986). Logarithmic normalization was also applied but did not prove to be successful, i.e., the resulting plots yielded no interpretable results. Finally, the data were mean-centered. All calculations were performed with the personal

U. Ja¨rnberg et al.

computer software SIRIUS, version 2.01 (Pattern Recognition Systems Ltd., Bergen, Norway).

Results and Discussion One important goal of this investigation was to establish a background value for PCN in Swedish sediment to be able to compare this with previously found values. Another goal was to provide an overview of CN congener profiles in samples of different geographical origin and pollution situation and to compare these with some potential source samples.

Levels The levels and CN congener composition of the supplementary samples analyzed for this study are given in Table 2. The levels of PCN in sediment samples varied over three orders of magnitude from 0.23 to 270 ng/g (dw 5 dry weight) (Table 1). The sediment from L. Storvindeln appears to be a suitable choice to background site for PCN since the levels found in this sample were the lowest so far, with only 0.23 ng/g (dw) (Table 1). L. Storvindeln is a pristine mountain lake in northern Sweden which may be expected to receive its input of pollutants via long range transport in air. It has been used as a background site within the Swedish environmental monitoring program as well. Compared to this value, clearly elevated levels were found in the surficial sediments at four sites; L. Ja¨rnsjo¨n (41 ng/g, dw), L. Munksjo¨n (18 ng/g, dw), L. Va¨nern (260 ng/g, dw), and L. Bengtsbroho¨ljen (24 ng/g, dw). As indicated in Table 1, these levels may be related to industrial activities at these sites such as processing of recycled paper and chlorine production. The PCN levels found in pike from L. Ja¨rnsjo¨n (360 ng/g, lw 5 lipid weight), L. Va¨nern (210 ng/g, lw), and L. Kyrksjo¨n (130–170 ng/g, lw) were considerably higher than that found in pike from L. Storvindeln (2.6 ng/g, lw), indicating point source pollution at these sites. The levels found in the Baltic sediment sample (7.6 ng/g, dw) were more than ten times higher than that found in the sediment sample from L. Storvindeln and about ten times higher than that found in the sample from the west coast of Sweden outside Gothenburg (0.62 ng/g, dw). The levels of PCN in both guillemot (84 ng/g, lw) and white-tailed sea eagle (120–130 ng/g, lw) egg samples were approximately ten times that found in the Baltic fish (8.4–26 ng/g, lw) samples examined. The otter samples showed much lower PCN levels with only one hexachlorinated congener detected (7.0 and 2.6 ng/g lw). For this reason these two samples were not included in the congener profile comparison study. Similar low levels have previously been found in other mammalian fish feeding species such as common porpoise (Phocaena phocaena) and Baltic grey seal (Halichoerus grypus) (Ja¨rnberg et al. 1993) and may indicate differences in uptake and elimination of these substances between the investigated avian and mammalian species. However, these findings need to be verified with more samples of different species.

Normalized Profiles The environmental samples investigated displayed several different CN congener profiles. From the sediment samples two

Polychlorinated Naphthalene Congeners in the Environment

237

Fig. 2. Labelling of major CN peaks. Total ion chromatogram of GC-MSD analysis of technical mixture Halowax 1014. Peaks marked with an asterisk are known or expected to consist of more than one congener. Column: J&W DB5-MS 60 m 3 0.25 mm 3 0.25 om, program: 90(1.5)–200/(1.5)25–280(6)/2.5 temp.°C(time, min.)/rate °C/min

