We went to general food markets in tsunami-stricken areas (the Aomori, Iwate, Miyagi, and Chiba prefectures) and purchased a total of 101 fish samples, comprising fat greenling (n = 29), flounder (n = 36), and mackerel (n = 36), in 2012–2013. The localities the fish originated from were confirmed by their labeling and by hearing from the seller that these fish were mostly caught near the above areas. Further informational details will be made available upon request. The fish were transported from the markets under refrigeration and were fileted as soon as possible using a kitchen knife to obtain their muscles. The muscle tissue of each fish was uniformly homogenized in a food processor (GM200, Knife mill GRINDOMIX, Verder Scientific Co., Ltd., Germany), and the homogenized fish samples were stored at –20°C until required for analysis.

Dioxin-analysis-grade n-hexane, acetone, toluene, dichloromethane, and alumina were obtained from Kanto Chemical Co., Inc. (Japan). A multilayer silica gel column filled (from bottom to top) with 2 g of anhydrous sodium sulfate, 0.9 g of silica gel, 3 g of 44% (w/w) sulfuric acid-impregnated silica gel, 0.9 g of silica gel, and 2 g of anhydrous sodium sulfate was purchased from GL Science Inc. (Japan). PCB standards for calibration (TPCB-CSL-A, CS1-A, CS2-A, CS3-A, CS4-A, and CS5-A) and ¹³C₁₂-labeled internal standards (TPCB-CL-A100 and TPCB-SY-A100) were obtained from Kanto Chemical Co., Inc. Retention times were checked by a PCB standard (a mixture of the 209 congeners) prepared by mixing equal amounts of M-1668A-1–0.01x, 2–0.01x, 3–0.01x, 4–0.01x, and 5–0.01x standards (AccuStandard Inc., CT, USA). High-resolution gas chromatography/high-resolution mass spectrometry (HRGC/HRMS) was performed with an HP-6890 Plus gas chromatograph (Hewlette Packard Co., CA, USA) coupled with a JMS-700 MStation mass spectrometer (JEOL, Japan).

The extraction and clean-up procedures of PCBs from fish samples are described in S1 Text. All 209 PCB congeners were determined by HRGC/HRMS with an HT8-PCB capillary column (0.25 mm × 60 m, Kanto Chemical Co., Inc.), as previously described [13], with minor modifications. The samples (2 μL) were injected in splitless mode. The measurements were performed in selected ion monitoring mode at a resolution of >10,000 (S1 Table). For each sample whose ¹³C₁₂-labeled recovery standard concentration satisfied a 40–120% recovery rate, the PCB concentration was calculated by setting PCB congener concentrations lower than the limit of quantitation (LOQ) equal to zero. The results were analyzed statistically using the Mann–Whitney U test and Pearson’s R test.

First, we evaluated the performance of the developed method. A recovery test was performed on a bigeye tuna sample spiked with 62 representative PCB congeners at concentrations of 0.005 ng/g and 0.05 ng/g. The mean recovery ranges of the 0.005 ng/g and 0.05 ng/g spike levels were 73–115% and 76–113%, respectively (S2 Table). The relative standard deviation was below 10% in the sample of 0.05 ng/g spike level. The limit of detection (LOD) and LOQ were conventionally determined by evaluating the signal-to-noise ratio (S/N) of a five-fold dilution of the lowest standard. The LOD and LOQ for congeners that were not included in the standard were determined by the mean of S/N of the congeners with the same number of chlorine atoms. When the congener was detected in the blank sample, the standard deviation (SD) of the analytical results for the corresponding congener in the five-replicated blank sample was calculated, and the LOD (3 × SD) and LOQ (10 × SD) were determined if these values were higher than the values obtained from the S/N. In this study, the LODs and LOQs of the PCB congeners were 0.035–1.0 pg/g and 0.12–3.4 pg/g, respectively (S3 Table). These results suggest that the developed method can adequately determine trace levels of PCB in fish samples.

