Prediction of long-term polysorbate degradation according to short-term degradation kinetics

ABSTRACT

Polysorbates (PSs) are a class of surfactants commonly used in the formulation of protein therapeutic agents to provide protection against denaturation and aggregation. When the PS in these drug formulations degrades, loss of stabilization of the protein therapeutic and formulation may occur, resulting in particulate formation or other undesirable changes in product critical quality attributes. Here, we present a simplified platform to predict long-term PS20 and PS80 degradation for monoclonal antibody drugs containing the PS-degrading enzyme lysosomal acid lipase. The platform was based on a temperature-dependent equation derived from existing PS20 degradation stability data. Accurate prediction of both PS20 and PS80 hydrolysis for as long as 2 years was achieved through short-term kinetics studies performed within 2 weeks. This platform substantially shortens the time required to determine the long-term stability of PS degradation and therefore can be used to guide the purification process and optimization of antibody formulations.

KEYWORDS:

• Free fatty acids
• lipase activity
• long-term stability
• polysorbate hydrolysis
• short-term kinetics

Introduction

Polysorbates (PSs) are surfactants that have a high hydrophilic–lipophilic balance and low critical micellar concentration and are widely used in therapeutic products. Aggregation and denaturation of protein therapeutic agents can be prevented by the addition of PS to drug products (DPs), thereby enhancing DP stability.^1–3 However, PS is prone to oxidation and hydrolysis, and thus is susceptible to degradation.^3–8 PS degradation can lead to undesirable particle formation and unstable DPs.^8–15

Enzymatic hydrolysis is a major cause of PS degradation in recombinantly expressed proteinaceous DPs.^1–3,5,16 Host cell proteins (HCPs), including PS-degrading enzymes, from the Chinese hamster ovary (CHO) cell lines typically used to produce clinically relevant recombinant proteins may co-purify with DP proteins, and thus may be present in trace amounts in the DPs. The enzymes that degrade PS are usually categorized as lipases or esterases, which catalyze cleavage of the ester bond of PS and release free fatty acids (FFAs) as a byproduct.^4,6,16–24 Some FFAs are insoluble in drug formulations, and therefore precipitate and form visible or subvisible particles.^{7,8,11,12,14,15} The released FFAs can also induce the formation of proteinaceous particles, as suggested in a recent study by Zhang et al. ²⁵ Furthermore, the loss of PS can result in protein aggregation in DPs or during intravenous admixture, thereby negatively affecting DP quality.^7,8

Several enzymes from CHO cell lines capable of degrading PS have been detected and reported, including lipoprotein lipase (LPL), lysosomal acid lipase (LAL), group XV lysosomal phospholipase A2 isomer X1 (LPLA2), liver carboxylesterase, sialate O-acetylesterase (SIAE) and palmitoyl-protein thioesterase 1 (PPT1).^{1,4,6,16–23,26–28} The activity of each lipase markedly differs; for example, lipases/esterases such as LPLA2 and carboxylesterase cause substantial PS degradation at levels below parts per million (ppm),^20,23,29 whereas other lipases/esterases cause PS degradation at ppm levels or higher.^17,19 SIAE targets only PS20 but not PS80,^16,18 whereas other enzymes degrade both PS20 and PS80. Despite their different activities, these PS-degrading enzymes are commonly identified in DPs from different cell lines prepared through different processes used across the biopharmaceutical industry.^26,27 Therefore, understanding the properties of each individual lipase/esterase would be helpful and have broad applications.

A variety of methods have been developed to track PS changes in formulated drug substances and DPs as part of drug quality assurance. PSs can be characterized and quantified through direct measurement with a liquid chromatography-charged aerosol detector (LC-CAD) or an evaporative light scattering detector.^30–35 PS degradation can also be monitored by tracking changes in degradation byproducts, such as FFAs from hydrolysis,^9,13,14,36 aldehydes derived from FFA esters, and other byproducts derived from oxidation.^3,7,37,38 Lipase substrate-based enzyme activity assays also provide information on whether the PS in a DP is subject to degradation.^1,4,6 Monitoring of released FFAs, compared with direct measurement of PS or lipase substrate-based enzyme activity assays, is a more sensitive and accurate method of detecting PS hydrolysis, as described Zhang et al. ⁹ In an accelerated PS stability study, a 0.0002% decrease in PS20 or PS80 degradation can be detected within 1 day by monitoring the increase in lauric acid or oleic acid concentration, thus enabling early detection of lipase activity in DPs.⁹

Although an accelerated study can provide useful information on whether PS will be degraded, the acceptable level of PS degradation that supports the targeted long-term shelf life of DPs remains unclear.¹ In this study, we generated an equation for converting PS degradation levels in accelerated studies (37°C) to the PS degradation levels in long-term stability studies (4–8°C), on the basis of LAL-catalyzed PS20 degradation. We first quantified PS degradation by monitoring released FFAs in both stability and accelerated studies. We then developed an equation to convert degradation measured at 37°C over the course of days to that at 5°C over the course of months. The conversion equation was then used to calculate the level of PS degradation in a different stability study, and the results matched the measured levels well. Therefore, our method enables prediction of PS degradation during the refrigerated shelf life of DPs through measurement of FFA levels in accelerated studies.

Results

Host cell protein analysis of monoclonal antibodies

HCP analysis was performed for all six monoclonal antibodies (mAbs) used in this study (Table 1). A total of 100–250 HCPs were identified for each mAb with the ultra-sensitive ProteoMiner with limited digestion method, with a detection limit as low as 0.002 ppm (data not shown). Table 1 summarizes the identified lipases and their quantification in each of these six mAb samples. For mAb-1 process A, mAb-1 process B, mAb-2 process A, and mAb-2 process B, both LAL and lipoprotein lipase were identified. However, a subsequent study indicated that LAL was the only lipase responsible for PS degradation in mAb-1 and mAb-2. This conclusion was based on two previously reported findings^17: (1) the degradation of both PS20 and PS80 is completely inhibited by the LAL-specific inhibitor Lalistat-2, and (2) the relative concentration of LAL and the PS degradation level are positively correlated. In mAb-3, three lipases – PPT1, LAL and SIAE – were identified; however, which lipase was primarily responsible for PS80 degradation was unclear, given that both LAL and PPT1 may contribute to PS80 degradation. In mAb-4, SIAE was the only lipase responsible for PS20 degradation.

