The production efficiency of 10,000- and 30,000-MW cutoff hollow-fiber bioreactor cartridges using the Cellex, Inc., Acusyst Jr. and the Unisyn Technologies, Inc., CP2000 bioreactor systems was investigated. Both cartridge types produced significant quantities of mouse monoclonal antibody, but the production efficiency (milligrams of antibody produced per liter of medium consumed) was higher for the 30,000-MW cartridge. Also, the smaller pore size cartridges displayed a lengthened lag phase in antibody production and cell growth of 26 to 33 days versus 11 to 20 days for the 30,000-MW cartridge. High intracapillary medium feed rates (5.0 L/day for 30,000 MW and 8.4 L/day for 10,000 MW) may reduce overall antibody production efficiency, and lower feed rates (3.5 and 7.2 L/day, respectively) may produce more antibody per liter of medium consumed.

The periodic removal of excess cells from confluent bioreactor cartridges can augment antibody production and extend the length of the run. Batch harvesting of antibody may lead to more highly concentrated antibody product than continuous harvesting.

Bioreactor systems that use hollow-fiber cartridges as a matrix for the growth of hybridoma cells are an efficient way of producing multigram quantities of monoclonal antibody.^1­3 Hollow-fiber cartridges can produce highly concentrated monoclonal antibody relatively free from contamination by cellular or serum proteins.⁴ These cartridges are available primarily in molecular weight pore sizes of 10,000 (e.g., renal dialysis cartridges) or in larger pore size configurations manufactured exclusively in bioreactor systems.

Using the Cellex, Inc. (Minneapolis), Acusyst Jr. and the Unisyn Technologies, Inc. (Tustin, CA), CP2000 bioreactor systems, we investigated the ability of several types of hollow-fiber cartridges to sustain the growth of mouse hybridoma cells and to produce antibody. We also studied the effects of several key operating parameters, including medium feed rate, antibody harvest method, cell removal, and glucose utilization rate (GUR) on antibody production.

Materials and Methods

Cell Lines. Mouse hybridomas VA1 18D7.3.2.4 (antivalproic acid) (IgG1), MOR 9B1 (antimorphine) (IgG1), METH7-4B6 (IgA) (antimethamphetamine), and PCP1.9D10 (IgG1) (antiphencyclidine) were grown in Iscove's Modified Dulbecco's Medium (IMDM) (Mediatech, Inc., Herndon, VA) supplemented with 2 mM L-Glutamine (Gibco, Grand Island, NY) and 10% fetal bovine serum (FBS) (Hyclone Labs, Logan, UT). All hybridomas were generated using the NSO fusion partner.

Baseline static culture productivity for each cell line was established by replicate seeding of 5 x 10⁴ cells/ml in growth medium (described above) in standard T25 flasks (Corning, Corning, NY). The flasks were incubated at 37°C and 7% CO_² for 72 hours; then the medium was removed and centrifuged at 1000 rpm for 10 minutes. The resulting supernatant was assayed for total and specific immunoglobulin by ELISA (Table I), as described below in the section "Determination of Antibody Concentration." (Tables not yet available on-line.)

Bioreactor Systems and Hollow-Fiber Cartridges. The Acusyst Jr. was run with the Acucell 1100 (Cellex, Inc.) 10,000-MW cartridge with 19 sq ft of surface area and 120 ml of extracapillary space (ECS) (Table I). The cartridge uses a reverse-flow cycling process to actively transport medium back and forth between the intracapillary (ICS) and the ECS. The system was unique to the Acusyst Jr. and its cultureware and was not available for the other cartridges or systems tested.

The CP2000 was used to compare two other 10,000-MW cartridges: the Clirans T220 cuprammonium rayon kidney dialysis cartridge (19-sq-ft surface area, 100-ml ECS) (Terumo Corp., Somerset, NJ) and the BR2010 ampholytic copolymer membrane cartridge (35-sq-ft surface area, 210-ml ECS) (Unisyn Technologies, Inc., Tustin, CA). These were also compared to the BR3530 cellulose acetate 30,000-MW bioreactor cartridge from Unisyn Technologies, Inc. (Table I).

General Bioreactor Run Parameters. The mouse hybridoma cultures were grown in the IMDM formulation. Cells in log phase growth (>85% viable by trypan blue exclusion) were centrifuged at 400 x g for 5 minutes, then resuspended in fresh growth medium. Approximately 10⁹ cells were then introduced into the ECS of each bioreactor cartridge by a large syringe.

The hollow-fiber bioreactor cartridges were tested in the two bioreactor systems with IMDM as described above. The FBS concentrations of the ICS medium for all the bioreactor experiments except those using MOR 9B1 and PCP1.9D10 were 3% for the first 20 L of media and 1% for the remainder of the experiment. The FBS concentration in the ICS for the MOR 9B1 and PCP1.9D10 runs was 3% throughout the course of the experiment. The serum concentration for the MOR 9B1 and PCP1.9D10 experiments was kept at 10% throughout. The FBS concentration of the ECS for the other cell lines was gradually reduced from 10 to 1% by introducing the ICS feed medium throughout the experiment.

The operating parameters of each bioreactor run were manipulated to maximize productivity. They varied according to the requirements of the cell line, bioreactor system, and bioreactor cartridge. The initial intracapillary recirculation rates for each of the CP2000 runs ranged between 350 and 550 ml/min and were gradually increased to the maximum setting of 999 ml/min. The recirculation rate of the Acusyst Jr. was between 200 and 400 ml/min. It could not be run with a recirculation rate >400 ml/min because of pump system limitations. Temperature was automatically maintained by the two systems at 37 ±0.5°C. The 100% air and carbon dioxide flow rates were held at approximately 100 and 5­10 cm³/min, respectively, throughout the course of each CP2000 run. The pH for the runs ranged between 6.9 and 7.3.

Determination of Glucose Utilization Rate. The GUR is an indicator of cellular metabolism and cell growth.⁵ The relative ability of the cartridges to sustain the growth of cells was assessed by measuring the amount of glucose from the ICS used by the cells. Glucose concentration in the circulating bioreactor ICS medium was determined using Roche Glucose Reagent and Roche Calibrator for Serum on the COBAS BIO instrument (all from Roche Diagnostic Systems, Inc., Somerville, NJ).

A bioreactor with a low GUR probably has too few metabolically active cells to demonstrate adequate antibody production. A high GUR may lead to rapid overgrowth and premature senescence. We have, therefore, established a target GUR for the different bioreactor configurations. At the target GUR, run parameters such as medium feed rates and recirculation rates are optimized to maximize antibody production. The target GUR is determined empirically by growing each cell line at different medium feed rates; it depends on the cell line and cartridge. A target GUR of 250 mg/hr for each bioreactor system was established.

The GUR (expressed as mg/hr) was determined by the formula:

GUR = {feed rate (ml/hr) x [glucose in (mg/L)­glucose out (mg/L)]}/1000

where glucose in is the concentration of glucose in the feed medium entering the bioreactor system, and glucose out is the concentration of glucose in the recirculating medium.

Determination of Antibody Concentration. The concentrations of secreted antibody in the bioreactor systems were determined by ELISA. Ninety-six-well microtiter plates (Costar Corp., Cambridge, MA) previously coated with drug­bovine serum albumin (BSA) conjugates were blocked to prevent nonspecific binding with 1% BSA in phosphate-buffered saline (PBS)/ azide. Serial dilutions of culture supernatant drawn from the bioreactor ECS were applied to the specific plate and a standard curve for assay quantitation was generated by incorporating known concentrations of purified antibody in parallel wells on the plate.

After incubation of the samples and standards, unbound antibody was removed by washing the plates in PBS/ Tween 20 (Sigma Biochemicals, St. Louis). The bound antibody was detected by the addition of alkaline phosphatase­labeled antimouse antibody (Zymed, San Francisco). The concentration of antibody was determined spectrophotometrically at 405 nm with a CR340 plate reader (SLT Corp., Salzburg, Austria), and data analysis was performed using the ELISA version 3.0 computer software (Meddata, Inc., New York City).

Criteria for Cell Growth and Antibody Production. The rates of antibody production and cell growth in the Clirans T220, BR2010, and the Acucell 1100 hollow-fiber cartridges were compared to those of the Unisyn BR3530 cartridge. For consistency, arbitrary criteria were used for between-run comparisons. One criterion was the number of days required to reach a target GUR level of 250 mg/hr; another was the number of days required to produce 1 g of antibody. These criteria defined the lag phase in comparisons of the same cell line grown in different cartridges or different cell lines or conditions for a given cartridge.

The effect of pore size on antibody production was determined by measurement of the amount of antibody produced (in milligrams) per liter of ICS feed medium consumed over a period of time (production efficiency). Because the runs varied in duration, day 40 was chosen arbitrarily to be representative for each run. The last criterion was the total antibody produced (in grams) by day 40 of the run.

Using the VA1 18D7 cell line, we compared the rate of antibody production and cell growth in the CP2000 bioreactor system using the Clirans T220 (Terumo Corp., Somerset, NJ), the BR3530, and the BR2010 (Unisyn Technologies, Inc.) bioreactor cartridges. The MOR 9B1 cell line was used to compare the Acusyst Jr. bioreactor system and Acucell 1100 cartridge with the rates of antibody production and cell growth in the CP2000 bioreactor system using the BR3530 cartridge (Table I).

Effect of Medium Feed Rate on Antibody Production. The effect of medium feed rate on antibody production was examined in both bioreactor systems. In the CP2000 bioreactor system that effect was studied using the BR3530 cartridge inoculated with the METH7-4B6 cell line.

Cell Removal. The effect of cell removal on antibody production and on bioreactor run length was determined in the CP2000 with the BR3530 cartridge and the PCP1.9D10 cell line. Mammalian cell cultures grown in hollow-fiber cartridges have a finite life. It has been speculated that accumulating cells decrease the productivity of the system by inhibiting the transport of nutrients across the hollow-fiber membrane, causing cell death. Whether periodic removal of excess cells from the bioreactor cartridge ECS would extend run length and improve faltering run parameters such as GUR and antibody production was investigated.

Harvest Methods. Antibody was harvested from the ECS regularly beginning within two weeks after the cartridges were inoculated. Two harvest modes were used, both using the same equipment. In a continuous mode of harvest, a peristaltic pump (Cole-Parmer, Chicago) withdrew medium from the ECS at a slow, constant rate (50 to 200 ml/day). At the same time an equal volume of fresh medium was introduced into the ECS by the same pump by using a separate piece of tubing run through the pump. Harvest bottles were changed periodically, and their contents centrifuged at 1000 * g for 10 minutes to remove cells and debris. The antibody-containing supernatant was then stored at ¾­20°C before testing. The second harvest mode was a batch method: A fixed volume of medium (50­250 ml) was pumped at once into the ECS, displacing an equal volume of antibody-rich medium from the chamber. This was repeated two to three times per week. The harvests were then processed in the same way as they were with the continuous mode.

Results

Cell Growth and Antibody Production. For the MOR 9B1 cell line and the BR3530 cartridge, it was determined in two separate runs that an average of 19 days were required to produce the first gram of antibody. This same cell line in two experiments with the Acucell 1100 cartridge required 31 days to produce 1 g of antibody (Table I).

Similar experiments were performed with the VA1 18D7 cell line. The Clirans T220 cartridge needed 33 days to produce 1 g of antibody, an amount comparable to the other 10,000-MW cartridge (BR2010), which required 32 days to produce the same amount. But the same VA1 18D7 cell line grown in the BR3530 (30,000-MW) bioreactor cartridge required only 17 days to produce 1 g of antibody. Likewise, the 30,000-MW bioreactor cartridge required 17 days to reach the target GUR of 250 mg/hr, whereas the two 10,000-MW cartridges required 31 and 30 days, respectively, to reach this level.

Similar results--13 and 26 days, respectively--were obtained with the MOR 9B1 cell line in the BR3530 and Acucell 1100 cartridges. The average time a 10,000-MW-pore-size cartridge required to reach the target level (1 g of antibody production) was 32 days. This contrasted with an average of 16 days for the 30,000-MW cartridge. Similarly, the average time needed by a 10,000-MW-pore-size cartridge to reach the target GUR of 250 mg/hr was 28 days versus 18 for the BR3530 cartridge.

Antibody Production Efficiency. We compared the ability of each of the hollow-fiber cartridges to produce antibody as a function of the amount of intracapillary medium consumed by the system during the experiment. These values were then expressed as milligrams of antibody produced per liter of medium consumed over the course of the entire run.

Using the MOR 9B1 cell line, we determined the optimum production efficiency to be 50 mg/L for the BR3530 cartridge and 24 mg/L for the Acucell 1100 cartridge. The VA1 18D7 cell line grown in the 10,000-MW cartridges had a production efficiency of 35 mg/L in the Clirans T220 cartridge and 30 mg/L with the BR2010. However, when the same cell line was grown in the BR3530 (30,000-MW) cartridge, the efficiency increased to 44 mg/L of medium consumed.

Comparison of antibody production in the BR3530 hollow-fiber cartridge to that in 72-hour static tissue culture (Table I) reveals a correlation between the amount of antibody a cell line produces in 72-hour culture and the quantity produced in the 30,000-MW cartridge.

Feed Rate. The effect of feed rate on antibody production was measured in two experiments using the METH7-4B6-8C3.2.1 cell line grown in the BR3530 cartridge. In the first experiment, the intracapillary feed rate was slowly increased from 0.5 L/day to a maximum of only 3.5 L/day (Figure 1). In the second experiment, the intracapillary feed rate was increased more rapidly from 1 L/day to a maximum of 5.0 L/day. The run with the lower feed rate produced more antibody in less time than the run with the higher feed rate (Figure 2). At day 40, the production efficiency of the low-feed-rate experiment was 118 mg/L and the production efficiency of the high-feed-rate scheme was 58 mg/L (Table I).

