Chem. Res. Toxicol. 1992,5, 512-519

512

A Comparative Toxicological Investigation of Perfluorocarboxylic Acids in Rats by Fluorine-19 NMR Spectroscopy Carol M. Goecke,tJ Bruce M. Jarnot,*l§and Nicholas V. Reo*ltv* Department of BiochemistrylKettering Scott Magnetic Resonance Laboratory, Wright State University and Kettering Medical Center, Dayton, Ohio 45429, and Toxicology Division, Armstrong Lab, Wright-Patterson Air Force Base, Dayton, Ohio 45433 Received October 28, 1991

Male Fischer-344 rata administered a single intraperitoneal dose of perfluoro-n-octanoic acid (PFOA) or perfluoro-n-decanoic acid (PFDA) display a similar "wasting toxicity" characteristic of perfluorocarboxylic acids, with marked differences in temporal expression. Food/water consumption and urine output were monitored daily in PFOA-treated, PFDA-treated, and control rats. Fluorine- 19nuclear magnetic resonance (NMR) spectroscopy was used to monitor these fluorocarbons and possible fluoro metabolites in vivo, and to correlate differences in elimination with differences in effective toxicity. The data reveal a prolonged hypophagic response to PFDA and a more acute but transient response associated with PFOA treatment. PFOA causes a greater decline in food consumption than PFDA within the first 24 h postdose. PFOA-treated rata also show a ca. 2.5-fold increase in urine output on day 1,with only a slight increase in water consumption. In contrast to PFDA, PFOA-treated rats recover from hypophagia within 8 days. Fluorine-19 NMR spectra of various bodily fluids and liver in vivo display resonances of the parent PFOA or PFDA compounds and do not reveal any evidence of metabolism. Inorganic fluoride from dietary sources is detected in urine from both exposed and control rata. Differences in the route of excretion of PFOA vs PFDA are apparent from the spectral signal-to-noise ratio. The data suggest that PFOA is more readily excreted in the urine while PFDA is preferentially carried in bile. These apparent differences in elimination may account for their observed differences in effective toxicity. The acute transient toxicity and higher LDm associated with PFOA may result from its rapid renal clearance. In contrast, the more delayed protracted toxicity and lower LDw associated with PFDA may result from persistent enterohepatic recirculation with little net excretion. Introduction Perfluorocarboxylic acids have many commercial and industrial uses based on their surfactant and antiwetting properties and their thermal and chemical stability (1). Related neutral compounds used as nonflammable fluids, lubricants, and degreasers have been shown to be metabolized to analogous n-carboxylic acids (2). Similar fluorocarbon compounds are under investigation as potential blood substitutes and emulsifiers ( 3 , 4 ) . Although these perfluorinated materials were generally believed to be biologically inert, their unexpected metabolism and toxicological effects are now being investigated. The purpose of this study was to investigate the metabolism and excretion of perfluoro-n-octanoic acid (PFOA)' and perfluoro-n-decanoic (PFDA) in rata by nuclear magnetic resonance (NMR) spectroscopy. Previous work has shown the liver to be the primary target for fluorocarboxylic acid toxicity (5-8). Species and sexrelated differences have previously been observed in the excretion of PFOA and PFDA (9,10). The present study Correspondence and requests for reprints should be addressed to this author at KSMRL, Cox Institute, Wright State University, 3525 Southern Blvd., Kettering, OH 45429. + Kettering Scott Magnetic Resonance Laboratory. 1 Department of Biochemistry, Wright State University. I Armstrong Lab, Wright-Patterson Air Force Base. I Abbreviations: PFOA, perfluoro-n-octanoic acid; PFDA, perfluoron-decanoic acid; NMR, nuclear magnetic resonance; ip, intraperitoneal; S/N, signal-to-noise ratio; Au1/2, NMR linewidth measured at half-height.

0893-228s/92/2705-0512$03.00/0

exploits the unique advantages of NMR for investigating metabolism in vivo and provides a direct comparative examinationof the metabolic fate and pattern of excretion of these perfluorocarboxylic acids. The nondestructive/ noninvasive nature of NMR spectroscopy enables in vivo and circumvents problems associated with the chemical isolation of potentially labile metabolites. Fluorine-19 NMR was used to identify possible fluorinated metabolites in rat liver in vivo as well as in various bodily fluids. A second objective was to investigate and correlate differences in the pattern of elimination of these acids with observed differences in the temporal expression of their toxicity. The utility of 19FNMR to investigate the metabolism and pharmacokinetics of fluorocarbon compounds has been demonstrated (11-1 7).The stable fluorine-19isotope is 1005% in natural abundance,yields high NMR sensitivity (83.3% of the proton nucleus), and has a low endogenous background in biological systems. The wide chemical shift range for this nuclide (approximately400 ppm) provides high spectral resolution and facilitates the assignment of fluorinated metabolite resonance (18). The compounds chosen for this study were PFOA [CF&Fd&OzHI, a straight-chain eight-carbon carboxylic acid, and PFDA [CF&F2)&02H], a straight-chain ten-carbon carboxylic acid. The established 30-day intraperitoneal (ip) LDm levels in male Fischer-344 rais for PFOA and PFDA are 189 and 41 mg/kg, respectively (5). 0 1992 American Chemical Society

Toxicity of Perfluorocarboxylic Acids by

19F

NMR

Previous work (5, 6) has demonstrated that a single ip exposure to these perfluorocarboxylic acids a t their LDw doses results in a characteristic "wasting syndrome". Symptoms include pronounced hypophagia, marked body weight loss, and hepatomegaly. Hepatic studies also reveal peroxisome proliferation, an increase in peroxisomal fatty acyl-CoA oxidase activity (19,20),changes in the ratio of saturated to unsaturated fatty acids ( 5 , 6 ) ,decreased ketogenesis: and a significant elevation in liver triglyceride levels (21). While the underlying pattern of toxicity for these perfluorocarboxylic acids is similar, they display marked differences in the onset and duration of toxicity. PFOA causes an acute lethality but transient toxicity. Rats dosed at the 30-day ip LDw level die within 5 days postexposure, or recover rapidly. In contrast, PFDA induces a more delayed lethality and persistent wasting toxicity. Rats dosed a t the 30-day ip LDw level die between 10 and 16 days postexposure or recover slowly, displaying symptoms past 60 days (5). The NMR studies presented herein investigate the potential metabolism and biodistribution of these fluorocarbon compounds in vivo and provide new insight into the characterization of their toxicological effects. The results support an increasing body of evidence which suggests that PFOA and PFDA do not undergo biotransformation.

