JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY APRIL 1991 VOL. 6 225 High-performance Liquid Chromatography-Atomic Absorption Spectrometry Interface for the Determination of Selenoniocholine and Trimethylselenonium Cations Application to Human Urine Jean-Simon Blais Alexis Huyg hues-Despointes Georges Marie Momplaisir and William D. Marshall Department of Food Science and Agricultural Chemistry Macdonald College 2 1 1 1 1 Lakeshore Road Ste-Anne de Bellevue Quebec H9X- 1 CO Canada The operation of a prototype high-performance liquid chromatography-atomic absorption spectrometry (HPLC- AAS) interface based on thermochemical hydride generation (THG) was characterized for the determination of sel- enonium compounds. Methanolic solutions of analytes containing selenium were nebulized by a thermospray effect pyrolysed in a methanol-oxygen kinetic flame in the presence of excess of hydrogen and atomized in a micro-diffusion flame maintained at the entrance to an unheated quartz T-tube.Factorial models for predicting the performance of the interface-detector combination at different levels of five interface operating variables indicat- ed that the interface is compatible with both reversed- and normal-phase HPLC eluents and that variations in the five operating parameters within relatively wide ranges does not affect the analyte response appreciably (less than 50% variation in response). Co-injection of trimethylselenonium iodide [(CH,),Sel] with a 1 0-fold excess of other potential interferent onium ions did not affect the THG process significantly. A modified apparatus was used to study the composition of the gases produced from the pyrolysis of (CH,),Sel or SeO in the presence of either H or He. The product gases were condensed acidified and channelled through two consecutive trapping solutions to recover SeIV and hydrogen selenide (H,Se) separately from the product mixture.The analysis of these trapping solutions by HPLC-AAS demonstrated that in the presence of H both (CH,),Sel and SeO were thermochemical- ly reduced to H,Se whereas SeIV was the major product in post-pyrolysis atmospheres of He. A rapid isocratic HPLC separation was developed for the determination of selenoniocholine and trimethylselenonium cations and applied to human urine. Recoveries of both analytes when present at levels of between five- and ten-times the normal background level of total Se were 77% or better. The low cost high reproducibility and robust nature of this system make it a good candidate for the routine determination of several selenium compounds including other selenonium cations selenoamino acids and SeIV organometalloid species. Keywords High -performance liquid chroma tograp hy-a tomic absorption spectrometry; the rmochemical hydride generation; human urine; selenoniocholine; trimethylselenonium Several studies have demonstrated that the trimethylselenoni- um cation [(CH,),Se+] is a major metabolite of selenium (as selenide selenite or selenate) in mammals.’“ In controlled metabolic studies (CH,),Se+ has been determined as 7sSe by y- radiation counting or by autoradiography after ion- exchange purification and precipitation as the Reineckate paper chromatography3-” or high-performance liquid chromatography (HPLC).6.x,y This organometalloid has also been determined indirectly by wet oxidation and fluorimetric quantification of Sel” as the 2,3-naphthalenediamine deriva- tive.’@I2 However this approach does not discriminate between (CH,),Se+ and other potential selenonium metabo- lites such as selenoniobetaine [(CH,),Se+CH,COOH] or selenoniocholine [(CH3)2Se+CH2CH20H].Neutron activation analysis of ion-exchange chromatographic eluates provides both selectivity and a low limit of detection,’,-’-” but because of its technical complexity and high cost this technique may not be readily available to all researchers. High-performance liquid chromatography coupled with de- tection by atomic absorption spectrometry (HPLC-AAS) repre- sents a valuable tool for the determination of metals and metalloids.16 However the requirements for the successful on- line coupling of HPLC and AAS instruments are demanding.The atomization cell e.,?. kinetic flame electrothermal furnace or plasma must be capable of handling voluminous flows of organic or aqueous eluents (typically 0.1-3 ml min-I of liquid is vaporized to about 400-700 1 min-I of gases at normal detector operating temperatures). Ideally the interface module should permit the detection of low- to sub-nanogram amounts of trace elements emerging from the HPLC column yet it should remain robust and inexpensive to construct and operate. Post-column hydride generation (using sodium tetrahydrobo- rate and dilute acid) circumvents problems associated with sample introduction in conventional AAS techniques.This ap- proach has been developed successfully for the determination of ar~enic’~.’