different types of profiles may be distinguished. One profile contains low to medium chlorinated congeners with a specific ratio of pentaCN that is similar both to technical PCB and PCN mixtures (Figures 3a and b). The other is dominated by higher chlorinated congeners displaying a relatively higher proportion of congeners constituting 5a, 6a, and 7a peaks and is similar to the profile found in graphite sludge from one chloroalkali process. In most of the biological samples the congeners corresponding to peaks 4a, 5a, 5c, and 6a (1,3,5,7-tetraCN, 1,2,4,6,7- and 1,2,3,5,7-pentaCN, 1,2,3,4,6,7- and 1,2,3,5,6,7-hexaCN) were more abundant than the others (Figure 3c). For the guillemot and eagle samples this tendency is particularly pronounced. This finding may indicate that these congeners are amenable to bioaccumulation as well as biomagnification in the investigated fish and piscivorous birds examined as was suggested in Ja¨rnberg et al. (1993). The investigated fish samples however, showed profiles that varied according to their exposure situation and more closely resembled the profile found in their corresponding sediment sample. The fly ash sample displayed some similarities with the chloroalkali sample in that they both showed an increased proportion of the 5a, 6a, and 7a peaks compared to the other peaks. The CN congener profile of the fly ash sample in this investigation is in very good agreement with those found by Takasuga et al. (1994) and Imagawa et al. (1993) for several

investigated MWI samples from similar plants (Stoker type). There is therefore good reason to believe that the CN congener profile included in this investigation is representative of the CN congener profiles generated by MWI plants of this type. In addition to the quantified peaks, several other congeners were abundant in the fly ash sample which were not present or had low abundance in technical PCB and PCN mixtures. According to Imagawa et al. (1993), these are substituted in the 2,3,6,7positions, i.e., 1,2,6,7-, 1,3,6,7- and 2,3,6,7-tetraCN, 1,2,3,6,7and 1,2,4,6,7-pentaCN and 1,2,3,6,7,8-hexaCN. The presence of 1,2,3,6,7,8-hexaCN was also indicated in the graphite sludge sample in this investigation.

Principal Component Analysis The above findings were obvious from a visual inspection of the normalized profiles in Figures 3a–c. To further study the relationships, a first preliminary screening of the entire data set was performed with PCA on normalized data using no weighting of the variables (unscaled). The first three PCs from this analysis explained 70% of the total variance (PC1:34%, PC2:21%, PC3: 15%). The third PC did not provide any additional information that was not obvious from the plot of the first two PCs. The resulting plot of the sample scores for PC1 versus PC2 is displayed in Figure 4.

5.2 12 2.2 1.6 31 36 118 13 n.a. n.a. % % % % % % %

4b

4c

0.73 0.12 0.035 1.3 0.18 0.05 No tetraCN detected No tetraCN detected 0.42 2.6 4.6 0.85 2.7 0.92 0.011 0.03 n.d. 0.42 1.3 0.56 1.3 1.3 0.16 0.36 3.7 0.55 0.026 0.12 0.089 2.6 15 6.5 3.4 19 7 3.9 22 6.4 1.4 23 5.8 0.18 1.8 0.78 n.d. n.d. n.d.

4a

n.d. n.d.

4f

0.022 0.029

4g

0.03 0.03

4h

5a

5b

5c

5d

3 0.18 1.5 0.13 7.2 0.35 2.2 0.17 No pentaCN detected No pentaCN detected 3.1 1.9 0.3 0.46 0.99 0.92 0.22 0.36 2.4 1.9 0.9 0.17 0.73 1.6 1.2 0.16 1.2 0.56 0.014 n.d. n.d. 0.007 0.014 0.018 n.d. 0.01 0.01 0.81 0.11 0.066 0.19 0.54 1.2 0.13 0.54 0.22 0.88 0.39 n.d. n.d. 0.78 6 0.71 2.1 6.2 1.9 3.7 0.27 0.22 0.36 4.9 0.42 0.28 3.3 0.10 0.011 0.011 0.048 0.14 0.027 n.d. 0.048 0.022 6.7 0.92 0.82 5.2 12 6.9 0.48 5.7 2.6 9.6 1.2 1.5 7.4 15 5.8 0.72 4.9 2 16 2.5 2.3 15 24 0.5 0.063 1.3 0.29 9.4 1 1.1 9.7 33 0.7 0.074 2.2 0.43 0.81 0.17 0.17 0.67 2.1 5.7 0.3 4.4 1.3 n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d.

4e

0.024 n.d. 0.032 n.d.

