Surface Features of Human Natural Killer Cells and Antibody-Dependent Cytotoxic Cells

J. Cell Sd. 77, 27-46 (1985) 27 Printed in Great Britain © The Company of Biologists Limited 1985

SURFACE FEATURES OF HUMAN NATURAL KILLER CELLS AND ANTIBODY-DEPENDENT CYTOTOXIC CELLS

CLAIRE M. PAYNE1*, ALEC LINDE1, RUTH KIBLER2, BONNIE POULOS2, LEWIS GLASSER1 AND ROGER FIEDERLEIN1 'Department of Pathology and 2 Department of Microbiology and Immunology, University of Arizona, College of Medicine, 1501 N. Campbell Avenue, Tucson, Arizona 85724, U.SA.

SUMMARY The purpose of the present study was to examine the surface features of purified large granular lymphocytes (LGLs) (natural killer (NK) cells, antibody-dependent cytotoxic lymphoid (ADCL) cells, K-cells, Fcy( + ) third population (non-T, non-B) lymphoid cells, Tr cells) by scanning electron microscopy (SEM) and to compare their surface features with granulocytes, monocytes and Fcy(—) lymphoid cells that were all fixed for SEM under identical conditions. We have determined that 72-80 % of LGLs enriched by rosette formation with sensitized erythrocytes or using Percoll gradients, have a complex microvillous surface (CMS) pattern identical to that of lymphocytes. The LGL fraction appears by SEM to represent a morphologically homogeneous population of cells. Monocytes prepared for SEM under identical conditions had distinct surface folds and granulocytes displayed numerous broad-based ridge-like profiles. The majority of lymphoid cells in an unfrac- tioned population have a CMS pattern when incubated at room temperature (25 °C) before fixation, and a sparse microvillous surface (SMS) pattern when incubated at body temperature (37°C). Ficoll-Hypaque (FH) also had a direct effect on the cell surface pattern. Over half of the unfractionated lymphoid cells displayed a CMS pattern after cells were washed free of FH and incubated at 37 CC before fixation. The CMS pattern is therefore not unique to LGLs but can be produced by the surface alteration of non-LGLs found in unfractionated buffy coat and mononuclear fractions. The interactions between LGLs and sensitized erythrocytes in an antibody-dependent cytotoxic assay system, and LGLs and K562 target cells in an NK assay system, were also examined. This is the first report that describes the surface features of human LGLs interacting with K562 target cells in an NK assay system. The LGL populations studied by SEM were determined to have a high percentage ofLeu-ll( + )and Leu-7 ( + ) cells. These same populations were also shown to have high antibody-dependent cytotoxicity and NK activity using the Cr release assay.

INTRODUCTION Human natural killer (NK) cells (Herberman & Ortaldo, 1981), antibody- dependent cytotoxic lymphoid cells (K cells) (Perlmann, 1976), Fcy ( + ) third population (non-T, non-B) lymphoid cells (Winchester, Fu, Hoffman & Kunkel, 1975) and Ty cells (Ferrarini et al. 1980) appear to comprise a unique leucocyte population that is characterized at the light-microscopic level by the presence of azurophilic granules (Timonen, Ortaldo & Herberman, 1981) and at the ultrastructural level by the presence of parallel tubular arrays (PTAs) (Burns, Zucker-FrankJin & Valentine,

• Author for correspondence. Key words: SEM, NK cells. 28 C. M. Payne and others 1982; Henkart & Henkart, 1981; Payne & Glasser, 1981; Payne, Glasser, Fiederlein & Lindberg, 1983; Smit, Blom, van Luyn & Halie, 1983). This cell type is an important part of the immune system and is believed to function in tumour sur- veillance, resistance to infection by viruses and other microbes, and rejection of bone marrow grafts (Herberman & Ortaldo, 1981). Although they are referred to as large granular lymphocytes or LGLs (Timonen ef a/. 1981), recent evidence indicates that LGLs may have a myeloid origin (Gallief al. 1982; Kay & Horwitz, 1980; Lohmann- Mattes, Domzig & Roder, 1979; Neighbour, Huberman & Kress, 1982; Oehl et al. 1977). (The term 'myeloid' refers in this paper to non-lymphoid cells that include granulocytes and monocytes). Since LGLs represent a functionally heterogeneous population of cells with different surface antigens (Fitzgerald, Evans, Kirkpatrick & Lopez, 1983), it is conceivable that the LGL population may be mixed, with some subpopulations lymphoid in origin and others myeloid in origin (Zarling, Clouse, Biddison & Kung, 1981). On the other hand, other investigators believe that they may represent a unique population of cells with a separate haematopoietic lineage (Fitz- gerald et al. 1983; Horwitz et al. 1978; Ortaldo, Sharrow, Timonen & Herberman, 1981). Combined surface-marker and transmission electron microscopic studies from our laboratory have indicated that the Fcy( + ) lymphoid cells that characterize the LGL population (NK, K and Tr cells) are ultrastructurally unique (Payne & Glasser, 1981; Payne et al. 1983). The cells appear lymphoid in nature by both light microscopy and transmission electron microscopy (TEM), but possess unique cytoplasmic inclusions called PTAs. The PTAs can be used as an ultrastructural marker for the K (Henkart & Henkart, 1981; Payne & Glasser, 1981; Smit etal. 1983), NK (Babcock& Phillips, 1983; Burns et al. 1982; Henkart & Henkart, 1981) and TY (Payne et al. 1983) cells. We have shown, however, that at least some LGLs are capable of phagocytosing complement-coated bacteria (Payne & Nagle, 1980) and in this respect appear more myeloid than lymphoid in nature. Contrary to the enzymic content of myeloid cells, PTA-containing cells lack peroxidase activity (Payne & Nagle, 1980) that is so characteristic of granulocytes and monocytes. The scanning electron microscopic (SEM) features of human granulocytes (Hattori, 1972; Polliack, 1978), monocytes (Dantchev & Belpomme, 1977; Polliack, 1978) and lymphocytes (Dantchev & Belpomme, 1977; Hattori, 1972; Kelly & Nockolds, 1977; Newell, Roath & Smith, 1976; Polliack, 1978; Roath et al. 1978) are quite distinctive, but no SEM studies have been done to compare the surface features of these cells with LGL-enriched populations. Previous SEM studies of human LGLs have described the interaction between effector and sensitized target cells (Alexander & Henkart, 1976; Biberfeld, Wahlin, Perlmann & Biberfeld, 1975; Inglisef al. 1975; Kelly & Nockolds, 1977; Perlmann, 1976), but different experimental conditions used in various laboratories make it difficult to compare the surface features of these antibody-dependent cytotoxic effector cells with purified cell populations. The surface features of human LGLs interacting with K562 target cells in an NK assay system have not previously been reported. In the present study, the surface features of LGLs were studied and compared with Quantitative SEM study of natural killer cells 29

Fcy(—) lymphoid cells, a granulocyte-enriched fraction and monocytes (that were fixed for SEM under the same experimental conditions) in order to determine if the LGLs have a unique surface appearance or represent a morphologically heterogeneous population of cells. A quantitative SEM evaluation of the effect of several experimental parameters used in the purification of LGLs is reported.

