Biochemical and Biophysical Research Communications 419 (2012) 529–536

Contents lists available at SciVerse ScienceDirect

Biochemical and Biophysical Research Communications

journal homepage: www.elsevier .com/locate /ybbrc

Characterization of cancer stem cell properties of CD24 and CD26-positivehuman malignant mesothelioma cells

Hiroto Yamazaki a, Motohiko Naito a, Farhana Ishrat Ghani a, Nam H. Dang b, Satoshi Iwata a,Chikao Morimoto a,⇑a Division of Clinical Immunology, Institute of Medical Science, University of Tokyo, Tokyo, Japanb Division of Hematology/Oncology, University of Florida Shands Cancer Center, Gainesville, FL 32610, USA

a r t i c l e i n f o

Article history:Received 9 February 2012Available online 17 February 2012

Keywords:CD24CD26Cancer stem cellsMalignant mesothelioma

0006-291X/$ - see front matter � 2012 Elsevier Inc. Adoi:10.1016/j.bbrc.2012.02.054

⇑ Corresponding author. Address: 4-6-1 Shirokane8639, Japan. Fax: +81 3 6409 2098.

E-mail address: morimoto@ims.u-tokyo.ac.jp (C. M

a b s t r a c t

Malignant mesothelioma (MM) is an asbestos-related malignancy characterized by rapid growth andpoor prognosis. In our previous study, we have demonstrated that several cancer stem cell (CSC) markerscorrelated with CSC properties in MM cells. Among these markers, we focused on two: CD24, the com-mon CSC marker, and CD26, the additional CSC marker. We further analyzed the CSC properties ofCD24 and CD26-positve MM cells. We established RNAi-knockdown cells and found that these markerswere significantly correlated with chemoresistance, proliferation, and invasion potentials in vitro. Inter-estingly, while Meso-1 cells expressed both CD24 and CD26, the presence of each of these two markerswas correlated with different CSC property. In addition, downstream signaling of these markers wasexplored by microarray analysis, which revealed that their expressions were correlated with several can-cer-related genes. Furthermore, phosphorylation of ERK by EGF stimulation was significantly affected bythe expression of CD26, but not CD24. These results suggest that CD24 and CD26 differentially regulatethe CSC potentials of MM and could be promising targets for CSC-oriented therapy.

� 2012 Elsevier Inc. All rights reserved.

1. Introduction

Malignant mesothelioma (MM) is an asbestos-related malig-nancy characterized by rapid growth, late metastasis, and poorprognosis [1]. MM consists of three major histological subtypes:epithelioid, sarcomatoid, and mixed types [2]. Unfortunately, theresults of standard therapies (surgery, radiotherapy, and chemo-therapy) remain disappointing. Although combination chemother-apy with cisplatin/pemetrexed (Alimta) has recently emerged [3],the effectiveness is far from satisfactory. Therefore, new therapiesbased on an improved molecular understanding are desired.

Cancer stem cell (CSC) theory is widely accepted at the presenttime [4]. CSC proliferates in an asymmetric cell division-likemanner, exhibits various stem cell signatures, and is believed tobe potential reasons for therapy-resistance. Recent studies haveshown the existence of CSCs not only in hematologic malignancies[5] but also solid tumors [6]. We have also demonstrated theexistence of CSC in lymphoid leukemia [7–10].

In the case of solid tumors, much work has been done to iden-tify the specific cell surface antigen markers of CSCs. CD24 [11,12],CD44, CD90, and CD133 [13] are now considered to be common

ll rights reserved.

dai, Minato-ku, Tokyo 108-

orimoto).

markers in a variety of tumors. Particular antigens are also usedas additional markers within the CSC fraction for further enrich-ment of CSCs. For example, CD26 correlated with the tumorigenicand metastatic capacities within CD133-positive human colorectalCSCs [14].

In our previous report, we found that CD26 was highlyexpressed in surgical specimens of MM, and treatment withhumanized anti-CD26 antibody drastically inhibited tumor growthin mouse xenograft model, suggesting that CD26 is a newtherapeutic target for MM [15].

