Influence of Assay Methodology on the Measurement of Free Serum Ceftriaxone Concentrations

Abstract

The influence of assay methodology on the measurement of the active free fraction of ceftriaxone in plasma was determined. The free fraction was measured by three methods: agar diffusion bioassay, precipitation of plasma protein with methanol followed by high-performance liquid chromatography (HPLC) of the supernatant, and ultrafiltration of plasma followed by HPLC of the filtrate. In human serum, the free ceftriaxone levels were significantly lower (P = 0.03) when measured on ultrafiltrates compared to the other two methods. This difference disappeared when dolphin serum was studied. After ultrafiltration, human serum was shown, by Scatchard plot analysis, to have two ceftriaxone binding sites. Species differences were also demonstrated. Hence, in humans, determination of free plasma ceftriaxone varies with the assay method employed.

Many newer antimicrobial agents demonstrate high degrees (>80%) of protein binding that limit the amount of free drug present. The extent of protein binding may show significant interspecies variation, thus complicating veterinary dosing regimens and the indiscriminate extrapolation from animal models to humans (4). Although only free drug is microbiologically active (4), National Committee for Clinical Laboratory Standards guidelines are still set with total drug levels. Overestimates of the amount of highly protein-bound drugs to be administered can ensue, particularly in hypoalbuminemic states related to illness or renal impairment (8, 9). Techniques to determine the amount of free drug in plasma include agar diffusion bioassay, protein precipitation with subsequent high-performance liquid chromatography (HPLC) assay of supernatant, and membrane ultrafiltration with HPLC assay of the ultrafiltrate (4). We determined the amount of free plasma ceftriaxone in human and other animal sera.

MATERIALS AND METHODS

Ceftriaxone standard powder was a gift from P. F. Sorter, Hoffmann-La Roche, Inc., Nutley, N.J. HPLC-grade acetonitrile was purchased from J. T. Baker, Inc., Phillipsburg, N.J. Bacto Antibiotic Medium 1 and Bacto Agar were purchased from Difco Laboratories, Detroit, Mich. Fatty acid-free human and horse albumins were purchased from Sigma Chemical Co., St. Louis, Mo. All other chemicals were reagent grade and purchased from Sigma. Ultrafree-MC 10,000 nominal molecular weight limit filter units were purchased from Millipore Corp., Bedford, Mass.

HPLC ceftriaxone concentration analysis was based on the methods of Granich and Krogstad (7). Chromatography was carried out with Beckman 110A pump; an Econosphere C₁₈ 5-μm silica gel column purchased from Alltech Associates, Inc., Deerfield, Ill.; a Hitachi L-4250 UV-VIS detector set at 280 nm and either 0.02 or 0.002 absorbance unit full scale; and a Hewlett-Packard HP 3396 series II integrator. The mobile phase consisted of 3 g of hexadecyltrimethylammonium bromide, 10 ml of 1 M potassium phosphate buffer (pH 7.0), 600 ml of acetonitrile, and double-distilled water to make 1 liter. Chromatography was carried out at room temperature and a 1-ml/min flow rate. Ceftriaxone eluted at approximately 5 min under these conditions. Samples were prepared for HPLC analysis by either cold-methanol precipitation or ultrafiltration.

For precipitation of serum proteins by cold methanol, 0.1 ml of well-mixed serum with drug was added to 0.9 ml of HPLC-grade methanol stored at 4°C. The mixture was vortexed and then centrifuged at 3,500 rpm in a CRU5000 floor model IEC centrifuge with an 8-in. radius (2,780 × g) for 5 min to precipitate the protein. The supernatant was assayed for ceftriaxone. Calculation of serum ceftriaxone concentrations included a correction for the 1:10 dilution introduced in the sample preparation.

For ultrafiltration, 0.2 ml of well-mixed serum with drug was placed in the upper reservoir of the Millipore Ultrafree-MC filter apparatus. The sample was then centrifuged for 10 min in the HF-120 benchtop centrifuge supplied by Millipore for use with the filter (12). The filtrate was assayed for ceftriaxone. Binding of ceftriaxone to the filter apparatus itself was not distinguishable from zero (recovery, 100% ± 5%).

