Oncogenic epidermal growth factor receptor mutants with tandem duplication: gene structure and effects on receptor function

Oncogenic epidermal growth factor receptor mutants with tandemduplication: gene structure and e�ects on receptor function

Michael J Ciesielski1,2 and Robert A Fenstermaker*,1,2

1Department of Neurosurgery, State University of New York at Bu�alo School of Medicine and Biomedical Sciences; 2RoswellPark Cancer Institute, Elm and Carlton Streets, Bu�alo, New York, NY 14263, USA

A number of epidermal growth factor receptor (EGFR)deletion mutants have been identi®ed in gliomas, in whichthe EGFR gene is frequently ampli®ed and rearranged.We have previously characterized the structure of a genein A-172 human glioma cells that encodes a 190-kDaEGFR mutant with tandem duplication of the tyrosinekinase (TK) and calcium-mediated internalization(CAIN) domains. Here we describe a 185-kDa tandemduplication mutant (TDM) that is expressed in KE andA-1235 glioma cells, along with certain functionalcharacteristics of the mutants. The correspondingtranscripts in KE and A-1235 cells contain 1053additional nucleotides representing an in-frame duplica-tion of exons 18 through 25 which encode the entire TKregion and a portion of the CAIN domain. As withduplication of the entire TK/CAIN region (exons 18 ±26) in A-172 cells, duplication of exons 18 ± 25 isassociated with a speci®c genomic rearrangementbetween ¯anking introns. Involved introns containhomology to recombination signal sequence (RSS)heptamers present in the V(D)J region of the Tlymphocyte receptor gene. In de®ned medium, bothoncogenic TDM are constitutively autophosphorylatedand ine�ciently downregulated. High-a�nity binding isreduced in EGFR.TDM/18 ± 26, although the t12 ofreceptor internalization is not prolonged. Oncogene(2000) 19, 810 ± 820.

Keywords: epidermal growth factor receptor; glioma;intron; oncogene; phosphorylation; recombination

Introduction

Two of the most consistent genetic alterationsidenti®ed in human gliomas include EGFR geneampli®cation and rearrangement (Libermann et al.,1985; Humphrey et al., 1988; Bigner et al., 1990; Wonget al., 1987). In gliomas, levels of wild type EGFR maya�ect prognosis and are also proportional to tumorinvasiveness (Etienne et al., 1998; O'Rourke et al.,1997; Lund-Johansen et al., 1990). Expression of wildtype EGFR induces ligand-dependent cellular transfor-mation of astrocytes, indicating a signi®cant role forthe receptor in oncogenesis or tumor progression (Frisaet al., 1996). EGFR expression in gliomas is associated

with gene ampli®cation, which is present in up to 50%of these tumors. This leads to simultaneous expressionof both wild type and mutant receptors (Libermann etal., 1985; Humphrey et al., 1988; Bigner et al., 1990;Wong et al., 1987; Ekstrand et al., 1991).

Mutant EGFR in malignant gliomas arise from generearrangements with internal deletions (Sugawa et al.,1990; Humphrey et al., 1991; Wong et al., 1992). Mostmutants contain a deletion of speci®c exons encodingpart of the extracellular portion of the EGFR molecule.They may contribute to tumor progression via constitu-tive receptor activation, impaired receptor downregula-tion, activation of alternate signaling cascades,abrogation of apoptotic mechanisms, and perhaps othermechanisms as well (Nishikawa et al., 1994; Huang et al.,1997; Chu et al., 1997; Nagane et al., 1996). One mutantreceptor with an extensive N-terminal truncation(EGFRvI) is similar to the viral erbB-1 gene productwhich induces malignant transformation and potentconstitutive receptor activation (Wong et al., 1992; Haleyet al., 1990). The EGFRvII mutation has an in-framedeletion of 83 amino acids in domain IV of theextracellular region (amino acids 520 ± 603; exons 14and 15). This 150 kDa mutant receptor is capable ofbinding ligand and has enhanced tyrosine kinase activity(Humphrey et al., 1991). EGFRvIII is the most commonvariant yet identi®ed in malignant gliomas. It consists ofa 140-kDa EGFR mutant containing unique epitopesthat are of interest for receptor-targeted antitumorstrategies (Humphrey et al., 1990). EGFRvIII containsan in-frame loss of a large portion of the extracellularregion of the molecule (amino acids 6 ± 273; exons 2 ± 7).Each of these EGFR deletion mutants appears to arisefrom loss of speci®c exons from the genes that encodethem. In the case of EGFRvIII, the mutant protein mayarise from intragenic recombination between introns 1and 7 (Sugawa et al., 1990). EGFR mutants such asEGFRvIII are capable of oncogenic transformation inthe absence of ligand and their expression may suggest apoor prognosis and resistance to certain forms ofchemotherapy (Nishikawa et al., 1994; Schlegel et al.,1994; Batra et al., 1995; Nagane et al., 1998).

In addition to the small EGFR deletion mutantsfound in human gliomas, a 190 kDa EGFR-like formhas been described in A-172 human glioma cells that islarger than the wild type (170 kDa) molecule (Pan-neerselvam et al., 1995). Other large EGFR forms arefound in KE and A-1235 human glioma cell lines(Panneerselvam et al., 1995; Steck et al., 1988). The A-172 mutant contains tandem duplication of sequencesthat encode the tyrosine kinase and calcium inter-nalization domains of the molecule (Fenstermaker etal., 1998). Together with the mutant receptor found in

Oncogene (2000) 19, 810 ± 820ã 2000 Macmillan Publishers Ltd All rights reserved 0950 ± 9232/00 $15.00

www.nature.com/onc

*Correspondence: RA Fenstermaker, Department of Neurosurgery,Roswell Park Cancer Institute, Elm and Carlton, Bu�alo, New York,NY 14263, USAReceived 1 September 1999; revised 23 November 1999; accepted 9December 1999

both A-172 cells and glioma biopsy specimens, andevidence presented here regarding the related KE andA-1235 mutants, results suggest the existence of adistinct class of oncogenic EGFR tandem duplicationmutants (TDM) present in some human gliomas. LikeEGFRvIII, these mutants arise from rearrangementbetween introns. Rather than deletion of interveningsequences, the mutants in A-172, A-1235, KE cells andcertain solid tumors contain a duplication of thosesequences. The resulting mutants are oncogenic andexhibit constitutive autophosphorylation, as well asimpaired ligand binding and receptor downregulation.

