| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Department of Pathology and Molecular Medicine (L.J., B.D.), Wellington School of Medicine and Health Sciences, Wellington 6002, New Zealand; Departments of Medicine (B.M., N.L.E., S.K.G.G.), Biochemistry and Molecular Biology (N.L.E.), and Laboratory Medicine and Pathology (S.K.G.G.), Mayo Foundation and Clinic, Rochester, Minnesota 55905
Address all correspondence and requests for reprints to: Stefan K. G. Grebe, M.D., Endocrine Lab, Hilton 730, Department of Laboratory Medicine and Pathology, Mayo Foundation and Clinic, 200 First Street SW, Rochester, Minnesota 55905 E-mail: grebe.stefan{at}mayo.edu.
 |
Abstract |
|---|
 Top Abstract IntroductionPatients and MethodsResultsDiscussionReferences |
|---|
|
Introduction |
|---|
|
TopAbstractIntroductionPatients and MethodsResultsDiscussionReferences |
|---|
To be convincing, tumor clonality studies should ideally demonstrate tumor monoclonality by showing concordant somatic genetic changes in all tumor cells. However, because of technical limitations, the majority (but not all) of human tumor clonality studies, including those addressing endocrine tumor clonality, have instead been based on X-chromosome inactivation analysis. During early embryonic development of mammalian female embryos, one of the two X-chromosomes in each cell is randomly deactivated by the methylation of deoxycytosine residues. Once established, genetic imprinting retains X-chromosome inactivation through all subsequent cell divisions. This allows investigators to distinguish active and inactive X-chromosomes in tissues derived from female subjects by combining selective restriction digestion of genomic DNA with methylation sensitive enzymes and subsequent PCR of polymorphic microsatellite loci in proximity of these restriction sites. In a polyclonal tissue such an experiment results in PCR products from both alleles being detectable because some cells will have inactivated one X-chromosome, and other cells will have inactivated the other. By contrast, in a monoclonal tissue, all the cells will be descendent from a single progenitor cell; therefore, all will have inactivated the same X-chromosome, resulting in the finding of the PCR product from only one allele after methylation-sensitive restriction digestion. Based on these principles, monoclonal origins have been confirmed for most human tumors studied (20, 21, 22), including thyroid tumors (8, 12, 15, 16, 19).
However, clonality can be interpreted correctly only as a marker of neoplastic transformation in relation to the clonality of surrounding normal tissue of the same embryological origin. It has been reported that normal epithelial tissues are often organized in monoclonal patches, with all cells within a patch having inactivated the same X-chromosome (23, 24, 25, 26, 27). This makes X-chromosome inactivation-based analysis of lesions arising within such a patch unsuitable for determining whether the lesion is neoplastic or hyperplastic (28, 29). If the embryonic patch is particularly large, then several separate tumors can arise within the patch and appear as a single monoclonal or multifocal tumor, when in fact they are multicentric and have arisen independently from quite distinct portions of the patch (Fig. 1
). Essentially the probability that a tumor is of monoclonal origin is inversely proportional to the number of cells in the embryonal patch. Consequently, as the patch size increases and the number of cells in the patch increases, the probability that a tumor is truly of monoclonal origin decreases because of the uncertainty that more than one cell within the patch contributed to tumorigenesis. Unfortunately, most X-chromosome inactivation studies of human tumors have not included an assessment of the patch size of the surrounding tissues. The few studies that have examined this question have often demonstrated a large patch size, raising questions about the validity of the conclusions of most published tumor clonality studies (23, 24, 25, 26).
