Background: Retinoids, analogues of vitamin A, are required for the normal growth and differentiation of human bronchial epithelium. They are also able to reverse premalignant lesions and prevent second primary tumors in some patients with non-small-cell lung cancer (NSCLC). These effects are thought to result from modulation of cell growth, differentiation, or apoptosis (programmed cell death). When certain retinoid receptors in the cell nucleus (i.e., retinoic acid receptors [RARs] and retinoid X receptors [RXRs]), which mediate most retinoid actions, are suppressed, abnormal activity may result that could enhance cancer development. Purpose: This study was designed to determine whether there are abnormalities in the expression of retinoid receptors in surgical specimens from patients with NSCLC. Methods: Transcripts of nuclear retinoid receptors were detected in formalin-fixed, paraffin-embedded specimens by use of digoxigenin-labeled riboprobes specific for RARα, RARβ, RARγ, RXRα, RXRβ, and RXRγ for in situ hybridization to histologic specimens from 79 patients with NSCLC and as control from 17 patients with non-lung cancer. The quality and specificity of the digoxigenin-labeled probes were determined by northern blotting, and the specificity of the binding of antisense riboprobes was verified by use of sense probes as controls. Results: All receptors were expressed in at least 89% of control normal bronchial tissue specimens from 17 patients without a primary lung cancer and in distant normal bronchus specimens from patients with NSCLC. RARα, RXRα, and RXRγ were expressed in more than 95% of the NSCLC specimens. In contrast, RARβ, RARγ, and RXRβ EXPRESSION WAS DETECTED IN ONLY 42%, 72%, and 76% of NSCLC, respectively. Conclusions: These data suggest that the expression of RARα, RXRα, and RXRγ is not altered in NSCLC; however, expression of RAR and possibly also of RAR and RXR is suppressed in a large percentage of patients with lung cancer. Implications: The loss of expression of one or more of these nuclear retinoid receptors may be associated with lung carcinogenesis. [J Natl Cancer Inst 1997;89:624-9]
Lung cancer is still the leading cause of cancer death. The incidence of this cancer continues to increase in both men and women, and mortality from lung cancer has surpassed that from breast cancer in women. It has been estimated that there will be 178 100 new cases and 160 400 deaths from lung cancers in the United States in 1997 ( 1 ) . Despite advances in therapy, the overall 5-year survival rate of patients with lung cancer is still under 13%. Therefore, the identification and use of novel approaches for the prevention and treatment of lung cancer are urgently needed. One such approach is to use retinoids, structural and functional analogues of vitamin A, for chemoprevention ( 2 ) . Retinoids are suitable for this strategy because they regulate differentiation in airway epithelium ( 3 ) and suppress carcinogenesis in a variety of animal models for lung cancer ( 4 , 5 ) . Interestingly, vitamin A deficiency has been associated with increased lung cancer incidence ( 6 ) . Furthermore, certain retinoids suppress premalignant oral lesions and prevent the development of second primary cancers among patients with head and neck and lung cancer ( 7 , 8 ) .
The regulation of cell growth and differentiation of normal, premalignant, and malignant cells by retinoids is thought to result from their effects on gene expression. These effects are mediated by nuclear retinoid receptors, which are ligand-activated transcription factors and members of the steroid hormone receptor superfamily ( 9-11 ) . Two types of receptors have been identified: retinoic acid (RA) receptors (RARs) and retinoid X receptors (RXRs), which differ in the sequence of the amino-and carboxyl-terminal domains and in retinoid-binding specificity. RXR-RAR heterodimers bind to specific DNA sequences— RA response elements (RAREs)— that are usually located in the 5 upstream regions of genes that are regulated by retinoids. Each receptor type includes three subtypes designated α, β, and γ, which exhibit specific and distinct spatial and temporal expression patterns during embryonal development and different distributions in adult tissues. These receptors are thought to regulate the expression of distinct genes ( 9-11 ) .
The association between vitamin A deficiency and cancer incidence suggests that retinoids are physiologic suppressors of carcinogenesis. The development of cancer may, therefore, require the initiated and premalignant cells to escape the control of natural retinoids. One way by which this can be accomplished is by aberrant expression of one or more of the retinoid receptors, which could result in an abrogated retinoid signaling. Indeed, several studies ( 12-17 ) have demonstrated that RAR expression is suppressed in cultured lung cancer cell lines and have suggested that the expression of this receptor is associated with suppression of cancer. The suppression of tumorigenicity in cultured human lung squamous cell carcinoma transfected with RAR supports this contention ( 18 ) . However, these results were obtained by use of cultured human lung cancer cell lines; neither their relationship to expression of receptors in normal lung epithelial mucosa cells nor their relevance to the in vivo status of receptors in lung cancer has been established.
This study was designed to determine whether abnormalities exist in the expression of retinoid receptors in surgical specimens from patients with non-small-cell lung cancer (NSCLC).
