CLIN. CHEM. 35/1, 74-76 (1989)
Preparationof Fecal Samples for Assay of Volatile Fatty Acids by Gas-Liquid Chromatography
and High-PerformanceLiquidChromatography
Huel-Mel Chen1
andCanoeH. Llfschltz
We describe a procedure for preparing fecal samples for
determination of volatile fatty acids (VFAs) by gas-liquid.
chromatography (GLC) and “high-performance” liquid chromatography (HPLC). The simple, one-step procedure involves only ultrafiltrationthrough a membrane with a molecular-mass cutoff of 3000 Da. As revealed by the GLC chromatograms, ultrafiltration appears to be as effective as steam
distillation in sample clean-up. It also enables higher, more
reproducible analytical recoveries of long-chain VFAs. The
VFA content of the filtrate can also be measured by HPLC.
Use of the ion-exclusion mechanism completely resolves
isobutyric acid and butyric acid on a cation-exchange column. The mean (± SD) percentage distribution values of
VFAs (measured by GLC) from five healthy subjects were
56.0 ± 3.5 (acetic acid), 17.0 ± 5.3 (propionic acid), 2.9 ±
1.5 (isobutyric acid), 18.8 ± 5.8 (butyric acid), 2.3 ± 1.2
(isovaleric acid), and 2.9 ± 0.8 (valeric acid).
Aminex HPX-87H (Bio-Rad Laboratories,
Richmond, CA) and ORH-801 (Interaction Chemicals
Inc.,
Mountain View, CA), reportedly
have the capacity to separate VFAs (8,9). The ORH-801 column, packed with sulfonated polystyrene
divinylbenzene
in the hydrogen form, was
produced to separate
various organic acids found in the
fermentation
of dairy products (C2 to C6 in the Krebs cycle)
and in certain metabolic disorders,
namely, acidw-ia and
acidemia (9). We used this column for our study because it
withstands
extreme pH conditions (pH 0 to 14) and separates VFAs according to their respective pKa values, with
dilute acid as the only mobile phase.
We undertook this study (a) to develop a simple procedure
for preparing fecal samples for GLC, (b) to compare the
results with those for samples prepared by steam distillation, and (c) to determine the applicability of the present
sample-preparation
procedure
to HPLC.
columns,
Materials and Methods
Additional Keyphrases: ultrafiltration
isobutynclbutyric acids
referenceinteival
steam-distillationprocedure compared
.
Volatile fattyacids(VFAs), including acetic (C2), propionic (C3), isobutyric (iC4), butyric (C4), isovaleric (iC5), and
valeric (C5) acids, are among the predominant
anions produced in the gastrointestinaltractsof mammals, including
humans
(1).2 They result
from the anaerobic microbial
fermentation
of various dietary carbohydrates
and proteins
(2). The VFA content of feces is associated
with changes in
stoolpH, with changes in the excretion of breath hydrogen
that accompany carbohydrate malabsorption-associateddiarrhea, and subsequently, with metabolic acidosis (3). The
conversion of malabsorbed
carbohydrates
to absorbable
VFA by colonic flora can reduce the osmotic load in the
lumen of the colon. Watkins (4) has suggested that this
conversion
provides a mechanism
of energy salvage.
VFAs in physiological fluids such as urine and serum
have been determined
by gas-liquid
chromatography
(GLC), alone or in combination with mass spectrometry
(5,
6). Steam
distillation,
a time-consuming
procedure,
has
often been used to prepare
samples for analysis by GLC.
Because of the many steps involved, this sample-preparation procedure frequently results in lower analytical
recoveries of VFAs.
Based on different modes of separation, “high-performance” liquid chromatography
(HPLC) also has been applied
to analyses for carboxylic acids (7). Two cation-exchange
USDA/ABS
Children’sNutritionResearch Center,Department
of Pediatrics, Baylor College of Medicine,and Texas Children’s
Hospital, Houston, TX 77030.
‘Address correspondence to thisauthor at: Children’s Nutrition
Research Center, 1100 Bates St., Houston, TX 77030.
2Nonstandard abbreviations: VFA, volatilefatty acid; GLC, gasliquid chromatography(-ic);
C2, acetic acid; C3, propionic acid; iC4,
isobutyric
acid; C4, butyric acid; iC5, isovaleric
acid; and C5,valeric
acid.
Received August 31, 1988; accepted October 29, 1988.
74 CLINICALCHEMISTRY, Vol. 35, No. 1, 1989
Ultrafiltration procedure. Fecal specimens
were collected
from five healthy subjects: one bottle-fed infant (age <6 mo),
one child (age <2 y),and three adults (ages 30,34, and 43y).