Experimental Procedures Untrapure (99+%) PFOA and PFDA were obtained from Technolube Products Co., Los Angeles, CA. Male Fischer-344 rata (200-250 g) were obtained from Charles River Breeding Laboratories (Wilmington, MA) and Harlan (St. Louis, MO). The animal housing area was maintained at 22 "C with a 12 h/12 h light/dark cycle, and animals were fed Purina Formulab Chow No. 5008. Treated animals received either a single ip injection of 150 mg/kg PFOA or 50 mg/kg PFDA, dissolved in 1:l (v/v) propylene glycol/water. Control animals received a single ip injection of 1:l propylene glycol/water (vehicle). Stock solutions were prepared such that the ip dose did not exceed 0.5 mL. Previous studies indicate that this volume of vehicle solution causes no distress or peritoneal reaction. Data presented in Figures 1-3 were analyzed using the appropriate Student's t-test for paired and unpaired data and are considered to be statistically significant at a value of p I 0.05. Some p values are explicitly stated in the text. Experimental procedures are discussed in three subcategories, each describing experiments involving separate groups of animals. Food, Water, and Urine Monitoring. Animals were divided into four experimental groups as follows: (i) PFDA-treated rata (n = 4) which received ad libitum access to food and water; (ii) PFOA-treated rata (n = 3) which received ad libitum access to food and water; (iii) An ad libitum-fed control group (n = 6) in which rata were vehicle-treated and received ad libitum access to food and water; (iv) A pair-fed control group (n = 4) in which rata were weight-paired to the PFDA-treated rata and vehicletreated. These rata received ad libitum access to water, but were fed the same amount of food as their PFDA-treated partners had consumed during the previous 24-h period (pair-feeding). Pairfeeding simulates the prolonged hypophagic effects associated with PFDA. Following exposureto either fluorocarbonor vehicle, animals were individually housed in metabolism cages. Food consumption,water consumption,and urine output were recorded daily for all groups. Bodily Fluid and Liver Homogenate Analyses. Following exposure to either fluorocarbon or vehicle, animals were indi-

* G . D. Pilcher, unpublished results.

Chem. Res. Toxicol., Vol. 5,No. 4, 1992 513 vidually housed in metabolism cages. Urine samples were collected daily for up to 8 days into plastic cups maintained at 4 OC. Urine samples from PFDA-treated, PFOA-treated, and pair-fed control rats were centrifuged to remove particulate contamination and stored frozen at -20 "C for subsequent NMR and fluoride analyses. Fluorine-19 NMR spectra were acquired from urine samples collected on days 1,2,3,4,5,and 8 postdose (n = 4/group/day). Potentiometrically measured urine fluoride levels were determined on separate groups of PFDA-treated, PFOA-treated, and pair-fed control rata using a fluoride ionselective electrode (Orion Model 960999) as described by Neefus et al. (22). Fluoride levels were measured prior to perfluorocarbon or vehicle dosing, and daily for up to 6 days thereafter (n = 2/group/day). Bile was collected from both PFDA- and PFOA-treated rata on days 1and 3 postexposure (n= 5/day)at approximately 1000 a.m. Rata were anesthetized with an ip dose of 70% ketamine/ 30% xylazine (1 mL/kg for induction and 0.1 mL/30 min for maintenance). Bile ducta were surgically isolated and cannulated with 22-gauge catheters, and bile was collected over a 2-3-h period (bile flow -1 mL/h). Blood samples were collected from both groups prior to termination. Liver homogenates were prepared from a separate group of PFDA-treated rata on days 3,7, and 8 postexposure (n = 2/day) and PFOA-treated rata on day 3 postexposure (n = 2). Livers were perfused in situ with ice-cold heparinized saline, surgically removed, weighed, and homogenized using a Potter-Elvehjem Teflon/glass homogenizer. Crude liver homogenates were fractionated using standard differential centrifugation techniques. Gross cellular debris were removed (2000g pellet), and mitochondrial/peroxisomal (1oooOg pellet), microsomal (11oooOg pellet), and cytosolic (11oooOg supernatant) fractions were isolated. All biological samples were stored frozen at -20 O C for subsequent NMR analyses. In vitro 19FNMR analyses of bodily fluids, liver homogenates, and cell fractions were performed on a Bruker AM 360 highresolution spectrometer (8.5 T) using a commercial Bruker 5mm fluorine probe operating at a centerband frequency of 338.86 MHz. Samples (approximately 0.5 mL) were placed in 5-mm NMR tubes (Wilmad Glass Co., catalog no. 507PP8) and several drops of DzO added to provide a deuterium signalfor the magnetic field lock. The initial spectrum of each sample was acquired over the entire chemical shift range (ca. 400 ppm) to ensure detection of all possible metabolites. The spectral sweep width was subsequently decreased to provide better resolution over the range of interest. Spin-lattice relaxation times (2'1) for specific resonances in urine from a PFOA-treated rat were measured at 8.5 T and 27 OC by the inversion-recovery technique (23)and were found to range from0.9 to 1.1s. The Ernst equation was used to calculate acquisition pulsing parameters to optimize the signal/noise ratio per unit time. Except where noted, all spectra were acquired at 27 "C using a 30" pulse and an interpulse repetition time of 0.151 s. Chemical shifts for all spectra are relative to the CF3 resonance of the parent perfluorocarboxylic acid, which is set at 0 ppm. The IUPAC convention for the chemical shift scale is used. In reference, trifluoroacetic acid (TFA) resonates at 5.4 ppm on this chemical shift axis. Samples were also prepared in which PFDA was added to a bile specimen from a control animal, and PFOA was added to a urine specimen from a control animal. Liver homogenates and serum were also prepared from control animals and spiked with PFOA or PFDA. NMR spectra of these sampleshelped to identify and confirm chemicalshift assignments for the fluoro compounds in these media. NMR of Liver in Vivo. Following ip exposure to the fluorocarbon compounds, animals were individually housed, received ad libitum accessto food and water, and were maintained for up to 51 days. Fluorine-19 NMR spectra of liver in vivo were acquired on a Bruker Biospec 2.35/400 NMR system (2.35 T) operating at a centerband frequency of 94.44 MHz. Rata were anesthetized with ketamine/xylazine (as described above) and a 3-cm diameter 19F surface coil (constructed in-house) was

