ORIGINAL PAPER

Loss of NF2/Merlin expression in advanced sporadic colorectalcancer

Tamara Čačev & Gorana Aralica & Božo Lončar &

Sanja Kapitanović

Accepted: 21 November 2013 /Published online: 10 December 2013# International Society for Cellular Oncology 2013

AbstractPurpose NF2 /Merlin was first identified through its associa-tion with neurofibromatosis type 2 (NF2). However, accumu-lating evidence suggests a more general involvement in tu-morigenesis and, in particular, a broader role in tumor sup-pression. The aim of this study was to examine NF2 /Merlininvolvement in sporadic colorectal cancer.Methods This study is the first to examine the role of NF2 /Merlin in sporadic colorectal cancer through LOH analysis attheNF2 locus and mRNA expression analysis via quantitativeRT-PCR of total NF2, NF2 isoform I and II. In addition,Merlin protein expression was assessed by immunohisto-chemistry and Western blotting.Results NF2 LOH was detected in 20.0 % of heterozygouscases and was found to be more frequent in tumors larger than5 cm in diameter (p =0.041) and in tumors with a less differ-entiated phenotype (p =0.027). No differences were observedin total NF2 and NF2 isoform I/isoform II mRNA expressionbetween the tumors and their corresponding normal mucoustissues. NF2 isoform II was the most predominant isoform inall samples analyzed. mRNA expression levels of total NF2and isoforms I and II were significantly lower in poorlydifferentiated tumors (p =0.033, p =0.036 and p =0.044, re-spectively). Weak Merlin immunostaining was more frequentin poorly differentiated tumors (p =0.034) and tumors classi-fied as Dukes’ C (p =0.023). A distinct pattern of Merin

phosphorylation was observed in tumors compared to normalmucous tissues.Conclusion Our data indicate that NF2 /Merlin may serve as apotential target in the management of colorectal cancer.

Keywords NF2 .Merlin .NF2 isoform I and II . LOH .

mRNA expression . Immunohistochemistry .Western blot .

Colorectal cancer

1 Introduction

Extensive research over the past three decades has led to anincreased understanding of the genetic and epigenetic changesunderlying the pathogenesis of colorectal cancer, one of themost common types of cancer and one of the leading causes ofcancer death worldwide [1]. Nevertheless, as our knowledgeincreases, novel roles and functions of known genes emerge indifferent physiological and pathophysiological settings, there-by raising new questions that need to be answered [2].

The NF2 tumor suppressor gene was first identified inassociation with neurofibromatosis type 2, an autosomal dom-inantly inherited disorder characterized by the development ofmultiple benign tumors of the central nervous system, such asschwannomas and meningiomas [3]. Since then, mutations ofthe NF2 gene were also found in renal cell cancer, thyroidcancer, mesothelioma and melanoma, suggesting a more gen-eral tumor suppressive role of NF2 in different cell types[4–6]. The human NF2 gene comprises 17 exons and encodesa protein termed Merlin. Merlin is closely related to themembrane–cytoskeleton associated ERM (ezrin, radixin andmoesin) proteins and is involved in the regulation of severalfundamental biological processes such as contact-dependentinhibition of proliferation, cell to cell communication, receptortrafficking, actin cytoskeleton remodeling and membrane or-ganization, all of which are crucial for tumor initiation andprogression [7]. There are at least 10 known isoforms of

Electronic supplementary material The online version of this article(doi:10.1007/s13402-013-0164-2) contains supplementary material,which is available to authorized users.

T. Čačev (*) : S. KapitanovićDivision of Molecular Medicine, Ruđer Bošković Institute,Bijenička cesta 54, 10000 Zagreb, Croatiae-mail: tcacev@irb.hr

G. Aralica :B. LončarClinical Hospital Dubrava, Avenija Gojka Šuška 6,10000 Zagreb, Croatia

Cell Oncol. (2014) 37:69–77DOI 10.1007/s13402-013-0164-2

human Merlin, the two most common isoforms being I and II.Isoform I represents the longest splice variant of Merlin and isencoded by exons 1–17, with splicing out of exon 16. IsoformII retains exon 16, which results in a premature stop codonand, consequently, a shorter protein [8]. Merlin isoform I canundergo an intramolecular head-to tail association, and evi-dence suggests that the growth inhibitory function ofMerlin is associated only with this closed conformation[9, 10]. The conformational switch between the “closed” and“open” state of Merlin is at least in part regulated by phos-phorylation of its C-terminal serine residue S518 by p21-activated kinases (PAK) that act downstream of the smallGTPases Rac and Cdc42 [11, 12]. The posphorylation statusof Merlin can be regulated through alterations in cell cultureconditions, including cell density, cell to substrate attachment,and the availability of growth factors [13]. At low cell density,phosphorylation of Merlin at S518 weakens self-association,which leads to the “open” conformation and inactivation of itsgrowth-suppressing function. At high cell density, Merlin ishypophosphorylated, self-associated and active in mediatingcontact-dependent inhibition of proliferation [14]. In addition,through a negative feedback loop Merlin inhibits Rac and PAKactivation [15], as well as Rac recruitment to themembrane [16].In contrast to Merlin isoform I, Merlin isoform II is believed tobe incapable of forming the head-to-tail association and, there-fore, to lack tumor suppressive activity [9]. Furthermore, it hasbeen found that tumor-associated missense mutations in theFERM or C-terminal domain may disrupt the closed conforma-tion of Merlin and, concomitantly, may lead to a loss of cellulargrowth suppression [4]. Additionally, recent studies have point-ed at NF2 /Merlin’s involvement in the regulation of the Wntand EGFR/MAPK signaling pathways, both of which arederegulated during colorectal tumorigenesis. It has been sug-gested that Merlin negatively regulates receptor tyrosine kinase(RTK) signaling by either directly controlling the turnover ofRTKs from the cell membrane, or by interfering with the acti-vation of signaling components downstream of RTKs [17–20].

