Novel remediation of per- and polyfluoroalkyl substances (PFASs) from contaminated groundwater using Cannabis Sativa L. (hemp) protein powder

Chemosphere 229 (2019) 22e31

Contents lists available at ScienceDirect

Chemosphere journal homepage: www.elsevier.com/locate/chemosphere

Novel remediation of per- and polyfluoroalkyl substances (PFASs) from contaminated groundwater using Cannabis Sativa L. (hemp) protein powder Brett D. Turner*, Scott W. Sloan, Glenn R. Currell Centre of Excellence for Geotechnical Science and Engineering, Civil Surveying and Environmental Engineering, The University of Newcastle, University Drive, Callaghan, N.S.W., 2308, Australia

h i g h l i g h t s

g r a p h i c a l a b s t r a c t


Hemp protein powder has been shown to be highly effective at removing PFAS from actual contaminated groundwater.
Hemp was found to remove more PFAS compared to soy, lupin, whey, pea, and egg proteins when normalised for protein content.
Reaction kinetics show rapid PFAS removal with very good removals (>98%) attained in approximately one hour for PFOS.
In the presence of hemp protein powder, increasing salinity appears to favour the PFAS removal.
FTIR analysis shows hydrogen bonding and hydrophobic interactions are significant to the PFASprotein reaction.

a r t i c l e i n f o

a b s t r a c t

Article history: Received 12 December 2018 Received in revised form 16 April 2019 Accepted 18 April 2019 Available online 28 April 2019

Perfluoroalkyl and polyfluoroalkyl substances (PFASs) are a group of environmentally persistent, manmade chemicals used in many industrial products and everyday consumer items. Of the plant proteins trialled, those of hemp (Cannabis sativa L.) were found to be far superior for PFAS removal than the next best protein, soy. The use of hemp plant proteins as a possible pump-and-treat solution to PFAS remediation from groundwater has been successfully demonstrated with very good removals (>98%) of the main contaminants of PFOS and PFHxS in approximately 1 h of contact time, with salinity enhancing removal of short chain PFAS. Changes to the secondary structure of hemp proteins was found using FTIR spectroscopy analysis and calculated based on the integrated areas of the amide I component bands. The amount of b-turns increased from ~9.3% (control) to 44.1% (undiluted groundwater); with a decrease in random coils (25.6e8.6%); a-helix (19.3e8.6%) and b-sheets (38.8e23.1%). These changes indicate that hemp proteins partially unfold during the reaction with PFAS with other FTIR evidence suggesting sorption at hydrophobic sites of the protein as well as with the side chains of the amino acids aspartic and glutamic acid. The absence of these side chains in soy protein, as evidenced from FTIR and amino acid analysis, being part of the reason why soy removed less (approx. half) of the S(PFHxS þ PFOS) load

Handling Editor: Tsair-Fuh We dedicate this paper to Professor Scott Sloan, friend and mentor (1954e2019). Keywords: PFAS PFOS PFOA

* Corresponding author. E-mail address: [email protected] (B.D. Turner). https://doi.org/10.1016/j.chemosphere.2019.04.139 0045-6535/Crown Copyright © 2019 Published by Elsevier Ltd. All rights reserved.

B.D. Turner et al. / Chemosphere 229 (2019) 22e31 Sorption Protein Hemp

23

when compared to hemp. The findings reported here will lead to new, environmentally friendly methods for PFAS remediation. Crown Copyright © 2019 Published by Elsevier Ltd. All rights reserved.

1. Introduction

1.1. Background: plant proteins for PFAS removal

Perfluoroalkyl and polyfluoroalkyl substances (PFASs) have been widely used for various purposes, including food wrappers, nonstick cooking utensils, carpet and furniture protectants, and in fire-fighting foams as they are highly effective against hydrocarbon fuel fires. Consequently, significant numbers of fire-fighting training facilities around the world have been identified as having soil, ground and surface water contaminated by PFASs. Between 2009 and 2017, studies have discovered 455 new PFAS compounds (Xiao, 2017) with PFOA (perfluorooctanoic acid), PFOS (perfluorooctanesulfonic acid), and PFHxS (perfluorohexane sulfonic acid) the only species to have regulated concentrations for drinking water. PFOA is found in the blood of an estimated 98% of Americans (Nicole, 2013), and is often found along with PFOS (Steenland et al., 2010) and PFHxS with blood serum half-lives of 3.5, 4.8 and 7.3 years respectively (Olsen et al., 2007). PFOA has been shown to cause liver, testicular, and pancreatic tumors in rats, however human/PFOA studies are sparse (Barry et al., 2013) with human/PFOS/ PFHxS studies are even more limited. Consequently, current official health guidelines state that there is no consistent evidence that PFOS or PFOA cause adverse human health effects (NSW Government Health, 2017) despite PFOA (and PFOS) being classified as “likely to be carcinogenic in humans” by the USEPA Science Advisory Board (Steenland et al., 2010). Due to the long blood serum half-lives of these chemicals, continued exposure to even relatively low levels in drinking water, food packaging, carpets etc. can substantially increase PFAS concentrations in humans over time and therefore may increase the risk of health effects (Post et al., 2012). To minimise the risk to the population, many water authorities and governments around the world are implementing treatment systems for municipal water supplies. One process commonly implemented for PFAS removal from solution utilises granulated activated carbon (GAC), an effective substrate for the removal of long-chain PFASs. GAC is less effective for the treatment of the shorter chain PFASs, for example PFBS (4 carbon chain length (Eschauzier et al., 2013);). Accordingly, GAC filtration is often used in conjunction with other treatment methods such as reverse osmosis (RO) resin to increase the number of PFASs removed during treatment. Combining GAC filtration with reverse osmosis adds significantly to the complexity and costs of PFAS remediation with the process generating by-products including PFAS contaminated GAC, and a hyper-saline PFAS contaminated waste solution created during RO resin regeneration. Consequently, the problem remains on how to deal with the PFAS containing waste materials left over from the filtration technologies (Eschauzier et al., 2013). Considered almost non-degradable in nature and highly mobile, PFAS contamination is extremely difficult and expensive to remediate with the high costs exacerbated by the fact there are few natural or engineered processes that can treat the contamination (Higgins and Field, 2017). The author proposes here, for the first time, the use of plant proteins (albumin, edestin etc; provisional patent granted) as an effective, natural method for the treatment of PFAS contaminated waters.

