PCB pollution continues to impact populations of orcas and other dolphins in European waters

Organochlorine (OC) pesticides and the more persistent polychlorinated biphenyls (PCBs) have well-established dose-dependent toxicities to birds, fish and mammals in experimental studies, but the actual impact of OC pollutants on European marine top predators remains unknown. Here we show that several cetacean species have very high mean blubber PCB concentrations likely to cause population declines and suppress population recovery. In a large pan-European meta-analysis of stranded (n = 929) or biopsied (n = 152) cetaceans, three out of four species:- striped dolphins (SDs), bottlenose dolphins (BNDs) and killer whales (KWs) had mean PCB levels that markedly exceeded all known marine mammal PCB toxicity thresholds. Some locations (e.g. western Mediterranean Sea, south-west Iberian Peninsula) are global PCB “hotspots” for marine mammals. Blubber PCB concentrations initially declined following a mid-1980s EU ban, but have since stabilised in UK harbour porpoises and SDs in the western Mediterranean Sea. Some small or declining populations of BNDs and KWs in the NE Atlantic were associated with low recruitment, consistent with PCB-induced reproductive toxicity. Despite regulations and mitigation measures to reduce PCB pollution, their biomagnification in marine food webs continues to cause severe impacts among cetacean top predators in European seas.

Scientific RepoRts | 6:18573 | DOI: 10.1038/srep18573 1970s-1980s and some terrestrial wildlife populations recovered [1][2][3] . Experimental studies have established that OC pesticides and the more persistent polychlorinated biphenyls (PCBs) have a range of species-specific and dose-dependent toxic effects such as immunosuppression and reproductive impairment in all mammalian species tested, including marine mammals [4][5][6][7] . However, the impact of historic and current exposure to marine pollutants on European cetacean top predators is still largely unknown.
To investigate this issue we conducted a meta-analysis of European data on summed 18-25 chlorobiphenyl congeners (∑PCB) concentrations (mg/kg lipid weight (lw)) in cetaceans that were stranded and necropsied or free-living and biopsied. Data came from 1,081 animals of four species:-harbour porpoise (Phocoena phocoena, HP)(n = 706), striped dolphin (Stenella coeruleoalba, SD)(n = 220), bottlenose dolphin (Tursiops truncatus, BND) (n = 131) and killer whale (Orcinus orca, KW)(n = 24). The ∑PCB was mainly derived from two different laboratories, but with both using internationally standardised methodology enabling full integration of data. The analysis included statistical evaluation of long-term temporal trends in ∑PCB concentrations in UK-stranded HPs  and Mediterranean SDs (1990SDs ( -2009. We compared mean ∑PCB concentrations in all four species with a range of established PCB toxicity thresholds for marine mammals [5][6][7] . The lowest PCB toxicity threshold was a widely used and relatively low PCB toxicity threshold for onset of physiological impacts and equivalent to 9 mg/kg lipid as ∑PCB 6 in this study. We also used the very highest PCB toxicity threshold reported for marine mammals, for profound reproductive impairment in ringed seals (Phoca hispida) in the Baltic Sea, and equivalent to 41mg/kg lipid as ∑PCB in this study 6 .
We compared mean PCBs concentrations in UK-stranded HPs that were "healthy" and died of acute physical trauma (n = 345) (control group) with mean PCBs concentrations in HPs that died of infectious diseases (n = 252) (case group) or starvation (n = 42). Where data permitted, sex differences in mean PCB concentrations were investigated to assess maternal offloading of PCBs via lactation. Finally, we reviewed the current population and reproductive status of the extant odontocete populations in European waters using systematic local and international observational survey and other data collected over the past 15-20 years (e.g. International Union for Conservation of Nature assessments) 8 .

Results
Spatial and temporal trends and sex differences in ∑PCB exposure. The temporal trend in ln ∑PCB lipid concentrations in UK HPs  and SDs from the western Mediterranean Sea (1990Sea ( -2009 is shown in Fig. 1A,B. The temporal trend in both populations was highly statistically significant (p < 0.001). In UK-stranded HPs, ln ∑PCB concentrations declined slowly from 1990 to 1998 and then remained relatively stable from 1998 to 2012 against the null hypothesis of no trend (p < 0.001, F = 11.76, residual df = 701.97, trend df = 3.03) (Fig. 1A). In Mediterranean SDs, ln ∑PCB concentrations showed a marked decline from an initial peak in 1990, but then stabilized from 2003 to 2008, but with blubber ∑PCB concentrations still consistently exceeding all mammalian toxicity thresholds. The trend is statistically significant (p < 0.001, F = 55.45, residual df = 212.03, trend df = 6.97) against the null hypothesis of no trend (Fig. 1B).