Activation of the Wnt signaling pathway is one of the keyevents in colorectal cancer development. A recent study ofBosco et al. [21] has shown that loss of NF2 expression inmouse embryonic fibroblasts (MEFs) increases the transcrip-tional activity of the nuclear pool of β-catenin in a Rac1-dependent manner. Additional evidence for a role of Merlin inthe regulation of Wnt signaling was presented in a recentstudy by Zhou et al. [22] who proposed a model in which lossof Merlin in primary human tumor cells from NF2 patientsmay lead to increased Wnt/β-catenin signaling.

In summary, there is accumulating evidence on Merlin’scomplex involvement in tumor initiation and progression, aswell as its emerging connection to signaling pathwaysderegulated in colorectal tumorigenesis. Nevertheless, the roleof NF2 /Merlin in tumors not associated with neurofibroma-tosis type II has so far rarely been studied and not been

reported for most common human cancers. Our study is thefirst to examine the role of NF2 /Merlin in colorectal tumori-genesis through the analysis of loss of heterozygosity (LOH)at the NF2 gene locus, mRNA expression analysis of totalNF2 and NF2 isoforms I and II, as well as Merlin proteinexpression in sporadic colorectal cancers.

2 Materials and methods

2.1 Tumor samples and normal controls

Tumor and adjacent normal mucous colon tissues from patientswith sporadic colorectal cancer included in our study wereobtained from the Croatian Tumor Bank [23]. All specimenswere obtained during routine surgery performed in patients (n =185) with colon adenocarcinomas. Fresh samples of resectedcolon carcinomaswere snap frozen in liquid nitrogen and storedin the Human Tumor Bank at −80 °C until use. Before use inthis study, each specimenwas verified by a histopathologist. Allspecimens were examined by routine haematoxylin-eosin stain-ing to determine the proportion of tumor cells present in thesample (≥ 80 %). Control normal DNA was extracted fromhistologically normal colon mucous adjacent (15 cm from thetumor) to the adenocarcinoma. Frozen tissue DNA extractionwas performed using a routine proteinase K digestion andphenol chloroform extraction protocol.

2.2 Polymerase chain reaction (PCR)

For LOH analysis at the NF2 gene locus, three primer setswere used. Sequences of specific oligonucleotides used were:TETF (5′-GAG AAT CGC TTA AAC CTG -3′) and TETR(5′-CCT TAT GCC ATG TTC TTG -3′); CAIVF (5′-CTCCCT CTC AGG CTA AAC CT-3′) and CAIVR (5′-GAAATA TGT GGA GGT TCA GAG -3′); CAVF (5′-GGAGAA AAT TGG AGA AGA ACT T-3′) and CAVR (5′-CAG AAG TGA CCA CTG CAC TA-3′) [24].

Genomic DNA (100 ng) was used as a template in areaction volume of 25 μl containing 5 pmol of each primer,50 mM of each dNTP, and 1 U Taq Gold DNA polymerase(Applied Biosystems, Foster City, California, USA). PCRreactions were carried out in an Applied BiosystemsGeneAmp PCR System 2400 for 30 cycles. Annealing tem-peratures for each primer set were optimized in pilot studiesbefore processing the experimental samples.

2.3 Variable number of tandem repeats (VNTR) and LOHanalysis

Polymorphic marker analysis was performed by non-denaturing polyacrylamide gel electrophoresis. For VNTRanalysis, 5 μl of each PCR product was mixed with 3 μl of

70 T. Čačev et al.

loading buffer and loaded onto a 1 mm thick, 35×30 cm, 10%nondenaturing polyacrylamide gel. Electrophoresis was per-formed in 1xTBE buffer for 16 h at 10 V/cm at room temper-ature. Next, the gels were silver stained. LOH was defined bya visible change in allele:allele ratio in tumors compared tomatching normal tissues. Allelic loss of the NF2 gene wasdefined by a positive LOH score at any of the three sites.

2.4 RNA extraction and reverse transcription

Total RNAwas extracted from 60 pairs of resected colorectalcarcinomas and corresponding normal tissues using Trizolreagent (Invitrogen, Carlsbad, USA). Ten micrograms ofRNA were used for reverse transcription (High-CapacitycDNA Archive Kit, Applied Biosystems, Foster City, USA)according to the manufacturer’s protocol. The number oftumor samples analyzed was dependent on the availability ofsufficient RNA quality for real-time RT-PCR analysis.

2.5 Real-time RT-PCR analysis of NF2 mRNA expression

Real-time RT-PCR analysis of totalNF2 ,NF2 isoform I,NF2isoform II and 18S rRNA (control) was performed using anABI PRISM 7300 SDS (Applied Biosystems, Foster City,USA) and predeveloped TaqMan assay reagents,Hs_00738978 (NF2 ), Hs00966310_m1 (NF2 isoform I),Hs0000966295_m1 (NF2 isoform II) and Hs99999901_s1(18S rRNA ) (Applied Biosystems, Foster City, USA). ThePCR reactions were carried out according to the manufac-turer’s protocol. To compensate for inter-PCR variation nor-malization of the targets (total NF2, NF2 isoform I or NF2isoform II) with an endogenous control (18S rRNA ) wasperformed. The levels of 18S rRNA expression in both tumorand adjacent normal mucous tissues were found to be compa-rable. All reactions were performed in duplicate. Results arepresented as ΔCt value, which is the difference in cyclenumber required to raise the amount of PCR product abovethreshold. ΔCt is, therefore, inversely proportional to theamount of mRNA [25].

