Cancer Characterization of Phosphoglycerate Kinase-1 ... · glycolytic pathway and is directly involved in CXCL12-CXCR4 signaling. Normal fibroblasts overexpressing PGK1 resembled

Published OnlineFirst January 12, 2010; DOI: 10.1158/0008-5472.CAN-09-2863

Microenvironment and Immunology
Cancer
Research
Characterization of Phosphoglycerate Kinase-1 Expressionof Stromal Cells Derived from Tumor Microenvironmentin Prostate Cancer Progression
Jianhua Wang1,4, Gigi Ying3, Jingchen Wang4, Younghun Jung4, Jian Lu1, Jiang Zhu2,Kenneth J. Pienta3, and Russell S. Taichman4

Abstract

Authors' AInstitute oUniversityInternal MMichigan SOral MedicMichigan

Note: SuppResearch

CorresponShanghai JP.R. Chinajianhuaw20odontics anAnn Arbor,rtaich@umi

doi: 10.115

©2010 Am

www.aacr

Downl

Tumor and stromal interactions in the tumor microenvironment are critical for oncogenesis and cancerprogression. Our understanding of the molecular events by which reactive stromal fibroblasts—myofibroblastor cancer-associated fibroblasts (CAF)—affect the growth and invasion of prostate cancer remains unclear.Laser capture microdissection and cDNA microarray analysis of CAFs in prostate tumors revealed strong up-regulation of phosphoglycerate kinase-1 (PGK1), an ATP-generating glycolytic enzyme that forms part of theglycolytic pathway and is directly involved in CXCL12-CXCR4 signaling. Normal fibroblasts overexpressingPGK1 resembled myofibroblasts in their expression of smooth muscle α-actin, vimentin, and high levels ofCXCL12. These cells also displayed a higher proliferative index and the capability to contribute to prostatetumor cell invasion in vitro, possibly through expression of MMP-2 and MMP-3 and activation of the AKT andERK pathways. Coimplantation of PGK1-overexpressing fibroblasts with prostate tumor cells promoted tumorcell growth in vivo. Collectively, these observations suggest that PGK1 helps support the interactions betweencancer and its microenvironment. Cancer Res; 70(2); 471–80. ©2010 AACR.

Introduction

Prostate cancer (PCa) is a common neoplasm and is re-sponsible for ∼27,000 deaths yearly in American males (1).Neoplastic epithelial cells, which result mainly from the ac-cumulation of somatic mutations (2), coexist in carcinomaswith a highly complex stroma composed of a number of stro-mal cell types and extracellular matrix, both of which con-tribute to the complexity of the tumor microenvironment (3).It is now well-documented that microenvironmental inter-

actions in tumors are crucial in oncogenesis and cancer pro-gression in a number of settings (4), including breast (5, 6),colon (7), lung (8), and prostate (9) cancers. Overall, in-creased numbers of fibroblasts are found in tumor stroma,which commonly have a myofibroblastic phenotype, and

ffiliations: 1Institute of Medical Sciences and 2Shanghaif Hematology, Rui-Jin Hospital, Shanghai Jiao-TongSchool of Medicine, Shanghai, P.R. China; 3Department ofedicine, Division of Hemotology/Oncology, University ofchool of Medicine, and 4Department of Periodontics andine, University of Michigan School of Dentistry, Ann Arbor,

lementary data for this article are available at CancerOnline (http://cancerres.aacrjournals.org/).

ding Author: Jianhua Wang, Institute of Medical Sciences,iao-Tong University School of Medicine, Shanghai 200025,. Phone: 021-63846590-776684; Fax: 021-63842157; E-mail:[email protected] or Russell S. Taichman, Department of Peri-d Oral Medicine, University of Michigan School of Dentistry,MI 48109. Phone: 734-764-9952; Fax: 734-763-5503; E-mail:ch.edu.

8/0008-5472.CAN-09-2863

erican Association for Cancer Research.

journals.org

on May 2, 2020cancerres.aacrjournals.org oaded from

have been described as cancer-activated or cancer-associatedfibroblasts (CAFs) as distinguished from normal fibroblasts(NSFs; 10). The presence of a reactive stroma has been shownto enhance capillary density and deposition of type I collagenand fibrin (10). In the prostate, activation of stromal fibro-blasts has been shown to lead to cancer progression, andhas a prognostic value (11–13).Our previous studies have shown that stromal cell–derived

factor 1 (SDF-1 or CXCL12) and its receptors (CXCR4 andRDC1/CXCR7) play important roles in PCa metastasis (14,15). More recently, Orimo and colleagues showed that CAFsplay a central role in promoting the growth of tumor cellsthrough their ability to secrete CXCL12 (5, 6). In fact, Begleyand colleagues, showed that NSFs derived from older men ex-press more CXCL12 than NSF derived from younger men, sug-gesting that the stromal fibroblasts of older men may be“primed” to support cancer growth (16). However, whetherstromal cells promote or inhibit prostate growth remains un-settled. For example, Degeorges and colleagues, developedstromal cell cultures from normal human prostate and benignprostatic hyperplasia composed of NSFs and myofibroblastsand studied their roles in modulating LNCaP, PC3, andDU145 PCa cell lines (17, 18). The benign prostatic hyperplasiacell line decreased LNCaP growth by 50% andDU145 growth by40%, but did not inhibit the growth of PC3 cells (17, 18). Collec-tively, understanding of the molecular events for which the re-active stromal cells directly affect PCa progression, however,remains unclear.Phosphoglycerate kinase-1 (PGK1) is an ATP-generating

glycolytic enzyme that forms part of the glycolytic pathwayand has been shown to be a downstream target of CXCL12/CXCR4 signaling in PCa cells (19). PGK1 is secreted

471

. © 2010 American Association for Cancer Research.

http://cancerres.aacrjournals.org/

Wang et al.