Materials and Methods
Specimens from 79 patients who had surgery from 1984 through 1992 for NSCLC (Table 1) were used in this study. Of the 79 specimens, 37 were obtained from the Department of Pathology, The University of Texas M. D. Anderson Cancer Center, Houston, and 42 were obtained from the Istituto Nazionale Tumori, Milan, Italy. These specimens were selected from pathology archival material to include, in addition to the tumor, adjacent normal and distant normal bronchial epithelial tissue to enable a comparison between receptor expression in tumor and nontumor tissues. Processing of these specimens included a routine fixation in 10% neutral formalin or Bouin'ss fixative and embedding in paraffin. All of the specimens were cut into 4-m sections and stained with hematoxylin- eosin for classification.
In addition to cancers, these samples also contained adjacent normal lung and bronchial epithelium (n 40) or distant normal bronchial epithelium (n 36). As normal lung controls, we used normal lung specimens that were obtained at the Istituto Nazionale Tumori from 17 patients (eight females and nine males) who had undergone lobectomy for lung metastases from primary cancers in other sites of the body (five breast cancers, six sarcomas, two kidney cancers, two melanomas, one laryngeal cancer, and one colon cancer).
In Situ Hybridization
A previously described method for nonradioactive in situ hybridization ( 19 ) was used without modifications to analyze nuclear retinoid receptors in formalin-fixed, paraffin-embedded histologic sections. The quality and specificity of the digoxigenin-labeled probes were determined by northern blotting, and the specificity of the binding of antisense riboprobes was verified by use of sense probes as controls ( 19 , 20 ) . To determine whether the expression of retinoid receptors is different between specimens from normal bronchus and a lung adenocarcinoma, we analyzed consecutive sections by in situ hybridization. To determine whether the expression of retinoid receptors is different between distant normal bronchial epithelium and NSCLC in the same patient, we used available paired specimens obtained after lobectomy from the resection margin of a subset of the 79 patients with NSCLC. To determine whether the expression of the nuclear retinoid receptors is different between the tumor and its surrounding adjacent normal bronchial tissue, we analyzed the expression of the receptors in specimens from available cases that contained both tissue types in the same histologic section.
Frequency and summary data are given whenever appropriate. The chi-squared test for equality of proportions between normal and tumor tissues was performed for association of each pair. The McNemar test was performed to determine the association between matched pairs, such as the association between distant or adjacent normal tissues and tumors. All P values were generated from two-sided statistical tests.
Expression of RARs and RXRs in Normal Bronchial Lung Tissue From Patients With Non-Lung Cancer and in NSCLC Specimens
An analysis of consecutive sections of specimens from normal bronchus and a lung adenocarcinoma revealed that antisense probes for all of the receptors (RARα, RARβ, RARγ, RXRα, RXRβ, and RXRγ) hybridized to epithelial cells in all sections (, A, D, G, J, M, and P), whereas the sense probes failed to hybridize to any of the sections (, B, E, H, K, N, and Q), indicating that the binding of the antisense probe is specific. The binding of the antisense probes to consecutive sections of a lung adenocarcinoma (, C, F, I, L, O, and R) was distinct from the binding to the normal bronchial mucosa in that RAR and RAR were either absent or barely detectable; RAR'ss staining intensity was lower than that in normal control specimens, and only minor changes were detected in the staining of RXRs in normal and carcinoma sections.
The results of an analysis of bronchial specimens from 17 patients without primary lung carcinoma (normal lung) and tumor specimens from 79 patients with NSCLC are shown in Table 2. The most dramatic finding was the loss of RARβ expression in 58% of the specimens (53% of squamous cell carcinomas [SCCs] and 63% of adenocarcinomas). The difference in RARβ expression between normal tissue and NSCLC was statistically significant. Furthermore, about 25% of the NSCLCs also lost RARγ and RXRβ expression. In contrast, the distribution of RARα, RXRα, and RXRγ expression was very similar in normal control and NSCLC specimens. There was no apparent relationship between the differentiation status of NSCLC and RARβ expression because six of 12 well-differentiated tumors expressed RARβ. Likewise, the loss of RARβ expression was not associated with the stage of disease when stage I was compared with stages II-IV, most likely because this event may have occurred already in premalignant lesions and was maintained throughout the carcinogenesis process, as we had observed in head and neck carcinogenesis ( 20 , 21 ) . Because only seven of the 79 patients with NSCLC were never smokers and three of them did lose RARβ expression, it was not possible to relate smoking status to RARβ expression. This lack of association was maintained, even after combining never smokers with remote former smokers (quit smoking =2 years ago).
Expression of RARs and RXRs in Paired Specimens of Distant Normal Bronchial Lung Tissue and NSCLC From the Same Patient
Table 3 shows that more than 90% of the distant normal specimens expressed all six retinoid receptor messenger RNAs (mRNAs), whereas only 47% of the corresponding tumors continued to express RARβ. In addition, about 39% of the tumors have lost RARγ expression. The differences in RARβ and RARγ expression between tumor and corresponding distant normal bronchial tissue were statistically significant. The in situ hybridization results of a representative case of SCC are shown in . All of the receptor mRNAs were detected in the distant normal tissue (, C, G, K, O, S, and W), whereas the SCCs failed to express RARβ and RARγ (, H and L, respectively).