Results

Expression of large EGFR-like proteins in human gliomacell lines and biopsy material

EGFR-related proteins that are larger than the170 kDa wild type molecule have been observedpreviously in human glioma cell lines (Panneerselvamet al., 1995; Steck et al., 1988). One such 190 kDaprotein is recognized in A-172 cells by antibodies toboth intracellular and extracellular regions of theEGFR, but not by antibodies to HER2, HER3 orHER4 (Panneerselvam et al., 1995; Fenstermaker et al.,1998). In addition to the wild type 170 kDa EGFRmolecule, KE and A-1235 human glioma cells expressEGFR-reactive species of approximately 185 kDa(Figure 1a). In the absence of exogenous ligand, levelsof the wild type and mutant EGFR forms are similarin each of the three, cultured cell lines. As with the190-kDa EGFR mutant in A-172 cells, neither p185cross-reacts with the other HER-family proteins (datanot shown). Similarly, analysis of a panel of glioma

biopsy specimens revealed a tumor with 140-, 170- and190-kDa EGFR forms (Figure 1b; RPG-77). Otherhigh molecular weight forms of the EGFR are presentin some glioma biopsy samples as well (our unpub-lished observations), indicating that such large EGFRspecies are not artifacts of cell culture.

Glioma cells produce large EGFR-like transcripts

The major EGFR-related transcript present in normalsyncytiotrophoblast is 10.5 kb in size (Ullrich et al.,1984). Size di�erences between deglycosylated wild typeand mutant EGFR protein cores in KE cells suggestthat approximately one kilobase of additional mRNAsequence would be required, beyond that present in thewild type EGFR transcript, to encode the EGFR-likeprotein present in A-172 cells (Steck et al., 1988). Inaddition to the 11.5 kb species present in A-172 cells,we and others have identi®ed similar mRNA species inKE and A-1235 cells (Figure 1c; Steck et al., 1988;Fenstermaker et al., 1998). We have not detected anytranscripts larger than 10.5 kb in A-431 carcinoma cellsor in any other cell line that does not express either185 kDa or 190 kDa EGFR-reactive species. Thus,expression of the 11.5 kb transcripts correspondsclosely to expression of the high molecular weightEGFR-like proteins. Moreover, the di�erence in sizebetween wild type and mutant transcripts is su�cientto account for the apparent di�erence in weightbetween wild type and mutant proteins.

Mutant EGFR transcripts in human A-172, KE andA-1235 glioma cells

We used several di�erent sets of EGFR-speci®coligodeoxyribonucleotide (ODN) pairs in reverse tran-

Figure 1 (a) Western blot of lysates from A-431 carcinoma and A-172, A-1235, and KE human glioma cell lines using anti-EGFRantibody 1005 (Santa Cruz) and equal amounts of total protein per lane. (b) Western blot of lysates from A-172 cells (lane 1) andmalignant glioma biopsy samples (lanes 2 ± 5). (c) Northern blot with total cellular RNA from indicated cell lines. Blot was probedwith a DIG-labeled riboprobe spanning the sequence between nucleotides +1347 and +1805 of the human EGFR cDNA (Ullrichet al., 1984)

Oncogene

EGFR tandem duplication mutantsMJ Ciesielski and RA Fenstermaker

811

scription-polymerase chain reactions (RT ±PCR) tocharacterize EGFR-related transcripts in glioma andA-431 carcinoma cells. With all of the cell lines tested,each ODN pair produced a speci®c fragment ofuniform molecular weight that corresponds to the sizeof wild type EGFR sequence (Figure 2a ± d; Ullrich etal., 1984). ODN that are homologous to sequences¯anking the extracellular region of the EGFR moleculeproduced 2035 bp bands in each case (Figure 2a; +187to +2221). No signi®cant secondary bands related tothe extracellular region were detectable in any of thefour cell lines. ODN pairs binding to the EGFR cDNAat +2101 and +2292 produced secondary major bandsin A-172, A-1235 and KE glioma cells (Figure 2b). Thesecondary band in A-172 cells is slightly larger (48 bp)than that in either KE or A-1235 cells. Similarly, ODNspanning the region from +2101 to +3380 producedone PCR fragment (1280 bp) in each of the four celllines and a second major band in A-1235, KE and A-172 glioma cell lines (2333 bp; Figure 2c). AnotherODN combination binding at +3116 and +3819produced a 704 bp band consisting of wild typesequence in each of the cell lines including A-431carcinoma and larger secondary bands in the gliomacell lines (Figure 2d). These results suggest the presenceof four distinct species of EGFR mRNA (wild typeand mutant) in the human glioma cell lines with smallsize di�erences between the mutant transcript in A-172cells and those in KE and A-1235 cells. RNA fromRPG-77 gave a set of RT ±PCR fragments identical insize to those obtained with A-172 cells, as well as thoseexpected for the EGFRvIII transcript (data notshown).

Nucleotide sequences of the mutant EGFR transcripts

The nucleotide sequences of the smaller major PCRfragments derived from both A-431 and glioma cells

were identical to the published EGFR cDNA sequenceas reported by Ullrich et al., 1984. Sequence data fromthe secondary PCR fragments in A-1235 and KE cellsrepresented in Figure 2c (1857 bp) and Figure 2d(2333 bp) reveal a tandem duplication of nucleotides+2248 to +3300. This corresponds to a duplication ofexons 18 through 25 as derived from the avian EGFRexon-intron boundary structure (Callaghan et al.,1993). This mutation creates an in-frame, tandemduplication of the tyrosine kinase (TK) domain as wellas a portion of the calcium-mediated receptor inter-nalization (CAIN) domain (Figure 3). The tandemduplication of EGFR exons 18 through 25 present inKE and A-1235 cells contains one fewer exon than themutant EGFR transcript in A-172 cells (i.e. exon 26;Fenstermaker et al., 1998). Sequencing of RT ±PCRfragments from RPG-77 gave results identical to thoseobtained with fragments from A-172 cells (i.e. tandemduplication of exons 18 through 26).