The fresh feces were stored frozen in sealed containers
within
1 h of collection.
After the feces were thawed, --3 g was homogenized for 2
miii in 30 mL of 0.15 mmolfL H2S04 in Milli-Q-purified
water (resistance >18 Mfl; Millipore, Bedford, MA) with a
Stomacher Lab Blender 80 (Tekmar Co., Cincinnati, OH). A
5-mL portion of the homogenate was centrifuged at 9000 x
g at 2 #{176}C
for 20 miii. The supernatant
fluid was then ifitered
through
a microconcentrator
(Centricon-3;
Amicon, Danvers, MA) with a molecular-mass cutoff of 3000 Da, by
centrifligation
(7000 x g, 2#{176}C,
1 h). For GLC analysis, the
filtrate was acidified to pH 2.0 with 6 mol/L 112504,2 to 4 pL
being required for 1.5 mL of filtrate. Filtrate to be used for
analysis by HPLC
received no further
treatment.
The
percentage of dry weight in the stool homogenates
was
determined by drying --0.5 g of sample in an oven (>95 #{176}C)
until the weight became constant.
Steam-distillation
procedure for GLC. Samples for GLC
analysis of VFA were prepared as in the method described
by Fleming and Rodriguez (10). Briefly, 5 mL of the fecal
homogenate was acidified to pH 2.0 with -25 1zLof 6 mol/L
H2S04. Steam distillation into 10 mL of 0.2 mollL NaOH
was continued until 500 mL of distillate had been collected.
After freeze-drying,
the sodium salts (nonvolatile) of VFA
residues were redissolved in 5 mL of Milli-Q-purified water
and acidified to pH 2.0 with -150 1zL of 6 moIIL H2S04.
Calibration standards
with known amounts of acetic, propionic, butyric (Aldrich Chemical Co., Milwaukee, WI), isobutyric, isovaleric, and valeric acids (Sigma Chemical Co., St.
Louis, MO) were treated in an identical manner, either
alone or after addition to fecal samples. All samples were
ifitered through a 0.2-tim (pore size) membrane (Nalgene
Co., Rochester, NY) before GLC analysis.
Gas-liquid chromatography.
We used a 1.82 m x 2 mm
(i.d.) glass column packed with 80/120 Carbopack B-DA
(deactivated for acidic compound)/4% Carbowax 20 M (Supelco Inc., Bellefonte, PA), preconditioned by heating at
245 #{176}C
for 19 h in nitrogen (flow rate 10 mL/min). The
column
ends were plugged with phosphoric acid-treated
C4
glass wool. The glass wool and the septum at the injection
U
end were replaced after every 20 injections to remove
w
0)
accumulated impurities and salts and to maintain a stable
z
0
0.
baseline. VFAs were determined with a flame ionization
CO
I0
detector (Model 1400; Varian Aerograph, Walnut Creek,
CA) linked to a reporting integrator (Model 3390 A; Hewlett-Packard,
Avondale, PA). Chromatographic
settings
were as follows: injection port at 200 #{176}C,
detector at 200 #{176}C,
and oven at 175 #{176}C.
The flow rates for nitrogen, hydrogen,
and air were set at 30, 30, and 300 mllmin,
respectively.
Analyses were performed under isothermal conditions for 20
mm.
1’O
20
MINUTES
The linearity of the standard
calibration curve was determined from the average of five 1-iL injections of each
Fig. 1. GLC chromatograms of VFA in fecal samples
standard acid solution (0.1,0.25, 1.0,2.0, and 4.0 minollL for
Top sampleprepared by steam distillation;bottom:sample preparedby ultrafilC2 to C5). External standardization procedures with estabtration.U unidentifiedcomponent
lished response factors from the average of five 1-4 injecthe concentration gradient of the samples. The GLC analytions of standard acid solutions were used to calculate the
ses of the VFA contents of fecal samples prepared by both
concentrations of VFA in the samples.
ultrafiltration
and steam distillation are presented in Table
Separation
by HPLC. The isocratic HPLC apparatus
1.
(Waters
Associates, Milford, MA) consisted of a mobileFigure 2 shows HPLC chromatograms
for the standard
phase reservoir;a Model 510 pump; an automatic sample
mixture (10-tL injection, 1 mmol of VFA per liter) and for
injector,Model 712 (WISP, Waters Intelligent Sample Prothe focal samples. The resolution was reproducible,
and the
cessor); a cooling module; a precolumn with Guard-PAK
ifiter (“Resolve” C18 insert); the ORH-801 organic acid
Table 1. QuantItatIve Analysis of Volatile Fatty Acids in
chromatographic
column, 30 x 0.65cm (i.d.); a temperatureFeces as Measured by GLC
control module and a column heater; a conductivity detector,
Sample preparation
Model 431; a recorder and integrator, Model 840 data and
chromatography
control station; a system interface module;
VFA
Ultraflltration
Steam dlstlllation
% difference
and (from Digital Equipment
Co., Bedford, MA) a Model
Acetic
213.2
205.4
3.8
LA5O printer.