472


extracellularly by tumors, where it acts as a disulfide reduc-tase that can cleave plasminogen, generating the vascularinhibitor angiostatin (20, 21). Downregulation of PGK1 se-cretion by tumor cells at sites rich in CXCL12 provides amolecular mechanism to generate an angiogenic switchduring metastasis (19).Here, tissues obtained from a rapid autopsy program were

used to identify PGK1 as a molecule expressed by CAFs, andwas able to regulate PCa growth. Importantly, overexpressionof PGK1 and CXCL12 in NSFs facilitated their conversion in-to cells bearing the CAF phenotype. These findings suggestthat CXCL12 and PGK1 may serve as molecular switches thatare able to drive cancer progression from both a stromal en-vironment and cancer cell perspective.

Materials and Methods

Cell lines. The cell lines used in this study were derivedfrom two human cancerous glands obtained from patientswith PCa. The glands were dissociated, and the various celltypes were then separated to obtain CAFs. NSFs were takenfrom a histologically noncancerous region of the prostate atleast 2 cm away from the outer tumor margin. Their identi-ties were then characterized by immunohistochemical stain-ings for CAF markers α-smooth muscle actin (α-SMA) andvimentin (25). The PC3 cells were cultured in growth medi-um, RPMI 1640 with 10% FCS, and antibiotics (Invitrogen,Corp.) at 37°C, 5% CO2.Isolation of NSF and CAFs from normal and PCa tissues.

All tissues were obtained from the radical prostatectomy se-ries at the University of Michigan and from the Rapid Autop-sy Program (22). Stromal nodules of benign prostatichyperplasia and stroma adjacent to PCa foci (n = 2) wereminced and enzymatically dissociated in mammary epithelialgrowth medium (Cambrex) supplemented with 2% bovine se-rum albumin (Fraction V; Fisher Scientific), 10 ng/mL ofcholera toxin, 300 units/mL of collagenase (Invitrogen), and100 units/mL of hyaluronidase (Calbiochem) at 37°C for 18 h.Later, tissue cells were recovered by differential centrifuga-tion to separate the epithelial cells and fibroblasts (23). NSFsin the supernatant were pelleted and centrifuged at 800 rpmfor 10 min followed by two washes with DMEM/F12 medium.The cell pellet was resuspended in DMEM/F12 mediumsupplemented with 5% fetal bovine serum (Invitrogen) and5 μg/mL of insulin, and plated in 25 cm2 tissue culture flasks.The culture was incubated and undisturbed for 2 to 3 d at37°C, 5% CO2.Laser capture microdissection and cDNA microarrays.

Laser capture microdissection (LCM) was performed fromfrozen tissue sections with the SL Microtest device usingCUT software (MMI). Approximately 5,000 to 10,000 cellswere isolated from frozen prostate tissue specimens fromseparate cases of Gleason 3 + 3 tumors. Additional method-ologies are provided in the Supplementary sections.Construction of lentiviral vectors. For small interfering

RNA knockdown of PGK1 or CXCL12, the iLenti siRNA vectorwas purchased from Applied Biological Materials, Inc., anddesigned with convergent PolIII promoters. Five micrograms

Cancer Res; 70(2) January 15, 2010

on May 2, 2020cancerres.aacrjournals.org Downloaded from

of each vector and 2.5 μg of each packaging vector were co-transfected into 293T cells using Lentifectin (Applied Biolog-ical Materials). Supernatants were collected at 36 to 48 hafter purification and were used to infect NSFs and CAFs(NSFsiPGK1, NSFsiCXCL12 and CAFsiPGK1, CAFsiCXCL12). A PGK1lentiviral vector was similarly prepared using a 1.33 kb hu-man cDNA isolated by reverse transcription-PCR from totalRNA extracted from PC3 cells. A 300-bp hCXCL12 cDNA wasrecovered from a pEB6-SDF-1 vector provided by Dr. DaisukeUchida (Second Department of Oral and Maxillofacial Sur-gery, Tokushima University School of Dentistry, Tokushima,Japan; ref. 24) and was used to generate CXCL12 lentiviruses.NSFs or CAFs were infected with the PGK1 or CXCL12 expres-sing vector (NSFPGK1, CAFPGK1, NSFCXCL12, and CAFCXCL12)or empty control vectors (NSFControl or CAFControl). Controlcells were transduced with lenti-β-gal vector or with theempty vectors. All transfected cells were selected in G418(Invitrogen) and cells expressing high GFP levels were iso-lated by fluorescence-activated cell sorting.ELISAs. Antibody sandwich ELISAs were used to evalu-

ate PGK1 and CXCL12 expression in the NSFs or CAFsconditioned medium (CM; R&D Systems) as previously de-scribed (19).Proliferation assays. After 24 h of serum withdrawal, PC3

cells were digested and washed three times in PBS and 1 × 104

cells were incubated with a 1:1 mixture of CM derived fromCAFβ-gal, CAFsiPGK1, CAFsiCXCL12, NSFControl, NSFPGK1, orNSFCXCL12 and fresh growth medium in triplicate 96-wellflat-bottomed tissue culture plates in a total of 0.1 mL. Ad-ditional methodologies are provided in the Supplementarysections.Invasion assays. Cell invasion was examined using a re-

constituted extracellular matrix membrane (BD Biosciences).PC3 cells were placed in the upper chamber (1 × 105 cells/well) in serum-free medium, and CM derived from alteredexpression of CXCL12 and PGK1 in NSFs, CAFs, and the re-spective controls were added to the bottom chambers. Addi-tional methodologies are provided in the Supplementarysections.Western blot analyses. PCa, NSFs, or CAFs were cultured

to confluence, washed, and then serum-starved in RPMI with0.1% bovine serum albumin for 24 h. The cells were lysed inice-cold radioimmunoprecipitation assay buffer. Cell lysateswere clarified by centrifugation at 14,000 rpm for 10 minand protein concentrations were determined (Bio-Rad Labo-ratories). Additional methodologies are provided in the Sup-plementary sections.In vivo assay for the effects of NSFs on tumor develop-

ment. To determine whether PGK1 activated fibroblast-sup-ported PCa progression in vivo, 1 × 104 PC3luc cells alone ormixed 1:1 with either CAFControl, NSFControl, NSFPGK1, orNSFCXCL12 were implanted s.c. into the backs of severe com-bined immunodeficient mice (CB.17.SCID; Taconic) underisofluorane. After 4 wk, bioluminescence imaging was usedto monitor the tumor growth. The mice were injected i.p.with luciferin (100 μL at 40 mg/mL in PBS) before imaging.This dose and route of administration has been shown to beoptimal for rodent studies within 10 to 20 min after luciferin