The effect of feed rate on antibody production was measured in two other experiments using the MOR 9B1 F11 P1.1.1 cell line grown in the Acucell 1100 cartridge using the Acusyst Jr. bioreactor system. When the feed rate began at 1 L/day and was rapidly accelerated to 8.4 L/day, the efficiency of the run was increased by only 9 mg/L (Figures 3 and 4). With a lower feed rate strategy beginning at 0.6 L/day and accelerating slowly to 3.3 L/day, the production efficiency almost tripled to 24 mg/L (Table I). Total antibody production from each run was approximately the same.

Cell Removal. When the PCP1.-9D10 cell line was grown in the CP2000 bioreactor system using a BR3530 cartridge, the cells reached confluence after only 25 days. Then antibody production began to drop precipitously (Figure 5). Beginning on day 30, cells were removed from the cartridge twice weekly by attaching syringes filled with IMDM (without FBS) to the extracapillary ports and vigorously pushing and pulling the plunger of the syringes. Large numbers of cells, cell debris, and a significant amount of antibody were obtained with this procedure.

Immediately after each of the nine cell-removal procedures performed during this one bioreactor experiment, antibody production increased significantly. The GUR also had a corresponding elevation. (Figure 5).

Batch versus Continuous Harvests. How did the kind of harvest mode affect total antibody production? The batch mode of harvesting used either syringes or a faster pump rate to remove antibody. When the VA1 18D7 cell line grown in the CP2000 system was switched from continuous to batch mode harvesting, the concentration of antibody increased from an average of 413 to 3387 µg/ml (Figure 6).

Discussion

The 10,000-MW-pore-size hollow- fiber cartridges from Unisyn Technologies, Inc. (BR2010), Terumo Corp. (Clirans T220), and Cellex, Inc. (Acucell 1100), are each capable of producing significant amounts of monoclonal antibody if used in the bioreactor systems described above. However, when compared with the Unisyn Technologies, Inc., BR3530 cartridge, the BR2010 and Clirans T220 cartridges were shown to be less efficient in terms of amount of antibody produced per liter of medium consumed. Likewise, the BR3530 was also found to be more efficient than the Cellex, Inc., Acucell 1100.

All three of the lower-molecular- weight-pore-size cartridges experienced a significant lag phase (approximately 1 month) for antibody production and cell growth, as measured by GUR. The lag phase with the BR3530 cartridge was consistently about one-half that of the 10,000-MW cartridges. The cost of medium consumed during the lag phase and of the labor used to run the bioreactor for an additional two weeks may offset the savings derived from using the lower-cost Clirans T220 cartridge in a bioreactor system.

The shorter lag period for the 30,000-MW pore size cartridge may come from the ability of medium components to more easily traverse the hollow-fiber membrane. Increased mass transport associated with a larger pore size hollow-fiber membrane is well documented.¹ The larger pore size fibers have higher ultrafiltration rates, which facilitate medium component exchange.⁶ The BR3530 cartridge provides a larger surface area for the growth of cells that may enhance cell growth and antibody production.

When low- and high-medium feed-rate strategies were compared, both the 10,000- and 30,000-MW cartridges showed that it was possible to overfeed cells in these bioreactor systems. High feed rates led to poorer cell growth and antibody production. By gradually increasing the feed rate throughout the run, more antibody was produced at a more efficient rate.

The high feed rates may dilute soluble factors that these cells require for growth. Some cell lines of the B cell lineage can stimulate autocrine growth.^7,8 The factors secreted from cells may be small enough to pass through the hollow fibers into the ICS. Some manufacturers of hollow-fiber bioreactor equipment propose high feed rate strategies to maximize antibody production,^4,6,9,10 but experience has shown that these strategies may not work well with all hybridoma cell lines.

The removal of cells from confluent hollow-fiber cartridges is useful to maintain the antibody production level of the bioreactor run. The cell-removal procedure rids the cartridge of large numbers of dead cells and cellular debris. This gives space for additional cell growth and clears the semipermeable fibers of cell components that may interfere with the efficient exchange of nutrients.

Harvesting the antibody in batch rather than in a continuous mode can increase the concentration of the antibody product if the batch harvests use small volumes of medium. This provides a reasonable alternative to continuous harvesting, especially for cell lines that do not secrete much antibody or if there is demand for a more highly concentrated product.

In the authors' experience, rapidly accelerating the medium feed rate in hollow-fiber bioreactor experiments may have an adverse effect on antibody production. It is recommended that once the GUR has begun to decrease, cells be periodically removed from the bioreactor cartridge to enhance antibody production. Inexpensive hollow-fiber cartridges, for kidney dialysis, can be a substitute for hollow-fiber cartridges used in antibody production with these instruments.

References

1. Evans TL, and Miller RA, "Large Scale Production of Murine Monoclonal Antibodies Using Hollow Fiber Bioreactors." Biotechniques, 6:762­767, 1985.

2. Lowrey D, Murphy S, and Goffe RA, "A Comparison of Monoclonal Antibody Production in Different Hollow Fiber Bioreactors," J Biotech, 1994, in press.

3. Lowrey DM, Meslovich K, Murphy S, et al., "Small Scale in Vitro Antibody Production Using a Disposable Hollow Fiber Device," presented to the 12th Annual Meeting of the European Society for Animal Cell Technology (ESACT), Wurzberg, Germany, May 1993.

4. Muragachi A, Nishimoto H, Kawamura N, et al., "Cell Derived BCGF Functions as an Autocrine Growth Factor in Normal and Transformed Lymphocytes," J Immunol, 137:179, 1986.

5. Knazek R, Gullino PM, Kohler PO, et al., "Cell Culture on Artificial Capillaries: An Approach to Tissue Culture Growth in Vitro," Science, 178: 65­67, 1972.

6. Gordon J, Ley SC, Melamed MD, et al., "Soluble Factor Requirements for the Autostimulatory Growth of B Lymphoblasts Immortalized by Epstein Barr Virus," J Exper Med, 159:154, 1984.

7. Hiefetz AH, Brantz JA, Wolfe RA, et al., "Monoclonal Antibody Production in Hollow Fiber Bioreactors Using Serum Free Medium," Biotechniques, 7:192­199, 1989.

8. Lowrey D, Murphy S, and Goffe RA, "The Effect of Intracapillary Media Feed Protocols on Hollow Fiber Cell Culture," Biotech Lett, 15:1025­1030, 1993.

9. Caple MV, Fletcher TR, Owens WJ, , et al., "A Dual Formulation Serum Free Media System," Biopharm, 5:52­60, 1992.

10. Hirschel M. and Keznoff S, "The Effect of Perfusion Rates on Hollow Fiber Bioreactors," presented to the Federation of Applied Scientists and Experimental Biologists, New Orleans, 1990.

Rudolph J. Czirbik, PhD, and Steven M. Rosen, PhD, are senior scientists; Diane M. Trunfio is an associate scientist; Ellyn W. Fischberg-Bender is research group manager; and Stuart M. Palmer, PhD, is director of research and development for the Therapeutic Drug Monitoring Business Unit of Roche Diagnostic Systems, Inc. (Somerville, NJ).

Molecular imprinting is becoming increasingly recognized as a technique for the ready preparation of polymeric materials containing recognition sites of predetermined specificity. The technique of molecular imprinting allows the formation of specific recognition and catalytic sites in macromolecules by the use of templates. Molecularly imprinted polymers (MIPs) have been used in an increasing number of applications. Several useful reviews of the molecular imprinting field have recently been published.^1­3 Of special interest to the developers of diagnostic assays is the potential use of MIPs in sample preparation as antibody or receptor binding site mimics in recognition and assay systems, and as recognition elements in biosensors.

History of Molecular Imprinting

The concept of molecular imprinting has its roots in Linus Pauling's early theory of the formation of antibodies. Pauling suggested that antibodies were formed when serum proteins assembled around template antigen molecules. The assembled antibodies were thought to have specificity-endowing binding pockets complementary in shape to the antigens. Furthermore, strong antibody-antigen binding energy would result from multiple noncovalent binding interactions including hydrogen bonds, ionic bonds, and van der Waals forces. This theory led to the hypothesis by Pauling and Campbell that artificial antibodies could be assembled using these basic principles.⁴ Although Pauling's theory of antibody formation was later disproved, several groups subsequently tried to apply it to synthetic systems. Pauling's student, Frank Dickey, attempted to form specific absorbents using silica.⁵ In the 1970s Wulff, at the University of Düsseldorf (Germany), formed covalent bonds between a monomer and the template molecule, followed by polymerization and template cleavage to yield a specific binding site. This method is limited by the synthetic necessities of first preparing a monomer-template molecule conjugate, and later chemically cleaving the template molecule from the polymer.⁶

Pauling's original concept was finally applied to the synthesis of artificial antibodies with the development of molecular imprinting by Klaus Mosbach's group at the University of Lund (Sweden).⁷ Mosbach's approach of preassembling a noncovalently associated monomer-template complex in solution prior to polymer formation was the breakthrough that enabled molecular imprinting to be used in a variety of applications. Mosbach's group continues to lead in new developments including studies on various polymer systems, classes of template molecules, aqueous imprinting systems, and novel physical formats, and has extended the potential usefulness of molecular imprinting. Imprinting has now reached a high level of sophistication, and patent coverage in the field is extensive.

MIPs are achieving commercialization through a collaboration between Mosbach and IGEN, Inc. (Gaithersburg, MD). IGEN's Separations Business Group is researching new applications, developing molecular-imprinted products, and offering molecular imprinting services for use by diagnostic and drug manufacturers. IGEN holds the rights to numerous patents and patent applications in molecular imprinting.

Making an Imprint

The concept of molecular imprinting is shown in Figure 1. To make the imprinted polymer, the molecule to be imprinted is dissolved in an organic solvent, such as chloroform, with a functional monomer, a cross-linking monomer, and a polymerization initiator. The functional monomer is chosen to have a chemical functional group that will interact and preassociate with the imprint molecule (Figure 1A­B). Ionic, hydrogen bonds, ¼-¼, hydrophobic, metal coordination, and covalent bond interactions are typical. Functional monomers that have been used for noncovalent interactions are shown in Table I. Following preassociation, polymerization occurs by UV irradiation or mild heating (Figure 1C). Once the solid polymer has formed, it is ground in a mortar and pestle and sieved to obtain a desired size, and the print molecule is extracted by incubation in a solvent capable of disrupting the specific interactions between the imprint molecule and the polymer (Figure 1D). This extraction step often involves including an acid or base in the solvent. What remains are rigid stable polymer particles that have pockets complementary in shape and electron density to the imprint molecule (Figure 1E). Shape complementarity results in high specificity while multiple interactions between the polymer and individual imprint molecules yield high affinity. (Figures and tables not yet available on-line.)

Several advantages make noncovalent imprinting the usual choice when a new method is being designed. For covalent imprinting, chemical modification of the print molecule with, for example, a vinyl group is necessary before forming the imprint. A wide range of monomers are available for noncovalent imprinting. In some cases the monomers can be combined to increase the strength of binding. In noncovalent interactions, association and dissociation of the imprint molecule to the imprinted polymer occur by the imprint molecule simply diffusing in and out of the complementary sites.

Recent developments in the formation of MIPs include imprinting of beaded polymers⁸ or silica resins³ and surface imprinting of membranes,⁹ which open additional methods of use, especially in diagnostic analysis.

Benefits and Current Limitations of Molecular Imprinting

Although molecular imprinting has some limitations, MIPs provide a combination of mechanical and chemical robustness with highly selective molecular recognition. They can be stored in the dry state at ambient temperatures for several years without loss of recognition capabilities, and are inexpensive, simple, and easily prepared. Generation of molecular imprints does not involve the use of laboratory animals or any material of biological origin other than the imprint molecule. MIPs are much more resistant to matrix effects than are biological antibodies. Highly lipophilic analytes can be assayed in organic extraction solvents using MIPs, which are formed in those solvents (Table II). Also, imprinted polymers can be made against print molecules that are too toxic for immunization in animals to raise antibodies. MIPs can also be made against molecules that are difficult to raise antibodies against--for example, short peptides. The preparation of antibodies against small organic recognition elements (haptens) requires hapten conjugation to a carrier protein before immunization. Preparing a molecular imprint avoids the need for derivatization of haptens.

Molecular imprints have been demonstrated against many classes of molecules. These include drugs,¹⁰ hormones,¹¹ pesticides,^2,12 proteins,¹³ amino acids,¹⁴ peptides,¹⁵ carbohydrates,¹⁶ coenzymes,¹⁷ nucleotides,¹⁸ nucleotide bases,¹⁹ steroids,²⁰ dyes,²¹ and metal ions.²² Some examples of interest for developers of in vitro diagnostics are shown in Table III.

Because noncovalent interactions are strongly dependent on the polarity of the solvent, the best imprints are made in organic solvents such as chloroform or toluene. Subsequent specific recognition of the imprint molecule by the imprint polymer is strongest under conditions that most closely resemble the cocktail used for polymer synthesis. If the MIP is transferred to aqueous solution, binding strength can be reduced significantly. Mosbach and others are currently working to improve aqueous imprints. As with polyclonal antibodies, the individual imprint cavities have varying degrees of selectivity. The distribution is usually around an average figure of high selectivity but with some sites of low selectivity. In chromatography, the low-selectivity sites do not influence the separation as long as the column is not overloaded with sample. In an assay format, specificity depends on the high- selectivity sites.

