Report

Gain-of-Function Mutations of c-kit in Human Gastrointestinal Stromal Tumors

See allHide authors and affiliations

Science  23 Jan 1998:
Vol. 279, Issue 5350, pp. 577-580
DOI: 10.1126/science.279.5350.577

Abstract

Gastrointestinal stromal tumors (GISTs) are the most common mesenchymal tumors in the human digestive tract, but their molecular etiology and cellular origin are unknown. Sequencing of c-kit complementary DNA, which encodes a proto-oncogenic receptor tyrosine kinase (KIT), from five GISTs revealed mutations in the region between the transmembrane and tyrosine kinase domains. All of the corresponding mutant KIT proteins were constitutively activated without the KIT ligand, stem cell factor (SCF). Stable transfection of the mutant c-kitcomplementary DNAs induced malignant transformation of Ba/F3 murine lymphoid cells, suggesting that the mutations contribute to tumor development. GISTs may originate from the interstitial cells of Cajal (ICCs) because the development of ICCs is dependent on the SCF-KIT interaction and because, like GISTs, these cells express both KIT and CD34.

The c-kit proto-oncogene encodes a type III receptor tyrosine kinase (KIT) (1), the ligand of which is SCF (2). SCF-KIT interaction is essential for development of melanocytes, erythrocytes, germ cells, mast cells and ICCs (3, 4). Gain-of-function mutations of the c-kit gene have been found in several tumor mast cell lines of rodents and humans (5, 6) and in mast cell tumors of humans (7). Here we investigate the mutational status of c-kit in mesenchymal tumors of the human gastrointestinal (GI) tract.

We collected 58 mesenchymal tumors that developed in the GI wall (4 in the esophagus, 36 in the stomach, 14 in the small intestine, and 4 in the large intestine). KIT expression was examined by immunohistochemistry (8). Eight authentic leiomyomas and an authentic schwannoma did not express KIT. The remaining 49 mesenchymal tumors were diagnosed as gastrointestinal stromal tumors (GISTs), and 94% (46/49) of these expressed KIT. Examination of these tumors for expression of CD34, which is a reliable marker for GISTs (9), revealed that 82% (40/49) were CD34-positive, and 78% (38/49) were positive for both KIT and CD34 (Fig.1, A to I). Three of five KIT-negative GISTs were also CD34-negative.

Figure 1

Coexpression of KIT and CD34 in human GISTs. (A to C) Serial sections of a tumor that developed in the muscle layers of the stomach. (D toF) Higher magnification of (A) to (C). (A) and (D) are stained with hematoxylin and eosin. The boundary of the tumor is indicated by arrows in (A). (B) and (E) are stained with anti-KIT and (C) and (F) with anti-CD34 (8). Arrowheads in (A) to (C) indicate that endothelial cells of blood vessels express CD34 but not KIT. lm; longitudinal muscle layer. (Gto I) Coexpression of KIT and CD34 in another GIST, also from the stomach, demonstrated by confocal laser scanning microscopy (8). (G) Binding of rabbit anti-KIT to tumor cells demonstrated by FITC-labeled anti-rabbit IgG (green). (H) Binding of mouse anti-CD34 to tumor cells demonstrated by RPE-labeled anti-mouse IgG (red). (I) Merged confocal image of (G) and (H) showing coexpression of KIT and CD34 in tumor cells (yellow). Bars in (A) to (C), 50 μm; in (D) to (I), 100 μm.

We compared the immunohistochemical characteristics of GISTs with those of ICCs, cells that regulate autonomous contraction of the GI tract (4). ICCs are located in and near the circular muscle layer of the stomach (10), small intestine (11), and large intestine (12). Because ICCs of the small intestine surrounding myenteric ganglion cells are easily identified by their specific localization, we examined the immunohistochemical characteristics of these cells and found that they were double-positive for KIT and CD34 (Fig. 2, A to G). ICCs in the circular muscle layer of the stomach and small intestine and ICCs in the myenteric plexus region and circular muscle layer of the large intestine were also double-positive for KIT and CD34.