Cancer Research



Tumor Microenvironment and Prostate Cancer


injection (14). Mice were anesthetized with 1.5% isoflurane/air, and the Xenogen IVIS cryogenically cooled imaging sys-tem was used as described (14). Selected mice were imagedweekly after tumor injection to monitor tumor development.Bioluminescence generated by the luciferin/luciferase reac-tion served as a locator for cancer growth and was usedfor quantification using the LivingImage software on a red(high intensity/cell number) to blue (low intensity/cellnumber) visual scale. A digital grayscale animal image wasacquired followed by acquisition and overlay of a pseudoco-

www.aacrjournals.org


lor image representing the spatial distribution of detectedphoton counts emerging from active luciferase within the an-imal. Signal intensity was quantified as the sum of all de-tected photons within the region of interest during a 1-minluminescent integration time. All animal procedures wereperformed in compliance with institutional ethical require-ments and were approved by the University of MichiganCommittee for the Use and Care of Animals.Immunohistochemistry. For immunostaining with

α-SMA, vimentin, and fibroblast activation protein (FAP;

Figure 1. Staining of PCa tissuemicroarrays for α-SMA shows a CAFtransition. A, tissue microarray wasconstructed from 30 patients torepresent benign, primary, and PCametastasis sites. Three cores(0.6 mm in diameter) were taken fromeach representative tissue block. Thetissue microarray was immunostainedfor α-SMA. Few cells in the stromaof normal prostates stained positive forα-SMA (A1, arrows). In primary PCa,however, virtually all of the stromalcells stained positive for α-SMA(A2, arrows). B, flow chart forexperimental design showing LCM andtwo rounds of T7 amplification.ssDNA, single-stranded cDNA;ds cDNA, double-stranded cDNA.

Table 1. Partial summary of genes upregulated in CAFs vs. NSFs derived from the prostate peripheralzone

Gene symbol
Average differencebetween NSFs vs. CAFs
Overexpressionrank

Gene function

PDLIM3
5.2695 2 α-Actin–associated LIM protein—binds α-SMA THBS1 4.4057 8 Thrombospondin—cell adhesion, inhibits angiogenesis ACTC 4.2650 9 α-SMA—epithelial-to-mesenchymal transition ADAMTS1 4.1438 12 Disintegrin—inhibits angiogenesis PGK1 3.9983 13 Glycolysis—inhibits angiogenesis FN 3.8904 14 Fibronectin—a direct role in promoting angiogenesis COL4A1 3.6211 29 Collagen—inhibits angiogenesis and hemorrhage VIM 3.4635 49 Vimentin—expression marker of epithelial-to-mesenchymal transition
Cancer Res; 70(2) January 15, 2010 473



Wang et al.

474


Abcam), tissue sections were blocked with Sniper for 5min, and incubated overnight at 4°C with 2.8 μg/mL rabbitanti-human α-SMA, vimentin, and FAP diluted in PBS. Thesections were incubated with the appropriate secondary



antibodies for 30 min, followed by processing with a Lin-coln Label 41 Detection System (Biocare Medical). Addi-tional methodologies are provided in the Supplementarysections.

Figure 2. Reciprocal expression of CXCL12 and PGK1 by tumor stromal cells. A, expression of α-SMA is higher in CAFs than NSFs (green). Representativestaining of vimentin and FAP (brown) in cultures of NSFs and CAFs. Hematoxylin was used as a counterstain. Original magnification, ×40 (bar, 50 μm).B, CXCL12 and PGK1 levels in NSFs and CAFs CM. CXCL12 and PGK1 levels (1 × 106 cells/well) were evaluated by ELISA at 24 and 48 h. Columns,mean for triplicate determinations normalized against total protein; bars, SD. *, P < 0.05, significant difference from respective controls (ANOVA; n = 6in groups). C and D, ELISA assays for CXCL12 and PGK1 levels in CM derived from NSFs and CAFs which overexpress or have reduced levels ofCXCL12 and PGK1 (denoted as NSFCXCL12, NSFPGK1/NSFControl, CAFCXCL12, CAFPGK1/CAFControl, CAFsiCXCL12, or CAFsiPGK1/CAFβ-gal). A β-gal sequencewas incorporated into the small interfering RNA as a vector control. Columns, mean for triplicate determinations normalized against total protein;bars, SD. *, P < 0.05, significant difference from respective controls (ANOVA; n = 6 in groups).

Cancer Research





Statistical analyses. Numerical data are expressed asmean ± SD. Statistical differences between the means forthe different groups were evaluated with Instat 4.0 (Graph-Pad software) using one-way ANOVA, with the level of signif-icance set at P < 0.05.

Results

PGK1 expression is upregulated in CAF versus NSF de-rived from the prostate peripheral zone by LCM and cDNAmicroarray assay. To explore the role of tumor stroma in PCadevelopment, immunohistochemical stainings for α-SMA inPCa tissues was performed and compared with normal pros-tate. We found that a few cells in the stroma from the normal



prostate tissues stained positively forα-SMA (Fig. 1A1, arrows).In primary PCa, however, virtually all stromal cells expressedα-SMA (Fig. 1A2, arrows), as has been reported elsewhere (23, 25).α-SMA was next used to stain serial tissue sections and

CAFs were isolated by LCM. cDNA microarrays were per-formed using RNA recovered from CAFs compared with genesexpressed in normal NSFs (Fig. 1B). SAM analysis and averagelinkage hierarchical clustering were performed and the out-put was visualized with TreeView software (26) to comparethe expression profiles of CAFs with NSFs. The expressionof select genes upregulated in stroma during the conversionto cancer is presented in Table 1. Among those genes whoseexpression was most highly upregulated in CAFs versus NSFsderived from the prostate peripheral zone was PGK1.

