Improved Purification of Human Granzyme A/B and Granulysin Using a Mammalian Expression System

Cytotoxic lymphocytes release proteins contained within the cytoplasmic cytolytic granules after recognition of infected or tumor target cells. These cytotoxic granular proteins (namely granzymes, granulysin, and perforin) are key immunological mediators within human cellular immunity. The availability of highly purified cytotoxic proteins has been fundamental for understanding their function in immunity and mechanistic involvement in sepsis and autoimmunity. Methods for recovery of native cytotoxic proteins can be problematic leading to: 1) the co-purification of additional proteins, confounding interpretation of function, and 2) low yields of highly purified proteins. Recombinant protein expression of individual cytolytic components can overcome these challenges. The use of mammalian expression systems is preferred for optimal post-translational modifications and avoidance of endotoxin contamination. Some of these proteins have been proposed for host directed human therapies (e.g. - granzyme A), or treatment of systemic infections or tumors as in granulysin. We report here a novel expression system using HEK293T cells for cost-effective purification of high yields of human granzymes (granzyme A and granzyme B) and granulysin with enhanced biological activity than previous reports. The resulting proteins are free of native contaminants, fold correctly, and remain enzymatically active. Importantly, these improvements have also led to the first purification of biologically active recombinant human granulysin in high yields from a mammalian system. This method can be used as a template for purification of many other secreted cellular proteins and may lead to advances for human medicine.


INTRODUCTION
The immune response to various intracellular pathogens and tumors includes cytotoxic T lymphocytes (CTLs) and natural killer (NK) cells which recognize and directly kill infected or malignant cells. They are involved in cell mediated immunity and play a major role in host defense against infections by intracellular pathogens including bacteria, viruses and fungi replicating in host cells (1). The effector molecules which kill host cells and intracellular microbial pathogens include a family of serine proteases (Granzymes or Gzms) and a small antimicrobial protein (Granulysin or GNLY) delivered by the pore-forming protein perforin (2).
There are five human granzymes (granzyme A, B, H, K, and M), and ten mouse counterparts. Based on substrate specificity studies, other groups have shown species specific immunological functions of granzymes highlighting a potential divergent evolution process and suggesting that the human counterparts ought to be used for translational work to human medicine (3). Thus, our work cited here will focus on human granzymes and GNLY. Furthermore, granzymes are ubiquitously expressed in CTL and NK cells, particularly GzmB which is even detected in non-cytotoxic immune cells (4,5). As GNLY is not expressed in mice and therefore in vivo experiments to date have been performed only using human GNLY transgenic mice (6)(7)(8). Granzyme A (GzmA) was previously thought to only induce apoptosis of target cells in concert with perforin. Death induction by GzmA involves a complex sequence of events, ultimately leading to the activation and nuclear transfer of two nucleases (NM23-H1 and Trex1) that trigger lethal DNA damage (9)(10)(11)(12)(13). GzmA also induces monocytes to produce pro-inflammatory cytokines (14)(15)(16) and to inhibit the intracellular replication of mycobacterial growth within infected primary human monocytes (15,17). Due to its proinflammatory potential, there have been recent reports of its involvement during bacterial sepsis (16,18,19). To further study these effects, it is imperative that researchers are careful to avoid potential endotoxin contamination in the final purified products: the use of a bacterial expression system will directly contaminate purified proteins, while any other system will contaminate the product if the researcher is not careful throughout the process. In contrast to human GzmA, Granzyme B (GzmB) induces apoptosis of the target cells either by direct or indirect activation of caspase 3 and 7 (20,21). Furthermore, GzmB efficiently activates the mitochondrial death pathway by truncating the pro-apoptotic protein Bid (22) and induces mitochondrial outer membrane permeabilization (23). Ultimately, activated caspases trigger the release of an active DNase (CAD), responsible for DNA fragmentation and nuclear changes during apoptosis (24).
GNLY is a prokaryotic membrane-disrupting, lymphocytic effector protein (25). GNLY alone can kill a wide array of microbial pathogens when micromolar concentrations are used in cell culture (1). However, its intracellular antimicrobial activity is synergistically enhanced by GzmB (26,27). Pathophysiological roles of these proteins have been reported demonstrating their broad implications for understanding immune homeostasis and pathogenic inflammatory diseases (28). Known functions of these lymphocytic effectors heavily relied on in vitro studies using purified proteins in killing assays as well as biochemical and morphological studies.
GzmA/B and GNLY were expressed and purified from various recombinant systems including bacteria (29)(30)(31), yeast (32) and insect cells (33). These systems produce proteins without the post-translational modifications identical to native cytotoxic effectors. Native purifications of granzymes (14,34) and granulysin from NK cells and CTL cell lines like YT Indy and NK92MI yield limited amounts of purified proteins contaminated with other cytolytic granule components (35). The existing recombinant protein expression methods (e.g., bacteria, yeast, and insect) can yield higher concentrations and purity, but also alter tertiary structures and unnaturally absent glycosylations (36,37). The mammalian expression system with HEK293 T cells was shown to produce recombinant GzmA, GzmB and GzmM proteins with higher molecular weights than their bacterial counterparts that can be cleaved using an endoglycosidase, suggesting that glycosylated proteins can be produced in a mammalian system (38).
We present an improved purification process for the purification of GzmA/B and GNLY that builds on previous protocols (35,38). This was accomplished by modifying the transfection method, medium and buffer conditions, and enterokinase (EK) activation and purification steps. Our yields were consistently ten times higher with at least similar purity as compared to previous results (38). Our purified proteins are biologically active with higher specific activity as measured in diverse and complex immunological systems. We also highlight steps that minimize endotoxin contamination by using mammalian expression.

