Introduction

The epidermal growth factor receptor tyrosine kinase (EGFR-TK) is currently considered one of the most interesting molecular targets for cancer therapy [1]. EGFR is differentially dysregulated, overexpressed, mutated, or amplified in many types of cancer, and this may be associated with more aggressive disease, therapy resistance, and shorter survival time [2–5]. Two major therapeutic strategies have been developed to inhibit EGFR pathways in cancer: monoclonal antibodies (Abs) that target the external binding domain (ligand binding domain) of the receptor or small molecular weight inhibitors that target the intracellular TK domain [6]. Some molecular markers for selecting patients who may benefit from these drugs and predict response to EGFR-targeted therapy are under investigation. Thus, an attempt to quantify EGFR in tumors in vivo has become one of the most pressing challenges in cancer research.

Different imaging approaches have been devised for EGFR detection, ranging from optical imaging modalities to single photon emission computed tomography and positron emission tomography (PET) technologies [7]. New PET probes, including labeled monoclonal Abs and small molecules such as TK inhibitors, have been developed and evaluated in the preclinical setting for EGFR visualization [8–25]. Although most of these TK inhibitor tracers showed promising and potential characteristics in vitro, none of them proved successful in the clinical setting. The main drawbacks of these EGFR PET agents stem from their rapid clearance from blood, moderate and nonspecific binding in tumors, and high accumulation in metabolic organs. To overcome these limitations, a newer generation of labeled irreversible EGFR inhibitors was developed [26, 27]. This group of compounds contains a polyethylene glycol (PEG) group at the 7-position of the quinazoline ring to increase solubility and decrease (logP) and a dimethylcrotonylamide at the 6-position to form covalent binding with the receptor. These compounds were labeled either with carbon-11 on the dimethylamine moiety (11C-1), fluorine-18 on the F-PEG moiety (18F-2), or iodine-124 at the anilino moiety (124I-3; Fig. 1). Herein, we report the micro-PET results of three of these compounds in tumor-bearing mice.

Fig. 1
figure 1

Chemical structure of the 11C-1, 18F-2, and 124I-3 compounds.

Full size image

Materials and Methods

Cells

Two human glioblastoma cell lines were used: U138MG lacks EGFR (HER-1) expression, whereas U87MG.wtEGF-R (U87MG) was transfected with an overexpressing human wild-type EGFR gene [17]. Cells were routinely cultured in DMEM supplemented with 10% fetal bovine serum and were maintained at 37°C in a humidified 5% CO2 atmosphere. The U87MG cells were also routinely supplemented with G418 at a final concentration of 500 μg/ml. All medium constituents were purchased from Invitrogen, Milan, Italy.

Cytofluorometric Studies

Glioblastoma cell line phenotype was studied by means of indirect immunofluorescence and cytofluorometric analysis. The following primary mouse monoclonal antibodies were used: antihuman HER-1 (EGF-R) clone 528 (Oncogene Research Products, Uniondale, NY, USA), antihuman HER-2 clone MGR-3, and antihuman HER-3 clone SGP1 (NeoMarkers, Fremont, CA, USA). The secondary antibody was Alexa Fluor 488 F(ab’)2 fragment of goat antimouse IgG (Invitrogen, Milan, Italy). After the final washings, cells were resuspended in phosphate-buffered saline containing 1 μg/ml of ethidium bromide to gate out dead cells followed by cytofluorometric analysis with a FACScan (Becton Dickinson, St. Jose, CA, USA).

Mice

Athymic Crl:CD-1-Foxn1 nu/nu mice (here referred to as nude mice) were purchased from Charles River Italy and kept under sterile conditions. Experiments were authorized by the institutional review board of Bologna University and performed according to Italian and European guidelines.

U138MG and U87MG cells (3 × 106 and 5 × 106, respectively) were injected subcutaneously (s.c.) in the hind legs. Mice carrying a single tumor derived from a subcutaneous injection of each cell line or carrying both tumors (one on each side) were obtained. Tumor incidence and growth were evaluated weekly. Neoplastic masses were measured with calipers; tumor volume was calculated as π × [√(a × b)]3/6, where a = maximal tumor diameter and b = a tumor diameter perpendicular to a. Animals underwent PET imaging 5–6 or 10–15 days after cell injection. At sacrifice, tumor samples were mechanically and enzymatically dissociated using Trypsin–EDTA to obtain a single cell suspension for evaluation of EGFR expression by cytofluorometric analysis. Additional samples for histological determinations and biomolecular studies were fixed in 10% phosphate-buffered formalin or frozen in liquid nitrogen.

