Clinical and Tumor Characteristics of Patients with High Serum Levels of Growth Differentiation Factor 15 in Advanced Pancreatic Cancer

Simple Summary Growth differentiation factor 15 (GDF-15) is a stress responsive cytokine that mediates food intake, energy consumption, and body weight. We aimed to evaluate whether circulating GDF-15 level could be associated with cachexia symptoms, which include loss of skeletal muscle mass, systemic inflammatory reaction, poor performance status, anorexia, shortened survival time and biological tumor activity in advanced pancreatic cancer (APC). The cut-off for serum GDF-15 was 3356.6 pg/mL, as the mean plus two standard deviations in patients with benign pancreatic disease. APC patients with high serum GDF-15 showed worsened performance, anorexia and elevations of inflammatory and tumor burden, signatures of cachexia, and activation of Akt and JNK in tumor GDF-15-producing pathways. This study identified tumor-driven GDF-15 as a potential cause of cachexia symptoms in APC. Abstract We aimed to evaluate the association of circulating growth differentiation factor 15 (GDF-15) with cachexia symptoms and the biological activity of advanced pancreatic cancer (APC). Treatment-naïve patients with liver metastasis of APC or with benign pancreatic disease were retrospectively analyzed. Clinical data, blood samples, and biopsy specimens of liver metastasis were collected prior to anti-cancer treatment. Serum GDF-15 levels and multiple protein expressions in lysates extracted from liver metastasis were measured by enzyme-linked immuno-sorbent assay and reverse-phase protein array, respectively. The cut-off for serum GDF-15 was determined as 3356.6 pg/mL, the mean plus two standard deviations for benign pancreatic disease. The high-GDF-15 group was characterized as showing low Karnofsky performance status (KPS) (p = 0.037), poor Eastern Cooperative Oncology Group performance status (ECOG-PS) (p = 0.049), severe appetite loss (p = 0.011), and high serum levels of carbohydrate antigen 19-9 (p = 0.019) and C-reactive protein (p = 0.009). Tumors of the high-GDF-15 group expressed high levels of phosphorylated (p)JNK (p = 0.007) and pAkt (p = 0.040). APC patients with high serum GDF-15 showed signatures of cachexia and activation of the signaling pathways involving Akt and JNK in the tumor. This study indicated circulating GDF-15 could be associated with cachectic symptoms in APC.

Expression of GDF-15 is maintained at low levels in most tissues [21] and is increased by p53 activation due to various cellular stressors, such as inflammation, oxidative stress, hypoxia, and oncogene activation [22]. Serum GDF-15 levels in the patients with chronic pancreatitis were higher, 2248 pg/mL in mean [1] compared to healthy subjects (416.8 pg/mL [23], 546 [24], 639 [25] in mean). Pancreatic disease compressing the pancreatic duct is considered to increase the GDF-15 level, and thus can result from benign pancreatic diseases including chronic pancreatitis, intraductal papillary mucinous tumor, and serous cystic tumor. The cut-off level to characterize a high serum level of GDF-15 in APC thus should discriminate cancer-related upregulation of GDF-15 from the effects of benign pancreatic disease. We therefore planned to determine the upper limit of serum GDF-15 levels in benign pancreatic diseases, in order to set an appropriate cut-off to discriminate low and high serum GDF-15 groups for APC.
Solid stress increased GDF-15 expression by transcriptionally regulating GDF-15 expression through activation of the Akt pathway in pancreatic cancer cells [26]. Stress responses involving the Akt pathway can become dominant in cases with pancreatic cancer complicated by high levels of circulating GDF-15, providing biological validation for a cut-off for serum GDF-15 in pancreatic cancer. The clinical and tumoral characteristics of APC patients with high serum GDF-15 levels are expected to provide useful information for elucidating the causes of clinical symptoms triggered by stress responses in the tumor microenvironment.
We undertook the present study to investigate the following in APC patients: (A) an appropriate cut-off value for serum GDF-15 levels; (B) the relationship between serum GDF-15 levels and clinical data; and (C) tumor characteristics associated with high serum levels of GDF-15.

Ethics
This study was approved by the ethics review committee of National Cancer Center Hospital East (approval no. 2020-373).

Patients
Treatment-naïve patients with liver metastasis from pathologically proven APC or with benign pancreatic disease were eligible to participate in this study. Subjects were Cancers 2021, 13, 4842 3 of 13 registered between 5 August 2011 and 7 January 2015 at the National Cancer Center Hospital East. Benign pancreatic diseases comprised chronic pancreatitis, intraductal papillary mucinous tumor, non-invasive intraductal papillary mucinous carcinoma, and serous cystic tumor. Written informed consent was obtained from all subjects before enrollment.

