¹⁸F-FDG PET SUVmax Correlates with Osteosarcoma Histologic Response to Neoadjuvant Chemotherapy: Preclinical Evaluation in an Orthotopic Rat Model

Osteosarcoma Model and Treatment Schedule

Care of and procedures for animals were performed according to institutional and national guidelines. Animals were anesthetized throughout all surgical and imaging procedures with isoflurane/oxygen (2.5%/2.5%, v/v) (Minerve). The transplantable orthotopic and metastatic rat osteosarcoma has been described elsewhere (14,16). This model mimics its human counterpart in terms of aggressiveness, metastatic spreading, and chemoresistance phenotype (15,16). Tumors were grafted on 25-d-old Sprague–Dawley rats (Charles River Laboratories) as previously described (14). Fourteen days after tumor transplantation, animals underwent a first ¹⁸F-FDG PET scan and were randomly assigned to a control (saline, n = 10) or a treatment (ifosfamide; Baxter) group (n = 10). Treated animals received 2 subcutaneous doses (20 mg/kg each) of ifosfamide 7 d apart (at days 15 and 22 after tumor transplantation). A second ¹⁸F-FDG PET scan was obtained 7 d after the second administration of ifosfamide (i.e., day 29 after tumor implantation). Total tumor volume and metabolic volume were measured using a volume of interest (VOI)–based method on reconstructed PET images. All animals were euthanized if they showed any signs of distress. At the time of necropsy, tumors and lungs were collected for histologic examinations.

Metabolic Response to Chemotherapy Using ¹⁸F-FDG Small-Animal PET Scan

The initial ¹⁸F-FDG PET scan was obtained 14 d after tumor implantation (i.e., 24 h before treatment initiation, or prechemotherapy PET). A second ¹⁸F-FDG PET scan was acquired 7 d after the completion of chemotherapy (i.e., 29 d after tumor transplantation, or postchemotherapy PET). After the animals had fasted for 4 h, they received intravenous injections of ¹⁸F-FDG (30 MBq/kg; CERMEP) 2 h before image acquisition. During acquisition, animals were kept under general anesthesia with isoflurane/oxygen (2.5%/2.5%, v/v), and animals' body temperature was maintained using a warming pad. PET scans were obtained at the Animage core facility on a small-animal PET camera (Clearpet; Raytest). Whole-body images were acquired with 2 bed positions (field of view, 110 mm): one centered on the tumor and the other one on the lungs, with an acquisition time of 15 min per bed position.

Data were processed without attenuation correction, and images were reconstructed using an iterative method (maximum-likelihood expectation maximization, 20 iterations) using the Amide software (http://amide.sourceforge.net). ¹⁸F-FDG PET images were semiquantitatively analyzed using a VOI-based method. Ellipsoid VOIs were drawn over the whole tumor. Three-dimensional isocontour VOIs using a 40% maximum threshold defined the total tumor volume. Similarly, a 50% maximum threshold defined the metabolic tumor volume. The 40% and 50% threshold values applied to define the total tumor volume and the metabolic tumor volume in 3-dimensional isocontour VOIs were chosen on the basis of published studies (17,18). As internal negative control, a reference nontumor VOI was positioned in a corresponding location to the tumor region of interest (e.g., in the contralateral unaffected paw). An SUVmax normalized to body weight was calculated for all volumes using the formula (18):

The 50% and 40% threshold values allowed us to calculate a metabolic volume corresponding to all the voxels having an SUV greater than or equal to 50% and 40% threshold values and to give an SUV mean value for the concerned volume. With this method, the necrotic areas (for which SUVs were below the threshold values) did not affect the metabolic volume measure.

According to recommendations by the European Organization for Research and Treatment of Cancer (EORTC) (19), we graded metabolic responses on the basis of the changes of SUVmax measured on prechemotherapy PET and postchemotherapy PET reconstructed images (Table 1).

Histology

All animals were euthanized 24 h after the postchemotherapy PET scan. Primary tumor and lungs were fixed in 10% buffered formalin and then processed and embedded in paraffin. Viable tumor remaining after chemotherapy was expressed as the percentage of whole tumor volume using the Huvos histologic response (Table 2) (4,20). Tumors were oriented, and series of whole transverse sections were cut in the distal fourth, middle, and proximal fourth of the tumor. After hematoxylin and eosin coloration, tissue slides were analyzed using a DM4500 B microscope (Leica). We examined all cases to evaluate the following histologic features: mitotic rate expressed as the number of mitotic figures per 10 high-power fields (1 field, 0.237 mm²), necrosis, and bone. For each tumor, mitotic rate and necrosis were estimated on whole transverse sections from the 3 areas (e.g., distal fourth, middle, and proximal fourth of the tumor).

