The determination of multiple per- and polyfluoroalkyl substances (PFAS) migrating from food contact material gained in importance as an increasing range of PFAS has been found migrating from food contact material into food. In this study, an integrated analytical approach that combines high performance liquid chromatography–tandem mass spectrometry (HPLC–MS/MS) and gas chromatography–tandem mass spectrometry (GC–MS/MS) was established for detecting 36 PFAS migrating from food contact materials into four food simulants (3% acetic acid, 10% alcohol, 50% alcohol, and olive oil). The response surface methodology was applied to optimize the selection of solvents in sample treatment. This target analytical method was appropriate for the determination of multiple PFAS, with recoveries ranging from 81.8 to 118.7%. The relative standard deviations (RSDs) ranged from 2.4 to 7.8%, and detection limits were in the range of 0.3 to 10 µg/kg in relevant food simulants.

Per- and polyfluoroalkyl substances (PFAS) are a family of synthetic substances that do not occur naturally in the environment. They are a concern due to their stable physical and chemical properties with strong C-F bonds, including their chemical inertness, thermal stability, high surface activity, and hydrophobic-oleophobic properties (1–3). PFAS are widely used in consumer goods, household products and food contact materials (4–5). In food contact material (FCM), PFAS are mainly used in nonstick cookware, as well as the coatings of paper and other resistant materials, due to their oil- and water-repellent properties (6–7). Studies have indicated that the migration of PFAS from FCM into food is likely to be the main route of consumer exposure to these substances (8–11).

PFAS have been found to be highly resistant, and could persist in the environment for long periods of time, as well as in human serum, milk, and tissues (12–17). Certain type of PFAS, such as perfluorocarboxylic acids (PFCA) and perfluorosulfonic acids (PFSA), are likely to be toxic and bioaccumulate, posing adverse effects on human health (18–21). Perfluorooctane sulfonic acid (PFOS) and perfluorooctanoic acid (PFOA) are two of the most studied PFCA and PFSA. There is suggestive evidence from human epidemiology studies that PFOS and PFOA may cause abnormal liver enzymes, antibody response, and cancer (21–23). To reduce the occurrence of PFOS and PFOA, a number of PFAS, such as perfluoro‑alcohol and its derivatives, have been used as substitutes to replace PFCA and PFSA in FCM. However, studies indicated that these substitutes could also pose health risks to humans, as PFSA and PFCA precursors are more toxic than the PFSA or PFCA themselves (24). In addition, these precursors could be absorbed into the human body, and degrade into PFOA and PFOS by an oxidation mechanism (25). For example, 8:2FTOH (flurotelomer alcohol) and 10:2FTOH could be converted into PFOA, and PFDA, and N-methyl perfluorooctane sulfonamido ethanol (N-MeFOSE) and N-ethyl perfluorooctane sulfonamido ethanol (N-EtFOSE) could be capable of conversion into PFOS. Therefore, besides known harmful PFAS, the migration of PFAS precursors from food contact material must also be taken into account to ensure its safety.

To protect consumers from exposure to PFAS migrated from FCM, stringent regulatory approaches have been adopted to control their production, application, and migration. The Danish Food and Veterinary Administration set a recommended limit for the total organic fluorine content in paper and cardboard at 0.35 μg per square decimeter of paper. The U.S. Food and Drug Administration (FDA) finalized an amendment to regulations that certain type of PFAS were not permitted as additives used in the manufacture of FCM (26). In China, the National Food Safety Standard GB 9685, which is the regulation for the use of additives in FCM, added an amendment in 2016 that no longer authorized PFAS as additives in the manufacture of FCM (27). In 2017, The European Chemicals Agency (ECHA) added seven types of PFCAs and PFSAs to the Substances of Very High Concern (SVHC) list (28), which attracted much attention concerning the occurrence of PFAS, and their effects on the environment and human health.

Comparable results for measurement of PFAS migration from FCM is crucial for official control purposes. To achieve this objective, the guidance for choice of food simulants and migration test conditions for plastics have been provided in relevant regulations (29), which define the food simulants that should be used to mimic a specific foodstuff, and what standardized conditions of time, temperature, and contact should be performed. Various analytical approaches have been investigated for measuring the migration of PFAS from FCM (30–35). GC–MS is usually used for detecting volatile fluorine-containing compounds, such as perfluoroalcohols and perfluoroalcohol acrylates. LC–MS/MS, a highly sensitive and selective tool, was applied to detect more than 10 PFAS that mainly refer to straight‑chain perfluorinated carboxylic acids or perfluorinated sulfonic acids. However, an analytical approach that is suitable for determination of multiple PFAS to detect migration from food contact materials has not been well established, especially for the increasing range of precursors, including PFCAs and PFSAs. In addition, the previous studies mainly focused on measuring the residue of PFAS in different matrices of FCM, such as extracts of paper. Few studies have been carried out on detecting migration of PFAS by using conventional simulants that represent the specific foodstuff.

To meet official control purposes, this present study aims to establish an effective method for simultaneously measuring multiple PFAS migrated from FCM, and possibly containing both PFAS and their precursors, including fluorinated carboxylic acid, hydrogen‑substituted fluorinated carboxylic acid, and hydrogen‑substituted fluorinated alkyl acid amides. To achieve this objective, the optimization of sample treatment was carried out by using a response surface methodology. An integrated analytical approach of combining high performance liquid chromatography–tandem mass spectrometry (HPLC–MS/MS) and gas chromatography–tandem mass spectrometry (GC–MS/MS) was established in four conventional food simulants (3% acetic acid, 10% alcohol, 50% alcohol, and olive oil) that are considered to represent specific foodstuffs.

Ultrapure water was puriﬁed using a Milli-Q system (Millipore, Milford, Massachusetts, USA). Alcohol and methanol (HPLC‑grade) were purchased from TEDIA Company. Ammonium acetate and acetic acid were purchased from Guangzhou Chemical Reagent Factory (Guangzhou, China), and olive oil was purchased from Sinopharm (Shanghai, China).

A total of 36 PFAS standards (Table 1) were purchased from Wellington Laboratories (Guelph, Ontario, Canada), Dr. Ehrenstorfer Company (Augsburg, Germany), Chiron Company (Trondheim, Norway), and TRC Company (North York, Ontario, Canada). An intermediate standard solution containing 27 PFAS (0.1 μg/mL, Group 1) was prepared by dissolving the commercial standards in methanol. An intermediate standard solution containing nine PFAS standards (1.0 μg/mL, Group 2) was prepared by dissolving the commercial standards in dichloromethane.

The standard working solutions used for LC–MS/MS analysis were prepared by transferring the intermediate standard (1.0 μg/mL), containing 27 PFAS (Group 1), into a 10 mL volumetric flask, spiked with 50 μL (1.0 μg/mL) 1,2,3,4-

 perfluorooctanoic acid (MPFOA) of internal standard, and then made up to 10 mL with the food simulants (10% [v/v] ethanol, 3% [w/v] acetic acid, and 50% [v/v] ethanol), respectively. The standard working solutions used for GC–MS/MS analysis were prepared by transferring the intermediate standard (1.0 μg/mL) containing nine PFAS (Group 2) into a 10 mL volumetric flask, spiked with 50 μL (0.2 μg/mL) methyl margarate-d33 of internal standard, and then made up to 10 mL with the food simulants (10% [v/v] ethanol, 3% [w/v] acetic acid, and 50% [v/v] ethanol), respectively. Then, 2 mL dichloromethane was added, vortexed for 5 min, allowed to separate, and the lower solvent layer was collected for further analysis.

The standard working solutions for the food simulant (olive oil) were prepared by transferring the intermediate standard of Group 1 (0.1 μg/mL) and Group 2 (1.0 μg/mL) into 10 g olive oil, spiked with internal standard of 10 μL MPFOA (1.0 μg/mL) and 50 μL (0.2 μg/mL) methyl margarate-d33. Then, 2 mL acetonitrile was added, vortexed for 5 min to allow stratification, and the upper solvent layer was taken for further analysis.

