Abstract
High-temperature/short-time (HTST) treatment of cell culture media is one of the proven techniques used in the biopharmaceutical manufacturing industry for the prevention and mitigation of media viral contamination. With the HTST method, the formulated media is pasteurized (virus-deactivated) by heating and pumping the media continuously through the preset high-temperature holding tubes to achieve a specified period of time at a specific temperature. Recently, during the evaluation and implementation of HTST method in multiple Amgen, Inc. manufacturing facilities, media precipitates were observed in the tests of HTST treatments. The media precipitates may have adverse consequences such as clogging the HTST system, altering operating conditions and compromising the efficacy of viral deactivation, and ultimately affecting the media composition and cell growth. In this study, we report the identification of the composition of media precipitates from multiple media HTST runs using combined microspectroscopic methods including Raman, Fourier transform infrared spectroscopy, and scanning electron microscopy with energy-dispersive X-ray spectroscopy. The major composition in the precipitates was determined to be metal phosphates, including calcium phosphate, magnesium phosphate, and iron (III) phosphate. Based on the composition, stoichiometry, and root-cause study of media precipitations, methods were implemented for the mitigation and prevention of the occurrence of the media precipitation.
LAY ABSTRACT: Viral contamination in cell culture media is an important issue in the biopharmaceutical manufacturing industry and may have serious consequences on product quality, efficacy, and safety. High-temperature/short-time (HTST) treatment of cell culture media is one of the proven techniques used in the industry for the prevention and mitigation of media viral contamination. With the HTST method, the formulated media is pasteurized (virus-deactivated) by heating at preset conditions. This paper provides the identification and root-cause study of the media precipitates that adversely affected the HTST process and discusses the possible solutions to mitigate the precipitation problem.
- Cell culture media
- Viral contamination mitigation
- High-temperature/short-time (HTST)
- Media precipitation
- Metal phosphates
- Root cause analysis
Introduction
Prevention of viral contamination in mammalian cell culture media is an important issue in the biopharmaceutical manufacturing industry. It is widely recognized in the industry that viral contamination may have serious consequences (1, 2), including mandatory product recalls, lot release holding, product distributing delays, customer complains, legal investigations, manufacturing plant shut-down, and loss of sales. One of the potential origins of viral entry is the raw materials that are used as the ingredients in the cell culture media. There are dozens of raw materials used as the media ingredients, and their sourcing and delivery routes may be quite diverse. It is challenging to have complete control of these raw materials in terms of their production standards and transportation safety. Therefore it is desirable to establish a methodology to mitigate the risks of media viral contamination introduced by these raw materials. One of the proven methods that has been implemented in the manufacturing facilities is to pasteurize the formulated media by high-temperature/short-time (HTST) processing (2⇓–4) prior to its use. HTST involves processing cell culture media through a thin tube at high pressure and holding the media for a specified amount of time at a specified temperature (4). The robustness and effectiveness of HTST media treatment in the deactivation of viruses such as MMV (mouse minute virus) and CVV (Cache Valley virus) have been addressed and reported (2, 4). A typical and effective HTST treatment is to heat the media to 102 °C and hold it for 10 s. Other combinations of heating temperatures and holding times may also be used (2, 4). Currently at Amgen Inc., the HTST technology is being rapidly introduced into commercial, clinical, and pilot facilities to mitigate viral contamination risks.
However, it was reported during the evaluation of HTST treatments that there was an issue of media precipitation in the process. This issue needs to be addressed because the media precipitates may have adverse consequences, such as fouling or clogging the HTST heating tube, altering operating temperatures, and compromising the efficacy of viral deactivation, and ultimately affecting the media composition and cell growth. In particular, we have observed multiple consecutive precipitating occurrences in the HTST treatment tests at our pilot plant facilities, as shown in Figure 1. It was observed that the media precipitates were usually from the heating tube and could be collected by rinsing off or scraping off the wall of the heater, shown in Figure 2, as well as from the turbid media solution after going through the HTST processing. To mitigate the media precipitation problem and prevent it from reoccurrence, we have collected and analyzed these media precipitates from multiple HTST media treatments.