Figure 3. The role of CXCL12 and PGK1 on NSFs and PCa proliferation and invasion. A, expression of CXCL12 or PGK1 in NSFs induces proliferation. NSFsand CAFs with altered levels of CXCL12 and PGK1 were evaluated for the expression of Ki-67 as an indicator of proliferation. Representative Ki-67 staining(brown, nuclear) is presented with hematoxylin used as a counterstain. Original magnification, ×20; bars, 100 μm. B, expression of CXCL12 or PGK1 inNSFs and CAFs induces PCa cell proliferation. After 24 h of serum starvation, PCa cells (PC3) were washed and 1 × 105 cells were plated. CM derived fromCAFβ-gal, CAFsiPGK1, CAFsiCXCL12, NSFCXCL12, NSFPGK1, and NSFControl cells were overlaid onto the PCa cells. Proliferation was evaluated by XTT stainingover a 5-d period. Points, mean; bars, SEM; *, P < 0.05, significant difference from NSF and CAF controls (ANOVA; n = 5 samples per condition). C, PCainvasion regulated by CXCL12 or PGK1 expressed by NSFs and CAFs. PC3 cells were placed in the top chamber of invasion plates containing a reconstitutedextracellular matrix in serum-free RPMI medium, and CM derived from NSFCXCL12, NSFPGK1, and NSFControl; CAFCXCL12, CAFPGK1, and CAFControl; orCAFsiCXCL12, CAFsiPGK1, and CAFβ-gal cells were added to the lower chambers. Invasion was determined at 48 h by MTT staining. Columns, percentage ofinvasion; bars, SD; *, P < 0.05, significant difference from invasion between alterations of CXCL12 or PGK1 expression versus controls (ANOVA; n = 5).




Wang et al.

476


A facilitating relationship between CXCL12 and PGK1exists in tumor stromal cells. In previous studies, the rela-tionship between CXCL12 and CXCR4 signaling and theexpression of PGK1 in PCa was explored (19). Here, todetermine how CXCL12 and PGK1 expressed by CAFs reg-ulate tumor progression, stromal cells were first character-ized for the expression of the CAF markers α-SMA, FAP,and vimentin. As shown in Fig. 2A, expression of α-SMA,vimentin, and FAP were higher in CAFs than NSFs. Next,the levels of CXCL12 and PGK1 in the CM of CAFs andNSFs were determined. Higher levels of CXCL12 were se-creted by the CAFs relative to the NSFs (Fig. 2B). Likewise,the levels of PGK1 secreted from CAFs were higher thanlevels produced by NSFs (Table 1).To further explore the role of CAF-secreted CXCL12 and

PGK1 in PCa progression, lentiviral vectors were used to over-express or reduce CXCL12 and PGK1 expression in NSFs andCAFs. After clonal selection, individual clones were pooled andassayed by ELISA for CXCL12 and PGK1 expression (Fig. 2D).Interestingly, overexpression of CXCL12 in NSFs and CAFs en-hanced the secretion of PGK1 compared with controls. In con-trast, reduced CXCL12 expression inhibited PGK1 expressionby CAFs. These data suggest that expression of CXCL12 instromal cells directly regulates PGK1 levels (Fig. 2D).To determine if PGK1 alters CXCL12 levels, overexpression

and knockdown of PGK1 in NSFs and CAFs were performed.Overexpression of PGK1 increased the expression of CXCL12compared with controls. In contrast, reducing PGK1 expres-sion decreased the expression of CXCL12 levels by CAFs(Fig. 2D). Together, these data further suggest a direct andfacilitating relationship between CXCL12 and PGK1 levelsin the tumor stroma (Fig. 2D).Previously, we found that expression of PGK1 in tumor

cells was inversely correlated with CXCR4 level (19). To de-termine the relationship between the expression of PGK1and CXCR4 in tumor stromal cells, quantitative reversetranscription-PCR and Western blot assays were performed.Supplementary Fig. S1 shows that reduced CXCR4 levelswere observed in NSFs that overexpressed PGK1 andCXCL12, consistent with the results shown in CAFs. Similarresults were also observed for CXCR7 levels, a secondCXCL12 cognate receptor. These data suggest that expres-sion of PGK1 in tumor stromal cells was also inversely reg-ulated by CXCL12 receptors.CXCL12/PGK1 expression in NSFs stimulates PCa prolif-

eration and invasion. Next, we evaluated whether the ex-pression of CXCL12 or PGK1 in NSFs and CAFs affectstheir phenotype. As shown in Supplementary Fig. S2, overex-pression of CXCL12 and PGK1 in NSFs induced morphologicchanges reminiscent of the CAF cell phenotype. In addition,overexpression of CXCL12 or PGK1 inNSFs resulted in a higherproliferative rate, in which reduced expression of CXCL12or PGK1 produced a reduction in Ki-67 levels (Fig. 3A).To determine if the expression of CXCL12 and PGK1 by

NSFs and CAFs was able to alter PCa growth, CM was col-lected and was coincubated with PC3 cells. PC3 growth rateswere enhanced in the presence of CM derived from NSFPGK1

and NSFCXCL12 at day 5, compared with the CM derived from



NSFControl cells. In keeping with the previous results, PC3growth rates were significantly attenuated in the presenceof CM derived from CAFsiCXCL12 and CAFsiPGK1 cells, com-pared with the CM controls (Fig. 3B).Given that alterations in the expression of PGK1 and

CXCL12 by stromal constituents were able to alter the pro-liferation of PCa cells, it was next determined if similar ef-fects could be observed on the invasive activity of PCa.Using reconstituted extracellular matrices in porous cul-ture chambers, PC3 cells were placed in the upper chamber(1 × 105 cells/well) in serum-free medium, and CM derivedfrom the panel of NSFs and CAFs expressing different levelsof CXCL12, and PGK1 was added to the bottom chambers.Invasion into the matrix was determined after 48 hours ofculture. CM containing high levels of CXCL12 expressed by

Figure 4. Expression of CXCL12 or PGK1 alters the phenotype of NSFs.A, immunofluorescence staining for α-SMA in NSFCXCL12, NSFPGK1,NSFControl, and CAFcontrol cells. Expression of α-SMA (green) washigher in NSFPGK1 and CAFcontrol cells than in NSFControl cells.Original magnification, ×40; bars, 50 μm. B, vimentin staining ofNSFCXCL12, NSFPGK1, NSFControl, and CAFcontrol cells. Representativeimmunohistochemistries for vimentin (brown, nuclear) are presented.Original magnification, ×40; bars, 50 μm.

