Expression of major histocompatibility antigens and leucocyte infiltration in benign and malignant human breast disease

MHC/HLA Class I Loss in Cancer Cells

In this chapter I describe Tumour Immune Escape mechanisms associated with MHC/HLA class I loss in human and experimental tumours. Different altered HLA class-I phenotypes can be observed that are produced by different molecular mechanisms. Experimental and histological evidences are summarized indicating that at the early stages of tumour development there is an enormous variety of tumour clones with different MHC class I expression patterns. This phase is followed by a strong T cell mediated immune-selection of MHC/HLA class-I negative tumour cells in the primary tumour lesion. This transition period results in a formation of a tumour composed only of HLA-class I negative cells. An updated description of this process observed in a large variety of human tumors is included. In the second section I focus on MHC/HLA class I alterations observed in mouse and human metastases, and describe the generation of different tumor cell clones with altered MHC class I phenotypes, which could be similar or different from the original tumor clone. The biological and immunological relevance of these observations is discussed. Finally, the interesting phenomenon of metastatic dormancy is analyzed in association with a particular MHC class I negative tumor phenotype.

HLA-Dr Expression by Tumor Cells Compared with Survival in High Grade Astrocytomas

Frozen samples from 78 high grade astrocytomas were reacted with a monoclonal antibody directed against HLA-Dr invariant chain. Survival data was obtained for all 78 cases. HLA-Dr was expressed by a proportion of tumor cells in 65/78 (83 %). Comparison of the survival of positive and negative cases showed that the difference was not statistically significant (p = 0.4). The relevance of the finding is discussed in the context of the immunoreaction to brain tumors.

Mononuclear Cell Infiltrate and Hla-Dr Expression in 28 Pituitary Adenomas

Frozen sections from 28 pituitary adenomas were reacted with a panel of monoclonal antibodies to macrophages, lymphocytes and HLA-Dr invariant chain. A low number of macrophages were demonstrated in all tumors, mainly perivascular. CD8 and CD4 lymphocytes were detected in even smaller numbers in 80% and 14% of tumors respectively. B lymphocytes were present in only 1 case. An occasional NK cell was present in 1/13 cases studied. HLA-Dr antigen was expressed by macrophages in all cases and by tumor cells in 2 growth hormone-producing adenomas/19 adenomas. These findings may represent evidence for a low degree of cellular immune response to pituitary adenomas.

Mechanisms of immune evasion by tumors

In the past decade, basic studies in animal models have begun to elucidate the physiological barriers which impede a successful antitumor immune response. These barriers operate at a number of levels, and involve the tumor, the tumor microenvironment and various components of the innate and adaptive immune systems. In this review, we discuss the multiple mechanisms by which tumors evade an immune response, with an emphasis on clinically relevant strategies to overcome these inhibitory checkpoints.

HLA-DRB alleles are differentially expressed by tumor cells in breast carcinoma

The biologic and prognostic significance of HLA-DR expression and T-cell infiltration in breast carcinoma are presently controversial. To test the hypothesis that these factors are influenced by particular HLA-DRB alleles, 52 breast tumor samples, composed of 26 DRB1*04 and 26 non-DRB1*04 tumors, were assessed using immunohistochemistry for expression of DR and its associated invariant chain (Ii) and for infiltrating CD3+ T cells. While DR expression by tumor cells was significantly associated with T-cell infiltration, DRB1*04 tumors were more frequently DR+ Ii+ and contained smaller CD3+ infiltrates than non-DRB1*04 tumors. This difference was largely attributable to DRB1*07 tumors, which were typically DR- Ii-, although they contained similar numbers of T cells to DR+ Ii+ tumors. Further analysis of DR+ tumors using allotype discriminating antibodies revealed that DRB1*04 alleles were always expressed, while non-DRB1*04 alleles were inconsistently expressed. The results of this study provide the first reported evidence that DRB alleles influence DR expression and T-cell infiltration in breast carcinoma and suggest that multiple factors contribute to DR expression. Ongoing studies aimed at elucidating the molecular and immunologic mechanisms controlling differential DR expression and implications for prognosis and outcome should further our understanding of the antitumor immune response and evasion strategies employed by tumor cells.

Does the immune system see tumors as foreign or self?

Given the vast number of genetic and epigenetic changes associated with carcinogenesis, it is clear that tumors express many neoantigens. A central question in cancer immunology is whether recognition of tumor antigens by the immune system leads to activation (i.e., surveillance) or tolerance. Paradoxically, while strong evidence exists that specific immune surveillance systems operate at early stages of tumorigenesis, established tumors primarily induce immune tolerance. A unifying hypothesis posits that the fundamental processes of cancer progression, namely tissue invasion and metastasis, are inherently proinflammatory and thus activating for innate and adaptive antitumor immunity. To elude immune surveillance, tumors must develop mechanisms that block the elaboration and sensing of proinflammatory danger signals, thereby shifting the balance from activation to tolerance induction. Elucidation of these mechanisms provides new strategies for cancer immunotherapy.

Immunotherapy for peritoneal ovarian carcinoma metastasis using ex vivo expanded tumor infiltrating lymphocytes

Surgery and chemotherapy have contributed to a modest overall survival in patients with epithelial ovarian carcinoma. It is therefore important to pursue novel therapy strategies for this disease that are different from conventional chemotherapy. Tumor infiltrating lymphocytes (TILs) from patients with ovarian carcinoma may represent an active immune response of the host directed against the tumor cells. These TILs can be expanded in vitro in low concentrations of recombinant interleukin-2 (rIL-2) by a few thousandfold. The resulting T-cell lines comprise CD3+CD4+TCR alpha beta + or CD3+CD8+TCR alpha beta + cells, or mixtures of both. These T-cell lines may exhibit either tumor-specific cytotoxicity against autologous tumor cells, or produce cytokines (interferon-gamma, tumor necrosis factor, and granulocyte stimulating factor) either in antigen-dependent (tumor-specific) or an antigen-independent manner. T-cell lines exhibiting primarily autologous tumor-specific cytotoxicity were developed from approximately 50% of the patients. Blocking experiments using appropriate monoclonal antibodies revealed that the CD3/TCR complex on the effector cells and the MHC class I antigens on the tumor cells were involved in the cytolytic process. We have developed a four-step method for the expansion of TILs to large numbers (1 x 10(10) to 1 x 10(11)) sufficient for clinical trials in patients with ovarian cancer. We have conducted a pilot clinical trial to examine the feasibility and clinical effects of intraperitoneal TILs and low-dose rIL-2 in patients with advanced ovarian carcinoma who were refractory to platinum-based chemotherapy. More recently, procedures have been developed for obtaining large numbers of purified CD8+ rIL-2-expanded TILs for the treatment of patients with ovarian carcinoma. The evolution of clinical trials and correlative studies necessary to develop an effective adoptive immunotherapy approach were discussed.

