J. Pathol. 187: 217–222 (1999)

CLINICOPATHOLOGICAL IMPLICATIONS OFPARATHYROID HORMONE-RELATED PROTEIN IN

HUMAN COLORECTAL TUMOURS

1*, 2, 1, 3, 4 1

1Department of Surgery II, Nagasaki University School of Medicine, Nagasaki, Japan2Department of Molecular Pathology, Nagasaki University School of Medicine, Nagasaki, Japan

3Department of Nature Medicine, Nagasaki University School of Medicine, Nagasaki, Japan4Department of Pathology II, Nagasaki University School of Medicine, Nagasaki, Japan

SUMMARY

The purpose of the present study was to clarify the relationship of parathyroid hormone-related protein (PTHrP) to the oncogenesisand progression of colorectal adenocarcinoma. A total of 108 colorectal tumours, including 12 adenomas, six adenocarcinomas inadenomas, and 90 adenocarcinomas, were studied. Immunohistochemistry, in situ hybridization, and reverse transcription-polymerasechain reaction (RT-PCR) techniques were used to evaluate the expression of PTHrP. Positivity of immunostaining for PTHrP wasdefined as highly positive (+ +), slightly positive (+), and negative ("). None of the adenomas of background non-neoplastic mucosalepithelia showed immunostaining of PTHrP. In contrast, PTHrP was expressed in 85 (94·4 per cent) of 90 colorectal adenocarcinomas.Immunoreactivity of PTHrP was greater in poorly differentiated adenocarcinomas than in well-differentiated ones. Furthermore,advancing margins of primary tumours stained more intensely than other sites. Highly positive immunoreactivity of PTHrP, classifiedby histological invasiveness, was 22·6 per cent within the muscularis propria and 69·5 per cent beyond the muscularis propria. PTHrPexpression was significantly correlated with differentiation, depth of invasion, lymphatic invasion, lymph node metastasis, hepaticmetastases, and Dukes’ classification. In carcinoma, PTHrP mRNA expression was evident in tumour cells by in situ hybridization.PTHrP transcripts were also detected in two resected human colorectal adenocarcinomas by RT-PCR. These findings suggest thatPTHrP is related to carcinogenesis, differentiation, progression, and prognosis of colorectal adenocarcinomas. Copyright ? 1999John Wiley & Sons, Ltd.

KEY WORDS—PTHrP; colorectal cancer; tumour progression; invasion; prognosis

INTRODUCTION

Parathyroid hormone-related protein (PTHrP) wasinitially characterized as a causative factor of humoralhypercalcaemia of malignancy.1,2 However, this proteinwas subsequently found to be expressed in many normaland neoplastic tissues without hypercalcaemia.3 PTHrPis a product of normal tissues4–7 and plays diversephysiological roles in a paracrine/autocrine fashion.8These include (1) regulation of cellular proliferation anddifferentiation; (2) development of the fetus, especiallyof cartilage and the skeletal system; (3) relaxation ofvascular and non-vascular smooth muscle; and (4) main-tenance of the reproductive system.9 The expression ofthe PTHrP gene in normal tissues is tightly regulatedby a number of factors, involving both transcriptionaland post-transcriptional mechanisms. It has also beenreported that PTHrP plays a role in the growth anddifferentiation of neoplastic and non-neoplastic cells.10

Increased expression of PTHrP has been detected inadvanced prostatic cancer11 and in the skeletal metasta-ses of breast cancer.12 From our laboratories, it wasfound that in vivo transfection of the antisense oligo-nucleotide for PTHrP mRNA decreased tumour growth

CCC 0022–3417/99/020217–06$17.50Copyright ? 1999 John Wiley & Sons, Ltd.

in metastatic pituitary clonal cells.13 These findingssuggest that PTHrP is widely involved in tumour pro-gression, irrespective of humoral hypercalcaemia ofmalignancy.

In the present study, the expression of PTHrP insurgically resected colorectal tumours was investigatedusing immunohistochemical and molecular biologicalapproaches; special attention was paid to the relation-ship of PTHrP expression to clinicopathologicalparameters.

MATERIALS AND METHODS

*Correspondence to: Minoru Nishihara, MD, Department ofSurgery II, Nagasaki University School of Medicine, 1-7-1 Sakamoto,Nagasaki 852-8523, Japan.

