immunohistochemical analysis of thymidylate synthase, p16ink4a, cyclin-dependent kinase 4 and cyclin...

Pathology International

2004;

54

: 564–575

et al

.

Correspondence: Shingo Kamoshida, BS, Department of Pathology,Fujita Health University School of Medicine, Toyoake, Aichi 470-1192, Japan. Email: [email protected]

Received 18 December 2003. Accepted for publication 14 March2004.

Original Article

Immunohistochemical analysis of thymidylate synthase, p16

INK4a

, cyclin-dependent kinase 4 and cyclin D1 in colorectal cancers receiving preoperative chemotherapy: Significance of p16

INK4a

-mediated cellular arrest as an indicator of chemosensitivity to 5-fluorouracil

Shingo Kamoshida,

1

Hiroshi Matsuoka,

2

Kazuya Shiogama,

1

Atsuji Matsuyama,

1

Ryoichi Shimomura,

1

Ken-ichi Inada,

1

Morito Maruta

2

and Yutaka Tsutsumi

1

Departments of

1

Pathology and

2

Surgery, Fujita Health University School of Medicine, Toyoake, Japan

High expression of thymidylate synthase (TS) is allegedlyassociated with the chemoresistance to 5-fluorouracil(5-FU) in colorectal cancers. However, low TS expressiondoes not necessarily imply chemosensitivity. Inactivation ofp16

INK4a

correlates with poor prognosis in various cancers.We immunohistochemically evaluated the relationshipbetween the expression of TS, p16

INK4a

, CDK4 and cyclin D1and the effect of 5-FU-based chemotherapy in colorectalcancers. After antigen retrieval, immunoperoxidase stainingwas performed on the paraffin-embedded, biopsy andsurgical specimens of 37 advanced colorectal cancerspreoperatively treated with peroral administration of 5-FUderivatives. As a control group, 31 colorectal cancers with-out preoperative treatment were analyzed. High TS expres-sion was found in 23 (74%) of 31 tumors resected fromhistological non-responders and in 19 (61%) of 31 controlsbut in none of six responders. High p16

INK4a

expression wasseen in 83% of the responders, 52% of the non-respondersand 32% of the controls. The TS-low/p16

INK4a

-high phenotypewas noted in 83% of the responders, but only in 3% of thenon-responders (

P

=

0.0001). Induction of p16

INK4a

expres-sion after chemotherapy was predominantly seen in theresponders. Neither CDK4 nor cyclin D1 expression wasrelated to the chemotherapeutic effects. In conclusion, thecombination of low expression of TS and induction ofp16

INK4a

after chemotherapy can be important indicators ofthe sensitivity to 5-FU-based chemotherapy in colorectalcancers.

Key words:

5-fluorouracil, colorectal cancer, cyclin D1, p16

INK4a

,thymidylate synthase

Colorectal cancer is one of the most common malignanciesin developed countries. Chemotherapy is the major treatmentmodality when the diagnosis is made at an advanced stage.5-Fluorouracil (5-FU) has been used widely in the treatmentof solid tumors including colorectal cancer.

1

However, theresponse rate to this agent does not exceed 30% in colorec-tal cancer, and the overall survival remains low. Analysis ofthe indicators for identifying patients who will be responsiveor unresponsive to 5-FU is one of the approaches for theindividualization of treatment, leading to the appropriate deci-sion of choice for a 5-FU-based regimen, and thus the avoid-ance of unpleasant side-effects.

Thymidylate synthase (TS) catalyzes the methylation ofdeoxyuridine monophosphate (dUMP) to deoxythymidinemonophosphate (dTMP), an essential step for DNA syn-thesis.

2

The anti-cancer effects of 5-FU are biochemicallymediated through its metabolic conversion by two additionalenzymes (thymidine phosphorylase and thymidine kinase)into fluorodeoxyuridine monophosphate (FdUMP), whicheventually forms a ternary covalent complex with TS andreduced folic acid, resulting in the inhibition of TS activity.

It has been suggested that high expression of TS is asso-ciated with the 5-FU resistance of colorectal cancer cellsand the poor prognosis of colorectal cancer patients.

3–7

Werecently assessed the immunohistochemical expression ofTS and p53 protein in the same biopsy and the surgicalsamples of advanced colorectal cancers preoperativelytreated with 5-FU derivatives.

8

High TS expression inresected tumors was demonstrated in 74% of the histologicalnon-responders but none in the responders, whereas therewas no association of the expression of p53 with the chemo-therapeutic effects. Thymidylate synthase expression in thepreoperative biopsies and resected tumors was concordantin 80% of cases when two or more biopsy fragments were

p16

INK4a

as an indicator of 5-FU chemosensitivity 565

available, suggesting that TS immunostaining in preoperativebiopsy specimens can be used for the prediction of 5-FUchemoresistance. However, it was noted that 26% of the non-responders also showed low TS expression, indicating thatthis phenotype is not strictly specific for the prediction of thesensitivity to 5-FU-based chemotherapy. Thus, the authorsare interested in finding other sensitivity determinants thatcan be used for distinguishing the responders from the non-responders with TS-low tumors.

p16

INK4a

protein, a product of the

CDKN2/MTS1

genelocated at the chromosomal region 9p21, is a member of theINK4 family of cyclin-dependent kinase (CDK) inhibitors.

9

p16

INK4a

binds specifically to CDK4 to inhibit the activity ofCDK4-cyclin D1 complex, thereby blocking phosphorylationof the retinoblastoma tumor suppressor protein (pRb), whichallows the dissociation of E2F and subsequent transit to theS phase. The fact that high expression of p16

INK4a

results inG

1

arrest and the inactivation of p16

INK4a

is associatedwith the outgrowth of human tumors suggest that p16

INK4a

acts as not only a cellular cycle regulator but also a tumorsuppressor.

10–13

Alterations of the elements of p16

INK4a

-mediated cellulararrest have frequently been reported in various types of can-cers.

14–28

In particular, inactivation of the

CDKN2/MTS1

geneleading to loss of p16

INK4a

protein expression is associatedwith the biological aggressiveness of cancer cells and thepoor prognosis of cancer patients.

19,20,22–25,28

As for the col-orectal cancers, there have been several studies thatassessed the relationship between the expression ofp16

INK4a

-associated cell cycle proteins and TS expression orpatient outcome.

