High CD133 expression levels in gastrointestinal stromal tumors


  • How to cite this article: Bozzi F, Conca E, Manenti G, Negri T, Brich S, Gronchi A, Pierotti MA, Tamborini E, Pilotti S. High CD133 expression levels in gastrointestinal stromal tumors. Cytometry Part B 2011; 80B: 238–247.



Gastrointestinal stromal tumours (GISTs) have activating KIT or PDGFRA gene mutations. Imatinib mesylate, which targets KIT and PDGFRA, is effective in treating GISTs, but 90% of GIST patients become imatinib-resistant as a result of acquiring secondary KIT mutations. Recent findings suggest that tumour growth can be driven by mutated self-renewing progenitors known as cancer stem cells (CSCs), which are believed to be present in all neoplastic proliferations and are thought to accumulate mutations. It is therefore possible that the acquisition of secondary KIT mutations during imatinib treatment may occur in putative GIST CSCs.


Using flow cytometry, in vivo murine xenografts and molecular characterization, we tried to identify putative GIST CSCs by looking for the occurrence of common CSC markers such as KIT, CD133, CD90, CD44, and CD34 in 18 surgical samples obtained from nine untreated and nine imatinib-treated KIT-mutated GIST patients.


The results indicated the homogeneous and previously unreported expression of CD133 (18/18), CD90 (15/16), and CD44 (12/14), together with KIT (18/18) and CD34 (13/18). This profile is similar to that identified in bone marrow mesenchymal progenitors and does not seem to be significantly modified by imatinib as only marginal changes in KIT and CD133 expression (P ≤0.05, Mann-Whitney test) were found in the treated samples.


These findings suggest that GISTs are a clonal expansion of quite primitive cells that strictly depend on KIT oncogenic addiction, and have no cancer/stem cell component that can be detected by means of the antigens used in this study. © 2011 International Clinical Cytometry Society

Gastrointestinal stromal tumors (GISTs) are the most frequent mesenchymal malignancies of the gastrointestinal tract and are characterized by the presence of the constitutively activated tyrosine kinases (RTKs) KIT (CD117) or platelet derived growth factor receptor alpha (PDGFRα) in respectively, 80 and 5% of cases (1). The mechanism responsible for the constitutive activation of the two receptors is the presence of gain of function mutations in the corresponding genes, and it has recently been suggested that insulin-like growth factor receptor (IGF-IR) activation is the main oncogenetic event in the remaining cases (2).

Imatinib mesylate (Gleevec; Novartis, Basle, Switzerland), a KIT and PDGFR tyrosine kinase inhibitor, blocks the activation of both receptors in vitro, and induces tumor control in ≥90% of patients with advanced GISTs. Activating KIT mutations involving tyrosine kinase (TK) domains and imatinib-insensitive mutations characterize other human disorders such as mastocytosis (3) and seminomas (4).

Approximately 60–70% of GISTs arise in the stomach, 20–30% in the small bowel, and 5% in the colon and rectum. Primary c-kit Exon 11 mutations are mainly found in gastric GISTs, whereas GISTs located in the small bowel are characterized by primary c-kit Exon 9 mutations. In keeping with these different genotypes, there are some differences in gene expression between stomach and small bowel GISTs (5).

Primary c-kit Exon 11 mutations respond better to imatinib than the other reported mutations (Exons 9, 13, 14, and 17) or the wild-type (1, 6), but 90% of the patients with c-kit Exon 11 mutations become imatinib resistant after long-term treatment, mainly because of gains in secondary c-kit mutations in exons encoding for the imatinib binding pocket domains (Exons 13, 14, or 17) (7). Other mechanisms of secondary resistance include the activation of alternative RTKs (8), BRAF mutations (9), neurofibromin loss-mediated RAS up-regulation in NF1 syndrome (10), GISTs with rhabdomyosarcomatous transdifferentiation and null KIT (11), and c-kit gene amplification/loss (12, 13).

The primitive nature of GISTs first suggested by its hallmark (i.e., the expression of KIT in addition to CD34) has been recently supported by the discovery of rare cells expressing KIT together with CD34 and CD44 in post-natal mouse gastric smooth muscle layers, which may be the local progenitors of both Cajal interstitial cells and GISTs (14).

The cancer stem cell (CSC) hypothesis suggests that only a small sub-population of tumor cells are self-renewing and capable of maintaining tumor growth (15). It is also believed that CSCs are present in all neoplastic proliferations and accumulate mutations. It is therefore possible that the acquisition of secondary c-kit mutations during imatinib treatment may occur in the putative CSCs of GISTs.

We used flow cytometry (FC), murine xenografts and molecular biology to investigate the profile of GIST tumoral cells in a series of untreated c-kit Exon 11-mutated and wild-type GISTs in a search for cells expressing antigens that are generally recognized as markers of CSCs (15, 16), including CD133, CD90, and CD44. We also analyzed the two well-established GIST markers (KIT and CD34) which are expressed in hematopietic progenitor cells too. Finally, on the basis of the assumption that imatinib selective pressure acts by eliminating sensitive cells and inducing the formation of resistant clones, we analyzed CSC markers in tumor samples taken from imatinib-treated patients who were molecularly characterized in terms of secondary/resistant mutations.