Examples of Molecular Imprints

MIPs have the potential for use in assay formats in a manner similar to the use of antibody-conjugated microspheres. For example, a radiolabeled ligand-binding assay, called the molecularly imprinted sorbent assay (MIA), was reported by Mosbach's group.¹⁰ Molecular imprints were made against two chemically unrelated drugs, theophylline and diazepam. Theophylline is a bronchodilating drug commonly used in the prevention and treatment of asthma. The assay accurately measured drug levels in human serum, with results comparable to those obtained using the established enzyme-multiplied immunoassay (EMIT). Specifically, the MIAs for theophylline and diazepam were linear over the ranges of 14­224 and 0.44­28 µm with detection limits of 3.6 and 0.2 µm, respectively, which are both satisfactory for therapeutic monitoring of the drugs. The specificity of the imprints was tested by determination of the cross-reactivity between major metabolites and of structurally related drugs. The imprints were shown to be highly selective, similar to those reported using commercial antibody­ based immunoassays (Table IV).

MIPs for leu-enkephalin and morphine have been made.¹¹ Because it is a closely related structure, codeine is a difficult cross-reactant for antimorphine antibodies. Cross-reactivity of the MIP with codeine is less than with most of the antimorphine antibodies (including monoclonal antibodies) reported to date. Although prepared in organic solvents, these MIPs were also efficient in aqueous solution with a performance sufficient for screening assays for drugs of abuse.

A molecularly imprinted membrane was recently developed that specifically bound theophylline over other purine bases.⁹ Such artificial affinity membranes should have applications for rapid analytical method development or for sample preparation methods. An additional benefit could be the combination of sample preparation and detection on the membrane.

MIPs that specifically bind particular metal ions (such as calcium and magnesium) have been produced using polymerizable metal-binding complexes. These have been used to make ion- selective electrodes.³ Imprints against pesticides and industrial pollutants have been reported by several groups.^2,12,16 Since many of these compounds are only soluble in organic solution, a solvent extraction step is necessary before the assay. Because MIPs can act as artificial antibodies in organic solution, they provide the possibility for developing formats similar to immunoassays, simplifying the current procedures.

Assays using molecular imprints with dissociation constants in the order of 10^­7­10^­8 M for cortisol and corticosterone were recently demonstrated.¹³ When the imprints were used in water systems, the interactions were disrupted because of the strong hydrogen bonding capacity of the water molecules, and the selectivity was severely reduced. Although many antibody-based steroid analytical methods have been developed for immediate use in aqueous environments such as plasma or urinary samples, several techniques use an extraction step for the steroids with organic solvents to avoid interferences with native binding proteins. In this perspective, anticorticosteroid MIPs may be an alternative for analytical applications, either by direct analysis or as a sample preparation step.

In some studies by Mosbach's group, target molecules have been chosen against which it is difficult or expensive to prepare natural antibodies. One example involves imprints against various macrolide antibiotics, such as erythromycin. Another example is imprints against immunosuppressants, such as cyclosporin.

Besides the many studies on resolution of the enantiomers of amino acids and sugars, molecular imprints against ß-blockers²⁴ and antiinflammatories²⁵ have also been reported.

Future Directions

The development of sample preparation methods and assays for small organic molecules using molecular imprints is expected to be completed in the near future. The preparation of MIPs for the isolation and detection of biomolecules is under development. Preparing an artificial antibody by molecular imprinting may be far more efficient in time and effort than is raising and producing a new antibody when none exists. Still needed are improvement of the binding strength of molecular imprints in aqueous solution and further development of surface imprinting and beaded imprints for higher-molecular-weight biopolymers such as proteins.

Molecular imprints have been included in several experimental biosensor designs.^1,3 Special advantages of MIPs in this application is their long-term stability at ambient conditions and in harsh environments. However, one obstacle to such biosensors is the current lack of a suitable interface between the MIP and the sensor element. Further developments in molecular-imprinted membranes may help in this area.

The creation of artificial antibodies through the combination of polymer chemistry and biochemistry is an ambitious goal, and one that promises considerable benefits. After many years of experimental development, MIPs are examples of progress toward this goal.

References

1. Mosbach K, and Ramstrom O, "The Emerging Technique of Molecular Imprinting and Its Future Impact on Biotechnology," Bio/Technol, 14:163­170, 1996.

2. Muldoon M, and Stanker L, "Plastic Antibodies: Molecularly-Imprinted Polymers," Chem Indust, pp 204­207, 1996.

3. Wulff G, "Molecular Imprinting in Cross-Linking Materials with the Aid of Molecular Templates--A Way Towards Artificial Antibodies," Angew. Chem Indust Ed Engl, 19:9­14, 1995.

4. Pauling L, and Campbell D, "The Manufacture of Antibodies in Vitro," J Exper Med, 76:211­220, 1942.

5. Dickey FH, "Specific Adsorption," J Phys Chem, 59:695­707, 1955.

6. Wulff G, Sarhan A, and Zabrocki K, "Enzyme-Analogue Built Polymers and Their Use for the Resolution of Racemates," Tetrahedron Lett, 44:4329­4332, 1973.

7. Andersson L, Sellergren G, and Mosbach K, "Imprinting of Amino Acid Derivatives in Macroporous Polymers," Tetrahedron Lett, 25:5211­5214, 1984.

8. Mayes A, and Mosbach K, "Molecularly Imprinted Beads: Suspension Polymerization Using a Liquid Perfluorocarbon as the Dispersing Phase," J Molec Recogn, in press.

9. Kobayashi T, Wang HY, and Fujii N, "Molecular Imprinting of Theophylline in Acrylonitrile-Acrylic Acid Copolymer Membrane," Chem Lett, 10:927­ 928, 1995.

10. Vlatakis G, Andersson L, Müller R, et al., "Drug Assay Using Antibody Mimics Made by Molecular Imprinting," Nature, 361:645­647, 1993.

11. Andersson L, Müller R, Vlatakis G, et al., "Mimics of the Binding Sites of Opioid Receptors Obtained by Molecular Imprinting of Enkephalin and Morphine," Proc Natl Acad Sci, 92:4788­4792, 1995.

12. Siemann M, Andersson L, and Mosbach K, "Selective Recognition of the Herbicide Atrazine by Noncovalent Molecularly Imprinted Polymers," J Ag Food Chem, 44: 141­145, 1996.

13. Kempe M, Glad M, and Mosbach K, "An Approach towards Surface Imprinting Using the Enzyme Ribonuclease A," J Molec Recogn, 8:35­39, 1995.

14. Kriz D, Ramström O, Svensson A, et al., "Introducing Biomimetic Sensors Based on Molecularly Imprinted Polymers as Recognition Elements," Anal Chem, 67:2142­2144, 1995.

15. Ramström O, Nicholls IA, and Mosbach K, "Synthetic Peptide Receptor Mimics: Highly Stereoselective Recognition in Nonconvalent Molecularly Imprinted Polymers, Tetrahedron: Assymetry, 5(4):649­656, 1994.

16. Mayes AG, Andersson LI, and Mosbach K, "Sugar Binding Polymers Showing High Anomeric and Epimeric Discrimination by Noncovalent Molecular Imprinting," Anal Biochem, 222:483­488, 1994.

17. Andersson LI, and Mosbach K, "Molecular Imprinting of the Coenzymer Substrate Analogue N-Pyridoxyl-L-Phenylalaninanalide," Makromolec Chem, Rapid Commun, 10:491­495, 1989.

18. Norrlöw O, Mänsson MO, and Mosbach K, "Improved Chromatography: Prearranged Distances between Boronate Groups by the Molecular Imprinting Approach," J Chromatog, 396:374­377, 1987.

19. Shea KJ, Spivak DA, and Sellergren B, "Polymer Complements to Nucleotide Bases: Selective Binding of Adenine Derivatives to Imprinted Polymers," J Am Chem Soc, 115:3368­3369, 1993.

20. Ramström O, Ye L, and Mosbach K, "Artificial Antibodies to Corticosteroids Prepared by Molecular Imprinting," Chem Biol, in press.

21. Norrlöw O, Glad M, and Mosbach K, "Acrylic Polymer Preparations Containing Recognition Sites Obtained by Imprinting with Substrates," J Chromatog, 299:29­ 41, 1984.

22. Rosatzin T, Andersson LI, and Mosbach K, "Preparation of Ca²⁺ Selective Sorbents by Molecular Imprinting Using Polymerisable Ionophores," J Chem Soc, Perkin Trans, 2:1261­1265, 1990.

23. Andersson L, Nicholls I, and Mosbach K, "Antibody Mimics Obtained by Noncovalent Molecular Imprinting," in Immunoanalysis of Agrochemicals, Nelson J, Karu A, and Wong R (eds), Washington, DC, American Chemical Society, pp 89­96, 1995.

24. Fisher L, Müller R, Ekberg B, et al., "Direct Enantioseparation of ß-Adrenergic Blockers Using a Chiral Stationary Phase Prepared by Molecular Imprinting," J Am Chem Soc, 113:9358­9360, 1991.

25. Kempe M, and Mosbach K, "Direct Resolution of Naproxen on a Non-Covalently Molecularly Imprinted Chiral Stationary Phase," J Chromatog A, 664:276­279, 1994.

Gary P. Henricksen is vice president, Separations Business Group, and Mark T. Martin, PhD, is director, Discovery Research Group, IGEN, Inc. (Gaithersburg, MD).

A company that can produce instruments faster than its competitors will not just have the first product on the market. It will also have the best product because it can receive customer feedback and produce second- or even third-generation products in the time it takes competitors to get their first-generation product out.

This article describes a system developed by LJL BioSystems, Inc. (Sunnyvale, CA), for completing instrument development projects in half the usual time and with a reliability rate well above the industry average. For example, the automated chemiluminescence reader for DNA probe assays that LJL developed with Chiron Corp. (Emeryville, CA) took just 13 months from start to full production and worldwide shipment. Similar projects typically require 24 to 26 months at other facilities. The reader's mean time between failures is well above the industry average of 10,000 to 15,000 hours for this type of instrument. In fact, it has been so reliable that, unfortunately for LJL, Chiron has had no incentive to buy a maintenance contract at the end of the instrument's warranty period.

Most of LJL's secrets for rapid but thorough instrument development are not so much secrets as basic processes that could be applied by other instrument manufacturers as well. This article presents a few of these basic processes as LJL practices them.

Program Management

One of the keys to rapid instrument development is having a program management system that is both clearly defined and strictly enforced. Program management should be thought of as a discipline that is every bit as important as the various engineering disciplines that are required for instrument development.

The development process should be comprehensively documented, and the document should be provided to all employees. In addition, LJL presents a customized version of this document to every manufacturing partner (the term LJL prefers to use instead of customer). It provides a detailed outline of all the activities and milestones involved in the development process, broken down into four clearly defined phases: product definition, concept and feasibility, prototype development, and manufacturing transfer and production.

During phase one, product definition, the instrument concept is defined in terms of instrument attributes and engineering specifications. This instrument concept is then used to evaluate and define key instrument functions, human factors issues, industrial design issues, and an estimated development schedule and budget.

A clear product definition is vital to delivering a high- quality instrument. Specifications must be agreed upon and approved by all functional groups involved in the project (including marketing, development engineers and chemists, and manufacturing) before proceeding to phase two.

During phase two, concept and feasibility, LJL develops detailed concepts for the package industrial design, mechanical subsystems, user interface, electrical subsystems, and firmware/ software components. It builds an instrument that is functionally similar to the final production instrument but not necessarily identical in form. If the form is substantially different from that of the planned production instrument, a nonfunctional model of the packaging may also be created during this phase to obtain user feedback on human factors and the industrial design.

In phase three, prototype development, functional prototypes are built that look and perform like the final production instrument and may be used for preclinical or beta trials. The engineering efforts during this phase focus on achievement of the reliability, manufacturability, and cost targets of the instrument.

The goal of phase four is to transfer a quality product into production. Pilot production instruments are built during this phase, the manufacturing transfer process is completed, and final production instruments are manufactured. The final production instruments completed during this phase are built by manufacturing rather than by the engineering team. All engineering hardware, firmware, and software designs, and all fabrication, assembly, and test documentation are upgraded at this time to meet production and good manufacturing practices (GMP) requirements.

By having a defined program management process that is broken down into clear phases, we not only know exactly what the project involves and have a schedule that we follow meticulously, but we can also maintain momentum by completing the project in short bursts rather than in one long haul. Each phase of the project has a predefined and strictly enforced start and stop time, so team members know they have to accomplish a specific task in a set time. This allows the team to focus on an immediate goal rather than try to do everything at once. By meeting a series of interim goals, team members not only keep the project moving along rapidly, they also have the reward of a continuous string of short-term successes to keep them motivated. LJL offers additional motivation in the form of a bonus for on-time completion of all tasks. If any interim goal is not met on time, all team members forfeit a portion of their bonus. This provides an incentive for team members to cooperate, be creative, work hard--to do whatever it takes to get the task done on time.

Concurrent Product Development

Beginning with phase one, our project teams include staff from all functional groups: marketing, research and development, product design, manufacturing engineering, and quality control. These permanent project teams include experts in each critical engineering technology and are led by a project manager with a technical background and project management experience. On average, teams contain between 5 and 10 members. We also consider our customer to be a part of the team. All functional groups have input at all stages of the project, providing concurrent product development. This avoids the possibility that marketing will decide that a product needs significant modification during research and development or, even more costly and time-consuming, during scale-up for manufacturing. With this method, instead of passing the project from team to team, one integrated team moves from milestone to milestone.

The entire team is required to attend all design review meetings, which are held weekly during phase one and less frequently during subsequent phases, with a minimum of three to four meetings per phase. Meetings may take all day, but they save weeks over the course of the project because issues are resolved and specifications updated and agreed on before the meeting ends. We bring a computer attached to an overhead projector and printer into all design review meetings. Team members can clearly see what is being discussed and, when they leave, always take with them a printout of the agreed-upon design specifications and action items.

An additional value to having all team members present at all meetings is that it enforces communication. People who tend to be task rather than people oriented sometimes need this inducement to discuss their individual projects in the context of the completed instrument. One of the biggest challenges in instrument development is to integrate the different engineering disciplines and systems with the chemistry requirements. Through the scheduling of frequent design review meetings, communication and preintegration are automatically built into the project. This ensures that the various elements that have been designed separately (e.g., packaging, mechanical, electrical, optical, software) will fit together more elegantly and integrate more easily with the chemistry.

