Expression of Folate Receptor Type α in Relation to Cell Type, Malignancy, and Differentiation in Ovary, Uterus, and Cervix¹

The FR³ is represented by a homologous family of glycoproteins (1, 2, 3, 4), two of which (FR-α and FR-β) are attached to the cell surface by a glycosyl-phosphatidylinositol anchor (5); the third isoform (FR-γ) and its truncated version (FR-γ′) are constitutively secreted because of a lack of an efficient signal for glycosyl-phosphatidylinositol modification (4). FR-α is expressed in some normal epithelial cells and is elevated in certain carcinomas, whereas FR-β is a myeloid differentiation marker and is elevated in some nonepithelial malignancies (6, 7). FR-γ/γ′ is expressed in hematopoietic tissues (3, 4).

At present, FR is a major focus as a tumor target for multiple experimental approaches in cancer therapy. One novel approach uses bifunctional antibodies to target T cells to the FR on the surface of ovarian carcinoma cells. Selective growth inhibition of the tumor cells was obtained by this approach (8). The chimeric antibodies, which bind to both FR and either CD3 or CD28, produced impressive results in a xenogeneic model (9, 10) and in patients with advanced ovarian cancer (11, 12, 13). Similarly, a chimeric molecule consisting of a single-chain, Fv of anti FR antibody and interleukin 2 was effective in inhibiting tumor growth in vivo (14). Alternatively, folic acid conjugates of single chain anti-T-cell receptor antibody could mobilize T-cell response against FR-rich tumors (15). Taking advantage of the nondestructive nature of FR-mediated internalization of folate-coupled macromolecules (16, 17), cytotoxins such as momordin, Pseudomonas exotoxin, and maytansinoids were shown to produce selective killing of FR-rich cells (18, 19, 20, 21). Furthermore, the toxicity of such conjugates was dependent upon receptor density on the cell surface (20). Folate-conjugated radiopharmaceuticals also appear to offer a means of tumor imaging/radiation therapy (22, 23, 24, 25). Folate-coated liposomes were shown to selectively target FR-rich tumor cells (26), and selective killing of the malignant cells was obtained by encapsulating doxorubicin in the liposomes (27). By a similar strategy, it was possible to deliver antisense oligonucleotides against the epidermal growth factor receptor to FR-rich tumor cells. Furthermore, selective targeting of an adenoviral vector to FR-rich tumor cells has been achieved in the presence of an antibody to ablate the endogenous viral tropism (28). Finally, several studies have shown that FR, when expressed at high levels, could offer the preferred uptake route of novel classes of antifolate drugs that target glycineamide ribonucleotide formyltransferase and thymidylate synthase (29).

A soluble form of FR has been detected in the serum and ascites of patients with ovarian cancer (30). Thus, receptor-rich tumors that shed FR-α into the circulation may potentially be detected by a serological test.

Among the available tumor targets/markers, FR has certain distinctive advantages in that it binds to a ligand (folate) with high affinity, is either absent or poorly expressed in most normal tissues that are accessible to the circulation, and is vastly overexpressed in certain malignant tumors. The occurrence of multiple tissue-specific FR isoforms provides an added level of tissue selectivity in FR-mediated cancer therapy/prognosis/diagnosis. The differential expression of FR isoforms in various tissues, their differential up-regulation in specific malignancies, and the potential to modulate their expression in tumors are, therefore, issues of major importance in their clinical utility. In this context, fundamental questions that need to be adequately addressed are: (a) the relative levels of FR expression associated with each malignant subtype; (b) relationships between FR expression and malignant transformation for different cell types; and (c) the basis for heterogeneity in FR expression among malignant cells within a tumor. These questions are best addressed by inspection of tissues in vivo because the receptor expression patterns in established cell lines do not necessarily reflect its in vivo expression (6).