We compared ∑PCB concentrations among regions within the NE Atlantic and Mediterranean Sea. Since our ∑PCB data is lognormal, we are able to use the sample geometric mean to estimate the population median (Tables 1  and 2), and that this estimator is unbiased for doing this 10 . Mean ∑PCB concentrations and geometric means (with upper and lower 95% confidence intervals) were generated for males and females of each cetacean species and region for all available data (n = 1,009) ( Table 1) and for adults only (n = 401) ( Table 2). These tabulated data were presented along with minimum and maximum values (i.e. range) for males and females for each species and region (Tables 1 and 2). In general, the ∑PCB sample geometric means (as estimator of population median values) were 25-65% lower than the (arithmetic) means for the same species, but still exceeded even the very highest marine mammal PCB toxicity threshold (∑PCB = 41 mg/kg lipid) for some BND, SD and KW groups in both NE Atlantic and Mediterranean Sea regions (Tables 1 and 2).
For males and females (all ages), high mean ∑PCB concentrations (> 100.0g/kg lw) were found for BNDs off Iberian Peninsula (NE Atlantic), western Mediterranean and northern Adriatic Sea; for SDs in western Mediterranean and for KWs (NE Atlantic) (Fig. 2). Among adults, high individual female ∑PCB concentrations (> 100mg/kg lw) were found in KWs from all NE Atlantic regions, in BNDs from England and Wales, and in SDs from the western Mediterranean Sea (  The toxicological data show that these populations greatly exceed concentrations at which severe toxic effects are known to occur. Pathological findings on necropsy consistent with immunosuppression and increased susceptibility to disease included macro-parasitic and bacterial pneumonias, high lung and gastric macro-parasite burdens, and generalised bacterial infections (septicaemias). These were regularly found on necropsy in stranded HPs 5 , BNDs and KWs. In Mediterranean SDs distemper due to cetacean morbillivirus (CeMV) infection 9 was frequently seen. Several stranded KWs also had multiple dental infections leading to large mandibular abscesses identified on necropsy 12 . High ∑PCB contamination can cause immunosuppression 7 and may be a significant contributing factor in the death of many of the stranded individuals that had fatal infectious diseases on necropsy.
∑PCB lipid concentrations in subgroups of "healthy" male UK-stranded HPs that were generally in good nutritive condition and died of acute physical trauma (sample mean = 13.8 mg/kg lw; sample median = 9.57 lw; n = 201) had significantly lower ∑PCB than male HPs that were in poorer condition and died of a range of infectious diseases (sample mean = 26.5 mg/kg lw; sample median = 18.9 mg/kg lw; n = 120) (ANOVA, P < 0.001). In female UK-stranded HPs that died of acute physical trauma, ∑PCB lipid concentrations (sample mean = 8.50 mg/kg lw; sample median = 6.41 mg/kg lw; n = 144) were significantly lower than female HPs that died due to a range of infectious diseases (sample mean = 16.8 mg/kg lw; sample median = 12.2 mg/kg lw; n = 132) from the same period (ANOVA, P < 0.001) (Fig. 2). Male and female HPs that died of starvation were all in a poor or emaciated condition. Both male HP starvation cases (sample mean = 20.7 mg/kg lw; sample median = 12.5 mg/kg lw; n = 26) and female HP starvation cases (sample mean = 27.9 mg/kg lipid; sample median = 9.74 mg/kg lw; n = 16) had higher ∑PCB concentrations than the physical trauma group. For UK-stranded HPs (both sexes), mean ∑PCB mg/kg lipid was 150-328% greater in infectious disease or starvation cases as compared to trauma cases. The increase in ∑PCB on a sample median basis was 130-198% greater for HP disease or starvation cases as compared to HP trauma cases. Population Dynamics. Such sustained and elevated PCB burdens are likely to have significant effects at the population level in European BNDs, SDs and KWs. Although we cannot directly and causally link high ∑PCB exposures to cetacean population declines, many of these populations are either very small, or show evidence of major and long-term declines or a significant contraction of range. Only very small KW populations are now found in industrialised regions of Europe [13][14][15] . There is only one resident population in southern Europe 14 : a tuna-feeding KW subpopulation in the strait of Gibraltar comprising two pods totalling 36 individuals 13 . These animals have been monitored every year since 1999. Only six mature females were reported to have calved between 1999 and 2011 and these produced only five calves living more than 1 year. This 6.4% annual female fecundity rate over a 13-year period is one of the lowest recorded reproductive rates for KWs globally and the subpopulation was recommended for listing as "Critically Endangered" due to small population size 14 . Around NW Scotland and Western Ireland, a small group of only nine KWs is regularly seen 15 . This group of marine-mammal-eating KWs comprises four adult males, two adult females and three adult female or sub-adult males. Although studied over a 19-year period, no calves have ever been reported within this group 15 . No other resident or coastal KW groups are found off the Iberian Peninsula or in the Bay of Biscay, although occasional offshore KW sightings occur (e.g. 16 ). Long-term cetacean stranding records in the Netherlands show that KWs were first recorded on the Dutch North Sea coastline in 1783, observed regularly from 1918 to 1963 and completely ceased thereafter until two single KW strandings in 2009 and 2010 17 .
