Stabilization and Functional Studies of High-Molecular-Weight Murine Lymphotoxins

High levels of lymphotoxin-like activity (LT) were found in supernatants from secondarily stimulated immune mouse splenocytes activated with concanavalin A (Con A) in vitro. Sple-nccytes obtained from C57B1/6 mice immune to the P815 mastocytoma were restimulated in vitro with mitomycin C-treated P815 cells, and then stimulated with Con A. High levels of unstable LT activity are rapidly (2-4 hr) released by these lectin-stimulated splenocytes. The introduction of a crosslinking agent, glutaraldehyde, was found to stabilize this LT activity and allowed us to perform more defined biochemical studies and to examine the functional activities of the LT classes. The lytic activity in these supernatants resided in the high-mo-lecular-weight classes, termed Complex (Cx > 200,000 daltons) and alpha-heavy (an, 130,000- 160,000 daltons). It was found that the Cx and (in LT classes from the secondarily stimulated immune splenocytes cause lysis of allogeneic target cells, P815 and EL-4, in a 16-hr “Se- methionine release assay, and in some cases, this lysis was specific for the sensitizing target cell.


INTRODUCTION
Lymphokines are a family of biologically active molecules released by immune and nonimmune lymphocytes, when they are stimulated in vitro by antigens or lectins. One lymphokine, lymphotoxin (LT), has been shown to inhibit cell growth and cause target cell lysis in vitro (1,2). Murine lymphocytes release various molecular weight classes of LT that can be resolved by molecular sieving, termed Complex (Cx, >200,000 daltons), alpha-heavy ((mu 130,000-160,000 daltons), alpha (a, 70,000-90,000 daltons), beta (8,35,000-50,000 daltons), and gamma (y, 15,000-20,000 daltons), (3,4). These same molecular weight classes have been observed in supernatants from activated lymphocytes from many animal species that have been studied (5).
Physical-chemical and immunological studies have demonstrated that the Cx and (YH classes are composed of the LT forms (a, 8, and y), indicating that these classes are interrelated. Under conditions of high-ionic-strength buffers, Cx and &, from several animal species dissociate into toxins similar to the lower-molecularweight forms (3,5,6). In the case of human, murine, and guinea pig LT, antisera have been raised against a smaller-molecular-weight LT class which also neutralizes the larger-molecular-weight forms (3,5,7). It has been proposed that the small LT forms may be either subunits or proteolytic fragments of the larger molecules. Recent evidence indicates Cx and (Yn LT molecules, but not the smaller forms, may be associated with nonclassical antigen-binding receptors. Human Cx and an classes are neutralized by antisera to the Fab region of human IgG molecules, while antisera to the constant region of IgG or IgM heavy chains did not cause significant inhibition (6,8,9). In addition, lytic activity in supernatants from both human and murine alloimmune lymphocytes can be specifically adsorbed by the allogeneic cell line used for stimulation (10,11). Also, it has recently been reported that LT released by lymphocytes from alloimmunized mice causes the specific lysis in vitro of the tumor cell line used for immunization ( 11). The report of antigenspecific cell-lytic forms of murine LT potentially originating from cytotoxic T lymphocytes was interesting; however, lysis of the tumor cells by the LT-containing supernatants was low, rarely exceeding 20% even after being concentrated 5 to 20 times. In addition, these forms were difficult to study because of their extreme instability.
We have developed a new in vitro method to produce highly toxic LT forms from alloimmune lymphoid cells. We are able to stabilize the lytic activity with the crosslinking agent glutaraldehyde. These advances have allowed preliminary studies of both the biochemical and functional activities of individual molecular weight classes of murine LT.
Immunization protocol and harvesting murine spleen cells. The mice used were 8-to 12-week old C57B1/6 (H-zb) (ARS, Madison, Wise.). Animals were immunized by intraperitoneal injection of 10' P815 mastocytoma. After 21 days, the spleens were aseptically removed, and a single-cell suspension was prepared. Spleen cells from nonimmune animals were routinely placed in 32-0~ bottles at a density of 5 X lo6 cells/ml for 1 to 2 hr at 37°C in medium containing 10% FCS, to remove glass-adherent cells.