~ and tin” using electr~therrnally~~~~~ or flameIx heated quartz tube atomizers. Nevertheless this technique is limited to the reducible physico-chemical forms of the ele- ments which can be volatilized as stable hydrides. Several bio- genic organometalloid compounds are fairly inert and are not predicted to react with reducing agents to form volatile hydride derivatives. Included among these are compounds in which the metalloid occurs in its lowest oxidation state such as selenoni- um species selenoamino acids”) and selenonucleosides.2’.22 Recently a thermochemical hydride generator was devel- oped as an interface for HPLC-AAS and optimized for the de- termination of low-nanogram amounts of arsenobetaine arsenocholine and tetramethylarsonium cations.’ This on-line interface was based on thermospray nebulization of the HPLC methanolic eluent pyrolysis of the analyte in a methanol- oxygen flame thermochemical derivatization using excess of hydrogen and cool diffused flame atomization of the product(s) in a quartz cell mounted in the AAS optical beam.A thermoc hemical hydride generation (THG) mechanism was suggested by indirect evidence. In this paper direct evidence supporting the gas-phase hydride generation mechanism for selenium is presented and a method for the determination of selenoniocholine and (CH,),Se+ in human urine is reported.Experimental Reagents and Standards All solvents were ‘distilled in glass’ grade or ‘pesticide analy- sis’ grade (Caledon Georgetown Ontario Canada). Certified226 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY APRIL 1991 VOL. 6 ACS-reagent grade hydrochloric and acetic acids were used. Triethylamine was purified gold-label grade and other chemicals were laboratory-reagent grade or better (Aldrich Chemicals Milwaukee WI USA). Water was doubly dis- tilled and deionized. The synthesis purification and charac- terization of trimethylselenonium iodide [(CH,),SeI] and (2- hydroxyethyl)dimethylselenonium (selenoniocholine) tetra- phenylborate standards have been described elsewhere.’? Naphtho[ 2,3-c-][ 1,2,5]selenadiazole (naphthalenepiazselenol) and bis(2,4-dinitrophenyl) selenide were prepared and purified as described previou~ly.’~.” Stock solutions of (CH,),SeI ( 1.01 x lo4 g ml-I) and selenoniocholine tetraphenylborate (2.06 x lo4 g ml-I) were prepared in methanol and stored at -40 “C. The addition of 10% (v/v) acetone was necessary to dissolve selenoniocholine tetraphenylborate.Dilution of these stan- dards in methanol containing 1% (v/v) acetic acid and 0.05% (v/v) triethylamine provided working standards. Instruments The instrumentation used for this study consisted of an HPLC system (Beckman Fullerton CA USA Model 100 A pump) an autosampler (LKB Stockholm Sweden Model 2157) and an atomic absorption spectrometer set at 196.0 nm (Philips Cambridge UK Model PU9100) which was equipped with a high energy selenium hollow cathode lamp (Photron Super Lamps System Victoria Australia) and a deuterium back- ground correction system.The optimization experiments (without chromatography) were performed with deuterium background correction. Because the background correction almost tripled the background signal of the AAS detector the chromatographic calibrations were performed without deuteri- um background correction. Narrow-bore stainless-steel tubing (i.d. 0.007 cm) was used as a post-injector. The silica transfer line (i.d. 50 pm) was connected to the HPLC tubing through a capillary reducing union (Chromatographic Specialties Brock- ville Ontario Canada). HPLC Conditions Naphthalenepiazselenol and bis(2,4-dinitrophenyl) selenide were separated on a Nucleosil C column (0.46 x 15 cm 3 pm particles CSC Montrkal Canada) using 100% methanol as the mobile phase (0.5 ml min-I). For the interface optimiza- tion the selenonium standards (100 p1 injections) were separ- ated on a cyanopropyl bonded phase [5 pm silica support 0.46 mm i.d.x 15 cm LC-CN (Supelco Bellefonte PA USA)] with methanol (0.65 ml min-I) containing 0.05% (v/v) triethylamine and 1% (v/v) acetic acid. For the determination of these compounds in urine diethyl ether (29% v/v) and tri- methylsulphonium iodide (0.2 mg ml-I) were added to the mobile phase. THG Interface The construction of the quartz interface for the HPLC-AAS method was as described previously.’3 Briefly it consisted of a quartz T-tube (an upper optical tube and a lower analytical- flame tube) with an added side arm which met the analytical- flame tube at an angle of 45”.This side arm contained a pyrol- ysis chamber which was fitted with separate inlets for 0 and H,. During the operation of the THG the liquid HPLC column eluate was thermosprayed into and combusted in the upstream region of the pyrolysis chamber. The product gases were reacted with H2 and entrained through a small O2 supported an- alytical flame maintained just below the optical beam of the spectrometer. [Caution The thermospray region of the inter- face must be heated to normal operating temperature prior to turning on the HPLC pump. Vapours of methanol (the result of an unsuccessful ignition of the thermospray) must not be allowed to accumulate in the interface.] Optimization The interface was optimized for the detection of (CH,),Se+ using a response surface methodology (half-replicate 2s com- posite design) as described previously.’3 In this approach 2n + 1 (n = number of independent variables = 5) data points were added to the half-replicate 2s