Table 1. Buffer and lipase information of mAb-1 process A and process B, mAb-2 process A and process B, mAb-3 and mAb-4.

Sample information	Buffer	mAb subtype	Lipase identified	Lipase Conc. (ppm)
mAb-1 process A	10 mM Histidine, 5% Sucrose, 70 mM Arginine-HCl, 0.05% polysorbate 20, pH 6.3	IgG4	LAL	.3
			LPL	1.4
mAb-1 process B	10 mM Histidine, 5% Sucrose, 70 mM Arginine-HCl, 0.05% polysorbate 20, pH 6.3	IgG4	LAL	.1
			LPL	0.8
mAb-2 process A	10 mM L-histidine, 8% (w/v) sucrose, and 0.1% (w/v) polysorbate 80, pH 6.0	IgG1	LAL	1.2
			LPL	0.5
mAb-2 process B	10 mM L-histidine, 8% (w/v) sucrose, and 0.1% (w/v) polysorbate 80, pH 6.0	IgG1	LAL	2.0
			LPL	0.2
mAb-3	25 mM L-arginine hydrochloride, 20 mM L-histidine, 12.5 mM sodium acetate, 5% (w/v) sucrose, 0.2% (w/v) polysorbate 80, pH 5.9	IgG4	LAL	.4
			PPT-1	1.9
			SIAE	1.7
mAb-4	10mM Acetate, 25mM Arginine, 5% Sucrose pH5.3	IgG4	SIAE	1.0

Establishing the PS20 degradation conversion correlation between 5°C and 37°C for mAb-1 process A

Establishing the correlation between the accelerated degradation conditions and long-term storage conditions was key for successful prediction. To this end, we incubated 60 mg/mL mAb-1 process A with 0.05% PS20 at 5°C for 12 months and collected an intermediate time point sample at 6 months. For accelerated degradation conditions, 200 mg/mL mAb-1 process A was incubated with 1% PS20 at 37°C for 6 days, and an intermediate time point sample was collected at day 3 for accelerated study. 60 mg/mL mAb-1 process A was chosen for the long-term stability study to match the concentration of final DP and 200 mg/mL mAb-1 process A was chosen for the short-term stability study to accelerate the PS20 degradation reaction. The formulation of mAb-1 remained the same regardless of the protein concentration. The free lauric acid concentration due to PS20 degradation was plotted against the incubation time. A linear regression line with an equation of y = 0.8233 × D_37C −0.08 and y = 0.3285 × M_5C −0.072 was observed for PS20 degradation at 37°C (Figure 1a, Equation1a) and 5°C (Figure 1a, Equation 2a), respectively. The conversion of the increased lauric acid concentration at 5°C over the course of months (M_5C) and at 37°C over the course of days (D_37C) was established by equalizing Equation 1a and Equation 2a. The resultant equation (Equation 3a below) can be used to convert the PS20 degradation determined at 37°C to that at 5°C for mAb-1 process A.

Figure 1. (a). Increased released lauric acid due to PS20 degradation at 37ºC in days (D_37C) for mAb-1 process A (200 mg/mL formulation, Eq.1) and 5ºC in months (M_5C) for mAb-1 process A (60 mg/mL formulation, Eq.2). Dark blue: mAb-1 process A (200 mg/mL formulation) incubated at 37ºC up to 6 days; light blue: mAb-1 process A (60 mg/mL formulation) incubated at 5ºC up to 12 months. Equation M_5C= 2.51× D_37C-0.02 (Eq. 3) was established by equalizing Eq.1 and Eq.2. (b). Estimated relationship between increased lauric acid concentration and incubation time M_5C for mAb-1 process B (60 mg/mL formulation). Increased lauric acid was measured for mAb-1 process B (200 mg/mL formulation) at 37ºC at day 0, 3, 6. Incubation time M_5C was converted from D_37C by Eq.3. (c). Comparison of the released lauric acid concentration due to PS20 degradation in mAb-1 process B (60 mg/mL formulation) in M_5C measured by LC-MRM with the predicted value. (d). Increased released oleic acid due to PS80 degradation at 37ºC in days (D_37C) for DS-1 process A (120 mg/mL formulation, Eq.1b) and 5ºC in months (M_5C) for DS-1 process A (120 mg/mL formulation, Eq.2b). Dark blue: DS-1 process A (120 mg/mL formulation) incubated at 37ºC up to 7 days; light blue: DS-1 process A (120 mg/mL formulation) incubated at 5ºC up to 6 months. Equation M_5C= 1.92× D_37C + 0.162 (Eq. 3b) was established by equalizing Eq.1b and Eq.2b. (e). Estimated relationship between increased oleic acid concentration and incubation time M_5C for DS-1 process B (150 mg/mL formulation) based on data in Table 2. Increased oleic acid was measured for DS-1 process B (200 mg/mL formulation) at 37ºC at day 0, 3 and 7 and converted to 150 mg/mL formulation by multiply 0.75. Incubation time M_5C was converted from D_37C by Eq.3b. (f). Comparison of the released oleic acid concentration due to PS80 degradation in DS-1 process B (150 mg/mL formulation) in M_5C measured by LC-MRM with the predicted value. Blue column: measured value; Orange column: predicted value. Duplicates were performed for mAb-1 process B stability samples.

M_5C = 2.51 × D_37C −0.02(3a)

Prediction of the long-term PS degradation profile at 5°C, on the basis of short-term kinetic data from accelerated (37°C) study

To demonstrate the feasibility of this method, we selected mAb-1 process B as a verification sample. The lipase responsible for PS20 degradation in mAb-1 process A was also found in mAb-1 process B, at a slightly lower concentration (Table 1). To test whether the conversion equation derived from mAb-1 process A could also be applied to mAb-1 process B, we incubated 200 mg/mL of mAb-1 process B at 37°C for 6 days and collected a sample at an intermediate time point at day 3. Lauric acid concentrations were measured at days 0, 3 and 6. The incubation time of D_37C was converted to M_5C with Equation 3a obtained from mAb-1 process A. The measured lauric acid concentration from accelerated study (37°C) was plotted against the converted incubation time M_5C to establish the PS20 degradation kinetics (Equation 4a) for mAb-1 process B (Figure 1b).