Human Serum Albumin is the major transport protein in blood plasma and is capable of binding and transporting a wide variety of drugs, and as such, has been intensively studied by the biomedical/ pharmacological industry as an effective mode of delivery to target cells (Ng and Hungerbühler, 2014). With a molecular weight of ~66,500 Da, HSA contains 585 amino acids and is a flexible, heart shaped macromolecule composed of three structurally similar globular domains (I, II, & III), each of which contains two subdomains (A & B). According to Sudlow's nomenclature (Sudlow et al., 1975) bulky anions bind to site I, whereas site II is preferred by carboxylates (Baroni et al., 2001) (e.g. ibuprofen). Having both charged and uncharged surface sites, proteins can undergo interactions with ligands via electrostatic or non-electrostatic (hydrophobic/hydrophilic) physisorption as well as site specific chemical adsorption (chemisorption (Parsons et al., 2017);). These properties therefore indicate that proteins may be used to target PFAS contaminants. Han et al. (2003) and Wu et al. (2009) reported that PFOA associates with Sudlow site III at the internal cavities as well as the outer surfaces. Luo et al. (2012) in crystallographic experiments, and Salvalaglio et al. (2010) in computer modelling simulations, reported that perfluorooctane sulfonic acid (PFOS) formed a strong salt bridge (bond) to HSA (Sudlow site II) being fixed to the carboxylic head group of Myristate (a fatty acid) as well a hydrogen bonding, and hydrophobic interaction of the PFOS tail to the methylene tail of another Myristate binding site. For the sulfonate head group and trifluoromethyl tail of PFOS also bonded in a similar manner but at the interface of subdomains IIA and IIB. Luo et al. (2012) did not find that Sudlow site I could bind PFOS possibly because of the number of polar (hydrophilic) residues which repulse the non-polar (hydrophobic) fluorinated PFOS tail. Zhang et al. (2009) assessed the binding of PFOS to HSA and found that PFOS acts as a mild protein denaturing agent producing an inhibition of the natural transport function of HSA in the blood. The partial denaturation of the protein may be in part responsible for the extremely high PFOS adsorption ratio (45:1, PFOS to HSA molecules) found (Zhang et al., 2009). This supports experimental work (Wu et al., 2009) which shows that PFOS bioaccumulates more than PFOA due to the higher energy of the non-polar interactions of PFOS with HSA (due to the fact that PFOS contains one CF2 group more than PFOA). Albumin and edestin are the two primary proteins found in the seed of the hemp (Cannabis Sativa L.) plant. With low (<0.3%) concentrations of Tetrahydrocannabinol (THC), a psychoactive chemical commonly found in the cannabis plant family, the seed has a very high protein content at approximately 25% (RodriguezLeyva and Pierce, 2010) of which approximately 75% is made up of edestin, and 25% albumin (Aluko, 2017). Although the dominant protein in hemp, little information is available concerning the physiochemical and functional properties of edestin (Tang et al., 2006). This is distinctly different from the case of Soybean protein which is usually composed of glycinin (legumin) and b-conglycinin (vicilin) with a similar (50:50) content with very little

24

B.D. Turner et al. / Chemosphere 229 (2019) 22e31

albumin present (Shewry et al., 1995). Soy protein (Glycinin) in its native state is compactly folded with the hydrophobic amino acid side chains located in the interior of the molecule forming a hydrophobic region, whereas the hydrophilic ones are located on its surface forming a tube like structure. Thus, it is expected that the properties of hemp protein might be remarkably different from those of soy protein. In this paper we determine the ability of six different plant protein powders to remove PFAS chemicals from contaminated groundwater for potential application to a pump-and-treat engineered application. Fourier Transform Infrared (FTIR) spectroscopy, a technique very sensitive to biomolecular changes (Li et al., 2005) is used to investigate the possible mechanism of PFAS-protein interactions. 2. Materials & methods 2.1. PFAS nomenclature

whey) were sourced from various commercial suppliers. The total protein percentage and amino acid content of each protein powder used was determined by the NATA accredited National Measurement Institute (NMI) Laboratories, Melbourne, Australia. To directly compare the PFAS removal efficiency of each protein, results were normalised for protein content as follows. The percentage removal (S) of each PFAS was calculated using:

 S¼

  100

(1)

where Ci and Ceq the initial and equilibrium PFAS concentrations (mg/L). PFAS removal data was normalised for protein content and subsequently referenced to hemp removal to enable direct comparison of PFAS removal between samples. The normalised protein sorption (SP (mg/g)) was calculated using:

Sp ¼

The term PFAS describes all per- or polyfluoroalkyl species, however this can be further divided into classes (perfluorocarboxylic acids (PFCAs), perfluorosulfonic acids (PFSAs), sulfonamides, and telomeres) and then individual chemicals as shown in Tables SIe1 (supplementary information). In this study no chemicals belonging to the latter two classes were detected i.e. all were below the laboratory limit of reporting (
Ci  Ceq Ci

  Ci  Ceq M  %P

(2)

where M is the solid to liquid ratio (g/L) and %P the protein percentage of the solid as reported above. Sp was then used to calculate the protein removal efficiency (PRE; Eqn (4)) which compares the removal efficiency of each protein powder to that of hemp.

PRE ¼

SP SPðHempÞ

(3)

2.2. PFAS contaminated groundwater With permission from the Australian Department of Defence, approximately 50 L of groundwater known to have some of the highest PFAS groundwater concentrations was supplied in PFAS approved polypropylene drums. The monitoring well, located down gradient of the base is located on a private residential property and is therefore unable to be identified and will therefore be referred to as MWx herein. The groundwater samples (Table 1) were thoroughly mixed by decanting 50% of each drum into the next etc. several times. Aliquots of the groundwater was then diluted into 5 L polypropylene jars using de-ionised water (18 MUcm1) to achieve a range of dilutions (10%, 25%, and 100% (undiluted)). 2.3. Proteins Four plant protein powders from hemp (Cannabis Sativa L.) protein powder (HPP), soy protein isolate (SPI), pea protein isolate (PPI), and lupin along with two animal protein powders (egg and

2.3.1. Binding forces Proteins, as biomacromolecules, have cavities and surfaces which give rise to hydrophobic forces, electrostatic interactions, hydrogen bonding, Van der Waals forces etc. (Li et al., 2014) all of which are encompassed by the terms physisorption or chemisorption. The term “adsorption” as defined by Sposito (1986) is the interfacial accumulation of a surface complex without development of a three-dimensional arrangement. This definition is not used here as it is unknown at this stage whether a three dimensional (multilayer) adsorption occurs in the presence of proteins. Therefore, due to the multitude of possible protein-ligand interactions we use the generic term “sorption” to encompass the terms physisorption and chemisorption. 2.4. Batch tests Batch tests were conducted in PFAS approved (polypropylene)

Table 1 Major PFAS analytes found in groundwater (natural, undiluted) from monitoring well MWx, Williamtown Airforce Base, Australia. Initial pH ~6.8. Electrical conductivity ~0.2 mS/cm (natural salinity) and ~49 mS/cm (high salinity). *indicates the removal of a single outlier data point.