Definition of Cancer Cachexia and Sarcopenia
In this study, we defined cancer cachexia as either weight loss >5% or weight loss >2% with a body mass index (BMI) < 20 kg/m 2 [16]. Base cachexia was defined as weight loss within 6 months before starting chemotherapy. Sarcopenia was diagnosed using the Japan Society of Hepatology criteria [27] as low handgrip strength (<26 kg for males; <18 kg for females) and low muscle mass (<42 cm 2 /m 2 for males; <38 cm 2 /m 2 for females). Muscle mass was evaluated as the skeletal muscle index (SMI) at the level of the third lumbar vertebra (L3) on computed tomography.

Data Collection
Body weight, Karnofsky performance status (KPS), Eastern Cooperative Oncology Group performance status (ECOG-PS), laboratory tests, severity of symptoms (the Japanese version of the MD Anderson Symptom Inventory; MDASI-J) [28] and a blood sample were collected before anti-cancer treatment. MDASI-J is a 13-item symptom scale (pain, fatigue, nausea, disturbed sleep, distress, shortness of breath, memory, lack of appetite, drowsiness, dry mouth, sadness, vomiting, and numbness or tingling) rated numerically from 0 ("not present") to 10 ("as bad as you can imagine") based on symptoms present over the preceding 24 h. Severe appetite loss was defined as MDASI-J ≥ 2, as per our previous study [29]. The cut-off for serum levels of C-reactive protein (CRP) was determined as 2.0 mg/dL in accordance with our previous study [20]. The cut-off for carbohydrate antigen (CA)19-9 was defined as above median serum CA19-9 level, which was 4150 U/mL in our study.

Enzyme-Linked Immuno-Sorbent Assay (ELISA)
Blood samples were collected in a serum-separating tube on the morning after overnight fasting and completely clotted for 30-60 min at room temperature. After centrifugation at 1500 g, 25 • C for 15 min, serum samples were separated and frozen at −80 • C until analysis. Serum GDF-15 levels were quantitatively determined in duplicate using an ELISA kit in accordance with the protocol from the manufacturer (R&D Systems, Minneapolis, MN, USA).

Protein Expression Analysis
Tissue cores were obtained from liver metastases of APC with endoscopic ultrasonographyguided fine-needle aspiration biopsy (FNAB) [30]. Immediately after the procedure, a tissue core specimen was homogenized with tissue-Lyser II (QIAGEN, Hilden, Germany) in the lysis buffer of a NucleoSpin TriPrep kit (Macherey-Nagel, GmbH & Co., KG, Düren, Germany), and was filtered with the NucleoSpin TriPrep kit. After protein purification with a Pierce™ SDS-PAGE Sample Prep Kit (Thermo Fisher Scientific, Waltham, MA, USA), the lysate was snap-frozen and stored at −80 • C until needed. Reverse-phase protein array (RPPA) analysis of lysates was performed at the functional proteomic RPPA core facility of the MD Anderson Cancer Center. Briefly, tissue lysates were serially diluted two-fold for five dilutions (Undiluted, 1:2, 1:4, 1:8, and 1:16) and arrayed onto nitrocellulose-coated slides [31,32]. RPPA slides immunostained with 435 primary antibodies were scanned with a TissueScope scanner (HURON Digital Pathology, St. Jacobs, ON, Canada), and the signal intensity of each spot on the arrays was quantified using Array-pro Analyzer 6.3 software (Media Cybernetics, Silver Spring, MD, USA). The relative protein levels of analytes in each sample were then determined using the SuperCurve 1.5.0 R package via SuperCurveGUI 2.1.1. [33]. RPPA analysis was performed in the patients whose protein samples could be extracted, purified, and detected as a reproducible band of β-actin in western blotting.

Statistical Analyses
Continuous data were compared using the Mann-Whitney U test. Associations between serum GDF-15 levels and each categorical variable were examined using a chisquare test or Fisher's exact test. Fisher's exact test and chi-square test were used for analysis of 2 × 2 contingency tables and more than 2 × 2 consistency tables, respectively. Overall survival (OS) was calculated from the starting date of first-line chemotherapy or BSC. Progression-free survival (PFS) was calculated only in patients with chemotherapy from the starting date of first-line chemotherapy. Survival curves were drawn using the Kaplan-Meier method. OS and PFS were compared using the log-rank test between lowand high-GDF-15 groups of APC patients with liver metastasis. For all statistical tests, values of p < 0.05 were considered statistically significant. Median survival times were calculated with 95%CIs, as determined using the Brookmeyer and Crowley method. Data analyses were performed using JMP ® version 11 software (SAS Institute, Cary, NC, USA).
Array-pro Analyzer 6.3 software (Media Cybernetics, Silver Spring, MD, USA). The relative protein levels of analytes in each sample were then determined using the SuperCurve 1.5.0 R package via SuperCurveGUI 2.1.1. [33]. RPPA analysis was performed in the patients whose protein samples could be extracted, purified, and detected as a reproducible band of β-actin in western blotting.

