Different degrees of protein purity have been observed in immobilized metal affinity chromatography ranging from extremely high purity to moderate and low purity. It has been hypothesized that the host cell protein composition and the metal ligands are factors governing the purity of a protein obtained after immobilized metal affinity chromatography (IMAC).
Journal of Chromatography A 1633 (2020) 461649 Contents lists available at ScienceDirect Journal of Chromatography A journal homepage: www.elsevier.com/locate/chroma Proteomics analysis of host cell proteins after immobilized metal affinity chromatography: Influence of ligand and metal ions Nico Lingg a,b, Christoph Öhlknecht a,c, Andreas Fischer a, Markus Mozgovicz a, Theresa Scharl a,d, Chris Oostenbrink a,c, Alois Jungbauer a,b,∗ a Austrian Centre of Industrial Biotechnology, Muthgasse 18, A-1190 Vienna, Austria Department of Biotechnology, University of Natural Resources and Life Sciences, Vienna, Muthgasse 18, A-1190 Vienna, Austria Institute of Molecular Modeling and Simulation, University of Natural Resources and Life Sciences, Vienna, Muthgasse 18, A-1190 Vienna, Austria d Institute of Statistics, University of Natural Resources and Life Sciences, Vienna, Peter-Jordan-Straße 82, A-1190 Vienna, Austria b c a r t i c l e i n f o Article history: Received 28 August 2020 Revised 20 October 2020 Accepted 26 October 2020 Available online 29 October 2020 Keywords: Metal chelate Nitrilotriacetic acid Iminodiacetic acid E coli Green fluorescent protein a b s t r a c t Different degrees of protein purity have been observed in immobilized metal affinity chromatography ranging from extremely high purity to moderate and low purity It has been hypothesized that the host cell protein composition and the metal ligands are factors governing the purity of a protein obtained after immobilized metal affinity chromatography (IMAC) Ni nitrilotriacetic acid (NTA) has become the first choice for facile His-tagged protein purification, but alternative ligands such as iminodiacetic acid (IDA) with other immobilized metal ions such as Zn, Cu and Co are valuable options when the expected purity or binding capacity is not reached Especially Cu and Zn are very attractive, due to their reduced environmental and safety concerns compared to Ni Co and Zn are more selective than Ni and Cu This increased selectivity comes at the cost of weaker binding In this work, the influence of ligand choice on protein purity after IMAC was evaluated by several methods, including peptide mapping His-tagged GFP was used as model protein We found that host cell protein (HCP) content varies drastically between ligands, as IDA eluates generally showing higher HCP concentrations than NTA The relative content of the key amino acids His, Cys and Trp in the sequence of the co-eluted protein does not suffice to explain coeluting propensity The co-elution of HCPs is mostly influenced by metal binding clusters on the protein surface and not by total content or surface concentration of metal interacting amino acids Prediction of co-elution is not dependent on these clusters alone, due to protein-protein interactions, indicted by a relative low metal binding cluster score but high co-elution propensity and in a lot of cases these proteins are often part of complex such as ribosome and chaperones The different co-eluting proteins were presented by a heatmap with a dendrogram Ward’s linkage method was used to calculate the distance between groups of co-eluting proteins Clustering of co-eluting HCPs was observed according to ligand and by metal ions, with Zn and Co forming one cluster and Ni and Cu another The co-elution of host cell proteins can be explained by clusters of metal interacting amino acids on the protein surface and by protein-protein interactions While Ni NTA still appears to be highly advantageous, it might not be the cure-all for all applications © 2020 The Authors Published by Elsevier B.V This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/) Introduction Immobilized metal affinity chromatography (IMAC) has long been established as a prime candidate for facile purification of fusion-tag proteins [1–4] but also for natively metal binding proteins [5–7] It offers many advantages when combined with a His- ∗ Corresponding author at: Department of Biotechnology, University of Natural Resources and Life Sciences, Vienna, Muthgasse 18, A-1190 Vienna, Austria E-mail address: alois.jungbauer@boku.ac.at (A Jungbauer) tag on the protein of interest, such as affordable stationary phases, mild elution conditions, high capacity, high selectivity, and a large knowledge base The biggest advantage is the plug-and-play like ease with which purification can be achieved Compared to other types of affinity chromatography though, the purity can be inferior As such, an IMAC capture step is often combined with additional purification steps to achieve a desirable purity of the protein of interest The type of downstream processing unit operations will depend on the type and quantity of impurities still present after IMAC capture https://doi.org/10.1016/j.chroma.2020.461649 0021-9673/© 2020 The Authors Published by Elsevier B.V This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/) N Lingg, C Öhlknecht, A Fischer et al Journal of Chromatography A 1633 (2020) 461649 It has been hypothesized that the host cell protein composition and the metal ligands are the important factors governing the purity of a protein obtained after immobilized metal affinity chromatography A systematic evaluation of the influence of ligands and metals on the selectivity was done mainly for human serum proteins and oligo- and polynucleotides [8,9] Jerker Porath and others established serum fractions based on metal chelate chromatography [10–12] The vast body of information