4d

n.a. not analyzed n.d. not detected % Only percentage composition available for this sample

EaNo EaOe MiLo MiHi SBe SMu SSt SBa Gra Mwi WaJa AiA AiH H99 H13 H14 H51

Lip. Cont./ Sample Ignition Identity Loss (%)

Peak ID

0.24 0.2

5f

0.056 0.061

5g

0.044 0.068

5h

6a

0.44 1.6 0.15 0.041 0.42 0.35 1.1 0.69 0.49 0.75 0.94 0.9 1.2 0.17 0.004 0.029 0.012 0.0095 0.019 0.31 0.24 0.23 0.2 0.17 0.54 1.1 0.95 0.66 16 0.27 0.18 0.31 n.d. 0.85 0.039 0.057 0.056 0.086 n.d. 4.1 4.2 5.1 8 0.41 2 2.6 2.8 4.5 0.27 0.58 1.1 0.92 2.9 0.0055 1.2 2.1 1.8 7 0.018 8.7 12 8.9 14 0.62 n.d. n.d. n.d. n.d. 1.3

0.16 0.16

5e

Table 2. Levels of chlorinated naphthalenes in environmental and source related samples (ng/g fresh weight, dry weight, or L)

0.041 0.12 n.d. n.d. 0.38 0.16 0.003 0.074 7 0.16 n.d. 0.54 0.28 0.025 0.051 2.2 0.71

6b

0.019 0.12 n.d. n.d. 0.25 0.23 0.003 0.11 2.8 0.075 n.d. 1.2 0.48 0.061 0.1 7.8 1.3

6c

0.041 0.061 n.d. n.d. 0.2 0.36 0.007 0.062 0.71 n.d. n.d. 1.8 0.8 0.11 0.3 18 0.78

6d

n.d. n.d. n.d. n.d. 0.27 0.071 0.002 0.02 3.6 0.072 n.d. 0.26 0.11 0.016 0.041 2.3 0.51

6e

n.d. n.d. n.d. n.d. 0.24 0.084 0.026 0.14 76 0.15 n.d. n.d. n.d. 0.0015 n.d. 0.85 41

7a

n.d. n.d. n.d. n.d. 0.28 0.072 n.d. n.d. 21 n.d. n.d. n.d. n.d. 0.011 0.024 6.6 54

7b

238 U. Ja¨rnberg et al.

Polychlorinated Naphthalene Congeners in the Environment

239

(a) Fig. 3. Comparison of CN congener profiles by normalized (% on y-axis) data in some selected sediment (a), source-related (b), and biological samples (c). The herring (He6s) and cod (Com) samples may display lower levels of the tetraCN congeners than whould be expected due to losses in the second cleanup step. The labelling on the x-axis corresponds to the labelling in Figure 1. For explanation of sample codes, see Table 1

240

U. Ja¨rnberg et al.

(b) Fig. 3. Continued

The corresponding loading plot, i.e., the relative importance of the different peaks to PC1 and PC2 is also shown. Most of the distribution found in this score plot is explained by differences in chlorination degree between the samples and

this in turn is mainly caused by concentration differences in the 5a, 6a, and 7a peaks. The samples with higher chlorinated CN profiles, Aroclor 1260, Halowax 1051, graphite sludge and the two sediments samples from L. Va¨nern, are found in the

Polychlorinated Naphthalene Congeners in the Environment

241

(c) Fig. 3. Continued

uppermost part of the plot due to an increased abundance of 7a. The samples with lower chlorinated CN profiles, Clophen A30, Aroclors 1016, 1232, Halowax 1099 and 1013, appear to the right of this plot due to a higher abundance of most of the tetrachlorinated congeners.

The profiles from the Baltic guillemot samples and to some extent also the pike samples from Kattfjorden, L. Va¨nern have a higher abundance of the 6a peak which displaces these groups to the left of the plot. The eagle profiles appear from this plot to be more similar to those of most of the fish samples with an

242

U. Ja¨rnberg et al.

Fig. 4. Score plot of first two principal components from PCA of normalized data on all samples and all variables. The corresponding loading plot is inset in the left, lower corner. See Table 1 for explanation of sample codes

