Optical imaging predicts tumor response to anti-EGFR therapy

Abstract

To evaluate cetuximab treatment in head and neck squamous cell carcinoma xenografts and cell lines, we investigated a preclinical model of head and neck squamous cell carcinoma. Head and neck squamous cell carcinoma cell lines SCC-1, FaDu, CAL27, UM-SCC-5 and UM-SCC-22A were used to generate subcutaneous flank xenografts in SCID mice. Mice were divided into control and cetuximab treatment groups, mice in the latter group received 250 µg cetuximab once weekly for four weeks. After completion of therapy, SCC-1 (p < 0.001), UM-SCC-5 (p < 0.001), UM-SCC-22A (p = 0.016) and FaDu (p = 0.007) tumors were significantly smaller than control, while CAL27 tumors were not different from controls (p = 0.90). Mice were systemically injected with 50 µg of the Cy5.5-cetuximab bioconjugate and imaged by stereomicroscopy to determine if tumor fluorescence predicted tumor response. Intact tumor fluorescence did not predict response. Tissue was harvested from untreated xenografts to evaluate ex vivo imaging. Cell lines were then evaluated in vitro for fluorescence imaging after Cy5.5-cetuximab bioconjugate labeling. The location of fluorescence observed in labeled cells was significantly different for cell lines that responded to treatment, relative to unresponsive cells. Tumors from cell lines that showed low internalized signal in vitro responded best to treatment with cetuximab. This preclinical model may aid in determining which cancer patients are best suited for cetuximab therapy.

Introduction

Head and neck squamous cell carcinoma (HNSCC) affects more than 40,000 people in the United States every year. Despite aggressive medical and surgical treatment, functional outcomes and overall survival have remained poor for patients with advanced stage disease.^1–4 The recent development of antiepidermal growth factor receptor (EGFR) targeted therapy has generated potential for improved outcomes without additional treatment related morbidity.

Cetuximab (Erbitux; ImClone Systems, Inc., New York, NY) is the first anti-EGFR targeted agent approved for the treatment of HNSCC both in combination with radiation⁵ or as single agent therapy for patients with platinum-resistant metastatic or recurrent disease.⁶ Unfortunately, only a small fraction of head and neck tumors respond to anti-EGFR therapy⁷ and predicting which patients will benefit from treatment remains a challenge. Considerable cost is incurred with cetuximab treatment for patients who in the end fail to respond.⁵ This has led investigators to evaluate markers which could predict treatment response.

Previous studies have demonstrated that EGFR expression does not correlate with response to cetuximab therapy.⁶ In fact one study demonstrated response in tumors without EGFR expression.⁸ A correlation between treatment response and the development of a low grade acneiform rash has been reported,⁵ however by definition this occurs only after treatment has been initiated. Antibody penetration and subsequent response to treatment likely depends on a variety of intra-tumoral factors that are difficult to measure including tumor blood flow, intratumoral hydrostatic pressures and receptor availability to the targeting antibody. We hypothesize that tracking antibody-tumor interactions in vivo by administration of fluorescently labeled cetuximab can predict tumor response to cetuximab therapy. Furthermore, we propose to measure intracellular trafficking of the antibody to determine if this can predict response.

Results

In vivo assessment of tumor fluorescence.

Mice bearing HNSCC flank xenografts (SCC-1, UM-SCC-5, UM-SCC-22A, FaDu and CAL27) were divided into control and cetuximab treatment groups and tumor growth was measured over the course of 4 weeks (n = 10 per cell line, except for CAL27 where n = 8). These cell lines were chosen based on their ability to grow successful tumors in mice. At the end of treatment, SCC-1 (19% of control, p < 0.001), UM-SCC-5 (28% of control, p < 0.001), UM-SCC-22A (29% of control, p = 0.016) and FaDu (57% of control, p = 0.007) tumors were significantly smaller than untreated controls (Fig. 1A). CAL27 tumors were not significantly larger than controls (110% of control, p = 0.90) and were considered “non-responders” to cetuximab therapy. To evaluate whether treatment response could be characterized by tumor fluorescence intensity, mice with untreated head and neck cancer xenografts were systemically injected with Cy5.5-cetuximab and imaged 72 hours later using fluorescence stereomicroscopy. SCC-1 tumors demonstrated the greatest maximum fluorescence (36 AU) followed by FaDu (27 AU), CAL-27 (9 AU), UM-SCC-5 (5 AU) and UM-SCC-22A (1 AU). Although maximal fluorescent intensity varied among xenografts, intensity of whole tumor fluorescence as measured in vivo did not correlate with response to cetuximab therapy (p = 0.80, Fig. 1B). In addition, EGFR expression (as determined by flow cytometry) was not found to be predictive of treatment response (data not shown).