16,18,21,23,27,29

However, to the best of ourknowledge, no studies have examined the significance of theelements in p16

INK4a

-mediated cellular arrest as indicators ofthe chemotherapeutic effects in colorectal cancers.

In the present study, we immunohistochemically investi-gated the relationship between the chemotherapeutic effectsof 5-FU derivatives and the expression of p16

INK4a

, CDK4,cyclin D1 and TS in biopsy specimens and resected tumorsof advanced colorectal cancers preoperatively treated withperoral administration of 5-FU derivatives. We have previ-ously examined and reported the TS expression levels inrelation to the chemotherapeutic effects.

8

We demonstratehere that 5-FU-induced high p16

INK4a

expression in tumors ofthe low TS expression group can be a relevant and usefulindicator in determining the chemosensitivity to 5-FU.

MATERIALS AND METHODS

Patients and specimens

The study population consisted of 37 patients (23 menand 14 women; mean age, 60.3 years; age range, 34–

81 years; median, 60). All patients had undergone surgicalresection for advanced colorectal cancer in 2001 at theDepartment of Surgery, Fujita Health University Hospital,Toyoake, Japan. The clinicopathological features are sum-marized in Table 1. Histological grade, depth of invasion,lymph node metastasis, distant metastasis and the clinicalstage were classified according to the

TNM Classificationof Malignant Tumors

.

30

All patients received the peroraladministration of 5-FU derivatives, tegafur and uracil (UFT,600 mg/m

2

per day for 2 weeks) or 5

¢

-deoxyfluorouridine(5

¢

-DFUR, 800 mg/m

2

per day for 2 weeks) until 1 daybefore operation. No other treatment was givenpreoperatively. Informed consent for the administration ofpreoperative chemotherapy and also for the use oftumor tissue for analyzing antigen expression wasobtained from all patients.

Preoperative biopsied specimens and surgicallyresected materials were routinely fixed in 10% formalinand embedded in paraffin wax. Sections, 3

m

m-thick, werecut and mounted on aminopropyltriethoxysilane slidesand stained with hematoxylin and eosin (HE) in orderto assess the histopathological features and the respon-siveness to preoperative chemotherapy under a lightmicroscope.

Preoperative biopsied specimens and surgically resectedmaterials from 31 control patients who did not receive pre-operative chemotherapy (16 men and 15 women; mean age,63.4 years; age range, 43–92 years; median, 63) were alsosimilarly analyzed.

Histological assessment of chemotherapeutic effects

Chemotherapeutic effects were histologically evaluatedon the basis of the

Japanese Classification of ColorectalCarcinoma

.

31

Major grading (grades 0–3) and an addi-tional minor grading for grade 1 (grades 1a and 1b)were used, based upon the degree of necrosis or disap-pearance of tumor cells in the lesion: grade 0, nochange; grade 1, mild change (grade 1a, necrosis ordisappearance of the tumor seen in less than 1/3 of theentire lesion; and grade 1b, necrosis or disappearanceof the tumor seen in more than 1/3 but less than 2/3 ofthe entire lesion); grade 2, moderate change (necrosisor disappearance of the tumor seen in more than 2/3of the entire lesion, but still with remaining viable tumorcells); and grade 3, severe change (no viable tumorcells remaining). Patients with grades 1b and 2 werecategorized as the histological responders, and patientswith grades 0 and 1a as the histological non-responders, on the basis of the contents of the previousreport.

32

No grade 3 lesion was observed in the currentseries.

566 S. Kamoshida

et al

.

Immunohistochemistry

In addition to paraffin blocks of the biopsy specimens, oneor two representative blocks of the resected tumor principallycontaining the tumor-non-tumor junction were chosen forimmunohistochemical analysis. Consecutive sections wereevaluated for the expression of TS, p16, CDK4 and cyclin D1,using the universal immuno-enzyme polymer method. Thesections were deparaffinized with xylene, and rehydrated ingraded ethanol. Endogenous peroxidase activity wasquenched with 0.03% hydrogen peroxide in methanol atroom temperature for 30 min. Heat-induced antigen retrievalwas applied using a pressure cooker (Delicio 6L, T-FAL,Rumily, France) for 10 min. Optimal soaking solutions wereselected for the respective markers determined by prelimi-nary experiments: 10 mmol/L citrate buffer, pH 7.0 forp16

INK4a

and CDK4; and 1 mmol/L ethylenediaminetetraace-tate (EDTA) solution, pH 8.0 for TS and cyclin D1.

33

Afterpressure cooking, the sections were left for cooling in the

soaking solution for 30 min. The sections were then incu-bated with the primary antibody, which reacts specifically withTS (rabbit polyclonal, 1:200 dilution; Taiho Pharmaceutical,Saitama, Japan), p16

INK4a

(mouse monoclonal, clone 6H12,1:20 dilution; Novocastra Laboratories, Newcastle, UK),CDK4 (mouse monoclonal, clone DCS-156, 1:100 dilution;Santa Cruz Biotechnology, CA, USA) or cyclin D1 (mousemonoclonal, clone AM29, prediluted; Zymed Laboratories,CA, USA), overnight at room temperature. After rinsing in10 mmol/L phosphate-buffered saline (PBS), pH 7.2, the sec-tions were incubated with the universal immunoperoxidasepolymer, antimouse and -rabbit (Histofine

a

Simple Stain MAXPO; Nichirei, Tokyo, Japan) at room temperature for 30 min.The reaction products were visualized in 50 mmol/L Tris-HClbuffer, pH 7.6, containing 50 mg/dL diaminobenzidine tet-rahydrochloride (DAB) and 0.003% hydrogen peroxide for TSand p16

INK4a

, or commercially available DAB liquid systemcontaining a sensitizer (DakoCytomation, CA, USA) forCDK4 and cyclin D1. The nuclei were lightly counterstained

Table 1

Relationship between expression of thymidylate synthase, p16

INK4a

, CDK4 or cyclin D1 and clinicopathological features of patientswho underwent surgery after 5-fluorouracil-based chemotherapy

Clinicopathological featuresTS p16

INK4a

CDK4 Cyclin D1L H

P

-value L H

P

-value L H

P

-value L H

P

-value

Age (years)

