Introduction

Top Abstract Introduction Materials and Methods Results Discussion References

Lung cancer is the leading cause of cancer death in both males and females, accounting for nearly one in three cancer deaths in the United States (1). Previous efforts to reduce lung cancer mortality through earlier cancer detection by using a combined chest X-ray and sputum cytomorphology failed to show a significant reduction in lung cancer-related mortality (2). A pilot study of archival sputum specimens from a high-risk cohort identified a monoclonal antibody (mAb) that specifically reacted with "normal"-appearing bronchial epithelial cells from individuals who subsequently developed lung cancer, providing a more sensitive means for early detection of lung cancer (3). This antibody was later found to specifically recognize human heterogeneous nuclear ribonucleoprotein (hnRNP) A2/B1 (4). Preliminary results of prospective studies in two new high-risk lung cancer cohorts showed that immunocytochemical assessments of hnRNP A2/B1 expression is sputum specimens with the same antibody accurately predicted the outcome in 32 of 40 (80%) and in 69 of 94 (73.4%) subjects who subsequently developed lung cancer, whereas cytologic changes suggestive of lung cancer were found in less than 10% of the subjects (5). In that report, many eventual cancers were found in cases with hnRNP A2/B1 overexpression that showed minimal cytologic abnormalities during the initial examination (5). Subsequent studies in tissue samples with the same antibody, however, detected dynamic patterns of hnRNP A2/B1 expression during mammalian lung development (6), and also found hnRNP A2/B1 overexpression in both neoplastic and surrounding non-neoplastic lung tissues of patients with Stage I non-small cell lung cancer (7). In addition, overexpression of hnRNP A2/B1 was detected in 41% of normal-appearing and in 37% of metaplastic respiratory epithelial cells from multiple biopsy specimens of chronic smokers with metaplasia obtained from a previously published trial of 13-cis-retinoic acid chemoprevention for bronchial metaplasia (8). In that study hnRNP A2/ B1 expression status was not different between current smokers and recent former smokers.

Inasmuch as only about 1% of smokers with metaplasia would be expected to develop lung cancer, the frequent expression of hnRNP A2/B1 in the central airway is hard to reconcile with the reported degree of diagnostic precision of hnRNP A2/B1 expression in sputum specimens, raising an important question about the actual significance of hnRNP A2/B1 in the airway.

We have previously suggested that part of the diagnostic utility of immunocytochemical evaluation of bronchial epithelial cells recovered in the sputum compared with immunohistochemical evaluation of biopsies of intact bronchial epithelium related to the distinct biology of the types of cells recovered in the different specimens (8). The bronchial epithelial cells that comprise the "cells of interest" in the sputum are cells that exist as individual cells not connected to neighboring epithelial cells as is the case with the bronchial epithelial cells recovered from a bronchial biopsy. The solitary cells-of-interest from the sputum may represent a population of cells that are caught up in lung carcinogenesis related to the loss of contact with neighboring cells and the basement membrane, accounting for why such cells may be more diagnostically informative than the interconnected bronchial epithelial cells recovered with a bronchial biopsy.

In this study, our question was whether focusing on cells that overexpress hnRNP A2/B1 actually contributes to the identification of individuals in the process of developing a lung cancer. To evaluate the significance of hnRNP A2/B1 overexpression, we compared the frequency and pattern of several recognized markers of lung carcinogenesis, including microsatellite alterations (MA) and loss of heterozygosity (LOH), as well as clonality in phenotypically different respiratory epithelial cells that did or did not express high levels of hnRNP A2/B1. We conducted this analysis for two reasons: (1) MA and LOH are early signs of defective DNA repair or replication, which precede morphologic abnormalities and are changes suggestive of a tissue involved in carcinogenesis (9); (2) clonality analysis has been successfully applied to distinguish the neoplastic or reactive nature of several types of human tumors (15).

hnRNP A2/B1 appears not only to play important roles in messenger RNA (mRNA) processing but also to participate actively in post-transcriptional events (20), so that the overexpression of this protein could be causally involved in the cancer process. If the overexpression of hnRNP A2/B1 correlates with MA, LOH, and clonality, this association strengthens the possibility of hnRNP A2/ B1 involvement in the cancer process. Conversely, if no correlation is found between hnRNP A2/B1 overexpression and genetic alterations, it would then be unlikely that hnRNP A2/B1 expression was contributing to the process or identification of carcinogenesis.