The estimated released lauric acid for 60 mg/mL of mAb-1 process B at 4–8°C was then calculated with Equation 4a at 0, 6, 12, 18 and 24 months, and compared with the measured values for mAb-1 process B from the stability study. As shown in Figure 1c, the calculated lauric acid concentration from a 6-day experiment accurately matched the measured value from a 24-month stability study, with less than 20% variation (less than 0.001% difference in PS20% change based on Equation 10).

The conversion equation that translates a 37°C kinetic to a 5°C kinetic was not limited by the concentration of DP in long- and short-term stability studies or whether the drug is formulated with PS20 or PS80. In another example, 120 mg/mL of DS-1 process A was formulated with 0.1% PS80 and used to conduct long-term and short-term stability studies. The free oleic acid concentration due to PS80 degradation was plotted against the incubation time. A linear regression line with an equation of y = 7.4505 × D_37C −0.2151 and y = 3.8696 × M_5C −0.8379 was observed for PS80 degradation at 37°C (Figure 1d, Equation 1b) and 5°C (Figure 1d, Eq 2b), respectively. The conversion of the increased oleic acid concentration at 5°C over the course of months (M_5C) and at 37°C over the course of days (D_37C) was established by equalizing Equation 1b and Equation 2b. The resultant equation (Equation 3b(3b) ${\mathrm{M}}_{5\mathrm{C}}=1.92\times {\mathrm{D}}_{37\mathrm{C}}+0.162$(3b) below) can be used to convert the PS80 degradation determined at 37°C to that at 5°C for DS-1 process A where long- and short-term stability studies were conducted at the same concentration and formulation.

(3b) ${\mathrm{M}}_{5\mathrm{C}}=1.92\times {\mathrm{D}}_{37\mathrm{C}}+0.162$(3b)

Equation 3b(3b) ${\mathrm{M}}_{5\mathrm{C}}=1.92\times {\mathrm{D}}_{37\mathrm{C}}+0.162$(3b) can be used to convert incubation time for DS-1 process B from D₃₇ to M_5C as long as the long- and short-stability studies were conducted at the same concentration and formulation. For DS-1 process B, the long- and short-term stability studies were conducted at 150 mg/mL and 200 mg/mL, respectively. To equalize the concentration used in the long- and short-term stability studies, we calculated how much oleic acid would increase if the accelerated study was carried out at 150 mg/mL by multiplying the measured oleic acid concentration increase at 200 mg/mL by 0.75 (Table 2, column 3). The calculated oleic acid increase at 150 mg/mL was then plotted against the incubation time M_5C which can be calculated by Equation 3b(3b) ${\mathrm{M}}_{5\mathrm{C}}=1.92\times {\mathrm{D}}_{37\mathrm{C}}+0.162$(3b) (Table 2, column 4) to establish the PS80 degradation kinetics (Eq 4b) for DS-1 process B (Figure 1e).

Table 2. Converting measured oleic acid concentration increase at 37°C for DS-1 process B with 200 mg/mL formulation to calculated oleic acid concentration increase at 37°C for a 150 mg/mL formulation. Incubation time in M_5C was converted from D_37C by Equation 3b(3b) ${\mathrm{M}}_{5\mathrm{C}}=1.92\times {\mathrm{D}}_{37\mathrm{C}}+0.162$(3b) (M_5C = 1.92× D_37C +0.162).

DS-1 process B incubation time (D37C)	Measured oleic acid concentration (200 mg/mL formulation) (µg/mL)	Calculated oleic acid at 150 mg/mL formulation) (µg/mL)	Converted incubation time (M5C)
0	0.00	0.00	−0.08
3	22.43	16.82	1.47
7	59.69	44.77	3.55

The estimated released oleic acid for 150 mg/mL DS-1 process B at 4–8°C was then calculated with Equation 4b at 0, 3, 6, 9 and 12 months, and compared with the measured values for DS-1 process B from the stability study. As shown in Figure 1f, the calculated oleic acid concentration from a 7-day experiment accurately matched the measured value from a 12-month stability study, with less than 10% variation (corresponding to less than 0.005% difference in PS80% change).

Establishment of the long-term PS80 degradation profile at 5°C, on the basis of short-term kinetic data from accelerated study of PS20 for mAb-1

Many DPs are formulated with either PS20 or PS80 surfactants. Using existing PS20 stability data to predict both PS20 and PS80 degradation, or vice versa, would be ideal. If both PS20 and PS80 are degraded by a single enzyme (LAL), and the PS degradation is proportional to the enzyme level, we reasoned that the conversion equation for PS20 and PS80 might be correlated with an adjustment factor.

To determine whether a correlation existed between the LAL-catalyzed PS20 and PS80 degradation rates, we incubated 5 µg/mL LAL with 1% PS20 or 1% PS80 for as many as 12 days and collected intermediate time point samples at days 2, 5 and 8. The levels of lauric and oleic acid due to PS20 and PS80 degradation were determined for each time point (Supplementary Table S1) and are plotted against each other in Figure 2a. A strong correlation was observed between the increased lauric acid concentration and increased oleic acid concentration, thus yielding Equation 5(5) $Y=1.2085\mathrm{x}+15.55$(5) :

Figure 2. Estimate PS80 degradation in long-term stability study based on PS20 degradation data. (a). 5 µg/mL LAL was incubated with 1% PS80 and 1% PS20 at 37ºC respectively for up to 12 days. Positive correlation between released oleic acid due to PS80 degradation and released lauric acid due to PS20 degradation was established with R² over 0.98 (data in Supplementary Table 1). (b). 0, 0.1, 0.5, 1, 2 5 µg/mL LAL was spiked in mAb-2 process A and incubated with 0.1% PS80 at 37ºC for 5 days. Released oleic acid was positively related to spiked-in LAL concentration. (c). Calculated increased released oleic acid due to PS80 degradation at 37ºC in days for mAb-1 process A (120 mg/mL formulation, Eq.6) based on Table 3-a and 5ºC in months for mAb-1 process A(120 mg/mL formulation, Eq.7) based on Table 3-b. Blue dot: mAb-1 process A (120 mg/mL formulation) at 37ºC up to 6 days; Orange dot: mAb-1 process A (120 mg/mL formulation) at 5ºC up to 12 months. Equation M_5C= 0.752 × D_37C+ 0.14 (Eq. 8) was established from equation of Eq.6 and Eq.7. (d). Estimated relationship between increased oleic acid concentration and incubation time in month @ 5ºC for mAb-2 process B based on data in Table 3-a. Increased oleic acid was measured for mAb-2 process B at 37ºC at 0, 3, 4 day. Incubation time was converted from day to month by Eq.8. (e). Comparison of the rate of released oleic acid due to PS80 degradation in mAb-2 process B at 5ºC in months measured by LC-MRM with the predicted value by Eq.9. Blue column: measured value; Orange column: predicted value.

