Automated determination of nitrogen in milk products - Analytical

Sandra L. Larson and Harvey P. Peterson. Analytical Chemistry ... Influence of Curd Firmness at Cutting on Yield and Recovery of Milk Constituents. D...
0 downloads 0 Views 623KB Size
(9) J. Pavel, R. Kuebler, and ti. Wagner, Microchem. J., 15, 192 (1970).

fully fluorinated fluoro-organic compounds. Furthermore, benzoic acid was found t o be as efficient as NaN03 when used in the PbClF method. The procedure developed gave satisfactory results (Table 11) for the whole series of fluoro-organic compounds analyzed. The absolute error is f0.60%, and the total average recovery is 99.30%. Although these results are generally less accurate than those expected (11)for the fluoride ion-sensitive electrode methods, yet the present work offers two new indirect polarographic methods that can also serve valuably in other cases.

(IO) D.A. Shearer and G. F. Morris, Microchem. J., 15, 199 (1970). (11) W. Selig, Fresenius'Z. Anal. Chem., 249, 30 (1970). (12) N. I. Larina and N. E. Gel'man, Zh. Vses. Khim. Obshchest., 15, 231 (1970). (13) E. C. Olson and S. R. Shaw, Microchem. J., 5, 101 (1961). (14) M. E. Fernandopulle and A. M. G. Macdonald, Microchem. J., 11, 41 (1966). (15) G. Ingram, "Methods of Organic Elemental Microanalysis," Reinhold, New York, N.Y., 1962, p 209. (16) I. M. Kolthoff and J. J. Lingane. "Polarography," 2nd ed., lnterscience Publishers, New York and London, 1952. (17) Y. A. Gawargious, G. M. Habashy, and E. N. Faltaoos, lndian J. Chem., 7, 610 (1969). (18) G. M. Habashy, Y. A. Gawargious, and E. N. Faltaoos, Talanta, 15, 403 (1968). (19) A. M. G. Macdonald. Analyst (London), 86,3 (1961). (20) W. I. Awad, Y. A. Gawargious, S. S. M. Hassan, and N. E. Milad, Anal. Chim. Acta, 36, 339 (1966). (21) W. I. Awad. S.S. M. Hassan, and M. E. Elsayes, Mikrochim. Acta, 1969, 688.

LITERATURE CITED A. M. G. Macdonald. Advan. Anal. Chem. instrum., 4, 100-103 (1965). T. S.Ma, Anal. Chem., 30, 1557 (1958). T. S.Ma and M. Gutterson, Anal. Chem., 42, 105R (1970). R. Belcher and A. M. G. Macdonald, Mikrochim. Acta, 1957, 510. T. S.Ma, Microchem. J., 2, 91 (1958). W. Selig. Fresenius'Z. Anal. Chem., 234, 261 (1968). R. Belcher, E. F. Caldas, S. J. Clark, and A. M. G. Macdonald, Mikrochim. Acta, 1953, 283. W. Schoniger, Mikrochim. Acta, 1956, 869.

RECEIVEDfor review April 22, 1974. Accepted September 11, 1974.

Automated Determination of Nitrogen in Milk Products H. W. Schafer and N. F. Olson Department of Food Science, University of Wisconsin-Madison, Madison, Wis. 53706

Procedures were developed to determine concentrations of nitrogen in milk and cheese and fractions of these products by a modlflcatlon of the Kjeldahl method using the Technicon AutoAnalyzer. Mean percentage differences between concentrations of nitrogen determined by the automated and AOAC methods ranged between 0.7 and 2.9% of the nitrogen concentratlon measured in the various types of samples. Much better precision between replicate measurements was obtained with the automated method. No statistically significant differences were observed between methods in measurement of total nitrogen in milk and cheese and of noncasein and nonprotein nitrogen in milk. Statistically significant differences were found in measurement of noncasein nitrogen of cheese and total nitrogen of cheese whey. Differences between the methods were equal to 2.9% of noncasein nitrogen concentration in cheese and 1.1 % of nitrogen concentration In whey.