factorial design (which is suitable only for estimating linear and interaction effects) allowing the fitting of a second-order model (to include quadratic effects) to search for the optimum response.Thus (j x 25) + [(2 x 5 ) + I ] = 27 data points were required. The five variables studied were the flow-rates of (i) 0’ (OT 500-4300 ml min-I) and (ii) H (1.00-2.40 1 min-I) to the pyrolysis chamber (iii) 0 (OA 100-240 ml min-I) to the analytical flame (iv) HPLC mobile phase flow-rate (0.3&1 .OO ml min-I) and (v) the percentage of diethyl ether (040%) or water (040%) in the methanolic mobile phase which also contained 1 % (v/v) glacial acetic acid and 0.05% (v/v) triethylamine.The peak area of the atomic ab- sorption signal for (CH,),SeI (2 pg) was recorded in triplicate for each of the 27 operating conditions (HPLC column removed) spaced equally across the domain of the ranges described earlier. The AAS response to 1.2 nmol of trimethyl- selenonium iodide selenoniocholine tetraphenylborate sele- nomethionine selenium dioxide and sodium selenate standards were also recorded under these same operating conditions. The possible interference with the detection of Se by other orga- nometallic/metalloid cations was evaluated by co-injecting 1.2 nmol of (CH,),SeI together with a I - or 10-fold molar excess of interferent [(CH,),AsI (CH,),SI or (CH,),PbCI].Cal- ibration graphs for selenoniocholine and (CH,),Se+ were ob- tained by HPLC-THG-AAS analysis of sequential dilutions of a fresh standard. The limit of detection (LOD) was determined from these calibration graphs using a first order error propaga- tion model with base-line noise normally di~tributed.?~ Trapping Apparatus A modified interface (Fig. 1 ) was used to study the products of reaction between the crude pyrolysis gases and either Hz or He. In operation the sample dissolved in methanol was nebu- lized by the thermospray effect combusted in an atmosphere of O2 and the crude pyrolysis products were reacted with either Hz or He.The products were acidified cooled and passed through two scrubbing solutions to trap Se’” and HzSe separately. The design of this modified interface was similar to the THG interface described except that the pyrolysis chamber tube and the quartz outlet tube were smaller (4mm i.d. x 6 mm 0.d. and 2 mm i.d. x 4 mm 0.d. x 45 cm respec- tively). In addition to the gas inlets for 0 and H a third inlet to the pyrolysis chamber was added 2.5 cm downstream from the H inlet to permit the introduction of 2 ml min-’ of HCI ( 1 mol dm->) using a peristaltic pump [Eyela (Tokyo Japan) Model MP-31. The outlet of the modified THG ‘train’ was in- serted into a water-cooled condenser [E 1.27 0.d. x 37 cm copper tube equipped with a water inlet and outlet and sealed with brass Swagelok unions (J 1.27 x 0.64 cm)].One end of a polytetrafluoroethylene (PTFE) tube (F 2.48 mm i.d. x 4 mm 0.d. x 50 cm) was thermally sealed to the quartz exit transfer line and the other end was connected to a 250 ml Erlenmeyer filtering flask (G trap 1 ) which contained 190 pmol of 2,3- naphthalenediamine in 150 ml of 1 mol dm- HCI. The gases emerging from trap 1 were channelled iio a PTFE tube (I) to a second trap (H trap 2) containing 18 mmol of NaHCO and 500 pmol of l-fluoro-2,4-dinitrobenzene (FDNB) dissolved in 150 ml of dimethylformamide and water [7 + 3 (v/v)]. The 2,3 -naph t halened iam i ne and the 1 -fl uoro-2,4-dini trobenzene solutions were used to convert Se’” into naphthalenepiazsele- no1 and to transform H,Se into bis(2,4-dinitrophenyI) selenide respectively.The trapping solutions were stirred magnetically. Experiments were carried out by injecting 20 volumes (0.2 ml) of (CH,),SeI dissolved in methanol (0.537 pmol ml-I) orJOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY APRIL I99 I VOL. 6 227 D 0 H HCI I I Water E H 4 -G Fig. 1 Thermochemically generated hydride trapping apparatus consist- ing of A capillary transfer line; B silica guide tube; C thermospray heating coil surrounding a thermospray-pyrolysis quartz tube 4 mm i.d. x 14 cm with inlets (2 mm i.d. x 3 cm) for 02 H and 1 rnol dm-3 HCI; D Swagelok reducing unions to position the transfer line and the guide tube within the thermospray tube; E condenser jacket (1.27 x 37 cm); F and I (4 mm 0.d.) PTFE transfer lines; G and H 250 ml filtering flasks contain- ing 2,3-naphthalene diamine (trap I ) and 1 -ftuoro-2,4-dinitrobenzene (trap 2); and J modified Swagelok fittings to provide a watertight seal between the condenser and the silica exit tube (2 mm 0.d.x 45 cm) SeOz (0.955 pmol ml-I) into the system via an HPLC injection valve under the following conditions methanol flow-rate 0.5 ml min-I; heating element current 6 A; O2 flow- rate 600 ml min-'; H2 flow-rate 1.7 I min-'; and I rnol dm-3 HCI flow-rate 2 ml min-I. At the termination of the experiment the trapping solution from trap 1 was stirred for 30 min then extracted three times with 50 ml of benzene. The organic phases were combined washed three times with 50 ml of 1 rnol dm-3 HCI dried with generous amounts of Na,SO filtered through a Whatman No.1 filter-paper and the filtrate was evaporated to dryness under vacuum. The residue