<

60 5 12 0.50 7 10

>

0.99 16 1

>

0.99 17 0 0.23

≥

60 9 11 9 11 18 2 17 3Gender

Male 7 16 0.30 11 12 0.52 22 1 0.54 21 2

>

0.99Female 7 7 5 9 12 2 13 1

Tumor locationColon 8 12

>

0.99 9 11

>

0.99 18 2

>

0.99 18 2

>

0.99Rectum 6 11 7 10 16 1 16 1

Histological grade†‡G1 6 13 0.50 8 11 0.74 17 2

>

0.99 16 3 0.23G2 8 9 6 11 16 1 17 0

Depth of invasion†T2 3 5 0.35 1 7 0.11 8 0 0.25 7 1 0.54T3 2 8 6 4 8 2 10 0T4 9 10 9 10 18 1 17 2

Lymph node metastasis†N0 5 12 0.45 7 10 0.72 15 2 0.67 14 3 0.15N1 6 9 6 9 14 1 15 0N2 3 2 3 2 5 0 5 0

Distant metastasis†M0 11 20 0.65 14 17 0.68 28 3

>

0.99 29 2 0.42M1 3 3 2 4 6 0 5 1

Clinical stage†I and II 5 10 0.78 7 8 0.86 13 2 0.56 13 2 0.28III 6 10 7 9 15 1 16 0IV 3 3 2 4 6 0 5 1

Dukes’ classificationA 1 3 0.42 2 2 0.96 3 1 0.56 4 0 0.50B 4 7 5 6 10 1 9 2C 6 12 7 11 17 1 17 1D 3 1 2 2 4 0 4 0

Agent administeredUFT 9 15

>

0.99 11 13 0.74 22 2

>

0.99 22 2 >0.995

¢

-deoxyfluorouridine 5 8 5 8 12 1 12 1

H, high expression; L, low expression; UFT, tegafur and uracil. †According to the

TNM Classification of Malignant Tumors

.

30

‡Histological grade wasevaluated in 36 of 37 tumors treated with preoperative chemotherapy. Two cases of G3 were seen in the control group.

p16

INK4a

as an indicator of 5-FU chemosensitivity 567

with Mayer’s hematoxylin. No immunoreactions were seen inthe negative control slides, in which the primary antibody wasreplaced by PBS.

The HE-stained and immunostained sections were inde-pendently reviewed by three investigators (S.K., K.S. andR.S.), without prior knowledge of the clinical data of thepatients. When discrepancies existed, discussion on a mul-tiheaded microscope was made to adjust the differences. Forevaluating p16

INK4a

, CDK4 and cyclin D1 immunoreactivity,nuclear staining of the target cells was regarded as positive,regardless of cytoplasmic staining. Immunostained sectionswere classified into two groups in a semiquantitative manner.Namely, low expression (‘negative or positive but in less than30% of tumor cells’ for TS and p16

INK4a

; ‘negative or positivebut in less than 10% of tumor cells’ for CDK4 and cyclin D1)versus high expression (‘positive in 30% of tumor cells ormore’ for TS and p16

INK4a

; ‘positive in 10% of tumor cells ormore’ for CDK4 and cyclin D1), irrespective of the intensityof staining.

Data on TS expression were already reported in our pre-vious study using the same series of colorectal cancers.

8

Statistical analysis

The Fisher’s exact probability test was used for determiningthe statistical significance of correlations between the antigenexpression and the histological chemotherapeutic effects, theage or gender of the patients, tumor location, histologicalgrade, distant metastasis or the kind of agent administered.Correlation between the antigen expression and the depth ofinvasion, lymph node metastasis, clinical stage or Dukes’classification was analyzed using the Chi-squared test. TheMann–Whitney’s

U

-test was used for evaluating the associ-ation of the chemotherapeutic effects with the antigen expres-sion in the preoperatively biopsied and surgically resected

specimens. Categories consisting of less than 5% (i.e. twocases or less) of the total number of cases were combinedinto appropriate categories (Stages I and II), or excluded fromthe analysis (‘mucinous carcinoma’ in histology).

P

-values

<

0.05 were considered statistically significant.

RESULTS

Clinicopathological features and expression of thymidylate synthase, p16

INK4a

, CDK4 and cyclin D1

Thymidylate synthase immunoreactivity was observed in thecytoplasm of cancer cells. p16

INK4a

usually showed simulta-neous distribution in both the nuclei and cytoplasm, but cyto-plasmic expression of p16

INK4a

without nuclear staining wasfocally observed in only two control cases. CDK4 and cyclinD1 were invariably localized in the nuclei. The antigens wereexpressed with a heterogeneous distribution of positivity.p16

INK4a

occasionally showed predominant positivity in theinvasive component of the tumors, but in general, there waslittle difference seen in the staining pattern between the inva-sive and non-invasive components within the same tumors,as was so in the other three antigens. Non-neoplastic cryptepithelial cells located in the basal half of the mucosa, aswell as nonepithelial cells such as germinal center lympho-cytes, plasma cells, endothelial cells, fibroblasts and smoothmuscle cells, were also immunoreactive for TS. Immunore-activity of p16

INK4a

was occasionally seen in macrophages,fibroblasts and endothelial cells in the tumor tissue, and onlya small number of crypt epithelial cells. Non-neoplasticmucosa was negative for CDK4 and cyclin D1, while fibro-blasts were frequently positive for cyclin D1.

The relationship between clinicopathological features andthe expression of TS, p16

INK4a

, CDK4 or cyclin D1 in 37 casesreceiving chemotherapy are summarized in Tables 1 and 2.

Table 2

Relationship between expression of TS, p16

INK4a

, CDK4 or cyclin D1 and the effects of 5-fluorouracil-based chemotherapy insurgically resected tumors

Antigenexpression

Histologicalresponders (%)*

(

n

=

6)

Histologicalnon-responders (%)**

(

n

= 31)

Control untreatedcases (%)***

(

n

=

31)

P

-value

* versus ** * versus *** ** versus ***

TSLow 6 (100)† 8 (26) 12 (39) 0.001‡ 0.008‡ 0.42High 0 (0) 23 (74) 19 (61)

p16

INK4a

Low 1 (17) 15 (48) 21 (68) 0.21 0.03‡ 0.20High 5 (83)† 16 (52) 10 (32)

CDK4Low 6 (100)† 28 (90) 25 (81)

>

0.99 0.56 0.47High 0 (0) 3 (10) 6 (19)

Cyclin D1Low 6 (100)† 28 (90) 19 (61)

>

0.99 0.15 0.02‡High 0 (0) 3 (10) 12 (39)

TS, thymidylate synthase. †Two cases with histological grade 2 effects are included. ‡Statistically significant.