An LC–triple quadrupole mass spectrometer and a GC–triple quadrupole mass spectrometer (6410 Triple Quad LC–MS, 7000c Triple Quad GC–MS, Agilent Technologies, Palo Alto, California, USA) equipped with automatic injectors, were employed for the identiﬁcation and quantiﬁcation of PFAS.

The LC column employed was a 15 cm Poroshell120 EC-C18 column (2.7‑µm particle diameter and 2.1‑mm i.d., Agilent Technologies), while the GC column employed was 30 m DB-5 column (0.25‑µm particle diameter and 0.25‑mm i.d., Agilent Technologies). The LC mobile phase consisted of water (solvent A) and HPLC grade methanol (solvent B). The gradient elution procedure was as follows: 0–3 min, 10% B; 3–4 min, 20% B; 4–5 min, 45% B; 5–11 min, 70% B; 11–18 min, 85% B; 18–19 min, 100% B; 19–20 min, 75% B; 20–21 min, 50% B; 21–24 min, 20% B; 24 min, 10% B. The injection volume was 5 μL, and the flow rate was 0.2 mL/min. ESI was used in negative mode with the following conditions: spray ion voltage, 4000 V; nebulizer, 20 psi; gas flow, 8 L/min; and capillary temperature, 350 °C. The GC temperature program was as follows: 75 °C for 3 min, 30 °C/min to 250 °C for 0 min, 50 °C/min to 300 °C for 6 min. The injection volume was 2 μL and pulsed splitless. The gas flow was 1.5 mL/min, with an electron ionization (EI) source and multiple reaction monitoring (MRM) mode. All of the ionization parameters and collision cell parameters were optimized for each analyte (Table 1).

Migration experiments were carried out in accordance with European Union (EU) regulation 10/2011 for plastic materials and articles intended to be in contact with food (29).

Each sample was exposed to the food simulants with a ratio of contact area‑to‑volume at 1000 mL:6 dm

, and then placed in an incubator at 70 °C. The food simulants were collected from each sample at 2 h, followed by cooling at room temperature, and then further analysis.

The majority of PFAS being investigated are fatty acid compounds that can ionize protons, and negatively charged parent ions in aqueous solution. For this reason, LC–MS/MS was used to detect 27 of this type of PFAS (Group 1). For the rest of the nine PFAS being investigated (Group 2), these are generally considered volatile compounds in which perfluoroalcohols and perfluoroalcohol acrylates would not easily be ionized in aqueous solutions; these compounds are more susceptible to ionization in an EI source than in an ESI source, therefore GC–MS/MS was preferred for their analysis.

The PFAS analytes were extracted from the relevant food simulant collected following an exposure step. For the determination of 27 PFAS (Group 1), the food simulants (10% ethanol, 3% acetic acid, and 50% ethanol) were transferred into a 10 mL volumetric flask, and filtered by a 0.22 μm syringe filter, followed by direct LC–MS/MS analysis.

The aqueous simulants containing nine PFAS (Group 2) were extracted with dichloromethane, and a volume of 10 mL of aqueous food simulants was transferred into the headspace bottle, with 50 μL (0.2 μg/mL) methyl margarate-d33 added as the internal standard. Dichloromethane was then added, and vortexed for stratification. The bottom layer was filtered, and used for GC–MS/MS analysis.

For treatment of the olive oil, 10 g of olive oil was transferred into the headspace bottle, and 10 μL (1.0 μg/mL) MPFOA and 50 μL (0.2 μg/mL) methyl margarate-d33 were added as the internal standards. After acetonitrile was added and vortexed for stratification, the supernatant liquor was filtered and analyzed by LC–MS/MS and GC–MS/MS.

A mixture of 27 PFAS comprising a standard at a concentration of 100 ng/mL was used to check the mass spectrometric response and separation power of four different mobile phases, including methanol–water solution, methanol–10 mmol/L ammonium acetate solution (pH = 7), methanol–(containing 0.1% ammonia) water solution, and acetonitrile–10 mmol/L ammonium acetate solution.

The results indicated that the weakest MS signal for target analytes was observed for acetonitrile–10 mmol/L ammonium acetate solution, and the strongest MS signal was observed for methanol–aqueous solution (containing 0.1% ammonia). This is likely due to the MS signal of PFAS analytes being related to the formation of the negative parent ion of PFAS, and the degree of ionization of the proton as well. Methanol allows analytes to generate target ions more easily, and be atomized on the electrospray ionization process, while acetonitrile may decrease the ionization of analytes. The poor peak shapes and baseline disturbances were observed for certain analytes, such as PFBA and PFPeA when methanol–water (containing 0.1% ammonia) was applied. Therefore, methanol–water (containing 0.1% ammonia) was not chosen as the mobile phase for further experiments.

Either methanol–water solution or methanol–10 mmol/L ammonium acetate solution was considered satisfactory as a suitable mobile phase, given that a similar ion peak response was obtained, and good chromatographic peak shapes were observed. It can be explained that the ionization of a proton occurs more often in a neutral or weak alkaline mobile phase, leading to an increase of formation of parent ions. Considering the methanol–water solution has the advantage of making maintenance of instrument pipelines simpler and more effective, the methanol–water solution was preferred for further experiments.

Three different LC columns, including Agilent Zorbax SB-C18 (150 mm × 4.6 mm, 5‑μm), Poroshell 120 EC-C18 (150 mm × 3.0 mm, 2.7‑µm), and Eclipse XDB C-18 (150 mm × 4.6 mm, 3.5‑µm), were compared in the separation of a mixture of standards (27 PFAS), at a concentration solution of 100 ng/mL.

The results indicated that all three columns could effectively separate the target analytes, particularly n-alkane PFAS. However, due to the occurrence of a high number of target PFAS analytes, the chromatographic peaks of various analytes may overlap each other, causing difficulty in further analysis. The Poroshell 120 EC-C18 was shown to be able to completely separate 27 PFAS analytes in 25 min with the same gradient elution program. Therefore, the Poroshell 120 EC-C18 was chosen as the stationary phase for further experiments.

For the selection of a GC column, baseline separations could be achieved for nine target PFAS using a low or medium polarity column. Therefore, the DB-5, DB-17, or DB-35 column could be chosen to perform the analysis.

The MRM chromatogram of 36 PFAS compounds is shown in Figures 3 and 4.

The Choice of Extraction Solvent for Olive Oil

Olive oil, as a fatty simulant, cannot be directly analyzed due to interference from compounds it contains. Therefore, it must be converted into a suitable solution before analysis. To achieve the objective, a recovery experiment was conducted in an olive oil blank spiked with 0.1 mL PFAS standards (10 μg/mL), aimed at selecting a suitable solvent to extract the PFAS from the olive oil. Four different solvents—methanol, acetonitrile:water (1:1), acetonitrile:water (3:1), and acetonitrile—with serial extraction volumes (1 mL, 1.5 mL, 2 mL, 2.5 mL, 3 mL), were investigated. The recovery rate for each solvent is indicated in Figure 2a.

The results indicated that acetonitrile:water (3:1) and acetonitrile had the highest extraction efficiency, and a slightly higher recovery rate was obtained for acetonitrile compared with that of acetonitrile:water (3:1). One of the reasons might be that the PFAS being investigated are long‑chain aliphatic compounds, and the pure organic phase is more favourable for the extraction of the target analytes. For methanol, the separation of the methanol and olive oil layers was not clearly observed after mixing because of the emulsification effect, although the fluorinated fatty alcohol and some perfluorinated carboxylic acids had the highest recovery rates. In the case of acetonitrile:water (1:1) solvent, delamination was not clear, and the recovery rate was low, therefore acetonitrile:water (1:1) was not suitable for GC–MS analysis. Given the test results indicted, acetonitrile was chosen as the solvent to extract the PFAS from the olive oil.

