Proteomic detection of a large amount of SCGFα in the stroma of GISTs after imatinib therapy

Gastrointestinal stromal tumors (GISTs) are the most frequent mesenchymal tumors to develop in the digestive tract. They typically arise from the stomach (40-70%) or small intestine (20-40%) but can also occur in the colon-rectum (10-15%) and rarely in the esophagus. At least 10-30% of tumors are discovered incidentally during laparotomy, endoscopy, or other imaging studies; 15-47% of patients present with overt metastatic disease and common sites of metastases include liver, peritoneum, and omentum [2‐4]. GISTs are thought to originate from the interstitial cells of Cajal or their precursor cells [5]. Most GISTs are characterized by gain-of-function mutations in the genes encoding KIT and platelet-derived growth factor receptor alpha (PDGFRA) [6], mutations that appear to be mutually exclusive. The emerging role of stem cell factor (SCF) as the ligand of the receptor tyrosine kinase KIT [7‐9] suggests that an autocrine-paracrine loop serves as a possible further mechanism of action [10]. However, the SCF/KIT system plays an important role not only in the differentiation and proliferation of interstitial cells of Cajal and the development of GISTs but also in the development of hematopoietic cells such as mast cells, erythroblasts, and melanocytes [11].

GISTs are highly resistant to conventional chemotherapy [12]; in the past decade, the introduction of imatinib mesylate (Gleevec^®, Novartis Pharmaceutical Corporation, NJ, USA), a KIT receptor blocker, has significantly improved the prognosis of GIST patients. Tumor response depends on the presence/absence and type of mutations in KIT or PDGFR. Unfortunately, the major problem with imatinib treatment is resistance, mainly secondary resistance that generally evolves in most patients after a median of two years of therapy [13, 14].

Eighty to eighty-five percent of patients with advanced GISTs exhibit an initial benefit from imatinib treatment; however, the response level varies from rapid and gross reduction in tumor volume to little or no tumor shrinkage (described as stable disease) [6]. Size-based response criteria such as the World Health Organization criteria or the current international Response Evaluation Criteria in Solid Tumors are thus thought to underestimate the response and are not appropriate tools to assess tumor response to imatinib [15]. Consequently, the clinical management of these patients and the criteria used to assess clinical response to imatinib therapy have recently been redefined [16]. In accordance with previously reported criteria, GISTs can be classified using the following scores: high responders, 0 to <50% residual viable tumor cells with no mitosis, and no obvious Ki-67 immunostaining; moderate responders, >50% to 90% tumor cells, no mitosis, and Ki-67 immunostaining in 0 to <10% of cells; low responders, >50% to 90% tumor cells, mitotic index > 10/50 high-power fields, Ki-67 immunostaining in 20-30% or >30% of cells; and non-responders, >90% tumor cells [17, 18].

The histological/pathological response of GIST to imatinib therapy is also variable and heterogeneous from nodule to nodule within the same resection, as well as within the same lesion, and does not correlate well with clinical response. Limited studies of the histopathological changes in imatinib-treated patients indicate a significant change in the appearance of the tumor tissue following preoperative systemic imatinib therapy in GISTs; alterations particularly occur in the stromal compartment and are visible during standard pathological assessment. The density of the tumor and the number of intratumoral vessels decrease significantly and areas of cystification and hemorrhage are clearly visible. Tumors become homogeneous and hypodense. However, even in highly responsive tumors, microscopic foci of viable cells are seen either as isolated tumor cells or as distinct micronodules embedded in an extensively hyalinized background.

The molecular effects of imatinib on responding GISTs are currently being explored. Apoptosis (programmed cell death type I) has been frequently described in GIST cell lines treated with imatinib mesylate [19, 20]; importantly, McAuliffe et al. [21] indicate that the action of imatinib may be both cytotoxic (by evidence of apoptosis) and cytostatic. Recently, autophagy (programmed cell death type II) has been suggested as a possible alternative mechanism of response to the imatinib mesylate treatment in clinical biopsies [22]. Autophagy is the major self-degradative process in eukaryotic cells and has multiple physiological functions, including protein degradation, organelle turnover, and response of cancer cells to chemotherapy [23].

In this report, we used a proteomic approach in combination with other analyses to investigate the protein composition of highly responsive, resected GISTs after imatinib mesylated neoadjuvant therapy. Our aim was to identify several molecular components of the stroma with expression patterns possibly related to tumor response/behavior.

