'1906
Anal. Chem. 1991, 63,1906-1909
Operation of Ion-Selective Electrode Detectors in the Sub-NernstiaWLinear Response Range: Application to Flow-Injection/Enzymatic Determination of L-Glutamine in Bioreactor Media Wojciech Matuszewski,' Sara A. Rosario, and Mark E. MeyerhofP Department of Chemistry, The University of Michigan, Ann Arbor, Michigan 48109
A novel approach for ellmlnatlng podtlve errors from endogenous lonlc Interferences when uslng ion-selective eiectrodes as detectors in IkwkrJectkn enqme-based bksenskrg wnflgwatbns Is described. The method involves uslng a hlgh background level of lnterferlng Ions in the sample dlluent/ carrier stream to convert the normally logarlthmic potentlomelrlc sensor Into a llnear detector over a wen concentratkm range of primary Ions. A split-stream slngle-detector arrangement provides a convenlent means to compensate for varying levels of background interferent Ions in the InJected samples. One portlon of the spllt stream passes directly to the lonalectrode detector, yleldlng a signal linearly related to the concentratlon of endogenous primary Ions in the sample. The second portlon d the spllt sample Is delayed whlle passing through an hnmoblllzed enzyme that generates electrode detectable primary Ions In proportlon to the concentratlon of the substrate analyte In the sample. Two linear equatbns wlth two unknowns descrbe the twln potentbmetrk responses observed. The concept Is demonstrated by the accurate detennlnatlon of Lglutamlne In h y b t " a Moreactor media via the use of an ammonium-Ion-selectlve membrane electrode detector and lmmoblllzed glutaminase enzyme.
INTRODUCTION Over the past 2 decades, considerable effort has been expended on coupling ion- and gas-selective membrane electrode detectors with immobilized enzymes for sensing specific biomolecules including amino acids (1-5). Whether in the form of probe-type enzyme electrodes or flow-through configurations in which the enzyme is immobilized upstream from the potentiometric detector, such systems have failed to gain wide use owing to the analytical errors caused by interfering endogenous ion/gas species present in complex samples of interest (e.g., blood, urine, bioreactor media, etc.). Previous efforts in this laboratory (6-9) and elsewhere (9-15) have focused on examining ways to reduce or eliminate such interferences, particularly with respect to using a large number of known enzymes that liberate ammonia/ammonium ions as the electrochemically detectable product. In such instances, the presence of endogenous ammonia nitrogen (ammonia-N; total ammonia gas and ammonium ions) can cause positive errors. Most recent approaches have included on-line removal, exclusion, or consumption of endogenous ammonia-N (and some other ionic interfering species) through the use of gaspermeable tubes, anion-exchange membranes (6-8), or additional enzymes/reagents (e.g., glutamate dehydrogenase) (10-13). While efforts to correct for background sample am'On leave from the Department of Chemistry, University of
Warsaw, Warsaw, Poland.
0003-2700/91/0363-1906$02.50/0
monia-N via a completely separate ammonia measurement (no enzyme) have been suggested (9, 14,15), this increases the complexity of the bioanalytical method. Moreover, the innate logarithmic response of the ammonia-gas-selective or ammonium-ion-selectiveelectrode detectors creates significant precision and accuracy problems when the background level of ammonia-N approaches or exceeds the concentration of anal* present in the sample. We now describe an alternate approach for addressing this problem through operation of an ammonium-ion-selective electrode as a low-signal linear detector in a split-stream flow-injection analysis (FIA) arrangement incorporating an immobilized enzyme. This concept is further applied by devising a system suitable for the direct determination of L-glutamine, the key amino acid nutrient in hybridoma bioreactor media (16). Figure 1 illustrates the experimental arrangement that can be used for the FIA determination of L-glutamine. The injected sample is split into two streams; one flows directly to the potentiometric ammonium ion electrode detector with only a short delay in dispersion loop L,. The second portion of the sample first passes through a packed controlled pore glass reactor (CPG) containing immobilized glutaminase enzyme, through a longer delay loop L2, and then to the electrode detector. The enzyme catalyzes the conversion of glutamine to ammonium ions and glutamate. This arrangement results in two sequential detector responses (see Figure 1). If the electrode detector is operated under conditions that yield a linear response to the endogenous sample ammonium-N (first peak), as well as to the total ammonium ions present after passing through the enzyme reactor (second peak), two simultaneous linear equations can be solved to obtain the concentration of glutamine in the injected unknown sample (see below).