Statistical Analyses
Continuous data were compared using the Mann-Whitney U test. Associations between serum GDF-15 levels and each categorical variable were examined using a chisquare test or Fisher's exact test. Fisher's exact test and chi-square test were used for analysis of 2 × 2 contingency tables and more than 2 × 2 consistency tables, respectively. Overall survival (OS) was calculated from the starting date of first-line chemotherapy or BSC. Progression-free survival (PFS) was calculated only in patients with chemotherapy from the starting date of first-line chemotherapy. Survival curves were drawn using the Kaplan-Meier method. OS and PFS were compared using the log-rank test between lowand high-GDF-15 groups of APC patients with liver metastasis. For all statistical tests, values of p < 0.05 were considered statistically significant. Median survival times were calculated with 95%CIs, as determined using the Brookmeyer and Crowley method. Data analyses were performed using JMP ® version 11 software (SAS Institute, Cary, NC, USA).
The 40 APC patients who underwent tumor analysis by RPPA were separated into two groups according to median expression levels (pJNK: −0.0256 and pAkt: 0.0036).
The 40 APC patients who underwent tumor analysis by RPPA were separated into two groups according to median expression levels (pJNK: −0.0256 and pAkt: 0.0036).

Discussion
Mean circulating GDF-15 levels have been suggested to be higher in patients with pancreatic ductal adenocarcinoma, as 1731 pg/mL [23], 2428 [24], 7694.58 [1] and 2990.2 (this study), compared to patients with colorectal adenocarcinoma (CRC) (1371.0 pg/mL [23] and 2030 pg/mL [25]), non-small-cell lung carcinoma (mean, 1258.0 pg/mL [23] and median, 828.8-2000 pg/mL [34,35]), and gastric adenocarcinoma (1154.0 pg/mL [23]). An atlas of the human transcriptome reported that GDF-15 mRNA expression was relatively low in pancreatic cancer compared to prostate, urothelial, renal, melanoma, and colorectal cancers [36]. The discrepancy between tumor mRNA expression and circulating level of GDF-15 may involve the ratio of patients with metastasis in each study population. A large proportion of CRC patients in a previous study [23] showed early-stage disease, although CRC patients with distant metastasis showed higher levels of serum GDF-15 (mean, 3531 pg/mL) [25] and all APC patients in the present study showed liver metastasis. GDF-15 levels in various cancers should thus be assessed in concert with the ratio of metastasis in the study population. The diurnal pattern of serum GDF-15 levels needs to be clarified for evaluation of serum GDF-15. Circadian oscillation led to variations in serum GDF-15 levels near the midline estimating statistic of rhythm (MESOR) of about ± 10%, with a MESOR of 381 ± 15 pg/mL in healthy men [37]. In the present study, blood samples were obtained the morning after overnight fasting, and all APC patients showed