was diluted to 1 ml with methanol. Excess of FDNB in the trapping solution from trap 2 was reacted at room temperature with 1 mmol of glycine for 2 h diluted to 400 ml with 0.01 rnol dm-.' NaOH solution and ex- tracted with three successive 50 ml portions of benzene which were combined and back-extracted three times with 100 ml of 0.01 rnol dm-.' NaOH solution. The benzene extract was treated as described for the first trapping solution. The final ex- tracts were analysed (20 p1 injections) by HPLC-THG-AAS. (Caution Benzene is highly toxic and appropriate precautions should be taken.) Isolation of Selenonium Standards From Urine Urine (10 ml) which had been spiked with 10 pg each of sele- noniocholine tetraphenylborate and trimethylselenonium iodide was diluted with 100 ml of absolute ethanol and chilled in a dry ice-acetone bath for 20 min according to the method of Kraus et aLx The resulting precipitate was separated from the liquid by refrigerated (-15 "C) centrifugation at about 40008 for 15 min.The supernatant was evaporated to dryness and the residue dissolved in 10 ml of water was applied to the head of an anion-exchange column (0.5 x 10 cm of Dowex 2x8). The column was washed with additional distilled water to give a total volume of 30 rnl of column eluate which was acidified to pH 3 with HCI. The acidified aqueous solution was extracted four times with liquified phenol ( I x 10 and 3 x 5 ml). The phenol extracts were combined and back-extracted three times with water ( 1 x 10 and 2 x 5 ml). The phenol layer was diluted with 75 ml of diethyl ether then extracted three times with 5 ml of distilled water to recover the analytes.The aqueous washes were combined and back extracted with 3 x 5 ml of diethyl ether. The aqueous phase was evaporated to dryness and the residue was re-dissolved in methanol and re-concentrated to 1 ml. Aliquots (50 pl) were analysed by HPLC-AAS. Results and Discussion Post-HPLC Hydride Generation The prototype thennochemical hydride generation HPLC- AAS interface was developed2? as an analytical tool for study- ing the fate of potential biogenic arsonium and selenonium compounds. The HPLC-inductively coupled plasma (ICP) spectrometry,2N HPLC-ICP mass spectrometry29 or HPLC- graphite furnace AAS") instruments which have been used to determine arsonium compounds would probably be very suit- able for the analysis of selenonium species.However their higher purchase price and operating costs limit their availabil- ity to many researchers. As was observed previously for arso- nium compounds,2' the severe spectral interference of the kinetic flame at low AAS wavelengths precluded the use of a direct thermospray-micro-atomizer interface3' for the detec- tion of trace amounts of Se compounds. Alternative detection techniques for Se were sought. The THG interface23 was con- sidered to include at least three steps the thermospray induced nebulization pyrolysis and atomization of the organometallic analyte;" a postulated THG using hydrogen radicals; and dif- fusion-flame atomization.'? " The last technique has been re- ported to provide an absolute atomization of H,Se.'3 The atomization process is considered to be mediated by reactions with hydrogen radicals which are generated in the active zone of the diffusion flame.This spatially limited cloud of free radi- cals which does not extend to the AAS optical beam virtually eliminates spectral background noise. The thermochemical H2Se generation mechanism of the in- terface was corroborated by the fact that no AAS signal was observed in the absence of the diffusion flame or in the absence of post-thermospray H,. A confirmation of the ther- mochemical derivatization of (CH,),SeI and SeO to H2Se was obtained by chemical-trapping experiments using the apparatus presented in Fig.1. The hot product gases from the pyrolysis chamber were reacted with either H or He then mixed with 1 rnol dm-3 HCl cooled to room temperature in a condenser (E) and channelled to an acidic solution containing 2,3-naphthalenediamine (G trap I ) which converted Se'" into naphthalenepiazselenol.2h Pyrolysis products which were carried through the acidic trap 1 were channelled to an alka- line solution (H trap 2) containing FDNB to convert H,Se into bis(2,4-dinitrophenyI) selenide. After the trapping experi- ments excess of FDNB was reacted with glycine to form the base-soluble 2,4-dinitroanilinoacetic acid.M The organo- soluble content of trap 1 was extracted from the acid solution into benzene and the residue obtained after solvent removal was dissolved in methanol.The postulated bis(2,4- dinitrophenyl) selenide derivative" formed in trap 2 was ex- tracted from the basic aqueous solution with benzene evapo- rated to dryness and the residues were re-dissolved in methanol. The extracts from both traps were analysed directly by HPLC-THG-AAS. Chromatograms of the standards and of the products isolated from the trapping solutions after ther- mospray pyrolysis of (CH,),SeI are presented in Fig.2. The recoveries of naphthalenepiazselenol and bis(2,4- dinitrophenyl) selenide when the pyrolysis products of (CH,),SeI or SeOz injections were reacted with H2 or