Increase in cysteine-mediated multimerization under attractive protein–protein interactions

Abstract The CASPON enzyme became an interesting enzyme for fusion protein processing because it generates an authentic N-terminus. However, the high cysteine content of the CASPON enzyme may induce aggregation via disulfide-bond formation, which can reduce enzymatic activity and be considered a critical quality attribute. Different multimerization states of the CASPON enzyme were isolated by preparative size exclusion chromatography and analyzed with respect to multimerization propensity and enzymatic activity. The impact of co-solutes on multimerization was studied in solution and in adsorbed state. Furthermore, protein–protein interactions in the presence of different co-solutes were measured by self-interaction chromatography and were then correlated to the multimerization propensity. The dimer was the most stable and active species with 50% higher enzymatic activity than the tetramer. Multimerization was mainly governed by a cysteine-mediated pathway, as indicated by DTT-induced reduction of most caspase multimers. In the presence of ammonium sulfate, attractive protein–protein interactions were consistent with those observed for higher multimerization when the cysteine-mediated pathway was followed. Multimerization was also observed under attractive conditions on a chromatographic stationary phase. These findings corroborate common rules to perform protein purification with low residence time to avoid disulfide bond formation and conformational change of the protein upon adsorption.


Introduction
Circularly permuted caspase-2, in particular CASPON enzyme, became a potentially interesting industrial enzyme for fusion protein processing. [1][2][3] This cysteine-dependent protease is the essential element of a platform process for the production of recombinant proteins. [3] The CASPON enzyme is a mutant of human caspase-2 (wtCasp2) [1,4] and exhibits increased enzymatic activity as well as manufacturability compared to wtCasp2. [1,2,5] The high manufacturability of the CASPON enzyme is partially due to the use of a solubility tag. [2] Figure 1 shows the wtCasp2 crystal structure (Figure 1(A), PDB accession number: 1PYO) [4] and an AlphaFold prediction of the CASPON enzyme structure (Figure 1(B & C)). [6][7][8] The most notable difference is the large, relatively unstructured solubility tag of the CASPON enzyme (Figure 1(C)). The core is structurally similar to wtCasp2, differing in only four amino acids. [1] Human caspases are commonly active as dimers of heterodimers ( Figure 1(A)). The multimerization state of the CASPON enzyme is expected to be dimeric due to circular permutation of the heterodimers into a single chain ( Figure 1(B & C)). [9] As the CASPON enzyme dimer contains 26 cysteines, cysteine-mediated multimerization may cause product loss during manufacturing as previously reported for other cysteine-bearing proteins. [10][11][12] The CASPON enzyme contains six free, exposed cysteine residues on its surface (Figure 1(C)). [4] These exposed cysteine residues can potentially form intermolecular disulfide bridges to caspase molecules and other cysteine-bearing molecules alike, resulting in the formation of multimers or aggregates. Moreover, since caspases are cysteine-dependent aspartate-directed proteases, the oxidation state of the cysteine in the active site critical to the enzymatic activity. Generally, solubility of large solutes, such as multimers, is decreased due to increased free energy of cavity formation. [13,14] Thus, the question arises whether cysteine-mediated multimerization and disulfide bond-based aggregates are a critical quality attribute for the CASPON enzyme.
Contrary to aggregation due to protein unfolding, [15][16][17] cysteine-mediated aggregation may not result in changes to the protein tertiary structure. [18] As a result, the native fold and enzymatic activity can be maintained. This also means that disulfide-linked multimers or aggregates can be reduced to their smallest functional multimerization state using reducing agents such as dithiothreitol (DTT). [19,20] Therefore, protein aggregation through the cysteine-mediated pathway can be reversed during the manufacturing process or before use. Nonetheless, proteins recovered by reversing disulfide-linked multimerization may have lower product quality as native disulfide bonds may undergo chemical modifications such as b-elimination. [21,22] Disulfide bond formation is a highly pH-dependent process due to higher nucleophilicity of cysteines' deprotonated thiolates. [23,24] Hence, conditions favoring the formation of aberrant disulfide bonds should be avoided when processing cysteine bearing peptides or proteins, that is, incubation of cysteine-bearing proteins above pH 7. In many bioprocesses, however, slightly basic conditions are prerequisites for numerous unit operations such as the purification of polyhistidine-tag bearing proteins using immobilized metal ion affinity chromatography (IMAC) and the purification of proteins with an isoelectric point in the neutral range using anion-exchange chromatography (AEX).