MATERIALS AND METHODS Enrichment of LGLs (K, NK cells) by rosette formation ivith sensitized erythrocytes Unfractionated buffy coat fractions. Venous blood was collected by venipuncture from healthy human adults and drawn into heparinized evacuated tubes. Whole blood was spun at 200 gfor 10 min in a swinging-bucket clinical centrifuge. Most of the platelet-rich plasma was removed and discarded. The loose buffy coat layer was removed and mixed 1:1 with phosphate-buffered saline (PBS). A sample of this cell suspension was allowed to stand at room temperature (25 °C) for 1 -5 h before fixing for SEM (fraction 1). A second sample was incubated at 37 °C for 1 -5 h before fixing for SEM (fraction 2). Mononuclear fraction. A PBS-buffy coat cell suspension prepared as above was incubated at 37 °C for 1 -5 h (fraction 2) and then layered over an equal amount of Ficoll-Hypaque (Lymphocyte Separating Medium (LSM), Litton Bionetics, Kensington, MD). The cells were centrifuged at 800gfor 20 min in a swinging-bucket clinical centrifuge and the cells at the interface were removed. One sample was immediately fixed for SEM (fraction 3); another was rinsed with PBS and allowed to incubate at 37CC in PBS for 3 h before fixing for SEM (fraction 4). Monocyte-depleted lymphoid fraction. A mononuclear fraction obtained after Ficoll-Hypaque density-gradient centrifugation was incubated at 37 °C for 30 min with LSR (Technicon Instrument Corp., Tarrytown, NY). The LSR contains /im-sized magnetic particles sensitized with poly-L- lysine. Following incubation, the cells were passed through Tygon tubing wrapped around a magnet that effectively removed iron-laden phagocytes and excess iron filings. LGL/EAiu, rosettes. This procedure selects lymphoid cells that have high-avidity Fcr receptors for cytophilic antibody. Cells that function in an antibody-dependent cytotoxic assay system and in an NK assay system are both enriched by this procedure. The monocyte-depleted fraction (106 cells/ml) was mixed with an equal volume of 1 % sensitized RiRa human erythrocytes (Payne & Glasser, 1981), centrifuged for 5 min at 200 £ and placed upright in tubes at room temperature for 30 min. The cell pellet was gently resuspended and a sample prepared for SEM (fraction 5). LGL(Fcr(+) cells)- and non-LGL(Fcr(—))-enriched fractions. A sample of the rosetted cell preparation from above (fraction 5) was layered over Ficoll-Hypaque and centrifuged at 400 £ for 30 min. The interface cells were removed, allowed to stand at room temperature for 1 h and then fixed for SEM (fraction 6). The cell pellet was resuspended in PBS and incubated at 37 °C for 30 min to dissociate the IgG (attached to the EAhu marker particles) from the lymphoid cell surface (Kumagai, Abo, Sekizawa & Sasaki, 1975). Examination under the light microscope revealed that the vast majority of lymphoid cells had no attached erythrocytes. The cell suspension was then layered over Ficoll-Hypaque and centrifuged at 800 g for 20 min to pellet the erythrocytes. The interface cells (enriched LGLs) were removed and fixed for SEM (fraction 7). Fractions 6 and 7 were fixed at the same time.

Enrichment of LGLs using Percoll gradients Although NK activity is present in the enriched cell fraction isolated by rosette formation with sensitized erythrocytes (fraction 7), investigators studying NK cell activity typically isolate these effector cells using Percoll gradients. For completeness, the LGL effector cells isolated on Percoll gradients were also studied. Cell separation for LGL enrichment. Mononuclear cells were isolated from heparinized venous blood on Ficoll-Hypaque gradients. The mononuclear cells were then separated by passage through nylon wool columns to remove B cells and adherent monocytes (Julius, Simpson & Herzenberg, 1973). The non-adherent cells were further purified by rosetting with sheep red blood cells at 29 °C, which permits binding of high-affinity rosetting T cells but not the low-affinity rosetting LGLs 30 C. M. Payne and others

Preparation of cell fractions: flow chart

Whole blood

FHS Fix o

IFC/PBS (37°C, 3h) -•Fix O IFC/PBS+LSR (37°C, 30min) I 'Mo MDC fraction I MDCs + EAhu (25 °C, 30min) I LGL/EAhu rosettes 0 (In suspension)

FHS IFC(25°C, lh)

Pellet/PBS (37°C, 30 min)