He are recorded in Table I . These results indicate that both SeO and (CH,),SeI are converted into H,Se but only in the presence of Hz. As a result of the high gas flow-rates occurring in the trapping apparatus a quantitative recovery of the products was not anticipated.Under an inert post-pyrolysis atmo- sphere (He) a portion of injected selenium (Se'") was re- covered in trap 1 but an appreciable amount remained deposited in the condenser tube as metallic red selenium. In both instances a portion of the naphthalenepiazselenol was en-228 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY APRIL 1991 VOL. 6 r- o! Time - Fig. 2 HPLC-THG-AAS chromatograms resulting from the trapping ex- periments with (CH,),SeI. Chromatogram (a) bis(2,4-dinitrophenyl) sele- nide (3.94 min) and naphthalenepiazselenol (5.28 min) standards; chromatogram (h) pyrolysis performed under He atmosphere; chromato- gram (c) pyrolysis performed under H atmosphere. Chromatograms A and B are extracts from trap 1 (Se" specific) and trap 2 (Se" specific) re- spectively Table 1 Amounts and percentage recoveries of naphthalenepiazselenol (NPSe) and bis(2.4-dinitrophenyl) selenide [(DNP),Se] from chemical traps after the pyrolysis of (CH,),SeI or SeO in the presence of H or He Analyte Atmosphere Trap Amount Recovery Amount Recovery of (%I of (%I NPSe/ (DNP),Se/ pmol pmol (CH,),SeI* H 1 NDt - ND - 2 ND - 0.67 31.2 He 1 0.36 16.7 ND - 2 0.20 9.3 ND - - ND - I ND - 1.13 29.6 H? 2 ND 2 0.11 2.9 ND - He 1 0.78 20.4 ND * 2.15 pmol injected.t ND = none detected. $ 3.82 pmol injected. supply to the pyrolysis chamber by dinitrogen oxide (produc- ing a white flame >1665 "C) resulted in a complete loss of the AAS analyte (500 x LOD) signal which may reflect a rapid thermal degradation of the hydride at this higher post- thermospray temperature.Replacing the thermospray 0 by air decreased the AAS response by about 40%. In this instance a massive air flow-rate (2 I min-I) was required to maintain the thermospray flame and the lower response may be attributed to a lower residence time of the H,Se in the analytical flame. Under optimum conditions for selenonium compounds the in- terface resulted in equivalent responses for selenomethionine selenoniocholine and selenium(1v) oxide (Table 2) however sodium selenate (Na,SeO;-) was less efficiently derivatized and detected with the THG procedure. The relative AAS response of Se [as (CH,),SeI 1.2 nmol] co-injected with an equimolar amount and a 10-fold molar excess of potential interferents As (as tetramethylarsonium iodide) S (as trimethylsulphonium iodide) and Pb (as trimeth- yllead chloride) are presented in Table 3.In each instance the response to selenium remained virtually unaffected. As was the situation for the earlier thermospray-micro-atomizer inter- face,,' this THG device was not compatible with alkaline earth metals (degradation of the quartz) halogenated solvents (reduced response) or a high proportion of water (>60%) in the HPLC mobile phase (disruption of the thermospray). Table 2 Relative AAS responses to Se compounds of equimolar amounts (1.2 nmol) recorded under optimum conditions Anal yte Relative response (%) Trimethylselenonium iodide Selenomethionine Selenoniocholine Selenium(1v) oxide Sodium selenate 100 f 2* 99 k 2 104+3 101 + 2 18k I * Standard deviation based on three replicate analyses.Table 3 nmol) recorded under optimum conditions Effect of potential interferents on the response to (CH,),SeI (1.2 Analyte Interferent Relative amount Response (molar ratio) (%I (CH,),SeI - - 100f2* l00f 1 10 9 6 k 2 (CH3),AsI 1 (CH,),SI 1 99 f 0.5 10 9 4 k 2 10 95 k 0.8 (CH,),PMII 1 97 k 2 trained into trap 2. These data corroborate the postulated THG mechanism which is most probably mediated by hydrogen radicals. The fact that no Selv was detected in trap I in an H atmosphere suggests a direct thermochemical derivatization of an oxidized species (SeIV in this instance) to its hydride. One mechanism which may explain this phenomenon is initiated by hydrogen radicals H'+OSeO 4 OH'+SeO' ( 1 ) SeO'+H + HSe'+OH' (2) HSe'+HZ 4 H,Se+H' (3) Such a reaction sequence is considered to occur in a hot spa- tially limited volume around the H inlet.In this hypothetical process the temperature-sensitive final product (H,Se) has to be stabilized rapidly by the cooling effect of the massive flow- rate of Hz. This supposition has been corroborated by addition- al experiments with the THG-AAS interface. Replacing the 0 * Standard deviation based on three replicate analyses. Optimization of the THG Interface The interface operating parameters were optimized using a multivariate methodology based on a half-replicate 25 compo- site These experiments were carried out by record- ing the instrumental response to (CH,),SeI under different combinations of interface operating parameters which includ- ed the flow-rates of oxygen (OT) to the pyrolysis chamber; hydrogen to the pyrolysis chamber; oxygen (OA) to the analyt- ical flame; and the HPLC mobile phase.A fifth variable was introduced into the model to mimic typical normal- or re- versed-phase HPLC eluents (040% diethyl ether or 040% water in a methanolic mobile