The extraction conditions of PFAS in olive oil were optimized by applying a single‑factor experiment derived from the response surface methodology. Perfluorononanoic acid (PFNA) was taken as the sample to be investigated, and the acetonitrile concentration (A, %), volume of extraction solvent (B, mL), and the vortex time (C, min) were assigned as independent variables. The recovery rate of PFAS were taken as the response value based on the principle of the Box-Behnken central composite design. A secondary polynomial regression model was obtained from 15 experiments with three factors and three levels: recovery = 72.46 + 7.77 × A + 1.47 × B + 5.23 × C − 1.35 × A × B + 0.50 × A × C + 1.75 × B × C + 3.70 × A

. The results of the recovery of PFNA obtained by the response surface methodology are shown in Table 2 and Figure 1.

It can be found from the F-value (Table 2) that the sequence of influence of a single factor on the recovery rate was A > C > B, suggesting that the influence of A

 was significant in the quadratic term, while B

 were not. Thus, 100% acetonitrile was considered an appropriate extraction solution. The highest extraction efficiency was observed when the extraction time was 5 min, and the extraction volume was 2 mL.

The Choice of Extraction Solvent for Aqueous Simulants

In this study, GC–MS was employed for detecting fluorinated fatty alcohols, fluorinated fatty alcohol acrylates, and fluoro-sulfonamide fatty alcohols in the extraction of aqueous simulants. A comparison experiment was conducted on two aqueous simulants (3% acetic acid and 50% ethanol) to check the extraction efficiency of the solvents n-hexane, dichloromethane, and ethyl ether:n-heptane (1:1).

The results showed that all selected solvents were well able to extract the analytes of the aqueous simulants. However, n-hexane was not a suitable solvent to extract fluorinated fatty alcohols, due to its weak polarity. A mix of solvents consisting of ethyl ether and n-heptane did not demonstrate sufficient extraction efficiency. In addition, ethyl ether may cause a hazard to the analyst during vortexing at high pressures. The data indicated that the highest recovery rate was obtained when dichloromethane was applied for measuring fluorinated fatty alcohols and fluorinated fatty alcohol acrylates, which may be due to its medium polarity and strong density that is lower than water‑based solutions. Thus, dichloromethane was chosen as the extraction solvent. Figure 2b shows the extraction efficiency for three different solvents by measured recovery rate. The highest extraction efficiency was achieved with a 5 min extraction time with a 2 mL extraction volume.

The mass spectrometry methods for 27 PFAS in Group 1 and 9 PFAS in Group 2 were optimized to directly detect each PFC (1 mg/L) in methanol. Standard solutions were scanned by performing a first‑stage mass spectrometry full scan in negative ion ESI mode with corresponding transmission voltages. The results showed that the strongest peaks of the perfluorocarboxylic acid, hydrogen substituted fluorocarboxylic acid, or perfluorosulfonamide compounds were the quasi‑molecular ion peaks [M-1]

 or the molecular ion peaks following the loss of the carboxyl group [M-45]

. The strongest signal of the perfluorosulfonic acid or hydrogen substituted perfluorosulfonic acid compounds observed were quasi‑molecular ion peaks [M-1]

 and sulfonic acid group molecular ion peaks [SO

. Nine types of PFAS from Group 2 were scanned by conducting a first‑stage mass spectrometry full scan under the positive EImode. The results indicated that the fluorinated fatty alcohols and fluorinated fatty alcohol acrylates formed smaller fragmentary ion peaks [C

 (M = 131), which are shown in Figures 3 and 4.

Ion beam scanning was performed with a collision energy of 0–60 eV targeting each quasi‑molecular ion to obtain the daughter ion information and the collision energy values that were suitable for the daughter ion response sizes. Daughter ions of perfluorocarboxylic acid compounds are a series of molecular ions following the loss of the carboxyl group, and there are several CF

 fragments between each molecule, including the ionic fragment of [C

, having a general structural formula of [CnF

. The ions of the perfluorosulfonic acid compounds were those molecular ions that lost several [CnF2n+1]– saturated carbon chain groups and had sulfonic acid functional groups, such as [CnF

 ions. Both the fluorinated fatty alcohols and fluorinated fatty alcohol acrylates contained the ion-pair [C

 (M = 69). Therefore, the fragment ions with the largest response and least interference were chosen as the quantitative and qualitative ions. To obtain better sensitivities, the parameters including the selection of daughter ions, fragmentation voltage, and collision energy were optimized (see Table 1).

Using the method described above, the 36 PFAS compounds were analyzed in the MRM mode. The concentration and peak area of each compound was taken as the abscissa and ordinate, and a standard curve was drawn. The linear range of the recovered curve was between 1 and 1000 ng/mL, and the limits of quantification (10 S/N) were 0.35–9.72 μg/kg in 3% acetic acid, 0.29–8.43 μg/kg in 10% alcohol, 0.28–8.73 μg/kg in 50% alcohol, and 0.43–9.86 μg/kg in olive oil.

An experiment was carried out where samples were spiked with 10 and 100 ng standards in 3% acetic acid, 10% alcohol, 50% alcohol, and olive oil, respectively. The recovery rates for the four different food simulants ranged from 81.8 to 118.7%, and the relative standard deviations (RSDs) ranged from 2.4 to 7.8%.

A total of 70 food contact materials were collected from local supermarkets, which include coated paper board, multilayer paper packaging, multilayer plastic food packaging, heat‑resistant rubber articles, and coated heat‑resistant metallic containers. The novel analytical approach was applied to perform the analysis. Some specific PFAS compounds were found in paper samples with concentrations of 0.01 mg/kg to 0.05 mg/kg of food simulants, which accounted for 5% of the total number of samples analyzed. Multilayer cardboard and coated paper board were the main types of packaging in which PFAS compounds were found with average concentrations of 0.02 mg/kg. The highest migration of PFAS was 0.08 mg/kg. The type of PFAS compounds found were mainly straight‑chain perfluorinated acids and perfluorinated alcohol compounds. The chromatogram profiles of the PFAS in typical samples are shown in Figure 5.

A target approach was established for the determination of 36 PFAS migrating from food contact material into food simulants of 3% acetic acid, 10% alcohol, 50% alcohol, and olive oil. A total of 36 PFAS were divided into two groups, and measured by LC–MS/MS and GC–MS/MS. LC–MS/MS was used to detect 27 fatty acid PFAS, while 9 PFAS with volatile properties, including fluorinated fatty alcohols and fluorinated fatty alcohol acrylates were measured by GC–MS/MS. The response surface methodology was a useful tool to simplify the selection of solvents with optimized conditions. This integrated analytical approach was appropriate for the determination of multiple PFAS with recovery rates that ranged from 81.8 to 118.7%, the RSD ranged from 2.4 to 7.8%, and the detection limits ranged from 0.35 to 9.72 μg/kg in 3% acetic acid, 0.29 to 8.43 μg/kg in 10% alcohol, 0.28 to 8.73 μg/kg in 50% alcohol, and 0.43 to 9.86 μg/kg in olive oil. This target approach had the advantage of simultaneously measuring the migration of multiple PFAS from food contact materials with satisfactory sensitivity, accuracy, and reliability. The test results showed that PFAS were found in coated paper and board at mg/kg levels, suggesting that further investigation is needed for the migration of PFAS from coated paper and board.

The authors acknowledge funds given by national quality infrastructure program (No. 2016YFF0203705), and the General Administration of Quality Supervision Inspection and Quarantine Science-Technology Programs (Grant 2014IK078, Grant 2015IK049). The authors have declared no conﬂict of interest.

 (Marcel Dekker Inc., New York, 2nd ed., 2001).

J. Jeon, K. Kannan, H.K. Lim, H.B. Moon, and S.D. Kin, 

T.H. Begley, K. White, P. Honigfort, M.L. Twaroski, R. Neches, and R.A. Walker, 

DuPont. Information on PFOA.2008. http://repanet.de/PFOA2/en_US/index.html. April 07. 2008.