Historic strandings data suggest that multiple BND resident or coastal groups in Europe became depleted or locally extinct in the late-1960s to mid-1970s, including those in the UK (e.g. Morecambe Bay; East coast of England) 18 and the North Sea Dutch coast 19 . The last member of a resident BND population at Arcachon, France, died in 2003 20 and the small resident BND group (current census n = 27) in the Sado Estuary, Portugal, showed decline due to low calf survival over several decades 21 . Blubber PCB levels for two females and one male BND (sampled between 1995 and 1997) from this population ranged from 37.1 to 114.0 mg/kg lw. The Mediterranean Sea remains a global PCB hotspot 1 where most of the extant marine mammal species, including SDs, BNDs and short-beaked common dolphins (CDs)(Delphinus delphis) have declined over decades 14,22 . The current IUCN Red List conservation status of BNDs and SDs in the Mediterranean Sea is "Vulnerable". The Mediterranean CD subpopulation is now classified as "Endangered" after experiencing major and generalized decline over the last 40-50 years, particularly in the northern Adriatic Sea and in the eastern Ionian Sea 8 . Blubber PCB concentrations are significantly higher in CDs and SDs in the Mediterranean Sea compared to the much more abundant NE Atlantic populations 23 .
For HPs in the NE Atlantic, a single continuous population ranges from France to northern Norway. A separate much smaller Iberian population is estimated to comprise only 4,398 (CV = 0.92) porpoises 24 . In the Baltic Sea, two separate HP subpopulations exist: one in the western waters (Kattegat, the Belt Seas) estimated at 40,475 individuals (CV = 0.235) and listed as "Vulnerable" (by HELCOM) 25 , and another in the Baltic Sea proper, which is now very small and has suffered decades of decline and is listed as "Critically Endangered" 8    Collectively, these necropsy and fisheries bycatch observer studies from approximately 1990 onwards indicate that bycatch is unlikely to be driving population declines of BNDs or KWs around the UK or in the wider NE Atlantic during this period.
The potential impacts of anthropogenic high-intensity acoustic sources on cetaceans include mass stranding events (MSEs) that have occurred on a global basis, predominantly involving beaked whales exposed to mid-frequency active sonars in naval exercises [34][35][36] . Non-beaked whale MSEs with a probable acoustic cause have occurred, such as a 2008 CD MSE in the UK, but are very rare in European waters 32 . No acoustically-induced MSEs were recorded in HPs, SDs, BNDs or KWs in European waters. Acoustically-induced cetacean MSEs have been recorded in the Mediterranean Sea but also primarily involved beaked whales 36,37 .
