Expression of nuclear retinoid receptors in normal, premalignant and malignant gastric tissues determined by in situ hybridization

Retinoids exhibit multiple functions through interaction with nuclear retinoid receptors and have growth-suppressive activity on gastric cancer cells. To better understand the roles of nuclear retinoid receptors during gastric carcinogenesis, we have used in situ hybridization to investigate expression of retinoic acid receptors (RARs) and retinoid x receptors (RXRs) in premalignant and malignant formalin-fixed paraffin-embedded gastric tissues. Histological sections of eight normal, 17 distal normal and nine gastric cancer tissues were hybridized with non-radioactive RNA probes for subtypes of RAR and RXR. Expression of RARα, RARβ, RARγ, RXRα and RXRβ was found in most cell types in gastric mucosa tissues from normal individuals as well as in distal normal tissues from cancer patients. Expression of RARα and RARβ were found in three and seven cancer tissues, respectively, and levels of RXRα mRNA were significantly decreased in poorly differentiated cancer tissues. Among the five investigated nuclear retinoid receptors, only expression of RARα mRNA was significantly decreased in intestinal metaplasia, dysplasia and cancer tissues when compared to adjacent normal tissues. In conclusion, normal gastric mucosa expressed both RARs and RXRs, which supports the physiological role of retinoic acid on normal gastric mucosa. The decrease in RARα expression in premalignant and malignant gastric tissues suggests a significant role of RARα during gastric carcinogenesis. © 1999 Cancer Research Campaign

nuclear retinoid receptor may be altered distinctly in different tissues during carcinogenesis. This is further evidence that each receptor subtype has a unique function in different tissues.
The role of retinoids in gastric cancer treatment and prevention has been studied previously. Epidemiological and animal studies have demonstrated the activities of retinoids in prevention (Haenszel et al, 1985;Miasoedov et al, 1989) and treatment (Fujii et al, 1991) of gastric cancer. These studies support our recent observation of the growth suppressive activity of all trans RA and 13-cis RA on gastric cancer cells in vitro and in vivo (Shyu et al, 1995;Jiang et al, 1996). To further understand the roles of nuclear retinoid receptors in gastric mucosa, we have used an in situ hybridization method to detect expression of subtypes of RAR and RXR in histological sections of formalin-fixed, paraffinembedded gastric tissues from normal individuals and gastric cancer patients.

Specimen collection and preparation
Four pairs of biopsy specimens from the gastric body and antrum regions of normal individuals undergoing health check-ups were obtained using a pandoscope. In addition, nine gastric cancer tissues and 17 distal normal tissues (eight from the body and nine from the antrum regions) from nine patients were obtained by pandoscope. Tissues were fixed in 10% neutral formalin and embedded in paraffin. The specimens were sliced into 4-µm-thick sections. Haematoxylin and eosin (H&E)-stained tissue sections were screened by the same pathologist to identify adjacent normal tissues, intestinal metaplasia, dysplasia and carcinoma. Gastric cancer tissues were classified based on the criteria of World Health Organization (Watanabe et al, 1989).

Subcloning of RAR and RXR cDNA
The 1.9 kb of human RARα (Petkovich et al, 1987), 1.9 kb of human RXRα (Mangelsdorf et al, 1990) and 2.2 kb of human RXRβ (Hamada et al, 1989) cDNA fragments were cloned into the Eco RI site of the plasmid pBSK + / -(Stratagene, La Jolla, CA, USA) through blunt end ligation. The 1.6 kb of human RARβ (Brand et al, 1988) and 1.7 kb of human RARγ (Zelent et al, 1989) cDNA fragments were cloned into the Not I site of the plasmid pRC/CMV (Invitrogene Co., San Diego, CA, USA) through blunt end ligation.

Preparation of digoxigenin-labelled RNA probes
The digoxigenin-labelled RNA probes spanning the entire openreading frames of cDNAs of RARα, RARβ, RARγ, RXRα or RXRβ cDNA were synthesized using an in vitro transcription kit obtained from Boehringer Mannheim (Germany) (Xu et al, 1994b). The RNA probes were precipitated, washed and dissolved in diethylpyrocarbonate-treated water containing RNAsin. The concentration was adjusted to 100 ng ml −1 using a DNA Dipstick Kit (Invitrogene) and stored at -70°C. The length of the RNA probes was confirmed by RNA gel electrophoresis followed by visualization using the fluorescence substrate CSPD as described below. The specificity of the probes was determined by Northern blotting.