phase). Both mathematical models [reversed-phase (RP) and normai-phase (NP) eluents] resulting from the statistical analysis of the data were fairly ac- curate with average relative deviations between the observed229 JOURNAL OF ANALYTlCAL ATOMIC SPECTROMETRY APRIL 1991. VOL. 6 and predicted responses of 8.5 and 5S% respectively. The linear regressions of the observed on the predicted response were correlated at r = 0.9164 and 0.9425 for RP and NP models respectively.Unmodelled variations in the perfor- mance of the interface at extreme parameter values resulted in some outlier predictions. These unmodelled variations were thought to be caused by a rapid accumulation of carbon depos- its (which could be removed by increasing the OA flow-rate temporarily) that occurred for 3040% diethyl ether and at 500 ml min-I OT and the necessity to re-adjust the position of the capillary to obtain a stable thermospray for 40% water and at 500 ml min-I OT. The accuracy of the models was consid- ered sufficient to estimate the effect of individual variables and to determine optimum parameters visually using surface- response plots.The effects of the five variables were character- ized by plotting two selected variables versus response while keeping the three others constant at the centre of the design. The curvature of these plots (Fig. 3) provided valuable infor- mation for a tentative characterization of the effects. In the NP model (methanolic mobile phase containing diethyl ether) the combination of mobile phase flow-rate and thermospray oxygen (OT) flow-rate [Fig. 3(a); proportion of diethyl ether = 20% H flow-rate = 1.7 1 min-I and OA = 170 ml min-'1 predicted a maximum response at intermediate values (OT = 650 ml min-I; mobile phase flow-rate = 0.65 ml min-I). A minimum response occurred at low OT and high mobile phase flow-rate; these conditions resulted in a short thermospray flame and a slow accumulation of carbon deposits in the THG combustion chamber. A similar pattern was ob- served when the predicted response was plotted as a function of the diethyl ether content in the mobile phase and as a func- tion of OT [Fig.3(h)]. In this instance the THG-AAS re- sponse was decreased by about 40% (relative to maximum) at high diethyl ether content and low OT reflecting an incom- plete combustion of the mobile phase. When combined with a proper level of OT the presence of diethyl ether appeared to be beneficial to the response. Thus the optimum response for (CH,),Se+ was obtained when the OT mobile phase flow-rate and proportion of diethyl ether were at intermediate values close to the centre of the design. Relative to the NP eluent the presence of water in the mobile phase was generally detrimen- tal to the response. In the presence of 20% water a maximum response was observed at low mobile phase flow-rate and high OT [Fig.3 ( c ) ] . The last condition resulted in a relatively cool post-thermospray atmosphere. The combined effects of pro- portion of water and OT somewhat corroborated this observa- tion [Fig. 3(d)]. The selenonium analyte was most efficiently derivatized in a cooler (high water content) post-thermospray environment. Optimum THG parameters for the analysis of selenonium analytes were determined from surface response plots of OT versus H flow-rate at different levels of analytical oxygen (OA) with the two other variables fixed at optimum chromato- graphic values (mobile phase flow-rate = 0.65 ml min-I; pro- portion of diethyl ether = 0%) for the selenonium compounds. These plots are presented in Fig.4. At low OT the thermo- spray flame was short and the response was generally lower (with a maximum occurring around a 1.7 1 min-I H flow- rate) presumably reflecting an inefficient pyrolysis of the analyte. At high OT and low H flow-rate this flame was vigo- rous but a minimum response was still observed most prob- ably because of a rapid consumption of H by excess of 0,. Increasing the flow-rate of H resulted in higher responses up to a maximum after which a secondary H,-O flame was ignited at the H inlet tube thereby affecting the efficiency of the THG. The detrimental effect of this secondary flame cor- roborated previous observations that excessive post- thermospray temperatures were detrimental to the sensitivity of the THG-AAS. Again the maximum response occurred at intermediate levels of H and thermospray 0 flow-rates.The second-order interaction between H and OA (level of analyti- cal oxygen not presented) corroborated the previously report- ed characteristics of this atomization mechanismJ3 which requires a fuel-rich flame. A response minimum observed at a 51 < 46 v I 2 41 36 \ Y 8 31 0 2 6 21 v) 57 a 37 Fig. 3 Surface response plots of the predicted variation in peak area with (u) and (c) oxygen flow-rate (OT) to the pyrolysis chamber and with the HPLC mobile phase flow-rate and ( h ) and (d) with the proportion of eluent modifier (diethyl ether or water) \~ers~ts OT. The other variables were maintained con- stant at the centre of the experimental design230 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY APRIL I99 1 VOL.6 49 2 45 f 41 2 3 37 t! 0 33 * 3 29 CL 25 1. 