X. Trier, K. Granby, and J.H. Christensen, 

E. Zafeiraki, D. Costopoulou, I. Vassiliadou, E. Bakeas, and L. Leondiadis, 

G.W. Olsen, K.J.Hansen, L.A. Stevenson, J.M. Burris, and J.H. Mandel, 

K. Inoue, F. Okada, R.S. Kato, S. Sasaki, S. Nakajima, A. Uno, Y. Saijo, F. Sata, Y. Yoshimura, R. Kishi, and H. Nakazawa, 

G.W. Olsen, H-Y Huang, K.J. Helzlsouer, K.J. Hansen, J.L. Butenhoff, and J.H. Mandel, 

M.K. So, N. Yamashita, S. Taniyasu, Q. Jiang, J.P. Giesy, K. Chen, and P.K.S. Lam, 

L. Liu, J. She, X. Zhang, J. Zhang, M. Tian, Q. Huang, S.A. Musstjab, A.S. Eqani, and H. Shen,

T. Zhang, H.W. Sun, Q. Wu, X.Z. Zhang, S.H. Yun, and K. Kannan, 

C. Lau, J.R. Thibodeaux, R.G. Hanson, M.G. Narotsky, J.M. Rogers, A.B. Lindstrom, and M.J. Strynar, 

N. Johansson, P. Eriksson, and H. Viberg, 

Health Effects Support Document for Perfluorooctanoic Acid (PFOA)

, U.S. Environmental Protection Agency EPA822R16003. https://www.epa.gov/sites/production/files/2016-05/documents/pfoa_hesd_final-plain.pdf. (May, 2016; accessed November 17, 2020).

Health Effects Support Document for Perfluorooctane Sulfonate (PFOS)

, U.S. Environmental Protection Agency EPA822R16002. https://www.epa.gov/sites/production/files/2016-05/documents/pfos_hesd_final_508.pdf. (May, 2016; accessed November 17, 2020).

I. Zabaleta, E. Bizkarguenaga, and D. Bilbao, 

M.J.A. Dinglasan, Y. Ye , E.A. Edwards, and S.A. Mabury, 

Indirect Food Additives, Paper and Paperboard Components

, https://www.federalregister.gov/articles/2016/01/04/2015-33026/indirect-food-additives-paper-and-paperboard-components (2016).

Standard for the Use of Additives in Food Contact Materials and Articles

, http://www.nhfpc.gov.cn/sps/s7891/201611/06ed87a09dad4cf6aee
48cd89efbef35.shtml (2016).

Background Document to the Opinion on the Annex XV Dossier Proposing Restrictions on Perfluorooctanoic Acid (PFOA), PFOA Salts and PFOA-related Substances

, ECHA/RAC/RES-O-0000006229-7002/F, ECHA/SEAC/RES-O-0000006229-70-03/F (2015).

Commission Regulation (EU) No. 10/2011 of 14 January 2011 on Plastic Materials and Articles Intended to Come into Contact with Food

, https://www.fsai.ie/uploadedFiles/Reg10_2011.pdf (2011).

W.M. Henderson, E.J. Weber, S.E. Duirk, J.W. Washington, and M.A. Smith,

M. Stadalius, P. Connolly, K.L. Empereur, J.M. Flaherty, T. Isemura, M.A. Kaiser, W. Knaup, and M. Noguchi, 

F. Gosetti, U. Chiuminatto, D. Zampieri, E. Mazzucco, E. Robotti, G. Calabrese, M.C. Gennaro, and E. Marengo, 

Articles

Peer Reviewed

An increasing range of per- and polyfluoroalkyl substances (PFAS) has been found migrating from food contact material into food. This article establishes an integrated analytical approach combining HPLC–MS/MS and GC–MS/MS to detect 36 such PFAS.

The Determination of Trace Per- and Polyfluoroalkyl Substances and Their Precursors Migrated into Food Simulants from Food Contact Materials by Liquid Chromatography–MS/MS and Gas Chromatography–MS/MS

Q. Interaction polymer chromatography (IPC) is regarded as an “umbrella” term that covers a variety of separation techniques. Can you define IPC and describe what techniques this term includes?

 The term “interaction polymer chromatography” or IPC encompasses gradient polymer elution chromatography (GPEC, in both traditional and interactive modes), temperature gradient interaction chromatography (TGIC), and barrier methods, such as liquid chromatography at limiting conditions of absorption, and desorption and SEC-gradient analysis.

Q. Is there a distinct retention mechanism for IPC techniques?

One generally distinguishing feature of almost all IPC methods is that the retention mechanism is predominantly governed by enthalpic interactions between the analyte and the column packing material, as mediated by the mobile phase and, in TGIC, also by temperature. This distinguishes IPC techniques from so‑called “size‑based” separations, such as size‑exclusion chromatography, hydrodynamic chromatography and flow field‑flow fractionation where retention is primarily a result of changes in the standard entropy of transfer between phases. An exception, which is still included under the IPC rubric, are liquid chromatography at the critical condition (LCCC) methods where, at least in theory, the entropic and enthalpic forces of transfer balance each other resulting in the standard Gibbs free energy of transfer between phases equalling zero.

Q. Are there specific types of applications that IPC techniques are used in, and what benefits do they offer?

 IPC is primarily used to determine the chemical composition distribution of copolymers. The methods are also used to separate multi‑component mixtures according to the chemical composition of the samples’ constituents, to separate by tacticity, by end group distribution, and so on. Individually, they provide information not generally obtainable by size‑based techniques. More powerfully, they can combine with these latter techniques, in a two‑dimensional fashion, to provide, for example, the combined chemical composition and molar mass distributions (CCD×MMD) of a copolymer or blend in a single analysis.

Q. Can you describe an example that illustrates the benefits of IPC in practice?

When I worked in the chemical industry, I developed an interactive GPEC method to determine the vinyl alcohol (VOH) distribution in poly(vinyl butyral) or PVB terpolymer (1). PVB is primarily used as the plastic interlayer in automotive and architectural safety glass because of its excellent penetration resistance and adhesion to glass, and the VOH content is a key parameter in controlling this adhesion; good adhesion prevents the glass on the inside of your windshield from flying into your face when a rock hits the windshield! There are existing titration, nuclear magnetic resonance- (NMR) spectroscopy, and near infra-red (NIR) spectroscopy methods to measure the VOH content, but all these only provide average values, whereas a chromatographic method, such as the normal-phase (NP) GPEC method developed, provides the entire VOH distribution and accompanying statistical averages and dispersity (1). For example, the two PVB samples in Figure 1 differed from each other with respect to their adhesive properties when, in theory, they were supposed to be identical in this regard. VOH content was the immediate suspect, but both titration and NIR analysis showed no quantitative difference in average VOH content. Employing NP-GPEC, the difference in performance behaviour of the samples could be immediately attributed to their different VOH distributions, shown in the overlays in Figure 1.

Q. IPC is more commonly used in research and specialized industrial laboratories. Why is this and do you think the techniques could be adopted more routinely? What are the main obstacles preventing this?

The biggest obstacle towards more widespread use of IPC methods is that little guidance has been provided in the literature with respect to method development for the various techniques. Most publications showcase what I refer to as “bespoke” methods, that is, they are applicable only to the analyte(s) studied in that particular paper. It is not clear how these particular bespoke methods were developed nor how they can be applied or adapted to the analysis of a different polymer. There is, however, ample potential for these techniques for the analysis of industrial polymers intended for both specialized and everyday applications because these are usually complex macromolecules or blends possessing distributions in a host of physicochemical properties.

Q. You recently published a paper on method development in IPC (2). What was the aim of this paper and what were the main conclusions?

 My main aim was to address the shortcomings mentioned previously with respect to guidance in IPC method development. Essentially, I tried to provide a self‑contained guide to the fundamental chromatographic and polymer science principles that need to be considered when developing methods, as well as to provide guidance specific to each of the individual IPC methods mentioned above.