Discussion
Mean circulating GDF-15 levels have been suggested to be higher in patients with pancreatic ductal adenocarcinoma, as 1731 pg/mL [23], 2428 [24], 7694.58 [1] and 2990.2 (this study), compared to patients with colorectal adenocarcinoma (CRC) (1371.0 pg/mL [23] and 2030 pg/mL [25]), non-small-cell lung carcinoma (mean, 1258.0 pg/mL [23] and median, 828.8-2000 pg/mL [34,35]), and gastric adenocarcinoma (1154.0 pg/mL [23]). An atlas of the human transcriptome reported that GDF-15 mRNA expression was relatively low in pancreatic cancer compared to prostate, urothelial, renal, melanoma, and colorectal cancers [36]. The discrepancy between tumor mRNA expression and circulating level of GDF-15 may involve the ratio of patients with metastasis in each study population. A large proportion of CRC patients in a previous study [23] showed early-stage disease, although CRC patients with distant metastasis showed higher levels of serum GDF-15 (mean, 3531 pg/mL) [25] and all APC patients in the present study showed liver metastasis. GDF-15 levels in various cancers should thus be assessed in concert with the ratio of metastasis in the study population. The diurnal pattern of serum GDF-15 levels needs to be clarified for evaluation of serum GDF-15. Circadian oscillation led to variations in serum GDF-15 levels near the midline estimating statistic of rhythm (MESOR) of about ± 10%, with a MESOR of 381 ± 15 pg/mL in healthy men [37]. In the present study, blood samples were obtained the morning after overnight fasting, and all APC patients showed liver metastasis. The effects of circadian oscillation and metastatic rate on circulating GDF-15 levels were considered to have been minimized in our study.
Our results showed that patients in the high-GDF-15 group experienced severe appetite loss compared with those in the low-GDF-15 group. In a previous study, median serum GDF-15 levels were higher in cancer patients with anorexia (1224.1 pg/mL) than in patients without anorexia (812.4 pg/mL) [34], where the cut-off value for anorexia was set at the lowest Functional Assessment of Anorexia/Cachexia Treatment (FAACT) score tertile, similar to the consensus definition of anorexia (FAACT ≤ 24) [38]. Our results showed that median MDASI scores in the high-and low-GDF-15 groups were 4 and 1 on a 0-10 scale, respectively. National Comprehensive Cancer Network (NCCN) guidelines for palliative care have proposed defining non-pain symptoms scoring >4 on the 0-10 scale as severe [39]. In validation studies of MDASI, severe, moderate, and mild symptoms are defined as scores ≥7, 5 or 6, and 1-4, respectively [28,40]. The degrees of anorexia in highand low-GDF-15 groups corresponded to the upper and lower limits of mild anorexia, respectively. A prior report proposed the minimal clinical difference in numerical rating scale for anorexia as 1.5 points [41], and the median MDASI-J score in our study was 3 points higher in the high-GDF-15 group than in the low-GDF-15 group. On this basis, we concluded that patients in the high-GDF-15 group showed a clinically meaningful increase in the severity of anorexia compared with those in the low-GDF-15 group.
Our RPPA analysis revealed that expressions of pJNK and pAKT were significantly higher in the high-GDF-15 group than in the low-GDF-15 group. Tumor-associated stresses such as solid stress have been reported to induce phosphorylation of JNK and AKT [42][43][44]. Growth-induced solid stress in a tumor reflects an excessive tumor burden that compresses surrounding tissues [45,46]. APC patients with high serum GDF-15 levels also had high serum CA19-9 levels, suggesting high tumor burden [47]. Intriguingly, Akt and JNK have been shown to activate transcription of the GDF-15 gene [26,48] and to promote secretion of the GDF-15 protein [5,49]. We therefore speculate that increased solid stress due to a large tumor burden augments the expressions of pJNK and pAKT and also increases levels of circulating GDF-15. In addition, we observed that p-forkhead box-containing protein (FOX)O3a (serine 318_serine 321), p-tuberin (threonine_1462), and pElk-1 (serine 383) were highly expressed in the high-GDF-15 group compared to the low-GDF-15 group (Table S1). Activation of FOXO3a and tuberin was noted to promote tumor growth [50,51], which was inhibited by Akt-mediated phosphorylation [52]. On the other hand, Elk1 was activated by JNK-mediated phosphorylation, leading to cell proliferation and oncogenesis [53]. These observations suggest that signaling pathways involving Akt and JNK play important roles in tumor growth in APC patients with high serum GDF-15 levels.
Evans' diagnostic criteria for cachexia include body weight loss, sarcopenia, systemic inflammation, appetite loss and fatigue, and are related to poor PS [54]. In our study, patients with high serum levels of GDF-15 showed severe appetite loss, systemic inflammation and poor PS, but not sarcopenia or body weight loss. We consider that an increase in tumor burden contributed to GDF-15-related anorexia via solid stress and was related to systemic inflammation and worsened PS prior to the onset of body weight loss and sarcopenia. A phase 1, first-in-humans, two-part, open-label clinical trial was recently started for CTL-002, a GDF-15 neutralizing antibody, recruiting subjects with advancedstage, relapsed, or refractory solid tumors (NCT04725474). The target patients will be evaluated for not only antitumor effect and safety, but also cachexia-related outcomes, since inhibition of GDF-15 biological activity has been shown to reverse body weight loss and restore muscle and fat tissue mass in several cachectic animal models [55]. This clinical trial of CTL-002 is expected to clarify the relationship between GDF-15 and cachexia.
The present study had a limitation that requires consideration when interpreting the results. This study had no validation cohort for the cut-off value of serum GDF-15 levels. Our cut-off value was able to discriminate the biological activity of JNL and AKT signaling in tumors as pathways producing GDF-15. In terms of biological activity, our cut-off value for serum GDF-15 was confirmed. Further studies to validate our cut-off for serum GDF-15 are warranted.

Conclusions
In conclusion, APC patients with high serum GDF-15 showed worsened performance status, anorexia and elevations of inflammation and tumor burden as signatures of cachexia, as well as activation of Akt and JNK in tumor GDF-15-producing pathways. This study indicated that circulating GDF-15 could be associated with cachectic symptoms in APC.