Receptor for hyaluronan-mediated motility isoform B ...Several RIP-Tag; RIP-tva mice infected with RCASBP– ... after introduction into the general circulation. For this experi-ment,

Receptor for hyaluronan-mediated motility isoform Bpromotes liver metastasis in a mouse model of multisteptumorigenesis and a tail vein assay for metastasisYi-Chieh Nancy Dua,b,1, Chen-Kung Chouc, David S. Klimstrad, and Harold Varmusa,e,1

aProgram in Cancer Biology and Genetics, Memorial Sloan-Kettering Cancer Center, New York, NY 10065; bDepartment of Pathology and LaboratoryMedicine, Weill Cornell Medical College, New York, NY 10065; cDepartment of Biomedical Sciences, Chang-Gung University, Tao-Yuan 333, Taiwan;dDepartment of Pathology, Memorial Sloan-Kettering Cancer Center, New York, NY 10065; and eCancer Biology and Genetics Section, Cancer GeneticsBranch, National Human Genome Research Institute, Bethesda, MD 20892

Contributed by Harold Varmus, August 30, 2011 (sent for review July 14, 2011)

The gene encoding the receptor for hyaluronan-mediated motility(RHAMM) is overexpressed in many human cancers. However, itis unclear whether RHAMM plays a causal role in tumor initiationor progression. Using somatic gene transfer in a mouse model ofislet cell tumorigenesis, we demonstrate that RHAMM isoform B(RHAMMB) promotes tumor growth and metastases to lymphnodes and the liver. The propensity of RHAMMB-expressing cellsto metastasize to the liver was confirmed using an experimentalmetastasis assay in which cells were injected into the tail vein ofimmunodeficient mice. However, RHAMMB did not increase cellmigration or proliferation in culture. In initial efforts to identifysignaling pathways activated by RHAMMB, we found thatRHAMMB induced phosphorylation of epidermal growth factor re-ceptor (EGFR), Erk1/2, and STAT3 and conferred susceptibility toapoptosis after treatment with an EGFR inhibitor, gefitinib. Takentogether, the results indicate that RHAMMB promotes hepaticmetastasis by islet tumor cells, perhaps through growth factorreceptor-mediated signaling.

We have previously reported a bitransgenic mouse model,RIP-Tag; RIP-tva, in which the rat insulin promoter (RIP)

drives production of both the SV40 T antigen (RIP-Tag) andthe receptor for subgroup A avian leukosis virus (RIP-tva) inpancreatic β cells (1). Coding domains of genes suspected ofcontributing to tumor progression can be introduced into pre-malignant lesions by infection with the avian retroviral vector,RCASBP, after intracardiac injection. RIP-Tag transgenic micedevelop islet tumors through well-defined stages that resemblethe progression of several kinds of human cancers; for this rea-son, we and others have used these mice, with or without addi-tional transgenes, to identify and validate mechanisms oftumorigenesis that may operate in multiple tissues. For instance,we have used RIP-Tag; RIP-tva mice to show that RCASBP-mediated delivery of Bcl-xL or E-cadherin, factors implicated invarious neoplasms, promotes tumorigenesis and invasion in isletcells (1).High-throughput genomic technologies have identified many

genes that may be critical in tumor initiation and progression.However, it remains difficult to distinguish causative and pas-senger mutations and to assign specific biological functions toaltered genes in human cancers, and the RIP-Tag; RIP-tva mousemodel of multistage tumorigenesis offers an opportunity to ad-dress such issues. To that end, we have assessed the oncogenicfunctions of a small number of incompletely characterized genesthat are up-regulated in human hepatocellular carcinomas(HCC) and during mouse liver regeneration (2). One of thecandidate genes, a gene encoding a receptor for hyaluronan-mediated motility (RHAMM) is overexpressed in many types ofhuman cancers, including pancreatic ductal carcinoma, hepato-cellular carcinoma, multiple myeloma, breast cancer, gliomas,colon cancer, and prostate cancer (2–7); but the functions of atleast four proteins encoded by its alternatively spliced messengerRNAs and their roles, if any, in tumorigenesis are unclear.

Using the RIP-Tag; RIP-tva model, we find that isoform B ofRHAMM (RHAMMB) enhances the growth of mouse islettumors and promotes metastasis exclusively to the liver and locallymph nodes. Furthermore, we show that mouse pancreatic islettumor cells programmed to express RHAMMB form hepatic me-tastasis when injected into the tail vein of mice in a traditionalassay for metastasis. The cells also show evidence that RHAMMhas enhanced signaling via the epidermal growth factor receptor(EGFR). These observations and others suggest that RHAMMB

may be an important factor in tumor growth and progression andthat a better understanding of the RHAMM gene might offerinsights into the organotropism of metastatic cancer cells.

