Biotechnol. Prog. 1990, 6, 12 1- 128
121
Chemical Decomposition of Glutamine in Cell Culture Media: Effect of Media Type, pH, and Serum Concentration Sadettin S. Ozturk and Bernhard 0. Palsson* Department of Chemical Engineering, University of Michigan, Ann Arbor, Michigan 48109
T h e chemical decomposition of glutamine to ammonia and pyrrolidonecarboxylic acid was studied a t 37 " C in a p H range of 6.8-7.8 in different media preparations containing various amounts of fetal bovine serum. The media type influenced the decomposition rate, and the first-order rate constants increased with increasing p H values. T h e serum concentration had little or no effect on the decomposition rate. T h e importance of chemical decomposition of glutamine on the analysis of glutamine and ammonia metabolism was illustrated by a n example of batch cultivation of a hybridoma cell line. T h e difference between the apparent uptake rate of glutamine and the actual uptake rate (which is corrected for the chemical decomposition) is shown t o be as high as 200%. Similar discrepancy between the apparent and actual ammonia production rate is observed. Mathematical analysis was carried out t o develop the relationship between the apparent and actual glutamine uptake and ammonia production rates. T h e analysis reveals that there are three important dimensionless parameter ratios t h a t govern the difference between the apparent and actual glutamine uptake and ammonia production rates.
Introduction Cell culture media typically contain glutamine since it is a major energy and nitrogen source for mammalian cells in culture (e.g., Thilly, 1986). Under typical culture conditions that are used for hybridoma growth, glutamine becomes the limiting nutrient, thereby determining the extent of growth and monoclonal antibody production (Glacken et al., 1986; Miller et al., 1989; Ozturk and Palsson, 1990). The kinetics of glutamine uptake and the influence of environmental factors on glutamine metabolism are important in understanding the growth and energetics of mammalian cells in culture. However, unlike other amino acids, glutamine is not chemically stable in cell culture media, complicating the analysis of experimental data. The disappearance of glutamine in the cell culture media is not only due to the uptake by the cells but is also due to its chemical decomposition. The chemical, or nonenzymatic, decomposition of glutamine produces ammonia and pyrrolidonecarboxylic acid. In some instances, the serum used in the media contains glutaminase, and glutamine can then be degraded enzymatically to ammonia and glutamate (Wein and Goetz, 1973). Thus, both chemical and enzymatic degradation should be taken into account in the evaluation of specific glutamine uptake rates. The chemical decomposition of glutamine has been studied in the past for phosphate buffer saline (PBS) by Tirsch and Moore (1962) and for some media formulations: Eagle's medium (Wein and Goetz, 1973), DMEM (Seaver et al., 1984; Glacken et al., 1986; Miller et al., 1989, Lin and Agrawal, 1988). It was observed that the rate of glutamine disappearance is first order with respect to glutamine: d [ Gln] ---k[Gln] dt
WnI or -[Glnl,
8756-7938/90/3006-0121$02.50/0
- exp(-kt)
(1)
where [Gln] is the glutamine concentration, t is time, k is the first-order rate constant, and the subscript 0 denotes initial concentration. It has been shown that the rate constant, K , is influenced by media pH and ionic strength (Bray et al., 1949),temperature (Tirsch and Moore, 1962), and the composition of the media. Phosphate and arsenate ions have been reported to accelerate the decomposition rate (Gilbert et al., 1949). Since the different media preparations have different composition and different phosphate concentrations, differences in glutamine decomposition rate are expected. However, even for the same media, contradictory results have been reported. The half-life of glutamine in DMEM due to the chemical decomposition was found to be 6.7 days (Seaver et al., 1984) and 13 days (Lin and Agrawal, 1988). Further, Lin and Agrawal(1988) reported a strong dependency of the first-order rate constant on serum concentration while Seaver et al. (1984) reported only a slight difference. Here, we studied the decomposition of glutamine in four commonly used cell culture media: IMDM, RPMI1640, DMEM, and OPTI-MEM. For each media type, the influence of serium and pH was experimentally determined. For typical batch growth of a hybridoma cell line in IMDM, the apparent and actual uptake rates of glutamine were determined experimentally. We then present a mathematical analysis that accounts for the disappearance of glutamine via both chemical decomposition and cellular uptake. We use the analysis to obtain a relationship between the true glutamine uptake and the apparent one that is determined without consideration of chemical decomposition.