More recently [16], we identified SP, CD9, CD24, and CD26 asCSC markers of MM that correlated with primary stem cell signa-tures. SP cells of H226 and H2452 cell lines, and CD24-positivecells of JMN and H226 cell lines proliferated in an asymmetric celldivision-like manner. The expression of CD26 closely correlatedwith that of CD24 in sarcomatoid type cell lines. In addition, CD9and CD24-positive cells had higher potential to generate spheroidcolony than negative cells in the stem cell medium. Moreover,these marker-positive cells had a clear tendency to generate largertumors in mouse transplantation assay.

In this study, we focused on CD24 and CD26 in JMN (sarcoma-toid), H226 (epithelioid), and Meso-1 (epithelioid) cell lines. CD24is considered to be one of the common CSC markers in variousmalignant tumors [11–13], and CD26 is also the additional CSCmarker [14]. RNAi-knockdown of these markers revealed that they

530 H. Yamazaki et al. / Biochemical and Biophysical Research Communications 419 (2012) 529–536

significantly correlated with several CSC signatures in vitro. Theexpressions of these markers also correlated with several cancerrelated genes. Of interest was that while Meso-1 cells expressedboth CD24 and CD26, different CSC property was associated withthe expression of different marker. Furthermore, phosphorylationof extracellular signal regulated kinase (ERK) by epidermal growthfactor (EGF) stimulation was significantly regulated by the expres-sion of CD26, but not CD24. These results suggest that a combina-tion of CD24 and CD26 expression can be used to efficiently purifyCSCs of MM, and these markers can also be potential molecular tar-gets for novel therapeutic approaches in the future.

2. Materials and methods

2.1. Cell lines and culture

Details of cell lines and culture had been described previously[16].

2.2. Antibodies

Monoclonal antibodies against CD24, CD26, and CD49e wereobtained from BD-Pharmingen (San Jose, CA). Anti-pan ERK andphospho-ERK1/2 (pT202/pY204) antibodies were obtained fromBD. Anti-EGF receptor, phospho-EGF receptor (pY1068), MEK1,and phospho-MEK1/2 (pS217/221) antibodies were obtained fromCell Signaling Technology (Danvers, MA). Anti-b-actin antibodywas obtained from Sigma–Aldrich (St. Louis, MO).

2.3. Flow cytometry and cell sorting

Flow cytometry and cell sorting assays had been described pre-viously [16].

2.4. shRNA lentiviral transduction

CD24 and CD26 expressions were down-regulated usingshRNA-lentiviral vectors (MISSION; Sigma–Aldrich) as describedpreviously [10]. In the case of Meso-1, CD24-enriched cells andCD26-enriched cells were used for each knockdown. The shRNA-target sequences were as follows:

CD24: ccggcttctgcatctctactcttaactcgagttaagagtagagatgcagaag-tttttgCD26: ccgggactgaagttatactccttaactcgagttaaggagtataacttcagtc-tttttg

2.5. EGF stimulation and western blotting

Cells were seeded on a 24-well plate (5 � 104 cells/well in0.5 ml) and incubated for overnight. After the addition of EGF (final300 ng/ml), cells were washed and suspended directly with 50 llof 2 � SDS–PAGE sample buffer at 3 time points (0, 5, 20 min).The cell lysate was boiled and used for western blotting usingECL-advance system (GE, UK). The experiments were repeated atleast 3 times.

2.6. Cell cycle and growth rate analysis

Cell cycle analysis was performed as described previously [10].For growth rate analysis, shRNA cells were seeded on 6well plates(5 � 104 cells per well) and serial passages were done until day 9–10 prior to the cells reaching confluence. The total cell numberswere counted using Countess Automated Cell Counter (Invitrogen).

2.7. Drug sensitivity assay

For preliminary studies, a total of 13 chemotherapeutic drugs(Alimta, cisplatin, dexamethasone, cyclophosphamide, mitomycinC, daunomycin, doxorubicin, methotrexate, Ara-c, 5-FU, etoposide,5-azacitidine, and vincristine) were employed to test drug sensitiv-ity, as described previously [10]. For FACS analysis, each drug towhich the cells were sensitive was added to its respective 50–70% inhibitory concentration, and cells were incubated in a similarmanner.