Rearrangement of introns 17, 25 and 26 in glioma celllines and in RPG-77

Based on the in-frame arrangement of duplicatedsequences in the various mutants, we hypothesized thattandem duplication of exons 18 through 25 arose viarecombination between ¯anking introns 17 and 25, muchas with EGFR introns 17 and 26 in A-172 cells(Fenstermaker et al., 1998). ODN homologous to exonsequences near the exon-intron junctions of the avianEGFR gene were used to isolate wild type EGFR introns17, 25 and 26 from A-431, A-172, A-1235 and KEgenomic DNA (Figure 4a, lanes 1 ± 12). Using an ODNbinding to the 3'-end of exon 26 and an ODN binding tothe 5'-end of exon 18, a single PCR fragment wasampli®ed from A-172 genomic DNA template (Figure4a; lane 14). No such fragment could be ampli®ed fromA-431 cells (Figure 4a; lane 13), or from either of the

Figure 2 PCR products using four di�erent pairs of ODN on cDNA template mixes from RT reactions with A-431, A-172,A-1235 and KE cellular RNA. ODN pairs and resulting PCR products correspond to: (a) the extracellular region (nucleotides +187to +2221); (b) the transmembrane and 5'-end of the intracellular region (nucleotides +2101 to +2292); (c) the intracellular region(nucleotides +2101 to +3380); and (d) the 3'-end of the intracellular region (nucleotides +3116 to +3819). (e) Human EGFRcoding region with functional domains. Numbers correspond to the 5'-end of each ODN relative to the coding strand (Ullrich et al.,1984). Lines with arrows indicate expected wild type PCR fragment lengths in corresponding panels a ± d respectively

EGFR tandem duplication mutantsMJ Ciesielski and RA Fenstermaker

812

Oncogene

other two, glioma cell lines with these ODN (Figure 4a;lanes 15 and 16). Similarly, using an ODN that binds tothe 3'-end of exon 25 and one that binds to the 5'-end ofexon 18, PCR fragments were ampli®ed from A-1235and KE genomic DNA templates (Figure 4a; lanes 19and 20). PCR using A-172 genomic DNA and these sameODN produced a larger fragment that also contains theintervening exon 26 (Figure 4a; lane 18). No `chimeric'intron fragments could be detected in genomic DNAfrom A-431 cells (Figure 4a; lanes 13 and 17). A 26/17chimeric intron was also detected in DNA from theglioma biopsy specimen RPG-77 (Figure 4b), consistentwith the ®nding of a 190-kDa EGFR form in that tumor(Figure 1b).

Wild type introns 17, 25 and 26 were also isolatedfrom human renal genomic DNA using PCR. Compar-ing the DNA sequence of glioma-speci®c PCRfragments to that of wild type introns 17 and 25revealed elements of both introns incorporated intochimeric (25/17) introns. In KE cells, the mutant introncontains 339 bp of intron 25 (wild type intron 25 has380 bp) and 443 bp of intron 17 (wild type intron 17has 792 bp). In A-1235 cells, the mutant introncontains 296 bp of intron 25 and 417 bp of intron17. Therefore, in KE and A-1235 cells, rearrangementhas occurred at distinct sites in both introns 25 and 17.Similarly, in A-172 cells a third site of rearrangement ispresent in intron 17 (Figure 4d).

Analysis of regions ¯anking each of the variousrecombination sites in wild type introns 17, 25 and 26reveals sequences with strong homology to the

recombination signal sequence (RSS) heptamers pre-sent in the V(D)J region of the T lymphocyte receptorgene (Figure 4c). Homology to the potential mybbinding sites present in introns 17 and 26 of A-172cells is not evident at recombination sites in introns 17and 25 in either A-1235 or KE cells (Fenstermaker etal., 1998). Also present in proximity to recombinationsites in each intron is a relatively conserved TGTTTCmotif at 5 of 6 rearrangement sites (Figure 4d). Thissequence is most consistent in intron 17 and ishomologous to a portion of the nonamer site foundin the V(D)J region.

NR6 fibroblast transformation by the mutant receptors

Murine NR6 ®broblasts do not express detectableEGFR (Figure 5a) and do not exhibit anchorage-independent growth in agar, either with or withoutligand (Figure 6a,e). NR6 cells were transfected withan expression vector containing the wild type humanEGFR cDNA or the tandem duplication mutants.Clonal cell lines were selected for relatively equal levelsof expression (Figure 5a). NR6 cells with the 170-kDawild type EGFR formed colonies in soft agar, but onlyin the presence of ligand (Figure 6b,f). In contrast,both of the tandem mutants induced ligand-indepen-dent colony formation after 14 days in agar(EGFR.TDM/18 ± 26; Figure 6c,g and EGFR.TDM/18 ± 25, Figure 6d,h).

NR6 cells transfected with the neo-resistant expres-sion vector containing no EGFR gene failed to

Figure 3 Structure of wild type EGFR (top), EGFR.TDM/18 ± 26 transcript in A-172 cells (middle), and the EGFR.TDM/18 ± 25transcript in A-1235 and KE cells (bottom) with partial nucleotide and amino acid sequences at predicted exon boundaries(Callaghan et al., 1993).

Oncogene

EGFR tandem duplication mutantsMJ Ciesielski and RA Fenstermaker

813

produce detectable tumors in nude mice (Figure 5b),while NR6 cells expressing wild type EGFR producedsmall subcutaneous nodules that did not exhibitprogressive growth over the period of observation.In contrast, both tandem duplication mutantsproduced large tumors with rapid growth beginning30 ± 40 days after implantation.