HLA-D Antigen Expression and Langerhans' Cell Infiltrates in Thyroid Tumors

Papillary carcinomas (PCs) of thyroid are among the most common but least aggressive human malignancies. The factors explaining the indolence of these tumors are unknown but host-tumor immune interactions may play a role. This study was designed to determine if there is morphologic evidence of these. Frozen tissues collected from 21 PCs, 4 follicular adenomas (FAs), 4 follicular carcinomas (FCs), and 11 nodular hyperplasias (NHs) were stained immunohistochemically for HlLA-D antigens, lymphocyte, and macrophage markers; results were graded numerically. Paraffin-embedded tumors (35 PCs, 10 FAs, and 10 FCs) were stained for 5-100 protein to detect Langerhans' cells (LCs). Diffuse staining for HLA-D antigens and heavy mononuclear infiltrates were found more commonly in PCs compared to follicular neoplasms (FNs) or NHs. No consistent relationship was found between lymphocyte/macrophage infiltrates and expression of HLA-D antigens. The largest number of LCs was in PCs (median 11.8 cells/standard microscopic field [c/smf]), fewer cells were found in FA (3.7 c/smf), and the least in FC (0.05 c/smf). Features of host-tumor interaction including HLA-D expression and infiltrates with lymphocyte macrophages and LC are more strongly expressed in PC than other tumors. This may play a role in explaining their biological behavior.

Prognostic significance of T-cell infiltrates, expression of beta 2-microglobulin and HLA-DR antigens in breast carcinoma

In this study immunohistological staining was used to assess the presence of T-cell infiltrates and the expression of beta 2-microglobulin (beta 2-m) and HLA-DR antigens on tumor cells of 75 ductal invasive carcinomas. The results were compared with the morphometric prognostic index (MPI) that seems to be the most accurate prognostic predictor. The extent of T-cell infiltrates differed widely between tumors, but statistically significant correlation was found only with the lymph node status, namely, tumors with a high degree of infiltration had predominantly negative lymph nodes and vice versa (p < 0.05). Only 19 (25.3%) out of 75 carcinomas were beta 2-m+, 34 cases (45.3%) showed heterogeneous staining pattern and 22 tumors (29.3%) were completely negative. We could not find any significant correlation between beta 2-m expression and MPI or T-cell content. While normal breast epithelium was always HLA-DR negative, tumor cells displayed positivity in 25 cases (33.3%), 5 tumors (6.7%) were completely positive and 20 tumors (26.7%) displayed only focal expression of class II antigens. This expression did not correlate with any single prognostic parameter, nor with MPI. The results suggest that T-cell infiltrates and the expression of histocompatibility antigens can not be accepted as prognostic indicators in breast carcinoma.

Expression of major histocompatibility class I and class II antigens and intercellular adhesion molecule-1 on operable non-small cell lung carcinomas: frequency and prognostic significance

Major histocompatibility complex (MHC) antigens and adhesion molecules, such as the intercellular adhesion molecule-1 (ICAM-1), appear to play an important role in the immunological recognition and destruction of tumour cells. We, therefore, examined the expression patterns of these proteins on primary tumours of 91 patients with operable non-small cell lung cancer (NSCLC). Applying immunohistochemistry with monoclonal antibody (MAb) W6/32 against a common framework determinant of HLA class I antigens revealed a deficient expression in 33.0% of the cases analysed, while neo-expression of either HLA class II antigens (MAb TAL.1B5) or ICAM-1 (MAb PA3.58-14) was observed in 26.4 or 29.7% of tumours, respectively. Analysis of consecutive tumour specimens indicated that HLA antigens and ICAM-1 were frequently coexpressed. With regard to clinicopathological risk factors, we could demonstrate a preferential expression of those markers in patients with locally restricted and well-differentiated tumours or no lymph node metastases, which was more pronounced in adenocarcinomas than in squamous cell carcinomas. In contrast, the presence versus the absence of HLA antigens and ICAM-1 was not correlated with the rate of tumour recurrence or overall survival in patients with NSCLC. In conclusion, the co-ordinated expression of immunologically relevant cell surface molecules on primary NSCLC is a frequent event that correlates with distinct parameters of favourable prognosis. However, we have no evidence that the immune response facilitated by these molecules can effectively influence the clinical course of the disease.

Roles of Langerhans' cells and T-lymphocytes infiltrating cancer tissues in patients treated by radiation therapy for cervical cancer

Correlations between infiltration of immunologic cells in tumor tissues and prognosis of radiation therapy were investigated for 275 patients with cervical cancer who were treated with radiation therapy alone, including 216 patients with Stage III squamous cell carcinomas and 59 with adenocarcinomas of all stages. Langerhans' cell (LC) and T-cell were stained immunohistochemically on the specimens excised from the cervical cancer. In squamous cell carcinoma, 5-year survival rates for patients with LC infiltration were significantly better than those without LC (78% versus 60%; P < 0.01). The 5-year survival rate of patients with T-cell infiltration also was significantly better than that of patients without such infiltration (83% versus 61%; P < 0.05). Similar trends were observed in patients with adenocarcinoma; 5-year survival rates for patients with LC infiltration and those without LC infiltration were 49% and 25%, respectively (P < 0.025). The survival rates for patients with T-cell infiltration and those without were 50% and 33%, respectively (P < 0.1). An analysis of patterns of failure of radiation therapy demonstrated that the favorable prognosis in LC infiltration was attributable mainly to improvement of local control rates, but that in T-cell infiltration was not. T-cells infiltrated into tumor specifically in the patients with LC infiltration in both cell types. The authors suggest that the host anti-cancer immune response of individual patients may be remarkably different at the first step of antigen recognition by LC. The LC may induce T-cell-mediated antitumor response and improve local response in radiation therapy.