In the Mediterranean and Black Seas, all cetacean species where population assessments have occurred are considered to be declining 8,14,22 . A range of threats has been proposed including hunting (historically); bycatch; overfishing/prey depletion; habitat degradation; morbillivirus infection and chemical and acoustic pollution 8,14,22 . Empirical evidence for the effects of prey depletion (e.g. due to overfishing or climate change effects) leading to nutritional limitation is limited for cetaceans, but diet/nutrition effects have been linked to recruitment in KWs and fin whales (Balaenoptera physalus) [reviewed in 26 ]. Such processes would potentially increase PCB toxicity in malnourished animals by mobilising lipophilic PCBs from blubber to other body compartments 7,23 . A recent global review of Cetacean Morbillivirus (CeMV) found that SDs have been negatively impacted by CeMV epizootics

Discussion
The stabilised temporal trends in blubber PCB concentrations in cetaceans in Europe contrast with other industrialised northern hemisphere regions, such as North America, where PCBs have continued to decline in BND 41 and KW 11,23 populations. PCBs were banned in closed systems in the USA in 1979 1 and in 1981 in the UK 4 , but were not phased out until 1987 in European countries that border the Mediterranean Sea 1,42 . Mean ∑PCB concentrations in adult female KWs in our study substantially exceeded concentrations in both southern resident and transient female KWs off British Columbia 11,23 . Where females are reproducing normally, blubber PCB concentrations are known to fall significantly with successive pregnancies due to maternal offloading through gestation and lactation 7,41,42 . The relatively high ∑PCB concentrations in many adult female KWs in this study is consistent with reproductive failure in at least some individuals, due to either adult females failing to offload PCBs to calves during pregnancy and lactation 7 or re-accumulation of dietary PCBs in some post-parturient and reproductively senescent KWs 11 . Mean blubber ∑PCB concentrations in male KWs in this study were higher than male "southern resident" KWs off British Columbia (NE Pacific) in the 1990s and were only slightly lower than male "transient" (marine mammal eating) KWs from the same region 11,41 . Very high mean PCB concentrations have been recorded in blubber biopsies of adult male (n = 4) marine mammal-eating West Coast "transient" KW ecotype off California, USA (mean ∑PCB = 630 mg/kg lipid) 43 ; in individual stranded KWs in Oregon State 23 ; and in a transient adult female that stranded in Washington State in 2002 with ∑PCB = 1,100 mg/kg lipid 44 . For BNDs, only a population in a highly polluted Superfund site in Brunswick, Georgia (USA) exposed to a major point source of PCB pollution with Aroclor 1268 45 had mean exposures similar to many BND groups in our study. HP blubber ∑PCB concentrations in this study were similar to those associated with immunosuppression and infectious disease mortality in previous multivariate studies of UK-stranded HPs (statistically independent of age, sex, sexual maturity and two quantitative indices of nutritional status) 5 and similar to those associated with impaired reproduction in adult female HPs in UK waters 26 .
A large number of factors can influence PCB exposure and subsequent PCB concentrations in blubber including, but not limited to, blubber sample condition, diet, degree of maternal offloading, year of sampling, season, age, gender, location, species, body condition, and whether or not an animal was sampled in good health 1,4,7 . Some of these we can control for statistically as potentially confounding factors, others we cannot. Nevertheless, for SDs, BNDs and KWs around the Iberian Peninsula (both NE Atlantic and Mediterranean regions), mean and median ∑PCB concentrations levels were among the highest recorded for cetaceans globally, regardless of sample type/ nutritional condition/sex/year of sampling, etc. In terms of blubber sample quality, we are able to obtain 954/1081 ( = 88.3%) fresh or slightly decomposed stranded or biopsied blubber samples, effectively negating the potential bias of sample degradation on PCB determination. Nor did we see any significant overall differences in mean ∑PCB concentrations between stranded and biopsied BNDs, SDs or KWs. This shows that the ∑PCB concentrations in stranded animals were not fundamentally biased in comparison to the free-living biopsied cetaceans.
In relationship to nutritional status, we can show that both male and female HPs that died of trauma had the lowest ∑PCB concentrations and this increased by 190-198% for infectious disease cases and by 130-328% for starvation cases. Although we lack data to test this "nutritional loss" effect rigorously on ∑PCB concentrations in SDs, BNDs and KWs, we can show that the excessively high ∑PCB concentrations occurred in virtually all individual SDs and BNDs around the Iberian Peninsula and for virtually all individual KWs across the NE Atlantic region, irrespective of blubber sample quality, year of sampling, season, gender, nutritive condition and whether or not an animal was sampled in good health.
There are also marked species-specific variations in susceptibility to PCB exposure. The lower PCB toxicity threshold used in this study was predominantly derived from non-cetacean species such as seals, otters, and mink 7 . It is well recognized that mink are one of the most PCB-sensitive species and so it is highly possible that cetaceans are less sensitive to PCB exposure than species like mink. If so, the use of the Kannan threshold 7 is likely to overestimate the true PCB risk to cetaceans. However, this lower PCB toxicity threshold is equivalent to only 9 mg/kg lipid (as ∑PCB) -whereas many BND, SD and KW groups in this study had one order of magnitude greater mean or median ∑PCB concentrations (50-350 mg/kg lipid). So, even if the lower PCB toxicity threshold does significantly over-estimate the likely toxicity of PCBs on cetaceans -the very high ∑PCB exposures recorded empirically in European cetaceans in this study (> 1,000% greater than 9 mg/kg lipid) still provides compelling evidence for the inherent PCB toxicity risk in this study.