Autophosphorylation and downregulation of mutant andwild type receptors

Neither of the two EGFR forms identi®ed in A-172cells (p170 and p190) appears to be constitutivelyautophosphorylated following complete serum starva-tion (Fenstermaker et al., 1998). In contrast, mutantreceptors in A-172, KE and A-1235 cells, that havebeen cultured in de®ned medium (Optimem) withoutadded ligand, exhibit constitutive autophosphoryla-

tion (Figure 7a,c,e). In each of the glioma cell lines,addition of ligand leads to rapid phosphorylation ofthe wild type receptor. In addition, the mutantreceptors in A-172 and A-1235 cells (Figure 7a,c)exhibited a more modest, but de®nite, induction inphosphorylation. In contrast, NR6 cells transfectedwith the EGFR.TDM/18 ± 26 showed no suchinduction in the phosphorylated form relative tototal EGFR levels (Figure 8c,d) suggesting that co-expression of the wild type receptor, or other celltype-speci®c factors, may be necessary for thisresponse to occur. Wild type 170 kDa EGFRprotein levels are rapidly downregulated in each ofthe glioma cell lines (KE, A-1235 and A-172)following stimulation by ligand (Figure 7b,d,f). Withcontinuous exposure to ligand, levels of the wildtype receptor in A-172 cells do not return to basalvalues even by 24 h.

Figure 4 (a) PCR with genomic DNA from A-431, A-172, A-1235, and KE cells. ODN are from exons 17 and 18 (intron 17; lanes1 ± 4), exons 26 and 27 (intron 26; lanes 5 ± 8); exons 25 and 27 (introns 25 and 26; lanes 9 ± 12); exons 26 and 18 (chimeric intron 26/17; lanes 13 ± 16); and exons 25 and 18 (chimeric intron 25/17; lanes 17 ± 20). (b) PCR with genomic DNA from A-431 cells andglioma biopsy sample RP-G77 using ODN from exons 26 and 18 showing introns 17, 26 and a chimeric (26/17) intron. (c)Homology to V(D)J-like recombination signal sequence (RSS) heptamers in each intron. Numbers represent the distance innucleotides of each motif from the 5'-end of the indicated intron. (d) Structure of the 26/17 and 25/17 chimeric introns andsequences at sites of recombination. Introns 17, 25 and 26 contain a relatively conserved TGTTTC motif (boxes) adjacent to sites ofrecombination

EGFR tandem duplication mutantsMJ Ciesielski and RA Fenstermaker

814

Oncogene

Receptor binding and internalization of ligand

Although Western blots revealed that each of the NR6transfectants express wild type and mutant receptors atsimilar levels (Figure 5a), the total binding capacity ofthe 190 kDa, EGFR.TDM/18 ± 26 was only about 5%of that displayed by the wild type receptor (Figure 9a).At the concentration of [125I]EGF used here neitherbinding in NR6.neo5 cells with no detectable EGFR,nor binding in NR6.185.B8 cells with p185, could bemeasured above background. Therefore, ligand inter-nalization rates were measured in NR6.170.A5 (p170)and NR6.190.A3 (p190) cell lines only. The t12 ofinternalization was 30 min in each of the two cell lines(Figure 9a). Thus, both EGFR and EGFR.TDM/18 ±26 receptors appear to undergo internalization atsimilar rates, although the 190 kDa mutant receptorexhibits greatly reduced binding capacity (*95%)relative to the wild type receptor.

Scatchard analysis of binding data from transfectedNR6 cells showed the presence of both high- and low-a�nity binding sites (Figure 9b,c). The high-a�nitysites of the wild type receptor (KD=4.4610711 M;

1.76106 receptors per cell) were much greater innumber than high-a�nity sites of EGFR.TDM/18 ±26 (KD=1.1610711 M; 1.06105 receptors per cell).Western blots of the two NR6 cell lines reveal similarnumbers of total receptors overall (Figure 5a). There-fore, although both wild type EGFR andEGFR.TDM/18 ± 26 receptors undergo internalizationat similar rates, the 190-kDa mutant receptor exhibits asubstantial reduction in ligand binding, in associationwith a large decrease in the number of high-a�nityreceptor sites per cell.

Discussion

High molecular weight EGFR-like proteins have beendescribed previously in human glioma cells and certainstructural and functional characteristics of theseproteins have also been reported (Panneerselvam etal., 1995; Steck et al., 1988; Fenstermaker et al., 1998).To date, however, the potential relationship of thesetandem duplication mutants to tumorigenesis has notbeen reported. We have previously shown the genetic

Figure 5 (a) Western blot of EGFR proteins from individual NR6 clones transfected with the pLXIN bicistronic expression vectorwith: no EGFR (NR6.neo.5), wild type EGFR (NR6.170.A5), EGFR.TDM/18 ± 26 (NR6.190.A3), and EGFR.TDM/18 ± 25(NR6.185.B8) cDNA. (b) Nude mice were injected with 16107 NR6 transfectants and tumor volume was calculated as described.NR6.neo.5 parent vector (*, n=6); p170 wild type EGFR NR6.170.A5 (&, n=7); p190 TK/CAIN EGFR tandem mutantNR6.190.A3 (^, n=5); and p185 TK/CAIN EGFR tandem mutant NR6.185.B8 (~, n=3)

Figure 6 Growth of NR6 transfectants in soft agar in the absence (top row) and presence (bottom row) of TGFa (40 ng/ml). NR6cells transfected with expression vectors include: the pLXIN parent vector (NR6.neo.5 cell line; a, e); the 170 kDa, wild type EGFR(NR6.170.A5 cell line; b, f); the 190 kDa, EGFR.TDM/18 ± 26 (NR6.190.A3 cell line; c, g); and the 185 kDa, EGFR.TDM/18 ± 25(NR6.185.B8 cell line; d, h)

Oncogene

EGFR tandem duplication mutantsMJ Ciesielski and RA Fenstermaker

815

structure of one such EGFR mutant found in A-172human glioma cells (Fenstermaker et al., 1998). Thecurrent study describes a closely related mutant EGFRpresent in two other human glioma cell lines.Additional structural and functional characteristics ofthese two TDM are described.