Materials and Methods Glutamine Decomposition. Cell culture media IMDM (Gibco Laboratories, Grand Island, NY), RPMI-1640 (Sigma Chemical, St. Louis, MO), DMEM (Sigma), and OPTI-MEM (Gibco) were obtained from commercial
0 1990 American Chemical Society and American Institute of Chemical Engineers
122
sources and prepared according to the instructions provided by the manufacturer. Fetal bovine serum (FBS; Gibco) was used to prepare media with different serum concentrations. Four different serum concentrations were used: 0 % , 1.25%, 590, and 10%. The pH was adjusted with sodium bicarbonate (Sigma) and by addition of 1 N HC1 or 1 N NaOH. The media was supplemented by 100 units/mL potassium penicillin G and 100 pg/mL streptomycin sulfate (Sigma). T h e solutions were transferred into 50-mL T-flasks with 10-mL working volume and placed in a humidified incubator with 5% CO, at 37 "C. Samples (0.5 mL) were taken every 2-3 days and kept frozen for later analysis. The pH values were measured for each sample with a pH meter (Beckman). The pH values were different in the incubator than the initial values adjusted in room temperature without CO, but were constant to k0.02 pH units after the first sample. The equilibration of samples to the incubator environment was fast as compared to the time constant of glutamine decomposition. Hence, the samples were assumed to be at the pH values measured after the equilibration, and these pH values were used in reporting the results. Glutamine was measured by a gas-sensing electrode after enzymatic reaction by glutamase (Sigma), using a method described by Ozturk et al. (1989). The results were tested against an HPLC method (Ozturk et al., 1989), and a good agreement was found. The samples were also analyzed for glutamate to check glutaminase activity in the serum-containing solutions with use of HPLC. The OPA (o-phthaldialdehyde) precolumn derivatization method was employed for this purpose (Hill et al., 1982). The mobile phase was a mixture of solvent A [90:9.5:0.5% 0.1 M sodium acetate (pH 7.2)/methanol/THF] and solvent B (100% methanol). The samples were diluted by 20-fold in solvent A, and 100 pL of this mixture was reacted with 200 pL of OPA for 2 min. A 25-pL fraction of the derivatized sample was injected onto a Microsorb Model C18 column (Rainin Instrument Co., Inc., Emeryville, CA). The glutamate concentration was constant in all the samples, indicating the absence of glutaminase activity. We also analyzed the samples for glucose to assess the losses due to evaporation. Glucose concentrations were found to be constant, indicating the absence of evaporation losses. Glutamine Metabolism. We cultivated hybridoma cell line, 167.4G5.3, in batch mode at 37 "C. The antibody produced by this cell line is an IgG,, directed against phosphorylcholine (Briles et al., 1984). A Celligen bioreactor (New Brunswick, NJ) was used with a working volume of 1.5 L. The pH and dissolved oxygen concentrations were kept constant at 7.2 and 50% air saturation, respectively. These parameters could be controlled by adjusting the headphase gas composition via a microprocessor. The cells were cultivated in IMDM containing 10% FBS supplemented with 100 units/mL potassium penicillin G and 100 pg/mL streptomycin sulfate (Sigma). The cells were kept at 37 "C and 5% CO, atmosphere in humidified incubators. The cultures were maintained by passing them every 2-3 days with a dilution factor 1-4 with fresh media. Cells thus maintained in the exponential growth phase were centrifuged at 1000 rpm (200 g) for 10 min. They were inoculated at an initial cell density of 4 X lo4 cells/mL into the reactor. A l-mL sample was taken twice a day. After cell counts were performed, the samples were centrifuged and the supernatants were stored a t -80 "C for subsequent determination of metabolite concentrations. Viable and dead cells in suspension were counted with a hemacytometer.
Biotechnol. Prog., 1990, Vol. 6, No. 2
Trypan blue exclusion method was used to differentiate dead cells from viable cells. Ammonia and glutamine were measured with a gas-sensing electrode (Ozturk et al., 1989).