Standard curves were constructed by twofold serial dilution of standard ceftriaxone powder dissolved in phosphate-buffered saline (PBS) at pH 7.0. For the bioassay only, a standard curve was also constructed in serum. Bioassay was performed in 4% Bacto Antibiotic Medium 1 seeded with Bacillus subtilis ATCC 6633. Wells were cut in the agar, and 50 μl of standard, controls, or samples to be assayed was placed in the wells and incubated at 37°C for 18 h. Diameters of the clear zones, measured with calipers, were linear over the range studied (1 to 1,000 μg/ml). The HPLC standard curve was based on the integrator reported area under the curve and was linear over the range studied (6.3 to 800 μg/ml). The lower limits of detection by the HPLC assay were 5 μg/ml for the diluted, methanol-precipitated sample and 0.5 μg/ml for the filtered sample. Interday and intraday coefficients of variation measured with ceftriaxone diluted in PBS were 6.1 and 5.8%, respectively, at clinically relevant concentrations (25 and 100 μg/ml). The precision and coefficient of variation of the bioassay were 0.6 and 4.7% in either H₂O or serum.

Antibiotic-free horse serum and horse ventricular cerebrospinal fluid were supplied by N. C. Ringger, Department of Veterinary Medicine, Oregon State University, Corvallis. Antibiotic-free dolphin serum was supplied by Robert Ulrich, School for Pharmacy, University of Southern California, Los Angeles. Antibiotic-free human middle ear fluid was supplied by Leslie Serchuck, Department of Pediatrics, Boston University School of Medicine, Boston, Mass. Antibiotic-free chinchilla serum was supplied by Pilar Tam, Department of Pediatrics, Boston University School of Medicine. Antibiotic-free human serum was supplied by S.J.K. Ceftriaxone was added to all of the above antibiotic-free biological fluids at concentrations of 6.3 to 800 μg/ml for recovery experiments. Serum from patients receiving ceftriaxone was provided by James Leggett after informed consent had been obtained in accordance with institutional guidelines.

Statistical analysis was performed with the WinSTAR program purchased from Anderson Bell Corp., Arvada, Colo. Groups were compared by using analysis of variance or Students’ t test, and P values of 0.05 or less were required for significance. A Scatchard plot was constructed in the usual fashion (1).

RESULTS

The three assay methods differed in the amount of free ceftriaxone they measured in human serum to which ceftriaxone had been added. In Fig. 1a, the agar diffusion bioassay and the HPLC assay of methanol-precipitated serum yielded similar results (P = 0.8) at all of the concentrations tested. Recovery of added drug was constant at 70 to 80% for ceftriaxone concentrations of 12.5 to 800 μg/ml. Free ceftriaxone levels, assayed by HPLC, in filtered serum were significantly lower (P = 0.03). Drug recovery was concentration dependent, between 5 and 25% over the range of concentrations studied. Drug levels measured by ultrafiltration were similar to those reported in previous studies (13) that used an equilibrium dialysis method (Fig. 1a). Free ceftriaxone levels were also measured in 60 patients receiving treatment with 1 or 2 g of the drug administered intravenously per day. Free drug levels in serum determined by agar diffusion correlated well with levels measured by HPLC assay of methanol-precipitated serum (r² = 0.802) over a range of 4 to 210 μg/ml. In contrast, HPLC assay of serum submitted to ultrafiltration differed significantly from the agar diffusion bioassay (P = 0.03). The agar diffusion bioassay performed by using serum underestimated ceftriaxone concentrations determined by using PBS by 12% over a concentrations range of 6.25 to 100 μg/ml. The shapes of the curves in Fig. 1 were shifted vertically when serum was used as the standard curve.

FIG. 1.

Ceftriaxone recovery from human (a) and dolphin (b) sera to which known amounts of ceftriaxone had been added, determined by agar diffusion (▵), HPLC following methanol precipitation of denatured proteins (□), and HPLC following ultrafiltration (○). All values represent means ±1 standard deviation calculated from a minimum of two experiments performed in triplicate. ■, historical values of free ceftriaxone separated by equilibrium dialysis (14).

Figure 1b depicts the measurement of free ceftriaxone in dolphin serum to which ceftriaxone had been added. No significant differences were revealed among the three assay methods (P = 0.89). At each drug concentration, approximately 70% of the added drug was measured. Methanol precipitation of protein and membrane ultrafiltration (data not shown) were compared in horse and chinchilla sera to which ceftriaxone had been added. There were no differences between the two methods. It is of interest that in human spinal fluid and middle ear fluid, the assay method made no difference in the determination of the amount of free ceftriaxone (data not shown).

