Gut and Liver, Vol. 13, No. 3, May 2019, pp. 356-365 ORiginal Article

Preoperative Metabolic Tumor Volume2.5 Associated with Early Systemic Metastasis in Resected Pancreatic Cancer: A Transcriptome-Wide Analysis

Sung Hwan Lee1,2, Ho Kyoung Hwang2,3, Woo Jung Lee2,3, Mijin Yun3,4, and Chang Moo Kang2,3 1Department of Systems Biology, University of Texas MD Anderson Cancer Center, Houston, TX, USA, 2Division of Hepatobiliary and Pancreatic Surgery, Department of Surgery, Yonsei University College of Medicine, 3Pancreatobiliary Cancer Center, Yonsei Cancer Center, Severance Hospital, and 4Department of Nuclear Medicine, Yonsei University College of Medicine, Seoul, Korea

with MTV ≥4.5 demonstrated significantly different expres- See editorial on page 225. 2.5 sion of cancer-related reflecting aggressive tumor biol- ogy. Conclusions: Resectable pancreatic cancer with high 18 Background/Aims: F-fluorodeoxyglucose-positron emis- MTV2.5 is not only associated with lymph node metastasis but 18 sion tomography ( F-FDG-PET) reflects biological aggressive- also early systemic metastasis. The molecular background

ness and predicts prognoses in various tumors. Evaluating of resectable pancreatic cancer with high MTV2.5 may be as- the oncologic significance of the preoperative metabolic phe- sociated with aggressive biologic behavior, which might need notype might be necessary for planning the surgical strategy to be considered when managing resectable pancreatic can- in resectable pancreatic cancers. Methods: From January cers. Further study is mandatory. (Gut Liver 2019;13:356- 2010 to December 2015, a total of 93 patients with patho- 365) logic T3 (pT3) pancreatic cancer were included in this study. Clinicopathological parameters and PET parameters were Key Words: Pancreatic neoplasms; Positron-emission tomog- evaluated, and transcriptome-wide analysis was performed raphy; Metabolic phenotype; Transcriptome-wide analysis to identify the oncologic impact and molecular landscape of the metabolic phenotype of resectable pancreatic cancers. INTRODUCTION

Results: Preoperative metabolic tumor volume (MTV)2.5 was significantly higher in the pN1 group compared to the pN0 In the early 1920s, Otto Warburg reported that cancer cells

group (11.1±11.2 vs 6.5±7.8, p=0.031). Higher MTV2.5 val- preferentially metabolize glucose by “aerobic glycolysis,” even 1 ues (MTV2.5 ≥4.5) were associated with multiple lymph node in the presence of ample oxygen. The increased glucose de- metastasis (p=0.003), and the lymph node ratio was also mand has been considered one of the fundamental features of significantly higher in resected pT3 pancreatic cancer with cancer and has become the foundation of the clinical applica- 18 MTV2.5 ≥4.5 compared to those with MTV2.5 <4.5 (0.12±0.13 tion of F-fluorodeoxyglucose-positron emission tomography vs 0.05±0.08, p=0.001). Disease-specific survival of pa- (18F-FDG-PET) in oncology.2 Cancer cells show an increased up- 2 tients with MTV2.5 <4.5 was better than that of patients with take of glucose and FDG. Both glucose and FDG are phosphor-

MTV2.5 ≥4.5 (mean, 28.8 months; 95% confidence interval ylated by hexokinase (HK) to prevent their release from cancer [CI], 40.1 to 57.0 vs mean, 32.6 months; 95% CI, 25.5 to cells. In contrast to glucose, FDG cannot be further metabolized

39.7; p=0.026). Patients with MTV2.5 ≥4.5 who received due to the lack of the 2’-OH group, resulting in its accumulation postoperative adjuvant chemotherapy showed better survival in cells proportional to the amount of uptake and HK activity.3 outcomes than patients with MTV ≥4.5 who did not receive FDG-PET has emerged as an important clinical tool for cancer Received 는 2.5 Date Submitted 날짜로 adjuvant treatment in resected pT3 pancreatic cancers detection and staging and the monitoring of the response of 수정일은 마지막 날짜로 (p<0.001). Transcriptome-wide analysis revealed that tumors cancer to therapy.4 However, there are some important critiques 논평은 전부 삭제