Pilot plant scale HTST system used for cell culture media treatment at Amgen, Thousand Oaks, CA.
Media precipitate collection from the heating tubes of the HTST system following the media treatment.
In this study we present the detailed composition of the media precipitates, obtained by the combined microspectroscopic techniques (5⇓–7). These include Fourier transform infrared (FTIR) microscopy, Raman microscopy, and energy-dispersive X-ray spectroscopy (EDS) (6). Infrared (IR) and Raman are two branches of vibrational spectroscopy, and both can be used for molecular identification and structural analysis. The FTIR technique along with an IR spectral library search is powerful in the identification of unknown materials due to the readiness of spectral collection and the commercial availability of comprehensive IR spectral databases. However, for inorganic samples, their IR spectra are usually characterized with a few broad and less-resolved bands that are less informative for a conclusive identification. In contrast, Raman spectra of inorganic compounds can be routinely obtained in situ and in a noninvasive mode (6), and the Raman bands are much better resolved and defined so that a more conclusive spectral assignment may be achieved. However, Raman measurement for color samples is more challenging due to the interference from the fluorescence background (8), while color sample does not pose an additional problem for IR measurement. Though the fluorescence background may be reduced when using near-IR laser for excitation, the sensitivity of Raman detection is also impaired due to the nature of Raman scattering (9). As such the IR and Raman techniques complement each other very well and are often used together to solve challenging analytical problems (6). Scanning electron microscopy (SEM)/EDS are techniques that provide additional analytical power to capture details of sample morphology and to identify composition of sample elements, even in small amount of quantities. In particular, the EDS technique is able to detect the elementary differences among inorganic samples or compounds with the same structure but that are different in metal ions. In contrast, both IR and Raman techniques are not able to differentiate the elementary composition for these types of samples because they are either inactive in the common IR/Raman spectral region or they show similar spectral features. Therefore, in this study we have taken full advantages of these complementary techniques (6) for the identification of precipitate compositions.
Furthermore, the root cause of media precipitates is also investigated, and feasible solutions are discussed for mitigating or eliminating the media precipitation problem. The solutions are proposed based on the composition information of the precipitates and the media ingredients. These solutions include the modifications or replacement of media formulations as well as the changes of HTST treatment conditions. The implementation of media modifications has led to the improved performance of HTST media treatments at our multiple manufacturing facilities for cell cultures.
Materials and Methods
HTST System
The HTST system used in our pilot plant for media pilot studies is a custom-fabricated module that was designed by Amgen, as shown in Figure 1. The design team adhered to specific design principles in order to align key parameters with the previously installed commercial systems at Amgen Manufacturing Limited (AML). The heater is a steam-heated, shell-and-tube stainless steel heat exchanger with four parallel tubes in eight tube-side passes with 16 square feet of overall heat transfer area. The heating and cooling rates are aligned with the commercial systems in AML. The media is delivered through the system using an Alfa Laval SX2/013 rotary lobe positive displacement pump (Richmond, VA). After the media is heated to a temperature of 102 °C in the heater, it travels with fully turbulent flow through a stainless steel residence coil for 10 s before entering the cooler, where it is cooled to 25 °C. The system is controlled by a DeltaV system that automatically maintains the flow rate at 10 liters per minute by controlling the pump speed. The system also controls the heater and cooler outlet temperatures and the system backpressure by manipulating a control valve to ensure that the liquid media does not change state in the system at elevated temperature.
