What is the value of proton magnetic resonance in detecting early malignant changes?
P. Barry, C. Wadström, G.L. Falk, P. Shehan,
P. Russell, C.L. Lean, C.E. Mountford (Sydney)
In the Western world, the incidence of adenocarcinoma of the lower esophagus has risen almost 10% per year since the mid- 1970s and now accounts for over 40% of esophageal carcinomas in males . The estimated relative risk of developing adenocarcinoma of the lower esophagus in patients with "Barrett's epithelium" is increased 40-50 fold compared to the normal population. The prevalence of adenocarcinoma in Barrett's metaplasia correlates with the degree of dysplasia, ranging from 21-95% . Histological surveillance of patients with Barrett's epithelium is, as with all biopsies, subject to sampling error, and changes correlating with the presence of in situ adenocarcinoma have not yet been defined. If optimal surveillance was achieved, the chances for curative resection of early adenocarcinoma of the esophagus and long-term survival would be increased [3, 4].
Within Barrett's epithelium, dysplasia is considered to be a pre-neoplastic change most frequently seen in specialized (i.e. intestinal) columnar epithelium and is subdivided into high and low-grade. The criteria for distinguishing reactive or regenerative hyperplasia due to inflammation from low-grade dysplasia are poorly defined and often described as indeterminate . High-grade dysplasia, while shown to be an important harbinger of malignancy in patients with Barrett's metaplasia , should be confirmed by at least two experienced pathologists according to Altorki et al. . Flow cytometric studies may provide a useful adjunct to cytological and histological surveillance whereby neoplastic progression is associated with aneuploidy or an increase in the G2/tetraploid DNA content .
Proton magnetic resonance spectroscopy (1H-MRS) has successfully documented changes in cellular chemistry associated with tumour development and progression in other organs [9-15] and can distinguish pre-invasive from invasive cancer of the uterine cervix with a sensitivity and specificity of 98% and 94% respectively (p < 0.0001) . The technique also distinguishes genuinely benign from malignant follicular lesions in human thyroid  and discriminates degrees of loss of cellular differentiation in ovarian tumors . As such, the 1H-MRS method offers an adjunct to routine histopathology (for review see ). Thus MR technology may offer both independent and objective assessment of dysplasia in Barrett's epithelium with the further potential to identify a predisposition towards adenocarcinoma prior to histological declaration and thus facilitate early intervention.
The aims of this pilot study were to determine whether 1H-MRS could distinguish between the normal esophageal epithelium, Barrett's epithelium, dysplasia and adenocarcinoma of the esophagus, and to identify chemical markers for tumor progression.
Materials and methods
Patient and tissue handling
Tissue specimens were obtained from 8 patients who underwent gastroesophagectomy for adenocarcinoma of the lower esophagus. All patients had preoperative biopsy-confirmed Barrett's epithelium. Mucosal biopsies (2 x 10 mm) from macroscopically normal (stratified squamous) esophageal epithelium, Barrett's epithelium and the edge of the adenocarcinoma were obtained for 1H-MRS analysis and correlative histopathology. The specimens were placed in polypropylene vials containing phosphate buffered saline in deuterated water (PBS/D20), immediately frozen in liquid nitrogen and stored at -70°C for up to 6 weeks until analysis. Prior to the 1H-MRS experiment, each tissue specimen was thawed and bisected. Tissue for MRS analysis was washed twice in PBS/D20 and transferred to a 5 mm MRS tube containing 400 ml PBS/D20, while the adjacent piece of tissue was prepared for routine histological analysis (see below).