Measurement of recovery of milk proteins during cheese manufacture and monitoring subsequent proteolysis of cheese are important observations in quality control and research functions of the cheese industry. The time and expenditures required for manual nitrogen analysis have led to the development of automated Kjeldahl nitrogen analyses (1-7) and alternatives to the Kjeldahl method (8-11). Automated Kjeldahl analyses of organic materials using the Technicon AutoAnalyzer generally required adjustment of digestion conditions and choice of a suitable standard for the type of sample being analyzed. The standards included @-alanineand freeze-dried meat for meat (12-14 ), 2-benzyl-2-pseudo-thiourea for milk (12-14 ), mixtures of nicotinamide and urine for urine, feces, and homogenates of hospital patients' diets ( 1 5 ) , and glycine for urine and feces samples ( 1 6 ) . Various ammonium salts have been

used as standards for soil and feed samples which were predigested in a block heater and analyzed with the Technicon AutoAnalyzer ( 7 ) . In the present study, total nitrogen content of milk, whey, and cheese and the amounts of noncasein nitrogen in cheese, and milk, and nonprotein nitrogen in milk were determined using the Technicon AutoAnalyzer I1 (17). Subsequent to initiation of this study, Kramme et al. ( 1 8 ) described a modified Technicon AutoAnalyzer system which was used for analysis of amino acid mixtures, casein, urea, and nicotinamide. The modifications included higher digestion temperatures, longer time for color development after adding phenol and hypochlorite, and expansion of the range of the recorder. This procedure allowed use of ammonium sulfate as a standard for all types of samples tested. Ammonium sulfate was used in another study, for analysis of a variety of foods, by increasing residence time of the sample in the digestor helix and adjusting the perchloric acid concentration and digestor temperature (19). However, agreement between this procedure and AOAC Kjeldah1 method was not very satisfactory for samples of condensed milk. The Kjedahl nitrogen analysis system of the Technicon AutoAnalyzer I1 was used without modification in our investigations. A stock standard of acid-precipitated, whole casein in distilled-ionized water or 0.1M sodium citrate was diluted with various solutions to make working standards for analysis of various milk-based products. We did not choose 2-benzyl-2-pseudo-thiourea as a standard as suggested by Brisson (13), because of nonlinearity of standard curves obtained during preliminary trials.

EXPERIMENTAL Apparatus. Analyses were made with the basic Technicon AutoAnalyzer I1 system consisting of sampler IV, proportioning pumps I and 111, continuous digestor, single-channel colorimeter,

ANALYTICAL CHEMISTRY, VOL. 47, NO. 3, M A R C H 1975

505

~

Table I. Comparison of Automated and Official Methods for Analysis of Total N and Noncasein N of a Solution of Cheese in 48% HzS04 (v/v) Replicate observations N Concentration, yg K I m P

Sample

Official

Standard deviation, ug N i m l

Difference

Automated

between methods, %

Trial

Trial

Official

Coefficient of variability, %

Automated

Official

Automated

Total N

1 2 3 4 5 6 7 8 mean

5 00 5 04 477 481 518 502 514 516

Trial

1

2

1

2

500 519 488 492 526 509 521 530

496 504 487 482 51 5 506 513

0.0 3.0 2.3 2.3 1.5 1.4 1.4 2.7 1.8

0.8 0.0 2.1 0.2 0.6 0.8 0.2

...

...

0.7

5.68 2.38 4.99 4.11 4.43 12.19 1.71 5.29 5.10

Trial

1

2

0.00 1.15 0.58 1.63 1.15 3.21 1.53 0.58 1.23

0.58 0.58 0.58 0.58 1.00 1.15 5.86

.. .

1.48

1

2

1.14 0.47 0.94 0.85 0.86 2.43 0.00 1.03 0.97

0.00 0.22 0.12 0.33 0.22 0.63 0.29 0.11 0.24

0.12 0.12 0.12 0.12 0.19 0.23 1.14

2.59 0.59 4.48 1.79 3.54 2.14 4.62 1.62 2.67

0.73 0.17 0.12 0.26 0.48 0.29 0.24 0.44 0.34

. ..

0.29

Noncasein N

1 70.2 67 . I 72.6 73.2 2 51.6 50.5 3 45.6 43.6 4 41.4 41.5 5 58.4 59.1 6 68.4 62.2 7 65.3 66.0 8 mean a Means of replicate analyses.