Current threats to cetaceans from persistent organic pollutants (POPs) in Europe appear to be restricted to PCBs. Marked and ongoing declines in tissue concentrations of organochlorine pesticides (e.g. DDTs) have occurred in UK-stranded HPs 4 and in western Mediterranean SDs 3 and BNDs 42 . Similar marked declines in tissue contaminants have occurred in UK HPs for butyltins 46 and brominated flame retardants following a 2004 EU-ban 4 . Other newer chemical compounds including the alternative brominated and chlorinated flame retardants such as tetrabromo-p-xylene (TBX), tetrabromo-o-chlorotoluene (TBCT) and 2,3-dibromopropyl-2,4,6-tribromophenyl ether (TBP-DBPE) and Dechlorane Plus (DDC-CO) isomers 47 , perfluoroalkyl substances (PFASs) such as perfluorooctane sulphonate (PFOS) and perfluorooctanoic acid (PFOA) in 48 and emerging organophosphorus flame retardants (PFRs) and plasticisers 49 were detected in some UK-stranded HP samples, and with the possible exception of PFOS, were at either very low levels or occurred below levels of analytical detectability.
Many European seas (e.g. Baltic, Mediterranean and North Seas) are surrounded by highly industrialised land masses with high human population densities which may at least partly explain the high PCB exposures we have detected 1 . Mass production of PCBs commenced in 1929, reaching peak levels in the 1960s and 1970s, when many European cetacean populations already appear to be declining. Nearly 97% of the historical use of PCBs occurred in the northern hemisphere 50 , with the marine environment acting as the ultimate PCB "sink". By the late 1990s, however, only 1% of globally manufactured PCBs were estimated to be in sea water, with 6-7% in marine sediments (reviewed in 1 ), while as of 2005, 1.1 million tons of PCB contaminated material still required disposal by EU Member States, most notably France and Spain 51 . PCB effects are likely to be particularly pernicious in highly exposed European BND, SD and KW populations by markedly suppressing reproduction and probably subsequent recruitment at individual and population levels. EU regulations to mitigate PCB pollution currently appear insufficient to protect cetacean top predators in the NE Atlantic and Mediterranean Sea. As sentinels of marine ecosystems, monitoring PCB concentrations in marine mammals should be considered for inclusion as an indicator under descriptor 8 within the European Marine Strategy Framework Directive (MSFD) (Directive 2008/56/ EC). Our results indicate that European marine mammal habitats are unlikely to meet Good Environmental Status within the MSFD by 2020, and that several dolphin species are unlikely to achieve Favourable Conservation Status now under EC Habitats Directive (Council Directive 92/43/EEC). Further steps to reduce PCB inputs into the European marine environment should include stricter controls on disposal of PCBs, e.g. in buildings with sealants containing PCBs 4,52 . Also, measures should be taken to limit PCB bioavailability to marine food webs, such as improved management of dredging of harbours/ports, as well as improved containment of PCB environmental contamination from industrial buildings/equipment and domestic waste disposal (e.g. landfill).
In conclusion, this pan-European meta-analysis of stranded or biopsied cetaceans demonstrates that several European cetacean species, specifically BNDs, SDs, and KWs, currently have markedly elevated blubber PCB concentrations. Particular "PCB hotspots" included the western (SDs and BNDs) and central (BNDs) Mediterranean Sea and SW Iberia, the Gulf of Cadiz (BNDs) and the Strait of Gibraltar (BNDs and KWs). Despite an EU ban on the use and manufacture of PCBs in the mid-1980s, blubber PCB concentrations are still very high, possibly having reached a "steady state" between environmental input and degradation, meaning that high PCB exposures are set to continue for the long-term in cetacean top predators in Europe. These high and stable PCB exposures are associated with small populations, long-term population declines or contraction of range in several dolphin species in Europe (NE Atlantic and Mediterranean Seas) that were not adequately explained by other factors (e.g. bycatch or other anthropogenic causes of mortality). Bycatch is common in the most abundant cetacean species in Europe, but is comparatively rare in BNDs and virtually unrecorded in recent years for KWs, suggesting that the ongoing population declines in these two species are predominantly driven by other processes, with bioaccumulation of PCBs through marine food chains being the predominant factor. A lack of recruitment in monitored KW and BND populations is also consistent with PCB toxicity as the likeliest cause of their declines. In the Mediterranean Sea, the SD has suffered recurrent CeMV mortalities, which may have been exacerbated by the high and immunotoxic level of PCB exposure. Without significant mitigation, PCBs will continue to drive population declines or suppress population recovery in Europe for many decades to come. Measures to significantly reduce inputs of PCBs into the marine environment from terrestrial and other sources are urgently needed. Further studies are also needed to better assess PCB exposure and quantify toxic effects in marine apex predator populations in Europe. Finally, the potential impact of PCB bioaccumulation in marine ecosystems may extend beyond European waters, particularly in globally distributed marine apex predators such as KWs, false killer whales (Pseudorca crassidens) and great white sharks (Carcharodon carcharias).