Figure 1.

Response to treatment and maximal fluorescence of HNSCC flank xenograft in vivo. (A) Flank xenograft response to treatment with cetuximab. Treated xenografts responded to cetuximab with decreased growth except CAL27 (110% of control, p = 0.90). SCC-1 (19% of control, p < 0.001), UM-SCC-5 (28% of control, p < 0.001), UM-SCC-22A (29% of control, p = 0.016) and FaDu (57% of control, p = 0.007) tumors were significantly smaller than untreated controls. (B) Maximum fluorescence including representative fluorescent images of xenografts from each cell line. Each cell line exhibited a unique level of fluorescence (SCC-1 = 36 AU, FaDu = 27 AU, UM-SCC-22A = 1 AU), except CAL27 and UM-SCC-5, which had similar fluorescence intensity (9 and 5 AU, respectively; p = 0.08). Error bars indicate standard errors of the mean.

Antibody internalization corresponds with cetuximab treatment response.

In order to determine if cellular processing of the anti-EGFR monoclonal antibody played a role in tumor response to cetuximab therapy, the same cell lines were evaluated in vitro. Untreated SCC-1, FaDu, CAL27, UM-SCC-5 and UM-SCC-22A cells were incubated with Cy5.5-cetuximab and imaged by fluorescence microscopy. All cell lines demonstrated membrane-bound and internalized cetuximab monoclonal antibody, however internalization levels varied greatly between the five different cell types (Fig. 2). In order to simplify and create repeatable criteria for evaluation of the flourescent patterns observed in these cell lines, the signal was categorized as ‘internalized’ only if it showed one large area of fluorescence intensity within the cell. CAL27 displayed the highest level of antibody internalization (68%) while SCC-1 (15%), FaDu (19%), UM-SCC-5 (24%) and UM-SCC-22A (12%) demonstrated lower levels of internalized cetuximab monoclonal antibody. CAL27 demonstrated a significantly higher percentage of antibody internalization when compared to all other cell types (p = 0.015). When cells were isolated from untreated HNSCC xenografts and grown in culture, antibody internalization levels were maintained (Fig. 3). Ex vivo and in vitro CAL27 cells labeled with Cy5.5-cetuximab demonstrated similar internalization of the cetuximab monoclonal antibody (Fig. 3). High levels of antibody internalization appeared to correlate with resistance to cetuximab therapy (p = 0.027).

Figure 2.

Percentage of internalized signal in vitro. Each cell line was evaluated in vitro for cetuximab-Cy5.5 labeling. Cells exhibiting an area of concentrated fluorescence were counted and that number was divided by the total number of cells in the image. CAL27 cells show a significantly higher percentage of internalized label than do the other four cell lines (p = 0.015). Error bars indicate standard errors of the mean.

Figure 3.

Cells labeled in vitro with cetuximab-Cy5.5 exhibit patterns unique to a specific cell line. This pattern is conserved even after cells have been grown in vivo and harvested from the xenograft ex vivo.

Discussion

Although many patients with HNSCC benefit from cetuximab therapy, there is a subset of patients that fail to respond to treatment.^5,7–9 Researchers have thus far failed to demonstrate a consistent marker which will predict treatment response.^5,6,10,11 The ability to identify the likelihood of a positive response to cetuximab treatment would prevent patients not likely to respond to this treatment from receiving an unnecessary agent. In the present study we demonstrate that antibody internalization may be used to determine response to cetuximab therapy. Tumor cell lines that had less internalization of labeled antibody in vitro had a greater response to cetuximab treatment in tumor xenografts in vivo.

Consistent with findings from other investigators,^6,9,10 EGFR expression did not correlate with treatment response. Further studies are required to determine the significance of receptor internalization—specifically, of the Cy5.5-cetuximab bioconjugate. One possible explanation is that certain cell lines sequester the antibody through receptor internalization, thus rendering it less effective at inhibiting EGFR signaling. Without the blocking effects the antibody is designed to produce, the tumor continues to grow unimpeded.