in this field can only restrictively be applied to purification of recombinant proteins expressed in E coli Within IMAC, the use of nitrilotriacetic acid (NTA) as the ligand and Ni2+ ions as the immobilized metal has become standard [13] Other ligands, such as iminodiacetic acid (IDA) or 1,4,7triazacyclononane (TACN) [14] are chemically similar but differ in complexation sites Consequently, this affects the binding and interaction with the immobilized metal ions Other metal ions, such as Cu2+ , Zn2+ and Co2+ can offer varying binding strengths and selectivity Cu2+ and Zn2+ in particular would be valuable alternatives to Ni, owing to their lower toxicity compared to Ni2+ and Co2+ Both Ni2+ and Co2+ are assigned to class 2A of the ICH Q3D guideline, with permitted daily exposure (PDE) of 22 μg/day and μg/day respectively [15] Ni2+ is considered genotoxic and carcinogenic, while Co2+ is a possible carcinogen Cu2+ is assigned to class with 340 μg/day PDE and can produce adverse effects in gastrointestinal tract, liver, and kidney when daily limits are exceeded These PDE values refer to parenteral exposure, with oral exposure limits being ten times higher Zn2+ is not classed at all, with no PDE established, but with a recommended dietary allowance of 8–11 mg/day [16] Zn can be a concern in patients with reduced hepatic function [15] It is clear that for large scale production, Zn would be preferable over Ni, considering the lack of adverse effects that residual Zn ions would cause in a drug product A deeper insight into co-elution of other impurities can be made by identification of the individual proteins by mass spectrometry The collected fractions can be enzymatically digested and the present proteins can be identified through data bank search for the peptide fragments Then, purely quantitative host cell protein concentration gained through ELISA can be enhanced by qualitative information about those host cell proteins [17] The proteins can be grouped for certain properties such as metal binding domains or other features Green fluorescence protein is a popular model protein, because it can be readily overexpressed in soluble form in E.coli [18,19] It has been also shown that the his-tagged form can be easily purified by immobilized metal chelate chromatography In order to judge the applicability of other metal/ligand combinations, a methodical investigation of critical process parameters, such as yield, and impurity profile is needed Here, we attempt a systematic characterization of yield and impurity pattern of a histagged protein on agarose based stationary phases with either IDA or NTA ligands with Cu, Zn, Co, and Ni as the immobilized metal ions His-tagged green fluorescent protein (GFP) produced in E coli was used, as it has well documented compatibility with a variety of metals [18,20,21] 2.2 Chromatography GFP mut3b carrying the mutations S65G, S72A [22] and carrying an N-terminal 6-His-tagwas produced in fed-batch fermentation with E coli strain BL21(DE3) (New England Biolabs, Ipswich, USA) For protein purification runs, 205 g of frozen cells were re-suspended in homogenization buffer (50 mM sodium phosphate, 300 mM NaCl, pH 8.0) to a final concentration of 200 g cell wet mass/L Cells were lysed in a high-pressure homogenizer (Panda Plus 20 0, GEA, Düsseldorf, Germany) in two passages at 10 0/10 bar The homogenate was centrifuged for h at 10,0 0 rpm, the pellet was discarded The supernatant was supplemented with imidazole (8 M) to a final concentration of 10 mM Prior to chromatographic capture, samples were 0.22 μm filtered (Millipore Millex-GV, Merck KGaA, Darmstadt, Germany) All chromatographic experiments were performed on an Äkta Pure 25 system (Cytiva, Uppsala, Sweden), with fraction collector F9-C (and sample pump S9) The stationary phases were WorkBeads IDA or NTA (BioWorks, Uppsala, Sweden) loaded with Ni, Cu, Zn or Co ions according to the manufacturer’s recommendation packed in Tricorn 10 columns (Cytiva) The columns had a column volume of approximately 1.5 mL and were operated with a constant residence time of The mobile phases were 10 mM imidazole, 50 mM sodium phosphate, 300 mM sodium chloride and pH 8.0 as equilibration buffer and 500 mM imidazole, 50 mM sodium phosphate, 300 mM sodium chloride and pH 8.0 as elution buffer The column was equilibrated with column volumes and loaded with mL of the clarified homogenate The resin was washed in 15 column volumes using equilibration buffer Elution was performed in a column volumes step-gradient of 100% elution buffer The column was cleaned in place using 10 mM NaOH for 30 2.3 Protein quantification All quantification experiments were performed on the Infinite 200Pro plate reader (Tecan Trading AG, Männedorf, Switzerland) For establishing an internal standard curve, His-tagged GFP, which was purified using a nickel-affinity and subsequently an ion exchange chromatography, was used The protein concentration of the standard, 10.3 g/L, was measured using UV–Vis spectroscopy at 280 nm The dilution series with GFP quantities ranged from 515 mg/L to mg/L Regression analysis were performed to associate GFP quantity and fluorescent intensity A slope of 92 (RFU × L/mg) was obtained for the standard curves For each resin, the flow-through, the wash and the elution fractions were pooled according to the volumetric content respectively 2.4 Peptide mapping The elution fractions of the eight chromatography runs were digested in solution according to the manufacturer’s instructions The proteins were S-alkylated with iodoacetamide and digested with Trypsin (Promega, Madison, WI, USA) 30 μl of each sample was transferred into a 1.5 ml screw cap micro-tube and cysteines were reduced by the addition of 30 μl 15 mM dithiothreitol in 100 mM ammonium bicarbonate buffer pH 7.8 for 45 at 56 °C 30 μl