Fractionation of perforin and granzymes by immobilized metal affinity chromatography (IMAC)

JOURNAL OF IMMUNOLOGICAL METHODS ELSEVIER

Journal of Immunological

Methods

191 (1996) 11-20

Fractionation of perforin and granzymes by immobilized metal affinity chromatography ( IMAC) Ulrike Winkler *, Timothy M. Pickett, Dorothy Hudig Cell and Molecular

Biology Program.

Department

oj’Microhiology

Agriculture.

University of Nevada

Received 26 September

und School of Veterinary Rem.

199.5; revised 5 December

NV 89557.004,

Medicine,

School of Medicine and Collage of

USA

1995; accepted 7 December

1995

Abstract Cytotoxic lymphocytes and natural killer cells kill their targets by releasing pore-forming granules or by Fas ligand-Fas initiated death. The granules contain the pore-forming protein perforin, proteoglycan and multiple serine proteases termed granzymes. In this paper we describe two options for isolating perforin and granzymes. Both options separate the proteins by their ability to bind to immobilized metal affinity chromatography (IMAC) columns. The first option, with Cu2+ as the metal (CU’+-IMAC), separates both perforin and granzymes while the second, with Co2+ as the metal (Co2+-IMAC), separates only perforin. After Cu2+-IMAC perforin is > 20-fold enriched with excellent recovery of lytic activity. Only two proteins are substantial contaminants. After Cu 2+-IMAC , the perforin is dilute and requires concentration before additional steps of purification. The second option, with Co2+ as the metal (Co2+-IMAC), yields perforin that is concentrated in a sharp peak. The concentrated perform is immediately suitable for further purification. The first option, with Cu’+, isolates the granzymes while the second option, Co’+ -IMAC, does not. After isolation, the perforin lytic and granzyme activities are stable for weeks at 4°C an advantage to previous isolation methods for these proteins. The excellent recoveries of perforin and granzymes also indicate that these proteins are less than 4% and 15% of the total lymphocyte granule protein, respectively. Keywords:

Chromatography,

immobilized

metal affinity; Perforin;

Granzyme

1. Introduction

Granule exocytosis and Fas-initiated death are the major means by which cytotoxic lymphocytes kill other cells. The importance of perforin in the cy-

Abbreviations: CTL. cytotoxic T lymphocyte; NK, natural killer cells; -SBzl, thiobenzyl; SDS-PAGE, sodium dodecyl sulfate-polyacrylamide gel electrophoresis. * Corresponding author. OO22- 1759/96/$15.00 SSDI 0022.1759(95)00290-

0

1996 Elsevier Science B.V. All rights reserved 1

tolytic mechanism was recently established by reports in which mice lacking the perform gene showed greatly diminished T or NK cytotoxicity (Kagi et al., 1994; Walsh et al., 1994; Lowin et al., 1994). Perforin and granzymes (serine-dependent proteases) are found together in granules unique to cytotoxic lymphocytes (Ojcius et al., 1991; Peters et al., 1991). Better methods are needed to isolate both proteins. Perform has yet to be quantified to determine how many perform molecules are necessary to lyse a

12

(1. Win&r

et ui./Journal

oj’lmmunolo~icul

single target cell. Many investigators, including the authors, have been unable to get full length recombinant perforin to show substantial lytic activity, indicating that rigorously purified native perforin will be necessary to address fundamental questions. While over ten different granzyme genes have been cloned and the enzymes are predicted to vary greatly in their substrate specificity (Odake et al., 1991) only four granzymes have been purified. There are tryptases or trypsin-like enzymes (that cleave after the amino acids Lys and Arg), chymases or chymotrypsin-like enzymes (that cleave after amino acids Phe, Trp or Tyr) and enzymes that cleave after Asp and Met which are termed Asp-ase and Met-ase, respectively. One predicted chymase (human granzyme H) remains to be identified as a protein (encoded by the granzyme H gene), while the substrate specificity of four murine granzymes also predicted to be chymases (granzymes D, E, F, and G) remains to be experimentally determined. For perforin and granzyme isolation, we selected IMAC procedures because a previous report indicated that neutrophil granule serine proteases adhere to chelated copper (Torres et al., 1979). The binding of specific proteins to IMAC columns is essentially unpredictable, though the binding appears to be influenced by the number and relative positions of histidine residues (Kagedal, 1989). The high salt concentration (1 M NaCl) and betaine (10% w/v) in our buffers dissociate the granule proteins from the proteoglycan matrix. Here we report two options using either copper or cobalt metal affinity chromatography. Both yield perforin with excellent enrichment and depletion of granzymes. To isolate both perforin and granzyme copper metal affinity is the better option. To isolate only perforin or to continue purification of perforin, cobalt metal affinity is preferable because the perforin product is markedly more concentrated.