Analysis of distribution of tumor- and preneoplasia-infiltrating lymphocytes using simultaneous Hoechst 33342 labeling and immunophenotyping

Hoechst 33342 in vivo staining was combined with immunofluorescent staining of cell surface antigens to quantify the distribution, relative to blood supply, of lymphocytes in a preneoplastic mammary lesion, the murine C4 hyperplastic alveolar nodule (HAN), and the C4 adenocarcinoma which develops from C4 HAN. The vascular supply to lymphocytes expressing Thy 1.2, L3T4, Ly2, and ASGM1 cell surface antigens was evaluated in both tissues. The distribution of ASGM1+ cells, which include natural killer cells, differed between the two tissues, being significantly increased in the 20% brightest Hoechst-stained lymphocyte fraction in HAN but not in C4 tumor. Distribution of T lymphocytes did not differ between the two tissues. The combination of in vivo Hoechst 33342 with in vitro immunofluorescence provides a simple method to evaluate the distribution with regard to blood supply of lymphocyte subsets in solid tumors and preneoplastic lesions.

Modulation of the antigenic phenotype of human breast carcinoma cells by modifiers of protein kinase C activity and recombinant human interferons

In the present study we have analyzed the effect of a synthetic protein kinase C (PKC) activator 3-(N-acetylamino)-5-(N-decyl-N-methylamino)-benzyl alcohol (ADMB) and the natural PKC-activating tumor-promoting agents 12-O-tetradecanoylphorbol 13-acetate (TPA) and mezerein on the antigenic phenotype of T47D human breast carcinoma cells. All three agents increased the surface expression of the tumor-associated antigen BCA 225 and various cellular antigens, including HLA class II antigens, intercellular adhesion molecule 1 (ICAM-1) and c-erbB-2. Expression of the same antigens was also upregulated to various extents in T47D cells by recombinant fibroblast (IFN beta) and immune (IFN gamma) interferon. Shedding of BCA 225 from T47D cells was induced by TPA, mezerein, IFN beta and IFN gamma, whereas ADMB did not display this activity. The ability of ADMB, TPA and mezerein to modulate the antigenic phenotype of T47D cells appears to involve a PKC-mediated pathway, since the PKC inhibitor, H-7, eliminates antigenic modulation. In contrast, the ability of IFN beta and IFN gamma to enhance the synthesis, expression and shedding of BCA 225, as well as to enhance HLA class II antigens, c-erbB-2 and ICAM-1 expression, was either unchanged or modestly reduced by simultaneous exposure to H-7. Analysis of steady-state mRNA levels for HLA class I antigens, HLA class II-DR beta antigen, ICAM-1 and c-erbB-2 indicated that the ability of H-7 to inhibit expression of these antigens in TPA-, mezerein- and ADMB-treated cells was not a consequence of a reduction in the steady-state levels of mRNAs for these antigens. The results of the present investigation indicate that the biochemical pathways mediating enhanced antigenic expression in T47D cells induced by TPA, mezerein and the synthetic PKC activator ADMB are different from those induced by recombinant interferons. Furthermore, up-regulation of antigenic expression in T47D cells can occur by a PKC-dependent or a PKC-independent pathway.

Natural killer-cell activity under conditions reflective of tumor micro-environment

Natural-killer(NK) activity was examined in the presence of low oxygen tension, low glucose concentration and acidic pH, to determine whether physical conditions present in the tumor micro-environment could play a role in down-regulating cytolytic activity of tumor-infiltrating lymphocytes with NK phenotype. Anoxia (0% O2), but not hypoxia (1% O2), significantly reduced NK activity, as did acidic pH (6.4 or 6.7). Low glucose concentration (6 mg/dl) did not impair NK activity. Combinations of either moderate (1% O2, 26 mg/dl glucose, pH 6.7) or extreme (0% O2, 6 mg/dl glucose, pH 6.4) alteration of physical conditions significantly reduced NK activity. This study indicates that the physico-chemical conditions present within solid tumors are capable of down-regulating NK activity.

Study of the relationship between immunohistologically demonstrated lymphocytes infiltrating human breast carcinomas and patients' survival

Eighty-five breast carcinomas were immunostained for CD3-, CD4-, CD8-, CD16-, CD22-, CD38- and CD57-positive lymphocyte subpopulations. The results were related to follow-up data (median follow-up 46 months) of 74 patients regarding overall survival and 73 patients in respect to disease-free survival. Whereas the number of axillary lymph node metastases (P less than 0.01) and the hormone receptor status (P less than 0.01) resulted in significantly different survival curves for overall survival, not one of the lymphocyte subset infiltrats correlated significantly which overall survival. For disease-free survival, pT stage (P less than 0.01) and nodal (P less than 0.01) and hormone receptor status (P less than 0.05) proved to be prognostically important. However, disease-free survival was not influenced by the infiltration of any lymphocyte subset.

Application of isoelectric focusing for studies of major histocompatibility complex class I antigen expression on human carcinomas and sarcomas

By one-dimension isoelectric focusing we analysed the major histocompatibility complex class I antigen expression on human tumours. Blood lymphocytes of the patients, processed in parallel, served as a basis for comparison. The prerequisite for the analysis is the preparation of metabolically active tumour cell suspensions devoid of significant leucocyte contamination. The method was found to be suitable for study of the expression of HLA alleles on ex vivo tumour cells and allowed the detection of changes imposed by in vitro treatment with interferon gamma and tumour necrosis factor alpha.

Expression of HLA-D subloci DR and DQ by breast carcinomas is correlated with distinct parameters of favourable prognois

The expression of HLA-D region products HLA-DR, DQ and DP by primary breast carcinomas was examined for its relationship to standard prognostic parameters. A positive correlation was found between the expression of HLA-DR and the differentiation state of the tumour (P = 0.02) and the expression of progesterone receptors (P = 0.002), two parameters which are associated with good prognosis and with each other. No correlation was seen between these parameters and the expression of HLA-DQ or HLA-DP. In contrast, tumour diameter was inversely correlated with the expression of HLA-DQ (P = 0.0004) although no association was observed between this parameter and HLA-DR expression. Essentially all HLA-DQ positive tumours had a diameter of less than 2 cm although these represented only 50% of the tumours of this size examimed. These data show that in breast carcinomas HLA class II expression is correlated with several distinct parameters of good prognosis and suggest that HLA-DQ expression may define a subtype of T1 tumours.

Mononuclear cell infiltrate and HLA-Dr expression in intra- and extradural secondary carcinomas

Frozen sections from 30 secondary carcinomas (22 intra- and 8 extradural) were reacted with a panel of monoclonal antibodies to macrophages, lymphocytes, NK cells and HLA-Dr invariant chain. A moderate number of macrophages was demonstrated in 91% of tumours. CD 8 and CD 4 lymphocytes were detected in smaller numbers in 68% and 28% of tumours respectively. B lymphocytes were present in only one tumour and NK cells were absent. There was no significant difference between the mononuclear cell infiltrate in the intra- and extradural tumours and in respect to the histological tumour type. HLA-Dr antigen was expressed by macrophages in most cases and by tumour cells in 5/22 (22%) (4 adeno and 1 anaplastic carcinoma; of which one adenocarcinoma was extradural and the remaining intradural). The results may represent evidence for a degree of cellular immune response to secondary carcinomas which is independent of the intra- or extradural location of the tumour.

