Validation of the BacT/ALERT® 3D System for Rapid Sterility Testing of Biopharmaceutical Samples

Results and Discussions

Because of the importance of sterility testing to determine the safety of pharmaceutical and biopharmaceutical products, an alternative method protocol and validation study must be thorough and reproducible. Several studies have reported the validation of alternative microbiological methods for the analysis of sterile pharmaceutical products (6, 7, 9). These studies were based upon the recovery of microorganisms from samples using different types of assays and detection systems. Specific regulatory informational chapters described the recommended parameters needed to demonstrate the suitability of an alternative microbiological system for sterility testing (12–14).

The BacT/ALERT® is a multi-component system based upon a control and incubator module connected to a database to manage and control the information generated by the system. Therefore, the first step in the validation of the system was the installation qualification to verify the manufacturer's specifications and procedures. The actual installation was visually observed and inspected to verify that it agrees with the system documentation.

The installation qualification of the BacT/ALERT® system was initiated by the collection and verification of all related documents supplied by the vendor. These documents were the BacT/ALERT® user guide and certificate of calibration. A list of specifications, manuals, standard operating procedures (SOPs), reports, and certifications associated with the system was completed and verified (Table I). An equipment component list based upon information from nameplate, manufacturer name, model number, calibration date, purchase orders, serial numbers, and descriptions of each component was completed and verified.

All calibrations were verified to be in compliance with National Institute of Standards and Technology (NIST) standards. Recalibration intervals for critical instrumentation were established to be yearly. All information regarding the manufacturer name, serial numbers, utility verification, and software verification was documented and verified (Table I). The installation of the system and the software were successfully completed and verified based upon the manufacturer's specifications (Table I).

The operation qualification was initiated by the development of a SOP specifically written for the operation, maintenance, and documentation of the system (Table II). This SOP covered not only these areas but also the back-up of archived reports, instrument data, operating software, user access, and audit trails. All the analysts were assigned confidential user names and passwords to access the different levels of the system to establish a secure and stable operational environment. The configuration of the system's software was determined to maximize efficiency and protect data integrity (10). The original access levels created were based upon the ability of an analyst to load samples into the instruments and generating the final reports. However, the software did not have the capability of protecting the data integrity from manipulations from the analysts who loaded the samples. Therefore, a segregation of user access allowed the protection of the data integrity by restricting the analyst to loading the samples and entering the test parameters and sample information. Once the test was initiated, the same analyst cannot generate the final report nor has any access to modify the report. A different analyst who did not load the samples into the system was capable of the report generation. Furthermore, because all analysts have a specific access to the system by passwords and user name, the system's data trail allowed the constant audit of analyst's behavior preventing the possibility of data manipulation. All the alarms system incorporated into the system such as Power/Loss and System Function were verified and found to be functional (Table II). Similar results were obtained with all the other parameters described in Table II.

Because the system is based upon the continuous incubation and reading of the samples, it was very important to map the incubator chambers to determine the stability of the set temperature. Empty and full chambers temperature measurements were performed to determine the stability of the set point temperature. The chosen temperature for sample's incubation was determined to be 33 °C. Preliminary studies demonstrated that this temperature allowed the optimal growth of bacteria, yeast, and mold both for clinical and environmental isolates. Temperature readings were stable over a 24 h period under continuous operation for both empty and full chamber runs. Furthermore, ongoing, daily use of the system confirmed the stability of the set point temperature. A range of 0.5 °C was chosen for the set point temperature. A SOP was established to recalibrate the temperature if an out-of-range reading was recorded.

To determine that the WFI, UB, and FB samples did not interfere with the detection platform of the system by generating background CO₂ enough to trigger a false positive result, aliquots of the three different sample types were added into iFA (aerobic) and iFN (anaerobic) media. The bottles were incubated for 14 days. Table V shows the incubation of the biopharmaceutical samples inoculated into the media used for sterility testing with the BacT/ALERT®. No CO₂ production was detected by any of the samples, indicating the absence of microbial growth. Furthermore, during the spiking studies additional negative samples were also incubated. A positive result was never obtained. This demonstrated that the samples did not have any interference with the assay and that there was no CO₂ generated by the different cell culture and purification materials.

