Safety of DEHP: role of the bureau of biologics.

The position of the Bureau of Biologics (BoB) relative to the issue at hand is unique on several counts. First, Bureau of Biologics is formally interested in the phthalate issue only in connection with containers for blood and blood products. The authority to regulate the nature of such containers derives from specific regulations. For example, regarding plasma, 21 CFR, Section 648.68(b) states: "Final containers and their components shall not interact with the plasma contents under conditions of storage and use so as to alter the safety, quality, purity or potency of the plasma." Second, the subjects involved are exposed to the highest doses of DEHP, and it is delivered via the intravenous route. Most of these subjects are patients undergoing treatment, although the increased interest in pheresis procedures will result in similar exposure for more and more normal donors. Finally, BoB has considered the DEHP in vinyl plastics to be a problem for a number of years. That is, we have been aware that DEHP, a biologically active material, was leached by blood and plasma in considerable quantities from the PVC bags. We were also aware that the available data on toxicity did not support a regulatory demand for immediate change. Members of the blood, banking community were similarly concerned, and this concern was shared by the manufacturers of blood bags, as indicated by the considerable research expenditures dedicated to the quest for a substitute plastic. At BoB the concerns which derived from the presence of DEHP were balanced by an appreciation of the need for a flexible container. The entire science of blood component preparation in a sealed system is dependent upon a flexible container. The Bureau of Biologics would therefore consider that reversion to a system based on glass bottles is not a viable alternative.

Summary.-A line of human leukaemia-derived cells is described that kills a wide range of human and animal cell lines, whether normal or malignant, even at a ratio of 1: 1. During exposure to the target cells, the killer cells released a factor into the culture medium which destroyed target cells in the absence of the killer cells. This phenomenon occurs without exogenous complement and requires no pre-treatment of target or killer cells. The humoral factor is a protein precipitable by 60% saturation of ammonium sulphate and has a mol. wt. of approximately 70,000. It prevented the growth of a fibrosarcoma in mice.
CELL killing by lymphoid cell lines, either mediated by antibody or by direct cell-to-cell contact (K-cells and T-cells), is well documented (reviewed by Cerottini and Brunner, 1974). In addition, cell killing of certain target cells only can be induced by lymphokines secreted by activated T-and B-cells. This report describes the properties of a killer cell line which differs from both T-and K-cellmediated killing. This unique culture has been derived from the white blood cells of a leukaemic patient, but the killer cells do not represent the malignant cell population. During the contact of the killer with target cells a humoral substance is released by the killer cells, which by itself can kill cells.

MATERIALS AND METHODS
Origin of the cell culture.-A 32-year-old male patient presented with a mediastinal mass and lymphadenopathy in the right supraclavicular fossa, without hepatoor splenomegaly. Peripheral blood and bone marrow films were normal. He was diagnosed as a non-Hodgkin's lymphoma. Six weeks later, meningeal symptoms developed, but the blood picture remained normal.
During the 12th week his blood became leukaemic with 105 x 109/1 white blood cells (WBC), platelets 40 x 109/1 and haemoglobin 8-3 g/dl. It was during this week that peripheral blood was obtained for setting up the cell cultures of WBC. The patient died a day later.
Heparinized blood was left at room temperature for 2 h: the buffy coat was then taken off and spun at 500 g for 10 min at 50C.
The WBC were re-suspended in RPMI-1640 culture medium supplemented with antibiotics (100 iu/ml penicillin; 100 ,ug/ml streptomycin) and foetal bovine serum at a final concentration of 10%. The cells were then seeded in 125-ml Erlenmeyer's flasks with 50 ml of culture medium, and placed in a humidified CO2 incubator in an atmosphere of 5% C02 95°% air at 370C. The medium was changed every 5 days by aspirating about half the medium and replacing it with fresh medium. For large-scale growth, the cells were transferred into waterjacketed spinner cultures (Wingent Ltd., Cambridge) of 1-5 litres' capacity. To test the cytotoxic action of the cultured cells, they were counted, separated from the growth medium and re-suspended in fresh RPMI-1640 with or without serum to give the desired concentration, before being added to cultures of target cells.
Morphological and immunological proper-tie8.-For morphological studies cytocentrifuged smears of the cultured cells were stained at various intervals with May-Griinwald-Giemsa dyes. Ultrathin sections of the packed, fixed cells were prepared and examined with the electron microscope by routine methods (Cawley and Hayhoe, 1973).