The Right People

We have found that smaller teams work best because they generally require less management, which means less red tape, which means faster movement of the project. In addition, small permanent teams have a greater sense of ownership of the project, and all members share their expertise to solve whatever problems arise.

Basic to having a small team is hiring the right people. We prefer to hire high-end, experienced people and have fewer of them. Although well-seasoned people are expensive, they can make the initial product definition much easier and contribute significantly to the thought process that avoids last-minute issues. An additional value is that, if last-minute issues do arise, well-seasoned employees are much better equipped to resolve them with minimum risk because they know from experience what the risk parameters are. Thus, the price for these employees may be higher, but the overall costs to the company are actually lower. We prefer employees who have been through some product development scrapes, and we always ask when interviewing prospective employees whether they have been through a full product launch, all the way to market, dealing with the various issues that turn up during each separate phase of the process.

The right people are also versatile. An engineer who understands both hardware and software is invaluable because there is never any need for an outside arbitrator to help decide whether the instrument in development has a software or a hardware problem. If the same person is responsible for solving either problem, problem ownership is clear. Also, having a mechanical engineer with experience in design avoids the common situation of having the designer say, "It should look like this," and the engineer replying, "Well, it won't fit in that box." With these two disciplines combined, form will always follow function.

Inside/Outside

The right people do not always have to be in-house (see sidebar, "When and how to outsource"). Having a network of consultants can be far more cost-effective than hiring specialists who may not be needed for every project. Here it is critical to locate and maintain contact with technology centers that are a rich source of technical expertise. For example, LJL is fortunate to be located in the heart of Silicon Valley, one of the premier technology centers in the United States. We benefit greatly from regular contact with the many vital, innovative technology leaders in the area.

In addition, not every part needs to be or even should be manufactured in house. A good balance must be maintained between what is made in house and what is bought from a network of specialized and cost-effective vendors who supply both custom and commercial parts.

Deciding what should be made and what should be bought can be difficult. During phase one, we involve all of our functional groups in defining which components are high-risk, low-risk, high-return, and low-return, and then give each component a make or buy designation. We consider our own in-house capabilities when making these decisions. If a vendor can provide a component more quickly, less expensively, and with better quality than we can make it in house (all three criteria must be met), we buy it rather than make it.

It is critical to decide at the outset of the project who will make what, in order to know how much each component will cost and what the lead times will be. Plus, by making these decisions in advance, we can include our vendors in the project team. They are often involved in the design of the instrument and always in its transfer to manufacturing. This way, manufacturing knows whom to go to with questions, and the vendor knows exactly what manufacturing requires. As a part of the team, vendors see the instrument as it is being created and can better integrate their efforts with those of the in-house staff.

Using vendors for standard components also leaves in-house staff free to focus on elements of the instrument design that can make a difference in the marketplace. Tough projects are good because they provide a motivating challenge; out of them come new solutions and technologies. For example, given the challenge of designing a low-cost, highly reliable controller for one instrument, we used PIC (programmable integrated circuit) technology to create an electronic brain for this instrument that we can modify and transplant into any instrument we design in the future.

State-of-the-Art Equipment

To develop technological innovations rapidly, in-house staff needs the right equipment. Powerful advanced mechanical and electrical computer-aided design (CAD) packages allow sophisticated computer modeling that reduces the number of design iterations and the duration of the engineering programs. For example, we have nine CAD workstations for mechanical and package design: five Sun systems and four Silicon Graphics systems. All of these systems currently use the Pro-E software package from Parametric Technology Corp. (Waltham, MA), all are connected to the same network, and so all are able to share the same files. File sharing allows one engineer to access another's design, ensuring better integration of separate components.

These CAD systems also offer paperless prototyping. We have a direct link via modem to several vendors that produce models or even finished parts directly from our CAD files within 48 hours. All electronics design is done on CAD systems as well. We have an in-house system, Quick Circuit, linked to our CAD network that makes prototype circuit boards right from our CAD files, so we can go directly from the design of mechanical or electrical systems to prototypes without the need for any intermediate drawings or schematics (Figure 1). This allows production of prototypes in a matter of hours or days, not weeks or months. Further, building tooling directly from CAD databases is a significant money saver.

Modular Design Approach

Perhaps most important to LJL's rapid development capability is that when we develop a subsystem (such as a robotic arm, or a printed circuit board, or an optics system, or a piece of software), we think ahead to other applications for it in other markets. By doing this, we are developing a subsystem not only for the current customer but for the next one as well. LJL makes it a condition that we, rather than the customer, retain the manufacturing rights to these subsystems, which we can then modify for use in a wide range of instruments for various markets. This modular design approach is of value to any instrument manufacturer because it not only accelerates the design process but also improves reliability by allowing subsystems to be pretested independently of the completed instrument. Figure 2 illustrates how technology developed by LJL for one product was adapted for use in several others. Having a base of these core subsystems or technologies gives us an edge in getting an instrument to market quickly.

Conclusion

There are no real secrets to rapid instrument development. All it takes is small, focused, sharing, highly competent, well-equipped technical teams and a clear, proven process. Having a clearly defined process speeds development by reducing the amount of change: change of mind, change of personnel, change of project. Although a company must always accommodate some level of change, the rate of change should never exceed the rate of progress. Controlling change by having and enforcing a well-defined process reduces the risk of running out of time and ensures being able do all that was planned and more.

Lev J. Leytes is president and chief executive officer of LJL BioSystems, Inc. (Sunnyvale, CA). Amer El-Hage is director of program management for LJL. Ann C. Petersen is a technical writer specializing in the diagnostics industry.

When and how to outsource

Sometimes the quickest way to get something done is to give it to an outside vendor. The engineering teams of OEM (original equipment manufacturer) companies are experienced with a wide variety of projects, are usually in tune with the latest engineering technologies, and get a great deal of practice in the rapid development process. A vendor that has kept abreast of the new technologies and business processes can often provide cost-effective and technologically advanced products quite rapidly.

In many cases, the decision to outsource can be based solely on the amount of practice your company has or wants to have in a particular field. If you have a great deal of experience producing certain components, they should be manufactured in house. However, if expertise in a particular area does not reside in house and is not needed on a continuing basis, the project should be outsourced.

When choosing a vendor, size should not be a ruling factor. The largest vendor candidate may not automatically guarantee dependable service, because your project may be less important to a larger company with multiple projects than to a smaller one focused on a few important projects. Concentrate on the quality of the company's people and its track record with projects similar to yours.

Keep in mind, too, that the lowest bidder is not a bargain if it can't deliver to your quality standards and you have to change vendors. Check the prospective vendor's record of on-time deliveries with other customers. If a vendor misses shipments to you, your revenues from new instrument installations will be delayed and may be lost. Also, look beyond the quoted prices and evaluate the total costs over the lifetime of the program. Consider the costs of maintenance, repair, and downtime. Make sure that the quoted price includes adequate testing of the instrument during the development phases. Correcting problems that arise in the field because they were not tested for and corrected during development can be extremely costly in both dollars and reputation. Check for built-in loops that allow the vendor to come back for more money.

Choose vendors that can demonstrate a commitment to both price stability and quality. Look for those that will link their success with yours rather than merely play back your requirements to get your order. You do not want an outside contractor to come back and tell you that the program has failed because they took direction from you.

Consider offering the supplier exclusive rights to specific technologies it develops for your project but may want to use in future projects for other customers. This can motivate a supplier to invest its full creativity into the project and also allow you to get better pricing.

By outsourcing, you can dedicate precious internal assets to those activities that generate the best return and the highest shareholder value, while distributing functions that are peripheral to or do not match your company's core competence to other companies that specialize in those functions.

We learned at an early age that the first-place finisher gets the gold medal. In the IVD industry, the winners are those that are first to market. These competitors are rewarded not only with profits but also with increased chances of long-term survival. Miles Laboratories with its dipstick technology, Ortho Diagnostics with its HCV ELISA technology, and Perkin-Elmer with its PCR technology are a few examples of companies that have achieved market dominance by being first.

The time required to develop a new product affects its profitability more than do its initial costs. A McKinsey study reports that, on average, companies lose 33% of after-tax profits when they ship products six months late, compared with 3.5% when they overspend 50% on product development.

Many companies try to streamline the new product introduction process by improving their business organizations. Toward this end, they often establish cross-functional development teams and specific company guidelines on departmental interrelationships. Collaboration and communication are effective ways to improve the development process, but only if the product or project has been chosen carefully. A company may spend a lot of time developing the perfect organization and never launch a new product in time to capture the market.

Even the best-run company can improve its product introduction time by choosing a new technology or product carefully. Indeed, this is the most important factor in becoming the first to market.

The industry in recent years has had to weigh the appeal of new technologies against the inevitability of regulatory roadblocks. Assay formats with no historical precedent require more than the average product development time, for both in-house quality assurance and regulatory licensure. Assays that are similar to previously marketed products, however, have limited market potential if they offer no advantage over those existing products. In the blood glucose monitoring field, assay formats were predictable for many years. Newcomers in the marketplace could expect a 1 to 2% market share and a predictable lackluster product lifespan. Johnson & Johnson's introduction of LifeScan One Touch Glucose Monitor and Test Strips changed the rules and found high customer acceptance. The product development process unifying reagents and handheld instrumentation required longer-than-average product development time but in the long term is increasing the company's market share and market stability.

Having a clear mission is one key to timely product launch. The mission should be matched to a specific need of the intended customer. "Mission creep"--brought on by new marketing data, management redirection, or attractive technological advances--can bog down new product development. The development team must stay focused on the goal while still being able to react to the dynamic product development en-vironment. Project leaders may find themselves in the role of "coaching" when strategies must change to meet the desired goal.

After choosing the product or technology, the investigation or feasibility stage comes next. Within a predetermined time, the company should decide whether to commit to developing the product. During this phase, the company should outline the product's desirable features, demonstrate its feasibility, and estimate its cost and price. Marketing and R&D should collaborate to determine what features the customers want and how to include them in the product. New product ideas may be generated at any level within or even outside an organization, but the allocation of funds should be approved by the highest levels of management. Formal presentations and documentation by the designated product development groups are recommended. The technology should be understood and agreed upon before any money is allocated. Early allocations for research feasibilities are usually made after idea generation and before a formal contract or proposal.

The development phase begins only after a thorough investigation of feasibility. The developed product must be made correctly the first time to ensure a timely introduction and long life cycle. At this point, the technology transfer and manufacturing personnel carry the ball. The development phase should include all the stages of research and development including scale-up and product stability and clinical studies. The main focus should be to determine how to make the product at the desired price. Strategic project staffing at this time is critical. Management must provide adequate staff to meet deadlines and keep development costs down.

Processes and test methods that become carved in stone at this stage must be easily scaled up for manufacturing and must be accepted by FDA. Reengineering of licensed products is not cost-effective in the 1990s IVD market. Any process changes made to an existing product require that the manufacturer reevaluate validated standard operating procedures, specifications, stabilities, or clinical studies. In "Changes to Be Reported for Product and Establishment License Applications," published on April 6, 1995, in the Federal Register (60 FR:17535­17538), FDA provides guidance to manufacturers who are attempting to correct problems or address process changes with minimal expense and without unnecessary waste. Although this document has sorted process changes into distinct categories requiring different levels of compliance, the marketplace and the time factor control the actuality of manufacturing process changes. As technology has advanced and new antigens have been discovered, for example, infectious disease ELISA manufacturers have found it easier to license new versions of their assays rather than try to improve their earlier formats.

The manufacturing phase is the last step where new product introduction can proceed quickly or become snarled. Successful manufacturing is the result of training and collaboration between technical personnel and manufacturing personnel.

New technologies may be moved into the market faster if the pro-cesses used to make the reagents or the instrumentation are tried and true. Technology transfer professionals must determine not only the best technical processes and test methods to use in manufacturing a product but also whether FDA has previously cleared products made with those processes. Raw materials, chromatography resins, viral clearance techniques, and cleaning procedures are a few examples of reagents and processes regarding which novelty should be avoided if possible. All FDA-licensed intravenous antibodies marketed today are purified by ion-exchange chromatography. There are other techniques for purifying antibodies, but they require extra time and money. Unless a company really wishes to be a biotech pioneer, it should follow accepted practice.

At this stage, validation of the manufacturing process should occur. A successful validation should be a sure thing, since so much of the investigational work was performed at the development phase. First-place finishers in the marketplace know that validation is not experimentation, and they are confident of success by the time the product reaches manufacturing.

Long-term survival of an IVD company depends on its ability to beat competitors to market with new products. Improved profits, prestige, legal position, and regulatory acceptance are the prizes awarded to the winner in the IVD marketplace.

Alfred W. Schweikert, PhD, is a member of IVD Technology's editorial advisory board and manager of process development at Ortho Diagnostics Systems, Inc., in Raritan, NJ.

Conventional medical wisdom holds that finding cancer early is the next best thing to a cure. But the U.S. government has decreed that when it comes to bladder cancer, this doesn't apply. The Guide to Clinical Preventive Service, 2nd Edition issued in 1995 by the U.S. Preventive Services Task Force specifically recommends against routine testing for early detection of bladder cancer. The guidelines are available from the US Government Printing office.

Where does that leave two new bladder-cancer diagnostic tests? It keeps them confined to checks for recurrence of disease, a realm only a fraction as effective as more widespread screening.

As the 10-member task force was formulating its new guide, two diagnostics were getting ready to hit the market: the Bard BTA bladder test (Bard Diagnostic Sciences, Redmond, WA), currently available and Matritech's nuclear-matrix protein test (Matritech, Inc., Newton, MA).

But physicians on the task force found little evidence to support urine testing as a routine screen. The reason: like the prostate-specific antigen test, the most reliable marker for prostate cancer, bladder-cancer diagnostic screening doesn't have a strong track record of documented benefit. To be included in the recommendations, such tests need proof of improved morbidity and mortality.