568 S. Kamoshida

et al

.

There was no significant association of the expression of TS,p16

INK4a

, CDK4 or cyclin D1 with the patient’s age, gender,tumor location, histological grade, depth of invasion, lymphnode metastasis, distant metastasis, clinical stage, Dukes’classification or the kind of agent administered.

Chemotherapeutic effects and expression of thymidylate synthase, p16

INK4a

, CDK4 and cyclin D1 in resected colorectal tumors with 5-fluorouracil-based preoperative chemotherapy

Chemotherapeutic responders were histologically selected ina total of six (four cases of grade 1b and two cases of grade2) (16%) out of 37 cases (Fig. 1a). Diagnostic imaging bycomputed tomography and endoscopic examination showedmarked regression of the tumors in two cases with histolog-ical grade 2. Thirty-one cases (22 cases of grade 0 and ninecases of grade 1a) (84%) were categorized in the non-responders (Fig. 2a).

Table 2 shows the relationship between the chemothera-peutic effects and the expression of TS, p16

INK4a

, CDK4 orcyclin D1 in the resected tumors. High TS expression wasdetected in none of the responders, in 23 lesions (74%) ofthe non-responders and in 19 lesions (61%) of the controluntreated cases; the differences between the respondersand the non-responders, and between the responders and

the control cases were statistically significant (

P

=

0.001 and

P

=

0.008, respectively). All the responders showed low TSexpression, whereas eight (26%) of the 31 non-responderswere also categorized in the low TS group, as reportedpreviously.

8

High expression of p16

INK4a was observed in five (includingtwo cases with histological grade 2) (83%) of the six respond-ers in 16 lesions (52%) of the 31 non-responders and in 10lesions (32%) of the 31 control cases; a significant differencewas found between the responders and the control cases(P = 0.03). High CDK4 expression was detected in none ofthe responders, in three lesions (10%) of the non-responders(one with high p16INK4a expression and two with low p16INK4a

expression) and in six lesions (19%) of the control cases(three with high p16INK4a expression and three with lowp16INK4a expression); there were no significant differencesamong the three groups. Cyclin D1 was highly expressed asfollows: in none of the responders; in three (10%) of the non-responders (two with high p16INK4a expression and one withlow p16INK4a expression); and in 12 (39%) of the control cases(five with high p16INK4a expression and seven with low p16INK4a

expression); the difference between the non-responders andthe control cases was significant (P = 0.02).

Table 3 summarizes the relationship between the chemo-therapeutic effects and the expression of p16INK4a, CDK4 orcyclin D1 in relation to TS expression in the resected tumors.

Figure 1 Expression of thymidylate synthase (TS) and p16INK4a in a representative resected colon cancer responsive to 5-fluorouracilderivative (grade 2). (a) Hematoxylin–eosin staining; (b) TS immunoreactivity; and (c) p16INK4a immunoreactivity. Chemotherapy-inducedhistological changes include the disintegration of glandular structures with marked fibrosis in the stroma and nuclear degradation of the tumorcells. Only a few cancer cells (arrow) and plasma cells (arrowheads) in the stroma are positive for TS in the cytoplasm. In contrast, p16INK4a

is expressed in the nuclei and cytoplasm of a number of cancer cells.

p16INK4a as an indicator of 5-FU chemosensitivity 569

Figure 2 Expression of thymidylate synthase (TS) and p16INK4a in two resected colon cancers unresponsive to 5-fluorouracil (5-FU) derivative.(a,d) Hematoxylin–eosin staining; (b,e) TS immunoreactivity; and (c,f) p16INK4a immunoreactivity. Little histological changes are discernibleafter 5-FU-based chemotherapy. In the case shown in the top panels (a–c), a number of cancer cells express both TS and p16INK4a, butcharacteristically showing a reciprocal localization pattern. In the case shown in the bottom panels (d–f), there are no TS-positive cells andonly a few p16INK4a-positive cells (arrows). Normal crypt epithelial cells (asterisks) are positive for TS but negative for p16INK4a. Plasma cellsin the stroma (arrowhead) and endothelial cells (double arrow) are positive for TS.

Table 3 Expression of p16INK4a, CDK4 or cyclin D1 and effects of 5-fluorouracil-based chemotherapy in relation to TS expression in surgicallyresected tumors

ExpressionHistological

responders (%)*Histological

non-responders (%)**Control untreated

cases (%)***

P-value


Low TS (n = 6) (n = 8) (n = 12)Low p16INK4a 1 (17) 7 (88) 7 (58) 0.03‡ 0.15 0.32High p16INK4a 5 (83)† 1 (13) 5 (42)Low CDK4 6 (100)† 7 (88) 10 (83) >0.99 0.53 >0.99High CDK4 0 (0) 1 (13) 2 (17)Low cyclin D1 6 (100)† 7 (88) 6 (50) >0.99 0.05 0.16High cyclin D1 0 (0) 1 (13) 6 (50)

High TS (n = 0) (n = 23) (n = 19)Low p16INK4a 8 (35) 14 (74) 0.02‡High p16INK4a 15 (65) 5 (26)Low CDK4 21 (91) 15 (79) 0.38High CDK4 2 (9) 4 (21)Low cyclin D1 19 (83) 13 (68) 0.47High cyclin D1 4 (17) 6 (32)

Low TS and high p16INK4a 5 (83)† 1 (3) 5 (16) <0.0001‡ 0.003‡ 0.20High TS or low p16INK4a 1 (17) 30 (97) 26 (84)

TS, thymidylate synthase. †Two cases with the effects of histological grade 2 are included. ‡Statistically significant.