Tumor infiltrating leukocytes (tils) during progressive tumor growth and BCG-mediated tumor regression

Tumor regression was induced by intralesional injection with BCG, 7 days after inoculation of line 10 hepatocellular carcinoma cells into strain 2 guinea pigs. Tumor-infiltrating leukocytes (TILS) were characterized immunohistochemically with 11 monoclonal antibodies (MoAbs) during the induction phase of line 10-immunity, and during immune-mediated regression of the tumor, at days 12 and 28 after tumor cell inoculation, respectively. At day 5 after BCG-injection (day 12 after tumor cell inoculation), there were no major differences between the TIL subpopulations of the BCG-treated and untreated tumors. The TILS were mainly T-cells, as identified by MoAbs against Pan T-cells (CT5), T-cytotoxic/suppressor cells (CT6) and T-helper/inducer cells (H155). A limited number of macrophages was also present. However, at day 21 after BCG-treatment (28 days after tumor cell inoculation), the fibrous stroma was increased dramatically in most of the BCG-treated tumors, and as a result, the tumor cell islets were smaller than in control tumors. In the BCG treated tumors, the numbers of T-cells and macrophages were increased. In growing and regressing tumors, MHC class I and II antigens were strongly expressed in TILS and in the tumor stroma. Line 10 tumor cells prior to inoculation expressed no MHC class I or II antigens. In the centers of the tumor islets at days 12 and 28, expression of these antigens was not found. However, MHC class I and II antigens were expressed on tumor cells at sites where they lay close to the fibrous stroma or TILS. This observation was made in progressively growing tumors and was most apparent in BCG-treated tumors.

Flow cytometric analysis of tumour infiltrating lymphocytes in breast cancer

In 31 patients with carcinoma of the breast the phenotype and activation status of tumour infiltrating lymphocytes (TILs) was analysed by flow cytometry. The predominant cells, in all patients, were T lymphocytes and in the majority of cases CD8+ (cytotoxic/suppressor) T lymphocytes were present in greater numbers than CD4+ (helper) T lymphocytes. There was no relationship between the degree of lymphocytic infiltration and either tumour stage or grade but there appeared to be an inverse correlation with the levels of oestrogen receptor (ER) in the tumour (P less than 0.01). Both populations of T cells had significantly higher numbers of cells carrying HLA DR (class II major histocompatibility antigen) than the equivalent populations in peripheral blood from the same patient group (P less than 0.001). The transferrin receptor was found on similar numbers of CD8+ T cells in peripheral blood and among the tumour infiltrating lymphocytes while more of the CD4+ T cells infiltrating the tumour were found to carry this receptor (P = 0.034). The Tac (CD 25) antigen was also on similar numbers of CD8+ T cells from both peripheral blood and the tumour but was on fewer of the CD4+ T cells in the tumour with respect to peripheral blood (P = 0.029). In both TILs and blood lymphocytes, the Tac antigen was consistently present on greater numbers of CD4+ T lymphocytes than on the CD8+ T lymphocytes (P less than 0.001) and as this is a component of the interleukin 2 (IL-2) receptor this may be of relevance to the use of IL-2 in TIL cancer therapy.

An evaluation of the prognostic significance of HLA-DR expression in gastric carcinoma

The prognostic value of histochemical staining for HLA-DR was assessed retrospectively in 52 "curative" gastrectomy specimens from patients with gastric carcinoma. In addition other tumor variables, including histologic type, degree of differentiation, extent of local spread, prominence of mononuclear infiltrate, tumor edge, and lymph-node metastases, were examined. Sixty-five percent of the tumors showed positive staining for HLA-DR, and these tumors had a higher mean and median survival at 5 years compared with negative tumors. However, on multivariate analysis the difference was not statistically significant. The authors conclude that, although HLA-DR staining is associated with tumors of better prognosis, it cannot be used as an independent prognostic factor. Of the other tumor variables, only lymph-node status was of prognostic significance.

Lacking prognostic significance of beta 2-microglobulin, MHC class I and class II antigen expression in breast carcinomas

To evaluate the impact of MHC antigen expression on the survival of patients with cancer, 77 human breast carcinomas were investigated for the expression of beta 2-microglobulin (beta 2m), HLA-A,B,C and HLA-DR. Thirty-one benign breast tumours were stained for comparison. The results for the carcinomas were related to the survival data of the cancer patients. The expression of beta 2m, HLA-A,B,C and HLA-DR was significantly lower in malignant tumours compared to the benign lesions. Whereas all benign tumours were positive for beta 2m and HLA-A,B,C and 28/31 positive for HLA-DR the following positivity rates were found in carcinomas: 74/77 for beta 2m, 57/77 for HLA-A,B,C and 10/77 for HLA-DR. The follow-up (median 45 months) of 66 cancer patients for overall survival and of 65 patients for disease-free survival revealed no influence of beta 2m, HLA-A,B,C or HLA-DR expression on the prognosis of this cancer. In conclusion, experimental data indicating the importance of MHC antigens in anti-tumour responses are not confirmed by the analysis of cancer patient survival data.

Antigenic profile of mammary fibroadenoma and cystosarcoma phyllodes. A study using antibodies to estrogen- and progesterone receptors and to a panel of cell surface molecules

Using serial frozen sections, monoclonal antibodies and an indirect immunoperoxidase method, 13 fibroadenomas (FA) and 3 cystosarcomas phyllodes (CSP) were analyzed for the expression of Egp34, HEA319-antigen, leucocyte differentiation antigens CD10, CD30, CD57, CD72, CDw75, and CD77, epidermal growth factor receptor (EGFR), estrogen (ER) and progesterone receptor (PR), and transferrin receptor (CD71). Egp34, CDw75, HEA319 antigen, CD10, and CD30 turned out to be consistently expressed in different cell types constituting FA and CSP and revealed that in malignant CSP the myoepithelial compartment acquires the ability to invade the stroma. Phenomenologically, the variable mode of expression of CD57 in myoepithelial cells, of CD77 in ductal epithelium, and of CD72 in both epithelial and stromal cells is suggestive for reflecting differences in their functional state but cannot be further interpreted at present. Expression of PR and ER was restricted to duct cells and was relatively independent, non-systematical. However, expression of ER and EGFR was inverse. This was also true for EGFR and CD71 in both duct cells and myoepithelial cells of FA. In contrast, stromal cells of FA were able to co-express EGFR and CD71 in the absence of PR and ER. This suggests a hormone-independent stimulation of the stromal cell compartment, possibly leading to local proliferation as the primary event in tumorigenesis of FA. In malignant CSP, however, the main proliferating cell is an abnormally mobile, HEA319 antigen-, CD10- and CD30-positive myoepithelial cell found to co-express ERFR and CD71 which is abnormal for this cell type but encountered in (myo-)fibroblasts of FA.