Propionic
69.7
66.1
5.4
The mobile phase, a 0.15 mmol/L solution of “ultrapure”
Isobutyric
10.1
10.2
-1.0
sulfuric acid (VWR Scientific,
Houston,
TX) in Milli-QButyric
65.2
58,4
11.6
lsovaleric
8.4
7.6
purified water (pH 3.1, background conductance <100 MS),
10.5
Valeric
11.1
9.6
15.6
was ifitered through
a 0.2-tm (pore-size) membrane and
8Values are means (n = 5), jmol/g dry weight.
degassed before use. Separation of VFAs (C2 to C5) was best
with a flow rate of 0.5 mL/min and a column temperature
of
IC,
37#{176}C.
A single run required 40 mm. To regenerate the
column,
we increased the column temperature to 65 #{176}C
C2
while washing with 25 mmol!L ultrapure sulfuric acid at a
flow rate of 0.6 mL/min for 6 h. Calibration was based on
mean values of the response factors from five 10-ML injections of the same external standard
solution used for GLC.
Cs
To prepare the standard calibration curve the procedures
we
w
used were those for GLC. Standard calibration data deterz
0
mined by HPLC were subjected to linear-regression analysis
aCo
w
(11).
(I)
Results
Typical gas-liquid
chromatograms
for the fecal samples
prepared by conventional steam distillation and those samples prepared by ifitration with the Centricon-3 microconcentrator are shown in Figure 1. Except for unidentified
peaks eluted before acetic acid, the elution profiles shown in
the two chromatograms are almost identical. The minor
baseline drift during the run did not affect accurate quantification of VFA peaks.
For VFAs in the standard solution and added to fecal
samples, analytical recovery by steam distillation for GLC
analysis ranged from 96% to 106%. The flame ionization
detector response for C2 to C5 acids was linear (r >0.99) over
40
10
MINUTES
Fig.2. HPLC chromatograms of VFA in (top) a standard mixture (VFA
C1 toC5 each at 1 mmol/L),and (bottom) a fecal filtrate
C,: formic acid (off-scale). The shaded areas contain unidentifiedcomponents,
whichvaried in concentrationfromrun to run. Identification of C2toC5,however,
was unaffectedby these components
CLINICALCHEMISTRY,Vol. 35, No. 1, 1989 75
Table 2. Analytical Recovery and Intra- and Interrun
Reproducibility of Volatile Fatty Acids in the
Ultrafiltrate of Fecal Homogenates
Acetic
Propionic
Mean (and SD)
recovery, %
103 (3.8)
95 (4.2)
Isobutyric
96 (3.6)
VFA
Butyric
Isovaleric
Variability’ (CV, %)
intra-run
Inter-run
2.6
1.4
2.4
4.8
2.6
4.5
102 (4.3)
1.3
98 (4.8)
4.1
Valeric
97 (5.1)
‘Based on n = 5 determinationseach.
4.6
51
coefficients of variation (CVs) for retention times after nine
successive injections of standard solutions of all VFA peaks
were <0.5%. In runs involving fecal samples, however, all
VFAs except acetic acid were eluted faster than the respective standards.
The consistent shift in retention times in the
VFA-supplemented
fecal samples enabled us to confirm the
identification of the VFA peaks.
The response of the conductivity detector varied linearly
(r >0.99) with the amounts of VFA injectedover a range of 1
to 40 nmol, with good reproducibility
as demonstrated
by
the low CV (0.2% to 4.8%). Analytical recoveries of VFAs
from the standard solutions and the supplemented
focal
samples
by ultrafiltration
through
microconcentrators
ranged from 95% to 103% (CV <5%) for C2 to C5 acids as
analyzed by HPLC. Intra-run
and inter-run reproducibilities (Table 2) were good.
In the five subjects, the mean (± SD) percentage distribution values of VFAs measured by GLC were 56.0 ± 3.5 for
acetic acid, 17.0 ± 5.3 for propionic acid, 2.9 ± 1.5 for
isobutyric acid, 18.8 ± 5.8 for butyric acid, 2.3 ± 1.2 for
isovaleric acid, and 2.9 ± 0.8 for valeric acid.
Discussion
One-stepultrafiltration
of fecal aqueous extract through a
membrane with a 3000-Da cutoff appeared to be as effective
as multiple-step distillation for sample clean-up. The effectiveness of our procedure is further supported by the similarities in the amounts of residues accumulated in the
injection port for the two types of sample preparations.