(5) $Y=1.2085\mathrm{x}+15.55$(5)

With the obtained linear regression equation (Equation 5(5) $Y=1.2085\mathrm{x}+15.55$(5) ), we were able to convert LAL-catalyzed PS20 degradation measured at 37°C, represented by increased lauric acid concentration, to LAL-catalyzed PS80 degradation, represented by increased oleic acid concentration. We performed this procedure with mAb-1 at 200 mg/mL and 60 mg/mL, wherein the lauric acid concentrations measured at 37°C and 5° were converted to oleic acid concentrations with Equation 5(5) $Y=1.2085\mathrm{x}+15.55$(5) (Table 3, column 3).

Table 3. Converting measured lauric acid concentration increase at 37°C for mAb-1 process A with 200 mg/mL formulation to calculated oleic acid concentration increase at 37°C for a 120 mg/mL formulation using Equation 5(5) $Y=1.2085\mathrm{x}+15.55$(5) and multiplication factor.

mAb-1 process A Incubation time (D37C)	Measured Lauric acid (200 mg/mL formulation) (µg/mL)	Calculated oleic acid (200 mg/mL formulation) (µg/mL)	Calculated oleic acid (120 mg/mL formulation) (µg/mL)	Calculated oleic acid increase (120 mg/mL formulation) (µg/mL)
0	0.37	16.00	9.60	0.00
3	2.60	18.69	11.22	1.62
6	5.31	21.97	13.18	3.58

We also performed an enzyme spike-in experiment to determine the correlation between LAL concentration and oleic acid release. LAL concentrations of 0, 0.1, 0.5, 1, 2 or 5 µg/mL were spiked in mAb-1 process A, and the samples were then incubated with 0.1% PS80 at 37°C for 5 days. A strong positive correlation was established between the spiked-in LAL concentration and the increased oleic acid concentration (Figure 2b). Because the LAL concentration in the DPs is proportional to the protein concentration, the protein concentration conversion could be determined simply by multiplication or division between the concentration of mAb-1 and the targeted concentration. As shown in Supplementary Figure S1, we calculated the lauric acid concentration in the 200 mg/mL formulation by multiplying 200/60 of the increased lauric acid concentration in the 60 mg/mL formulation. The calculated lauric acid concentration matched well with the measured value. Therefore, a concentration correction factor of 120/200 and 120/60 was applied to obtain the calculated oleic acid concentration for the 120 mg/mL formulation at 37°C and 5°C, respectively (Tables 3 and 4, column 4). The calculated oleic acid increase at 37°C and 5°C showed a linear relationship with incubation time (Equation 6 and Equation 7 in Figure 2c). By equalizing these two linear equations, we obtained Equation 8 (M_5C = 0.752 × D_37C +0.14), which converts PS80 degradation at 37°C in the 120 mg/mL formulation to 5°C in the 120 mg/mL formulation.

Table 4. Converting measured lauric acid concentration increase at 5°C for mAb-1 process A with 60 mg/mL formulation to calculated oleic acid concentration increase at 5°C for a 120 mg/mL formulation using Equation 5 and multiplication factor.

mAb-1 process A Incubation time (M5C)	Measured Lauric acid (60 mg/mL formulation) (µg/mL)	Calculated oleic acid (60 mg/mL formulation) (µg/mL)	Calculated oleic acid (120 mg/mL formulation) (µg/mL)	Calculated oleic acid increase (120 mg/mL formulation) (µg/mL)
0	0	15.55	31.10	0
6	1.76	17.68	35.35	4.25
12	3.94	20.31	40.62	9.52

Prediction of the long-term PS80 degradation profile at 5°C, on the basis of short-term kinetic data from accelerated study of PS20 for different mAbs containing the same lipase

Because PS degradation is driven by trace levels of lipase, we further examined whether the correlations and conversions obtained with mAb-1 containing a trace level of LAL might potentially be applied to a different antibody (mAb-2) preparation also containing trace levels of LAL (Table 1).

The relationship between the oleic acid increase and incubation time at 5°C for the mAb-2 process B was established by plotting the increased oleic acid measured on days 3 and 4 at 37°C from the accelerated study against the incubation time in months, converted by Equation 8 (Table 5). As shown in Figure 2d, Equation 9 was established to estimate the long-term stability of PS80 in mAb-3 at 5°C.

Table 5. Establishment of the rate of released oleic acid due to PS80 degradation in mAb-2 process B (120 mg/mL formulation) at 5°C in months.

Incubation time (D37C)	Measured oleic acid (µg/mL)	Converted incubation time (M5C)
0	0	0.15
3	10.17	2.41
4	12.65	3.16

The estimated oleic acid increase for the mAb-2 process B at 5°C was then calculated with Equation 9 at 0, 3, 4.3 and 6 months, and compared with the measured values from the stability study (Table 6). As shown in Figure 2e and Table 6, the calculated increased oleic acid concentration from a 4-day experiment accurately matched the measured values from a 6-month stability study.

Table 6. Comparison of the concentration of released oleic acid due to PS80 degradation in mAb-2 process B (120 mg/mL formulation) at M_5C measured by LC-MRM with the estimated value.

Incubation time (M5C)	Measured oleic acid (µg/mL)	Estimated oleic acid (µg/mL)	Accuracy
0	0	−0.54	-
3	11.30	12.27	7.93%
4.3	15.65	17.82	12.21%
6	22.99	25.08	8.36%

Conversion of released FFA concentration to percentage of PS remaining

Released FFAs are well known to be positively correlated with the amount of PS degradation, but the ability to directly convert the released FFA concentration to the percentage of degraded PS would aid in understanding the results from fatty acid measurements. Therefore, we measured the lauric acid concentration increase and percentage of PS20 remaining for mAb-1 process A (200 mg/mL formulation) stability samples at 0, 6, 12, 18 and 24 months at 5°C, and established equation (Equation 10) to convert the released lauric acid concentration to the percentage of PS20 degraded (Figure 3, left panel). Similarly, we measured the oleic acid concentration increase and the percentage of PS80 remaining for mAb-3 stability samples at every 3 months from 3 to 36 months at 5°C, and established equation (Equation 11) to convert the released oleic acid concentration to the percentage of PS80 degraded (Figure 3, right panel).