Histocompatibility antigen expression in pulmonary carcinomas as indication of differentiation and of special subtypes

In 70 human carcinomas of the lung the expression of histocompatibility antigens (HLA) was investigated by using monoclonal antibodies in frozen sections. The immunmoreactive tumor volume was determined morphometrically. The tumor types showed a different pattern of HLA expressions. In squamous cell carcinomas (SQC) and small cell carcinomas (SCC) the HLA-A,B,C expression varied in a great range from 0% to 100% of positive tumor volume. For the most part, HLA-DR was not demonstrable. In adenocarcinomas (AC) and large cell carcinomas two different populations were observed. One showed a pattern of HLA expression as seen in SQC. Electron microscopically these AC (type 1) (AC/1) were characterized by intracytoplasmic mucin granules indicating a phenotypical relationship to goblet cells. The other--nearly 50% of adenocarcinomas and one third of large cell carcinomas--expressed both HLA antigens homogeneously in the total tumor volume. Electron microscopically the characteristics of type II pneumocytes and Clara cells were seen in tumor cells of these AC (type II) (AC/II), consisting in lamellar bodies and apically located electron-dense granules. In SQC and AC/I a significant correlation was found between grades of differentiation and mitotic activity on the one side, and expression of HLA-A,B,C on the otrher. The loss of HLA-A,B,C expression seemed to be an indication of a lower grade of differentiation; the smalles expression of HLA-A,B,C occurred in SCC.

Risk factors for breast recurrence in premenopausal and postmenopausal patients with ductal cancers treated by conservation therapy

Risk factors for local failure were evaluated for 496 clinical Stage I-II patients with infiltrating ductal carcinomas (median follow-up, 71 months) treated by conservative surgery and radiotherapy. Monofactorial analysis identified the following factors to be correlated with increased risk: moderate/marked mononuclear cell reaction (MCR), high histologic grade (G), extensive intraductal component (EIC), tumor necrosis, macroscopic multiplicity, estrogen receptor negativity, anatomic tumor size, age younger than 40 years, and vascular invasion. Only MCR, G, and EIC proved significant in Cox multivariate analysis. These risk factors were highly age dependent, with EIC markedly more prevalent in women younger than 50, MCR and G in women younger than 40. Separate Cox analysis for premenopausal patients showed that MCR/EIC determined risk independent of resection margins: tumors with MCR had a 28%, and with EIC a 22% probability of recurring locally by 5 years. Premenopausal patients with neither risk factor had a very low failure rate (2.6% at 5 years), regardless of age. For postmenopausal patients risk of breast recurrence was determined both by adequacy of resection margins and grade, with a high local failure rate for patients having G3 tumors with positive or indeterminate margins (31% at 5 years). The authors conclude that the microscopic examination is the only useful tool for assessing the risk of local failure, which is quite low for the majority of patients treated with breast conservation. High-risk patients can be recognized morphologically. The age dependence of morphologic risk factors appears to explain the high local failure rate seen in patients younger than 40.

The relationship of class I MHC antigen expression to stage IV-S disease and survival in neuroblastoma

Cultured human neuroblastoma cells express low levels of class I (MHC) surface antigen. In order to determine if this low expression is representative of the clinical tumor, this study investigates class I expression in archival human neuroblastoma. Whereas stages I to IV neuroblastoma expressed low levels of class I antigen, stage IV-S tumor cells expressed normal levels, similar to control tissues. Expression of class I antigen in tumors from survivors of stage III neuroblastoma was significantly greater than in tumors from nonsurvivors. Tumors comprised predominantly of ganglion cells expressed significantly more class I antigen than neuroblasts. These data suggest that class I MHC expression may play a role in the natural history of human neuroblastoma.

Lymphocytic infiltration and cytotoxicity under hypoxic conditions in the EMT6 mouse mammary tumor

Infiltration of lymphocytes, neutrophils and macrophages was evaluated in hypoxic and well-oxygenated areas of the EMT6 mouse mammary adenocarcinoma, by in vivo staining with the fluorescent dye Hoechst 33342 followed by cell sorting on the basis of fluorescence intensity. Tumors were grouped by days post-injection (days 11-14, 15-17 and 20-27). As lymphocytes are the only host cell population in this tumor model to possess lytic activity against EMT6 tumor cells, the ability of sensitized T lymphocytes to lyse syngeneic EMT6 cells was examined under conditions of varying oxygen concentrations. Infiltrating lymphocytes were detected to the same extent in cell fractions from both areas in all tumors. In contrast, neutrophils were found in significantly higher percentages in the hypoxic population than in the well-oxygenated cell fraction of all but the largest tumors. Macrophages were present in significantly higher percentages in the well-oxygenated fraction than in the hypoxic fraction of day-11 to -14 tumors. Extreme radiobiological hypoxia (0% O2) resulted in a significant decrease in T-cell-mediated lysis of EMT6 tumor cells, compared to lysis in room air (20% O2), but lysis was not impaired under conditions of mild radiobiological hypoxia (1% O2). Our study indicates that host-cell infiltration into areas of differing oxygenation may be quantitated via in situ Hoechst staining followed by cell sorting; in the EMT6 tumor, lymphocytes appear to infiltrate hypoxic areas to the same extent as well-oxygenated areas, and T-lymphocyte killing of syngeneic tumor cells is significantly reduced, although still present, under these hypoxic conditions.