Cancer Research





both NSFs and CAFs supported the invasion of PC3 as com-pared with the respective controls, consistent with previousreports (refs. 16, 27; Fig. 3C). CM derived from cells engineeredto overexpress PGK1 increased the invasiveness of PC3(Fig. 3C), whereas decreasing PGK1 and CXCL12 expressionreduced the invasive abilities of PC3 compared with controls(Fig. 3C). Moreover, adding a CXCL12-neutralizing antibodyin CM from the NSFPGK1 cells decreased the effects of PC3proliferation relative to an IgG antibody control. Similarresults were also observed in PC3 cells pretreated withCXCR4 inhibitors before the addition of CM from theNSFPGK1 cells. These data further suggest that the CXCL12secreted from NSFPGK1 cells activates CXCR12 receptors inPCa cells, which is associated with proliferation and inva-sion (Supplementary Fig. S3).Expression of PGK1 and CXCL12 by NSF contributes to

expression of CAF phenotype. Given the effects of over-



expression of PGK1 and CXCL12 in NSFs and CAFs on tumorproliferation, we hypothesized that expression of these genescontributes to the expression of a CAF phenotype. Figure 4Aand B show that the expression of α-SMA and vimentin washigher in CAFs than NSFs under basal conditions. Overex-pression of PGK1 and CXCL12 in NSFs induced higher ex-pression of vimentin compared with controls. Similarly,overexpression of PGK1 in NSFs increased a higher level ofα-SMA but not in NSFs overexpressing CXCL12. These datasupport the notion that expression of PGK1 may induce thedifferentiation of NSFs into a stage that resembles CAFs.CXCL12 and PGK1 expression in NSFs promotes PCa

growth in vivo. To determine whether overexpression ofPGK1 in NSFs supports PCa progression in vivo, an equal num-ber of PC3luc cells alone or mixed with either CAFcontrol,NSFPGK1, or NSFCXCL12 cells were implanted s.c. into the backsof nonobese diabetic/severe combined immunodeficient mice.

Figure 5. Overexpression of CXCL12 or PGK1 in NSFs facilitates PCa growth in vivo. A, PC3luc cells were implanted alone or mixed with NSFCXCL12,NSFPGK1, NSFControl, or CAFControl cells at a ratio of 1:1 cells into the backs of nonobese diabetic/severe combined immunodeficient mice. At 4 wk, imagingwas performed with a Xenogen IVIS imaging system. Luciferase signals from the representative mice are shown. B, quantification of tumor burdensidentified by bioluminescence imaging. Increasing tumor size was observed after coinjection of NSFCXCL12 or NSFPGK1 compared with NSFControl cells.Coimplantation of CAFControl cells served as a positive control. *, P < 0.05, significant difference from controls. #, P < 0.05, significant difference frompresence of CAFControl or PC3 alone. C, immunofluorescence staining for CAF markers α-SMA and vimentin in tumor stroma from NSFControl, NSFCXCL12,and NSFPGK1 mixed with PC3luc. Nuclei were identified by 4′,6-diamidino-2-phenylindole. All images were captured on Zeiss LSM 510 meta confocallaser scanning microscope. Original magnification, ×60; bars, 30 μm. D, quantitative evaluation of the expression of PGK1 in human PCa. Left,immunostaining intensity was scored as either absent (1), weak (2), moderate (3), or strong (4). The mean expression scores multiplied by percentage ofpositive cells in the field (Quick's combined score system) for all benign, prostatic intraepithelial neoplasia (PIN), and localized PCa are presented ingraphical format using error bars with 95% confidence intervals. *, P < 0.05, statistically significant differences were noted between benign hyperplasiaand prostatic intraepithelial neoplasia, or localized cancer. Right, ELISA analysis of PGK1 level in serum of normal controls and patients with PCa.Points, mean for triplicate determinations normalized against total protein; bars, SD. *, P < 0.05, significant difference from respective controls(ANOVA; n = 3 in groups).




Wang et al.

478


As shown in Fig. 5A and B, the presence of CAFcontrol clearlysupported tumor growth when compared with PC3luc alone.Surprisingly, when PCa cells were implanted with NSFControl

cells, growth was reduced compared with when the PCa cellswere grown alone. However, when the PCa cells were im-planted with NSFs that overexpressed PGK1 or CXCL12, thegrowth of the tumor was significantly enhanced comparedwith PCa cells grown with NSFControl cells (Fig. 5A and B).At the end of the observation period, the tumors were ex-

cised and immunofluorescence staining for α-SMA, vimentin,and FAP were performed. Greater staining for α-SMA, vimen-tin, and FAP were observed in tumors derived from the PCacells mixed with NSFPGK1 compared with the NSFControl cells.A similar pattern of staining for vimentin was shown in tu-mors derived from the PCa cells grown with NSFCXCL12, butnot with α-SMA (Fig. 5C; Supplementary Fig. S4). These re-sults are consistent with observations shown in Fig. 4A and B,



and further support the notion that PGK1-activated NSFs ex-pressing myofibroblast/CAF markers contribute to tumorgrowth when commingled with PC3 human PCa cells.Primary human tissues were next used to verify the results