2. Materials

and methods

2.1. Granule preparation Cytolytic granules were obtained from rat RNK- 16 NK-like leukemia cells (Reynolds et al., 1981). The cells were grown as an ascites line in F344 weanling

Methods 191 (1996) II-20

rats (Harlan Sprague Dawley, Frederick, MD). The rats were primed with pristane (2,6.10,14- tetramethyl-pentadecane, Sigma Chemical Co., St. Louis, MO) and injected with 2 x 10’ tumor cells (i.p.). The ascites cells were washed with relaxation buffer (Borregard et al., 1983) that contained 130 mM KCI (Fisher Scientific, Pittsburgh, PA), 5 mM NaCl (Fisher Scientific), 1 mM ATP (Sigma, St. Louis, MO, #A-2383), 2 mM MgClz (J.T. Baker Chemical Co., Phillisburg, NJ), 10 mM piperazine-N.N’-bis(2-ethanesulfonic acid) (Pipes, Sigma) and 1.25 mM EGTA (Sigma), pH 6.8. The cells (at 1 X 10’ cells/ml) were disrupted using a nitrogen cavitation bomb (Parr Instrument Co., Moline, IL) pressurized to 450 psi for 25 min. The cell lysate was fractionated using a Percoll (Sigma) gradient (Borregard et al., 1983). The gradient was formed during centrifugation for 20 min at 45 000 X g of samples layered over 54% Percoll using a Beckman Ti50.2 rotor at 4°C in a Beckman L5-50 ultracentrifuge. The high density fraction, up to 1.068 g/cm3, was collected. To remove nuclei, the fraction was passed through a 3 pm Nucleopore (Millipore, Bedford, MA) filter (Henkart et al., 1984). To remove the Percoll, the filtered fraction was spun for 4 h at 145 000 X g. The granules above the Percoll pellet were collected. To disrupt the granule membranes, the samples were subjected to at least three freeze/thaw cycles after adding dry NaCl to make the final concentration 1 M NaCl (Borregard et al., 1983). The extracts were aliquoted and stored at - 20°C. Granule protein concentrations were determined by BCA assay (Pierce, Rockford. IL) using bovine serum albumin for calibration. 2.2. IMAC fractionation

of granule components

All columns were attached to a FPLC work station with LC-500 pumps (Pharmacia, Piscataway, NJ). The granule proteins were initially applied to a P6-DG (Bio-Rad Laboratories, Hercules, CA) size exclusion column in order to remove EGTA. In this process the granules are exchanged into the starting buffer for the IMAC column procedure: 20 mM Hepes (Fisher), 1 M NaCl (Fisher) and 10% betaine (Sigma), pH 7.5. After removal of EGTA, granule extracts were loaded onto an immobilized metal affinity column