PFAS Grouping Perfluoroalkane Sulfonates (PFSAs)

Perfluoroalkyl Carboxylates (PFCAs)

PFAS PFBS PFPeS PFHxS PFHpS PFOS S(PFHxS þ PFOS) PFBA PFPeA PFHxA PFHpA PFOA SPFAS (TOTAL)

Concentration (Natural Salinity) (mg/L)

Concentration (High Salinity) (mg/L)

-TOP (n ¼ 8) 3.19 ± 0.51 3.23 ± 0.54 29.7 ± 9.12 3.52 ± 0.83 101.11 ± 21.08 130.81 ± 30.04 1.14 ± 0.19* 1.88 ± 0.68* 9.95 ± 2.11* 2.49 ± 0.45 6.92 ± 1.64 165.91 ± 34.50

-TOP (n ¼ 3) 3.48 ± 0.60 4.02 ± 0.75 30.87 ± 6.20 4.23 ± 0.45 93.57 ± 15.87 124.4 ± 20.74 1.00 ± 0.00 2.00 ± 0.31 11.77 ± 1.96 2.94 ± 0.47 6.53 ± 0.64 196.67 ± 25.33

þTOP (n ¼ 6) 4.61 ± 1.72 4.86 ± 0.85 25.72 ± 2.91 4.59 ± 0.77 93.97 ± 12.85 119.67 ± 12.29 6.98 ± 3.32 10.20 ± 2.49 38.3 ± 14.06 3.66 ± 0.44 8.11 ± 1.13 201.15 ± 24.52

þTOP (n ¼ 1) 4.06 3.96 31.9 4.5 129 161 <0.01 6.43 50 3.73 10.6 244

B.D. Turner et al. / Chemosphere 229 (2019) 22e31

plastic vials, capped and left for at least 3 days in an end-over-end stirrer to equilibrate at ~20  C. As all tests were preliminary proofof-concept investigations (i.e. that plant proteins can be used as a PFAS remediation substrate) only one nominal solid-to-liquid ratio (S:L; x±0.01 g/L)) was used for all experiments, however due to patent restrictions this cannot be defined here. Control samples were included in each batch test using de-Ionised (DI) water or DI water made up to ~45 mS/cm with KCl for high ionic strength tests. For comparison activated carbon (manufactured by Norit® (CAS number 7440-44-0) and supplied by Sigma Aldrich) was used at identical solid to liquid ratios at low and high ionic strength. At the end of the equilibration period, samples were centrifuged at 20  C and the supernatant decanted into clean polypropylene jars. Residual solid plant material was filtered and freeze dried for FTIR analysis.

25

Nicolet iS50 FTIR spectrometer (Thermo Fisher Scientific, Sydney, Australia). No sample preparation is required for ATR-FTIR and provides unique molecular chemical information on the bonding structure within a sample (Thumanu et al., 2015). The ATR-FTIR spectra of the freeze dried protein samples were determined over the 4000-400 cm1 wavenumber range with 64 scans of both background and sample in the transmittance mode, at a resolution of 4 cm1. The resultant spectra were analysed using OMNIC™ software (Thermo Fisher Scientific) including pre-processing baseline and advanced ATR correction. Spectral processing was achieved using Gaussian peak fitting and second order derivatisation with the Savitzky-Golay six datapoint algorithm using OMNIC software (Thermo Fisher). This process allows for the interpretation of protein secondary structure (Kong and Yu, 2007). 3. Results and discussion

2.4.1. Batch kinetics Batch kinetic experiments under the same conditions as described were also completed at 1, 3, 24, 72 and 142 h to determine PFAS equilibration time. Final pH was obtained on an aliquot of supernatant using an Orion Star A215 benchtop meter with an Orion (9165BN) pH electrode and Orion Star Com software (Thermo Fisher Scientific, Sydney, Australia). pH electrode (Orion 9165BN) calibration was completed using pH 4, 7 and 10 NIST buffers until a slope of 92e102% was obtained as per manual instructions. EC calibration was done using an Orion Star A322 m and a 1413 mS/cm standard as per manual instructions. All PFAS analyses were done at ALS laboratories, Sydney (NATA accredited) using modified USEPA method 315 for a standard suite of 28 PFAS analytes, however only those detected as listed in Table 1 are reported here. Reported analyte concentrations were corrected by the laboratory from recovery analysis using 13C isotopically labelled surrogates (ALS laboratories). All experimental samples were refrigerated until transfer ALS laboratories typically the same day (or <24 h). Where indicated, the Total Oxidisable Precursor (TOP) analysis was also done by the analytical laboratory. The TOP analysis was developed to quantify the unknown precursor PFAS compounds which can potentially transform to PFCA or PFSA end-products. The method uses aggressive oxidation which converts PFAS precursors into PFCAs (Houtz and Sedlak, 2012) however the standard method used by laboratories are limited to the detection of currently identifiable species. Consequently currents methods do not allow for a full quantitative conversion to known measurable species (Robel et al., 2017) and as a result a total PFAS mass balance cannot be accurately quantified. 2.4.2. Dissolved protein partitioning To determine if any PFAS are partitioned to possible soluble proteins an experiment was done whereby hemp protein powered (100 g/L) was added to DI water and mixed in an end-over-end stirrer for 72 h. The solution was then centrifuged and the supernatant decanted. A solution of PFOS of a known concentration (~760 mg/L) was made from analytical grade Heptadecaflourooctanesulfonic acid (40%; Sigma Aldrich). This was added (50% v/v) to the hemp supernatant solution to give a final concentration of ~340 mg/L PFOS. Control samples (50% v/v) PFOS:DI water; and hemp supernatant:DI water were also made. All three samples were left to further mix for an additional 3 days, centrifuged and analysed at ALS laboratories, Sydney. 2.5. ATR-FTIR analysis Attenuated Total Reflectance (ATR) Fourier Transform Infrared Spectrophotometry (FTIR) experiments were recorded using a

3.1. PFAS contaminated groundwater Table 1 shows the analysis results of undiluted groundwater MWx for high salinity (~49 mS/cm) and natural (0.2 mS/cm) samples. Concentrations pre- and post- Total Oxidisable Precursor (TOP) analysis (herein, -TOP and þTOP respectively) are also shown as the average and standard deviation of the number of samples (n) analysed. The relative standard deviation (%RSD) can be calculated from equation (4) and, for the natural salinity sample, ranged up to ~23.7% for S(PFHxS þ PFOS) (-TOP); and ~20.7% for SPFAS (-TOP). Of note, PFBA, PFPeA and PFHxA all had %RSD >50, however application of Grubb's Test (Graphpad Prism (Graphpad, 2017)) for statistical outliers indicated that at the P < 0.01 significance level, there was a single outlier. Removal of this resulted in the reduction of % RSD to 16.7% (PFBA), 36.2% (PFPeA), and 21.2% (PFHxA).