Methods
Between January 1990 and December 2012, HPs (n = 706), SDs (n = 220); BNDs (n = 131) and KWs (n = 24) in NE Atlantic and Mediterranean Sea regions of Europe were necropsied (HP n = 706; SD n = 119; BND n = 93; KW n = 9) or biopsied (SD n = 101; BND n = 38; KW n = 15) for skin and blubber samples using standardised methodologies 5,9 . Necropsies of stranded cetaceans (n = 929). A total of 929 cetaceans were sampled or necropsied after stranding on a beach or being found dead at sea. These comprised 706 HPs; 121 SDs 93 BNDs and 9 KWs. In the UK, stranded or bycaught HPs (n = 706), BNDs (n = 38) and KWs (n = 7) were necropsied using methods based on a standardised European necropsy protocol 53 . Necropsies were also conducted on stranded SDs (n = 119) from western Mediterranean, BNDs from Galicia, Spain (n = 11) and Portugal (n = 5) and a single KW stranded in the Strait of Gibraltar, Spain using similar methods. Wherever possible, stranded cetaceans in fresh or slight decomposition were prioritized for necropsy to limit decomposition of tissue samples including blubber. Fresh or slightly decomposed carcasses comprised 802/929 (86.3%) of all necropsies and included 660 HPs; 103 SDs; 33 BNDs and 6 KWs. Moderately decomposed carcasses accounted for 72/929 (7.75%) of necropsied carcasses including 37 HPs; 21 BNDs; 11 SDs; and 3 KWs. Only 18/929 (1.94%) carcasses were in advanced decomposition comprised 8 HPs; 5 BNDs and 5 SDs. State of decomposition could not be determined for 37/929 (3.98%) carcasses comprising 34 BNDs; 2 SDs and 1 HPs. Sex was determined on necropsy. Sex was not recorded or could not be determined (e.g. due to loss of gonads from scavenging animals) for 65 SDs and 6 BNDs. Causes of death in UK-stranded animals were determined by specific and established diagnostic criteria 5 . Additional chlorobiphenyl congener and other data were included from a single KW that stranded at Roches Point, Cork harbour, Ireland and was sampled on 8 th July 2001 12 .
Blubber samples from biopsied cetaceans (n = 152). Blubber biopsies of free-living SDs from the western Mediterranean Sea (n = 101), BNDs from the Gulf of Cadiz (n = 11), Strait of Gibraltar (n = 11) and northern Adriatic Sea (n = 6), and skin/blubber biopsies of KWs from Canary Islands (n = 8) and Strait of Gibraltar (n = 7) were obtained using a range of standardised techniques including use of a cross-bow 54 . For BNDs and SDs in the western Mediterranean Sea, the skin and blubber tissue biopsy was excised from bow-riding dolphins using a sterile biopsy dart of the butterfly valve type, shot with a spear gun or a compressed air pistol from a boat, a non-destructive technique commonly used in cetacean research 54 . Darts were aimed at the region posterior to the dorsal fin. The precise age or sex of the individual samples was unknown. The samples collected contained about 0.8 g of blubber. Once collected, the skin and blubber biopsy samples were wrapped in aluminum foil and preserved at − 20 or − 80C until analysis. Sex of biopsied animals was determined using molecular techniques 55 . Additional chlorobiphenyl congener and other data were included from BNDs (n = 8) biopsied in the Shannon estuary, Ireland in September 2000 56 . Assessment of sexual maturity in cetaceans. Sexual maturity was determined in necropsied animals by analysis of gonadal material 57 . For those individuals where sexual maturity status data based on gonadal assessments were unavailable, published data on age and length at both sexual and physical maturity were used as criteria for assessing maturity status -BNDs 58   Cefas data. The blubber samples at the Cefas lab were analysed for 25 individual chlorobiphenyl congeners (IUPAC numbers: 18,28,31,44,47,49,52,66,101,105,110,118,128,138,141,149,151,153,156,158,170,180,183,187,194) (sum25CBs mg/kg lipid) using internationally standardised methodology 4,5 . After thawing, the homogenised subsamples (~5g) were dried by mixing with anhydrous sodium sulphate and storing in a freezer for a minimum of 12 hours prior to further analysis. The samples were subjected to Soxhlet extraction using of acetone: n-hexane 1:1 (v:v) for 5.5 hours. The total extractable lipid content was determined gravimetrically after evaporation of the solvent from an aliquot of the uncleaned extract. Depending on the lipid content of the samples, varying volumes of the biota extracts were cleaned to have ~50 mg of lipid in the samples for PCB analysis. For marine mammal blubber samples, this was usually 1mL out of the 100mL extract.