il. Winklrr et al. / Journd

of Immunologicul

(IMAC) (4.6 x 50 mm prepacked, PerSeptive Biosystems, Cambridge, MA) that had previously been charged with CuCI, (Sigma) for option 1 or CoCl, (Sigma) for option 2. To get consistent results it was important to remove all the remaining metal and recharge the column completely after each run. The chelating buffer contained 1 M NaCl, 20 mM Hepes and 50 mM EDTA, pH 7.5. To remove the metal, 14 column volumes of chelating buffer were passed over the column. This wash was followed by lo-20 column volumes of distilled water before reapplying metal. The Cu*+ and Co*+ metal buffers consisted of 20 mM sodium acetate (Baker) and 0.1 M CuCl, or CoCl,, pH 3.0 and pH 6.0, respectively. The columns were charged with metal by applying 14 column volumes of the desired metal buffer followed by washing with lo-20 column volumes of distilled water. The column was next washed with starting buffer (20 mM Hepes (Fisher), 1 M NaCl (Fisher) and 10% betaine (Sigma), pH 7.5) and then loaded with the sample. Bound fractions were eluted with an elution buffer containing 20 mM Hepes (Fisher), 1 M NaCl (Fisher), 10% betaine (Sigma) and an imidazole gradient (Sigma Prod. #I-0250), pH 7.5. The concentration of imidazole ranged from 0.1 to 0.5 M. 2.3. Protease assays Protease activities were measured using colorimetric assays with peptide thiobenzyl (-SBzl) ester substrates. Upon cleavage, these substrates release a thiobenzyl group. Ellman’s reagent (Ellman, 1959), dithiobis-(2-nitrobenzoic acid), which absorbs at 412 nm, was used to detect the free thiol leaving group. Ellman’s reagent was 0.63 mM in the assay. The assay buffer contained 100 mM Hepes and 0.5 M NaCl, pH 7.5. The substrates were Suc-Phe-LeuPhe-SBzl (Bachem Biosciences, Philadelphia, PA), Boc-Ala-Ala-Asp-SBzl (ESP, Enzyme Systems Products, Dublin, CA), Boc-Ala-Ala-Met-SBzl (ESP) and Na-benzyloxycarbonyl-L-lysine thiobenzyl ester (Calbiochem, San Diego, CA, ‘BLT’) (Green and Shaw, 1979) to detect chymase, Asp-ase, Met-ase and tryptase activities, respectively. The assay concentrations for the substrates were 30, 90, 90 and 150 PM for chymase, Asp-ase, Met-ase and tryptase activities, respectively. The assays were performed

Methods 191 (1996) I I-20

13

in 96-well flat bottom plates. A Molecular Devices Corp. (Menlo Park, CA) Thermomax microplate reader was used in the kinetics mode at 412 nm to measure the rates of substrate hydrolysis. 2.4. Perforin cytolytic activity Cytolytic activity was determined by the hemoglobin released from lysed rabbit red blood cells (RBC) (Henkart, 1993). Dilutions of the granule extracts were incubated with 0.5% (v/v) RBC at room temperature for 20 min in a volume of 0.2 ml in round bottom microtiter plates (Falcon 3910, Becton Dickinson Labware, Lincoln Park, NJ). The assay buffer contained 10 mM Hepes, 0.15 M NaCl, and 10 pug/ml bovine serum albumin (Sigma A4503), pH 7.5. The RBC were diluted in this buffer with enough CaCl, (Baker) so that the final calcium concentration was 1 mM during incubation (Henkart et al., 1984). The reaction was halted by acidification with 25 ~1 of 0.15 M pH 6.0 buffer, 2-[N-morpholinolethane-sulfonic acid (ME& Sigma M-8250) containing 0.15 M NaCl (Hudig et al., 1988). The microtiter plates were spun at 1500 X g for 10 min and the cell-free supernatants transferred to a second microtiter plate. The hemoglobin released into the supernatant was detected with the microplate reader at a wavelength of 412 nm. The percent specific lysis was calculated as [(% experimental hemolysis % spontaneous hemolysis)/(% maximal hemolysis % spontaneous hemolysis)] X 100. Addition of 0.01% saponin (Sigma) to RBC produced maximal hemolysis. 2.5. Proteoglycan

assay.

We followed the instructions for a calorimetric assay for sulfated glycosaminoglycans which has been described previously (Famdale et al., 1986). The color reagent consisted of 4.6 PM 1,9-dimethylmethylene blue (Aldrich prod. #34,108-g), 40 mM NaCl (Fisher) and 40 mM glycine (Fisher), pH 3.0. To assay for proteoglycan, a 25 ~1 sample was mixed with 200 ~1 of color reagent and the mixture was read immediately at 525 nm using a Thermomax microplate reader. Chondroitin sulfate A (Sigma) was used to create a calibration curve.

14

A

CL?+-IMAC

120

2

- 140

010

- 120

100 c 0.08 I I

/-

I1

P 8

s 0.06 2 a, n m * 0.04 d .!z n 5

-100

A I

-80

- 60

? E a 0 E % 2

-.40

2

0.02

20

0

- 20

0

0.00 0

70

30

20

40

LO

50

ml

Co*+-IMAC

B

180

100

~ 160 I-0.4

80

140 120

3

100 80

20

0 40

50

60

70

ml Fig. 1. IMAC separation of perform and granzymes. A: option I. Copper as the immobilized metal. Granzyme activity eluted in the second absorption peak at 280 nm. Lytic activity followed the granzymes; however, there was little absorption at 280 nm. The chymase activity is illustrated as milliOD,, Z of substrate hydrolyzed per minute per 20 /LI of each I ml fraction. The other granzyme activities followed the same elution pattern. B: option 2. Cobalt as the immobilized metal. The lytic peak eluted in the imidazole gradient and coincided with the second absorption peak at 280 nm. The chymase activity is illustrated as milliOD I,2 of substrate hydrolyzed per minute per 25 ~1 of each 1 ml fraction.