570 S. Kamoshida et al.

In the low TS group, high expression of p16INK4a was observedin only one (13%) of the non-responders and five (42%) inthe control group. Five (83%) of the responders and only one(3%) of the non-responders were categorized into the TS-low/p16INK4a-high phenotype (Fig. 1). In 10 of 15 tumors withhigh expression of both TS and p16INK4a, cells expressingthese antigens revealed a reciprocal localization pattern(Fig. 2). Expression of CDK4 and cyclin D1 was not associ-ated with the chemotherapeutic effects, even in the tumorswith low TS expression (Fig. 3).

Comparative expression of thymidylate synthase, p16INK4a, CDK4 and cyclin D1 in preoperative biopsied specimens and surgically resected materials

Adequate biopsy material containing two or more pieces wasavailable in 25 of the 37 tumors receiving the preoperativechemotherapy and in 17 of the 31 control untreated tumors.Table 4 shows the relationship between the chemotherapeu-tic effects and the expression of TS, p16INK4a, CDK4 or cyclin

D1 in the preoperative biopsied specimens. High TS expres-sion was seen in 13 (65%) of 20 non-responders but noneof five responders; the difference between the two groupswas significant (P = 0.01). Twelve (71%) of the 17 controluntreated cases also showed high TS expression. However,there were no significant differences in the expression ofp16INK4a, CDK4 and cyclin D1 in the preoperative biopsiesamong the responders, the non-responders and the controlgroup.

We also assessed the comparative expression of antigensin paired samples of the preoperative biopsies and the sur-gically resected tumors (Table 5). High expression of p16INK4a

in the preoperative biopsy specimens and the surgicallyresected materials was observed in one (20%) and four(80%) in the responders, five (25%) and 10 (50%) in the non-responders, and six (35%) and five (29%) in the controlgroup, respectively. The data indicated that p16INK4a expres-sion was induced by the chemotherapy in the responders: Asignificant difference was noted between the responders andthe control group (P = 0.01). However, for the other threeantigens, no significant difference between the biopsy spec-

Figure 3 Expression of CDK4 and cyclin D1 in two resected colon cancers from non-responder (top panels, a–c) and control untreatedcases (bottom panels, d–f). (a,d) Hematoxylin–eosin staining; (b,e) CDK4 immunoreactivity; and (c,f) cyclin D1 immunoreactivity. Both thenon-responder and control tumor show negligible nuclear expression of CDK4. Cyclin D1-positive nuclei are observed diffusely in the controlcase but focally in the non-responder.


imens and the resected materials was detected among thethree groups. Figure 4 shows the immunostaining patterns ofTS and p16INK4a in the preoperative biopsy specimen and theresected material in the responder.

DISCUSSION

The preoperative chemotherapy may provide information onthe responsiveness of a given tumor to a particular drugregimen. Fujii used a short administration period (10 days)of UFT for the preoperative treatment of colorectal cancer,and demonstrated that the histopathological analysis of theresected specimens following preoperative chemotherapycan predict the chemosensitivity to postoperative 5-FU-based chemotherapy.32 Matsuura et al. described that preop-erative 5-FU suppository administration induced apoptosisand suppressed angiogenesis in colorectal cancers in a p53-

independent manner.34 In our previous study, based on thesame series of colorectal cancers that were examined in thepresent study, high expression of TS was an excellent pre-dictor of the resistance to preoperative 5-FU-based chemo-therapy, but low TS expression did not necessarily indicatethe responsiveness to the preoperative chemotherapy.8

p16INK4a, a product of the CDKN2/MTS1 gene located atthe chromosomal region 9p21, is a negative regulator of G1/S-phase transition that promotes cell-cycle arrest via the Rbtumor suppressor pathway, and its expression increasesalong with the onset of cellular senescence.9–13 In colorectalcancers, homozygous deletion and mutation of the CDKN2/MTS1 gene are reportedly very rare,35–37 while methylationof the gene is detected in 18–53% of the cancerseries.19,20,23,25,37 Reportedly, inactivation of the CDKN2/MTS1gene is generally associated with an aggressive phenotypeand poor prognosis in a wide range of neoplasms, includingcolorectal cancers,19,20,23,25,28 and is usually paralleled with

Table 4 Relationship between expression of TS, p16INK4a, CDK4 or cyclin D1 and the effects of 5-fluorouracil-based chemotherapy inpreoperative biopsies

Antigen expression

Histologicalresponders (%)*

(n = 5)


(n = 20)


(n = 17)P-value


TSLow 5 (100) 7 (35) 5 (29) 0.01† 0.01† >0.99High 0 (0) 13 (65) 12 (71)

p16INK4a

Low 4 (80) 15 (75) 11 (65) >0.99 >0.99 0.72High 1 (20) 5 (25) 6 (35)

CDK4Low 4 (80) 19 (95) 15 (88) 0.37 >0.99 0.58High 1 (20) 1 (5) 2 (12)

Cyclin D1Low 5 (100) 14 (70) 13 (76) 0.29 0.54 0.72High 0 (0) 6 (30) 4 (24)

TS, thymidylate synthase. †Statistically significant.

Table 5 Comparative expression of TS, p16INK4a, CDK4 or cyclin D1 and chemotherapeutic effects in preoperative biopsy specimens andsurgically resected materials

Antigenexpression

Histological responders(%)*

(n = 5)


(n = 20)


(n = 17)P-valueBiopsied

specimensResectedspecimens

Biopsiedspecimens

Resectedspecimens

Biopsiedspecimens

Resectedspecimens * versus ** * versus *** ** versus ***

TSLow 5 (100) 5 (100) 7 (35) 5 (25) 5 (29) 8 (47) 0.59 0.91 0.23High 0 (0) 0 (0) 13 (65) 15 (75) 12 (71) 9 (53)

p16INK4a

Low 4 (80) 1 (20) 15 (75) 10 (50) 11 (65) 12 (71) 0.27 0.01† 0.09High 1 (20) 4 (80) 5 (25) 10 (50) 6 (35) 5 (29)

CDK4Low 4 (80) 5 (100) 19 (95) 19 (95) 15 (88) 13 (76) 0.28 0.29 0.61High 1 (20) 0 (0) 1 (5) 1 (5) 2 (12) 4 (24)

Cyclin D1Low 5 (100) 5 (100) 14 (70) 18 (90) 13 (76) 11 (65) 0.36 >0.99 0.11High 0 (0) 0 (0) 6 (30) 2 (10) 4 (24) 6 (35)

TS, thymidylate synthase. †Statistically significant.


loss or markedly reduced expression levels of p16INK4a proteinand mRNA.28,37 In the present immunohistochemical investi-gation, p16INK4a expression was low in 68% of the resectedtumors in the untreated control patients and in 71% of thepreoperative biopsy specimens. The difference between thefrequency of low expression of p16INK4a protein (68–71%) andthe frequency of p16INK4a methylation (18–53%) in the presentstudy seems to be related to the high threshold definitionused (i.e. high expression was defined as 30% or more cellsstained).