The need for the present study arose from the limited and sometimes conflicting reports of FR expression patterns among various normal and malignant cell types of gynecological tissues from previous immunocytochemical analyses (31, 32, 33, 34, 35). The limitations of the immunological approach are documented by Stein et al. (32) and include: (a) the nonquantitative nature of the results; (b) the nonreactivity of the anti-FR monoclonal antibody probes with paraffin-embedded sections and the variability in their reactivities (including false negatives) in different preparations of frozen sections; (c) the lack of a suitable control for the integrity of the antigen; (d) inability to take into account the differences in cell densities between different malignant tissues and the corresponding normal tissues; and (e) possible recognition of nonreceptor protein by the primary antibody.

The present study overcomes the above limitations by in situ analysis of the relative levels of FR mRNA. The approach allows the probing of morphologically intact cells in paraffin-embedded sections and has a demonstrable high specificity for specific FR transcripts. The analysis is quantitative, and the signal intensities are expressed relative to that of the β-actin transcript, thus taking into account differences in cell densities and the integrity of the target mRNA.

The cDNAs for FR-α (nucleotides 558–900) and FR-β (nucleotides 591–933) were placed between BamHI and PstI sites in the polylinker region of the expression vector pBluescript SK (Invitrogen), and the resulting plasmid was amplified in Escherichia coli XL1 blue. The cDNA for FR-γ (nucleotides 1–358) was placed into the BamHI site in the polylinker region of the same expression vector and amplified. A fragment (nucleotides 820–946) of cDNA encoding β-actin was inserted into pBluescript SK (Invitrogen) at KpnI/EcoRI sites. The cDNA fragments were transcribed in either the sense (negative control) or the antisense orientation in vitro in the presence of [³⁵S]CTP and [³⁵S]UTP (Amersham). Either T3 or T7 RNA polymerase (Promega) was used in the in vitro reverse transcription reactions in the different orientations. One μl (10 units) of RNase-free DNase I (Boehringer Mannheim) was added to each reaction incubated for 15 min at room temperature. The labeled cRNA probes were purified on Sephadex G50/50 (Sigma) columns. The probes were eluted in 0.1 m Tris-HCl (pH 7.5), 12.5 mm Na₄EDTA, and 0.15 m NaCl.

Paraffin-embedded surgical blocks of various normal and malignant tissues were obtained either from Cooperative Human Tissue Network or from the Department of Pathology at the Medical College of Ohio. The tissues were sectioned and transferred onto polylysine (Sigma)-coated slides. The sections were deparaffinized by heating at 65°C for 20–30 min, followed by sequential washes in xylene and absolute alcohol and air dried. Deproteination of the tissue sections was carried out with 40 μg/ml proteinase K (Boehringer Mannheim) in 100 mm Tris-HCl (pH 8) and 50 mm EDTA for 10 min at 37°C, followed by a brief wash in distilled water. The slides were incubated for 10 min at room temperature in 0.1 m triethanolamine (TEA; Sigma) with 0.25% acetic anhydride (Sigma). The slides were then washed in 2× SSC buffer (0.3 m NaCl, 0.03 m sodium citrate; Sigma), followed by dehydration in graded ethanol (50% to 100%) and air dried.

cRNA probes labeled with [³⁵S]UTP and [³⁵S]CTP were diluted with hybridization buffer [50% formamide, 10% dextran sulfate, 3× SSC buffer, 50 mm sodium phosphate buffer (pH 7.4), 1× Denhardt’s solution, and 0.1 mg/ml yeast tRNA], and 2 × 10⁶ dpm of radioactive probe were used for each slide. Freshly prepared DTT (10 mm) was added to the hybridization buffer. The probe/hybridization mixture was applied to a coverslip, which was then applied on the section. The slides were placed on wet foam in sealed boxes and incubated at 55°C for 18 h.

The coverslips were removed from the slides, which were then washed twice in 2× SSC buffer. Sections were then incubated for 1 h at 37°C in 10 mm Tris-HCl (pH 8) containing 0.5 m NaCl and 200 μg/ml RNase A (Boehringer Mannheim). The sections were washed in graded SSC buffer (2× to 0.5×) and incubated in 0.1× SSC buffer for 1 h at 70°C, followed by dehydration (50–100% ethanol), and air dried.