For Quality Assurance and Quality Control the Cefas laboratory participates biannually in proficiency testing scheme Quasimeme (Quality Assurance of Information for Marine Environmental Monitoring in Europe) as external quality assurance .All analyses were carried out under full analytical quality control procedures that included the analysis of certified reference material(s) and a blank sample with every batch of 10 samples analysed so that the day-to-day performance of the methods could be assessed. If levels of target analytes in the samples were outside of the range of the instrument calibration, extracts were diluted to be within range and re-analysed. Reference material used was BCR349 (cod liver oil; European Bureau of Community reference), which has been used in the Cefas lab since 1993. The results obtained for the reference materials were plotted as Shewhart quality control charts for each compound determined. The charts had previously been created by the repeated analysis of the above certified reference materials in the Cefas Lowestoft Laboratory using the North West Analytical Quality Analyst software ™ (Northwest Analytical Inc., USA). Warning and control limits had been defined for the charts as 2σ and 3σ -2x and 3x the standard deviation from the mean for each compound. The results obtained for all samples analysed were accepted as valid as the results for the certified reference materials were within the limits set by the control charts.
IRBio data. Lipid 42,62 . Approximately 1 g of tissue was ground with anhydrous sodium sulphate using a mortar. The mixture was extracted with n-hexane for 4 h in a Soxhlet apparatus with a capacity of 125 ml. The solution obtained was concentrated to 40 ml. A portion of this extract (10 ml) was used to gravimetrically determine the quantity of extractable fat per gram of blubber. A further quantity was mixed with sulphuric acid for the purpose of cleaning it, and the resulting extract was concentrated to 1 ml and centrifuged for five minutes.
A sample volume of 0.5 μ L was injected in splitless mode (1 min splitless vent time) into a Thermo Scientific ITQ 900 GC/MS system equipped with a 30-m DB-5 MS (95% metylsilicone and 5% phenyl) fused-silica capillary column with an ID of 0.25 mm and a film thickness of 0.25 μ m, with helium as the carrier gas at constant flow of 1 ml min −1 . Injection port and transfer line temperatures were 270 and 300 °C, respectively. The oven temperature program was as follows: 60 °C for 1 min, an increase to 315 °C at 6 °C min −1 , and a final hold of 20 min. Source temperature was 200 °C. Mass spectrometry was performed in electron impact mode. Full-scan spectra were acquired over the m:z range 50-550.
The total PCB concentration (tPCB) was calculated as the sum of the 23 congeners known as IUPAC# 95 ,  101, 136, 110, 151, 144, 149, 118, 153, 141, 138, 187, 183, 128, 174, 177, 202, 171, 180, 170, 201, 203 and 195. Identification and quantification of the individual compounds were performed by comparison with external reference standards calibrated with a 6-point calibration curve encompassing the entire concentration range. The detection limits were between 0.1 to 1 μ g kg −1 for the PCBs standards. Analyses were performed in a series of 10 samples and 1 blank. The recoveries of the PCBs were calculated by adding known amounts of a standard to 12 homogenised replicates of the same sample; recovery levels ranged from 82-101%.
Marine Institute (Galway) data. The Marine Institute, in Galway, Ireland examined blubber biopsies collected in 2000 from 8 BNDs from the Shannon estuary, Ireland [see 56 for Methods].

Data integration
A conversion factor [0.9] was generated to integrate the different sets of PCB data (Cefas, IRBio and Marine Institute) using a statistical analysis of the 13 CB congeners common to both methods [i.e. sum25CBs = sum18CBs*0.9]. The combined and converged PCB data was referred to as "sum18-25CBs (mg/kg lipid)" or \xAD\xF4PCB (Supplementary Dataset).