FHS

IFC Enriched LGL fraction 0 This flow chart depicts the overall scheme for the enrichment of cell fractions and indicates the specific points at which cell fractions were fixed for SEM studies. Fractions 1-7 (numbers encircled) were obtained from the experimental protocol used to enrich LGLs by rosette formation with sensitized erythrocytes. Fraction 8 represents Percoll-enriched LGLs. Fraction 9 represents Percoll-enriched LGLs that were reacted with KS62 target cells. Fraction 10 represents a granulocyte-enriched preparation. Abbreviations used: BCF, buffy coat fraction; EAhu, human group O, type R1R2 erythrocytes sensitized with Ripley anti-CD serum; FHS, Ficoll-Hypaque Separation; Gran, granulocytes; IFC, interface cells; K562, target cells used in NK assay system; LGL, large granular lymphocytes; LSR, lymphocyte separator reagent, MDC, monocyte-depleted cell; Mo, monocytes; PBS, phosphate buffered saline; PG, Percoll gradient; PRP, platelet-rich plasma. •Additional steps present (see Materials and Methods for details). Quantitative SEM study of natural killer cells 31 (Timonen et al. 1981). The resetted cells were removed from the population by centrifugation on discontinuous density gradients of Percoll (Pharmacia Fine Chemicals, Piscataway, NJ). The gradients were prepared by diluting Percoll with RPMI 1640 medium to final concentrations of 375%, 42-5%and47-5% (Timonen etal. 1981). After centrifugation at 300 g for 30 min at room temperature, the cells that were in the 37-5 % and 42-5 % layers, as well as those cells at the interface between these two layers, were collected, washed and resuspended in RPMI 1640 medium supplemented with 10 % heat-inactivated foetal calf serum (FCS; Irvine Scientific, Santa Ana, CA) and 100/Jg/ml gentamicin (HEM Research, Inc., Rockville, MD). Characterization of effector cells (fraction 8). Identification of cells as LGLs was accomplished by fixing air-dried cytocentrifuge preparations for 5 min in absolute methanol and then staining with May—Grunwald— Giemsa stain. Differential counts were made at X 1000 magnification under oil immersion and at least 200 cells were analysed for each preparation. The LGLs were identified as large lymphocytes with a high cytoplasmic/nuclear ratio and a non-vacuolar cytoplasm containing azurophilic granules. An indirect immunofluorescence procedure using the monoclonal antibodies anti-Leu-1, anti- Leu-7 and anti-Leu-11 (Becton-Dickinson, Mountain View, CA) was also used to evaluate the cell types present in the LGL preparation. The monoclonal antibodies were added to cell pellets for a 30-min incubation on ice. The unbound primary antibody was removed by a single wash in FCS, and the cell pellets were resuspended in rhodamine-conjugated F(ab')2 fragment of goat anti-mouse IgG (heavy and light chain-specific, Cappel Laboratories, Cochranville, PA) at a predetermined dilution. After 30 min incubation on ice, unbound conjugate was removed by two washes in FCS. The cell pellet was resuspended in FCS, placed on slides, air dried, fixed in 95 % (v/v) ethanol and mounted in Gelvatol. At least 200 cells per slide were scored for the percentage of cells with membrane fluorescence by reading on a Zeiss microscope equipped with epifluorescence. LGL-target cell preparations. K562 cells, originally derived from a patient with chronic mye- logenous leukaemia, served as the NK-sensitive target cell line. K562 cells at a concentration of 7 X 10s cells/ml of supplemented RPMI medium were mixed with Percoll-enriched LGLs at a concentration of 3-5 X 106 cells/ml of supplemented RPMI medium and incubated at 37°C. Samples were removed at incubation times of 30, 60 and 180 min and fixed for SEM (fraction 9). Natural killer cell assay. K562 target cells (5 X 106) were labelled with lOOflCi of sodium 51chromate solution (New England Nuclear, Boston, MA; sp. act. 100-400 mCi/mg Cr) for 1-5 h (Timonen et al. 1981). After washing, 6 X 103 cells in 01 ml of RPMI 1640 medium supplemented with 10% FCS were pipetted into round-bottom microtitre wells. LGL effector cells in 0-1 ml medium were added to triplicate wells at an effector to target cell ratio (E:T) of 10:1. The plates were centrifuged at 200 £ for 5 min and incubated for 4h at 37 °C. Supernatant fluid was then collected and counted in a gamma counter. The percentage specific 51Cr release (%Sp.R.) was calculated from the following equation: X experimental release - X spontaneous release vb op.R. =—• • X luu. X total incorporation - X spontaneous release Spontaneous release was determined by counting the supernatant fluid from target cells in medium without effector cells, while total incorporation was measured by counting a suspension of the target cells in medium. Antibody-dependent cellular cytotoxicity assay. Human group O, type R1R2 erythrocytes were labelled with 100 ^Ci of sodium "chromate solution per 5 X 10* cells for 1-5 h. After washing, 5 X 104 51Cr-labelled erythrocytes in 0-1 ml were sensitized with 0-05 ml Ripley anti-CD serum (diluted 1:400) for 1 h in round-bottom microtitre wells. Various effector cell populations (5 X 105 in 0-05 ml) were then added to triplicate wells and incubated for 18 h at 37 °C. Supernatant fluid was then collected and counted in a gamma counter. The % Sp.R. was calculated in the same manner as the NK assay. Spontaneous release was determined by counting the supernatant fluid from slCr- labelled erythrocytes incubated with effector cells in the absence of specific antiserum, while total incorporation was measured by counting a suspension of 51Cr-labelled erythrocytes in medium. Granulocyte fraction A PBS-buffy coat cell suspension was layered over an equal amount of Ficoll-Hypaque and centrifuged at 800 £ for 20 min. The pellet (which was rich in granulocytes), was washed with PBS and fixed for SEM (fraction 10). 32 C. M. Payne and others

Scanning electron microscopy procedures Preparative techniques. All cellular fractions were fixed in an identical manner. One drop of a concentrated cellular suspension (approx. 2 X 106 cells) was placed into a beaker containing 10 ml of 1 % glutaraldehyde made up in 0-1 M-phosphate buffer (pH7-2). The fixative was allowed to come to room temperature before this initial fixation step. The cells were fixed in suspension for 1 h at room temperature and then placed in the refrigerator and fixed overnight at 4°C. The cells were allowed to settle overnight onto double-sticky tape or placed onto coverelips that were pre-coated with poly-L-lysine (Tsutsui, Kumon, Ichikawa & Tawara, 1976). The attached cells were rinsed in 0-1 M-phosphate buffer, post-fixed for 1 h with 1 % osmium tetroxide made up in 0-1 M-phosphate buffer, dehydrated through a graded series of ethanols and critical-point dried from liquid carbon dioxide. The specimens were then sputter-coated with a thin layer of gold and examined at 40 kV in a JEOL 100 CX II TEMSCAN electron microscope. Quantitation of the surface features of lymphoid cells. The surface features of 100 cells from fractions 1-4, 6 and 7 were scored under the TEMSCAN. A total of 14 LGL/EAhu rosettes (fraction 5) and 13 LGL/KS62 conjugates (fraction 9) were photographed and scored. The surface features of the majority of the lymphoid cells were easily evaluated and had either a complex micTOvillous surface (CMS) pattern or a sparse microvillous surface (SMS) pattern. A CMS pattern was defined as an exposed cellular area that was covered with microvilli and microridges with very little cell surface visible. An SMS pattern was defined as an exposed cellular area that contained relatively few micTOvilli and microridges, making visible large areas of the cell surface. In the minority of cells (< 5 %) in which the surface features appeared equivocal, the following criteria for scoring were used. If more than half of the exposed cellular area was covered with microvilli and microridges with very little cell surface visible, the cell was considered to have a CMS pattern.