of the animal studies. High-density tissue microarrays werestained with antibody to PGK1. Quantitative analysis of the ex-pression of PGK1 shows that the total expression of PGK1 wasfound to be higher in prostatic intraepithelial neoplasia andPCa than in the normal epithelium (Fig. 5D). In addition,PGK1 serum levels were higher in patients with PCa comparedwith age-matched controls (Fig. 5D). These findings suggestthat PGK1 may indeed play a role in PCa progression (20, 28).Molecular mechanisms of PCa progression modulated

by PGK1. Based on these in vitro and in vivo findings, we fur-ther explored the molecular mechanisms by which PGK1 in-duces CAF formation, and subsequently, PCa progression. Thedifferences in mRNA levels of select genes were explored inNSFControl, NSFPGK1, or NSFCXCL12 cells using microarray as-says. Supplementary Table S1 shows that the top seven path-ways enriched in common genes induced by overexpression ofCXCL12 and PGK1 in NSFs were involved in resistance to ap-optosis (e.g., CADM1, MAPK10, and HIPK3), promoted cellproliferation (e.g., CXCL12, ADAMTS1, and E2F1), were in-volved in metastasis [e.g., matrix metalloproteinases (MMP-2, MMP-3, MMP-9, MMP-14), MTA1, and CD44], activatedMAPK and Wnt pathways (e.g., MAP2K7, PPP2CA, andDKK1), or associated with cell-cell adhesion, including VIM,SMARC1, CAV2, and others. These results were partiallyvalidated by focusing on α-SMA, MMP-2 and MMP-3 usingWestern blots. As shown in Fig. 6A, expression of MMP-2and MMP-3 levels were higher in NSFs overexpressing PGK1and CXCL12 compared with controls. Expression of α-SMAwas higher in NSFPGK1 or NSFCXCL12 cells relative to NSFControl

cells (Fig. 6A). Yet, overexpression of PGK1 in CAF did notgreatly alter the expressions of α-SMA, MMP-2, and MMP-3.Finally, the pathways activated in PCa cells themselves

when they interact with stromal cells were explored. PCacells are cultured with CM derived from NSFs overexpressingPGK1 or CXCL12. Here, activation of the AKT pathway inPC3 cells was observed at 10 and 30 minutes compared withCM derived from NSFcontrol (Fig. 6B). Similarly, the ERK path-way was also activated in PCa when the cells were treatedwith CM derived from NSFPGK1 or NSFCXCL12 cells (Fig. 6B).

Discussion

Tumors arise from cells that have sustained genetic muta-tions resulting in disregulation of normal growth control me-chanisms (3). Research focused on the origins of cancer hasidentified that genetic mutations within tumor cells have re-sulted in the concept that neoplastic progression is a cell-au-tonomous process. However, it has become increasinglyapparent that the microenvironment influences many stepsin tumor progression (29). Not only do the cancer cellsinteract with each other, they are also influenced by the ex-tracellular matrix and other microenvironmental cells in-cluding fibroblasts, endothelial cells, and inflammatorycells, etc. (9, 29). Recent work has shown that alterations in

Figure 6. Molecular mechanisms of PCa progression modulated byPGK1. A, CXCL12 or PGK1 altered the expression of α-SMA, MMP-2,and MMP-3 in NSFs. Western blots were used to confirm the changesobserved in selected mRNA levels identified in Supplementary Table S1.Whole cell lysates were immunoblotted with antibody to α-SMA, orantibodies to MMP-2 or MMP-3. Expression of α-SMA, MMP-2, andMMP-3 in CAFs were used as positive controls. The blots were strippedand reprobed with anti–α-tubulin antibody to confirm equal proteinloading. B, evaluation of the effects of expression of CXCL12 or PGK1 inPC3 on ERK or AKT signaling. After 24 h of serum withdrawal, PCacells treated with CM derived from NSFCXCL12, NSFPGK1, and NSFControl

were assayed for ERK or AKT signaling by Western blotting at varioustime points. Whole cell lysates were separated on SDS-PAGE andimmunoblotted with antibodies to p-ERK (p-p44/p42) or p-AKT. Theblots were stripped and reblotted with antibodies against total ERK(p44/p42) and AKT, respectively (top).

Cancer Research





the stromal compartment often accompany, and may evenprecede, the malignant conversion of epithelial cells. Theconversion of normal tissue-resident fibroblasts into CAFsseems to be a critical step in neoplastic progression (10, 30).Here, LCM was performed in conjunction with cDNA mi-

croarray analysis of noncancerous prostate tissues, primaryPCa, as well as PCa metastases to differentiate the contribu-tions of cancer epithelial cells versus supporting stromal cells(Fig. 1A). Our data showed significant alterations in the ex-pression of genes associated with the transition of NSFs toCAFs. In conjunction with the transition of the stroma to aCAF phenotype, there was a significant shift in the expres-sion of a multitude of genes associated with angiogenesisinhibition, including PGK1, and ADAMTS1, which by itselfhas been shown to stimulate PCa growth (ref. 27; Table 1).Additional studies exploring the mechanisms by which

CAFs induce tumor progression and stromal-epithelial inter-actions were modeled in vitro and in vivo. We found that ex-pression of PGK1 in NSFs supported the expression of CAFdifferentiation including the expression of α-SMA, vimentin,FAP, and CXCL12. Furthermore, overexpression of CXCL12or PGK1 in NSFs enhanced proliferation and modulatedPCa invasion in vitro. These data suggest that although theexpression of PGK1 has a role as an inhibitor of angiogenesis(19, 21), it also supports CAF-like features and regulatesCXCL12 secretion. Together, these features may create a mi-croenvironment that supports tumor growth and invasion.We have also shown that expression of PGK1 and CXCL12