ResultsRHAMMB Promotes Tumor Growth and Metastasis to PancreaticLymph Nodes and the Liver in a Mouse Model of Pancreatic IsletTumors. To evaluate the malignant potential of genes reportedto be up-regulated both in HCC and during mouse liver re-generation (2), we examined five candidate genes, includingRHAMMB, paternally expressed 10 (PEG10), FLJ10540, FLJ11252,and FLJ11164. FLJ stands for the “full-length long Japan” col-lection of human cDNAs (8).The cDNAs of the candidates were cloned into avian retroviral

vector, RCASBP, with a FLAG epitope tag added to the Nterminus of PEG10, FLJ10540, and FLJ11252. We injected hightiter virus stocks (0.1 mL; >108 infectious units per milliliter) into7-wk-old RIP-Tag; RIP-tva mice by the intracardiac route. At thispoint, many islets show evidence of hyperplasia, allowing in-fection with oncoretrovirus vectors, which are dependent oncell division for successful infection (1). RCASBP–ALPP (Al-kaline Phosphatase) and RCASBP–Bcl-xL were chosen as con-trols. ALPP encodes a protein unlikely to contribute to tumori-genesis, serving as a negative control for effects of viral infection.We have previously shown that infection with RCASBP–Bcl-xLpromotes tumor growth and lymph node metastasis in RIP-Tag;RIP-tva mice (1), so infection with this virus provided a pos-itive control.Nine weeks after infection, the pancreas and other organs

were harvested for histological staging and grading of the lesions.Human RHAMMB significantly increased pancreatic tumorburden in 8 of 12 mice, but not all, compared with mice infectedwith RCASBP–ALPP (Fig. 1, P = 0.0097, Wilcoxon rank sumtest). RCASBP–Bcl-xL induced a small increase in pancreatictumor burden (P = 0.0087), whereas none of the other vec-

Author contributions: Y.-C.N.D. and H.V. designed research; Y.-C.N.D. performed re-search; C.-K.C. provided new reagents; D.S.K. contributed pathology analysis; Y.-C.N.D.analyzed data; and Y.-C.N.D. and H.V. wrote the paper.

The authors declare no conflict of interest.

Freely available online through the PNAS open access option.1To whom correspondence may be addressed. E-mail: [email protected] [email protected].

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1114022108/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1114022108 PNAS | October 4, 2011 | vol. 108 | no. 40 | 16753–16758

MED

ICALSC

IENCE

S

Dow

nloa

ded

by g

uest

on

Dec

embe

r 16

, 202

0

tors (RCASBP–FLAG-PEG10, RCASBP–FLAG-FLJ10540,RCASBP–FLAG-FLJ11252, or RCASBP–FLJ11164) causedany significant increase in tumor burden (Table 1 and Table S1).To aid the search for metastasis of islet tumors in RIP-Tag;

RIP-tva mice, tissue sections were subjected to immunohisto-chemical staining for synaptophysin, a neuroendocrine marker,and for insulin, a β-cell marker. Local lymph node metastaseswere detected in mice infected with RCASBP–RHAMMB (8 of11 mice, P= 0.001, Fisher’s exact test), RCASBP–Bcl-xL (7 of 15mice, P = 0.013), RCASBP–FLAG-PEG10 (2 of 8 mice, P =0.183), RCASBP–FLAG-FLJ10540 (4 of 7 mice, P = 0.015), andRCASBP–FLJ11164 (1 of 9 mice, P = 0.474) (Fig. 2, Table 1,and Table S1). The sizes of lymph node metastases in miceinfected with RCASBP–FLAG-PEG10, RCASBP–FLAG-FLJ10540, and RCASBP–FLJ11164 were small and were noteasily detected by hematoxylin and eosin staining, unlike those inmice infected with RCASBP–RHAMMB and RCASBP–Bcl-xL.No lymph node metastases were found in mice receiving thenegative control virus or RCASBP–FLAG-FLJ11252 (Table 1and Table S1).Importantly, liver metastases were found in 8 of 12 RCASBP–

RHAMMB infected mice (Fig. 2, Table 1, and Table S1, P =0.002, Fisher’s exact test), and the appearance of metastasis wasnot closely associated with the aggregated size of primarytumors. Several RIP-Tag; RIP-tva mice infected with RCASBP–RHAMMB with less or similar pancreatic tumor burden thancontrol mice still developed metastasis (Table S1), suggestingthat RHAMMB does not merely promote metastasis as a directconsequence of promoting tumor growth. In addition, 2 of 7mice infected with RCASBP–FLAG-FLJ10540 developed smallliver metastases, and 3 of 9 mice infected with RCASBP–FLJ11164 developed liver micrometastases with fewer than fivetumor cells on the histological sections (Table 1 and Table S1).These findings suggest that PEG10, FLJ10540, and FLJ11164have a modest effect on the promotion of metastasis in thismodel, but only RHAMMB had a dramatic effect on both tumorsize and metastasis to lymph nodes and the liver. No metastaseswere found at other sites, including lungs, heart, thymus, kidney,and spleen.In view of the profound effects of RHAMMB on tumor size

and metastasis 9 wk after infection, we asked whether sucheffects could be detected at earlier times. RIP-Tag; RIP-tva micewere euthanized 5 wk after being infected with RCASBP–RHAMMB at 7 wk of age. Although these RIP-Tag; RIP-tva micehad twofold more tumor burden than did the mice infected withRCASBP–ALPP at the same age, no metastases were found aftermicroscopic examination of tissue slices stained with reagentsthat detect insulin and synaptophysin (Table 1), indicating thatmore than 5 wk were required to observe any effects onmetastasis.To verify that RHAMM was indeed produced in the tumors

and metastases of RCASBP–RHAMMB infected mice, we gen-erated rabbit polyclonal antibodies against a region entirelyconserved among human, mouse, and rat RHAMM proteins,and present in all four human isoforms (Table S2 and Fig. S1).