The immunological characteristics of this and 24 other newly established haemic cell lines from leukaemic patients have been described in detail in a separate paper (Gordon et al., 1977). Briefly, rosetting techniques were used for the identification of cells with receptors for gamma Fc, the C3 component of complement, and sheep red blood cells (SRBC). Fluorescein-conjugated antisera to immunoglobulin (heavy and light chains) were used to determine SIg and intracellular Ig. Newly synthesized immunoglobulin was estimated by the sandwich technique of immunoprecipitation. Tests for the presence of Epstein-Barr viral nuclear antigen (EBNA) were kindly performed by Dr D. Crawford. Cellular killing.-The killer cells were seeded on a wide range of stationary human and animal cell lines and also co-cultivated with suspension cultures of human leukaemiaderived cells. Human cell lines used as target cells included dermal fibroblasts from normal human foetus; the A-204 line from a rhabdomyosarcoma (Giard et at., 1973); the KHOS line from an osteosarcoma (Rhim, Cho and Huebner, 1975); a suspension culture of SIg-positive cells (Line 139) derived from a patient with acute myelomonocytic leukaemia; and a suspension culture of lymphoid cells (line 45) derived from a child with a mediastinal lymphoid neoplasm. The lastnamed cells are probably T-cells, since they form spontaneous SRBC rosettes and are SIg negative (Smith et al., 1973). Animal cell lines derived from dog thymus, rabbit cornea (SIRC), mink lung, bat lung and fibrosarcomatous mouse cells (Balb MSVDNA) (Karpas and Kleinberg, 1974) were also used as target cells.
Similar killing ability was sought in 17 other haemic cell lines established by the author from various haematological malignancies (7 ALL, 10 AML). In each case the cell line under test was seeded as an effector target cell ratio of 10: 1. Cellular killing was monitored by visual examination of the stationary target cells after removal of the killer cells by gentle washing, and by staining the mixed populations of killer and target cells. It was estimated by two methods: (1) The loss of [125I]5-iododeoxyuridine (125ldUrd) from pre-labelled (0-1 ,uCi 125IdUrd/2 x 105 cells) confluent layers of target cells after exposure to killer cells.
(2) 51Cr release from pre-labelled (10 uCi 51Cr/106 cells) target cells after various intervals of exposure to killer cells. The percentage of 56Cr release was calculated according to the following formula: Specific releasecontrol release Total available labelcontrol release X 1 Effect of conditioning on cell killing.-Since there was a delayed killing effect by fresh killer cells, it was ofinterest to determine whether conditioning would enhance killing and whether this conditioning was specific. The term " conditioned cells " was applied to killer cells which had already been in contact with, and had killed target cells. The rate of killing by conditioned and non-conditioned killer cells was monitored and estimated using the human osteosarcoma-derived cells (KHOS line) as the target. KHOS and SIRC cells were used for conditioning. 51Cr release and 125JdUrd loss were used to estimate the effect of conditioning on cell killing.
The effect of conditioning on killer cells was also tested in an experiment with the target cells grown in suspension. Killer cells were co-cultivated with a suspension culture of human leukaemic lymphoblasts (Line 45). This line is a homogeneous population of small lymphoid cells with a high nuclearcytoplasmic ratio, and is easily distinguished morphologically from the killer cells. After 80 h of co-cultivation, the only viable cells left in the culture were killer cells, and these were then separated from the dead cells by centrifugation over a Ficoll-Triosyl gradient.
These conditioned killer cells, and also fresh killer cells, were then seeded with fresh 51Crlabelled (Line 45) lymphoid cells. Incubation of the 51Cr-labelled cells in microtitre wells was continued at 37°C in duplicate, under 5 different conditions as outlined in Table I. Conditioned and non-conditioned killer cells as a control were each added to target cells at a ratio of 10 : 1 (killer : target).
Humoral killing.-The term " conditioned medium " is used for culture fluid obtained after 72 h co-cultivation of the killer with target cells. Conditioned medium was centrifuged for 15 min at 1000g and the supernatant re-centrifuged at 10,000g for 20 min at 4°C. 153 In order to avoid a non-spE inhibition by conditioned medi in nutrients, each preparation o medium was dialysed overn 10 volumes of fresh medium.
filtered through a 0-45-ptm mi The fresh medium used for filtered and used as a contro experiments. Target cell cultures were set' 30-mm plastic petri plates with: of each of the 5 cell lines listed Each of the cell lines tested M 12 plates. The following dav fluid from each of the cultures w duplicate with 2 ml of either (1' from non-conditioned killer cel logous conditioned medium, (3) (6) heterologous conditioned me ed from conditioning with the lines. All the conditioned me from the 5 cell lines were supplI 10% fresh foetal bovine serum.
