Mononuclear-cell infiltration in ovarian cancer. I. Inflammatory-cell infiltrates from tumour and ascites material.

Malignant effusions and tumour tissue obtained at surgery provided material for a study of the prognostic value of the various inflammatory cells in the prognosis of human ovarian cancer. Ascitic fluids predominantly contained inflammatory cells; tumour cells, both singly and in clusters, were a minor component. Tumour cells were usually in excess in dispersed solid material. Some patients had significant proportions of lymphocytes and macrophages in their solid tumour, and these patients invariably responded to therapy. Sedimentation-velocity separation at unit gravity provided tow populations of inflammatory cells. One consisted of mononuclear cells similar in size to those in the patients blood: the other consisted of one or more large macrophage populations, distinct in morphology and enzymatic markers from both blood monocytes and each other. T lymphocytes were enriched in ascites fractions (78%) and in the tumour-derived mononuclear fraction (71%) compared to patient blood (60%). The T-cell subset characterized by ANAE reactivity was markedly depleted in the tumour-infiltrating fraction (17%) compared to patient blood (62%) or patient blood (51%). Esterase-positive monocyte-like cells were more frequent in the tumour-infiltrating fraction (17%) than ascites (7%) or blood (12%). B lymphocytes were infrequent in solid tumours and difficult to assess in ascites. Histiocyte-like macrophages were present in the higher-velocity tumour-cell containing fractions of both solid and ascitic material. The variation in infiltrating cells between patients and between tumour and ascites of the same individual was marked.

and difficult to assess in ascites. Histiocyte-like macrophages were present in the higher-velocity tumour-cell containing fractions of both solid and ascitic material. The variation in infiltrating cells between patients and between tumour and ascites of the same individual was marked.
OVARIAN CANCER is the foremost gynaecoligical cause of death from cancer. Survival after surgery depends upon a multitude of factors, including residual tumour burden, degree of differentiation and staging (Newman et al., 1980). Little is known about the possible relevance of mononuclear-cell infiltration to patient survival. Although there is a consensus that, for a variety of cancers in situ, immunity may be important to clinical response (Underwood, 1974;loachim, 1976), there has yet to be a study which directly implicates immune-dependent cellular mechanisms in growth control of cancer. Recently, a number of reports have described attempts to characterize the immunological competence of lymphocytes or macrophages isolated from human cancers or ascites (Mantovani et al., 1979;Werkmeister et al., 1979;Totterman et al., 1980;Klein et al., 1980;Herberman et al., 1980;Vose, 1978).
We have initiated a programme designed to assess the relevance of mononuclearcell infiltration to the prognosis of patients with cancer of the ovary. In this first report, we will describe the methodology which we have used to fractionate both ascites and solid material into two basic fractions. In subsequent studies, these fractions will be used to compare and contrast the activity of blood-derived, ascites, and tumour-associated lymphocytes and macrophages.

MATERIAL AND METHODS
Human subjects.-Thirty-eight patients with histologically confirmed epithelial cancer of the ovary were used in this study. All but one were classified as Stages III or IV. Most patients provided both ascites and solid tumour material for analysis. In several cases, both ovarian and omental tumours were obtained. Because surgical removal of the ovary left only the extensive omental tumour to be treated, this tissue was used, when available, for analysis as it represented the target for further therapy.
Heparinized venous blood was collected from patients at surgery, when available. Healthy laboratory workers supplied control blood.
Enzymatic digestion of tumours.-The general methodology has recently been extensively described (Haskill, 1981a). Briefly, tumour material obtained directly from surgery was placed in cold sterile saline and maintained on ice until processing.