in NSFs alters a number of pathways which are associatedwith tumor progression, including a number of the MMPs(e.g., MMP-2, MMP-3, MMP-9, and MMP-14). Moreover,AKT and ERK pathways were activated in PC3 cells whencultured with CM from NSFs overexpressing PGK1. MMPshave been implicated in the promotion of tumor growth, an-giogenesis, invasion, and metastasis, in which elevated MMPlevels were associated with poor prognosis (31–33). In thisstudy, the observation that PGK1 induced MMP-2 andMMP-3 expression in NSFs may provide the activation ofcritical molecules required to promote epithelial-to-mesen-chymal transition, including cleavage of the extracellular do-main of E-cadherin (34). Intriguingly, activation of AKT andERK pathways in PCa by NSFs overexpressing PGK1 indi-cated that stromal cells support tumor growth and invasion.One of the interesting aspects of this work was found

when coimplantation studies were performed. Here, PC3luc

cells alone or mixed 1:1 with either CAFcontrol, NSFcontrol,NSFPGK1, or NSFCXCL12 were injected into the backs of severe



combined immunodeficient mice. The data showed thatcompared with the CAFcontrol cells, NSFPGK1 or NSFCXCL12

cells did not support robust in vivo tumor growth. One pos-sibility is that PGK1-activated fibroblasts may not have com-pleted the transition to CAF. However, in the presence of thefibroblast group, NSFPGK1 or NSFCXCL12 facilitated tumorgrowth to a higher extent than NSFcontrol cells (Fig. 5A andB). Moreover, it was interesting to note that NSFs inhibitedPCa growth when the PCa cells were grown without NSFs(Fig. 5A and B). These data suggest that early NSFs may beprotective, but as they move towards a CAF phenotype, theexpression of PGK1 or SDF-1 in tumor stroma may drive tu-mor development.Previously, we reported that PGK1 was a critical target of

the CXCL12/CXCR4 chemokine axis, and suggested that a“Warburg effect” is likely to be a fundamental property of can-cer cells and was known to correlate with enhanced tumorprogression (19). Here, our observations indicate that PGK1also plays a central role in inducing fibroblasts towards theCAF phenotype in the tumor microenvironment. One possiblerole of PGK1 in stromamay have the ability to increase metab-olism and cell cycling, thus eliciting a CAF phenotype. Thissuggests that a “Warburg effect” may also be associated withtumor stromal cells; however, further studies are needed.

Disclosure of Potential Conflicts of Interest

No potential conflicts of interest were disclosed.

Grant Support

Department of Defense grants PC073952 and PC060857(R.S. Taichman); NIH grant PO1 CA093900 (R.S. Taichmanand K.J. Pienta); American Cancer Society Clinical ResearchProfessorship, NIH Specialized Programs of Research Excel-lence in prostate cancer grant P50 CA69568, and the CancerCenter support grant P30 CA46592 (K.J. Pienta). The work issupported by National Natural funding of China (30973012)and Shanghai Education Committee Key Discipline andSpecialties Foundation Project Number:J50208.The costs of publication of this article were defrayed in

part by the payment of page charges. This article musttherefore be hereby marked advertisement in accordancewith 18 U.S.C. Section 1734 solely to indicate this fact.Received 8/5/09; revised 11/3/09; accepted 11/5/09;

published OnlineFirst 1/12/10.

References
1. Concato J, Jain D, Uchio E, et al. Molecular markers and death from
prostate cancer. Ann Intern Med 2009;150:595–603.2. Hida K, Hida Y, Amin DN, et al. Tumor-associated endothelial cells

with cytogenetic abnormalities. Cancer Res 2004;64:8249–55.3. Bissell MJ, Radisky D. Putting tumours in context. Nat Rev Cancer

2001;1:46–54.4. Micke P, Ostman A. Exploring the tumour environment: cancer-as-

sociated fibroblasts as targets in cancer therapy. Expert Opin TherTargets 2005;9:1217–33.

5. Orimo A, Gupta PB, Sgroi DC, et al. Stromal fibroblasts present ininvasive human breast carcinomas promote tumor growth and an-giogenesis through elevated SDF-1/CXCL12 secretion. Cell 2005;121:335–48.

6. Orimo A, Weinberg RA. Stromal fibroblasts in cancer: a novel tumor-promoting cell type. Cell Cycle 2006;5:1597–601.

7. Tsujino T, Seshimo I, Yamamoto H, et al. Stromal myofibroblastspredict disease recurrence for colorectal cancer. Clin Cancer Res2007;13:2082–90.




Wang et al.

480


8. Micke P, Ostman A. Tumour-stroma interaction: cancer-associatedfibroblasts as novel targets in anti-cancer therapy? Lung Cancer2004;45 Suppl 2:S163–75.

9. De Wever O, Demetter P, Mareel M, et al. Stromal myofibroblastsare drivers of invasive cancer growth. Int J Cancer 2008;123:2229–38.

10. Ostman A, Augsten M. Cancer-associated fibroblasts and tumorgrowth-bystanders turning into key players. Curr Opin Genet Dev2009;19:67–73.

11. Ao M, Franco OE, Park D, et al. Cross-talk between paracrine-actingcytokine and chemokine pathways promotes malignancy in benignhuman prostatic epithelium. Cancer Res 2007;67:4244–53.

12. Yang F, Tuxhorn JA, Ressler SJ, et al. Stromal expression of connec-tive tissue growth factor promotes angiogenesis and prostate cancertumorigenesis. Cancer Res 2005;65:8887–95.

13. Yanagisawa N, Li R, Rowley D, et al. Stromogenic prostatic carcino-ma pattern (carcinomas with reactive stromal grade 3) in needle bi-opsies predicts biochemical recurrence-free survival in patients afterradical prostatectomy. Hum Pathol 2007;38:1611–20.

14. Sun YX, Schneider A, Jung Y, et al. Skeletal localization and neutral-ization of the SDF-1(CXCL12)/CXCR4 axis blocks prostate cancermetastasis and growth in osseous sites in vivo. J Bone Miner Res2005;20:318–29.

15. Wang J, Shiozawa Y, Wang J, et al. The role of CXCR7/RDC1 as achemokine receptor for CXCL12/SDF-1 in prostate cancer. J BiolChem 2008;283:4283–94.

16. Begley L, Monteleon C, Shah RB, et al. CXCL12 overexpression andsecretion by aging fibroblasts enhance human prostate epithelialproliferation in vitro. Aging Cell 2005;4:291–8.