To clarify ceftriaxone protein binding, free and bound forms of the drug were separated by membrane ultrafiltration and Scatchard plot analysis was performed. Human albumin and serum displayed two binding sites for ceftriaxone (Fig. 2). In contrast, horse albumin and serum showed a single binding site (data not shown). Table 1 lists ceftriaxone binding constants in a variety of mammalian sera. Dog, dolphin, and horse sera analyzed by Scatchard plotting displayed only one binding site, while baboon, human, rabbit, and rat sera displayed two sites.

FIG. 2.

Scatchard analysis of ceftriaxone bound to human serum (▴) and human albumin (•). Free-ceftriaxone levels were determined by HPLC quantitation after ultrafiltration to separate free from bound drug.

TABLE 1.

Binding constants of ceftriaxone in different biological fluids determined by Scatchard analysis

No. of sites and mammal	Ceftriaxone binding constant
	105 M−1	104 M−1
Single site
Doga	1.7
Dolphinb	5.0
Horseb
Serum	2.7
Albumin	0.6
Double site
Baboona	0.8	2.8
Humana	2.7	1.6
Humanb
Serum	2.9	7.2
Albumin	0.4	29
Rata	1.7	1.4
Rabbita	0.86	0.38

^aFree, separated by dialysis (11). 

^bFree, separated by filtration. 

DISCUSSION

Accurate determination of the active free fraction of highly protein-bound antibiotics is important in determining antimicrobial diffusion into, and activity at, sites of infection such as the middle ear and cerebrospinal fluid. Although equilibrium dialysis is the classic procedure for separating free from bound drug, it is too cumbersome for routine clinical use (6). There is no widely accepted alternative technique. Published assay methods advocate agar diffusion bioassay or HPLC assay after separation of free from bound drug by either protein precipitation or membrane ultrafiltration.

We compared three methods to determine the free fraction of ceftriaxone in serum and body fluids. The apparent free fraction in human serum, over a clinically relevant range of concentrations, was higher when measured by either bioassay or protein precipitation followed by HPLC compared to membrane ultrafiltration followed by HPLC. The results are consistent with the concept that there are two binding sites for ceftriaxone on human albumin. Further, it appears that the binding avidity is more tenuous for one of the two sites. We postulate that since ultrafiltration is more gentle than precipitation, ultrafiltration preserves both the lower-capacity, high-affinity binding site and the higher-capacity, lower-affinity binding site. On the other hand, precipitation with methanol preserves only the more tightly bound site, as reflected by higher measured free ceftriaxone levels. The methanol results are similar to those reported for acetonitrile. Precipitation of serum proteins with acetonitrile resulted in supernatant-free drug levels comparable to those found by bioassay (0.99 correlation coefficient) (7).

Our study confirms previous reports of two binding sites in human serum (11, 14) and concentration-dependent protein binding of ceftriaxone (2, 3, 6). We assume that the assays for free drug in middle ear and cerebrospinal fluids reflect the paucity of albumin in these fluids. The mechanism of variable degrees of binding of ceftriaxone by serum proteins from different species is unclear, but the heterogeneity of drug protein binding among mammalian species is well recognized (4).

In the past, agar diffusion and HPLC after serum protein precipitation have been used to determine total (free and bound) ceftriaxone levels. Our data suggest that these methods underestimate the total concentration of ceftriaxone by approximately 20 to 30%, as recovery of ceftriaxone added to serum was measured at 70 to 80%. Others report a 90 to 100% recovery of added drug (8, 10). The disparity stems from methodological differences. We constructed our standard curves in both aqueous and serum buffers. We used an aqueous buffer to ensure that all protein binding could be detected. Previously reported total drug levels do not account for the approximately 20% of bound drug that fails to precipitate with methanol or does not diffuse freely in agar. The combination of undetected, bound drug and concentration-dependent shifts in free drug levels might contribute to the difficulty some investigators have encountered in trying to fit ceftriaxone pharmacokinetics to standard two-compartment pharmacokinetic models using either first- or second-order kinetics (5, 10, 11).

In conclusion, the determination of free ceftriaxone levels in humans is dependent upon the assay technique used. Use of methods that overestimate the free-drug concentration may result in misinterpretation of data (9) or inappropriate calculations of dosing regimens (12). Further, the magnitude of drug-protein binding varies from animal species to animal species. Although our experiments focused on ceftriaxone, the results indicate the potential importance of assay methodology for other highly protein-bound drugs.