2.6 Immunohistochemistry

Immunohistochemistry was performed on formalin-fixed,paraffin-embedded tissues. The expression of Merlin in 73sporadic colorectal cancer samples was analyzed using arabbit polyclonal anti-NF2 antibody (A-19) (Santa Cruz Bio-technology, Santa Cruz, USA). Paraffin-embedded tissue sec-tions of normal human colon mucosa stained with anti-NF2antibody A-19 served as a positive control. Negative controlswere performed through omission of the primary antibody.After deparaffinization in xylene, slides were rehydrated inethanol, washed in dH2O and immersed in 1 mmol/L EDTA,pH=9.0 in a water bath at the required temperature (95–99 °C)

for 20min. The endogenous peroxidase activity was quenchedthrough a 5 min incubation in peroxidase blocking solution(DAKO, Carpenteria, USA). Next, the slides were washed inbuffer (tris-buffered saline solution) and a primary rabbitpolyclonal antibody at a concentration of 200 μg/mL (1:50)was applied for 30 min at room temperature. Slides were thenwashed again in tris-buffered saline solution. The DAKOEnVision™ + System, HRP (DAB) (DAKO, Carpenteria,USA) was used for visualization of a positive reaction accord-ing to the manufacturer’s instructions. The slides were coun-terstained with haematoxylin for 30 s, dehydrated andmounted in Canada balsam. Each slide was evaluated for theentire tumor area. The relative intensity of the immunostainingwas defined, semi quantitatively, as weak staining (1),moderate staining (2) or strong staining (3).

2.7 Western blot analysis

For Western blot analysis, frozen tissues from 15 pairs oftumors and corresponding normal mucous tissues werehomogenized in RIPA lysis buffer containing protease(Complete Mini Protease inhibitor cocktail, Roche,Mannheim, Germany) and phosphatase (PhosSTOP phospha-tase inhibitor cocktail, Roche, Mannheim, Germany) inhibitors.The protein concentration was determined using a Bio-Radprotein assay (Bio-Rad Laboratories, Munchen, Germany)and equal amounts (30 μg) of protein were separated by 10 %SDS-PAGE electrophoresis, including β-actin as loading con-trol. Proteins were then transferred to a nitrocellulose mem-brane (Bio-Rad Laboratories, Hercules, USA) and this mem-brane was blocked with 5 % nonfat milk—Tris-buffered salinewith Tween 20 (TBST) for 1 h at room temperature. Next, themembrane was incubated with rabbit polyclonal anti-NF2 an-tibody A-19 (Santa Cruz Biotechnology, Santa Cruz, USA)overnight at 4 °C. The following day the membrane wasincubated with an anti-rabbit HRP-conjugated secondary anti-body (Santa Cruz Biotechnology, Santa Cruz, USA) for 1 h atroom temperature, and protein bands were visualized by en-hanced chemiluminescence using the Western Lightning PlusECL (Perkin Elmer, Waltham, USA) system.

2.8 Statistical analysis

Correlations betweenNF2 LOH status or NF2 protein expres-sion (immunostaining) and clinicopathological characteristicswere explored from contingency tables. These tables wereanalyzed using the χ2 test or Fisher’s exact test. Correlationswere considered significant when the two tailed p-value wasless than 0.05. Correlations between mRNA expression innormal and tumor tissues were analyzed with the Student’st-test. Correlations between mRNA expression in tumor tis-sues and clinicopathological characteristics were analyzedwith Student’s t-test or with the analysis of variance where

NF2 /Merlin in sporadic colorectal cancer 71

appropriate. A p-value less than 0.05 was considered statisti-cally significant. All evaluations were performed usingGraphPad Prism version 4.03 for Windows, GraphPad Soft-ware, San Diego California USA, www.graphpad.com.

3 Results

3.1 Loss of heterozygosity (LOH) at the NF2 gene locus

DNAs from 185 sporadic colorectal tumors were examinedfor the occurrence of loss of heterozygosity (LOH) at the NF2gene locus using 3 intragenic polymorphic VNTR markers:TET, CAIV and CAV. In this VNTR analysis DNA extractedfrom normal tissues showed either one (homozygous) or two(heterozygous) bands at the specific VNTR site. NF2 LOHwas defined as positive when tumor DNA from heterozygous(informative) patients showed loss of one of these two bands(alleles). At the TET locus, 139 of the 185 patients were foundto be heterozygous (75.1 %) and 23 tumors in these 139 cases(16.5 %) showed LOH. The heterozygosity rate of the CAIVmarker was 55.7 % (103 of the 185 patients were heterozy-gous) and 11 tumors in these 103 cases (10.7 %) showedLOH. At the CAV locus, 128 of the 185 patients were hetero-zygous (69.2 %) and 21 tumors in these 128 cases (16.4 %)showed LOH. Using these three intragenic polymorphicmarkers, the total heterozygosity rate was 91.9 % (170 of the185 samples tested) and in 34 tumors of 170 heterozygouscases (20.0 %) LOH was detected.

3.2 Associations between NF2 LOH and clinicopathologicalcharacteristics

The results of the LOH analyses were correlated to the clini-copathological characteristics of the patients and their tumors(age, sex, tumor size, tumor grade and Dukes’ stage) (Table 1).By doing so, no statistically significant correlations werefound betweenNF2 LOH and either age or sex of the patients.Nevertheless, NF2 LOH was found to be more frequent inmales (25.3 %, 23 out of 91 patients) than in females (13.9 %,11 out of 79 patients) and in patients younger than 70 years(20.6 %, 21 out of 102 patients) as opposed to those older than70 years (18.7 %, 12 out of 64 patients) (Table 1). In addition,NF2 LOH was found to be more frequent in tumors largerthan 5 cm in diameter (18 out of 61 tumors; 29.5%) than in thesmaller ones (15 out of 95 tumors; 15.8%), and this differencewas statistically significant (p =0.041) (Table 1).