RESULTS Surface features of enriched lymphoid fractions LGLs enriched by rosette formation with sensitized erythrocytes. A total of 97 % of the cells isolated in fraction 7 ranged from 4*1 to 5-0 pan. in diameter and had a predominantly microvillous surface pattern (Table 1), and 3% of the cells were distinctly larger (>6-0/UM), had a ruffled surface and are believed to be contaminating monocytes. The enriched LGLs revealed a variation in the density of surface projections. A total of 65 % had a CMS pattern that consisted of numerous microvilli and microridges (Fig. 1), and 25% of the cells had an SMS pattern (Fig. 2). A microridge is defined as a narrow, strip-like surface projection whose base is approximately two to ten times larger than the base of a microvillous projection. In some cells the microvilli and microridges were confluent so that the cell surface at the base of the projections was not visible (lymphoid cell in Fig. 3). A small percentage of cells (7%) had microvilli, microridges and a few surface ruffles (Table 1). These cells were distinctly smaller than monocytes (Figs 3, 4). Non-LGLs {Fcy{ — ) lymphoid cells). The cells isolated in fraction 6 ranged from 4-1 to 4-8/^1 in diameter. No contaminating monocytes were seen. A total of 82% of these Fcy(-) cells had an SMS pattern and 18% had a CMS pattern (Table 1). No surface ruffles were seen on these small lymphoid cells. Quantitative SEM study of natural killer cells 33

Table 1. Quantitation of surface features of monocyte-depleted, enriched lymphoid fractions

Surface features of cells (%)

Enriched fractions SMSf SMS/SRJ CMS§ CMS/SR SR LGLs (F^ (+) enriched)* 25 2 65 5 3 Non-LGLs (Feu, (-) enriched)|| 82 0 18 0 0

• Cell fraction 7 (sec Materials and Methods). f Smooth microvillous surface pattern. \ Surface ruffles. § Complex microvillous surface pattern. [| Cell fraction 6 (see Materials and Methods).

Experimental variables present during LGLfractionation: effect on lymphoid surface features. There appeared to be a predominantly CMS pattern associated with the LGL- enriched fraction. It is apparent that before fraction 7 was obtained, the cells were exposed to different temperature conditions and isolation media. The effect of temperature alone was ascertained by incubating an unfractionated buffy coat preparation at 25 °C (fraction 1) and 37 °C (fraction 2), respectively, and quantitating the surface features of the leucocytes in this mixed population that were in the same size range as the purified LGLs. A striking effect of temperature was noted. A total of 70 % of the unfractionated lymphoid cells had a CMS pattern when the cells were incubated

Figs 1-4. Composite scanning electron micrographs of human leucocytes. X 7680. Fig. 1. Typical LGL enriched by rosette formation with sensitized erythrocytes (fraction 7). This cell has a complex microvillous surface pattern. The microvilli, some of which are branched, are abundant. Microridges are also present on the surface but no surface ruffles are present. Fig. 2. Typical lymphoid cell obtained from the unfractionated buffy coat preparation that was incubated at 37 °C for 1-5 h before fixation (fraction 2). This cell has a sparse microvillous surface pattern in which large areas of the cell surface are visible. Microvilli and microridges can be seen but no surface ruffles are present. Fig. 3. Comparison of the size and surface features of four distinct leucocytes obtained from an unfractionated buffy coat preparation (fraction 1). The platelet (p) is the smallest and has surface microvilli and ridge-like profiles. The lymphoid cell (/) is next in size and displays a highly complex microvillous surface. The microvilli and microridges are so confluent that the cell surface at the base of the projections is not visible. The monocyte (m) is larger than the lymphoid cell and smaller than the granulocyte {g). The monocyte has numerous surface ruffles, microridges and larger broad-based, ridge-like profiles. The granulocyte is largest and displays typical microvilli, branched microvilli, microridges and broad-based, ridge-like profiles. Fig. 4. Typical monocyte obtained from the mononuclear fraction after Ficoll- Hypaque density gradient centrifugation (fraction 3). The delicate surface ruffles are most apparent. 34 C. M. Payne and others at 25 °C (Table 2). On the other hand, 87 % of the unfractionated lymphoid cells had an SMS pattern when incubated at 37 °C. The purification of LGLs also involved centrifuging the LGL/EAhu rosettes through Ficoll-Hypaque (FH), warming the rosettes to 37°C to remove the adherent sensitized erythrocytes and centrifuging again through FH to obtain the enriched LGLs at the interface. To ascertain the effects of FH on surface features,

Figs 1-4. For legend see p.33 Table 2. Quantitation of sutface features of lymphoid cells in van'ous cell fractions obtained from nonnal human blcwd

Experimental variables Surface features

Exposure to 8 Lymphoid % Lymphoid Exposure Exposure IgG-coated Exposure to Exposure to cells with cells with Cell fractions Fraction* to FHt to LSRI erythrocytes nylon wool Percoll SMS patterns CMS patternll "B3 P. Buffy coat ii 25 "C 1 - - - - - 30 70 =. 37 "C - - - - - 87 13 E 2 Co Mononuclear 37"C, FH, fix 3 + - - - - 49 5 1 B 37OC, FH, wash, 4 + - - - - 38 62 37°C (3 h), fix k5 LGL/EAh, rosettes 5 + + + - - 21 79 % 3 LGLs (Fey (+) enriched) 7 + + + - - 28 72 i Non-LGLs 6 + + + - - 82 18 (Fey (-) enriched) 5 5.c LGL/K562 aggregates 9 + - - + + 20 80 3 Y See flow chart and Materials and Methods for details of fractionation procedures. 5 t Ficoll-Hypaque (lymphocyte separating medium). 2 1Lymphocyte separator reagent. Sparse microvillous surface. R Complex microvillous surface. 36 C. M. Payne and others an unfractionated buffy coat preparation was incubated for 15 h at 37 °C, layered over FH and then centrifuged to obtain a mononuclear fraction. The percentage of lymphoid cells with a CMS pattern increased four times over the cells that had been incubated only at 37 °C (Table 2). When cells were washed free of FH and incubated at 37 °C for 3 h, the percentage of cells with a CMS pattern increased almost fivetime s over the cells that had been incubated only at 37 °C (Table 2). This was interpreted to mean that FH had a direct effect on lymphoid cell surfaces. Decreasing the temperature and using Ficoll—Hypaque, however, did not induce any unique surface projections, but only increased the frequency of the microvilli and microridges already present on these small leucocytes.