	Materials and Methods

Top Abstract Introduction Materials and Methods Results Discussion References

Case Selection and Classification

In this initial study, the focus was to explore the correlation of hnRNP A2/B1 expression status with genetic alterations in tissues with defined histologic differentiation. To do this, paraffin-embedded respiratory tissue samples from 20 human subjects were used under an institutionally approved tissue procurement protocol. This included lung tissue from six cancer patients. To obtain a range of well-preserved non-neoplastic tissues, we used lung tissue harvested from lung transplants and autoptic tissues obtained from trauma victims. In the 12 cases from individuals whose lungs were harvested during the course of lung transplantation, and two cases from young victims who died of trauma- related causes, extensive clinical information was not available. In previous publications the expression of hnRNP A2/B1 was not related to either gender or type of lung cancer, so in this study our analysis was directed at the coincidence of lung cancer- related molecular events with hnRNP A2/B1 expression (3, 5). Serial sections of 5 to 7 µm were placed on poly-L-lysine-coated or positively charged microscopic slides. Sections were incubated overnight at 37°C, deparaffinized with three changes of xylene, washed with descending concentrations of ethanol and tap water, and stained with hematoxylin and eosin for morphologic assessment. The morphologic designations were determined by independent reviews at the National Cancer Institute and the Armed Forces Institute of Pathology using published criteria (20). Of these, six were normal-appearing; eight were hyperplastic bronchial and/or alveolar epithelium with no or mild atypia; two were well-differentiated, in situ lung carcinomas; and four were well-differentiated invasive lung carcinomas with mixed (endocrine, spindle, and squamous) cell types. A total of 11 cases contained both bronchial and alveolar regions, five contained only bronchial epithelium, and four contained only alveolar tissues. None of the tissues from the six cancer patients contained clear-cut normal epithelium, but all contained normal-appearing stromal element. Only two cases, one with normal-appearing and the other with hyperplastic alveolar tissue, were nonsmokers.

Immunohistochemical Staining

A previously characterized mouse mAb that recognizes human hnRNP A2/B1 was used (3). Detection kits with a biotinylated secondary antibody to mouse immunoglobulin G, streptavidin-conjugated alkaline phosphatase, and Fast Red chromogen, were purchased from BioGenex (San Ramon, CA). Immunostaining was performed using our published protocol (6). Briefly, deparaffinized sections were incubated with the primary antibody solution (4 µm/ml or 1:100 dilution) or normal serum (4 µg/ml) overnight at 4°C. After two 5-min washes in 1× phosphate-buffered saline (PBS) that followed each step, sections were sequentially incubated with the secondary antibody, streptavidin-conjugated alkaline phosphatase, and chromogen. After chromogen reaction, sections were counterstained with hematoxylin, briefly washed with tap and distilled water, and mounted in 3% gelatin mounting medium for microscopic evaluation. A staining distribution score (0: no immunoreactive cells; 1: 1 to 10% immunoreactive cells; 2: 11 to 50% immunoreactive cells; 3: 51 to 100% immunoreactive cells) and staining intensity score (0: nonimmunoreactive; 1: +; 2: ++; 3: +++) were obtained for each sample. Using the sum of these two scores, a staining index (SI) was established for each case. A given sample was considered immunoreactive when the SI > 2 (7, 8). The immunoreactive cases were subclassified as cytoplasmic reactive if  85% of the immunoreactive cells showed localization of hnRNP A2/B1 in their cytoplasm, or nuclear immunoreactive when  85% of the reactive cells displayed hnRNP A2/B1 immunoreactivity in their nuclei. Positive controls included two hnRNP A2/B1 immunoreactive bronchial and alveolar samples identified in previous studies (6, 7). Negative controls included substitution of the primary antibody with nonimmune serum, or omission of the primary or secondary antibody from the immunostaining sequence. Immunostaining for each of the samples was repeated three times under the same conditions.

Assessments for MA and LOH

The coverslips of immunostained sections were removed by soaking the slides in 50-to-60°C hot water. After a 3-to-5-min wash in warm tap water, the sections were soaked in 1× PBS (pH 7.4) containing 10% glycerin. For each case, multiple foci of epithelial cell clusters with and without hnRNP A2/B1 expression, and stromal elements near or far away from the epithelium were microdissected under a standard microscope. The dissected sample from each focus was placed in a separate tube and subjected to proteinase K digestion as previously described (16, 24). DNA extracts of these samples were assessed for MA and LOH by polymerase chain reaction (PCR) amplification with 14 polymorphic DNA markers, which were selected on the basis of the following criteria: (1) they were located on chromosomes that harbor proven or putative tumor suppressor genes; and (2) they were found to have high frequencies of MA and LOH in a variety of malignant and premalignant lung lesions (9). The primers were purchased from Research Genetics (Huntsville, AL) and labeled with fluorescent dyes, and are listed in Table 1. The cases selected for assessments of MA and LOH and the types and foci of the epithelial cells microdissected are listed in Table 2. Gene Amp PCR kits, Taq gold DNA polymerase, and DNA size standard were obtained from Perkin-Elmer (Foster City, CA). PCR amplification was carried out in a programmable thermal cycler (Perkin-Elmer) at the following settings: after a denaturation at 94°C for 14 min, the samples were amplified for 35 to 40 cycles at 94°C, 55 to 60°C, and 72°C, each for 1 min, with a final extension at 72°C for 10 min. Amplified products were subjected to electrophoresis in 5 to 6% polyacrylamide gels (Bio-Rad, Foster City, CA) and the signal was detected with an automated 377 DNA sequencer (Perkin- Elmer). The expected PCR products were located and examined by comparing the mobility of the predominant DNA bands in each case with that of the molecular weight size marker, and by comparing the mobility and intensity of the amplified specific DNA bands from the tumor cells with those or normal controls (normal stromal or epithelial cells). MA was defined as an addition or deletion of one or more repeat units that results in the shift of microsatellite alleles; LOH was defined as the complete absence or at least 95% reduction of one allele (16, 24).