Q: Are there any specific differences that need to be taken into account when developing an IPC method compared to other approaches?

The biggest differences between IPC methods and size-based methods regard the use of mixed solvents in the former and, therefore, solvent gradients; the role of temperature and, in TGIC, temperature gradients the use of columns with a functionalized surface that interacts with the analytes. There is also the fact that one is often trying to separate a multi-component mixture containing many more components than are usually separated in SEC, HDC, or flow FFF. In the cases of solvent and temperature, one must understand (or, for temperature, continue to try to understand because systematic studies of the effect of temperature on IPC separations other than TGIC are almost non-existent) how these mediate interactions between analytes and stationary phase chemistry to achieve preferential retention. Column functionality, gradient design, and multi-component separations all involve a knowledge of fundamental separation science greater than is usually needed when designing a size-based separation.

 I would very much like to see these methods in the hands of a more general, yet simultaneously more applied, audience, such as industrial non-specialists. As method development thrives, so will the information obtained and benefits derived there from, in terms of superior products and more versatile applications. To do this, non-specialists need to know where to start with respect to method development. Another thing that is needed is columns specific to IPC. Currently, except for when size‑exclusion chromatography columns can be repurposed for IPC analysis, IPC employs columns designed for small‑molecule separations. There is great product variability amongst manufacturers of what should be the same columns; for example, the performance of diol- or cyano columns from different manufacturers can vary greatly when the columns are used in interactive GPEC or other IPC analyses. Other issues such as surface inhomogeneities in “small‑molecule” columns can compromise macromolecular analyte recovery and lead to biased results regarding molar mass, chemical composition, and so on.

W. Radke, J. Chromatogr. A., 1335, 62–79 (2014).

Y. Brun and C.J. Rasmussen, in: Liquid Chromatography: Fundamentals and Instrumentation, S. Fanali, P.R. Haddad, C.F. Poole, and M.-L. Riekkola, Eds. (Elsevier, Amsterdam, 2nd ed., 2018) pp. 275–318.

W. Staal, Ph.D. Dissertation, “Gradient Polymer Elution Chromatography”, Eindhoven University of Technology, Netherlands (1996).

 is a scientific advisor in the Chemical Sciences Division of the Material Measurement Laboratory at the National Institute of Standards & Technology (NIST). He received his Ph.D. in analytical chemistry in 1996 and his B.S. in chemistry in 1991, both from the University of New Orleans (USA). From 1996 to 1998 he performed postdoctoral research for the U.S. Department of Agriculture, at the National Center for Agricultural Utilization Research. For the next six years he worked for Solutia (now Eastman Chemical), achieving the rank of Research Specialist. From 2004 to 2011 he was assistant professor of both analytical and materials chemistry in the Department of Chemistry and Biochemistry at Florida State University (FSU, USA). In September 2011 he joined NIST. His research interests are in the area of polymer characterization, in particular applying separation science to determining structure‑property relations of complex macromolecules, and in the fundamental aspects of separation and detection methods.

Featured Interview

LCGC Europe spoke to André Striegel about the role and benefits of interaction polymer chromatography (IPC) for polymer analysis.

Evolving Interaction Polymer Chromatography

Charged aerosol detection (CAD) is a powerful complement to ultraviolet (UV) absorbance and mass spectrometric (MS) detection for liquid chromatography (LC), particularly for analytes that have no UV chromophore, or do not ionize well by electrospray ionization. As with all detector types, successful implementation requires an understanding of behaviour and characteristics of the detector. In this instalment, we discuss the characteristic nonlinear response of CAD, and approaches to deal with this, as well as mobile-phase requirements to consider for successful separations involving CAD.

Ultraviolet (UV) absorbance and mass spectrometry (MS) detectors are by far the most commonly used detectors for liquid chromatography (LC). They work exceptionally well for both qualitative and quantitative methods in many application areas. However, a significant weakness of UV-absorbance detection is that not all molecules of interest have chromophores (that is, a part of a molecule that absorbs UV light) that absorb strongly enough to make UV detection a viable option, especially in applications where detection sensitivity is an important performance metric. Over the years this weakness of UV detection has motivated research and development aimed at more “universal” detection schemes for LC (1), including refractive index detection (RID), evaporative light scattering detection (ELSD), and charged aerosol detection (CAD). Although these detection schemes have the tremendous upside that they are less dependent on the presence of a chromophore in our analytes compared to UV detection, they have their own weaknesses, including significant response dependencies on temperature, nebulization process, analyte volatility, and mobile-phase composition, as well as occasional nonlinear dependence of the detector response on analyte concentration. For users who were trained using UV or MS detection, these characteristics of RID, ELSD, and CAD can be challenging. For this month’s instalment of “LC Troubleshooting,” I have invited Dr. Imad Haidar Ahmad, Associate Principal Scientist at Merck & Co., Inc., to join me in sharing some tips and tricks he has found useful for the routine use of CAD in the research and development laboratories where he has worked.

Charged aerosol detection (CAD) is a detection method that is widely and routinely used in the pharmaceutical and biopharmaceutical industries for analyses involving analytes with weak or no UV chromophore. Some well-known applications of CAD in these contexts are analyses for organic and inorganic counterions (2), sugars (3), and lipids (4). In addition to the characteristic of responding more uniformly to a broad range of analytes, CAD has other strengths, including ease of use, reliability, and wide dynamic range. These characteristics have earned CAD a place on the benches of many research and development laboratories. Moreover, CAD-based methods are also used in regulated quality control laboratories where methods are validated according to demanding ICH guidelines. Success in this environment requires good performance as measured by metrics—including accuracy, repeatability, sensitivity (as measured by limits of detection [LOD] and quantitation [LOQ]), and response linearity. However, as stated earlier, there are some characteristics of CAD in particular that require familiarization for those of us who were trained on UV and MS detectors. An excellent source of information on CAD is the recently published book edited by Gamache (5).

Coping with the Nonlinear Response of CAD

Modern UV-absorbance detectors used for LC exhibit a linear response over an analyte concentration range of more than four orders of magnitude (powers of ten). This makes quantitation very straightforward, as simple two‑parameter equations (the slope and intercept of a line) can be used to describe the calibration curve. With CAD, however, the detector response to analyte concentration is not linear, and more complex relationships must be used to describe calibration curves. Several approaches have been used to overcome this limitation for quantitative applications. These include using localized linear fitting over a narrow concentration range, and plotting the response data on a logarithmic scale (log[response] vs. log[concentration]), followed by a linear fit of the data over the entire concentration range. Neither of these are particularly attractive from a practical point of view, especially in regulated environments. The former approach is limited to a narrow concentration range, or establishing multiple calibration functions and choosing one that is most appropriate to the analyte concentration in each sample that is analyzed. The latter approach involves electronic data processing and transformation in spreadsheets, and these processes need to be reviewed, validated, and approved.

An alternative approach to expand the linear dynamic range of CAD is to transform the signal using a power function. Typically, a power function is implemented by raising the signal stream of the detector to a power 

. This approach can yield highly linear plots of peak area vs. analyte concentration; however, in practice, it has been a challenge to determine the value of n (referred to hereafter as the 

 [PFV]) that yields the best quantitative performance in terms of accuracy and precision. The optimum PFV can be obtained by studying the variation in the detector response factor (RF = peak area/analyte concentration) with analyte concentration. Specifically, we can look at the slope of RF vs. analyte concentration, and the relative standard deviation (RSD) of RF values over the range of concentrations studied. If the PFV were to perfectly linearize the detector response, we would expect both the slope and RSD to be zero. Figure 1 shows experimental results from a study of the dependence of this slope and RSD on PFVs. We see that the optimum PFV in this case is about 1.4, corresponding to the lowest RSD value and a slope that is closest to zero.

Figure 2 shows that, at the optimum PFV of 1.4, the relationship between peak area and analyte concentration is clearly more linear than when a PFV of 1.0 is used (that is, the default value in the software).