The nuclear and cytoplasmic localization of p16INK4a,including a small number of tumors without nuclear staining,is in agreement with previous studies.14,15,28 As it is unclearwhether or not the cytoplasmic reactivity of p16INK4a is non-specific, we did not evaluate the cytoplasmic staining ofp16INK4a in the present study.

Interestingly, we found that the responder group frequentlyshowed an increase in p16INK4a expression after preoperative5-FU-based chemotherapy. The relationship betweenp16INK4a expression and the sensitivity of cancer cells to anti-

cancer agents including topoisomerase I and II inhibitors,adriamycin, cyclophosphamide and cisplatin has been doc-umented,38–40 but there is no report investigating p16INK4a

expression in relation to 5-FU sensitivity. Shapiro et al. dem-onstrated that the G1 arrest response induced by a widevariety of DNA-damaging agents in non-small cell lung can-cer cells depends on p16INK4a expression, accompanied by adecrease in cyclin D1 levels.38 The association of p16INK4a

expression with increased sensitivity of the non-small celllung cancer cells to topoisomerase I inhibitors was alsoreported by Fukuoka et al.39 Schmitt et al. demonstrated thatp16INK4a levels in lymphomas transplanted in mice markedlyincreased after cyclophosphamide therapy, and claimed thatthe cellular arrest leading to an antitumor effect was induciblefollowing chemotherapy.40 The exact molecular mechanismsof the induction of p16INK4a by administration of 5-FU deriva-tives are not known and require further investigation. How-ever, observations in the present study suggest that p16INK4a

plays an important role in the sensitivity to 5-FU-basedchemotherapy in colorectal cancer.

Figure 4 Expression of thymidylate synthase (TS) and p16INK4a in a preoperative biopsy specimen (top panels, a–c) and the resected material(bottom panels, d–f) of the responder (grade 2). (a,d) Hematoxylin–eosin staining; (b,e) TS immunoreactivity; and (c,f) p16INK4a immunore-activity. Both the preoperative biopsy specimen and resected tumor show low expression of TS. Increased expression of p16INK4a is noted inthe resected tumor when compared with the preoperative biopsy specimen. Thymidylate synthase immunoreactivity is observed in plasmacells (arrows) and endothelial cells (arrowhead) in the stroma. Macrophages (double arrows) are positive for p16INK4a.


When we analyzed the combination of TS and p16INK4a

expression in the resected tumor after chemotherapy, the TS-low/p16INK4a-high phenotype was seen in 83% of the respond-ers but in only 3% of the non-responders. In tumors with ahigh expression of both TS and p16INK4a, the cancer cellsfrequently expressed antigens with a reciprocal localizationpattern. The in vitro study identified TS as an early responsegene in the transition from the G1 to S phase in the cellcycle.41 Thymidylate synthase transcription is at least partlycontrolled by the transcription factor E2F, as it has beenshown that levels of TS and E2F expression are closelylinked in colorectal cancers.42 Lukas et al. demonstrated thatE2F overcame a G1 arrest mediated by p16INK4a protein.43

These results enable us to postulate that the high expressionof TS indirectly inhibits p16INK4a-mediated cellular arrest, andthat low expression of TS is essential for the responsivenessto 5-FU, but the cellular arrest is not induced when thep16INK4a expression remains low. Contrary to our view, Omuraet al. reported that in gastric and colorectal cancers, expres-sion of TS was significantly high in the high p16INK4a expres-sion subgroup when compared with the low p16INK4a

expression subgroup, and pointed out the possibility thatp16INK4a regulates TS expression directly, not through G1checkpoint regulation.21

The authors did not find any association of CDK4 andcyclin D1 expression with the chemotherapeutic effects, butobserved a significant reduction of cyclin D1 expression inthe non-responders when compared with the control group.A significant difference was undetectable in the responders,possibly due to the small number of tumor specimens avail-able. No significance was shown between the paired samplesof the preoperative biopsies and the surgically resectedtumors from the non-responders, but it is plausible that cyclinD1 expression is suppressed by the chemotherapy with 5-FUderivatives. As mentioned above, the DNA-damaging agents,although no reports on 5-FU, have shown to decrease thecyclin D1 levels in non-small cell lung cancer cells.38 Moreextensive studies using a large number of cases should beundertaken in the responder group to confirm such findings.

Hypothetically, the overexpression of CDK4 and cyclin D1causes increased phosphorylation of pRb, resulting inincreased cell proliferation.16,28 The overexpression of CDK4in the absence of p16INK4a has been reported in several kindof tumors, suggesting a possible direct interaction betweentwo cell cycle proteins and an aggressive growth potential ofthe tumors.22 In the present study, the incidence of highexpression of CDK4 and cyclin D1 in the resected materialswas low (13% and 22%), and there were no significant asso-ciations between the expression of CDK4 or cyclin D1 andthat of p16INK4a. The results in the present study suggest thatthe expression of CDK4 and cyclin D1 in colorectal cancersis not so excessive as to be regulated by p16INK4a. Recently,Zhao et al. reported the inverse correlation between the

expression of p16INK4a and CDK4; strong immunostaining ofp16INK4a was a predictor for a better prognosis, whereasstrong immunostaining of CDK4 was a predictor for a poorprognosis.29 However, they mentioned that the localization ofp16INK4a and CDK4 was almost always ‘cytoplasmic’, conflict-ing with most of the other reports. One report has describedthat there was no association between the overexpression ofcyclin D1 and the patient’s survival,18 whereas another reportsuggests that the expression of cyclin D1 correlated with theimproved outcome of the patients.27