Fig. 5. Score plot of first two principal components from PCA of normalized and variance scaled (Autoscaled) data on all samples and all variables. The corresponding loading plot is inset in the right, upper corner. See Table 1 for explanation of sample codes

elevated abundance of 5a placing them closer to the Clophen A50 and Aroclor 1254 than to the lower chlorinated PCB mixtures. The Clophen A50 CN profile appears to be an outlier in this data set. This may be due to the influence of a large number of non detected congeners in this sample. It is difficult from this plot to explore the relationships between the profiles of the samples in the middle region of the plot and in particular to relate these profiles to the sourcerelated profiles, which was one important aim of this investigation. To solve this, it was necessary to reduce the influence of differences in chlorination degree and the impact of the three peaks with the largest variance, i.e., 5a, 6a, and 7a. The latter may be accomplished by scaling the normalized data set using the total variance of each variable (autoscaling to unit variance; Sharaf et al. 1986), so that each variable exhibits unit variance. To reduce the influence of differences in chlorination degree the data set may be normalized so that for each sample, the concentration of each peak, e.g., the 4a, is expressed as a fraction of the total for that homologue group i.e. the tetraCNs. The former procedure will retain some information related to the chlorination degree while the latter will completely eliminate this influence. After the autoscaling was applied, a new PCA was performed on the entire data set. This time, considerably less variance was accounted for by the first PCs (54% for the first four PCs) and information related to the MWI sample was found in PC4, with 4e and 5d peaks being important to distinguish this sample from the others. Using the third PC did not improve the clarity of the model. In the score plot using PC1 and PC2 of this second PCA (Figure 5), the herring and guillemot samples have been

represented by a single sample each in order to simplify the plot. The first PC in this model is still influenced by differences in chlorination degree, as illustrated in the inset showing the loading plot for PC1 versus PC2. Samples showing a high abundance of hexa- and heptaCN congeners are found in the lower left quadrant of the plot, while samples in which the tetraCN levels are elevated, are found in the right lowermost corner (lower chlorinated Aroclors, Clophens and Halowaxes). PC2 is mainly affected by differences in the hexaCN profiles. Samples showing a higher abundance of the 6a congeners, i.e., graphite sludge, guillemot, Aroclor 1260, Clophen A50, and Halowax 1051, are grouped together in the lower left corner of the plot close to the sediment and pike samples from Kattfjorden, L. Va¨nern. Opposite to these, in the uppermost part of the plot, Halowax 1014 appears as an outlier. In this sample the 6d congeners are relatively more abundant than the other hexaCNs. The background samples are found close to the central part of the plot: pike and sediment from L. Storvindeln, burbot from Pajala, and the air sample from Ammarna¨s. In close proximity to these samples the sediment samples from the PCB contaminated lakes (L. Ja¨rnsjo¨n and L. Munksjo¨n) as well as the Baltic sediment sample are found. To the left of the origin are found the remaining biological samples including the pike sample from L. Ja¨rnsjo¨n and L. Kyrksjo¨n and some of the sediment samples from the Gothenburg area. Two sediment samples appear as outliers in this plot as compared to the other sediment samples. One of the sediment samples from Gothenburg harbour (SGo1) showed a higher abundance of the 6b, c, d, and e peaks compared to the other sediments samples. After studying the raw data for this sample, no evidence was found that this could have been due to an