PCB was detected in all 101 samples, and the concentrations of 109–198 congeners exceeded the LOQ. A histogram of the ∑PCB concentrations (ng/g) is shown in Fig 1. In approximately 90% of the samples, the ∑PCB concentrations were below 15 ng/g. Most of the fat greenling and flounder samples contained up to 5 ng/g ∑PCB, whereas most of the mackerel samples contained 5–10 ng/g. Very few samples contained high levels of ∑PCB, which was consistent with typical concentration patterns of hazardous chemicals in foods [14]. The maximum concentration obtained from a mackerel sample (Table 1), was 86 ng/g, much lower than the provisional regulatory limit for ocean fish in Japan (500 ng/g) [15].

Fig 2 presents box plots of the ∑PCB concentrations in the three fish groups. The ranges of the ∑PCB concentrations between the 25^th and 75^th percentiles were very similar in fat greenling (3.6 g/g and 10 ng/g, respectively) and mackerel (5.4 ng/g and 11 ng/g, respectively). According to reports by the Ministry of the Environment of Japan (MOE) [16–20], the ∑PCB concentrations in fat greenling around the northern areas of Japan (including the investigated areas in the present study) ranged from 1.6 to 16 ng/g (n = 12) before the Great East Japan Earthquake in 2010 and 2011, and from 5.0 to 16 ng/g (n = 8) after the event. Most of the present ∑PCB levels in fat greenling were within the MOE-reported concentrations after the event, and there was no substantial difference in the ∑PCB concentrations between the MOE reports and the present results. The effect of the tsunami on the PCB contamination of marine fish was difficult to evaluate because few data were collected prior to the event in the present study areas. As shown in Fig 2, the ranges of the ∑PCB concentrations between the 25^th and 75^th percentile values were lower in the flounder samples (0.83 ng/g and 2.7 ng/g, respectively) than in the fat greenling (3.6 g/g and 10 ng/g, respectively) and mackerel (5.4 ng/g and 11 ng/g, respectively) samples. In an investigation in 2002, the ∑PCB concentrations in four flounder samples ranged from 0.92 to 19 ng/g, although but the investigation area differed from the present study area [21]. Other types of marine fish yielded higher values; sea eel contained 46–280 ng/g ∑PCB (n = 4), whereas flathead mullet contained 8.3–13 ng/g (n = 4). The 2002 report also calculated the ranges of the ∑PCB concentrations per fat weight: flounder, 150–5,900 ng/g; sea eel, 490–1,600 ng/g; flathead mullet, 1,200–1,700 ng/g [21]. These results suggest that ∑PCB is probably concentrated in the fatty tissues, as reported previously [21, 22].