The conventional approach to obtain the optimum PFV involves acquiring peak areas for a standard at several concentrations, using several different PFVs, as shown in Figure 1. In our own work we have demonstrated the ability to predict the optimum PFV using only the CAD signal acquired with PFV =1.0 (6). Recently, changes to commercially–available software for use with CAD have made determination of the optimum PFV much more user friendly.

Mobile Phase Considerations that Promote Success with CAD

As discussed above, a significant virtue of CAD is that it can “see” some compounds that are practically invisible to UV‑absorbance detectors. The flipside of this characteristic of CAD, though, is that it will also “see” impurities in the mobile phase, or interferences in the sample, that would otherwise not be detected by UV‑absorbance detectors if they do not contain chromophores. This is similar to the dilemma we often face with MS detection with electrospray ionization for LC separations—that it is exquisitely sensitive for detection of low concentration ionogenic compounds, but it is also very prone to detection of mobile‑phase impurities that can give rise to high levels of detector background. Given that CAD is most sensitive to non-volatile compounds, the background signal in CAD is strongly affected by the concentration of non-volatile impurities in the mobile phase. Therefore, many of the recommendations made regarding use of LC systems with MS detection and preparation of mobile phases for LC–MS also apply when using CAD. These are listed below:

When switching from the use of non-volatile buffers with UV‑absorbance detection to volatile buffers and CAD, be sure to thoroughly flush (30–60 min) the LC system with the new, volatile buffers before connecting the CAD to the column outlet.

Use ultrapure water with low conductivity (18.2 MΩ∙cm and less than 5 ppb total organic carbon [TOC]) for mobile-phase preparation to minimize the concentration of non-volatile impurities. It is recommended to use fresh water rather than water stored in bottles because of the possibility of sodium leaching from glass.

Use high purity organic solvents to prepare the mobile phase to minimize the concentration of non-volatile impurities. We find that LC–MS-grade solvents are generally adequate.

Use volatile, low molecular weight acids and bases for buffer constituents (examples include, but are not limited to, formic acid, acetic acid, ammonium formate, and ammonium acetate).

Filter samples prior to analysis whenever large particulate matter may be present in the sample.

Use flow rates within the recommended range of the detector.

Adjust the evaporation temperature of the detector to ensure efficient and consistent nebulization for different mobile-phase conditions.

Choose chromatographic conditions (such as temperature, mobile phase pH, or organic solvent) that minimize the likelihood of stationary phase bleed from the column.

Avoid sources of contamination from glassware, filters, stir‑bars, pH electrodes, and laboratory equipment.

If a high background signal is observed, we recommend to start troubleshooting by removing the column while running LC–MS-grade mobile phase without additives. Once the background has returned to normal, the other components (buffer and column) can be added back one at a time while tracking the signal until the source of the higher-than-expected background is discovered.

Figure 3 shows the impact on the CAD background signal of flushing the LC system and detector with high purity methanol and water after a period of use with buffered mobile phase. Comparing the background after flushing (black trace) to that before flushing (red trace), we see that there are decreases in short-term noise, magnitude of the background signal, and long-term drift of the signal.

Many LC practitioners were trained using UV absorbance or MS detectors. Other detectors, including CAD and ELSD, which are attractive for the analysis of semi- and non-volatile compounds because they do not require that analytes have a light-absorbing chromophore, have increased in prominence in recent years. In this instalment of “LC Troubleshooting,” we have highlighted some of the ways that CAD in particular is different from UV-absorbance detection, including nonlinear response to analyte concentration, and sensitivity to mobile-phase impurities. We hope that these tips are especially helpful for those new to CAD, and lead to development and implementation of successful methods involving this powerful detection technique.

K. Zhang, K.L. Kurita, C. Venkatramani, and D. Russell, 

J.P. Hutchinson, T. Remenyi, P. Nesterenko, W. Farrell, E. Groeber, R. Szucs, G. Dicinoski, and P.R. Haddad, 

Charged Aerosol Detection for Liquid Chromatography and Related Separation Techniques

 (John Wiley & Sons, Inc., Hoboken, New Jersey, 2017).

I.A. Haidar Ahmad, A. Blasko, J. Tam, N. Variankaval, H.M. Halsey, R. Hartman, and E.L. Regalado, 

 is Associate Principal Scientist and Supervisor in the Analytical Chemistry Enabling Technology group at Merck & Co., in Rahway, New Jersey, USA.

 is the editor of “LC Troubleshooting.” Stoll is a professor and the co-chair of chemistry at Gustavus Adolphus College in St. Peter, Minnesota, USA. His primary research focus is on the development of 2D-LC for both targeted and untargeted analyses. He has authored or coauthored more than 60 peer‑reviewed publications and four book chapters in separation science and more than 100 conference presentations. He is also a member of LCGC’s editorial advisory board.

Direct correspondence to: amatheson@mjhlifesciences.com

LC Troubleshooting

LC users that are new to charged aerosol detection (CAD) will find behaviours that are different from ultraviolet (UV) detectors. This article discusses some of the most important of these behaviours and provides tips to shorten the path to usable data.

Tips for LC Coupled with Charged Aerosol Detection

Because of advances in the fundamental understanding of gene therapy treatments and the 2017 regulatory approval of a recombinant adeno-associated viral (rAAV) therapy, interest in developing rAAV-based therapies has increased. However, for patients worldwide to truly benefit from these therapies, significant improvements in the efficacy and manufacturing efficiency of rAAV are essential. We have recently seen substantial increases of inquiries regarding liquid chromatography (LC) and liquid chromatography–mass spectrometry (LC–MS) methods for the characterization and analysis of rAAV from scientists who are supporting advancements in product development. This edition of “Column Watch” highlights selected LC and LC–MS methods that can be applied to the analysis of these complex products.

Gene therapy involves altering of the genome of selected cells within a patient (1). Although CRISPR/Cas9 has received widespread attention because of the groundbreaking work done by Emmanuelle Charpentier and Jennifer A. Doudna that resulted in their receiving the 2020 Nobel Prize in Chemistry, there are a number of other gene therapy methodologies currently being studied and developed (1).

One such approach is the in vivo delivery of a genetic payload to specific patient cells through the use of a non-replicating virus (viral vector). In this approach, the primary functions of the virus capsid are to shield the encapsulated RNA or DNA from enzymatic degradation and to deliver it to the interior of the cell. The predominant virus types adopted for therapeutic development are recombinant adeno-associated virus (rAAV), lentivirus, as well as adenovirus. In this article, we focus on liquid chromatography (LC) and liquid chromatography–mass spectrometry (LC–MS) characterization and analysis of rAAV, which continues to be the most widely used viral gene therapy platform (2). This focused perspective is predicated on the wide use of rAAV and our direct experience with rAAV samples.

Before considering analytical approaches, it is not only useful to study the nature of rAAV structure, but to also understand the general manufacturing process. A rAAV consists of a ~25 nm diameter icosahedral protein shell or capsid that is composed of three different viral proteins (VP1, VP2, and VP3) arranged in pentameric substructures and in its entirety each capsid contains 60 VP proteins in a 1:1:10 proportion of VP1:VP2:VP3 (3). The VP proteins are highly homologous, where VP2 (~65 kDa) is composed of the entire amino acid sequence of VP3 (~60 kDa) with an N‑terminal extension, and VP1 (~80 kDa) is an N‑terminal extension of VP2. The entire rAAV protein capsid shell has a molecular weight of approximately 3.7 MDa. Although also homologous, the varied serotypes of AAV are defined by variation in their capsid protein primary structures. It is these differences that can dictate viral tropism resulting in a gene therapy more selectively targeting the desired cells of a patient. The single‑stranded DNA (ssDNA) of therapeutic rAAV is located in the interior of the capsid structure and is composed of only the therapeutic gene and sequence regions that are required for the expression of the protein encoded by the therapeutic gene. Because of its size, the natural ssDNA packaging limit of rAAV is restricted to 4.7 kilobases (kb) with a molecular weight of approximately 1.55 MDa and extending that limit has been shown to be problematic (4). With the protein and ssDNA the total molecular weight of rAAV is approximately 4.3 MDa.