Figure 3. Left panel: mAb-1 process A (200 mg/mL formulation) was incubated with 0.05% PS20 at 5ºC for up to 24 months. Intermediate time point was taken at 6, 12, 18 months. Degraded PS20% was measured by LC-CAD and increased lauric acid concentration was measured by LC-MRM for each time point. Degraded PS20% was plotted against increased lauric acid concentration as Eq. 10. Right panel: mAb-3 was incubated with 0.2% PS80 at 5ºC for up to 36 months. Degraded PS80% was measured by LC-CAD and increased oleic acid concentration was measured by LC-MRM for each time point. Degraded PS80% was plotted against increased oleic acid concentration as Eq. 11.

Polysorbate hydrolysis kinetics

The kinetics of enzymatically mediated PS degradation follows the Michaelis–Menten equation $v=\frac{\mathrm{V}\mathrm{m}\mathrm{a}\mathrm{x}\left[\mathrm{S}\right]}{\mathrm{K}\mathrm{M}+\left[\mathrm{S}\right]}$. As shown in Figure 4, the PS20 degradation rate remained constant, on the basis of the Vmax in the first 12 months at 5°C for mAb-1, when the percentage of PS20 remaining was above 0.03% (below 17.8 µg/mL lauric acid released from PS20 degradation; left panel), and the PS80 degradation rate remained constant, on the basis of the Vmax in the first 15 months for mAb-3 at 5°C, when the percentage of PS80 remaining was above 0.1% (below 171 µg/mL oleic acid released from PS80 degradation; right panel). The reaction rate began to decrease when the remaining percentage of PS20 decreased below 0.03%, or that of PS80 decreased below 0.1%, and showed a steep decline with decreasing substrate concentration when the percentage of PS20 remaining decreased below 0.022%, or the percentage of PS80 remaining decreased below 0.078%.

Figure 4. Left panel: mAb-1 process A (200 mg/mL formulation) incubated with 0.05% PS20 at 5ºC for up to 36 months. Increased lauric acid concentration was plotted against incubation time. Lauric acid increased linearly in the 12 months when the remaining PS20% was above 0.03%. Reaction rate decreased significantly after 24 months when remaining PS20% was below 0.022% Right panel: mAb-3 incubated with 0.2% PS80 at 5ºC for up to 36 months. Increased oleic acid concentration was plotted against incubation time. Oleic acid increased linearly in the 15 months when the remaining PS80% was above 0.1%. Reaction rate decreased significantly after 24 months when the remaining PS80% was below 0.078%.

We assumed that the K_M of PS degradation by LAL was relatively small with respect to the initial substrate concentration, which usually ranges between 0.05% and 0.2% PS in DPs. Therefore, at the beginning of the incubation, the reaction rate (v) was considered to be constant and independent of the substrate concentration [S], and the increased FFA concentration plotted against incubation time was linear. Linearity of the plot began to be lost only when the percentage of PS remaining had substantially decreased and became comparable to the K_M, where v began to be dependent on the substrate concentration and to decrease over time. The possible reasons for the decreased reaction rate included substrate (PS) consumption, product inhibition (FFA accumulation) or decreased enzyme activity. In our case, the reaction plot was no longer linear when the percentage of PS20 remaining decreased to 0.024%, or the percentage of PS80 remaining decreased to 0.078%, at 24 months.

Therefore, the prediction platform indicated the best performance when the lipase activity was relatively low, as in mAb-1 process A (60 mg/mL formulation) and mAb-1 process B (60 mg/mL formulation, Figure 1c), wherein the percentage of PS20 remaining was above 0.024% at 24 months, or when the DP was formulated with a higher percentage of PS. The prediction platform remained applicable in the early incubation stage for a DP with high lipase activity, as in mAb-1 process A (200 mg/mL formulation) or mAb-3 (Figure 4), but overestimated the PS degradation at later time points.

In contrast, the PS degradation rate was constant for at least 21 days in the accelerated condition. As shown in Figure 5a, the reaction rate remained unchanged with 0.1%, 0.2%, 0.5% and 1% PS20 present in the DP, thus suggesting that Vmax had been reached even with 0.1% PS20 present, in agreement with our findings from the stability study. In the accelerated study, protein samples were incubated with 1% PS, a much higher concentration than the 0.024% PS20 or 0.078% PS80; thus, the increased FFA concentration plotted against incubation time remained linear for an extended period of time, until more than 90% of the PS had degraded. As shown in Figure 5b,c the linear plot fitted perfectly for both mAb-1 process A (200 mg/mL formulation) and mAb-3 after as many as 42 days at 37°C.

Figure 5. Reaction rate remained the same with different substrate concentration. (a). mAb-1 process A (200 mg/mL) with ISTD incubated with 0.1%, 0.2%, 0.5% and 1% PS20, incubated for up to 6 days at 37ºC. Increased lauric acid was measured by LC-MRM and plotted against incubation time. (b). mAb-1 process A (200 mg/mL) was incubated with 1% PS20 at 37ºC for up to 21 days. Increased lauric acid was measured by LC-MRM and plotted against incubation time. (c). mAb-3 was incubated with 1% PS80 at 37ºC for up to 42 days. Increased oleic acid was measured by LC-MRM and plotted against incubation time.

Prediction of long-term PS stability based on accelerated degradation is enzyme specific

We also assessed whether the equation for converting PS20 degradation at 5°C over a period of months and 37°C in over a period of days for mAb-1 could be applied to DPs containing other lipases or esterases. As shown in Figure 6, the conversion factor from D_37C to M_5C was 0.83 in mAb-2 process B (left panel), whereas the conversion factor was 1.43 from D_37C to M_5C in mAb-3 (right panel). The lipase that caused PS80 to degrade in mAb-2 was LAL, whereas multiple lipases, including PPT1, LAL and lipoprotein lipase (LPL), potentially contributed to PS80 degradation in mAb-3. We concluded that the relationship between 5°C and 37°C is enzyme-specific. However, a platform that applies the temperature-dependent kinetics from one set of stability samples to other DPs containing the same lipase/esterase would still enable substantial shortening of the time required for risk assessment of PS degradation for a new program, because DPs produced from the same cell line and with the same downstream processing procedures usually share similar lipase/esterase profiles.