Magnetic resonance spectroscopy
One-dimensional (1D) MRS was carried out on an AM-360 wide-bore spectrometer (Bruker Analytische Messtechnik [BAM], Karlsruhe, Germany) operating at 360 MHz (or 8.5 Tesla) equipped with an Aspect 3000 computer and a standard 5 mm dedicated proton probehead. The sample was spinning at 20 Hz and the temperature maintained at 37 °C. Residual water signal was suppressed by selective gated irradiation using low power
(15 dB below 0.2 W). The chemical shifts of resonances were referenced to aqueous sodium 3-(trimethylsilyl)-propanesulphonate (TSPS) at 0.00 parts per million (ppm). One-dimensional spectra were acquired as previously described  over a sweep width of 3597 Hz (10.0 ppm) using a ninety degree pulse, 8192 data points, 256 accumulations, an acquisition time of 1.14 seconds and a relaxation delay of 2.00 seconds.
Data processing for each specimen was undertaken on a Silicon Graphics Indigo-2 workstation using proprietary Bruker software, xwinnmr. A line broadening of 3.0 Hz was applied to the data prior to Fourier transformation. Data were phase corrected using zero and first order phase correction. No baseline correction was applied.
Following 1H-MRS analysis, all tissue specimens were fixed in 10% buffered formalin, paraffin embedded, sectioned at 5 mm and stained with haematoxylin and eosin according to standard protocols. Tissue preservation, relative proportion of mucosa to submucosa, type of metaplasia, degree of dysplasia and presence of potentially confounding factors such as inflammatory cells were reported in addition to the principal diagnosis.
Twenty-one samples (normal esophageal epithelium: 7, Barrett's epithelium: 6, adenocarcinoma: 8) from 8 patients were assessed by 1H-MRS. Typical spectra of normal esophagus, Barrett's esophagus (BE) and adenocarcinoma are shown in Figure 1, where differences in the resonance intensity corresponding to the altered chemical content of these pathologies were observed.
The 1D spectra of normal esophagus is dominated by a resonance at 1.3 ppm which has contributions from the acyl chain of neutral lipid, lactate and threonine (methylene group; -CH2-). The spectra of the esophageal adenocarcinoma have a characteristic peak at 3.2 ppm due to the N-trimethyl (+N(CH3)3) of choline-containing metabolites as well as a peak at 2.03 ppm (acyl chain, -CH=CH-CH2-) as previously described by Lean et al. .
Ratios of the resonance intensity from methyl (-CH3-) groups from neutral lipid and amino acids (0.9 ppm), to the methylene resonance (1.3 ppm) increased progressively from normal through Barrett's metaplasia to adenocarcinoma. The same trend was observed in the intensity ratio of choline, phosphocholine and glycerophosphocholine resonance (3.2 ppm) to the resonance from creatine and lysine (3.0 ppm). Two directional plots of these ratios are shown in Figure 2. A further composite resonance from taurine and phosphoethanolamine at 3.4 ppm was consistently high in normal esophagus and low or absent in adenocarcinoma while resonances of the acyl chain at 2.03 and 2.25 ppm
(-CH2-CH2-COO-) were increased in intensity in Barrett's epithelium compared to normal esophagus and was even higher in adenocarcinoma.
Figure 1. Typical 1D proton MR spectra of tissue from (A) normal esophageal mucosa (B) Barrett's epithelium without dysplasia and (C) adenocarcinoma of the esophagus. Data (256 8K accumulations) were collected at 37°C at 360 MHz with the sample spinning at 20Hz on a Bruker 8.5T MR Spectrometer as previously described . A line broadening of 3 Hz and no baseline correction was applied to each spectrum before Fourier transformation.
Figure 2. Two directional plots of the ratio of the intensity of resonances at 3.2/1.3 ppm and 3.0/0.9 ppm in (A) and at 0.9/1.3 ppm and 3.0/0.9 ppm in (B) measured from one-dimensional MR spectra. Data are grouped on the basis of the final histopathology.
One-dimensional 1H-MRS distinguishes normal esophageal epithelium from adenocarcinoma on the basis of changes in neutral lipid composition and choline-containing metabolites. Choline, which is involved in phospholipid membrane turnover, has been shown to be a marker for increased cellular proliferation . In addition, spectral trends, include a significant increase in the intensity of resonances at 3.2 ppm (due to choline and phosphocholine) and 1.5 ppm (amino acids and lipid) are observed in the progression from Barrett's metaplasia to invasive adenocarcinoma.