514 Chem. Res. Toxicol., Vol. 5, No. 4, 1992

Goecke et al.

t

32t

c

It n

e

'i 0

t

0 ,

-4

,

,

-2

,

,

.

.

.

2

0 TIME

,

4

.

.

6

.

-2

-4

8

TIME

[Days)

Figure 1. Changes in mean daily food consumption for PFOA (triangles;n = 3), PFDA (open circles;n = 4), and ad libitum-fed controls (filled squares; n = 6). Control animals were monitored for 5 consecutive days prior to treatment (days -4 to 0)and for 8 days posttreatmentwith vehicle solution. Data acquisition for all other groups was initiated following treatment with the appropriate compounds on day 0. For some data points, the error bar (ASEM) is contained within the size of the symbol and is not shown. positioned on the abdomen over the liver. NMR data were acquired using a pulse that produced a 180' flip angle at the surface. This technique suppresses signals from the surface and provides greater NMR sensitivity from a volume which encompasses predominantly liver. Since these fluorocarbon compounds were dosed ip and predominantly concentrate in the liver, there is little concern for spectral contamination from stomach or muscle. (For more information about surface coils in general or the acquisition of 19FNMR data from liver, see refs 24 and 25). Spectra were obtained at 2, 3, 7, 9, 10, 13, 22, and 51 days postexposure with PFDA (n= l/day). Because of toxicity-related deaths, these experiments involved the use of three separaterats in order to obtain data at all the various time points. In vivo spectra were also acquired from a single PFOA-treated rat at 9 and 13 days postexposure.

Food, Water, and Urine Data. Changes in daily food consumption, water consumption, and urine output are shown in Figures 1,2, and 3, respectively. Ad libitum-fed control animals were monitored for 5 consecutive days prior to treatment with the vehicle solution (days -4 to 0). Data acquisition for all other groups was initiated following treatment on day 0. At this time, control animals were treated with the vehicle solution and the experimental groups were given the appropriate fluorocarbon compound. Data was collected for 8 days postdose for each group of animals. Both PFDA-treated and PFOA-treated rats show a significant decrease in food consumption (Figure 1)relative t o ad libitum-fed control animals within the first 24 h postdose (p 5 0.001). PFDA-treated rats consume 6.9 f 0.6 g of food on day 1 and show a continued decrease in food consumption to nearly 0 g by day 3 and the remaining days examined. PFOA-treated rats show a more pronounced decrease in food consumption than the PFDA group, consuming 0.5 f 0.5 g of food on day 1(p = 0.0006); however, this initial decrease in food consumption is followed by a rapid recovery t o 9.3 f 1.8 g on day 3, and

4

6

6

[Days)

Figure 2. Changes in mean water consumption for PFOA (triangles; n = 3), PFDA (open circles; n = 4), PFDA pair-fed controls (filled circles; n = 4), and ad libitum-fed controls (filled squares;n = 6). Experimental protocol is described in the legend of Figure 1. For some data points, the error bar (ASEM) is contained within the size of the symbol and is not shown. 2 8 '

I

7

r

t

'

'

'

'

'

'

'

T

"t

I

1

20

E

t

I-

a 3

.I

I

I

t

0

-4

Results

2

0

-2

0

2 T I E

4

6

8

(Days1

Figure 3. Changes in the mean daily volume of urine output for PFOA (triangles; n = 3), PFDA (open circles; n = 4), PFDA pair-fed controls (filledcircles; n = 41, and ad libitum-fed controls (filledsquares; n = 6). Experimental protocol is described in the legend of Figure 1. For some data points, the error bar (ASEM) is contained within the size of the symbol and is not shown.

a continuing gradual increase to normal control levels by day 7 (data at days 7 and 8 are not statistically different from control). Ad libitum-fed control rats show a relatively steady level of food consumption on all days examined. Relative t o the ad libitum-fed control group, PFDAtreated rata show a ca. 60% decrease in water consumption (Figure 2) to 9.5 f 1.3 mL within the first 24 h postdose (p = 0.OOOl). This is also significantly less than the water consumed by their pair-fed control partners at day 1 (95 f 1.3versus 15.0 f 1.5m l l d a y ) . A more gradual decrease in water consumption t o ca. 4 mL/day is observed for the remaining days examined. Pair-fed control rats show a ca. 35% decrease in water consumption within the first 24 h postdose and then display a further decline in water consumption which closely parallels the results seen with PFDA-treated rats. Water consumption for the pair-fed control rats remains significantly different from the ad