|
|
Patients and Methods |
|---|
|
TopAbstractIntroductionPatients and MethodsResultsDiscussionReferences |
|---|
We obtained 26 archival thyroid specimens from 22 female subjects. Of these, DNA samples from two subjects could not be PCR amplified and were excluded (see Results). The mean age at thyroid surgery of the remaining 20 subjects was 54.1 yr (range, 19–77). The final diagnoses of the lesions were benign nonneoplastic nodule or nodular hyperplasia in 14 cases, papillary thyroid carcinoma in two cases (one unifocal and one bifocal), follicular adenoma, medullary thyroid carcinoma, and granulomatous foreign body reaction in one case each, and one sample showed no diagnostic abnormalities. All samples had been fixed in 10% phosphate-buffered formalin and embedded in paraffin by standard methods. For each case, we selected either slides that contained significant amounts of morphologically normal-appearing thyroid epithelium in addition to any lesion or, if available, slides from the contralateral thyroid lobe. For the papillary thyroid carcinoma specimens, we also selected slides containing tumor tissue.
We used six samples of archival normal large bowel tissue, in which the known patch size is a single crypt (31), from human androgen receptor (HUMARA)-repeat heterozygote females (see below) as polyclonal controls. Two normal archival male thyroid samples, with their single unmethylated X-chromosome, served as hemizygous/digestion controls.
Microdissection, DNA extraction, and cell counting
We sectioned paraffin-embedded blocks of formalin-fixed tissues at 6- to 8-µm thickness and mounted several sections onto standard microscopic slides without adhesive. For each specimen one of these slides was stained with hematoxylin and eosin and the remainder were left unstained. An anatomical pathologist (B.D.) with extensive experience in microdissection manually microdissected normal follicular thyroid tissue from the unstained slides using a stereomicroscope and sterile surgical blades, avoiding any morphologically abnormal follicular epithelial structures and any regions containing significant amounts of stromal tissue. For the two papillary carcinoma specimens, tumor tissue was also selectively microdissected in the same fashion, in the case of the bifocal tumor separately for each focus. During microdissection the stained slides from the same specimens served as dissection templates. We collected the microdissected tissues into sterile microcentrifuge tubes and, without prior deparaffinization of the tissue, extracted the DNA by adding 200 µl extraction buffer [100 mM Tris-HCl and 2 mM EDTA (pH 8.0)] and 500 µg proteinase K, followed by 48 h of digestion at 55 C, with an additional 500 µg proteinase K added after the initial 12 h of incubation.
Following microdissection the remaining tissues on the unstained slides were stained with hematoxylin and eosin. We then compared these slides with the matching stained dissection-template slides to estimate the microdissected surface areas and numbers of normal follicular cells that had been microdissected. Using an integration grid of 5 x 5 = 25 boxes measuring 0.828 mm2 at x100 magnification, we counted 5 x 0.828 mm2 areas and calculated average cell n per square millimeter. The number of normal follicular cells in a sample was then calculated by multiplying the cells per square millimeter with the measured microdissected area.
Raw DNA purification
Crude archival male control DNA failed to be digested to completion with the methylation-sensitive restriction enzymes (HpaII and HhaI) used in our study, suggesting that the crude archival DNA samples contained restriction enzyme inhibitors. To remove these putative inhibitors, we purified the raw DNA samples after the proteinase K digestion. We extracted the crude DNA samples three times with Tris-buffer saturated phenol-chloroform (1:1, pH 8.0) and once with chloroform. Phase divider gel tubes (Sigma-Aldrich Corp., St. Louis, MO) were used during organic extraction to facilitate the recovery of samples without contamination by solvents or denatured proteins. After organic extraction, we added a 10% volume of 25% Chelex 100 resin (Bio-Rad Laboratories, Inc., Hercules, CA) in water to the samples and incubated them for 15 min at room temperature. The Chelex beads were removed by centrifugation, and we added one third volume of 7.5 M ammonium acetate and 2.5 volumes of ice-cold absolute ethanol to the supernatant of each tube and precipitated the DNA at -20 C for 24 h. Ten micrograms nuclease free glycogen (Roche, Basel, Switzerland) were added to each sample as a carrier, and the samples were centrifuged at 4 C at 21,000 x g for 60 min. The resulting precipitates were washed with 70% ethanol, air dried, and resuspended in ultrapure water. Finally, we determined the DNA concentration of the purified samples fluorometrically (PicoGreen assay, Molecular Probes, Inc., Eugene, OR) and stored them in sterile, sealed tubes at 4 C. The organic extraction and Chelex treatment allowed complete digestion of male archival control DNA samples.