Bard launched its product this spring after the final go-ahead from FDA. The dipstick like test is the "world's first three-minute bladder-cancer test" and the assay "requires no special training to carry out and interpret," according to company descriptions. The test relies on detection of proteins associated with developing cancer; a yellow color-change indicates positive green indicates negative.

Matritech, with approval pending, also has a urine test, also based on a protein marker. It includes antibodies that recognize nuclear-matrix proteins (NMPs). The advantage, say scientists who have worked with the assay, is that NMPs are specific to cancer cells -- they don't seem to appear without the presence of the carcinoma. Scientific literature shows that bladder cancer has increased by about a third over the past decade. The reasons probably are multifactorial, explains Donald Lamm, MD, professor and chair of urology at the Robert C. Byrd Health Sciences Center of West Virginia University in Morgantown.

One reason for the increase is the influence of environmental factors, such as smoking and exposure to chemicals, particularly in industries as varied as truck driving and hair dressing. But another reason is that more than a quarter of patients with bladder cancer -- which generally shows up first as blood in urine -- may yield a negative cytologic result during cytoscopy, the traditional confirmatory measure.

Clearly a better screening tool is needed, Lamm says. Although neither the Matritech nor Bard test is to be used for initial detection, that is precisely what is needed to find cancer early, when intervention is most likely to pay off, he adds.

Lamm and his colleagues are advocating a selective screening process, in which those at risk are carefully followed. A family history of disease, exposure to workplace chemicals, and smoking all fit into that category, Lamm says.

But, to treat this population, "we really need a precise way of detecting the cancer," he stresses.

An even newer, experimental immunoassay, Aura-Tek FDP by Perimmune, Inc. (Rockville, MD), is showing heightened sensitivity in detecting the cancer, when compared to both hemoglobin-based dipsticks and cytologic evaluations, Lamm says. At the University of Southern California School of Medicine, David Esrig, MD, Richard Cote, MD, and Peter Jones, PhD -- along with others -- have developed a histochemical stain that shows promise in predicting a recurrence, which could also aid in narrowing the patient population in need of close monitoring.

A mutation in the p53 gene, which leads to an increase in a related protein, occurs in significantly more patients who experience a recurrence. "These people are twice as likely to recur," Jones, who is director of the university's Norris Comprehensive Cancer Center.

In the future, genetic testing may be used to further divide that pool of higher-risk people, predicts Lance Willsey, MD, a urologist at Massachusetts General Hospital in Boston. Then one of these more-sensitive assays may be applied to actual screening.

This kind of "second-layer selection" is likely to be the first wave of a true clinical application of gene testing, a fit that may go hand-in-glove with in vitro diagnostics down the road. 

Microporous membranes have proven to be highly successful surfaces for the development of simple, rapid immunoassay delivery systems. Membranes made of nitrocellulose, currently the material of choice, can be produced with a wide range of pore sizes and treated with surfactant to optimize their performance. These membranes help create sensitive, stable, and reproducible assays. Since the first demonstration in 1979 that proteins could be transferred to microporous nitrocellulose membranes and detected using antibodies,¹ development of rapid immunoassays using these high-surface-area materials has proliferated.^2­7 Initially much investigation centered on understanding the interactions between proteins and polymers and the requirements for blocking nonspecific interactions on the membrane, and on developing a series of detection methodologies and strategies. This work has led to a variety of immunoassay delivery systems for detecting a large menu of analytes. Membranes are used in both sandwich and competitive immunoassays to detect both small and large molecules.^8,9

Key membrane characteristics include polymer type, porosity, surfactant content, and strength. Because the performance and capabilities of membrane-based immunoassay systems are influenced by specific properties of the membranes, a thorough characterization and understanding of these factors is essential.

Types of Membranes

The choice of a membrane for an immunoassay delivery system depends largely on three properties: protein-binding capacity, porosity, and strength. The ability of the membrane to immobilize macromolecules, in particular proteins, is paramount, since they are the solid phase used in the assay. The porosity of the membrane is important because reactants must flow through the matrix so that it can separate bound from free components. The strength of the membrane is important for the manufacture and eventual use of a device.

Polymers tested for their ability to bind proteins include cellulosics (nitrocellulose, cellulose acetate, regenerated cellulose), nylon, and polyvinylidene fluoride (PVDF), as shown in Figure 1. The binding capacities of cellulose acetate and regenerated cellulose are insufficient for most immunoassays, but a range of higher capacities are available with nitrocellulose, nylon, and PVDF membranes. (Figures not yet available on-line.)

Most current membrane-based immunoassays use either nylon or nitrocellulose. Both bind proteins noncovalently, although the mechanism of interaction is probably different. A polyamide, nylon binds proteins via electrostatic and charge interactions, while the interaction of nitrocellulose with proteins appears to be primarily hydrophobic in nature.¹⁰ Both polymers bind most proteins--in particular, antibodies--with sufficient density and avidity to support an immunoassay. They generally do not require covalent attachment of the immobilized reactant.

Although nylon was perhaps the first polymer membrane employed in a commercial immunoassay delivery system,¹¹ nitrocellulose has proven to be more versatile and is currently the membrane of choice. While the protein-binding capacity of nitrocellulose is lower than that of nylon, it is sufficient to support sensitive immunoassays. This allows nitrocellulose-based assays to avoid the high level of nonspecific interactions that can contribute to background in nylon-based assays. Nitrocellulose can be manufactured in a wide range of porosities, from 0.05 to about 12 µm, providing in turn a range of binding capacities and wicking or capillary-rise rates to aid in the flow of reactants through the assay.

Flow-Through and Lateral-Flow Immunoassays

In choosing a particular nitrocellulose membrane for an immunoassay system, the role the membrane plays in the assay must be considered. Membrane-based assays fall into two broad categories. Historically, the first such assays were flow-through filtration assays (Figure 2). In making this type of assay, one immunoreactant is immobilized to a defined area on a membrane surface. This membrane is then overlaid on an absorbent layer that acts as a reservoir to pump sample volume through the device. Following immobilization, the remainder of the protein-binding sites on the membrane are blocked to minimize nonspecific interactions. When the assay is used, a sample containing analyte is added to the membrane and filters through the matrix, allowing the analyte to bind to the immobilized antibody. In a second step, a tagged secondary antibody (an enzyme conjugate, an antibody coupled to a colored latex particle, or an antibody incorporated into a colored colloid) is added that reacts with captured analyte to complete the sandwich. Alternatively, the secondary antibody can be mixed with the sample and added in a single step. If analyte is present, a colored spot develops on the surface of the membrane. This type of assay has the advantage of simplicity both in manufacturing and in use.

The choice of a membrane for such an assay depends on several factors, such as the sensitivity of the assay, the time required to run the assay, and the type of readout employed in the assay. The pore size of the membrane is the key factor. Nitrocellulose membranes with pore sizes from 0.2 to 8 µm have been used in flow-through assays. The larger the pore size, the lower the surface area and, therefore, the lower the protein-binding capacity. But because membranes with larger pores support more rapid filtration rates, they can yield a faster assay. Conversely, a smaller-pore membrane can create a more sensitive assay.

The size of the tagged antibody complex may also dictate the pore size, which must be large enough to permit unreacted antibody to pass through. Assays that use an antibody coupled to a colored latex bead require a larger pore size.

More recently, lateral-flow immunoassay systems have been developed to allow for single-step assays that require only the addition of a sample (Figure 3). In these assays, the sample is added to one end of the device and flows by capillary action through the interstitial space of the materials in the device. While continuing along this flow path, the sample contacts dried reagents, usually tagged secondary antibodies, which then migrate with the analyte to a capture zone of immobilized antibody on the membrane. Unreacted tagged antibody continues to flow past this capture zone, normally to an end-of-assay indicator. Generally, absorbent material at the distal end of these devices helps draw the sample through the device.

Membranes in these types of systems must offer different properties than those best used in flow-through devices. Capillary rate becomes a more important property, since it not only dictates the total time required to run the assay but also determines the residence or reaction time at any given zone.

To accommodate these requirements, nitrocellulose membranes with larger pore sizes, up to 12 µm, have been developed. As the pore size increases, the speed of wicking or capillary rise through the matrices increases (Figure 4). However, the relationship between wicking rates and distance is not linear (Figure 5); thus reaction time at any point along a membrane may vary. Even though there is a considerable decline in the binding capacity of a large-pore-size nitrocellulose membrane, it retains sufficient capacity to support all immunoassays (Figure 6).

Pore size is not the only factor that can affect the capillary rate. Most nitrocellulose membranes contain a surfactant to aid in wetting and act as an antistatic agent. The amount and type of surfactant included in a membrane can be altered to impart different capillary rates to membranes with equivalent pore sizes (Figure 7).

Manufacturing Considerations

Tensile Strength of Nitrocellulose Membranes. A property of nitrocellulose that challenges its use in immunoassay formats is its low tensile strength. Three different solutions to providing sufficient strength have been used.

Some nitrocellulose membranes can be cast around a supporting, presumably noninteractive, material such as a nonwoven polyester. The most common approach, however, has been to laminate the membrane to a plastic backing with an adhesive. A third approach, intended to eliminate the potentially detrimental interactions between membrane and adhesive, has been to cast the nitrocellulose membrane directly on a plastic support. This eliminates the need for a lamination step in manufacturing as well as the problems that can be encountered with the use of adhesive systems.

Storage of Nitrocellulose Membranes. In order for nitrocellulose membranes to preserve their protein- interactive properties, they must be properly stored, to remain hydrated. Dried membranes will not rehydrate easily, and will become more brittle and perhaps lose binding capacity. Prior to device manufacturing, nitrocellulose membranes should be stored in an environment of controlled humidity (40­60%) and controlled temperature (18°­20°C).

Application of Reagents to a Nitrocellulose Membrane. In order to establish a reproducible, sensitive assay on a membrane, the various immunoreactants must be accurately applied to the membrane. Generally, antibodies or antigens are added as a line or a symbol covering a defined surface area. Most antibody lines have a width of 1­2 mm. The volume required to produce such a line is in the submicroliter range, with application rates between 0.5 and 1.0 µl/cm of membrane.

Instrumentation to attain such rates must be reproducible and accurate. One application method is airbrushing, which involves moving the membrane past an airbrush head at a controllable rate. A second and perhaps preferred method is to use a positive-displacement micropipetting system. Manufacturers that provide such systems include Ivek Corp. (North Springfield, VT), BioDot (Irvine, CA), and Ismeca USA (Carlsbad, CA).

Additional Treatments of the Membrane. Following application of the immunoreactants to the membrane, further treatment of the latter falls into two categories. One treatment sometimes required is a drying or curing step to maintain the immunoreactivity of reagents and to fix reactants to the membrane. This is often accomplished by desiccating the membrane, sometimes at an elevated temperature.

The second postapplication treatment sometimes employed is to block the membrane to prevent subsequent nonspecific interactions. A variety of membrane blockers have been successfully employed, most commonly proteins. The membrane is typically dipped in a high concentration of a protein, such as albumin, and allowed to dry. Other systems instead incorporate a blocker into a release pad to pass through the membrane with the sample. Surfactants are often used for this purpose. One problem that may be encountered when blocking a membrane with a protein is a significant decline in capillary rate.

Conclusion

Characterization and modification of the physical properties of microporous membranes to facilitate their use in immunoassay delivery systems has led to the development of a new generation of nitrocellulose membranes. These membranes are cast on different supports and have modified surfactant content and a range of pore sizes. They are currently used as components of human and veterinary diagnostics and food tests and are making their way into new measurement technologies in environmental and agricultural testing. They will also play a role in molecular-biology-based (genetic) assays.

References

1. Towbin H, Staehelin T, and Gordon J, "Electrophoretic Transfer of Proteins from Polyacrylamide Gels to Nitrocellulose Sheets: Procedure and Some Applications," Proceedings of the National Academy of Sciences of the United States of America, 76:4350­5354, 1979.

2. Rudolph JL, Henderson MB, Chow O, et al., "A New Microporous Membrane for Diagnostic Immunoassays," Biotechniques, 9:218­220, 222­223, 1990.

3. Snowen K, and Hommel M, "Antigen Detection Immunoassay Using Dipsticks and Colloidal Dyes," J Immunol Methods, 140:57­66, 1991.

4. Buechler KF, Moi S, Noar B, et al., "Simultaneous Detection of Seven Drugs of Abuse by the Triage Panel for Drugs of Abuse," Clin Chem, 38:1678­1684, 1992.

5. Van Amerongen A, Wichers JH, Berendsen L, et al., "Colloidal Carbon Particles as a New Label for Rapid Immunochemical Test," J Biotechnol, 30:185­195, 1993.

6. Rao DV, and Kashyap VK, "A Simple Dipstick Immunoassay for Detection of A and B Antigens," J Immunoassay, 13: 15­30, 1992.

7. Liu H, Yu JC, Bindra DS, et al., "Flow Injection Solid Phase Chemiluminescent Immunoassay Using a Membrane-Based Reactor," Anal Chem, 63:666­669, 1991.

8. Kudo T, Iqbal K, Ravid R, et al., "Ubiquitin in Cerebrospinal Fluid: A Rapid Competitive Enzyme-Linked Immunoflow Assay," Neuroreport, 5:1522­1524, 1994.

9. Sonnenberg A, Kolvenbag G, Al E, et al., "Immunobinding Procedure with Monoclonal Antibodies," J Immunol Methods, 72:443­450, 1984.

10. Gershoni JM, and Palade GE, "Electrophoretic Transfer of Proteins from Sodium Dodecyl Sulfate Polyacrylamide Gels to a Positively Charged Membrane Filter," Anal Biochem, 124:396­405, 1982.

11. Levinson PA, Owen CN, and Valkirs GA, Method and apparatus for immunoassays: binding enzyme-labeled antibody to antigen on porous membrane, U.S. Pat. 4,632,901.