We further determined the concentrations of the individual chlorinated congeners (S4 Table). Fig 3 shows graphs of the chlorinated congener concentrations plotted against the ∑PCB concentration on a log–log scale. In each case, the chlorinated congener and ∑PCB concentrations were positively correlated (the correlation coefficients are given at the bottom right of each panel in Fig 3). The correlation coefficients were lower for MoCBs, DiCBs, NoCBs and DeCB than for other chlorinated congeners, probably because their concentrations were lower, which would increase the uncertainty in their measurements [23]. In the present investigation, the concentration ratio of each chlorinated congener to ∑PCB was approximately constant, and insensitive to variations in the ∑PCB concentration. Fig 4 shows the concentrations of the chlorinated congeners as a percentage of the ∑PCB concentration. In all samples, the dominant PCBs were tetra- to hepta-chlorinated congeners (TeCBs, PeCBs, HxCBs, and HpCBs) (combined contribution >86%). In particular, the concentration proportions of the chlorinated congeners were as follows: MoCBs, ND–0.16%; DiCBs, 0.046%–2.1%; TrCBs, 1.0%–9.7%; TeCBs, 7.4%–27%; PeCBs, 21%–39%; HxCBs, 23%–43%; HpCBs, 5.9%–25%; OcCBs; 0.66%–5.2%; NoCBs, ND–0.98%; DeCB, 0.015%–2.1%. The pattern of these proportions was quite similar among the samples, although slight compositional differences were observed in some samples such as C-FG6 and A-F6. Both of these samples contained a relatively high percentage of TeCBs; however, it was difficult to determine that these differences were due to the tsunami-induced PCB contamination. We expected that the proportions of the chlorinated congeners in fish would differ significantly between conventional samples and samples freshly contaminated with PCBs. KC 300, 400, and 500, or their mixtures, were widely used as insulation in electrical condensers and transformers in Japan [4, 24]. Especially KC 300 and 400 contain a relatively large proportion of low-chlorinated PCBs (MoCBs, DiCBs, and TrCBs) at more than 15% of ∑PCB [24]. Furthermore, it has been reported that fish metabolize PCBs more slowly than do mammals [25], implying that the proportions would change slowly. If fish were contaminated with high levels of a new contaminant source of PCBs induced by the tsunami, the proportions of the low-chlorinated PCBs in fish should have been higher than those of conventional samples. Before the earthquake, the proportion of the sum of MoCBs–TrCBs to the ∑PCB in fat greenling was reported as 0.91–6.0% [16, 17]. This proportion range is close to that obtained from fat greenling in the present study (1.2–6.9%). In addition, the patterns of PCB proportions in the present investigation resemble those reported in other studies of marine fish around Japan [26, 27]. These findings imply that marine fish obtained from markets in tsunami-stricken areas in our study were unlikely to have been newly contaminated with PCBs as a result of the tsunami. It has been reported that the major congeners in KC 400, 500 and 600 were TrCBs–PeCBs, TeCBs–HxCBs, and HxCBs–HpCBs, respectively [21, 24]. In the present study, contaminating PCBs in the fish consisted mainly of TeCBs–HpCBs, and the PCB congener patterns were similar to those in fish that may have been contaminated with PCBs whose source presumably was a mixture of KC 500 and 600 [26]. In other words, the PCBs detected in marine fish in the present investigation were most likely derived from residual KCs in the environment rather than introduced by the tsunami.

In all samples, the most dominant congener was #153 (comprising 7.2%–18% of the ∑PCB concentration). Because they have high lipophilicity and persistence, the predominance of #153 and #138 in marine fish has previously been reported in Japan [26, 27] as well as in fish in Europe [28–30] although the PCB source of the contamination would be different from KC in Japan. The maximum collective concentration of six indicator PCBs (#28, #52, #101, #138, #153, and #180), which are classified as NDL-PCB and used to monitor ∑NDL-PCB levels in foods in the EU [31], was 28 ng/g in a mackerel sample (Table 2). This value is below the EU regulation limit (75 ng/g for ocean fish) [32]; moreover, the concentration of #52 might have been overestimated because it was incompletely separated from #69 on the chromatogram obtained in the analysis. The sum of these six indicator PCB concentrations, relative to the ∑PCB concentration, was 29%–42% in fat greenling, 26%–40% in flounder, and 24%–36% in mackerel. These proportions do not differ from those in the investigated fish groups in the present study.

Significant positive correlations (r = 0.995) were observed between the ∑NDL-PCB concentrations and the collective concentrations of the six indicator PCBs (Fig 5A). The average concentration proportion of the sum of the indicator PCBs to ∑NDL-PCB was 37% (range 26%–48%), although the European Food Safety Authority noted that these six indicators collectively account for approximately 50% of the ∑NDL-PCB in food [31]. We also calculated the collective concentration of seven indicators (the six indicators plus #118) (Table 2). In all samples, #118 was the most abundant DL-PCB, contributing 3.4–13% to the ∑PCB concentration. As observed in the six-indicator analysis, the collective concentration of the seven indicators was significantly correlated with the ∑PCB concentration (r = 0.996; see Fig 5B). The average concentration proportion of the sum of seven indicators to ∑PCB was 39% (range 29%–50%). Thus, we consider that the ∑PCB concentration can represent the net concentration of the specified congeners, and vice versa.