Quantitative immunohistochemical analysis of mononuclear infiltrates in breast carcinomas--correlation with tumour differentiation

Inflammatory infiltrates were analysed in tissue sections of 76 breast carcinomas by counting the percentage of macrophages, IgA+ and IgG+ plasma cells, T cells with their subpopulations, and natural killer cells, and by measuring postcapillary venules (PCVs, found in 12 cases) within the infiltrates. These parameters were correlated with nuclear grade and biochemically determined hormone receptor status, known markers of tumour differentiation. A direct correlation was found between the extent of inflammation and nuclear grade (P less than 0.0001), and an inverse correlation between inflammation and oestrogen receptor (OR) positivity (P less than 0.05) as well as inflammation and progesterone receptor (PR) positivity (P less than 0.05). The percentage of the OKT8+ suppressor/cytotoxic T cells increased when the inflammation expanded from scanty to moderate (P less than 0.02). The diameter of the PCVs also increased with increasing inflammatory infiltrate (P less than 0.02). In addition, a direct correlation exists between the diameter of the PCVs and both the percentage of the OKT8+ T cells (P less than 0.04) and the Leu-7+ natural killer cells (P less than 0.03).

Density and phenotype of tumour-associated mononuclear cells in colonic carcinomas determined by computer-assisted video image analysis

The density and phenotypes of tumour-associated mononuclear cells (TAMC) in tissue sections of colonic carcinomas was determined by the technique of video image analysis (VIA). This technique allowed an accurate and objective enumeration of both total mononuclear cells (MC) in H&E stained sections and individual types of cells as revealed by immunoperoxidase staining with monoclonal antibodies in frozen sections. This enumeration allowed reliable statistical analysis of the differences between sample groups. Using this technique it was found that the density of MC in histiologically normal tissue was significantly higher than in tumour tissue. Tumours from patients with the best prognosis (stage A) had significantly higher numbers of TAMC than stage B (P less than 0.02), C (P less than 0.002) and D (P less than 0.002) tumours. The differences in the density of TAMC between tumours obtained from stage B and C and that between C and D were not significant, whereas stage B had a significantly higher TAMC density than stage D tumours (P less than 0.05). Comparing tumour differentiation, well differentiated adenocarcinomas had a significantly higher (P less than 0.05) TAMC density than poorly differentiated tumours but not moderately differentiated tumours. Moderately and poorly differentiated adenocarcinomas did not differ significantly in the density of TAMC. In examining the phenotype of these cells, it was found that T lymphocytes formed the majority of the TAMC with the CD4+ subset predominating in 28 of 29 cases. Similarly, all sections of normal colon (taken at least 4 cm away from the tumour) had more CD4+ than CD8+ cells. The proportion of the total leucocyte population that was CD3+ was comparable in normal and tumour tissue. Generally, few macrophages were present in either tumour or normal tissues. B cells (CD21%) and subset of NK cells (CD57+) were not detected in the tumours. There were no significant differences in the proportion of leucocytes which were CD4+, CD8+ and CD14+ (macrophages) between the normal colon and the tumour tissues. The types of cells in the TAMC population did not differ with tumour stage or differentiation or with the density of the TAMC itself.

Expression of HLA ABC and DR antigens in thyroid neoplasia and correlation with mononuclear leukocyte infiltration

The rejection of tumor cells by the immune system depends on the production of tumor-associated antigens and the expression of HLA antigens on these cells. We therefore studied the expression of HLA ABC and DR antigens in malignant and benign thyroid disorders and correlated it with the types and extent of mononuclear cell infiltration. In the normal thyroid, HLA ABC expression was weak and focal, while it was diffusely present in benign disorders and in most but not all malignancies. HLA DR antigens, while absent or infrequently expressed in normal thyroid, were strongly but often focally expressed in all cases of autoimmune thyroid disease (AITD), as well as in most cases of malignant tumors and benign epithelium surrounding these tumors, and colloid nodule disease. There was a T cell predominance in all disorders, and the CD4+/CD8+ ratio was greater than 1 in most benign and malignant specimens. There was a direct correlation between the high expression of HLA antigens and dense inflammatory infiltration in AITD but not in most tumor specimens. Lack of such correlation suggests that the expression of HLA antigens is an autonomous event in tumors, independent of cellular infiltrate.

Expression of the adhesion molecule ICAM-1 and major histocompatibility complex class I antigens on human tumor cells is required for their interaction with autologous lymphocytes in vitro

In a group of 30 human tumors, comprising 12 lung, 14 ovarian, 2 breast carcinomas, 1 hypernephroma and 1 mid-gut carcinoid, the expression of major histocompatibility complex (MHC) class I molecules and the intercellular adhesion molecule 1 (CAM-1, CD54) was found to vary independently. Some tumors expressed both or neither of these molecules. Among 9/13 ICAM-1+ tumors, in which greater than 50% cells reacted with the anti-ICAM-1 monoclonal antibody (mAb) (LB-2), the class I antigen was also detected on greater than 50% of the cells. Only 2 ICAM-1+ tumors were class-I-. In 5/17 cases the tumors were MHC-class-I+ and ICAM-1-. Lymphocytes collected from the blood or from the tumor site were assayed for recognition on the tumor cells in the auto-tumor cytotoxicity test and in mixed lymphocyte tumor cell culture (MLTC). Positive results were obtained only with the MHC-class-I+/ICAM-1+ tumors. In vitro treatment of the tumor cell suspensions with interferon gamma and tumor necrosis factor alpha (TNF alpha) induced or enhanced the ICAM-1 and/or class I antigen expression in 8/12 cases. Of the tumor samples treatged, 8/9 acquired stimulatory capacity and 3/10 became susceptible to lysis by the lymphocytes. In 6/6 MLTC performed with the cytokine-treated tumor cells, cytotoxicity against the autologous tumor was generated. Three of these MLTC lymphocytes also lysed the untreated targets. mAb directed to class I antigens or to ICAM-1 inhibited both the stimulation by and the lysis of tumor cells when confronted with fresh lymphocytes. The cytotoxicity generated in the MLTC was also inhibited. If, however, the cytotoxic function was induced in MTLC containing interleukin-2 (5 U/ml), inhibition was obtained only by pretreatment of the targets with mAb against ICAM-1. The results show thus (a) that the lymphocytes react in vitro with tumor cells only if these express both MHC class I molecules and ICAM-1; (b) that expression of these molecules can be induced by interferon alpha and TNF alpha; (c) that cytotoxic effectors generated in the MLTC with cytokine-treated tumors can also act on the untreated tumor cells. The requirement of the two surface moieties for the interaction with lymphocytes was also substantiated by blockade with relevant mAb.