Co-solutes, such as ions, can induce attractive and repulsive protein-protein interactions, [25,26] potentially increasing spatial proximity of thiol moieties [24] and therefore cause multimerization. As proteins can reversibly associate under attractive conditions, the residence time and orientation of such reversible multimers dictate spatial proximity and residence time of intermolecular thiol/thiolate pairs. Therefore, protein-protein interactions can have a substantial impact on cysteine-mediated multimerization. Chaotropic salts such as guanidium hydrochloride can cause aggregation by simultaneously destabilizing the tertiary and quaternary structures. [15,16] The analysis of the aggregation state of a protein can be achieved by several techniques. [27][28][29][30] While native gel electrophoresis allows separation of proteins with different sizes, accurate determination of the size and quantity is difficult. [31] Sodium dodecylsulfate polyacrylamide gel electrophoresis (SDS-PAGE), on the other hand, is not suitable for the analysis of non-covalent aggregates due to the denaturing conditions used. SDS-PAGE may also have additional complications when analyzing proteins with high cysteine contents. For example, traces of oxidizing compounds can induce disulfide-bond formation and therefore obscure the native aggregation state. [32] High-performance size exclusion chromatography (HP-SEC) is the standard technique [30] that offers several advantages: elution of the protein in its native form and the possibility of integration with a multi-angle light scattering (MALS) detector, allowing determination of the molar mass irrespective of the shape of the protein. [31,33] In this work, we studied the multimerization of CASPON enzyme, as well as the stability and enzymatic activity of its multimerization variants. Isolated CASPON enzyme species were characterized with respect to their molecular weight and enzymatic activity. The most stable and active species was exposed to a kosmotropic, neutral, as well as a chaotropic salt and its multimerization state was monitored over time via HP-SEC. Multimerization rates were correlated to protein-protein interactions obtained by self-interaction chromatography experiments. Moreover, we investigated the difference between multimerization in the liquid phase versus in a chromatographic solid phase to elucidate the possible origin of CASPON enzyme multimerization variants. Overall, this study aims to assess whether multimerization is a critical quality attribute for the investigated protein in downstream processing, formulation and application as an industrial protease. In the broader context, this study can provide guidance in the manufacturing, formulation, and the of application of cysteine-bearing proteins.

Buffer preparation
All employed chemicals were obtained from Merck KGaA (USA). pH of all buffers was adjusted using either 10 M NaOH or 25% HCl to achieve the desired pH value with a maximum deviation of ± 0.05. Phosphate buffered saline (PBS) used as a reference buffer contained 20 mM Na 2 HPO 4 and 150 mM NaCl at pH 7. When investigating the impact of different salts on the multimerization behavior of the CASPON enzyme, solutions were buffered with 60 mM Na 2 HPO 4 .

Model protein: the CASPON enzyme
The model protein, the CASPON enzyme, was produced inhouse. After expression of the protein in E. coli, a two-step purification process was applied. The purification procedure has been adapted from Lingg et al. [1] We have employed an additional SEC polishing step on an € AKTA pure system (Cytiva, Sweden). Parameters of the SEC are shown in the supplementary materials (Table S1). For the preparation of CASPON enzyme dimer, SEC polishing loads were incubated for 15 min with either 25 mM TCEP or 25 mM DTT to maximize yield. After SEC polishing, the protein solution was either directly employed or further concentrated to up to 18 mg Ã ml À1 using Amicon Ultra-15 centrifugation tubes (Merck Millipore, USA) with a 10 kDa cutoff. The former preparation was used for the investigation of CASPON enzyme's multimerization behavior and the latter for further analysis of multimerization mechanisms, SIC experiments and the chromatographic binding experiments. For analysis of individual CASPON enzyme species, fractions at peak maximum (as indicated by 280 nm absorbance) were used. Extinction coefficients are 15.6 and 0.557 ml Ã mg À1 Ã cm À1 at 214 and 280 nm, respectively.

HP-SEC
A detailed description of HP-SEC parameters can be found in the supplementary materials (Table S2). In brief, a TSKgel G3000 SWXL analytical column was used at a flowrate of 0.4 ml Ã min À1 for all samples that did not contain imidazole. For imidazole containing samples, a Superdex 200 Increase 10/300 GL (Cytiva, Sweden) was used at a flowrate of 0.5 ml Ã min À1 . For the determination of molar mass, MALS was performed on a DAWN HELEOS (Wyatt, USA) and analyzed with the ASTRA software (Wyatt, USA). Samples were filtered with a 0.22 mm filter prior to all injections and injection volume was 100 ml. Absorbance at 214 as well as at 280 nm were monitored depending on the concentration of the analyzed CASPON enzymes species.