The case materials were 18 surgical samples obtained from nine untreated patients (17) and nine patients who underwent surgical resection after imatinib treatment (18); at the time of surgery, six of these patients were clinical responders and three had progressive disease. The histological diagnosis of the GISTs was confirmed using antibodies against KIT (CD117), CD34, PDGFRA, desmin, 1A4, and S100, as previously described (19). Informed written consent for the use of biological materials was obtained from all of the subjects upon admission.

Clinical Findings

Table 1 shows the tumor locations, the tumor-node-metastasis status, the dose and duration of imatinib treatment, and the patients' clinical status and molecular characteristics. Briefly, seven of the untreated patients carried a c-kit mutation in Exon 11 (no. 1, 2, 3, 4, 5, 6, and 7) and two had wild-type c-kit and PDGFRA genes (no. 8 and 9); the six imatinib-treated and clinically responding patients (no. 10, 11, 12, 13, 14, and 15) had only c-kit Exon 11 mutations, whereas the three patients with progressive disease included one with a secondary Exon 13 mutation (no. 16), and two (no. 17 and 18) with secondary Exon 17 mutations (coupled with a primary c-kit Exon 11 mutation).

Table 1. Tumor-Node-Metastasis Status, Imatinib Dose and Treatment Duration, Clinical Response, and c-kit Mutations in Untreated and Imatinib-Treated GISTs Patients
SamplesTumor siteDose of imatinib (mg daily−1)Duration on imatinib (months)Clinical responsePrimary c-kit mutationsSecondary c-kit mutations
  1. P: primary, M metastasis, Res: responding, Pr: progressing, Ex: exon, del: deletion, Wt: wild type, dup: duplication.

Untreated samples 
No. 1Stomach///Ex11: del K558 and I563/
No. 2Ileal + peritoneal M///Ex11: V560D/
No. 3Rectal P///Ex11: del W557 and K558/
No. 4Stomach///Ex11: del M552-V555/
No. 5Stomach P///Ex11: del W557 and K558/
No. 6Duodenal P///Ex11: del V560/
No. 7Duodenal P///Ex11: V560A/
No. 8Duodenal P///Wt/
No. 9Stomach///Wt/
Imatinib treated samples with c-kit exon 11primary mutations
No. 10Duodenal P4006ResEx11: V559D/
No. 11Rectal P4006ResEx11: V560D/
No. 12Stomach P4003ResEx 11 del P551-V555/
No. 13Stomach P4009ResEx11: del W557 and K558/
No. 14Stomach P4006ResEx11: del V557and K558/
No. 15Oesophagus P4009ResEx 11 del E554-K558/
Imatinib treated samples with c-kit primary and secondary mutations
No. 16Stomach + liver M40011PrEx 11: del G556-V560Ex13: V564A
No. 17Peritoneal M400 + 80039 (33+6)PrEx11: dup W577-R586Ex17: N822K
No. 18Peritoneal M40014PrEx11:V559AEx17: D816H


To ascertain the surface localization of the wild-type and mutant KIT proteins, the NIH3T3 cell line was stably transfected with human wild-type KIT, or the human Δ559 (the deletion of KIT Exon 11 V559, a common imatinib-sensitive primary mutation) or T670I KIT mutations (an imatinib-resistant KIT Exon 14 mutation).

NIH3T3 Transfection

The NIH3T3 cells were stably transfected and maintained in a 100 × 200 mm2 tissue culture dish (Cat. No. 35303, Becton Dikinson, Buccinasco, Italy) as previously described (20). When sub-confluence was reached (1 × 106 cells), the cells were detached using a 1% trypsin-EDTA solution (Cat. No. 15400, Invitrogen, San Giuliano Milanese, Italy) and underwent FC.

Flow Cytometry

The cells obtained after trypsin-EDTA detachment were washed twice in PBS (Cat. No. 1001003, Invitrogen), counted, and diluted to an appropriate concentration for antibody incubation. Briefly, 1 × 105 cells diluted in 50 μL of PBS plus BSA 1% (Cat. No. 340486, Roche, Milan, Italy) were placed in a 12 × 75 mm2 round-bottom polystyrene tube (Becton Dickinson), and incubated for 30 min at 4°C with KIT phycoerythrin (PE) (clone 130-090-853, Cat. No. PN IM1360U, Beckman Coulter, Monza, Italy). After antibody incubation, the cells were washed and resuspended in 300 μl of PBS. At least 5 × 104 cells were analyzed using a FACS-Scalibur flow cytometer (Becton Dickinson).

Protein Extraction and Immunoprecipitation/Western Blotting

The proteins were extracted, immunoprecipitated for KIT, and blotted as previously described (19).