NF2 LOHwas found to be less frequent in grade 1 (7 out of55 tumors; 12.7 %) and grade 3 (4 out of 24 tumors; 16.7 %)tumors than in grade 2 tumors (20 out of 61, 32.8 %), and alsothis difference was statistically significant (p =0.027)(Table 1). A positive trend was observed between NF2 LOHand Dukes’ stage of the tumor, i.e., in tumors classified as

Dukes’ A NF2 LOH was detected in 6 out of 34 cases(15.0 %), in tumors classified as Dukes’ B in 10 out of 53cases (18.9%) and in Dukes’C in 18 out of 71 cases (25.4%),but this trend was not statistically significant (Table 1).

3.3 Associations between NF2 mRNA expressionand clinicopathological characteristics

We also analyzed NF2 mRNA expression, as well as itsisoforms I and II, in 60 pairs of colorectal tumors and itscorresponding normal mucous tissues by real-time RT-PCR.The number of tumor samples analyzed was dependent uponthe availability of sufficient RNA with a quality suitable forreal-time RT-PCR analysis. No statistically significant differ-ence was observed in total NF2 mRNA expression, as well asthe expression of its isoforms I and II, between normal mu-cous tissues and tumor tissues (p =0.704, p =0.959 and p =0.412, respectively) (Fig. 1). However, a statistically signifi-cant correlation between total NF2 , isoform I and II mRNAexpression and tumor differentiation was observed (p =0.033,p =0.036 and p =0.044, respectively). The expression of totalNF2 mRNA, as well as its isoforms I and II, was higher inwell- and moderately-differentiated tumors compared to poor-ly differentiated tumors (Fig. 2). No correlation was foundbetween total NF2 , isoform I and II mRNA expression and

Table 1 Clinicopathological characteristics of 170 heterozygous patientswith colorectal cancer stratified by NF2 LOH status

NF2 gene LOH

Characteristics No. ofcases

NegativeNo. (%)

PositiveNo. (%)

p-Value

Age

<70 years 102 81 (79.4) 21 (20.6)

≥ 70 years 64 52 (81.3) 12 (18.7) 0.777

Sex

Male 91 68 (74.7) 23 (25.3)

Female 79 68 (86.1) 11 (13.9) 0.065

Tumor size

≤5 cm 95 80 (84.2) 15 (15.8)

>5 cm 61 43 (70.5) 18 (29.5) 0.041

Histological grade (differentiation)a

Well 55 48 (87.3) 7 (12.7)

Moderate 61 41 (67.2) 20 (32.8)

Poor 24 20 (83.3) 4 (16.7) 0.027

Dukes’ stagea

A 40 34 (85.0) 6 (15.0)

B 53 43 (81.1) 10 (18.9) 0.399

C 71 53 (74.6) 18 (25.4)

a p-value includes all data subcategories

72 T. Čačev et al.

Dukes’ stages of the tumors (p =0.174, p =0.116 and p =0.286, respectively) or tumor size (p =0.795, p =0.658 and p=0.714, respectively) (data not shown). The NF2 isoform IIwas the most predominant isoform in all samples analyzed,and it was of interest to note that, as the overall NF2 mRNAexpression decreased with higher histological grade, the de-crease of the NF2 isoform I was much more pronounced andthat, thus, the balance of the isoforms shifted towards isoformII in tumors with a higher histological grade.

3.4 Associations between Merlin protein expressionand clinicopathological characteristics

The expression of the Merlin protein in sporadic colorectalcancers was analyzed by immunohistochemistry in formalinfixed paraffin embedded tissue sections using an anti-NF2antibody (A-19). The results of the immunohistochemicalanalyses were correlated to the clinicopathological featuresof the patients and their tumors (i.e., age, sex, tumor size,tumor grade and Dukes’ stage) (Table 2). Paraffin-embeddedtissue sections of normal human colon mucosa stained withthe anti-NF2 antibody served as positive control. The

predominant pattern of positive staining for Merlin wascytoplasmatic. All analyzed sporadic colorectal cancersstained positively for Merlin. Eleven (15.0 %) adenocarci-nomas showed weak (1), 31 (42.5 %) showed moderate (2)and another 31 (42.5 %) showed strong (3) immunostainingfor Merlin (Fig. 3). No correlation was found between Merlinimmunostaining and the age or sex of the patients (p =0.771and p =0.403, respectively). In addition, no correlation wasfound between Merlin immunostaining and tumor size (p =0.289) (Table 2). However, a statistically significant correla-tion was found between Merlin positivity and the histologicalgrade of the tumors (p =0.034) (Fig. 4a). A gradual decreasein Merlin positivity from lower to higher tumor grade wasobserved, i.e., strong and moderate Merlin staining was morefrequently observed in well- and moderately-differentiatedtumors, while poorly differentiated tumors showed a higherproportion of samples with a weak Merlin immunostaining.Among the well-differentiated tumors (grade 1) 11 out of 18(61.1 %) showed a strong (3) and 5 out of 18 (27.8 %) showeda moderate (2) Merlin positivity, compared to 2 out of 18(11.1 %) well-differentiated tumors with a weak Merlin im-munostaining. Among the moderately-differentiated tumors

a b c

NF2 mRNA expression

p=0.704

NF2 isoform I mRNA expression

p=0.959

NF2 isoform II mRNA expression

p=0.412

N T0

5

10

15

20

25

Non-Outlier Range

25%-75%

Median

NF

2m

RN

Aex

pre

ssio

n (

Ct)

N T0

5

10

15

20

25

Non-Outlier Range

25%-75%

Median

NF

2is

ofo

rm I

mR

NA

exp

ress

ion

( C

t)

N T0

5

10

15

20

25

Non-Outlier Range

25%-75%

Median

NF

2is

ofo

rm II

mR

NA

exp

ress

ion

( C

t)

Δ

Δ

Δ

Fig. 1 NF2 mRNA expression in sporadic colorectal cancers (real time RT-PCR). a TotalNF2 mRNA expression. b NF2 isoform I mRNA expression.c NF2 isoform II mRNA expression. N , normal colon tissue; T, tumor tissue