RNA isolation and Northern blotting
SC-M1 and TSGH9201 human gastric cancer cell lines cells were grown in T75 cm 2 flasks in RPMI-1640 medium supplemented with 5% fetal bovine serum and antibiotics in a humidified atmosphere of 5% carbon dioxide and 95% air at 37°C. Polyadenosine (poly-A) + RNA was purified from cellular lysates using oligo dT cellulose as described by Badley and co-workers (Badley et al, 1987). RNA was then fractionated on a 1.1% agarose, 1.1% formaldehyde gel in 5 mM NaOAc, 1 mM EDTA, 20 mM 3-[Nmorpholino]propanesulphonic acid, pH 7.0 and transferred to a nylon membrane (Boehringer Mannheim) by capillary blotting in 20 × saline-sodium citrate (SSC; 3 M NaCl, 0.3 M Na 3 citrate, pH 7.0). Blots were UV-fixed, prehybridized and hybridized at 68°C in buffer containing 50% (v/v) formamide, 5 × SSC, 2% (w/v) blocking reagent (Boehringer Mannheim), 0.1% N-lauroylsarcosine and 0.2% (w/v) sodium dodecyl sulphate (SDS). The membranes were washed first with 2 × SSC containing 0.1% SDS and then washed with 0.1 × SSC containing 0.1% SDS at 68°C for 30 min. Specific hybridization was detected by a DIG luminescence detection kit using CSPD as the substrate and was recorded using Kodak XAR-5 film at room temperature. Probes were removed and membranes were then hybridized with the digoxigenin-labelled probe for glyceraldehyde-3-phosphate dehydrogenase (GAPDH).

In situ hybridization
A nonradioactive in situ hybridization using digoxigenin-labelled riboprobes was used as described by Xu and co-workers (Xu et al, 1994b). Briefly, sections were deparaffinized, rehydrated and deproteinized by digestion with protease K. The slides were fixed with 4% paraformaldehyde and then acetylated in 0.25% acetic anhydride in 0.1 M triethanolamine-HCl buffer. After washing and dehydration, the slides were prehybridized for more than 1 h at 42°C with a hybridization solution containing 50% deionized formamide, 2 × SSC, 2 × Denhardt's solution (0.02% Ficoll 400, 0.02% polyvinylpyrrolidone, 0.02% bovine serum albumin), 10% dextran sulphate, 400 µg ml −1 yeast tRNA, 250 µg ml −1 salmon sperm DNA and 20 mM dithiothreitol in diethylpyrocarbonatetreated water. Then, the slides were incubated with a 50 µl per slide hybridization solution containing 100 ng freshly denatured digoxigenin-labelled RNA probes at 42°C for 16 h in closed humid containers. After incubating with RNase A, the slides were washed and then blocked for non-specific binding using 2% normal sheep serum. The slides were incubated overnight at 4°C with sheep anti-digoxigenin antibody, and the in situ hybridization signal was visualized by incubating the slides in a chromogen solution containing nitroblue tetrazolium and X-phosphate (Boehringer Mannheim). Slides were observed under an AH-1 light microscope (Olympus, Japan) and then stored dark at 4°C.

Reviewing and scoring the sections
All sections from the same patient were stained on the same day with the same reagents to ensure a proper comparison of the different sections. The sections were reviewed by three independent researchers, including a pathologist. The staining of the sections was assigned scores ranging from 0 to 4, representing no staining (0), weak (1 +), positive (2 +), strongly positive (3 +), or very strongly positive (4 +). The intensity of staining in the surface mucus cells, mucus neck cells, parietal cells and chief cells was analysed in the body tissues. The intensity of staining in the surface mucus cells and regenerative and mature gland cells in the antrum tissues was analysed. Correlation of RXRα and RXRβ expression between cancer tissues with poor vs non-poor differentiation was analysed by Fisher's Exact Test. Differential expression of RARα mRNA in adjacent normal, intestinal metaplasia adjacent to gastric cancer, dysplasia adjacent to gastric cancer and gastric cancer tissues from different or the same gastric cancer patients was analysed by logistic regression. However, the estimated coefficients and their correspondent variances were obtained through the generalized estimating equations method (Liang and Zeger, 1986) and the SAS/IML macro program, GEE1.SAS, to take into account the within subject correlation.

Activity and specificity of digoxigenin-labelled RNA probes
The activity of digoxigenin-labelled probes was first tested by Northern hybridization on gastric cancer cells. Antisense RNA probes of each RAR or RXR subtype could detect specific mRNA subtypes of RAR and RXR ( Figure 1). The molecular weights of each subtype of RAR or RXR mRNA transcript in SC-M1 or TSGH9201 gastric cancer cells detected were similar to results described previously (Shyu et al, 1995). Therefore, the binding of antisense RNA probes was specific. No band was shown in the mRNA samples hybridized with five control sense RNA probes.