Expression of RARs and RXRs in Paired Specimens of Adjacent Normal Bronchial Lung Tissue and NSCLC From the Same Patient
Table 3 shows that there is a statistically significant decrease in the expression of RAR, RAR, and RXR in the NSCLC tumor compared with the adjacent normal bronchial tissue, whereas the expression of the other three receptors was similar in the adjacent normal and malignant tissues. shows representative in situ hybridization results of the analysis of adjacent normal and malignant tissue from the same specimen. All receptors were expressed in the adjacent normal tissue (, A, E, I, M, Q, and U), whereas the SCC failed to express RAR and RAR and showed a lower staining of RXR and RXR (, F, J, R, and V). Because both the adjacent normal and tumor tissue were present in the same histologic section, the comparison of the receptor expression in these tissues is ideal, since the experiment is internally controlled for factors, such as fixation and processing for analysis.
The premise of our investigation is that one of the functions of physiologic retinoids is to prevent carcinogenesis by regulating the differentiation of epithelial tissues. This contention is especially relevant for airway epithelium because vitamin A deficiency results in replacement of mucociliary epithelium with keratinizing squamous epithelium (squamous metaplasia) and is associated with increased lung cancer incidence ( 6 ) . Because nuclear retinoid receptors mediate most of the effects of retinoids on gene expression, reduced expression of one or more of these receptors may enhance cancer development.
The study reported here is the first to use in situ hybridization for the analysis of the mRNAs for RARs and RXRs in surgical specimens of normal bronchus, normal bronchial lung tissues distant and adjacent to lung tumors, and NSCLC carcinomas. We have demonstrated that RAR expression is markedly decreased in more than 50% of both adenocarcinomas and SCCs, and that the expression of RAR and RXR was lost in some cases. These findings are similar to our previous demonstration of a decrease in the expression of this receptor in premalignant oral lesions, dysplastic lesions, and head and neck squamous cell carcinomas ( 20 , 21 ) . However, unlike what was observed in head and neck premalignant and malignant lesions, the lung carcinomas showed suppression of RAR and RXR in some of the cases. Loss of RAR appears to be the major defect during lung carcinogenesis and this loss did not seem to predispose the tissue to lose also RAR or vice versa because 17 (22.4%) of 76 of the cases had lost both receptors, 26 (34.2%) of 76 expressed both receptors, 29 (38.2%) of 76 had lost RAR but retained RAR expression, and only four (5.3%) of 76 had lost RAR but not RAR.
Previous studies ( 12-17 ) with cultured lung carcinoma cell lines have demonstrated a decreased expression of RAR.We confirmed that this also occurs in vivo by use of specimens from 79 patients with NSCLC. The former studies employed northern blotting of extracted RNA for assessment of RAR expression. Thus, Gebert et al. ( 12 ) found decreased RAR expression in at least 50% of 33 lung cancer cell lines and in 30% of nine lung cancer specimens.
Although decreased expression of RAR appears to be a common event in lung carcinogenesis, the underlying molecular mechanism is unclear. The RAR gene promoter includes a RARE, which can be activated by RXR-RAR heterodimers. However, many lung cancer cell lines exhibit defects in transcription of RAR-RARE, possibly because of inactivation or absence of trans -acting co-factors that are required for the transcription ( 16 , 17 ) . The RAR gene promoter also has a RARE, and RXR has been shown to be regulated by retinoids in embryonal carcinoma cells, so their decreased expression in some of the lung cancer cases may also reflect defects in transcription activation.
The involvement of retinoid receptors in the regulation of gene expression in tracheal epithelial cells ( 22 ) and in suppression of growth of transformed tracheal epithelial cells ( 23 ) has been demonstrated. Recent evidence implicated RAR specifically in tumor suppression in vivo; transfection of a human epidermoid lung cancer in vitro with an RAR expression vector resulted in decreased tumorigenicity in nude mice ( 18 ) . In addition, transgenic mice expressing antisense RAR2 developed carcinomas 14-18 months after birth ( 24 ) . We have recently reported that blocking the expression of RAR in head and neck SCC cells resulted in resistance to retinoid-induced growth inhibition and a loss of the ability of RA to exert anti-AP-1 activity ( 25 ) . Thus, the loss of expression of either RAR or RAR in lung tissue could enhance cancer development. Future studies will be directed toward identifying the genes that are regulated specifically by RAR to gain a better understanding of the pathway by which this receptor may suppress carcinogenesis.
Supported by Public Health Service grant CA68437 from the National Cancer Institute, National Institutes of Health, Department of Health and Human Services, by a grant from the Lamar Fleming and O. H. Davenport Fund awarded by the Texas Medical Center, Houston, and by grants from the Associazione Italiana Ricerca Cancro and Consiglio Nazionale delle Ricerce Special Project ARCRO.
We thank Susan Cweren for preparing some of the tissue sections, Ren Yu for assistance in statistical analysis, and Dr. H. A. Pierotti for support and critical discussions.
Manuscript received November 11, 1996; revised February 25, 1997; accepted February 27, 1997.
Journal of the National Cancer Institute Vol. 89, No. 9, May 7, 1997 ARTICLES 629