2.8. Cell invasion assay

Cell invasiveness was determined by using Matrigel invasionchambers (BD) according to the manufacturer’s instructions.Briefly, cells were seeded into the upper chambers at 5 � 105 perchamber in serum-free RPMI1640. The outer chambers were filledwith RPMI1640 containing 10% FBS. After the cells were incubatedfor 24 h, the non-invading cells were removed by swabbing the toplayer. The membranes containing the invasive cells were stainedwith HE, and the invading cells were counted. The experimentswere done in triplicates.

2.9. Microarray analysis

cDNA was synthesized by Superscript III (Invitrogen) from totalRNA. Microarray analysis was performed using the 3D-GeneHuman Oligo-chip 25 k (TORAY, Japan).

2.10. RT-PCR

Thermal cycling was performed as described previously [10]. Thesequences of the primers were as described below: GAPDH,50-accacagtccatgccatcac-30 and 50-tccaccaccctgttgctgta-30. IGFBP7,50-atctggaatgtcactggtgccc-30 and 50-tgatgctgaagcctgtccttgg-30.IGFBP3, 50-aacttctcctccgagtccaagc-30 and 50-gcccatacttatccacacac-cagc-30. Wnt5A, 50-aggtcaacagccgcttcaactc-30 and 50-tagcagcacca-gtggaacttgc-30. CD127, 50-ggctatgctcaaaatggagacttg-30 and50-agacgacactcaggtcaaaaggag-30.

3. Results

3.1. Comparison of asymmetric cell division potential between CD24+

and CD26+ cells in Meso-1 cell line

In our previous report, CD24+ cells of JMN and H226 exhibitedasymmetric cell division potential [16]. As Meso-1 cells displayedheterogeneous expression of CD24 and CD26, we also examinedthis property in Meso-1 (Fig. 1A). Cells were first sorted intoCD24+ and CD24� populations, and then each cell population wasseparately cultured for additional days prior to being re-analyzedfor CD24 expression. Unexpectedly, CD24+ cells generated bothCD24+ and CD24� cells, while CD24� cells also repopulated consid-erable amount of CD24+ cells, indicating that CSC cannot be puri-fied solely through CD24 targeting.

We also evaluated CD26 as a marker for the Meso-1 cell popu-lations, and we found that each cell subset repopulated both neg-ative and positive populations. However, repeated sorting revealedwhile CD26+ cells repopulated both populations, CD26� cellshardly repopulated CD26+ cells, indicating that CD26+ cells havethe potential for asymmetric cell division.

Double-staining of Meso-1 cells for CD24 and CD26 revealedthat these markers were expressed independently with no signifi-cant correlation to each other (Fig. 1B). When cells were thensorted into 4 fractions and cultured separately, CD24�/CD26�

Fig. 1. (A) Assay of asymmetric cell division in Meso-1 cells. Cells were sorted according to the marker expression and cultured for several additional days. Both CD24+ andCD24� populations generated both negative and positive populations. On the other hand, CD26� cells generated almost only CD26� cells, whereas CD26+ cells repopulatedboth populations. (B) Meso-1 cells were double-stained for CD24/CD26 and each fraction was separated and cultured for 6 days. CD24+/CD26+ population regenerated allfractions most evenly than other populations, whereas CD24�/CD26� population regenerated almost only itself.

H. Yamazaki et al. / Biochemical and Biophysical Research Communications 419 (2012) 529–536 531

population regenerated almost only itself, while CD24+/CD26+

population regenerated all fractions most evenly than otherpopulations. These results indicate that CSC of Meso-1 can bepurified most efficiently by targeting the combination of CD24and CD26.

3.2. Drug resistance of CD24 and CD26-positive cells

As CD24 and CD26 were shown to be important markers for iso-lating CSC of MM, we next conducted studies to confirm the in vitroCSC signatures of the marker-positive cells. For this purpose, wechose the JMN, H226, and Meso-1 cell lines since they expressedboth CD24 and CD26. H226 cells displayed heterogeneous expres-sion of CD24 but homogeneous expression of CD26. JMN cells

expressed both markers heterogeneously, with their expressionbeing correlated to each other. In addition, small subpopulationsof CD24+ cell in JMN and H226 exhibited asymmetric cell divi-sion-like proliferation [16].