Michael A. Harvey, PhD, is vice president, research and development, and Charlene A. Audette is a research associate at Schleicher and Schuell, Inc. (Keene, NH). Richard McDonogh, PhD, is director of research and development, Schleicher and Schuell GmbH (Dassel, Germany).

Successful purification of recombinant proteins for use in commercial diagnostics entails efficient and cost-effective use of expensive equipment and components, in close quarters and on tight deadlines, while ensuring and validating cleanliness and purity. n the mid-1980s, Chiron Corp. (Emeryville, CA) entered the commercial diagnostic market with the purification of several recombinant proteins (primarily HIV). Soon afterward, the company formed a diagnostics purification group to produce these proteins and others, including HCV proteins. Today, the Diagnostics Purification Group purifies more than 25 proteins (antigens), licensed and nonlicensed.

I

Commercial purification of recombinant proteins is an extensive subject. To limit the scope of this article, we focus on three distinct but integral components of a manufacturing operation: process and product segregation, process transfer and scale-up, and in-process testing.

Process and Product Segregation

We may perform a given protein purification process as rarely as once a year or as often as 25 times a year, depending on commercial demand for the protein and on the process's yield. We prefer to perform a process at least once a year to keep the purification staff familiar with the procedures.

A typical process takes one week from cell breakage to final bulk. Yields can range from 20 mg to several grams, depending on the expression level and process scale.

The number of individual processes performed at our facility increased slowly over time, up to a point where we had to start processing two products a week instead of one. Today, we typically perform from two to four processes concurrently. Concurrent processing raises important good manufacturing practices (GMP) issues related to the prevention of contamination and the segregation of processes.

The movement of people, materials, equipment, and waste must be planned to prevent product contamination. Steps must also be taken to effectively segregate processes in the purification area to prevent product-to-product contamination.

Our facility consists of two purification labs and several accessory rooms. Laboratory areas have sealed vinyl floors and sealed ceilings. Air supply is controlled by an HVAC system. Purification areas are pressurized, and while air quality is not classified, environmental monitoring suggests that the air quality does meet Class 100,000 specifications.

The accessory rooms consist of centrifuge labs, cold rooms, and storage rooms. Once a company has established a suitable environment with adequate space for equipment and storage, the next consideration should be process and product segregation.

Careful planning maximizes the use of space and minimizes the opportunity for product contamination. Our purification labs have designated storage and work areas. Chemicals, intermediates, equipment, glassware, and reagents are stored in designated areas, the latter two in dedicated cabinets. Each lab has a reagent preparation area with balances, chemicals, and a pH meter nearby.

Lab benches in rows in the central area of the lab are kept clear of equipment and other supplies when not in use. Lab benches are separated by either an aisle or a divider. Each process is assigned to a specific bench.

The bench area is used for storing components and equipment during protein processing. The primary activity performed in the bench area is chromatography. When the process is finished, personnel clean and store the equipment and clear the area in accordance with written standard operating procedures (SOPs), in preparation for the next process.

The early stages of processing--cell breakage and extractions--are done in small, office-sized centrifuge labs. A centrifuge lab typically houses two to four centrifuges and about 6 ft of benchtop. A typical process starts on Monday and ends on Friday.

All processes have centrifugation steps. The availability of multiple centrifuge labs provides scheduling flexibility and minimizes the likelihood of product cross-contamination.

Processes are segregated spatially. Each handling step (reagent preparation, centrifuging, and chromatography) is performed in a separate designated area.

Another way to separate processes is by the use of dedicated personnel. We assign one or more purification operators to a specific process for its duration, reducing the likelihood of operator error. Process-dedicated personnel are responsible for setting up and performing the process as well as preparing process reagents.

Process reagents have a unique identification system. Each reagent is identified by a part number and a lot number. These numbers are assigned and noted within the body of the batch production record. The lot number of the process reagent is the same as that of the antigen being processed.

We retain samples of all reagents for potential troubleshooting. Once the run is completed and determined to be typical, the retained samples are discarded, as are unused portions. This procedure limits the number of reagents in the labs at any given time and lessens the chance that a reagent may be used in the wrong process.

Another consideration for a multiuse facility is the separation of the manufacturing of licensed and nonlicensed products. If space permits, separating the processing of licensed product from that of nonlicensed product is advisable in order to satisfy regulatory concerns. However, processing licensed and nonlicensed product together is allowable as long as adequate measures are taken to segregate the two.

About two years ago, we annexed an adjacent lab to better handle the increasing demand from Chiron's diagnostic business. We dedicated the original lab to the manufacturing of licensed products, the newly acquired lab to nonlicensed ones.

Previously we had been successfully manufacturing both licensed and nonlicensed proteins in one purification lab, following the segregation measures previously described. All processing, including nonlicensed processing, was performed according to GMPs with valid batch production records.

Process Transfer and Scale-Up

How easily and quickly a process is transferred to manufacturing and scaled up depends on variables such as the antigen's expression level in the starting material, the process methods employed, and the protein's stability. Protein purification processes can vary in complexity from those having simple wash steps and a single chromatography step to those requiring extensive extractions and several chromatography techniques. A typical process design is represented in Figure 1. A new process that necessitates considerable modifications before scale-up or one that requires novel in-process test procedures can slow down completion of the consistency series (production runs used to qualify the process).

Scale-Up Limitations. Some small-scale purification techniques have scale-up limitations. Ideally, development personnel are aware of them, but this is not always the case. Some development personnel have limited exposure to and understanding of the manufacturing environment. For this reason, our purification group established its own process transfer and scale-up group to review and test new processes before manufacturing. This group also serves to troubleshoot problems from any of our established processes.

Certain purification techniques yield good results at a development or small-scale level but may not be scalable. For example, following an ion exchange chromatography step, the product must be concentrated before being passed over a gel filtration column. A process development scientist who needs to concentrate 1 L of ion exchange material down to 100 ml decides to do so using an apparatus that relies on a membrane and nitrogen pressure. When the process design is completed, he or she transfers the process to manufacturing. Manufacturing finds that it needs to scale up the process tenfold to meet demand. However, the nitrogen concentration apparatus has a capacity of only 2 L, so manufacturing must find an alternative method of concentrating the material. The problem with this is that the results (e.g., shearing, foaming, and microbial growth) of using another method are unknown.

Another small-scale technique that is not practical for scaling up is ultracentrifugation. An ultracentrifugation step would have to be replaced by another technique in scale-up.

Bioburden. In addition to being aware of scale-up capabilities and limitations, development personnel should design processes using techniques that minimize microbial growth. For example, if a procedure requires a buffer exchange, developers should investigate the options of performing a diafiltration or performing a dialysis at 2°­8°C (as opposed to room temperature, which promotes microbial growth). Minimizing hold steps limits opportunities for microbial growth and also lessens the chance of product breakdown and aggregation. Using in-line filters during column chromatography steps helps to limit contamination from reagents.

All of our antigens are tested for microbial load before sterile filtration and filling. In almost all lots, the bioburden at that point is below the action limit, even though our processes are not meant to be sterile. Many of the buffers used during processing contain substances, like the detergent sodium dodecyl sulfate, that retard bacterial growth.

In the two or three cases where we had high bioburden in our prefiltration samples, we traced the source to a contaminated stock buffer or a container used to make the final formulation buffer. We eliminated the possibility of recurrence by making the buffer in an autoclaved container and by filtering the formulation or dilution buffer before use.

Optimal Yield. Yield does not always increase proportionally at scale-up; a tenfold increase in starting material does not always produce a tenfold increase in yield. In some cases, the yield increases at the expense of antigen purity. To maintain product quality in these cases, it becomes necessary to sacrifice yield per run. For this reason, the process development team must test the robustness of the purification procedure during the development phase.

Generally, the more of an antigen's physical characteristics (e.g., hydrophobicity, immunoaffinity, ionic strength, isoelectric point, and molecular weight) a purification process uses, the more likely it is to be successful. For example, a process in which gel filtration is the only chromatography step uses only the molecular weight of the antigen for separation. Any contaminant proteins of similar molecular weight will not be removed efficiently by such a process, and some will be left in the final product. These impurities may compromise the antigen's specificity and functionality.

Fermentation. Another key to the success of a purification process is the fermentor material. The higher the protein's expression level, the more likely the purification process is to be successful. Where the expression level for an antigen is low, the ratio of contaminants to antigen is relatively high. The process must remove a greater amount of extraneous proteins.

Process development must use the same fermentor material as manufacturing. Varying the fermentation conditions may result in the formation of different contaminants and affect the protein's attributes. Different constructs of the same antigen fermented in the same host strain can have different expression levels, different contaminant proteins, and different antigenicity. Likewise, different host strains with the same antigen construct can have different expression levels and contaminant profiles.

Fermentation also must be performed at the same scale for process development and manufacturing runs. Fermentor material from the same seed stock does not necessarily have the same expression level or behave in the same way when fermented in a shaker flask as it does in a 10-L or a 200-L fermentor.

Equipment Selection. Because the market life span of an IVD antigen may be very short and the quantities needed per year low, the use of expensive equipment and procedures (e.g., immunoaffinity chromatography) in its production is best avoided. Simpler, less expensive, and more widespread methods such as ion exchange chromatography and gel filtration chromatography usually suffice.

Automation of chromatography steps is not practical for production of most commercial diagnostic proteins because of the production scale and the low number of runs required per year. If equipment is dismantled when a process is completed (as is common in a multiuse facility), then much of the initial setup work is built into the procedure. Automation in such a case is not very practical because the amount of time saved is minimal and probably does not offset the cost of implementing and validating the automated system. Where most purification processes are performed only a few times a year and numerous processes are performed only on demand, manual techniques are more practical than automated systems.

In a multiuse facility, it is best when possible to choose equipment or components with wide-ranging capabilities, adaptable to more than one process. Doing so reduces the number of validation studies required, the number of pieces of equipment on which the operators must be trained, the number of spare parts that must be stored in inventory, and the number of vendors used for spare parts and repairs. If possible, a commercial operation should have a backup for every piece of critical equipment (e.g., centrifuges) to forestall loss of a run as a result of equipment failure. All critical equipment must be on preventive maintenance and calibration programs, as dictated by GMP requirements. Such programs should be handled by an outside department.

Validation. Another issue relevant to equipment and components is validation. Two major questions are:

* Should a piece of equipment, a system, or a component be dedicated to a particular process or validated for use with multiple products?

* Once the decision has been made to dedicate a component to a particular process, is it better to validate the removal of product carryover or to discard the component after use?

The question of whether to dedicate a piece of equipment or validate it for multiple uses depends on its cost, its frequency of use, the stage at which it is used in the process, and its interaction with product. For equipment that is very expensive and that can be used in many different processes, it is probably more cost effective to validate its cleaning than to dedicate it. This is frequently the case with equipment used in the early stages of the process, when extensive purification is to be done.

As mentioned previously, the early stages of processing in our production facility occur in small auxiliary centrifuge rooms. Dedication of equipment (e.g., bead mill, sonicator, or centrifuge) to individual processes at this stage would be expensive and pose a storage problem.

Later steps in the purification process usually involve chromatography. Because the interaction between the product and equipment and components (e.g., chromatography resin) is greater at this stage, more-innovative validation techniques are required to document cleanliness and product removal. We have found it more beneficial to dedicate the equipment and components at this stage than to validate.

Most equipment used for manufacturing (especially at later stages of purification) is too expensive to be discarded after use. Interaction between equipment and product is in most cases limited to surface contact. In such cases, validating cleanliness is not a major challenge. This is not always the case, however.

In the case of resin media, the product flows through and may interact or bind with the component. Whether to reuse or discard in this case depends on the interaction with the resin, the quantity of resin used, and the frequency of the process. For example, negligible binding occurs between a gel filtration resin that separates molecules by their size and the proteins we purify. Most processes use a lot of gel filtration media, which is quite expensive. The expense, quantity, and pass-through mechanism of separation weigh in favor of reuse of the media. Some media, such as ion exchange resins, bind protein, and validating the removal of bound proteins is a fairly extensive and expensive process. It may be more cost effective to discard these types of resins after each use, particularly if the quantities used per run are small and the annual number of runs low.

In-Process Testing

Quality control and in-process testing groups share a common goal: to ensure a consistently reproducible product. While quality control handles final product release testing and conducts stability studies for each product, the in-process group tests a variety of process intermediate samples, in accordance with GMP guidelines.

To ensure that the lot is being manufactured as specified in the batch production record, in-process tests are governed by SOPs. Analytical methods and instrumentation must be validated to demonstrate suitability for intended use.

An in-process testing group is likely to have faster turnaround times than quality control because their function is to service manufacturing. A quality control group may also be responsible for testing samples from more than one manufacturing department and therefore may have to prioritize incoming samples. Faster turnaround times are critical for minimizing hold times between steps. If in-process testing reveals problems during processing, the process can be halted immediately if necessary, limiting losses.

Specifications. The specification for an in-process assay should be tighter than quality control's specification to allow for changes that result from subsequent manufacturing steps and for lab-to-lab assay variability. For example, quality control's purity specification (by densitometric gel scanning) for a final bulk product might be set at 90%, but the in-process purity specification for pooling fractions at the final step would be higher, perhaps 95%. In-process acceptance criteria or specifications should ideally be 5­10% tighter than quality control's limits. In our operation, an out-of-spec in-process test result would prompt us to halt the process.

A successful process is one in which each production run consistently and reproducibly meets established product specifications. Typically, a purification process is qualified by the performance of a series of purification runs conducted according to GMPs. A minimum of three runs are needed to qualify a process at Chiron.

Analytical Methods. During the qualification period, a variety of off-line analytical methods (see Table I) are evaluated for antigen characterization. Other methods, known as in-line methods, are used for monitoring of the actual process. These include determinations of pH, conductivity, absorbance, and volume.