Staining of RAR and RXR mRNA in tissues
Expression of the mRNAs for RAR and RXRs in gastric cancer tissues was analysed by non-radioactive in situ hybridization. Figure 2 shows the mRNA localization of RARα, RARβ, RARγ, RXRα and RXRβ in consecutive sections of a specimen from the body region with moderately differentiated gastric cancer by digoxigenin-labelled RNA probes. Positive results appeared as a dark purple colour in the cytoplasm, where mRNA is expected to be localized. The adjacent normal gland had the highest expression of RXRα mRNA and the lowest expression of RARβ mRNA (see below). The sense probes did not bind to the adjacent sections, indicating that hybridization with the antisense probes was specific.

Expression of RARs and RXRs in non-malignant gastric body and antrum tissues
Four pairs of body and antrum tissues from individuals without cancer were analysed for in situ expression of nuclear retinoid receptor mRNAs. H&E staining showed that three specimens exhibited features of gastritis and one exhibited atrophic gastritis in both antrum and body tissues. In addition to gastritis, intestinal metaplasia and dysplasia were also found in two antrum tissues. Expression of RAR and RXR mRNAs in areas without intestinal metaplasia and dysplasia was analysed.
Expression of mRNAs of nuclear retinoid receptors in surface mucus, mucus neck, parietal and chief cells was analysed in specimens derived from the body tissues. In the antrum tissues, expression of nuclear retinoid receptors in surface mucus, regenerative gland and mature gland cells was analysed. Most cell types analysed from both body and antrum tissues expressed five subtypes of nuclear retinoid receptors (data not shown). No cell type-specific expression of a unique subtype of RAR or RXR was observed. However, mRNA of RXRα was found to express at levels equivalent to or higher than that of the other four tested nuclear retinoid receptors in all analysed cell types. Also, RARα mRNA appeared to be more abundant than RARβ or RARγ in some specimens. Similar results were obtained in eight distal normal body and nine distal normal antrum tissues from patients with gastric cancer.

Expression of RARs and RXRs in gastric cancer tissues
Nine gastric cancer specimens, four derived from the body region and five derived from the antrum region, were analysed. All tissues exhibited the morphology of adenocarcinoma. Two specimens exhibited the morphology of both tubular and papillary tumours. Four specimens were the tubular type and three were the signet-ring cell type (Table 1). Among nine investigated cancer tissues, RARγ, RXRα and RXRβ were expressed in all nine, while RARα and RARβ were expressed in three and seven cancer tissues respectively. Among tissues found to express RARα, two were well-differentiated and one was moderately differentiated. Both papillary and tubular types of adenocarcinoma cells were found to stain positively for RARα mRNA in the two cancer tissues that exhibited the morphology of both. None of two cancer tissues with poor differentiation expressed RARα. Among six tissues without RARα expression, three exhibited the morphology of tubular type and the other three were the signet-ring cell type tumours. Well and moderately differentiated gastric adenocarcinoma expressed higher levels of RXRα mRNA than the poorly differentiated adenocarcinoma, and the difference was significant (P = 0.03). A similar trend, although not a significant difference (P = 0.083), was observed for RXRβ.