The labeling patterns that we observed in these cell lines were consistent throughout multiple imaging sessions, and with multiple human observers. Each cell line exhibits an individual, predictable pattern when labeled with cetuximab. It is interesting that the four types of xenografts that showed a positive response to treatment also had a similar labeling pattern when the cells were labeled directly in vitro. This argument would be stronger if we had more than one type of xenograft show a lack of response to cetuximab treatment in combination with a pattern of labeling similar to that of CAL27 cells. None the less, these results are encouraging and may be the first step towards elucidating a reliable marker for treatment response. Because Cy5.5 has not been FDA approved for clinical use, imaging a tumor directly with the bioconjugate Cy5.5-cetuximab would currently be impossible in a clinical setting. However, isolating cells from a tumor biopsy and labeling them directly in vitro would be an excellent alternative, and likely be more informative without exposing the patient to additional therapies.

It is possible that the lack of whole tumor in vivo fluorescence seen in the CAL27 xenografts we initially observed is a result of antibody compartmentalization. However, UM-SCC-5 and UM-SCC-22A xenografts showed a similar lack of in vivo fluorescence, yet they both displayed a positive response to treatment with cetuximab. The outcome of labeling cells directly in vitro seems to suggest a correlation between the frequency of internalized antibody signal and response to treatment. Scatchard assay results demonstrated that binding affinity of cetuximab (labeled with Cy5.5 and unlabeled) is very high in SCC-1 cells, and essentially irreversible (data not shown). This could help explain the high fluorescence and treatment response seen in this group in vivo.

The graded values of fluorescence observed in xenograft images suggest the possibility of using fluorescence intensity to characterize tumors. Unfortunately this is not likely feasible due to a number of issues including: the influence of tumor size on the intensity of fluorescence, the location of many tumors and the differences in tissue density and depth. Additionally, cetuximab is typically used in combination with radiation or traditional chemotherapy.^5,9 Because we evaluated cetuximab alone, it is unknown whether this data will hold true in combined therapy. While we have yet to determine the precise impact of tumor size on fluorescence, we have demonstrated that the tumor characteristics affecting intensity of fluorescence go beyond size alone.

We demonstrate that fluorescently labeled cetuximab can be used to predict tumor response to anti-EGFR therapy. Internalization of the antibody appears to confer resistance to treatment. This pre-clinical model may aid in determining which patients are best suited for cetuximab therapy. Further, confocal imaging instruments are now available with small diameter probes (0.3–4.2 mm), such as the Leica FCM1000 endoscopic confocal microscope. The subcellular resolution of this instrument would enable detection of internalized cetuximab.

Materials and Methods

Reagents.

Cetuximab (Erbitux, ImClone Systems, Inc., New York, NY) is a recombinant human/mouse chimeric monoclonal antibody that binds specifically to the extracellular domain of human EGFR. It is composed of the Fv regions of a murine anti-EGFR antibody with human IgG1 heavy and kappa light chain constant regions and has an approximate molecular weight of 152 kDa.

Cy5.5 (CyDye deoxynucleotides, GE Healthcare, Piscataway, NJ) was used as the far-red fluorescent marker (the Cy5.5 bis-functional reactive dye that was used is in monofunctional NHS-ester form). It displays a broad absorption spectrum with maximum absorption at 683 nm. When coupled to IgG, Cy5.5 emits at a maximum of 707 nm and has a relative quantum yield of 0.28. Cetuximab was incubated with Cy5.5 reactive dye in 0.15 mol/L phosphate buffer (pH 7.8) for 1.5 hours and non-conjugated Cy5.5 was removed by a Centricon Centrifugal Filter Unit, YM-30 (Millipore, Billerica, MA). Mass spectrometry was used to determine the ratio of antibody to Cy5.5, on average 1:4.5.

Cells and culture conditions.

Five human HNSCC cell lines were evaluated: FaDu (ATCC, Manassas, VA), SCC-1 (Thomas Carey, University of Michigan, Ann Arbor, MI), CAL27 (ATCC, Manassas, VA), UM-SCC-5 and UM-SCC-22A (Kevin Raisch, University of Alabama at Birmingham, Birmingham, AL). Cells were maintained in complete media [Dulbecco's modified Eagle media (DMEM, Mediatech, Herdon, VA) supplemented with 10% (v/v) fetal bovine serum (FBS, Hyclone, Logan, UT) and 1% penicillin-streptomycin solution (10,000 units/mL penicillin and 10,000 µg/mL streptomycin, Mediatech, Herdon, VA)] in a 37°C humidified atmosphere containing 5% CO₂. Cells were labeled for fluorescent imaging by incubating each cell line for 1 hour with 1 µg/mL Cy5.5-cetuximab in complete media without phenol red at 37°C. Cells were then washed three times in the same media and imaged for Cy5.5 fluorescence 0, 4, 24 and 48 hours later.