 SD  100 Mean

 %RSD ¼

(4)

where SD is the standard deviation of the data from the mean. Comparison of natural and high salinity samples (eTOP) shows no significant changes in PFAS concentrations due to ionic strength effects. Addition of the þTOP analysis shows a large increase in the PFCA grouping with PFHxA and PFPeA increasing by a factor of ~3e4 (Fig. 1) indicating the presence of potential precursors (Casson and Chiang, 2018). Table 1 also shows that following TOP analysis, there was a decrease in PFOS concentrations in the natural groundwater sample (within error, Fig. 1) and at high salinity, the PFOS concentration increased from ~124 to ~161 mg/L. This may be due to known issues with TOP analysis including the fact that controlling the oxidation process is difficult due to high concentrations of PFAS or competition from organic carbon, resulting in the PFAS conversion being incomplete and therefore mixed results (ALS Environmental, 2016; Banzhaf et al., 2017). 3.1.1. Batch tests The order of increasing protein content was found to be lupin (38%), hemp (48%), pea (78.3%), egg (81.1%), whey (82.9%) and soy (84.2%). Soy, pea, egg, and whey all have a high protein content (~80%) as these samples are protein isolates meaning they have been commercially concentrated. The two main indicators of PFAS removal performance are (i) the total sum of all PFAS chemicals (SPFAS), and (ii) the sum of PFHxS and PFOS (S(PFHxS þ PFOS)). Fig. 2 shows that prior to protein normalisation (Fig. 2A), that hemp achieves the highest removals with a removal of 95.2% for (SPFAS) and 97.9% for S(PFHxS þ PFOS) with whey protein performing the worst. It appears as though soy and pea have similar removals to hemp at ~82% for (SPFAS) and ~97% for S(PFHxS þ PFOS). In

26

B.D. Turner et al. / Chemosphere 229 (2019) 22e31

Fig. 1. PFAS concentration changes due to Total Oxidisable Precursor (TOP) analysis. (A) Perfluoroalkyl Sulfonic Acids (PFSAs); (B) Perfluoroalkyl Carboxylic Acids (PFCAs). Low (natural) salinity. Pre- and post TOP analysis defined as eTOP and þTOP respectively.

Fig. 2. Sum of PFAS and PFHxS þ PFOS removal (%) from groundwater by plant protein powders (A); and Protein Removal Efficiency (PRE; unitless) normalised to hemp protein powder. All results for eTOP analysis.

comparison, Norit® GAC showed very similar results with 95.5% for (SPFAS) and 95.3% for S(PFHxS þ PFOS) or <3% difference to that of hemp under the same conditions. After normalisation for protein content (Eqns. (1)e(3); Fig. 2B), it becomes clear that with respect to hemp, the next best performing plant protein was that of soy with a PRE (Eqn. (3)) of 0.18 and 0.52 times that of hemp for SPFAS and S(PFHxS þ PFOS) respectively. Equilibrium pH was found to range from ~4.98 (hemp) to ~7 (soy). Given that the pKa values of the PFCAs with eleven or less carbon atoms (99% ionised; Eqn. (5)). Consequently, PFASs may undergo electrostatic sorption with proteins. Hemp protein has been observed to have a negative charge (zeta potential) at pH 7 which becomes increasingly negative with decreasing pH due to the increased exposure of hydrophobic amino acids (Teh

et al., 2016). Therefore, in the absence of bridging cations, the possibility of electrostatic interaction may be reduced because of the predominately negative protein surface charge at the observed pH. This is supported by Zhang et al. (2013c) in their study of PFAS removal with human liver fatty acid protein, where it was found that the driving force for the protein-PFAS reaction was predominately hydrophobic interactions.

 %I ¼

100 1 þ 10xðpHpKaÞ

 (5)

Equation (5) is derived from the Henderson-Hasselbalch equation where %I is the percentage ionisation of the acid (x ¼ 1) or base (x ¼ 1) and pH/pKa are as defined above. 3.1.2. Batch kinetics As hemp protein powder was determined to be the best

B.D. Turner et al. / Chemosphere 229 (2019) 22e31

substrate for PFAS removal from the groundwater sample, all subsequent experiments focus on its use. As such all results presented are “as is” and have not been corrected for protein content. Fig. 3 shows the post (þ)TOP analysis removal kinetics for PFAS from undiluted (100%) MWx groundwater over 144 h of contact time. The rate of PFSA removal (Fig. 3A) appears to increase with increasing carbon chain length (i.e. increasing hydrophobicity) with the shorter four carbon (4C) PFBS experiencing ~80% removal within 3 h followed by PFPeS (5C) with ~76% removal, PFHxS (6C) with ~85% removal and PFHpS (7C) & PFOS (8C) with ~96% removal in 1 h (the shortest time step measured) increasing to >99%. The S(PFAS) and S(PFHxS þ PFOS) attain equilibrium within 3 h. However, the rate effects may also be simply due to the much smaller initial concentrations of the shorter chain species. This will be addressed in a forthcoming publication. For the PFCAs (Fig. 3B) there does not appear to be a trend in the rate of removal as a function of carbon chain length as PFPeA (5C) appears to reach equilibrium removal (~92%) within 3 h in comparison to PFOA (C8) at ~86% removal. Of note is the apparent decreasing trend in removal of PFBA observed in Fig. 3B. Herein, solution pH was found to be 6.28 ± 0.08 over the first 20 h of contact time (not shown) following reduction to ~4.98. As discussed above, the low pKa values of PFASs (<3.5) would imply that at the observed pH range all detectable species would exist in their anionic state. The possibility of a decrease in PFBA removal due to anionic repulsion would be unlikely given that its pKa is ~0.4e0.7 (Goss, 2008) and would be >99% ionised (negative) and the protein surface would most likely be negative (Zhang et al., 2013c). The actual protein surface charge under the observed conditions would, of course, require further investigation. However as the experimental PFBS values in both the eTOP and þTOP analysis were all found to be considerably higher than the initial starting concentration, the observed removal is therefore attributed to issues with the TOP analysis (ALS Environmental, 2016; Banzhaf et al., 2017). Increasing the salinity (Fig. 4) of the PFAS solution appears to be favourable for the reaction as the equilibration times decrease, particularly for the shorter chain PFAS. For example after 1 h of contact time in the saline solution, the removal of PFBS (4C) increases by ~17.5% above that observed in the natural (non-saline) solution and, after 3 h of contact time, removal increases by more than 20%. Removal of PFPeS (5C) also increased in the saline solution by ~9% for the respective time periods. The increased removal

27

of shorter chain PFAS with increasing salinity supports the notion of ionic reactions dominating the short chain PFAS sorption mechanism supporting Higgins and Luthy (2006) and Hatton et al. (2018) which have shown that ionic effects are also significant for PFAS removal. No significant changes in removal due to increasing salinity was found for S(PFAS) and S(PFHxS þ PFOS) with <2 & 0.7% differences respectively at 1 h. For the PFCAs (Fig. 4B) with the exception of PFBA for which there was a ~9% increase in removal, all changes due to increased salinity were found to be <3%. This suggests that the hydrophobic reactions between the CF chains and amino acid residues of proteins, is the dominant mechanism for PFAS removal by hemp protein. PFAS removal as shown in Figs. 3 and 4 supports experimental work which shows that PFOS bioaccumulates more than PFOA due to the higher energy of the non-polar interactions of PFOS with human proteins (Salvalaglio et al., 2010; Zhang et al., 2013c) as PFOS contains one CF2 group more than PFOA. It is clear therefore that hydrophobic reactions with the hemp protein play a significant role in PFAS removal mechanism. This is also supported through ATRFTIR analysis of the protein solid samples.