Some islet tumors and the metastases in pancreatic lymph nodesand the livers in RIP-Tag; RIP-tva mice infected with RCASBP–RHAMMB stained positive for RHAMM (Fig. 2E). AlthoughRHAMM is negatively regulated by p53 in cell lines (9) and T-antigen represses p53 transactivation in RIP-Tag; RIP-tva mice,endogenous mouse RHAMM was not detectable by immuno-histochemical staining in most of the pancreatic islets and islettumors from uninfected RIP-Tag; RIP-tva mice (Fig. 2 C and D).Taken together, our results suggest that ectopic expression ofhuman RHAMMB increases tumor burden and independentlypromotes metastasis to pancreatic lymph nodes and the liver inour model of mouse pancreatic islet tumors.

RHAMMB Promotes Liver-Specific Metastasis in a Tail Vein Assay ofCancer Metastasis. The portal vein brings blood from the pan-creas, spleen, stomach, duodenum, and colon to the liver. Thus,it seemed possible that RHAMMB promotes metastasis of pan-creatic islet tumor cells to the liver because it promotes metas-tasis in general and the liver is the first target organ in the path ofblood drainage from the primary tumors. To distinguish morerigorously between enhancement of liver-specific versus gener-alized metastasis, we asked whether mouse islet tumor cell linesexpressing RHAMMB metastasized preferentially to the liverafter introduction into the general circulation. For this experi-ment, we injected established cell lines into the tail vein of re-cipient mice. Cells entering the venous circulation will firstencounter the capillaries of the lungs. If they are able to passthrough the lungs, they will then enter the arterial system andeventually pass through the portal circulation. Assuming thatpromotion of metastasis by RHAMMB affects steps followingintravasation, this approach should distinguish between liver-specific tropism and dependency on the circulatory route as anexplanation for the observed liver metastases.We infected a β-cell tumor cell line βTC-N134, (N134 for

brevity) (1) derived from an islet tumor in a RIP-Tag; RIP-tvamouse with RCASBP–Luciferase, RCASBP–RHAMMB, orRCASBP–Bcl-xL in vitro. To verify the expression of humanRHAMMB in N134 cells infected with RCASBP–RHAMMB, weperformed reverse transcription–PCR using primers specific forexon 16 of human RHAMM and readily detected humanRHAMMmRNA in cells infected with RCASBP–RHAMMB, butnot in uninfected cells (Fig. 3A).A total of 1 × 106 tumor cells were introduced into the tail vein

of immunodeficient mice, NOD/scid-lL2Rgc knockout (NSG).The mice receiving tumor cells infected with RCASBP–Lucifer-ase were monitored by in vivo bioluminescence imaging (Fig.3B). We observed signals from the thoracic cavity during the first

RCAS-RHAMMB control infection A B

Fig. 1. Significantly increased pancreatic islet tumor burden in 67% of RIP-Tag; RIP-tva mice infected with RCASBP–RHAMMB (n = 12). Representativepancreas and spleen from RIP-Tag; RIP-tva mice infected with RCASBP–ALPP(A) or RCASBP–RHAMMB (B). Mouse islet tumors are red due to angiogenesis.

Table 1. Impact of candidate genes on tumorigenesis in vivo

RCASBP–Age(wk)

Tumorburden(mm3)

Lymph nodemetastasis (%)

Livermetastasis (%)

ALPP 12 8.1 ± 4.5 0/5 mice (0) 0/5 mice (0)ALPP 16 99.7 ± 19.4 0/10 mice (0) 0/10 mice (0)RHAMMB 12 16.2 ± 8.9 0/6 mice (0) 0/6 mice (0)RHAMMB 16 298.3 ± 121.3 8/11 mice (73) 8/12 mice (67)Bcl-xL 16 150.5 ± 23 7/15 mice (46) 0/6 mice (0)FLAG-PEG10 16 57.8 ± 26.8 2/8 mice (25) 0/8 mice (0)FLAG-FLJ10540 16 126.1 ± 54.3 4/7 mice (57) 2/7 mice (28)FLAG-FLJ11252 16 41.2 ± 30.9 0/7 mice (0) 0/7 mice (0)FLJ11164 16 43.8 ± 25.1 1/9 mice (11) 0/6 mice (0)*

RCASBP retroviruses carrying indicated cDNAs were introduced to RIP-Tag; RIP-tva mice through intracardiac injection at 7 wk of age. Mice wereeuthanized at 12 or 16 wk of age for measurement of tumor burden and formetastasis survey. A standard formula for tumor volume was applied (vol-ume [mm3] = 0.52 × width2 × length). Tumor burden is the sum of the tumorvolume per mouse.*Mice having micrometastases with fewer than five cells were excluded.

16754 | www.pnas.org/cgi/doi/10.1073/pnas.1114022108 Du et al.