In colorectal cancer, loss or inactivation of the Rb gene isconsidered to be a rare event and high pRb expression iscommon.18,44 We immunohistochemically examined pRbexpression in colorectal tumors with low TS expression, butloss of the pRb expression was undetectable in those tumors(data not shown). Therefore, Rb status seems to be unrelatedto the effects of 5-FU-based chemotherapy in colorectalcancers.

p53 protein is involved in the cellular arrest, as well as DNAreplication and apoptosis in response to DNA damage.45 p53status has been studied as a prognostic factor or predictorfor the response to cancer chemotherapy. However, contro-versies exist on the prognostic significance of p53 over-expression and p53 mutations in colorectal cancers.46 Ourprevious study did not demonstrate the association of immu-nohistochemical expression of p53 with the effects of 5-FU-based preoperative chemotherapy.8 Furthermore, there wasno correlation between the chemotherapeutic effects andexpression of p53 in relation to TS or p16INK4a expression(data not shown). Immunohistochemically detected p53 over-expression has generally been used as a surrogate of p53mutations, but this assumption is not always correct.46 Inorder to accurately evaluate the prognostic role of p53 abnor-malities in colorectal cancers, further investigations includingthe definition of standardized methods of analysis areessential.

In conclusion, our results showed that a combination of lowexpression of TS and the induction of p16INK4a after chemo-therapy were indicative, but unfortunately not predictive, ofthe chemosensitivity to 5-FU in colorectal cancers. Althoughthe findings were based upon the analyses of a limited num-ber of patients, we believe they have important clinical impli-cations. The mechanism of induction of p16INK4a expressionafter 5-FU treatment is unknown at the present time. However,its elucidation should be the key point to the preoperativeselection of colorectal cancer patients responsive to 5-FU-based chemotherapy, and also to the possibility of enhancingthe chemosensitivity to 5-FU of colorectal cancer cells. Fur-ther studies are required to determine the factors that havea more direct and specific value than the current combinationof TS and p16INK4a. We are now examining the relationshipbetween the prognosis of colorectal cancer patients withpostoperative chemotherapy and the expression of TS and


p16INK4a in primary and recurrent tumors in order to confirmthe findings of the present study.

REFERENCES

1 Bleiberg H. Colorectal cancer: The challenge. Eur J Cancer1996; 32A: S2–6.

2 Pinedo HM, Peters GFJ. Fluorouracil: Biochemistry and phar-macology. J Clin Oncol 1988; 6: 1653–64.

3 Johnston PG, Fisher ER, Rockette HE et al. The role ofthymidylate synthase expression in prognosis and outcome ofadjuvant chemotherapy in patients with rectal cancer. J ClinOncol 1994; 12: 2640–47.

4 Johnston PG, Lenz H-J, Leichman CG et al. Thymidylate syn-thase gene and protein expression correlate and are associ-ated with response to 5-fluorouracil in human colorectal andgastric tumors. Cancer Res 1995; 55: 1407–12.

5 Peters GJ, van der Wilt CL, van Triest B et al. Thymidylatesynthase and drug resistance. Eur J Cancer 1995; 31A: 1299–305.

6 Leichman CG, Lenz H-J, Leichman L et al. Quantitation ofintratumoral thymidylate synthase expression predicts for dis-seminated colorectal cancer response and resistance to pro-tracted-infusion fluorouracil and weekly leucovorin. J Clin Oncol1997; 15: 3223–9.

7 Aschele C, Lonardi S, Monfardini S. Thymidylate synthaseexpression as a predictor of clinical response tofluoropyrimidine-based chemotherapy in advanced colorec-tal cancer. Cancer Treat Rev 2002; 28: 27–47.

8 Kamoshida S, Matsuoka H, Matsuyama A, Shimomura R,Maruta M, Tsutsumi Y. Immunohistochemical demonstration ofthymidylate synthase (TS), p53 protein and bcl-XL protein incolorectal cancer with preoperative peroral chemotherapy: TSas marker of unresponsiveness to 5-fluorouracil. Ann CancerRes Ther 2003; 11: 73–94.

9 Serrano M, Hannon GJ, Beach D. A new regulatory motif incell-cycle control causing specific inhibition of cyclin D/CDK4.Nature 1993; 366: 704–7.

10 Lukas J, Aagaard L, Strauss M, Bartek J. Oncogenic aberra-tions of p16INK4/CDKN2 and cyclin D1 cooperate to deregulateG1 control. Cancer Res 1995; 55: 4818–23.

11 Koh J, Enders GH, Dynlacht BD, Harlow E. Tumour-derivedp16 alleles encoding proteins defective in cell-cycle inhibition.Nature 1995; 375: 506–10.

12 Reddel RR. Genes involved in the control of cellular proliferativepotential. Ann N Y Acad Sci 1998; 854: 8–19.

13 Sherr CJ. The INK4a/ARF network in tumour suppression. NatRev Mol Cell Biol 2001; 2: 731–7.

14 Geradts J, Kratzke RA, Niehans GA, Lincoln CE. Immunohis-tochemical detection of the cyclin-dependent kinase inhibitor2/multiple tumor suppressor gene 1 (CDKN2 / MTS1) productp16INK4A in archival human solid tumors: Correlation with retin-oblastoma protein expression. Cancer Res 1995; 55: 6006–11.

15 Hu Y-X, Watanabe H, Ohtsubo K et al. Frequent loss of p16expression and its correlation with clinicopathological parame-ters in pancreatic carcinoma. Clin Cancer Res 1997; 3: 1473–7.

16 Maeda K, Chung Y, Kang S et al. Cyclin D1 overexpressionand prognosis in colorectal adenocarcinoma. Oncology 1998;55: 145–51.

17 Naka T, Kobayashi M, Ashida K, Toyota N, Kaneko T, KaibaraN. Aberrant p16 (INK4) expression related to clinical stage andprognosis in patients with pancreatic cancer. Int J Oncol 1998;12: 1111–16.

18 Palmqvist R, Stenling R, Öberg Å, Landberg G. Expression ofcyclin D1 and retinoblastoma protein in colorectal cancer. EurJ Cancer 1998; 34: 1575–81.

19 Liang JT, Chang KJ, Chen JC et al. Hypermethylation of thep16 gene in sporadic T3N0M0 stage colorectal cancers: Associ-ation with DNA replication error and shorter survival. Oncology(Basel) 1999; 57: 149–56.