49 v) a 45 * 41 37 0 Y 3 33 a 29 1 Fig. 4 Surface response plots (normal-phase model) of the predicted vari- ation in response with hydrogen (H2) and O2 (OT) flow-rates to the pyroly- sis chamber at different flow-rates of O2 (OA) to the analytical flame. A OA = 170 ml min-I; B OA = 205 ml min-I; C OA = 240 ml min-I. Both mobile phase flow-rate and diethyl ether content were maintained constant at 0.65 ml min-' and 0%. respectively low OA resulted from an appreciable reduction in the volume of the analytical cool diffusion flame. From these surface response plots the optimum operating parameters were determined to be mobile phase flow-rate = 0.65 ml min-I; proportion of diethyl ether = 0%; OT = 725 ml min-I; Hz flow-rate = 2.03 1 min-I; and OA = 170 ml min-I.Over the ranges of operating parameters studied the difference between low and high responses was generally less than 50%. Thus the performance of this interface was only moderately affected by appreciable variations in the levels of the five vari- ables. This valuable characteristic was reflected in the fairly high reproducibility of the instrument. HPLC of Selenonium Compounds Strong cation-exchange chromatography (sulphonate based stationary phase) appeared to be the method of choice for sep- arating selenonium compounds.x However this approach was incompatible with the THG interface because of the high pro- portion of water required in the mobile phase.With optimum flow-rates of 0.24.5 rnl rnin-I a micro-bore (0.21 cm i.d.) strong cation-exchange column coupled with post-column methanol enrichment of the eluate would have been desirable I I I 0 3 6 9 Ti me/m i n Fig. 5 HPLC-THG-AAS chromatogram of A selenoniocholine; and B trimethylselenonium standards (500 ng of each) recorded under optimum conditions in this particular instance. However this packing was not com- mercially available in a micro-bore format. The trimethylselenonium and selenoniocholine cations were separated on a cyanopropyl stationary phase with a methanolic mobile phase containing a silanol masking agent [triethyl- amine 0.05% (v/v) and acetic acid 1 % (v/v)]. A chromatogram of these standards recorded under optimum THG conditions is presented in Fig.5. Although the two selenonium analytes were totally resolved the (CH,),Se+ was eluted as a broader peak which substantially decreased its LOD. The addition of up to a 500-fold (m/m) excess of ammonium acetate in the standard solution did not appreciably affect the background signal nor the retention times of the analytes. Linearity Reproducibility and Limits of Detection The LOD for each analyte was calculated from the correspond- ing calibration graph under optimized conditions using a first order error propagation model .24 The linear calibration models were highly correlated [ I - = 0.9996 and 0.9986 for selenonioch- oline and (CH,),Se+ respectively] in the concentration range studied (100 ng - 2.5 pg as selenonium salt).The calculated LOD of each analyte (as free cation) was selenoniocholine = 31.3 ng and trimethylselenonium = 43.9 ng. Subsequent studies with the same column indicated that the addition of tri- methylsulphonium iodide to the mobile phase improved the chromatography of these analytes appreciably which lowered the LODs substantially. The short-term reproducibility of the THG interface (based on three replicate analyses) for different concentrations of each analyte is reported in Table 4 The long-term reproducibility (6 h n = 6 ) recorded at 10 x LOD was selenoniocholine tetraphenylborate = 3.1 % and (CH,',,SeI = 5.8%. Analysis of Human Urine The THG technique was then applied to the determination of selenoniocholine and (CH,),Se+ in human urine.Urine ( 10 ml) which had been spiked with 1 mg I-' each of selenoniocho- line tetraphenylborate and trimethylselenoniurn iodide was de-salted with ethanol according to the method of Kraus et a/.' After centrifugation the supernatant was evaporated toJOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY APRIL I99 I VOL. 6 23 1 Table 4 Reproducibility of the HPLC-THG-AAS response to various analyte levels recorded under optimum conditions Amount/ Selenoniocholine Trimethylselenonium Pt.2 tetraphenylborate iodide 0.1 10.7 11.6 0.5 2.5 1.4 I .o 1.9 2.8 2.5 0.6 1.2 RSD (%)* RSD (5%) * Relative standard deviation based on three replicate analyses. 0 3 6 9 12 Time/min Fig. 6 HPLC trace of selenoniocholine (at 7.05 min); and trimethylsele- nonium (at 8.39 min) standards which had been spiked (at 0.173 and 0.315 pg ml-' of Se respectively) into human urine.The solvent front peak appears at 3.29 min. The cyanopropyl column was eluted isocratically with 0.65 ml min-' of mobile phase containing methanol (70% v/v) diethyl ether (29% v/v) glacial acetic acid ( I % v/v) triethylamine (0.01 o/c v/v) and trimethylsulphonium iodide (20 mg per 100 ml) dryness and the residue was re-dissolved in water. The result- ing aqueous solution was passed through an ion-exchange column then acidified to pH 3 and extracted four times with liquified phenol. The phenol phases were combined back- extracted with water then diluted with diethyl ether. Finally the analytes were recovered by extracting