APPLIED AN ENVIRONMENTAL MICROBIOLGY, Sept. 1976, p. 451-454
Copyright X 1976 American Society for Microbiology
Vol. 32, No. 3 Printed in U.S.A.
Computer-Assisted Analysis of Adenosine Triphosphate Data' CARL W. ERKENBRECHER,* STERLING J. CRABTREE, JR., AND L. HAROLD STEVENSON The Belle W. Baruch Institute for Marine Biology and Coastal Research and the Department of Biology,* University of South Carolina, Columbia, South Carolina 29208 Received for publication 20 February 1976
A computer program has been written to assist in the analysis of adenosine 5'triphosphate data. The program is designed to calculate a dilution curve and to correct sample and adenosine 5'-triphosphate standard data for background and dilution effects. In addition, basic statistical parameters and estimates of biomass carbon are also calculated for each group of samples and printed in a convenient format. The versatility of the program to analyze data from both aquatic and terrestrial samples is noted as well as its potential use with various types of instrumentation and extraction techniques.
Adenosine 5'-triphosphate (ATP) was suggested as a reliable estimate of total microbial biomass by Holm-Hansen and Booth in 1966 (6). Since that time, ATP measurements have been reported in ecological studies from a variety of environments (1, 4, 5, 9, 11). For the past 3 years this laboratory has been monitoring estuarine creeks and sediments for ATP levels, using firefly lantern extract and an ATP-Photometer (JRB, Inc.). The procedures available for handling ATP data obtained with this technique involved graphic interpretations that were both time consuming and lacking in precision (8). Consequently, a computer program was written to handle the mathematical calculations, output the data in usable form, and calculate some basic statistical parameters on data obtained from either aquatic or terrestrial samples. Samples were extracted with boiling tris(hydroxymethyl)aminomethane buffer (Tris; 0.02 M, pH 7.75) (4) and assayed for ATP using firefly lantern extract (Sigma Chemical Co., FLE 50). For each determination, a 1-ml sample of enzyme preparation was dispensed into each reaction vial and a 6-s background count was recorded. An equal volume of either Tris buffer, standard, or sample was then injected directly into the enzyme-containing vial, mixed for 10 s, loaded into the Photometer, and counted for 1 min. The counts per minute were recorded after a 60-s integration of the emission
obtained when Tris buffer was substituted for sample. Assaying blanks periodically was necessary because of a loss of light emission from the enzyme preparation during assay of a batch of samples. In addition, the simultaneous effect of dilution resulted from the addition of sample or standard to the enzyme preparation. Both of these effects were corrected by the method of least-squares analysis. Crystalline ATP (disodium salt with MgSO4) was used to prepare standards in concentrations that ranged from 10-10 g of ATP per ml to 10-7 g of ATP per ml. The program outlined here was written in FORTRAN IV, used 32 K of memory, and required double precision. All input was provided on punched cards. The program consisted of three sections, the first of which was used to calculate a linear regression equation for the dilution curve. Instruction cards for section 1 allowed input for both the number of blanks to be used and selection of whether an analysis was performed on data from aquatic or terrestrial samples. The second section of the program was used to correct each series of standards for background and dilution effects. Provisions were made to allow for use of a variable number of standard curves as well as a variable number of individual standards. A corrected count for each ATP standard was calculated from the linear regression equation obtained for the dilution curve and printed along with the original input data. The corrected counts for the standcurve. An ATP standard curve and a dilution curve ards were then subjected to least-squares analwere prepared for each batch of enzyme. A ysis and the regression coefficients were blank was recorded as the counts per minute printed. If more than one standard curve was ' Contribution no. 152 of The Belle W. Baruch Library in given, a mean was calculated along with its corresponding regression coefflcients. Marine Science. 451
452
NOTES
APPL. ENVIRON. MICROBIOL.
NORTH INLET ESTUARY PROECT JUNE 13, 1975
DILUTION CURVE
NUMBER OF PAIRS OF BLANK DATA S BKG COUNTS
BLANKS
2229. 2163. 2112. 1713. 439.
843. 751. 716. 540. 185.