The percentage differences between the VFA amounts
from these two sample preparations indicate that ultrafiltration of feces enables greater analytical
recoveries of longchain VFAs. This result is consistent with the finding by
Edwards et al. (12) that the efficiency of distillation decreases as the VFA carbon-chain length increases.
Because of its higher selectivity, we chose a conductivity
detector, rather than an ultraviolet detector, as more appro-
priate for the determination of VFAs. Although there were
some differences in the absolute concentrations
of VFAs
measured by GLC and HPLC, the proportions of VFAs
measured in focal ifitrate by the two methods were similar
(Table 3). These data agree well with the data from a oneweek study of feces from 10 healthy subjects (13) and with
the reported (1) molar ratios of acetic acid:propionic acid:butyric acid (59:22:19) in fecal dialysate for a man on an
unrestricted
diet.
The simple, one-step ultrafiltration of fecal homogenate
described in our study minimizes
the loss of VFAs. Samples
prepared by this method are therefore suitable for routine
analysis by either GLC or HPLC, and this probably is true
for a variety of physiological samples that contain VFAs.
This prqject
was funded in part with funds from the U.S. Dept. of
Agriculture,AgriculturalResearch Service, under Cooperative
Agreement no. 58-7MNI-6-100.The contents of this publication do
not necessarily reflect the views or policies of the U.S. Dept. of
Agriculture, nor does mention of trade names, commercial products,
or organizations imply endorsement by the U.S. Government.
References
1. Cummings JH. Short chain fatty acids in the human colon. Gut
1981;22:763-79.
2. Zarling EJ, Ruchim MA. Protein origin of the volatile fatty acids
isobutyrate and isovalerate in human stool. J Lab Clin Med
1987;109:566-7.
3. Lifschitz CH. Breath hydrogen testing in infants with diarrhea.
In: Lifschitz F, ed. Carbohydrate intolerance
in infancy. New York:
Marcel Dekker, 1982:32-42.
4. Watkins JB. Developmental aspects of carbohydrate malabsorption in the premature infant. Ibid.: 61-74.
5. Chalmers RA, Lawson AM. Organic acids in man. New York:
Chapman and Hall, 1982:128-37.
6. Niwa T. Metabolic proffling with gas chromatography-mass
spectrometry and its application to clinical medicine. J Chromatogr
1986;379:313-45.
7. Schwarzenbach R. High-performance
liquid chromatography of
carboxylic acids. Ibid. 1982;251:339-58.
8. Guerrant GO, Lambert MA, Moss CW. Analysis of short-chain
acids from anaerobic bacteriaby high-performance
liquid chromatography.J Clin Microbiol1982;16:355-60.
9. Benson JR, Woo DJ. Polymericcolumns for liquid chromatography. J Chromathgr Sci 1984;22:386-99.
10. Fleming SE, Rodriguez MA. Influence of dietary fiberon feed
excretion of volatile fatty acids by human adults. J Nutr
1983;113:1613-25.
11. Ryan RF, Joiner BL, Ryan TA. Mimtab. Boston: Duxbury
Press, 1985:218-58.
12. Edwards GB, McManus WR, Bigham ML. Effect of carbon
chain length upon extraction
of volatile fatty acids from rumen
liquor. J Chromatogr 1971;63:397-401.
13. Hoverstad T, Bjorneklett A. Short-chain fatty acids and bowel
functions in man. Scand J Gastroenterol
1984;19:1059-65.
Table 3. DistrIbution1 of Volatile Fatty Acids (as Percent of Total) in Fecal Filtrates from Five Subjects, as Measured
by HPLC and GLC
1
VFA
Acetic
Propionic
Isobutyric
Butyric
Isovaleric
Valeric
2
3
4
5
HPLC
GLC
HPLC
GLC
HPLC
GLC
HPLC
GLC
HPLC
GL.C
54.9
57.0
54.7
59.7
50.2
51.2
58.4
54.5
22.7
20.8
13.2
14.7
21.9
17.1
53.6
18.7
58.4
21.1
7.9
9.0
3.5
3.1
3.8
1.7
4.6
5.4
1.6
1.6
2.6
2.9
27.3
3.8
2.6
27.4
1.8
1.7
13.9
2.7
2.2
13.7
1.6
2.5
20.2
8.2
3.6
20.3
4.5
3.9
13.5
2.7
2.9
13.2
1.8
3.1
19.9
3.9
2.5
19.6
2.0
3.2
‘Percent of total fecal VFA content. Results are the means of n
76 CLINICALCHEMISTRY, Vol. 35, No. 1, 1989
=
5 determinations
each.