1월 January Correspondence to: Chang Moo Kang (https://orcid.org/0000-0002-5382-4658)a and Mijin Yun (https://orcid.org/0000-0002-1712-163X)b 2월 February aDepartment of Surgery, Yonsei University College of Medicine, 50-1 Yonsei-ro, Seodaemun-gu, Seoul 03722, Korea 3월 March Tel: +82-2-2228-2100, Fax: +82-2-313-8289, E-mail: [email protected] 4월 April bDepartment of Nuclear Medicine, Yonsei University College of Medicine, 50-1 Yonsei-ro, Seodaemun-gu, Seoul 03722, Korea 5월 May Tel: +82-2-2228-6068, Fax: +82-2-312-0578, E-mail: [email protected] 6월 June Received on May 28, 2018. Revised on September 19, 2018. Accepted on October 1, 2018. Published online March 15, 2019 7월 July pISSN 1976-2283 eISSN 2005-1212 https://doi.org/10.5009/gnl18242 8월 August This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/4.0) 9월 September which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited. 10월 October 11월 November 12월 December Lee SH, et al: MTV2.5 in Resectable Pancreatic Cancer 357 of the routine use of FDG-PET in clinical oncology. It has been study protocol was approved by the Institutional Review Board observed that FDG-PET is not more effective than conventional of Severance Hospital with the waiver of informed consent due imaging technology at detecting early pancreatic cancer and to retrospective study (IRB number: 4-2017-0657). small metastatic lesions.5 In addition, FDG-PET is not able to 2. Clinicopathological parameters discriminate the oncologic process of pancreatic cancer from in- flammatory focal lesions including mass-forming pancreatitis.6 We retrospectively reviewed the medical records of the pa- However, with the clinical experiences of using FDG-PET in tients, and collected information, such as gender, age, tumor patients with pancreatic cancer, recent studies have suggested location, operation type, tumor size, tumor grade (differentia- that the characteristics of preoperative FDG-PET images are tion), pathologic tumor (pT) stage, presence of lymph node me- strongly associated with aspects of tumor biology,7 such as the tastasis, lymph node ratio (the total number of metastatic lymph pathological grade,8 recurrence,9 distant metastasis,10 and long- nodes divided by the total number of retrieved lymph nodes), term oncologic outcomes.11 Several parameters for FDG-PET are the presence of microscopic perineural invasion, the presence being developed to measure the metabolism of cancer lesions of lymphovascular invasion, the recurrence pattern, and the and to predict aspects of tumor biology and the resultant on- time to recurrence, which was defined as the time from surgical cologic outcomes. Unlike the maximum standard uptake value resection to recurrence or the last clinical follow-up visit at our

(SUVmax), which reflects the metabolic activity of a single pixel, medical center. the metabolic tumor volume (MTV) and total lesion glycolysis 3. PET/CT protocol and PET-related parameters (TLG) can measure the volume-based metabolic activity of the entire tumor lesion.11 In our previous study, the results of the 18F-FDG PET/CT scans were performed with the same protocol 12 propensity score-matched analysis demonstrated that preopera- described in our previous study. The SUVmax and MTV2.5 were tive MTV2.5 greater than 4.5 can predict tumor recurrence and measured on PET images using volume viewer software. Each long-term survival in patients with resected pancreatic cancer.12 tumor was examined with a spherical-shaped volume of interest In addition, it was shown that high FDG uptake in left-sided (VOI) that included the entire lesion in the axial, sagittal, and pancreatic cancer is strongly associated with loss of SMAD4, coronal planes. By using CT images, the 18F-FDG uptake values subsequently resulting in early systemic metastasis.13 From the of normal organs such as the bowel, stomach, and liver were viewpoint of the American Joint Committee on Cancer (AJCC) not included in the VOI. The SUVmax of the VOI was calculated 7th cancer staging system, pathologic T3 (pT3) lesions are the as (decay-corrected activity/tissue volume)/(injected dose/body most common type of resected pancreatic cancer, regardless of weight). MTV2.5 was defined as the total tumor volume with an the tumor size.14 Therefore, small proportions of T1 and T2 are SUV ≥2.5, and the MTV and mean SUV of the VOI were auto- potential confounding factors in the analysis of the oncologic matically calculated. The tumor to liver ratio (TLG) was calcu- impact of PET/computed tomography (CT)-related parameters in lated as (mean SUV) × (MTV). In patients with SUVmax <2.5, the patients with resected pancreatic cancer. MTV2.5 and TLG were not measured. The TLG was calculated as

In this study, we focused on patients who underwent po- the SUVmax/SUVmean of the liver parenchyma to adjust the values tentially curative pancreatectomy for pT3 pancreatic cancer of different PET scanners. without neoadjuvant treatment. We determined the value of 4. Sample preparation MTV2.5 in predicting the oncologic outcome of resected pancre- atic cancer, and investigated the differences in expression In total, six tumor samples from each pancreatic cancer pa- in patients with pT3 pancreatic cancer to explain the different tient were used for the transcriptomics study. Three samples oncologic outcomes according to the value of MTV2.5. This ap- were randomly selected from both the MTV high (MTV2.5 ≥4.5) proach may be a basis for a patient-oriented surgical approach group and the MTV low (MTV2.5 <4.5) group. Total RNA was to the treatment of resectable pancreatic cancer in the near fu- extracted from each formalin-fixed paraffin-embedded sample ture. using the TRIzol reagent (Molecular Research Center, Inc., Cin- cinnati, OH, USA) according to the manufacturer’s instructions. MATERIALS AND METHODS Following homogenization, 1 mL of solution was transferred to a 1.5-mL Eppendorf tube and centrifuged at 12,000 g for 10 1. Study period and population minutes at 4°C to remove any insoluble material. The super- From January 2010 to December 2015, 232 patients under- natant containing the RNA was collected, mixed with 0.2 mL went potentially curative pancreatectomy for pancreatic cancers of chloroform, and centrifuged at 12,000 g for 15 minutes at in Yonsei University Severance Hospital, Seoul, Korea. Of those 4°C. After nucleic acids in the aqueous phase were transferred 232 patients, 93 patients with pT3 pancreatic cancer (AJCC 7th) to a new tube, the RNA was precipitated by adding 0.5 mL of without neoadjuvant treatment who had undergone preopera- isopropyl alcohol and then recovered by centrifuging the tube tive PET/CT imaging were included in this investigation. This at 12,000 g for 10 minutes at 4°C. The RNA pellet was washed 358 Gut and Liver, Vol. 13, No. 3, May 2019 briefly in 1 mL of 75% ethanol and centrifuged at 7,500 g for mean target intensity of each array was arbitrarily set to 100. 5 minutes at 4°C. Finally, the total RNA pellet was dissolved The normalized and log transformed intensity values were then in nuclease-free water, and its quality and quantity were as- analyzed using GeneSpring GX 12.6.1 (Agilent Technologies, sessed on an Agilent Bioanalyzer 2100. was Santa Clara, CA, USA). Fold change filters included the require- analyzed with the GeneChip® U133 Plus 2.0 ment that the genes be present in at least 150% of the controls Array (Affymetrix, Santa Clara, CA, USA), which is composed for upregulated genes and fewer than 66% of the controls for of over 45,000 probe sets representing approximately 38,500 downregulated genes. Through hierarchical clustering, data well-characterized human genes. For each gene, eleven pairs of were clustered into groups that behaved similarly across experi- oligonucleotide probes are synthesized in situ on the arrays. ments using GeneSpring GX 12.6.1 (Agilent Technologies). The clustering algorithm used the Euclidean distance with average 5. Microarray linkage. Biotinylated cRNA was prepared according to the standard 6. Transcriptomics data analysis Affymetrix protocol from 200 ng of total RNA (Expression Analysis Technical Manual, 2001; Affymetrix). Following frag- The transcriptomics data obtained from the Affymetrix U133 mentation, 3 µg of RNA were hybridized for 16 hours at 45°C Plus 2.0 Arrays were used for the calculation of the fold ratio ® ® on the GeneChip Human Genome Array. The GeneChip was and p-value of gene expression levels between MTV2.5 high and washed and stained in the Affymetrix Fluidics Station 450. MTV2.5 low tumor samples. Genes showing significantly dif- ® ® The GeneChip was scanned using the Affymetrix GeneChip ferential expression between MTV2.5 high and MTV2.5 low tumor Scanner 3000 7G. The data were analyzed with Robust Mul- samples were selected and included in an integrative analysis tichip Analysis using the Affymetrix default analysis settings with ingenuity pathway analysis software (Qiagen Inc., Valen- and global scaling as the normalization method. The trimmed cia, CA, USA). The comprehensive analyses for canonical path-