THE LEAST SQUARES LINEAR COEFFICIENTS ARE V-INTERCEPT * 16.154755 SLOPE - 0.341292 CORRECTION OF ATP STANDARD CURVE DATA
GROUP 1 (9 ENTRIES) STANDARD
G ATP/NL
O.1lOE-09 O.50E-09 O.lOE-08 O.1lSE-08 O.50E-08 O.1lOE-07 O. 15E-07 O.50E-07
1 1 1 1 1 1 1 1 1
O.1lOE-06
BACKGRD COUNTS
STANDARD COUNTS
CORRECTED COUNTS
2124. 2009. 2017. 1872. 1981. 1941. 1728. 1737. 1695.
1006. 2218. 3730. 4266. 12758. 26844. 41297. 175926. 414768.
265. 1516. 3025. 3611. 12066. 26165. 40691. 175317. 414173.
THE LEAST SQUARES LINEAR COEFFICIENTS FOR GROUP 1 ARE Y-INTERCEPT a -8049.335938 SLOPE * 0.409216E 13
CORRECTION OF SAMPLE DATA
NUMER OF SAMPLE GROUPS TC BE CORRECTED 2 ESTUARINE STATION #4 REPLICATES 4 F
a
25.00
BKGRD COUNTS
SAPLE COUNTS
1429. 1512. 1368. 1406.
26517. 27846. 27619. 27518.
CORRECTED COUNTS
MEAN
MG BIO-C/M**3
0. 100SE-07 0. 1061E-07 0. 1053E-07 0. 1048E-07
0. 2009E-02 O. 2122E-02 0. 2107E-02 0. 2097E-02
526.62 524.15
226871.
0. 1042E-07 0. 2536E-09
0. 2084E-02 0. 5072E-04
520.90 12.68
CORRECTED COUNTS
19136.289 19623.398 MEAN 19379.844
STANDARD DEVIATION
G ATP/M*3
26013. 27314. 27136. 27022.
STADARD DEVIATION 584. COEFFICIENT OF VARIATION * 2.18% SEDIMENT STATION 12 REPLICATES 2 F * 0.885 BKGRD SAMPLE COUNTS COUNTS 2716. 19603. 2760. 20097.
G ATP/ML
344.436
G ATP/ML 0. 1596E-07 0. 1631E-07 0.1614E-07
0.2475E-09
502.27 530.48
G ATP/G OWT
MG BIO-C/G OWT
0. 2705E-06 0. 2764E-06
.00676 .00691 0.6837E-03
0.2735E-06
O.4195E-08
0.1049E-05 COEFFICIENT OF VARIATION a 1.78% END OF REPORT FIG. 1. Portion of a computer printout from the ATP program illustrating the calculation of the dilution curve, the correction ofATP standard curve data, and the correction of aquatic and terrestrial sample data.
VOL. 32, 1976
NOTES
The input for the third section of the program, used to calculate the concentration of ATP for each sample, specified the number of sample groups, the number of replicates in each particular group, and either a filtration volume or a dry-weight measurement. The regression equations for the dilution and standard curves were then used to compute the concentration of ATP for each water or sediment sample, respectively, as grams of ATP per meter. These concentrations of ATP were then converted to biomass (cellular) carbon by multiplying by a factor of 250 (7). In addition, the arithmetic mean, standard deviation, and coefficient of variation (as percent) of the parameters discussed above for each group of samples were calculated and printed. Figure 1 provides a portion of the computer printout illustrating the general format and type of information available from the program. The printout is composed of three sections: (i) the dilution curve, (ii) the ATP standard curve, and (iii) the sample data. The calculation of the dilution curve was a critical step in the manipulation of sample and standard data. Figure 1 illustrates a series of background and blank counts recorded throughout the assay of a single batch of samples, and a reference dilution curve, which was prepared from 329 observations of background and blank counts from several batches of enzyme, is shown in Fig. 2. Background counts in excess of 1,500 cpm at the start of an assay were not considered desirable because of the in-
453
creased scatter. The equation for the regression line was: blank counts per minute = 0.5627 (background counts per minute) - 23.7095. The Pearson correlation coefficient (r) was 0.93, with a P value of 0.0001. The correction of a typical series of ATP standards is illustrated in the second section of Fig. 1, and a reference standard prepared from 179 observations is shown in Fig. 3. The general equation for the regression line is: log,oATP 0.918 (log10 corrected counts per minute) 11.843; the Pearson correlation coefficient (r) was 0.97, with a P value of 0.0001. The linear response of ATP over the range from 10-10 g of ATP per ml to 10-7 g of ATP per ml was in agreement with other studies (2, 3, 10). Caution should be exercised when concentrations of ATP in the sample extract are less than 10-10 gI ml, because these low concentrations approach the lower limit of detection with the ATP-Photometer and the crude enzyme preparation. In summary, this computer program offers a time-saving, simple, and accurate procedure that has all but eliminated the problems associated with the mechanical manipulation of ATP data. Although the program was originally written for the analysis of estuarine samples, it is versatile enough to analyze other aquatic or terrestrial data from a variety of sources. Likewise, this program can be used in conjunction with any type of extraction technique. Copies of this program together with instructions for the input of data and control cards are available from us upon request.