The sections were initially placed on Kodak XAR-5 X-ray film and exposed at room temperature for ≈2 days before developing. The autorads were examined for the strength of the signal for β-actin. The cases that gave a strong specific signal for the antisense β-actin cRNA probe were dipped in Kodak NTB-2 emulsion and stored in desiccated light-tight boxes at 4°C. The emulsion-dipped sections were developed in Kodak EDP/EDF photochemicals and counterstained with Mayer’s H&E (Sigma). Slides containing contiguous sections were also counterstained with H&E without radioactive labeling or immunolabeling to facilitate inspection of cellular morphology.

The sections were inspected with a Nikon Labophot microscope, and 25–30 images were captured for each section using a ×40 objective lens and a Sony video camera. The captured images were analyzed using NIH Image 1.57 software and a Macintosh Quadra 840AV computer to determine the intensity of the signal (number of silver grains). Contiguous sections from each sample were probed with sense and antisense cRNA probes specific for FR-α, FR-β, FR-γ, and β-actin. Corresponding images from all of the slides were captured for analysis. The signal for the sense cRNA probes (negative control) was subtracted from the signals for the antisense cRNA probes. Specificity of the signals in the captured images was ensured by inspection of multiple sections on the same slide. The reactivity of the cRNA probes for FR was expressed in terms of the ratio to the signal for the β-actin probe.

The cRNA probes used to estimate mRNAs for the three FR isoforms, FR-α, FR-β, and FR-γ, were chosen for in situ hybridization experiments based on their specificity for the corresponding FR isoforms in RNase protection assays (results not shown). The lower limit of sensitivity of the cRNA probes for either the FR isoforms or for β-actin was established by measuring the signal for the sense (negative control) probes in ≈200 randomly chosen fields, including 40 randomly selected samples, and normalizing each value to the signal for the antisense cRNA probe for β-actin in the corresponding field. It was thus determined that the lower limit of detection of FR mRNA by in situ hybridization is represented by a value for the FR:β-actin ratio of 0.05 ± 0.02. As seen in the following sections, this limit of sensitivity afforded a good working definition of positivity for FR mRNA, because most of the samples analyzed either gave this background value or values that were significantly (at least 5-fold) above background (Table 1; Fig. 1). The quantitative analysis is also substantiated by the consistent trends in the FR expression pattern observed for each subset of normal and pathological tissue studied. Because significant expression of FR-β or FR-γ was not observed in the normal and malignant tissues included in this study, the expression patterns of FR-α alone are discussed.

Among the different cell types of the normal ovary, FR-α expression was detected only in germinal epithelium (Table 1; Figs. 1 and 2). Differentiation of the pluripotent germinal epithelium into benign mucinous or benign serous lesions caused down-regulation of FR-α (Table 1), as demonstrated by the absence or relatively insignificant levels of FR-α mRNA. Malignant transformation of the germinal epithelium into mucinous cystadenocarcinoma or clear cell carcinoma also resulted in down-regulation of FR-α (Table 1). However, high receptor mRNA levels were observed in all of the serous cystadenocarcinomas examined (Table 1; Fig. 1). Endometrioid adenocarcinomas of the ovary (Table 1; Fig. 1) showed moderate to high FR-α expression with the exception of one of five samples, which was distinguished as a well-differentiated tumor containing mostly stromal tissue. Thus, FR-α expression in the germinal epithelium is either retained or down-regulated, following specific pathways of differentiation or malignant transformation.

It was also observed that in both serous cystadenocarcinomas (Fig. 3) and endometrioid adenocarcinomas (results not shown), areas of the tumors that appeared to be poorly differentiated expressed higher levels of the receptor, in contrast to well-differentiated foci in the same sections. These observations suggest a phenotype basis for heterogeneity in the receptor expression in FR-α-positive ovarian tumors. The results are consistent with the flow cytometric analysis of ovarian tumors by Toffoli et al. (36), in which a positive correlation was established between FR expression and high histological grade, advanced stage, aneuploid status, and percentage of S-phase cells.