Differential expression of RARs and RXRs in distal normal tissues, intestinal metaplasia adjacent to gastric cancer, and dysplasia adjacent to gastric cancer tissues from different gastric cancer patients
Adjacent normal tissues, intestinal metaplasia, dysplasia and cancer tissues from the same tissue section of different gastric cancer patients were compared. Most of the gastric cancer cells were believed to be derived from neck cells which exhibited    a Compared with adjacent mucus neck cells in body tissues or adjacent regenerative gland in antrum tissues. b A > C: expression of mRNA was greater in adjacent tissues (AT) than in cancer (C). c A = C: expression of mRNA was similar in adjacent tissues (AT) than in cancer (C). d A < C: expression of mRNA was lower in adjacent tissues (AT) than in cancer (C). e P < 0.05; odds ratio (OR) = 0.06; 95% confidence Interval (CI), (0.01, 0.57). f P < 0.05; OR = 0.02, 95% CI, (0.00, 0.19).
highly proliferative potential. Therefore, mucus neck cells of the body tissues or the regenerative gland cells of the antrum tissues were chosen to compare to pre-neoplastic and cancer cells in levels of RAR and RXR expression. Representative staining of the tissues containing adjacent normal tissues as well as moderately differentiated or well-differentiated adenocarcinoma is shown in Figure 2 and Figure 3 respectively. In Figure 2, mucus neck cells in the tissue expressed RARα with an intensity of '++' (Figure 2A). However, the adjacent cancer cells did not express RARα. RARβ was not expressed in either normal or cancer cells ( Figure 2C). RARγ was expressed in both normal and cancer cells with an intensity of '++' (Figure 2E). RXRα and RXRβ were expressed in both normal and cancer cells with an intensity of '++++' in the adjacent tissues and '+++' in the cancer tissues ( Figure 2G, I).
Similarly, adjacent normal and cancer tissues from a specimen with well-differentiated gastric adenocarcinoma were stained positive for RARβ, RARγ, RXRα and RXRβ mRNAs (Figure 3). RARα mRNA was expressed in the adjacent normal tissue. However, it was not expressed in the cancer tissues. Figure 4 shows the representative RARα mRNA staining of the atrophic gastritis tissue with the morphology of normal glands, intestinal metaplasia and dysplasia. Expression of RARα was found only in adjacent normal glands but not in intestinal metaplasia and dysplasia. Table 2 summarizes the frequency of RAR and RXR mRNA expression from eight cancer tissues. One cancer tissue was not analysed due to a lack of adjacent non-malignant tissue. Mucus neck cells from body tissues and regenerative gland cells from the antrum tissues of normal individuals described above, and distal normal tissues obtained from cancer patients, expressed all five nuclear retinoid receptors (Table 2). In the cancer tissues, only two among eight specimens stained positive for RARα mRNA. Frequency of RARα expression in tissues with intestinal metaplasia, dysplasia or cancer was significantly lower than in the adjacent normal tissues (P < 0.05; Table 2). Loss of RARα expression was correlated with an increase in the morphological features from normal to cancer. Only one specimen did not have RARβ expression in the adjacent normal tissues, intestinal metaplasia, dysplasia and cancerous tissues. All eight cancer specimens with all histological features expressed RARγ, RXRα and RXRβ.

Expression of RARs and RXRs in adjacent normal tissue, intestinal dysplasia and cancer lesions from the same gastric cancer patient
The expression levels of RARs and RXRs in normal, premalignant and malignant tissues within the same slide section are compared ( Table 3). Levels of RARα mRNA in adjacent normal, intestinal metaplasia and dysplasia tissues were found to be higher than those in cancer tissues in seven, two and two cases respectively. Significantly more cases of adjacent normal tissues than intestinal metaplasia and dysplasia had levels of RARα mRNA expression higher than that of cancer cells (P < 0.05). No similar result was found for the other four nuclear retinoid receptors. mRNA levels of RARβ and RXRα in adjacent normal, intestinal metaplasia and dysplasia tissues were similar to cancer lesions in all cases. RARγ and RXRβ mRNA expression in adjacent normal tissues in two and three cases, respectively, was found to be greater than that in cancer lesions. However, most intestinal metaplasia, dysplasia and cancer lesions had similar levels of RaRγ mRNA.