1 2 30

5

10

15

20

25

Non-Outlier Range

25%-75%

Median

NF2

iso

form

II m

RN

A e

xpre

ssio

n (

Ct)

c

Histological grade

p=0.044

a

Histological grade

p=0.033

1 2 30

5

10

15

20

25

Non-Outlier Range

25%-75%

Median

NF2

mR

NA

exp

ress

ion

( C

t)

b

Histological grade

1 2 30

5

10

15

20

25

Non-Outlier Range

25%-75%

Median

NF2

iso

form

I m

RN

A e

xpre

ssio

n (

Ct)

p=0.036

Δ

Δ Δ

Fig. 2 NF2 mRNA expression (real time RT-PCR) and histological grade of tumors. a Total NF2 mRNA expression. b NF2 isoform I mRNAexpression. c NF2 isoform II mRNA expression

NF2 /Merlin in sporadic colorectal cancer 73

(grade 2) 16 out of 38 (42.1 %) showed a strong (3) and 19 outof 38 (50.0 %) showed a moderate (2) Merlin positivity,compared to 3 out of 38 (7.9 %) moderately-differentiatedtumors with a weak Merlin immunostaining. In contrast tothese findings, in poorly differentiated tumors a weak Merlinpositivity was found to be present in 6 out of 17 cases(35.3 %), and a moderate positivity in 7 out of 17 cases(41.2 %), while a strong positivity was observed in only23.5 % (4 out of 17) of the tumors analyzed. In addition, astatistically significant correlation was found between Merlinexpression and the Dukes’ stage of the tumors (p =0.023)(Fig. 4b). As tumors became more advanced, a gradual de-crease inMerlin positivity was observed. Strong andmoderateMerlin staining was more frequently observed in Dukes’ Aand B tumors, while Dukes’ C (metastatic) tumors showed ahigher proportion of samples with a weak Merlin immuno-staining. In tumors classified as Dukes’ A, 6 out of 10(60.0 %) showed a strong (3) and 4 out of 10 (40.0 %) showeda moderate (2) Merlin positivity, while none of the samplesshowed a weak Merlin immunostaining. In Dukes’ B tumors17 out of 40 cases (42.5 %) showed a strong (3) and 20 out of40 cases (50.0 %) showed a moderate (2) Merlin positivity,compared to 3 out of 40 (7.5 %) Dukes’ B tumors with a weakMerlin immunostaining. In contrast to these findings, in

metastatic (Dukes’ C) tumors a weak Merlin positivity wasobserved in 8 out of 23 cases (34.8 %), a moderate positivitywas observed in 7 out of 23 cases (30.4 %), and strongpositivity was observed in 34.8 % (8 out of 23) of the casesanalyzed. For this set of patients, also the data on TNMstaging was available, and the same trend was observed, i.e.,

Table 2 Clinicopathological characteristics of 73 patients with colo-rectal cancer stratified by Merlin status

Characteristics Merlin staining a p

1 (%) (n =11) 2 (%) (n =31) 3 (%) (n =31)

Age

<70 years 6 (18.2) 13 (39.4) 14 (42.4)

≥70 years 5 (12.5) 18 (45.0) 17 (42.5) 0.771

Sex

Male 7 (14.6) 18 (37.5) 23 (47.9)

Female 4 (16.0) 13 (52.0) 8 (32.0) 0.403

Tumor size

≤5 cm 3 (8.5) 15 (42.9) 17 (48.6)

>5 cm 8 (21.1) 16 (42.1) 14 (36.8) 0.289

Histological grade (differentiation)b

Well 2 (11.1) 5 (27.8) 11 (61.1)

Moderate 3 (7.9) 19 (50.0) 16 (42.1)

Poor 6 (35.3) 7 (41.2) 4 (23.5) 0.034

Dukes’ stageb

A 0 (0.0) 4 (40.0) 6 (60.0)

B 3 (7.5) 20 (50.0) 17 (43.5)

C 8 (34.8) 7 (30.4) 8 (34.8) 0.023

aweak staining was denoted (1), moderate staining was denoted (2) andstrong staining was denoted (3)b p-value includes all data subcategories

Fig. 3 Merlin expression (immunohistochemistry) in sporadic colorectalcancers. a Normal mucous tissue, positive control for Merlin immuno-histochemical reaction. b Strong positive immunostaining (3) in well-differentiated adenocarcimoma with most cancer cells expressing Merlinin the cytoplasm. c Poorly differentiated tumor with a weak positiveimmunostaing (1) for Merlin. (original magnification ×100)

74 T. Čačev et al.

Merlin expression was lower in higher TNM stages (p=0.025) (Supplementary Fig.1).

3.5 Western blot analysis of Merlin expression in sporadiccolorectal cancers

In 15 selected pairs of tumors and corresponding normalmucous tissues we have additionally examinedMerlin expres-sion by Western blot analysis. Merlin migrates as a doubletnear 70 kDa due to differential phosphorylation. A distinctivepattern was observed when corresponding normal and tumortissues were compared, i.e., in normal tissues the phosphory-lated form of Merlin was found to be the predominant one,while in the tumor tissues the unphosphorylated form wasfound to be more pronounced (Fig. 5)

4 Discussion

Despite extensive research on NF2 /Merlin’s role as a tumorsuppressor, its biological function and the mechanism bywhich its loss contributes to tumorigenesis are not completelyunderstood. Although NF2 was first identified as a geneinvolved in the etiology of neurofibromatosis type 2, thesubsequent discovery of NF2 /Merlin’s coordinating role incoupling cellular adhesion to growth factor receptor responsesvia transmembrane receptor tyrosine kinases points at a moreuniversal role in tumor development and progression [26].Indeed, heterozygous NF2-mutant mice spontaneously devel-op a range of highly invasive and metastatic tumors, includingosteosarcomas, lymphomas, lung adenocarcinomas, hepato-cellular carcinomas and fibrosarcomas [27]. More so, in highgrade human gliomas NF2 loss results in an increase in Wntsignaling, specifically TCF transcription factor activation,suggesting a possible link between Merlin and the Wnt sig-naling pathway, one of the key pathways deregulated in colo-rectal cancer [28]. Therefore, the aim of this study was toinvestigate a possible involvement of NF2 /Merlin in colorec-tal cancer tumorigenesis through loss of heterozygosity(LOH) analysis at the NF2 gene locus and analysis of NF2mRNA and protein expression in sporadic colorectal tumors.