The performance qualification experiments were designed to comply with the requirements described by the published pharmacopeial chapters and guidelines on the validation of alternative microbiological methods (Table IV) (12–14). Similar protocols were previously reported for the validation of the system to detect microbial contamination in cell therapy products (9). The protocol evaluated the system for sensitivity, limits of detection, ruggedness, equivalency with and without microbial spiking, time of detection, and repeatability (Table IV). Eleven different microbial species were used in this study (Table III). In addition to the compendial strains recommended by the regulatory agencies, Burkholderia cepacia was part of the study due its frequent presence in sterile product recalls (15). Previous studies reported that B. cepacia is the most frequently isolated microbial species in sterile product recalls (15). Four environmental isolates identified as Kocuria rhizophila, Staphylococcus haemolyticus, Methylobacterium radiotolerans, and Penicillium spp. were also spiked into individual samples. Of these four microbial species, M. radiotolerans exhibited the slower growth at either 25, 33, or 35 °C. The first nine microorganisms from Table III were spiked into the samples by three different analysts, while B. cepacia and M. radiotorerans studies were performed by a single analyst in triplicates and analyzed using the BacT/ALERT® 3D system.

The manufacturing of biopharmaceutical products comprises the processing of mammalian cell cultures in a bioreactor to produce the maximum quantities of drug product (16). All reagents and raw materials are thoroughly tested for the presence of microorganisms. Water is a very important raw material that is used in the processing and rinsing of materials and equipment. The performance qualification of the system was initiated by the three analysts inoculating aliquots of nine different types of microorganisms to samples of 10 mL of sWFI with an inoculum level of less than 100 CFU. The samples were tested using the standard sterility test by using TSB and FTM media and the BacT/ALERT® system with iFA (aerobic) and iFN (anaerobic) media. Additional studies were also performed in triplicate with a single analyst spiking B. cepacia and M. radiotolerans into the iFA and iFN media bottles containing sWFI samples.

The sensitivity, ruggedness, repeatability, and time of detection of the system were demonstrated by detecting all nine microorganisms in the water samples by the three analysts. Both the BacT/ALERT® system and the standard method were able to detect all the bacteria, yeast, and mold spiked into the samples within 14 days of incubation (Table VI). All anaerobic and aerobic microorganisms were also detected using the two different types of growth media. The limits of detection between the three analysts in sWFI samples were ranging from a low of 3 CFU by K. rhizophila to a high of 67 CFU by Penicillium sp. (Table VII). These limits of detection represent the lowest and highest CFU per plate obtained by any of the three analysts during the inoculation studies.

Both microbes were environmental isolates from different cleanroom sampling sites. K. rhizophila exhibited a slow growth rate when grown in regular TSB media at 25 °C, and detection time was 3 days; but it was detected by the BacT/ALERT® system within 1.57 days at 33 °C (Figure 3). Evidently, higher incubation temperatures and shaking accelerated the growth of K. rhizophila, allowing the faster detection by the BacT/ALERT® system.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

Detection of microorganisms by the BacT/ALERT® (BacT) and standard method (SM) in three different biopharmaceutical samples.

The ruggedness of the system was proven to be strong, as all the microorganisms were detected by different analysts on different days using different reagents and inoculum preparations. Time to detection for all microorganisms was faster with the BacT/ALERT®. Furthermore, the average detection time for all microorganisms in water by all three analysts was 1.16 days, while the average detection time using the standard method media was 2.4 days (Table VI, Figure 3). The fastest detection time by the BacT/ALERT® was observed by B. subtilis with 0.64 days, while A. niger showed the longest with 1.87 days (Table VI). The BacT/ALERT® was found to be equivalent to the standard method because both detected all microorganisms within 3 days (Figure 3). However, the higher incubation temperature and agitation of the samples in the BacT/ALERT® system accelerated the growth of all the microorganisms analyzed, decreasing the time of detection. The longest overall detection time in sWFI samples was obtained with M. radiotolerans. A single analyst study performed in triplicate showed an average detection time of 4.31 days.

Equivalency was also determined by analyzing unspiked samples with the BacT/ALERT® and the Steritest™ system. All the nine samples representing triplicate samples of the sWFI, UB, and FB categories were found to be negative by both methods (Table VIII). Furthermore, the Steritest™ testing was performed in a different facility and with different analysts. For repeatability, all replicates inoculated with microorganisms either using the BacT/ALERT® or standard media were positives as determined by the three different analysts (Figure 3).