16 *
.
14 .S *s
*
*
12 0
10
i
a.
ad
.*
*
2 C-)
*
8
0
.
*0
~
*
S 6.-
* S
* *
*
-
*0*0
0
~~~~
. @5
6
*o#-~~~~~0
4
*#4w0
.
e3~*
2 "0
2
4
10 16 is 12 24 14 20 22 BACKGROUND CPM 100 illustrating the linear relationship between blank and background counts 6
8
FIG. 2. Reference dilution curve minute of 239 observations from several batches of enzyme preparation.
per
454 106 L
APPL. ENVIRON. MICROBIOL.
NOTES 1111-
I
I
v
1
.-
...
I.I.
II
.
.
-r-,
0
t
;Ij
I
*
E
*I I 8 0
R802928-02-0 from the Environmental Protection Agency, and by The Belle W. Baruch Institute for Marine Biology and Coastal Research.
LITERATURE CITED 1. Ausmus, B. S., and M. Witkamp. 1974. Litter and soil microbial dynamics in a deciduous forest stand. EDFB-IBP-73-10, Oak Ridge National Laboratory, Oak Ridge, Tenn. 2. Chappelle, E. W., and G. V. Levin. 1968. Use of the firefly bioluminescent reaction for rapid detection and counting of bacteria. Biochem. Med. 2:41-52. 3. Cheer, S., J. H. Gentile, and C. *. Hegre. 1974. Improved methods for ATP analysis. Anal. Biochem. 60:102-114. 4. Erkenbrecher, C. W., and L. H. Stevenson. 1975. The influence of tidal flux on microbial biomass in salt marsh creeks. Limnol. Oceanogr. 20:618-625. 5. Holm-Hansen, 0. 1973. Determination of total microbial biomass by measurement of adenosine triphosphate, p. 73-89. In L. H. Stevenson and R. R. Colwell (ed.), Estuarine microbial ecology. University of South Carolina, Columbia. 6. Holm-Hansen, O., and C. R. Booth. 1966. The measurement of adenosine triphosphate in the ocean and its ecological significance. Limnol. Oceanogr. 11:510519. 7. Holm-Hansen, O., and H. W. Paerl. 1972. The applicability of ATP determination for estimation of microbial biomass and metabolic activity. Mem. Ist. Ital. Idrobiol. 29(Suppl.):149-168, Pallanza, Italy. 8. JRB, Inc. 1973. Methods for data analysis using the JRB ATP-Photometer. JRB, Inc., La Jolla, Calif. 9. Karl, D. M., and P. A. LaRock. 1975. Adenosine triphosphate measurements in soil and marine sediments. J. Fish. Res. Board Can. 32:599-607. 10. Levin, G. V., E. Usdin, and A. R. Slonim. 1968. Rapid detection of microorganisms in aerospace water systems. Aerosp. Med. 38:14-16. 11. Quammen, M. L., P. A. LaRock, and J. A. Calder. 1973. Some environmental effects of pulp mill wastes, p. 329-343. In L. H. Stevenson and R. R. Colwell (ed.), Estuarine microbial ecology. University of South Carolina, Columbia.
*~~~~~~~~~~~~ !;-~~~~~~5 S..
102: .-
K}l __u
I.I.
1 7 10 8 g ATP m-I'
lo-9
10-0
FIG. 3. Reference ATP standard curve illustrating typical values for several series of standards and batches of enzyme preparation. The linear relationship between counts and ATP was significant at the 0.01 level and had an r value of 0.97. We wish to acknowledge Henry Wills for writing the preliminary program and Bert Ely for critically reviewing the manuscript. This research was supported in part by grant NG-33-72 from the U.S. Department of Commerce, by grant no.