METHODS
Summary of the steps needed for purifying cytotoxic granular proteins: 1. Production of the Expression Plasmid pHLsec-Gzm a. Prepare total RNA from human NK cells using a suitable RNA isolation method and reverse transcribe using a firststrand cDNA synthesize kit. Amplify Gzm cDNA using PCR and clone into pHLsec (39) using the AgeI and KpnI sites. b. For GNLY shift from His tag at the N-terminus (but also applied to Gzms), the forward primer HisEk.consAge1For and the reverse primer GNLYSTOPkpn1Rev were used with the template DNA pHLsecGNLY. The transformation of DNA was done by the high efficiency transformation protocol NEB alpha with C2987I NEB 5alpha competent E. coli cells. After overnight incubation the colonies were harvested. The obtained DNA was sequenced to confirm the cloning process. c. Confirm correct inserts by sequencing. Expand the expression plasmids in DH5a cells and purify using an endotoxin-free plasmid isolation kit and follow the manufacturer's instructions. d. Resuspend the purified plasmids in endotoxin-free, sterile water at a concentration of 2 mg/ml and store at -80°C until use. ii. Prepare transfection medium, which is culture medium without penicillin-streptomycin. iii. Prepare the solutions described in Table 1 and filter with a 0.22 mm filter. Receiver bottle should be singleuse plastic. For water, use MilliQ water or other sterile and endotoxin-free equivalents to avoid endotoxin contaminations. b. Expanding HEK293T cells.

Expansion of HEK293T Cells
i. Grow HEK293T cells in 10 ml culture medium using 10 cm tissue culture dishes. ii. Split cells at 80% confluency (split-ratio of 1:4, usually every 3rd day). Cells loosely attach to dishes and can be mechanically detached without trypsinization by rigorously pipetting up and down. The following trypsinization method allows for a more accurate monitoring of cell heath during cell passaging and cell plating for the next step, so it is our preferred method for passaging cells. To do this, remove medium, wash once carefully with 5 ml of room temperature PBS, then add 2 ml of Trypsin 0.05% EDTA and incubate inside 37°C incubator for 2 min. Add 10 ml of complete medium to quench reaction, remove all cells from dish and collect in 50 ml conical tube. Spin down at 400 x g for 5 minutes at 22°C. Discard supernatant, count the cells and plate in 10 cm tissue culture dishes. iii. Plate cells the night before transfection to give 60-70% confluency at the day of transfection (seed around 5e 6 cells per 10 cm plate in transfection medium-no antibiotics). A typical preparation size consists of 20 to 25 culture dishes with the expected yield around 700 mg of pure protein per plate.