Clonality Analysis

Specimens from all six female subjects, including three with invasive carcinoma, two with hyplastic epithelium, and one with normal-appearing epithelium, were analyzed for clonality. Clonality analysis was carried out by assessing the DNA methylation pattern with a pair of fluorescent dye-labeled DNA markers (5'-GCTGTGAAGGTTGCTGTTCCTCAT-3' and 5'-TCCAGAATCTGTTCCAGAGCGTGC-3') located on exon 1 of the human androgen receptor (HUMARA) gene (19). Rsa 1 and Hpa II enzymes were purchased from GIBCO BRL/Life Technologies (Gaithersburg, MD). The primers were purchased from Research Genetics, and the same procedures described in our previous studies (16, 24) were used with some modifications. Briefly, in each case, two or three different foci of epithelial cells with and without hnRNP A2/B1 expression and stromal cells near or far away (at least 15 mm) from the epithelium were separately collected into different tubes. Duplicates of DNA extracts from each case were mixed with Rsa 1 or Hpa II enzymes and incubated overnight at 42°C. After incubation, the mixtures were treated at 95°C for 10 min to inactivate the enzymes. Mixtures were then subjected to PCR amplification and electrophoresis as described earlier. Monoclonality was defined as the presence of only one DNA band after Hpa II digestion, compared with two bands in normal controls and in samples treated with the control enzyme Rsa I.

Production of the Gel Images for LOH Assessment and Clonality Analysis

The production of gel images was carried out according to the User's Manual (Perkin-Elmer), and used protocol modifications developed in our laboratory (25).

Statistical Analysis

Tests of the equality of the probability of a genetic alteration (MA or LOH) across samples and across DNA markers were conducted by exact computation of the Fisher-Freeman-Malton test. Because significant heterogeneity was found, the data were analyzed using a generalized estimating equations technique as implemented in the GENMOD procedures of SAS version 6.12 (SAS Institute, Cary, NC). For each sample, the determinations of genetic alterations in the 14 DNA markers were considered to be repeated binary observations with an exchangeable correlation structure. Noninformative results were treated as missing values. A logistic regression model for the probability of genetic alterations as a function of sample and marker was used. This model diverges for observed proportions of 0 or 1, and some markers and some samples had no alterations. Therefore, indicator variables for each sample and each marker could not be used; instead, the samples and the markers were classified by their observed marginal genetic alteration proportions P into categories of 0 < P < 0.2, 0.2 < P < 0.4, and so on. In one analysis, a marker proportion on the 0.2 boundary was assigned to the lower category to prevent model divergence. For each two-group comparison, the relative risk parameters were estimated separately within each group being compared, a linear contrast was formed by weighing each parameter by the frequency of its category within the group, and the contrast was tested by the likelihood ratio method. Analysis of the residuals showed consistency with the modeling assumptions.

	Results

Top Abstract Introduction Materials and Methods Results Discussion References

hnRNP A2/B1 Expression in Normal-Appearing and Hyperplastic Epithelial Cells

Distinct epithelial cells demonstrating immunoreactivity for hnRNP A2/B1 were detected in five or six (83.3%) normal-appearing, and in five of eight (62.5%) hyperplastic cases (Table 2). The immunoreactive cells were often homogeneously distributed in certain areas (Figure 1A), but clusters with immunoreactive or nonimmunoreactive cells were also commonly seen in 8 of 10 immunoreactive cases (Figure 1B). This hnRNP A2/B1 distribution pattern made it possible to separate the immunoreactive and nonreactive cells by microdissection, allowing for the evaluation of coexpression with the panel of markers for genetic instability.

View larger version (118K): [in this window] [in a new window]   Figure 1.   Expression of hnRNP A2/B1 in normal-appearing, hyperplastic, and malignant cells of human respiratory epithelia immunostained for hnRNP A2/ B1. (A) Strong (+++) cytoplasmic and/or nuclear hnRNP A2/B1 immunoreactivity is seen in a vast majority of the normal-appearing alveolar cells, which are homogeneously distributed. Original magnification: ×400. (B) A cluster of hnRNP A2/B1-reactive hyperplastic cells is surrounded by normal-appearing alveolar cells, which show no detectable hnRNP A2/B1 immunoreactivity. Original magnification: ×200. (C ) Diffuse hnRNP A2/B1 immunoreactivity (+) is seen in the vast majority of the hyperplastic and normal-appearing bronchial epithelial cells. Some of the reactive stromal cells are also weakly reactive for hnRNP A2/B1. Original magnification: ×200. (D) Despite a high rate of mitotic figures in epithelial cells as seen under a higher magnification (not shown), both the epithelial and stromal cells are nonreactive for hnRNP A2/B1. Original magnification: ×200. (E) A vast majority of the infiltrating cancer cells display strong cytoplasmic hnRNP A2/B1 immunoreactivity. Original magnification: ×400. (F ) Punctuated hnRNP A2/B1 immunoreactivity is seen in many of the cancer cells. Original magnification: ×400.