In addition to the A-172 cell line, we haveestablished the presence of the EGFR.TDM/18 ± 26 ina human glioma (glioblastoma) tissue sample (RPG-77)

obtained directly at surgery. Thus, the TDM are clearlynot an artifact of cell culture, although their frequencyof occurrence is less than that of the most commondeletion mutant EGFRvIII (our unpublished observa-tions). In addition to the EGFR.TDM/18 ± 26 and wildtype EGFR, RPG-77 contains a 140-kDa EGFR formconsistent with EGFRvIII. The latter has beenobserved to be extremely unstable in cultured gliomacell lines, in apparent contrast to the two types of

Figure 7 Duplicate Western blots of EGFR proteins immunoprecipitated from A-172 (a, b), A-1235 (c, d), and KE (e, f) cell lines.Glioma cells were maintained in Optimem medium without (7) or with (+) EGF (100 ng/ml) for the indicated times. Anti-phosphotyrosine antibody PY99 (a, c, e) and anti-EGFR antibody (b, d, f) were used for immunoprecipitations. In each case anti-EGFR antibody was used for Western blot detection of the immunoprecipitated EGFR forms

Figure 8 Duplicate Western blots of EGFR proteins immunoprecipitated from NR6 transfectants. NR6.170.A5 cells (a, b) andNR6.190.A3 cells (c, d) were maintained in Optimem medium without (7) or with (+) EGF (100 ng/ml) for the indicated times.Anti-phosphotyrosine antibody PY99 (a, c) and anti-EGFR antibody (b, d) were used for immunoprecipitation. Anti-EGFRantibody was used for Western blot detection of phosphorylated and total EGFR

EGFR tandem duplication mutantsMJ Ciesielski and RA Fenstermaker

816

Oncogene

TDM. The presence of both deletion and tandemduplication mutations in the same tumor suggests thatthe molecular mechanisms underlying their formationcould be fundamentally related.

Nucleotide sequences of the tandem mutant cDNAobtained from KE and A1235 cell lines were comparedto the avian EGFR exon-intron structure (Callaghan etal., 1993). This revealed a tandem duplication of exons18 through 25. Based on this same numbering system,the tandem mutant in A-172 cells contains a duplica-tion of exons 18 through 26. Thus, the mutant EGFRtranscripts present in both A-1235 and KE humanglioma cells contain one complete copy of the TK/CAIN region and a second complete copy of the TKlinked to a portion of the CAIN domain. In contrast,the tandem mutant in A-172 cells contains twocomplete copies of the TK/CAIN region. Functionally,the two types of mutants appear to behave similarly.

The EGFR/18 ± 25 mutant described here arises froma genomic alteration in which introns 25 and 17 haveundergone recombination. Similarly, the EGFR/18 ± 26the TK/CAIN duplication found in A-172 cells, andthe associated recombination between ¯anking introns,strongly suggest the presence of a mechanism in

malignant gliomas by which genomic rearrangementmay occur between susceptible sites in distant introns.Such rearrangements lead to the creation of genes thatencode highly oncogenic EGFR-like proteins. Inaddition to tandem duplication of sequences, evidencesuggests that intragenic deletions within the EGFRgene that give rise to mutants such as EGFRvIII alsoarise from rearrangement between introns (Sugawa etal., 1990; our unpublished observations). It will beinteresting to determine whether the mechanismsunderlying intragenic deletion and tandem duplicationinvolve similar processes and molecular machinery.

The multi-domain organization of EGFR is presentin the other HER-family receptors as well. Studies inwhich various functional domains of EGFR and neu(HER2) have been exchanged reveal that the ligandbinding, transmembrane, and tyrosine kinase domainsare capable of functioning as independent modules(Lehvaslaiho et al., 1989). A detailed analysis of avianEGFR exon-intron junctions reveals that exon duplica-tion probably occurred within the ligand bindingregion of the EGFR gene at some time during itshistory (Callaghan et al., 1993). Thus, as one step inthe evolution of the EGFR gene, susceptible introns

Figure 9 (a) Time course of [125I]-EGF internalization following binding in NR6.170A5 and NR6.190A3 cell lines. Radioreceptordata were plotted by the method of Scatchard for NR6.170A5 cells (b) and NR6.190A3 cells (c) with receptor numbers and bindingconstants determined using the LIGAND program (Munson and Rodbard, 1980)

Oncogene

EGFR tandem duplication mutantsMJ Ciesielski and RA Fenstermaker

817

may have undergone genetic recombination withtandem duplication of speci®c exons. This raises thepossibility that susceptible genetic loci exist withinEGFR introns that could give rise to recombination inthe wild type EGFR gene in both normal cells and ingliomas, and that common protein factors could beinvolved in the two phenomena.

Comparison of the intron rearrangement sites ineach cell line reveal that di�erent loci are involved ineach individual case. Nucleotide sequences ¯anking thesites of recombination in both introns 17 and 25 di�erin KE and A-1235 cells. In addition, the site ofrecombinant in intron 17 of A-172 cells is di�erentfrom that of either KE or A-1235 cells. Hence, anumber of di�erent loci within introns 17, 25 and 26may be susceptible to recombination. The presence of aTGTTTC motif adjacent to ®ve of six recombinationsites may be of signi®cance and suggests the possibilityof sequence-speci®c protein-DNA interaction. Also ofpotential signi®cance is the ®nding in each involvedintron of at least two copies of V(D)J-like recombina-tion signal sequence (RSS) heptamers CAC(A/T)GTGthat ¯ank sites of rearrangement. Together theseobservations suggest the possibility that V(D)J-likerecombination could be involved in the formation ofthe tandem duplication mutants. Such a mechanism ofoncogenic activation has been postulated for theformation of lymphoid malignancies (Roth and Craig,1998).

Expression of EGF or TGFa occurs frequently inmalignant gliomas and may lead to autocrine-stimulated tumor cell proliferation and cellulartransformation (Blasband et al., 1990; Maruno et al.,1991). Ligand-dependent cellular transformation hasbeen reported in ®broblasts and immortalized murineastrocytes following high-level expression of the hu-man wild type EGFR (Di Fiore et al., 1987; Frisa etal., 1996). Hence, in the presence of other criticalgenetic alterations, ligand-dependent glial cell trans-formation may occur via a normally functionalEGFR.

The tandem mutants described here could causeligand-independent transformation of NR6 cells, andby extension glial cells, as a result of enhanced basalEGFR autophosphorylation. Either or both tyrosinekinase domains could phosphorylate substrates result-ing in activation of various signaling cascades. In bothof the TK/CAIN mutants (p185 and p190), the spatialrelationship of the upstream tyrosine kinase to thesingle regulatory domain is altered. Thus, at aminimum the upstream tyrosine kinase is likely to beunregulated in both of these mutants leading toactivation of associated substrates. Alternatively, orin addition, the downstream TK could be activated byspatial separation from the ligand-binding domain, thefunction of which is clearly altered.