Polychlorinated Naphthalene Congeners in the Environment

analytical artifact. However, the organic content of this sample was considerably lower (3%) than for the other samples from this area (10–13%) which may indicate that this sampling site may not have been an accumulation zone. The other outlier among the sediment samples was the sample from a second chloroalkali site (SBe). This profile showed a higher abundance of the 7b congener. This may be an indication that this sample site was affected by an additional unknown point source or by multiple sources or that other CN profiles are possible in graphite electrode sludge. Differrent binders, such as coal tar, linseed oil and technical PCN have been used in the manufacturing of graphite electrodes and this may be reflected in the CN profiles found in samples from different chloroalkali plants. One of the more important indications of this model was that the technical PCN mixtures Halowax 1099, 1013 and 1014 as well as the MWI sample appeared to be less important to the CN profiles found in the samples from non point-source polluted sites than the technical PCB mixtures Aroclor 1242, 1248, 1254, and Clophen A40. The profiles contained in the centre of the PC versus PC2 score plot may be regarded as constituting a PCB-related domain. Secondly, the only fish samples that showed a different profile were those associated with sediments from one of the sites with a chloroalkali point-source (PiKI, PiKm, SKa and SAn). It thus may be concluded that the profiles found in fish to some extent reflects their exposure situation. In order to investigate if there were any profile differences that were independent of the chlorination degree, PCA was performed on the homologue-normalized data as described above. This approach assumes that if certain congeners of a given homologue group behave differently from the others of that group, so do similarly substituted congeners of other homologue groups. In this PCA, all biological samples were excluded since they were already known to display some enhanced peaks (4a, 5a, c, 6a). From a preliminary PCA, some other samples were also judged to be sufficiently dissimilar from the remaining sample profiles to be excluded from the final version (all the Halowaxes, Aroclors 1016, 1232, 1248, 1254, Clophen A30, A50). The first three PCs of this model accounted for 56% of the total variance (PC:37%, PC2:13%, PC3:7%). In Figure 6, where sample scores for the first two PC:s have been used, the samples are separated into three major categories of profiles. The three samples related to chloroalkali activity (Gra, SAn, SKa), the background sediment samples (SSt and SBa) and Clophen A40 and the MWI sample are separated from the remaining samples along PC1, due the abundance of the ‘‘a’’ peaks. Similarly to Figure 5, this plot distinguishes the CN profile of the SBe sample as different than the other samples and this is seen to be due to an elevated abundance of the 7b peak. From this plot it appears that the CN profiles in the air samples are similar to the CN profiles of Aroclor 1242 and other PCB-related samples. It is interesting to note that the congener responsible for the difference between the background sediment profiles and the remaining sediment profiles is the 7a peak, or 1,2,3,4,5,6,7heptaCN. This congener has previously shown the strongest interaction with activated carbon and with the pyrene HPLC

243

Fig. 6. Score plot of first two principal components from PCA of homologue-normalized data and all variables excluding all biological samples, all PCN technical mixtures and some of the outliers from the two previous plots. The corresponding loading plot is inset in the right, lower corner. See Table 1 for explanation of sample codes

stationary phase (Asplund et al. 1986; Haglund et al. 1990). These findings may implicate that this congener is more strongly adsorbed to particles and thus may be enhanced in background sediments by wet and dry deposition. This phenomenon has been described previously for OCDD by Koester and Hites (1992). The overall conclusions from this investigation are that specific point sources for PCN pollution may be certain chloroalkali plants using graphite electrodes for chlorine production and co-contamination of PCN from accidental PCB pollution due to the recycling of waste paper contaminated with technical PCB formulations. The occurrence of technical PCB formulations exposed to atmospheric weathering are major sources for diffuse spread of PCN to the Swedish environment. Technical PCN formulations are less important sources in this respect, but may play a role in specific cases, such as waste disposal of used electrical equipment.

Acknowledgments. Dr. Peter Haglund at the Institute of Environmental Chemistry, Umeå University is gratefully acknowledged for providing the chromatographic identity of 2,3,6,7-tetraCN, 1,2,3,6,7-, and 1,2,3,5,8-pentaCN.

References Asplund L, Jansson B, Sundstro¨m G, Brandt I, Brinkman UATh (1986) Characterization of a strongly bioaccumulating hexachloronaphthalene. Chemosphere 15:619–628