3.1.3. ATR-FTIR spectroscopy ATR-FTIR is a technique widely used in the determination of protein structure because of its high sensitivity to small variations in bonding environments. The theory behind these will not be given here, however the reader is directed to works such as Kong and Yu (2007) and Yang et al. (2015) for the explanations regarding elucidation of protein bonding environments with FTIR. The major bands of the protein IR spectrum are the Amide I & II bands, the exact frequencies at which they occur are influenced by the strength of the bonding environment in the protein. Amide I absorption originates from the carbonyl (C]O) stretching vibration, and is evident in the ~1600 to 1700 cm1 region. The Amide II band results from NeH bending and CeN stretching vibrations and can be seen in the ~1480-1575 cm1 region. Although there have been nine amide bands described, these are very complex and have seen little practical use in protein conformational studies (Kong and Yu, 2007) and therefore only amide band I, which directly relate to the secondary structure of the protein is generally used. The amide I band is directly related to the protein “backbone” conformation and is therefore used to determine changes in protein secondary

Fig. 3. PFAS removal kinetics from groundwater using hemp protein powder at low (natural ~0.2 mS/cm) salinity. (A) PFSA's; (B) PFCA's.

28

B.D. Turner et al. / Chemosphere 229 (2019) 22e31

Fig. 4. PFAS removal kinetics from groundwater using hemp protein powder at high (~49 mS/cm) ionic strength. (A) PFSA's; (B) PFCA's.

structure. According to Kong and Yu (2007) the amide band can be further broken down into underlying secondary structures of the ahelix (1656 ± 2.0 cm1), b-sheets (1624e1642 ± 2.0 cm1), random coils (1648 ± 2.0 cm1) and b-turns (1667e1685 ± 2.0 cm1), and anti-parallel b-sheets (1691e1696 ± 2.0 cm1). To determine changes in the protein structure, the percentage areas of each secondary structure can be calculated by adding the areas of all peaks in the identified bands (Kong and Yu, 2007).

Fig. 6. FTIR spectra of soy protein powder exposed to 10%, 25%, and 100% (undiluted) PFAS contaminated groundwater (MWx). Spectra of soy protein powder and DI water only.

3.2. PFAS induced changes to hemp and soy protein secondary structure The ATR-FTIR spectra of hemp and soy protein powder exposed to DI water only (control), and increasing concentrations of PFAS contaminated groundwater (MWx 10, 25, & 100%) can be seen respectively in Fig. 5 and Fig. 6. The large peaks at ~3500 cm1 is the so called amide A band which does not depend on the backbone structure but is very sensitive to the strength of hydrogen bonds (Barth, 2007). With increasing PFAS concentration the intensity of the Amide A band appears to decrease, however they also become

Fig. 5. FTIR spectra of hemp protein powder exposed to 10%, 25%, and 100% (undiluted) PFAS contaminated groundwater (MWx). Spectra of hemp protein powder and DI water only (Control) is the average of 11 separate samples.

broader, increasing in overall area indicating an increase in the amount of hydrogen bonding within the protein. The amide I & II bands are labelled as 1645 cm1 and 1547 cm1 respectively. It is clear that both amide bands decrease with increasing PFAS concentration indicating that the hemp protein structure is changed by the interaction of PFAS. The percentages of each secondary structure of hemp proteins was calculated based on the integrated areas of the amide I component bands. Deconvolution of the amide I bands, determined that the amount of b-turns increased from ~9.3% (control) to 44.1% (MWx 100%); with a decrease in random coils (25.6e8.6%); a-helix (19.3e8.6%), bsheets (38.8e23.1%) and anti-parallel b-sheets increased (7e9%). This is supported by the work of Chi et al. (2018) who reported that the binding of PFOA and PFOS to human serum albumin decreased the a-helix content. These changes indicate unfolding of the protein structure (Deng et al., 2012) as a function of increasing PFAS concentration. In comparison, decovolution of the amide I band of the soy spectra showed that b-turns also increased (31.5e38.8%), random coils decreased (20.2%e18.8%) as did the anti-parallel bsheets (8.6e4.3%) whilst no change (<1%) was observed in the ahelix (~7.9%), b-turns (~23.4%) secondary structures. The presence of a strong band at ~1745 cm1 in the hemp control spectra (Fig. 5) is absent in the soy protein (Fig. 6). This band is due to the carbonyl (C]O) stretching vibration associated with amino acids (Li et al., 2005; Wei et al., 2018). Further, it has been found that the ~1700-1765 cm1 spectral region is influenced strongly by the charged amino acids glutamic (Glu) and aspartic (Asp) acid (Guler et al., 2016) of which hemp protein was found to have 240 and 1400 mg/kg Asp and Glu respectively (this study). In comparison, analysis of the soy protein showed less than the laboratory limit of reporting (<50 mg/kg) of these amino acids. It is