Surgical specimens

A total of 108 surgically resected colonic tumourswere examined including 90 primary colorectal adeno-carcinomas without bone metastasis, six adenocarcino-mas in adenomas, and 12 adenomas. The 90adenocarcinomas were from 84 patients, with 80patients having a single cancer, three a double, and onea quadruple cancer. All specimens were obtained frompatients operated on at Nagasaki University Hospitalbetween 1993 and 1994. Paraffin sections were routinelystained with haematoxylin and eosin and the histologicaltype of tumour was determined according to the WorldHealth Organization classification.14

Received 13 March 1997Revised 8 April 1998

Accepted 28 July 1998

218 M. NISHIHARA ET AL.

Immunohistochemical studies

Immunohistochemical studies were carried out usingthe alkaline phosphatase technique as previouslydescribed.15 The monoclonal antibody was raised to thecarboxy-terminal amino acid fragment (38–64) ofPTHrP with no homology to or cross-reactivity withparathyroid hormone.

Formalin-fixed and paraffin-embedded tissues werecut into 4 ìm sections, deparaffinized in xylene, andrehydrated in phosphate-buffered saline (PBS). Depar-affinized sections were preincubated with bovine serumto prevent non-specific binding and then incubatedovernight at 4)C with an optimal dilution (5 ìg/ml) of aprimary monoclonal antibody of PTHrP (OncogeneScience, Inc., Uniondale, NY, U.S.A.). The slides werethen incubated with alkaline phosphatase-conjugatedgoat anti-mouse immunoglobulin antibody. The sites ofalkaline phosphatase were detected using a mixture of5-bromo-4-chloro-3-indolyl phosphate and nitrobluetetrazolium chloride (BCIP/NBT; BRL, Gaithersburg,MD, U.S.A.).

Negative controls were prepared by replacing theprimary antibody with non-immunized mouse serum.An immunoabsorption test was performed to confirmthe specific immunoreactivity of the antibody. A speci-men of squamous cell carcinoma of the lung was used asa positive control.15 High immunostaining intensity wasdenoted by + + (Fig. 1a). Definitely lower stainingintensity was denoted by +. Absence of staining wasdenoted by " (Fig. 1b). The grade of positivity ofPTHrP expression was determined by the dominantimmunostaining intensity observed in each specimen. Allof the slides were independently examined by two inves-tigators and the results were compared. In some cases ofdiscrepancy, the slides were re-reviewed and a consensuswas reached.

In situ hybridization

Three cases of well-differentiated adenocarcinoma, 11cases of moderately differentiated adenocarcinoma, anda case of poorly differentiated adenocarcinoma, all ofwhich expressed PTHrP by immunohistochemistry, werestudied by in situ hybridization. In situ hybridization wasperformed as previously described with the followingmodifications.16 Tissues for in situ hybridization analysiswere fixed in 4 per cent p-formaldehyde, dehydrated,embedded in paraffin wax, and serially sectioned at4 ìm. Membrane protein digestion was performed with20 ìg/ml proteinase K (Sigma Chemical Co., St Louis,MO, U.S.A.) in PBS for 10 min at 37)C. PTHrP mRNAwas detected using a 343 bp Pvu II/Bgl II fragment ofrPLPm10 and a 569 bp HindIII/NotI fragment of R15B,

Copyright ? 1999 John Wiley & Sons, Ltd.

subcloned into pBluescript SK+ (Stratagene, La Jolla,CA, U.S.A.). Antisense and sense single-strand cRNAprobes were labelled with digoxigenin-labelled UTP(Boehringer Mannheim, Mannheim, Germany) andT3/T7 RNA polymerase (Ambion Inc., Austin, TX,U.S.A.). Sections were hybridized at 50)C for 16 h withthe digoxigenin-labelled riboprobe. After the slides weresubjected to RNase A treatment (10 ìg/ml in 10 mmol/lTris–HCl, pH 8·0; 550 mmol/l NaCl; and 1 mmol/lEDTA) at 37)C for 30 min, hybridized digoxigenin-labelled riboprobe was detected with a DIG DNAdetection kit (Boehringer Mannheim) according to themanufacturer’s guidelines.