Figure 6. Left panel: Released oleic acid due to 0.1% PS80 degradation at 5ºC in month and 37ºC in days for mAb-2 process B. The lipase/esterase in mAb-2 process B is lysosomal acid lipase (LAL). Equation M_5C= 0.83 × D_37C + 0.24 (Eq. 16) was established by equation of Eq.14 and Eq.15. Right panel: Increased released oleic acid due to 0.1% PS80 degradation at 5ºC and 37ºC in month for mAb-3. The lipase/esterase in mAb-3 includes palmitoyl-protein thioesterase 1 (PPT1), lysosomal acid lipase (LAL) and Sialate O-acetylesterase (SIAE). Equation M_5C= 1.43 × D_37C -0.39 (Eq. 19) was established by equation.

LAL is a lipase commonly detected in mAbs produced from CHO cell lines, and PS degradation by LAL has been reported by authors from several companies.^{16,17,19,27,29} Therefore, although the equation determined for PS20 degradation in this study can be applied to only PS20 and PS80 degradation by LAL, our results should have broad industrial applications. Notably, not all lipases/esterases show the same degradation activity toward PS20 and PS80. For example, SIAE is an esterase that targets PS20 but not PS80; therefore, the equation based on PS20 degradation in DPs containing SIAE cannot be applied to PS80 degradation. As shown in Figure 7, left panel, mAb-4 is an antibody drug containing SIAE only, and only PS20 degradation was observed; thus, the equation used to estimate PS20 degradation could not be applied to PS80 degradation in mAb-4. Evaluation of whether certain lipases or esterases show similar trends in degradation of PS20 and PS80 must be performed before application of the platform. We surveyed PS20 and PS80 degradation in eight different mAbs with different polysorbate degrading enzymes (PSDEs) presented. These eight mAbs are distinct samples produced through processes different from those used for mAb-1 to mAb-4. As shown in Figure 7, right panel, PS20 degradation and PS80 degradation were not always related to each other. PS20 degradation was related to PS80 degradation in five of eight mAbs, whereas PSDE in the other three mAbs led low PS20 degradation but no PS80 degradation.

Figure 7. (a) Particle formation in mAb-1 process A(200 mg/mL formulation) at 5ºC in 30 months. (b) PS20 degradation and free fatty acid concentration increased in mAb-1 process A (200 mg/mL formulation) at 5ºC in 36 months. Orange line: Released lauric acid due to PS20 degradation; Grey line: Released myristic acid due to PS20 degradation; Blue column: PS20 degradation was measured by PS20 species using LC-CAD. (c) PS20 degradation led to increased particle counts during IV admix. Particle levels for ≥10 and ≥25 µm increase (measured by HIAC) as PS20% levels decrease in the IV admixture for mAb-1 in 60 mg/mL formulation.

Particle formation in mAb-1 (200 mg/mL formulation)

PS degradation causes particle formation in two ways: 1) the FFAs released from PS degradation exceed their solubility and thus precipitate, or 2) the remaining PS can no longer support DP stability, thereby leading to the formation of proteinaceous particles. In our case, particles began to increase in the mAb-1 process A 200 mg/mL formulation after an 18-month incubation at 5°C (Figure 8a). The quantification of FFAs suggested that myristic acid exceeded its solubility (5.4 µg/mL) at 18 months, whereas lauric acid exceeded its solubility at 24 months (24.4 µg/mL), and therefore particles formed (Figure 8b)¹⁵. The continuous decrease in the percentage of PS20 in mAb-1 eventually led to a final percentage of PS20 in the DP below 0.025%, at which point the remaining PS20 could no longer stabilize the protein (Figure 8c).

Figure 8. Left panel: Comparison of the released free fatty acid due to PS degradation in mAb-4 at 37ºC. Blue dot and solid line: increased lauric acid concentration due to PS20 degradation in mAb-4; Orange dot and solid line: increased oleic acid concentration due to PS80 degradation in mAb-4. Right panel: Comparison of increased oleic acid due to PS80 degradation with increased lauric acid due to PS20 degradation in eight mAbs.

PSs are chemically diverse compounds made of ethoxylated sorbitan esterified to a series of fatty acids.³² Regarding the prediction of the earliest time of particle formation due to PS degradation, the release of FFAs other than lauric acid and oleic acid, such as myristic acid and palmitic acid for PS20 degradation, or palmitic acid and stearic acid for PS80 degradation, could also be measured.^11,15 In mAb-1 process A, particles first formed because of myristic acid rather than lauric acid precipitation; thus, a prediction based on the myristic acid increase, rather than the lauric acid increase, over the incubation time would provide a more accurate estimate of when particle formation might be expected. The conversion of the incubation time at 37°C over the course of days to 5°C over the course of months was determined on the basis of the myristic acid increase in an accelerated study and stability study (Figure 9a), and the myristic acid increase during the incubation time at 5°C for mAb-1 process B was plotted (Figure 9b). The estimated myristic acid concentration accurately matched the measured value after 24 months (Figure 9c). Given the known myristic acid solubility limit (5.4 µg/mL) at 2–8°C in mAb-1/mAb-2, we would not expect to observe particles due to PS20 degradation within 10 years for mAb-2 despite the detectable lipase activity.

Figure 9. (a). Establish the relationship between incubation at 37ºC and incubation at 5ºC for myristic acid based on mAb-1 process A. Increased myristic acid concentration was plotted against incubation time at days under 37ºC for mAb-1 process A (200 mg/mL formulation) (Eq.20) and at month under 5ºC for mAb-1 process A (60 mg/mL formulation) (Eq.21). The relationship between incubation at 37ºC and incubation at 5ºC was established by equalizing Eq.20 and Eq.21, resulting M_5C =1.31 × D_37C + 2.08 (Eq.22). (b). Estimated relationship between increased myristic acid concentration and incubation time in month @ 5ºC for mAb-1 process b (60 mg/mL formulation). Increased myristic acid was measured for mAb-1 process b (200 mg/mL formulation) at 37ºC at day 0, 3, 6, 10, 14 and plotted against calculated incubation time. Incubation time was calculated from D_37C to M_5C Eq.22. (c). Comparison of the increased myristic acid concentration due to PS20 degradation in mAb-1 process b (60 mg/mL formulation) at 5ºC in months measured by LC-MRM with the estimated value by Eq.23 (Supplementary Table 2-b). Blue column: measured value; Orange column: predicted value. Duplicates were performed for mAb-1 process b samples.