Scalpel and forceps were used to remove all normal tissue, and obvious heamorrhagic tissue. No attempt was made to dissect out small areas of necrosis contained within the mass of tumour, which was then gently diced into small 2mm cubes, washed and placed in a 100ml bottle containing a 15mm stirring magnet. To this, 0.14% collagenase and DNase (Sigma Type I, St Louis, Mo.) in calcium-and-magnesium-free EDTA-containing PBS were added to at least 4 times the volume of tumour. The mixture was slowly stirred for 15-20 min, and the resting solution gently removed. The process was usually repeated until the tumour was totally digested. Each fraction was washed in medium and held on ice until all the material was available. Usually the first digestion contained mostly damaged cells, and subsequent material contained predominantly viable cells ( > 95 %). When viability was poor, it was inevitably in the epithelial cells. Although cell aggregates were frequently seen in the tumour digests, they were almost entirely composed of tumour cells. Selective adherence of mononuclear cells to tumour cells was not significant. Sedimentation-velocity separation.-The methodology has been extensively discussed in a recent article (Haskill, 1981b) and is according to the method of Miller & Phillips (1969). 2 x 106 cells per ml were applied to the STAPUT chamber in medium without FCS. Then a 10-20% FCS gradient was formed under the sample. Sedimentation was carried out for 2 h at 4°C at 1 g. Two chamber sizes were used. If the cell number was < 40 x 106, a lOml/mm chamber was used. If the cell load was > 40 x 106, a 22ml/ mm chamber was used. Usually cells were applied to the chamber directly, without further fractionation (both tumour and ascites). When erythrocytes exceeded nucleated cells more than 3 :1, the entire cell suspension was first separated on Ficoll-Hypaque.
Cell counts were carried out on each fraction, velocities determined and fractions pooled as required. Cell viability exceeded 90%.
Adherence separation of macrophages.-Macrophages were separated from contaminating tumour cells and occasional plasma cells by adherence. Macrophage-rich fractions > 6 mm/h were allowed to adhere for 30 min at 37°C in 10% FCS RPMI 1640 medium. The cells were gently rinsed to remove nonadherent cells (checked by microscopy) and then the macrophages were removed by rinsing vigorously with medium. Viability exceeded 90% after this procedure. If tumour cells proved adherent under these conditions, macrophages were isolated on the basis of EA RFC properties.
EA RFC separation of macrophages.-In several cases, either the large macrophages initially isolated by sedimentation velocity were poorly adherent to plastic and/or the tumour cells were highly adherent. In this situation, rosette formation with IgG-coated erythrocytes (ox or sheep) for 5 min at 4°C, followed by rosette isolation over Percol (Becker & Haskill, 1980a) provided macrophages of similar purity to the adherence technique. Bound erythrocytes were lysed in 0.14% N HCI, 4°C for 1 min. We have never encountered EARFC clusters associated with clearly recognizable tumour cells.
Isolation of blood mononuclear cells.-Blood lymphocytes and monocytes were routinely isolated on Ficol-Hypaque (Boyum, 1968). Because few cells sedimented > 6-5 mmlh, blood mononuclear cells were not routinely separated at 1 g, as for ascites and tumour-derived inflammatory cells, which were said to be equivalent to blood mononuclear cells if they were in the same velocity range as blood mononuclear cells.
Immunocytochemical markers T lymphocytes: E RFC assay.-The percentage of E RFC was determined by incubating 2 x 105 test cells with 107 sheep erythrocytes previously treated with neuraminidase (Hoffman & Kunkel, 1976). The cells were incubated at 370C for 10 min and then centrifuged at 100 g for 5 min and left at 40C overnight. They were gently resuspended in 1% crystal violet and counted in a haemacytometer, or used for cytocentrifuge preparations without the crystal violet. Four or more bound erythrocytes were required for a rosette. At least 200 cells were counted per sample.