Reverse transcription-polymerase chain reaction(RT-PCR)

Total RNA was extracted from two fresh humancolorectal adenocarcinomas. As normal controls, non-

Fig. 1—PTHrP immunoreactivity in PTHrP-producing cells of humanlung squamous cell carcinoma as a control. (a) A positive control. (b)A negative control was prepared by replacing the primary antibodywith non-immunized mouse serum

Fig. 2—PTHrP immunohistochemistry in normal colonic mucosa and colorectal tumours. (a) Normal colonic mucosa. PTHrP expression is notfound. (b) Tubulo-villus adenoma. PTHrP antigen is not detected in adenoma cells. (c) Adenocarcinoma in adenoma. PTHrP expression is detectedin the carcinomatous component (arrows) but not in the adenomatous component. (d) Well-differentiated adenocarcinoma. PTHrP expression isweakly positive in a well-differentiated adenocarcinoma in the submucosa. (e) Poorly differentiated adenocarcinoma. PTHrP is strongly expressedthroughout the cytoplasm of tumour cells. (f) Subserosal invasion of a moderately differentiated adenocarcinoma. Strong expression of PTHrP isfound in the subserosal invasive site. (g) Invasive front. PTHrP expression is stronger in the advancing margins (arrows) than in the adjacent partof the moderately differentiated adenocarcinoma. (h) Tumour cells in a lymphatic vessel. Strong immunoreactivity is found in the carcinoma cellsinvading the lymphatic vessel

J. Pathol. 187: 217–222 (1999)

Copyright ? 1999 John Wiley & Sons, Ltd. J. Pathol. 187: 217–222 (1999)

220 M. NISHIHARA ET AL.

neoplastic fresh colonic tissues were obtained fromsurrounding normal colonic tissues in the primarycolonic adenocarcinomas of two patients. A PTHrP-producing lung cancer cell line served as a positivecontrol.15 Cellular RNA (1 ìg) was incubated at 37)Cfor 1 h in 50 ìl of reverse transcriptase buffer containing20 units of RNAsin (Promega Corp., Madison,WI, U.S.A.), 100 pmol of random hexamer primers(Boehringer Mannheim), and 400 units of Moloneymurine leukaemic virus reverse transcriptase (GIBCO/BRL). Reverse transcription was terminated by heatingat 95)C for 10 min and 20 per cent of the resultingcDNA was removed for PCR. PCR samples wereincubated with 50 pmol of each primer and 2·5 units ofTaq EX DNA polymerase. For PTHrP mRNA, syn-thetic primers (sense 5*-CTGGTTCAGCAGTGGAGCGTC-3* and antisense 5*-GTTAGGGGACCACCTCCGAGGT-3*) were used to amplify a 231 bp fragmentas previously described.17 The â-actin primers were

Table I—PTHrP expression in adenoma, adenocarcinoma in adenoma, and adenocarcinoma of thecolorectum

No. of cases

PTHrP expression

" + + +

Adenoma 12 12 (100%) 0 0Adenocarcinoma in adenoma 6 0 6 (100%) 0Adenocarcinoma 90 5 (5·6%) 37 (41·1%) 48 (53·3%)

Intensity of PTHrP staining: " =absence of signal; + =low immunostaining intensity; + + =high immunostainingintensity.

Correlation between histological type and PTHrP expression is significant by the cumulative chi-square method(P<0·01).

Table II—Relationships between PTHrP expression and clinicopathological variables in human colorectal adenocarcinomas

Pathological findings No. of cases

PTHrP expression

" + + +

Differentiation* Well differentiated 22 3 (13·6%) 12 (54·6%) 7 (31·8%)Moderately differentiated 65 2 (3·1%) 25 (38·5%) 38 (58·4%)Poorly differentiated 3 0 0 3 (100%)

Depth of tumour invasion* Within the muscularis propria 31 3 (9·7%) 21 (67·7%) 7 (22·6%)Beyond the muscularis propria 59 2 (3·4%) 16 (27·1%) 41 (69·5%)

Lymphatic invasion* Absence 18 2 (11·1%) 12 (66·7%) 4 (22·2%)Presence 72 3 (4·2%) 25 (34·7%) 44 (61·1%)

Lymph node metastasis* Absence 50 5 (10·0%) 23 (46·0%) 22 (44·0%)Presence 40 0 14 (35·0%) 26 (65·0%)

Venous invasion Absence 54 3 (5·6%) 26 (48·1%) 25 (46·3%)Presence 36 2 (5·5%) 11 (30·6%) 23 (63·9%)

Hepatic metastasis* Absence 75 5 (6·7%) 34 (45·3%) 36 (48·0%)Presence 15 0 3 (20·0%) 12 (80·0%)

Dukes’ classification* A 25 3 (12·0%) 17 (68·0%) 5 (20·0%)B 22 2 (9·1%) 6 (27·3%) 14 (63·6%)C 28 0 11 (39·3%) 17 (60·7%)D 15 0 3 (20·0%) 12 (80·0%)

*Significant correlation with PTHrP expression by the cumulative chi-square method (P<0·01).