GIST Surgical Samples

Sample selection criteria

A hematoxylin and eosin-stained frozen section of each untreated and imatinib-treated GIST was used to check tumor cellularity. Only the imatinib-treated GISTs with a low (i.e., 50–90% of viable tumoral cells) or moderate response (10–50% of viable tumoral cells) were considered (13).

Sample disaggregation

A single cell suspension was obtained from the surgical samples used to define the histological response by means of three hours of collagenase treatment (Cat. No. C6885, Sigma, Milan, Italy), after which the cells were filtered through a 45-μl nylon mesh, washed, and diluted with PBS/BSA 1% to an appropriate concentration for subsequent use.

Flow cytometry

To ensure the consistency of instrument set-up over time, the flow cytometer was calibrated daily using Calibrite fluorescent microbeads and Autocomp software (Cat. No. 340486, Becton Dickinson) as recommended by the manufacturer. Briefly, the instrument set-up is designed to maintain the unlabelled bead population within the first log decade of the scale and keep the separation of the labeled and unlabelled bead populations well within the accepted variability (typically less than 2.5 channels per month for the three fluorescence detectors).

The FC analysis was made using single cell suspensions obtained from the surgical samples and, in each case, fluorescence intensity was adjusted by means of isotypical FITC/PE controls (FITC-PE IgG1 isotype, clone MOPC-21, Cat. Nos. 555748 and 555749, Becton Dickinson); the nucleic acid dye 7-amino-actinomycin D (7AAD, Cat. No. 559925, Pharmingen, Buccinasco, Italy) was used to distinguish live (7AAD−) and dead cells (7AAD+). The immunological gate used to identify live GIST cells in the tumor suspensions (the R1 region of Fig. 1) was KIT PE+ and 7AAD−. All of the 7AAD+ cells were captured in the R2 region and eliminated from the analysis, and the KIT PE+ and 7AAD− cells were subsequently combined in a triple stain with CD45 (clone 2D1, Cat. No. 345808, Becton Dickinson), CD34 (clone 8G12, Cat. No. 345801, Becton Dickinson), CD44 (clone L178, Cat. No. 347943, Becton Dickinson), and CD90 (clone 5E10, Cat. No. 555595, Pharmingen) fluorescein isothiocyanate (FITC) conjugates. As both KIT and CD133 (clone 293C3, Cat. No. 130-090-853, Milteny Biotech, Calderara di Reno, Italy) were PE-conjugated, CD34 FITC (or CD90 FITC) was used as a secondary immunological gate to evaluate the CD133 expression of the CD34+/7AAD− (or CD90+/7AADv) cells (the R3 region of Fig. 1). Tumor infiltrating lymphocytes were used as internal negative antibody-staining controls.

Figure 1.

Gating strategy for identifying GIST cells in tumor suspensions. (A) Live GIST cells in the tumor suspensions (obtained from patient no. 9 in Table 1) were identified using the KIT+ and 7AAD− immunological gate (the R1 region, black); all of the 7AAD+ cells were captured in the R2 region (yellow), infiltrating KIT-/7AAD− cells were captured in the R3 region (blue). KIT−/CD45+ cells were considered lymphocytes (B); KIT−/CD90+ (C) and KIT−/CD44+ (D) cells were considered stromal cells; and KIT−/CD34+ cells were considered endothelial cells (E). As both KIT and CD133 were PE-conjugated, CD34 FITC 7AAD− (the R1 region in (F) was used as a secondary immunological gate to evaluate CD133 expression (G) of GIST cells.

Fifty microlitres (1 × 105 to 2 × 105 cells) of the single cell suspensions were placed in a 12 × 75 polystyrene tube, and incubated for 30 min at 4°C with the following antibody combinations: (i) isotype-matched controls to set the FITC-PE backgrounds; (ii) CD34FITC alone, KIT PE alone, and 7AAD alone to set the sample fluorescence compensation, (iii) CD45 FITC, KIT PE, and 7AAD to define live GIST cells and exclude hematopoietic cells; (iv) CD34 FITC, KIT PE, 7AAD; (v) CD44 FITC, KIT PE, 7AAD; (vi) CD90 FITC, KIT PE, 7AAD to evaluate the co-expression of CD34, CD44, and CD90 on live GIST KIT+ cells; and (vii) CD34 FITC, CD133 PE, 7AAD to evaluate CD133 expression on GIST cells. After antibody incubation, the samples were washed and resuspended in 300 μL of PBS plus BSA 1% (Roche, Milan, Italy). At least 5 × 104 cells were analyzed using a FACS-Scalibur.

Fluorescence intensity was evaluated by calculating the median fluorescence intensity (MFI) ratio obtained by dividing the geometric mean of the stained sample fluorescence histogram by the geometric mean of the respective isotype control (in triplicate experiments). The MFI ratio provides a normalized measurement of the fluorescence emission of each stained sample, regardless of histogram shape or fluorochrome type (21).