Immunohistochemistry

Glucose transporter 1 (Glut-1) immunostaining was performed on deparaffinized tumor sections with an anti–rat Glut-1 rabbit polyclonal antibody (Abcam; 1-h incubation at room temperature). Slides were then washed twice in 0.3% phosphate-buffered saline containing bovine serum albumin for 5 min. Slides were incubated with a biotinylated secondary antibody (Vector Laboratories) according to the manufacturer's instructions. After a final wash, tumor sections were stained with a Vectastain ABC Kit (Vector Laboratories), followed by counterstaining using hematoxylin (Sigma-Aldrich) and microscopic examination. Glut-1 staining was graded as positive or negative. Cases were considered negative when less than 10% of cells showed Glut-1 staining and positive when 10% or more of tumor cells showed Glut-1 staining. Variations in staining intensity of the cells were scored, and the following criteria were used: +, weak but unequivocal staining in some cells; ++, staining of moderate intensity; and +++, strong or intense staining.

Statistical Analysis

SUVs and tumor volumes were compared between control and treatment groups using a 2-tailed Student t test.

Relative mean tumor volume (RMTV) and tumor volume inhibition rate (IR) were calculated as indicated below:P values less than 0.01 were considered statistically significant. Correlation between tumor responses determined by histologic analysis or by ¹⁸F-FDG PET was assessed using the Pearson test. Statistical analysis was performed using the StatView 5.0 software package (SAS Institute Inc.).

Ifosfamide Induces Clinical and Histologic Tumor Response in Orthotopic Osteosarcoma in Rats

There was no difference in the mean (±SD) tumor volumes measured on ¹⁸F-FDG PET reconstructed images at the initiation of treatment between control and treated animals (1,909 ± 515 mm³ and 2,559 ± 1,046 mm³, respectively; P = 0.09). After the end of treatment, the mean tumor volume in the treated animals was significantly smaller than that in the control group (1,251 ± 701 mm³ vs. 9,172.1 ± 2,693 mm³; P < 0.01). Compared with the control group, in the treatment group the 2 cycles of ifosfamide induced a 53.34% decrease in tumor volume, which corresponded to a 10.2-fold inhibition rate (P < 0.01) (Table 3). According to Huvos histologic response, animals from the control group were all classified as grade 3 or 4 nonresponders (Table 4). In treated animals, histologic evaluation found only good (grade 1, n = 5) and partial responders (grade 2, n = 5) (Table 4).

¹⁸F-FDG PET Correlates with Histologic Observations in Osteosarcoma in Rats

For all animals, the SUVmax was observed within the tumor metabolic volumes. Before treatment initiation, the metabolic volume represented 86.9% (±13%) and 94.5% (±11%) of the whole tumor volume in the control and treatment groups, respectively (Table 3). After treatment, the metabolic volume still represented approximately 60% of tumor volume in both groups (Table 3). For all control animals (representative images shown in Fig. 1), when ¹⁸F-FDG PET images were superimposed onto the corresponding whole-mount tumor sections, areas of metabolic volume defined on ¹⁸F-FDG PET images were found to match with the histologically defined hyperproliferative regions (Figs. 1A and 1B). On histologic analysis, these areas presented with the highest cell density, an elevated mitotic score, and a strong staining of Glut-1 (Figs. 1C and 1E). Conversely the areas with the lowest SUV corresponded to tumor regions with necrotic or osteoid areas with low cellular density and weak Glut-1 staining (Figs. 1D and 1F).

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 1. 

Representative 3-dimensional maximal-intensity-projection reconstruction obtained for control group. Tumor coronal (A) and transaxial (B) slices visualized by ¹⁸F-FDG PET. Areas with higher fixation of ¹⁸F-FDG corresponded on hematoxylin- and eosin-stained tumor transverse sections (C) to regions with high cellular density, mitotic score, and high homogeneous expression of Glut-1 visualized by immunohistostaining (E). Areas with lower fixation of ¹⁸F-FDG corresponded on hematoxylin- and eosin-stained tumor transverse sections (D) to necrotic areas, or regions with low cellular density, mitotic score, and lower and heterogeneous expression of Glut-1 visualized by immunohistostaining (F). Ms = muscle; T = tumor; Ts = testis. All images are obtained from same rat. Magnifications are ×200 for C, D, and F; ×100 for E.