Lipofectamine 3000 Transient Transfection of Gzms and Granulysin
a. For transfection, ensure that cells are confluent around 70-90% as per L3000 product recommendations. b. Perform transfection using lipofectamine 3000 kit workflow.
Use Opti-Mem for diluting L3000, DNA, and P3000. Do not mix L3000 with DNA directly because the DNA will and 100 ml of An5 buffer is added to each well. After centrifugation for 3 min at 500 x g, the cells are resuspended in 100 ml An5 buffer containing APCconjugated Annexin V (1:33 dilution) and incubated for 10 min at room temperature in the dark. The cells are washed once in 100 ml An5 buffer and resuspend in An5 buffer containing 2 mg/ml propidium iodide and analyzed by flow cytometry (35). iv. Biological assay is the ability of GzmB in combination with GNLY to suppress the growth of extracellular E. coli. 250 nM of GzmB are incubated with 100 nM of GNLY and bacterial growth is measured over time by OD 600 readouts (26,27). d. For Granulysin: i. Western Blot that looks at His-tag presence using Anti-His antibody. Thus, EK cleavage efficiency can be monitored as final product will become Anti-His negative. ii. Biological assay is the ability of GNLY to suppress the growth of extracellular E. coli. 100 nM of GNLY is incubated and bacterial growth is measured over time by OD 600 readouts (26,27).

Statistical Analysis
For generation of graphs and statistical analysis we used GraphPad Prism version 9.0.0 for Mac, GraphPad Software, San Diego, California USA, www.graphpad.com.

Improved Purification Process for the Production of Gzms
We have modified several steps in the recombinant protein expression and purification protocol for GzmA and GzmB (38), with the process schematically represented in Figure 1.
As shown in Figure 1A, the switch from calcium phosphate to lipofectamine delivery allowed us to use significantly less cells (5e 8 cells vs 1.75e 8 cells), and correspondingly less plasmid DNA (2 vs 0.27 mg). In parallel, the previous transient transfection procedure included only 8-12 hours incubation before changing to a serum-free medium, while our new transfection protocol extends the incubation in medium with serum for 96 hours before adding fresh medium on D4. The previous method included the use of Ex-cell HEK293 serum-free medium (Sigma # 14571C-500ML), which allowed transfected cells to produce protein and proliferate (data not shown). For improved purification, D4 and D8 supernatants were combined, and cell debris removed by centrifugation and filtration. The combined supernatants were incubated overnight with nickel Sepharose beads and then washed with His-buffer A to remove weakly bound proteins. As shown schematically in Figure 1B and in a representative image of A 280 monitoring during His-buffer B elution in Figure 2A, the bound proteins are eluted with a linear gradient from 0-100% of His-buffer B, corresponding to 10-1000 mM imidazole. When 40% His-buffer B (~400 mM Imidazole) is achieved, Gzms start to elute. SDS-PAGE followed by Coomassie stain is used to identify which fractions to combine for the next steps as shown in Figure 2B.
Pooled and concentrated fractions are then incubated overnight with EK to activate the proteins as represented schematically in Figure 1C. It was originally noted that EK cleaves more efficiently at low salt concentrations (e.g., 50 mM NaCl) (38,40). However, we observed significant protein precipitation after overnight dialysis in hypotonic solution as shown in Figure 1D. Thus, we altered the EK buffer to an isotonic solution which prevented protein precipitation and allowed EK to cleave Gzms. The recombinant human EK from CHO cells used in our protocol can cleave and activate target proteins in isotonic buffer conditions as shown by the band shift of Gzms in Figure 2C and the loss of the His signal by western blot in Figure 2D. On the left of Figure 2D, silver stains show that samples pre and post-EK were comparable in terms of protein loading, while on the right, the EK-cleaved GzmB sample shows that only a significantly small portion of total prep remains uncleaved. The recombinant Gzm plasmids were engineered to express the 6x poly His Tag at the Nterminus of the DNA upstream of the EK cleavage site. This is an optional step that allows the researcher to monitor EK activation efficiency by probing post-EK protein with Anti-His antibody; a lower molecular weight confirms that the secretion signal containing the His tag has been cleaved. This shift in MW is more evident with homodimeric GzmA as two His tags are cleaved, while monomeric GzmB has only one.
After confirmation of EK cleavage, proteins were dialyzed into MonoS-buffer A, and then separated as shown in the representative chromatogram shown in Figure 3A. As shown in Figure 3B, the fractions contain predominantly the reduced GzmA monomers (~31 kDa) and were enzymatically active as measured by specific Gzm peptide substrates. The BLT assay is used for GzmA as shown in Figure 3C and for GzmB the AAD assay is used as shown in Figure 3D. Proteins are then concentrated, undergo endotoxin removal using Endotrap columns. As shown by silver stain and western blot in Figures 4A, B, GzmA is highly pure and able to form homodimers (as well as multimerspreviously reported in (14,17) and of unknown significance). For Figure 4C, the differences between native, recombinant GzmA purified using old protocol vs new protocol are analyzed for purity by silver stain. To confirm that all purified proteins are capable of cleaving substrates, the specific activity of each purified GzmA was compared and showed no difference at cleaving BLT as shown in Figure 4D. In Figures 4E, F, highly purified GzmB is shown by silver stain and western blot. Like GzmA, in Figure 4G we compared Native to New recombinant GzmB enzymatic activity at different protein concentrations, and it appears that recombinant GzmB performs better than native counterpart, probably owing to a purer final product. In Figure 4H, the biological activity of GzmA in a Mycobacterial Growth Inhibition Assay (MGIA). While the old purification protocol produced inhibitory GzmA [purified per (38)], the new protocol yields more biologically potent GzmA similar to native GzmA [purified per (35)]. Similarly, in Figure 4I, recombinant GzmB produced using this new protocol performed significantly better than native GzmB [purified per (35)] at suppressing extracellular bacterial growth after delivery into the organisms with GNLY addition. To further compare native vs new recombinant GzmB, we performed an Annexin V and PI staining of target cells incubated with and without perforin, and the results are shown in Figure 4J.
In Table 2, we compared the cell expansion times, total purification times, purity levels, advanced equipment necessary, yields per purification, DNA amounts needed, costs per purification, and the overall efficiency. As shown, our updated protocol provides many advantages compared to other protocols (35,38). As shown in Figure 4K, our system has allowed us to increase the final yields of purified Gzms compared to the yields obtained with other protocols (35,38).