HUMARA X-chromosome inactivation assay
The majority of recent work relating to assessment of clonality has been performed using the HUMARA (human androgen receptor gene, Xq13) assay (32). In the first exon of the X-linked HUMARA gene, there are two HhaI and two HpaII restriction sites, which are located less than 100 bp 5' to a highly polymorphic CAG microsatellite repeat. X-chromosome inactivation is associated with methylation of these restriction sites. Digestion with methylation-sensitive endonucleases followed by PCR amplification with primers flanking these restriction sites, and the repeat can be used to distinguish between transcriptionally active and inactive X-chromosomes in tissues of heterozygous female subjects.
We used a modified HUMARA assay similar to that described by Kopp et al. (33). We incubated 600–800 ng purified DNA in a final volume of 40 µl with 40 U of HhaI (Amersham Pharmacia Biotech AB, Uppsala, Sweden) in potassium acetate digestion buffer [50 mM K-acetate, 10 mM Mg-acetate, 20 mM Tris-acetate, 1 mM dithiothreitol, and 1 µg/ml BSA (pH 7.9)] at 37 C for 12 h. Heating to 70 C for 20 min terminated the reactions. For each sample we included a positive control reaction using MspI, a nonmethylation-sensitive isoschizomer of HhaI, and a negative control reaction, containing only DNA but no enzyme. In addition, each set of experiments also included both a polyclonal (female bowel) and a hemizygous (male thyroid) control sample.
We PCR amplified paired digested and control samples using 5'-hexachlorofluorescein phosphoaramidite-labeled primers. The PCR reactions contained 2.5 µl of digested or undigested DNA in a 25-µl PCR mix of sterile distilled H2O containing 50 mM KCl, 10 mM Tris-HCl, 1.5 mM MgCl2 (pH 8.3), 250 µM deoxynucleotide triphosphates, 500 nM of each primer, and 0.2 µl (1 U) Taq polymerase (Roche). The primer sequences were (32): TCCAGAATCTGTTCCAGAGCGTGC (sense, labeled with 5'-hexachlorofluorescein phosphoaramidite) and GCTGTGAAGGTTGCTGTTCCTCAT (antisense; all primers were synthesized by the Mayo Clinic Oligo-Nucleotide Core Facility). The cycling conditions were: hold at 95 C for 5 min; seven cycles 95 C (45 sec), 67 C (30 sec), 72 C (30 sec); five cycles 95 C (45 sec), 65 C (30 sec), 72 C (30 sec); 28 cycles 95 C (45 sec), 63 C (30 sec), 72 C (30 sec); and hold at 72 C for 15 min. We amplified all samples in duplicate.
We size separated the PCR products on an automated DNA sequencer (ABI-377, Applied Biosystems, Foster City, CA) and analyzed the results using GeneScan 3.1.2 software (Applied Biosystems), with the allele peak heights serving as a semiquantitative measure of the amount of PCR product amplified from each allele. To correct for possible preferential amplification of one allele, we calculated a corrected allele ratio for each HhaI digested and negative control sample pair. This corrected ratio was derived by dividing the ratio of the undigested allele1 and the digested allele1 by the ratio of the undigested allele2 and the digested allele2. We defined samples as monoclonal if this corrected ratio was less than 0.33 or above 3 (33). Polyclonal tissues with random X-inactivation patterns would be expected to have ratios equal or close to 1.0.