Toxicity of Perfluorocarboxylic Acids by

19F

Chem. Res. Toxicol., Vol. 5, No. 4, 1992 515

NMR

libitum-fed control group, with the latter showing a relatively steady consumption of water on all days examined. In contrast, PFOA-treated rats display a transient increase in water consumption relative to ad libitum-fed controls within the first 24 h postdose to 27.3 f 4.3 mL, although this difference is not statistically significant. This decreases to ca. 15 mL by day 4 and eventually returns to normal by day 7. Data for this group are statistically different from data for the ad libitum-fed control animals on days 4-6 only. On all posttreatment days examined, PFOA-treated rata consume a statistically higher volume of water per day as compared to PFDA-treated and their pair-fed controls. PFDA-treated rata show a gradual decrease in urine output to ca. 2 mL by day 5 (Figure 3). Pair-fed control rata also show a decrease in urine output which closely parallels the PFDA-treated group. In contrast, PFOAtreated rata show an approximate 2.5-fold increase in urine output relative to the ad libitum-fed controls @ = 0.04) within the first 24 h postdose to 22.3 f 4.1 mL. This is followed by a rapid decrease to control levels by day 2. Urine output remains slightly below ad libitum-fed control levels from days 3 to 7 and is significantly different on days 3 and 5-7. By day 8 urine output returns to control levels. Ad libitum-fed control rata show a relatively steady level of urine output on all days examined. NMR Data and FluorideAnalyses. A high-resolution 19F NMR spectrum of PFOA (50 mM in 1:l propylene glycol/H20) is shown in Figure 4. The insert is an expansion of the low-frequency region between -36 and -46 ppm. Spectral assignments were determined using a two-dimensional correlation spectroscopy 19Fhomonuclear (2D-COSY) NMR experiment (2D spectrum not shown) (23). All in vitro urine lgF spectra from PFOA-treated rata (1-5 and 8 days postdose) show a spectral pattern very similar to that of the parent dosing compound plus one additional peak, and a steady gradual decrease in the signal-to-noise ratio (S/N) over successive days. A representative spectrum is shown in Figure 5A. This spectrum shows 75 min of signal averaging from a sample of rat urine collected 4 days postdose with 150 mg/kg PFOA. A comparisonof this spectrum to that obtained from a control urine specimen spiked with PFOA (not shown) reveals that all the resonances can be assigned to the parent compound (PFOA), with the exception of a peak at -38.7 ppm. The identity of the resonance at -38.7 ppm (observed in urine samples from both the PFOA-treated rat and control) was investigated by acquiring a l9F homonuclear 2D-COSY spectrum of the PFOA-treated rat urine sample. The results reveal the absence of scalar couplings between the peak at -38.7 ppm and the fluorines within PFOA (2D spectrum not shown), indicating that this particular resonance originates from a separate molecule. All in vitro urine spectra acquired from PFDA-treated rata (1-5 and 8 days postdose) predominantly show a single resonance at approximately -37 ppm. Several of these samples also display much weaker signals which appear to be from the parent compound (barely detectable above the noise). The 19FNMR spectrum of rat urine collected 3 days postdose with 50 mg/kg PFDA is shown in Figure 5B. In contrast to Figure 5A, this spectrum reflects approximately 5 h of signal averaging and displays asingle peak at -37.1 ppm. No resonances of PFDA are observed. A 19FNMR spectrum of urine from a control (vehicle-

2

-38

-36

b

I

1

-l5

-10

-15

-20

-hO

I

-25 PPH

PPM

-h2

1

I

-30

-35

-44

- is

4 1

I

-40

-45

Figure 4. High-resolution 19FNMR spectrum of PFOA (50 mM in 1:lpropylene glycol/H*O). This spectrum was acquired at 8.5 T and 27 OC, using a 22' pulse, an interpulse repetition time of 2.33 s, 24 transients, and a sweep width of 24 kHz. The insert is an expansion of the low-frequency region between -36 and -46 ppm. Data were processed using 32K total data points and an exponential filter producing 2-Hz line-broadening. Spectral assignments were determined by a 19Fhomonuclear 2D-COSY experiment, and chemical shifts are relative to the CFSresonance which was set at 0 ppm. Numbers abovethe peaks refer to specific carbons in the carboxylicacid chain,with the carbonyldesignated as C1.

treated) rat is shown in Figure 5C. This spectrum required approximately 6 h of signal averaging and also displays a fluorine resonance at -38.9 ppm. This peak (observed from -37 to -39 ppm) was determined to arise from inorganic fluoride, as confirmed by the addition of NaF to all urine samples. In all cases the addition of NaF resulted in an increase in intensity of the peak in question. Fluorine-19 NMR spectra from control and exposed animals (both PFOA and PFDA) suggest similar urinary concentrations of free inorganic fluoride. The baee-line values of potentiometrically measured [PIobserved in urine from all animals (n = 6), prior to exposure to any fluoro compounds, were found to range from 1.96 to 4.08 pg/mL. For individual groups, the measured [PIfrom urine samples collected at all times postdose ranged from 1.86 to 4.09, from 1.43to 5.67, and from 2.44 to 4.77 pg/mL for control (n = 2), PFOA (n = 2), and PFDA (n = 2) groups, respectively. The 19FNMR spectra of bile from PFOA-exposed rata typically displayed poor S/N. Figure 6A is a spectrum collected 3 days postdose with 150mg/kg PFOA and shows approximately 27 h of signal averaging. In contrast, the l9F NMR spectrum of rat bile collected 3 days postdose with 50 mg/kg PFDA (Figure 6B) required only 4 h of signal averaging and displays roughly twice the S/N of Figure 6A. These spectra were compared to the corresponding appropriate controls (Le., bile from a control rat

516 Chem. Res. Toxicol., Vol. 5, No. 4, 1992

Goecke et al.