Media Precipitates
The three media precipitate samples from operations using the above HTST system were collected by our pilot plant colleagues following the completion of respective HTST runs. The first precipitate sample, registered as PS-A, was collected after processing media CM-A. The sample was obtained by centrifuging the treated media sample and decanting off the supernatant liquid. A representative fraction of the moist pellet from the bottom of the spin tube, a yellow to brown substance as shown in Figure 3a and 3b, was dried and solidified. The second precipitate sample (PS-B) was obtained after processing chemically defined media CM-B by scraping the inner wall of a tube inside the shell, a tube heat exchanger used to heat the media to 102 °C, as shown in Figure 2. Media formulation CM-B was a precursor to the production bioreactor batch media CM-C, which was later specified in the 2011 process platform. Media formulation CM-B had a higher calcium concentration than CM-C. The CM-B formulation had sodium hydroxide added at the end of the media preparation prior to HTST treatment to achieve a final pH of approximately 6.9. The third precipitate sample, PS-C, was obtained after treating an HYP (high-yield process) media in the HTST system. This sample was obtained by rinsing the inner wall of a tube inside the heat exchanger shell. In contrast to sample PS-A that was spun down from treated CM-A media, the representative media precipitate samples PS-B and PS-C collected directly from the heating tubing were fine powders, either as a light yellow powder, as shown in Figure 3b, or as a white powder, as shown in Figure 3c. The composition of the precipitates was directly analyzed with minimal sample manipulation to maintain the integrity of the samples. For this purpose, a small and representative portion of samples was transferred to a NaCl salt disc (International Crystal Laboratories, Garfield, NJ) for FTIR measurements, to a Raman slide (SpectRIM™, Tienta Sciences, Indianapolis, IN) for Raman measurements, or to a carbon adhesive tab (a common sample substrate) for SEM/EDS elementary analysis.
Optical micrographs of media precipitates following the respective HTST treatments: (a) concentrated from the turbid media solution, PS-A; (b) scraped off the tubing wall of the heater after the run was complete (the heating tubing was previously flushed with a large quantity of water), PS-B; (c) collected by rinsing off of the wall of the heater with water after the HTST run was complete, PS-C.
Optical Microscopy
The optical microscopic inspections were carried out with an optical stereomicroscope (Zeiss Stemi-2000, Thornwood, NY), and the optical micrographs of the precipitates were taken with an AxioCam MRC (Thornwood, NY) digital camera attached to the Zeiss stereomicroscope. The imaging white balance was carefully adjusted so that the color of the media precipitates was correctly captured.
FTIR Microscopy
The FTIR measurements were performed with a Hyperion 3000 IR microscope attached to a Tensor 27 FTIR spectrometer (Bruker Optics, Inc., Billerica, MA). The FTIR spectra were obtained in transmission mode from 4000 to 600 cm−1 with the precipitate samples pressed between two NaCl crystal discs in a microcompression cell. All IR spectra were collected for 128 scans at 4 cm−1 spectral resolution, and they were used for spectral library search and display without further spectral processing.
Raman Microscopy
Raman spectra were measured with a Senterra|Pr Raman microscope from Bruker Optics Inc. This Raman microscope is a combination of a confocal microscope module and a dual-laser Raman spectrometer equipped with a thermoelectrically cooled charge-coupled device detector. A near-IR (785 nm, 100 mW) laser was used for Raman excitation to reduce the interference from fluorescence background. A 20× magnification objective lens along with a 50 × 1000 μm slit aperture were used for all measurements from 70 to 3200 cm−1 at the resolution of approximately 12 cm−1. Whenever necessary, the raw Raman spectra were processed with baseline correction and/or cosmic ray removal prior to final display and spectral comparisons.
SEM and EDS
The precipitates were also imaged by SEM for the examination of their morphology at higher magnifications. The elementary composition of the precipitates was analyzed with an energy-dispersive X-ray spectrometer that is attached to the SEM system. The SEM images of the precipitates were taken with a Zeiss EVO MA 10 scanning electron microscope with the optimal filament current and electron probe current. The EDS spectra were recorded with the Oxford INCA PentaFET-x3 EDS detector (Concord, MA) at variable pressure mode with the electron beam acceleration voltage at 20 kV.
Spectral Library Search
The IR and Raman spectral library searches (6) were performed using KnowItAll|Pr Informatics System software (Bio-Rad Laboratories, Inc. Philadelphia, PA, enterprise edition, 2011) with the purpose of matching the sample spectra with the references from the respective IR or Raman databases. The first-derivative Euclidean distance algorithm was used for all spectral searches, as this algorithm is more sensitive to peak shifts and usually gives improved searching results (6). The returned search list is sorted in the order of the hit quality index (HQI), a grading system to quantify how an unknown spectrum matches a reference spectrum. The best matched spectrum is usually one of the top entries of the returned search list (though not necessary the first entry with the highest HQI) and is finally determined based on media information and complementary spectral data.