B.D. Turner et al. / Chemosphere 229 (2019) 22e31

apparent that following exposure of hemp protein to PFAS, the band at 1745 cm1 significantly decreases, almost disappearing implying PFAS binding at the site. This conclusion is supported by research which shows the binding of anionic surfactants (of which PFASs belong) to proteins induces the formation of multiple noncovalent bonds (e.g. hydrogen bonds, hydrophobic interactions and van der Waals forces) (Gao et al., 2006, 2008; Zhang et al., 2009). The bands observed at 3010, 2930 and 2857 cm1 for the control sample are related to CeH stretching vibrations (¼CH stretch at 3010; CH3 and CH2 at ~2855 and ~2925 cm1) (Forfang et al., 2017) and in biomolecules such as proteins, these bands have been shown to represent hydrophobic lipids (Simonova and Karamancheva, 2013; Ami et al., 2014; Forfang et al., 2017). The 3010 cm1 peak observed in the control disappears after exposure to PFAS indicating an association with protein ¼ CH sites possibly due to the electron withdrawing effect of the fluorinated tail on the carbon double bond. In fact it has been observed (Lu et al., 2010) that organofluorines such as PFASs display both electrophilic character along the CeF bonds and nucleophilic character perpendicular to these bonds. Due to their extreme electronegativity, fluorine atoms frequently act as good hydrogen bond acceptors forming bifurcated interactions i.e. F bonding in the “head on” orientation and hydrogen bonding “side on” with aromatic amino acid side chains (Lu et al., 2010; Parisini et al., 2011) such as those of Glu and Asp. The observed peak area at ~2925 cm1 (control) was found to decrease after PFAS (MWx 100%) exposure by ~93.3% indicating that the hydrophobic component of the hemp protein is interacting with PFAS. Similarly, we see no such strong bands in soy protein (Fig. 6) which may explain the reduction of the observed PFOS removal in soy compared to that of hemp (Fig. 2B). Given the very strong affinity of hemp protein for the hydrophobic PFOS (as demonstrated in Figs. 2e4), this spectral change supports the hypothesis that PFOS (and other species) bind to the hydrophobic sites on hemp proteins. This is corroborated with the appearance of bands at ~1000 to 1400 cm1 indicative of cabon-fluorine (CF) bonding. 3.3. CF FTIR band assignments In a study of the trifluoromethyl group, Beg and Clark (1962) reported CF3 vibrations at ~1100 cm and 1180 cm with some shift in frequency expected as the electronegativity of the F changes due to the bonding environment. Coates (2000) observed CF bands from 1150 to 1000 cm1; whilst others (Larkin, 2017) assign trifluorinated species (CF3) having multiple strong bands between 1350 and 1050 cm1 and CF2 with two bands between 1250 and 1050 cm1. Identification of specific groups associated with the PFCAs (i.e. carboxylates, C]O) becomes difficult in biomolecular systems due to the large amount of C]O groups already present in protein. Similarly for the PFSAs groups (i.e. sulfonates, SO 3 ) where the identified bands can overlap with the CF bands. For example, sulfonate group vibrations have been given (Zhang et al., 2013b) as 1250-1150 cm1 and 1075-1000 cm1 which overlap with the defined CF assignments. These overlapping bands coupled with the presence of sulphur containing amino acids such as methionine and cysteine makes it difficult to confidently assign spectral bands for carboxylic and sulfonate PFAS groups in protein spectra. Infrared difference spectra are the result of subtracting a spectrum of the protein in state A, from a spectrum in state B. In this way, only groups that actively participate in the reaction become evident and groups that do not participate in the reaction are cancelled in the subtraction (Kumar, 2014). The FTIR difference spectra of HPP samples exposed to PFAS over the CF “finger-print” region (~<2000 cm1) can be seen in Fig. SI-3. The figure clearly

29

shows a number of peaks not associated with the control sample from ~1750 to 900 cm1 that increase with increasing PFAS concentration. We observe multiple peaks from 1350 to 1050 cm1 which increase in intensity with PFAS concentration indicative of CF3 containing groups on the protein (Larkin, 2011). Bands at ~1425 cm1 coupled with ~1200 cm1 account (Huang et al., 2004) for the asymmetric stretching vibration of CF2. The appearance of a band at 1480 cm1 indicates the presence of CH3 groups (Coates, 2000) with the intensity increasing with PFAS concentration. This supports the evidence of unfolding proteins upon exposure to the PFAS solutions (as determined by the changes in the secondary structure a-helix structures discussed above) as we are observing more amino acid chains that were hidden in the initial folded protein. Three peaks can be seen to show a negative absorbance (1743 and 1050 cm1) with the peak at 1743 cm1 representative of a band associated with the amino acid carbonyl (C]O) stretching vibration (Service et al., 2010). The fact that it appears increasingly negative with increasing PFAS concentration indicates the association of PFOS molecules with this particular site on the protein. The spectra and subsequent interpretation is very complicated and further work on these is required before any definitive conclusions can be made as to the PFAS/HPP interactions. It is clear however that there is sorption occurring. 3.4. Protein solubility The solubility of hemp proteins has been found to be very poor (Tang et al., 2006; Wang et al., 2008) with the minimum solubility between pH 4e6. As the equilibrium pH the hemp protein solutions was determined to be 6.28 ± 0.08 over the first 20 h of contact time, the effect on PFAS removal by possible solublised proteins were determined (see methodology). No significant difference between the initial and final PFSA solution concentrations was observed (see Fig. SI-2; supplementary information). In fact, the actual difference between the initial and final concentrations of S(PFHxS þ PFOS) and SPFAS was found to be 2.1 and 3.7% respectively, far less than the smallest %RSD error of 10.3% as calculated for S(PFHxS þ PFOS) using Table 1 and Eqn. (4). The largest difference between initial and final concentrations was found for PFHpS at 15.1% (Fig. SI-2A); however, this is still less than the respective %RSD error of ~23.5% determined from the control sample (using Table 1 and Eqn. (4).) or the 13C surrogate recovery error of 16% reported by the analytical laboratory for this experiment. It is clear therefore, for both eTOP and þTOP analysis, there is no significant change in the concentrations prior to, and following, addition of the PFAS solution. Consequently, it can be conclude that no significant hemp proteins (that adsorb PFAS) are solubilised during the hemp protein powder/PFAS reaction supporting the findings of Tang et al. (2006) and Wang et al. (2008). 4. Conclusions The use of plant proteins as a possible solution to PFAS remediation from contaminated groundwater has been successfully demonstrated by very good removals (>98%) of the main contaminants of PFOS and PFHxS in less than 1 h of contact time. Of the plant proteins trialled, the proteins of Cannabis sativa L. were found to be far superior for PFAS removal than the next best protein, soy. Unlike GAC, hemp protein powders are a natural product that do not require pyrolysis to make it an active sorbent, however other varieties and sources of hemp may change the protein content which may impact on its performance as a PFAS substrate. In comparison to Norit® GAC the removals were found to be <1% for SPFAS and <3% for S(PFHxS þ PFOS), however on a commercial

