2.1. GI-NEN cases
Surgical specimens of GI-NENs and their metastatic lesions (lymph nodes and liver) from 2002 to 2019 were retrieved from surgical pathology files at Tohoku University Hospital (Sendai, Japan), Aichi Prefectural Cancer Center Hospital (Nagoya, Japan), Noe Hospital (Osaka, Japan), Tokyo Medical and Dental University Hospital (Tokyo, Japan), and Kansai Electric Power Hospital (Osaka, Japan). 10% formalin fixed and paraffin embedded tissue blocks (FFPE) were available for this study in the specimens from Tohoku University Hospital, but only unstained serial tissue slides for immunohistochemistry and hematoxylin and eosin stain were available in those from other institutions above. The clinicopathological features of these NEN cases were summarized in Table 1. The research protocol of this study was approved by the institutional review boards of Tohoku University Graduate School of Medicine (2020-1-7) and the institutions above.
Serial tissue sections of FFPE specimens were used for subsequent analyses. The cases included 70 GI-NET and 14 GI-NEC. The GI-NET cases were tentatively classified into foregut (n = 33) and hindgut (n = 37) GI-NETs according to their primary sites. These cases were further classified into G1 (n = 49), G2 (n = 19), and G3 (n=2) tumors based on the grading criteria of the 2019 WHO Classification [24].
2.2. Immunohistochemistry/IHC
After carefully reviewing the available hematoxylin and eosin (H&E)-stained slides, one representative section including the tumor area in its greatest dimensions was selected in each case. Serial tissue sections were prepared at 3-μm thickness. The IHC protocols were summarized in Table 2.
Immunostained slides were digitally scanned using Nanozoomer S360 (C13220-01; Hamamatsu Photonics, Shizuoka, Japan) for the subsequent imaging analysis.
2.3. Evaluation of Ki-67 labeling index
The Ki-67 labeling index (LI) was determined according to the counting method defined by WHO in 2019 [24, 25], using the HALO image analysis software (Indica Laboratories, Corrales, New Mexico, USA) with the CytoNuclear IHC v1.6 algorithm module. We performed imaging analysis, according to a previously reported study using this digital data to obtain the Ki-67 LI [26]. We analyzed nuclear immunoreactivity according to the gradients of brown color (3,3-diaminobenzidine [DAB]) spectrum intensity. Tumor cells with blue nuclei were negative, whereas cells with yellow (weak intensity), orange (moderate intensity), and red (strong intensity) nuclei were positive for Ki-67 immunoreactivity. The labeling index or LI was calculated based on the following formula: Number of all stained cells regardless of immunointensity/number of tumor cells (hot spot, at least 500 cells). Representative images obtained before and after the analysis were illustrated in Figure 1-1.
2.4. Evaluation of MGMT immunohistoreactivity
MGMT immunoreactivity was detected in the nuclei and evaluated independently using two different scoring systems, i.e., MGMT score and H-score.
The H-score was obtained using the HALO image analysis software with the CytoNuclear IHC v1.6 algorithm module. Parameter of “Cell detection” (nuclear contrast, optical density, size, and shape) and the thresholds of immunoreactivity in each section were set according to those previously reported [26]. The image analysis was performed by a single observer in the average areas. The average area was randomly selected and analyzed by counting more than 500 cells. All the parameters were set individually in each case. In the selected annotation areas, the HALO software automatically calculated the number of positive cells with weak, moderate, and strong immunoreactivity among the total cells. Representative images of analytical procedures were illustrated in Figure 1-2. The H-score was subsequently calculated based on the following formula: Σ (individual gradients of the positive tumor cells/all tumor cells × Score 1+, 2+, 3+).
The MGMT score was obtained by manual/eyeball analysis performed by three of the authors (H.W., F.F., and H.S.). The MGMT score was determined according to a previously reported study [27], incorporating the proportion of positive nuclear immunoreactivity in tumor cells as follows: score 0, absence of immunoreactivity; score 1, nuclear immunoreactivity in less than 20% tumor cells; score 2, nuclear immunoreactivity in greater than 20% but less than 50% tumor cells; and score 3, immunoreactivity in greater than 50% tumor cells. Representative images of MGMT scores were illustrated in Figure 1-3.