Kjeldahl nitrogen manifold, recorder, and vacuum pump. Air was used as the blank in the reference flowcell of the colorimeter. Reagents. Reagents were prepared according to the Technicon Industrial Methods (17, 2 0 ) , except that the concentration of perchloric acid in the digestion mixture was doubled. In preliminary trials, it was difficult to completely digest and obtain apparent complete nitrogen recovery from sodium citrate solutions of cheese. Increasing perchloric acid concentration corrected this problem. The greater recovery of nitrogen contrasts with previous observations in which recovery from various foods was improved by reducing the perchloric acid concentration (19). All reagents used were of reagent grade purity. Standards. Acid-precipitated, freeze-dried, whole casein containing 0.1360 g nitrogen/g was dispersed in deionized water or 0.1M sodium citrate with a Potter-Elvehjem homogenizer. This stock solution containing 2000 fig N/ml was subsequently diluted to the desired concentration of nitrogen in the working standards. Separate stock solutions were prepared for each type of sample analyzed. Subsequently, it was found that the stock solution could be prepared by dissolving casein in 70% H2S04. All working standards contained 48% sulfuric acid (H2S04) but composition was varied to simulate that of samples to be analyzed. Standards for cheese contained 0.05M sodium citrate whereas 0.5 and 1.75% lactose was added to standards for analysis of total nitrogen in milk and whey, respectively. Standards for analysis of noncasein nitrogen in milk contained 1.0%lactose. In addition to 1.0% lactose, 3.0%trichloroacetic acid (TCA) was added for determination of nonprotein nitrogen in milk. The following nitrogen concentrations were used in working standards in this study: 50fig/ml for measurement of noncasein N in cheese, 500 pg/rnl for total N in cheese and milk, 340 pg/ml for total N in whey, 250 pg/rnl for noncasein N in milk and 56 figlml for nonprotein N in milk. Sample Preparation. Samples of cheese, manufactured by a direct acidification method (21 ), were prepared and fractionated by the procedure of Vakaleris and Price (8) for subsequent analysis of total and noncasein nitrogen. The total N sample was a dispersion of 10 g of cheese in 200 ml of 0.1M sodium citrate; the noncasein nitrogen fraction was the filtrate from the cheese dispersion after adjusting the pH to 4.4 f 0.05. Skim milk samples were fractionated by the method of Aschaffmburg and Drewry ( 2 2 ) to determine noncasein nitrogen and nonprotein nitrogen. 506

1.82 0.43 2.31 0.79 1.47 1.25 3.16 1.07 1.54

4.4 0.8 2.1 4.6 0.2 1.2 9.0 1.1 2.9

0.49 0.12 0.06 0.12 0.20 0.17 0.15 0.29 0.20

Analysis of aqueous or citrate solutions of samples was not possible because of precipitation a t the point of injection of digestion mixture into the sample line of the AutoAnalyzer. The precipitation undoubtedly occurred in a zone where the mixture of aqueous sample and acidic digestion solution attained a p H equal to the iso-electric point of the protein. T o avoid precipitation, samples were solubilized in concentrated HzS04 and diluted so that the concentration of HzS04 in the final dilution was approximately 48% (v/v). The following ratios of sample volume to total volume of sample plus acid diluent were used: 1:4 for analysis of total N in cheese, 1:2 for noncasein N in cheese aged 1 week, 1:4 for noncasein N in cheese aged 1 and 3 months, 1:lO for total N in milk, 1:2 for noncasein N in milk, 6:25 for nonprotein N in milk, and 1:4 for total N in whey. AutoAnalyzer Procedure. The basic procedure as given in the Technicon Industrial Methods (17, 20) was used to analyze all samples. The 0-100 fig N/ml range was used for determination of noncasein N in cheese and nonprotein N in milk. The 0-1000 f i g N/ml range was used for analysis of total N in cheese, milk, and whey and for analysis of noncasein N in milk. Triplicate measurements were made of each sample with an appropriate standard, in triplicate, placed after every five samples (15 measurements). Samples were analyzed at the rate of 20 measurements per hour with a wash-sample ratio of 1:9; the wash solution was 48% HzS04 (v/v). The sample flow was doubled over that recommended in the published method for the 0-1000 fig N/ml range. Temperatures ("C) used for stage 1 and stage 2 of the digestor module for various sample types were: 240 and 220 for total and noncasein N of cheese, 280 and 260 for noncasein N of milk, and 300 and 280 for total N of milk, total N of whey, and nonprotein N of milk. Preliminary trials indicated that these temperatures were optima. Method Comparison. The AOAC Kjeldahl N method (23) and automated method were compared in analysis of total N of eight cheese samples in one trial; seven of these same eight cheese samples were compared in a second trial. Different stock casein standard solutions were used for each trial. The two methods were compared also for analysis of noncasein N of eight cheese samples, total N of four whey samples, and total N, noncasein N, and nonprotein N of four milk samples. All samples were randomly selected from a larger group of samples within each of the categories previously listed. Each randomized group of samples contained