In vitro restimulation of immune splenocytes. Spleen cells from immune animals were restimulated in vitro by P815 cells which had been treated with mitomytin C (Sigma Chemical Co. St. Louis, MO) (50 pg/lO' cells) for 1 hr at 37°C and washed three times with RPMI. These cells were cultured at 5 X lo6 cells/ml with a ratio of 25: 1 lymphocyte to tumor cells for 4 days (13). At the end of the 4 days, the cells were harvested by centrifugation at 300g for 5 min and employed in these studies.
Production of cell-lytic culture supernatants.
Production of nonimmune LTcontaining supernatants has been described previously (14). Briefly, lo8 nonadherent normal murine splenic lymphoid cells were placed on phytohemagglutinin P (PHA-P, Difco, Detroit, Mich.)-coated L-929 cells (4 X 106) in RPMI-10% at DEVLIN, YAMAMOTO, AND GRANGER a density of 2.5 X lo6 cells/ml. These cultures were then incubated 6-8 hr at 37'C at which time the culture supernatants were collected. TO generate immune LT, splenic lymphoid cells from a 4-day culture were resuspended at a density of 2 X lo6 viable cells/ml and stimulated with PHA-P or concanavalin A (Con A, Sigma), 20 pg/ml in RPM1 ,164O supplemented to 1% FCS for 2-24 hr at 37°C. At the end of each of the incubation periods, the supernatant was collected and cells were removed by centrifugation at 1OOOg for 10 min.
Treatment of LT-containing supernatant with glutaraldehyde. LT containing supernatants were treated with 25% glutaraldehyde (Sigma) so that the final concentration of glutaraldehyde was 0.025% at 25°C for 1 hr.
Lymphotoxin assay. Two types of assay employing the L cell were used. Supernatants were either tested on L cells to determine the number of units per milliliter of LT activity, or a fixed amount from each column fraction was assayed on L cells, and the number of viable cells remaining after 24 hr was used to determine where toxicity was eluting from molecular sieving columns. These techniques are described in detail elsewhere (6,15).
Physical-chemical separation of LT. The methods employed for fractionation of LT activity in supernatants from activated murine lymphocytes were identical to those employed for separated human LT molecules, all of which have been previously published (3,5). All procedures were performed at 4°C. Briefly, 2-ml samples of 25X concentrated murine LT supernatant containing from 500 to 5000 units of LT activity were chromatographed on molecular sieving columns containing Ultrogel AcA 44 (LKB, Bromma, Sweden). One of the columns used was 90 X 2.4 cm, from which we collected 6.5-ml fractions; the other column was 60 X 2.4 cm and 4.7 ml fractions were collected. Aliquots of the eluted fractions of between 0.05 and 0.2 ml were assayed on L-929 cells for LT activity. The molecular weight markers are described in Fig. 1.
75Se-Met release assay. P8 15 and EL-4 cell lines used as targets in these studies were radiolabeled using a modification of the procedure of Brooks et al. ( 16). Approximately 4 X lo6 cells were incubated at 37°C overnight in 10 ml of RPMI-10% containing 20 &i of 75Se-methionine ('Se-Met) (Amersham, Arlington Heights, Ill.). Cells were then washed four times (800g for 6 min) with RPMI-3%. Cells were resuspended in RPMI-10% and lo4 cells in 0.02 ml added to each well of a round-bottom microcytotoxicity plate (Flow Laboratories, Inglewood, Calif.). All wells contained a total volume of 0.22 ml, consisting of either 3% sodium dodecyl sulfate to determine total releasable counts, RPMI-l%, to determine spontaneous release, LT containing media, or effector cells from in vitro cultures. A 4-hr assay was used to test cell-mediated lysis, and a 16-hr assay was used for LTmediated lysis. Target cell destruction was determined with the following formula: % lysis = experimental release -spontaneous release total release -spontaneous release .

Stabilization of Murine LT by Glutaraldehyde
The instability of murine LT led to a series of attempts to stabilize the lytic activity. Numerous methods were tried, but we were only successful with the cross-linking reagent glutaraldehyde. The level of glutaraldehyde used in this work was determined from a dose curve of the effect of various concentrations of glutaraldehyde on LT activity. It was found that 0.025% glutaraldehyde resulted in retention of lytic activity for at least three weeks at 4°C. To remove the toxicity due to glutaraldehyde, the treated supernatants were either dialyzed or chromatographed over an AcA 44 molecular sieving column.
