Visceral adipose tissue remodeling in pancreatic ductal adenocarcinoma cachexia: the role of activin A signaling

Pancreatic ductal adenocarcinoma (PDAC) patients display distinct phenotypes of cachexia development, with either adipose tissue loss preceding skeletal muscle wasting or loss of only adipose tissue. Activin A levels were measured in serum and analyzed in tumor specimens of both a cohort of Stage IV PDAC patients and the genetically engineered KPC mouse model. Our data revealed that serum activin A levels were significantly elevated in Stage IV PDAC patients in comparison to age-matched non-cancer patients. Little is known about the role of activin A in adipose tissue wasting in the setting of PDAC cancer cachexia. We established a correlation between elevated activin A and remodeling of visceral adipose tissue. Atrophy and fibrosis of visceral adipose tissue was examined in omental adipose tissue of Stage IV PDAC patients and gonadal adipose tissue of an orthotopic mouse model of PDAC. Remarkably, white visceral adipose tissue from both PDAC patients and mice exhibited decreased adipocyte diameter and increased fibrotic deposition. Strikingly, expression of thermogenic marker UCP1 in visceral adipose tissues of PDAC patients and mice remained unchanged. Thus, we propose that activin A signaling could be relevant to the acceleration of visceral adipose tissue wasting in PDAC-associated cachexia.


Pancreatic ductal adenocarcinoma (PDAC) patients display distinct phenotypes of cachexia development, with either adipose tissue loss preceding skeletal muscle wasting or loss of only adipose tissue. Activin A levels were measured in serum and analyzed in tumor specimens of both a cohort of Stage IV PDAC patients and the genetically engineered KPC mouse model. Our data revealed that serum activin A levels were significantly elevated in Stage IV PDAC patients in comparison to agematched non-cancer patients. Little is known about the role of activin A in adipose tissue wasting in the setting of PDAC cancer cachexia. We established a correlation between elevated activin A and
Cancer cachexia is a devastating, multifactorial, and often irreversible syndrome that accounts for up to 20% of cancer deaths 1,2 . Unlike starvation where adipose tissue mass is lost while skeletal muscle mass is preserved, cachexia patients lose weight from both compartments 3 . The current international consensus for diagnostic criterion for cancer cachexia is weight loss > 5% over 6 months; or BMI < 20 and any degree of weight loss > 2%; or appendicular skeletal muscle index consistent with sarcopenia and any degree of weight loss > 2% 4 . Pancreatic ductal adenocarcinoma (PDAC) is currently the third leading cause of cancer death in the United States and is projected to become the second leading cause by 2030 5 . Due to lack of early clinical signs, more than 85% of patients present with advanced metastatic disease by time of diagnosis 6 . Prognosis for patients with metastatic disease remains dismal, with a 5 year survival rate of only 2% and median survival time of 3-6 months 7 . A major contributor to the low survival rate is cachexia, a frequent and prominent feature of pancreatic cancer. Notably, 80% of pancreatic cancer patients exhibit symptoms of cachexia at later stages of disease development 8 . Median weight loss close to time of death has been reported to correspond to 25% of pre-illness weight 9 .
Adipose tissue serves as a metabolically dynamic endocrine and energy-storage organ of the body and is primarily composed of adipocytes. In the current research paradigm, most studies and clinical trials have focused on skeletal muscle atrophy rather than adipose tissue loss in cancer cachexia. However, emerging reports indicate Ex-vivo culture of adipose tissue. Gonadal and subcutaneous adipose tissue were obtained from four C57BL/6 male mice aged between 9 and 11 weeks and used for the ex-vivo culture. Briefly, two gonadal and subcutaneous fat pads from one mouse were divided into four pieces by cutting vertically, and each piece was incubated in either RPMI-1640 (control), CM from KPC8069 cells or CM from KPC8069 cells supplemented with the activin-binding protein follistatin 288 (FST288, 100 ng/mL) for 48 h. The tissues were then fixed in Modified Davidson's Fixative (64133-50, Electron Microscopy Services, Hatfield, PA) and sectioned for Picro-Sirius Red Staining.