Copyright X 1976 American Society for Microbiology
Vol. 32, No. 3 Printed in U.S.A.
Computer-Assisted Analysis of Adenosine Triphosphate Data' CARL W. ERKENBRECHER,* STERLING J. CRABTREE, JR., AND L. HAROLD STEVENSON The Belle W. Baruch Institute for Marine Biology and Coastal Research and the Department of Biology,* University of South Carolina, Columbia, South Carolina 29208 Received for publication 20 February 1976
A computer program has been written to assist in the analysis of adenosine 5'triphosphate data. The program is designed to calculate a dilution curve and to correct sample and adenosine 5'-triphosphate standard data for background and dilution effects. In addition, basic statistical parameters and estimates of biomass carbon are also calculated for each group of samples and printed in a convenient format. The versatility of the program to analyze data from both aquatic and terrestrial samples is noted as well as its potential use with various types of instrumentation and extraction techniques.
Adenosine 5'-triphosphate (ATP) was suggested as a reliable estimate of total microbial biomass by Holm-Hansen and Booth in 1966 (6). Since that time, ATP measurements have been reported in ecological studies from a variety of environments (1, 4, 5, 9, 11). For the past 3 years this laboratory has been monitoring estuarine creeks and sediments for ATP levels, using firefly lantern extract and an ATP-Photometer (JRB, Inc.). The procedures available for handling ATP data obtained with this technique involved graphic interpretations that were both time consuming and lacking in precision (8). Consequently, a computer program was written to handle the mathematical calculations, output the data in usable form, and calculate some basic statistical parameters on data obtained from either aquatic or terrestrial samples. Samples were extracted with boiling tris(hydroxymethyl)aminomethane buffer (Tris; 0.02 M, pH 7.75) (4) and assayed for ATP using firefly lantern extract (Sigma Chemical Co., FLE 50). For each determination, a 1-ml sample of enzyme preparation was dispensed into each reaction vial and a 6-s background count was recorded. An equal volume of either Tris buffer, standard, or sample was then injected directly into the enzyme-containing vial, mixed for 10 s, loaded into the Photometer, and counted for 1 min. The counts per minute were recorded after a 60-s integration of the emission
obtained when Tris buffer was substituted for sample. Assaying blanks periodically was necessary because of a loss of light emission from the enzyme preparation during assay of a batch of samples. In addition, the simultaneous effect of dilution resulted from the addition of sample or standard to the enzyme preparation. Both of these effects were corrected by the method of least-squares analysis. Crystalline ATP (disodium salt with MgSO4) was used to prepare standards in concentrations that ranged from 10-10 g of ATP per ml to 10-7 g of ATP per ml. The program outlined here was written in FORTRAN IV, used 32 K of memory, and required double precision. All input was provided on punched cards. The program consisted of three sections, the first of which was used to calculate a linear regression equation for the dilution curve. Instruction cards for section 1 allowed input for both the number of blanks to be used and selection of whether an analysis was performed on data from aquatic or terrestrial samples. The second section of the program was used to correct each series of standards for background and dilution effects. Provisions were made to allow for use of a variable number of standard curves as well as a variable number of individual standards. A corrected count for each ATP standard was calculated from the linear regression equation obtained for the dilution curve and printed along with the original input data. The corrected counts for the standcurve. An ATP standard curve and a dilution curve ards were then subjected to least-squares analwere prepared for each batch of enzyme. A ysis and the regression coefficients were blank was recorded as the counts per minute printed. If more than one standard curve was ' Contribution no. 152 of The Belle W. Baruch Library in given, a mean was calculated along with its corresponding regression coefflcients. Marine Science. 451
452
NOTES
APPL. ENVIRON. MICROBIOL.
NORTH INLET ESTUARY PROECT JUNE 13, 1975
DILUTION CURVE
NUMBER OF PAIRS OF BLANK DATA S BKG COUNTS
BLANKS
2229. 2163. 2112. 1713. 439.
843. 751. 716. 540. 185.