of 55 mM iodoacetamide in 100 mM ammonium bicarbonate buffer pH 7.8 were added and in 100 mM ammonium bicarbonate and incubated for 30 at room temperature in the dark Proteins were subsequently precipitated with 360 μl acetone (30 incubation at −20 °C) and dried in a speed vac concentrator The samples were re-dissolved in 30 μl 100 mM ammonium bicarbonate and digested with 6.5 μl (=0.65 μg) trypsin on 37 °C over night The digested samples were loaded on a BioBasic C18 column (BioBasic-18, 150 × 0.32 mm, μm, Thermo Scientific, Waltham, Material and methods 2.1 Chemicals Imidazole, copper (II) sulfate and cobalt (II) sulfate heptahydrate were purchased in analytical grade from Sigma-Aldrich (Missouri, USA) NaCl, sodium dihydrogen phosphate, nickel (II) sulfate and zinc chloride were purchased in analytical grade from Merck KGaA (Darmstadt, Germany) N Lingg, C Öhlknecht, A Fischer et al Journal of Chromatography A 1633 (2020) 461649 MA, USA) using 80 mM ammonium formate buffer as the aqueous solvent A gradient from 5% B (B: 80% acetonitrile) to 40% B in 45 was applied, followed by a 15 gradient from 40% B to 90% B that facilitates elution of large peptides, at a flow rate of μL/min Detection was performed with QTOF MS (Bruker maXis G, Bruker, Billerica, MA, USA) equipped with the standard ESI source in positive ion, DDA mode and switching to MSMS mode for eluting peaks MS-scans were recorded (range: 150–2200 Da) and the highest peaks were selected for fragmentation Instrument calibration was performed using ESI calibration mixture (Agilent, Santa Clara, CA, USA) The analysis files were converted (using Data Analysis, Bruker) to mgf files, which are suitable for performing a MS/MS ion search with ProteinScape (Bruker, MASCOT embedded) and/or GPM The files were searched against a Uniprot database for the proteome of E coli (strain B / BL21-DE3) (Proteinscape) where m(i) is the number of clusters of i key amino acids within a single protein imax is the biggest cluster that was found within the respective protein In this MBCS, clusters with higher i score with higher weight This was done to take cooperative effects into account: if one key amino acid is bound, another key amino acid in the vicinity has a higher probability of binding Next to the proteins that were found in the peptide mapping, further cytosolic HCPs were evaluated if structural information was available in the Swiss-Model Repository This information was available for 102 proteins in total Results 3.1 Chromatography GFP capture chromatography was performed with eight different stationary phases The ligands NTA and IDA were used with the metals Co, Cu, Ni and Zn Clarified cell lysis supernatant was loaded, washed with equilibration buffer containing a low concentration of imidazole and eluted using a step gradient to high imidazole concentration The chromatograms with IDA stationary phases generally showed larger wash peaks, but in general all chromatograms were highly similar, as shown in Fig Only the Cu NTA stationary phase showed a very small elution peak at the conditions tested In order to compare the resulting eluates, SDS-PAGE was performed for all eight runs Fig shows the load material compared to pooled flow-through and wash fractions (FT+W) and eluate While all eight stationary phases can produce pure and concentrated GFP, the endogenous E coli protein pattern co-eluting with the different in all eight cases NTA ligands appear to have a higher specificity, owing to their lower number of available complexation sites, which requires a higher interaction affinity for a protein to adsorb Compared to the load material, the elution fractions of all eight conditions have high purity, with NTA resins, generally leading to less co-elution of impurities, while IDA resins leading to lower protein losses in the flow-through and wash fractions The impurity pattern itself also differs for the four different metal ions on the IDA resins, where host cell proteins (HCP) are visible in the SDS-PAGE No differences in purity can be determined from SDSPAGE analysis for the NTA resins In order to get more meaningful impurity data, more sensitive analytical methods were performed Table shows the yield and recovery of the chromatography runs Interestingly, some columns exhibited irreversible binding, that led to a total recovery of below 100% This strong binding could be confirmed visually, since GFP was seen in some columns after elution ELISA was used to quantify the HCP concentration in all eight eluates, as shown in Table The ELISA results confirm that the IDA eluates are less pure than the NTA eluates Interestingly, even though HCPs can be detected in the SDS-PAGE of the Zn IDA eluate, the HCP ELISA determined this fraction to have a relatively low concentration of 48 ng/mL Since the total number of HCPs is relatively low, it is possible that the ELISA underestimated the actual HCP concentration due to the nature of the assay [26] Since the ELISA is using a pool of antibodies against various HCPs, the measured concentration can suffer from bias if only a small number of HCPs are present The quantification of dsDNA was difficult to perform experimentally, since the most commonly used assays all rely on fluorescence measurements at the same wavelength as GFP for detection This required a different quantitative method for our samples, namely qPCR Table shows the dsDNA concentrations found in the eight eluate samples Only Ni and Co NTA eluates had dsDNA over the lower limit of quantification for the assay The endotoxin concentration was measured using a recombinant assay The eluate from all four IDA columns showed endotoxin concentrations over the upper limit of quantification For 2.5 Host cell protein (HCP) ELISA and endotoxin assay The E coli ELISA was purchased from Cygnus (Southport, North