First, we examined the relationship between the expression ofthese markers and drug resistance. We incubated the cells with13 different anti-cancer drugs (see Section 2) for 3 days and foundthat these cell lines displayed drug-resistance against some ofthese anti-neoplastic agents. In the case of cisplatin, all cell lineswere cisplatin-resistant. Alimta was effective for JMN andMeso-1, but the expressions of these markers were not affected.On the other hand, daunomycin (JMN and Meso-1) and etoposide(H226 and Meso-1) were effective and significantly affected theproportion of CD24-positive cells as documented by FACS analysis

532 H. Yamazaki et al. / Biochemical and Biophysical Research Communications 419 (2012) 529–536

(Fig. 2A). Although the expression of CD49e (Integrin a5) was notaffected by drug treatment in each experiment, survival of CD24+

cells resulted in a significant increase in their proportion, indicat-ing that CD24+ cells were relatively drug-resistant. The proportionof CD26+ cells was also increased in JMN, possibly due to the cor-relation between CD24 and CD26 expression in this cell line. How-ever, the level of CD26+ cells of Meso-1 did not increase followingtreatment by both drugs, indicating that CD24+ and CD26+ cellshave different drug-sensitivity profiles in Meso-1.

3.3. shRNA knockdown of CD24 and CD26 and invasion potential

To assess the functional significance of CD24 and CD26 in MMcells, their expressions were targeted by small interference RNA

Fig. 2. (A) Effect of anti-cancer drugs on CD24 and CD26 expressions. After 3 days incubatwith daunomycin or etoposide shifted the peaks of CD24+ cells rightward in all cell linesCD24+ cells were relatively more drug-resistant than CD24� cells. The proportions of CD2Meso-1 were not affected by both drugs. CD49e is a positive control marker of MM ceKnockdown of CD24 and CD26 expression by shRNA. Their expressions were significantly

using lentivirus expressing short hairpin RNA (shRNA). Treatmentof CD24 and CD26-shRNA caused more than 90% reductionrelative to non-targeting control-shRNA (Control vector) inMeso-1 (Fig. 2B). In JMN and H226, they were also down-regulatedat similar levels (data not shown).

Using these shRNA cells and cells sorted by the respective mark-ers, we examined the invasion potential of each population. Cellswere sorted by relative expression levels of CD24 or CD26, andapplied in Matrigel invasion chambers. The shRNA cells of Meso-1 were directly used for this assay. As shown in Fig. 3A, we foundthat JMN and H226 cells with higher expressions of CD24 andCD26 exhibited higher invasive activity than the cells with lowerexpressions. Intriguingly, CD26+ cells of Meso-1 exhibited higherinvasion potential than CD26� cells, whereas CD24+ and CD24�

ion of each drug, the proportion of CD24+ and CD26+ cells was measured. Incubation(arrows). This result indicated that the proportion of CD24+ cells was increased and6+ cells of JMN were also significantly increased by daunomycin, but CD26+ cells of

lls. Final concentrations of the drugs are described at the corner of each panel. (B)reduced by shRNA of CD24 and CD26 compared to the non-targeting control vector.

H. Yamazaki et al. / Biochemical and Biophysical Research Communications 419 (2012) 529–536 533

cells exhibited similar invasion potential. This difference was alsoconfirmed in the shRNA cells of CD24 and CD26. These results indi-

Fig. 3. (A) Invasion potential of MM cells. CD24+ and CD26+ cells of JMN, CD24+ cells oMeso-1 also had higher invasion potential, but no difference was observed between CDCont, control vector-shRNA. (B) Cell cycle analysis. CD24+ and CD26+ cells of JMN, CDproportion than negative cells. However, there was no difference between CD26+ and CCD24, shRNA-CD26, and control-shRNA were cultured and cell numbers were counted evsignificantly slower than the control-shRNA cells. In Meso-1, the growth of CD24-shRNAthe CD26-shRNA cells was observed. ⁄Statistically significant.

cate that both markers correlate with invasiveness, but only CD24+

cells of Meso-1 do not contribute to this potential.

f H226 showed significantly higher invasiveness than negative cells. CD26+ cells of24+ and CD24� cells. The same results were also obtained in the shRNA of Meso-1.24+ cells of H226, and CD24+ cells of Meso-1 showed significantly higher G2/M