Real-time reverse-transcription PCR (RT-PCR). Total RNA was collected using the RNeasy Plus Mini
Kit (74134, Qiagen Inc., Germantown, MD) or the RNeasy Lipid Tissue Mini Kit (74804, Qiagen Inc.). One microgram of total RNA from each of the samples were then converted to cDNA using the High-Capacity RNAto-cDNA™ Kit (4387406, Applied Biosystems, Foster City, CA). TaqMan® probe-based (4304437, Applied Biosystems) or SYBR® Green based (1725121, Bio-Rad Laboratories, Hercules, CA) real-time PCR was performed using primer sequences listed in Supplemental Table S1. The experiment was performed in triplicates using the CFX96™ Real-Time PCR Detection System (185-5096, Bio-Rad Laboratories).

Immunocytochemistry (ICC), immunohistochemistry (IHC), and immunofluorescence (IF)
assays. For ICC, 1 × 10 4 cells from each of the human pancreatic cancer and normal cell lines were seeded on cover glasses overnight and fixed with 2% formaldehyde. Antigen retrieval was carried out in 10 mM sodium citrate buffer. Samples were then blocked with normal goat serum blocking solution (S-1000-20, Vector Laboratories, Burlingame, CA) and treated with primary antibodies (Supplemental Table S2). Biotinylated or Alexa Fluor™-conjugated secondary antibodies were applied according to primary antibody host species (Supplemental Table S2). Sections for IHC were treated with horseradish peroxidase streptavidin (SA-5004-1, Vector Laboratories) and a DAB substrate kit (34065, ThermoScientific, Waltham, MA) and counter-stained with hematoxylin.
Adipocyte size measurement. Adipocyte size was measured by analyzing images of hematoxylin-eosin stained adipose tissue sections using Adiposoft 1.14 in Manual Mode as previously described 32 .
Picro-sirius red staining. To visualize collagen I and III fibers, adipose tissue sections were stained using the Picro-Sirius Red Stain Kit (MER PSR1, Mercedes Scientific, Lakewood Ranch, FL).
Light and fluorescence microscopy. All images were taken with the EVOS M7000 Imaging System (AMF700, Invitrogen, Carlsbad, CA).
Western blot analysis. Samples were homogenized in RIPA buffer (89900, Thermo Scientific) combined with cOmplete™ Protease Inhibitor Cocktail (CO-RO, Roche Diagnostics USA, Indianapolis, IN) and PhosS-TOP™ (PHOSS-RO, Roche Diagnostics USA). Equal amounts of protein were loaded onto a 4-20% polyacrylamide gel (4561094, Bio-Rad Laboratories) and transferred to a nitrocellulose or PVDF membrane. The primary and secondary antibodies used are described in Supplemental Table S2.
Statistical analysis. Data were presented by mean ± S.E.M. Serum activin A and IL-6 levels were compared between groups using a nonparametric Wilcoxon rank-sum test due to the departure from the normality. Pearson correlation was used to determine the correlation between activin A and IL-6 levels and activin A levels with metastasis sites and body mass index for patients with PDAC. All figures with statistical analyses were generated using PRISM version 9.2.0 (GraphPad Software Inc., San Diego, CA, www. graph pad. com). Values of p < 0.05 were considered statistically significant. *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001; n.s., not significant.

Results
Visceral adipose tissue loss can be correlated with increasing systemic activin A levels contributed by pancreatic ductal adenocarcinoma (PDAC) tumors. We first analyzed the correlation between INHBA (inhibin subunit βA) expression and the prevalence of cancer cachexia in 18 types of cancer using RNA-Seq data from The Cancer Genome Atlas (TCGA). Notably, the expression of INHBA was greater in PDAC tumor tissue in comparison to that in matched normal tissue (Fig. 1A). We next examined the expression of activin A (βA) in PDAC tumor specimens and non-cancerous pancreatic tissues. We also investigated the expression of keratin 17 (CK17), a marker of the most lethal molecular subtype of pancreatic cancer 33 , to evaluate potential correlation with activin A. Both activin A (βA) and CK17 appear to be highly expressed in PDAC tumor tissues in comparison to non-cancerous pancreatic tissues. A large portion of βA-positive cells were found to overlap with CK17 signal (yellow, Supplemental