Chemistry

The labeled compounds were prepared according to recent publications [26, 27].

Imaging Protocol

The PET imaging procedure was as follows: Mice were anesthetized (Sevofluorane 3–5% and oxygen 1 l min−1) and injected with variable doses of tracers (ranging from 10 to 37 MBq according to the isotope half-life) with an insulin syringe via the tail vein. All animals were evaluated in a static mode. This approach was chosen since the primary aim of this work was to evaluate tracer uptake related to the EGFR expression. Two animals injected with 18F-2 underwent dynamic scans to evaluate whether an early tracer uptake and a later washout were present.

For static acquisitions (repeated several times after injection to monitor tracer distribution), animals were subsequently allowed to awaken and move freely for the uptake period. Finally, anesthesia was induced a second time before the animal was placed prone on the scanner bed of the small animal PET tomograph (GE, eXplore Vista DR). Imaging lasted at least 15 min per bed position.

For dynamic acquisitions, the animal was anesthetized, placed prone on the scanner bed, centered in the field of view, and then injected intravenously with the tracer. Simultaneously, the dynamic list mode acquisition was started. Image acquisition lasted for 30 min.

Since the axial field of view was only 4 cm, one bed position was not sufficient to cover the whole body. Therefore, a group of animals underwent at least one static whole body scan (two bed positions) to verify the whole body distribution of the tracer. For the remaining animals, images were acquired on tumor areas only. Once the scan had been completed, gas anesthesia was interrupted and the animal was placed in a warm recovery box.

The images were reconstructed with iterative reconstruction 2D OSEM, matrix 175 × 175. Due to the small size of the animals, we did not apply attenuation correction or scatter correction as they do not significantly influence semiquantitative PET analysis since the very high energy of the emitted photons (511 KeV) led to a very high rate of nonabsorbed photons. Using this system, the best attainable spatial resolution was 1.4 mm as described by Wang et al. in their performance evaluation [20]. According to this procedure, the image quality was satisfactory for all of the scans. Although the image quality could have been slightly improved especially by increasing the acquisition time, as in every nuclear medicine test, it was unnecessary since a prolonged acquisition time reduces the number of scanned animals per tracer synthesis.

For static acquisitions, a semiquantitative analysis was performed by calculating the tumor-to-background ratio (TBR) on the pelvic area of increased uptake. TBR was calculated as follows:

$$ {\hbox{TBR}} = {\hbox{ Max Count in the Target Region of Interest }}\left( {\hbox{ROI}} \right)/{\hbox{Mean Count in the Background ROI}} $$

The background ROI was drawn on the subcutaneous tissues. For dynamic acquisitions, a ROI was drawn on the tumor and a time/activity curve was generated. No correction for decay was applied.

124I-3

Eleven animals underwent PET imaging with 10 to 30 MBq of the iodinated tracer. Three animals also underwent a small animal CT whole body acquisition that was used for accurate localization of activity uptake detected by the micro-PET. Table 1 shows the animal population and the image acquisition features.

Table 1 Image acquisition features for 124I-3
Full size table

18F-2

Six animals underwent PET imaging with approximately 37 MBq of the fluorinated tracers. Table 2 shows the animal population and the image acquisition features.

Table 2 Image acquisition features for 18F-2
Full size table

11C-1

Three animals underwent PET imaging with approximately 37 MBq of the carbon-11 tracer. Table 3 shows the animal population and the image acquisition features.