Spatial and temporal trends in ∑PCB exposure. To assess the temporal trends in HP and SD data we fitted a Generalised Additive Model (GAM) to the data using the R package mgcv 63 . We smoothed the data using thin plate regression splines and the degree of smoothing was determined by the integrated generalised cross validation function 64 . The smoothing parameter gamma was set at 1.4 for HPs and set at 2.0 for SDs. To assess the spatial distribution of HP and SD data we used ∑PCB data from 1996-2012 showing spatially smoothed mean values for the ∑PCB data. This period was chosen to maximise data most representing current level of ∑PCB exposure in these species based on their temporal trends. Figures showing the spatial distribution of ∑PCBs (lipid weight) were produced in Esri ArcMap 10.1 (www.esri.com). Data points are shown along with local averages. These averages were calculated by kernel smoothing using a polynomial order 5 kernel with power = 0, ridge parameter = 50 and bandwidth based on the spatial distribution of the observations for each species: bottlenose dolphin 0.75 degrees; harbour porpoise 0.5 degrees; killer whale 1.2 degrees; striped dolphin 0.5 degrees.

Comparison of ∑PCBs concentrations in relation to marine mammal toxicity thresholds. Two
PCB toxicity thresholds were used in this study. A lower PCB toxicity threshold was used for the onset of physiological endpoints in marine mammals of 17 mg/kg lipid weight (lw) (as Aroclor 1254) 7 , that was calculated to be equivalent to 9.0 mg/kg lw (as ∑PCB) in this study. A higher PCB toxicity threshold, the highest reported in marine mammal toxicology studies, of 77 mg/kg lw (as Clophen 50) for reproductive impairment in Baltic ringed seals (Pusa hispida) 6 was calculated to be equivalent to 41 mg/kg lw (as ∑PCB) in this study. Differences for sex and cause of death in mean ∑PCBs concentrations in HPs only were investigated using univariate analysis of variance (ANOVA). Since only 4 months of ∑PCB data was available for HPs and SDs in 1990, some statistical analyses were conducted from 1991 onwards for these two species. All ∑PCB data were natural logarithm transformed (ln) prior to statistical analysis to satisfy the requirement for the residuals of ∑PCB data to be normally distributed.
We determined long-term spatial and temporal trends in ∑PCB in two cetacean data sets being UK-stranded or bycaught harbour porpoises (HPs) (n = 706) from 1990-2012 and stranded or biopsied striped dolphins (SDs) (n = 220) from 1990-2009 in the western Mediterranean Sea. Secondly, we assessed mean, sample median and population median (predicted by the geometric mean of the sample ∑PCB data) for ∑PCB exposure in male and female HPs of all ages stranded in the UK (n = 706), and stranded/biopsied SDs (n = 220), bottlenose dolphins (BNDs) (n = 131) and killer whales KWs (n = 24) from NE Atlantic and Mediterranean waters. Since our ∑PCB data is lognormal -we can use the geometric mean (GEOMEAN) as the best estimate of the population median together with upper and lower 95% CIs 10 . ∑PCB concentrations in UK HPs that were "healthy" and died of acute physical trauma (control group) (n = 345) were compared to ∑PCB concentrations in HPs that died of a range of infectious diseases (case group) (n = 252). Sex differences in ∑PCB concentrations of all species were assessed for 3 subsets of animals: sexually mature individuals only; sexually immature animals only; and those individuals where sexual maturity status was undetermined. Finally, we compared mean ∑PCB concentrations between stranded and biopsied BNDs, SDs and KWs (all HPs were stranded).
Population Dynamics. We reviewed the current population dynamics data, mortality data (including bycatch) and reproductive status of the extant KW and other odontocete populations in European waters using strandings and necropsy data from this and other studies and systematic local and international observational surveys and other data on conservation status collected over the past 15-20 years (e.g. IUCN criteria) 8 .
Permits. The UK Cetacean Stranding Investigation Programme conducts work on UK strandings under contract to the UK Department of Environment, Food and Rural Affairs. It has appropriate licenses from the relevant authorities (Natural England, Scottish Natural Heritage and Natural Resources Wales) to allow it to collect and hold carcasses and samples from European Protected Species, in line with UK legislation enacted under the EU Habitats Directive. The blubber samples analysed at the University of Barcelona were obtained from the biological tissue bank of the University of Barcelona (BMA Tissue Bank) and originated either from biopsies or from naturally stranded individuals. In Spain, no specific permits are required to carry out research on samples stored at public institutions. The Spanish Government translates the competency for wildlife conservation to the regional governments of the various Autonomous Communities involved in this study, and the latter provided the permits for attending strandings, obtaining tissue samples, and using them for scientific studies. All the technicians that work on marine animal strandings from the Portuguese Wildlife Society are licensed for the capture, handling,