Dow

nloa

ded

by g

uest

on

Dec

embe

r 16

, 202

0

few days postinjection and the signals gradually became un-detectable within a week, suggesting that the tumor cells areinitially trapped in the capillary beds of the lungs.After 5 wk, organs of the recipient mice were harvested to

survey for metastases. The findings were dramatic: all five micereceiving RHAMMB-expressing tumor cells exhibited macro-metastases in the livers, with an average of 57 macrometastasesper mouse (Fig. 4 A and B). In contrast, three of five mice re-ceiving Bcl-xL–expressing tumor cells had liver macrometastases,with an average of one macrometastases per mouse. No mac-rometastases were found in mice receiving uninfected tumorcells or Luciferase-expressing tumor cells (Fig. 4 A and B). Theliver weight was increased about twofold in the mice receivingRHAMMB-expressing tumor cells compared with that in othermice (Fig. 4C). Moreover, the liver metastases in recipients of

RHAMMB-expressing tumor cells continued to express insulin asindicated by low blood glucose levels and by immunohisto-chemical staining for insulin (Fig. 4 D and E). To survey formicrometastases, immunohistochemical staining for synapto-physin was performed. Mice receiving uninfected tumor cells had1.8 ± 0.8 liver micrometastases and mice receiving Bcl-xL–expressing tumor cells had 26.6 ± 5 liver micrometastases (Fig.4F). Immunohistochemical staining for synaptophysin and in-sulin in sections of liver, lung, heart, thymus, brain, pancreas,spleen, kidney, bone marrow, and mammary gland revealed thatthe liver is the only organ with metastases (Fig. 4E). Our resultsindicate that overexpression of RHAMMB significantly promotesthe pancreatic islet tumor cells to establish large liver-specificmetastases after injection into the general venous circulation.In addition, we examined histological sections of several

organs for the existence of islet tumor cells at earlier time points(i.e., 1.5 h and 1, 3, 5, 7, and 14 d after tail vein injection).Consistent with in vivo bioluminescent results using Luciferase-expressing N134 cells in Fig. 3B, parental and RHAMMB-expressing N134 cells in clusters of various sizes were foundwithin pulmonary vessels associated with fresh fibrin during thefirst few days. Over the course of 7 d, we observed increasingorganization of fibrin, infiltrated neutrophils centered on pul-monary vessels, and gradual clearance of tumor cells. By 14 d,tumor cells were no longer found in pulmonary vessels. Im-munohistochemical staining for synaptophysin in liver sectionsrevealed a few positive cells. Thus, RHAMMB did not appear to

A B

C

D

E

isle

t tum

or (c

ontro

l)

n

orm

al p

ancr

eas

RHAMM

RHAMM

5x

5x

20x

20x

isle

t tum

or

20x

RHAMM

20x

20x

synaptophysin

liver

l

ymph

nod

e

liver

lym

ph n

ode

liver

l

ymph

nod

e

20x

20x

insulin

Fig. 2. Detection of metastases in RIP-Tag; RIP-tva mice infected withRCASBP–RHAMMB. (A and B) Representative lymph node and liver metastasesfound in RCASBP–RHAMMB infected mice. Immunohistochemical staining ofsynaptophysin (A) and insulin (B) is shown. (C–E) Immunohistochemicalstaining of RHAMM in pancreas from wild-type mouse (C), islet tumor froma control RIP-Tag; RIP-tva mice (D), and islet tumor, lymph node metastases,and liver metastases found in RCASBP–RHAMMB infected mice (E).

A B

Human actin Mouse actin

+RT -RT

HeL

a un

infe

cted

R

HA

MM

B

HeL

a un

infe

cted

R

HA

MM

B

Human RHAMM

1d

2d

3d

5d

7d

Fig. 3. Detection of human RHAMM mRNA or luciferase activity in themouse islet tumor cell line infected with RCASBP–RHAMMB or RCASBP–Lu-ciferase. (A) The expression of human RHAMMB in N134 cells infected withRCASBP–RHAMMB was confirmed by reverse transcription–PCR. mRNAs fromHeLa cell, N134 cells (uninfected), and N134 cells infected with RCASBP–RHAMMB (RHAMMB) were isolated for RT-PCR. Specific primers for humanRHAMM (exon 16) and human/mouse β-actin (as a control) were used. (B) Atotal of 1 × 106 N134 cells infected with RCASBP–Luciferase were injectedinto the tail vein of five recipient mice. The locations of tumor cells over-expressing luciferase were monitored by in vivo bioluminescent imaging 1, 2,3, 5, and 7 d after injection. RT, reverse transcriptase.

Du et al. PNAS | October 4, 2011 | vol. 108 | no. 40 | 16755

MED

ICALSC

IENCE

S

Dow

nloa

ded

by g

uest

on

Dec

embe

r 16

, 202

0

enhance infiltration of the lung by circulating tumor cells or byoutgrowth of micrometastasis in the liver at these early timepoints.