GNLY Purified With the Mammalian Expression System Is Biologically Active
GNLY was purified using the improved protocol as summarized in Figure 1 and production was monitored in Figure 5 using secreted protein supernatants. Similar to Gzms, GNLY elutes from Ni-IMAC column around 40% His-buffer B. In Figure 5A, the medium supernatant at D4, D6, and D8, and the Ni-IMAC fractions show GNLY presence after SDS-PAGE and Coomassie staining. The fractions containing GNLY were pooled for subsequent EK treatment and further purification. In Figure 5B, EK treatment cleaves the His-tag similarly as in Figure 2D. In Figure 5C, the GNLY MonoS fractions are shown as well as the native GNLY protein [purified per (35)] in the second lane for comparison. Figure 5D, pure GNLY is shown by SDS-PAGE and silver stain.

Finally, in
The final yield obtained was 7 mg from 10 -10 cm 2 plates of HEK293T cells (from 200 ml of total supernatant giving a final yield of 35 mg pure protein per ml of supernatant, which is in line with the Gzms purifications). Pooled GNLY fractions are then tested in the antimicrobial assay (26,27). E. coli were treated with 250 nM of native vs recombinant GNLY or left untreated before monitoring bacteria viability in growth assays. Recombinantly expressed GNLY appears to better inhibit bacterial growth as shown in Figure 5E.