After incubation at 370C cultures were examined micros ecific growth 125JdUrd (0-2 ,uCi) added to each plate. ium deficient Incubation was continued for a further 20 h If conditioned before re-examination for cell death microight against scopically, and quantitatively by measuring It was then incorporation of the radioactive label. Adillipore filter. herent cells were gently washed (x 2), dialysis was trypsinized and transferred into separate 1 for further vials (2 plates/vial) for assay of '25IdUrd uptake. Concentration of the cytotoxic substance and ioned Killer estimation of its molecular weight.-Ammonic T-cells in unm sulphate was added to several preparations of conditioned medium to give 60% saturation. These were then spun at 10,000 51Cr release rev/min for 20 min, and the pellets dissolved 100 of the fraction precipitated by 60% saturaukaemic T-cells tion of ammonium sulphate on target cells, was evaluated in the same manner as conditioned medium (humoral killing).
For an estimation of the mol. wt the up by seeding concentrated (x 100) cytotoxic fraction was 105 cells/plate spun at 100,000g for 1 h. One ml of the clear I in Table II. supernatant was then layered -on a 12-ml vas grown on sucrose gradient (50-20%) and centrifuged for r the culture 24 h at 40C in a swing-out rotor (MSE 43127as replaced in 111) at 30,000 rev/min. Using an identical ) culture fluid sucrose gradient, several known protein [Is, (2) homomarkers were added in 1-ml volumes. The ), (4), (5) and markers used were haemoglobin, albumin, -dium obtainferritin and IgM.
other 4 cell Fractions of 8 drops (0-45 ml) were collect-:dia prepared ed and the optical density (OD) of each fracemented with tion was measured. The fractions shown to contain the killing substance were diluted for 22 h the x 4 in phosphate-buffered saline, and diacopically and lysed overnight at 40C against 50 volumes of fresh culture medium. The dialysate was filtered through a 0 45-,um millipore filter and supplemented with foetal calf serum to give a final concentration of 10% serum. 0-2 ml of each fraction was added to each of 3 wells which a day earlier had been seeded with 5 x 103 mouse tumour cells/well. After 48 h of incubation at 37°C, 0-02 ,tCi of 125IdUrd in 20-jAu volume was added to each well. Following 20 h of incubation, the culture fluid was sucked off, the wells rinsed, and the uptake of 125JdUrd by cells in each well was measured (Fig. 5). Animal studies.-Fifteen 4-month-old BALB/c mice were each injected s.c. on their back with 5 x 105 fibrosarcoma cells (Balb MSV DNA line). This line had previously been transformed by DNA from cells carrying the murine sarcoma viral (MSV) genome (Karpas and Kleinberg, 1974). Conditioned medium, prepared after 3 days of co-cultivation of the killer with the malignant mouse cells at an effector-to-target-cell ratio of 10: 1 was used for injection into 9 of the mice. Each of the mice received 7 s.c. injections in their backs of 1 ml on Days 4, 10, 13, 18, 22, 25 and 28. The other 6 mice were injected with 1 ml offresh culture medium only, on the same days.

Morphological and immunological
properties The initial leukaemic cell population consisted mostly of small lymphocytes with a high nuclear/cytoplasmic ratio. The immunological findings of SJg negativity, combined with the ability to rosette with SRBC, indicated that this cell population was probably of T-cell origin. However, after 3 months in culture, the leukaemic T-cells disappeared, and a

155
A. KARPAS pleomorphic population of larger mononuclear cells continued to proliferate. Multinucleated giant cells were also frequently observed. In May-Grunwald-Giemsa-stained films the nuclei of most cells contained dense granular chromatin, staining reddish purple, and large blue nucleoli. The cytoplasm stained dark blue, except for the lightly staining area of the Golgi apparatus near the nucleus. EM examination of the cells revealed a very well-developed rough endoplasmic reticulum and Golgi apparatus (Fig. 1). Examination of both live and stained cultures revealed the presence of numerous cytoplasmic protrusions (blebs) and many of these appeared to be released from the cell surface as drops of varying sizes. Similar protrusions have been described in mature human myeloma cells (Hayhoe and Flemans, 1969). They probably represent a form of secretion of Ig by the cells, since large quantities of IgM(A) were found to be released into the culture medium by those cells (Gordon et al., 1977). All the cells also stained for co -J kK 0 surface and intracellular IgM(A) and expressed the Epstein-Barr viral nuclear antigen (EBNA). Some 50% of the cells possessed receptors for C3, and 25% receptors for gamma Fc (Gordon et al., 1977).