Results and Discussion
Morphology and Elementary Compositions
As mentioned above, the media precipitates from three respective HTST treatments have some differences in terms of their morphology, as shown in both optical (Figure 3) and SEM (Figure 4) micrographs, respectively. The precipitates were usually white or light yellow to brown powders when they were collected from the heating tubing, as shown in Figure 2 and Figure 3b and 3c. In another case, the precipitates were yellowish, gel-like substance concentrated from the turbid media after going through the HTST treatment, as shown in Figure 3a. It was considered that the precipitates from the turbid media were less pure and might be mixed with other nonprecipitate ingredients, likely a small amount of amino acids and sugars. Nevertheless, the EDS spectra of all precipitates, shown in Figure 4, demonstrate the presence of the four major elements: phosphorous (P), oxygen (O), calcium (Ca), and magnesium (Mg). The possible combinations of these elements suggest that the precipitates most likely consisted of calcium and magnesium phosphates, which were subsequently confirmed by their FTIR and Raman spectra, as discussed below. Interestingly, an additional element iron (Fe) was also detected in both precipitates that exhibited brown or light yellow color. This indicates that the iron compound might be accountable for the precipitate color, as both calcium and magnesium phosphates are white or colorless. In addition to the major elements mentioned above, other elements such as Na, K, Cl, and S were also presented in trace amounts in the precipitates that were concentrated from the turbid media solution, which was consistent with the precipitate origins. This is because that all four additional elements detected in the precipitate that was concentrated from the turbid media solution are normally present in the cell culture medium. It is very likely that these elements were introduced into the precipitate as the liquid medium residuals dried from evaporation. In contrast, little of these elements were observed in the precipitates that were collected directly from the heating tubing of the HTST system.
EDS spectra and SEM images of the media precipitates sampled from three different HTST runs. All EDS spectra demonstrate strong peaks from elements P, O, Ca, and Mg.
FTIR Microscopic Analysis
The FTIR analysis was performed with the gel-like, yellow to brown precipitate (Figure 3a) concentrated from the turbid media. The measured FTIR spectrum of the precipitate is shown in Figure 5 as a red trace, along with three reference spectra of iron (III) phosphate (blue trace), tricalcium phosphate (green trace), and water in KBr (black trace) extracted from the KnowItAll|Pr IR database. The three references were selected from the IR spectral library based primarily on the matches of peak positions and shapes, as well as the elementary composition from SEM/EDS analysis and the morphology of the precipitate. The spectral matching between the precipitate and the combination of three references is evident in Figure 5. It is noted that the gel-like precipitate concentrated from the turbid media solution may contain some water, thus the water bands were also present in the precipitate. In addition, magnesium and calcium are elements from the same alkaline earth metal group and they have the same valences and similar structural configurations in their phosphate compounds. Therefore their contributions to the overall IR spectrum of the precipitates are comparable; as such, only tricalcium phosphate is selected as the representative component for the following discussion. It is clear that both calcium and magnesium phosphates were not responsible for the yellow to brown color of the precipitate, as both are white. In contrast, iron (III) phosphate (or ferric orthophosphate, FePO4) and its hydrated forms have yellow to brown colors, thus this compound was determined as the color component in the media precipitates. For all the metal phosphates discussed here, their IR spectra were marked with common features (10) from the PO43− groups, which are characterized by the strong P-O stretching band around 1060 cm−1, as shown in Figure 5.
FTIR spectra of the precipitate shown in Figure 3a (red trace) and reference materials of iron (III) phosphate (blue trace), tricalcium phosphate (green trace), and water in KBr (black trace) from the KnowItAll|Pr IR databases.
It is also evident from Figure 5 that the IR bands from the phosphate groups were broad and not well resolved due, most likely, to the particle size effects. Additional sample preparation such as grinding with KBr matrix may improve the spectral quality, but this may introduce potential changes of sample integrity in addition to being more time-consuming. In contrast, the Raman bands of PO43− groups (10) are sharper and much better resolved due to the different mechanism of Raman scattering from the IR absorption. However, the Raman measurements of the yellowish precipitates concentrated from the turbid media were interfered seriously by the fluorescence background, not applicable even with the near-IR 785 nm laser used for excitation. Nevertheless, the Raman spectra of the media precipitates as white or light yellow powders collected directly from the heating tube (Figure 3b and 3c) were successfully measured, and they are discussed below.