In-process testing generally refers to off-line analytical methods. A combination of analytical methods may be used at critical intermediate steps. For instance, fractions from a final gel filtration chromatography step are selected for pooling based on results of the following analytical methods: SDS-PAGE, HPLC, and densitometric gel scanning. The acceptable fractions to pool would be the most conservative range allowable, based on all three methods.

If a qualifying run is inconsistent with the others or yields a product of unacceptable quality, then at a minimum the one run must be repeated. When a run fails to meet in-process specifications, either the process was not performed correctly or some of its factors were not controlled. If the inconsistency of the run has no discernible cause, then the process may not be reproducible and further process development may be required. Any process that is poorly controlled or yields a product that just barely meets the established in-house specifications is likely to show problems in future purification runs.

In-process test limits must initially be broad enough to account for the inherent variation of purification processes in general, and to overcome the lack of a significant number of runs to use as a basis for the test limits. However, they must be narrow enough to control the process and to ensure the integrity of the product.

As more and more runs are completed, the in-process data should be evaluated and the specifications adjusted if necessary. Depending on the yield of the purification process and the commercial demand for the antigen, it may take several years to produce enough lots to justify tightening an in-process specification. The confidence that one can have in a process increases with the number of runs performed. For example, if the purity specification for an early chromatography step is initially 35% but the subsequent 10 purification runs pooled fractions that were all >=50% pure at this step, then the specification should be tightened to reflect the actual results.

In-process testing data are continually trended and evaluated in order to assess the quality of the product and the reproducibility of the manufacturing process. Examples of trend analysis include percent recovery and total protein values for specific intermediate steps and for the final bulk. These data are especially important for determining run-to-run variation. Some in-process testing (such as SDS-PAGE on extract, pellet, and wash samples) is done to collect historical information that is helpful for troubleshooting in the event of product- or process-related problems.

Stability Studies. Performing short-term stability studies for new antigens is prudent to ensure that the formulation of the product, as well as storage conditions, do not result in product breakdown or aggregation/precipitation. Rarely is this information available before transfer. Because development material is often in short supply, material from the consistency series can be used for these studies. A short-term stability study might address the following questions:

* What is the stability when the product is stored at ­20°C compared with ­60°C at various time points (for hold steps)?

* How well does the product hold up when frozen and thawed several times?

* How does it do when heated to 37°C or to boiling?

* How long is it stable when stored at 2°­8°C, or at room temperature on the bench?

* How well does it handle being thawed quickly? slowly?

* Does it break down when spiked with a heavy load of microbes typically isolated in the manufacturing environment?

As soon as the antigen-purification-consistency series has been deemed successful and all initial processing parameters and final bulk antigen specifications have been established, the antigen needs to be entered in a long-term stability study, under the oversight of the quality control department. This study should ideally be initiated immediately following the consistency series in order to gather data for the longest time before introduction of the product into the market.

Conclusion

Process and product segregation, process transfer and scale-up, and in-process testing are important considerations for any multiuse protein purification facility. Manufacturing personnel involved in protein purification regularly encounter the problems discussed here and more. The setup, procedures, and processes we have shared provide insurance against many of these pitfalls, and tools for overcoming others.

Dan Gilbrech is associate manager of diagnostics purification, Fay Kim is associate manager of diagnostics analytical purification, and Cheryl Covert, MBA, is associate director of diagnostics purification at Chiron Corp. (Emeryville, CA).

The major goal of environmental toxicology research has been to predict and prevent human disease resulting from chemical exposure and other environmental risk factors. Until about 15 years ago, the health risk associated with exposure to environmental chemicals was determined by calculations of external exposure. Such exposure, however, correlates only roughly with the amount of internal exposure and individual risk.

The introduction of molecular approaches to toxicology research has led to the discovery of biological and biochemical markers, which are increasingly valuable for predicting and preventing diseases with environmental etiology. Through recombinant DNA techniques, researchers can detect the environmentally induced changes in the structure and sequence of DNA indicated by biomarkers. Automated immunoassay technology currently in development can detect biomarkers of specific disease-causing compounds in biological fluids. Also, clinical laboratories are beginning to use biomarker-detection tests to evaluate individual risk for environmental disease.

A biomarker is the result of a subclinical event in the body caused either directly or indirectly by exposure to an environmental factor, such as a chemical or radiation. It is an indicator of disease susceptibility. Ideally, the increased risk for disease associated with the presence of the biomarker should be reversible by appropriate medical intervention. Common sample materials for biomarker assays are blood, urine, feces, hair, and saliva. DNA samples are usually derived from white blood cells, and the presence of chemical metabolites is usually determined from urine.

Biomarkers have been categorized into three types: biomarkers of exposure, effect, and susceptibility. Biomarkers of exposure are foreign compounds or their metabolites found within the body--for example, an adduct formed by the interaction of polyaromatic hydrocarbons with DNA. The formation of these adducts has been linked to the transformation of cells to a cancerous phenotype.

Another type of biomarker of exposure is the metabolic product of an environmental toxin. The metabolites of many chemicals, such as some pesticides, are measured instead of the parent compound either because the metabolite is the more toxic or because the parent compound is completely metabolized in vivo.

Biomarkers of effect result from changes in the homeostasis of an organ system that precede clinical disease. Examples of such effects are liver necrosis, low birth weight, and changes in pulmonary function. A biomarker of effect may be specific to a particular environmental toxin or it may imply a mixed exposure to a number of environmental agents.

A biomarker of susceptibility indicates whether a person is more likely to be sensitive to a xenobiotic than are most other members of the population (Cullen et al., Clin Chem, 41:1809­1813, 1995). Degrees of susceptibility to environmental disease vary with hereditary and acquired metabolic variation, mutations, the functional status and the genetic composition of a person's immune system, and variation in his or her nutritional status. For example, the levels of ß-carotene, selenium, and retinoids have been associated with the predisposition to cancer. Some individuals are known to exhibit multiple chemical sensitivities.

Epidemiological studies indicate that many diseases and most cancers are due to human exposure to environmental toxins. In recent years, there has been a great deal of interest in defining the intermediate steps of carcinogenesis. This work has been facilitated by novel molecular diagnostic techniques that enable the quantitation of changes in cellular biochemistry.

Researchers are exploring the possibility that giving chemopreventative drugs to biomarker-positive persons might reverse damage to cellular biochemistry. For example, cytologic examinations of human sputum have found that the administration of folate and vitamin B12 suppressed the development of squamous metaplasia and atypia in smokers' airways (Kamei et al., Cancer, 71:2477­2483, 1993). In addition, folate has been associated with reduced risk for invasive cervical cancer in patients with cervical dysplasia (Potischeman et al., J Nutrition 123 [2 suppl], February, pp 424­429, 1993).

Some major reference laboratories, such as LabCorp, Inc., Analytics Laboratory (Richmond, VA), already offer tests for chemical-exposure biomarkers to help monitor exposure to toxic substances in the workplace (see Table I). The entry of leading reference laboratories into the environmental toxicology field suggest that this area will see accelerated growth in the future.

Several roadblocks obstruct the application and commercialization of biomarkers for the detection of toxic chemical exposure. The analytical, diagnostic, and etiologic validity of many of the new markers has yet to be established (Tockman et al., Cancer Res, 52, pp 2711s­2718s, 1992). Recognized disease end points need to be more clearly associated with them. Standardized criteria for the quantitative measurement of these new markers must be established, and the predictive values of each of them must be determined by population studies.

At the present time, little is known about how FDA will greet the use of the new biomarkers in the clinical laboratory to predict disease risk. In the past, the agency has approved tests that help predict the risk for heart disease, such as those for high- and low-density lipoproteins. However, it has been slower to accept changes in biomarker levels as end points in clinical trials for infectious-disease drugs. FDA's stance on biomarkers will depend upon their epidemiological significance.

In addition to the technical challenges, the social impact of biomarker assays must be considered. New diagnostic tests will have the capability of identifying predisposition to serious and life-threatening diseases. In many cases, there will be a lag between the time a marker is identified and the availability of treatment to prevent the onset of disease. Information on increased risk may be especially undesirable if potential employers or life insurance companies could access it. Legislation should be enacted to prevent the use of this technology in a manner contrary to societal goals. However, such legislation should not defeat the ability of this technology to predict and prevent environmental disease.

Further issues may exist in tort law. In Ayers v. Jackson Township (106 NJ 557), the courts ruled that persons who had been exposed to aromatic hydrocarbon solvents but at the time of the trial exhibited no disease were entitled to damages for medical surveillance and the enhanced, although not quantified, risk of disease. The availability of biomarker assays may influence litigation such as this case and cause potential polluters to be more cautious.

The potential for early treatment and disease prevention should justify the extensive cost and effort of commercializing biomarkers. The benefit of such tests greatly outweighs any negative societal impact. Biomarker-based diagnostic tests may offer opportunities to intervene effectively with relatively inexpensive treatments before environmen-tally induced disease advances to a stage where more costly therapy is required.

Steven M. Rosen, PhD, is a senior scientist for the Drug Monitoring Business Unit of Roche Diagnostic Systems, Inc. (Somerville, NJ).

It's little wonder that the automated Pap smear has found rapid acceptance in the clinical laboratory. The traditional Pap smear is notorious for its high false-negative rate. But now an FDA-approved assay for a cervical cancer marker has arrived, allowing much more accurate diagnosis. The assay detects infection by certain types of human papillomavirus (HPV) that are associated with increased risk of cervical cancer.

Digene, Inc. (Silver Spring, MD), developed the DNA-based diagnostic employing its Hybrid Capture method. In direct comparisons with Pap smears, the HPV DNA test has significantly greater sensitivity for premalignant lesions--what physicians refer to as grades II or III of cervical intraepithelial neoplasia, says Attila Lorincz, PhD, vice president and scientific director of Digene.

Indications for the HPV DNA test are expected to be included in new guidelines issued this year by the American Society for Colposcopy and Cervical Pathology (Washington, DC). The document, "Management Guidelines for Atypical Squamous Cells of Undetermined Significance," is expected to be published this spring.

In April, a scientific panel convened by the National Institutes of Health (NIH) concluded that HPV is a principal cause of cervical cancer. Some HPV types integrate with human DNA, a feature believed to account for the invasiveness of certain forms. The panelists stressed the importance of educating health-care providers about the causal link. Historically, this kind of pronouncement--official recognition of a cause-and- effect relationship--has preceded a call for changes in testing.

The traditional Pap smear is entrenched in the American health system, though, and a change will take years. The American College of Obstetricians and Gynecologists (ACOG), for example, does not endorse HPV DNA testing. When it examined the concept two years ago, the organization concluded that the "clinical expression of HPV is highly variable," and that the lesions are subject to "spontaneous regression and recurrence."

Lorincz agrees that certain HPV infections are transient and will disappear on their own. Digene's third-generation test, however, uses new, multiple DNA probes to detect only those types of HPV that cause the persistent infections most closely associated with increased risk for cervical cancer.

A year ago, ACOG's Committee on Gynecologic Practice designated women with current or prior HPV infection as being at high risk for cervical cancer. This, together with the recent findings by the NIH panel, may make the committee take a new look at the issue, said Alice Kirkman, public information specialist for ACOG. And if ACOG revises its recommendations, the Society of Gynecologic Oncologists (Chicago) is likely to follow suit.

HPV DNA detection has won widespread support among pathologists, if not OB/GYNs. HPV DNA testing is the wave of the future in cervical cancer screening, according to Carl Treling, MD, chair of pathology at Queen of Angels/Hollywood Presbyterian Hospital in Los Angeles. "If you have some reliable way to screen for HPV, you won't even need the Pap smear," he says. "But it will take ACOG cheering it on" to ensure acceptance into medical practice, he adds.

"In the short term, this will not replace the Pap smear in the United States," Lorincz agrees. In developing countries, where the Pap smear is not so well established, "we think HPV DNA will become the de facto test for cervical cancer," he says.

Last year, the Canadian Task Force on the Periodic Health Examination, a panel that evaluates tests and procedures for that nation's health system, reviewed the evidence on HPV DNA screening and decided not to recommend including it in routine exams. The task force cited as one major limitation the fact that hybridization techniques are relatively new.

Another barrier to widespread acceptance of the HPV DNA test is the recent FDA clearance of two automated Pap smear analyzers for QC use. Clinical laboratories that have invested in these systems, which are designed to reduce the false-negative rate of Pap smears, will be loath to find them less universally applicable.

The AutoPap 300QC, from Neopath, Inc. (Redmond, WA), is showing great promise. It reviews negative slides and selects the 10% with the most suspicious cells for cytotechnologists to rereview, to see if they contain a missed abnormality.

Lorincz hypothesizes that widespread use of the automated Pap smear analyzers could actually increase demand for the HPV DNA test. With the automated systems, the number of Pap smears judged to be equivocal would increase, swamping scarce personnel and resources. The HPV DNA test could be used for triage to determine which patients with equivocal Pap smear results are at highest risk and require immediate follow-up invasive diagnostic procedures.

How well the Hybrid Capture HPV test fares in today's competitive environment will probably depend on the way it is received by the medical community and managed-care providers. Good data are just not enough to ensure reimbursement, cautions Don White, spokesman for the American Association of Health Plans (Washington, DC). Reimbursement by HMOs often goes hand in hand with a medical consensus, says White. Support by an NIH consensus conference or multiple clinical trials specifically endorsing a testing protocol is what is compelling, he stresses.

Today's requirement for reducing the cost to produce a reportable patient result is driving the laboratory to consolidate as much testing as possible onto the fewest instruments. Manufacturers must respond by developing systems that provide a large menu of assays. he Access immunoassay system (Sanofi Diagnostics Pasteur, Inc., Chaska, MN) provides a tool for developing assays for a wide array of molecules. The instrument design has been described previously.^1,2 This article focuses on the two key chemistry parameters, paramagnetic particles (PMPs) and chemiluminescence, that provide the basis for developing high-sensitivity assays with large dynamic ranges.