Expression of FR-α in the normal uterus was restricted to epithelial cells, but curiously, there was a clear distinction between the surface epithelium and the glandular epithelium. FR-α was only detected in the surface epithelium (Table 2; Figs. 1 and 4). The segregation between surface and glandular epithelia in terms of FR-α expression also extended to benign lesions of the uterine endometrium (Table 2).

Uterine epithelial malignancies of small cell, clear cell, and squamous cell differentiation did not express significant amounts of FR-α (Table 2). Among endometrial adenocarcinomas, three of nine specimens that were distinguished by a relatively high level of differentiation and a predominant stromal content were FR-α negative (Table 2), whereas the relatively poorly differentiated tumors showed a high level of FR-α mRNA (Fig. 1). In sections of the FR-α-positive endometrial tumors that showed morphological heterogeneity, relatively higher receptor expression was observed in the more poorly differentiated foci (results not shown), as seen above for ovarian cancers. Thus, in contrast to the ovary, the pattern of FR-α expression in tumors derived from the glandular epithelium of the uterus reflects malignancy-associated de novo receptor synthesis, restricted to a particular malignant cell phenotype. However, as in the ovary, heterogeneity in FR expression within a uterine tumor may be linked to apparent heterogeneity in relative differentiation. Benign or malignant lesions in the uterus, of either stromal or smooth muscle origin, did not express significant amounts of FR (Table 2).

In contrast to the uterus, glandular epithelium in normal cervix showed a high expression of FR-α mRNA (Table 3; Fig. 1). Whereas the columnar surface epithelium of the endocervix was also strongly FR-α positive (Table 3; Fig. 1, Fig. 5), the squamous epithelium of the exocervix did not show detectable levels of FR (Table 3). Squamous cell carcinomas of the cervix consistently showed an absence of FR-α (Table 3). The receptor expression in cervical adenocarcinoma was variable, with four of the five specimens tested showing relatively poor expression (Table 3; Fig. 1). Interestingly, sections of cervical adenocarcinoma containing both normal glandular epithelium and adenocarcinoma showed a high expression of the receptor in the normal epithelial cells contrasted by poor expression in the corresponding malignant neoplasm (Fig. 5). It would, therefore, appear that malignant transformation of the glandular epithelium of the cervix may be associated with down-regulation of FR-α.

The results discussed in the preceding sections demonstrate that the ovary, the uterus, and the cervix have unique and contrasting expression patterns of FR-α in relation to both normal differentiation and pathogenesis. In all three organs, the receptor exhibits a very narrow cell phenotype specificity in both normal and malignant tissues. The quantitative analysis of FR-α mRNA in this study demonstrates regulatory patterns that were not apparent, based upon the immunocytochemical studies reported previously (31, 32, 33, 34, 35).