244

Asplund L, Jakobsson E, Haglund P, Bergman Å (1994) 1,2,3,5,6,7Hexachloronaphthalene and 1,2,3,4,6,7-hexachloronaphthalene— Selective retention in rat liver and appearance in biota. Chemosphere 28:2075–2086 Benfenati E, Mariani G, Fanelli R, Zuccotti S (1991) ‘‘De novo’’ synthesis of PCDD, PCDF, PCB, PCN and PAH in a pilot incinerator. Chemosphere 22:1045–1052 Brinkmann UATh, Reymer HGM (1976) Polychlorinated naphthalenes. J Chromatogr 127:203–243 Cash GG, Breen JJ (1992) Principal component analysis and spatial correlation: Environmental analytical software tools. Chemosphere 24:1607–1623 Crookes MJ, Howe PD (1993) Environment hazard assessment: halogenated naphthalenes. Department of the Environment TSD/ 13, London, Great Britain. Egeba¨ck A-L, Wideqvist U, Asplund L, Ja¨rnberg U, Strandell M, Alsberg T (1993) Organochlorine compounds in air samples from two Swedish locations. Poster presentation at the international symposium ‘‘On the ecological effects of Arctic airborne contaminants,’’ Reykjavik, Island 4–8 October 1993 Engwall M, Brunstro¨m B, Jakobsson E (1994) Ethoxyresorufin O-deethylase (EROD) and aryl hydrocarbon hydroxylase (AHH)inducing potency and lethality of chlorinated naphthalenes in chicken (Gallus domesticus) and eider duck (Somateria molissima) embryos. Arch Toxicol 68:37–42 Greenburg L, Mayers MR, Smith AR (1939) The systemic effects resulting from exposure to certain chlorinated hydrocarbons. J Ind Hyg Toxicol 2(1):2:29 Haglund P, Asplund L, Ja¨rnberg U, Jansson B (1990) Isolation of mono- and non-ortho polychlorinated biphenyls from biological samples by electron-donor acceptor high performance liquid chromatography using a 2-(1-pyrenyl)ethyldimethylsilylated silica column. Chemosphere 20:887–894 Haglund P, Jakobsson E, Asplund L, Athanasiadou M, Bergman Å (1993) Determination of polychlorinated naphthalenes in polychlorinated biphenyl products via capillary gas chromatography-mass spectrometry after separation by gel permeation chromatography. J Chromatogr 634:79–86 Hanberg A, Waern F, Asplund L, Haglund E, Safe S (1990) Swedish dioxin survey: Determination of 2,3,7,8-TCDD toxic equivalent factors for some polychlorinated biphenyls and naphthalenes using biological tests. Chemosphere 20:1161–1164 Imagawa T, Yamashita N, Miyazaki A (1993) Isomer specific analysis of tetra- and pentachloronaphthalene in fly ash and Halowax. J Environ Chem (Kankyo Kagaku) 3(2):221–230 Jakobsson E, Eriksson L, Bergman Å (1992) Synthesis and crystallography of 1,2,3,4,6,7-hexachloronaphthalene and 1,2,3,5,6,7hexachloronaphthalene. Acta Chem Scand 46:527–532 Jakobsson E, Lo¨nnberg C, Bergman Å (1994) Synthesis and crystallography of two hexachloronaphthalenes. Part II. 1,2,4,5,6,8hexachloronaphthalene and 1,2,4,5,7,8-hexachloronaphthalene. Acta Chem Scand 48:891–898 Jansson B, Andersson R, Asplund L, Bergman Å, Litze´n K, Nylund K, Reutergårdh L, Sellstro¨m U, Uvemo U-B, Wahlberg C, Wideqvist U (1991) Multiresidue method for the gas-chromatographic analysis of some polychlorinated and polybrominated pollutants in biological samples. Fres J Anal Chem 340:439–445 Ja¨rnberg U, Asplund L, de Wit C, Grafstro¨m A-K, Haglund P, Jansson B, Lexe´n K, Strandell M, Olsson M, Jonsson B (1993) Polychlorinated biphenyls and polychlorinated naphthalenes in Swedish sediment and biota: Levels, patterns and time trends. Environ Sci Technol 27:1364–1374 Jensen S, Renberg L, Reutergårdh L (1977) Residue analysis of sediment and sewage sludge for organochlorines in the presence of elemental sulphur. Anal Chem 49:316–318