Fig. S1A, B). In contrast to undetectable levels of activin A (βA) in non-cancerous pancreatic tissues, activin A is highly expressed in PDAC tumor tissues ( Fig. 1  To determine whether activin A detected in pancreatic tumor tissue could contrinute to changes in levels of activin A in the circulation, we measured activin A from sera from deceased donors with Stage IV PDAC (n = 34) and from healthy donors (n = 25). Serum activin A concentrations of PDAC donors had an average of 5.85 ng/ mL, while serum activin A concentrations of healthy donors had an average of 0.91 ng/ml (Fig. 1D). The mean serum IL-6 concentration of the same population of PDAC donors was 270.90 pg/mL, and the mean serum IL-6 concentration of healthy donors was 34.93 pg/ml (Supplemental Fig. S1E). No significant correlation could be observed between serum activin A and serum IL-6 concentrations in our cohort of PDAC patients (Supplemental Fig. S1F). No statistically significant gender differences in serum activin A and IL-6 levels could be observed within our PDAC patient population (Supplemental Fig. S1D, G and H). Attempts to correlate serum activin A and body mass index in PDAC patients yielded a Pearson correlation coefficient of − 0.29 with p = 0.10, while serum activin A had no correlation with the number of metastasis sites (Supplemental Fig. S1I, J). Interestingly, activin A expression could also be detected in pancreatic intraepithelial neoplasia (PanIN) lesions that are considered precursors to PDAC from KPC mice 34 (Fig. 1E). Accordingly, activin A expression was detected in the   . S2C). Visceral adipose tissue collected from the omenta of PDAC patients displayed a significant histological changes in correspondence with elevated serum activin A including possible fibrous connective tissue ( Fig. 2A). We first observed a decrease in white adipocyte diameter inversely correlating with serum activin A levels (Fig. 2B). We also noted a frequency shift towards smaller adipocytes (Fig. 2C) in patients with higher than average serum activin A levels. We also sought to assess fat mass in PDAC patients by quantifying tissue area in transverse CT-scan images from the third lumbar vertebrae. CT images were matched with serum activin A levels of respective donors, with subcutaneous (blue) and visceral (yellow) fat (Supplemental Fig. S3). While there appears to be decreasing body fat with increasing activin A levels, the activin A levels used were measured after death and the CT images were captured at time of diagnosis.

Elevated activin A levels observed in patients can be recapitulated in tumor-derived cell lines and an animal model of PDAC.
BxPC3, PANC-1 and MIA PaCa-2 cells derived from PDAC patients revealed comparable upregulation of activin A expression in contrast to low basal expression in normal pancreatic acinar-to-ductal metaplasia hTERT-HPNE cells (Fig. 3A). Elevated mature activin A could be detected in BxPC3 cell lysates, while PANC-1 and MIA PaCa-2 cell lysates revealed elevated levels of pro-activin A (Fig. 3B). BxPC3 cells possessed a three-fold change in INHBA mRNA in comparison to PANC-1 cells (Fig. 3C).
We also observed high expression of activin A in KPC tumor specimens in comparison to healthy control (Fig. 4A) and a significant elevation in measured circulating activin A levels in the serum of KPC mice (Fig. 4B). In contrast to virtually no pro-activin A and mature activin A detected in normal mouse pancreas tissue, the expression levels of pro-activin A and mature activin A were dramatically elevated in KPC8069 cell lysates (Fig. 4C). KPC8069 cells demonstrated 1000-fold more Inhba mRNA than healthy mouse pancreatic tissue (Fig. 4D). To determine whether PDAC cells directly secrete activin A into growth media, we measured activin A levels in conditioned media (CM) from KPC8060 and KPC8069 cells. CM from both KPC8060 and KPC8069 cells showed higher activin A secretion (Fig. 4E). CM from KPC 8069 cells collected at 60% and 100% confluency reflected an increase in activin A secretion dependent on cell density. Interestingly, activin A secretion remained just as high after replacement with fresh culture medium at 100% confluency (Fig. 4F).