Differences in the phenotypes of cells mediating anti-tumour immunity at various stages of tumour progression in mice

In Winn assays, T cells from donors immunized by tumour excision, or from mice with small tumours, mediate rejection of the metastasizing murine fibrosarcoma MC-2. As the mean size of primary tumours in spleen donors increases, the strength of anti-tumour activity declines, until it is frequently undetectable in spleen cells from mice with very large tumour burdens. Loss of splenic anti-tumour activity is coincident with the appearance of cells capable of suppressing an otherwise protective anti-tumour response in Winn assays. This paper defines the phenotypes of T cells mediating immunity against MC-2. Eleven or more days after tumour inoculation the proportions of tumour-bearer splenic leucocytes expressing Ly 1.2 (CD5), Ly 2.2 (CD8a) or L3T4 (CD4) surface antigens were significantly less than similar preparations from normal animals. Depletion of Ly 1.2+ or L3T4+ cells from spleen cells of donors with small tumours, or from donors immunized by tumour excision, diminished protection in the Winn assay. Depletion of Ly 2.2+ cells from these donors had no effect on immunity. In contrast, spleen cells taken from donors with large tumors lost all anti-tumour activity if pretreated with any one of anti-Ly 1.2 or anti-Ly 2.2 or anti-L3T4 antibodies in the presence of complement. These results suggest that cells bearing the Ly 2.2 marker may be important to weak immunity remaining in the spleens of mice with large tumours, but are not critical to strong immunity generated early in tumour growth, nor to that following tumour excision. That is, in addition to an Ly 1.2+, Ly 2.2-, L3T4+ spleen cell subset also seen early in the growth of the MC-2 tumour, a cell population which expresses the Ly 2.2 marker and which is important to anti-tumour immunity emerges late in tumour growth.

Correlation of HLA-D/Ii antigen expression in breast carcinoma with local lymphohistiocytic infiltration reveals considerable dysregulation in a subset of tumors

Non-neoplastic mammary glands and 203 carcinomas of the breast were examined immunohistochemically for expression of HLA-DR, -DP, and -DQ molecules and the HLA-D-associated invariant chain (Ii). According to the reaction of normal breast epithelium in the course of chronic inflammation, these antigens were found to be induced in 120 carcinomas. In 106 of these the order was Ii greater than or equal to HLA-DR greater than or equal to HLA-DP greater than or equal to HLA-DQ. The extent of induction of these sequentially expressed antigens correlated with the density of local lymphohistiocytic infiltration (Ii: p = 0.003; HLA-DR: p = 0.0005; HLA-DP: p = 0.0000005; HLA-DQ: p = 0.01). Therefore, Ii/HLA-D antigen expression of carcinoma cells in local association with inflammation was regarded as an adequate reaction. The presence of these antigens in the absence of inflammation was regarded as "inadequate hyper-expression", and was found in 14 cases involving Ii, in 8 involving HLA-DR, and in 1 case each involving -DP and -DQ molecules. In contrast, 20 cases expressed minimal/no Ii despite heavy inflammatory infiltration. This "inadequate hypo-expression" was also found in 25 cases involving HLA-DR, in 30 involving -DP and in 40 involving -DQ determinants. Combining the 2 variants showed that at least 54 cases (26.6%) were characterized as being dysregulated. "Adequate vs. inadequate Ii/HLA-D antigen expression" did not correlate with tumor grading and might therefore be an independent parameter.

What's new in the local immune response in cancer?

Malignant tumors in humans are commonly associated with an inflammatory infiltrate. The mechanisms that account for the accumulation of T-lymphocytes and macrophages--these cells comprise the major components of tumor infiltrates--in the vicinity of a growing tumor are not fully understood. Tumor specific and immunogenic antigens could not be demonstrated in most solid tumors of humans, in contrast to several experimental tumor models. Thus it is not proven in human malignancies that neoantigens expressed on malignant cells are the signal which initiates an inflammatory response that, immunohistologically, is comparable to mononuclear infiltrates present in allograft rejection. A variety of nonspecific factors including lymphokines released by tumor cells may also account for the accumulation of inflammatory cells at the tumor site. The difficulties to evaluate the functional role of the "local immune response" for tumor and host are even greater. Most tumors progress in the presence of mononuclear infiltrates. Do they progress in spite of or because of the action of the local immune response? Clinical, immunopathological, and experimental data suggest that both is right, and that at least four distinct properties of tumor-associated immune reactions exist: Regression, Selection, Modulation and Progression. These distinct properties will be discussed below, using mainly the malignant melanoma of the skin as a model for a malignant tumor in humans.

Expression of MHC class I and class II antigens in colonic carcinomas

Malignant and non-malignant ('normal') colonic tissues from patients with colonic carcinoma were examined for the expression of MHC class I and class II antigens by immunoenzymatic staining using monoclonal antibodies. The amount of class I antigen as detected by 2 monoclonal antibodies, FMC 16 or W6/32 was clearly diminished in 11 of 14 tumours when compared to the amount present on 'normal' colonic tissue from the same individual. The loss of class I antigen did not correlate with tumour stage or differentiation. The reactivities of FMC 16 and W6/32 with these tissues were not identical, which indicates that the 2 monoclonal antibodies may recognize different epitopes on the HLA class I molecule. Class II antigens were absent from 'normal' colonic epithelium but were present on 20 of 28 tumours, with DR being detected more often than DP, and DQ found only on 4 of 28 tumours. When present, staining for class II antigens was heterogeneous within the tumour, in that all tumour cells did not stain equally. DR and DP antigens were found more often on moderately or poorly differentiated adenocarcinomas and on stage B, C and D tumours in that order of frequency. Thus tumours with a better prognosis were less likely to express DR and DP. The expression of DQ was unrelated to staging or differentiation.

Lysis of human melanoma cells by autologous cytolytic T cell clones. Identification of human histocompatibility leukocyte antigen A2 as a restriction element for three different antigens

From the peripheral blood of the melanoma patient (AV), we derived cytolytic T lymphocyte (CTL) clones that lysed the autologous tumor line SK-MEL-29, but not autologous EBV-B cells, K562, and other tumor targets. By immunoselection experiments it was shown that the CTL clones recognized at least three different antigens on the autologous tumor cells. We demonstrate here that these melanoma antigens are presented to the CTL in association with HLA-A2. First, HLA-A2-reactive pregnancy sera as well as an mAb against HLA-A2 inhibited the CTL lysis. Second, immunoselected melanoma subclones that were resistant to lysis by CTL clones against the three antigens described were found to lack expression of HLA-A2. By sensitizing the patient's lymphocytes against an HLA-A2- melanoma clone, we established a new series of CTL clones recognizing autologous AV melanoma cells. However, efficient lysis was only seen when target cells were pretreated with IFN-gamma. The lytic activity of these CTL was selectively inhibited by an mAb against a common HLA-B determinant. These results indicate that in addition to HLA-A2, other class I antigens are involved in the recognition of AV melanoma cells by autologous CTL.