i A

0

-20

-10

-30

PPH

-50

-40

C

T

5

k b -5 -10 -15 -20 PPH-25 -30 -35 - 4 0 - 4 5 -50 Figure 5. High-resolution 19FNMR spectra of rat urine at 8.5 T and 27 "C, using a 30" pulse, an interpulse repetition time of 0.151 s, and a sweep width of 20 kHz. Data was processed using 4K total data points and an exponential filter producing 10-Hz line-broadening. (A) Urine from a PFOA-treated rat (150 mg/ kg) collected 4 days postdose; 75 min of signal averaging (29 669 transients). (B)Urine from aPFDA-treatedrat(50mg/kgPFDA) collected 3 days postdose; 5 h of signal averaging (115812 transients). (C) Urine from a control (vehicle-treated)rat; 6 h of signal averaging (146 669 transients). Chemical shifts are relative to the CF3 resonance in spectrum A which was set at 0 I

,

,

I

I

I

I

l

/

,

PPm.

spiked with either PFOA or PFDA). All resonances in Figure 6B can be assigned to the parent PFDA compound. Bile from the PFOA-treated rat yields a weak signal at -39.5 ppm which is not attributed to the parent compound, but was also seen in the control spectrum. This peak is presumed to be inorganic fluoride. All other resonances in this spectrum (Figure 6A) are assigned to the parent PFOA compound. The slight chemical shift differences observed between Figures 4 and 6A will be addressed in the Discussion. Fluorine-19 NMR spectra of serum, plasma, liver homogenates, and cell fractions were obtained from PFOAand PFDA-treated rats and analyzed for fluorinated metabolites. The serum and plasma samples reveal a relatively high S/N with broader linewidths (Av1p = 120170 Hz)than those observed in the spectra of urine and bile (Av1p = 25-70 Hz).The corresponding control spectra (serum or plasma spiked with the appropriate fluorocarbon compound) were directly comparable, and all peaks were attributed to the parent dosing compound. Representative spectra of serum from a PFOA-treated rat and the corresponding control are provided as supplementarymaterial (Figure 1 s ) . Spectra of liver homogenates and cell fractions were typically very noisy with relatively broad lines (not shown); no definitive resonances could be assigned even after 12 h of signal averaging.

i

0

,

-5

,

-10

,

-15

I

-20

PPM

I

I

I

I

-25

-30

-35

-40

I

-45

I

-50

Figure 6. High-resolutionlgFNMR spectra of rat bile acquired and processed using the same experimental parameters as indicated in Figure 5. (A) Bile from a PFOA-treated rat (150 mg/kg) collected 3 days postdose;27 h of signalaveraging(648 046 transients). (B) Bile from a PFDA-treated rat (50 mg/kg) collected 3 days postdose; 4 h of signal averaging (104000 transients). Chemical shifts are relative to the CF3resonance which was set at 0 ppm.

NMR spectra of liver in vivo (at 2.35 T) were also acquired from PFOA- and PFDA-treated animals. A representative 19Fspectrum collected 3 days postdose with PFDA is shown in Figure 7. Although there is asignificant decrease in resolution relative to the spectra obtained at 8.5 T, the resonances of the parent compound are evident. All spectra of in vivo liver solely display resonances representative of the parent dosing compounds (PFOA or PFDA) at all days postexposure investigated.

Discussion Data from this study reveal differences in the temporal expression of the toxicity associated with PFOA and PFDA. Although both compounds cause hypophagia in rats, as presented here and by others (5, 26), several distinctions are apparent in the present study. First, PFDA causes a prolonged hypophagic response while PFOA-treated rats display an acute decrease in food consumption on days 1 and 2, which rapidly recovers to normal by day 7. Second, PFDA and pair-fed controls show a relatively steady decrease in daily water consumption while PFOA-treated rats show a more transient response. Although daily water consumption for PFOAtreated rata falls below ad libitum-fed control levels, they are significantly greater than the PFDA group. Changes in water consumption observed with PFDA and pair-fed control rats closely parallel each other, suggesting that the decrease in water consumption is a result of hypophagia, rather than a direct influence of the fluorocarbon compound. Since PFOA-treated rats show a dramatic recovery from hypophagia, daily water consumption levels

Toxicity of Perfluorocarboxylic Acids by

19F

Chem. Res. Toxicol., Vol. 5, No. 4, 1992 517

NMR

I

10

0

-10

-20

PPM

-30

-40

-50

-60

Figure 7. Fluorine-19 NMR spectrum of rat liver in vivo at 2.35 T collected 3 days postdose PFDA. Data was acquired using a 2 O - p pulse, ~ an interpulse delay of 0.051 8, sweep width of 10 kHz,

and 25 min of signal averaging (30 161 transients). Data were processed using 4K total data points and an exponential filter producing 20-Hz line-broadening. Chemical shifts are relative to the CF3 resonance which was set at 0 ppm.

are not as low as those for PFDA and pair-fed control groups. Third, in contrast to the gradual and constant decrease in urine output observed with PFDA and pairfed control rats, PFOA-treated rats display a dramatic 2.5-fold increase in urine output on day 1. The corresponding increase in water consumption seen in PFOAtreated animals does not appear sufficient to account for the increased urine output. This suggests that the acute toxicity associated with PFOA may involveincreased renal activity and dehydration. High-resolution in vitro and in vivo 19FNMR spectra obtained from animals exposed to PFOA or PFDA reveal fluorine resonances which can be assigned to these parent compounds. Urine spectra, and the bile spectrum shown in Figure 6A, reveal an additional peak at -37 to -39 ppm, identified as free inorganic fluoride. The broader linewidths observed in I9F spectra of serum and plasma, as compared with bile and urine samples, are presumably due to an association of PFOA or PFDA with proteins (Le., albumin) (9, IO). The poor spectra from samples of liver homogenate5 and cell fractions may also be due to association with proteins and/or heterogeneity of the sample. Potentiometrically measured urine [PI varied daily over a wide range for control animals as well as treated groups. This urinary fluoride is believed to be of dietary origin, and therefore, the variation observed in [F-] is not unexpected. The laboratory water is defluorinated, but the animals’ food contains 30 ppm inorganic fluoride. Interestingly, biological systems are typically thought to be void of endogenous fluorine metabolites, thereby rendering 19FNMR spectra free of background signals. We caution, however, that dietary sources of inorganic fluoride at concentrations of a few micrograms per milliliter can be detected at high field.