U. Wink&

et ul./Journul

of Immunolo~icul

2.6. Protein analysis and characterization. 2.6. I. Protein assay A BCA (Pierce, Rockford, IL) assay was used to determine protein concentration. BSA was used as a standard to calibrate the assay. Microcon- 10 concentrators (Amicon, Beverly, MA) were used to concentrate and wash away imidazole. 2.6.2. SDS-PAGE and Western-blots For SDS-PAGE and Western-blots, samples were electrophoresed in precast 12% single percentage gels (Bio-Rad Laboratories, Hercules, CA) according to Laemmli (1970). Protein bands were visualized using a silver staining procedure (Bollag and Edelstein, 1991). For Western blots the proteins were transferred to nitrocellulose paper (Bollag and Edelstein, 1991), and perforin was detected using rabbit anti-rat perform antibodies. Antibodies to perforin were made in rabbits using synthetic peptide (Multiple Peptide Systems, San Diego, CA) representing amino acids 464-475 of rat perforin (Ishikawa et al., 1989). The antibodies were affinity purified using a BSA-peptide column. The Western blots were developed using a Vectastain ABC horseradish peroxidase kit (Vector Laboratories, Burlingame, CA) in conjunction with LumiGLO (Kirkegaard and Perry Laboratories, Gaithersburg, MD) a chemiluminescent substrate. Pictures were obtained using a Vmax scanner (Envisions Solutions Technology, Burlingame. CA) and Adobe Photoshop (Adobe Systems, Mountain Table 1 Granzyme

recovery

View, CA. ver. 2.5) capturing software in conjunction with Core1 Draw (Corel. Ottawa. Ontario, ver. 4.0).

3. Results 3.1. Separation

of perforin and granzymes

3.1.1. Option I to isolate both pe$orin and granzymes: Cu2 ‘-IMAC With immobilized copper IMAC perforin, granzymes and proteoglycan were separated into three fractions. The OD,,, abso~tion profile from this column shows two distinct peaks (Fig. 1A). The first peak was not retained by this column. The second peak eluted with 35 mM imidazole and contained the granzymes. Fig. IA depicts the chymase activity profile. The other granzymes follow the same elution pattern (Table 1). This granzyme peak represented about 14% of the total protein. Perforin eluted in a third peak starting at 60 mM imidazole (Fig. 1A). Near complete recovery of perforin lytic activity is possible (Table 2). In four different experiments there was subst~tial variation in lytic recovery (from 11.6 to 167% of the initial lytic units). The critical loss of lytic activity (when it occurred) happened during the PB-DG buffer exchange which removes EGTA from the granule extracts. The longer the sample was held without chelation of divalent cations, the less lytic activity was recovered. Immediate loading of the IMAC column after P6-DG

at each step of protein purification

Metal on IMAC

steps

Asp-ase

iW+

(I ) Granule extract (2) P6-DG exchange (3A) IMAC granzymes f3B) IMAC perform (1) Granule extract (2)P6-DG exchange (3A) IMAC granzyme (3B) IMAC perform

loo 87.7 38.9 0 100 64.5 b 80.3 0

CO”+

15

Methods 191 (1996) II-20

a (%I

Chymase 100 64 13.4 0.74 loo 64 19.4 0.56

a (%f

Met-ase a (%/c)

Tryptase

100 31.3 27.4 0.74 100 13.5 b 25.3 0

100 42.0 36.8 0.1 100 35 b 47.6 0.37

a (%I

a The recovery is based on the starting samples. For the Cuz+ IMAC, 1.6 mg of granule extract was used which contained 5650, 23 100, 6800 and 20 150 mOD rng- ’ of Asp-ase, chymase, Met-ase and tryptase activities respectively. For the Cu’+ IMAC, 3.6 mg of granule extract was used which contained 4300, 26@00, 10850 and 19050 mOD mg- ’ of Asp-ase, chymase, Met-ase and tryptase activities respectively. b Granzyme losses were increased by storage of the material without EGTA.