Carolina, USA) and performed according to Sauer et al [23] In brief, 96-well Nunc MaxiSorp Immuno plates (Thermo Fisher) were coated with anti-E coli HCP capture antibody and blocked with BSA E coli HCP antigen was used as a standard and concentrations from 0.4 to 50 ng/mL were transferred to the microtiter plate Samples were pre-diluted with sample buffer to be in the calibration range and serial dilutions were transferred to the plate After incubation and detection with horseradish peroxidase conjugated anti-E coli HCP antibody, the absorbance change after addition of TMB substrate was used to quantify the HCP concentration with an Infinite M200 Pro plate reader (Tecan) Endotoxin was quantified using EndoZyme R II recombinant Factor C (rFC) assay (Hyglos, Bernried, Germany) according to Sauer et al [23] 2.6 qPCR The DNA content was quantified using resDNASEQTM Quantitative E coli DNA Kit (Thermo Scientific) based on qPCR according to the manufacturer’s instructions Due to interaction with the matrix, the GFP eluate samples had to be buffer exchanged into PBS using kDa cut-off in Amicon Ultra spin vials (Merck) 2.7 Computational analysis Sequence for the proteome of E coli (strain B / BL21-DE3) were downloaded from the Uniprot Knowledgebase [24] Structures from homology modeling were downloaded from the Swiss-Model Repository [25] Calculations on structural information were performed using Pymol 2.3.4 (The PyMOL Molecular Graphics System, Version 2.0 Schrödinger, LLC) Three different analyses were performed to link information based on sequence or structure to protein selectivity of binding of immobilized metal ions The amino acids His, Cys and Trp were selected as key amino acids for metalion binding Calculated were a) the relative occurrence of the key amino acids in the sequence; b) the relative surface area of the key amino acids compared to the total surface area of the protein; c) the occurrence of clusters of key amino acids The occurrence of clusters that are made up of one or more different types of key amino acids were counted A cluster was defined as a minimum of key amino acids of the same or different kind within a sphere of nm diameter The total occurrence of metal binding clusters within a protein was transformed into a metal binding cluster score (MBCS) that was defined as imax i2 × m MBCS = (1) i=2 N Lingg, C Öhlknecht, A Fischer et al Journal of Chromatography A 1633 (2020) 461649 Fig Chromatograms of all eight runs, with metal ion and ligand denoted in the chromatogram The blue trace is the absorbance at 280 nm, the dashed green trace is the specific absorbance for GFP at 488 nm, and the gray dashed trace is the elution buffer concentration Elution profiles of His-tagged GFP look very similar across all metal/ligand combinations, while the behavior of host cell proteins during the wash step varies for each column (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.) Table Mass balance of chromatographic runs by ligand and immobilized metal-ion Yield refers to the amount of GFP in the elution fraction relative to the loaded amount, whereas recovery FTW refers to the amount of GFP lost in the non-binding flow through and wash steps HCP, dsDNA and Endotoxin concentrations were measured for the elution fractions Ligand Metal-ion Yield Recovery FTW HCP (ng/mL) dsDNA (ng/mL) Endotoxin (103 EU/mL) IDA Co Cu Ni Zn Co Cu Ni Zn 72% 81% 80% 84% 36% 51% 97% 57% 6% 3% 13% 2% 37% 17% 4% 23% 323 ± 69 247 ± 74 400 ± 108 48 ± 14 171 ± 18 20.3 ± 0.5 265 ± 33 27.1 ± 0.5 < 0.03 < 0.03 < 0.03 < 0.03 0.56 < 0.03 12.3 < 0.03 > 35 > 35 > 35 > 35 5.8 ± 0.7 0.5 ± 0.2 21 ± 3.1 ± 0.3 NTA the NTA stationary phases, the concentration depended strongly on the immobilized metal ion, with Ni showing the highest concentration and Cu the lowest, and Co and Zn in between These endotoxin results somewhat mirror the behavior in regard to HCP and dsDNA concentration by ELISA for comparison The qualitative and quantitative HCP results are similar, but not identical The ligand appears to have a larger influence on the specificity of the interaction, with IDA being less specific than NTA The immobilized metal has a lower influence on specificity: Co and Zn appear to be more specific in their protein interaction, since only around 50% of the total proteins identified were found on either Co or Zn Immobilized Cu and Ni on the other hand lead to co-elution of 72% and 65% of total HCPs, respectively 3.2 Peptide mapping The identity of the specific HCPs co-eluting with the His-tagged GFP was determined via peptide mapping In total 381 E coli proteins were identified, in addition to GFP, trypsin (from the digest) and human keratin (from the operator) A full list can be found in the supplementary information In total 360 HCPs were identified in at least one of the IDA eluates (94% of the total HCPs) and 109 HCPs were identified in at least one of the NTA eluates (29% of total HCPs), with 88 HCPs (23%) identified eluting from both ligands A total of 13 HCPs were identified in all conditions, which can be found in Table The HCPs found co-eluting from or stationary phases can be found in Tables and Fig 3A shows the number of identified HCPs from the IDA and NTA eluates depending on the immobilized metal Fig 3B shows the effect of immobilized metal on the number of HCPs co-eluting for IDA and NTA ligands Fig 3C shows the quantitative HCP results as determined 3.3 Computational analysis We hypothesized that metal interacting amino acids play an important role in the co-elution behavior in IMAC In order to test this hypothesis, we compared the characteristics of the co-eluting proteins with a random sample of cytosolic E coli proteins that were not found in any of our eluates We have evaluated systematic experiments varying the ligand (IDA or NTA) and immobilized metal ions (Co2+ , Cu2+ , Ni2+ or Zn2+ ) In total we have found 381 proteins which have co-eluted in at least one experiment We refer to the co-elution propensity as n, defined as the number of metal/ligand combinations in which co-elution was observed When n = a particular protein is found in all eluates and N Lingg, C Öhlknecht, A Fischer et al Journal of Chromatography A 1633 (2020) 461649 Fig Coomassie stained SDS-PAGE results of ligand (NTA or IDA) and immobilized metal ion (Ni, Co, Cu or Zn) combinations The FT + W lane corresponds to the pooled flow-through and wash fractions GFP with a 6-His-tag was overexpressed in E coli and cell lysate was loaded to WorkBeads columns when n = the respective protein could only be found in a single experiment For 99 different cytosolic proteins, which have not been found in the eluates, the co-elution propensity n = This served as a control group Three computational analysis methods were compared regarding their ability to explain differences in interaction specificities among E coli HCPs 1) the relative content of the key amino acids in the primary sequence, 2) the relative surface area of the key amino acids, and 3) a metal binding cluster score describing the presence and size of key amino acid clusters The relative content of the key amino acids His, Cys and Trp in the sequence does not suffice to explain differences between the individual groups, as shown by the lack of correlation in Fig 4A Moreover, no correlation of co-elution and the relative surface area of the three key amino acids could be found (Fig 4B) An alternative metal binding cluster score (MBCS) was constructed (see methods) that represents the fact that cooperative effects may occur when multiple key residues are in close vicinity The MBCS are available in the supplementary information for each protein and shown in Tables 2, and for the proteins were n = 8, 7, Scoring surface clusters of the key amino acids can be used to explain a part of the variability between different HCPs A protein with a higher MBCS has a higher probability to be co-eluting in a larger number of conditions (Fig 4C) In other words, a higher amount of different cluster sizes and differently orientated clusters increases the proteins tendency to bind to different immobi5 N Lingg, C Öhlknecht, A Fischer et al Journal of Chromatography A 1633 (2020) 461649 Table List of host cell proteins found co-eluting under all conditions Accession # Name MBCS pI Molecular mass (kDa) A0A140N3N3 A0A140N7Y4 A0A140N3D6 A0A140ND61 A0A140NHM8 A0A140N6W0 A0A140N6E5 A0A140NE13 A0A140N7J1 A0A140NFK2 A0A140N548 A0A140NE25 A0A140N587 tRNA (guanine-N(7)-)-methyltransferase Pseudouridine synthase Transcriptional regulator Crp/Fnr family Chaperone protein HtpG Soluble pyridine nucleotide transhydrogenase Elongation factor Tu D-tagatose-1 6-bisphosphate aldolase subunit GatZ Ferric uptake regulation protein 50S ribosomal protein L2 30S ribosomal protein S2 30S ribosomal protein S4 Glutamine–fructose-6-phosphate aminotransferase [isomerizing] Bifunctional polymyxin resistance protein ArnA 33 37 77 84 151 198 214 258 430 446 449 512 667 6.6 5.9 8.2 5.0 6.2 5.2 5.5 5.8 11.2 6.8 10.3 5.6 6.5 27.3 25.8 23.6 71.4 51.6 43.3 47.0 16.8 29.9 26.7 23.5 66.9 74.3 Table List of host cell proteins found co-eluting under seven conditions Accession # Name MBCS pI Molecular mass (kDa) Not found A0A140N6Z9 A0A140NBL1 A0A140N319 A0A140N598 A0A140N4M0 A0A140N811 A0A140NAY3 30S ribosomal protein S5 HAD-superfamily hydrolase subfamily IIA RNase adapter protein RapZ 50S ribosomal protein L13 50S ribosomal protein L17 30S ribosomal protein S15 Histidine biosynthesis bifunctional protein HisB 54 118 431 432 449 1105 10.5 5.1 6.9 10.2 11.3 10.7 5.9 17.5 27.1 32.5 16.0 14.4 10.3 40.2 Zn NTA Ni NTA Ni NTA Co NTA Cu NTA Zn IDA Co NTA Table List of host cell proteins found co-eluting under six conditions Accession # Name MBCS pI Molecular mass (kDa) Not found A0A140NFV3 A0A140SS47 A0A140NGK1 A0A140NF03 A0A140NHS0 A0A140NC35 A0A140NBE7 A0A140NB96 A0A140N6V1 A0A140N8K1 A0A140SS84 A0A140N4K1 A0A140NA80 Chaperone protein DnaK Uncharacterized protein RNA-binding protein Hfq Transcriptional regulator IclR family ATP synthase subunit beta Bifunctional aspartokinase/ homoserine dehydrogenase Formate acetyltransferase Transcriptional regulator LacI family Peptidyl-prolyl cis-trans isomerase Transcriptional regulator LysR family Acetylornithine deacetylase 30S ribosomal protein S3 Succinate dehydrogenase flavoprotein subunit 39 77 77 84 95 95 97 97 183 209 310 452 620 4.7 6.7 7.6 7.9 4.8 5.5 5.7 6.6 4.8 6.2 5.6 10.6 6.0 69.1 15.6 11.2 29.7 50.3 89.0 85.3 38.9 20.8 32.7 42.3 26.0 64.4 Co Cu Cu Cu Co Co Co Co Zn Cu Co Co Co lized metal ions However, this MBCS is not sufficient to explain the entire variability between the different groups Other effects such as protein-protein binding may have a significant role too: proteins that not bind the matrix directly but are believed to have high binding affinities towards other proteins Among the lower scoring proteins in n > in Fig 4C, chaperones and ribosomal subunits were found The most drastic examples of which is 30S ribosomal protein S5 (A0A140N6Z9) that was found coeluting in out of conditions but has an MBCS of Since other 30S ribosomal sub-units were found to have high MBCS (e.g S2, A0A140NFK2, score 446 and S4, A0A140N548, score 449) and coeluting with all eight metal/ligand combinations, it seems plausible that sub-unit S5 was bound merely through protein-protein interaction Another interesting protein is chaperone protein HtpG (A0A140ND61), which was found in all conditions, but has a relatively low MBCS of 50 It is likely that the main mechanism for chaperones is through protein-protein interaction, instead of direct metal binding Indeed, a variety of chaperones were identified (DnaK, DnaJ, ClpB, OmpH, ProQ) with varying MBCS, including Electrostatic interaction of the co-eluted proteins can be excluded because the experiments were performed at 0.3 M NaCl and electrostatic shielding can be expected at this high salt concentration IDA, Zn NTA IDA, Ni IDA IDA, Ni IDA NTA, Cu IDA NTA, Zn NTA NTA, Zn NTA NTA, Zn NTA NTA, Zn NTA NTA, Zn IDA IDA, Ni IDA NTA, Zn NTA NTA, Zn NTA NTA, Zn NTA For the visualization of the output of the experiments a heatmap was chosen (Fig 5) It is a two-dimensional representation of the data in which the outcomes of the experiments are color-coded (light gray – no co-elution, black – co-elution) Additionally, the different experiments as well as the proteins were reordered by hierarchical clustering The binary distance was computed to group similar objects next to each other Ward’s linkage method was used to calculate the distance between groups of objects On top of the heatmap a dendrogram of the experiments is given The dendrogram is a tree visualizing in which order groups are merged Experiments with similar outcome are grouped next to each other The dendrogram on the left of the heatmap gives the similarity between the proteins Instead of the names of the proteins a color key was used which is a combination of metal binding cluster score (MBCS), molecular mass and pI The co-eluting HCPs can be grouped by ligand first, and within those groups, the metals Cu and Ni make up a group that co-elute similar clusters of HCPs and Co and Zn forming a second group with their own cluster of HCPs, as shown in the dendrogram at the top of Fig 5A These clusters are not exclusive, and many overlaps exist This clustering is already apparent in Fig The proteins with a high MBCS are mostly found co-eluting under multiple conditions (green bars in Fig 5B), indicating that the MBCS has some N Lingg, C Öhlknecht, A Fischer et al Journal of Chromatography A 1633 (2020) 461649 Fig Venn diagram of E coli HCPs found in IMAC eluates, depending on metal and ligand The area of the ellipses is relative to the number of identified HCPs Numbers denote the number of HCPs that were identified in each sample Panel A is grouped by ligand (IDA and NTA) and panel B is grouped by immobilized metal ion (Co, Cu, Ni and Zn) 381 proteins were identified in total Panel C shows the quantitative HCP ELISA results (in ng/mL) in the same style predictive power for IMAC co-elution The factors molecular mass and isoelectric point not seem to influence the clustering ity comes at the cost of weaker binding as described previously [20] and further confirmed by the data in this work HCP content varies drastically between ligands, as IDA eluates generally showing higher HCP concentrations than NTA Additionally, the choice of metal ion also having an impact We have shown that HCP coelution cannot be explained simply by content or surface concentration of metal interacting amino acids (His, Cys, Trp), but depends on the presence of clusters of these proteins on the surfaces When using a MBCS for all proteins that were found co-eluting and comparing it to a sample of HCPs that were not found coeluting, the score correlated with the number of co-eluting conditions For some HCPs though, co-elution appears to be affected by protein-protein interactions, as is the case for ribosomal sub- Discussion While the combination of Ni2+ and NTA reigns supreme in the world of IMAC, other metal/ligand combinations can be viable alternatives From the peptide mapping data, it can clearly be deduced that the choice of ligand has an immense impact on the number of co-purified HCPs, with NTA generally resulting in a lower number of unique HCPs The choice of immobilized metal ion also affects the number of unique HCPs, with Co2+ and Zn2+ being more selective than Ni2+ and Cu2+ This increased selectiv7 N Lingg, C Öhlknecht, A Fischer et al Journal of Chromatography A 1633 (2020) 461649 Fig Results of the computation analyses on the key amino acids His, Cys and Trp Several analyses were used to compare differences in the co-elution propensity (n = 0…8) A) Relative content of the key amino acids based on primary sequence information for all HCPs versus co-elution propensity n B) Relative surface area of the key amino acids compared to the total surface area for all HCPs versus co-elution propensity n C) MBCS of the key amino acids for all HCPs versus co-elution propensity n Red spheres mark the median and the vertical red bars mark the standard deviation within the individual groups (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.) units and chaperone Quantitatively, stationary phases with immobilized Ni2+ and Co2+ leads to strong co-elution of HCP, with concentrations in the hundreds of ng/mL Immobilized Zn2+ on the other hand leads to lower HCP concentrations co-eluting with histagged GFP In the case of immobilized Cu2+ , the effect of ligand choice is strong, with NTA leading to ten times less HCP co-elution than IDA The effect on dsDNA depletion was very difficult to study, since the most common DNA assays rely on fluorescence measurements where GFP directly interferes Quantitative PCR was used to quantify the dsDNA concentration in the samples, but the results were much lower than expected, with only Co and Ni2+ NTA being quantifiable As such, it is uncertain if the measured concentrations are accurate Endotoxin concentrations varied widely between samples, with the IDA samples all being over the upper limit of quantification of the assay used Out of the NTA eluates, the ranking in purity is Cu2+ , Zn2+ , Co2+ and Ni2+ with a 40-fold range of endotoxin concentrations A wash step with organic solvent during IMAC can reduce this concentration [27], but an anion exchange step might be necessary for products where endotoxins are of concern When IMAC is sought as the sole