FRET activity assay
A FRET assay was used for the determination of proteolytic enzymatic activity of the CASPON enzyme. The procedure was already described in literature [5] . In brief, 1 mM CASPON enzyme was incubated with 75 mM Abz-VDVAD#SA-Dap(Dnp) (Bachem AG, Germany). Fluorescence was measured on an Infinite M200 Pro plate reader (Tecan Group AG, Switzerland) and the reaction buffer was 50 mM HEPES, 150 mM NaCl, pH 7.2. The initial slope of the fluorescence signal (excitation: 320 nm, emission 420 nm, gain: 110) was measured and enzymatic activity was calculated based on the slope of CASPON enzyme standards. Enzymatic activity units (U/g) correspond to the catalysis of 1 mmol substrate per minute and gram of the CASPON enzyme.

Multimerization in the presence of co-solutes
For analysis of the CASPON enzyme dimer multimerization over 7 days, two-fold stock solutions of buffers were used to achieve the desired salt concentration after 1:2 dilution, resulting in a protein concentration of 1 mg Ã ml À1 . After sample preparation, samples were incubated on a thermomixer (Thermo Fisher Scientific, USA) at 300 rpm at either 4 or 25 C, whereas the impact of the shaking frequency was not investigated. After incubation, samples were directly analyzed via HP-SEC. Due to the long analysis time of HP-SEC and overall variability of the absorbance signal in the HP-SEC analysis, the relative content of the different species instead of the concentration of each species was compared between the co-solute conditions. Dimer depletion rates were calculated based on the decrease of dimer content over time, in which the decrease over 7 days were used for all conditions except for ammonium sulfate. For ammonium sulfate, rates were calculated at different time points at which 30-50% of dimer could still be detected. For testing reducibility of CASPON enzyme multimers that were formed either in the presence of ammonium sulfate or guanidine, CASPON enzyme dimer preparations with a concentration of 14-18 mg Ã ml À1 were directly diluted to 1 mg Ã ml À1 in the corresponding buffer and incubated at room temperature for 168 hours. 50 mM DTT was added to the samples to investigate the reducibility of formed multimers, incubated for 15 min and then further diluted to 0.2 mg Ã ml À1 for HP-SEC analysis. As a reference, H 2 O was added instead of DTT.

Self-interaction chromatography
The CASPON enzyme was immobilized on a prepacked 1 ml HiTrap NHS-Activated HP affinity column (Cytiva, Sweden). The detailed coupling procedure can be found in the supplementary materials.
After coupling the protein to the column, unbound protein was determined photometrically to calculate surface coverage. For calculation of the surface coverage, we have followed the same procedure as earlier reported in literature [30] . The CASPON enzyme's radius was estimated combining the Einstein-Stokes equation and the relation derived by Tyn and Gusek [31] : Considering the molar mass of the CASPON enzyme dimer of 70.626 kDa (which is equivalent to a r h of 3.29 nm as determined by Equation 1) and a surface area of 43.7 m 2 Ã ml À1 [34] , the immobilized protein content of 18.1 mg Ã ml À1 is equivalent to a surface coverage of 12 %.
For each condition in the SIC experiments, 50 ml of CASPON enzyme with a concentration of 1.5 g Ã l À1 were injected in triplicates on a 1220 Infinity LC (Agilent Technologies, USA) HPLC system at a residence time of 10 min. The column was operated at either room temperature or 0-4 C. Throughout our experiments on the HPLC system, air was frequently entrapped in the system, leading to ripples in the UV signal. Since the entrapment could not be avoided, affected chromatograms were smoothed by applying a gaussian filter using Mathematica 12.1 (Wolfram Research, Inc., USA). Smoothed chromatograms are marked as such.
For analysis of the protein, absorbance at 280 nm was used, except for conditions with a retention time of >1.5 CV or low overall signal, in which case absorbance at 214 nm was used. Figures containing absorption data at 214 as well as 280 nm are normalized according to their corresponding extinction coefficient.
For the calculation of B 22 , the partition coefficient of the protein under interactive (K overall ) and non-interacting condition (K SEC ) is needed. Generally, the partition coefficient is equivalent to the retention factor and is given by Equation 2: Where V R is there retention volume of the protein in the mobile phase, V 0 is the extra particle volume, and V t is the total volume of the mobile phase. B 22 is calculated by the following Equation 3: where V i is the intra-particle pore volume, N is the number of immobilized proteins, and M w is molar mass of CASPON enzyme dimer (70.626 kDa). We assumed that the CASPON enzyme can access the entire mobile phase and therefore K SEC was set to 1. For the calculation of V 0 and V t , extraparticle porosity e and particle porosity e P are assumed to be 0.30 [35] and 0.84 [34] , respectively.