A B p=0.031 p=0.168 10.0 Nodal metastasis 5 8.0 No Yes 30 7.0 6.0 25 4 5.0

(cm) 4.0 20

size 3.5 2.5 V 3 size 3.0 15

MT

2.5 umor

T

umor 10 T 2.0 2 1.75 1.5 5 1.2 R=0.212 0 1 1.0 0.01 0.02 0.05 0.1 0.2 0.5 12 5102050 No Yes No Yes MTV Nodal metastasis Nodal metastasis C 2.5 1.0

0.8 p=0.017

survival 0.6

0.4 pN0

Disease-free 0.2 pN1

0 20 40 60 Time (mo)

Fig. 1. Tumor size and MTV2.5 for nodal metastasis. (A) Correlation analysis for tumor size and MTV2.5. (B) Comparison of the difference in tumor size and MTV2.5 according to nodal metastasis. (C) The oncologic outcome of lymph node metastasis in resected pT3 pancreatic cancer. MTV, metabolic tumor volume; pT3, pathologic T3. Lee SH, et al: MTV2.5 in Resectable Pancreatic Cancer 359 ways, upstream elements, diseases and functions, and regulatory 26.7 months; 95% CI, 18.5 to 34.8 vs mean, 17.7 months; 95% effects were performed. CI, 11.4 to 24.0; p=0.017) (Fig. 1C).

7. Statistics 3. Preoperative detectable clinical parameter to predict lymph node metastasis in patients with resected pT3 Continuous variables were described as the means±standard pancreatic cancer deviation, and categorical variables were described as frequen- cies (%). The Student t-test, chi-squared tests with Fisher exact Only the value of preoperative MTV2.5 was significantly tests, and linear regression analyses were performed. The Pear- higher in patients with pN1 than in other patients (6.5±7.8 vs son correlation method was used to calculate the correlation 11.1±11.2, p=0.031). SUVmax was also marginally significantly coefficients. Survival curves were estimated using the Kaplan- different between the two groups; however, when adjusted

Meier method to calculate the cumulative recurrence-free sur- by the SUVmean of the liver (tumor to liver ratio), there was no vival rates. Statistical analyses were performed using SPSS 20.0 significant difference between two groups (1.9±0.8 vs 2.2±1.1, for Windows (IBM Corp., Armonk, NY, USA). p-values <0.05 p=0.231) (Table 1). were considered statistically significant. 4. Preoperative MTV2.5 ≥4.5 is significantly associated with lymph node metastasis in patients with resected pT3 RESULTS pancreatic cancer 1. Characteristics of patients with resected pT3 pancreatic The previously determined preoperative PET parameter12 of a cancer

From January 2010 to December 2015, 232 patients under- went radical resection for pancreatic cancer. The final study Table 2. Preoperative Positron Emission Tomography Parameters to population was 93 patients after excluding the patients who Predict Pathologic Outcomes in Resected pT3 Pancreatic Cancer underwent neoadjuvant treatment or palliative resection. The Pathological MTV2.5 mean age of the study population was 63.6±9.4 years; 53 (57%) p-value variable <4.5 ≥4.5 were males, and 40 (43%) were females. pN-stage 0.019 2. Oncologic outcome of lymph node metastasis in patients pN0 21 18 with resected pT3 pancreatic cancer pN1 16 38 Among the clinically available pathologic variables, such as Retrieved LNs 19.3±9.4 19.9±10.9 0.76 the tumor size, degree of differentiation, presence of lympho- Metastatic LNs 0.003 vascular invasion, presence of perineural invasion, presence of 0 21 14 lymph node metastasis, and resection margin, the Cox propor- 1 10 15 tional hazard model showed that lymph node metastasis was ≥2 8 25 the strongest prognostic factor predicting tumor recurrence LNR 0.05±0.08 0.12±0.13 0.001 (Exp(β)=1.868, p=0.019). There was a significant difference in Data are presented as number or mean±SD. disease-free survival (DFS) according to the pN stage (mean, pT3, pathologic T3; MTV, metabolic tumor volume; LN, lymph node; LNR, lymph node ratio.