ANALYTICAL CHEMISTRY, VOL. 47, NO. 3, MARCH 1975

Table 11. Comparison of Automated a n d Official Methods for Analysis of Total N, Noncasein N, a n d Nonprotein N in Milk and Total N in Whey Dissolved in 48% H2S04 (v/v) Replicate observations N Concenh'ation, "g N / m l a Sample

Official

Automated

Standard deviation, ug N / m l

Difference between methods, %

Official

Coefficient of variability, %

Automated

Official

Automated

3.51 1.41 0.58 2.65 2.04

1.01 0.55 0.58 0.73 0.72

0.65 0.26 0.11 0.51 0.38

1.70 1.52 1.15 0.58 1.24

1.01 1.20 1.32 2.05 1.40

0.69 0.62 0.54 0.32 0.54

0.67 0.14 0.14 0.51 0.37

2.49 4.36 2.59 1.20 2.66

0.94 0.24 0.27 0.87 0.58

1.73 1.53 2.08 2.31 1.91

0.47 0.35 1.41 0.43 0.67

0.54 0.55 0.75 0.80 0.66

Total N in Milk

1 2 3 4

543 547 517 516

542 545 517 515

0.2 0.3 0.0 0.1 0.2

241 240 218 189

245 244 213 184

1.7 1.7 2.3 2.6 2.1

70.3 60.1 50.5 58.4

71.1 59.2 51.2 58.6

1.1 1.5 1.4 0.3 1.1

317 272 275 290

323 276 276 288

1.9 1.4 0.4 0.7 1.1

mean

5.51 2.99 2.99 3.77 3.82 Noncasein ru' in milk

1 2 3 4

mean

2.44 2.87 2.87 3.87 3.01 Nonprotein N in milk

1 2 3 4

mean

1.75 2.62 1.31 0.70 1.60 Total N in Whey

1 2 3 4 a

mean Mean of replicate analyses.

1.50 0.96 3.87 1.26 1.90

samples which may have been made from different lots of milk, milk pasteurized at different temperatures, or cheese of various ages. Quadruplicate analyses of each sample were made with the AOAC method; triplicate analyses were run on the automated system. Statistical comparison of the two analytical methods included calculation of standard deviations and coefficients of variability of replicate observations, numerical and percentage differences of nitrogen concentrations between the two methods, standard deviations of numerical differences, Q-test for equality of variances (241, and two-way analysis of variance with unequal subclass numbers.

For the analysis of variance, treatments (curing time of cheese and heat treatment of milk prior to cheese manufacture) were considered to have a random effect while the effect of the two analytical methods was considered to be fixed.

RESULTS A N D DISCUSSION As is indicated in Tables I and 11, mean percentage differences between nitrogen concentrations determined by the two methods for each type of sample ranged from 0.2 to 2.9%. A mean difference of less than 2% was found for four out of the six types of samples. These values fell within a range of differences observed in a previous comparison of the AOAC Kjeldahl and AutoAnalyzer methods for analysis of nitrogen in beef ( 1 4 ) . Mean numerical differences and standard deviations of numerical differences for each sample type are given in Table 111. The sign was disregarded in calculating percentage differences in Tables I and I1 but was taken into consideration in determining numerical differences in Table 111. Numerical differences indicate very close agreement between the two methods with the exception of the first trial of analysis of total N in cheese. The results of a Q-test for equality of variances indicated that only in the first and second trials of determination of total N of cheese and determination of noncasein N of cheese was there reason t o suspect inhomogeneous data. After reviewing the variances of trials of total N, the variance for sample No. 6 as determined by the AOAC method

.