To explore the protein expression pattern of GIST samples treated with imatinib mesylate, protein extracts from five GISTs were loaded onto one-dimensional SDS-PAGE and visualized by Coomassie staining. These five samples consisted of one untreated tumor (GIST 1), three tumors from high-responder patients (GISTs 3-5), and one tumor (GIST 2) from a patient with clinically stable disease but who was scored as a non-responder (Table 1). The protein patterns of the highly responsive tumors (Figure 1, lanes C-E) were uniform and contained fewer protein bands than the other samples (Figure 1, lanes A and B). The staining patterns of the GIST samples did not differ consistently from that of a pool of plasma samples from healthy donors (Figure 1, lane F). MS of gel bands obtained after OFFGEL fractionation of the extracts (data not shown) confirmed that the majority of proteins were blood components (Table 2).

In the GIST 5 extract, Coomassie staining revealed bands (bands 8-10 in Figure 1A, lane E, and Figure 1B) that were not present in the plasma sample. Bands 8 and 9 corresponded to SCGF, also known as C-type lectin domain family member 11A (CLEC11A), an important hematopoietic growth factor with burst-promoting activity for human bone marrow erythroid progenitors [29]. Band 10 corresponded to C1q, the initiator of the classical complement cascade [30].

Western blotting with anti-SCGF antibody confirmed the presence of SCGF-positive bands in GIST 5 extract (b and c in Figure 2, lane B) and highlighted the presence of a third isoform of ~46 kDa (a in Figure 2, lane B) that was not detected by the proteomic analysis. All three isoforms were undetectable in a plasma sample (Figure 2, lane C), consistent with the observation that the SCGF plasma concentration was too low to be analyzed by western blot. In the peripheral blood of adults, SCGF concentration can only be detected by enzyme-linked immunosorbent assay because the concentration is approximately 10-20 ng/ml [31]. Using Coomassie staining to visualize the electrophoresed SCGF proteins, we estimated at least 60 ng of protein in the GIST 5 sample. Since 30 μg of tissue extract was analyzed, we calculated the SCGF concentration to be 0.5 ng/μg in the GIST extract, which was much more protein than would be present in an equivalent amount of plasma protein (0.03-2 pg/μg).

To assess the possible origin and localization of SCGF, we performed immunohistochemistry on all five GIST samples. GIST 5 (Figure 3B) displayed a strong SCGF positivity, consistent with the western blot, that was restricted to the stroma compartment. The rare viable tumoral cells present in the responding area were negative in the cytoplasm and the nucleus. However, these cells retained weak cytoplasmic KIT reactivity (Figure 3C). The responding cases (GISTs 3 and 4) and the naïve GIST (GIST 1) were negative (Additional file 1 and data not shown). Interestingly, GIST 2 (Figure 3D-F) exhibited minimal areas of regression that were mostly depleted of tumoral cells and were SCGF-positive and KIT-negative. RT-PCR did not reveal the presence of SCGF transcripts in responding or non-responding GISTs (Additional file 2), corroborating the hypothesis that blood is likely the source of SCGF.

SCGF is a largely uncharacterized hematopoietic mediator that promotes enhanced erythroid progenitor formation from human bone marrow [32]. Stem cell transplantation-elevated serum SCGF levels are associated with enhanced hematopoietic recovery [32], and the differentiation of dendritic cells in mature type I inflammatory cells is accompanied by an increase in SCGF secretion [1]. Based on these observations, we hypothesized that an inflammatory reaction induced by imatinib treatment could be responsible for the SCGF positivity we observed in our GIST specimens. We therefore employed immunohistochemistry to investigate the presence of macrophage/dendritic cells using anti-CD68 antibodies in samples GIST 2 and 5. Interestingly, we observed areas of CD68 positivity in GIST 5, while GIST 2 presented a diffuse infiltration of CD68-positive macrophages. In both cases the corresponding stromal counterpart was SCGF-positive (Figure 4).