The NF2 gene is located on chromosome 22q12 [3, 29]and several studies have pointed at the long arm of the chro-mosome 22, more specifically 22q12.2–12.3, as a region thatis frequently lost in sporadic colorectal cancers [30, 31]. In ourcurrent study 3 intragenic polymorphic markers were used forNF2 LOH analysis. In total, heterozygosity was observed in91.9 % of the patients, and in 20.0 % of these patients LOHwas detected in the corresponding tumor samples. Our resultsare in agreement with those of Rustgi et al. [30] in which anallelic deletion of 22q was found in approximately 20–30% ofthe colorectal tumors tested, and those of Zhou et al. [31] whofound that the average LOH frequency on chromosome 22q insporadic colorectal cancer was 28.38 %. In our study LOHwas found to be more frequent in tumors larger than 5 cm indiameter and in less differentiated tumors, which is in agree-ment with the proposed role of Merlin in tumor differentiationand progression. We have also analyzed total NF2 mRNAexpression, as well as that of its isoforms I and II, in tumorsand corresponding normal mucous tissues. No difference inexpression was observed. Additionally, when NF2 mRNAexpression was analyzed in relation to the clinicopathologicalfeatures of the patients and their tumors, no significant differ-ences or correlations were found between total NF2 , isoform Iand isoform II mRNA expression and tumor size or Dukes’stage. However, a statistically significant correlation betweentotal NF2 , isoform I and isoform II mRNA expression andtumor differentiation was observed. The expression levels oftotal NF2 mRNA, as well as its isoforms I and II, were foundto be higher in well- and moderately-differentiated tumors

a

b

p =0.034

p =0.023

Fig. 4 Merlin expression and clinicopathological characteristics of spo-radic colorectal cancers. a Merlin expression and histological grade oftumors. b Merlin expression and Dukes’ stage of tumors. 1 weak staining;2 moderate staining; 3 , strong staining

N N N

merlin

T TT

-actin

70 kDa

47 kDa

Fig. 5 Western blot analysis of Merlin expression in sporadic colorectalcancers. Merlin migrates as a doublet of ~70 kDa due to differentialphosphorylation. The slower migrating (upper) band is the phosphory-lated form, while the faster migrating (lower) band is the hypo/nonphosphorylated form, N , normal colon tissue; T, tumor tissue

NF2 /Merlin in sporadic colorectal cancer 75

compared to poorly differentiated tumors. In proliferatingepithelial tissues, morphogenic processes involving the estab-lishment or remodeling of cell-cell contacts are coordinated byproliferation and differentiation signals. In addition, it hasbeen found that the establishment of cell polarity is crucialfor the differentiation and morphogenesis of many tissues and,conversely, that loss of polarity is one of hallmarks of tumorcells [32]. Since Merlin has been implicated in the regulationof mechanisms determining cell shape, polarity, proliferationand survival [7], it is not surprising that we find that itsexpression is lower in poorly differentiated tumors.

Alternative splicing is an important regulatory mechanismof gene expression and it is often deregulated in humandiseases, including cancer [33]. Also the human NF2 gene issubject to alternative splicing [8]. Merlin isoform I, as op-posed to isoform II, can form a closed protein conformation,which is essential for Merlin’s growth inhibitory function and,therefore, for its tumor suppressive activity [9, 10]. Throughthe analysis ofNF2 isoform I and II ratios in tumor tissues andtheir correlation with clinicopathological features of the tu-mors, we found that isoform II was the predominantlyexpressed isoform in all samples analyzed. In addition, asthe overall NF2 mRNA expression level decreased withhigher histological grades, the decrease in NF2 isoform Ibecame more pronounced, resulting in a shift of the ratiotowards isoform II in tumors with higher histological grades.These findings are in line with the previously proposed tumorsuppressor function of NF2 isoform II and its loss of activityin poorly differentiated tumors.

Finally, we have analyzed the expression of Merlin proteinin sporadic colorectal cancers by immunohistochemistry informalin fixed paraffin embedded tissue sections. All ana-lyzed sporadic colorectal cancers stained positive for Merlin.No correlation was found between Merlin immunostainingand the age or sex of the patients, nor the size of their tumors.Nevertheless, a statistically significant correlation was foundbetween Merlin positivity and the histological grade of thetumors. Overall, a gradual decrease in Merlin positivity fromlower to higher tumor grade was observed, i.e., strong andmoderate Merlin staining was more frequently observed inwell- and moderately-differentiated tumors, while poorly dif-ferentiated tumors showed a higher proportion of sampleswith a weak Merlin immunostaining. These findings are inaccordance with the results of our NF2 LOH and mRNAexpression analyses, indicating a general decrease in NF2 /Merlin level as the tumor becomes less differentiated.