Table 111. Numerical Differences between Official a n d Automated Methods of Nitrogen Analysis for E a c h Type of Sample Standard deviation Mean numerical Type of sample

Total N of cheese Trial 1 Trial 2 NCN of cheese Total N of milk NCN of milk NPN of milk Total N of whey

of differences,

difference, ug N / m l SD, ug h / m l

-9.1 -1 .o 1.1 1.o 0.5 4.2 2.3

4.76 4.76 2.51 0.82 5.20 0.77 3.50

was found to be much larger than the other values. Errors in sampling may account for the atypically large variance. When data for this sample were deleted, the results of the Q-test indicated no reason to suspect inhomogeneous data in trials 1 and 2 . A statistically significant difference between the two methods was detected by analysis of variance of the first trial in analyses of total N in cheese but not of the second (Table IV).This result was also found by analysis of variance when sample 6 was included in both trials. However, a significant interaction (trial 1, P = 0.01, trial 2 P = 0.025) was detected between methods and treatments for both trials. This indicates that values obtained by the AOAC method were not consistently larger or smaller than those obtained by the automated method over all treatments. The significant interaction will have no practical effect because of the lack of significant difference between methods in trial 2 .

ANALYTICAL CHEMISTRY, VOL. 47, NO. 3, MARCH 1975

507

lower for the automated analysis system for all samples except for total N determinations of one cheese sample and three whey samples. Recorder peaks which were typical for AutoAnalyzer I1 were obtained during experimental trials included in this study. Each peak had a plateau which was very consistent between replicate samples within a given nitrogen concentration range. The plateaus (region in which concentration changed less than 1 pg/ml) were 6-7 mm long for replicate determinations of 500 pg/ml standards and for samples containing 200-600 pg N/ml. Plateaus for replicate determinations of samples and standards containing 30-70 pg N/ml were 9-11 mm in length. Operating conditions had to be varied slightly with different samples. Digestion temperatures had to be increased for samples containing higher concentrations of lactose. High concentrations of lactose in whey also caused problems in preparing samples. Whey dissolved in 48% H2S04 was too viscous to pump through sample lines unless diluted as outlined in the Methods Section. The automated system was quite stable except for occasional drifting of the base line. Slight drifting( 0.10). A new standard was made for the second trial in determining total N concentration in cheese which suggests that the nitrogen content of the standard used in the first trial was higher than calculated. Errors in weighing and dilution and nonhomogeneity of the dried casein should not have contributed greatly to differences in the two standards. Greater error probably resulted from dispersion of casein with the homogenizer and transfer of the casein slurry to a volumetric flask. Dissolving casein directly in 72% or concentrated HzS04, with subsequent dilution to attain 48% H2S04,lessens this error. In using this system, it is recommended that a laboratory make up sufficient stock standard for several months’ usage. This dilution should be analyzed by the AOAC Kjeldahl procedure to accurately determine nitrogen concentration. Working standards of desired composition can be made by diluting a single stock standard with appropriate solutions. Stock solutions appeared to be quite stable during several months of refrigerated storage. In addition to the statistically significant difference between methods in determining cheese total N in one trial, the methods also differed significantly in determination of noncasein N in cheese and total N in whey. The reasons for these differences are not obvious, especially for the whey analysis since the percentage difference (Table I) between nitrogen concentrations was extremely, low. As indicated earlier, inequality of variances was observed for data on noncasein N concentrations in cheese. The inequality was eliminated by transforming the data into square roots. This was necessitated by much larger variances of the AOAC method as compared to the automated procedure. Analysis of variance of the transformed data yielded the same results (significant a t P I0.005) as obtained with the original data. The significant difference in cheese noncasein N trials probably resulted from the substantial differences in three samples ( I , 4, 7 ) . A positive interaction was noted between treatments and methods for the milk noncasein N samples. After examination of the data, it was found that the automated method overestimated noncasein N concentrations when levels in samples exceeded 230 pglml but underestimated concentrations below that level. However, this effect is probably inconsequential since there was no statistically significant difference between the two methods when all levels were considered (Table IV). Differences in the other two sample types were more consistent and, seemingly, quite low and within acceptable limits for analysis of such materials. Much better precision between replicate analyses was obtained with the automated as compared t o AOAC procedure as indicated by standard deviations and coefficients of variability in Tables I and 11. Standard deviations were 508