THE LEAST SQUARES LINEAR COEFFICIENTS ARE V-INTERCEPT * 16.154755 SLOPE - 0.341292 CORRECTION OF ATP STANDARD CURVE DATA
GROUP 1 (9 ENTRIES) STANDARD
G ATP/NL
O.1lOE-09 O.50E-09 O.lOE-08 O.1lSE-08 O.50E-08 O.1lOE-07 O. 15E-07 O.50E-07
1 1 1 1 1 1 1 1 1
O.1lOE-06
BACKGRD COUNTS
STANDARD COUNTS
CORRECTED COUNTS
2124. 2009. 2017. 1872. 1981. 1941. 1728. 1737. 1695.
1006. 2218. 3730. 4266. 12758. 26844. 41297. 175926. 414768.
265. 1516. 3025. 3611. 12066. 26165. 40691. 175317. 414173.
THE LEAST SQUARES LINEAR COEFFICIENTS FOR GROUP 1 ARE Y-INTERCEPT a -8049.335938 SLOPE * 0.409216E 13
CORRECTION OF SAMPLE DATA
NUMER OF SAMPLE GROUPS TC BE CORRECTED 2 ESTUARINE STATION #4 REPLICATES 4 F
a
25.00
BKGRD COUNTS
SAPLE COUNTS
1429. 1512. 1368. 1406.
26517. 27846. 27619. 27518.
CORRECTED COUNTS
MEAN
MG BIO-C/M**3
0. 100SE-07 0. 1061E-07 0. 1053E-07 0. 1048E-07
0. 2009E-02 O. 2122E-02 0. 2107E-02 0. 2097E-02
526.62 524.15
226871.
0. 1042E-07 0. 2536E-09
0. 2084E-02 0. 5072E-04
520.90 12.68
CORRECTED COUNTS
19136.289 19623.398 MEAN 19379.844
STANDARD DEVIATION
G ATP/M*3
26013. 27314. 27136. 27022.
STADARD DEVIATION 584. COEFFICIENT OF VARIATION * 2.18% SEDIMENT STATION 12 REPLICATES 2 F * 0.885 BKGRD SAMPLE COUNTS COUNTS 2716. 19603. 2760. 20097.
G ATP/ML
344.436
G ATP/ML 0. 1596E-07 0. 1631E-07 0.1614E-07
0.2475E-09
502.27 530.48
G ATP/G OWT
MG BIO-C/G OWT
0. 2705E-06 0. 2764E-06
.00676 .00691 0.6837E-03
0.2735E-06
O.4195E-08
0.1049E-05 COEFFICIENT OF VARIATION a 1.78% END OF REPORT FIG. 1. Portion of a computer printout from the ATP program illustrating the calculation of the dilution curve, the correction ofATP standard curve data, and the correction of aquatic and terrestrial sample data.
VOL. 32, 1976
NOTES
The input for the third section of the program, used to calculate the concentration of ATP for each sample, specified the number of sample groups, the number of replicates in each particular group, and either a filtration volume or a dry-weight measurement. The regression equations for the dilution and standard curves were then used to compute the concentration of ATP for each water or sediment sample, respectively, as grams of ATP per meter. These concentrations of ATP were then converted to biomass (cellular) carbon by multiplying by a factor of 250 (7). In addition, the arithmetic mean, standard deviation, and coefficient of variation (as percent) of the parameters discussed above for each group of samples were calculated and printed. Figure 1 provides a portion of the computer printout illustrating the general format and type of information available from the program. The printout is composed of three sections: (i) the dilution curve, (ii) the ATP standard curve, and (iii) the sample data. The calculation of the dilution curve was a critical step in the manipulation of sample and standard data. Figure 1 illustrates a series of background and blank counts recorded throughout the assay of a single batch of samples, and a reference dilution curve, which was prepared from 329 observations of background and blank counts from several batches of enzyme, is shown in Fig. 2. Background counts in excess of 1,500 cpm at the start of an assay were not considered desirable because of the in-
453
creased scatter. The equation for the regression line was: blank counts per minute = 0.5627 (background counts per minute) - 23.7095. The Pearson correlation coefficient (r) was 0.93, with a P value of 0.0001. The correction of a typical series of ATP standards is illustrated in the second section of Fig. 1, and a reference standard prepared from 179 observations is shown in Fig. 3. The general equation for the regression line is: log,oATP 0.918 (log10 corrected counts per minute) 11.843; the Pearson correlation coefficient (r) was 0.97, with a P value of 0.0001. The linear response of ATP over the range from 10-10 g of ATP per ml to 10-7 g of ATP per ml was in agreement with other studies (2, 3, 10). Caution should be exercised when concentrations of ATP in the sample extract are less than 10-10 gI ml, because these low concentrations approach the lower limit of detection with the ATP-Photometer and the crude enzyme preparation. In summary, this computer program offers a time-saving, simple, and accurate procedure that has all but eliminated the problems associated with the mechanical manipulation of ATP data. Although the program was originally written for the analysis of estuarine samples, it is versatile enough to analyze other aquatic or terrestrial data from a variety of sources. Likewise, this program can be used in conjunction with any type of extraction technique. Copies of this program together with instructions for the input of data and control cards are available from us upon request.