Figure 2

Coexpression of KIT and CD34 in ICCs surrounding myenteric ganglion cells in normal human small intestine. (A and B) Serial sections of the normal human small intestine. (C and D) Higher magnification of (A) and (B), respectively. (A) and (C) are stained with anti-KIT and (B) and (D) are stained with anti-CD34 (8). The localization of KIT and CD34 double-positive cells is consistent with that of ICCs, which are present between the circular muscle layer (cm) and the longitudinal muscle layer (lm), and surrounding the myenteric ganglion cells (gc) (11). (E to G) Demonstration by confocal laser scanning microscopy (8). (E) Binding of rabbit anti-KIT. (F) Binding of mouse anti-CD34. (G) Merged confocal image of (E) and (F). Arrows in (F) and (G) show endothelial cells and fibroblast-like cells that express CD34 but not KIT. Bars in (A) to (G), 50 μm.

We obtained the complete coding region of c-kit cDNA from six GISTs and control tissues using the reverse transcriptase–polymerase chain reaction (RT-PCR) (13). Ten independent c-kit clones were obtained from each sample. In 5 of the 6 GISTs (GIST 1 to 5), 4 to 6 clones out of 10 examined showed mutations in the region between the transmembrane and tyrosine kinase domains (hereafter called juxtamembrane domain) (Fig. 3). These mutations were located within an 11–amino acid stretch (Lys-550 to Val-560), but at nonidentical sites. No mutations were detectable in other domains of c-kit cDNA. Because ∼50% of the cDNA clones from each GIST did not show any mutations, we conclude that only one of the two c-kit alleles was mutated in each case. Direct sequencing of the PCR products confirmed the mutations in the juxtamembrane domain of c-kit cDNA in the five GISTs.

Figure 3

Mutations of c-kitin GISTs. GIST 1 showed an in-frame deletion of 6 base pairs (bp). GIST 2 showed an in-frame deletion of 15 bp. GIST 3 showed the same in-frame deletion as observed in GIST 2 and an additional point mutation at codon 550 (AAA to ATA) that resulted in a Lys550→Ile substitution. GIST 4 showed a point mutation at codon 559 (GTT to GAT) that resulted in a Val559→Asp substitution. GIST 5 showed an in-frame deletion of 27 bp. The c-kit mutations in the juxtamembrane domain of the HMC-1 human mast cell leukemia cell line (5) and the FMA3 murine mastocytoma cell line (6) are shown for comparison. Deleted amino acids are shown by dashes (–) and mutated amino acids by boxes. Murine and human KIT are of different lengths (1), so the amino acid numbering in the FMA3 KIT is different. Abbreviations used are as follows: SP, signal peptide; EC, extracellular domain; TM, transmembrane domain; JM, juxtamembrane domain; TK1 and TK2, tyrosine kinase domains; and KI, kinase insert. Abbreviations for the amino acid residues are as follows: D, Asp; E, Glu; G, Gly; H, His; I, Ile; K, Lys; L, Leu; M, Met; N, Asn; P, Pro; Q, Gln; T, Thr; V, Val; W, Trp; and Y, Tyr.

We next examined whether the c-kit mutations found in the GISTs resulted in constitutive activation of the c-kit receptor tyrosine kinase by transient introduction of the mutant c-kit cDNAs into the 293T human embryonic kidney (HEK) cell line (5, 14, 15). The wild-type c-kit cDNA was introduced as a negative control, and the tyrosine kinase domain mutant and the juxtamembrane domain mutant found in the HMC-1 human mast cell leukemia cell line (5) were introduced as positive controls. Wild-type KIT was phosphorylated on tyrosine only when recombinant human (rh) SCF was added to the culture medium (Fig.4A). In contrast, the gain-of-function KIT mutants found in HMC-1 cells were phosphorylated on tyrosine without the addition of rhSCF, as reported previously (5). The magnitude of the constitutive tyrosine phosphorylation was greater in the tyrosine kinase domain mutant than in the juxtamembrane domain mutant. The c-kitmutants found in GISTs also showed the constitutive tyrosine phosphorylation in 293T cells without rhSCF (Fig. 4A). The constitutive tyrosine phosphorylation of the juxtamembrane mutant of HMC-1 cells was of similar magnitude to that of the juxtamembrane mutants of GISTs. In in vitro kinase assays (16), the c-kit mutants found in the GISTs exhibited constitutive kinase activation that was similar in magnitude to that of the juxtamembrane domain mutant of HMC-1 cells (5) (Fig. 4B).