Macrophages: EA RFC assay.-Macrophages bearing FcR for IgG can be selectively detected by the EA RFC assay previously described, using coated sheep erythrocytes (SRBC) or ox erythrocytes coated with high dilutions of antibody (Korn et at., 1978). In each case 2 x 105 nucleated cells were combined with 107 coated erythrocytes, centrifuged at 4°C for 10 min (100 g) and immediately read as for the E RFC. Control experiments and those reported in Fig. 2a indicated that lymphocytes forming E RFC are not detected to a significant level by the use of coated SRBC, provided that the incubation time at 40C was restricted to 10 min, and the optimal dose of antibody was determined by titration as in our previous report (Korn et al., 1978). Recently, when only ox RBC have been used, we have not been able to detect differences in EA RFC numbers when either blood or tumour are used. B lymphocytes :SIg.-B cells were sometimes enumerated using fluoresceinated (Fabl)2 antihuman IgG, IgA, and IgM (Cappel Laboratories). B lymphocytes were readily detected in normal blood in this way by counting the proportion of cells with definite capping. Monocytes were negative. Monocyte-like cells and macrophages derived from ascitic and tumour material reacted intensely with the (Fabl)2 reagent, indicating the presence of either immune complexes or monomeric immunoglobulins. Because of the difficulty in clearly distinguishing small macrophages from larger B cells, no attempt was made to count ascitic B-cell levels routinely.
Cytochemical stains.-Acid naphthyl-acetate esterase (ANAE) and May-Grrunwald-Giemsa staining were carried out as described by Saksela et al. (1979a), acid phosphatase according to Burstone (1959) and myeloperoxidase according to Kaplow (1965). Cytocentrifuge preparations were immediately fixed in calcium acetate-formalin buffer, for all enzyme stains except on E RFC and EA RFC preparations. These were allowed to sit at room temperature overnight before fixation.

Proportions of inflammatory: tumour cells in ascites and tumours
Ascitic and tumour-derived cell suspensions were analysed for the relative proportions of lymphocytes, macrophages and tumour cells. Giemsa-stained cytocentrifuge preparations were scored by standard morphological criteria (Fig. 1) Survival data indicate that patients whose solid tumours contained a higher inflammatory response responded to chemo. therapy (Adriamycin, Alkeran). Granulocytes were rarely isolated from either source, but were commonly found in pleural effusions (not shown).
preponderance of inflammatory cells in ovarian ascites derived from patients not undergoing therapy. In contrast, lymphocytes and macrophages seldom exceeded the number of tumour cells recovered from enzymatically dispersed tumour tissue. Numerous attempts were made to quantitate inflammatory-cell proportions using rosette assays on unseparated disaggregated material. E RFC assays were found to be unreliable presumably because of interference with the cell debris associated with the unseparated material, the low lymphocyte proportions and the difficulty in pelleting the SRBCs in the proper levels with the lymphocytes when cells of various sizes were present together.
Relationship between inflammation and response to therapy A striking feature of this part of the study was that patients in whom inflammatory cells were common (ovary and/or omentum) inevitably responded to standard forms of chemotherapy (survival longer than 12 months) (Fig. 1). Because of the small number of patients involved it was not possible to subdivide them further according to the particular chemotherapy used, residual tumour burden or differentiation of the tumour. These data are provided as a preliminary indication that infiltration is somehow related to a response to chemotherapy, and that a thorough analysis of the functional properties of these inflammatory cells would be justified.

Sedimentation velocity separation of ascites and tumour material
Our separation approach was designed around three features. First, we did not wish to bias our experiments by selecting a particular inflammatory-cell type early in the analysis. Thus we chose a method which selected predominantly on the basis of size. Second, as we wished to compare systemic and in situ immune responses, tumour and ascites-infiltrating cells had to be isolated in as similar a way as possible to the same cells isolated from blood. Finally, these fractions would have to be free of tumour cells. Fortunately in few cases of highly differentiated adenocarcinoma (not used in this study) were tumour cells similar in velocity to the largest, most rapidly sedimenting inflammatory cells (< 6 mm/h). The results show that blood-like inflammatory cells can be isolated from ascites and tumour material, provided that velocity fractions are < 6 mm/h. By selecting cells sedimenting at <6 mm/h, we did not recover blast-like lymphocytes. However, as these were relatively infrequent and in any event were absent from both normal and patient blood, the comparison between blood, ascites and tumour was still valid. Rapidly sedimenting macrophages (> 6 mm/h) as distinct from blood monocytes, were subsequently isolated by adherence or EArosette depletion from the contaminating tumour cells in the high velocity regions.