Normal cells cease to proliferate upon establishing cell-cellcontacts, and loss of this contact-mediated inhibition ofgrowth enhances the ability of cancer cells to invade hosttissues and, ultimately, to metastasize [26]. Several lines ofevidence suggest that loss of Merlin expression in tumor cellsleads to adherens junction destabilization, loss of contactinhibition and an increase in cell proliferation, motility and

invasion [4]. In agreement with these findings, we found astatistically significant correlation between Merlin expressionand the Dukes’ stage of the tumors, i.e., as tumors becamemore advanced a gradual decrease in Merlin positivity wasobserved. Strong and moderate Merlin staining was morefrequently observed in Dukes’ A and B tumors, while Dukes’C (metastatic) tumors showed a higher proportion of sampleswith a weak Merlin immunostaining. In addition to the im-munohistochemical analyses, we have also examined Merlinexpression by Western blotting in tumors and correspondingnormal mucous tissues. By doing so, we found that in normalmucous colon tissue the dominantly expressed form of Merlinwas the phosphorlyated less active form, while in tumortissues the hypo/nonphosphorylated active forms were morepronounced. As it has previously been shown by Shaw et al.[14] that high cell density, serum starvation, or loss of adhe-sion result in dephosphorylation of Merlin, our results maysuggest a compensatory mechanism. Since the activity and theamount of different Merlin conformations may vary depend-ing on the model used, as well as on the specific experimentalconditions, further studies exploring this issue are clearlywarranted.

We did not examine the mutation status of the NF2 gene inthe tumors tested and, therefore, in case truncating mutations arepresent, no firm conclusions can be drawn based on the immu-nohistochemical and qRT-PCR analyses carried out. Accordingto data presented in the literature, however, these mutationsappear to be rare. For instance in the study by Arakawa et al.[34] and Rustigi et al. [30] mutations were detected in 2/44 and2/24 colorectal carcinomas examined, respectively, and in thestudies reported by Sugai et al. [35] and Yoo et al. [36] no NF2mutations were detected in 45 analyzed tumors.

In summary, our results show a negative correlation be-tween NF2 /Merlin expression and the histological grades ofthe colorectal tumors tested. This expression may, therefore,serve as a potential prognostic factor in colorectal cancer. Inaddition, NF2 LOH was observed more frequently in largerand less differentiated tumors. Finally, the observed decrease inMerlin expression in invasive and metastatic colorectal tumorssupports a more general tumor suppressor role, next to itspreviously described suppressor role in NF2-associated tu-mors. Taken together, our findings suggest a previously un-identified role ofNF2 /Merlin in colorectal cancer developmentThis putative role warrants further studies on larger sets ofsamples in order to establish its ultimate clinical significance.

Disclosure/conflict of interest All experiments were performed in fullagreement with ethics, Croatian laws and international conventions. Thiswork was supported by grant number 098-0982464-2508, from theMinistry of Science and Technology, Republic of Croatia. We confirmthat we are the sole authors of the contribution which is our original work.It has not been previously published in whole or substantial part and thereis no conflict of interest present.

76 T. Čačev et al.

References

1. A. Jemal, F. Bray, M.M. Center, J. Ferlay, E. Ward, D. Forman,Global cancer statistics. CA Cancer J. Clin. 61, 69–90 (2011)

2. E.R. Fearon, Molecular genetics of colorectal cancer. Annu. Rev.Pathol. 6 , 479–507 (2011)

3. J.A. Trofatter, M.M. MacCollin, J.L. Rutter, J.R. Murrell, M.P.Duyao, D.M. Parry, R. Eldridge, N. Kley, A.G. Menon, K. Pulaski,A novel moesin-, ezrin-, radixin-like gene is a candidate for theneurofibromatosis 2 tumor suppressor. Cell 72, 791–800 (1993)

4. T. Okada, L. You, F.G. Giancotti, Shedding light on Merlin’s wiz-ardry. Trends Cell Biol. 17 , 222–229 (2007)

5. Z.S. Morris, A.I. McClatchey, Aberrant epithelial morphology andpersistent epidermal growth factor receptor signaling signaling in amouse model of renal carcinoma. Proc. Natl. Acad. Sci. 106 , 9767–9772 (2009)

6. H. Yu, Y. Zhang, L. Ye, W.G. Jiang, The FERM family proteins incancer invasion and metastasis. Front. Biosci. 16, 1536–1550 (2011)

7. I. Stamenkovic, Q. Yu, Merlin, a magic linker between the extracel-lular cues and intracelular signaling pathways that regulate cellmotility, proliferation and survival. Curr. Protein Pept. Sci. 11 ,471–484 (2010)

8. A.B. Bianchi, T. Hara, V. Ramesh, J. Gao, A.J. Klein-Szanto, F.Morin, A.G. Menon, J.A. Trofatter, J.F. Gusella, B.R. Seizinger, N.Kley, Mutations in transcript isoforms of the neurofibromatosis 2gene in multiple human tumour types. Nat. Genet. 6 , 185–192 (1994)

9. L. Sherman, H.M. Xu, R.T. Geist, S. Saporito-Irwin, N. Howells, H.Ponta, P. Herrlich, D.H. Gutmann, Interdomain binding mediatestumor growth suppression by the NF2 gene product. Oncogene 15 ,2505–2509 (1997)

10. D.H. Gutmann, C.A. Haipek, L.K. Hoang, Neurofibromatosis 2tumor suppressor protein, merlin, forms two functionally importantintramolecular associations. J. Neurosci. Res. 58 , 706–716 (1999)

11. G.H. Xiao, A. Beeser, J. Chernoff, J.R. Testa, p21-activated kinaselinks Rac/Cdc42 signaling to merlin. J. Biol. Chem. 277 , 883–886(2002)

12. K. Alfthan, L. Heiska, M. Gronholm, G.H. Renkema, O. Carpen,Cyclic AMP dependent protein kinase phosphorylates Merlin atserine 518 independently of p21-activated kinase and promotesMerlin-Ezrin heterodimerization. J. Biol. Chem. 279 , 1859–1866(2004)

13. A.I. McClatchey, Merlin and ERM proteins: unappreciated roles incancer development? Nat. Rev. Cancer 3 , 877–884 (2003)

14. R.J. Shaw, A.I. McClatchey, T. Jacks, Regulation of the neurofibro-matosis type 2 tumor suppressor protein, merlin, by adhesion andgrowth arrest stimuli. J. Biol. Chem. 273 , 7757–7764 (1998)