Visceral adipose tissue loss can be correlated with increasing systemic activin A levels contributed by pancreatic ductal adenocarcinoma (PDAC) tumors in an orthotopic mouse model. We evaluated the expression of activin A in an orthotopic mouse model implanted with S2-013 cells. Tumor tissue sections from S2-013 mPDAC revealed greatly elevated activin A expression in tumors from tumor-bearing mice in comparison to sham (Fig. 5A). Serum levels of activin A also reflected the growth of S2-013-derived www.nature.com/scientificreports/ tumors (Fig. 5B). However, in accordance with human data, there was no significant correlation between circulating activin A and IL-6 levels (Supplemental Fig. S2A). In addition to reduced body weight, the mass of gonadal fat pads from S2-013 mPDAC (n = 7) was greatly reduced (Fig. 5C), with smaller white adipocytes (Fig. 5D) with possible fibrous connective tissue and a shift toward small-diameter adipocyte size in comparison to sham (Fig. 5E, F). We next investigated the metabolic changes in visceral adipose tissue in response to pancreatic cancer development. Expression level of adiponectin, a marker for insulin sensitivity, was significantly reduced, suggesting insulin-resistance. Paradoxically, there was no significant difference in leptin transcription levels, despite markedly reduced adipocyte size, signifying adipose tissue dysfunction in the cachectic state in S2-013 mPDAC (Fig. 6A). Next, we tested adipocyte-specific gene expression levels. Analyses of S2-013 mPDAC revealed a significant reduction in the expression of Pparg, a major transcription factor for adipocyte differentiation, and fatty acid synthetase (Fas) (Fig. 6B). To address the role of apoptotic programmed cell death in cachectic fat loss, we labeled DNA fragmentation on sections of adipose tissue from both sham and S2-013 mPDAC. The percentage of TUNEL-positive cells in adipose tissue from S2-013 mPDAC did not increase in comparison to that of sham (Fig. 6C, D). These results suggest that loss of adipose tissue in S2-013 mPDAC is not driven by apoptotic pathways.

Activin A secreted from PDAC tumor cells contributes to loss of visceral adipose tissue independent of WAT browning and instead contributes to fibrosis of adipose tissue as a mechanism for adipocyte remodeling.
Unexpectedly, we were unable to detect an increase in classical brown adipose tissue marker uncoupling protein 1 (UCP1) protein expression per adipocyte in S2-013 mPDAC samples from the gonadal fat pad (Fig. 6E, F). To investigate the downstream signaling cascades potentially involved with adipose tissue remodeling in favor of fibrosis, we analyzed the expression of phosphorylated SMAD2/3 (p-SMAD2/3) and ERK/MAP kinase signaling molecules. In addition to an increase in the expression of p-SMAD2/3 (Fig. 6G) and p-SMAD3 (Fig. 6H) in the adipose tissue of S2-013 mPDAC, we observed an increase in p-p38 and p-JNK expression and a marginal decrease in p-ERK expression, indicating that modulation of MAP kinase signaling pathways could be related with adipose tissue remodeling in favor of fibrosis.
Since we did not observe adipose tissue browning in the setting of tumor-bearing mice, we next examined if white visceral adipose tissue in PDAC patients underwent a transition towards brown adipose tissue at gene and protein levels. Accordingly, we were unable to detect any increase in UCP1 expression in adipose tissue of Stage IV PDAC patients (Fig. 7A). Consistently, gene expression of UCP1 in visceral adipose tissue revealed no significant change with increasing circulating activin A levels in PDAC patients and no significant change between PDAC patients and non-cancer patients (Fig. 7B). Accordingly, we were unable to detect an increase in UCP1 protein expression per adipocyte in PDAC patients (Fig. 7C). The gene expression levels of HSL measured in visceral adipose tissue from PDAC patients and non-cancer patients again revealed no significant change www.nature.com/scientificreports/ with increasing circulating activin A levels in PDAC patients and no significant change between patient groups (Fig. 7D).
To confirm ECM deposition in cachectic adipose tissue, we stained adipose tissue from S2-013 mPDAC and PDAC patients with Picro-Sirius Red (Fig. 8 A, B). The quantification of collagen deposits revealed an increase in area percentage in adipose tissues of S2-013 mPDAC than in that of their sham counterparts (Fig. 8C). We were also able to examine ECM deposition in visceral adipose tissue from PDAC patients. We quantified collagen deposition surrounding adipocytes and correlated the surge in pixel area percentage with increased circulating activin A levels (Fig. 8B, D). A significant increase in protein levels of fibronectin and collagen I/III was found within visceral adipose tissue from a patient with a greatly elevated level of circulating activin A in comparison to a patient with a relatively low level of circulating activin A, implying that adipose tissue could undergo remodeling in the cachectic state (Fig. 8E).