Surface features ofLGL effector cells in antibody-dependent and natural killing Antibody-dependent cytotoxic lymphoid cells (K cells). The surface features of the effector cells in an antibody-dependent cytotoxicity assay (fraction 5) were virtually identical to those of enriched LGLs (fraction 7) (Figs 5-12). A total of 79 % of the effector cells observed had a CMS pattern with numerous microvilli and microridges, comparable to the control cells at room temperature (Table 2). Since the LGL/ EAhu rosette incubation is performed at room temperature, one cannot evaluate the direct effect, if any, of immunoglobulin G (IgG)-coated cells on surface activation. Effector cells with a CMS pattern (Figs 5-8) and those with an SMS pattern (Figs 9-12) both participated in erythrocyte deformation. The effector cells interacted with the erythrocytes by the use of microvilli (Figs 6, 12), microridges (Figs 6-8) and lamellipodia (pseudopodal-type extensions, Fig. 10). No cells with surface ruffles or broad-based, ridge-like profiles formed rosettes with the sensitized erythrocytes. Natural killer (NK) cells. The surface features of the effector cells in fraction 9 were virtually identical to those of the enriched LGLs in fraction 7 (Figs 13-16). A total of 80% of the effector cells observed had a CMS pattern with numerous microvilli and microridges (Fig. 14), comparable to that of the control room-

Figs 5-8. Composite scanning electron micrographs of LGL/EAhu rosettes (fraction 5). Fig. 5. Low-magnification view showing very little deformation of the surrounding erythrocytes. X 4800. Fig. 6. Higher-magnification view of the surface of the same central rosette-forming cell shown in Fig. 5. Abundant microvilli and microridges are present. A couple of microvilli and the upper edge of one microridge can be seen touching an erythrocyte (left side of micrograph). X 14400. Fig. 7. High-magnification view of a different rosette. The central rosette-forming cell displays a complex microvillous surface pattern. Microridges and microvilli are abundant. The surrounding erythrocytes are greatly misshapen. MicToridges are observed to participate in erythrocyte attachment (lower right of micrograph). X 7680. Fig. 8. High-magnification view of a different rosette. Short stubby microvilli, branched microvilli and microridges are apparent. Two erythrocytes have broken away from the lymphoid surface to reveal the greatly deformed erythrocyte surface. At the point of previous attachment to the central rosette-forming cell, the erythrocyte surfaces have a puckered appearance. The lateral surfaces of two microridges are observed to participate in erythrocyte attachment (middle left of micrograph). X 9600. Quantitative SEM study of natural killer cells 38 C. M. Payne and others

Figs 9-12. For legend see p.40 Quantitative SEM study of natural killer cells

Figs 13-16. For legend see p.40 40 C. M. Payne and others temperature cells (Table 2). Effector cells with both a CMS pattern (Figs 13-15) and an SMS pattern (Fig. 16) were found attached to the K562 target cells. Effector cells attached to the target cells by both microvilli and microridges (Fig. 14). No cells containing surface ruffles or broad-based ridge-like profiles were found interacting with the target cells.

Surface features of granulocyte-enriched fractions The surface features of granulocytes (fraction 10) were quite distinctive (Fig. 3) and contained numerous broad-based ridge-like profiles in addition to microridges, microvilli and surface ruffles. The surface ruffles were never as extensive as monocytes (large cells in fractions 3, 4) (Fig. 4) and the surface never had confluent surface projections as observed with unfractionated lymphoid cells (fractions 1—4, Figs 2, 3), or enriched LGLs (fractions 7,9). The diameter of the granulocytes ranged between 5-4 and 7-

Figs 9-12. Composite scanning electron micrographs of LGL/EAhu rosettes (fraction 5). Fig. 9. Low-magnification view comparing the greatly deformed adherent erythrocytes with the normal biconcave shape of a non-adherent, non-rosetted erythrocyte. Two erythrocytes have broken away from the lymphoid surface and large cavities are revealed at the point of previous cell attachment. A large bleb is seen on the surface of the detached erythrocyte (bottom right of micrograph). X 5760. Fig. 10. Higher-magnification view of the same central rosette-forming cell shown in Fig. 9. Short, stubby microvilli, branched microvilli and microridges are present. In addition, lamellipodia are evident on this lymphoid cell and are seen adhering to an erythrocyte surface (upper left of micrograph). X 14400. Fig. 11. Low-power view of a different rosette. Some of the erythrocytes (top left) have a crenated appearance with multiple surface blebs present. One erythrocyte (middle right) is being attacked by two different lymphoid cells. Note the deformation of this erythrocyte at the left and bottom surfaces. X 4480. Fig. 12. Higher-magnification view of the same central rosette-forming cell as shown in Fig. 11. Long microvilli can be seen extending to two different erythrocytes (bottom left). X7360. Figs 13-16. Composite scanning electron micrographs of LGL/K562 conjugates (fraction 9). Fig. 13. Low-power view of an effector cell attacking a larger KS62 target cell. The target cell surface has many broad-based, ridge-like profiles and a number of long and short microvilli. 2h incubation at 37 °C. X 4800. Fig. 14. Higher-magnification view of the same effector cell as shown in Fig. 13. This lymphoid cell has a highly complex microvillous surface pattern that consists of microvilli and numerous microridges. The effector cell adheres to the target cell by means of both microvilli and the ridge-like profiles. X 9600. Fig. 15. Low-power view showing an effector cell simultaneously attacking two target cells. The effector cell has a complex microvillous surface pattern and the target cells show surface blebs. 1 h incubation at 37 °C. X 4800. Fig. 16. Low-power view of a K562 target cell with three adherent lymphoid cells. The two lymphoid cells at the bottom have a complex microvillous surface pattern whereas the lymphoid cell at the top has a sparse microvillous surface pattern. 2 h incubation at 37 °C. X384O. Quantitative SEM study of natural killer cells 41

Table 3. Characteristics and cytotoxic activity of enriched lymphoid populations

Enriched lymphoid populations

Fraction 6 Fraction 7 Fraction 8

Cell types (%) Non-LGLs* 97 21 26 LGLs 3 73 70 Monocytes 0 2 3 Polymorphonuclear leucocytes 0 3 0 Monoclonal antibody (% positive) Leu-1 92 NDt 21 Leu-7 3 53 52 Leu-11 0 88 81 Cytotoxic activity (% SpR)t NK (10:1, E:T§ ratio) 3 54 66 ADCCll 0 ND 53

• Large granular lymphocytes. fNot done. | Percentage specific slCr release. 5 Effector :target. [| Antibody-dependent cellular cytotoxicity.