the phenol-diethyl ether mixture with water.The water extracts were combined washed with diethyl ether evaporated to dryness and the residue was re-dissolved in methanol and analysed by HPLC- AAS using the cyanopropyl column and a mobile phase con- sisting of methanol [70% (v/v)] diethyl ether [29% (v/v)] and glacial acetic acid [ 1 % (v/v)] containing triethylamine [O.Ol% (m/v)] and trimethylsulphonium iodide (0.2 mg ml-I). Recoveries the average of three replicate determinations (k one standard deviation) were 77 f 4 and 85 f 0.4% for (CH,),Se+CH2CH,0H and for (CH,),Se+ respectively. The spiking level [173 and 315 ng ml-I of Se for (CH,),Se+CH,CH,OH and (CH,),Se' respectively] was delib- erately chosen to be from five- to ten-fold higher than back- ground levels for total Se encountered in normal human urine8 [30-50 ng ml-l of Se of which some 10% is considered to be (CH,),Se+]. Neither analyte was detected in several samples of control urine although the LODs using the same model as above and the modified solvent system were 5 and 7 ng (as Se) for (CH,),Se+CH2CH,0H and (CH,),Se' respectively.No difficulties with the chromatography (Fig. 6) or with the detec- tion of these analytes at this spiking level were encountered. The final extract could have been further concentrated and 100 pl injections of sample could have been used without compro- mising the method. Conclusion The concentrations of selenonium metabolites in biological samples remain to be determined. In order to provide a method with an LOD for Se in the 1-20 ng g-' range the ex- traction protocol to be developed for these analytes should be designed to allow the treatment of larger samples (5-25 g) and concentration of the final extract to less than 1 ml.Since the cool diffused flame atomizer has been shown to provide picogram sensitivites at gas flow-rates exceeding 5 I min-1,32.33 it is considered that the performance of this interface is limited by the efficiency of the hydride generation and hydride transport processes. This THG process appeared to be affected by high post-thermospray temperatures hence it is reasonable to anticipate that further research on this aspect will result in sub-nanogram LODs for Se and other hydride forming elements. Financial support from the Natural Science and Engineering Research Council of Canada is gratefully acknowledged. 1 2 3 4 5 6 7 8 9 10 1 1 12 13 14 1s 16 17 18 19 20 21 22 23 24 2s 26 27 References Byard J.L. Arch. Biochem. Biophys. 1969. 130 556. Palmer I. S. Fischer D. D.. Halverson A. W. and Olson 0. E.. Biochim. Biophys. Acta 1969 177 336. Palmer 1. S. Gunsalus R. P. Halverson A. W. and Olson 0. E. Biochim. Biophys. Arm 1970 208 260. Kiker K. W.. and Burk R. F. Am. J. Physiol. 1974,227,643. Nahapetian. A. T. Janghorbani M. and Young V. R. J . Nutr. 1983 113,401. Sun X. F.. Ting B. T. G. and Janghorbani M. Anal. Biochem. 1987,167,304. Ganther. H. E. Kraus R. J. and Foster S. J. Methods Enzvmol. 1987,143 195. Kraus J. K. Foster J. S. and Ganther H. E.. Anal. Biochem. 1985 147,432. Hoffmann. 1. L. and Connell. K. P. Arch. Biochem. Biophys. 1987 254 534.Watkinson J. H. Anal. Chem. 1966,38,92. Olson 0. E. Palmer 1. S. and Cary E. E. J . Assoc. Off. Anal. Chem.. 1975,58 117. Janghorbani M. Ting B. T. G. Nahapetian. A. and Young V. R.. Anal. Chem.. 1982,54 1188. Blotcky A. J. Hansen. G. T. Opelanio-Buancamino L. R. and Rack E. P. Anal. Chem. 1985,57 1937. Blotcky. A. I. Hansen G. T. Borkar N. Hebrahim A.. and Rack E. P. Anal. Chem. 1987,59 2063. Blotcky A. J. Ebrahim A. and Rack E. P.. Anal. Chem. 1988 60 2734. Ebdon L. Hill S. and Ward R. W.,Ana/yst. 1987 112. I . Ricci G. R.. Shephard L. S. Colovos G. and Hester. N. E. Anal. Chem. 1981,53,611. Haswell S . J.. O'Neil P. and Bancroft K. C. C.. Anal. Chim. Actu 1985.169 195. Bums D. T.. Glockling. F. and Harriott M. A n u / w 1981. 106,921. Stadtman. T. C. Ad\*. Eti:Fmo/. 1979.48. 1. Chen C. S. and Stadtman T. C.. Proc. Natl. Amd. &I. 1980 77 1403. Whitter. A. J. Stai. L.. Ching. W. M. and Stadtman T. C. Bio(,heni- istry 1984 23 4650. Blais. J. S. Mornplaisir. G. M. and Marshall W. D.. Anal. Chem. 1990.62. 1161. Foley. J. P.. and Dorsey J. G. Cht.omatogr.aphia 1984. 18.503. Huyghues-Despointes A. M.Sc. Thesis McGill University. 1990. Baytield R. F. and Romalis L. F. Atid. Biochent.. 1985. 144 569. Ganther H. E. and Kraus. R. J . A w I . B i o c h 3 m . . 1984. 138. 396.232 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY APRIL 1991 VOL. 6 28 Shiomi K.. Kakehashi. Y. Yamanaka H. and Kikushi T. App/. Or- ganomet. Chem. 1987 1 177. 29 Beauchemin D. Bednas M. E.. Berman S. S. McLaren. J. W. Siu K. W. M. and Sturgeon R. E. Anal. Chem. 1988,60,2209. 30 Lawrence J. F. Michalik P. Tam G.. and Gonacher H. B. S.. J . Agi'ic. Food Chem. 1986 34. 315. 31 Blais J. S. and Marshall W. D. J . A n d . At. Spectrom.. 1989. 4,271. Paper 01024070 32 Siemer D. D. Koteel P. and Jawiwala V. A n d . Chem. 1976 48 Received May 30th 1990 836. Accepted November 20th 1990 33 Dedina. J. and Rubeska I. Spectrochim. Actu Purr B. 1980 35 119. 34 Bunnett J. F. and Hermann D. H. Biochemistry 1970,9 816. 35 Hill W. J. and Hunter W. G. Technometrics 1966 8 571.