Phosphorylation studies of the TDM in glioma celllines maintained in de®ned medium revealed consti-tutive autophosphorylation, as well as an increase inphosphorylation produced by ligand. In contrast,studies of the EGFR.TDM/18 ± 26 in NR6 cellsfailed to show a detectable induction in basalphosphorylation by EGF. These results, along withthe ®nding of markedly reduced binding a�nity ofthe tandem mutants, strongly suggest that theinduction in glioma cells is not mediated directly by

way of the mutant receptor itself. Instead, enhancedEGFR.TDM/18 ± 26 phosphorylation is probably aresult of transphosphorylation by the activated wildtype receptor.

The mechanism by which EGFRvIII transduces anoncogenic signal remains under investigation. EGFR-vIII constitutively complexes with the adapter proteinsGrb2 and Shc (Moscatello et al., 1996; Chu et al.,1997; Montgomery et al., 1995). The mutant receptorhas also been noted to activate MAP kinase in some,but not all, studies and extracellular signal-regulatedkinase as well (Moscatello et al., 1996; Chu et al.,1997). Therefore, EGFRvIII may activate a di�erentcomplement of substrates than wild type EGFR. As aresult, co-expression of both wild type EGFR and oneof the deletion or tandem duplication mutants couldhelp to maintain a transformed state by simultaneousactivation of parallel signaling pathways.

There is some evidence that EGFRvIII may createan oncogenic signal via ine�cient receptor down-regulation, rather than by way of constitutive tyrosinekinase activation (Chu et al., 1997; Huang et al., 1997).Thus, unlike wild type receptors, EGFRvIII may beunable to attenuate mitogenic signaling by receptorinternalization (Huang et al., 1997). Within thecarboxyterminal region of EGFR is a functionaldomain designated CAIN, which modulates receptorinternalization (Chen et al., 1989). Although deletionof internalization sequences leads to increased onco-genicity, presumably via prolongation of cell-surfacereceptor half-life, it is unknown how CAIN duplicationby itself might a�ect receptor internalization andtransforming ability (Chen et al., 1989; Wells et al.,1990).

Evidence suggests that both tandem duplicationmutants may also be ine�ciently downregulated. Inall three glioma cell lines, wild type EGFR israpidly internalized and degraded. However, in thecase of EGFR.TDM/18 ± 26 the t12 for ligandinternalization was indistinguishable from that ofthe wild type receptor. EGFR.TDM/18 ± 26 has amarked reduction in high-a�nity binding sitescompared to the wild type receptor (*94%). Thisdecrease in the number of high-a�nity sites issu�cient to account for the reduction in totalbinding of ligand to EGFR.TDM/18 ± 26 comparedto EGFR (*95%). Thus, impaired downregulationmay result from decreased binding of ligand to themutant receptors through the loss of high-a�nitysites, rather than from an alteration in the e�ciencyof receptor internalization per se.

A precedent for mutations within the intracellularregion of EGFR a�ecting ligand binding and inter-nalization has been reported previously (Prywes et al.,1986). Here, both deletion and insertion mutationscreated by site-directed mutagenesis were shown toabolish high a�nity binding sites and impair the abilityof the EGFR to internalize. Hence, the tandemduplication mutants could promote oncogenesis byway of impaired ligand binding and receptor down-regulation, in concert with constitutive autophosphor-ylation. Additional studies will be required todetermine whether the tandem mutants activatesignaling pathways that are di�erent from the wildtype receptor, and whether one, or both, tyrosinekinases are constitutively active.

EGFR tandem duplication mutantsMJ Ciesielski and RA Fenstermaker

818

Oncogene

Materials and methods

Human glioma samples and glioma cell culture

Human glioma samples were ¯ash-frozen in liquid nitrogen atthe time of surgery and stored at 7708C. We obtained KEcells from the late Dr Peter Steck and A-1235 cells from DrStuart Aaronson. All other cell lines were obtained fromAmerican Type Culture Collection. Cultured glioma cell lineswere grown on 100 mm tissue culture plates in Dulbecco'sModi®ed Eagle Medium (DMEM) with 1500 mg/L glucose,2 mM L-glutamine, and 110 mg/L sodium pyruvate, supple-mented with 10% fetal bovine serum, 50 units/ml penicillinand 50 mg/ml streptomycin. Cells were grown at 378C in 5%CO2 with media changes 2 ± 3 times per week.

RNA isolation and Northern blotting

Total cellular RNA was isolated with RNeasy silica resincolumns (Qiagen). RNA samples were resolved on formalde-hyde-agarose gels and transferred to Nytran membranes witha Turboblotter (Schleicher and Schuell). Hybridizations wereperformed with a DIG-UTP-labeled riboprobe synthesizedfrom a plasmid (KS7; Stratagene) containing a 456 bpBamHI±HindIII restriction fragment obtained from thehuman EGFR cDNA (nucleotides +1348 to +1804; Ullrichet al., 1984). Speci®c hybridization of the EGFR riboprobewas detected using alkaline-phosphatase-conjugated, anti-DIG antibodies and CDP-STAR chemiluminescent substrate(Bio-Rad) according to the manufacturer's instructions.

Reverse transcription-polymerase chain reaction analysis ofEGFR mRNA

RT reactions were performed with total cellular RNA, oligodT, and MMLV reverse transcriptase (Perkin-Elmer). RTreactions were examined for EGFR cDNA by PCR usingseveral ODN pairs. ODN were prepared on an AppliedBiosystems DNA synthesizer in the Biopolymer Facility atRoswell Park Cancer Institute. PCR fragments weregenerated using a mixture of 1.5 units of Taq polymerase(Life Technologies) and 0.075 units Pfu polymerase (Strata-gene). Reactions were incubated in a thermal cycler (PerkinElmer) at 948C for 30 s, 608C for 30 s and 728C for 6 min,for 30 cycles. Reaction products were resolved on 1%agarose gels and eluted using Qiaquick gel elution columns(Qiagen). Isolated fragments were tailed with Taq polymeraseand dATP and introduced into the pCRII-Topo1 vector(Invitrogen). Sequencing was performed using automatedmethods.