purification step, considering our results it seems reasonable to choose Ni2+ NTA as the capture adsorbent Ni2+ NTA exhibits the highest yield and a relatively high purity, compared to other ligand and metal combinations The biggest downside of using Ni, is its inherent toxicity, which necessitates its removal for products intended for administration to humans Even if Ni2+ leakage can be quite low, in the range of ppm [28], the successful removal still has to be validated High concentrations of Ni2+ ions in wastewater for column cleaning further increase costs Implementing a downstream process without relying on Ni as part of a two-step process with removal of the affinity tag could be practicable Such a two-step process consists of IMAC capture, enzymatic tag removal and subsequent subtractive IMAC in which the product is in the flow-through fraction and the previously co-eluting HCPs can be removed along with the affinity tagged enzyme An important prerequisite of such a process is the absence of endogenous proteases, that could otherwise digest the protein of interest Indeed, of all 381 E coli proteins, only two proteases were identified: ATP-dependent zinc metalloprotease and ATP-dependent Clp protease, both of which should be inactive after capture due to their ATP dependence One effect not studied here, is the potential of displacement of HCPs by the protein of interest, especially in the case of very high titers and/or high binding affinity The exposure of the His-tag may vary with the protein of interest The co-elution of proteins can be explained by clusters of metal interacting amino acids, and by protein-protein interaction, where one protein binds to the metal ions and other proteins interact with the bound protein A total N Lingg, C Öhlknecht, A Fischer et al Journal of Chromatography A 1633 (2020) 461649 Fig Heat map of E coli HCP co-elution depending on stationary phase In the central graph, black denotes co-elution and gray denotes no co-elution The proteins are color coded by their MBCS (1 < 50, > 50, 325), by molecular mass (1 < 42.2 kDa, > 42.2 kDa), and by isoelectric point (1 < 7, > 7) Panel A shows all 381 proteins that were found co-eluting from at least one stationary phase, whereas panel B is zoomed in on those proteins co-eluting on at least one NTA stationary phase N Lingg, C Öhlknecht, A Fischer et al Journal of Chromatography A 1633 (2020) 461649 of 13 E coli HCPs (Table 2) were found co-eluting from all eight investigated stationary phases Our database may serve as a reference for others [4] W Jiang, M Prescott, R.J Devenish, L Spiccia, M.T.W Hearn, Separation of hexahistidine fusion proteins with immobilized metal ion affinity chromatographic (IMAC) sorbents derived from MN+ -tacn and its derivatives, Biotechnol Bioeng 103 (4) (2009) 747–756, doi:10.1002/bit.22302 [5] D.M Jandhyala, R.C Willson, B.T Sewell, M.J Benedik, Comparison of cyanidedegrading nitrilases, Appl Microbiol Biotechnol 68 (3) (2005) 327–335, doi:10.10 07/s0 0253-0 05-1903-8 [6] B Xu, U Hellman, B Ersson, J.C Janson, Purification, characterization and amino-acid sequence analysis of a thermostable, low molecular mass endoβ -1,4-glucanase from blue mussel, Mytilus edulis, Eur J Biochem 267 (16) (20 0) 4970–4977, doi:10.1046/j.1432-1327.20 0.01533.x [7] Z Zhang, K.T Tong, M Belew, T Pettersson, J.C Janson, Production, purification and characterization of recombinant human interferon γ , J Chromatogr A 604 (1) (1992) 143–155, doi:10.1016/0021- 9673(92)85539- [8] J Porath, B Olin, Immobilized metal ion affinity adsorption and immobilized metal ion affinity chromatography of biomaterials Serum protein affinities for gel-immobilized iron and nickel ions, Biochemistry 22 (7) (1983) 1621–1630, doi:10.1021/bi00276a015 [9] P Hubert, J Porath, Metal Chelate Affinity-Chromatography Group Separation of Mononucleotides and Dinucleotides, J Chromatogr 206 (1) (1981) 164– 168, doi:10.1016/S0 021-9673(0 0)82620-5 [10] J Porath, J Carlsson, I Olsson, G Belfrage, Metal chelate affinity chromatography, a new approach to protein fractionation, Nature 258 (5536) (1975) 598– 599, doi:10.1038/258598a0 [11] K Karkra, K.K.R Tetala, M.A Vijayalakshmi, A structure based plasma protein pre-fractionation using conjoint immobilized metal/chelate affinity (IMA) system, J Chromatogr B 1052 (2017) 1–9, doi:10.1016/j.jchromb.2017.02.032 [12] R.R Prasanna, M.A Vijayalakshmi, Characterization of metal chelate methacrylate monolithic disk for purification of polyclonal and monoclonal immunoglobulin G, J Chromatogr A 1217 (23) (2010) 3660–3667, doi:10.1016/j chroma.2010.03.058 [13] E Hochuli, H Döbeli, A Schacher, New metal chelate adsorbent selective for proteins and peptides containing neighbouring histidine residues, J Chromatogr A 411 (C) (1987) 177–184, doi:10.1016/S0 021-9673(0 0)93969-4 [14] W Jiang, B Graham, L Spiccia, M.T.W Hearn, Protein selectivity with immobilized metal ion-tacn sorbents: chromatographic studies with human serum proteins and several other globular proteins, Anal Biochem 255 (1) (1998) 47– 58, doi:10.1006/abio.1997.2395 [15] ICH, ICH Q3D elemental impurities, 2019 https://www.ema.europa.eu/en/ ich- q3d- elemental- impurities (Accessed 2020-02-11) [16] L.M Plum, L Rink, H Haase, The essential toxin: impact of zinc on human health, Int J Environ Res Public Health (4) (2010) 1342–1365, doi:10.3390/ ijerph7041342 [17] A.L Tscheliessnig, J Konrath, R Bates, A Jungbauer, Host cell protein analysis in therapeutic protein bioprocessing - methods and applications, Biotechnol J (6) (2013) 655–670, doi:10.10 02/biot.20120 018 [18] M Zimmer, Green fluorescent protein (GFP): applications, structure, and related photophysical behavior, Chem Rev 102 (3) (2002) 759–781, doi:10.1021/ cr010142r [19] M Chalfie, Y Tu, G Euskirchen, W.W Ward, D.C Prasher, Green fluorescent protein as a marker for gene expression, Science 263 (5148) (1994) 802–805, doi:10.1126/science.8303295 [20] L KågedalJ.