In order to test the consistency of the relationship between SCGF expression and the histological response to imatinib, we analyzed SCGF expression in tumor areas scored as acellular or with <10% residual tumoral cells and in those areas scored as non-responsive or scarcely responsive in three imatinib-treated patients (GISTs 6, 7 and 8 in Table 1). Figure 5A summarizes the results obtained in a comparative analysis of non-affected and affected tumor areas: the hematoxylin-eosin staining of the histological sections (panels 1-3; 6-8; and 11-13); the images of the protein lysates prepared from adjacent tumor areas that were separated onto SDS gels and stained with Coomassie-blu (panels 4, 9 and 14); and, the western blotting of the same protein lysates probed by with anti-SCGF antibody (panels 5, 10 and 15). Lysates from imatinib-affected areas displayed a simplified protein band profile; notably, the anti-SCGF antibody recognized an SCGF-related band in protein lysates from responsive tumoral areas that was absent in lysates from matched non-responsive or scarcely responsive areas (Figure 5A). These observations strongly support a relationship between SCGF positivity and low cellularity due to imatinib activity. In addition, we noted the presence of inflammatory cell infiltrate in the hematoxylin-eosin section of the responding area of GIST 6 (panels 2 and 3), of scattered lymphocytes in the hematoxylin-eosin section of the responding area of GIST 7 (panels 7 and 8), and of monocytes/macrophages in the hematoxylin-eosin section of the responding area of GIST 8 (panels 12 and 13). One SCGF-related signal was also detected by western blot in responding GISTs 14 and 16, but not in responding GIST 15 and non-responding GISTs 9-13 (Table 1 and Figure 5B).

SCGF is a secreted cytokine expressed in two distinct forms; SCGF-α is the full size form (323 amino acids, 35,695 Da), while SCGF-β is the shorter form (245 amino acids, 26,902 Da) characterized by a deletion within a conserved carbohydrate recognition domain [33]. These theoretical masses only partially explain the observed molecular weights in electrophoretic separations. The spectra from our MALDI-TOF-MS of tryptic peptides from two SCGF-positive bands (band 8 and 9 in Figure 1, lane E) were nearly identical in size and were attributed to the α form (lanes G and H in Figure 1A).

Since SCGF has been described as O-glycosylated and sulphated, we deglycosylated the GIST 5 lysate to determine the sizes of SCGF bands before and after digestion. The immunoreactive band with the highest molecular weight (band a in Figure 6, lane 1) completely disappeared after deglycosylation; the intermediate band (band b in Figure 6, lane 1) remained unmodified, and the lowest band (band c in Figure 6, lane 1) exhibited an appreciable, apparently incomplete reduction in its molecular weight (band d in Figure 6, lane 2). Isoform a also occurred in conditioned medium (Figure 6, lane 4), as compared with isoforms expressed by the TPC1 cell line (Figure 6, lanes 3-5). Isoform a was equivalent to the fully glycosylated and secreted form, while the low-molecular weight isoform in the cell extract (band b in Figure 6, lanes 3 and 5) was equivalent to an immature, cytoplasmatic, unglycosylated form. Interestingly, we observed SCGF positivity for the isoforms corresponding to bands a and b in Figure 2, which were also observed in the conditioned medium (Figure 2, lane E) and in the cell extract (Figure 2, lane D) of the TPC1 cell line.

To further define the nature of SCGF isoforms expressed in the GIST 5 sample, we tested the possibility that the other form corresponded to isoform β. We transfected the HEK293 cell line that scored negative for SCGF proteins (Figure 6, lanes 6 and 7) with SCGF-β cDNA. Western blotting detected a protein with the size expected for the SCGF-β backbone (Figure 6, lanes 8-10) that did not correspond to band d of the GIST 5 sample (Figure 6, lane 2). Thus, we concluded that band d was not the SCGF-β form.

Few studies have correlated clinical response with histological response in GISTs after prolonged imatinib treatment. However, it is widely recognized that clinical outcome in stable or partially responsive GIST patients does not seem to be influenced by the duration of imatinib treatment, the histological response, or the size of the tumor [18]. In this study, we analyzed the proteomic, histological, immunohistochemical, and clinical features of a small group of GISTs resected after prolonged imatinib treatment.