Table 1. Preoperatively Available Parameters to Predict Lymph Node Metastasis in Resected pT3 Pancreatic Cancer pN-stage Clinical variable p-value pN0 (n=37) pN1 (n=56) Age, yr 64.3±9.2 63.2±9.6 0.597 Sex, male/female 23/14 30/26 0.413 Preoperative CA 19-9 575.6±2207.6 886.2±2343.4 0.58 Radiologic tumor size, cm 2.4±0.8 2.8±1.2 0.168 Tumor location, proximal/distal 23/14 38/18 0.572

SUVmax 4.5±2.1 5.5±2.6 0.055 TLR 1.9±0.8 2.2±1.1 0.231

MTV2.5 6.5±7.8 11.1±11.2 0.031 Data are presented as mean±SD or number. pT3, pathologic T3; SUVmax, maximum standard uptake value; TLR, tumor to liver ratio; MTV, metabolic tumor volume. 360 Gut and Liver, Vol. 13, No. 3, May 2019

value of MTV2.5≥4.5 was again found to have a close relation- sensitivity of 60.7% and a specificity of 71.5%. The same cutoff ship with lymph node metastasis in patients with resected pT3 value could discriminate those with early systemic metastasis pancreatic cancer (p=0.019). In addition, higher MTV2.5 values from those without with marginal significance in patients with

(MTV2.5≥4.5) in patients with pancreatic cancer was found to be resected pT3 pancreatic cancer (p=0.078). However, significantly associated with metastasis to multiple lymph nodes (chi-square, more patients with MTV2.5 ≥4.5 than those with MTV2.5 <4.5 had linear-to-linear association, p=0.003), and the lymph node ra- early systemic recurrence after initial radical pancreatectomy tio was also significantly higher in patients with resected pT3- (p=0.008) (Fig. 2B). Subsequently, the disease-specific survival pancreatic cancer with MTV2.5 ≥4.5 than in those with MTV2.5 of the patients with MTV2.5 <4.5 was much better than those

<4.5 (0.05±0.08 vs 0.12±0.13, p=0.001) (Table 2). with MTV2.5 ≥4.5 (mean, 28.8 months; 95% CI, 40.1 to 57.0 vs mean, 32.6 months; 95% CI, 25.5 to 39.7; p=0.026) (Fig. 3). 5. Association between MTV2.5 and tumor size according to nodal metastasis 7. Postoperative adjuvant chemotherapy attenuates the adverse oncologic impact of MTV ≥4.5 in patients with To elucidate the confounding effect of tumor size on MTV , 2.5 2.5 resected pT3 pancreatic cancer a correlation analysis and statistical comparisons of tumor size and MTV2.5 according to nodal metastasis were performed. Approximately 80% of patients who underwent curative There was no marked correlation (Pearson correlation coeffi- resection received adjuvant treatment based on gemcitabine cient=0.212) between MTV2.5 and tumor size (Fig. 1A). Though or fluorouracil (5-FU). Among them, approximately 15% of pa- the difference in tumor size according to nodal metastasis was not statistically significant (p=0.168), the value of MTV was 2.5 1.0 significantly higher in the nodal metastasis group than in the group without nodal metastasis (p=0.031) (Fig. 1B). 0.8

6. Preoperative MTV2.5 ≥4.5 is significantly associated with

survival MTV <4.5 early systemic metastasis in patients with resected pT3 0.6 2.5 pancreatic cancer 0.4 MTV2.5>4.5 Among the 93 patients with resected pT3 pancreatic cancer, 28 patients (30.1%) experienced early (<6 months after sur-

Disease-specific 0.2 gery) systemic recurrence. The mean time interval from radi- p=0.026 cal pancreatectomy to early systemic recurrence was 3.9±2.6 months. The most frequent distant metastatic site was found 0 20 40 60 to be the liver (16/28, 57.1%). pN stage could not predict early Time (mo) systemic recurrence in patients with resected pT3 pancreatic Fig. 3. Disease-specific survival according to the preoperative MTV2.5 cancer (p=0.146) (Fig. 2A). A lymph node ratio cutoff value of parameter in resected pT3 pancreatic cancer. 0.07 was shown to determine early systemic recurrence with a MTV, metabolic tumor volume; pT3, pathologic T3.