RHAMMB Does Not Enhance Cell Migration in a Transwell Assay orProliferation in Vitro. To examine whether RHAMMB enhancedthe ability of mouse pancreatic tumor cells to migrate in vitro, weperformed a two-chamber cell migration assay. Uninfected N134cells, RHAMMB-expressing tumor cells, or Bcl-xL–expressingtumor cells were seeded in the upper chambers of Transwellinserts. Bcl-xL–expressing tumor cells was used as a positivecontrol, as we have previously demonstrated that these cells havegreater ability to migrate in the two-chamber migration assaythan do parental cells (1). We measured cell migration alonga serum gradient through the membrane to the bottom of thechambers. Although RHAMMB, but not Bcl-xL, strongly pro-moted liver metastasis of mouse islet tumors in mice, the numberof migratory RHAMMB-expressing tumor cells was similar tothat of uninfected cells (Fig. 5A), suggesting that expression ofRHAMMB does not affect migration in vitro. This result indi-cates that the effect of RHAMMB on hepatic metastasis is un-related mechanistically to cell migration or that the Transwellassay does not mimic the in vivo microenvironment sufficientlywell to demonstrate an effect of RHAMMB on migration.To determine whether RHAMMB promoted cell proliferation

in vitro, we performed immunocytochemistry using antiseraagainst a proliferation marker, Ki67. The frequency of Ki67-positive cells was slightly reduced in RHAMMB-expressing tu-mor cells compared with the uninfected cells (Fig. 5B). Thesefindings indicate that RHAMMB does not stimulate cell pro-liferation in vitro and may even have a slight inhibitory effect.

RHAMMB Increases Phosphorylation of EGFR, Erk, and STAT3. Little isknown about the biochemical properties of the RHAMMB. Ithas been shown that Erk1/2 activity is aberrant in RHAMM−/−

fibroblasts (10) and activation of EGFR is required for wound-ing-induced hyaluronan synthesis (11). As an initial effect toexplain the physiological consequences of overexpression ofRHAMMB, we measured the relative amounts and phosphory-lation status of EGFR, Erk1/2, and STAT3 by Western blotanalysis using antibodies against total and phosphorylatedEGFR, Erk, and STAT3. We observed increased phosphoryla-tion of EGFR (on Tyr1068), Erk1/2 (on Thr202/Tyr204), andSTAT3 (on Ser727) in RHAMMB-expressing tumor cells com-pared with uninfected N134 cells (Fig. 6A), suggesting thatoverexpression of RHAMMB activates signaling pathways thatinclude EGFR, Erk1/2, and STAT3. In addition, we observedslightly reduced levels of the cell adhesion molecule E-cadherinin RHAMMB-expressing tumor cells; reduced E-cadherin couldcontribute to invasiveness of the tumor cells (12).To gauge whether the EGFR signaling pathway might be re-

quired for the survival of N134 cells overexpressing RHAMMB,we treated cells with gefitnib, a small molecule inhibitor of theEGFR kinase (13). Two days later, gefitnib-treated and un-treated cells were labeled with Alexa Fluor 488 annexin V andpropidium iodide and then subjected to flow cytometry toidentify apoptotic cells. We found that (i) overexpression ofRHAMMB in untreated tumor cells provided modest protectionagainst apoptosis, and (ii) gefitnib induced an approximatelyfourfold increase in apoptosis of RHAMMB-expressing cells, butdid not induce significant apoptosis in parental N134 cellsin vitro (Fig. 6B). Gefitinib treatment also reduced phosphory-lation of both Erk and EGFR, without significant changes inphosphorylation of STAT3 or Bcl-xL protein levels, inRHAMMB-expressing tumor cells (Fig. 6C). These results sug-gest that RHAMM uses the EGFR signaling pathway to promotetumor cell survival, although this effect cannot explain the spe-cific increase in hepatotropic metastasis.

DiscussionMetastasis is the major cause of cancer mortality and appears tooccur in an orderly sequence of general steps: local invasion,intravasation, survival in the circulation, extravasation, and col-onization. Two in vivo assays, “spontaneous metastasis” in RIP-

A RHAMMB Luciferase uninfected

0

10

20

30

40

50

60

70

# of

mac

rom

ets

in th

e liv

er

liver

wei

ght (

mg)

0

500

1000

1500

2000

2500

bloo

d gl

ucos

e (m

g/dL

) 0

20

40

60

80

100

120

140

160 B D C

10x

RHAMMB

4x

uninfected RHAMMB

4x

Bcl-xL

4x

insulin synaptophysin E F

Bcl-xL

Fig. 4. RHAMMB greatly promotes liver metastasis of pancreatic islet tu-mor cells in the tail vein metastasis assay. A total of 1 × 106 parental mousepancreatic islet tumor N134 cells and tumor cells overexpressing Luciferase,RHAMMB, or Bcl-xL were injected into the tail vein of recipient mice. Fiveweeks later, the recipient mice (n = 5 for each group) were euthanized tosurvey for metastatic sites and incidence (A and B), to record liver weight(C), and to measure the blood glucose (D). (E) The liver metastases in micereceiving RHAMMB-expressing cells continued to express insulin. Liversections from recipient mice with N134 tumor cells or tumor cells over-expressing RHAMMB or Bcl-xL were subjected to immunohistochemicalstaining of synaptophysin (F) to reveal the presence of metastases. Originalmagnification is indicated.

# of

mig

rato

ry c

ells

0

500

1000

1500

2000

2500

3000 A B

90

80

0

100

% o

f Ki6

7 po

sitiv

e ce

lls

70

Fig. 5. RHAMMB does not promote migration in a Transwell assay in vitro.(A) Uninfected N134 cells and N134 cells infected with RCASBP–RHAMMB orRCASBP–Bcl-xL were plated in the upper chambers of Transwell plates. After72 h, cells were counted in the lower chambers. Data are presented as themean numbers of cells in five fields under 20× magnification, and are rep-resentative of three independent experiments. (B) Proliferative indices ofthe two cell lines. Cultures grown on chamber slides from uninfected N134cells and N134 cells infected with RCASBP–RHAMMB were stained withantisera against Ki67. Data shown are the mean percentage ± SD fromtriplicate experiments.