Western blotting and immunoprecipitation

Cells were maintained for 24 h in Optimem medium (Gibco)containing essential nutrients and attachment factors, butwithout EGF or TGFa. Next, fresh medium with EGF(100 ng/ml) was added. Cells were lysed in RIPA bu�er(50 mM Tris-HCl, pH=7,5, 0.25% sodium deoxycholate, 1%NP-40, 1 mM sodium orthovandate, 1 mM sodium ¯uoride,1 mM EGTA, 1 mg/ml each of pepstatin, leupeptin, andaprotinin). Lysates were sonicated for 30 s and centrifuged toremove debris. Total protein was measured using thebicinchoninic acid method (Sigma). Samples (50 mg protein)were boiled 5 min and applied to 5% SDS-polyacrylamidegels. Proteins were blotted onto Nytran membranes (Schlei-cher and Schuell) and then probed with anti-EGFR antibody1005 (Santa-Cruz). Speci®c binding was detected with CDP-STAR chemiluminescent substrate (Bio-Rad) and secondaryantibodies conjugated to alkaline phosphatase. Phosphory-lated EGFR was detected by immunoprecipitation with anti-phosphotyrosine antibody (PY99; Santa Cruz) and protein-Aagarose beads. Immunoprecipitates were washed with RIPAbu�er, boiled in Laemmli bu�er, run on 5% SDS-

polyacrylamide gels and detected by Western blotting withanti-EGFR antibody as described above.

Construction of expression vectors and transfection of NR6 cells

Expression vectors containing the full-length tandem duplica-tion mutants were created from TOPO TA (Invitrogen)cloned RT±PCR products derived from A-172 and A-1235cDNA. Fragments cloned into pCR1II-TOPO (Invitrogen)were sequenced using automated methods. RT±PCR usingODN from +3016 and +3819 generated fragments encom-passing the tandem duplications of the TK/CAIN regions inboth mutant genes, as well as the regulatory domain. Partialdigestion of these PCR fragments was performed with NsiI at+3149 along with complete digestion with XhoI at a 3'-engineered site produced by the +3819 ODN (Figure 2). Theresulting fragments were introduced into the NsiI and XhoIsites of the plasmid pHER.1 (a gift of Dr Hsing-Jien Kung)containing the wild type human EGFR cDNA. The resultingfull-length tandem duplication mutant cDNA (excluding the3'-untranslated region from pHER.1) was excised withEcoRV and XhoI and inserted into the HpaI and XhoI sitesof pLXIN (Clontech; a MMLV 5'-LTR-driven IRESbicistronic vector with neomycin resistance). A comparablewild type EGFR expression vector was constructed bycomplete digestion with NsiI and XhoI of a RT±PCRfragment from A-172 (+3016 to +3819) to exclude the 3'-untranslated region. The fragment was introduced into theNsiI and XhoI sites of pHER.1 and the full-length wild typemolecule was excised with EcoRV and XhoI and cloned intopLXIN. NR6 cells were transfected with pLXIN-basedvectors using LipofectAMINETM Reagent (Gibco) accordingto the manufacturer's instructions. Resistant cells wereselected with G418 (600 mg/ml) and individual clones wereisolated with cloning discs (Bel-Art Products). Clones werescreened by Western blot with anti-EGFR antibody asdescribed above.

Soft agar assays and tumor formation in nude mice

NR6 cell lines with stable EGFR expression were maintainedin DMEM with 10% serum and suspended in 0.35% lowmelting agarose at a density of 16105 cells per 35 mm well,with or without TGFa (40 ng/ml). After 14 days, colonieswere stained with crystal violet and photographed. NR6 cells(16107) cultured in fresh DMEM with 10% serum wereinjected subcutaneously into the ¯anks of nude mice. Tumorvolumes (V) were determined using the formula: V=p(D16D26D3)/6, where D1, D2 and D3 represent perpendi-cular tumor diameters.

Radioreceptor and ligand internalization assays

Cells for radioreceptor assays were plated at a density of56104 cells/well on 24-well plates with DMEM containing10% serum and grown overnight. DMEM was then replacedwith Optimem medium for 24 h. All subsequent steps wereperformed at 48C. Cells were washed twice in binding bu�erconsisting of DMEM with 0.2% bovine serum albumin(BSA). Various concentrations of unlabeled EGF (Sigma)were added in 0.2 ml binding bu�er (triplicate wells) andincubated for 2 h. Cells were washed three times with bindingbu�er and incubated with 125I-EGF (0.048 mM; speci®cactivity=1185 Ci/mmole; Amersham Pharmacia Biotech) in0.2 ml binding bu�er for 2 h. Cells were washed with bindingbu�er and solubilized with 1% Triton X-100 and BSA 1 mg/ml in phosphate-bu�ered saline (PBS), pH 7.4. Radioactivitywas measured with a gamma counter (Packard Cobra II).Cells used for ligand internalization were incubated in pre-chilled binding bu�er containing 125I-EGF (0.032 nM) for 1 hat 48C, then washed twice and incubated in binding bu�er atambient temperature. At each time point, cells were treated

Oncogene

EGFR tandem duplication mutantsMJ Ciesielski and RA Fenstermaker

819

twice for 5 min with PBS containing 0.01% MgCl2 and0.01% CaCl2, (pH 2.0). Acid-washed fractions were pooledfor counting and cell lysates were prepared with 1 N NaOH.

AcknowledgmentsThis project was supported by NIH grant CA 16056-22,The Troup Foundation and the Je�rey Wright Fund.

References

Batra SK, Castelino-Prabhu S, Wikstrand CJ, Zhu X,Humphrey PA, Friedman HS and Bigner DD. (1995).Cell Growth Di�., 6, 1251 ± 1259.

Bigner SH, Humphrey PA, Wong AJ, Vogelstein B, Mark J,Friedman HS and Bigner DD. (1990). Cancer Res., 50,8017 ± 8022.

Blasband AJ, Gilligan DM, Winchell LF, Wong ST,Luetteke NC, Rogers KT and Lee DC. (1990). Oncogene,5, 1213 ± 1221.