The rAAV production process can be quite varied and is a highly active area of study. A common production method is to transiently transfect a suspended host cell culture with plasmids that, when incorporated into the host cell, are designed to result in the host cell producing the rAAV capsid containing the desired therapeutic gene as ssDNA. In this approach, the plasmids used for the transient transfection may be considered critical reagents in the manufacturing process. As a result, these plasmid preparations may need to be well characterized and controlled (4). The harvesting and purification of rAAV generally includes cell lysis, clarification, and purification. Purification steps include a variety of extraction, immunoaffinity, chromatographic, and filtration steps (5). The influences that such a broad range of cell culture and purification steps can have on therapeutic rAAV quality are an important consideration when developing a structural characterization and product testing strategy.

In addition to the complexity of the rAAV structure, there are several additional challenges that are presented by rAAV samples compared to typical recombinant protein analyses. The low productivity of the rAAV manufacturing process results in limited sample quantities, and although this may not be a major issue for the final commercialized product, it can be very problematic when assessing process development samples where often only microgram quantities of protein and ssDNA components of rAAV are available to the analyst. Another challenge of rAAV samples is their low concentrations. As examples, a formerly approved intramuscular rAAV therapy was formulated at a concentration of 3 × 10

 vector genomes per millilitre (vg/mL) and the concentration of a current ocular rAAV therapy is 5 × 1011 vg/mL. This translates into protein concentrations of 20 μg/mL and 3 μg/mL, and ssDNA concentrations (assuming 4.7 kb ssDNA) of 13 μg/mL and 2 μg/mL, respectively. To add further complication, VP1 and VP2 each account for 8% of the total number of proteins, resulting in single to submicrogram levels of these two proteins in the formulated samples.

Although an in-depth review on all of the methods that may be included as part of rAAV product characterization and analysis is well beyond the scope of this article, we highlight several LC and LC–MS methods that we have applied to characterize rAAV, as well as some related studies. Included in this column are the anion‑exchange chromatography (AEC) and size‑exclusion chromatography (SEC) analyses of intact virus, the LC–MS analysis of intact VP proteins, and peptide mapping of VP proteins.

Anion-Exchange Chromatography for Intact Capsid Analysis

During the rAAV biomanufacturing process, empty capsids that do not contain the desired ssDNA are also produced. Without a consistent ssDNA payload, the efficacy and safety of rAAV therapies may be compromised (6). Several techniques have been used to quantify empty capsids in AAV samples, including analytical ultracentrifugation (AUC), static spectrophotometry, cryogenic electron microscopy (Cryo-EM), and charge detection MS (CDMS) (7–10). However, various challenges are presented by these methods including the cost of instrumentation, the expertise needed, the sample amounts required, low sample throughput, and susceptibility to interferences. As a result, the development of analytical methods to measure empty capsid levels remains an active area of research.

AEC was reported to have the potential of separating empty and full capsids based on differences in surface charge (9,11,12). It was also shown that compared to UV absorbance detection, fluorescence detection not only increases signal‑to‑noise ratio, but also has much less response bias between empty and full capsids, which is in part due to minimal intrinsic fluorescence of DNA (2). Figure 1 shows the optimized separation of AAV8 empty and full capsids using 5 µm non-porous strong anion-exchange particles with a quaternary ammonium ligand. It was also observed (data not shown) that AEC retention varied by rAAV serotype indicating that optimized method conditions are serotype dependent. In addition, the percentage of empty capsids in a sample can be determined once the relative intrinsic protein fluorescence response between the empty and full capsids is determined.

In the chromatographic overlay, we also observe that for the 98.5% empty sample (as confirmed by CDMS, data not shown), a substantial portion of the empty capsids were coeluted with the full capsid fraction (13). This is presumed to be because of the charge heterogeneity of the empty capsid proteins or the presence of oligonucleotide fragments on the empty capsid surface (14). These results indicate that an accurate assessment of partially full AAV8 capsid may not be possible to obtain via AEC. Despite this potential limitation, AEC may provide a useful approach to guide the process and product development and monitor product quality because of its simplicity, low sample demand, and precision.

SEC for Aggregate, Fragment, and Concentration Analysis

The levels of virus aggregation may be considered to be a critical quality attribute (CQA) in therapeutic rAAV preparations (4,15). SEC is most often employed for the analysis of aggregation and fragmentation of proteins such as IgG, due in large part to the precision and sample throughput of the analysis. There are limited reports of its use for the analysis of rAAV; in contrast, the routine use of dynamic light scattering (DLS) has been more reported (4,16,17). It is important to recognize that the upper analyte size limitation for SEC separations is approximately 100 to 200 nm in diameter, which is in the subvisible range (18). Above these limits, resolution will be low, and the analyte may be altered because of on-column shear forces, or trapped by the column. Therefore, these subvisible entities in rAAV preparations are analyzed using DLS, nanoparticle tracking analysis (NTA), and asymmetric flow field-flow fractionation (AF4) methods, among others (18,19).

As previously noted, rAAV has a total protein and ssDNA molecular weight of 4.3 MDa. However, the compact icosahedral structure has a diameter of only 25 nm (20). By comparison, IgG protein has a reported maximum lateral dimension of approximately 15 nm and its dimer and lower valency high molecular weight forms (HMW) are well resolved on a column packed with 200 Å pore‑size diol‑bonded silica‑hybrid SEC particles (21). Therefore, it was predicted that a high efficiency 450 Å pore-size version of that column should provide the required separation of rAAV HMW forms ranging in valency from dimer to tetramer. In addition to an increase in precision, SEC analysis can also provide information on the capsid fragments and other proteins in rAAV preparations in comparison to other methods.

Figure 2 shows the optimized non‑denaturing SEC separations for several rAAV-CMV-GFP serotypes containing ssDNA with CMV as the promoter sequence and green fluorescent protein (GFP) as the insert gene (22). To accommodate the low concentration of the rAAV samples, a detector monitoring intrinsic protein fluorescence was deployed (23). Fluorescence provides approximately 10-fold greater signal-to-noise in comparison to UV absorbance at 280 nm or 260 nm. Alternatively, the addition of a multiangle light scattering (MALS) detector to the SEC analysis of rAAV has also been shown to provide enhanced sensitivity for virus aggregates and in addition can provide accurate information on the capsid concentration and an estimate of empty capsid levels (17). Rapid estimates of these two sample characteristics can also be derived from an SEC separation using an UV absorbance detector or a fluorescence and UV absorbance detector in series (24,25).

Interrogation of proteins that compose the capsid of rAAV is an integral aspect of rAAV characterization. At a most basic level, the 1:1:10 proportion of VP1:VP2:VP3 proteins that comprise the capsid correlate with the effectiveness of an rAAV preparation (26). Changes in this proportion are not only observed for different cell culture systems but can also be a measure of process consistency. Traditionally, sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE) has been used to evaluate the relative VP protein abundance and more recently the use of SDS–capillary gel electrophoresis (SDS-CGE) has been adopted. In contrast, LC–MS separations of VP capsid proteins can not only confirm consistent VP1:VP2:VP3 ratios but can also provide useful information on VP degradation (such as deamidation and oxidation), post-translational modifications (PTMs) such as phosphorylation and SUMOylation, and engineered point mutations that may impact the efficacy of rAAV products (26,27). For rAAV subserotypes with engineered point mutations, LC–MS analysis of the VP proteins may also find utility as a subserotype identity test.