Raman Microscopic Analysis
The Raman spectra of the precipitates are shown in Figure 6, and both spectra of the two powder precipitates have simple feature characterized by a pronounced band (10, 11) centered at 968 cm−1. This band is originated from P-O stretching mode of PO43− groups (10, 11), and this feature can be used for the determination of several standard phosphate compounds (11). It is pointed out above that calcium and magnesium phosphates are similar in terms of their ionic structures and vibrational spectral features due to the same alkaline earth metal group and valences of the two elements. Therefore in the following, we use calcium phosphates as the representative compound for Raman spectral comparison. It has been shown in a study (11) that the Raman spectra exhibit distinct differences around the P-O stretching band for four different phosphate compounds that contain different numbers of calcium ions. The reported four compounds include (a) tricalcium phosphate or calcium orthophosphate [Ca3(PO4)2], (b) calcium tribasic phosphate [Ca10(PO4)6(OH)2], (c) calcium dibasic phosphate [CaHPO4], and (d) calcium monobasic phosphate [Ca(H2PO4)2]. It is very interesting to see that the Raman spectrum of the media precipitates resembles that of tricalcium phosphate, which has only one dominant Raman band (11) at 968 cm−1, as shown in Figure 2 in the reference (11). This indicates that the calcium phosphates in the media precipitates were mainly present as tricalcium phosphates (TCPs).
TCP itself has many industrial uses as the main ingredient of phosphate fertilizers, and in glass production, and in the food industry (12). Moreover, TCP is an important medical material and is used as the active ingredient in calcium drugs as well as the bone replacement materials for bone repair, transplant, and regeneration (13⇓–15). Commonly, TCP has two isomorphs (12) of α-TCP (monoclinic space group P21/a) and β-TCP (rhombohedral space group R3c) in terms of their crystal structures. It is noted that the β-form is more stable than α-form, though their thermal stability difference was not significant based on their respective standard enthalpy of formation (α-form: −982.3 kcal/mol; β-form: −984.9 kcal/mol) (16). In addition, β-TCP is less soluble than α-TCP (14) and is widely used as synthetic bone replacement material due to its high biocompatibility and osteoconductive properties (15). In order to further determine the dominant isomorphs of the TCP precipitates, both α-TCP and β-TCP samples were obtained (Aldrich, Saint Louis, MO), and their Raman spectra were measured and displayed in Figure 7 along with that of the precipitates. By spectral comparison, both alpha and beta forms of TCP were likely present in the precipitate, though the former form might be dominant, as the position and shape of the characteristic band centered at 938 cm−1 in the precipitate spectrum bear more resemblance to that of α-TCP.
Raman spectra of the media precipitates from HTST treatments and references of alpha and beta form tricalcium phosphates.
Media Precipitation and Mitigation
The combined EDS, Raman, and IR spectral data discussed above confirmed that the cell culture media precipitates were mainly composed of metal phosphates, namely calcium, magnesium, and iron (III) phosphates. In addition, the yellow to brown color of the media precipitates were associated with the presence of iron (III) phosphates.