D26� cells of Meso-1. (C) Growth inhibition by shRNA-knockdown. Cells of shRNA-ery 3–4 days. In JMN and H226, the growth rate of shRNA-CD24 and CD26 cells waswas also significantly slower than the control, whereas no significant difference in

534 H. Yamazaki et al. / Biochemical and Biophysical Research Communications 419 (2012) 529–536

3.4. Growth analysis

We next performed cell cycle analysis using cells sorted byexpression of CD24 and CD26 (Fig. 3B). In the case of CD24,CD24+ cells of all cell lines contained G2/M-phase cells at signifi-cantly higher levels than CD24� cells. CD24+ and CD26+ cells ofJMN, and CD24+ cells of H226 also demonstrated higher G2/M pro-portion than the negative cells, indicating the higher proliferatingpotential of marker-positive cells. In Meso-1, CD24+ cells containedmore abundant G2/M phase cells than CD24� cells, whereas no dif-ference between CD26+ and CD26� cells was observed.

We further analyzed the growth inhibitory effect of shRNA-knockdown (Fig. 3C). Cells of shRNA-CD24, shRNA-CD26, and con-trol-shRNA of each cell line were cultured and cell numbers werecounted every 3–4 days. As expected by the cell cycle analysis,the growth rate of shRNA-CD24 and CD26 cells became signifi-cantly slower than that of control-shRNA cells in JMN and H226.In Meso-1, proliferation of shRNA-CD24 cells was also significantlydelayed, whereas no growth inhibition was observed in shRNA-CD26 cells.

Taken together, these results indicate that both CD24 andCD26 are correlated with the CSC properties of JMN and H226.

Fig. 4. (A) RT-PCR analysis of the CD24/CD26-isolated cells of Meso-1. The expressions ofof Wnt5A and IL7-rceptor were correlated with not only CD24 but also CD26 expressionblot analysis of EGF-R signaling between CD26+ and CD26� cells of Meso-1. After stimulatanalyzed at 3 time points. No significant difference between CD26+ and CD26� cells wasstimulated cells. Phosphorylation of ERK2 was significantly suppressed in the CD26� cellpanels). In the case of CD24, phosphorylation of ERK1/2 was not affected in both experi

However, different CSC property is correlated with differentexpression of these alternative markers in Meso-1, suggesting thata combination of these markers is critical for the isolation of CSC inMM.

3.5. Gene expression analysis

To distinguish signals downstream of these cell surfaceantigens, we performed DNA microarray analysis of CD24+ andCD24� cells of JMN and H226, CD24+CD26+ and CD24�CD26� cellsof Meso-1, and control-shRNA and CD24-shRNA cells of Meso-1.We found that several genes related to cancer development andstem cell signatures were differentially regulated. Among thegenes up-regulated in the positive cells, insulin-like growth factorbinding protein 7 (IGFBP7) was commonly up-regulated in allexperiments (fold change, JMN-CD24+, 2.5x; H226-CD24+, 3.3x;Meso-1 CD24+/CD26+, 27.3x; Meso-1 control/CD24-shRNA, 1.9x).In CD24+CD26+/CD24�CD26� cells and control/CD24� shRNA cellsof Meso-1, IGFBP3 (another member of IGFBP, 7.9x and 5.8x,respectively), a cancer gene Wnt5A (5.2x and 6.3x), and hemato-poietic/lymphoid stem cell antigen CD127 (interleukin 7 receptor,IL7R, 10.5x and 3.0x) were also significantly up-regulated.

IGFBP7 and IGFBP3 were well correlated with that of CD26, rather than CD24. Thoses, and their expressions completely disappeared in CD24�/CD26� cells. (B) Westernion by EGF, phosphorylation of EGF-receptor (p-EGF-R) and MEK1/2 (p-MEK1/2) was

observed at each time point. (C) Phosphorylation of ERK1/2 (p-ERK1/2) in the EGFs at 5 min (rectangles). The same result was observed in the CD26-shRNA cells (leftments (right panels).