In the biopharmaceutical industry, downstream processing is a series of operations that starts with a large volume of a complex biological mixture and refines it to a small volume containing the purified product in bulk form (16). Once the mammalian cells in the bioreactor reached the density and activity required for optimal antibody production, the cells are harvested to separate them from the protein fraction. This is what is called UB. The UB samples were spiked with nine different microorganisms by the three analysts and B. cepacia and M. radiotolerans in triplicate by a single analyst. The limits of detection between the three analysts in UB ranged from a low of 4 CFU by S. haemolyticus and Penicillium spp. to a high of 67 CFU by C. albicans (Table VII). All microorganisms were detected by the three analysts within 14 days. The average detection time with the BacT/ALERT® for all microorganisms in UB as determined by all three analysts was 1.25 days. The fastest detection time was observed again by B. subtilis with 0.66 days, while C. albicans showed the longest with 2.91 days (Table VI). In the UB samples, the BacT/ALERT® was equivalent to the standard enrichment method. Similar to the sWFI samples, the longest detection time in the UB was found to be 6.24 days by M. radiotolerans.

After the antibody is purified out from the mixture, the FB drug goes into final processing formulation (16). For the FB samples, the limits of detection between the three analysts ranged from 1 CFU by C. albicans to 56 CFU by both C. sporogenes and S. aureus (Table VII). For the FB samples, the average detection time for all microorganisms by all three analysts using the BacT/ALERT® was 1.23 days. The fastest detection time was observed by B. subtilis with 0.65 days, while C. albicans showed the longest with 2.47 days (Table VI). Again in the FB samples, the BacT/ALERT® was equivalent to the standard enrichment method. The longest detection time in the FB samples was found to be 6.50 days by M. radiotolerans.

The BacT/ALERT® system always detected all the microorganisms faster than the standard method (Figure 3). Although a single incubation temperature, for example, 33 °C, was used, there was no negative impact on the sensitivity of the assay. All microorganisms were detected in the three different sample types analyzed. The system also agitated the samples, which was completely different from the static cultures used in the compendial tests (1, 2). Evidently, this sample agitation enhanced the detection of the microorganisms by the system. These results confirmed the studies reported by previous investigators using different biological samples such as autologous cultured chondrocytes (9).

To determine the minimum levels of microorganisms detected by the system, different microorganisms were diluted in sPBS to achieve and inoculum of 1 CFU and immediately spiked into the biopharmaceutical samples. Dilutions were made to try to get to a 1 CFU inoculum, which was obtained, in some cases but overall all the microorganisms spiked into different biopharmaceutical products were detected within a limit of 1–2 CFU (Table IX). All microorganisms were detected in less than 7 days regardless of the inoculum level spiked (Table IX). For example, P. aeruginosa was inoculated at three different inoculum levels, 23 CFU, 2 CFU, and 1 CFU. However, after incubation, the conditions were optimal enough to detect all the different inoculation levels within 1 day. On the other hand, M. radiotolerans inoculated with 12 CFU and 2 CFU was detected within 7 days. For all the microorganisms tested the enrichment conditions were favorable enough for the cells to multiply and provide a positive signal. The system is a rapid absence/presence test and does not provide quantitative information. Previous studies reported estimated limits of detection for the BacT/ALERT® 3D of 1–2 CFU in antibiotics formulations artificially contaminated using the most probable limit of detection method (17). We were able to verify these results with artificially contaminated biopharmaceutical samples. The use of enrichment procedures optimizing the detection of microbial contamination in pharmaceutical samples using the BacT/ALERT® 3D and ATP bioluminescence assays have been previously demonstrated (17, 18). The iFA and iFN growth media used in this study are basically the same used for the standard sterility test, TSB and FTM. However, the iFA and iFN media were modified by the addition of activated charcoal to neutralize the antimicrobial activity of a given formulation (9). The standard sterility test relies on visual endpoints, which are extremely subjective and create false readings in turbid samples or in samples with low turbidity (4). For instance, microorganisms with slow growth rate and low biomass such as M. radiotolerans and K. rhizophila do not produce high microbial biomass in enrichment broths during standard tests. However, the BacT/ALERT® 3D detection technology is based upon the production of carbon dioxide in the growth media, which is a completely objective analysis performed by the instrument.