DISCUSSION
This study analysed in situ expression of nuclear retinoid receptors in premalignant and malignant gastric tissues. Normal gastric mucosa tissues expressed all five nuclear retinoid receptors, and levels of RXRα mRNA were the highest among the five investigated nuclear retinoid receptors in both normal gastric mucosa and cancer tissues. No differences in expression of RARβ, RARγ, RXRα and RXRβ between normal tissues and cancer lesions were observed. However, expression of RARα was significantly decreased in intestinal metaplasia, dysplasia and cancer tissues.
Our data showed that tissues of normal gastric mucosa from the body and antrum regions expressed both RAR and RXR. This result suggests that the formation of a RAR/RXR heterodimer in the presence of RA may play a physiological role in the growth and differentiation of gastric mucosa. Alteration in levels of RA or nuclear retinoid receptors may perturb cell growth or differentiation and may have an impact on gastric carcinogenesis. This is supported by the finding that patients with gastric cancer have lower levels of serum vitamin A than normal controls (Miasoedov et al, 1989), and that individuals with low serum β-carotene levels have an increased risk for the development of gastric cancer (Haenszel et al, 1985). Among the five receptor subtypes, RXRα mRNA appeared to express at the highest levels in various cell types of gastric mucosa. RXR can regulate gene expression through the formation of a RXR/RXR homodimer in response to 9-cis RA (Zhang et al, 1992). The receptor can also mediate signal transduction of retinoids, thyroid hormone and vitamin D through heterodimer formation between RXR and RAR, thyroid hormone receptor or vitamin D receptor (Lablanc and Stunnenbery, 1995). In addition, other receptors such as peroxisome proliferative activating receptor and orphan receptors are known to form heterodimers with RXR. Due to generally higher levels of RXRα expression than RAR in normal gastric mucosa, hormones such as thyroid hormone or vitamin D may, therefore, also play a physiological role in gastric mucosa. The expression of RAR and RXR subtypes in normal and cancer tissues has been investigated in various tissues. Expression of RARα is decreased in tissues of squamous cell carcinoma (Issing and Wustrow, 1996) and mouse skin tumours (Darwiche et al, 1995(Darwiche et al, , 1996, and the decrease is associated with activation of the protein kinase C or H-ras oncogene in mouse skin tumours (Darwiche et al, 1996). RARβ expression is decreased in tissues of the head and neck (Xu et al, 1994a), breast (Xu et al, 1997b;Widschwendter et al, 1997), larynx (Castillo et al, 1997) and nonsmall-cell lung cancers (Xu et al, 1997a). A decrease in RARγ expression is found in mouse skin tumours (Darwiche et al, 1995(Darwiche et al, , 1996 and tissues of non-small-cell lung cancer (Xu et al, 1997b). An increase in RXRα and a decrease in RXRβ expression is found in tumours of skin (Darwiche et al, 1995) and non-small-cell lung cancer (Xu et al, 1997a) respectively. The difference in alteration of RAR and RXR subtypes among various tumours may be associated with tissue-specific functions of each retinoid receptor subtype as well as differences in mechanisms of carcinogenesis among tumours. In this study, we only observed a significant decrease in RARα expression in premalignant and malignant gastric tissues among the five investigated nuclear retinoid receptors. Mutation of K-and H-ras genes is detected, but uncommon, in some cancer tissues and is rarely detected in precancerous lesions (Soman et al, 1991;Tsuchiya et al, 1997). However, overexpression of K-and H-ras proteins are observed much more frequently in both premalignant and malignant gastric tissues (Czerniak et al, 1989). It is therefore possible that the decrease in RARα expression observed in gastric tissues may be related to the activation of the ras oncogene as observed in mouse skin tumours (Darwiche et al, 1996). However, ras mutation or overexpression is preferentially associated with the development of the welldifferentiated type of gastric cancer (Tahara, 1993;Yoshida et al, 1988), which is inconsistent with the decrease in RARα expression in poorly differentiated gastric cancer with signet-ring cell type, as observed in this study. It is therefore likely that other genetic alterations, in addition to ras or PKC, may also contribute to the abnormal RARα expression in gastric tissues.
Our study observed a decrease in RARα expression in both premalignant and malignant gastric tissues. Similarly, a decrease in RARα, RARβ or RARγ expression is observed in premalignant tissues of the breast, oral, head and neck and skin (Xu et al, 1994a;Darwiche et al, 1995Darwiche et al, , 1996Lotan et al, 1995;Widschewendter et al, 1997). Furthermore, McGregor et al (1997) have shown that the loss of RARβ expression occurs during the transition from senescent to immortal phenotypes. These data suggest that the decrease or loss of expression of nuclear retinoid receptors may occur before neoplastic transformation. The alteration may be reversible. Treatment with 13-cis retinoic acid can restore RARβ expression in premalignant oral tissues, and the restoration is correlated with a clinical response (Lotan et al, 1995). Therefore, some of the cancer-preventive activities of retinoids are mediated, first, through restoring the expression of nuclear retinoid receptors and, second, through bringing back the growth and differentiation control by retinoids in precancerous lesions. N-4-hydroxyphenyl retinamide has been found to prevent the progression of gastric dysplasia into cancer (Han, 1993). Whether the effect is associated with restoration of RARα expression requires further study.
During the process of gastric carcinogenesis, telomere reduction leading to genomic instability appears to occur at the earliest step (Tahara, 1993). Alterations in genes such as K-and H-ras, p53, c-met and tpr-met occur later in precancerous lesions. Activation of c-erbB and c-met oncogenes and loss of expression of genes like cadherin, DCC, transforming growth factor β and type I receptor of transforming growth factor β are involved in the later stages of gastric cancer. We observed a decrease in RARα expression in intestinal metaplasia, dysplasia and cancer tissues. Abnormality in expression of RARα appeared to occur at a similar stage to that seen in genes such as K-and H-ras, p53, c-met and trp-met. Currently, the possible link between these genes and RAR has not been investigated and requires further study.
In summary, this study observed the expression of RAR and RXR in normal gastric tissues, indicating the physiological function of RA on gastric mucosa. The decrease in expression of RARα in premalignant and malignant gastric tissues suggests that RARα has a significant role during processes of gastric carcinogenesis. Investigating the mechanism of the loss of RARα expression may provide further insight regarding the mechanism of growth control of gastric cancer cells.