Since the concentration and magnitude in variation of urine fluoride are similar for control and treated animals, this fluoride is not believed to be a product of fluorocarbon metabolism. These findingssupport those of Vanden Heuvel et al. (7, 81, in which [PIwas potentiometrically measured in urine and plasma from PFOA- and PFDAtreated rata. Their potentiometric analyses suggest that neither acid is biologicallydefluorinated. Defluorination would also result in the formation of new fluorinated compounds; however, our experiments detected no such metabolites. Detection of fluoro metabolites by NMR is subject to the following limitations: (i) the metabolites must have resonance frequencies which are distinct from the parent compound and the spectral resolution must be sufficient to depict these chemically shifted peaks; (ii) the metabolites must exist at concentrations above the level of detection for 19FNMR (estimated to be 100 pM in these studies); (iii) the metabolites must have sufficient molecular mobility. If they are bound to large macromolecules or incorporated into membranes, the resonance lines may become very broad and thus be rendered NMR invisible. Fluorine chemical shiftsare very sensitive to any changes in the molecular environment. Displacement of fluorine for H or OH (i.e., hydroxylation) is expected to result in significant chemical shifts of the fluorine resonances. For example, CF3(CF2)zCOzHversus CF2H(CF2)&02H shows a 36 ppm chemical shift difference between the fluorines on the terminal carbons (27). Thus, defluorination or fluorine displacement of the perfluorocarboxylic acids would be easily discernible by 19FNMR. A more likely biotransformation of PFOA or PFDA, however, is esterification,which results in considerably smaller 19Fchemical shifts. For example, the conversion of CF3CF2CF2C02H to its ethyl ester results in a 0.8 ppm chemical shift in the a-carbon fluorines, while its conversion to a thio ester results in a 4.8 ppm shift in the a-carbon fluorines (27). Thus detection of an esterified metabolite of PFOA or PFDA in urine or bile would only be possibleif the chemical shift is greater than about 1.5 ppm. Detecting such a metabolite from liver in vivo would require a larger shift due to the lower spectral resolution associated with in vivo spectra. The chemical shifts of these perfluorocarboxylic acids are solvent and concentration dependent. Spectra of bodily fluids from treated animals show slightly different (typically I f1.0ppm) chemical shifts than spectra of the parent dosing solutions (PFOA or PFDA in 1:lpropylene glycol/water; Le., Figure 4). Therefore, control solutions (samples of a particular bodily fluid spiked with PFOA or PFDA) were prepared with concentrations similar to the corresponding samples from the treated animals. This was done by estimating the concentration in the sample from the treated animal from the S/N in the NMR spectrum, and then adjusting the control solution by a series of dilutions. These spectral shifts may partly explain the slight variation observed for the P resonance between different samples. It is also likely that variations in pH and ionic concentrationsbetween urine samples may affect the P chemical shift (15,17). Within the boundaries of the aforementioned limitations, all the 19FNMR investigations show no evidence of fluoro metabolites. The 19Fspectra from bile, urine, and serum display resonance peaks in which the chemical shifts (relative to the CFBpeak) are comparable within fO.10

518 Chem. Res. Toxicol., Vol. 5, No. 4, 1992

ppm to the spectra from the appropriate controls, with the exception of bile from a PFOA-treated rat (Figure 6A). This spectrum has three peaks in the region between -40 and -42ppm, with the peak at -41.9 ppm having about twice the intensity as the other two downfield peaks. The corresponding control sample yields four peaks of approximately equal intensity in this same spectral region. The shift of the additional peak varies with concentration of the control solution from -41.6 to -41.8 ppm. This particular resonance is thought to overlap with the peak at -41.9 ppm in the spectrum from the PFOA-treated animal (Figure 6A), thus yielding a peak of about twice the normal (control)intensity at this position. This minor discrepancy between the sample from the treated animal and its corresponding control is thought to be due to slight differences in the composition of the bile samples from these animals, which may be important at very dilute concentrations of PFOA. The in vivo liver spectrum (Figure 7) displays the resonances of PFDA with overlapping peaks between -40 and -44 ppm. In comparison, a l9F liver spectrum from a control rat in vivo is void of any signal (data not shown). The reduction in spectral resolution relative to the spectra of urine and bile results partly from a lower magnetic field strength (2.35 T). Also, tissue in vivo typically yields lowresolution spectra due to restricted molecular motions (decrease in TZrelaxation) and the heterogeneous nature of the sample. Within the limits of this resolution, all resonances can be assigned to the parent compounds; no resonances attributable to fluoro metabolites were detected on any days postexposure to PFDA or PFOA. The purpose of this study was not to quantitate the distribution of PFOA and PFDA in liver and bodily fluids-this has already been reported (7,8,28). Rather, l9F NMR spectra were acquired following PFOA or PFDA exposure to investigate the possible formation of fluorinated metabolites. Qualitative assessment of the spectral data suggests that PFOA is rapidly excreted in the urine while PFDA is preferentially carried in the bile. Comparison of urine spectra from rats exposed to these perfluorocarboxylic acids shows relatively high S/N of the parent compound in PFOA-treated animals and insignificant levels of the parent compound in PFDA-treated animals (Figure 5A,B). Conversely, bile spectra from PFOA-exposed rats show poor S/N of this acid, while bile spectra from PFDA-exposed rats show relatively high S/ N of PFDA resonances (Figure 6). These NMR data, in conjunction with the finding that urine output in PFOA-treated rats significantly exceeds that of PFDA-treated rats (Figure 3), suggest that renal excretion is a major process for the detoxification of PFOA. The relative distribution of PFOA and PFDA in bodily fluids suggests that differences in effective toxicity may relate to their route and rate of clearance from the body. In general, the aqueous solubility of straight-chain carboxylic acids is inversely related to chain length. The greater aqueous solubility of PFOA relative to PFDA appears to facilitate its rapid urinary excretion, thereby yielding a higher LDw, more acute lethality, and transient toxicity. The relative hydrophobicity of PFDA appears to favor biliary enterohepatic recirculation over urinary elimination. The resulting biological persistence may account for ita lower LDm, delayed lethality, and protracted toxicity. The route of excretion is also known to be influenced by molecular weight, where compounds ex-