Mouse xenografts

All of the experiments were carried out using female athymic CD-1 Nu/Nu mice aged 8–11 weeks (Charles River, Calco, Italy). The experimental protocols were approved by the Ethics Committee for Animal Experimentation of the Istituto Nazionale Tumori (Milan, Italy) in accordance with the United Kingdom Coordinating Committee on Cancer Research Guidelines (22). Four primary GISTs (no. 2, 9, 11, and 13; Table 1) were aseptically minced into fragments (2 × 2 × 2 mm3) using a scalpel in RPMI medium (cat: 21875091, Invitrogen) containing penicillin/streptomycin (cat: 15140163, Invitrogen), and subcutaneously (s.c.) inoculated into the right flank of isoflurane-anesthetized mice (three mice for each tumoral specimen) using a micro-trochar. The growth of the s.c. tumors was followed by measuring their diameters every week using Vernier calipers. At the end of the observation period (6–9 months), the mice were anesthetized with isoflurane and killed by means of cervical dislocation. The tumors were excised and maintained in cold RPMI medium (Invitrogen) for subsequent characterization.

Relative quantification of c-kit, CD133, and CD90 mRNA

The RNA for molecular analysis was obtained from GIST samples immunomagnetically enriched in CD117+ cells (cat: 120-000-442, CD117 microbeads kit, Milteny Biotech) taken from seven untreated patients (no. 1, 2, 3, 4, 6, 7, and 9; Table 1), four responding imatinib-treated patients (no. 10, 11, 14, and 15; Table 1) and two imatinib-treated patients with progressive disease (no. 16 and 17; Table 1). The RNA obtained from 1 × 106 mobilized CD34+/CD133+/CD45+ hematopoietic stem cells (HPCs) purified immunomagnetically from peripheral blood (Cat. No. 120-002-250, CD133 microbeads kit, Milteny Biotech) was used as a calibrator for the relative quantification of KIT and CD133 mRNA transcription; and the RNA from cultured human bone marrow mesenchymal stem cells (BM MSCs), obtained as described by Pittenger et al. (23)., was used as a calibrator for the relative quantification of CD90 mRNA transcription. The HPCs and BM MSCs were obtained from the cryopreserved material of a patient participating in an investigative protocol (approved by our Institutional Ethics and Scientific Board) in which high-dose chemotherapy was used; informed consent for the use of the material was obtained as previously described (24). The c-kit (Hs00174029_m1 probe; Applied Biosystem), CD133 (Hs01009259_m1 probe; Applied Biosystem), CD90 (Hs00264235_m1 probe; Applied Biosystem), and beta 2 microglobulin cDNAs (Hs00187842_m1 probe; Applied Biosystem) used as endogenous controls were relatively quantified by means of real-time quantitative PCR (ABIPRISM 5700 PCR Sequence Detection Systems, Applied Biosystems) and a TaqMan-based analysis (User Bulletin No. 2, ABI prism 7700 sequence detection system, PE Applied Biosystems, December 1997) in accordance with the manufacturer's instructions. The relative changes in gene expression were calculated using the 2−ΔΔCt method.

Statistical Analysis

The data were analyzed using a non-parametric Mann-Whitney test (two-sided) and Stat View software. A P value of ≤0.05 was considered statistically significant.


Analysis of KIT Cell Localization

As it has been reported that c-kit mutations induce the intracellular retention of the receptor, which leads to its loss from the outer plasma membrane (25, 26), we first assessed whether combined FC and biochemistry can identify the surface localization of KIT on NIH3T3 cells stably transfected with human wild-type KIT, human Δ559 (a frequent primary Exon 11 mutation) or human T670I (an imatinib-resistant Exon 14 mutation). KIT surface expression was observed in all cases, but the cells tranfected with wild-type and T670I-mutated KIT had a higher MFI ratio than those transfected with Δ559-mutated KIT (Fig. 2A).

Figure 2.

KIT wild-type, Δ559 and T670I NIH 3T3 transfected cell lines: flow cytometry (FC) and immunoprecipitation/Western blot (IP/WB) analyses. (A) Membrane-localized KIT proteins (blue line; the dark lines are the isotype controls) were detected by FC in all cases. (B) After KIT-specific immunoprecipitation, and in line with the FC detection of KIT, the completely glycosylated 145 kDa form (indicated by arrows and an asterisk) was seen in all cases except for the non-transfected NIH. (C) Δ559 showed greater phosphorylation of the immature 125 kDa form than T670I. A c-kit Exon11-mutated GIST sample was used as the positive control (CTR+). The MFI ratios of KIT wt, Δ559 and T670I (expressed as mean values ± standard deviation) were respectively 68.5 ± 5.6, 32.8 ± 2.9, and 130 ± 3.8, and were obtained in triplicate experiments as described in Materials and Methods. As a loading control, equal volumes of unbound protein from the IP experiments were loaded onto a gel and hybridized with alpha actin; they showed comparable amounts of proteins along the lanes (not shown). M: marker.