2.5. Evaluation of GLUT2 immunoreactivity
We evaluated the status of GLUT2 immunoreactivity with a semiquantitative scoring system assessing both the proportion and relative immunointensity according to Kaemmerer et al [28], using manual or eyeball analysis performed by three of the authors (H.W., F.F., and H.S.). The proportion of immunopositive cells was tentatively classified into three different categories: proportion score 0, completely negative; 1, 1–50% cells positive; and 2, 51–100% cells positive. The relative immunointensity of positive cells was further sub-classified into 4 categories: intensity score 0, completely negative; 1, weak; 2, weak and strong (the tumor area presenting weak and strong immunointensity was respectively and simultaneously detected in more than 10% of positive tumor area); and 3, strong. Representative images for the GLUT2 intensity score were illustrated in Figure 1-3, and the GLUT2 score was subsequently calculated using the following formula: proportion score × intensity score.
2.6. Evaluation of MGMT promoter methylation with methylation-specific real-time PCR
49 FFPE specimens of GI-NENs including 35 NETs and 14 NECs were retrieved from the pathology files of Tohoku University Hospital. Serial tissue sections at 10-μm thickness were prepared following the macro-dissection of relevant tumor areas. DNA was extracted from these specimens above using the Cobas DNA preparation kit (Roche, Mannheim, Germany). Tumor DNA was treated with bisulfite using the MethylEasyTM Xceed Rapid DNA Bisulphite Modification Kit (Takara Bio Inc, Shiga, Japan), following the manufacturer’s instructions. 1.0μl of tumor DNA template (20ng/μl) treated with bisulfite was mixed with 10.0μl of LightCycler® 480 Probes Mater (Roche Diagnostics, Mannheim, Germany), 1.0μl of Forward Primer, 1.0μl of Reverse Primer, 1.0μl of TaqMan Probe and 6.0μl of H20. MGMT promoter methylation was evaluated by methylation-specific real-time PCR in a LightCycler 480 Real-Time PCR System (Roche) for preincubation (10 minutes, 95℃), amplification (15 seconds, 95℃ and 1 minute, 60℃) 50 cycles and cooling (30 seconds, 40℃), with reference to the method reported by Sonoda et al and Kitange et al [27, 29]. We used β-actin as Housekeeping gene. The primer sequences (obtained from Nihon Gene Research Laboratories INC, Sendai, Japan), 5′-TTCGCGGTGCGTATCGT-3′ (forward) and 5′-CACTCTTCCGAAAACGAAACGA-3′ (reverse), were used for the methylation reaction and 5′-TTTTATTTAGAGTGTAGGTGTGTGGAGATTTT-3′ (forward) and 5′-CAAAAACAAAAACCTAACCCCTAAACCT-3′ (reverse) for b-actin. The probe sequence 5′-FAM-ACACTCACCAAATCGC-MGB-3′ (TaqMan® MGB, Thermo Fisher Scientific, Tokyo, Japan) was used for the methylation reaction and 5′-FAM-CCCACCCTCTAAAACT-MGB-3′ (TaqMan® MGB) for β-actin. Cp Genome Universal Methylated DNA (Merck, Darmstadt, Germany) was used as MGMT methylation control DNA.
2.7. Statistical analysis
The differences of MGMT and GLUT2 immunoreactivity were analyzed using χ2 test or Mann-Whitney’s U test. The correlation between Ki-67 LI and MGMT (H- and MGMT scores) and GLUT2 (GLUT2 score) immunoreactivity was analyzed by Spearman’s test. P values of <0.05 were considered significant. The JMP Pro ver.14.3.0 software (SAS Institute, Inc., Cary, NC, USA) was used for statistical analysis.