Callaghan T, Anctczak M, Flickinger T, Raines M, Myers Mand Kung H-J. (1993). Oncogene, 8, 2939 ± 2948.

Chen WS, Lazar CS, Lund KA, Welsh JB, Chang CP,Walton GM, Der CJ, Wiley HS, Gill GN and RosenfeldMG. (1989). Cell, 59, 33 ± 43.

Chu CT, Everiss KD, Wikstrand CJ, Batra SK, Kung H-Jand Bigner DD. (1997). Biochem. J., 324, 855 ± 861.

Di Fiore PP, Pierce JH, Fleming TP, Hazan R, Ullrich A,King CR, Schlessinger J and Aaronson SA. (1987). Cell,51, 1063 ± 1070.

Ekstrand AJ, James CD, Cavenee WK, Seliger B,Petterson RF and Collins VP. (1991). Cancer Res., 51,2164 ± 2172.

Etienne MC, Formento JL, Lebrun-Frenay C, Gioanni J,Chatel M, Paquis P, Bernard C, Courdi A, Bensadoun RJ,Pignol JP, Francoual M, Grellier P, Frenay M and MilanoG. (1998). Clin. Cancer Res., 4, 2383 ± 2390.

Fenstermaker RA, Ciesielski MJ and Castiglia GJ. (1998).Oncogene, 16, 3435 ± 3443.

Frisa PS, Walter EI, Ling L, Kung H-J and Jacobberger JW.(1996). Cell Growth Di�., 7, 223 ± 233.

Haley JD, Hsuan JJ and Water®eld MD. (1990). Oncogene,4, 273 ± 283.

Huang H-J Su, Nagane M, Klingbeil K, Lin H,Nishikawa R, Xiang-Dong J, Huang C-M, Gill GN,Wiley HS and Cavenee WK. (1997). J. Biol. Chem.,272, 2927 ± 2935.

Humphrey PA, Wong AJ, Vogelstein B, Zalutsky MR, FullerGN, Archer GE, Friedman HS, Kwatra MM, Bigner SHand Bigner DD. (1990). Proc. Natl. Acad. Sci. USA, 87,4207 ± 4211.

Humphrey PA, Wong AJ, Vogelstein B, Friedman HS,Werner MH, Bigner DD and Bigner SH. (1988). CancerRes., 48, 2231 ± 2238.

Humphrey PA, Gangarosa LM, Wong AJ, Archer GE,Lund-Johansen M, Bjerkvig R, Laerum O-D, FriedmanHS and Bigner DD. (1991). Biochem. Res. Comm., 178,1413 ± 1420.

Lehvaslaiho J, Lehtola L, Sistonen L and Alitalo KA. (1989).EMBO J., 8, 159 ± 166.

Libermann TA, Nusbaum HR, Razon N, Kris R, Lax I,Soreq H, Whittle N, Water®eld MD, Ullrich A andSchlessinger J. (1985). Nature (London), 313, 144 ± 147.

Lund-Johansen M, Bjerkvig R, Humphrey PA, Bigner SH,Bigner D and Laerum O-D. (1990). Cancer Res., 50,6039 ± 6044.

Maruno M, Kovach JS, Kelly PJ and Yanagihara T. (1991).J. Neurosurg., 75, 97 ± 102.

Montgomery RB, Moscatello DK, Wong AJ, Cooper JA andStahl WL. (1995). J. Biol. Chem., 270, 30562 ± 30566.

Moscatello DK, Montgomery RB, Sundareshan P, McDanelH, Wong MY andWong AJ. (1996). Oncogene, 13, 85 ± 96.

Munson PJ and Rodbard D. (1980). Anal. Biochem., 107,220 ± 239.

Nagane M, Coufal F, Lin H, Bogler O, Cavenee WK andHuang H-J. (1996). Cancer Res., 56, 5079 ± 5086.

Nagane M, Levitzki A, Gazit A, Cavenee WK and HuangHJ. (1998). Proc. Natl. Acad. Sci. USA, 95, 5724 ± 5729.

Nishikawa R, Xiang-Dong J, Harmon RC, Lazar CS, GillGN, Cavenee WK and Huang HJ. (1994). Proc. Natl.Acad. Sci. USA, 91, 7727 ± 7731.

O'Rourke DM, Qian X, Zhang H-T, Davis JG, Nute E,Meinkoth J and Greene MI. (1997). Proc. Natl. Acad. Sci.USA, 94, 3250 ± 3255.

Panneerselvam K, Kanakaraj P, Raj S, Das M and BishayeeS. (1995). Eur. J. Biochem., 230, 951 ± 957.

Prywes R, Livneh E, Ullrich A and Schlessinger J. (1986).EMBO J., 5, 2179 ± 2190.

Roth DB and Craig NL. (1998). Cell, 94, 411 ± 414.Schlegel J, Merdes A, Stumm G, Albert FK, Forsting M,

Hynes N and Kiessling M. (1994). Int. J. Cancer, 56, 72 ±77.

Steck PA, Lee P, Hung M-C and Yung WKA. (1988). CancerRes., 48, 5433 ± 5439.

Sugawa N, Ekstrand AJ, James CD and Collins VP. (1990).Proc. Natl. Acad. Sci. USA, 87, 8602 ± 8606.

Ullrich A, Coussens L, Hay¯ick JS, Dull TJ, Gray A, TamAW, Lee J, Yarden Y, Liberman TA, Schlessinger J,Downward J, Mayes ELV, Whittle N, Water®eld MD andSeeburg PH. (1984). Nature (London), 309, 418 ± 425.

Wells A, Welsh JB, Lazar CS, Wiley HS, Gill GN andRosen®eld MG. (1990). Science, 247, 962 ± 964.

Wong AJ, Bigner SH, Bigner DD, Kinzler KW, Hamilton SRand Vogelstein B. (1987). Proc. Natl. Acad. Sci. USA, 84,6899 ± 6903.

Wong AJ, Ruppert JM, Bigner SH, Grzeschik CH,Humphrey PA, Bigner DS and Vogelstein B. (1992). Proc.Natl. Acad. Sci. USA, 89, 2965 ± 2969.

EGFR tandem duplication mutantsMJ Ciesielski and RA Fenstermaker

820

Oncogene