|
Results |
|---|
|
TopAbstractIntroductionPatients and MethodsResultsDiscussionReferences |
|---|
). First, we assessed DNA samples from all subjects for heterozygosity at the HUMARA locus. Samples from two subjects failed to amplify and were excluded from further studies. Four subjects were homozygous and noninformative (20%), and the remaining 16 (80%) were heterozygous and informative. For two of these 16 patients, we examined multiple sections representing different, physically distinct regions of the thyroid gland, so 20 heterozygous specimens in total were tested for clonality.
|
and Fig. 2). Moreover, only one of the samples with random inactivation patterns displayed a corrected allele ratio of close to 1.0; the other five samples displayed corrected allele ratios of between 0.3 and 0.5 or between 2.33 and 3. Previous studies have sometimes considered ratios of less than 0.5 or greater than 2 as indicative of monoclonality (34).
|
|
).
For one subject, normal thyroid epithelium from four separate tissue blocks removed from both lobes of the gland was assessed, and nonrandom X-inactivation patterns were observed in all of them (Table 2); in all four samples, the longer allele was amplified following digestion with HhaI. Two separate samples taken from the left and right lobes of the thyroid were available for another subject. Both samples had nonrandom X-chromosome inactivation patterns, but different alleles were inactivated in the two sections; in the sample taken from the left lobe, the longer allele was inactivated, but in the right-lobe sample, the smaller allele was present after HhaI digestion (Fig. 3).
|
).
|
|
Discussion |
|---|
|
TopAbstractIntroductionPatients and MethodsResultsDiscussionReferences |
|---|
). These results indicate that monoclonality in thyroid epithelium is not restricted to neoplastic processes but that large portions of normal tissue are also monoclonally derived. It is unlikely that these results are chance findings or related to experimental methodology (e.g. the use of paraffin-embedded specimens). The majority of previous clonality studies of tumors have also used such specimens, and the methodology has proved robust. Random allele dropout can occur in microsatellite PCR analysis of DNA derived from formalin-fixed, paraffin-embedded specimens, a potential problem we are well aware of (35). However, allele dropout occurs mainly with low amplifiable DNA inputs, which we avoided, and affects preferentially the longer allele, the reverse of what we observed in our study (36). Also, one would expect this to affect negative controls, which were treated exactly the same way as the digested samples. Finally and most compellingly, we had some initial problems with achieving complete digestion of samples, rather than spurious allele loss, and in fact had to optimize the published HUMARA assay to overcome this limitation (37). From this perspective it is interesting to speculate whether incomplete digestion may have been an unrecognized problem in some previous tumor clonality studies using the HUMARA assay and might have contributed to erroneous clonality assessment. We therefore believe that our data are reliable and the human thyroid gland exhibits a large embryonal patch size. Consequently, X-inactivation analysis may not be the appropriate tool to study clonal composition of thyroid tumors or distinguish thyroid hyperplasias from neoplasms.
This may call into question the relevance of thyroid tumor clonality studies with regard to the mechanisms of thyroid tumorigenesis, at least as far as studies of paraffin-embedded specimens are concerned (8). In particular, unless there is direct evidence of concordant tumor somatic genetic changes, suggestions that somatic mutations in unknown genes are the reason for the apparent frequent monoclonal origin of benign solitary thyroid nodules (38, 39, 40) may need to be reconsidered because the monoclonality in these lesions may simply reflect the clonal architecture of normal thyroid epithelium from which they have arisen. To truly demonstrate monoclonality, postulated somatic genetic changes have to be directly demonstrated, rather than inferred on the basis of apparent X-inactivation monoclonality, a gold standard that has not often been achieved in tumor clonality studies. It may well be that in endocrine glands the tissue context and variant abilities of individual cells to interpret growth-regulating signals derived from neighboring cells and through trophic hormones are more important factors in tumorigenesis than acquisition of a series of defined tumorigenic somatic genetic changes. X-chromosome inactivation assays have long been used to study clonal composition of human tumors. Unfortunately, the fact that epithelial tissues are often organized in patches consisting of groups of cells inactivating the same X-chromosome has largely been ignored. Erroneous interpretation of tumor clonality assessments will result if the question of patch size is not considered or if normal control tissue is derived from a sample that is significantly larger than a single patch, consists of several mixed tissues, or is derived from a completely different tissue. Previous studies performed on human thyroids have not been an exception to these methodological shortcomings, and although this has been discussed in some detail in two recent reviews (28, 29), it is not appreciated by most thyroidologists.