Figure 4

Constitutive activation of the mutant KIT found in GISTs. The mutant c-kit cDNAs found in five GISTs were transfected into 293T cells. Human wild-type c-kit cDNA and mutant c-kit cDNAs found in the HMC-1 cells (6, 21) were also transfected as controls. (A) Tyrosine phosphorylation was examined with or without rhSCF stimulation (15). (B) Immune complex kinase assay carried out without rhSCF stimulation (16).

To investigate the biological consequences of the mutant c-kit, we introduced the c-kit mutations found in the GISTs into the mouse c-kit cDNA (17) and then stably transfected the cDNA into the interleukin 3 (IL-3)–dependent Ba/F3 murine lymphoid cell line (18). As a control, mouse wild-type c-kit cDNA was also transfected into Ba/F3 cells. We estimated Ba/F3 cell proliferation using an MTT colorimetric assay (19). Ba/F3 cells with the wild-type murine c-kitgrew in the presence of either recombinant mouse (rm) IL-3 or rmSCF; Ba/F3 cells with the mutated murine c-kit grew autonomously without rmIL-3 and rmSCF (Fig. 5, A and B). Ba/F3 cells with the mutated murine c-kit also grew autonomously in nude mice (Fig. 5C) (20). The constitutive kinase activation of all KIT mutants found in the five GISTs was confirmed in Ba/F3 cells (Fig. 5D) (5, 21).

Figure 5

Autonomous proliferation of Ba/F3 cells transfected with mutated murine c-kit in culture and in nude mice. (A and B) MTT colorimetric assay (19) in the presence of rmIL-3 (A) or rmSCF (B). Data are expressed as the mean of four wells. Untransfected Ba/F3 cells (□), Ba/F3 cells with the murine wild-type c-kit (○), and Ba/F3 cells with the murine mutated c-kit [GIST 1 (▴), GIST 2 (•), GIST 3 (▾), GIST 4 (▪), and GIST 5 (x). (C) Development of tumors in nude mice (20) after injection of the Ba/F3 cells with each mutated c-kit. The original Ba/F3 cells and Ba/F3 cells with murine wild-type c-kit did not form tumors. Data are expressed as the mean of results in five mice. (D) Immune complex kinase assay (16). Ba/F3 cells with various c-kit cDNAs were cultured without rmSCF.

Although various cells including hematopoietic stem cells express both KIT and CD34 (22), ICCs are the only cells that are double-positive for KIT and CD34 in normal GI wall of humans. This strongly suggests that KIT and CD34 double-positive GISTs might originate from ICCs, although we cannot exclude the possibility that ICCs and GISTs simply show common undifferentiated characteristics such as those observed in multipotential hematopoietic stem cells.

The mechanism by which KIT becomes constitutively activated appears to be different for the tyrosine kinase domain mutant and the juxtamembrane domain mutant (6, 21). The former is constitutively activated without forming dimers (21), whereas the latter constitutively dimerizes without binding SCF (6, 21). The tyrosine kinase domain mutation of KIT has been found only in mast cell neoplasms (7) and its juxtamembrane domain mutation only in GISTs. The mechanisms by which these different mutations cause malignant transformation of different cell types remain to be investigated.

  • * These authors contributed equally to this work.

  • To whom correspondence should be addressed. E-mail: kitamura{at}patho.med.osaka-u.ac.jp

REFERENCES AND NOTES

View Abstract

Navigate This Article