16756 | www.pnas.org/cgi/doi/10.1073/pnas.1114022108 Du et al.

Dow

nloa

ded

by g

uest

on

Dec

embe

r 16

, 202

0

Tag; RIP-tva mouse model and “tail vein assay of cancer me-tastasis,” provide complementary information to dissect thespecific metastatic steps. Using the avian retroviral vector todeliver five candidate genes that are up-regulated in human HCCinto mouse pancreatic hyperplastic cells in RIP-Tag; RIP-tvamice, we have shown that RHAMMB among the five genesdramatically promotes metastasis in an organ-specific pattern,with the appearance of liver metastasis regardless of whethertumor cells enter the portal circulation endogenously in a mousemodel or whether they are injected into a tail vein to enter themajor venous circulation.In addition to RHAMMB, we found that FLJ10540 promotes

metastasis to lymph nodes and the liver, but the sizes and num-bers of metastases are less profound than RHAMMB. PEG10,FLJ11252, and FLJ11164 were not able to promote liver metas-tasis in the RIP-Tag; RIP-tva model. FLJ10540 has been shown tobe up-regulated in HCC and oral cavity squamous cell carcinomaand is associated with poor survival (14, 15). Overexpression ofFLJ10540 in 3T3 cells was also shown to promote survival in softagar and in low serum medium and induces tumor formation innude mice (14).

The functions and subcellular localizations of the variousproducts of the RHAMM locus are controversial (16). RHAMMproteins have been implicated in multiple functions due to theirassociation with hyaluronan, BRCA1, BARD1, CD44, Erk1/2,mitotic spindle, microtubules, and microfilaments (17). Theyhave been reported to act as cell-surface receptors but have alsobeen reported in the cytoplasm or the cell nucleus. Theseobservations may be attributed to the existence of different iso-forms generated by alternative splicing, as it is not clear whichisoform(s) was examined in many reports due to the uncertainspecificity of the antibodies and nucleic acid probes. Even thoughRHAMM proteins have been shown to be overexpressed invarious cancer cells, their oncogenic potentials and growth pro-moting signals have not been rigorously demonstrated eitherin vitro or in vivo. One possible explanation is that the oncogenicactivity of RHAMMB is too modest to be observed in conven-tional cell transformation assays but becomes evident in a moresensitive context, such as our RIP-Tag; RIP-tva mouse model.Further investigation is needed to decipher the mechanism bywhich metastasis is enhanced and to determine whether otherRHAMM isoforms have similar effects.Because hepatotropic metastasis promoted by RHAMMB is

also observed with tail vein injection, the effect cannot be me-diated directly through intravasation or be dependent on accessto the liver through the portal vein. The effect is not a conse-quence of size of the primary tumors, because abundant hepaticmetastases developed even in RCASBP–RHAMMB infected RIP-Tag; RIP-tva mice that had relatively small tumors. Increasedproliferation also appears not to explain the enhanced hepaticmetastasis, because RCASBP–RHAMMB infected N134 cells hada lower Ki67 index than uninfected cells. Furthermore, the effectis unlikely to be attributable to enhanced migration, becauseRHAMMB does not promote tumor cell migration in a two-chamber Transwell assay. Our findings suggest that RHAMMB

may allow islet tumor cells to invade and grow specifically in themicroenvironment of the liver. It will be important to determinethe cell specificity of this phenomenon and to identify the factor(s)in the liver that might be recognized by hepatotropic cancer cellsand thus serve as targets for interventions that prevent metastasis.In a preliminary effort to characterize the biochemical effects

of RHAMMB on pancreatic islet tumor cells, we showed that itenhances phosphorylation of EGFR, Erk1/2, and STAT3, im-plying activation of signaling pathways that are often affected incancers. Furthermore, treatment with an inhibitor of EGFRsignaling, gefitnib, induced apoptosis in RHAMMB-expressingcells, although we cannot exclude the possibility that this effectmay be mediated by inhibition of other uncharacterized kinases.Additionally, it has been shown that EGFR signaling contributesto tumorigenesis in RIP-Tag mice (18). The RHAMM full-lengthisoform, RHAMMA, enhances serum-induced Erk1/2 phos-phorylation in embryonic fibroblasts from RHAMM knockoutmice (10), and here we showed that overexpression ofRHAMMB phosphorylates Erk1/2 in mouse pancreatic islet tu-mor cells. Thus, Erk1/2 seem to be common targets of twoRHAMM isoforms. Conditional STAT3 knockout mice havebeen used to show that STAT3 is important for the function ofinterleukin-6 (IL-6) in liver regeneration (19). Because IL-6 isenriched in the liver, tumor cells with activated STAT3 may havea growth advantage in the liver.We previously showed that normal, untransformed mammary