30

B.D. Turner et al. / Chemosphere 229 (2019) 22e31

price basis, hemp protein powder retails at ~ A$45/kg compared to Norit® GAC at ~ A$125/kg. Therefore at the solid to solution ratio used in this study hemp protein powder appears to be ~2.7 times cheaper than treatment with an equivalent ratio of Norit® GAC (Note: the sorption isotherms and hence, optimal dose rate for each has not been determined and therefore the cost benefit is for the results presented only.) FTIR spectroscopy analysis revealed that hemp proteins partially unfold during the reaction with PFAS, undergoing H-bonding possibly with the amino acids Asp and Glu. Also evident from FTIR analysis was the association of PFAS with the hydrophobic sites of the protein, however, due to the complexity of this system it is difficult to determine the specific sites involved due to overlapping bands. This will be explored with future experiments including the thermal destruction of spent protein powders. The findings reported here however, will lead to new, environmentally friendly methods for PFAS remediation with many other options particularly for phytoremediation of PFAS contaminated soils. Acknowledgements This work was supported by a Research Attraction and Acceleration Program grant from the NSW Department of Primary Industry and the Office of the NSW Chief Scientist and Engineer, and the University of Newcastle, NSW, Australia. We would like to thank Prof. Mary O'Kane for her support and vision for this work, as well as the people of Williamtown, NSW, who struggle on a daily basis with the fall-out of PFAS contamination. Appendix A. Supplementary data Supplementary data to this article can be found online at https://doi.org/10.1016/j.chemosphere.2019.04.139. References ALS Environmental, 2016. PFOS Identifying Hidden PFAS Chemicals in Environmental Samples and Firefighting Foams. EnviroMail Issue #110. Aluko, R.E., 2017. Chapter 7 - hemp seed (cannabis sativa L.) proteins: composition, structure, enzymatic modification, and functional or bioactive properties A2 nadathur, Sudarshan R. In: Wanasundara, J.P.D., Scanlin, L. (Eds.), Sustainable Protein Sources. Academic Press, San Diego, pp. 121e132. Ami, D., Posteri, R., Mereghetti, P., Porro, D., Doglia, S.M., Branduardi, P., 2014. Fourier transform infrared spectroscopy as a method to study lipid accumulation in oleaginous yeasts. Biotechnol. Biofuels 7, 12-12. Banzhaf, S., Filipovic, M., Lewis, J., Sparrenbom, C.J., Barthel, R., 2017. A review of contamination of surface-, ground-, and drinking water in Sweden by perfluoroalkyl and polyfluoroalkyl substances (PFASs). Ambio 46, 335e346. Baroni, S., Mattu, M., Vannini, A., Cipollone, R., Aime, S., Ascenzi, P., Fasano, M., 2001. Effect of ibuprofen and warfarin on the allosteric properties of haem-human serum albumin. A spectroscopic study. Eur. J. Biochem. 268, 6214e6220. Barry, V., Winquist, A., Steenland, K., 2013. Perfluorooctanoic acid (PFOA) exposures and incident cancers among adults living near a chemical plant. Environ. Health Perspect. 121, 1313e1318. Barth, A., 2007. Infrared spectroscopy of proteins. Biochim. Biophys. Acta Bioenerg. 1767, 1073e1101. Beg, M.A.A., Clark, H.C., 1962. Chemistry of the trifluoromethyl group: part v. infrared spectra of some phosphorus compounds containing CF3. Can. J. Chem. 40, 393e398. Casson, R., Chiang, S.-Y., 2018. Integrating total oxidizable precursor assay data to evaluate fate and transport of PFASs. Remed. J. 28, 71e87. Chi, Q., Li, Z., Huang, J., Ma, J., Wang, X., 2018. Interactions of perfluorooctanoic acid and perfluorooctanesulfonic acid with serum albumins by native mass spectrometry, fluorescence and molecular docking. Chemosphere 198, 442e449. Coates, J., 2000. Interpretation of infrared spectra, a practical approach. Encycl. Anal. Chem. 12, 10815e10837. Deng, F., Dong, C., Liu, Y., 2012. Characterization of the interaction between nitrofurazone and human serum albumin by spectroscopic and molecular modeling methods. Mol. Biosyst. 8, 1446e1451. Eschauzier, C., Raat, K.J., Stuyfzand, P.J., De Voogt, P., 2013. Perfluorinated alkylated acids in groundwater and drinking water: identification, origin and mobility. Sci. Total Environ. 458e460, 477e485. Forfang, K., Zimmermann, B., Kosa, G., Kohler, A., Shapaval, V., 2017. FTIR spectroscopy for evaluation and monitoring of lipid extraction efficiency for

oleaginous fungi. PLoS One 12, e0170611 e0170611. Gao, H.W., Xu, Q., Chen, L., Wang, S.L., Wang, Y., Wu, L.L., Yuan, Y., 2008. Potential protein toxicity of synthetic pigments: binding of poncean S to human serum albumin. Biophys. J. 94. Gao, H.W., Zhao, J.F., Yang, Q.Z., Liu, X.H., Chen, L., Pan, L.T., 2006. Non-covalent interaction of 2',4',5',7'-tetrabromo-4,5,6,7-tetrachlorofluorescein with proteins and its applications. Proteomics 6. Goss, K.-U., 2008. The pKa values of PFOA and other highly fluorinated carboxylic acids. Environ. Sci. Technol. 42, 456e458. Graphpad, 2017. Detecting Outliers with Grubbs' Test. Guler, G., Gartner, R.M., Ziegler, C., Mantele, W., 2016. Lipid-protein interactions in the regulated betaine symporter BetP probed by infrared spectroscopy. J. Biol. Chem. 291, 4295e4307. Han, X., Snow, T.A., Kemper, R.A., Jepson, G.W., 2003. Binding of perfluorooctanoic acid to rat and human plasma proteins. Chem. Res. Toxicol. 16, 775e781. Hatton, J., Holton, C., DiGuiseppi, B., 2018. Occurrence and behavior of per- and polyfluoroalkyl substances from aqueous film-forming foam in groundwater systems. Remed. J. 28, 89e99. Higgins, C.P., Field, J.A., 2017. Our stainfree future? A virtual issue on poly- and perfluoroalkyl substances. Environ. Sci. Technol. 51, 5859e5860. Higgins, C.P., Luthy, R.G., 2006. Sorption of perfluorinated surfactants on sediments. Environ. Sci. Technol. 40, 7251e7256. Houtz, E.F., Sedlak, D.L., 2012. Oxidative conversion as a means of detecting precursors to perfluoroalkyl acids in urban runoff. Environ. Sci. Technol. 46, 9342e9349. Huang, K.P., Lin, P., Shih, H.C., 2004. Structures and properties of fluorinated amorphous carbon films. J. Appl. Phys. 96, 354e360. Kong, J., Yu, S., 2007. Fourier transform infrared spectroscopic analysis of protein secondary structures. Acta Biochim. Biophys. Sin. 39, 549e559. Kumar, S., 2014. Infrared difference spectroscopy as a physical tool to study biomolecules. Appl. Spectrosc. Rev. 49, 187e200. Larkin, P., 2011. Chapter 7 - general outline and strategies for IR and Raman spectral interpretation. In: Larkin, P. (Ed.), Infrared and Raman Spectroscopy. Elsevier, Oxford, pp. 117e133. Larkin, P., 2017. Infrared and Raman Spectroscopy: Principles and Spectral Interpretation. Elsevier Science. Li, Q.-B., Xu, Z., Zhang, N.-W., Zhang, L., Wang, F., Yang, L.-M., Wang, J.-S., Zhou, S., Zhang, Y.-F., Zhou, X.-S., Shi, J.-S., Wu, J.-G., 2005. In vivo and in situ detection of colorectal cancer using Fourier transform infrared spectroscopy. World J. Gastroenterol. 11, 327e330. Li, Q., Yang, W.-y., Qu, L.-l., Qi, H.-Y., Huang, Y., Zhang, Z., 2014. Interaction of warfarin with human serum albumin and effect of ferulic acid on the binding. J. Spectrosc. 2014, 7. Lu, Y., Wang, Y., Zhu, W., 2010. Nonbonding interactions of organic halogens in biological systems: implications for drug discovery and biomolecular design. Phys. Chem. Chem. Phys. : Phys. Chem. Chem. Phys. 12, 4543e4551. Luo, Z., Shi, X., Hu, Q., Zhao, B., Huang, M., 2012. Structural evidence of perfluorooctane sulfonate transport by human serum albumin. Chem. Res. Toxicol. 25, 990e992. Ng, C.A., Hungerbühler, K., 2014. Bioaccumulation of perfluorinated alkyl acids: observations and models. Environ. Sci. Technol. 48, 4637e4648. Nicole, W., 2013. PFOA and cancer in a highly exposed community: new findings from the C8 science panel. Environ. Health Perspect. 121. A340-A340. NSW Government Health, 2017. PFOS and PFOA Factsheet. New South Wales Governement Health, Australia. Olsen, G.W., Burris, J.M., Ehresman, D.J., Froehlich, J.W., Seacat, A.M., Butenhoff, J.L., Zobel, L.R., 2007. Half-life of serum elimination of Perfluorooctanesulfonate,Perfluorohexanesulfonate, and perfluorooctanoate in retired fluorochemical production workers. Environ. Health Perspect. 115, 1298e1305. Parisini, E., Metrangolo, P., Pilati, T., Resnati, G., Terraneo, G., 2011. Halogen bonding in halocarboneprotein complexes: a structural survey. Chem. Soc. Rev. 40, 2267e2278. Parsons, D.F., Duignan, T.T., Salis, A., 2017. Cation effects on haemoglobin aggregation: balance of chemisorption against physisorption of ions. Interfac. Focus 7, 20160137. Post, G.B., Cohn, P.D., Cooper, K.R., 2012. Perfluorooctanoic acid (PFOA), an emerging drinking water contaminant: a critical review of recent literature. Environ. Res. 116, 93e117. Robel, A.E., Marshall, K., Dickinson, M., Lunderberg, D., Butt, C., Peaslee, G., Stapleton, H.M., Field, J.A., 2017. Closing the mass balance on fluorine on papers and textiles. Environ. Sci. Technol. 51, 9022e9032. Rodriguez-Leyva, D., Pierce, G.N., 2010. The cardiac and haemostatic effects of dietary hempseed. Nutr. Metabol. 7, 32-32. Salvalaglio, M., Muscionico, I., Cavallotti, C., 2010. Determination of energies and sites of binding of PFOA and PFOS to human serum albumin. J. Phys. Chem. B 114, 14860e14874. Service, R.J., Hillier, W., Debus, R.J., 2010. Evidence from FTIR difference spectroscopy for an extensive network of hydrogen bonds near the oxygen-evolving Mn(4)Ca cluster of photosystem II involving D1-glu65, D2-glu312, and D1glu329. Biochemistry 49, 6655e6669. Shewry, P.R., Napier, J.A., Tatham, A.S., 1995. Seed storage proteins: structures and biosynthesis. Plant Cell 7, 945e956. Simonova, D., Karamancheva, I., 2013. Application of fourier transform infrared spectroscopy for tumor diagnosis. Biotechnol. Biotechnol. Equip. 27,