Among the treated animals, ¹⁸F-FDG PET scans obtained after the completion of treatment revealed tumors with areas of intense ¹⁸F-FDG fixation (Figs. 2A and 2B) amid areas of lower intensity (Figs. 2A and 2E). The superimposition of ¹⁸F-FDG PET transverse sections with the corresponding whole-mount paraffin slides revealed that the areas showing the least ¹⁸F-FDG fixation intensity (i.e., areas with SUV < SUV mean) corresponded on histologic analysis to necrotic areas. Interestingly, areas of active metabolic activity as visualized by ¹⁸F-FDG PET corresponded on histologic slides to tumor foci as small as 1.2 mm in diameter (Fig. 2C). These remaining tumor foci presented with the same characteristics as the metabolic volume areas of tumors from the control group: high cell density, high mitotic index, and intense expression of Glut-1, all being evidence of the aggressiveness of remaining tumor cells (Figs. 2C and 2D). Areas presenting with an intermediate fixation of ¹⁸F-FDG (areas with SUVs approaching the SUV mean) on ¹⁸F-FDG PET were found to correspond histologically to regions of chemotherapy-induced modifications (Figs. 2E and 2F). No viable tumor cells were found in these areas, but fibrous scarring and infiltration by immune and giant cells were observed (Fig. 2F). These cells presented a Glut-1 staining of moderate intensity. Indeed, normal immune cells are known to express Glut-1, albeit at a lower level than tumor cells, which may explain this nonspecific ¹⁸F-FDG uptake (Fig. 2G).

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 2. 

Representative 3-dimensional maximal-intensity-projection coronal reconstruction of ¹⁸F-FDG PET images from ifosfamide-treated group (A). ¹⁸F-FDG accumulates heterogeneously in tumor. Transaxial slice from region of 1.2-mm thickness showing high levels (B) of ¹⁸F-FDG uptake corresponded on hematoxylin and eosin histologic analysis to remaining hyperproliferative osteosarcoma (C) with strong Glut-1 immunostaining (D). Transaxial slice from region showing low and heterogeneous fixation of ¹⁸F-FDG (E) corresponded on hematoxylin and eosin histologic analysis to chemotherapy-modified regions without viable tumor cells but with infiltration of giant polynuclear immune cells (arrows) (F) expressing Glut-1 (G). Ms = muscle; T = tumor; Ts = testis. Magnifications: ×50 for C; ×100 for D and F; ×400 for G and insets from C and F.

Evaluation of Tumor Response by ¹⁸F-FDG PET

There was no correlation between changes in tumor volumes (whole or metabolic; Table 4) and histologic responses (Fig. 3), even if a 55% reduction in tumor volume could identify good responders. When applying the criteria of the EORTC to evaluate the tumor metabolic response in our tumor model, all tumors in the control group qualified as progressive disease, whereas all animals in the treated group were considered partial responders (Table 4). No correlation was found between the metabolic response graded according to the recommendations by the EORTC and the histopathologic gradation. Thus, using the criteria of the EORTC for SUV changes and staging of metabolic response, we could not discriminate good from partial responders (Table 4). A closer analysis of the SUVmax revealed that a 40% decrease in the SUVmax between prechemotherapy and postchemotherapy ¹⁸F-FDG PET could differentiate good from partial responders with a 100% sensitivity and specificity (Fig. 4A). We also found that an SUVmax less than 15 on ¹⁸F-FDG PET performed after chemotherapy distinguished good from partial responders (Fig. 4A). Indeed, 5 of the 5 good responders had a postchemotherapy SUVmax less than 15, whereas animals presenting a partial response to ifosfamide had an SUVmax between 15 and 20 (P < 0.01). The SUVmax in rats with progressive tumors (control group, n = 10) was above 20; the most aggressive tumor presented with an SUVmax greater than 25 (P < 0.01). Interestingly, there was a significant correlation between the SUVmax on ¹⁸F-FDG PET after chemotherapy and histologic response (P < 0.01, R² = 0.94) (Fig. 4B).

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 3. 

Scatter plot of histologic response and total tumor volume changes measured using ¹⁸F-FDG PET shows absence of correlation between parameters measured by metabolic imaging and histologic response (R² = 0.56). Tumor volumes were measured by VOI on maximal-intensity-projection reconstructed images.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 4. 

Correlation of tumor histologic response (Huvos) with SUVmax measured by ¹⁸F-FDG PET. (A) Scatter plot of relative SUVmax changes and tumor histologic response shows strong correlation of SUVmax change with histologic response (R² = 0.86): a 40% decrease in SUVmax could distinguish good from partial responder (grade 1 from grade 2). (B) Scatter plot of SUVmax measured by ¹⁸F-FDG PET scan after chemotherapy correlates with histologic grade (R² = 0.94). Postchemotherapy SUVmax less than 15 indicated good response to chemotherapy. For this analysis, histologic response, relative SUVmax change, and postchemotherapy SUVmax from all animals studied were considered.