cells, when introduced to the tail vein of the recipient mice,colonize to the lungs, a major organ site for breast cancer me-tastasis (20). Although RHAMMB promotes hepatotropic me-tastasis of islet tumor cells in the tail vein assay, it remainspossible that the hepatotropism is an inherent property of theislet tumor cells rather than determined by RHAMM protein.Further studies are required to determine whether uninfectedpancreatic islet tumor cells also reach the liver, but remaindormant or are unable to survive and/or grow. AlthoughRHAMMB does not provide a proliferation advantage in vitro,the slight protection against apoptosis observed in RHAMMB-

total Erk

p-Erk

WCE

total STAT3

p-STAT3

p-EGFR

E-cadherin

total EGFR

Gefitnib ( M): 0 0.5 1 10 10 Time (hr): 2 2 2 2 24

total Erk

p-Erk

BcL-xL

p-EGFR

total EGFR

total STAT3

p-STAT3

A B

C

0

10

20

30

40

50

60

% o

f apo

ptot

ic c

ells

RHAMMB: - - + +

Gefitnib: - + - +

Fig. 6. Overexpression of RHAMMB activates EGFR in pancreatic tumor cells,and EGFR small molecule inhibitor induced apoptosis of mouse pancreaticislet tumor cells in vitro. (A) Western blot analysis revealed the elevatedphosphorylated EGFR (p-EGFR), phosphorylated Erk (p-Erk), and phosphor-ylated STAT3 (p-STAT3), and decreased E-cadherin in N134 cells over-expressing RHAMMB (lane 2) compared with uninfected N134 cells (lane 1).(B) N134 cells and N134 cells overexpressing RHAMMB were cultured in thepresence of DMSO (vehicle control) or 10 μM gefitnib. Two days later, cellswere harvested and labeled with Alexa Fluor 488 annexin. The apoptoticcells were distinguished using a flow cytometer with a 488-nm laser to excitethe dye. Data shown are the mean percentage ± SD from triplicate experi-ments. (C) N134 cells overexpressing RHAMMB were cultured in the presenceof DMSO, 0.5, 1, or 10 μM gefitnib, and whole cell extracts were prepared 2or 24 h later for Western blot analysis of indicated antibodies. WCE, wholecell extracts.

Du et al. PNAS | October 4, 2011 | vol. 108 | no. 40 | 16757

MED

ICALSC

IENCE

S

Dow

nloa

ded

by g

uest

on

Dec

embe

r 16

, 202

0

expressing tumor cells might help them survive better in ectopicand hemodynamically stressful sites, such as the liver.The liver is the most common organ for the metastases in

human pancreatic neuroendocrine tumors and cancers of theintestines, and almost all patients will succumb to liver failurefrom the metastases. Because RIP-Tag mice (21) develop islettumors that do not metastasize, whereas such tumors in RIP-Tag;RIP-tvamice infected with vectors encoding genes like RHAMMB

form liver metastases readily, the model described here shouldbe useful for preclinical studies of cancers that preferentiallymetastasize to the liver.Numerous microarray-based screens and immunohistochemi-

cal screens have been performed to identify genes differentiallyexpressed in human tumors and normal cells. Some havereported overexpression of RHAMM in various tumors anda negative prognostic significance for its expression in breastcancer (4), multiple myeloma (22), and colon cancer (7). Ourstudy provides support for these observations by demonstratinga causal role for RHAMMB in liver metastasis in a mouse modelof multistep tumorigenesis.

Materials and MethodsGeneration of RIP-Tag; RIP-tva mice and N134 cell line has been described(1). NSG mice were generated by The Jackson Laboratory. All mice werehoused in accordance with institutional guidelines. All procedures involvingmice were approved by the institutional animal care and use committee.

RCASBP is a replication-competent avian leukosis virus with a splice acceptorand the Bryan-RSV pol gene. RCASBP–ALPP has been described previously(23). RCASBP–Luciferase was generated by Yi Li in the Varmus laboratory.RCASBP–RHAMMB, RCASBP–FLAG-PEG10, RCASBP–FLAG-FLJ10540, RCASBP–FLAG-FLJ11252, and RCASBP–FLJ11164 were generated in the Chou labora-tory. Viral propagation and titer determination were described (1). In-tracardiac injection was performed as described (24).

Tail Vein Injection of Tumor Cells and in Vivo Bioluminescent Imaging. Singlecell suspension of tumor cells was prepared before tail vein injection. Micewere i.v. injected in the tail vein using insulin syringes with 1 × 106 tumor cellsin 150 μL PBS. Mice were subjected to in vivo bioluminescent imaging asdescribed previously (25). All other experimental details are provided in SIMaterials and Methods.

ACKNOWLEDGMENTS. We thank members in the H.V. laboratory, especiallyJennifer Demers, Mary Ann Melnick, Andreas Giannakou, Gabriela Sanchez,and Raymond Dematteo for technical assistance; Katrina Podsypanina andRomel Somwar for insightful discussions; Levi Beverly for 293T cell line andreagents. We thank Danny Huang for mouse database design; and IrinaLinkov for protocols. We thank Leigh Selesner in the Y.-C.N.D. laboratory fortechnical assistance; Hua-Chien Chen at Chang-Gung University for RHAMMexpression data from HCC samples; Mouse Genetics Core Facility andResearch Animal Resource Center for foster service and animal husbandry;and Memorial Sloan-Kettering Cancer Center’s Molecular Cytology, FlowCytometry, and Small-Animal Imaging Cores for technical assistance. Thiswork was funded in part by a grant from the National Institutes ofHealth (5P01CA094060).