DISCUSSION
The purification in high yields of cytotoxic granular proteins in a reliable and consistent manner has been a significant barrier in the field of immunology. Early native purification protocols did not exclude contamination with other granular proteins in the final products (32). For instance, GzmA was routinely co-purified with Granzyme K, and could not be distinguished as they both cleave the BLT substrate. Only with the commercialization of high-quality monoclonal antibodies against GzmA and GzmK have researchers been able to separate fractions to avoid potential co-contaminants from native purifications (35). The sequencing of the human genome and the development of cDNAs that include all human Gzms and GNLY allowed the recombinant expression of these proteins first in bacteria (29)(30)(31), later in yeast (41,42), and finally in insect cells (33,43). While all these methods allow protein production, bacterial expression is associated with contamination of the final product with endotoxins (44), skewing assay results particularly when working with cells that are sensitive to this stimulus. Pichia pastoris expression does not produce endotoxin, however, like bacterial expression systems, the granzyme products do not normally have native glycosylations known to be important for normal protein physiology (45). While monitoring and comparing post-translation modifications in Gzms between native, old recombinant, and new recombinant protocols was beyond the scope of this work, it is important to point out that there continues to be a knowledge gap on their role in biology as highlighted in (4,46,47). In this work, we have used transient transfection in a mammalian expression system using HEK293T cells, lipofectamine 3000, and isotonic buffers. Compared to previous work (38), our protein products (GzmA, GzmB and GNLY) are purified in large yields, with a cost-efficient method, are more biologically active, and are therefore useful in complex immunological assays. Previous expression of GzmB through stable transfection gave purification yields of~4 mg/L of culture supernatant (45), while our new method yields much larger amounts up to 40 mg/L of culture supernatant. Our process involves a faster purification, results in high purity products, does not involve the use of advanced equipment, provides larger yields, needs less plasmid DNA, and overall is significantly more cost-effective ( Table 2). The lower amount of DNA allowed us to use Maxi instead of Giga DNA preps to purify plasmid DNA, which further decreased our protein expression cost. The main improvements in our protocol are: 1) the transition from calcium phosphate transfection to lipofectamine 3000, 2) the use of transfection medium that contains serum proteins for optimal cell viability, 3) the extended production over 96 hours, 4) the use of isotonic buffers throughout the purification process, and 5) the translocation of the His tag upstream of EK site for monitoring protein activation (Figure 1). Use of lipofectamine 3000 or similar reagents are highly recommended for higher transfection efficiencies than lipofectamine 2000 (48). These changes have significantly improved the yield of our protein purifications as shown in Figure 4G and allowed for the purification of recombinant human GNLY ( Figure 5). Our recombinant system also facilitates site-directed mutagenesis and has been successfully employed to mutate a key amino acid within the active site of Granzyme A (GzmA-S195A) (17). There were previous reports of GNLY expression in bacteria (1,49), yeast (32), and insect cells (33), but to the best of our knowledge, this is the first time that GNLY has been successfully purified in a mammalian expression system. As shown in Figure 5E, GNLY expressed with our method has greater antimicrobial activity compared to previous purifications. Similarly, as shown in Figures 4E, F, GzmA and GzmB purified with this updated protocol give a better biological response at inhibiting intracellular mycobacterial growth (GzmA) and suppressing extracellular bacterial growth (GzmB) than the previous purification protocol.
The construct that was used in our purification process uses an EK cleavage site at the N-terminus of the proteins. The overnight EK cleavage in hypotonic solution was believed to be required for optimal enzymatic activity (38). However, as shown in Figure 1D, most of the protein produced precipitates in this solution, suggesting that, at least for Granzyme A, salt concentration is key for proper protein folding. Previous studies hinted at this possibility (34), in which GzmA activity dropped by 66% when treated with hypotonic detergent washes. For GzmB, the activity drop was even greater (by 94%). Given the presence of granzymes in serum, we hypothesized that isotonic conditions would favor promotion of native protein structure. When comparing the old protocol to this new protocol, the role of hypotonicity in lowering activity is shown by decreased inhibition of intracellular mycobacteria as presented in Figure 4E.
Previous failed attempts at purifying active GNLY from a mammalian expression system could also have been due to the hypotonic conditions, and studies have also shown GNLY's antimicrobial activity is influenced by salt concentration (25). Thus, we recommend maintaining human serum physiological conditions during the purification process (154 mM NaCl and pH 7.4) and storing the final product in 50 mM HEPES buffer. Our proteins are concentrated to~700 mM and for most of our applications, the concentration needed is between 20 pM and 200 nM (1:3,500-35,000,000 dilution into final reagent).
The increase in biological activity for Gzms and GNLY ( Figures 4E, F, 5E) could be due to an increased purity of the final proteins, as well as the increased protein stability by use of this updated protocol. In fact, it is known that high protein concentration allows protein to remain more stable and in solution (50).
In conclusion, we have established an improved cytotoxic granular protein production process using a mammalian expression system that allowed us to purify large yields of Gzms and express human GNLY for the first time. This robust purification system allows the researcher to obtain enough protein material for in-depth studies to unravel unknown mechanisms involved in protection against infections and cancers, while also opening new doors for therapeutic applications. GNLY is considered as a potential alternative to antibiotics (51)(52)(53), while GzmB was found to be highly effective in limiting human tumor progression (28). The use of readily available equipment and reagents will allow researchers of all backgrounds to use these tools and significantly contribute to a better understanding of the function of these proteins and their translation to human medicine.

DATA AVAILABILITY STATEMENT
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

AUTHOR CONTRIBUTIONS
VR and OH designed and conducted experiments under the supervision of DH, MW, and SS. SB, DG, and PM provided logistical and technical support for the experiments represented in this article. VR, OH, DG, DH, and MW wrote and finalized this manuscript. All authors contributed to the article and approved the submitted version.