A B 1.0 1.0

pN0 MTV <4.5 0.8 0.8 2.5

pN1 0.6 0.6 MTV2.5 >4.5

p=0.146 recurrence-free p=0.008

survival 0.4 survival 0.4

systemic-recurrence-free systemic- 0.2 0.2

Early Early

0 12 3 45 6 0 12 3 45 6 Time (mo) Time (mo)

Fig. 2. Preoperative MTV2.5 parameter and early systemic metastasis in resected pT3 pancreatic cancer. (A) Early systemic recurrence-free survival according to nodal metastasis. (B) Early systemic recurrence-free survival according to MTV2.5 value. MTV, metabolic tumor volume; pT3, pathologic T3. Lee SH, et al: MTV2.5 in Resectable Pancreatic Cancer 361 tients who received adjuvant treatment after surgery underwent 34.7 months; 95% CI, 29.9 to 45.2 vs mean, 48.4 months; 95% chemotherapy with radiation therapy. There was no significant CI, 39.9 to 56.9; p=0.241) (Fig. 4B). In addition, it was found that difference in the proportion of patients receiving adjuvant treat- among the 28 patients who experienced early (<6 months after ment according to the level of MTV2.5 (Table 3). Postoperative surgery) systemic recurrence, only 17 patients (60.7%) received adjuvant chemotherapy could modify the oncologic outcome of postoperative adjuvant chemotherapy, while most patients (58/65 patients with resected pT3 pancreatic cancer. There was no sig- patients, 89.2%) without early systemic recurrence received post- nificant difference in the oncologic outcomes (log-rank p=0.352 operative adjuvant chemotherapy (p=0.001) (Table 4). in DFS, p=0.142 in overall survival) between patients who re- 8. Genetic background of resected T3 pancreatic cancer ceived adjuvant treatment and those who did not in the MTV 2.5 with MTV ≥4.5 is associated with aggressive biologic low group (MTV <4.5). However, among patients with MTV 2.5 2.5 2.5 behavior; transcriptome-wide analysis discriminating ≥4.5, those who received postoperative adjuvant chemotherapy gene expression had longer DFS times than those who did not receive postopera- tive adjuvant chemotherapy (mean, 5.7 month; 95% CI, 2.2 to In total, 85 genes (upregulated in MTV2.5 high: 14 genes; 9.2 vs mean, 17.4 month; 95% CI, 11.3 to 23.6; p=0.026) (Fig. downregulated: 71) were identified in the transcriptome-wide 4A); furthermore, the DFS of the former group approached that analysis for differentially expressed genes according to metabol- of patients with MTV2.5 <4.5 (mean, 28.3 months; 95% CI, 18.9 ic phenotype (MTV high; MTV2.5 ≥4.5 vs MTV low; MTV2.5 <4.5) to 37.6; p=0.075). Moreover, the overall disease-specific sur- (Fig. 5A and B, Supplementary Tables 1 and 2, Supplementary vival of patients with MTV2.5 ≥4.5 who underwent postoperative Figs 1 and 2). Causative analyses using regulator effectors asso- chemotherapy became similar to those with MTV2.5 <4.5 (mean, ciated with the phenotype of the tumors were performed. Master regulators (NFKB1, ABL1, EDN1, ABCG2, ADIPOQ, MYD88, FOS, and IL1B) were associated with the phenotype of the tu- mor in terms of the proliferation and invasion of tumor cells Table 3. Details of Adjuvant Treatment According to MTV2.5 via several downstream effectors (AKR1B1, NOTCH1, CDKN1A, MTV <4.5 MTV ≥4.5 Adjuvant treatment 2.5 2.5 p-value (n=37) (n=56) CDKN1B, CD44, CCND1, VEGFA, NFkB, ERK1/2, BCL2, TGFB1, Akt, and PI3K). These molecular pathways were significantly Adjuvant treatment 29 (78.4) 45 (80.4) 0.812 Chemoradiation 4 (13.8) 8 (17.8) 0.788 Chemotherapy only 25 (86.2) 48 (82.2) Table 4. Early Systemic Recurrence According to Postoperative Adju- vant Chemotherapy Radiation dose, range, Gy 40.3±15.4 41.2±12.3 0.886 Early systemic recurrence Chemotherapy regimen Adjuvant chemotherapy (<6 mo after surgery) p-value Gemcitabine base 16 (55.2) 24 (53.3) 0.728 No Yes 5-Fluorouracil base 13 (44.8) 21 (46.7) Postoperative adjuvant No 7 11 0.001 Data are presented as number (%). MTV, metabolic tumor volume. chemotherapy Yes 58 17

A B 1.0 1.0

0.8 0.8

survival MTV2.5<4.5 survival 0.6 0.6

0.4 0.4 MTV2.5 >4.5 with Adj-CTx MTV2.5 <4.5

Disease-free MTV2.5 >4.5 with Adj-CTx

0.2 * Disease-specific 0.2

MTV2.5 >4.5 without Adj-CTx MTV2.5 >4.5 without Adj-CTx 0 20 40 60 0 20 40 60 Time (mo) Time (mo)

Fig. 4. The adverse oncologic impact of MTV2.5 ≥4.5 according to postoperative adjuvant chemotherapy in resected pT3 pancreatic cancer. (A) Dis- † ‡ ease-free survival according to MTV2.5 subgroup with or without adjuvant chemotherapy (*p=0.00, p=0.075, p=0.026). (B) Disease-specific overall § Ⅱ ¶ survival according to MTV2.5 subgroup with or without adjuvant chemotherapy ( p<0.001, p=0.241, p<0.001). MTV, metabolic tumor volume; pT3, pathologic T3; Adj-CTx, adjuvant chemotherapy. 362 Gut and Liver, Vol. 13, No. 3, May 2019 A