The proteins extracted from wild-type and Δ559 or T670I-mutated KIT were immunoprecipitated using an anti-KIT antibody and blotted; in line with the FC results, the mature (completely glycosylated) 145 kDa form of KIT was observed in all cases (Fig. 2B). The Δ559 mutation showed a higher level of phosphorylation in the immature 125 kDa form of KIT than the T670I mutation (Fig. 2C); no phosphorylation was observed in the immature wild-type form.

Taken together, these results show that, although the mutated KIT proteins can undergo intracellular phosphorylation, they can also be completely glycosylated and exposed to the cell surface. This membrane localization justifies the use of KIT as an immunological gate for the detection of KIT+ cells in GIST samples.

GIST Surgical Samples

Untreated GISTs are characterized by a primitive profile

To identify putative GIST CSCs, the presence of the stem cell markers CD133, CD90, CD44, and CD34 was evaluated in untreated c-kit Exon 11-mutated and wild-type samples (Table 2).

Table 2. Frozen Sample Histological Evaluation and FC Characterization of the Untreated and Imatinib-Treated Samples
SamplesTumor cellularity on frozen samplesFC percentage of KIT + cellsKITCD34CD133CD90CD44
  1. Fluorescence intensity (reported in round brackets as mean values ± SD) was evaluated by calculating the median fluorescence intensity (MFI) ratio, obtained by dividing the geometric mean of the stained sample fluorescence histogram by the geometric mean of the respective isotype control (in triplicate experiments). +: FC positive sample, −: FC negative sample, Nd: not done.

Untreated samples
No. 1>50%90%+ (52.4 ± 17.2)+ (50.6 ± 1.9)+ (125 ± 5.5)+ (109 ± 11)+ (10.6 ± 3.2)
No. 2>50%50%+ (250 ± 23)-+ (157 ± 5.2)+ (80.5 ± 18.3)+ (84.5 ± 19)
No. 3>50%70%+ (58.3 ± 13.4)+ (85.5 ± 5.1)+ (55.3 ± 4.3)+ (85.7 ± 6.2)+ (72.5 ± 5.7)
No. 4>50%50%+ (19.5 ± 5.6)+ (56.6 ± 3.1)+ (137 ± 12)+ (53.2 ± 4.5)+ (23.8 ± 2.4)
No. 5>50%60%+ (43.5 ± 4.1)+ (48.5 ± 4.6)+ (130 ± 6.1)+ (75.6 ± 5.6)-
No. 6>10%<50%30%+ (269 ± 26)-+ (124 ± 17)+ (116 ± 16)+ (60.7 ± 17)
No. 7>50%70%+ (5.0 ± 0.4)-+ (70.6 ± 10)+ (74.5 ± 12.6)+ (6.6 ± 0.5)
No. 8>10%<50%10%+ (54.7 ± 6.2)+ (46.5 ± 5.7)+ (107 ± 10)+ (52.5 ± 4.5)Nd
No. 9>50%50%+ (12.2 ± 1)+ (48.8 ± 8.1)+ (155 ± 7)-+ (10.8 ± 2.2)
Imatinib-treated samples with primary c-kit mutations
No. 10>50%70%+ (37.8 ± 1.9)-+ (30.5 ± 2.1)+ (54.6 ± 7.1)+ (11 ± 1.3)
No. 11>50%70%+ (5.0 ± 0.7)+ (10.3 ± 2.4)+ (9.1 ± 2.1)+ (8.4 ± 0.7)+ (6.8 ± 0.9)
No. 12>50%60%+ (43.2 ± 3.7)+ (72.6 ± 5.5)+ (120 ± 9.4)NdNd
No. 13>10%<50%30%+ (4.2 ± 1.1)+ (54.5 ± 7.4)+ (9.7 ± 1.2)NdNd
No. 14>10% <50%35%+ (6,5±0,8)+ (228 ± 7.8)+ (59.3 ± 6.1)+ (90.3 ± 11)Nd
No. 15>10% <50%20%+ (7.2 ± 0.2)+ (145 ± 4.6)+ (52.3 ± 3.1)+ (3.6 ± 0.5)+ (4.3 ± 0.3)
Imatinib treated samples with primary and c-kit secondary mutations
No. 16>50%80%+ (12.1 ± 2.2)+ (9.6 ± 2.4)+ (16.8 ± 1)+ (147 ± 2)-
No. 17>50%90%+ (4.5 ± 0.7)+ (228 ± 7.5)+ (105 ± 7.4)+ (242 ± 11.3)+ (11.3 ± 1.5)
No. 18>50%60%+ (60.4 ± 15.5)-+ (10.4 ± 1.6)+ (30.5 ± 2.1)+ (12.3 ± 2.5)

In line with the results obtained in the KIT-transfected NIH3T3 cells, KIT+ 7AAD− (live) cells could be identified in all of the samples (see Materials and Methods, and Fig. 1 for a detailed description of the FC strategy). All of the KIT+ GIST cells were negative for the hematopoietic marker CD45, but expressed CD133. CD90 was expressed in 8/9 samples, CD44 in 7/8 (not done in patient no. 8), and CD34 in 6/9 (Table 2). The unimodal (single-peak) fluorescence distribution of these antigens suggested a homogeneous cell population (Fig. 3A).