Multimerization in IMAC
IMAC Sepharose FastFlow (Cytiva, Sweden) 50% resin slurry was incubated with the same volume of 50 mM NiSO 4 (thus two-fold resin volume). After incubating the IMAC resin with Ni 2þ , the resin was kept in 20% ethanol for storage. For binding experiments, the slurry was washed with equilibration buffer, which was either 60 mM Na 2 HPO 4 , 1 M ammonium sulfate, 20 mM imidazole, pH 8 or 60 mM Na 2 HPO 4 , 1 M sodium chloride, 20 mM imidazole, pH 8. 20 mM imidazole was used to suppress unspecific binding. The CASPON enzyme dimer stock (17 mg Ã ml À1 ) was diluted with equilibration buffer to achieve a protein concentration of 2.78 mg Ã ml À1 , which was further diluted to a final protein concentration of 2.5 mg Ã ml À1 by addition of the 50% slurry. Final resin concentration was 5% and total volume was 200 ml. After addition of the chromatographic resin to the protein solution, the solution was incubated for 30 min on a tube rotator at 300 rpm (Antylia Scientific, USA) for binding. After removal of 100 ml of the supernatant, the mixture was washed two times with 700 ml equilibration buffer and then incubated without withdrawing the liquid from the second wash for 6 h on the tube rotator at 300 rpm. Finally, 750 ml of the supernatant were withdrawn and 200 ml of IMAC elution buffer (300 mM imidazole, 20 mM Na 2 HPO 4 , 300 mM NaCl) were added to the slurry to elute the bound protein. After filtration of the eluted protein, HP-SEC analysis was performed using Superdex 200 Increase as stated above. Each experimental condition was performed in triplicates.

Results & discussion
Caspon enzyme multimerization states: stability, enzymatic activity and reducibility Similarly, the CASPON enzyme tetramer could be identified (experimental: 143,000 Da, theoretical: 141,252 da, Figure 2(B & C)). Higher order multimers were also identified; however, they could not be resolved and showed a broad distribution of molar masses, indicating the presence of hexameric and higher forms of the CASPON enzyme ( Figure 2(C)). Separation of higher order multimers could probably be achieved by HP-SEC columns with a larger pore diameter. However, the focus of this study was on the dimeric species since it is the most active and hence most relevant multimerization state. For the isolated fractions, SDS-PAGE could not differentiate different size variants and showed mostly monomer with a minor dimer band ( Figure  S3, supplementary material).
To determine whether heterogeneous multimerization of the product affects product quality, CASPON enzyme dimers, tetramers, and the broad higher order multimer fractions were separated via preparative SEC. The resulting CASPON enzyme fractions were immediately transferred to 4 C, stored at the initial elution concentration and analyzed for their enzymatic activity and stability over a time course of 48 h. Figure 3(A) shows the time course of the multimerization state and the corresponding recovery. SEC profiles of the CASPON enzyme dimer do not change significantly over 48 h, whereas the multimer profiles of the tetramer fraction change significantly. After 48 h, residual dimeric and tetrameric CASPON enzyme in the mainly tetrameric fraction multimerized to higher order multimers. This suggests that dimers have a stable multimerization state, while higher order multimers are more transient in nature. This is also reflected in the recovery, as the recovery of dimeric fraction is high (96%), while recoveries of tetrameric and higher order fraction decreases to under 40% after 48 h. Since SEC only shows soluble multimers or aggregates up to a certain size, it is likely that the observed loss in area stems from the formation of very large or insoluble aggregates. Interestingly, multimerization of tetrameric and higher order species was higher even though the sample concentration was lower compared to that of the dimeric fraction. It appears that increasing multimer size correlates with decreased stability in the investigated phosphate buffer. This could be due to higher free energy of cavity formation. It was shown that free energy of cavity formation in water correlates for length scales >1 nm with the surface area, [13] which could explain decreased stability of higher order multimers.