Our results are not surprising when taken in the context of the known embryonic development of the human thyroid gland. The thyroid is a small organ, arising from a small, tightly clustered group of submandibular embryonic thyroid stem cells, which migrate caudally along the thyroglossal duct early in embryogenesis. Given the small pool of cells present at the time of X-inactivation and restricted cell mixing, there is a high probability that large portions of follicular tissue will have the same X-chromosome inactivated. The exact size of the thyroid epithelial patch size is difficult to estimate because, unlike other epithelial tissues, thyroid follicular tissue does not form epithelial sheets. In cross-bred animals with X-chromosome-encoded enzyme polymorphisms, histochemistry can nonetheless directly delineate the two-dimensional boundaries of a thyroid epithelial patch (19, 41). Unfortunately, this is usually not possible in humans, but occasionally the patch boundaries can be inferred. One example is when normal tissue abutting a tumor has inactivated a different HUMARA allele than the tumor tissue. We identified one such case when we examined a small number of thyroid tumors along with matching normal thyroid epithelium, with one allele being inactivated in the normal gland, while the bifocal tumor had inactivated the other (Fig. 4). However, even in this case completely accurate assessment of patch size is impossible.
Follicles are three-dimensional structures, formed by the budding of new follicles in different directions from preexisting ones. With the lack of microanatomical structures to guide the direction that tissue is cut during pathological sectioning and microdissection, distant portions of the thyroid may be included on a single slide and consequently in a microdissected sample. As a result, a random X-chromosome inactivation pattern may be observed in a sample, despite large portions of tissue having inactivated the same X-chromosome. Bearing these limitations in mind, which will likely result in an underestimation of clonal unit size, we still detected that areas on slides containing monoclonal-derived tissue were very large, between 48 and 128 mm2, equating to approximately 1–4 x 105 thyrocytes. Of course, neighboring patches have a 50% likelihood of inactivating the same X-chromosome, and the actual patch size could be somewhat smaller. On the other hand, the thyroid gland is a three-dimensional structure and the actual number of cells in a patch may well be significantly larger. Without serial sections this cannot be estimated accurately. In any case, in combination with the fact that we observed so few cases with random X-inactivation, our data suggest that the embryonal patch size within the thyroid exceeds significantly the dimensions of embryonal patches in most other studied organs, such as cervical tissue (20), vascular smooth muscle tissue (25), urothelium (24), skin (27), and breast (23, 26).
The limited availability of archival normal human thyroid tissue did not allow us to test the clonality patterns in whole glands or narrow the thyroid patch size down more precisely. However, for two thyroid glands tested, specimens from both lobes were available. Nonrandom X-inactivation pattern was found in both cases in all sections studied. In one patient, different parental alleles were inactivated in left and right lobes, suggesting a minimum of two stem cells. In the other patient, the same allele was inactivated in all four sections examined from both lobes. The latter case implies that this thyroid gland may have been derived from a single progenitor cell or more likely represents a case of extreme skewing in X-inactivation. Preferential inactivation of one parental X-chromosome has been observed in different tissues of healthy individuals as well as a cause of human disease. Guo et al. (20) reported three patients with the same, nonrandom pattern of X-inactivation present in normal admixed cervical tissue and matched epithelium sampled from three different sites, and Gale et al. (42) found that 20% of hematologically normal subjects showed greater than 75% inactivation of the same X-chromosome in all tissues tested (granulocytes, T-lymphocytes, skin, and muscle). Such extreme preferential inactivation probably just reflects the randomness of the inactivation process; although the majority of females have relatively equal proportions of maternal and paternal X-chromosomes inactivated, a small number of individuals will, just by chance, demonstrate predominately maternal or paternal X-chromosome inactivation. Finding of extreme preferential inactivation in one of the normal bowel samples, supposed to serve as a polyclonal control, can be explained in this way.