16 *
.
14 .S *s
*
*
12 0
10
i
a.
ad
.*
*
2 C-)
*
8
0
.
*0
~
*
S 6.-
* S
* *
*
-
*0*0
0
~~~~
. @5
6
*o#-~~~~~0
4
*#4w0
.
e3~*
2 "0
2
4
10 16 is 12 24 14 20 22 BACKGROUND CPM 100 illustrating the linear relationship between blank and background counts 6
8
FIG. 2. Reference dilution curve minute of 239 observations from several batches of enzyme preparation.
per
454 106 L
APPL. ENVIRON. MICROBIOL.
NOTES 1111-
I
I
v
1
.-
...
I.I.
II
.
.
-r-,
0
t
;Ij
I
*
E
*I I 8 0
R802928-02-0 from the Environmental Protection Agency, and by The Belle W. Baruch Institute for Marine Biology and Coastal Research.
LITERATURE CITED 1. Ausmus, B. S., and M. Witkamp. 1974. Litter and soil microbial dynamics in a deciduous forest stand. EDFB-IBP-73-10, Oak Ridge National Laboratory, Oak Ridge, Tenn. 2. Chappelle, E. W., and G. V. Levin. 1968. Use of the firefly bioluminescent reaction for rapid detection and counting of bacteria. Biochem. Med. 2:41-52. 3. Cheer, S., J. H. Gentile, and C. *. Hegre. 1974. Improved methods for ATP analysis. Anal. Biochem. 60:102-114. 4. Erkenbrecher, C. W., and L. H. Stevenson. 1975. The influence of tidal flux on microbial biomass in salt marsh creeks. Limnol. Oceanogr. 20:618-625. 5. Holm-Hansen, 0. 1973. Determination of total microbial biomass by measurement of adenosine triphosphate, p. 73-89. In L. H. Stevenson and R. R. Colwell (ed.), Estuarine microbial ecology. University of South Carolina, Columbia. 6. Holm-Hansen, O., and C. R. Booth. 1966. The measurement of adenosine triphosphate in the ocean and its ecological significance. Limnol. Oceanogr. 11:510519. 7. Holm-Hansen, O., and H. W. Paerl. 1972. The applicability of ATP determination for estimation of microbial biomass and metabolic activity. Mem. Ist. Ital. Idrobiol. 29(Suppl.):149-168, Pallanza, Italy. 8. JRB, Inc. 1973. Methods for data analysis using the JRB ATP-Photometer. JRB, Inc., La Jolla, Calif. 9. Karl, D. M., and P. A. LaRock. 1975. Adenosine triphosphate measurements in soil and marine sediments. J. Fish. Res. Board Can. 32:599-607. 10. Levin, G. V., E. Usdin, and A. R. Slonim. 1968. Rapid detection of microorganisms in aerospace water systems. Aerosp. Med. 38:14-16. 11. Quammen, M. L., P. A. LaRock, and J. A. Calder. 1973. Some environmental effects of pulp mill wastes, p. 329-343. In L. H. Stevenson and R. R. Colwell (ed.), Estuarine microbial ecology. University of South Carolina, Columbia.
*~~~~~~~~~~~~ !;-~~~~~~5 S..
102: .-
K}l __u
I.I.
1 7 10 8 g ATP m-I'
lo-9
10-0
FIG. 3. Reference ATP standard curve illustrating typical values for several series of standards and batches of enzyme preparation. The linear relationship between counts and ATP was significant at the 0.01 level and had an r value of 0.97. We wish to acknowledge Henry Wills for writing the preliminary program and Bert Ely for critically reviewing the manuscript. This research was supported in part by grant NG-33-72 from the U.S. Department of Commerce, by grant no.