Discussion

PS degradation is an important area of study, because of its extensive use as a stabilizer for clinical and commercial biopharmaceuticals, and the reported degradation^{2,4,6,11,12,15,19,21,27} of PS during long-term storage. Here, we report a platform enabling the prediction of long-term PS stability for new programs by using stability data collected from accelerated studies, as well as different molecules containing the same lipase at trace levels. With this platform, we first established an equation enabling the conversion of incubation time at 37°C over the course of days to 5°C over the course of months for PS degradation, on the basis of one program with known lipase and stability data. We demonstrated that this conversion equation can be applied to other DPs sharing the same lipase and a similar buffer system. By monitoring the FFA increase at 37°C within 1 week, we were able to establish the PS degradation kinetics for the new DP at 5°C for as long as 24 months.

It is important to note that the prediction model used in this study did not take into account PS oxidation. Our data indicated that no PS oxidation was observed in long-term stability DPs stored under 4–8°C for up to 24 months or under 37°C for up to 28 days (Supplementary Figure S2).

The prediction is more accurate when the percentage of PS remaining is above 0.024%, whereas PS degradation is overestimated if the percentage of PS remaining decreases below 0.024% after longer periods of incubation. Despite this possible overestimation, this platform should be useful for predicting worst-case scenarios. Because the reaction rate decreases over time, the predicted FFA concentration would be higher than the actual FFA concentrations in the stability samples. If the predicted FFA released from PS degradation is below the solubility of the corresponding fatty acid, the projected percentage of PS remaining at the end of the stability study would still exceed the required amount for stabilizing the protein, and particle formation due to PS degradation would not be a concern. Further investigations would be required only if the predicted FFA release were to exceed the solubility limit, or the percentage of PS remaining would not prevent protein aggregation.

Materials and methods

Materials

Super-refined PS20 and super-refined PS80 were purchased from Croda (East Yorkshire, UK). All mAbs and recombinant LAL were produced and purified at Regeneron Pharmaceuticals, Inc. An Oasis Max 30 μm, 2.1 × 20 mm column (Waters, catalog # 186002052) and glass vials were purchased from Waters. ProteoMiner kits (catalog # 1633006) were purchased from Bio-Rad (Hercules, CA). Sodium deoxycholate (SDC) (catalog # D6750), sodium lauroyl sarcosinate (SLS) (catalog # L5777), ammonium acetate (catalog # 238074), lauric acid (catalog # 1356949), myristic acid catalog #1448990), palmitic acid (catalog # 1492007), oleic acid (catalog # 1478130), stearic acid (catalog # 1621008), linoleic acid (catalog # 62230), lauric-d23 acid (catalog # 451401), myristic-d27 acid (catalog # 366889), Oleic acid-¹³C₁₈ (catalog # 490431) are purchased from Sigma Aldrich. Water with 0.1% Formic Acid (v/v), Optima™ LC/MS Grade, Acetonitrile with 0.1% Formic Acid (v/v) (catalog # LS118), Optima™ LC/MS Grade (catalog # LS120), methanol (MeOH) (catalog # A456), Isopropanol (IPA) (catalog # 149320025) and UltraPure 1 M Tris-HCl pH 8.0 (catalog # 15568–025) were purchased from Fisher Scientific. Trypsin (Sequencing Grade Modified) (catalog # V5111) was purchased from Promega.

Methods

Host cell protein analysis by ProteoMiner with limited digestion

HCPs in mAbs were analyzed with the ProteoMiner with limited digestion method, as previously described by Zhang et al. ³⁹ In brief, a ProteoMiner bead slurry was added into 15 mg of 50 mg/mL mAb sample in 10 mM histidine, pH 6.0, buffer. Each sample was incubated at room temperature for 2.5 h with rotation, and HCP-enriched ProteoMiner beads were then loaded onto an in-house-made tip with a 9.5 µm pore size frit. The enriched HCPs were eluted from the beads with elution buffer (12 mM sodium deoxycholate and 12 mM sodium lauroyl sarcosinate) after the beads had been washed three times with water. Limited digestion was applied to the collected eluate by addition of 75 ng trypsin, digestion at 28°C overnight, reduction by dithiothreitol at 90°C for 20 min, alkylation by iodoacetamide at room temperature in the dark for another 20 min and acidification by 10% trifluoroacetic acid to pH 2–3. The peptide mixture was centrifuged at 14,000 × g for 10 min, then desalted for nano-LC MS analysis.

Stability study of formulated mAb

For each formulation, material was filter-sterilized through a syringe and Millex GV (PVDF Durapore) 0.22 µm filter in a laminar flow hood. The storage conditions were at either 4–8°C for as many as 36 months or 25°C/60% relative humidity (RH) for as many as 6 months. Intermediate time point samples were collected and stored at −80°C for PS quantification and FFA analysis or analyzed within days for particle counts by Micro-Flow Imaging (MFI) and High Accuracy liquid particle counter (HIAC).

Accelerated PS degradation study of formulated mAbs

The accelerated PS degradation study was set up as previously described by Zhang et al. ⁹ Briefly, 3 µL of internal standard (200 µg/mL lauric-d₂₃ acid, 40 µg/mL myristic-d₂₇ acid and 40 µg/mL oleic acid-¹³C₁₈ in 20% MeOH) was added into 120 µL of mAb formulation (200 mg/mL for mAb-1 process A, mAb-1 process B, mAb-3 and mAb-4; 120 mg/mL for mAb-2 process A and mAb-2 process B) and incubated with 1% PS20 or 1% PS80 at 37°C for as many as 14 days. One aliquot (10 µL) from an intermediate time point was collected and stored at −80°C for FFA measurement.