Characteristics and cytotoxic activity of enriched lymphoid populations studied by scanning electron microscopy

To ensure that the SEM features of the surfaces of the Fcy( + )-enriched lymphoid fractions were indeed those of the effector cells responsible for cell killing in the immune system, functional assays, surface marker analysis and an assessment of the percentage of LGLs by light microscopy were performed. The lymphoid cells enriched by rosette formation with sensitized erythrocytes (fraction 7) and the Percoll-enriched lymphoid cells (fraction 8) were both characterized by a high percentage of LGLs and a high reactivity with Leu-7 and Leu-11 monoclonal antibodies (Table 3). The Leu-11 antibody stained approximately 30% more cells than the Leu-7 antibody in both fractions. The Fcy(—) cells (fraction 6) were almost devoid of LGLs, did not react with Leu-11 antibody and reacted minimally with Leu-7 antibody. The majority or these cells were characterized as Leu-1 ( + ), non-granular lymphocytes. Significant degrees of cell killing as determined by % 51Cr release, were observed with both the Fcy( + )-enriched LGLs (fraction 7) and the Percoll-enriched LGLs (fraction 8). LGLs enriched by rosette formation with sensitized erythrocytes (Fcy (+ )-enriched lymphoid fraction; fraction 7) lysed K562 target cells with a similar degree of 51Cr release as LGLs enriched using Percoll gradients (fraction 8). The latter also effectively lysed sensitized erythrocytes in an antibody-dependent cytotoxic assay system. The non-LGL (Fcy (—); fraction 6) population was not active either in the NK or antibody-dependent cytotoxicity assays. 42 C. M. Payne and others

DISCUSSION Previous work from our laboratory has indicated that the vast majority (89 %) of human lymphoid cells that have Fcy receptors for cytophilic antibody (non-T, non-B lymphoid cells) have unique cytoplasmic inclusions called parallel tubular arrays (PTAs) (Payne & Glasser, 1981). The specific nature of these inclusions can be ascertained only by using transmission electron microscopy. PTAs have not been observed in other cell types, including monocytes and granulocytes, and most prob- ably represent the ultrastructural correlate of the azurophilic granules that characterize the LGLs at the light-microscopic level (Timonen et al. 1981). The PTA is therefore a marker for lymphoid cells that have Fcy receptors for cytophilic antibodies and includes K cells (Payne & Glasser, 1981), NK cells (Babcock & Phillips, 1983; Burns et al. 1982; Henkart & Henkart, 1981) and Ty cells (Payne et al. 1983). Although this unique killer cell population appears by TEM to represent a morphologically homogeneous population, it is possible that: (1) LGLs may all have a lymphoid origin; (2) LGLs may all have a myeloid origin; (3) some subpopulations of LGLs may be lymphoid in origin and others myeloid in origin; or (4) the LGL population may be unique and represents a distinct leucocyte with a separate haematopoietic lineage. Work from our laboratory has emphasized the uniqueness of this cell population. They are steroid-resistant (Payne & Glasser, 1978) and can phagocytoze complement-coated bacteria (Payne & Nagle, 1980), two features that are not characteristic of the majority of lymphocytes. Since lymphocytes, monocytes and granulocytes have unique surface features, the purpose of the present study was to determine by SEM if the LGL population appears predominantly lymphoid or myeloid in nature, and whether it is a mixed cell population or possesses unique surface features. This study for the first time compares the surface features of purified LGLs with Fcy(—) lymphoid cells, granulocytes and monocytes using identical SEM fixation procedures. The surface features of human LGLs interacting with K562 target cells in an NK assay system are also described for the first time. We have determined that the surface features of these LGL effector cells are virtually identical and seem to represent a morphologically homogeneous population. They all have a predominantly CMS pattern that consists of abundant microvilli and microridges characteristic of lymphocytes (Abugaber et al. 1981; Alexander & Wetzel, 1975; Dantchev & Belpomme, 1977; Hattori, 1972; Kelly & Nockolds, 1977; Newell et al. 1976; Polliack, 1978; Roath et al. 1978; Schneider, Pockwinse & Billings-Gagliardi, 1978; Thornthwaite ef a/. 1975). No effector cells were observed to have predominantly surface ruffles characteristic of monocytes, or broad-based ridge-like profiles characteristic of granulocytes. The results of our SEM studies are contrary to the observations of Horwitz & Bakke (1984), who state that the membrane features of LGLs (L cells) resemble those of monocytes (possess lamellipodia and filopodia) and not lymphocytes (few short filopodia). It was not specified, however, whether light microscopy or SEM was used in this evaluation. We have determined that the CMS pattern was not a unique feature, however, of LGLs. Room temperature Quantitative SEM study of natural killer cells 43 and Ficoll-Hypaque can independently induce a CMS pattern in the majority of unfractionated lymphoid cells. These experimental variables, however, did not induce any unique surface projections, but only increased the frequency of the microvilli and microridges already present on these small leucocytes. Since LGLs comprise only 8-25 % of the lymphoid cell population from normal blood (Payne & Glasser, 1981) it is obvious that other lymphoid populations can display a CMS pattern. Temperature effects may help to explain the difference in surface projection density reported by different investigators (Roath et al. 1978). A CMS pattern or abundant microvilli was consistently observed on lymphocytes that were maintained and fixed at room temperature without the use of silver membrane substrates (Augaber et al. 1981; Alexander & Wetzel, 1975; Hattori, 1972; Kelly & Nockolds, 1977; Newell et al. 1976; Schneider et al. 1978; Thornthwaite et al. 1975). Dantchev & Belpomme (1977) suggested that cells with a completely smooth or smooth undulated surface may be characteristic of the null cell population but stated that this needed experimental confirmation using purified null cells. We could not confirm their suspicions but have determined instead that the LGL population (which corresponds to their purified null cells) has a highly villous surface. The smooth surfaces of the type 1 and type 2 cells observed by Dantchev & Belpomme (1977) may have been caused by the toxic effects of Flotronic silver membranes (Alexander & Wetzel, 1975), which they used and, or, by the short fixation times. Kelly & Nockolds (1977) reported that only 22-31 % of the K cells in their EA rosettes had a complex villous surface. On the other hand, we found that 79 % of the effector cells in our LGL/EAhu rosettes had a CMS pattern similar to the results of others (Alexander & Henkart, 1976; Biberfeld et al. 1975; Inglis et al. 1975; Perl- mann, 1976). These differences can also be explained by differences in temperature. Kelly & Nockolds (1977) incubated their rosettes at 37 °C whereas we incubated ours at room temperature. Newell et al. (1976) state that little difference in surface morphology was exhibited between lymphocytes fixed at 37 °C and those fixed at 22 °C. We have demonstrated, however, a dramatic difference in surface morphology between lymphocytes incubated at 37 °C and those incubated at 25 °C. These differences may be a reflection of the length of incubation at these respective temperatures since our incubation times were three times longer than those of Newell et al. (1976). Kelly & Nockolds (1977) state that the presence of numerous microvilli on lymphocytes is a true reflection of the morphology of lymphocytes in their natural, fully suspended state in blood. We feel that the SMS pattern is more characteristic of the in vivo state of lymphocytes, since an SMS pattern is produced at body temperature (37 °C), whereas a CMS pattern is produced at room temperature (25 °C). This report also describes for the first time the surface features of human LGLs interacting with K562 target cells in an NK assay system. The CMS pattern of these human effector cells is similar to the highly villous surface on mouse NK cells described by Roder, Kiessling, Biberfield & Andersson (1978). These investigators also noted a temperature effect on the surface features of NK cells. They noted that the villous nature of the cells became more obvious in preparations incubated at 20 °C 44 C. M. Payne and others instead of 37 °C. In our LGL/K562 system, the effector cells had a CMS pattern, although the conjugates were incubated at 37 °C, which produces an SMS pattern in unreacted cells. This may indicate surface activation by specific target antigens or may reflect differences between human and mouse NK cells. Roder et al. (1978) similarly concluded that mouse NK cells are not morphologically distinguishable from normal lymphocytes. Similarities in surface features do not, however, prove that LGLs (NK and K cells) have the same haematopoietic origin as lymphocytes. Other studies such as the evaluation of colony-forming-units derived from bone marrow precursor cells will determine if the LGL is really 'a phagocyte in lymphocyte's clothing' (Babior & Parkinson, 1982). The enriched killer cell populations that were analysed by SEM were further characterized in this study using independent criteria to ensure the homogeneity and killer cell nature of these cells. The majority of the killer cells studied by SEM were classified as LGLs (Timonen et al. 1981) by light microscopy, stained with Leu-11 antibody (the most specific monoclonal antibody for NK cells; Lanier, Phillips, Warner & Babcock, 1983) and exhibited both antibody-dependent cytotoxicity and NK activity (Kay, Bonnard, West & Herberman, 1977; Landazuri, Silva, Alvarez & Herberman, 1979; Parrillo & Fauci, 1978). Much evidence has now accumulated to suggest that human LGLs represent a distinct leucocyte population, although the haematopoietic precursor cell is not known. They are best characterized by the presence of numerous azurophilic granules (by light microscopy), have unique cytoplasmic inclusions called parallel tubular arrays (by TEM), possess Fcy receptors (using EA-rosetting techniques), possess Leu-7 and Leu-11 surface antigens (using monoclonal antibody techniques) and functionally exhibit both antibody-dependent cytotoxicity and NK activity. We have now shown that human LGLs have a complex microvillous surface pattern.