MTV low 1

min

Relative MTV low 2 Row MTV low 3

x MTV high 1

ma MTV high 2

Row MTV high 3

EIF1B AKR1C2 TMC8 DPT SYNPO2 SNPH MYH1 SYNPO2 SP3 MYH1 BFRS14 PYCR2 MYH1 DES MMP28 PCDHGA1 TMEM50B TNK2 PCDHGA1 CNOT6L METTL8 PCDHGA10 C6ort58 KYNU MAF TSC22D3 HIST1H4L LYZ SYNM SPRR1A FOXF1 ZNF592 GJB2 ELN PLN H19 CRELD1 MYLK PKP1 ETS2 NFKBIA APOL1 MAN2C1 MYH1 PDK4 HPGD MYH1 UTRN MGA ACTG2 RASSF5 LYZ MF CMBL LOC338799 DUSP3 TMED4 ETS2 PCDHGA12 ELL2 A AA905821 PDXDC1 T FXYD6 MLPH TFF2 VAPA GNA1 MUC6 L SNW1 CPT1A HIPK2 PGC PGC TST KCNE3 MUC3A KCNE3 SLC12A2 ALOX5AP L AGR2 BF476613

Y6E

TBP4

ADA2B

W51

AP4

T5

1319

1

1 1 1

1 1

+

+

1

+ B 1.5 1.0

low) 0.5 change vs p-value<0.05) 0.0 fold high 0.5

(MTV 1.0 Relative transformed,

CPT1 TSC22D3 PDK4 H1 UTRN PCDHGA1 SYNM TMC8 TMEM50B PCDHGA1 MUC6 SPRR1A CRELD1 NFKBI ELN GJB2 MAN2C1 ALOX5AP MYL PKP1 CNOT6L PCDHGA1 MUC3 SP3 ELL MLPH SYNPO2 HIST1H4L ETS2 VAPA MF SNPH LOC151 PLN KCNE MMP2 T DUSP KYNU CMBL APOL ZNF592 FOXF1 PYCR2 SYNPO2 TST TNK2 HIPK2 DPT METTL TMED4 AW51 C6orf58 RASSF5 ACTG2 LY6E TFF GNA1 LYZ AA90582 MYH1 FXYD6 MYH1 KCNE MYH1 PGC AGR2 MYH1

LYZ DES PGC BF476613 MYH1

L AKR1C2 PCDHGA1 ETS2 SFRS1 SLC12A LOC338799 HPGD MAF EIF1 PDXDC1 1.5 SNW1

TBP4

ADA2

9

AP4

2

2

K

B (log

A

1319

1

1

1 1 1

1

1

3 3

3

A

8

A

B

4

8

162

2

1

+

+

1

0

2

+

C ABCG2 ADIPOQ EDN1 MYD88 ABL1 FOS

NFKB1 1L1B

NF B PI3K AKR1B1 NOTCH1 CDKN1A CDKN1B CD44 CCND1 VEGFA k ERK1/2 BCL2 TGFB1 Akt (complex) (complex)

Proliferation Invasion of of tumor cells tumor cells

Fig. 5. Hierarchical clustering & significant genes heat map comparing metabolic tumor volume (MTV) high versus low (14 upregulated genes and 71 downregulated genes). (A) Hierarchical clustering of significantly expressed genes (fold change >2 or <0.5 with p<0.05). (B) Relative log- transformed fold changes of MTV high versus low. (C) Causative analysis of regulatory effectors and associated tumor phenotype. Master regula- tors (NFKB1, ABL1, EDN1, ABCG2, ADIPOQ, MYD88, FOS, and IL1B) were associated with the tumor phenotype in terms of the proliferation and invasion of tumor cells via several downstream effectors (AKR1B1, NOTCH1, CDKN1A, CDKN1B, CD44, CCND1, VEGFA, NFkB, ERK1/2, BCL2, TGFB1, Akt, and PI3K) (data shown in Supplementary Tables 4 and 6). correlated with the aggressive tumor phenotype (Fig. 5C, data lism obtained from the GeneCards database (Supplementary Figs shown in Supplementary Tables 4 and 6). 2-6).15,16 The upregulated canonical pathways predicted from the A supervised analysis for cancer-specific genes related to the gene expression pattern of the MTV high tumors were associat- hallmarks of cancer was performed using gene set analysis with ed with aberrant immune response, impaired stromal signaling, gene sets for oncogenes, tumor suppressor genes, proliferation, and alternative metabolic pathways (Fig. 6). MTV high tumors apoptosis, extracellular remodeling, angiogenesis, immune es- were strongly associated with the aggressive tumor phenotype cape, epithelial-mesenchyme transition, metastasis, and metabo- in terms of the proliferation and invasion of tumor cells, with Lee SH, et al: MTV2.5 in Resectable Pancreatic Cancer 363

Canonical pathway for MTV2. 5 high tumor B cell development Hepatic fibrosis/hepatic stellate cell activation Hematopoiesis from pluripotent stem cells Primary immunodeficiency signaling LXR/RXR activation Caveolar-mediated endocytosis signaling GP6 signaling pathway Clathrin-mediated endocytosis signaling Inhibition of matrix metalloproteases IL-7 signaling pathway Tight junction signaling Glucocorticoid receptor signaling p53 signaling Agranulocyte adhesion and diapedesis HIF 1a signaling FXR/RXR activation Atherosclerosis signaling Epithelial adherens junction signaling Macropinocytosis signaling Fig. 6. Significantly upregulated Renal cell carcinoma signaling canonical pathways in MTV2.5 high Alanine biosynthesis II tumor samples (data shown in Sup- Alanine degradation III plementary Table 3). Cysteine biosynthesis/homocysteine degradation MTV, metabolic tumor volume; Glutamine degradation I LXR, liver X receptor; RXR, retinoid X receptor; IL, interleukin; HIF, 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 hypoxia-inducible factor 1; FXR, Log (p-value) farnesoid X receptor.