Supernatants containing nonimmune murine LT in RPMI-3% were split into two samples. Glutaraldehyde was added to one sample, and the other was left untreated. Both samples were then concentrated and chromatographed over an AcA 44 molecular sieving column, and assayed for toxicity on L cells (Fig. 1A). The treated sample shows more toxicity in the column profile. This is because the toxicity in the untreated sample has been decaying in the time between supernatant collection and assay of the column fraction. If a sample of supernatant is assayed immediately, it is just as toxic as a glutaraldehyde-treated sample. The major peak of toxicity in both samples eluted in the 130,000-to 160,000 molecular-weight range, which is characteristic of (Y u. The fractions were stored at 4°C for 2 weeks and reassayed (Fig. 1B). All lytic activity in the untreated sample was lost during storage, while the treated supernatant retained toxicity. Control medium containing 3% FCS and glutaraldehyde, when treated in an identical manner, was not toxic (Fig. IA). Treatment with glutaraldehyde did not detectably alter the molecular weight of these supernatant LT forms. Glutaraldehyde is known to form covalent crosslinks between lysine residues in proteins ( 17). However, even in the presence of 3% FCS, intermolecular crosslinking did not seem to occur, as the peak of toxicity eluted in the same fraction in both the treated and nontreated sample (Fig. 1A).
Glutaraldehyde treatment did not make RPMI-3% toxic or detectably change the molecular weight of the toxic molecules in these supernatants. This indicates that preexisting LT is being stabilized rather than glutaraldehyde forming new toxins. Additionally, glutaraldehyde does not increase the toxicity of supernatants (data not shown).

Glutaraldehyde Reacts Directly with LT Molecules
Glutaraldehyde stabilization of cell-lytic activity appears to be a direct action on the LT molecules, rather than the inactivation of a separate degradative enzyme. To demonstrate this, we glutaraldehyde-treated a supernatant containing 100 units/ ml of nonimmune LT activity. FCS was added to a final concentration of lo%, and the sample was dialyzed overnight to remove remaining reactive glutaraldehyde. A sample of this treated supernatant was then mixed with an equal volume of fresh untreated supernatant (which also had an LT activity of 100 units/ml). After 48 hr, the treated supernatant retained its original activity, while the activity in the untreated supernatant had fallen to ~10 units/ml. The mixture of treated and fresh untreated LT had slightly more than half the original activity after 48 hr. Thus it seems that the activity due to the untreated LT in the mixture was lost, but the activity from the treated LT remained. So it would appear that the untreated supernatant did not contain a degradative enzyme responsible for loss of activity. This experiment was repeated three times with similar results.

Kinetics of Immune LT Production
The production system employed in this study generates LT in half the time required by procedures previously used in this laboratory (14). Figure 2 shows LT activity in the supernatant reaching a maximum in 8-10 hr, when alloimmune splenocytes are stimulated on a PHA-coated monolayer of L cells, as was done previously. However, we found that immune lymphocytes, restimulated in vitro, released peak levels of LT in just 3-4 hr, when placed on a lectin-coated monolayer of L cells. In an effort to avoid contamination of supernatants with material from lysed L cells, studies were performed to examine the action of lectin alone. We found Con A stimulation led to the release of maxmal levels of LT within 3-4 hr in the absence of L-929 cells (Fig. 2). PHA alone would also stimulate rapid release, but not as reproducibly as Con A. In agreement with prevous work (14), Con A would not stimulate rapid release of LT from alloimmune splenocytes which had not been restimulated in vitro by the immunizing cell line.
Molecular weight classes of immune LT. A 60 X 2.4-cm column was used for molecular weight characterization of immune LT. This column allowed us to work with these supernatants quickly, yet still gave adequate separation of the molecular weight markers. The lytic activity from these supernatants chromatographed as a FIG. 2. Immune lymphocytes were cultured with P815 cells in vitro and stimulated with Con A, 10 rg/ml, at a density of 2.5 x 106/ml, or nonadherent nonimmune C57Bl/6 murine spleen cells were incubated at a 25:l ratio with PHA-coated L-929 cells. After various time intervals, the supernatants were harvested, cleared of cells by centrifugation, and frozen at -70°C. After all the supernatants were collected, they were simultaneously tested for LT activity on L-929 cells for LT activity. Triplicate values varied by no more than 5%. The number of units of LT activity per milliliter in each sample is depicted on the ordinate, and the time the sample was collected on the abscissa. (0) Restimulated lymphocytes; (0) nonimmune C57Bl/6 murine spleen cells.