H. Yamazaki et al. / Biochemical and Biophysical Research Communications 419 (2012) 529–536 535

The microarray data of these genes were further confirmed byRT-PCR in the CD24/CD26-isolated Meso-1 cells (Fig. 4A). Interest-ingly, the expressions of IGFBP3 and IGFBP7 were well correlatedwith that of CD26, rather than CD24. Those of Wnt5A and IL7Rwere correlated with not only CD24 but also CD26 expressions,and their expressions completely disappeared in the CD24�/CD26� cells. These data suggest that the expressions of CD24 andCD26 correlated with several cancer and stemness genes.

3.6. EGF-receptor signal

Both IGFBP3 and IGFBP7 have been reported to be cancer-re-lated genes in a variety of cancers [17,18]. In addition, IGFBP7inhibited extracellular ERK signaling pathways and activated apop-tosis [19,20]. Moreover, EGF-receptor (EGF-R) signal cascadeplayed a critical role in the generation of MM by asbestos exposure[21]. Therefore, we examined downstream signaling of EGF-R,including mitogen-activated protein kinase (MEK) and ERK.

Preliminary experiments revealed that response to EGF stimula-tion was quite poor in JMN and H226 cells (data not shown) butMeso-1 cells were highly responsive. Following separation ofMeso-1 cells according to CD26-expression, each population wasstimulated by EGF. Phosphorylation of EGF-receptor (p-EGF-R)and MEK1/2 (p-MEK1/2) was analyzed at 3 time points (pre, 5,20 min) by western blotting (Fig. 4B). Although EGF-R and MEKwere clearly phosphorylated immediately, no significant differencebetween CD26+ and CD26� cells was observed at each time point.This result indicates that the signal cascade between EGF-R andMEK is transduced normally.

Next, phosphorylation of ERK1/2 (p-ERK1/2) was examined inthe EGF-stimulated cells (Fig. 4C). Different from the other compo-nents, phosphorylation of ERK2 was significantly suppressed in theCD26� cells at 5 min. The same result was obtained in the CD26-knockdown cells (left panels). In the case of CD24, phosphorylationof ERK1/2 was not affected in both CD24+/�-sorting and knock-down experiments (right panels). These results suggest thatCD26, rather than CD24, is involved in the ERK signaling pathway.Therefore, we conclude that CD24 and CD26 share and regulatevarious cellular functions of CSC, and may give new clues to CSC-oriented therapeutic approaches in MM.

4. Discussion

CSC markers of solid tumors vary depending on the histologicaltypes. Common CSC markers that many types of CSC express havebeen also reported, including SP, CD24, CD44, and CD133. CSC isusually identified and purified by a combination of a main markerand a few additional particular markers.

However, recent studies revealed that these markers could notalways distinguish CSCs even in the same histological types. Theopposite expressions of CSC markers were also observed in differ-ent types of CSC [13]. Although still controversial at the presenttime, these common markers are now considered to be key mark-ers in most cases.

CD24 is known as a GPI-anchored membrane protein, and pre-vious studies have reported that CD24 was a main CSC marker forpancreatic cancer [11] and ovarian cancer [12]. In MM, CD44 wasexpressed homogeneously in all cell lines. Although a recent studyreported that CD133 was detectable in MM by RT-PCR [22], wecould not detect membranous CD133 by FACS in all cell lines[16]. Therefore, CD24 was considered to be the main marker ofMM in many cases.

CD26 is another CSC marker in MM. It is known as dipeptidylpeptidase 4 (DPP4) and is associated with immune regulationand signal transduction [23]. It is also involved in the progression

and recurrence of gastrointestinal stromal tumors [24], which areimportant CSC-mediated signatures. In fact, a recent study showedthat a subpopulation of CD26-positive cells correlated with tumor-igenic and metastatic capacities within human colorectal CD133-positve CSC [14].

In our previous studies, we found that CD26 was highlyexpressed in most of the MM patient specimens [15]. Cell lineanalysis also revealed that it is expressed heterogeneously in somecases [16]. Therefore, CD26 is considered to be an additionalmarker of CSC in MM. Moreover, anti-CD26 monoclonal antibodyinhibited growth of various malignant tumors including MM inmouse xenograft model [15,25,26].

In the current study, we analyzed in more details the correlationbetween CSC properties and associated cell surface markers. Ourprevious paper mainly described the heterogeneous expression ofthese markers in JMN and H226 [16]. Both cell lines expressedCD24 as well as CD26, but these markers were heterogeneously ex-pressed and significantly correlated with each other in JMN. InH226, CD24 was heterogeneously expressed, while CD26 was ex-pressed homogeneously at a relatively high level. Therefore,CD24-positive cells of both cell lines were always CD26-postive,and these markers were correlated with every CSC property inJMN and H226.