Cellular killing
When the killer cells were co-cultivated with any of the cell lines tested, even with a killer: target cell ratio of 1: 1, microscopic examination revealed that the entire target cell population was destroyed within a few days. Killing of target cells also occurred when serum-free medium was used. The capacity to destroy other cells was found to be a unique property of this particular line, as 17 other cell lines that have been established by the author from various haematological malignancies failed to affect the target cells when seeded in an effector-to-target-cell ratio of 10: 1. Fig. 2 and Table III illustrate   Loss of 1251 monitored at daily intervals expressed by the decrease in ct/min relative to control cultures without killer cells. sarcoma genome and which grows rapidly to a high density, appears to be very susceptible to the killing effect. Over two thirds of the cells became detached within the first 48 h when grown with killer cells at a killer: target cell ratio of 10: 1, and one third detached when the ratio of killer cells was 1: 1. On the other hand, a smaller percentage of dog thymus cells detached within the first 3 days when at a ratio of 10: 1.
The killing of target cells growing in suspension, e.g. the human leukaemia T-cell line, was assessed by the staining of samples from the mixed cultures. Since the cultured T-cell line (Line 45) was made up of small cells with high nuclear-cytoplasmic ratio, they were easily distinguished from killer cells with their particular morphological characteristics. After 3 days of co-cultivation of killer with the T-cells (at a ratio of 1: 1) only the killer cells remained viable.
Killing of target cells also occurred when the medium was not supplemented with serum. Total destruction of KHOS target cells (as measured microscopically) also occurred when fresh growth medium was added daily to the mixed cultures, indicating that the death of target cells was not due to the exhaustion of nutrients in the culture fluid. On the other hand, when the killer cells were washed off the target cells (KHOS, mink lung, SIRC, Balb MSV DNA) before their complete destruction, the viable target cells which remained again proliferated when fresh growth medium was added.

Effect of conditioning on cell killing
The result of the first experiment in which 51Cr was used to label the KHOS target cells is outlined in Fig. 3 and Table IV. It indicates that following the exposure of killer to target cells, the killer cells will kill newly exposed target cells more rapidly. Conditioning is not specific, since killer cells conditioned with normal rabbit cornea cells killed human malignant KHOS cells as efficiently as did killer cells conditioned with KHOS cells.
In addition, the 51Cr labelling was found to be a far more sensitive indicator for the onset and course of the killing phenomenon than was 1-25IdUrd. Thus, using 51Cr labelling, the minimum time required for the initiation of killing by nonconditioned cells was about 20 h, whereas conditioned cells started to kill by the 5th hour ( Fig. 3 and Table IV). When, however, the target cells were labelled with 125IdUrd, the onset of killing appeared to be delayed (Fig. 4). This may be due to the fact that cell death is registered only when the cells (and their 1251) come off the plate surface. Several hours may elapse between the time the cell is irreversibly damaged and its complete detachment. However, the difference in the onset and course of killing between conditioned and non-conditioned cells in any of the 3 ratios of killer-to-target-cells tested was between 15 and 20 h, a similar time interval to that obtained in the 51Crrelease experiment. The effect of conditioning on killer cells was also shown to occur in suspension cultures of the human T-cell (Line 45). WVhen 5'Cr release was assayed after 4 h of incubation, there was no significant increase in any of the wells. As can be seen in Table I, a significant 51Cr release could be detected by 10 h, but only from target cells exposed to conditioned killer cells. By the 24th hour there was already about 80% spontaneous release. Therefore only the results at the 10th hour are illustrated.

Humoral killing
In the experiment outlined in Table II, the inhibition of 125IdUrd uptake in each of the different lines by the various conditioned media was expressed in relation to the 125IdUrd uptake by cells in the presence of culture fluid from killer cells alone. The malignant mouse (Balb MSV DNA) and normal rabbit cornea cells (SIRC line) were found to be the most susceptible lines, with extensive cellular degeneration and detachment of cells during the second day. 125IdUrd uptake by the normal mink cell line was also minimal in the presence of the various conditioned media, but the mink cells did not degenerate or detach as quickly as the mouse and rabbit cells. The uptake of 125JdUrd by the dog thymus and KHOS cells was also inhibited by several heterologous conditioned media, but the highest degree of inhibition was recorded by the homologous conditioned medium.