Flow cytometry was performed on all cell lines to evaluate EGFR expression using PE mouse anti-Human EGF Receptor antibody (BD Biosciences, San Jose, CA). PI/RNase staining buffer (BD Biosciences, San Jose, CA) was used to control for cell death.

Animal models.

Severe combined immunodeficient (SCID) mice, age 4–6 weeks (Charles River Laboratories, Wilmington, MA and NCI-Frederick, Frederick, MD) were obtained and housed in accordance with the University of Alabama at Birmingham's Institutional Animal Care and Use Committee (IACUC) guidelines. All experiments and euthanasia protocols were approved by IACUC.

SCC-1, FaDu, CAL27, UM-SCC-5 and UM-SCC-22A cell lines (2 million cells per injection) were used to generate subcutaneous flank xenografts in SCID mice. Mice (n = 10 per cell line, except for CAL27 where n = 8) were divided into control and cetuximab treatment groups, which received 250 µg cetuximab by intraperitoneal injection once weekly for 4 weeks. Tumors were measured with digital calipers twice weekly for 5 weeks and tumor size was determined by multiplying the two largest diameters measured. Prior to treatment, average tumor size was 42 mm² and there was no statistical difference between control and treatment groups (p = 0.74). The treatment schedule was based on previous experiments using combined therapy¹² and designed to ensure that tumors persisted after completion of therapy to allow for fluorescence imaging. After completion of therapy, mice were systemically injected with 50 µg Cy5.5-cetuximab and imaged 72 hours later. Measurements were taken of the overlying skin following euthanasia to determine if differences in tumor depth affected fluorescent values. All tumors were between 0.75 and 0.86 mm deep, with only overlying skin covering subcutaneous tumors.

Imaging.

Tumor xenografts were imaged using a custom-built Leica MZFL3 fluorescent stereomicroscope (Leica Microsystems, Bannockburn, IL) fitted with a Cy5.5 filter and an ORCA ER charge coupled device (CCD) camera (Hamamatsu, Bridgewater, NJ) to allow for real-time imaging of Cy5.5 fluorescence. A Cy5.5 filter (Chroma filter set 41023) provided excitation between 630 and 670 nm and emission from 685 to 735 nm. The angle of incidence for imaging was 90°. Gross, brightfield and fluorescent images were obtained for each tumor. All analyses were done using 600 ms exposure images. Fluorescence was quantified after digital capture using ImageJ (http://rsb.info.nih.gov/ij/). The mean fluorescence intensity of the tumor was determined by outlining the tissue as a region of interest in ImageJ and then subtracting the background fluorescence.

To evaluate in vitro fluorescence, cell lines were imaged with a Nuance spectral camera connected to a Leica DMIRE2 inverted microscope (Leica Microsystems, Bannockburn, IL) fitted with a Cy5.5 filter different than the above-mentioned Cy5.5 filter. Cells exhibiting concentrated internalized antibody were counted and divided by the total number of cells per image to determine the percentage of cetuximab internalization.

Cell Isolation from xenografts.

Tumor cells were isolated from untreated xenografts to evaluate the future efficacy of applying cell labeling and imaging techniques to human tissue biopsies and to strengthen the concept that antibody internalization promotes cetuximab resistance as seen in vitro. Following euthanasia, untreated HNSCC xenografts were isolated, minced and allowed to incubate for 2 hours at 37°C in 2 mg/mL Type I collagenase in DPBS (Dulbecco's modified phosphate buffered saline, pH 7.4). After incubation, tumors were washed and cultured in 12 well tissue culture-treated plates containing complete media. Media was changed after 24 hours, observable tissue was removed and remaining cells were allowed to grow for three to seven days. At the end of this growth period, cells were treated with Cy5.5-cetuximab in the manner described above and imaged to evaluate antibody internalization patterns.

Statistical analyses.

Fluorescent values obtained from in vivo specimens, as well as tumor sizes were compared using a single factor ANOVA (SAS® version 9.2). The bias present between fluorescent values, caliper measurements and antibody internalization is expressed as standard error of the mean. Linear relationships between treatment response and fluorescence and internalization of fluorescence were evaluated with Pearson correlation coefficients. An alpha level of p < 0.05 was considered significant.