Our current study mainly analyzed Meso-1 cells, whose expres-sions of CD24 and CD26 are independent of each other. For thisreason, it is speculated that these markers have distinct correlationwith each CSC property. In fact, CD24 correlated with drug-resis-tance and proliferation potentials, whereas CD26 correlated withasymmetric cell division and invasion potentials. Therefore, pat-terns of marker-combination are also important for a full under-standing of the biological characteristics of CSC.

On the other hand, it is suggested that signals downstream ofthe markers can be new targets for CSC-oriented therapy, sincethey most likely have important roles in determining CSC proper-ties. Gene expression analysis identified the up-regulation ofmany cancer-related genes such as IGFBP3 and IGFBP7. Whilesome of them correlated with either CD24 or CD26, others corre-lated with both of these markers. Although IGFBP7 has beenidentified as a tumor suppressor, a recent paper reported thatits high expression was associated with stem cell features andpoor prognosis in leukemia [27]. In addition, IGFBP7 has been re-ported to be involved in ERK-phosphorylation. Therefore, it isplausible that CD26 is involved in EGF-R signaling cascadethrough IGFBP7.

In fact, ERK-phosphorylation is a key step in the development ofMM by asbestos exposure [21], and the EGF-R inhibitor Gefitinib(Iressa) is now widely used for some type of cancers. Phosphoryla-tion of ERK can therefore be affected by the relative expression ofCD26. Although only weak responses were detected in JMN andH226, Meso-1 cells exhibited prompt phosphorylation of ERK,which correlated with CD26 rather than CD24.

Taken together, the expression patterns of CD24 and CD26 dif-fer depending on the cell types in MM. The CSC signatures and ERKsignaling pathways were also different, depending on the markerexpressions. Consequently, although these markers are promisingtargets for novel CSC-oriented therapies in MM, it is important totake into account the relative expression pattern of the combina-tion of CD24 and CD26 in each clinical case.

Acknowledgments

This work was supported by grants-in aid from the Ministry ofEducation, Science, Sports and Culture, and Ministry of Health,Labor and Welfare, Japan, and by the Program for Promotion ofFundamental Studies in Health Sciences of the National Instituteof Biomedical Innovation (C. Morimoto).

536 H. Yamazaki et al. / Biochemical and Biophysical Research Communications 419 (2012) 529–536

References

[1] R. Ismaril-Khan, L.A. Robinson, C.C.J. Williams, et al., Malignant pleuralmesothelioma, a comprehensive review, Cancer Control 13 (2006) 255–263.

[2] J.M. Corson, Pathology of mesothelioma, Thrac. Surg. Clin. 14 (2004) 447–460.[3] N.J. Vogelzang, J.J. Rusthoven, J. Symanowski, et al., Phase III study of

pemetrexed in combination with cisplatin versus cisplatin alone in patientswith malignant pleural mesothelioma, J. Clin. Oncol. 21 (2003) 2636–2644.

[4] T. Reya, S.J. Morrison, M.F. Clarke, et al., Stem cells, cancer, and cancer stemcells, Nature 414 (2001) 105–111.

[5] J.C.Y. Wang, J.E. Dick, Cancer stem cells: lessons from leukemia, Trends CellBiol. 15 (2005) 494–501.

[6] N.Y. Frank, T. Schatton, M.H. Frank, The therapeutic promise of the cancer stemcell concept, J. Clin. Invest. 120 (2010) 41–50.

[7] H. Kayo, H. Yamazaki, H. Nishida, et al., Stem cell properties and the sidepopulation cell as a target for interferon-a in adult T-cell leukemia/lymphoma,Biochem. Biophys. Res. Commun. 364 (2007) 808–814.

[8] H. Nishida, H. Yamazaki, T. Yamada, et al., CD9 correlates with cancer stem cellpotentials in human B-acute lymphoblastic leukemia cells, Biochem. Biophys.Res. Commun. 382 (2009) 57–62.