Enzymatic activity was monitored over the course of two days for the dimeric and tetrameric fraction (Figure 3(B)). For the higher order multimers, enzymatic activity could not be determined since protein concentration was too low for the assay. Concentration via membrane concentrators of this fraction resulted in multimerization itself (data not shown), hence the activity measurement of the original higher order multimer fraction was not feasible. Generally, the CASPON enzyme dimer exhibits a higher enzymatic activity than the tetramer fraction. Furthermore, enzymatic activity decreases can be observed for both fractions, where the relative loss in enzymatic activity is more pronounced in the tetrameric fraction. Oxidation of cysteines in the active sites could be the cause of reduced enzymatic activity. The cysteine in the active site can be oxidized via dissolved oxygen, but can also be involved in cross-linking between the tetrameric or higher order multimers, therefore decreasing the total number of functional active sites. Since tetramer and higher multimer fractions show decreased enzymatic activity and stability, maintaining the CASPON enzyme in the dimeric state is beneficial.
In order to investigate the multimerization pathway, dimer, tetramer, and higher order multimer fractions were incubated with 10 mM DTT. Enzymatic activity and SEC profiles were compared to those of untreated samples, shown in Figure 4. Note that peaks after 30 min are DTT associated ( Figure S2, supplementary materials) and the protein concentration varied for each sample. All investigated samples showed a higher dimer content after DTT treatment as tetramer and higher order multimer content decrease in the normalized chromatograms ( Figure 4) and the area of the dimer fraction in the raw chromatograms increases ( Figure S1). This effect is more pronounced for the tetramer and higher order multimer fractions. Enzymatic activity of the dimeric and tetrameric fraction was increased after addition of DTT. This increase was most pronounced for the dimeric fraction (188%) and moderate for the tetrameric fraction (13%). Protein concentration was again too low for the determination of the enzymatic activity of the higher order multimer fraction. The CASPON enzyme monomer cannot be detected throughout the experiment. Moreover, we have analyzed all fractions with and without DTT using SDS-PAGE ( Figure S3, supplementary material). In the SDS-PAGE analysis, only the CASPON enzyme monomer and very low amounts of the dimer could be detected for all investigated samples. Altogether, we believe that SDS-PAGE is not a suitable analytical method for investigating disulfide-bond linked CASPON enzyme multimers.
Reduction of CASPON enzyme tetramers and higher order multimers is a strong indication for multimerization formed by covalent disulfide bridges. As indicated by the AlphaFold prediction (Figure 1(C)), the cysteine close to the hexa-histidine tag appears to be exposed in a flexible region, which could be the most prominent site for disulfide bond formation. However, tetrameric and higher order multimers were only partly reduced, hence it is not clear whether cysteine-mediated multimerization is accompanied by another multimerization mechanism or more reducing agent would be needed. Other recombinant proteins have also been reported to undergo cysteine-dependent multimerization. [10,11,20,36] For monoclonal antibodies, cysteine-dependent multimerization can occur during manufacturing [36] and after exposure to thermal stress, [11] correlating overall to free thiol content. [11,36] Cysteine-dependent multimerization in recombinant proteins derived from E. coli seems to be more prominent overall. [10,20] Schweizer and coworkers reported that a disulfide bridge in the dimer interface stabilized the dimeric state in wtCasp2 (4). Interestingly, no monomerization occurred upon DTT addition. We hypothesize that the reduction of the internal disulfide bridge is either sterically hindered or non-covalent forces between the monomers are strong enough to keep the dimer intact in the employed buffer. With regards to enzymatic activity, the dimeric fraction experienced a higher replenishment of the enzymatic activity compared to the tetrameric fraction. This could be explained by preferential reduction of disulfide bonds compared to oxidized cysteine in the active site. SDS-PAGE is not suitable for the analysis of the size variants since tetrameric or higher species could not be detected irrespective of DTT addition. We hypothesize that harsh sample preparation and non-native conditions could alter the native multimerization state. Disulfide reshuffling and oxidation to higher order multimer species could occur which might be visible as slight smeared bands in the SDS-PAGE. Nevertheless, SDS-PAGE confirmed the high purity of the investigated samples and the absence of host cell impurities.