CONCLUSIONS Methods described for automated analysis of nitrogen in milk and cheese involve preparation of casein standards of similar composition to that of the unknown samples, adjustment of digestion temperatures for each sample type, dilution of samples to approach the nitrogen concentration of working standards, and increased perchloric acid content of the digestion mixture. These modifications can be employed with conditional assurance of obtaining a mean percentage difference of 2.0 for the 50-70 pg N/ml range and 1.0 for the 200-500 p g N/ml range when compared to the official Kjeldahl method. Statistically significant differences were found in this study between the two analytical methods for determination of noncasein N of cheese and total N of whey. Significant interactions between treatments and methods were detected for total N and noncasein N of cheese, noncasein N of milk, and total N of whey. In these cases, the individual researcher must decide if the possibility of error and magnitude of error are of concern for the type of study to be undertaken. Significant interactions are of less concern unless there is a trend of over- or underestimating concentrations above and below a given concentration range.

ANALYTICAL CHEMISTRY, VOL. 47, NO. 3, MARCH 1975

ACKNOWLEDGMENT The authors wish to acknowledge the assistance of J. N. Senturia, Departments of Statistics and Agronomy, in statistical analyses of data.

LITERATURE CITED (1) A. Ferrari, Ann. N. Y. Acad. Sci., 7 6 , 792 (1960). (2)Th. Schrnidhofer, H. R. Egli, and R. Weber, Alirnenfa, 12, 109 (1973)

(3) Th. Schmidhofer, H. R. Egli, E. Hauser, and W. Kunzler. Alimenta. 12, 145 (1973). (4) M. J. Brennan, FoOdEng., 46, 102 (1974). (5) S. J. Toma and S. Nakai, J. FoodSci., 35, 507 (1971). (6) R. S. White, N. R. Ghandi, and G. H. Richardson, J. Dairy Sci., 56, 630 (1973). (Abstr). (7) G. E. Schman, M. A. Stanley, and D. Knudsen, Soil Sci. SOC. Amer. Proc., 37, 480 (1973). (8) D. G. Vakaleris and W. V. Price, J. Dairy Sci., 42, 264 (1959). (9) C. E. Childs and E. 6. Henner, Microchem. J., 15, 590 (1970). (10) R. Fiedler, G. Proksch, and A. Koepf, Anal. Chim. Acta, 63, 435 (1973). (11) W. T. Greenway, CerealChem, 49, 609 (1972). (12) Technicon Corporation, Tarrytown, N.Y., Industrial Method 31-69A (1969). (13) J. G. Brisson, "Determination of Total Nitrogen in Milk Using a Techni-

con Autoanalyzer System," presented at the Technicon Symposium, "Automation in Analytical Chemistry," Sept. 9. 1965. (14) W. M. Gantenbein, J. Ass. Offic. Anal. Chem., 56, 31 (1973). (15) S. C. Jacobs, J. Clin. Pathol., 21, 218(1968). (16) J. L. Cox and B. G. Harmon, Automat. Anal. Chem. Technicon Symp. 1, 149 (1967). (17) Technicon Corporation, Tarrytown, N.Y., industrial Method 103-70A/ Preiiminary (1972).

Kramme, R. H. Griffen, C. G. Hartford, and J. A. Corrado, Anal. Chem., 45, 405 (1973). H. G. Lento and C. E. Daugherty, Food Prod. Develop., 5,86 (1971). Technicon Corporation, Tarrytown, N.Y., Industrial Method 146-71A/ Preliminary (1971). E. L. Quarne, W. A. Larson, and N. F. Olson, J. Dairy Sci., 51, 527 (1968). R. Aschaffenburg and J. Drewry, Proc. X V lnt. Dairy Congr., 3, 1631 (1959). W. Horwitz, Ed., "Official Methods of Analysis of the Association of Official Analytical Chemists." 1 Ith ed.. The Collegiate Press, Menasha. Wis., 1970, p 858. I. W. Burr and L. A. Foster, "A Test for Equality of Variances." Purdue Univ. and Valparaiso Univ., Dept. of Statistics, Div. of Math. Sci., Mimeograph Series No. 282, April 1972. D. G.

(21) (22) (23) (24)

RECEIVEDfor review July 24, 1974. Accepted November 1, 1974. This research was supported by the College of Agricultural and Life Sciences, University of Wisconsin-Madison, and by the Cooperative State Research Service, U.S. Department of Agriculture.