17. Degeorges A, Tatoud R, Fauvel-Lafeve F, et al. Stromal cells fromhuman benign prostate hyperplasia produce a growth-inhibitory fac-tor for LNCaP prostate cancer cells, identified as interleukin-6. Int JCancer 1996;68:207–14.

18. Degeorges A, Wang F, Frierson HF, Jr., et al. Human prostate cancerexpresses the low affinity insulin-like growth factor binding proteinIGFBP-rP1. Cancer Res 1999;59:2787–90.

19. Wang J, Wang J, Dai J, et al. A glycolytic mechanism regulatingan angiogenic switch in prostate cancer. Cancer Res 2007;67:149–59.

20. Zieker D, Konigsrainer I, Traub F, et al. PGK1 a potential marker forperitoneal dissemination in gastric cancer. Cell Physiol Biochem2008;21:429–36.

21. Lay AJ, Jiang XM, Kisker O, et al. Phosphoglycerate kinase acts in



tumour angiogenesis as a disulphide reductase. Nature 2000;408:869–73.

22. Rubin MA, Putzi M, Mucci N, et al. Rapid (“warm”) autopsy study forprocurement of metastatic prostate cancer. Clin Cancer Res 2000;6:1038–45.

23. Mishra PJ, Mishra PJ, Humeniuk R, et al. Carcinoma-associated fi-broblast-like differentiation of human mesenchymal stem cells. Can-cer Res 2008;68:4331–9.

24. Uchida D, Onoue T, Tomizuka Y, et al. Involvement of an autocrinestromal cell derived factor-1/CXCR4 system on the distant metasta-sis of human oral squamous cell carcinoma. Mol Cancer Res 2007;5:685–94.

25. Tuxhorn JA, Ayala GE, Smith MJ, et al. Reactive stroma in humanprostate cancer: induction of myofibroblast phenotype and extracel-lular matrix remodeling. Clin Cancer Res 2002;8:2912–23.

26. Eisen MB, Spellman PT, Brown PO, et al. Cluster analysis and dis-play of genome-wide expression patterns. Proc Natl Acad Sci U S A1998;95:14863–8.

27. Wang J, Wang J, Sun Y, et al. Diverse signaling pathways throughthe SDF-1/CXCR4 chemokine axis in prostate cancer cell lines leadsto altered patterns of cytokine secretion and angiogenesis. Cell Sig-nal 2005;17:1578–92.

28. Hwang TL, Liang Y, Chien KY, et al. Overexpression and elevatedserum levels of phosphoglycerate kinase 1 in pancreatic ductal ad-enocarcinoma. Proteomics 2006;6:2259–72.

29. Wels J, Kaplan RN, Rafii S, et al. Migratory neighbors and distantinvaders: tumor-associated niche cells. Genes Dev 2008;22:559–74.

30. Hill R, Song Y, Cardiff RD, et al. Selective evolution of stromal mes-enchyme with p53 loss in response to epithelial tumorigenesis. Cell2005;123:1001–11.

31. Curran S, Murray GI. Matrix metalloproteinases in tumour invasionand metastasis. J Pathol 1999;189:300–8.

32. de Vicente JC, Lequerica-Fernandez P, Lopez-Arranz JS, et al. Ex-pression of matrix metalloproteinase-9 in high-grade salivary glandcarcinomas is associated with their metastatic potential. Laryngo-scope 2008;118:247–51.

33. Kahari VM, Saarialho-Kere U. Matrix metalloproteinases and their in-hibitors in tumour growth and invasion. Ann Med 1999;31:34–45.

34. Lochter A, Galosy S, Muschler J, et al. Matrix metalloproteinasestromelysin-1 triggers a cascade of molecular alterations thatleads to stable epithelial-to-mesenchymal conversion and a prema-lignant phenotype in mammary epithelial cells. J Cell Biol 1997;139:1861–72.

Cancer Research



2010;70:471-480. Published OnlineFirst January 12, 2010.Cancer Res Jianhua Wang, Gigi Ying, Jingchen Wang, et al. Prostate Cancer ProgressionStromal Cells Derived from Tumor Microenvironment in Characterization of Phosphoglycerate Kinase-1 Expression of

Updated version

10.1158/0008-5472.CAN-09-2863doi:

Access the most recent version of this article at:

Material

Supplementary

http://cancerres.aacrjournals.org/content/suppl/2010/01/12/0008-5472.CAN-09-2863.DC1

Access the most recent supplemental material at:

Cited articles

http://cancerres.aacrjournals.org/content/70/2/471.full#ref-list-1

This article cites 34 articles, 14 of which you can access for free at:

Citing articles

http://cancerres.aacrjournals.org/content/70/2/471.full#related-urls

This article has been cited by 12 HighWire-hosted articles. Access the articles at:

E-mail alerts related to this article or journal.Sign up to receive free email-alerts

Subscriptions

Reprints and

[email protected] at

To order reprints of this article or to subscribe to the journal, contact the AACR Publications

Permissions

Rightslink site. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC)

.http://cancerres.aacrjournals.org/content/70/2/471To request permission to re-use all or part of this article, use this link

on May 2, 2020. © 2010 American Association for Cancer Research. cancerres.aacrjournals.org Downloaded from


http://cancerres.aacrjournals.org/lookup/doi/10.1158/0008-5472.CAN-09-2863

http://cancerres.aacrjournals.org/content/suppl/2010/01/12/0008-5472.CAN-09-2863.DC1

http://cancerres.aacrjournals.org/content/70/2/471.full#ref-list-1

http://cancerres.aacrjournals.org/content/70/2/471.full#related-urls

http://cancerres.aacrjournals.org/cgi/alerts

mailto:[email protected]

http://cancerres.aacrjournals.org/content/70/2/471


Documents

Cancer Characterization of Phosphoglycerate Kinase-1 ... · glycolytic pathway and is directly involved in CXCL12-CXCR4 signaling. Normal fibroblasts overexpressing PGK1 resembled