Multimerization behavior of dimeric CASPON enzyme
Since the dimeric species of the CASPON enzyme is most stable and active, we further investigated its multimerization Figure 4. Effect of DTT on tetramers and higher order fractions: SEC-HPLC chromatograms of the dimeric fraction (0.9 g Ã l À1 ) that exhibited a 188% increase in enzymatic activity (A), tetrameric (0.4 g Ã l À1 ) with a 12.5% increase (B) and heterogenous higher order multimer fraction (0.2 g Ã l À1 ) (C). Samples were either incubated with 10 mM DTT or were directly analyzed. Size variants identified with SEC-MALS are highlighted and marked: higher order multimer (HO), tetramer (TE) and dimer (DI). Corresponding raw data can be found in the supplementary materials ( Figure S1). behavior in the presence of kosmotropic, neutral and chaotropic salts (ammonium sulfate, sodium chloride and guanidium hydrochloride, respectively). We have selected a pH range of 6-8 to study multimerization behavior. At this practically relevant pH range, the protonation state of cysteines varies from partially protonated and fully protonated to partially deprotonated and fully deprotonated at pH 6 and 8, respectively [23] . Temperatures for incubation were set to either 4 C or 25 C to emulate conventional temperatures in downstream processing. During this experiment, minor peaks occurred at higher retention times than the dimeric species and due to their later retention time compared to the dimer in SEC, we identified the species as the CASPON enzyme monomer. Figure 5 shows the evolution of the multimeric state of the CASPON enzyme with respect to different salt exposure, pH, and temperature. As expected, multimerization is higher at 25 C compared to 4 C for all salts. In the presence of 1 M ammonium sulfate and 1 M sodium chloride, multimerization increases with solution pH, with ammonium sulfate inducing more pronounced multimerization. Highest multimerization was observed in the presence of 1 M ammonium sulfate at pH 8 and 25 C, where multimerization was over 90% after 3 days. In comparison, multimer content was below 20% after incubation for 3 days with sodium chloride at pH 8 and 25 C. For 1 M guanidium chloride, temperature had a tremendous impact on multimerization. Almost no multimerization was observed at 4 C, whereas multimerization can be observed for all tested pH values at 25 C. Ammonium sulfate and sodium chloride also induce monomerization compared to the reference. However, monomer content was lower than 20% throughout the experiment for those two salts. Multimerization behavior does not correlate with pH when the enzyme is incubated with guanidium hydrochloride. pH 6 induced pronounced monomerization in the presence of guanidium hydrochloride as indicated by a maximum monomer content of over 20%. While a high monomer content was observed throughout the experiment at 4 C, monomer content decreases abruptly after one day at 25 C. After one day, decreased monomer content coincides with increased multimerization.
In the presence of 1 M ammonium sulfate and 1 M sodium chloride, multimerization increases at higher pH and thus higher thiolate content. This indicates a predominantly cysteine-dependent multimerization because of higher reactivity of thiolates. However, multimerization rates differ substantially between ammonium sulfate and sodium chloride. These effects are generally in accordance with the Hofmeister series. [37] High monomer content at pH 6 indicates an overall destabilization of the monomer-monomer interface that is further enhanced by disruption of noncovalent interactions by guanidine. [16,17,38] Interestingly, guanidine induced monomerization followed by multimerization at 25 C. Since guanidine can potentially cause unfolding of the protein, we hypothesize that formed multimers could have an altered tertiary structure and hence follow a denaturing multimerization pathway. At 4 C however, monomerization occurs in the presence of guanidine while multimerization does not occur. Thus, the disruption of the dimer interface is not necessarily accompanied by unfolding of the CASPON enzyme which is potentially increased at 25 C. Lastly, the reference experiment shows that the CASPON enzyme dimer is stable for at least one day in PBS with a comparably low ionic strength.
As previously mentioned, cysteine-dependent multimerization can also occur in IgG1 and IgG2; however, it is much lower compared to the multimerization of CASPON enzyme in the reference PBS buffer. Even after thermal stress exposure for 12 weeks at 45 C, total aggregate content of an anti-streptadivin antibody preparation was below 10% and reducible aggregate content was 2-5%. [11] On the other hand, cysteine-mediated multimerization appears to be more prominent when overexpressing recombinant, cysteine-bearing proteins in E. coli. Depending on the protein of interest [10,20] and E. coli strain, [20] disulfide-linked aggregate content directly after purification varies from under 10% to a considerably higher aggregate content. [10,20] When incubating the cysteine-bearing extracellular domain of human CD83 at room temperature, higher cysteine-dependent multimerization can be observed, where disulfide-linked aggregate content reaches 30-70%. [10] This is considerably higher compared to the multimerization of the CASPON enzyme in the low ionic strength reference conditions. In PBS, the CASPON enzyme preparation contains under 20% aggregate content after 7 days.