These preliminary data show a clear delineation between normal esophagus and adenocarcinoma (Figure 2). However, while the Barrett's epithelium is distinguished from normal tissue it is not distinguished from adenocarcinoma by 1D MRS. It is possible that chemical changes present in Barrett's epithelium represent a "field change" in the metaplastic region. Alternatively, it must be considered that the 8 patients had unidentified microfoci of invasive adenocarcinoma included in the biopsy samples. MR has shown, in several organs [10-13] changes to cellular chemistry which are not morphologically manifest. To further evaluate Barrett's epithelial changes, a cohort of patients with Barrrett's epithelium but without adenocarcinoma, will need to be studied to examine MR changes of progressive dysplasia prior to the onset of histologically detectable adenocarcinoma.
Further analysis of the chemical groups will be possible after two-dimensional correlation spectra (2D COSY) has been undertaken. If chemical groups detected at various stages in the progression from Barrett's epithelium to invasive adenocarcinoma indicate a neoplastic predisposition prior to morphological changes detected by histology as shown for thyroid  and colon , then the potential exists for earlier detection of adenocarcinoma with resultant improvement in survival, or even preventive surgery in such patients. Refinement of surveillance protocols allowing targeting of at-risk individuals would thus be possible.
Current directions being pursued are:
1) analysis of 2D COSY data to further resolve the observed chemical differences,
2) a longitudinal proton MRS study of endoscopic biopsies to examine metaplasia and dysplastic progression in a cohort of patients with Barrett's esophagus,
3) correlation of this data with genetic and histological changes,
4) the application of linear discriminate multivariate analysis of spectroscopic data as described by Somorjai et al. .
These preliminary results of 1D MR spectroscopy on esophageal tissue indicate significant chemical changes between normal epithelium and Barrett's epithelium and adenocarcinoma.
8. Reid BJ, Blount PL, Rubin CE, Levine DS, Haggitt RC, Rabinovitch PS. Flow-cytometric and histological progression to malignancy in Barrett's esophagus: prospective endoscopic surveillance of a cohort. Gastroenterology 1992;102:1212-1219.
10. Delikatny EJ, Russell P, Hunter JC, Hancock R, Atkinson K, van Haaften-Day C, Mountford CE. Proton MR and human cervical neoplasia. I. Ex vivo spectroscopy allows distinction of invasive carcinoma of the cervix from carcinoma in situ and other preinvasive lesions. Radiology 1993;188:791-796.
11. Russell P, Lean CL, Delbridge L, May G, Dowd S, Mountford CE. Proton magnetic resonance and human thyroid neoplasia. I. Discrimination between benign and malignant follicular thyroid neoplasms by magnetic resonance spectroscopy. Am J Med 1994;96:383-388.
12. Ende DA, Lean CL, Mackinnon WB, Chapuis P, Newland R, Russell P, Bokey EL, Mountford CE. Human colorectal adenoma-carcinoma sequence documented by 1H MRS (ex vivo). Proc Soc Magn Reson Med 1993;2:1033.
14. Lean CL, Delbridge L, Russell P, May GL, Mackinnon WB, Roman S, Fahey III TJ, Dowd S, Mountford CE. Diagnosis of benign follicular thyroid lesions by proton magnetic resonance on fine needle biopsy. J Clin Endocrinol Metab 1995;80:1306-1311.
16. Lean CL, Mackinnon WB, Delikatny EJ, Whitehead RH, Mountford CE. Cell-surface fucosylation and magnetic resonance spectroscopy characterization of human malignant colorectal cells. Biochemistry 1992;31:11095-11105.
18. Somorjai RL, Nikulin AE, Pizzi N, et al. Computerized consensus diagnosis: a classification strategy for the robust analysis of MR spectra. I. Application to 1H spectra of thyroid neoplasms. Magn Reson Med 1995;33:257-263.