20 Wiencke JK, Zheng S, Lafuente A et al. Aberrant methylationof p16INK4a in anatomic and gender-specific subtypes of spo-radic colorectal cancer. Cancer Epidemiol Biomarkers Prev1999; 8: 501–6.

21 Omura K, Uno Y, Kawakami K et al. Expression of thymidylatesynthase in human gastric and colorectal adenocarcinomas isupregulated by p16/INK4. Hepatogastroenterology 2000; 47:742–5.

22 Sui L, Dong Y, Ohno M et al. Inverse expression of cdk4 andp16 in epithelial ovarian tumors. Gynecol Oncol 2000; 79: 230–37.

23 Esteller M, Gonzalez S, Risques RA et al. K-ras and p16aberrations confer poor prognosis in human colorectal cancer.J Clin Oncol 2001; 19: 299–304.

24 Sturm L, Petrowsky H, Volz R et al. Analysis of p53/BAX/p16ink4a/CDKN2 in esophageal squamous cell carcinoma: HighBAX and p16ink4a/CDKN2 identifies patients with good prognosis.J Clin Oncol 2001; 19: 2272–81.

25 Yi J, Wang ZW, Cang H et al. p16 gene methylation in colorec-tal cancers associated with Duke’s staging. World J Gastroen-terol 2001; 7: 722–5.

26 Marsh S, McKay JA, Curran S, Murray GI, Cassidy J, McLeodHL. Primary colorectal tumour is not an accurate predictor ofthymidylate synthase in lymph node metastasis. Oncol Rep2001; 9: 231–4.

27 McKay JA, Douglas JJ, Ross VG et al. Analysis of key cell-cyclecheckpoint proteins in colorectal tumours. J Pathol 2002; 196:386–93.

28 Namazie A, Alavi S, Olopade OI et al. Cyclin D1 amplificationand p16 (MTS1/CDK4I) deletion correlate with poor prognosisin head and neck tumors. Laryngoscope 2002; 112: 472–81.

29 Zhao P, Hu Y-C, Talbot IC. Expressing patterns of p16 andCDK4 correlated to prognosis in colorectal carcinoma. World JGastroenterol 2003; 9: 2202–6.

30 Sobin LH, Wittekind Ch, eds. UICC: TNM Classification ofMalignant Tumours, 5th edn. New York: John Wiley, 1997.

31 Japanese Society for Cancer of the Colon and Rectum. Japa-nese Classification of Colorectal Carcinoma, 1st English edn.Tokyo: Kanehara Shuppan, 1997.

32 Fujii M. Using preoperative UFT to predict sensitivity to fluoro-pyrimidines in colorectal cancer. Oncology (Huntingt) 1999; 13:55–60.

33 Kamoshida S, Matsuoka H, Matsuyama A, Shimomura Y,Maruta M, Tsutsumi Y. Reproducible and reliable immunohis-tochemical demonstration of thymidylate synthase in formalin-fixed, paraffin-embedded sections: Application of antigenretrieval in ethylenediaminetetraacetic acid solution. Acta His-tochem Cytochem 2003; 36: 115–18.

34 Matsuura T, Fukuda Y, Fujitaka T, Nishisaka T, Sakatani T, ItoH. Preoperative treatment with tegafur suppositories enhancesapoptosis and reduces the intratumoral microvessel density ofhuman colorectal carcinoma. Cancer 2000; 88: 1007–15.

35 Ohhara M, Esumi M, Kurosu Y. Activation but not inactivationof the MTS1 gene is associated with primary colorectal carci-nomas. Biochem Biophys Res Comm 1996; 226: 791–5.

36 Pollock PM, Pearson JV, Hayward NK. Compilation of somaticmutations of the CDKN2 gene in human cancers: Non-randomdistribution of base substitutions. Genes Chromosomes Cancer1996; 15: 77–88.


37 Trzeciak L, Hennig E, Kolodziejski J, Nowacki M, Ostrowski J.Mutations, methylation and expression of CDKN2a/p16 gene incolorectal cancer and normal colonic mucosa. Cancer Lett2001; 163: 17–23.

38 Shapiro GI, Edwards CD, Ewen ME, Rollins BJ. p16INK4A par-ticipates in a G1 arrest checkpoint in response to DNA damage.Mol Cell Biol 1998; 18: 378–87.

39 Fukuoka K, Adachi J, Nishio K et al. p16INK4 expression isassociated with the increased sensitivity of human non-smallcell lung cancer cells to DNA topoisomerase I inhibitors. Jpn JCancer Res 1997; 88: 1009–16.

40 Schmitt CA, Fridman JS, Yang M et al. A senescence programcontrolled by p53 and p16INK4a contributes to the outcome ofcancer therapy. Cell 2002; 109: 335–46.

41 Mirjoret J-F, Barberi-Heyob M, Merlin J-L et al. Thymidylatesynthase expression and activity: Relation to S-phase param-eters and 5-fluorouracil sensitivity. Br J Cancer 1998; 78: 62–8.

42 Banerjee D, Gorlick R, Liefshitz A et al. Levels of E2F-1 expres-sion are higher in lung metastasis of colon cancer as comparedwith hepatic metastasis and correlate with levels of thymidylatesynthase. Cancer Res 2000; 60: 2365–7.

43 Lukas J, Petersen BO, Holm K, Bartek J, Helin K. Deregulatedexpression of E2F family members induces S-phase entry andovercomes p16INK4a-mediated growth suppression. Mol Cell Biol1996; 16: 1047–57.

44 Chetty R, Subramoney T, Singh J, Harilal P. Retinoblastoma(pRb) protein immunoexpression in colorectal cancer. Anti-cancer Res 1997; 17: 2593–7.

45 Harris CC. Structure and function of the p53 tumor suppressorgene: Clues for rational cancer therapeutic strategies. J NatlCancer Inst 1996; 88: 1442–55.

46 Graziano F, Cascinu S. Prognostic molecular markers for plan-ning adjuvant chemotherapy trials in Dukes’ B colorectal cancerpatients: How much evidence is enough? Ann Oncol 2003; 14:1026–38.

immunohistochemical analysis of thymidylate synthase, p16ink4a, cyclin-dependent kinase 4 and cyclin...

Documents