In addition, Picro-Sirius Red staining of both gonadal and subcutaneous adipose tissue obtained from C57BL/6 mice and cultured ex vivo with CM rich in activin A collected from KPC8069 cells revealed significant collagen deposition and adipocyte diameter reduction, similar to phenomena observed in vivo, which could be attenuated when supplemented with FST288 (Fig. 8F). www.nature.com/scientificreports/

Discussion
Pancreatic ductal adenocarcinoma (PDAC) is a notoriously lethal disease with indications that both adipose tissue and skeletal muscle wasting occurs early in disease progression 35 . Recent studies in mouse models have demonstrated a systemic increase of activin A, highlighting the atrophy of skeletal muscle cells 36 . Strikingly, current data from retrospective studies suggest no difference in survival between pancreatic cancer patients who lost skeletal muscle and adipose tissue mass and patients who lost only adipose tissue mass 10 . This challenges the current paradigm that muscle loss is the major complication in cancer cachexia and that cachectic progression is similar across different types of cancer. Our work provides new insight into the significance of visceral adipose tissue loss in PDAC and establishes the correlation between cachectic fat loss and autocrine/paracrine/endocrine signaling of activin A. We detected activin A expression in precursor PanIN lesions as well as PDAC tumor cells, which suggests that activin A from neoplastic cells predisposes PDAC patients to pre-cachexia status. We further revealed that activin A signaling is connected to the reduction of adipocyte size and the promotion of fibrosis in visceral adipose tissue from PDAC patients. The determinants of dramatic wasting of visceral adipose tissue in our patient population appear to transcend lipolysis and browning. Interestingly, the upregulation of SMAD and ERK signaling pathways in adipose tissue seem to indicate that activin A signaling mainly participates in visceral adipose tissue wasting. Although the metabolic crosstalk between the severity of PDAC and adipose tissue requires further investigation, our study demonstrates that activin A may be a key contributor among cancer-driven mediators for visceral adipose tissue loss in cachexia.
The first animal study regarding activin A and cancer cachexia was in the inhibin-deficient mouse model 37 . The deletion of inhibin elevated circulating activin A by 20-fold and manifested symptoms akin to human cancer cachexia 38 . A clinical trial in patients with cancer of the lung or the pancreas demonstrated that blockade of ACTRIIB using monoclonal antibody bimagrumab (BYM338) exerts an effect on preserving body weight and baseline of thigh muscle volume, total lean body mass, bone mineral density, and physical activity levels (NCT01433263). While BYM388 appears promising for recovery of skeletal muscle mass, there is no published clinical data regarding the effects of BYM338 on fat mass in patients with pancreatic cancer. To our knowledge, our study is the first to reveal the dose-dependent association between serum activin A levels and loss of visceral adipose tissue in PDAC patients.
The recently published report by Zhong et al. investigates human PDAC tumor expression of activins and related factors in relation to their effects on skeletal muscle loss. Our data concur with conclusions from their report regarding the choreography of systemic effects of activin A by PDAC tumors for the induction of cachectic pathogenesis in patients. The imbalance in activin A expression between PanIN lesions and tumor cells lends credence to a correlation between secretion of activin A and acceleration of cachexia in the disease state. The lack of change in activin A secretion between BxPC3 and PANC-1 cells suggests there may not be a relationship between KRAS or p53 mutation status and the production of activin A for promotion of cancer growth (data not shown). Intriguingly, the maintenance of high levels of activin A secretion in KPC8069 cells after changing culture medium implies that there is continuous production of activin A from tumor cells with both KRAS and p53 mutations.
Unlike phenomena observed in other studies, visceral WAT in PDAC patients and our mouse model displayed progressive loss of adipocyte function and size without undergoing the browning process. Our results instead demonstrate fibrotic remodeling of human and mouse visceral adipose tissue with elevated serum activin A, suggesting that activin A could induce trans-differentiation of white adipocytes into fibrotic cells. This is parallel to the implication of TGFβ signaling with fibrosis of subcutaneous fat in patients with cancer cachexia 25 . These shed new light on activation of SMAD-independent signaling pathways that has been described in non-adipose tissue [39][40][41] . Previous reports have posited that targeting of molecules downstream of activin A such as Twist1 or JNK ameliorates cancer-induced muscle cachexia and may prolong survival in animal models of PDAC 42,43 . In consideration of our findings, the activity of these molecules may also be targetable in adipose tissue. Given that we showed that browning of white adipose tissue is not the primary mechanism of cachectic fat loss in PDAC patients and mice, activin A seems to mediate an alternative mechanism for triglyceride degradation in visceral adipose tissue. Although our study design is limited to determine the effects of secreted activin A from PDAC tumor cells on visceral adipose tissue, further investigation is currently ongoing to examine the specificities of activin A signaling and the mechanisms by which activin A mediates cachectic fat degradation in PDAC. In summary, our study reveals a novel role of activin A in relation to the loss of visceral adipose tissue in the cachectic state in pancreatic cancer and may open new possibilities for the development of targeted therapies for the treatment of cachexia. www.nature.com/scientificreports/