T

Access Solid Phase

Like most available immunoassay systems, the Access system uses heterogeneous immunochemistry. The solid phase is a submicron-size paramagnetic particle. The particles are polystyrene with incorporated iron particles and carboxyl functional groups on the surface. Antibodies, antigens, or other ligands are coupled covalently to the particle during reagent manufacturing. Standard sandwich or competitive assay formats are used; unbound enzyme label is washed from the particles before initiation of the chemiluminescent substrate reaction.

Why choose paramagnetic particles? Particles are commercially available in a variety of sizes and with a variety of functional groups for covalent coupling of biologicals. Covalent attachment of antibodies and antigens to the surfaces of PMPs yields reagents with shelf lives of a year or more at 2°­10°C. Since PMPs can be suspended in solution, coating them is an easier manufacturing process than coating classic solid phases, such as microplates or 1/4-in. beads. We use a fully automated liquid-handling process to couple ligands to particles. Particle concentrates are easily stored in bottles until they are needed for production. One liter of coupled PMP concentrate is enough to manufacture 200,000 tests. By comparison, over 50 L of 1/4-in. beads plus additional coating solution are required to manufacture the same number of tests.

Convenience and stability aside, PMPs provide exceptional assay performance and ease of integration into an immunoassay analyzer. The high surface area provided by submicron particles and the ability to suspend them near the target analyte enable fast reaction times. Our test results show that reactions come to equilibrium approximately four times faster with PMPs than with a standard 1/4-in. bead.

Processing the PMPs as a liquid reagent allows the assay developer to optimize the mass of solid phase for each analyte. Assays requiring a large dynamic range, such as hCG, employ 100 mg of PMPs, whereas other assays may use only 25 mg.

The Luminescence Phenomenon

Close observation of the glowing firefly is fascinating, whether the observer understands the phenomenon of luminescence or not. The firefly's glow is just one example of luminescence. Thousands of species, including insects, bacteria, fungi, worms, and countless sea animals, use light emission as a tool.

In the broad sense, luminescence is simply the conversion of energy into light. The energy is stored in chemical bonds and released by a chemical reaction. Just as the chemical reaction of burning wood releases chemical bond energy to produce heat, enzymes in the firefly help to release chemical bond energy from adenosine triphosphate (ATP) to produce light.

In living organisms, this luminescence is referred to as bioluminescence. The reactions in such species require enzyme action. Through the study of bioluminescence, researchers were eventually able to synthesize molecules that could emit light in a test tube without an enzyme catalyst. This phenomenon has been termed chemiluminescence.

In the last 15 years, the utility of luminescent labels has steadily increased. The phenomenon has been applied to biomedical science in immunoassays, DNA probe assays, and measurement of important enzymes and metabolites.

Immunoassays based on chemiluminescence have substantially greater sensitivity and dynamic range than those based on earlier-generation detection techniques. Efficient light emission with low background is coupled with the high sensitivity and broad range of the photomultiplier detector. For every photon of light striking the surface of the photomultiplier, there is a 10⁶-fold electronic amplification of the signal. Photomultipliers have very low background noise and inherent dynamic ranges of 5 to 6 orders of magnitude.

Performance Characteristics

Many excellent review articles and reference books cover performance characteristics.^3­7 The various luminescent reactions do not all result in the same sensitivity or the same ease of use. Their utility is governed by several key variables.

Quantum efficiency is perhaps the most important variable. It is defined as the percentage of those molecules undergoing the chemical reaction that emit photons. The maximum obtainable luminescence quantum efficiency would be 100%, although in practice this level has never been obtained. Some bioluminescent reactions have efficiencies of up to approximately 90% in vivo. Most chemiluminescent reactions have well below 15% quantum efficiency. The percentage of chemical reaction energy that is not dissipated by light emission is dissipated as heat or as kinetic collisional energy with other molecules in solution.

The impact of luminescence quantum efficiency on the sensitivity of the immunoassay depends on how the luminescent molecule is used for detection. For chemiluminescent labels that are coupled directly to protein (e.g., acridinium esters), the sensitivity is directly proportional to the quantum yield. For reactions that are amplified by an enzyme, sensitivity depends on both the quantum efficiency and enzyme turnover. Enzyme amplification can turn a reaction system with moderate quantum efficiency into an ultrasensitive reaction.

Background luminescence is another key variable. The background comes from two major sources:

One is simply nonspecific binding of the label. Nonspecific binding is not directly related to the chemiluminescent reaction. Its effect on performance characteristics depends on how well a given immunoassay procedure has succeeded in washing away nonspecifically bound proteins or in inhibiting nonspecific binding in the first place.

The second important source of background is nonspecific light emission of the chemiluminescent molecule. This phenomenon can occur before or during the initiation of the specific reaction and is caused by unwanted oxidants, metal catalysts, pH differences, enzymatic activity, and other variables. A review of various chemiluminescent compounds and their potential interferences is outside the scope of this article. Washing unbound material from a solid phase before triggering the specific luminescent reaction helps keep these interferences to a minimum.

Thermal degradation is another mode of unwanted light emission. The level of thermal-initiated background is specific for the chemiluminescent label or substrate in question and for the particular temperature.

For a sandwich immunoassay format, the detection limit is a function of signal-to-noise ratio. The considerable disparity in the quantum efficiency of different luminescent reactions makes the choice of a label particularly important. Once the label is chosen, however, proper control of background can make the difference between moderate and optimal sensitivity.

The Access system uses LumiPhos 530 (Lumigen, Inc., Southfield, MI), a chemiluminescent enzyme substrate for alkaline phosphatase. LumiPhos has a one-year shelf life when stored at 4°C and a one-week shelf life once a bottle of it is plumbed into the instrument. The LumiPhos reaction combines the inherent sensitivity of chemiluminescent light detection with the signal amplification of an enzyme label. As little as 10^­21 M of alkaline phosphatase can be detected with this substrate.⁸ Figure 1 depicts the detection reaction for this system.

Low-Level-Light Detection

Detection of the low levels of light produced by chemiluminescence generally requires the use of a photomultiplier tube (PMT). A PMT consists of an evacuated glass tube containing a photocathode, a number of dynodes, and an anode. As photons strike the photocathode, electrons are ejected and accelerate toward the first dynode. Each electron striking the dynode then ejects several more electrons. This process is repeated for each dynode in the chain. Thus, each impact of a photon upon the photocathode results in production of an electron current pulse at the anode. Typical amplification throughout the dynode chain is 10⁶ to 10⁷. If the incident light is intense enough, the individual pulses overlap and merge. The resulting signal then becomes a fluctuating dc current.

The output from a photomultiplier may be processed in either digital or analog mode. In digital (or photon-counting) mode, discrete photoelectron pulses are counted. Photon counting is the natural choice for the low light levels produced by chemiluminescence. It affords greater immunity to noise, drift, and variations in supply voltage and components. Its output is inherently digital, so there is no extra analog-to-digital conversion step (see box below). In analog mode, the fluctuating photocurrent is amplified, filtered, and read as dc signal. This detection mode is appropriate for higher light levels.

High Voltage and Discriminator Threshold

Even in the absence of light, PMTs produce some noise pulses (or dark pulses) stemming from a variety of sources.^9,14 The number and size of both dark and signal pulses increase with increasing temperature and high voltage. On average, the noise pulses are smaller than the signal pulses. The system may be programmed to reject them by counting only pulses greater than a fixed threshold. This ability to reject noise is a chief advantage of photon counting over analog mode.

The system designer may select a particular high voltage and adjust the discriminator threshold to maximize the signal-to-noise ratio. Another option is to select the discriminator threshold first and then adjust the high voltage. We used several mathematical techniques to visualize the functional dependence and locate the optimum operating point.

To ensure a robust design, system developers must take into account the wide variability among PMTs. We obtained 11 randomly selected PMTs from the manufacturer and located the optimum operating point for each. From these 11 optimums, we then chose the one that gave the best signal-to-noise ratio across all 11 PMTs.

Individual tubes can vary in gain by as much as a factor of 10. For ease in manufacturing and calibration we wanted the system to require only a single adjustment--not both high voltage and discriminator threshold. We chose a constant discriminator threshold with a variable high voltage. We found that varying the high voltage gives a wider range of gain adjustment than varying the discriminator.

Temperature Control

Optimal system performance requires precise temperature control. Antibody binding, alkaline phosphatase activity, the LumiPhos reaction, PMT gain and dark counts, luminometer electronics, and reference LED output are all sensitive to temperature. We found it necessary to control the temperature of the PMT, luminometer electronics, reference LED, and antibody reactions at 37° ±1°C, and the temperature of the LumiPhos addition and reaction at 37° ±0.25°C.

Performance Characteristics

Figure 3 shows the response of the luminometer to serial dilutions of alkaline phosphatase reacting with LumiPhos 530. The plot shows that the luminometer is linear up to about 10 million counts per second (cps), with a usable range extending to 30 million cps. The dark-pulse counts for the system are generally less than 100 cps. This gives a total dynamic range of 5.5 decades. This range is not fully realized for assays, however, because of the background signal from the substrate, which in this case is at about 6000 cps. The usable range for assays is therefore 3.7 decades. The lower limit of detection (signal equal to two standard deviations above the substrate background) is 0.032 attomoles of enzyme. The luminometer thus covers 5.7 orders of magnitude in enzyme concentration.

The response plateaus at about 60 million cps. Other experiments have shown that the response actually decreases with higher enzyme concentrations. This "hook" effect is due primarily to pulse overlap at high count rates.

Signal Processing and Calculations

As mentioned previously, samples may be located in every third carousel position. When these locations are presented to the PMT, the system takes 10 1-second readings and computes the median. If a sample is present, the median is used to compute the sample response; otherwise, it is stored as the dark-pulse count. The median gives better rejection of noise than the average does.

Sample readings are adjusted for dark pulses and luminometer drift in the following way: The reference LED is read periodically in positions adjacent to samples. A drift correction factor (DCF) is calculated by comparison of the current reading to a previously stored value set at the factory.

The DCF is used to normalize the sample reading after the dark pulses are subtracted.

Normalized net sample reading = k * DCF (reading ­ dark-pulse count)

The constant k adjusts for the prescaler in the circuit.

Calibration and Luminous Standards

To aid in development and allow matching between instruments, we employed various luminous standards. Early on, we used standards containing from 1 to 500 µCi of tritium-labeled palmitic acid in a liquid scintillation cocktail. Its spectral output peaks at 440 nm. Later, we developed standards using a proprietary solid-phase tritium-labeled scintillation polymer with an emission maximum at 545 nm. The latter more closely matches the spectral output of LumiPhos 530. This match is important, since variability in spectral sensitivity among PMTs may cause calibration and mismatching errors.

Conclusion

We have succeeded in building an immunoassay system that makes possible the development of a wide range of immunoassays. By combination of paramagnetic-particle, solid-phase, enhanced chemiluminescence and a sensitive luminometer, the system delivers assays with rapid kinetics and high sensitivity.

References

1. Patterson W, Werness P, Payne WJ, et al., "Random and Continuous-Access Immunoassays with Chemiluminescent Detection by Access Automated Analyzer," Clin Chem, 40(11):2042­2045, 1994.

2. Access Immunoassay System," Clin Instr Sys, 13(5):1­9, 1994.

3. Kricka LH, "Chemiluminescent and Bioluminescent Techniques," Clin Chem, 37: 1472­1481, 1991.

4. McCapra F, Watmore D, Sumun F, et al., "Luminescent Labels for Immunoassay--From Concept to Practice," J Biolum Chemilum, 4:51­58, 1989.

5. Wood WG, "Routine Luminescence Immunoassays--Dream or Reality?" J Biolum Chemilum, 4:79­87, 1989.

6. Van Dyke K, and Van Dyke R (eds), Luminescence Immunoassay and Molecular Applications, Boca Raton, FL, CRC Press, 1990.

7. Scholmerich J, Andreesen R, Kapp A, et al. (eds), Bioluminescence and Chemiluminescence: New Perspectives, Chichester, UK, John Wiley, 1987.

8. Schaap AP, Akhavan H, and Romano LJ, "Chemiluminescent Substrates for Alkaline Phosphatase: Application to Ultrasensitive Enzyme-Linked Immunoassays and DNA Probes," Clin Chem, 35:1863­1864, 1989.

9. How to Perform Photon Counting Using Photomultiplier Tubes, Bridgewater, NJ, Hamamatsu Corp., 1990.

10. Turner GK, "Measurement of Light from Chemical or Biochemical Reactions," in Bioluminescence and Chemiluminescence: Instruments and Applications, Van Dyke K (ed), Boca Raton, FL, CRC Press, pp 43­78, 1985.

11. Van Dyke K, "Commercial Instruments," in Bioluminescence and Chemiluminescence: Instruments and Applications, Van Dyke K (ed), Boca Raton, FL, CRC Press, pp 83­128, 1985.

12. Wampler JE, and Gilbert JC, "The Design of Custom Radiometers," in Bioluminescence and Chemiluminescence: Instruments and Applications, Van Dyke K (ed), Boca Raton, FL, CRC Press, pp 129­150, 1985.

13. Berthold F, "Instrumentation for Chemiluminescence Immunoassays," in Luminescence Immunoassay and Molecular Applications, Van Dyke K, and Van Dyke R (eds), Boca Raton, FL, CRC Press, pp 11­25, 1990.

14. Photomultiplier Tubes, Bridgewater, NJ, Hamamatsu Corp., 1990.

Richard Creager, PhD, is vice president, research and development; David Knoll, PhD, and Curtis Shellum, PhD, are scientists; and Peter Werness, PhD, is director of advanced technology for Sanofi Diagnostics Pasteur, Inc. (Chaska, MN). They work together in Access research and development. Dr. Creager is a member of the Editorial Advisory Board of IVD Technology.