To further investigate the differences of multimerization caused by ammonium sulfate and guanidine, we incubated the CASPON enzyme over 168 h (7 days) at room temperature with these salts. For guanidine, pH 6 was selected due to prominent monomerization followed by multimerization ( Figure 5). For ammonium sulfate, pH 8 showed highest multimerization content after 7 days and thus represented an interesting condition. After incubation over 7 days, both preparations were analyzed with and without DTT. Figure 6 shows the resulting HP-SEC chromatograms. After incubation with ammonium sulfate, the overall signal is lower than 90% compared to all other samples. Nevertheless, addition of DTT reduces higher order multimers back to dimers (and to a much smaller extent, monomers), indicating strictly cysteine-mediated multimerization. Guanidine on the other hand causes less multimerization compared to ammonium sulfate and incubation with reductant cannot reduce all higher order multimers. Furthermore, monomer content is highest of all investigated samples, indicating that after monomerization of the CASPON enzyme, monomers can covalently link via disulfide bridges to non-native dimers or higher order multimers. After DTT addition, the total UV area of the SEC chromatogram of guanidine treated samples is considerably lower compared to that of the samples incubated with ammonium sulfate, indicating a higher proportion of non-reducible higher order multimers that are too large to be detectable after 0.22 mm filtration. We suggest that guanidine induces a denaturing pathway and mixed pathway due to its salting in effect, leading to the formation of multimers with an altered tertiary structure.

Protein-protein interactions
In this section, we investigate the correlation between protein-protein interactions and multimerization rates. Proteinprotein interactions were determined by SIC experiments, where dimeric CASPON enzyme was present as an analyte in the mobile phase and as a ligand in the stationary phase. As a reference, highly interactive conditions were also tested with a column that carried immobilized Tris instead of the CASPON enzyme.
Firstly, 1 M ammonium sulfate induces strong attractive protein-protein interactions that cannot be quantified using SIC due to backbone interactions ( Figure 7) and low recovery (Table 1), respectively. As seen in Figure 7, protein retention is higher at increasing pH values. Furthermore, retention increases with increasing temperature with the CASPON enzyme as a ligand and with Tris as a ligand. Overall, retention on NHS Sepharose with immobilized CASPON enzyme is higher compared to that on the bare column with Tris as a ligand in addition to recoveries dropping below 22%. It cannot be stated whether the enzyme binds irreversibly or reversibly to the SIC column. Initially, formed multimers could be reversible due to protein-protein interactions as reported in literature [39] and become irreversible as disulfide bonds form over time. Once again, higher pH values likely increase reaction rates of disulfide bond formation due to higher concentration of thiolate anions on the surface of the protein.
Moreover, protein-protein interactions may additionally contribute to higher multimerization rates at higher pH. Interestingly, binding of the CASPON enzyme to Tris-Sepharose at pH 8 and 25 C appears to change the interactions with protein remaining in the mobile phase. After each injection, retention time increases and recovery decreases, suggesting successive deposition of the CASPON enzyme and increased interactivity of the stationary phase to the CASPON enzyme in the mobile phase. Secondly, 1 M sodium chloride and 1 M guanidine buffers induce weak repulsive protein-protein interactions and dimer depletion showed no correlation to B 22 in the presence of those two salts ( Figure S4, supplementary materials). Protein-protein interactions were comparable at different temperatures ( Table 2). In the presence of sodium chloride, protein-protein interactions are increasingly repulsive with increasing pH. When guanidine is employed, protein-protein interactions are equally repulsive irrespective of pH.
While multimerization was highly temperature dependent in the presence of guanidinium hydrochloride and sodium chloride, increasing protein-protein interactions did not correlate with increasing temperature. For sodium chloride, increasing temperature could simply allow faster kinetics and thus increase dimer depletion rates. For guanidine, proteinprotein interactions were not indicative of multimerization whatsoever. This suggests that protein-protein interactions might not be a useful parameter in predicting multimerization via a denaturing multimerization mechanism.
The impact of protein-protein interactions on multimerization in the adsorbed phase was investigated by batch adsorption of CASPON enzyme on Ni Sepharose FF resin. The protein was adsorbed for 6 h under either highly attractive (1 M ammonium sulfate, pH 8) or slightly repulsive (1 M    NaCl, pH 8) conditions and subsequently eluted and evaluated. This contact time is typical for a purification process using IMAC. [40][41][42] Figure 8 shows normalized HP-SEC chromatograms (raw data can be found in supplementary materials, Figure S5) of the eluted CASPON enzyme after adsorption in the presence of different salts. The relative dimer content in the eluate is equal to that of the load at 57% when incubated with 1 M sodium chloride. In the presence of 1 M ammonium sulfate, the relative dimer content decreases from 57% in the load to 45% in the eluate. Multimerization still occurs in the adsorbed phase under attractive conditions. Immobilized proteins can still interact with other proteins in close proximity. Unfortunately, mass balances were difficult to determine and were unreliable due to the low volume of stationary phase, leading to low recoveries (Table S3).