Figure 3.

Immunophenotype profiles of GIST cells from untreated and imatinib-treated patients. The KIT 7AAD, KIT CD34, and CD34 CD133 bivariate dot plots are shown for each case. The R1 region was the immunological gate used to identify live KIT+/7AAD− GIST cells; the R2 regions were used to identify dead cells (blue). CD133, CD90, CD44 and CD34 expression (evaluated in the corresponding R1 regions) showed unimodal fluorescence distribution (green histograms) in all cases. (A) Untreated sample (no.1 in Table 1). (B) Treated sample harbouring only a primary c-kit Exon 11 mutation (no.15 in Table 1) with lower levels of KIT, CD133, CD90, and CD44 expression. (C) Treated sample harboring a primary c-kit Exon 11 and a resistant c-kit Exon 17 secondary mutation (no. 17 in Table 1) with lower KIT and higher CD133 and CD90 expression levels. The normalized MFI ratios are shown in Table 2. The dark lines are the FITC and PE isotype controls. (D) Comparison of KIT, CD133, CD90, CD44, and CD34 expression in untreated (UT), treated harboring only a primary c-kit Exon 11 mutations (T) and treated samples harboring a primary c-kit Exon 11 and a resistant c-kit mutation (T1). FITC: fluorescein isothiocyanate; PE: phycoerythrin.

Overall, with the exception of CD34, these results showed that naïve c-kit Exon 11-mutated and wild-type GIST cells have an immunophenotypical profile that is similar to that identified in primitive CD133+, CD34−, KIT+, and CD90+ stromal bone marrow mesenchymal progenitors (27). This particular immunophenotype led us to hypothesize that GISTs are a clonal expansion of quite homogeneous and primitive cells without any sign of a more primitive CSC component, at least when using this set of antigens.

Imatinib treatment does not select cells with a more primitive profile

The main escape from imatinib treatment is the acquisition of resistance mutations in c-kit Exons 13, 14, or 17. According to the CSC hypothesis, these mutations occur in putative GIST CSCs that may be selected by imatinib pressure. For this reason, we analyzed the FC profiles of samples obtained from molecularly characterized imatinib-treated patients: six carrying only primary c-kit mutations and three carrying both primary and resistant mutations (Table 1). All of the samples bearing only primary c-kit Exon 11 mutations showed CD133, four showed CD90 (not done in patients no. 12 and 13), three showed CD44 (not done in patients no. 12, 13, and 14), and 5/6 showed CD34; all three samples with primary and resistant mutations showed CD133 and CD90, and 2/3 showed CD44 and CD34 (Figs. 3B and 3C, and Table 2).

In comparison with the nine untreated samples, the six treated c-kit Exon 11-mutated samples showed a significant reduction in the staining intensity of surface KIT and CD133 (P ≤ 0.05, Mann-Whitney test), a trend toward a reduction in CD90, and an increase in CD34 expression. As sample-autofluorescence, spectral overlapping and undesirable antibody binding were respectively controlled by normalizing MFI, compensation and the evaluation of negative cells, the observed antigens expression variability (Fig. 3D and Table 2) among GISTs samples can be attributed biological rather than a technical differences ones. In addition, the reduction in KIT staining intensity after imatinib therapy was already described (28), and was confirmed in our samples by immunohistochemistry (Supporting Information images 1, 2, 3, 4, 5, and 6). Although FC profile modulation could be observed in the treated GIST samples without resistant mutations, all of the antigens retained a unimodal fluorescence distribution like that of the untreated samples, which was also retained in the samples with secondary mutations.

Murine Xenografts

To explore further the biological significance of the imatinib-induced GIST phenotype, we established a murine xenograft model. Two untreated (no. 2 and 9, Table 1) and two treated tumor samples carrying c-kit Exon 11 mutations (no. 11 and 13, Table 1) were transplanted into athymic CD-1 Nu/Nu mice (three 2 × 2 × 2 mm3 fragments of each sample). All of the tumors from the untreated patients (n = 6) had disappeared one month after transplantation, whereas those from the treated patients remained stable and, as there was no evidence of growth, were excised after six (no. 11) and eight months (no. 13) and analyzed. FC and immunoprecipitation/western blot (IP/WB) analyses showed increased KIT activation coupled with increased CD133 expression in comparison with the corresponding imatinib-treated surgical samples (Fig. 4). Nude CD-1 Nu/Nu mice have been successfully used to engraft the non-commercially available GIST 882 cell line (29, 30). Although the choice of this mouse model does not fully explain the lack of in vivo growth in our primary GIST xenografts, we cannot exclude the possibility that other strains of immunodeficient mice may be a more suitable model for GIST xenograft experiments.