SCGF, a novel cytokine, exerts its action on primitive hematopoietic progenitor cells. In combination with other hematopoietic growth factors such as granulocyte-macrophage colony-stimulating factor and erythropoietin, SCGF stimulates the formation of erythroid and granulocyte/macrophage colonies, although SCGF alone cannot induce colony formation [29]. Using a proteomic approach, we detected a large amount of SCGF in the protein extract of one imatinib-treated GIST sample (Figure 1). This sample also had a residual cell component of <10% and was negative or very slightly positive for CD117 staining. Biochemical analyses revealed the presence of a small number of KIT receptors with very low activation (data not shown). In parallel, we found extensive SCGF positivity in the abundant stromal component, which appeared homogenous, hypodense, and eosinophilic. This observation was replicated in part of the progressive lesion of another treated GIST case in which we identified strong SCGF positivity exclusively in CD117-negative areas. Interestingly, SCGF expression occurred in the imatinib-affected areas of three GISTs that was not detected in the unaffected or scarcely affected areas of the same tumors; SCGF-positive bands were also identified in two out three responding tumors but were absent in tumors from five non-responder patients.

A study carried out on in vivo material demonstrated that SCGF is strongly expressed in bone marrow and only faintly in lymphoid organs; in bone marrow, SCGF is concentrated in the cytoplasm of immature neutrophils, but not in myeloblasts, mature neutrophils, or the extracellular bone marrow fluid [34]. Since immature neutrophils play a role in tumor-induced immuno-inflammatory responses, SCGF may impact mechanisms regulating these responses. These observations are consistent with RNA expression patterns in mouse and human protein-encoding transcriptomes [35] and are attributed as follows: high expression levels to CD34+ cells, low expression levels to CD33+ (myeloid) cells, cardiomyocytes, and smooth muscle cells, and very low expression levels to all other cell lines or tissues.

Hiraoka [36] recently demonstrated that leukemia cell lines require self-secreted SCGF for their proliferation in tumors, indicating a putative autocrine SCGF mechanism, and that loop blockage with neutralizing antibody prevents extracellular SCGF from inducing apoptosis. Levina et al. [37] demonstrated that the high tumorigenic and metastatic potentials of lung cancer stem cells correlated with superior production of angiogenic factors and growth factors involved in cell proliferation and angiogenesis, describing increased levels of SCGF, stroma-derived factor 1α, and SCF in tumors from cancer stem cells in association with the stem cell phenotype. Gene-expression profiles from 35 childhood acute lymphoblastic leukemia matched diagnosis/relapse pairs, as well as 60 uniformly treated children at relapse, indicated that SCGF is significantly overexpressed at relapse [38].

The presence of SCGF in the CD117-negative stromal compartment of imatinib-treated GISTs suggests that its expression is associated with the histological response of GISTs to imatinib therapy. Our RT-PCR investigation revealed that SCGF is not actively transcribed in GIST samples, and thus it is difficult to determine the possible sources of SCGF in these specimens. A recent study described a subgroup of GISTs surgically resected after neoadjuvant imatinib treatment that exhibited reduced numbers of tumor cells in the hypocellular myxohyaline stroma, with small numbers of scattered atypical nuclei and occasional stromal hemorrhages [39]. A separate investigation assessed a GIST case treated with imatinib therapy for four weeks in which most of the tumor cells were replaced by myxoid stroma and the remaining tumor cells did not appear to be actively dividing [40]. However, to date no study has reported data on SCGF in GIST samples.

In our study, SCGF appeared as part of the stromal GIST component and, in particular, as part of the eosinophilic proteinaceous matrix described as myxoid, collagenous, or hyaline. Our proteomics experiment uncovered high plasma-protein content in treated tumors, and immunohistochemistry revealed the SCGF positivity in CD117-negative and CD68-positive areas. These observations could link SCGF positivity with the imatinib-induced inflammatory response that elicits monocyte/macrophage tissue migration, promoting scarring and removal of cell debris. CD68 positivity confirms macrophage infiltration, which may also explain the high level of C1q [41] in our GIST 5 proteomic analysis. Recent data support the hypothesis that induced type I maturation of dendritic cells is associated with a peak of SCGF production [1], supporting a pro-inflammatory role for this cytokine. It is therefore plausible that these are immunological reactions, and that liquidation of the dead tumor cells via macrophages leads to lesion shrinkage.

SCGF function may be related to the imatinib-induced inflammation response in responding GIST patients. Further studies are necessary to identify the receptor of this cytokine, to further clarify its origin, and to determine the reason for its accumulation in some imatinib-treated GISTs. These investigations may answer fundamental questions about the composition of the stromal matrix after imatinib therapy and identify proteins related to desirable tumor response/behavior.