Table 3 Image acquisition features for 11C-1
Full size table

Molecular and Pathological Analyses

After sacrifice, tumor necrosis, microvessel density (MVD) analysis, and total and activated (pY1068) EGFR quantification were evaluated. Immunostains were adopted to evaluate endothelial cells. Anti-CD31 antibody (Santa Cruz, CA, USA) at a 1:20 dilution was used. The sections underwent antigen retrieval in EDTA (pH = 6.0) in a microwave at 750 W (four cycles lasting 5 min each) and revealed by the alkaline phosphatase–anti-alkaline phosphatase technique. Two separate methods were used to estimate MVD. In the first method, visual microvessel grading, the slides were visually scanned at ×100, ×200, and × 400 magnification and semiquantitatively graded for the extent of CD31 staining. To ensure the accuracy of the grading method, each sample was reviewed by two observers. Morphologic analysis was performed carefully to ensure vessel specificity of the CD31-stained stroma considered for the analysis. Four different microvessel grades (MVG) were considered as previously described [28]: MVG 1, normal or slightly increased MVD; MVG 2, microvessels easy to detect and definitely increased in respect to normal; MVG 3, abundant microvessels; and MVG 4, strongly increased MVD. Nonneoplastic human nervous tissues were used as control reference. The second method, visual count, involved counting microvessels according to previously described methods [28, 29]. In performing this visual count, each of the slides was first scanned at ×100 magnification, and three areas with abundant microvessels were chosen and defined as “hot spots”. The number of microvessels in each of these hot spots was then determined at ×400 magnification. The final MVD number was determined by taking the average of the three separate visual counts. During counting, vessels in the tumor capsule were excluded. Areas of staining with no discrete breaks were counted as single vessels, and the presence of a lumen was not required.

The two-tailed Student T test was used to compare U87GM and U138GM mice. Differences were considered significant for p values < 0.05. Pearson’s correlation was used to compare MVD and MVG methods.

Total and activated (pY1068) EGFR quantification was performed using an enzyme-linked immunosorbent assay (ELISA). For protein extraction, frozen tissues were homogenized using lysis buffer (50 mM Tris, pH 7.4, 150 mM NaCl, 2 mM MgCl2, 1% Triton X-100, 10% glycerol, 2 mM EGTA, 1 mM DTT) containing protease inhibitors (10 mg ml−1 aprotinin and leupeptin, 5 mg ml−1 pepstatin, 1 mM phenylmethylsulfonyl fluoride) and phosphatase inhibitors (50 mM NaF, 10 mM Na4P2O7, 1 mM Na3VO4, 3 mM H2O2). Homogenates were centrifuged at 13,000×g for 15 min at 4°C, and supernatants were stored at −80°C until analysis. Total and activated EGFR concentrations were assessed using ELISA kits purchased from Biosource International Inc. (Camarillo, CA, USA). Relative activated (pY1068) EGFR was defined as pY1068/total EGFR.

Results

Animal Model Development

The human glioblastoma cell lines U138MG (lacking EGFR expression) and U87MG (transfected with an overexpressing human wild-type EGFR gene) were used to establish a subcutaneous xenograft tumor model in nude mice for the assessment of EGFR labeling with innovative PET bioprobes. Indirect immunofluorescence and cytofluorometric analyses were used to study the growth factor receptor expression profile of EGFR (HER-1) and HER/erbB family receptors HER-2 and HER-3 in U138MG and U87MG cells. As indicated by the cytofluorometric profile, U87MG cells in continuous culture showed EGFR overexpression on the cell surface, whereas U138MG cells were negative for EGFR membrane expression. Surface expression of HER-2 proteins was almost negative on both cell lines, while HER-3 was expressed only in the U138MG cell line. When the cell lines were grown as subcutaneous tumors in nude mice, the HER/erbB family receptor expression profile was similar to that seen when the cells were grown in continuous liquid culture. Most importantly, EGFR expression in U87MG xenografts remained very high, whereas U138MG xenografts remained completely negative for EGFR expression.

The xenograft tumor model was established by s.c. injection of 3 × 106 U138MG or 5 × 106 U87MG cells into the hind legs of nude mice. After subcutaneous injection, tumor growth became evident after a median latency time of 4 days for both cell lines. U138MG tumor growth was exponential whereas U87MG growth was very slow and sometimes indolent. However, no differences in tumor latency, incidence, or growth were seen in mice carrying one subcutaneous tumor or both (one on each hind leg). Tumor-bearing animals underwent PET analysis 5 to 6 days after cell injection when tumor dimensions were 0.161 ± 0.011 and 0.104 ± 0.004 cm3 for U138MG and U87MG, respectively (dimensions from one experiment representative of other similar experiments). PET analysis was also performed 10–15 days after cell injection when tumor dimensions were 3.368 ± 0.956 and 0.785 ± 0.309 cm3, respectively (dimensions from one experiment 14 days after cell injection representative of other similar experiments).