Pyridine Catalyzed Reaction of Volatile N- Nitrosamines with Heptafluorobutyric Anhydride Terry A. Gough, Keith Sugden, and Kenneth S. Webb Laboratory of the Government Chemist, Cornwall House, Stamford Street, London SE 1 9N0, England

A procedure has recently been devised for the direct conversion of trace quantities of volatile N-nitrosamines to fluorinated anhydride derivatives via a pyridine catalyzed reaction with heptafluorobutyric anhydride (HFBA). The objective of the present study is to investigate the nature of these derivatives by use of combined gas-liquid chromatography and mass spectrometry. For this purpose, a series of dialkyl and heterocyclic nltrosamlnes were treated with HFBA In the manner previously described. The derivatives were formed in hexane solution, separated on a 2 % OV-1 gas chromatographlc column at 135 O C and detected by flame ionization. Kovats retention indices are presented. The gas chromatograph was linked to the mass spectrometer via a silicone membrane separator. Low resolution mass spectra were obtained for the derivatives, and parent and other characteristic ions were accurately mass-measured at resolution 18,000. Structures for the derivatives are postulated. The N-nitrosodimethylamine-HFBA reaction is shown to be unique.

The carcinogenic properties of many of the N - nitrosamines are well known, and sensitive methods for the detection of the steam volatile nitrosamines are available ( I ) . Most procedures rely on mass spectral confirmation after preliminary separation and detection by gas chromatography (2). A frequently adopted means to quantitatively determine any trace constituent is to prepare a derivative which can be separated from interferants by gas chromatography and detected by electron capture. This approach has been used indirectly for nitrosamine determination ( 3 ) . Heptafluorobutyric anydride (HFBA) is a well known acylating agent and has been employed to form derivatives with amines (4, 5 1, N - methylcarbamate insecticides (6) and various drugs found in biological samples ( 7 ) . A method (8) has been described whereby volatile N - nitrosamines are converted into electron capturing derivatives by a pyridine catalyzed reaction with HFBA.

The mechanism for the reaction of HFBA and pyridine with nitrosamines has not been established. There is little information in the literature regarding the nature of the reaction products, and only a N - nitrosodimethylamine derivative was studied by mass spectrometry (8). Furthermore the HFBA derivatives of other nitrosamines were identified by gas-liquid chromatography only, and the results were reported to be tentative. I t is the aim of this work to investigate the nature of these other derivatives by combined gas chromatography and mass spectrometry (GC-MS) and to describe a second N - nitrosodimethylamine derivative. EXPERIMENTAL Apparatus. A Pye 104 chromatograph fitted with a flame ionization detector was used. The mass spectrometer was an AEI Model MS902 double focusing instrument fitted with peak matching facilities. The GC-MS interface was a silicone membrane separator (9) housed in the GC oven between the column exit and the detector. The separator was connected to the mass spectrometer via a heated 0.5-mm i.d. stainless steel line. The operating conditions are given in Table I. Reagents. N-Nitrosodimethyl-, diethyl-, dipropyl-, and dibutylamines and N-nitrosopiperidine (NDMA, NDEA, NDPA, NDBA, and NPIP) were purchased from Eastman Chemical Company; N-nitrosopyrrolidine (NPYR) from K and K Laboratories, Inc.; heptafluorobutyric anhydride (HFBA) from Pierce Chemical Company; pyridine (Spectrograde) and chloroform (Analytical Reagent grade) from Fisons, and hexane (Puriss) from Kochlight. Procedure. The 50 J/ml solutions of NDMA, NDEA, NDPA, NPIP, and NPYR in chloroform were prepared, and 60-pl aliquots of these solutions were treated with pyridine and HFBA. The procedure employed was identical to that in the literature (8). To minimize losses due to evaporation, hexane (200 ~ 1 was ) used as the final solvent instead of diethyl ether although it is recognized that the derivatives are more soluble in the latter. Five-pl portions of the hexane solutions were injected onto the chromatograph for analysis.

RESULTS AND DISCUSSION Preliminary gas chromatography demonstrated that the nitrosamine-HFBA derivatives are completely and irreversibly adsorbed by stainless steel. Quantitative passage

ANALYTICAL CHEMISTRY, VOL. 47, NO. 3, MARCH 1975

509