broad peak with approximately 70% of the lytic activity, as determined on L cells, present in the (Yu class. A shoulder of high-molecular-weight toxicity eluted in the void volume of the column. This was in contrast to nonimmune LT, where we rarely find Cx forms. Fractions from the void volume of the column were pooled and termed Cx. Fractions eluting with the molecular weight marker IgG (150,000 daltons) were pooled and termed (YH.

Glutaraldehyde-Treated Cx and (YH from Alloimmune C57B1/6 Lymphocytes Lyse both P815 and EL-4 Targets
Alloimmune lymphocytes restimulated with P815 cells for 4 days in vitro were assayed on "Se-Met labeled P8 15 and EL-4 target cells. Lysis of EL-4 cells never exceeded 5%, and 45-60s lysis of the P815 targets was observed ( Table 1). The resensitized lymphocytes were then stimulated for 3 hr with Con A. These supernatants were immediately assayed on L-929 cells, and the number of units per milliliter of LT activity was determined (  a Percentage lysis of PSI5 and EL-4 radiolabeled target cells by in vitro restimulated C57B1/6 anti-P8 I5 effector cells as described under Materials and Methods. Percentage release are expressed as the mean of triplicate values, and the standard deviations are within 5% of the mean. A 25:l effector to target ratio was used in all cases.
b The Cx and 0~" MW LT classes were obtained from fractioned glutaraldehyde-treated supernatants derived from lectin-stimulated in vitro-restimulated murine lymphocytes as described under Materials and Methods. 200 pl of either Cx or (Y" was incubated with the radiolabeled allogeneic target cells for I6 hr, percentage release calculated by the formula given under Materials and Methods, and the triplicate values exhibited ~10% variation.
' Supernatants were obtained from lectin-stimulated MLC murine lymphocytes and immediately assayed on L-929 cells to determine units of LT activity/ml. The LT units/200 ~1 of the Cx and (Y" fractions were calculated using the concentration factors, the dilution by the column, and the distribution of toxicity in the column profile. d ND, not done.
of P8 15 and EL-4 targets. In addition, the number of units/200 ~1 did not correlate with the specificity of the LT. However, it is possible that L cells are not equally sensitive to both Cx and aH LT molecules (see Discussion). This caveat precludes the use of LT units, as measured on L cells, as a basis for a definitive quantitative comparison of the amount of lysis Cx and (YH LT cause on P8 15 and El4 targets. The column fractions from the different cultures also varied in target specificities. These data do not unequivocally support either antigen-specific lysis or nonspecific lysis; indications of both are seen. Further work will determine if the source of this variability is the lymphocyte populations or whether Cx and ffH can exist in both antigen-specific and nonspecific stages.

Stabilization
of these high-molecular-weight murine LT forms with glutaraldehyde is a technical breakthrough which will permit biochemical studies of these materials. The lytic activity of these untreated forms has a 24-hr half-life, and the half-life of antigen-specific toxicity for allogeneic tumors is less than 1 hr. Thus, without stabilization, the lytic activity due to these forms is lost, leaving only the stable lower-molecular-weight forms normally found when longer production sys-terns are used (4,18). We have tried numerous techniques to stabilize murine LT lytic activity, e.g., storage at different temperatures and pH, the addition of polycations, protein carriers, glycerol, and protease inhibitors. None of these techniques were successful. Previous biochemical studies of murine LT forms suggested some of the instability of these molecules may be due to dissociation of Cx and (Yu forms into the weakly active smaller forms (3,5,11). With this possibility in mind, we attempted crosslinking fresh supernatants with glutaraldehyde. The present results demonstrated that we are able to maintain the high levels of LT lytic activity as assayed on L cells for several weeks at 4°C. In further biochemical studies of these materials, the use of reversible crosslinkers should be very useful.