Of the five normal-appearing cases with hnRNP A2/B1 overexpression, two displayed hnRNP A2/B1 immunoreactivity in the nucleus and three showed immunoreactivity in the cytoplasm (Table 2). The five hyperplastic cases with hnRNP A2/B1 overexpression showed the same hnRNP A2/B1 distribution pattern (Table 2) as that seen in normal-appearing cases. The number of hnRNP A2/B1 immunoreactive cells and the intensity of immunoreactivity varied substantially among normal-appearing and hyperplastic cases. Five to six normal-appearing cases displayed distinct hnRNP A2/B1 immunoreactivity in about 20 to 40% of the epithelial cells. The remaining one case showed no detectable hnRNP A2/B1-immunoreactive cells. Among eight hyperplastic cases, three contained 5 to 35% hnRNP A2/ B1 immunoreactive cells, two showed positive cells in almost the entire epithelium (Figure 1C), and the remaining three showed no immunoreactive cells (Figure 1D).

hnRNP A2/B1 Expression in Malignant Epithelial Cells

Tissue sections of all malignant cases contained distinct hnRNP A2/B1 immunoreactive cells, which accounted for  50% of the epithelial cell population in each of the six cases. Immunoreactive malignant cells were also either homogeneously distributed in certain areas or distributed as clusters (Figures 1E and 1F). In contrast to the hnRNP A2/B1 immunoreactive, normal-appearing, and hyperplastic cells, malignant cells displayed only cytoplasmic hnRNP A2/B1 immunoreactivity (Figures 1E and 1F).

Cellular Localization of hnRNP A2/B1

The subcellular localizations of hnRNP A2/B1 often varied substantially in different cell clusters of the same case. Figure 2 shows a cross-section profile of a small bronchiole, in which the normal-appearing bronchial epithelial cells (short arrows) are devoid of hnRNP A2/B1 expression, whereas the hyperplastic cells (long arrows) show cytoplasmic hnRNP A2/B1 immunoreactivity. In addition, some of the cells located between the normal and hyperplastic cell compartments (Figure 2, arrowheads) display nuclear localization of hnRNP A2/B1.

View larger version (103K): [in this window] [in a new window]   Figure 2.   Variable subcellular localizations of hnRNP A2/B1 among histologically different cells in the longitudinal section of a small bronchiole. The normal-appearing epithelial cells (small arrows) are devoid of hnRNP A2/B1 immunoreactivity; the hyperplastic cells (long arrows) show cytoplasmic staining; and some of the intermediate cells show nuclear staining (arrowheads). Original magnification: ×800.

Clonality Analysis

Samples for five of six female subjects were informative for clonality analysis. Of these, two were malignant with heterogeneous expression of hnRNP A2/B1 (Figure 3A); one contained normal-appearing epithelium with uniformly hnRNP A2/B1 immunoreactive cells (data not shown); one contained immunoreactive hyperplastic cells that constituted almost the entire epithelium (Figure 3D); and the remaining one contained hyperplastic epithelium with no detectable hnRNP A2/B1-expressing cells (Figure 3G). hnRNP A2/B1 immunoreactive cells from the normal-appearing and hyperplastic epithelia and cells from the two malignant cases showed the same monoclonal composition after Hpa II digestion (Figure 3C, lanes 4-6; Figure 3F, lane 4 ). The stromal cells near the hnRNP A2/ B1 immunoreactive, normal-appearing, and hyperplastic epithelium showed the same monoclonality as did the epithelial cells (Figure 3F, lanes 4 and 6 ), whereas the stromal cells far away from the epithelium displayed a polyclonal composition (Figure 3F, lane 5 ). Cells from the hyperplastic lesion that lacked detectable hnRNP A2/B1 expression (Figure 3J, lanes 1-6 ), the normal controls, and the samples treated with the control enzyme Rsa 1 all showed a polyclonal composition. Figure 3 shows the analysis of three representative samples.

View larger version (120K): [in this window] [in a new window]   Figure 3.   Examples of microdissected cells and their clonal composition. (A-C) Invasive lung cancer with mixed cell types. (A) Before microdissection; (B) after microdissection; arrows indicate cells microdissected. hnRNP A2/B1 nonimmunoreactive cells were also microdissected, but the location from which the cells were harvested is not indicated in the figure (c) Lanes 1 and 2: hnRNP A2/B1 immunoreactive cells; lane 3: hnRNP A2/ B1 nonimmunoreactive cells; lanes 4-6: duplicates of lanes 1-3. Lanes 1-3: treated with Rsa 1 (control enzyme); lanes 4-6: treated with Hpa II (a methylation-sensitive restriction endonuclease). Both the hnRNP A2/B1-positive and -negative cells are monoclonal. (D-F ) Homogeneous hnRNP A2/B1 immunoreactive hyperplastic epithelium. (D) Before microdissection; (E) after microdissection; arrows indicate cells microdissected. (F ) Lane 1: hnRNP A2/B1 immunoreactive cells; lane 2: stromal cells far away from the epithelium; lane 3: stromal cells surrounding the epithelium; lanes 4-6: duplicates of lanes 1-3. Lanes 1-3: treated with Rsa 1 (control enzyme) lanes 4-6: treated with Hpa II. The hnRNP A2/B1 reactive epithelial cells (lane 4) and the adjacent stromal cells (lane 6) are monoclonal, whereas the normal stromal cells far away from the epithelium are polyclonal. (G-I ) hnRNP A2/B1 nonimmunoreactive hyperplastic epithelium. (G) Before microdissection; (H ) after microdissection; arrows indicate cells microdissected. (I ) Lane 1: hnRNP A2/B1 nonreactive epithelial cells; lane 2: stromal cells far away from the epithelium; lane 3: stromal cells surrounding the epithelium; lanes 4-6: duplicates of lanes 1-3. Lanes 1-3: treated with Rsa 1 (control enzyme); lanes 4-6: treated with Hpa II. All the cells are polyclonal.