Copyright ? 1999 John Wiley & Sons, Ltd.

predicted to amplify 313 bp DNA fragments.15 Sampleswere subjected to 35 cycles of PCR amplification using athermocycler. Each cycle included denaturation at 94)Cfor 1 min, annealing at 45)C for 1 min, and primerextension at 72)C for 1 min. An aliquot of each ampli-fication mixture was subjected to electrophoresis on a2·5 per cent agarose gel and DNA was visualized byethidium bromide staining.

Statistical analysis

The cumulative chi-square method was used to com-pare the staining intensity of tissue PTHrP and variousclinicopathological parameters.

RESULTS

None of the cases in this study exhibited humoralhypercalcaemia. Expression of PTHrP was observed

J. Pathol. 187: 217–222 (1999)

221PTHrP EXPRESSION IN COLORECTAL TUMOURS

Fig. 3—In situ hybridization of PTHrP mRNA. (a) Cytoplasmic PTHrP mRNA expression is detected in colorectal adenocarcinoma cells.(b) The sense labelled probe showed no specific hybridization signals

Copyright ? 1999 John Wiley & Sons, Ltd.

Fig. 4—RT-PCR of PTHrP mRNA. Two human colorectal adeno-carcinomas (1, 2) expressed PTHrP mRNA (231 bp) (top). Non-neoplastic colonic mucosae (3, 4) did not express PTHrP mRNA.â-Actin mRNA was detected in all samples (313 bp) (bottom). Lanes 1,2: human colorectal adenocarcinomas; lanes 3, 4: normal colorectalmucosae

throughout the cytoplasm of the tumour cells. Non-neoplastic colorectal mucosae were negative for PTHrP(Fig. 2a). None of the 12 adenomas showed positiveimmunoreactivity for PTHrP (Fig. 2b). In six cases withadenocarcinoma in adenoma, PTHrP expression wasdetected in the carcinomatous but not in the adeno-matous component (Fig. 2c). Forty-eight (53·3 per cent)of the 90 adenocarcinomas showed strong immuno-reactivity for PTHrP (Table I). Among the PTHrP-positive cases were 19/22 cases (86·4 per cent) ofwell-differentiated, 63/65 cases (96·9 per cent) of moder-ately differentiated, and 3/3 cases (100 per cent) ofpoorly differentiated adenocarcinoma. The intensityof PTHrP immunoreactivity was correlated with thedifferentiation of colorectal adenocarcinoma (P<0·01).Strong positivity of PTHrP expression classified by thedepth of tumour invasion was found in 7/31 cases (22·6per cent) within the muscularis propria and 41/59 cases(69·5 per cent) showing deeper invasion beyond themuscularis propria. There was a significant correlationbetween the intensity of PTHrP expression and thedepth of invasion (P<0·01). The intensity of PTHrPexpression was significantly correlated with lymphaticinvasion, lymph node metastasis, hepatic metastasis, andDukes’ classification (Table II). The cases showingstrong intensity in the primary lesion are likely to havemetastases to the liver or lymph nodes.

The Figures show representative examples of PTHrPimmunoreactivity. PTHrP expression was weaker inwell-differentiated (Fig. 2d) than in poorly differenti-ated adenocarcinomas (Fig. 2e). Strong positivity wasobserved in carcinomas invading the subserosa (Fig. 2f).Furthermore, advancing margins stained more intenselythan adjacent sites of the primary carcinoma (Fig. 2g).Invasive tumour cells in lymphatic vessels also showedstrong positivity (Fig. 2h).

In accordance with immunostaining, PTHrP mRNAexpression by in situ hybridization was positive in thecytoplasm of carcinoma cells (Fig. 3a). No specifichybridization signals were observed with the senselabelled probe (Fig. 3b). RNase treatment of sectionshybridized with the PTHrP RNA probe yielded nopositive signals.