U. Ja¨rnberg et al.

Koester CJ, Hites RA (1992) Wet and Dry deposition of Chlorinated Dioxins and Furans. Environ Sci Technol 26:1375–1382 Kover F (1975) Environmental hazard assessment report: Chlorinated naphtalenes. US Environmental Protection Agency (EPA 560/8-75001). NTIS Publ PB 248–834 Nakano T, Fujimori K, Takaishi Y, Umeda H (1993) Isomer specific analysis of polychloronaphthalenes. Report of the Hyogo Prefectural Institute of Environmental Science, Japan, 25:34–41 Nikoforov VA, Wightman RH, Auger A, Malaiyandi M, Williams DT (1992) Synthesis and characterization of polychlorinated naphthalenes. Poster presentation at Dioxin ’92, 12th International Symposium on Dioxins and Related Compounds, Tampere, Finland, August 24–28, Organohalogen Compounds 8:123–124, and personal communication Oehme M, Manø S, Mikaelsen A (1987) Formation and presence of polyhalogenated and polycyclic compounds in the emissions of small and large scale municipal waste incinerators. Chemosphere 16:143–153 Schlabach M, Biseth A, Gundersen H, Knutzen J (1994) Congener specific determination and levels of polychlorinated naphthalenes in cod liver samples from Norway. Organohalogen Compounds 24:489–492 Schwartz TR, Stalling DL (1991) Chemometric comparison of polychlorinated biphenyl residues and toxicologically active polychlorinated biphenyl congeners in the eggs of Foster’s terns (Sterna fosteri). Arch Environ Contam Toxicol 20:183–199 Sharaf MA, Illman DL, Kowalski BR (1986) Exploratory data analysis. In: Chemometrics, John Wiley and Sons, NY, Vol 82, pp 179–295 Shelley WB, Kligman AM (1957) The experimental production of acne by penta- and hexachloronaphthalenes. AMAArch Dermat 75:689– 695 Shigeishi R, Nikoforov VA, Wightman RH (1993) Varian Instrument Application 22:10 Takasuga T, Inoue, Ohi E, Ireland P, Suzuki T, Takeda N (1994) Development of an all congener specific, HRGC/HRMS analytical method for polychlorinated naphthalenes in environmental samples. Organohalogen Compounds 19:177–182 Theisen J, Maulshagen A, Fuchs J (1993) Organic and inorganic substances in the copper slag ‘‘Kieselrot.’’ Chemosphere 26:881–896 Tirey DA, Taylor PH, Kasner J, Dellinger B (1992) Gas phase formation of chlorinated aromatic compounds from the pyrolysis of tetrachloroethylene. Combust Sci and Tech 74:137–157 Tysklind M, Fa¨ngmark I, Marklund S, Lindskog A, Thaning L, Rappe C (1993) Atmospheric transport and transformation of polychlorinated dibenzo-p-dioxins and dibenzofurans. Environ Sci Technol 27:2190–2197 de Voogt P, Brinkman UATh (1989) Production, properties and use of polychlorinated biphenyls. In: Kimbrough RD and Jensen AA (eds) Halogenated biphenyls, terphenyls, naphthalenes, dibenzodioxins and related products. Elsevier Sci Publ, Amsterdam, pp 161–193 Weistrand C, Nore´n K, Lunde´n Å (1992) Leakage of polychlorinated biphenyls and naphthalenes from electronic equipment in a laboratory. Chemosphere 24:1197–1206 Wenning RJ, Paustenbach DJ, Harris MA, Bedbury H (1993) Principal component analysis of potential sources of polychlorinated dibenzop-dioxins and dibenzofuran residues in surficial sediments from Newark Bay, New Jersey. Arch Environ Contam Toxicol 24:271– 289 Wiedmann T, Ballschmiter K (1992) Halogenated naphthalenes as precursors and products in flame chemistry. Poster presentation at Dioxin ’92, 12th International Symposium on Dioxins and Related Compounds, Tampere, Finland August 24–28, Organohalogen Compounds 9:331–334, and personal communication

Polychlorinated Naphthalene Congeners in the Environment

Wiedmann T, Ballschmiter K (1993) Quantification of chlorinated naphthalenes with GC-MS using the molar response of electron impact ionization. Fres J Anal Chem 346:800–804 Williams DT, Kennedy B, LeBel GL (1993) Chlorinated naphthalenes in human adipose tissue from Ontario municipalities. Chemosphere 27:795–806

245

Wold S, Esbensen K, Geladi P (1987) Principal component analysis. Chemom Intell Lab Systems 2:37–52 Yasuhara A, Morita M (1988) Formation of chlorinated aromatic hydrocarbons by thermal decomposition of vinylidene chloride polymer. Environ Sci Technol 22:646–650