We thank Dr Jack M. Layton (Head of the Department of Pathology) whose encouragement, support and advice has made this work possible. We also thank Mrs Katie Eckinger for typing the manuscript. This work was supported in part by the Southwestern Clinic and Research Institute, Tucson, Arizona.

REFERENCES ABUGABER, A. A., LALAGUE, E. D., ROY, A., CHAMPAGNE, M. R. & COUSINEAU, G. H. (1981). The use of various buffers in the preparation of human lymphocytes for SEM observation, J. Micmsc. 122, 59-64. ALEXANDER, E. & HENKART, P. (1976). The adherence of human Fc receptor-bearing lymphocytes to antigen-antibody complexes. II. Morphologic alterations induced by the substrate. J. exp. Med. 143, 329-347. ALEXANDER, E. L. & WETZEL, B. (1975). Human lymphocytes: Similarity of B and T cell surface morphology. Science 188, 732—734. BABCOCK, G. F. & PHILLIPS, J. H. (1983). Human NK cells: light and electron microscopic characteristics. Surv. Immun. Res. 2, 88-101. BABIOR, B. M. & PARKINSON, D. W. (1982). The NK cell: a phagocyte in lymphocyte's clothing? Nature, Lond. 29S, 511. Quantitative SEM study of natural killer cells 45 BIBERFELD, P., WAHLIN, B., PERLMANN, P. & BIBERFELD, G. (1975). A plaque technique for assay and characterization of antibody-dependent cytotoxic effector (K) cells. Scand. J. Immun. 4, 859-864. BURNS, E. R., ZUCKER-FRANKLIN, D. & VALENTINE, F. (1982). Cytotoxicity of natural killer cells: correlation with emperipolesis and surface enzymes. Lab. Invest. 47, 99—107. DANTCHEV, D. & BELPOMME, D. (1977). Critical study of the mononuclear leukocyte morphology based on scanning electron microscopy in normal subjects and in patients with lymphoid or monocytoid proliferative disorders. Comparison with the T, B or null cell membrane phenotypes. Biomedtdne 26, 202-222. FERRARINI, M., CADONI, A., FRANZI, A. T., GHIGLIOTTI, C, LEPRINI, A., ZICCA, A. & GROSSI, C. E. (1980). Ultrastrucrure and cytochemistry of human peripheral blood lymphocytes. Similarities between the cells of the third population and TG lymphocytes. Eur. J. Immun. 10, 562-570.- FITZGERALD, P. A., EVANS, R., KIRKPATRICK, D. & LOPEZ, C. (1983). Heterogeneity of human NK cells: comparison of effectors that lyse HSV-1 infected fibroblasts and K562 erythroleukemia targets. J. Immun. 130, 1663-1667. GALLI, S. J., DVORAK, A. M., ISHIZAKA, T., NABEL, G., DER SIMONIAN, H., CANTOR, H. & DVORAK, H. F. (1982). A cloned cell with NK function resembles basophils by ultrastructure and expresses IgE receptors. Nature, Land. 298, 288-290. HATTORI, A. (1972). Scanning electron microscopy of human peripheral blood cells. Ada haemat. jap. 35, 457-482. HENKART, M. P. & HENKART, P. A. (1981). Lymphocyte mediated cytolysis as a secretory phenomenon. Adv. exp. Med. Biol. 146, 227. HERBERMAN, R. B. & ORTALDO, J. R. (1981). Natural killer cells: Their role in defenses against disease. Science 214, 24-30. HORWITZ, D. A. & BAKKE, A. C. (1984). An Fc receptor-bearing, third population of human mononuclear cells with cytotoxic and regulatory function. Immun. Today 5, 148—153. HORWITZ, D. A., NIAUDET, P., GREAVES, M. F., DORLING, J. & DETEIX, P. (1978). Surface markers and electron microscopy of human blood L cells: A comparison with T and B lymphocytes. J. Immun. 121, 678-684. INGLIS, J. R., PENHALE, W. J., FARMER, A., IRVINE, W. J. & WILLIAMS, A. E. (1975). Observa- tions on the antibody-dependent cytotoxic cell by scanning electron microscopy. Clin. exp. Immun. 21, 216-225. JULIUS, M. H., SIMPSON, E. & HERZENBERG, L. S. (1973). A rapid method for the isolation of functional thymus-derived murine lymphocytes. Eur.jf. Immun. 3, 645-649. KAY, H. D., BONNARD, G. D., WEST, W. H. & HERBERMAN, R. B. (1977). A functional comparison of human Fc-receptor-bearing lymphocytes active in natural cytotoxicity and antibody- dependent cellular cytotoxicity. J. Immun. 118, 2058-2066. KAY, H. D. & HORWITZ, D. A. (1980). Evidence by reactivity with hybridoma antibodies for a probable myeloid origin of peripheral blood cells active in natural cytotoxicity and antibody- dependent cell-mediated cytotoxicity. J. clin. Invest. 66, 847-851. KELLY, G. E. & NOCKOLDS, C. E. (1977). Morphological differences between subpopulations of human lymphocytes revealed by scanning electron microscopy. Scand. J. Haemat. 19, 172—184. KUMAGAI, K., ABO, T., SEKIZAWA, T. & SASAKI, M. (1975). Studies of surface immunoglobulins on human B lymphocytes. I. Dissociation of cell-bound immunoglobulins with acid pH or at 37°C7- Immun. 