15. R.J. Shaw, J.G. Paez, M. Curto, A. Yaktine, W.M. Pruitt, I. Saotome,J. O’Bryan, V. Gupta, N. Ratner, C.J. Der, T. Jacks, A.I. McClatchey,The NF2 tumor suppressor, merlin, functions in Rac-dependent sig-naling. Dev. Cell 1 , 63–72 (2001)

16. T. Okada, M. Lopez-Lago, F.G. Giancotti, Merlin/NF2 mediatescontact inhibition of growh by suppressing recruitment of Rac tothe plasma membrane. J. Cell Biol. 171 , 361–371 (2005)

17. D. Lallemand, J. Manent, A. Couvelard, A. Watilliaux, M. Siena, F.Chareyre, A. Lampin, M. Niwa-Kawakita, M. Kalamarides, M.Giovannini, Merlin regulates transmembrane receptor accumulationand signaling at the plasma membrane in primary mouse Schwanncells and in human schwannomas. Oncogene 28, 854–865 (2009)

18. M. Curto, A.I. McClatchey, Nf2/Merlin: A coordinator of receptorsignalling and intercellular contact. Br. J. Cancer 98, 256–262 (2008)

19. M. Curto, B.K. Cole, D. Lallemand, C.H. Liu, A.I. McClatchey,Contact-dependent inhibition of EGFR signaling by Nf2/Merlin. J.Cell Biol. 177 , 893–903 (2007)

20. A.I. McClatchey, R.G. Fehon, Merlin and the ERM proteins: regula-tors of receptor distribution and signaling at the cell cortex. TrendsCell Biol. 19, 198–206 (2009)

21. E.E. Bosco, Y. Nakai, E.F. Hennigan, N. Ratner, Y. Zheng, NF2-deficient cells depend on the Rac1-canonical Wnt signaling pathwayto promote the loss of contact inhibition of proliferation. Oncogene29 , 2540–2549 (2010)

22. L. Zhou, E. Ercolano, S. Ammoun, M.C. Schmid, M.A. Barczyk,C.O. Hanemann, Merlin-deficient human tumors show loss of con-tact inhibition and activation of Wnt/β-Catenin signaling linked tothe PDGFR/Src and Rac/PAK pathways. Neoplasia 13, 1101–1112(2011)

23. R. Spaventi, L. Pecur, K. Pavelic, Z.P. Pavelic, S. Spaventi, P.J.Stambrook, Human tumour bank in Croatia: a possible model for asmall bank as a part of future European tumour bank network. Eur. J.Cancer 30A , 419 (1994)

24. P. Legoix, M.-F. Legrand, E. Ollagnon, G. Lenoir, G. Thomas, J.Zucman-Rossi, Characterisation of 16 polymorphic markers in theNF2 gene: application to hemizygosity detection. Hum. Mutat. 13 ,290–293 (1999)

25. T. Cacev, S. Radosevic, R. Spaventi, K. Pavelic, S. Kapitanovic, NF1gene loss of heterozygosity and expression analysis in sporadic coloncancer. Gut 54, 1129–1135 (2005)

26. D. Hanahan, R.A. Weinberg, Hallmarks of cancer: The next genera-tion. Cell 144 , 646–674 (2011)

27. A.I. McClatchey, I. Saotome, K. Mercer, D. Crowley, J.F. Gusella,R.T. Bronson, T. Jacks, Mice heterozygous for a mutation at the Nf2tumor suppressor locus develop a range of highly metastatic tumors.Genes Dev. 12, 1121–1133 (1998)

28. Y.K. Lau, L.B. Murray, S.S. Houshmandi, Y. Xu, D.H. Gutmann, Q.Yu, Merlin is a potent inhibitor of glioma growth. Cancer Res. 68 ,5733–5742 (2008)

29. G.A. Rouleau, P. Merel, M. Lutchman, M. Sanson, J. Zucman, C.Marineau, K. Hoang-Xuan, S. Demczuk, C. Desmaze, B. Plougastel,Alteration in a new gene encoding a putative membrane-organizingprotein causes neurofibromatosis type 2. Nature 363 , 515–521(1993)

30. A.K. Rustgi, L. Xu, D. Pinney, C. Sterner, R. Beauchamp, S.Schmidt, J.F. Gusella, V. Ramesh, Neurofibromatosis 2 gene inhuman colorectal cancer. Cancer Genet. Cytogenet. 84 , 24–26(1995)

31. C.Z. Zhou, Z.H. Peng, F. Zhang, G.Q. Qiu, L. He, Loss of heterozy-gosity on long arm of chromosome 22 in sporadic colorectal carci-noma. World J. Gastroenterol. 8 , 668–673 (2000)

32. D. Hanahan, R.A. Weinberg, The hallmarks of cancer. Cell 100 , 57–70 (2000)

33. K. Miura, W. Fujibuchi, I. Sasaki, Alternative pre-mRNA splicing indigestive tract malignancy. Cancer Sci. 102 , 309–316 (2011)

34. H. Arakawa, N. Hayashi, H. Nagase, M. Ogawa, Y. Nakamura,Alternative splicing of the NF2 gene and its mutation analysis ofbreast and colorectal cancers. Hum. Mol. Genet. 3 , 565–568 (1994)

35. T. Sugai, W. Habano, S. Nakamura, N. Uesugi, S. Sasou, C. Itoh, Aunique method for mutation analysis of tumor suppressor genes incolorectal carcinomas using a crypt isolation technique. Arch. Pathol.Lab. Med. 124 , 382–386 (2000)

36. N.J. Yoo, S.W. Park, S.H. Lee, Mutational analysis of tumour sup-pressor gene NF2 in common solid cancers and acute leukaemias.Pathology 44, 29–32 (2012)

NF2 /Merlin in sporadic colorectal cancer 77