Mechanism of action
Understanding multimerization mechanism can facilitate process development and improvement for cysteine-bearing proteins, specifically for the CASPON enzyme. In a manufacturing process, reductants could be added to buffers in certain downstream unit operations (ion-exchange chromatography, hydrophobic interaction chromatography, buffer exchange, filtration) if a strictly cysteine-mediated multimerization mechanism is followed ( Figure 9). This could ultimately increase product yield due to reduction of disulfide-bond linked multimers. If target protein aggregates are reversed during the process, potential chemical modifications of the product must be monitored. In the final formulation, low reductant concentration could reduce the active center to replenish the enzymatic activity and reverse newly formed multimers. Reductants show pH-dependent activity which could limit their application to defined pH ranges. [43] The addition of chaotropic and denaturing reagents such as guanidine is rare except for refolding processes. [44] Nevertheless, the enzyme could enter a denaturing or mixed pathway ( Figure 9) in several unit operations that can cause partial unfolding. Partial unfolding of proteins can be observed upon adsorption in HIC [45,46] or ion-exchange chromatography (IEX). [15,47] In general, multimers obtained from the denaturing pathway are more problematic than reducible multimers as their tertiary structure and hence activity might be affected.
Aside from the CASPON enzyme, some general considerations must be made for other cysteine-bearing proteins. Firstly, in-silico tools could help to identify potentially redundant cysteines and avoid cysteine-dependent multimerization a priori when conducting protein design. Secondly, cysteine removal might not always be possible, in which case special emphasis should be put on stability and activity of the protein in presence of the selected reductant. Dedicated stability experiments should be performed by employing analytical techniques to monitor changes of the native state (e.g., SEC) and the denatured state (e.g., reversed-phase HPLC).

Conclusion
Dimeric CASPON enzyme is the most stable and active multimerization species. Under standard process conditions, i.e., in PBS at 25 C for up to 24 h, multimerization of the dimer is negligible, indicating that multimerization is not a critical quality attribute in its application as a protease for tag removal. Generally, multimerization of the CASPON enzyme follows two different pathways or a combination thereof, namely a cysteine-mediated pathway or a denaturing pathway. In the cysteine-mediated pathway, CASPON enzyme multimers are linked via disulfide bonds and the formed multimers are reduceable. Here, protein-protein interactions correlate with dimer depletion rates, where 1 M ammonium sulfate induced highest dimer depletion. In the denaturing pathway, the CASPON enzyme partly unfolds and binds to other proteins to form non-reducible multimers. In the presence of chaotropic guanidine, the CASPON enzyme follows both the denaturing pathway and the cysteine-mediated pathway in a weakly repulsive regime. The multimerization also occurs when binding CASPON enzyme to IMAC Sepharose FF under attractive conditions, indicating that multimerization occurs during downstream processing. Multimerization propensity must be considered when processing such cysteine-bearing proteins. The addition of reductants such as DTT can be considered to reverse multimerization and restore the native state of the protein. This work is in line with common rules to execute protein purification with low contact time and low residence time to avoid disulfide bond formation and conformational change of the protein upon adsorption. Table S1: Preparative SEC parameters with HiLoad Superdex 200 pg (Cytiva, Sweden). Table S2: Detailed HPLC parameters for samples containing imidazole (Superdex 200 Increase 10/300 GL) and without imidazole (TSKgel G3000 SWXL). Figure S1: Chromatogram raw data for Figure 3. Figure S2: Injection of 10 mM DTT solution after incubation at RT for 160 min. Figure S3: SDS-PAGE analysis of dimeric, tetrameric and higher order multimer fraction (NuPAGE (Invitrogen, USA) 4-12% Bis-Tris with MES running buffer). 1: Protein standards. 2: dimer fraction. 3: dimer fraction, 10 mM DTT. 4: dimer fraction, 180 mM DTT. 5: tetramer fraction. 6: tetramer fraction, 10 mM DTT. 7: tetramer fraction, 180 mM DTT. 8: higher order multimer fraction. 9: higher order multimer fraction, 10 mM DTT. 10: higher order multimer fraction, 180 mM DTT. 11: Protein standards. Figure S4: B 22 plotted over dimer depletion rate for sodium chloride and guanidium hydrochloride at 4 and 25 C. Table S3: Mass balances and binding capacities for IMAC binding experiments. Figure S5: All SEC chromatograms of IMAC eluates after 6 h incubation. Top: raw data. Bottom: normalized data. Ammonium sulfate induces strong protein-protein interactions, whereas sodium chloride induces weakly repulsive interactions.