The stabilizing effect of glutaraldehyde is apparently due to its direct action on the LT molecule(s). Inactivation of an external factor(s) by glutaraldehyde, e.g., a degradative enzyme, appears to be unlikely, because the addition of fresh untreated supernatant did not inactivate the glutaraldehyde-treated LT. Some of the remaining possible mechanisms of stabilization of the LT molecules by glutaraldehyde are: (i) intramolecular crosslinking, preventing dissociation of subunits or denaturation; (ii) inactivation of a portion of the LT molecule which functions as an autolytic enzyme; or (iii) masking site(s) on the LT molecule recognized and clipped by a degradative enzyme. It seems that intermolecular crosslinking of LT to other components in these supernatants did not occur, for had other materials been crosslinked to the LT, the increased molecular weight would be expected to alter the elution profile of the treated LT.
Studies by Hiserodt et al. suggested that the forms in these supernatants, which lysed cell lines other than the L cell, were the high-molecular-weight classes (11). However, because of the unstable nature of this lymphokine(s), only minimal testing of the different molecular weight classes of LT on allogeneic target cells, other than the L cell, was performed. Stabilization of murine LT has allowed us to further study the functional activity of molecular weight column fractions. It is now obvious that both Cx and aH are able to cause high levels of lysis when assayed on both P815 and EL-4 targets. When the fractions of the molecular sieving columns are assayed, the majority of the lytic activity, as determined by L cells, resides in the (YH pool. When assayed on P815 or EL-4 cells, however, the Cx pool contained as much if not more toxicity than the (Y" pool. This could be due to the P8 15 and EL-4 cells being more sensitive to Cx than (II H. Alternatively, L cells may be less sensitive to Cx than (Yu. This cannot be resolved in the absence of method to quantitate LT on a molar basis. Thus, the various molecular weight classes of LT may differ in their ability to lyse any particular cell line. There is evidence to suggest that column fractions are more lytically active than unfractionated supernatants, for when supernatants were used at the same concentration as the (YH or Cx fractions, they rarely caused greater than 10% lysis on the P815 or EL-4 targets (data not shown). We obtained these results regardless of whether we used the present production method, including glutaraldehyde stabilization, or the method previously used in our laboratory to produce LT from alloimmune murine lymphocytes (14). This finding suggests that an inhibitor may be present in the whole supernatant which is removed by the molecular sieving columns. Alternatively, these supernatants may be activated by passage over the column. This observation may explain past difficulties in demonstrating lysis of various cell lines with unfractionated supernatants and suggests a mechanism for limiting the activity of these forms to the microenvironment between effector and target cell(s).
The high levels of lysis caused by Cx and (YH LT led us to compare the lytic capacities of these column fractions with that of intact lymphocytes. For example, in one experiment the number of lymphocytes used. to generate the LT for each well of microtiter plate was equivalent to a 375:l effector to target cell ratio. In an 8-hr assay, the (Yn fraction from these lymphocytes caused 46% lysis of the P8 15 targets, so approximately IO-fold more lymphocytes and an assay twice as long were required for tiH to cause the same level of lysis as that found in intact lymphocytes. While LT-mediated lysis is obviously less effective than the intact lymphocyte, it cannot be excluded as a possible mediator of lysis, for a variety of reasons. For example, degradation of LT before glutaraldehyde treatment, adsorption to surfaces during concentration and chromatography, and splitting the LT into separately assayed Cx and (YH pools could all reduce the total LT activity recovered from the stimulated lymphocytes. Additionally, it is known that LT is expressed on activated lymphocyte membranes (19), and that cell contact is required for cell-mediated lysis. Thus, LT may only be delivered in a highly active form during cell contact, when a high local concentration on the membrane of the effector could be delivered to a target, or LT may require a lipid environment for full activity.
The LT-mediated lysis of the target cells did not always display the pattern of specificity observed when intact lymphocytes were assayed on the target cells. Only in Experiments IV and VI was there a marked preferential lysis of the P8 15 targets by the fractionated LT. In the other four positive experiments, two yielded strong lysis and two gave weak lysis of both targets. It is possible that low levels of activity and instability kept this nonspecific LT below the threshold of detection in previous work, which only reported specific lysis (14). Alternatively in the present study, glutaraldehyde may have stabilized some of the LT in a nonspecific form, which was not detected previously.
The stabilization of murine LT reported in this study has removed one of the most inconvenient facets of this experimental system. Already stabilization has allowed us to demonstrate that the larger-molecular-weight forms are capable of both selective and nonselective lysis.