Expression of HLA-A, -B, -C, -DR, -DP, -DQ, and of HLA-D-associated invariant chain (Ii) in non-neoplastic mammary epithelium, fibroadenoma, adenoma, and carcinoma of the breast

Non-neoplastic mammary gland, 20 benign tumors and 206 carcinomas of the breast were immunohistochemically examined for expression of HLA-A, -B, -C, HLA-DR, -DP, and -DQ molecules and the HLA-D associated invariant chain (Ii). In contrast to cells from benign lesions, tumor cells of 51.2% of carcinomas had an abnormally low content of HLA-A, -B, and -C determinants ranging from reduction of antigenic density per cell (28.8%) over an incomplete (15.6%) to complete loss of antigens (6.8%). Associated with lymphohistiocytic stromal infiltrates, HLA-D/Ii determinants were found to be induced in benign duct and acinar epithelium after the order Ii greater than or equal to HLA-DR greater than or equal to HLA-DP greater than or equal to HLA-DQ. These antigens were also expressed, mostly noncoordinately, in 55.5% of carcinomas, and in 98 cases according to the above order. In 28.6%, Ii expression clearly exceeded HLA-D antigen expression; conversely, 6.2% contained HLA-DR+/Ii- tumor cell subsets. In breast carcinoma, the association of reduced HLA-A, -B, and -C expression and a noninduction of HLA-DR was highly significant (P less than 0.0009), suggesting an abnormal signal acting down-regulating on the expression of both classes of antigens. Because the modality of HLA-A, -B, and -C and HLA-D/Ii expression correlated with neither tumor type nor grade, it might be an independent parameter.

Auto-tumor recognition following in vitro induction of MHC antigen expression on solid human tumors: stimulation of lymphocytes and generation of cytotoxicity against the original MHC-antigen-negative tumor cells

Expression of major histocompatibility complex (MHC) class I antigens was induced in eight out of nine freshly prepared tumor cell suspensions by exposure to interferon (IFN gamma) and tumor necrosis factor (TNF alpha) in vitro. The untreated, class-I-antigen-negative, and the treated, antigen-positive, cells of three tumors (one breast carcinoma, one plasmocytoma and one ovarian carcinoma) were compared for the capacity to stimulate autologous and allogeneic blood lymphocytes, to generate auto-tumor cytotoxicity and for sensitivity to the lytic effect induced in autologous mixed lymphocyte tumor cell culture (MLTC). The MHC class I-negative cells did not stimulate, while the cells induced for expression of antigens did. On the other hand, when the autologous cytotoxic cells were generated in the MLTC by the class I antigen-positive tumor cells the class I-negative tumor cells were also damaged. Lysis of the class-I-positive tumor cells was abrogated by the W6/32 monoclonal antibody directed against the monomorphic part of the class I molecules.

Tumor-infiltrating lymphocytes and histocompatibility antigens in primary intracranial germinomas

Subpopulations of tumor-infiltrating lymphocytes (TIL's) and the major histocompatibility complex (MHC) antigens of neoplastic cells were examined in three intracranial germinomas by an immunohistochemical method using monoclonal antibodies. About 70% to 80% of TIL's were T lymphocytes which were either infiltrating diffusely or in clusters, whereas 20% to 30% of TIL's were B lymphocytes which tend to cluster in tumor tissues. Examination of T lymphocyte phenotypes revealed both the cytotoxic/suppressor and helper/inducer T lymphocytes, as in other tumors. However, the existence of a considerable number of B lymphocytes in the TIL population was uncommon and seemed to be a characteristic feature of the intracranial germinoma, which might suggest a difference of host immune response to this neoplasm as compared to other tumors. On examination of the MHC antigens, no MHC class I or II antigens in the neoplastic cells were stained, while positive staining for both antigens was seen in the TIL and stroma tissues. From these findings, it was suggested that the degree of TIL infiltration might not be correlated with the expression of MHC antigens in neoplastic cells in cases of primary intracranial germinoma.

Expression of class I and II major histocompatibility complex antigens in Wilms' tumour and normal developing human kidney

The Wilms' tumour is a solid childhood tumour of the kidney, consisting of blastema, tubules and mesenchyme. Embryonic tumours, such as Wilms', may arise as a result of a developmental disturbance in differentiation. The expression of class I and II major histocompatibility complex (MHC) antigens was investigated on 6 Wilms' tumours and related to that in the developing human kidney in this immunohistological study, using a panel of monoclonal antibodies. The Wilms' tumour blastemal cells were class I MHC antigen negative, but differentiated structures were positive. Class II MHC antigens were not observed in Wilms' tumours. In the developing human kidney class I MHC antigen expression was observed on glomeruli from 8 weeks and on tubules from 13 weeks gestational age. Class II MHC antigen expression was observed on glomeruli from 11 weeks and on tubules from 13 weeks gestation. These results suggest that the blastemal cells within the Wilms' tumour may reflect an early stage of development with respect to the expression of MHC antigens.

HLA class II sublocus expression in benign and malignant breast epithelium

Using immunohistochemistry and a panel of five monoclonal antibodies, the epithelial expression of HLA class II sublocus products by benign and malignant breast has been studied. The magnitude of the stromal mononuclear inflammatory cell infiltrate was assessed. There was expression of HLA class II by 75 per cent of epithelial cells in the benign tissues, with little variation in intensity and between antibodies. There was coordinate expression of DR and DQW1. Epithelial expression by carcinomas was more complex and variable. Most (61 per cent) carcinomas exhibited variable loss of epithelial expression of class II products, as detected by three antibodies recognizing epitopes on DP, DQ, and DR together. Thirteen (28 per cent) carcinomas were completely negative or had very occasional positive cells. The extent of this loss was unrelated to the magnitude of the inflammatory infiltrate and axillary lymph node status. No well-differentiated carcinomas exhibited complete loss. Furthermore, non-coordinate expression of DR and DQW1 was present in 8 out of 40 carcinomas, with the proportion of DQW1 positive epithelium always being less than that of DR. Carcinomas exhibiting non-coordinate expression were never well differentiated; there was no relationship with the extent of the inflammatory infiltrate. This is the first study to detail HLA class II expression in breast, and our results suggest that alterations in expression of these products may modify or reflect tumour behaviour.