Figure 4.

KIT reactivation after imatinib discontinuation in a mouse xenograft: representative flow cytometry (FC) and immunoprecipitation/Western blot (IP/WB) analyses. Three fragments of the surgical samples obtained from patient no. 13 were subcutaneously implanted into three CD-1 Nu/Nu mice (one fragment per mouse), and then explanted and analyzed by means of IP/WB and FC. (A) KIT IP/WB analysis before (Line 1) and after mice implantation (Line 2). Increased KIT activation (i.e., phosphorylation) was observed after the discontinuation of imatinib. The reduced intensity of the completely glycosylated 145 kDa form of KIT in the excised sample was due to the smaller percentage of KIT+ cells in comparison with the sample obtained from the patient (5% vs. 30%). As a loading control, equal volumes of unbound protein from the IP experiments were loaded onto a gel and hybridized with alpha actin, and showed comparable amounts of proteins along the lanes. The FC profile of the imatinib-treated surgical sample before (B) and after mouse implantation (C). KIT reactivation was coupled with increased KIT (MFI ratios: 2 ± 0.8 vs. 6.8 ± 0.44) and CD133 (MFI ratios 9.7 ± 0.9 vs. 25.4 ± 2.3) expression, and reduced CD34 (54 ± 7.1 vs. 21.16 ± 2) expression. M: molecular weight marker; Ctr-: negative control (a KIT-negative leiomyosarcoma sample); Ctr+: positive control (a KIT-positive GIST sample).

Although obtained in a limited number of cases (and thus lacking statistical support), and with small changes in the MFI ratios, these results suggest that c-kit Exon 11-mutated GIST cells re-acquire the profile characterized by KIT activation and CD133 expression when imatinib treatment is discontinued, which may indicate that the profile modulation observed in the treated samples does not reflect the selection of putative GIST CSCs. Furthermore, these findings confirm that at least the GISTs harboring a c-kit Exon 11 mutation are made up of primitive cells that strictly depend on the oncogenic addiction of KIT.

c-kit, CD133, and CD90 Transcription Levels in the GIST Samples

To support these indications of the primitive nature of GIST cells further, we compared the transcription levels of c-kit, CD133, and CD90 with those observed in known pluripotent cells.

RNA from CD34+ hematopoietic precursor cells (HPCs) was used as calibrators to quantify KIT and CD133, and RNA from cultured bone marrow-derived (23) mesenchymal stem cells (BM MSCs) that do not express KIT or CD133 were used as calibrators to quantify CD90. HPCs and BM MSCs were used because they express the antigens intensely. All of the GIST samples had higher c-kit mRNA levels than the HPC calibrator, and there was no difference in c-kit mRNA transcription between the untreated and treated GISTs (Fig. 5A). The levels of CD133 mRNA in all of the untreated samples, as well as in the samples carrying secondary c-kit mutations, were comparable with those measured in the HPC calibrator, whereas the CD133 mRNA content in three of the untreated samples (no. 1, 3, and 6) was higher than in the HPC calibrator (Fig. 5B). Interestingly, there was a trend towards lower CD133 transcription rates in the treated samples carrying primary c-kit mutations alone than in the untreated samples. The highest CD90 transcription rate was measured in two untreated c-kit Exon 11-mutated samples (no. 1 and 7), and in the two samples with primary c-kit Exon 11 and secondary c-kit Exon 13 and 17 mutations (Fig. 5C).

Figure 5.

c-kit, CD133 and CD90 transcription levels: comparison with hematopoietic and mesenchymal bone marrow-derived stem cells. The RNA extracted from mobilized CD34+/CD133+/CD45+ hematopoietic stem cells (HPCs) and cultured human bone marrow mesenchymal stem cells (BM MSCs) were used as calibrators. (A) Relative quantification of c-kit mRNA. (B) Relative quantification of CD133. (C) Relative quantification of CD90. Sample no. 9 was not used for the relative quantification of CD90 RNA as FC did not reveal the presence of CD90. Grey columns: untreated samples; dark grey columns: treated samples carrying only c-kit Exon 11 mutations; white columns: treated samples carrying both the c-kit Exon 11 primary mutation and resistance mutations.

Taken together, these results suggest that GISTs have c-kit, CD133, and CD90 transcription rates similar to those of other primitive (hematopoietic and mesenchymal) cells, thus supporting the primitive nature of GIST cells. Furthermore, the trend towards a decrease in CD133 transcription supports the lower CD133 MFI values found in the treated samples bearing primary c-kit Exon 11 mutations alone in comparison with those observed in the untreated samples.