Frequenceis of MA and LOH in Normal-Appearing Epithelial Cells

The numbers of MA and LOH seen in each sample at each chromosomal locus were summed and the combined frequencies of MA and LOH seen in hnRNP A2/B1 immunoreactive or nonimmunoreactive cells are listed in Table 3. Among a total of 85 informative foci in hnRNP A2/B1 immunoreactive samples, a total of 34 MA and LOH were detected (0.40 MA and LOH/focus) and each of the 14 (100%) chromosomal loci displayed at least one MA and LOH; whereas among 65 informative loci in hnRNP A2/ B1 nonreactive samples, a total of only 12 MA and LOH were found (0.18 MA and LOH/focus), and 8 of the 14 (57%) chromosomal loci showed genetic alterations. The hnRNP A2/B1 immunoreactive cells had a significantly higher frequency (P < 0.01) of MA and LOH than did hnRNP A2/B1 nonreactive cells (Table 3).

Frequency of MA and LOH in Hyperplastic Epithelial Cells

The numbers of MA and LOH seen in each sample at each chromosomal locus were summed and the combined frequencies of MA and LOH seen in hnRNP A2/B1 immunoreactive or nonimmunoreactive cells are listed in Table 3. Among a total of 75 informative foci in hnRNP A2/B1 immunoreactive samples, a total of 22 MA and LOH were detected (0.29 MA and LOH/focus), and MA and LOH were distributed in 10 of the 14 (71.4%) chromosomal loci; whereas among 89 informative foci in hnRNP A2/B1 nonreactive samples, a total of only 13 MA and LOH were found (0.15 MA and LOH/focus), and only 5 of 14 (36%) loci showed genetic alterations (Table 3). The hnRNP A2/ B1 immunoreactive cells had a significantly higher frequency of MA and LOH than did their hnRNP A2/B1 nonreactive counterparts (P < 0.03).

Frequencies of MA and LOH in Malignant Epithelial Cells

The numbers of MA and LOH seen in each sample at each chromosomal locus were summed and the combined frequencies of MA and LOH in hnRNP A2/B1 immunoreactive and nonimmunoreactive cells are listed in Table 3. Among 95 informative foci in hnRNP A2/B1 immunoreactive samples, a total of 47 MA and LOH were detected (0.49 MA or LOH/focus), and MA and LOH were seen in 13 of the 14 (93%) chromosomal loci; whereas a total of only 28 MA and LOH were found in 91 informative foci (0.30 MA or LOH/focus) hnRNP nonreactive samples, and MA and LOH were seen in 11 of 14 (78%) chromosomal loci (Table 3). The difference in the frequency of MA and LOH between the reactive and nonreactive cells was significant (P < 0.01).

The frequency and pattern of MA and LOH seen among phenotypically different cells in replicated analyses were consistent.

Concurrent MA or LOH in Normal-Appearing, Hyperplastic, and Malignant Cells

Concurrent MA or LOH was defined as the simultaneous presence of either MA or LOH of either allele (of note for the majority of LOH, the loss is from the same allele) at a given chromosomal locus in both malignant and nonmalignant (including both normal-appearing and hyperplastic) tissues. MA and LOH detected in each cell phenotype of different cases were combined, and the combined frequency and pattern was used for comparison. For hnRNP A2/B1 immunoreactive nonmalignant cells compared with immunoreactive malignant cells, concurrent MA and LOH were found in 13 (92.9%) of the 14 chromosomal loci assessed (Table 4). In a parallel analysis, for the nonmalignant samples without hnRNP A2/B1 overexpression compared with the malignant samples with hnRNP A2/B1 overexpression, concurrent genetic alterations were found in 8 (57.1%) of the 14 loci (Table 4). For hnRNP A2/B1 immunoreactive, nonmalignant samples compared with malignant samples without hnRNP A2/B1 overexpression, concurrent MA and LOH were found in 11 (78.6%) of 14 loci (Table 4). In the final comparison of nonmalignant and malignant cells without hnRNP A2/B1 overexpression, the frequency of shared genetic alterations was 57.1% (8 of 14 loci).