buffer exchange is crucial. (An example of lytic loss is indicated for the Co”+-IMAC. See fifth row of Table 2.1 In contrast to the variable recovery of lytic activity, perform protein recovery was consistently 23-fold or better (last column, Table 2). The perforin contained some residual granzyme activity (Table 1). Perforin antigen, as determined by slot immunoblots (not illustrated), coincided with only those fractions that had lytic activity. The perforin contained only traces of granzyme activity (Table 1). Furthermore, perforin is less than 4% of the total granule protein. 3.1.2. Option 2: Co’ + -IMAC COpurify only perforin With immobilized cobalt there were two fractions. One contained both granzymes and proteoglycan and the other contained perforin. Using cobalt as the immobilized metal increased the separation between the granzymes and perforin (Fig. Is>. Fig. IB shows a typical profile of a separation of rat granule proteins using cobalt on the column. Two peaks absorbing at 280 nm are seen, the first peak was not retained by the column and the second peak eluted at 75 mM imidazole. The majority of the chymase activity (95.5% of the activity loaded onto the column) was not retained on this column. The other enzyme activities that were assayed (Asp-ase, Metase and tryptase) followed a similar pattern and are summarized in Table 1. The perforin purified by cobalt coincides with the second absorption peak. In contrast to the broad elution of perforin with imidazole from a copper column, when cobalt was used. perforin eluted in a small, sharp peak of protein. The recovery of lytic activity as determined by lytic units was 24.9% for the preparation illustrated. In four Co”+-IMAC experiments there was variation in lytic

Table 2 Recovery

recovery (11.8. 16.1, 24.9 and 42.7% of the start). The specific enrichment of lytic units per mg of protein was 1.9-, 2.8, 3.7- and 6.4-fold, respectively. A major variable that affected lytic recovery occurred after the P6-DG removal of EGTA from the starting granules used for both columns. In contrast, the perforin enrichment, based on the amount of protein in the perforin peak, was quite reproducible. usually 2 1 to 25-fold respectively (last column, Table 2). The protein enrichment was similar to that for the Cu’+-IMAC. However, with Co”+-IMAC there was always some (- 20%) loss of lytic activity associated with the IMAC step as well as the P6-DG step. 3.2. Proteoglycan

depletion

Both Co’+ or Cu’+ -1MAC columns removed most of the proteoglycan from the bound perforin or granzynies. Fig. 2A shows the proteoglycan profile when using copper. Fig. 2B shows the proteoglycan profile of a cobalt IMAC column run. With either metal. the proteoglycan does not bind to the column. Proteoglycan depletion facilitates further purifications that require low salt conditions, for example ion exchange chromatography. 3.3. Prtiteins in the perforin and grarl~y~e tractions Fig. 3 shows a silver stain of the granzyme and lytic fractions from the cobalt and copper IMAC procedures. Lane 1 contains the unbound proteoglycan/protease fraction from cobalt IMAC. Lane 2 represents the perforin lytic peak from the cobalt separation. There was marked enrichment of a - 66 kDa protein that may represent perforin, with deple-

of perforin lytic activity after each step of purification Total LU at each step

Metal bound to IMAC

steps

C””

f 1) Granuleextract (2) I%-DG exchange (3) IMAC (1) Granule extract (2) P6-DG exchange (3) IMAC

C02'

’ This granule

preparation

had unusually

16266 25 923 16778 23 592 5 188 5880 high lytic activity.

LU recovered from IMAC (%)

Protein recovered (mg) 5.88

Specific activity LU/mg

(LU/mg)

Enrichment

Perforin enrichment (protein)

22.8-fold

2%fold

6.3-fold

25.fold

2766

103 0.255 0.552

65 798 42 739 a

0.022.