B.D. Turner et al. / Chemosphere 229 (2019) 22e31 4200e4207. Sposito, G., 1986. Distinguishing adsorption from surface precipitation. In: Davis, J.A., Hayes, K.F. (Eds.), ACS Symposium Series 323. Geochemical Processes at Mineral Surfaces. American Chemical Society. Division of Environmental Chemistry, Chicago, Illinois, pp. 217e228. Steenland, K., Fletcher, T., Savitz, D.A., 2010. Epidemiologic evidence on the health effects of perfluorooctanoic acid (PFOA). Environ. Health Perspect. 118, 1100e1108. Sudlow, G., Birkett, D.J., Wade, D.N., 1975. The characterization of two specific drug binding sites on human serum albumin. Mol. Pharmacol. 11, 824e832. Tang, C.-H., Ten, Z., Wang, X.-S., Yang, X.-Q., 2006. Physicochemical and functional properties of hemp (cannabis sativa L.) protein isolate. J. Agric. Food Chem. 54, 8945e8950. Teh, S.-S., Bekhit, A.E.-D.A., Carne, A., Birch, J., 2016. Antioxidant and ACE-inhibitory activities of hemp (Cannabis sativa L.) protein hydrolysates produced by the proteases AFP, HT, Pro-G, actinidin and zingibain. Food Chem. 203, 199e206. Thumanu, K., Sompong, M., Phansak, P., Nontapot, K., Buensanteai, N., 2015. Use of infrared microspectroscopy to determine leaf biochemical composition of cassava in response to Bacillus subtilis CaSUT007. J. Plant Interact. 10, 270e279. Wang, X.-S., Tang, C.-H., Yang, X.-Q., Gao, W.-R., 2008. Characterization, amino acid composition and in vitro digestibility of hemp (Cannabis sativa L.) proteins. Food Chem. 107, 11e18.

31

Wei, W., Hu, W., Zhang, X.-Y., Zhang, F.-P., Sun, S.-Q., Liu, Y., Xu, C.-H., 2018. Analysis of protein structure changes and quality regulation of surimi during gelation based on infrared spectroscopy and microscopic imaging. Sci. Rep. 8, 55665566. Wu, L.-L., Gao, H.-W., Gao, N.-Y., Chen, F.-F., Chen, L., 2009. Interaction of perfluorooctanoic acid with human serum albumin. BMC Struct. Biol. 9, 31. Xiao, F., 2017. Emerging poly- and perfluoroalkyl substances in the aquatic environment: a review of current literature. Water Res. 124, 482e495. Yang, H., Yang, S., Kong, J., Dong, A., Yu, S., 2015. Obtaining information about protein secondary structures in aqueous solution using Fourier transform IR spectroscopy. Nat. Protoc. 10, 382e396. Zhang, C., Yan, H., Li, F., Hu, X., Zhou, Q., 2013a. Sorption of short- and long-chain perfluoroalkyl surfactants on sewage sludges. J. Hazard Mater. 260, 689e699. Zhang, K., Huang, J., Yu, G., Zhang, Q., Deng, S., Wang, B., 2013b. Destruction of perfluorooctane sulfonate (PFOS) and perfluorooctanoic acid (PFOA) by ball milling. Environ. Sci. Technol. 47, 6471e6477. Zhang, L., Ren, X.-M., Guo, L.-H., 2013c. Structure-based investigation on the interaction of perfluorinated compounds with human liver fatty acid binding protein. Environ. Sci. Technol. 47, 11293e11301. Zhang, X., Chen, L., Fei, X.-C., Ma, Y.-S., Gao, H.-W., 2009. Binding of PFOS to serum albumin and DNA: insight into the molecular toxicity of perfluorochemicals. BMC Mol. Biol. 10, 16.