Figure 4 shows examples of co-occurrence of MA and LOH at different chromosomal loci in hnRNP A2/B1 immunoreactive, normal-appearing, and hyperplastic epithelial cells and their malignant counterparts.

View larger version (27K): [in this window] [in a new window]   Figure 4.   Examples of co-occurrence of MA and LOH in normal-appearing (NA), hyperplastic (HP), and malignant (Malig) respiratory epithelial cells analyzed with different DNA markers. M: DNA size marker; NS: normal stromal cells; NE: normal epithelial cells; A: hnRNP A2/B1 nonimmunoreactive alveolar cells; A+: hnRNP A2/B1 immunoreactive alveolar cells; B: hnRNP nonimmunoreactive bronchial epithelial cells; B+: hnRNP immunoreactive bronchial epithelial cells. Arrows indicate genetic alterations (either MA or LOH).

Frequencies of MA and LOH in Cells with Cytoplasmic or Nuclear hnRNP A2/B1 Expression

Of the 13 hnRNP A2/B1 immunoreactive, normal-appearing, and hyperplastic samples, nine showed hnRNP A2/B1 overexpression in the cytoplasm and four in the nucleus (Table 2). A total of 54 MA and LOH were detected in the nine samples with cytoplasmic hnRNP A2/B1 immunoreactivity (6 MA and LOH/sample), whereas only eight MA and LOH were found in the four samples with nuclear hnRNP A2/B1 immunoreactivity (2 MA and LOH/sample) (Table 5). The cells with cytoplasmic hnRNP A2/B1 immunoreactivity had a significantly higher frequency of MA and LOH than did cells with nuclear hnRNP A2/B1 immunoreactivity (P < 0.01).

Of the seven hnRNP A2/B1 immunoreactive malignant samples, all showed cytoplasmic immunoreactivity and no distinct nuclear immunoreactivity was found in any of these samples. A total of 47 MA and LOH were detected in these seven samples (6.7 MA and LOH/sample) (Table 5).

	Discussion

Top Abstract Introduction Materials and Methods Results Discussion References

Although hnRNP A2/B1 overexpression in sputum specimens has been consistently correlated with the eventual development of lung cancer in our previous studies in sputum specimens from over 8,000 patients (3, 5), the mechanism for hnRNP A2/B1 involvement in lung carcinogenesis is unknown. To evaluate the correlation of hnRNP A2/ B1 overexpression with molecular markers of lung carcinogenesis, we compared the frequency and pattern of MA and LOH among phenotypically different epithelial cells with and without hnRNP A2/B1 overexpression. In this study, we found that hnRNP A2/B1 immunoreactive cells had a significantly higher frequency of MA and LOH. Over 80% of MA and LOH seen in hnRNP A2/B1 immunoreactive, normal-appearing, and hyperplastic cells persisted in malignant cells. In addition, in the preliminary analysis, the hnRNP A2/B1 immunoreactive, normal- appearing, and hyperplastic cells tested were monoclonal, whereas comparable cells without hnRNP A2/B1 expression were polyclonal. We also found that cytoplasmic hnRNP A2/B1 immunoreactive cells had 3-fold higher rate of MA and LOH than did cells with nuclear immunoreactivity. These data collectively suggest that (1) although normal-appearing or hyperplastic cells among different cases or in the same case are morphologically similar, they often display a substantially different genetic profile, depending on their hnRNP A2/B1 expression status; and (2) cells with cytoplasmic hnRNP A2/B1 overexpression appear to have accumulated more genetic alterations than other cells, which may account for their more aggressive biologic behavior.

The DNA markers selected for this study were from chromosomes 3, 11, and 17, which harbor known or putative tumor suppressor genes, including fragile histidine triad, Wilm's tumor gene 1, and p53 (9). A number of previous studies with markers from these loci have shown high frequencies of MA and LOH, ranging from 15 to 100%, in different malignant lung tissues (9). The frequencies of MA and LOH seen in this study ranged from 14.3 to 57.1%, with an average of 48% in hnRNP A2/B1 immunoreactive and 38.6% in nonimmunoreactive malignant cells. It is interesting to note that high frequencies of MA and LOH have also been reported in about 50% of histologically normal specimens of smokers (11), in 31 to 76% of specimens from smokers with metaplasia or dysplasia (26), in 76 to 86% of nonmalignant lung tissues form lung cancer patients (27), and in 27% of patients with no cytologic or radiologic evidence of respiratory neoplasia (28). In this study, MA and LOH were seen in normal- appearing and hyperplastic cells without detectable hnRNP A2/B1 with an averaged frequency of 18 and 15%, respectively, and in cells with hnRNP A2/B1 overexpression with an averaged frequency of 40 and 29%, respectively (Table 3). The consistent detection of high frequencies of MA and LOH in nonmalignant respiratory epithelial cells by different studies on different populations suggests that molecular biologic abnormalities precede morphologic abnormalities. Therefore, assessment for MA and LOH may be a more sensitive approach for the detection of early lung cancer. MA and LOH assessment alone, however, can only suggest the general tendency but cannot precisely predict when and which of the premalignant lesions may progress to lung cancer, for the following reasons: (1) the presence of LOH at a given locus is not always a definitive sign of tumor suppressor gene inactivation because the remaining allele may retain normal function; (2) there is currently no consensus as to inactivation of which tumor suppressor gene defines the aggressive behavior of cells; and (3) LOH occurs at a low level ( 10%) randomly throughout the genome. Using a single molecular marker for lung cancer can be criticized because any individual marker suffers from low specificity for lung cancer, which is why we used a panel of markers frequently reported as being altered in lung cancer. Additional studies to better define the relationship of molecular events to field carcinogenesis relative to both synchronous and metachronous primary lung cancer would be of value.