Our findings indicate a previously unreported profile in a series of untreated and imatinib-treated GISTs (wild-type for c-kit and PDGFRA, or carrying primary and secondary c-kit mutations) which, in addition to expressing KIT and CD34, also co-expressed CSC markers such as CD133, CD90, and CD44. The unimodal fluorescence distribution revealed by the analyzed antigens, together with their modulation during imatinib treatment in the absence of any selection of more primitive cells, strongly supports the hypothesis that GISTs are made up of homogeneous and primitive cells that strictly depend on KIT oncogenic addiction. The primitive nature of GISTs was further confirmed by the high levels of c-kit, CD133, and CD90 mRNA.

The primitive nature of GISTs is supported by the recent discovery in post-natal mouse smooth muscle layers of rare, small, oval-shaped cells expressing KIT, CD34, and CD44 that may be the local progenitors of GISTs (14), and by our own findings showing that GIST cells co-express CD133 and CD90, two biomarkers that have been reported in primitive CD133+ bone marrow-derived stromal cells (27). At pre-clinical level, it has been shown that CD133+ stromal cells have a potentially wide range of differentiation that encompasses mesenchymal (osteoblastic, adipocytic, and myocytic), endothelial, hematopoietic and ectodermic (neurogenic) cell lineages (27). Interestingly, it has also been recently reported that imatinib-resistant rhabomyosarcoma transdifferentiation parallels KIT expression/activation loss in GISTs treated with imatinib (11).

It is well known that GIST patients develop imatinib resistance after an initial response, and highly sensitive mutation screening has led to the detection of secondary mutations in up to 90% of cases (7). The mutations may also be heterogeneous, and up to four different secondary mutations may be detected in surgical samples of advanced GISTs obtained from a single patient (31). This is particularly true for GISTs carrying c-kit Exon 11 mutations which, in addition to being the most frequent, are the most responsive to imatinib and therefore more likely to undergo long-term treatment and develop imatinib resistance (1).

As one of the key concepts of the CSCs hypothesis is that tumour progression is driven by the accumulation of mutations in CSCs, and it is well known that imatinib resistance is due to selective drug pressure leading to the survival of resistant clones, the latter are expected to have a different and possibly more stem-cell oriented FC profile. It is also expected that the presence of resistant clones (not targeted by imatinib) will be highlighted by FC and characterized by a “bimodal” antigen fluorescence pattern (i.e., CD133, CD90, CD44, or CD34); however, we found that imatinib-treated GISTs with primary c-kit Exon 11 mutations or secondary c-kit mutations retained a “unimodal” pattern that was similar to that observed in the untreated samples. Nevertheless, there was a trend toward reduced KIT, CD133, and CD90 expression, and increased CD34 expression in the treated GISTs harboring primary c-kit Exon 11 mutations alone, which suggests that the primitive FC profile may be modulated by treatment. On the basis of gene expression (5), it cannot be excluded that anatomical location (stomach vs. small bowel) may have an effect on FC profile modulation but the expression/activation of KIT and its downstream targets is maintained in responsive GISTs, albeit at a decreased level (32), which supports the persistence of KIT addiction in differently located tumors. Furthermore, the increased KIT expression/activation and increased CD133 expression observed in the two samples harboring primary c-kit Exon 11 mutations derived from imatinib-responding GIST tissue that were engrafted into CD-1 Nu/Nu mice (and thus mirroring imatinib discontinuation) suggest that the effects of imatinib are reversible, and that GIST cells may revert to their primitive profile when the drug is withdrawn.

In terms of secondary c-kit mutations, all of the evaluated antigens showed a unimodal distribution coupled with low KIT expression. Interestingly, two out of three samples showed high CD90 expression, which further underlined the primitive nature of the GISTs.

Taken together our findings support the idea that secondary mutations may represent the clonal evolution of “common” tumoral cells (31, 33, 34) rather than the selection of the CSCs since no variation in the FC profile of GIST was observed also after Imatinib treatment. Clones harboring primary and secondary c-kit mutations might be present at the time of tumor onset or might develop during imatinib treatment but remain hidden by a sort of “ frozen shared FC profile” driven by KIT activation which characterizes all the tumoral cells (KIT oncogene addiction).

It is important to point out that, on the basis of published data indicating that KIT signaling is essential for the development of Cajal interstitial cells (35), we evaluated the expression of CD133, CD90, CD44, and CD34 in KIT-positive cells. However, we cannot exclude the possibility that a minor sub-population of GIST CSCs that are KIT negative or express only low levels of KIT survive the inhibition of critical oncogenic pathways, and may have escaped this immunological gate (36).

In conclusion, our findings suggest that naïve KIT Exon 11-mutated and wild-type GISTs consist of homogeneous and primitive cells with a stem cell-like profile (CD133, CD90, CD44, and CD34), and that their imatinib resistance may be due to time-dependent clonal evolution closely related to KIT expression. However, these findings were obtained in a limited number of c-kit Exon 11-mutated GISTs and require further studies, possibly including GISTs with other genotypes.


The authors are grateful to Doctor Bruno Brando for his guidance and support. The authors declare that they have no financial, professional, or personal conflict of interest.