Finally, our results should rekindle the discussion about the role of trophic hormone effects in endocrine tumor promotion. Whereas in its classical form the trophic hormone theory is most likely incorrect, trophic hormones may play a permissible and tumor-promoting role. For instance, in rodents, elevated TSH levels alone are sufficient for thyroid tumor development, and humans with chronically elevated TSH levels, as is, for example, seen in thyroid dyshormonogenesis syndromes, have a higher-than-average risk of developing thyroid neoplasms (43, 44). It is now widely acknowledged that the normal thyroid gland is characterized by a degree of somatic genetic heterogeneity (29). Constant trophic hormone stimulation may well exaggerate existing intrinsic differences in growth potential between cells and in turn lead to accumulation of further somatic genetic changes through accelerated proliferation of clones that possess slightly increased trophic hormone responsiveness or intrinsic growth potential. Eventually this may lead to tumor formation. However, despite the plausibility of this scenario, even the modified trophic hormone theories have been largely abandoned in recent years because multiple X-chromosome inactivation studies seemed to show that endocrine neoplasms are monoclonally derived.
In light of our findings of a very large patch size, at least in the thyroid gland, it appears that the relevance of these studies as to endocrine tumorigenesis needs to be questioned. It is interesting to consider that other small endocrine glands, in particular the pituitary with its several distinct cell types occurring in a very small volume, might well also have large embryonal patch sizes, raising the question of whether such tumors are monoclonal or polyclonal. This in turn opens the door for reconsidering both a cancer field effect and a modified trophic stimulus theory for the etiology of some thyroid or pituitary tumors. Such a hypothesis could account for the finding that many papillary thyroid carcinomas are multicentric. A similar hypothesis of trophic tumorigenesis could explain why many cases of functioning pituitary tumors, particularly in Cushing’s disease, show no distinct tumor and, perhaps, why recurrence occurs so frequently, despite apparent complete removal of the tumor (45, 46). In pituitary neoplasms, supportive evidence for this hypothesis exists in the form of feedback tumors, such as is observed sometimes in long-term undertreated Addison’s disease, and in thyroid neoplasia, the tumorigenic effects of trophic hormones may be at the root of the increased thyroid tumor incidence in patients with thyroid dyshormonogenesis.
|
Acknowledgments |
|---|
|
Footnotes |
|---|
Abbreviation: HUMARA, Human androgen receptor gene.
Received October 4, 2002.
Accepted April 4, 2003.
|
References |
|---|
|
TopAbstractIntroductionPatients and MethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
 |
|
 |
|
  R. P. McCarthy, M. Wang, T. D. Jones, R. W. Strate, and L. Cheng Molecular Evidence for the Same Clonal Origin of Multifocal Papillary Thyroid Carcinomas Clin. Cancer Res., April 15, 2006; 12(8): 2414 - 2418. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 T. M. Shattuck, W. H. Westra, P. W. Ladenson, and A. Arnold Independent Clonal Origins of Distinct Tumor Foci in Multifocal Papillary Thyroid Carcinoma N. Engl. J. Med., June 9, 2005; 352(23): 2406 - 2412. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 K. Krohn, D. Fuhrer, Y. Bayer, M. Eszlinger, V. Brauer, S. Neumann, and R. Paschke Molecular Pathogenesis of Euthyroid and Toxic Multinodular Goiter Endocr. Rev., June 1, 2005; 26(4): 504 - 524. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| Endocrinology | Endocrine Reviews | J. Clin. End. & Metab. |
| Molecular Endocrinology | Recent Prog. Horm. Res. | All Endocrine Journals |