THEORY The potentiometric response of the ammonium-ion-selective electrode detector can be described by the well-known Nikolsky-Eisenmann equation, which in a simplified form can be expressed as follows: Eeiect
=K
+ S log (CNH,++ L )
(1)
where S is the slope of the electrode, K is a constant, C-+ is the concentration of ammonium ions in the solution in contact with the electrode, and L is expressed as
L = CkRff4,+$Cj1Izj j
(2)
In this case, L represents the contribution of the measured electrode potential attributable to all background interferent ions (j)present in the solution in contact with the electrode. Equation 1 can be rewritten in natural logarithm form as
Eelect= K
+ RT - In (CNH,++ L ) F
0 1991 American Chemical Society
(3)
ANALYTICAL CHEMISTRY, VOL. 63, NO. 18, SEPTEMBER 15, 1991
L C
R
,
CPG 6.0
h"
~
W
I
U
-0 1. Schematic diagram of the sptit-stream FIA system used to determine glutamine in bkreactor medla: (C) debnired water carrier stream: (R) buffer diiuent reagent stream; (S) sample Injection point: (L, and L,) detlay coils; (CW) controikd pore glass enzyme reactor: ( I S ) ammonlumkn-setedve membrane electrode and reference electrode; (W) waste.
The In term can be approximated by an infiiite Taylor series
as
+
In (Cm+ + L) = 1 n L + 2
For cases where the total concentration of interfering ions is much greater than the concentration of detected ammonium ions (i.e., 2L >> , )+C then
em4+ >>!(
2L+c,,+
CNb+
3 2L+C"
)'+A(
5
cm+
)"+
2L+cm+
... ( 5 )
and, therefore,
Suhtituting back into eq 4 and 3 yields a simplified expression for the electrode potential as a function of ammonium ion concentration:
E,,,
K
RT + RT In L + F FL 'm+
(7)
Since L is held constant by operating the electrode with a high concentration of interfering ions in the carrier stream electrolyte, the electrode will respond linearly to the concentration of ammonium ions up to the point where L is no longer significantly greater than the sample ammonium ion concentration. That is,
Eel,
A
+ BCm+
(8)
where A and B are constants. Since the dispersion coefficient (17)of the injectad sample slug will be unequal in each path of the split stream, there will be different sensitivitiea in the mponse of the ammonium ion electrode detector to the endogenous ammonium ions in each portion of the split sample. F'urthermore, for the portion of sample that passe% through the enzyme reactor, additional potentiometric response will be observed due to the production of ammonium ions from Gglutamine in the sample. Therefore, two linear equations describe the two potentiometric signals that result:
peak 1 = Hl = Kl[NH4+]
1807
(9)
+
peak 2 = H2= K2[NH4+] K3[glutamine]
(10) where Kl and K2are the sensitivities of the detector to endogenous ammonium ions in each split segment of the FIA manifold and K3 is the sensitivity to glutamine. The sensitivity to glutamine will be highly dependent on the dispersion coefficient of the portion of the sample slug that flom through the enzyme reactor as well as the enzyme loading and concomitant glutamine conversion efficiency of the immobilized enzyme column. In practice, K1,K2,and K s can readily be determined by injections of ammonium chloride and glutamine standard solutions into the system prior to analysis of unknown samples. EXPERIMENTAL SECTION Apparatus. The flow-injection manifold (Figure 1) was assembled by using a Rainin Rabbit peristaltic pump (Woburn, MA) to deliver the carrier buffer and a Rheodyne six-port low-pressure rotary valve for sample injection. The sample (typically 20 pL), injected at point S into a carrier stream of pure water (C), was subsequently merged with a reagent buffer stream (R)( t y p i d y 0.1 M lithium acetate buffer, pH 4.9,containing high background concentrations of NaCl and/or KCl). This diluted sample stream was then split into two channels. One portion of the sample stream passed through a short length (100cm) of a narrow bore (0.26mm4.d.) Teflon tubing loop (L1)and then on the membrane electrode detector. The other portion of the split sample stream flowed through an immobilized enzyme reactor (3-mm i.d. by 8o-mm length; see below) and then a longer Teflon tubing delay loop (200 cm, L2)before passing through the electrochemical detector. The fraction of sample flowing through each channel of the FIA system was controlled by the relative resistance to flow of each segment of the manifold. The ammonium-ion-selectivemembrane electrode was prepared by incorporating nonaction into a plasticized poly(viny1 chloride) (PVC) membrane (18) and then mounting a piece of this membrane within a Phillips electrode body (ISE-661, Glasblaserei Maller, Zurich). The electrode was fitted with a special cap for use as a flow-through potentiometric detector in a large-volume wall-jet confiiation (19). The membrane electrode, connected to the FIA manifold via a short length