The most commonly used LC–MS separation method for the analysis of proteins is reversed-phase LC (RPLC), which has been effectively applied to the analysis of rAAV (28). Building on this work, a method was developed for the RPLC-MS analysis of several rAAV serotypes as shown in Figure 3 (29). The separation was optimized for a set of rAAV serotypes using a 300 Å pore size C4 column and difluoroacetic acid (DFA) as a mobile‑phase modifier and analyzed using a compact time‑of‑flight LC–MS system (LC–TOF‑MS). When deployed, fluorescence optical detection was favoured over UV absorbance due to the nearly 50‑fold greater sensitivity. Figure 3 also shows representative mass spectra for the major components observed, including significant levels of phosphorylation of the VP1 and VP2 proteins.

A well-executed alternative approach to LC–MS analysis of the VP proteins using a hydrophilic interaction liquid chromatography (HILIC) separation with MS detection has also been reported (30). This separation makes use of a 300 Å pore size, silica‑hybrid based amide HILIC column that was optimized for the separation of glycosylated peptides and proteins (31,32). The reported HILIC separation can be applied to a wide range of rAAV serotypes, and although trifluoroacetic acid (TFA) is used as a mobile‑phase modifier, MS sensitivity is enhanced by modifying the MS desolvation gas with propionic acid and isopropanol.

As discussed for intact VP proteins, the LC–MS identification and analysis of PTMs and protein degradation can provide useful information to both guide rAAV product development and monitor product quality (27). However, to evaluate site‑specific variants, peptide mapping may likely be required. Additionally, a peptide mapping procedure can also be deployed to monitor for host cell proteins (HCP) that may be of concern in rAAV products (33). A peptide mapping procedure involves the proteolytic digest of a protein and may also include disulfide bond reduction and alkylation steps. The LC or LC–MS analysis of the digested sample can be comprehensive to cover as much of the protein amino acid sequence as possible or be focused on the peptides and peptide variants of interest.

In comparison to typical recombinant protein therapeutics, rAAV samples present several significant challenges during peptide map development. As noted previously, these samples are much lower in concentration in addition to presenting the difficulties presented by the homology and relative abundance of the VP proteins. Furthermore, the amount of sample that can be used for these analyses is often limited and rAAV samples are often formulated with surfactants that can significantly degrade rAAV MS signal quality. Another choice that can be made for this analysis is whether to map the capsid proteins as a single digest or to purify VP1, VP2, and VP3 prior to digestion.

Given the limited amounts of sample that are often available, a single‑digest mapping method was developed. The disadvantage of this approach is that variants in the conserved sequence domains cannot be attributed to a VP protein type. The use of a single digest also presents a dynamic range challenge for UV absorbance detection of the peptide map given the greater than 10‑fold abundance of VP3 domain peptides vs. the peptides that comprise the N-terminal extensions of VP2 and VP1. With these considerations, a tryptic digest of 1.25 μg of rAAV was optimized and then analyzed using an LC–TOF‑MS system without a UV absorbance detector and resulted in sequence coverages of greater than 95% for VP1, VP2, and VP3 (Figure 4).

As the clinical interest in rAAV gene therapies increases it will be critical to optimize the product efficacy and the efficiency of the rAAV manufacturing process. LC and LC–MS technologies have a long history of use in the development, characterization, and quality testing of recombinant protein therapies and will likely be valuable tools in the advancement of rAAV‑based gene therapies, as well as other viral and nonviral gene therapy development. Indeed, as fairly recently‑introduced therapy modalities, the regulatory analytical requirements for these products are only starting to mature (34). This article has provided a brief introduction to the potential application of some of these LC and LC–MS methods to rAAV characterization and analysis. In addition, we have also seen that in comparison to recombinant protein analysis, the deployment of LC and LC–MS methods is further complicated by numerous factors such as more complex samples and limited sample amounts.

The authors would like to thank Benjamin E. Draper of Megadalton Solutions for providing CDMS results for the rAAV8 samples used for AEC method development.

C.E. Dunbar, K.A. High, J.K. Joung, D.B. Kohn, K. Ozawa, and M. Sadelain, 

C. Wang, S.H.R. Mulagapati, Z. Chen, J. Du, X. Zhao, G. Xi, L. Chen, T. Linke, C. Gao, A.E. Schmelzer, and D. Liu, 

Y.Z. Tan, S. Aiyer, M. Mietzsch, J.A. Hull, R. McKenna, J. Grieger, e

D. Blessing, N. Deglon, and B.L. Schneider, in 

Scalable Production and Purification of Adeno-Associated Viral Vectors (AAV) 

(Springer, New York, 2018), pp. 259–274.

K. Gao, M. Li, L. Zhong, Q. Su, J. Li, S. Li, R. He, Y. Zhang, G.Hendricks, J. Wang, and G. Gao, 

B. Burnham, S. Nass, E. Kong, M. Mattingly, D. Woodcock, A. Song, S. Wadsworth, S.H. Cheng, A. Scaria, and C.R. O’Riordan, 

J.M. Sommer, P.H. Smith, S. Parthasarathy, J. Isaacs, S. Vijay, J. Kieran, S.K. Powell, A. McClelland, and J.F. Wright, 

S. Subramanian, A.C. Maurer, C.M. Bator, A.M. Makhov, J.F. Conway, K.B. Turner, J.H. Marden, L.H. Vandenberghe, and S.L. Hafenstein, 

E.E. Pierson, D.Z. Keifer, A. Asokan, and M.F. Jarrold, 

E.D. Horowitz, K.S. Rahman, B.D. Bower, D.J. Dismuke, M.R. Falvo, J.D. Griffith, 

X. Fu, W.C. Chen, C. Argento, P. Clarner, V. Bhatt, R. Dickerson, et al. 

Anion-exchange chromatography for determining empty and full capsid contents in adeno-associated virus 2020.

G. Qu, J. Bahr-Davidson, J. Prado, A. Tai, F. Cataniag, J. McDonnell, 

AAV critical quality attribute analysis by SEC-MALS. 

J.F. Carpenter, T.W. Randolph, W. Jiskoot, D.J. Crommelin, C.R. Middaugh, and G. Winter, et al. 

S.K. Vernon, J.T. Stasny, A.R. Neurath, and B.A. Rubin, 

L.J. Harris, E. Skaletsky, and A. McPherson, 

W. Xiong, A.E. MacColl Garfinkel, Y. Li, L.I. Benowitz, and C.L. Cepko,

Size-Exclusion Chromatography Analysis of Adeno-Associated Virus (AAV) Preparations Using a 450 Å Diol-Bonded BEH Column and Fluorescence Detection 2020. 

Rapid AAV Concentration Determination Using Size-Exclusion Chromatography with Fluorescence and UV Dual Detection 2020. 

Gene Therapy and AAV: Advancing Analytical Characterization and Product Understanding for Late-Stage Success 

S. Pacouret, M. Bouzelha, R. Shelke, E. Andres-Mateos, R. Xiao, A. Maurer, et al. 

B. Mary, S. Maurya, S. Arumugam, V. Kumar, and G.R. Jayandharan, 

X. Jin, L. Liu, S. Nass, C. O’Riordan, E. Pastor, and X.K. Zhang, 

X. Zhang, S.M. Koza, Y.Q. Yu, and W. Chen, 

Optimizing Adeno-Associated Virus (AAV) Capsid Protein Analysis Using UPLC and UPLC-MS. 

Waters Application Note. 2020, EN72000686.

A.P. Liu, S.K. Patel, T. Xing, Y. Yan, S. Wang, and N. Li, 

M.A. Lauber, S.M. Koza, P.C. Iraneta, and K.D. Wyndham, 

A look at the LC and LC–MS methods that are helping to overcome challenges in sample analysis, and how this can make adeno‑associated virus (AAV)–based gene therapy more accessible.

LCGC Europe-12-01-2020

The Determination of Trace Per- and Polyfluoroalkyl Substances and Their Precursors Migrated into Food Simulants from Food Contact Materials by Liquid Chromatography–MS/MS and Gas Chromatography–MS/MS

Evolving Interaction Polymer Chromatography

Tips for LC Coupled with Charged Aerosol Detection

Applying LC and LC–MS Methods to the Characterization and Analysis of rAAV