For the root cause analysis, we have examined the formulation ingredients used to make cell culture media. As expected, all relevant parent compounds were found and each of them is very soluble by itself in water prior to being mixed with others in aqueous solution. For example, calcium was from its nitrates and chlorides, magnesium was from its chlorides and sulfates, iron (III) was from its nitrates and citrates, and finally phosphate ions were from sodium phosphates; all of them are very soluble in water. However, when they are mixed and dissolved in water together, the circumstance was changed completely. This is because that the dissolved ions freed from their original compounds will undergo regrouping in solution according to their solubility product constants (Ksps). For example, the Ksps (17, 18) for calcium, magnesium, and iron (III) phosphates at 25 °C are 2.07 × 10−33, 1.04 × 10−24, and 1.3 × 10−22, respectively, and they are so small as compared to, for example, calcium chloride (Ksp is greater than 3 × 102 at 20 °C). As such, when positive metal ions (Ca2+, Mg2+, and Fe3+) encounter negative phosphate ions (PO43+) in the solution, they have great propensity to form calcium, magnesium, and iron (III) phosphates, respectively, as they are essentially insoluble in water and will precipitate out. This is just like a common precipitation reaction where the reaction may easily go to completion as long as the products are much less soluble than the reactants. In fact, in the formulation of culture media, the concentrations of the above ingredients are controlled to be low so that when they mixed in aqueous solutions, the precipitation does not occur or is negligible at room temperatures. However, during the HTST treatments, the conditions were changed as the media was heated to high temperature (102 °C) and forced to pass through the long and thin holding tube. Unlike most other compounds, calcium phosphate is increasingly insoluble at higher temperatures (19), thus heating may cause it to precipitate. This is considered as the root cause that the calcium phosphates were present in all submitted media precipitates and were most significant in terms of their relative quantities. In contrast, iron (III) phosphate has a larger Ksp compared to calcium and magnesium phosphates, therefore it had the least amount of qualities in the precipitates, and in the last precipitate no iron was even detected. The root cause of precipitation for magnesium and iron (III) phosphates, however, might not be directly associated with heating but rather with the local concentration changes inside the heater from the nonsmooth flow (jamming) of the media caused by calcium phosphate precipitates.
Based on the above analysis, the feasible methods to mitigate the media precipitations in the HTST treatment were evaluated, and they are outlined below. First, it is helpful to further reduce the concentrations of the affected media ingredients that may lead to the combinations of metal phosphate, which are insoluble in water. In particular, the calcium ingredients should be minimized because very likely they will precipitate as phosphates first due to their smaller Ksps compared with other precipitates and their unusual temperature dependence. Moreover, the calcium phosphates precipitate may jam the heating tubes and result in more turbulences of media flow, thus it might be accountable for the precipitation of other components such as magnesium and iron (III) phosphates. Second, the pH value of the media may be adjusted to increase its acidity before the HTST treatment because calcium, magnesium, and iron phosphates are much more soluble in acidic solutions. Thus acidic media is helpful in the prevention of precipitation during the HTST treatment. The required pH value in the final media may be restored by adding NaOH solution right after the completion of the HTST process. The NaOH solution added after HTST processing shall be of sufficient concentration to inactivate virus due to its high pH, obviating the need for the HTST treatment of NaOH itself. The last but more drastic method is to remove some or all affected ingredients completely and develop a novel media for HTST treatment. Upon the completion of the analysis of media precipitate and root cause study, all three methods discussed above for mitigating the HTST media precipitation issue were being tested and evaluated at our development facilities, and significant progress has been achieved towards improved HTST performance.
Conclusions
In this work, we have first identified the composition of media precipitates from multiple cell culture media HTST runs. The major compositions in the precipitates were metal phosphates, including TCP, trimagnesium phosphate, and iron (III) phosphate. All of them have very small Ksps, and they are essentially insoluble in water. Particularly for TCP, not only is its Ksp the smallest among the three components but it is also even less soluble at elevated temperatures and may also be responsible for the precipitations of magnesium and iron (III) phosphates. Based on the understanding of the composition and root causes of media precipitations, a number of mitigation methods were implemented to prevent the occurrence of media precipitation. The common goals of these methods are either the reduction of the interactions between the positive metal ions (Ca2+, Mg2+, and Fe3+) and the negative phosphate ions (PO43+) or the suppression of precipitating propensity by changing chemical conditions. At present, much improved performance has been obtained for our new HTST treatments based on the combination of modified media formulation, chemical conditions, and process designs.
Acknowledgements
We thank the cell culture media HTST team for their helpful and inspiring discussions. Drs. Ping Yeh and Joseph Phillips are gratefully acknowledged for their support in cell culture media projects and technology development. Drs. Linda Narhi and David Brems are thanked for their comments and revisions in the preparation of this paper. Dr. Gary Li from forensic analysis group is appreciated for his helpful reviews and data verification in the preparation of internal technical reports.
- © PDA, Inc. 2013