The average detection time for all microorganisms in the three different types of biopharmaceutical samples studied using the BacT/ALERT® was 1.62 days. Differences between sample types were not significant for six of the 11 microorganisms analyzed. For instance, it took the same average time to detect P. aeruginosa in sWFI, UB, and FB samples. It did not matter that the three samples were derived from different manufacturing processes or that steps or were performed by different analysts on different days. However, B. cepacia, C. albicans, M. radiotolerans, and K. rhizophilla were detected faster in sWFI than in UB and FB samples. When all sample types were combined in all studies, the longest overall detection time was found with M. radiotolerans, 5.77 days, and the fastest with B. subtilis, 0.65 days (Table VI). With the standard method, some of the slow-grow microorganisms such as K. rhizophila and A. niger were very hard to detect by using a visible endpoint to be considered positive. Therefore, an additional day of incubation was needed to obtain a positive visual endpoint. All the three sample types inoculated with B. cepacia showed positive detection by both systems. The average detection time for B. cepacia in all three samples was 1.05 days, while the standard method took 2 days to demonstrate a distinct visual endpoint (Table VI, Figure 3). However, B. cepacia detection in sWFI was faster than in the UB and FB samples. This was very significant because B. cepacia is mostly associated with water contamination during production and processing (6). For M. radiotolerans, the detection times for sWFI samples were significantly faster than in the UB and FB samples (Table VI). Methylobacterium species are known oligotrophic bacteria present in environments with low carbon concentration, which were previously isolated from pharmaceutical manufacturing waters (19). They are slow growers capable of growing on single carbon compounds more reduced than carbon dioxide such as methanol, formaldehyde, and formate, and they demonstrated resistant to chlorine (20, 21). We wanted to see if lowering the carbon concentration in the iFA media will lower the detection time of M. radiotolerans down to the times obtained by the other microorganisms. M. radiotolerans exhibited faster detection time in samples of FB when an equal volume of the iFA media was aseptically replaced with 5 mL or 10 mL of sPBS (Table X). By diluting the media in sPBS the carbon concentration was lowered, allowing the earlier detection in the most diluted sample, for example, 10 mL, by almost a day. The manipulation of the commercial media, for example, iFA, by removing and adding any volume from the bottles prior to testing can increase the risk of analysts contamination during testing. However, these results indicated that media optimization for rapid technologies based upon broth enrichments can enhance the microbial growth in the contaminated samples, leading to faster detection times. Similar results were previously reported using ATP bioluminescence technology by enrichment broths and solid agar media (7, 18).

When the bacteriostasis and fungistasis—the validation test—of the FBP sample A was tested, A. niger and C. albicans showed the longest detection time with the BacT/ALERT® with an average of 2.1 days (Table XI). The fastest detection times were found with B. subtilis with an average of 0.68 days. The FBP sample B showed very similar results, with the longest detection by A. niger, 2.01 days, and again B. subtilis with the fastest, 0.68 days (Table XII).

These results were very similar to the detection times obtained with the UB, sWFI, and FB samples. The positive control samples exhibited similar detection times when compared to the inoculated samples, indicating that the samples did not have any antimicrobial activity. In general none of the compendial microorganisms tested required more than 3 days to be detected by the system using a single temperature, 33 °C, with agitation. This study confirmed the results reported by Kielpinski et al. (9) demonstrating the feasibility of the BacT/ALERT® system for rapid sterility testing of cell therapy products. They reported the rapid detection of different types of microorganisms in autologous cultured chondrocytes. No interference was found by those samples and the system. Similar results were obtained in this study with a more chemically diverse set of samples. We analyzed four different types of biopharmaceutical samples and did not see any inhibitory effects or interference with the system. All these samples came from different steps in the production, processing, purification, and filling of biopharmaceutical products. Regardless of the sample type, all the microorganisms were detected within 7 days in a reproducible and robust way. There were no differences between analysts in the total detection times for the 11 microorganisms analyzed. If a sample was contaminated, the analysts were able to detect the microbial contaminant regardless of the day of test, sample preparation, inoculum preparation, reagent's lot, inoculum level, etc. If a sample was not contaminated, the analysts did not find any microbial contamination by using either the BacT/ALERT® system, the Steritest™, or the standard TSB and FTM media.

To date the continuous on-line, non-destructive monitoring of biopharmaceutical samples represents the best-case scenario for real-time detection of microbial contamination. The system provided on-line monitoring of the inoculated bottles without any disturbance of the samples. Because the system did not rely on a visual endpoint analysis of the incubated samples, slow-growing microorganisms such as K. rhizophilla and M. radiotolerans can be detected faster than in the standard method without any subjective mistake. Optimization of the sterility testing of biopharmaceutical samples provided rapid on-line, non-destructive detection of microbial contamination during the three processing stages of biopharmaceutical products, UB, FB, and FBP, by eliminating the need of separate incubators and other supportive equipment. Daily readings and manipulations of samples were not needed because the system simultaneously incubated and analyzed the samples for the on-line detection of microbial growth. Faster detection of microbial contamination during the processing stages of a biopharmaceutical product provides a faster response for the implementation of corrective actions, leading to the optimization of process control and manufacturing.