Results Glutamine Decomposition. The time profiles of glutamine are presented in Figure 1for the four media types used in this study a t different pH values. In each case, no serum was used. The rate constants, 12, were evaluated by fitting eq 1 to the data. The solid lines are the calculated curves based on exponential decay. The decomposition follows first-order kinetics in these media formulations a t all the pH's, and the decomposition rate is strongly influenced by the pH. The serum concentration was found not to influence the glutamine decomposition rate significantly as seen in Figure 2. We tested for glutaminase activity in the serum by checking the increase in glutamate concentration and found no activity. Hence, for serum without glutaminase activity, we conclude that there is no effect of serum components. The previous data on glutamine decomposition in DMEM reported by Lin and Agrawal (1988) indicated an increase in the decomposition rate with increasing serum concentration. For the same media (DMEM), we found no influence of serum, which is consistent with the results of Seaver et al. (1984). Hence, the increase reported by Lin and Agrawal could possibly be due to the glutaminase activity in the lot of serum that they used. Unfortunately, these authors did not check the glutaminase activity in their serum. The glutamine decomposition rate constants were found to be a function of pH and the media type. Figure 3 presents the summary of the rate constants. The decomposition rate constants for IMDM, OPTI-MEM, and DMEM are similar while the rate constant is higher in RPMI-1640 media as compared to the other three. For comparison, also included in Figure 3 are the values obtained in PBS and the values in DMEM reported by other investigators. The data of Trisch and Moore (1962) for PBS was very close to the measured values for RPMI1640. The rate constant that we obtained for DMEM at pH 7.8 was in good agreement with that of Seaver et al. (1984). The data by Lin and Agrawal(l988) for DMEM was consistantly lower than all other reported values. The data by Glacken et al. (1986) and Miller et al. (1989) could not be compared as the pH values were not reported. These investigators both used DMEM, but the rate constants reported were very different [ k = 0.0048 h-' by Glacken et al. (1986) and k = 0.000 23 h-' by Miller et al. (1989)l. The differences in the values of the rate constants could be due to the differences in pH. However, as can be seen from the slopes in Figure 3, this l-order of magnitude difference in k values needs a pH difference of more than 1.5. This large pH variation in media preparation is not reasonable, so is the attribution to the pH differences. The quality of water used in media preparation may influence the decomposition rate constant. As it was mentioned above, phosphate and arsenate catalyze the decomposition. Similarly, some trace ions in water may influence the rate of decomposition. Although deionized water is used in most cases, the ion composition at very low concentration may vary from laboratory to laboratory. More experiments are needed to verify this hypothesis. The pH dependency of the gluamine decomposition rate constant k in the four different media can be repre-
Biotechnol. Prog., 1990,Vol. 6, No. 2
01
0
.
123
'
.
.
'
'
'
.
3.0
2.0
1.0
400
0
zoo
10.
Time, hr
Time, hr
I
0' 0
40.
3.0
zoo
1.0
3.0
4.0
0
100
T h e , hr
3 00
2.0
400
Time, hr
Figure 1. Chemical decomposition of glutamine in (A) IMDM, (B) RPMI-1640, (C) DMEM, and (D) OPTI-MEM media at the pH values indicated on the graphs. No serum was used. The solid lines present the calculated decomposition rates from eq 1 with the fitted first-order rate constant. k . 10
10
10
6.6
6.8
7.0
7.2
7.4
7.6
7.8
8.0
PH
20' 0
1
1
100
200
300
400
Time, hr Figure 2. Influence of serum on the decomposition rate for IMDM and RPMI-1640. The same behavior was observed for the other pH values and for other media formulations.
sented by the following mathematical equation: Ink = a + b p H (2) This equation was fitted to our data, and the numerical values for the constants a and b are presented in Table I. All curve fits had regression coefficients higher than 95%. It is known that phosphate increases the decomposition rate of glutamine (Gilbert et al., 1949). Table I1 summarizes the phosphate concentrations in different media used and the k values a t pH 7.4 for comparison. The values tabulated suggest that there is a correlation
Figure 3. Influence on the decomposition rate constant k by pH for the four different media formulations: RPMI-1640 (open circles), IMDM (closed circles), OPTI-MEM (open squares), and DMEM (closed squares). The arrows point to the values obtained by earlier investigators: (1)value of Seaver et al. (1984) in DMEM and (2) value of Trisch and Moore (1962) in PBS. The solid triangles are the data of Lin and Agrawal (1988).