The results of RT-PCR for PTHrP mRNA expressionin two human colorectal adenocarcinomas and twonon-neoplastic colorectal mucosae are shown in Fig. 4.Two human colorectal adenocarcinomas expressedPTHrP mRNA. In contrast, non-neoplastic colorectalmucosae did not express PTHrP mRNA. â-ActinmRNA was used as an internal control.

DISCUSSION

Hypercalcaemia associated with malignant tumours isone of the most common signs of a paraneoplasticsyndrome.18,19 In our study, regardless of humoralhypercalcaemia of malignancy, elevated PTHrP expres-sion was detected at the mRNA and protein levels inmost of the patients, suggesting that this protein

J. Pathol. 187: 217–222 (1999)

222 M. NISHIHARA ET AL.

has some pathophysiological role in colorectaladenocarcinoma.

It has been suggested that activation of the PTHrPgene is closely related to malignant transformation ofnormal mammalian cells.20 In our study, neither adeno-mas nor adjacent non-neoplastic mucosae of the colonshowed positive PTHrP immunoreactivity. In contrast,immunostaining of PTHrP was recognized in mostcolorectal adenocarcinomas. These results suggestthat PTHrP is related to carcinogenesis of theintestinal mucosa and to malignant transformationof the adenoma–carcinoma sequence, and that PTHrPis a critical marker for distinguishing adeno-carcinoma from adenoma. All the cases with adenocar-cinoma in adenoma showed a diverse expression pattern.The greater expression in poorly differentiated adeno-carcinoma suggests that PTHrP may be an oncofetalprotein.

The most important prognostic indicators for colo-rectal carcinoma are extension of the tumour into thebowel wall and the presence of lymph node metastasis,as described in Dukes’ classification. There have beenseveral studies indicating that PTHrP expression isrelated to tumour invasion and metastasis. The invasiveparts of primary tumours were intensely stained com-pared with other parts in thyroid carcinomas.15

Luparello et al.21 indicated that the cell line of PTHrP-producing breast carcinoma produced extracellular pro-teolytic enzymes, and PTHrP was considered tointervene in local control of the invasive process inbreast carcinoma. In our study, the immunoreactivity ofPTHrP was closely correlated with tumour invasion andmetastasis. The intensity of staining had a significantcorrelation with increasing grades of Dukes’ classifi-cation. These results suggest that the expression ofPTHrP may also have prognostic value in colorectaladenocarcinomas.

PTHrP appears to be involved in an autocrine/paracrine regulation of human renal cell carcinoma.22

Similarly, PTHrP has been reported to be a potentialautocrine/paracrine growth factor in prostate cancercells in culture.11 To elucidate the autocrine/paracrinemechanism, a demonstration of PTHrP receptor expres-sion in tumour cells is necessary. PTHrP gene expressionis regulated by many different cytokines and regulatoryproteins in several kinds of carcinoma cells.23 Theinteraction with such factors might regulate the prolifer-ation and local invasion of colorectal adenocarcinomacells.

In conclusion, PTHrP is related to the carcinogenesis,differentiation, and progression of colorectal cancer.Furthermore, this protein may have potential as aprognostic marker for colorectal tumours.

ACKNOWLEDGEMENTS

We wish to thank Dr Yoshisada Shibata (RadiationEffects Research Foundation, Nagasaki) for statistical

Copyright ? 1999 John Wiley & Sons, Ltd.

analysis and M. S. Razzaque, MD (Department ofPathology II) for editorial assistance.

REFERENCES

1. Moseley JM, Kubota M, Kubota M, et al. Parathyroid hormone-relatedprotein purified from a human lung cancer cell line. Proc Natl Acad Sci USA1987; 87: 5048–5052.

2. Suva LJ, Wiaslow GA, Wettenhall REH, et al. A parathyroid hormone-related protein implicated in malignant hypercalcemia: cloning andexpression. Science 1987; 237: 893–896.

3. Campos RV, Asa SL, Drucher DJ. Immunocytochemical localization ofparathyroid hormone-like peptide in the rat fetus. Cancer Res 1991; 51:6351–6357.