Several cell-derived factors are known to affect the incorporation of radiolabelled nucleotides; therefore the incorporation experiments were controlled by monitoring target cell deaths microscopically. A direct, inverse correlation between 158 observed cell death and incorporation of 125JdUrd (i.e. DNA synthesis) was seen to exist.
Concentration of killing substance and estimation of molecular weight The cytotoxic fraction was readily precipitated in 60% saturation of am-, monium sulphate, since dilution of the concentrated precipitate to the original volume gave a similar killing efficiency to the undiluted culture fluid. The killing fraction banded at a sirnilar sucrose densitv to haemoglobin, and it assumed to have an apl of 70,000. As can be E distribution of the c represents only part of band. Animal studies After 2 weeks, 5/6 co ed tumours which incre and killed the animals bi 7th week. The 6th n tumour by the 10th week and died 2 weeks later. Of the 9 mice injected with conditioned medium, one developed a visible tumour by the 24th day and another by the 27th day after implantation of the malignant cells. The tumours gradually increased in size, and these 2 animals died in the 9th week. The other 7 mice remained free from any obvious malignancy 20 weeks after the injections. DISCUSSION t can therefore be The properties of a universal killer cell proximate mol. wt. line which has been derived from the white seen in Fig. 5, the blood cells of a leukaemic patient is *ytotoxic fraction described. The killer culture is made up the major protein of a pleomorphic population of mononuclear cells of variable sizes, together with multinucleate giant cells. The cells contain a verv well-develoned rouah endoplasmic reticulum. They are probably B-cells, since 100% of the cells stain for SIg (IgM(A)). These results demonstrate that the cells which are actively proliferating in vitro are very different from the original malignant T-cell population of the patient, which was uniform small lymphoblasts. It is likely that they were derived from one of the other leucocytes present at the time of the fatal disease in the patient. Cells with similar properties have never before been derived from leukaemic patients, and they are obviously not just another EBNA-positive line. They may represent a subpopulation of cells which have been transformed by the same agent that caused the T-leu- Kemiaii, oUt were mUore suiuaUle TL proIiieration in vitro. killing factor in The mode of cell killing described in -rose gradient as this paper differs in several ways from nt.
other forms of in vitro cell-mediated killing actor assay (for a recent review see Cerottini and Brunner, 1974). After a lag period, it can kill in vitro normal as well as malignant cells, even at a ratio of 1: 1, irrespective introl mice developof whether they grow as stationary, conased rapidly in size fluent, monoor multilayer, or in suspenletween the 6th and sion, in the absence of exogenous comnouse developed a plement. The killer cells do not require 159 160 A. KARPAS prior stimulation by exogenous chemical agents such as phyto-haemagglutinin (PHA), nor do they require the prior coating of the target cells with specific, preformed antibodies. The spontaneous secretion by these B-cells of a humoral factor capable of killing a wide range of cell types appears to distinguish this factor from the various lymphotoxins which are produced by short-term cultures of T-cells with limited range of toxicity. The minimum time required for the initiation of killing by non-conditioned cells was about 20 h, while conditioned cells started to kill by the 5th hour.
During the incubation of the killer with target cells, a humoral, as yet unidentified, substance is discharged into the culture fluid of killer cells after contact with target cells. This substance is probably a protein, since it can be precipitated by ammonium sulphate. It banded on sucrose gradients in the same region as haemoglobin and therefore its mol. wt. is probably around 70,000.
The in vitro experiment does not point to a highly specific cytotoxic factor, since only 2 cell lines were more susceptible to the homologous conditioned medium, while the growth of the other 3 cell lines was equally inhibited by heterologously conditioned medium. This might reflect the presence of a common cellular target site. However, the first animal experiments suggest a certain degree of specificity, since the humoral factor prevented the development of malignant fibrosarcoma in 7/9 mice which had received implants of fibrosarcoma cells. It also delayed the development of the tumour in the 2 mice which developed fibrosarcoma, while 5/6 control mice which received no injections of conditioned medium developed sarcomata and died within 6 weeks, the 6th dying in the 10th week.
Therefore the factor appears to have an effect on the growth of the tumour cells without any ill effect on the host. The nature of this factor and its potential use as an anti-tumour agent is being investigated.
It is a pleasure to acknowledge and thank Dr C. Milstein for his continuous interest and stimulating discussions, and for reviewing the manuscript. Also Dr J. C. Cawley for the electron microscopic examination of the cells. The work was supported by the Leukaemia Research Fund, U.K.