between the phosphate content of the media and the decomposition rate constant. RPMI-1640 has higher phosphate concentration than the other three media and has the highest decomposition rate constant. Although they have the same phosphate concentration, the difference in rate constants for IMDM, OPTI-MEM, and DMEM was statistically significant. These media differ in their osmolarity, and our values for the rate constant decrease with increasing osmolarity. A possible explanation for this behavior would be the increase in ionic strength with the increase in osmolarity. Thus, the increased ionic
124
Biotechnol. Prog., 1990,Vol. 6, No. 2
Table I. Summary of the Decomposition Rates in Different Media Preparations' media a b IMDM -18.31 (f1.21) 1.685 (f0.095) OPTI-MEM -16.76 (fl.11) 1.458 (f0.065) DMEM -17.07 (21.12) 1.478 (f0.073) RPMI-1640 -13.85 (11.19) 1.133 (f0.055) "The rate constant, k, is in h-'; the values given are for In k = a + b pH. Table 11. Phosphate Content and the Decomposition Rate Constant, k,in Different Media Preparations a t pH 7.4 phosphate osmolarity, media (Pod3-),mM mOsm rate constant, h-' IMDM 1.01 276 0.002 91 (fO.000 13) OPTI-MEM 1.01 290 0.002 55 (fO.000 18) DMEM 1.01 331 0.002 17 (fO.000 16) RPMI-1640 7.98 288 0.004 22 (fO.000 25)
strength could cause a decrease in rate constant (e.g., Bray et al., 1949). Glutamine Metabolism. The growth of hybridoma cell line 167.4G5.3 in a 1.5-L Celligen bioreactor is shown in Figure 4. The cells doubled every 17 f 1 h until a final cell density of 1.2 X lo6 cells/mL. Antibody concentration continued to rise into the decline phase of growth. Glutamine was the limiting nutrient, and by day 5, it was depleted. Cell growth ceased a t this point. Ammonia concentration increased following glutamine depletion, and accumulation of ammonia stopped after the depletion of glutamine. Only half of the glucose was consumed by the cells, and the consumption of the glucose stopped after the glutamine depletion (data not shown).
Data Analysis The kinetics of growth and glutamine metabolism in batch culture are described by simple differential equations. Cell growth is described by
dX,= PX,
(3) dt where X, is the viable cell concentration, t is time, and p is the apparent growth rate (the actual growth rate minus the death rate). The glutamine consumption and ammonia production are described by (4) and
if both the first-order chemical decomposition of glutamine to ammonia and metabolic activity are considered. Here, [Gln] and [NH,+] are the glutamine and ammonia concentrations, and qGh and pNH,+ are the actual uptake rate of glutamine and the actual production rate of ammonia. If the chemical decomposition of glutamine is ignored, then these equations read
and (7)
'w)
. 20
. IO 0
100
I50
?OU
Time, hr 41
'
3
'
"
'
E
"0
50
100
150
10
Time, hr
F i g u r e 4. Growth and glutamine metabolism of hybridoma cell line 167.4G5.3 in 1.5-L Celligen bioreactor. The pH and dissolved oxygen were kept constant a t 7.2 and 50% air saturation, respectively. Key: (A) viable cell and monoclonal antibody concentration; (B) glutamine (open circle) and ammonia (closed circle) concentrations.
where the prime denotes an apparent rate. These parameters may be evaluated from the data by various methods. Differential Method. The differential method uses a discrete approximation to the time derivatives between the data points obtained. Then, the parameters in the equations are evaluated as functions of time (transient kinetics). Here, we evaluate the time derivatives a t each sampling time and calculate the rates at each point from these derivatives and the concentrations using the above equations. The time profiles so obtained for cell growth, glutamine uptake, and ammonia production rates are presented in Figure 5. The values for the cell growth, actual glutamine uptake, and ammonia production rates remained constant for 3 days and then decreased monotonically. We define this period over which these values are constant as the exponential phase. The values obtained for the actual rates are p = 0.040 f 0.002 h-l, qGln = 0.042 f 0.003pmol/(106cells/h), andpNH4+Of0.022f 0.002pmOl/ (lo6cells/h). The yield coefficient of ammonia from glutamine is then 0.52 mol of ammonia produced/mol of glutamine consumed. The "apparent" glutamine uptake and ammonia production rates can also evaluated in this way. The apparent rates are indicated by the dashed lines in Figure 5. These apparent rates are higher than the actual or corrected rates. The error in apparent uptake as compared to the corrected rate is given in Figure 5D. The error is initially as high as 200% for glutamine uptake and 300% for ammonia production and decreases later as the cell concentration, and hence the uptake by the cells, increases. Integral Method. Here, one assumes that the parameters are invariant in time. If the values are not truly
125
Biotechnol. Prog., 1990, Vol. 6,No. 2 0 . 0 5 .