Inasmuch as our previous immunocytochemical studies of sputum specimens from high-risk cohorts have shown that 73.4 to 80% of the individuals with exfoliated hnRNP A2/B1 immunoreactive epithelial cells eventually developed lung cancer (6), it is likely that these cells may have accumulated additional abnormalities during the exfoliation process. Therefore, comparisons of the molecular biologic profile of hnRNP A2/B1 immunoreactive cells in random bronchial biopsies with that of comparable exfoliated cells recovered in sputum specimens may lead to the identification of specific genetic markers that are associated with the clinical progression of lung cancer. From the current analysis, the expression of hnRNP A2/B1 in the airway should be considered not as a benign event but as an indication that at least some molecular events consistent with lung carcinogenesis have occurred. This statement reflects our current finding of a significant correlation between hnRNP A2/B1 immunoreactivity and molecular changes associated with lung cancer even in tissue sections. On the basis of our current findings, we speculate that additional critical molecular events must occur when the bronchial epithelial cells are shed from the basement membrane.

The further elucidation of a specific role of hnRNP A2/ B1 and steps in the development of metastatic competence will be of considerable interest, leading either to better understanding of the mechanism of tumorigenesis and progression or to more effective prevention and management of lung cancer (28).

Clonality analysis based on the Lyon/Beutler's hypothesis of random X chromosomal inactivation is thought to be a reliable method of distinguishing between benign and neoplastic lesions (15). During the early stage of embryogenesis in the female, either the maternally or paternally derived X chromosome in each cell is randomly and permanently inactivated, and this pattern of chromosomal inactivation is stable through subsequent cell cycles (15- 19), which leads to somatic mosaicism of normal females with half of the normal cells expressing the maternal and the other half expressing the paternal allele. In contrast, tumors arising from the clonal proliferation of a single neoplastic cell will display only one, either the maternal or the paternal phenotype. A highly polymorphic trinucleotide CAG repeat in the X-linked HUMARA gene has been used to distinguish between the inactivated and active chromosomes, and to determine the clonal composition of a variety of lesions (15). In this study, two or three foci of phenotypically different epithelial cells with and without hnRNP A2/B1 overexpression in each case were microdissected for clonality analysis. This preliminary result showed that hnRNP A2/B1 immunoreactive cells from non-neoplastic cases were monoclonal, whereas comparable cells without hnRNP A2/B1 were polyclonal. Although the small sample size has prevented us from drawing a general conclusion about the clonal composition of non-neoplastic cells with hnRNP A2/B1 overexpression, our results suggest that some of the hnRNP A2/B1 immunoreactive, normal-appearing, and hyperplastic epithelial cells may be neoplastic and/or precursors of the malignant lesions. Further, stromal cells near the hnRNP A2/ B1 immunoreactive, normal-appearing, and hyperplastic epithelium shared the same monoclonal composition with the epithelial cells, consistent with the assumption that an interaction between stromal and epithelial cells may play an important role in carcinogenesis and tumor progression (24, 29, 30). Further investigation in this regard is ongoing.

The frequency and extent of hnRNP A2/B1 expression with molecular alterations appear to be independent of the cell differentiation status in non-neoplastic samples. Of six normal-appearing cases, five (83.3%) showed hnRNP A2/ B1 overexpression, compared with five of eight (62.5%) cases with hyperplastic cells. Also, in microdissected hnRNP A2/B1 immunoreactive, normal-appearing samples, a total of 34 MA and LOH were detected in a total of 82 informative foci (0.4 MA and LOH/focus), and all 14 (100%) chromosomal loci showed genetic abnormalities; whereas in hnRNP A2/B1 immunoreactive hyperplastic samples, a total of 22 MA and LOH were found in a total of 75 informative foci (0.29 MA and LOH/focus), and only 10 of the 14 (71%) chromosomal loci displayed molecular alterations. The slightly higher frequency of hnRNP A2/B1 coexpression with genetic alterations in normal-appearing compared with hyperplastic cells is consistent with our previously reported finding with immunohistochemical analysis of 1,078 foci of phenotypically different respiratory cells in biopsy specimens from smokers with metaplasia, where overexpression occurred in 41% of the normal-appearing and in 37% of squamous metaplastic samples (8). Our early detection report also showed that a majority of the individuals with hnRNP A2/B1 overexpression in their sputum cells who subsequently developed lung cancer had minimal cytomorphologic abnormalities in their sputum cells (5). Other studies also showed that some histologically normal biopsy specimens from bronchial epithelium of current and former smokers showed allelic loss equal to or greater than that seen in in situ carcinoma lesions (11), and that the cancerous and adjacent normal- appearing mammary epithelial cells share a similar frequency and pattern of genetic alterations (29).