Diseases or functions for MTV2. 5 high tumor

Metabolism of carbohydrate Invasion of tumor cells Cell viability Invasion of tumor cell lines Binding of DNA Survival of organism Metabolism of nucleic acid component or derivative Metabolism of nucleotide Cell movement of pancreatic cancer cell lines Activation of antigen presenting cells Migration of tumor cells Quantity of carbohydrate Genital tumor Proliferation of tumor cells Cell spreading Invasion of tumor Inflammation of body cavity Contraction of muscle cells Glomerulosclerosis Apoptosis of tumor cell lines Inflammation of organ Cell death of tumor cell lines Apoptosis Inflammation of absolute anatomical region Cell death Necrosis Morbidity or mortality Organismal death 4.0 3.5 3.0 2.5 2.0 1.5 1.0 0.5 0.0 0.5 1.0 1.5 2.0 2.5 Activation z-score

Fig. 7. Diseases or functions associated with gene expression pattern of MTV2.5 high tumor samples (data shown in Supplementary Table 5). MTV, metabolic tumor volume.

364 Gut and Liver, Vol. 13, No. 3, May 2019 M

FO

P F

T TS TFF2 A

K S

XF1 P

P

NFKBI P 4

1 3 K2 2 IA B G JB TN K JB A G F 3 2 N SP

UBB Angiogenesi ECM ALOX5AP A TFF2 G CRELD1 R 2 Immune A TFF2 FO E NFKBI XF1 M s PGC SP3 T Apoptosis PKP1 PKP1 SP3 Metastasis n FOXF1 ALOX5AP TFF2

METTL8 1 Proliferatio DMBT KYNU SCGB3A1 APO GJB2 L1 H19 LY6E Metabolism TFF2 TSG Fig. 8. Circos plot for the landscape ALOX5AP SGK1 TNK2 MLPH H19 of discriminating gene expression Oncogene NFKBIA SP3 related to cancer-specific genes ac- cording to metabolic phenotype in resected pT3 pancreatic cancer. MTV, metabolic tumor volume; pT3, MTV low MTV high pathologic T3; EMT, epithelial mes- enchymal transition; ECM, extracel- lular matrix; TSG, tumor suppressor gene. the predicted correlation with disease and function (Fig. 7). The Therefore, considering the potential role of systemic chemo- landscape of discriminating gene expression levels of cancer- therapy in treating pancreatic cancer, even resectable pancreatic specific genes according to metabolic phenotype in resected pT3 cancer, neoadjuvant chemotherapy might be reasonable in 17 pancreatic cancer visualized by Circos plots revealed that MTV patients with pancreatic cancer with high MTV2.5. Our results high tumors had markedly higher expression levels of cancer- also showed that high metabolic tumors correlated with higher specific genes than MTV low tumors, reflecting the biologically node positive rate, early systemic recurrence within 6 months aggressive behavior of resectable pancreatic cancer (Fig. 8). after curative resection, and favorable oncologic outcomes in response to systemic chemotherapy. It is thought that this popu- DISCUSSION lation can be a potential candidate population for neoadjuvant treatment in resectable pancreatic cancer. Pancreatic cancer is an intractable carcinoma with a dismal In this study, we performed a transcriptome-wide analysis prognosis due to early systemic metastasis. Even after margin- using patients’ tumor tissue to reveal the genetic background of negative resection of primary pancreatic cancer, the majority of the differences in metabolic phenotypes identified on PET scans patients undergo systemic metastasis. Early systemic metastasis preoperatively. The result demonstrated broad enhancement of is a major feature of pancreatic cancer even after margin-neg- cancer-related gene expression, especially genes related to can- ative resection of a primary cancer lesion. Therefore, the identi- cer progression and metastasis in the MTV high group (MTV2.5 fication of a high-risk group showing aggressive tumor biology ≥4.5) compared to the low group (MTV2.5 <4.5). To the best of even among patients with resectable pancreatic cancer is needed our knowledge, this is rare translational study using clinical to improve their oncologic outcomes after curative resection. samples based on metabolic phenotype measured by preopera- Current evidence shows that metabolic phenotypes that can tive images to reveal differences in the molecular landscape ac- be obtained clinically by PET scans before surgical resection can cording to metabolic phenotype. predict the oncologic outcomes of patients with various solid This retrospective observational study was performed in a tumors.18 There is a high likelihood that a primary tumor with single institution with a limited study population. The transcrip- high metabolic activity has at least a minimal amount of re- tome-wide analysis was also conducted in small number based sidual disease that cannot be detected on preoperative imaging on metabolic phenotype. In spite of these limitations, our pilot studies.19,20 translational study showed the prognostic value of metabolic In the treatment of pancreatic cancer, systemic chemotherapy phenotype that can be identified by preoperative PET scan. In is fundamental for improving oncologic outcomes after resec- addition, the potential correlation between gene expression tion.19 In fact, in the present study, 15 of 28 patients (53.6%) related to aggressive tumor behavior and high metabolism pan- showing early recurrence were found to have systemic metas- creatic cancer can be utilized to suggest neoadjuvant treatment tasis before adjuvant treatment because of their poor general and could be beneficial for patients with “apparently” resectable condition and their refusal of treatment (11 out of 15 patients). pancreatic cancer with high FDG uptake. Lee SH, et al: MTV2.5 in Resectable Pancreatic Cancer 365