[9] H. Yamazaki, H. Nishida, S. Iwata, et al., CD90 and CD110 correlate with cancerstem cell potentials in human T-acute lymphoblastic leukemia cells, Biochem.Biophys. Res. Commun. 383 (2009) 172–177.

[10] H. Yamazaki, C.W. Xu, M. Naito, et al., Regulation of cancer stem cell propertiesby CD9 in human B-acute lymphoblastic leukemia, Biochem. Biophys. Res.Commun. 409 (2011) 14–21.

[11] C. Li, D.G. Heidt, P. Dalerba, Identification of pancreatic cancer stem cells,Cancer Res. 67 (2007) 1030–1037.

[12] M.Q. Gao, Y.P. Choi, S. Kang, et al., CD24+ cells from hierarchically organizedovarian cancer are enriched in cancer stem cells, Oncogene 29 (2010)2672–2680.

[13] J.E. Visvader, G.J. Lindeman, Cancer stem cells in solid tumours: accumulatingevidence and unresolved questions, Nature Rev. Cancer 8 (2008) 755–768.

[14] R. Pang, W.L. Law, A.C.Y. Chu, et al., A subpopulation of CD26+ cancer stem cellswith metastatic capacity in human colorectal cancer, Cell Stem Cell 6 (2010)603–615.

[15] T. Inamoto, T. Yamada, K. Ohnuma, et al., Humanized anti-CD26 monoclonalantibody as a treatment for malignant mesothelioma tumors, Clin Cancer Res.13 (2007) 4191–4200.

[16] F.I. Ghani, H. Yamazaki, S. Iwata, et al., Identification of cancer stem cellmarkers in human malignant mesothelioma cells, Biochem. Biophys. Res.Commun. 404 (2011) 735–742.

[17] T. Benatar, W. Yang, Y. Amemiya, et al., IGFBP-7 reduces breast tumor growthby induction of senescence and apoptosis pathways. Breast Cancer Res. Treat,doi: 10.1007/s10549-011-1816-4.

[18] H.H. Mehta, Q. Gao, C. Galet, et al., IGFBP3 is a metastasis suppressor gene inprostate cancer, Cancer Res. 71 (2011) 5154–5163.

[19] N. Wajapeyee, R.W. Serra, X. Zhu, et al., Oncogeneic BRAF induces senescenceand apoptosis through pathways mediated by the secreted protein IGFBP7,Cell 132 (2008) 363–374.

[20] Y. Amemiya, W. Yang, T. Benatar, et al., Insulin like growth factor bidingprotein-7 reduces growth of human breast cancer cells and xenograftedtumors, Breast Cancer Res. Treat. 126 (2011) 373–384.

[21] M.E. Ramos-Nino, C.R. Timblin, B.T. Mossman, Mesothelial cell transformationrequires increased AP-1 binding activity and ERK-dependent Fra-1 expression,Cancer Res. 62 (2002) 6065–6069.

[22] L. Cortes-Dericks, G.L. Carboni, R.A. Schmid, et al., Putative cancer stem cells inmalignant pleural mesothelioma show resistance to cisplatin and pemetrexed,Int. J. Oncol. 37 (2010) 437–444.

[23] C. Morimoto, S.F. Schlossman, The structure and function of CD26 in the T-cellimmune response, Immunol. Rev. 161 (1998) 55–70.

[24] U. Yamaguchi, R. Nakayama, K. Honda, et al., Distinct gene expression-defined classes of gastrointestinal stromal tumor, J. Clin. Oncol. 26 (2008)4100–4108.

[25] K. Ohnuma, T. Ishii, S. Iwata, et al., G1-S cell cycle arrest provoked in human Tcells by antibody to CD26, Immunology 107 (2002) 325–333.

[26] L. Ho, U. Aytac, L.C. Stephens, et al., In vitro and in vivo antitumor effect of theanti-CD26 monoclonal antibody 1F7 on human CD30+ anaplastic large cell T-cell lymphoma Karpas 299, Clin. Cancer Res. 7 (2001) 2031–2040.

[27] S. Heesch, C. Schlee, M. Neumann, et al., BAALC-associated gene expressionprofiles define IGFBP7 as a novel molecular marker in acute leukemia,Leukemia 24 (2010) 1429–1436.