‘67 212

24.9

ofImmunnlogicalMet~~ods 191

U. WinkleretuI./Jourml

A

17

(1996) II-20

CL?+-IMAC 0.10

0.6 0.08

1 I , 0.06 z E .$ ?! 0.04 ; Ei G E

0.5 + 0.4 2 " H 0.3 p a 0.2 a

:

0.02 0.1 0.00 0

5

10

15

20

25

30

35

00

40

ml

Co’+-IMAC

B 1.0

/ /

/---

-.-

0.5 0.12

/I

1’

0.8

I

0.6

z N 8

04

0.4

/I

I

/I /I

I I 0.3 s v E a, P & 0.2 8

I/ // //

: :: :: :: ::

/I

!I P E

0.10 ( 0.08 ; z 3 0.06 8 a, b Q 0.04 2

0.1 0.02

.? 10

20

30

40

50

0.0

0.00

60

ml

Fig. 2. Proteoglycan removal. of the proteoglycan-dimethylene granule extract and purified accounting for the interassay @g/ml sample in the copper

A: option

1. Copper as the immobilized metal. B: option 2. Cobalt as the immobilized metal. The absorbance blue complex at 525 nm is depicted. The raw data are presented because the calibration curves of the chondroitin A were nonparallel. The absorbance also decreased within minutes after the complexes form, variability. For example, the chondroitin sulfate A standard gave an absorbance reading of 0.37 for a 50 run and an absorbance reading of 0.18 in the cobalt run.

U. Widlrr

18

I

2

3

94 kD -

94 kD

66

66 kD

45

45 kD

_?l

31 kD

20

14

et ul./Journul

oflmmunoloyicul

4

14kD

visualI, the separaseparaIMAC from a

tion of other proteins (e.g.. - 31 kDa and - 18 kDa) from the lytic fraction. Lane 3 shows the granzymes from a copper IMAC separation and lane 4 shows the lytic component from this separation. (The proteins that were unbound by copper IMAC are not illustrated.) The multiple proteins in the granzyme fraction correspond to more than seven serine proteases identified by their reactivity with a biotinylated, general, mechanism-based irreversible isocoumarin serine protease inhibitor, Bi-Aca-AcaIC-OMe. Bi-Aca-Aca-IC-OMe inhibits all four granzyme activities so that the biotinylated proteases can be detected on protein blots (Winkler et al., manuscript submitted). What is striking is that the Cu*+ 1 tic peak contains fewer proteins than does the Co’ lytic peak. The predominant bands are at 66 kDa (the M, of perforin) and slightly higher. 3.4. Three additional

191

f 1996)

I I-LO

eluted at 60 mM imidazole (not illustrated), later than when copper IMAC was used as an initial separation. 3.4.2. Chymase losses can be reduced Chymase activity is the most unstable when EGTA is omitted from buffers. To enhance protease recovery, copper IMAC may be run in the presence of 0.1 mM EGTA, but only with small volumes of granule extracts ( < 2 ml> that contain 1.2 mM EGTA. We found that even at low concentrations of EGTA lytic activity will not bind to the cobalt IMAC column.

20 kD

Fig. 3. SDS-PAGE analyses. The reduced proteins where ized by silver staining. The samples are: lane proteoglycan/granzyme component from a cobalt IMAC tion; lane 2, the lytic component from the cobalt IMAC tion; lane 3, the granzyme component from a copper separation; and lane 4. concentrated perforin lytic fractions copper IMAC separation.

Mrthds

obseruations

3.4.1. Granzymes can be recovered by sequential cobalt and copper IMAC There was no need to exchange the buffer prior to the second IMAC column. Curiously, the granzymes

3.4.3. Fractionated perforin is stable The lytic activity using either method is relatively (- 90%) stable for several weeks at 4°C and it is also stable if frozen at - 20°C. Addition of EGTA to 0.1 mM in the IMAC fractions further increases stability. The lytic activity of perforin upon further purification, by ion exchange or hydrophobic interaction chromatography, decreases rapidly over several days. Thus it is recommended to store perforin after IMAC fractionation and perform further separations immediately before use. Granzyme activities are partially depleted after freezing and thawing.

4. Discussion In order to study the granzyme involvement in perforin-mediated lysis it was necessary to develop a method for fast separation of perforin from the granzymes. The separation method must produce active perforin with high lytic recovery (as determined by lytic units) and must simultaneously provide good enrichment of perforin from other granule proteins. Another requirement for this method is that the granzymes be rigorously separated from perforin, to optimize studies of granzyme control of perform lysis. Furthermore. proteoglycan should be depleted as it frequently carries cationic proteins (such as granzymes) with it. Perforin and granzymes form a complex with proteoglycan (Masson et al., 1990) at neutral pH and in physiological salt concentrations which can impact subsequent methods of protein separation that utilize less than 0.5 M salt concentrations. Previously published methods for the purification of perforin used MonoQ anion exchange chro-