FFA extraction from mAb formulations

FFAs released through PS degradation were extracted from the formulated DPs through a protocol previously described by Zhang et al. ⁹ Briefly, 90 µL of precipitation buffer (80% isopropanol (IPA)/20% methanol (MeOH)) containing 0.55 µg/mL of lauric-d₂₃ acid, 0.11 µg/mL myristic-d₂₇ acid and 0.11 µg/mL oleic acid-¹³C₁₈ was added into 10 µL of formulated stability samples, vortexed for 1 min and maintained at room temperature for 2.5 h. The samples were then centrifuged at 25°C, 14000 × g for 30 min. Protein was precipitated, and 40 µL supernatant was collected and transferred to a 96-well plate for LC-multiple reaction monitoring (MRM) analysis. For the accelerated study samples, protein was precipitated by precipitation buffer containing 80% IPA/20% MeOH without heavy isotope labeled internal standards.

FFA quantification with LC-MRM

Extracted FFAs were measured with LC-MRM on an Agilent 6495 QQQ mass spectrometer (Agilent, Wilmington, DE) equipped with an Agilent 1290 Infinity UHPLC (Agilent, Wilmington, DE). Subsequently, 5 µL of supernatant from FFA extraction was injected into the LC-MS system. FFAs were analyzed on an Acquity CSH C18 column (2.1 × 50 mm, 1.7 mm) at 40°C, with 20 mM ammonium acetate in water as mobile phase A and methanol as mobile phase B. The column was first equilibrated with 50% mobile phase B in 0.1 min after injection, then linearly increased to 90% B in 5.5 min, held for 2 min and re-equilibrated at 50% mobile phase B for 1.9 min. Elution was performed at 0.4 mL/min, and peaks at 1–8 min were analyzed with an electrospray ionization source operating in negative mode, with a nebulizer gas pressure of 20 psi, sheath gas temperature of 300°C, sheath gas flow of 11 L/min, capillary voltage of −3000 V and nozzle voltage of 500 V. Lauric acid and lauric-d₂₃ acid were monitored at 199.1/199.1 and 222.3/222.3, respectively; myristic acid and myristic-d₂₇ acid were monitored at 227.2/227.2 and 254.4/254.4, respectively; and palmitic acid, oleic acid, oleic acid-¹³C₁₈, stearic acid and linoleic acid were monitored at 255.2/255.2, 281.2/281.2, 299.2/299.2, 283.2/283.2 and 279.2/279.2, respectively, with CE 5. Data analysis was performed with Skyline, and FFA concentrations were calculated on the basis of a calibration curve created from the spiked-in fatty acid concentration plotted against $\frac{PeakareaofFFA}{PeakareaofISTD}$.

Quantification of PS degradation by LC-CAD

Degradation of PS20 or PS80 in a formulated antibody drug substance was analyzed with a UPLC-CAD system. PSs and associated degradants were separated from formulated mAbs with an Oasis MAX column (2.1 × 20 mm, 30 µm), with the initial gradient held at 90% solvent A (0.1% formic acid in water) and 10% solvent B (0.1% formic acid in acetonitrile). The gradient was increased to 20% solvent B in 1 min, held for 2.4 min, increased to 100% solvent B in 0.1 min and held for 1 min. This was followed by equilibration with 10% B for 1.5 min. The flow rate was kept at 1 mL/min, and the column temperature was at 30°C.

The UPLC system was set up with a Waters UPLC H-class system coupled to a Corona^Ⓡ Ultra/Veo CAD detector operated under a nitrogen pressure of 75 psi for quantification. The peak of interest from each chromatogram was identified and integrated with Empower software from Waters and quantified on the basis of a calibration curve based on PS stock solutions.

Particle measurement

Sub-visible particle analysis by light obscuration

Particle size and concentration were determined by light obscuration with an HIAC 9703+ Liquid Particle Counting System equipped with an HRLD 400 CE/Standard Sensor with a theoretical size range of 2–400 µm and a Tecan 1 mL Sample Syringe from Beckman Coulter Life Sciences (Indianapolis, IN). The performance of the instrument was verified with 15 µm polystyrene counting and size standards purchased from Thermo Fisher Scientific (Waltham, MA). Before particle analysis, samples were vacuum degassed for 15 min. Subvisible particles were reported for ≥10 µm and ≥25 µm particles per container. Four 1 mL aliquots were collected from the samples, and the average of the last three draws is reported.

Sub-visible particle analysis by flow imaging

Particle size, morphology and concentration were characterized through MFI™ 5200 flow microscopy with a Bot1 Autosampler equipped with a 100 µm flow cell of 1.6 mm with silane coating (Bio-Techne. Minneapolis, MN; Protein Simple, Inc., Santa Clara, CA). System suitability was ensured with COUNT-CAL™ 3000/mL, 5 µm concentration standard, purchased from Thermo Fisher Scientific (Waltham, MA). MFI Image Analysis was performed with the “remove stuck” and “remove edge” filters to view images of the particles, and the number of particles per mL for 2–10 µm, ≥10 µm and ≥25 µm are reported.

Abbreviations

Name	=	Abbreviation
acetonitrile	=	ACN
Chinese hamster ovary	=	CHO
day at 37C	=	D37C
dithiothreitol	=	DTT
drug product	=	DP
drug substance	=	DS
free fatty acids	=	FFA
group XV lysosomal phospholipase A2 isomer X1	=	LPLA2
High Accuracy liquid particle counter	=	HIAC
host cell protein	=	HCP
iodoacetamide	=	IAM
isopropanol	=	IPA
lipoprotein lipase	=	LPL
liquid chromatography-charged aerosol detector	=	LC-CAD
lysosomal acid lipase	=	LAL
methanol	=	MeOH
Micro-Flow Imaging	=	MFI
monoclonal antibody	=	mAb
month at 5C	=	M5C
Multiple Reaction Monitoring	=	MRM
palmitoyl-protein thioesterase 1	=	PPT1
polysorbate	=	PS
polysorbate 20	=	PS20
polysorbate 80	=	PS80
polysorbate degrading enzyme	=	PSDE
relative humidity	=	RH
sialate O-acetylesterase	=	SIAE
sodium deoxycholate	=	SDC
sodium lauroyl sarcosinate	=	SLS
Trifluoroacetic acid	=	TFA

Disclosure statement

No potential conflict of interest was reported by the author(s).

Supplemental material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/19420862.2023.2232486

Additional information

Funding

The author(s) reported there is no funding associated with the work featured in this article.