115, 982-987. LANDAZURI, M. O., SILVA, A., ALVAREZ, J. & HERBERMAN, R. B. (1979). Evidence that natural cytotoxicity and antibody-dependent cellular cytotoxicity are mediated in humans by the same effector cell populations. J. Immun. 123, 252-258. LANIER, L. L., LE, A. M., PHILLIPS, J. H., WARNER, N. L. & BABCOCK, G. F. (1983). Sub- populations of human natural killer cells defined by expression of the Leu-7 (HNK-1) and Leu- 11 (NK-15) antigens. J. Immun. 131, 1789-17%. LOHMANN-MATTES, M. L., DOMZIG, W. & RODER, J. (1979). Promonocytes have the functional characteristics of natural killer cells, jf. Immun. 123, 1883-1886. NEIGHBOUR, P. A., HUBERMAN, H. S. & KRESS, Y. (1982). Human large granular lymphocytes and natural killing: ultrastrucrural studies of strontium-induced degranulation. Eur.jf. Immun. 12, 588-595. 46 C. M. Payne and others NEWELL, D. G., ROATH, S. & SMITH, J. L. (1976). The scanning electron microscopy of normal human peripheral blood lymphocytes. Br.jf. Haemat. 32, 309-316. OEHL, S., SCHAEFER, K. W., BOECKER, W. R., KALDEN, J. R. & PETER, H. H. (1977). Human peripheral blood "null-lymphocytes", a highly enriched pool for granulocytic stem cells (CFU- C). Z. ImmunForsch. exp. Ther. 152, 423-430. ORTALDO, J. R., SHARROW, S. O., TIMONEN, T. & HERBERMAN, R. B. (1981). Determination of surface antigens on highly purified human NK cells by flow cytometry with monoclonal antibodies. J. Immun. 127, 2401-2409. PARRILLO, J. E. & FAUCI, A. S. (1978). Comparison of the effector cells in human spontaneous cellular cytotoxicity and antibody-dependent cellular cytotoxicity: Differential sensitivity of effector cells to in vivo and in vitro corticosteroids. Scand.J. Immun. 8, 99-107. PAYNE, C. M. & GLASSER, L. (1978). The effect of steroids on peripheral blood lymphocytes containing parallel tubular arrays. Am.J. Path. 92, 611—618. PAYNE, C. M. & GLASSER, L. (1981). Evaluation of surface markers on normal human lymphocytes containing parallel tubular arrays: a quantitative ultrastructural study. Blood 57, 567—573. PAYNE, C. M., GLASSER, L., FIEDERLEIN, R. & LINDBERG, R. (1983). New ultrastructural observations: Parallel tubular arrays in human Tr lymphoid cells. J. Immun. Meth. 65, 307-317. PAYNE, C. M. & NAGLE, R. B. (1980). Complement receptors on normal human lymphocytes containing parallel tubular arrays. Am.J. Path. 99, 645—666. PERLMANN, P. (1976). Cellular immunity: antibody-dependent cytotoxicity (K-cell activity). Clin. Immunobiol. 3, 107-132. POLLIACK, A. (1978). Surface morphology of lymphoreticular cells: review of data obtained from scanning electron microscopy. Recent Res. Cancer Res. 64, 66-93. ROATH, S., NEWELL, D., POLLIACK, A., ALEXANDER, E. & LIN, P-S. (1978). Scanning electron microscopy and the surface morphology of human lymphocytes. Nature, Land. 273, 15-18. RODER, J. C, KIESSLING, R., BIBERFELD, P. & ANDERSSON, B. (1978). Target-effector interaction in the natural killer (NK) cell system. II. The isolation of NK cells and studies on the mechanism of killing. J. Immun. 121, 2509—2517. SCHNEIDER, G. B.. POCKWINSE, S. M. & BILLINGS-GAGLLARDI, S. (1978). Morphological changes in isolated lymphocytes during preparation for SEM: A comparative TEM/SEM study of freeze drying and critical-point drying. Scan. Electron Microsc. II, 77-84. SMIT, J. W., BLOM, N. R., VAN LUYN, M. & HALIE, M. R. (1983). Lymphocytes with parallel tubular structures: morphologically a distinctive subpopulation. Blut 46, 311-320. THORNTHWAITE, J. T., THORNTHWAITE, B. N., CAYER, M. L., HART, M. A. & LEIF, R. C. (1975). A new method for preparing cells for critical point drying. Scan. Electron Microsc. I, 387-392. TIMONEN, T., ORTALDO, J. R. & HERBERMAN, R. B. (1981). Characteristics of human large granular lymphocytes and relationship to natural killer and K cells. J. exp. Med. 153, 569—582. TSUTSUI, K., KUMON, H., ICHIKAWA, H. & TAWARA, J. (1976). Preparative method for suspended biological materials for SEM by using of polycationic substance layer. J. Electron Microsc. 25, 163-168. WINCHESTER, R. J., Fu, S. M., HOFFMAN, T. & KUNKEL, H. G. (1975). IgG on lymphocyte surfaces: technical problems and the significance of a third cell population. J. Immun. 114, 1210-1212. ZARLING, J. M., CLOUSE, K. A., BIDDISON, W. E. & KUNG, P. C. (1981). Phenotypes of human natural killer cell populations detected with monoclonal antibodies. J. Immun. 127, 2575-2580. {Received 18 October 1984 -Accepted 3 January 1985)