r(
i
P
i
Time, hr
Time, hr
am
mb*,
Lo(
-
'-7,
J
Actual
Apparent
?'.
b. 0.M
0
SO
100
IS0
2.0
Time, hr Time, hr Figure 5. Time profiles of (A) growth rate, (B) glutamine uptake rate, and (C) ammonia production rate. (D)The error introduced by ignoring the decomposition rate. The dashed line presents the results from ignoring the glutamine decomposition whereas the solid lines are generated by including the chemical decomposition. constant, then we obtain time-averaged values. Integrating eqs 2-4, one obtains (Glacken et al., 1988) X, = Xovert
(8)
@Gin [Gln] = [GlnIo- -(Xv CC
and [NH;]
where X",, [Gln],, and [NH,+], are the initial cell, glutamine, and ammonia concentrations, respectively. The parameters p , qGln, and pNH4+ can be found by linear and nonlinear regression techniques. We use a linear regression by arranging eq 9 as e-kt
[Gln] - [Gln]oe-kt= qoe(XoV-)
- ,rt
k+P
(11)
Hence, a plot of [Gln] - [Gln],e-kt versus Xov(e-kt- ePt)/ (k + p ) should give a straight line with a slope of qGln. Only the data confined to the exponential growth phase is used. The results from this rocedure for our data is qGln = 0.041 f 0.003 pmol/(lO cells/h). Similar linear transformation for eq 10 can be constructed, and it gives of 0.021 f 0.003 pmol/(106 cells/h). a value for pNH4+ These values are close to the ones obtained from the differential method (Figure 5). The apparent uptake rates can be evaluated by equating k to zero in eqs 8-10. By taking the ratio between the cell density and the concentrations of glutamine and ammonia given in these equations (with k = 0), one obtains
B
= [NH,+]o
P'NH~+
+ -(Xv P
- Xov)
- X",)
(12)
Hence, a plot of [Gln] and [NH,+] versus X , should be straight lines with slopes equal to qtGln/pand P ' ~ ~ , + / P , respectively. The plots for glutamine and ammonia are presented in Figure 6B. From the slopes of these lines, one can obtain q'Gln = 0.078 f 0.003 Pmol/(1O6 cells/h) and pINH4+ = 0.053 f 0.002 pmol/(106 cells/h). These two apparent rates are more than 200% higher than the actual ones. The yield coefficient is then 0.68 mol of ammonia produced/mol of glutamine consumed. This apparent yield is 31% higher than the true one. Parametric Sensitivity of t h e Error. Since glutamine is an important substrate for the cultivation of many cell lines in vitro and its chemical decomposition is frequently ignored in the kinetic analysis of cell culture data (see, for instance, Dean et al., 1984),it is important to assess the seriousness of such an omission and how the culture conditions effect the magnitude of this error. The error made by ignoring glutamine decomposition will depend on a number of factors, such as the initial cell and glutamine concentrations and the growth and decomposition rates. We now analyze this error quantitatively. Using the equations for the glutamine time profile (eq 9) and using eq 12 for the apparent glutamine uptake rate, we obtain [Gln], - [Gln] - [Gln],(l- e-kt) qGhXov e p t - -kt e (13) - X", x, - X " , k + p X,-X",
x,
+-
126
Biotechnol. Prog., 1990, Vol. 6, No. 2
= 4.365984
1bA(1.7317e.1x)
R”2 = (396
10‘
0
20
60
40
80
Time, hr
zE
47
1
m“
. I
C
0
E E
I
I
c)
3
c
Oe+O
2e+S
4e+S
6e+S
8e+S
ues of a. For our experimental conditions, k = 0.0023 h-I and the growth rate is p = 0.04 h-l, and hence, we have a = 0.058. (b) The parameter, P, is comprised of initial cell and glutamine concentrations and by the growth and apparent glutamine uptake rates. In our experiments, the initial cell density was 4 X lo4 cells/mL and the initial glutamine concentration was 3.6 mM. The apparent glutamine uptake rate was determined above to be q’Gln = 0.079 pmol/(106 cells/h), and with a growth rate of p = 0.04 h-l, we calculate P to be 46. (c) The third parameter, y, represents the decrease in the glutamine concentration during the period of exponential growth. It is typically between 0.5 and 0.2. For our