Tumor Uptake

The time elapsed between the injection of the tracers and image acquisition is listed in Tables 1, 2, 3, and 4. Table 4 shows the results of the static imaging procedures with the tumor-to-background ratio. Some tumors presented a slight increase in tracer uptake, but in general, there was no retention over time and no significant difference between U138MG and U87MG masses. Static acquisitions and dynamic scans with the tested tracers showed no statistically significant differences between tumor xenografts (Figs. 2, 3, and 4). Dynamic scans did not demonstrate any significant uptake of 18F-2 over time (Fig. 5).

Table 4 Imaging results for all tested tracers
Full size table
Fig. 2
figure 2

PET/CT images of an animal carrying a U87MG (EGFR+) tumor, 4 h after the tracer injection. The image fusion between the CT and the PET image shows no specific tracer uptake in the tumor of 18F-2.

Full size image
Fig. 3
figure 3

The segmental 11C-1 static acquisition image (10 min after the tracer injection) does not show any visual significant difference between U87MG (EGFR+) and U138MG (EGFR−) mass tracer uptake. TBR turned out to be 1.7 for both masses.

Full size image
Fig. 4
figure 4

The segmental 18F-2 static acquisition image (5, 60, and 150 min after the tracer injection) does not show visual significant difference between U87MG (EGFR+) and U138MG (EGFR−) mass tracer uptake. The cold area indicated by the solid arrow in the U138MG EGFR− mass is related to necrosis.

Full size image
Fig. 5
figure 5

The segmental 18F-2 static acquisition image (60 min after the tracer injection) does not show any visual significant difference between U87MG (EGFR+) and U138MG (EGFR−) mass tracer uptake. TBR turned out to be <1 for both masses. The two time-activity curves do not reveal any specific uptake.

Full size image

Molecular and Pathological Analyses

No necrosis was present except in four animals with bilateral tumors where it was found both in U87MG and U138MG masses because of high tumor size. Results from the two angiogenesis evaluation methods were absolutely comparable (Pearson correlation, r = 0.88). Moreover, there was a very high interobserver consensus with more than 95% agreement. Median MVD was 28.5 microvessels/×400 field (range 6–70.8) in tumor samples and six microvessels/×400 field (range, 4–6) in nonneoplastic tissues. When the two tumor xenografts were compared, mean MVD was 9.6 and 48 microvessels/×400 field in the U138GM and U87GM masses, respectively (p = 0.00008; Fig. 6). Similarly, the MVG scores were generally higher in the U87GM group (p = 0.002). Total EGFR amount was higher in U87MG than U138MG masses (p = 0.001), but the ratio of activated (pY1068) to total EGFR did not differ (p = 0.95; Fig. 7).

Fig. 6
figure 6

Micro vessel density evaluation in U87MG (EGFR+) and U138MG (EGFR−) masses.

Full size image
Fig. 7
figure 7

Total EGFR and ratio of activated (pY1068) EGFR protein level quantification by ELISA in U87MG (EGFR+) and U138MG (EGFR−) masses.

Full size image

Discussion

The present study reports on the development of PEG-anilinoquinazoline derivatives labeled with 11C, 124I, and 18F as potential probes for in vivo imaging of EGFR in U138MG and U87MG glioblastoma xenograft mice using small animal PET. The most important chemical features of these PEGylated compounds were increased stability and solubility and decreased lipophilicity which together may increase specific tracer uptake in tumors. In vitro studies showed that these PEGylated derivative compounds had a high and irreversible potency to inhibit EGFR autophosphorylation [26, 27]. However, the in vivo imaging protocol of 11C-1, 18F-2, and 124I-3 showed only a slight increase in uptake in some tumors and no significant difference between U138MG and U87MG glioblastoma xenograft mice. The lack of specific tumor uptake of these tracers could be due to factors such as the target status in vivo, nonselective binding, and their metabolism in vivo.

The molecular quantification of total EGFR by ELISA differed between U138MG and U87MG tumor mass, but no significant differences were found when the ratio of activated/total EGFR was measured. Thus, tracer binding to EGFR could be influenced by the activation of the tyrosine kinase domain of the receptor which was similar in U138MG and U87MG.