Goecke et al.

ceeding approximately 300 g/mol are largely eliminated through the bile and intestine (29). Since both PFOA and PFDA exceed 300 in molecular weight, the observed differences in the route of excretion are more likely due to the factors relating to solubility. These comparative studies of PFOA and PFDA were made at their respective 30-day LDm doses in order to investigate their metabolic fate and biodistribution at a level of similar toxicity. It is not known whether the biodistribution of these compounds is dose dependent; however, both perfluorocarbon compounds have been shown to accumulate in the liver (6,301. Also, preliminary data from our laboratory (not shown) in which PFOA was administered at lower doses (Le., 2, 20, and 50 mg/kg) show a similar preference for urinary versus biliary excretion as seen at the higher LD50 dose. The mechanism of toxicity associated with these perfluorinated carboxylic acids remains unclear. Olson et al. (5)hypothesize that these compounds may be biologically recognized as fatty acid analogues and interfere with hepatic mitochondrial @-oxidationof endogenous fatty acids. Bronfman (31) further suggests that the biotransformation of these compounds yields CoA conjugates which may be responsible for their toxic effects. Peterson and coworkers have recently demonstrated, however, that a CoA derivative of PFOA or PFDA could not be detected in rat hepatocytes or purified microsomes in vitro (321, and neither PFOA nor PFDA is incorporated into lipids (8). Therefore, they hypothesize that the effects of PFDA on fatty acid metabolism are due to the parent compound and not to PFDA entering the lipid metabolic pathways as an activated CoA conjugate (33). Our results further corroborate this proposed theory since data from this study provide no evidence for the detectable metabolism of either PFOA or PFDA in vivo. In conclusion,the results of this work suggestthat PFOA and PFDA are not metabolized in vivo and that differences in their toxicity may be related to the excretory processes which clear them from the body. Further studies are currently in progress to investigate the effects of PFOA and PFDA on hepatic metabolic processes using NMR in vivo.

Acknowledgment. We thank Dr. Phillip Cruz (Department of Biochemistry, Wright State University) for his assistance with 2D-COSY NMR experiments. This work was sponsored by the Air Force Office of Scientific Research, Air Force Systems Command, USAF, under grant or cooperative agreement number AFOSR 90-0148. Supplementary Material Available: Fluorine-19 NMR spectra of (i) serum from a PFOA-treated rat and (ii) serum obtained from a control rat and spiked with PFOA (Figure 1s) (1page). Ordering inform8tisn is given on any current masthead page.

References (1) Guenther, R. A.,and Victor, M. L. (1962) Surface active materials

from perfluorocarboxylic and perfluorosulfonic acids. Ind. Eng. Chem. Prod. Res. Deu. 1,165-169. (2) Greene, R. J.,Brashear, W. T., Auten, K. L., and Mahle, D. A. (1992) Confirmation of a carboxylic acid metabolite of polychlorotrifluoroethylene and a method for GC/ECD analysis from biological matrices. J . Anal. Toxicol. 16, 28-32. (3) Riess, J. G., and Le Blanc, M. (1988) Preparation of perfluorochemical emulsions for biomedical use: Principles, materialsand methods.