Cr. Winkler et al./ Journal of Immunological

matography or phenyl-Superose hydrophobic interaction chromatography. MonoQ as the initial chromatography step (Young et al., 1986; Podack et al., 1985) yielded good enrichment of perforin with low recovery of lytic activity. Furthermore. perform prepared by MonoQ has substantial chymase activity associated with it (Hudig et al., unpublished data). Phenyl-Superose hydrophobic interaction chromatography yields perform with good lytic activity (38% when used as an initial step) and good enrichment; however, the lytic activity is not stable. Major advantages for using IMAC as an initial step in perforin and granzyme purification are that this method is very fast and that it is suitable for large amounts of granule extracts. The actual time required to separate a standard sample of 5 mg granule extract by IMAC can be as little as 60 min. Even faster separations are possible because we were not using the maximal flow rates for the IMAC columns. The most time-consuming part of this separation is removal of EGTA. The EGTA in the granule extract is needed to retain lytic and chymase activities. The P6-DG size exclusion column is faster than dialysis. It is critical to minimize the time the granule proteins are without EGTA. The granzymes, especially the chymases, spontaneously lose activity in buffers without EGTA. Lytic activity is also reduced without EGTA. Sephadex matrices are unsuitable for the initial buffer exchange, because even greater amounts of chymase activity are lost. Perforin recovery, as assessed by lytic units, is excellent when using copper and almost as good when using cobalt IMAC (Table 2). The copper separation has the advantage that the specific enrichment per mg of protein is usually greater than that achieved by the cobalt method. The copper procedure gives perform activity that elutes from the column in a large volume, which presents a disadvantage. When these samples (with low protein concentration) are concentrated prior to other separation procedures, lytic activity is lost. The advantage of copper IMAC (for rat perforin preparation) is that there are few other co-purifying proteins. Cobalt IMAC separation has other advantages, including better separation between granzymes and perforin. This method also serves to concentrate perform into a small sample volume. The small volume permits easier transitions to other chromatographic methods.

Methods 191 (1996) I I-20

19

For example, samples can be diluted to reduce salt concentrations immediately prior to ion exchange chromatography. Either copper or cobalt method is suitable for removal of the proteoglycan component from perform. Granzymes behave as a single species in the copper IMAC method. The copper affinity matrix offers extreme enrichment of all the granzyme activities, with the exception of the chymases which appear to be unstable. Both granzyme activities and perforin lytic activity are stable in the high salt buffers with low concentrations of imidazole. The imidazole protects these activities from losses that appear to be associated with very low concentrations of calcium. In summary, these IMAC methods isolate the granzyme and lytic components. They are a good starting point for further separations of these proteins. Both granzymes and perform are depleted of their proteoglycan component. These methods are simple, effective, fast and highly reproducible. The cobalt IMAC procedure has recently been modified for the preparation of human perforin from small numbers of lymphocytes. High salt extracts of granule fractions, prepared by several centrifugation steps, replaced organelle separation by Percoll gradients (personal communication, C.J. Froelich, M.D., Northwestern University). Acknowledgements We thank Susan L. Woodard, Ph.D., University of Nevada, Reno, and Christopher J. Froelich, M.D., Northwestern University, Chicago, for their suggestions for the manuscript. This work was supported in part by NIH grants ROl CA38942 and T32 CA09563. References Bollng, D.M. and Edelstein, S.J. (1991) Protein methods. WileyLiss, New York. Borregard, N., Heiple, J.M., Simons, E.R. and Clark, R.A. (1983) Subcellular localization of the beta-cytochrome component of the human neutrophil microbicidal oxidase: translocation during activation. J. Cell Biol. 97, 52. Ellman. G.L, (1959) Tissue sulfhydryl groups. Arch. Biochem. Biophys. 82, 70. Farndale, R.W., Buttle, D.J. and Barrett, A.J. (1986) Improved quantitation and discrimination of sulphated glycosaminogly-