4. Ikeda K, Weir EC, Mangin M, et al. Expression of messenger ribonucleicacid encoding a parathyroid hormone-like peptide in normal human andanimal tissues with abnormal expression in human parathyroid adenomas.Mol Endocrinol 1988; 2: 1230–1236.

5. Thiede MA, Rodan GA. Expression of a calcium mobilizing parathyroidhormone-like peptide in lactating mammary tissue. Science 1988; 242:278–280.

6. Rodda CP, Kubota M, Heath JA, et al. Evidence for a novel parathyroidhormone-related protein in fetal lamb parathyroid glands and sheep pla-centa: comparisons with a similar protein implicated in humoral hypercal-cemia of malignancy. J Endocrinol 1988; 117: 261–271.

7. Burton PBJ, Moniz C, Quirke P, et al. Parathyroid hormone-relatedpeptide: expression in fetal and neonatal development. J Pathol 1992; 167:291–296.

8. Enomoto H, Yamashita S, Usa T, et al. Autocrine function of parathyroidhormone-related peptide in rat osteoblast-like cells. Biochem Biophys ResCommun 1994; 191: 1261–1269.

9. Ikeda K. Molecular biology of parathyroid hormone-related peptide(Review). Jpn J Clin Med 1995; 53: 835–843.

10. Roskams T, Willems M, Campos RV, et al. Parathyroid hormone-relatedpeptide expression in primary and metastatic liver tumours. Histopathology1993; 23: 519–525.

11. Iwamura M, di Sant’Agnese PA, Wu G, et al. Immunohistochemicallocalization of parathyroid hormone-related protein in human prostatecancer. Cancer Res 1993; 53: 1724–1726.

12. Vargas SJ, Gillespie MT, Powell GJ, et al. Localization of parathyroidhormone-related protein mRNA expression in breast cancer and metastaticlesions by in situ hybridization. J Bone Miner Res 1992; 7: 971–979.

13. Akino K, Ohtsuru A, Yano S, et al. Antisense inhibition of parathyroidhormone-related peptide gene expression reduces malignant pituitary tumorprogression and metastasis in the rat. Cancer Res 1996; 56: 77–86.

14. Jass SR, Sobin LH. Histological Typing of Intestinal Tumours. WHO, 2ndedn. Berlin: Springer-Verlag, 1989.

15. Nakashima M, Ohtsuru A, Luo WT, et al. Expression of parathyroidhormone-related peptide in human thyroid tumours. J Pathol 1995; 175:227–236.

16. Ozeki S, Ohtsuru A, Seto S, et al. Evidence that implicates the parathyroidhormone-related peptide in vascular stenosis. Arterioscler Thromb 1996; 16:565–575.

17. Husong L, Patricia K, Mary L, et al. Widespread expression of theparathyroid hormone-related peptide and PTH/PTHrP receptor genes inintestinal epithelial cells. Lab Invest 1995; 73: 864–870.

18. Stewart A, Horst R, Defros L, et al. Biochemical evaluation of patients withcancer-associated hypercalcemia: evidence of humoral and nonhumoralgroups. N Engl J Med 1980; 303: 1377–1383.

19. Omenn G, Roth S, Baker W. Hyperparathyroidism associated with malig-nant tumors of nonparathyroid origin. Cancer 1969; 24: 1004–1012.

20. Motokura T, Endo K, Kumaki K. Neoplastic transformation of normal ratembryo fibroblasts by a mutated p53 and an activated ras oncogene inducesparathyroid hormone-related peptide gene expression and causes hyper-calcemia in nude mice. J Biol Chem 1995; 270: 30857–30861.

21. Luparello C, Burtis WJ, Raue F, et al. Parathyroid hormone-related peptideand 8701-BC breast cancer cell growth and invasion in vitro: evidence forgrowth-inhibiting and invasion-promoting effects. Mol Cell Endocrinol1995; 111: 225–232.

22. Burton PBJ, Moniz C, Knight DE. Parathyroid hormone-related peptidecan function as an autocrine growth factor in human renal cancer cellcarcinoma. Biochem Biophys Res Commun 1990; 167: 1134–1138.

23. Kiriyama T, Matthew TG, Jane AG et al. Transforming growth factorstimulation of parathyroid hormone-related protein (PTHrP): a paracrineregulator? Mol Cell Endocrinol 1993; 92: 55–62.

J. Pathol. 187: 217–222 (1999)