1. Du YC, Lewis BC, Hanahan D, Varmus H (2007) Assessing tumor progression factors by

somatic gene transfer into a mouse model: Bcl-xL promotes islet tumor cell invasion.

PLoS Biol 5:e276.2. Yang CW, et al. (2005) Integrative genomics based identification of potential human

hepatocarcinogenesis-associated cell cycle regulators: RHAMM as an example. Bio-

chem Biophys Res Commun 330:489–497.3. Akiyama Y, et al. (2001) Hyaluronate receptors mediating glioma cell migration and

proliferation. J Neurooncol 53:115–127.4. Assmann V, et al. (2001) The pattern of expression of the microtubule-binding protein

RHAMM/IHABP in mammary carcinoma suggests a role in the invasive behaviour of

tumour cells. J Pathol 195:191–196.5. Grützmann R, et al. (2004) Gene expression profiling of microdissected pancreatic

ductal carcinomas using high-density DNA microarrays. Neoplasia 6:611–622.6. Gust KM, et al. (2009) RHAMM (CD168) is overexpressed at the protein level and may con-

stitute an immunogenic antigen in advanced prostate cancer disease. Neoplasia 11:956–963.7. Zlobec I, Baker K, Terracciano LM, Lugli A (2008) RHAMM, p21 combined phenotype

identifies microsatellite instability-high colorectal cancers with a highly adverse

prognosis. Clin Cancer Res 14:3798–3806.8. Ota T, et al. (2004) Complete sequencing and characterization of 21,243 full-length

human cDNAs. Nat Genet 36:40–45.9. Sohr S, Engeland K (2008) RHAMM is differentially expressed in the cell cycle and

downregulated by the tumor suppressor p53. Cell Cycle 7:3448–3460.10. Tolg C, et al. (2006) Rhamm-/- fibroblasts are defective in CD44-mediated ERK1,2 mo-

togenic signaling, leading to defective skin wound repair. J Cell Biol 175:1017–1028.11. Monslow J, Sato N, Mack JA, Maytin EV (2009) Wounding-induced synthesis of hya-

luronic acid in organotypic epidermal cultures requires the release of heparin-binding

egf and activation of the EGFR. J Invest Dermatol 129:2046–2058.12. Perl AK, Wilgenbus P, Dahl U, Semb H, Christofori G (1998) A causal role for E-cad-

herin in the transition from adenoma to carcinoma. Nature 392:190–193.

13. Ciardiello F, et al. (2000) Antitumor effect and potentiation of cytotoxic drugs activityin human cancer cells by ZD-1839 (Iressa), an epidermal growth factor receptor-se-lective tyrosine kinase inhibitor. Clin Cancer Res 6:2053–2063.

14. Chen C-H, et al. (2007) FLJ10540-elicited cell transformation is through the activationof PI3-kinase/AKT pathway. Oncogene 26:4272–4283.

15. Chen C-H, et al. (2009) Expression of FLJ10540 is correlated with aggressiveness of oralcavity squamous cell carcinoma by stimulating cell migration and invasion throughincreased FOXM1 and MMP-2 activity. Oncogene 28:2723–2737.

16. Hofmann M, et al. (1998) Problems with RHAMM: A new link between surface ad-hesion and oncogenesis? Cell 95:591–592, author reply 592–593.

17. Maxwell CA, McCarthy J, Turley E (2008) Cell-surface and mitotic-spindle RHAMM:Moonlighting or dual oncogenic functions? J Cell Sci 121:925–932.

18. Nolan-Stevaux O, et al. (2010) Differential contribution to neuroendocrine tumori-genesis of parallel Egfr signaling in cancer cells and pericytes. Genes Cancer 1:125–141.

19. Li W, Liang X, Kellendonk C, Poli V, Taub R (2002) STAT3 contributes to the mitogenicresponse of hepatocytes during liver regeneration. J Biol Chem 277:28411–28417.

20. Podsypanina K, et al. (2008) Seeding and propagation of untransformed mousemammary cells in the lung. Science 321:1841–1844.

21. Hanahan D (1985) Heritable formation of pancreatic beta-cell tumours in transgenicmice expressing recombinant insulin/simian virus 40 oncogenes. Nature 315:115–122.

22. Maxwell CA, et al. (2004) RHAMM expression and isoform balance predict aggressivedisease and poor survival in multiple myeloma. Blood 104:1151–1158.

23. Fekete DM, Cepko CL (1993) Replication-competent retroviral vectors encoding al-kaline phosphatase reveal spatial restriction of viral gene expression/transduction inthe chick embryo. Mol Cell Biol 13:2604–2613.

24. Kang Y, et al. (2005) Breast cancer bone metastasis mediated by the Smad tumorsuppressor pathway. Proc Natl Acad Sci USA 102:13909–13914.

25. Du YC, Klimstra DS, Varmus H (2009) Activation of PyMT in beta cells induces irre-versible hyperplasia, but oncogene-dependent acinar cell carcinomas when activatedin pancreatic progenitors. PLoS ONE 4:e6932.

16758 | www.pnas.org/cgi/doi/10.1073/pnas.1114022108 Du et al.

Dow

nloa

ded

by g

uest

on

Dec

embe

r 16

, 202

0