20

IJ. Winkler er ul./Jourd

oj’immunological

cans by use of dimethylmethylene blue. Biochim. Biophys. Acta 883, 173. Green, G.D.J. and Shaw, E. (1979) ~jobenzyl benzyloxycarbonyl-L-lysinate, substrate for the sensitive calorimetric assay for trypsin-like enzymes. Anal. Biochem. 93, 223. Henkart, P. (1993) Measurement of cytolysin hemolytic activity. in: M.V. Sitkovsky and P.A. Henkart @ids.), Cytotoxic Cells: Recognition, Effector Function, Generation and Methods. Birkhauser, Boston, MA, p. 484. Henkart, P.A., Millard, P.J., Reynolds, C.W. and Henkart, M.P. (1984) Cytolytic activity of purified cytoplasmic granules from cytotoxic rat large granular lymph~yte tumors. J. Exp. Med. 160. 75. Hudig, D.. Callewaert, D.M., Redelman, I)., Allison, N.J., Krump, M. and Tardieu, B. (1988) Lysis by RNK-16 cytotoxic Iymphocyte granules: Rate assays and conditions to study control of cytolysis. J. Immunol. Methods 115,169. lshikawa, H., Shinkai, Y., Yagita, H., Yue, CC., Henkart, P.A., Sawada, S., Young, H.A., Reynolds, C.W. and Okumura, K. (1989) Molecular cloning of rat cytolysin. J. lmmunol. 143, 3069. Kagedal, L. (1989) Immobilized metal ion affinity chromatography. In: J.-C. Janson and L. Ryden @ds.), Protein purification. Principles, high resolution methods, and application. VCH Publishers (UK), Cambridge, Kagi, 0.. Ledermann, B., Burki, K., Seiler, P., Odermatt, B., Olsen, K.J., Podack, E.R., Zinkernagel, R.M. and Hengartner, H. (1994) Cytotoxicity mediated by T cells and natural killer cells is greatly impaired in perforin-deficient mice. Nature 369, 31. Laemmli, U.K. (1970) Cleavage of structural proteins during the assembly of the head of bacteriophage T4. Nature 227, 680. Lowin, B., Beermann, F., Schmidt, A. and Tschopp, J. (1994) A nul1 mu~tioa in the perforin gene impairs cytolytic T lymphocyte- and natural killer cell-mediated cytotoxicity. Proc. Natl. Acad. Sci. USA 91, 11571.

Methods 191 (19961 I I-20

Masson, D., Peters, P.J., Geuze, H.J., Borst. J. and Tschopp, J. (1990) interaction of chondroitin sulfate with perforin and granzymes of cytolytic T-cells is dependent on pH. Biochemistry 29, I 1229. Odake, S., Kam, C., Narasimhan, L., Poe. M.. Blake, J.T., Krahenbuhl, 0.. Tschopp, J. and Powers, J.C. (1991) Human and murine cytotoxic T lymphocyte se&e proteases: Subsite mapping with peptide thioester substrates and inhibition of enzyme activity and cytolysis by isocoumarins. Biochemistry 30, 2217. Ojcius, D.M., Zheng, L.M., Sphicas, E.C., Zychlinsky, A. and Young, J.D. (1991) Subcellular localization of perforin and serine esterase in iymphokine-activated killer cells and cytotoxic T cells by immunogold labeling. J. lmmunol. 146, 4427. Peters, P.J., Borst, 1.. Oorschot, V.. Fukuda, M., Krahenbubl, 0.. Tschopp, J., Slot, J.W. and Geuze, H.J. (1991) Cytotoxic T lymph~yte granules are secretory lysosomes, containing both perforin and granzymes. J. Exp. Med. 173, 1099. Podack, E.R., Young, J.D. and Cohn, Z.A. (1985) Isolation and biochemical and functional charactedzation of perforin 1 from cytolytic T-cell granules. Proc. Natl. Acad. Sci. USA 82, 8629. Reynolds, C.W., Timonen, T. and Herberman, R.B. (1981) Natural killer (NK) cell activity in the rat I. Isolation and characterization of the effector cells. J. Immunol. 127, 282. Torres, A.R., Peterson, E.A., Evans, W.H., Mage, M.G. and Wilson, S.M. (i979) Fractionation of granufe proteins of grauulccytes by copper chelate chromatography. Biochim. Biophys. Acta 576, 385. Walsh, C.M., Matloubian, M., Liu. CC., Ueda. R., Kurabara, C.G., Christensen, J.L., Huang, M.T., Young, J.D., Ahmed, R. and Clark, W.R. (1994) Immune function in mice lacking the perforin gene. Proc. Natl. Acad. Sci. USA 91, 10854. Young, J.D., Hengartuer, H., Podack, E.R. and Cohn, Z.A. (1986) P~~~cation and ch~acte~zation of a cytolytic ~~-fo~~ng protein from granules of cloned l~ph~ytes with natural killer activity. Cell 44, 849.