The earliest and the most exhaustive report of the distribution of FR-α in normal and malignant tissues of the female genital tract is the immunocytochemical study of Veggian et al. (31). The same laboratory had previously (37) developed two monoclonal antibodies, termed MOv18 and MOv19, that recognized a glycoprotein antigen, subsequently termed GP38 (38) in ovarian carcinoma. GP38 was found to be identical to FR-α. Veggian et al. (31) used MOv18 to determine the tissue distribution of FR-α. The present study contradicts several of their major observations: (a) Veggian et al. (31) reported an absence of FR-α in the normal ovary and therefore associated the expression of FR-α in serous cystadenocarcinoma and endometrioid carcinoma with a derepression event. In contrast, our study demonstrates a relatively high expression of FR-α in the germinal epithelium of the ovary, implying down-regulation of the receptor during transformation into mucinous adenocarcinoma but continued or increased expression of FR-α in serous and most endometrioid carcinomas; (b) whereas our study is in agreement with that of Veggian et al. (31) in concluding that derepression of FR-α expression occurs during neoplastic transformation of the glandular epithelium of the uterus, they did not observe the expression of FR-α in the surface epithelium of the normal uterine endometrium, as reported here; (c) Veggian et al. (31) did not observe FR-α expression in the endocervical surface epithelium, in contrast to the significant expression observed in both the glandular and surface epithelium of the cervix in the present study, which implies possible down-regulation of FR-α in cervical adenocarcinoma; (d) Veggian et al. (31) reported FR-α expression in squamous carcinomas of the cervix, contrary to the lack of significant receptor expression quantitated in squamous carcinomas in our study; and (e) the quantitative analysis in this study demonstrates differential expression of FR-α in morphologically distinct foci of ovarian and uterine tumors, an observation not reported by Veggian et al. (31). Subsequent immunohistological findings using either MOv18 and/or Mov19 disagree in part with either Veggian et al. (31) or our present study. Thus, Boerman et al. (33) reported FR-α expression in mucinous adenocarcinoma of the ovary and its absence in squamous carcinoma of the cervix; Stein et al. (32) described apparent MOv18 reactivity in the normal epithelium of both the uterus and the endocervix. Weitman et al. (39) did not observe FR expression in the surface epithelium of normal ovary but observed staining of epithelial cells of the endometrium. The results of Buist et al. (34) are in agreement with the present study in respect to FR expression in normal ovary and normal uterus but contradict those of Garin-Chesa et al. (35). Garin-Chesa et al. (35) also reported a lack of association between FR expression and histological type and grade in ovarian carcinomas. The present study shows a quantitative increase in FR expression associated with foci of poor differentiation within a tumor in both ovarian and uterine carcinomas that express FR-α. The latter results are consistent with the flow cytometric analysis of ovarian carcinoma by Toffoli et al. (36). An association between FR expression and differentiation state was also reported in a squamous cell carcinoma cell line (40), but interestingly, those cells expressed more receptors when they were induced to differentiate.

Several methodological limitations are associated with the immunocytochemical studies discussed above using MOv18/MOv19 antibodies; they include both inherent limitations of immunocytochemistry and technical problems associated with the MOv18 and MOv19 monoclonal antibodies as described by Stein et al. (32). The qualitative nature of immunocytochemical methods and the subjective evaluation of positive antibody reactivity without considering relative cell density may, in large part, account for differing results. The reactivity of MOv18/MOv19 is known to be influenced by sample preparation; these antibodies are only reactive with unfixed cryosections (31) in which many morphological features are obscured, possibly contributing to the discrepancies between studies. In some instances (32), reacting a section with MOv18 and MOv19 yielded contradicting results, suggesting that these antibodies could potentially react with nontarget protein. The present study offers an alternative approach that circumvents the aforementioned problems by quantitating the FR-α-specific transcript relative to an internal standard (β-actin mRNA) in paraffin-embedded sections in which cellular morphology is relatively well preserved. Previous studies from this laboratory (6) have indicated that the relative levels of FR mRNA in normal and malignant tissues reflect relative protein levels.

This study, which demonstrates expression of FR-α in normal progenitor cells and the progressive changes of its regulation associated with malignant transformation, has obvious clinical implications. Applications of FR-α as a serum marker, for tumor imaging (22, 23, 24, 25), and in determination of prognostic value may be developed. It will also aid in therapeutic decision making regarding FR-mediated tumor targeting of specific tumor phenotypes. In some instances, FR-α may also serve as a marker to identify the primary site of origin of metastatic tumor. At the molecular level, the results of this study provide an essential basis for additional studies of the mechanisms determining the narrow tissue specificity of FR-α and its overexpression in certain malignant cell types. The data reported here are essential for the extrapolation of specific regulatory mechanisms established in vitro to in vivo situations.

We thank Jenny Zak and Ann Chlebowski for typing the manuscript. We are grateful to Dr. James Watson at the University of Michigan, Ann Arbor, MI, for initially teaching us the in situ hybridization technique. Most of the tissue specimens were obtained through the Cooperative Human Tissue Network.