Variation in inflammatory pattern between tumour and ascitic material of the patient and between different patients The variation in sedimentation-velocity profiles of ascites and tumour-cell suspensions, both within individual patients and between patients, was marked. One example is given in Fig. 3. The tumourderived material was composed mostly of tumour cells; T lymphocytes and macrophages were present in similar proportions. In the ascites, however, few tumour cells were detected. Most of the cells were large macrophages which could be separated into two classes on the basis of sedimentation velocity and myeloperoxidase and acid-phosphatase reactions Although the tumour contained macrophages similar in velocity to those present in the lower velocity region of the ascites sample (.< 8 mm/h), the cytochemical differences were marked. Twenty per cent of tumour-derived macrophages were weakly peroxidase-positive, whereas 60% of ascitic macrophages stained intensely. Although tumour macrophages reacted with the acid-phosphatase reagent after 1 h, ascitic macrophages stained brilliantly within 10 min, indicating marked quantitative differences. Finally, many of the highestvelocity macrophages in the ascites sample had ingested neutrophils and/or lympho- cytes and other cellular debris. The T lymphocytes in the ascites were also heterogeneous. A sizeable proportion sedimented up to 8 mm/h, a value in keeping with the characteristics of dividing or blast lymphocytes.
These data suggest that heter( within both T-lymphocyte and m macrophage populations exists bot tumour and ascites material fro vidual patients, and between (cf . Figs 2 & 3).

Characterization of infiltration: lym monocyte fraction
The cell isolation outlined ab( performed on the series of patier adenocarcinoma of the ovary to in Fig. 1. For technical and I reasons, not all cell-maker assa; carried out on each patient or cell i In particular, B-cell assays w frequent. SIg values were usually to determine as ascites and derived monocyte-macrophge c acted strongly with the (Fabl)2 even when there was no reactio normal blood monocytes.
The results (Fig. 4)  ogeneity isolating tumour and ascitic inflammatory Dnocyte-cells. First, both ascitic and tumour-,h within infiltrating cells (< 6 mm/h) had T-cell m indivalues which averaged at least as high as patients normal or patient blood. Second, the T-cell subset identified on the basis of ANAE staining of E RFC preparations was tphocyteabnormally low in frequency in the tumour derived material. Third, monocyte levels ove was assessed by nonspecific-esterase staining nts with were usually much lower in the ascitic referred fraction than in the other groups. Fourth, practical FcR rosette assays, which detected monoys were cytes only in blood, (Fig. 2, top) were fraction. markedly higher in patient material, rere in-and apparently measured either esterasedifficult negative monocytes and/or lymphocytes tumour-with abnormally high FcR levels, Finally, ells re-plasma cells were rarely seen in either reagent ascites or tumour material. In summary, in with' the marker studies indicated that the fractionation procedure yields preparaseveral tions adequate for comparison with bloodblishing derived cells for tests of immune com-[ogy for petance, though the proportions of the various cell types vary greatly between TUMOUR patients. The use of collagenase on all the solid-tumour material and several of the ascites which had gelled is unlikely to have had any effect on these results as we (Haskill et al., 1982;Hayry & Totterman, 1978) have shown this enzyme to be without marked influence on either functional or membrane studies.  Our results are based upon cell fractions which are as representative a mixture in those in the ascites, and show a as possible of all the inflammatory cells.
greater variation in reactivity for Thus, apparent differences between our tdphosphatase.
data and others probably result from differences in isolation sequences.