The subcellular localization of hnRNP A2/B1 expression correlates with both the cell differentiation status and molecular alterations. Among 14 cases with normal-appearing and hyperplastic epithelium, four (28.5%) contained no detectable hnRNP A2/B1 immunoreactive cells, four (28.5%) displayed nuclear immunoreactivity, and six (43%) showed cytoplasmic immunoreactivity. In contrast, all six malignant cases showed exclusive cytoplasmic localization of hnRNP A2/B1. The cytoplasmic localization of hnRNP A2/B1 may have useful diagnostic implications. Assessments for MA and LOH revealed a 3-fold higher frequency of MA and LOH in cells with cytoplasmic hnRNP A2/B1 overexpression than in comparable cells with nuclear hnRNP A2/B1 overexpression. Also, clonality analysis of normal-appearing and hyperplastic cells with cytoplasmic hnRNP A2/B1 overexpression showed a monoclonal composition, whereas the comparable cells that lacked detectable hnRNP A2/B1 expression displayed a polyclonal composition. These findings suggest that cytoplasmic overexpression of hnRNP A2/B1 may be an important factor that is closely associated with or signifies the progression of lung neoplasia.

This study was initiated because of the complexity in understanding the significance of hnRNP A2/B1 overexpression. In previous studies, the expression of hnRNP A2/B1 in the exfoliated bronchial cells recovered in the sputum was highly associated with the eventual development of cancer (3, 5), but hnRNP A2/B1 overexpression in the intact bronchial epithelial tissues harvested by bronchial biopsy showed frequent expression in smokers with bronchial metaplasia. In the original study of the mAb to hnRNP A2/B1, we did not find significant immunoexpression in the airways of young, healthy nonsmokers who died of acute trauma. Relatively few of the smokers with metaplasia would be expected to progress to lung cancer (8). In light of this disparate result, we wanted to better understand how hnRNP A2/B1 expression is permitting effective early lung cancer identification. To do this it is important to consider why there may be serious molecular changes suggestive of lung cancer at a higher frequency than the disease frequency actually observed in the general population. The first potentially confounding factor is that the most important risk for mortality in individual with tobacco consumption is cardiovascular diseases. In smokers, death due to heart disease is a dominant competing risk which would decrease the number of people with hnRNP A2/B1-associated abnormalities that would ultimately manifest as lung cancer. Similarly, because developing a cancer is a multistep process, initiated or genetically altered cells may be present for many years in smokers; only with protracted follow-up will the true rate of development of clinically significant lung cancer be established.

Another possible explanation as to why the sputum test is informative but the analysis of cells from intact bronchial mucosa obtained by biopsy is not related to the biology of cancer progression. Evolving cancer cells lose contact with neighboring cells and the basement membrane as they acquire the ability to invade or metastasize. Consequently, these evolving cancer cells are also prone to exfoliate into the sputum. Solitary bronchial epithelial cells recovered in the sputum may preferentially indicate involvement with field carcinogenesis (3, 5, 8). The shed cells-of-interest in the sputum have presumably escaped normal cellular clearance mechanisms because they are recovered intact. Although this type of bronchial epithelial cell is present in low frequency, when these cells display detectable levels of hnRNP A2/B1 a high correlation exists with the eventual development of an evident clinical cancer. This utility presumably reflects not only the biology of hnRNP A2/B1 but also the biology of exfoliation (31). Comparisons of the genetic profiles of hnRNP A2/ B1-immunoreactive cells in random bronchial biopsies with those of comparable cells in sputum specimens may lead to the identification of more specific genetic markers that are associated with true commitment to clinically significant cancer.

The mechanism underlying a correlation between hnRNP A2/B1 overexpression with genetic alterations is not evident, inasmuch as a definite function for hnRNP A2/B1 in the progression of lung cancer has not been elucidated. It has been reported that telomere and telomerase are associated with the regulation of diverse biologic events such as cell growth, immortalization, and chromosomal stability, and there is a growing literature about the potential for interactions between hnRNP A2/B1 and closely homologous family members with telomeric sequences (21, 32, 33). Inasmuch as hnRNP A2/B1 has binding domains for pre-mRNAs and shuttles between the nucleus and cytoplasm, it is also possible that hnRNP A2/ B1 may also function through the mechanism that regulates the biogenesis, localization, and metabolism of certain mRNAs (20). A recent study demonstrated that hnRNP A2 not only selectively binds but also remains associated with the cytoplasmic transport sequence of myelin basic protein mRNA (22). In oogenesis and embryonic development, hnRNP family members, including hnRNP A2/B1, selectively target mRNAs encoding many essential morphogenic proteins to different subcellular domains and other proteins, and mislocalization of these mRNAs results in aberrant embryogenesis (23). For all these reasons, further evaluation of the biology of hnRNP A2/B1 related to carcinogenesis is warranted.