Toxicity of Perfluorocarboxylic Acids by 19F NMR In Blood Substitutes: Preparation, Physiology and Medical Applications (Lowe, K. C., Eds.) pp 94-129,VCH Publishers, New York. Riess, J. G., Santaella, C., and Vierling, P. (1991)New perfluoroalkylated phosphatidylcholines as surfactants for biomedical applications. Proceedingsof theXXIInd CED Meeting on Surfactants, Palma de Mallorca, pp 157-171,Comit&EspagnolDetergencia, Tensioactivos y Mines, Ed. Barcelone. Olson, C. T., and Andersen, M. E. (1983)The acute toxicity of perfluorooctanoic and perfluorodecanoic acids in male rats and effects on tissue fatty acids. Toxicol. Appl. Pharmacol. 70,362-372. George, M. E., and Andersen, M. E. (1986)Toxic effects of nonadecafluoro-n-decanoic acid in rata. Toxicol. Appl. Pharmacol. 86, 169-180. Vanden Heuvel, J. P., Kuslikis, B. I., Van Rafelghem, M. J., and Peterson, R. E. (1991)Tissue distribution, metabolism, and elimination of perfluorooctanoic acid in male and female rats. J. Biochem. Toxicol. 6,83-92. Vanden Heuvel, J. P., Kuslikis, B. I., Van Rafelghem, M. J., and Peterson, R. E. (1991)Disposition of perfluorodecanoic acid in male and female rats. Toxicol. Appl. Pharmacol. 107,1-10. Ylinen,M.,and Auriola, S. (1990)TissueDistribution andelimination of perfluorodecanoic acid in the rat after single intraperitoneal administration. Pharmacol. Toxicol. 66,45-48. Hanhijarvi, H., Ophaug, R. H., and Singer, L. (1982)The sex-related differences in perfluorooctanoate in the rat. Proc. SOC. Exp. Biol. Med. 171,50-55. Nicholson, J. K., and Wilson, I. D. (1987)High resolution nuclear magnetic resonance spectroscopy of biological samples as an aid to drug development. Prog. Drug Res. 31,427-479. Malet-Martino, M. C., Bernadou, J., Martino, R., and Armand, J. P. (1988)19F NMR spectrometry evidence of bile acid conjugates of a-fluoro+alanine as the main biliary metabolites of antineoplastic fluoropyrimidines in humans. Drug Metab. Dispos. 16,7684. Malet-Martino, M. C., and Martino, R. (1989)The application of nuclear magnetic resonance spectroscopytodrugmetabolism studies. Xenobiotica 19,583-607. Malet-Martino, M. C., and Martino, R. (1991)Uses and limitations of nuclear magnetic resonance (NMR) spectroscopy in clinical pharmacokinetics. Clin. Pharmacokinet. 20,337-349. Martino, R., Lopez, A., Malet-Martino, M. C., Bernadou, J., and Armand, J. P. (1985)Release of fluoride ion from 5’-deoxy-5-fluorouridine, an antineoplastic fluoropyrimidine, in humans. Drug Metab. Dispos. 13, 116-118. Martino, R., Bernadou, J., Malet-Martino, M. C., Roche, H., and Armand, J. P. (1987)Excretion of doxifluridine catabolitesin human bile assessed by 19F NMR spectrometry. Biomed. Pharmacother. 41, 104-106. Tecle, B., and Casida, J. E. (1989)Enzymatic defluorination and metabolism of fluoroacetate, fluoroacetamide, fluoroethanol, and (-)-erythro-fluorocitrate in rata and mice examined by ‘9F and 13C NMR. Chem. Res. Toxicol. 2,429-435. Beteille, J. P., Lopez, A., Bon, M., Malet-Martino, M. C., and Martino, R. (1985)Simultaneous assay for organic fluorine compounds

Chem. Res. Toxicol., Vol. 5, No. 4, 1992 519 in biological fluids by fluorine-19 nuclear magnetic resonance spectrometry over a large spectral width. Anal. Chim. Acta 171, 225-231. (19) Harrison, E. H., Lane, J. S.,Luking, S., Van Rafelghem, M. J., and Andereen, M. E. (1988)Perfluoro-n-decanoic acid: Inductionof peroxisomal 8-oxidation by a fatty acid with a dioxin-like toxicity. Lipids 23, 115-119. (20) Intrasuksri, U.,and Feller, D. R. (1991)Comparison of the effects of selected monocarboxylic and perfluorinated fatty acids on peroxisome proliferation in primary culture rat hepatocytes. Biochem. Pharmacol. 42, 184-188. (21) Van Rafelghem, M. J., Vanden Heuvel, J. P., Menahan, L. A., and Peterson, R. E. (1988)Perfluorodecanoic acid and lipid metabolism in the rat. Lipids 23,671-678. (22) Neefus, J. D., Cholak, J., and Saltzman, B. E. (1970)The determination of fluoride in urine using a fluoride specific ion electrode. Am. Ind. Hyg. Assoc. J. 31 (Jan-Feb), 96-99. (23) Bovey, F. A. (1988)in Nuclear Magnetic Resonance Spectroscopy, Academic Press, New York. (24) Ackerman, J. J. H. (1990)Surface (local) coils aa NMR receivers. Concepts Magn. Reson. 2, 33-42. (25) Wolf, W., Albright, M. J., Silver, M. S., Weber, H., Reichardt, U., and Sauer, R. (1987)Fluorine-19 NMR spectroscopic studies of the metabolism of 5-fluorouracil in the liver of patients undergoing chemotherapy. Magn. Reson. Imaging 6, 165-169. (26) Pilcher, G. C. (1985)The effects of perfluoro-n-decanoic (PFDA) on the rat heart, Ph.D. Thesis, Wright State University, Dayton, OH. (27) Emsley, J. W., and Phillips, L. (1971)Fluorine Chemical S h i h . In Progress in Nuclear Magnetic Resonance Spectroscopy (Emsley, J. W., Feeney, J., and Sutcliffe, L. H., Eds.) Vol. 7,pp 142,146,149, and 175, Pergamon Press, New York. (28) Davis, J. W., 11, Vanden Heuvel, J. P., Kuslikis, B. I., and Peterson, R. E. (1991)Castration increases elimination of perfluorooctanoic acid (PFOA) in male rats. The Toxicologist 11, Abstract 296. (29) Gibson, G. G., and Skett, P. (1986)in Introduction t o Drug Metabolism, pp 183-185,Chapman and Hall, New York. (30) Griffith, F. D., and Long, J. E. (1980)Animal toxicity studies with ammonium perfluorooctanoate. Am. Ind. Hyg. Assoc. J. 41,576583. (31) Bronfman, M. (1986)Activation of hypolipidaemic drugs to acyl coenzyme-A thioesters. Biochem. J. 239,781-784. (32) Kuslikis, B. I., Vanden Heuvel, J. P., and Peterson, R. E. (1991) Lack of evidence of perfluorodecanoyl- or perfluorooctanoylcoenzyme-A formation in male and female rata. The Toxicologist 11, Abstract 677. (33) Vanden Heuvel, J. P., Kuslikis, B. I., Shrago, E., and Peterson, R. E. (1991)Inhibition of long-chain acyl-CoA synthetase by the peroxisomeproliferator perflourodecanoicacid in rat hepatocytes. Biochem. Pharmacol. 42,295-302.

Registry No. PFOA, 335-67-1; PFDA, 335-76-2.