Though nonselective binding of tracers to other proteins could reduce EGFR-specific uptake, this hypothesis is unlikely. A higher affinity and selectivity of 11C-1 and 18F-2 to EGFR compared to other tyrosine kinases (ERBB2, IGF1R, KIT, PDGFRA, PDFGRB, VEGFR1, VEGFR2, VEGFR3) has been described [26, 27]. The same results were obtained for 124I-3 (unpublished data; Table 5). However, these compounds may bind to targets other than tyrosine kinase receptors, thus influencing their uptake. Future genomic and proteomic studies of tumor masses and cell cultures could be designed to define the molecular background of these models. As U138MG and U87MG are known to have different proliferation and growth rates in mice, different molecular pathways may be up- and downregulated in vivo.

Table 5 Affinity and selectivity studies of 124I-3 to EGFR
Full size table

Tracer distribution in tumor masses should not have influenced the results since all animals had a small mass at the time of scan acquisition and no necrosis, except for four animals with large bilateral necrotic tumors. Moreover, MVD data showed that angiogenesis was more apparent in U87GM tumors, suggesting sufficient tracer access to tumor masses.

Although the results of the present study showed no specific tumor uptake in micro-PET images for 11C-1, 18F-2, and 124I-3, the promising in vitro characteristics of these compounds merit further investigation in other animal models. Cell lines characterized by both high receptor expression and activation are deemed ideal for in vivo imaging of EGFR tumors. Pal et al. reported the first in vitro and in vivo studies of a morpholino-124I-IPQA TK irreversible inhibitor with a chemical structure allowing covalent binding only to the ATP site of the phosphorylated EGFR-TK in xenograft mice with brain tumors expressing EGFRvIII mutant receptor [18]. Recently, Memon et al. reported a higher uptake of 11C-Iressa in HCC827 xenograft mice which overexpress a mutant EGFR harboring an in-frame deletion mutation in exon 19 that likely confers a constitutive activation to the receptor and erlotinib sensitivity compared with nonmutant A549 and NCI358 cell lines [24].

These new approaches are encouraging because they better reflect the current clinical scenario. In recent years, the importance of detecting EGFR “expression” as a predictive biomarker to discriminate between patients responding and not responding to anti-EGFR drugs has lost its significance for various reasons [30, 31]. It is well known that EGFR inhibitors have clinical activity irrespective of total EGFR quantity. For example, there is accumulating evidence of the value of EGFR mutations in lung cancer and of k-ras mutations in codons 12 and 13 in colorectal cancer as predictive markers of sensitivity to anti-EGFR treatment [31–33].

Thus, this approach should also be considered in molecular imaging research, with more attention paid to imaging receptor “function” which is influenced by mutations or an aberrant downstream signaling pathway [34]. A recent study identified distinct biological features underlying the various EGFR mutations in nonsmall-cell lung cancer (NSCLC) that may confer different transforming potential or TK activity and consequently different sensitivity to EGFR inhibitors [35]. Specific chemical structures of the tracers and animal models should be developed to improve the imaging of EGFR status in cancer. Developing PET probes for imaging membrane receptors may have more rationale in cases of receptor amplification, such as HER-2 in breast cancer.

A remaining question is: if molecular predictive markers are available in everyday practice, why still study EGFR imaging? In vivo detection of EGFR may overcome several limitations of tissue specimen analysis [36]. Firstly, the information yielded by analysis of small tissue specimens may not be representative of the whole tumor or all metastases. Secondly, biological changes may occur during the natural history of metastatic disease especially in long survival or heavily pretreated patients. Lastly, invasive, nonrepeatable, and time-consuming procedures are required for both patients and physicians. As a consequence, the in vivo and noninvasive detection of EGFR function with a repeatable approach during disease progression may still represent a major potential tool for many oncologists.

Conclusions

PEG-anilinoquinazoline tracers should be investigated in animal models with an aberrant receptor function such as EGFRvIII mutants and HCC827 NSCLC models. Moreover, novel chemical modifications are warranted to further improve binding with the ATP pocket of the mutated kinase domain underlying receptor activity. Lastly, it would also be interesting to use an imaging approach to identify changes related to the downstream signaling pathway of EGFR.