-C Janson (Ed.), Immobilized metal ion affinity chromatography, Prot Purif (2011) 183–201, doi:10.1002/9780470939932.ch7 [21] M Peterka, M Jarc, M Banjac, V Frankovicˇ , K Bencˇ ina, M Merhar, V GabercPorekar, V Menart, A Štrancar, A Podgornik, Characterisation of metal-chelate methacrylate monoliths, J Chromatogr A 1109 (1) (2006) 80–85, doi:10.1016/ j.chroma.2005.08.057 [22] B.P Cormack, R.H Valdivia, S Falkow, FACS-optimized mutants of the green fluorescent protein (GFP), Gene 173 (1) (1996) 33–38, doi:10.1016/ 0378-1119(95)00685-0 [23] D.G Sauer, M Mosor, A.C Frank, F Weiss, A Christler, N Walch, A Jungbauer, A Durauer, A two-step process for capture and purification of human basic fibroblast growth factor from E coli homogenate: yield versus endotoxin clearance, Protein Expr Purif 153 (2019) 70–82, doi:10.1016/j.pep.2018.08.009 [24] C UniProt, UniProt: a worldwide hub of protein knowledge, Nucleic Acids Res 47 (D1) (2019) D506–D515, doi:10.1093/nar/gky1049 [25] A Waterhouse, M Bertoni, S Bienert, G Studer, G Tauriello, R Gumienny, F.T Heer, T.A.P de Beer, C Rempfer, L Bordoli, R Lepore, T Schwede, SWISSMODEL: homology modelling of protein structures and complexes, Nucleic Acids Res 46 (W1) (2018) W296–W303, doi:10.1093/nar/gky427 [26] J.Y Baik, J Guo, K.H Lee, Host cell proteins during biomanufacturing, Cell Culture Eng (2019) 295–311, doi:10.1002/9783527811410.ch12 [27] K.L Franken, H.S Hiemstra, K.E van Meijgaarden, Y Subronto, J den Hartigh, T.H Ottenhoff, J.W Drijfhout, Purification of his-tagged proteins by immobilized chelate affinity chromatography: the benefits from the use of organic solvent, Protein Expr Purif 18 (1) (20 0) 95–99, doi:10.1006/prep.1999.1162 [28] V Gaberc-Porekar, V Menart, Potential for using histidine tags in purification of proteins at large scale, Chem Eng Technol 28 (11) (2005) 1306–1314, doi:10.10 02/ceat.20 050 0167 [29] V.M Bolanos-Garcia, O.R Davies, Structural analysis and classification of native proteins from E coli commonly co-purified by immobilised metal affinity chromatography, Biochim Biophys Acta 1760 (9) (2006) 1304–1313, doi:10.1016/j bbagen.2006.03.027 Conclusion In general, we can conclude that NTA ligands result in lower co-elution of proteins irrespective of the chelated metal ion In case of single step purification, NTA is still preferable In the case of a more complex downstream process the lower purity in the capture step can be compensated in subsequent unit operations For large scale manufacturing the toxic metal ion Ni2+ can be replaced by other metal ions Zn2+ IDA, in particular had a similar HCP, DNA, and endotoxin profile as Ni2+ NTA Unfortunately, there is no simple prediction of co-elution with proteomics tools Co-elution appears to be determined by either clusters of metal interacting amino acids on the surface of the protein or through protein-protein interaction of proteins adsorbed on the stationary phase The limitation of the prediction is knowledge about the interactome and knowledge about the 3D structure of proteins Our results match closely with the results from Bolanos-Garcia et al [29] who identified 18 commonly co-eluting proteins from E coli in 2006 We were able to identify 15 out of the 18 proteins described Out of those, 14 had an MBCS higher than 75 Future studies using other model proteins than GFP might be able to elucidate which protein-protein interactions are specific to the protein of interest and which are host specific Declaration of Competing Interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper Acknowledgements We thank Clemens Grünwald-Gruber for performing the MS analysis The MS equipment was kindly provided by the EQ-BOKU VIBT GmbH and the BOKU Core Facility for mass spectrometry We thank our company partners at Boehringer-Ingelheim RCV Process Bioscience for their collaboration and fruitful discussions This work has been supported by the Federal Ministry for Digital and Economic Affairs (bmwd), the Federal Ministry for Transport, Innovation and Technology (bmvit), the Styrian Business Promotion Agency SFG, the Standortagentur Tirol, Government of Lower Austria and ZIT - Technology Agency of the City of Vienna through the COMET-Funding Program managed by the Austrian Research Promotion Agency FFG The funding agencies had no influence on the conduct of this research Supplementary materials Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.chroma.2020.461649 References [1] H Block, B Maertens, A Spriestersbach, N Brinker, J Kubicek, R Fabis, J Labahn, F Schäfer, Chapter 27 immobilized-metal affinity chromatography (IMAC): a review, in: R.R Burgess, M.P Deutscher (Eds.), Methods in Enzymology, Academic Press, 2009, pp 439–473, doi:10.1016/S0076- 6879(09)63027- [2] L Fanou-Ayi, M Vijayalakshmi, Metal-chelate affinity chromatography as a separation tool, Ann N Y Acad Sci (1983) 300–306 [3] M Kenig, S Peternel, V Gaberc-Porekar, V Menart, Influence of the protein oligomericity on final yield after affinity tag removal in purification of recombinant proteins, J Chromatogr A 1101 (1–2) (2006) 293–306, doi:10.1016/j chroma.2005.09.089 10 ... obtained after immobilized metal affinity chromatography A systematic evaluation of the influence of ligands and metals on the selectivity was done mainly for human serum proteins and oligo- and polynucleotides... recombinant proteins expressed in E coli Within IMAC, the use of nitrilotriacetic acid (NTA) as the ligand and Ni2+ ions as the immobilized metal has become standard [13] Other ligands, such as... affects the binding and interaction with the immobilized metal ions Other metal ions, such as Cu2+ , Zn2+ and Co2+ can offer varying binding strengths and selectivity Cu2+ and Zn2+ in particular