of narrow bore Teflon tubing, and a saturated calomel reference electrode (SCE) were placed in a large beaker of reagent buffer (see ref 19). Potentiometric response of the working electrode relative to the reference was deteded with an Accumet Model 910 pH/mV meter (Fisher Scientific, Romulus, MI) and recorded on a Fisher Recordall Series 5000 strip-chart recorder. The immobilized enzyme reactor was prepared according to the procedure reported by Masoom and Townshend (20).Twohundred milligrams of controlled pore glass (170& 2OlHOO me&) from Sigma Chemical Co. (St. Louis, MO) was employed for the immobilization of 50 unita of glutaminase enzyme (grade V from Sigma). The porous glass beads with the immobilized enzyme were packed into an Wmm length (3-mm id.) of glass tube (with glass wool on one end). The ends of the reador were fitted with plastic sleeves for easy connection of the Teflon tubing of the FIA manifold The enzyme reactor was filled with 0.1 M acetate buffer, pH 4.9,and kept at 4 O C when not in use. RESULTS AND DISCUSSION Preliminary experiments focused on demonstrating that the ammonium-ion-selective electrode could function effectively as a low-signal linear detector toward ammonium ions in the presence of high levels of interfering ions. To this end, both single-line and splibstream FIA manifold arrangement8were used. Figure 2 illustrates the electrode's response (recorded peak height in mv) toward different concentxations of injected ammonium chloride standards using the split-stream configuration and two different carrier stream buffer reagenta. In the absence of high levels of sodium and potassium in the lithium acetate carrier buffer, each of the two resulting signals exhibits the expeded Nernstian-type behavior (see Figure 2A).
1008
ANALYTICAL CHEMISTRY, VOL. 83, NO. 18, SEPTEMBER 15, 1991 (A)
150
A B C D E
20 min
- 5
-3
- 4
- 2
log C, M
Figure 2. Potentiometric response of the split-stream FIA system to varying concentrations of injected ammonium chlorMe using two different diiuent reagent streams: peak 1 ( 0 )and peak 2 (A)for pH 4.9
(8)
0.5 mM glulamlnb
acetate buffer with no interfering ions added: peak 1 (W) and peak 2 (A)for the same reagent buffer with 0.2 moi/L NaCi and 0.01 moVL KCI added. (A) Nernst equation plots: (B) data plotted on linear coordinates. The magnitude of the two signals is not the same due to the different effective dispersion coefficients of the sample slug in each of the two channels of the manifold (D= 8.3 and 15.8, respectively). That is, there is more dilution for the portion of sample traversing the longer segment containing the packed reactor. As expected, when the reagent stream buffer contains 0.2 mol/L NaCl and 0.01 mol/L KCI, a dramatic decrease in the total potentiometric response toward the ammonium chloride standards is observed. It should be noted that the potentiometric selectivity coefficients (Kf&+j)for the nonactin-based ammonium ion electrode with respect to Na+, K+, and Li+ are approximately lO-l, and lod respectively (21). Consequently, the presence of Na+ and K+ ions in the sample stream yields a constant increase in the background of interfering ions equivalent to 1.2 mmol/L ammonium (see eq 2). This results in a response to injected ammonium ions for both peaks that is linear with concentration over the range 0-8 mmol/L ammonium (see Figure 2B). It is important to recognize that the actual concentration range in which linear response is observed can be varied for a given application merely by changing the concentration of added interfering ions and/or the manifold setup (e.g., flow-rates, tubing lengths, etc.) to obtain different dispersion coefficients in each segment of the system. With the glutaminase enzyme reactor in place, the FIA manifold shown in Figure 1 can be used to generate linear responses to both glutamine and ammonium ions within injected samples in accordance with eqs 9 and 10. Figure 3A shows a typical strip-chart recording obtained for injections of glutamine standards alone (in the range 0.5-4.5 mmol/L) and in the presence of a fixed level of background ammonium ions (2 mmol/L). Note that when glutamine is present alone, only one detector signal is observed, and this is related linearly to glutamine concentration. When high levels of ammonium ions are present, two signals for each injected sample are observed, the first related only to the ammonium ions present (constant in this case), and a variable second signal directly proportional to the sum of ammonium ion and glutamine levels. While the net signals for both species are quite small in voltage terms, they are clearly above the background signal (actual sensitivities vary day to day: see below). Indeed, the presence of high background levels of interfering ions in the carrier/diluent reagent actually helps to reduce the baseline noise (to