In conclusion, resectable pancreatic cancer with high MTV2.5 7. Kubota K. From tumor biology to clinical PET: a review of posi- is associated with lymph node metastasis and early systemic tron emission tomography (PET) in oncology. Ann Nucl Med metastasis. Nonetheless, postoperative adjuvant chemotherapy 2001;15:471-486. attenuated this adverse oncologic impact of high MTV2.5 in re- 8. Torizuka T, Tamaki N, Inokuma T, et al. In vivo assessment of sectable pancreatic cancer. The genetic background of resectable glucose metabolism in hepatocellular carcinoma with FDG-PET. J pancreatic cancer with high MTV2.5 might be associated with Nucl Med 1995;36:1811-1817. aggressive biologic behavior. Taken together, neoadjuvant treat- 9. Chung HH, Jo H, Kang WJ, et al. Clinical impact of integrated ment can be a potential alternative even in resectable pancreatic PET/CT on the management of suspected cervical cancer recur- cancer with high MTV2.5. A well-designed prospective clinical rence. Gynecol Oncol 2007;104:529-534. trial with confirmative mechanism study is needed to confirm 10. Shinoto M, Yamada S, Yoshikawa K, et al. Usefulness of 18F- the choice of therapeutic strategy according to metabolic phe- fluorodeoxyglucose positron emission tomography as predictor notype in pancreatic cancer. of distant metastasis in preoperative carbon-ion radiotherapy for pancreatic cancer. Anticancer Res 2013;33:5579-5584. CONFLICTS OF INTEREST 11. Chong JU, Hwang HK, Lee JH, Yun M, Kang CM, Lee WJ. Clini- cally determined type of 18F-fluoro-2-deoxyglucose uptake as No potential conflict of interest relevant to this article was an alternative prognostic marker in resectable pancreatic cancer. reported. PLoS One 2017;12:e0172606. 12. Kang CM, Lee SH, Hwang HK, Yun M, Lee WJ. Preoperative ACKNOWLEDGEMENTS volume-based PET parameter, MTV2.5, as a potential surrogate marker for tumor biology and recurrence in resected pancreatic This study was supported by Korea Institute of Planning and cancer. Medicine (Baltimore) 2016;95:e2595. Evaluation for Technology in Food, Agriculture, Forestry and 13. Kang CM, Hwang HK, Park J, et al. Maximum standard uptake Fisheries (IPET, 31-316055-3). value as a clinical biomarker for detecting loss of SMAD4 expres- Author contributions: Study design: S.H.L., C.M.K. Data col- sion and early systemic tumor recurrence in resected left-sided lection: H.K.H., W.J.L., M.Y. Data analysis: S.H.L., C.M.K. Article pancreatic cancer. Medicine (Baltimore) 2016;95:e3452. writing: S.H.L., C.M.K. Critical revision: C.M.K., M.Y. Obtaining 14. Allen PJ, Kuk D, Castillo CF, et al. Multi-institutional validation fund: C.M.K. study of the American Joint Commission on Cancer (8th Edition) changes for T and N staging in patients with pancreatic adenocar- REFERENCES cinoma. Ann Surg 2017;265:185-191. 15. Rebhan M, Chalifa-Caspi V, Prilusky J, Lancet D. GeneCards: in- 1. Hsu PP, Sabatini DM. Cancer cell metabolism: Warburg and be- tegrating information about genes, and diseases. Trends yond. Cell 2008;134:703-707. Genet 1997;13:163. 2. Bares R, Klever P, Hauptmann S, et al. F-18 fluorodeoxyglucose 16. Stelzer G, Rosen N, Plaschkes I, et al. The GeneCards Suite: from PET in vivo evaluation of pancreatic glucose metabolism for de- gene data mining to disease genome sequence analyses. Curr Pro- tection of pancreatic cancer. Radiology 1994;192:79-86. toc Bioinformatics 2016;54:1.30.1-1.30.33. 3. Lindholm P, Minn H, Leskinen-Kallio S, Bergman J, Ruotsalainen 17. Krzywinski M, Schein J, Birol I, et al. Circos: an information U, Joensuu H. Influence of the blood glucose concentration on aesthetic for comparative genomics. Genome Res 2009;19:1639- FDG uptake in cancer: a PET study. J Nucl Med 1993;34:1-6. 1645. 4. Czernin J, Phelps ME. Positron emission tomography scanning: 18. Kim BK, Kang WJ, Kim JK, et al. 18F-fluorodeoxyglucose uptake current and future applications. Annu Rev Med 2002;53:89-112. on positron emission tomography as a prognostic predictor in lo- 5. Kinkel K, Lu Y, Both M, Warren RS, Thoeni RF. Detection of he- cally advanced hepatocellular carcinoma. Cancer 2011;117:4779- patic metastases from cancers of the gastrointestinal tract by using 4787. noninvasive imaging methods (US, CT, MR imaging, PET): a meta- 19. Müller V, Alix-Panabières C, Pantel K. Insights into minimal analysis. Radiology 2002;224:748-756. residual disease in cancer patients: implications for anti-cancer 6. Ahn SJ, Park MS, Lee JD, Kang WJ. Correlation between 18F- therapies. Eur J Cancer 2010;46:1189-1197. fluorodeoxyglucose positron emission tomography and pathologic 20. Kostakoglu L, Goldsmith SJ. 18F-FDG PET evaluation of the re- differentiation in pancreatic cancer. Ann Nucl Med 2014;28:430- sponse to therapy for lymphoma and for breast, lung, and colorec- 435. tal carcinoma. J Nucl Med 2003;44:224-239.