data, we had about 1.9 mM glutamine at the end of the exponential phase (at 75 h), and with an initial glutamine concentration of 3.6 mM, we get a y value of about 0.5. Thus, if one prepares a plot of [Gln] versus the cell concentration (as in Figure 6), one can evaluate the apparent uptake rate of glutamine and then calculate the numerical values for a , P, and y. Then, to obtain the actual glutamine uptake rate, one can calculate the correction directly from eq 14 or estimate it from Figure 7. For instance, for our the experimental data, with (Y = 0.058, P = 46, and y = 0.5, eq 14 gives the ratio of 0.55. Hence, the actual uptake rate is q G l n = (0.55)0.079 = 0.048 pmol/ (lo6 cells/h). This value is close to the one obtained above when both chemical decomposition and glutamine metabolism are considered. A similar mathematical analysis can be used to get the relationship between the apparent and true ammonia production rates:
le+6
Viable cell/ml Figure 6. Evaluation of apparent glutamine uptake and ammonia production rates by graphical method for the experimental data obtained in the growth of hybridoma cell: (A) cell growth rate; (B) glutamine uptake and ammonia production rates.
After some mathematical manipulations, this leads us to a relationship between the actual uptake rate, q G l n , and the apparent one, qlGln,by
where
The ratio of actual uptake rate to the apparent one, gives the error introduced by ignoring the decomposition rate of glutamine. From eq 14, one sees that the error is governed by the three dimensionless parameters given in eq 15. Figure 7 shows the variation of the error with the three dimensionless quantities. The meaning and representative numerical values for these parameters are as follows: (a) The parameter a is the ratio of two rate constants: decomposition rate constant to the growth rate. For values of pH 7.0-7.6, we see from Figure 3 that representative values fork = 0.001-0.004 h-’ (with RPMI-1640 being the exception). Hybridoma growth rates typically fall in the range 0.02-0.05, so a is on the order of 0.02-0.2. Figure 7 indicates that the error in ignoring the chemical decomposition of glutamine is very large for these valq’Gln/qGln,
The same dimensionless parameters appear in this equation as in eq 14. An additional parameter appears in this equation, which is the apparent yield coefficient of ammonia from glutamine, Y’ = P’NH,+/q’Gln. This value is readily obtainable from the apparent rates and varies between 0.5 and 1 in most cases. The case of Y’ = 1 is special because the ratio of ammonia production rates becomes equal to the ratio of glutamine uptake rates. Here, we analyze the error in the ammonia production rate for the other limit of Y’ = 0.5. Equation 16 was used to prepare the curves in Figure 8. The trend is the same as is in the case of glutamine uptake rates. Both the value of a and p decrease the ratio and hence increase the error. The error is more pronounced in the case of ammonia as evident from the ratios at the same parameter values. In the hybridoma metabolism example, we had the parameters of a = 0.058, 0 = 46, and y = 0.5. The ratio q G l n / q ‘ G l n was obtained as 0.55 from the analysis of the glutamine uptake rate. The apparent yield coefficient was 0.68 obtained from the data on Figure 6. + = 0.34. Hence, Then, eq 16 gives the ratio of pNH4+/pfNH the actual production rate is pNH+ = (0.?35)0.053 = 0.019 pmol/(106 cells/h). This value is close to the one obtained above when both chemical decomposition and ammonia metabolism are considered.
Discussion The inclusion of the rate of chemical decomposition of glutamine is important for kinetic studies of mammalian cell metabolism in vitro. We have measured the
Biotechnol. frog., 1990, Vol. 6, No. 2
-.-u.00
0.05
127
0.10
0.15
-.-
0.20
0.W
0.05
a
0.15
0.10
0.20
a
3 U
u.00
0.05
a.10
0.15
0 . 0 1 ' . W.00 0.05
0.20
.