DISCUSSION
A number of groups have investigated both cytological and functional activity Histological grade, stage, and amount of inflammatory cells in either tumour residual tumour are likely to be the or ascites-derived material from a variety )st important prognostic indicators. of human tumours. Several features of )wever, there is considerable evidence, these reports warrant discussion, in view th in animal models and clinical of our subsequent reports concerning idies, that the degree of mononuclearfunctional activity associated with these 1 infiltration may be relevant to survival. fractions. cently there has been an increase in Klein et al. (1980) have made an extenerest in the possible functions of sive study of lymphocyte function in a mune cells isolated from solid tumours. number of types of human tumours. though a variety of effector mechanisms Using a combination of approaches, in wluding ADCC, NK, CTL, cytostatic volving velocity, density and adherence, d cytolytic macrophages have been for cell isolation, these authors isolated lated from highly immunogenic animal lymphoid fractions which averaged 51 % mours (Russell et al., 1980; Haskill E RFC and 220% EA RFC cells for their al., 1979; Herberman et al., 1980), functional studies. They seldom encounter-:re are fewer instances where active ed B lymphocytes in these preparations.
Is have been isolated from human Hayry & Totterman (1978) have inmours (Mantovani et al., 1979;Werk-vestigated the types of cell infiltrating ister et al., 1979; Totterman et al., several classes of human tumour, in-80; Klein et al., 1980;Herberman et al., cluding two with ovarian cancer. Infil-30; and Vose, 1978). trating cells were isolated through a [solation of a representative sample of series of steps similar to those described lammatory cells from tumour and in this report. T cells, monocytes and itic material for functional studies is macrophages, polymorplis and plasma cells were common isolates of these tumours, and there was considerable variation between tumours. The values for the percentage of T cells varied from 35% for several tumour classes to 83% for 3 seminoma patients. The ANAE T-cell subset was the most prominent subset in the isolate.
Recently Mantovani et al. (1980b) have reported their findings on NK activity associated with malignant effusions of the ovary. Enriched preparations of ascitesassociated lymphocytes were obtained by stepwise density and velocity separations similar to those used by Vose et al. (1977) followed by repeated adherence steps or discontinuous Ficoll-Hypaque gradients. The percentage of lymphoid cells forming E RFC (39%) averaged less than that in either blood from patients or normal blood. In contrast, the values we have found for E RFC in similar fractions average 78%. The 2-fold difference does not appear to be related to the E RFC assay, as values for normal and patient bloods are similar to ours.
This report demonstrates that mononuclear cell fractions containing > 70% T cells are routinely obtained by sedimentation-velocity separation of collagenasedispersed tumours. Similar separations on ascitic material yielded mostly T cells. Our data however, indicate differences in the T-cell characteristics of these fractions. Several studies have clearly demonstrated that T cells can be further subdivided by a highly localized esterase "dot" staining on E RFC preparations (ANAE positivity) (Saksela et al., 1979a;Moretta et al., 1979;Bevan et al., 1980). We have found that the T subset characterized by ANAE reactivity was noticeably smaller in the tumour infiltrate than in patient blood or ascites. Whether this indicates a selective localization or degranulation in vivo (Moretta et al., 1979) is not known. This result may, however, provide evidence of a preferential subset localization in ovarian tumours.
The present study also clearly indicates the heterogeneity of the macrophage component of tumour and ascites preparations. Esterase-positive monocyte-sized macrophages were frequently absent from ascites isolates, but were commonly found in tumour-derived material. In contrast, macrophages high in esterase and acidphosphatase but low in myeloperoxidase activity were isolated from the highervelocity fractions. These cells varied greatly between patients in their peroxidase and acid-phosphatase activity. In addition, many of the larger ascites macrophages contained phagocytosed leuccytes. Tumour-derived macrophages frequently contained debris, but were seldom peroxidase-positive, and usually had less acid phosphatase. This variation between patients will undoubtedly explain in part the variation in cytolytic activity reported for ascites macrophages derived from ovarian-cancer patients (Mantovani et al., 1980a) and serves as an indication that quantitative and qualitative differences in infiltration must be taken into account in assessing intra-tumour immune reactivity.
In the accompanying reports, the NK, ADCC, PHA and suppressor-cell responses of the mononuclear cell fractions described above will be outlined. These data indicate the frequent dissociation between systemic and in situ immunity, as well as the decreased reactivity of some classes of tumour-infiltrating cells isolated from ovarian cancers.